Reprogrammed FoxP3⁺ T Regulatory Cells Become IL-17⁺ Antigen-Specific Autoimmune Effectors In Vitro and In Vivo¹

Immunotherapy aims to harness the immune system to impact the treatment of a wide variety of diseases. However, for any specific therapy to be successful, it is critical that it direct a specific immune attack on the disease while protecting the host from aggressive autoimmunity. This effect is particularly a challenge in the case of cancer, where tumor growth is driven by mutations or abnormal expression of normal cellular proteins. To the host immune system, these “tumor-associated Ags” are likely recognized as an extension of self, allowing the tumor protection via active immune tolerance.

T regulatory cells (Tregs)³ actively suppress both physiological and pathological immune responsiveness and are thus central players in the establishment of self-tolerance and the maintenance of immune homeostasis (1, 2, 3, 4). Natural Tregs originate in the thymus early in life as CD4⁺ T cells (5, 6), constitutively display the IL-2R α-chain (CD25), and characteristically express FoxP3, a master regulator of Treg functions (7, 8, 9, 10, 11). Tregs also arise in response to newly encountered T cell epitopes in the absence of costimulatory signals (12, 13, 14). Mutation of the FoxP3 gene results in Treg deficiency and autoimmune inflammatory disorders, referred to as Scurfy disease in mice and IPEX (immune dysregulation, polyendocrinopathy, enteropathy X-linked syndrome) in humans (15, 16, 17). Treg-mediated suppression of proliferation and secretion of cytokines by effector-type T cells involves cell-to-cell contact, engagement of the CTLA-4 receptor on T cells, killing of effector cells and APCs, as well as the secretion of IL-10 and TGFβ (4).

It is becoming increasingly clear that development of effective therapeutic approaches to manipulate Treg functions and release immune effector mechanisms from stringent regulatory constraints will be crucial to advancing immune-based treatments of autoimmunity or cancer (18, 19). However, many details are not clear, such as how Tregs communicate with other cell types like dendritic cells (DC), how the balance of tolerance and immunity is maintained, and what becomes of Tregs as immune tolerance shifts toward active immunity. We recently characterized a human B7-DC cross-linking Ab (B7-DC XAb) that binds to B7-DC/PD-L2 molecules on both murine and human DC and cross-links the molecules on the cell surface (20). B7-DC XAb-activated DCs (DC^XAb) can redirect established Th2 cells toward a Th1 phenotype in an Ag-specific manner (21). In addition, administration of B7-DC XAb to mice prevents tumor growth (22) and allergic asthma (23). This therapeutic effect can be mimicked by adoptive transfer of DC^XAb, suggesting that the Ab acts directly on DC, not by preventing B7-DC-PD-1 interaction (24, 25).

DCs also present Ag to the T regulatory compartment of the immune system. Recent studies have documented the down-regulation of FoxP3 in the Tregs (9, 26). However, the functions acquired by those converted cells have not been explored. Our experiments using DC^XAb to induce antitumor immunity suggested to us that activated DC may modulate Treg functions. In testing this hypothesis, we found that DC^XAb not only caused Tregs to down-regulate their suppressive functions, but also to adopt an effector phenotype that promotes Ag-specific autoimmunity in animal models.

C57BL/6J, BALB/CJ, DO11.10, OT-II, B6.129S2-IL6^tm1Kopf/J, C57BL/6-Tg(Ins2-OV)59Wehi/WehiJ (high), and C57BL/6-Tg(Ins2-OV)307Wehi/WehiJ (low) mice and rat insulin promoter (RIP)-OVA (27) mice were purchased from The Jackson Laboratory. Female mice were used in diabetes experiments and in experiments using BALB-neuT mice (H-2^d). Otherwise, male mice were used. All experiments were conducted with Institutional Animal Care and Use Committee oversight. The mouse breast cancer cell line transfected with rat Her-2/neu (A2L2), a mock-transfected parental cell line (66.3neo), and a cloned cell line from established carcinoma that spontaneously arose in a BALB-neuT mouse (TUBO) (28, 29, 30) were gifts from E. Celis (Moffitt Cancer Center, Tampa, FL). Peptide p66 (TYVPANASL) derived from rat Her-2/neu was synthesized in Mayo Protein Core Facility. Hybridoma clone PC61 (anti-CD25) was a gift from W.-Z. Wei (Wayne State University, Detroit, MI). Anti-mouse CD4-PE (RM4-5) Ab and anti-mouse IFN-γ-PE (XMG1.2) Ab were from BD Biosciences. Anti-mouse FoxP3-allophycocyanin (FJK-16s), anti-DO11.10 TCR-FITC (KJ1-26), anti-mouse TNF-α-PE (MP6-XT22), anti-mouse IL-10-PE (JES5-16E3), and anti-mouse IL-17A-PE (clone eBio17B7) Abs were from eBioscience.

