Lectin-Like Molecules of Lactobacillus rhamnosus GG Inhibit Pathogenic Escherichia coli and Salmonella Biofilm Formation

Objectives Increased antibiotic resistance has catalyzed the research on new antibacterial molecules and alternative strategies, such as the application of beneficial bacteria. Since lectin molecules have unique sugar-recognizing capacities, and pathogens are often decorated with sugars that affect their survival and infectivity, we explored whether lectins from the probiotic strain Lactobacillus rhamnosus GG have antipathogenic properties. Methods The genome sequence of L. rhamnosus GG was screened for the presence of lectin-like proteins. Two genes, LGG_RS02780 and LGG_RS02750, encoding for polypeptides with an N-terminal conserved L-type lectin domain were detected and designated Llp1 (lectin-like protein 1) and Llp2. The capacity of Llp1 and Llp2 to inhibit biofilm formation of various pathogens was investigated. Sugar specificity was determined by Sepharose beads assays and glycan array screening. Results The isolated lectin domains of Llp1 and Llp2 possess pronounced inhibitory activity against biofilm formation by various pathogens, including clinical Salmonella species and uropathogenic E. coli, with Llp2 being more active than Llp1. In addition, sugar binding assays with Llp1 and Llp2 indicate specificity for complex glycans. Both proteins are also involved in the adhesion capacity of L. rhamnosus GG to gastrointestinal and vaginal epithelial cells. Conclusions Lectins isolated from or expressed by beneficial lactobacilli could be considered promising bio-active ingredients for improved prophylaxis of urogenital and gastrointestinal infections.


Introduction
Antibiotic resistance is a growing issue representing a major challenge for healthcare. The inadequate and inappropriate use of antibiotics in the last decades has led to an increasing incidence of multidrug-resistant bacteria both in hospitals and the community [1]. The situation becomes even more severe taking the capacity of bacterial pathogens to form biofilms on biotic and abiotic surfaces into account, thereby limiting the effect of the available antibiotics [2,3]. Therefore, novel antibacterial agents with the capacity to inhibit bacterial biofilms are important for the treatment of bacterial infections. A promising class of antimicrobial molecules is the family of lectins [4]. Lectins are defined as carbohydrate binding proteins without catalytic activity that are prevalent in all organisms. They often have important functions in cell signaling and cellular interactions [5]. Currently, lectins are especially explored as antiviral agents, since the discovery that infection of heavily glycosylated viruses, such as human immunodeficiency virus (HIV-1) and hepatitis C virus (HCV), can efficiently be blocked by a variety of lectins in cell and animal models [4].
Similar to plants and animals, bacteria can also express lectins on their surface, but reports on their characterization, and especially antipathogenic potential, are very scarce. In the present study, we aimed at the characterization of L-type lectin-like proteins with antimicrobial potential in the model probiotic strain Lactobacillus rhamnosus GG [6]. Genome mining of L. rhamnosus GG resulted in the identification of two genes encoding a putative L-type lectin-like protein. Knock-out mutant analysis and characterization of the purified lectin domains revealed that these novel bacterial lectins exert a unique pathogenic biofilm inhibitory activity and promote adhesion to host cells, suggesting that they might be suitable for future strategies to topically fight bacterial pathogens.
Construction of knock-out mutants in the lectin-like proteins in L. rhamnosus GG To determine the role of the LGG_RS02780 gene, a corresponding knock-out mutant termed CMPG10701 (Table 1) was constructed by double homologous recombination as previously described [7]. Subsequently, a knock-out mutant in LGG_RS02750 was constructed by PCR using primers Pro5841 and Pro5842 and subsequent cloning the amplicon into plasmid pFAJ5301 resulting into plasmid pCMPG10705. The resulting plasmid containing LGG_RS02750 gene was isolated and transferred to highly competent L. rhamnosus GG wild type by electroporation as described above. Plasmid insertion into the L. rhamnosus GG genome was checked by PCR using primers Pro5610-M13 and Pro5611-M13. A putative knock-out mutant of LGG_RS02750 gene was selected by its resistance to erythromycin and confirmed by PCR. Finally, a double mutant was constructed by transferring the plasmid pCMPG10705 into the llp1 mutant CMPG10701 strain by electroporation, resulting in the double mutant llp1-llp2 strain CMPG10707, which was selected by its ability to grow in medium containing erythromycin and confirmed by PCR. Construction of overexpression constructs of the lectin-like proteins in E. coli BL21 (DE3) For heterologous expression of Llp1 and Llp2 proteins in E. coli, the pET 28 a(+) system (Novagen) was used. The LGG_RS02780 and LGG_RS02750 genes from L. rhamnosus GG wild type were amplified by PCR using the corresponding primers listed in Table 2. The LGG_RS02780 and LGG_RS02750 genes were cloned into the pET-28 a(+) vector (Novagen) resulting in plasmids pCMPG10708 and pCMPG10709, respectively. The plasmids were then transformed in competent E. coli strain BL21 (DE3) cells resulting into strain CMPG10708 and CMPG10709. In addition, the L-type lectin domains of Llp1 and Llp2 were also successfully amplified and cloned into pET-28 a(+) resulting in plasmids pCMPG10712 and pCMPG10713 for the L-type lectin domain from Llp1 and Llp2, respectively. pCMPG10712 and pCMPG10713 were successfully transformed to E. coli BL21 (DE3) and designated CMPG10712 and CMPG10713 respectively.

