Involvement of Flagellin in Kin Recognition between Bacillus velezensis Strains

ABSTRACT Kin discrimination in nature is an effective way for bacteria to stabilize population cooperation and maintain progeny benefits. However, so far, the research on kin discrimination for Bacillus still has concentrated on “attack and defense” between cells and diffusion-dependent molecular signals of quorum sensing, kin recognition in Bacillus, however, has not been reported. To determine whether flagellar is involve in the kin recognition of Bacillus, we constructed Bacillus velezensis SQR9 assembled with flagellin of its kin and non-kin strains, and performed a swarm boundary assay with SQR9, then analyzed sequence variation of flagellin and other flagellar structural proteins in B. velezensis genus. Our results showed that SQR9 assembled with flagellin of non-kin strains was more likely to form a border phenotype with wild-type strain SQR9 in swarm assay than that of kin strains, and that non-kin strains had greater variation in flagellin than kin strains. In B. velezensis, these variations in flagellin were prevalent and had evolved significantly faster than other flagellar structural proteins. Therefore, we proposed that flagellin is an effective tool partly involved in the kin recognition of B. velezensis strains. IMPORTANCE Kin selection plays an important role in stabilizing population cooperation and maintaining the progeny benefits for bacteria in nature. However, to date, the role of flagellin in kin recognition in Bacillus has not been reported. By using rhizospheric Bacillus velezensis SQR9, we accomplished flagellin region interchange among its related strains, and show that flagellin acts as a mediator to distinguish kin from non-kin in B. velezensis. We demonstrated the polymorphism of flagellin in B. velezensis through alignment analysis of flagellin protein sequences. Therefore, it was proposed that flagellin was likely to be an effective tool for mediating kin recognition in B. velezensis.


RESULTS
The lack of flagellin impairs kin recognition of B. velezensis SQR9. The flagella are very important motility organs for the bacterium, and the bacteria without flagella are not able to swim and swarm (3,25). B. velezensis SQR9 is a plant-growth promoting rhizobacteria (PGPR) strain, isolated from cucumber rhizosphere soil. B. velezensis SQR9 mutant strain Dhag lacked flagellar filaments and did not swarm on a Semi-solid medium ( Fig. 1A, B, D, and E). The strain SQR9wt (wild type) merged with itself on the swarming plate (Fig. 1G, white arrows), while cannot merge with mutant Dhag without flagellar filaments, and even it avoided the surrounding area of Dhag to grow (Fig. 1G, black arrows). Moreover, we used the Dsrf (the surfactin synthetic gene mutant) as a control with motility mutants different than Dhag. Results showed that Dsrf mutant of B. velezensis SQR9 has the complete and normal flagella as the wild-type strain (Fig. 1C), but loses its swarming ability (Fig. 1F). When the Dsrf mutant meets the SQR9 wild-type during swarm assay, they merged on the swarm plate (Fig. 1H). These results indicated that flagellin is potentially involved in kin recognition for B. velezensis SQR9. Additionally, we observed that the Dsrf mutant with impaired swarming ability could exploit surfactin from wild-type as biosurfactant to restore its swarming ability (Fig. 1H), and these observations were consistent with the formerly report by Nicholas A. Lyons and Roberto kolter (33). Based on the results, kin recognition and cooperation were hindered between the mutant strain and the wild-type strain, suggesting that bacterial flagellar filaments might play a role in kin recognition in B. velezensis.
B. velezensis SQR9 assembled with "non-kin" flagella now behaved as non-kin when confronted with the wild-type in swarming assay. To study how flagellar filaments affect kin recognition ability in B. velezensis, we collected 20 strains belonging to B. velezensis, constructed a phylogenetic tree, and performed a swarm boundary assay between SQR9 and other 19 strains, to characterize kinship distance between Flagellin Involved in Kin Recognition mSystems 20 strains (Fig. 2). The phylogenetic tree was constructed basing the housekeeping gene gyrA (2264 bp), and 9 strains had the same gyrA gene sequences as SQR9 and on a phylogenetic tree branch, the other 10 strains had different gyrA gene sequences from SQR9 and were located on different tree branches ( Fig. 2A). 9 strains on the same branch with SQR9 formed merge with SQR9 on swarming plates (defined as kin), while 10 strains on different branches without SQR9 formed boundary phenotype with SQR9 (defined as non-kin) (Fig. 2B). Interestingly, among the non-kin strains and SQR9, the boundary width was positively correlated with kinship distance (Fig. S1).
