Withanone from Withania somnifera May Inhibit Novel Coronavirus (COVID-19) Entry by Disrupting Interactions between Viral S-Protein Receptor Binding Domain and Host ACE2 Receptor CURRENT STATUS: POSTED

Background Newly emerged COVID-19 has been shown to engage the host cell ACE2 through its spike protein receptor binding domain (RBD). Here we show that natural phytochemical from a medicinal herb, Withania somnifera, have distinct effects on viral RBD and host ACE2 receptor complex. Methods We employed molecular docking to screen thousands of phytochemicals against the ACE2-RBD complex, performed molecular dynamics (MD) simulation, and estimated the electrostatic component of binding free energy, along with the computation of salt bridge electrostatics. Results We report that W. somnifera compound, Withanone, docked very well in the binding interface of AEC2-RBD complex, and was found to move slightly towards the interface centre on simulation. Withanone significantly decreased electrostatic component of binding free energies of ACE2-RBD complex. Two salt bridges were also identified at the interface; incorporation of Withanone destabilized these salt bridges and decreased their occupancies. We postulate, such an interruption of electrostatic interactions between the RBD and ACE2 would block or weaken COVID-19 entry and its subsequent infectivity. Our data, for the first time, show that natural phytochemicals could well be the viable options for controlling COVID-19 entry into host cells, and W. somnifera may be the first choice of herbs in these directions to curb the COVID-19 infectivity.


Introduction
In December 2019, a mysterious virus causing pneumonia was first reported in China [1], and it is now growing globally into a deadly disease. In the first week of January 2020, the Chinese Centre for Disease Control and Prevention (CCDCP) identified a novel coronavirus strain that has not been earlier identified in humans [2][3][4]. Coronaviruses are characterized as zoonotic which can transmit between animals and people [5]. These are large positive-strand RNA viruses [6], which can cause infection in a variety of avian as well as mammalian species. These can cause diseases related to the central nervous system, upper and lower respiratory and gastrointestinal tracts [7,8]. Till now, seven coronaviruses have been identified that can cause human diseases, four of these are mild viruses: OC43, 229E, HKU1 and NL63 [9]. Whereas, other three viruses can have more serious consequences in human, and these are SARS-CoV (causes Severe Acute Respiratory Syndrome) which appeared in November 2002 [10,11], in China. Another outbreak of MERS-CoV (causes Middle East Respiratory Syndrome) emerged in 2012 in Saudi Arabia [12]. Now, Coronavirus COVID-19 has gradually crossed the wall of China, and has spread throughout the world. Stringent efforts are being placed globally to contain COVID-19 spread. However, epidemiological information and clinical features of illness caused by COVID-19 is still at infancy at present [13,14].
In a viral infection, the viral entry in the host cell is a critical step that can be exploited for anti-viral therapy [15]. Coronaviruses get access into the target animal cells via binding to cell-surfaceassociated receptors, and its entry can be barred by targeting the viral receptor-binding site with neutralizing antibodies (nAbs). There are also certain small molecules (like RFI-641 and VP-14637) which inhibit the entry of several viruses including respiratory syncytial virus [16,17]. In the case of Coronavirus, the viral entry is mediated by the Receptor-Binding Domain (RBD) of its spike (S) glycoprotein, which binds to the host cell receptor Angiotensin-Converting Enzyme-2 (ACE2) [18,19].
The coronavirus S-protein, a structural protein responsible for the crown-like shape of the viral particles, is ~ 1200 aa long S-protein belonging to class-I viral fusion proteins, and contributes to the cell receptor binding, tissue tropism and pathogenesis [20]. It contains several conserved domains and motifs, and the trimetric S-protein is processed at the S1/S2 cleavage site by host cell proteases.
The protein is divided (cleavage, or priming) at a conserved sequence AYT↓M (located 10 aa downstream of SLLR-ST) into an N-terminal S1-ectodomain that recognizes a cognate cell surface receptor and a C-terminal S2-membrane-anchored protein involved in viral entry [20][21][22]. The SARS-CoV S1-protein contains a conserved RBD, which recognizes the host ACE2. The RBD surface of S1/ACE2 implicates 14 aa in the S1 of SARS-CoV [23], among them, 8 residues are strictly conserved in COVID-19, supporting the hypothesis that ACE2 is also the receptor of this newly emerged coronavirus [24]. The RBD of COVID-19 differs largely from the SARS-CoV at the C-terminus, and it has been reported that such difference did not result in drastic changes in its capability to engage ACE2 receptor [25]. Therefore, RBD has been an attractive target for the researchers to abrogate coronavirus infection. Reports suggested that certain human antibodies recognized RBD on the S1 domain and inhibited the viral infection by blocking the attachment of ACE2 [26,27]. Three possible mechanisms have been proposed where COVID-19 infection can be abrogated by blocking the interaction of spike protein and ACE2 (Fig. 1), along with the strategy followed in the present work.
Withania somnifera (L.) Dunal (Solanaceae), commonly known as Ashwagandha, one of the most valued medicinal plants of the traditional Indian systems of medicines, is used in more than 100 formulations of Ayurveda, and is thought to be therapeutically equivalent to Ginseng [28]. W. somnifera has been used as antiviral herb for the treatment of genital disease caused by Herpes Simplex Virus among African tribes [29], and shown to anti-influenza properties [30]. The aim of present study is to check the antiviral potential of W. somnifera ingredients against COVID-19 by means of computational methods. The core rationale of this study is to inhibit or weaken the interactions between the receptor and RBD by using phytocompounds that could block or hamper viral entry into the host cells.

