key: cord-0964754-flh6jxo2
authors: Milani, Mario; Donalisio, Manuela; Bonotto, Rafaela Milan; Schneider, Edoardo; Arduino, Irene; Boni, Francesco; Lembo, David; Marcello, Alessandro; Mastrangelo, Eloise
title: Combined in silico docking and in vitro antiviral testing for drug repurposing identified lurasidone and elbasvir as SARS-CoV-2 and HCoV-OC43 inhibitors
date: 2020-11-12
journal: bioRxiv
DOI: 10.1101/2020.11.12.379958
sha: f5316aa85732305636f5617b26c5a6aa78fdab1e
doc_id: 964754
cord_uid: flh6jxo2

The current emergency of the novel coronavirus SARS-CoV-2 urged the need for broad-spectrum antiviral drugs as the first line of treatment. Coronaviruses are a large family of viruses that already challenged humanity in at least two other previous outbreaks and are likely to be a constant threat for the future. In this work we developed a pipeline based on in silico docking of known drugs on SARS-CoV RNA-dependent RNA polymerase combined with in vitro antiviral assays on both SARS-CoV-2 and the common cold human coronavirus HCoV-OC43. Results showed that certain drugs displayed activity for both viruses at a similar inhibitory concentration, while others were specific. In particular, the antipsychotic drug lurasidone and the antiviral drug elbasvir showed promising activity in the low micromolar range against both viruses with good selective index.

The growth in human and animal population density through urbanization and agricultural development, combined with increased mobility and commercial transportation, land perturbation and climate change, all have an impact on virus emergence and epidemiology.

Over the past decades, emerging zoonotic RNA viruses continuously gripped the world's attention, either briefly (like the severe acute respiratory syndrome coronavirus SARS-CoV-1 in 2003), or continuously. Many RNA virus threats were considered as re-emerging including Dengue, Zika, Ebola, and Chikungunya virus, and current consensus predicts that novel and potentially highly pathogenic agents will continue to emerge from the large, genetically variable natural pools present in the environment. Coronaviruses (CoVs) are of particular concern due to high case-fatality rates, lack of therapeutics as well as the ability to seed outbreaks that rapidly cross geographic borders. A large number of highly diverse CoVs have been identified in animal hosts and especially in bat species, where they may have the potential to diffuse in other species including humans (Fan et al., 2019) .

Coronaviruses consist of a large and diverse family of viruses that cause multiple respiratory, gastrointestinal and neurologic diseases of varying severity, including the common cold, bronchiolitis, and pneumonia (Weiss & Leibowitz, 2011) . The CoV family is divided into four genera (alpha, beta, gamma, and delta) and thus far human CoV are limited to the alpha (HCoV-229E and HCoV-NL63) and beta genera (HCoV-OC43, HCoV-HKU1); the latter includes SARS-CoV-1 and the Middle East respiratory syndrome coronavirus (MERS-CoV).

A new previously unknown coronavirus, named SARS-CoV-2, was discovered in December 2019 in Wuhan (Hubei province of China) and sequenced by January 2020 (Lu et al., 2020) .

SARS-CoV-2 is associated with an ongoing outbreak of atypical pneumonia , and was declared as 'Public Health Emergency of International Concern' on January 30 th , 2020 by the World Health Organization (www.who.int).

Currently, for the COVID19 outbreak, many known drugs are under clinical investigation (Kupferschmidt, 2020) (Magro, 2020) , following different general principles and mechanisms of action: 1. the control of cytokine storms due to the hyper-reaction of the immune system against the virus (e.g. corticosteroids, (Salton et al., 2020) ); 2. the control of coagulopathy (e.g. heparin) 3. the inhibition of viral RdRp (e.g. prodrugs favipiravir and remdesivir); 4. the inhibition of viral entry (e.g. hydroxychloroquine); 5. the inhibition of the viral main protease (e.g. lopinavir and ritonavir); 5. the inhibition of viral attachment (the viral receptor ACE2 antagonist losartan).

