In Vitro Screen of Lactobacilli Strains for Gastrointestinal and Vaginal Benefits

Traditional probiotics comprise mainly lactic acid bacteria that are safe for human use, tolerate acid and bile, and adhere to the epithelial lining and mucosal surfaces. In this study, one hundred commercial and non-commercial strains that were isolated from human feces or vaginal samples were tested with regards to overall growth in culture media, tolerance to acid and bile, hydrogen peroxide (H2O2) production, and adhesion to vaginal epithelial cells (VECs) and to blood group antigens. As a result, various of the tested lactobacilli strains were determined to be suitable for gastrointestinal or vaginal applications. Commercial strains grew better than the newly isolated strains, but tolerance to acid was a common property among all tested strains. Tolerance to bile varied considerably between the strains. Resistance to bile and acid correlated well, as did VEC adhesion and H2O2 production, but H2O2 production was not associated with resistance to bile or acid. Except for L. iners strains, vaginal isolates had better overall VEC adhesion and higher H2O2 production. Species- and strain-specific differences were evident for all parameters. Rank-ordered clustering with nine clusters was used to identify strains that were suitable for gastrointestinal or vaginal health, demonstrating that the categorization of strains for targeted health indications is possible based on the parameters that were measured in this study.


Introduction
Probiotics are defined as "live microorganisms that, when administered in adequate amounts, confer a health benefit on the host" [1]. Strains of lactic acid bacteria and bifidobacteria are the most frequently used probiotics and have been recognized as safe for human consumption [2]. In addition to safety aspects, a good probiotic, if taken orally, must survive passage through the gastrointestinal tract. Thus, an effective oral probiotic should tolerate and endure the harsh acidic conditions of the stomach and bile in the small intestine [3]. Generally, probiotics confer their benefits by improving epithelial barrier function, increasing adherence to the mucosa, and competing with pathogens for sites of adherence, resulting in competitive pathogenic exclusion, the production of antimicrobial substances and bacteriocins, and modulation of the immune system [4]. Traditionally, probiotics have been used for their gastrointestinal effects and immunomodulatory functions, but increasing evidence has shown their benefits in improving the general health and well-being of a host beyond the gastrointestinal tract [5].
The vaginal microbiota has the lowest taxonomic diversity of any site in the body [6]. In most women of reproductive age, the vagina harbors the highest relative abundance of lactobacilli, comprising more than 20 species, with a predominance of only one or two species in a given sample, the most common being Lactobacillus (L.) crispatus, L. iners, Table 1. List of lactobacilli strains that were screened in the study, as well as their commercial category and source. For strains marked with an asterisk (*), the taxonomic classification to a species was not absolutely certain. Strains marked with a hashtag (#) did not grow under laboratory conditions and were excluded from future studies. The category in the table indicates whether the strain is commercial (Com.) or non-commercial (Non-Com.). Source in the table refers to origin of the strain as fecal (F), unknown (U), dairy (D), plant (P), or vaginal (V). The color coding and number for the various parameters measured in this study indicate the rank order of the strains in the measured parameters, with 1 (red) indicating the best-performing strain and 98 (blue) indicating the worst-performing strain.    Fifty isolates were sourced from the fecal samples of healthy Finnish adult volunteers (ethical permission 229/E0/07, Helsinki University Hospital EC). A total of 35 strains were obtained from the Danisco Global Culture Collection (DGCC, Niebüll, Germany), containing commercialized (15 strains) and non-commercial strains (20 strains) with dairy, plant, fecal, or unknown origin, all of which were arbitrarily selected based on their lactobacilli status.
A total of 15 strains were isolated from the vaginal tract of asymptomatic healthy Swedish women according to Amsel criteria with a pH of 4.1 (kindly donated by Dr. Inger Mattsby-Baltzer, University of Gothenburg, Gothenburg, Sweden) ( Table 1): 5 L. crispatus, 5 L. iners, and 5 L. jensenii. Of the 5 L. iners strains, 2 failed to grow under laboratory conditions and were therefore omitted from the data analysis. In total, 98 strains were tested for general growth, tolerance to acid and bile, VEC adhesion, H 2 O 2 production, and adherence to blood group antigens.
The comparator strains were as follows: L. rhamnosus GG (VTT E-96666, VTT Culture Collection, Espoo, Finland) for acid and bile tolerance; L. jensenii DSM20557 (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ), Braunschweig, Germany) for H 2 O 2 production; and L. crispatus LMG18199 (JCM8778) (Laboratorium voor Microbiologie, Universiteit Gent, Ghent, Belgium), which is known to adhere well to the A antigen [15], for adhesion to the A, B, and H antigens. In a screen, L. jensenii LX11796 adhered well to VECs and was therefore selected as a comparator strain for the VEC adhesion assay.