After knowing the kinship of those 20 B. velezensis strains, 4 strains with different kinship distances to strain SQR9 were selected: SQR9, FZB42, NB20, and ACCC02961. These 4 strains all possessed complete pericyte flagella and superior swarm motility (Fig. S2A, G-I), and were all merged with themselves on swarming plates ( Fig. S3E-H). Four plasmids containing hag genes amplified from genome DNA of these 4 strains were transformed into mutant Dhag of SQR9, respectively. The Dhag carrying corresponding plasmids were respectively named Dhag-hag SQR9 , Dhag-hag FZB42 , Dhag-hag NB20 , and Dhag-hag ACCC02961 . Transmission electron microscope (TEM) images showed that the flagellar filaments were recovered after the transformation of the plasmid, which contains a different hag gene (Fig. S2C-F). Interestingly, the Dhag-hag SQR9 , Dhag-hag FZB42 , Dhag-hag NB20 , and Dhag-hag ACCC02961 restored a certain motility ability on semi-solid plates, however, the first 2 strains recovered considerably more than the latter 2 on swarm assay medium with 0.5% and 0.7% agar (Fig. 3A to F and Fig. S4). In swarm boundary assay, Dhag-hag SQR9 and Dhaghag FZB42 could merge with SQR9wt ( Fig. 3G and H and Fig. S3J and K), Dhag-hag NB20 was an intermediate phenotype with SQR9wt ( Fig. 3I and Fig. S3L), Dhag-hag ACCC02961 even formed boundary phenotype with SQR9wt ( Fig. 3J and Fig. S3M). In summary, when Dhag mutant of SQR9 was complemented different flagellar filaments from kin and nonkin strains, the swarm phenotypes against wild-type SQR9 are similar to donor strains of hag gene.
Production of bacillunoic acid by Dhag mutant and its flagellin gene complementary strains is similar to SQR9 wild type. We demonstrated in previous studies that changes in flagella (deletion or replacement with non-kin flagella) resulted in a change in the recognition phenotype (from merge to boundary) of the mutant and complemented strains with SQR9wt on swarm plate. Whether this phenomenon is caused by changes in the flagella or by changes in the secretion of antibiotics needs to be further explored. Our previous work shows that B. velezensis SQR9 can secrete a FIG 2 The recognition phenotype of B. velezensis SQR9 and its relative on swarm plate varied from merging to the boundary with their phylogenetic distance. (A) The tree was constructed on gyrA gene sequences (2264 bp) using MEGA (v.5.05) for Neighbor-Joining, the B. subtilis 168 (NC_000964.3) was selected as the outgroup. And the reliability of clades was tested by the 1000 bootstrap replications. (B) displayed swarm phenotype of B. velezensis SQR9 and 20 strains, which were sorted according to the phylogenetic distance between 20 strains and SQR9 on the tree from near to far. The first two rows were the kin of SQR9, which merged with SQR9; the last two rows were the non-kin of SQR9, which formed a boundary with SQR9, and the boundary tends to widen with the increase of the phylogenetic distance between the non-kin strains and SQR9. The results are representative of three experiments. showed similar boundary phenotype with non-kin strains ACCC02961 as SQR9wt strain ( Fig. S5H to M). Only the DGI mutant (the bacillunoic acid synthetic gene mutant) formed boundary phenotype with SQR9wt ( Fig. S5G) and formed a weakened boundary phenotype with non-kin strains FZB42, NB20 and ACCC02961 ( Fig. 4A to C), indicated that bacillunoic acid is partly involved in the kin discrimination of SQR9. The bacillunoic acids secreted by SQR9 into the fermentation supernatant can effectively antagonize B. velezensis FZB42, making it form an antagonistic circle around the Oxford cup with the fermentation supernatant. We compared the production of bacillunoic acids by using the method of the antagonistic circle diameter measurement (Fig. 4D). Results showed that deletion of flagella or complementation for the various hag genes in SQR9 does not affect bacillunoic acid production (Fig. 4E). In addition, we tested the surfactin production of SQR9wt and Dhag mutant in liquid culture using high-performance liquid chromatography. Again, similar production of surfactin was observed between the 2 strains (Fig. S6). These results are consistent with our conclusion that flagellin is partly involved in kin recognition in B. velezensis SQR9 without affecting the antibacterial compounds production.
