Lactic acid bacteria (LAB) are applied worldwide in the production of a variety of fermented food products. Additionally, specific Lactobacillus species are nowadays recognized for their health-promoting effects on the consumer. To optimally exert such beneficial effects, it is considered of great importance that these probiotic bacteria reach their target sites in the gut alive.
In the accompanying manuscript by Bron et al. the probiotic model organism Lactobacillus plantarum WCFS1 was cultured under different fermentation conditions, which was complemented by the determination of the corresponding molecular responses by full-genome transcriptome analyses. Here, the gastrointestinal (GI) survival of the cultures produced was assessed in an in vitro assay. Variations in fermentation conditions led to dramatic differences in GI-tract survival (up to 7-log) and high robustness could be associated with low salt and low pH during the fermentations. Moreover, random forest correlation analyses allowed the identification of specific transcripts associated with robustness. Subsequently, the corresponding genes were targeted by genetic engineering, aiming to enhance robustness, which could be achieved for 3 of the genes that negatively correlated with robustness and where deletion derivatives displayed enhanced survival compared to the parental strain. Specifically, a role in GI-tract survival could be confirmed for the lp_1669-encoded AraC-family transcription regulator, involved in capsular polysaccharide remodeling, the penicillin-binding protein Pbp2A involved in peptidoglycan biosynthesis, and the Na+/H+ antiporter NapA3. Moreover, additional physiological analysis established a role for Pbp2A and NapA3 in bile salt and salt tolerance, respectively.
Citation: van Bokhorst-van de Veen H, Lee I-C, Marco ML, Wels M, Bron PA, Kleerebezem M (2012) Modulation of Lactobacillus plantarum Gastrointestinal Robustness by Fermentation Conditions Enables Identification of Bacterial Robustness Markers. PLoS ONE 7(7): e39053. https://doi.org/10.1371/journal.pone.0039053
Editor: Arnold Driessen, University of Groningen, Netherlands
Received: April 12, 2012; Accepted: May 14, 2012; Published: July 3, 2012
Copyright: © 2012 van Bokhorst-van de Veen et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: PAB is partly employed within the research program of the Kluyver Centre for Genomics of Industrial Fermentation which is part of the Netherlands Genomics Initiative/Netherlands Organization for Scientific Research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: All authors are affiliated to NIZO food research, which participates in the Kluyver Centre for Genomics of Industrial Fermentation and TI Food & Nutrition consortia. TI Food & Nutrition has patented the technology and genes described in these papers that are shown to be involved in bacterial robustness, but has no products in development or marketed products to declare. This does not alter the authors’ adherence to all the PLoS ONE policies on sharing data and materials, as detailed online in the guide for authors.
According to the world health organization (WHO) probiotics are defined as live microorganisms which, when administered in adequate amounts, confer a health benefit on the host . The most widely applied probiotic strains belong to the genera Lactobacillus and Bifidobacterium , . Probiotics are most commonly provided as freshly fermented food products, non-fermented food products to which probiotics are added, or as dried bacterial preparations , . The viability of bacteria is considered an important trait for probiotic functionality, justifying the interest to unravel the mechanism(s) involved in gastrointestinal (GI)-tract survival at the molecular level , , , .
During passage through the GI-tract, probiotics encounter several stresses including acidity in the stomach which may reach a pH as low as 1 during fasting . This low extracellular pH affects the proton motive force of the bacterial cells, thereby disrupting the energy supply required for processes such as membrane transport . In addition, lower intracellular pH values caused by acidic conditions may inhibit specific pathways by damaging acid-sensitive associated enzyme functions . After stomach passage probiotic strains reach the small intestine, where bile acids act primarily as a surfactant that can disrupt bacterial membranes  and damage macromolecules such as RNA and DNA through the generation of free oxygen radicals . Moreover, protonated bile acids can freely pass bacterial cell membranes and release protons intracellularly which might lead to lowering of the intracellular pH, analogous to acid stress .
Among the lactobacilli, Lactobacillus plantarum is encountered in a plethora of fermentations, ranging from vegetables to dairy, meat and sourdough , . L. plantarum is also frequently encountered as a natural inhabitant of the GI-tract of several mammals, including humans . In addition, L. plantarum NCIMB8826 was demonstrated to effectively survive passage of the human stomach, reached the ileum in high numbers, and was detected in the colon . A single colony isolate of this strain (designated L. plantarum strain WCFS1) was the first Lactobacillus strain of which the full genome sequence was published . Subsequently, sophisticated bioinformatics tools were developed for this LAB strain, including an advanced genome annotation , genome-based metabolic models , as well as effective mutagenesis tools . This enables the molecular investigation of gene-regulatory mechanisms underlying the observed GI-tract persistence of L. plantarum WCFS1.
The availability of full genome sequences has also enabled the exploration of genomic diversity among L. plantarum strains and its association to differential phenotypes , , , , . To enable the identification of genes of which the relative expression level is correlated to the phenotype of interest, we recently developed a complementary transcriptome-phenotype matching strategy for L. plantarum (see accompanying paper by Bron et al.). Here, we employed this fermentation genomics platform to correlate transcriptome data to GI-tract survival. These correlations led to the identification of 13 candidate effector molecules for GI-tract persistence. A subsequent gene deletion strategy established a definite role in GI-tract persistence for the AraC-family transcription regulator encoded by lp_1669, the penicillin-binding protein Pbp2A involved in peptidoglycan biosynthesis, and the Na+/H+ antiporter NapA3.
Gastric Acidity is a Critical Determinant of L. plantarum Survival
An in vitro assay was developed that allows high-throughput assessment of bacterial GI-tract survival (fig. 1A). Two independent reference L. plantarum WCFS1 cultures that were harvested during logarithmic phase of growth (OD600 = 1.0) displayed a 6-log decrease in CFU·ml−1 in the GI-tract assay (fig. 1B). The survival curves of these reference cultures demonstrated the major impact on survival exerted by gastric juice on L. plantarum viability and the relatively minor effect of the conditions which resembled the small intestine (fig. 1B). This differential effect on survival during the two stages within the GI-tract assay was consistently observed for all cultures tested, irrespective of the fermentation conditions applied or the growth phases from which bacterial cells were harvested.
L. plantarum WCFS1 cultures were grown aerobically at 28°C in 2×CDM containing normal acid concentration, at a pH of 5.8 and without NaCl. The cultures were harvested at mid-exponential phase (OD600 = 1.0) and subjected to an upper GI-tract mimicking assay (A): After 60 min incubation in gastric juice containing pepsin and lipase at a pH of 2.4 (logarithmic cells) or 2.3 (stationary cells), cultures were neutralized with NaHCO3 and pancreatic juice containing pancreatin and bile acids was added and incubation continued for 60 min (see materials and methods for details). Preceding and during incubation, samples were taken for CFU determination (aligned arrows). Panel B shows the relative survival of two independent cultures in logarithmic phase (solid lines) and stationary phase (dashed lines) during the GI-tract mimicking assay. Input (CFU determination immediately prior to the GI assay) is set at 0 Log10 CFU ml−1, data presented are averages of technical sextuplicates (+standard deviation).
