Fermentation patterns of L. plantarum strains were tested with an API 50 CHL test (BioMérieux, France) toward 49 selected carbohydrate sources following the manufacturer’s guidelines. Briefly, overnight cultures of lactobacilli were suspended in 10 mL API 50 CHL medium (BioMérieux, France) and each applied to cupels containing different carbohydrates on an API 50 CH test strip. Fermentation patterns were monitored for up to 72 h at 37°C.

To test heat resistance of L. plantarum strains, overnight broth-cultured strains were heated in a water bath at 70°C for 20, 40, 60, and 80 s. Viable bacteria from each time point were measured by 10-fold serial dilution and spread-plate method.

Antibacterial effects of L. plantarum strains were tested by the disc diffusion method. Four Gram-positive bacteria (SA, LM, SM, SS) and three Gram-negative bacteria (EC, PA, CS) were used as target bacteria. Overnight cultures of target bacteria were suspended in 0.9% saline with an optical density (OD) of 0.8 at 600 nm wavelength (OD₆₀₀). Each bacterial suspension was spread on mixed agar medium of BHI and MRS at a 1:1 ratio in weight. After spreading bacterial suspensions, 8 mm paper discs (Advantec, Japan) were placed on agar plates, and 35 μL of each overnight lactobacillus broth culture was inoculated on the paper discs. After 24–48 h of incubation at 37°C, diameters of inhibition zones were measured. The diameter of the paper disc was subtracted from the total diameter of each measured inhibition zone.

BSH activity of L. plantarum strains was determined using the method described by Dashkevicz and Feighner (1989). Briefly, each strain was streaked on MRS agar medium supplied with 0.5% w/v sodium taurocholic acid (TDCA, Sigma-Aldrich, Germany). Inoculated agar plates were placed in an anaerobic jar and incubated at 37°C for 48–72 h.

H₂O₂ production ability of L. plantarum strains was determined with 0.25 mg/mL tetramethylbenzidine (TMB, Sigma-Aldrich, Germany) and 0.01 mg/mL horseradish peroxidase (Sigma-Aldrich, Germany) in MRS agar medium as described by McGroarty et al. (1992). Additives were filtered through 0.2 μm size pore filter (Satorious, Germany) and added to autoclaved MRS agar medium when the medium reached approximately 50°C. Test specimens were incubated at 37°C for 48–72 h in an anaerobic jar, and exposed to air at 25°C to monitor color change of colonies.

Radical scavenging activity of each L. plantarum strain was tested by using a 2, 2′-azino-bis 3-ethylbenzthiazoline-6-sulfonic acid (ABTS, Sigma-Aldrich, Germany) colorimetric assay. The test was conducted with lysates from each strain to avoid growth and to examine both intracellular and extracellular contents of the bacteria. To prepare bacterial lysates, biomass from overnight broth cultured strains were obtained by centrifugation at 4,000 × g for 30 min at 4°C. Bacterial pellets were washed twice with 1× phosphate buffered saline (PBS) and concentrated to 10× by resuspending in PBS. The samples were treated with lysozyme (Sigma-Aldrich, Germany) at 37°C for 2 h, then lysed with sonication. Solid content of resulting lysates was measured with a moisture analyzer (A&D Co., Ltd., Japan), and the final stock solution concentration was adjusted to 50 mg/mL. ABTS was prepared by mixing 14 mM ABTS and 4.9 mM potassium persulfate at a 1:1 ratio and incubating in a darkroom overnight. The OD of the solution was adjusted to 0.7 at 734 nm wavelength to prepare a working solution. Lysate samples were added into the working solution at a 1:10 ratio on microplates and incubated for 10 min in a darkroom. The ABTS-reacted samples were centrifuged at 13,000 rpm for 5 min at 4°C, and the resulting supernatants were measured at 734 nm wavelength with an Epoch microplate spectrophotometer (BioTek Instruments, United States). Percent ABTS radical scavenging activity was calculated using following equation, where ODsample is the measured OD value of ABTS-reacted lysate and ODcontrol is the OD of non-reacted ABTS control:

To determine the safety of L. plantarum strains in the present study, hemolysis activity was first examined with 5% sheep blood supplied in tryptic soybean agar (TSA, Difco Laboratories, United States) by streaking each strain onto the medium followed by overnight incubation at 37°C under anaerobic conditions. SA was also streaked onto a sheep blood agar medium as positive control for the experiment.

