Molecular Switch Controlling Expression of the Mannose-Specific Adhesin, Msa, in Lactobacillus plantarum

Probiotic strains possess adhesive properties enabling colonization of the human intestinal tract through interactions between molecules present on the probiotic bacteria and components of the epithelial surface. In Lactobacillus plantarum, interaction is mediated through bacterial surface proteins like Msa, which binds to mannose residues present on the intestinal cells. Such interactions are believed to be important for the health-promoting effects of probiotics, including displacement of pathogens, immunomodulation, and protective effects on the intestinal barrier function. In this study, we have identified a new molecular switch controlling expression of the msa gene in L. plantarum strain WCFS1. Strains with increased msa expression could be valuable in the development and manufacture of improved probiotic products.

ABSTRACT

Some lactic acid bacteria, especially Lactobacillus spp., possess adhesive properties enabling colonization of the human gastrointestinal tract. Two probiotic Lactobacillus plantarum strains, WCSF1 and 299v, display highly different mannose-specific adhesion, with L. plantarum 299v being superior to L. plantarum WCFS1 based on a yeast agglutination assay. A straightforward correlation between the mannose adhesion capacity and domain composition of the mannose-specific adhesin (Msa) in the two strains has not been demonstrated previously. In this study, we analyzed the promoter regions upstream of the msa gene encoding a mannose-specific adhesin in these two strains. The promoter region was mapped by primer extension and DNA sequence analysis, and only a single nucleotide change was identified between the two strains. However, Northern blot analysis showed a stronger msa transcript band in 299v than in WCFS1 correlating with the different adhesion capacities. During the establishment of a high-throughput yeast agglutination assay, we isolated variants of WCFS1 that displayed a very strong mannose-specific adhesion phenotype. The region upstream of the msa gene in these variants showed an inversion of a 104-bp fragment located between two perfectly inverted repeats present in the untranslated leader region. The inversion disrupts a strong hairpin structure that otherwise most likely would terminate the msa transcript. In addition, the ribosome binding site upstream of the msa gene, which is also masked within this hairpin structure, becomes accessible upon inversion, thereby increasing the frequency of translation initiation in the variant strains. Furthermore, Northern blot analysis showed a higher abundance of the msa transcript in the variants than in the wild type, correlating with a strong-Msa phenotype.

IMPORTANCE Probiotic strains possess adhesive properties enabling colonization of the human intestinal tract through interactions between molecules present on the probiotic bacteria and components of the epithelial surface. In Lactobacillus plantarum, interaction is mediated through bacterial surface proteins like Msa, which binds to mannose residues present on the intestinal cells. Such interactions are believed to be important for the health-promoting effects of probiotics, including displacement of pathogens, immunomodulation, and protective effects on the intestinal barrier function. In this study, we have identified a new molecular switch controlling expression of the msa gene in L. plantarum strain WCFS1. Strains with increased msa expression could be valuable in the development and manufacture of improved probiotic products.

INTRODUCTION

Lactobacilli are widely distributed in nature and in the gastrointestinal tract (GIT) of both humans and animals. Some strains of Lactobacillus are considered beneficial for humans and are attributed a major role in the positive health effects of probiotics (1–4). The health claims associated with intake of lactobacilli include maintenance of a balanced immune response in the mucosa-associated lymphatic tissue (MALT) (5), support of the GIT endothelial barrier by upregulating host production of mucin (6), inhibition of pathogenic bacteria by competing for space and nutrients (7), and production of antimicrobial compounds such as bacteriocins (8). The molecular mechanisms behind these effects of lactobacilli are not elucidated in detail, but the efficacy of probiotics has been connected with their adherence to the mucosal surface of the GIT. The molecular factors responsible for adhesion and colonization of these organisms have therefore been investigated using tissue samples containing mucus, human cell lines, and components of the extracellular matrix. The receptors of the adhesins have been identified as sugar components (9), mucin (10, 11), and extracellular matrix (ECM) proteins, such as fibronectin (12) and collagen (13). The adhesins of lactobacilli constitute a chemically diverse group including both protein and nonproteinaceous components located in the cell wall or attached to the bacterial cell surface. In many cases, the mucus layer protecting the human epithelial cell layer is the target of these bacterial surface proteins. A mucus adhesion-promoting protein, MapA, targeting the mucus layer was identified in Lactobacillus reuteri and was also found to promote direct binding to Caco-2 cells in vitro (11). Other mucus-binding proteins have been identified in L. reuteri (14) and in Lactobacillus acidophilus (15). A mannose-specific adherence mechanism in L. plantarum strain 299v confers binding to the human colonic cell HT-29 in vitro (9). This capacity of binding to HT-29 cells correlates with the ability to agglutinate Saccharomyces cerevisiae cells allowing an easy and semiquantitative screening method for mannose adhesion capacity among probiotic bacteria. The lp_1229 gene responsible for this adhesion in L. plantarum strains WCFS1 and 299v was later identified (16, 17) and found to encode a large cell wall-anchored protein (mannose-specific adhesin [Msa]) with a number of features, such as a concanavalin A (ConA)-like domain and different mucus-binding domains. The Msa protein belongs to a subgroup of surface proteins anchored to the peptidoglycan layer through the LPQTNE motif located in the C terminus. This motif is recognized by the sortase enzyme gene (srtA), leading to cleavage and subsequent anchoring of the surface protein to the peptidoglycan layer.

