The probiotic strains involved in this study were provided by the Lactobacillus Bank of National Engineering Research Center of Dairy Health for Maternal and Child of Beijing Sanyuan Foods Co., Ltd. Escherichia coli (ATCC 25922), Staphylococcus aureus (ATCC 25923), Salmonella enteritidis (ATCC 14028), Bacillus cereus (ATCC 14579), Listeria monocytogenes (ATCC 19115), and Salmonella typhimurium (CMCCB 50071) were purchased from Guangdong Microbial Culture Collection Centre. All probiotic strains were cultured in MRS broth (Beijing Land Bridge Technology Co., Ltd.) at 37°C, and pathogenic bacteria were cultured in Luria-Bertani medium (Beijing Land Bridge Technology Co., Ltd.) at 37°C.

Bromocresol violet, an indicator, was added into folic acid casei medium (FACM, Difco) at pH 6.8, and the activated strains was cultured at 37°C for 1–3 days. FACM is a medium for the determination of folic acid, consisting of 10 g/L acid hydrolyzed casein (without vitamins), 40 g/L glucose, 40 g/L sodium acetate, 1 g/L dipotassium phosphate, 1 g/L potassium dihydrogen phosphate, etc. If the medium solution turns yellow, it indicates positive, indicating that the strain is capable of synthesizing FA (Laiño et al., 2012). The FA content of the samples was detected according to the method described in the previously published literature (Qiao et al., 2022), Aluminum foil was used to avoid exposing the samples to light for all the following steps. The pH of 4 mL samples was adjusted to 1.7 with 1 M HCl then kept in the dark for 2 min. The pH was then adjusted to 4.7 using 1 M NaOH and placed in the dark for 2 min. After making up to 10 g with water, the samples were centrifuged and defatted at 10000 × g for 20 min 4°C (Sigma, Gottingen, Germany). Finally, the supernatants were removed and filtered through a 0.22-μm syringe filter. The samples were analyzed by an UltiMate 3000 (Accela)-Q Exactive mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with an ACQUITY UPLC HSS T3 column (HSS T3, 50 mm × 2.1 mm i.d., 1.8 μm) being used for separation. The MS conditions were as follows: source temperature 110°C, desolvation temperature 350°C, desolvation gas flow (N2) 700 L h⁻¹, cone gas flow 30 L h⁻¹, and collision gas flow (Ar) 0.13 mL min⁻¹. Thermo Scientific Xcalibur software was used for system control and data management. Three parallel samples in each group.

In this study, L. reuteri B1-28 was selected to construct a knockout vector using a gene editing system. Primers and upstream and downstream homology arms were designed according to the target gene folA, and the knockout vector was constructed accordingly. A recombinant plasmid was constructed using a temperature-sensitive plasmid, pk18mobsacB (Harighi, 2009; Wang et al., 2018), and then introduced into the target strains for the clone screening. The knockout strain was obtained by electroporation transformation (Wang et al., 2018; Yamamoto et al., 2018; Yao et al., 2021). The transformed L. reuteri B1-28 were sequentially screened in an LB medium containing 50 μg/mL kanamycin to remove the target gene, and then in an MRS solid medium containing 10% sucrose to remove the plasmid. The resulting knockout strain was named ΔFolA.

The activated B1-28 and ΔFolA were passed to the third generation, then the bacterial solutions were diluted to 10⁸ CFU/mL with sterile PBS. They were inoculated into sterilized MRS broth and cow’s milk at 6%, respectively, and then cultured at 37°C for 30–35 h. The pH values of the culture medium were measured using a pH meter (Five Easy Plus, Mettler Toledo Instruments (Shanghai) Co., Ltd., Shanghai, China) before and after fermentation. The pH meter is calibrated prior to use, the electrode is placed in the standard solution sample, the measurement button is pressed, the electrode is held in the solution for 1–2 min to ensure that an accurate reading can be made, and the pH level is set once the reading has stabilized. The electrode is then placed into the sample solution to be tested to determine its pH level. FA content in the samples was determined by the same method as 2.2. Three parallel samples in each group.

