Marine drugs as putative inhibitors against non-structural proteins of SARS-CoV-2: an in silico study

Abstract

Introduction

Coronavirus disease 2019 (COVID-19) is an unprecedented pandemic, threatening human health worldwide. The need to produce novel small-molecule inhibitors against the ongoing pandemic has resulted in the use of drugs such as chloroquine, azithromycin, dexamethasone, favipiravir, ribavirin, remdesivir and azithromycin. Moreover, the reports of the clinical trials of these drugs proved to produce detrimental effects on patients with side effects like nephrotoxicity, retinopathy, cardiotoxicity and cardiomyopathy. Recognizing the need for effective and non-harmful therapeutic candidates to combat COVID-19, we aimed to develop promising drugs against SARS-COV-2.

Discussion

In the current investigation, high-throughput virtual screening was performed using the Comprehensive Marine Natural Products Database against five non-structural proteins: Nsp3, Nsp5, Nsp12, Nsp13 and Nsp15. Furthermore, standard precision (SP) docking, extra precision (XP) docking, binding free energy calculation and absorption, distribution, metabolism, excretion and toxicity studies were performed using the Schrӧdinger suite. The top-ranked 5 hits obtained by computational studies exhibited to possess a greater binding affinity with the selected non-structural proteins. Amongst the five hits, CMNPD5804, CMNPD20924 and CMNPD1598 hits were utilized to design a novel molecule (D) that has the capability of interacting with all the key residues in the pocket of the selected non-structural proteins. Furthermore, 200 ns of molecular dynamics simulation studies provided insight into the binding modes of D within the catalytic pocket of selected proteins.

Conclusion

Hence, it is concluded that compound D could be a promising inhibitor against these non-structural proteins. Nevertheless, there is still a need to conduct in vitro and in vivo studies to support our findings.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00894-023-05574-9.

Introduction

A new case of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was first exposed in Wuhan, China [1, 2]. It is a deadly virus which disseminates amongst humans and other mammals causing a broad range of infections from common cold to fatal diseases such as respiratory syndrome. SARS-CoV-2 is a round-shaped, single-stranded RNA virus consisting of 30,000 bp of genomic sequence incorporating approximately eleven open reading frames (ORFs) encoding both structural and non-structural proteins intricate in viral life cycle [3–5]. The first and foremost protein, playing a key role in the viral infection and adaptive immunity, is spike glycoprotein (SGP). SGPs usually project from the surface of mature virions which play a role in fusion, attachment and entry into the host cell. The virions bind to the angiotensin-converting enzyme 2 (ACE-2) receptor via SGPs and mediate viral infection. Thus, SGPs’ surface location renders it as a direct anti-CoV-2 target for most of the therapeutic interventions. After the entry of virus into the host cell, it releases its RNA into the cell cytoplasm followed by the translation of its replicase gene. The two overlapping ORFs (1a and 1b) of replicase gene translated to polyproteins 1a and 1b, which are subsequently cleaved by main protease/3-chymotrypsin-like protease (Mpro/3CLpro) and papain-like protease (PLpro), respectively. The cleavage of polyproteins results in the formation of 16 non-structural proteins (Nsp1–16). Amongst them, papain-like protease (Nsp3) [6, 7], main protease (Nsp5) [8, 9], RNA-dependent RNA polymerase (RdRp/Nsp12) [10, 11], helicase (Nsp13) [12, 13] and endoribonuclease/XendoU (Nsp15) [14, 15] are considered to be viable anti-CoV-2 targets (Fig. 1) [4, 16]. The details of these five non-structural proteins are represented in Table 1.

Fig. 1.

Formation of non-structural proteins of SARS-CoV-2

Table 1.