Materials And Methods Structures
We screened thousands of phytocompounds by in-silico method. The strategy was to find phytocompounds that bind at the protein complex interface, and perturb their interactions. In this regard, we observed that the phytocompounds from W. somnifera could fit into our strategy. For W. somnifera root extracts, Chaurasiya et al. [31] have utilized a reversed-phase HPLC method for the simultaneous analysis of nine structurally similar withanolides: 27-hydroxy withanone, 17-hydroxy withaferin A, 17-hydroxy-27-deoxy withaferin A, withaferin A, withanolide D, 27-hydroxy withanolide B, withanolide A, withanone and 27-deoxywithaferin A. Our lab data (unpublished results) also showed significant concentration of Withanolide A, Withanolide B, Withaferin A and Withanone in W. somnifera sapling (15-20 days old). Therefore, we studied these compounds in detail after preliminary screening. The 3D structures of all the phytocompounds were sourced from PubChem database (https://pubchem.ncbi.nlm.nih.gov).

Molecular docking
The structure of the ACE2 complexed with spike protein receptor binding domain (RBD) of 2019-nCoV is not yet available. Therefore, RBD sequence of 2019-nCoV spike protein was obtained from NCBI (NCBI accession: QHD43416), and the mutations were considered [32]. The protein complex model was built using the MODELLER program as implemented in Structuropedia (http://structuropedia.org) [33] and by using SWISS-Model (http://swissmodel.expasy.org) based on a clean 3D0G (2.8 Å resolution; Protein Data Bank, http://www.rcsb.org) template. The modelled 2019-nCoV RBD complexed with ACE2 A-chain was used for further processing after editing on PyMol [34]. Energy minimization was performed by 100 steps of steepest descent, followed by 500 steps of conjugate gradient using UCSF Chimera-1.13.1 [35], and the stereo-chemical quality of the energy minimized model was checked using VERIFY 3D [36], ERRAT [37], PROCHECK [38] and RAMPAGE (for Ramachandran plot) [39].
The SDF files of all the ligands downloaded from the PubChem database were then converted into PDB files using OpenBabel 2.4.1 [40]. The binding site in the ACE2-RBD complex was determined by blind docking using AutoDock Vina-1.1.2 (ADV-1.1.2) [41], and then the focused/targeted docking was performed based on the binding location of the ligands derived post blind docking. For intermediary steps, such as PDBQT files for protein and ligands preparation and grid box creation were performed using Graphical User Interface program AutoDock Tools-1.5.6 (ADT-1.5.6) [42]. ADT was used to assign polar hydrogens and Gasteiger charges. 'Choose ligand' option was used to set map file types.
AutoDock Tool was used to save the prepared files in PDBQT format. For blind docking, Grid maps were prepared using a grid box size of 60 × 60 × 60 xyz points and the protein centre (x = 50.541, y = -1.366, z = 105.966). To obtain the maximum number of poses, we set num_modes to 20, energy range to 9, and exhaustiveness to 8. The pose with lowest energy of binding was extracted and aligned with the receptor for further analysis by Discovery Studio 2017 R2 Client [43] and PyMol [34].