The bacterial solution in the early stage of the platform was centrifuged, washed three times with PBS, then resuspended and fixed overnight with 2.5% glutaraldehyde phosphate buffer. Then, the samples were fixed with 1% osmic acid solution for 1–2 h. After washing, the bacteria cells were dehydrated using a graded ethanol series, and then dried in a critical point dryer (Quorum k850, Quorum UK Ltd., Nottingham, UK). The samples were fixed on the sample stage using conductive carbon adhesive, gold sprayed for about 30 s using an ion sputtering instrument (Hitachi MC1000, Hitachi Ltd., Tokyo, Japan), and observed under a scanning electron microscope (Hitachi SU3050) to collect images (Ali et al., 2022).

The activated B1-28, ΔFolA, Staphylococcus aureus and Bacillus cereus were passed to the third generation and then the bacterial solutions were diluted to 10⁸ CFU/mL with sterile PBS. The inoculation ring was used to dip a small amount of bacterial solution on the blood agar plate containing 5% defibrinated sheep blood (Solarbio, China) and incubated at 37°C for 48 h. The obvious transparent circle around the pathogenic indicator Staphylococcus aureus and Bacillus cereus showed a positive hemolysis, which was used as a positive control for the lactic acid bacteria hemolysis test. B1-28 and ΔFolA were inoculated at 2% to the MRS broth medium, then the strains were cultured at 37°C with shaking at 180 rpm. The optical density (OD) at 600 nm was measured using a UV–visible spectrophotometer (TU-1810, Beijing, China) at intervals of 2 h, and recorded continuously for 48 h.

The viability of B1-28 and ΔFolA in low pH and 0.3% bile salt environments was determined by measuring the number of viable colonies (Ali et al., 2023; Jiang et al., 2023). The concentration of cultured bacterial solution was diluted to 10⁸ CFU/mL. The bacterial solutions were centrifuged, then the resulting precipitate were resuspended in sterilized MRS broth medium with pH of 1.5, 2.0, 3.0, 6.25 (original medium pH, control) and 0.3% bile salt. The strains were incubated at 37°C with shaking at 180 rpm. Samples were taken at 0, 1, and 3 h, respectively, and the number of viable bacteria was recorded. The viable bacterial count of the bacterial solution in MRS broth medium at pH 6.25 was used as a control to calculate the survival rate of the bacterial solution at different pH values with the following formula: Survival rate % = N 1 N 0 × 100 % N0: number of viable bacterial count in 0 h.

N1: number of viable bacterial count in 1 or 3 h.

The adhesion ability of B1-28 and ΔFolA was assessed using the Caco-2 cell line (Li et al., 2023; Liu et al., 2023). Caco-2 cells were inoculated into 6-well plates at a density of 1 × 10⁵ cells per well and cultured to monolayer in complete medium containing Dulbecco’s modified eagle medium (DMEM, Gibco, 11965092), fetal bovine serum (Gibco, 10099141C), penicillin–streptomycin (Gibco, 15140122) and MEM non-essential amino acids solution (Gibco, 11140050). At 37°C and 5% CO₂, 1 mL of bacterial solution (10⁸ CFU/mL) per well was added to a six-well plate full of Caco-2 cells, then 1 mL DMEM was added into each well. After incubation at 37°C and 5% CO₂ for 2 h, the non-adherent bacteria were removed by washing with PBS 5 times, adding 0.5 mL 0.25% Trypsin–EDTA (Gibco, 25200072) for digestion for 3 min, adding 1 mL DMEM for repeated blowing, and recording the number of viable bacteria by plate counting method. The adhesion rate of B1-18 and ΔFolA to Caco-2 cells was determined by dividing the number of viable adherent bacteria (N2 h) by the initial number of Caco-2 cells or initial number of bacteria (N0 h).

