The Plasticity of Regulatory T Cell Function

Regulatory T cells (T_regs) play a key role in maintaining immune tolerance, preventing autoimmune diseases, and limiting inflammatory conditions (1–3). A unique and important feature of T_regs is the brevity and flexibility of their regulatory capacity. T_regs can suppress an array of different cell types [including CD4⁺ T cells (Th1/Th2/Th17) (4), CD8⁺ T cells (5), B cells (4, 6), dendritic cells (7), and osteoclasts (8)] in a variety of inflammatory conditions and in distinct tissue locations. The ability of T_regs to suppress a broad range of targets in a variety of scenarios can be attributed to the numerous mechanisms employed by T_regs to mediate their function (2, 9). However, it is not clear whether all these mechanisms are equally important or whether they have nonredundant roles under different inflammatory settings. Indeed, it was recently reported that T_regs may have specialized mechanisms for controlling specific cell types, as T_regs appear to require IFN regulatory factor-4, T-bet, and STAT3 to suppress Th2, Th1, and Th17 cells, respectively (10–12). However, it is unclear which T_reg mechanisms are used under specific conditions, how many mechanisms are required for maximal T_reg function, and whether there is any crosstalk between the various regulatory mechanisms used. Although it is well established that Foxp3 is a key transcription factor critical for the stability of T_regs (13), whether there is stability or plasticity in the regulatory mechanisms used by T_regs is unclear.

T_regs use multiple mechanisms to mediate their function, with the immunosuppressive cytokines TGF-β, IL-10, and IL-35 contributing significantly (14–16). IL-10 is important for T_reg function in vitro and in vivo, especially in the gut (17, 18). IL-35 is a recently discovered heterodimeric cytokine composed of Ebi3 (also part of IL-27) and Il12a/p35 (also part of IL-12) that is uniquely expressed by T_regs, but not by T conventional (T_conv) cells, and is required for maximal T_reg function (16). Whereas the loss of either IL-10 or IL-35 significantly reduces T_reg function, they do not become completely dysfunctional and deficient mice do not exhibit the lethal multiorgan inflammatory disease seen in Scurfy or Foxp3^−/− mice that lack T_regs (19, 20). Thus, in the current study, we speculated that T_regs that lacked both IL-10 and IL-35 might exhibit a more profound functional defect and that this approach could be used to assess the relative contributions of different suppressive mechanisms. Alternatively, given the importance of T_regs in the maintenance of immune homeostasis, as yet unknown compensatory mechanisms may be triggered that attempt to restore immune balance. These possibilities were tested in this study.

Ebi3^−/− mice (C57BL/6; now 100% C57BL/6 by microsatellite analysis performed by Charles River) were provided by R. Blumberg and T. Kuo (Brigham and Women's Hospital, Boston, MA); Il10^−/− mice were provided by T. Geiger (St. Jude Children's Research Hospital, Memphis, TN); Tnfsf10^−/− mice were provided by D. Green (St. Jude Children's Research Hospital, Memphis, TN); and Foxp3^−/− were provided by J. Ihle (St. Jude Children's Research Hospital, Memphis, TN) with permission from A. Rudensky (Sloan Kettering Institute, New York, NY). TGFβRII.DN, Il12a^−/−, Rag1^−/−, C57BL/6, BALB/c, and B6.PL mice were purchased from The Jackson Laboratory. All animal experiments were performed in American Association for the Accreditation of Laboratory Animal Care-accredited, Helicobacter-free, murine norovirus-free, specific pathogen-free facilities in the St. Jude Animal Resource Center following national, state, and institutional guidelines. Animal protocols were approved by the St. Jude Animal Care and Use Committee. Spleens and lymph nodes from Tnfrsf10b^−/− (death receptor 5 [DR5]^−/−) mice were provided by T. Ferguson (Washington University, St. Louis, MO), with approval of the Washington University Animal Care and Use Committee.

T_conv (CD4⁺CD25⁻CD45RB^high) cells and T_regs (CD4⁺CD25⁺CD45RB^low) from spleens and lymph nodes of either knockout or wild-type C57BL/6 mice were positively sorted by FACS following staining with fluorochrome-conjugated Abs: anti-mouse CD4, anti-mouse CD45RB, and anti-mouse CD25 (BioLegend, San Diego, CA). The cells were sorted on a Reflection (i-Cyt, Champaign, IL) or on a MoFlo (DakoCytomation, Fort Collins, CO). For flow cytometric analysis, purified T_regs were cultured as described and stained with anti-TRAIL PE Ab (eBioscience, San Diego, CA) at the indicated time points after culture in presence of rIL-2 (1000 IU/ml). Cathepsin E (CTSE) intracellular staining was performed, as previously described (21). Briefly, freshly isolated T_regs were fixed with formaldehyde and permeabilized with Triton X-100 for 2 min prior to staining with anti-CTSE Ab (R&D Systems). Cells were analyzed on a FACSCalibur (Beckman Coulter, Brea, CA), and data were analyzed using FlowJo.

