Mechanism of TLR4-Mediated Anti-Inflammatory Response Induced by Exopolysaccharide from the Probiotic Bacillus subtilis

This article is featured in Top Reads, p. 1171

Microorganisms are essential for shaping and developing vertebrate immune systems (1). The hygiene hypothesis suggests that as societies become urbanized and people are less exposed to microorganisms, the immune system fails to develop properly, and inflammatory diseases such as multiple sclerosis, lupus, and allergy develop (2). Probiotics are a class of microorganisms, usually bacteria, known to have health benefits (3), but how they function remains a mystery. Bacillus subtilis is a Gram-positive (Gm(+)) probiotic soil bacterium often found in fermented foods (4). Previous studies showed that B. subtilis has anti-inflammatory properties, and the molecule responsible for this anti-inflammatory effect is its exopolysaccharide (EPS) (5–11).

Intraperitoneal administration of EPS protects against exacerbated disease in murine models of T cell–mediated diseases, including colitis, sepsis, and graft-versus-host disease (GVHD) in a TLR4-dependent manner (5–11). Paynich et al. (7) demonstrated that i.p. administration of EPS 1 d prior to infection with the intestinal pathogen Citrobacter rodentium was sufficient to prevent inflammation and diarrhea. Furthermore, they showed that EPS can suppress the activation of T cells when EPS is administered 3 d prior to activation of T cells with anti-CD3 in vivo (7). In this study, we used bone marrow–derived dendritic cells (BMDCs) to further explore the mechanism by which EPS induces an anti-inflammatory response and found that it requires signaling through the canonical NF-κB pathway as well as the immunosuppressive enzyme IDO. The inhibition of T cell proliferation by IDO-expressing BMDCs utilizes the kynurenine/aryl hydrocarbon receptor (AhR) circuit. Furthermore, unlike LPS, EPS does not use the TLR4 coreceptor, CD14, and does not induce inflammatory cytokines upon injection in vivo.

All studies using mice were reviewed and approved by the Institutional Animal Care and Use Committee at Loyola University Chicago Health Sciences Division (Maywood, IL). C57BL/6, OTII, Tlr4^−/−, and Ido^−/− mice were initially purchased from The Jackson Laboratory and colonies were maintained at Loyola University Chicago. BMDCs were derived from bones of AhR^−/− (Liang Zhou, University of Florida); RELB^−/− (Christoph Vogel, University California at Davis), NF-κB–inducing kinase (NIK)^−/− (Irving Allen, Virginia-Maryland College of Veterinary Medicine), and CD11b^−/− (David Corry, Baylor College of Medicine). RAW264.7 cells were purchased from the American Type Culture Collection and maintained using DMEM and 10% FCS.

Abs were as follows: IDO (Cell Signaling Technology, 51851), p65 (BioLegend, 14G10A21), GAPDH (BioLegend, FF26A/F9), p52 (Cell Signaling Technology, 4882), RELB (Cell Signaling Technology, 10544), lamin B1 (Santa Cruz Biotechnology, sc-374015), phospho-p65 (Cell Signaling Technology, 3033), peroxidase goat anti-rabbit Ig (Jackson ImmunoResearch Laboratories, 111-035-144), and peroxidase goat anti-mouse Ig (Jackson ImmunoResearch Laboratories, 115-035-003). LPS from Escherichia coli, 055:B5 was from Alexis Biochemicals/Enzo Life Sciences.

Splenocytes from OTII mice were harvested and labeled with 5 μM CellTrace Violet (CTV) (Life Technologies) according to the manufacturer’s instructions. BMDCs were loaded with OVA peptide 323–339 (Sigma-Aldrich) for ≥4 h. OVA-loaded BMDCs were cocultured with CTV-labeled OTII splenocytes for 4 d with or without EPS (5 μg/ml). To inhibit IDO, 1-methyl-l-tryptophan (Sigma-Aldrich) (1 mM) was added to the culture. After 4 d, cells were analyzed by flow cytometry.

