Structure-based virtual screening of highly potent inhibitors of the nematode chitinase CeCht1

Abstract Nematode chitinases play vital roles in various physiological processes, including egg hatching, larva moulting, and reproduction. Small-molecule inhibitors of nematode chitinases have potential applications for controlling nematode pests. On the basis of the crystal structure of CeCht1, a representative chitinase indispensable to the eggshell chitin degradation of the model nematode Caenorhabditis elegans, we have discovered a series of novel inhibitors bearing a (R)-3,4-diphenyl-4,5-dihydropyrrolo[3,4-c]pyrazol-6(2H)-one scaffold by hierarchical virtual screening. The crystal structures of CeCht1 complexed with two of these inhibitors clearly elucidated their interactions with the enzyme active site. Based on the inhibitory mechanism, several analogues with improved inhibitory activities were identified, among which the compound PP28 exhibited the most potent activity with a Ki value of 0.18 μM. This work provides the structural basis for the development of novel nematode chitinase inhibitors.

The importance of nematode chitinases indicates that they may be promising nematicide targets for the development of small-molecule inhibitors for nematode pest control 12 . Many GH18 chitinase inhibitors with diverse scaffolds have been reported so far, and some showed potential applications as antifungal agents, pesticides, and drugs 6,[13][14][15] . However, the inhibition of nematode chitinases is rarely studied, and only few inhibitors have been reported to be effective on nematode chitinases, including allosamidin, closantel, b-carboline, and 4-hydroxy-1,2,3-triazoles. Allosamidin, a natural product derived from the mycelium of Streptomyces sp., is a broad-spectrum GH18 chitinase inhibitor 16 . As a substrate analogue, allosamidin showed inhibitory activity against nematode chitinases from Heligmosomoides polygyrus, Brugia malayi, Loa, and Wuchereria bancrofti 10,17 . Allosamidin could also retard egg hatching and inhibited exsheathment. However, the polysaccharide scaffold of allosamidin makes it difficult to synthesise and has poor druggability. Closantel, a known anthelmintic drug, was previously discovered as a potent inhibitor against OvCht1 from O. volvulus and BmCht1 from B. malayi 18 . Closantel and its derivates were capable of affecting O. volvulus L3 molting [19][20][21] . In continued studies to discover OvCht1 inhibitors, 4hydroxy-1,2,3-triazoles were identified through bioisosteric modulation and scaffold hopping approaches 22 , and b-carbolines were obtained by screening a commercial library of natural products 23 . b-carbolines were capable of penetrating the worm cuticle and preventing filaria moulting. However, the binding modes of these compounds have not been elucidated, which imposes restrictions on their further optimisation and application.
The scarce nematode chitinase inhibitors may be, to a great extent, attributed to the lagged research on the structure of nematode chitinases. The availability of structure information could facilitate both structure-based virtual screening for inhibitor development and elucidation of inhibitory mechanism for inhibitor optimization [24][25][26][27] . Recently, we resolved the crystal structure of CeCht1 (PDB ID: 6LDU), a chitinase from the model nematode C. elegans 28 . In this study, exploiting the structure of CeCht1, we identified a series of inhibitors bearing a (R)-3,4-diphenyl-4,5-dihydropyrrolo [3,4-c]pyrazol-6(2H)-one (PP) scaffold. In addition, we demonstrated the binding mechanism by X-ray crystallographic analysis, which facilitated the further optimisation of these compounds and led to the identification of several compounds with improved inhibitory activity. This work provides a solid basis for the development of nematode chitinase inhibitors.

Protein expression and purification
The DNA encoding the target protein with a C-terminal 6 Â His affinity tag was cloned into pPIC9 vector and transformed into Pichia pastoris GS115 (Invitrogen, Carlsbad, CA). After 120 h of fermentation, the culture supernatant was collected and subjected to ammonium sulphate precipitation. The precipitate was dissolved and purified with a HisTrap FF affinity column (GE Healthcare, Uppsala, Sweden). Then the protein was deglycosylated by PGNase F and the deglycosylase was removed through HisTrap FF affinity column. The protein was further purified by anion-exchange chromatography.

