Virtual Screening of Repurposed Drugs as Potential Spike Protein Inhibitors of Different SARS-CoV-2 Variants: Molecular Docking Study

Like most of the RNA viruses, SARS-CoV-2 continuously mutates. Although many mutations have an insignificant impact on the virus properties, mutations in the surface protein, especially those in the receptor-binding domain, may lead to immune or vaccine escape variants, or altered binding activities to both the cell receptor and the drugs targeting such a protein. The current study intended to assess the ability of different variants of interest (VOIs) and variants of concern (VOCs) of SARS-CoV-2 for their affinities of binding to different repurposed drugs. Seven FDA approved drugs, namely, camostat, nafamostat mesylate, fenofibrate, umifenovir, nelfinavir, cefoperazone and ceftazidime, were selected based on their reported in vitro and clinical activities against SARA-CoV-2. The S1 protein subunit from eleven different variants, including the latest highly contiguous omicron variant, were used as targets for the docking study. The docking results revealed that all tested drugs possess moderate to high binding energies to the receptor-binding domain (RBD) of the S1 protein for all different variants. Cefoperazone was found to possess the highest binding energy to the RBD of the S1 protein of all the eleven variants. Ceftazidime was the second-best drug in terms of binding affinity towards the S1 RBD of the investigated variants. On the other hand, fenofibrate showed the least binding affinity towards the RBD of the S1 protein of all eleven variants. The binding affinities of anti-spike drugs varied among different variants. Most of the interacting amino acid residues of the receptor fall within the RBD (438–506).


Introduction
Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) belongs to the family Coronaviridae. The structural proteins of the SARS-CoV-2 include the spike (S) protein, nucleocapsid protein (N), envelope protein (E) and matrix protein (M). The S protein is translated as an inactive precursor (S0) that requires post-translation cleavage at the furin polybasic cleavage site by the host-derived serine protease enzymes into S1 and S2 [1]. Viral neutralizing antibodies are directed to S1. S1 is also responsible for binding of the virus to the cell surface receptor. The receptor-binding domain is the part of the S1 that binds to human angiotensin converting enzyme-2, ACE2, before entering the cells by clathrinmediated endocytosis [2]. SARS-CoV-2 was found to have ten-fold higher binding affinity to the ACE2 in comparison to other SARS-CoVs [3]. More recently, metabotropic glutamate receptor subtype 2 (mGluR2) was suggested to be an important factor for internalization of SARS-CoV-2 into the cell after cell binding to ACE2 [4].
Similar to other coronaviruses, SARS-CoV-2 mutates continuously. Most of these mutations possess an insignificant role in virus virulence or transmissibility [5]. However, a small proportion could result in the emergence of different types of variants of interest (VOI) and variants of concern (VOC). VOC are proved to be associated with increase transmissibility, virulence and/or decreased effectiveness to available diagnostics and vaccines, while VOI are variants that harbour a genetic constitution that are supposed to affect the virus transmissibility, virulence, break through immunity, diagnostics and spread in many countries. Currently, five VOC, alpha (B.1.1.7), beta (B.1.351), gamma (P.1), delta (B.1.617.2) and omicron (B.1.1.529), and two VOI, lambda (C.37) and mu (B.1.621), are circulating in different parts of the world [6]. The recent omicron VOC contains 15 amino acid substitutions in the RBD [7]. Hindering the S/ACE2 receptor binding by neutralizing antibodies or antiviral drugs could inhibit viral replication by preventing viral entry to the host cells [2,8,9]. However, the presence of accelerating genetic variation of the S1 and RBD could be a real challenge against using this type of antiviral strategy [7].
In the present study, seven repurposed FDA approved drugs were selected to investigate their potential inhibitory activities of the S1 unit RBD domain of the S1 protein of different SAR-CoV-2 variants via molecular docking. The selection of the drugs under investigation was based on their recently reported activity against SARS-CoV-2 as potential repurposed drugs. Fenofibrate, a hyperlipidaemic drug, was reported to significantly reduce SARS-CoV-2 infection in cell culture models [10]. Both camostat mesylate, used as a treatment of chronic pancreatitis, and nafamostat mesylate, an anticoagulant drug, are serine protease inhibitors that inhibit TMPRSS2. Camostat was first speculated to possess antiviral activity against SARS-CoV-2 [1]. Afterwards, camostat mesylate and nafomostat were considered as potential repurposed drugs against SARS-CoV-2 [11,12]. The anti-HIV drug nelfinavir has been reported as a potential inhibitor of the cell fusion of the SARS-CoV-2 S-glycoprotein [13]. Umifenovir is a broad-spectrum antiviral drug approved in Russia and China for treatment of influenza, SARS and Lassa viruses. In the current COVID-19 pandemic, umifenovir underwent several clinical trials as a potential repurposed drug for treatment of SARS-CoV-2 infection [14,15]. Ceftazidime is an antibiotic that inhibits SARS-CoV-2 infection in vitro. It inhibits the SARS-CoV-2 main protease and blocks the binding of SARS-CoV-2 S1 to the ACE2 [16,17]. Cefoperazone is another third-generation cephalosporin broad-spectrum antibiotic used for treatment of both mild and severe cases of SARS-CoV-2 infections in combination with sulbactam, with significant cure rates [18].
The current study intended to screen and compare the binding affinity of different potential SARS-CoV-2 antiviral agents to the S1 of omicron and other variants.