The changing habitat that lactobacilli experience going from food products to the GI tract may require the expression of essential genes that support colonization or survival in the new environment. Selection systems that capture promoters from GI tract-induced genes have been developed (18). Application of the resolvase-based in vivo expression technology (R-IVET) in L. plantarum WCFS1 revealed 72 genes that were induced during passage of the GI tract of mice. As expected, genes involved in stress and metabolism were identified; also, four genes encoding cell surface proteins were induced, illustrating that L. plantarum changes its surface architecture during passage in the GI tract of mice. Lactobacilli also regulate the surface protein architecture through other mechanisms than by induction of promoter activities. The S-protein is a major surface protein found in some lactobacilli, which crystallizes to form a two-dimensional layer on the bacterial cell surface. L. acidophilus carries two S-protein-encoding genes, slpA and slpB, in opposite orientations on a 6-kb chromosomal segment. The majority of cells in a bacterial culture only express the slpA gene, but due to an inversion of the 6-kb segment, the two slp genes are interchanged, leading to positioning of the slpB gene behind the S promoter. This inversion is mediated through recombination via a 26-bp sequence showing similarity to the recognition site for the Din family invertases (19). Genome rearrangements and inversions as a mechanism to control the expression of especially surface proteins have been described in other bacteria as well. Escherichia coli, for instance, shows phase variation in its expression profile of type 1 fimbriae (20), which are adhesins that mediate adhesion of the bacterial cell to mannose receptors present on epithelial cells in the host. The phase-variable expression of type 1 fimbriae is regulated by a mechanism that involves site-specific DNA inversion of the fragment containing the major promoter, fimA, responsible for expression of the entire fim operon. When the fimA promoter is in the on orientation, the fimA gene is transcribed, resulting in bacteria with fimbriae, whereas inversion of the fimA promoter changes the bacteria to a nonfimbrial phenotype (off orientation). This inversion is controlled by global regulatory proteins and the fimB and fimE gene products, which are members of the integrase family of site-specific recombinases (21, 22).

In this study, we isolated variants of L. plantarum strain WCSF1 that show a high mannose adhesion capacity similar to that found in strain 299v. The variants occur by inversion of a DNA fragment located between the promoter region and the msa coding sequence. The increased mannose adhesion capacity in WCSF1 can be explained by the formation of alternative secondary structures in the mRNA, leading to increased transcription and translation of the msa gene. This type of molecular switch is intriguing. A better understanding of the components involved could lead to novel genetic control systems suited for the design of bacteria displaying phase-variable expression of gene products.

RESULTS

Identification of a putative promoter controlling msa gene expression in L. plantarum strain 299v.

The region upstream of the msa gene in L. plantarum 299v was determined by DNA sequencing and compared to the published DNA sequence present in the same region of L. plantarum WCSF1 (23, 24). The DNA sequencing showed an identical genetic organization in 299v and in WCSF1, both with an intergenic 306-bp region situated between the lp_1228 gene and the msa gene, only differing in one nucleotide position (C→T) (see Fig. S1 in the supplemental material). The upstream region was analyzed for motifs present in σ70-dependent promoters, but no consensus −10 (5ʹ-TATAAT-3ʹ) or −35 (5ʹ-TTGACA-3ʹ) motifs were identified. As no obvious σ70-dependent promoter elements could be identified by in silico analysis, primer extension analysis was used to map the transcriptional start site and to localize the putative promoter region. The transcriptional start site was mapped to an adenine (A) (Fig. S1) located 184 bp upstream of the msa start codon. A putative −10 region (TATGAT) was found 2 bp upstream of the transcriptional start site, and a partial consensus −35 region (TTGTAG) was found 17 bp further upstream. Another −10 motif (5ʹ-TACCAT-3ʹ) with a TG extension was present 6 bp immediately upstream of the first −10 motif. The corresponding −35 motif (5ʹ-TTTACT-3ʹ) was somewhat similar to the consensus −35 hexamer of the σ70 promoters. Notably, the single nucleotide difference observed between strains 299v and WCFS1 was located in this putative −35 region. The region upstream of the putative −35 hexamers included a typical UP element rich in AT nucleotides (25).

Overexpression of the mannose-specific adhesion gene was toxic to L. plantarum 299v.

An msa mutant 299v strain (299vΔMsa) was constructed by homologous recombination and contained a large deletion of approximately 1,500 bp covering the entire concanavalin A (ConA)-like region. The resulting msa knockout strain was completely unable to agglutinate yeast (Table 1). Interestingly, 299vΔMsa had also lost the clumping phenotype normally seen for Lactobacillus cultures. Complementation of strain 299vΔMsa with msa expressed on a low-copy-number plasmid (pAMJ1296) restored the agglutinating phenotype (Table 1). However, introduction of the msa gene on a high-copy-number plasmid (pAMJ1297) resulted in only a few viable transformants (data not shown), most likely due to high and lethal msa expression levels as a result of increased gene dosage. None of the few isolated transformants were able to agglutinate yeast cells, demonstrating a loss of this phenotype. Isolation of plasmid DNA from surviving transformants after the introduction of msa on a high-copy-number plasmid (pAMJ1297), followed by sequence analysis, showed that both the promoter region and the msa gene were missing in the plasmid (data not shown). Analysis of the DNA sequence on the original plasmid showed a short repeated DNA sequence on both sides of the insert, correlating well with deletion of the promoter and the msa gene by homologous recombination through these repeats (data not shown). Also, the introduction of msa on either a low-copy-number plasmid (pAMJ1296) or on a high-copy-number plasmid (pAMJ1297) into the wild-type 299v strain resulted in no surviving transformants. These results clearly showed that overexpression of the msa gene in L. plantarum 299v is toxic and affects cell viability negatively. As a consequence, mutants that have lost the msa gene and thereby their ability to agglutinate yeast cells have been selected.

TABLE 1.

Agglutination phenotypes of Lactobacillus strains in this study

We also tested msa complementation in a lactic acid bacterium that normally lacks the ability to agglutinate yeast cells. Introduction of the msa expression plasmids (pAMJ1296 and pAMJ1297) into L. lactis MG1363 resulted in a mannose-specific adhesion phenotype, whereas cells transformed with a control plasmid without the msa gene did not adhere to yeast cells (Table 1). These results showed that the msa gene product is responsible for the mannose-specific adhesion of L. plantarum 299v, and that this phenotype can be transferred to other bacteria by cloning of the msa gene. The msa knockout and the gene complementation strains indicate that msa alone encoded this phenotype and that the mechanisms for secretion and anchoring of Msa seem conserved among different lactic acid bacteria. There was no toxic effect in L. lactis MG1363 of overexpressing the msa gene from L. plantarum 299v.

Quantitative measurement of cell surface-located Msa.