For B. velezensis, the flagellin heterogeneity of non-kin strains is much higher than kin. The results above suggest that flagellin in B. velezensis strains might be involved in the recognition of kin, we next want to investigate whether the structure of flagellin is related to the kinship. We sequenced the hag gene sequences of the 20 B. velezensis strains Results showed that both ends of the sequences were relatively conservative, and there were only several variant bases, but the middle part of the sequences was very different (Fig. S7). To observe the details of the variable region of the flagellin sequences more clearly, we cut and display the variable region separately (Fig. 5A). The sequences belonging to different strains were sorted according to the gyrA gene similarity of the strains and SQR9, that was, the strain ACCC02961 at the bottom had the farthest kinship with SQR9. The strains on a tree branch with SQR9 have the same flagellin sequence as SQR9. Other strains had an extra sequence in the variable region of the flagellin sequence (the length of sequences was 52-57 aa), except for strain FZB42 ( Fig. 2A and Fig. 5A). The flagellin sequence variant region of FZB42 was very similar to SQR9, compare to strains on adjacent branches (Fig. 5A). This may be due to individual differences in strains, or the evolutionary rate of the hag gene was not strictly consistent with the gyrA gene. To clarify what difference the variable region of the flagellin sequence caused, we selected 4 strains that were known genomes: SQR9, FZB42, NB20, and ACCC02961, then applied the entire hag gene sequences to predict the tertiary structure of flagellin and FIG 4 The deletion and complementation of the hag gene of SQR9 had no significant effect on the production of antibacterial substances Bacillunoic acids. Compared with SQR9wt, the DGI mutant (the bacillunoic acid synthetic gene mutant) formed a weakened boundary phenotype with non-kin strains FZB42 (A), NB20 (B) and ACCC02961 (C). The DGI is a deletion mutant of the gene island in strain SQR9 for the synthesis of bacillunoic acids. The pictures of the plates were acquired 48 h after inoculation, and the results are representative of three experiments. (D) The photo of the antagonism circle assay of B. velezensis FZB42 by the fermentation supernatant of strains SQR9wt, Dhag, and four hag gene complemented strains (Dhag-hag SQR9 , Dhag-hag FZB42 , Dhaghag NB20 and Dhag-hag ACCC02961 ). (E) The production of Bacillunoic acids in SQR9 wild-type, Dhag mutants, Dhag-hag SQR9 , Dhaghag FZB42 , Dhag-hag NB20 and Dhag-hag ACCC02961 , was assessed using the antagonism assay of the fermentation supernatant of the tested strain to B. velezensis FZB42. The antagonism assays for each strain included nine replicates. The box plots were drawn using R (v.4.0.3), and the analysis of significant differences was performed using Duncan's multiple range tests (P , 0.05) on SPSS (v. 25).
Flagellin Involved in Kin Recognition mSystems flagellar filaments by using comparative modeling method based on Swiss-model database. The most similar template matched by the flagellin sequences of four strains was the same template 6t17.1.A, and the details of template matching information were placed in Table S1 (see Table S1 at [https://zenodo.org/record/7131344#.Yzei8thBxPY]). Based on the prediction results of the flagellin monomer, the variable region of the flagellin monomers of the NB20 and ACCC02961, which had more b-strands and random coils than that of SQR9 and FZB42 (Fig. 5B to E). After the flagellin monomers were assembled into flagellar filaments, the part of b-strands and random coils was exposed on the periphery of the columnar flagellar filaments (Fig. 5F to I). In addition, we also carried out the protein structure prediction of the flagellar cap structure (Table S1; at [https://zenodo.org/record/7131344#.Yzei8thBxPY]) and found that the flagellar cap structure of the 4 strains was very similar, and they were all composed of 5 protein monomers ( Fig. 5J to M). The flagellar cap was first assembled on the flagella, and then helps and regulates the flagellin monomers to gradually assemble into flagellar filaments (37,38). Therefore, SQR9 can assemble different flagellin monomers into flagellar filaments. Taken together, these results above suggest that flagellin in B. velezensis strains might be involved in the recognition of kin.
