Lactobacillus rhamnosus GG was routinely grown on commercial MRS agar (BD, Franklin Lakes, United States) prior to propagation in modified MRS broth for planktonic or biofilm cultivations. The composition of this modified MRS was as follows: 10 g/L tryptone pancreatic digest of casein, 10 g/L beef extract powder, 5 g/L Bacto^TM yeast extract, 1 g/L Tween 80, 2 g/L ammonium citrate tribasic, 5 g/L sodium acetate, 0.1 g/L magnesium sulfate heptahydrate, 0.05 g/L manganese II sulfate monohydrate and 2 g/L dipotassium hydrogen phosphate trihydrate (pH 6.5–7). When appropriate, MRS was supplemented with 2% glucose (glc) (MRS-G) or 2% fructose (frc) (MRS-F) (Sigma-Aldrich, St. Louis, MO, United States) or less (1.75, 1.5, 1.25, 1.0, 0.75, 0.5, 0.25, 0.1, or 0.05%).

GG colonies from MRS agar plates were suspended in MRS supplemented with glc or frc at desired concentrations to obtain an OD₆₀₀ = 0.15–0.2, from which 200 μL was removed and added to flat-bottomed 96-well plates (FALCON; Tissue Culture Treated, polystyrene, Becton Dickinson), and the plates were incubated at 37°C under anaerobic conditions (5% CO₂) for 24 h, 48 h or 72 h. Biofilm formation efficiency was assessed with crystal violet staining essentially as previously described (Sandberg et al., 2008). Briefly, the non-adherent cells were removed from the wells, and the biofilm cells were washed twice with deionized H₂O. Adherent cells were stained with 200 μL of the crystal violet solution (0.1%, w/v) (Sigma–Aldrich, Munich, Germany) for 30 min at RT. Excess stain was removed by washing the cells twice with deionized H₂O, and the stained cells were suspended in 200 μL of 30% acetic acid by shaking at RT (400 r.p.m. for 30 min). The density of the biofilms was recorded at 540 nm using an ELISA reader (LabSystems Multiskan EX, Thermo Scientific, Wilmington, DE, United States). Biofilm experiments were performed several times, using at least eight technical replicates.

GG cells were suspended in MRS-G (2% glc) or MRS-F (2% frc) to obtain OD₆₀₀ ∼ 0.15–0.2, from which 4 mL was added to uncoated 35 mm glass bottom dishes (MatTek, Ashland, MA, United States). Dishes were incubated at 37°C under anaerobic conditions (5% CO₂) for 48 h for biofilm formation. Non-adherent cells were removed, and adherent cells were subjected to LIVE/DEAD viability staining using 5 μM of Syto9 and 30 μM of propidium iodide (PI) according to the manufacturer’s instructions (LIVE/DEAD^® BacLight^TM, Molecular Probes, Life Technologies, Thermo Scientific, Wilmington, DE, United States). Fluorescence images were captured with Zeiss LSM510 META confocal microscope using Zeiss LSM 3.2 software (Zeiss GmbH, Oberkochen, Germany). All biofilms were analyzed in duplicates.

Samples from planktonic and biofilm cells for 1-DE (SDS polyacrylamide gel electrophoresis) and immunoblotting were prepared as follows. GG cells were suspended in MRS-G or MRS-F to obtain an OD₆₀₀ = 0.15–0.2. The cell suspensions were split in two; half (2.5 mL) of the suspension was cultured under planktonic conditions (Falcon tubes) and half (2.5 mL/ tube) was cultured under biofilm-formation conditions (2.5 mL per well in a flat-bottomed 24-well plate; FALCON – Tissue Culture Treated, Polystyrene, Becton Dickinson) at 37°C in the presence of 5% CO₂. The cell densities of planktonic cultures and biofilms, after suspending the adherent cells in fresh MRS, were measured at 600 nm. Surfaceome proteins from planktonic cells (1.5 mL) after overnight incubation were harvested by centrifugation (4000 × g, 3 min, 4°C). Cells were washed once with ice cold 100 mM sodium acetate buffer (pH 4.7). Cells from biofilm cultures after 48 h of incubation were harvested by removing non-adherent cells and washing adherent cells with ice cold acetate buffer as above. Cells normalized to equal cell densities were centrifugated (4000 × g, 3 min, 4°C) and then suspended gently in Laemmli buffer (pH 7.0) as described earlier (Jakava-Viljanen et al., 2002). Cells were incubated in Laemmli buffer on ice for 15 min and then an additional 5 min at RT. Supernatants were recovered by centrifugation (4000 × g, 3 min, 4°C) and equal volumes from each sample were subjected to 12% TGX^TM Gel (Bio-Rad, Hercules, CA, United States) electrophoresis (1-DE) using 1× Tris-glycine-SDS as the running buffer. After 1-DE, proteins were transferred onto a PVDF membrane using the TransBlot Turbo^TM Transfer System (Bio-Rad, Hercules, CA, United States) according to the manufacturer’s instructions. The membrane was then probed first with antibodies (1:4000) detecting surface-associated factors of GG (Espino et al., 2015) and then with IRDye^® 800CW goat anti-Rabbit IgG (LI-Cor^® Biosciences, Lincoln, NE, United States) (1:20000). After probing, the membrane was blocked using Odyssey Blocking buffer and washed with PBS (phosphate buffered saline, pH 7.4) according to the instructions provided by LI-Cor^® Biosciences. The cross-reacting antigens were detected and quantified using an Odyssey^® infrared imaging system (LI-Cor^® Biosciences, Lincoln, NE, United States) and the AlphaView Alpha View 3.1.1.0 software (ProteinSimple, San Jose, CA, United States), respectively. The experiment was repeated three times (each with two technical replicates).

