Identification of a Gene Cluster for the Biosynthesis of a Long, Galactose-Rich Exopolysaccharide in Lactobacillus rhamnosus GG and Functional Analysis of the Priming Glycosyltransferase^{▿ †}

Bacterial strains and culture conditions.

L. rhamnosus GG and its derivatives were grown at 37°C in MRS (Difco) or Lactobacilli AOAC medium (Difco) in nonshaking conditions (Table 1). Escherichia coli cells, used as cloning hosts, were grown in Luria-Bertani medium with aeration at 37°C (48). When appropriate, antibiotics (Sigma-Aldrich) were used at the following concentrations: tetracycline (Tc) at 10 μg/ml, ampicillin (Ap) at 100 μg/ml, and erythromycin (Ery) at 10 μg/ml for L. rhamnosus GG and at 100 μg/ml for E. coli.

DNA manipulations.

Routine molecular biology techniques were performed according to standard procedures (48). Restriction and modification enzymes (from New England Biolabs or Roche) were used as recommended by the manufacturer. Plasmid DNA was prepared from E. coli cells with Qiagen miniprep kits. Chromosomal DNA and plasmid DNA were isolated from L. rhamnosus GG as previously described (12).

Sequencing and annotation of the putative EPS biosynthetic cluster.

The wzb gene of L. rhamnosus GG (accession no. EF690379) was identified as previously described (32), starting from a PCR with primers based on the published wzb sequence of the closely related strain L. rhamnosus ATCC 9595 (43). This gene provided the starting point for the sequencing of the remainder of the EPS gene cluster of L. rhamnosus GG. Primers Pro-0224, Pro-0225, Pro-249, and Pro-250 (see Table S1 in the supplemental material) were designed for chromosome walking. DNA sequencing was performed using the chain termination dideoxynucleoside triphosphate method (48) (BigDye Terminator V3.1 CycleSequencing kit, ABI 3100-Avant Genetic Analyzer; Applied Biosystems). This yielded a 2.8-kb fragment including the sequences for the wzb gene and the wzr gene. This information was used to develop a digoxigenin-labeled probe with PCR primers Pro-0300 and Pro-0278 (see Table S1 in the supplemental material) based on an internal fragment of the wzr gene to screen a lambda phage genomic library of L. rhamnosus GG (λCMPG5317). This library was made with the Packagene lambda DNA packaging system (Promega) with E. coli LE392 as a host and phage EMBL3 containing inserts of ca. 15 to 20 kb of L. rhamnosus GG genomic DNA. Southern hybridization resulted in the identification of several phages containing a fragment that hybridized with the wzr probe. One phage that contained a ∼15-kb fragment upstream of wzr was selected. Phage DNA was prepared with the Qiagen lambda DNA extraction minikit and subsequently digested with SalI, BamHI, KpnI, and HindIII. The restriction fragments obtained were subcloned and sequenced. This yielded the DNA sequence of a ca. 15-kb fragment containing the genes encoding glycosyltransferases (welJIHGFE), proteins involved in dTDP-rhamnose precursor biosynthesis (rmlACB), the oligosaccharide transporter (wzx), and polymerase (wzy). To complete the EPS gene cluster, primers Pro-0708, Pro-0839, and Pro-0989 (see Table S1 in the supplemental material) were subsequently used to sequence wzd-wze with genomic DNA of L. rhamnosus GG as a template. Finally, all fragments were assembled in a ca. 20-kb contig containing the DNA sequence for the putative EPS gene cluster of L. rhamnosus GG. The Entrez nucleotide database was screened for similarities using BLASTx (1), and alignment of overlapping fragments was performed with the VectorNTI Advance 10 ContigExpress software (Informax, Oxford, United Kingdom). Annotation of the open reading frames (ORFs) was performed with the Vector NTI Advance 10 software (Informax, Oxford, United Kingdom) and the Sequin sequence submission software. Membrane-spanning regions of translated gene products were predicted using the TMpred program (23). Additionally, the Pfam database (19) was searched for conserved motifs in the translated gene products. The genes were annotated according to Péant et al. (43). These genes were named based on the bacterial polysaccharide gene nomenclature system (45).

Insertional inactivation of the welE gene (CMPG5351).

