Foxp3 Expression Is Required for the Induction of Therapeutic Tissue Tolerance

Immune homeostasis and maintenance of self-tolerance depends upon constant vigilance by CD4⁺Foxp3⁺ regulatory T cells (Treg). Commitment to the Treg lineage occurs primarily in the thymus (1, 2), but also in the periphery in a TGF-β–dependent manner (3–6). One of the major goals of modern immunosuppression, be it in autoimmune disease or transplantation, is to harness tolerance mechanisms such as those used by Treg to minimize the duration and extent of drug immunosuppression. Short-term coreceptor blockade provided the first demonstration that induction of tolerance could be achieved using low-impact intervention in a mature immune system (7). Studied in a transplantation setting, this form of tolerance is totally dependent on the ability of TGF-β to signal to T cells (6) and is also associated with de novo induction of Ag specific Foxp3⁺ induced Treg (iTreg) (4, 8). This raises the possibility that the absolute need for TGF-β is simply to guarantee conversion of naive CD4⁺ T cells to stable Foxp3 expression.

However, TGF-β signaling not only induces expression of Foxp3, but also many other effector molecules, including CD39, CD73, CTLA4, CD103, neuropilin, perforin, and IL-10 (9–12). There are also claims that TGF-β is needed for the effector arm of suppression (13), and if so, this, too, could explain the need for TGF-β signaling to T cells.

Despite a large literature on Foxp3⁺ Treg in self-tolerance, it is not known whether Foxp3 expression is essential for dominant tolerance induced to foreign Ags or whether other genetic/expression modalities (e.g., those potentially necessary for Th3, Tr1, and iTr35 cells) can operate in its absence. Using a combination of genetically manipulated mouse strains unable to express Foxp3, retroviral constructs facilitating conditional nuclear localization of Foxp3, and a dominant-negative TGF-βRII to ablate TGF-β signaling in T cells, we have addressed the contributions of TGF-β and Foxp3 to the induction and function of iTreg. We show that Foxp3 expression is indispensible for tolerance induction to transplanted tissue, and its continued nuclear expression is necessary for maintenance of tolerance. In contrast, once tolerance is established, prevention of tissue damage does not depend on TGF-β signaling to naive T cells. This indicates that the major role of TGF-β in this form of acquired tolerance largely resides in the induction of Foxp3 expression in naive CD4⁺ T cells. Comparative transcriptome analysis of in vitro-generated, TGF-β–experienced Foxp3⁺ and Foxp3⁻ CD4⁺ T cells with TGF-β–experienced Foxp3^−/− CD4⁺ T cells not only confirms TGF-β signaling in all the populations, but also demonstrates that, although crucial for maintenance of tolerance, the proportion of the transcriptome controlled specifically by Foxp3 in iTreg appears to be relatively limited.

CBA/Ca (CBA), C57BL/6J (B6), CBA.Rag1^−/−, B6.Rag1^−/−, Marilyn.Rag1^−/− (Marilyn) [derived from Marilyn.Rag2^−/− (14)], Marilyn.Foxp3^hCD2 (8), Marilyn.Foxp3^−/− (9), CD4-dnTGFβRII (15), CD4-dnTGFβRII.Rag1^−/−, and Marilyn.CD4-dnTGFβRII.Rag1^−/− mice were bred and maintained under specific pathogen-free conditions at the Sir William Dunn School of Pathology (University of Oxford, Oxford, U.K.). CD4-dnTGFβRII mice backcrossed to B6 for >10 generations were crossed with B6.Rag1^−/− mice for two generations to generate CD4-dnTGFβRII.Rag1^−/− animals, which were further crossed with Marilyn mice to introduce the Dby-H2A^b specific TCR-transgenic element. All experimental procedures received local ethics committee approval and were conducted in accordance with the Home Office Animals (Scientific Procedures) Act of 1986.

Skin grafting was performed as described previously (16). Maintenance of nuclear Foxp3 in cFoxp3-transduced T cells adoptively transferred into mice was enforced by daily i.p. injections of tamoxifen in sunflower oil (1 mg daily).