Ligand Preparation
The 2D structure of all drugs used in the study, including camostat, fenofibrate, nafamostat mesylate, nelfinavir, umifenovir, cefoperazone and ceftazidime, were compiled by us using ChemDraw Professional. The 3D structures of all drug ligands were constructed using Chem 3D ultra 17.0 software molecular modelling and analysis (CambridgeSoft Corporation, Cambridge, MA, USA (2017)), then they were energetically minimized using MOPAC (semi-empirical quantum mechanics0, Job Type with 100 iterations and minimum RMS gradient of 0.01, and saved as an MDL MolFile [*.mol] (Figure 1). Curr. Issues Mol. Biol. 2022, 2, FOR PEER REVIEW 3 Figure 1. Potential SARS-CoV-2 S1 glycoprotein repurposed FDA approved drugs.

S1 Protein Retrieval and Homology Modelling
Sequences of the S1 RBD protein of the different SARS-CoV-2 variants, including Wuhan

S1 Protein Retrieval and Homology Modelling
Sequences of the S1 RBD protein of the different SARS-CoV-2 variants, including Wuhan were retrieved from GISAID and GenBank databases. Amino acid deduced amino acid se-quences of the S1 protein were aligned using Mega 5 software (Supplementary Material File S1). Sequences were loaded into the SWISS-MODEL server (http//swissmodel.expasy.org/) on 9 January 2022 using the default settings of the server to create 3D homology models of the S1 protein variants. The online server created three different models of each variant. The top ranked homology models were downloaded as PDB files. The downloaded PDB files underwent protein preparation and optimization including removal of all cofactors and ligands using standard protein preparation protocol in Molecular Virtual Docker (MVD) v 6.0 software.

Molecular Docking
The molecular docking between the FDA approved potential S1 inhibitor drugs and SARS-CoV-2 S1 target proteins of different variants were studied using Molegro Virtual Docker (MVD, Molexus IVS, Molexus IVS, Rørth Ellevej 3, Rørth, DK-8300 Odder, Denmark) 2013.6.0 software [19]. The docking process was carried out using a 20 Å grid radius adjusted to contain amino acids residues of the S1 protein variants, which were identified as the RBD domain of SARS-CoV-2 S1 protein interacting with the ACE2 receptor [20]. The grid resolution was 0.30 Å. The number of runs for each docking process was 10, and the max iterations were 1500 with an energy threshold of 100; the maximum population size generated was 50; the maximum number of poses generated was 5. The docking score in the MVD algorithm, is presented as an arbitrary unit (MolDock Score). The best conformations for each docking process were selected based on the lowest score [21]. Gaps and insertions were detected in the omicron variant [7,22], which were described based on Wuhan-Hu-1 numbering.