The strongest determinant in the loss of survival during the gastric juice treatment appeared to be pH. For screening log-phase cells of L. plantarum, a pH of 2.4 was used for cells, as lowering or increasing of the gastric juice pH by 0.1 pH unit resulted in death or survival of almost all cells, respectively (over 7-log differences, data not shown). L. plantarum cells harvested at the stationary phase of growth consistently displayed a higher tolerance to the gastric juice treatment, which is exemplified by their higher survival rate in the GI-tract assay when a reduced pH of 2.3 was used (fig. 1B) at which the cells harvested from the logarithmic phase of growth were nearly all killed within 60 minutes of incubation.
Fermentation-enhanced Digestive Tract Survival
We examined the effects of different growth conditions on L. plantarum WCFS1 GI-tract survival by applying samples derived from the fermentation-genomics platform described in the accompanying paper by Bron et al. to our in vitro GI-tract assay. The results demonstrate that fermentation conditions used to culture L. plantarum WCFS1 conferred a profound influence on the GI-tract survival. Variable fermentation conditions resulted in major differences (a reduction of 7 logs for the logarithmic population and 5 logs for stationary cells) in L. plantarum WCFS1 survival after incubation in gastric juice (fig. 2). Notably, survival of cultures grown in different fermentation conditions strongly exceeded the levels of variation in survival observed in independent GI-tract assays (fig. 1B).
Log10 CFU ml−1 determination of L. plantarum WCFS1 in logarithmic phase (A) and stationary phase (B) after 20 (light grey), 40 (dark grey), and 60 min (black) gastric juice incubation. Input is set at 0 Log10 CFU ml−1, # = fermentation number, cultures were grown in 2× CDM with (300 mM) or without (0) NaCl; with normal amino acid concentration (2) or reduced (1.1); at 28°C or 37°C; medium buffered at a pH of 5.2, 5.8, or 6.4; and aerobically (O2) or anaerobically (N2). Data presented are averages of technical sextuplicates (+ standard deviation).
To identify the fermentation conditions that significantly affected the survival rate in the simulated GI-tract conditions, a Mann-Whitney U test-based correlation analysis was performed in FermDB on all time points measured (See accompanying manuscript by Bron et al. for correlation analyses details). The presence of 300 mM additional NaCl in the growth medium resulted in a significant (P<0.05) negative influence on L. plantarum GI-tract survival irrespective whether cells were analyzed after collection from either logarithmic or stationary phase of growth (shown for 60 min incubation in fig. 3A and B). L. plantarum grown in more acidic conditions (pH 5.2 instead of pH 6.4) and harvested in stationary phase showed a significantly (P<0.05) enhanced the gastric juice survival rate (shown for 60 min in fig. 3C).
Box plots of NaCl and 60 min GI-tract survival of logarithmically (A) and stationary (B) grown cultures and of pH and 60 min GI-tract survival of stationary cells (C). Results are based on data from all fermentations used in this study (see fig. 2A). * P-value <0.05 compared with 0 mM NaCl (A and B) or pH 6.4 (C).
Transcriptome to Phenotype Association Identifies Candidate Effector Molecules for GI-tract Survival
In parallel with the GI-tract survival patterns, transcriptome profiles were obtained for logarithmic cells harvested from all fermentation conditions employed in this study (see accompanying paper by Bron et al.). To investigate whether high- and low-rate surviving cultures in the GI-tract assay could be distinguished based on the expression of specific genes, the cultures were first ranked on their GI-tract survival after gastric juice incubation (t = 60 min). For cultures that had retained undetectable survival rates after 60 min of gastric incubation, the relative survival rates after 20 min and 40 min of gastric incubation, were employed to refine their relative survival ranking (fig. 2A).
The transcriptomes of the eight cultures with the highest survival rates and the eight cultures with the lowest survival in the GI-tract assay were clearly distinguishable according to principal component analysis (PCA) (fig. 4). This result indicated that the transcriptomes contained information (genes) within the first two components of the PCA which might allow the discrimination between high- and low-survival rates in the GI-tract assay. To identify specific transcripts that discriminate between low and high GI-tract survival, and thus can be regarded as candidate robustness markers, the random forest algorithm was applied (see the accompanying paper by Bron et al.). This allowed the identification of transcripts that have a high contribution to accurately predict the low- and high-survival outcomes (table S1).
Validation of Target GI-tract Survival Effector Molecules by Mutagenesis
To validate the association of the expression level of specific genes in L. plantarum with GI-tract survival, the 13 genes with the highest ranking based on the criteria described in the Materials and Methods section were targeted by genetic engineering (table 1), aiming to improve GI-tract survival beyond the levels observed with the wild-type strain. Therefore, the direction of the correlation between transcript intensity and survival in the GI-tract assay determined whether a gene would be targeted for overexpression (positive correlation, see fig. 5A for an example) or gene-deletion (negative correlation, fig. 5B).
The eight best and eight worst fermentations (see fig. 2A) are ranked with increasing GI-tract survival. *Data was normalized to correct for between slide variation . R2 thrC = 0.71, R2 pbp2A = 0.70.
Genes targeted for sakacin-inducible overexpression were thrC, lp_0149, hicD3, and lp_1357 (table 1). For folB overexpression, we used a previously constructed mutant that overexpresses the entire folB-folK-folE-folC2-xtp2-folP cluster , . Sakacin P induced overexpression of the cytoplasmic hicD3 and thrC and the downstream lp_2759 gene products could readily be confirmed by SDS-PAGE analysis of cell-free extracts of induced cultures (figure S1). In contrast, overproduction of the membrane-anchored (lp_1357) and transmembrane proteins (lp_0148–0150) were not distinguishable by SDS-PAGE (data not shown). Although overexpression could only be demonstrated for two of the genetic loci, we applied all overexpression strains to our GI-tract assay. The constructed overexpression and gene deletion mutants were grown until the logarithmic growth phase and subjected to the GI-tract assay. The survival of the Sakacin P induced overexpression mutants was anticipated to improve when compared to a control strain harboring the empty induction plasmid (fig. 6). Although not significant, the contrary seemed to be the case, since the slight effects that were observed in some of the experiments suggested that the expression of the cloned genes reduced the survival capacity of these cells rather than improved.
