In this study, four bifidobacterial strains were investigated: BB-69, Bbm-19, BX-18, and B8762. The BB-69 and B8762 strains were both isolated from the intestinal tract of healthy infants in 2017, while the BX-18 strain was isolated in 2018 from the feces of a long-lived elderly person in Enshi, Hubei. In contrast, the Bbm-19 strain was isolated from healthy breast milk in 2017. All strains were preserved and obtained from the Key Laboratory of Dairy Biotechnology and Engineering, Ministry of Education, Inner Mongolia Agricultural University, China. The complete genome sequences of the four Bifidobacterium strains have been deposited in the iLABdb database,¹ a specialized database for lactic acid bacteria genomes (Jin et al., 2023). The genomic data, including genome assembly and annotation, can be accessed through the website or by email request. These strains are also available to researchers through formal material transfer agreement procedures.

In the modified MRS liquid medium sterilized at pH 2.5 (adjusted with 1 mol/L HCl), 3.5 g/L pepsin was added, and the sterilization was filtered and sterilized with a 0.22 μm microporous filter membrane to make a simulated gastric juice.

In the modified MRS liquid medium sterilized at pH 8.0 (adjusted with 0.1 mol/L NaOH), 0.10% trypsin and 1.80% bovine bile salt were added and sterilized by 0.22 μm microporous filter membrane to prepare a simulated intestinal fluid.

We added 5 × 10⁷ CFU/mL of the prepared bacterial solution to the simulated gastric juice, incubated at 37°C for 3 h, and measured the number of viable bacteria at 0 h and 3 h by the modified MRS agar medium pouring method. The bacterial artificial gastric juice with a pH of 2.5 was taken and cultured for 3 h, transferred into the artificial intestinal fluid, and the number of viable bacteria was measured by sampling after 8 h of anaerobic culture at 37°C, all anaerobic operations are carried out at the anaerobic workstation and the survival rate was calculated as follows.

N₀—0 h number of viable bacteria; N₁—the number of viable bacteria after 3 h of digestion by simulated gastric juice; N₂—Number of viable bacteria after 8 h digestion of simulated intestinal fluid.

High-performance liquid chromatography-grade methanol and acetonitrile (Merck, Germany) and formic acid (Shanghai Aladdin Bio-Chem Technology Co., Ltd., Shanghai, China) were used in the study. The modified MRS liquid medium comprised the following components: (10 g/L), beef extract (5 g/L), yeast extract (4 g/L), glucose (20 g/L), Tween-80 (1.08 g/L), potassium dihydrogen phosphate (2 g/L), sodium acetate (5 g/L), ammonium citrate tribasic (2 g/L), magnesium sulfate heptahydrate (0.2 g/L), manganese sulfate tetrahydrate (0.05 g/L), and L-cysteine hydrochloride (0.5 g/L). The final pH of the medium was adjusted to 6.20 ± 0.02 using 1 mol/L hydrochloric acid or 1 mol/L sodium hydroxide.

The test strains were obtained as frozen stock stored at −80°C. They were inoculated and reactivated in the modified MRS liquid medium, grown under constant temperature at 37°C. Following reactivation, the test strains were continued to grow as seed culture (2% cell density) at 37°C for 18 h. Subsequently, they were inoculated into fresh MRS liquid medium (5% cell density) for further cultivation at 37°C. Constant temperature anaerobic culture at 37°C to stable period. All strains were maintained under anaerobic conditions throughout the entire cultivation period. Microbial morphology was observed during the passage to ensure the purity of the test strains.

All strains were cultured until the end of the logarithmic phase, at which point the cultivation was terminated. They were then sterilized and inactivated at 90°C for 15 min. Following this, the inactivated broth was centrifuged at 14,000 rpm to pellet the cells, which were washed three times with sterile purified water and then subjected to vacuum freeze-drying. The samples were stored at −80°C until extraction of intracellular metabolites.

The frozen samples were thawed on ice. Approximately 40 ± 1 mg of each sample was weighed into a centrifuge tube, homogenized using a ball mill, and then centrifuged at 3,000 rpm for 30–50 s at 4°C. Then, 400 μL of 80% methanol–water internal standard extraction agent was added, and the mixture was vortexed for 3 min to ensure complete suspension of the sample. The sample was rapidly frozen in liquid nitrogen for 5 min, thawed on ice, vortexed for another 3 min, and this process was repeated three times. After centrifugation at 12,000 rpm for 10 min at 4°C, 300 μL of the supernatant was transferred to a new centrifuge tube. The extraction process was repeated under the same conditions to obtain an additional 200 μL of supernatant for analysis. Samples were filtered and transferred to appropriate sample vials for subsequent metabolomics analysis (Grün et al., 2019).

The analysis was performed on an ultra-performance liquid chromatography (UPLC) (ExionLC AD; AB Sciex LLC, Framingham, MA, United States) and tandem mass spectrometry (MS/MS) (QTRAP; AB Sciex LLC, Framingham, MA, United States).

