Four distinct sauerkraut fermentations were performed and pH, acidity, and LAB profiles over time were observed (Fig 1). Overall, the majority of chemical and microbiological changes occurred during the first days of fermentation, in particular until day 7. All four fermentations revealed comparable patterns in the measured parameters, with results being most similar in fermentations using the same type of cabbage.
Profiles based on the measured parameters were generally in agreement with those previously described for sauerkraut fermentations, though some differences should be noted: fermentations from the present work reached a lower percentage of acidity (0.3 to 0.4%) and a higher pH value (4.1 to 4.9) than traditionally obtained at similar temperatures (1.6 to 2.3% and 3.5 or less, respectively) [11,25].
This discrepancy can be explained by differences in sauerkraut fermentation conditions, particularly the substrates used. Alternative varieties of cabbage may have a differing composition of nutrients, such as soluble sugars. In fact, it has been described that white cabbage has a higher quantity of sugar than some portuguese cabbage cultivars . The level of this nutrient may be particularly important, since low quantities of sugar in the substrate will lead to a lesser amount of acid production during the fermentation process. However, despite differences observed in the acidity and pH values obtained, the evolution of parameters was similar to previous reports .
Isolation, characterization and identification of lactic acid bacteria
Bacteria were isolated from the four distinct sauerkraut fermentations at selected timepoints (0, 2, 5, 7, 16, 23 or 30 days). A collection of 114 isolates was obtained: from the portuguese cabbage fermentations, 29 (fermentation with herbs) and 35 (fermentation without herbs) LAB were recovered; while from the pointed-head cabbage fermentation, 50 (25 from each recipe) were isolated. Following phenotypic characterization, 95 isolates displaying phenotypic characteristics associated with LAB (Gram-positive, oxidase-negative and catalase-negative/weakly positive) were selected for further characterization.
Subsequently, to allow the selection of genomically distinct isolates, PCR-fingerprinting was performed (Fig 2). Isolates with similarity above the reproducibility level (83.3%) were considered genomically similar, leading to the selection of 63 representative LAB for each different time-point/fermentation.
Fig 2. Dendrogram built using M13 and OPC-15 PCR-fingerprinting profiles of the 95 isolates from the four sauerkraut fermentations.
The vertical line represents the reproducibility level, which was used as a cut-off value for the definition of genomically distinct LAB. Isolates highlighted with a dot (●) were chosen as representatives for further studies.
Selected LAB were then identified to the genus level by multiplex PCR, using two genus-specific primer sets, targeting Lactobacillus and Leuconostoc genera. From the 63 isolates, 21 presented a 250 bp amplicon and were identified as Leuconostoc sp. (33%), while 33 presented a 613 bp amplicon and were identified as Lactobacillus sp. (52%); nine isolates remained unidentified (14%) by this method. A reproducibility of 100% was achieved for this technique.
Considering genus allocation and source of isolates, it became evident that distribution of the two genera was not uniform among sauerkraut fermentations (Fig 3). While most isolates in pointed-head cabbage fermentations were identified as Leuconostoc (n = 16/22), in portuguese cabbage fermentations the majority belonged to the Lactobacillus genus (n = 31/41). Fisher’s exact test was used to compare these two groups of fermentation, showing statistically significant differences (P<0.001). Furthermore, isolates from the fermentations performed with aromatic herbs were also compared to those performed without their addition using the same test, and the results were not statistically different (P>0.05). These results indicate that the distribution of Lactobacillus and Leuconostoc genera in sauerkraut fermentations is dependent on the type of cabbage used as substrate, but not on the addition of aromatic herbs to the recipe.
Fig 3. Incidence of Leuconostoc and Lactobacillus isolates among representatives from the four sauerkraut fermentations.
Fisher’s exact test was used to determine statistically significant differences between fermentations. *—P<0.001.
After identification, new dendrograms were created for each genus (Fig 4), and groups of genomically similar bacteria were defined as previously explained. Restricting the analysis to isolates within the same genus clustered four LAB with other isolates of the same fermentation/time-point at high similarity levels, leading to their exclusion from additional studies.
Fig 4. Dendrograms built using the PCR-fingerprinting profiles of Lactobacillus (A) or Leuconostoc (B) isolates.
The vertical line represents the reproducibility level, which was used as a cut-off value for the definition of genomically similar groups. Groups containing more than one isolate are represented. Isolates indicated with an arrow were considered genomically similar to others within the same fermentation/time-point, and were removed from subsequent characterization.
