Computational Screening of FDA Approved Drugs from ZINC Database for Potential Inhibitors of Zika Virus NS2B/NS3 Protease: A Molecular Docking and Dynamics Simulation Study

Zika virus is a mosquito borne pathogen with a single strand RNA genome. Human infection with this virus is usually asymptomatic, however outbreaks reported in both Pacific region and Latin America have been associated with increase in frequency of microcephaly in newborns and fetuses of infected mothers. Also, the incidence of Guillain-Barré syndrome had also increased among adults with Zika virus infection. Currently, neither vaccine nor antiviral drug has been developed against Zika virus. Structure based virtual screening can be employed, through drug repurposing strategy, to accelerate the identification of potential anti-Zika virus candidates. As such, virtual screening of approved drugs against Zika virus NS2B/NS3 protease can help to recognize new hits capable of hindering viral ability to replicate and evade immune system of the host. In this computational study, we have screened 1615 FDA approved drugs against NS2B/NS3 protease enzyme of Zika virus by using both molecular docking and dynamics simulation. Our virtual screening results indicate that the anti-muscarinic agent Darifenacin and the anti-diarrheal agent Original Research Article Odhar et al.; JPRI, 33(39B): 308-319, 2021; Article no.JPRI.72038 309 Loperamide may have a promising capacity to inhibit Zika virus NS2B/NS3 protease. According to molecular docking and dynamics simulation, these two approved drugs have good binding capacity to NS2B/NS3 as reported by docking energy of binding and MM-PBSA binding energy. In addition, both Darifenacin and Loperamide were able to maintain close proximity to protease crystal throughout simulation period. However, invitro evaluation of these two drugs against Zika virus NS2B/NS3 protease is required to confirm these computational results.


