Aflatoxin standards were purchased from the Food and Drug Administration (Lot number: 61001-201703, Beijing, China). Coumarin medium (magnesium sulfate 0.25 g, potassium dihydrogen phosphate 0.25 g, ammonium sulfate 0.5 g, potassium nitrate 0.5 g, calcium chloride 0.05 g, agar 20 g, coumarin 1 g, and deionized water 1 L) and LB solid medium (tryptone 10 g, yeast extract 5 g, sodium chloride 10 g, agar 20 g, and deionized water 1 L) were self-prepared.

Based on previous reports, we found that the changes of Aspergillus flavus in Polygalae radix from different regions in China showed regularity (Chang et al., 2021). There may be microorganisms in Polygalae radix that can inhibit Aspergillus flavus, and even have potential activity in degrading AFB₁. In addition, few studies have isolated strains that degrade AFB₁ from Polygalae radix. Therefore, we isolated potential functional bacteria from the surface of Polygalae radix from different regions. These samples of Polygalae radix came from Yuncheng, Shanxi province, China (seven samples), Xingtai, Hebei province, China (one sample), Weinan, Shaanxi province, China (one sample), Luoyang, Henan province, China (one sample), Chifeng, Inner Mongolia, China (one sample), and Bijie, Guizhou province, China (one sample), respectively.

Aflatoxin B₁-degrading strains were isolated from the surface of the Polygalae radix of various origins. Specifically, 1 g of Polygalae radix was placed in a sterile centrifuge tube, 9 mL of sterile water was added, vortexed and homogenized, and 500 μL of it was evenly coated in coumarin medium, with an equal amount of sterile water as a control, and incubated for 72 h at 37°C. Coumarin medium was used to screen for AFB₁ degrading strains (Tang et al., 2024). Coumarin was the only carbon source and the strains that could grow in coumarin medium were considered to be able to degrade AFB₁. These bacterial strains grown in coumarin medium were purified to single colonies using LB solid medium (tryptone 10 g, yeast extract 5 g, sodium chloride 10 g, agar 20 g, deionized water 1 L) and preserved.

The bacterial strains grown in coumarin medium were re-screened. The purified bacterial strains were inoculated into 50 mL of liquid LB medium and incubated for 48 h at 160 rpm at 37°C. The culture broth was collected by centrifugation at 8,000 rpm for 15 min at 4°C. 10 μL of AFB₁ standard was added to 490 μL culture broth for the experimental group. The same amount of culture broth was used as the control group, and culture broth containing the same amount of methanol solution was used as the blank group. All treatments were incubated for 72 h and the reaction was stopped by adding 500 μL of methanol. After incubation, the reaction was centrifuged at 12,000 rpm for 30 min and analyzed by UPLC–MS. The degradation rate of AFB₁ was calculated as follows: Degradation rate (%) = (1 − AFB_{1experimental group}/AFB_{1control group}) × 100%.

The screened bacterial strain was designated SX1-1. Gram staining, colony morphology, size, and acid and salt tolerance were observed, and starch hydrolysis, lipid hydrolysis and glucose utilization, indole production, and methyl red reactions were determined. Identification was also carried out by 16S rDNA sequence analysis. Universal primers 27F (AGAGTTTGATCCTGGCTCAG) and 1492R (GGTTACCTTGTTACGACTT) were used for PCR amplification of the strains under investigation. A amplification products were analyzed by 1.0% agarose gel electrophoresis and products were sequenced by Sangon Biotech (Shanghai) Co., Ltd. The sequencing results were proofread and spliced using ContigExpress software, and the resulting sequences were used to build a clustering tree based on neighbor-joining (NJ) using MEGA 6.0 software (molecular evolutionary genetics analysis), and the support test for each branch was performed by bootstrap (1000 repetitions) to test the support of each branch.