Adhesion to mucus was carried out as described previously (Vesterlund et al., 2006; Kainulainen et al., 2013). GG was grown in MRS-G and MRS-F with H³-thymidine for metabolic labeling in planktonic and biofilm forms as described above in section 2.4 (1-DE immunoblotting). Briefly, the planktonic and biofilm cells were first washed with a low pH buffer (100 mM sodium acetate, pH 4.7), which prevented the release of adhesive moonlighting proteins (cytoplasmic proteins) from the cell surfaces. Then, the washed cells were allowed to bind the porcine mucin immobilized onto microtiter plate wells. After washing the cells to remove non-adherent cells, the radioactivity of the adherent cells was measured by liquid scintillation, and the percent of bacterial adhesion was determined by calculating the ratio between the radioactivity of the adherent bacteria and that of the added bacteria. The experiment was repeated twice with four to five technical replicates. An unequal variance t-test was used to determine significant differences between selected samples.

Planktonic and biofilm GG cells were grown on MRS-G and MRS-F and washed with sodium acetate buffer as above. Washed cells at the same density were treated with 10 μM 4,6′-diamidino-2-phenylindole (DAPI; Molecular Probes, Thermo Scientific, Wilmington, DE, United States) to stain all bacterial populations and with antiserum raised against His_6–SpaC to detect SpaC expression. Antiserum against purified His_6–SpaC of L. rhamnosus GG was raised in rabbits using routine immunization procedures as described previously (Tytgat et al., 2016). The staining was carried out by using indirect immunofluorescence as described previously (Kainulainen et al., 2012) with minor modifications. Briefly, GG cells were collected by centrifugation, washed once with 100 mM Na-acetate buffer (pH 4.7) and fixed with 4% (wt/vol) paraformaldehyde in phosphate buffered saline (PBS; pH 4.0) prior to detection with anti-His₆-SpaC primary antibody and Alexa-488 (Invitrogen)-conjugated goat anti-rabbit IgG (1 μg/ml) secondary antibody and DAPI counterstain. After staining, the optical density (OD₆₀₀) of the cells was adjusted to 1.0 with washing buffer, and the intensity of the fluorescence in the samples was measured using the Victor3 1420 multilabel counter (PerkinElmer, Waltham, MA, United States). The experiment was repeated twice, each with five technical replicates. An unequal variance t−test was used to determine significant differences between selected samples.

Surfaceome analyses were conducted with GG cells cultured in planktonic (24 h; stationary phase) and biofilm states (48 h) in MRS-G and MRS-F at 37°C under microaerophilic (5% CO₂) conditions. Biofilms were formed using flat-bottomed 24-well plates with 2.5 mL of the indicated media containing suspended GG cells (OD₆₀₀ = 0.1) per well as described above. After 48 h of incubation, non-adherent cells were discarded, and the adherent cells were washed gently with ice-cold 100 mM Tris–HCL (pH 6.8). Pelleted cells suspended gently in 100 μl of TEAB were mixed with 50 ng/μL of sequencing grade modified porcine trypsin (Promega, Madison, WI, United States), and digestions were incubated at 37°C for 15 min. Released peptides and trypsin were recovered by filtration through a 0.2 μm pore size acetate membrane by centrifugation (7000 × g, 2 min, +4°C), and the digestions were further incubated for 16 h at 37°C. The cell samples from planktonic cultures were harvested by centrifugation (7000 × g, 2 min, +4°C) and subsequently washed and treated with trypsin as described above for biofilm cells. Digestions were stopped by adding trifluoroacetate (TFA) to a final concentration of 0.6%. The peptide concentrations were measured using a Nano-Drop (ND 1000, Thermo Scientific, Wilmington, DE, United States) at 280 nm. For each growth mode, three biological replicate cultures were used.

Trypsin-digested peptides were purified using ZipTips (C18) (Merck Millipore, Burlington, MA, United States), and equal amounts of the peptide samples were subjected to LC-MS/MS using an Ultimate 3000 nano-LC (Dionex) and QSTAR Elite hybrid quadrupole TOF mass spectrometer (Applied Biosystems/MDS Sciex) with nano-ESI ionization as previously described (Espino et al., 2015). The MS/MS data were searched against the concatenated GG protein database composed of the target (Acc. No. FM179322; 2944 entries) (Kankainen et al., 2009) and the decoy protein sequences using the Mascot (Matrix Science, version 2.4.0) search engine through the ProteinPilot software (version 4.0.8085). The Mascot search criteria were based on trypsin digestion with one allowed mis-cleavage, fixed modification of carbamidomethyl modification of cysteine, variable modification of oxidation of methionine, 50 ppm peptide mass tolerance, 0.2 Da MS/MS fragment tolerance and 1+, 2+, and 3+ peptide charges. The false discovery rate (FDR) percentages calculated using the formula 2×nreverse/ (nreverse + nforward) (Elias and Gygi, 2007) indicated FDRs < 4% in each data set. The raw Mascot proteomic identifications and associated files are freely downloadable from the Dryad Digital Repository (doi: 10.5061/dryad.15534h9).