The susceptibility of B1-28 and ΔFolA to 15 antibiotics, including penicillin, ampicillin, and tetracycline, etc., were performed according to the literature-adapted test methods (Wenzler et al., 2023). Briefly, the B1-28 and the ΔFolA were passed to the third generation, then the bacterial solutions were diluted to 10⁸ CFU/mL. 100 μL of the diluted bacterial solutions were taken and evenly spread on the MRS solid medium with a coating rod. After letting them stand for 20 min, pieces of drug-sensitive paper were placed equidistantly on the MRS solid medium. After anaerobic incubation at 37°C for 18–24 h, the diameter (mm) of inhibition zones was measured using a vernier caliper.

The antimicrobial capacities of the strains against six pathogenic bacteria were tested by the agar diffusion method (Chavez-Esquivel et al., 2021). The third-generation activated cultures of B1-28 and ΔFolA were diluted to 10⁸ CFU/mL, inoculated with 3%, and cultured for 36 h. After centrifugation, the supernatant was collected. The third-generation activated indicator strain were also diluted to 10⁸ CFU/mL, evenly spread on nutrient agar plates and placed in sterile air for 20–30 min. Subsequently, four Oxford cups were placed on each plate, with 200 μL of supernatant added to three Oxford cups, while an equal amount of sterile MRS broth medium was added to the fourth cup as a control. The plates were incubated at 37°C for 20–24 h, and the diameter of the inhibition circle (mm) was measured by vernier calipers.

Strains B1-28 and ΔFolA (10⁸ CFU/mL) were inoculated into sterilized MRS broth and cow’s milk at 6%, respectively, and cultured and fermentation at 37°C for 30 h. Metabolites in MRS Medium supernatant and cow’s milk after fermentation were detected. Six parallel samples in each group. The samples were placed in EP tubes and resuspended with prechilled 80% methanol by well vortex. Then the samples were melted on ice and whirled for 30 s. After the sonification for 6 min, they were centrifuged at 5,000 rpm, 4°C for 1 min. The supernatant was freeze-dried and dissolved with 10% methanol. Untargeted metabolic analysis was performed using a Vanquish UHPLC system (ThermoFisher, Germany) coupled with an Orbitrap Q Exactive^™ HF mass spectrometer (Thermo Fisher, Germany) in Novogene Co., Ltd. (Beijing, China) platform as described elsewhere (Sellick et al., 2011; Yuan et al., 2012). Samples were injected onto a Hypersil Goldcolumn (100 × 2.1 mm, 1.9 μm) using a 12-min linear gradient at a flow rate of 0.2 mL/min. The eluents for the positive and negative polarity modes were eluent A (0.1% FA in Water) and eluent B (Methanol). The solvent gradient was set as follows: 2% B, 1.5 min; 2–85% B, 3 min; 85–100% B, 10 min;100–2% B, 10.1 min; 2% B, 12 min. Q Exactive^™ HF mass spectrometer was operated in positive/negative polarity mode with spray voltage of 3.5 kV, capillary temperature of 320°C, sheath gas flow rate of 35 psi and aux gas flow rate of 10 L/min, S-lens RF level of 60, Aux gas heater temperature of 350°C. The raw data files generated by UHPLC–MS/MS were processed using the Compound Discoverer 3.3 (CD3.3, ThermoFisher) to perform peak alignment, peak picking, and quantitation for each metabolite.

RNA was extracted from the logarithmic strains using the RNAprep Pure Cell/Total RNA Extraction Kit (Centrifugal Column Type, Catalog No. DP430), and RNA integrity and quantity were accurately detected using the Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA). RNA sequencing was performed in the Illumina sequencing platform of Novogene Co., Ltd. (Beijing, China). Differential expression analyses for both conditions/groups were performed using the DESeq2 R package (version 1.20.0). GO enrichment analysis of differentially expressed genes was achieved by GOseq R package software. Statistical enrichment of differentially expressed genes in the KEGG pathway was analyzed using KOBAS software (Mao et al., 2005; Kanehisa et al., 2007).