Virtual screening
A hierarchical virtual screening strategy was used as described previously 13,27,29 . First, structural analogues to active hits were identified from a subset of commercially available compounds from ZINC database 30 employing substructure search and shape similarity calculations. Substructure search was performed using OEChem toolkit (OpenEye Scientific Software, Santa Fe, NM). Shape similarity calculations were performed using ROCS 31 . The conformational database used for shape similarity calculations was prepared using OMEGA 32 . Compounds in the screening library were scored using "TanimotoCombo" score. Structural analogues were then prioritised for the evaluation of CeCht1 inhibitory activity using molecular docking. The crystal structure of CeCht1-CAD was used for molecular docking calculations. The protein structure for molecular docking was prepared using Maestro, where all water molecules were removed, hydrogens were added and protonation states of all charged residues were assigned at neutral pH. Ligands for molecular docking were prepared using LigPrep. Tautomeric and ionisation states of all ligands were determined using Epik program 33 at neutral pH. Molecular docking was performed using Glide program in extra precision mode [34][35][36] . Grids for molecular docking calculation were prepared by including the catalytic residues and residues in both "þ" and "À" GlcNAc binding subsites. Ligands were scored using Glidescore with Epik penalties and a single pose per compound was saved.

Inhibitory activity assays
Compounds selected by virtual screening were purchased from Topscience (Shanghai, China; http://www.tsbiochem.com) for inhibitory activity assays. The inhibitory activity were assayed in end-point experiments using 4-methylumbelliferyl b-D-N,N'-diacetylchitobioside hydrate (4MU-(GlcNAc) 2 , Sigma, St. Louis, MO) as a substrate. The reaction mixture containing 20 mM sodium phosphate buffer (pH 6.0), 1% (v/v) DMSO, 10 nM CeCht1-CAD, 4 lM 4MU-(GlcNAc) 2 and inhibitor was incubated in a final volume of 100 lL at 25 C for 20 min. The reaction was stopped by adding 100 lL 0.5 M sodium carbonate, and fluorescence of the released 4-MU was quantified (excitation 366 nm, emission 440 nm). Experiments were performed in triplicate unless otherwise specified. The inhibition constant (K i ) was calculated using Dixon plots by changing the compound concentration at several fixed concentrations of 4MU-(GlcNAc) 2 (2 lM, 4 lM, and 8 lM).

Crystallisation, data collection, and structure determination
The purified protein was desalted in a buffer containing 20 mM Tris-HCl (pH 7.5), 20 mM NaCl and spin-concentrated to 15.0 mg/ mL. For crystallisation of CeCht1 with bound inhibitors, the protein was incubated with inhibitor at a final concentration of 0.1 mM overnight. Then co-crystallisation experiments were performed by vapour diffusion in hanging drops at 4 C. The volume ratio of protein to reservoir was 1:1 and the reservoir solution contained 0.1 M Bis-Tris, pH 6.0, and 25% PEG3350. Crystals were cryoprotected by gently increasing the cryoprotectant concentration in the drops (up to 22% glycerol) and directly flash frozen by immersion in liquid nitrogen before data collection.
The diffraction data were collected on the BL18U1 and BL19U1 at the Shanghai Synchrotron Radiation Facility in China 37 , and the diffraction data were processed using the HKL-3000 package 38 . Structures were determined by molecular replacement with Phaser using native CeCht1 (PDB ID: 6LDU) as the search model 39 . Iterative molecular models were manually built and extended using Coot 40 , and the X-ray structure was refined by PHENIX suite of programs 41 . Structural figures were prepared by PyMOL (DeLano Scientific, San Carlos, CA). The data collection and structure refinement statistics are summarised in Table 2. 3. Results and discussion

