Immunostimulation by Lactobacillus kefiri S-layer proteins with distinct glycosylation patterns requires different lectin partners

S-layer (glyco)-proteins (SLPs) form a nanostructured envelope that covers the surface of different prokaryotes and show immunomodulatory activity. Previously, we have demonstrated that the S-layer glycoprotein from probiotic Lactobacillus kefiri CIDCA 8348 (SLP-8348) is recognized by Mincle (macrophage inducible C-type lectin receptor) and its adjuvanticity depends on the integrity of its glycans. However, the glycan´s structure has not been described so far. Herein, we analyze the glycosylation pattern of three SLPs, SLP-8348, SLP-8321, and SLP-5818, and explore how these patterns impacts their recognition by C-type lectin receptors (CLRs) and the immunomodulatory effect of the L. kefiri SLPs on antigen-presenting cells. HPAEC-PAD performed after β-elimination showed glucose as the major component in the O-glycans of the three SLPs, however, some differences in the length of hexose chains were observed. No N-glycosylation signals were detected in SLP-8348 and SLP-8321, but SLP-5818 was observed to have two sites carrying complex N-glycans based on a site-specific analysis and a glycomic workflow of the permethylated glycans. SLP-8348 was previously shown to enhance LPS-induced activation on both RAW264


Introduction
Glycosylation is considered the most popular posttranslational modification targeting proteins.Despite until the mid 70´s it was assumed that the ability to glycosylate proteins was restricted to https://www.jbc.org/cgi/doi/10.1074/jbc.RA120.013934The latest version is at eukaryotic cells and archaebacteria (1), nowadays it is known that this post-translational modification is not an exception in eubacteria, although it is early to predict the full extent of prokaryotic glycosylation (2).From the studies performed on eukaryotic cells, it is clear that the presence of glycans affect expression, localization and lifetime of numerous proteins thus affecting its functions as well as several downstream biological events (3).
In eubacteria, protein glycosylation has been extensively studied in pathogens, emphasizing its relevance in virulence and pathogenicity.On the contrary, the nature and function of glycoproteins in non-pathogenic bacteria, including gut commensal species, remains largely unexplored (4).Since interactions between commensal bacteria, intestinal epithelial and immune cells play a crucial role in the maintenance of gut homeostasis, the study of protein glycosylation in gut commensal bacteria has become an expanding field of research due to the importance of the role of gut microbiota in health and disease (5,6).This is particularly relevant in the case of surface proteins, which act as mediators of several interactions between microorganisms and their host.
The S-layer is a nanostructured proteinaceous envelope constituted by subunits that self-assemble to form a two-dimensional lattice that covers the surface of different species of Bacteria and Archaea (7).Given their ubiquitous presence, S-layers are considered as the result of evolutionary changes of the microorganisms to survive in harsh environments (8) although there are no reports on common functions for all of them (9).The presence of S-layer has been found in both Gram-negative and Gram-positive bacteria, including pathogenic and non-pathogenic species (10).Regarding the members of the genus Lactobacillus, microorganisms commonly retrieved in the gut of different mammalian hosts as well as in fermented foods, the S-layer has been detected in many but not all species (11).Due to their unique self-assembly ability, presenting repetitive identical physicochemical properties down to the subnanometer scale, the S-layer proteins have gain interest in distinct areas of biotechnology, biomimetic and biomedicine.Indeed, different researchers have been conducted experiments focusing on the application of S-layer proteins for the development of new antigen/hapten carriers, adjuvants or vaccination vesicles (11,12).
After the presence of S-layer proteins in several strains of L. kefiri formerly isolated from kefir grains and characterized as potentially healthpromoting microorganisms was described (13), several studies have been performed in our laboratory to gain knowledge about structural and functional properties of these surface proteins.In particular, the S-layer proteins from L. kefiri strains show differences in their amino acid sequences, mainly in the C-terminal region of the polypeptide (14).Moreover, they are part of the small group of glycosylated S-layer proteins reported within the genus Lactobacillus (11), and until now, the detailed glycan structures have been described just for the strain L. kefiri CIDCA 83111 (15).Noteworthy, although the S-layer proteins were the first glycoproteins described in prokaryotes, the studies of the role of the glycan residues in the functional properties of these proteins are scarce and are usually focused on some archaea (16) or pathogenic bacteria (17,18).Regarding the genus Lactobacillus, there are some reports describing the involvement of the carbohydrate receptor DCspecific ICAM-3-grabbing nonintegrin (DC-SIGN) in the functional activity of the S-layer proteins from L. acidophilus NCFM (19), L. plantarum (20) and L. kefiri JCM 5818 (21).In this sense, we have recently demonstrated that the S-layer glycoprotein from L. kefiri CIDCA 8348 is able to enhance the LPS-induced response in murine macrophages in a carbohydrate-receptor mediated process (22).Moreover, this SLP improves the OVA-specific cell immune response by triggering maturation of antigen presenting cells through the recognition of glycan moieties by Mincle (23).However, the structure of its glycans has not been determined so far.
Considering all these evidences, in the present work we aim to determine the glycosylation patterns of the S-layer proteins expressed on different strains of L. kefiri and to analyze the impact on the immunomodulatory activity on antigen presenting cells.