Unless otherwise stated, the data from this study are expressed as mean ± standard deviation (SD). Two-tailed unpaired t-tests were used for two groups and one-way analysis of variance (ANOVA) was used for multiple groups. Excel 2010 software was used to organize basic data, and GraphPad Prism 8.0.2 software was used for data analysis and graphing. Calculated values of p < 0.05 were considered statistically significant.

Bromocresol violet was added into FACM to observe the FA production of the tested strains. Among 42 strains of human breast milk-derived strains, the medium solution of 20 strains turned significantly yellow, indicating that the strain had the capacity of producing FA. The 20 strains of human breast milk-derived strains obtained from the initial screening were tested by liquid chromatography mass spectrometry (LC–MS). The results were shown in Figure 1A, all strains produced folic acid with varying yields, and the strain with relatively high FA content was L. reuteri B1-28 (60.72 ng/mL), which could be used for further experiments.

The human breast milk-derived strains B1-28 and ΔFolA were cultured in FACM and in cow’s milk to test the FA content. As shown in Figure 1B, the FA production of knockout strain ΔFolA was approximately 150 ng/mL in FACM, which was about three times higher than that of the wild-type strain. After fermentation in milk, the folate content of the knockout strain was as high as 1,700 ng/mL, which was about 2.5 folds that of the wild-type strain. This indicates a significant improvement in FA production after folA gene knockout of B1-28 strain. Scanning electron microscopy images revealed that both the B1-28 and ΔFolA strains exhibited rod-shaped morphology (Figures 1C–H). However, the surface of the ΔFolA (Figures 1F–H) was rougher than that of the B1-28 (Figures 1C–E), suggesting that knockout of the folA gene affected the surface morphology of the strains.

In Figure 2A, both Staphylococcus aureus and Bacillus cereus showed obvious hemolytic circles, whereas none of the human breast milk-derived strains exhibited hemolytic activity. The OD600 nm values of B1-28 and ΔFolA during 40 h of continuous incubation in MRS broth medium are shown in Figure 2B. The B1-28 grew faster from 2 to 6 h than the ΔFolA (p < 0.05). The pH values of the broth before and after fermentation were measured as shown in Figure 2C. The pH value of the milk fermented by the ΔFolA was significantly lower than that of the milk fermented by the B1-28 (p < 0.05). Based on the above experimental results, both strains were tested for acid and bile salt tolerance, and the number of viable bacteria was recorded, as shown in Figure 2D. No difference in the number of viable bacteria were observed between the two strains. Figure 2E shows that the survival rates of the B1-28 and the ΔFolA in MRS broth medium containing 0.3% bile salt were higher than 70%, while the B1-28 had a better ability to tolerate bile salts than the ΔFolA. Based on the previous scanning electron microscopy results, ΔFolA had a rougher bacterial surface than B1-28. ΔFolA’s adhesion rate and the number of adherent cells were approximately 1.4 folds higher than those of B1-28 (p < 0.01), indicating that ΔFolA had a stronger adhesion ability to the human colon cancer cells Caco-2 than B1-28 (Figure 2F).

Antibiotic sensitivity of probiotics is one of the most important indicators of their safety for use in food (Pierce et al., 2023; Wenzler et al., 2023). In this study, the sensitivities of B1-28 and ΔFolA strains to 15 common antibiotics were tested, and the results are presented in Table 1. There was no statistically significant difference between B1-28 and ΔFolA in sensitivity to tetracycline and amoxicillin. The diameters of the drug-sensitive circles of B1-28 were significantly smaller than those of ΔFolA in penicillin, cefazolin, erythromycin, chloramphenicol, and roxithromycin (p < 0.05). Conversely, the diameters of drug-sensitive circles of B1-28 were significantly larger than those of ΔFolA in ampicillin, ceftriaxone, augmentin, and enrofloxacin (p < 0.05). The folA gene knockout strain showed resistance to oxacillin, gentamicin, and neomycin, which was significantly different from B1-28 (p < 0.05). Both the B1-28 and the ΔFolA showed resistance to kanamycin, streptomycin, norfloxacin, enoxacin and polymyxin B, suggesting that up-regulation of folA expression as a target of drug action affects microbial susceptibility to antibiotics (Shi et al., 2021; Angermayr et al., 2022).

