The crystal structures of the SARS-CoV-2 selected enzymes were downloaded from the Protein databank at https://www.rcsb.org. For M^pro enzyme (PDB: 5RH4, 1.34A^o), the structure was identified by X-ray diffraction as the crystal structure of SARS-CoV-2 main protease in complex with Z1530425063³; for Papain-like protease (PDB: 6wx4, the structure was determined by X-ray diffraction as Crystal structure of the SARS CoV-2 papain-like protease in complex with peptide inhibitor VIR251,1.66 Å (Rut et al. 2020)) and RNA dependent RNA polymerase (nsp 12, PDB: 7bv2 as The nsp12-nsp7-nsp8 complex bound to the template-primer RNA and triphosphate form of Remdesivir; 2.5 Å (Yin et al. 2020)) . All proteins were prepared by removing water molecules, any additional RNA or non-specific protein structures, followed by protonation and automatic correction to check for any errors in the atom’s connection and type using molecule preparation tool found in MOE software. Potential and charges were fixed, and dummy atoms were added instead of legend atoms.

The active site of the used enzyme was prepared, hydrogen atoms were added, charges were fixed, dummy atoms were introduced in ligand position, compounds were docked, and possible interactions with amino acid residues were computed within the active binding site. Poses were studied and selected according to the best energy scores and binding interactions observed.

The 3D structure of the selected coumarin compounds, N3, chloroquine and Remdesivir, were built using a builder interface, energy was minimised to an RMSD gradient of 0.01 kcal/ mol and 0.1 A^o. All compounds were added to a database and saved as an mdb file.

The selected coumarins; aminocoumarin antibiotics 1–3 (Fig. 1), coumarin drugs 4–13 (Fig. 2), and natural coumarins with reported antiviral activity 14–37 (Fig. 3) were docked into the active site of SARS-CoV-2 main protease (PDB: 5RH4, 1.34Å) using Molecular Orbital Environment (MOE, 2019) software. The observed binding interactions alongside energy scores for the obtained enzyme compound complexes are listed in Tables 1 and 2.

All data obtained compared to a potent standard M^pro inhibitor called N3, Fig. 4. N3 was redocked into the active site of M^pro, and the observed interactions were similar to that reported implying the validity of the used method (Baz and Boivin 2019; Jin et al. 2020). Validation of the methodology used was tested via redocking of hydroxychloroquine, a reported inhibitor for M^pro. The energy score recorded was -6.78 kcal/mol compared to a reported -6.9 kcal/mol (Baildya et al. 2020). The nearby residues also were similar to that reported (Fig. 4b). The used method showed a potential formation of hydrogen bonding with THR 190, CYS 44, and hydrophobic interactions with GLU 166 and GLN 189 with proximity THR 25, GLN 192, ASN 142, and MET 165. THR 25, MET 49, PHE 140, LEU 141, ASN 142, GLY 143, SER 144, CYS 145, MET 165, HIS 164, GLU 166, GLN 189 were reported residues of proximity (Baildya et al. 2020). The docking was also validated by following up with the MD simulation study mentioned later.

Figure 4. 

(A) Structure of standard inhibitor of SARS-CoV-2 main protease, N3, (B) 2D pose for the interaction of hydroxychloroquine into the active site of SARS-Cov-2 main protease enzyme.

