key: cord-0709286-g5ugrmax
authors: Steuten, Kas; Kim, Heeyoung; Widen, John C.; Babin, Brett M.; Onguka, Ouma; Lovell, Scott; Bolgi, Oguz; Cerikan, Berati; Cortese, Mirko; Muir, Ryan K.; Bennett, John M.; Geiss-Friedlander, Ruth; Peters, Christoph; Bartenschlager, Ralf; Bogyo, Matthew
title: Challenges for targeting SARS-CoV-2 proteases as a therapeutic strategy for COVID-19
date: 2020-11-23
journal: bioRxiv
DOI: 10.1101/2020.11.21.392753
sha: 6c9066e49e80083880d2156993fbfa135a653d4f
doc_id: 709286
cord_uid: g5ugrmax

Two proteases produced by the SARS-CoV-2 virus, Mpro and PLpro, are essential for viral replication and have become the focus of drug development programs for treatment of COVID-19. We screened a highly focused library of compounds containing covalent warheads designed to target cysteine proteases to identify new lead scaffolds for both Mpro and PLpro proteases. These efforts identified a small number of hits for the Mpro protease and no viable hits for the PLpro protease. Of the Mpro hits identified as inhibitors of the purified recombinant protease, only two compounds inhibited viral infectivity in cellular infection assays. However, we observed a substantial drop in antiviral potency upon expression of TMPRSS2, a transmembrane serine protease that acts in an alternative viral entry pathway to the lysosomal cathepsins. This loss of potency is explained by the fact that our lead Mpro inhibitors are also potent inhibitors of host cell cysteine cathepsins. To determine if this is a general property of Mpro inhibitors, we evaluated several recently reported compounds and found that they are also effective inhibitors of purified human cathepsin L and B and showed similar loss in activity in cells expressing TMPRSS2. Our results highlight the challenges of targeting Mpro and PLpro proteases and demonstrate the need to carefully assess selectivity of SARS-CoV-2 protease inhibitors to prevent clinical advancement of compounds that function through inhibition of a redundant viral entry pathway.

The emergence of the novel coronavirus, SARS-CoV-2 in late December 2019 1 created a global pandemic, which has prompted unprecedented efforts to combat the virus using diverse vaccine and therapy strategies. One of the more promising therapeutic approaches involves repurposing existing drugs that can be rapidly advanced into clinical studies. Other strategies build on existing knowledge and lead molecules that were developed in response to earlier coronavirus outbreaks 2 .

Two promising targets that emerged from the SARS-CoV-1 outbreak in 2003 were the essential main protease (M pro ) and papain-like protease (PL pro ) 3 . These two cysteine proteases are encoded in the viral polyprotein as non-structural protein (Nsp) 3 and Nsp5. They are responsible for cleavage of the viral polyprotein into several structural and non-structural proteins prior to formation of the replication organelle that is established in close proximity to virus assembly sites 3 .

Therefore, inhibition of one or both of these enzymes effectively blocks viral RNA replication and thus virus transmission.

Several covalent inhibitors containing various electrophilic warheads including αketoamides, aldehydes and α,β-unsaturated ketones have been developed as inhibitors of M pro , with α-hydroxy ketone PF-07304814 recently entering human clinical trials (ClinicalTrials.gov, NCT04535167) 4 . The development of covalent small molecule inhibitors of PL pro has been more challenging, perhaps due to a dominant non-proteolytic function and preference for relatively large ubiquitin-like protein substrates 5, 6 . This premise is further supported by the fact that, while some small peptide-based inhibitors have been reported 7 , the most successful inhibitors target exosites involved in ubiquitin recognition 6, 8, 9 . While both M pro and PL pro are considered to be promising therapeutic targets, several properties of these proteases, combined with the past history of efforts to develop protease inhibitors for other RNA viruses such as hepatitis C virus (HCV) 10 portend multiple challenges for drug discovery efforts. Like other proteases from RNA viruses, the M pro protease liberates itself from a large polyprotein through N-terminal autocleavage before the mature, active dimer can be formed [11] [12] [13] . This initial event is difficult to inhibit due to the favorability of the intramolecular reaction. After maturation, the dimeric protease is likely localized to defined regions inside the cytosol or at membrane surfaces in proximity to its viral protein substrates resulting in relatively high local substrate concentrations. In addition, a number of viral proteases have been found to undergo product inhibition where they retain their cleaved substrates within the active site, thus requiring displacement for effective inhibitor binding 14, 15 . Additionally, inhibition of M pro prior to formation of its semi-active monomer is likely impossible due to the fact that this early stage intermediate lacks a properly formed active site 11, 12, 16 . Thus, inhibitors must be highly bioavailable and cell permeant such that they can reach local concentrations that are sufficient to compete with native substrates and inhibit the viral protease early in the infection cycle.

