key: cord-0015064-gf7ya2bd
authors: Balasubramaniam, Meenakshisundaram; Reis, Robert J. Shmookler
title: Computational target-based drug repurposing of elbasvir, an antiviral drug predicted to bind multiple SARS-CoV-2 proteins
date: 2020-04-08
journal: ChemRxiv
DOI: 10.26434/chemrxiv.12084822
sha: 00a92808a365d7f1fb9cbf7f64ac953e71e9d71b
doc_id: 15064
cord_uid: gf7ya2bd

Coronavirus disease 19 (COVID-19) is a severe acute respiratory syndrome caused by SARS-CoV-2 (2019-nCoV). While no drugs have yet been approved to treat this disease, small molecules effective against other viral infections are under clinical evaluation for therapeutic abatement of SARS-CoV-2 infections. Ongoing clinical trials include Kaletra (a combination of two protease inhibitors approved for HIV treatment), remdesivir (an investigational drug targeting RNA-dependent RNA polymerase [RdRP] of SARS-CoV-2), and hydroxychloroquine (an approved anti-malarial and immuno-modulatory drug). Since SARS-CoV-2 replication depends on three virally encoded proteins (RdRP, papain-like proteinase, and helicase), we screened 54 FDA-approved antiviral drugs and ~3300 investigational drugs for binding to these proteins using targeted and unbiased docking simulations and computational modeling. Elbasvir, a drug approved for treating hepatitis C, is predicted to bind stably and preferentially to all three proteins. At the therapeutic dosage, elbasvir has low toxicity (liver enzymes transiently elevated in 1% of subjects) and well-characterized drug-drug interactions. We predict that treatment with elbasvir, alone or in combination with other drugs such as grazoprevir, could efficiently block SARS-CoV-2 replication. The concerted action of elbasvir on at least three targets essential for viral replication renders viral mutation to drug resistance extremely unlikely.

The COVID-19 pandemic has created an urgent need for effective and rapidly deployable therapeutics [1, 2] .

Several studies employed in silico or in vitro drug screens to identify FDA-approved drugs that could be repurposed to treat patients infected with the SARS-CoV-2 virus. Among them, lopinavir/ritonavir (Kaletra TM ), chloroquine or hydroxychloroquine, and favilavir (Favipiravir TM ) were top candidates and are in clinical trials for the treatment of SARS-CoV-2 [3, 4] . Other studies have identified novel candidate drugs by computational screening of small-molecule libraries, but these would require extensive clinical trials prior to FDA approval. SARS-CoV-2 belongs to the same family as SARS-CoV, coronaviridae, with the largest genome among known RNA viruses. Previous studies have indicated several key protein that can be targeted for intervention in RNA-virus diseases [5, 6] . The viral RNA-dependent RNA polymerase (RdRP) and helicase are necessary for SARS-CoV-2 replication [7, 8] . Papain-like proteinase, another critical protein required for viral replication, is also involved in blocking the host's innate immune response [8] . Although RNA viruses are particularly known for their high mutation rates, the active-site structures of essential proteins remain highly conserved.

After the initial COVID-19 outbreak in China, the Zhang laboratory generated theoretical models of SARS-CoV-2 proteins based on the viral-genome sequence [9] . We screened libraries of 54 FDA-approved antiviral drugs and 3300 investigational drugs, against three key proteins (RdRP, helicase, and papain-like proteinase) to seek novel candidates that could be repurposed for COVID-19 therapy. Unlike other drug discovery efforts, we performed parallel screens of drugs against these three enzyme targets, in the hope of devising a highly effective cocktail to combat the disease.

Targeting proteins essential for viral replication. We selected three viral proteins that play essential roles in replication of the RNA genome of SARS-CoV-2: (1) RNA-dependent RNA polymerase (RdRP), (2) papain-like proteinase, and (3) helicase. Previous studies had shown that antiviral drugs targeting any one of these proteins will inhibit replication of RNA viruses generally, including coronaviruses [5, 7, 10] . We therefore hypothesized that disrupting these three proteins in SARS-CoV-2 should quite effectively block its replication, with little possibility of mutationally generated resistance.

Beginning with the full-length structural models of SARS-CoV-2 proteins developed by the Zhang group [9] , we used Discovery Studio Suite TM to predict the druggable pocket for each target protein (Figure 1, a-f ).

Molecular-dynamic (MD) simulations of SARS-CoV-2 proteins (RdRP, papain-like proteinase, and helicase) confirmed that the predicted drug-binding sites were stably maintained over time. We therefore used these predicted druggable pockets as targets to screen against libraries of approved and investigational antiviral drugs.