SX1-1 culture with OD_600nm = 1.0 was inoculated into fresh LB medium and incubated for 12 h. The bacterial culture was then divided into two tubes and centrifuged at 12,000 rpm at 4°C for 10 min to separate the supernatant and the sediment. The supernatant of one tube was filtered by a 0.22 μm water filtration membrane to obtain the fermentation supernatant sample. The sediment was washed three times with phosphate balanced solution (PBS) (0.1 M, pH 7.0) buffer and then resuspended in PBS as the active cell solution. The sediment in the other tube were obtained as described above and were subsequently treated by ultrasonic treatment (400 W, 20–25 kHz) for 15 min, centrifugation was then carried out at 12,000 rpm at 4°C for 10 min and the supernatant was passed through a 0.22 μm filtration membrane as the cell lysate (Deng et al., 2022). 490 μL of supernatant, active cell solution and cell lysate were taken separately, 10 μL of AFB₁ solution was added and incubated at 37°C for 72 h. The toxin residues were measured in the control group with equal amounts of PBS solution added to equal amounts of AFB₁ solution and a blank group with LB medium without any treatment.

Another supernatant was taken and treated with 1 mg·mL⁻¹ proteinase K and heat treatment, and the untreated supernatant was used as control (Chen et al., 2022; Deng et al., 2022) 490 μL of the each group were taken and added to 10 μL of AFB₁ solution at a concentration of 0.005 μg·mL⁻¹, and incubated at 37°C for 72 h under protected from light. The reaction was terminated with 0.5 mL of methanol and the untreated LB medium was used as the blank group to measure the toxin residue. All experiments were repeated three times.

SX1-1 cultures with OD_600nm = 1.0 were inoculated into fresh LB medium and sampled at 0, 12, 24, 48, and 72 h under the same conditions. The samples were centrifuged at 10,000 rpm for 10 min to extract the supernatant and the biomass of the flora was measured by OD_600nm and the free protein content was measured using the BCA kit (R23183, Shanghai Yuanye Biotechnology Co., Ltd.). To assess the effect of treatment time, the culture systems were sampled at 0, 12, 24, 48, 72, and 96 h to determine the AFB₁ residue. To assess the effect of treatment temperature, culture systems were incubated in the dark at 27, 37, 47, 57, and 67°C and AFB₁ levels were determined. To evaluate the effect of pH, the initial pH of the incubation system was adjusted to 5.0, 6.0, 7.0, 8.0, and 9.0 with sterile 0.1 mol·L⁻¹ HCl solution or 0.1 mol·L⁻¹ NaOH solution, respectively, and the reactions were carried out at the optimum time and temperature to investigate the effect of pH on the degradation process.

The instrument was methodologically tested before use and samples were determined after passing the test. The chromatographic conditions were as follows: column BEH C18 (100 mm × 2.1 mm, 1.7 μm, Waters, United States), mobile phase 0.1% formic acid in water (A) and methanol (B), separation conditions as follows: 0 min, 90% A, 0.5 min, 90% A, 6 min, 10% A, 7 min,10% A, 9 min, 90% A, and 10 min, 90% A. The sample volume was 2 μL. Samples were detected by triple quadrupole mass spectrometry using an electrospray ionization (ESI) source in positive ion mode with an AFB₁ parent ion of (m/z) 313.2, a quantitative daughter ion of (m/z) 241.1, a collision voltage of 36 V and a daughter ion of 285.1, and a collision voltage of 24 V.

SX1-1 cultures with OD_600nm = 1.0 were inoculated into fresh LB medium, the bacterial broth was centrifuged at 4,000 × g for 10 min and the organisms were washed twice with sterile water. Genomic DNA was extracted by the SDS method, and DNA purity and integrity were checked by 1.0% agarose gel electrophoresis and quantified by Qubit. Qualified DNA samples were randomly fragmented into approximately 350 bp fragments using a Covaris ultrasonic fragmentation machine and genomic libraries were prepared using the NEBNext®Ultra™ DNA Library Prep Kitfor Illumina (NEB) kit. Sequencing was then performed on the Illumina NovaSeq PE150.