Acid and Bile Tolerance
The tolerance of all strains to acid was tested at pH 2.5 and pH 3.5 for 1.5 h, and tolerance to bile was measured in the presence of 0.9% and 0.3% oxgall (Difco, Fisher Scientific, Vantaa, Finland) for 24 h [16]. In these assays, L. rhamnosus GG was included for comparison based on its ability to survive and proliferate at gastric pH and in bilecontaining medium [17].
Briefly, the strains were cultured in duplicate in MRS broth under anaerobic conditions at 37 • C for 18 h. Next, the bacteria were collected by centrifugation at 2800× g for 10 min. The pelleted cells were washed twice with 10 mL phosphate-buffered saline (PBS 0.01 mol/L; pH 7.2, Thermo Fisher Scientific, Waltham, MA, USA) and resuspended in PBS to an optical density at 600 nm (OD 600 ) of 1 (equaling approximately 1 × 10 8 colony-forming units (CFU)/mL).
Acid tolerance was tested by incubating cells in PBS at pH 2.5, 3.5, and 7.2 for 90 min at 37 • C. Tenfold dilution series were grown on MRS agar, and colonies were counted after a 48 h incubation under anaerobic conditions at 37 • C. The results are expressed as log reduction in growth in CFU at pH 2.5 and 3.5 versus pH 7.2.
Bile tolerance was tested by incubating diluted (1:10) strain cultures in MRS broth that contained 0%, 0.3%, or 0.9% oxgall at 37 • C for 24 h under anaerobic conditions. Growth was measured before and after incubation as OD 595 on a Multiskan RC (Labsystems Oy, Vantaa, Finland). The results are expressed as % growth (OD 600 ) in MRS with 0.9% or 0.3% oxgall versus without bile.

Hydrogen Peroxide Production
The strains were cultured overnight in MRS broth under microaerophilic conditions at 37 • C. OD 600 absorbance was then measured on a Multiskan RC, averaging 2.2 (range: 1.8 to 2.4). For the H 2 O 2 assay, 1 mL of the culture was inoculated in 10 mL MRS broth and incubated under aerobic conditions for 3 h at 37 • C to induce H 2 O 2 production. A 50 µL sample was taken from the aerated culture at the start of the incubation (0 h) and after 1.5 and 3 h. H 2 O 2 in the 50 µL sample was measured using a hydrogen peroxide fluorometric detection kit (AH diagnostics Oy, Helsinki, Finland) according to the manufacturer's instructions. Briefly, 50 µL of reaction cocktail was added to the 50 µL sample and incubated for 10 min in the dark at room temperature, after which the fluorescence was measured on a Wallac Viktor 2 1420 multilabel counter (Perkin Elmer, Turku, Finland). Results are expressed as H 2 O 2 production after 0 h, 1.5 h, and 3 h fermentation versus the reference strain, L. jensenii DSM20557. All measurements were performed in duplicate.

Adhesion to A, B, and H Antigens
The blood group antigens A, B, and H (H antigen corresponding to blood group O phenotype) (Elicityl, Crolles, France) were suspended in PBS at pH 7.2 and biotinylated. The strains were cultured on MRS plates under anaerobic conditions at 37 • C for 48-72 h. A single colony was used to reinoculate 10 mL MRS broth and cultured overnight under anaerobic conditions at 37 • C. The cells were washed twice with 10 mL PBS (pH 7.2) and resuspended in PBS to an OD 600 of 1 (approximately 1 × 10 8 CFU/mL).
Then, 1 mL of bacterial sample and 1 mL 10 µg/mL biotinylated antigen solution were mixed and incubated with slow shaking for 30 min at room temperature, and 100 µL of this mixture was transferred to Delfia streptavidin-coated 96-well plates (Perkin Elmer). The plates were washed twice with 200 µL PBS at pH 7.2 three times with SuperBlock (twice with 200 µL and once with 100 µL) (ThermoFisher Scientific, Pierce™, Waltham, MA, USA) and once with 200 µL sterile water. The plates were then incubated for 30 min at room temperature with slow agitation, after which each well was washed 3 times with 200 µL sterile water for 5 min each. To detect the attached bacteria, 200 µL Syto9 dye (diluted 1:6) (Invitrogen) was added to each well and incubated for 15 min in the dark. The intensity of the dye was measured on a Wallac Viktor 2 1420 multilabel counter; all measurements were performed in quadruplicate, and the results were repeated at least twice.

Adhesion to VECs
Primary VECs were obtained from healthy adult donors (age not specified) (CelProgen, San Pedro, CA, USA), maintained at 37 • C in a 5% CO 2 atmosphere, and expanded in human vaginal epithelial expansion ECM T75 flasks (CelProgen) in human VEC growth media with serum (CelProgen). The cells were used in the adhesion assays at the earliest passage possible.
For the adhesion assays, 150,000 VECs were seeded in human vaginal epithelial cell culture ECM 24-well plates (CelProgen). After overnight incubation at 37 • C in 5% CO 2 , the cells were washed 2 times with Dulbecco's Modified Eagle Medium (DMEM, Thermo Fisher Scientific) without supplements. Radioactively labeled bacteria (see below) were then applied to the cells and incubated for 1 h to attach. The cells were washed 4 times with PBS at pH 7.2 to remove unattached bacteria, after which 100 µL DMEM and 1 mL Optiphase Supermix (Perkin Elmer) were pipetted onto the cells.
The bacterial strains were labeled radioactively by transferring 1% inoculation from overnight cultures into 1.5 mL MRS with 10 µL methyl-1.2 [ 3 H]thymidine (4.4 TBq/mmol) (Perkin Elmer) and anaerobically cultured overnight at 37 • C. The next day, the bacteria were collected by centrifugation at 2800× g for 5 min, and the bacterial pellet was suspended in PBS and washed twice with PBS. The bacterial quantity was determined by flow cytometry as previously described [18]. Then, 10 bacteria per vaginal cell were centrifuged and diluted with DMEM (Thermo Fisher Scientific™). The number of bacteria in the adhesion experiments was determined in an initial optimization screen with 7 strains (data not shown).
The radioactivity was counted on a 1450 Microbeta Trilux liquid scintillation and luminescence counter (Perkin Elmer). The adhesion for each bacterium was calculated as follows: where: • Control 1 contained VECs but no bacteria; • Control 2 contained the same number of bacteria but no VECs, which represents the maximum radioactive count that can be obtained from a sample; • Control 3 comprised empty wells without bacteria or VECs.
Because L. jensenii LX11796 (Table 1) showed good adhesion, with 5.9% (±2.7) of bacteria adhering to VECs, it was selected as a comparator strain. Thus, relative adhesion was calculated as the ratio of the % adhesion of the sample to the mean % adhesion value of the comparator.
Spearman correlation was performed to assess the statistical dependence between measurement pairs. The statistical significance of Spearman correlation coefficients was calculated using GraphPad Prism, version 9.2.0 (GraphPad Software, La Jolla, CA, USA), and p values < 0.05 were considered significant.
The adhesion data for antigens A, B, and C were standardized separately to 0 mean and unit variance for descriptive purposes.
Part of the analysis was performed using rank-transformed data that were derived by rank ordering the strains based on a measurement, then by replacing the measured value by the rank of the strain such that the best-performing strain was assigned a value of 1 and the worst-performing strain was assigned a value of 98. As not all strains produced H 2 O 2 , the maximum value of H 2 O 2 production either at 0 h, 1.5 h, or 3 h was taken for ranking. The benefit of this approach includes the ability to compare and visualize the measurements. The strains were clustered by Euclidean distance in the rank-transformed data using Ward's method [19]. The resulting tree was pruned into 9 clusters for descriptive purposes.
The statistical significance of comparisons between strain groups using ranked data was calculated by the Kruskal-Wallis rank sum test [20] (Supplementary Materials Figures S7-S10).
Statistical analyses were performed in R (version 4.0.3) and with GraphPad Prism (version 9.2.0).