Another significant challenge for targeting M pro and PL pro is the potential for any lead molecule to target host proteases with similar substrate preferences. This is compounded by the diverse set of cellular systems used to evaluate lead molecules, which express different levels and types of proteases. There also remains controversy about which cell type best represents primary sites of infection in vivo 17, 18 . In particular, priming of the receptor binding domain (RBD) of the S-glycoprotein of SARS-CoV-2 by host proteases is required after binding to the angiotensin converting enzyme-2 (ACE2) entry receptor 19 . This process can be mediated by multiple proteases including cysteine cathepsins B and L or the transmembrane protease serine 2 (TMPRSS2) 20, 21 . While high expression levels of both cathepsins and TMPRSS2 have been confirmed in lung tissue 22 , cell lines commonly used for viral infection assays have varying expression levels of both protease classes which can have a dramatic impact on the mechanism used by the virus for entry 18 . The redundancy of these pathways not only poses a challenge for antiviral drugs that are targeted towards host factors such as cathepsins or TMPRSS2 (K11777, E64d or camostat [23] [24] [25] ), but also for drugs that display off-target activity towards these enzymes.

In this work, we screened a highly focused library of ~650 cysteine reactive molecules against PL pro and M pro using a fluorogenic substrate assay to identify novel lead molecules as potential antiviral agents. From this screen, we identified six inhibitors containing various electrophiles, which demonstrated time-dependent inhibition of recombinant M pro . Notably, we did not identify any viable hits for PL pro . Two of the six lead M pro inhibitors were active in cellular infectivity assays using A549 epithelial lung cells, but their potency decreased significantly upon expression of TMPRSS2 as was the case for established cysteine cathepsin inhibitors (E64d and K11777) and multiple previously reported M pro inhibitors. This loss of potency could be best explained by the fact that TMPRSS2 expression provides an alternate entry pathway for the virus and therefore any lost antiviral activity was likely mediated by cathepsin inhibition. Indeed, we confirm cathepsin cross-reactivity of our newly discovered M pro inhibitors as well as for several of the reported M pro inhibitors. These results highlight the challenges for selection of M pro inhibitors based on antiviral activity without complete understanding of their target selectivity as it can result in advancement of compounds based on disruption of redundant entry pathways rather than on direct antiviral effects.

To identify potential inhibitors of M pro and PL pro , we developed fluorogenic substrate assays that allowed us to screen a focused library of cysteine reactive molecules. We based the design of internally quenched-fluorescent M pro substrates on recent specificity profiling of the P1-4 residues using non-natural amino acids with a C-terminal 7-amino-4-carbamoylmethylcoumarin (ACC) reporter 26 . However, because the reported structures have relatively low turnover rates, we decided to make extended versions of these substrates that combine the optimal P1-4 residues with the native cleavage consensus of the P1'-P3' residues (i.e., residues C-terminal of the scissile bond) 26, 27 . This required synthesis of substrates using a quencher/fluorophore pair rather than an ACC reporter ( Fig. 1A; Fig S1) . A dramatic increase was observed in the catalytic rate of substrate conversion by M pro as we incorporated more prime site residues into the substrate sequence (Fig 1B-C) . This result explains the reason for the overall low kinetic rate constants for reported ACC substrates which lack any prime side residues. As a substrate for the PL pro protease, we synthesized the reported ACC peptide derived from the ubiquitin consensus sequence LRGG (N-terminal acetylated substrate referred to as: Ac-LRGG-ACC) 7 . For activity assays we used recombinant M pro and PL pro that were cloned for expression in E. coli and subsequently purified (Fig S2A-B) . We then optimized enzyme and substrate concentrations such that the Z-factors for each assay were consistently above 0.5. We found that substrate turnover by PL pro required the presence of reducing agent DTT whereas it could be omitted in the M pro assay ( Fig S3) .