The variation of flagellin is higher than other flagellar proteins, and it is not conservative in B. velezensis. In the above analysis, we observed a wider variation in a specific area of the hag gene among 20 B. velezensis strains, we next want to explore During alignment, it was found that the middle region of the flagellin sequences (170-293 sites) showed the most variation ( Fig. 6), with many large insertions or deletions, roughly divided into 3 lengths: two amino acids, 21 amino acids, and 110 amino acids ( Fig. S9; at [https://zenodo.org/record/7086025#.YyR7PaRBxPY]). The divergence of the flagellin sequences was greater than that of the 20 B. velezensis strains above (Fig. S7). In contrast with the intermediate variable region of flagellin, both ends of flagellin have fairly conservative sequences, with only a few amino acid sites having variation, and the tail (C-terminal) was more conservative than the head-end (N-terminal) (Fig. 6).
In addition, we also analyzed sequence variations of other flagellar proteins, including extramembrane flagellar structural protein: filament cap protein (encoded by fliD gene) (Fig. S8A), junction protein (encoded by flgK and flgL genes) ( Fig. S8B and C), hook cap protein (encoded by flgD gene) (Fig. S8D), hook structure protein (encoded by flgE gene) (Fig. S8E); intramembrane protein: flagellar rod structure protein (encoded by flhO gene) (Fig. S8F). The analysis results showed that the primary sequence homology of these proteins was high (Fig. S8B to F), and only the individual amino acid residues of filament cap protein differed (Fig. S8A). The results above indicate that these structures of the flagella of the intraspecies strains of B. velezensis were all extremely conservative, with little variation.
In summary, the specific area of flagellin has large variation within B. velezensis strains and the variation degree of flagellin is higher than that of other structural proteins of flagella, whether the N-terminal conserved region or the central variable region, indicating that flagellin evolved faster than other structural proteins of flagella in B. velezensis.

DISCUSSION
In the natural environment, bacteria will use a variety of methods to maximize the benefit of progeny, and kin discrimination is one of them (21). In this study, we investigated whether flagella were involved in kin recognition of B. velezensis strains. To clarify the role that flagella play in kin recognition, we performed an exchange experiment of kin and non-kin strain's flagellin on B. velezensis SQR9, then detected their swarm phenotype. Our results showed that flagellin heterogeneity (from kin or non-kin strains) affected recognition of B. velezensis strains.
The hag gene encodes Bacillus flagellin, and its absence makes SQR9 unable to synthesize flagellar filaments (39). Electron micrographs showed that mutant Dhag has no Flagellin Involved in Kin Recognition mSystems flagella (Fig. 1B). Strains cannot swarm without flagella (Fig. 1E) but can slide, a shortdistance migration movement that does not depend on flagella (40), which accounts for the ability of the Dhag community to spread outward from the inoculation site ( Fig. 1E and G). When SQR9wt with flagellar filaments encounter mutant Dhag without flagellar filaments, it will detour (Fig. 1G), instead of the same strain as kin to merge (10), which implied that the lack of flagella filaments changed inherent patterns of communication and cooperation between the 2 populations. The next stage of investigation revolves around 20 strains of B. velezensis, and the strains with farther kinship distance from SQR9 had wider swarm boundaries with SQR9 ( Fig. 2 and Fig. S1), which was consistent with previous reports (13,33). Here, we need to note that we used B. velezensis as a model given the numerous strains that have been collected. Many plant-growth promoting strains (Bacillus amyloliquefaciens, among others) were reclassified as B. velezensis (41). Then, flagellin of 4 strains (SQR9, FZB42, NB20, and ACCC02961) with different kinship distances from SQR9 was swapped to SQR9, and the results showed that the swarming phenotype between SQR9wt and SQR9 with heterologous flagellin filaments was related to the flagellin divergence of the 2 strains ( Fig. 3G to J). The hag mutant strain of SQR9 assembling flagellin from its kin merged with wild type when they encountered each other on swarming plates; however, the hag mutant strain of SQR9 assembling flagellin of non-kin strains (NB20 and ACCC02961) forms boundary with wild type. Flagellin region interchange experiments further confirmed that flagella are involved in the kin recognition in B. velezensis SQR9.