Anti-CD4 (L3T4), anti-CD25 (7D4), and anti-CTLA4 (BN13) were purchased from BD Biosciences. Anti-Foxp3 (FJK-16s), anti-CD39 (24DMS1), and anti-CD73 (TY/11.8) were purchased from eBioscience. Anti-GITR (YGITR-765) was purchased from Serotec.

Lyophilized recombinant human TGF-β1 was purchased from R&D Systems and reconstituted to 10 μg/ml in 4 mM HCl containing 1 mg/ml BSA. Lyophilized peptide HYA^b (NAGFNSNRANSSRSS) (14) was reconstituted in PBS and used at a maximum working concentration of 100 nM.

Bone marrow-derived dendritic cells (BMDCs) were prepared as described previously (17). Murine GM-CSF–enriched supernatant was harvested from the transfected cell line X63 (provided by D. Gray, Edinburgh, Scotland) and used at an equivalent of 5 ng/ml.

Splenic CBA or B6 CD4⁺ T cells were isolated by negative selection using an AutoMACS CD4⁺ T cell isolation kit and AutoMACS separator (Miltenyi Biotec).

RBC-depleted splenocytes from Marilyn, Marilyn.Foxp3^hCD2, and Marilyn.Foxp3^−/− mice were cultured at a 5:1 ratio with mitomycin-treated B6 female BMDCs in RPMI 1640 containing 100 nm HYA^b peptide for 7 d. For TGF-β–conditioned iTreg cultures, recombinant human TGF-β was also added at 2 ng/ml. Foxp3 expression was typically 50–90% for TGF-β–conditioned cultures (i.e., 50–90% of TGF-β–conditioned cultures were Foxp3⁺ iTregs). For microarray analysis, MoFlo sorting (DakoCytomation, Glostrup, Denmark) was employed to sort Foxp3⁺ and negative cells from the cultures based on human CD2 (hCD2) expression.

Cells were washed twice in PBS containing 0.5% w/v BSA and incubated for 10 min at room temperature in PBS with 10 μg/ml anti-FcR block (2.4G2) and then with primary Ab for 30 min in the dark at 4°C. Following washing and fixation with 2% paraformaldehyde, analysis was performed on an FACSCalibur (BD Biosciences) with dual laser excitation (488 and 633 nm). Data acquisition was performed with interchannel compensation using CellQuest version 3.1 software (BD Biosciences) and data analysis performed with FlowJo 7.2.4 software (Tree Star). Foxp3 staining was performed using a commercial kit (eBioscience) according to the manufacturer’s directions.

Flow cytometry sorting was performed using a MoFlo sorter (DakoCytomation).

All vectors used except pCL-ECO have previously been described (18). The conditional Foxp3 construct cFoxp3 was made by cloning sequence-encoding ERT2, a modified estrogen-binding domain, in-frame with the C terminus of Foxp3. GFP coding sequence was cloned directly after the fifth codon of Foxp3 to produce the fusion protein GFP-Foxp-ERT2. This was then cloned into a Moloney murine leukemia virus backbone, a circular plasmid that also contained a GPI-linked and internal ribosome entry site-driven rat CD8α (rCD8) gene. The empty control vector contained only the GFP and rCD8 segments, but no Foxp3. The packaging vector, pCL-ECO, was purchased from Addgene (Cambridge, MA). The day before transfection, a confluent plate of HEK 293eT cells were split at one in three and plated in six-well plates at ∼40% confluency, with 2 ml/well. Approximately 16 h later, the cells were transfected with 2 μg pCL-ECO packaging plasmid and 2 μg retroviral vector cFoxp3 using the CaCl₂ method. Six hours after transfection, the medium in the well was replaced with fresh medium. A further 24 h later, the viral supernatant was harvested and filtered through a 0.45-μm filter. Protamine sulfate (600 μg/ml) was added to the viral supernatant.