Antibiotic susceptibility of lactobacilli was tested with ampicillin, gentamicin, kanamycin, streptomycin, clindamycin, erythromycin, tetracycline, and chloramphenicol E-test strips (BioMérieux, France) to determine minimum inhibitory concentration (MIC) values. Briefly, overnight cultures of lactobacilli strains were suspended in 0.9% saline at an OD₆₀₀ of 0.8. Bacterial suspensions were spread onto counting agar plates (PCA, Difco Laboratory, United States), and E-test strips were placed on the agar plates. MIC cut-off values were determined using European food safety authority (EFSA) guidelines on antibiotic susceptibility (EFSA, 2012).

Production of the potentially harmful biogenic amines histamine, tyramine, putrescine, and cadaverine was examined using a test medium described by Bover-Cid and Holzapfel (1999). The test medium was made according to the composition of Bover-Cid and Holzapfel and each strain was streaked on the medium. The inocula were incubated at 37°C for 72 h, and colorimetric change of bromocresol purple in the medium from yellow to purple due to pH increase by decarboxylase activity (indicating a positive result) was monitored.

Growth inhibition effects of cell free culture supernatant (CFCS) of L. plantarum strains toward CA and GV were measured. CFCS was prepared from overnight broth cultures of each strain by centrifugation at 5,000 rpm for 30 min. CFCS was filtered through 0.2 μm pore syringe filter for sterilization. CFCS was freeze dried and resuspended in 1× PBS to make a 10× stock solution. No additional process of CFCS was done after stock preparation to preserve all substances in the CFCS. CA was inoculated at 1 × 10⁵ CFU/mL in YM broth supplied with CFCS at 1× final concentration. GV was inoculated at 1 × 10⁶ CFU/mL in BHI broth supplemented with 20% heat inactivated horse serum and CFCS. Samples without CFCS were used as controls. CA was incubated at 37°C for 24 h and GV for 48 h depending on growth properties of each pathogen. Endpoint OD was measured at 600 nm wavelength by a spectrophotometer. Growth inhibition rates were calculated using following equation, where ODcontrol and ODcfcs are the OD of the control and CFCS-treated groups, respectively.

Coaggregation activity to pathogens, another potential antimicrobial ability, was determined for the L. plantarum strains by a method described by Handley et al. (1987). Briefly, overnight broth-cultured microorganisms were collected by centrifugation at 13,000 rpm for 5 min at 4°C, and cell pellets were washed twice with 1× PBS. Each microorganism was resuspended in 1× PBS and adjusted to an OD value of 1.0 at 600 nm wavelength. L. plantarum strain and pathogen suspensions were mixed in equal volumes, vortexed for 5 min and left at room temperature for sedimentation. The upper part of the mixture was carefully removed and its OD was measured at 1, 4, and 8 h time points for each sample. Single suspensions of each microorganism were used as controls. Measured OD values were calculated with following equation to determine coaggregation rate, where ODpatho, ODlp, and ODmix are the ODs of the pathogen groups, L. plantarum strains, and strain and pathogen mixture, respectively.

