Biological functions of Lacticaseibacillus rhamnosus JB-1 membrane vesicles: lipoteichoic acid, immune activity, and gut epithelial endocytosis

Intestinal bacteria have diverse and complex inuence on their host. Evidence is accumulating that this may be mediated in part by bacterial extracellular membrane vesicles (MV), nanometer-sized particles important for intercellular communication. Little is known about the composition of MV from gram-positive benecial bacteria nor how they interact with intestinal epithelial cells (IEC). Here we demonstrate that MV from Lacticaseibacillus rhamnosus JB-1 are endocytosed in a likely clathrin-dependent manner by both mouse and human IEC in vitro and by mouse IEC in vivo. We further show that JB-1 MV contain lipoteichoic acid (LTA) which activates Toll-like receptor 2 (TLR2) and induces immunoregulatory interleukin-10 expression by dendritic cells in an internalization-dependent manner. By contrast, neither LTA nor TLR2 appear to be required for JB-1 MV endocytosis by IEC. These results demonstrate a novel mechanism by which bacterial MV can inuence host physiology and suggest one potential route for benecial inuence of certain bacteria and probiotics.

Introduction associated polysaccharide A. Similarly, we have previously shown that the gram-positive bacterium Lacticaseibacillus rhamnosus JB-1 (JB-1; recently reclassi ed from Lactobacillus 14 ) and its MV can promote the number and functions of regulatory T cells, activate TLR2, and induce an immunoregulatory phenotype in dendritic cells (DCs) [15][16][17] . Substantial evidence now exists that MV from bene cial bacteria can independently in uence the host 18 , though their mechanisms of action in many cases remain unclear.
Bacteria express conserved features (microbe-associated molecular patterns; MAMPs) that are recognized by pattern recognition receptors (PRRs) of many host cells, including epithelial, endothelial, and immune 19 . These MAMPs are thought to be important signalling molecules also present on MV, which can modulate responses in host intestinal epithelial cells (IEC) and immune cells 9 and may be involved in their immunoregulatory in uence. Indeed, interactions between PRRs and MAMPs from commensal microbes are necessary to maintain homeostasis between host and microbiota, and thus contribute to healthy host development and immune responses 20 . Whether MAMPs are involved in activity associated with gram-positive MV is unknown.
Since we remain uncertain as to the exact mechanisms whereby gut bacteria in uence the host, we sought to extend our previous work with JB-1 MV to learn more about their interaction with the gut epithelium and to further characterize their in uence on dendritic cells. Here, we show that MV from JB-1 are endocytosed in a likely clathrin-mediated manner by both mouse and human IEC lines in vitro and by mouse IEC in vivo, using OMV from Escherichia coli Nissle 1917 (EcN) as a positive control as they are known to be endocytosed by IEC 21 . We further demonstrate that JB-1 MV contain lipoteichoic acid (LTA), which is recognized by TLR2 and induces interleukin 10 (IL-10) expression in DCs in vitro. Moreover, our data suggest that internalization of JB-1 MV by DCs is involved in their induction of IL-10 expression.

L. rhamnosus JB-1 MV are internalized by gut epithelial cells in vitro
To test whether JB-1 MV are internalized by cells in vitro, we rst examined whether CFSE-labelled MV and OMV were successfully stained and retained activity by incubating them with bone marrow-derived dendritic cells (BMDCs) for 1 h. We assessed MV and OMV uptake by measuring CFSE signal in ow cytometry and found that > 85% of BMDCs were CFSE-positive for both EcN OMV and JB-1 MV ( Supplementary Fig. S1), consistent with phagocytosis and suggesting that vesicles remained immunologically active after CFSE labelling.
Since EcN OMV are known to be endocytosed by HT-29 cells (a human IEC line) 21 , we used this as a positive control in exploring whether JB-1 MV would also be internalized by IEC in vitro. We incubated HT-29 cells with CFSE-labelled JB-1 MV or EcN OMV and measured internalization by association with CFSE uorescence. Flow cytometry analyses revealed > 96% of cells were positive for MV-related uorescence (Fig. 2a, left panels) and clear puncta were visible within cells when viewed with uorescence microscopy (Fig. 2a, right panels). We repeated these experiments using the mouse duodenal cell line MODE-K, and again found that both JB-1 and EcN MV were internalized to similar extents (Fig. 2b). These MV appear to be intracellular as CFSE uorescence was consistently found near nuclei and dispersed through the cell when examined by z-stacking (Fig. 2c).