Glycan structures of L. kefiri S-layer proteins
We have previously reported the composition and structure of the glycans present in the S-layer protein from L. kefiri CIDCA 83111 (15).In this work, glycosylation of the S-layer protein from L. kefiri JCM 5818, CIDCA 8348 and CIDCA 8321 (SLP-5818, SLP-8348 and SLP-8321, respectively) were characterized.In this regard, the analysis of both O-and N-glycans was performed.

O-glycome analysis
The SDS-PAGE bands corresponding to the purified SLPs of each strain were subjected to reductive β-elimination.Sugar composition was achieved by HPAEC-PAD.For this purpose, the released oligosaccharides were hydrolyzed, and monosaccharides were determined.In the three SLPs examined glucose was detected as the major component (Figure 1A, B and C).Also, in the three cases, when an alditol analysis was performed, sorbitol was detected.Taking into account that during the treatment the oligosaccharides linking the peptide are released and the reducing end of the oligosaccharides are reduced by NaBH4 to an alditol, the result indicates that a glucose unit is the linkage between the O-linked oligosaccharides and the peptides.(Figure 1D, E and F).
The O-glycoproteomic analysis of the SLPs was performed by nanoHPLC-ESI-Orbitrap-HCD.An in-gel trypsin digestion of the purified glycoprotein from each strain was performed followed by a glycopeptide enrichment step using cotton HILIC chromatography.In SLP-5818, the analysis of the extracted ion chromatogram showed a defined region of O-glycosylation detected by EICs of oxonium ions where up to four hexose units were detected (m/z =163.06;325.11, 487.16, 649.21).It is known that glycosidic bonds are more labile under the HCD dissociation than peptides.However, notably in this case, a high coverage level of peptide fragmentation was achieved by detecting b and y ions from the glycosylated peptide.Moreover, peptide-glycan fragments (Y ions) and the ion corresponding to the naked peptide were also detected (Figure 2A).Deconvolution of the ions allowed the attribution of the signals eluting at retention time between 10-14 min to peptide 147 SASASSASSASSAEQTTALTDAQK 170 carrying from 10 to 35 hexoses.Different glycoforms were resolved in the operating conditions of the RP nanoHPLC employed.It is clearly evidenced that the chromatographic elution of O-glycopeptides is based on the hydrophilic nature of the attached glycans, therefore on their decreasing size, from the largest to the smallest one.Interestingly, it was also possible to detect glycosylation from just one hexose unit co-eluting with the non-modified peptide (Figure 2B, Table 1), indicating that this O-glycosylation site is not fully occupied in this protein.
On the other hand, it was interesting to detect in EICs from both, SLP-8321 and SLP-8348 a region where several clusters of ions with charge +4, +5, +6 and +7 corresponding to peptide 125-170 containing the SASSASASA motif bearing up to 26 hexose units (Figure 3A, B, Tables 2 and 3).MS/MS spectrum of the different ions showed band y-peptide fragments, confirming the peptide identification (Figure 3C, D).