Log10 CFU ml−1 determination of mid-exponentially grown in batch L. plantarum mutants after 60 min gastric juice incubation (white bars) and subsequent 60 min pancreatic juice incubation (grey bars). Input is set at 0 Log10 CFU ml−1. Empty vectors are pSIP411 (A) and pNZ7021 (B). L. plantarum harboring pNZ3430 (over-lp_1357), pNZ3431 (over-hicD3), pNZ3432 (over-thrC+lp_2759), pNZ3433 (over-lp_0148∼0150), and pNZ7026 (over-lp_3294∼3299). Data presented is the average of technical sextuplicates (+ standard deviation).
In contrast, the L. plantarum Δpbp2A::cat, Δlp_1669::cat, and ΔnapA3::cat mutants showed significantly improved survival in the GI-tract assay, as compared to their parental strain (fig. 7). These strains harbored disruptions in genes associated with poor survival in gastric stress. Notably, we have combined the individual mutants described here to construct Δpbp2A-ΔnapA3::cat and ΔnapA3-Δlp_1669::cat. However, these double gene deletion derivatives displayed robustness phenotypes comparable to the single ΔnapA3::cat gene deletion derivative, indicating that the positive effect on GI robustness of these mutations appeared not cumulative (data not shown). Nevertheless, these results establish the involvement of certain fermentation-condition dependent gene products in GI survival.
Log10 CFU ml−1 determination of logarithmic (OD600 = 1.0) batch L. plantarum mutants after 60 min gastric juice incubation (white bars) and subsequent 60 min pancreatic juice incubation (grey bars). Input is set at 0 Log10 CFU ml−1. Δpbp2A::cat = L. plantarum NZ3412CM, Δ1669::cat = L. plantarum NZ3417CM, Δ1817::cat = L. plantarum NZ3414CM, ΔpacL3::cat = L. plantarum NZ3415CM, and ΔnapA3::cat = L. plantarum NZ3416CM. * P-value <0.05, ** P-value <0.01 compared with wild type (wt). Representative of two independent experiments, data presented are averages of technical sextuplicates (+ standard deviation).
Pbp2A is annotated as a penicillin-binding protein involved in peptidoglycan biosynthesis, Lp_1669 is predicted to be a transcription regulator, and NapA3 is homologous to Na+/H+ antiporters. To gain more insight in the mechanisms by which these proteins influence robustness, growth of the parental strain and the Δpbp2A::cat, Δlp_1669::cat, and ΔnapA3::cat derivatives was monitored under standard- and stress-conditions. At 28°C in complex culture medium (MRS), the growth rates of the mutants did not differ from the wild-type, nor did the addition of H2O2 (1 to 5 mM), lysozyme (0.025 to 3.2 g/ml), or SDS (0.9 to 30 g/l) induce differences in growth rate of the mutants compared with the wild type strain. However, the presence of bile salts (10 to 50 mM) in the culture medium reduced the maximum growth rate of Δpbp2A::cat to 20% as compared to the parental strain (data not shown). This result indicates that Pbp2A contributes to the survival capacity of L. plantarum in low-pH, stomach like conditions, but also improves bile tolerance, but not to tolerance to detergents in general.
The addition of NaCl to the growth medium reduced the growth rate of ΔnapA3::cat to 20% (400 mM) and 80% (1 M) of the wild type (data not shown). As NapA3 is a Na+/H+ antiporter which might be affected by extracellular pH, the growth of the ΔnapA3::cat mutant was monitored under different starting pH conditions (pH 4.6 to 6.4) in the presence and absence of NaCl (300 mM). The growth rate of the mutant appeared unaltered during growth in the absence of salt. Only the presence of NaCl reduced the growth rate of ΔnapA3::cat under all measured conditions (data not shown). These results support a role of this function in salt tolerance, which in our experiments, appeared to be independent of the pH.
Contrary to ΔnapA3::cat and Δpbp2A::cat, a specific phenotype was not established for the transcription regulator Lp_1669. To elucidate the regulon associated with this regulator, the transcriptome profile of the NZ3417CM (Δlp_1669::cat) strain was compared to that of the wild-type strain grown in 2× CDM  or MRS. The results showed that the Lp_1669-deficient strain displayed enhanced expression of genes belonging to the main functional class of cell envelope associated functions, and more specifically to its subclass of surface polysaccharides, lipopolysaccharides, and antigens. This effect of the mutation was observed independent of the medium used (fig. 8 and table S2 and S3). Analysis at the individual transcript level revealed that the capsular polysaccharide (CPS) clusters cps2, cps3, and cps4 were induced in the MRS-grown Lp_1669-deficient strain as compared to the wildtype, suggesting that the regulatory function encoded by lp_1669 is involved, either directly or indirectly, in the regulation of CPS biosynthesis. Notably, especially the expression of the cps2 cluster was induced in 2× CDM grown Lp1669 deficient cells (table S2 and S3). Analysis of monosaccharide composition revealed minor changes in CPS sugar composition of the Lp_1669-deficient strain in comparison to the wild type strain (table 2). Galactosamine was only detected in the mutant strain, whereas arabinose was found only in the wild-type strain. Rhamnose and glucosamine also tended to be slightly more abundant in the wild type L. plantarum WCFS1. Moreover, the average molar mass of Δlp_1669::cat strain-derived CPS was 1.5-fold higher compared to the wild type (Table 2). This indicates that Lp_1669 seems to be involved in subtle CPS modification, specifically in chain length determination. These observations might also (partially) explain the observed increased gastrointestinal survival of the L. plantarum Lp_l669-deficient strain.
Yellow rectangle nodes represent growth in 2× CDM (left node) or in MRS (right node). Round nodes represent single genes with their corresponding lp_number, which is the number as annotated in the sequenced parental L. plantarum WCFS1 strain , green lines represent down-regulation, and red lines represent up-regulation of the gene compared to the parental strain. The colors of the round nodes represent the annotated main class. See figure S2 of the supplemented materials for the original cytoscape file.
This study demonstrates that the production method, medium composition, and stage of growth strongly influenced the GI-survival efficacy of L. plantarum WCFS1. Combining the fermentation and survival data pinpointed to specific fermentation conditions that may enhance robustness (low salt and low pH), whereas genome association analysis of the transcriptome and survival data revealed 13 genes potentially involved in GI-survival.
As reported previously , cells harvested from stationary phase generally were more robust than logarithmically growing cells, and in particular, those cells displayed enhanced survival in gastric juice which overall had a dramatically larger impact on survival compared to pancreatic juice. The influence of acidity on GI-tract survival was also emphasized by the observation that lowering the gastric juice pH by as little as 0.1 unit had a pronounced impact on survival. Differences among bacterial species in their sensitivity to gastric and intestinal secretions have been observed before , ,  and a higher sensitivity for acid than bile stress was also noted for L. rhamnosus, as well as for other L. plantarum strains , . The finding that exposure to low pH during growth enhances GI-survival is in agreement with earlier observations that pre-adaptation to sublethal stress conditions enhances the subsequent robustness of bacteria to lethal stress conditions , supporting the suitability of the fermentation genomics platform and bioinformatics tools employed in this study and the accompanying paper by Bron et al. For salt it is known that it can protect against, but also increase susceptibility to, other stresses , . Moreover, these results clearly establish that fermentation conditions have a major impact on the GI-tract associated stress tolerance of bacterial cultures, and that specifically mild salt stress and lower pH adaptation may elicit adaptive responses that reduce and support such stress tolerance, respectively.
