Inhibition of nonstructural protein 15 of SARS‐CoV‐2 by golden spice: A computational insight

Abstract

The quick widespread of the coronavirus and speedy upsurge in the tally of cases demand the fast development of effective drugs. The uridine‐directed endoribonuclease activity of nonstructural protein 15 (Nsp15) of the coronavirus is responsible for the invasion of the host immune system. Therefore, developing potential inhibitors against Nsp15 is a promising strategy. In this concern, the in silico approach can play a significant role, as it is fast and cost‐effective in comparison to the trial and error approaches of experimental investigations. In this study, six turmeric derivatives (curcuminoids) were chosen for in silico analysis. The molecular interactions, pharmacokinetics, and drug‐likeness of all the curcuminoids were measured. Further, the stability of Nsp15‐curcuminoids complexes was appraised by employing molecular dynamics (MD) simulations and MM‐PBSA approaches. All the molecules were affirmed to have strong interactions and pharmacokinetic profile. The MD simulations data stated that the Nsp15‐curcuminoids complexes were stable during simulations. All the curcuminoids showed stable and high binding affinity, and these curcuminoids could be admitted as potential modulators for Nsp15 inhibition.

Significance statement

We implemented advanced computational approaches to highlight the inhibitory mechanism of bioactive molecules of turmeric against nonstructural protein 15 (Nsp15) of SARS‐CoV‐2. In this study, we reported that the turmeric compounds could interact effectively with Nsp15 of SARS‐CoV‐2 and inhibit its further progression. The reported molecules from turmeric could be developed as potential inhibitors against the Nsp15 of SARS‐CoV‐2.

1. INTRODUCTION

The coronavirus (CoV) epidemic arose at the end of 2019. WHO announced an international emergency with the disclosure of the primary transmission. After a month, the deaths raised significantly in 53 countries. The situation is mostly due to the absence of a specific drug against the CoV. Confined within a shielding cover and a helical capsid, the CoV genome is ~26–32 kb in length, encoding four proteins: envelope, spike, nucleocapsid, and the membrane protein. Although some nonstructural polyproteins (Nsps) have been classified to hold a part in RNA transcription and replication, the functions of some are unknown. ¹ The nonstructural protein (Nsp15), also termed as uridylate‐specific endoribonuclease (NendoU), favorably cuts the 3′ end of uridines. ² The enzymes of this family are specific for uridine. They are responsible for playing a role in the functions related to RNA processing and RNA endonuclease activities that develop cyclic 5′‐hydroxyl termini and 2′‐3′ phosphodiester via transesterification. ³ Nsp15 from SAR‐CoV‐2 shares similarities with both MERS‐CoV and SARS‐CoV, comprising dimer of trimers ultimately forming a double ring hexamer critical for the enzymatic activity. ⁴ Each monomer is formed of ~345 amino acids divided into a C‐terminal NendoU catalytic domain, a middle, and an N‐terminal domain. NendoU further comprises two β‐sheets (antiparallel), forming a shallow groove between them that houses the main active site, whereas the concave part in between the sheets comprises five α‐helices. ⁵ Earlier, Nsp15 was deemed to have a direct involvement in virus replication, however, its interventions with the innate immune response is disclosed in recent studies, ⁶ , ⁷ therefore declaring its biological relevance. ⁸ , ⁹

To treat COVID‐19, doctors prescribe antiretroviral, anti‐malarial, anti‐influenza pills and their determinate‐dose combination, but these medicines have no durable remedies for SARS‐CoV‐2 infection. ¹⁰ Several bioactive molecules in various studies were shown to hold probable antiviral action against SARS‐CoV‐2 and could be considered as an option to limit the replication cycle of coronavirus. ¹¹ , ¹² , ¹³ Natural molecules have less side effects, less production prices, and greater chemical variety than synthetically manufactured drugs. Therefore, using computational analysis, we evaluated the ability of turmeric‐derived (curcuminoids) compounds targeting Nsp15 (SARS‐CoV‐2). Turmeric (golden spice) has been generally used as a spice and healing agent, mainly in Ayurveda. It has already been published in previous literature that turmeric‐derived compounds inhibit several viral infections by intervening with their replication cycles. ¹⁴ In another piece of literature, it was confirmed that the curcumin obstructed Zika and Chikungunya infections by preventing virus binding to the cell membrane. ¹⁵ In some recent studies, it was reported that the bioactive phytoconstituents were able to inhibit Nsp15 of SARS‐CoV‐2. ¹⁶ , ¹⁷ We selected a library of six curcuminoids to dock them into the binding site of Nsp15. All the curcuminoids were further analyzed via pharmacokinetic evaluation and structural parameters of molecular dynamics (MD) simulations, hence, furnishing a foundation for further experimental analysis.