L. rhamnosus JB-1 MV are internalized by gut epithelial cells in vivo
Given that MV are internalized in vitro, we wished to see if this also occurred in vivo. We gavaged BALB/c mice with CFSE-labelled JB-1 MV, then collected jejuna after 2 h. To differentiate IEC from phagocytes, we physically separated and independently isolated IEC and lamina propria cells by differential gradient separation. By ow cytometry we identi ed IEC as CD45-negative (marker of differentiated hematopoietic cells) and A33-positive (intestinal epithelial cell marker 22 ). We identi ed lamina propria DCs by positivity for CD11c and MHC II. Analyses of these populations indicated appreciable and similar uorescence in both IEC (Fig. 2d) and lamina propria DCs (Fig. 2e), suggesting that each internalized MV. This is consistent with active internalization of JB-1 MV by IEC and DCs in vivo.
L. rhamnosus JB-1 MV are likely internalized by clathrin-mediated endocytosis OMV may be internalized by epithelial cells through a variety of mechanisms 23 . When this work was undertaken, one published study had shown that MV from the gram-positive bacterium Staphylococcus aureus are internalized by HeLa cells via cholesterol-dependent membrane fusion 24 . Since EcN OMV are internalized by HT-29 in a clathrin-dependent manner 21 , we tested if the same mechanism is active here and focused on JB-1 MV.
We pre-incubated cells with the dynamin inhibitor dynasore, which blocks phagocytosis 25 and clathrinmediated endocytosis 26 , then added CFSE-labelled MV as before. We encountered an unexplained interaction between dynasore and CFSE causing an artifactual increased uorescent signal that we could not prevent. We therefore labelled the MV with DiO, a lipid-soluble membrane marker previously used with dynasore in analogous experiments 27 . DiO-labelled MV were internalized by 80% of BMDCs, and this was prevented by preincubation with dynasore (Fig. 3a). Similarly, both HT-29 (Fig. 3b) and MODE-K cells (Fig. 3c) were prevented from internalization of DiO-labelled JB-1 MV by dynasore, suggesting that internalization of JB-1 MV by IEC is an active and likely clathrin-mediated process.
L. rhamnosus JB-1 MV contain immunologically active lipoteichoic acid Recent work suggests that MV from some lactic acid bacteria contain LTA 28 , a known ligand for TLR2 that can induce IL-10 production by DCs 29 . As we previously showed JB-1 MV to have these same effects 15 , we questioned whether LTA could mediate them and additionally serve as a ligand to induce receptor-mediated endocytosis. Using western blot, we rst demonstrated the presence of LTA in JB-1 and its MV (Fig. 4a). We then performed antibody neutralization experiments to determine whether LTA is involved in MV-related effects in vitro. Anti-LTA antibodies, but not isotype control, inhibited MV interaction with TLR2 in a reporter cell assay (t = 10.1, d = 2.0, p = 0.0048; Fig. 4b). They also inhibited internalization of DiO-labelled JB-1 MV to a similar extent as dynasore (Fig. 4c), and abolished MVinduced production of IL-10 ( Fig. 4d). Interestingly, decreased IL-10 correlated with decreased internalized MV in experiments with both anti-LTA and dynasore (Fig. 4e), suggesting that internalization of whole MV is involved in the induction of IL-10.
Given that LTA mediates immune phenotypic change in DCs, we tested possible involvement of LTA in internalization by MODE-K cells but found no effect of anti-LTA antibodies ( Supplementary Fig. S2). We further attempted to block the mouse pattern recognition receptors TLR2 and SIGN-R1 with neutralizing antibodies as we have done previously with JB-1 MV in DCs 15 , but again found no effect ( Supplementary  Fig. S2). This suggests that some other ligand-receptor systems are involved in inducing clathrinmediated endocytosis of JB-1 MV.

