Glycoantigens Induce Human Peripheral Tr1 Cell Differentiation with Gut-homing Specialization

Abstract

The carbohydrate antigen (glycoantigen) PSA from an intestinal commensal bacteria is able to down-regulate inflammatory bowel disease in model mice, suggesting that stimulation with PSA results in regulatory T cell (Treg) generation. However, mechanisms of how peripheral human T cells respond and home in response to commensal antigens are still not understood. Here, we demonstrate that a single exposure to PSA induces differentiation of human peripheral CD4⁺ T cells into type-Tr1 Tregs. This is in contrast to mouse models where PSA induced the production of Foxp3⁺ iTregs. The human PSA-induced Tr1 cells are profoundly anergic and exhibit nonspecific bystander suppression mediated by IL-10 secretion. Most surprisingly, glycoantigen exposure provoked expression of gut homing receptors on their surface. These findings reveal a mechanism for immune homeostasis in the gut whereby exposure to commensal glycoantigens provides the requisite information to responding T cells for proper tissue localization (gut) and function (anti-inflammatory/regulatory).

Introduction

Over the past decade, the importance of the indigenous microbiota in the development and maintenance of intestinal health has risen to the fore. Colonization of the gut regulates intestinal angiogenesis via recognition of microbial molecules by Paneth cells (1). Upon monocolonization of gnotobiotic animals, capsular polysaccharides from Bacteroides thetaiotaomicron induce IgA production, leading to reduced proinflammatory signals in the gut (2). The symbiotic relationship between host and commensal flora educates normal immune system development in that at least one commensal antigen, PSA from the capsule of Bacteroides fragilis, can correct the abnormal T_H1²/T_H2 balance in gnotobiotic animals (3). In murine models of human disease, this MHCII (class II major histocompatibility complex)-dependent (4) T cell-activating commensal carbohydrate antigen (glycoantigen; GlyAg) reduces inflammation and ameliorates signs of bowel disease in colitis models (5, 6) and prevents demyelination in the experimental autoimmune encephalomyelitis model of multiple sclerosis (7) through a T cell-dependent mechanism. These findings suggest that exposure to GlyAg results in tolerogenic or suppressive T cells in mice, and possibly humans.

It is well established that regulatory T cells suppress systemic and mucosal immune activation to control intestinal inflammation. The mechanism remains a matter of intensive investigation, but maintenance of peripheral tolerance has been demonstrated to be critical for maintaining balance of immunity. The three best characterized CD4⁺ Tregs are CD4⁺CD25⁺Foxp3⁺ thymically selected naturally occurring Tregs (nTregs), CD4⁺CD25⁺Foxp3⁺ peripherally induced T regs (iTregs) and CD4⁺ Foxp3⁻ peripherally induced T regulatory type 1 (Tr1) cells (8). The earliest descriptions of Tr1 cell properties link them to prevention of colitis in the T cell transfer model of colitis (9, 10) and the maintenance of tolerance to enteric flora (11, 12). Human Tr1 cells, isolated from tolerant transplant patients, were shown to produce large amounts of IL-10, are typically defined by their high IL-10:IL-4 ratio, and do not express constitutive levels of Foxp3 (13). Because suppressive Tr1 cells can be differentiated from IPEX patients who genetically lack functional Foxp3, it is known that Foxp3 is not required for all suppressive T cell activity in humans (14). Work done by Papadakis et al. (15) has demonstrated that mucosal T cells that express the CC chemokine receptor 9 (CCR9) have a suppressive Tr1 cell phenotype. CCR9, the receptor for CCL25, which is differentially expressed by T lymphocytes of the small intestine and colon, plays an important role in the migration of leukocytes to the intestinal mucosa. This and other gut-homing receptors have been shown to be present on lymphocytes found within the gut (15–17). Tr1 cells have also been shown to suppress responses toward nonharmful foreign antigens to which exposure but nonreactivity is critical. Anergic gluten antigen-specific Tr1 cells that produce large amounts of both IL-10 and IFNγ and that exhibit bystander suppressive activity have been isolated from the intestinal mucosa of patients with celiac disease (18), implicating gut-associated Tr1 cells with additional regulation of gut immune homeostasis.

Over the past decade, disparate diseases ranging from cancer and autoimmunity to microbial infections have been tied together through the common involvement of sugar moieties. Although carbohydrates have generally been considered mediators of innate immunity, MHCII-dependent GlyAgs (4, 19) such as PSA that contain a zwitterionic charge motif within an extended helical conformation (20–22) are rising to prominence due to the aforementioned ability to suppress inflammatory responses in animal models of autoimmunity in a T cell-dependent fashion (5, 7, 23–25). Experiments have shown that adoptive transfer of CD4⁺ T cells from PSA-immunized animals can provide nonspecific immunologic protection against bacterial-induced abscesses (25), surgical adhesion formation (26), inflammatory bowel disease (5), and experimental autoimmune encephalomyelitis-mediated demyelination (7) in naïve recipient animals, strongly suggesting that repeated stimulation of T cells with PSA results in suppressive T cells. Within the murine system, these T cells have been described as typical Foxp3⁺ iTregs, which produced IL-10 and TGF-β (6, 7); however, phenotypic and functional analyses of GlyAg-driven T cells in humans remains untouched.

