Samples of fermented food (pickled Chinese cabbage, pickled chili, pickled cabbage, pickled carrot, pickled leaf mustard, pickled cowpea, and pickled ginger) used for the isolation of LAB were purchased from different local markets in Shaanxi, China. Samples were put in sterilized tubes and transported to the laboratory at Shaanxi Institute of Microbiology (Xi’an, China) within 12 h. The isolation was performed using the method described before (Pringsulaka et al., 2012) with slight modification. The fermented food (1 g) was weighed, minced with sterile scissors, and was added to 10 mL sterile water and mixed thoroughly. Each sample was serially diluted with 0.85% (w/v) sterile saline solution and cultured on the de Man, Rogosa Sharpe (MRS) agar supplemented with 2% (w/v) CaCO₃ at 37°C for 48 h. The colonies, which were round, milky, and had a clear halo surrounding them, were picked and purified by streaking on MRS plates. Cell morphology was examined by optical microscopy (100X, Olympus BX41, Japan) and Gram staining. The isolates were stored at –80°C in MRS broth containing glycerol (25%, v/v).

The strains were preliminarily selected by incubating them in MRS with and without 0.3% bile salts (0% as control) at an inoculum size of 1% (v/v). The strains were then incubated at 37°C for 4 h. The OD values at 600 nm were measured after 4 h cultivation with and without 0.3% bile salts, respectively. Bile tolerance was expressed as follows:

The tolerance against low pH was determined on the strains with high bile salt tolerance. The strains were incubated in MRS with pH 2.0, pH 2.5, and pH 6.4 (as a control) for 24 h. The OD values at 600 nm were measured at 0 h and 24 h cultivation, respectively. The survival rate was calculated as follows:

Molecular identification of selected strains was performed according to the method described before (Wang et al., 2020). Briefly, microbial genomic DNA was extracted by using TaKaRa Universal DNA extraction kit (Cat# 9765) (TaKaRa Bio Inc., Shiga, Japan). Bacterial 16S rRNA gene sequences were PCR-amplified from each sample using the 27F-1492R primers. PCR was carried out with an automated thermal cycler (Biometra Germany) using Taq polymerase (TaKaRa Bio, Inc., Shiga, Japan). Sequences were determined at BGI Biotechnology (Shenzheng, China) and analyzed using BLAST (NCBI, Bethesda, MD, United States). Sequences obtained in this study were submitted to the GenBank database with accession numbers from OK021606 to OK021632 and the neighbor-joining phylogenetic tree was constructed using MEGA 5.05.

The survival of selected LAB through simulated gastrointestinal juices was examined according to a previous report with slight modification (Jamyuang et al., 2019). After two generations of cultivation, the LAB strain was inoculated with 1% (v/v) inoculum in MRS liquid medium and incubated for 18 h at 37°C. Cells were harvested by centrifugation (8000 rpm at 4°C for 5 min) and cell number was adjusted to 10⁹ CFU/mL by adding PBS solution. The obtained bacterial suspensions (1 mL) were inoculated to 9 mL of filter sterilized simulated gastric juice solution [3 g/L pepsin (p7000, sigma), pH 2.0] at 37°C for 60 min. The gastric juice solution was removed by centrifugation (8000 rpm, 5 min) and subsequently re-suspended in 9 mL of filter sterilized simulated small intestinal juice solution [1 g/L trypsin (T105532, Aladdin), 0.3% bile salt, pH 8.0]. Samples were incubated statically at 37°C for 120 min. Then, the intestinal juice was removed by centrifugation (8000 rpm at 4°C for 5 min). The cell pellet was suspended in a sterile 0.85% NaCl (w/v) solution. Survival cells were counted on MRS plates. The survival rate (%) was calculated by the following equation:

where N₀ and N_t represent viable bacterial cells before and after growth in the simulated gastrointestinal tract, respectively.

