Antagonism of Two Plant-Growth Promoting Bacillus velezensis Isolates Against Ralstonia solanacearum and Fusarium oxysporum

Plant growth promoting rhizobacteria (PGPR) provide an effective and environmentally sustainable method to protect crops against pathogens. The spore-forming Bacilli are attractive PGPR due to their ease of storage and application. Here, we characterized two rhizosphere-associated Bacillus velezensis isolates (Y6 and F7) that possess strong antagonistic activity against Ralstonia solanacearum and Fusarium oxysporum under both laboratory and greenhouse conditions. We identified three lipopeptide (LP) compounds (surfactin, iturin and fengycin) as responsible for the antimicrobial activity of these two strains. We further dissected the contribution of LPs to various biological processes important for rhizosphere colonization. Although either iturin or fengycin is sufficient for antibacterial activity, cell motility and biofilm formation, only iturin plays a primary role in defense against the fungal pathogen F. oxysporum. Additionally, we found that LP production is significantly stimulated during interaction with R. solanacearum. These results demonstrate the different roles of LPs in the biology of B. velezensis and highlight the potential of these two isolates as biocontrol agents against phytopathogens.

Ralstonia solanacearum is an economically important bacterial phytopathogen 13 and infects a broad range of host plants including crops such as potato (Solanum tuberosum), eggplant (Solanum melongena), and tomato (Solanum lycopersicum) 14,15 . Tomato wilt caused by R. solanacearum is one of the most damaging tomato diseases, particularly during the hot-humid summer season in subtropical regions such as South China. No completely effective chemical controls are available 16,17 . A few resistant tomato cultivars have been developed such as FL7514 and BHN466 18,19 ; however, host resistance is often linked with unwanted traits such as reduced fruit size 20 . Some LP-producing Bacillus species have been reported as promising biocontrol agents [21][22][23] . However, the role of LPs in their biocontrol activity against R. solanacearum is unclear.
F. oxysporum f. sp. cubense 4 (Foc), a soil-borne fungus, is the causative agent of banana Fusarium wilt (also called Panama disease). Fusarium wilt is a global threat to the banana industry. It has seriously affected banana plantations in many countries and caused severe crop losses 24,25 . In South China, Fusarium wilt is the major disease faced by many banana plantations 26 . The disease incidence is in the range of 20 to 40%, with some plantations reaching a rate of 90% 27 . Most commercial fungicides are ineffective at controlling the disease. Currently, the best solution is the use of genetically resistant cultivars, but these plants are not resistant to all races of F. oxysporum. Emerging evidence has shown that LP-producing PGPR offers a potential environmentally-friendly biocontrol method against F. oxysporum 9,12 . However, PGPR isolated from non-native locations may not perform well, since they may not survive in the local soil environment due to circumstantial variables (e.g. pH, moisture, temperature), may be outcompeted by indigenous bacteria, or may not be effective against local pathogens 28,29 . These considerations motivated us to isolate and characterize PGPR endemic to the area where they will be applied.
Here we isolated and characterized two rhizosphere-associated B. velezensis strains (Y6 and F7) from a local farm in South China. These strains exhibit potent biocontrol activity against tomato bacterial wilt and banana Fusarium wilt under greenhouse conditions. We identified and quantified the antimicrobial LPs secreted by the two strains and defined their roles in biological processes relevant to rhizosphere colonization and plant protection. Either iturin or fengycin is sufficient for antibacterial activity, cell motility and biofilm formation, whereas iturin but not fengycin plays an essential role in antagonism against Foc. These two potential biocontrol agents offer a promising strategy to combat plant pathogens.

Results
Isolation and identification of two Bacillus strains with significant antibacterial activity. Many rhizosphere-associated Bacillus isolates exhibit antimicrobial activity towards phytopathogens 10,12,30 . To isolate antagonistic Bacillus species for use in the subtropical regions of China, we collected rhizosphere soil from healthy tomato plants in a local farm in Guangzhou. Among 60 isolated strains, two exhibited the strongest inhibitory effects against R. solanacearum GMI1000 on Casamino acid-Peptone-Glucose (CPG) plates using an optimized spot-on-lawn assay, where a high-density culture of the test strain is spotted on a low-density lawn of the indicator strain (R. solanacearum) (Fig. 1A). These two isolates were named Y6 and F7. Phylogenetic trees, constructed based on 16S rRNA and gyrA gene sequences, indicate that Y6 and F7 cluster with plantarum subspecies of B. velezensis such as S499, UMAF6614, SQR9 and FZB42 (Fig. S1).
