The bacterial strains used in this study are listed in Table 1. These isolates were selected based on their known capacity to protect plants against fungal phytopathogens. They were routinely cultivated on 868 medium plates (yeast extract 16 g/l, casein peptone 10 g, glucose 20 g/l, agar 17g/l) at 26°C and maintained at 4°C before use. The maize variety used for experiments was Eckavel currently broadcast in Kivu by the Haves Plus program of CGIAR-IITA. Seeds displayed an average 90% germination rate under standard conditions. They were stored in aluminum bags at 4°C, relative humidity (RH) 50%.

The fungal pathogen was isolated from infected seeds and ears of maize collected in South Kivu (DR Congo, the province extends between 1° 44′13″ and 4° 51′32″ east longitude and between 26° 10′30″ and 29° 14′10″ south longitude). This infected plant material was placed on PDA (potato dextrose agar supplemented with chloramphenicol) plates during 18 days at 25°C. For further purification, the fungus was subcultured five times (every 2 weeks) on the same medium using spores as inoculum. The fungus was then maintained at 25°C by sub-culturing every 2 weeks on PDA medium. Fungal species identification was first based on morphology and microscopy observation (five replicates, independent cultures on PDA plates) and further identified by the laboratory of mycology at the Catholic University of Louvain-La-Neuve in Belgium. It was performed by sequencing the internal transcribed spacer (ITS) gene. Fungal DNA was extracted using EZ-10 Spin Column Fungal Genomic DNA Mini-Preps Kit (Bio Basic, Markham, ON, Canada). Amplification of ITS genes using universal primers ITS1 (TCCGTAGGTGAACCTGCGG)/ITS4 (TCCTCCGCTTATTGATATGC) as well as sequencing was achieved by Macrogen- Europe and DNA sequences were compared against NCBI database using BLAST alignment¹. In order to evaluate the pathogenicity of the fungus, the pathogen was grown on PDA plates for 4 days at 25°C. The mycelium was collected and placed in peptone water to recover spores. Maize seeds were dipped in a solution of 10⁸ spores/ml for 1 h at 25°C. Four seeds were then placed in Petri dish containing moistened Whatman paper for germination. The assessment of pathogenicity on seedling was evaluated by the incidence of diseases on seed rot, root decay, or reduction in both the number and mass of roots after 25 days.

Maize seeds were surface-disinfected by successively soaking in 3% sodium hypochlorite for 3 min and then in 75% ethanol for 2 min. The seeds were then washed extensively with sterile distilled water prior to use. Bacterial inocula were prepared as follows. Strains were cultured on 868 medium at 25°C for 48 h. A fresh colony culture was placed in 9 ml of peptone water (1 g/l Bactopeptone, 9 g/l NaCl, 0.02% Tween 80), and centrifuged at 4000 rpm during 20 min at 4°C. The supernatant was removed and the pellet re-suspended in peptone water and cell concentration was adjusted to 5 × 10⁸ cells/ml (based on the fact that an OD₆₀₀ value of 1 corresponds to 10⁸ cells/ml). For the preparation of fungal inoculum, the pathogens were grown on PDA medium for 10 days. Spores were scrapped from the mycelium in peptone water and centrifuged at 3000 rpm. For pathogen inoculation, the seeds were soaked for 40 min in a fungal suspension (1 ml per seed) at a concentration of 1.4 × 10⁷ spores/ml (Bürker cell chamber counting) and at the same time in a bacterial solution with a concentration of 5 × 10⁸ bacterial cells/ml. Treated seeds were placed in Petri dishes containing watered Whatman paper and incubated at 20°C for 10 days. The efficacy of Bacillus to control infection was determined using an arbitrary scale of Disease Reduction Index (DRI) established according to the symptoms observed on the seeds, and on hypocotyl/roots of young seedlings. This scale is as follows: (4) total protection of seed and seedling, no symptoms of pathogen damages, (3) intermediate protection, no damage to the seed or on hypocotyls and roots but few traces of spores on parts of the plant, no clear adverse effect on plant health, (2) serious damages on seeds/seedling due to obvious pathogen growth and infection, and (1) dead seeds and/or total decay of the seedlings.