Discussion
Membrane vesicles are promising mediators of bacterial-host communication because they enable the delivery of diverse signaling molecules, including proteins, lipids, carbohydrates, and nucleic acids, to diverse recipient cells, potentially allowing for more complex signalling and protecting contents from degradation 12 . Though well-studied in multicellular eukaryotes, only recently have MV been considered as possible mediators of communication between bene cial bacteria and their host.
MV from intestinal microbes are thought to affect the local gut environment. Most work thus far has examined the role of MV in pathogenic effects of especially gram-negative bacteria 23 , while more recent work has considered the interaction between intestinal epithelial cells and MV from bene cial bacteria. Some of the rst work to this end focused on the gram-negative bene cial bacterium E. coli Nissle 1917 (EcN), whose OMV were shown to be endocytosed by IEC in a clathrin-dependent manner 21 . While the current paper was in preparation, two recent reports found similar results in gram-positive non-pathogenic bacteria. Rubio and colleagues found that Bacillus subtilis MV were internalized and apparently transcytosed by Caco-2 cells in vitro 30 , while Bajic and colleagues demonstrated that MV from Lactiplantibacillus plantarum BGAN8 are endocytosed by HT-29 cells in a clathrin-dependent manner 31 .
Here we show that CFSE-labelled MV from the gram-positive bene cial bacterium L. rhamnosus JB-1 are internalized by both murine and human IEC within 2h in vitro, as evidenced both by ow cytometry experiments and by the presence of distinct puncta when viewed under routine uorescence microscopy and z-stacking. We further showed evidence of JB-1 MV being internalized in vivo within 2h after oral consumption of CFSE-labelled MV by both small intestinal gut epithelial cells and mononuclear dendritic cells in the lamina propria. While we have shown in a previous publication that labelled MV were internalized by cells in Peyer's patches 18 h after feeding 15 , in the present study visible Peyer's patches were excised prior to isolation of cells for analysis and ow cytometric analysis revealed their dendritic cell nature.
Classically, internalization of MV by IEC is thought to occur via several different mechanisms, including macropinocytosis, clathrin-dependent endocytosis, clathrin-independent endocytosis, and membrane fusion 23 . Interestingly, though rarely acknowledged, phagocytosis (e.g., of pathogens like Salmonella typhimurium and Staphylococcus aureus) can also occur in IEC 32 , though there is no evidence for this with non-pathogenic bacteria. Other investigations of the mechanism of internalization of MV from potentially bene cial bacteria have concluded that endocytosis is primarily clathrin-mediated, as internalization was inhibited by dynasore and chlorpromazine but not lipin III or nystatin 21,31 .
In microscopy experiments, we found that internalization of JB-1 MV was indistinguishable from that of EcN OMV, which suggested that similar mechanisms might be involved. Indeed, pre-incubation of both human and mouse IEC lines with the dynamin inhibitor dynasore almost entirely abolished endocytosis of JB-1 MV, suggesting that their endocytosis is clathrin-mediated. It is important to note, however, that in addition to clathrin-mediated endocytosis, dynasore also inhibits activity of dynamin required for phagocytosis 25 . Thus, we cannot distinguish between these mechanisms. We were also unable to determine which ligand-receptor systems were involved in internalization, as antibodies to LTA, TLR2, and SIGN-R1 were without inhibitory effect in MODE-K cells.
MV from several lactic acid bacteria have been shown to contain distinct cargo 31,33 which in some cases associate with their functional effects 15 . LTA is one promising candidate, as it is present on MV of some lactic acid bacteria 28 , including the L. rhamnosus strain ATCC 7469 34 . Using western blot, we found that LTA is also present on MV from L. rhamnosus JB-1, and that it appears to be a major TLR2 agonist associated with MV as antibody neutralization experiments with anti-LTA reduced TLR2 activation in a reporter cell line. Moreover, anti-LTA inhibited internalization of JB-1 MV by BMDCs and simultaneously reduced their induction of IL-10 expression, suggesting that LTA is involved in immunoregulatory effects of JB-1 MV and that these are magni ed by MV internalization.
Lipoteichoic acids are amphiphilic membrane-anchored polymers associated with the cell wall of grampositive bacteria 35 . They are important for bacterial physiology and host-bacteria interaction, and are commonly considered analogous to the gram-negative lipopolysaccharide, as both are studied as highly immunologically active molecules in host-bacteria interactions 36 . LTA are structurally variable between species 35 , and evidence is accumulating that LTA from some bacteria are immunoregulatory 29,37−40 .
The most extensively studied receptor for LTA is TLR2, though interactions with other receptors have been documented 35 . Interestingly, structurally distinct LTAs appear to interact with TLR2 to different extents and are thought to contribute to varied response magnitudes associated with different bacteria 41 . The potential role for TLR2 signalling in MV-associated LTA is interesting considering this receptor's known role in microbiota-host homeostasis. Round and colleagues demonstrated that colonization of the gut by Bacteroides fragilis requires TLR2 stimulation by polysaccharide A resulting in mucosal tolerance to the bacterium 42 . Subsequent experiments found that similar effects were seen using B fragilis OMV alone, which induced IL-10 production by DC via OMV-associated polysaccharide A interacting with TLR2 13 .