Mass spectra from each replica sample were searched both separately and combined against the indicated protein database. Proteins with Mascot score (ms) ≥ 30 and p < 0.05 and identified in at least 2/3 replicates were considered high quality identifications and used to indicate condition-specific identifications and relative protein abundance changes. The emPAI (exponentially modified Protein Abundance Index) values of the high confidence identifications were used to estimate protein abundance changes; emPAI is considered roughly proportional to the logarithm of the absolute protein concentration, which allows label-free and relative quantitation of the protein pairs (Ishihama et al., 2005). Principle component analysis (PCA) of the emPAI values was performed with IBM SPSS Statistics v.24. Missing values were substituted with half minimum emPAI value for each protein identified in at least 2/3 replicates in at least one of the conditions using MetImp 1.2¹ (Wei et al., 2018a, 2018b) (Supplementary Table S1). Two proteins with zero variance (YP_003172193.1; YP_003172584.1) were excluded. PCA of the imputated emPAI data was carried out utilizing Varimax rotation with Kaiser normalization.

Theoretical molecular weights (MW) and isoelectric points (pI) for the identified proteins were acquired using the Protein Manipulation Suite Server v. 2.0 at https://www.genscript.com/sms2/proteiniep.html (Stothard, 2000). The presence of classical secretory motif (signal peptide) and lipobox was predicted using the SignalP Server v. 5.0 at http://www.cbs.dtu.dk/services/SignalP/abstract.php (Almagro Armenteros et al., 2019). The cellular location for each protein was predicted using the PSORTb Server v. 3.0 at https://www.psort.org/psortb/ (Yu et al., 2010) and the number of potential transmembrane domains (TMDs) with the TMHMM Server v. 2.0 at http://www.cbs.dtu.dk/services/TMHMM/ (Krogh et al., 2001). The identified proteins with plausible moonlighting function/entering the extracellular milieu via non-classical route were predicted using the SecP Server v. 2.0 at http://www.cbs.dtu.dk/services/SecretomeP/ (Bendtsen et al., 2005). UniProt/Proteomes² with the GG proteome ID UP000000955 (UniProtKB, 2,832) with Gene Ontology (GO) IDs was used to link biological process, molecular function and cellular component to each identified protein.

Cell surface-associated aminopeptidase activity (PepN and PepC) was determined from planktonic and biofilm cells cultured in the presence of 2.0 and 0.5% glc and frc for 24 h (planktonic and biofilm cells) and 48 h (biofilm cells) using a previously reported method (Varmanen et al., 1994) with the following modifications. Briefly, biofilm and planktonic cells (1.8 mL) obtained from 10 mL cultures were washed as described above and then suspended in 1 mL of cold 100 mM Tris–HCl buffer, pH 6.8. The optical density of each sample was measured at 540 nm. The incubation mixture contained 1 mM L-leucine-para-nitroanilide (Leu-pNA) substrate in 200 μl of the cell suspension. The inhibitory effect of the metal-ion chelator EDTA was examined by incubating the cell suspension with 5 mM EDTA for a few minutes prior to the addition of Leu-pNA. Reactions were incubated at 37°C for 5–25 min and stopped with 800 μl of 30% (v/v) acetic acid. After centrifugation (8000 × g, 5 min), the absorbance was measured at 410 nm. The specific aminopeptidase activity was calculated by dividing the absorbance change at 410 nm by the reaction time used for Leu-pNA hydrolysis and the optical density of the cells (OD₅₄₀). An unequal variance t-test was used to determine significant differences between selected samples.

Biofilm formation of GG was first tested at three time points (24, 48, and 72 h after inoculation, hpi) in MRS-G and MRS-F (2% glc and 2% frc) and MRS without carbohydrate, which indicated that frc stimulated biofilm formation by approximately 2-fold compared to glc at each time point tested (Figure 1A). The highest biofilm masses in MRS-F and MRS-G were detected at 48 and 72 hpi, respectively (Figure 1A). The cell viability analysis of biofilm cells grown on glass-bottom dishes with both carbon sources revealed that biofilms formed on frc at 48 hpi contained proportionally more living cells than those grown on glc (Figure 1B). In addition, the biofilm formation efficiency was higher on hydrophilic polystyrene than hydrophilic glass under the conditions used (data not shown).