L. kefiri JCM 5818 N-glycome analysis
Further on, taking into account that the S-layer protein of L. kefiri 83111 presented N-glycosylation (15) we searched for diagnostic oxonium ions for HexNAc + (m/z 204.09, m/z 168.07 and m/z 138.05) in the MS/MS spectra.No signals were detected for SLP-8348 and SLP-8321, however SLP-5818 evidenced the presence of the reporter ions.Therefore, we first tried an affinity-based purification of the SLP of L. kefiri JCM 5818 using immobilized Wheat Germ Agglutinin (WGA) to evidence this type of posttranslational modification.Interestingly, although whole protein fraction showed that the purification was not complete (Figure 4A, lane 2), the SDS-PAGE profile from the lectin-bound fraction showed a protein band with MW 67 kDa in accordance with the reported MW of the SLP-5818 (24,25) (Figure 4A, lane 1).This band was excised, trypsin-digested and subjected to mass spectrometry analysis confirming the S-layer protein identity and thus, the presence of Nglycosylation in this protein (Supplemental 1 shows protein identification data).
In order to determine the N-glycosylation sites in SLP-5818, we used PNGase F to release the N-glycans.It is known that the glycosylated asparagine residue is converted to aspartic acid during the enzymatic reaction (26).As a result, a mass increase of 0.984 Da is obtained in the modified peptide.Based on this strategy, a sample of enriched glycopeptides from SLP-5818 was Ndeglycosylated using PNGase F and analyzed by nanoHPLC-ESI-Orbitrap.Due to the fact that spontaneous deamidation might occur, the ratio asparagine/aspartic acid was considered in order to avoid misassignments (Supplemental 2 shows peptide identification data).Thus, two Nglycosylation sites at NVAVN 55 GTNALYTK and at DAALTSLN 457 NTVK were determined.As shown in Figure 4B, MS2 spectrum of peptide NVAVN 55 GTNALYTK showed fragment ion y9 with a m/z = 982.48(calc.m/z = 981.50, 0.984 Da) confirming the N-glycosylation site at N 55 .In the same way, MS2 spectrum of peptide DAALTSLN 457 NTVK showed y5-fragment ion m/z = 576.30(calc.m/z = 575.31, 0.984 Da) evidencing an N-glycosidic linkage in N 457 (Figure 4C).Further on, in order to determine the N-glycan structures attached to SLP-5818, we performed an enzymatic hydrolysis of the N-oligosaccharides using PNGase F. Glycans were separated, permethylated and analyzed by nanoHPLC-ESI-Orbitrap.
The extracted ion chromatograms corresponding to m/z = 464.24(HexNAc-Hex + ), and m/z = 189.11(dHex + ) allowed to detect a family of complex permethylated N-glycans eluting between Rt 40-45 min (Figure 5A).Main structures, m/z and mass errors are detailed in Table 4. Figure 6 shows two examples of MS2 spectra of selected ions assigned to two bi-antennary glycans differing in the composition of one antennae.The permethylated N-glycans were also examined using MALDI-TOF MS and structures corresponding to bi-antennary species were also detected (Figure 5B).

Stimulation of murine macrophages by L kefiri SLPs
Regarding that we have previously demonstrated that SLP-8348 can enhance LPS-induced activation of antigen presenting cells (22,23), we decided to test the ability of SLP-8321 and SLP-5818 to activate murine macrophage, using RAW 264.7 cells as first model.As it was previously observed for SLP-8348 (22,23), the SLP-8321 and SLP-5818 were not able to induce activation of RAW 264.7 cells by themselves.However, respective significant increments in secreted IL-6 as well as in surface expression of CD40 and CD86 were observed for macrophages simultaneously exposed to E. coli LPS and L. kefiri S-layer glycoproteins in comparison to LPS-treated cells (Figure 7A, B).To note, no significant differences were observed among the stimulatory activity of strains despite the differences in their glycosylation patterns.

Interaction of SLPs with different C-type lectin receptors
Considering that it has been previously shown for SLP-8348 that the recognition of the glycan residues by a C-type lectin receptor (CLR) is a key event in the immunomodulatory activity of the protein (23), we decided to test the ability of SLP-8321 and SLP-5818 to interact with different CLRs.As shown previously for SLP-8348 (23), SLP-8321 and SLP-5818 bound to the CLRs DC-SIGN, SingR3, Langerin and Mincle (Figure 8A).As expected, a reduction of approximately 70% was observed in the SLP-CLR interactions in the presence of the Ca +2 chelating agent EGTA (Figure 8B).

