3.1 Docking between ligands and monomeric SARS-CoV–2 Mpro
Docking studies between ligands and SARS-CoV–2 Mpro showed that all ligands: indomethacin (Fig. S1A), naftazone (Fig. S1B), ofloxacin (Fig. S1C), zopiclone (Fig. S1D), sofosbuvir (Fig. S1E), pitavastatin (Fig. S1F), eszopiclone (Fig. S2A), perampanel (Fig. S2B), fenoterol (Fig. S2C), azelastine (Fig. S2D), celecoxib (Fig. S2E), nelfinavir (Fig. S2F), praziquantel (Fig. S3A), ondansetron (Fig. S3B), and lemborexant (Fig. S3C) reached the catalytic binding site of SARS-CoV–2 Mpro (Supplementary material, Figs. S1-S3). These ligands were mostly stabilized by H41, F140, N142, C145, H163, H164, M165, E166, Q189 and R188 residues through nonpolar interactions. H41, S46, Y54, F140, L141, N142, G143, S144, C145, H163, H164, E166 and D187 established polar interactions through backbone or side chain atoms with some of the compounds: indometachin (Fig. 1A), naftazone (Fig. S1B), ofloxacin (Fig. S1C), zopiclone (Fig. S1D), sofosbuvir (Fig. S1E), pitavastatin (Fig. S1F), perampanel (Fig. S2B), fenoterol (Fig. S2C), azelastine (Fig. S2D), praziquantel (Fig. S3A), ondansetron (Fig. S3B), lemborexant (Fig. S3C), and ritonavir (Fig. S3E). The residues stabilizing the ligands were mostly distributed between domains 1 (residues 8–101) and 2 (residues 102–184), and the interactions established were similar to those observed in the co-crystallized complex between the SARS-CoV–2 Mpro ligand and the inhibitor N3 (PDB ID: 6LU7), highlighting the interactions with conserved residues (H41 and C145) involved in the catalytic activity of the enzyme [23].
3.2 Docking of lopinavir or ritonavir with monomeric SARS-CoV–2 Mpro and SARS-CoV Mpro
Docking studies show that lopinavir and ritonavir on SARS-CoV–2 Mpro and SARS-CoV Mpro reached the catalytic site of both systems (Supplementary material, Figs. S3D-G). On SARS-CoV–2, lopinavir (Fig. S3D) and ritonavir (Fig. S3E) were mostly stabilized by T25, T26, H41, F140, L141, N142, G143, H163, E166, D187, Q189 and R188 residues through nonpolar interactions, whereas ritonavir established polar interactions with the side chain of S46 (Fig. S3E). On SARS-CoV Mpro, lopinavir (Fig. S3F) and ritonavir (Fig. S3G) are mostly stabilized by T25, A46, M49, L141, S144, E166 and Q189 through nonpolar interactions, while ritonavir formed polar interactions with the side chain of Q189 (Fig. S3G).
Comparative analysis between the coupling of lopinavir or ritonavir on SARS-CoV–2 Mpro and SARS-CoV Mpro showed that T25, S/A46, Y/M49, L141, S144, E166 and Q189 are present in the stabilization of ligands on SARS-CoV–2 Mpro and SARS-CoV Mpro. In addition, these compounds are better stabilized on SARS-CoV–2 Mpro than on SARS-CoV Mpro. All these docking-predicted complexes were submitted to MD simulation in the monomeric and dimeric states to validate their stabilization at the catalytic sites of SARS-CoV–2 Mpro and SARS-CoV Mpro.
3.3 Convergence of MD simulations
RMSD and Rg studies showed that monomeric SARS-CoV–2 Mpro and SARS-CoV Mpro in their free and bound states reached equilibrium between 10 and 20 ns with average values that oscillated between 1.6 ± 0.2 and 3.8 ± 0.2 Å for RMSD and 21.9 ± 0.2 and 23.1 ± 0.2 Å for RG (Table S1, Supplementary material). Dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro in their free and bound states reached equilibrium among 10 to 30 ns with average values that ranged between 1.5 ± 0.1 and 2.2 ± 0.2 Å for RMSD and 25.8 ± 0.2 and 26.1 ± 0.2 Å for RG (Table 2, Supplementary material). Therefore, for further analyses, the first 30 ns were discarded from the 100 ns simulation for each monomer and dimer simulations.
