Purification and identification of metabolites produced by Bacillus cereus and B . subtilis active against Meloidogyne exigua , and their in silico interaction with a putative phosphoribosyltransferase from M . incognita

To contribute to the development of products to control Meloidogyne exigua, the bacteria Bacillus cereus and B. subtilis were cultivated in liquid medium to produce metabolites active against this plant-parasitic nematode. Fractionation of the crude dichloromethane extracts obtained from the cultures afforded uracil, 9H-purine and dihydrouracil. All compounds were active against M. exigua, the latter being the most efficient. This substance presented a LC50 of 204 μg/mL against the nematode, while a LC50 of 260 μg/mL was observed for the commercial nematicide carbofuran. A search for protein-ligand complexes in which the ligands were structurally similar to dihydrouracil resulted in the selection of phosphoribosyltransferases, the sequences of which were used in an in silico search in the genome of M. incognita for a similar sequence of amino acids. The resulting sequence was modelled and dihydrouracil and 9H-purine were inserted in the active site of this putative phosphoribosyltransferase resulting in protein-ligand complexes that underwent molecular dynamics simulations. Calculation of the binding free-energies of these complexes revealed that the dissociation constant of dihydrouracil and 9H-purine to this protein is around 8.3 x 10 and 1.6 x 10 M, respectively. Consequently, these substances and the putative phosphoribosyltransferase are promising for the development of new products to control M. exigua.


INTRODUCTION
Plant-parasitic nematodes are a constant source of problems for Brazilian farmers due to the negative impact these animals cause to the production of several crops in Brazil (Dias-Arieira et al. 2010).For example, the production of coffee is severely affected by species of Meloidogyne (Campos and Villain 2005).In the State of Minas Gerais, where ~50% of this commodity in Brazil is produced (Ministério da Agricultura, Pecuária e Abastecimento -Mapa 2012), Meloidogyne exigua Goeldi occurs in 22% of the coffee farms in the southern region of this state (Castro et al. 2008).Although exclusion methods and plant resistance to nematodes should be emphasised, synthetic nematicide application is the most often used method to control M. exigua on coffee farms (Campos and Silva 2008) contaminates humans and the environment with harmful substances (Chitwood 2002).A possible alternative to circumvent such problems comprises the use of rhizobacteria, which colonise plant roots and can reduce the population of plant-parasitic nematodes.For example, Pseudomonas fluorescens (Flügge) Migula and Pseudomonas putida Trevisan can reduce the population of Meloidogyne spp.and Radopholus similis (Cobb) Thorne in tomato and banana plants (Aalten et al. 1998).The use of a combination of Bacillus thuringiensis Berliner and Lysinibacillus sphaericus (Meyer and Neide) Ahmed et al. to control plant-parasitic nematodes was protected by a patent (B'Chir 2000).One of the mechanisms by which rhizobacteria can act against nematodes consists of the production of nematicidal substances, for example the production of 2,4-diacetylphloroglucinol by P. fluorescens that can control Globodera rostochiensis (Wollenweber) Behrens in potato plants (Cronin et al. 1997).Also worthy of mention is the ability of Bacillus amyloliquefaciens (Fukumoto) Priest et al. to produce a cyclic peptide active against nematodes.The use of this substance as a nematicide was protected by a patent some years ago (Bendzko et al. 1998).Thus, in order to select rhizobacteria for the development of new products to control plant-parasitic nematodes and to elucidate their mechanism of action, in a previous work several rhizobacterial strains were screened for the ability to produce substances active against M. exigua (Oliveira et al. 2007).Then, the most active crude metabolites underwent a fractionation process monitored by in vitro assays with M. exigua second-stage juveniles (J2), which resulted in the identification of common amino acids (Oliveira et al. 2009) that were probably formed in the culture medium by hydrolysis of proteins or peptides that may be used as nutrients by rhizobacteria.Although the low nematicidal properties of amino acids preclude such substances from being commercially used to control M. exigua, they are sufficiently active to cause false positive results in a screening program aimed to detect bacterial metabolites with nematicidal activities.Consequently, in a preliminary work they were removed from crude bacterial metabolites by solvent extraction before the screening process.To continue such work, the most active amino-acid-free crude metabolites, produced by Bacillus subtilis Cohn and Bacillus cereus Frankland and Frankland, underwent fractionations in the present study to isolate and identify those substances with nematicidal properties against M. exigua.Furthermore, these compounds were also submitted to a computational study to elucidate their mode of action.