These two isolates showed strong antimicrobial activity against R. solanacearum in vitro. Thus, we tested their ability to suppress tomato wilt caused by R. solanacearum in pot experiments under greenhouse settings during spring and autumn. After treatment with either Y6 or F7, the wilt incidence was significantly reduced in both sets of experiments (Fig. 1B,C). For example, in pot experiments done in spring, the wilt incidence in the control group (~65%) was reduced to 29% or 35% for Y6-or F7-treated groups, respectively (Fig. 1B,C). These results reveal that the disease control ability (biocontrol efficacy) of these two isolates approaches 50%.
Identification of the lipopeptide compounds secreted by Y6 and F7. In some PGPR, the biocontrol activity is linked to their ability to produce lipopeptide (LP) compounds 21,31 . We isolated the secreted metabolites produced by the Y6 and F7 strains grown in LB medium and both extracts showed very strong activity against R. solanacearum with a clearance zone of 13-14 mm in diameter ( Fig. 2A). We then identified the LP compounds from the extracts using ultra-performance liquid chromatography coupled with mass spectrometry (UPLC-MS). Three major LPs were detected: iturin, fengycin, and surfactin (Fig. 2B). Multiple isoforms of each LP were observed, including five for surfactin (C12 to C16), three for iturin (C14 to C16), and six for fengycin (C12 to C16) ( Fig. 2B and Table S3). Consistent with the observation that Y6 showed higher antimicrobial activity against R. solanacearum than F7 (Fig. 1A), Y6 produces significantly higher levels of iturin and fengycin than F7 (Fig. 2C,D and Table S3).
To further evaluate the antimicrobial potency of these isolated LPs, we tested the minimum inhibitory concentration (MIC) of the LP standards and compared them with the extracts from Y6 or F7. Both iturin and fengycin standards showed relatively strong antibacterial activity against R. solanacearum with a MIC of 100 µg ml −1 (Fig. S2B,C), while surfactin showed moderate activity with a MIC of 500 µg ml −1 (Fig. S2A). By contrast, the MIC of the extract from either Y6 or F7 is 75 µg ml −1 (Fig. S3D,E), considerably lower than that of any single LP ( Fig. S2A-C), suggesting that these three LPs may have synergistic antagonistic activity against pathogenic organisms.
LP production is strongly stimulated in the presence of R. solanacearum. LP production by PGPR can be modulated by the presence of a pathogenic organism 32 . Thus, we evaluated the influence of R. solanacearum on the production of LPs by Y6 and F7. CPG agar samples were harvested from the clearance zone (Fig. 1A) and LP compounds were extracted using an acetonitrile/water based method and then subjected to UPLC-MS analysis. Co-culture with R. solanacearum significantly stimulated LP production, particularly SCiEntifiC REPORTs | (2018) 8:4360 | DOI:10.1038/s41598-018-22782-z fengycin and iturin. When encountering the pathogen, iturin production increased 11-and 20-fold (F7 and Y6 respectively) and fengycin production increased 15-and 38-fold (F7 and Y6, respectively) (Fig. 3A,B and Table 1). A moderate increase of ~4-fold was observed for surfactin production by Y6 and F7 (Fig. 3A,B and Table 1).
Each LP contains multiple isoforms (Fig. 2B, Table 1 and Table S3) that may possess different antimicrobial activity and which may change upon the presence of other bacteria 33,34 . Indeed, upon R. solanacearum challenge, the isoform profile of all three LPs changed: the C15 isoform increased in all three LPs and became the dominant iturin and surfactin species (more than 50%) ( Table 1). Whether or not the C15 isoform is most effective against R. solanacearum is unknown.