Activation of BMDC by SLP-8321 and SLP-5818
Regarding the ability of SLP-8321 and SLP-5818 to interact with different CLRs and considering that both glycoproteins enhanced LPS-induced response in murine macrophages, we decided to use BMDC from C57BL/6 mice as antigen presenting cells to assess both internalization of SLPs and cell activation.As shown for SLP-8348 (23), we observed that both SLP-8321 and SLP-5818 were internalized by BMDC in a dose-dependent way (Figure 9A).However, a significant decrease in the uptake of SLP-8321 was observed when BMDC from Mincle -/-mice were used.On the contrary, the internalization of SLP-5818 was only reduced in BMDC from SignR3 -/-mice and was not affected by the absence of Mincle (Figure 9A).
Following the same methodology previously reported (23), we decided to test the immunomodulatory activity of SLP-8321 and SLP-5818 on BMDCs.Unlike in the case of RAW 264.7 cells, both SLPs were able to induce the maturation of BMDC since a significant increment in the level of secreted IL-6 and TNF- as well as in the surface expression of CD40 and CD80 were observed (Figure 9B, C).The same effect was observed when BMDC were stimulated by a combination of SLPs and LPS.The stimulatory activity of SLP-8321 was lost when BMDC from Mincle -/-mice were used, whereas SLP-5818 was not able activate BMDC from SignR3 -/-mice (Figure 9B, C).
To further investigate the T-cell stimulatory function of the SLP-activated BMDCs, OVAtreated BMDCs from wild type, Mincle -/-, SignR3 -/- and CARD9 -/-were incubated in the presence or absence of SLP-8321 or SLP-5818 with purified T cells from OT-II mice, as was previously performed with SLP-8348 (23).Both SLPs could enhance the OVA-specific T cell response since a higher IFN- secretion as well as higher CD4 + T cell proliferation were observed upon simultaneous stimulation of wild type BMDCs with OVA and SLP-8321 or SLP-5818 compared to OVA-treated antigen presenting cells (Figure 10A,B).The absence of Mincle abrogated the immunostimulatory activity of SLP-8321 while SLP-5818 was not able to enhance CD4 + T cell response when BMDCs from SignR3 -/-mice were used (Figure 10A,B).As was expected, none of the SLPs assessed were able to potentiate CD4 + T cell activation when assays were performed using BMDC from CARD9 -/-mice, a downstream signaling protein of Mincle and SignR3 (27) (Figure 10A, B).