2. COMPUTATIONAL METHODOLOGY

2.1. Datasets

The 1.90 Å resolution PDB structure of Nsp15 was acquired from the RCSB database (ID: 6W01). ¹ The natural form of six turmeric rhizome curcuminoids (curcumin derivatives) were acquired from the PubChem database. ¹⁸ , ¹⁹ Discovery Studio's module of “prepare protein” was practiced for the protein preparation. ²⁰ The B3LYP hybrid DFT (minimization protocols) with 6‐31 + G(d, p) basis set of Gaussian16 was used for geometry optimization of every ligand. ²¹

2.2. Molecular docking

Molecular docking was conducted by the CDOCKER module of Discovery Studio v2018. CDOCKER uses a CHARMm force field for semi‐flexible docking, providing rigorous docking results. ²² It provides ligand‐receptor interaction mode and interaction energy. The ligand binding site defined according to co‐crystallized citrate molecule, generating a 12 Å radius range. The XYZ coordinates values were ‐64.966163, 72.265539, and 29.143208. Ten best conformations were obtained after docking for each compound. Finally, the interaction energies obtained through CDOCKER were utilized to estimate the binding affinities between the target and the selected molecules. We retained all the remaining CDOCKER parameters as default for docking executing studies.

2.3. Physicochemical profile

The drug‐like and pharmacokinetics properties of curcuminoids were calculated by adopting the SwissADME web server. ²³ The vNN‐ADMET online platform accomplished the predictions of the toxicological potential of curcuminoids. ²⁴

2.4. Molecular dynamics simulations

The stability and flexibility of the curcuminoids‐Nsp15 complexes were evaluated by executing 100 ns of MD simulations. The whole simulations were conducted by employing the Gromacs 4.6 suite operating GROMOS96 43a1 (force field). ²⁵ , ²⁶ , ²⁷ PRODRG server 2.0, used to produce the topology files of ligands. ²⁸ A sufficient number of ions were added to the refined protein systems, and a cubic box was created to solvate with SPC water particles for system neutrality. The energy minimization was done to eliminate any steric clashes within atoms applying the steepest descent method with a confluence model of fewer than 1000 kJ/mol/nm. PME method was used to measure large‐scale electrostatic interactions and a 9 Å cut‐off radius for Van der Waal and coulombic interactions, equilibrated for two‐states. ²⁹ In an initial state, ions and solvents were cached free for 1000 ps in the NVT ensemble. In the next state, the control load from complexes was continually reduced for 1000 ps in the NPT ensemble. Each hydrogen bonds, held obligated to accept the LINCS algorithm. ³⁰ , ³¹ The systems temperature and pressure were maintained at 300 K and 1 atm by Parrinello‐Rahman pressure coupling and Berendsens temperature. ³² , ³³ The binding free energies of the complexes were determined using MM‐PBSA method. The polar portion of solvation energy, estimated via determining the Poisson‐Boltzmann equation. In contrast, non‐polar solvation, calculated by linear connection to the solvent accessible surface area. The g_mmpbsa script, executed to evaluate the complexes binding affinity. ³⁴ The free energy landscape analysis (FEL) study was carried out to achieve the low‐energy state of selected complexes. We used the g_sham module to calculate FEL.

