Roles for TGF-β and Programmed Cell Death 1 Ligand 1 in Regulatory T Cell Expansion and Diabetes Suppression by Zymosan in Nonobese Diabetic Mice

The development of type 1 diabetes in the NOD mouse is a spontaneous autoimmune process influenced by genetic and environmental factors, with significant roles for CD4 and CD8 T cells (1–4), B cells (5, 6), and innate APCs (7–9). APCs have been suggested to be involved during both initiation and effector stages of disease. For instance, roles for both B cells (5, 6) and CD11c⁺CD11b⁺ dendritic cells (DCs) (8) have been reported for priming autoreactive T cells in diabetes. Consistent with this, surgical removal of the pancreatic draining lymph nodes effectively prevents the onset of diabetes when performed on young NOD mice, but not in older mice (>10 wk of age) that already show substantial cellular infiltrates into the islets of Langerhans cells (10, 11). T cells that infiltrate the pancreas recruit CD11b⁺ macrophages (Mϕs) (12), which are required within the pancreatic lesion alongside T cells to effect the destruction of the insulin-producing β cells (7).

Altering the environment in which autoantigens are presented strongly influences the balance of T cell responses and the progression of autoimmunity in NOD mice. Th1 cells are widely thought to mediate type 1 diabetes in NOD mice, and, accordingly, increases in cytokines such as IL-10, IL-2, or IL-4 can favor other T cell populations, including Th2 or Foxp3⁺ regulatory T cells (Tregs) and inhibit autoimmune diabetes (13–16). TGF-β potently reduces the maturation status of APCs and inhibits Th1 responses (17, 18) and, accordingly, suppresses the onset of diabetes when expressed as a transgene in the islets or in DCs (19–22). Foxp3⁺ Tregs have been invoked to explain the inhibitory effects of TGF-β on diabetes progression (18, 22, 23), and it has been shown in multiple systems that effector cells need to be able to respond to Treg-produced TGF-β for the Tregs to control diabetes development (17, 24, 25). Perhaps uniquely among regulatory cytokines, TGF-β may even assist in establishing tolerance in overtly diabetic mice, as has been suggested in the context of islet transplantation (23) and anti-CD3 therapy (26, 27).

Specialized tolerogenic DCs that express low levels of costimulatory molecules, or negative costimulatory molecules such as programmed cell death 1 ligand 1 (PD-L1), may preferentially induce Foxp3⁺ Tregs (28–32). In the face of potent inhibitory signals, T cells may even become unresponsive, as has been shown to occur when PD-L1 interacts with its receptor, programmed cell death 1 (PD-1), on activated T cells (33, 34). In NOD mice, PD-L1 provides a critical check on effector T cells in the pancreas of NOD mice, and its absence results in aggressive infiltration of the pancreas and rapid onset of diabetes (35–38).

Exposure of NOD mice to active infections or antigenic preparations of bacteria, viruses, or helminth parasites can alter both APC and T cell phenotypes, suppressing the progression of autoimmunity (39), but until recently the effects of fungal Ags on diabetes onset were unknown (40). Interactions between fungal Ags and the cells of the innate immune system are capable of strongly polarizing T cell responses, giving rise to either proinflammatory subsets such as Th1 and Th17 cells or suppressive populations such as Tregs or IL-10–producing T cells (40–44). At the crux of this dichotomy lies the ability of fungal Ags to trigger a complex mixture of cytokine secretion from APCs, including IL-1, IL-6, TNF, and members of the IL-12 family alongside the suppressive cytokines IL-10 and TGF-β (40, 42, 43). The induction of TGF-β is particularly important for driving Th17 and Foxp3⁺ Treg responses, which have been implicated in the eradication (45–47) and persistence (48, 49), respectively, of fungal pathogens. Fungal Ags can also act as adjuvants, and zymosan, a complex fungal cell wall component, drives Th17 responses that promote autoimmunity in mice (44, 50). Recently, however, zymosan has been shown to initiate events favoring the generation of Foxp3⁺ Tregs and suppressing the development of experimental autoimmune encephalomyelitis (EAE) (41, 42), and it has also been shown to prevent diabetes onset in NOD mice (40). In this paper, we investigate the ability of zymosan to drive innate immune responses favorable to Foxp3⁺ T cell development in NOD mice, and we assess the contributions of specific molecules in the regulation of diabetes onset by zymosan and Tregs.