Analysis of groups shared between the fermentations (Table 1) allowed several observations. First, one specific cluster of Leuconostoc, Le9, was present at early time-points in every fermentation. Furthermore, microorganisms from this cluster persisted in every time-point in the pointed-head cabbage fermentation without herbs, dominating the fermentation. On the contrary, the other three fermentations showed a more diverse distribution of microbial clusters. Several groups were present in both portuguese cabbage fermentations, with one group of Lactobacillus, La18, being present at all time-points from day 16 onwards. This may indicate the importance of microorganisms within this group to the fermentation process.
Results observed for the portuguese cabbage fermentations are in accordance with the work of Plengvidhya and coworkers , which found that most microorganisms isolated until the third day of three sauerkraut fermentations belonged to Weissella and Leuconostoc genera, while those isolated at the seventh and fourteenth day were from the Lactobacillus genus. In the present work, the same distribution was observed in the portuguese cabbage fermentations, with Leuconostoc being isolated at the start (T0) and second day of fermentation, and Lactobacillus from the fifth day onwards.
Results for the pointed-head cabbage fermentations showed a different LAB distribution, with a predominance of Leuconostoc spp. at every time-point, which is not usually reported for sauerkraut fermentations. However, a 16S metagenomic study also showed that Leuconostoc remained a significant part of the microbiota throughout sauerkraut fermentation . Moreover, Plengvidhya and coworkers  observed a different pattern of microbial groups in one of the fermentations studied, with both hetero- and homofermentative species being present at every time-point, which may indicate that distinct patterns of microorganisms can occur in sauerkraut fermentations. Variations found between the various types of fermentations are probably due to the differences in chemical, biochemical and/or microbiological characteristics between the varieties of cabbage used as substrate, affecting the microbial succession. In fact, the different substrates and recipes were used to increase the diversity of LAB and potentially find better probiotic candidates.
Safety evaluation and assessment of probiotic potential
Hemolytic ability is a relevant virulence factor that can be present in pathogenic microorganisms. Sauerkraut isolates were screened for hemolytic activity (n = 59) and only one was β-hemolytic, with 18 presenting α-hemolysis and 40 showing a γ-hemolytic phenotype. α-hemolytic non-enterococcal LAB have been considered safe by other authors [22,27], suggesting that the majority of the sauerkraut isolates may harbor low virulence potential and could potentially be safe for use as probiotics.
Another important safety concern is the presence of mobile antimicrobial resistance genes. Sauerkraut LAB isolates were assessed for antimicrobial resistance and results are shown in Fig 5. A low percentage of LAB isolates were classified as resistant to ampicillin (12%), chloramphenicol (15%) and clindamycin (19%). For the other antimicrobial compounds tested (erythromycin, gentamicin, kanamycin, streptomycin and tetracycline), statistical analysis showed that the results were genus-dependent (P<0.05). Lactobacillus rhamnosus GG, a widely studied probiotic strain, showed resistance to kanamycin.
Fig 5. Percentage of isolates resistant to the studied antimicrobial compounds.
AMP- Ampicillin; C- Chloramphenicol; DA- Clindamycin; E- Erythromycin; CN- Gentamicin; K- Kanamycin; S- Streptomycin; TE- Tetracycline. Fisher’s exact test was used to determine statistically significant differences between genera. *—P<0.05; **—P<0.005; ***-P<0.0005.
A high percentage of resistance to gentamicin, kanamycin and streptomycin was detected in Lactobacillus spp. These antimicrobials are aminoglycosides, to which lactobacilli have been described as having a high natural resistance . Likewise, the Leuconostoc genus is usually reported to be resistant to aminoglycosides , but no resistance to gentamicin was found in isolates of this genus in the present work, as observed by other authors . The risk of transmission of aminoglycoside resistance is negligible, so the presence of this characteristic was not applied as criteria for the exclusion of LAB isolates as probiotic candidates.
For the other antimicrobials analyzed, Lactobacillus isolates showed a low level of resistance. Resistance to these antimicrobials is not widespread in this genus, although it has been linked to transmission to other microorganisms through genes found in plasmids or transposons [29,31–33], therefore, resistant isolates may act as reservoirs for dissemination. Taking this into account, 42% of the isolates (n = 25/59), found to be resistant to at least one of these antimicrobial compounds, were removed from further characterization.
Study of antimicrobial resistance involves the use of breakpoint values for the classification of microorganisms as resistant or susceptible. Neither Clinical and Laboratory Standards Institute nor EUCAST have defined breakpoints for the study of antimicrobial resistance in Lactobacillus or Leuconostoc species by disc diffusion , for this reason breakpoints were defined for each antimicrobial based on the resistance level of all bacteria included in this study. Isolates presenting an inhibition halo diameter equal or below the mean minus standard deviation of all isolates were considered resistant, while those above this value were considered sensitive or intermediate (non-resistant).