Production of recombinant lectins and lectin domains and their purification
The recombinant E. coli BL21 (DE3) cells expressing the full length lectins or the corresponding lectin domains of LGG_RS02780 and LGG_RS02750 (CMPG10708, CMPG10709, CMPG10712 and CMPG10713) were grown overnight in LB medium with 50 μg/ml kanamycin. The production of recombinant protein was induced with 1 mM isopropyl β-D-thiogalactopyranoside (IPTG) and the cultures were incubated at 25°C with shaking until an OD of 0.8 to 1 was reached. The pellets were suspended in non-denaturing lysis buffer (NaH 2 PO 4 50 mM, NaCl 300mM, imidazole 20 mM) and incubated for 30 minutes at room temperature while swirling and sonicated to release the soluble recombinant lectins from the cells. The full length lectins or the corresponding lectin domains were purified from the cell lysate using affinity chromatography with a HisTrap™ HP column (GE Healthcare). The bound lectin (domain) was eluted using an elution buffer (NaH 2 PO 4 50 mM, NaCl 300mM, imidazole 250mM) and further purified by using size exclusion chromatography. Hereto the sample was applied on a Highload™ 16/60 column packed with a matrix of Superdex™ prep grade (GE Healthcare). Fractions containing the lectin (domain) were collected and analyzed using SDS-PAGE.

SDS-PAGE and Western blot
To verify the expression of recombinant proteins, as well as the presence of pure lectin (domain) after purification steps, each fraction was separated by SDS-PAGE using Bolt 12% Bis-Tris Plus gels (Life sciences). The gels were run submerged in morpholinepropanesulfonic acid (MOPS) buffer for 45 minutes at 400 mA and 200 V. Hereafter, the gels were used for a Western blot or stained with Coomassie Brilliant Blue R-250 (Bio Rad) or Sypro 1 Ruby protein gel stain (Invitrogen).

Bacterial growth assays in suspension
The antimicrobial effect of the lectins on pathogenic growth was assessed by using 100-well microtiter plates (Honeycomb, Oy Growth Curves Ab Ltd.) (Bioscreen) as previously described [29]. Overnight cultures of E. coli UTI89 and S. Typhimurium ATCC14028 were 200-fold diluted in 1/20 TSB and 200 μl was added to sterile wells of 100-well microtiter plates (Honeycomb, Oy Growth Curves Ab Ltd). The purified lectin domains were added at concentration of 200 μg/ml. The bacteria were incubated for 3 days at 37°C with agitation in a Bioscreen (Oy Growth Curves Ab Ltd.), which measured the OD at 600 nm every 10 minutes. Each strain and lectin domain was tested in triplicate.