After overnight incubation, bacterial cultures were centrifuged at 8000 rpm for 5 min (4°C). The cell pellets were washed twice and re-suspended in PBS buffer (pH 7.20) to achieve OD₆₀₀ value of approximately 0.4 (labeled as A₀). The cell suspension (3 mL) was blended with 1 mL chloroform and the mixture was vortexed for 30 s. Then it was left at room temperature for 30 min to separate the aqueous and organic phases. The aqueous phase (1 mL) was removed carefully, and the absorbance (A_X) was measured at 600 nm. The hydrophobicity was calculated according to following equation:

The aggregation abilities of selected LAB were evaluated in a previous study (Choi et al., 2018). Overnight cultured LAB strains were collected by centrifugation (8000 rpm, 4°C, 15 min). The cell pellets were washed twice and re-suspended to 10⁸ CFU/mL with PBS buffer (pH 7.2).

For the auto-aggregation assay, 4 mL LAB suspension was incubated at 37°C for 6 h without agitation. Then 0.5 mL of suspension was transferred to a new tube and mixed with 1.5 mL PBS buffer. The OD values at 600 nm were measured and the auto-aggregation rate was calculated as follows:

where A₀ and A_t represent the OD₆₀₀ values after 0 and 6 h incubation, respectively.

For co-aggregation assay, Shigella flexneri CMCC51574, Salmonella paratyphi B CMCC50094, and Escherichia coli CMCC44102 were used as pathogenic strains. The LAB strains and the pathogenic strains liquid concentration were adjusted to 10⁸ CFU/mL as described above. Equal volumes (2 mL) of LAB strains and pathogenic strains were mixed and vortexed for 10 s followed by an incubation at 37°C for 4 h without shaking. Cell suspensions of each single strain were used as controls. Based on the method proposed by Handley et al. (1987), co-aggregation percentage was calculated as follows:

whereA_LAB, A_pat, and A_mix represents the OD₆₀₀ values of control tubes and mixture after 4 h incubation, respectively.

The adhesion ability to Caco-2 cells was evaluated following the method described by Srimahaeak et al. (2021) with some modifications. The Caco-2 cells were cultured in DMEM-high glucose medium (HyClone, United States) supplemented with 10% (v/v) fetal bovine serum (FBS) (Bioexplorer, United States), 1% (v/v) nonessential amino acids solution (AAS) (Solarbio, China) and 1% (v/v) penicillin-streptomycin (Bioexplorer, United States) until they approached 80–90% confluence. The cells were trypsinized with 1% (w/v) of trypsin-EDTA (Bioexplorer, United States) solution to make new passages. For the adhesion assay, cells were seeded into 24-well cell culture plates and incubated at 37°C, 5% CO₂ for 15 days in a humidified atmosphere to obtain cell differentiation. One milliliter of LAB suspension (1 × 10⁸ CFU/mL) in DMEM was added to each well of the 24-well cell culture plate and co-incubated with Caco-2 monolayers for 1 h at 37°C with 5% CO₂. The wells were washed twice with a pre-warmed Dulbecco’s phosphate buffered saline (DPBS) (pH 7.2, without Ca and Mg) (Procell, China) to remove non-attached cells of LABs and adherent bacteria were detached with DPBS containing 1% (v/v) Triton X-100. The serially diluted lysates were then plated onto MRS agar to determine the number of adherent bacteria and the experiment was carried out in triplicate. Results were expressed as the percentage of bacteria adhered with respect to the number of bacteria added.

Well diffusion was performed to evaluate the inhibitory effects of LAB strains on the growth of indicator strains. In this study, Shigella flexneri CMCC51574, Salmonella paratyphi B CMCC50094, and Escherichia coli CMCC44102 were used as indicator bacteria. The indicator bacteria suspension (50 μL, approximately 10⁸ CFU/mL) was added to 200 mL MRS agar, mixed thoroughly, and poured plate. Eight-millimeter wells were punched on plates using sterile borer. Then, each well was filled by 80 μL of filtered supernatant and plates incubated overnight at 37°C. The inhibition zone was measured in millimeters around the well.

Antibiotic sensitivity was determined using the agar disk diffusion assay method based on a previous study (Nami et al., 2019b). Eight clinically important antibiotics were used to determine the antibiotic susceptibility of the isolated strains. The strains were incubated in MRS broth at 37°C for 18 h. Then, 50 μL of the suspension (approximately 10⁸ CFU/mL) was mixed with 200 mL MRS agar thoroughly and was poured on plate. Antibiotic disks were manually placed on the plates by using sterilized forceps. The diameter of inhibition (mm) around each disk was measured after overnight incubation at 37°C. The strains were grouped into sensitive, moderately sensitive, and resistant according to the Clinical Laboratory Standards Institute (CLSI) recommendations (CLSI, 2020).

