Comprehensive genomic analysis of Bacillus subtilis 9407 reveals its biocontrol potential against bacterial fruit blotch

Bacillus subtilis, a plant-beneficial bacterial species exhibiting good biocontrol capabilities, has been widely used in agricultural production. The endophytic strain 9407 can efficiently control bacterial fruit blotch (BFB) caused by the gram-negative bacterium Acidovorax citrulli. However, the mechanism underlying its biocontrol ability remains poorly understood. Given the genomic diversity of B. subtilis, strain 9407 was sequenced and assembled in this study to determine the genome information associated with its biocontrol traits. A combination of core genome phylogenetic analysis and average nucleotide identity (ANI) analysis demonstrated that the 9407 strain belonged to B. subtilis. Various functional genes related to biocontrol traits, i.e., biofilm formation, motility, pathogen inhibition, plant growth promotion, and induction of systemic resistance, were identified in B. subtilis 9407. Four secondary metabolite biosynthesis gene clusters with antibacterial ability were also found in the B. subtilis 9407 genome, including newly identified subtilosin A, bacilysin, and bacillaene, and the previously reported surfactin. Mutants lacking sboA or bacG, which are defective in synthesizing subtilosin A or bacilysin, showed decreased inhibitory activity against A. citrulli MH21, and the triple mutant with deleted sboA, bacG, and srfAB almost completely lost its inhibitory activity. The biofilm formation and swarming motility of the sboA and bacG mutants also decreased, in turn leading to decreased colonization on melon roots and leaves. Under greenhouse conditions, the biocontrol efficacy of the sboA and bacG mutants against BFB on melon leaves decreased by 21.4 and 32.3%, respectively. Here, we report a new biocontrol pathway of B. subtilis 9407 against BFB, in which subtilosin A and bacilysin contributed to the biocontrol efficacy by improving antibacterial activity and colonization ability of the strain. The comprehensive genomic analysis of B. subtilis 9407 improves our understanding of the biocontrol mechanisms of B. subtilis, providing support for further research of its biocontrol mechanisms and field applications.


Background
Bacillus subtilis has been widely used in agricultural production due to its environmental safety, straightforward industrial production, and good biocontrol efficacy (Wang et al. 2020). The development of genome sequencing technology and bioinformatic analysis have made it convenient to obtain genomic information on B. subtilis, allowing us to comprehensively understand its life activities (Moszer 1998). Genomic comparison of different B. subtilis strains can elucidate the genetic variation, evolutionary classification, and genomic diversity of this species (Rahimi et al. 2018). To date, the genome sequences of 389 strains of B. subtilis have been deposited in the National Center for Biotechnology Information (NCBI) genome assembly database. These genomic data have revealed important information about the development, sporulation, and metabolism of B. subtilis strains (Kunst et al. 1997), and can be further used to obtain new information related to their biocontrol traits (Sulthana et al. 2019;Franco-Sierra et al. 2020). In particular, the genomic sequence of the well-known model organism B. subtilis strain 168 has allowed the formulation of useful inferences for studies of other B. subtilis strains (Barbe et al. 2009).
Accumulating evidence indicates that B. subtilis possesses biocontrol traits, including plant colonization, pathogen inhibition, and plant growth promotion abilities, and activation of induced systemic resistance (Hashem et al. 2019). In B. subtilis, motility towards plant roots and biofilm formation on the root surface are crucial for its colonization of plant roots and biocontrol efficacy against plant pathogens (Gao et al. 2013;Allard-Massicotte et al. 2016;Al-Ali et al. 2018). The production of active substances is an important indicator for assessing the biocontrol efficacy of a beneficial strain (Zeriouh et al. 2014;Gao et al. 2016). B. subtilis produces various substances with broadspectrum antibacterial activity, including lipopeptide antibiotics, bacteriocins, and antibacterial proteins (Stein 2005). Importantly, it has been suggested that different antibacterial substances can act synergistically to inhibit phytopathogen growth (Koumoutsi et al. 2004;Alanjary and Medema 2018). Furthermore, the production of phytohormones, siderophores, lipopeptides, volatile compounds, and phytases allows B. subtilis to promote plant growth and induce plant immune responses (Franco-Sierra et al. 2020).
B. subtilis strain 9407, isolated from healthy apple fruit, has exhibited broad-spectrum antimicrobial activities (Fan et al. 2017a). Previously, we found that B. subtilis 9407 controls bacterial fruit blotch (BFB) through surfactin-mediated antibacterial activity (Fan et al. 2017b). BFB is a serious melon disease caused by Acidovorax citrulli and poses a serious threat to the melon industry (Bahar et al. 2008;Adhikari et al. 2017;Rahimi-Midani and Choi 2020). It is characterized by symptoms such as water-soaked disease spots. The main control strategy against BFB is the application of antibiotics and chemicals; therefore, more effective and environmentally-friendly control strategies are urgently needed (Rahimi-Midani and Choi 2020). B. subtilis, as a well-known environmentally-friendly biocontrol bacterial species, can exert its biocontrol effects via several pathways (Fira et al. 2018;Hashem et al. 2019). Although we have confirmed that B. subtilis 9407 controls BFB by producing surfactin, its biocontrol mechanism against BFB remains poorly understood. The objective of this study was to shed light on the underlying biocontrol mechanism of B. subtilis 9407, especially those pathways that could effectively control BFB. We performed a comprehensive genome analysis of B. subtilis 9407 to reveal the biocontrol mechanism of this specific strain, and to determine which substances play a direct role against BFB. The results will contribute to the development of new biocontrol agents with original modes of action against specific plant diseases.