The elevated LP production by F6 and Y7 in response to R. solanacearum might be due to an increase in expression of LP biosynthesis genes. In fact, the expression of LP genes is highly upregulated in several Bacillus isolates during interaction with phytopathogens 35,36 , although LP production was not quantified in these studies. To test this notion, we monitored LP gene expression in the presence of R. solanacearum. Indeed, the expression of fenD (encoding fengycin synthetase D) in both Y6 and F7 was significantly upregulated (5-6-fold), whereas ituC (encoding iturin A synthetase C) and srfAB (encoding surfactin synthetase A) expression increased 2-3-fold upon encountering R. solanacearum (Fig. 3C). These results are consistent with the LP quantification results The biocontrol ability of the two isolates to suppress tomato bacterial wilt was evaluated in pot experiments during spring and autumn using nonsterile local soil (see Additional Methods in the SI materials). Three groups were included: control (CK), only inoculated with R. solanacearum (Rs); Y6 + Rs, inoculated with Y6 and Rs; and F7 + Rs, inoculated with F7 and Rs. The wilt incidence was calculated on the 30 th day after transplanting as described in Materials and Methods. The data are expressed as the mean ± SD (n = 24). Significant differences between the control and treated group (Y6 or F7) are determined by Tukey's Studentized Range (HSD) Test as indicated: α = 0.05, n = 24. (C) Representative photographs of the tomato plants to show typical wilt symptoms in CK group while only minorto-negligible symptoms in the two treated groups after 25 days of transplanting in the spring pot experiments. The methanol extracts from the two isolates showed strong antibacterial activity against R. solanacearum, as judged by an agar diffusion assay. Five µl of the methanol extracts of Y6, F7, B. megaterium X3 (X3; control), and methanol only (control) were tested. The latter two served as controls and did not show any inhibitory effects. The clearance zone indicated by the black lines is evident when treated with the methanol extract from either isolate (Y6, 14 ± 0.1 mm; F7, 13 ± 0.1 mm; mean ± SD with n = 3). The scale bar is 5 mm. (B) Identification of LP compounds from the methanol extract of Y6 or F7 using UPLC-MS analysis. Mass spectrum [M+H] + of LPs are shown, surfactin ((i, Y6) and (iv, F7)), fengycin((ii, Y6) and (v, F7)), and iturin((iii, Y6) and (vi, F7)). (C) Representative chromatograms of the LPs (iturin, fengycin and surfactin) in methanol extract (3 μl) from strain Y6 or F7 using UPLC-MS analysis. (D) Quantification of the LP compounds (surfactin, iturn and fengycin) secreted by Y6 or F7. Cell were grown in LB medium at 30 °C for 48 h, harvested by centrifugation, and subjected to methanol extraction for LP analysis. The LPs were then quantified based on the standard curves of commercial standards surfactin, iturin and fengycin. Significant differences between Y6 and F7 were determined by two-tailed t-test as indicated: **P < 0.01, n = 3.  1A) and quantified by UPLC-MS analysis as described in Materials and Methods. Briefly, after the inhibition zone against R. solanacearum (Rs) became evident after 24 h incubation at 30 °C, a 300 mg of agar sample was harvested from the inhibition zone, mixed with 1 ml of acetonitrile/water (1:1 v/v), and sonicated for 30 s followed by centrifugation and filtration. The supernatant was collected as the acetonitrile/ water extract and subjected to UPLS-MS analysis. The control was performed in a similar way but without the pathogen Rs in the top soft agar layer. (A) 300 mg of agar sample was collected around the well (2-4 mm). The two isolates produce more LPs, particularly fengycin and iturin, when they encounter the pathogen R. solanacearum. B. Representative chromatograms illustrate the differences in LP production by strain Y6 or F7 in the presence or absence of the co-culturing pathogen R. solanacearum. (C) The presence of R. solanacearum affects expression of LP biosynthesis genes. Expression of the LP biosynthesis genes (srfAB, ituC and fenD for synthesis of surfactin, iturin, and fengycin, respectively) was monitored in the two isolates in the presence or absence of the co-culturing pathogen Rs. RNA extraction and qPCR were performed as described in Materials and Methods. The relative gene expression (in fold change) was calculated using the 2 −ΔΔCt method. For (A,C), significant differences between the Rs infected and control groups were determined by two-tailed t-test as indicated: **P < 0.01, *P < 0.05, n = 3.  Table 1) and further confirm that interaction with R. solanacearum induces expression of LP genes resulting in stimulated LP production.