The LAB strains were cultivated in MRS broth at 37°C for 16 h∼18 h. The fermentation broth was centrifuged at 8000 rpm for 5 min under 4°C and the supernatants were collected and filtered through a Millex^® 0.22-μm filter (Millipore) to obtain cell-free supernatant (CFS). The harvested cell pellets were washed twice with 0.1 mM phosphate buffer solution (PBS, pH 7.4), re-suspended in PBS, and adjusted to 10⁹ CFU/mL. The re-suspended solutions were divided into two aliquots. One aliquot was used as intact cells (IC), and the other aliquot was ultra-sonicated at 800 W for 30 min (5 s sonication, 7 s interval) in an ice bath. The resulting supernatant was harvested and was filter sterilized to obtain intra-cellular cell-free extracts (CFE).

The stable radical DPPH (2,2-Diphenyl-1-picrylhydrazyl) and the stable cation radical ABTS 2, 2’-Azino-bis (3-Ethylbenzothiazoline-6-sulfonic acid) were used to measure the free radical scavenging activity of LAB strains as reported previously (Cao et al., 2019; Ragul et al., 2020). Intact cells or CFE or CFS (1 mL) were mixed with 1 mL DPPH solution in methanol (200 μM) vigorously and kept at room temperature in the dark for 30 min. Trolox (150 μg/ml) was used as a positive control. The absorbance was measured at 517 nm and DPPH scavenging ability was calculated as:

The solution of 7 mM ABTS⁺ was prepared in 2.45 mM potassium persulfate followed by incubation at room temperature for 12 h. The ABTS⁺ solution was diluted in deionized water to achieve OD₇₃₄ value of approximately 0.7. Twenty microliters of intact cells, CFS, or CFE were vortexed with 2 mL of ABTS⁺ working solution for 30 s and incubated at room temperature for 6 min. Trolox (350 μg/ml) was used as a positive control. The absorbance was read at 734 nm and ABTS⁺ scavenging ability was defined as follows:

The α-glucosidase inhibition assay was performed following Chen’s method (Chen et al., 2014) with slight modification. Twenty-five microliter samples (CFS/IC/CFE) were added to 96-wells microplate containing 50 μL of 20 mM p-nitrophenyl α-D-glucopyranoside (pNPG) and 50 μL of 0.1 M PBS (pH 6.8). The samples were then incubated at 37°C for 10 min and then 30 μL of α-glucosidase solution (20 U/mL) (Solarbio, China) was added and incubated at 37°C for 20 min. The reaction was stopped by adding 50 μL of 1 M Na₂CO₃. p-nitrophenol released from pNPG was measured spectrophotometrically at 405 nm. The percent inhibition of α-glucosidase was calculated by the following equation:

where A is the absorbance of the reactants with sample and B is the absorbance of blank group containing reactants without sample solution.

MRS-CHOL broth was prepared by supplementing MRS broth with cholesterol at a concentration of 0.1 g/L. Specifically, cholesterol (0.1 g), bile salt (2.0 g), and sucrose octaacetate (0.1 g) were mixed with 1.0 mL of tween 80 and followed by adding 5 mL glacial acetic acid. The mixture was dissolved at 60°C and exposed to ultrasound of 130 W (2 s sonication, 3 s interval) for 30 min. The prepared MRS liquid medium was quickly added to the above mixture. The mixture was kept stirring until a homogeneous colloidal solution was formed.

After autoclaving, the MRS-CHOL broth was inoculated with 1% (v/v) bacterial culture (approximately 10⁸ CFU/mL) at 37°C for 24 h. Bacterial broth was centrifuged at 10,000 rpm for 15 min to remove cells. Cholesterol concentration was determined (Rudel and Morris, 1973; Nami et al., 2019b) and cholesterol removal percentage was calculated as follows:

where C₀ is the concentration of cholesterol at the initial medium and C_t is the concentration of cholesterol at the end of inoculation.