B. subtilis strain DK7019 (provided by D.B. Kearns and A.M. Burrage, Indiana University, Bloomington, IN) with an isopropyl β-d-thiogalactoside–inducible eps promoter was grown to midlog phase and plated on Luria–Bertani agar plates with 0.1 M isopropyl β-d-thiogalactoside (Teknova). After 4 h at 37°C, bacteria were harvested in endotoxin-free saline containing 50 μg/ml DNase (Sigma-Aldrich) and 34 μg/ml RNase. After centrifugation, the supernatant was incubated ≥4 h with 40 μg/ml proteinase K (Sigma-Aldrich) at 56°C and carbohydrates were precipitated in cold 80% ethanol. Carbohydrate was resuspended in endotoxin-free sterile water and loaded on a Sephacryl S-500 column. Carbohydrate-containing fractions were identified by a phenol–sulfuric acid assay and concentrated using a 30,000 Da molecular mass cutoff Vivaspin filter (Millipore). EPS purity was confirmed by undetectable OD readings at 260 and 280 nm, and by undetectable endotoxin contaminants (<10 ng/ml) tested by dot blot using anti–lipid A of LPS from E. coli (Bio-Rad). All EPS preparations were sterile filtered before use. Mice were injected i.p. with 2.5 mg/kg.

Immune colonic cells were isolated Yamamoto by removing fecal matter, fat, mesentery lymph nodes, and Peyer’s patches and opening longitudinally. Tissue pieces (1 cm) were incubated in Ca²⁺- and Mg²⁺-free 10% HBSS supplemented with FCS, 10 mM HEPES, and 1 mM DTT. After shaking 30 min at 37°C, tissues were vortexed vigorously and passed through a 100-μm filter. To isolate lamina propria lymphocytes, the remaining tissues were incubated in 10% HBSS supplemented with 0.1 mg/ml DNAse (Roche, Basel, Switzerland), 2 mg/ml collagenase D (Sigma-Aldrich), and 0.17 U/ml Dispase (Life Technologies/Thermo Fisher Scientific). After rocking 30 min at 37°C, 10% HBSS with 5 mM EDTA was added and cells were obtained using a 100-μm filter. Lamina propria lymphocytes were obtained from 40/80% Percoll gradient centrifugation. Flow cytometry gating strategy is illustrated in Supplemental Fig. 1.

For T cell activation assays, C57BL/6 splenocytes were labeled with 5 μM CTV (Life Technologies), and 0.25 × 10⁶ cells were cultured in anti-CD3–coated (10 μg/ml) 96-well plates with soluble anti-CD28 (2 μg/ml). BMDCs were generated from 6-d cultures of femur and tibia cells using 20 ng/ml GM-CSF in 10% FBS. BMDCs were cocultured with T cells in the presence or absence of EPS (5 μg/ml) for 4 d, and T cell activation was analyzed by flow cytometry.

Cell lysates were homogenized in lysis buffer containing 20 mM Tris-HCl (pH 7.4), 0.14 M NaCl, 1% Triton X-100, and protease inhibitors, sodium orthovanadate (1 mM), and sodium fluoride (10 mM). Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes. Immunoreactive bands were detected by ECL. For reverse transcription-quantitative real-time PCR (RT-qPCR), RNA was isolated with TRIzol (Thermo Fisher Scientific), and cDNA was synthesized using the GoScript reverse transcription system (Promega). RT-qPCR was conducted using primer sets for ido1 (forward primer, 5′-TTGCTGTTCCCTACTGCGAG-3′, reverse primer, 5′-AGAAGCTGCGATTTCCACCA-3′) and iTaq Universal SYBR Green master mix (Bio-Rad).

EPS was fluorescent labeled with CF aminooxy-reactive group (Biotium, San Francisco, CA). Fluorescent-labeled EPS (fl-EPS) was washed with PBS using 3-kDa ultrafiltration vials. Cells treated with fl-EPS (5 μg/ml) were examined by microscopy with a DeltaVision widefield fluorescence microscope or by flow cytometry.