Antimicrobial assays for pathogens grown in biofilms
Biofilm formation assays on static pegs were performed as previously described with minor modifications [30]. Hereto, E.coli K12, E. coli UTI98, S. aureus SH1000, S. aureus Rosenbach, P. aeruginosa PA14, different Salmonella strains and different Lactobacillus strains were grown on polystyrene pegs in the presence of purified full length lectins or lectin domains at a final concentration of 50 or 200 μg/ml. E. coli UTI98 and S. Typhimurium ATCC14028 were also grown in biofilms when adding different lectins concentrations, namely 200μl/ml, 150μl/ml, 100μl/ml, 50μl/ml, 40μl/ml, 30μl/ml, 20μl/ml, 10μl/ml, 5μl/ml, 1μl/ml, 500ng/ml, 250ng/ml. After 72 hours of growth, the biofilm formation was quantified by staining with crystal violet (0.1 w/v% in 5% methanol, 5% isopropanol and 90% PBS). For each strain and lectin domain, the experiment was performed at least three times with 8 technical repeats. The total cell count of biofilms was also determined as previously described [17]. Briefly, S. Typhimurium ATCC14028 and E. coli UTI89 were allowed to form biofilms at the bottom of polystyrene wells of 12-well culture plates (Cellstar 1 ). The lectin domains were added at 50 μg/ml. After incubation for 48h at 25°C or 37°C for S. Typhimurium ATCC14028 and E. coli UTI89, respectively, the biofilms were detached from the bottom of the wells using scrapers (Greiner bio-one) and pushed through a needle (25G, 0.5 x 16 mm, Terumo) to separate cellular aggregates. The dissolved biofilms were serially diluted in PBS and plated on LB. For each strain, the experiments were performed at least three times with three technical repeats.
For the visualization of S. Typhimurium biofilms, S. Typhimurium ATCC14028 carrying the pFPV25.1 plasmid was used, which constitutively expresses the gfpmut3 gene [17]. Microscopic epifluorescence imaging was performed using a Zeiss Axio Imager Z1 microscope with an EC Plan Neofluar (X40 magnification/0.3 numerical aperture) objective (excitation 488 nm, emission 511 nm). Pictures were acquired with an AxioCam MRm and the AxioVision software. Alternatively, wild type S. Typhimurium ATCC14028 or E. coli UTI89 and FITC-labeled lectin domains were used to visualize the biofilms.
Pull-down carbohydrate binding assay using sepharose beads Sepharose 1 6B beads (Sigma-Aldrich) were coated with 20% D-glucose, D-mannose, Dfucose, GlcNAc and mannan of S. cerevisiae as previously described with little modification [31,32]. For the sugar-binding assay, 25 μl of each functionalized bead was washed with binding buffer (25mM MES, 25mM NaCl and 1% polyvinylalcohol) as previously described [31]. Briefly, 1 ml of binding buffer containing 50 μg/ml of the purified lectin domain was added to each bead. Hereafter, the mixture was incubated at 4°C for 2 h. The beads were washed twice with 1 ml of wash buffer and bound lectin domains were eluted by boiling the beads in SDS-PAGE loading buffer (Fermentas, Life Sciences) for 10 min at 95°C. The bound lectin domains were resolved by SDS-PAGE through 12% polyacrylamide gels (Life Sciences), which were stained with Sypro 1 Ruby protein gel stain (Invitrogen) and scanned by using the Typhoon scanner (GE Healthcare Life Sciences).

Glycan array analysis
The mammalian glycan array version 5.2 was used to explore the carbohydrate binding specificity of the lectin domain of Llp1 and Llp2. The array consists of 609 glycan targets of natural and synthetic mammalian glycans with amino linkers and it is printed onto N-hydroxysuccinimide (NHS)-activated glass microscope slides (SCHOTT Nexterion), forming covalent amide linkages. The purified lectin domains of Llp1 and Llp2 were labeled with FITC by using FluoReporter 1 FITC Protein Labeling Kit (Life Technologies) according to the producer's manual. 200 μg/ml of FITC labeled protein was used to analyze the carbohydrate binding activity. The experiment was performed by the Consortium for Functional Glycomics (CFG, www. functionalglycomics.org).

In vitro adhesion assay to a human epithelial cell lines
Adhesion assays using the Caco2 (ATCC HTB-37TM) and VK2/E6E7 (ATCC CRL-2616™) cell lines were performed as previously described [33,34]. Alternatively, an immunofluorescence assay was performed as previously described [35] with minor modifications. Briefly, FITC labeled lectin domains were suspended in the DMEM medium in the absence of serum and antibiotics, and incubated for 1 h with the monolayer of Caco2 and VK2/E6E7 cells grown on the 13-mm coverslips. After incubation, cells were sequentially washed three times with PBS, and fixed with 4% paraformaldehyde for 10 min. Slides were examined with a Zeiss Axio Imager Z1 microscope with an EC Plan Neofluar (X40 magnification/0.3 numerical aperture) objective (excitation 488 nm, emission 511 nm). Pictures were acquired with an AxioCam MRm monochrome digital camera.

Statistical analysis
To determine significant differences the unequal variance t-test was used. A P-value below 0.05 was considered as statistically significant

Results
The LGG_RS02780 and LGG_RS02750 genes encode Lectin-like protein 1 (Llp1) and Lectin-like protein 2 (Llp2) To identify genes encoding putative lectin-like proteins, the genome sequence of L. rhamnosus GG [10] was screened for the presence of open reading frames (ORFs) containing a lectin Legume (L)-type domain (PF00139). Two genomic regions encoding two putative cell wall proteins, i.e.
The LGG_RS02780 and LGG_RS02750 genes are 2041 bp and 2078 bp long, respectively, encoding polypeptides of 679 and 688 amino acid residues with a similar predicted protein domain organization (Fig 1B). The two proteins contain two conserved protein domains, i.e. an N-terminal Legume-type (L-type) lectin domain (PF00139) and a C-terminal WxL anchoring domain (PF13731). Because of the lectin domain, these gene sequences were annotated as llp1 and llp2, encoding the putative lectin-like protein 1 and 2 (Llp1 and Llp2), respectively. The L-type lectin domain of Llp1 and Llp2 represents approximately 250 amino acid residues and their amino acid sequences show only 35% sequence similarity, suggesting that Llp1 and Llp2 result from a gene duplication event followed by diversification. Sequence comparisons with sequences in the NCBI protein database revealed a number of significant matches with proteins from lactobacilli. Unfortunately, all these proteins remain functionally uncharacterized, except for Msa from L. plantarum WCFS1 [28], which shows 22% and 16% sequence identity to Llp1 and Llp2, respectively.

