In silico identification of natural antiviral compounds as a potential inhibitor of chikungunya virus non-structural protein 3 macrodomain

Abstract Chikungunya Virus (CHIKV) is having a major impact on humans with potentially life-threatening and debilitating arthritis. The lack of a specific antiviral drug against the CHIKV disease has created an alarming situation to identify or develop potent chemical molecules for its remedial measures. Antiviral therapies for viral diseases are generally expensive and have adverse side effects. Plant-based antiviral natural compounds are the most suitable and best alternative of current antiviral drugs because of less toxicity. In the present study, non-structural protein 3 macrodomain (nsP3MD) of the CHIKV that is essential for virus replication has been selected for anti CHIKV drug target. The compounds were identified using molecular docking, virtual screening and further evaluated by molecular dynamics (MD) simulation studies. The binding mechanism of each compound was analyzed considering the stability and energetic parameter. We have found six plant-based natural antiviral compounds Baicalin, Rutaecarpine, Amentoflavone, Apigetrin, Luteoloside, and Baloxavir as strong inhibitors of nsP3MD of CHIKV. ADMET prediction and target analysis of the selected compounds showed drug likeliness of these compounds. MD simulation studies indicated energetically favorable complex formation between nsP3MD and the selected antiviral compounds. Furthermore, the structural effects on these substitutions were analyzed using the principles of each trajectory, which validated the interaction studies. Our analysis suggests a very high probability of these compounds to inhibit nsP3MD of CHIKV and could be evaluated for Chikungunya fever drug development. Communicated by Ramaswamy H. Sarma

Presently the treatment of Chikungunya fever is based on palliative care, using analgesics for pain and non-steroidal anti-inflammatory drugs to reduce arthralgia in chronic infections (Ghildiyal & Gabrani, 2020;Subudhi et al., 2018). Nevertheless, some of these drugs can have serious side effects upon prolonged use (Kovacikova & van Hemert, 2020). The lack of clinically approved therapeutics and adequate control measures for Chikungunya fever warrants the development of safe and effective antiviral therapy .
CHIKV has approximately 11.7 kb positive-sense, single strand (þss) RNA genome that codes for four non-structural proteins (nsP1-4) and five structural proteins, (one capsid (C), three envelope glycoproteins (E1-E3) and one 6 K peptide) (Abraham et al., 2018;Eckei et al., 2017;Malet et al., 2009;McPherson et al., 2017). All four nsP3 proteins in combination with host factors form the replication machinery, tethered to cytoplasmic vacuoles and have distinct enzymatic activities responsible for viral RNA replication (Ghildiyal & Gabrani, 2020;Pietil€ a et al., 2017;Rupp et al., 2015). The nsP3 has 3 domains: a very preserved N-terminal macrodomain (MD) that is present in numerous þ ss RNA viruses, a zinc-binding oligomerization domain, and a poorly conserved, unstructured, acidic, and highly Ser/Thr phosphorylated C-terminal hypervariable domain (Abraham et al., 2018). The nsP3 encompasses the N-terminal of nsP3 MD has ADPr 1"-phosphate phosphatase activity, and it can bind to DNA, RNA, poly(ADPr) and ADPr (Malet et al., 2009;McPherson et al., 2017). Recent work has demonstrated that CHIKV nsP3 MD possesses mono(ADPr) hydrolase activity, which specifically removes the ADPr moiety from glutamate and aspartate residues but not from lysine residues in substrate proteins (Abraham et al., 2018). The protein fold in the macrodomain is remarkably conserved. Mutants affecting ADPr binding and hydrolase activity slow viral replication in mammalian cells and reduce CHIKV virulence in mice (Abraham et al., 2018). These results suggest that the nsP3 MD is critical for CHIKV replication and can be a good target for anti-CHIKV drug discovery.