Results
General genome description of B. subtilis 9407 In this study, we sequenced the genome of B. subtilis 9407 to explore its biocontrol mechanism. Genomic assembly of B. subtilis 9407 produced 16 scaffolds, with an N 50 of 2,111,374 bp. The whole-genome sequence of B. subtilis 9407 is 4,062,615 bp in length with a G + C content of 43.7% (Table 1 and Additional file 1: Figure S1). The number of predicted protein-coding genes in B. subtilis 9407 is 4033. Among these, 2853 were assigned a putative function, and 1180 were predicted to encode hypothetical proteins. The protein-coding genes had an average length of 884 bp and accounted for 89.1% of the genomic sequence. A total of 79 tRNA-coding genes and 9 rRNA genes were predicted in the chromosome sequence.
Phylogenetic analysis of B. subtilis 9407 As a molecular marker, the 16S rRNA gene has been widely used for strain identification, but microbial taxonomies based on 16S rRNA gene relationships still have limitations, including low phylogenetic resolution. Phylogeny construction based on the core genome has progressed in recent years towards a standardized bacterial taxonomy (Parks et al. 2018). In this study, a phylogenetic tree was constructed using core genome analysis to understand the evolutionary relationships of B. subtilis strain 9407. A phylogenetic tree of 16 Bacillus genomes was constructed based on the concatenation of 662 single-copy core genes present in all genomes using the maximum likelihood (ML) method and rooted in Paenibacillus polymyxa M1. As shown in Fig. 1, B. subtilis 9407 is in the same clade with other B. subtilis strains and a sister group of B. subtilis SEM-9. Average nucleotide identity (ANI) is the average identity value calculated from a pair-wise comparison of homologous sequences between two genomes; this indicator is frequently used in species definition (Lee et al. 2016). In this study, we conducted a heatmap analysis based on the ANI values of different strains to confirm the findings of our phylogenetic analysis. The ANI values of representative Bacillus strains are summarized in Fig. 2. B. subtilis 9407 and other B. subtilis strains showed ANI values of > 98%, suggesting that they are the same species. A pan-genome analysis indicated that the selected B. subtilis strains contain 3153 common genes, comprising 92.7-99.3% of all genes (Fig. 3). To intuitively illustrate the results above, we performed a comparative genomic analysis using BLAST Ring Image Generator (BRIG) software to evaluate synteny between B. subtilis strains 9407 and 168 (Sulthana et al. 2019). The results showed that there was a high genomic similarity between these two strains, suggesting that their genetic information is very similar (Additional file 1: Figure S2).