In contrast to the mouse models, this study reveals that in humans, the commensal-derived GlyAg PSA induces Treg cells of the CD25⁺ Tr1 subset, which lack Foxp3 expression. These cells were found to express potent bystander suppressive activity mediated by IL-10, but not TGF-β. A single exposure to this antigen provoked CCR9 and other homing receptor up-regulation in the responding T cells, providing the required signals to facilitate Tr1 cell migration and longevity at the site of chronic commensal antigen exposure, the gut. These findings highlight the differences in the Treg compartment in the mouse and human immune systems, and provide the missing link between the reported Tr1 cell population found in the human gut, their association with intestinal homeostasis, and commensal bacterial antigens that coordinate these events to the benefit of the host.

EXPERIMENTAL PROCEDURES

Culture Medium and GlyAgs

Advanced RPMI (Invitrogen), supplemented with 10% Australian-produced, heat-inactivated fetal bovine serum was used for generation and maintenance of all cells. For culture and expansion of T cells, 20 units/ml IL-2 (Invitrogen) was used. B. fragilis PSA glycoantigen was prepared in-house as described previously (27). Circular dichroism (Jouan) analysis as described (21) demonstrated that the PSA displayed appropriate helical structure, and NMR using a 900-MHz wide field magnet and cryoprobe (Brucker BioSpin) at the Cleveland Center for Structural Biology confirmed the purity of the PSA.

Cell Isolation

Peripheral blood mononuclear cells (PBMCs) were isolated from healthy volunteers by Ficoll density gradient centrifugation (Lymphoprep, Sigma). Informed consent was received from subjects, and this work was approved by the Case Western Reserve University Institutional Review Board for human subjects. Cells were sorted into CD4⁺ and CD4⁻ using antibody-coated magnetic beads, according to the manufacturer's directions (Miltenyi Biotec, Bergisch Gladbach, Germany). CD4⁺ cells were used as unstimulated T cells. CD4⁻ cells were further sorted using antibody-coated CD3 beads (Miltenyi), and CD4⁻CD3⁻ cells were used as antigen-presenting cells (APCs). APCs were treated with 20 mg/ml mitomycin C (Sigma) for 30 min at 37 °C to inhibit cell division and washed with PBS prior to use.

Generation of CD4⁺ Tr1 T Cells

A single round of antigen stimulation, followed by a rest period, was performed on naïve cells to induce the Tr1 phenotype. PBMCs were isolated from healthy volunteers by Ficoll density gradient centrifugation and used immediately. For each Tr1 cell preparation, cells were cultured in Advanced RPMI plus 10% FBS, 20 units/ml IL-2, and 50 μg/ml PSA or 50 amount ovalbumin control (OVA, Sigma) for 7 days. At day 7, cells were washed in PBS and transferred to six-well plates at 3 × 10⁶ cells in 4 ml per well in media (without IL-2 or antigen) to rest for 10 days. CD4⁺ cells were isolated using antibody-coated magnetic beads, according to the manufacturer's directions (Miltenyi) and used for subsequent assays.

Proliferation and Anergy Assays

To assess T cell proliferation, 2 × 10⁵ APCs were cultured with 2 × 10⁵ CD4⁺ T cells (naïve or previously stimulated Tr1) in 200 μl Advanced RPMI with PSA or OVA antigen in 96-well plates to generate anergic or optimally stimulated control T cells. T cell proliferation was measured after 6 days with a 6-h pulse of [³H]thymidine (Amersham Biosciences). The medium was removed prior to pulse and stored at −150 °C for cytokine analysis. Cells were harvested to glass filter mats and analyzed in a scintillation counter (Microbeta, PerkinElmer Life Science). For coculture experiments to assess suppressive activity, an additional 2 × 10⁵ CD4⁺ Tr1 cells (or prior-stimulated OVA controls) was included in the 200-μl culture volume. All tests were carried out in triplicate. For blocking experiments, monoclonal anti-human IL-10 neutralizing antibody (eBioscience) was added at both 10 μg/ml and 20 μg/ml. To overcome anergy, 20 units/ml IL-2 was added to the coculture experiments.