All samples were subjected to three biological replicates, and the data are presented as mean ± standard deviation.

The mass spectra data underwent peak extraction, noise reduction, deconvolution, and peak alignment. Comprehensive library searching and identification were performed using the MassBank,² ReSpect,³ and GNPS⁴ databases. Utilizing principles of chemometrics and multivariate statistics, the multidimensional data were subjected to dimensionality reduction and classification analysis. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) of the three-dimensional data matrix were conducted using R software to identify differential metabolites. Hierarchical cluster analysis was also performed with R software, generating heatmaps (using the pheatmap package) to characterize the accumulation patterns of metabolites among samples (Chong and Xia, 2018). The identified compounds were further analyzed for metabolic pathways using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (Ogata et al., 1999).⁵ The sample preparation and analysis workflow are illustrated in Figure 1.

In this study, the absorbance value (OD_600nm) was used to construct the growth curves of four strains of Bifidobacterium, we strictly controlled the sampling points to ensure the reliability and reproducibility of the data, and we performed three biological replicates at each time point, and the results were plotted as mean ± standard deviations, as can be seen from Figure 2 (Supplementary Table S3), BB-69 entered the stable phase at 15 h, Bbm-19 entered the stable phase at 12 h, BX-18 and B8762 entered the stable phase at 14 h, and the nutrients in the culture medium were gradually depleted, and the metabolites began to accumulate in large quantities. Therefore, the stable period was selected to harvest the bacteria for follow-up research.

We performed gastrointestinal fluid tolerance tests on Bbm-19 and BX-18. Bifidobacteria were inoculated in simulated gastrointestinal fluids, and their survival rate was monitored to assess their tolerance in the environment of gastric acid and bile salts. The specific results are as follows:

Bbm-19 also demonstrates good acid and bile salt tolerance, with a survival rate of 61.16% after 3 h in gastric juice at pH 2.5 and a survival rate of 95.09% after 8 h in intestinal fluid at pH 8.0.

BX-18 exhibits good acid and bile salt tolerance, with a survival rate of 42.02% after 3 h in gastric juice at pH 2.5 and a survival rate of 37.50% after 8 h in intestinal fluid at pH 8.0.

Previous research by the laboratory team has shown that: BB-69 exhibits good acid and bile salt tolerance, with a survival rate of 73.54% after 3 h in gastric juice at pH 2.5 and a survival rate of 91.29% after 8 h in intestinal fluid at pH 8.0. These physiological characteristics indicate that this strain possesses essential probiotic properties. Additionally, results from an animal study suggest that it alleviates symptoms of experimental colitis in mice through mechanisms such as enhancing the structure and function of the intestinal mucosal barrier, reducing oxidative stress damage in the intestine, inhibiting excessive immune activation, and lowering inflammation levels (Han et al., 2024).

B8762 also shows good acid and bile salt tolerance, with a survival rate of 61.26% after 3 h in gastric juice at pH 2.5 and a survival rate of 87.70% after 8 h in intestinal fluid at pH 8.0. Furthermore, a study indicates that both live Bifidobacterium longum subsp. infantis B8762 and heat-killed cells can alleviate DSS-induced colitis in rats (Li et al., 2024).

In summary, the four strains of Bifidobacterium have potential probiotic properties and growth characteristics, which provide basic research for the study of their physiological and metabolic properties.

A total of 12 samples from four bifidobacterial strains in triplicate were investigated using untargeted metabolomics analysis. To assess the intra- and intergroup statistical differences in total metabolites, the metabolomics data were first normalized, followed by multivariate statistical analysis, including PCA and OPLS-DA.

In the PCA, the symbols representing different samples showed distinct strain-based clustering patterns on the PCA score plot (Figure 3), with principal components PC1 and PC2 accounting for 40.03 and 28.10% of the total variance, respectively. Afterwards, OPLS-DA was employed to identify differential metabolites between bifidobacterial strains. The predictive parameters of the OPLS-DA model include R²X, R²Y, and Q², where R²X and R²Y represent the explanatory rates of the model, respectively, and Q² indicates the predictive ability of the model. Values closer to 1 for these indicators reflect a more stable and reliable model. A Q² value greater than 0.5 is considered effective, while a value exceeding 0.9 is considered excellent (Zhang et al., 2015). In the OPLS-DA model, the pairwise comparisons between samples of BB-69, B8762, BX-18, and Bbm-19 were yielded high Q² (0.984–0.994), R²X (0.809–0.852), and R²Y (range = 1; Supplementary Table S1), demonstrating high predictability and strong fitness of the models.

The S-plots of the pairwise comparisons by OPLS-DA are shown in Figure 3. A total of 1,340 metabolites (B8762 vs. BX-18), 1,340 metabolites (B8762 vs. Bbm-19), 1,340 metabolites (BX-18 vs. BB-69), 1,338 metabolites (B8762 vs. Bbm-19), 1,339 metabolites (BX-18 vs. Bbm-19), and 1,338 metabolites (BB-69 vs. Bbm-19) displayed VIP values greater than 1, and were positioned furthest from the origin. Therefore, these metabolites can serve as biomarkers for distinguishing different Bifidobacterium strains.