Overexpression and gene deletion derivatives of the parental strain were constructed, depending on the direction of the predicted correlation to GI-tract survival and aiming to enhance this trait. Three of the five constructed gene deletion derivatives displayed enhanced GI-tract survival, confirming the predicted role of the targeted genes. By contrast, none of the overexpression derivatives displayed improved robustness behavior in the GI-tract assay, and all had survival characteristics that were virtually identical to those of the parental strain. A possible explanation for these observations may be found in the potential disruption of a gene-regulatory network by the deletion of a single gene in that network, while overexpression of a single element from a complementary gene-function network may not provide the same effect as the enhanced expression of all elements in the network. The fact that 3 out of 5 gene deletion derivatives displayed enhanced survival is in line with earlier observations , ,  and can be explained by the fact that the random forest algorithm also leads to the identification of non-causal relationships, reiterating the importance of follow-up mutagenesis approaches to establish a definite role for candidate biomarkers identified with this algorithm.
The 3 genes for which the importance in GI-tract survival could be confirmed by gene deletion encode a AraC family regulator (Lp_1669), a Na+/H+ antiporter; NapA3, and a penicillin binding protein; Pbp2A. Notably, all three proteins are associated with cell wall modification and transport, and their mutation may lead to cell envelope modulation. This finding per se, may not be qualified as surprising, because the cell envelope is the first line of defense against stresses . Moreover, the resistance to acid and adaptation to bile stress in L. plantarum WCFS1 has been associated with membrane integrity and cell envelope modifications, respectively , . The AraC family of regulators to which Lp_1669 belongs  is characterized by transcriptional regulators that act mostly as activators. However, in some cases these regulators serve as repressors of transcription or as both activators and repressors . The observed effect of Lp_1669 on GI-tract survival is likely to be indirect, possibly via CPS remodeling, because the Lp_1669-deficient strain had CPS with a higher molar mass that might result in a thicker CPS layer around the cells. It has been demonstrated that the presence of EPS/CPS improved the in vivo GI survival of L. rhamnosus GG . The Na+/H+ antiporter NapA3 might affect GI survival via a role in pH homeostasis. Because disruption of napA3 improved GI-tract survival, it seems likely that NapA3 exports sodium ions associated with the influx of protons, thereby decreasing its internal pH and proton motive force due to the acid stomach conditions. This is also in line with our observation that the gene deletion derivative is only reduced during growth in the presence of sodium salts. Finally, pbp2A encodes the penicillin binding protein 2A which is annotated to be involved in peptidoglycan biosynthesis . Disruption of pbp2A improved the acid stomach condition survival, while it decreased the growth rate in the presence of bile. Noteworthy in this respect is the finding that the compositions of peptidoglycan directly affects the integrity of the cells and can influence the acid- and bile-tolerance , , , . Moreover, transcriptome analysis of L. acidophilus NCFM and L. plantarum WCFS1 demonstrated that many genes related to cell membrane and peptidoglycan biosynthesis displayed altered expression profiles during exposure to bile , . An increased acid sensitivity by the inactivation of penicillin binding proteins is found in Lactococcus lactis and L. reuteri , . However, we found the deletion of pbp2A improves the GI-tract survival, which suggests that disruptions in peptidoglycan biosynthesis genes could either improve or decrease the survival of probiotics, reiterating the general concept of subtle inter-strain and species differences in survival mechanisms.
In conclusion, this study demonstrated that fermentation conditions have a large influence on the GI-tract survival of L. plantarum. We showed that transcriptome-trait matching enables the identification of genetic loci involved in gastrointestinal robustness and this approach can also be employed to rationally design fermentation and process conditions that aim for the production of probiotics with improved GI survival and consequently have a higher potential to achieve their desired health-beneficial effects on the consumer.
Materials and Methods
GI-tract Assay and Correlation to Transcriptome Data
Cells were harvested from the fermentation-genomics platform at OD600 = 1.0 for full-genome transcriptome profiling (see accompanying paper by Bron et al.), while the GI-tract survival was determined in the same cells, as well as in cells that were harvested 25 h after inoculation. One set of fermentations (F19–F24, see accompanying paper by Bron et al.) was excluded from the data analysis as the GI-tract survival data appeared unreliable, likely caused by minor deviations in the pH of the batch of GJ applied which is known to heavily influence GI survival. For GI-tract survival analysis, cultures were washed with prewarmed (37°C) PBS and resuspended in prewarmed (37°C) filter sterilized gastric juice [53 mM NaCl, 15 mM KCl, 5 mM Na2CO3, 1 mM CaCl2, 0.1 mg ml−1 lipase (Fluka; derived from Aspergillus niger), and 1.2 mg ml−1 pepsin (Sigma; derived from porcine) that had a pH adjusted to 2.4 with HCl (logarithmic cells) or 2.3 (stationary cells)]. The gastric juice enzymes were added immediately prior to the treatment. After 60 min incubation while rotating at 10 rpm in a Hybridization oven/shaker (Amersham pharmacia biotech, Little Chalfont, UK) at 37°C, the cultures were neutralized to pH 6.5 with 0.5 M NaHCO3, and prewarmed (37°C) pancreatic juice [85 mM NaCl, 5 mM KH2PO4, 2 mM Na2HPO4, 10 mM NaHCO3, 30 mg ml−1 pancreatin (Sigma; derived from porcine stomach) and bile acid mixture (latter two components were added fresh to pancreatic juice immediately prior to the treatment)] was added, followed by continued incubation for another 60 min (agitation at 10 rpm, 37°C). The bile acid mixture consisted of 3.0 mM (final concentration in assay) sodium glycocholate hydrate, 1.3 mM sodium glycodeoxycholate, 2.4 mM sodium glycochenodeoxycholate, 1.0 mM taurocholic acid sodium salt hydrate, 0.4 mM sodium taurodeoxycholate hydrate and 1.0 mM sodium taurochenodeoxycholate to mimic human bile components and concentrations . Preceding and during GI-tract assay incubation (t = 0, 20, 40, 60, 90, and 120), samples were taken for colony forming unit (CFU) enumeration by spot-plating . A reduction of 8 logs could be detected with this method. Relative GI-tract survival of the different cultures was expressed as the fraction of the corresponding input numbers of viable cells (t = 0 was set at 1.00). The transcriptome and GI-tract survival data are available in FermDB (www.cmbi.ru.nl/fermdb).