Discussion
Considering the ubiquitous presence of S-layercarrying microorganisms and the abundance of the S-layer proteins, it is evident that these structures reflect the evolutionary adaptation of the organisms to natural habitats and must have provided them with advantages in specific environmental and ecological conditions (8).Glycosylation is the posttranslational modification most frequently found in SLPs, a feature shared with other surface exposed proteins such as flagellin and pilin.Within the genus Lactobacillus, the glycoprotein nature of the S-layer proteins has been reported for strains of L. buchneri (28,29), L. plantarum 41021/252 (28) L. acidophilus NCFM (19), L. acidophilus ATCC 4356 (30), and several strains of L. kefiri (14,24).In this last species, we have previously described the composition and structure of the O-and N-glycans present in the SLP of the strain L. kefiri CIDCA 83111 (15).Herein, we show that SLP-5818 also presents O-and N-glycosylated chains.In this case, peptide 147-170 is O-glycosylated in average with eight hexose units, but at difference with the SLP from L. kefiri CIDCA 83111, is not decorated with galacturonic acid (15).Regarding N-glycans, both SLP 5818 and SLP 83111 present two Nglycosylation carrying bianntenary chains.On the other hand, SLP-8321 and SLP-8348 show 100% of homology at primary sequence level (14) and despite they present six potential Nglycosylation sites, no evidence of bearing Nglycans was found.On the contrary, O-glycosidic chains constituted by 4 to 12 glucose units were detected in both SLPs.
Inter-strain differences in the glycosylation patterns of SLP were also seen some years ago in Thermoanaerobacter thermohydrosulfuricus and Thermoanaerobacterium thermosaccharolyticum (31).Moreover, in Clostridium difficile the presence of a glycosylated SLP was only demonstrated in the strain Ox247 which encodes a putative glycosylation locus (18), while normally most C. difficile strains do not possess a glycosylated S-layer (32).On the other hand, Anzengruber and coworkers reported that SlpB from L. buchneri CD034, SlpN from L. buchneri NRRL B-30929 and L. buchneri SlpB from 41021/251 showed identical O-glycans consisting of on average seven glucoses at four serine residues within the sequence S152-A-S154-S155-A-S157 without any evidence of the presence of N-glycosidic chains in those SLP (29).In this sense, SLP-8321 and SLP-8348 show a higher similarity to the SLPs from L. buchneri strains than SLP-83111 and SLP-5818.Further studies are needed in order to elucidate the mechanisms through which the glycan chains are coupled to the peptide backbone in these surface proteins.
As surface exposed structures, the SLPs could be a key mediator between bacteria and host immune system contributing to the shaping of the immunological response to pathogenic, commensal and probiotic microorganisms.By using murine macrophages RAW 264.7 cells as model, we have demonstrated that despite the differences between glycosylation patterns of SLP-8321 and SLP-5818, both proteins are able to enhance LPS-induced cell activation (Figure 7), similarly to previously demonstrated for SLP-8348 (22).These results suggest that O-glycans present in these three mentioned L. kefiri SLPs could be responsible, at least in part, of their immunostimulatory ability.
On the other hand, as previously described with SLP-8348 (23), the assays using BMDCs from Mincle -/-and SignR3 -/-mice allowed us to determine that the internalization of SLP-8321 and SLP-5818 as well as their ability to activate BMDCs were mainly mediated by Mincle and SignR3 (the murine ortholog of human DC-SIGN), respectively (Figure 9), which could be expected considering the differences in the glycosidic moieties present in those proteins.Moreover, the incubation of BMDCs with SLPs from different L. kefiri strains leading to the enhancement of activation of OVA-specific CD4+T cells from OT-II mice, is also dependent of the presence of Mincle and SignR3 for SLP-8321 and SLP-5818, respectively (Figure 10).Some years ago, a pioneering work performed using a SlpA-knockout mutant L. acidophilus NFCM revealed that recognition of SlpA by DC-SIGN in presence of LPS induce an anti-inflammatory response in human monocyte-derived dendritic cells (19) suggesting the involvement of glycans in the immunomodulatory properties of SLPs.More recently, it was demonstrated that L. acidophilus NCK2187, a strain which solely expresses SlpA, and its purified SlpA binds to SignR3 to trigger regulatory signals that result in mitigation of experimental colitis, maintenance of healthy gastrointestinal microbiota, and mice protection (33).In addition, the ability of DC-SIGN to recognize the SLP-5818 was recently suggested by Prado Acosta and colleagues (21), which agrees with our findings.However, the role of glycan residues as well as the CLR involved in the internalization and immunostimulatory activity of the SLP-5818 had not been reported so far.On the contrary, is not surprising that Mincle, the CLR responsible for the recognition and adjuvanticity of the SLP-8348 (23), mediate the immunostimulatory effect of SLP-8321 on BMDCs since both glycoproteins present not only the same amino acid sequence (14) but also the same O-glycans.It is known that Mincle recognizes diverse sugar-containing ligands including trehalose dimycolate glycolipid from mycobacteria, mannose-, glucoseor fucose-containing glycoconjugates, and Lewis antigen from Helicobacter pylori LPS (34).To note, it was recently reported that Mincle also recognize the O-linked oligosaccharides of the SLP from the Gram-negative oral pathogen Tannerella forsythia, and that interaction induces both pro-and anti-inflammatory cytokine secretion in macrophages (17).
To resume, despite the differences in glycosylation patterns as well as the involvement of different CLRs, the glycosylated SLPs from distinct L. kefiri strains can induce activation of antigen presenting cells through the recognition of their glycans.Taken together, these findings encourage us to further investigate the potential application of these surface proteins derived from probiotic bacteria to the development of new adjuvants or carrier for vaccine antigens as well as to get deeper insight the protein glycosylation mechanisms in these bacteria.

S-layer protein extraction
Extraction from bacterial cells at stationary phase was performed using 5M lithium chloride (LiCl) as described previously (22).The samples were centrifuged and the protein concentration in the supernatant was determined according to Bradford.The homogeneity of the protein extracts was tested by SDS-PAGE, stained with Coomassie blue (36).S-layer proteins were filtrated through a membrane of 0.22 µm pore diameter for cellular stimulation assays.Lectin-based N-glycoprotein purification.The enrichment of glycoproteins was performed using Wheat Germ Agglutinin (WGA) lectin affinity chromatography.Briefly, 1 mL of WGA -Sepharose (Sigma) was loaded onto a 5 mL column.Prior to sample purification, stationary phase was equilibrated with 20mM Tris pH 7.5, 0.5M NaCl, 1mM CaCl2, 1mM MgCl2.Then, 50 L of protein sample was loaded using the latter buffer solution.Non-glycosylated proteins were washed using 20mM Tris pH 7.5 containing 1M NaCl, 1mM CaCl2, 1mM MgCl2.Finally, the retained glycoproteins were eluted using the above buffer solution containing 0.5M N-Acetyl-D-Glucosamine.The glycoprotein fraction was dialyzed against ultra-pure water using Spectra/Por ® (Repligen) molecular porous membrane tubing cut-off 3.5 kDa.The lectin-based purification was monitored by SDS-PAGE and stained with Colloidal Coomassie Blue.