Potential functional genes involved in biocontrol traits of B. subtilis 9407
Bacillus harbors various functional genes associated with biocontrol traits, thereby ensuring its biocontrol efficacy (Ashwini and Srividya 2014). In this study, potential functional genes related to biocontrol traits of B. subtilis 9407 were analyzed based on whole-genome annotation and pan-genome analysis results. The results showed that B. subtilis 9407 possesses several functional genes involved in biofilm formation, motility, pathogen inhibition, plant growth promotion, and induced systemic resistance, sharing 91-100% identity and 98-100% genome coverage with B. subtilis 168 (Additional file 2: Tables S1 and S2). Genes related to biofilm formation, Fig. 1 Phylogenetic tree showing the relationship of the sequenced Bacillus spp. strains. The tree was generated based on the 662 single-copy core genes using RAxML 8.2.10. P. polymyxa M1 was used as the out-group. Percent bootstrap values (from 100 replicates) are indicated at the nodes motility, and flagellum biosynthesis were found in B. subtilis 9407, including biofilm synthetic genes (eps operon, tapA-sipW-tasA operon, blsA, pgs operon), regulatory genes (spo0A, abrB, sinR, sinI, etc.), and flagellum biosynthesis genes (cheY, motA, motB, flg and fli operon), which are well known to be involved in colonization ability. Several genes were found to be involved in the production of volatile organic compounds (VOCs), phytohormones, and siderophore bacillibactin, suggesting that B. subtilis 9407 has the potential to promote plant growth. For example, ysnE, ywkB, phyC, and dhb operon are involved in the synthesis of indole-3acetic acid (IAA), auxin, phytase, bacillibactin, respectively. Various genes encoding proteins associated with induced plant systemic resistance, i.e., srf operon, als operon, and bdhA encoding surfactin, acetoin, and 2, 3butanediol, respectively, were detected in B. subtillis 9407. TasA and lipopeptide surfactin are also well known for their antimicrobial activity. Synthetic genes of other antibacterial and antifungal substances were also found, such as pps operon encoding fengycin, bac encoding bacilysin, sboA encoding subtilosin A, and pks encoding bacillaene. These findings demonstrate that B.
subtilis 9407 has the potential to colonize plants, inhibit pathogens, promote plant growth, and induce plant systemic resistance.

Prediction of biosynthesis gene clusters in the genome sequence of B. subtillis 9407
Various widely reported secondary metabolites produced by B. subtilis are beneficial to the survival of this bacterial species in the complex and changeable natural environment (Stein 2005). In this study, six biosynthesis gene clusters (BGCs) were found in the genome of B. subtilis 9407, including four nonribosomal peptide synthetases (bacillibactin, bacilysin, fengycin, and surfactin), one trans-acyl transferase polyketide synthetase (bacillaene), and one sactipeptide (subtilosin A), all sharing a high degree of sequence similarity with those of B. subtilis 168 (Table 2 and Additional file 1: Figure S3). The amino acid sequence identity of each gene in the BGCs between B. subtilis strains 9407 and 168 was 95-100%. However, bacE within the bacilysin cluster in B. subtilis 9407 seemed to be partially missing, and an additional gene of unknown function was found in the bacillaene cluster of B. subtilis 9407, but not in that of B. subtilis 168. Whether these differences were caused by sequencing and assembling errors or natural variation, and whether they can influence the production of substances related to the life activity of B. subtilis 9407 remains unknown. Subtilosin A, bacilysin, bacillaene, and surfactin have been reported to possess antibacterial activity, suggesting that multiple substances may be involved in the antagonism of B. subtilis 9407 to A. citrulli MH21.
Subtilosin A and bacilysin participate in biocontrol efficacy of B. subtilis 9407 against BFB The synthetic genes srfAB, sboA, and bacG are essential for the synthesis of surfactin, subtilosin A, and bacilysin, respectively (Zheng et al. 2000;Stein 2005;Rajavel et al. 2013). Our previous studies showed that a lack of srfAB decreases surfactin production and colonization ability of B. subtilis 9407, thereby decreasing its biocontrol efficacy against BFB (Fan et al. 2017b). To verify whether   . 4), suggesting that subtilosin A, bacilysin, and surfactin have a synergistic effect in inhibiting A. citrulli MH21. Compared with the wild-type strain, the sboA and bacG mutants simultaneously showed weaker biofilm formation (Fig. 5a, b) and swarming ability (Fig. 5c (Table 3 and Fig. 6). In summary, subtilosin A and bacilysin affected the biocontrol efficacy of B. subtilis 9407 against BFB by influencing its antibacterial activity and colonization ability.

