This strain of B. subtilis BS-Z15 was isolated from native soil and preserved by Xinjiang Normal University’s Laboratory of Applied Biochemistry. B. subtilis BS-Z15 was cultured in beef paste liquid medium at 37°C for 18 h in a shaking incubator. The ratios of culture medium ingredients are in Supplementary Table S1 .

The extraction method of then-butanol part of the metabolites was based on Zhao et al. (Qian-Qian et al., 2018). Briefly, Liquid fermentation culture of B. subtilis BS-Z15 was centrifuged at 9000g for 15min to obtain sterile solution, precipitated with hydrochloric acid for 12h and continued centrifugation, and collect the precipitate dissolved in water, then have n-butanol extraction to obtain BS-Z15 antifungal crude extract and volatilize the liquid phase to collect the solid and weigh. The dried metabolite was fully dissolved in dimethyl sulfoxide (DMSO) and microporous filter membranes with pores of 0.2 mm are used to remove impurities. Mycosubtilin was purified using HPLC as described by Lin et al. (Lin et al., 2022). The elution procedure is provided in Supplementary Table S2 .

The isolated and purified mycosubtilin was weighed and compared to the n-butanol extracts to determine the mycosubtilin content in n-butanol extracts.

Four-week-old Kunming white male mice, weighing 22–25 g. Animals were raised in accordance with the conditions at the Xinjiang Medical University Animal Experiment Center.

Normal feeding for 7 days to acclimatize to the new environment, a total of 30 mice were randomly selected into 3 groups: Blank control group (Group A), gavage-administered with saline (0.5 mL per day); N-butanol site control group (Group B), gavage-administered n-butanol crude extract metabolite (a dose of 90 mg/kg per day); and mycosubtilin treated group (Group C), gavage-administered with mycosubtilin (a dose of 6 mg/kg). Drug treatment for 14 days in each group. The gavage administration dose of mycosubtilin was determined by quantifying its content in the dose of n-butanol site metabolites, having a significant modulating effect on the body weight of mice as described previously (Yang et al., 2023). Water and food were sufficient during the experiment, and mice were fed freely. The mouse feed was purchased from Xinjiang Tianyifeng biotechnology co, and its main ingredients are shown in Supplementary Table S4 . The body weight of the mice was weighed daily at 17:30.

Mice were transferred to new clean cages 10 hours prior to euthanasia to stop them from eating, and they were weighed and intraperitoneally injected with sodium barbiturate for anesthetization. Sampling of blood specimens from their ventral vein and then sacrificed immediately by cervical dislocation. Their livers, spleens, hearts, and epididymal and intestinal fat tissues were removed by dissecting their abdominal cavities on the bench. Blood was rinsed with saline, which was absorbed using filter paper and weighed and recorded. The epididymal fat rate, intestinal fat rate, and the corresponding organ indices were calculated. The Fat rate=(Fat weight/Mouse body weight)*100. The epididymal fat tissues were stored in 10% neutral phosphate buffered formalin for 24 h and then observed after hematoxylin-eosin (H&E) staining (Cardiff et al., 2014). The remaining organs were snap-frozen in liquid nitrogen and saved in a –80°C refrigerator. Blood samples were rested at 4°C for 3h and centrifuged at 3000g for 10 min. 50-uL of plasma was stored at -80°C in the refrigerator after quick-freezing with liquid nitrogen.

Mouse liver tissue was weighed (0.5 g) and mechanically homogenized in 4.5 mL anhydrous ethanol. The homogenized solution was centrifuged at 3000 g for 10 min. A supernatant of 50-µL was collected and TC (Richmond, 1973) and TG (Sullivan et al., 1985) concentration were determined. A 50-µL frozen serum sample was taken and melted on ice for the detection of LDL (Rifai et al., 1998) and HDL (Schectman et al., 1996) concentration. The assay was performed using enzyme-labeled instrument, and the method steps were strictly in accordance with the instructions of the Nanjing built kit (http://www.njjcbio.com).

A 0.3-g mouse liver tissue was taken for RNA extraction. Extracted total RNA from liver tissues using liquid nitrogen grinding using TRIzol reagent, 1 µl of total RNA solution per sample was taken for quantification and integrity analysis using a micro-nucleic acid detector and then reverse-transcribed into cDNA using a Tiangen reverse transcription kit. The mRNA expression levels of lipid synthesis-related genes, including peroxisome proliferator-activated receptor-alpha (PPARα) and transcription factor sterol regulatory element binding protein-1 (SREBP-1), and lipid metabolism-related genes, including hormone-sensitive lipase (HSL) and adipose triglyceride lipase (ATGL). The fluorescence real-time quantitative PCR (RT-qPCR) method was used to determine the results, the PCR reaction conditions were as follows: initially denaturing at 95°C for 5 min, then annealing at 58°C for 30 s and extending at 72°C for 30 s; and finally extending at 4°C for 10 min. The specific primers were designed according to the gene sequences ( Supplementary Table S3 ), and Bioengineering (Shanghai) Co. synthesized the primers.