Although yeast agglutination assays are useful to determine the amount of Msa on the bacterial surface, a more quantitative method for measuring Msa was established using an enzyme-linked immunosorbent assay (ELISA). To obtain antiserum specific against Msa, we expressed the ConA-like domain of the msa gene from strain 299v in L. lactis using a plasmid vector that secreted the product. The lectin function of the ConA-like domain was used for purification. The culture supernatant containing the expressed ConA domain was applied to a mannan column. This resulted in a pure fraction of ConA after elution with mannose (data not shown). The predicted ConA domain is therefore indeed responsible for the mannose affinity of this protein. The purified ConA domain was used to raise ConA-specific antibodies in a rabbit.

Quantitative measurement of cell surface Msa was done using whole bacteria in an ELISA-like reaction. Dilutions of overnight cultures adjusted for optical density at 600 nm (OD₆₀₀) were distributed in microtiter plates preblocked with PBS containing skim milk. After incubation with antiserum (anti-ConA), cells were washed, incubated with secondary antibodies, and quantified by a colorimetric reaction (see Materials and Methods). As expected, there was a clear correlation between the number of 299v cells and the amount of Msa, resulting in a maturation curve (Fig. 1). The Msa knockout strain (299vΔMsa) showed low interaction with ConA-specific antibodies, whereas complementation of this strain with Msa (299vΔMsa/pAMJ1296) restored the surface display of Msa. The weak-agglutinating phenotype of the WCFS1 strain correlated well with the ELISA that showed smaller amounts of cell surface-displayed Msa in this L. plantarum strain than that in 299v (Fig. 1). This ELISA-like assay thus seems to be suitable for assaying the surface display of the Msa protein on Lactobacillus strains.

FIG 1.

Quantification of surface-displayed Msa by ELISA on whole L. plantarum cells. The amount of Msa was analyzed in four different L. plantarum strains using an antiserum directed against the ConA domain of Msa. The absorbance at 595 nm is plotted against the cell density (OD₆₀₀). The mean results of three independent determinations are shown, with error bars representing the standard deviation.

Msa mediates adhesion to epithelial cells.

The role of Msa in bacterial adherence to epithelial cells was analyzed by coincubating bacteria and human Caco-2 cells in vitro. Caco-2 cells express many of the markers associated with normal small intestine villus cells and are therefore often used to study bacterial adherence. Indeed, strain 299v had significantly higher Caco-2 cell-binding capacity than did the 299vΔMsa strain (Fig. 2). The reduced adhesion to Caco-2 cells by deletion of the msa gene indicates that Msa plays a role in adhesion to epithelial cells.

FIG 2.

Quantitative evaluation of the adhesion properties of L. plantarum 299v (wild type) and mutant strain 299vΔMsa. Adhesion was determined by counting the number of wild-type and mutant bacteria adhering to Caco-2 cells after Giemsa staining in randomly selected fields of view (FoV) using a ×100 objective (n = 50). ****, significant difference in the adhesion capability of the two strains (P < 0.0001) using an unpaired t test.

Identification of a phase-variable event leading to enhanced yeast agglutination in L. plantarum WCFS1.

During the establishment of a high-throughput yeast agglutination assay using a microtiter plate format, we observed microcultures of L. plantarum WCFS1 that, at a low frequency, displayed a strong-agglutination phenotype visible without microscopy. The phenotype (termed Agg⁺) was similar to the agglutination phenotype observed for L. plantarum 299v but was only represented in approximately 2% of the 1,000 independently analyzed microcultures. To analyze if this Agg⁺ phenotype is locked and maintained during cell divisions, 1,000 individual colonies (obtained by plating an Agg⁺ culture of approximately 30 generations) were subsequently screened in microtiter plates using the high-throughput yeast agglutination assay. Now, only about 20% of the microcultures displayed the strong-agglutination phenotype (Agg⁺). The observed phenomenon could indicate that msa gene expression in L. plantarum WCFS1 is controlled by a mechanism resembling phase-variable expression seen in other bacterial systems. The agglutinating mutant and the corresponding wild-type strain were analyzed as described above. The mutant showed a very strong agglutination phenotype (Table 1). L. plantarum 299v was also analyzed for variation in the ability to display the mannose-binding phenotype. However, analysis of 1,000 independent microcultures displayed no heterogeneity with respect to yeast agglutination, showing that the variable msa expression is less frequent or absent in strain 299v compared to WCFS1 under the growth conditions used. Alternatively, the existing yeast agglutination assay is not able to discriminate between the “normal” level and an enhanced level of the agglutinating phenotype.

Variable msa expression in L. plantarum WCFS1 is controlled by DNA inversion.

To investigate if a DNA inversion mechanism is involved in the variable expression of the msa gene in L. plantarum WCFS1, the msa promoter region from a number of WCFS1 variants showing the Agg⁺ phenotype was sequenced and compared to the DNA sequence present in the wild-type L. plantarum strains WCSF1 and 299v. Sequence analysis of the intergenic region located between lp_1228 and msa of WCFS1 and WCFS1-Agg⁺ revealed an inversion of a 104-bp DNA fragment (Fig. S1). The inverted segment is flanked by two perfectly inverted repeats (IRs), left IR (IR_L) and right IR (IR_R), of 14 bp (Fig. 3).

FIG 3.

(A) Genetic organization of the intergenic region controlling expression of the msa gene. The open reading frames containing lp_1228, msa, and lp_1230 are shown by open rectangles. The inversion region is indicated by a double-headed arrow and the two inverted repeats by hatched boxes. The stem-loop structure is shown by inverted dotted arrows. The direction of transcription is indicated by solid arrows. The promoter (P) and the ribosome binding site (RBS) controlling msa translation are indicated. (B) Sequences of the two 14-bp perfectly inverted repeats flanking the 104-bp invertible region.