The execution of outer membrane exchange in M. xanthus requires the initial recognition of 2 cell surface proteins, TraA and TraB, and successful outer membrane exchange can only be achieved when they both are present and have the same or similar structures (42,43). Some strains of a-, b-, and gammaproteobacteria use the CDI (contact-dependent inhibition) system to secrete adhesins for primary recognition with surface receptors of neighboring cells (44,45). In our study, flagellin played a role in the kin recognition of B. velezensis SQR9. However, the same SQR9-body means that they have the same virulence and immune system (11), therefore even if they are identified as non-kin as the difference in flagellin, they cannot attack and kill each other.
Given this, we analyzed the differences between the hag gene and flagellin structure between kin and non-kin strains. The results showed that the central region length of hag gene sequences of kin and non-kin strains varied greatly, ranging from 52 aa to 57 aa (Fig. 5A), which encodes the b-sheet and coils on the D2 domains of the flagellin monomer. Research has shown that the D2 domains contribute to the stability of flagellar filaments and that deletion of the domain also affects the primary anti-flagellin responses (46). In the study, the 4 Bacillus strains of flagellin did not contain the D3 domain and had 2 types of flagellin: without D2 domain (SQR9 and FZB42), containing D2 domain (NB20 and ACCC02961) (Fig. 5B to I). The lengths of amino acid sequences of flagellin vary widely among bacteria, especially the variable region at the center, ranging from 8 amino acid residues in Clostridium tetani to 247 amino acid residues in Helicobacter pylori (29). This part of the structure does not participate in the polymerization of flagellin monomers. After the flagellin monomer is assembled into filaments, the structure will be exposed on the periphery of the filaments and directly in contact with the outside environment (29,47). The wide variation in flagellin structures within species makes them potentially capable of mediating bacterial-bacterial recognition.
Our results showed that flagellin heterogeneity (from kin or non-kin strains) affected recognition of B. velezensis strains. (Fig. 5). Such high variability and rapid evolution of flagellin were reported to exist in many species (48,49), including 2 levels: the first level is the large difference in flagellin variable regions, which may involve gene transfer across phyla; the second level is that there are evolutionary variations in species, similar to the evolution of other conserved genes (49). For Bacillus, sequence diversity of flagellin within the same species has been widely investigated. In Bacillus cereus and Bacillus thuringiensis, flagellin sequence variation is one of the bases for classification (48). Here we found that flagellin sequence variation may plays an important role in kin recognition between B. velezensis strains. These indicate that the hag gene has a stronger response to environmental pressure. This intraspecific flagellin differential phenotype is likely to be a differential expression of intraspecific species communication.
In conclusion, our research showed that the flagellin contributes to the kin recognition between B. velezensis strains. However, the specific identification mechanism remains to be explored. In follow-up experiments, it would be interesting to investigate further whether the communication mechanism is flagellar-flagellar or flagellin-receptor specific in Bacillus spp.