Llp1 and Llp2 inhibit biofilm formation of the key gastrointestinal pathogen Salmonella enterica serovar Typhimurium ATCC14028
To determine the antimicrobial activity of Llp1 and Llp2, their sequences were overexpressed in E. coli BL21 DE3. Despite numerous attempts, we succeeded to purify the full length Llp1 and Llp2 only in very low concentrations. Fortunately, overexpression of only the predicted Ltype domains from Llp1 and Llp2 resulted in much higher yields and allowed us to purify the recombinant lectin domain from the soluble protein fraction of E. coli BL21 DE3 cells in sufficient amounts.
To monitor interactions between the putative lectins and pathogens, biofilm assays were performed. First, varying concentrations of recombinant lectins were tested. These lectins were added at the start of the static peg biofilm assay (Fig 2A). Both Llp1 and Llp2 were able to significantly reduce S. Typhimurium ATCC14028 biofilms at the highest tested concentration 200 μg/ml with~90% for Llp2 and with~50% for Llp1. Llp2 was still able to inhibit S. Typhimurium ATCC14028 biofilm at a concentration 10 μg/ml with significant decrease of 60%, but the inhibition was lost at concentration of 5 μg/ml. Llp1 did not show a significant inhibition below 50 μg/ml (Fig 2A). Therefore, 50 μg/ml was used for both of the lectins for further experiments to be able to compare the activities of Llp1 and Llp2. At a concentration of 50 μg/ml, the reduction in biofilm formation by S. Typhimurium ATCC14028 was on average with 40% for Llp1 and 90% for Llp2 (Fig 2B). Alternatively, the lectins were added after the adhesion phase (after 1.5 h) in which S. Typhimurium ATCC14028 was first grown on pegs without lectins allowing the bacterial cells to adhere. This resulted in a significant reduction in biofilm formation by S. Typhimurium ATCC14028, which was on average 20% for Llp1 and 92% for Llp2 at a concentration of 50 μg/ml (Fig 2B). No significant decrease in biofilm formation was observed when the lectin domains were added at the exponential growth phase after 8 hours or after 24 hours (Fig 2B). Nevertheless, when the lectin domains were added continuously, i.e. supplemented in the fresh medium added in the beginning and once after 24 hours, a significant reduction in the biofilm formation was also observed amounting to 70% for Llp1 and 93% for Llp2 (Fig 2B).
Absolute CFU counts of S. Typhimurium ATCC14028 biofilms grown in the presence of the lectin domains confirmed the antibiofilm activity of Llp1 and Llp2. In these experiments, Llp1 resulted in a reduction of biofilm growth by S. Typhimurium of 50% compared to the biofilm grown without lectins, while Llp2 resulted in 75% reduction (Fig 2C). Since in their natural context, the lectin domains are part of full length lectins, we also assessed the activity of the full length proteins, which exerted resulted in average reductions in biofilm formation of 79% and 77% for Llp1 and Llp2, respectively (Fig 2D). These inhibitions are similar to those observed for the corresponding lectin domains, though the full length Llp1 seems to be more active than its lectin domain alone.
Given the capacity of the lectins to prevent S. Typhimurium ATCC14028 biofilms, bioscreens were performed with the nutrient-poor medium (1/20 TSB) to provide the same growth conditions as during the biofilm experiments and to investigate whether the lectins also have an antimicrobial effect on growth in suspension. Interestingly, no inhibitory effect on planktonic growth was observed at concentrations of 200 μg/ml (Fig 2E). The growth of S. Typhimurium ATCC14028 was even increased when the lectin domains of Llp1 and Llp2 were added. However, this was not the case when S. Typhimurium ATCC14028 grown in nutrientricher TSB medium (S1 Fig). These results suggest S. Typhimurium ATCC14028 is able to Of note, ConA and HHA two well-known plant lectins were also included in the biofilm assay of S. Typhimurium ATCC14028 as a controls but no inhibition was observed when added at the 0 time point at concentration of 50 μg/ml (S1 Fig). Since the lectins dramatically affected the biofilm formation of the model strain S. Typhimurium ATCC14028, other clinical Salmonella strains were included to investigate whether the lectin activity is strain-and species-specific (Fig 2F). Addition of the lectin domain of Llp2 resulted in a significant reduction (between 50% for SGSC2221 and 90% for most other strains) in the biofilm formation of all of the tested strains. In contrast, the lectin domain of Llp1 only inhibited the biofilm formation of three of the tested strains (SGSC2199, SGSC2221 and SGSC2459) (Fig 2F).