Growth
The ability to grow in MRS broth was tested for the entire panel of strains. At the species level, most species grew well in MRS, although high variation was observed in certain species ( Figure 1A). The only species that did not grow in MRS were two strains of L. iners (Li25-34 and Li22-26), which were therefore excluded from further studies. In addition, according to optical density measurements, the other L. iners strains grew more poorly than the other species, with the exception of L. gasseri (Table S1). At the strain level (Supplementary Materials Figure S1), L. rhamnosus LX11888 showed the highest variation in growth amongst individual strains. The 10 worst-growing strains in MRS were the 3 L. iners strains, L. amylovorus LX11898, L. mucosae LX11854, Latilactobacillus curvatus 360, L. fermentum 508, L. rhamnosus LX11870, L. gasseri LG11876, and L. crispatus LX11211. None of the poor performers was a commercial strain. The 10 best growers in MRS were the commercial strains L. salivarius Ls-33, L. plantarum Lp-115, L. rhamnosus HN001, and Lacticaseibacillus casei Lc-11, in addition to the non-commercial strains L. plantarum LX11878 and LX11861, L. rhamnosus 1704 and LR1049, L. fermentum LX1312, and Lacticaseibacillus paracasei LX11858.

Acid Tolerance
Acid tolerance of the 98 strains in MRS was tested by growing them at pH 2.5 and pH 3.5; the logarithmic reduction in growth was normalized to the growth at pH 7.2 ( Figure 1B,C). Acid tolerance of L. rhamnosus GG was used for comparison. The acid tolerance at pH 2.5 correlated positively with that at pH 3.5 (pairwise Spearman correlation, r = 0.2782 and p = 0.0066 (data not shown)).
The acid tolerance varied widely between species, but many tolerated the milder acidic conditions at pH 3.5 well ( Figure 1B), particularly L. rhamnosus, L. acidophilus, L. fermentum, L. paracasei, L. gasseri, and L. crispatus (with the exception of one strain). L. jensenii had the highest variation among species, with lower tolerance to acid, and did not significantly differ from L. fermentum (Table S2). The other significant differences were between L. acidophilus and L. rhamnosus, as well as between L. fermentum and L. plantarum (Table S2), which indicates that the various species tolerated mild acid quite well.
At pH 2.5 ( Figure 1C), all six L. gasseri strains were among those that tolerated acid the best (log reduction < 1). Indeed, L. gasseri was significantly more tolerant than all other species (Table S3). L. acidophilus species tolerated acid at pH 2.5 quite well and were significantly more tolerant than L. crispatus, L. plantarum, and L. jensenii (Table S3). Similarly, L. paracasei and L. rhamnosus were more tolerant than L. crispatus and L. jensenii (Table S3). Visually, L. amylovorus and L. brevis strains also performed poorly, and L. johnsonii was the best-performing in terms of acid tolerance, although the number of strains was too low for proper statistical analysis ( Figure 1C and Table S3).

Bile Tolerance
To test the bile tolerance of the strains, their growth in MRS was compared with and without bile using L. rhamnosus GG as the comparator. The tolerance to 0.3% and 0.9% bile correlated positively (pairwise Spearman correlation, r = 0.858 and p < 0.0001 (data not shown)). Furthermore, in the Spearman correlation analysis, the tolerance to acid at pH 2.5 correlated positively to 0.3% and 0.9% bile tolerance (r = 0.223, p = 0.027 and r = 0.216, p = 0.033, respectively (data not shown)).
At the species level, the variation was much higher at milder (0.3%) ( Figure 1D) versus 0.9% bile ( Figure 1E), and the bacteria generally tolerated 0.3% bile well and similarly to the comparator strain. The weakest-performing species in 0.3% bile ( Figure 1D) was L. jensenii, with significantly worse performance than the rest of the species, with the exception of L. iners (Table S4). Likewise, L. crispatus grew poorly in 0.3% bile and performed signifi-cantly worse than every other species, except L. paracasei, L. gasseri, and L. iners (Table S4). The best-performing species in 0.3% bile was L. mucosae, with its two strains, but only visually. Statistically, the best-performing species was L. acidophilus, which performed better than L. paracasei, L. rhamnosus, L. crispatus, and L. jensenii ( Figure 1D and Table S4).
At the strain level, in 0.3% bile (Supplementary Materials Figure S3A), 40 strains performed better than L. rhamnosus GG, and survival in 31 strains decreased by more than 50%. Twelve strains had a survival rate of 10% or lower.
In 0.9% bile, 36 strains performed better than L. rhamnosus GG, and survival in 69 strains decreased by more than 50%. A total of 22 strains had survival rates of 10% or less (Supplementary Materials Figure S3B).

