key: cord-0757489-sywk7014 authors: Liu, Sen; Zheng, Qiang; Wang, Zhiying title: Potential covalent drugs targeting the main protease of the SARS-CoV-2 coronavirus date: 2020-04-01 journal: Bioinformatics DOI: 10.1093/bioinformatics/btaa224 sha: bc9f1832749dffe4d9f28ad4d077414c5af3cbf0 doc_id: 757489 cord_uid: sywk7014 MOTIVATION: Since December 2019, the newly identified coronavirus SARS-CoV-2 has caused a massive health crisis worldwide and resulted in over 70,000 COVID-19 infections so far. Clinical drugs targeting SARS-CoV-2 are urgently needed to decrease the high fatality rate of confirmed COVID-19 patients. Traditional de novo drug discovery needs more than 10 years, so drug repurposing seems the best option currently to find potential drugs for treating COVID-19. RESULTS: Compared with traditional non-covalent drugs, covalent drugs have attracted escalating attention recent years due to their advantages in potential specificity upon careful design, efficiency, and patient burden. We recently developed a computational protocol named as SCAR for discovering covalent drugs. In this work, we used the SCAR protocol to identify possible covalent drugs (approved or clinically tested) targeting the main protease (3CLpro) of SARS-CoV-2. We identified 11 potential hits, among which at least 6 hits were exclusively enriched by the SCAR protocol. Since the preclinical or clinical information of these identified drugs is already available, they might be ready for being clinically tested in the treatment of COVID-19. Starting December 2019, an outbreak of pneumonia of unknown cause took place in Wuhan, Hubei province of China (Chaolin Huang et al., 2020) . In a short time, Chinese authorities rapidly isolated and characterized a novel coronavirus closely related to the SARS-CoV that caused the outbreak of a severe acute respiratory syndrome 18 years ago in China (Zhou et al., 2020) . The newly identified coronavirus was initially represented by 2019-nCoV, but formally named as SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) by the International Committee on Taxonomy of Viruses (ICTV) on February 12 th , 2020. Meanwhile, the disease caused by this virus was named as COVID-19 by the World Health Organization (WHO). Although China has adopted unprecedented policies to control the spread of the virus including temporarily "shutting down" the Wuhan City on January 23 rd , 2020, SARS-CoV-2 has resulted in over 70,000 COVID-19 patients and more than 2,000 fatalities worldwide by February 20 th , 2020. With an estimated case fatality rate of 2%-3% and the growing patient numbers, SARS-CoV-2 poses a serious health threat to the whole world (https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). Unfortunately, specific clinical treatments for the rapidly escalating international crisis caused by SARS-CoV-2 are very limited, which is one of the reasons of the high mortality rate. Due to the deficiency of targeted drugs, the current clinical treatment of COVID-19 focuses on supportive care and symptom relief. Remdesivir, an experimental drug developed to treat the Middle East respiratory syndrome coronavirus (MERS-CoV), has been reported to be effective in treating several COVID-19 cases, but a systematic clinical trial of this drug is still ongoing in Wuhan, China (ClinicalTrials.gov Identifier: NCT04257656). Similarly, chloroquine, lopinavir/ritonavir, and many other drugs have also been reported to be potentially effective, but supportive clinical data are not available for all (Maxmen, 2020) . Therefore, it is of great value to identify more potential drugs targeting SARS-CoV-2 to save COVID-19 patients. In the traditional process, the development of a drug will need over 15 years from target identification, target validation, hit discovery, lead optimization, and preclinical and clinical trials (Kaitin, 2010) . Therefore, it is virtually not possible to develop de novo drugs in the time frame needed to impact the current SARS-CoV-2 crisis. Hence, the most feasible approach is to find potential cures from clinical drugs by drug repurposing (also known as drug repositioning, reprofiling, redirecting, or rediscovering). Drug repurposing can rapidly expand target disease indications of an existing drug while saving time and money, since the data for human pharmacokinetics, safety and the preclinical results are already available (Cha et al., 2018) . Successful examples of drug repurposing include the use of sildenafil in treating erectile dysfunction and the anti-cancer uses of thalidomide (Pushpakom et al., 2019) . Previously, we described a computer-aided drug discovery protocol named as SCAR (steric-clashes alleviating receptors) (Ai et al., 2016) for the screening of covalent inhibitors of target proteins enlightened by the in silico protein design strategy (Liao et al., 2015) . Realizing the potential use of SCAR in discovering both covalent and non-covalent inhibitors, we recently demonstrated