Bacillus cereus 892-1 and its derivatives were cultured in tryptic soy broth (TSB; Guangdong Huankai Co., Ltd., Guangzhou, China) at 37°C, 200 rpm, or on nutrient agar plates (Guangdong Huankai Co., Ltd.) at 37°C. Escherichia coli strains were grown at 37°C in luria-bertani broth (LB; Guangdong Huankai Co., Ltd.). When needed, antibiotics were added at the following concentrations: 5 μg/mL of erythromycin, 17 μg/mL chloramphenicol for the growth of B. cereus, and 100 μg/mL of ampicillin for the growth of E. coli. A list of strains and plasmids used in this work is provided in Supplementary Table 1. Oligonucleotides are listed in Supplementary Table 2.

The construction and screening steps of a transposon mutagenesis library were depicted in Figure 1. The plasmid pMarA (Gao et al., 2019) carrying a mariner-based transposon TnYLB-1 was used as the backbone. To replace the selectable marker kanamycin resistance cassette, the chloromycin resistance cassette was amplified from pBAD33 (Guzman et al., 1995). The newly generated plasmid, named pMarA-cat, was transformed into the strain 892-1 by electroporation, followed by the selection for both Erm^R (erythromycin-resistant) and Chlo^R (chloramphenicol-resistant) colonies at 28°C. Positive transformants were inoculated at 37°C, 200 rpm overnight to induce transposon-mediated mutagenesis. Fifty microliters of diluted cultures (1:100,000, v:v) were then spread onto LB agar plates containing chloramphenicol and incubated at 48°C for 10 h to induce plasmid suicide. The biofilm phenotypes of each mutant was screened by a microplate reader (Gen5™, BioTek, Winooski, VT, United States). Then, potential mutants with altered phenotypes were verified by antibiotic selection, which are resistant to chloramphenicol (Chlo^R) and sensitive to erythromycin (Erm^S). To select mutants containing a transposon insertion, a quick DNA extraction method was used. Briefly, a single colony was suspended into 30 μL ddH₂O in a 1.5 mL tube and then ultrasonically treated at 40 kHz at 25°C for 5 min. Then, the mixture was centrifuged at 10,000 g, 25°C for 1 min. Afterward, the upper aqueous phase was carefully transferred to a new 1.5 mL tube without disturbing the pellet. The DNA samples was tested by the polymerase chain reaction (PCR). To confirm the transposon insertion site, restriction endonuclease Taq I (ER0671, Thermo Fisher Scientific Inc., Waltham, MA, United States) was used to cut the genomic DNA. The genome fragments were self-ligated by T4 DNA ligase at 25°C for 10 min (EL0014, Thermo Fisher Scientific, Waltham, MA, United States), and reverse PCR was performed using primer OIPCR-1/2 to amplify the sequence inserted by TnYLB-1 transposon. After sequencing, the insertion sites were identified by local blast (ncbi-blast-2.12.0+ -win64).

Different strains was constructed as described previously (Qi et al., 2011). Plasmid pHT304-TS (Zhu et al., 2011) was used to construct mutants by homologous recombination. Recombinant plasmids were generated by in-fusion cloning (Ma et al., 2019). Briefly, the vector was linearized by restriction endonuclease EcoR I and Sal I (1611 and 1636, Takara, Shiga, Japan) digestion. The upstream and downstream fragments were amplified by using genome DNA as the template. 5′ end of the forward and reverse primers of inserts were amplified by PCR with 15-20 bp homolog fragments of linearized vector. Then, the mixture of inserts, linearized vector, and 2 × Hieff Clone^® Enzyme Premix (10912 and 10911, Yeasen, Shanghai, China) was incubated at 50°C for 1 h by using a thermo-cycler (C1000 Touch, Bio-Rad, Hercules, CA, United States). The mixture was then transformed into E. coli DH5α by thermal shock directly.