Results
Camostat showed low binding affinities to beta (−88.193) and kappa (−91.036) variants, as well as the original Wuhan strain (−94.755), with higher binding affinities to other variants ranging from −102.745 to −123.722 ( Nelfinavir showed good binding affinities to most of the SARS-CoV-2 variants; however, it showed the lowest binding affinity to the beta variant, followed by the alpha variant. Nelfinavir showed higher affinities to SARS-CoV-2 variants in comparison to umifenovir for the Wuhan original strain, in addition to delta, eta, iota, lambda and mu variants, while the reverse was detected with alpha and gamma variants ( Table 2). Higher hydrogen bonds and more amino acids residues were detected in the interaction of nelfinavir with different variants in comparison to umifenovir ( Table 2, Supplementary Material Files S5 and S6). Interestingly, umifenovir binds to the different variants with hydrogen bonds to one or two amino acids only. Cefoperazone showed the highest binding affinities to SARS-CoV-2 strains and variants, followed by the ceftazidime (Table 3) Table 3).   The docking results revealed that all tested drugs possess moderate to high binding energies to the binding site (RBD) of the S1 protein for all different variants. Among all tested drugs, cefoperazone was found to possess the highest binding energy towards the RBD of the S1 protein of all eleven variants, ranging from a −109.913 MolDock score against the BetaGHB.1.351 variant, to a −185.011 MolDock score against the GammaGRP.1 variant. The binding interaction of cefoperazone includes multiple hydrogen bonds with the amino acid. Residues Arg451, Arg454, Lys455, Phe456, Ser466, Ile469, Gln471 and Arg509 are common to the S1 RBD for all eleven variants. Ceftazidime was the second-best drug in terms of binding affinity towards S1 RBD of the investigated variants scoring binding energy, ranging from a −109.913 MolDock score for the BetaGHB.1.351 variant, to a −144.263 MolDock score for the GammaGRP.1 variant. On the other hand, fenofibrate showed, on average, the least binding affinity towards the RBD of the S1 protein of all eleven variants, The docking results revealed that all tested drugs possess moderate to high binding energies to the binding site (RBD) of the S1 protein for all different variants. Among all tested drugs, cefoperazone was found to possess the highest binding energy towards the RBD of the S1 protein of all eleven variants, ranging from a −109.913 MolDock score against the BetaGHB.1.351 variant, to a −185.011 MolDock score against the GammaGRP.1 variant. The binding interaction of cefoperazone includes multiple hydrogen bonds with the amino acid.
Residues Arg451, Arg454, Lys455, Phe456, Ser466, Ile469, Gln471 and Arg509 are common to the S1 RBD for all eleven variants. Ceftazidime was the second-best drug in terms of binding affinity towards S1 RBD of the investigated variants scoring binding energy, ranging from a −109.913 MolDock score for the BetaGHB.1.351 variant, to a −144.263 MolDock score for the GammaGRP.1 variant. On the other hand, fenofibrate showed, on average, the least binding affinity towards the RBD of the S1 protein of all eleven variants, ranging from a −68.199 MolDock score for the AlphaGRYB.1.1.7 variant, to a −114.814 MolDock score against the S1 RBD of the omicron variant. The binding interaction of fenofibrate showed, on average, three to four hydrogen bonds with the amino acid residues in the RBD of the S1 protein of the different investigated variants.
Overall, all tested repurposed drugs possess moderate to high binding affinities towards the RBD of the S1 protein subunit of all variants. Most of the interacting amino acid residues of the receptor fall within the RBD (438-506). This finding suggests that these drugs are capable of hindering the interaction of the spike protein to the human ACE2 receptor, consequently inhibiting the viral activity.

