In-silico study for the identification of potential destabilizers between the spike protein of SARS-CoV-2 and human ACE-2

Abstract

The emergence of the new SARS-CoV-2 virus, which causes the disease known as COVID-19, has generated a pandemic that has plunged the world into a health crisis. The infection process is triggered by the direct binding of the receptor-binding domain (RBD) of the spike (S) protein of SARS-CoV-2 to the angiotensin-converting enzyme 2 (ACE2) of the host cell. In the present study, virtual screening techniques such as molecular docking, molecular dynamics, calculation of free energy using the GBSA method, prediction of drug similarity, pharmacokinetic, and toxicological properties of various ligands interacting with the RBD-ACE2 complex were applied. The ligands radotinib, hinokiflavone, and ginkgetin were identified as potential destabilizers of the RBD-ACE2 interaction, which could produce their pharmacological effect by interacting at an allosteric site of ACE2, with affinity energy values of −10.2 ± 0.1, −9.8 ± 0.0, and −9.4 ± 0.0 kcal/mol, indicating strong receptor affinity. The complex with hinokiflavone showed the highest conformational stability and rigidity of the dynamic simulation and also obtained the best binding free energy of the three molecules, with an energy of −215.86 kcal/mol.

1. Introduction

In December 2019, a new coronavirus (CoV) emerged and rapidly spread throughout the world, causing an unprecedented pandemic with these types of viruses. This virus was named SARS-CoV-2 by the World Health Organization. The respiratory disease caused by the virus was designated as coronavirus disease 2019 (COVID-19) [1].

CoVs have four structural proteins, namely the envelope protein (E), the membrane protein (M), the nucleocapsid protein (N), and the spike protein (S) [2]. The S protein plays the most crucial roles in viral attachment, fusion, and entry, and is the primary determinant of CoV tropism [3]. Viral entry depends on the binding of the S1 subunit to the angiotensin-converting enzyme 2 (ACE2) via the receptor-binding domain (RBD) on the S1 subunit, which facilitates viral attachment to the surface of target cells [4]. Genome sequence data has revealed at least 70% similarity between SARS-CoV-1 and SARS-CoV-2 [[5], [6], [7], [8]]. Despite the similarities, the S protein of SARS-CoV-2 has 20 times higher affinity with human ACE2 than the S protein of SARS-CoV-1, which leads to faster spread from one human to another [5,[7], [8], [9], [10]].

Given the severity of the disease and the rapid increase in the number of affected individuals, therapies are needed to address COVID-19. Scientists, governments, regulators, and pharmaceutical companies from all over the world have joined forces to develop drugs and vaccines against SARS-CoV-2 [11]. Currently, effective vaccines are being administered. However, especially since a large fraction of the world population remains unvaccinated, the potential of the constantly mutating virus to become at least partially resistant to the vaccines exists, for this reason the development of drugs must continue [12]. The first two oral antivirals, molnupiravir and nirmatrelvir-ritonavir, are now available in many countries. These antivirals must be prescribed within five days of symptom onset and after SARS-CoV-2 infection has been confirmed. However, the availability of these antivirals will be limited for some time due to manufacturing constraints [13].

In silico methodologies have opened new avenues of research and are now widely accepted as a useful tool for shortening delivery timelines, understanding and predicting drugability in early drug discovery [14]. Computers have a relevant role in this process, as they are capable of analysing large compound collections and predicting which molecules are more likely to interact with the target and become a lead compound or successful and promising [15]. The idea is to narrow down the search as much as possible to avoid the expense of large-scale screening [16].

Molecular docking aims to predict the structure of the intermolecular complex formed between two or more constituent molecules, typically a small molecule and a target protein [17]. It is a useful method for finding the best therapeutic candidates from an existing library [18]. However, despite docking programs being fast, one of the main methodological problems is obtaining accurate results [19]. A widely used practice to optimize results is pairing with molecular dynamics (MD) simulations. By performing MD simulations, the dynamic behavior of molecular arrangements can be monitored and tested on different timescales [20]. Combining both approaches is a practical habit to enhance computational drug design projects [21].

The SARS-CoV-2 protein S and human ACE2 are key players in the virus infection process. This study aims to identify potential drugs or phytochemicals that can destabilize the RBD-ACE2 interaction through molecular docking and simulation of the molecular dynamics. A virtual screening approach was applied to identify candidates that exhibit suitable drug-like properties and can serve as alternative treatments for COVID-19 pandemic and in general, coronaviruses.