BACTERIA
This work was carried out with Bacillus cereus (strain 137JC) and B. subtilis (strain 18JC), which were previously isolated from tomato (Solanum lycopersicum L.), and pepper (Capsicum annum L.) plants, respectively (Silva et al. 2008).Both microorganisms have been deposited in the Department of Plant Pathology-Federal University of Lavras.In order to identify these microorganisms, their DNA was extracted according to the work by Ausubel et al. (1997).Then, the gene coding the ribosomal RNA 16S underwent amplification by polymerase chain reaction (PCR) using the primers fD1 and rP1 (Weisburg et al. 1991).The products of such amplifications were purified by the Qiaquick PCR purification kit (Qiagen) following the manufacture's recommendations.The sequencing reaction was carried out by the chain termination method, using Big Dye 3.1 (Applied Biosystems), followed by analysis in an automated sequencer ABI 377 (Applied Biosystems).The resulting sequences were compared with those deposited in the Ribosomal Database Project, release 10 (http:// rdp.cme.msu.edu)(Cole et al. 2009).BACTERIAL METABOLITES WITH NEMATICIDAL PROPERTIES PRODUCTION OF BACTERIAL METABOLITES Each bacterium was grown in 2.5 L of tryptic soy broth (TSB, Merck) for seven days, at 28°C, under constant stirring (100 rpm), in the dark.After bacterial cell removal by centrifugation (10,000 g, 15 min), the supernatant liquids were freeze-dried and washed with dichloromethane (4 x 1.0 L).The resulting dichloromethane-soluble fractions were combined and concentrated to dryness in a rotary evaporator to afford the crude extracts.An aliquot (0.5%) of each extract was dissolved in 2.0 mL of an aqueous 0.01 g/mL Tween 80 solution to be tested in vitro for activity against M. exigua J2.

IN VITRO ASSAY
The test was performed as described by Amaral et al. (2002).Briefly, M. exigua eggs were extracted from coffee (Coffea arabica L.) roots infected with the nematode in accordance with the Hussey and Barker (1973) technique, modified by Boneti and Ferraz (1981).J2 were hatched from the eggs and collected to be employed in the in vitro assays.Only less than two-day old J2 were used.Into each 300 μL well of a 96-well polypropylene plate, 20 μL aqueous suspension containing approximately 25 J2 and 100 μL of samples dissolved in aqueous 0.01 g/ mL Tween 80 solution, was poured.To evaluate the crude extracts, 30 μL of 3.0 mg/mL Pentabiótico (a mixture of antibacterial substances produced by Fort Dodge, Brazil) suspension was also poured into each well to prevent bacterial growth.After 48 h at 28°C, one drop of an aqueous 1.0 M NaOH solution was added and J2 which changed their body shape from straight to curled or hookshaped within 3 min were considered to be alive, whereas the nematodes not responding to the addition of NaOH were considered dead.This experiment was performed with four replicates per treatment, employing aqueous 0.01 g/mL Tween 80 solution and aldicarb [(2-methyl-2-(methylthio)propanal O-(N-methylcarbamoyl) oxime] (50 μg/mL) as negative and positive controls, respectively.For the aldicarb solution preparation, 8.6 g of Temik 150 (150 g of aldicarb/ kg), from Rhône-Poulenc AgroBrasil Ltda, was suspended in water and filtered through filter paper.
The resulting solution was diluted with water to the desired concentration.All values of dead J2 were converted to percentage before analysis of variance (ANOVA) and means separation according to the Scott and Knott (1974) test (P ≤ 0.05), which were performed using the SISVAR 5.1 software (Sistema para Análises Estatísticas, UFLA, Lavras, 2006).