3.4 MD simulations of ligands with monomeric SARS-CoV–2 Mpro and SARS-CoV Mpro
MD simulations show that indomethacin, ofloxacin, fenoterol, nelfinavir, praziquantel and ritonavir) lost interactions at the catalytic site of SARS-CoV–2 Mpro. In contrast, naftazone (Fig. 1A), zopiclone (Fig. 1B), sofosbuvir (Fig. 1C), pitavastatin (Fig. 1D), eszopiclone (Fig. 1E), perampanel (Fig. 1F), azelastine (Fig. 2A), celecoxib (Fig. 2B), ondansetron (Fig. 2C), and lemborexant (Fig. 2D) maintained interactions with the catalytic site of SARS-CoV–2 Mpro. These compounds were mainly stabilized by M49, M165 and Q189 residues through nonpolar interactions. However, S46, G143, S144, H163, M165, C145, E166, P168, D187, T190 and Q192 formed polar interactions with backbone or side chain atoms with some of the compounds, including naftazone (Fig. 1A), zopiclone (Fig. 1B), T190 (Fig. 1C), pitavastatin (Fig. 1D), eszopiclone (Fig. 1E), perampanel (Fig. 1F), azelastine (Fig. 2A), celecoxib (Fig. 2B), ondansetron (Fig. 2C), and lemborexant (Fig. 2D).
3.5 MD simulations of lopinavir or ritonavir with monomeric SARS-CoV–2 Mpro and SARS-CoV Mpro
MD simulations showed that ritonavir lost interactions with the catalytic site of SARS-CoV–2, whereas lopinavir maintained the interactions with the catalytic site (Fig. 2E). Lopinavir was mostly stabilized by hydrophobic residues (M49, M165 and Q189) similar to those present in the fifteen repositioned compounds (Figs. 1 and 2), while it established polar interactions with the side chain of S46 (Fig. 2E). On SARS-CoV Mpro, lopinavir and ritonavir were mainly stabilized by L27, H41, A46, M49 and C145 through hydrophobic interactions, whereas lopinavir formed polar interactions with backbone atoms of A46 (Fig. 2F), and ritonavir formed polar interactions with the side chain of N142 (Fig. 2G).
Analyses between the coupling of lopinavir or ritonavir on SARS-CoV–2 Mpro and SARS-CoV–2 Mpro showed that only hydrophobic contacts with M49 were shared in the stabilization of the fifteen repositioned compounds on SARS-CoV–2 Mpro and SARS-CoV Mpro. In addition, the stabilization of these compounds was better on SARS-CoV–2 Mpro than on SARS-CoV–2 Mpro.