The small intestinal contents were collected and stored at –80°C in a refrigerator. Bacterial DNA was extracted from the small intestinal contents. The V3-V4 region of the 16S rDNA gene was amplified using PCR. The PCR amplicons were quantified, and a pool was created to construct the MiSeq library for MiSeq sequencing. Paired-end raw sequence reads were obtained. The overlapping sequences were spliced followed by quality control and filtering of sequences. As a result of differentiating the samples, an OTU (operational taxonomic unit) clustering analysis was conducted. This allowed the analysis of different diversity indices based on OTU sets and detected the depth of high-throughput sequencing. The composition of gut microbiota was statistically analyzed at different taxonomic levels, referring to the corresponding taxonomic information. Based on the above analysis, the information on the composition of intestinal flora in the mycosubtilin treatment group was visualized by multivariate analysis and Student’s t-test. Bioinformatics analyses were performed on the high-quality sequencing data, and functional prediction of the 16S amplicon sequencing results was performed using the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) software package, in which the OTU abundance table was standardized by PICRUSt. The OTU abundance table was standardized by PICRUSt, and then the COG family information and KEGG Ortholog (KO) information corresponding to each OTU were obtained from the greengene id of each OTU; the abundance of each COG and the KO abundance were calculated. Based on the information of COG database, the description information of each COG, as well as its functional information, can be resolved from eggNOG database, so as to obtain the functional abundance spectrum; based on the information of KEGG database, the KO, Pathway, and EC information can be obtained, and the abundance of each functional category can be calculated based on the OTU abundance.

SPSS 19 software was used to perform the statistical analyses. The metric data were mainly expressed as mean ± standard deviation. Differences among multiple groups were analyzed using one-way analysis of variance (ANOVA). Different lowercase letters represent significant differences at P <0.05. In addition, graphs were plotted with Origin 8.5. OTU clustering was performed at a 97% similarity level using USEARCH software. The obtained OTUs were taxonomically annotated based on the silva138/16s_bacteria (https://www.arb-silva.de/) taxonomic databases. Using R language tools, community structure maps were plotted at the animal taxonomic level. Information on the analytical software and database used for the analysis and the version number is in Supplementary Table S5 . Majorbio Cloud Platform was used to analyze the data (www.majorbio.com).

In order to determine the concentration of mycosubtilin in the n-butanol extract of B. subtilis BS-Z15, the metabolites in the DMSO-dissolved n-butanol site were detected using HPLC separation. Mycosubtilin was characterized as peak IV with a peak time of 40–48 min. The concentration of mobile phase acetonitrile was 50%, which was similar to that used by Lin et al. ( Figure 1 ). A total of 20 mg of mycosubtilin was isolated and purified from 300 mg of the n-butanol site. According to a previous study, 90 mg/kg of n-butanol site metabolite could significantly affect the body weight gain of mice (Yang et al., 2023).

Current research suggests that, gavage of B. subtilis BS-Z15 and its metabolites were consistent with the effects of pregavage of B. subtilis BS-Z15 on slowing the body weight of mice and did not affect the amount of food they ate, so the effect of diet of mice on the results was not considered in this case. Both gavage of B. subtilis BS-Z15 n-butanol site metabolites and purified mycosubtilin reduced body weight gain in mice compared to controls; the body weights of mice in groups B and C were significantly lower as compared to the control group mice from day 4 and lasting until day 14 (P <0.05) ( Figure 2 ). The heart, kidney, and spleen indices of the mice at the end of the experiment were also observed, showing no significant differences in groups B and C as compared to group A. However, at the end of the experiment, the liver index of mice in group B was significantly lower than that in group A (P <0.05). Moreover, the liver indices in group C tended to decrease as compared to group A; however, the difference was insignificant ( Table 1 ).

The present study found, after gavage of B. subtilis BS-Z15 n-butanol site metabolite and purified mycosubtilin in mice, the obesity rate of epididymis in groups B and C was remarkably less than that in group A (P <0.05), while the percentage of intestinal fat had no significant differences in the treatment and untreated groups ( Figure 3A ). Investigating the H&E-stained sections of epididymal fat tissues further revealed that the mean size of epididymal adipocytes decreased significantly in treatment groups B and C compared with control group A ( Figures 3B, C ) (P <0.05). These results suggested that the metabolites of mycosubtilin and n-butanol sites could reduce epididymal fats in mice by reducing the size of epididymal adipocytes.