Animals
Male 8-to 10-week-old speci c pathogen-free BALB/c mice were purchased from Charles River (Montreal, Canada) and maintained on a 12-hour light-dark cycle free access to food and water. Mice were euthanized by decapitation. All experiments involving mice were approved by the McMaster Animal Research Ethics Board and followed both the Canadian Council on Animal Care guidelines and the ARRIVE guidelines.

Bacteria and MV Preparation
Lacticaseibacillus rhamnosus JB-1 (JB-1) was grown from stock in Man-Rogosa-Sharpe (MRS) medium at 37°C in anaerobic conditions. Escherichia coli strain Nissle 1917 (EcN) was a gift from Ardeypharm GmbH (Herdecke, Germany) and was grown in LB (Lennox) in aerobic conditions at 37°C with shaking.
After 24 h, cultures were centrifuged at 4°C and 1900 × g for 45 min to pellet bacteria. Supernatants were vacuum ltered through 0.20 μm lter units. The resulting ltrates were ultracentrifuged at 42,000 RPM (138,000 × g) for 3h at 4°C in a Type 45 Ti xed-angle rotor (Beckman Coulter, Mississauga, Canada), pellets resuspended in cold PBS, then ultracentrifuged again at 42,000 RPM (121,000 × g) for 3h at 4°C in a Type 70 Ti xed-angle rotor (Beckman Coulter). Pellets were nally resuspended in 5 μL PBS for every 1 mL of ultracentrifuged supernatant (i.e., concentrated 200x), aliquoted, and frozen at -80°C until further use. Protein concentrations of MV preparations were determined using the Pierce Rapid Gold bicinchoninic acid assay (Thermo Scienti c, Mississauga, Canada).
To uorescently label them, MV were incubated with 20 μm CFSE (CFDA SE; Invitrogen, Burlington, Canada) or 20 μm DiO (Invitrogen) in the dark for 20 min at 37°C. Samples were diluted in cold PBS, ultracentrifuged to wash, then resuspended in equal volume PBS and stored at -80°C. To ensure the absence of nanoparticles in the original dye stocks, negative controls for all experiments were created by incubating the same concentration of CFSE or DiO with an equal volume of sterile PBS, then ultracentrifuging and resuspending as above. These negative controls did not produce uorescence when incubated with any cell lines.
Nanoparticle tracking analysis MV were characterized by nanoparticle tracking analysis (NTA) using a NanoSight NS300 (Malvern Panalytical, Montreal, Canada) at the Structural & Biophysical Core Facility at the Hospital for Sick Children (Toronto, Canada). MV were diluted in PBS to 30-100 particles per frame then continuously owed by syringe pump through a 532 nm laser. Five 60 sec recordings (camera level 16) were analysed using NTA software (v. 3.2; Malvern Panalytical) with a detection threshold of 5.

Electron microscopy
Electron microscopy was performed by the Canadian Centre for Electron Microscopy (McMaster University, Hamilton, Canada). Samples were deposited (3.5 μL) onto formvar-coated copper grids and incubated for 10 min. Excess liquid was blotted, samples were negatively stained with 1% aqueous uranyl acetate (3.5μL) to each grid, incubated for 1 min, then blotted and dried by evaporation. Grids were viewed in a 1200 EX TEMSCAN transmission electron microscope (JEOL, Peabody, USA) operating at an accelerating voltage of 80 kV. Images were acquired with a 4-megapixel digital camera (Advanced Microscopy Techniques, Woburn, USA).
BMDCs were derived as previously described 15,44 using tibia and femurs from BALB/c mice. Cells were plated in 100 mm dishes at 10 6 /mL in 20 mL growth medium (day 0), refreshed on days 2 and 6, and harvested on day 7.
For uorescence microscopy, cells on coverslips were xed for 15 min in 4% formaldehyde in PBS, washed twice, mounted with ProLong Glass antifade mountant with NucBlue nuclear stain (Hoechst 33342; Invitrogen), allowed to cure for 24 h, then imaged using a Zeiss Axio Imager Z1 microscope and processed using AxioVision software (v. 4.8; Zeiss, Toronto, Canada).
TLR2 assay TLR2 ligand presence was determined using a mouse TLR2 reporter cell line (HEK-Blue-mTLR2; InvivoGen) following manufacturer's directions and as described previously 15 . Where appropriate, cells were pre-incubated with anti-LTA antibody (5 μg/mL) for 1 h then incubated with 10 μL sample in 90 μL media at 37°C for 20 h. Positive control wells were incubated with the TLR2 agonist Pam3CSK4 (300 ng/mL). Cell-free supernatants (20 μL) were then added to the detection reagent (180 μL), incubated at 37°C for 1 h, and measured spectrophotometrically at 650 nm.