Female NOD/Tac mice were housed and barrier bred under specific pathogen-free conditions in the Pathology Department, University of Cambridge animal facilities (Cambridge, U.K.). Mice were used between 10 and 12 wk of age. All animal work was conducted under United Kingdom Home Office project license regulations after approval by the Ethical Review Committee of the University of Cambridge. Mice were routinely monitored for glucosuria using Diastix (Bayer Diagnostics, Basingstoke, U.K.).

All cells were cultured in IMDM supplemented with 2 mM l-glutamine, 50 μM 2-ME, 100 U/ml penicillin, 100 μg/ml streptomycin, and 10% heat-inactivated FCS. Anti-CD3 clone 145.2C11 (2 μg/ml) was grown and affinity purified in house. Cells were stimulated in the presence of zymosan from Saccharomyces cerevisiae (50 μg/ml; Sigma-Aldrich, St. Louis, MO), LPS from Samonella enteritidis (1 μg/ml; Sigma), FSL-1 (1 μg/ml; InvivoGen, San Diego, CA), or Pam₃CysK₄ (1 μg/ml; InvivoGen). Fluorescently conjugated Alexa 488–zymosan A bioparticles were purchased from Invitrogen (Paisley, U.K.). The retinoic acid receptor antagonist LE540 (1 μM; Wako Pure Chemicals, Osaka, Japan) and the ERK inhibitor U0126 (10 μM, Invivogen) were dissolved in DMSO, whereas the p38 MAPK inhibitor SB202190.HCl (5 μM; Enzo Life Sciences, Farmingdale, NY) was prepared in aqueous solution.

The MLE/PAI-L cell line is derived from mink lung epithelial cells and contains firefly luciferase under PAI-1 promoter control (51). MLE cells were cultured with 50 μl tissue culture supernatants or a double-dilution standard curve derived from recombinant human TGF-β1 (R&D Systems, Minneapolis, MN) for 18 h. Cells were then lysed and luciferase activity was measured using a luciferase assay kit (Biotium, Hayward, CA) and a BetaLux scintillation counter (Wallac, PerkinElmer, Wellesley, MA).

Anti–TGF-β (1D11.16), anti–IL-6 (6B4), anti–IL-6R (15A7), anti–IL-10R (1B1.3a), and anti–PD-L1 (MIH5) were grown in-house and were precipitated from hybridoma supernatants and dialyzed against PBS before use. For in vivo cytokine neutralization experiments, Abs were injected i.p. 2 h prior to zymosan injection, then again at 48 h and on day 5. Other Abs were purchased, including anti-TLR2 (T2.5; eBioscience, San Diego, CA), anti-SIGNR1 (ER-TR9; Hycult Biotechnology, Uden, The Netherlands), and anti–dectin-1 (2A11; AbD Serotec, Oxford, U.K.), and were used at 20 μg/ml in vitro.

Whole pancreas samples were harvested and processed from individual mice. Briefly, pancreases were torn into smaller pieces in cold PBS containing 5% FCS, 56 mM glucose, and Complete Mini protease inhibitors (Roche Diagnostics, Basel, Switzerland). The tissues were washed twice in cold PBS before incubation in 2 ml prewarmed PBS containing 15% FCS, 10 μg/ml DNase I (Sigma-Aldrich), and 330 μg/ml Liberase CI (Roche Diagnostics). After digestion, tissues were washed and cell suspensions were prepared by forcing through a 70-μm cell strainer. Cells were then centrifuged through a 33% Percoll (GE Healthcare, Chalfont St. Giles, U.K.) gradient to obtain leukocyte populations.

