2.1. Inhibitory activities against HCV NS3/4A protease
A total of twelve compounds with the sulfonamidobenzamide scaffold were selected for comparison, with differing substituents at positions A and C (Fig. 2). These twelve compounds exhibited various degree of inhibitory activity (IC50) against the HCV NS3/4A protease as summarized in Fig. 2. The IC50 values of seven compounds were less than 10 µM varying from 1.2 µM to 8.0 µM. Of the seven, F0325-0125 possessed the best inhibitory activity at 1.2 µM. Four compounds (F1822-0567, F2708-0139, F2708-0140, and F2730-0247) showed IC50 values in the range of 12.7–32.9 µM, while F0325-0077 did not have any activity. Based on these observations, methyl, methoxy and chloro groups are tolerated at the para positions of the phenyl in ring A, while large aromatic substituents are favored for moiety C. Positions 6 and 7 in the benzothiazole ring are more tolerant of large groups than positions 4 and 5, which could be seen from F0328-0092 and F0328-0093. F0725-0019 also shows lower IC50 values with both positions 6 and 7 in the benzothiazole ring substituted. Methoxy is not tolerated at position 4 of the benzothiazole when a methyl was substituted at the para position of the phenyl ring A in F0325-0077.
2.2. Molecular docking of the sulfonamidobenzamide compounds to HCV NS3/4A
All twelve sulfonamidobenzamide compounds were subjected to molecular docking to predict the possible binding poses and to evaluate the binding affinities. The docking protocol was benchmarked by successfully redocking the crystallized ligand F9K [44] into the active site with an RMSD of 0.4 Å (Fig. S1). The docking pose with the highest ChemPLP [52] score for each compound was extracted for comparison. The results show that all inhibitors shared a similar binding pose, with certain differences for aromatic ring A and sulfonamide moieties in the catalytic sites of HCV NS3/4A protease. Ring A tends to occupy two different pockets, S1 or S1’, separated by the sidechain of Q41 (Fig. 3a). For example, ring A of F0325-0077, F0325-0086, F0325-0092, F0325-0093, F0325-0125, F0725-0019, F0816-0111, and F1813-0711 occupies the S1 pocket, and can form face-to-face stacking interactions with the amide sidechain of Q41, and edge-to-face stacking interactions with the imidazole ring of H57 and the phenyl ring of F43 (Fig. 3b). However, only edge-to-face stacking with the amide sidechain of Q41 was detected if ring A flipped to the S1’ pocket for the binding of F1822-0567, F2708-0139, F2708- 0140 and F2730-0247 (Fig. 3b). The sulfonamide moiety forms hydrogen bond interactions with G137 and S139, but different hydrogen bond networks were observed for inhibitors with ring A occupied in the S1 or S1’ pocket. For example, for F0325-0077, F0325-0086, F0325-0092, F0325- 0093, F0325-0125, F0725-0019, F0816-0111, and F1813-0711, with ring A occupying in S1 pocket, both sulfonamide oxygens are involved in hydrogen bonds with the sidechain OH of S139 and the backbone NH of residue G137. However, this sulfonamide moiety slightly rotates away accompanied by the occupancy of the S1’ pocket for ring A in the less active compounds F1822- 0567, F2708-0139, F2708-0140 and F2730-0247, and only one oxygen retains the hydrogen bond interactions with G137 and S139. Therefore, the sulfonamide functional group seems important for the binding of these compounds, which is also shown in the binding of the crystallographic ligand F9K (Fig. S2). The connecting phenyl ring B occupies the S3 pocket and remains buried into the hydrophobic pocket for all the compounds’ binding predictions, with the binding driven by van der Waal interactions with sidechains of L135, F154, A156, and A157 (Fig. 3) based on their docking poses. The stacking interactions of the inhibitor amide moiety and aromatic ring C with both sidechains of the catalytic H57 and the helicase domain residue Q526 of HCV NS3/4A were also well-preserved in all binding predictions. Steric bulky groups were tolerated at ring C, since the broad space of pocket S2 provides space for large groups, allowing for the introduction of more extensive interactions (Fig. 3a). An additional helicase residue, M485, was also involved in van der Waals contacts with ring C in the inhibitors. Therefore, residues M485 and Q526 from the helicase domain of HCV NS3/4A are found to be important for the binding of this inhibitor series from our docking calculations. A hydrogen bond interaction was also found between the backbone carbonyl oxygen of R155 and the amide NH of inhibitors F0325-0077, F0325-0086, F0325-0092, F0325-0093, F0325-0125, F1813-0711, and F1822-0567.