2. Methodology

2.1. Preparation of the receptor

The crystal structure of the RBD SARS-CoV-2/ACE2 complex (PDB ID: 6M0J) with a resolution of 2.45 Å was obtained from the Protein Data Bank repository (https://www.rcsb.org). The structure was prepared using AutoDock Tools (ADT; version 1.5.6) [22] following these steps: removal of water molecules and N-acetyl-d-glucosamine, addition of hydrogen atoms and Kollman charges, nonpolar hydrogen fusion, and rotational bond assignment. Zinc was not removed as it is an essential cofactor for mechanism of proteins [23]The generated file was saved in pdbqt format for further analysis.

Next, contact residues in the RBD-ACE2 interface, as described by Lan et al. [24], were identified to generate a grid box large enough to contain the ligand structures. The residues for the SARS-CoV-2 RBD were Lys417, Gly446, Tyr449, Tyr453, Phe456, Ala475, Phe486, Asn487, Tyr489, Gln493, Gly496, Gln498, Thr500, Asn501, Gly502, and Tyr505, and the residues for ACE2 were Glln24, Thr27, Phe28, Asp30, Lys31, His34, Glu35, Glu37, Asp38, Tyr41, Gln42, Leu79, Met82, Tyr83, Asn330, Lys353, Gly354, Asp355, Arg357, and Arg393.

2.2. Selection and preparation of ligands

A literature review was conducted in the Clinical trials, Clinical Key, Drugbank, Science Direct, and Pubmed databases to identify drugs and phytochemicals with potential activity for the treatment of COVID-19. The molecules were analyzed for drug similarity, synthetic accessibility, and ADMET using the SwissADME[25] and pkCSM [26] web servers, taking into account their SMILES code available in PubChem (https://pubchem.ncbi.nlm.nih.gov). Molecules were selected following eligibility criteria: meeting Lipinski's rules and having a synthetic accessibility ≤5.00, according to SwissADME predictions, and not showing AMES toxicity or hepatotoxicity results, according to pkCSM predictions.

The key benchmark used in selecting drug candidates in the drug discovery process is the estimation of their similarity parameters with drugs. This was achieved by correlating the physicochemical properties of a given molecule with its biopharmaceutical aspect within the human body, mainly, its impact on oral bioavailability [27]. Lipinski's "Rule of five" predicts that poor absorption or permeability is more likely when there are more than five H-bond donors, 10 H-bond acceptors, the molecular weight (MWT) is greater than 500, and the calculated Log P (CLogP) is greater than five (or MlogP >4.15) [28]. A compound satisfies the rule of five if it does not violate more than two of the rules. The candidates selected in this study, by meeting the Lipinski rule, are estimated to have good biopharmaceutical properties for oral administration.

On the other hand, the parameter of synthetic accessibility is reported on a scale ranging from 1 (very easy to synthesize) to 10 (very difficult to synthesize). Accessibility ≤ five indicates that the compounds can be easily synthesized. This selection strategy in the ligand repository aims to have candidates that are easy to manufacture and therefore more readily available in the market.

The Ames test is a short-term bacterial reverse mutation assay specifically designed to detect a wide range of chemicals that can cause genetic damage leading to genetic mutations. The Ames test is used worldwide as an initial screening method to determine the mutagenic potential of new chemicals and drugs [29]. The selected ligands have a negative in silico estimation of the Ames test, which reduces the risk of generating genetic mutations in the population.

Hepatotoxicity is one of the most frequently observed adverse effects resulting from exposure to a xenobiotic. In pharmaceutical product research and development, hepatotoxicity is one of the major reasons for drug withdrawals, clinical failures, and discontinuation of drug candidates [30]. The molecules evaluated in this study do not present in silico hepatotoxicity, in order to identify the safest candidates for the treatment of COVID-19.

The 3D structures of the selected molecules were retrieved from PubChem in SDF format and converted into MOL2 format using Open Babel software version 2.3.1 [31]. The conversion was necessary for visualizing the molecules in AutoDock Tools [22]. Each ligand was prepared in AutoDock Tools [22] by adding hydrogens, Gasteiger charges, and removing non-polar hydrogens. Finally, the generated file was stored in pdbqt format for further analysis.

2.3. Molecular docking

Molecular docking was performed using AutoDock Vina version 1.1.2 [32]. These were executed by triplicate using a grid space of 18.75 Å × 43.12 Å x 22.50 Å, with dimensions of 50.00 Å × 115.00 Å x 60.00 Å. The best pose of each molecule with the RBD-ACE2 complex was stored during the molecular docking, yielding values expressed in Kcal/mol. Additionally, the interactions between the ligands and RBD-ACE2 complex were analyzed using Discovery Studio Visualizer program version 21.1.