Peritoneal Mϕs were flushed from the peritoneal cavity by lavage with cold PBS and isolated to >99% CD11b⁺ Mϕs by adherence to plastic plates over 4 h at 37°C in IMDM with 10% FCS. Pancreatic Mϕs from zymosan-treated mice were separated from total pancreatic mononuclear cells similarly, reaching >95% CD11b⁺CD11c⁻ cells.

Cells were washed and resuspended in staining buffer (PBS, 2.5% FCS, and 0.05% NaN₃). Nonspecific binding was blocked using 2.4G2 anti-FcγR supernatant (prepared in-house). Cells were stained with appropriate combinations of the following Abs: CD3ε (2C11), CD4 (RM4-5), CD11b (M1/70), CD11c (HL3), Gr-1 (RB6-8C5), Foxp3 (FJK-16s), PD-L1 (MIH5), PD-1 (J43), IL-4 (11B11), IL-17 (TC11-18H10), and IFN-γ (XMG1.2). Biotinylated polyclonal goat anti-human latency-associated peptide (LAP) and biotinylated normal goat serum (R&D Systems) were used in conjunction with streptavidin-PE (BD Biosciences, San Diego, CA). For live cell discrimination, 7-aminoactinomycin D (BD Biosciences) was used.

Intracellular Foxp3 staining was performed according to the manufacturer’s instructions (Foxp3 anti-mouse/rat Foxp3 staining set; eBioscience). For intracellular cytokine staining, cells were stimulated for 4 h with 500 ng/ml phorbol 12,13-dibutyrate and ionomycin, and cytokine secretion was blocked using 1 μg/ml brefeldin A (all from Sigma-Aldrich). Cells were fixed promptly in 1% paraformaldehyde-PBS and permeabilized with 0.5% saponin (Sigma-Aldrich) in staining buffer.

For CFSE dye dilution experiments, splenocytes (2 × 10⁷) were labeled in 1 ml PBS containing 5 μM CFSE (Invitrogen) by incubating for 15–20 min at 37°C. The reaction was stopped and excess CFSE removed by washing with 10% FCS IMDM.

For detection of phosphorylated epitopes, Mϕs were stimulated for 45 min with Ags and then immediately fixed with 4% paraformaldehyde-PBS for 30 min at room temperature. Cells were permeabilized with 70% ice-cold methanol for at least 30 min. Methanol was later removed by two wash steps, and cells were exposed to Abs in staining buffer for 30 min using anti–phospho-p38 MAPK Thr¹⁸⁰/Tyr¹⁸² (28B10)-Alexa 488 and anti–phospho-p42/p44 Erk1/2 Thr²⁰²/Thr²⁰⁴ (E10)-Alexa 647 (Cell Signaling Technologies, Danvers, MA). General staining procedures were adapted from work by Krutzik et al. (52).

Cells were acquired using a BD FACSCalibur, a BD FACScan (BD Biosciences), and a CyAn ADP (Beckman Coulter, Fullerton, CA) and were analyzed using FlowJo (Tree Star, Ashland, OR) software.

Total RNA was isolated using an RNeasy kit, converted to cDNA using a reverse transcription kit, and quantified in real-time using Fast SYBR Green PCR (all from Qiagen, Valencia, CA). RNA was normalized prior to conversion into cDNA, and equal amounts of cDNA were analyzed in real-time. cDNA was analyzed in duplicate reactions with amplification of target gene and housekeeping gene (hypoxanthine phophoribosyl transferase 1 [Hprt1]) transcripts performed on the same reaction plate on a 7500 Fast real-time PCR system (Applied Biosystems). Proprietary QuantiTect primer assays for all genes were purchased from Qiagen. Relative gene expression was normalized to Hprt1 according to the change in threshold cycle (C_T) method, C_T^Gene1 − C_T^Hprt1.