IDENTIFICATION OF ISOLATED SUBSTANCES
To obtain mass spectra, ~0.5 mg of each substance was dissolved in 0.5 mL water/methanol (1:1) solution and 20 μL of the resulting solutions were directly infused at a flow rate of 5.0 μL/min into an Agilent 1100 LC/MS Trap mass spectrometer equipped with an electrospray ionization source in positive and negative ion modes.Probe and cone were maintained at ±3.5 kV and ±25 V, respectively.Nitrogen at 250°C was used as nebuliser (200 L/h) and gas drier (20 L/h), while selected ions underwent fragmentation by collisions with helium at 6 x 10 -6 bar.Substances were also introduced into a Shimadzu mass spectrometer model PQ5050 through a probe, which was heated from 60°C to 300°C for 20 min.Ionization was carried out by electron impact at 70 eV.
Substances were also dissolved in 0.8 mL of hexadeuterated dimethyl sulphoxide to obtain hydrogen ( 1 H) and carbon-13 ( 13 C) nuclear magnetic resonance (NMR) spectra on a Bruker AVANCE DPX 200 spectrometer ( 1 H at 200 MHz and 13 C at 50 MHz).Solvent peak was used as reference.

CONCENTRATION LETHAL TO 50% OF THE INDIVIDUALS (LC 50 )
Solutions of carbofuran (Sigma-Aldrich, 98%) and dihydrouracil in aqueous 0.01 g/mL Tween 80 solution were prepared and used in the in vitro assay with M. exigua J2 as described above, employing the Tween 80 solution as control.The final concentrations inside the wells were 224, 244, 265 and 296 µg/mL for carbofuran, and 171, 192, 213 and 233 µg/mL for dihydrouracil.Average values of dead J2 were converted to a percentage, corrected {correct value = 100 [(value -control value)/100control value]} and submitted to probit analyses on POLO-PC software (LeOra Software 1987).

PROTEIN MODELLING
Initially, a search for ligands similar to dihydrouracil was carried out using the ReverseScreen3D web service (Kinnings and Jackson 2011) (Christoffersen et al. 2009), 1BD4 (Schumacher et al. 1998), 1ORE (Silva et al. 2004a), 1L1Q (Shi et al. 2002), 1MZV BACTERIAL METABOLITES WITH NEMATICIDAL PROPERTIES (Silva et al. 2004b), 1QB7 (Phillips et al. 1999), 2H3D (Wang et al. 2006a), 1VQU (Joint Center for Structural Genomics, unpublished data) and 3LAR (Kang et al. 2011).The sequences with the highest scores (prot: Minc06801, contig: MiV1ctg193; prot: Minc10020, contig: MiV1ctg367; and prot: Minc03376, contig: MiV1ctg69) obtained when the amino acid sequence of 1BZY was searched against the MincV1A1.fasdatabase, were combined to form the sequence of the putative phosphoribosyltransferase from M. incognita (PPTM_A).Then, using the SWISS-MODEL Workplace (Arnold et al. 2006), this sequence was submitted to the SWISS-MODEL automated protein homology-modelling service (Schwede et al. 2003), which used chain A of the homotetramer 1BZY (1BZY_A) as the template.Employing Swiss PDBViewer 4.0.4(Guex and Peitsch 1997), the generated tridimensional structure was aligned with each one of the four chains of 1BZY to generate a tetramer (PPTM).Multiple sequence alignments of the putative phosphoribosyltransferase from Meloidogyne incognita (PPTM_A), the amino acids sequences of the genome from this nematode (prot: Minc06801, contig: MiV1ctg193; prot: Minc10020, contig: MiV1ctg367; and prot: Minc03376, contig: MiV1ctg69), chain A of 1BZY and another putative phosphoribosyltransferase from M. incognita that was described in the literature (0186631, Kloek et al. 2005)  Initially, both dihydrouracil and 9H-purine underwent optimisation with the Hamiltonian RM1, using the software MOPAC 2009 (James J. P. Stewart, Stewart Computational Chemistry, Version 11.052W).In this step, water was implicitly considered using the conductor-like screening model (COSMO).Four molecules of each optimised structure were aligned with four molecules of the ligand {phosphoric acid mono- [5-(2-amino-4-oxo-4,5-dihydro-3h-pyrrolo[3,2-d]pyrimidin-7-yl)-3,4dihydroxy-pyrrolidin-2-ylmethyl] ester} that was complexed to 1BZY.Using Swiss PDBViewer 4.0.4(Guex and Peitsch 1997), each chain of PPTM was aligned with the corresponding chain in 1BZY and merged with the optimised structures, resulting in two complexes containing four molecules of each ligand: PPTM-dihydrouracil and PPTM-9H-purine.The optimised structures of dihydrouracil and 9H-purine were also used for the generation of Charmm topology and parameter files based on the Charmm General Force Field (top_all36_cgenff.rtf and par_all36_cgenff.prm,v. 2a5) by using MATCH software (Yesselman et al. 2012).The PPTM-dihydrouracil and PPTM-9Hpurine complexes underwent optimisation of the protein side-chains by SCWRL4 software (Krivov et al. 2009) and hydrogen atoms were added to the protein by the autoPSF plugin of VMD 1.9.1 software (Humphrey et al. 1996), which employed the generated topology files for the ligands and the Charmm topology file top_all27_prot_lipid.rtf for protein.Finally, the resulting complexes were minimised by NAMD 2.8 software (Phillips et al. 2005) using the generated parameter files for the ligands and the Charmm parameter file par_all27_ prot_lipid.prm for protein.Solvent was implicitly considered in this step (Generalized Born), which comprised 1000 iterations.