A knockout mutant of the welE gene, encoding the priming glycosyltransferase, was constructed by double homologous recombination as previously described (30). DNA fragments containing ca. 1 kb up- and downstream of welE were amplified by PCR with primers Pro-0722 and Pro-0723 (see Table S1 in the supplemental material). These fragments were cloned into the pFAJ5301 suicide vector (30). Subsequently, welE was inactivated by insertion of a tet(M) Tc^r cassette from plasmid pMD5057 of Lactobacillus plantarum 5057 (11), amplified with primers Pro-221 and Pro-222 containing EcoRI sites. The tet(M) marker gene fragment was blunted and ligated into the blunted BglII site of the welE gene. The resulting suicide vector pCMPG5351 was transformed to L. rhamnosus GG wild-type cells. At 3 days after the electroporation, transformants were screened for resistance to 10 μg/ml Tc and sensitivity to 10 μg/ml Ery to select for the occurrence of a double homologous recombination event. Confirmation of DNA recombination was obtained by PCR using different primer pairs (Pro-249/Pro-0722 and Pro-0788/Pro-0789) (see Table S1 in the supplemental material). One colony showing the correct insertion pattern of the tet(M) marker and absence of the suicide vector was selected for further analyses and designated CMPG5351. The insertion of the tet(M) marker by double homologous recombination was shown to be 100% stable after 100 generations. Therefore, the mutant was grown in medium without antibiotics for all phenotypic assays, to limit possible interference of Tc with observed phenotypes.

Construction of the complemented strain (CMPG5354).

The welE knockout mutant was complemented by insertion of a wild-type copy including the genuine promoter in a single copy at another locus in the chromosome of L. rhamnosus GG. The self-integrating plasmid pEM40 (2) was used, which integrates at the tRNA^Leu locus in the L. rhamnosus GG genome (12). The welE gene of L. rhamnosus GG was amplified with primers Pro-0722 and Pro-0723 (see Table S1 in the supplemental material). The resulting PCR fragment was digested with the restriction enzymes Ecl136II/EcoRI and ligated in the Ecl136II/EcoRI sites of pEM40, resulting in plasmid pCMPG5353. pCMPG5353 was electroporated to the welE knockout mutant CMPG5351 for complementation, resulting in strain CMPG5354.

Isolation and characterization of CW-PS.

CW-PS were isolated and quantified as described previously (32, 51). Briefly, total CW-PS was extracted from L. rhamnosus GG cells by mild sonication followed by ethanol precipitation and dialysis against water (6,000- to 8,000-Da dialysis membrane [Spectra/Por, VWR International]). The total amount of carbohydrate was estimated by the phenol-sulfuric acid method (16). The sugar monomer composition of the isolated polysaccharides was determined according to the method of Englyst and Cummings by gas chromatography after hydrolysis and derivatization to alditol acetates (17). β-d-Allose was used as an internal standard, and calibration samples (glucose, galactose, rhamnose, and mannose) containing the expected monosaccharides were included with each set of samples. For an estimation of the molecular weights of the isolated polysaccharides, samples were analyzed by size exclusion chromatography as previously described (6). A volume of 50 μl of EPS sample (1 mg/ml) was injected, and the detection was performed with a refractive index detector. The results were compared using a dextran standard series of 8 × 10⁴, 1.5 × 10⁵, 2.7 × 10⁵, 6.7 × 10⁵, and 1.4 × 10⁶ Da (Sigma-Aldrich). Alternatively, differences in molecular weight profiles of the isolated polysaccharides were compared on 5% polyacrylamide gels (10 by 10 cm) (Hoefer miniVE; GE Healthcare) using Tris-borate-EDTA buffer (50 mM Tris, 13 mM EDTA, 15 mM boric acid). Samples were run for 120 min at 350 V. The polysaccharides were stained with the Pro-Q Emerald 488 glycoprotein stain (Molecular Probes, Invitrogen). Gels were imaged with a Typhoon 9400 variable-mode imager (GE Healthcare) using an excitation laser at 488 nm and an emission filter at 520 nm (band-pass).

TEM and SMFS.

To observe the polysaccharides on the cell surface produced in situ, cell surface structures of L. rhamnosus GG were visualized by negative staining in transmission electron microscopy (TEM) experiments using published methods (29). Additionally, SMFS experiments were performed to detect and localize single polysaccharide molecules on live bacteria as described recently (21). Briefly, atomic force microscopy (AFM) tips functionalized with the carbohydrate-binding lectins concanavalin A (mannose and glucose specificity) and PA-I (galactose specificity) were used to scan the surfaces of wild-type L. rhamnosus GG and mutant CMPG5351 for the localization, abundance, and polymer length of mannose/glucose-rich and galactose-rich polysaccharides, respectively. Adhesion forces and rupture distances between the functionalized AFM tips and surface polysaccharides were recorded as described previously (20).