Iturin and fengycin play a redundant role in antagonism against R. solanacearum. Other com-
pounds in addition to the LPs produced by Y6 and F7 may contribute to antagonistic activity against phytopathogens. Moreover, it is difficult to accurately assess the level of LP production and activity in the soil environment. Therefore, to dissect the contribution of each individual LP compound to antimicrobial activity, we constructed mutants lacking each LP individually or in combination by disrupting LP biosynthesis genes required for iturin, fengycin and surfactin production (ituA, fenC, and srfAA, respectively) in Y6. Y6 was chosen for further study because F7 is naturally resistant to some antibiotics such as spectinomycin, which makes mutant construction more difficult. Loss of any individual LP (ituA, fenC, or srfAA) did not affect the capacity of Y6 to suppress R. solanacearum growth on CPG plates, whereas the ituA fenC double mutant completely lost the ability to inhibit R. solanacearum growth (Fig. 4). These results suggest that iturin and fengycin are the two major antimicrobial compounds and are functionally redundant in suppression of R. solanacearum.
Iturin and fengycin have been proposed to have synergistic antagonistic activity against phytopathogens 9 . However, neither ituA nor fenC single mutants showed reduced antibacterial activity against R. solanacearum, suggesting either a lack of synergy in this case or that deficiency in biosynthesis of one LP might be compensated for by increase in production of another LP upon encountering pathogens. Indeed, quantification of LP production in wild-type (Y6), ituA, and fenC single mutants revealed that iturin production in the fenC mutant increased 2-fold ( Fig. S3B,C). To test if the crosstalk between LPs occurs at the transcriptional level, we monitored Total LPs 502 ± 54 36 ± 14 220 ± 24 22 ± 7 Table 1. LP production is significantly stimulated during interaction with R. solanacearum (RS). Note: 300 mg of agar sample was harvested from the inhibition zone (Fig. 1A), mixed with 1 ml of acetonitrile/water (1:1 v/v), and sonicated for 30 s followed by centrifugation, filtration, and UPLC-MS analysis. Commercial LP standards were used to generate standard curves for LP quantification. Data are expressed as the mean ± SD (n = 3). Ratio is indicated as the percentage of each individual isoform out of the total LP compound. the expression of ituC and fenD genes in wild-type (Y6), ituA, and fenC single mutants by qPCR during co-culture with R. solanacearum. Expression of fenD in the ituA mutant was upregulated about 2-fold compared to wild-type, and expression of ituC in the fenC mutant also increased about 2-fold (Fig. S3A). These results suggest that interactions between the LP biosynthesis pathways at the transcriptional level can allow one LP to compensate, at least in part, for the defective production of another.

Iturin is the primary factor responsible for antifungal activity against Foc in vitro and in vivo.
PGPR can also provide protection from fungal pathogens 8,32 . To evaluate the antifungal ability of these two isolates, we tested their antagonistic activity against several common fungal plant pathogens using a plate confrontation assay. Both Y6 and F7 significantly inhibited the growth of Fusarium oxysporum f.sp. cubense (Foc), F. oxysporum f.sp. cucumerinum, and Colletotrichum gloeosporioides on Potato-Dextrose-Agar (PDA) plates (Fig. S4). We then assessed the contribution of iturin and fengycin to antifungal activity using an optimized spot-on-lawn assay (Fig. 5A), which is more sensitive compared to the traditional plate confrontation assay (see Materials and Methods). The fenC single mutant did not affect the capacity of Y6 to suppress the wildgrowth of the fungal pathogen Foc (Fig. 5A). Surprisingly, both the ituA single mutant and the ituA fenC double mutant almost completely lost the ability to inhibit Foc growth, although there is a sensitivity zone with reduced cell growth around the well where the ituA single mutant was inoculated (Fig. 5A). Similar results were observed in a plate confrontation test (Fig. S5).