Llp1 and Llp2 structurally disrupt Salmonella biofilms
The divergent antibiofilm activity of Llp1 and Llp2 was also apparent when S. Typhimurium biofilms were visualized. As shown in Fig 3C, incubation in the presence of the lectin domain of Llp2 resulted in biofilms with large holes when compared to the negative control ( Fig 3A). The biofilms treated with Llp1 appeared denser, but small holes were also observed (Fig 3B). In a second assay, biofilms of S. Typhimurium ATCC14028 were grown in the presence of FITClabeled lectin domains. As shown in Fig 3F and 3G, both lectin domains appeared to cause the formation of holes, particularly visible even with a naked eye for Llp2. The results were also confirm by plotting the fluorescent frequency and intensity for each of the tested conditions for GFP-expressing Salmonella (Fig 3D). The fluorescent intensity of biofilms treated with Llp2 showed to be lower compared to the control (Fig 3D). The GFP-expressing Salmonella biofilms treated with Llp1 also showed lower fluorescence intensity as compared to the control, but with increased frequency.

Llp1 and Llp2 have biofilm inhibition capacity beyond the gastrointestinal tract
As for S. Typhimurium ATCC14028, various biofilm assays were performed for the uropathogenic species E. coli UTI89 to investigate whether the lectins from L. rhamnosus GG can affect this pathogen's biofilm formation. First, different concentrations of Llp1 and Llp2 were tested (Fig 4A). Similar to S. Typhimurium ATCC14028, both lectins showed an inhibitory activity at the highest tested concentrations 200 μg/ml with~95% for Llp2 and with~90% for Llp1 ( Fig  4A). Llp2 was still able to inhibit E. coli UTI89 biofilm at a concentration of 10 μg/ml, with a significant decrease of 80%, but the inhibition was lost at concentration of 5 μg/ml. In comparison to S. Typhimurium ATCC14028, Llp1 showed a significant inhibition of the biofilms of E. coli UTI89 also at concentrations of 40 μg/ml, 30 μg/ml and 20μg/ml, but the inhibition was lost at 10 μg/ml. A concentration of 50 μg/ml was used for both of the lectins for further experiments. When purified lectin domains were added at a concentration of 50 μg/ml at the onset of biofilm development, Llp1 and Llp2 reduced the biofilm on average by 88 and 90%, at a concentration of 50 μg/ml. Absolute CFU counts of the treatments was normalized towards the control, which was grown in medium without lectins. (D) Effect of full length (FL) lectin (50 μg/ml) on S. Typhimurium ATCC14028 biofilms added at zero-time point to the biofilms. (E) Growth of S. Typhimurium ATCC14028 in presence of lectin domain of Llp1 and Llp2 (200 μg/ml) in 1/20 TSB medium. (F) Effect of lectin domains (50 μg/ml) on various Salmonella biofilms added at zero-time point. The error bars represent standard deviation of three independent experiments. The dataset comparisons are considered significant (p < 0.05 indicated with one asterisk in the picture, p < 0.01 indicated with two asterisks in the picture or p< 0.001 indicated with three asterisks on the picture).
doi:10.1371/journal.pone.0161337.g002 respectively (Fig 4B). When the lectin domains were added after 1.5, 8 or 24h, both lectins could still reduce the biofilm formation as compared to the control, indicating that they can still inhibit later phases of biofilm development (Fig 4B). In agreement with the previous results, when determining the CFU counts, Llp1 and Lp2 caused significant reductions of E. coli UTI89 biofilm development, by on average 42% and 60% respectively (Fig 4C). Moreover, a bioscreen of E. coli UTI89 pointed out that the lectin domains did not affect the planktonic growth (Fig 4D), confirming that they have specific biofilm inhibitory activity without  possessing anti-bacterial capacity. The activity of the well-known plant lectins ConA and HHA was also investigated as controls, but no inhibition was observed when added at the zero time point at concentration of 50 μg/ml (S1 Fig).
The E. coli UTI89 biofilms were also visualized to explore how the lectin domains structurally interfered with the biofilm formation. As shown in Fig 4F and 4G, incubation in the presence of the FITC-labeled lectin domain of Llp1 and Llp2 at a concentration 50 μg/ml resulted in biofilms with large holes when compared to the negative control (Fig 4E). Of interest, the FITC-labeled lectin domain of Llp1 and Llp2 were clearly distributed across the biofilm, but did not bind to the single cells (Fig 4F and 4G zoomed images).

