Dominant Role of Antigen Dose in CD4⁺Foxp3⁺ Regulatory T Cell Induction and Expansion¹

Type 1 diabetes is an autoimmune disease characterized by the destruction of insulin-producing β cells of the islets of Langerhans (1). The nonobese diabetic (NOD)³ mouse provides a useful model for human type 1 diabetes, since it shares many of the genetic and immunological features of the human disease (2). Recent studies in NOD mice have revealed an age-related progressive decrease in the number and/or function of CD4⁺Foxp3⁺ regulatory T cells (Treg) that is associated with the onset of diabetes (3, 4). Treg play an important role in the maintenance of self-tolerance and therapeutic strategies aimed at increasing the number and/or function of Treg have proven effective in preventing autoimmune disease (5, 6, 7). The induction and maintenance of Treg in vivo is dependent on a variety of factors that include costimulatory molecules such as CD80 and CD86 (8, 9, 10, 11) and cytokines such as TGF-β (6) and IL-2 (12).

Dendritic cells (DC) play an important role both in initiating immunity and in maintaining self-tolerance. DC subsets specifically stimulate the activation and differentiation of Treg types (9, 10, 13, 14). Recent reports have suggested that immature DC can induce the development of functional suppressor cells that express CD4 and CD25 (13) and it has been proposed that immature DC might be a useful therapeutic option in the case of autoimmunity (9, 10). However, the requirement for high levels of costimulatory molecules for the differentiation of Th2 cells (15) as well as suppressor CD4⁺CD25⁺ T cells (8) suggests that immature DC might not effectively generate these populations of Treg in vivo. Indeed, we have previously shown that immature DC, grown in GM-CSF alone (GMDC), are ineffective at preventing diabetes in NOD mice (16), whereas mature DC, grown in GM-CSF plus IL-4 (G4DC), can reproducibly protect these mice (16, 17, 18). In addition, we have observed that G4DC therapy causes an increase in CD4⁺CD25⁺ T cells (19). More recently, it was shown that mature DC populations phenotypically similar to therapeutic G4DC were able to induce expansion of sorted CD4⁺CD25⁺ T cells from NOD and BDC2.5 TCR-transgenic (Tg) mice (20, 21). These DC-expanded CD4⁺CD25⁺ T cells expressed Foxp3 and were able to prevent diabetes development in vivo.

In this report, we have further examined this phenomenon in the BDC2.5 TCR-Tg system using GMDC and G4DC. We found that both DC populations induced the expansion of CD4⁺Foxp3⁺ Treg in the absence of exogenous TGF-β and that this expansion was Ag dose dependent. Low doses of cognate Ag preferentially expanded Treg and this was observed for two different peptide mimetopes known to stimulate BDC2.5 T cells (22). Furthermore, we provide evidence that G4DC may naturally present low doses of self-Ags to self-reactive T cells in a manner that favors Treg development. The induction of Treg in this in vitro culture system was inversely correlated with the level of signaling via the Akt/mammalian target of rapamycin (mTOR) pathway. Analysis of the cytokine requirement of this phenomenon revealed that IL-6 production by T cells, but not DC, inhibited the induction of Treg at low Ag doses. These results suggest that it is possible to enhance Treg development and function in vivo by manipulating both the dose of specific Ag and the type of DC that is targeted.

NOD, C57BL/6, and BALB/c mice were purchased from The Jackson Laboratory. BDC2.5 and BDC2.5 Foxp3-GFP TCR-Tg mice were obtained from Dr. C. Benoist (Joslin Diabetes Center, Boston MA). DO11.10 mice were obtained from Dr. M. Jenkins (University of Minnesota, Minneapolis, MN). OT-II mice and IL-6^−/− mice were obtained from Dr. L. Falo and Dr. A. J. Demetris, respectively (University of Pittsburgh, Pittsburgh, PA). All mice were housed in a specific pathogen-free facility at the University of Pittsburgh and were treated under Institutional Animal Care and Use Committee-approved guidelines in accordance with approved protocols.