3.6 MD simulations of ligands with dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro
In contrast, with the observations with monomeric SARS-CoV–2 and SARS-CoV Mpro, MD simulations for most of the dimeric systems showed that all the ligands remained on both subunits of the dimer, except for the complexes between indomethacin, ofloxacin and lemborexant with SARS-CoV–2 Mpro, in which these compounds only remained at one of the catalytic sites of SARS-CoV–2 Mpro. Indomethacin coupled to subunit 2 (Fig. 3A), naftazone bound to subunits 1 and 2 (Fig. 3B and 3C), ofloxacin bound to subunit 1 (Fig. 3D), and zopiclone coupled to subunits 1 and 2 (Fig. 3E and 3F). Sofosbuvir bound at subunits 1 and 2 (Fig. 4A and 4B), pitavastatin bound at subunits 1 and 2 (Fig. 4C and 4D), and eszopiclone bound at subunits 1 and 2 (Fig. 4E and 4F). Perampanel coupled at subunit 1 or 2 (Fig. 5A and 5B), fenoterol bound at subunit 1 or 2 (Fig. 5C and 5D), and azelastine bound at subunits 1 and 2 (Fig. 5E and 5F). Celecoxib coupled at subunits 1 and 2 (Fig. 6A and 6B), nelfinavir bound at subunits 1 and 2 (Fig. 6C and 6D), and praziquantel bound at subunits 1 and 2 (Fig. 6E and 6F). Ondansetron bound at subunits 1 and 2 (Fig. 7A and 7B), and lemborexant bound at subunit 2 (Fig. 7C). These compounds were mainly stabilized by L27, H41, M49, N142, C145 and M165 through nonpolar interactions. T25, H41, T45, S46, L141, N142, G143, F140, S144, H163, H164, M165, E166, Q192, and Q189 formed polar interactions with backbone or side chain atoms with some of these compounds: naftazone (Fig. 3B), ofloxacin (Fig. 3D), zopiclone (Fig. 3E and 3F), sofosbuvir (Fig. 4A and 4B), pitavastatin (Fig. 4C and 4D), eszopiclone (Fig. 4E and 4F), perampanel (Fig. 5A and 5B), fenoterol (Fig. 5C and 5D), azelastine (Fig. 5F), celecoxib (Fig. 6A and 6B), nelfinavir (Fig. 6C and 6D), praziquantel (Fig. 6E and 6F), ondansetron (Fig. 7A and 7B) and lemborexant (Fig. 7C). Comparison of the residues stabilizing these ligands in the monomeric (Fig. 1 and 2) versus dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro (Figs. 3 and 7) revealed that the repositioned compound was better stabilized in the dimeric state than in the monomeric state. In addition, only in the complexes using the dimeric system, the presence of interactions with conserved residues (H41 and C145) involved in the catalytic activity was observed (Huang et al., 2004).
3.7 MD simulations of lopinavir or ritonavir with dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro
MD simulations showed that lopinavir at subunits 1 (Fig. 7D) and 2 (Fig. 7E) and ritonavir at subunits 1 (Fig. 7F) and 2 (Fig. 8A) were maintained interactions at the catalytic site of SARS-CoV–2. Ritonavir and lopinavir were generally stabilized by four residues (M49, M165, L167 and Q189), whereas only lopinavir formed polar interactions with backbone atoms and side chain atoms of T90 and Q189 (Fig. 7E). On SARS-CoV Mpro, lopinavir coupled at subunits 1 (Fig. 8B) and 2 (Fig. 8C) and ritonavir coupled at subunit 1 (Fig. 8D) were mostly stabilized by H41, M49, M165 and Q189 through hydrophobic interactions. Similar nonpolar and polar interactions were observed for the fifteen repositioned compounds (Figs. 3–7), except for L167. Comparative analysis of the residues stabilizing ritonavir and lopinavir in the monomeric (Fig. 2) versus dimer SARS-CoV–2 Mpro and SARS-CoV Mpro (Figs. 7 and 8) showed that ritonavir is stabilized by similar hydrophobic residues (M49 and M165) in the monomeric and dimeric states, whereas only two residues (H41 and M49) are shared in the stabilization of lopinavir in the monomeric and dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro.
3.8 Binding free energy calculations
Differences in affinity for the complexes between ligands and monomeric and dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems were calculated using the MM/GBSA approach, showing that all the bindings are energetically favorable and guided through nonpolar interactions, van der Waals energy (ΔEvdw) and the nonpolar free energy of desolvation (ΔGnpol,sol).. Binding free energy (ΔGbind) values for the ligands coupled at the monomeric SARS-CoV–2 Mpro show the following tendency: perampanel > lopinavir > ondansetron > pitavastatin > zopiclone > azelastine > sofosbuvir = eszopiclone > celecoxib > lemborexant (Table 1). However, a higher affinity towards monomeric SARS-CoV Mpro was exhibited by lopinavir than by ritonavir. Comparison of ΔGbind values for the affinity of repositioned compounds with ritonavir or lopinavir shows that perampanel was able to inhibit monomeric SARS-CoV Mpro in a similar manner to lopinavir and ritonavir, which diffuses in the first nanoseconds of MD simulations (see section 3.5). Comparisons between the affinity of lopinavir or ritonavir for SARS-CoV–2 Mpro and SARS-CoV Mpro systems showed that these two compounds exhibit a higher affinity by SARS-CoV than by SARS-CoV–2.