Purified T_regs from wild-type C57BL/6 mice or knockout mice were activated in the presence of anti-CD3– and anti-CD28–coated beads or cultured in combination with T_conv cells at 4:1 (T_conv:T_regs) ratio for 48 h. Where indicated for TRAIL expression, cells were activated in presence of IL-2 (1000 IU/ml), and cells were collected from RNA isolation at indicated time points. Cells were resorted based on their congenic markers, where indicated. RNA was isolated using Qiagen micro or mini RNA kit following manufacturer's instructions. RNA was quantified using a nanodrop spectrophotometer, and equal amount of total RNA in each sample was reverse transcribed with the high-capacity cDNA reverse-transcription kit (Applied Biosystems) following the manufacturer's guidelines. TaqMan primers and probes were designed with Primer Express software and were synthesized by the St. Jude Hartwell Center for Biotechnology and Bioinformatics. The primers for CTSE were CTSE forward, 5′-CAACCTCTGGGTCCCTTC-3′; CTSE reverse, 5′-TGATTCCCTACCTCCGTG-3′; and CTSE probe, 5′-CATGCAAGGCACACCCAG-3′. The primers for TRAIL-R2 (DR5) were TRAIL-R2 (DR5) forward, 5′-TGCTGCTCAAGTGGCGC-3′; DR5 reverse, 5′-GGCATCCAGCAGATGGTTG-3′; TRAIL forward, 5′-CCTCTCGGAAAGGGCATTC-3′; TRAIL reverse, 5′-TCCTGCTCGATGACCAGCT-3′; actin forward, 5′-ACCCACACTGTGCCCATCTAC-3′; actin reverse, 5′-AGCCAAGTCCAGACGCAGG-3′; and actin probe, 5′-AGGGCTATGCTCTCCCTCACGCCA-3′. Equal volume of cDNA samples was used in a quantitative real time PCR (qPCR) reaction with primer probes and amplified for 40 cycles using an Applied Biosystems Prism 7900 Sequence Detection System instrument, according to the manufacturer's protocol. Relative quantification of mRNA expression was carried out using the comparative critical threshold method, as described in the Applied Biosystems User Bulletin number 2 (http://docs.appliedbiosystems.com/pebiodocs/04303859.pdf). In this study, the amount of target mRNA is normalized to the endogenous β-actin expression and is calculated by the equation 2^ΔΔCT.

Immunoprecipitation and immunoblotting for CTSE were performed, as previously described (21). T_conv, wild-type, or knockout T_regs were purified from spleens, as described. Equal numbers of natural T_regs and T_conv cells were lysed with lysis buffer. To all supernatants, lysis buffer containing 0.1% Tween 20, 50 mM HEPES, 150 mM NaCl, 1 mM EDTA, 2.5 mM EGTA, and 1 complete protease inhibitor tablet (Roche, Indianapolis, IN) per 50 ml lysis buffer was added. Supernatants were immunoprecipitated with anti-mouse CTSE (R&D Systems) and protein G-Sepharose beads. Immunoprecipitates were resolved by 10% SDS-PAGE (Invitrogen Life Technologies), and blots were probed with an anti-goat HRP secondary Ab (Amersham Biosciences). Blots were developed using ECL (Amersham Biosciences) and autoradiography.

The 4 μM sulfate latex beads (Molecular Probes) were incubated overnight at room temperature with rotation in a 1:4 dilution of anti-CD3 and anti-CD28 Ab mix (13.3 μg/ml anti-CD3 [murine clone 145-2c11 and 26.6 μg/ml anti-CD28 murine clone 37.51]; eBioscience). Beads were washed three times with 5 mM phosphate buffer (pH 6.5) and resuspended at 5 × 10⁷/ml in sterile phosphate buffer with 2 mM BSA.

In vitro T_reg suppression assays were performed, as described previously (16, 17, 22). Anti-CD3– and anti-CD28–coated beads used for T cell stimulation in these assays were prepared, as described previously (17, 22). Briefly, T_conv and T_regs from wild-type C57BL/6, Ebi3^−/−, Il10^−/−, Ebi3^−/−Il10^−/−, and Il12a^−/−Il10^−/− mice were purified by FACS. Purified T_regs were titrated into a 96-well round-bottom plate starting at a 2:1 ratio (T_conv:T_regs) with 5 × 10⁴ T_conv cells. The cells were stimulated with anti-CD3/anti-CD28–coated latex beads for 72 h. Cultures were pulsed with 1 μCi [³H]thymidine for the final 8 h of the 72-h assay and harvested with a Packard harvester. The cpm were determined using a Packard Matrix 96 direct counter (Packard Biosciences).

In vitro Transwell suppression assays were performed, as described previously, to assess the ability of T_regs to suppress via soluble mediators (17, 22). T_conv and T_regs from wild-type C57BL/6 mice, and T_regs from Ebi3^−/−, Il10^−/−, Ebi3^−/− Il10^−/− mice were purified by FACS. Wild-type T_conv and wild-type or knockout T_regs were cultured at a 2:1 ratio in the Transwell insert with a pore size of 0.4 μM (Millipore). Target wild-type T_conv or Tnfrsf10b^−/− (DR5^−/−) T_conv cells were activated in the bottom compartment of the Transwell plate with anti-CD3– and anti-CD28–coated latex beads for 72 h. Where indicated, neutralizing IL-10 mAb (JES5-2A5; BD Biosciences), neutralizing IL-35 mAb (V1.4C4.22), isotype control, or DR5-Fc was added to standard T_reg assays and Transwell experiments at the concentrations indicated. Where indicated, T_conv cells were fixed at a 1:5 dilution of 20% formaldehyde in culture medium, incubated at room temperature for 20 min, and washed three to five times with medium prior to culture. After 64 h in culture, top Transwell inserts were removed and [³H]thymidine was added directly to the responder T_conv in the bottom chambers of the Transwell plate for the final 8 h of the 72-h assay. Cultures were harvested with a Packard harvester. The cpm were determined using a Packard Matrix 96 direct counter (Packard Biosciences).