Pathogenicity of non-structural proteins

Non-structural proteins	Mechanism of pathogenicity	References
Nsp3	Possesses ~1945 amino acids in SARS-CoV-2 and is a papain-like protease (PLpro) (multi-pass membrane protein) which acts on polyprotein to release Nsp1, Nsp2 and Nsp3. It displays deISGylating and deubiquitinating activities (removal of ubiquitin and interferon-stimulated gene 15). Interacts with Nsp4 and Nsp6 to promote virulence	Báez-Santos et al. [6], 2015
Nsp5	Also called 3CLpro; possesses ~306 amino acids in SARS-CoV-2. It is involved in the maturation of Nsps. Furthermore, it acts on 11 sites of polyprotein to release Nsp4 to Nsp16	Wu et al. [9], 2020
Nsp12 (RNA-dependent RNA polymerase (RdRp))	Consists of ~932 amino acids in SARS-CoV-2. With the aid of Nsp7 and Nsp8, Nsp12 involves in both replication and transcription of the viral genome. It also exhibited >95% similarity with that of SARS-CoV polymerase	Snijder et al. [11], 2016
Nsp13 (multifunctional superfamily 1 helicase)	Consists of ~601 amino acids in SARS-CoV-2. It uses both dsDNA and dsRNA with 5′–3′ polarity in viral genome replication. Furthermore, it plays a key role in mRNA splicing and capping	Jang et al. [12], 2020
Nsp15 (endoribonuclease)	It consists of ~346 amino acids in SARS-CoV-2 and is involved in RNA cleavage at the 3′ ends of uridylates. It plays a major role in the replication and evasion of dsRNA sensors	Deng et al. [14], 2017

dsRNA double-stranded RNA

Because of the COVID-19 outbreak, pharmaceutical hubs, private and government organisations, institutions and biotechnology firms are under burdened to produce high-quality, safety and low-cost drugs or vaccines against the ongoing unprecedented pandemic. Therefore, drugs such as hydroxychloroquine, ivermectin, azithromycin and various antiviral drugs have been compassionately used for the management of COVID-19 [17]. Although significant data are available from the randomized clinical and preclinical studies, there were many shreds of evidence on the adverse effects of these therapies on virus-infected patients [18]. The major side effects that are encountered after administering the anti-COVID trial drugs are cardiomyopathy, nephrotoxicity, retinopathy, cardiotoxicity and hepatotoxicity [18–20]. As a result, the need for safer and more easily available solutions to combat this viral scourge has intensified.

Naturally occurring compounds are now amongst the most common sources of prototypes for antimicrobial and antiviral drugs that continue to be developed today. Thousands of substances derived from aquatic animals are currently being studied in depth for use in medicine, with more than forty now in the medicine market. Several chemical groups of naturally biologically active substances, such as peptides, flavonoids and alkaloids, have been successfully tested against SARS-CoV-2.

The global history in marine pharmacy attests to the enormous potential of marine species as raw materials for the production of novel medicinal substances. In the course of evolution, aquatic species have evolved several anti-infectious mechanisms and molecules to shield them from assaults of microbes and viruses that inhabit the marine environment. Compounds isolated from aquatic animals that suppress DNA and RNA viruses, including coronaviruses, have been discovered in a variety of structural groups, including polysaccharides, terpenoids, steroids, alkaloids and peptides. The various mechanisms used by these chemical groups to suppress coronaviruses account for their diversity.

Recognizing the uniqueness of their possibilities to treat a variety of diseases, we looked at a subset of its derivatives in this article to learn more about how they function as inhibitors against the above-mentioned viral non-structural proteins using in silico approaches [21–23]. Recent studies on the SARS-CoV-2 targets, such as Mpro, PLpro and RdRP, have highlighted the value of structure-based drug design in the search for powerful compounds from a variety of chemical libraries, including azadirachtins, ceramicines, withanolides and nucleotide analog medicines [24–27]. As a result, the creation of new medications for the control and treatment of this continuing pandemic may benefit from unique alternative evidence based on a molecular docking approach. In this paper, virtual screening and molecular docking were employed to screen a large comprehensive marine database library (https://cmnpd.org/) against non-structural proteins of SARS-CoV-2 such as Nsp3, Nsp5, Nsp12, Nsp13 and Nsp14. Furthermore, pharmacological and toxicological parameters were evaluated for the best hit compounds. The final hits were used for the designing of a novel compound (D), and its potential of inhibiting the Nsps through computational approaches was evaluated.

Materials and methods

Protein crystal structure preparation

The protein X-ray crystal structures of five non-structural proteins (Nsp3, Nsp5, Nsp12, Nsp13 and Nsp14) were retrieved from the protein databank, with accession IDs of 6LU7 [7], 6W9C [8], 6M71 [10], 6XEZ [13] and 6VWW [15], respectively. Protein preparation wizard available in the Glide module was used for the rectification of specific errors like missing loops and hydrogen atoms in the structure of the proteins during crystallographic studies [28–30]. Before optimization, water molecules located >3 Å were removed from the protein crystal structure. Furthermore, hydrogen bonds and missing side chain atoms were added and repaired using the module Prime. Minimization of protein was carried out using OPLS3e force field with heavy atom convergence criteria to a root-mean-square deviation (RMSD) of 0.3 Å [31]. Ramachandran plots of all five proteins were assessed to determine stereospecificity and favourable and unfavourable regions in the protein crystal structure. A 3D grid box of 10 Å was created, defining the co-crystallized ligand, active residues and dimensions of the catalytic site (Table 2) [32].