Hydrogen Peroxide Production
H 2 O 2 production by each strain was measured after 1.5 h and 3 h of incubation, expressed as a percentage versus the comparator L. jensenii DSM 20557 [21]. H 2 O 2 production correlated strongly between the two time points (pairwise Spearman correlation, r = 0.788, p < 0.0001 (data not shown)). Of the 98 tested strains, only 50 produced H 2 O 2 after 1.5 h; therefore, only the H 2 O 2 production after 3 h was more closely investigated at the species level. Notably, H 2 O 2 production at 1.5 h correlated negatively with acid tolerance at pH 2.5 (r = −0.230, p = 0.023 (data not shown)) and pH 3.5 (r = −0.312, p = 0.002 (data not shown)) and with bile tolerance at 0.3% (r = −0.229, p = 0.023 (data not shown)) and 0.9% oxgall (r = −0.259, p = 0.01 (data not shown)) and positively with VEC adhesion (r = 0.201, p = 0.048 (data not shown)). Similarly, H 2 O 2 production at 3 h had an inverse relationship with acid tolerance at pH 2.5 (r = −0.227, p = 0.025 (data not shown)) and pH 3.5 (r = −0.304, p = 0.003 (data not shown)) and with bile tolerance at 0.3% (r = −0.207, p = 0.041 (data not shown)). H 2 O 2 production varied highly within species, indicating strain-specific differences. All L. acidophilus, L. jensenii, L. gasseri (excluding L. gasseri Lg-36), and L. crispatus strains (excluding L. crispatus LX11211) produced H 2 O 2 at 3 h, whereas L. rhamnosus, L. johnsonii, L. paracasei, and L. plantarum strains generated little, if any ( Figure 1F). L. fermentum strains were variable in production, as some strains produced relatively high amounts, whereas some did not produce at all ( Figure 1F). Furthermore, vaginal isolates of L. iners strains synthesized limited H 2 O 2 compared to other vaginal isolates. L. crispatus and L. jensenii. L. jensenii produced the highest mean concentration of H 2 O 2 at 3 h, which was significantly better than almost every other species, except for L. acidophilus and L. crispatus ( Figure 1F and Table S6). Additionally, apart from low-producing outliers, L. crispatus generated substantial amounts of H 2 O 2 and performed better than every other species except L. acidophilus, L. fermentum, L. gasseri, and L. jensenii (Table S6).
The number of strains that produced H 2 O 2 increased from 50 to 62 after 3 h incubation (Supplementary Materials Figure S4). H 2 O 2 production was undetectable in 36 strains. A total of 23 strains performed better than the comparator at 3 h: 6 L. jensenii, 5 L. acidophilus, 5 L. crispatus, 5 L. gasseri, Ligilactobacillus ruminis LX11894, L. amylovorus LX11898, L. delbrueckii subsp. bulgaricus Lb-64, Lactococcus lactis Ll-23, and L. fermentum LX11865 strains (Supplementary Materials Figure S4). Almost all vaginal isolates produced H 2 O 2 and were among the top 23 strains performing better than L. jensenii DSM 20557, with the exception of L. crispatus LX11211 and all L. iners strains, which produced limited H 2 O 2 , and L. crispatus LX12212, which generated 97% of the H 2 O 2 levels produced by the control strain.

Adhesion to VECs
The adhesion of bacteria to primary VECs was measured by applying them to VECs at a ratio of 10 tritium-labeled lactobacilli to 1 VEC and by incubating them for 1 h, after which adhered bacteria were counted after washing away the unadhered bacteria. To compare the results between experiments, adhesion in each assay was normalized to that of the comparator strain, L. jensenii LX11796. The adhesion to VECs correlated positively with H 2 O 2 production at 1.5 h (Spearman r = 0.201, p = 0.048 (data not shown)) and negatively with tolerance to 0.3% bile (Spearman r = −0.222, p = 0.029 (data not shown)).
At the species level ( Figure 1G), the adhesion was quite uniform. L. brevis, with its three strains, seemed to adhere the best to the cells, whereas L. iners represented the least adherent species. However, when analyzed statistically, no statistical differences were detected among strains (Table S7).
At the strain level, the bacteria exhibited wide range of adhesion (Supplementary Materials Figure S5). The top 15 strains that adhered the best to VECs included 3 L. rhamnosus strains (LR1049, LX11875, and LX11881), 3 L. jensenii strains (LX11795, 911, and LX11796), and 3 L. crispatus strains (LX1220, LX11797, and LX11799). L. reuteri 1753, two L. brevis strains (LX11860 and LX11864), L. paracasei LX11887, and L. crispatus LX11799 adhered better than or equal to the comparator strain. As the comparator strain in the screen was not a commercial or type strain, it was difficult to determine how well the bacteria adhered compared with a benchmark. However, the strain utilized as comparator was adhered better than the commercial strains in the panel.