MATERIALS AND METHODS
Strains information and cultural condition. A list of strains in the study can be found in Table 1, including 20 strains belonging to B. velezensis and several mutants of SQR9 (the strain accession number is 5808 in the China General Microbiology Culture Collection Center, CGMCC, and the genome accession number CP006890 in the National Center for Biotechnology Information [NCBI]). All strains came from Laboratory stock or were isolated from soil ( Table 1). All strains were grown at 37°C in low-salt LB (LLB) medium, including 10 g of Tryptone, 5 g of yeast extract, and 3 g of NaCl per L.  B. velezensis SQR9, hag::Spc, pNW33N-hag ACCC02961 (Zeo R Spc R Cm R ) This study resistance gene fragments was: 18 mL of water, 1 mL of 1 Â Phanta Max Master Mix DNA polymerase (Vazyme), 1 mL of dNTP mix, 25 mL of Buffer, 2 mL of the forward primer, 2 mL of the reverse primer, and 1 mL of DNA template. The PCR program was performed under the following conditions: 98°C for 2 min and 32 cycles at 98°C for 10 s, 55°C for 10 s, and 72°C for 4 min. Upstream, downstream, and spc resistance gene fragments were fused using the method of twostep overlapping PCR (50). The mixture volume (25 mL) for the first step was: 5 mL of water, 0.5 mL of 1 Â Phanta Max Master Mix DNA polymerase (Vazyme), 1 mL of dNTP mix, 12.5 mL of Buffer, 2 mL (100 ng) of the upstream fragment, 2 mL (100 ng) of downstream fragment, and 2 mL (100 ng) of resistance gene fragment. The PCR program was performed under the following conditions: 98°C for 2 min and 12 cycles at 98°C for 10 s, 50°C for 10 s, and 72°C for 4 min. For the second step, the mixture volume (50 mL) contained 18 mL of water, 1 mL of 1 Â Phanta Max Master Mix DNA polymerase (Vazyme), 1 mL of dNTP mix, 25 mL of Buffer, 2 mL of the primer up-F (the forward primer of the upstream fragment), 2 mL of the primer up-R (the reverse primer of the downstream fragment), and 1 mL of product from the first PCR step. In addition, the PCR program was performed under the following conditions: 98°C for 2 min and 32 cycles at 98°C for 10 s, 55°C for 10 s, and 72°C for 4 min.
After purification of the fused fragment of 3 genes (upstream, downstream, and Spc resistance gene), the transformation was conducted by the artificial induction of genetic competence. When SQR9 with plasmid pUBXC (carrying the xylose-inducible comK expression cassette) was cultivated to an OD 600 of 0.5 in LB medium, 1% (wt/vol) xylose was added. After 1 h of incubation, 20 mL fused fragment was mixed with 200 mL SQR9 cells in a 2 mL centrifuge tube and incubated at 37°C for 7 h. Then, cells were plated on LB agar plates including 100 mg mL 21 Spc, and the correct mutants were verified by sequencing (51).
The hag gene fragments and plasmid pNW33N (carry chloramphenicol [Cm] resistance gene) were digested with restriction endonucleases: XbaI and BamHI (TaKaRa), and the reaction system and conditions were referred to the instructions on the website of the TaKaRa bio (https://www.takarabiomed .com.cn/Product.aspx?m=20150106133447710028). After that, the hag gene fragment with the 2 sticky ends exposed and the plasmid pNW33N was ligated overnight at 16°C, and the enzymatic ligation system was as follows (10 mL): 5 mL Solution I (TaKaRa), 1 mL (100 ng) plasmid pNW33N and 4 mL (400 ng) hag gene fragment. The ligation mixture was transferred into Dhag mutant with plasmid pUBXC using the xylose induction method mentioned above. Then, cells were plated on LB agar plates including 100 mg mL 21 Spc and 5 mg mL 21 Cm, and the correct complement strains were verified by extracting plasmid and sequencing. All these mutants and flagellin gene complementary strains are listed in Table 1.
Swarm assay. Swarm assays were performed on the 9 cm plates containing B-medium with 0.7% agar at 37°C (13). Strains were grown on solid LLB plates at 37°C for 12 h before use and then transferred to 3 mL of liquid B-medium and shaken overnight at 37°C. The overnight cultures were then diluted to an optical density (OD 600 ) of 0.5, and 2 mL was spotted on the agar plates. The plates with a cover were dried in a laminar flow hood for 30 min, sealed and incubated for 2 days at 37°C, and photographed. Regarding the determination of boundary widths of non-kin strains in swarm boundary assays, each pair of strains included 6 replicates of the swarm boundary phenotype, and each replicate was measured three times using the Image J (v.1.53c) (53). The point plots were drawn using the ggplot2 package and linear regression analysis is performed using the lm() function in R (v.4.0.3).