The human monoclonal IgM Abs sHIgM12 (B7-DC XAb) and sHIgM39 (isotype-matched control) arose from a screen to identify Abs that bound mouse DC using a pool of sera from patients with monoclonal gammopathies (20). The Abs were purified as described (20). Due to the B7-DC-dependence of its biologic properties, the requirement for the pentameric form, and the observed signals (S. Radhakrishnan, L. N. Arneson, J. L. Upshaw, C. L. Howe, S. J. Felts, M. Colonna, P. J. Leibson, M. Rodriguez, and L. R. Pease. TREM-2-mediated signaling induces antigen uptake and retention in mature dendritic cells. Submitted for publication.) in DC elicited by Ab binding (20, 21), we refer to this novel reagent as B7-DC cross-linking Ab (B7-DC XAb). DC treated with the B7-DC XAb reagent are referred to as DC^XAb.

Peptide and CpG immunization strategy (see Fig. 1) and analysis of the ensuing immune response was conducted as previously described (28). For the in vivo Treg depletion experiments, 0.5 mg of anti-CD25 mAb (PC61) was injected i.p. on days −3, −2, and −1 before immunization with peptide on day 0. Other groups of mice were injected i.p. or s.c. with 10 μg of sHIgM39 control or B7-DC XAb on days −1, 0, and +1 relative to immunization. For all other experiments, mice received Ab, Ag, pretreated DC, or T cells i.v. as indicated.

Splenocytes were isolated from pooled spleens harvested from at least three mice in each treatment group. Tregs were isolated by positive selection using Mouse T Regulatory Isolation kit from Miltenyi Biotec (31), as per the manufacturer’s protocol. Briefly, splenocytes were incubated with anti-CD25 Ab coupled to magnetic beads for 15 min before binding to the MACS column. Unbound cells were washed three times with RPMI 1640 plus 10% FBS and used as non-Tregs. Adherent cells (Tregs) were eluted and washed before use. Flow cytometry analysis for FoxP3 expression showed the Treg population to be >95% pure.

DCs were generated from the mouse bone marrow (32). Bone marrow cells were plated (1 × 10⁶/ml) in RPMI 10 containing 10 μg/ml murine GM-CSF and 1 ng/ml murine IL-4 PeproTech. The culture medium was refreshed on day 2 and pulsed with Ag (1 mg/ml), isotype control Ab, or B7-DC XAb (10 μg/ml) on day 6, followed by overnight incubation. Cells were washed on day 7 before use.

Bone marrow-derived wild-type (WT) or IL-6^−/− immature DCs (2 × 10⁶) were pulsed with Ag and treated with isotype control Ab or B7-DC XAb then used to stimulate naive DO11.10 Tregs or OT-II Tregs in culture (together in 24-well plates or separated into the bottom and top partitions of Boyden chambers) at a 1:1 ratio in vitro for 48 h. For in vivo studies, 1.5 × 10⁶ Tregs were adoptively transferred into histocompatible mice along with 3 × 10⁶ Ag-pulsed, Ab-treated bone marrow-derived DC. In experiments involving modulation of endogenous DC to activate Treg cells, CFSE-labeled OT-II Treg cells were transferred into mice along with isotype control Ab sHIgM39 or B7-DC XAb Ab (10 μg/ml) and OVA Ag (1 mg/ml). After 48 h, spleens were removed, pooled, and cells were prepared for analysis.

DO11.10 Tregs, non-Tregs, or combinations of the two were stimulated with serially diluted Ag-pulsed, Ab-treated DC for 72 h. Proliferative responses were monitored by the addition of [³H]thymidine to the cultures for the last 18 h and measuring incorporation as previously described (20).

The frequency of CD8 T cells secreting IFN-γ was measured using ELISPOT (Mabtech) as described (28). Secretion of TGFβ1 into Treg:DC culture supernatants was measured using an ELISA kit from R&D Systems as per the manufacturer’s protocol. Multiplex cytokine assay was performed using a mouse cytokine panel and Bio-Plex Manager software version 4.0, according to the manufacturer’s protocol (Bio-Rad). For cytokine expression by intracellular staining, cells were permeabilized with CytoFix/CytoPerm kit (BD Biosciences) and incubated with the appropriate conjugated Ab at 4°C according to the manufacturer’s suggestions before analysis by flow cytometry as described (20, 21).