DC were generated as previously described (16, 23) by growing bone marrow precursors for four (NOD) or six (C57BL/6 and BALB/c) days in RPMI 1640 containing 5 ng/ml GM-CSF (GMDC) or 5 ng/ml GM-CSF and 2 ng/ml IL-4 (G4DC). GM-CSF was purchased from R&D Systems. IL-4 was from PeproTech. GMDC were purified using CD11c magnetic beads (Miltenyi Biotec), yielding a population of largely immature cells. G4DC were purified using a 13.5% Histodenz gradient (Sigma-Aldrich), as previously described (16, 23), to enrich for the mature population. Mature DC derived from these cultures expressed high levels of CD86, CD80, CD40, and MHC class II as previously described (16) (supplemental Fig. 1⁴).

CD4⁺ T cells were purified from dissociated spleens and lymph nodes of NOD or BDC2.5 mice by negative selection using a CD4⁺ T cell isolation kit (Miltenyi Biotec). For naive T cells, MACS-purified CD4⁺ T cells were stained with anti-CD4-PerCP, anti-CD45RB-FITC, anti-CD62L-PE (BD Biosciences), and anti-CD25-allophycocyanin (eBioscience) and sorted on a FACSAria (BD Biosciences) to yield the CD4⁺CD62L⁺CD45RB^highCD25^neg naive fraction. Before stimulation, 5–50 × 10⁶ CD4⁺ T cells were labeled with 2.5 μM CFSE (Invitrogen/Molecular Probes) diluted in PBS containing 1% FBS, according to the manufacturer’s instructions.

DC and T cells were cocultured in 96-well U-bottom plates (BD Falcon). DC were seeded at 2 × 10⁴ cells/well and peptides were added at least 2 h before the addition of CFSE-labeled CD4⁺ T cells (1 × 10⁵/well). Cultures were maintained in complete DMEM containing penicillin and streptomycin, l-glutamine, sodium pyruvate, nonessential amino acids, HEPES buffer, 2-ME, and 10% FBS. Cells were fed on days 3, 4, 5, and 6 by replacing half of the medium. Cells were harvested at the indicated times for flow cytometric analysis.

Cell surface staining was performed in PBS containing 2% FBS with the following Abs: anti-CD11c-PE, anti-CD25-allophycocyanin (eBioscience), and anti-CD4-PerCP (BD Biosciences). Intracellular Foxp3 was stained with anti-Foxp3-Pacific Blue Ab and staining buffers from eBioscience according to the manufacturer’s instructions. Staining with Abs against phospho-S6 and phospho-Akt (Cell Signaling Technologies) was performed in conjunction with the Foxp3 Ab and staining buffers. Stained cells were analyzed on an LSR II (BD Biosciences) using FACSDiva software (BD Biosciences).

CD4⁺ T cells from TCR-Tg BDC2.5Foxp3GFP reporter mice were stimulated with G4DC and low-dose (10 pg/ml) AV10 peptide for 5 days to induce Treg proliferation. After 5 days, T cells were stained with PerCP-anti-CD4 and allophycocyanin-anti-CD25 mAbs (BD Biosciences). GFP-Foxp3⁺CD25^high Treg were then sorted on a FACSAria (BD Biosciences). CD4⁺GFPFoxp3⁺CD25^high T cells were also isolated from the spleen of BDC2.5 Foxp3GFP reporter mice to serve as a freshly isolated Treg control. Treg purified from fresh spleen or in vitro DC:T cell cultures were added at various ratios to freshly isolated, PKH26-labeled (24) CD4⁺ BDC2.5 responder T cells. The mixed T cells were then stimulated with G4DC plus 1 ng/ml AV10 for another 5 days, then stained with a Pacific Blue-anti-CD4 mAb (eBioscience) and analyzed on an LSR II flow cytometer (BD Biosciences). Proliferation of responders was measured by recording PKH mean fluorescence intensity (MFI) in gated CD4⁺GFP^neg T cells.

