Toll-like receptor 2 and nucleotide-binding oligomerization domain-2 play divergent roles in the recognition of gut-derived lactobacilli and bifidobacteria in dendritic cells

Abstract

The gut microbiota is vital in the maintenance of homeostasis in the gut immune system. Its diversity and composition play major roles in relation to allergies and inflammatory bowel diseases, and administration of lactic acid bacteria (LAB), such as lactobacilli and bifidobacteria, has positive effects on these pathologies. However, the mechanisms behind the beneficial effects are largely unknown. Here we reveal divergent roles played by Toll-like receptor-2 (TLR2) and nucleotide-binding oligomerization domain-2 (NOD2) in dendritic cell (DC) recognition of LAB. Murine bone-marrow-derived DC lacking NOD2 produce higher levels of interleukin-10 (IL-10) and reduced levels of IL-12 and tumour necrosis factor-α (TNF-α) in response to LAB. This indicates that peptidoglycan is partly responsible for the T helper type 1 skewing effect of certain LAB. Dendritic cells that are TLR2^−/− produce less IL-12 and TNF-α and more IL-10 in response to some strains of lactobacilli, while they produce more IL-12 and less IL-10 in response to bifidobacteria. The same tendency was found in human monocyte-derived DC. We have previously reported that the weak IL-12-inducing and TNF-α-inducing bifidobacteria inhibit the T helper type 1 skewing effect induced by strong immunostimulatory lactobacilli. Here we show that this immunoinhibitory effect of bifidobacteria is dependent on TLR2 and independent of NOD2. Moreover, independently of the cytokine pattern induced by intact LAB, cell wall fractions of all LAB, as well as synthetic lipoproteins possess immunoinhibitory capacities in both human and murine DC. These novel findings suggest that LAB act as immunoregulators through interaction of lipoprotein with TLR2 and as immunostimulators through interaction of peptidoglycan with NOD2.

Introduction

The influence of microbiota on mucosal immune function and gut health has emerged as an area of scientific and clinical importance. Mounting evidence stresses the pivotal role of microbiota in the maintenance of gut homeostasis and immunological tolerance. Oral administration of lactic acid bacteria (LAB), primarily strains of the Lactobacillus or Bifidobacterium genera, also termed ‘probiotics’, has proven beneficial in a variety of immunopathologies, such as inflammatory bowel diseases^1,2 and atopic diseases.^3–5 Little attention has been given to differences in how the immune system recognizes lactobacilli and bifidobacteria or the molecular patterns which mediate their immunomodulatory effects. Dendritic cells (DC) are considered to be gate keepers of the immune system, and contact between DC and microbiota is essential for proper immune system development and regulation. In a healthy state, subepithelial DC sample the microbiota by passing their dendrites between epithelial cells into the gut lumen^6,7 and by interacting directly with bacteria that have gained access to lymph nodes and Peyer’s patches via M cells.⁸ As DC are potent stimulators of naive T cells and express several pattern recognition receptors (PRRs), they serve as an important link between the microbiota and polarization towards T helper type 1 (Th1), Th2- or regulatory T-cell-dominated environments. Cell surface-expressed and intracellularly expressed PRRs collectively recognize lipid, carbohydrate, protein and nucleic acid structures that are broadly expressed by different groups of microorganisms. The most studied PRRs are the Toll-like receptor (TLR) family comprising at least 12 receptors triggering immune responses through the nuclear factor-κB dependent pathway. The expression of TLRs is tightly regulated and differs between DC populations at different anatomical sites,⁹ e.g. intestinal lamina propria CD11c⁺ DC express TLR5 but not TLR4, which might prevent the induction of immune responses towards commensal Gram-negative bacteria.¹⁰ Another family of PRRs, the C-type-lectin receptors, recognize mannose residues and play an important role in binding and uptake of microbial components.¹¹ A third family of PRRs is the nucleotide-binding oligomerization domain (NOD) -like receptor family, including the intracellularly expressed NOD1 and NOD2, which recognize diaminopimelic acid of Gram-negative bacterial peptidoglycan (PGN)¹² and muramyl dipeptide (MDP) of Gram-positive and Gram-negative bacterial PGN,¹³ respectively.

The LAB are Gram-positive bacteria and their cell walls comprise a complex mixture of glycolipids, lipoproteins and phosphorylated polysaccharides embedded in a thick layer of PGN, a polymer of β(1-4)-linked N-acetylglucosamine and N-acetylmuramic acid, crosslinked by short peptides. TLR2 has been shown to recognize lipoteichoic acid (LTA), lipoarabinomannan, lipoprotein (LP) and PGN, and is likely to be involved in the recognition of LAB. Recognition of these diverse microbial structures demands complex and finely tuned TLR2 signalling upon ligand binding. It has been found that TLR2 forms dimers with both TLR1 and TLR6, resulting in recognition of tri- and diacylated LP respectively, which partly explains this promiscuity.^14,15 As a result of contaminating molecules in preparations of PGN and LTA, contradicting reports exist on the ligands recognized by TLR2. Travassos et al. showed that PGN recognition by TLR2 was lost after removal of contaminating LP and LTA, indicating that PGN is recognized independently of TLR2;¹⁶ however, a later study refutes these findings.¹⁷ Newer studies suggest that the TLR2 stimulating effect of LTA is the result of contaminating LP.^18,19 The role of TLR2 and its coreceptors in the recognition of pathogenic microbial motifs and their involvement in infectious disease have received enormous attention compared to the role of TLR2 in the recognition of LAB commensals and possibly in tolerance. Mice that are deficient in TLR2, TLR4 or MyD88 are more susceptible to dextran sulphate sodium-induced colitis than wild-type (WT) mice, and this indicates a protective role of the microbiota. In contrast, mice deficient in NOD2 were as protected as WT mice.²⁰ C-type lectin receptors have also been reported to alter TLR-mediated signalling, e.g. several pathogens use DC-specific intercellular adhesion molecule-grabbing non-intregrin (DC-SIGN) as an escape mechanism.²¹ The DC-SIGN may also play a role in the regulatory function of certain strains of LAB.²² No studies on the involvement of NOD2 in the recognition of LAB have been reported, but Foligne et al. have shown that, whereas transfer of LAB-treated WT DC mediated protection from 2,4,6-trinitrobenzenesulphonic acid-induced colitis, transfer of LAB-treated TLR2-deficient or NOD2-deficient DC did not.²³