At the end of the gavage treatment, the gavage administration of B. subtilis BS-Z15 n-butanol site metabolite and purified mycosubtilin had no significant effect on liver TC concentration in mice in groups B and C as compared with control group A. Compared with the control group A, groups B and C had significantly lower liver TG concentration and plasma LDL levels (P <0.05). However, there was no significant change in plasma HDL levels in group B mice and a significant decrease in plasma HDL in group C mice compared to control group A (P <0.05) ( Figure 4 ).

To further investigate the role of mycosubtilin on adipogenesis, the mRNA expression levels of key liver lipid metabolism-related genes were detected using RT-qPCR. The results revealed that mycosubtilin treatment could noticeable enhancement the expression levels of lipolysis gene ATGL (P <0.05) in group C and decrease that of lipolysis gene SREBP. On the other side, the mycosubtilin treatment had no obviously impact on the expression levels of HSL and PPARa ( Figure 5 ).

Based on bacterial 16S rRNA gene sequencing, a total of 208,780 valid sequences, including 88,621,609 bases, were obtained for all the groups. The average number of valid sequences in Groups A and C are 35,515 and 34,078.33, respectively, and the number of bases is 15,066,529 and 14,474,007, respectively. Three samples per group, the sequencing coverage was above 0.99, which suggest that the probability of undetected sequences in the specimen is low and well-reflected the total gut microbiota in each sample ( Figure 6A ). Rank-Abundance curves were used to compare the differences in species composition of the mouse gut microbiome after MYCO treatment. The results showed that the shapes of the curves for the bacterial communities were roughly similar in the different groups, but deviated from the same abundance thresholds ( Figure 6B ). Shannon and chao indices were found to increase the diversity of the mouse gut community and community richness after MYCO treatment ( Figures 6C, D ). The number of shared taxa at the genus level between the groups are shown in Venn diagram. An average of 96 genera were detected in mycosubtilin-treated group C, and 70 genera were detected in group A of the blank control ( Figure 6E ). The principal coordinates analysis (PCoA) was executed to explore the dissimilarity among microbial communities obtained from different treatment of mouse small intestine. Mycosubtilin treated group C explained 68.36% and 27.33% of bacterial variation by PCoA1 and PCoA2 compared to blank control group A, respectively. ( Figure 6F ). The mycosubtilin-treated group C showed a significantly higher abundance of gut microbiota in mice as compared to that in the control group.

At the phylum level, Firmicutes were the most dominant group of microbes based on relative abundances. The gavage administration of mice with mycosubtilin reduced the relative abundance of Firmicutes (A: 95.65% vs. C: 74.40%) and increased that of Bacteroidota (A:2.04% vs. C:2.26%) as compared to the blank group A. Those results were consistent with those from previous studies investigating B. subtilis BS-Z15 metabolites’ n-butanol effects ( Figures 7A–C ). The mycosubtilin treatment could increase the relative abundance of Desulfobacterota (A: 1.73% vs. C: 16.41%). The relative abundance of Lactobacillus decreased with mycosubtilin at the genus level (A:93.76% vs. C:58.23%) and increased Desulfovibrio (A: 1.71% vs. C: 18.03%), Lachnospiraceae_NK4A136_group (A:0.06%, C:2.05%) abundance, thereby altering the composition of gut microbiota ( Supplementary Figure 1 ). Heatmap and Spearman’s correlation analyses of the top 30 genera in terms of genus-level abundance showed that Lactobacillus and Candidatus_Arthromitus were positively correlated with Fat, TC and body weight, and Lachnospiraceae_NK4A136_group and Desulfovibrio were negatively correlated with Fat, TC and body weight ( Figure 7D ).

Calculation of abundance profiles for each functional category by PICRUSt on OTU abundance revealed that the predicted relative abundance of COG functional compositions was more similar in all samples compared to the species composition, with Amino acid transport and metabolism, Carbohydrate transport and metabolism. Amino acid transport and metabolism, Carbohydrate transport and metabolism, and Energy production and conversion all showed high abundance in different groups. However, the function with the highest abundance was set as unknown, indicating that the COG database was not rich enough to fully reveal the functions of the microbial community ( Figures 8A, B ). Comparison of KEGG database showed that Metabolism was the most abundant at Level 1 and Membrane Transport was the most abundant at Level 2 ( Supplementary Table S4 ). For comparison, we further analyzed the COG-based functions in all samples showing the relative abundance of different functions, where the top three up-regulated relative expression abundance of functions in the treatment group compared to the control group were: Energy production and conversion (A: 5.22%, C: 6.97%), Signal transduction mechanisms (A: 3.59%, C: 5.59%), Inorganic ion transport and metabolism (A: 4.89%, C: 5.43%), The top three functional relative expression abundances were down-regulated, respectively: Replication, recombination and repair (A: 8.65%, C: 7.19%); Translation, ribosomal structure and biogenesis (A: 8.79%, C: 7.36%); Nucleotide transport and metabolism (A: 4.62%, C: 3.32%) ( Figure 8C ).