Culture supernatants were harvested after 48 h of DC/T cell coculture and assayed for the presence of 17 cytokines/chemokines (IL-1β, IL-1α, IL-2, IL-4, IL-6, IL-9, IL-10, IL-12p40, IL-2p70, IL-13, IL-17, TNF-α, IFN-γ, KC, MIP-1β, MCP-1, and RANTES) using a Milliplex Luminex kit (Millipore) according to the kit instructions. TGF-β concentrations were measured by ELISA using a Ready-Set-Go Human/Mouse TGF-β ELISA kit (eBioscience). IL-6 production by purified DC was determined by ELISA using supernatants from DC cultured for 24 h with or without LPS.

We have previously established that diabetes can be effectively prevented in NOD mice with a single i.v. injection of mature bone marrow-derived G4DC, whereas immature GMDC failed to prevent diabetes (16, 18, 23). As previously shown (16), G4DC and GMDC show marked differences in phenotype, consistent with their differing degree of maturity (supplemental Fig 1). In this study, we compared the ability of therapeutic G4DC and nontherapeutic GMDC to induce or expand islet-specific Foxp3⁺ Treg. To do this, we used CD4⁺ T cells from BDC2.5 TCR-Tg mice. Although the specific Ag recognized by BDC2.5 T cells has not yet been identified, several peptide mimetopes have been described (22). We chose two of the peptides that differ in their apparent affinity for the BDC2.5 TCR: the low-affinity KV11 peptide (KVAPVWVRMME), which has an intermediate EC₅₀ of 75 nM, and the high-affinity AV10 peptide (AVRPLWVRME) that has a much lower EC₅₀ of 0.7 nM (22) (supplemental Fig. 1B). We tested the ability of therapeutic G4DC and nontherapeutic GMDC, presenting these two peptides, to induce proliferation and Foxp3 expression in CD4⁺ T cells purified from BDC2.5 TCR-Tg mice. After 7 days of stimulation with G4DC or GMDC and increasing doses of KV11 or AV10 peptides, CD4⁺ BDC2.5 T cells were stained for expression of Foxp3 and analyzed for proliferation by CFSE dilution. G4DC loaded with high doses (40 μM) of the low-affinity KV11 peptide induced robust proliferation of islet-specific BDC2.5 CD4⁺ T cells, which were almost exclusively Foxp3⁻ (Fig. 1,A, left panels). The same G4DC loaded with a low dose (0.4 μM) of KV11 induced less CD4⁺ T cell proliferation overall, but a substantial number of Foxp3⁺ Treg were observed within the proliferating T cell population. A similar Ag dose-dependent pattern was observed with the high-affinity AV10 peptide, with maximal Treg proliferation being observed at low Ag doses and Th proliferation being favored at high Ag doses. However, ∼500-fold lower doses of AV10 were required to induce the same Treg expansion that was obtained with KV11 (0.8 nM vs 0.4 μM, respectively; Fig. 1,B). The shift in the dose-response curves for KV11 and AV10 peptides is consistent with the reported difference in the EC₅₀ of these two peptide mimetopes (22). Interestingly, nontherapeutic GMDC also induced the proliferation of Foxp3⁺ Treg at low Ag doses and Foxp3⁻ T cells at high peptide doses. Reflecting their lower expression of MHC class II (16), GMDC required higher doses of AV10 and KV11 peptides (Fig. 1,A, right panels) to elicit the same relative proliferation of Th and Treg obtained with G4DC (Fig. 1 A, left panels).

To determine whether expansion of Treg at low Ag doses is a unique feature of autoantigen-specific T cells, we examined the response of CD4⁺ T cells from DO11.10 (OVA-specific) mice when stimulated with autologous DC presenting their cognate peptide Ags (OVA₃₂₃). As seen with BDC2.5 T cells, DO11.10 T cells responded in an Ag dose-dependent fashion, with Treg proliferation being favored at low Ag doses (Fig. 1 C). This suggests that proliferation of Treg in response to low level stimulation is a general feature of CD4⁺ T cells, regardless of their specificity for self- or non-self Ags.