The fermentation broths of strains B1-28 and ΔFolA were used to evaluate their antibacterial activities (Table 2). Both B1-28 and ΔFolA exhibited specific inhibitory effects on different pathogenic bacteria, with varied degrees of antibacterial activities. Specifically, B1-28 was significantly better than ΔFolA in inhibiting Staphylococcus aureus and Salmonella typhimurium (p < 0.05). Conversely, ΔFolA exhibited stronger inhibitory capabilities than B1-28 against Escherichia coli, Salmonella enteritidis, and Bacillus cereus (p < 0.05).

The PCA plots of the B1-28 and ΔFolA in positive and negative ion modes during the growth period in MRS broth medium and cow’s milk are displayed in Figures 3A,B. They both showed obvious regular differences. A total of 619 positive ion mode metabolites and 251 negative ion mode metabolites were identified in the 24 samples (Supplementary Table S1). Differential metabolites were screened by setting threshold values as VIP > 1.0, 1.2 or FC < 0.833 and p-value <0.833 (Wen et al., 2017). The values of the screened differential metabolites are shown in Table 3 and the top 20 differential metabolites after up- and down-regulation are shown in Figures 3C–F. The B1-28 and ΔFolA produced metabolic differential products dominated by organic acids in MRS broth in positive and negative ion modes, including citric acid, L-serine, adenine, α-ketoglutaric acid and other differential metabolites. And, in milk, the B1-28 and ΔFolA produced 60 differential metabolites, such as citric acid and indole-3-acetic acid, in negative ion mode, and 143 significantly up-regulated differential metabolites, such as hippuric acid, in positive ion mode.

Metabolomics analysis is an effective method to compare the differential metabolite expression between wild-type and knockout strains. Hierarchical clustering analysis of the two sets of differential metabolites obtained showed that strains B1-28 and ΔFolA fermented different media into different differential metabolites, including organic acids and their derivatives, organic oxids, indoles and their derivatives, carboxylic acids and their derivatives of purines. These differences in metabolite abundance are shown in the clustered heatmap (Figures 4A–D). It can be seen from the heat map that the metabolic patterns were similar within the same group. After fermentation, the ΔFolA produced higher levels of citric acid, pyridoxine, 2-furanic acid, L-serine, and L-tyrosine than B1-28 (p < 0.01). Moreover, the strains produced more metabolites by fermentation in cow’s milk than those in the basic medium, suggesting that some metabolites are more readily activated in cow’s milk. Fermented milk is considered to be the most ideal vehicle for delivering probiotics into the body, which can perfectly combine the health benefits of probiotics and milk (Sun et al., 2023). By comparing the different metabolites, no toxic metabolites were found (Supplementary Table S1), which need to be evaluated in further safety experiments.

KEGG is a powerful tool for metabolic analysis and metabolic network studies in organisms. Pathway enrichment enabled the identification of the most prominent biochemical metabolic pathways in which the differential metabolites are involved, as shown in Figures 5A–D, including the citrate cycle, biosynthesis of amino acids, FA biosynthesis, purine metabolism, and butyric acid metabolism, and other metabolic pathways. In MRS broth, significant differences were observed in the citrate cycle (p = 0.008), glyoxylate and dicarboxylate metabolism (p = 0.045), and C5-branched dicarboxylic acid metabolism (p = 0.045) in positive ion mode, while in the negative ion mode, differences observed in methane and carbon metabolism (p = 0.025) were related to citric acid, L-serine and L-tyrosine, which was consistent with the results of the above differential metabolites. In cow’s milk, significant differences were observed in phenylalanine metabolism (p = 0.028) and aminobenzoic acid degradation (p = 0.028), and the enriched metabolites included phenylglyoxylic acid, phenylacetylglutamine, hippuric acid, p-hydroxybenzaldehyde, hydroquinone, and 4-hydroxybenzoic acid in the positive ionic mode. In negative ionic mode, the metabolites enriched in FA biosynthesis included dihydrofolate and 4-aminobenzoic acid, and the reason for no significant difference is that FA is unstable and prone to oxidative decomposition (Mahara et al., 2019). These metabolic pathways for the folA gene to exert probiotic efficacy in vivo.