The initial list of genes predicted by the random forest algorithm  (see the accompanying paper by Bron et al. for details on correlation analyses) to be associated with GI-tract survival was further refined by application of several selection criteria that are based on transcript ranking. Firstly, only transcripts with an importance factor higher than 1 according to the random forest algorithm were selected for further analysis. Secondly, the quantitative correlation of individual transcripts with the survival rate observed in individual cultures was evaluated, selecting those transcripts (genes) that had the highest quantitative correlation with survival (expressed in R2 in table S1, see figure 5 for two examples). Lastly, genes encoding prophage associated functions that are typically hypervariable among L. plantarum strains were discarded , . The remaining transcripts and their associated genes (table 1) were considered to have the strongest correlation with the measured gastric juice tolerance and were therefore selected for validation by gene deletion or overexpression.
Deletion Mutant Construction
Gene deletion mutants were constructed using the mutagenesis vector pNZ5319 according to Lambert et al. . The L. plantarum WCFS1 pbp2A, lp_1669, lp_1818, pacL3, and napA3 genes were replaced with a lox66-P32-cat-lox71 cassette resulting in strains NZ3412CM (Δpbp2A::cat), NZ3417CM (Δlp_1669::cat), NZ3414CM (Δlp_1817::cat), NZ3415CM (ΔpacL::cat), and NZ3416CM (ΔnapA3::cat), respectively. Primer sequences used to construct the gene-targeted knock-out vectors for L. plantarum WCFS1 are provided in table S4. In short, upstream and downstream flanking regions (left flank, LF; right flank, RF, respectively) of the target genes (i.e., pbp2A, lp_1669, lp_1817, pacL3, and napA3) were amplified with primer pair combinations as listed in table S5. Primers at the 3′-end of the upstream and 5′-end of the downstream flanking regions (A3, A4, B3, B4, C3, C4, D3, D4, E3, and E4 ) were extended with an overlap-sequence complementary to the 5′ and 3′ end of the lox66-P32-cat-lox71 cassette (amplified with primers I and J ), to enable knock-out construction by a Splicing by overlap extension (SOE) PCR  with primer pairs as listed in table S5. The obtained (SOE-ing) amplicons were blunt-ligated into Ecl136II-SwaI digested pNZ5319  resulting in plasmids pNZ3412, pNZ3417, pNZ3414, pNZ3415, and pNZ3416 (see table 3). Escherichia coli was used as an intermediate cloning host and after introduction of the mutagenesis plasmids into competent L. plantarum WCFS1, cells were plated on MRS containing 10 µg ml−1 chloramphenicol. After 48 h, grown colonies were plated on MRS with and without 30 µg ml−1 erythromycin. Colonies from each mutant displaying the anticipated erythromycin sensitive phenotype were selected for colony-PCR using primer pairs as listed in table S5. Mutant colonies with the expected genetic organization were selected for each of the knock-out target loci; NZ3412CM (Δpbp2A::cat), NZ3417CM (Δlp_1669::cat), NZ3414CM (Δlp_1817::cat), NZ3415CM (ΔpacL::cat) and NZ3416CM (ΔnapA3::cat). The L. plantarum WCFS1 pbp2A plus napA3 and napA3 plus lp_1669 double-mutants were constructed in the NZ3412CM (Δpbp2A::cat) and NZ3416CM (ΔnapA3::cat) background, respectively, in a two-step procedure. Firstly, strains NZ3412 (Δpbp2A) and NZ3416 (ΔnapA3) were constructed by excision of the lox66-P32-cat-lox71 cassette by transient expression of the Cre resolvase enzyme from pNZ5348 according to methods described by Lambert et al. . In these deletion mutant strains, pNZ3416 and pNZ3417 were introduced and double mutant strains were selected using the approach described above, resulting in the isolation of strains NZ3419CM (Δpbp2A-ΔnapA3::cat) and NZ3418CM (ΔnapA3-Δlp_1669::cat), respectively (table 3).
Overexpression Mutant Construction and SDS-PAGE Analysis
Gene overexpression mutants were constructed using the expression vector pSIP411 . For the candidate genes selected for overexpression that were part of a predicted operon , the whole operon was cloned in the sakacin induction vector (table 1). Primers were designed (table S5) to introduce a restriction enzyme site for cloning the target gene(s) into the expression vector pSIP411 at the NcoI site. The lp_1357 and thrC+lp_2759 overexpression mutants were designed with BspHI site, which has compatible ends with NcoI site. The target gene(s) were amplified by PCR using corresponding primers for each mutant (F1/F2, G1/G2, H1/H2 and I1/I2 for lp_1357, lp_2349, thrC+lp_2759, and lp_0148∼0150 mutants, respectively). The reactions were carried out with KOD polymerase (Novagen, Darmstadt, Germany) according to the instructions of the manufacturer. The purified PCR products were digested by restriction enzymes (Invitrogen, Molecular probes, Inc, USA) for which sites were introduced in the primers (see table S4) and cloned in NcoI-SmaI digested pSIP411. Ligation mixtures were transformed to E. coli, and re-isolated from primary transformants. Correctly assembled overexpression plasmids were identified by PCR, restriction and sequence analysis. Re-isolated plasmids were propagated into L. plantarum WCFS1 and transformants were selected on MRS containing 30 µg·ml−1 erythromycin (table 3).
For protein analysis of the overexpression mutants, the induction and sample preparation procedures were modified from the description by Sørvig et al. . The 19-amino-acid inducing peptide (of Met-Ala-Gly-Asn-Ser-Ser-Asn-Phe-Ile-His-Lys-Ile-Lys-Gln-Ile-Phe-Thr-His-Arg ) was custom-synthesized by BACHEM (Budendorf, Switzerland). The inducing peptide was dissolved in degassed water, as recommended by BACHEM to avoid oxidation of the peptides. The overnight cultures of the overexpression strains were diluted 50-fold and then incubated at 37°C. After OD600 had reached 0.3, the inducing peptide was added to the cultures at varying concentrations of 0, 0.1, 1, 10, and 50 ng/ml. Incubation was continued at 37°C for another 4 h until the OD600 had reached approximately 1.8. Bacterial cells were collected by centrifugation at 5,200×g for 10 min, followed by resuspension of the cell pellet in 50 mM Sodium-phosphate buffer pH 7. The cells were disrupted with 1 g zirconium beads by using a FastPrep™ (Qbiogene Inc, Cedex, France). After the disruption, the samples were centrifuged 5 min at 20,800×g to obtain cell-free extracts for analysis by SDS-PAGE.