Despite their species diversity, CoVs share key genomic elements that are essential for viral replication, suggesting the possibility to design broad spectrum therapeutic agents to address the current epidemic and manage possible future outbreaks. The target considered in this work to identify new inhibitors is the highly conserved RNA dependent RNA polymerase (RdRp), that plays a crucial role in CoV replication cycle, catalyzing the synthesis of new viral RNA (Te Velthuis et al., 2012) . The cryo-EM structure of RdRp of SARS-CoV-1 and of SARS-CoV-2, bound to nsp7 and nsp8 co-factors, have been recently solved (PDB codes: 6NUR (Kirchdoerfer & Ward, 2019) , and 6M71 , respectively). The two proteins share a sequence identity of 96% (98% conservative substitution) and a structural r.m.s.d. of 0.54 Å (considering 788 Cas).

The exploration of libraries of molecules already in use as human drugs and well characterized in terms of human metabolism might allow the identification of antivirals that could be, in principle, rapidly tested in patients. Accordingly, we chose to analyze in silico the public database of approved drugs (DrugBank library, https://go.drugbank.com/), targeting a wide region around the active site of SARS-CoV RdRp. The computational work allowed the selection of 13 commercially available compounds with predicted high affinity for the protein and favorable solubility properties. These potential inhibitors (together with suramin, known to inhibit several RNA viruses) have been tested in cell based assays against SARS-CoV-2 and CoV-OC43 (Su et al., 2016) , revealing moderate to high antiviral activities for seven of them.

Our results confirm antiviral properties already described for some of the selected compounds, and, more importantly, show new interesting properties for the compounds lurasidone and elbasvir as betacoronavirus inhibitors.

For the purpose of known drugs repurposing, a total of 6996 molecules were downloaded from the DrugBank library (https://www.drugbank.ca/) to target a wide region (~13,300 Å 3 ) around the active site of SARS-CoV RdRp (PDB-ID 6NUR; (Kirchdoerfer & Ward, 2019) ). The in silico screening was divided into two runs: a fast procedure for the selection of the best 2% of the library (118 compounds), with predicted binding free energy values (ΔG) from -8.9 to -7.6 kcal/mol, followed by a more accurate analysis with AutoDock4.2. In this way we ranked 118 known drugs based on the ΔG value of the best pose for every compound (between -11.7 kcal/mol (predicted Ki=2.7 nM) and -0.12 kcal/mol (predicted Ki=819 mM). The list of the first best 60 compounds is reported in supplemental material (Table S1) .

From our list a reasonable number of compounds was selected for cell-based assays (Table 1) , taking into account commercial availability and solubility properties. Suramin was added to the list, since it was already known to inhibit several RNA viruses such as flavivirus (Basavannacharya & Vasudevan, 2014) (Albulescu et al., 2017) , norovirus (Mastrangelo et al., 2012) , but also chikungunya and Ebola viruses (Albulescu et al., 2017) (Henß et al., 2016) . The best in silico docking pose of each of the selected thirteen compounds, in the RdRp active site, is reported in Figure 1A . The protein region explored is located between thumb, fingers and palm domains and would host growing dsRNA during polymerase activity. Such region defines a wide, complex and variable hydrophilic protein surface, and it is therefore able to host very different types of ligands. In Figure 1B , C, D we report the lurasidone and elbasvir best docking sites, between the thumb and fingers domains and in palm domain, respectively. Since, among the selected compounds, known inhibitors of viral proteases (like simeprevir and grazoprevir) were also present, we performed an additional in silico analysis 8 targeting the active site of main protease (PDB-ID: 6LU7 (Jin et al., 2020) ) obtaining the results listed in Table 1 . Eight of the selected compounds showed predicted binding affinity for the protease lower than 50 nM suggesting that such known drugs could be in principle active against multiple targets.

The antiviral activity of the selected compounds was assessed against two pathogenic CoVs strains: SARS-CoV-2 and HCoV-OC43.

test antiviral drugs against SARS-CoV-2 in vitro. The assay was established using Huh7 cells engineered with the human ACE-2 receptor (Huh7-hACE2) to promote viral infection. The readout of the assay was accessed using immunofluorescence to quantify the number of Spikepositive cells to measure infected cells, and number of nuclei to measure cell viability, as shown in Figure 2 . The panel of compounds was tested in dose response, and the assay was validated using Hydroxychloroquine as reference compound. Results are reported in Table 2 , Figure 3 and Figure S2 . Among all, lurasidone grazoprevir and elbasvir showed the best antiviral profile against SARS-CoV-2 ( Figure 3) . 