INTRODUCTION
Zika virus (ZIKV) is a single stranded and positive sense RNA virus, Zika virus is a member of the family Flaviviridae along with other viruses like dengue virus and yellow fever virus [1,2]. Zika virus was isolated from a rhesus monkey in the Zika forest for the first time in Uganda in 1947 [3]. For a half century, sporadic cases of human infection with Zika virus were reported in both Africa and southeast Asia [4,5]. Then, in 2007 and 2013, two outbreaks of Zika virus were recognized in Yap island and French Polynesia respectively [6,7]. Later on, several outbreaks of ZIKV were recorded in Brazil between 2014 and 2016. These outbreaks of Zika virus were accompanied by a national increase of congenital neurological abnormalities [8]. As a result, the World Health Organization (WHO) have considered Zika virus infections and its possible neurological complications a worldwide public health emergency in 2016 [9]. Recent resurgence of Aedes mosquito, the known vector of Zika virus, in Mediterranean and Black sea regions can contribute significantly to future outbreaks of this virus in these areas [10].
Most cases of Zika virus infections are asymptomatic however some patients may be present with mild symptoms like fever, rash, fatigue and arthralgia. These mild symptoms can last for 2 to 7 days, and the incubation period of Zika virus infection can range between 3 and 12 days [3,11]. However, in utero infection with Zika virus during pregnancy can result in serious fetal neurological malformations like microcephaly [12]. Also, it is believed that infection with Zika virus can increase the incidence of Guillain-Barré syndrome. This syndrome is considered an autoimmune disease that influence peripheral nervous system leading to muscles weakness and possible paralysis [13]. As Zika virus can cause illness with nonspecific symptoms, diagnosis of ZIKV infection is still challenging. The most accurate diagnostic method depends on detection of Zika virus RNA in human body fluids by using reverse transcription polymerase chain reaction (RT-PCR). The application of serological tests like enzyme-linked immunosorbent assay (ELISA) is considered challenging due to ZIKV antibodies cross reactivity with those antibodies generated against other viruses of Flaviviridae family [14].
Zika virus is considered a mosquito borne disease, the virus can also be transmitted through blood transfusion and sexual contact. Vertical transmission of Zika virus is also possible where the mother can pass the virus to her fetus during pregnancy or to her newborn during delivery [15].
The RNA genome of Zika virus encodes a single polypeptide chain with a length of 3423 amino residues, this large polypeptide can be then cleaved by both host and viral protease enzymes into structural and non-structural proteins. Zika virus has three structural proteins and these are envelope (E), capsid (C) and pre-membrane (prM) proteins. These structural proteins are important for the creation of viral particles. Zika virus has also seven non-structural proteins including NS1, NS2A, NS2B, NS3, NS4A, NS4B, NS5. These non-structural proteins are essential for replication and assembly of viral particles. These proteins can also help the virus for evasion of immune system [1,16].
To date, neither vaccine nor drug has been developed against Zika virus. Different vaccine platforms like DNA plasmid and adenovirus based vector have been applied in animal models to induce protective immune response against Zika virus envelope (E) protein [3]. Regarding the development of anti-ZIKV drug, the NS2B/NS3 protease seems to be a potential molecular target to design and discover a candidate capable of inhibiting Zika virus ability to replicate inside host cells and escape immune defenses. This protease enzyme is composed of two main subunits, the first one is the nonstructural protein NS3 with a protease activity at its N-terminus and RNA helicase at its C terminus. The second subunit is another non structural protein, namely NS2B, membrane bound. NS2B subunit is considered important for proper orientation and functioning of Zika virus protease enzyme [16,17] dimensional view for Zika virus NS2B/NS3 protease can be seen in Fig. 1.
Structure based virtual screening is a computer based approach capable of identifying hit candidates with potential activity against specific targets by using molecular docking and dynamics simulation [18]. Virtual screening approach can be applied to facilitate drug repurposing efforts. In drug repurposing, new uses are identified for an already approved drugs and b original indications of these drugs. Thus, virtual drug repurposing can save both time and cost in drug development projects [19]. In this trend, two 310 terminus and RNA helicase at its Cterminus. The second subunit is another nontructural protein, namely NS2B, that is membrane bound. NS2B subunit is considered ation and functioning [16,17]. A threedimensional view for Zika virus NS2B/NS3 Structure based virtual screening is a computerbased approach capable of identifying hit al activity against specific targets by using molecular docking and dynamics . Virtual screening approach can be applied to facilitate drug repurposing efforts. In drug repurposing, new uses are identified for an already approved drugs and beyond the original indications of these drugs. Thus, virtual drug repurposing can save both time and cost in . In this trend, two attempts for computer aided drug repurposing were able to recognize both the antihistamine chlorcyclizine and the antimalarial hydroxychloroquine as potential anti candidates.
Both chlorcyclizine and hydroxychloroquine were predicted to have an inhibitory capacity against NS2B/NS3 protease according to virtual screening of approved drugs. Also, these two approved drugs appeared to be able to inhibit Zika virus viability in cell culture assay [17,20].
In this computational study, we have screened FDA approved drugs database against NS2B/NS3 protease crystal of Zika virus by using molecular docking and dynamics simulation. Our aim is to identify potential inhibitors of ZIKV protease enzyme through virtual drug repurposing.
dimensional illustration for NS2B/NS3 protease crystal (PDB: 5H4I) of Zika NS2B subunit is colored in violet while NS3 subunit is colored with rainbow gradient. The C terminus of NS3 is shown in red and the N-terminus is colored by blue. Key residues of NS3 protease active site are represented as labeled sticks ; Article no.JPRI.72038 attempts for computer aided drug repurposing were able to recognize both the antihistamine cyclizine and the antimalarial hydroxychloroquine as potential anti-ZIKV candidates.
Both chlorcyclizine and hydroxychloroquine were predicted to have an inhibitory capacity against NS2B/NS3 protease according to virtual screening of approved drugs.
these two approved drugs appeared to be able to inhibit Zika virus viability in cell culture In this computational study, we have screened FDA approved drugs database against NS2B/NS3 protease crystal of Zika virus by using ng and dynamics simulation. Our aim is to identify potential inhibitors of ZIKV protease enzyme through virtual drug dimensional illustration for NS2B/NS3 protease crystal (PDB: 5H4I) of Zika virus. NS2B subunit is colored in violet while NS3 subunit is colored with rainbow gradient. The Cterminus is colored by blue. Key residues of NS3 2. METHODOLOGY