AutoMACS-sorted CD4⁺ cells were cultured at 2 × 10⁵/well in a 96-well plate precoated with anti-CD3 (KT3; 4 μg/ml in PBS, 1 h at 37°C). The cells were activated for 24 h and then resuspended in a 1:2 mixture of viral supernatant and complete medium (IMDM containing 10% v/v FCS and 10 μM 2-ME). This mix was supplemented with 5 ng/ml recombinant murine IL-2 (PeproTech) and 5 ng/ml recombinant murine IL-7 (PeproTech). This was followed by centrifugation at 500 × g for 2 h at 32°C. Cells were analyzed 40 h later. Typical transduction efficiencies were 30–50%. Following transduction, in most experiments, the cells were MoFlo sorted (DakoCytomation) to achieve 99.9% GFP-positive cells, as indicated in figure legends. For induction of nuclear localization of Foxp3, transduced cells were cultured in the presence of 50 nM of 4′hydroxytamoxifen (4HT) for at least 24 h.

Confocal microscopy was performed on a Zeiss LSM 510 META laser scanning microscope (Carl Zeiss MicroImaging, Göttingen, Germany) using a ×20 objective.

A 23-component multiplex bead-based assay specific for eotaxin, G-CSF, GM-CSF, IFN-γ, IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12p40, IL-12p70, IL-13, IL-17, eotaxin, keratinocyte chemoattractant, MCP-1, MIP-1α, MIP-1β, RANTES, and TNF-α (Bio-Rad) was used according to the manufacturer’s recommended protocol and samples analyzed using a Bio-Plex System and Bio-Plex Manager Software 4.0 (Bio-Rad).

Effector cells were Marilyn CD4⁺ cells enriched by negative selection using AutoMACS. Suppressor cells were Marilyn CD4⁺ cells that had been transduced with retroviral vectors for 40 h and then sorted based on CD4 and GFP expression. BMDCs were prepared from female B6 mice. Suppressor cells and BMDCs were incubated with mitomycin C (25 μg/ml) for 30 min at 37°C and washed three times before being added to the assay. Each well of a 96-well plate contained 1 × 10⁴ BMDCs, 2 × 10⁴ suppressors, and 2.5 × 10⁴ effectors. HYA^b peptide was added as a dilution series (100, 10, 1, and 0 nM). Cultures were incubated for 72 h. To measure proliferation, [³H]thymidine (GE Healthcare) was added (0.5 μCi/well) for the final 18 h of culture. [³H]thymidine incorporation was then measured by scintillation counting.

Cultured splenocytes were used to generate highly purified populations of Marilyn.Foxp3^hCD2-activated (HY) and TGF-β–induced (HYT) cells sorted as CD4⁺hCD2⁺ and CD4⁺hCD2⁻ and Marilyn.Foxp3^−/− HY and HYT cells sorted as CD4⁺. Three to five independent biological replicates of the five sorted cell populations above were processed (RNA isolation, labeling, and hybridization to mouse exon 1.0 ST Affymetrix arrays; Affymetrix) by Asuragen (Austin, TX). Total RNA was isolated from frozen cell pellets by Asuragen according to the company’s standard operating procedure for automated isolation on a KingFisher magnetic particle separator (Thermo Scientific). The procedure included a DNAse treatment step and cleanup prior to elution from the magnetic beads. The purity and quantity of total RNA samples were determined by absorbance readings at 260 and 280 nm using a NanoDrop ND-1000 UV spectrophotometer (NanoDrop). The integrity of total RNA was qualified by Agilent Bioanalyzer 2100 microfluidic electrophoresis using the Nano Assay (Agilent Technologies).

Samples for mRNA profiling studies were processed by Asuragen according to the company’s standard operating procedures. Biotin-labeled sense-strand cDNA was prepared from 30 μg cRNA generated from either 500 ng or 1 μg total RNA per sample using a modified Affymetrix GeneChip Whole Transcript Sense Target Labeling Assay (Affymetrix). Intermediate cRNA and resulting cDNA yields were quantified by spectrophotometry. Fragmentation and labeling of cDNA was performed using 5 μg for exon arrays. Hybridization to arrays was carried out at 45°C for 16 h in an Affymetrix Model 640 hybridization oven (Affymetrix). Arrays were washed and stained on an Affymetrix FS450 Fluidics station (Affymetrix). The arrays were scanned on an Affymetrix GeneChip Scanner 3000 7G (Affymetrix). For every array scanned, .DAT, .CEL, .jpg, and .xml flat files were provided. Data generated from Asuragen for the arrays were further preprocessed in the Affymetrix Expression Console software version 1.2 (Affymetrix), which was used to normalize .CEL files using the quantile normalization method, and the data were summarized at both the exon and gene level by using the robust multichip analysis algorithm. All further data analysis was performed in JMP Genomics software version 5.1 (JMP) using the gene level log₂ data. Low-value intensities were filtered out. Differential gene expression analysis was performed on transcript values based on core probe sets. The microarray data discussed in this publication have been deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus and are accessible through Gene Expression Omnibus Series accession number GSE39529 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE39529).

