About 1 g of fermented food was mixed with 10 ml of sterile saline and then diluted to 10^–3, 10^–4, 10^–5, 10^–6, and 10^–7 levels. All dilutions were spread evenly on coumarin medium (CM: 1% coumarin, 0.025% KH₂PO₄, 0.1% NH₄NO₃, 0.1% CaCl₂, 0.025% MgSO₄⋅7H₂O, 0.0001% FeSO₄, and 1.5% agar) and cultured at 37°C for 4 days. Single colonies were isolated and transferred to fresh CM plates three times. Colonies growing on CM plates were selected and tested for AFB1 degradation.

Overnight cultured bacterial cells were diluted to an optical density at 600 nm (OD₆₀₀) of 0.01 with fresh Luria-Bertani (LB) medium, and then AFB1 purchased from J&K Scientific (Beijing, China) was added into 1 ml of dilution for a final concentration of 2 μg/ml. Sterile LB medium with AFB1 was used as the control. After 3-day incubation at 37°C by shaking at 180 rpm, the supernatant was extracted using chloroform according to previous reports (Guan et al., 2008), and the chloroform fractions were evaporated and dissolved using dimethyl sulfoxide (Sigma-Aldrich, St. Louis, MO, USA). The redissolved solution was filtered using the 0.22-μm pore filter (Merck-Millipore, Darmstadt, Germany) and stored at −20°C for high-performance liquid chromatography (HPLC) detection. About 94–96% of AFB1 could be recovered from the liquid culture using chloroform extraction.

Aflatoxin B1 was analyzed by HPLC according to the procedure reported by Fang et al. (2020). The percentage of AFB1 degradation was calculated using the following formula: the percentage of AFB1 degradation = (1 − C_a/C_b) × 100%, where C_a and C_b are the concentration of remaining AFB1 in the sample and total AFB1 in the control sample, respectively.

Aflatoxin B1 metabolites were extracted with chloroform after a 72-h incubation of AFB1 degrading bacterium in LB medium with 2 μg/ml of AFB1 and analyzed by HPLC-Q-TOF-MS according to the procedure described by Fang et al. (2020). Extractions from the AFB1-degrading bacterium in LB and sterile LB media with AFB1 were used as controls.

Genomic DNA was extracted using Wizard^® Genomic DNA Purification kit (Promega, Beijing, China) according to the manufacturer’s protocol and sequenced using a combination of PacBio RS II Single Molecule Real Time (SMRT, Pacific Biosciences, MenloPark, CA, United States) and Illumina sequencing platforms (Hiseq X Ten; Illumina, San Diego, CA, United States). The PacBio reads and Illumina reads were used to assemble the complete genome sequence into a contig using the hierarchical genome assembly process (HGAP) and CANU (Version 1.7¹). The last circular step was manually checked and finished, generating a complete genome with seamless chromosomes and plasmids. Finally, error correction of the PacBio assembly results was performed with Illumina reads using Pilon. Sequence data were deposited at the US National Center for Biotechnology Information (NCBI) under accession number CP092778.

The coding sequences (CDSs) were predicted with Glimmer (Version 3.02²) and annotated from the databases of Non-Redundant (NR Protein Sequence Database), Swiss-Prot, Pfam, Gene Ontology (GO), Clusters of Orthologous Group (COG), and Kyoto Encyclopedia of Genes and Genomes (KEGG) using sequence alignment tools such as Basic Local Alignment Search Tool (BLAST, Version 2.3.0³), Diamond (Version 0.8.3⁴), and HMMER (Version 3.1b2⁵), and annotations were obtained from the best-matched subjects (E-value < 10^–5) for gene annotation. All data were analyzed on the free online Majorbio Cloud Platform⁶.

Disk diffusion susceptibility tests were conducted according to the procedure reported by the National Committee for Clinical Laboratory Standards. Briefly, a bacterial dilution (OD₆₀₀ = 0.01) was spread on the Mueller-Hinton agar (MHA: 0.2% beef dehydrated infusion, 1.75% casein hydrolyzate, 0.15% starch, and 2% agar) plates and the disks with 2 μg lincomycin, 5 μg ciprofloxacin or rifampin, 10 μg gentamicin, streptomycin, ampicillin penicillin, imipenem, or norfloxacin, 15 μg erythromycin, or 30 μg tetracycline, cefalexin, kanamycin, chloramphenicol, or vancomycin were put on the plates. After 12 h of incubation at 37°C, the diameters of the inhibition zones were recorded.