This strategy could lead to a bias in the incidence of resistance, but the observed results were supported by similar findings from LAB isolated from vegetable fermentations, despite the use of different techniques [22,24]. Additionally, the resistance profile obtained for L. rhamnosus GG was comparable to the observed by Argyri and coworkers , indicating that resistance profiles observed in the present study may be comparable to those obtained using different methodologies.
After evaluating the safety of probiotic candidates, isolates were tested for resistance to low pH and bile, important characteristics to survive transit though the human GI tract . For this purpose, an agar-based screening protocol was performed, with results showing that few isolates were resistant to low pH conditions (20%, n = 12/59), all belonging to the Lactobacillus genus. Furthermore, a high number of isolates were resistant to bile (88%, n = 52/59). L. rhamnosus GG was used as a probiotic control and was resistant to both 0.5% bile and a pH value of 3.5.
Based on hemolytic activity, antimicrobial resistance and resistance to low pH and bile (Fig 6), six Lactobacillus sp. (L54, L59, L61, L71, L80 and L89) were selected and tested for antimicrobial activity against Listeria monocytogenes and resistance to a lower pH than previously applied. All six isolates were shown to harbor antimicrobial activity against L. monocytogenes in a spot-on-lawn assay, yet when the inhibitory activity of culture supernatants was tested in an agar well diffusion assay, inhibition was non-existent or very weak. The six isolates were also tested for resistance to lower pH values (pH = 2.5) using a broth-based assay. While viability was observed in four of the isolates after 3 h of incubation, only three (L54, L61 and L89) were still viable after 24 h.
Fig 6. Summary of results for the 95 isolates with a LAB-like phenotype.
Dendrogram built based on PCR-fingerprinting profiles, with the red vertical line representing the reproducibility level. Information regarding hemolytic activity, antimicrobial resistance, low pH (3.5) and bile resistance, genus identification, group attributed after PCR-fingerprinting analysis and source of the isolate is also shown. Isolates written in red were chosen as representatives after PCR-fingerprinting, and isolates written in blue were also chosen, but excluded after further analysis. Isolates marked with a box were selected for subsequent analysis based on the information presented in the figure. Black squares represent the presence of AMP—Ampicillin; C—Chloramphenicol; DA—Clindamycin; E—Erythromycin; CN—Gentamicin; K—Kanamycin; S—Streptomycin; or TE—Tetracycline resistance.
Overall, results showed that probiotic candidates inhibited the growth of L. monocytogenes, a Gram-positive pathogen representing an important cause of foodborne outbreaks and related mortality , but the nature of this effect was not completely established. The antibacterial effect cannot be attributed to soluble compounds present after growth of the isolates in liquid media, such as organic acids or bacteriocins, since these would have inhibited growth in the well diffusion assay. However, antimicrobial activity was observed on the spot-on-lawn test, performed in solid media. A possible explanation is that the production of inhibitory compounds was induced by the presence of the pathogen, since inhibition was only shown when the isolates and the pathogen where in direct contact. In fact, in some cases, co-culture of lactic acid bacteria with target cells can be a requirement for bacteriocin production . Therefore, there is evidence of the antibacterial effect of all the tested isolates against L. monocytogenes, which further indicates their probiotic potential, although the precise mechanism by which this effect occurred is still not understood.
Results from previous reports regarding acid resistance vary greatly, and this is probably due to the use of different methodologies, which hinder the comparison between results from other studies and the present work. Although there is no established protocol for assessing resistance to low pH, the agar-based methods used in the present study allowed the selection of a small number of isolates with probiotic potential for further testing. The broth-based method allowed to further assess this characteristic, and the fact that three of the six selected isolates were resistant to pH values as low as 2.5 is a good indicator of their suitability as probiotic candidates. Nonetheless, conditions closer to those found in the GI tract, such as the presence of digestive enzymes, should be tested for further confirmation.
Through this work, we were able to identify three lactobacilli showing characteristics associated with probiotics. Isolates belonging to Leuconostoc and other genera were also recovered but were excluded as candidate probiotics. In agreement with these findings, Beganović and coleagues  isolated strains belonging to these two genera from brines sampled during the course of sauerkraut fermentation of white cabbage, but the two most promising probiotic candidates belonged to the Lactobacillus genus. Likewise, Yu and coleagues  recovered Lactobacillus plantarum from chinese sauerkraut, and identified two potentially probiotic strains. Other studies focusing on kimchi, a fermented vegetable product similar to sauerkraut, showed comparable results [16,36,37].