2.4. Molecular dynamics simulation

The complexes formed by the three ligands with the best coupling scores underwent MD simulations. The E chain of the 6M0J structure was removed, and the entire MD process was run on the A chain. Unconstrained MD simulations were performed for all atoms using AMBER20 software [33]. The DM minimization, equilibrium, and production protocols and analyses were performed according to Alviz-Amador et al. [34] with some modifications. The ff14SB and GAFF2 force fields were applied to the ACE2 receptor and ligands, respectively.

The structures solvated with the TIP3P water model were minimized using 1000 steps of steepest descent followed by 1000 steps of conjugate gradient minimization, applying a constraint force constant of 25 kcal/mol-Å2 to the entire solute molecule. Heating was performed in 5000 MD steps from 100 to 300 K with a two fs time step, using a weakly coupled thermostat at constant pressure and restraining bonds involving hydrogen using SHAKE with the tolerance set to 0.00001. An 8.0 Å unbound cut-off was used. Long-range electrostatics were handled using Particle Ewald Mesh (PME), with default PME parameters for Amber and automatic update of the pair list. Following heating, constraints applied to peptides and proteins were slowly decreased from 5 to 0.5 kcal/mol-Å2 in five intervals, first minimizing each step using 1000 steepest descent steps followed by 500 conjugate gradient minimization steps, and two fs time step, followed by 50 ps MD at 300 K, constant pressure, and temperature, both with Berendsen constants of 0.2 ps. The total production time was 200 ns for each system.

Subsequently, RMSD analyses were performed using the average structures as initial references to study the stability of the free and coupled receptor. In the same way, a mobility analysis was carried out using RMSF calculations, as well as the analysis of the solvent-accessible surface area (SASA) of ACE2 and the degree of compaction of the complexes using the radius of gyration (Rg) throughout the simulation. All these analyses were performed using the CPPTRAJ program [35] found in the AMBER20 package [33].

2.5. Calculation of the binding free energy

The binding free energy of each ligand bound to the ACE2 receptor was calculated by the MM-GBSA method using the MMPBSA.py tool [36]. In total, 500 trajectory frames (200 ns) were extracted from the DM simulations. The binding free energy of each complex was calculated using the following equation:

ΔG = G_COMPLEX − (G_RECEPTOR + G_LIGAND)

Where G_COMPLEX indicates the free energy of the receptor-ligand complex, G_RECEPTOR (ACE2) and G_LIGAND (radotinib, hinokiflavone and ginkgetin) are the free energies of isolated protein and ligand in solvent, respectively, following the methodology proposed by Alviz-Amador et al. [34] calculating the free energy (ΔG), the entropy contribution of the protein was ignored, since here the binding energy was used to determine the relative binding strength of each complex.

2.6. Prediction of drug similarity, pharmacokinetic and toxicological parameters

Some drug similarity, pharmacokinetic, and toxicological parameters were determined for the three selected ligands. Drug similarity parameters such as molecular weight, mLogP, number of hydrogen acceptors and donors, Lipinski compliance, rotatable bonds, and bioavailability were determined using the SwissADME web server [37]. The parameters ADMET intestinal absorption, Pg-p substrate, unbound fraction, CYP450 family inhibitor, total clearance, maximum tolerated dose, and acute oral toxicity in rats (LD50) were estimated using the pkCSM web server [26].

3. Results

A total of 304 molecules were identified in the literature review, of which 174 met the eligibility criteria in the results of the prediction using SwissADME [37]. Of these, only 79 also met the eligibility criteria in the prediction results by pkCSM [26]. These 79 molecules were selected as ligands for the docking study, as shown in Fig. 1 .

Fig. 1.

Screening criteria for potential destabilizers of the RBD-ACE2 interaction.

3.1. Molecular docking

Table 1 shows the molecular docking scores of the 10 molecules with the best score obtained from the simulation conducted in Autodock Vina [32], along with the interactions analyzed in Discovery Studio Visualizer. The molecular docking was performed in triplicate. The average values and standard deviations were obtained for each molecule. The molecules with the best docking scores, that is, lower energy values, were radotinib, hinokiflavone, and ginkgetin, with scores of −10.2 ± 0.1, −9.8 ± 0.0, and −9.4 ± 0.0 kcal/mol, respectively. The other molecules listed in the table were morusin, silymarin, prunin, escularetin, bicuculline, puerarin, and nafamostat.