In the present study, we selected natural plant-based compounds which were reported to have antiviral properties against different viruses, but not for CHIKV. The present study utilizes a systematic approach to find natural anti CHIKV compounds. These compounds might act as promising inhibitors against nsP3 MD of CHIKV. Through an extensive in silico approach, the aim of this study is to understand the underlying inhibitory mechanisms of these compounds. In order to accomplish this, molecular docking and MD simulation studies have been used to calculate various structural parameters, including the estimated binding free energy (DG) of the compounds, Root Mean Square Deviation (RMSD), Root Mean Square Fluctuation (RMSF), Radius of Gyration (Rg), Solvent Accessible Surface Area (SASA), Principal Component Analysis (PCA) and the intermolecular hydrogen bonds (H-bonds) for free and inhibitor bounded CHIKV nsP3 MD . Further in vivo and in vitro studies of these compounds will provide inroads for the development of novel anti-CHIKV nsP3 MD inhibitors that emerge as good candidate drugs for Chikungunya fever.

Docking
The 3 D SDF structure library of 801 antiviral, flavonoids and alkaloids compounds were downloaded from the SELLEKCHEM database (https://www.selleckchem.com). Other 19 Plant-based natural compounds known to have reported antiviral and antibacterial activity (Withanolide, Withaferin A, Aloesin, Aloenin Cannabidiol, Piperine, Curcumin, Sitosterol, Berberine, Astragalin, Nicotiflorin, Lupeol, Quercetin, Ursolic Acid, Apigenin, Gingerol, Shogaol, Dehydroandrographolide, and Nimbin) were also selected for molecular docking and download from the PubChem database (https://pubchem. ncbi.nlm.nih.gov/). These compounds were imported into OpenBabel using the PyRx Tool (Dallakyan & Olson, 2015) and subjected to energy minimization(EM). The EM was performed with the universal force field (UFF) using the conjugate gradient algorithm. The number of steps was set to 200, and the number of steps for the update was set to 1. In addition, the minimization was set to stop at an energy difference of less than 0.1 kcal/mol. The structures were converted to the PDBQT format for docking.
The structure of the CHIKV nsP3 MD protein complex with ADPr (PDB-ID: 3GPO at a resolution 1.90 Å) was used in the present study. The ligand ADPr was selected and removed from the structure for further docking studies. The receptor preparation was done using the protein preparation wizard tool of maestro Schr€ odinger (Sastry et al., 2013). Hydrogens were added and optimized by a hydrogen-bonding network in order to find the Histidine protonation state.

In silico ADMET and target prediction
Pharmacokinetic parameters of ADMET (absorption, distribution, metabolism, excretion, and toxicity) play a significant role in the discovery of new drug candidates, as many invented drugs fail in the development process. Hence, the in silico ADMET evaluation was computed using online SwissADME (https://www.swissadme.ch) (Daina et al., , 2017Daina & Zoete, 2016) and PreADMET (Lee et al., 2003(Lee et al., , 2004 web tools. The pkCSM online database (http://biosig.unimelb.edu.au/ pkcsm/prediction)was used for toxicology prediction analysis (Pires et al., 2015). The website provides details of toxicology effects in the fields of AMES Toxicity, Human Maximum Tolerance Dose, hERG-I Inhibitor, hERG-II Inhibitor, LD50, LOAEL, Hepatotoxicity, Skin Toxicity, T. pyriformis toxicity and Minnow Toxicity.

MD simulations
MD simulations of CHIKV nsP3 MD -apo and complexed with identified two natural antiviral compounds (Baicalin and Amentoflavone binding score À10.8 and À10.1 kcal/mol respectively) were performed using the GROMACS software package using the GROMOS96 54a7 force field (Pronk et al., 2013;Van Der Spoel et al., 2005). The best pose of each CHIKV nsP3 MD complex was used for MD simulation studies. The ligand topology was generated by the PRODRG server (Schuttelkopf & van Aalten, 2004). Each complex was centered in a dodecahedron box with a minimum edge distance of 10 Å from the edge. The dodecahedron box was solvated with water molecules and 0.15 M NaCl. After system preparation, the system was energy minimized to avoid any steric hindrance by choosing the steepest descent method for a maximum of 5,000 steps (Fmax < 100 kJ/mol/nm).