Llp1 and Llp2 show species-specific activity and increase biofilm formation of Lactobacillus species
Since Llp1 and Llp2 can clearly inhibit the biofilm formation of S. Typhimurium ATCC14028 and E. coli UTI89, the activity against other bacterial species was also explored. Both Llp1 and Llp2 could not prevent biofilm formation of the other important pathogens Staphylococcus aureus SH1000 and Rosenbach (Fig 5A), nor of Pseudomonas aeruginosa PA14 (Fig 5B), suggesting that the Llp1 and Llp2 lectins have pathogenic species and strain-specific activity.
In addition, potential new anti-bacterial agents should not affect the beneficial bacteria of the human microbiota. Therefore, the activity of Llp1 and Llp2 against the biofilm formation of the beneficial E. coli K12 species and various Lactobacillus strains was also investigated. Of interest, biofilm formation of E. coli K12 was not significantly affected by Llp1 and Llp2 ( Fig  5C), while the biofilm of the Lactobacillus strains was significantly increased with approximately 2 fold for most of the strains in the presence of Llp1 and Llp2 (Fig 5C).

Llp1 and Llp2 show sugar specificity for complex glycans
As an indication of the mode of action of Llp1 and Llp2, their sugar specificity was determined by pull-down sugar-binding assays using Sepharose beads coated with D-glucose, D-mannose, mannan, D-fucose and N-acetylglucosamine (GlcNAc) (Fig 6B and 6C). Hippeastrum hybrid lectin (HHA), a well-known mannose-specific plant lectin with molecular weight of 12.5 kDa, was used as positive control. The purified lectin domain of Llp1 showed the highest binding to mannan (Fig 6B, lane 2), whereas the lectin domain of Llp2 was binding to mannan and to Dmannose (Fig 6C, lane 2 and 3). For both lectin domains, there was no significant binding to any of the other sugars tested (glucose, fucose and GlcNAc).
Additionally, the purified FITC-labeled lectin domains were used for glycan array screening against a panel of more than 600 mammalian N-glycan structures. These data indicate that the lectin domains of Llp1 and Llp2 both recognize some complex N-glycans (Fig 7A and 7B), such as: i) Fucα1-4(Galβ1-3)GlcNAcβ1-2Manα1-6(Fucα1-4(Galβ1-3)GlcNAcβ1-2Manα1-3) Manα1-4GlcNAcβ1-4(Fucα1-6)GlcNAcβ-Sp22; ii) Galα1-3Galβ1-3(Fucα1-4)GlcNAcβ1-2Manα1-6(Galα1-3Galβ1-3(Fucα1-4)GlcNAcβ1-2Manα1-3)Manα1-4GlcNAcβ1-4GlcNAc-Sp19; iii) Fucα1-2Galβ1-4(Fucα1-3)GlcNAcβ1-2Manα1-6(Fucα1-2Galβ1-4(Fucα1-3) GlcNAcβ1-2Manα1-3)Manβ1-4GlcNAcβ1-4GlcNAβ-Sp20.  Llp1 and Llp2 modulate the adhesion of L. rhamnosus GG to epithelial cells Since we envisaged that the in situ mucosal activity of the Lactobacillus lectins could be enhanced if they promote adhesion to the mucosa, we also studied the role of Llp1 and Llp2 in the adhesion capacity of L. rhamnosus GG cells. Hereto, we created corresponding knock-out mutants, including a double mutant of llp1 and llp2. Functional analysis showed that CMPG10701 (llp1 mutant) and CMPG10706 (llp2 mutant) showed a minor but statistically significant reduction of 17% (p = 0.03) in adhesion to the intestinal epithelial cell line CaCo2. In contrast, CMPG10707 (double mutant in which both the llp1 and llp2 genes are knockedout) showed a more pronounced reduction in adhesion by 34% (p < 0.01), suggesting a partially redundant role for Llp1 and Llp2 in adhesion to CaCo2 cells (Fig 8A). The complemented strains CMPG10702 and CMPG10715 showed complete restoration of the adhesion phenotype, while the double mutant in which only the llp2 gene was re-introduced (CMPG10773) reached the same adhesion capacity as the llp1 mutant CMPG10701 (S1 Fig). The role of Llp1 and Llp2 in the adhesion to vaginal epithelial cells VK2/E6E7 was also investigated. Only the CMPG10706 (llp2 mutant) showed a significant reduced adhesion to VK2/E6E7 cells by 24% compared to the L. rhamnosus GG wild-type, suggesting a divergent role in adhesion to vaginal cells for the two lectins. Similarly, also biofilm formation assays suggest a different role for Llp1 and Llp2, with only the biofilm capacity of the llp2 mutant CMPG10706 being altered compared to wild-type. In fact, this capacity was even increased by 30%, suggesting a suppressive role for Llp2 in biofilm formation (Fig 8C). However, studying the role of the lectins in the adhesion and biofilm capacity of L. rhamnosus GG is difficult in the context of the bacterial cells, because various other cell surface molecules may interfere and also have a role in adhesion. Therefore, the adhesion capacity of the purified FITC-labeled lectin domains of Llp1 and Llp2 after incubation with CaCo-2 and VK2/E6E7 cells grown on cover slips was also explored. Both lectin domains appeared to recognize and bind to the epithelial cells (Fig 8B).