Discussion
Microbial biocontrol strategies against BFB have been widely reported. For instance, B. subtilis R14, B. megaterium pv. cerealis RAB7, B. pumilus C116, and Bacillus sp. MEN2 show antibiosis against A. citrulli by producing bioactive compounds that are partially characterized as lipopeptides (Santos et al. 2006). B. amyloliquefaciens 54 significantly controls BFB by increasing the expression of an important defense-related gene, PR1 (Jiang et al. 2015). Bacteriophages effectively control BFB by translocating from soil to leaf tissue and killing A. citrulli (Rahimi-Midani and Choi 2020). Bacillus strains exert biocontrol efficacy through several biocontrol mechanisms (Fira et al. 2018;Hashem et al. 2019). In Fig. 4 Antimicrobial activity of the sboA and bacG mutants against A. citrulli MH21. a The antimicrobial activity of the sboA and bacG mutants was decreased compared with that of the wild-type B. subtilis 9407 strain. b The inhibition zone of B. subtilis 9407, the sboA and bacG mutants against A. citrulli MH21. Antibacterial activity was indicated by the diameter of the inhibition halo minus the diameter of the colony. The statistical analysis was performed using GraphPad Prism 7 software by one-way ANOVA with the Dunnett test (P < 0.05). The variation was recorded as mean ± SE (****: P < 0.0001, ***: P < 0.001, **: P < 0.01, *: P < 0.05) this study, we report that B. subtilis 9407 has the potential to colonize plants, inhibit pathogens, promote plant growth, and induce plant systemic resistance; it also produces bacilysin and subtilosin A in addition to the previously reported surfactin, all of which are active against BFB.
Many studies have shown that swarming motility helps bacteria to migrate to plant roots (Allard-Massicotte et al. 2016;Gao et al. 2016) and that biofilms contribute to plant root colonization (Verstraeten et al. 2008).
Swarming motility relies on a swinging flagellum, encoded by the fli and flg operons (Kearns 2010). The eps and tapA-sipW-tasA operons are responsible for synthesizing the main biofilm components, and spo0A, sinI, abrB, sinR are the main regulatory genes (Verstraeten et al. 2008). The genes alb, bac, pks, pps, and srf are responsible for the synthesis of antimicrobial substances subtilosin A, bacilysin, bacillaene, fengycin, and surfactin, respectively (Stein 2005;Moldenhauer et al. 2007;Amrouche et al. 2010;Rajavel et al. 2013). The Fig. 5 The impact of deleting sboA or bacG on colonization ability of the B. subtilis 9407 strain. a, b Biofilm formation assays. Biofilm assays were detected in Msgg plate incubated at 28°C for 96 h. c, d Swarming ability assays. Swarming assays were detected in LA plate (containing 0.7% agar) incubated at 37°C for 6 h. After drying the water, the plates were incubated overnight at 37°C. e Colonization on the melon roots (R) and leaves (L) of B. subtilis 9407 and mutant strains. To facilitate the screening, all these strains were transformed into a plasmid pC-1 with a chloramphenicol resistance gene. The y-axis was log 10 CFU and the x-axis was the number of days. The statistical analysis was performed using GraphPad Prism 7 software by one-way ANOVA with the Dunnett test (P < 0.05). The variation was recorded as mean ± SE (****: P < 0.0001, ***: P < 0.001, **: P < 0.01, *: P < 0.05)  (Ongena et al. 2007;Wang et al. 2020). The genes related to plant growth promotion, such as ysnE, ywkB, phyC, dhb operon, and bltD, synthesize IAA, auxin, phytase, siderophore, and spermidine, respectively (Kerovuo et al. 1998;May et al. 2001;Baichoo et al. 2002;Quentin et al. 2002). The presence of these genes in B. subtilis 9407 implies its biocontrol potential for controlling BFB. The antibacterial compounds subtilosin A and bacilysin, which are produced by Bacillus strains, allows these strains to antagonize pathogens (Khochamit et al. 2015;Wu et al. 2015). B. subtilis KKU213 produces subtilosin A against various gram-positive bacteria and B. amyloliquefaciens FZB42 exhibits biocontrol activity against gram-negative bacterium Xanthomonas strains by producing bacilysin (Khochamit et al. 2015;Wu et al. 2015). Previously, we have determined that surfactin produced by B. subtilis 9407 is crucial for this strain to control BFB (Fan et al. 2017b). In this study, subtilosin A and bacilysin BGCs were identified in the genome of B. subtilis 9407. To verify whether these two compounds play a role in the B. subtilis 9407-mediated control of BFB, we conducted verification tests. We found that the loss of sboA or bacG in B. subtilis 9407 decreased its biocontrol efficacy against BFB by affecting both of its inhibitory activity against A. citrulli MH21 and ability to colonize plant tissues. Subtilosin A, bacilysin, and surfactin showed a synergistic effect on the inhibition of A. citrulli MH21. However, the absence of sboA or bacG slowed the swarming motility and biofilm formation of these strains without affecting their growth (Additional file 1: Figure S4). Surfactin triggers biofilm formation and plant root colonization in B. subtilis, which is crucial for its biocontrol efficacy (Zeriouh et al. 2014). However, there have been no reports of the effects of subtilosin A and bacilysin on biofilm formation and swarming motility. Based on the results of the present study, we speculate that subtilosin A and bacilysin may have similar functions with surfactin.