Molecular dynamics (MD) simulation
The Ligand-ACE2-RBD complex was obtained after the targeted molecular docking. The simulation systems for ACE2-RBD complex without or with the Withanone were prepared using the VMD software (Humphrey et al., 1996). Ligand parameterization was done with CHARMM-GUI web interface (http://www.charmm-gui.org) [44]. MD simulation was performed with CHARMM36 force field using the NAMD package [45]. The protein complex without or with Withanone was solvated with TIP3P water molecules 10 Å from the protein. The systems were ionized and neutralized with 100 mM of NaCl. The systems contained 67890 and 67851water molecules in the protein complex without and with the Withanone, respectively. NPT ensemble was used with periodic boundary conditions. Pressure was fixed at 1 atm, while the temperature was 310 K. The particle-mesh Ewald method was used to evaluate the Coulomb interactions. 2 fs of time step was used in all MD simulations. Initially, water was equilibrated for 200 ps at 310 K after fixing the protein and energy minimization of 1000 steps. 1000 steps of energy minimization of the whole system were performed, and further equilibration for 400 ps at 310 K after releasing the protein was done. Production run was of 2000 ps. The trajectory data were saved at every 0.5 ps to analyze the change in the dynamics of ACE2-RBD binding interface. The results for flexibility were analyzed by plotting the non-H atoms RMSD values against the 1000 conformations (stride 4). Trajectory clustering was performed by UCSF Chimera-1.13.1 [35], using the step size of 1 and default parameters.

Salt bridge analysis
Salt bridge was defined at a cut off distance of 3.2 Å of O-N atoms of the oppositely charged amino acid side chains using VMD [46]. Electrostatic free energies up on salt bridge formation were computed in initial and final trajectories. All the computations were performed according to the protocol of Hendsch and Tidor [47], with slight modifications. Briefly, it was calculated relative to a mutation of its salt-bridging side-chains to their hydrophobic isosteres; they are identical with the charged residue side-chains, with the exception that their partial atomic charges were set to zero. The protonation states of all the charged residues were assigned at pH 7.4 using ProteinPrepare module in PlayMolecule (https://www.playmolecule.org). Continuum electrostatic calculations were performed with the DelPhi v8.4.3 [48]. The PARSE partial atomic charges and atomic radii [49] were used. The

Molecular docking
The model of protein complex was built by SWISS-MODEL and the energy was minimized. The stereochemical quality was checked and confirmed (refer to Supporting Information1 for the details).
Withanolides present in roots and leaves of W. Somnifera were docked against ACE2-RBD complex.
The phytocompound which bound to the interface was subjected to targeted/focused docking. The phytocompounds were bound at the ACE2-RBD complex tightly (see Table 1 for Vina score). Of these compounds docked, only the Withanone bound at the interface of the receptor and RBD (Fig. 2).
Therefore, Withanone was analyzed further to study its role in blocking or weakening the interactions between the ACE2 receptor and RBD. On targeted docking of the Withanone, it was found to be wellbound at the ACE2-RBD interface by two H-bonds (Tyr16 of ACE2 and Tyr175 of RBD to Withanone), alkyl and van der waals interactions (Fig. 2). These Tyrosines were mutated by alanine, and re-docked with Withanone.

Molecular dynamics (MD) simulation Interaction analysis
The RMSD of the simulated molecule (Withanone) was 5.08 Å compared to starting position. At the end of simulation, it moved slightly towards the binding interface centre (Fig. 3A). On analysing, the ligand interaction, it was found that ACE2 Y16 H-bonding to Withanone was preserved in the simulated coordinates, whereas RBD Y175 forms Carbon H-bond to Withanone. Additionally, there is formation of three more H-bonds (ACE2 N15, ACE2 Q19 and RBD R78 to Withanone) in the simulated ligand-ACE2-RBD complex (Fig. 3B).

Salt bridge analysis
We detected two inter-chain (binding interface) salt bridge interactions, Glu12 OE2-Lys87 NZ (2.75 Å) (aa 404) and Glu20 OE2 -Arg73 NZ (2.67 Å) (aa 390). Val residue (aa 404) in SARS-CoV is substituted by Lys in COVID-19 S protein RBD, and Lys (aa390) is substituted by Arg (Fig. 4). Protein surfaces have many hydrophilic residues, and salt bridges present in the surface play an important role in protein-protein association or binding [52]. Hence the protein interface (binding interface) is generally more hydrophilic than the protein interiors. Xu       Electrostatic contribution upon salt bridge formation of the two salt bridges (E12-K87 and E20-R73) calculated in the initial and the final frames of simulations. E12-K87 salt bridge was energetically favorable in the first frame; gradually it got destabilized as seen in final frame. Similarly, E20-R73 salt bridge was more favorable in the initial frame, and then its extent decreased as seen in the final frame.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.