For the construction of mutants, including ΔflgE, ΔmotA, and ΔmotB, recombinant plasmids were transformed into B. cereus 892-1 by electroporation as mentioned above. One milliliter SOC media was immediately added to suspend the bacteria and the mixture was then incubated at 30°C, 200 rpm for 3 h. After incubation, the bacteria were collected by centrifugation at 25°C 5,000 g for 2 min and then spread on an LB agar plate with erythromycin. Plates were incubated overnight in an incubator at 30°C and potential transformants were confirmed by PCR. For gene knockout, positive transformants were transferred to media with erythromycin and incubated at 42°C, 200 rpm for 6–12 h, and this process was repeated 6 times. Then, strains were incubated at 30°C, 200 rpm for 6–12 h for 9–12 times and spread on LB agar plates without antibiotics. Colonies were then transferred on LB agar plates with antibiotics, and the ones that could not grow on the plates were detected by PCR and sent for sequencing. For the construction of complementary strains, plasmid pHT304 (Arantes and Lereclus, 1991) was used. The gene was cloned into plasmid pHT304 by in-fusion cloning as described above. Recombinant plasmids and electroporation were followed as described above and positive transformants were used for the following experiments.

To evaluate the effect of gene deletion and complementation on bacterial growth, overnight cultures were diluted 1,000-fold (v:v) into fresh TSB broth, and then 200 μL of bacterial suspension was added to the wells of a 96-well plate. In total, three biological repeats with six technical repeats each were performed. Bacterial growth was monitored by measuring the optical density at OD₆₀₀ of each well at 37°C every 30 min for 12 h by using a microplate spectrophotometer (EPOCH2, Biotek, Vermont, United States) and the data were analyzed by GraphPad Prism (v8.0.2) to generate XY plots.

For pellicle formation analysis, B. cereus strains were grown overnight at 37°C, 200 rpm. Five microliters of overnight culture were inoculated into 5 mL TSB medium in a test tube, which was then statically incubated at 37°C for 12 h. The formation of the pellicle was recorded by a Nikon D750 camera. Evaluation of the ring part was performed as previously described with minor modifications (Stepanović et al., 2000). Briefly, Bacteria (3.3 × 10⁵ cfu/mL) were inoculated into 200 μL fresh TSB medium in 96-well polystyrene plates (Costar, Washington, DC, United States), and incubated at 37°C for 12 h statically. Then, planktonic cells were poured out, and plates were washed three times with ddH₂O. The remaining attached biofilms were dried and fixed with 210 μL of 95% methanol per well for 15 min. After drying, 210 μL of crystal violet (0.1%; w/v) was added to each well and incubated for 15 min to stain biofilms attached to the well surface. After briefly washing three times and drying, the crystal violet was dissolved in 220 μL of 30% acetic acid, and staining levels were assessed by measuring absorbance at 590 nm (A₅₉₀).

Overnight cultures were diluted to an OD₆₀₀ of 0.001 in TSB broth and biofilms were grown on 8 mm × 8 mm glass coverslips (WHB-48-CS, WHB, Shanghai, China) in 12-well plates (Costar, Washington, DC, United States) for 12 h at 37°C. Biofilms formed on the surface of the cell slide were fixed with 3% (w/v) glutaraldehyde overnight at 4°C. Samples were then dehydrated with a graded ethanol series, dried, sputter-coated with gold, and imaged by a scanning electron microscope (Hitachi S-3000N, Tokyo, Japan) operating at 20 kV and 83 μA.

The biofilm structure was visualized by confocal laser scanning microscopy (CLSM) as described previously (Zhao et al., 2022) with minor modifications. Overnight cultures were diluted 1,000 (v:v) times in 50 mL fresh TSB broth and biofilms were grown in a beaker (100 mL volume), which were then observed using a confocal laser scanning microscope (ZEISS LSM700, Oberkochen, Germany). Biofilms without planktonic cells were stained using SYTO^® 9 (Thermo Fisher Scientific Inc., Waltham, MA, United States) at 25°C in the dark for 2 min. Biofilms were then visualized using a CLSM by a 20× objective lens with excitation at 488 nm and emission at 500–550 nm. The images were processed by using the Zeiss ZEN (v3.5).

Different samples were examined by transmission electron microscopy (TEM; Tecnai G2 F20 S-TWIN, Thermo Fisher Scientific Inc., Waltham, MA, United States) for the appearance of flagella. Overnight cultures were diluted 1,000 (v:v) times with fresh TSB broth and statically cultured at 37°C for 7 h. The bacterial suspension was spotted onto a copper grid and air-dried. Then, the samples were stained using 3% phosphotungstic acid for 2 min and observed using the TEM.

