Mucosal surfaces are in continuous contact with microbes. Toll-like receptors (TLRs) mediate recognition of microbial molecules to eliminate pathogens. In contrast, we demonstrate here that the prominent gut commensal, Bacteroides fragilis, activates the TLR pathway on T lymphocytes to establish host-microbial symbiosis. TLR2 deletion on CD4+ T cells results in anti-microbial immune responses that reduce B. fragilis colonization of a unique mucosal niche during homeostasis. A symbiosis factor (PSA) of B. fragilis activates TLR2 directly on Foxp3+ regulatory T cells through a novel process to engender mucosal tolerance. Remarkably, B. fragilis lacking PSA is unable to restrain host immune responses and is defective in niche-specific mucosal colonization. Therefore, unlike pathogens whose TLR ligands trigger inflammation, some commensal bacteria exploit the TLR pathway to actively suppress immune reactions. We propose that the immunologic distinction between pathogens and the microbiota is mediated not solely by host mechanisms, but also through specialized molecules evolved by symbiotic bacteria that enable commensal colonization.

The mucosal immune system employs an arsenal of responses against enteric pathogens including Th17 cells. Intriguingly, Th17 cell development is intimately linked to the presence of the microbiota, as germ-free mice lack these cells in the small intestine and colon (refs. (12–13) and Fig. 1A). Unlike most gut microbes, segmented filamentous bacteria (SFB) appear to have the unique property of inducing Th17 cells (14–15). Since most bacteria share common microbial patterns, how do other symbiotic bacteria avoid triggering intestinal immunity? Based on recent findings that specific members of the microbiota have potent immunomodulatory capacities (2), we explored the hypothesis that the prominent gut commensal Bacteroides fragilis evolved molecular mechanisms to suppress Th17 responses during homeostatic colonization. As predicted by previous studies (12, 15–16), we find that B. fragilis mono-associated animals do not appreciably induce Th17 cell development in the colon compared to germ-free controls (Fig. 1A). Remarkably however, colonization with B. fragilis in the absence of PSA (B. fragilisΔPSA) results in significant Th17 cell responses in the gut (Fig. 1A and 1B). Cells isolated from the colonic lamina propria (LP) of B. fragilisΔPSA mono-associated animals display increased secretion of IL-17A in vitro (fig. S1) and elevated transcription of il17a and rorγc, encoding for the Th17-specific lineage differentiation factor RORγt (Fig. 1C and 1D, respectively). Immune reactions were specific, as Th1 cell-associated responses were not observed during B. fragilisΔPSA colonization (fig. S2). We purified CD4+ T cells from MLNs and assayed IL-17A production during in vitro Th17 skewing assays (TCR ligation in the presence of TGF-β and IL-6) (17). Cells from B. fragilis mono-associated animals produced low levels of IL-17A, even in the presence of TGF-β and IL-6, while CD4+ T cells from B. fragilisΔPSA animals displayed elevated levels of IL-17A similar to conventionally colonized mice (Fig. 1E). We next determined whether purified PSA can suppress mucosal immunity to commensal bacteria. Oral treatment with PSA reduced Th17 cell proportions in SFB mono-associated animals (Fig. 1F). Most significantly, administration of PSA to B. fragilisΔPSA mono-associated animals greatly suppressed Th17 cell responses compared to control (PBS) treatment (Fig. 1G). These data demonstrate that unlike SFB, B. fragilis has evolved a molecular mechanism to actively restrain Th17 cell responses during homeostatic intestinal colonization.

Recent studies have shown that certain gut bacteria can promote regulatory T cell (Treg) induction (16). Furthermore, Tregs expressing the transcription factor Foxp3 (forkhead box P3) suppress pro-inflammatory Th17 cell reactions. Since B. fragilis has evolved the ability to induce anti-inflammatory gene expression from Foxp3+ Tregs in the intestine (9), we speculated Tregs may prevent immune responses during B. fragilis colonization. To test this hypothesis, germ-free RAG−/− mice (lacking T and B cells) were reconstituted with bone marrow from Foxp3-DTR donors. Under these conditions, all Foxp3-expressing Tregs also express the diphtheria toxin receptor (DTR) and can be specifically ablated by administration of diphtheria toxin (DT) (18). Reconstituted mice were mono-associated with B. fragilis to induce development of Tregs (Fig. 2A). Animals were subsequently treated with DT and colonic lamina propria Treg and Th17 cell development was assessed. DT treatment of mice results in almost complete ablation of Foxp3+ T cells (Fig. 2, A and B). Interestingly, when Tregs are ablated in B. fragilis mono-associated mice, Th17 cellular responses are increased (Fig. 2, C and D and fig. S3), demonstrating that Foxp3+ Tregs are required for suppression of Th17 cells during B. fragilis colonization.