Cytokine Analysis

For assessment of cytokine production, supernatants were collected 6 days after stimulation or restimulation of T cells with mature DCs and stored at −150 °C. Amounts of IL-10 and IFNγ were measured by ELISA using commercially available antibody pairs and standards according to the manufacturer's protocol (BD Pharmingen). All ELISAs were performed in triplicate. Quantification of the ELISA signal from the biotinylated detection antibody was through europium-conjugated streptavidin and time-resolved fluorescence detection in a Victor multilabel plate reader (PerkinElmer Life Science). Analysis of other cytokines was by simultaneous detection in a Bio-Plex cytokine array analysis (Bio-Rad).

Antibodies

Direct-labeled mAbs were used for flow cytometry: FITC-conjugated CD4; phycoerythrin-conjugated CD25, CD28, CD40L, CD62L, CD69, CD122, CTLA-4 (eBioscience), and CCR9 (BD Pharmingen). Appropriate mouse subclass-specific isotype control antibodies were used (eBioscience).

Flow Cytometry

Immunofluorescence staining was performed after washing the cells twice with PBS plus 0.5% FBS and 0.05% sodium azide (Sigma). Cells were incubated with FITC- and phycoerythrin-conjugated mAbs for 20 min at 4 °C, washed two times, and analyzed by flow cytometry (Accuri C6 cytometer, Ann Arbor, MI) with FCS Express analysis software (version 3; De Novo Software, Los Angeles, CA). For intracellular analysis of Foxp3 expression, T cells (or Panc-1 cells as a positive control; ATCC) were collected, washed, fixed/saponin-permeabilized (perm/fix solution, eBioscience), and stained with Foxp3-specific or IgG control antibodies included with the Foxp3 analysis kit (eBioscience).

Statistics

All data were analyzed using GraphPad Prism 5. Analysis of variance or Students t tests, as applicable, were used, and S.D. was calculated using GraphPad Instat 3.

RESULTS

GlyAg T Cell Stimulation Leads to a Nonproliferative Anergic State

GlyAg-specific CD4⁺ T cell responses have been described previously as T_H1 in rodent intraabdominal abscess and adhesion studies (23, 26, 28), whereas more recent findings in mice suggest that GlyAg-stimulated T cells serve a regulatory function in gut homeostasis (3, 5). Human CD4⁺ T cells are less well studied but are also known to proliferate (29) and produce high concentrations of IFNγ when stimulated with GlyAg. The canonical GlyAg is the capsular polysaccharide PSA expressed by the commensal organism B. fragilis, which is found in the colon of most mammals.

To understand the immunologic effects of chronic exposure to PSA, human PBMCs were used as the source for both CD4⁺ T cells and APCs. The APC population (CD3⁻CD4⁻) was treated with mitomycin C to prevent proliferation, cultured with untreated CD4⁺ T cells, and incubated with 50 μg/ml of either OVA or PSA. On day 6, supernatants were collected for cytokine and ATP analyses, and the cells were pulsed with [³H]thymidine for 6 h to measure proliferation. T cells showed strong proliferative responses to both OVA and PSA over no-antigen media control (Fig. 1, A and B; naive). In parallel assays, the cells were allowed to be stimulated for a total of 10 days, followed by 8 days of rest without antigen, and then restimulated with fresh mitomycin-treated APCs and antigen. After 6 days, supernatants were collected, and the cells were pulsed with thymidine as before. The second stimulation of these PSA-derived and OVA-derived T cell lines (PSA-T and OVA-T) showed significantly less proliferation in response to antigen (Fig. 1A). For PSA-T, the response was ∼5-fold less than the initial response; however, when cultured with 20 units/ml IL-2, the anergic phenotype of the PSA-T cells was partially broken (Fig. 1B). Addition of PSA with IL-2 further increased proliferation to a significant degree (Fig. 1B).

FIGURE 1.

GlyAg stimulation yields anergic CD4⁺ T cells. A, PSA- and OVA-activated naive T cells as measured by proliferation (raw cpm, left; fold change, right) upon initial exposure, yet PSA- and OVA-stimulated cell lines showed significantly decreased proliferation upon a second antigen exposure. B, addition of recombinant IL-2 partially reduced proliferation anergy of PSA-raised cell line. C, ATP production of naive CD4⁺ cells stimulated with PSA or OVA for 6 days as compared with fresh PBMCs demonstrates that neither PSA nor OVA stimulation induces cell death under these conditions. ATP production by PSA-T cells (D) or OVA-T cells (E) at days 0, 4, and 7 following restimulation with either PSA or OVA showed that these cells also remained viable upon second stimulation despite the reduced ability to proliferate. Fold change was calculated using the respective media-only controls. Error bars, S.D. with at least n = 3 for all panels.