In addition to their fungicidal activity, LPs are known to inhibit fungal growth by inhibiting spore germination. We conducted a spore germination assay using LP commercial standards and Y6 extract. As expected, the Y6 extract showed strong antagonistic activity against Foc and 75 μg ml −1 of the LP extract prevented ~60% of the spores from germinating. A similar level of inhibition was observed for the iturin standard (Fig. S6). By contrast, fengycin only showed minor antifungal activity even with the highest concentration tested (200 μg ml −1 ) and surfactin showed no evident inhibitory activity (Fig. S6). These results indicate that iturin is the major LP responsible for antifungal activity in vitro and fengycin plays a minor role.
To further investigate the biocontrol activity of Y6 and its derived LP-defective mutants against banana Fusarium wilt, we set up banana pot experiments in greenhouse settings. Treatment with WT (Y6) significantly reduced the wilt incidence, 48% by day 30 after transplanting compared to 86% in the untreated group (Fig. 5B-D). Deletion of fenC did not significantly affect the wilt incidence (51% by day 30). By contrast, the ituA single and ituA fenC double mutants only retained minor suppression ability (Fig. 5B-D). The biocontrol efficacy of Y6 was 44% against banana Fusarium wilt and iturin alone contributes about 90% of the total biocontrol efficacy. These results suggest that iturin is the major contributor against Foc in vivo under controlled-environmental conditions.

Either iturin or fengycin is sufficient for biofilm formation. Bacteria often exist in the environment
as cell communities called biofilms, which are required to effectively colonize plant roots and protect cells against unfavorable conditions 37,38 . Thus, biofilm formation is a prerequisite for effective Bacillus PGPR activity 22,39 . Surfactin is well known for its role in biofilm formation in the phylloplane and rhizosphere 22,40,41 . However, the contribution of the other LPs to biofilm formation is unclear. Therefore, we assessed the contribution of all three LPs produced by Y6 to this multicellular process by monitoring pellicle formation and colony morphology in biofilm stimulating media (LBGM and MSgg), where no evident growth defects were observed for all the strains tested (Fig. S7). As expected, WT (Y6) forms robust, wrinkled pellicles on LBGM (Fig. 6A) and MSgg (Fig. 6C) liquid medium. WT also forms complex colony patterns on both LBGM (Fig. 6B) and MSgg plates (Fig. 6D). The srfAA single mutant, which is defective in surfactin production, showed apparent defects in colony morphology on both LBGM and MSgg plates, but no noticeable defects in pellicle formation (Fig. 6). Mutation of ituA or fenC alone has no evident effects in biofilm formation. By contrast, the ituA fenC double mutant displays a flat and non-wrinkled biofilm architecture and an unstructured pellicle in comparison to the wild-type strain (Fig. 6), indicating iturin and fengycin, like surfactin, play an important but redundant role in facilitating biofilm formation.
Either iturin or fengycin is sufficient for cell motility. Many bacteria move on surfaces towards nutrient-rich niches. In the rhizosphere, the ability of Bacillus to sense and move towards secreted plant compounds is critical for root colonization 42 . LPs and biosurfactants such as surfactin play important roles in the motility of some Pseudomonas and Bacillus isolates 43 . To understand the contribution of LPs to cell motility, we tested the surface motility of WT (Y6) and its derived mutants on LBGM medium plates prepared with varying levels of agar (Fig. S7). On "swim plates" (LBGM plates containing 0.3% agar) 44 , the wild-type (Y6) strain moved rapidly and completely spread over the whole plate within 4 h with a diameter of 85 mm. The ituA or fenC single mutants showed moderate defects in swimming with a diameter of 73 or 80 mm, respectively (Fig. 7A). In contrast, the ituA fenC double mutant almost completely lost its ability to swim with a diameter of 22 mm by 4 h, very similar to that of a surfactin-deficient strain (srfAA) (Fig. 7A). On plates to assess swarming motility (LBGM plates containing 0.7% agar) 44 , WT or either single mutant (ituA or fenC) was able to completely colonize the surface with a diameter of 85 mm by 15 h, whereas the ituA fenC double mutant showed a significant defect in swarming with a diameter of 60 mm by 15 h (Fig. 7B,C), close to that of srfAA mutant (50 mm in diameter by 15 h). These results suggest that in addition to surfactin, iturin and fengycin play important roles in surface movement, and either one is sufficient to facilitate cell motility.