Statistical analyses were routinely performed using GraphPad Prism 4 software (GraphPad Software, San Diego, CA). Diabetes progression was analyzed by plotting Kaplan-Meier survival curves and performing log-rank tests between the cohorts. The Mann-Whitney U test was used to assess the significance of nonparametric data, while unpaired t tests were used for parametric data sets. For comparisons between multiple samples, one-way ANOVA (for independent data sets) or repeated measures ANOVA (for paired data) was used. Tests performed and calculated two-tailed p values are indicated in the individual figure legends.

Female NOD mice in our colony develop an incidence of diabetes that exceeds 70% penetrance by 20 wk of age, while substantial pancreatic infiltration and insulitis are already detectable by 10 wk of age. When 10- to 11-wk-old female prediabetic NOD mice received a single i.p. injection of zymosan, diabetes onset was significantly delayed, whereas earlier administration had little effect on the kinetics of diabetes onset (Fig. 1 and data not shown). Previous studies indicated a need for multiple immunizations to achieve tolerance (40); however, we found that a single treatment was sufficient to considerably delay diabetes progression.

The anti-inflammatory cytokines IL-10 and TGF-β have been shown to be produced in response to zymosan by both APCs and T cells (40), and IL-10 was found to be involved in the suppression of EAE by zymosan (41). Overexpression of TGF-β can ameliorate or prevent diabetes in the NOD mouse (21), as can expansion or activation of Tregs in vivo (53), and therefore we examined the contributions of TGF-β and IL-10 to the inhibition of diabetes onset by zymosan. We found that Ab neutralization of TGF-β abrogated the ability of zymosan to delay diabetes (Fig. 2A). IL-10 has also been suggested to assist in Treg induction by zymosan (41, 42), but IL-10R blockade did not affect the onset of diabetes in zymosan-treated NOD mice (Fig. 2B). In addition to driving Tregs via TGF-β (41), zymosan has been shown to initiate autoimmune pathologies by inducing TGF-β– and IL-6–dependent Th17 cells (44, 50). IL-6 had no role in zymosan-mediated inhibition of diabetes onset (data not shown), which suggested a lack of involvement of a Th17 response.

PD-L1–PD-1 interactions limit the progression of diabetes in NOD mice (35), and therefore we examined whether PD-L1 was involved in the delay of diabetes in zymosan-treated mice. We observed that irrespective of previous zymosan treatment, neutralization of PD-L1 triggered rapid onset of diabetes in NOD mice (Fig. 2C).

As substantial islet infiltrates were already present in pancreas of zymosan-injected mice, its mode of action was unlikely to be prevention of infiltration, but rather it suggested alterations in immune regulation within the pancreas involving rapidly responding innate immune cells. We therefore examined the composition and phenotypes of the cells recovered from the pancreas of zymosan-treated mice. The most remarkable and consistent changes we observed were to CD11b⁺CD11c⁻ cells, which increased dramatically within hours of zymosan treatment and remained elevated for weeks (Fig. 3A). Consistent with our findings of PD-L1–dependent diabetes inhibition, zymosan treatment markedly upregulated PD-L1 expression on CD11b⁺ APCs recovered from the pancreas during the same period (Fig. 3B, 3C). Interestingly, zymosan treatment also induced surface expression of the LAP of TGF-β, and using fluorescently conjugated zymosan, we were able to determine that LAP expression was largely dependent on internalization of zymosan (Fig. 3D, 3E). Upon in vitro restimulation, mononuclear cells from the pancreas of zymosan-treated, but not untreated, NOD mice secreted substantial quantities of biologically active TGF-β (Fig. 3F).