Statistical significance was determined using one-way ANOVA tests with Dunnett’s multiple comparison posttest or Student t test (unpaired, two-tailed) using GraphPad Prism software (http://www.graphpad.com; GraphPad). In the figures, *p < 0.05, **p < 0.01, and ***p < 0.001.

It is well established that T cells expressing Foxp3 are critical for maintenance of tolerance to self-Ags. The requirement for Foxp3 in the induction of peripheral tolerance to foreign Ags has not, however, been unequivocally demonstrated. It is possible that other cell types such as Th3, Tr1, or iTr35 might emerge and operate in its absence. Transplantation tolerance can be induced in TCR-transgenic mice lacking natural Treg (nTreg). This is associated with the development of Foxp3⁺ iTreg in the periphery. It cannot be induced, however, if TGF-β signaling to T cells is prevented (6). We asked if alternative routes to tolerance might engage in a situation in which Foxp3 induction was prevented by genetic ablation, but in which TGF-β signaling pathways were left intact. To address this in circumstances in which autoimmune pathology could be avoided, we generated Foxp3^−/− Marilyn TCR-transgenic mice. All of the T cells in normal Marilyn mice are monospecific. They recognize a specific peptide of the male Ag Dby (HYA^b) presented in the context of H2-A^b, and naive T cells do not express Foxp3. Female Marilyn and Marilyn.Foxp3^−/− mice were each grafted with male B6.Rag^−/− skin under the influence of a short regimen of anti-CD4–blocking Ab previously shown to be sufficient to generate tolerance and the induction of peripheral Foxp3⁺ iTreg in Marilyn mice (8). As expected, the majority of the Foxp3-sufficient mice accepted their male grafts indefinitely, whereas all of the Marilyn.Foxp3^−/− mice had rejected their grafts by 60 d (Fig. 1A). Thus, the tolerance induced by short-term coreceptor blockade requires a functional Foxp3 gene, despite an intact TGF-β signaling axis.

TGF-β induces potent anti-inflammatory adenosine production by CD4⁺ T cells via cell-surface–expressed CD39 and CD73 ectonucleotidase activity (9). Regulation of these enzymes is independent of Foxp3 expression, and it has been suggested may confer suppressive activity in the absence of Foxp3 (19). We asked whether TGF-β–conditioned T cells might still be able to suppress rejection of grafts if they could not express Foxp3 and also whether TGF-β was required as a mechanism of suppression. We generated sets of TGF-β–conditioned T cells in vitro by stimulating Marilyn or Marilyn.Foxp3^−/− splenocytes with BMDCs and cognate peptide in the absence (HY) or presence (HYT) of TGF-β. These cultures typically yielded ∼70% Foxp3⁺ cells in the Foxp3-sufficient cells after 7 d (Fig. 1B). CD4⁺ T cells from both wild-type and Foxp3^−/− animals upregulated CD73 and CD103, indicating effective TGF-β signaling, consistent with their expression of equivalent surface levels of TGF-βRII (Supplemental Fig. 1A). Previous data (Supplemental Fig. 1B) indicated that as few as 2500 Marilyn T cells were sufficient to reject male grafts on adoptive transfer into Rag^−/− recipients. One million Marilyn or Marilyn.Foxp3^−/− HY or HYT splenocytes were injected into female B6.Rag^−/− recipients, followed by grafting with male B6.Rag^−/− skin. All seven mice that received TGF-β–experienced Foxp3-sufficient splenocytes accepted their grafts, despite cotransfer of >30 times more Foxp3⁻ cells than should be necessary for rejection (Fig. 1C). Mice that received either Ag-activated or Ag-activated and TGF-β–experienced splenocytes from Foxp3^−/− animals rejected their grafts within the same rapid time frame as that of wild-type Ag-activated but non–TGF-β–experienced (Foxp3-negative) splenocytes (Fig. 1C). Even as few as 4 × 10⁴ HYT Marilyn.Foxp3^−/− cells were still able to reject male skin grafts at the same rate as 1 × 10⁶ cells (Fig. 1D). In short, TGF-β conditioning in the absence of Foxp3 expression has little impact on the ability of the T cells to reject grafts, showing the critical role of Foxp3 in extinguishing that function and thus ruling out a major role for alternative TGF-β–induced pathways [such as adenosine generation (9)] in iTreg-induced long-term graft acceptance in this setting.

