In silico drug repositioning against human NRP1 to block SARS-CoV-2 host entry

Despite COVID-19 turned into a pandemic, no approved drug for the treatment or globally available vaccine is out yet. In such a global emergency, drug repurposing approach that bypasses a costly and long-time demanding drug discovery process is an effective way in search of finding drugs for the COVID-19 treatment. Recent studies showed that SARS-CoV-2 uses neuropilin-1 (NRP1) for host entry. Here we took advantage of structural information of the NRP1 in complex with C-terminal of spike (S) protein of SARS-CoV-2 to identify drugs that may inhibit NRP1 and S protein interaction. U.S. Food and Drug Administration (FDA) approved drugs were screened using docking simulations. Among top drugs, well-tolerated drugs were selected for further analysis. Molecular dynamics (MD) simulations of drugs-NRP1 complexes were run for 100 ns to assess the persistency of binding. MM/GBSA calculations from MD simulations showed that eltrombopag, glimepiride, sitagliptin, dutasteride, and ergotamine stably and strongly bind to NRP1. In silico Alanine scanning analysis revealed that Tyr297, Trp301, and Tyr353 amino acids of NRP1 are critical for drug binding. Validating the effect of drugs analyzed in this paper by experimental studies and clinical trials will expedite the drug discovery process for COVID-19.

S protein of SARS-CoV-2 interacts with angiotensinconverting enzyme 2 (ACE2) that has been accepted as the main receptor for the entry of the virus Yan et al., 2020). ACE2, a membrane protein, is expressed in lung, upper and stratified epithelial cells of the esophagus, heart, kidney, testis, colon, ileum, and intestine (Donoghue et al., 2000;Zhang et al., 2020). Despite the attenuated expression of ACE2 in elderly people, increased or unaffected severity of COVID-19 symptoms (Walls et al., 2020) in these people (Liu et al., 2020;Omori et al., 2020;Singh et al., 2020a), and high viral loads in the throat (Wolfel et al., 2020) suggest the existence of an alternative receptor for the virus entry.
A polybasic sequence (Arg 682 Arg-Ala-Arg 685 ) at the S1/ S2 region of S protein of the SARS-CoV-2 may explain its higher transmission rate and tissue tropism. This sequence is also conserved in the S protein of various pathogenic human viruses such as Ebola, influenza, and HIV-1 (Tse et al., 2014). Cleavage of the S protein within the polybasic sequence, which is primed by transmembrane serine protease 2 (TMPRSS2) via furin produces S1 and S2 (Hoffmann et al., 2020b). FURIN mediated cleavage of S protein results in increased tropism and the infection rate of SARS-CoV-2, most probably because of the formation of new cell surface binding regions (Hoffmann et al., 2020a;Wrapp et al., 2020). Knocking down FURIN gene or using its inhibitor decreases the virus entry to cells and syncytia formation in the infected cells, respectively (Shang et al., 2020;Walls et al., 2020).
Recent studies showed that interaction between NRP1 and protein S of SARS-CoV-2 is required for the SARS-CoV-2 cell entry (Cantuti-Castelvetri et al., 2020;Daly et al., 2020). While coexpression of NRP1 with ACE2 and TMRPSS2 increased the infection, monoclonal antibody targeting the b1b2 extracellular domain of NRP1 significantly inhibited the infection. The stimulating effect of NRP1 on the virus infection diminished when the furin cleavage site of SARS-CoV-2 is mutated (Cantuti-Castelvetri et al., 2020). Analysis of direct interaction between CendR peptide of protein S and NRP1 using isothermal titration calorimetry method showed that binding affinity of the complex is between 20 and 13µM depending on the pH. Mutating the critical arginine residue (Arg 685 ) on the CendR of protein S prevented the NRP1 binding (Daly et al., 2020). The crystallographic analysis confirmed that the binding mode of protein S' CendR to the NRP1 is similar to the previously resolved NRP1-VEGF-A structure (Daly et al., 2020;Parker et al., 2012).