Discussion
Tomato bacterial wilt and banana Fusarium wilt are two destructive diseases to crops. Most chemical controls are ineffective against these diseases. Some lipopeptide (LP)-antibiotic producing Bacillus species offer a promising strategy in plant disease management. Here, we isolated two rhizosphere-associated B. velezensis strains (Y6 and SCiEntifiC REPORTs | (2018) 8:4360 | DOI:10.1038/s41598-018-22782-z F7) that exhibit potent antagonistic activity against bacterial and fungal pathogens in both laboratory and greenhouse settings. We demonstrated that the two LPs (iturin and fengycin) secreted by these two isolates are, chiefly if not solely, responsible for their antimicrobial activity. We then observed that expression of LP biosynthesis genes is upregulated and production of LPs is greatly stimulated in the presence of R. solanacearum. Modulation of LP production by pathogenic organisms may be a common natural phenomenon 32,45,46 . Indeed, the expression of LP genes including ituC, fenD, and srfAA in B. amyloliquefaciens D29 is significantly upregulated during interaction with R. solanacearum 45 and in B. amyloliquefaciens Bk7 in response to Pseudomonas fuscovaginae 46 . Production of iturin and fengycin is strongly enhanced in B. subtilis 98S encountering fungal phytopathogens 32 . Interestingly, only fenC but no other LP genes is upregulated by the fungal pathogen Rhizomucor variabilis 47 . Thus, modulation of LP production might be a species-specific response between antagonist and pathogen. Although the triggering mechanism is largely unknown, it may involve the DegS/U two component sensing system, which is involved in The relative contribution of different LPs to antimicrobial activity may be dependent on the species of plant pathogen encountered. Our results showed that iturin and fengycin are functionally redundant in antagonism against R. solanacearum. By contrast, we observed that iturin, not fengycin, produced by Y6 plays the primary role, both in vitro and in vivo, in defense against the fungal pathogen Foc. Similar observations have been reported previously in other Bacillus isolates. In B. velezensis FZB42 and SQR9, bacillomycin D (a LP antibiotic related to iturin) is major contributor to antifungal activity against Foc while fengycin plays a minor role 9,12 . Fengycin has been observed to play an important role in antagonism against various fungal phytopathogens in B. subtilis UMAF6639 and EA-CB0015 50,51 . Additionally, in B. subtilis UMAF6639, iturin but not fengycin inhibits the pathogens Pectobacterium carotovorum and Xanthomonas campestris 31 .
Surfactin has been shown to have antagonistic activity in vitro and play protective roles in the rhizosphere 22,39,40 . We investigated the contribution of surfactin to the protective effects of the isolate Y6. Our results show that surfactin plays a minor role in antagonism against R. solanacearum (Fig. 4). We observed that fengycin and iturin production by Y6 is dramatically induced during interaction with R. solanacearum while surfactin levels remain low (Figs 2 and 3). Furthermore, surfactin is the least active antibacterial compound among all three LPs tested (Fig. S2).
The contribution of LPs to biocontrol outside of their direct antimicrobial activity is not well understood. We constructed LP deficient mutants and dissected the contribution of each individual LP to different biological processes. Surfactin may contribute to traits required for effective rhizosphere colonization such as biofilm formation and cell motility as reported previously 39,40,52,53 . Indeed, surfactin is important for biofilm formation, particularly on a solid surface (Fig. 6B,D), and is critical for cell motility (Fig. 7). Our results also show that iturin and fengycin contribute to biofilm formation and motility, as observed in B. subtilis 916 48 , B. velezensis LL3 and SQR9 12,54 . Thus, we propose that the biocontrol activity of Y6 is the result of coordinated actions of all three LPs.
Collectively, this study demonstrates that two LP-producing B. velezensis isolates (Y6 and F7) have potent antagonistic activity against R. solanacearum and Foc under both laboratory and greenhouse conditions. LP production in these two isolates is strongly stimulated in the presence of the pathogenic strain R. solanacearum and is critical for antagonism and colonization in controlled-environmental conditions. Most importantly, these two isolates, as potential biocontrol agents, may provide an effective strategy to combat plant pathogens in the target environment.