Hydrogen bond interactions with G137 and S139 seem more dominant than that with R155, since they were found in all docked inhibitor binding modes. The stacking interactions also played an important role for the binding of these compounds. However, it was still difficult to estimate which interactions predominate in binding, or which interactions were not stable based on docking results alone. Furthermore, poor correlation is observed between the ChemPLP score and the experimental values (Fig. 4a), which is a common and major problem with current docking approaches because of protein flexibility and lack of accounting for solvation effects. Thus, MD simulations were applied to offset this disadvantage, to explore the key interactions, and to explain the different activities by calculating the binding free energies using the MM/PBSA approach.
2.3 Binding free energy calculations by MM/PBSA approach based on 20-ns MD simulations and correlation with experimental results.
Twelve inhibitors were chosen to probe their various interactions with HCV NS3/4A at the atomic level and to explore the correlation between calculated binding free energies and the experimental values. In the MM/PBSA calculations, the explicit water molecules were stripped off, and a continuum implicit water model was used. Interaction entropy was estimated and combined with MM/PBSA computations to obtain more reasonable binding affinities. Our calculations reveal that MM/PBSA showed good correlation (R2 = 0.92) with experimental data for the 12 inhibitors (Fig. 4b) based on the 20-ns MD simulations and improved obviously comparing with that from docking scores (R2 = 0.56), which confirm that incorporating protein flexibility as well as solvent effects are essential for protein-inhibitor analysis. In addition, other atomic radii related to the PB [55] model and various GB models, including GBNeck, GBHCT, GBOBC, and GBOBC2, were systematically investigated in this study, all of which showed good correlation with experimental results (See Supplementary Information Figs. S3 and S4).
2.4. Hydrogen bond analysis.
Hydrogen bonds usually play an important role for the efficacy and specificity in protein- inhibitor interactions. Since docking suggests that several hydrogen bond interactions are involved in these compounds binding with HCV NS3/4A, these interactions were further explored by analyzing the 20-ns MD trajectory of representative compounds to elucidate their stability and contributions. Table 1 lists the three most prominent hydrogen bond interactions for HCV NS3/4A with four representatives, F0325-0125, F0325-0086, F2708-0140, and F0325-0077, possessing different inhibitory activities. Both sulfonamide oxygens were involved in the formation of hydrogen bonds for the most active compounds, F0325-0125 and F0325-0086, with residues G137 and S139 of HCV NS3/4A acting as hydrogen bond donors from docking prediction. These interactions were well preserved with around 90% occupancy throughout the 20-ns MD simulations. Interestingly, the sulfonamide of F2708-0140, showing only one oxygen atom maintaining the hydrogen bond interactions with G137 and S139 from docking, tended to move back with the second oxygen atoms getting involved in hydrogen bond interactions during MD simulations. However, the new joint interaction was not stable, with occupancy much lower than that of F0325-0125 and F0325-0086. For the inactive compound, F0325-0077, despite a network of two oxygen hydrogen bonds with G137 and S139 observed in docking similar to other active compounds, these interactions thoroughly collapsed during MD simulations, which could be the result of an unstable predicted docking pose. The results suggest that the hydrogen bond interactions for the sulfonamide moiety with residues of G137 and S139 are important for maintaining their interactions and inhibitory activity toward HCV NS3/4A. In addition to the hydrogen bond for the sulfonamide core, the hydrogen bond involving the amide linker with the backbone of R155 for F0325-0077, F0325-0086, and F0325-0125 shows only partial occupancy during MD simulations and was predicted not to be pivotal for stabilizing the binding conformations for inhibitors, which can also be seen from the docking prediction that only seven of the inhibitors could form hydrogen bond with R155.