In a further attempt to establish a linkage between the Agg⁺ phenotype and the orientation of the 104-bp fragment, we performed a yeast agglutination assay followed by a genomic analysis of the invertible region. A dilution of an overnight culture of WCSF1 was spread on de Man-Rogosa-Sharpe (MRS) agar plates. After colony formation, we randomly picked 960 individual colonies into microtiter well plates using a robotic picking instrument. Using a yeast agglutination assay, we identified 13 wells in which agglutination had occurred. After isolation of the respective clones, we confirmed yeast agglutination in four of the clones using the standard protocol (9). Two primers, P2-A and P6, were used for PCR amplification of the DNA region covering the invertible element, resulting in an 817-bp fragment. As both primers are located outside the inverted sequence, we could not distinguish between the two genotypes directly by PCR. However, an asymmetrically positioned restriction enzyme site, BsmFI, in the PCR fragment allowed us to orientate the fragment. After digestion with BsmFI, the wild-type orientation resulted in two fragments of 455 bp and 362 bp, while an inverted fragment gave two fragments of 431 bp and 386 bp, demonstrating the wild-type orientation in WCSF1 and the inverted orientation in the four Agg⁺ clones, respectively (Fig. 4). Furthermore, DNA sequence analysis confirmed inversion of the 104-bp element in the four highly agglutinating variants (data not shown). This clearly points to a strong linkage between the Agg⁺ phenotype and inversion of the 104-bp element located between the two inverted repeats.

FIG 4.

Detection of inversion pattern upstream of the msa gene. PCR products covering the inverted-repeat region upstream of the msa gene from L. plantarum WCFS1 and from four isolates with inverted sequences were digested with BsmFI and separated by agarose gel electrophoresis, followed by ethidium bromide staining. Lane M, 100-bp DNA ladder (New England BioLabs, Beverly, MA); lane 1, band pattern obtained from wild-type strain WCFS1; lanes 2 to 5, band pattern from the four highly agglutinating WCFS1 isolates (strains AMJ1632 to AMJ1635).

To analyze the effect of DNA inversion on msa expression, the promoter-probe vector pAK80 (26) was used in a transcriptional analysis of msa. This vector carries a promoterless β-galactosidase reporter gene (lacLM) and replicates in L. lactis but not in L. plantarum. A 499-bp fragment covering the distal end of lp_1228 and the proximal end of msa was PCR amplified from strains WCFS1 and WCFS1-Agg⁺ and inserted into the pAK80 vector. The on and off orientations of the msa gene were tested in transformed L. lactis on agar plates supplemented with the chromogenic substrate 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-Gal). The off orientation gave light-blue colonies, while the on orientation gave dark-blue colonies. The difference in β-galactosidase activities on agar plates was confirmed in liquid culture assays where the on orientation resulted in low, but significantly (2- to 4-fold) higher, β-galactosidase activity compared to that in the off orientation.

Northern blot analysis using a probe that covered the ConA domain of msa showed an intense transcript of approximately 3.6 kb in WCFS1-Agg⁺ and 299v, whereas only a weak transcript was detected in WCFS1 (Fig. 5). The promoter-probe analysis and Northern blot analysis demonstrated that inversion of the 104-bp element affects the transcription of msa, explaining the different phenotypes of strains 299v, WCFS1, and WCFS1-Agg⁺. Also, the size of the hybridization product proved that the msa gene is cotranscribed with lp_1230 as part of an operon structure, as has been suggested by others (17).

FIG 5.

Northern blot analysis. Expression of the msa gene was analyzed by Northern blot analysis using a ConA-specific gene probe. Total RNA was extracted from the wild-type L. plantarum strains WCFS1 and 299v and from a highly agglutinating WCFS1 isolate. Lane 1, WCSF1-Agg⁺; lane 2, WCFS1 wild type; lane 3, 299v wild type. The arrow indicates the position of the 3.6-kb msa transcript.

Alternative secondary mRNA structures mediate variable msa expression in L. plantarum WCFS1.

Because the putative msa promoter is located outside the invertible DNA element, a simple promoter inversion can be ruled out as a mechanism for controlling the expression of the msa gene. However, expression may also be regulated by alternative secondary mRNA structures. Interestingly, two additional inverted repeats are located on each side of the right inverted repeat (IR_R), when the invertible element is oriented in the off position (Fig. 3, 6, and S1). These two 14-bp inverted repeats, located in the untranslated leader of the msa transcript, may give rise to an mRNA hairpin structure (ΔG = −29.3 kcal/mol) leading to the termination of transcription. Furthermore, the hairpin structure seems to mask the putative ribosome binding site upstream of the msa gene, leading to inhibition of translation initiation. During inversion of the 104-bp DNA fragment, formation of the hairpin structure becomes less energetically favorable (ΔG = −12.7 kcal/mol), and the ribosome binding site becomes exposed, corresponding to the on position (Fig. 6). The formation of such a stem-loop structure in the untranslated leader of the msa transcript could therefore explain the lack of msa expression in the off position in strain WCFS1. Although identical mRNA structures potentially can be formed in strain 299v, msa is highly expressed in this strain. The basis for the different expression in strain 299v is unknown but may be due to the presence/absence of different factors in the two strains.

FIG 6.

Prediction of the secondary structure of the untranslated leader sequence in the msa mRNA. The structure to the left shows the stem-loop structure leading to low msa gene expression, whereas the structure to the right occurs after inversion of the 104-bp DNA sequence. The structures and free energies were calculated using the mfold web server. The ribosome binding site is indicated in both structures.

DISCUSSION

Mannose-specific adhesins, found in a number of L. plantarum strains (16, 17), have been suggested to mediate the exclusion of pathogens in the human GI tract. Two L. plantarum strains, WCSF1 and 299v, have been of special interest with respect to mannose adhesion capacity. Both strains exhibit mannose-specific adhesion mediated through the Msa surface protein, but the capacity is much higher in strain 299v (16). Sequence analysis showed that mannose-specific adhesins are composed of common domains, whereas the C-terminal part especially shows varied domain composition (16). In this context, 299v contains three MUB type II domains and 16 proline-rich PxxP repeats in the C terminus, while WCFS1 only has two MUB domains and seven PxxP repeats. Despite these differences, no clear correlation between mannose adhesion capacity and domain composition has been found (16). In this study, we investigated if gene expression variations could explain the diversity in mannose-binding capacity between L. plantarum strains WCSF1 and 299v. Northern blot analysis showed a strong msa transcript in 299v, whereas only a weak transcript was detected in WCFS1, most likely explaining the increased mannose-binding capacity of 299v. DNA sequencing of the region upstream of the msa coding sequence in both strains showed putative −10 and −35 hexamers with a low degree of consensus sequence to typical promoters recognized by the σ70 factor. Except for a single nucleotide difference between the two promoters in the −35 region or in the UP element just upstream of the −35 hexamer, the two promoter regions were identical. Whether the identified single nucleotide difference is implicated in the increased msa expression in 299v and therefore in the higher mannose adhesion capacity in 299v versus WCSF1 is unknown. The difference in yeast agglutination in strains WCFS1 and 299v could also be caused by the different domain compositions in their Msa proteins. However, our results show that the Msa protein properties cannot alone be responsible for high yeast agglutination capacity, as a high capacity similar to that of 299v can be obtained by overexpressing the Msa protein from WCFS1, as demonstrated in the highly agglutinating variants derived from WCFS1.