The surface antigenicity of GG cells grown in biofilm and planktonic states in MRS-G and MRS-F was studied by subjecting surface-associated proteins extracted in 1× Laemmli buffer to 1-DE combined with immunoblotting using antisera raised against intact GG cells. Figure 1C reveals that in each sample, the most intense/abundant cross-reacting antigen signals migrated to approximately 65, 55, 40, and 35 kDa. Comparing the total antigen intensity profiles, reflecting antigen abundances (Figure 1D), showed that the presence of frc in the growth medium increased the surface antigenicity by ∼two-fold in planktonic and ∼ 2.5-fold in biofilm cells compared to cells grown on glc. In glc-associated cells, switching from planktonic to biofilm growth had no effect on surface-antigenicity, whereas the surface-antigenicity increased by ∼20% after switching from planktonic to biofilm growth in the presence of frc. The presence of Frc in the growth medium resulted in the appearance of unique protein bands at 15 and 20 kDa in both the planktonic and biofilm cell samples, while a higher-molecular weight protein (∼200 kDa) was specific to glc- and frc-biofilm samples. Thus, growing cells in planktonic or biofilm forms on frc increases the overall surface-antigenicity of GG along with the specific production of small-molecular-weight antigens, while a larger protein was specifically produced by biofilm cells growing on both carbon sources.

We previously demonstrated that the biofilm formation of GG is mainly protein-mediated (Savijoki et al., 2011). To explore this further, the surfaceomes of cells cultured in biofilm state were analyzed. For obtaining high amounts of biomass needed for proteome analysis without compromising cell viability, the biofilm samples were collected at 48 hpi (Figures 1A,B). According to previous studies the physiology of mature biofilms resembles that of stationary phase planktonic cultures (Stoodley et al., 2002; Folsom et al., 2010). Therefore, the planktonic cell samples for comparative analyses were harvested from cultures at stationary phase (24 hpi). LC-MS/MS based protein identifications (p < 0.05) are listed in Supplementary Table S2. Four surfaceome catalogs [(ms) ≥ 30, p < 0.05] were generated with each data set showing extensive overlap; 70–94% of the identified proteins were shared between at least two of the replica samples (Supplementary Table S3). The total number of high-quality identifications was 245 [(ms) ≥ 30, p < 0.05], consisting 94, 117, 179, and 227 proteins from the glc-planktonic, frc-planktonic, glc-biofilm and frc-biofilm cells, respectively (Supplementary Table S4). As no decrease in the colony forming units (CFUs) of the shaved cells was observed (data not shown), we concluded that no significant cell lysis had occurred under the trypsin-shaving conditions used. Thus, comparison of the surfaceome catalogs indicated that the number of proteins at the cell surface of GG increased with frc as the carbon source compared to glc. Similarly, growth in the biofilm state increased the number of protein identifications compared to cells grown in planktonic form, regardless of the carbon source used.

A principal component analysis (PCA) was performed to identify carbon source- and growth mode-dependent patterns in the surfaceome data. Figure 2A shows that all biological replicate samples clustered together well, with four clearly separated and identifiable groups. PC1, correlating with the change in growth mode-, and PC2 with the carbohydrate-dependent changes, suggest that the carbon source had a greater effect on GG during biofilm growth than planktonic growth. This was further explored by Venn diagrams that compared the number of specifically identified proteins (unique to growth mode and/or carbon source) and proteins with emPAI values showing ≥ 2-fold change (Supplementary Table S4). Venn digrams in Figure 2B compares the growth-mode and carbon source-associated surfaceomes. The number of specifically identified proteins (not identified from planktonic cells) was 90 and 113 in biofilm cells grown on glc and frc, respectively. In total, 5 and 3 identifications were specific for planktonic cells grown on glc and frc, respectively. In addition, the abundance of 60 and 67 proteins increased ≥ 2-fold in biofilm cells, whereas 2 and 3 proteins were more abundant (≥2-fold) in planktonic cells with glc and frc, respectively. With frc as the carbon source, the number of specific identifications (not identified from glc-grown cells) from the planktonic and biofilm cells was 32 and 61, respectively. For glc, 9 and 13 specific identifications were made from the planktonic and biofilm cells, respectively. Four and 19 proteins displayed ≥ 2-times higher abundances in planktonic cells grown of glc and frc, respectively. The number of more abundant proteins with ≥ 2-fold change was 11 and 33 on glc- and frc-associated biofilms. Altogether, the presence of frc in the growth medium resulted in a higher number of specifically identified proteins as well as proteins with increased abundances in comparison to cells grown on glc.

Because Venn diagrams implied that the carbon source affected the number and amount of proteins attached to the biofilm surfaces, we also explored the biofilm formation efficiency in the presence of varying carbohydrate concentrations. Figure 2C shows that frc enhanced the biofilm formation over a wider concentration rate compared to glc. However, 0.5% glc produced the thickest biofilms at 48 hpi, while over two-fold higher concentrations of frc were required to achieve the thickest biofilms under the same conditions. Furthermore, optical densities (at 600 nm) of the formed biofilms grown on different carbohydrate concentrations revealed only marginal changes in cell densities, indicating that increased biofilm formation was not accompanied by an increased number of cells (data not shown).