Further, to equilibrate the system at a constant temperature of 300 K in a two-step ensemble process, i.e. NVT and NPT were used for 100 ps. Finally, all systems were introduced to MD simulation. For analysis, conformations of CHIKV nsP3 MD in each system were studied for the whole 200 ns trajectory. The trajectories were analyzed using g_gyrate, g_hbond, g_rmsf, g_rms and trajconv tools of GROMACS. All presentation was prepared using UCSF Chimera, Origin 6.0 and Xmgrace (Deschenes, 2000;Pettersen et al., 2004).

Docking
The structure-based molecular docking approach was performed on 820 compounds (347 antiviral,185 flavonoids, 269 alkaloids and 19 plant-based natural compounds) to identify potential CHIKV nsP3 MD inhibitor. These selected compounds have therapeutic potential against many infectious diseases, including anti-malarial and antiviral activities.
Blind docking resulted in docking of all of the 820 compounds and ADPr in the same cavity of the CHIKV nsP3 MD chain A (Figure 1 and Table S1). Primary screening of the docked complexes showed that out of the 820 compounds, 42 compounds (23 compounds from antiviral, flavonoids, and alkaloids library and 19 plant-based natural compounds) had a docking score ranging from À11.1 to À10.0 kcal/mol and À9.5 to À6.8 kcal/mol, respectively. The observed binding energy of ADPr with the CHIKV nsP3 MD complex was À8.6 kcal/mol (Table S1). Generally a compound is predicted to have activity against a protein when it has binding energy less than À6.0 kcal/mol (Shityakov & F€ orster, 2014).
The binding energies and interaction with CHIKV nsP3 MD with remaining six natural compounds (Narirutin, Scutellarin, Complanatuside, Eriocitrin, Oroxin B and Homoorientin) and other four plant-based natural compounds, Berberine, Quercetin, Curcumin, Apigenin showed anti CHIKV activity in vivo also given in the Table S1.

In silico ADMET prediction
ADMET predictions of the twelve selected plant-based natural compounds were carried out. Parameters such as MW (molecular weight), HBA (hydrogen bond acceptor), HBD (hydrogen bond donor), n-rot-b (number of rotatable bonds), TPSA (topological polar surface area), LogPo\w (partition coefficient), LogS (aqueous solubility), BBB (blood-brain barrier permeability), CYP2D6(cytochrome-P2D6 inhibitor), Lipinski rule of 5, hERG inhibition, HIA (human intestinal absorption) and PAINS alert were studied. The identified compounds were predicted to be non-toxic and showed lead likeliness. The ADMET predictions of all selected natural compounds are summarized in Tables S3 and S4 and Figures S4  and S5. Figure 1. 2 D structure of the best twelve natural compounds identified through CHIKV nsP3 MD docking study visualized using maestro (Schr€ odinger). Compound highlighted with blue color used for MD simulation study. All 2 D structures are representing with compound number, compound name and binding affinity in (kcal/mol)).
Of the twelve natural compounds studied, one compound (Rutaecarpine) showed less than ten hydrogen bond acceptors (HBA). In addition, less than five hydrogen bond donors (HBD) were found in Rutaecarpine and Baloxavir. The number of rotatable bonds was less than 10. The topological polar surface area (TPSA) was observed between 47.53 to 257.68 Å2 (Table S3). The partition coefficient between n-octanol and water (LogPo/w) is the standard descriptor for hydrophobicity. All twelve natural compounds studied were found to have Log P values less than 5 (ranging between À3.02 and 3.62), showing good permeability across the cell membrane (Table S3). These parameters showed that the studied compounds obeyed Lipinski's rule of five (Lipinski, 2000) with only three violations and, therefore, could be used as orally active antiviral agents.
Drug solubility is another significant property that influence absorption and a water-soluble compound facilitates the ease of formulation and oral administration in order to deliver enough quantity of the active ingredient (Daina et al., , 2017(Daina et al., , 2019Daina & Zoete, 2016;Gfeller et al., 2014). All selected compounds were found water soluble. Human intestinal absorption (HIA) is the sum of the absorption and  bioavailability that is estimated from the ratio of excretion or cumulative excretion in bile, urine and feces. Drugs are considered well absorbed if the intake lies between 70-100%. All the selected natural compounds showed an intestinal absorption between 1.42-91.64%, as shown in Table S3.