Cells were treated with anti-CD16/32 Fc Block (BioLegend) and after Ab staining were analyzed on FACSCanto II or LSRFortessa flow cytometers (BD Biosciences). Data were analyzed using FlowJo software (Tree Star, Ashland, OR). Surface receptor expression was quantified by mean fluorescence intensity (MFI). For TLR4, CD14, and CD11b receptor modulation studies, surface receptor percentage at the indicated time points was determined as the ratio of the MFI values measured from the stimulated cells to those measured from the unstimulated cells.

The percentage surface expression and time were plotted to reflect receptor endocytosis. To measure the TLR4/MD-2 dimerization, the percentage of the TLR4/MD-2 dimer was calculated by 100% minus the percentage of the TLR4/MD-2 monomer. The percentage of the TLR4/MD-2 monomer was measured by the ratio of MFI values (clone MTS510) of the stimulated cells to those of unstimulated cells. Levels of proinflammatory cytokines and chemokines were measured using cytokine bead array (CBA) (BD Biosciences), according to the manufacturer’s protocol. Samples were measured on an LSRFortessa (BD Biosciences) and analyzed using FCAP Array software v3.0 (BD Biosciences).

After splenocytes from OTII mice (CD45.2) were incubated with biotinylated anti-CD25, CD25⁺ cells were removed with streptavidin magnetic beads. CD25⁻ OTII splenocytes (10⁷) were transferred i.v. to CD45.1 wild-type (WT) mice, and 2 d later mice were gavaged with 10⁸ CFU of OVA–C. rodentium. At the peak of the disease (day 11), mice were sacrificed and the lamina propria cells were analyzed for the presence of OTII T cells by flow cytometry. The OVA–C. rodentium strain was gifted by Daniel Mucida at the Rockefeller University (New York, NY).

Previous data showed that EPS protects mice from C. rodentium–induced colitis (7). As colitis is a T cell–mediated disease, we hypothesized that EPS inhibits excessive T cell responses in these mice. To test this, we asked whether EPS inhibits activation of pathogen-specific T cells in infected mice. Because of the difficulty to identify C. rodentium–specific T cells, we transferred CD25⁻ splenocytes from CD45.2 OTII transgenic mice (in which most T cells recognize a peptide of chicken OVA) into CD45.1 naive WT mice. Two days later, we injected EPS (2.5 mg/kg) i.p. and then infected mice with OVA–C. rodentium 1 d later. During the peak of disease, 11 d postinfection, colonic lamina propria cells were analyzed by flow cytometry for OVA–C. rodentium-specific (CD45.2⁺) T cells. Mice treated with EPS had significantly fewer of these T cells compared with control (nontargeting [NT]) mice that received only OVA–C. rodentium (Fig. 1A), suggesting that EPS blocks the expansion of OVA–C. rodentium-specific T cells. These findings are consistent with the previous observation that EPS prevents or alleviates T cell–mediated diseases, as well as inhibits the release of proinflammatory cytokines by anti-CD3 in vivo (7).

To understand how EPS inhibits T cell activation and proliferation, we tested which colonic cells take up EPS. We injected fl-EPS i.p. into WT mice and analyzed colonic DCs (CD11c⁺), macrophages (F4/80⁺), and lymphocytes for EPS uptake 16 h later. By flow cytometry, lymphocytes were EPS⁻, whereas macrophages and DCs were EPS⁺ (Fig. 1B, 1C). This was recapitulated ex vivo by isolating colonic cells and incubating them with fl-EPS for 16 h. Again, by flow cytometry, lymphocytes were EPS⁻ (data not shown), whereas most (65–80%) macrophages (F4/80⁺) and DCs (CD11c⁺) were EPS⁺ (Fig. 1D). These data indicate that EPS is not bound to or taken up by B or T lymphocytes, but instead interacts with other immune cells, including macrophages and DCs.