Table 2.

Dossier on receptors, grid box parameters and binding pocket residues

Name of the receptors (PDB ID)	Number of residues	XYZ coordinates		Resolution; r-factor; r-free	References
		Centre grid box (XYZ), Å	Dimension (XYZ), Å
Nsp3 (6LU7)	306	−39.05 × 39.62 × 32.04	33.50 × 28.55 × 29.87	2.16; 0.202; 0.235	Jin et al. [7], 2020
Nsp5 (6W9C)	317	−14.85 × 14.92 × 69.59	25.02 × 27.98 × 30.87	2.70; 0.235; 0.279	Osipiuk et al. [8], 2021
Nsp12 (6M71)	942	123.94 × 136.89 × 128.54	20.84 × 18.27 × 25.00	2.90	Gao et al. [10], 2020
Nsp13 (6XEZ)	932	−13.89 × 15.19 × −73.25	29.92 × 25.83 × 24.80	3.50	Chen et al. [13], 2020
Nsp15 (6VWW)	370	−91.15 × 21.86 × −30.63	20.62 × 25.00 × 24.23	2.20; 0.158; 0.178	Kim et al. [15], 2020

Comprehensive marine database library preparation

Approximately 47,451 compounds from marine origin were downloaded from Comprehensive Marine Natural Products Database (https://www.cmnpd.org). Ligands were prepared using the LigPrep module available in the Maestro interface of Schrödinger suite. Low-energy conformers were segregated, and possible ionization states were generated for the 2D structures of all the ligands at a physiological pH of 7.2 ± 0.2. All the ligands were minimized using an optimized OPLS3e force field by keeping all other options as default.

High-throughput virtual screening and molecular docking studies

High-throughput virtual screening (HTVS) was carried out in the binding pocket of Nsp3, Nsp5, Nsp12, Nsp13 and Nsp15 via the Glide module of the Schrӧdinger suite. Prior to docking, the non-structural proteins were made rigid and the ligands were made flexible. Before the analysis, the physicochemical characteristics of the database library (47,451 compounds) acquired from CMNPD were filtered. Virtual screening is divided into three categories: HTVS, SP and XP docking. These three docking methods were used to take out a possible lead molecule in a short period. Although HTVS quickly screens a huge number of molecules, the sampling techniques were limited, and the results could not be immediately evaluated. As a result, the ligands identified by HTVS were docked using SP, which selects an appropriate binding posture from a vast pool of ligands [33]. Furthermore, the top 10% of SP molecules was chosen for docking in the XP mode and examined using an XP visualizer. Based on their Glide score, Glide E-model and Glide energy, the top 10% of molecules was picked for free energy estimates [34–37].

Molecular mechanics with generalized Born surface area calculation

Binding free energy (BFE) calculations were carried out for XP-docked complexes using molecular mechanics with generalized Born surface area (MM-GBSA) approach. Docked complexes were minimized using local optimization feature in the module Prime. OPLS3e force field was employed to evaluate the BFE for a set of protein-ligand complexes [31].

$$\mathrm{\Delta }\text{Gbind}=\mathrm{\Delta }\text{GSA}+\mathrm{\Delta }\text{Gsolv}+\mathrm{\Delta }\text{EMM}$$

where ΔGSA is the difference in surface area energies in the protein and ligand and the complexes, ΔGsolv is the difference between the solvation energies of the complexes and individual proteins and ligands and ΔEMM is the variation between the minimized energy of the protein-ligand complexes.

Pharmacological and toxicological parameters

After molecular docking studies, SwissADME (http://www.swissadme.ch/) was utilized to estimate the absorption, distribution, metabolism, excretion and toxicity (ADMET) characteristics of the compounds. This module calculates numerous attributes and offers values for descriptors that forecast ligand drug likeness. It also generates functional groups, rings and a variety of other components that are used to predict ligand descriptors [38]. This approach is also used to connect the physicochemical and pharmacokinetic features of the test ligands to support the discovery of small-molecule inhibitors.