During the course of this study, we most unexpectedly observed changes in the yeast agglutination phenotype of WCFS1, as we (albeit with a low frequency) isolated subcultures that displayed a very strong agglutination phenotype (Agg⁺). DNA sequence analysis showed a 104-bp DNA element located downstream of the msa promoter but upstream of the msa coding sequence (i.e., in the nontranslated leader mRNA region), which was inverted in the WCFS1-Agg⁺ clones. This type of phase-variable expression has been found in other bacterial systems, but most often, phase-variable gene expression is regulated by the inversion of promoter regions. In Bacteroides spp. which are able to colonize the intestinal tract, surface antigenicity is modulated through phase-variable expression of surface-localized capsular polysaccharides (27, 28). The variable expression of the polysaccharides is controlled by DNA inversions of the promoter regions located between the biosynthetic loci. Different families of site-specific recombinases mediate the DNA inversions. Also, in E. coli, the expression of fimA is controlled by promoter inversion. Several protein factors, including an integrase, bind to recognition sites based on inverted repeats on the target DNA (20). In WCFS1, we believe that msa gene expression is controlled by the formation of a strong stem-loop structure followed by a stretch of Us around the right inverted repeat (IR_R) leading to the termination of transcription. By inversion of the DNA fragment, the stem-loop structure is disrupted, allowing transcription through the alternative much weaker stem-loop structure. In addition to the putative rho-independent transcriptional terminator present in the wild-type sequence, it is noteworthy that the ribosome binding site upstream of msa is masked within the stem-loop structure. After DNA inversion, this RBS becomes accessible to ribosome binding. This indicates that the regulation of msa expression in WCFS1 is controlled at both the transcriptional and translational levels. The increase in msa transcription was clearly demonstrated by Northern blot analysis, where an increase in msa transcripts was detected in a highly agglutinating WCFS1 clone.

The mechanism that controls msa expression through the stem-loop structure in WCFS1 is abolished in 299v. The expression of msa seems to be constitutively high in 299v, which could be due to the presence of a regulatory transcription factor in strain 299v allowing read-through of the stem-loop structure. We did not observe any DNA inversion in 299v, despite the presence of identical inverted repeats with the same free energy in the two strains. We have not identified a putative recombinase or other factors necessary for inversion of the 104-bp fragment in WCFS1. Such factors may be missing or inactive in 299v, thereby explaining the lack of this inversion phenomenon in this L. plantarum strain. There may also be a selection against 299v clones with increased mannose-binding capacity, because overexpression of msa using a high-copy-number vector is lethal to this strain, whereas expression on a low-copy-number plasmid was tolerated. Overexpression of the msa gene from WCFS1 in strain WCFS1 was not described to be lethal (17), indicating that differences in the primary amino acid sequences of the two Msa proteins, such as the number of type II MUB domains and PxxP domains, may cause the toxic effects. However, as different promoters were used for Msa overexpression in the two studies, we cannot exclude that the described toxicity is due to gene dose effects rather than the properties of the two proteins. Supporting the first hypothesis, i.e., that the Msa toxicity is due to the protein characteristics and not gene dosage, is the fact that high Msa expression is well tolerated in the highly agglutinating isolates of WCFS1. Interestingly, we also saw no toxic effect of overexpressing the msa gene in L. lactis. We hypothesize that a very high density of Msa in the cell wall upon msa overexpression could be lethal in L. plantarum 299v, while the Msa cell wall-anchoring motif (LPQTNE) is less efficiently recognized by the sortase in L. lactis, as the cell wall-anchoring motif is different (LPXTGX) in this strain. Therefore, we speculate that Msa from 299v is primarily secreted in L. lactis and to a much lesser degree anchored to the cell wall. In order to identify any particular region in the Msa protein causing the toxicity in L. plantarum 299v, further experiments are required allowing Msa domain exchange, use of the same expression vector, and testing expression of Msa variants in the two strain backgrounds simultaneously.

The promoter region upstream of the msa gene was analyzed by a transcriptional fusion to the promoterless β-galactosidase gene present on a plasmid that only replicates in L. lactis. Rather low β-galactosidase activity was measured, which is most likely due to a weak Lactobacillus promoter, which is poorly recognized by the transcriptional machinery present in L. lactis. This could suggest a trans-acting factor, such as a transcriptional activator, which is missing in L. lactis but is required for high expression levels. Preferably, the promoter studies should have been done in L. plantarum. Another aspect of the regulation of msa expression is whether gene expression is controlled on the transcriptional or translational level or a combination thereof. In the current study, we only analyzed the expression at the transcriptional level, as the promoter-probe vector contains stop codons in all three reading frames preventing translation of the β-galactosidase (β-Gal) reporter using the translational initiation region upstream of the msa gene. It could therefore be of interest to create transcriptional and translation msa gene fusions to study the expression of the msa promoter containing the inverted region in both the on and off positions in L. plantarum.

The importance of Msa for the binding to epithelial cells was proven by construction of an msa mutant. We showed that the adhesion capability of Lactobacillus cells to Caco-2 cells was significantly reduced for an msa mutant strain (299vΔMsa) versus the wild-type 299v strain. This confirms the data obtained with the spontaneous and uncharacterized mutant strain named adh 299v (6). Our data suggest a role for Msa as a factor important for competitive exclusion of intestinal pathogens that can bind to the same mannose receptor sites present in the gastrointestinal tract. Furthermore, it would be interesting to compare the wild-type WCFS1 strain and the isogenic strongly agglutinating isolate in intestinal studies in order to analyze whether increased mannose-adhesive properties are advantageous with respect to competitive exclusion of pathogens or, for example, immunomodulation, as shown by others (29). Additionally, the highly agglutinating WCFS1 isolate is a non-genetically modified organism (non-GMO), in contrast to conventional msa-overexpressing clones made by genetic engineering, enabling its use in animal and human trials.