To confirm that the antigen-stimulated T cell lines were viable, culture supernatants were assayed for ATP production as a measure of metabolic activity. Fresh PBMCs showed equal ATP production compared with naive cells stimulated with PSA or OVA for 6 days (Fig. 1C), indicating that the initial stimulation did not cause significant cell death. PSA-T and OVA-T cells following 8 days of rest were also assayed for ATP production at days 0, 4, and 7 following restimulation with antigen. Neither the PSA-T nor the OVA-T cells showed significant losses in cell viability (Fig. 1, D and E), indicating that the loss of proliferation upon second stimulation was due to anergy rather than cell death.

GlyAg Stimulation Yields Anergic IFNγ⁺CD4⁺ T Cells with Common Surface Markers

Supernatants from the naive, PSA-T, and OVA-T experiments described for Fig. 1 were analyzed for cytokine production using IFNγ ELISA. Naive T cells produced high concentrations of IFNγ upon stimulation with either OVA or PSA (Fig. 2A). Remarkably, following the first stimulation and 8 days of rest, the PSA-T cells show constitutive expression of IFNγ that did not change upon second stimulation (Fig. 2. A and B); however, addition of IL-2 to the culture media doubled IFNγ levels from the PSA-T line in the absence of further antigen-specific stimulation and 3-fold more IFNγ when PSA and fresh APCs were also added to the cells (Fig. 2B). In contrast, OVA-T cells strongly restimulated with OVA as measured by IFNγ (Fig. 2A). These data demonstrate that PSA-T cells but not OVA-T cells are anergic to antigen re-stimulation but that IL-2 breaks this anergy and induces IFNγ by at least 2-fold beyond the constitutive IFNγ levels.

FIGURE 2.

GlyAg-stimulated CD4⁺ T cells are IFNγ⁺. PSA- and OVA-activated naive T cells as measured by IFNγ production (raw pg/ml, left; fold change over no antigen, right) upon initial exposure. PSA-T cell lines maintained significant IFNγ production after first exposure and were unable to produce more cytokine upon second stimulation with PSA. In contrast, OVA-T cells did not maintain IFNγ production after the initial stimulation and restimulated in an antigen-specific manner as measured by IFNγ secretion when exposed a second time. B, addition of recombinant IL-2 only modestly reduced cytokine production anergy of the PSA-raised cell line. For a more complete cytokine profile (shown in log₂ scale), Bio-Plex assays were performed on culture supernatants from the first T cell exposure to antigen (C) and restimulation of the PSA-T (D) and OVA-T (E) cells. On first stimulation, both antigens stimulated production of varying amounts of IL-1β, IL-17, GCSF, GMCSF, and TNFα. OVA also stimulated a modest amount of IL-5 over background. As seen with the IFNγ, the PSA-T cells maintained significant cytokine production and failed to show significant increases in any of the cytokines upon second stimulation with PSA. The OVA-T cells showed a similar trend, only IL-1β, GCSF, and TNFα were increased upon a second OVA stimulation. These data demonstrate significant cytokine anergy of GlyAg-stimulated T cell populations. Fold change was calculated using the respective media-only controls. Where indicated, error bars are S.D. with at least n = 3.

To provide a comprehensive profile of the naive, PSA-T, and OVA-T cells, the culture supernatants were used in Bio-Plex Multianalyte Analysis (Bio-Rad) experiments. The initial naive T cell activation experiments with PSA and OVA showed IL-1β, IL-17, GCSF, GMCSF, and TNFα production (Fig. 2C). A number of cytokines including IL-6, IL-7, IL-8, IL-13, MCF1, and MIP1β were not found in significant quantities (data not shown). Of specific note, IL-2, IL-4, and IL-5 were also not detected in these experiments (Fig. 2C). The ratio of IL-10:IL-4 for the PSA-T cells (137 to 1) is consistent with a Tr1 cell classification, which requires a ratio of at least 8 to 1 (13). Following rest, the PSA-T cells show constitutive production of IL-1β, IL-17, GCSF, GMCSF, and TNFα (compare levels to naive T cells without antigen; Fig. 2C), and these levels did not change upon restimulation with PSA (Fig. 2D). As seen with the IFNγ ELISA, the OVA-T cells responded well to OVA restimulation by producing IL-1β, GCSF, and TNFα (Fig. 2E). Collectively, these data show significant anergy of PSA-driven T cells to multiple exposures to PSA but constitutive production of IFNγ and no IL-2, IL-4, or IL-5.