Molecular dynamics simulation studies

Molecular dynamics simulation (MDS) studies were performed for the compound D with the selected non-structural proteins using OPLS3e force field. The systems were solvated using TIP4P water in orthorhombic boxes within the Desmond module [39, 40]. Counter ions were added to the systems to achieve neutralization. A tolerance of 1e−09 for long-range interactions and a 9-Å cut-off radius for short-range columbic were applied using smooth particle mesh Ewald method [41]. The MDS was performed for 200 ns each for the selected non-structural protein-ligand complexes under NPT conditions, which were maintained by Martyna-Tobias-Klein (1 bar pressure) and Nose-Hoover thermostat (300 K). A steepest descent method was used to minimize the gradient energy of optimized structures [42–44].

Results and discussion

The SARS-CoV-2 infection continues to disseminate like wildfire, demanding urgent attention from scientists across the globe to discover a possible therapeutic against viral infection.

Molecular docking studies, post docking analysis and binding free energy calculation

Molecular docking plays a crucial role in the designing of novel drugs against non-structural proteins of interest. It predicts the experimental binding mode and binding affinity of molecules within the catalytic site of the selected proteins. In the first mode of HTVS, the screening of a marine database library was carried out using the selected non-structural proteins. Nearly 30% of obtained hits was used for SP docking and XP docking analysis. The results of the molecular docking studies along with the 2D structures and IUPAC names of the predicted ligands are depicted in Table 3 and Fig. 2. The compounds having high negative Glide score and which are observed in common with all the proteins were described briefly.

Table 3.

Docking and binding free energy results of compounds within the catalytic pocket of respective non-structural proteins

Gs Glide score, Gb binding free energy

Fig. 2.

Two-dimensional diagram of a CMNPD5804/Nsp3, b CMNPD20924/Nsp5, c CMNPD20924/Nsp12, d CMNPD15988/Nsp13 and e CMNPD5804/Nsp15 docking complex exposing key interactions in the catalytic pocket of non-structural proteins

Nsp3 (PDB ID: 6LU7)

From the docking results, it is clear that CMNPD5804 had the most favourable Gscore of −7.804 kcal/mol. The compound also well fitted within the catalytic pocket of Nsp3 and exhibited four hydrogen bonding (HB) interactions with the key amino acid residues. The –OH of the phenolic moiety formed a HB interaction with the Lys102. Furthermore, the –OH present at positions 4 and 8 of the quinoline nucleus formed two HB interactions with the Gln110 and Tyr154, respectively. In addition, the carbonyl group present at position 5 of the quinoline moiety also showed a HB interaction with the protonated Asn151. Furthermore, the compound has been stabilized in the catalytic pocket of Nsp3 by the formation of two π-π stacking interactions between the fused rings of quinoline nucleus and electron cloud of Phe294. Moreover, the quinoline nucleus of the compound was found to be crucial for the binding affinity as evident by its ΔGbind −75.56 kcal/mol and key interactions.

Nsp5 (PDB ID: 6W9C)

From the docking result, it is clear that CMNPD20924 had the most favourable Gscore of −8.928 kcal/mol against Nsp5. The compound also well fitted within the catalytic pocket of Nsp5 and exhibited four HB interactions with the key amino acid residues. The two –OH molecules of the catechol moiety formed two HB interactions with Gly100. Furthermore, the two –OH groups of resorcinol formed a HB interaction with Ser278 and the other –OH present at the meta position of this structure formed two HB interactions with Lys279 and Thr277.

Nsp12 (PDB ID: 6M71)

With the help of docking results, it can be seen that CMNPD20924 had the most suitable Gscore of −9.002 kcal/mol. The compound also interacted well with the catalytic pocket of Nsp12 and showed four HB interactions with the essential amino acid residues. The resorcinol moiety exhibited three HB interactions with Thr394, Arg249 and Arg457. Furthermore, the –OH moieties of the catechol moiety showed two HB interactions with Glu350 and Ser318. In addition, the carbonyl group present at position 2 of the dihydrofuran moiety showed a HB interaction with Phe396. Moreover, the catechol nucleus of the compound exhibited a π-cationic interaction with Arg349.

Nsp13 (PDB ID: 6XEZ)

With the aid of docking outcomes, CMNPD15988 had the most favourable Gscore of −8.569 kcal/mol with Nsp13. The compound also connected most precisely with catalytic residues with four HB interactions and one π-π interaction. The –OH of the imidazole moiety formed a HB interaction with Tyr38. The carbonyl group present at position 2 of the imidazole moiety formed a HB interaction with Asn209. The –NH of the pyrrole moiety and carbonyl group formed two HB interactions with Asp221 and Arg733. Moreover, the compound has been stabilized in the catalytic pocket of Nsp13 by the formation of salt bridge with Mg1003.