3. RESULTS

3.1. Structure‐based molecular docking

The six curcuminoids were efficiently screened against Nsp15 target protein. The docked molecules interaction evaluation unveiled that they all furnish strong bonds with the binding pocket of Nsp15. Bis‐demethoxycurcumin formed hydrogen bonds with Thr341, Gly248, Lys290, His250, Glu340, and His235, while hydrophobic interactions with Thr341, Trp333, Tyr343, and Val292. Curcumin formed hydrogen bonds with Thr341, His235, Lys290, His250, and Gly248 while residues Val292 and Tyr343 interacted via hydrophobic interactions. Dimethylcurcumin formed only hydrogen bonds with Gln245, Gly248, Lys290, and His250. Tetrahydrocurcumin formed hydrogen bonds with Lys290, His235, Thr341, His250, and Glu340, while residue Trp333 interacted via Pi‐stacking interaction. Residues Thr341 and Tyr343 formed hydrophobic interactions with tetrahydrocurcumin. Demethoxycurcumin formed hydrogen bonds with Lys290, His235, Glu340, His250, and Gly248, whereas hydrophobic interactions with residues Trp333, Thr341, and Tyr343. Diacetylcurcumin formed hydrogen bonds with Gln245, His250, Lys290, Gly248, Asn278, and Leu346. Aromatic and carboxylate group of diacetylcurcumin formed one Pi‐cation and one salt bridge with residue His235. Hydrophobic interactions were observed with residues Glu340, Thr341, Tyr343, and Lys345 (Figure 1).

Figure 1.

3D interactions of nonstructural protein 15 (Nsp15) of SARS‐CoV‐2 with the selected docked curcuminoids (A) demethoxycurcumin, (B) dimethylcurcumin, (C) tetrahydrocurcumin, (D) diacetylcurcumin, (E) curcumin, (F) Bis‐demethoxycurcumin

3.2. Pharmacokinetic/ADMET predictions

The ADMET and drug‐likeness properties of selected molecules were shown in Tables 1 and 2. All the selected molecules were highly absorbable in the gastrointestinal tract. The dimethylcurcumin and Bis‐demethoxycurcumin were BBB permeant, while tetrahydrocurcumin, demethoxycurcumin, curcumin, and diacetylcurcumin were impermeant. Moreover, the bioavailability radar plot confers a functional assessment of a molecules drug‐likeness by reflecting six Physicochemical features: weight, saturation, polarity, lipophilicity, flexibility, and solubility. All the curcuminoids were inside the permissible range of physiochemical properties (Supporting Information: Figure S1). The curcuminoids were predicted to be orally bioavailable, less toxic, and highly absorbable. The physiochemical predictions disclosed that these curcuminoids possess required qualities to act as suitable lead candidates for drug development.

Table 1.

Physicochemical and drug‐likeness properties of selected molecules

Molecules	MW (g/mol)	LogP	HBD	HBA	MR	TPSA (Ų)	PAINS alert	Lipinski/Ghosh violation	Bioavailability score
Tetrahydrocurcumin	372.41	1.62	2	6	102.17	93.06	0	0	0.55
Demethoxycurcumin	338.35	1.80	2	5	96.31	83.83	0	0	0.55
Dimethylcurcumin	396.43	1.91	1	6	112.64	74.22	0	0	0.85
Curcumin	368.38	1.47	2	6	102.80	93.06	0	0	0.55
Bis‐demethoxycurcumin	308.33	2.13	2	4	89.82	74.60	0	0	0.55
Diacetylcurcumin	452.45	2.22	0	8	121.75	105.20	0	0	0.55

Note: MW, molecular weight = 50–500; LogP, octanol/water partition coefficient = _2‐10; TPSA, topological polar surface area = 20–130; HBA, number of H‐bond acceptors = 0–10; HBD, number of H‐bond donors = 0–5; MR, molar refractivity = 40–130.

Table 2.

Pharmacokinetics and water solubility of selected molecules

Molecules	Log S (Esol)	BBB	GI absorption	CYPD26	AMES	hERG Blocker
Tetrahydrocurcumin	S	No	High	No	No	No
Demethoxycurcumin	S	No	High	No	No	No
Dimethylcurcumin	MS	Yes	High	No	No	No
Curcumin	S	No	High	No	No	No
Bis‐demethoxycurcumin	S	Yes	High	No	No	No
Diacetylcurcumin	MS	No	High	No	No	No

Abbreviations: AMES, mutagenicity; BBB, blood‐brain barrier; CYP, cytochrome P450; GI, gastrointestinal; hERG, cardiotoxicity; MS, moderately soluble; S, soluble.