For in vitro assays with transfected 293T cells cocultured with T_conv cells, 293T cells were transfected with Ctse (murine CTSE in pIYneo [pCIneo (Promega) with an IRES-YFP expression cassette]; provided by B. Chain, University College London, London, U.K.) or Tnfsf10 (murine TRAIL in pIGneo [pCIneo with IRES-GFP]; provided by T. Griffith, University of Minnesota, Minneapolis, MN) alone or in combination. Posttransfection (48 h), the cells were irradiated (3000 rad) and seeded at a density of 7 × 10³ cells/well in the 96-well flat-bottom plate. Purified C57BL/6 T_conv cells were added to the seeded plate at 8 × 10⁴ per well and stimulated with anti-CD3– and anti-CD28–coated beads for 72 h with [³H]thymidine added during the last 8 h of culture. T cell proliferation was calculated by subtracting the basal [³H]thymidine incorporation of irradiated 293T plus unstimulated T_conv cells.

Homeostasis assays were performed, as described previously (16, 23). Briefly, naive Thy1.1⁺ T_conv cells from B6.PL mice, which were used as target cells, and Thy1.2⁺ wild-type or knockout T_regs were purified by FACS. T_conv cells (2 × 10⁶) and T_regs (5 × 10⁵) were resuspended in 0.5 ml PBS plus 2% FBS and injected i.v. into Rag1^−/− mice. Where indicated, the mice were injected on days 0 and 3 with anti-TRAIL Ab (0.3 mg; provided by T. Griffith, University of Iowa) or isotype control Ab (0.3 mg; R&D Systems). Mice were euthanized 7 d post-transfer, and splenocytes were counted, stained, and analyzed by flow cytometry using Abs against CD4, Thy1.1, Thy1.2 (BioLegend), and Foxp3 (BD Biosciences). For each group, six to eight mice were analyzed.

A recovery model of colitis/inflammatory bowel disease (IBD) was used, with some modifications (14, 23). Briefly, Rag1^−/− mice were injected i.v. with 0.5 × 10⁶ wild-type or DR5^−/− (CD4⁺CD45RB^highCD25⁻) naive T_conv cells to induce IBD. Mice were weighed at the time of injection (time 0) and every week on the same day. At the onset of clinical symptoms of colitis (∼4 wk post-T_conv cell transfer), the mice were divided into T_reg recipient or no T_reg control groups. Purified T_regs from wild-type, Ebi3^−/−, Il10^−/−, or Ebi3^−/−Il10^−/− were injected i.p. All mice were weighed weekly and euthanized 32 d after the initial T cell transfer. In experimental mice, the colons were collected and fixed in 10% neutral-buffered formalin 4 wk after T_reg injection. The tissues were further processed, and 4-μm sections were cut and stained with H&E. Pathology of the large intestine was scored in a blinded manner using a semiquantitative scale, as described previously (23). In summary, grade 0 was assigned when no changes were observed; grade 1, minimal inflammatory infiltrates present in the lamina propria with or without mild mucosal hyperplasia; grade 2, mild inflammation in the lamina propria with occasional extension into the submucosa, focal erosions, minimal to mild mucosal hyperplasia, and minimal to moderate mucin depletion; grade 3, mild to moderate inflammation in the lamina propria and submucosa occasionally transmural with ulceration and moderate mucosal hyperplasia and mucin depletion; grade 4, marked inflammatory infiltrates commonly transmural with ulceration, marked mucosal hyperplasia and mucin depletion, and multifocal crypt necrosis; grade 5, marked transmural inflammation with ulceration, widespread crypt necrosis, and loss of intestinal glands.

The Foxp3^−/− rescue model was performed, as described previously (23). Briefly, wild-type or knockout T_regs purified by FACS were injected (10⁶) i.p. into 2- to 3-d-old Foxp3^−/− mice. Recovery from disease was monitored weekly and reported as a clinical score. Five macroscopic categories were used to generate a 6-point scoring system. Mice were scored on the first four categories based on whether they showed (score of 1) or did not show (score of 0) the following characteristics: body size runted; tail is scaly and/or with lesions; ears small and scaly with or without lesions; and eyelids scaly and/or not fully open. The final scoring parameter was monitoring the activity level of the mouse. A score of 0 was assigned if the mouse was normal. A score of 1 was assigned if the mouse’s activity was moderately impaired, and a score of 2 was assigned if the mouse was immobile. A combined score of 4 or greater was assigned moribund for longevity. Mice were euthanized 25 d posttransfer, spleen cells were counted and stained, and cell numbers were determined by flow cytometry. Lung, liver, and ear pinna were prepared for H&E analysis, and the severity of inflammation was assessed and scored in a blinded manner by an experienced veterinary pathologist. The scoring system used for assessing inflammation was based on a simple algorithm for expressing inflammatory infiltrates in the lungs, liver, and ear. The scores allotted to these three tissues were 0–9, 0–11, and 0–8, respectively, giving a maximum possible total of 28. Scoring criteria for each organ was as follows. The lung score was based upon inflammation in the peribronchiolar region, perivascular region, or interstitium. A score of 0–3 was assigned to each category, with 0 being minimal or no inflammation, and scores of 1, 2, or 3 indicative of <10, 10–50, or >50%, respectively. The liver was scored based on three criteria. First was the degree of portal tract inflammation, with a score of 0 assigned to minimal or no inflammation. A score of 1, 2, or 3 was assigned if inflammation was associated with <25, 25–75, or >75% of the liver portal tracts, respectively. The second criteria related to portal/periseptal interface hepatitis. A focus of interface hepatitis associated with either a few or most of the portal tracts was scored 1 and 2, respectively. Two or more foci of interface hepatitis surrounding <50 or >50% of the portal tracts or periseptae were scored 3 and 4, respectively. Third, foci of granulocytes and/or lymphocytes with or without necrotic hepatocytes that expand the sinusoid were considered foci of inflammation. The number of inflammatory foci in 10 contiguous original magnification ×10 objective fields was counted and recorded as the average number of foci per ×10 field and given a score of 0–4. A score of 0 was assigned when sinusoidal foci of inflammatory cells were absent. One focus or less per ×10 field, 2–4 foci per ×10 field, 5–10 foci per ×10 field, and >10 foci per ×10 field were scored 1, 2, 3, and 4 respectively. The ear pinna was similarly scored based on two parameters, as follows: the percentage of the ear dermis with inflammatory infiltrates and the intensity of the dermal inflammation. For percentage analysis, a score of 0 was assigned when the inflammatory cells in all segments were not beyond that of normal background level. A score of 1, 2, 3, or 4 was assigned when the average percentages for the segments were <25, 25–50, 51–75, or >75%, respectively. The intensity of the inflammatory infiltrate in the dermis was assessed as being of a loose or dense nature. A score of 0 was assigned when inflammatory cells in the dermis were not beyond the normal background level. When all of the inflammation was of the loose nature, a score of 1 was assigned. When there was a mixture of loose and dense inflammatory cell infiltrates, a score of 2 was assigned when the loose form was dominant. A score of 3 was assigned when the dense form was dominant. A score of 4 was assigned when all of the inflammation was of a dense nature.

