J Mol Model (2016) 22: 163 DOI 10.1007/s00894-016-3021-1 ORIGINAL PAPER Structure-based virtual screening of hypothetical inhibitors of the enzyme longiborneol synthase—a potential target to reduce Fusarium head blight disease E. Bresso1 & V. L ero ux 2 & M. Urban3 & K. E. Hammond-Kosack3 & B. Maigret2 & N. F. Martins1 Received: 17 December 2015 /Accepted: 27 May 2016 /Published online: 21 June 2016 # Springer-Verlag Berlin Heidelberg 2016 Abstract Fusarium head blight (FHB) is one of the most compounds from a library of 15,000 drug-like compounds. destructive diseases of wheat and other cereals worldwide. These putative inhibitors of longiborneol synthase provide a During infection, the Fusarium fungi produce mycotoxins that sound starting point for further studies involving molecular represent a high risk to human and animal health. Developing modeling coupled to biochemical experiments. This process small-molecule inhibitors to specifically reduce mycotoxin could eventually lead to the development of novel approaches levels would be highly beneficial since current treatments to reduce mycotoxin contamination in harvested grain. unspecifically target the Fusarium pathogen. Culmorin pos- sesses a well-known important synergistically virulence role Keywords Fusarium mycotoxins . Culmorin . Inhibitors . among mycotoxins, and longiborneol synthase appears to be a Homology modeling . Molecular dynamics . Ensemble key enzyme for its synthesis, thus making longiborneol syn- docking thase a particularly interesting target. This study aims to dis- cover potent and less toxic agrochemicals against FHB. These compounds would hamper culmorin synthesis by inhibiting Introduction longiborneol synthase. In order to select starting molecules for further investigation, we have conducted a structure- Fusarium head blight (FHB), caused by Fusarium based virtual screening investigation. A longiborneol synthase graminearum and 16 other Fusarium species, is one of the structural model is first built using homology modeling, most important wheat diseases in the world [1, 2]. Wheat followed by molecular dynamics simulations that provided infection during flowering (anthesis) results in grain contam- the required input for a protein–ligand ensemble docking pro- ination by mycotoxins. The latter is a well-known phenome- cedure. From this strategy, the three most interesting com- non observed before or during silage storage, reducing quality pounds (hits) were selected among the 25 top-ranked docked and yield while constituting an obvious risk factor for human and animal health [3]. Current management of this problem relies on multiple fungicide applications in combination with Electronic supplementary material The online version of this article semi-resistant cultivars. This increases production costs, fur- (doi:10.1007/s00894-016-3021-1) contains supplementary material, ther fuels public concerns over abuse of chemicals such as which is available to authorized users. pesticides in agriculture, and raises the chances that pathogen * N. F. Martins resistance will eventually occur and spread [4, 5]. The discov- [email protected] ery of new compounds that could block mycotoxin production with exclusivity and no environmental hazard would certainly 1 EMBRAPA Genetic Resources and Biotechnology Parque Estação provide a welcome alternative to control the pathogen. Biológica - PqEB, Av. W5 Norte, Brasília, DF, Brazil Among the mycotoxins produced by Fusarium [6], 2 CNRS, LORIA, Campus Scientifique, BP 239, culmorin has been found in relatively high level in contami- 54506 Vandœuvre-lès-Nancy, France nated wheat grains [7, 8]. Several studies on contaminated 3 Department of Plant Biology and Crop Science, Rothamsted grains established a clear link between culmorin (and Research, Harpenden, Hertfordshire AL5 2JQ, UK various hydroxy-colmorin) levels and the trichothecene 163 Page 2 of 13 J Mol Model (2016) 22: 163 deoxynivalenol (DON) mycotoxin [9, 10]. Moreover, a grow- possible templates for building a three dimensional (3D) model ing body of evidence indicates that culmorin actually en- of this protein, it was first necessary to search protein databases hances DON toxicity [10, 11]. Culmorin is produced by the in order to identify proteins that could be suitable templates for biotransformation of farnesyl diphosphate through a complex obtaining a robust and convincing homology 3D model. Our pathway. A key element of this pathway is longiborneol syn- selection was based on close relatednesses in both structure and thase, which produces longiborneol (Fig. 1a), which differs function within all members of the family. We focused mostly from culmorin by the lack of a single hydroxyl group on the isoprenoid domain characterized by a well-conserved (Fig. 1b), but was also recently identified as being required structural organization that is mostly α-helical with a core bun- to complete culmorin biosynthesis in possible association with dle of antiparallel α-helices. The superposition of 33 structures an oxygenase [12]. of this family shows clearly the conservation of this structural This work describes initial efforts to identify potential organization within the whole family (http://www.cathdb. longiborneol synthase inhibitors, aiming to develop a new info/version/latest/superfamily/1.10.600.10). We checked suite of agrochemicals for crop protection with reduced risks carefully that this structural signature had been conserved for human and animal toxic contamination. For this purpose, during our homology modeling process. virtual screening represents a valuable in silico approach as it Once the protein class was identified and the possible allows screening large chemical libraries in order to detect a orthologs detected, we searched for similar structures at the limited number of compounds applicable for experimental Protein Data Bank (PDB), [16] to look for the most suitable testing, at a sensible cost [13, 14]. template for homology modeling. This step is crucial as the Since no experimentally derived 3D structure exists for quality of any homology model depends strongly on the per- longiborneol synthase, we first built a stable homology model centage of similarity, identity and sequence coverage. When that was further subjected to molecular dynamics (MD) sim- low sequence identities are detected, as in this work (less than ulations. This resulted in the selection of a conformational 20 %), the modeling task is not trivial and requires a suitable ensemble encompassing longiborneol synthase flexibility in strategy to improve the quality of the proposed models. The its most stable structural state. This data was then used to use of multiple templates has been shown to generally in- perform ligand-to-protein docking calculations [15]. The pres- crease model quality over a single template [17]. Despite re- ent study presents the docking simulation for 15,000 com- cent improvements [18], this approach still needs to use rules pounds that allowed the selection of three putative to combine information from all the templates. In our case, we longiborneol synthase inhibitors to be submitted to experi- chose to use several well-known homology web servers and mental validation. compared their results. Hence, a comparative dendrogram showing the relationship between the proposed templates and our target protein was established using the phylogeny.fr Materials and methods web server [19, 20], in which MUSCLE 3.8.31 was used for multiple sequence alignment with default parameters (find Homology modeling diagonals option disabled, a maximum number of iterations: 16, no duration limitation and no more than 200 sequences) The F. graminearum longiborneol synthase coded by the locus [21]. To draw the dendrogram, based on the protein alignment, FGSG_10397 (ACY69978.1) [12] is described in the Uniprot we used the phylogeny.fr web server set PhyML3.1/3.0 aLRT database as being encoded by the CML1 gene (see http://www. with default parameters (Substitution model: WAG for pro- uniprot.org/uniprot/D1M8S2) and belongs to the fungal teins, aLRT test: SH-like, number of substitution rate catego- sesquiterpene synthase superfamily of enzymes. To identify ries: 4, gamma parameter: estimated, proportion of invariable Fig. 1 a Longiborneol A biosynthesis pathway. b HO Longiborneol and culmorin O O - chemical structures O P O P O + H2O + diphosphate O- O- (2E,6E)-farnesyl diphosphate longiborneol B OH HO OH culmorin longiborneol J Mol Model (2016) 22: 163 Page 3 of 13 163 sites: estimated and the transition/transversion ratio: 4; the tree embedded in a box of 100 Å × 100 Å × 100 Å with TIP3P branch support was evaluated by an approximate likelihood- explicit water molecules [25]. Next, Na+ ions were added to ratio test) [22]. ensure electrostatic neutrality. The NAMD program version The closest template was retained to perform the homology 2.6 was employed in conjunction with the CHARMM27 force modeling task using the MODELLER program with its de- field [26, 27] in order to simulate the ensemble of the 94,678- fault settings [23] (see the MODELLER manual and tutorial). atom system. The initial state for dynamics was generated Additionally, the automatic loop refinement method available from the model after 64,000 steps of conjugate gradient min- in MODELLER was used. The crude longiborneol synthase imization. The conformational behavior of the protein, water 3D0 model was achieved via this route. The model was vali- and counter ions system was obtained by running 100 ns of dated