Lifespans were monitored as described previously (Donato et al., 2017). Briefly, L4 worms of C. elegans N2 were grown on a nematode growth medium (NGM) agar plate seeded with Escherichia coli OP50 at 20°C and treated with alkaline hypochlorite to collect embryos. Embryos were cultivated to obtain a synchronized population. Synchronized L4 worms were transferred to fresh NGM plates seeded with the tested bacterium or E. coli OP50 or E. coli OP50 plus AFB1 or degradation metabolites every 2 days. Worms were considered dead when they stopped pharyngeal pumping and did not respond to prodding with a platinum wire. The number of dead/live worms was recorded every day.

To evaluate the mutagenicity of the degradation metabolites, the Salmonella (Ames) test was conducted with the S9 Enzyme Activation kit (Iphase Pharma Service, Beijing, China) according to the manufacturer’s instructions and the procedure described by Fang et al. (2020). Briefly, the degradation metabolites extracted from a 96-h culture co-incubated with the AFB1-degrading bacterium and AFB1 were incubated with Salmonella typhimurium TA98 or TA100 at 37°C for 48 h. The number of S. typhimurium colonies was recorded, and the data were given as the number of reversed colony-forming units (CFUs). Samples extracted from LB medium with AFB1 were used as positive controls, and extracts from LB medium were used as negative controls.

The dilution of bacterial cells (OD₆₀₀ = 0.02, the same below unless specified) was cultured in LB medium with shaking for 48 h at 37°C, and the supernatant and cells were collected, respectively, after centrifugation at 12,000 rpm for 5 min at 4°C. The supernatant filtered with a 0.22-μm pore filter served as the extracellular extracts for AFB1 degradation. After washing with 10 mM phosphate buffer (pH 8.0) three times, the cells were broken by ultrasonication (25 kHz, ultrasound for 4 s interval 1 s, 15 min) in the ice bath, and centrifuged at 12,000 rpm for 15 min at 4°C. The supernatant filtered with 0.22-μm pore filter served as the intracellular extracts for AFB1 degradation. Meanwhile, cells washed with phosphate buffer were boiled for 20 min, resuspended in an equal volume of 10 mM phosphate buffer (pH 8.0), and served as dead cells for AFB1 degradation. Extracellular extracts, intracellular extracts, and dead cells were co-incubated with 2 μg/ml of AFB1 at 37°C with shaking at 180 rpm for 72 h, respectively. Cultures of LB medium or phosphate buffer supplemented with 2 μg/ml AFB1 were used as the control, and all variables of control groups were similar to those of the corresponding extracts and dead cells. Residual AFB1 was tested as described above.

Extracellular extracts were divided into four fractions to investigate the influence of proteinase K, SDS, and heat on AFB1 degradation. One fraction was boiled for 20 min, and other fractions were treated with proteinase K (1 mg/ml), SDS (1%), or SDS plus proteinase K for 6 h, respectively. Subsequently, each fraction was incubated with 2 μg/ml AFB1 at 37°C with shaking at 180 rpm, and phosphate buffer with 2 μg/ml AFB1 was used as the control. After 24 h, residual AFB1 was monitored as described above.

To investigate the effects of AFB1 concentrations on AFB1 degradation mediated by the AFB1-degrading bacterium, bacterial cells were incubated with 1, 2, 5, and 8 μg/ml AFB1, respectively, at 37°C for 96 h by shaking at 180 rpm. LB medium with the corresponding concentration of AFB1 was used as the control. The effects of temperature, pH, and metal ions were determined by setting the cultivation temperature at 25, 30, 37, 40, 45, or 50°C, adjusting the initial pH values to 5.0, 6.0, 7.0, 8.0, or 9.0, and adding MgSO₄ (0.5 mg/ml), ZnSO₄⋅7H₂O (0.5 mg/ml), CuSO₄⋅5H₂O (0.5 mg/ml), MnSO₄⋅H₂O (0.5 mg/ml), FeSO₄⋅7H₂O (0.5 mg/ml), or CaCl₂ (0.5 mg/ml). Bacterial cells were incubated in LB medium with 2μg/ml AFB1 at 37°C for 24, 48, or 72 h by shaking at 180 rpm. Correspondingly, LB medium with 2 μg/ml AFB1 in each incubation was used as the control. Residual AFB1 was detected by the HPLC described as above. In addition, bacterial growth was also investigated by measuring the OD₆₀₀ value.