Conclusions
Previous studies showed that B. subtilis 9407 produces surfactin against BFB. Whether B. subtilis 9407 possesses other pathways to control BFB remains unknown at present. This study is the first to report that B. subtilis 9407 can control BFB by producing subtilosin A and bacilysin. Subtilosin A and bacilysin contributed to the biocontrol efficacy of B. subtilis 9407 against BFB through their antibacterial activities and plant colonization abilities. Comprehensive genomic analysis of B. subtilis 9407 suggests that this strain still has unrevealed biocontrol mechanism, highlighting its potential as a biocontrol agent. Further research on the biocontrol mechanisms of this beneficial strain will aid the development of biocontrol agents for specific plant diseases.

Bacterial growth and construction of mutant strains
A list containing all strains and plasmids used in this study is presented in Additional file 2: Table S1. A. citrulli MH21 was incubated at 28°C, 200 rpm in LB broth containing 100 μg/mL of ampicillin.
Deletion mutants of B. subtilis were constructed by homologous recombination and screened on LB plates with erythromycin (5 μg/mL), chloramphenicol (5 μg/ mL), or kanamycin (20 μg/mL). Briefly, the plasmid containing the resistance cassette flanked by 1 kb DNA sequences corresponding to the upstream and downstream regions of the target genes was cloned into E. coli Fig. 6 Biocontrol efficacy of the sboA and bacG mutants against BFB under greenhouse conditions. Blank: PBS buffer was used to immerse the germinated melon seeds, and a saline solution (0.9% NaCl) was used to spray the seedlings; Control: PBS buffer was used to immerse the germinated seeds, and the seedlings were sprayed with A. citrulli; 9407: the germinated seeds were soaked in a suspension of B. subtilis 9407 and the seedlings were sprayed with A. citrulli; sboA: the germinated seeds were soaked in the sboA mutant and the seedlings were sprayed with A. citrulli; bacG: the germinated seeds were soaked in the bacG mutant and the seedlings were sprayed with A. citrulli DH5α, and then was introduced into E. coli EC135 without endogenous limiting modification system by chemical conversion method. Finally, the plasmids were introduced into the competent cells of B. subtilis 9407 by electroporation (1.8 kV, 200 Ω, 25 μF) with a time constant of 4.5 to 5.5 msec, and the mutants were obtained by screening with the corresponding antibiotic. In sboA mutant, the sboA coding sequence was replaced with a kanamycin resistance cassette. In bacG mutant, the bacG coding sequence was also replaced by a kanamycin resistance cassette. In srfABsboAbacG mutant, the srfAB, sboA, and bacG coding sequences were replaced by the tetracycline, kanamycin, and chloramphenicol resistance cassette, respectively. Transformants were verified by PCR amplification and DNA sequencing.