RdRP displayed a single large pocket (Figure 1, a-b) ; based on an experimental model of another viral RdRP (PDB-ID 6K32), we postulated that this was likely to be the region of interaction with template RNA.

Simulations of RNA docking to SARS-CoV-2 RdRP support stable RNA binding to the predicted pocket (see Figure 2a ), making this a suitable target for drugmediated disruption of viral replication. Similarly, we predicted the druggable pockets for papain-like proteinase (Figure 1, c-d) , and helicase (Figure 1 , e-f).

The predicted helicase pocket displayed nucleic-acid binding capability (Figure 2b) , supporting previous reports of nucleic-acid interactions with other viral helicases [11] . All three predicted binding cavities in candidate proteins were used as druggable targets in our subsequent screens.

Elbasvir interacts with RdRP, proteinase, and helicase of SARS-CoV-2. Because repurposing an approved drug is the most efficient strategy to quickly deploy new therapeutics against a novel pathogen, we screened FDA-approved antiviral drugs for docking potential with SARS-CoV-2 proteins. Target-based computational docking using Glide, followed by MMGBSA to assess interaction stability (see Methods), was performed sequentially against the three candidate-protein binding pockets. Lopinavir was the top-ranked drug for binding to SARS-CoV-2 RdRP, with elbasvir a very close second (Figure 3a) . In preliminary clinical trials for patients with severe COVID-19, the lopinavir-ritonavir combination therapy had shown no significant benefit over the standard of care without an antiviral drug [3] . Other drugs predicted to have stable binding to RdRP include daclatasvir, delavirdine, ribavirin and sofosbuvir, but all showed lower affinity than elbasvir.

We therefore focused on elbasvir as the most promising candidate for the RdRP target. Molecular modeling of the docking pose indicated that elbasvir fits well in the predicted RNA-binding pocket of RdRP (Figure 3b) . The third target protein is SARS-CoV-2 helicase. Drug targeting of helicase was shown to efficiently inhibit replication of other coronaviruses [8, 10] . Since helicase structure and function are conserved across the SARS family of viruses, including SARS-CoV and MERS [12] , we postulated that targeting the helicase protein should inhibit viral replication, independent of other inhibitory activities. Glide docking, followed by solventbased calculation of ΔG for drug : helicase binding, predicts that elbasvir should be a singularly effective candidate drug to target the helicase nucleic-acid-binding pocket; ΔG for elbasvir exceeds the next best drug, daclatasvir, by at least 15 kCal/mol, and zanamivir by 20 kCal/mol (Figure 3e ). Docking-pose modeling supports elbasvir binding to the predicted nucleic-acid binding pocket of the helicase protein (Figure 3f ). An experiment-based model of the SARS-CoV Nsp13 (helicase) protein indicates that the site of elbasvir binding coincides with the single-strand DNA (ssDNA) binding pocket established by X-ray crystallography [11, 12] .

To assess the validity of these predictions, we performed ssDNA-protein docking studies with SARS-CoV-2 helicase. Results corroborate that the nucleic-acid binding site of SARS-CoV Nsp13 is likely to be structurally conserved in SARS-CoV-2. Indeed, the same DNA-interacting residues (e.g., Ser539) are predicted to interact with RNA or ssDNA, indicating conservation of helicase structure at the protein-sequence level (Figure 2b) . We next modeled the elbasvir-helicase complex, which appeared to stabilize after ~40 ns (Figure 4c) , and retained crucial interactions such as Ser539 (Figure 4d) . We retrieved the structure of helicase bound to elbasvir at 100 ns, and attempted to dock it with single-stranded DNA. Elbasvir prevented helicase from binding to ssDNA, in contrast to stable binding in the absence of drug (Figure 5, c & d) . SsDNA was used in place of RNA, for consistency with structures previously reported for helicase : ssDNA [11, 12] . (a, b) , and preventing ssDNA from binding to viral helicase (c, d) .

Papain-like proteinase showed somewhat more chaotic behavior in the presence of elbasvir, throughout the majority of the 100-ns simulation (Figure 4e) ; both protein and ligand appeared to stabilize only after 70-80 ns.

That an elbasvir:proteinase complex was formed during modeling is supported by analysis of the docking pose at 100 ns, which shows elbasvir seated within the proteinase-docking cavity (Figure 4f) .