The sequencing results were assembled and corrected to ensure the quality of the genome. Genewise software (version 2.4.1) was used for coding gene prediction, and the protein sequences of the predicted genes were compared with each functional database by Diamond (evalue ≤1e⁻⁵), and for each sequence (Chen et al., 2022), the comparison result with the highest score (default identity ≥40%, coverage ≥40%) was annotated. The main common functional databases currently available for annotation were Gene Ontology (GO), Clusterof Orthologous Groups of proteins (COG) (Tatusov et al., 2003), Kyoto Encyclopedia of Genes and Genomes (KEGG), Carbohydrate Active enZYmes DatabaseS (CAZy), Virulence Factor Database for Pathogenic Bacteria (VFDB), etc.

We collected coix seeds of different origins from the Chinese herbal market, stored them at 25°C, and tested the AFB₁ content according to the above UPLC-MS analytical conditions every half month until the AFB₁ content exceeded the limit (5 μg·kg⁻¹) specified in the Chinese Pharmacopeia. We used the coix seeds with AFB1 content exceeding the limit as our subsequent raw material for in situ degradation.

12 g of AFB₁-exceeding coix seeds was weighed, 40 mL of fermentation supernatant of SX1-1 strain (OD_600nm = 1.0, pH 8.0) was added to coix seeds for incubation as experimental group, and an equal amount of LB liquid medium was added as control group, which was combined with the actual samples stored in the selected samples to be incubated at 37°C for 72 h. At the end of the incubation, the liquid was discarded, and the coix seeds was dried, pulverized, and extracted for the determination of the content of AFB₁ contained in the coix seeds.

All experiments were repeated three times and data were expressed as mean ± standard deviation (RSD) and analyzed using one-way ANOVA or Tukey’s test (SPSS 22.0), p < 0.05 was considered to be statistically significant.

The molecular structure of aflatoxin, which has a coumarin ring as its parent nucleus, is very stable, and thus bacteria that can grow on coumarin as the sole carbon source are expected to degrade and utilize AFB₁. In a preliminary screening of degrading bacteria, we obtained 55 strains capable of growing in plates with coumarin as the sole carbon source. The ability of the strains to degrade AFB₁ was further confirmed using UPLC-MS. The results showed that five strains had good degradation activity against AFB₁ (Figure 1A). The degradation ability of strain SX1-1 was significantly higher than other strains, with a degradation rate of 81.34%. Meanwhile, the confirmation results of UPLC-MS spectra further showed that the typical absorption peak of AFB₁ appeared at 4.6 min in the AFB₁ control solution, and that of the control also appeared around 4.6 min, while the peak area and peak height of the sample were significantly lower than those of the control (Figure 1B), indicating the efficient removal ability of SX1-1 for AFB₁. This was the highest degradation capacity of AFB₁ among the Bacillus megaterium reported to our knowledge (Liu R., et al., 2013).

Further identification of this SX1-1 was required for the effective use of this strain. Colonies of SX1-1 on LB plates were creamy white, round colonies with clean and thin edges and a slightly raised, bumped center (Figure 1C). Under the microscope, the cell morphology was rod-shaped and positive to Gram staining (Figure 1D). Sequences of approximately 1,500 bp were obtained by amplification and sequencing with 16S rDNA universal primers. Homology analysis of the sequenced PCR products showed that the sequences of SX1-1 were grouped together with Bacillus aryabhattai, Bacillus megaterium, and Bacillus huizhouensis (Figure 1E). This suggests that these strains may be more conservative in the primer fragments we used. Based on the biochemical reaction characteristics (Supplementary Table S1), combined with morphological and phylogenetic features, we finally identified SX1-1 as Bacillus megaterium.