Nsp15 (PDB ID: 6VWW)

Given that, CMNPD5804 had the stronger Glide score of −8.639 kcal/mol with Nsp15. The compound was well locked with the catalytic pocket of Nsp15 and demonstrated three HB interactions with the amino acid residues. The –OH and –COOH moiety of the quinoline moiety formed two HB interactions with Lys290 and Ser294. Furthermore, the carbonyl group exhibited a HB interaction with Ser294. The phenolic moiety made a π-π interaction with Trp333.

ADMET properties

The SwissADME was used to calculate the pharmacological and toxicological (ADMET) characteristics, as well as to theorize on the acceptance of title compounds (CMNPD5804, CMNPD20924, CMNPD15988, CMNPD26614, CMNPD2414) (Table 4). The total amount of overall polar atoms or molecules in a compound is within the acceptable range (<120 Ų) of topological surface area (TSA) (primarily oxygen and nitrogen, including their bonded hydrogen atoms). None of the compounds was permeable through the blood-brain barrier. Amongst these five compounds, only CMNPD20924 was considered as a P-glycoprotein substrate and is constantly pumped out from the brain. Log Kp is a skin penetration indicator that shows the absorption of a substance via the skin. The results show that all drugs are absorbed through the skin. All these compounds had a substantial bioavailability score (BS), indicating that they had good permeability and bioavailability. Pains alert (PA) (PAN assay interference chemicals) were absent in all the compounds, which implicates a specific interaction with the chosen target and no contact with other unintended biological targets.

Table 4.

ADMET parameters of compounds

Compound	TSA (Å2)	BBB	P-gp	CYP1A2a inhibitor	CYP2c19a inhibitor	CYP2c9a inhibitor	CYP2D6a inhibitor	CYP3A4a inhibitor	Log Kp (cm/s)b	BS	PA
CMNPD5804	127.95	No	No	No	No	No	No	No	−6.63	0.56	0
CMNPD20924	107.22	No	Yes	No	No	No	No	No	−6.86	0.55	0
CMNPD15988	123.32	No	No	No	No	No	No	No	−8.56	0.55	0
CMNPD26614	86.71	No	No	No	No	No	No	No	−7.35	0.55	0
CMNPD2414	71.98	No	No	Yes	No	No	No	No	−6.71	0.55	0

TSA (<120 Ų) topological surface area (sum of overall polar atoms or molecules), BBB blood-brain barrier, P-gp P-glycoprotein substrate, BS bioavailability score, PA pains alert

^aCytochrome P450 enzymes intricate xenobiotic metabolism

^bSkin permeation constant

Design of novel compound

From the docking results, it is clear that the three compounds (CMNPD5804, CMNP15988 and CMNPD20924) exhibited significant interactions with the key amino acid residues of all non-structural proteins. The nuclei responsible for the interactions were isolated and conjoined to generate one molecule (D) (Fig. 3). Furthermore, docking studies and MM-GBSA were also performed to the designed molecule. The results were astonishing, and all the non-structural protein (D) complexes established greater binding affinity when compared with the hits obtained (Table 5, Fig. 4).

Fig. 3.

Design of novel compound (D) by conjoining the nuclei of top-score hits

Table 5.

Docking, MM-GBSA results against selected non-structural proteins and ADMET properties of compound D

Non-structural proteins	Gs	ΔGb	ADMET properties
Nsp3	−7.234	−90.562	TSA (Å2)	152.61	CYP1A2a	Yes
Nsp5	−7.191	−67.119	BBB	No	CYP2C19a	No
Nsp12	−6.514	−98.478	P-gp	No	CYP2C9a	Yes
Nsp13	−7.331	−119.69	Log Kpb	−6.80	CYP2D6a	No
Nsp15	−8.274	−67.771	BS	0.11	PA	0

TSA (<120 Ų) topological surface area (sum of overall polar atoms or molecules), BBB blood-brain barrier, P-gp P-glycoprotein substrate, BS bioavailability score, PA pains alert

^aCytochrome P450 enzymes intricate xenobiotic metabolism

^bSkin permeation constant (cm/s)

Fig. 4.