Adhesion to Blood Group Antigens A, B, and H
The adhesion of the strains to blood group antigens A, B, and H was measured by in vitro assay using biotinylated antigens and compared with the comparator strain, L. crispatus LMG18199, which adheres well to the A antigen [22]. The specificity for the antigens was low, and most strains adhered similarly to all tested antigens. The adhesiveness also correlated strongly between antigens (pairwise Spearman correlation r = 0. 819 between adhesion to A and B, r = 0.700 between A and H, and r = 0.788 between B and H; p < 0.0001 for all (data not shown)). These values did not correlate well with the other measured parameters in the Spearman correlation analysis.
At the species level, all species adhered to the antigens at similar levels ( Figure 1H-J). L. fermentum, L. paracasei, and L. plantarum species showed the highest variation between strains; therefore, some strains from these species displayed differences in adherence to the antigens. However, statistical significance was only observed between L. rhamnosus and L. crispatus, L. gasserii, L. jensenii, and L. iners, with higher adhesion to all blood type antigens (Tables S8-S10). Furthermore, L. rhamnosus was significantly more adhesive compared to L. paracasei in the case of antigen H adhesion (Table S10).
Some strains showed high variation in their adherence to antigens, whereas others did not adhere or did so to a limited extent (Supplementary Materials Figure S6A-C). Five strains showed better adhesion to the A antigen than the comparator strain: L. paracasei LX11858, L. fermentum LX11866, L. fermentum LX11865, L. fermentum LX11852, and L. fermentum LX11853 (Supplementary Materials Figure S6A), of which the latter four also adhered to B and H antigens better than the comparator strain (Supplementary Materials Figure S6B,C).

Differences in Commercial Strains and Vaginal Isolates
Next, the data were rank-ordered and analyzed as groups (fecal versus non-fecal isolates, commercial versus non-commercial, and vaginal versus non-vaginal). In the rank ordering, the best-performing probiotic was assigned a value of 1, increasing in value for each strain to 98, with the exception of ranking based on H 2 O 2 production (see Section 2.3). The rank-ordered data were also correlated to gastrointestinal and vaginal parameters, and in each correlation analysis, the 10 best-performing probiotics were identified. Generally, with regard to gastrointestinal probiotic characteristics, fecal lactobacilli were the top performers, whereas vaginal isolates predominated when vaginal probiotic characteristics (H 2 O 2 production and adhesion to vaginal epithelial cells) were analyzed.
When commercial strains were compared to non-commercial strains, the 15 studied commercial strains grew better in MRS (p = 0.009). Interestingly, the current commercial strains showed less adhesion to VECs than the non-commercial strains (p = 0.030) (Supplementary Materials Figure S8A).
We also projected the rank-ordered data in two dimensions to determine the bestperforming strains in each projection. When the rank-ordered data for general growth in MRS and acid tolerance at pH 2.5 were projected, three commercial strains-L. casei Lc-11, L. rhamnosus HN001, and L. rhamnosus Lr-32-were among the 10 best-performing strains (Supplementary Materials Figure S8B). L. rhamnosus strains were generally well represented in growth versus acid tolerance, contributing 6 of the top 10 strains (HN001, Lr-32, LX11882, LX11877, LX11879, and LX11881), with L. gasseri 811, L. johnsonii LG0884, L. casei Lc-11, and L. acidophilus LA11883 constituting the remainder.
Similarly, tolerance to 0.3% and 0.9% bile was projected to acid tolerance at pH 3.5 and pH 2.5 (data not shown; Supplementary Materials Figure S9A). Under the milder conditions (pH 3.5 and 0.3% bile (data not shown)), the best-performing strains were four L. fermentum (238, 508, SBS-1, and 1924), three L. acidophilus (LA11883, LX11873, and LX11850), L. gasseri LG11859, L. salivarius Ls-33, and L. mucosae LX11893. However, when the tolerance data under stronger conditions (pH 2.5 and 0.9% bile) were projected, eight of the best-performing strains were L. acidophilus (NCFM, LA11883, LA11890, LA11892, LA11897, 74-2, LA1076, and LX11873), and the remaining two were L. fermentum 508 and L. reuteri 1753 (Supplementary Materials Figure S9B). Thus, with regard to acid and bile tolerance, L. acidophilus strains were among the best-performing. In this comparison of the rank-ordered data, the top 10 strains at pH 2.5 and 0.9% bile did not show any significant differences in other parameters versus the rest of the strains, and among these strains, only 2 were commercial, i.e., L. acidophilus NCFM and 74-2 (Supplementary Materials Figure S9A).  In addition, the parameters were examined separately for the 13 strains isolated from the vaginal tract and compared with the rest of the strains in the rank-ordered data (Supplementary Materials Figure S10A). Tolerance to acid (p = 0.003 for pH 2.5 and p < 0.001 for pH 3.5) and bile (p = 0.004 for 0.3% and p = 0.002 for 0.9%) was lower in vaginally isolated strains versus the other strains in the panel. Furthermore, these strains produced more H 2 O 2 and adhered better to VECs, albeit insignificantly.
When VEC adhesion was correlated to H 2 O 2 production in the rank-ordered data in the full dataset (Supplementary Materials Figure S10B), the 10 best-performing strains were more diverse in origin, with 4 of vaginal origin (L. crispatus LX11797, L. crispatus LX11798, L. jensenii LX11794, and L. jensenii LX11796). The remainder of the 10 best-performing strains comprised L. crispatus LX1220, L. jensenii 911, L. gasseri 811, L. johnsonii LG0883, L. fermentum LX11865, and L. reuteri 1753. The top 10 strains showed good VEC adhesion and H 2 O 2 production, but they had poor tolerance to mild acid (p = 0.042) and bile (p = 0.003 for 0.3% and p = 0.022 for 0.9%). Thus, if a strain shows good VEC adhesion and H 2 O 2 production or is isolated from the vaginal tract, it does not necessarily have good tolerance to acid or bile.