These assays were performed on maize plants grown in potting soil (commercial soil DCM-ECOTERA with the following characteristics: 38% dry matter; 20% organic matter; pH 6.5; hydraulic conductivity 1.5 μS/cm; NPK: 7-7.5-8; 1.5 Kg/m³) to the third-leaf stage at 25°C and 50% RH. Prior to sowing, sterilized seeds were inoculated with the bacterium as described above while the pathogens were introduced into the growth substrate by mixing 15 g of maize flour inoculated with 4 × 10⁸ spores with 2 kg of potting soil. Both potting soil and maize flour were autoclaved prior to use. Mortality and typical symptoms on surviving plants were used as parameters to evaluate disease reduction according to the following arbitrary scale: (0) wilting and death of the entire plant, (1) 100% of the sheet exhibit symptoms but restricted to few basal leaves of the plant, (2) 50–75% of the sheet displays typical symptoms of the disease, (3) <25% of the leaves show symptoms of the disease, (4) <10% of the leave exhibit the symptoms of the disease, and (5) No visible symptoms on plants.

Surface-sterilized maize seeds were soaked for 10 min in the bacterial cell suspension at 10⁴ CFU/ml and placed into a 4-cm-diameter glass tube containing 3 g of perlite substrate and 9 mL of Hoagland solution. All this material was sterilized by autoclave. Plantlets were grown at 25°C with a photoperiod of 16 h. Cultures were stopped after 15 days, and S499 cells colonizing the maize rhizosphere were quantified using 1 g of root samples, after vigorous vortexing in the presence of glass beads. Vegetative cell concentrations were determined by plate counts of serial dilutions in parallel on LB medium, based on typical colony morphology, and on blood agar on the basis of the hemolytic halo typically formed around Bacillus colonies after 3 days of incubation at 28°C, as a result of lipopeptide activity. In parallel, lipopeptides were extracted from the rhizosphere by adding 6 mL of acetonitrile/formic acid 0.1% and 2 g of beads to each tube. Tubes were vortexed (5 min) and incubated overnight at 30°C with agitation at 2.6 g. The tubes were centrifuged, and the supernatants were vacuum dried (SpeedVac). Dried residues were suspended in the same solvent and processed as described elsewhere for analysis via UPLC–MS.

Sterilized seeds were placed in Petri dishes (five seeds per plate) containing moistened Whatman paper and incubated for 14 days in the dark at 22°C. Maize root exudates (MRE) were collected from these seedlings with 5 ml of sterile distilled water for each plate. The solution was used as such after filtration through 0.22 μm. The tomato exudate mimicking (EM) medium was adapted from Nihorimbere et al. (2012) and contained per liter: (NH4)₂SO₄ 1 g, CuSO₄ 1.6 mg, Fe₂(SO₄)₃ 1.2 mg, KCl 0.5 g, KH₂PO₄ 0.7 g, MgSO₄⋅7H₂O 0.5 g, MnSO₄ 0.4 mg, MOPS 21 g, Na₂MoO₄⋅2H₂O 4 mg, casamino acids 0.5 g, citrate 2 g, fructose 1.7 g, fumarate 0.5 g, glucose 1 g, malate 0.5 g, maltose 0.2 g, oxalate 2 g, ribose 0.3 g, succinate 1.5 g, yeast extract 1 g, pH 6.5. The LB medium contained tryptone 10 g, yeast extract 5 g, NaCl 10 g, pH 7. Cultures were performed in 24-well plates by inoculating 2 ml of fresh medium (LB, MRE, or EM) with 25 μl of a bacterial cell suspension (OD₆₀₀ = 0.1) corresponding to 1 × 10⁷ CFU/ml. Plates were incubated with continuous shaking for 40 h at 28°C. The supernatants obtained after centrifugation were filtered through a 0.2 μm membrane (Sartorius AG), after pooling the content of three wells (repetitions) for each strain.

To conduct the experiments, a small section of mycelium (3 cm) of an 8-day old fungal culture grown on PDA medium was taken and placed on agar-supplemented (15 g/L) MRE medium plate. Afterward, 2 days-old colonies of Bacillus grown on 868 medium were collected using a sterile toothpick and deposited on opposite edges of the Petri dish (approximately 3 cm from the piece of mycelium). Antagonistic activity was evaluated by measuring the distance (in mm) between the fungus and the bacterial colony after 3 days growth at 25°C. In order to extract lipopeptides from the inhibition zone on this gelified medium, liquid–solid extraction from 1-cm diameter agar plugs was performed. Briefly, agar samples were mixed with 500 μL of 50% acetonitrile homogenized by vortexing and incubated at 4°C for 4 h. The samples were centrifuged (13,000 g for 5 min) and the supernatant was filtered through 0.22 μm before further analysis.