It was important to show that the Foxp3⁺CD4⁺ T cells expanded in these cultures were functionally suppressive. To this end, CD4⁺ T cells from BDC2.5 Foxp3GFP reporter mice were stimulated for 7 days with G4DC plus low-dose AV10 peptide and the expanded CD25^highFoxp3⁺ Treg were FACSorted and tested in an in vitro suppression assay. The in vitro-expanded CD4⁺CD25^highFoxp3⁺ Treg had suppressor activity that was equivalent to, if not better than, that of CD4⁺CD25^highFoxp3⁺ Treg sorted from fresh spleen (Fig. 1 D). As expected, control CD4⁺CD25⁻Foxp3⁻ T cells, either in vitro-cultured or freshly isolated, did not suppress responder T cell proliferation (data not shown).

It was possible that DC presenting low Ag doses may have 1) expanded preexisting Treg present in the starting CD4⁺ population and/or 2) induced de novo Foxp3 expression in naive CD4⁺ T cells. To address these possibilities, naive CD4⁺CD25^negCD62L⁺CD45RB^high T cells were sorted to >99.5% purity from BDC2.5 splenocytes and stimulated by G4DC loaded with varying doses of AV10 peptide, as described above. After 7 days of stimulation by G4DC and low-dose peptide (0.8 nM AV10), Foxp3 expression was induced in 14.7% of the sorted naive CD4⁺ T cells, compared with 20.4% Foxp3⁺ expression when total CD4⁺ T cells were similarly stimulated (Fig. 1 E). Thus, DC presenting low levels of Ag can induce de novo expression of Foxp3 in naive CD4 T cells.

During the course of these experiments, we noticed that NOD G4DC were able to induce expansion of BDC2.5 Treg in the absence of any exogenous peptide (Fig. 2,A). This phenomenon was more prominent when stimulating whole compared with naive CD4⁺ BDC2.5 T cells (data not shown) and was also observed when using unloaded C57BL/6 DC (supplemental Fig. 2), which can also present the AV10 peptide and stimulate BDC2.5 Th proliferation. In contrast, OVA-specific CD4⁺ T cells from DO11.10 or OT-II mice failed to proliferate in response to unloaded BALB/c or C57BL/6 G4DC, respectively (Fig. 2,A and supplemental Fig. 2). Unloaded GMDC did not induce significant proliferation of either BDC2.5 or DO11.10 CD4⁺ T cells (Fig. 2 A). These results suggest that autoreactive BDC2.5 T cells can recognize low levels of endogenous autoantigens presented by mature G4DC, whereas the DO11.10 or OT-II T cells, which are specific for the foreign Ag OVA, do not cross-react with any self-Ags expressed by unloaded DC.

Since unloaded G4DC were able to present endogenous autoantigen to islet-specific BDC2.5 CD4⁺ T cells, it was possible that they could also present peptides from other endogenous autoantigens and expand autoreactive Treg from the polyclonal NOD T cell repertoire. When CD4⁺ T cells from NOD mice were stimulated with G4DC, robust expansion of Treg was observed, whether or not the DC were loaded with the BDC-specific antigenic peptides (Fig. 2,B). However, unloaded GMDC failed to elicit a significant response from either BDC2.5 or NOD T cells (Fig. 2, A and C). When GMDC were loaded with KV11 peptide, proliferation of BDC2.5 T cells was observed (Fig. 1,A), but NOD T cells failed to respond to peptide-pulsed GMDC (Fig. 2 C). These results are consistent with our previous reports that G4DC, but not GMDC, can prevent diabetes in NOD mice, whether or not they are preloaded with islet autoantigens (16). The robust proliferation of polyclonal Treg from NOD mice in response to unloaded G4DC but not GMDC is possibly related to the previously described increased frequency of autoreactive T cells in NOD mice, which has been related to defects in negative selection (25).