The ability of the different fractions of SX1-1 (fermentation supernatant, cell lysate, and active cell solution) to degrade AFB₁ was further evaluated. It was found that the degradation efficiency of the fermentation supernatant was significantly higher than that of the active cell solution and cell lysate at 72 h, with a degradation rate of 97.45% (Figure 2A). This result indicated that the removal of AFB₁ by SX1-1 was mainly dependent on the fermentation supernatant, similar to previous studies. To further investigate whether the fermentation supernatant degraded AFB₁ as a protein or an enzyme, we subjected the fermentation supernatant to proteinase K and heat treatment. It was found that the degradation rate of AFB₁ in SX1-1 fermentation broth after proteinase K treatment (degradation rate was 25.17%) and after heat treatment (degradation rate was 18.72%) was significantly low than in the control group (Figure 2B). Since Proteinase K was a potent proteolytic enzyme and heating can also denature and inactivate proteins, although it is also possible that a small amount of SX1-1 is adsorbed to the cell membrane, we speculate that extracellular enzymes or proteins in the supernatant of SX1-1 fermentation are still involved in the degradation of AFB₁ as a major fraction.

We measured the OD_600nm and free protein content of SX1-1 for 72 h after inoculation into LB (Figure 3A), and the results showed that the strain multiplied rapidly within 12 h, with a large increase in biomass. Therefore, 12 h was considered as the optimal incubation time for SX1-1, when the OD_600nm was 1.46 and the protein content was 198.11 μg·mL⁻¹. In addition, pH, temperature, and incubation time may affect the ability of SX1-1 fermentation supernatant to degrade AFB₁. The results of the degradation ability of SX1-1 fermentation supernatant after coincubation with AFB₁ for 96 h showed that the degradation ability of the strain reached its optimum after 72 h and remained stable (Figure 3B). As the incubation temperature increased, the degradation rate of AFB₁ by SX1-1 gradually increased, and AFB₁ could be completely degraded at 57°C and beyond (Figure 3C). As the initial pH increased, the degradation rate of SX1-1 gradually increased, with complete degradation at pH 8 (Figure 3D), indicating that the degradation of AFB₁ by SX1-1 was more favorable under alkaline conditions. We therefore founded that the optimum incubation conditions for SX1-1 were 12 h and the optimum degradation conditions were 72 h, 57°C and an initial pH of 8.0. These results indicated that SX1-1 was more stable and would increase the practical application of SX1-1 in the future.

We sequenced the genome of SX1-1 to investigate its genomic properties that cause AFB₁ degradation. High quality data of 1,006 Mb was obtained for the SX1-1 genome, with an estimated genome size of 5.41 Mb, 201 contigs larger than 500 bp, a value of 370,411 for N50 and an average GC content of 37.1%. The genome contained 8,633 genes, 7,059 coding genes, and 147 non-coding RNAs, including 127 tRNAs, 13 rRNAs (115 srRNAs, 116 srRNA, and 123 srRNA) and seven sRNAs. In the ANI analysis, SX1-1 had a matrix value of 97.16% (over 95% considered to be the same species) with Bacillus megaterium (Figure 4A), confirming it as Bacillus megaterium.

The annotations in the GO database (Figure 4B), where genes were assigned functions, were divided into three main categories, including cellular component, molecular function, and biological process, with the largest number of genes involved in metabolic processes being annotated in biological process (2,591). In the molecular function category, the largest number of genes involved in catalytic reactions was 2,521, and in the cellular component category, the largest number of genes related to cells and cellular components was 1,809, respectively. Overall, the number of genes involved in bacterial metabolic processes and cellular processes was the highest. This implied the presence of a large number of metabolic and catalytic genes in SX1-1, suggesting a strong metabolic capacity and possible involvement in enzymatic catalysis. The functions annotated in the KEGG database were consistent (Figure 4C). The genes involved in carbohydrate metabolism, amino acid transport, and membrane transport pathways were the most abundant, with 335, 286, and 213 genes, respectively. We then analyzed the secondary metabolite gene clusters using antiSMASH, and found 11 biosynthetic gene clusters in the SX1-1 (Supplementary Figure S1). There were three gene clusters for terpene synthesis and the number of annotated genes was 65. These BGCs may contribute to the survival and better functioning of SX1-1.