After having established optimal assay conditions, we screened a library of approximately 650 compounds designed to inhibit cysteine proteases 28, 29 . Because this set of compounds contains a diverse but highly focused set of cysteine-reactive molecules, we have found that it produces viable lead scaffolds for virtually all the cysteine protease targets that we have screened. The library contains molecules with electrophiles including aza-peptide epoxyketones, aza-peptide vinylketones, epoxides, halomethylketones, acyloxymethylketones and sulfones. We screened the library by measuring residual enzymatic activity after 30 min incubation of M pro substrate 2 and Ac-LRGG-ACC for PL pro . We set a threshold of maximum 10% residual M pro activity and identified 27 hits. In subsequent time-dependent inhibition assays, the hits were further narrowed down to six validated reproducible covalent M pro inhibitors (Fig 2A) . Surprisingly, when we screened the same compound library for inhibition of PL pro we identified only one compound that initially made the 10% cutoff, but this compound proved to be a false positive and we therefore ended up with no viable lead molecules for PL pro (Fig 2B) . An explanation for the absence of PL pro lead scaffolds in our library most probably relates to the DUB like character of the protease together with its extremely narrow substrate specificity. The six validated M pro hits can be categorized based on their electrophile class into aza-epoxyketones, chloro-and acyloxymethylketones and chloroacetamides ( Fig 2C) . We measured the kinetic inhibition parameter kinact/KI, for each compound (Fig 2D, Fig S4) and found that the aza-peptide epoxide, JCP474, was the most potent inhibitor of M pro with a kinact/KI value of 2526 ± 967 mol·sec -1 .

Interestingly, this compound was previously identified as a covalent inhibitor of SARS-CoV-1 M pro (kinact/KI: 1900 ± 400 mol·sec -1 ) 30 .

To probe the therapeutic potential of our M pro inhibitors, we tested all of the compounds for inhibition of SARS-CoV-2 infection using a cellular model. A number of different types of host cells have been used in SARS-CoV-2 infection assays, with the most common cell type being Vero E6 cells of primate origin. However, as Vero E6 cells are not an accurate mimic of the human airway and lung epithelial cells that are the primary site of SARS-CoV-2 infection, we chose to instead use the lung adenocarcinoma cell line A549. This cell line is a more relevant lung derived human cell system, but it lacks sufficient expression of the ACE2 receptor to allow efficient infections by SARS-CoV-2 20 . Hence, we stably expressed the ACE2 entry receptor in A549 cells and achieved high level infection (typically greater than 50% infection) and replication during a 24-hours observation period. Using this cell system, we found that only two out of the six initial lead compounds blocked viral replication in these cells ( Fig 3A) . The two most potent inhibitors of M pro in vitro, JCP474 and JCP543 were inactive in the cellular infection assay, likely due to the fact that they are both tripeptides with a polar P1 glutamine or asparagine residue resulting in poor cell permeability. The only two compounds that demonstrated activity were the chloromethylketone JCP400 and the acyloxymethylketone JCP403. These compounds showed relatively weak potency with greater than 75% inhibition only when applied at concentrations above 20 µM, which is well below cytotoxic concentrations (Fig. S5 ). This drop in potency of compounds in the cellular infection assay is consistent with what has been reported for other M pro inhibitors 2,13 , and is likely due to poor cellular uptake and the difficulty in achieving complete inhibition of M pro inside the host cell.