The Ag dose-response curves in both BDC2.5 and DO11.10 T cells suggested that Treg vs Th expansion is related to the strength of signaling via the TCR, such that strong TCR signaling favors Th expansion and low TCR signaling allows Treg expansion. Indeed, two recent reports described a role for the Akt/mTOR pathway in suppressing Foxp3 expression in CD4⁺ T cells (26, 27). To examine the role of Akt/mTOR signaling in our system, we measured intracellular levels of phosphorylated S6 ribosomal protein (pS6), a direct target of p70 S6 kinase that is regulated by mTOR (28). BDC2.5 or DO11.10 CD4⁺ T cells were stimulated with G4DC, as before, and levels of intracellular Foxp3 and pS6 were analyzed by flow cytometry at 18 h, 3 days, and 7 days. After 18 h of stimulation with G4DC presenting high Ag doses, BDC2.5 T cells displayed high levels of intracellular pS6 that decreased as the Ag dose was lowered (Fig. 3,A, upper panels). A similar pattern was observed in DO11.10 T cells, with a distinct population of Foxp3^neg cells expressing high levels of pS6 emerging as the Ag dose increased (Fig. 3 A, lower panels).

We examined the time course of pS6 and Foxp3 expression in BDC2.5 T cells activated with increasing doses of AV10 peptide presented by G4DC (Fig. 3,B). The highest levels of pS6 (Fig. 3,B, left panel) were seen at 18 h and there was a marked drop in pS6 levels once the Ag dose fell below 10 nM. The levels of pS6 diminished over time and, by day 7, no longer correlated with Ag dose. Examination of Foxp3 expression at these same time points (Fig. 3,B, right panel) revealed a transient increase, across the whole Ag dose range, in the percentage of Foxp3⁺CD4⁺ T cells after 18 h of coculture with G4DC. In T cells stimulated with high Ag dose, the frequency of Foxp3⁺ T cells was reduced at day 3 and Foxp3 expression was completely absent by day 7. Under low Ag stimulation conditions, the percentage of Foxp3⁺CD4⁺ T cells was increased by day 3 and this population was further expanded by the seventh day. The inverse correlation between pS6 expression at 18 h and Foxp3 expression at day 7 (Fig. 3,C) suggests that differences in the strength of TCR signaling detected after 18 h of DC/T cell coculture result in sustained differences in Foxp3 expression. Results from three independent experiments, plotted on the same graph, show that this inverse correlation of pS6 and Treg frequency is remarkably consistent (Fig. 3 D).

We examined this same relationship through experiments in which the high-affinity (AV10) or low- affinity (KV11) peptide was presented by either GMDC or G4DC. Strikingly, despite the two-log difference in the AV10 or KV11 dose required to induce maximal numbers of BDC2.5 Foxp3⁺ Treg by G4DC (Fig. 4,A), the inverse relationship between pS6 MFI and frequency of Foxp3⁺ T cells was the same (Fig. 4,B). A similarly consistent relationship between pS6 and Treg frequency was observed in the stimulation of BDC2.5 T cells by either G4DC or GMDC presenting the AV10 peptide (Fig. 4 C). These results suggest that a threshold level of signaling via the Akt/mTOR pathway may be required to inhibit Foxp3 expression and this can be achieved with either G4DC or GMDC.

From the results described above, the strength of signal 1 delivered via the TCR appears to be critical for controlling the relative expansion of Treg and Th. To test whether Treg induction by weak TCR stimulation could be recapitulated in the absence of APCs, purified CD4⁺ BDC2.5 T cells were stimulated in vitro with anti-CD28 mAb and decreasing doses of anti-CD3. Similar to the Ag dose response observed with the DC-stimulated T cells, expansion of Foxp3⁺CD4⁺ T cells was favored at low anti-CD3 doses, whereas high concentrations of anti-CD3 led to robust Th proliferation (Fig. 4,D). At very low concentrations of anti-CD3, no stimulation was observed. Consistent with our results using DC stimulation (Fig. 3,C), the dose of anti-CD3 stimulation correlated positively with pS6 levels at 18 h and inversely with the frequency and proliferation of Foxp3⁺ Treg after 5 days of culture (Fig. 4 E). The results from these experiments indicate that the level of TCR signaling via the Akt/mTOR pathway is a primary determinant of Treg vs Th commitment.