In vitro biofilm formation and adhesion assays.

In vitro biofilm formation by L. rhamnosus GG was assessed by crystal violet staining after 72 h as previously described (32). Additionally, adhesion to immobilized mucins (pig mucin type II; Sigma-Aldrich) was tested in a microtiter plate-based assay (3). Cells were labeled with N-hydroxysuccinimidobiotin (Sigma-Aldrich), followed by detection with streptavidin conjugated to alkaline phosphatase (Roche) and addition of 4-nitrophenylphosphate (Fluka) as a substrate. The percentage of adherence was calculated based on the optical density at 405 nm of the labeled adherent bacteria compared to the optical density at 405 nm of the original labeled suspension before adhesion. Adhesion was tested each time in eight replicates, and each experiment was repeated at least twice. Adhesion to bovine serum albumin (Sigma-Aldrich) was included to account for nonspecific adhesion. Finally, adhesion to Caco-2 cell line was investigated as previously described (29). The adhesion ratio (percent) was calculated by comparing the number of adherent cells to the cell number of the added original bacterial suspension (10⁷ CFU/ml). All strains were tested in triplicate in three independent experiments.

Nucleotide sequence accession number.

The entire sequence determined in this study was submitted to the NCBI database under GenBank accession number FJ428614.

Identification of the EPS biosynthetic gene cluster of L. rhamnosus GG.

As the genome sequence of L. rhamnosus GG is not publicly available, the data for the EPS gene cluster of the phylogenetically related strain L. rhamnosus ATCC 9595 (43) were used as starting point to determine the DNA sequence of the putative L. rhamnosus GG EPS gene cluster. Following the strategy described in Materials and Methods, we obtained an L. rhamnosus GG EPS gene cluster containing 17 putative ORFs (Fig. 1A), of which 16 putatively encode proteins involved in the biosynthesis of various bacterial polysaccharides and 1 (orf1) encodes a putative transposase (Table 2). All ORFs except wzr are located in the same orientation. The EPS gene cluster of L. rhamnosus GG has a modular organization, which is typical for surface heteropolysaccharides (26). The genes putatively encoding regulatory proteins are located at both extremities of the cluster. The first two ORFs (wzd and wze) encode proteins that have predicted amino acid sequences with homology to the putative Wzd-Wze tyrosine kinase complex of L. rhamnosus ATCC 9595 (43) (Table 2). At the other end of the cluster, the wzb gene is located 145 bp downstream of wzr in reverse orientation. The Wzb protein of L. rhamnosus GG shows 99% amino acid identity with Wzb of L. rhamnosus ATCC 9595, which was recently biochemically characterized as a copper-dependent O-phosphatase (28). A regulatory role in CPS biosynthesis and polymer export has been experimentally demonstrated in streptococci for the Wzb ortholog CpsB (48% amino acid similarity) in combination with the autophosphorylating tyrosine kinase complex Wzd-Wze, designated CpsC-CpsD in streptococci (44 to 55% amino acid similarity) (4, 40-42). Wzr of L. rhamnosus GG shows homology to transcriptional regulators of the LytR-CpsA-Psr family involved in cell envelope-related functions, and it contains a short putative N-terminal cytoplasmic domain and a transmembrane domain forming a signal-anchor. Mutation of cpsA, the ortholog of wzr (encoded proteins show 61% similarity), has been shown to affect CPS production in certain streptococci (8, 40).