that SCAR is also quite efficient in drug repurposing (Y. Zhang et al., 2020) . Compared with non-covalent drugs, covalent drugs have the following advantages (Mah et al., 2014) : (i) covalent drugs have better biochemical efficiency since they are more competitive than many non-covalent endogenous substrates and cofactors; (ii) covalent drugs cause lower patient burden and delay the emergence of drug resistance due to lower and less frequent dosing; (iii) covalent drugs might have better target specificity by reacting with a non-conserved nucleophilic amino acid with careful designs (Cuesta et al., 2020) . Therefore, recent years have witnessed the resurgence of the discovery of covalent drugs. As a result, we set out to use SCAR to identify possible covalent drugs targeting SARS-CoV-2 by drug repurposing in this work. The protein pocket is shown in surface and the inhibitor in sticks. The covalent bond is indicated by the arrow. (B) The binding poses of the identified hits. The putative reactive atoms in the drugs are indicated by black arrows. The protein without the SCAR mutation (C145G) was used for representation, and the yellow surface close to the arrow represents Cys145. The steric conflicts between the reactive atoms and Cys145 are expected, which can be eliminated by the covalent bonding between the drug and Cys145. Table 1. The potential drugs that might be repurposed as covalent inhibitors of the SARS-CoV-2 3CLpro from the SCAR strategy. Docking score and conformation rank are listed for the identified pose. Atom distance is the distance between the putative reactive atom of the drug and Cys145-SG. SCAR enriching score is calculated as the top docking score of the drug docked to the wild-type 3CLpro minus the listed docking score. The structure files of the compounds were downloaded as mol2 files from the ZINC15 database (http://zinc15.docking.org). As described by ZINC15, the 3D conformations were protonated at physiological pH and biologically relevant tautomers were generated for each molecule (Sterling and Irwin, 2015) . The "in-trials" catalog (2019-04-22 version) was downloaded, which contains 5811 approved or investigational (clinically tested but not approved) drugs worldwide. MGLTools (version 1.5.6) was used to generate the PDBQT files from the mol2 files for docking. The SARS-CoV-2 3CLpro structure was downloaded from PDB (PDB ID: 6LU7). This is a complex structure of the SARS-CoV-2 3CLpro and an inhibitor covalently bonding to Cys145. The structure was energetically minimized in Rosetta (Leaver-Fay et al., 2011) by applying harmonic distance and angle constraints on the bonding atoms in the inhibitor and Cys145. The backbone of the protein was fixed during the minimization. The lowest score model was chosen from 1000 models. For SCAR docking, the Cys145 was mutated to Gly. MGLTools (version 1.5.6) was used to generate the PDBQT file for docking. The computational docking process was similar as previously described (Ai et al., 2016; Liao et al., 2015) . Briefly, AutoDock Vina (Trott and Olson, 2010) (version 1.1.2) was used to dock the small molecules to the substrate binding pocket of the SARS-CoV-2 3CLpro. The docked poses were then manually evaluated by docking score, ranking, and the distance between the reactive atom and the sulfur atom of Cys145 in the original structure as previously described (Ai et al., 2016) . In SARS-CoV, the 3C-like proteinase (3CLpro) is the main protease, which cleaves the large replicase polyprotein 1a (pp1a) and pp1ab to produce non-structural proteins (NSPs) for the transcription and replication of the virus (Zumla et al., 2016) . Therefore, 3CLpro is a key drug target in inhibiting SARS-CoV. The recent data showed that the SARS-CoV-2 3CLpro is highly similar with the SARS-CoV 3CLpro both in sequence and structure (PDB ID: 6LU7). Previous studies demonstrated that Cys145 is a key residue in the active site of 3CLpro (Changkang Huang et al., 2004) , which makes this residue be an attractive target for covalent bonding of covalent 3CLpro inhibitors. Cysteine is also a most popular target of covalent inhibitors because its high intrinsic reactivity at physiological pH permits the use of relatively unreactive electrophiles (Cuesta et al., 2020) . Meanwhile, targeting cysteine renders the high selectivity of covalent inhibitors due to the low prevalence of cysteine in the proteome (Cuesta et al., 2020) . To repurpose potential covalent drugs targeting the SARS-CoV-2 3CLpro, we firstly optimized the X-ray complex structure of this protein with its inhibitor using the macromolecule modeling suite Rosetta (Leaver-Fay et al., 2011) ( Figure 1A ). As demonstrated in our previous work, this structural optimization could benefit the docking result (Ai et al., 2016) . Next, we mutated Cys145 to Gly as required by the SCAR protocol to prepare the docking target (Ai et al., 2016) . The "in-trials" dataset (10083 compounds/conformations) obtained from