Identification of CeCht1 inhibitors with novel scaffold
Screening of an in-house collection of compounds accumulated in various chitinase inhibitor discovery projects in our laboratory resulted in the identification of several compounds that showed moderate CeCht1 inhibitory activity. Among these compounds, there were three compounds bearing a similar (R)-3,4-diphenyl-4,5-dihydropyrrolo[3,4-c]pyrazol-6(2H)-one scaffold (Figure 1), and  the best one (PP3) inhibited CeCht1 with a K i value of 38.3 lM ( Table 1). As compounds with this scaffold have not been previously described to possess activity against any chitinase, we decided to proceed with compound PP3. To identify compounds with improved CeCht1 inhibition, a hierarchical virtual screening was performed ( Figure 2). Initially, structural analogues were identified employing substructure search and shape similarity calculations using compound PP3 as the starting structure. Finally, molecular docking was used to prioritise compounds for the evaluation of CeCht1 inhibitory activity. A set of compounds (PP4-PP26) were identified, and most of these compounds showed improved inhibitory activity over the starting compound and the reported inhibitor closantel (Table 1).

Structure-activity relationship analysis
As shown in Table 1, all the compounds with better activity than PP3 had a pyridine group at R 1 position, indicating that the increase of hydrophobicity in this position may be of benefit to the inhibitory activity. It is worth noting that the position of the nitrogen atom in pyridine group had a marked impact on the inhibitory activity because the inhibitory activity of PP7 increased nearly 10-fold over PP6. The para-substitution was obviously superior to the meta-substitution, suggesting that the nitrogen atom may form important interactions with CeCht1. The substituent at R 2 position seemed to have little effect on the inhibitory activity as compounds PP8 and PP24 exhibited similar K i values. Compounds PP7, PP9, and PP12-PP17 only differed in the R 3 position, but their inhibitory activities showed significant difference. A bulky group at R 3 position may facilitate the increase of inhibitory activity. Comparison of PP9-PP11 or PP22-PP24 revealed that the methyl substituent at R 4 and R 5 positions is not conducive to inhibit CeCht1, especially for a mono-substitution at R 4 position. The differences in bioassay results between PP7 and PP8, or PP25 and PP26, also supported this inference.