Setting up Virtual Screening Scheme
The project steps for virtual screening are outlined in Fig. 2. In summary, FDA approved drugs library were screened by docking software against Zika virus NS2B/NS3 protease crystal. Then the best 25 hits were selected and subjected to safety profile assessment by using online clinical reference. After that, only those safe hits were submitted to molecular dynamics (MD) simulation study. The ligands th maintain both good binding energy and close proximity to protease crystal throughout simulation period were then selected for further elaboration.

Structure Based Virtual Screening
The steps of structure based virtual applied in this study is similar to those used in our previous projects [21][22][23]. Concisely, a library of FDA approved drugs was downloaded as SDF format from ZINC 15 database on June 01, 2021 [24]. This library contains 1615 chemical structures, and it was then uploaded to Mcule.com online discovery website for

ing up Virtual Screening Scheme
The project steps for virtual screening are . In summary, FDA approved drugs library were screened by docking software virus NS2B/NS3 protease crystal. Then the best 25 hits were selected and subjected to safety profile assessment by using online clinical reference. After that, only those safe hits were submitted to molecular dynamics (MD) simulation study. The ligands that can maintain both good binding energy and close proximity to protease crystal throughout simulation period were then selected for further

Structure Based Virtual Screening
The steps of structure based virtual screening applied in this study is similar to those used in . Concisely, a library of FDA approved drugs was downloaded as SDF format from ZINC 15 database on June . This library contains 1615 and it was then uploaded to Mcule.com online discovery website for screening against NS2B/NS3 protease crystal of Zika virus [25]. The NS2B/NS3 protease crystal with code 5H4I was downloaded from Protein Data Bank as PDB format [26,27] uploading the target crystal to Mcule.com, the drug discovery platform automatically prepared the crystal for docking by using AutoDock tools [28]. Preparing the uploaded crystal by Mcule.com did include removing water molecules and any bound ligand from PDB file. It d involve adding polar hydrogen and Gasteiger charges to the crystal. AutoDock Vina is embedded in Mcule.com to carry out structure based virtual screening [29]. For docking FDA approved drugs against protease crystal, we have used default parameters on Mcule.com website. The employed docking coordinates were (X: -8.5, Y: 4.0, Z: -17.0) and the area of binding site was (22 * 22 * 22) Angstrom. After docking, the hits were ranked according to their least binding energy to protease crystal. Finally, the top 25 drugs were selected for further consideration. Both PyMOL version 2.3 and Protein-Ligand Interaction Profiler (PLIP) web tool were used to visual three-dimensional and two-dimensional conformations for ligand protease complex with least binding ene [30][31][32].

Assessment of Safety Profile
We have used Medscape.com online clinical reference to evaluate the safety profile and the legal status for dispensing of the best 25 drugs of the virtual screening study [33]. Only those drugs with acceptable safety profile were then subjected to molecular dynamics (MD) simulation.