To investigate the effects of initial pH values, temperature, and metal ions on AFB1 degradation by the cell-free supernatant of AFB1-degrading bacterial culture, the cell-free supernatant was collected as described and exposed to 2 μg/ml of AFB1, and the mixture was incubated at 37°C by shaking at 180 rpm. The effects of initial pH values were analyzed by adjusting the mixture to 5.0, 6.0, 7.0, 8.0, or 9.0. In the temperature test, the mixture was incubated at 20, 30, 37, 40, 50, 60, or 70°C, respectively. In terms of metal ions, the reaction mixture was supplemented with 0.5 mg/ml of MgSO₄, ZnSO₄⋅7H₂O, CuSO₄⋅5H₂O, MnSO₄⋅H₂O, FeSO₄⋅7H₂O, and CaCl₂, respectively. Correspondingly, LB medium with 2 μg/ml of AFB1 in each incubation was used as the control. After a 48-h incubation, residual AFB1 was analyzed by HPLC as described above.

To investigate an effect of the AFB1-degrading bacterium on the fungal growth of A. flavus, the antagonistic effect and dry weight were tested after the AFB1-degrading bacterium was co-incubated with A. flavus for 2 days in potato dextrose agar (PDA) and potato dextrose broth (PDB), respectively. About 1 μl of the bacterial cell dilution was spotted in the upper part of a PDA plate and 1 μl of a conidial suspension (1.0 × 10⁶ conidia/ml) of A. flavus was spotted in the lower part of the plate. The plates were incubated at 30°C for 2 days, and fungal growth was observed. Meanwhile, 100 μl of a conidial suspension (1.0 × 10⁹ conidia/ml) of A. flavus was added into 100 ml of PDB supplemented with bacterial cells (OD₆₀₀ = 0.02) and incubated at 30°C by shaking at 180 rpm. PDB with A. flavus conidia was set as the control. After 2 days, fungal mycelia and the supernatant were collected by centrifugation. The collected mycelia were dried at 60°C and weighted. The supernatant was filtered by the 0.22 μm pore filter and analyzed by HPLC to detect the content of AFB1.

About 100 μl of a conidial suspension (1.0 × 10⁹ conidia/ml) of A. flavus was added to 100 ml of PDB supplemented with bacterial cells (OD₆₀₀ = 0.02) and incubated at 30°C by shaking at 180 rpm. PDB with A. flavus conidia was set as the control. After 2 days, fungal mycelia were harvested and used to extract total RNA using the RNAiso™ Plus reagent (TaKaRa, Dalian, China). Total RNA was reversely transcribed to cDNA using the PrimeScript^® RT reagent kit (TaKaRa). Transcripts of targeted genes were quantified via real-time quantitative polymerase chain reaction (qRT-PCR) with paired primers (Supplementary Table 1) under the action of SYBR^® Premix Ex TaqTM (TaKaRa). The transcript of the fungal β-tublin gene was used as an internal standard. The relative transcript level of each gene was calculated as the ratio of its transcript in the group of A. flavus plus bacterial cells to the control group, using the threshold-cycle (2–ΔΔ^Ct) method.

All the above experiments were conducted three times. The results of three replicates were expressed as mean ± standard deviation (SD), and statistical analysis was subjected to one-factor analysis of variance (ANOVA) performed with SPSS software. It is considered statistically significant when p < 0.05 in all the experiments.

After primary screening using coumarin as the sole carbon source and secondary screening by addition of 2 μg/ml AFB1 in LB medium, four isolates showed the ability to degrade AFB1 after a 48-h incubation (Supplementary Figure 1A). Among the four isolates, WF2020, which was isolated from naturally fermented pickles, displayed the maximum degradation ability up to 70.22% (Supplementary Figure 1A). When AFB1 concentration was not more than 5 μg/ml, except those in the first 24 h, the percentages of degrading AFB1 for WF2020 during the 96-h incubation were nearly similar among the treatments at the same cultivation time and nearly reached the maximum at 72 h where the percentage of AFB1 degradation was more than 84% (Figure 1A). Moreover, when AFB1 concentration was up to 8 μg/ml, WF2020 could degrade AFB1 in a time-dependent manner, and a reduction of more than 75% was observed at 96 h (Figure 1A).