In addition to regulatory proteins, the EPS cluster of L. rhamnosus GG contains the genes encoding the enzymes for the actual biosynthesis of EPS. Glycosyltransferases synthesize blocks of repeating units that are linked to a lipid carrier at the inner side of the cytoplasmic membrane (Fig. 1B). Six genes (welE to -J) encoding putative glycosyltransferases are contained in the central portion of the putative EPS locus of L. rhamnosus GG. The protein encoded by the welE gene displays 61% identity with YP_001271961 from Lactobacillus reuteri F275, annotated as a galactose phosphotransferase (Table 2). The welE gene putatively encodes the priming glycosyltransferase that transfers the first sugar of each subunit of an EPS molecule of L. rhamnosus GG. welF to welJ probably encode the glycosyltransferases transferring the other sugars of the EPS subunit in an ordered and sugar- and glycosidic linkage-dependent fashion (Fig. 1B). In general, heteropolymeric polysaccharide subunits are flipped across the cytoplasmic membrane by a Wzx-type exporter and polymerized into long polysaccharides by a Wzy-type polymerase at the outer side of the cytoplasmic membrane (Fig. 1B) (13, 56, 58), which has been experimentally best studied in E. coli (57). The wzx and wzy genes of L. rhamnosus GG are located in the 5′ region of the putative EPS cluster (Fig. 1A). In contrast to the conserved Wzd/Wze and Wzb homologs, Wzx and Wzy proteins are transmembrane proteins that are specific for the associated EPS repeating unit (13, 56, 58), as reflected by their lower level of similarity among orthologs (Table 2).

The EPS cluster of L. rhamnosus GG also contains genes for the synthesis of specific nucleotide sugars that cannot be obtained from the central metabolism. The glf gene putatively encodes a UDP-galactopyranose mutase for the conversion of UDP-galactopyranose to UDP-galactofuranose (Table 2). This is in agreement with the presence of galactofuranose residues in the repeating units of the EPS molecules of L. rhamnosus GG, as previously reported (27). Genes for dTDP-rhamnose biosynthesis (rmlACB) are also located in the EPS gene cluster (Fig. 1A). However, the rmlA gene, putatively encoding the glucose-1-phosphate thymidylyl transferase, appears to be inactivated by insertion of ISLrh2. The rmlB and rmlC genes putatively encode a dTDP-glucose-4,6-dehydratase and a dTDP-4-keto-rhamnose-3,5-epimerase, respectively. No rmlD gene encoding the final-acting dTDP-rhamnose synthase was found within the L. rhamnosus GG EPS gene cluster, in contrast to the case for the previously identified EPS gene cluster of the phylogenetically related strain L. rhamnosus ATCC 9595 (43) (Fig. 1A). Nevertheless, Southern hybridization experiments suggest the presence of a complete rmlACBD operon at another locus in the L. rhamnosus GG genome distinct from the EPS gene cluster (data not shown).

Mutation of the priming glycosyltransferase reduces the total level of CW-PS.

As welE putatively encodes the priming glycosyltransferase, i.e., an important control point of EPS biosynthesis, we hypothesized that deletion of this gene should result in a drastic reduction of EPS production. To test this hypothesis, a welE insertion mutant of L. rhamnosus GG, designated CMPG5351, was constructed. The growth capacity of CMPG5351 was not significantly different from that of wild-type L. rhamnosus GG. Wild-type L. rhamnosus GG and welE mutant CMPG5351 both showed a generation time of 1.6 ± 0.1 h and 1.9 ± 0.1 h in MRS and AOAC media, respectively. For subsequent analyses, cells were grown in AOAC medium, as we previously observed that this medium induces a high level of EPS production by L. rhamnosus GG (32). L. rhamnosus GG wild-type cells that are grown in AOAC medium have a typical outer layer formed by CW-PS (29). TEM analysis showed that this layer is absent in the CMPG5351 mutant, although some isolated zones with accumulation of, presumably, polysaccharides could still be detected (Fig. 2A). Additionally, fimbria-like structures appear to become exposed in welE mutant cells at the poles opposite the septum (Fig. 2A). Subsequently, total CW-PS material was extracted from L. rhamnosus GG wild-type and CMPG5351 cells as described in Materials and Methods. The quantity of CW-PS that could be extracted from welE mutant cells was ca. 3-fold lower than that from L. rhamnosus GG wild-type cells (Fig. 2B). Importantly, cis complementation of CMPG5351 with a functional welE gene containing its upstream promoter region (corresponding to strain CMPG5354) could restore CW-PS production to the wild-type level (Fig. 2B), indicating that the welE mutation is not polar and that the welE gene in the wild-type strain is transcribed from its own promoter (see also below).

Mutation of the priming glycosyltransferase specifically abolishes the production of the long, galactose-rich EPS molecules.