One of our concerns about screening for M pro inhibitors in our cysteine protease inhibitor library was the potential for hits to have cross-reactivity with other cysteine proteases. This becomes particularly problematic if compounds are only active against the virus at relatively high concentrations. The most likely family of off-target host proteases are the cysteine cathepsins, which are broadly expressed in many cell types and are accessible to small molecule and peptidebased inhibitors because of their lysosomal localization. Furthermore, recent studies have shown that SARS-CoV-2 can utilize multiple entry pathways into the host cell that depend on a variety of cellular proteases among which are cathepsin B and L, TMPRSS2 and furin 20, 21 . One of the primary routes involves processing of the viral spike protein by the TMPRSS2 protease. This pathway is highly redundant with a pathway involving processing by cathepsin L (recent work has shown that Cat B is unable to independently process the spike protein 31 ). Therefore, cathepsin inhibitors such as E64d and K11777 are highly potent inhibitors of viral entry in some cell lines but this activity is lost upon expression of TMPRSS2 20 . Hence, we sought to assess if either of our two lead M pro inhibitors were active in the cellular assay as a result of inhibition of host cathepsins rather than as a result of inhibiting the virus encoded M pro enzyme.

To address this issue, we generated A549+ACE2 cells that also express TMPRSS2, which is not expressed to a detectable level in regular A549 cells (data not shown), and investigated if expression of this alternate protease resulted in any change in antiviral activity. We first tested remdesivir and E64d and found that remdesivir was equipotent in both cell lines, while E64d completely lost its potency upon expression of TMPRSS2, consistent with previous studies 20 ( Fig   3B) . Following a recent report showing that the cathepsin inhibitor K11777 is a highly potent SARS-CoV-2 antiviral compound 31 , we included this molecule in our analysis and found that it too lost all of its activity upon expression of TMPRSS2. For our two lead M pro inhibitors, we found that their apparent EC50 values dropped by two to three-fold upon expression of TMPRSS2 ( Fig 3C) .

Notably, both compounds displayed some signs of cytotoxicity at concentrations above 50 µM ( Fig S5) .

To confirm that the observed drop in potency of lead molecules upon TMPRSS2 expression was due to off-target reactivity of the compounds with cysteine cathepsins, we performed competition inhibition studies using the covalent cathepsin activity-based probe (ABP) BMV109. This ABP has been used to quantify levels of cathepsin activity in various cell-based systems [32] [33] [34] [35] [36] . Using this labeling approach, we found that JCP400 and JCP403 are both able to compete with BMV109 labeling of Cat B and L in A549+ACE2 cells ( Fig 4A) . As further validation of the off-target activity of the two lead molecules, we also tested the compounds for their ability to inhibit purified Cat B and L enzymes. These results confirmed that both are relatively potent inhibitors of cathepsins with IC50 values in the low micromolar range (Fig 4A-B) .

Having confirmed that our newly identified compounds were cross-reactive with host cathepsins and that this activity was responsible for the bulk of their antiviral activity, we questioned whether previously reported M pro inhibitors might have similar properties. We first evaluated five reported M pro inhibitors for inhibition of human recombinant Cat B and L using a fixed time point fluorogenic substrate in vitro assay ( Fig 5A) . Surprisingly, the three aldehydecontaining inhibitors GC373, 11a, and 11b were highly potent with nanomolar IC50 values for both Cat L and Cat B. Rupintrivir, on the other hand, displayed no inhibition toward Cat B (tested up to 250 µM) and had only weak micromolar activity against Cat L.

We next evaluated whether the inhibitors were active against Cat B and L in A549+ACE2 cells. In-cell competition of the selected compounds with cathepsin labeling by BMV109 demonstrated that all of the reported M pro inhibitors modified the active site residues of Cat B and L ( Fig 5B, Fig S6) . Consistent with the recombinant enzyme data, compounds 11a and 11b were active against cellular Cat B and L in the micromolar range with complete competition at 20 µM.

The inhibitor GC373 and its pro-drug form GC376 show similar competition of Cat L between 5-10 µM and were slightly less potent toward Cat B with competition beginning between 20-50 µM.

Rupintrivir was active against Cat L starting at 20 µM and showed only slight inhibition of Cat B labeling even at 100 µM.

Finally, we tested the reported M pro inhibitors for activity in the A549+ACE2 cells with and without expression of TMPRSS2 to determine if cross reactivity with cathepsins was contributing to their antiviral activity. Indeed, we found that all five inhibitors showed a loss in potency upon TMPRSS2 expression similar to what we observed for our newly identified M pro inhibitors. The effect appeared to be most prominent for aldehyde 11b, which showed an 11-fold drop in potency.