Table 1.

Docking scores and molecular interactions of selected ligands.

Ligands/Pubchem CID	Kcal/mol ±SD	Hydrogen bridge interactions	Hydrophobic and electrostatic interactions
Radotinib/16063245	−10.2 ± 0.1	Arg393, Gly395, Glu398	Phe40, Asp350, Leu351, His378, Phe390, Leu391, His401
Hinokiflavone/5281627	−9.8 ± 0.0	His345, His374, Asp382, His401	His374, His378, His401, Glu402, Glu406, Zn
Ginkgetin/5271805	−9.4 ± 0.1	Ala348, Asp350, His378	Phe40, Trp69, His378, Phe390, Arg393, His401
Morusin/5281671	−9.0 ± 0.1	Ser44, Asp350, Asp382	Phe40, Asp382, Phe390, His401
Silymarin/5213	−8.6 ± 0.1	Ala348, Arg393, Asn394	Phe40, Ala348, Asp350, His378, Phe390, Arg393
Prunin/92794	−8.4 ± 0.0	Ser47, Asp350, Asp382, Tyr385	Thr347, Trp349
Scalarane/45480583	−8.2 ± 0.0		Phe40, Trp349, Phe390
Bicuculline/10237	−8.2 ± 0.0	Asp350, His401	Trp349, His401
Puerarin/5281807	−8.2 ± 0.0	Ser44, Pro346, Asp350	His378, His401
Nafamostat/4413	−8.0 ± 0.0	Leu391, Arg393,	Phe40, Asp382, His401

The structure identified with PDB ID: 6M0J was used as the pharmacological target in the molecular docking. This structure consists of two protein subunits: chain A of the structure corresponds to the ACE2 receptor, and chain E corresponds to the RBD of SARS-CoV-2. The 10 ligands reported in Table 1 interacted with amino acid residues of chain A, and none of these ligands showed interaction with chain E. The identified amino acid residues were: Phe40, Ser44, Ser47, Trp69, His345, Pro346, Thr347, Ala348, Trp349, Asp350, Leu351, His374, His378, Asp382, Tyr385, Phe390, Leu391, Arg393, Asn394, Gly395, Glu398, His401, Glu402, and Glu406. The ASsp350 residue was observed more frequently in hydrogen bond interactions, and the His401 residue, in hydrophobic and electrostatic interactions. hinokiflavone interacted with the zinc molecule.

The 3D interaction of radotinib with the 6M0J molecule (RBD-ACE2) is shown in Fig. 2 A. Radotinib was the ligand with the best docking score and showed hydrogen bond interactions with the amino acid residues Arg393, Gly395, and Glu398; hydrophobic interactions with Phe40, His378, Phe390, A Leu391, and His401, and electrostatic interactions with Asp350 and Leu351. Fig. 2B shows the 3D of hinokiflavone in the interaction site with RBD-ACE2. hinokiflavone was the ligand with the second-best docking score and showed hydrogen bond interactions with the amino acid residues His345, His374, Asp382, and His401; hydrophobic interactions with His374, His378, and His401, and electrostatic interactions with Glu402, Glu406, and zinc. Fig. 2C shows the interaction of ginkgetin with RBD-ACE2. Ginkgetin was the third ligand with the best docking score and showed hydrogen bond interactions with the amino acid residues Ala348, Asp350, and His378; hydrophobic interactions with Phe40, Trp69, His378, Phe390, Arg393, and His401, and unfavourable interactions with Asp382.

Fig. 2.

3D representations of selected ligands at the site of interaction (A) Radotinib (B) Hinokiflavone (C) Ginkgetin.

The residue His401 showed hydrophobic interaction with all three candidate ligands. The residue His378 showed hydrophobic interaction with radotinib and hinokiflavone, and hydrogen bonding interaction with ginkgetin. The residues Phe40 and Phe390 interacted with radotinib and ginkgetin through hydrophobic bonds. The residue Asp382 showed hydrogen bonding interaction with hinokiflavone and unfavourable bonding with ginkgetin. The residues Asp350 and Arg393 interacted with radotinib and ginkgetin; however, the interaction of Asp350 with radotinib was through hydrogen bonding and with ginkgetin through hydrophobic bonding, and the interaction with Arg393 was through hydrophobic bonding with radotinib and through hydrogen bonding with ginkgetin. The Arg393 was the only interacting residue of the ligands with the RBD-ACE2 active site.