It remains a puzzle how DC can differentiate between pathogens and commensals, which have several pathogen-associated molecular patterns in common. When DC encounter a microorganism, distinct TLRs act in synergy^24,25 and TLRs cooperate with other PRRs resulting in a coordinated sum of signals. The interrelationship between TLR2 and NOD2 is complicated by the postulation that different forms of PGN are recognized by both receptors^15,17 and, not surprisingly, contradictory data exist on this matter. Stimulation by PGN of splenocytes that are deficient in NOD2 produces more T helper type 1 (Th1) cytokines compared to WT splenocytes, and MDP inhibits Pam3CSK4-induced Th1 cytokines only in WT mice,²⁶ indicating a negative regulation mediated by NOD2 on TLR2. However, Kobayashi et al. reported synergistic effects of MDP and TLR ligands, including Pam3CSK4, in WT, but not in NOD2-deficient bone marrow-derived macrophages.²⁷ Synergistic effects of TLR2 ligands on MDP-induced activation in human monocytes^28,29 and in human mononuclear cells³⁰ have also been reported, while synergistic effects between NOD2 and TLR3, TLR4 and TLR9, but not TLR2, have been reported in human monocyte-derived DC (moDC).³¹ These diverse findings suggest that cooperation between TLRs and NOD2 is cell-type specific and dependent on the structure of the agonist.

We have previously shown that strains of lactobacilli differentially mature both human moDC and murine bone marrow-derived DC (bmDC) in vitro with vast differences in the expression of interleukin-12 (IL-12), tumour necrosis factor-α (TNF-α) and surface markers, while all bifidobacteria induce low levels of these Th1 skewing cytokines and surface markers.^32,33 The effect on DC maturation is further transmitted on to allogeneic naive T cells, which differentiate into interferon-γ (IFN-γ)-producing Th1 cells, and to natural killer cells, which produce IFN-γ in response to DC matured by strong IL-12-inducing and TNF-α-inducing lactobacilli, while bifidobacteria are much less potent.^34–36 Bifidobacteria are able to inhibit DC maturation induced by strong Th1 skewing strains.^32–35 In this study we set out to elucidate the mechanisms behind these large differences by focusing on bacterial components and their receptors. Our results reveal an unexpected role of TLR2 in the anti-inflammatory effects of LAB in DC maturation, while NOD2 appears to be involved in the Th1 skewing effects of LAB, providing a new perspective on the cross-talk between TLR2 and NOD2 in immune recognition of the microbiota.

Materials and methods

Preparation of ultraviolet-light-killed LAB and cell wall fractions

The microbiota-derived bacteria used are listed in Table 1. Lactobacilli and bifidobacteria were grown anaerobically overnight at 37° in de Man, Rogosa and Sharpe broth (Merck, Darmstadt, Germany). The cultures were harvested, washed twice in sterile phosphate-buffered saline (PBS; Cambrex Bio Whittaker, East Rutherford, NJ) and resuspended in one-tenth the growth volume of PBS. The bacteria were killed by a 40-min exposure to ultraviolet (UV) light and stored at − 80°. Concentration was determined by lyophilization. Cell wall (CW) fractions were prepared by sonicating overnight cultures of LAB (washed as described) for 45 min followed by 15 min of heating to 60° to inactivate autolytic enzymes. Whole cells were removed by centrifugation at 1000 g at 4° for 10 min, and the CW fraction was obtained by centrifuging the resulting supernatant at 40 000 g at 4° for 30 min. The CW fractions were exposed to UV light for 15 min and kept at − 80°. The dry weight of the CW fractions was determined after lyophilization. LAB-spent growth media (SM) was prepared by growing LAB in RPMI-1640 supplemented with 2 mm l-glutamine, 10% (v/v) heat-inactivated fetal calf serum and 50 μm 2-mercaptoethanol (all from Cambrex Bio Whittaker) until the stationary growth phase. Cells were removed by centrifugation at 2000 g at 4° for 10 min, the SM was sterile-filtered (0·2 μm) and stored at − 80°. The SM was added to DC at 10% of the culture volume and CW fractions and intact LAB as indicated. Endotoxin levels in LAB preparations and CW fractions were determined using the Pyrochrome kit (Association of Cape Cod, East Falmouth, MA) to < 0·05 EU/ml in the highest concentration of stimuli used in cell culture experiments.

Table 1.

Strains used in this study

Name	Origin	DC maturation pattern
Lactobacillus acidophilus X37	Adult, biopsy2	High levels ofIL-12, TNF-α andsurface markers. Promotes Th1 polarization
Lactobacillus paracasei Z11	Adult, biopsy2
Lactobacillus casei CRL431	Child, faeces1
Lactobacillus rhamnosus GG	Adult, faeces3
Bifidobacterium longum Q46	Adult, biopsy2	Low levels of IL-12,TNF-α and surfacemarkers. Inhibits Th1 polarizationinduced bylactobacilli
Bifidobacterium bifidum Z9	Adult, biopsy2
Bifidobacterium breve 20091	Child, faeces2
Bifidobacterium bifidum20082	Child, faeces2

¹Chr. Hansen A/S, Hørsholm, Denmark.

²Faculty of Life Science, University of Copenhagen, Denmark.

³Valio, Helsinki, Finland.

IL-12, interleukin-12; Th1, T helper type 1; TNF-α, tumour necrosis factor-α.