Two-dimensional diagram of a D/Nsp3, b D/Nsp5, c D/Nsp12, d D/Nsp13 and e D/Nsp15 docking complex exposing key interactions in the catalytic pocket of non-structural proteins. Furthermore, MDS studies were performed to get insights into binding modes of the complexes D/6LU7, D/6W9C, D/6M71, D/6XEZ and D/6VWW by analysing the trajectory frames

The complex D/6LU7 exhibited a Gscore of −7.239 kcal/mol and a ΔGbind score of −90.562 kcal/mol with four HB interactions. The –OH groups of resorcinol formed two HB interactions with Thr169 and Gly170, while the carbonyl oxygen linker showed a HB interaction with Lys137. On the other side of the compound, quinoline exhibited two HB interactions with Gln127 (–OH of the carboxyl moiety) and Lys137 (–OH at position 8). Furthermore, the compound was stabilized by the π-cationic interaction between the π-cloud of quinoline and the protonated Lys5 residue.

The complex D/6W9C showed a Gscore of −7.191 kcal/mol and a ΔGbind score of −67.119 kcal/mol and exhibited eight HB interactions. The –OH of resorcinol nucleus had a HB interaction with Gln97 whereas the –NH at position 3 of resorcinol exhibited a HB interaction with Gly100.The pyrrole –NH exhibited a HB interaction with Gln121, while the carbonyl oxygen linker showed a HB interaction with Lys279. The carboxyl moiety at position 5 and –OH at position 8 of the quinoline nucleus established four HB interactions with Gln122, Glu124 and Lys279.

The complex D/6M71 established a Gscore of −6.514 kcal/mol and a ΔGbind score of −98.478 kcal/mol and exhibited six HB interactions. The carboxyl moiety of quinoline nucleus exhibited two HB interactions with Ile266 and Trp268. The –NH of pyrrole exhibited a HB interaction with Phe321. The –OH groups of resorcinol formed two HB interactions with Asn459 and Pro677, while carbonyl oxygen linker showed a HB interaction with Arg349. Moreover, the compound was stabilized by π-π stacking interaction between the π-cloud of quinoline nucleus and the phenol ring of Tyr265.

The complex D/6XEZ exhibited a Gscore of −7.331 kcal/mol and a ΔGbind score of −119.69 kcal/mol and exhibited five HB interactions. The –OH and –COOH on quinoline nucleus exhibited two HB interactions with Thr324 and Pro378, respectively. The –NH of pyrrole nucleus exhibited a HB interaction with Phe326. The carbonyl oxygen of amide linker exhibited a HB interaction with Arg349, while the –OH of resorcinol exhibited a HB interaction with Val675.

The complex D/6VWW displayed a Gscore of −8.274 and a ΔGbind score of −67.771 and five HB interactions. The –OH group of resorcinol exhibited two HB interactions with Ser294 and Leu346. The carbonyl oxygen of amide linker exhibited a HB interaction with Lys290. Furthermore, nitrogen of quinoline nucleus exhibited a HB interaction with His235, while the –OH group present at position 8 of the quinoline nucleus formed a HB interaction with Asp240. The –NH of pyrrole nucleus exhibited a HB interaction with Thr341 while the carbonyl oxygen of amide linker exhibited a HB interaction with Lys290. The compound was stabilized by the formation of the π-cationic interaction between the π-cloud of pyrrole and the protonated Lys290 residue. In addition, ADMET properties of compound D using SwissADME revealed that the compound D possesses significant druggable properties with TSA, complying with the recommended range (<120 Ų). Furthermore, it exhibited non-permeability through the BBB and also non-substrate for poly-glycoprotein, indicating it cannot be pumped out of the brain or lumen of gastrointestinal tract. Compound D is also an inhibitor for only two cytochromes: CYP1A2 and CYP2C9. From the value of log Kp (−6.80 cm/s), it is clear that the compound D is permeated through the skin. Furthermore, it also exhibited zero PA, indicating that the interaction with the desired target is a specifically bypassing interaction with other targets.

RMSD (Fig. 5a–e) of initial structures of D/6LU7, D/6W9C, D/6M71 and D/6XEZ increased during equilibration and converged till 15 ns of a 200-ns MDS study, while the complex D/6VWW exhibited equilibration till 120 ns which may be attributed to the structural alterations of the protein residues.

Fig. 5.

RMSD (Å) of the simulated positions of Cα, backbone and heavy atoms of a D/6LU7, b D/6W9C, c D/6M71, d D/6XEZ and e D/6VWW complexes during MDS

After equilibrium, the RMSD values of D/6LU7 (Cα: 1.47–2.62 Å; BB: 1.47–2.17 Å; HA: 1.89–2.82 Å), D/6W9C (Cα: 1.86–2.81 Å; BB: 1.91–2.80 Å; HA: 2.35–3.18 Å), D/6M71 (Cα: 2.69–3.10 Å; BB: 2.61–3.01 Å; HA: 2.88–3.36 Å), D/6XEZ (Cα: 2.53–3.32 Å; BB: 2.52–3.31 Å; HA: 2.93–3.65 Å) and D/6VWW (Cα: 2.34–4.63 Å; BB: 2.35–4.61 Å; HA: 2.72–4.87 Å) indicated minimal fluctuations in the protein structure.

