The MRS medium adjusted to different pH values (5.50, 4.90, 4.30, 3.60, and 3.30) and the MRS medium supplemented with bile salt (0.10–0.60%, w/v) were prepared for the tests of acid and bile salt tolerance of strains, respectively. Cells from the overnight culture of L. plantarum strains were collected and resuspended in acid MRS medium and bile salt medium to reach an OD_{600 nm} of 0.10 ± 0.05. Then, a 200 μl volume of each culture was transferred to 96-well plates covered with lids. The plates were incubated in a Logphase 600 (BioTek, Agilent, United States) at 37°C, 500 rpm for 24–96 h. Control experiments containing only MRS medium and cell inoculum were also performed. The OD_{600 nm} of cultures was read at 10 min intervals. The lag phase for each strain to adapt to the inhibition conditions was used to indicate the resistance of strains to acid and bile salt. All assays were conducted in triplicate.

The adhesion ability of strains was evaluated using cells’ hydrophobicity and auto-aggregation index. Protocols developed by Lee and Kim (2019) were adopted to measure these two indexes. All experiments were conducted in triplicate. Briefly, 1 ml of xylene was added to 3 ml of ICS and mixed well. Then the mixtures were kept at 37°C until a phase separation was observed. The bottom phase was used to measure the hydrophobicity of strains. For the assay of the hydrophobicity of strains, the absorbance of bottom phases was read at 600 nm, which was defined as the value of A_t. The hydrophobicity of cells was expressed using the following equation:

where A₀ is the initial OD_{600 nm} of samples measured at t = 0 h.

Briefly, 4 ml of ICS was incubated at 20°C for 24 h. Next, 1 ml of the upper layer solution was read at OD_{600 nm} to determine the auto-aggregation of strains. The auto-aggregation index was defined as follows:

where A₂₄ is the OD_{600 nm} measured of the upper layer at t = 24 h and A₀ is the initial OD_{600 nm} measured at t = 0 h.

The simulated gastric juice (SGJ) was composed of 27 mg/ml of pepsin (Solarbio, Beijing, China) dissolved in 0.50% (w/v) saline solution (pH 2.50), whereas the simulated small intestinal juice (SSIJ) was composed of 0.27 mg/ml of pancreatin (Solarbio, Beijing, China) in 0.50% of saline at pH 8.00. Cells from the pre-cultivation of L. plantarum strains were harvested and resuspended in SGJ with an initial cell density of 10⁸ CFU/ml and incubated at 37°C for 3 h. Then, 0.50 ml of incubated SGJs was added to 4.50 ml SSIJ and cultivated at 37°C for another 3 h. The survival rates of SGJ and SSIJ were determined by counting viable cells on MRS agar plates after inoculation with cultures from simulating gastrointestinal transit tests (Fei et al., 2018). The analyses for each strain were performed in triplicate. The survival rate was expressed using the following equation:

The antioxidant activity of strains was measured with the 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging activity assay and 2,2′-azino-bis-(3-ethylbenzthiazoline-6-sulfonic acid) (ABTS) radical cation scavenging activity assay as described by Son et al. (2017) and Tang et al. (2021), respectively. The experiment for each strain was performed in triplicate. In brief, 0.50 ml of ICS (sample) or PBS (control) was mixed with 0.50 ml of 0.20 mmol/L of DPPH solution. An equal volume of methanol was used to replace the DPPH solution as the blank. Then the mixtures were incubated at dark for 30 min at 37°C. After centrifugation (12,000 rpm, 10 min), the absorbance of mixtures was measured at 517 nm using a microplate reader (M5, SpectraMax, United States). The antioxidant activity was calculated as follows:

For the ABTS radical cation scavenging assay, a working solution composed of 7.40 mmol/L of ABTS⁺ and 40 mmol/L of potassium persulfate was prepared. After reacting for 12 h in the dark at room temperature, the working solution was diluted with absolute ethanol to an absorbance of 0.70 ± 0.02 at 734 nm and stored at 30°C. Next, 0.25 ml of ICS (sample) or PBS (control) was mixed with 4 ml of working solution. The blank samples were prepared with 4 ml of absolute ethanol to replace the working solution. All the samples were kept at 30°C for 6 min of incubation. Absorbance against the blank samples was read at 734 nm. Results were calculated as follows:

The antibacterial activity of selected strains was determined using the agar diffusion test based on the method previously reported by Wang et al. (2018). Briefly, 100 μl of CFS was added into the holes with a 7-mm diameter in Luria Bertani (LB) agar plates that had been spread with indicator bacteria (Escherichia coli, Staphylococcus aureus, Shigella flexneri, and Salmonella typhimurium). The plates were cultured at 30°C or 37°C for 48 h. The diameter of the inhibitory zone formed on the plates was considered an indication of antibacterial activity.