RNA was isolated using TRIzol reagent (Invitrogen). The 300 ng of RNA was used for quantitative RT-PCR (7900HT real-time PCR; Applied Biosystems) using FoxP3 forward 5′-CTACTTCAAGTACCACAATATGCGAC-3′, reverse 5′-CGTTGGCTCCTCTTCTTGCGAAACTC-3′ and actin control forward 5′-CGTCTGGACTTGGCTGGCCGGGACCT-3′, reverse 5′-AGTGGCCATCTCCTGCTCGAAGTCTA-3′. Intracellular staining for FoxP3 protein was performed as described for cytokine staining.

B16 tumor cells (5 × 10⁵) were injected s.c. into the flank of WT or IL-6^−/− mice along with 10 μg of control or B7-DC XAb Ab (i.v.) on days −1, 0, +1 as described (22). After 7 days, tumor draining lymph nodes were removed, and effector cells were prepared for in vitro cytotoxicity assays as previously described (33). Additional animals were monitored regularly for the development of tumors and euthanized if and when tumors measured 17 × 17 mm.

DC (2 × 10⁶) derived from WT or IL-6^−/− mice were pulsed with 100 μg of OVA and stimulated with 10 μg of control Ab or B7-DC XAb before injection into RIP-OVA mice. Mice were monitored regularly for blood sugar levels using a glucometer. Pancreata were excised, fixed in 2% formalin, and processed by Mayo Immunohistochemistry Core Facility for staining with anti-insulin Ab.

Eight-week-old male C57BL/6J mice, housed individually, received 300 μg of B7-DC XAb in 100 μl of PBS i.v. or 100 μl of PBS. Mice were observed and weighed periodically. Blood samples were collected on day 17 for hematology and chemistry analyses, including the number and percentage of white blood cells, lymphocytes, granulocytes, monocytes, and RBC as well as hematocrit, mean corpuscular volume, hemoglobin, mean corpuscular hemoglobin, mean corpuscular hemoglobin with concentration, red cell distribution width, mean platelet volume, with platelet counts and blood urea nitrogen, glucose, creatine L, alkaline phosphatase and alanine aminotransferase, and T-Pro measurements levels. Following the blood collection, the mice were euthanized and the heart, brain, spleen, kidney, liver, and lungs were analyzed for gross pathology and sectioned for histopathology in a blinded manner.

Normally distributed quantitative data are represented graphically and in tabular form with SEs of the mean. Qualitative data were analyzed using the χ² distribution.

Host Treg cells that normally function as helpful guardians of immune tolerance become detrimental suppressors of antitumor immune responses in the context of cancer (19). Depletion of host Tregs by systematically treating mice with CD25-specific Abs can allow antitumor responses to become effective. In earlier studies (28) we observed that DC^XAb elicited a robust immunity against tumors expressing rat Her-2/neu. These data suggested that DC^XAb had rendered Tregs in the NeuT mice functionally defective and thus had broken their tolerance of the Her-2/neu (self) Ag. As a test, we set up a direct comparison of the two immunomodulatory therapies. The treatment protocol is outlined in Fig. 1 A. BALB/c-NeuT (rat Her2neu transgenic) animals were treated with CD25-specific Abs or B7-DC XAb in the presence of CpG costimulation then challenged with an immunodominant Her-2/neu peptide determinant p66 (28). Generation of an immune reaction to the peptide challenge was determined by measuring in vitro IFN-γ production by CD8⁺ T cells isolated from the spleens of treated animals when challenged with various tumor cells.

CD8⁺ T cells from Treg-depleted (Fig. 1,B) and from XAb-treated (Fig. 1 C) animals were highly reactive to tumor cells that express Her-2/neu (TUBO or A2L2 tumor cell lines) and were much less reactive to either a non-Her-2/neu tumor (P815) or mock transfected cells (66.3). The similarities in the response patterns between mice treated with Treg-depleting Ab and mice treated with B7-DC XAb were consistent with our hypothesis that B7-DC XAb induces endogenous DC to modulate the function of endogenous Tregs. Because there is the possibility that this type of immune modulation might provide an Ag-specific method to affect self-tolerance, further characterization of this DC^XAb-Treg interaction seemed warranted.