In vitro generation of human DC from monocytes

Human peripheral blood mononuclear cells were obtained after Ficoll density-gradient centrifugation of buffy coats from healthy donors. Monocytes were isolated by magnetic activated cell sorting with hCD14⁺ microbeads (Miltenyi Biotech, Bergisch Gladbach, Germany) and were cultured at a density of 6 × 10⁶ cells/3 ml/well in six-well tissue culture plates (Nunc, Roskilde, Denmark) for 6 days in culture medium [RPMI-1640 supplemented with 2 mm l-glutamine, 10% (v/v) heat-inactivated fetal calf serum, 100 U/ml penicillin, 100 μg/ml streptomycin and 50 μm 2-mercaptoethanol (all from Cambrex Bio Science)] containing 30 ng/ml IL-4 (corresponding to 15% supernatant from an IL-4-transfected cell line) and 20 ng/ml recombinant human granulocyte–macrophage colony-stimulating factor (GM-CSF) from Biosource (Nivelles, Belgium). After 3 days, media containing full doses of cytokines were added. On day 6, 90–95% of the cells expressed the DC-marker CD1a and trypan blue exclusion indicated that cultures contained ≥ 95% viable cells. The immature DC were harvested and reseeded in 48-well tissue culture plates (Nunc) at 6 × 10⁵ cells/500 μl/well in culture medium. The UV-killed bacteria or TLR ligands [Staphylococcus aurus LTA and PGN (Sigma-Aldrich, St Louis, MO), Pam2CSK4, Pam3CSK4 and MDP (Invivogen, San Diego, CA)] were added at 100 μl/well in final concentrations as indicated, and cells were incubated for 18 hr at 37° in a 5% CO₂ humidified atmosphere. Preliminary experiments had shown that timing was essential for the DC response; therefore stimuli were premixed when multiple stimuli were tested simultaneously. Neutralization with TLR2 (eBioscience, San Diego, CA) and DC-SIGN antibodies (R&D Systems, Minneapolis, MN) was performed 1 hr before stimulation at 50 and 20 μg/ml respectively with matched isotype control antibodies. No effect was observed for matched isotype control antibodies on DC maturation so isotype controls were not included in the figures. Neutralization assays were performed in 96-well tissue culture plates (Nunc) at 1·5 × 10⁵ cells/150 μl/well in culture media.

Generation of murine bmDC

Bone marrow cells were isolated and cultured as described earlier.³³ Briefly, femurs and tibias from sex-matched and age-matched WT C57BL/6, TLR2^−/− (B6.129-TLR2^tm1Kir/J) and CARD15^−/− (B6.129S1-Card15^tm1Flv/J) mice (Charles River Breeding Laboratories, Portage, MI), were removed and stripped of muscles and tendons and the marrow was flushed. The resulting cell suspension was washed and cells were resuspended in culture medium supplemented with 15 ng/ml murine GM-CSF [3% (v/v) culture supernatant harvested from a GM-CSF-transfected Ag8.653 myeloma cell line] at 3 × 10⁵/ml, 10 ml/100-mm Petri dish and incubated for 8 days at 37° in 5% CO₂. Fresh media containing 15 ng/ml GM-CSF was added on days 3 and 6. On day 8, immature DC were harvested and reseeded in 48-well tissue culture plates (Nunc) at 9 × 10⁵ cells/500 μl/well in culture medium without GM-CSF. The DC were > 90% CD11c-positive. Stimuli were added in 100 μl/well at final concentrations as indicated, and cells were incubated for 16 hr at 37° in a 5% CO₂ humidified atmosphere.

Immunostaining and flow cytometry

Human DC were harvested and resuspended in cold PBS containing 1% (v/v) fetal bovine serum and 0·15% (w/v) sodium azide (PBS-Az), while murine DC were resuspended in PBS-Az containing anti-mouse FcγRII/III (3 μg/ml; BD Biosciences, San Jose, CA) to block non-specific binding of antibody reagents. The following antibodies were used for staining: phycoerythrin (PE)-conjugated anti-human CD1a, allophycocyanin (APC)-conjugated anti-human CD83 (both from BD Biosciences), PE-conjugated anti-human HLA-DR, PE-conjugated anti-human CD86, APC-conjugated anti-human CD40, APC-conjugated anti-mouse CD86, PE-conjugated anti-mouse MHCII, PE-conjugated anti-mouse CD11c (all from Southern Biotech, Birmingham, AL), PE-conjugated anti-human TLR2 and PE-conjugated anti-mouse CD40 (eBioscience), non-specific binding was evaluated by including matched isotype controls for all the antibodies used. DC were analysed using a BD FACSarray flow cytometer (BD Biosciences) based on counting 10 000 cells. The level of expression was expressed as the geometric mean fluorescence intensity (MFI).

Cytokine quantification in culture supernatants

The production of human and murine IL-12(p70), IL-10, IL-6 and TNF-α was analysed using commercially available enzyme-linked immunosorbent assay kits (R&D Systems).

Statistical analysis

Statistical analysis (one-way analysis of variance with Tukey post-test) was performed using the GraphPad Prism version 4.03 (GraphPad Software, San Diego, CA). Differences were considered significant if P < 0·05.

Results

Only intact LAB fully mature human moDC

The first approach to elucidate the mechanism behind the differential DC maturation patterns induced by different strains of LAB was to fractionate selected strains and characterize the effects of CW fractions on moDC maturation along with SM from the bacterial cultures. We selected two strains, Lactobacillus acidophilus X37 and Bifidobacterium longum Q46, as representatives of the Lactobacillus and Bifidobacterium genera. Lactobacillus acidophilus X37 induces high levels of TNF-α and IL-12 in DC and high levels of IFN-γ in T cells, while B. longum Q46 is not only weak in driving this Th1 polarization but also possesses the ability to inhibit the Th1 polarization induced by L. acidophilus X37.^32,33,35 The cytokine levels and expression of surface markers in moDC after stimulation with CW fractions are shown in Fig. 1. The IL-12, TNF-α and IL-6 stimulatory capacity was reduced more than 20-fold when the bacteria were ruptured, while IL-10 production was only reduced about fourfold, demonstrating that only intact LAB can fully mature moDC. The SM-matured DC produced IL-10, but no or very low amounts of IL-12, TNF-α and IL-6 and they resembled immature DC in their expression of surface markers (flow cytometric data not shown). The CW fractions induced lower levels of cytokines compared to intact bacteria even at concentrations 100 times higher, at which they induced comparable expression of surface markers. Expression of CD83 is shown in Fig. 1(b); CD80, CD40 and CD86 followed the same pattern. The CW fractions maintained the ability to induce IL-10, but at lower levels than intact bacteria. The CW fraction of L. acidophilus X37 was the most potent fraction for maturing DC.

Figure 1.