Inhibitory mechanism
To gain molecular insights into the inhibitory mechanism of the PP series of compounds, we solved the structures of CeCht1 in complex with two inhibitors, PP7 and PP21 (Table 2), which potently inhibited CeCht1 with K i values of 0.76 lM and 4.51 lM, respectively. The crystal structure of CeCht1 in complex with PP21 was determined at a resolution of 1.86 Å. The electron density map showed that PP21 was well-anchored in the substrate-binding cleft of CeCht1 from subsites À1 to þ2, and stabilised by hydrophobic interactions and hydrogen bonds (Figure 3(A)). The nomenclature for substrate-binding subsites was named according to Davies et al., where subsite -n represents the non-reducing end and subsite þ n represents the reducing end 42 . The structure provided an explanation for the above structure-activity analysis. The dihydropyrrolopyrazol-6-one skeleton bound in a hydrophobic pocket lined with several aromatic residues and formed a 2.5-Å hydrogen bond with the backbone of Trp138 at the þ1 subsite. The pyridine moiety penetrated into the active site pocket and stacked well with Trp394 at the À1 subsite. Besides, the nitrogen atom formed a hydrogen bond with Tyr247. These interactions elucidated the reasons why pyridine group at R 1 position could significantly increase inhibitory activity. The phenol moiety interacted with Trp62 via T-shaped p-p contacts, and methyl substituents at R 4 and R 5 positions of the benzene ring would cause steric hindrance, resulting in a decrease of inhibitory activity. The 3methoxy-4-propoxyphenyl moiety of PP21 interacted with Trp138 with hydrophobic contacts. The methoxy group formed hydrogen bonds with Asp248 and Arg304, while the propoxy group extended to Trp253 at the þ2 subsite.
PP7 is the most potent among these compounds, with a K i value of 0.76 lM. The structure of the complex was also obtained and refined to 1.40 Å. The electron density map of the ligand was unambiguous in the substrate-binding cleft, which could easily be used to reconstruct the conformation of PP7 and clearly showed details of the interactions (Figure 3(B)). The dihydropyrrolopyrazol-6-one skeleton of PP7 was anchored in the hydrophobic pocket and formed a hydrogen bond with Trp138 while the pyridine moiety interacted with Trp394 and Tyr247, which was similar with those observed in PP21-CeCht1 complex. The phenol moiety bound in a small hydrophobic cave constructed by residues Val334, Tyr302, and Phe398, and it was further stabilised by forming a 2.6-Å hydrogen bond with Glu331. The ethoxyphenyl moiety hydrophobically interacted with Trp138.
The main chemical structure difference between PP21 and PP7 was the substitution at R 4 position. Compared with PP7, PP21 had a methoxy at R 4 position, which led to a 6.7-fold decrease of the K i value. Structural comparison showed differences in the binding modes of these two inhibitors (Figure 3(C)). First, although the lack of a methoxy group at R 4 position of PP7 abolished the formation of hydrogen bonds with Asp248 and Arg304, it pulled the inhibitor closer to the active site pocket. As a result, the side chain of Trp138 rotated 180 and stacked well with the ethoxyphenyl moiety of PP7. Second, the dihydropyrrolopyrazol-6-one skeleton of PP7 rotated about 51 and also got closer to the protein. This rotation resulted in a bigger conformation change of the phenol moiety which formed more stable hydrophobic contacts and induced a conformation shift of Glu331 to form a hydrogen bond. Finally, the dihydropyrrolopyrazol-6-one skeleton and phenol moiety was coplanar, which could enhance the hydrophobic  interactions between PP7 and CeCht1. Therefore, these structural differences made PP7 a more potent inhibitor than PP21.

Structure-guided discovery of more potent inhibitors
Structure-activity relationship analyses indicated that a bulky group at R 3 position is beneficial to the increase of inhibitory activity. The two crystal structures of inhibitor complexes showed that both the R 3 group (the propoxy group of PP21 and the ethoxy group of PP7), extended to the edge of Trp253 at the þ2 subsite. However, the interaction between these inhibitors and Trp253 was weak and there was still a plenty of space to accommodate a bigger group. Therefore, we hypothesised that PP7 derivatives with bulkier substituent groups at R 3 position would have better inhibitory activity. To confirm this hypothesis, we performed another round of virtual screening and obtained several derivatives (PP27-PP32). The inhibitory activity analysis showed that four compounds were better than PP7 (Table 3). Among these, compound PP28 exhibited the most potent activity as a competitive inhibitor, with a K i value of 0.18 lM (Figure 4(A)), which was a 4-fold increase than that of PP7. The docking calculation indicated that the benzyloxy phenyl group of PP28 extended to the cavity between Trp138 and Trp253 (Figure 4(B)). The benzyloxy phenyl group hydrophobically interacted with Trp253 and Trp138, forming a sandwich structure, which further improved the affinity of PP28. Besides, these two tryptophans together with the cavity they formed are conserved among different GH18 chitinases, and many GH18 chitinase inhibitors have taken advantage of this structural feature. Therefore, further optimisation at this position in the compounds might lead to better inhibitors.

Conclusion
In summary, we have identified a series of CeCht1 inhibitors bearing a novel scaffold. Structure-activity relationship analyses and crystallography studies clearly elucidated the inhibitory mechanism of these compounds. The crystal structures of enzyme-inhibitor complexes provided clues to develop compounds with improved inhibitory activity. This work presents an efficient strategy, which combined computational and experimental studies, to discover potent inhibitors. In addition, this work may promote further development of nematicides to deal with the increasing damages caused by nematode pests.