To date, there is no consensus on the treatment of COVID-19 yet. Since developing a new drug is costly and requiring a long period, repurposing of U.S. Food and Drug Administration (FDA) approved drugs may provide an alternative approach to find a therapeutic for the COVID-19 treatment within a reasonable time and cost. Multiple targets are available to neutralize COVID-19. For example, virus entry proteins on the host (ACE2, TMPRSS2, NRP1), replication machinery of the virus (3CL pro RNA-dependent RNA polymerase), virus proteins taking roles on the assembly mechanism (protein E) and the release of the virus (protein M and N) are candidates for drug development and drug repurposing ( Venkatagopalan et al., 2015;Schoeman and Fielding, 2019;Li et al., 2020a;Sarma et al., 2020). Given the crucial role of NRP1 for the SARS-CoV-2 entry into the host specifically in elderly people, this in silico study aimed to repurpose FDA-approved drugs against b1 domain of NRP1 to find a promising drug to block SARS-CoV-2 infection. For this purpose, initially docking simulations were run to find drugs with high binding affinity to b1 domain of NRP1. As a control, previously identified NRP1 inhibitor, EG01377, was docked to the same pocket. While docking binding energy of the inhibitor is around -6.0 kcal/mol, over 250 drugs have -7.4kcal/mol or lower AutoDock Vina binding energy. Drugs were sorted based on docking binding energies, and an extensive literature search was done to select well-tolerated drugs for further analysis. One hundred ns molecular dynamics simulations (MD) of drug-NRP1 complexes showed that thrombopoietin receptor agonist eltrombopag, migraine drug ergotamine, drugs for type 2 diabetes sitagliptin and glimepiride, and antiandrogen dutasteride can stably interact with NRP1. Crystal structure of NRP1 in complex with EG01377 (PDB ID: 6FMF) was simulated to compare the affinity of selected drugs with its inhibitor. Binding free energy calculation using MM/GBSA method showed that selected drugs have a comparable or better affinity to NRP1 than EG01377. Alanine scanning calculations revealed Tyr 297 , Trp 301 , and Tyr 353 as the critical amino acid residues for drug-NRP1 interaction. These findings may be used in experimental studies and clinical trials to test the effect of promising drugs alone or in combination with current COVID-19 treatment protocols.

Materials and methods
High-resolution human NRP1 structure (PDB ID: 6FMC) was retrieved from The Protein Data Bank 2 (Powell et al., 2018). Protein was prepared for docking simulations via Dock Prep module of UCSF Chimera as described previously ( Tardu et al., 2016;Doruk et al., 2020;Gul et al., 2020). Crystal water molecules were removed and if alternate locations are available for residues, the ones with the higher occupancies were selected. Rotamer library developed by Shapovalov and Dunbrack (2011) was used to complete missing side chains of amino acids, polar hydrogens were added, and nonpolar hydrogens were merged with bound atoms. Atom charges should be defined to run docking simulations. Thus, Gasteiger charges to each atom in the protein were assigned using Auto Dock Tools (ADT) (v. 1.5.6). Structures of drugs were downloaded from Zinc15 database catalogue of FDA-approved drugs. Set of drugs used in this study were imported from the U.S. Environmental Protection Agency's (EPA) distributed structure-searchable toxicity (DSSTox) database. The 3948 drugs and NRP1 inhibitor (EG01377) were prepared for docking by using ADT suite. Inhibitor binding pocket of NRP1 was targeted for docking simulations in which grid center was placed on the center of Tyr 297 , Glu 348 , and Tyr 353 side chains with 6400 Å 3 of the grid box. AutoDock Vina (version 1.1.2) was used for the docking (Trott and Olson, 2010).
With the aid of VMD, NRP1 structure (PDB ID: 6FMF) was solvated using TIP3P water molecules in a rectangular box with edge lengths of x = 75 Å, y = 80 Å, z = 75 Å and a size of 4.50 × 10 5 Å 3 . Solvated protein was neutralized and then ionized with sodium-chloride salt (Na + and Cl -) to 150 mM final concentration to mimic the physiological conditions (Humphrey et al., 1996). Twenty thousand steps of energy minimization (via conjugate gradient) was performed. The minimized system was gradually heated and then equilibrated for 1.4 ns (NPT ensemble) with constraints on the protein. Constrains starting from 2 kcal/mol/Å 2 were reduced by 0.5 kcal/mol/Å 2 for each 0.4 ns equilibration run. Production simulation of 100 ns for each equilibrated drug-NRP1 complex was run using 2 fs time step at 310K and 1atm pressure. Langevin thermostat and Langevin barostat maintained the temperature and pressure, respectively. To calculate the force acting on the system van der Waals (12Å cut-off) and long-range electrostatic interactions ( calculated. NAMD (Phillips et al., 2005) software and CHARMM36m force field (Huang et al., 2017) were used for all MD simulations. CHARMM-GUI server was used to generate parameters of drugs (Jo et al., 2008;Kim et al., 2017).