Our findings of biologically active TGF-β secretion by pancreatic mononuclear cells from zymosan-treated mice are intriguing in view of the potent ability of TGF-β to inhibit diabetes development in NOD mice (19–21), as well as in view of our findings of a requirement for TGF-β in the delay of diabetes by zymosan (Fig. 2A). Relatively little is known about the pathways driving biologically active TGF-β production in response to fungal Ags. Transcription and translation of TGF-β are under complex regulation, and although latent TGF-β is expressed at high levels, little active TGF-β is available to act on cellular receptors (17). Posttranslational processing of TGF-β involves proteolytic cleavage and extraction of active TGF-β from the latent complex with LAP (17), which we found expressed on CD11b⁺ Mϕs in the pancreas of zymosan-treated mice (Fig. 3D). Consistent with our ex vivo findings of bioactive TGF-β production by pancreatic mononuclear cells from zymosan-treated mice (Fig. 3E), Mϕs stimulated in vitro with zymosan secreted robust amounts of biologically active TGF-β, which was not seen in response to the TLR2/1 ligand Pam₃CysK₄, the TLR4 agonist LPS, or the TLR2/6 agonist FSL-1 (Fig. 4A).

Many of the immune mediators produced by APCs in response to zymosan are induced following ligation of both TLR2 and dectin-1 (54, 55). SIGNR1 has also been implicated in the production of TNF (56), and, recently, dectin-2 has been shown to play a partially redundant role with dectin-1 in driving Syk-dependent cytokine secretion (46). Both the TLR2/MyD88 and dectin/Syk pathways trigger downstream activation of p38 and ERK MAPK, among other signaling molecules, leading to transcription of cytokine genes (43, 57). Using flow cytometric analysis of NOD peritoneal Mϕs (pMϕs), we confirmed that zymosan stimulation resulted in strong phosphorylation of both p38 and ERK MAPK, as assessed by flow cytometry (Fig. 4B–D). Phosphorylation of p38 MAPK was greater than seen with the TLR4 agonist LPS (Fig. 4C). TLR2 ligation has been shown to activate both p38 and ERK MAPK (57), and inhibition of either kinase reduced the production of TGF-β (Fig. 4E).

A recent report detailing the ability of zymosan to suppress autoimmunity found that this effect was due to interactions with TLR2, whereas proinflammatory cytokines could be induced through dectin-1 (41). Therefore, we investigated the requirements for PRR ligation in Mϕ secretion of bioactive TGF-β. Ab blockade of TLR2 significantly reduced the production of TGF-β (Fig. 4F), whereas blockade of SIGNR1 or dectin-1 modestly increased the secretion of TGF-β (Fig. 4E).

Zymosan has been shown to promote Th17 responses (44, 50), which are dependent on TGF-β and IL-6 (58), as well as driving Foxp3⁺ Tregs through a combination of TGF-β, IL-10, and retinoic acid (41). Because activation of Tregs or immune deviation to Th17 might forestall a Th1-driven pathology, we examined the expression of relevant genes in the cellular infiltrates recovered from the pancreas of zymosan-treated NOD mice. Along with TGF-β, expression of IL-6, IL-10, and IL-17 were increased at 18 h and 2 wk after injection (Fig. 5). Ab neutralization of IL-6 and IL-10 (Fig. 2B and data not shown) indicated little impact of these cytokines or a Th17 response on the progression of diabetes. The induction of TGF-β at both time points supports the critical role for this cytokine in suppression of diabetes by zymosan (Fig. 2A), and it is consistent with our findings of LAP expression by CD11b⁺ Mϕs at 18 h after treatment (Fig. 3E) as well as the T cell secretion of TGF-β found by Karumuthil-Melethil et al. (40) in the adaptive phase.

Conversion of retinoic acid from retinaldehyde is controlled by multiple enzymes, including retinaldehyde dehydrogenases 1 and 2 (Aldh1a1 and Aldh1a2), which have been recently found to be inducible by TLR ligands (41, 59) as well as being constitutively expressed in APCs from GALTs (28, 60). Although zymosan has been shown to upregulate expression of both Aldh1a1 and Aldh1a2 in splenic DCs from C57BL/6 mice (41), we did not find increased mRNA for either enzyme in the pancreatic infiltrating cells of zymosan-treated NOD mice, and in fact Aldh1a1 levels diminished (Fig. 5), suggesting that retinoic acid induction may not be involved in the establishment of a benign pancreatic infiltrate by zymosan.