To examine whether PSA directly impacts Treg function through cell-intrinsic TLR2 signaling, we treated purified CD4+Foxp3+ T cells with various TLR2 ligands. The proportion of Foxp3+ T cells was equivalent under all in vitro conditions (fig. S7). To measure Treg suppressive functions, proliferation of non-Treg responder T cells was assayed following co-culture with Tregs stimulated with media, PSA, Pam3CysK, or Pam2CysK. PSA-treated Tregs display increased suppressive capacity compared to media or other TLR2 ligand-treated Treg cultures (Fig. 3D and E). Remarkably, PSA treatment during the culture period maintained Treg activity similar to freshly isolated Foxp3+ Tregs (Fig. 3E; compare PSA to naïve). IL-2 was not produced by Tregs during treatment with any of the TLR2 ligands (data not shown), and IL-2 neutralization had no effect on in vitro suppression (fig. S8). Importantly, the suppressive capacity of PSA-treated Tregs is completely lost when Foxp3+ T cells are deficient in TLR2, indicating that PSA can uniquely induce Treg activity through TLR2 signaling (Fig. 3D and E). PSA directs the development of inducible Foxp3+ Tregs by promoting the expression of IL-10, TGF-β2 and Foxp3 (11). We show here that PSA induced these genes from purified CD4+Foxp3+ Tregs, unlike the prototypical TLR2 ligand Pam3CysK (Fig. 3F and fig. S9). PSA did not however promote anti-inflammatory gene expression from TLR2−/− Foxp3+ Tregs (Fig. 3F and fig. S9). Collectively, these studies show that unlike other TLR2 ligands, PSA potently enhances Treg function and gene expression in the absence of APCs through TLR2 signaling directly on CD4+Foxp3+ Treg cells.

As PSA is required for suppression of Th17 cell responses in vivo, we wished to determine if PSA signals through TLR2 on Foxp3+ Tregs during B. fragilis colonization. Germ-free RAG−/− animals were reconstituted with purified CD4+ T cells from WT or TLR2−/− animals. Animals were subsequently mono-associated with either B. fragilis or B. fragilisΔPSA, and initially Treg profiles were analyzed. Unlike WT animals (TLR2 sufficient) that up-regulate IL-10 expression from Foxp3+ Tregs upon B. fragilis colonization (fig. S10 and (11)), IL-10-producing Tregs are not induced in the gut of B. fragilis mono-colonized animals lacking TLR2 on CD4+ T cells (fig. S10). We sought to determine whether TLR2 signaling on T cells is required for suppression of Th17 responses during commensal colonization. While we observed minimal Th17 cell development in WT mice mono-associated with B. fragilis, significantly increased Th17 cell responses were elicited during colonization of animals with T cell-specific TLR2 deletion (Fig. 4, A and B). Indeed, Th17 cell proportions were similar in animals colonized with B. fragilisΔPSA, regardless of whether the cells have TLR2 expression (Fig. 4, A and B). Consistent with our findings in the colon, considerably more IL-17A is produced in vitro by T cells from MLNs of mice reconstituted with TLR2-deficient T cells compared to animals containing WT CD4+ T cells (Fig. 4C). Collectively, these data demonstrate that B. fragilis actively suppresses mucosal immunity during colonization through engagement of TLR2 directly on CD4+ T lymphocytes.

The microbiota occupies both mucosal and luminal niches during normal colonization of the gut; however, the biogeographic distributions of specific microbial species are poorly characterized. Based on our findings, we reasoned that the functions of PSA were driven by an evolutionary impetus to prevent mucosal immune reactions against B. fragilis, thus enabling a population of bacteria to contact or associate with host tissue. We directly tested this hypothesis. When colonization levels of B. fragilis and B. fragilisΔPSA were determined by microbiological plating of feces for live bacteria, there appears to be no difference in bacterial levels in the gut lumen up to two months post-colonization (fig. S11) (19). Remarkably, we find a population of bacteria that intimately associates with intestinal crypts (Fig. 5A) and therefore in close proximity to host tissue. Whole mount preparations of medial colons were probed for bacteria following labeling with B. fragilis specific anti-sera (fig. S12). 3-dimensional reconstruction of confocal microscopic images revealed micro-colonies of B. fragilis residing within colonic crypts (Fig. 5A). The levels of B. fragilis associated with host tissue represent a fraction of total bacteria as determined by bacterial plating and qRT-PCR (Fig. 5, B and C, respectively), but likely are an important population (approximately 1 million bacteria per animal) that are in direct contact with the host immune system. We speculated that due to close proximity to host tissues, only mucosally-associated bacterial levels would be affected by the host immune responses (such as Th17 cell induction following B. fragilisΔPSA colonization). Colonization levels of B. fragilis at the mucosal surface were determined in WT B. fragilis and B. fragilisΔPSA mono-associated animals. Interestingly, animals colonized with B. fragilisΔPSA display profoundly reduced numbers of tissue-associated bacteria when compared to animals colonized with WT B. fragilis (Fig. 5D). Additionally, treatment of B. fragilisΔPSA colonized animals with purified PSA complemented this defect and increased colonization of the B. fragilisΔPSA mutant to wild-type levels (Fig. 5D). Finally, to determine B. fragilis colonization dynamics in the presence of other Th17 inducing organisms, animals that were co-associated with SFB and WT B. fragilis or B. fragilisΔPSA. Indeed, B. fragilis colonized the mucosal tissues of animals even in the presence of SFB, while the B. fragilisΔPSA mutant remained poorly associated with host tissue (Fig. 5E). Collectively, these data identify a previously unappreciated mucosal niche for B. fragilis, and reveal that PSA is required for maintaining host-bacterial symbiosis in the colon.