Wild-type or knockout T_regs were purified by FACS and mRNA isolated using the Qiagen micro RNA kit (Qiagen). Quality was confirmed by UV spectrophotometry and by analysis on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Total RNA (100 ng) was processed in the Hartwell Center for Biotechnology and Bioinformatics according to the Affymetrix eukaryote two-cycle target-labeling protocol and arrayed on a Mouse-430v2 GeneChip array. The expression data from the Affymetrix U133 plus two arrays were analyzed as MAS 5.0 signal log-start transformed using the following formula: log signal = natural log (signal + 20). This transform improves data dispersion and normality, and stabilizes the variance of the data (24). Statistical tests and batch effect removal were performed using Partek Genomics Suite (St. Louis, MO). The log₂ ratio of Ebi3^−/−Il10^−/− T_reg to wild-type T_reg was calculated, and the 20 most positively induced named genes were selected. The log₂ ratios are calculated in STATA/SE 11.0 (College Station, TX) by the following formula: log ratio A over B = log(exponentiation[mean log signal A]/exponentiation[mean log signal b])/log (2). Minimum selected gene had a log ratio of 1.65, which is 3.14-fold induced. Log ratios of the Il10^−/− T_reg and the Ebi3^−/− T_reg with respect to wild type were also defined and plotted with the log ratio of Ebi3^−/−Il10^−/− T_reg to wild type as a heat map using Spotfire Decision Site software (Fig. 3A). The t tests were then applied to each probe set to compare the Ebi3^−/−Il10^−/− T_reg with wild-type T_reg and single knockout T_reg samples and log₂ ratios were calculated. The p value from the t tests were then −log₁₀ transformed to create the significance score seen in the x-axis of the volcano plot (Fig. 3B). A second series of t tests was performed to compare T_reg with T_conv and to develop a T_reg signature. Probe sets that had a p value <10⁻⁵, an absolute value log ratio of T_reg versus T_conv of at least 3 (log₂), and a defined gene name were selected for each category in the signature that the mean was found. If a gene name appeared more than once, then the mean data were averaged for that gene. The scores were calculated by finding the maximum and minimum values for each gene and then rescaling them from 0 to 1 by the following formula: score_g = (observed mean_g − minimum mean_g)/(maximum mean_g − minimum mean_g) for each gene g. These gene scores were then sorted in descending order by the T_reg:T_conv log ratio that includes activated and resting cells and graphed as a heat map in Spotfire Decision Site (data not shown). The microarray data from this study have been submitted into the Gene Expression Omnibus repository, accession number GSE29262 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE29262).

Unless otherwise stated, a Student t test was used to determine statistical significance. All calculations were done using GraphPad software. A p value <0.05 was considered significant.

We first assessed the functional capacity of T_regs that lacked the ability to secrete IL-10 or IL-35 by generating Ebi3^−/−Il10^−/− and Il12a^−/−Il10^−/− mice (note that both Ebi3 and Il12a/p35 are required for IL-35 production) (16, 17). Purified wild-type, Ebi3^−/−, Il10^−/−, Ebi3^−/−Il10^−/−, and Il12a^−/−Il10^−/− T_regs were assessed in a standard T_reg assay [note that these double-deficient T_regs would not be able to secrete IL-10 or IL-35, and although Ebi3 is also used by IL-27 and Il12a/p35 is also used by IL-12, these cytokines are not produced by T_regs (16)]. Surprisingly, Ebi3^−/−Il10^−/− and Il12a^−/−Il10^−/− T_reg function was comparable or slightly better than wild-type T_regs in suppressing their target T_conv (Fig. 1A). We have previously shown that if T_regs are optimally stimulated by anti-CD3– and anti-CD28–coated beads and in contact with T_conv cells in the upper chamber (insert) of a Transwell plate, they can suppress third-party T_conv cells in the lower chamber across a semipermeable membrane (17). Importantly, this suppression requires, and is limited to, IL-10 and IL-35. Thus, we anticipated that the loss of IL-10 and IL-35 would render Ebi3^−/−Il10^−/− T_regs unable to suppress across a Transwell. Strikingly, Ebi3^−/−Il10^−/− T_regs suppressed T_conv cells across a Transwell comparable to their wild-type counterparts, even though Ebi3^−/− and Il10^−/− T_regs were partially defective (Fig. 1B). This equivalency in function was further supported by experiments with CFSE-labeled target T_conv cells and Transwell experiments with titrated Ebi3^−/−Il10^−/− T_regs in presence of fixed or unfixed T_conv cells (Supplemental Fig. 1A, 1B). These data suggest that Ebi3^−/−Il10^−/− T_regs, unlike Ebi3^−/− and the Il10^−/− T_regs, are functionally intact in in vitro suppression assays.