Discussion
In this study, we explored the role of two novel lectin-like proteins (Llp1 and Llp2) isolated from the probiotic L. rhamnosus GG and focused on their activity against various gastrointestinal and urogenital pathogens and beneficial species. Using a combined strategy of knock-out  mutagenesis and heterologous expression of the lectin (domains) with advanced glycan specificity characterization assays and relevant bacterial functional characterization assays, we provide novel in vitro insights in the role of lectins in probiotic-pathogen-host cell interactions. To our knowledge, this is one of the first reports on the detailed characterization of a lectin-like proteins isolated from a probiotic strain.
L-type lectins are well documented and characterized in plants and animals [37,38], but not in bacterial species. However, the L-type lectin domain is part of several cell surface proteins of Gram-positive bacteria [39,40]. Interestingly Llp1 and Llp2 show only 35% sequence similarity at the protein level suggesting a possible gene duplication event followed by diversification. Of note, dynamic evolutionary diversification of genes encoding a L-type lectin receptor kinase (L-type LecRKs) and L-type lectin domain proteins (LLPs) has also been reported in plants [41]. The genes encoding L-type LecRKs and LLPs have been subject of strong positive selection, suggesting functional specialization in plant immunity. The differences observed in this work in the activity of Llp1 and Llp2 from L. rhamnosus GG, as further discussed below, suggest functional specialization of these lectins in lactobacilli.
Most intriguingly, we could show that the lectin domains of Llp1 and Llp2 can have a major impact on the biofilm development of multiple clinically relevant pathogens. First, both Llp1 and Llp2 lectin domains were able to prevent S. Typhimurium ATCC14208 biofilms, but with Llp2 showing the highest activity up to 90% reduction. This biofilm inhibition was only observed when the lectin domains were added at the onset or after 1.5 h of the biofilm formation and not on well-established biofilms. In contrast, both lectin domains were still significantly active against E. coli UTI89 even when added 24h after the onset of biofilm formation, indicating that they can also destroy established biofilms of this important uropathogenic pathogen. We even observed clear differences in biological activity between Llp1 and Llp2 when testing their activity against various clinical strains of Salmonella. Llp2 was active against most Salmonella species tested while the lectin domain of Llp1 only showed a much narrower spectrum. Thus, Llp1 and Llp2 clearly possess a different activity against different pathogens.
The localization of Llp1 and Llp2 within Salmonella and E. coli biofilms as observed after FITC-labeling suggests that they interact with components of the biofilm matrix. This matrix is composed of extracellular polymeric substances, including polysaccharides, proteins such as fimbriae and lectins, DNA and lipids [42] forming a cohesive network that plays an important role in stabilization of the biofilm, adherence of the bacterial cells to surfaces and cell interconnections. The biofilm matrix varies among strains, which may also explain the observed strainspecific and species-specific activity of the lectins against various clinical strains. For example, the biofilm matrix of P. aeruginosa contains the Psl and the Pel exopolysaccharides (rich in Dmannose, glucose and L-rhamosus) and alginate, while the biofilm matrix of S. aureus contains mainly poly-N-acetyl β-(1,6)-glucosamine [43,44]. However the biofilms of these species were not affected by Llp1 and Llp2. Both S. Typhimurium and E. coli biofilm matrices contain the polysaccharides cellulose (β-1,4-D-glucose polymer) and colanic acid (heteropolysaccharide of glucose, galactose, fucose and glucuronic acid) [45,46], which can be a target for Llp1 and Llp2 from L. rhamnosus GG. Of interest, the composition of the colanic acids closely resembles the composition of the complex N-type glycans to which Llp1 and Llp2 show specificity. Therefore, Llp1 and Llp2 might be able to bind to the glycosylated colanic acid and in this way destabilize the biofilm structure. This would explain the observed holes in the biofilms and the unstable biofilms formed by S. Typhimurium and E. coli strain after adding the lectins. In addition, cell surface-associated and extracellular lectins have been shown to play a role in the cross-linking of the polysaccharides and connecting the cells of pathogenic bacteria with the matrix of the biofilm. For example, outer membrane lectins LecA and LecB of P. aeruginosa are known to stabilize the biofilm [47,48], while aggregative fimbriae and proteins (BapA, Yeej and Bap) of S.