To evaluate directly the hypothesis that treatment with B7-DC XAb down-regulates Treg function, we used a nontumor model (DO11.10) in which we could track reactive T cells with a specific TCR Ab (KJ-126). Treg function was initially monitored by the expression of the Treg-specific transcription factor FoxP3. Enriched CD25⁺ and CD25⁻ splenic T cells from DO11.10 mice were cocultured with OVA-pulsed DC^XAb or DC treated with isotype control Ab (DC^cntrl) for 48 h and assayed for FoxP3 expression. Only CD25⁺ T cells expressed FoxP3 mRNA, and when cultured with DC^XAb, markedly down-regulated FoxP3 mRNA in comparison to Tregs cultured with Ag-pulsed DC^cntrl (Fig. 2,A). Down-regulation of FoxP3 expression was also observed at the protein level (Fig. 2,B). Enriched CD4⁺ DO11.10 T cells (Tregs) or CD25⁻CD4⁺ D011.10 T cells (non-Tregs) were stimulated with OVA-pulsed DC^XAb or DC^cntrl for 48 h, then stained for intracellular FoxP3 and analyzed by flow cytometry. All non-Tregs were FoxP3⁻. Ag-pulsed DC^XAb suppressed the expression of FoxP3 protein by CD4⁺ DO11.10 Tregs. Importantly, there was a requirement for contact between the DC^XAb and the Tregs (Fig. 2,C). When DC and T cells were cocultured in Boyden chambers in which the two cell types were separated by a permeable membrane, FoxP3 expression in the Tregs remained unchanged (Fig. 2 C, compare top right quadrants). DC^XAb not pulsed with Ag were unable to suppress FoxP3 protein levels; time course analysis of the down-regulation of FoxP3 showed only a modest reduction at 36 h with complete down-regulation occurring by 48 h; and DC matured with CpG did not down-regulate FoxP3 expression by CD4⁺CD25⁺ DO11.10 Tregs (data not shown).

We next sought to determine whether down-regulation of Treg function by DC^XAb was restricted only to Tregs specific for a given Ag presented. Therefore, we mixed OVA-specific DO11.10 Tregs with BALB/cJ Tregs that express a wide range of Ag specificities. To this mix we added OVA-pulsed DC^cntrl or DC^XAb. After 48 h, T cells were harvested from the cultures and analyzed for expression of FoxP3. BALB/c Tregs (KJ1-26⁻) and the OVA-specific DO11.10 Tregs (KJ1-26⁺) were distinguished using the KJ1-26 clonotypic marker (Fig. 3,A, top histogram). As shown in Fig. 3,A, recovered DO11.10 Tregs treated with Ag-pulsed DC^cntrl retained FoxP3 expression, whereas DO11.10 Tregs recovered from culture receiving DC^XAb no longer expressed FoxP3 (Fig. 3,A, middle panels). Importantly, expression levels of FoxP3 were normal in the BALB/c CD4⁺CD25⁺ Tregs exposed to either DC (Fig. 3 A, bottom panels), demonstrating that FoxP3 was down-regulated only in those Tregs specific for OVA Ag. We also detected CD4⁻ FoxP3⁺ cells, which may represent previously described CD8⁺ regulatory cells (34).

We next asked whether endogenous DC could mediate the down-regulation of exogenous Tregs in vivo when activated by B7-DC XAb injection. BALB/cJ mice received adoptively transferred DO11.10 Tregs, OVA, and isotype control Ab or B7-DC XAb. Splenic T cells were recovered 48 h later and stained for expression of the DO11.10 clonotypic TCR (KJ1-26), CD4, and FoxP3. As shown in Fig. 3 B, coadministration of Ag with B7-DC XAb, but not with isotype control resulted in the down-regulation of FoxP3 expressed by the transferred DO11.10 T (KJ1-26⁺) cells, whereas the endogenous Tregs (KJ1-26⁻) retained FoxP3 expression. Thus, DC in the host animal can be activated by B7-DC XAb injection and, in turn, regulate FoxP3 expression by Tregs in vivo in an Ag-specific manner.

FoxP3 expression is a marker for Treg phenotype, but it was important to establish whether there was any alteration in Treg function when FoxP3 was down-regulated by DC^XAb. Tregs were isolated from DO11.10 mice, stimulated in vitro with Ag-pulsed DC treated with activating or control Ab, then tested for their ability to suppress an Ag-specific response by effector T cells. Cells were incubated for 72 h and pulsed with [³H]thymidine during the last 18 h (Fig. 4,A). As expected, Tregs activated with Ag-pulsed DC^cntrl incorporated very little [³H]thymidine (Fig. 4,A, left, open circles), whereas CD4⁺CD25⁻ T effector cells responded to Ag and DC (Fig. 4,A, left, filled circles). When Tregs were mixed in increasing numbers with CD4⁺CD25⁻ effectors, the effector proliferation in response to Ag was suppressed (Fig. 4,A, right, open circles). However, when B7-DC XAb was used to modulate the mixed culture, the T effectors responded vigorously (Fig. 4 A, right, filled circles), thus no longer suppressed by the Tregs, and demonstrating that DC^XAb can alter the function of Tregs.