3.3. Molecular dynamics simulation analysis

Post‐docking, all the six curcuminoids were passed on to 100 ns MD simulations. The MD simulations were undertaken to access the stability and binding interactions of the curcuminoids with Nsp15. The stability of the Nsp15‐curcuminoids complexes was evaluated via root mean square deviation (RMSD) and root mean square fluctuation (RMSF) interpretations. The RMSD graph of backbone c‐alpha atoms implied that all the complexes gained stability after 40 ns and manifested fewer fluctuations with low RMSD values within the range of 0.2‐0.62 nm. This indicated that the curcuminoids‐Nsp‐15 complexes have a stable conformation (Figure 2A). We estimated the RMSF to measure the fluctuations of the active site residues of Nsp15 protein upon binding with curcuminoids. The average RMSF estimation of all the complexes was witnessed within the range of 0–0.55 nm. The RMSF of the residues of the binding pocket was significantly low, which shows the stability of the protein‐ligand complex (Figure 2B). Additionally, we scrutinized the hydrogen bond numbers to acquire insight into the stability and the interactions between protein‐ligand systems. An average of 4–6 hydrogen bonds were observed between Nsp15 and curcuminoid molecules (Figure 3).

Figure 2.

Graphical representation of (A) root mean square deviation (RMSD). (B) Root mean square fluctuation (RMSF) of the backbone C‐α‐atoms for nonstructural protein 15 (Nsp15) of SARS‐CoV2 in complex with curcuminoids: demethoxycurcumin (cyan), dimethylcurcumin (green), tetrahydrocurcumin (brown), diacetylcurcumin (blue), curcumin (red), and Bis‐demethoxycurcumin (yellow).

Figure 3.

Hydrogen bond profiles of nonstructural protein 15 (Nsp15) in complex with curcuminoids: demethoxycurcumin (cyan), dimethylcurcumin (green), tetrahydrocurcumin (brown), diacetylcurcumin (blue), curcumin (red), and Bis‐demethoxycurcumin (yellow)

3.4. Binding free energy and FEL

MM‐PBSA based binding energies (kJ/mol) for all the Nsp15‐curcuminoid complexes were measured. The diacetylcurcumin, Bis‐demethoxycurcumin, dimethylcurcumin, tetrahydrocurcumin, demethoxycurcumin, and curcumin showed −150.757, −113.174, −121.991, −83.513, −91.246, and −95.233 kJ/mol binding free energy. These results intimated that the selected curcuminoids had the most eminent affinity against Nsp15 of SARS‐CoV‐2 (Supporting Information: Figure S2). Further, we conducted the FEL analysis for all the Nsp15‐curcuminoid complexes to measure the low‐energy (minima) acquired during the simulations. The FEL investigation pointed out that all the Nsp15‐curcuminoid complexes achieved minimum energy with (2.25 ± 0.05 nm and 0.17 ± 0.05 nm) low Rg and RMSD scores during the simulations. It was clear that an RMSD value below 2.0 Å corresponds to less deviation in the protein backbone, which leads to stability of the protein structure. Each Nsp15‐curcuminoid complex had a similar pattern of the FEL. The graphical representation of FEL analysis demonstrated that all the complexes exhibited a single energy basin with a deep valley (Figure 4). The FEL interpretation reported that all the complexes gained minimum energy that corresponds to the most stable conformational states.

Figure 4.

The free energy landscape plot of six nonstructural protein 15 (Nsp15) complexes (A) Bis‐demethoxycurcumin, (B) curcumin, (C) demethoxycurcumin, (D) tetrahydrocurcumin, (E) diacetylcurcumin, and (F) dimethylcurcumin. The yellow color region depicts the minimum energy conformation

4. DISCUSSION

In March 2020, a crystal structure of Nsp15 SARS‐CoV‐2 was reported by Kim and colleagues. ¹ This protein is specific for uridine and debases the viral dsRNA, thereby restricting host identification. ³⁵ The enzymatic degeneration of RNA occurs via a transesterification reaction by constructing a (20–30) cyclic phosphodiester edge. Earlier studies of Nsp15 have classified residues Thr341, His250, Tyr343, Lys290, Ser294, and His235 as the binding pocket residues engaged in protein‐ligand interactions. ¹² , ³⁶ The site of co‐crystallized citrate was adopted as an active site for docking studies. The docking result revealed that the selected curcuminoids had significant CDOCKER interaction energies (Supporting Information:  Figure S3).