The above-described switch mechanism controlling the expression of msa in WCSF1 is an interesting phenomenon that is different from other cases where the switched part includes the promoter region (19, 21) and not the region encompassing the untranslated mRNA region.

It would be intriguing to apply Msa as a carrier to surface display vaccine epitopes, antigens, or proteins and enzymes of interest. This could be done by creating genomic translational fusions to, e.g., the 5ʹ end of the msa gene. This would allow a certain fraction of the culture to express the fusion protein at high titers due to the switch mechanism; alternatively, if a locked construct is created, it would allow the entire culture to express the fusion protein at high titers. Using techniques like CRISPR/Cas9, it would be possible to make genomic fusions rather than a plasmid-based construction, leading to high-yield expression and high genetic stability without the need for an antibiotic selection marker.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The bacterial strains used are listed in Table 2. L. plantarum was grown at 37°C on de Man-Rogosa-Sharpe (MRS) medium (Oxoid, Hampshire, United Kingdom) without aeration and supplemented with 3 μg/ml erythromycin (Merck, Darmstadt, Germany) when appropriate. L. lactis was grown at 30°C in M17 medium (Oxoid) supplemented with 0.5% glucose (GM17). 5-Bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) was used at a concentration of 160 μg/ml in GM17 agar plates. E. coli DH10B (Invitrogen, Carlsbad, CA) was used as an intermediate cloning host and grown in LB medium with or without 250 μg/ml erythromycin as appropriate.

TABLE 2.

Strains, plasmids, and cell line used in this study

Strain, plasmid, or cell line	Relevant characteristic(s)	Reference or source
Strains
Lactobacillus
L. plantarum 299v	Human intestine	36
L. plantarum WCFS1	Human saliva	23
L. plantarum 299vΔMsa	299v Msa deletion mutant	This study
AMJ1632	Highly agglutinating WCSF1 isolate	This study
AMJ1633	Highly agglutinating WCSF1 isolate	This study
AMJ1634	Highly agglutinating WCSF1 isolate	This study
AMJ1635	Highly agglutinating WCSF1 isolate	This study
L. lactis
MG1363	Wild type	37
AMJ1294	MG1363 containing pAMJ1294	This study
E. coli
DH10B′	E. coli cloning host	38
TOP10F	E. coli cloning host	Invitrogen
Yeast strain
Saccharomyces cerevisiae	Wild type	Bioneer collection
Plasmids
pCR2.1-TOPO	TA Cloning vector	Invitrogen
pTRKL2	Low-copy-number plasmid	39
pNZ8048	High-copy-number plasmid	40
pTN1	Lactobacillus integration vector, Ts replicona	33
pAK80	L. lactis promoter-probe vector with lacLM reporter	26
pKWY2589-773	P170-based L. lactis expression vector	This study
pAMJ1271	msa gene (299v) with ConA deletion inserted into pTN1	This study
pAMJ1294	ConA domain (299v) inserted into pKWY2589-773	This study
pAMJ1296	msa gene (299v) inserted into pAMJ1309	This study
pAMJ1297	msa gene (299v) and Lactobacillus promoter inserted into pNZ8048	This study
pAMJ1309 (pTRKL2v1)	pTRKL2 with constitutive L. plantarum promoter	This study
pAMJ2052	msa gene (299v) with ConA deletion inserted into pCR2.1-TOPO	This study
pAMJ2075	msa gene (299v) inserted into pCR2.1-TOPO	This study
pUP747	msa promoter (WCFS1) and IR element (wild-type orientation) in pAK80	This study
pUP751	msa promoter (WCFS1) and IR element (inverted orientation) in pAK80	This study
Cell line
Caco-2	Human Caucasian epithelial colorectal adenocarcinoma	Sigma-Aldrich

^aTs, temperature sensitive.

DNA isolation and manipulation.

Chromosomal DNA from L. plantarum (30) and plasmid DNA from L. lactis and from L. plantarum (31) were prepared as described. E. coli plasmids were isolated using a plasmid extraction kit from Genomed (Bad Oeynhausen, Germany), as recommended by the manufacturer. L. lactis and L. plantarum were made electrocompetent and transformed as described previously (32), while competent E. coli DH10B cells were purchased (Invitrogen, Groningen, The Netherlands) and transformed as recommended by the manufacturer. DNA restriction and modification enzymes (New England BioLabs, Beverly, MA) were used as recommended by the manufacturer. DNA was sequenced with a Thermo Sequenase fluorescent-labeled primer cycle sequencing kit (Amersham Pharmacia, Uppsala, Sweden), Cy5-labeled primers, and an ALFexpress DNA sequencer (Amersham Pharmacia).

Yeast agglutination assay.

Bacteria were tested for their ability to agglutinate yeast cells, as described previously (9), but with minor modifications. Briefly, overnight cultures (diluted to the same density; see Fig. 1) were washed in PBS (pH 7.4) and resuspended in the same volume of PBS. Five microliters of bacterial solution was transferred to microplates with flat-bottom wells already containing 50 μl baker yeast suspended in PBS. After shaking on a microplate shaker for 2 min, agglutination was assessed by visual inspection of the wells. Intense agglutination gives a clear solution with one big clump. Wells were also inspected under the microscope after transferring the solution to glass slides.

Overexpression of msa and construction of a deletion mutant in L. plantarum 299v.

Cloning of the complete msa gene from L. plantarum 299v was done using PCR and primers lp_1229 up overexpression and lp_1229 down overexpression (Table 3 ), which introduce terminally situated BamH1 restriction sites. Purified genomic DNA from L. plantarum 299v was used as the template. PCR products were purified using the GFX purification kit (Merck KGaA, Darmstadt, Germany) and cloned into the pCR2.1-TOPO vector (Invitrogen). The resulting pAMJ2075 contains the full reading frame of msa and the ribosome binding site. The msa gene, including the ribosome binding site, was isolated on a BamH1 fragment from pAMJ2075 and inserted into similar sites of the low-copy-number expression plasmid pTRKL2v1 (pAMJ1309) carrying a constitutive Lactobacillus promoter, resulting in plasmid pAMJ1296. The msa gene, including the promoter from pTRKL2v1, was isolated on a PstI-FspI fragment and inserted into compatible PstI and BsaAI sites of pNZ8048, resulting in the high-copy-number plasmid pAMJ1297. The two plasmids pAMJ1296 and pAMJ1297 were used for complementation studies.