Up to 60% of intestinal DCs take up EPS (Fig. 1D), and because DCs are critical APCs that activate T cells, we tested whether EPS could inhibit DC activation of T cells. Similar to colonic DCs, BMDCs also take up EPS (Fig. 2A), and we loaded these cells with OVA peptide and cocultured them with CTV-labeled OTII splenocytes with and without EPS. After 4 d, T cell activation and proliferation were analyzed by flow cytometry, and we found, as expected, that CD4⁺ OTII T cells were activated (CD25⁺) and had proliferated (Fig. 2B, 2C) after culture with OVA-loaded BMDCs. However, in cultures with EPS, CD4⁺ OTII T cells did not proliferate (Fig. 2B, 2C) and T cell activation (CD25 and CD44) was significantly decreased (Fig. 2D, 2E). These data indicate that EPS inhibits BMDC-mediated Ag-specific T cell activation and proliferation.

To ensure that EPS was not acting directly on T cells, we stimulated T cells with anti-CD3 and anti-CD28 with and without EPS and found that T cells were similarly activated and proliferated in both conditions (Fig. 2F, 2G), confirming that EPS does not act directly on T cells. T cell proliferation was not inhibited by BMDCs alone, but it was inhibited with both BMDCs and EPS present (Fig. 2F, 2G). We conclude that EPS induces inhibitory BMDCs, which inhibit pan–T cell activation/proliferation.

We sought to understand the mechanism by which EPS induces inhibitory BMDCs. EPS requires IDO and TLR4 to inhibit MLRs (10). The expression of IDO can be upregulated through either the canonical or noncanonical NF-κB pathway downstream of TLR4 signaling (12). By Western blot analysis, phosphorylation of the canonical pathway transcription factor p65 increased in EPS-treated BMDCs in a time-dependent manner (Fig. 3A), indicating that EPS activates the canonical NF-κB pathway. To determine whether EPS also activates the noncanonical NF-κB pathway, we tested whether the transcription factor RelB translocates to the nucleus after BMDC stimulation with EPS. Western blot analysis of nuclei from EPS-treated BMDCs revealed that EPS treatment led to nuclear translocation of RelB (Fig. 3B), showing that EPS activates the noncanonical NF-κB pathway. Because the noncanonical NK-κB pathway can induce the expression of anti-inflammatory molecules such as IDO (12, 13), we tested whether this pathway was required for T cell inhibition. We used BMDCs from RelB and NIK knockout (KO) mice, and using the OVA/OTII culture system, we found that EPS retained the capacity to inhibit T cell proliferation (Fig. 3C, 3D). These data indicate that the noncanonical NF-κB pathway is not required for EPS induction of inhibitory BMDCs.

DCs can inhibit T cell activation and proliferation by several mechanisms that include the secretion of anti-inflammatory cytokines (14–16), the induction of regulatory T cells (17–19), the generation of inhibitory metabolites (20), or the downregulation of stimulatory receptors (21). Based on our previous studies, we know that EPS-treated BMDCs do not require IL-10 or PD-L1 for inhibiting the proliferation of alloreactive T cells, but they do require expression of immunosuppressive IDO1 (10). In this study, we tested whether IDO is also needed for inhibiting proliferation of Ag-specific T cells. Using the OVA/OTII activation system, we added the IDO inhibitor 1-methyl l-tryptophan to BMDCs and found by flow cytometry that EPS lost the capacity to inhibit T cell proliferation (Fig. 4A, 4B). We also used KO mice and found that in contrast to EPS-treated WT BMDCs, EPS-treated IDO KO BMDCs failed to inhibit T cell proliferation (Fig. 4C, 4D), indicating that EPS makes use of IDO expression by BMDCs for inhibiting T cell proliferation in our model.