DNA Microarray Analysis and Data Visualization
DNA microarray analysis were performed to compare global transcriptome profiles of NZ3417CM (Δlp_1669::cat) and the wild-type. RNA isolation from L. plantarum, subsequent cDNA synthesis and indirect labeling, as well as DNA microarray hybridizations were performed as described in the accompanying paper by Bron et al. The hybridization scheme is presented in figure S3. Genes of the Lp_1669 regulon with FDR-adjusted p-values less than 0.05 together with a fold-change higher than 2.0 or lower than 0.5 were considered to be significantly differently expressed. All microarray data is MIAME compliant and is available in the GEO database under accession number GSE31254. The biomolecular interaction network of the Lp_1669 regulon in 2× CDM and MRS was visualised using the Cytoscape software (version 2.8.1) , and the Biological Networks Gene Ontology (BiNGO) tool  was employed to detect significantly overrepresented categories in the regulon of Lp_1669. See the accompanying paper by Bron et al. for details.
Phenotypic Assays of Mutant Strains
Gene deletion mutants were analyzed for their gastrointestinal survival characteristics in a procedure identical to that described for the wild-type (see above). To evaluate the relative GI-tract survival of the overexpression mutants, the mutant strain SIP411B (empty vector) and the overexpression mutants were sakacin-induced (50 ng/ml) (see above). Additionally, to measure the relative GI-tract survival of the folate overexpression strain, strains NZ7021 (empty vector) and NZ7026 (folate overproducing strain)  were inoculated at OD600 = 0.1 in MRS containing 80 mg/ml chloramphenicol and 0 or 10 mg/ml p-aminobenzoic acid (pABA) according to Wegkamp et al. , grown at 37°C until OD600 was 1.0, and subjected to the GI-tract survival assay. To evaluate relative growth efficiency of the deletion mutants, the parental strain (WCFS1) and mutant strains NZ3412CM (Δpbp2A::cat), NZ3417CM (Δlp_1669::cat), and NZ3416CM (ΔnapA3::cat) were inoculated at OD600 = 0.1 in 96-wells plates and incubated in MRS broth at 28°C. OD600 of the cultures was monitored spectophotometrically (GENios, Tecan Austria GmbH, Grödig, Austria).
Capsular Polysaccharide Isolation and Determination
Capsular polysaccharide (CPS) was purified and chain lengths and sugar composition were determined essentially as described before . Briefly, 500 ml cultures of L. plantarum WCFS1 and NZ3417CM (Δlp_1669::cat) were grown in 2× CDM at 37°C until stationary phase (25 h). After 1 h incubation at 55°C, the cells were separated from the CPS containing growth medium by centrifugation for 15 min (6000×g) and to prevent overgrowth during dialysis, erythromicine was added to the supernatant to a final concentration of 10 µg/ml. A dialyzing tube 12–1400 Da (Fisher Scientific) was prepared by boiling twice 2% NaHCO3/2 mM EDTA, and once in reverse osmosis water. After overnight dialysis against running tap water followed by 4 h dialysis using reverse osmosis water, the samples were freeze-dried and stored at −20°C until further analysis.
The samples were dissolved in eluent (100 mM NaNO3+0.02% NaN3), filtered over 0.2 µm, and placed in a thermally controlled sample holder at 10°C and 200 µl was injected (model 231 Bio, Gilson) on the columns connected in series and remained at 35°C with a temperature control module (Waters, Milford, USA) to perform size exclusion chromatography (SEC) [TSK gel PWXL guard column, 6.0 mm×4.0 cm, TSK gel G6000 PWXL analytical column, 7.8 mm×30 cm, 13.0 µm and TSK gel G5000 PWXL analytical column, 7.8 mm×30 cm, 10 µm (TosoHaas, King of Prussio, USA)]. Light scattering was measured at 632.8 nm at 15 angles between 32° and 144° (DAWN DSP-F, Wyatt Technologies, Santa Barbara, USA). UV absorption was measured at 280 nm (CD-1595, Jasco, de Meern, The Netherlands) to detect proteins. The specific viscosity was measured with a viscosity detector (ViscoStar, Wyatt Technologies, Santa Barbara, USA) at 35°C and sample concentration was measured by refractive index detection, held at a fixed temperature of 35°C (ERC-7510, Erma Optical Works, Tokyo, Japan).
During the analysis with SEC the polysaccharide peak was collected (2 min×0.5 mL/min = 1 mL). The acid hydrolyses of the collected polysaccharide was carried out for 75 min at 120°C with 2 M trifluoro acetic acid under nitrogen. Following hydrolyses, the solutions were dried overnight under vacuum and dissolved in water. High Performance Anion Exchange Chromatography with Pulsed Amperometric Detection (HPAEC-PAD) on a gold electrode was used for the quantitative analyses of the monosaccharides rhamnose, galactosamine, arabinose, glucosamine, galactose, glucose, mannose, xylose, galacturonic acid, and glucuronic acid. The analyses were performed with a 600E System controller pump (Waters, Milford, USA) with a helium degassing unit and a model 400 EC detector (EG&G, Albuquerque, USA). With a 717 autosampler (Waters, Milford, USA), 20 µl of the sample was injected on a Dionex Carbopac PA-1, 250×4 mm (10–32), column thermostated at 30°C. The monosaccharides were eluted at a flow rate of 1.0 mL/min. The monosaccharides were eluted isocratic with 16 mM sodium hydroxide, followed by the elution of the acid monosaccharides starting at 20 min with a linear gradient to 200 mM sodium hydroxide +500 mM sodium acetate in 20 minutes. Data analysis was done with Dionex Chromeleon software version 6.80. Quantitative analyses were carried out using standard solutions of the monosaccharides (Sigma-Aldrich, St. Louis, USA).
SDS-PAGE of cell-free extracts logarithmic L. plantarum strains overexpressing hicD3 (lp_2349) and overexpressing thrC (lp_2758) and lp_2759. The arrows indicate protein bands increasing with increasing amounts of Sakacin P (inducing peptide, IP). Empty vector = pSIP411B. L. plantarum harboring pNZ3431 (over-hicD3), and pNZ3432 (over-thrC+lp_2759). Marker sizes are indicated in kDalton (kDa).
Cytoscape version of figure 8; regulatory network of Lp_1669 grown in 2× CDM or MRS. Yellow rectangle nodes represent growth in 2× CDM (left node) or in MRS (right node). Round nodes represent single genes with their corresponding lp_number, which is the number as annotated in the sequenced parental L. plantarum WCFS1 strain , green lines represent down-regulation, and red lines represent up-regulation of the gene compared to the parental strain. The colors of the round nodes represent the annotated main class.
Lp_1669 regulon hybridization scheme. Tail and head of the arrow represent Cy3 and Cy5 labeling, respectively.
Candidate genes associated with GI-tract survival of L. plantarum WCFS1.
Differentially regulated genes in NZ3417CM (Δ lp_1669::cat ) grown in 2× CDM.