The top 20 differential metabolites with the highest VIP values for each strain are presented in Figure 4 (Croux et al., 2000). In the comparison between B8762 and BB-69, notable metabolites included 3-epideoxycholic acid, 8-methylnonenoate, and hydroxyphenyllactic acid (Figure 5A). For the B8762 and Bbm-19 comparison, key metabolites were equilenin, tropate, and phenyllactate (Figure 5B). In the B8762 and BX-18 comparison, significant metabolites included Lys-Ser, Phe-Hyp, and Ser-Val (Figure 5C). The comparison between Bbm-19 and BB-69 highlighted N-arachidonoylglycine, L-cystine, and 3-epideoxycholic acid (Figure 5D). In the BX-18 and BB-69 comparison, the main metabolites were L-cystine, isocytosine, and cytosine (Figure 5E), while the BX-18 and Bbm-19 comparison revealed tiglylglycine, equilenin, and 2-O-methylcytidine (Figure 5F).

We then created dynamic plots of differential metabolite distributions to illustrate metabolites of greatest differences between compared samples (Figure 6). In the comparison between B8762 and BB-69, the metabolite with the highest fold change (FC) was (24S)-24,25-dihydroxyvitamin D3, while the one with the lowest FC was Glu-Val. For B8762 versus BX-18, the highest FC was observed in D-maltopentaose, and the lowest in Ser-Gln. In the B8762 and Bbm-19 comparison, equilenin exhibited the highest FC, with Ser-Gln having the lowest. In the BX-18 versus BB-69 comparison, 3-epideoxycholic acid showed the highest FC, while L-theanine had the lowest. For BX-18 compared to Bbm-19, For BX-18 compared to Bbm-19, cytosine had the highest FC, and D-maltopentaose had the lowest. Finally, in the Bbm-19 versus BB-69 comparison, (24S)-24,25-Dihydroxyvitamin D3 again showed the highest FC, with glycerol 2-phosphate having the lowest.

To gain deeper insights into the metabolic differences among the four bifidobacterial strains, we screened differential metabolites from a total of 1,340 metabolites (VIP value >1) based on the threshold of FC ≥2 or ≤0.5 and a p-value <0.05 from the t-test. The results are displayed in volcano plots (Figure 7). The number of significantly different metabolites identified between pairwise comparisons: in the comparison between B8762 and BB-69, 838 significantly different metabolites were detected, with 532 upregulated and 306 downregulated; in the B8762 and Bbm-19 comparison, 907 significantly different metabolites were identified, including 520 upregulated and 387 downregulated; in the B8762 versus BX-18 comparison, 863 significantly different metabolites were found, with 525 upregulated and 338 downregulated; in the Bbm-19 versus BB-69 comparison, 828 significantly different metabolites were identified, with 412 upregulated and 416 downregulated; for the BX-18 versus BB-69 comparison, 824 significantly different metabolites were noted, with 405 upregulated and 419 downregulated; and in the BX-18 versus Bbm-19 comparison, 822 significantly different metabolites were found, consisting of 389 upregulated and 433 downregulated. These findings provide important information for further functional analysis and exploration of metabolic pathways.

To visually represent the variations in differential metabolite profiles by biochemical classes, the original data for the selected metabolites were processed using the unit variance scaling method and were subjected to hierarchical clustering analysis, with results displayed in a heatmap (Figure 8). The level in the figure indicates the name of the sample, the longitudinal information is the differential metabolite, the red represents the concentration of the high-content substance, and the green represents the concentration of the low-content substance.

The differential metabolites among the four bifidobacterial strains primarily belonged to the biochemical classes of amino acids and their metabolites, benzene and substituted derivatives, organic acids and their derivatives, nucleotides and their metabolites, carbohydrates and their metabolites, heterocyclic compounds, alcohols and amines, fatty acids, coenzymes and vitamins, glycerophospholipids, aldehydes, ketones, and esters, and hormones and hormone-related compounds. Notably, the distribution of biochemical class of differential metabolites exhibited substantial variations between strains.

In order to study the trend of the relative content of metabolites in different groups, the relative contents of all differential metabolites identified according to the screening criteria in all comparison groups were treated with UV (unit variance scaling) (the mean of the original data minus the mean of the variable and divided by the standard deviation of the variable), and then the K-means cluster analysis was performed. The abscissa represents the sample grouping, the ordinate represents the normalized relative amount of metabolites, the subclass represents the metabolite class number of the same trend, and total: represents the number of metabolites in the class. The results indicated that the trends in these metabolite changes could be categorized into two distinct clusters. In cluster 1, a total of 524 metabolites were ranked in the following order of abundance: Bbm-19 > BB-69 > BX-18 > B8762 (Figure 9A). Conversely, in cluster 2, the order for the remaining 721 metabolites was: B8762 > BX-18 > BB-69 > Bbm-19 (Figure 9B).