The PSA-T cells were also compared with naive CD4⁺ T cells using flow cytometry to discover the cell surface phenotype of PSA-responsive T cells (Fig. 3). CD11a, a component of LFA-1, which is involved with cellular adhesion and costimulation, was found on the cell surface to the same extent as naive CD4⁺ T cells (96.8 and 94.7% positive, respectively). CD40L is a member of the TNF superfamily and a common marker of activated T cells, whereas PD-1 functions as a down-regulator of T cell receptor-mediated responses. We found modest, but repeatable increases in CD4⁺ T cells expressing CD40L (1.1% naive and 6.0% PSA-T) and PD-1 (12.2% naive and 14.4% PSA-T). CD69 is a common marker of activated T cells, whereas CD71 plays a critical role in iron transport and is also known to be expressed by activated lymphocytes. We found large increases in the number of T cells expressing CD69 (0.9% naive and 12.1% PSA-T) and CD71 (2.5% naive and 20.8% PSA-T) in PSA-T cells compared with naive controls. CD28 is the ligand partner of the B7 molecules expressed on APCs and is critical for activation costimulatory signals in T cells. The PSA-T cells showed no significant change over naive controls but were positive for CD28 (data not shown).

FIGURE 3.

PSA-T cells show surface markers of activation. To determine the surface phenotype of GlyAg-raised T cells, flow cytometric analysis of surface markers was performed on naive and PSA-raised CD4⁺ T cells. Bar graphs show the percent change of PSA-T cells as compared with control and are representative of n ≥ 3 experiments. Of greatest note, the general T cell activation markers CD69 and CD71 as well as the Treg marker CD45Rb^low significantly increased in PSA-stimulated cells compared with controls, whereas CD11a, PD1, and CD40L were essentially unchanged. Error bars are S.D. for all panels.

CD45Rb^high T cells are associated with an effector phenotype, whereas CD45Rb^low T cells are generally thought to be either regulatory or memory T cells. In comparing naive T cells to PSA-T cells, we found that the number of CD4⁺CD45Rb^low T cells increased sharply in PSA-T cells compared with naive controls (17.0% naive and 34.4% PSA-T).

PSA-T Cells Show Bystander Suppressive Activity

Because it has been demonstrated that PSA is capable of alleviating T cell-driven colitis (5) and demyelination (7) in mice, and we observed a significant increase in CD4⁺CD45Rb^low T cells, we investigated whether PSA-responsive T cells display suppressive activity. CD4⁺ PSA-T cells (8 days of activation and 10 days of rest) were added in 1:1 ratio to a standard proliferation assay of mitomycin-treated APCs, CD4⁺ T cells, and antigen or media control. Fig. 4A shows that the PSA-activated T cells (PSA-T) were able to suppress the proliferation of naïve T cells in response to PSA or conventional OVA peptide antigen. IFNγ production as a marker of cell activation was also suppressed when PSA-activated T cells were present (Fig. 4B). Bio-Plex detection was performed to discover how PSA-T cells influence production of other key cytokines. PSA-T cells strongly suppressed the secretion of many cytokines, including IL-1β, IL-17, and TNFα from naïve T cells in response to PSA or conventional OVA peptide antigen (Fig. 4C). Finally, the addition of exogenous IL-2 was able to break the suppressive activity of the PSA-T cells and restore IFN-γ production to normal levels (Fig. 4D).

FIGURE 4.

PSA-T cells are suppressive. Naive CD4⁺ T cells were stimulated in the presence and absence of PSA-T cells at a 1:1 ratio to determine if the PSA-T cells showed bystander suppressive activity. The PSA-T cells were able to inhibit both proliferation (A) and IFNγ production (B) nonspecifically for both PSA and OVA stimulations (raw values, left; fold change, right). C, Bio-Plex cytokine analysis extended the cytokine results to show remarkable suppression of all tested naive cell cytokines for both antigens, although addition of exogenous IL-2 (D) was able to reduce T cell suppression and increase IFNγ production for naïve T cells. Fold change was calculated using the respective media-only controls. Error bars are S.D. with at least n = 3 for all panels.

PSA-T Cells Are Foxp3-negative Tr1 Cells

Further analysis of surface markers by flow cytometry was performed to define the phenotype of these suppressive PSA-T cells. Consistent with IL-2 efficacy to partially relieve PSA-T anergy, a significant increase in CD25⁺CD4⁺ T cells was observed (Fig. 5A; 12.6% naive and 30.8% PSA-T). Because two types of regulatory T cells from the periphery have been identified, Foxp3⁻ Tr1 cells and Foxp3⁺ iTregs, PSA-T cells were compared with naive CD4⁺ T cells to look at expression of Foxp3, the transcription factor responsible for differentiation and maintenance of iTregs. We found that the CD25⁺ PSA-T cells lacked Foxp3 expression (Fig. 5A; 0.1% naive and 0.4% PSA-T). As a positive staining control for Foxp3 expression, Panc-1 cells, which constitutively express Foxp3 (30, 31), were shown to have high levels of Foxp3 production (data not shown).

FIGURE 5.