MOLECULAR DYNAMICS SIMULATIONS
The minimised protein-ligand complexes were solvated with the Solvate plugin of VMD 1.9.1, to result in rectangular systems in which the minimum distance between protein and sides of the rectangle was 19 Å.These were neutralised by the Autoionize plugin of the same software, which added 32 sodium cations (Na + ) to each system.The final systems DENILSON F. OLIVEIRA et al.
underwent minimisation by NAMD 2.8 with the Charmm force field (5000 iterations), followed by NPT (isothermal-isobaric ensemble) molecular dynamics simulations for 3 ns, at 300 K, employing 2 fs time step, Particle Mesh Ewald (PME) algorithm to take electrostatic interactions into account, constant temperature control by Langevin dynamics, constant pressure dynamics by Nosé-Hoover Langevin piston and periodic boundary conditions.NAMDPlot (a plugin of VMD) was used to extract data from the output files generated by NAMD, while trajectory was analysed employing RMSD Trajectory Tool (another plugin of VMD), which aligned all frames to the starting structure before calculating root-mean square deviation (RMSD).

LIGAND BINDING AFFINITIES
A frame with the smallest total energy in the last 300 ps of each of the above-mentioned molecular dynamics simulation was taken, and water as well as Na + ions were removed.Both resulting pdb files containing only the protein-ligand complexes were manually altered and finally corrected by Swiss PDBViewer 4.0.4 in order to get them to the appropriate format for calculation of the binding free-energy by the use of PEARLS software (Han et al. 2006).
Using VMD, the last 200 ps of the dcd files generated by NAMD were converted to dcd files containing only protein and ligands, which were transformed into Amber trj trajectory files by the use of Simulaid software (Mezei 2010).The last frame of each simulation was converted by VMD to a pdb file containing only protein and the ligand.Then, the resulting pdb files were transformed into Amber pdb files by Simulaid software, to be used in the generation of Amber prmtop files by Chimera 1.6.1 software (Pettersen et al. 2004), which employed Amber Force Field ff99SB for protein and General Amber Force Field (GAFF; Wang et al. 2004) for ligands.Charges of ligands were computed through the AM1-BCC method, using the Antechamber (Wang et al. 2006b) module of Chimera.Both trj and prmtop files were used by Sietraj-20-03-2012 software (Cui et al. 2008, Naïm et al. 2007) to carry out solvated interaction energy (SIE) calculations in order to obtain the binding free-energies of ligands to protein.