Table 2 describes the medicinal properties and potential effects against coronavirus disease 2019 (COVID-19) of the three ligands selected as candidates for destabilizing the interaction between the RBD of the spike protein of SARS-CoV-2 and human ACE2. The three studied molecules, radotinib, hinokiflavone, and ginkgetin, report antiviral activity in the literature and they all have studies with the SARS-CoV-2 3CL^PRO protease. Radotinib is the only synthetic ligand, and it was found that only this molecule has studies on genomic biomarkers; however, there are no reports on the spike protein. On the other hand, hinokiflavone and ginkgetin are bioflavonoids found in various plants. Both molecules have studies on the spike protein of SARS-CoV-2.

Table 2.

Candidate ligands along with their medicinal properties and possible effects against coronavirus disease 2019 (COVID-19) reported in the literature.

Ligands	Source	Medicinal properties	COVID-19 Studies
Radotinib	Synthetic	Anticancer, antiviral	3CLPRO Protease, Genomic Biomarkers
Hinokiflavone	Chamaecyparis obtusa Endlicher, Dacrydium balansae, Metasequoia glyptostroboides, Rhus succedanea, Rhus coriaria, Podocarpus elongatus y Garcinia multiflora	Antioxidant, anti-inflammatory, antiviral (dengue and influenza)	3CLPRO protease, Protein S (S2 subunit)
Ginkgetin	Ginkgo biloba, Torreya nucifera, Cephalotaxus drupacea,	Anticancer, antiinflammatory, antibacterial, antiviral, antifungal, antiadipogenic, antiplasmodial, antitrypanosomal, leishmanicidal, antiatherosclerotic, antioxidant, pancreatic lipase inhibitor, and neuroprotective	3CLPRO protease, Spike Protein (RBD)

3.2. Molecular dynamics simulation

Molecular dynamics simulations were performed on complexes of the ACE2 receptor and the selected ligands: radotinib, hinokiflavone, and ginkgetin. Fig. 3 shows the results of root mean square deviation (RMSD), root mean square fluctuation (RMSF), solvent-accessible surface area (SASA), and radius of gyration obtained, respectively. The analysis of the RMSD trajectory of the carbon skeleton showed that the native ACE2 receptor had RMSD values between 1.0. and 2.0 Å and that the hinokiflavone molecule was more stable with RMSD values between 1.0 and 2.0 Å, showing few differences in the trajectory with the native ACE2 receptor in contrast to radotinib and ginkgetin, which presented values of up to 3.0 Å, with the highest peaks of radotinib at 110 ns and ginkgetin at 15 ns. This indicates that hinokiflavone was the molecule that generated fewer conformational changes in its interaction with the ACE2 receptor compared to the other two molecules (Fig. 3a). The flexibility of residues was evaluated by the RMSF of the carbon skeleton atoms. Fig. 3b shows that the fluctuation trends of individual residues with the three ligands and the native ACE2 receptor were approximately the same. The regions of highest fluctuation were observed in residues 130 and 320 with values close to 4.5 and 4.0 Å, respectively.

Fig. 3.

Molecular dynamics of ACE2 native, Radotinib, Hinokiflavone and Ginkgetin complexes. (A) RMSD, (B) RMSF, (C) SASA, and (D) Rg analysis of all trajectories.

In Fig. 3c, the SASA of the selected ligands is shown. The three ligands had values ranging between 25000 and 29000 Å2 following a central and similar trend among the molecules. ginkgetin presented the highest values, suggesting greater expansibility, while hinokiflavone had the lowest values, indicating greater contraction. All three molecules, over the 200 ns evaluated, tended to have a lower SASA value than the native ACE2 receptor.

Regarding the radius of gyration, Fig. 3d shows that radotinib was the only molecule that had values of radius of gyration above the native ACE2 receptor, with higher values compared to the other molecules, suggesting that it has a lower degree of compactness and rigidity. On the other hand, hinokiflavone reported the lowest values, indicating that it presented more rigid conformations during the simulations.

Binding free energy calculation of the ACE2 receptor and selected ligands.

The total binding energy was calculated in terms of solvation energy, gas phase energy, and entropy contributions (G). To determine the contribution of binding site residues, the systems were decomposed. The contribution of the energy components of Van der Waals (VDWAALS), electrostatic component of internal energy (EEL), polar component of solvation energy (EGB), non-polar component of solvation energy (ESURF), Delta G gas, Delta G solvent, and the total energy of each complex are shown in Table 3 . The most important contributions to the binding free energy are the Van der Waals and electrostatic interactions for radotinib and hinokiflavone and the Van der Waals and polar interactions for ginkgetin. The receptor complex with hinokiflavone showed the best binding free energy interaction with −215.86 kcal/mol, followed by ginkgetin −149.10 kcal/mol and finally radotinib with 24.31 kcal/mol.