Molecular Dynamics (MD) Study
The top hits of the virtual screening study, with acceptable clinical safety profile, were then subjected to molecular dynamics (MD) analysis by using YASARA Dynamics version 20.12.24 [34]. The steps for MD study used here are similar to what we employed in our previous works [22,23,35]. For each hit, the ligandprotease complex with the least binding energy pose was submitted to MD simulation for 10 nanoseconds. The protocol of MD simulation did include optimization of hydrogen bonds and prediction of pKa to fine-tune protonation of amino acids at pH = 7.4 [36]. A 0.9% sodium chloride was used in MD simulation. To neutralize ligand-protease complex, an excess of sodium or chloride were added. To get rid of any clashes, both steepest descent and simulated annealing minimizations were utilized. The force fields applied during simulation were TIP3P for water, AMBER14 for solute, AM1BCC and GAFF2 for ligand [37][38][39]. Default parameters were used by AMBER, and the cutoff border for van der Waals forces was eight Angstrom [40]. For electrostatic forces, no cutoff limit was used due to the application of Particle Mesh Ewald algorithm [41]. Equations of motions were employed as multiple timestep of 1.25 femtoseconds and 2.5 femtoseconds for bonded and non-bonded interactions respectively at temperature of 298K and pressure of 1 atm [42]. After evaluation of solute Root Mean Square Deviation (RMSD) as a function of simulation time, the first 10 nanoseconds period was considered the equilibrium time and excluded from any further analysis. MD results for ligand movement RMSD and ligand conformation RMSD throughout simulation period were then plotted and visualized by using GraphPad Prism 8.0.2 [43].
The trajectories of MD simulation were then subjected for calculation of Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) binding energy [44]. A built in YASARA macro was employed to facilitate estimation of MM-PBSA binding energy by using AMBER14 force field. According to YASARA guideline, the more positive binding energy indicates better interactions [45]. YASARA macro for binding energy calculation employs the following equation: Binding Energy = EpotRecept + EsolvRecept + EpotLigand + EsolvLigand − EpotComplex − EsolvComplex

RESULTS AND DISCUSSION
For this structure based virtual screening study, the output hits were ordered according to their least energy of binding to Zika virus protease crystal. As seen in Table 1, we have selected only the top 25 hits and their docking energy was ranging between -8.5 and -7.3 Kcal/ mol. Based on this table, we can notice that all these hits are prescription only medications (POM) except the antidiarrheal agent Loperamide which is considered an over the counter (OTC) drug. According to a published cell culture-based project, low micromolar concentration of Loperamide was sufficient to inhibit the replicative capacity of Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV), Middle East Respiratory Syndrome Coronavirus (MERS-CoV) and human coronavirus 229E [46]. Of interest, is the appearance of the antiretroviral drug Raltegravir and the penicillin antibiotic Nafcillin among these 25 hits. It is also worth mentioning that both Conivaptan and Azelastine were predicted to be potential inhibitors of SARS-CoV-2 main protease by our research group in a previous project [22]. Both Conivaptan and Azelastine were later proved to have activity against SARS-CoV-2 according to in vitro studies [47,48]. Additionally, Darifenacin is reported among the 25 hits in Table 1. The antimuscarinic agent Darifenacin was previously predicted to have inhibitory capacity against several conserved molecular targets of SARS-CoV-2 by using both docking and dynamics simulation programs [49]. Then by reviewing the safety profile of these 25 drugs in Medscape clinical reference, only 14 hits were selected for further consideration and analysis. We have excluded 11 drugs that are approved by FDA for carcinoma, leukemia, schizophrenia, epilepsy and acne due to our concerns about the safety of these agents among general population. In the next step, only those 14 top hits with acceptable safety profile were then subjected to molecular dynamics (MD) simulation. MD study is considered an important tool in virtual screening field because it can predict ligandtarget interactions by considering the flexibility of target [50]. For each hit, the ligand-protease complex with least binding energy pose was submitted to MD simulation for 10 nanoseconds.
The summary of molecular dynamics study results can be seen in Table 2. In this table, we have reported the average binding energy by using Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) calculation. According to YASARA Dynamics guideline, the more positive binding energy means better binding of the ligand. Also, we have presented in Table 2 the mean, minimum and maximum ligand movement Root-Mean-Square Deviation (RMSD) in Angstrom. The ligand movement RMSD was generated by superposing the receptor on its reference structure as a function of simulation period. Low ligand movement RMSD usually indicate close proximity to the target and hence better binding of the ligand. As such, potential inhibitor of NS2B/NS3 protease should have both low ligand movement RMSD and more positive MM-PBSA binding energy. By applying this condition to Table 2, only the anti-muscarinic agent Darifenacin and the anti-diarrheal agent Loperamide were selected as final hits for further evaluation.