The true role for TGF-β in Treg-mediated suppression is controversial, with studies arguing for and against its requirement. We have previously shown that TGF-β signaling to T cells is essential for the development of dominant tolerance (6). To establish whether in this case TGF-β was operating in the inductive phase of tolerance or at a later stage to sustain dominant tolerance, female Marilyn mice were made tolerant to male B6.Rag^−/− skin using anti-CD4 Ab treatment. On regrafting at 95–120 d after the first graft, the tolerized mice also received 1 × 10⁶ rejection-competent CD4⁺ T cells from either female Marilyn or female Marilyn.dnTGFβRII mice. In both cases, the mice were able to resist rejection by the transferred cells (Fig. 1E), demonstrating that TGF-β signaling to T cells was not essential for suppression of rejection once regulatory T cells had been induced.

We then asked what had become of CD4⁺ T cells that had not converted to Foxp3 expression by comparing their pattern of cytokine secretion to that of their Foxp3⁺ counterparts. Marilyn.Foxp3^hCD2 CD4⁺ T cells were activated for 7 d with dendritic cells and HY peptide in the presence of TGF-β. Cells sorted into hCD2-positive and -negative populations were then stimulated for 2 d with anti-CD3/anti-CD28–conjugated beads. Culture supernatants were assayed for cytokine content. T cells that had not transcribed Foxp3 in response to TGF-β exposure produced 5–20-fold increased amounts of ILs 2, 3, 4, 5, 6, 13, and 17 and IFN-γ (Fig. 2). Thus, in the absence of Foxp3, TGF-β–treated cells are clearly capable of producing proinflammatory cytokines commensurate with their capacity to reject grafts.

As TGF-β–induced Foxp3 expression is crucial to the continuing function of iTreg in vivo and so closely linked to the biology of induced tolerance, we asked to what extent the gene transcription program induced by TGF-β was related to Foxp3 expression. To answer this question, we performed microarray analysis comparing transcriptomes of CD4⁺ T cells from Marilyn.Foxp3^hCD2 or Marilyn.Foxp3^−/− mice treated (HYT) or untreated (HY) with TGF-β in vitro (see 2Materials and Methods). In each case, CD4⁺ cells were MoFlo sorted (DakoCytomation) and, in the case of the Marilyn.Foxp3^hCD2-derived cells, sorted into Foxp3 expressors and nonexpressors on the basis of surface expression of hCD2.

We asked whether the signature of transcripts regulated by TGF-β exposure in CD4⁺ T cells was dependent on Foxp3 expression. A global suppression effect was observed upon TGF-β treatment in cells of Marilyn.Foxp3^hCD2 mice, identified by comparative analysis between TGF-β–treated Marilyn.Foxp3^hCD2 cells (including both HYT hCD2⁺ and HYT hCD2⁻ populations) and untreated HY cells (Fig. 3A). A total of 3046 transcripts were found differentially expressed at p < 0.01 (Supplemental Table I). A similar analysis performed on Marilyn.Foxp3^−/− cells indicated 469 transcripts differentially expressed at p < 0.01 (Fig. 3B, Supplemental Table II).

To examine if the presence of Foxp3 wild-type transcripts could generate additional response following TGF-β treatment, the expression profile of the TGF-β signature in Marilyn.Foxp3^hCD2 mice was examined in Marilyn.Foxp3^−/− mice (Fig. 3C). The high correlation (R² = 0.61) suggested that there was not a significant additional detectable effect, positive or negative, from the genomic copies of Foxp3. Thus, the TGF-β signature seems to be independent of Foxp3. These data also show that loss of Foxp3 was not inhibiting TGF-β responsiveness of these cells.