WF2020 is a Gram-positive bacterium with the typical colony characteristics of Bacillus sp. (Supplementary Figure 1B). According to genome sequences obtained using Illumina Hiseq and a PacBio system, the complete genome sequence of WF2020 comprises a 4,043,726 bp circular chromosome, consisting of 4,133 predicted genes, 27 rRNA genes, and 86 tRNA genes (Figure 1B). In addition, no plasmid was observed in the genome. Based on the sequence analysis of 16s rRNA and other 31 housekeeping genes, including dnaG, frr, infC, nusA, pgk, pyrG, rplA, rplB, rplC, rplD, rplE, rplF, rplK, rplL, rplM, rplN, rplP, rplS, rplT, rpmA, rpoB, rpsB, rpsC, rpsE, rpsI, rpsJ, rpsK, rpsM, rpsS, smpB, and tsf, the closest relative of WF2020 was B. amyloliquefaciens strain (Supplementary Figures 1C,D). Therefore, this isolate was termed B. amyloliquefaciens WF2020.

Adsorption and degradation are the two main approaches in the removal of mycotoxins by microbes (Hathout and Aly, 2014). Here, cell-free supernatant (i.e., extracellular extracts) of B. amyloliquefaciens WF2020 was more effective than dead cells and intracellular extracts in reducing AFB1 during a 72-h incubation (Figure 2A). The percentage reduction of AFB1 for cell-free supernatant, intracellular extracts, and dead cells is 60.67, 14.11, and 26.95% at 24 h, 71.01, 19.80, and 27.87% at 48 h, and 71.01, 20.95, and 27.85% at 72 h (Figure 2A), respectively. Additionally, bacterial cells harvested from the cultivation of 48-h incubation in LB medium removed 12.25% of AFB1 on average after incubated with 2 μg/ml AFB1 for 1 h at 37°C by shaking at 180 rpm. These findings suggested that the removal of AFB1 mediated by B. amyloliquefaciens WF2020 was mainly dependent on the degradation and the cell-free supernatant was the main active ingredient during AFB1 degradation. Moreover, AFB1 degradation capacity of the cell-free supernatant decreased by 20.50, 93.40, and 100% after pretreatment with proteinase K, SDS, and SDS plus proteinase K (Figure 2B), respectively. Furthermore, cell-free supernatant still could degrade AFB1 by 37.16% after boiling for 20 min (Figure 2B). These results indicated that thermostable extracellular proteins or enzymes secreted by B. amyloliquefaciens WF2020 were involved in AFB1 degradation.

pH, temperature, and metal ions affected the AFB1 degradation ability of the cell-free supernatant from bacteria and fungi (Zhang et al., 2014; Rao et al., 2016; Wang et al., 2017; Xu et al., 2017; Shu et al., 2018). Here, AFB1 degradation studies with different incubation temperatures after 48-h incubation showed that the cell-free supernatant of B. amyloliquefaciens WF2020 could degrade AFB1 at temperatures ranging from 20°C to 70°C and the percentage of AFB1 degradation at 70°C remained more than 70% (Figure 2C), implying that the active constituents or components of the cell-free supernatant were thermostable and could work well within a wide range of working temperature. Moreover, the percentage of AFB1 degradation increased with the increase of temperature up to 60°C which was the optimum temperature for 100% AFB1 degradation (Figure 2C). In addition, the cell-free supernatant of B. amyloliquefaciens WF2020 could degrade AFB1 over a broad pH from 5.0 to 9.0 and the maximum percentage displayed at pH 8 (Figure 2D). Lastly, the effects of metal ions on the AFB1 degradation ability of the cell-free supernatant were evaluated (Figure 2E). Mn²⁺, Mg²⁺, Fe²⁺, and Cu²⁺ stimulated AFB1 degradation by 26.52, 15.19, 8.29, and 5.69%, respectively, whereas Ca²⁺ had no significant effect, but Zn²⁺ inhibited the degradation by 6.73% (Figure 2E), inferring that Mn²⁺, Mg²⁺, Fe²⁺, and Cu²⁺ may act as enzyme activators, membrane stabilizers, and help to maintain the structural integrity of proteins.