Analyses of protein-drug interactions were done by following these steps: 1) MD trajectory of each drugprotein simulation was visualized to inspect the position of the drug. If the drug leaves the binding pocket, that one is eliminated. 2) For simulations that drug-protein interaction was maintained for 100 ns, root mean square deviation (RMSD) of C α atoms was calculated to verify the successful equilibration of the system. 3) Contact frequency between drug and nearby amino acid residues were carried out for determining the binding residues. 4) Binding free energy (BFE) of drugs was calculated using molecular mechanics generalized born surface area (MM/GBSA) method. 5) Amino acid residue with a high contact frequency was mutated to Alanine to calculate their contribution to BFE of drugs. All MD analyses were performed and protein figures were prepared by using VMD and Pymol, respectively (Humphrey et al., 1996;DeLano, 2009). RMSD trajectory tool and timeline function in VMD were used to calculate RMSD of protein and contact frequencies between drug and protein, respectively. Unstructured and highly dynamic first 9 N-terminal amino acid residues were excluded from the RMSD calculation.
For MM/GBSA and Alanine scanning calculations, MMPBSA.py script of AmberTools20 was used (Case et al., 2020). For each calculation, 25,000 frames from 100 ns simulations were used. BFE between drug and protein is calculated based on Equation 1: ΔG binding = G complex -G receptor -G ligand (1) G complex : Energy of protein-drug complex, G receptor : Energy of protein only, G ligand : Energy of unbound drug.
NRP1 is a recently identified receptor that facilitates the SARS-CoV-2 cell entry in coordination with ACE2 and TMPRSS2 (Cantuti-Castelvetri et al., 2020;Daly et al., 2020). S protein of SARS-CoV-2 attaches to the host cells for the virus entry Yan et al., 2020). When the S protein is cleaved by the host protease, S1 and S2 proteins are generated. S1 has a polybasic sequence (Arg 682 -Arg-Ala-Arg 685 ) at the C-terminal which interacts with b1 domain of NRP1 (Cantuti-Castelvetri et al., 2020;Daly et al., 2020) (Figure 1). Mutating the critical Arg 685 diminishes protein S and NRP1 binding (Daly et al., 2020) and antibody against b1b2 domain of NRP1 relieve the SARS-CoV-2 infection (Cantuti-Castelvetri et al., 2020). Drugs having a high affinity to CendR binding pocket of NRP1 can be used in the COVID-19 treatment by blocking SARS-CoV-2 entry into cells.

Docking simulations
AutoDock Vina program was used to calculate binding energy and predict the binding mode of drugs to the target pocket as described before (Tardu et al., 2016;Gul et al., 2020). During the docking simulations, receptor (NRP1) was treated as a rigid body and drugs were allowed to sample different conformations in the CendR binding pocket. NRP1 inhibitor, EG01377 (Powell et al., 2018), was docked to NRP1 as a control to evaluate the binding affinities of drugs against NRP1. Vina binding energy of EG01377 was calculated as -6.0 kcal/mol (Table 1). The docked conformation of EG01377 is in contact with Tyr 297 , Trp 301 , Lys 351 , and Tyr 353 similar to observed in the crystal structure (6FMF) (Figure 2A). Docking simulations revealed that over 100 molecules have Vina binding energies -7.6 kcal/mol or lower (Table S1). Histogram representation of Vina binding energies of all drugs are provided in ( Figure S1). Top 15 of those drugs having the Vina binding energies between -8.5 and -8.0 kcal/mol were given in (Table 1) with their ZINC IDs, structures, and types.