Fig. 3. Three-dimensional and two Loperamide against Zika virus NS2B/NS3 protease crystal. Hydrogen bond is presented as blue line while hydrophobic bond is displayed as dashed line
When examining docking images for Darifenacin and Loperamide in Fig. 3, we can notice that both drugs occupy almost same location within active site of NS2B/NS3 protease. According to Fig. 3

dimensional and two-dimensional views for the docking of (A) Darifenacin and (B) Loperamide against Zika virus NS2B/NS3 protease crystal. Hydrogen bond is presented as blue line while hydrophobic bond is displayed as dashed line
When examining docking images for Darifenacin , we can notice that both drugs occupy almost same location within active site of NS2B/NS3 protease.
, both Darifenacin and Loperamide are involved in hydrophobic interactions with valine 155 and tyrosine 161 dues of NS2B/NS3 protease complex. However, Darifenacin seems to be able to form hydrogen bonding with alanine 132. The alanine 132 is very close to serine 135 residue, the later residue is considered a key residue for the protease activity of NS2B/NS3 com the ability of Darifenacin to form hydrogen bond with alanine 132 may influence access of substrate to serine 135 and adversely influence protease activity of NS2B/NS3 enzyme complex.

Darifenacin and (B) Loperamide against Zika virus NS2B/NS3 protease crystal. Hydrogen bond is presented as
hydrogen bonding with alanine 132. The alanine 132 is very close to serine 135 residue, the later residue is considered a key residue for the protease activity of NS2B/NS3 complex [17]. So, the ability of Darifenacin to form hydrogen bond with alanine 132 may influence access of substrate to serine 135 and adversely influence protease activity of NS2B/NS3 enzyme Later on, we analyzed both ligand movement RMSD and ligand conformation RMSD throughout simulation time as seen in Fig. 4 and Fig. 5 respectively. As evident in Fig. 4, Loperamide can maintain a closer proximity throughout 10 nanoseconds than Darifenacin can do. In fact, it is obvious that Darifenacin movement RMSD has more fluctuations than that for Loperamide.
Regarding ligand conformation RMSD as a function of simulation duration, as seen in Fig. 5, it is clear that Darifenacin has undergo sharp fluctuation in conformation at the beginning of simulation and around 8 nanosecond point. This may explain the deviation experienced by Darifenacin in ligand movement RMSD in the same simulation points as can be noticed in Fig. 4. It is worth to clarify that ligand conformation RMSD was calculated by superposing the ligand on its reference structure throughout duration of simulation.

CONCLUSION
In this study, we have applied structure based virtual screening approach to repurpose FDA approved drugs against Zika virus protease. Our project scheme did involve docking of 1615 approved drugs against NS2B/NS3 protease crystal, followed by evaluation of safety profile and dispensing regulations for the top 25 hits of docking output. Finally, only those top hits with acceptable safety records were subjected to molecular dynamics simulation. Here, we report that both antimuscarinic agent Darifenacin and antidiarrheal agent Loperamide may have potential capacity to inhibit NS2B/NS3 protease of Zika virus. Both drugs were able to maintain close proximity to protease active site throughout simulation. In addition, MM-PBSA calculation showed that both Darifenacin and Loperamide have favorable binding energy during simulation period. The antimuscarinic drug Darifenacin had more positive average MM-PBSA binding energy than did Loperamide. Also, Darifenacin was involved in hydrogen bonding with alanine 132 and this binding behavior may negatively impact substrate access to serine 135. The later residue is considered important for activity of protease enzyme. When compared to Darifenacin, Loperamide displayed a closer proximity to protease active site with lower tendency to fluctuate in terms of ligand movement RMSD and ligand conformation RMSD. However, these computational findings must be confirmed through invitro evaluation.

DISCLAIMER
The products used for this research are commonly and predominantly use products in our area of research and country. There is absolutely no conflict of interest between the authors and producers of the products because we do not intend to use these products as an avenue for any litigation but for the advancement of knowledge. Also, the research was not funded by the producing company rather it was funded by personal efforts of the authors.