We next asked whether Ebi3^−/−Il10^−/− T_regs were functionally equivalent in several in vivo models. The adoptive transfer of T_regs into neonatal Scurfy or Foxp3^−/− mice has been shown to restore normal immune homeostasis and prevent the lethal, systemic autoimmune disease that develops in these mice (19, 25, 26). Two-day-old neonatal Foxp3^−/− mice were injected with 10⁶ wild-type, Ebi3^−/−, Il10^−/−, or Ebi3^−/−Il10^−/− T_regs. Clinical symptoms, histological analysis, and CD4⁺ T cell numbers were determined when the mice were ∼4 wk old. Although no defects were observed with the Ebi3^−/− T_reg recipients, increased histological scores were observed with Il10^−/− T_reg recipients. In contrast, Ebi3^−/−Il10^−/− T_regs were clearly capable of fully restoring immune homeostasis despite the loss of these two key regulatory cytokines (Supplemental Fig. 2A–C). We also assessed the ability of these T_reg populations to rescue immune homeostasis in mixed bone marrow chimeras generated using a 50:50 mixture of bone marrow from Foxp3^−/− mice and either wild-type, Ebi3^−/−, Il10^−/−, or Ebi3^−/−Il10^−/− mice transferred into Rag1^−/− mice. Interestingly, significant defects were observed in the ability of Ebi3^−/− and Il10^−/− bone marrow to rescue the Foxp3^−/− phenotype (Supplemental Fig. 2D, 2E). In contrast, the Foxp3^−/− bone marrow recipients of Ebi3^−/−Il10^−/− T_regs were largely intact and comparable to their wild-type T_reg, Foxp3^−/− recipient counterparts (Supplemental Fig. 2D, 2E).

T_regs have been shown to regulate the homeostatic expansion of T_conv cells in lymphopenic Rag1^−/− mice (27–29). Purified wild-type Thy1.1 T_conv cells, either alone or in presence of wild-type, Ebi3^−/−, Il10^−/−, Ebi3^−/−Il10^−/−, or Il12a^−/−Il10^−/− Thy1.2⁺ T_regs, were adoptively transferred into Rag1^−/− mice, and splenic Thy1.1 T_conv and Thy1.2 T_reg numbers (data not shown) were determined 7 d later. In the presence of wild-type, but not Ebi3^−/− or Il10^−/− T_regs, T_conv cell expansion was significantly reduced (Fig. 1C). Surprisingly, the capacity of Ebi3^−/−Il10^−/− and Il12a^−/−Il10^−/− T_regs to control T_conv cell expansion was comparable to wild-type T_regs.

T_regs cure colitis in mice, a model for inflammatory bowel disease (IBD) in humans, in an IL-10– and IL-35–dependent manner (16, 30). Colitis in mice is induced experimentally by transferring low numbers of naive CD4⁺CD45RB^highCD25⁻ T_conv cells into Rag1^−/− mice (31). Recovery from disease, marked by weight gain and decreased histopathology, is observed only in mice that receive purified T_regs ∼4 wk after the initial T_conv cell transfer (14). We used this recovery model of colitis to assess the functional capacity of Ebi3^−/−Il10^−/− T_regs in vivo. Approximately 4 wk post-T_conv cell transfer, recipients developed clinical symptoms of colitis (monitored by weight loss) and were either left untreated or treated with either wild-type, Ebi3^−/−, Il10^−/−, or Ebi3^−/−Il10^−/− T_regs. As expected, mice that did not receive T_regs continued to lose weight, and exhibited substantial histiocytic infiltration and goblet cell destruction during the subsequent 4 wk (Fig. 1D, 1E, Supplemental Fig. 2F). In contrast, the wild-type T_reg recipients started to gain weight within 1 wk of transfer. Despite previous studies clearly demonstrating the inability of Ebi3^−/− or Il10^−/− T_regs to cure colitis, weight gain and improved histological parameters were evident in the Ebi3^−/−Il10^−/− T_reg recipients, suggesting that these double-inhibitory cytokine-deficient T_regs had regained their regulatory potential (Fig. 1D, 1E, Supplemental Fig. 2F).

To rule out the possibility that this regulatory restoration had occurred as a consequence of their development in the absence of IL-10 and IL-35 and/or due to alternate cell-extrinsic mechanisms, we directly compared the suppressive capacity of wild-type and Ebi3^−/−Il10^−/− T_regs that had developed in the same environment. To address this possibility, we generated mixed bone marrow chimeras with a 1:1 ratio of congenically marked Thy1.1⁺ wild-type bone marrow with Thy 1.2⁺ wild-type or Ebi3^−/−Il10^−/− bone marrow into sublethally irradiated Rag1^−/− mice. Eight weeks posttransfer, T_regs were purified by FACS from the mixed bone marrow chimeras and assessed in in vitro Transwell and in vivo homeostasis assays. Chimera-derived Ebi3^−/−Il10^−/− T_regs and wild-type T_regs suppressed third-party T_conv cells comparably across a Transwell (Fig. 2A). In contrast, similarly prepared Ebi3^−/− and Il10^−/− T_regs were defective (data not shown). Furthermore, Thy1.2⁺ Ebi3^−/−Il10^−/− T_regs and wild-type T_regs that had developed in the bone marrow chimeras suppressed T_conv expansion comparably in homeostasis assay (Fig. 2B). Taken together, these data suggest that a cell-intrinsic modification had occurred in the Ebi3^−/−Il10^−/− T_regs to render them functionally comparable to wild-type T_regs to compensate for their inability to secrete IL-10 and IL-35.