We then tested whether EPS upregulates IDO expression in cultures of OVA-loaded WT BMDCs and splenocytes from OTII mice. By Western blot analysis, IDO expression increased at both 24 and 48 h when coincubated with EPS (Fig. 4E). To test whether this increase was due to EPS acting on BMDCs, EPS was incubated solely with BMDCs, and we found an increase of both IDO protein and mRNA expression by 7 h posttreatment (Fig. 4F, 4G).

IDO has been shown to inhibit T cell proliferation by two mechanisms: tryptophan degradation and/or kynurenine generation/accumulation. The depletion of tryptophan activates the ribosomal kinase GCN2, which in turn triggers the integrated stress response in response to amino acid withdrawal. In TCR-activated T cells, the integrated stress response blocks cell cycle entry and thus inhibits proliferation of activated T cells (22). Kynurenines, in contrast, suppress T cells by binding to the PDK1 or AhRs in T cells, APCs, or other immune cells (23). Metabolism of tryptophan by IDO can also activate the IDO/kynurenine/AhR circuit. This circuit consists of catabolism of tryptophan by IDO into kynurenine, which is an activating ligand for the transcriptional factor AhR in DCs. This transcriptional factor, in turn, transcribes ido1 and increases the expression of IDO in APCs, resulting in a positive feedback loop of immunosuppression.

To determine which mechanisms are involved in EPS-induced inhibition of T cell activation, we first tested whether tryptophan depletion is responsible for this inhibition in vitro. We added tryptophan to the cultures of OTII cells and OVA-loaded BMDCs and found that EPS still inhibited T cell proliferation (Fig. 5A), suggesting that degradation of tryptophan by IDO is not the primary mechanism by which EPS-treated BMDCs inhibit T cells. To test whether AhR expression in BMDCs is needed for the inhibition, we generated BMDCs from AhR KO mice and found that after treatment with EPS, these BMDCs did not inhibit T cell activation (Fig. 5B, 5C). These data suggest that EPS utilizes this IDO/kynurenine/AhR pathway to induce immunosuppression.

We tested whether IDO is needed for EPS inhibiting T cell stimulation in vivo by measuring levels of serum T cell cytokines after injection of IDO KO mice with EPS followed with a low dose (0.25 mg/kg) of anti-CD3 3 d later. CBAs of serum IFN-γ, IL-2, and IL-6 2 h after injection of anti-CD3 showed that in contrast to the reduced levels of T cell cytokines in WT mice treated with EPS, the IDO KO mice treated with EPS had unchanged levels of serum T cell cytokines (Fig. 6). These data indicate that in this model system, EPS makes use of IDO to suppress T cell cytokine production in vivo.

EPS and LPS both use TLR4 to activate cells, but they lead to different outcomes (5–11, 24). To directly compare the cytokine responses, we injected i.p. LPS or EPS and measured serum cytokines 1–24 h later (Supplemental Fig. 2). As expected, LPS induced the release of several inflammatory cytokines in a time-dependent manner. In contrast, EPS induced little to no detectable levels of these cytokines, clearly showing the different outcomes of EPS and LPS administration, even though both molecules signal through TLR4. Both EPS and LPS induced IL-10, as previously shown for LPS (25).

LPS activation of cells through TLR4 results in TLR4 dimerization (26) and internalization of CD14 (27). Dimerization of TLR4 is often measured by the loss of the monomeric form of TLR4 using an Ab that recognizes only monomeric TLR4 (28). Because many of these studies used the RAW246.7 macrophage cell line, we also used these cells incubated with EPS in a time-course study. The loss of the monomeric form as measured by flow cytometry was used to calculate dimerization of TLR4 as described (28) (Fig. 7A). We found that as with LPS, TLR4 appears to be dimerized by EPS. Internalization of CD14 was also measured by the loss of surface CD14 expression after treatment with EPS or LPS. In contrast to LPS, CD14 was not internalized by EPS (Fig. 7B), suggesting that LPS and EPS differentially activate cells through TLR4. CD14 is essential for the proinflammatory response of LPS, and we hypothesize that EPS uses other coreceptors that shape the TLR4 signaling response into an anti-inflammatory response.