Differentially regulated genes in NZ3417CM (Δ lp_1669::cat ) grown in MRS.
Primer pair combinations used for LF and RF amplification and for the SOE step of the deletion mutants.
We thank Roger Bongers, Anne Wiersma, and Guido Staring (NIZO food research, Ede, The Netherlands) for excellent technical assistance, Lex Overmars (Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands) for extending the FermDB database to allow inclusion of our gastrointestinal survival data, and Yu Zhao (Wageningen University and Research Centre, Wageningen, The Netherlands) for performing the phenotypic assays. We acknowledge the managerial activities executed by Dr. Paul de Vos (University Medical Centre Groningen, The Netherlands), who was the project leader of this project.
Conceived and designed the experiments: HvB-vdV MW MLM PAB MK. Performed the experiments: HvB-vdV IL. Analyzed the data: HvB-vdV IL MW PAB MK. Wrote the paper: HvB-vdV IL MW MLM PAB MK.
- 1. FAO/WHO (2001) Evaluation of health and nutritional properties of powder milk with live lactic acid bacteria. Report of FAO/WHO expert consultation 1–4 October.
- 2. Marco ML, Pavan S, Kleerebezem M (2006) Towards understanding molecular modes of probiotic action. Curr Opin Biotechnol 17: 204–210.
- 3. Lebeer S, Vanderleyden J, De Keersmaecker SC (2010) Host interactions of probiotic bacterial surface molecules: comparison with commensals and pathogens. Nat Rev Microbiol 8: 171–184.
- 4. Bron PA, van Baarlen P, Kleerebezem M (2012) Emerging molecular insights into the interaction between probiotics and the host intestinal mucosa. Nat Rev Microbiol 10: 66–78.
- 5. Ma D, Forsythe P, Bienenstock J (2004) Live Lactobacillus reuteri is essential for the inhibitory effect on tumor necrosis factor alpha-induced interleukin-8 expression. Infect Immun 72: 5308–5314.
- 6. Gobbetti M, Cagno RD, De Angelis M (2010) Functional microorganisms for functional food quality. Crit Rev Food Sci Nutr 50: 716–727.
- 7. Corcoran BM, Stanton C, Fitzgerald G, Ross RP (2008) Life under stress: The probiotic stress response and how it may be manipulated. Current Pharmaceutical Design 14: 1382–1399.
- 8. van Bokhorst-van de Veen H, Bron PA, Wels M, Kleerebezem M (2011) Engineering robust lactic acid bacteria. In: Tsakalidou E, Papadimitriou K, editors. pp. 369–394. Stress Responses of Lactic Acid Bacteria: Springer US.
- 9. van de Guchte M, Serror P, Chervaux C, Smokvina T, Ehrlich SD, et al. (2002) Stress responses in lactic acid bacteria. Antonie Van Leeuwenhoek 82: 187–216.
- 10. Watson D, Sleator RD, Hill C, Gahan CG (2008) Enhancing bile tolerance improves survival and persistence of Bifidobacterium and Lactococcus in the murine gastrointestinal tract. BMC Microbiol 8: 176.
- 11. Begley M, Gahan CG, Hill C (2005) The interaction between bacteria and bile. FEMS Microbiol Rev 29: 625–651.
- 12. Weckx S, Van der Meulen R, Allemeersch J, Huys G, Vandamme P, et al. (2010) Community dynamics of bacteria in sourdough fermentations as revealed by their metatranscriptome. Appl Environ Microbiol 76: 5402–5408.
- 13. Siezen RJ, Tzeneva VA, Castioni A, Wels M, Phan HT, et al. (2010) Phenotypic and genomic diversity of Lactobacillus plantarum strains isolated from various environmental niches. Environ Microbiol 12: 758–773.
- 14. Ahrne S, Nobaek S, Jeppsson B, Adlerberth I, Wold AE, et al. (1998) The normal Lactobacillus flora of healthy human rectal and oral mucosa. J Appl Microbiol 85: 88–94.
- 15. Vesa T, Pochart P, Marteau P (2000) Pharmacokinetics of Lactobacillus plantarum NCIMB 8826, Lactobacillus fermentum KLD, and Lactococcus lactis MG 1363 in the human gastrointestinal tract. Aliment Pharmacol Ther 14: 823–828.
- 16. Kleerebezem M, Boekhorst J, van Kranenburg R, Molenaar D, Kuipers OP, et al. (2003) Complete genome sequence of Lactobacillus plantarum WCFS1. Proc Natl Acad Sci U S A 100: 1990–1995.
- 17. Teusink B, Wiersma A, Molenaar D, Francke C, de Vos WM, et al. (2006) Analysis of growth of Lactobacillus plantarum WCFS1 on a complex medium using a genome-scale metabolic model. J Biol Chem 281: 40041–40048.
- 18. Teusink B, van Enckevort FH, Francke C, Wiersma A, Wegkamp A, et al. (2005) In silico reconstruction of the metabolic pathways of Lactobacillus plantarum: comparing predictions of nutrient requirements with those from growth experiments. Appl Environ Microbiol 71: 7253–7262.
- 19. Lambert JM, Bongers RS, Kleerebezem M (2007) Cre-lox-based system for multiple gene deletions and selectable-marker removal in Lactobacillus plantarum. Appl Environ Microbiol 73: 1126–1135.
- 20. Molenaar D, Bringel F, Schuren FH, de Vos WM, Siezen RJ, et al. (2005) Exploring Lactobacillus plantarum genome diversity by using microarrays. J Bacteriol 187: 6119–6127.
- 21. Pretzer G, Snel J, Molenaar D, Wiersma A, Bron PA, et al. (2005) Biodiversity-based identification and functional characterization of the mannose-specific adhesin of Lactobacillus plantarum. J Bacteriol 187: 6128–6136.
- 22. Meijerink M, van Hemert S, Taverne N, Wels M, de Vos P, et al. (2010) Identification of genetic loci in Lactobacillus plantarum that modulate the immune response of dendritic cells using comparative genome hybridization. PLoS ONE 5: e10632.
- 23. van Hemert S, Meijerink M, Molenaar D, Bron PA, de Vos P, et al. (2010) Identification of Lactobacillus plantarum genes modulating the cytokine response of human peripheral blood mononuclear cells. BMC Microbiol 10: 293.
- 24. Wegkamp A, de Vos WM, Smid EJ (2009) Folate overproduction in Lactobacillus plantarum WCFS1 causes methotrexate resistance. FEMS Microbiol Lett 297: 261–265.
- 25. Wegkamp A (2008) Modulation of folate production in lactic acid bacteria. Wageningen: Wageningen University, The Netherlands.