Recent reports have suggested that inflammatory cytokines, such as IL-6, produced by DC can inhibit TGF-β-mediated Treg induction (29, 30, 31). However, our culture system did not include exogenous TGF-β and it could not be detected in the cultures by ELISA (data not shown). Nonetheless, cytokines such as IL-2 and IL-6 are known to affect Akt signaling (32, 33, 34), which could potentially modulate low-dose Treg induction. CD4⁺ BDC2.5 T cells were stimulated by G4DC or GMDC in the presence of AV10 peptide, and the levels of various cytokines in the supernatants were examined after 48 h by multiplex Luminex assay. Consistent with our previously published results (16, 23), G4DC induced higher levels of type 2 cytokines than GM DC (data not shown). IL-2, IL-6, IL-10, IL-17 (Fig. 5,A), IL-13, IFN-γ, and TNF-α (data not shown) increased in an Ag dose-dependent fashion. IL-2, IL-6, IL-13, and IL-17 concentrations were higher in cultures with the more mature G4DC, compared with the less mature GMDC, whereas IL-10, TNF-α, and IFN-γ production levels were similar regardless of the maturation state of the DC used in these cultures (Fig. 5,A and data not shown). IL-2, IL-6, and TNF-α were the only cytokines observed in cultures of naive T cells stimulated with G4DC and the AV10 peptide, and the pattern of secretion was similar to that seen with cultures of total CD4⁺ T cells (Fig. 5,B and data not shown). In contrast, no IL-13, IL-17, IL-10, or IFN-γ was detected in cultures of naive T cells (Fig. 5,B and data not shown). Because we observed a dose-dependent induction of Foxp3⁺ T cells in cultures of naive T cells (Fig. 1 E), we conclude that IL-13, IL-17, IFN-γ, and IL-10 do not influence the induction of Treg by low Ag doses.

Previous studies have suggested that IL-6 production by DC in response to TLR ligands inhibits Treg induction (30, 31). We observed an inverse correlation between IL-6 production and the induction of Foxp3⁺ T cells (Fig. 5,C, right panel), and IL-6 levels were highly sensitive to the maturation level of the DC and the dose of peptide, with the highest levels being produced in cultures with G4DC and AV10 peptide (Fig. 5,C, left panel). Therefore, we sought to rule out the possibility of LPS contamination in our system. Examination of cytokine production in DC/T cell cocultures revealed that no IL-1α, IL-1β, IL-12p40, or IL-12p70 was produced at any dose of peptide used (data not shown). In addition, exposure of DC to peptide alone failed to induce increased expression of CD86 or CD40 (data not shown). Furthermore, exposure of GMDC and G4DC to LPS revealed that GMDC produce significantly higher levels of IL-6 than G4DC (Fig. 5,D). This is in contrast to the higher IL-6 levels seen in T cell cocultures with G4DC vs GMDC (Fig. 5 C). These data indicate that the peptides and other solutions used in these cultures were not contaminated with LPS and that the levels of IL-6 detected in the supernatants are due to Ag-specific DC-T cell interactions.

The inverse correlation between IL-6 levels and Foxp3 induction was intriguing and warranted further investigation. To determine whether IL-6 in these cultures was produced by DC, T cells, or both and whether this may affect Treg induction, we performed experiments using DC or T cells isolated from IL-6 knockout mice. Wild-type (WT) CD4⁺ T cells stimulated with anti-CD3 and anti-CD28 produced low, but detectable, amounts of IL-6, which increased in a dose-dependent fashion relative to the concentration of stimulatory plate-bound anti-CD3 (Fig. 6,A). Furthermore, the frequency of Foxp3⁺ Treg induced by low doses of anti-CD3 was enhanced in T cells lacking IL-6 (Fig. 6, B, left panels, and C). As seen earlier (Fig. 4,D), high doses of anti-CD3 stimulation prevented Treg induction and also abrogated any differences between WT and IL-6^−/− T cells (Fig. 6, B, right panels, and C).