Docking simulations showed that, according to the descriptions in DrugBank (Wishart et al., 2006), various types of drugs such as anticancer, antipsychotics, antiinflammatory, antibiotics, antidiabetics, and estrogen hormone may bind to NRP1 (Table S1). For further analysis, widely used drugs eltrombopag (-8.5 kcal/mol), glimepiride (-8.2 kcal/mol), dutasteride (-8.2 kcal/mol), sitagliptin (-8.2 kcal/mol), ergotamine (-8.1 kcal/mol) at the top of the docking result list, and two antimalarial drugs, mefloquine (-8.0 kcal/mol) and atovaquone (-7.9 kcal/mol), having similar binding energies to top drugs were selected. Two-dimensional diagram of interaction   Figure 2B). Among those, hydrogen bond and pi interactions between the inhibitor and Tyr 297 , Tyr 353 , and Trp 301 ; and van der Waals interactions generated through Ser 298 , Ser 346, Thr 316 , Ile 415 , Gly 414 , Thr 349 , Lys 351 , and Asn 300 were reproduced in docking pose as well. However, some hydrogen bond and pi interactions between the inhibitor and amino acids such as inhibitor-Asp 320 and Glu 348 in the crystal could not be observed in the docking pose. Hydrogen bond interactions between the inhibitor and Ser 298 , Thr 349 , Ser 346 in the crystal were generated as van der Waals type interaction in docking conformation ( Figure  2C). Note that to compare docking position of EG01377 with that of in crystal, structure with 6FMF PDB ID was used since the binding mode of inhibitor in 6FMF was stated as free of artifact in the original study (Powell et al., 2018). To solidify the idea of using NRP1 inhibitor as an antiviral drug, the binding mode of S protein to NRP1 was analyzed (Daly et al., 2020) and compared with that of EG01377.

MD simulations and binding free energy calculation
In short MD simulations (20 ns) eltrombopag, glimepiride, dutasteride, sitagliptin, and ergotamine stably interacted with NRP1, however, atovaquone and mefloquine did not. These two antimalarial drugs left their initial docking positions and moved to the solution even in two independent trials. Then we extended 20 ns simulations for those who could interact with the protein in these short simulations to 100 ns to assess the stability of their interaction with NRP1. In a control simulation, the EG01377-NRP1 complex obtained from crystal (6FMF) was simulated for 100 ns. Visual inspection of long simulations showed that all selected drugs and the inhibitor retained their docking positions on the NRP1 through the entire simulation. Root mean square deviation analysis (RMSD) of C α atoms showed that simulations reached equilibrium around 2Å after an initial jump within 1-2 ns (Figure 3). To examine the interaction between drugs and critical residues, the contact frequency between drugs and selected amino acids around the binding pocket were visualized. Thus, the persistency of drug-amino through 100 ns simulation ( Figure 4A). As a result, MD simulation of EG01377-NRP1 analysis verified interacting residues in the docking. Contact frequency analysis for eltrombopag-NRP1 simulation indicated that similar to EG01377, Tyr 297 , Trp 301 , Thr 316 , and Tyr 353 are the frequently interacting residues. The same residues were also identified in docking simulation ( Figure 4B). In addition to these highly interacting amino acids, interaction between Asp 320  and eltrombopag that is identified in the EG01377-NRP1 crystal structure (6FMF) makes eltrombopag drug a prime candidate. One minor difference between EG01377 and eltrombopag simulations is that while EG01377 interacted with Gly 318 , eltrombopag interacted more with Gly 414 . Glimepiride and sitagliptin frequently interacted with the same amino acids that are Tyr 297 , Trp 301 , Thr 316 , Thr 349 , and Tyr 353 (Figures 4C and 4E) in MD simulations. To note some interactions altered throughout the simulations. For example, interaction with highly interacting amino acids Thr 349 and Lys 351 fluctuated as simulations progressed. Despite slight differences in their docking positions, interacting residues in docking simulations are very similar. While glimepiride interacted with Tyr 297 , Trp 301 , Thr 316 , Glu 348 , Thr 349 , and Tyr 353 , sitagliptin fitted a deeper part of the pocket and, in addition to glimepiride's interacting amino acids, interacted with Thr 316 ( Figures  4C and 4E). Overall, nearby residues observed in docking simulations of glimepiride and sitagliptin were verified by MD simulations.
The docking position of dutasteride showed that the drug interacts with Tyr 297 , Thr 349 , Lys 351 , and Tyr 353 . MD simulation of dutasteride-NRP1 showed that dutasteride behaved less similarly with the inhibitor and did not interact with Trp 301 and Tyr 353 after a certain time ( Figure  4D). Possibly dutasteride translated in the binding pocket and started to interact with new residues such as Gly 318 , Glu 319 , Asp 320 , and Ile 415 .