To assess the specificity of interactions between elbasvir and the targets we had selected as essential to virion replication, we compared elbasvir docking to those targets, vs. 100 proteins retrieved at random from the PDB database. Elbasvir showed surprising specificity for affinity to a variety of SARS-CoV-2 proteins: of the 100 random proteins screened, only 3 bound to elbasvir as well or better than either the papain-like proteinase or the viral S protein complex with the ACE2 host protein (S_ACE2); only 5 exceeded the Guanine-N7 methyltransferase; 7 exceeded RdRP; and 9 surpassed the helicase (Figure 6) . 

Through an unbiased in silico screen, supported by molecular modeling, we have unveiled an FDA-approved drug with remarkably broad affinity for SARS-CoV-2 proteins. This drug, elbasvir, is marketed in combination with grazoprevir as Zepatier TM , which was developed and approved for treatment of chronic infections with the hepatitis C virus.

We began with structure-based drug screening against viral proteins (RNA-dependent RNA polymerase, helicase, and papain-like proteinase) that are required for viral replication. Elbasvir was predicted to interact with all three proteins, with exceptionally high affinity (see Figures 3, 6 and 7) . Its stability of binding to a wide variety of proteins from the same virus is rather surprising. Because elbasvir interacts with 5-10% of proteins taken at random, with binding energies comparable to those for SARS-CoV-2 proteins (Figure 6 ), we infer that elbasvir is a highly reactive molecule -in common with many of the most effective therapeutic drugs -and is thus likely to have many off-target interactions [13] . However, it is encouraging that it has been reported to have a relatively low incidence of deleterious side effects at the recommended antiviral dosage [14] .

We note that the other component of Zepatier TM , grazoprevir, showed only modest affinity for SARS-CoV-2 proteins (Figure 7) , and is thus unlikely to contribute directly to therapeutic targeting of viral replication in COVID-19 despite the demonstrated benefit of combining the two drugs to treat hepatitis C [14] . On the other hand, it is possible for effective drug combinations to derive their synergy from indirect effects. In treatment of HIV, for example, the combination of protease inhibitors lopinavir and ritonavir (Kaletra TM ) derives most of its potency from lopinavir, with ritonavir contributing indirectly by delaying clearance of lopinavir [15] . Therefore, we propose clinical tests of elbasvir alone as well as in combination with grazoprevir or zanamivir.

The propensity of elbasvir to bind avidly to 6 distinct SARS-CoV-2 proteins, all within the top 10% of predicted affinities toward random proteins assessed for ΔGbinding (Figure 6 ), remains unexplained -and would occur by chance as 6 independent events at a frequency less than 1 per million (0.1 6 ). This may reflect the drug's dimensions (similar to a short stretch of RNA or ssDNA) fitting into nucleic-acid binding pockets, along with polar atoms appropriately oriented to form electrostatic interactions with target proteins. Binding-residue analysis indicates that elbasvir interacts chiefly with positively-charged residues including arginine and lysine, and the polar residue serine. A coronavirus must assemble its constituent proteins around the RNA core prior to exit from a host cell; this essentially obliges all viral proteins to have affinity for either RNA (typically via electrostatic binding of positively-charged amino acids to the strongly negative phosphates of the RNA backbone), or for another RNA-bound viral protein. We note in this regard that the 3 viral proteins having the strongest interactions with elbasvir are guanine-N7 methyltransferase, RdRP, and helicase -all of which have nucleic-acid binding affinity. A second plausible explanation would depend upon the evolution of this RNA virus by gene duplication [16] , resulting in viral protein structures that are all related and well conserved.

The strategy of targeting multiple vulnerabilities of a virus (or indeed, a cancer cell) is well established, and confers several important benefits: more effective lethality to the targeted entity, due to a multipronged attack; reduction in the requisite dose of each component, thereby reducing side effects; and elimination the potential of the pathogen to evolve resistance via mutation -an especially likely outcome for RNA viruses in view of their high mutation rates.

Based on target-based computational drug screening, we predict elbasvir to have very high affinity for key SARS-CoV-2 proteins, including RNA-dependent RNA polymerase, helicase, papain-like proteinase, and the viral S protein used for entry to cells decorated with ACE-2. Molecular-dynamic simulations and protein docking predict stable interactions of elbasvir with a wide range of SARS-CoV-2 targets. We therefore propose clinical tests of elbasvir as a therapeutic drug for those with demonstrated COVID-2, comparing the efficacy of elbasvir alone to its combination with grazoprevir as Zepatier TM .