One receptor known to bind carbohydrates and inhibit proinflammatory signaling through TLR4 is the integrin CD11b (29–32). Using fl-EPS, we first showed that EPS is taken up by RAW264.7 macrophages and localized to the endosomal compartment (Lamp1⁺) (Fig. 7C). We then tested whether CD11b was internalized by using Abs to detect changes in levels of expression of surface and cytoplasmic CD11b after treatment with EPS (Fig. 7D, 7E). We found that the expression of surface CD11b decreased whereas the amount of intracellular CD11b increased upon EPS treatment. These data show that both EPS and CD11b are internalized after treatment with EPS.

Commensals and probiotics have beneficial effects on health. However, the molecular mechanisms by which these organisms modulate the immune system are largely unknown. The most prominent example of a known molecule from commensals or probiotics is the polysaccharide A (PSA) from Bacteroides fragilis, which has been shown to be protective and therapeutic in murine colitis, multiple sclerosis, and viral infection (33–37). In this study, we investigated the mechanism of the molecule EPS from the probiotic B. subtilis, which has been shown to prevent allergic eosinophilia, GVHD, colitis caused by an enteric pathogen, and disease caused by systemic infection of Staphylococcus aureus (5–11). This study adds to our understanding of how microorganisms modulate the immune system to suppress inflammation.

Our study reveals a new mechanism of immune suppression whereby an EPS from a Gm(+) probiotic induces tolerogenic DCs that inhibit T cell activation. Previously, we showed that EPS prevented enteric inflammation following infection with C. rodentium (5, 7, 8), and that EPS inhibits in vivo activation of T cells following injection of anti-CD3 (7). EPS does not directly act on T cells, as evidenced by the absence of detectable EPS binding to T cells as well as the findings that in vitro activation of T cells by anti-CD3 or of alloreactive T cells in MLRs (10) were only inhibited by EPS in cultures in which BMDCs were present. In this study, we showed that BMDCs, as well as lamina propria DCs and macrophages, directly bind or take up EPS, and that in vivo, EPS inhibits the expansion of C. rodentium–specific T cells postinfection with C. rodentium. We think that B. subtilis EPS prevents colonic inflammation postinfection with C. rodentium by inducing anti-inflammatory lamina propria DCs and/or macrophages, which inhibit activation and expansion of C. rodentium–specific T cells.

EPS activates both the canonical and noncanonical NF-κB signaling pathways, and it increases expression of the immunosuppressive enzyme IDO. This anti-inflammatory signaling appears to use the IDO/kynurenine/AhR circuit, as EPS did not induce inhibitory BMDCs when these cells lack IDO or the transcriptional factor AhR. The immunosuppressive function of IDO is generally associated with its enzymatic activity, but it can also mediate immunosuppression independent of enzyme activity (38). In the case of IDO inhibition of T cell proliferation by EPS in vitro, this inhibition is likely due primarily to the enzymatic function because T cell proliferation is restored in the presence of the enzyme inhibitor 1-methyl l-tryptophan. In vivo, EPS also needed IDO to inhibit T cells activated with anti-CD3, but in this case we do not know whether this is due to the enzymatic or nonenzymatic function of IDO. In other studies, the expression of IDO has been shown to limit T cell function and engage in immune tolerance (23), and in cancer, IDO generates a suppressive environment that limits T cell activation. We suggest that EPS can generate inhibitory DCs that block or diminish T cell activation, which could be beneficial in inflammatory diseases where there is aberrant activation of T cells.