Table 3.

Binding free energy components (Kcal/mol) of ACE2 receptor complexes and selected ligands using the GBSA method.

Energy component	Radotinib	Hinokiflavone	Ginkgetin
VDWAALS	−46.80	−162.33	−184.71
EEL	−1167.96	−951.24	69.4474
EGB	1197.89	854.24	−79.25
ESURF	−7.43	−22.65	−22.45
ΔG gas	−1214.76	−1.047.45	−47.40
ΔG solv	1190.45	831.59	−101.70
ΔG TOTAL	−24.30	−215.86	−149.10

3.3. Prediction of drug similarity, pharmacokinetic and toxicological parameters

The drug similarity and ADMET properties of the three candidate ligands were simulated using the SwissADME and PkCSM web servers, and this information was condensed in Table 4 . It is clear that all selected compounds are well absorbed, permeable and have pharmacological properties; of the three characterized molecules, only one of Lipinski's rules is violated; by having a molecular weight greater than 500 g/mol. Radotinib presented a higher number of rotatable bonds with eight, followed by ginkgetin with five and then hinokiflavone with four. Additionally, it is observed that all selected compounds have a calculated bioavailability score of 0.55.

Tabla 4.

Prediction of druglikeness and ADMET properties with SwissADME and PkCSM.

Properties	Radotinib	Hinokiflavone	Ginkgetin
Molecular Weight (g/mol)	530.50	538.46	566.51
mLogP	1.78	0.52	0.63
H-bond acceptors	9	10	10
H-bond donors	2	5	4
Lipinski's violations	1	1	1
Num. rotatable bonds	8	4	5
Bioavailability Score	0.55	0.55	0.55
Intestinal absorption (%)	95.33	79.91	95.38
P-glycoprotein substrate	Si	Si	No
Fraction unbound	0.26	0.23	0.27
CYP1A2 inhibitor	No	No	No
CYP2C19 inhibitor	Yes	No	No
CYP2C9 inhibitor	Yes	No	Yes
CYP2D6 inhibitor	No	No	No
CYP3A4 inhibitor	Yes	No	No
Total Clearance (ml/min/kg)	0.34	0.51	0.65
Max. tolerated dose (human) (log mg/kg/día)	0.24	0.44	0.43
Oral Rat Acute Toxicity (LD50) (mol/kg)	2.49	2.65	2.73

On the other hand, Table 4 shows that all selected compounds have good values of human intestinal absorption, ranging from 79.91% for radotinib to 95.38% for ginkgetin. Only ginkgetin is not a substrate for P-glycoprotein. The unbound fraction (free fraction) of the three ligands is quite similar, all below 0.3. Regarding the results of inhibition of the enzymes CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4, hinokiflavone does not inhibit any of the enzymes, radotinib inhibits the enzymes CYP2C9, CYP2C19, and CYP3A4, and ginkgetin inhibits only CYP2C9. The three candidate ligands presented favourable renal clearance. Radotinib has the lowest values of maximum tolerated dose in humans and acute oral toxicity in rats (LD₅₀), so the ligands hinokiflavone and ginkgetin could be used in higher doses than radotinib.

4. Discussion

79 molecules that met the eligibility criteria of the ligand repository for molecular docking studies were identified. All selected molecules met Lipinski's rules and have a synthetic accessibility value ≤ five according to the SwissADME prediction result, and none had AMES toxicity or hepatotoxicity according to the prediction made in pkCSM.

Of the 79 molecules selected as ligands in the molecular docking study, radotinib, hinokiflavone, and ginkgetin exhibited the best docking scores with values of −10.2 ± 0.1, −9.8 ± 0.0, and −9.4 ± 0.0 kcal/mol, respectively. All interactions in the three candidate molecules were presented with ACE2, and interaction with the active site RBD-ACE2 was observed only in ARG393 with radotinib and ginkgetin.