After in vivo exposure to zymosan, DCs have been shown to expand Foxp3⁺ T cells through a combination of TGF-β, IL-10, and retinoic acid (41). PD-L1 expression on specialized DC subsets has also been suggested to enhance the conversion of naive T cells into Foxp3⁺ Tregs (31, 32) and possibly maintain higher levels of Foxp3 expression in Tregs (31).

To examine the ability of infiltrating APCs to expand Tregs, we cocultured CFSE-labeled NOD splenocytes with the infiltrating monocytes in the presence of anti-CD3 and initially monitored the ability of these cells to influence division of Foxp3-expressing cells. We found that these infiltrating cells significantly enhanced the ratio of Foxp3⁺-to-Foxp3⁻ T cells in vitro (Fig. 6A). This increase in Foxp3⁺ Tregs was reduced by neutralization of TGF-β (Fig. 6B), but not by anti–PD-L1 (Fig. 6C). In contrast to the data obtained by others using zymosan-exposed DCs, expansion of Foxp3⁺ T cells by Mϕs from the pancreas of zymosan-treated NOD mice was not affected by inhibition of the retinoic acid receptor using LE540 (Fig. 6D). Although expression of PD-L1 and production of retinoic acid have been shown to contribute to DC-mediated induction of Foxp3⁺ Tregs (28, 29, 32), less is known about the function of these molecules in Mϕs (60). To determine whether the CD11b⁺TGF-β⁺PD-L1⁺ APCs were the cells responsible for expanding Tregs, we fractionated the cells recovered from the pancreas of zymosan-treated mice by adherence, isolating a 95% pure population of adherent CD11b⁺CD11c⁻ cells. These adherent cells displayed a significantly greater ability to expand Foxp3⁺ T cells (Fig. 6E).

A previous report on zymosan treatment of NOD mice showed activation of Tregs without altered proportional representation of Foxp3⁺ Tregs in the spleen or pancreatic draining lymph nodes, but it did not examine the cells within the pancreatic lesion (40). We found that a single injection of zymosan dramatically increased both proportional representation and the absolute number of Foxp3⁺CD4⁺ T cells in the pancreatic infiltrate as compared with untreated mice (Fig. 7). Despite increases in mRNA for IL-17 and IL-6 (Fig. 5), we did not observe significant changes in the frequencies of Th17, Th1, or Th2 cells in the pancreas following zymosan treatment (Supplemental Fig. 1 and data not shown).

Previously it was shown that NOD mice could be protected from diabetes past 30 wk of age with a long course of repeated injections of zymosan (40). In this study, we have found that a single zymosan treatment evokes a significant delay, although diabetes does eventually occur (Fig. 1 and data not shown). We have used this treatment to understand how zymosan modifies the immune system and slows the autoimmune attack. Studies from the EAE model of multiple sclerosis suggested that effects of zymosan on DCs, particularly splenic DCs, suppressed the onset of autoimmunity via TLR2-dependent activation of ERK, leading to production of TGF-β, retinoic acid, and IL-10 (41, 42). We found important effects on pancreatic Mϕ populations in NOD mice, whereas the effects on CD11c⁺ DCs were far less marked (data not shown). Considerable increases in CD11b⁺ APCs expressing LAP and PD-L1 correlated with enhanced TGF-β mRNA and protein levels in the mononuclear cell infiltrates of zymosan-treated mice and suggested that these cells contributed to the TGF-β–dependent delay of diabetes onset. We therefore explored the mechanisms by which zymosan alters Mϕ phenotype and the onset of autoimmunity.