Discussion
Drug repurposing has proven to be a key strategy for finding FDA approved drugs as potential SARS-CoV-2 inhibitors, and different approaches have been used in this respect. For example, virtual screenings of FDA approved drug libraries were applied to investigate the binding energies of these drug libraries against different virus protein targets. In addition, molecular docking of specific FDA approved drugs against specific viral protein targets was also reported as a second approach to identify potential anti SARS-CoV-2 repurposed drugs. Some of the screened drugs were found to possess potential activity in-vitro, as well as in clinical studies. The spike glycoprotein was identified as a prime drug target in numerous studies using virtual screening methodologies. In this study, we focused on the potential inhibition of the selected drugs to the reported binding site of the S1 subunit of the different variants of the SARS-CoV-2 spike glycoprotein using a molecular docking approach.
In the current study, we investigated whether or not there is a variability among different virus variants in binding to different repurposed drugs. The latest VOC, omicron, possesses 15 amino acid substitutions in the RBD; these substitutions include G339D, S371L, S373P, S375F, K417N, N440K, G446S, S477N, T478K, E484A, Q493R, G496S, Q498R, N501Y and Y505H [7]. Three such amino acid substitutions, K417N, E484A and N501Y, were shared with those present in the beta (B.1.351) VOC that was dominant in South Africa [23]. Meanwhile, the gamma VOC, which spread in the Brazil and Amazon region, shared the N501Y amino acid substitution with many VOCs, including omicron and beta. It also possesses two other amino acid substitutions in the same positions in omicron, but with different amino acids: T417 and K484 [24]. Many gamma VOC strains had S477N, which increased the infectivity and spread of the virus [25]. Meanwhile, the delta variant possesses S477G (present in all VOC and VOI except iota and omicron) and T478K. Interestingly, serine at the position of 477 is frequently exposed to changes, and S477G and S477N are the most common mutations detected; both lead to increased binding to the ACE2 receptor [26]. The delta VOC possesses T478K and S477G, and was found to partially escape from neutralizing antibodies from patients vaccinated with either Pfizer or AstraZeneca vaccines, since it is neutralized 3-5-fold lower than that of the alpha VOC [27]. E484K was reported in Mu and eta, while E484Q was detected in the kappa VOC. Both mutants were found to escape neutralization by bamlanivimab [28,29]. Partial neutralization/resistance was associated with E484A/K [30,31] and Q493R [32]. Immune pressure was assumed to be responsible for the evolution of both E484K and E484Q [33]. Omicron was found to get rid of the K417N mutation [22] as it attenuates its binding to the ACE by about 4-fold [34]. N440 is an interface residue that plays a role in receptor recognition. The N440K variant is an escape mutant that was found to be evolved under selective pressure by using human monoclonal antibody C135 and showed resistance to it [35], which might increase reinfection and decrease vaccine efficiency. Such a mutant has been observed during viral passaging experiments in the presence of convalescent plasma and provides additional immune escape in vitro [4], and has been detected in 3940/17,046 omicron VOC to date. However, it is also detected in 47 isolates of delta VOC, 19 lambda VOI, 6 gamma VOC, a single beta VOC and 37 alpha VOC [36]. Interestingly, none of the 15 amino acids in the omicron variant were incorporated with the docking drugs tested in the current study; however, the neighboring amino acids in the RBD were found to be involved in the binding to the different drugs. This finding suggests that these drugs can hinder the interaction of the spike protein to the human ACE2 receptor, consequently inhibiting the viral activity. To the best of our knowledge, only two out of the seven drugs that were used in this study have been previously reported to virtually screen, via molecular docking, against the S1 of the wild-type virus. Umifenovir was found to possess considerable binding affinity to the S glycoprotein (−7.47 kcal/mol), forming a 1H bond with the Lys462 amino acid, in addition to several other hydrophobic interactions to the RBD of the S1 subunit. Meanwile, ceftazidime showed a binding affinity of −6.36 kcal/mol, forming three hydrogen bonds with LYS403, GLY504 and TYR505 amino acid residues of the RBD of the S1 protein [37,38].