Only intact lactic acid bacteria (LAB) induce fully mature human monocyte-derived dendritic cells (moDC). Human moDC (6 × 10⁵ cells/500 μl/well) were matured by 2 μg/ml intact or ruptured LAB, 10% LAB spent medium (SM), or 2, 20 or 200 μg/ml cell wall (CW) fractions of Lactobacillus acidophilus X37 and Bifidobacterium longum Q46. (a) DC-derived interleukin-12 p70 (IL-12p70), tumour necrosis factor-α (TNF-α), IL-10 and IL-6 in culture supernatant produced in response to the various stimuli. Data are means and SD of triplicate cultures. ‘#’ indicates: not detectable. Immature DC did not produce detectable amounts of any of the four cytokines. (b) Expression of CD83 in response to intact LAB [left, solid histogram represents DC stimulated with 2 μg/ml intact LAB, dotted histogram represents immature DC (iDC) and grey-filled histogram represents binding of isotype control antibody] or CW fractions of LAB (right, solid histogram represents 200 μg/ml, dotted histogram 20 μg/ml and grey-filled histogram represents 2 μg/ml). Numbers indicate the geometrical mean fluorescent intensity (grey refers to grey histograms, black to solid histograms, and black italic to dotted histograms). Data are representative of four experiments with cells from different donors.

Neutralization of TLR2 in human moDC affects LAB-induced maturation

Since TLR2 is involved in the recognition of Gram-positive cell wall components, LAB-induced maturation was assessed after neutralization of TLR2 in moDC (Fig. 2). Production of IL-12 and TNF-α was reduced only in DC matured by either intact or the CW fraction of L. acidophilus X37, while IL-10 production was reduced only in DC matured by the intact or CW fraction of B. longum Q46. Interleukin-6 was only reduced in response to intact LAB. In response to Pam3CSK4, production of TNF-α, IL-10 and IL-6 was only modestly reduced by blocking TLR2 (27%, 62% and 19%, respectively) while IL-12 was undetectable. This indicates that the antibody does not fully block the receptor, and hence we chose to repeat these experiments in murine DC that lacked TLR2 (see Figs 6 and 7). Neutralization of TLR2 reduced the expression of CD40 and CD86 in DC matured by CW fractions of LAB (Fig. 2b), but not in DC matured by intact LAB (data not shown). The expression of TLR2 on moDC was followed during culture with the strains of LAB listed in Table 1 and was found to be upregulated on LAB-matured DC compared to immature DC with only minor differences between strains (L. acidophilus X37 and B. longum Q46 are shown in Fig. 2c). Pam3CSK4 also induced higher expression of TLR2, suggesting that LP was responsible for the LAB-induced upregulation of TLR2 in DC.

Figure 2.

Neutralization of Toll-like receptor 2 (TLR2) in human monocyte-derived dendritic cells (moDC) affects maturation induced by lactic acid bacteria (LAB). Human moDC (1·5 × 10⁵ cells/150 μl/well) were matured by various stimuli. (a) DC-derived interleukin-12 (IL-12) p70, tumour necrosis factor-α (TNF-α), IL-10 and IL-6 in culture supernatant in response to intact LAB (2 μg/ml), cell wall (CW) fractions (200 μg/ml), Pam3CSK4 (1 μg/ml) and muramyl dipeptide (MDP; 10 μg/ml) with and without neutralization with anti-TLR2 antibody (white bars) or matched isotype control (filled bars). Data are means and SD of triplicate cultures. Percentages represent reduction in cytokine levels by blocking TLR2. ‘#’ indicates: not detectable. (b) Expression of CD40, CD83 and CD86 in response to 200 μg/ml CW fractions of LAB [solid histogram represents DC without and dotted histogram with neutralization of TLR2, grey-filled histogram represents immature DC (iDC) and light grey-filled histogram represents binding of isotype control antibody]. Numbers indicate the geometrical mean fluorescence intensity (MFI; black refers to solid histograms, and black italic to dotted histograms). (c) TLR2 expression on DC matured by LAB (2 μg/ml) and Pam3CSK4 (1 μg/ml) as indicated (solid histograms), iDC (dotted histogram) and isotype control (grey-filled histogram). Numbers indicate the geometrical MFI (grey refers to grey histograms, black to solid histograms, and black italic to dotted histograms). Data are representative of four experiments with cells from different donors. Blocking of TLR2 did not change lipopolysaccharide-induced cytokine production or expression of surface markers (data not shown).

Figure 6.

Toll-like receptor 2 (TLR2) and nucleotide-binding oligomerization domain-2 (NOD2) play divergent roles in lactic acid bacteria (LAB)-induced cytokine production by murine bone-marrow-derived dendritic cells (bmDC). Murine bmDC (9 × 10⁵ cells/500 μl/well) from wild-type (WT), TLR2^−/− and CARD15^−/− mice were matured with 50 μg/ml LAB, 3 μg/ml Pam2CSK4, 3 μg/ml Pam3CSK4, 30 μg/ml muramyl dipeptide (MDP), 30 μg/ml Staphylococcus aureus lipotecichoic acid (LTA) or 30 μg/ml S. aureus peptidoglycan (PGN). DC-derived interleukin-12 (IL-12) p70, tumour necrosis factor-α (TNF-α), IL-10 and IL-6 in culture supernatant in response to the various stimuli are shown. Data are means and SD of triplicate cultures. The levels of cytokine produced by DC lacking NOD2 or TLR2 were significantly different from the levels produced by WT DC (P < 0·01) when marked with ‘*’. ‘#’ indicates: not detected. Data are representative of three experiments. Note the different values on the x axes. DC lacking TLR2 or NOD2 responded to lipopolysaccharide (LPS) as WT DC (data not shown).

Figure 7.

Toll-like receptor 2 (TLR2) and nucleotide-binding oligomerization domain-2 (NOD2) play distinct roles in lactic acid bacteria (LAB)-induced expression of maturation markers in murine bone-marrow-derived dendritic cells (bmDC). Murine bmDC (9 × 10⁵ cells/500 μl/well) were matured by strains of LAB (5 μg/ml), Pam2CSK4 (0·3 μg/ml), Pam3CSK4 (0·3 μg/ml) or muramyl dipeptide (MDP; 3 μg/ml). Expression of CD86, MHC II and CD40 is shown for WT DC [grey-filled histograms, mean fluorescence intensity (MFI) in grey], TLR2^−/− DC (black solid histograms, MFI in black) and CARD15^−/− DC (black dotted histograms, MFI in italic black). Filled light grey histograms represent isotype control. This is one representative experiment out of three.