Furthermore, root-mean-square fluctuation (RMSF) of protein residues is presented in Fig. 6a–e. The D/6LU7 complex established lower RMSF values (<2 Å) at the ligand contacts Arg4 to Thr26, Gln69 to Gln74, Ala116 to Gln127, Lys137 to Gly143 and Glu166 to His172, indicating lower fluctuations of the residues. Furthermore, no ligand contacts were observed after His172 except Leu286, Glu288 and Glu290, whereas the complex D/6W9C exhibited lower RMSF values (<2 Å) at the ligand contacts Tyr95 to Thr102, Leu120 to Leu125, Ala139 to Ala144 and Thr277 to Glu280, indicating minimal fluctuations of the residues. In addition, no ligand contacts were observed between Arg3 and Lys94. The complex D/6M71 also exhibited lower RMSF values with the contacting residues Thr248 to Ser225, Pro264 to Trp268, Val315 to Gly327, Tyr456 to Thr462 and Tyr674 to Gly679, indicating minimal residual fluctuations. No ligand contacts were found between Thr680 and Leu931. Furthermore, the complex D/6XEZ exhibited lower RMSF values (<2 Å) with the contacting residues Tyr265 to Tyr273, Phe321 to Lys332, Pro378 to Ser397, Arg457 to Leu460 and Leu663 to Met666. Though significant fluctuations were found for some residues, they are not in contact with the ligand. The other complex D/6VWW was found to possess lower RMSF values (<2 Å) with the ligand contacting residues Glu229 to His250, Gly287 to Ser294, Val314 to Ser316 and Trp333 to Gln347, indicating minimal residual fluctuations and greater stability of protein-ligand interactions. Besides, no ligand contacts were observed between Leu3 and Leu228.

Fig. 6.

RMSF (Å) of the simulated positions of Cα, backbone and heavy atoms of a D/6LU7, b D/6W9C, c D/6M71, d D/6XEZ and e D/6VWW complexes during MDS

Analysis of MD trajectory of the compound D exposed hydrogen bonding, π-π stacking and π-cationic interactions binding pocket residues of the proteins Nsp3, Nsp5, Nsp12, Nsp13 and Nsp15, depicted in Supplementary Figs. S1a–e and S2a–e. All the complexes exhibited similar binding modes of interactions as speculated by the molecular docking study. The complex D/6LU7 established five hydrogen bonding interactions with Glu14 (15% and 25% MD trajectory), Gln19 (13% MD trajectory), Ty118 (43% MD trajectory), Asn119 (11% MD trajectory), Ser123 (13% MD trajectory; water-bridge interaction) and Ser139 (35% MD trajectory). Furthermore, the complex was stabilized by forming the π-π interaction with Phe140, whereas the complex D/6W9C exhibited stable hydrogen bonding interactions with the polar residues of the proteins Gln97 (31% MD trajectory), Thr102 (77% MD trajectory), Gln121 (36% MD trajectory) and Glu124 (71% MD trajectory) and the hydrophilic residue Tyr95 (48% MD trajectory; water-bridge interaction). Furthermore, the complex D/6M71 also exhibited significant hydrogen bonding interactions with most of hydrophilic residues. The –NH of the pyrrole linker established a stable hydrogen boding interaction with Phe321 (99% MD trajectory). The hydroxy groups of the resorcinol and –COOH moiety of quinoline established four hydrogen bonding interactions with Asn459 (85% MD trajectory), Glu350 (68% MD trajectory), Leu460 (25% MD trajectory; water-bridge interaction) and Ile266 (63% MD trajectory). Moreover, the complex was also stabilized by forming the π-π stacking interaction between the electron cloud of quinoline and the aromatic ring of Tyr265.

The complex D/6VWW established moderately favourable hydrogen bonding interactions with the residues Glu234 (38% MD trajectory), Asp240 (35% MD trajectory), Cys334 (23% MD trajectory), Glu340 (11% MD trajectory) and Pro344 (10% MD trajectory). Furthermore, the complex was stabilized by forming π-π stacking and π-cationic interactions with His234, His235 and Lys335, respectively.