RESULTS AND DISCUSSION
The phylogenetic analysis of the partial 16S ribosomal DNA sequences revealed that the bacterial isolates corresponded to B. cereus and B. subtilis, which seems reasonable since species in the genera Bacillus are among the most beneficial rhizobacteria described in the literature (Benizri et al. 2001).B. cereus is widely distributed in nature, being commonly found in soil (Oka et al. 1993), while B. subtilis is described as a plant growthpromoting rhizobacteria (Gupta et al. 2000, Ongena et al. 2005).
According to Bunch et al. (2003), amino acids present low solubility in liquids like dichloromethane.Thus, this was the solvent of choice to eliminate these substances, since they could produce false positives during the in vitro assay with M. exigua J2 (Oliveira et al. 2009).When submitted to the previously mentioned assay, dichloromethanesoluble metabolites produced by both B. cereus and B. subtilis presented activity against the nematode (Table I).This result seems reasonable since these bacteria can reduce the population of root-knot nematodes (Meloidogyne spp.) in plant roots (Araújo andMarchesi 2009, Xiao et al. 2012).According to the literature, this activity may be a result of the production of macromolecules active against nematodes by these bacteria (Germida et al. 2000, Sela et al. 1998).
After the fractionation by elution through a silica gel column, both dichloromethane-soluble metabolites afforded six fractions (Table II), amongst which those eluted with methanol presented the largest amounts.Thus, they were used in the following step that comprised HPLC analyses.The other fractions were stored for future work.BACTERIAL METABOLITES WITH NEMATICIDAL PROPERTIES The methanol fractions (Table II) were still very complex according to HPLC analyses.Thus, during the semi-preparative HPLC process, the main goal was the purification of substances present in larger amounts that could not be detected in the dichloromethane-soluble fraction of TSB not exposed to the bacteria.Such a process resulted in four fractions that underwent NMR and mass spectrometry analyses to elucidate their chemical structures.Interpretation and comparison of data with those reported in the literature allowed the identification of three substances: uracil, dihydrouracil and 9H-purine (Figure 1).The metabolite produced by B. cereus that was purified and identified was identical to one of the substances obtained from B. subtilis (see Experimental procedures: Fractionation of bacterial metabolites).
Chemical shifts observed (Table III) during the NMR analyses are in perfect agreement with NMR data published for dihydrouracil (Roberts