Samples of the TGF-β–treated and -untreated, sorted hCD2⁺ and hCD2⁻ groups exhibited the expected relative level of Foxp3 transcript expression (Fig. 4A). To identify a TGF-β–induced Foxp3 effect between the two sorted, TGF-β–influenced groups (HYT hCD2⁺ and HYT hCD2⁻), a comparative analysis was first carried out to identify transcripts differentially expressed between the two populations. A total of 183 transcripts (p < 0.01) were found differentially expressed (Fig. 4B, 4C, Supplemental Table III). Among them, a subset of 34 transcripts was then identified as uniquely associated with the TGF-β–induced hCD2⁺ cells (Fig. 4C, indicated in blue, orange, and turquoise). Finally, upon further eliminating general TGF-β effects, 27 transcripts were considered to comprise a unique TGF-β–induced-Foxp3 signature (Fig. 5A, 5B).

Our demonstration that Foxp3 was indispensible for tolerance induced by TGF-β led us to ask whether the need for Foxp3 expression in tolerance was transient or sustained. Many groups have demonstrated that ectopic expression of Foxp3 leads to suppressive function, and in vitro cre-lox deletion of Foxp3 has been shown to reverse the suppressive properties of Treg. However, it is not known whether iTreg need to express Foxp3 for prolonged periods in order for tolerance to be maintained nor what effect turning off Foxp3 in vivo following tolerance induction would have. For example, Foxp3-influenced T cells may, over time, acquire suppressive features that are Foxp3 independent. We therefore asked two questions: namely, is prolonged Foxp3 expression needed for iTreg to remain rejection incompetent, and is continuous Foxp3 expression required for iTreg to retain their suppressive ability? To address these questions, we used a retroviral construct, cFoxp3, encoding a fusion protein GFP-Foxp-ERT2, for which entry to the nucleus, and therefore Foxp3 activity, requires addition of tamoxifen (18) (see 2Materials and Methods). We first established the ability of the construct to translocate to the nucleus following tamoxifen treatment and to induce attributes consistent with a Treg phenotype, namely anergy, upregulation of Treg cell-surface markers, and suppression of proinflammatory cytokine production (Fig. 6). We showed that after 64 h, GFP-Foxp3 fusion protein had clearly translocated into the nuclei of >95% of transduced AutoMACS-sorted CD4⁺ splenocytes, FACS-sorted for GFP and treated with 4HT, but remained predominantly cytoplasmic in transduced but noninduced cells (Fig. 6A, 6B). Tamoxifen exposure rendered the T cells anergic (Fig. 6C) and was associated with upregulation of the Treg-associated surface markers CTLA4, CD25, and GITR (Fig. 6D). Foxp3 negatively regulates transcription of many cytokines (20, 21), and we found that 4HT-treated transduced T cells also secreted minimal quantities of proinflammatory cytokines, such as IL-1, IL-2, IFN-γ, IL-4, IL-9, and IL-17, when compared with cells transduced with empty vector or with cFoxp3, but without 4HT treatment (Fig. 7). We term such transduced cells induced to exhibit a regulatory phenotype by using 4HT as conditional Treg (cTreg).

We next tested whether, in the absence of exogenous addition of TGF-β, tamoxifen exposure would suffice to render T cells unable to reject grafts. Naive Marilyn CD4⁺ T cells, which do not express Foxp3, were transduced with cFoxp3; half were treated and half untreated with 4HT for 2 d in vitro to induce nuclear localization of the GFP-Foxp-ERT2 fusion protein. CD4⁺GFP⁺ cells (Fig. 8A) were purified by FACS and 2 × 10⁴ or 1 × 10⁶ cells transferred into female B6.Rag^−/− mice. These mice then each received a male B6.Rag^−/− skin graft; half received daily tamoxifen injections, and the other half daily injections of vehicle for 28 d posttransplant. All of the mice that received 4HT-facilitated cTreg but without further tamoxifen injections rejected their grafts within 20 d (Fig. 8B). Mice that received either 2 × 10⁴ or 1 × 10⁶ transduced cells treated in vitro with 4HT and with further daily injections of tamoxifen accepted their grafts up to and following cessation of the tamoxifen injections, but rejected their grafts 14–17 d following the last tamoxifen injection. Once the tamoxifen injections ended the speed of rejection was similar to that experienced on transfer of cells with no tamoxifen. Thus, persistent expression of nuclear Foxp3 was associated in vivo with T cells maintaining a rejection-impotent phenotype.