Release of N-glycosidic chains by PNGase F treatment.
The SDS-PAGE S-layer glycoprotein band was cut out from the gel, frozen for 3 h, and successively washed with (a) acetonitrile, (b) 20 mM NaHCO3, pH 7, and (c) acetonitrile.The gel pieces were dried and the N-glycans were released by incubation with PNGase F (20 milliunits) (New England Biolabs Inc., Beverly, MA) overnight at 37 °C in 20 mM NaHCO3, pH 7. The gel pieces were thoroughly washed, the supernatants were separated and filtered through an Ultrafree McFilter (Mr 5000), dried, resuspended in 0.1% (v/v) formic acid (20 μL), and left at room temperature for 40 min.Finally, glycans were dried and suspended in water.
Glycoprotein digestion.The protein band corresponding to the S-layer glycoproteins were cut out from the gel and washed with acetonitrile.The gel pieces were reduced with 10 mM DTT in 50 mM NH4HCO3 at 55 °C for 30 min.They were further washed with acetonitrile and alkylated with 55 mM IAA in 50 mM NH4HCO3 for 20 min at room temperature in darkness.After washing the gel pieces with 50 mM NH4HCO3 for 10 min and with acetonitrile for 5min, they were dried in a SpeedVac and submitted to digestion with 20 ng/μL trypsin (Sigma) in 40 mM NH4HCO3, 9 % acetonitrile at 37 °C overnight.Afterwards, the supernatants were separated and dried.
In gel-reductive β-elimination and sugar analysis.O-linked glycans were released from each SDS-PAGE band of the S-layer glycoproteins using 0.05M NaOH / 1M NaBH4 and incubated overnight at 50°C.The solution was separated, acetic acid was added until pH 7 followed by repeated evaporation with methanol.The sample was dissolved in water, desalted in a Dowex 50 W (H + ) (Fluka) column and dried in SpeedVac.The β-eliminated samples were further hydrolyzed in 2 N TFA for 4 h at 100 °C.The acid was eliminated by evaporation and the hydrolysates were re-suspended in water for HPAEC-PAD (high performance anion exchange chromatography).Analysis was performed using a DX-500 Dionex BioLC system (Dionex Corp.).For neutral and amino sugars, a CarboPac PA20 with the corresponding guard column was used.Isocratic separations were achieved with 18mM NaOH and flow rate set to 0.4 mL/min.For alditols, a CarboPac MA 1 column was used with a 0.4 M NaOH isocratic program and a flow rate of 0.4 mL/min.Permethylation of released oligosaccharides.After PNGase digestion, the oligosaccharides were methylated with NaOH in DMSO followed by addition of CH3I as described (37).
Glycopeptide enrichment.Sample purification and glycopeptide enrichment was achieved by solidphase extraction Cotton-HILIC microtips as described (38).Briefly, HILIC microtips were equilibrated with three volumes of 85% ACN.Dried sample was resuspended in 85% acetonitrile.Prior to SPE clean-up step, sample was loaded onto the stationary phase by aspirating and dispensing the protein mixture.Peptides were washed off from the column with 85% ACN, 0.5% TFA, and the retained glycopeptides were eluted using water.The glycopeptide fraction was freeze-dried and stored at -20°C until mass spectrometry analysis.
Data Analysis.Proteome Discoverer 2.1 (Thermo Scientific) assisted protein identification.The mass tolerances for MS and MS/MS were 10 ppm and 0.02 Da respectively.Missed cleavages allowed for tryptic digestion was 1. Carbamidomethylation of cysteine residues was set as static modification and oxidation of methionine as dynamic modification.For N-glycosite analysis conversion of asparagine to aspartic acid was included as dynamic modification.Precursor mass tolerance was 10 ppm, and fragment mass tolerance was 20 ppm.Confident identification for peptides was set to a score > 100.Glycopeptide search was performed with MaxQuant version 1.6.5,Byonic™ from ProteinMetrics, Proteome Discoverer 2.2 and using a semi-automatic home-made software.(A0A1C3S3I4 S-layer protein OS=Lactobacillus kefiri OX=33962 GN=S-layer A0A1C3S3T6 Slayer protein OS=Lactobacillus kefiri OX=33962 GN=S-layer. All glycopeptides identifications from both searches were manually curated.MS-Convert Tool (from ProteoWizard platform tools) converted raw data from enriched glycopeptides into text files for manual annotation.Deconvolution was assisted by Xtract on Thermo Xcalibur 3.0.63.