Ergotamine showed similar interaction frequency patterns with the inhibitor and constantly interacted with Tyr 297 , Trp 301 , Thr 316 , and Tyr 353 ( Figure 4F). In addition to these, ergotamine interacted with Asp 320 , Thr 349 , and Ile 415 in certain periods of the MD simulation. In docking simulation ergotamine interacted with Tyr 297 , Trp 301 , Thr 316 , Ser 346 , Glu 348 , Thr 349 , Lys 351 , and Tyr 353 . While some of the amino acids recognized in the docking lost contact with ergotamine such as Ser 346 and Glu 348 , most of them interacted with ergotamine in the whole MD simulation.
To determine the binding free energy (BFE) between drugs and NRP1 MM/GBSA method was used. The BFE of eltrombopag, glimepiride, dutasteride, sitagliptin, and ergotamine was calculated as -17.11, -12.55, -9.47, -13.34, and -14.95 kcal/mol, respectively ( Table 2). BFE of EG01377 was calculated as -16.19 kcal/mol. A well-tolerated drug eltrombopag stimulates platelet synthesis in patients having low blood platelet (Erickson-Miller et al., 2009). In comparison to EG01377, eltrombopag has comparable Vina binding energy (-8.5 kcal/mol) and BFE (-17.11 kcal/mol) calculated from docking and MD simulations, respectively. During the MD simulations interacting residues of EG01377 and eltrombopag with NRP1 are very similar to each other ( Figure 4B). A previous drug-repurposing study against 3CL pro and RdRp of SARS-CoV-2 from our group showed that eltrombopag may bind to the active site of 3CL pro and nsp8 binding site of RdRp which may attenuate the virus activity (Gul et al., 2020). Eltrombopag exhibits in vitro antiviral properties with IC 50 lower than 10µM (Jeon et al., 2020). This report shows that eltrombopag can strongly interact with NRP1, as well. If these in silico findings can be verified by in vitro and cell-culture experiments, eltrombopag can be a very strong antiviral drug by blocking SARS-CoV-2-NRP1 binding to host and alleviating the virus replication.
Two antidiabetics, glimepiride and sitagliptin are welltolerated drugs used to treat type2 diabetes (Bautista et al., 2003;Raz et al., 2008). Both drugs have the same Vina binding energies (-8.2 kcal) in docking simulations and very similar BFEs calculated from MM/GBSA analysis that are -13.34 and -12.55 kcal/mol for sitagliptin and glimepiride, respectively. Interacting amino acid residues and BFEs of these antidiabetics are quite alike to EG01377 (Figures 4C and 4E). Sitagliptin is an inhibitor of dipeptidyl-peptidase IV (DPP-4) that is the host receptor of the Middle East respiratory syndrome (MERS)-CoV (Karasik et al., 2008;Raj et al., 2013) and in silico studies suggested that protein S of SARS-CoV-2 can bind to DPP4 as well (Li et al., 2020b;Vankadari and Wilce, 2020). Administration of sitagliptin to patients suffering from diabetes and COVID-19 does not only help to lower the blood glucose level but may also reduce the SARS-CoV-2 entry to the cells.
Dutasteride, another well-tolerated drug, inhibits type I and II 5α-reductase and shows antiandrogenic activity (Roehrborn et al., 2002;Andriole and Kirby, 2003). 5α-reductase converts testosterone to dihydrotestosterone and enlarges the prostate glands. Thus, dutasteride is used to treat prostatic conditions by blocking the 5α-reductase activity (Makridakis et al., 2000). Besides the effect on the prostate gland, dutasteride suppresses the transmembrane serine protease 2 (TMPRSS2) expression (Mostaghel et al., 2014) which primes protein S of the SARS-CoV-2 (Hoffmann et al., 2020b). Docking binding energy and MM/GBSA calculations show that dutasteride has a high affinity to NRP1 (Tables 1 and 2). Relatively higher BFE with a high standard deviation of dutasteride-NRP1 may stem from the loss of interaction between dutasteride and two aromatic residues Trp 301 and Tyr 353 as shown in contact frequency analysis ( Figure 4D). However, dutasteride has still a high affinity to the NRP1 and interacts with amino acids observed in docking simulations. Despite further in vivo analysis needed, dutasteride may be a preventive drug that shuts down the virus entry by blocking the NRP1 and downregulating the TMPRSS2.