In order to establish whether sustained expression of Foxp3 in cTreg was needed for their suppressive function, immune-competent naive female Marilyn mice, which lack endogenous Treg, were adoptively transferred with 5 × 10⁵ transduced Marilyn, 4HT-treated cTreg (Fig. 8C, 8D). The mice were then grafted with a male CBA.Rag^−/− skin graft and injected daily with tamoxifen for 28 d. All Marilyn mice without cTreg cells rejected their grafts within 11 d, as expected. Mice with grafts together with cTreg tolerated their grafts for 40 d (12 d beyond the last injection of tamoxifen). Thus, cTreg can suppress rejection of an allogeneic graft, but this requires sustained exposure to tamoxifen.

We had set out to test the extent to which induction of Foxp3 expression explained the need for TGF-β signaling to T cells in eliciting tolerance to transplanted tissues (6). We made use of TCR-transgenic mice on a Rag^−/− background. Untreated, these mice have no Foxp3⁺CD4⁺ T cells, so that any Foxp3⁺ cells emerging from treatment are de facto iTreg cells. Furthermore, the use of TCR-transgenic Foxp3^−/− mice sidesteps any potentially confounding issues of autoimmune pathology. We used Foxp3-reporter mice on the Marilyn background, Marilyn.Foxp3^−/− mice, and microarray analysis to demonstrate that although the number of genes affected by TGF-β exposure is large, the set of genes in iTreg for which expression is correlated specifically with Foxp3 expression is surprisingly low. Finally, ectopic inducible nuclear Foxp3 expression was used to demonstrate that continuous expression of Foxp3 on a per-cell basis was required for maintenance of induced tolerance.

We present five novel findings. First, the potential to express Foxp3 is essential for this form of induced peripheral tolerance. Second, a TGF-β experience in the absence of Foxp3 is insufficient for tolerance despite inducing a significant TGF-β gene signature. Third, TGF-β signaling to T cells, although essential for induction of tolerance, is not essential for suppression by iTreg. Fourth, sustained functional expression of Foxp3 in vivo is necessary to ensure prolonged graft survival. Fifth, iTreg have independent TGF-β and Foxp3 gene signatures, the latter consisting of a relatively small number of genes.

The literature describes many different CD4⁺ T cell subsets exhibiting regulatory properties. These include TGF-β–secreting Th3 cells (22, 23), IL-10–secreting Tr1 cells (24), and IL-35–secreting iTr35 cells (25). We have previously shown that maintenance of tolerated allogeneic skin grafts required constant suppression by Foxp3⁺ cells (8). In this study, we show for the first time, to our knowledge, that Foxp3 expression is essential for the induction of tolerance. Any other potentially suppressive mechanisms such as Tr1 cells did not emerge as effective substitutes for Foxp3⁺ T cells.

In Marilyn mice, tolerance by coreceptor blockade relies on the TGF-β–dependent de novo differentiation of iTreg (6). We propose, therefore, that failure to induce tolerance in the Marilyn.Foxp3^−/− mice was due to failure to generate Foxp3-expressing iTreg cells. Consistent with that, Ag-activated Marilyn.Foxp3^−/− CD4⁺ T cells cultured in the presence of TGF-β (9) remained responsive to Ag in vitro and could still reject grafts in vivo, suggesting that any other TGF-β–induced mechanisms were insufficient to both calm the cells and to convert them to potent regulatory function.

It is still unclear whether iTreg play a major role in the control of immune homeostasis. Numerous studies have shown that Foxp3⁺ iTreg can arise in a number of experimental settings. These include exposure of T cells to cognate Ag in the absence of microbial or endogenous danger signals as in the case of mucosally (26), as well as parenterally delivered Ags (27). It has, however, been difficult to separate the need for nTreg away from that for iTreg in experimental models. Whatever emerges as the normal physiological role for iTreg, it is clear that they play a crucial role in therapeutic tolerance in TCR-transgenic mice, as demonstrated in this study. This coupled with the absolute need for TGF-β signaling to T cells for transplantation tolerance in conventional mice suggests that iTreg are also very relevant in that context.