All experiments were performed in triplicate. Means and standard deviations were calculated. Statistical analysis of data was carried out using SPSS (Ver. 19.0 SPSS, Chicago, IL, United States). The comparisons of means among different treatments were performed by one-way ANOVA at a significance level of P < 0.05.

A total of 74 strains, Gram-positive, acid-producing, and rod-shaped under the microscope, were isolated from 16 samples of spontaneously fermented vegetables, including fermented cabbage (13 isolates), fermented carrot (14 isolates), fermented chili (4 isolates), fermented Chinese cabbage (32 isolates), fermented cowpea (4 isolates), fermented cucumber (2 isolates), fermented ginger (2 isolates), and fermented leaf mustard (3 isolates), in Shaanxi, China. All 74 strains were subjected to a bile tolerance prescreening (Supplementary Table 1) and 27 of them, which showed below 50% inhibition rate in MRS medium containing 0.3% bile for 4 h, were selected for pH tolerance test. The results showed 26 out of 27 LAB strains survival rate were higher than 100% at both pH 2.5 and pH 2.0 condition (Supplementary Table 2). The 26 isolates were identified based on 16S rRNA sequencing and the results are shown in Table 1. Phylogenetic tree of the 26 LAB strains is shown in Figure 1. All 16S rRNA gene sequences showed 98.05% ∼ 100% similarity with the Lactobacillus spp. Fifteen strains of Lactobacillus plantarum accounted for 57.69% of all strains that passed the initial screening, followed by Lactobacillus brevis with 9 strains (34.62%), and the remaining 2 strains belonged to Weissella viridescens, as shown in Table 1. Since the genus Lactobacillus was generally considered as GRAS (generally recognized as safe) (Min et al., 2019), we focused on the Lactobacillus strains for further study.

After sequential exposure to simulated gastrointestinal juices, survival rate of all test organisms remained above 68% and L. plantarum 115-4 had the highest survival rate of 96.70% (Table 2). Results showed that 20 out of 24 strains had a survival rate higher than 95% after 1 h incubation in simulated gastric juice and after the following 2 h incubation in simulated small intestinal juice there were 10 isolates with a survival rate higher than 90%. The results indicated that the test strains have better acid tolerance compared to the bile tolerance. The survival rate of L. plantarum ranged from 75.01 to 96.7% while L. brevis’s survival rate ranged from 68.17 to 96.6%. There was no significant difference between the two species in terms of tolerance to simulated gastrointestinal. The tolerance differences were strain-dependent rather than species-dependent, which was consistent with the low pH and high bile salts test results.

The isolates were tested for their cell surface hydrophobicity to estimate their adhesion ability by chloroform. Significant difference (p < 0.05) in hydrophobicity was found among different test strains. The five strains with the highest hydrophobicity were L. plantarum 117-5 with a value of 90.09%, and four strains of L. brevis (strains 116-2, 120-4, 114-5, and 114-1) with values of 98.03, 91.45, 90.31, and 83.94%, respectively. At the species level, the hydrophobicity values of L. plantarum ranged from 20.95 to 90.09% while those ranged from 41.87 to 98.03% for L. brevis (Figure 2).

Auto-aggregation of probiotic strains could affiliate adhesion of LAB to intestinal epithelium. The results of cell auto-aggregation assay are shown in Figure 2. The cell auto-aggregation rates of the isolates ranged from 24.73 to 75.98% (5 out of 9 above 60%) for L. brevis and 9.11 to 86.12% (12 out of 15 above 60%) for L. plantarum.

The results of co-aggregation of Lactobacillus isolates in the presence of target pathogens are shown in Figure 3. The highest co-aggregation rates to S. flexneri, S. paratyphi B, and E. coli were obtained for isolate 115-4 (37.19, 45.93, and 28.19%, respectively). No obvious co-aggregation with all three pathogens was exhibited by L. plantarum 132-1 and 032-1. In addition, L. brevis 117-3, 119-2, and 119-8 showed no co-aggregation with S. paratyphi B.

The adhesion capacity to human colon carcinoma cell line, Caco-2, was determined (Figure 2). Adhesion capacity to Caco-2 cells varied significantly among the tested L. plantarum, with adhesion ratio ranging from 2.42 to 34.52%. However, L. brevis showed adhesion capacity ranging from 1.98 to 8.53%. The top five strains with the highest adherence capacity all belonged to L. plantarum, namely isolates 118-4, 122-1, 132-2, 132-1, and 117-5, with mean values of 34.52, 20.27, 18.00, 16.69, and 11.26%, respectively.