Enzymatic reactions were considered to be the main mechanism for microorganisms degradation of AFB₁ (Watanakij et al., 2020; Zhao et al., 2020; Zhang et al., 2021b). We further mined the enzymes annotated in the SX1-1 genome and showed that more than 27% of the enzymes were annotated to the glycoside hydrolase (GH) family (56 genes), 31.7% to the CBM family (65 genes), 30.7% to the GT family (63 genes), and 9.7% to the CE family (20 genes). Further analysis revealed the presence of alkaline phosphatase, peroxidase, superoxide dismutase, and laccase in the COG database (Supplementary Table S2). Enzymes such as laccase (Sun et al., 2023), manganese peroxidase (Xu et al., 2023), and aflatoxin oxidase (Deng et al., 2022) were reported to be the main enzymes involved in the degradation of AFB₁, among which laccase, a class of polyketide oxidases, is considered as a green tool for xenobiotic degradation, which converts AFB₁ into less toxic compounds through 3-R-hydroxylation. The abundant presence of these enzymes suggests that they may be responsible for the effective degradation of AFB₁ by SX1-1. However, since the structure of aflatoxin B₁ includes phenolic, keto, alcohol, and ether groups, it is reasonable to speculate that the degradation of AFB1 may be complex, involving not only oxidation by oxidases and peroxidases, but also reduction by reductases or catalysis by oxidoreductases. However, whether these enzymes have a degradative function needs to be further verified.

In general, Bacillus megaterium is considered to be a safe and non-toxic microorganism for agriculture and medicine use (Rios-Ruiz et al., 2023). We compared the genome of SX1-1 with the VFDB and found 248 annotated genes, which are not virulence factors but important regulators involved in the microbial growth process (Supplementary Table S3). This suggests that the strain may be safe and green. In addition to the safety of the degrading bacteria, their safety against AFB₁ degradation products needs to be considered. Unfortunately, we do not currently have a clear resolution of their structure. Although some studies have shown that AFB₁ can be degraded into smaller non-toxic molecules (Yue et al., 2022), further experimental validation is needed to assess the safety of AFB₁ degradation products of AFB₁ by SX1-1.

To verify the effect of SX1-1 on contaminated medicinal coix seeds, the SX1-1 supernatant was co-incubated with AFB₁ contaminated coix seeds. After 72 h, the degradation rate was determined. The results showed that SX1-1 was able to degrade AFB₁-contaminated coix seeds and the degradation rate on the herb was 50.06% compared to the control (Figure 5). The surface of coix seeds does not appear to have any significant changes. This suggested that SX1-1 was able to better degrade aflatoxin-contaminated coix seeds. Compared to some other strains, such as Microbacterium protolyticum B240 reported to degrade AFB₁ on corn (Yan et al., 2022), the degradation of AFB₁ in coix seed by SX1-1 was not too high, which may be due to the own quality of coix seed as well as the amount of SX1-1 added to coix seed. In addition, the enzyme activity only reaches its maximum value in a certain pH range (Xu et al., 2023a,b). When the pH-adjusted SX1-1 supernatant was added to the herbs, the change of pH might affect the degradation of SX1-1, and it is necessary to continue to explore the optimal degradation of SX1-1 in co-culture with herbs.

Nevertheless, there are potential risks associated with the use of microbial degradation and the safety of the degradation by-products needs to be fully evaluated to explain the degradation mechanism of microbial degradation of AFB₁ and to lay the groundwork for the development of future strains. It must also be taken into account that the current amount of enzyme is insufficient for application in practical production, and that further purification and enzymatic studies of the enzyme contained in the supernatant will be needed in the future to sufficiently increase the enzyme activity. This is easier, more reproducible, more specific and more effective than using live microorganisms. Although researchers have attempted to address this issue using genetically engineered strains, the safety and efficacy of engineered bacteria remains to be thoroughly investigated.