Interestingly, the α,β-unsaturated ketone rupintrivir, which has low micromolar activity in the cells lacking TMPRSS2, completely lost its antiviral activity when TMPRSS2 was expressed even though it showed minimal cathepsin cross reactivity (Fig 5B) . Together with a lack of inhibitory activity against recombinant M pro (Fig S7) , this strongly suggests that rupintrivir derives all of its activity in cellular assays from weak inhibition of Cat L or possibly activity against other redundant proteases that can process the RBD to facilitate viral entry. The other compounds 11a, GC373

and GC376 displayed a 4-5-fold decrease in potency upon expression of TMRPSS2 in the host cell ( Fig 5C) . Taken together, these results suggest that all of the tested M pro inhibitors have some level of antiviral activity that is due to inhibition of host derived cathepsins and which is overcome to varying degree by the use of an alternate spike protein processing pathway employed by SARS-CoV-2.

In conclusion, inhibition of the M pro and PL pro proteases is considered to be a potentially viable therapeutic strategy for the treatment of COVID-19. However, because animal models of SARS-CoV-2 infection are still being optimized and controversy remains about cell systems that most accurately mimic aspects of the human infection (likely including viral entry pathways), it will be critical to assess key parameters of target selectivity of drug leads prior to clinical testing in humans. Furthermore, variability within the cellular systems used for antiviral testing can lead to flawed conclusions about lead candidate efficacy. The majority of current approaches only use inhibition of viral replication as a metric for efficacy of lead molecules without any direct confirmation of target inhibition. Only recently, has inhibition of processing of a genetically expressed M pro substrate or labeling of active M pro enzyme been established as a measure of M pro activity in cells 26, 37 . In this work, we describe our efforts to screen a library of approximately 650 diverse covalent inhibitor scaffolds against the two primary SARS-CoV-2 cysteine proteases, M pro and PL pro . We failed to identify any inhibitors of PLpro and ultimately found only two inhibitors of M pro that exerted antiviral activity in cell infection models, but only at relatively high concentrations.

However, we found that the antiviral activity of these lead molecules as well as several previously reported M pro inhibitors was related to their ability to inhibit host cathepsins, thus highlighting the importance of understanding compound selectivity and verifying target engagement. Taken together, our results point out the challenges for developing inhibitors of SARS-CoV-2 proteases and suggest that using strategically chosen cell lines for antiviral testing can help to prevent selection of compounds whose mechanisms of action can be easily overcome by redundant viral entry pathways. We strongly believe that our findings are of particular importance in light of drugs that are widely suggested for advancement into clinical trials such as rupintrivir 38 Chemistry methods. All reactions were performed exposed to atmospheric air unless noted otherwise and with solvents not previously dried over molecular sieves or other drying agents.

Reactions containing light sensitive materials were protected from light. Synthesis of internally quenched and fluorogenic substrates. Fmoc-ACC-OH was synthesized as described 43 . Standard Fmoc chemistry was performed on Rink AM resin as described 42 . Internally quenched peptide substrate sequences were synthesized on 2-Chlorotrityl resin using standard Fmoc chemistry as previously described 44 . Peptides were cleaved from resin Data Availability. All data and information necessary to reproduce the results reported in this manuscript are provided. Any additional data that support the findings of this study is available upon reasonable request. In-cell competition labeling with BMV109. A549+ACE2 cells were subjected to 1h treatment with inhibitor at indicated concentrations followed by 1h incubation with 1 µM BMV109. Cells were lysed and ran on SDS-PAGE gels that were scanned for in-gel fluorescence. Bar graphs represent relative densitometric quantification of two replicate experiments ± SD. C) Plots of EC50 curves of reported M pro inhibitors in A549+ACE2 cells +/-TMPRSS2. Data are means ± SD of two replicate experiments.