None of the compounds crosses the blood-brain barrier with the exception of Rutaecarpine, thereby decreasing the possibility of neurotoxicity (Table S3 and Figure S4). In addition, the results showed that none of the twelve selected natural compounds inhibited CYP2D6, a crucial isoenzyme belonging to the superfamily of cytochrome P450 (CYP450) ( Table S3). CYP2D6 is responsible for metabolizing most of the drugs in the liver (Rydberg & Olsen, 2012). Inhibition of CYP2D6 causes drug-drug interactions are leading to toxic effects (Daina et al., 2017). Thus, the observation that none of the studied compounds inhibited CYP2D6 is a significant finding ruling out metabolic toxicity.
In related toxicity correlate, diabetic patients are at higher risk of developing cardiovascular disease (Aldossari, 2018). The inhibition of human hERG leads to fatal ventricular tachyarrhythmia via a prolonged QT interval (Guth & Rast, 2010). Of the twelve selected compounds, Rutaecarpine and Amentoflavone were found to be at medium risk of hERG inhibition while Luteoloside was found to be at high risk, and the rest were ambiguous (Tables S3 and S4 and Figure S3).
The pKCSM program-based in silico toxicity scores revealed that none of the twelve selected natural compounds showed AMES toxicity (Table S4). Likewise, none of the compounds showed inhibition of hERG-I and Seven compounds (Narirutin, Scutellarin, Complanatuside, Eriocitrin, Amentoflavone, OroxinB and Luteoloside showed hERG-II inhibition potential (Table S4). For hepatotoxicity, Rutaecarpine and Baloxavir were predicted to have toxicity potential. All of the compounds showed a maximum T. pyriformis toxicity range (0.284 to 0.285log mg/L). The lowest Fathead Minnow toxicity was observed for Amentoflavone at À0.801 mM (Table S4).
Finally, the predictions of selected natural compounds human intestinal absorption (HIA) and brain permeation (BBB) determined using the tPSA/WLogP-based graphical BOILED-Egg method suggested only six compounds Baicalin, Rutaecarpine, Amentoflavone, Apigetrin, Baloxavir and Luteoloside were in the acceptable range ( Figure S4).
The in silico potential target prediction of selected natural compounds in the human proteome are summarized in Table S5 and Figure S6. All the six potential CHIKV inhibitors (Baicalin, Rutaecarpine, Amentoflavone, Apigetrin, Baloxavir and Luteoloside) selected in the present study predicted AG protein-coupled receptors as a potential target although with a very low probability of having off-target activity in humans ( Figure S6 and Table S5). Further, electrochemical transporter was predicted second most common target for the selected compounds (Rutaecarpine and Baloxavir).
Baicalin showed a potential target with enzymes (20%) but with a very low probability of 0.15%. The targets of Rutaecarpine were predicted to be the protease, kinase, primary active transporter and voltage-gated ion channel (6.7%), with a probability of below 1%. The targets of Amentoflavone were predicted to be the unclassified proteins (13.3%), kinase (13.3%), enzymes (13.3%), primary active transporters, secreted proteins, ligand-gated ion channels, phosphatases and proteases (6.7%) with a probability of 1%. This indicates that for Amentoflavone, there is the probability of having off-target activity in humans.
In Apigetrin, there was a potential target with the secreted proteins (20%), enzyme (13.3%) and other cytosolic protein (6.7%) with a probability of less than 1%. In Baloxavir, there was a potential target with the kinase (33.3%), ligand-gated ion channels, phosphatases, isomerase, and enzyme (6.7%) with a probability of below 1%. In Luteoloside, showed a potential target of secreted proteins (13.3%) with a probability of 0.81%, enzymes (20%) with a probability of 0.33% and oxidoreductases (13.3%) with the probability of 0.16%. These data indicate that there is a very low probability of having off-target activity of these selected compounds in humans. The in silico potential target prediction of reaming six natural compounds in the human proteome are summarized in supplementary data (Table S5).