Both CW fractions and SM of LAB efficiently inhibit maturation in human moDC induced by L. acidophilus X37

Next we evaluated the inhibitory effect of CW fractions and SM of the two selected strains of LAB on L. acidophilus X37-matured DC both in relation to cytokine production and expression of surface markers (Fig. 3). Surprisingly, we found that the CW fraction and SM from both strains inhibited the production of IL-12, TNF-α and levels of surface markers (only the inhibitory effect of CW fractions on CD83 expression is shown) induced by intact L. acidophilus X37, indicating that CW and SM from both an immunostimulatory and an immunosuppressive strain have similar immunosuppressive properties. The CW fraction, SM and intact LAB all induced an additive production of IL-10 and IL-6 in combination with L. acidophilus X37 as expected from our earlier studies.³² While only CW fractions induced expression of surface markers when added at high concentrations (Fig. 1b) both the CW fraction and SM efficiently inhibited L. acidophilus X37-induced DC maturation, indicating a competitive engagement of PRRs by stimulatory and suppressive components found in the CW of LAB.

Figure 3.

Both cell wall (CW) fractions and spent medium (SM) of lactic acid bacteria (LAB) efficiently inhibit maturation in human monocyte-derived dendritic cells (moDC) induced by Lactobacillus acidophilus X37. Human moDC (6 × 10⁵ cells/500 μl/well) were matured by 2 μg/ml intact LAB, 10% LAB SM, or 2, 20 or 200 μg/ml CW fractions of L. acidophilus X37 and Bifidobacterium longum Q46 in combinations with 2 μg/ml intact L. acidophilus X37. (a) Production of interleukin-12 (IL-12) p70, tumour necrosis factor-α (TNF-α), IL-10 and IL-6 by DC matured with combinations of fragmented LAB and L. acidophilus X37 as indicated. The level of cytokine induced by intact L. acidophilus X37 alone is indicated with a dotted line. Data are means and SD of triplicate cultures. Cytokine levels induced by L. acidophilus X37 in combination with intact bacteria, SM or CW were significantly different from the levels induced by L. acidophilus X37 alone (P < 0·001) when marked with ‘*’. ‘#’ indicates: ‘irrelevant’. This graph shows the inhibition of maturation induced by intact L. acidophilus X37. Therefore the combination L. acidophilus X37 and L. acidophilus X37 is irrelevant. (b) The CW fractions alone slightly increased expression of CD83 (left and right, solid histogram: 20 μg/ml, dotted histogram: 2 μg/ml) and efficiently reduced maturation induced by L. acidophilus X37 (right, light grey-filled histogram: 2 μg/ml L. acidophilus X37 alone). Grey histograms represent isotype control. Numbers indicate the geometrical mean fluorescence intensity (grey refers to grey histograms, grey italic to light-grey histograms, black to solid histogram, and black italic to dotted histogram). Data are representative of four experiments with cells from different donors.

The immunosuppressive effect of LAB on L. acidophilus X37-induced maturation in human moDC is independent of DC-SIGN

It has been suggested that DC-SIGN expressed by DC plays a role in the regulatory function of certain strains of LAB in vitro,²² so we investigated whether it plays a role in the inhibitory effect of LAB. Neutralization of DC-SIGN on DC reduced production of IL-12 and TNF-α induced by L. acidophilus X37 but no significant difference was observed for DC matured by B. longum Q46, nor was there an effect on IL-10 and IL-6 production (Fig. 4). The inhibitory effect of B. longum Q46 was independent of DC-SIGN. Since both IL-10 and IL-6 were induced in an additive manner, when DC were matured by strong and weak IL-12/TNF-α-inducing LAB simultaneously, we examined whether these cytokines played a role in the immunosuppressive effect of LAB. Even though neutralization of IL-10 and IL-6 increased the levels of IL-12 and TNF-α, respectively, the inhibitory effect of LAB was unaffected (data not shown).

Figure 4.

The suppressive effect of lactic acid bacteria (LAB) on Lactobacillus acidophilus X37-induced maturation in human monocyte-derived dendritic cells (moDC) is independent of DC-specific intercellular adhesion molecule-grabbing non-intregrin (DC-SIGN). The DC-derived interleukin-12 (IL-12) p70, tumour necrosis factor-α (TNF-α), IL-10 and IL-6 in culture supernatant of human moDC (1·5 × 10⁵ cells/150 μl/well) in response to L. acidophilus X37, Bifidobacterium longum Q46 and combination hereof (5 μg/ml) with and without neutralization with anti-DC-SIGN antibody and matched isotype control. Data are means and SD of triplicate cultures and are representative of three experiments with cells from different donors.

TLR2 ligands efficiently suppress the maturation induced by L. acidophilus X37 in human moDC

Since CW fractions of both lactobacilli and bifidobacteria had an inhibitory effect, likely candidates for the immunosuppressive component of LAB are PGN and LP. After trypsinization of the CW preparations we found a reduction in the inhibitory effect, indicating that protein at least partly constituted the active bacterial component (data not shown). Commercial PGN preparations are contaminated by LP¹⁶ and hence we evaluated the inhibitory effect of synthetic MDP, which is the smallest unit of PGN reported to have immunomodulatory effects via NOD2.¹³ To mimic LAB-derived LP we evaluated the inhibitory effect of synthetic di- and tripalmitoylated LP, Pam2CSK4 and Pam3CSK4, in combination with L. acidophilus X37. As shown in Fig. 5, synthetic LP possessed a dose-dependent inhibitory effect on the levels of IL-12, TNF-α and surface markers induced in L. acidophilus X37-matured DC. However, the efficacy of the ligands differed, with Pam2CSK4 being an extremely potent inhibitor of both IL-12 and TNF-α even at 3 ng/ml. Pam3CSK4 possessed the inhibitory effect only at the highest concentrations tested (33–300 ng/ml), while MDP did not reduce IL-12 and TNF-α production. All ligands alone induced only low amounts of TNF-α relative to LAB and no IL-12, while Pam3CSK4 and Pam2CSK4 induced high levels of IL-10 and IL-6 relative to MDP. When DC were matured by L. acidophilus X37 in combination with any of the three ligands the expression of CD83 was reduced (Fig. 5b) resembling the effect of CW fractions shown in Fig. 3(b).

Figure 5.