Table 1
Hydrogen bond interactions of sulfonamidobenzamide based compounds with HCV NS3/4A during MD simulations.
ID | Interactions | Occupancya (%) |
Acceptors | Donors |
F0325-0125 | sulfonamide O1 | S139: sidechain OH | 99 |
sulfonamide O2 | G137: backbone NH | 96 |
sulfonamide O1 | G137: backbone NH | 88 |
F0325-0086 | sulfonamide O1 | G137: backbone NH | 94 |
sulfonamide O2 | G137: backbone NH | 86 |
sulfonamide O2 | S139: sidechain OH | 83 |
F2708-0140 | sulfonamide O1 | G137: backbone NH | 93 |
sulfonamide O2 | S139: sidechain OH | 39 |
benzothiazole N | Q526: sidechain OH | 40 |
F0325-0077 | benzothiazole N | Q526: sidechain OH | 14 |
H57: sidechain N | amide NH | 11 |
sulfonamide O1 | K136: sidechain NH3+ | 22 |
aPercentage occupancy during the analysis time period. The hydrogen bond is defined by a distance shorter than 3.5 Å and an angle larger than 120.0°.
2.5. Per-residue energy decomposition to explore the binding hot spots.
Per-residue energy decomposition of HCV NS3/4A with representative compounds (F0325-0125, F0325-0086, F2708-0140, and F0325-0077) was investigated to gain further insights into the energetic contributions of individual residues to complex formation using the MM/PBSA (bondi radii) approach. The most favorable interactions are formed by residues in the catalytic sites of the protease and part of the helicase. These are mainly by either hydrophobic interactions or hydrogen bonds. The energy contributions of the more important residues in the HCV NS3/4A - inhibitor interactions are shown in Fig. 5. The decomposition of the enthalpy revealed that all of these complexes have similar interaction networks, which further supports the hypothesis derived from docking that the inhibitors adopt similar binding modes. The most favorable interactions are formed by residues Q41, F43, H57, R109, K136, G137, S138, S139, and A156 in the protease domain as well as M485 and Q526 in the helicase domain. The significant contributions of G137 and S139 indicate their critical role of hydrogen bond interactions with these inhibitors. In addition, residues Q41, F43, H57, and Q526 probe the important function of the stacking interactions. The phenyl ring B occupies the hydrophobic pocket in the HCV NS3/4A protease and was deeply buried in the hydrophobic pocket during the simulations, which can be observed from the A156 contribution. M485 made large contributions to the binding free energies through hydrophobic interactions with ring C. It is notable that residues K136, S139, G137, and F43 contribute much more for F0325-0125 and F0325-0086 than for F2708-0140, and even more than that for F0325-0077. That is consistent with the hydrogen bond analysis stated above for these four compounds during MD simulations with residues G137 and S139, since persistent hydrogen bond interactions are pivotal for their binding. The edge-to-face stacking interaction between the phenyl ring of F43 and ring A is also important to discriminate among the distinct binding activities of the various inhibitors. The interactions and contributions of K136, which exhibits the most favorable binding energy residue for all complexes will be further discussed below.
2.6. Alanine scanning mutagenesis for HCV NS3/4A - F0325-0125 complex.
Computational alanine scanning was employed to determine the role of active site residues on binding of these compounds. The trajectories of the wild-type HCV NS3/4A binding with the most active compound, F0325-0125, were used to generate the structures of the mutated complex. This method depends on the assumption that the effects of mutating a residue to alanine only propagates locally and will result in insignificant entropic changes. For alanine scanning, the above important residues shown in Fig. 5 (except for G137 and A156) were mutated individually, and the results of the mutagenesis are presented in Fig. 6. Changes in the inhibitor-residue interactions associated with the alanine scan show that, in general, mutations of active site residues are highly unfavorable. Significant losses in binding free energies were observed for the catalytic residue of S139, as the S139A mutation removes a key hydrogen bond between the sidechain of S139 and the F0325-0125, responsible for stability and potency. In addition, the mutation of K136A exhibits a substantial destabilization. The hydrophobic and stacking interactions also contribute significantly to the interactions, which can be observed from the energy change by the mutation of Q41, F43, H57, M485 and Q526 to alanine. The mutation of S138A does not alter the binding from this result, since only the backbone is involved in the binding and the sidechain of S138 is away from the binding pocket.