MD simulation
MD simulation studies were carried out to infer Inhibitor-CHIKV nsP3 MD stability; fundamental properties like deviation, fluctuation, and compactness of the structure during a simulation provides insights into protein stability in thermodynamics. The RMSD for each system was computed considering 200 ns trajectories. For CHIKV nsP3 MD , the RMSD calculation of the Ca backbone was selected. All of the systems showed minimal deviations and remained stable during the simulation period. The average RMSD for the apo-CHIKV nsP3 MD , CHIKV nsP3 MD -Amentoflavone, and CHIKV nsP3 MD -Baicalin complexes was raised to be 0.29 nm, 0.298 nm and 0.29 nm, respectively. All Inhibitor CHIKV nsP3 MD systems attained equilibrium at 50 ns since the apo-CHIKV nsP3 MD was stabilized after 60 ns. The CHIKV nsP3 MD -Amentoflavone complex showed a peak of 187 ns at 0.41 nm. The RMSD of Inhibitor CHIKV nsP3 MD -Baicalin complex was the same as the apo-nsP3MD, suggesting a stable conformation of CHIKV nsP3 MD upon binding to an inhibitor (Figure 4). The RMSD of each ligand was also calculated to elucidate changes in the binding pattern. The average RMSD of Amentoflavone, and Baicalin, was 0.166 nm and 0.11 nm, respectively. All of the systems attained equilibration after 50 ns of simulation. Further analysis of stability parameters for each complex was carried out for the last 150 ns.
In order to understand the residue-wise fluctuations between the apo-nsP3 MD and Inhibitor-nsP3 MD complexes, the RMSF values were plotted. The average RMSF values for apo-nsP3MD, nsP3 MD -Amentoflavone, and nsP3 MD -Baicalin complex were found to be 0.12, 0.122 nm and 0.124 nm, respectively. In nsP3 MD -Amentoflavone complex, the loop region Leu 28 to Gly 30 showed significant fluctuations of .48 nm and .40 nm, respectively. No significant fluctuations were observed with Inhibitor nsP3 MD -Baicalin complex (Figure 4). The RMSF of the nsP3 MD -Baicalin complex was the least among the Amentoflavone complex, suggesting a stable binding.
Intermolecular polar interaction between protein and ligand plays a crucial role in the binding affinity. The hydrogen bond between nsP3 MD and each of the compounds was calculated. The maximum number of hydrogen bonds in nsP3 MD -Amentoflavone and nsP3 MD -Baicalin complexes were 5 and 6, respectively ( Figure 5). Baicalin formed the highest number of hydrogen bonds. Also, protein compactness is a significant factor in determining the folded state. The Rg was calculated to measure the compactness of nsP3 MD in all systems. It was observed that all systems had comparable Rg values with no significant changes in the simulation. The average Rg for the apo-nsP3 MD , nsP3 MD -Amentoflavone, and nsP3 MD -Baicalin complex was 1.43 nm, 1.53 nm, 1.44 nm, respectively ( Figure 5). Compactness analysis suggested that identified compounds formed attractive and stable interactions with nsP3 MD .
In the presence of the identified compounds, the binding pocket of nsP3 MD remains stable, suggesting that the identified drugs molecules have a high affinity towards nsP3 MD and can be used as potential binders nsP3 MD . The SASA was then calculated to decipher protein stability and folding during the simulation ( Figure 5). The SASA plot showed lower SASA values for nsP3 MD -Amentoflavone, and nsP3 MD -Baicalin, while for apo-nsP3 MD , it showed higher SASA values. The average value for the apo-nsP3 MD , nsP3 MD -Amentoflavone, and nsP3 MD -Baicalin, the complex was found to be 94 nm2, 91.5 nm2 and 93 nm2, respectively. These results concluded that Amentoflavone, and Baicalin, showed good binding to nsP3 MD .

Conclusion
Recently, plant-based natural compounds have received attention as novel antiviral therapeutics (Hussain et al., 2020;Khan et al., 2021;Lani et al., 2016). Thus, identifying natural ingredients to advance antiviral treatments holds good possibilities for new therapies. Molecular docking allows in silico screening of compounds before testing experimentally and has gained popularity to save time and resources in the drug discovery and development process (Agarwal et al., 2019).