Both camostat and nafamostat were found to potently inhibit SARS-CoV-2 infection in cultured human airway epithelia. Nafamostat showed a more potent effect than camostat. It also inhibited SARS-CoV-2 infection and improved disease outcomes in two COVID-19 mouse models [39]. The action is related to inhibition of the host serine protease TMPRSS2. The latter is responsible for priming of the SARS-CoV-2 S, which is necessary for virus fusion and entry to the host cell [1]. However, in our study, nafamostat was not superior to camostat, and both showed high affinities to the SARS-CoV-2 S1 protein, with slight differences of the affinities between the two compounds. The determinant of affinities is based on the SARS-CoV-2 variant type.
Fenofibrate is used as an anti-hyperlipidemic, and was found to inhibit the binding of the SARS-CoV-2 spike protein to the human ACE2 receptor and reduces the virus replication by ∼60% in Vero cells after 24 h post-infection [10]. It possesses an anti-inflammatory effect by reduction the activities of CXCL10, IL17, CCL2 and CCL20. It inhibits the phospholipid in infected cells as it possesses peroxisome proliferator-activated receptor alpha (PPARα) agonist activity, thus affecting the pathways of lipid metabolism in the lung cells of COVID-19 patients [40,41].
Dual effects on both the spike and ACE2 receptor are suggested to be among the potential antiviral activities of fenofibrate [10]. In the current study, fenofibrate showed variable activities with different SARS-CoV-2 variants, with the highest activity recorded in omicron and eta variants, and the lowest in the alpha variant. No link was found between the antiviral effect and the inhibition of cholesterol synthesis when compared with different anti-hyperlipidemic drugs. Davides et al. suggested that fenofibrate is less likely to have resistance against newly emerging strains of SARS-CoV-2; however, the current study suggested that wide variation does exist in regard to the affinity of fenofibrate to different variants [10].
Nelfinavir is a protease inhibitor that is used in the triple therapy combination used for HIV. It was found to inhibit post-entry antiviral activity against both SARS-CoV infection [42] and SARS-CoV-2 [43,44] by inhibiting the main protease. Nelfinavir mesylate might bind inside the S protein with subsequent inhibition virus entry and spike mediated cell fusion [13]. Umifenovir, known as arbidol, is involved in the reduction of the replication of SARS-CoV-2 at both viral entry and post-entry stages [45,46]. Umifenovir alone did not improve the clinical outcome of COVID-19 patients [45]. Our finding that umifenovir efficiently binds to most of the current SARS-CoV-2 variants, except the delta and beta variants, agrees with the study that reported a positive RNA test of shorter duration with umifenovir treatment in comparison to the lopinavir/ritonavir treated group [47]. Meanwhile, in comparison to the nelfinavir, the latter was found to bind more efficiently than umifenovir.
Both cefoperazone and ceftazidime are bactericidal antibiotics that contain a betalactam ring. The latter renders them susceptible to the beta-lactamase enzyme. Recently, ceftazidime (over 300 µM) was found to inhibit SARS-CoV-2 infection by binding to the S1 RBD of the SARS-CoV-2 [16]. In the same study, efficacy of different cephalosporins were compared, and ceftazidime was found to be more potent than other cephalosporins, including cefoperazone. However, in the current study, we found the reverse, since cefoperazone was found to more potently bind to the RBD than ceftazidime. Interestingly, high binding variabilities of cefoperazone were detected among different variants.

Conclusions
In conclusion, all tested repurposed drugs possess moderate to high binding affinities towards the RBD of the S1 protein subunit of all variants. Most of the interacting amino acid residues of the receptor fall within the RBD (438-506). The nature of the neighboring amino acids to RBD of the S1 provide important clues for the design of targeted inhibitors and/or peptidyl disruptors. Cefoperazone showed the highest binding affinities to the SARS-CoV-2 S1 protein subunit, followed by ceftazidime, nelfinavir, camostat, nafamostat mesylate and fenofibrate.