In order to evaluate the anti-HCoV-OC43 activity of the selected compounds, focus reduction assays were performed on MRC-5 cells, as described in the Materials and Methods section and elsewhere . Results on antiviral activity and cell toxicity are reported in Table 3 , Figure 4 and Figure S3 . Among the tested compounds, alectinib showed the strongest inhibitory activity against HCoV-OC43, with an EC50 in the low micromolar range (0.6 µM). Lurasidone and elbasvir also 14 exerted high antiviral activity against HCoV-OC43, exhibiting EC50 in the low micromolar range: 1.1 µM and 1.5 µM, respectively. A moderate antiviral activity was shown by tedizolid, carbenoxolone and suramin, with EC50 ranging from 11.0 µM to 94.0 µM. The afore-mentioned compounds' antiviral effect was not a consequence of cytotoxicity, since none of the screened compounds significantly reduced cell viability at any concentration used in the antiviral assays (i.e. up to 100 µM), exhibiting CC50 values higher than 1000 µM. By contrast, the remaining compounds did not exhibit interesting features as anti-HCoV-OC43 molecules, due to either no antiviral activity (cefoperazone, teniposide, irinotecan, natamycin), or low-moderate selectivity index (ponatinib, venetoclax, simeprevir, grazoprevir). In summary, these data showed that alectinib, lurasidone and elbasvir were endowed with strong anti-HCoV-OC43 activity (Figure 4) , with minimal toxicity and selectivity indexes higher than 600.

An accurate in silico docking search within a wide region around the SARS-CoV RdRp active site, allowed us to select 13 known drugs from the DrugBank library to be experimentally tested. We added suramin to the list, a well-known compound able to inhibit different RNA viruses (Mastrangelo et al., 2012 ) (De Clercq, 1979 ) (Albulescu et al., 2015) .

Unexpectedly, a rather high percentage (>60%) of the selected compounds showed some in vitro activity against one or both of the tested CoV strains. A possible explanation for such a positive result is related to the characteristics of the protein region selected for the in silico docking. Such a portion of the protein is a wide, complex and rather hydrophilic surface with many conformational degrees of freedom, allowing it to adapt to the growing dsRNA during translation. Accordingly, the average crystallographic (or cryoEM) conformation of this region must be capable to accommodate different kind of ligands (as shown in Figure 1) , with a 15 preferential affinity for large compounds possessing polar/charged moieties and planar aromatic groups: i.e. compounds that generally mimics RNA backbone and bases. In other words, the in silico docking on RdRp not only selects compounds potentially capable of interfering with the polymerase activity but could also act as a molecular filter for the selection of properties generally favorable for protein binding/inhibition. This explanation is supported by the predicted high affinity for the main protease of most of the selected compounds (Table   1 ).

Among the tested compounds lurasidone and elbasvir displayed higher activity and lower cytotoxicity against both SARS-CoV-2 and HCoV-OC43 strains. Lurasidone lead to complete inhibition of both strains with EC50 values in the micromolar range (18 and 1.1 µM, respectively) and favorable selectivity indexes. Lurasidone is an antipsychotic drug for treatment of acute depression and schizophrenia, known to bind with a low nanomolar affinity to Dopamine-2, 5-HT1A, 5-HT2A, and 5-HT7 receptors, and with slightly lower affinity to alpha-2C adrenergic receptors (Greenberg & Citrome, 2017) . Lurasidone was already identified as a potential inhibitor of SARS-CoV-2 main protease (Elmezayen et al., 2020) and in our in silico analysis it showed good predicted binding affinity for both RdRp and main protease. Since SARS-CoV-2 and HCoV-OC43 share a high level of protein sequence conservation (Vijgen et al., 2005) we hypothesize that mechanisms of action of lurasidone against these viruses might be the same.