The docking study demonstrated good interactions between natural compound Baicalin and CHIKV envelope protein with the binding affinity of À9.7 kcal/mol (Oo et al., 2018). CHIKV nsP3 MD specific Baicalin, a dominant flavonoid used for adjuvant therapy of hepatitis in traditional Chinese medicine (Tao et al., 2018).This compound has various pharmacological activities, including anti-oxidative, antiviral, anti-inflammatory, anti-HIV and anti-proliferative properties (Huang et al., 2000;Huang et al., 2006;Lee et al., 2008;Li et al., 1993;Oo et al., 2018;Seyedi et al., 2016;Tao et al., 2018;Wang et al., 2020). Baicalin was tolerated by Vero, BHK-21 and HEK 293 T cells with maximal nontoxic doses >600 lM, % 350 lM and %110 lM, respectively, indicating very low cytotoxicity of this compound (Oo et al., 2018). Anti CHIKV assays indicated that baicalin was the most effective inhibitor when tested for its direct virucidal activity with EC50 % 7 lM, followed by inhibition of virus entry into the host cell, attachment of virus particle to cellular receptors and finally intracellular replication of CHIKV RNA genome. Quantitative RT-PCR analysis showed that baicalin had very high effect on  the synthesis of viral À ve stand RNA with EC50 % 0.4 lM, followed by the inhibition of synthesis of þ ve strand genomic (EC50 % 13 lM) and subgenomic RNAs (EC50 % 14 lM) (Oo et al., 2018;Seyedi et al., 2016;Tao et al., 2018;Wang et al., 2020). Subsequent studies further confirmed that baicalin inhibits the early stages of CHIKV replication and has strong virucidal activity (Kovacikova & van Hemert, 2020).
Rutaecarpine is a bioactive alkaloid and has been used to treat various cerebrovascular, metabolic and cardiovascular diseases (Tian et al., 2019). Amentoflavone a, biflavonoid found in several natural plants, has been shown to exhibit anti-oxidation, antitumor, anti-inflammatory, neuroprotective and cardiovascular protective properties (Chiang et al., 2019;Funakoshi-Tago et al., 2015;Gan et al., 2020;Saroni Arwa et al., 2015;Zhang et al., 2015;Zheng et al., 2013). Apigetrin, a flavonoid, has been shown to exhibit antimutagenic, anticancer, antioxidant and anti-inflammatory properties (Hadrich & Sayadi, 2018). Luteoloside is a Flavonoid found in some dried fruits and various other plant sources such as Lonicera japonica (Shi et al., 2020;Zhao et al., 2019).Luteoloside exhibits several bioactivities, including anti-microbial and anti-cancer activities, and was also shown to act as a 3 C protease inhibitor of EVA71 in vitro (Cao et al., 2016). Baloxavir is a polymerase acidic protein endonuclease inhibitor (Influenza A and influenza B cap-dependent endonuclease enzyme inhibitor). Recently a study conducted a multicenter, doubleblind, randomized, placebo-controlled trial to evaluate the postexposure prophylactic efficacy of baloxavir in Japan (Ikematsu et al., 2020).
However, few studies about safety in pregnant women and people with underlying medical conditions are present. Drug-likeness factor rules were obeyed accordingly with no violation by these plant-based natural compounds. Thus, these natural compounds can act as a drug in biological systems. The toxicity prediction says that they are safe and can be given as drugs with the value of tolerance prescribed for human consumption. All of the twelve natural compounds six compounds Baicalin, Rutaecarpine, Amentoflavone, Apigetrin, Baloxavir and Luteoloside showed favorable ADMET properties. MD simulation studies showed stable conformation dynamics upon both the compounds binding to CHIKV nsP3 MD . Hence, out of twelve, these six compounds can be considered for their exceptional binding energy and ADMET properties. However, further investigation and validation of these inhibitors against CHIKV are needed to claim their candidacy for clinical trials.