Down-regulation of Gfi-1 expression by TGF-β is important for differentiation of Th17 and CD103⁺ inducible regulatory T cells

Th2 cells differentiated from Gfi1 cKO (Gfi1^fl/fl-CD4Cre) mice have a severe defect in cell expansion and also produce some IFN-γ, suggesting that Gfi-1 is important for sustaining the Th2 cell phenotype. A recent report suggests that Gfi-1, as a transcriptional repressor, is associated with the histone lysine-specific demethylase 1 (LSD1), which results in demethylation of histone 3 lysine 4 (H3K4) at target promoters (46). LSD1 demethylates mono- and dimethylation but not trimethylation of H3K4. However, because of the balance of different methylation states, active gene promoter regions usually contain higher levels of H3K4 trimethylation than of mono- and dimethylation (47). To search for Gfi-1 target genes in T cells, the chromatin immunoprecipitation (ChIP)–Seq technology was applied. We used H3K4 trimethylation, the end product of H3K4 methylation, as an indicator, reasoning that such modification would not accumulate in Gfi1^−/− cells at Gfi-1 target sites because of a lack of the mono- or dimethylated forms of H3K4. As expected, both wild-type and Gfi1 cKO Th2 cells displayed a similar pattern of H3K4 trimethylation at the Gata3 locus (Fig. 1 A).1 Because H3K4 methylation is associated with active chromatin, this indicates that Gfi-1, although important for Th2 expansion, is not essential for the Th2 polarization process. Interestingly, at the Cd103 locus, whose expression is associated with a subset of Treg cells, Gfi1 cKO Th2 cells showed high levels of H3K4 modification that was absent in wild-type Th2 or Th17 cells (Fig. 1 B). At the Rorc locus, which encodes RORγt, critical for Th17 cell differentiation, Gfi1 cKO Th2 cells displayed levels of H3K4 modification similar to Th17 cells (Fig. 1 C). H3K4 modification was also detected at the Il23r locus in Gfi1 cKO Th2 cells (Fig. 1 D). Such a modification was absent in wild-type Th2 cells. Gfi1 cKO Th2 cells did not appear to be contaminated with Th17 cells, because H3K4 trimethylation at the Il21 and Il17f loci was not observed the Gfi1 cKO Th2 cells, although it was prominent in Th17 cells (Fig. 1, E and F). These results indicate that Gfi-1 plays an important role in chromatin remodeling of Th17 and iTreg cell–related genes, including Rorc, Il23r, and Cd103, in Th2 cells. Although Gfi1 cKO Th2 cells displayed an active Rorc locus, expression of RORγt in Gfi1 cKO Th2 cells is much lower than that in Th17 cells (unpublished data), indicating that additional factors, possibly induced by TGF-β and IL-6, are required for Rorc expression as well as for the induction of IL-17.

To test the significance of the chromatin modification change at the Rorc locus, wild-type or Gfi1 cKO CD4 T cells were primed under Th2 conditions for one round, and cells were then either restimulated under Th2 conditions or switched to Th17 priming conditions. Gfi1 cKO Th2 cells expanded less well than wild-type Th2 cells, as previously reported (41); the expansion defect of Gfi1 cKO Th2 cells was smaller when they were cultured under Th17 conditions (Fig. 2 A). In the wild-type group, 24.5% of the cells from the culture restimulated under Th2 conditions produced IL-4, whereas, only 8% of the cells that were switched from Th2 to Th17 conditions produced IL-4, implying that TGF-β suppresses IL-4 production in differentiated Th2 cells (Fig. 2 A). Nevertheless, neither cell population was capable of producing IL-17, suggesting that TGF-β and/or IL-6 is not effective in inducing IL-17 production in differentiated Th2 cells. In contrast, in the Gfi1 cKO group, although IL-4 production was also reduced (from 16.3 to 5.6%) when cells were switched from Th2 to Th17 conditions, 8.2% of the switched cells were capable of making IL-17. These data indicate that Gfi-1 expression in Th2 cells, induced by IL-4, prevents the induction of IL-17 by TGF-β and IL-6.