Elbasvir inhibited SARS-CoV-2 and HCoV-OC43 with EC50 values in the micromolar range (about 23 and 1.5 µM, respectively). Previous in silico studies predicted elbasvir as a high affinity compound for the RdRp, the papain-like protease and the helicase of SARS-CoV-2 (Balasubramaniam & Shmookler Reis, 2020), whereas our in silico investigation suggested a preferential binding for the main protease (1.2 nM). Elbasvir is an inhibitor of the HCV NS5A 16 protein that has not homologues in coronaviruses: in light of our and previous work results, it has the potential to inhibit different viral proteins.

Alectinib (with the best in silico Ki against RdRp) showed no activity against SARS-CoV-2 but it is the best compound against HCoV-OC43 (EC50=0.6 µM, CC50 value >1000 µM). Alectinib inhibits the anaplastic lymphoma kinase (ALK) tyrosine kinase receptor in the nM range (Kinoshita et al., 2012) , binding to the ATP binding site. Therefore, we can speculate that it might inhibit other kinases essential for HCoV-OC43 replication in MRC-5 cells.

Moreover, we can speculate that the compound could act as a competitive inhibitor of RdRp, interfering with the nucleotide binding site(s) of the enzyme. In this case a possible explanation for the different results obtained for SARS-CoV-2 and HCoV-OC43 could be related to the differences between the two polymerases (sequence identity of 55.1%) (Elfiky, 2020) .

Grazoprevir inhibited SARS-CoV-2 with an EC50 around 16 µM (CC50 value >100 µM). Published computational studies suggested grazoprevir as a potential inhibitor of the nucleocapsid protein or the papain-like protease of SARS-CoV-2 (Behera et al., 2020) .

Grazoprevir is an inhibitor of HCV protease and it is often used for therapy in combination with elbasvir (in the drug named zepatier).

Simeprevir, another inhibitor of HCV protease, has been previously shown to inhibit SARS-CoV-2 in synergy with remdesivir (Lo et al., 2020) . In our experiments it showed a similar potency against both SARS-CoV-2 (EC50 about 9.3 µM) and HCoV-OC43 (EC50 about 2.1 µM) but with low SI. From our docking results its effect is likely directed against the protease, but RdRp inhibition cannot be excluded.

ABT-199, also known as venetoclax, is a potent selective Bcl2 inhibitor, which induces the apoptosis pathway. An early work showed that Bcl2 expression prevents SARS-CoV induced apoptosis (Bordi et al., 2006) . In addition, previous reports demonstrated that SARS-CoV 7a protein was dependent on Bcl2 to induce apoptosis, suggesting Bcl2 as an important host factor for virus replication and pathogenesis (Tan et al., 2007) . However, despite its good EC50 (about 6.2 µM) against SARS-CoV-2, venetoclax shows high toxicity in the tested cells.

Ponatinib, an oral drug for the treatment of chronic myeloid leukemia and Philadelphia chromosome-positive acute lymphoblastic leukemia, was already proposed as SARS-CoV-2

inhibitor (Nguyen et al., 2020) (Sauvat et al., 2020) (Gordon et al., 2020) , and it is shown here to inhibit SARS-CoV-2 and HCoV-OC43 with a poor selective index.

All the other compounds, although some of them have been described in the literature as potential inhibitors of SARS-CoV-2 (i.e. teniposide (Kadioglu1 et al., n.d.) and irinotecan (B, 2020), did not show any relevant activity in either of the two viruses tested.

In our work we have: 1. excluded SARS-CoV-2 antiviral activity for teniposide (Kadioglu1 et al., n.d.) and irinotecan (B, 2020), selected from previous computational studies;

2. showed the ability of some of the already described anti-SARS-CoV-2 compounds to inhibit also coronavirus HCoV-OC43 causing the common cold (suramin, ponatinib -although with a low SI); and most importantly 3. showed the capability of some of the selected drugs to selectively inhibit HCoV-OC43 (alectinib) or SARS-CoV-2 (grazoprevir) or be active against both CoV strains (lurasidone and elbasvir). Treatment of CoV infections with drugs that could inhibit different viral targets, as predicted for lurasidone and elbasvir, would be an effective way to lower chances of the emergence of drug resistant viral strains.