Changes in Treg suppressive activity correlated with changes in Treg cytokine production (Table I). Using a bead-based cytokine assay, we found that IFN-γ, TNF-α, and IL-17 levels were significantly increased in cultures containing DC^XAb and DO11.10 Tregs, whereas the level of the immunosuppressive cytokine IL-10 was low compared with Treg or DC^cntrl cultures. To validate that these cytokines were emanating from the Tregs and not from other cell types in the cultures, DO11.10 Tregs were recovered and analyzed by intracellular flow cytometry. In concordance with the multiplex analysis, IFN-γ, TNF-α, and IL-17 were up-regulated in the CD4 T cells cocultured with DC^XAb, whereas IL-10 was down-regulated compared with controls (Fig. 4,B). The flow cytometry profiles suggest that the recovered T cells could be simultaneously producing all three cytokines. This pattern was confirmed in double-labeling experiments demonstrating that many cells secreting IFN-γ were also secreting TNF-α or IL-17 (Fig. 4,C). TGFβ in the culture supernatants was also down-regulated in DC^XAb-Treg cocultures but not in DC^cntrl-Treg mix and was low in all non-Treg controls (Fig. 4 D). Taken together, these studies indicate that when Tregs interact with DC^XAb, they lose the expression of FoxP3, production of TGFβ or IL-10, and the ability to suppress T effector cell responses and gain a T effector phenotype with the ability to produce the proinflammatory cytokines IL-17, IFN-γ, and TNF-α.

We used enriched Treg populations for the experiments described. Thus, it was possible that the apparent modulation of the Treg phenotype was the consequence of the outgrowth of non-Treg effectors during the course of the experiments. To test for such proliferation, Tregs were labeled with CFSE and stimulated with Ag-pulsed, unlabeled DC^cntrl or DC^XAb for 48 h before characterization. As shown in Fig. 5 A, CFSE high Tregs cultured with DC^cntrl were still FoxP3⁺ after 48 h. In contrast, CFSE^high Tregs cultured with DC^XAb were FoxP3⁻. A similar number of T cells was recovered from each culture, indicating that the phenotype changes were not due to the death of FoxP3⁺ Tregs.

The lack of significant CFSE dilution in cells displaying down-regulation of FoxP3 strongly suggested that the change in Treg phenotype was not due to the outgrowth of contaminating cell types. However, it was still possible that some small amount of non-Tregs in the Treg preparation were developing into effector cells. To evaluate this possibility, we followed the change in both Treg and effector cell markers in response to DC^XAb. First, FoxP3 and IL-17 levels were measured in the purified but unstimulated Tregs and non-Tregs (Fig. 5,B). The Tregs were 97% FoxP3⁺ and >99% IL-17⁻; the non-Tregs were 99% FoxP3⁻ and >99% IL-17⁻. The Treg or non-Tregs (2 × 10⁶ each) were then labeled with CFSE and incubated with DC as in Fig. 5,A. After 48 h, the total number of cells recovered (∼70% of the initial number plated) was the same for each group. CFSE analysis of those cells showed that some non-Tregs (CD25⁻ cells) did undergo at least one cell division in response to Ag-pulsed DC^XAb compared with DC^cntrl (Fig. 5,C, upper left). Yet, the non-Tregs remained FoxP3⁻ and failed to express IL-17 (Fig. 5,C, lower panels). In contrast, CFSE analysis of Tregs (CD25⁺) showed very little dilution upon stimulation with DC^XAb (Fig. 5,C, upper right). Intracellular staining showed that Tregs stimulated with DC^cntrl remained FoxP3⁺ and IL-17⁻, whereas Tregs stimulated with DC^XAb became FoxP3⁻ and IL-17⁺ (Fig. 5 C, middle panels). Because the preparation of non-Tregs did not develop into IL-17-producing cells, whereas the Treg preparation clearly produced IL-17 in the absence of significant CFSE dilution, we conclude that the induced effector phenotype was not due to outgrowth of CD25⁻ cells, but rather that Tregs were converted into IL-17 effectors by Ag-pulsed DC^XAb.