Based on the proteome bioinformatics the identified proteins were next divided according to their predicted secretion motifs, subcellular location and cellular functions (Supplementary Table S4). Based on this categorization 48 proteins were predicted to be exported via the classical and 68 via the non-classical secretion pathway. For majority of the identifications the proteome bioinformatics tools imply cytoplasmic or cytoplasmic/membrane location. Next, the most distinctive protein abundance variations, including condition-specific identifications and proteins with abundance change ≥ 2-fold between the indicated conditions were selected for visual illustration by heatmaps; Figure 3 shows relative abundance changes estimated for classical surface proteins and known/adhesive moonlighters and Figure 4 candidate moonlighting proteins. Altogether, these proteins could be categorized into three groups: (i) the classically secreted surface adhesins, transporters and enzymes, (ii) known surface-associated moonlighters with adherence features, and (iii) predicted and new candidate moonlighters lacking established extracellular function. The known and candidate moonlighters were the dominant protein group among the detected surfaceomes. Among the identified putative moonlighters, the ribosomal proteins r-proteins (42 proteins), PTS/ABC-type transporter proteins (17 proteins; several OppA paralogs) and amino acid tRNA synthetases (10 proteins) formed the largest protein groups.

Because the identification data implied that the growth mode and carbon source in the growth medium could affect the adherence features of GG, we investigated the mucus-binding ability of the planktonic and biofilm cells grown on glc and frc. Figure 5A indicates that the biofilm cells grown on frc displayed the highest level of adherence, while planktonic cells grown on glc were the least adherent. The frc biofilms demonstrated a mucus-binding ability that was ∼2-fold more efficient than that of the glc-associated planktonic cells; p = 0.037). The possible role of the SpaC-adhesin, the tip pilin of the SpaCBA pilus known to mediate key interactions with the human mucus (Kankainen et al., 2009), in mediating the frc-stimulated adherence was assessed next by testing the adherence of the GG cells in the presence of anti-SpaC antiserum. Figure 5A indicates that the presence of SpaC antibodies markedly decreased the adherence of each type of cell to mucus. Although the final level of adherence detected for these cells was somewhat similar, the biofilm cells grown on frc displayed the greatest SpaC-mediated inhibition, whereas the level of adherence decreased the least in glc-associated planktonic cells. To readdress the possible role of the pilus structure in the observed differences, we also monitored the SpaC abundances on each GG cell type. As shown in Figure 5B, the highest abundance of SpaC was detected on the glc-associated biofilm cells, which was > 2-fold (p = 4.46e−09) higher than that on the planktonic cells grown on glc and ∼2-fold (p = 1.15e−07) higher than that on the frc-biofilm cells. Planktonic cells grown on frc displayed the lowest SpaC abundance.

Varying abundances detected for predicted cell-surface moonlighters, such as PepN, PepC, PepA, ClpP, and several OppA paralogs, implied differences in the ability of the GG cells to utilize casein/casein-derived peptides for growth. PepN was specific to frc-biofilm cells. PepC was predicted to be equally abundant on the biofilm cells grown on frc and glc and displayed over a 4-fold higher abundance on biofilm than planktonic cells. Comparison of PepN and PepC implies that PepC is produced ∼2-fold more on the frc-biofilm cells than PepN on the same cells. To verify these results, we monitored the surface-associated aminopeptidase activity by exposing washed and intact GG cells to a chromogenic substrate, Leu-pNA, a specific substrate for both aminopeptidases (Varmanen et al., 1994). Figure 5C confirms the identification findings by showing that the surface-associated aminopeptidase activity of frc-biofilm cells was ∼4-fold higher (p = 0.001) than that of planktonic cells growing on frc. Changes from glc to frc increased the aminopeptidase activity on both the planktonic and biofilm cell surfaces but led to a proportionally greater increase in the biofilm cell surface. Lowering the carbohydrate concentration to 0.5% increased the aminopeptidase activity in planktonic cells by ∼4-fold, while only a slight decrease in the enzyme activity was observed in biofilm cells on both carbon sources. Enzyme activities were also examined in the presence of 5 mM EDTA, a metal-chelator that inhibits the activity of metallopeptidases such as PepN while not affecting other types of peptidases such as PepC (Varmanen et al., 1994). Since EDTA inhibited only marginally the total aminopeptidase activity only in frc-biofilm cells, PepC is the likely candidate for the hydrolysis of the substrate.

Environmental signals, such as O₂, CO₂, bile, mucins, non-digestible polysaccharides and carbohydrates present in the human GIT, have been shown to affect the GG biofilm formation in vitro (Lebeer et al., 2007b; Savijoki et al., 2011). In MRS growth medium, GG produces biofilm, which is completely disintegrated by treatment with proteolytic enzyme (Savijoki et al., 2011), indicating the presence of a large amount of proteins in the biofilm matrix and/or that biofilm formation is protein-mediated. Since both carbon metabolism (Lebeer et al., 2007b) and cell-surface exposed proteins (Savijoki et al., 2011) appear to play a role in biofilm formation, we explored this link further by investigating the effects of glc and frc as carbon sources on the surfaceome composition of GG. Our findings indicated that changing the carbon source from glc to frc resulted in enhanced biofilm formation efficiency of GG in vitro. GG is known to exert high in vitro adhesion to Intestinal Epithelial Cells (IECs) and biofilm formation capacity on polystyrene and glass (Doron et al., 2005; Lebeer et al., 2007b). Here, our findings demonstrated more efficient biofilm growth on polystyrene than on glass and more efficient biofilm growth with frc than with glc on both inert materials. Crystal violet provides a good detection of biofilm mass, but this dye stains both the bacterial cells and extracellular matrix, including proteins (Santhanalakshmi and Balaji, 2001; Welch et al., 2012). As the tested carbohydrates had only a marginal effect on the cell density, we suggest that the increase in crystal violet staining, instead of showing increased biofilm formation, related directly to changes in the protein abundance of biofilm cells. A similar observation was also made when growing the biofilm cells with varying carbohydrate concentrations; decreased carbohydrate concentrations resulted in thicker biofilms, which did not result from higher cell density (data not shown). Thus, changes in carbon sources and their concentrations regulate the changes in protein abundance on the biofilm cell surfaces.