A Pneumonia Outbreak Associated with a New Coronavirus of Probable Bat Origin

Structure of Mpro from SARS-CoV-2 and Discovery of Its Inhibitors

Recent Insights into Emerging Coronaviruses

Characterization and Noncovalent Inhibition of the Deubiquitinase and DeISGylase Activity of SARS-CoV-2 Papain-like Protease

Activity Profiling and Structures of Inhibitor-Bound SARS-CoV-2-PLpro Protease Provides a Framework for Anti

No. October

A Noncovalent Class of Papain-like Protease/Deubiquitinase Inhibitors Blocks SARS Virus Replication

Disulfiram Can Inhibit MERS and SARS Coronavirus Papain-like Proteases via Different Modes

Hepatitis C Virus: From Molecular Virology to

Mutation of Gly-11 on the Dimer Interface Results in the Complete Crystallographic Dimer Dissociation of Severe Acute Respiratory Syndrome Coronavirus 3C-like Protease

Mechanism of the Maturation Process of SARS-CoV 3CL Protease

Crystal Structure of SARS-CoV-2 Main Protease Provides a Basis for Design of Improved a-Ketoamide Inhibitors

Product Inhibition of the Hepatitis C Virus NS3 Protease

Substrate Inhibition Kinetic Model for West Nile Virus NS2B-NS3 Protease

Liberation of SARS-CoV Main Protease from the Viral Polyprotein: N-Terminal Autocleavage Does Not Depend on the Mature Dimerization Mode

Chloroquine Does Not Inhibit Infection of Human Lung Cells with SARS-CoV-2

A Mouse-Adapted Model of SARS-CoV-2 to Test COVID-19 Countermeasures

Characterization of Spike Glycoprotein of SARS-CoV-2 on Virus Entry and Its Immune Cross-Reactivity with SARS-CoV

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Article SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Cell Entry Mechanisms of SARS-CoV-2

SARS -CoV-2 Receptor ACE 2 and TMPRSS 2 Are Primarily Expressed in Bronchial Transient Secretory Cells

Protease Inhibitors Targeting Coronavirus and Filovirus Entry

Candidate Drugs against SARS-CoV-2 and COVID-19

Camostat and The Role of Serine Protease Entry Inhibitor TMPRSS2

SARS-CoV-2 Mpro Inhibitors and Activity-Based Probes for Patient-Sample Imaging

A SARS-CoV-2 Protein Interaction Map Reveals Targets for Drug Repurposing A SARS-CoV-2 Protein Interaction Map Reveals Targets for Drug Repurposing

Covalent Ligand Discovery in Native Biological Systems

Identification of Proteases That Regulate Erythrocyte Rupture by the Malaria Parasite Plasmodium Falciparum

Crystal Structures of the Main Peptidase from the SARS Coronavirus Inhibited by a Substrate-like Aza-Peptide Epoxide

Chien-Te Tseng, 3 Aleksandra Drelich, 3 Pavla Fajtová, 4, 5 Bala C

Departments. A Cysteine Protease Inhibitor Blocks SARS-CoV-2 Infection of Human and Monkey Cells

Live Cell Imaging and Profiling of Cysteine Cathepsin Activity Using a Quenched Activity-Based Probe

Dual-Modality Activity-Based Probes as Molecular Imaging Agents for Vascular Inflammation

Cysteine Cathepsin Inhibitors as Anti-Ebola Agents

Improved Quenched Fluorescent Probe for Imaging of

Discovery of Small Molecules That Normalize the Transcriptome and Enhance Cysteine Cathepsin Activity in Progranulin-Deficient Microglia

Rational Design of a New Class of Protease Inhibitors for the Potential Treatment of Coronavirus Diseases

GC-376, and Calpain Inhibitors II, XII Inhibit SARS-CoV-2 Viral Replication by Targeting the Viral Main Protease

A Nanoluciferase SARS-CoV-2 for Rapid Neutralization Testing and Screening of Anti-Infective Drugs for COVID-19

Production of Authentic SARS-CoV Mpro with Enhanced Activity: Application as a Novel Tag-Cleavage Endopeptidase for Protein Overproduction

SARS-CoV-2 Structure and Replication Characterized by in Situ Cryo-Electron Tomography

Design of Protease Activated Optical Contrast Agents That Exploit a Latent Lysosomotropic Effect for Use in Fluorescence-Guided Surgery

Expedient Solid-Phase Synthesis of Fluorogenic Protease Substrates Using the 7-Amino-4-Carbamoylmethylcoumarin (ACC) Fluorophore