Data showed that most of the tested drugs could form stable complexes with the active site of SARS-CoV-2 main protease with energy scores ranging from -9.30 to -4.62 compared to -8.5 for N3, the standard ligand for M^pro enzyme. As in Umbelliferone, 4, a primary coumarin nucleus fits into the receptor with potential binding with Methionine 165 and Glycine 189, 192 residues found in the active site actively pointing at the potential of coumarins to bind to M^pro. More stable complexes were observed with extended coumarins such as the NMDA receptor antagonist, Ensaculin, 13 with an energy score of -7.42, and potential binding with the essential cysteine residue CYS 145. Similarly, Acenocoumarol, 11, and Carbocromen, 9, also showed relatively stable complexes with energy scores of -7.1 and -7.01 and showing interactions with MET 165 and CYS 145, respectively, Table 1. The most stable complexes were observed with larger structures of aminocoumarin antibiotics, suggesting a possible high activity against SARS-CoV-2. Three aminocoumarin antibiotics were examined: Novobiocin, Clorobiocin, and Coumermycin. They fit into the N3 binding pocket and show stable complexes with energy scores of -8.17 for Novobiocin, -8.69 for Clorobiocin, and the most stable complex was observed with Coumermycin showing higher stability (-9.30) even than that observed with N3. They all showed potential binding interaction with the CYS 145 residue in M^pro active site; Table 1, Figures 5 and 6. Coumermycin also showed additional hydrogen bonding with THR 24, GLU 166, and one hydrophobic interaction with GLY 143 residues found in the active site, Fig. 6. Though Novobiocin showed more interactions (4 hydrogen bonds and one hydrophobic interaction, Table 1), the energy score suggests a more stable complex with Coumermycin. These data suggest the relative importance of the molecular size of the coumarin derivative in such settings. Surface mapping of M^pro active site reflected stable areas where bonding is formed with Coumermycin,3, Fig. 6A.

Figure 5. 

2D poses of (A) Novobiocin; (C) Clorobiocin; (B) hydroxychloroquine; (D) N3; docked into the active site of SARS-CoV-2 main protease.

Figure 6. 

(A) surface map (B) 2D poses showing ligand interactions of Coumermycin docked into the active site of SARS-CoV-2 main protease (PDB: 5rh4).

Previous reports suggested an antiviral activity of coumarin antibiotics. Novobiocin suppresses the replication of cytomegalovirus (Sekiguchi and Shuman 1997) and is used successfully in experiments to inhibit Zika virus with an EC₅₀ of 25 µg/ml (Baz and Boivin 2019). It also blocks vaccinia viral assembly and morphogenesis and shows excellent activity against the herpes virus (Palu et al. 1986)., Ferrazzi et al. reported the antiviral activity of the Coumermycin, which is a carbohydrate with two coumarin groups, showing that Coumermycin inhibits the DNA polymerase activity resulting in suppressing the replication of herpes simplex virus type 1 (Palu et al. 1986; Ferrazzi et al. 1988). Reports also mentioned its ability to inhibit murine retrovirus replication (Varnier et al. 1984). Later evidence supported its potential use against HIV though the mechanism was not understood at that time (G. Tachedjian 1990). It was also used to control the African Swine Fever Virus (ASFV) (Coelho and Leitao 2020). Current docking results go following these previous data suggesting the possibility of using these drugs in the fight against COVID-19 worldwide struggle. Moreover, the high safety profile of Coumermycin also encourages taking these studies for further steps into in vitro testing to prove the anti-SARS-CoV-2 activity.

Meanwhile, docking naturally found coumarins revealed a formation of less stable complexes (energy scores of -4.79 to -7.61; Table 2). 4-Phenyl coumarins isolated from Marila pluricostata; Mesuol; 18 and Isomesuol; 19 reported earlier for their anti-HIV activity; showed the highest potential activity against M^pro. Mesuol 18 formed the most stable complex, possibly binding with CYS 145, GLU 166, and GLN 189; Fig. 7A. Switching the two-side chain into Isomesuol decreased the complex stability, and the number of observed interactions decreased with the disappearance of the CYS 145 interaction, Fig. 7B. The four tested Calanolides 20–23 also form relatively stable complexes, mainly binding with CYS 145 or GLY 143. Glycyrin 15 and Baykangelicin 33 also showed relatively stable complex with the binding pocket with potential binding with ARG 188, GLU 166, GLU 166 for Glycyrin and ASN 142, CYS 145, HIS 163, GLY 143 for Baykangelicin. Smaller sized coumarins gave far less stable complexes with the order of compound potential activity is 24, 25, 26, 37, 28, 32, 30, 34, 31, 27, 29, 35, 36, 14, 22, 16, 21, 33, 15, 20, 23, 17, 19 and 18, Table 2.