Cluster Analysis of Rank-Ordered Data
To better visualize the characteristics of the 98 strains, 9 clusters were created by hierarchical clustering according to Ward's method ( Figure 3A). Each cluster contained various Lactobacillus species, and none contained just one species. Cluster 8 had the most 'classical' probiotic characteristics ( Figure 3B), with high tolerance to acid and bile and good growth in MRS. In terms of H 2 O 2 production and adhesion to blood group antigens and VECs, the strains in cluster 8 were among the lowest-ranking. This cluster contained the most commercial strains (5 of 14), but otherwise, the commercial strains were distributed across seven clusters, with cluster 5 being the only one without commercial strains. Cluster 1 contained the second highest number of commercial strains (3 of 12), with good acid and bile tolerance but moderate to good rankings for other characteristics, apart from average adherence to VECs. This cluster contained most of the L. acidophilus strains (10 of 15 L. acidophilus strains in the panel).
With regard to the properties of vaginal probiotics, the most notable clusters were clusters 7, 5, and 9, showing high adhesion to VECs and/or high H 2 O 2 production. Additionally, cluster 3 contained strains that produced significant H 2 O 2 and adhered to VECs ( Figure 3B). Cluster 5 (five strains) showed the best adherence to blood group antigens A, B, and H ( Figure 3B). This cluster contained strains that were low in growth but tolerated bile and mild acidic conditions rather well and produced H 2 O 2 moderately. Cluster 5 comprised L. amylovorus LX11898, L. brevis LX11864, and three L. fermentum (L. fermentum LX11852, L. fermentum LX11853, and L. fermentum LX11865). All of these strains were of fecal origin. The other cluster with good adherence to blood group antigens A, B, and H was cluster 9 (see below), which, likewise, did not contain any vaginal isolates. Strains from cluster 9 also adhered well to all blood group antigens, with good growth acid resistance and moderate bile resistance, but failed to produce H 2 O 2 . This cluster contained more than half of the L. rhamnosus strains (11 of 20 in the panel) and two L. paracasei and two L. plantarum strains; however, likewise, none was of vaginal origin. Cluster 3, containing six different species (L. acidophilus, L. fermentum, L. helveticus, L. delbrueckii bulgaricus, L. jensenii, and L. crispatus) had the opposite profile to cluster 9, producing H 2 O 2 well but adhering poorly to blood group antigens. Most of the L. jensenii strains (four of six strains) resided in this cluster. The other two L. jensenii strains were in cluster 7, with good H 2 O 2 production and VEC adhesion, implying that the L. jensenii species has potential as a probiotic for vaginal health.  The vaginal isolates in the screening panel were distributed among clusters 2, 3, 6, and 7, indicating that lactobacilli from the vaginal tract have varying characteristics. Cluster 7, which was also the largest of the clusters, contained the most vaginal isolates (7 of 16 strains) (L. crispatus LX11211, L. crispatus LX12212, L. crispatus LX11797, L. crispatus LX11798, L. crispatus LX11799, L. jensenii 11794, and L. jensenii LX11795). These strains showed rather good adhesion to blood group antigens and VECs and good H 2 O 2 production but moderate tolerance to acid and bile and growth in MRS.