Toll-like receptor 2 (TLR2) ligands efficiently suppress maturation induced by Lactobacillus acidophilus X37 in human monocyte-derived dendritic cells (moDC). Human moDC (6 × 10⁵ cells/500 μl/well) were matured by 5 μg/ml L. acidophilus X37 in combinations with graded doses of Pam2CSK4, Pam3CSK4 and muramyl dipeptide (MDP). (a) DC-derived interleukin-12 (IL-12) p70, tumour necrosis factor-α (TNF-α), IL-10 and IL-6 in culture supernatant in response to the various stimuli. The level of cytokine induced by L. acidophilus X37 alone is indicated with a dotted line. Data are means and SD of triplicate cultures. Cytokine levels induced by L. acidophilus X37 in combination with ligands were significantly different from the levels induced by L. acidophilus X37 alone (P < 0·001) when marked with ‘*’. (b) Expression of CD83 in response to ligands as indicated (black histograms: 300 ng/ml, dotted histograms: 33 ng/ml) and ligands in combination with 5 μg/ml L. acidophilus X37 (right, L. acidophilus X37 alone: filled light grey histogram). Filled grey histograms represent isotype control. Numbers indicate the geometrical mean fluorescence intensity (grey refers to grey histograms, grey italic to light grey histograms, black to solid histograms, and black italic to dotted histograms). Data are representative of four experiments with cells from different donors.

TLR2 and NOD2 play divergent roles in LAB-induced maturation of murine bmDC

To further characterize the components responsible for the immunostimulatory effects we investigated the maturation pattern induced by LAB in bmDC from TLR2^−/− and CARD15^−/− mice, lacking TLR2 and NOD2, respectively. The cytokine production of DC matured by four selected strains of lactobacilli and four strains of bifidobacteria along with relevant ligands is presented in Fig. 6. Lack of NOD2 resulted in reduced IL-12 and TNF-α production, enhanced IL-6 production and similar production of IL-10 when compared with WT DC in response to strains of lactobacilli. The DC lacking NOD2 produced lower levels of IL-12 and TNF-α, and higher levels of IL-10 in response to bifidobacteria, while IL-6 production was only modestly reduced. The TLR2^−/− DC also responded differently to the two bacterial genera. Bifidobacteria induced much higher levels of IL-12 and lower IL-10 and IL-6 in the absence of TLR2, while TNF-α levels resembled the levels produced by WT DC. The TLR2-deficient DC were only partially impaired in their recognition of lactobacilli with a small reduction in production of IL-12, TNF-α and IL-6. It is not surprising that the responses to various strains of lactobacilli were differentially affected by lack of TLR2 because lactobacilli are a heterogeneous genus in respect to their DC maturation patterns.^33,32 In response to MDP, CARD15^−/− DC produced no or very low amounts of the four cytokines, while TLR2^−/− DC responded in the same way as the WT DC. TLR2^−/− DC did not respond to Pam2CSK4 and Pam3SK4, but surprisingly these synthetic LPs, like the lactobacilli, induced more IL-6 in CARD15^−/− DC compared to WT DC. The cytokine production in response to LTA and PGN from S. aureus in DC lacking NOD2 and TLR2 was lower compared to that in WT DC for IL-12 and TNF-α, while a small decrease was seen in IL-10 in DC lacking TLR2 and a small increase was seen in DC lacking NOD2. While lack of NOD2 and TLR2 affected the cytokine profiles in DC matured by bifidobacteria differently from those in DC matured by lactobacilli, the pattern in induction of surface markers was independent of the genus of LAB. The expression of CD86, MHCII and CD40 in WT DC, TLR2^−/− DC and CARD15^−/− DC matured by L. acidophilus X37 and B. longum Q46 and relevant ligands is shown in Fig. 7. The expression of all markers in response to LAB was higher in TLR2^−/− DC and lower or equal in CARD15^−/− DC compared to WT DC. As expected, the expression of surface markers was lower in TLR2^−/− DC in response to Pam2CSK4 and Pam3CSK (only Pam3CSK4 is shown), but lack of NOD2 also resulted in lower expression of CD86 in response to LP. DC lacking NOD2 were impaired in their response to MDP. The expression of all markers in immature DC from the two knockout mice resembled the expression in immature WT DC, with a tendency to higher expression in immature TLR2^−/− DC.

The immunosuppressive effect of LAB in bmDC is independent of NOD2 and dependent on TLR2

Synthetic LP and CW fractions of LAB were potent inhibitors of L. acidophilus X37-induced DC maturation so we examined the inhibitory effect of bifidobacteria on L. acidophilus X37-induced DC maturation in WT, TLR2^−/− and CARD15^−/− DC to further elucidate the role of CW molecules interacting with TLR2 and NOD2, respectively. As shown in Fig. 8(a), a dose-dependent reduction in IL-12, TNF-α and IL-6 was seen when WT and CARD15^−/− DC were matured with B. longum Q46 or B. bifidum Z9 in combination with the strong Th1 polarizing LAB strains L. acidophilus X37. This inhibitory effect was absent in TLR2^−/− DC, strongly indicating an involvement of TLR2 ligands in the inhibitory action of LAB. To support these findings, we tested whether the ligands of the two PRRs of interest would inhibit the Th1-skewed maturation induced by L. acidophilus X37. As expected, we found an inhibitory effect of Pam3CSK4 on IL-12, TNF-α and IL-6 production in DC from WT and CARD15^−/− mice, which was lacking in TLR2^−/− DC (Fig. 8b), but no inhibitory effect of MDP was observed (data not shown). This ligand-dependent inhibition resembles the pattern observed in human moDC (Fig. 5a). In general, the maturation pattern induced by LAB in human moDC was paralleled in murine bmDC. However, the immunosuppressive strains of LAB reduced IL-6 production only in murine bmDC (Figs 8 and 3a). In mice, IL-6 has been shown to promote IL-4 production³⁷ and to negatively regulate IFN-γ production³⁸ in CD4⁺ cells, categorizing IL-6 as a strong Th2 skewing cytokine. Hence, a suppressive effect of the inhibitory strains of LAB on both Th1 cytokines and Th2 cytokines in murine bmDC was observed.

Figure 8.