Celik et al. [38] evaluated the molecular docking of hydroxychloroquine (HCQ) and chloroquine (CQ) in the SARS-CoV-2 spike protein-ACE2 complex (PDB code: 6LZG) obtaining values of −9.178 and −8.488 kcal/mol, respectively. These results corroborate in vitro findings that suggest HCQ exhibits stronger anti-SARS-CoV-2 activity compared to CQ [39]. This interaction did not occur at the active site of the SARS-CoV-2 spike protein-ACE2 complex; however, it was established at an allosteric site of ACE2, causing inhibition of the spike protein's binding to ACE2. HCQ appears to prevent the binding of the SARS-CoV-2 spike protein by interacting with the amino acids PHE40, ALA348, ASP350, ASP382, PHE390 [38]. In our study, radotinib and ginkgetin showed interaction with PHE40, ASP350, and PHE39, and hinokiflavone with ASP382, suggesting that radotinib and ginkgetin may have a higher affinity for the allosteric site than the RBD-ACE2 active site, and hinokiflavone would only bind to the allosteric site. The three selected ligands interacted at the allosteric site, obtaining better docking scores than HCQ. Therefore, the results obtained in our study indicate that radotinib, hinokiflavone, and ginkgetin could destabilize the binding of the spike protein with ACE2 by interacting at an allosteric site of ACE2, presenting docking scores denoting strong receptor affinity.

On the other hand, it was evidenced that all three molecules have been attributed with antiviral activity and all have studies on COVID-19. Two studies reported that radotinib can be used for the prevention, relief, or treatment of viral respiratory diseases [40,41]. Novak et al. [42] established radotinib as a potential allosteric inhibitor of the main protease of SARS-CoV-2. Two studies of radotinib with COVID-19 genomic biomarkers [43,44] suggest that this molecule could play a vital role in the treatment against different variants of SARS-CoV-2 infections. Additionally, Radotinib is an oral multi-targeted inhibitor of receptor tyrosine kinases (RTKs) that by causing inhibition of the Bruton's tyrosine kinase (BTK) pathway could reduce the excessive and harmful immune response in the severe form of COVID-19 and the resulting respiratory complications [45,46].

In the case of hinokiflavone, antiviral activity has been reported against dengue and influenza [47], and it has been proposed as an inhibitor of the main protease of SARS-CoV-2 in three studies [[48], [49], [50]]. On the other hand, Mondal et al. proposed in their study that hinokiflavone has inhibitory activity against the invasion of SARS-CoV-2 spike protein into target cells by membrane fusion. Additionally, as an antioxidant, it promotes improvement in coronavirus-induced complications in mice. This protective effect is thought to occur through the reduction of oxidative stress, cerebral lipid peroxidation, and regulation of inflammation [51].

Ginkgetin, is a natural biflavone that has shown activity against herpes simplex virus type one and type two and activity against influenza virus [47,52]. Like radotinib and hinokiflavone, it has also been proposed as a candidate for inhibiting the main protease of SARS-CoV-2 [[53], [54], [55], [56]]. Xiong et al. [57] demonstrated that gingketin isolated from Ginkgo biloba leaf extract (GBLE) exhibits relatively strong SARS-CoV-2 3CL^PRO enzyme proinhibitory activity (IC50 < 10 μM), with a mean inhibitory concentration of 2.98 ± 0.86 μM. Ginkgetin is the only one of the three compounds to have a study in the RBD-ACE2 complex, Patel et al. [58], like in our study, they selected the 6M0J structure, however, for molecular docking and MD, ACE2 was removed, in contrast to our study, where only RBD was removed in the MD simulation. In Patel et al. [58] study, ginkgetin was the second-best compound based on binding energy, with a binding energy of −7.848 kcal/mol. In our study, it showed greater affinity with an energy of 9.4 ± 0.0 kcal/mol, which could suggest that this molecule is more affine with ACE2 than with RBD.

Regarding the results of the molecular dynamics obtained in our study, the RMSD values indicate that the molecule with the highest conformational stability in its interaction with the ACE2 receptor was hinokiflavone, reporting values that indicate a folding state in contrast to radotinib and ginkgetin, RMSD values indicate an intermediate folding state according to the criteria set by Deng et al. [59]. The flexibility of the ACE2 residues was similar in the complexes of the 3 molecules; no significant changes in mobility were observed in the selected ligands compared to the native ACE2 receptor. In relation to the SASA analysis, a higher SASA value denotes protein expansion; and according to what was evidenced, the tendency of the complexes of the ACE2 receptor with the 3 molecules evaluated was lower SASA values than those evidenced in the native ACE2 receptor, therefore, the interaction of the receptor with the ligands tends to contraction in nature [60]. Similarly, the protein's dimensional calculation performed by the radius of gyration indicated that the hinokiflavone complex was less compact than the other complexes during the simulation, and the radotinib complex was the only one to present radius of gyration values above what was reported by the native ACE2 receptor in the intervals 0–20 ns and 85–110 ns. This parameter is a parameter that describes the reorganization of the structures in the unfolded region, confirming that hinokiflavone formed the most rigid conformation during the simulation, and radotinib the most flexible conformation.