Antimigraine ergotamine is a tolerable drug with mild and transient side effects (Diener et al., 2002;Christie et al., 2003). Several in silico drug repurposing studies reported that ergotamine may inhibit the RdRp and 3CL pro of SARS-CoV-2 ( Barage et al., 2020;Gul et al., 2020). Ergotamine has the second-best BFE among all drugs analyzed in this study (Table 2). Comparable BFEs of ergotamine and EG01377 to NRP1 suggests that ergotamine can strongly bind to NRP1, in turn, can attenuate NRP1protein S binding. If these computational analyses can be experimentally verified, ergotamine can be used to mitigate the COVID-19. To note, it has been reported that ergotamine may cross-react with antiviral drugs (Rosenthal et al., 1999;Mortier et al., 2001;Ayarragaray, 2014). Thus, further studies related to ergotamine should be designed accordingly.
Comparable BFE and interaction pattern of inhibitor and selected drugs indicate that these drugs may act as NRP1 inhibitors.

H-bond and Alanine scanning analysis
To determine the number of hydrogen bonds drug made during the MD simulation, the hydrogen bond (H-bond) generated by drugs to any amino acids through 100 ns were analyzed ( Figure 5). Calculation of the average number of H-bonds showed that each drug, including the inhibitor, produced less than one H-bond per frame. Interaction types between selected drugs and protein were analyzed and represented in a two-dimensional diagram ( Figure 6). Analyses showed that the inhibitor and all drugs interact with the protein through hydrophobic interactions, and conduct hydrogen bond and at least one type of pi interaction. Since drugs of interest possess aromatic rings, charged or aromatic amino acids on NRP1 generate those interactions. Tyr 297 consistently generated pi interactions, Despite drugs suggested to behave like NRP1 inhibitor have diverse structures, they have some common properties. For example, all drugs analyzed in details have multiple ring structures at least one with aromatic. Second, all these organic compounds bear polar groups such as carbonyl group or halogen atoms that have ability to generate hydrogen bonds or pi interactions with amino acids. Third, drugs have 70 to 110 atoms in their structure that allow them to generate high number of van der Waals interaction with the protein. These common properties of drugs cause tight binding to NRP1.  To reveal the critical amino acids for drug binding and quantify their contribution to BFE, in silico Alanine scanning mutation was performed. The most frequently and commonly interacting amino acids e.g., Tyr 297 , Trp 301 , Thr 316 , and Tyr 353 in all drug/inhibitor-NRP1 simulations were analyzed (Figures 4A-4F). These residues were mutated to Alanine one by one and BFE between drugs and NRP1 was calculated using the MM/GBSA method (Table 3). Then the difference between BFEs of drugs against mutant and wild-type NRP1 is calculated (ΔΔG). ΔΔG values larger than 0 means that mutation is destabilizing the interaction, ΔΔG lower than 0 means mutation is stabilizing the interaction. Free energy change in protein-protein interaction at equilibrium is calculated by using the following formula: ΔG ᵒ = -RTlnK eq where ΔG ᵒ is the standard free energy change, R is the gas constant, T is the absolute temperature. Using R = 1.987 × 10 -3 kcal/mol, T = 298 K in that formula gives that increase in 0.4-0.5 kcal/mol in ΔG ᵒ (ΔΔG in our data), leads to 2-fold decrease in ligand bound protein state. Thus, mutations causing ΔΔG > 0.5 kcal/mol were evaluated as critical amino acid for drug-protein binding. According to ΔΔG values, Tyr 297 , Trp 301 , and Tyr 353 are very critical for all drugs binding to NRP1. Interestingly, despite high interaction with drugs, Thr 316 minimally contributes to BFE. Tyr 297 has the highest contribution to BFE, -4.03, -4.12, -2.24 kcal/mol, in NRP1-EG01377, eltrombopag, and dutasteride simulations, respectively. Trp 301 is the second highest contributor to BFE in EG01377 (-3.1 kcal/ mol), eltrombopag (-2.36 kcal/mol), and sitagliptin (-2.60 kcal/mol) simulations and top contributor to BFE in ergotamine-NRP1 simulation (-2.43 kcal/mol). Tyr 353 has a similar effect on the binding of eltrombopag to NRP1 with Trp 301 and contributes -2.37 kcal/mol to BFE. Tyr 353 has also the highest effect on BFE in sitagliptin (-3.06 kcal/ mol) and glimepiride (-2.39 kcal/mol) simulations. (Table  3). The contribution of each mutated amino acid to drug-NRP1 BFE is given (Figure 7).