Given that Ebi3^−/−Il10^−/− T_regs can suppress T_conv cells across a Transwell, they had clearly acquired a suppressive mechanism that operated via a soluble mediator. Beyond IL-10 and IL-35, TGF-β is the only other known cytokine or soluble factor that would likely function across a Transwell that has been suggested to play a role in T_reg function (note that cAMP and adenosine are highly labile inhibitors that are only active in very close proximity) (2, 3, 9). We assessed any potential role for TGF-β by comparing the capacity of wild-type and Ebi3^−/−Il10^−/− T_regs to suppress across a Transwell using third-party T_conv cells from CD4–dominant negativeTGF-βRII transgenic mice that are resistant to TGF-β–mediated suppression (32). The data clearly show that the suppressive capacity of wild-type and Ebi3^−/−Il10^−/− T_regs was comparable when T_conv cells resistant to TGF-β–mediated suppression were used as target cells (Supplemental Fig. 1C). This suggested that the compensatory suppressive mechanism used by Ebi3^−/−Il10^−/− T_regs was not TGF-β.

To identify this compensatory suppressive mechanism, we compared the gene expression profile of wild-type, Ebi3^−/−, Il10^−/−, and Ebi3^−/−Il10^−/− T_regs using Affymetrix GeneChip arrays. We first generated a list of highly differentially expressed wild-type T_reg signature genes, by comparison of the array profile with wild-type T_conv, to determine whether there were any notable global changes in gene expression in wild-type versus Ebi3^−/−Il10^−/− T_regs. Minimal variations were observed in the expression (up or down) of 47 highly modulated T_reg signature genes (data not shown). Indeed, global analysis revealed very few differences between wild-type and Ebi3^−/−Il10^−/− T_regs (Fig. 3A, 3B, data not shown). The two notable exceptions were Ap1S3 (adaptor-related protein complex 1, σ 3 subunit) and Ctse (Fig. 3A, 3B). AP1S3 is the σ subunit of the adaptor protein-1 complex that is a component of the clathrin-coated vesicles associated with the trans-Golgi network that mediate vesicular formation and transport (33). The significance of its upregulation in Ebi3^−/−Il10^−/− T_regs is unknown and was not selected for further study in this work. CTSE is an intracellular aspartic protease of the endolysosomal pathway that has been primarily implicated as a component of the Ag-processing machinery for the MHC class II pathway (34). qPCR, immunoprecipitation/Western blot analysis, and intracellular staining with purified T_regs confirmed that CTSE mRNA and protein are highly upregulated in Il10^−/− and Ebi3^−/−Il10^−/− T_regs compared with wild-type and Ebi3^−/− T_regs (Fig. 3C–E, Supplemental Fig. 3A–C).

Interestingly, CTSE has been implicated in the cleavage and/or processing of TRAIL (Tnfsf10; TNF [ligand] superfamily, member 10) and its release from the cell surface (35, 36). TRAIL is a suppressive molecule of the TNF superfamily that can function in its surface-bound form or as a soluble trimer (37, 38). TRAIL can mediate apoptosis and programmed regulated necrosis (necroptosis) or suppress proliferation (37, 39). Furthermore, activated CD4⁺Foxp3⁺ T_regs and CD8⁺ T_regs may express and use TRAIL as a suppressive mechanism (40, 41). Thus, we speculated that increased CTSE in Ebi3^−/−Il10^−/− T_regs might result in an increase in the functional capacity of surface TRAIL and/or an increase the release of soluble TRAIL. To directly examine this possibility, 293T cells were transfected with expression plasmids encoding Ctse and/or Tnfsf10 and used to assess the ability of TRAIL to limit T cell proliferation. TRAIL transfectants limited T cell proliferation, and this was further enhanced in the presence of CTSE (Fig. 4). These data suggest that CTSE may play a role in enhancing the function of TRAIL by either increasing its activity via processing or increasing the generation of soluble TRAIL. These data also raised the possibility that Ebi3^−/−Il10^−/− T_regs are dependent on TRAIL for their suppressive activity, whereas wild-type T_regs are not.

We first assessed whether there were any changes in the level or rate of TRAIL expression during activation of wild-type, Ebi3^−/−, Il10^−/−, or Ebi3^−/−Il10^−/− T_regs. Minimal alterations in Tnfsf10 (TRAIL) mRNA expression were observed over time or between the four T_reg populations (Supplemental Fig. 3D). Whereas all T_reg populations exhibited increased TRAIL surface expression following activation, Ebi3^−/−Il10^−/− T_regs expressed significantly higher levels of TRAIL after 16 h, but not 24 h, postactivation (Fig. 5A, 5B, Supplemental Fig. 3E). This suggested that the kinetics of TRAIL expression is accelerated in Ebi3^−/−Il10^−/− T_regs. Interestingly, although IL-10 appeared to influence CTSE expression (Fig. 3C, 3E, Supplemental Fig. 3A–C), IL-35 may influence other parameters that influence TRAIL expression, as Ebi3^−/− T_regs expressed slightly higher levels of TRAIL at 16 h compared with wild-type or Il10^−/− T_regs (Supplemental Fig. 3E).

We then used various approaches to determine the extent to which this accelerated TRAIL expression meant that the Ebi3^−/−Il10^−/− T_regs were dependent on TRAIL-mediated suppression. TRAIL mediates its suppression in part via caspase-mediated apoptosis (37). Thus, we asked whether Ebi3^−/−Il10^−/− T_regs mediated suppression in a caspase-dependent fashion by performing a Transwell suppression assay in the presence of the general caspase inhibitor z-VAD-Fmk or a vehicle control (42). Although wild-type T_reg suppression was unaffected by z-VAD-Fmk or its DMSO vehicle control, Ebi3^−/−Il10^−/− T_reg-mediated suppression was blocked (Fig. 5C). These data suggest that Ebi3^−/−Il10^−/− T_regs suppress T_conv proliferation via a caspase-dependent pathway.