Genome sequencing, assembly, and annotation
The draft sequences of the Bacillus strains were produced by using Illumina paired-end sequencing technology at the company of BerryGenomics, Beijing. Assemblies were performed using SOAPdenovo v.2.04 (Luo et al. 2012), resulting in 16 scaffolds for 9407. Predictions of protein-coding genes were implemented using Prokka v.1.11 (Seemann 2014). Functional annotation was carried out using the Basic Local Alignment Search Tool (BLAST) against the Cluster of Orthologous Groups of proteins (COG), NCBI nr protein database, Kyoto Encyclopedia of Genes and Genomes (KEGG) database, and InterPro database. Ordering of contigs of the strain 9407 was achieved using the Java-based graphical interface program Mauve (Rissman et al. 2009). The genome sequence of B. subtilis 168 was used as a reference for the strain 9407. The final annotated chromosome was plotted using CIRCOS to show gene locations, GC-skew, and GC content (Krzywinski et al. 2009). A comparative circular genome map was constructed by BRIG v.0.95 to evaluate the synteny of the assembled genome of B. subtilis 9407 with that of B. subtilis 168 (Alikhan et al. 2011).
The whole-genome shotgun data of B. subtilis 9407 have been deposited at GenBank under the accession number PISO00000000.1. The genomic sequence of B. subtilis 168 was deposited under the accession number AL009126.3. All of the bacterial strains used in this study and their accession numbers in GenBank are listed in Additional file 2: Table S3.

Phylogenetic analysis of B. subtilis 9407
All genomes used in this analysis were downloaded in FASTA format from the NCBI database. A maximumlikelihood phylogenetic tree of Bacillus species was constructed based on 662 single-copy core proteins shared by 18 Bacillus genomes and the genome of Paenibacillus polymyxa M1 according to the following methods: (1) multiple alignments of amino acid sequences were carried out by MAFFT v.7.310 (Katoh and Standley 2013); (2) conserved blocks from multiple alignments of test protein were selected by using Gblocks (Castresana 2000); (3) ML tree was constructed using RAxML v.8.2.10 (Stamatakis 2014) software using the PROTGAMMALGX model with 100 bootstrap replicates. The tree was displayed by molecular evolutionary genetic analysis (MEGA) (Kumar et al. 2018). Then, ANI values between two genome sequences were calculated using the original ANI function of OrthoANI (Lee et al. 2016). The heat maps were generated using CIMminer (https://discover.nci.nih.gov/cimminer/) based on ANI values (Scherf et al. 2000). The pan-genomic analysis was performed by the PGAP analysis pipeline (Zhao et al. 2018).

Analysis of functional genes and secondary metabolite biosynthesis gene clusters
Amino acid sequence identity was compared by the Blastp program between genes of B. subtilis 9407 and 168. The BGCs were predicted using the antiSMASH bacterial v.5.1.2 (Blin et al. 2019) and further analyzed by the 2ndFIND (http:// biosyn.nih.go.jp/2ndfind/) program to confirm more accurate information of BGCs, which was performed via the Web servers with the default parameters.

In vitro antagonism test
Colonies of A. citrulli MH21 were inoculated into 5 mL LB broth and incubated on an orbital shaker (200 rpm) until an OD600 of 0.8 was reached, corresponding to a bacterial concentration of approximately 10 8 CFU/mL. The bacterial suspension was then added to melted and cooled LA medium, mixed and poured into Petri dishes (9 cm in diameter), and allowed to re-solidify. Colonies of B. subtilis were initially inoculated into 5 mL of LB broth and incubated at 37°C and 200 rpm, for 12 h. The bacteria were then adjusted to an OD600 of 0.8 (10 8 CFU/mL) using LB broth and 2 μL aliquots of this suspension were added to the surface of the above-prepared LB solid plates. Plates were cultured at 28°C for 5 d and the zone of inhibition was observed and measured. There were five repetitions for each tested B. subtilis strain. The experiments were repeated three times independently.

Biofilm formation assay
The biofilm formation assay was performed as previously described (Fan et al. 2017b). Colonies of B. subtilis were initially inoculated into 5 mL of MSgg liquid culture medium (5 mM potassium phosphate buffer pH 7, 100 mM Mops pH 7, 2 mM MgCl 2 , 700 μM CaCl 2 , 50 μM MnCl 2 , 50 μM FeCl 3 , 1 μM ZnCl 2 , 2 μM thiamine, 0.5% glycerol, 0.5% glutamate, 50 μg/mL tryptophan, 50 μg/ mL phenylalanine), and incubated at 28°C and 200 rpm until an OD600 of 0.8 was reached. Then, the bacterial suspension was inoculated into a 12-well microtiter plate (Corning) containing 4 mL MSgg liquid medium in each well. Four μL aliquots of the bacterial suspension were inoculated into each well. After inoculation, the microtiter plates were incubated statically at 28°C for 96 h. Photos were then taken to record the biofilm phenotypes of different strains. For each tested strain, three replicates were included.
The biofilm quantification assay was performed in 96-well polystyrene microplates (Corning) as described previously (Ma et al. 2017). Each well contains 150 μL aliquots of MSgg liquid medium, and eight independent replicated wells were used for each tested strain. Then 1.5 μL of bacterial suspension were inoculated into each well. After inoculation, the plates were incubated at 28°C for 96 h. Then, the bacterial cells were collected separately from each well and washed twice with 200 μL of sterile ddH 2 O, and stained with 200 μL of 0.1% (w/v) crystal violet (CV) solution. After staining for 15 min, the staining CV solution was removed, the bacterial cells were washed twice with 200 μL sterile ddH 2 O and then 200 μL ethanol was used to elute the CV. The optical density of the eluate was measured with a microplate reader (Tecan Infinite F200) at 595 nm. All experiments were repeated three times independently.