TABLE 3.

Primer sequences used in this study

^a Underlining indicates a restriction recognition sequence; lowercase letters indicate homology regions.

The L. plantarum 299vΔMsa mutant was constructed by homologous recombination, as described below. First, two distinctly located fragments of the msa gene were PCR amplified and joined by the use of overlapping primer sequences. Using primers P1 and P2 and primers P4 and P5, the regions upstream and downstream of the ConA-like domain, respectively, were amplified by PCR. (Primers P2 and P4 have cDNA sequences in their 5ʹ ends.)

The two PCR products were mixed, annealed, and applied in a third PCR using the two outer primers P1 and P5. The final PCR product carrying part of the msa gene with an internal 1,500-bp deletion was cloned in pCR2.1-TOPO, leading to plasmid pAMJ2052. Besides the flanking EcoRI restriction sites present in pCR2.1-TOPO, the assembled PCR fragment contains an internal EcoRI restriction site close to the 5′ end. After digestion of pAMJ2052 with EcoRI, the fragment containing the ConA deletion was isolated and inserted into the EcoRI site of integration vector pTN1 (33). Cloning was performed in L. lactis MG1363, leading to plasmid pAMJ1271. The structures of the primary transformants were verified using PCR with primers pTN1-up and pTN1-down. The pTN1 vector replicates at 30°C but not at 41°C, allowing use of the vector for single-copy integration in L. plantarum 299v. Plasmid pAMJ1271 was isolated from L. lactis and introduced into competent L. plantarum 299v cells by electroporation, followed by plating on MRS agar containing erythromycin and incubation for 48 h at 30°C. A colony was picked and cultured overnight at the permissive temperature (30°C) in MRS with erythromycin, diluted 10³-fold in fresh MRS without erythromycin, and shifted to the nonpermissive temperature (41°C) overnight. This culture was diluted 10⁴-fold, plated on MRS with erythromycin, and incubated for 48 h at 41°C to obtain single-copy integration of pAMJ1271 into the chromosome. Plasmid integration was verified as erythromycin-resistant clones and by PCR using primers P5 and P6 (Table 3). Excision of the integrated plasmid by a single-crossover event was induced by incubation in nonselective MRS broth at the permissive temperature (30°C). Bacteria were spread on MRS (without erythromycin, 30°C), and individual colonies were inoculated as duplicates into microwells (96-well plates) containing 100 μl MRS with and without erythromycin, respectively. After incubation at 30°C, cultures sensitive to erythromycin were checked for their ability to agglutinate yeast cells. Erythromycin-sensitive clones represent either msa deletion mutants or the wild-type strain. A deletion mutant named 299vΔMsa was identified by its lack of the agglutinating phenotype. DNA sequencing confirmed that the DNA region corresponding to amino acid positions 96 to 598 was deleted.

Recombinant expression of the lp_1229 ConA domain and affinity purification.

The ConA domain of msa was PCR amplified using primers lp_1229 ConA frw and lp_1229 ConA rev and using L. plantarum 299v as the template DNA. The primers introduced terminal BglII and SalI restriction sites, allowing insertion into the P170-based L. lactis expression vector pKWY2589-773 (Table 2). The resulting plasmid, pAMJ1294, was introduced into L. lactis MG1363. The recombinant L. lactis strain, AMJ1294, expressing the ConA domain, was fermented in a 1-liter bioreactor (34). The ConA domain was affinity purified by applying crude culture supernatants to prepacked mannose columns (Amersham). Nonbinding proteins were removed by washing with PBS, and mannose-binding proteins were eluted by the addition of 20 mM mannose in PBS. Samples (0.5 ml) were collected and analyzed using 14% SDS-PAGE gels.

Quantification of surface-displayed Msa using ELISA.

Purified recombinant ConA-like domain was mixed with Freund’s incomplete adjuvant (FIA) and used for immunization of a rabbit (DakoCytomation A/S, Copenhagen, Denmark). One rabbit was immunized with 0.15 mg ConA protein and boosted five times at regular intervals. A blood sample was collected for an antibody titer check 10 to 14 days after each booster injection. When the desired titer was attained (60 to 70 days), total serum collection was obtained in two bleeds. The Msa-specific ELISA was performed as follows. Cultures grown overnight of the same cell density were washed in PBS (pH 7.4) and resuspended in 1/10 volume of PBS. Serial 2-fold dilutions were made in PBS. Fifty microliters of each cell dilution was transferred in doublets into microtiter plates containing 50 μl PBS with 3% skim milk and a 1/1,000 dilution of antiserum raised against the ConA-like domain. Samples were incubated for 1 h. After 2 washings, the cells were pelleted to remove unbound antibodies, and the cell pellets were resuspended in 100 μl PBS buffer containing 3% skim milk and a 1/4,000 dilution of alkaline phosphatase-conjugated anti-rabbit IgG, according to the manufacturer (KPL, Tåstrup, Denmark). After incubation for another hour, cells were washed 3 times in PBS and resuspended in 100 μl alkaline phosphatase substrate (BluePhos Microwell phosphatase substrate system; KPL). The reaction was stopped by the addition of 100 μl of 2.5% EDTA, typically after 10 min at room temperature. Cells were centrifuged, and 100 μl of the supernatant was transferred to another microtiter plate. The absorbance at 595 nm was measured and plotted as a function of cell density (OD₆₀₀).

Adhesion of lactobacilli to Caco-2 cells.