ΔGbind values for the ligands coupled on the first subunit of dimeric SARS-CoV–2 Mpro show the following tendency: perampanel > lopinavir > praziquantel > ritonavir > ofloxacin > azelastine > zopiclone > eszopiclone > fenoterol > pitavastatin> nelfinavir = celecoxib > sofosbuvir > ondansetron > naftazone. The ligands coupled at the second subunit showed the following order: nelfinavir > lopinavir > praziquantel > perampanel > azelastine > ritonavir > eszopiclone > fenoterol > ondansetron > pitavastatin> zopiclone > sofosbuvir > celecoxib > lemborexant > indomethacin > naftazone (Table 2). Based on this analysis, it is evident that perampanel and praziquantel can be proposed as anti-COVID–19 clinical drugs, whereas nelfinavir could also exhibit moderate activities against COVID–19. Interestingly, perampanel and praziquantel also exhibit a similar affinity to lopinavir and a higher affinity than ritonavir, both known inhibitors of SARS-CoV Mpro (Vastag et al., 2003). A comparison of the ΔGbind values of lopinavir and ritonavir on SARS-CoV Mpro versus SARS-CoV–2 Mpro indicated that these compounds exhibit a higher affinity for SARS-CoV–2 Mpro than for SARS-CoV–2 Mpro. In addition, a comparison between the monomeric versus dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems shows that although the employment of the monomeric system allowed us to identify perampanel and lopinavir as good inhibitors of SARS-CoV–2, it did not permit to the identification with praziquantel nelfinavir and ritonavir, highlighting the suitability of employing the dimeric system for drug discovery.
3.9 Per-residue free energy decomposition
An analysis of the residues contributing to the ΔGbind values for complexes with monomeric and dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems resulted in 5 to 11 residues (Table 3–6). An analysis of the residue stabilizing complexes between ligands and monomeric SARS-CoV–2 and SARS-CoV Mpro systems showed that H41, M49, M165 and Q189 were present in most of the complexes (Table 3), but only H41 and M165 were present for perampanel, the compound with the highest affinity for monomeric SARS-CoV–2 (Table 1); instead, it was stabilized by N142, G143, S144 and C145, which together with M165, contributed the most to the ΔGbind value. M49 and M165 were present in the stabilization of lopinavir, the second-best compound, in the monomeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems.
For the dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems, H41, M49 and M165 were present in the stabilization of almost all the complexes (Table 4–6). From these three residues, the energetic contribution of H41 and M49 was only present in one of the subunits for the complex between SARS-CoV–2 Mpro and perampanel (Table 5). H41 was present in both subunits of the complex between SARS-CoV–2 Mpro and praziquantel, and M49 was present only in subunit 1 for perampanel (Table 5). H41 and M49 were present in the complexes of SARS-CoV–2 Mpro with nelfinavir (Table 5) and lopinavir (Table 6). M49 was present in the complex of SARS-CoV–2 Mpro with ritonavir (Table 6).