As for BSH activity, 100 μl of ICS from cultures of different strains were added into the holes (7 mm diameter) in the MRS agar plates supplemented with 0.37 g/L of CaCl₂ and 5 g/L of sodium taurocholate (Park et al., 2016). Precipitation observed in the holes can be determined as capable strains with BSH activity.

One loop of pre-culture of each strain was streaked on Columbia blood agar plates containing 5% of sheep blood (Sharma et al., 2018). If transparent cycles formed around strain colonies, it meant strains hold the hemolytic activity.

To further identify the probiotic properties of the best-performed strains (B652, C232, D444, and E932), the whole genome sequencing analysis was performed. The overnight culture of L. plantarum strains was collected by centrifugation at 3,800 rpm for 5 min and submitted to Shanghai Majorbio Bio-Pharm Tech Co., Ltd., (China) for DNA extraction and sequencing analysis using the Illumina HiSeq2000 sequencing platform at 2 × 150 paired-end reads.

For genome assembly, after removing low-quality data obtained from the sequencing platform, the clean reads of each strain were assembled using SOAPdenovo 2.0 and SPAdes 3.11. Glimmer version 3.02 was used for the prediction of coding sequence and GC content. Functional annotation of the genome was performed by blasting genes with multiple public databases including Kyoto Encyclopedia of Genes and Genomes (KEGG), Clusters of Orthologous Groups of proteins (COG), Non-redundant protein database (NR), and Gene Ontology (GO) under an E-value cutoff of 1e-5. A certificated probiotic L. plantarum strain ST-III (accession number CP002222 in NCBI) was used as a reference to identify the probiotic potentials of candidate strains on genomic levels (Huang et al., 2021). Antismash software was used to predict the gene clusters of secondary metabolite synthesis. The core and pan genome of selected strains were analyzed using PGAP 1.2.1 and CG-HIT 4.6.1, respectively. Based on genome mapping and sequencing, the MUMmer 3.23 with default parameters was used to compare known genes and genome structures of each strain at the nucleotide level (Li et al., 2016a).

The acid resistance assay showed that the lag phase of 44 L. plantarum strains increased significantly (P < 0.05) as pH decreased (Figure 1A). Although most tested strains proliferated within 48 h at pH 3.60, C232 started an exponential phase at pH 3.30. Similarly, the lag phase duration of the strains was prolonged as the increase in bile salt concentration in the cultivation medium (Figure 1A). As shown, 75% of the 44 strains could tolerate 0.30% bile salt, the average bile concentration in the human gastrointestinal tract, but only 23% of tested strains could enter the exponential phase when the concentration of bile salt reached up to 0.60% in the medium. No significant differences were observed in terms of the lag phase of D531, E932, and L117 at this condition (P > 0.05). This result indicated that L. plantarum had a considerable tolerance toward bile salt. Lin et al. (2020) also found that three L. plantarum strains isolated from kimchi have a stronger resistance against bile salt than those of gastric acid.

As demonstrated in Figure 1B, the cell surface hydrophobicity of all tested strains ranged from 12.97 to 83.63%. Strains, whose cell surface hydrophobicity index was higher than 70%, were classified as hydrophobic ones with good adhesion activity (Shangpliang et al., 2017). Among the tested strains, the hydrophobicity of B652, D2510, and K125 was all above 70%. Specifically, the number for B652 reached up to 83.63%, making it an ideal candidate for probiotics. In addition, auto-aggregation activity is a crucial index, reflecting the ability of strains to colonize in gastrointestinal environments. In the study, the auto-aggregation capability of tested strains showed significant differences (P < 0.05), ranging from 24.03 to 91.88%. The highest auto-aggregation value (91.88%) was found in L11, followed by A232 with 86.54%. A recent study showed that L. plantarum MYSRD 71, isolated from fermented Vellappam, exhibited a high auto-aggregation value (88.5%) and good adhesion performance (Divyashree et al., 2021).

The survival rate of L. plantarum in gastrointestinal transit has been evaluated (Figure 1C). Among the tested 44 strains, 15 strains survived in artificial gastrointestinal transit. All strains showed an obvious reduction in viable cells after 3 h exposure to artificial gastric juice (pH 2.50). The loss of viability ranged from 3 log to 5 log and the survival rate was about 50%. However, it was found that 3 h incubation in simulated intestinal juice did not result in a significant decrease (73.71–99.46%) in the survival rate of test strains (P > 0.05). The result was aligned with our observation that L. plantarum had a stronger tolerance to bile salt than that of acid. After a total of 6 h of digestion in the simulated gastrointestinal tract, D444 and C232 still keep high viable counts (around 4 log CFU/ml). In general, C232 had better performance in tolerant acid and bile salt than other strains.