Figure 7. 

2D poses of (A) Mesuol; (B) Isomesuol; (C) suksdorphin; (D) Calanolide; docked into the active site of SARS-CoV-2 main protease (PDB: 5rh4).

The studied compounds were docked against the active site of papain-like protease (PL^pro) (PDB:6wx4). Docking against the active site of papain-like protease resulted in the formation of relatively stable complexes. The highest Amino coumarin antibiotics 1–3 and ensaculin 13 were the most potentially active inhibitors of the enzyme. They showed highly stable complexes with energy scores of -7.19, -7.54, -9.42 and -7.29 kcal/mol, respectively. Coumermycin showed the highest energy score with the highest number of possible interactions, including hydrogen bonds with ASP 164 and LYS 157 and various hydrophobic interactions with HIS 89, TYR 264, ASP 108, LEU 162, and TYR 268, Table 3, Fig. 8. Other coumarin drugs demonstrated less stable complexes with energy scores of -4.38 to -6.64 kcal8/mol, Table 3. A similar docking was obtained for a virtual docking study for novobiocin against the active site of papain-like protease (Mitra et al. 2020), supporting the potential role of coumarin antibiotic hypothesised in the current study. An independent report also suggested using Coumermycin A1 to treat COVID-19 via blocking SARS-CoV-2 papain-like protease (PL^pro). They reported the ability of coumermycin A1 to bind to PL^pro with an energy binding score of -12 kcal/mol(ARULANANDAM 2020) according to data obtained in the current research.

Figure 8. 

(A) surface map (B) 2D poses showing ligand interactions of Coumermycin; docked into the active site of SARS-CoV-2 papain-like protease (PDB: 6wx4).

Figure 9. 

(A) surface map (B) 2D poses showing ligand interactions of Coumermycin; docked into the active site of SARS-CoV-2 RNA-dependent RNA polymerase (PDB: 7bv2).

Naturally occurring coumarins studied also yielded relatively stable complexes with the active site of PL^pro. Mesoul 18 and Isomesoul 19 formed the most stable complexes formed with energy scores of 6.12 -6.05 kcal/mol, Table 3. Collectively the ability of the two natural products to inhibit the two proteases strongly supports their potential use in the fight against SARS-Cov-2.

The studied compounds were also docked against the active site of RNA-Dependent RNA (RdRP) (PDB:7bv2). The docking methodology was validated via redocking the co-crystallised RdRp inhibitor Remedisivir with data obtained similar to that reported for Remedisivir (Rut et al. 2020). The results showed that compounds 1–13, 18, 19 and 33 demonstrated the most stable complexes and were reported in Table 4. Results showed that coumermycin 3 successfully fit into the Remedesivir binding pocket, forming hydrogen bonds with ASP 760 and neighbouring ARG 553 or ARG 555 residues. An -10.33 kcal/mol energy score implied a highly stable complex formation between compound 3 and amino acid residues present in the RdRP active site. The stable complex formed lacked the metal interactions formed by Remedisivir, Table 4, (Suppl. material 1). Unfortunately, the other coumarin-carrying drugs did not show similar stable complexes with energy scores ranging from -4.8 to -6.8 kcal/mol compared to -9.3 for Remedisivir. Natural products 14–37 also demonstrated an excellent fit to the same active site with energy scores of -5.2 to -8.1 kcal/mol. The compounds were most stable. Complexes formed were reported in Table 4 for compounds 33>19>18. Compound 33 showed a hydrogen bond formed with ARG 553, metal interactions with Mg ion ana hydrophobic interaction with ARG 555 similar to that observed with remedesivir. Mesuol 18 and Isomesoul 19 consistently demonstrated reliable interactions against this third protein, Table 4, reflecting their potential use in the fight against COVID-19.