- 26. Corcoran BM, Ross RP, Fitzgerald GF, Stanton C (2004) Comparative survival of probiotic lactobacilli spray-dried in the presence of prebiotic substances. J Appl Microbiol 96: 1024–1039.
- 27. Marteau P, Minekus M, Havenaar R, Veld J (1997) Survival of lactic acid bacteria in a dynamic model of the stomach and small intestine: Validation and the effects of bile. Journal of Dairy Science 80: 1031–1037.
- 28. Jacobsen CN, Rosenfeldt Nielsen V, Hayford AE, Moller PL, Michaelsen KF, et al. (1999) Screening of probiotic activities of forty-seven strains of Lactobacillus spp. by in vitro techniques and evaluation of the colonization ability of five selected strains in humans. Appl Environ Microbiol 65: 4949–4956.
- 29. Matto J, Malinen E, Suihko ML, Alander M, Palva A, et al. (2004) Genetic heterogeneity and functional properties of intestinal bifidobacteria. J Appl Microbiol 97: 459–470.
- 30. Succi M, Tremonte P, Reale A, Sorrentino E, Grazia L, et al. (2005) Bile salt and acid tolerance of Lactobacillus rhamnosus strains isolated from Parmigiano Reggiano cheese. FEMS Microbiol Lett 244: 129–137.
- 31. Georgieva RN, Iliev IN, Chipeva VA, Dimitonova SP, Samelis J, et al. (2008) Identification and in vitro characterisation of Lactobacillus plantarum strains from artisanal Bulgarian white brined cheeses. J Basic Microbiol 48: 234–244.
- 32. Prasad J, McJarrow P, Gopal P (2003) Heat and osmotic stress responses of probiotic Lactobacillus rhamnosus HN001 (DR20) in relation to viability after drying. Appl Environ Microbiol 69: 917–925.
- 33. Kim WS, Perl L, Park JH, Tandianus JE, Dunn NW (2001) Assessment of stress response of the probiotic Lactobacillus acidophilus. Curr Microbiol 43: 346–350.
- 34. Jordan S, Hutchings MI, Mascher T (2008) Cell envelope stress response in Gram-positive bacteria. FEMS Microbiol Rev 32: 107–146.
- 35. Bron PA, Marco M, Hoffer SM, Van Mullekom E, de Vos WM, et al. (2004) Genetic characterization of the bile salt response in Lactobacillus plantarum and analysis of responsive promoters in vitro and in situ in the gastrointestinal tract. J Bacteriol 186: 7829–7835.
- 36. Ingham CJ, Beerthuyzen M, van Hylckama Vlieg J (2008) Population heterogeneity of Lactobacillus plantarum WCFS1 microcolonies in response to and recovery from acid stress. Appl Environ Microbiol 74: 7750–7758.
- 37. Gallegos MT, Schleif R, Bairoch A, Hofmann K, Ramos JL (1997) Arac/XylS family of transcriptional regulators. Microbiol Mol Biol Rev 61: 393–410.
- 38. Lebeer S, Claes IJ, Verhoeven TL, Vanderleyden J, De Keersmaecker SC (2011) Exopolysaccharides of Lactobacillus rhamnosus GG form a protective shield against innate immune factors in the intestine. Microb Biotechnol 4: 368–374.
- 39. Pfeiler EA, Azcarate-Peril MA, Klaenhammer TR, Prouty AM, Brodsky IE, et al. (2007) Characterization of a novel bile-inducible operon encoding a two-component regulatory system in Lactobacillus acidophilus. J Bacteriol 189: 4624–4634.
- 40. Wall T, Bath K, Britton RA, Jonsson H, Versalovic J, et al. (2007) The early response to acid shock in Lactobacillus reuteri involves the ClpL chaperone and a putative cell wall-altering esterase. Appl Environ Microbiol 73: 3924–3935.
- 41. Whitehead K, Versalovic J, Roos S, Britton RA (2008) Genomic and genetic characterization of the bile stress response of probiotic Lactobacillus reuteri ATCC 55730. Appl Environ Microbiol 74: 1812–1819.
- 42. Bron PA, Molenaar D, de Vos WM, Kleerebezem M (2006) DNA micro-array-based identification of bile-responsive genes in Lactobacillus plantarum. J Appl Microbiol 100: 728–738.
- 43. Govers MJ, Termont DS, Van Aken GA, Van der Meer R (1994) Characterization of the adsorption of conjugated and unconjugated bile acids to insoluble, amorphous calcium phosphate. J Lipid Res 35: 741–748.
- 44. Sieuwerts S, de Bok FAM, Mols E, de Vos WM, Vlieg J (2008) A simple and fast method for determining colony forming units. Letters in Applied Microbiology 47: 275–278.
- 45. Breiman L (2001) Random forests. Machine Learning 45: 5–32.
- 46. van Bokhorst-van de Veen H, Abee T, Tempelaars M, Bron PA, Kleerebezem M, et al. (2011) Short- and long-term adaptation to ethanol stress and its cross-protective consequences in Lactobacillus plantarum. Appl Environ Microbiol 77: 5247–5256.
- 47. Horton RM, Cai ZL, Ho SN, Pease LR (1990) Gene splicing by overlap extension: tailor-made genes using the polymerase chain reaction. Biotechniques 8: 528–535.
- 48. Sorvig E, Mathiesen G, Naterstad K, Eijsink VG, Axelsson L (2005) High-level, inducible gene expression in Lactobacillus sakei and Lactobacillus plantarum using versatile expression vectors. Microbiology 151: 2439–2449.
- 49. Wels M, Francke C, Kerkhoven R, Kleerebezem M, Siezen RJ (2006) Predicting cis-acting elements of Lactobacillus plantarum by comparative genomics with different taxonomic subgroups. Nucleic Acids Res 34: 1947–1958.
- 50. Eijsink VG, Brurberg MB, Middelhoven PH, Nes IF (1996) Induction of bacteriocin production in Lactobacillus sake by a secreted peptide. J Bacteriol 178: 2232–2237.
- 51. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, et al. (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13: 2498–2504.
- 52. Maere S, Heymans K, Kuiper M (2005) BiNGO: a Cytoscape plugin to assess overrepresentation of gene ontology categories in biological networks. Bioinformatics 21: 3448–3449.
- 53. Looijesteijn PJ, Hugenholtz J (1999) Uncoupling of growth and exopolysaccharide production by Lactococcus lactis subsp. cremoris NIZO B40 and optimization of its synthesis. J Biosci Bioeng 88: 178–182.
- 54. Zhou M, Boekhorst J, Francke C, Siezen RJ (2008) LocateP: genome-scale subcellular-location predictor for bacterial proteins. BMC Bioinformatics 9: 173.
- 55. Casadaban MJ, Cohen SN (1980) Analysis of gene control signals by DNA fusion and cloning in Escherichia coli. J Mol Biol 138: 179–207.