ZINC15 was filtered by the generally used warhead groups targeting cysteine (5010 compounds/conformations) before it was used in the docking process (represented as "SCAR-dock"), among which 1253 were left for manual checking after score filtering (-8.0). As listed in Table 1 , we identified 11 potential covalent inhibitors of the 3CLpro of SARS-CoV-2 following the SCAR protocol described previously (Ai et al., 2016) . These hits contain five different covalent warhead groups suitable for targeting cysteine, which represents diverse structural options. As shown in Figure 1B , the putative covalent poses of the identified hits fit in the binding pocket reasonably. In each drug, the distance between the putative nucleophilic atom and the SG atom of Cys145 is shown in Table 1 . In 6LU7 (PDB ID), the distance between the covalent carbon atom of the inhibitor and the SG atom of Cys145 is 1.8 Å. Therefore, the proximity between these two atoms indicates high potential for covalent bonding between the drug and Cys145. To investigate how the SCAR strategy helped enriching these hits, we re-docked these compounds to the SARS-CoV-2 3CLpro without mutating Cys145 (represented as "regular-dock"). For quantitative comparison, we took the lowest score of each compound in the regulardock results. Then we compared the difference between this lowest score with the score of the same compound in the SCAR-dock results (represented as "SCAR enriching score"). A note is that we used the score of the pose oriented for covalent bonding in SCAR-dock, which is the lowest score only if the rank of the conformation is 1 ( Table 1) . As shown in Table 1 , for 6 out of these 11 compounds (54.5%), the SCARdock protocol generated lower scores (higher affinities, corresponding to positive SCAR enriching scores). This result might indicate that the SCAR protocol can generate better binding poses than the regular docking protocol if the compounds can covalently bind to Cys145, although experimental validations will be necessary. Base on the parameters in Table 1 , we suggest the following ones might have higher priorities: Telcagepant, Vidupiprant, Poziotinib, and Fostamatinib. In summary, we have identified eleven approved or investigational drugs that might be repurposed to covalently inhibit the 3CLpro of SARS-CoV-2. According to our previous studies, the hits from the SCAR protocol could have different reactive (bonding) activities. One reason is that the SCAR protocol does not consider the reactive activity of the warhead in a compound. The other reason is that the reactive activity of a warhead is affected by the attached scaffold, and the prediction of a warhead's reactive activity is still a profoundly challenging issue. Considering the worldwide health risk posed by SARS-CoV-2, although wet experiments are needed to validate our computational hits, we hope our work will provide more options for the rapid identification of targeted drugs to inhibit SARS-CoV-2 and save COVID-19 patients in time. Discovery of Covalent Ligands via Noncovalent Docking by Dissecting Covalent Docking Based on a 'Steric-Clashes Alleviating Receptor (SCAR)' Strategy Potent and selective P2-P3 ketoamide inhibitors of cathepsin K with good pharmacokinetic properties via favorable P1', P1, and/or P3 substitutions Drug repurposing from the perspective of pharmaceutical companies Ligand Conformational Bias Drives Enantioselective Modification of a Surface-Exposed Lysine on Hsp90 Cathepsin C inhibitors: property optimization and identification of a clinical candidate 3C-like proteinase from SARS coronavirus catalyzes substrate hydrolysis by a general base mechanism Clinical features of patients infected with 2019 novel coronavirus in Wuhan Deconstructing the drug development process: the new face of innovation ROSETTA3: an object-oriented software suite for the simulation and design of macromolecules Discovery of novel inhibitors of human S-adenosylmethionine decarboxylase based on in silico high-throughput screening and a nonradioactive enzymatic assay Drug discovery considerations in the development of covalent inhibitors More than 80 clinical trials launch to test coronavirus treatments Drug repurposing: progress, challenges and recommendations Investigating the proteome reactivity and selectivity of aryl halides ZINC 15 -Ligand Discovery for Everyone AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading Optimize the interactions at S4 with efficient inhibitors targeting 3C proteinase from enterovirus 71 Repurposing Clinical Drugs as AdoMetDC Inhibitors Using the SCAR Strategy A pneumonia outbreak associated with a new coronavirus of probable bat origin Coronaviruses -drug discovery and therapeutic options The authors want to thank the reviewers and all the members in the Liu Lab, Hubei University of Technology for kind suggestions. This work has been supported by the grants from National Natural Science Conflict of Interest: none declared.