3.1 Molecular Modeling, refinement and evaluation of Leptospiral collagenase model
Collagenase plays a pivotal role in the pathogenesis and it is known to be an important virulence factor responsible for the invasion and transmission [13, 22, 28, 33]. Therefore, it is important to develop drugs or inhibitors against Leptospiral collagenases to inhibit the pathogenesis of Leptopsira. We decided to search a molecular inhibitor for the collagenase to inhibit the pathogenesis of Leptospira. However, no crystal structure could be found for Leptospiral collagenase. When we searched sequence of collagenase from Leptospira interrogans Icterohaemorrhagiae serovar copenhageni (Fiocruz L1 – 130) against PDB database using different PAM and BLOSUM matrices, we did not find significant similarity with the crystal structure. The sequence similarity and query coverage identified for Leptospiral collagenase was found to be less than 30%, which makes it not suitable for the homology modeling. Therefore, we computationally modeled the structure of Leptospiral collagenase by employing the threading method. Secondary structure of the model was predicted by using PSIPRED online server [31]. The model of Leptopsiral collagenase was found to have 24 α helices, 19 β sheets and 50 turns (Figure 1). I-TASSER server [37] was used for threading the molecular structure of Leptospiral collagenase with the crystal structures of collagenase G (2Y3U and 4ARE) from Clostridium histolyticum as templates. Based on C-value generated from I-TASSER server, ranking of models was worked out. Model1 which showed the highest C-values was ranked as first. This model was used for further refinement of structure and evaluation. Model1 was subjected to loop building by using SPDB viewer software [16] to bring the amino acids from disallowed region to allowed region in Ramachandran plot. SPDB viewer was also used for energy minimization of the refined model of Leptospiral collagenase using GROMOS96 force field in vacuum. This structure was further re-refined using 3Drefine server [5] (Figure 2). The z-score was found to be -9.5 (Figure 3A) which indicated that the model lying in the range of available structures in the PDB that were resolved by X-ray crystallography. The residue energy was found to be largely negative (Figure 3B) indicating that the model has significantly less error prone regions. The amino acid distribution of Leptospiral collagenase in the Ramachandran plot was assessed by PROCHECK [24]. This analysis revealed that, phi-psi torsion angle for 81.8% of residues of Leptospiral collagenase are in the most favorable region (A, B and L), 5.5 % in additionally allowed region (a, b, l, p), 2.3 % in the generously allowed region (~a,~b,~l,~p) and 0.5% in the disallowed region of the Ramachandran plot (Figure 3C). Verify 3D [8] score was found to be 0.67 (Figure S1), which was greater than zero indicated that the environment profile of the model falls in the satisfactory level. The overall quality of the model was analyzed based on correctly and incorrectly positioned amino acids distribution by ERRAT2 [8]. The results showed that the value of 74.54 (Figure 3D), which indicates that the quality of model is good. The refined predicted structure of Leptospiral collagenase (model1) was superimposed with crystal structures of colG (2Y3U and 2Y50) crystal structures (Figure 4A, B and C). The RMSD values for 2Y3U, 2Y50 and top ranked model1 were found to be 0.99, 1.06 and 0.63, respectively.
3.2 Prediction of ligand binding sites in Leptospiral collagenase
Leptospira is known to secrete variety of extracellular proteases in order to digest the extracellular matrix of the host cells. Among them, collagenase is an important protease that targets and digests collagen which is the most abundant protein in the extracellular matrix of the host cells. Moreover, it is also found to be conserved among the different strains of Leptospira. Therefore, collagenase might act as a best drug target for Leptospira. In order to work out suitable inhibitors for the Leptospiral collagenase, we predicted ligand binding sites by using Metapocket 2.0 server [17]. From the predicted binding cavities, top three ligand binding sites were chosen, each one resided at N- terminal, catalytic site and C-terminal which were named as A, B and C, respectively (Figure 5).
3.3 Lead identification
From the available information in literature, it was found that Funalenone (ChEBI 65932) [18] could be a potent type-1 microbial collagenase inhibitor. Therefore, we looked for the structurally similar compound to Funalenone in ChEBI and Zinc databases. In ChEBI database, Pinoquercetin (CHEBI 8224) was found to be a structurally similar compound showing 70% similarity. These two molecules were further searched in the ZINC database to evaluate the similarity indices. 65 (4 molecules for Funalenone and 61 for Pinoquercentin) structurally similar compounds were found (Table S1) with 70% and 90% similarity, respectively. Overall 67 (2+4+61) molecules were taken further for docking studies to find out the potential inhibitor for Leptospiral collagenase.