We have previously shown that DC^XAb secrete IL-6 (35). IL-6 plays a dominant role in preventing newly activated T cells from differentiating into Tregs and favoring the development of the Th17 phenotype (26, 36). The importance of IL-6 expression in modulating the phenotype of established Tregs is less clear. Therefore, we evaluated the importance of IL-6 in mediating DC-Treg interactions using DC isolated from an IL-6-deficient subline of the C57BL/6 mouse lineage. We measured the down-regulation of FoxP3 expression in OVA specific OT-II CD4⁺CD25⁺ Tregs, an IL-6 WT subline of C57BL/6. FoxP3 expression was unaffected in Tregs cultured with WT or IL-6^−/− DC pulsed with OVA and treated with control Ab (Fig. 6,A, left panels). As observed previously, FoxP3 was down-regulated in Tregs cultured with WT DC^XAb, but not if the DC were derived from IL-6 deficient mice (Fig. 6,A, right panels). These results suggested that Treg modulation by B7-DC XAb required the ability of the DC to make IL-6. To test this IL-6-dependence further, we adoptively transferred CFSE-labeled OT-II Tregs into C57BL/6 WT or IL-6^−/− mice along with B7-DC XAb or control Ab. After 48 h, spleens were removed and recovered CFSE-labeled cells were analyzed for FoxP3 expression. As observed in vitro, FoxP3 was down-regulated in Tregs transferred into WT mice receiving Ag and B7-DC XAb, but not in Tregs transferred into IL-6^−/− mice (Fig. 6,B, compare right panels). The percentage of CFSE⁺ cells recovered from each treatment group in WT and knockout mice was comparable. Consistent with the data in Fig. 5, FoxP3 was down-regulated in most of the CFSE-marked cells within 48 h. This down-regulation was IL-6-dependent not only using isolated DC in vitro, but also using endogenous DC in vivo.

We reasoned that the modulation of Treg activity by B7-DC XAb or similar treatments would only have the potential to be clinically relevant if the converted T cells could break self-tolerance in model systems that had physiologic relevance. Our initial experiments in Her-2/neu mice suggested this effect was the case (Fig. 1). So, we set out to extend those studies in the context of IL-6^−/− mice and test for specific responses toward tumor Ags. C57BL/6 (WT) and IL-6^−/− mice received B16 melanoma grafts and either B7-DC XAb or control Ab. After 7 days, T cells were isolated from draining lymph nodes and tested ex vivo for cytolytic activity toward B16 tumor or EL4 control cells. Tumor-reactive CTL were found in both groups of animals (Fig. 7, left). Cells from IL-6^−/− mice were only slightly less reactive than WT mice. Despite the licensing of CTL in mice engrafted with tumor, only treatment with B7-DC XAb protected the C57BL/6 mice from otherwise lethal B16 melanoma grafts (Table II); similar treatment of IL-6 knockout mice was not protective. All five WT animals treated with control Ab developed tumors and were sacrificed within 14 days, whereas all five B7-DC XAb-treated mice remained tumor-free for 6 mo (p < 0.05). In contrast, all of the seven IL-6^−/− mice treated with control Ab or with B7-DC XAb succumbed to the tumor grafts. Thus, immunomodulation by B7-DC XAb in this tumor model was also IL-6-dependent.

We next addressed the functional relevance of modulating Tregs in another Ag-defined model. RIP-OVA mice express chicken OVA in their pancreas due to transcriptional control by the RIP. The mice treat the OVA Ag as a self-protein and develop normal pancreas morphology without evidence of inflammation (27). We first tested whether adoptive transfer of DC^XAb modulated tolerance by injecting RIP-OVA^low and RIP-OVA^high mice with OVA-pulsed DC^cntrl or DC^XAb. We also tested IL-6-dependence by using DC derived from WT or IL-6^−/− mice. Activated OVA-specific OT-I effector T cells were administered i.p. to RIP-OVA mice as a positive control (data not shown). All groups of animals were monitored periodically for their blood glucose levels.

As shown in Fig. 8 A, RIP-OVA^high mice that received Ag-pulsed DC^cntrl retained normal glucose levels, whereas mice that received Ag-pulsed DC^XAb became diabetic within 10 days (glucose levels higher than 250 mg/dL). Similar phenotypes were observed using RIP-OVA^low hosts; RIP-OVA^low mice that received OT-I T cells and OVA in CFA also became diabetic in that time frame (data not shown). Importantly, administration of IL-6-deficient DC^XAb failed to induce diabetes in any of the mice.