Examining the cause-effect relationships between the identified surfaceomes revealed that the carbon source played a more important role in the biofilm mode of growth. Conditions of glc limitation have been shown to enhance the biofilm formation of GG (Velez et al., 2010). Here, decreasing the concentration of both carbon sources in growth medium resulted in similar outcomes. The presence of frc enhanced biofilm formation over wider concentrations than glc. Putative moonlighting proteins, including predicted as well as new candidate moonlighters, were the dominant protein group in the biofilm-associated surfaceomes. Comparing the frc- and glc-associated surfaceomes implied that these non-classical surface proteins are more abundantly present in both the biofilm and planktonic cells growing on frc. These results could be explained by increased cell lysis within the biofilm and the fructose containing cultures or alternatively, be a result of non-classical protein export by a yet unknown pathway. Cell lysis in biofilms is a well-reported phenomenon for model bacteria like S. aureus (Rice et al., 2007; Houston et al., 2011; Mashruwala et al., 2017), but remains to be studied in L. rhamnosus GG. In this study, three proteins potentially involved in hydrolysis of peptidoglycan and autolysis were identified, NlpC/P60, MltA and a CHAP-domain containing peptidoglycan hydrolase. Of these, NlpC/P60 was previously identified from the surface of exponentially growing L. rhamnosus GG cells, while the other two proteins were not detected in that study (Espino et al., 2015). Here, the results indicate constant expression of NlpC/P60 and CHAP-domain containing peptidoglycan hydrolase under the conditions used, whereas MltA appears to be more abundant on stationary phase planktonic cells than on biofilm cells. Thus, while the presence of MltA and the CHAP-domain containing peptidoglycan hydrolase could be linked to increased autolysis in stationary phase and biofilm cells compared to exponentially growing cells, these protein identification results do not support increased autolysis activity as an explanation of higher number of cytoplasmic proteins on biofilm cells compared to stationary phase planktonic cells. The presence of moonlighting proteins on the cell surface or in the culture medium has been reported for several microbial species, and for many of these, the moonlighting activity outside of the cells has been demonstrated (Kainulainen and Korhonen, 2014; Chen et al., 2018). It has been proposed that bacteria recycle conserved cytoplasmic proteins in a pH-dependent manner in the matrix to facilitate interspecies interactions without specifically recognizing the dedicated matrix components of the other species (Foulston et al., 2014). In a recent study, the strongly positively charged r-proteins were shown to be embedded in S. aureus biofilm matrix under acidic conditions (Graf et al., 2019). This was proposed to depend on the pH, coordinated by the formation/release of acidic fermentation end-products in the biofilm cells facing oxygen limitation (Graf et al., 2019). In the present study, the pHs of the spent culture supernatants were clearly acidic, pH < 5.0 (data not shown). Thus, while r-proteins were identified as the most abundant moonlighters in the biofilm matrices, we suggest that low pH could have promoted the interactions between the r-proteins and the biofilm cells also in the present study.

In general, the mechanisms underlying export/transport of moonlighting proteins remain unknown (Wang and Jeffery, 2016). In view of this, it is tempting to speculate that the specific appearance of HlyD/MFP in frc-associated cell surfaces indicates the role of this protein in directing moonlighting proteins out of the cell. HlyD is part of a translocon that comprises HlyB, HlyD, and TolC, which is proposed to coordinate the transport of potential novel proteins by quite different mechanisms (Holland et al., 2016). Overexpression of PrsA, a known surface-associated chaperone or foldase acting on secreted proteins (Kontinen and Sarvas, 1993), on biofilm cells and frc-associated cells is another plausible factor contributing to protein secretion under these conditions. This chaperone/foldase is known to enhance protein secretion efficiency in bacteria (Chen et al., 2015) and act on various substrates, ranging from 20 to 80 kDa in size (Jakob et al., 2015). The serine-type surface-protease, HtrA, follows the same abundance trend as PrsA and is another classical surface-protein that could enhance protein secretion in frc-associated samples. In bacteria, HtrA is involved in degrading abnormal proteins, processing secreted proproteins and the maturation of native proteins (Poquet et al., 2000), but this protease can also affect biofilm formation and control the presence of surface-associated moonlighters, such as enolase (ENO) and glyceraldehyde-3-phosphate dehydrogenease (GaPDH) (Biswas and Biswas, 2005).