Having established that the L. rhamnosus GG welE mutant CMPG5351 contains less CW-PS than the wild-type strain, we subsequently investigated whether the welE mutation specifically affects the production of the long, galactose-rich type of L. rhamnosus GG CW-PS molecules. First, the sugar monomer composition of the CW-PS extracts was determined by gas chromatography as described in Materials and Methods. The extract of wild-type L. rhamnosus GG was found to be composed of approximately 70% galactose, 19% rhamnose, and 10% glucose (Fig. 3A). On the basis of the L. rhamnosus GG EPS structure determined by Landersjö et al. (27), this implies that the CW-PS extract of L. rhamnosus GG wild-type cells contains mainly the known galactose-rich EPS molecules. The sugar monomer composition of the CW-PS extract of the CMPG5351 mutant was drastically changed to approximately 63% glucose, 31.5% rhamnose, 3% galactose, and 2.5% mannose (Fig. 3A). This indicates that the galactose-rich CW-PS molecules are strongly reduced in the welE mutant CMPG5351, while the relative proportion of the (yet-unknown) glucose-rich CW-PS has increased. Importantly, sugar monomer analysis also showed that complementation of the welE mutation in strain CMPG5354 could fully restore synthesis of the galactose-rich type of CW-PS at the L. rhamnosus GG surface (Fig. 3A). Additionally, the size of the CW-PS molecules was estimated by gel electrophoresis of the CW-PS extracts on 5% polyacrylamide gels. These analyses revealed the absence of the largest band in the CW-PS extract of the welE mutant (Fig. 3B), corresponding to the high-molecular-mass (>1.4 × 10⁶ Da) CW-PS molecules based on size exclusion chromatography (data not shown). Finally, SMFS for surface polysaccharides was applied on live bacteria, using functionalized AFM tips containing single lectins specific for end-standing galactose (PA-I lectin) as recently described (21). In comparison to the wild type, the welE mutant CMPG5351 showed a marked reduction in the galactose-rich CW-PS molecules (reduced adhesion frequency with the PA-1 functionalized AFM tip) and a marked reduction in polymer length (reduced elongation and rupture distance) (Fig. 3C). On the other hand, SMFS experiments with mutant CMPG5351 and concanavalin A-functionalized AFM tips showed an increase in the glucose-rich type of polysaccharides, with especially an increase in elongation and rupture length (data not shown).

Taken together, the data on monomer composition, polymer size, and SMFS for the L. rhamnosus GG wild-type, welE mutant, and complemented welE mutant strains demonstrate that the WelE enzyme has a crucial and specific role in the biosynthesis of the high-molecular-weight, galactose-rich type of CW-PS molecules of L. rhamnosus GG, referred to as the EPS molecules of L. rhamnosus GG in this study. Moreover, abolishment of the galactose-rich EPS molecules in welE mutant CMPG5351 seems to increase the presence of the glucose-rich type of CW-PS molecules, which were described for the first time in reference 21.

Inactivation of the priming glycosyltransferase WelE results in increased biofilm formation and adhesion to mucus and epithelial cells by L. rhamnosus GG.

Prior experiments showed that L. rhamnosus GG has an intrinsic high biofilm formation capacity (29, 32). The welE mutant with a specific reduction in the long, galactose-rich EPS molecules allowed us to investigate the relative contribution of these EPS molecules to the biofilm-forming capacity of L. rhamnosus GG. Surprisingly, biofilm formation by the EPS mutant CMPG5351 was increased threefold compared to that by the wild type in AOAC medium (Fig. 4A). Importantly, cis complementation with a functional welE gene in CMPG5351 (strain CMPG5354), restored wild-type levels of biofilm formation (Fig. 4A). Additionally, the in vitro adherence capacity of the EPS mutant CMPG5351 was investigated with mucus as a substrate, as L. rhamnosus GG has been previously reported to display high mucus-adhering properties (52). In our assay, the welE mutant CMPG5351 grown in AOAC medium showed a ca. 1.5-fold-increased adhesion to commercially available pig gastric mucus compared to wild-type L. rhamnosus GG grown under the same conditions (Fig. 4B). Again, complementation restored the adherence capacity to wild-type levels (Fig. 4B). Finally, adhesion of wild-type L. rhamnosus GG, the welE mutant CMPG5351, and the complemented strain CMPG5354 grown in AOAC medium was studied using the human gut epithelial cell line Caco-2. A marked increase in adherence by the welE mutant could be observed, while complementation restored wild-type levels of adherence (Fig. 4C). Overall, these results indicate that the EPS-deficient mutant CMPG5351 has an increased in vitro adherence capacity to several substrates.