C-type Lectin Receptor recognition of S-layer proteins
The CLR-reactivity on S-layer proteins from L. kefiri CIDCA 8321 (SLP-8321) and L. kefiri JCM 5818 (SLP-5818) was evaluated by an ELISA type assay as described previously (23,39).A half-area microplate (Greiner Bio-One GmbH, Frickenhausen, Germany) was coated with 0.25 µg of SLP per well for 16 h at 4ºC and blocked with 1%BSA (Thermo Fisher Scientific, Darmstadt, Germany) for 2 h at room temperature (RT).Then, 0.25 µg of each CLR-hFc fusion protein in lectinbinding buffer (50 mM HEPES, 5 mM MgCl2, and 5 mM CaCl2) was added and incubated for 1 h at RT.For inhibition assays, CLR-hFc fusion proteins were pre-incubated with 5mM of EGTA (Sigma-Aldrich, USA).After washes, HRP-conjugated anti-human IgG anti-body (Dianova) was added to each well for 1 h at RT. Finally, plates were incubated with chromogenic substrate (ophenylenediamine dihydrochloride substrate tablet (Thermo Fisher Scientific), 24 mM citrate buffer, 0.04% H2O2, 50 mM phosphate buffer in H2O).The reaction was stopped with 2.0 M sulfuric acid and the product was read at 495 nm using a Multiskan Go microplate spectrophotometer (Thermo Fisher Scientific).

Mice
The source of the CLR-and CARD9-deficient mice was described previously (40,41,42).Mincle −/− , SignR3 −/− and CARD9 −/− mice were backcrossed on C57BL/6 background over ten generations in the animal facility of the Federal Institute for Risk Assessment (Berlin, Germany).All mouse lines (including wild-type and C57BL/6-Tg (TcraTcrb)425Cbn/J (OT-II) mice) were kept in the animal house of the University of Veterinary Medicine (Hannover, Germany) with water and food supplied ad libitum.Mice were sacrificed for the isolation of spleen cells (OT-II transgenic mice) or the preparation of bone marrow for BMDC generation (Mincle −/− , CARD9 −/− , SignR3 −/− , and WT mice).Sacrificing of mice for scientific purposes was approved by the Animal Welfare Officers of the University of Veterinary Medicine Hannover (AZ 02.05.2016).

BMDCs/T cell co-culture assay
Splenocytes were isolated from OT-II transgenic mice by flushing the spleen with complete IMDM medium.After erythrocyte lysis, cells were resuspended in MACS buffer (0.5% BSA, 2 mM EDTA in PBS).T cells were obtained by magneticactivated cell separation (MACS) using Pan T cell isolation kit II, mouse (Miltenyi Biotech, Bergisch Gladbach, Germany) according to the manufacturer's instructions.T cells were labelled with CFSE (carboxyfluorescein diacetate succinimidyl ester, Sigma-Aldrich,USA) and seeded on a 96-well round bottom culture plate (7x10 4 cells/well).After 30 minutes, BMDCs treated with OVA (Endo Grade® Ovalbumin, LIONEX GmbH, Germany) or OVA/SLPs were added and incubated for 72 h.Culture supernatants were collected for IFN-γ secretion.Proliferation index of CD4 + T cells within each experimental group was calculated as the ratio between the percentage of CFSE low CD4 + cells and CFSE low CD4 + cells from OVA-treated BMDCs.

Cytokine quantification in culture supernatants
Production of IL-6 by macrophages was analyzed by ELISA from BD-Pharmingen (San Diego, USA).Secretion of IL-6 and TNF-α by BMDCs and IFN-γ from purified T cells were analyzed by ELISA from PeproTech (USA).The assays were performed according to the manufacturer's instructions.After determining optical densities, cytokine levels in cell culture supernatants were calculated using the GraphPad Prism 7.0 program.