The DPPH and ABTS scavenging ability of 44 L. plantarum is summarized in Figure 2. All strains showed antioxidant activity. The DPPH radical scavenging ability of the strains ranged from 0.88 to 28.50%, while a much higher number for ABTS radical cation scavenging was found from 16.02 to 54.24%. The highest DPPH scavenging (28.02%) was found in strain D444, followed by 20.33% found in B652. As for ABTS radical scavenging activity, B652 held the highest value at 54.24% among all tested strains.

The antibacterial activity of 44 L. plantarum strains against 4 pathogens (Escherichia coli, Staphylococcus aureus, Shigella flexneri, and Salmonella typhimurium) was evaluated (Table 1). A total of 14 L. plantarum strains were capable of inhibiting the growth of more than 3 kinds of pathogens. L117 could inhibit the growth of the 4 pathogens, especially Shigella flexneri and Salmonella typhimurium. A182 and A463 displayed the highest inhibitory to Salmonella typhimurium, with inhibitory zone diameters of 6.01–10.00 mm. As the CFS was directly used in the study without any pretreatment, the antibacterial compounds present in CFS should be further identified. Wang et al. (2018) treated the CFS with catalase and 1 mol/L NaOH to eliminate the influence of organic acids and hydrogen peroxide, which resulted in an inhibition zone with a diameter of 29.60 mm. The antibacterial capacity of L. plantarum is possibly related to the antibacterial substances generated by strains, such as organic acids, hydrogen peroxide, and antibacterial peptides (Sikorska and Smoragiewicz, 2013).

The generation of BSH in strains is suggested as one of the factors to screen probiotics for lowering serum cholesterol (Pereira et al., 2003). A BSH screening medium was used to characterize the BSH activity of strains, in which insoluble unbound bile salt derived from the taurine bile salt hydrolyzing under the catalysis of BSH can react with CaCl₂ in the culture medium to form precipitation circles (Meng et al., 2021). The precipitation observed in the BSH screening medium was used to indicate the generation of BSH in strains. The precipitate circles were observed on 44 L. plantarum strains (data not shown), implying that these 44 L. plantarum strains might be promising probiotic candidates to decrease cholesterol levels.

Hemolytic activity is a decisive index to evaluate the safety of probiotic candidates. If strains show a positive hemolytic activity, it is forbidden to be used as probiotics. In the study, the Columbia blood agar plates were used to assay the hemolytic activity of the strains. The red blood cells in the Columbia blood agar plates could be dissolved by the hemolysin produced by strains and form a transparent hydrolysis circle around colonies of strains. If there are no hydrolysis circles formed around colonies, the strain is defined as γ-hemolytic and considered a safe one (Sharma et al., 2018). All the tested strains displayed negative results for this experiment (data not shown).

The basic genomic features of B652, C232, D444, and E932 compared with ST-III are illustrated in Figure 3. It was found that the genomic size of 5 strains ranged from 3.25 to 3.78 Mb and the number of coding genes was about 3,013–3,488 with the GC content of 44.58–47.31%. Notably, the genome size and GC content of B652 were obviously higher than those of others (Figure 3A). Furthermore, from the distribution of common and unique genes of the 5 strains (Figure 3B), it could be seen that there were 2,323 genes shared by the 5 strains and B652 had the most unique genes (814). The result was aligned with the bigger genome size found in B652. The unique genes occupied by B652 might be related to its special properties, such as good antioxidant activity. The higher GC content may indicate that microorganisms consume more energy during reproduction, while microorganisms with a low GC content may be easier to maintain genomic stability due to low energy metabolism (Rocha and Danchin, 2002).

According to the annotation result of KEGG, the genomic function enrichment of 5 strains is illustrated in Figure 4A. It was noted that 15 pathways of the 5 strains were significantly enriched in nucleotide and amino acid metabolism. Among them, genes responsible for replication and repair accounted for the highest proportion. Figure 4B illustrates that, except for unknown functions, most of the annotated genes were enriched in metabolism, especially carbohydrate transport and metabolism, amino acid transport and metabolism, and energy production and conversion, which was consistent with the previous study reported by Huang et al. (2021). The result indicated that tested strains could metabolize different types of carbon sources and had good propagation ability. In addition, genes of all 5 strains were equally distributed in several COG categories, such as transcription, cell wall/membrane/envelope biogenesis, replication, recombination and repair, translation, ribosomal structure, and biogenesis. The results implied that the strains had a high similarity on genomic levels.