Based on the genomic sequence analysis, there are 12 secondary metabolic gene clusters via an antiSMASH analysis, but only six gene clusters harbored 100% similarity to those of known secondary metabolites (Table 1). The metabolites of the six gene clusters were macrolactin, bacillaene, fengycin, difficidin, bacillibactin, and bacilysin, respectively (Table 1), which are active substances with antibacterial, antifungal, anticancer, antiviral, anti-biofilm activities, biocontrol activity, etc. (Chen et al., 2008; Ryohei et al., 2015; Wu et al., 2015; Cochrane and Vederas, 2016; Zhou et al., 2018, 2021; Catherine et al., 2020; Erega et al., 2021; Kaushik et al., 2021). Moreover, a total of 35 genes with up to 50% similarity were found after blasting in the database of virulence factors, but they were not virulence genes but rather regulatory genes that played important roles in regulating biological processes, including virulence in other bacteria (Table 2). Additionally, a total of 19 genes with up to 50% similarity were found after blasting in the Comprehensive Antibiotic Resistance Database, and there is only one gene, imrB important for the resistance to lincosamide antibiotics, with up to 85% similarity (Table 3). Susceptibility to the corresponding antibiotics showed that, except for lincomycin belonging to a member of lincosamide antibiotics, B. amyloliquefaciens WF2020 was sensitive to other 14 antibiotics, including tetracycline, penicillin, cefalexin, ampicillin, streptomycin, kanamycin, gentamicin, ciprofloxacin, chloramphenicol, vancomycin, imipenem, rifampin, erythromycin, and norfloxacin (Figure 3A), suggesting their lower likelihood of being antibiotic-resistant bacterium.

Moreover, C. elegans has emerged as an invertebrate model to study host–pathogen interactions since its first documentation by Sydney Brenner (Brenner, 1974; Kumar et al., 2020). In the current study, C. elegans was used to evaluate the toxicity of B. amyloliquefaciens WF2020 to animals. C. elegans fed on B. amyloliquefaciens WF2020 cells showed significantly increased longevity compared with the effect of the laboratory-feeding bacterium E. coli OP50 cells when used as a food source (Figure 3B). The survival of worms fed on WF2020 cells increased by an average of 20.78% (average survival: 14.58 days, 95% confidence interval (CI): 13.83–15.32) compared with the strain OP50 (average survival: 17.61 days, 95% CI: 16.47–18.75). Maximum lifespans of worms fed on WF2020 were prolonged by 6 days compared with the strain OP50.

Except for the safety of the AFB1-degrading bacterium, the toxicity of AFB1 degradation products should not be neglected as some degradable products might be toxic like AFB1. In this case, B. amyloliquefaciens WF2020 might degrade AFB1 into C₁₅H₁₁O (m/z 207.08), C₁₅H₁₅O₂ (m/z 227.11), and C₁₅H₁₉O₄ (m/z 263.13), according to the HPLC-Q-TOF-MS analysis of the 72-h co-incubation culture of AFB1 and B. amyloliquefaciens WF2020 (Supplementary Figure 2), compared with those of AFB1 solution and the fermentation culture of B. amyloliquefaciens WF2020. Firstly, the Ames test was used to assess the mutagenicity of AFB1 degradation products by B. amyloliquefaciens WF2020. Compared with the control group, an approximately twofold increase in the number of revertant CFUs from S. typhimurium TA98 and TA100 was observed in the AFB1 group, but there was no significant difference in revertant CFUs of the degradation products and the control group (Figure 3C), indicating that B. amyloliquefaciens WF2020 converted AFB1 to the metabolites with a loss of mutagenicity. Except for mutagenicity, AFB1 decreased the lifespan and increased the mortality rate of C. elegans (Yang et al., 2015). Therefore, the effect of AFB1 and its degradation products on the lifespan of C. elegans was performed to further evaluate the toxicity of AFB1 degradation products mediated by B. amyloliquefaciens WF2020. The mean lifespan exposed to AFB1 significantly decreased by 25.14% compared with the control, but there was no significant difference in the survival rates of C. elegans exposed to degradation products and the control (Figure 3D), implying that AFB1 degradation products mediated by B. amyloliquefaciens WF2020 were not toxic to the lifespan of C. elegans. These findings demonstrated that B. amyloliquefaciens WF2020 degraded AFB1 into metabolites, which exhibited no mutagenicity or toxicity to the lifespan of C. elegans.

These collective results demonstrated that B. amyloliquefaciens WF2020 might be used as a potential probiotic to degrade AFB1 in food and feed.

To evaluate the effects of fermentation conditions on AFB1 degradation mediated by B. amyloliquefaciens WF2020, incubation temperature, the initial pH of the culture, and metal ions were chosen as the tested fermentation conditions. In this study, AFB1 was degraded by B. amyloliquefaciens WF2020 at all incubation temperatures after 72-h incubation. The percentage of AFB1 degradation was 31.20, 46.99, 86.53, 85.16, 89.24, and 48.79% on average at 25, 30, 37, 40, 45, and 50°C, respectively (Figure 4A). However, the degradation rate showed no significant difference in the range of 37–45°C (Figure 4A). The growth of B. amyloliquefaciens WF2020 at 25, 30, 45, and 50°C decreased by 32.94, 19.47, 22.44, and 32.23%, respectively, compared with that at 37°C, and bacterial growth at 37 and 40°C showed no significant difference (Figure 4B). Combined with the effects of different temperatures on the AFB1 degradation ability of the active component of B. amyloliquefaciens WF2020, we speculated the lower degradation of AFB1 mediated by B. amyloliquefaciens WF2020 at 25 and 30°C might be due to the lower bacterial growth and lower activities of the active components at 25 and 30°C, and the lower degradation of AFB1 at 50°C might be attributed to the lower bacterial growth of B. amyloliquefaciens WF2020.