PSA-T cells are induced regulatory T cells of the Tr1 subset. A, given the observed suppressive activity, flow cytometry was use to determine the expression of both CD25 and Foxp3, markers of Treg cells. PSA-T cells were found to be CD25⁺ and Foxp3⁻. ELISA analysis further revealed that PSA-T cells do not make TGFβ over background (B) but secrete high concentrations of IL-10 (C). Together, these data place PSA-T cells in the Tr1 subset of inducible Treg cells. Error bars are S.D. with at least n = 3 for all panels.

Both TGF-β and IL-10 have been extensively implicated in suppressive functions of regulatory CD4⁺ T cells. ELISAs of culture supernatants demonstrate there was no increase in TGF-β production in naïve, PSA-T cells, or the PSA-T cells upon second stimulation above background (Fig. 5B). In contrast, PSA-T cells following the resting period but not PSA-stimulated naïve T cells produced high concentrations of IL-10 (Fig. 5C). Interestingly, a second exposure to PSA did not further increase IL-10 production (Fig. 5C). Given the lack of IL-10 in the first stimulation, it is clear that the IL-10 is produced by the added PSA-T population. Taken together with the cytokine profile and anergic state, these data demonstrate that the commensal GlyAg PSA induces CD4⁺ T cells that fall within the Tr1 (CD25⁺Foxp3⁻IL-10⁺TGFβ⁻IL2^lowIL4⁻) subset of inducible Treg cells in humans.

To establish the mechanism for the suppressive activity of PSA-T cells, the suppression assays were performed as before only with the addition of anti-IL-10 neutralizing antibody to block the IL-10 production from the PSA-T cells (Fig. 6). We found that the addition of anti-IL-10 antibody strongly inhibited the suppressive action of PSA-T cells and restored naïve T cell activation via IFNγ production. These findings demonstrate that the suppression of naïve T cell activation by the PSA-T cells is mediated by IL-10 secretion.

FIGURE 6.

Suppressive activity of PSA-T cells is mediated by IL-10. Suppression assays were performed with naive and PSA-T cells stimulated with PSA or OVA. To determine the mechanism of suppressive activity, anti-IL-10 monoclonal antibody was added at two concentrations. Anti-IL-10 mAb neutralized PSA-T cell suppression of naive cell activation and enabled robust IFNγ production (raw pg/ml, top; fold change, bottom). These data demonstrate the PSA-T exert regulatory function via secreted IL-10. Fold change was calculated using the respective media-only controls. Error bars are S.D. with at least n = 3 for all panels.

PSA Induces Tr1 Cells to Express Gut-homing Receptors

As a gut commensal, B. fragilis is primarily found in the colon and PSA-specific T cells have been implicated in commensal tolerance and the prevention of an artificially-induced model of colitis in the mouse (5). To determine whether PSA exposure could influence migration or cellular localization to the gut to facilitate this tolerogenic function, PSA-T cells were analyzed for receptors known to mediate homing to the gut and gut-associated lymphoid tissues. Strikingly, stimulation of T cells with PSA induces a 48.4% increase in the number of cells expressing CCR9 and a 17.9% increase in the number of cells expressing CD103, both specific markers of gut homing (Fig. 7, A and C). Remarkably, all of the cells in the sample that were expressing CCR9 were also Foxp3⁻ (Fig. 7A). Expression of gut-associated lymphoid tissue-homing markers, such as CCR7, CD29, and CXCR4 were also up-regulated, but in more modest amounts (Fig. 7, B–F). As an internal control, expression of CD62L, which is involved with homing to high endothelial venules of peripheral lymphoid tissues, was not altered (data not shown), demonstrating that the commensal GlyAg PSA can initiate a Tr1 cell phenotype with specific gut-homing markers from a single exposure in peripheral human CD4⁺ T cells.

FIGURE 7.

PSA initiates Tr1 cells to express gut-homing receptors from the periphery. PSA is expressed by colonic commensal bacterium B. fragilis, which is thought to play an important role in gut immune homeostasis. To determine whether human PSA-T cells expressed the appropriate markers for gut localization to block inappropriate intestinal inflammation against the normal flora, flow cytometry was performed (A–F). Compared with unstimulated T cells, PSA-T cells showed robust CCR9, CCR7, and CD103 expression. Smaller changes were seen with CD29 and CXCR4. Unstimulated and PSA-T cells were equally positive for CD62L. These findings demonstrate that PSA-T cells express homing markers for gut localization, which is the site of chronic exposure to PSA under normal conditions. Error bars are S.D. with n = 3 for all panels.