Typhimurium and E. coli strains were demonstrated to stabilize the biofilm matrix via longdistance intercellular connections [49]. The exogenous added lectin domains from L. rhamnosus GG, as shown in this work, may therefore also compete with these crosslinking interactions and destabilize the biofilm. Of interest, Llp1 and Llp2 did not inhibit the biofilm formation of beneficial or commensal model species of the gastrointestinal and the vaginal environment. On the contrary, the Llp1 and Llp2 lectins were even found to increase the capacity of the Lactobacillus strains to form biofilms under the tested conditions.
To determine the sugar-dependent mode of action of the lectins, we aimed at characterizing their sugar specificity by using Sepharose beads binding assays and mammalian glycan array, which revealed specific binding of Llp1 and Llp2 to some complex N-glycan structures. This is in agreement with recent studies on plant lectins using glycan arrays showing that their specificity is manifold and cannot be merely described by single sugar monomers [50]. Plant L-type lectins have been reported as a family of lectins with diverse carbohydrate binding specificities, including mannose/glucose, galactose/GalNAc, GlcNAc, fucose and sialic acid. Interesting, the tested plant lectins HHA and ConA, showing overall mannose-specificity similar to Llp1 and Llp2, were not able to inhibit S. Typhimurium ATCC14208 and E. coli UIT89 biofilms. Therefore, the inhibitory effect of Llp1 and Llp2 on biofilm formation is probably based on binding to specific complex sugars in specific configurations. This is well in agreement with knowledge about plant lectins, that monomer specificity is insufficient to characterize specificity and avidity of lectins [50].
For bacterial lectins, only for a limited number of species, the sugar specificity has been determined, and glycan array analysis is rarely used. One important example of a well-characterized bacterial lectin is the soluble lectin from P. aeruginosa LecB (also known as PA-IIL). LecB binds to a large variety of fucosylated oligosaccharides, such as α-Fuc 1-2 Gal and β-Gal 1-4 α(Fuc1-3)GlcNAc, as confirmed with glycan array [51]. The sugar specificity of lectin-like adhesins of E. coli, present on the tip of pili or fimbriae, was also reported to include alphalinked mannosides, galabiose (disaccharide of two galactose residues), sialylated galactose such as 2-3 sialyllactose and GalNac-β-1-4-Gal epitopes [52]. However, to the best of our knowledge, none of the studied bacterial lectins belong to the L-type lectins, which have -in contrast -been well characterized in plants and animals. Nevertheless, the exact sugar specificity of Llp1 versus Llp2 needs further exploration, since only mammalian glycans were screened with the glycan array. Importantly, bacterial glycans are generally far more diverse than eukaryotic glycoconjugates, as they can show an enormous diversity in monosaccharides building blocks, configuration, conformation and stereochemistry [53].
Finally, we could show that Llp1 and Llp2 play a modulatory role in the adhesion capacity of L. rhamnosus GG. It is well-known that SpaCBA pili are the key adhesins providing initial contacts, while other proteins play a secondary role [24,31], as shown in this work for Llp1 and Llp2. Of note, Caco2 cells were shown to have a rich glycosylation profile containing N-acetyl-D-lactosamine (LacNAc or Galβ(1,4)GlcNAc), Sia, GlcNAc glycoconjugates, branched mannosylated structures, as well as terminal αGal/αGalNAc-containing structures [54]. Based on the results of our biochemical analyses and glycan array screening, these sugar residues can form possible binding sites for Llp1 and Llp2. The role of lactobacilli lectins in host cell interaction is in agreement with the known function of lectins isolated from pathogenic bacteria such as E. coli and P. aeruginosa, which are important virulence factors mediating host-pathogen interaction [55]. However, the lectins from L. rhamnosus GG would then play a beneficial function by supporting the probiotic action of the strain and its adaptation to the gastrointestinal tract. The gastro-intestinal tract is the major site of application for L. rhamnosus GG, but the vagina has also been shown as (temporary) target site of L. rhamnosus GG [56]. In agreement, we also observed a role for the lectins in adhesion to vaginal epithelial cells. Hereby, mutant analysis suggests Llp2 has a more important role than Llp1 and is thus possibly being a vaginal nichespecific factor. Nevertheless, these roles need to be further explored in vivo.
In conclusion, the pronounced inhibiting effect of the isolated lectins on biofilm formation of common bacterial pathogens is worth to be explored in more detail and more complex models in further studies. Given the prevalence of problems associated with biofilms and the increased resistance of various bacteria against antibiotics, they probably hold the best potential for topical applications for pathogen exclusion, either alone or in combination with other antibacterials.