Blood glucose levels correlated with the integrity of the pancreas in each of the treatment groups. The number of islets was greatly reduced in the diabetic mice (Fig. 8 B). Furthermore, the immunohistochemical staining of insulin in the pancreas revealed only traces of insulin in mice that were diabetic. Mice that were nondiabetic had intact islets. We conclude from these studies that IL-6 expression by the adoptively transferred DC was required for the induction of diabetes in the RIP-OVA mice. The requirement for IL-6 is consistent with the interpretation that the Ag-pulsed DC^XAb induced disease, in part, by down regulating Treg function (36).

We also approached the adoptive transfer in the opposite way, asking whether DC^XAb-converted Tregs caused pathology in the RIP-OVA model. As shown in Fig. 8,C, when reprogrammed OT-II CD4⁺CD25⁺ Tregs were adoptively transferred into RIP-OVA mice with or without OT-I CD8 CTL precursors, hyperglycemia was evident by day 5 or 6, whereas blood glucose levels for all of the recipients of control Tregs was normal. Over the course of three experiments, 10 of 14 mice receiving reprogrammed OT-II Tregs developed diabetes, whereas none of the 10 recipients receiving Tregs treated with DC^cntrl developed the disease (Table III; p < 0.001). To ensure that the pathogenesis was not caused by the outgrowth of contaminating CD4⁺CD25⁻ effector cells, mice were treated with enriched CD4⁺CD25⁻ OT-II cells that had been activated with OVA-pulsed DC^XAb (n = 8 mice). or alternatively, OVA-pulsed DC^cntrl (n = 7). None of these animals developed diabetes, demonstrating that any small number of CD4⁺CD25⁻ OT-II cells present in the population of enriched Tregs was not likely the source of effectors induced by DC activated with B7-DC XAb and that autoimmune effectors were derived from the Treg compartment and not from naive T cell precursors in this model of diabetes.

Taken together, these studies indicate that Ag-pulsed DC^XAb causes Tregs to down-regulate FoxP3 expression, stops Treg IL-10 and TGFβ production, and inhibits Treg ability to suppress CTL responses, whereas reprogramming the Tregs to produce effector cytokines IL-17, IFN-γ, and TNF-α. These reprogrammed Tregs effectively break tolerance, attacking tumors or causing autoimmune pathology, depending on the Ags used during reprogramming.

Because treatment of mice with B7-DC XAb rendered otherwise tolerant animals responsive to Ags associated with self-proteins in both the Her-2/neu and RIP-OVA transgenic models, we assessed whether the Ab might induce generalized autoimmunity when administered systemically to animals. Male C57BL/6J mice (n = 20), housed individually, were randomized and received i.v. either 300 μg of B7-DC XAb or PBS. The animals were scored over a 2-wk-period in a blinded manner for weight. Blood, heart, brain, spleen, kidney, liver, and lungs from each of the coded animals were examined for signs of abnormality by a board certified veterinary pathologist. During the course of the 2-wk observation period, one PBS-treated mouse died. There were no significant differences in body weight, blood chemistry, and cellularity, and no changes in organ morphology (including an absence of inflammation) among the surviving animals (data not shown). Moreover, B7-DC XAb treatment in other strains of mice (e.g., BALB/cJ) did not induce autoimmunity but protected mice from tumor (our unpublished observations). Therefore, treatment with B7-DC XAb does not induce a generalized autoimmunity.

In this study, we demonstrate that Ag-pulsed DC^XAb reprogram Tregs into functional effector cells and cause a break in tolerance in a number of Ag model systems. The reprogrammed Tregs down-regulated FoxP3 expression, secreted proinflammatory cytokines (IFN-γ, IL-17, and TNF-α), and mounted potent responses against self-Ag in vivo. This phenotype conversion required DC-Treg contact, IL-6 secretion by the DC, and occurred in an Ag-specific manner. Surprisingly, this phenotypic conversion did not require cell division, though it took roughly 48 h to reach completion. Another striking finding was that reprogramming of Tregs did not cause a generalized autoimmunity, but rather, whether the model system used tumor Ags or neo-self-Ags expressed in the pancreas, a specific immune attack was elicited. This response raises the hypothesis that some pathogens can mimic the activation state of DC, which we have described using B7-DC XAb, to reprogram Tregs into autoimmune effectors.