Zymosan is a complex Ag capable of interacting with multiple pattern recognition receptors, and we importantly identify TLR2 as a major receptor for driving bioactive TGF-β production by Mϕs. Zymosan induced Mϕ production of bioactive TGF-β through TLR2-involved activation of p38 and ERK MAPK. Strong activation of ERK in APCs has been suggested to induce cytokines that inhibit Th1 responses, and in some cases Th17 responses (41), while activation of p38 MAPK is known to be essential to TLR-driven activation of IL-12p70 (61). ERK-dependent cytokine induction by zymosan has been widely investigated, whereas p38 MAPK has been assumed to contribute to a lesser degree (41, 62). We found that bioactive TGF-β secretion by pMϕs was equally dependent on ERK and p38 MAPK, and it required TLR2 without assistance from dectin-1. However, Ab blockade of TLR2 did not completely abrogate TGF-β production, which may reflect incomplete receptor blockade or contributions of other receptors.

Different APC subsets regulate progression of diabetes in NOD mice (8), and treatments that enhance the frequency of immature or tolerogenic APC subsets, particularly those that promote Foxp3⁺ Tregs, can suppress diabetes onset (63–65). In this study we found that TGF-β and PD-L1–expresssing APCs appeared in the pancreas in response to zymosan, and that delay of diabetes required TGF-β and PD-L1. We demonstrated that Mϕs isolated from the pancreas of zymosan-injected mice were able to expand Foxp3-expressing cells in vitro and that TGF-β was required for this expansion. As TGF-β is also known to inhibit Th1 responses and T cell proliferation (17), and to mediate aspects of Treg suppression (24), its effects may not be limited to Treg expansion in vivo.

In addition to shaping potentially tolerogenic T cell and APC phenotypes, TGF-β promotes the repair of damaged tissues through restructuring of the extracellular matrix (66), and it is conceivable that TGF-β induced by zymosan treatment may also act directly or indirectly to enhance the survival or function of residual β cells. During development, TGF-β promotes the development and differentiation of the endocrine pancreas over exocrine tissue (21, 67–69). In the adult pancreas, however, TGF-β signals have been shown to repress insulin transcription in β cells via Smad3 (70), suggesting that the direct effects of TGF-β on the islets would disfavor glucose regulation. In contrast, the immunological roles of TGF-β in suppressing type 1 diabetes are well established (18–27).

Interactions of PD-L1 with its receptor PD-1 are required to maintain peripheral tolerance at all stages in the development of diabetes in NOD mice (35). It has been suggested that the critical site of PD-L1 expression for suppression of diabetes development is in fact on the islets of Langerhans cells, and that the contributions of PD-L1 on hematopoietic cells are negligible (37). Epithelial cells not of hematopoietic origin can also function as APCs in tissues, however (71), and not all of the CD11b⁺PD-L1^high APCs present in the pancreas following zymosan treatment express CD45, the common leukocyte Ag (data not shown). Furthermore, in zymosan-treated NOD mice, PD-L1 expression on CD11b⁺ APCs is induced to far higher than unstimulated levels, and T cell expression of PD-1 also increases (Supplemental Fig. 2). Since a relatively benign pancreatic infiltrate exists in zymosan-treated NOD mice, it is possible that PD-L1 is actively required to reinforce diabetes prevention. We did not find any role for PD-L1 in the induction of Foxp3 expression by zymosan, but it has been shown that PD-L1 can potentiate and maintain Foxp3⁺ Treg suppressive function (31) as well as inducing Foxp3 expression (31, 32). We cannot rule out a link between PD-L1 and Foxp3⁺ Tregs in NOD mice in vivo following zymosan treatment.

In this study we provide insight into the innate and adaptive immune modulation that occurs in NOD mice treated with a fungal Ag, and we demonstrate the importance of effects on pancreatic Mϕs. We provide further evidence that innate activation through TLR2 can have tolerogenic effects, and that induction of TGF-β and PD-L1 significantly delays the onset of overt diabetes in mice with advanced insulitis.

We thank Dr. Nick Holmes and Dr. Stephen Newland for helpful discussions and Yvonne Sawyer for technical assistance with the project.

Disclosures The authors have no financial conflicts of interest.