In B. velezensis DY3108, AFB1 degradation capability decreased in parallel with a decrease in initial pH (Shu et al., 2018), suggesting that the initial pH of the medium might be a critical factor in AFB1 degradation mediated by Bacillus. Here, AFB1 degradation mediated by B. amyloliquefaciens WF2020 was also sensitive to the initial pH of the medium. The percentage of AFB1 degradation was 47.36, 43.20, 46.08, 62.02, and 27.01% on average at an initial pH of 5, 6, 7, 8, and 9, respectively (Figure 4C), indicating that an initial pH value at 8.0 favored degradation when AFB1 was co-incubated with B. amyloliquefaciens WF2020. Moreover, the growth of B. amyloliquefaciens WF2020 at an initial pH of 5, 6, and 9 was inhibited by 25.36, 19.70, and 94.50%, respectively, compared with that at an initial pH of 7, and bacterial growth at an initial pH of 7 and 8 showed no significant difference (Figure 4D). Thus, the significant decrease in the percentage of AFB1 degradation mediated by B. amyloliquefaciens WF2020 at pH 9.0 might be due to the severe bacterial growth defects at pH 9.0.

With respect to the effect of metal ions on AFB1 degradation mediated by B. amyloliquefaciens WF2020, it was observed that, compared with the control, Mn²⁺, Ca²⁺, Fe²⁺, and Cu²⁺ stimulated degradation by 30.24, 25.35, 24.14, and 15.36%, respectively, and Mg²⁺ showed no significant difference though the percentage of AFB1 degradation increased by 8.61%, whereas Zn²⁺ inhibited degradation by 30.39% (Figure 4E). Moreover, compared with the control group, the growth of B. amyloliquefaciens WF2020 treated with Mg²⁺, Fe²⁺, Mn²⁺, and Ca²⁺ increased by 63.22, 67.84, 94.00, and 116.13%, respectively, but Zn²⁺ inhibited bacterial growth by 29.05% and Cu²⁺ had no significant effect on bacterial growth (Figure 4F). Combined with the effects of different metal ions on the AFB1 degradation ability of the active component of B. amyloliquefaciens WF2020, we speculated that changes in AFB1 degradation by B. amyloliquefaciens WF2020 caused by Fe²⁺, Mn²⁺, and Zn²⁺ might be due to the effects of corresponding metal ions on bacterial growth and active component capacities, and the increase in AFB1 degradation by B. amyloliquefaciens WF2020 caused by Ca²⁺ and Cu²⁺ might be attributed to the increase in bacterial growth and active component capacities caused by the corresponding metal ions, respectively.

Except for AFB1 degradation, B. amyloliquefaciens WF2020 could inhibit the fungal growth of A. flavus and reduce AFB1 production. Pairwise interaction on agar plates proved that B. amyloliquefaciens WF2020 inhibited the fungal growth of A. flavus (Figure 5A). Moreover, the dry weight of the co-incubation culture of B. amyloliquefaciens WF2020 and A. flavus was reduced by 6.55% compared with that of A. flavus culture (Figure 5B). Additionally, B. amyloliquefaciens WF2020 completely inhibited AFB1 production when co-incubated with A. flavus in PDB for 2 days (Figure 5C). Accompanied by a reduction of AFB1 production, B. amyloliquefaciens WF2020 suppressed the transcriptional expression of 10 aflatoxin pathway genes (aflA, aflB, aflE, aflG, aflH, aflJ, aflK, aflL, aflO, and aflQ) and 2 gene encoding transcription factor aflR and aflS by 22.44–100% but increased the expression of aflM, an aflatoxin pathway gene, by 146.98% (Figure 5D). The downregulated expression of 10 aflatoxin pathway genes and 2 transcription factors suggested that AFB1 synthesis might be inhibited by B. amyloliquefaciens WF2020, which might result in reduced AFB1 production caused by B. amyloliquefaciens WF2020.