DISCUSSION

In this study, we demonstrate that a single exposure of the B. fragilis GlyAg PSA induced human peripheral CD4⁺ T cell differentiation into cells that constitutively express IFNγ, IL-10, and CD25, but not TGF-β or Foxp3, classifying them as Tr1 cells. The high IL-10:IL-4 ratio further defines these cells as Tr1 cells (13). These PSA-specific cells were profoundly anergic to secondary stimulation as measured by both cytokine production and proliferation, although addition of recombinant IL-2 partially relieves the blockade. In concordance with mouse data showing that PSA immunization ameliorates experimental inflammatory bowel disease and experimental autoimmune encephalomyelitis demyelination in a T cell-dependent manner (6, 7), human PSA-T cells exhibited nonspecific bystander suppression mediated by IL-10 secretion, providing a mechanism by which PSA immunizations could protect against inflammatory conditions (5, 23, 25). However, differing dramatically with previous reports in mice, PSA exposure to human cells generates Foxp3⁻ Tr1 suppressive T cells, instead of the Foxp3⁺ iTregs reported in mice (6, 7). Perhaps most importantly, PSA induced the expression of gut homing receptors, most notably CCR9, on responding T cells, thereby providing an explanation for how commensal antigen-specific T cells achieve and maintain proper localization within the gut to assist in immune homeostasis with the normal microbiota.

Tregs are well established key players in mediating antigen tolerance and preventing autoimmune disease by suppressing immune activation and maintaining tolerance to self and commensal-derived antigens. The three classes of CD4⁺ Tregs are distinct cell types thought to arise from different lineages and may have divergent functions in vivo. iTregs and Tr1 cells are inducible from naïve CD4⁺ cells both in vitro and in vivo and are thus part of the adaptive immune response to an activating antigen. Conversely, CD4⁺CD25⁺ nTregs are not inducible because their antigen specificity is predefined from their selection in the thymus (32).

It has been demonstrated by several groups that the induction of CD4⁺CD25⁺ nTregs and Tr1 cells is independent. Tr1 cells differentiate from peripheral naive CD4⁺ precursors in an IL-10-rich environment (10) following multiple rounds of activation, whereas the same cells cultured with TCR stimulation and TGF-β gives rise to Foxp3⁺ iTregs (33). Both cell types acquire suppressive activity and secrete IL-10, which is important for Tr1 cell differentiation, their suppressive function, and maintenance of anergy (10, 34, 35). One prominent difference between iTregs and Tr1 cells is their expression of the transcription factor Foxp3. Foxp3 is thought to be the major regulator of iTreg cell development but is not expressed in significant quantity or frequency by Tr1 cells, providing strong evidence that these cell population are independent lineages. Because TGF-β induces Foxp3 expression in TCR-activated naïve CD4⁺ cells (33), it is not surprising that Foxp3⁺ iTregs require TGF-β for induction but Foxp3⁻ Tr1 cells do not (36). Tr1 cells also have a cytokine profile distinct from the classical helper paradigms because they produce low amounts of IL-2 and no IL-4 (37). Tr1 cells are known to sometimes (but not always) produce TGF-β, IL-5, and IFNγ, depending on the conditions used to produce the cells (8). However, Tr1 cells are always strong producers of IL-10 and are highly anergic (37). Taken together, the anergic, Foxp3⁻, CD25⁺, and IL-10 secreting PSA-specific human T cells described here are consistent with the Tr1 subset.

Although Foxp3 appears to be a specific marker for nTregs, reports indicate that it is not a universal marker for all suppressive T cells. Tr1 cells consistently demonstrate suppressive activity with a lack of Foxp3 expression, though transient Foxp3 expression in Tr1 cells has also been reported (38). Recent findings also show that Foxp3 is differentially regulated in mice and humans (39), and as a result, the observation that the human CD4⁺ cells induced by PSA are Foxp3⁻ is highly significant. Although PSA induced the induction of FoxP3⁺IL-10-producing cells from naive mouse CD4⁺ cells in the experimental autoimmune encephalomyelitis model of multiple sclerosis (7), a thorough phenotypic analysis of the T cells was not performed. A subsequent study was reported using PSA to stimulate production of mouse IL-10⁺Foxp3⁺ iTregs, which were able to reverse the signs of colitis in mouse models of inflammatory bowel disease (6), but as before, a detailed phenotype of these cells was not determined. In contrast, we have found that human PSA-specific T cells do not express Foxp3 yet are highly suppressive. This observation is supported by a number of human studies on Treg cells. For example, Tr1 cells isolated from tolerized human transplant patients do not express Foxp3 but produce high levels of IL-10 (13). In addition, IPEX patients with loss of function mutations in the Foxp3 gene were shown to have suppressive T cell activity through their Tr1 cells, demonstrating that Foxp3 is not required for suppressive Treg cell function in humans (14). The present results are consistent with these human studies, which collectively suggest that the mouse studies showing PSA-specific Foxp3⁺ iTregs may have limited applicability to human immunology.