Caco-2 cells were cultured in Dulbecco’s modified Eagle medium (DMEM; Lonza, Vallensbæk Strand, Denmark) supplemented with 10% fetal calf serum (Lonza), 1% penicillin-streptomycin (PS) (Lonza), 1% GlutaMAX (Lonza), and 1% nonessential amino acids (NEAA; VWR, Rødovre, Denmark) for 4 to 6 h at 37°C until settled. The medium was changed, and the cells were incubated further. After 2 days, 3/4 of the medium was exchanged with fresh medium, and at 70% confluence, cells were treated with trypsin by washing 2 times in PBS without Ca²⁺ and Mg²⁺, followed by the addition of 1 ml trypsin in 9 ml DMEM for 3 to 7 min until cells were released. Caco-2 cells at a concentration of 4 × 10⁵ cells per well were transferred to Nunclon 6-well multidishes (Nunc, Roskilde, Denmark) for growth prior to the adhesion assay. Cells were incubated 16 days with exchange of medium every second day. Lactobacilli were grown for 2 days in MRS medium and diluted to an OD₆₀₀ of 0.5 using conditioned medium corresponding to approximately 5 × 10⁸ CFU/ml. Prior to the adhesion assay, the Caco-2 cells were washed twice in 3 ml PBS and subsequently incubated at 37°C for 30 min with 750 μl PBS. An equal volume of the bacterial suspension was then added. For each strain, 5 wells were used. Adhesion was allowed to occur for 1 h, after which the nonadhesive bacteria were removed by 4 washes in PBS. To visualize the bacteria, each well was fixed for 5 to 10 min with methanol, followed by staining with Giemsa (Sigma-Aldrich, Brøndby, Denmark). For each well, the bacteria were inspected visually under bright-field microscopy using a ×100 objective. Ten randomly selected regions of each well corresponding to a field of view (FoV) were converted into micrographs, and the bacteria in each were counted visually (n = 50).

Cloning of invertible fragments in the promoter-cloning vector pAK80 and β-galactosidase assays.

A DNA fragment of 499 bp, containing the 104-bp invertible element upstream of the msa gene, was PCR amplified using P12 and P13, which introduced terminal BamH1 restriction sites. Genomic DNA from L. plantarum WCFS1 and WCFS1-Agg⁺ was used as the template. The amplified fragments were cloned into the promoter-probe vector pAK80 (26) in fusion with the promoterless lacLM gene, allowing analysis of the msa promoter. The ligation mixtures were transformed into L. lactis MG1363 and plated on GM17 supplemented with erythromycin, and clones were sequenced to determine the orientation of the promoter insertion. The resulting plasmid, harboring the wild-type orientation of the invertible element, was named pUP747, and the plasmid with the invertible element in the inverted orientation was named pUP751. L. lactis transformants were grown on GM17-erythromycin plates supplemented with X-Gal, and β-galactosidase activity was measured as described previously (26).

PCR amplification and restriction enzyme digestion of the DNA region containing the inverted repeats.

Colony PCR was performed using a single colony from wild-type L. plantarum WCFS1 and from the four highly agglutinating isolates derived from WCFS1. The four isolates were named AMJ1632 to AMJ1635 (Table 2). Primers P2-A and P6 were used in a standard PCR. The amplified PCR products of approximately 800 bp were purified using MicroSpin S-400 HR columns (Merck KGaA, Darmstadt, Germany). The purified products were subsequently digested with BsmFI, and the DNA fragments were separated by 2% (wt/vol) agarose gel electrophoresis.

Transcriptional analysis by Northern blotting and primer extension.

Overnight cultures of L. plantarum were harvested, and total RNA was extracted with Pure RNA isolation kit (Roche, Hvidovre, Denmark), following the instructions of the manufacturer. Total RNA was separated on 1% (wt/vol) agarose gels. Blotting, hybridization, and washing conditions were as previously described (35). An 0.8-kb PCR fragment, obtained by use of primers lp_1229 ConA frw and lp_1229 ConA rev, covering the ConA part of the msa gene, was labeled with [γ³²P]ATP and T4 polynucleotide kinase and used as a probe in Northern hybridization.

The 5ʹ end of the msa transcript was determined by primer extension analysis. Ten micrograms of total RNA was mixed with 10 pmol primer P-1229-R-1050-Extension (5ʹ-CCACATTTATAAAGCTTATAGTGTG-3ʹ), 100 nmol dithiothreitol (DTT), 10 nmol dinucleoside triphosphate (dNTP), and PowerScript reverse transcriptase in first-strand buffer (1×) (Clontech, Mountain View, CA). The 10-μl reaction mixture was incubated for 1 h at 42°C and stopped by incubation at 70°C for 15 min. The RNA template was degraded by the addition of RNase H and incubation for 20 min at 37°C. The mixture was ethanol precipitated and dissolved in 10 μl diethyl pyrocarbonate (DEPC) H₂O. A DNA linker was ligated to the 3ʹ end of the cDNA strand (the linker ligation site corresponds to the 5ʹ end of the mRNA) by adding 10 pmol Linker-2 (5′-pCACTCGGGCACCAAGGA-3ddC) and 20 units of T4 RNA ligase to 6 μl of the cDNA mix in a reaction buffer containing 1 μg bovine serum albumin (BSA). The mixture was incubated overnight at 16°C, ethanol precipitated, and dissolved in 10 μl DEPC H₂O. A PCR was done in 100 μl by mixing 6 μl of the cDNA-Linker-2 product with 10 pmol primer P-1229-R-1050-Extension (msa specific), 10 pmol primer F-oligo (5ʹ-GTCCTTGGTGCCCGAGTG-3ʹ), 20 nmol dNTP, 5 units of Taq DNA polymerase in ammonium reaction buffer (Ampliqon, Odense, Denmark), and H₂O. The profile was as follows: 30 s at 95°C, and then 30 PCR cycles of 95°C for 30 s, 50°C for 30 s, and 72°C for 30 s, followed by 7 min of extension at 72°C. A 250-bp PCR product was purified and subsequently inserted into pCR2.1-TOPO (Invitrogen). The recombinant plasmids were transformed into E. coli TOP10F′ (Invitrogen), and the DNA sequence of the inserts was determined. The 5ʹ mRNA start site was identified as the transition point between the Linker-2 sequence and the msa gene sequence.

Data availability.

The nucleotide sequence located upstream of the msa gene in L. plantarum strain 299v was submitted to GenBank under accession number MK249370.