To determine the contribution of IL-6 from DC, CD4⁺ T cells from OT-II mice were stimulated with OVA_323–339 peptide presented by G4DC from WT or IL-6^−/− mice. Consistent with our observations in the BDC2.5 and DO11.10 TCR-Tg systems, Foxp3⁺ OT-II Treg were induced by G4DC presenting low doses of cognate peptide (Fig. 6, D and E). However, the efficiency of low-dose Treg induction by G4DC from IL-6^−/− mice was similar to that of WT G4DC (Fig. 6, D and E). Thus, low-dose Treg induction by G4DC does not appear to be affected by IL-6 from DC. In addition, we observed low levels of IL-6 production in cultures of IL-6^−/− DC with OT-II T cells (Fig. 6 F), demonstrating that T cells can be a source of IL-6 in this system.

The experiments presented here demonstrate that Treg can be induced and expanded by DC that present low doses of cognate Ag and initiate a program of weak TCR signaling via the Akt/mTOR pathway. This phenomenon was observed in three different TCR-Tg systems: NOD-BDC2.5, BALB/c-DO11.10, and C57BL/6-OT-II. The induction of Treg by DC at low Ag doses did not require the addition of exogenous TGF-β or IL-2. We have also shown that low-level Ag presentation by DC can drive both the expansion of preexisting Treg and the conversion of naive T cells into de novo Treg. In addition, our results show that unloaded G4DC, but not GMDC, can expand Treg from BDC2.5 mice and from the polyclonal NOD T cell repertoire, suggesting that G4DC may present low levels of endogenous autoantigens, thus bypassing the need for exogenous delivery of autoantigens.

Despite very different phenotypes in terms of expression of maturation markers and cytokine production, both GMDC and G4DC could expand either Treg or Th, provided that an appropriate dose of Ag was applied. Comparison of these two distinct DC populations in combination with two peptides with different affinities for the TCR suggested that the strength of the TCR signal is a critical factor in determining the balance between Th and Treg. This was supported by our ability to induce Treg expansion by stimulating purified CD4⁺ T cells with low doses of anti-CD3, plus CD28, in the absence of any DC. Our data suggest that there is a low level of TCR stimulation that favors Treg development and this can be achieved when DC present low doses of Ag. Below this threshold, no stimulation occurs. In the context of strong TCR stimulation, full T effector cell activation and differentiation occur.

We observed a strong inverse correlation between pS6 levels induced after 18 h of stimulation and the resulting development and proliferation of Foxp3⁺ Treg. Furthermore, the inverse relationship between Akt/mTOR and Foxp3 remained constant, even though the Ag dose required for maximal Treg expansion varied depending on peptide affinity and the phenotype of the stimulating DC. Our results are consistent with recent reports that Akt/mTOR signaling negatively regulates Foxp3 expression in CD4⁺ T cells (26, 27). In studies by Sauer et al. (27), interruption of the TCR signaling after 18 h and blockade of the Akt/mTOR pathway, using rapamycin, resulted in increased Treg induction. In addition, this group reported an increase in Foxp3 expression at 18 h in ∼10% of T cells, suggesting that early TCR signaling leads to enhanced Foxp3 expression in a subpopulation of T cells. Similarly, in our studies, an increased percentage of Foxp3⁺ cells was observed after 18 h of stimulation with all doses of Ag stimulation. Over time, the number of Foxp3⁺ T cells decreased at the higher Ag doses and increased at lower Ag doses. It has been suggested that a second round of Akt/mTOR signaling via cytokine receptors such as IL-2R regulates the expression of Foxp3 in Treg and effector T cells (35). In our study, the expansion of Treg was inversely correlated with the levels of IL-2 produced and we are exploring the role of STAT5 in this system.