Furthermore, we have also determined the drug‐likeness and pharmacokinetics of the selected curcuminoids. SwissADME and vNN‐ADMET are free and validated web servers for estimating the drug‐likeness and pharmacokinetics of the molecules. These servers build on Lipinski's rule and other pharmacokinetics standards with a convenient interface. Tables 1 and 2 represent the drug‐likeness, Physicochemical, and pharmacokinetic properties of the six curcuminoids. The ADMET data disclosed physicochemical attributes of the selected molecules, which covers Lipinski's rule (MW, HBA, HBD, LogP, and MR), other features such as TPSA, and the alerts for unwanted properties (PAINS alert), amongst others as depicted in Table 1. According to Ghosh and Lipinski's rule, all the molecules followed the regulations by showing no violation. Also, there were no PAINS alerts for any molecule. Concerning this, all the examined curcuminoids conferred a 0.55 bioavailability score, except dimethylcurcumin (0.85). All the curcuminoids were predicted to have high gastrointestinal absorption (GIA), which proved that the molecules had high bioavailability. The critical property used in predicting compounds oral bioavailability was polarity by TPSA. Molecules between 20 and 130 Ų TPSA have high oral bioavailability. As displayed in Table 2, the six curcuminoids did not show inhibition of CYPD26, a part of the drug‐metabolizing cytochrome P450 group of enzymes. The function of the cytochrome P450 group of enzymes in drug metabolism is well defined. Accordingly, the interaction of drug/active molecules with any of CYP isoenzymes points to either bioaccumulation or fast metabolism of drugs in the body. In silico strategies are beneficial in predicting these interplays as an element of the drug development method. The prediction of enduring GIA was high for all six curcuminoids. The GIA result was based on the Brain and IntestinaL EstimateD permeation (BOILED‐Egg) standard, which utilizes polarity and lipophilicity of compounds to define the blood‐brain barrier permeation and enduring gastrointestinal absorption. Molecules found in the yellow area (yolk) had a large penetration probability of BBB. In contrast, molecules in the white area reflect the capacity for resistless absorption into the GI tract (Figure 5). Only dimethylcurcumin and Bis‐demethoxycurcumin exhibited permeability for biological barriers. The BBB shields the central nervous system by limiting some harmful substances from invading the brain tissues. The BBB restricts most of the external molecules entrance to manage the central nervous system at a steady‐state. Besides, adverse side effects and weak pharmacokinetics due to molecule toxicity are the leading motives for drug failure in the later stage. In this regard, our outcomes disclosed that all the curcuminoids were nonmutagenic and noncardiotoxic for AMES and the hERG blocker predictions. Tetrahydrocurcumin, demethoxycurcumin, curcumin, and Bis‐demethoxycurcumin molecules were soluble, while dimethylcurcumin and diacetylcurcumin were moderately soluble (Table 2).

Figure 5.

Boiled‐Egg plot generated by swissADME web server to predict gastrointestinal absorption and brain penetration of six curcuminoids