The antagonistic activity against foodborne pathogenic bacteria displayed by the 24 selected Lactobacillus strains is shown in Table 3. In this experiment, after 18 h incubation, the L. plantarum strains showed significant antibacterial activity against all the enteric pathogens while L. brevis strains showed varied antagonistic activity. The pH values of L. brevis CFS were in the range of 4.31∼4.98, except for the strain 119-2 (pH 3.89), which showed the highest antagonistic activity among all L. brevis strains. The pH values of L. plantarum CFS were in the range of 3.64∼3.83. After pH neutralization to 6.5, all CFS showed minimal activity against all the pathogens tested, proving the role of organic acids for their antimicrobial activity.

The isolates were tested for their antibiotic susceptibility against different antibiotics (Table 4). Overall, all the isolates showed the ability to resist vancomycin, streptomycin, kanamycin, and ciprofloxacin while sensitive to ampicillin, cefixime, tetracycline, and erythromycin varied among strains, even species. All tested L. plantarum were susceptible to tetracycline while 3 out of 9 L. brevis strains were resistant to tetracycline. No resistance to erythromycin was found for both L. brevis (2M7S) and L. plantarum (4M11S). In addition, L. brevis showed more resistant toward ampicillin (2M7R) and cefixime (1M8R) than L. plantarum, which showed 1R1M13S toward ampicillin and 1R14S toward cefixime.

DPPH and ABTS⁺ free radical scavenging experiments are commonly used to evaluate natural antioxidants (the antioxidant activity of LAB) (Sun et al., 2020).

As shown in Figure 4, the 24 strains tested showed varied degrees of radical scavenging activity. The intact cells of test strains showed the highest DPPH scavenging activity compared with corresponding CFS and CFE. The five highest scavenging activity (67.33, 79.54, 76.93, 86.29, and 82.66%) belonged to strains 032-1, 117-5, 119-1, 122-1, and 132-3, which were all L. plantarum. The lowest five values (37.46, 32.83, 24.51, 30.34, and 32.17%) belonged to L. brevis 114-5, 117-3, and 119-2, and L. plantarum 118-10 and 118-4.

For the ABTS⁺ radical scavenging assay, as shown in Figure 5, the scavenging activity of CFS is significantly higher than IC and CFE. The highest scavenging value of CFS, IC, and CFE were obtained from strains 115-4 (69.26%), 116-2 (12.18%), and 118-4 (5.93%). The ABTS⁺ scavenging rate of CFS ranged from 16.22 to 55.70% for L. brevis. Interestingly, the CFS of L. plantarum 115-4 had the highest (69.26%) ABTS⁺ scavenging rate and that of L. plantarum 121-5 had the lowest rate (1.82%).

The results of α-glucosidase inhibitory activities are shown in Figure 6. The inhibitory activities of CFS ranged from 1.74 to 84.01% for L. brevis while those of L. plantarum presented much higher values of 48.57–86.57%. The inhibitory activities of IC ranged from 2.55 to 9.45% for L. plantarum, which is similar to the corresponding CFE (between 1.13 and 23.34%) and much lower than the corresponding CFS. However, L. brevis exhibited similar levels for all the groups (IC: 1.47 to 8.01%; CFE: 1.87% ∼ 23.92%) except for the CFS of strain 117-3 and 116-1 with the much higher values of 84.01 and 49.06%, respectively. Out of 24 CFS of the isolates, 15, which all belonged to L. plantarum except for L. brevis 117-3, had inhibition rates higher than 50%.

Figure 7 presents the levels of cholesterol assimilation by isolates in the presence of 0.3% bile oxgall at 37°C for 24 h. The percentage of cholesterol assimilated varied (P < 0.05) in different strains and ranged from 6.53 to 50.64% for L. brevis and 13.59 to 43.06% for L. plantarum. The highest percentage of cholesterol assimilation was observed in isolates L. brevis 119-8 (50.64%), 114-1 (42.84%), L. plantarum 132-1 (43.06%), 119-5 (36.20%), and 121-2 (34.01%).