TRAIL signaling in the mouse is mediated through DR5 (Tnfrsf10b; TNFR superfamily, member 10b; also known as TRAIL-R2) (43). Therefore, we first asked whether the Ebi3^−/−Il10^−/− T_regs were able to suppress Tnfrsf10b^−/− T_conv cells (hereafter referred to as DR5^−/−) in conventional and Transwell suppression assays. As previously shown, wild-type and Ebi3^−/−Il10^−/− T_regs suppressed wild-type T_conv cells comparably (Fig. 6A). Furthermore, wild-type T_regs could effectively suppress DR5^−/− T_conv cells. However, Ebi3^−/−Il10^−/− T_regs were less effective at suppressing DR5^−/− T_conv cells in a standard T_reg assay (Fig. 6A) and completely failed to suppress across a Transwell (Fig. 6B). Secondly, we assessed whether a DR5-Fc fusion protein or an anti-TRAIL blocking Ab was able to inhibit Ebi3^−/−Il10^−/− T_reg-mediated suppression of wild-type T_conv cells. Although DR5-Fc had a minimal effect on the suppression mediated by wild-type T_regs across a Transwell, it blocked suppression by Ebi3^−/−Il10^−/− T_regs in a dose-dependent manner (Fig. 6C). Similarly, anti-TRAIL, but not an isotype control Ab, reduced the suppressive capacity of Ebi3^−/−Il10^−/−, but not wild-type, T_regs [note that this TRAIL mAb is known to block activity weakly in vitro, but very effectively in vivo (44)] (Supplemental Fig. 4A). These results suggest that Ebi3^−/−Il10^−/− T_regs mediate suppression across a Transwell in vitro via soluble TRAIL.

We then assessed the contribution of TRAIL-mediated suppression by Ebi3^−/−Il10^−/− T_regs in vivo. First, congenic Thy1.1⁺ wild-type T_conv cells were injected either into Rag^−/− mice alone or in the presence of Thy1.2⁺ wild-type or Ebi3^−/−Il10^−/− T_regs. Isotype control Ab or anti-TRAIL was injected on days 0 and 3, and homeostatic expansion of the Thy1.1⁺ T_conv cells was determined 7 d later. T_conv cell expansion, wild-type T_reg-mediated suppression, and T_reg numbers were unaffected by the anti-TRAIL treatment (Fig. 6D, data not shown). In striking contrast, TRAIL inhibition blocked the ability of Ebi3^−/−Il10^−/− T_regs to suppress T_conv cell expansion in vivo.

Second, we assessed the extent to which the Ebi3^−/−Il10^−/− T_regs could cure colitis induced by DR5^−/− T_conv cells. The development and severity of colitis induced by wild-type or DR5^−/− T_conv cells in Rag^−/− mice were comparable (Fig. 6E, 6F, Supplemental Fig. 4B). At the onset of clinical symptoms (5% loss of body weight; ∼4 wk), mice were treated with wild-type or Ebi3^−/−Il10^−/− T_regs. Wild-type T_reg recipients gained weight and recovered from the clinical symptoms of colitis regardless of whether the disease had been induced by wild-type or DR5^−/− T_conv cells (Fig. 6E, 6F). In contrast, Ebi3^−/−Il10^−/− T_regs could cure colitis caused by wild-type, but not DR5^−/− T_conv cells. Histological analysis of the colon 4 wk post-T_reg treatment confirmed that Ebi3^−/−Il10^−/− T_regs were unable to reverse DR5^−/− T_conv cell-induced colitis (Supplemental Fig. 4B).

Third, if TRAIL was essential for Ebi3^−/−Il10^−/− T_reg-mediated suppression, then its genetic deletion should abrogate their regulatory capacity. Our data suggest that although wild-type and Ebi3^−/−Il10^−/− T_regs could effectively mediate suppression of T_conv cells across a Transwell, Ebi3^−/−Il10^−/−Tnfsf10^−/− T_regs could not inhibit T_conv target cell proliferation (Fig. 6G). Taken together, these data clearly demonstrate that Ebi3^−/−Il10^−/− T_regs require TRAIL for maximal suppressive function, and that soluble TRAIL appears to be their only mechanism of suppression. In contrast, wild-type T_regs exhibit minimal TRAIL dependence and use IL-35 and IL-10 as their soluble mediators of suppression.

Loss of IL-10 and IL-35 production by T_regs led to increased CTSE expression and subsequent dependence on TRAIL-mediated suppression. We questioned the extent to which unmanipulated examples of this T_reg functional plasticity might exist. Differential CTSE expression has been reported in different inbred mouse strains (21). In particular, C57BL/6 mice express low levels of CTSE, whereas expression in BALB/c and 129 mice is high. We first confirmed these observations by assessing Ctse expression by qPCR and intracellular staining (Fig. 7A, 7B). The results clearly indicate that BALB/c T_regs express higher levels of CTSE, consistent with previous observations (21). Next, we assessed the kinetics of TRAIL surface expression on BALB/c T_regs following activation. Interestingly, BALB/c T_regs expressed slightly higher levels of surface TRAIL than C57BL/6 T_regs, particularly at 16 h postactivation (Fig. 7C). Indeed, the pattern of CTSE and TRAIL expression exhibited by BALB/c T_regs was analogous to observations made with Ebi3^−/−Il10^−/− T_regs (compare Figs. 3C, 3D, 5A with Fig. 7A–C), and was consistent with previous suggestions (45). We then examined the suppressive capacity of BALB/c and C57BL/6 T_regs in presence or absence of reagents that block IL-10, IL-35, or TRAIL. Whereas anti–IL-10 and the isotype control Ab had little effect on the suppression mediated by either T_reg population in a Transwell assay, IL-35 neutralizing mAb blocked suppression mediated by C57BL/6, but not BALB/c, T_regs (Fig. 7D). In contrast, DR5-Fc partially inhibited suppression mediated by BALB/c, but not C57BL/6, T_regs. Thus, C57BL/6 T_regs seem to be more dependent on IL-35, whereas BALB/c T_regs are more dependent on TRAIL-mediated suppression. This raises the possibility that genetic variations predispose T_regs to preferential modes of immunosuppression.