Coculture experiments of L. plantarum strains and pathogens were conducted to examine direct killing effects. Approximately 2.5 × 10⁶ CFU/mL of CA and of each strain of L. plantarum were inoculated in a 1:1 ratio of mixed YM and MRS broth and incubated at 37°C for 48 h. For the anti-GV test, approximately 1.0 × 10⁸ CFU/mL of GV and of each L. plantarum strain were inoculated in BHI broth supplemented with 20% horse serum, 1% yeast extract, and 0.1% maltose and incubated at 37°C for 48 h in an anaerobic jar. Each test inoculum was sampled and checked for viable pathogens by spread-plate method. For CA growth medium, 100 μg/mL ampicillin supplemented YM agar medium was used to grow CA selectively. For GV growth medium, 5% rabbit blood BHI agar medium was used and GV was selected based on white, slight translucent colony morphology and hemolysis zones formed around GV colonies.

To determine whether L. plantarum strains are capable of immunomodulation, in vitro cell experiments with RAW264.7 murine macrophage cells were performed. Cells were cultured in Dulbecco Modified Eagle Medium (DMEM, Gibco, United States) supplemented with 10% fetal bovine serum (Gibco, United States) and 1% penicillin/streptomycin cocktail (Sigma-Aldrich, Germany) throughout the experiment. RAW264.7 cultured up to 80–90% confluency was collected and 1 × 10⁶ cells were seeded in each well of a 24 well cell culture plate (SPL Life Science, Korea). Seeded cells were stabilized at 37°C in 5% CO₂ atmosphere for 24 h. Then, 1 μg/mL of lipopolysaccharide (LPS, Sigma-Aldrich, Germany) was administered was as administered to 100 or 500 μg/mL of lysates from each L. plantarum strain. Non-treated cells were used as a negative control and LPS-treated cell as a positive control. After treating with each substance, cells were incubated at 37°C in 5% CO₂ atmosphere for 24 h. Resulting culture supernatants of each treatment group were collected and processed with IL-6 enzyme linked immunosorbent assay (ELISA) with a Mouse IL-6 Quantikine ELISA Kit (R&D Systems, United States) and TNF-α ELISA with a Mouse TNF-alpha Quantikine ELISA Kit (R&D Systems, United States) for quantification of produced cytokines.

Using the same experimental procedure described previously in the present study with some modification, IL-10 production enhancement in macrophages by the lysate of L. plantarum strains was examined, and 1 μg/mL LPS and 100 μg/mL of L. plantarum lysate co-treated groups were added in this experiment. After treatment with each substance, cells were incubated for another 24 h, and culture supernatant samples were collected. IL-10 concentrations of each culture supernatant of the experimental groups were determined with an IL-10 Quantikine ELISA Kit (R&D Systems, United States). The dose of lysate (100 μg/mL) was selected as a representative dose based on the previous experiment of IL-6 and TNF-α quantification.

A single colony from each bacterial strain was inoculated in 3 mL MRS broth in a shaking incubator at 30°C for 6 h. Genomic DNA was extracted from the culture broths after 6 h using a Wizard Genomic DNA Purification Kit (Promega, Madison, WI, United States) according to the manufacturer’s instructions. Genomic DNA concentration was determined using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, United States) and the DNA quality and integrity were checked by electrophoresis on an 0.8% agarose gel. Whole genomes were sequenced on a Pacific Biosciences (Menlo Park, CA, United States) single molecule real-time (SMRT) sequencing platform with P6-C4 chemistry on a PacBio RS II instrument (Eid et al., 2009). The raw reads were generated with SMRT Cells and de novo assembled by the hierarchical genome-assembly process (HGAP) (Chin et al., 2013) protocol RS HGAP Assembly 2 in SMRT analysis version 2.3.0 (Pacific Biosciences) ¹. De novo assembly parameters were applied as follows: PreAssembler v2: Minimum Seed Read Length: 5000, Celera Assembler v1 Genome Size (Bp): 3,500,000, Target Coverage: 30, Overlapper Error Rate: 0.06, Overlapper Min Length: 40, Overlapper K-mer: 14.