Collectively, the literature supported the obtained results; virtual docking suggested an anti-protease role against SARS-Cov-2 M^pro for coumarins found in Salvadora persica (Ferrazzi et al. 1988). Flavonoids isolated can bind successfully to the N3 binding site with the ability to bind to CYS 145, as observed in the current study. Binding energies were similar to those surveyed here, ranging from -7.4 to -8 kcal/mol (Bhuiyan et al. 2020). Additionally, 17 coumarin derivatives demonstrated stable interaction against RdRp with binding affinities of less than -10 and regular interactions were made with ASP 623, THR 556, LYS 621, and Pro 620 residues (Ozdemir et al. 2020). Other natural and synthetic coumarins were studied and showed similar capabilities in inhibiting SARS-Cov-2 main protease. They were theoretically bound to GLY 143 and GLN 198 with hydrogen bonds while forming hydrophobic interactions with ASN 142, CYS 145, HIS 164 and MET 165 (Chidambaram et al. 2020). The same set of amino acid residues was embodied in the binding poses obtained in the current study. On the experimental level, a coumarin derivative isolated from the Marine Sponge Axinella cf. corrugata inhibited M^pro of the earlier version of SARS viral, inhibiting its replication (de Lira et al. 2007). In addition, another in silico study showed that novobiocin has a potential inhibition against RdRp (Wu et al. 2020a), and Coumarins isolated from A. keiskei showed inhibitory effects against the SARS-CoV M^pro and PL^pro (Park et al. 2016).

The configuration and vigorous properties of protein/ligand complexes during the simulation time of 50 ns were studied as the backbone RMSDs. The RMSD was calculated as the mean distance of complexes between atoms presents in the spine and is obtained from the following equation (Liu and Kokubo 2017).

$\underset{¯}{\mathrm{RMSD}}=\frac{1}{N}\sum _{i=0}^N\delta \frac{2}{i=0}$ 1).

In equation (Al-Khafaji et al.), the N is the complete no. of atoms present in an equation, and d indicates the subsequent distance of particles between the N pairs. The backbones RMSD of the three complexes are shown in Fig. 10.

Figure 10. 

The RMSD plot of Coumermycin complex with SARS-CoV-2 (A) Mpro (B) PLpro (C) RdRp, at 50 ns simulation.

The RMSD of the complex of Coumermycin/M^pro detected a minimal deviation at 0.05 nm from 5 to 35 ns. After 35ns, the complex showed stabilisation during the 50 ns MD simulation.

Similarly, the RMSD of complex Coumermycin/PL^pro represented a minor deviation at 0.05 nm from 1 to 5 ns, and it stabilised throughout the 50 ns of simulation. The RMSD of complex Coumermycin/RdRp was steady in 50 ns simulation from 0 to 20 ns, with a slight deviation of 0.015 nm observed throughout the simulation. While overall, the simulation was kept stable throughout the 50 ns simulation. Initially, the RMSD values increased steadily and stayed converged over the simulation time for the three complexes studied (Fig. 10). It is worth mentioning; all three complexes had RMSD descriptors that did not exceed 2.5 Å, which ascertains the rigid conformation of the formed complexes.

The RMSF calculated the flexibility of common protein and showed an unpaid parameter to examine residual protein’s flexibility over the simulation era (Alamri et al. 2020; Chinnasamy et al. 2020). The RMSF of complexes solvent-accessible surface area were observed between the range of 0.5 ± 1.8 Å, 0.6 ± 2.4 Å and 0.5 ± 1.5 Å. The RMSF plot of the complex with both proteases showed minor variations in both complexes. Simultaneously, the RdRp complex showed a higher fluctuation at the C and N terminals of protein (Fig. 11). During the simulation α, the β elements of the secondary structure make the protein structure more rigid.

Figure 11. 

The RMSF plot of Coumermycin complex with SARS-CoV-2 (A) Mpro (B) PLpro (C) RdRp, at 50 ns simulation.