The immunosuppressive effect of lactic acid bacteria (LAB) in bone-marrow-derived dendritic cells (bmDC) is independent of nucleotide-binding oligomerization domain-2 (NOD2) and dependent on Toll-like receptor 2 (TLR2). Production of interleukin-12 (IL-12) p70, tumour necrosis factor-α (TNF-α), IL-10 and IL-6 of DC from wild-type (WT), TLR2^−/− and CARD15^−/− mice in response to various stimuli is shown. Data are means and SD of triplicate cultures and are representative of three experiments. (a) Murine bmDC (9 × 10⁵ cells/500 μl/well) were matured by 50 μg/ml Lactobacillus acidophilus X37 in combination with 5 or 50 μg/ml Bifidobacterium longum Q46 or B. bifidum Z9. The line represents cytokine production induced by L. acidophilus X37 alone. (b) Murine bmDC (9 × 10⁵ cells/500 μl/well) were matured by 10 μg/ml L. acidophilus X37 in combination with 0, 0·03, 0·3 or 3 μg/ml Pam3CSK4.

Discussion

Studies on the involvement of PRRs in immune recognition of bacteria have helped us to understand how cells, especially DC, sense and differentiate between different classes of microorganisms. However, we still know little about how immune cells differentiate between closely related bacteria and understanding the mechanisms behind the immunostimulatory and immunosuppressive effects of LAB is important. Here we demonstrate, in both human and murine DC, that maturation by lactobacilli and bifidobacteria is dependent on TLR2 and NOD2, and that the two PRRs play different roles in the recognition of the two genera.

We have shown that LAB-induced DC production of IL-12, TNF-α and IL-6 requires intact bacteria, as CW fractions were much less potent in maturing DC. Divergent effects of neutralization of TLR2 were seen in response to weak and strong Th1 skewing strains. The strong Th1 skewing effect of L. acidophilus X37 was reduced by the blocking of TLR2, while only IL-10 and IL-6 were reduced in DC matured by B. longum Q46. Neutralization of TLR2 also resulted in reduced levels of costimulatory molecules and IL-10 production in response to CW fractions. Neutralization of DC-SIGN on moDC resulted in a reduction in IL-12 and TNF-α production in response to L. acidophilus X37, but not in response to B. longum Q46. As DC-SIGN has been shown to be involved in the binding of certain strains of LAB²² the strong Th1 skewing effect of lactobacilli may be mediated by DC-SIGN binding, leading to effective phagocytosis and interaction with intracellular PRRs, such as NOD2. Bifidobacteria, as well as CW fractions and SM, might primarily interact with DC through surface expressed PRRs. Spent medium from LAB induced high levels of IL-10, which were reduced upon TLR2 neutralization (data not shown). These observations are in accordance with a study by Hoarau et al., demonstrating a TLR2-dependent induction of IL-10 in human moDC by supernatant of B. breve C50.³⁹ As previously shown, the strong Th1 skewing effect of certain strains of lactobacilli is abrogated by bifidobacteria inducing little IL-12 and TNF-α.^32,33 Here we show that CW fractions and SM of bifidobacteria are also competent in abrogating the strong induction of IL-12, TNF-α and surface marker expression in moDC. Surprisingly, this effect was also observed for the CW fraction and SM from L. acidophilus X37, indicating that the responsible component is a general structural motif found in strains of both lactobacilli and bifidobacteria. Hence, we tested if synthetic variants of the major LAB CW structures, PGN and LP, possessed these immunosuppressive effects and indeed found that synthetic LP, especially the diacylated form, totally abrogated the DC maturation induced by L. acidophilus X37, while MDP had no effect. These observations pointed towards an involvement of TLR2 in the immunosuppressive effects of LAB. Blocking studies of DC-SIGN, IL-10 and IL-6 excluded these in the mechanisms of the immunosuppression mediated by LAB.

TLR2 and TLR4 have recently been suggested to play a role in the maintenance of intestinal epithelial homeostasis,²⁰ but no studies have reported on DC ligation of LAB-derived TLR2 components leading to a regulatory DC phenotype. However, it has been proven that TLR2 ligation combined with T-cell receptor stimulation triggers regulatory T-cell expansion and a temporal loss in the suppressive regulatory T-cell phenotype,^40,41 suggesting that strong TLR2 stimulation during an acute infection attenuates the suppressive activity of regulatory T cells, but as the infection subsides they regain their suppressive activity and help to attenuate the immune response. In the intestine, TLR2 ligation on regulatory T ells might partly explain how the microbiota helps to maintain tolerance, but it is likely that other cell types in direct contact with the microbiota are equally important. To elucidate the role of TLR2 in the DC recognition of LAB we investigated the maturation pattern in bmDC from TLR2^−/− mice. Lack of TLR2 on bmDC resulted in increased expression of maturation markers independent of the genus of LAB. The TLR2^−/− bmDC produced lower levels of IL-10 and IL-6 and higher levels of IL-12 in response to strains of bifidobacteria, while stimulation with different lactobacilli revealed a tendency towards lower production of IL-12, TNF-α and IL-6 and higher production of IL-10. These results agree well with the results of neutralizing TLR2 on human moDC (which only partially block the receptor) and suggest a difference in the mechanism of TLR2-dependent recognition of LAB, which is genus- and species-dependent. In the recognition of bifidobacteria, which induce low levels of IL-12, TNF-α and maturation markers and possess the immunoinhibitory effect,³² TLR2 seems to play a regulatory role that has not been previously described, while it plays an opposite role in the recognition of the strong Th1 skewing strains of lactobacilli. The ability of bifidobacteria to prevent the Th1-skewed DC maturation induced by lactobacilli was absent in bmDC lacking TLR2. These novel findings help us to understand why a microbiota dense in bifidobacteria may prevent the development of allergy⁴² and contribute a novel regulatory role to TLR2 signalling. The only other study addressing TLR2-dependent LAB recognition shows that L. fermentum YIT0159 and LTA purified from L. fermentum induce reduced levels of TNF-α in splenocytes from TLR2^−/− mice compared to WT mice, while modest or no reduction was seen in response to L. casei YIT9029.⁴³ This supports our finding that the TLR2 dependency of lactobacillus recognition varies between species. The divergent dependency on TLR2 signalling of bifidobacteria and lactobacilli might be explained by the existence of different complexes of TLR2, namely TLR2 alone, TLR1/TLR2 and TLR2/TLR6/CD36, which are known to pre-exist on the cell surface and are not induced by ligands.⁴⁴ It is still controversial how LTA is recognized because of the possible contamination with LP as reported by Hashimoto et al.^18,19 However, an early report supports the view that commercial LTA is recognized by TLR2 alone because HEK293 cells transfected with CD14 and TLR2 encoding plasmids were shown to be highly responsive to LTA,⁴⁵ which would not be the case if the immunostimulatory component was LP, the recognition of which requires the presence of TLR1 or TLR6 and CD36. At least some strains of lactobacilli contain LTA, which induces TNF-α in splenocytes in a TLR2-dependent manner.⁴³ Therefore, LTA of lactobacilli may signal through TLR2 inducing proinflammatory cytokines, while at the same time LP signals through TLR2 in receptor complexes, resulting in a reduction in proinflammatory cytokines.