3.4 Docking of validated lead molecules using Autodock vina
In order to find out the best lead molecules, 67 molecules were employed for docking study. The docking program was set to generate the ten best poses for each molecule at each predicted binding site (A, B and C) for the predicted Leptospira collagenase structure. After docking, the best poses with the lowest binding energy was chosen for each small molecule. However, the docking analysis of the selected small molecules revealed that there are variations in their binding energies. Our results indicate that four selected small molecules (Protohypericin, Hypericin, Protopseudohypericin and Pseudohypericin) efficiently bind to the Leptospiral collagenase (Figure 6 & 7) with minimum negative binding energy (kCal/mol) of -8.4, -8.2, -8.0 and -7.9 at A-site; -9.7, -9.6, -8.7 and -8.4 at B-site; -9.6, -9.2, -9.0 and -8.5 at C-site (Table S1). For Funalenone, the binding energy values are -5.9,-6.7 and -8.0 respectively for A, B and C- sites for best binding poses. The surface electrostatic potential surface of Letptospira collagenase as shown in figure 7 revealed the ligand binding cavities of superimposed four ligands at each predicted ligands binding pockets (Figure 7). The selected ligands occupied and were found to bind with the surrounding residues in the space between Cys23, Ser30, Asn32, Leu34, Thr39, Ala42, Gln43, Gln46, Gln47, and Glu62 at A-site; Asn481, Gly483, Arg497, Ser502, Ile503, Glu508, Leu509, His 512, Glu544, Ser563, Leu564, Glu567, and Tyr592 at B-site; Thr405, Tyr407, Asp410, Ala696, Phe697, Gly698, Lys838, Leu839, Gly841, Glu842, Leu843, and Leu846 at C-site (Figure 8). Interestingly, one of the predicted binding sites (B-site) was found to be the active site of the Leptospiral collagenase with HEXXH and EXXXE domains of GluZincin superfamily [39]. The docked molecule at the B-site has shown interaction with catalytic residues of Glu544 and Glu548 which are surrounded by His and Glu at 512th and 513th position [40] (Figure 8).
All four selected molecules were subjected to further analysis on the basis of binding energy and drug-like properties (Table S1). The selected molecules were found to be cyclic and possessed significant properties (xlogP, Apolar desolvation (kcal/mol), Polar desolvation (kcal/mol), H-bond donors, H-bond acceptors, net charge, tPSA (Ų), molecular weight (g/mol), and rotatable bonds) at pH 7. After comparative binding energy analysis at the three predicted binding sites, protohypericin was predicted to be the best inhibitor (-9.7 kCal/mol) among the 67 molecules that were screened. All those four lead molecules followed some of the Lipinski’s rule of five (Lipinski, 2004) at satisfactory level. Therefore, the rule of five can be bypassed by delivering these molecules through non-oral routes (Dermal, intravenous or pulmonary) [27]. It has been reported that, pulmonary permeability is less sensitive to the polar hydrogen bonding because generally pulmonary drugs have higher polar surface area [41] Therefore, it would be possible to increase the permeability nature of the selected inhibitors to act as drug by modifying their functional groups.
3.5 Pharmacophore Modeling
It is pertinent to mention that Leptospiral collagenase does not possess any crystal structure. Therefore, we used ligand-based Pharmacophore modeling using LigandScout [42]. Pharmocophore (active ligand) was generated based on alignment algorithm by superimposition of top four lead molecules binding at B site. Superimposition of ligands based on conformation of H-bond donors, H-bond acceptors, aromatic, hydrophobic and ionizable groups in 3-D space are shown in figure 9.
3.6 Molecular dynamics of collagenase with Protohypericin, Hypericin, Protopseudohypericin and Pseudohypericin
GROMACS molecular dynamics package was used to validate the stability and interaction of the collagenase with protohypericin, hypericin, protopseudohypericin and pseudohypericin. The lower RMSD values indicate the stability of ligands with collagenase. The RMSD of collagenase – hypericin complex started to fluctuate at 0.24 nm and completed at 1 nm for the span of 20 ns with average of 1 nm RMSD throughout the simulation duration (Figure 10a). The RMSD of collagenase-protohypericin complex started to fluctuate at 0.24 nm and completed at 1.4 nm for the span of 20 ns with average of 1.25 nm throughout the simulation period (Figure 10b). The RMSD of collagenase- protopseudohypericin complex started to fluctuate at 0.24 nm and completed at 1.8 nm for the span of 20 ns with average of 1.5 nm throughout the simulation period (Figure 10c). The RMSD of collagenase-pseudohypericin complex started to fluctuate at 0.24 nm and completed at 1.3 nm for the span of 20 ns with average of 1 nm throughout the simulation period (Figure 10d). The RMSD values of protein – ligand complexes revealed stronger binding of protohypericin, hypericin, protopseudohypericin and pseudohypericin with collagenase. These results are also inconcordance with the docking results of collagenase with protohypericin, hypericin, protopseudohypericin and pseudohypericin ligands.