Measuring JB-1 MV-CFSE internalization by IEC and DCs in vivo
Mice were gavaged with approx. 3x10 10 CFSE-labelled JB-1 MV in 200 μL PBS or with PBS alone, then 2 h later were euthanised and jejuna removed. Single-cell suspensions of IEC and lamina propria DC were then prepared as previously described 45,46 and kept dark where possible to limit photobleaching. Tissues were stripped of mesentery and visible Peyer's patches excised, ushed with cold PBS, cut into segments, suspended in 30 mL Hank's balanced salt solution (Ca 2+ and Mg 2+ free) with 5% FBS, 1 mM DTT, and 5 mM EDTA, then incubated in a shaking water bath for 30 min at 37°C to dissociate epithelial cells. Suspensions were then ltered successively through 70 μm and 40 μm cell strainers, separating dissociated IEC from lamina propria. IEC suspensions were washed twice then incubated with rabbit antimouse A33 polyclonal antibody (1:100; Invitrogen) for 1 h and goat anti-rabbit IgG (APC, 1:50; Invitrogen) and anti-mouse CD45-APC-Cy7 (1:100; Invitrogen) for 45 min, then analysed by ow cytometry.
To isolate lamina propria, non-dissociated jejunal segments were collected from the rst 70 μm strainers and cut into smaller pieces. These were then suspended in 20 mL RPMI with 5% FBS, 1 mg/mL collagenase IV/dispase (Invitrogen), and 40 μg/mL DNAse I (Roche, Mississauga, Canada) and incubated in a shaking water bath at 37°C for 45 min. Resultant suspensions were ltered though a 40 μm strainer, washed once with cold PBS, then applied to a Percoll (GE Healthcare, Mississauga, Canada) gradient (top layer 30%, bottom layer 75%) and centrifuged at 540 × g for 20 min at room temperature. The cells at the interface were collected, washed twice with PBS, then antibody-labelled with anti-mouse CD11c-PerCP-Cy5. 5

Declarations Data Availability
Data generated during the current study are present in the supplementary section or are available from the corresponding author on request. (d, e) Approx. 3x1010 CFSE-labelled JB-1 MV or phosphate-buffered saline (PBS) vehicle were orally gavaged to BALC/c mice, then 2 h later jejuna were isolated, and uorescence was measured by ow cytometry of (d) A33+ CD45-intestinal epithelial cells or (e) CD11c+ MHC II+ lamina propria dendritic cells. Scale bars represent 10 μm. Green colour represents CFSE signal, while blue represents nuclear stain (Hoechst 33342).

Figure 4
Lipoteichoic acid in L. rhamnosus JB-1 MV is responsible for immunomodulatory effects. (a) Western blot analysis was used to measure LTA associated with JB-1 MV (MV) or in lysates of whole JB-1 bacteria (Bact.). (b) Independent preparations of JB-1 MV were incubated with a TLR2 reporter cell line with or without preincubation with anti-LTA antibody (+ αLTA) or isotype control (+ Iso), and TLR2 activity was expressed as a percentage of that measured for the synthetic TLR2 ligand Pam3CSK4 (300 ng/mL).
(c-e) DiO-labelled MV or vehicle were incubated with BMDCs for 18 h either with or without preincubation with dynasore or anti-LTA antibody, then (c) DiO-related or (d) IL-10-related uorescence were measured by ow cytometry. (e) The extent to which MV internalization was associated with IL-10 expression was