Discussion
To elicit a health benefit, an oral probiotic strain should tolerate acid and bile, show good adhesion to intestinal or vaginal epithelia, depending on the target site, and possess antimicrobial properties. Furthermore, to be produced commercially, the strain should be culturable on a large scale. The aim of this study was to characterize the in vitro probiotic properties of 100 Lactobacillus sensu lato strains that were isolated primarily from the intestinal and vaginal tracts and rank them into various functional groups to determine their suitability for gastrointestinal and vaginal indications. Although important probiotic characteristics, such as H 2 O 2 production, adhesion to vaginal epithelial cells, and bile and acid resistance were measured in this preliminary screening, other important probiotic safety-related properties, such as hemolysis and, most notably, antibiotic resistance [23], were not. Thus, further in-depth safety evaluation of the strains is needed for any future clinical studies, as has been done, for instance, by Pino and coworkers [24]. However, the results obtained in the current study can be utilized as a primary strain selection tool for specific health indications. The results show high strain-specificity in each characteristic tested, and the strains fell into nine variable clusters. The results confirm the widely held notion that probiotic properties are strain-specific [11]. None of the clusters was dominated by a single species. However, the clustering differentiated strains that tolerated bile and acid, indicating their potential for gastrointestinal indications. Furthermore, the clustering identified strains that adhered to vaginal cells and produced H 2 O 2 , conferring on them vaginal benefits.
Lactobacilli possess differential surface characteristics and express various enzymes, creating strain and species specificity in their response to environmental conditions and stresses [25]. Bile salts have strong antimicrobial potential, and tolerance to bile determines the ability to survive in the small intestine [26]. Resistance to bile parallels that to other stresses, such as acid and oxidative stress [25][26][27], which we also noted in our study. This tolerance differed based on the strength of such conditions, and our study included strains that varied in their tolerance. The 'traditional' probiotic species-L. acidophilus and L. rhamnosustolerated the stronger acid and bile conditions consistently, as expected, because the genetic machinery for bile and acid tolerance is well described for Lactobacillus [26]. All 15 L. acidophilus strains and most (16/20) L. rhamnosus strains were fecal in origin, and logically, the fecal isolates tolerated bile better than vaginal isolates in the rank-ordered data because the fecal strains had adapted to the conditions in the gastrointestinal tract.
Overall, the commercial strains were easier to culture, and 4 of the 10 easiest strains to culture were commercial, whereas none of 10 worst performers was. Ease of production, viability, and stability, are the key factors in the selection of commercial probiotics and are important for industrial-scale production. In addition, probiotics need to tolerate various stresses during production [3,28]. Thus, as expected, many of the easiest strains to culture were commercial strains. Conversely, L. iners, L. brevis, L. amylovorus, and L. mucosae, all of which are non-commercial strains, grew poorly in MRS. However, the small number of these strains in the panel might have biased this result; for instance, numerous L. rhamnosus, L. crispatus, and L. fermentum contained both strains that grew well and grew poorly.
Traditionally, adherence has been examined with intestinal epithelial cells, such as Caco-2, or mucus, such as that from pigs or humans [29]. In our study, we used a different approach, measuring the adherence of lactobacilli to histo-blood group antigens A, B, and H, as well as to VECs. The ABO histo-blood group system consists of two antigens (A and B) and four blood types (types A, B, AB, and O), of which group O expresses only H antigen, the biosynthetic precursor of antigens A and B. These antigens are widely expressed in red blood cells and many tissues and secretions, including the gastrointestinal and vaginal mucosae [30]. However, ABH antigens are not present in all individuals; non-secretors lack functional fucosyltransferease-2 and do not express these antigens in their secretions or mucosa, instead expressing Lewis A antigen [31].
Some pathogens and their toxins can bind blood group antigens directly [32], such as H. pylori [33], Norwalk virus [34], norovirus GII.4 genotype [35], and rotavirus [36]. Lactobacilli strains of L. crispatus, L. mucosae, L. plantarum, and L. paracasei express blood group antigen-binding adhesins and therefore may compete in adhesion mechanisms that impede pathogenesis [22,[37][38][39][40]. Moreover, blood group antigens, glycans on ABO antigens, and Lewis antigen shed into the intestinal lumen can be fermented by intestinal bacteria such as bifidobacteria and Bacteroides spp.; thus, these glycans can serve as energy sources for bacteria and affect the composition of the gut microbiota [14]. Research shows that the composition of intestinal mucosal microbiota is affected by the ABO blood type and the secretor status of the host [14,41], especially the presence of B antigen [39], although contradictory results have also been reported [42]. However, information on the adhesion properties of probiotics to different blood group antigens could be relevant in the design of personalized probiotics for the market because specific adhesins toward a certain blood group antigen could further increase the colonization potential of probiotics. In our study, strains were differentiated and ranked as having antigen-specific adhesion, non-specific adhesion (similar affinity to all), or equal affinity (highest median adhesion). This categorization could be used to tailor probiotics based on adherence toward a certain blood group antigen or implement a more generic approach that is suitable for most consumers. Among the top 10 strains with the most variability in their adhesion to blood group antigens, 2 were commercial (L. rhamnosus HN001 and L. rhamnosus Lr-32).
It was previously reported that bacterial survival in an upper gastrointestinal experimental model depended on the secretor and non-secretor status of the donor, with bacteria from the latter being more vulnerable to acid and bile [43]. However, we do not know whether the donor from which the fecal strains were isolated was a non-secretor or secretor; thus, ideally, future studies should harvest bacteria from individuals with different blood group antigens and secretor statuses. Notably, fecal strains adhered better to all blood group antigens compared with strains sourced elsewhere. Furthermore, clustering of the rank-ordered data showed that adhesion to antigens A and B by some strains was associated with poor acid and bile tolerance, indicating that they would survive poorly in the upper gastrointestinal tract, although there were still some clusters that had both good acid and bile tolerance and good adhesion to blood group antigens. However, identifying new probiotic candidates solely from among those that best tolerate acid and bile would overlook other beneficial properties. For instance, the production of H 2 O 2 , an asset of vaginal lactobacilli, correlated negatively with tolerance to acid and bile. Considering women's health, such tolerance might be dispensable for intravaginal probiotics, whereas the bacteria in oral supplements would first need to travel through the gastrointestinal tract, resisting the acidity of the stomach and the bile from liver [26]. To this end, encapsulation technologies can be used to overcome such poor survival [44].