The swimming assay was performed according to a previous study (Singh et al., 2016) with some modifications. Swimming plates contained 1% tryptone, 0.5% NaCl, and 0.25% agar. For conducting the swimming assays, 1 μL overnight cultures were spotted on the agar plate and incubated at 37°C statically for 12 h. After that, the plate was imaged by using a camera (Nikon D750, Japan).

Query amino acid sequences of MotA (Houry et al., 2010) (B. cereus ATCC14579) and MotB (Cairns et al., 2013) (B. subtilis NCIB3610) by BLASTP (ncbi-blast-2.12.0+ -win64). Alignment of amino acid sequences (Supplementary Figure 1) was performed by CLUSTALW¹ and ESPript 3.0² (Robert and Gouet, 2014).

RNA isolation and purification were performed using the RNeasy Mini Kit (74104, Qiagen, Hilden, Germany) according to the manufacturer’s instructions. RNA concentration and purification were measured with a NanoDrop One Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). For RT-qPCR (reverse transcription-qPCR), purified RNA was used to synthesize cDNA according to the instructions of the PrimeScript™ RT reagent Kit with gDNA Eraser (RR047A, Takara, Shiga, Japan). Primers listed in Supplementary Table 2 were designed by SnapGene^® 2.3.2. udp (encoding a UDP-N-acetylglucosamine 2-epimerase) was used as a reference gene (Reiter et al., 2011). TB Green^® Premix Ex Taq™ II (Tli RNaseH Plus) (RR820A, Takara, Shiga, Japan) was used for all qPCR reactions. qPCR reactions were performed on a Roche LightCycler^® 96 in eight tubes (PCR-0108-LP-RT-C, Axygen, Glendale, AZ, United States) using three-step PCR amplification reaction as follows: 30 s preincubation at 95°C by 1 cycle followed by 45 cycles of denaturation at 95°C for 5 s, annealing at 58°C for 30 s and elongation at 72°C for 30 s, for melting at 95°C for 10 s, 65°C for 60 s, 97°C for 1 s by 1 cycle. The specificity of the reactions was affirmed by melting peaks analysis of the amplified products. Relative expression of flgE was calculated by the 2^–ΔΔCT (Livak) method (Livak and Schmittgen, 2001) using the difference in Cq (quantification cycle) values of the sample and a calibrator for the target gene and udp. Triplicates RT-qPCR reactions were performed for each sample with negative control for three biological repetitions.

Cereulide was extracted as described previously with some modifications (Tian et al., 2019). In brief, overnight cultures of B. cereus 892-1 were inoculated into 50 mL of LB medium (1:1,000; v:v) and cells were grown at 30°C, 200 rpm for 24 h. Bacteria were collected by centrifugation at 4°C, 8,000 g for 5 min, and resuspended in 5 mL methanol (HPLC grade, Guangdong Huankai Co., Ltd., Guangzhou, China). The suspension was cultivated in a shaker at 28°C, 200 rpm overnight. Then, the supernatant was filtered through a 0.22 μm filter, filled with methanol into equal volume, and diluted into a suitable concentration for LC-MS analysis. A Q Exactive Plus Orbitrap LC-MS/MS System (Thermo Fisher Scientific., Waltham, CA, United States) was equipped with an H-ESI (electrospray ionization) II probe source and positive mode was chosen to determine cereulide concentration according to a previous method (In’t Veld et al., 2019). Mass spectrometric characterization of cereulide was performed using a C18 column (ACQUITY UPLC^® Peptide BEH, 300A, 1.7 μm, 2.1 mm × 100 mm, 1/pkg). Mass spectrometric detection the ammonium adducts of cereulide at m/z 1,170.7 ([M + NH₄]⁺) and potassium adducts of valinomycin at m/z 1,128.6 ([M + K]⁺) (Supplementary Figure 4; Seyi-Amole et al., 2020). Methanol and ultrapure water containing 10 mM ammonium formate, both of which contained 0.1% formic acid, were used as eluents A and B. The gradient elution conditions were exhibited in Supplementary Table 3. An injection volume of 5 μL and a flow rate of 10 μL/min were used. MS runtime was 13 min and the retention time (RT) for cereulide and valinomycin was 5.00 and 5.10, respectively. The spray voltage was 3.50 kV. The flow rate for sheath gas was 45 and 10 for aux gas. The temperature for capillary and aux gas heater was 300 and 350°C, respectively. The concentration of cereulide in analytes is calculated by calibration curve, which was obtained by plotting the area ratios of cereulide to valinomycin (internal standard) for different dilutions. Linear regression was applied to give the equation y = 0.00783839x + 0.0686906 with R² = 0.9952; y is the area ratios of cereulide to valinomycin; x is the concentration of cereulide; R² determined the coefficient of the linear regression. The data was acquired and processed by Thermo Xcalibur (v3.5).