For the estimation of lipopeptide concentration in all supernatant samples and extracts, the filtrates obtained were analyzed using UPLC–MS with UPLC (Acquity H-class, Waters s.a., Zellik, Belgium) coupled to a single quadrupole mass spectrometer (SQD mass analyzer, Waters) using an C18 column (Acquity UPLC BEH C18 2.1 mm × 50 mm, 1.7 μm). Ten μl was injected and elution was performed with a constant flow rate of 0.6 ml min^-1 using water/acetonitrile acidified with 0.1% formic acid gradient and compounds were identified/quantified based on their retention times and masses compared to commercial standards (98% purity, Lipofabrik Society, Villeneuve d’Ascq, France) as described by Debois et al. (2014).

Inhibition of the mycelial growth and sporulation of R. variabilis by pure fengycin (LipoFabrik, Villeneuve d’Ascq, France) was performed in two ways. Firstly, well-diffusion assay in PDA Petri dish with using different concentration (73, 37, and 18.5 μM diluted in 50% methanol) of pure fengycin was performed in order to evaluate effect on mycelial growth. Briefly, fengycin (100 μL) at different concentrations was put into well and 5-mm diameter plug of mycelium was placed in the middle of the dish. The control treatment consisted of 50% MeOH; after that, plates were incubated at 25°C for 48 h. Mycelium inhibition was determined by measuring the mycelium growth diameter compared with the control. Secondly, in order to estimate minimal inhibitory concentration (MIC) in liquid culture, 96-well microplates containing 150 μL of PDB medium (Potato Dextrose Broth) were inoculated with 30 μL of fungal spore suspension (10⁸ spores/mL) supplemented with different concentration of pure fengycin (18.5; 9.25; 4.5 and 2.5 μM, final concentrations). Plates were then incubated at 25°C for 48 h and the inhibition of pathogen growth was determined by measuring the absorbance at 620 nm (OD 620 nm) using a microplate reader (Beckman). Each experiment was repeated three times.

A plug of mycelium from gelified PDA plates was sampled and placed in 10 mL of salt peptone water (5 g/L NaCl, 1 g/L peptone, and 0.2% Tween 80). After centrifugation (10,000 rpm for 5 min) 10 μL of the solution was used to inoculate 10 times diluted MRE medium Petri dishes and incubated for 18 h at 25°C. Afterward, a bacterial colony from an overnight culture on solid 868 medium was picked up and striated on both sides of the already inoculated fungus (2 cm between both microorganisms). Agar plugs were sampled in the inhibition zone after 8, 12, 14, 16, 18, and 24 h of culture for further lipopeptide quantification via UPLC–MS while bacterial cells were scrapped from the producing micro-colonies at the same time-points for gene expression measurements by RT-qPCR.

RNA purification was done using a NucleoSpinRNA isolation kit (Macherey-Nagel, Germany) from cells coming from 15 g of agar in the inhibition region. The concentration of RNA was measured using a Thermo Fisher Scientific NanoDrop 2000c UV-Vis spectrophotometer. A quantity of 50 ng RNA was used as a template for RT-qPCR (StepOne Real-Time PCR thermal cycler system, Thermo Fisher Scientific) for fengycin gene expression (primers fenC_for: CTGAATCTCTTGCGCCATGT and fenC_rev: TGATCTGCTGTGCTCCTTCA) using qPCRBIO SyGreen 1-Step Hi-Rox (NIPPON GENETICS, Germany) according to manufacturer’s protocol. Experiment was performed in triplicate, and negative controls were included for each sample. Gyrase gene expression was used as the reference gene (gyrA_for: GAGACGCACTGAAATCGTGA gyrA_rev: GCCGGGAGACGTTTAACATA).

The effect of fungal VOCs on CLPs production was evaluated on MRE medium using two-compartments Petri plates (Fernando et al., 2005; Chaves-Lopez et al., 2015). In the first compartment, the bacterium was inoculated by depositing a 10 μl drop of a cell suspension adjusted to an OD of 0.1, whereas in the second compartment, a mycelial disk of freshly pre-cultured R. variabilis was deposited. Plates were incubated for 4 days at 25°C and the influence of VOCs on the production of LPs was determined by comparing as described above, the amount of compounds released under these conditions (see above for extraction and quantification by UPLC–MS) with those formed in control plates not inoculated with the pathogen.

Statistical analysis was carried out with Excel and Statistix 8.0 software. Data from biocontrol assays in vitro and in growth chambers were subjected to one-way analysis of variance (ANOVA) and significant differences between the various treatments were detected using (LSD) (least significant difference of means test) at α 0.05.