Foxp3 has been shown to have binding regions on ∼700 genes in nTreg, as well as forming a supramolecular complex with other transcription factors and chromatin modifiers (21). Persistent Foxp3 expression is likely needed to maintain the transcriptional changes and epigenetic alterations of target genes that it controls. We performed extensive transcript analysis using exon tiling microarrays and demonstrated that T cells have ∼3000 transcripts significantly altered in expression by TGF-β. However, subtractive comparisons of the transcriptome of TGF-β–experienced T cells that did not convert to Foxp3 expression (HYT CD4⁺hCD2⁻ cells from Marilyn.Foxp3^hCD2) or TGF-β-experienced Foxp3^−/− CD4⁺ T cells showed a very limited signature of 27 genes influenced solely by Foxp3 in iTreg. Thus, although Foxp3 can bind hundreds of genes in thymically derived Treg, resulting in up and down gene regulation, in our populations of iTreg, the number of genes significantly altered in expression by Foxp3 was much more limited. Foxp3 expression is controlled in part by variable DNA methylation at the Treg-specific demethylated region (28, 29). The transcriptional and epigenetic modifications initiated by Foxp3 may not be sufficient to render the cell regulatory in the absence of continued Foxp3 expression. Cre-lox deletion of the Foxp3 gene in mature Treg was found to result in reversal of their regulatory characteristics and acquisition of ability to produce proinflammatory cytokines (30).

We have shown that sustained nuclear Foxp3 expression is required to maintain a cellular program compatible with regulation. For a regulatory cell to suppress, one desirable function would be that it limits its own production of proinflammatory cytokines. We observed such behavior for IL-1, IL-2, IL-4, IL-9, IL-17, and IFN-γ, all of which were reduced to low levels following tamoxifen-induced nuclear localization of GFP-Foxp3-ERT2 in cFoxp3-transduced Marilyn CD4⁺ T cells. The requirement for sustained suppression to maintain tolerance has already been argued where tolerance was lost when Foxp3-expressing cells were deleted (8, 31). We have shown that simple withdrawal of tamoxifen in this system is sufficient to revert cTreg to an effector phenotype. Tamoxifen in an equivalent setting in human T cells has been shown to induce nuclear localization as well as increase the stability of the fusion protein (32). Withdrawal of tamoxifen resulted in reversal of the inhibition of IFN-γ production (32). Thus, lowering the expression and changing the subcellular localization of Foxp3 from nucleus to cytoplasm is a potent switch to mediate loss of suppressive activity.

Two recent reports have shown that N-terminal addition of GFP to Foxp3 can alter the interaction of Foxp3 with HIF-1α, Eos, HDAC7, and TIP60, whereas interaction with NFAT is unaltered (33, 34). This led to a decrease in proteasomal degradation of Foxp3 and subtle alterations of the genes inhibited by Foxp3. The result of these alterations was increased suppression of Th17 responses, resulting in increased susceptibility to diabetes on the NOD background, consistent with previous reports showing Th17 cells to be protective in this model. We have shown that Th17 cells mediate damage in allogeneic skin graft rejection (35), thus the potency of our conditional Treg, which have an N-terminal GFP tagged Foxp3 may be explained in part by this phenomenon.

These studies reinforce and extend our earlier findings that Foxp3⁺ cells are critical for clinically relevant modes of tolerance induction. The role of TGF-β can be adequately accounted for by the induction of Foxp3 gene expression and a narrow set of Foxp3 influenced genes and no other feature of TGF-β signaling. The implications of these findings are that clinical efforts to harness induced Foxp3⁺ cells for tolerance will need to ensure the stable expression of Foxp3 as its role in Treg is not a transient one, but one requiring persistent expression.

We thank Dr. Nigel Rust for expert FACS sorting, the Pathology Support Building staff for excellent animal husbandry, and Dr. Amin Moghaddam for expert assistance with cytokine assays.

The authors have no financial conflicts of interest.