The radius of gyration (Rg) is a framework for assessing the biological molecule’s nature and stability during MD time by calculating the macromolecule’s structures (L and Soliman 2016). The Rg values show the MRSD of an atom distance from the common centre of mass. The Rg may also be used to measure whether the complex can stay folded during MD simulation. The Rg values through the simulation at 50 ns of the Coumermycin/M^pro, PL^pro and RdRp complexes were 0.96 nm ± 1.12 nm, 0.96 nm ± 1.12 nm, 1.05 nm ± 1.14 nm, respectively, that confirming the structures had entered a steady-state. The average Rg values of the three complexes stayed reasonably constant for the 50 ns, suggesting a robust folded arrangement. (Fig. 12).

Figure 12. 

The time frame of evolution against the radius of gyration (Rg) of Coumermycin complexes with SARS-CoV-2 (A) Mpro (B) PLpro (C) RdRp, during 50 ns MD simulation.

SASA or solvent-accessible surface area is a method that is used to calculate the water-accessible area of macromolecules (6). Monitoring the SASA value is a crucial method to estimate the conformational changes that result from dynamic interactions. The estimated average range of SASA values of three complexes solvent-accessible surface area for 50 ns simulation was between the 5 ± 10 nm², 5 ± 10 nm^2, 4.5 ± 6 nm², respectively. These findings indicated no improvements were found in all three systems’ usability regions during 50 ns simulation time. Consequently, the relative constancy of our protein/ligand complexes has been derived from the SASA analysis (Fig. 13). It is worth mentioning that the SASA profile for the RdRp complex was lower than that observed with the other 2 proteins.

Figure 13. 

The time frame of evolution against SASA of Coumermycin complexes with SARS-CoV-2 (A) Mpro (B) PLpro (C) RdRp, during 50 ns simulation.

Atomic-level knowledge is crucial to forecast the binding pocket of Coumermycin to the target protein’s binding site and to validate docking data obtained earlier. The different intermolecular interactions such as hydrogen bonds, water bridges, hydrophobic and ionic interactions were investigated over 50 ns of MD simulation studies for critical mode evaluation. The study stated that the Coumermycin complex with SARS-CoV-2 M^pro made strong hydrogen bonding with the amino acid’s residues THR 24, THR 25, THR 26, ASN 142, GLU 166, and GLN 189. It also formed strong water bridges interaction with amino acid residues CYC 22, CYS 24, THR 26, HIS 41, ASN 142, HIS 163, GLU 166, and GLN 189. The amino acid’s residues GLU 47, HIS 164, GLU 166, and LEU 167 showed the hydrophobic interaction (Fig. 14A, Suppl. material 1). Similarly, the report stated that the Coumermycin/PL^pro showed the hydrogen bonding with amino acid’s residues ASN-88, HIS 89, TRP 106, ASP 108, VAL 159, GLU 161, TYR 268, and CYS 270. The Coumermycin/PL^pro formed ionic interaction interactions with the amino acid’s VAL 159 and GLY 160. Further, the amino acid’s residues ALN 39, LYS 105, ASN 109, and ASN 267 were also implied in hydrophobic interaction with Coumermycin/PL^pro (Fig. 14B, Suppl. material 1). The study showed that the Coumermycin/RdRp demonstrated strong hydrogen bonding with the amino acid’s residues VAL 495, AR 553, TRP 617, LYS 621, CYS 622, and HIS 810. Along with this, Coumermycin/RdRp formed strong ionic interaction with amino acid residues TRP 617, ASP 618, and SER 759. The amino acid’s residues ALA 550, THR 565, and CYS 799 were involved in hydrophobic interaction (Fig. 14C, Suppl. material 1).

Figure 14. 

Protein interaction analysis. The green colour = hydrogen bonding, pink color = ionic interaction, grey colour = hydrophobic interaction and blue colour = water bridges showed in Coumermycin complexes with SARS-CoV-2 (A) Mpro (B) PLpro (C) RdRp during 50 ns MD simulations.