As numerous studies suggest cross-talk between TLR2 and NOD2 signalling^26–31 the immunosuppressive effect of LAB might involve both signalling pathways. Hence, we addressed the role of NOD2, the intracellular sensor of PGN, by using DC lacking expression of NOD2. The IL-12 and TNF-α production in bmDC in response to both lactobacilli and bifidobacteria was reduced in the absence of NOD2, while IL-10 production was enhanced at least in response to bifidobacteria. The inhibitory effect of LAB was not affected by the lack of NOD2. This clearly suggests that signalling through NOD2 in response to the microbiota, or components hereof, suppresses IL-10 production and enhances production of Th1 skewing cytokines. A schematic outline of our current model on how DC recognize LAB is presented in Fig. 9. Induction of strongly Th1 polarizing DC by strains of lactobacilli is to a certain extent dependent on intracellular NOD2 sensing of PGN, and internalization may depend on binding through DC-SIGN. Lactobacilli might further enhance the Th1-skewed maturation process through LTA interaction with TLR2 (Fig. 9a). The driving of DC maturation into a regulatory T-cell-promoting phenotype by bifidobacteria seemingly depends on LP interaction with TLR2 receptor complexes giving an inhibitory signal to the nuclear factor-κB pathway preventing high levels of TNF-α, IL-12 and costimulatory molecules and this signal overcomes the PGN activation of NOD2 (Fig. 9b). When a strong Th1 skewing strain of lactobacilli is present simultaneously with either intact bifidobacteria, CW components of LAB, or synthetic LP, a competitive situation, as depicted in Fig. 9(c), occurs. If the strength of the inhibitory signal of LP through TLR2 complexes is strong enough, it will outcompete the activation signals derived from NOD2 sensing of PGN and possibly from LTA activation of TLR2 and result in low production of IL-12, TNF-α and surface markers and high production of IL-10.

Figure 9.

Schematic representation of suggested mechanisms of dendritic cell (DC) recognition of lactobacilli and bifidobacteria. (a) Lactobacilli (black) induce T helper type 1 (Th1) polarizing DC partly by interactions of peptidoglycan (PGN) with nucleotide-binding oligomerization domain-2 (NOD2). Internalization might depend on binding to DC-specific intercellular adhesion molecule-grabbing non-intregrin (DC-SIGN). Lipoteichoic acid (LTA) might interact with Toll-like receptor 2 (TLR2). (b) Bifidobacteria (white) induce regulatory T cells (Treg) polarizing DC by interactions of lipoprotein (LP) with TLR2 in complex with TLR1 or TLR6, this induces an inhibitory signal and prevents PGN-induced NOD2 activation of nuclear factor-κB (NF-κB). (c) When both stimuli are present simultaneously, a competitive interaction of ligands with NOD2 and TLR2 determines the DC maturation pattern. If the LP-induced TLR1/TLR2 or TLR2/TLR6/CD36 stimulation (either as intact bifidobacteria, lactic acid bacteria-derived cell wall material, spent medium or synthetic LP (black stars) outcompetes the PGN stimulation, DC will mature into a Treg polarizing phenotype as depicted.

Mutations in NOD2 have been implicated in a predisposition to Crohn’s disease with 10–15% of patients having homozygous mutations of CARD15 in the region encoding the leucine-rich repeat, but the precise role of mutated NOD2 is unknown.^46,47 The most common frame shift mutation, 3020insC, abrogates MDP detection by NOD2.¹³ Crohn’s disease is caused by a Th1-dominated inflammation as the result of hyperresponsiveness to certain components of the microbiota.⁴⁸ It is not known whether NOD2 is important for defence against potential pathogens or in maintaining homeostasis by sensing commensal-derived PGN. Mucosal DC in patients with Crohn’s disease express increased levels of CD40, TLR2 and TLR4 compared to DC from healthy controls, indicating that DC are initiators of the inflammatory response.⁴⁹ The data presented here reveal NOD2 as a PRR involved in the Th1 skewing effects of LAB and because the inflammation is characterized by a Th1-dominated environment and because a group of patients with Crohn’s disease carries NOD2 impaired in PGN-mediated nuclear factor-κB activation, other bacteria more aggressive than lactobacilli are likely to be involved in the pathogenesis. Our finding that recognition of the bacterial structure of LAB by TLR2 can inhibit signalling through NOD2 raises the possibility that local changes in the microbiota may alter the balance of positive and negative signals received by DC in inflammatory bowel diseases and might help explain the beneficial effects of administration of certain strains of LAB.²

In particular, our knowledge is advanced by the finding that different structural motifs of LAB simultaneously interact with different PRRs in DC, giving a coordinated sum of signals and determining the DC maturation pattern and consequently the polarization of naïve T cells. This finding gives us information about the mechanisms controlling gut homeostasis and helps us comprehend the complexity of the interplay between the microbiota and the gut immune system.

Abbreviations

APC

allophycocyanin

bmDC

bone-marrow-derived DC

CW

cell wall

DC

dendritic cell

DC-SIGN

DC-specific intercellular adhesion molecule-grabbing non-intregrin

GM-CSF

granulocyte–macrophage colony-stimulating factor

IFN-γ

interferon-γ

IL-12

interleukin-12

LAB

lactic acid bacteria

LP

lipoprotein

LTA

lipoteichoic acid

MDP

muramyl dipeptide

moDC

monocyte-derived dendritic cell

NOD

nucleotide-binding oligomerization domain

PBS

phosphate-buffered saline

PE

phycoerythrin

PGN

peptidoglycan

PRRs

pattern recognition receptors

SM

spent medium

TLR

Toll-like receptor

TNF-α

tumour necrosis factor-α

WT

wild-type