Immunocytostaining and flow cytometry
After stimulation experiments, cells were washed twice with PBS containing 2% (v/v) FBS and then labelled with specific monoclonal antibodies for 30min at 4•C.Cells were washed twice and then fixed with 1% (v/v) formaldehyde.Cells were analyzed using a FACSCalibur Analyzer (BD Biosciences) or Attune NxT Flow Cytometer (Thermo Fisher Scientific).Data analysis was performed using the FlowJo Software (FlowJo, Ashland, OR, USA).

Statistical analysis
Values from at least three independent experiments were analyzed by using a one-way or two-way ANOVA with Tukey's post hoc test (P<0.05was considered statistically significant).Statistical analysis was performed with the GraphPad Prism program (GraphPad Software, San Diego, California, USA).

Data availability:
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [44] partner repository with the dataset identifier PXD020358.

Authors contribution
MM performed the immunomodulatory assays, analyzed data, contribute with manuscript´s writing and revised the manuscript; GC performed the structural analysis, wrote the manuscript and revised it; AC contributed with the structural determination; BL conceived and supervised immunomodulatory assays, and revised the manuscript; MAS conceived the work, supervised immunomodulatory assays, and wrote the manuscript; and ASC conceived the work, supervised the structural analysis, and wrote the manuscript.          .

Figure 2 .
Figure 2. O-glycopeptide analysis by nanoHPLC-ESI-Orbitrap of SLP-5818.A, Selected MS2 spectra, m/z 1537.8352.(y and b peptide fragments and Y glycan-peptide fragments are indicated).B, Full MS spectrum at retention time 19.89 min showing signals corresponding to 147-170 unmodified peptide and to the same peptide bearing one or two hexose units.

Figure 4 .
Figure 4. N-Glycosylation analysis of SLP-5818.A, WGA-Sepharose analysis: lane 1: Lectin bound glycoprotein, lane 2: SLP extract, lane 3: MW markers.MS2 spectra of enzymatically deglycosylated peptides: B, peptide 51-63 and C, peptide 450-461, containing the 0.9846 Da mass increase derived from Asn/Asp conversion.Differences between the in silico digested peptides (b and y ions in the inset) and the experimental masses are highlighted in blue.

Figure 7 .
Figure 7. Modulation of LPS-induced maturation of macrophages by L. kefiri SLP-8321 and SLP-5818.A, IL-6 concentration (pg/mL) by capture ELISA in the supernatant of the murine RAW 264.7 cultures after 24 h of stimulation.B, Percentage of CD40+ and CD86+ RAW 264.7 cells after 24 h of stimulation.A representative figure of three independent experiments is shown (±SD, indicated by error bars).*P< 0.05; n.s: not significant.

Figure 8 .
Figure 8. C-type Lectin Receptor recognition of SLP-8321 and SLP-5818.A, Reactivity of CLR-hFc fusion proteins on SLP was evaluated by an ELISA type assay (using the same experimental model as in Ref. 23).B, Inhibition assays were performed using EGTA, a Ca 2+ chelating agent.These are representative figures of three independent experiments (±SD, indicated by error bars).

Figure 9 .
Figure 9. SLP-8321 and SLP-5818 are recognized by specific CLRs expressed on BMDCs.A, Binding (4°C, squares) and internalization (37°C, circles) of Atto 647N-labeled-SLP at different concentrations on CD11c+ cells from C57BL/6 and CLR-deficient mice.B, Level of IL-6 and TNF- in culture supernatant of BMDCs from C57BL/6 and CLR-deficient mice after stimulation with SLPs, LPS or the combination of both for 24h.C, Cell surface expression of CD40 and CD86 on CD11c+ cells from C57BL/6 and CLRdeficient mice after stimulation with SLPs, LPS or the combination of both for 24h.A representative figure of three independent experiments is shown, following the same experimental lines described in Ref. 23 (±SD, indicated by error bars).*P<0.05;n.s: not significant; KO: knock-out.

Figure 10 .
Figure 10.CLR-SLP specific interactions impact on T-cell stimulatory function of the SLP-activated BMDCs.Cells from C57BL/6 and CLR-deficient mice were stimulated with SLP, OVA or the combination of both for 24h and then exposed to CFSE-labelled CD4+ T cells from OT-II mice using the same experimental model as in Ref. 23. A, IFN- level in supernatants measured after five days.B, Proliferation index of CD4+ T cells.A representative figure of three independent experiments is shown (±SD, indicated by error bars).*P<0.05;n.s: not significant; KO: knock-out.

Table 4 :
Summary of main N-glycan structures determined in L. kefiri JCM 5818 surface layer protein.