Of note, in previous works it was demonstrated that alectinib (Song et al., 2015) could penetrate the blood-brain barrier (BBB) exerting its activity in the central nervous system (CNS). Since HCoV-OC43, as other coronaviruses, is able to invade the CNS (Dubé et al., 2018) , alectinib might be an interesting candidate for the treatment of HCoV-OC43 persistent infections in the brain. Moreover, the free levels of alectinib found in both plasma and cerebrospinal fluid are similar (Herden & Waller, 2018) and its EC50 against HCoV-OC43 (0.6 uM) is lower than the maximum level attainable in human serum with daily recommended dosage (676 ng/mL corresponding to 1.4 µM) (Ly et al., 2018) .

In conclusion, our approach allowed the identification of lead-drugs for further in vitro and clinical investigation to contain the present outbreak. Furthermore, it could contribute to the identification of broad spectrum anti-CoV inhibitors / therapies that would allow for a rapid and effective reaction to future epidemics.

In silico docking. The virtual Library of DrugBank (https://www.drugbank.ca/) employed for the docking analysis (6996 compounds) includes commercially available FDAapproved drugs as well as experimental drugs going through the FDA approval process. The atomic coordinates of SARS-CoV RdRp (PDB-ID 6NUR) bound to NSP7 and NSP8 cofactors, were chosen as docking model for CoV polymerase. Hydrogen atoms and Kollman charges (Singh & Kollman, 1984) were added using the program Python Molecule Viewer 1.5.4 (MGL-tools package http://mgltools.scripps.edu/). The protein model was then used to build a discrete grid within a box of dimensions 22.5x26.3x22.5 Å 3 (program autogrid (Goodford, 1985) ) as the explored volume for the docking search. The grid was centered near the side chain of Lys545, to include a wide region around the protein active site. During the computational analysis, the protein was constrained as rigid, whereas the small molecules were free to move. The in silico screen was divided into two runs: a fast procedure using AutoDock

Vina (Trott & Olson, 2009) for the selection of the best compounds, followed by a more accurate screen using AutoDock4.2 (Morris et al., 2009) . The Autodock Vina docking search produced a ranked list of all compounds, with predicted binding free energy values (ΔG) ranging between -0.9 kcal/mol and -8.9 kcal/mol. The best 118 compounds (~2% of the library, ΔG between -7.6 to -8.9 kcal/mol) were further analyzed using AutoDock4.2 (Morris et al., 2009) , with 80 hybrid GA-LS genetic algorithm runs. Among the molecules with higher predicted affinity for RdRp (ΔG values varying between -0.12 and -11.7 kcal/mol), 13 FDA approved drugs were selected, taking into account commercial availability and solubility properties, for in vitro assays. Since among such drugs were present known inhibitor of viral protease we investigated their binding affinity for CoV main protease (PDB-ID 6LU7; (Jin et al., 2020) ). Briefly, we explored with AutoDock4.2 a region of 15x22.5x22.5 Å 3 (after mutating the active site Cys145 to Ala) centered between the side chains of Asn142 and

Gln189.

Cell lines and viruses. Vero E6 cells (ATCC-1586) , the human hepatocarcinoma Huh7 cells kindly provided by Ralf Bartenschlager (University of Heidelberg, Germany) and Huh 7 engineering by lentivirus transduction to overexpress the human ACE2 (Huh-7hACE2) were cultured in Dulbecco's modified 7 Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Gibco). Working stocks of SARS-CoV-2 ICGEB-FVG_5 isolated in Trieste, Italy, were routinely propagated and titrated on Vero E6 cells (Licastro et al., 2020) . 

Reagents. Dimethyl sulfoxide (DMSO) was purchased from Sigma-Aldrich (Saint Louis, MO). incubator, and titrated by standard plaque method on MRC-5 cells, as described elsewhere ; titers were expressed in terms of plaque forming units per ml (PFU/ml).

Cell viability assay. Cell viability was measured using the MTS assay, as described elsewhere (Lembo et al., 2014) . MRC-5 cells were seeded at a density of 2×10 Statistical analyses. All data were analyzed using GraphPad Prism 5.0 (GraphPad Software, San Diego, CA). All results are presented as means ± standard deviations.