Here, ENO, EfTU, TDPA, PGK, and GroEL, each with previously reported adhesive functions (Kainulainen and Korhonen, 2014; Chen et al., 2018), were detected as the most abundant moonlighters in the biofilm cells growing on both carbon sources. In addition, stress response proteins (GroES and DnaK), glycolytic proteins (PGM, PYK, PGK, FBA, GPI, L/D-LDH, GMR, and PDCE2), the elongation factors EfTS and EfG, the RNA polymerase subunit RpoA, the translation initiation factor IF1, and the trigger factor (TF), were also detected as relatively abundant on biofilms. The r-protein moonlighters (a total of 43 proteins) were the largest protein group among the biofilm-associated surfaceomes, independent of the carbon source used. In addition to their proposed role as biofilm integrity/stability enhancing role, coordinated by the production of acidic fermentation end-products (Graf et al., 2019), some of the identified r-proteins may also have other moonlighting functions. The most abundant r-proteins were RplX, RpsE, RpsG, RplL, and RplO, from which RpIL (the 50S ribosomal protein L7/L12) was predicted to be the most abundant. The high abundance of this r-protein during biofilm growth could explain the concomitant presence of another moonlighting protein, EfG, on the biofilm cells at high abundances, as the GTPase activity of EfG is reported to require RplL for ribosomal translocation (Koch et al., 2015). For several r-proteins (e.g., L5, L11, L23, L13a, S3, S19, and S27), non-canonical functions ranging from gene expression regulation to subverting pro-inflammatory actions, apoptosis and protection against abiotic stress (Liu et al., 2014; Henderson et al., 2016; Chen et al., 2018) have been proposed. The increased abundance of the TF during the biofilm mode of growth could be explained by the concomitant presence of the ribosomal protein L23. This r-protein has been reported to have binding sites for both the TF and the signal recognition particle, which is thought to aid in protein export to the cell membrane (Hoffmann et al., 2010). The specific appearance of several moonlighting aminoacyl-tRNA synthetases (ATRSs) in biofilm cells was also interesting, as these enzymes, unrelated to their primary function, also regulated gene expression, signaling, transduction, cell migration, tumorigenesis, angiogenesis, and/or inflammation (Son et al., 2014; Chen et al., 2018), thereby forming a novel group of moonlighters in probiotic species.

Several components of the proteolytic system (several OppA paralogs, PepC, PepN, and PepA) involved in the utilization of milk casein (Savijoki et al., 2011) were identified here as candidate moonlighters with higher abundances on biofilm than planktonic cell surfaces. Among these PepN and the OppA paralogs were more abundant on or specific to the frc-associated cell surfaces. The enzymatic assay performed on whole cells using a substrate specific to PepC and PepN confirmed the identification results and indicated that more PepC than PepN is associated with biofilm cells. In addition, the classical surface proteases PrtP and PrtR were predicted to be more abundantly produced on the biofilm cell surfaces. The two identified aminopeptidases are expressed in the cytoplasm, where they act on the oligopeptides that are taken up by the Opp-transport system. Thus, efficient surface-aminopeptidase activity could provide the probiotic with means to speed up growth and adaptation in conditions involving oligopeptides as the carbon source. In a recent study by Galli et al. (2019) GG was utilized as an adjunct starter culture in Camembert-type cheese production, which was found to improve the sensory characteristics of the cheese. While the mechanism behind the GG-mediated flavor formation in cheese remains to be elucidated, the efficient and diversified proteolytic system of the strain could be involved (Galli et al., 2019). It is highly likely that the robust GG cells do not undergo autolysis during cheese making, which makes the cell-surface located moonlighting aminopeptidases C and N interesting objects to be studied further in this context.

Several cytoplasmic proteins have been shown to be selectively sorted into membrane vesicles (MVs), which in some bacteria were shown to contribute to approximately 20% of the whole biofilm matrix proteome, demonstrating that MVs are also important constituents of the biofilm matrices (Couto et al., 2015). Protein targeting to MVs provides an important mechanism exploited by both gram-negative and gram-positive bacteria to export proteins in a protected and concentrated manner to aid neighboring cells or modulate the host immune system. Whether MVs play a role in L. rhamnosus GG biofilm formation and surface protein export, remain to be studied.

Comparison of the total antigenicity profiles of GG cells grown in planktonic and biofilm forms on glc and frc indicated that frc as the carbon source increases the abundances of many antigens (∼23, 28, 40, and 60 kDa in size) during both growth modes. We previously demonstrated the presence of several adhesive and antigenic moonlighters at the cell surface of planktonic GG cells grown on glc (Espino et al., 2015). In that study, protein moonlighters, such as DnaK (67.2 kDa), GroEL (57.4 kDa), PYK (62.8 kDa), TF (49.8 kDa), ENO (47.1 kDa), GaPDH (36.7 kDa), L-LDH (35.5 kDa), EfTu (43.6 kDa), Adk (23.7 kDa), and UspA (16.8 kDa), were identified as highly abundant and antigenic on planktonic GG cells. The present study suggests that PYK, ENO, GaPDH, TF, L-LDH, EfTU, UspA, DnaK, and GroEL could be produced more during biofilm than planktonic growth, which is also in line with an earlier proteomic study showing that a switch from planktonic to biofilm growth in L. plantarum increases the abundances of several stress responses and glycolytic moonlighters (De Angelis et al., 2015). From the identified moonlighters, ENO, IMP, FBA, TDPA, GreA, GroES, and GroEL are plausible factors that aid in biofilm formation in the presence of frc, as evidenced by surfaceome predictions.