Phenotypical characterization of the fungal phytopathogen grown on PDA plate showed a typical cotton candy like texture. At early stage of growth, fungal colonies appeared white before turning grayish to gray after few days of culture. Microscopic observation (Supplementary Figure S1) shows hyaline, broad, non-septate, and branched hyphae. Erected unbranched sporangiophores (up to 12.7 to 16.2 μm wide), which appeared sparingly branched later, was also observable. In addition, sporangia appear globose, yellowish, and up to 70–75 μm in diameter sporangia with ellipsoidal to spherical columellae. Sporangiospores were ellipsoidal and often flattened on one side, measuring 7 to 9 μm × 2 to 4 μm. Sequencing of ITS of the fungal pathogen reveal 100% identity with R. variabilis type strain CBS 103.93. Pathogenicity of the fungus was evaluated in planta where a reduction of root number of 70.3 ± 15.9% occurred 25 days after infection, similarly a reduction of maize plants mass of 29.2 ± 18.6% was observable compared to uninfected maize plants.

First, we wanted to evaluate in vitro the biocontrol potential of different Bacillus amyloliquefaciens/subtilis strains on maize seeds and seedlings infected by R. variabilis. As shown in Figure 1, fungal pathogenicity was confirmed, since all maize seeds infected with R. variabilis and not treated with Bacillus strains (N.T) were unable to germinated and were fully covered with fungal mycelium (DRI = 1, see Materials and Methods). Interestingly, all tested Bacillus strains exhibit good antifungal activity against R. variabilis since all of them show DRI comparable to not infected (N.I) plants but B. amyloliquefaciens S499 appears to be the most active strain (albeit not significant) since no visual symptoms of fungal disease was observable (DRI = 4) (Figure 1 and Supplementary Figure S2). Strain S499 was further tested for its ability to control the pathogen and to prevent from take-all and leaf infection on plants grown in infested potting soil to the third-leaf stage. The results showed that bacterial treatment reduced significantly not only the mortality rate (respectively 55 ± 12% and 18 ± 4% for N.T and S499-treated plants) but also leaf infection compared to N.T plants with a high and significant (α < 0.05) reduction of disease severity (respectively 2 ± 0.7 and 4.2 ± 0.9 for N.T and S499-treated plants according to an arbitrary scale of DRI, Supplementary Figure S2).

Additional experiments were performed on 20 days-old maize plants elevated under gnotobiotic conditions (sterilized perlite as growth substrate) in order to evaluate S499 colonization and possible antibiotic production in planta. Data revealed that the bacterium efficiently colonized maize roots to reach populations of 1.1 ± 0.2 CFU/g of root (mean and SD calculated from five plants) and produced surfactins in relevant amounts (2.3 ± 0.4 μg/ml of rhizosphere) as well as some fengycins and iturins but in non-measurable amounts. We used a second approach to appreciate the ability of the B. amyloliquefaciens isolate to possibly form secondary metabolites such as lipopeptides under a more realistic nutritional context according to the rhizosphere ecology by growing S499 in exudates collected from maize seedling elevated under gnotobiotic conditions as sole nutritional source (MRE) medium. These naturally produced maize exudates represented an adequate food source as they sustained growth of the S499 strain to an OD₆₀₀ value of approximately 0.8. Lipopeptides of the three different families, i.e., surfactins, fengycins, and iturins (structures in Supplementary Figure S3) were also readily formed in these conditions as detailed in Figure 2 showing the whole range of typical ions detected by UPLC–MS analysis of the culture supernatant (Debois et al., 2014). Interestingly, these data also revealed high relative amounts of fengycins compared to the blend of lipopeptides released by S499 after growth in other media such as the artificial LB medium or in the EM medium containing C and N sources reflecting root-exuded products from other plant species such as tomato (Figure 3).

Plate confrontation assays were thus performed on gelified MRE medium and a significant inhibition zone (7.2 mm ± 1.7 mm) occurred after 4 days of incubation. Compounds accumulating in that zone were extracted from 1 cm-diameter plugs, and analyzed by UPLC–MS after solid–liquid extraction. Data were compared to those obtained from pure cultures of the bacterial strain performed under the same conditions. Again, lipopeptides were the only compounds detected in significant quantities in those extracts but, interestingly, the amounts of fengycins in the inhibition zone was higher when B. amyloliquefaciens S499 was grown in confrontation with the fungus, whereas the presence of the pathogen did not impacted positively or negatively the production of iturins nor surfactins (Figure 4). R. variabilis did not display any anti-bacterial activity against the Bacillus strain (data not shown).