After showing that injection of EPS induces little or no inflammatory cytokines compared with LPS, we asked how this could happen when both EPS and LPS signal through TLR4. Both LPS and EPS dimerize TLR4, but the dimerization of TLR4 by LPS is dependent on its lipid components, and we have no evidence that our EPS preparation contains lipids. Furthermore, whereas LPS internalizes the coreceptor CD14, which is required for LPS activation (27), EPS does not internalize CD14, indicating that EPS activates TLR4 signaling via a different mechanism than for LPS. We hypothesize that EPS either does not use a TLR4 coreceptor, or it uses a coreceptor other than CD14 for activation. Several such receptors, including CD83 (39), dectin-1 (40), CD11b (30), and RP105 (41, 42), have been shown to dampen the proinflammatory signaling of TLR4. We tested two of these, RP105 and CD11b, and whereas RP105 did not cointernalize with EPS on BMDCs (data not shown), we found that surface CD11b expression on RAW264.7 cells was modulated by EPS. LPS also internalizes CD11b in bone marrow–derived macrophages (27). CD11b generates an anti-inflammatory response by suppressing the proinflammatory signaling of TLR receptors, including TLR4 (30, 31). This process could be part of an autoinhibitory mechanism that prevents the overactivation of a proinflammatory response. We hypothesize that LPS signals through CD14 and CD11b, but CD14-dependent signaling dominates. In contrast, we speculate that EPS activates CD11b and not CD14, leading to CD11b-dependent signaling that results in an anti-inflammatory response instead of a proinflammatory response. Studies to directly test this idea are needed.

The injection of both EPS and LPS into mice upregulated the expression of IL-10. Whereas IL-10 induced by LPS may serve to attenuate inflammation induced by LPS, EPS does not induce an inflammatory response. Although IL-10 induced by EPS might contribute to the anti-inflammatory activity of EPS, in previous studies, neutralizing anti–IL-10 Ab did not attenuate the inhibitory activity of EPS in GVHD (10), or of M2 macrophages induced by EPS (7). It remains unclear whether IL-10 contributes to the anti-inflammatory effect of EPS.

The molecular mechanisms of how probiotics function and the active molecules required for their health benefits are not well understood. To our knowledge, this is the first study to demonstrate a molecular mechanism by which Gm(+) bacterial polysaccharide can generate an anti-inflammatory response. Moreover, this study demonstrates that a polysaccharide can activate TLR4. EPS induces the expression of the immunosuppressive enzyme IDO in BMDCs, and supplementation of tryptophan in the medium did not reverse the effects of EPS. We hypothesize that this IDO metabolizes the essential amino acid tryptophan into kynurenine, and that kynurenine is responsible for the anti-inflammatory activity of EPS. Although studies to test for EPS-induced increases in kynurenine are needed, several species of colonic commensals have been shown to produce kynurenines, and further that they can activate the transcription factor AhR in host cells (43). However, it was previously not known whether commensal microorganisms could stimulate the expression of IDO in host cells. Our study suggests that B. subtilis can activate the expression of IDO in host cells, and we speculate that by expressing IDO and generating kynurenines, DCs can readily control T cell activation and, potentially, regulatory T cell formation.

We conclude that EPS from B. subtilis induces inhibitory BMDCs that inhibit T cell activation by activating TLR4 signaling and upregulating IDO expression. In contrast to LPS, EPS signaling does not appear to use CD14 as a coreceptor, and instead, we suggest that EPS uses other coreceptors such as CD11b to induce the anti-inflammatory response. By identifying the active molecules from probiotics and understanding their molecular mechanisms of action, we will be able to generate a new class of therapeutics to treat and prevent inflammatory diseases.

The authors have no financial conflicts of interest.

We thank Dr. Liang Zhou at the University of Florida, Dr. Christoph Vogel at the University of California at Davis, Dr. Irving Allen at Virginia-Maryland College of Veterinary Medicine, and Dr. David Corry from Baylor College of Medicine for the generosity and willingness to share with us bones from KO mice for our studies. We also thank Dr. Derek Wainwright (Loyola University Chicago) and Maile Hollinger (The University of Chicago, Chicago, IL) for helpful discussions.