The rmlABCD operon products, involved in the synthesis of O-antigen lipopolysaccharide (Reeves, 1993), could have conferred increased surface antigenicity to GG cells during biofilm growth. Here, all of the operon gene products RmlA, RmlB, RmlC, and RmlD corresponding to glucose-1-phosphate thymidylyltransferase, dTDP-4-dehydrorhamnose 3,5-epimerase, dTDP-4-dehydrorhamnose reductase and dTDP-glucose-4,6-dehydratase, respectively, were detected only in biofilm cells and with ∼2-fold higher abundances in the presence of frc. In some pathogens, L-rhamnose is required for virulence, and the enzymes of the Rml pathway are considered potential targets in drug design (Trent et al., 2006). In view of this, detection of the Rml enzymes as moonlighters is interesting and points toward a novel role of these enzymes at the cell surface of non-pathogenic bacteria. The rhamnose moiety has also been shown to bind specific moonlighters (e.g., ENO and 58 other potential moonlighters) at the bacterial cell surface, proposing that rhamnose-mediated anchoring is a general mechanism for anchoring the moonlighting proteins to the cell surface in bacteria (Daubenspeck et al., 2016).

The identified DltD, encoded by part of the dltABCD operon involved in the formation of lipoteichoic acid (LTA), could have increased antigenicity in frc-associated GG cells. DltD is a single-pass membrane protein involved in the final transfer of D-alanine residues to LTA on the outside of the cell (Reichmann et al., 2013) that contributes to adherence and biofilm formation (Spatafora et al., 1999; Gross et al., 2001). LTA has been shown to be essential for biofilm matrix assembly and bulk accumulation over time (Castillo Pedraza et al., 2017), and the early stages of biofilm formation necessitates efficient expression of dltD (Klein et al., 2012). In addition, a lack of DltD has been linked to poor acid survival in planktonic cells and an inability to form biofilms in vitro in the presence of sucrose or glucose (Quivey et al., 2015). As DltD could not be detected in the glc-associated planktonic or biofilm cells, we hypothesize that such deficiency might also explain the lower biofilm formation efficiency on glc compared to frc.

Many of the classical surface proteins may have also contributed to the total antigenicity and/or adherence in the GG cells. The most plausible candidates include lipase and lipoproteins (Aes, AbpE, Cad, and CamS), peptidoglycan hydrolases (e.g., the NlpC/P60), MBF, a lectin-binding protein, penicillin-binding proteins and the pilus protein SpaC. The identification data implies that SpaC is specifically produced during planktonic growth on frc. From these proteins, the SpaC adhesin and MBF are best known; the first is produced through the spaCBA operon coding for the pilus, the key factor promoting biofilm formation in vitro as well as colonization and adherence in vivo (Lebeer et al., 2007b, 2012), while MBF has been shown to bind intestinal and porcine colonic mucus, laminin, collagen IV, and fibronectin (von Ossowski et al., 2011; Nishiyama et al., 2015). Additional mucus-adherence assays with each GG cell type in the presence of anti-SpaC antibodies complemented with indirect ELISA monitoring of the SpaC abundances indicated the importance of the pilus adhesin in coordinating the adherence in both the planktonic and biofilm cells. Surprisingly, the glc–associated biofilm cells with the highest SpaC abundance did not shown the highest mucus binding ability, which was detected for frc grown biofilm cells. Possible explanations to this discrepancy include presence of specific factors in biofilm cells grown on frc, which could enhance or strengthen the pilus mediated mucus binding. The inability to identify SpaC by LC-MS/MS from biofilm-cells implied that the complex structure of the pilus adhesin together with the overwhelming presence of moonlighting proteins on biofilm cell surfaces could be the reason why the pilus protein was identified only from the planktonic cells. The same reason applies for the lack of MabA among the identified surfaceomes. MabA (LGG_01865), the modulator of adhesion and biofilm formation, is proposed to strengthen the biofilm structure following the pilus-mediated interaction with the surface (Velez et al., 2010).

A recent study highlighted an important role of glc and frc, carbohydrates prevalent in the Western diet, in regulating the colonization of a beneficial microbe independent of supplying these carbohydrates to the intestinal microbiota (Townsend et al., 2019). In that study, both frc and glc, but not prebiotics such as fructo-oligosaccharides, were found to silence the two-component system sensor histidine kinase/response regulator Roc that activates transcription of clustered polysaccharide utilization genes in a widely distributed gut commensal bacterium B. thetaiotaomicron. Roc was also suggested to promote gut colonization by interacting with moonlighting proteins such as GPI or glucose 6-phosphate dehydrogenase (Sonnenburg et al., 2006). From these two moonlighters, GPI was detected here as a more abundant protein in biofilm cells with a slight increase in cells grown on glc. As the present study compared surfaceomes of cells grown only in the presence of simple carbohydrates, we cannot exclude the possibility that corresponding response regulators (e.g., WalK) could have been modulated by these carbohydrates and that the two tested carbohydrates could coordinate the colonization of GG in vivo in an analogous manner.