The ligand properties of the compound D (Supplementary Fig. S3a–e) with respect to the selected non-structural proteins revealed that the compound D established an RMSD of 0.75–2.83 Å (6LU7), 0.41–1.93 Å (6W9C), 0.66–0.69 Å (6M71), 0.71–1.96 Å (6XEZ) and 0.81–2.79 Å (6VWW), indicating less conformational changes and increased stability of compound D during the MD study. Furthermore, the radius of gyration was found to be in the range of 5.04–5.71 Å (6LU7), 4.83–5.81 Å (6W9C), 5.51–5.66 Å (6M71), 5.36–5.76 Å (6XEZ) and 4.83–5.81 Å (6VWW). In addition, polar surface area, solvent accessible surface area and molar surface area values indicated the stabilization of MD study.

MM-GBSA was performed for the MD complexes D/6LU7, D/6W9C, D/6M71, D/6XEZ and D/6VWW, and the results are depicted in Table 6 and Supplementary Fig. S4a–e. The complex D exhibited significant binding energy with the selected non-structural proteins in the range of −33.28 to −67.66 kcal/mol. ΔGcou energy term was found to be favourable for 6LU7, 6W9C and 6XEZ and unfavourable against 6M71 and 6VWW. Besides, ΔGcov and ΔGlip were found to be moderately favourable for the selected non-structural proteins. From the results, it is clear that ΔGvdW (−30.94 to −81.10 kcal/mol) and ΔGb were found to be the key driving forces for the inhibitory action of compound D against the selected non-structural proteins.

Table 6.

MM-GBSA calculations for the selected complexes

Complex	ΔGb	ΔGcou	ΔGcov	ΔGlip	ΔGvdW	Interactions
						H bonding	π-π stacking
D/6LU7	−67.66	−53.20	7.306	−16.32	−52.64	Glu14, Gln19, Gly71, Asn119, Ser121, Asn142	–
D/6W9C	−43.98	−36.30	−2.78	−11.06	−30.62	Thr102, Gln121, Glu124	–
D/6M71	−40.82	13.77	10.15	−22.72	−81.10	Arg249, Phe321, Glu350, Asn459	Tyr265
D/6XEZ	−49.90	−45.32	14.50	−13.86	−30.94	Phe326, Arg349, Lys391, Arg457	Phe396
D/6VWW	−33.28	−7.35	9.77	−11.34	−40.92	Glu234, Glu340	His235

ΔGb is the binding free energy, ΔGcou is the coulombic interaction energy, ΔGcov is the covalent energy, ΔGlip is the lipophilic-solvation energy and ΔGvdW is the van der Waals energy

Conclusion

In summary, HTVS, SP docking, XP docking and binding free energy calculation studies were performed using the Comprehensive Marine Natural Products Database against five non-structural proteins: Nsp3, Nsp5, Nsp12, Nsp13 and Nsp15. Amongst the top hits, CMNPD5804, CMNPD20924 and CMNPD1598 compounds were utilized to design a novel molecule (D) which has the capability of interacting with all the key residues in the pocket of the selected non-structural proteins. The compound D exhibited greater binding affinity with the values of Gscore (−7.234 kcal/mol) and Gbind (−90.562 kcal/mol) for Nsp3, Gscore (−7.191 kcal/mol) and Gbind (−67.119 kcal/mol) for Nsp5, Gscore (−6.514 kcal/mol) and Gbind (−90.478 kcal/mol) for Nsp12, Gscore (−7.331 kcal/mol) and Gbind (−119.69 kcal/mol) for Nsp13 and Gscore (−8.274kcal/mol) and Gbind (−67.771kcal/mol) for Nsp15. Furthermore, MDS studies for the complexes D/6LU7, D/6W9C, D/6M71, D/6XEZ and D/6VWW revealed the greater stability of the compound D within the binding pocket of the selected non-structural proteins. These results indicated that compound D could be a promising inhibitor against these non-structural proteins. However, still the study requires further in vitro and in vivo studies to support our findings.

Author contribution

Simran Patel: writing of original draft and data curation; Haydara Hasan: writing of original draft and methodology; Divyesh Umraliya: formal analysis; Bharat Kumar Reddy Sanapalli: writing including reviewing and editing, and conceptualization; Vidyasrilekha Yele: writing including reviewing and editing, software, supervision and conceptualization.

Data availability

Not applicable.

Code availability

Not applicable.

Declarations

Conflict of interest

The authors declare no competing interests.