Vaginal bacteria can be classified into five main types by community. Four are predominated by Lactobacillus spp.-L. crispatus, L. gasseri, L. iners, and L. jensenii-whereas the fifth comprises a mixed community of strictly anaerobic bacteria [45]. Lactobacilli protect the vaginal mucosa through adherence to the vaginal epithelia and their antimicrobial properties [46]. The production of lactic acid, H 2 O 2 , and antimicrobial agents is considered a beneficial attribute of vaginal lactobacilli [9]. Whereas lactic acid inhibits the growth of pathogenic bacteria by lowering pH, the function of H 2 O 2 is not fully elucidated, but it has been suggested to destroy bacteria owing to a lack of H 2 O 2 -degrading enzymes such as catalase peroxidase [9,47]. Nevertheless, 70% to 95% of lactobacilli in the vaginal microbiota of healthy women produce H 2 O 2 , and the levels of these lactobacilli decrease in women with vaginal infections [48][49][50]. L. gasseri, L. jensenii, and L. crispatus species commonly predominate in the vaginal tracts of healthy women [10], and more than 90% of the strains in these genera produce H 2 O 2 [48]. For instance, at least 80% of isolates of L. jensenii, L. salivarius, L. rhamnosus, and L. vaginalis generate high amounts of H 2 O 2 , as does L. gasseri, although to a lesser extent [51]. In our study, 53% (52/98) of strains produced H 2 O 2 , the most prominent of which were L. gasseri, L. jensenii, and L. crispatus, whereas L. rhamnosus, L. johnsonii, and L. brevis synthesized lower amounts. However, most strains in the screening panel were of fecal rather than vaginal origin. Most of the vaginal isolates (10/13) produced H 2 O 2 , with the exception of L. iners. This was expected because L. iners strains are reported to generate limited amounts of H 2 O 2 [52]. Furthermore, L. iners is not well adapted to grow in MRS, preferring sheep blood agar. L. iners also produce L-lactic acid but not D-lactic acid, which is regarded as a more protective isomer for vaginal health [52]. The function of L. iners in vaginal health is controversial; it is commonly found in healthy women but is also often encountered during vaginal dysbiosis [52][53][54]. Moreover, the L. iners strains adhered poorly to VECs in our study compared with the L. crispatus and L. jensenii strains. Notably, L. iners lack adhesins that are common to other lactobacilli, instead expressing fibronectin-binding protein, which is similar to pathogenic Staphylococcus aureus [52].
In dysbiosis of the vaginal microbiota, such as during BV, lactobacilli are depleted from the vaginal microbiota and are overrun by various facultative anaerobes [55,56]. Probiotic lactobacilli with or without antibiotics administered orally or intravaginally have shown promise in reducing the risk of vaginal infections [55,57,58]. Good adherence of lactobacilli to vaginal epithelia is a potential mechanism for protecting the vaginal mucosa from pathogenesis by inhibiting and interfering with the adhesion of pathogenic microorganisms [46]. In our study, the adhesion of lactobacilli to VECs correlated positively with good H 2 O 2 production. In the rank-ordered data, when the entire dataset was projected according to VEC adhesion and H 2 O 2 production, 4 of the top 10 strains were vaginal in origin, highlighting their potential as probiotics for vaginal health. These vaginal isolates, especially L. jensenii, also showed lower tolerance to acid and bile compared with the rest of the strains. Future studies should compare the responses of less tolerant species, such as L. jensenii, to those with greater tolerance, such as L. acidophilus and L. plantarum, and determine whether gradual exposure to increasing stress improves tolerance in L. jensenii [26].
One drawback of our study is the use of bovine bile (Oxgall), which slightly reduces the pH of culture media [59]. We did not control the pH of the culture medium in the bile tolerance assays, as has been done in some other studies [60]; thus, the lower viability in bile could have been caused by the accumulation of lactic acid and other organic acids that are produced by the bacteria [26]. To survive passage through the gastrointestinal tract, the adaptation of microorganisms to a stressor might enhance survival under another stressor that is encountered [61]. This species-specific cross protection can be beneficial when cells are exposed to a combination of stresses [61], and these adaptive responses can be addressed further in examining the probiotic properties of lactobacilli.
In conclusion, we screened 98 strains for their overall growth in MRS, tolerance to acid and bile, H 2 O 2 production, and adhesion to VECs and blood group antigens. There were clear differences in these characteristics depending on the source of the strain, but there were also strain-dependent properties between species. We found that strains that were of fecal origin adapted better to acidic and high-bile conditions, whereas vaginal strains, with the exception of L. iners, were good H 2 O 2 producers and adhered well to VECs, although some of these strains were less tolerant to acid and bile. Future encapsulation technologies and an improved understanding of the genetic background in the development of tolerance and beneficial properties can guide and improve the selection of probiotics.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/microorganisms11020329/s1, Figure S1: Growth on MRS at the strain level. Figure S2: Acid tolerance at the strain level. Figure S3: Bile tolerance at the strain level. Figure S4: Hydrogen peroxide production at the strain level at 3 hours. Figure S5: Adhesion to vaginal epithelial cells at the strain level. Figure S6: Adhesion to blood group antigens at the strain level. Figure S7: Box plots of measured characteristics of the lactobacilli isolated from feces in comparison to strains of non-fecal origin. Figure S8: Characteristics of the 16 studied commercial probiotics in comparison to non-commercial strains. Figure S9: Characteristics of the 10 best-performing strains with rank-ordered acid (pH 2.5) and strong bile (0.9% oxgall) tolerances in comparison to the other strains in the dataset. Figure S10: Characteristics of the 13 vaginally isolated lactobacilli in comparison to strains from other origins. Table S1: Statistical analysis of growth in MRS OD600. Table S2: Statistical analysis of acid tolerance at pH 3.5 (log reduction). Table S3: Statistical analysis of acid tolerance at pH 2.5 (log reduction). Table S4: Statistical analysis of bile tolerance, growth in 0.3% oxgall, % of growth in MRS w/o bile. Table S5: Statistical analysis of bile tolerance, growth in 0.9% oxgall % of growth in MRS w/o bile. Table S6: Statistical analysis of hydrogen peroxide production at 3 h% of comparator strain. Table S7: Statistical analysis of epithelial cell adhesion % of adhesion of comparator. Table S8: Statistical analysis of average A antigen adhesion % of adhesion of comparator. Table S9: Statistical analysis of average B antigen adhesion % of adhesion of comparator. Table S10