Additionally, in this study, the binding free energy of the ACE2 receptor and selected ligands was calculated. All three complexes reported negative entropies, indicating the favourability of the formation of each ligand complex with the receptor. The complex with hinokiflavone showed the best interaction among the three molecules, with an energy of −215.86 kcal/mol. In contrast with radotinib, which showed the worst values for the binding free energy, given that, compared to the hinokiflavone complex, radotinib obtained a value 9 times less favourable.

On the other hand, regarding drug-likeness properties, the three molecules only violated the Lipinski rule due to their molecular weight, fulfilling the main requirement of a potential drug compound based on their different molecular properties [28]. Radotinib was the ligand with the highest number of rotatable bonds, making it the most flexible compound, an important property for determining its oral bioavailability [61]. This property was confirmed in the molecular dynamics results, as radotinib was the ligand with the most flexible conformations during the simulation. The simulated bioavailability for the candidates in SwissADME was 0.55. This calculation is predicted based on the probability value that a molecule possesses an optimal profile of permeability and bioavailability, where 0.55 designates the compliance with the Lipinski rule, and the average bioavailability value is 55%, which is a probabilistic value greater than 10% [37,62].

In general, the three compounds showed a good ADMET profile, with appropriate values of human intestinal absorption. The presented values exceed the recommended minimum of 30%. Ginkgetin is the only molecule that is not a substrate for P-glycoprotein, Therefore, would be the ligand with the highest permeability. The free drug fraction of all three candidates is quite similar, ranging from 0.23 to 0.27, and this fraction is responsible for the drugs' biological effects. Regarding metabolism, hinokiflavone does not inhibit any hepatic enzymes evaluated, and radotinib is the only inhibitor of the CYP3A4 enzyme, which is responsible for half of the metabolic clearance of marketed drugs, so caution should be taken when using it [63]. All three molecules showed a positive renal clearance value, with ginkgetin having the highest rate. Hinokiflavone and ginkgetin bioflavonoids present higher values of maximum tolerated dose in humans and acute oral toxicity in rats (LD₅₀) compared to radotinib, indicating that radotinib should be used in lower doses than bioflavonoids to avoid toxic effects on the body. Radotinib shares the recently reported cardiovascular toxicity of Nilotinib [64], so its use should be monitored clinically.

Based on the results of molecular docking, molecular dynamics, drug similarity properties, pharmacokinetic and toxicological characteristics, and the literature, the ligands radotinib, hinokiflavone, and ginkgetin could have an antiviral effect by destabilizing the interaction between the RBD of the SARS-CoV-2 protein and human ACE2, and are candidates to be used in the treatment of COVID-19. However, it is important that in the future, the results obtained in the present study will be contrasted with in vitro/in vivo experimental studies to explore the possible preclinical and clinical efficacy of these compounds and to verify the mechanism of destabilization of the ACE2-RBD binding of the 3 ligands proposed in this research.

5. Conclusions

The study identified 304 potential antiviral molecules for the treatment of COVID-19, and 79 of them were evaluated for molecular docking with the pharmacological target. Radotinib, hinokiflavone, and ginkgetin showed the best molecular docking scores, indicating a strong affinity to the receptor. The compounds could cause destabilization of the S protein-ACE2 binding by interacting with an allosteric site of ACE2. The selected compounds also showed satisfactory physicochemical and pharmacokinetic ADMET properties, making them good candidates for use in COVID-19.

The results of the molecular dynamics simulations indicated that all 3 molecules exhibit good conformational stability in their binding to the ACE2 receptor and the free energy calculations observed in the study demonstrate the favorability of complex formation of each ligand with the receptor.

Finally, the importance of this study in the field of COVID-19 is the recognition of 3 molecules that could act as ligands by binding to the ACE2 receptor and destabilizing the ACE2-RBD binding, especially radotinib, which has been postulated to be useful in several variants of SARS-COV-2. Additionally, these molecules, due to their pharmacological properties, could contribute to the modulation of the inflammatory cascade caused by COVID-19.

Funding

The present work was supported by the call of resolution 00416 of 2021 of the University of Cartagena. Which orders the opening of the process of strengthening plans to obtain financial resources to support the strengthening and sustainability of research groups.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.