Many factors contribute to lineage commitment of a naive T cell. Prevailing paradigms stipulate linear differentiation programs driving T cell lineage commitment, beginning with naive T cells that become Th1, Th2, Th17, or Tregs depending on the cytokine milieu the T cells encounter at the time of antigenic stimulation. In vitro studies have shown that the presence of IL-12 causes naive T cells to differentiate into Th1 cells; IL-4 drives naive T cells to become Th2 cells; TGFβ drives them to become Tregs, and the combination of IL-6 and TGFβ directs differentiation toward the Th17 lineage (37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49). In this study, we demonstrate an alternative pathway in which supposedly mature Tregs are reprogrammed. Although this pathway has been suggested by in vitro studies (50), we show that this reprogramming can take place in animal models using the endogenous Treg or DC. Together with our previous finding that committed Th2 cells can be reprogrammed to express Th1 cytokines during recall responses, identification of effector cells generated from Tregs provides new insights into the plasticity of T cell lineages.

T cell lineage specification is thought to be dependent on cell proliferation. Initial experiments reported by Bird and colleagues (51) demonstrated that expression of IL-2, IFN-γ, and IL-4 tracks with T cell division. Adoptive transfer of labeled transgenic T cells into Rag^−/− mice revealed a correlation between cell division and up-regulation of activation markers, with secretion of IFN-γ and IL-2 (52). Moreover, blockade of T cell entry into early S phase of the cell cycle resulted in abrogation of the secretion of IFN-γ and IL-10 (53). Finally, asymmetric cell division was characteristic of cells acquiring effector or memory phenotypes (54). The asymmetric fate of the daughter cells derived from first cell division following prolonged T cell interaction with APC suggests the possibility that the asymmetric nature of cell division may be mechanistically tied to lineage commitment. The possibility that T cells can acquire new traits without cell division suggested by our study will need to be addressed further. We have not yet determined the mechanism through which FoxP3 was down-regulated and whether reprogramming Tregs coincides with expression of T effector regulators such as T-bet, RORα, or RORγ (55, 56, 57). Although we found that Treg modulation was IL-6-dependent, that signal may not be the defining character of DC^XAb because TLR-activated DC also produced IL-6 (26), but did not reprogram mature Tregs in our experiments. Finally, we recovered reprogrammed cells from lymph nodes and spleens. More direct evidence of where reprogramming takes place remains to be determined.

In demonstrating the plasticity of Tregs, we have also described an approach for selectively depleting Tregs in an Ag-specific manner without promoting generalized autoimmunity. The ability to control Treg function is important for immune interventions in autoimmunity, cancer, and chronic viral infection. For example, during the course of tumor progression, changes in gene expression and the accumulation of mutations result in new antigenic profiles that shape T cell responses. Given that the appearance of new antigenic profiles in tissues is a normal process during the life of an individual, tolerance mechanisms are critical for restraining immune attack against emerging Ags. But because of the absence of overt danger signals in emerging cancers, the normal mechanisms of tolerance blunt potentially protective immune responses and interfere with efforts to use standard adjuvants to induce antitumor immunity. Strategies to eliminate Treg functions by blocking Treg effector mechanisms or by depleting Tregs from the immune repertoire have had mixed success (19), and the development of a generalized autoimmunity becomes increasing problematic.

An important aspect of this work is that the ability of T cells to be reprogrammed was revealed using DCs activated with the immune modulator, B7-DC XAb. How B7-DC XAb modulates DC function is only now emerging, but the biologic effects are significant. In short, matured DC regain their ability to take up Ag upon treatment with B7-DC XAb (58). A model now emerges in which DC^XAb causes the up-regulation of Ag presentation of proteins reaching the draining lymph nodes, and the repolarization of Ag-specific Th2 effectors and Tregs into effectors with Th1 or Th17 characteristics. At the same time, there is a rapid mobilization of Ag-specific CD8 cytolytic cells that is also Ag-specific (33). The net change that results in the loss of down-modulatory effects of the Tregs and the alignment of the polarity of Th cells with the activated CTL can result in an effective antitumor response that can clear transplanted tumors (22, 59).

The B7-DC XAb Ab also binds and activates human DC (60). Accumulating studies demonstrate that just as in the laboratory mouse, this immune modulator rapidly activates immunity against human self-proteins (E. L. Schenk and L. R. Pease, unpublished observations). Important remaining questions include how effective this strategy will be in treating spontaneous tumors more characteristic of human cancers and indeed how effective this approach will be for treating human patients. As a first step toward clinical applications, a phase I trial is now in progress to evaluate B7-DC XAb as an immunotherapeutic immune modulator in advanced melanoma.

The authors have no financial conflict of interest.