It is sometimes possible to induce antigen-independent proliferation of Tr1 cells through the addition of exogenous IL-2 (40) because these cells express the IL-2 receptor-α chain CD25. As a result, the ability to induce proliferation of the PSA-T cells with IL-2 further supports their characterization as Tr1. In contrast to TGF-β-dependent differentiation and Foxp3 expression in the iTreg population (33), IL-10 is thought to be the critical cytokine for Tr1 cell differentiation, suppressive function, and even maintenance of their anergic state. The present data strongly supports these observations in that PSA-induced suppressive T cells produced high concentrations of IL-10, essentially no TGFβ, and no detectable Foxp3. Moreover, antigen-specific Tr1 cells have been produced previously in vitro by culturing naïve T cells with APCs, antigen, and exogenous IL-10 (10) after multiple rounds of activation. Based on these previous findings, one of the most striking observations in this report is that an antigen present on the surface of commensal organisms already known to mediate T cell tolerance in the gut, induced this Tr1 subset of regulatory T cells after only a single exposure and without exogenous IL-10.

A hallmark characteristic of Tr1 cells is nonspecific bystander suppression of other nearby T cells though release of IL-10. This activity and to some extent their anergic state can be reversed with neutralizing anti-IL-10 monoclonal antibody (10, 40), which is consistent with our observations that anti-IL-10 antibody ameliorated the bystander suppression effect of PSA-T cells 25-fold. The ability to suppress naive T cell activation by another antigen (i.e. OVA) demonstrates a lack of suppression specificity and provides a mechanism by which PSA is able to stimulate T cells that ameliorate inflammatory bowel disease in three different animal models (5). In fact, Tr1 cells have even been used in preliminary studies for use as therapy against inflammatory bowel disease (41).

Mucosal sites such as the gastrointestinal tract are constantly barraged with an abundance of both potential pathogens that must be surveyed and eliminated and commensal bacteria, which must be maintained via tolerance. In recent years, it has become clear that the maintenance of gut homeostasis is strongly influenced by lymphocyte localization and the chemokines that regulate migration. For example, the chemokine TECK/CCL25 is highly expressed in the small intestine and the colon (17) and has been localized by in situ hybridization to the intestinal epithelium. CCR9 is the only known receptor for TECK (42) and is thought to be the central regulatory chemoattractant receptor for the intestine, as it is ubiquitously expressed by essentially all resident CD4⁺ and CD8⁺ cells (17, 43). CCR9 is also expressed on human T cells found in the intestine (44), and mucosal T cells expressing CCR9 tend to fall within the Tr1 subset (15). Likewise, CD103 binds integrin β7 to form the integrin molecule αEβ7, whose main ligand is E-cadherin, an adhesion molecule found on epithelial cells (45) that also plays an important role for homing to sites within the intestine (46). Some CD4⁺/Foxp3⁻ cells with suppressive activity have been shown previously to express CD103 (47), although it was unclear whether they represented a specialized feature for Tregs. More importantly, CD4⁺CD25⁺ T cells were also not able to prevent colitis in immune-deficient recipients lacking CD103, demonstrating a functional role for CD103 on host cells in Treg-mediated regulation of intestinal inflammation (48). These observations support an emerging paradigm in which epithelial cell-expressed chemokines provide specific “addressin” signals that control cellular recruitment and thus the character of immune responses at different epithelial surfaces (16). Consistent with this model, we have found that a natural commensal GlyAg known to induce protective responses in inflammatory bowel disease models strongly induces expression of CCR9 and CD103 on responding human Tr1 cells and these T cells show powerful tolerogenic potential via expression of IL-10.

The maintenance of tolerance to commensal bacteria in the gut is essential, and it is widely thought that defects in this tolerance are an underlying cause of colitis. The role for commensal antigens has therefore become a focal point for understanding this complex balance. It is interesting to note that gnotobiotic pigs selectively colonized with commensal bacteria increased cellular expression of TECK/CCR9 but not Foxp3 mRNA compared with germ-free controls (49). In contrast to observations in mice (6, 7), this study reveals that a carbohydrate expressed by a commensal organism induces regulatory T cells of the CD25⁺Foxp3⁻ Tr1 subset in humans. Perhaps most striking is the observation that a single exposure to this antigen provides the required signals to induce responding peripheral T cells to begin expressing homing receptors that facilitate migration to the site of chronic antigen exposure, the gut. These findings provide a direct link between the reported Tr1 cell population found in the human gut, their association with immunologic homeostasis, and commensal bacterial antigens that coordinate these events to the benefit of the human host. Furthermore, these data show that microbial products from the intestinal microbiota can induce suppressive CCR9⁺CD103⁺ gut-homing regulatory T cells that lack Foxp3 and secrete high concentrations of IL-10 in the periphery, leading to a model in which a single commensal antigen can program the tolerogenic phenotype of peripheral human T cells and direct their proper localization in the body.