Swarming assay
The swarming assay was performed as previously described (Fan et al. 2017b). B. subtilis 9407 and mutant strains were grown in LB broth until an OD600 of 0.8 was reached. The cells were collected by centrifugation at 5000 rpm and resuspended in PBS buffer (10 mM, pH 7.4). For each tested strain, 2 μL aliquots of cell suspension were pipetted onto the surface of a LB medium plate (0.7% agar, w/v), after that, the plate was incubated at 37°C for 4-6 h to allow the bacteria to swim. Then, the plate was placed in a laminar flow hood with a constant flow of dry air to reduce the water content in the medium and therefore to terminate the swimming process. Subsequently, the plate was incubated overnight at room temperature. Five replicates were included for each strain, and all experiments were repeated three times independently. Photos were taken to record the swarming phenotypes of different strains.

Colonization assay
The colonization assay was performed in an artificial climate chamber. Melon seeds were incubated in water at 55°C for 30 min and then transferred to a Petri dish with wet gauze and kept at 28°C for 36 h to allow germination. B. subtilis 9407 and mutant strains were cultured in LB broth to an OD600 of 0.8. The bacterial cells were collected by centrifugation at 5000 rpm and washed three times with sterile ddH 2 O, and then resuspended with PBS buffer. The germinated seeds were soaked in the bacterial suspensions for 30 min and then sown in pots filled with a mixture of vermiculite and organic soil (1:2 v/v), with six seeds per pot. The pots were then placed in an artificial climate chamber at 25°C with a 16 h light and 8 h dark photoperiod. At 6, 12, and 18 days after sowing, the bacterial population colonizing melon roots and leaves was determined by plate counting as described previously (Fan et al. 2017b). Each straintreated seeds were sown in three pots. The experiment was repeated three times independently.

Evaluation of biocontrol efficacy under greenhouse conditions
Melon seeds and bacterial suspensions of B. subtilis 9407 and mutant strains were prepared as described above for the colonization assay. Treatment of the germinated seeds with bacterial suspensions of B. subtilis 9407 and mutant strains was the same as in the colonization assay. Inoculum of the pathogen A. citrulli MH21 was produced in shake culture in 50 mL LB broth at 28°C for 36 h. The bacterial cells of A. citrulli MH21 were resuspended in a saline solution (0.9% NaCl) and adjusted to an OD600 of 0.8. Three days after sowing when melon seeds grew two cotyledons, the seedlings were spray-inoculated with A. citrulli MH21. Two controls were included in this experiment. In one control (referred to as "Blank"), PBS buffer instead of B. subtilis 9407, and a saline solution (0.9% NaCl) instead of A. citrulli were used; in the other control (referred to as "Control"), the germinated seeds were soaked in PBS buffer, and the seedlings were sprayed with A. citrulli. Pots with different treatments were randomized across the experimental area in a greenhouse with a light/dark period of 14/10 h at 20-35°C, and were sprayed with sterilized water every 2 days. Each tested B. subtilis strain had three pots as a repetition, and each pot had six seedlings. Disease incidence and severity were recorded within 3-5 d of the appearance of the first symptoms. The experiment was repeated three times.
Additional file 1: Figure S1. Genome map of Bacillus subtilis 9407. Figure S2. BRIG based on homology with B. subtilis 168. Figure S3. Schematic diagram of secondary metabolite clusters in the B. subtilis 9407 and 168 genomes. Figure S4. Growth curve of B. subtilis 9407 and mutant strains.
Additional file 2: Table S1. Strains and plasmids used in this study.