It is now well established that TGF-β can drive the de novo induction of Ag-specific Treg by DC presenting high Ag doses (6, 36, 37) and that this is antagonized by IL-6, which instead promotes Th17 development (29, 38). However, we observed conversion of naive T cells into Foxp3⁺ Treg by low Ag doses without the addition of exogenous TGF-β and found no evidence of TGF-β production in these cultures. We conclude that induction of Treg by low Ag dose and weak TCR/Akt/mTOR signaling occurs independently of soluble TGF-β. This is consistent with a recent report that TCR^low BDC2.5 T cells mediate long-term protection from diabetes in NOD.SCID mice independently of TGF-β, whereas TCR^high BDC2.5 T cells induce diabetes (39). Nonetheless, we observed a negative correlation between IL-6 production and Foxp3 expression. Because IL-6 is known to signal via the Akt/mTOR pathway (34, 40, 41), this cytokine could also antagonize Treg induction by increasing Akt/mTOR signaling. We therefore examined the ability of IL-6, produced by either T cells or DC, to inhibit Treg induction. These studies revealed that T cell-derived IL-6 could inhibit Treg induction at the very lowest Ag dose. Thus, we observed increased low-dose Treg expansion in IL-6^−/− T cells. At higher doses, the absence of IL-6 did not prevent the expansion of Th cells. In addition, we did not observe a significant difference in pS6 expression at the high doses but IL-6^−/− T cells expressed lower levels of pS6 at the lowest dose (data not shown). Interestingly, stimulating WT T cells with IL-6^−/− DC did not alter the dose-dependent induction of Treg. This would suggest that the amount of IL-6 produced by T cells in these cultures was sufficient to inhibit Treg induction at the higher Ag doses or that other signals delivered by DC compensated for the lack of IL-6. Thus, in the absence of any TLR ligand-mediated DC activation, IL-6 production by T cells may play a more important role than DC-derived IL-6 in the inhibition of Treg expansion.

DC can function in apparently contradictory ways, either to induce strong Th responses, to induce deletional tolerance, or to activate Treg. The challenge is to identify a population of DC with the desired function (42, 43, 44, 45). Work in recent years suggests that these diverse functions are not the property of defined lineages of DC, but rather that the same DC can function in diverse ways depending on surrounding stimuli. In the absence of pathogen-derived stimuli, resident steady-state DC, present in lymphoid organs and the periphery, function to maintain self-tolerance (42, 46). Furthermore, evidence of increased levels of TCR phosphorylation in CD4⁺CD25⁺ T cells suggests that Treg recognize self-Ags in vivo (47). Thus, DC may function to induce/maintain Treg populations in vivo by presenting low levels of endogenous autoantigens. This is consistent with our observation that unloaded NOD G4DC could induce low-level Akt/mTOR signaling and expand Treg from BDC2.5 mice in the absence of any exogenous peptide Ag, TGF-β, or IL-2. Interestingly, immature GMDC, reflecting their lower expression of MHC class II, could only expand BDC2.5 Treg if loaded with peptide. This result correlates with a previous report from our group that loaded or unloaded G4DC, but not GMDC, can prevent diabetes when adoptively transferred to NOD mice (16).

Our results show a T cell-intrinsic correlation between weak TCR signaling and expansion of Foxp3⁺ Treg. Previous studies have shown that administration of low-dose peptide Ag in vivo expands Treg in a DC-dependent manner (48, 49), and this successfully induces tolerance in mouse models of autoimmunity (50, 51). Hence, administration of peptide autoantigens presents an attractive method of inducing Ag-specific tolerance. However, it has been shown that repeated administration of self-peptides derived from insulin or GAD65 can induce fatal anaphylaxis in NOD mice (52, 53) and a similar phenomenon was observed in experimental autoimmune encephalomyelitis models (54). The insulin study draws parallels with our own studies, as anaphylaxis was observed when a higher dose (100 μg/mouse) was used, whereas the lower dose (10 μg/mouse) instead protected NOD mice from diabetes (52). In order for such therapies to be effective, it will be important to establish accurate protocols for the delivery of specific Ags to the appropriate DC at the correct dose and under optimal conditions for the generation of Ag-specific Treg.

We thank Dewayne Falkner for assistance with flow cytometry and cell sorting, Huijie Sun for Luminex assays, and Ee Wern Su for assistance with cell signaling studies.

The authors have no financial conflict of interest.