T_regs can function in diverse anatomical locations and in a wide variety of immunological and disease settings (46). Consequently, the large array of suppressive mechanisms that T_regs are reported to possess may help them maintain immune homeostasis under diverse scenarios. Indeed, T_regs may have specialized mechanisms for controlling specific cell types as T_regs appear to require IFN regulatory factor-4, T-bet, and STAT3 to suppress Th2, Th1, and Th17 cells, respectively (10–12). However, this may have a greater influence on their migratory behavior than the mechanisms they use to mediate suppression. Importantly, the relative importance of specific mechanisms of T_reg function and whether T_regs possess mechanistic flexibility have not been elucidated. Previous studies have reported that deficiency of IL-10 or IL-35 alone results in defective T_reg function (16, 18). Thus, our finding that T_regs lacking IL-35 and IL-10 are fully functional, instead of relying on TRAIL-mediated suppression as a primary mechanism of action, was very surprising. This implies that T_regs can exhibit remarkable functional plasticity and possess control mechanisms to compensate for the loss of key regulatory tools.

There is a reciprocal relationship in the expression of IL-10 and CTSE (47). Our data clearly show that Ebi3^−/−Il10^−/− T_regs are dependent on TRAIL for their regulatory function in vitro and in vivo. Furthermore, our studies suggest that increased expression of CTSE enhances the rate and extent of TRAIL surface expression and TRAIL function in mediating T cell suppression. It is possible that CTSE may process full-length TRAIL to enhance its ligand binding and/or may mediate the cleavage of cell surface TRAIL to generate a soluble version. Soluble TRAIL is thought to be either secreted into microvesicles (48) or cleaved from the cell surface (49). Whereas the precise mechanism by which CTSE enhances TRAIL function requires further elucidation, consistent with our results, previous studies have shown that proteolytic cleavage of TRAIL from the cell surface can be mediated by CTSE (35, 36). Thus, in Ebi3^−/−Il10^−/− T_regs, CTSE upregulation may play a role in the generation of soluble TRAIL. In contrast, expression of IL-10 by wild-type T_regs may suppress CTSE expression and thus reduce the contribution of TRAIL-mediated killing. These data also support the capacity of activated T_regs to use TRAIL (40, 41), and further highlight the complex interregulatory pathways modulated by inhibitory cytokines. However, TRAIL is clearly not used by Il10^−/− T_regs, emphasizing that loss of IL-35 expression also contributes to the ability of Ebi3^−/−Il10^−/− T_regs to mediate suppression via TRAIL. Although the contribution of IL-35 in minimizing TRAIL-mediated suppression remains to be defined, it is noteworthy that Ebi3^−/− T_regs exhibit accelerated TRAIL expression following activation, raising the possibility that IL-35 may suppress a distinct component of the TRAIL-processing machinery.

An important question is whether the extent of the physiological impact of the T_reg functional plasticity revealed in our study has applicability. As shown in this study and previous studies, substantial differences in CTSE expression occur in different mouse strains with BALB/c mice expressing high levels of CTSE and C57BL/6 mice expressing low levels (21, 45). Interestingly, BALB/c T_regs appeared to phenocopy Ebi3^−/−Il10^−/− T_regs in terms of their pattern of CTSE and TRAIL expression and, thus, their dependence on TRAIL-mediated suppression. Although there are certainly multiple genetic factors that might underlie differences in the function of T_regs from distinct genetic backgrounds, our data suggest differential CTSE expression may be one contributing factor. Whether this is related to the necessity of T_regs to adapt to the different Th cell bias exhibited in different mouse strains remains to be determined (50, 51). Given that previous studies have shown that T_regs can use different transcription factors to tackle different Th environments (10–12), it is possible that these may underlie the differential utilization of T_reg-suppressive mechanisms observed in this study. This remarkable T_reg functional plasticity may also be important in providing a backup mechanism in scenarios in which IL-10 and IL-35 production and/or signaling may be perturbed, and thus may empower T_regs with the ability to adjust to different environmental settings. Lastly, the possibility that TRAIL may be a legitimate target for the treatment of diseases impacted by excessive T_reg function, such as cancer, requires further study.

We thank Doug Green and Benny Chain for advice; Rick Blumberg, Doug Green, Sasha Rudensky, Terry Geiger, and Jim Ihle for mice; Amanda Burton and Kate Vignali for technical assistance; Creg Workman for help with Affymetrix analysis; Karen Forbes, Amy Krause, and Ashley Castellaw for mouse colony maintenance and breeding; Richard Cross, Greig Lennon, and Stephanie Morgan for FACS; the staff of the Shared Animal Resource Center at St. Jude for the animal husbandry; and the Hartwell Center for Biotechnology and Bioinformatics at St. Jude for real-time PCR primer/probe synthesis.

D.A.A.V. and L.W.C. have submitted patents that are pending and are entitled to a share in net income generated from licensing of these patent rights for commercial development. The other authors have no financial conflicts of interest.