This work was funded by the #FarmaCovid crowd founding initiative 

Suramin inhibits Zika virus replication by interfering with virus attachment and release of infectious particles

Suramin inhibits chikungunya virus replication through multiple mechanisms

The AI-Discovered Aetiology of COVID-19 and Rationale of the Irinotecan+Etoposide Combination Therapy for Critically Ill COVID-19 Patients

Computational Target-Based Drug Repurposing of Elbasvir, an Antiviral Drug Predicted to Bind Multiple SARS

Suramin inhibits helicase activity of NS3 protein of dengue virus in a fluorescence-based high throughput assay format

Simultaneous Inhibition of Entry and Replication of Novel Corona Virus by Grazoprevir: A Computational Drug Repurposing Study

Bcl-2 inhibits the caspase-dependent apoptosis induced by SARS-CoV without affecting virus replication kinetics

Suramin: A potent inhibitor of the reverse transcriptase of RNA tumor viruses

Axonal Transport Enables Neuron-to-Neuron Propagation of Human Coronavirus OC43

Ribavirin, Remdesivir, Sofosbuvir, Galidesivir, and Tenofovir against

SARS-CoV-2 RNA dependent RNA polymerase (RdRp): A molecular docking study

Drug repurposing for coronavirus (COVID-19): in silico screening of known drugs against coronavirus 3CL hydrolase and protease enzymes

Bat Coronaviruses in China

Structure of the RNA-dependent RNA polymerase from COVID-19 virus

A Computational Procedure for Determining Energetically Favorable Binding Sites on Biologically Important Macromolecules

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

Pharmacokinetics and Pharmacodynamics of Lurasidone Hydrochloride, a Second-Generation Antipsychotic: A Systematic Review of the Published Literature

Suramin is a potent inhibitor of Chikungunya and Ebola virus cell entry

Alectinib

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

Identification of novel compounds against three targets of SARS CoV-2 coronavirus by combined virtual screening and supervised machine learning

Design and synthesis of a highly selective, orally active and potent anaplastic lymphoma kinase inhibitor (CH5424802)

Structure of the SARS-CoV nsp12 polymerase bound to nsp7 and nsp8 co-factors

WHO launches global megatrial of the four most promising coronavirus treatments

Auto-associative heparin nanoassemblies: A biomimetic platform against the heparan sulfate-dependent viruses HSV-1, HSV-2, HPV-16 and RSV

Isolation and Full-Length Genome Characterization of SARS-CoV-2 from COVID-19 Cases in Northern Italy

Simeprevir potently suppresses SARS-CoV-2 replication and synergizes with remdesivir. BioRxiv, 2020.05.26.116020

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

Alectinib for advanced ALK -positive nonsmall-cell lung cancer

COVID-19: Review on latest available drugs and therapies against SARS-CoV-2. Coagulation and inflammation cross-talking

The cholesterol metabolite 27-hydroxycholesterol inhibits SARS-CoV-2 and is markedly decreased in COVID-19 patients

Structure-based inhibition of norovirus RNA-dependent RNA polymerases

Software news and updates AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility

Potentially highly potent drugs for 2019-nCoV

Repurposing of Miglustat to inhibit the coronavirus Severe Acquired Respiratory Syndrome SARS-CoV-2

Suramin inhibits SARS-CoV-2 infection in cell culture by interfering with early steps of the replication cycle

Prolonged Low-Dose Methylprednisolone in Patients With Severe COVID-19 Pneumonia

On-target versus off-target effects of drugs inhibiting the replication of SARS-CoV-2

An approach to computing electrostatic charges for molecules

Alectinib: A novel second generation anaplastic lymphoma kinase (ALK) inhibitor for overcoming clinically-acquired resistance

Epidemiology, Genetic Recombination, and Pathogenesis of Coronaviruses

Induction of Apoptosis by the Severe Acute Respiratory Syndrome Coronavirus 7a Protein Is Dependent on Its Interaction with the Bcl-XL Protein

The SARScoronavirus nsp7+nsp8 complex is a unique multimeric RNA polymerase capable of 30

both de novo initiation and primer extension

AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading

Complete Genomic Sequence of Human Coronavirus OC43: Molecular Clock Analysis Suggests a Relatively Recent Zoonotic Coronavirus Transmission Event

Coronavirus pathogenesis

We would like to thank all the people that contributed to the #FarmaCovid crowd founding initiative and in particular Mrs. Paola Allegretti, who with great commitment has advised many people from Perugia to support our research work.