The structures of SARS-CoV-2 proteins used in this study were retrieved from I-TASSER (neural networkbased structure predictions) [9] . Proteins were preprocessed using the Protein Preparation Wizard module of Schrödinger Suite [17] before docking. Druggable pockets in target proteins, including RNA-dependent RNA polymerase, helicase, and papain-like proteinase, were predicted using BIOVIA Discovery Studio BIOVIA Schrödinger's Prime module by the MMGBSA method, using molecular mechanics, the generalized Born solvent model, and solvent accessibility [18] [19] [20] .

Models from MMGBSA were used for atomistic molecular-dynamic (MD) simulations using the Desmond package (Schrödinger). Each protein-drug complex was immersed in a solvent box containing Single Point-Charge (SPC) water models, neutralized with NaCl, and brought to physiological conditions by further addition of NaCl to 0.15M. The system was next equilibrated using the NPT protocol (constant number of particles, pressure and temperature) with default temperature (300 o K) and pressure (1 bar). MD simulations were then conducted for 100-200 ns using Desmond's GPU-based simulation method [18] [19] [20] [21] [22] . Results were analyzed using the Simulation Interactions Diagram module from Schrödinger Maestro. Drug-protein complex stabilities presented in this paper were also confirmed using a web-based GROMACS simulation tool for protein-ligand complexes (https://simlab.uams.edu/ProteinWithLigand/, data not shown).

Unbiased or whole-protein docking was conducted using AutoDock-Vina molecular docking package [18, 23, 24] . Briefly, proteins and drugs were prepared in Vina format prior to docking. Unlike targeted docking protocols, the docking box was constructed to include the entire protein structure [18] . Drugs were then docked against each protein of interest using the Raccoon2 graphical user interface for Vina [25] . The resulting large datasets were analyzed using an in-house developed pipeline for analysis and interpretation of docking results.

Nucleic-acid structures used in this study were retrieved from the PDB databank (RNA, PDB-ID 6k32; ssDNA, PDB-ID 4N0O). The Hex docking program was used to predict the structures of complexes comprising RNA or single-strand DNA bound to a protein [26] . Docking parameters were set to "shape+electrostatic" with all other parameters left at default values. The top-ranked docking poses were retrieved and visualized using the BIOVIA Discovery Studio.

MB planned, designed and executed the study with guidance from RJSR. RJSR & MB wrote the manuscript. 

Drug targets for corona virus: A systematic review

A Novel Coronavirus from Patients with Pneumonia in China

A Trial of Lopinavir-Ritonavir in Adults Hospitalized with Severe Covid-19

Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro

Approved Antiviral Drugs over the Past 50 Years

The RNA-dependent RNA polymerase of the influenza A virus

Coronavirus Susceptibility to the Antiviral Remdesivir (GS-5734) Is Mediated by the Viral Polymerase and the Proofreading Exoribonuclease. mBio

Multiple enzymatic activities associated with severe acute respiratory syndrome coronavirus helicase

Protein Structure and Sequence Reanalysis of 2019-nCoV Genome Refutes Snakes as Its Intermediate Host and the Unique Similarity between Its Spike Protein Insertions and HIV-1

Helicases as antiviral drug targets

Structural basis for the regulatory function of a complex zinc-binding domain in a replicative arterivirus helicase resembling a nonsense-mediated mRNA decay helicase

Crystal structure of Middle East respiratory syndrome coronavirus helicase

Computational polypharmacology: a new paradigm for drug discovery

Safety and Efficacy of Elbasvir/Grazoprevir in Patients With Hepatitis C Virus Infection and Compensated Cirrhosis: An Integrated Analysis

Pharmacological and therapeutic properties of ritonavir-boosted protease inhibitor therapy in HIV-infected patients

Gene duplication is infrequent in the recent evolutionary history of RNA viruses

Protein and ligand preparation: parameters, protocols, and influence on virtual screening enrichments

MMB triazole analogs are potent NF-kappaB inhibitors and anti-cancer agents against both hematological and solid tumor cells

A novel microtubulebinding drug attenuates and reverses protein aggregation in animal models of Alzheimer's Disease

FCX-146, a potent allosteric inhibitor of Akt kinase in cancer cells: Lead optimization of the second-generation arylidene indanone scaffold

Structural insights into pro-aggregation effects of C. elegans CRAM-1 and its human ortholog SERF2

Apolipoprotein E4 inhibits autophagy gene products through direct, specific binding to CLEAR motifs

A novel tetrazole analogue of resveratrol is a potent anticancer agent

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

Computational protein-ligand docking and virtual drug screening with the AutoDock suite

Accelerating and focusing protein-protein docking correlations using multi-dimensional rotational FFT generating functions

We are grateful for grants from the U.