In order to investigate the role of each lipopeptide family in the control of R. variabilis, we first compared the growth inhibition activity on MRE medium of several strains and mutants of B. amyloliquefaciens/subtilis producing different combinations of the three families of CLPs. Strains FZB42 and S499 both co-producing iturins, surfactins and fengycins, were able to inhibit the growth of the fungus even if S499 displayed a slightly higher antifungal activity (Figure 5). This is in accordance to seedling protection data (Figure 1). Strains that do not produce lipopeptides or only very low amounts such as B. subtilis BNO1 and B. subtilis 168 were not able to inhibit growth of the pathogen highlighting the role of lipopeptides in the antifungal bioactivity. FZB42 derivatives specifically repressed in the synthesis of CLPs (Koumoutsi et al., 2004; Chen et al., 2006; Jourdan et al., 2009; Cawoy et al., 2015) were also included in the experiment. Obtained from the natural strain FZB42, the mutants CH1 and CH2 both producing iturins but not able to form surfactins, only differ from each other in fengycin synthesis which is repressed in CH2 but conserved in CH1. Comparison of antifungal activities of CH1 and CH2 revealed that fengycin production had a strong impact on R. variabilis growth indicated a clear involvement of this CLP in fungal inhibition (Figure 5). Also surfactin does not play any role in antifungal activity, based on the rather low inhibitory effect of the mutant AK3 also obtained from FZB42, which produces surfactins but is repressed in the synthesis of the two other families. Interestingly, CH1 is more active than the WT strain FZB42, suggesting that surfactins could somehow interfere with fengycin activity. This arises from the fact that the two lipopeptides may interact to probably form a kind of precipitate yielding a “white line” visible around the colonies co-producing all three LP families during antagonism. This was observed for FZB42 and other strains co-producing the three CLP families (Cawoy et al., 2015). Such a phenomenon has also be reported for Pseudomonas CLPs involving the lipopeptide so-called WLIP (With Line Inducing Principle) (Rokni-Zadeh et al., 2013). Since B. amyloliquefaciens CH2 is still able to slightly inhibit the growth of the phytopathogen, iturins may contribute albeit to a much lower extent, to antifungal activity. The importance of fengycins in the global antifungal effect of Bacillus toward R. variabilis was also supported by the inhibitory activity of B. subtilis 2504, a derivative of the non-active B. subtilis 168, in which efficient production of fengycins was restored (Ongena et al., 2007; Nihorimbere et al., 2009).

Based on these results, pure fengycins were tested both in a well-diffusion assays for mycelium growth inhibition and in microtiter plates allowing to test the lowest inhibitory concentrations. As shown in Figure 6A, fengycin concentrations ranging from 73 to 18.5 μM were inhibitory for mycelium growth on MRE agar plates in a dose response manner. Smaller concentrations from 18.5 to 2.5 μM were also tested in liquid cultures for inhibition of conidia germination and a significant decrease in OD620 was observed by adding pure fengycins at concentrations as low as 4.5 μM, which roughly corresponds to the MIC (Figure 6B).

The dynamic of lipopeptide production by strain S499 in confrontation with R. variabilis was further investigated on MRE agar plates with 18 h-delayed inoculation of the bacterium in the vicinity of expanding fungal mycelium. Time-course UPLC–MS profiling of CLPs in the inhibition zone revealed a clear stimulation of fengycin production (up to 10-fold higher amounts) between 8 and 18 h following Bacillus inoculation compared to control plates (Figure 7A). This increase in fengycin production correlated quite well with an enhanced expression of the corresponding biosynthesis genes as analyzed by RT-qPCR from the producing bacterial colonies. Indeed, evaluation of the expression of the fenC gene showed that fengycin synthesis was overexpressed (Figure 7B) at 12 h after bacterial inoculation suggesting an early perception of some signal emitted by the fungus. This lipopeptide-inducing molecule emitted by R. variabilis most probably corresponds to a soluble compound diffusing in the gelified medium since fengycin stimulation was not observed by growing the bacterium and the fungus under the same conditions but in two-compartments plates commonly used to test the role of volatiles in interspecies interactions (data not shown).

The same experiments were also performed with two other well-described B. amyloliquefaciens isolates in order to appreciate the strain specificity regarding such communication between Bacillus and R. variabilis. It seems that the stimulation of fengycin may also occur for other strains since a similar effect of the fungus, albeit lower in amplitude (5.2-fold increase in fengycin amounts recovered at 18 h post-inoculation), was observed with strain FZB42 but doesn’t seems to apply for all members of this species since no stimulation was detected in the case of strain QST713 (Figure 7C).