2.7. Different contributions for K136 and S139 of HCV NS3/4A with F0325- 0125 complex.
K136 and S139 appear to be the two most important residues from per-residue decomposition and their sidechains are indicated to be pivotal for the binding of F0325-0125 to HCV NS3/4A by computational alanine scanning mutagenesis. However, the sidechain roles in nonpolar contributions and polar contributions clearly differ (Fig. 7a). K136 contributes predominantly by van der Waals interactions, which can be distinguished from the snapshot of the most dominant cluster complex from 20-ns MD simulations (Fig. 7b). The K136 sidechain forms tight hydrophobic interactions with ring B of F0325-0125 by its sidechain atoms of Cγ and Cδ. However, there is no Cγ, Cδ in the sidechain of K136A, resulting in a dramatic loss of binding affinity for the K136A mutant as well as losing the polar interactions of Nζ with the amide group of F0325-0125. On the other hand, S139 contributes more polar interactions than van der Waals interactions, similar to the other pivotal residue, G137, and dramatically loses ΔΔG for the S139A mutant, which cannot retain this polar hydrogen bond interaction with the sulfonamide oxygen. Therefore, K136A and S139A exhibit the same trend of losing binding affinity, but the losses are derived from different pivotal interactions.
2.8. Structural basis for potency differences between F0325- 0125 and F0325-0077.
The analysis of the 20 ns MD simulation trajectories of F0325-0125 and F0325-0077 docked HCV NS3/4A complexes followed by the comparison of the binding pocket residues of the final frame of 20 ns MD simulation of F0325-0125 with that of the docked receptor complex shows major conformational changes in the side chain orientation of residue R155 (Fig. 8a). For F0325-0125, the flexibility of the thiazole linked 2,5-dichlorophenyl moiety allows the compound to “induce fit” around this flipped orientation (i.e., state II) along with retaining the critical hydrogen bond interactions of the sulfonamide oxygens with the S139 and G137 (Fig. 8b). F0325-0077, which contains the rigid benzothiazole moiety, doesn’t fit well around this conformational flip, resulting in a destabilization of the critical H-bonding interactions and as a result, the compound completely collapsed during the 20 ns MD simulation run. Furthermore, the presence of a methoxy at position 4 of the benzothiazole in F0325-0077 results in a potential steric hinderance with the R155 side chain orientation (state II) (Fig. 8c), a conformation found to be stabilized during the 20 ns MD simulation with the F0325-0125 docked complex.
Comparison with the other crystal structures of HCV NS3/4A complexes also revealed that the R155 residue adopts two different side chain conformations (i.e., orientation around χ3) (Fig. 8d). For the macrocyclic ligand bound complexes (PDB IDs 3M5L [56], 6NZT [57], 6NZV [57] and 4A92 [44]), the orientation of the R155 sidechain is different (state I) than that with the PDB structures complexed with non-cyclic compounds (PDB IDs 2FM2 [58], 3LON [59], 3LOX [60]). A closer inspection of the binding pocket reveals that this flipped orientation (state II) is most likely triggered by the nearby protonated sidechain of the H57 residue. A protonated histidine side chain group is known to exhibit a high propensity for forming like-charged high-strength contact pairs with protonated arginine [61]. With the crystal structures bound to macrocyclic compounds, the H57 is involved in pi-pi interaction (4A92 [44]) or hydrogen bonding (3M5L [56], 6NZT [57] and 6NZV [57]) with the ligand, which is further stabilized by backbone hydrogen bonding interactions with G137 and R155.