For more validation, MD simulations were carried out using GROMACS package. Proteins are intricate biological macromolecules delivering a wide variety of roles in living systems. Computational and mathematical modeling strategies can also be used besides experimental procedures to get insight into the atomic arrangements and conformational space of protein complexes. Through MD simulations, we can gather details regarding protein complexes at the microscopic level and formulate macroscopic attributes. MD creates a trajectory by mathematical assimilation of classical equations of motion. The generated trajectory holds all dynamical data required for calculating several features. The RMSD of the protein backbone calculated from the MD simulations illustrated the stability of the protein‐ligand complexes. ³⁷ , ³⁸ Every amino acid residue related to the protein‐ligand complex was accountable for the endurance of the dynamic conformity. The fluctuation of any selective amino acid regarding the native structure can be estimated by the RMSF parameter determined from the MD simulation trajectories. The RMSF for all residues belonging to the six complexes was low and showed a nearly similar style during the simulations. Assuredly, the low RMSF intimated the increase in thermodynamic stability of every complex during the MD simulations. Further, we conducted a thorough investigation at the atomic level, where hydrogen bonds between protein‐ligand were observed. The endurance of the subsequent symmetry can be associated with the formation of hydrogen bonds. ³⁹ , ⁴⁰ Hydrogen bonds interpretation inferred that all the complexes showing favorable binding sustained strong interactions through hydrogen bonding during the simulation run. The MM‐PBSA is a cutting‐edge and extensively adopted strategy that pairs the continuum solvent states and molecular mechanics to estimate free energy of binding between receptor and ligands. ⁴¹ The binding free energy was examined in terms of SASA, polar solvation, electrostatic, and van der Waal energy. It was evident without ambiguity that all the proposed molecules hold a significant binding affinity towards Nsp15 of SARS‐CoV‐2.

Moreover, we generated the 2D and 3D FEL plots to obtain Gibbs free energy, as conferred in Figure 4. The range of Gibbs free energy for diacetylcurcumin, Bis‐demethoxycurcumin, dimethylcurcumin, tetrahydrocurcumin, demethoxycurcumin, and curcumin was between 0 and 18.60 kJ/mol. The yellow locality expressed the localized minima energy and stabilized conformations by shielding eminent free energy attributes with intense basins. Each Nsp15‐curcuminoid complex had a similar shape to the FEL with stable conformations. The results of the FEL approach also facilitated the results obtained from binding free energy estimations.

5. COMPARATIVE ANALYSIS WITH EXISTING LITERATURE ON NSP15 OF SARS‐COV‐2

To classify the curcuminoids as influential inhibitors of the Nsp15 protein, we determined to compare the selected curcuminoids with co‐crystallized molecule (citrate) calculated by Chandra et al. ⁴² The cocrystallized citrate was interacting with key residues Thr341, His250, Lys290, Val292, Gln245, His235, and Tyr343. All the residues listed in the study by Chandra et al. were similar to the residues observed in binding profiles of curcuminoids selected in the present study. The RMSD and RMSF results were alike, while the intermolecular hydrogen bonds were higher in case of curcuminoids that the cocrystallized molecule. Furthermore, the authors estimated the molecules binding free energy through the MM‐PBSA method and proclaimed that the citrate showed −54 kJ/mol binding free energy, while in our study the calculated binding free energy for curcuminoids were −150.757, −113.174, −121.991, −83.513, −91.246, and −95.233 kJ/mol. The results of our study conducted with the six curcuminoids pointed out that they had a strong binding affinity with Nsp15. The tight connection between curcuminoids and binding site of Nsp15 compared to citrate could evade the infection by suppressing the Nsp15 more significantly.

6. CONCLUSION

The current research was designed to discover assuring molecules that could inhibit Nsp15 of SARS‐CoV‐2 with influential binding energy. Initially, six curcuminoids were implicitly selected to dock with Nsp15. All the chosen molecules were lower ‐CDOCKER energy. The pharmacokinetic and ADMET tests were conducted, and all the six curcuminoids were shown to exhibit lead‐like properties. Moreover, to estimate the curcuminoids endurance toward Nsp15, we conducted MD simulations and binding free energy computations through the MM‐PBSA method. Concurrently, outcomes gained via molecular docking, dynamics, and MM‐PBSA interpretations, suggested that all the six curcuminoids may perform as prospective inhibitors of Nsp15 and deserve in vivo/in vitro testing.

AUTHORS CONTRIBUTIONS

Rituraj Purohit conceived of and designed the study. Rituraj Purohit and Rahul Singh analyzed and interpreted the data. Rahul Singh and Vijay K. Bhardwaj critically revised it for important intellectual content. All the authors have read and approved the manuscript in all respects for publication.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information

Supporting information.

Singh R, Bhardwaj VK, Purohit R. Inhibition of nonstructural protein 15 of SARS‐CoV‐2 by golden spice: a computational insight. Cell Biochem Funct. 2022;40:926‐934. 10.1002/cbf.3753

DATA AVAILABILITY STATEMENT

All the related data are provided in the supplementary material.