Dichloromethane-soluble metabolites
Dead J2 (%) a B. cereus at 1.22 mg/mL 51b B. subtilis at 0.72 mg/mL 60c Tween 80 at 0.01 g/mL (negative control) 6a Aldicarb at 50 μg/mL (positive control) 83d a Means followed by the same letter do not differ significantly  1978) and 9H-purine (Therese et al. 1975).Such results were corroborated by the mass spectrometry analyses, since 9H-purine afforded peaks at m/z (mass-to-charge ratio) 121 [M+H] + and 119 [M-H] + during the positive and negative ion mode analyses, respectively, performed by direct infusion of the substance into the electrospray ionization source.
Regarding dihydrouracil and uracil, peaks at m/z 114 (M + ˙) and 112 (M + ˙) were respectively obtained after the electron impact ionization process.When tested for in vitro activity against M. exigua J2, all the isolated substances significantly increased nematode mortality (Table IV).For dihydrouracil, which was the most active structure, a LC 50 of 204 µg/mL was obtained while the commercial nematicide carbofuran (2,2-dimethyl-2,3-dihydro-1-benzofuran-7-yl methylcarbamate) presented a LC 50 of 260 µg/mL under the same conditions, suggesting that dihydrouracil could be a more efficient nematicide than carbofuran.Although various biological properties have already been described for dihydrouracil (Roberts and Poulter 1978), uracil (Roberts and Poulter 1978) and 9H-purine (Therese et al. 1975), this is the first time that their nematicidal properties have been demonstrated.Thus, in order to understand their mode of action, an in silico study was carried out to identify the nematode enzyme that these substances may be affecting.
This part of the work initiated with a search for protein-ligand systems in which the ligand was structurally similar or identical to dihydrouracil.The best-matched system, with a 3Dscore of 0.80 (Kinnings and Jackson 2011), comprised an enzyme for telomere protection.As no relation with nematodes for this protein could be found in the literature, it was discarded.The next system in the rank (3Dscore = 0.72) consisted of a phosphoribosyltransferase complexed to xanthine (PDB code: 1A95; Parry et al. 1998), which was interesting since most parasites are unable to synthesise purine bases by de novo pathways.Thus, enzymes in the salvage pathways of preformed bases, like the phosphoribosyltransferases, are potential targets for the development of new products to control parasites (Craig III and Eakin 2000), especially those of the Meloidogyne genus (Kloek et al. 2005, Liu et al. 2006).Consequently, a search for analogous sequences of amino acids produced by M. exigua was carried out, but no good match could be found.Since the genomic information about Meloidogyne incognita (Kofoid and White) Chitwood is more comprehensive, it was used instead of M. exigua, resulting in some sequences that could correspond to a phosphoribosyltransferase.However, the scores for these sequences were very low (28 bits; Altschull et al. 1997).Therefore, the sequences of other phosphoribosyltransferases deposited in the Protein Data Bank (http://www.rcsb.org)were also used in this search and the best result was obtained for the PDB code 1BZY (Shi et al. 1999), which is a human homotetrameric enzyme.Scores of 143, 142 and 140 bits were obtained for the Minc06801 (contig: MiV1ctg193), Minc10020 (contig: MiV1ctg367) and Minc03376 (contig: MiV1ctg69) amino acids sequences, respectively, when the search was carried out in the MincV1A1.fasdatabase.As these sequences were very similar to each other, they were combined to form the sequence of the putative phosphoribosyltransferase from M. incongita (PPTM_A; Figure 2).The score of the alignment a Means followed by different letters differ significantly (P ≤ 0.05).BACTERIAL METABOLITES WITH NEMATICIDAL PROPERTIES of PPTM_A with 1BZY was 36.0%, while for another putative phosphoribosyltransferase from M. incognita (0186631, Kloek et al. 2005) the scores of the alignments with PPTM_A and 1BZY were 8.0 and 10.0%, respectively, suggesting the production by M. incognita of enzymes with similar functions, but with large differences in their amino acid sequences.In the following step, PPTM_A underwent homology modelling, which was carried out using chain A of 1BZY (1BZY_A) as the template.This process resulted in a three-dimensional structure (PPTM) with a Z-score of -2.217 according to the Qmean4 global score (SWISS-MODEL Workplace; Benkert et al. 2011).As 1BZY is a tetramer, four chains of PPTM were used to build the complete protein complexed to the bacterial metabolites.In this study, only dihydrouracil and 9H-purine were used.To reduce any clashes or torsion problems in the protein structure, the complexes PPTM-dihydrouracil and PPTM-9H-purine had the positions of their side chains corrected and both systems were minimised before the molecular dynamics simulation process, which was carried out until stabilisation of their RMSD (Figure 3).
Frames with the lowest total energy in the last 300 ps of each simulation (Figure 4) were used for the free-energy calculation with PEARLS (Han et al. 2006), which is a force-field-based scoring function.The calculated binding freeenergy for dihydrouracil (Table V) corresponded to a dissociation constant (K d ) of ~1.3 x 10 -5 M [ln (1/K d ) = -ΔG o /2476.38;where ΔG o : binding Gibbs free-energy, in Joules], while the K d for 9H-purine was 3.5 x 10 -5 M, which is in accordance with the greater in vitro nematicidal activity of dihydrouracil (Table IV).obtained by the authors of the software employed in this calculation (Cui et al. 2008, Naïm et al. 2007).The binding free-energy of dihydrouracil to PPTM (Sietraj, Table V) corresponded to a K d of ~8.3 x 10 -7 M, which suggests that this substance is a better inhibitor of PPTM than 9H-purine, for which the K d was ~1.6 x 10 -6 M.
In summary, this study has demonstrated that uracil produced in vitro by B. cereus and B. subtilis, as well as dihydrouracil and 9H-purine produced by the latter rhyzobacterium, were active against M. exigua J2.Specifically for dihydrouracil, the LC 50 against M. exigua was less than that observed for the commercial nematicide carbofuran, suggesting that dihydrouracil is a promising substance for the control of plant-parasitic nematodes.According to the in silico study, these substances may be acting against M. exigua through inhibition of its phosphoribosyltransferase, which may be a promising enzyme for the development of new products to control Meloidogyne spp., since this macromolecule appears to be important for the salvage pathways of preformed bases in these pathogens.

Figure 3 -
Figure 3 -Root mean square deviation (RMSD) of the putative phosphoribosyltransferase from Meloidogyne incognita, during the molecular dynamics simulation of the complex of this enzyme with dihydrouracil (PPTM-DHU) and 9H-purine (PPTM-PUR).

Figure 4 -
Figure 4 -Structures of the complexes PPTM-dihydrouracil (right side) and PPTM-9H-purine (left side), in which the protein is represented by Tube format and ligands are represented by the VDW format.This image was created by VMD 1.9.1 software.

TABLE II Fractions of dichloromethane-soluble metabolites produced by Bacillus cereus and B. subtilis, obtained after elution through a silica gel column.
and Poulter 1978), uracil(Roberts and Poulter