H-bond and Alanine scanning calculations provide a deeper understanding of drug-NRP1 interactions. Future structure-activity relationship (SAR) studies aim to design a more potent NRP1 inhibitor may benefit from the critical amino acids identified here. Adding an H-bond donor or acceptor to a new molecule can increase the binding affinity to NRP1.

Conclusion
Despite substantial progress has been made by several companies for the SARS-CoV-2 vaccine Mulligan et al., 2020;Sahin et al., 2020) only several governments approved for emergency use authorization. Therefore, medication is still an essential element to combat COVID-19. Vital enzymes of SARS-CoV-2 such as proteases (3CL pro and M pro ) and RNA polymerase (RdRp), or virus binding receptors on the host such as ACE2 or TMPRSS2 are generally targeted in drug repurposing studies (Bagheri and Niavarani, 2020;Busnadiego et al., 2020;Carino et al., 2020;Gul et al., 2020;Kumar et al., 2020;Li et al., 2020c;Singh et al., 2020b). Recent studies reported protein S binds to NRP1 which facilitates SARS-CoV-2 host entry (Cantuti-Castelvetri et al., 2020;Daly et al., 2020). Here a drug repurposing study was done against NRP1 in terms of docking and MD simulations using the FDA approved drugs. Furthermore, key amino acids on NRP1 for drug binding were identified by running Alanine scanning analysis. Several well-tolerated drugs showed comparable and even better affinity to NRP1 than its inhibitor EG01377. Eltrombopag had the best binding affinity to NRP1 among all drugs. In addition to eltrombopag, two antidiabetics, sitagliptin and glimepiride stably interacted with and exhibited high affinity to NRP1. Since patients having chronic diseases such as diabetes are at more risk against COVID-19, those suffering from diabetes may benefit from sitagliptin and glimepiride not only as antidiabetics but also their preventive effects, if proven by experiment, against SARS-CoV-2 infection.  (Gul et al., 2020). Third, this study shows that dutasteride can bind to and strongly interact with NRP1, thus, may decrease SARS-CoV-2 entry. Ergotamine is another top candidate molecule for the NRP1 binding. Previous calculations showed that ergotamine has the potential to bind to 3CL pro and RdRp of SARS-CoV-2 (Gul et al., 2020;Rahman et al., 2020). Therefore, ergotamine may inhibit SARS-CoV-2 at the infection and replication stages. Alanine scanning calculations uncovered that Tyr 297 , Trp 301 , and Tyr 353 are the most critical amino acid residues for drug binding to NRP1. H-bond analysis and Alanine scanning results may be exploited by further inhibitor design studies.
To expedite the discovery of therapeutics for COVID-19 drug repurposing studies may provide a rational starting point. If the findings of this study can be verified by in vitro, in vivo, and clinical trials these well-tolerated and cost-effective drugs can be adapted to current protocols used to treat COVID-19.

Acknowledgment/disclaimer/conflict of interest
I would like to thank to Dr. İ. Halil Kavaklı from Koç University for his critical reading of the manuscript.
The numerical calculations reported in this paper were partially performed at TÜBİTAK ULAKBİM, High Performance and Grid Computing Center (TRUBA resources).
-7.5 ZINC000000004893 Asenapine Atypical antipsychotic for treatment of schizophrenia and acute mania associated with bipolar disorder.
-7.5 ZINC000035342789 Tolcapone Used in the treatment of Parkinson's disease as an adjunct to levodopa/ carbidopa medication.
-7.5 ZINC000019796168 Sildenafil Prevents or minimizes the breakdown of cyclic guanosine monophosphate (cGMP) by inhibiting cGMP specific phosphodiesterase type 5 (PDE5) pulmonary hypertension -7.5 ZINC000003830836 Doxepin A psychotropic agent with antidepressant and anxiolytic properties