Complete genomes were annotated by NCBI Prokaryotic Genomes Automatic Annotation Pipeline (PGAAP) (Tatusova et al., 2016). PathogenFinder was used to determinate the pathogenic potential (Cosentino et al., 2013). Further, Rapid Annotation using Subsystem Technology (RAST) version 2.0 (Aziz et al., 2008) was applied for genes of interest. The annotated genes were inspected for probiotic potential. The de novo assembled complete genomes were compared using eggnog (evolutionary genealogy of genes: Non-supervised Orthologous Groups)-mapper (Huerta-Cepas et al., 2017) and the associated default settings and Average Nucleotide Identity (ANI) values were obtained using the ANI Calculator tool on EzBiocloud² (Yoon et al., 2017).

Fermentation patterns of carbohydrate sources by each L. plantarum strain are noted in Supplementary Table S1. Notable differences in fermentation capability were observed: K6 weakly fermented L-arabinose resulting in a green to blue colorization transition of the API indicator medium instead of yellow, while K2 and K8 were completely negative; K6 fermented raffinose and melizetose while K8 only fermented melizetose, while K2 could not ferment either of the carbohydrates; and turanose was fermented by K6 and K8. Based on the pattern identification through the APIWEB database of BioMérieux, identities (%) of each strain were as follows: K2, 52.0% to L. plantarum group 1; K6, 99.4% to L. plantarum group 1; K8, 99.5% to L. plantarum group 1.

The most resistant strain to 70°C heat treatment for 80 s was K8, followed by K6 and then K2 (Figure 1). As shown in Figure 1, K2 and K6 were unstable and showed low reproducibility in replicate experiments as the error bars indicate. Except for the 80 s time point, all strains showed stable heat resistances up to 60 s. Notably, maximum viable CFU counts of L. plantarum strains when cultured overnight in MRS broth were as follows: K2, 9.15 × 10⁸ CFU/mL; K6, 2.83 × 10⁹ CFU/mL; and K8, 3.57 × 10⁹ CFU/mL.

The three L. plantarum strains showed antibacterial effects against all representative opportunistic pathogens in the disc diffusion assay (Table 1). K2 had the largest diameters of inhibition zones except for tests against EC and PA. Tests were conducted with live bacteria suspensions or CFCS, but only live bacteria suspensions showed inhibition activities.

All three strains of L. plantarum showed positive BSH activity which was identified by the formation of convex, shell-like, hardened colonies (Figure 2A). H₂O₂ production was detected in all three strains of L. plantarum based on the blue pigment of O⁻ from TMB reactions in formed colonies and surroundings (Figure 2B). All lysates from the L. plantarum strains showed radical scavenging activity and K6 had the highest radical scavenging rate, which reached a plateau at 20 mg/mL concentration of the bacterial lysate, followed by K8 and then K2 (Figure 2C).

CFCS of each L. plantarum strain showed growth inhibition toward CA and GV (Figure 4A,B). Against CA at the 24 h mark, K2, K6, and K8 showed 87.05 ± 0.38%, 82.11 ± 0.25%, and 81.34 ± 0.42%, respectively (Figure 4A). After 48 h of incubation, GV growth had inhibition rates by CFCS of K2, K6, and K8 of 37.49 ± 3.17%, 28.25 ± 5.33%, and 44.60 ± 1.51%, respectively (Figure 4B). The coaggregation rate of K2, K6, and K8 were 70.73 ± 3.41%, 58.57 ± 1.32%, and 57.88 ± 2.86%, respectively, when reacted with CA after 8 h (Figure 4C). In comparison to coaggregation with CA, coaggregation rates between GV and each L. plantarum strain were lower than that in the corresponding CA experiment. The coaggregation rates of K2, K6, and K8 were 32.72 ± 0.19%, 30.39 ± 0.13%, 20.45 ± 0.42%, respectively, when reacted with GV after 8 h (Figure 4D).