Our studies suggest that PSA induces Tregs through TLR2 signaling to suppress Th17 cell responses and promote mucosal colonization by B. fragilis. To test this model, we measured colonization levels of B. fragilis in animals with TLR2-deficiency in T cells (RAG−/− mice reconstituted with CD4+ T cells). Remarkably, when animals lack TLR2 expression on CD4+ T cells, tissue association by WT B. fragilis in the colon was reduced to the levels of B. fragilisΔPSA (Fig. 5F and fig.S13), demonstrating that TLR2 expression on T cells promotes microbial symbiosis. Accordingly, Foxp3+ Treg ablation in B. fragilis mono-associated animals resulted in significantly reduced mucosal levels of B. fragilis when compared to animals that contain Tregs (Fig. 5G). Bacterial numbers in the lumen of the gut were not altered by Treg ablation (fig. S14), showing the sensitivity of only tissue-associated bacteria to mucosal immune responses. Finally, to functionally determine the role of IL-17 responses on mucosal association, B. fragilisΔPSA mono-associated animals were treated with a neutralizing antibody to IL-17A. While levels of B. fragilisΔPSA in isotype control treated animals remained low, neutralization of IL-17A results in a 1,000-fold increase in mucosally-associated bacteria (Fig. 5, H and I). These data indicate that IL-17 suppression by PSA is a critical mechanism by which B. fragilis associates with its host. Therefore, unlike pathogens that trigger inflammatory responses through TLRs that result in immune responses to clear infections, symbiotic colonization by B. fragilis is actually enhanced by signaling via the TLR pathway promote suppression of Th17 immunity. We conclude that PSA evolved to engender host-bacterial mutualism by inducing mucosal tolerance through TLR2 activation of Treg cells.

As the gastrointestinal tract represents a primary portal for entry by numerous pathogens, the mucosal immune system has evolved mechanisms to detect and respond to infectious agents. Toll-like receptors recognize MAMPs expressed by bacteria and coordinate a cascade of innate and adaptive immune responses that lead to clearance of intestinal pathogens (3). Accordingly, mutations in the TLR pathway result in susceptibility to viral and bacterial infections. Although TLRs have classically been studied on innate immune cells, recent reports have demonstrated expression of these pattern recognition molecules by T cells in both mice and humans (6, 22–23, 25). As bacteria contain universally conserved MAMPs, how do commensal microbes avoid triggering TLR activation unlike pathogens? It is historically believed that the microbiota is excluded from the mucosal surface (26). However, recent studies have shown that certain symbiotic bacteria tightly adhere to the intestinal mucosa (14–16), and thus immunologic ignorance may not explain why inflammation is avoided by the microbiota. Therefore, the mechanism(s) by which the immune system distinguishes between pathogens and symbionts is unknown. We report herein that the symbiosis factor of Bacteroides fragilis, PSA, can directly induce the anti-inflammatory function of Tregs by signaling directly through TLR2 on CD4+ T cells. Modulation of Treg activity by PSA restrains intestinal Th17 cell responses during commensal colonization. The functional activity of PSA on Tregs contrasts with the role of TLR2 ligands of pathogens, which elicit inflammation, and thus reveals an entirely new function for TLR signaling during homeostatic intestinal colonization by the microbiota. Paradoxically, while engagement of TLR2 by previously identified ligands is known to stimulate microbial clearance of pathogens, signaling by PSA actually allows B. fragilis persistence on mucosal surfaces. These results identify PSA as the incipient member of a new class of TLR ligands termed ‘symbiont-associated molecular patterns (SAMPs)’ that function to orchestrate immune responses to establish host-commensal symbiosis. Based on the importance of the microbiota to mammalian health (27), evolution appears to have created molecular interactions that engender immunologic tolerance to symbiotic bacteria. In summary, our findings reveal the novel paradigm that the host is not ‘hard-wired’ to intrinsically distinguish symbionts from pathogens, and microbial-derived mechanisms have evolved to actively promote mutualism between mammals and beneficial bacteria of the microbiota.