Materials and Methods
Isolation of antagonistic Bacillus species. Samples were collected from the rhizosphere soil of tomato plants in Yuejin Farm, Guangzhou, China (23° 08′N 113° 16′E). Bacillus spp. were isolated as reported previously 55 . Briefly, a soil sample (10 g) was shaken in 90 ml of sterilized water for 30 min, heated for 30 min at 80 °C, serially diluted and then spread over lysogeny broth (LB) 56 plates. Single bacterial colonies were streaked onto fresh LB plates after 48 h of incubation at 30 °C. Frozen stocks of the purified colonies were prepared using 15% glycerol and kept at −80 °C for further study.
Antibacterial activity test. The antibacterial activity of these isolates and their derived mutants against R. solanacearum GMI1000 57,58 was evaluated using an optimized spot-on-lawn assay. 200 μl of R. solanacearum cell culture (OD 600 ~0.4) grown in Casamino acid-Peptone-Glucose (CPG, 0.1% peptone, 0.01% casamino acids, 0.05% glucose) was mixed with 4 ml of 0.7% CPG soft agar and directly poured onto a CPG plate (1.5% agar). Plates were dried for 50 min and a well (5 mm in diameter) was made in the center of each plate, and 50 μl of each Bacillus strain (OD 600~0 .4) grown in LB medium was added into each well. Antagonistic activity was evaluated based on the inhibition zone after 24 h incubation at 30 °C. Antibacterial activity of the methanol extracts of Y6 and F7 was tested using an agar diffusion assay. Briefly, 200 μl of R. solanacearum CPG culture (OD 600 ~0.4) was mixed with 4 ml of 0.7% CPG soft agar and directly poured onto a CPG plate (1.5% agar), and 5 µl of methanol extract was added into each well (5 mm in diameter). The methanol extracts (5 µl) of B. megaterium X3 (Genbank accession number KJ526881) and methanol alone were used as controls. The inhibition zone was measured after 24 h of incubation at 30 °C. The experiments were performed at least three times.
Antifungal activity test. We performed two assays, a plate confrontation assay and a spot-on-lawn assay, to test activity of the isolates (Y6 and F7) and derived mutants against common fungal pathogens including Fusarium oxysporum f.sp. cubense 4 strain XJZ2 (Foc4, GenBank accession number JX090598) 59 , F. oxysporum f.sp. cucumerinum (isolated from cucumerium rhizosphere, no accession number available yet), and Colletotrichum gloeosporioides (Genbank accession number MWUF00000000) 60 . In the plate confrontation assay, the fungi were cultivated on potato-dextrose-agar plates (PDA, 20% potato infusion, 2% dextrose, and 1.5% agar) at 30 °C for 5 days. A 5-mm-diameter block of mycelium agar was cut and transferred into the center of a fresh PDA plate. After 1 day of incubation, 3 µl of Bacillus spp. cells (F7, Y6, or mutants derived from Y6) (OD 600 ~0.4) grown in LB medium was spotted on the PDA plate 2.5 cm away from the center, where the mycelium agar block was placed. The antifungal activity was evaluated by measuring the diameter of the inhibition zone (the distance between the Foc mycelium and the bacterial colony) after 7 days of incubation at 30 °C.
We also developed a spot-on-lawn assay, which is more sensitive compared to the plate confrontation assay and only requires one-day incubation instead of seven days. Briefly, the fungal hyphae Foc4 were streaked and inoculated with 5 ml of PDA broth. After 2-day incubation at 30 °C with 180 rpm shaking, 50 μl of fungal culture was re-inoculated into 5 ml of fresh PDA broth, incubated for additional 12 h, filtered using a cheesecloth to remove hyphae, and then the concentration of spores was determined using a hemocytometer under a light microscope. Fifty µl of a spore suspension (~10 6 spores ml −1 ) was mixed with 4 ml of 0.7% soft PDA agar and directly poured onto a PDA plate (1.5% agar). Plates were dried for 50 min and a well (5 mm in diameter) was made in the center of each plate, and 50 µl of Bacillus spp. cells (OD 600 ~0.4; F7, Y6, or mutants derived from Y6) grown in LB medium was added into each well. The antifungal activity was evaluated by measuring the diameter of the inhibition zone (mm) after 24 h of incubation at 30 °C. The experiments were performed at least three times.