Motility test. To test the motility ability of different strains, 3 kinds of B-medium with 0.3%, 0.5%, and 0.7% agar were selected. The strain preparation and culture conditions were the same as above. Pictures of the plates were acquired 24 h after inoculation. The halo area of the strains was measured in Image J (v.1.53c).
Determination of surfactin production. Forty milliliters of sterile supernatant of the tested strains were cultured in Landy medium (54) at 30°C for 60 h, the pH was adjusted to 2.0 with 6 mM HCl, and sat at 4°C overnight. It was then centrifuged to retain the pellet, 4 mL of methanol (LC/MS, Merck) was added to soak for 5 h, and the sample was filtered through a 0.22 mm membrane to obtain a sterile crude extract, which was stored at 4°C for testing.
Determination of surfactin production of the strains was performed using an HPLC 1200 apparatus (1200 series; Agilent). A high performance liquid chromatography (HPLC) system equipped with an Agilent ZORBAX Eclipse XDB-C18 (250 Â 4.6 mm, 5 mm) column was operated and maintained at 30°C. A mobile phase mixture consisting of an Acetonitrile and 0.1% (vol/vol) CH 3 COOH solution (ratio of 88:12) was pumped in an isocratic mode with a flow rate of 0.84 mL min 21 . The injection volume of the sample was set at 20 mL and was detected through a VWD detector at 210 nm. Each analysis was completed within 20 min.
Surfactin standard solution (1000 mg L 21 ) was prepared from 99% pure surfactin (shyuanye). The surfactin substance peaks in the sample to be tested were determined by comparing the chromatographic peak of the sample with the surfactin standard solution, and the yield of surfactin in the sample was characterized by the sum of the absorbance values of the last 3 well-separated surfactin substance peaks. Each sample was tested in triplicate. Analysis of significant differences was performed using independent sample T-Test (P , 0.05) on SPSS (v. 25).
Five milliliters of a diluted overnight culture of FZB42 (;10 5 CFU mL 21 ) was spread onto LLB plates (10 Â 10 cm) to be grown as a bacterial lawn. The supernatant of the strain to be tested that was cultured in medium B for 48 h (37°C, 170 rpm), was concentrated 3 times using a centrifugal filter (10 kDa, Amicon Ultra-15), then 180 mL was added to the Oxford cup on the bacterial lawn, and then the plate was placed at 22°C until a clear zone formed around the Oxford cup. It was then photographed, and the antagonistic circle was measured. The antagonism assays for each strain included 9 replicates. The box plots were drawn using R (v. 4 Phylogenetic analysis. In this study, the phylogenetic analysis of genes was conducted using MEGA (v.5.05) for Neighbor-Joining (55). The 1000 bootstrap replications tested the clades' reliability. Furthermore, annotation and beautification of trees were achieved through the iTol online site (https:// itol.embl.de) (56).
Electron microscopy. We cultivated the strains on the LLB solid medium at 37°C for 8 h, placed the plate at an angle, and soaked the fresh colony in sterile deionized water for 2 h, during which we gently shook the plate every 20 min. Then, the strain suspension on the copper net was air dried, the flagella of strains were observed with a HT7700 transmission electron microscope (TEM) that operated at 80 kV, and photographed.
Protein structure prediction. The hag and fliD completed sequences gene of strains SQR9, FZB42, NB20, and ACCC02961 were obtained from the NCBI genome database (SQR9 and FZB42) and sequenced draft genomes (NB20 and ACCC02961). The protein tertiary structure of flagellin monomer, flagellin homomer, and flagellar cap were predicted on the Swiss-Model website (https://swissmodel .expasy.org/).
Statistics. Duncan's multiple range tests (P , 0.05) of the SPSS version 25.0 (IBM, Chicago, IL, version 25.0) was used for statistical analysis of differences among treatments.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. We declare that we have no conflicts of interest.