In total, 500 chloromycetin-resistant and erythromycin-sensitive mutants were identified, which were then screened for the identification of potential mutants with defects in biofilm formation. Among them, one mutant, designated 3-86, presented an obvious biofilm formation defect (Figure 2A). After reverse PCR, sequencing, and local blast, the transposon was proved to insert into the gene flgE (Figure 2B).

To further illustrate the role of flgE in biofilm formation, we compared the pellicle formation of wild-type strain with ΔflgE and complementary strain. Wild-type cells can form a pellicle in the air-liquid interface, while ΔflgE strain cannot (Figure 3A). As expected, the complementary strain restored a comparable level in its ability to form a pellicle. Furthermore, the amount of the ring was significantly reduced in ΔflgE; however, the ring of the complementary strain (ΔflgE:pHT304-flgE) was largely restored, although it did not reach the wild-type level (Figure 3B). Through observation by an SEM, the wild-type and complementary strains showed a dense biofilm community (Figure 3C). In contrast, only sparse cells of the mutant strain remained on the grid. Besides, biofilms were imaged by CLSM. The results are the same as SEM. The wild-type and complementary strains had dense biofilm structure, while ΔflgE only had scattered bacteria (Figure 3D).

To exclude the possibility that the biofilm formation defect is due to the differential growth rate, we monitored the growth of different bacteria for 12 h and found that the growth rate of ΔflgE had no obvious difference compared with wild-type and ΔflgE:pHT304-flgE at the early stage and was a little bit faster than other strains at the stationary stage (Figure 4A). To verify the role of flgE in flagella synthesis or assembly in B. cereus 892-1, the cells of wild-type, ΔflgE, and ΔflgE:pHT304-flgE were inspected by a TEM. No flagella could be found in ΔflgE. In contrast, either the wild-type or ΔflgE:pHT304-flgE had obvious flagella (Figure 4B), and there was no difference in quantity between the wild-type and ΔflgE:pHT304-flgE (Figure 4C). Due to the loss of flagella, ΔflgE could not swim as no outward movement can be observed (Figure 4D).

To test the swimming ability or flagella itself is important for biofilm formation in emetic B. cereus, two flagellar paralytic strains were constructed and the biofilm formation ability was measured. As expected, ΔmotA and ΔmotB lost swimming ability in motility assay (Figure 5A). Surprisingly, pellicle could also be formed in ΔmotA instead of in ΔmotB (Supplementary Figure 2). The amount of ring in ΔmotA was also significantly higher than that in ΔflgE or ΔmotB, but lower than that in the wild-type strain (Figure 5B), demonstrating that swimming ability contributes to biofilm formation but is not necessary for biofilm formation. ΔmotB completely lost the ability to form a biofilm, and has no significant difference in biofilm formation compared with ΔflgE, indicating that flagellar structure itself does not play a scaffold-role in emetic B. cereus 892-1.

Since 892-1 is an emetic strain, we also evaluated cereulide production in different strains by LC-MS/MS. The concentration range of cereulide was 19.327–22.757, 8.1586–10.312, and 32.417 to 35.294 μg/mL in WT 892-1, ΔflgE and ΔflgE:pHT304-flgE, respectively (Figure 6A; Supplementary Table 4). In conclusion, cereulide production was significantly decreased in ΔflgE, with a reduction of approximately 60% compared with the wild-type strain. In addition, ΔflgE:pHT304-flgE produced more cereulide; therefore, the transcriptional levels of flgE were monitored in the bacterial logarithmic phase. The relative expression level of flgE was obviously increased in ΔflgE:pHT304-flgE compared with wild-type (Figure 6B).