To investigate the expression of the strains at the RNA level after folA knockout, B1-28 and ΔFolA were analyzed by transcriptome sequencing. PCA analysis was conducted on the gene expression values (FPKM) of all samples, as shown in Figure 6A. Samples were scattered between groups and clustered within groups. The cDNA libraries were constructed by Illumina sequencing with three or more replicates per strain. The Q20 of each library was >97.84%, and the Q30 was >93.71%, indicating good sequencing results. The correlation coefficients of samples within and between groups were calculated based on the FPKM values of all the genes in each sample, where a higher correlation coefficient between samples (closer to 1) indicates higher similarity in expression patterns between the samples. The B1-28 and ΔFolA samples were divided into two distinct categories with good intra-group correlation (Figure 6B), where the Pearson correlation coefficient quadratic sum R2 exceeding 0.898 indicates relatively ideal sampling and experimental conditions for subsequent transcriptome analyses. The differential genes with changes in expression were screened in the sequencing library, and the complete list of differential genes is presented in Supplementary Table S2. The results are represented in the Volcano plot in Figure 6C. The ΔFolA showed significant differences in 460 genes compared to the B1-28, of which 229 genes were up-regulated and 231 genes were down-regulated, suggesting that knockout of the folA gene influenced the transcriptional level of the entire gene. Cluster analysis of the differential gene set is illustrated in Figure 6D. The major up-regulated genes were associated with the ABC transporter protein family, while the downregulation of the expression of genes encoding dihydrofolate reductase and glutamine synthetase was associated with knockout of the strain’s dihydrofolate reductase-encoding gene, folA.

The Gene Ontology (GO) function was analyzed on 460 differentially expressed genes, as shown in Figures 7A–C. In B1-28 and ΔFolA, differentially expressed genes (DEGs) was mainly involved in RNA processing, ribosomal structural components, organic acid transport and aminoacyl-tRNA ligase activity. The results showed that the indirect transcriptional regulation target of folA affected the formation of nucleotides and purines, and the significant increase of citric acid and other organic acid metabolites could improve the strain’s ability to inhibit Escherichia coli (Ji et al., 2023). The down-regulated genes may be the result of the target-dependent or compensatory response to the normal function of folA, which can not directly prove that folA promotes or suppresses the expression of these genes. Therefore, further experimental validation and functional studies are needed to determine the exact role of folA in the metabolic process.

To further analyze the impact of folA gene knockout on metabolic pathways of strains, 229 up-regulated genes and 231 down-regulated genes were analyzed for metabolic pathways by KEGG (Supplementary Table S3). Up-regulated genes were primarily in pathways such as amino acid biosynthesis, and secondary metabolite biosynthesis, while down-regulated genes were primarily in in the microorganisms in the pathways of nucleotide metabolism, pyrimidine metabolism, purine metabolism, carbon metabolism, and ABC transport protein. On this basis, the differential expression genes annotated by KEGG were further explored (Figure 7D). The enzymes encoded by differential genes were mainly involved in ribosome, aminoacyl-tRNA biosynthesis, and propanoate metabolism, and the expression of EDGs was down-regulated in ribosomal and aminoacyl-tRNA biosynthesis (Padj <0.001). The DEGs enrichment in propionic acid metabolism was significantly higher than that in the control group (Padj = 0.027), which was consistent with the functional analysis of GO. These genes and metabolic processes discovered through screening are related to amino acid, ribosomal, and purine synthesis pathways. The results indicate that folA gene knockout had a great impact on amino acids metabolic pathways and folate biosynthesis pathways in human breast milk-derived L. reuteri B1-28 and had a regulatory effect on genes involved in FA synthesis.