Lipopeptide (LP) extraction from the isolated strains. LPs were extracted according to previous protocol 61 . Briefly, 150 μl of Y6 or F7 cell culture (OD 600 ~0.4) was added into 15 ml of fresh LB medium. After 2 days of incubation at 30 °C with 180 rpm shaking, 10 ml of cell culture was spun down by centrifugation for 5 min at 14000 × g, and the cell-free supernatant was collected and subjected to filtration using 0.22-μm filters, and then loaded into a C 18 syringe cartridge column (Bond Elut C 18, Agilent, USA). The column was washed with 10 ml of ddH 2 O and the LP compounds were eluted with 2 ml of methanol. To understand the influence of pathogen on LP production, we extract LPs from the inhibition zone (Fig. 1A) 32 . 200 μl of R. solanacearum cell culture (OD 600 ~0.4) grown in CPG medium was mixed with 4 ml of 0.7% CPG soft agar and directly poured onto a CPG plate (1.5% agar). Plates were dried for 50 min and a well (5 mm in diameter) was made in the center of each plate, and 50 μl of Bacillus isolates (Y6 or F7) (OD 600 ~0.4) grown in LB medium was added into the well. After the inhibition zone became evident with 24 h incubation at 30 °C, a 300 mg of agar sample was harvested from the inhibition zone, mixed with 1 ml of acetonitrile/water (1:1 v/v), and sonicated for 30 s followed by centrifugation and filtration. The supernatant was collected from acetonitrile/water extract 32 . As a control, LPs were isolated in a similar way but without R. solanacearum in the top soft agar layer. Agar sample (300 mg) was collected around the well (2-4 mm).
Identification and quantification of LPs by UPLC-MS. Both methanol and acetonitrile/water extracts were analyzed by reverse phase Ultra-Performance Liquid Chromatography coupled with a triple quadrupole MS (UPLC-MS) (Waters, Acquity, XEVO-TQD). The lipopeptide compounds were identified based on the mass-to-charge ratio (m/z) and quantitated based on the standard curves generated using commercial LP standards (Sigma-Aldrich, USA). The column temperature was maintained at 40 °C and a gradient elution with (A) acetonitrile (containing 0.1% formic acid) and (B) water (containing 0.1% formic acid) was used. The gradient program was used as follows: 0-0.5 min, 40% A; 0.5-3.5 min, 40-80% A; 3.5-4.0 min, 80% A; 4.0-6.0 min, 80-95% A; 6.0-7.0 min, 95-98% A. The flow rate was set at 0.4 ml min −1 . The Electrospray Ionization (ESI) source was set in positive ionization mode with 3.26 kV of capillary voltage, and the source temperature was maintained at 150 °C. Nitrogen flow was 600 L h −1 . Argon flow was 50 L h −1 .
Swarming and swimming motility assays. Swimming and swarming motility of Y6 and its derived mutants were tested using standard protocols with minor modification 63,64 . LBGM plates containing 0.7% (swarming) or 0.3% agar (swimming) were dried in a laminar flow hood for 30 min and then 5 μl of LB precultures (OD 600 ~0.4) was spotted on the center of each plate. The plates were then dried for another 15 min and incubated overnight at 37 °C.

Biofilm formation assay.
To evaluate the contribution of the LP compounds to biofilm formation, we monitored colony morphology on LBGM and MSgg plates and pellicle formation on LBGM and MSgg liquid media. For colony morphology analysis, 3 μl of LB preculture (OD 600 ~0.4) was spotted onto LBGM or MSgg agar plates, which were dried for 30 min in a laminar airflow prior to spotting, and incubated at 30 °C for up to 36 h. To monitor pellicle formation 65 , 10 μl of LB precultures (OD 600 ~0.4) was inoculated into 2.5 ml of LBGM or MSgg medium in a 24 well plate and incubated at 30 °C for up to 36 h 65,66 .
Additional methods. Additional methods are described in the Supplementary Information, including plant pot experiments, identification of the isolates Y6 and F7 by phylogenetic analysis, minimal inhibition concentrations (MIC) determination of LPs and the extracts from the isolates, RNA extraction and quantitative PCR (qPCR), growth curves, and spore germination assay. Data availability. All data generated during this study are included in this article (and its Supplementary Information file).