As observed for the complex between perampanel and monomeric SARS-CoV–2 Mpro (Table 3), N142, G143, S144 and C145, together with M165, contributed the most to the ΔGbind value (Table 2) on both subunits of dimeric SARS-CoV–2 Mpro (Table 5). For praziquantel, the energetic contribution of M49 was only observed for one of the subunits, whereas participation of H41 and M165 was observed for both subunits, and as observed for perampanel, in which N142, G143, S144, C145 and M165 contributed importantly to the ΔGbind value (Table 2). For nelfinavir, the participation of H41, M49 and M165 was seen only in one of the subunits, the one with the higher affinity (Table 2), where it was also observed for the energetic contribution of D187, Q189, T190, A191 and Q192, which contributed importantly to the ΔGbind value (Table 2). Energetic contributions of H41, M49 and M165 residues were observed for complexes of lopinavir with the dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro (Table 6). Significant participation of P168, D187, Q189 and T190 was also observed but only for interactions at subunit 2 of the dimeric SARS-CoV–2 Mpro in complex with lopinavir (Table 6), whereas Q189 contributed importantly to the ΔGbind value in both subunits of the dimeric SARS-CoV Mpro (Table 6).
Energetic contributions of M49 and M165 were observed for complexes of ritonavir with dimeric SARS-CoV–2 Mpro and of H41 M49 and M165 with dimeric SARS-CoV Mpro (Table 6). It was also observed that there was significant participation of P168, Q189 and A191 for interactions of ritonavir at subunit 1 of dimeric SARS-CoV–2 Mpro and of D166, L167, P168, and Q189 for ritonavir at subunit 1 of dimeric SARS-CoV Mpro. Overall, this analysis supports the importance of two conserved residues (H41 and C145) (Nukoolkarn et al., 2008) in the stabilization of different inhibitors and highlights the importance of other residues (M49, N142, G143, S144, M165, D187, Q189, T190, A191 and Q192) in ligand stabilization.
3.10 Principal component analysis
PCA was performed to provide a quantified estimation of the differences in mobility. To this end, the trace of the diagonalized covariance matrix of the backbone atomic positional fluctuations was determined for the free and bound SARS-CoV–2 Mpro and SARS-CoV Mpro systems (Table 7). Based on this analysis, the values for free and bound monomeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems suggested that only the binding of sofosbuvir and lopinavir to monomeric SARS-CoV–2 Mpro was not coupled to conformational changes of monomeric SARS-CoV–2 Mpro. The binding of naftazone, pitavastatin, eszopiclone, perampanel, azelastine, celecoxib, ondansetron and lemborexant was linked to a decrease in the conformational mobility of monomeric SARS-CoV–2 Mpro, and the conformational reduction would be coupled to an increase in the ΔGbind value (Table 1), due to an unfavorable entropy component. The binding of zopiclone was coupled to an increase in the conformational mobility, which contributed to a decrease in the ΔGbind value due to a favorable entropy component. The binding of lopinavir and ritonavir was linked to a decrease in the conformational mobility of monomeric SARS-CoV Mpro (Table 7), which would also be linked to an increase in the ΔGbind value observed in Table 1.
Analysis of the covariance values for free and bound dimeric SARS-CoV–2 Mpro and SARS-CoV Mpro systems (Table 7) indicates that the binding of naftazone, zopiclone, sofosbuvir, eszopiclone, perampanel, azelastine, nelfinavir, praziquantel, lemborexant and lopinavir was not linked to important conformational changes of dimeric SARS-CoV–2 Mpro, which means that their coupling with receptors would not impact the affinity observed in Table 2. The binding of ofloxacin, pitavastatin and fenoterol contributed to a decrease in conformational mobility, and the coupling of indomethacin, celecoxib, ondansetron and ritonavir was linked to an increase in dimeric SARS-CoV–2 Mpro. However, the binding of lopinavir and ritonavir on dimeric SARS-CoV Mpro was not linked to conformational changes for lopinavir and an increase in the mobility of this receptor, which also means that their coupling on dimeric SARS-CoV Mpro did not impact the affinity observed in Table 2. Overall, this analysis shows that the binding of the best compounds on monomeric SARS-CoV–2 Mpro or SARS-CoV Mpro more importantly impacts the ΔGbind value estimated for each ligand due to the conformational changes coupled to the binding, whereas the affinity trends observed for the best compounds on the dimeric systems were not affected.