Complete disappearance of CA after 48 h of coculture was observed for all strains (Figure 5A). At the 24 h mark, the CA reduction rate by K2, K6, and K8 were 99.03 ± 0.06%, 99.49 ± 0.04%, and 99.53 ± 0.02%, respectively (Figure 5B). K6 and K8 showed statistically significant CA reduction rates compared to K2. GV was reduced by K2, K6, and K8, at rates of 87.37 ± 1.03%, 92.50 ± 0.54%, and 93.71 ± 1.10%, respectively, and no significant differences between each test group were found (Figure 5D). However, GV growth was detected at the 48 h mark in a roughly doubled CFU/mL after being reduced to 90% on average by each L. plantarum strain (Figure 5C).

RAW264.7 cells produced significantly increased IL-6 and TNF-α upon treatment with each L. plantarum strain lysate compared to the control and was significantly lower than that of LPS-treated group (Figure 6). Each cytokine concentration was increased to both 100 and 500 μg/mL lysate concentrations. However, there was no significant difference between induced IL-6 concentrations by the same dose of lysates (Figure 6A). For 100 μg/mL lysate treatment, the K8 lysate-treated group (523.47 ± 23.78 pg/mL of TNF-α, p< 0.05) showed significantly higher concentrations of TNF-α compared to that of the K2 lysate (443.25 ± 10.50 pg/mL of TNF-α) and K6 lysate (477.81 ± 7.57 pg/mL of TNF-α)-treated groups (Figure 6B). In the 500 μg/mL lysate dose groups, there were no significant differences between any L. plantarum lysate treated groups for both cytokines.

All three lysates of each L. plantarum strain were induced with a significantly higher concentration of IL-10 when treated with LPS compared to the control and LPS-only treated group (Figure 7). Significantly increased IL-10 production was observed in K2 and K6 lysates treated with 1 μg/mL LPS compared to K8 lysate treated with LPS. There was no statistically significant difference between K2 and K6 lysates treated with LPS.

The genomes of L. plantarum strains were sequenced with the PacBio SMRT platform with resulting 330× coverage for K2, 384× for K6, and 432× for K8. The genome of K2 was assembled into one circular chromosome and three plasmids, while the genomes of K6 and K8 were assembled into one circular chromosome and one plasmid (Figure 8). The total genome size of K8 (3,275,764 bp) is slightly larger than those of K2 (3,175,098 bp) and K6 (3,262,505 bp). The protein coding sequences were predicted using NCBI PGAAP and deposited in GenBank with accession numbers GCA_003597635.1 (K2), GCA_003597595.1 (K6), and GCA_003597615.1 (K8). A total of 3,114 open reading frames (ORFs) were predicted from the genome of the strain K2; 2,857 are protein coding genes and 85 are RNA genes. A total of 3,178 ORFs were predicted from the genome of the strain K6; 2,999 are protein coding genes and 87 are RNA genes. Finally, a total of 3,186 ORFs were predicted from the genome of the strain K6; 2,999 are protein coding genes and 87 are RNA genes (Table 3).

Based on ANI analyses, K2 showed 96.30 and 96.12% sequence identity with K6 and K8, respectively. Strain K6 and K8 showed 99.92% sequence identity. All three strains were predicted to be non-human pathogens with PathogenFinder. Coding sequences were identified and grouped in COG classes (Figure 9). The coding sequences were categorized as by their involvement in carbohydrate transport and metabolism (7.9–8.8%); transcription (8.0–8.5%); amino acid transport and metabolism (6.9–7.2%); translation, ribosomal structure and biogenesis (4.9–5.2%); cell wall, membrane, envelope biogenesis (5.6–5.9%); and replication, recombination and repair of nucleic acids (5.0–6.2%). K2 (counted number: 223) showed a lower gene number for carbohydrate transport and metabolism compared with K6 (counted number: 261) and K8 (counted number: 263), while K2 (counted number: 175) showed a higher gene number for replication, recombination and repair compared with K6 (counted number: 149) and K8 (counted number: 154).