To exclude the possibility that the induction of IL-17 in Gfi1 cKO cells is from a small fraction of non–Th2 cells, we bred our Gfi1 cKO mice to heterozygous G4 mice, in which one Il4 allele was replaced by eGfp (48). After one round of Th2 priming, GFP⁺ cells (Th2 cells) were sorted and restimulated under Th2 or Th17 conditions. Similar to the results shown in the previous paragraph, 6.5% of the GFP⁺ cells from the Gfi1 cKO group were able to produce IL-17 after reculturing under Th17 conditions. Again, the effect of Gfi1 in cell expansion under Th17 culture conditions was minimal (Fig. 2 B).

IL-4 has been reported to suppress the induction of IL-17 production (49, 50). The detailed mechanism is not clear. To test whether Gfi-1 mediates the suppression of IL-17 by IL-4, we performed a Th17 to Th2 switching experiment. Indeed, IL-4 suppressed IL-17 production (from 3.7 to 0.5%) and induced IL-4 production (from 0.1 to 6.4%) in wild-type cells (Fig. 2 C), but IL-4 failed to suppress IL-17 production in Gfi1 cKO cells even though it still induced its own production. Gfi1 cKO cells showed an approximately twofold decrease in cell expansion, compared with wild-type cells, when switched from Th17 to Th2 conditions. Strikingly, ∼1% of the cells can produce both IL-17 and IL-4 when Gfi1 cKO Th17 cells were switched to Th2 conditions, suggesting that Gfi-1 plays an important role in IL-4–mediated suppression of IL-17 production.

Because Gfi-1, induced by IL-4, has an effect on both RORγt and IL-17, we tested the normal expression levels of Gfi-1 during Th17 cell differentiation. Gfi-1 expression was assessed by quantitative PCR after naive CD4 T cells were primed under Th1, Th2, Th17, and iTreg cell conditions for 4 d. As expected, cells primed under different conditions specifically expressed their hallmark genes (IFN-γ for Th1 cells, IL-4 for Th2 cells, IL-17a for Th17 cells, and Foxp3 for iTreg cells; Fig. 3 A). Th2 cells expressed approximately twice as much Gfi-1 as Th1 cells, as previously reported (37). Th17 and iTreg cells expressed very low levels of Gfi-1. To test when the difference in Gfi-1 expression occurred during the differentiation process, cells were harvested at various time points after priming under different conditions. As early as 24 h after stimulation, Gfi-1 was strikingly underexpressed in cells primed under Th17 and iTreg cell conditions (Fig. 3 B).

IL-4 rapidly induces Gfi-1 expression in activated but not in naive CD4 T cells. As reported earlier (37), cells previously activated in the presence of anti–IL-4 for 24 h expressed Gfi-1 mRNA in response to IL-4 stimulation within 2 h (Fig. 3 C). In the presence of TGF-β, IL-4 failed to up-regulate Gfi-1. In addition, treatment with TGF-β suppressed the basal level of Gfi-1 expression, suggesting that TGF-β is directly involved in suppressing Gfi-1 expression. In contrast, TGF-β did not block the induction of suppressor of cytokine signaling 1 by IL-4, suggesting the specificity of Gfi-1 down-regulation by TGF-β. It did not appear that the TGF-β treatment mediated its effects by increasing cell death. No major differences in cell yields were observed, and there were no differences in the expression levels of Bcl-2 and Bim. The suppression of Gfi-1 expression by TGF-β may explain the low levels of Gfi-1 expression in cells that have undergone Th17 and iTreg cell differentiation.

To test the importance of Gfi-1 down-regulation by TGF-β, Gfi-1 was introduced by retroviral infection into activated CD4 T cells primed under Th17 conditions. 4 d after infection, IL-17 production was assessed upon restimulation with PMA plus ionomycin. A similar proportion of the cells infected by control GFP–retroviral vector (RV) and uninfected cells, in the same culture, produced IL-17 (17.8 vs. 15%; Fig. 4 A, left). However, only 3.9% of the cells infected by the Gfi-1–GFP-RV, compared with 14.1% of the uninfected cells in the same culture, produced IL-17. Three independent experiments showed significant suppression of IL-17 production by Gfi-1 (Fig. 4 A, right).

Because RORγt has been reported to be the master regulator for Th17 cells and TGF-β plus IL-6 are required for Th17 cell differentiation (17–19, 28), we tested the expression of RORγt, TGF-βRs, and IL-6Rα among Th1, Th2, and Th17 cells. RORγt was selectively expressed in Th17 cells, consistent with previous reports (Fig. S1 A, available). Interestingly, both TGF-βRI and TGF-βRII were expressed at higher levels in Th17 cells than in Th1 and Th2 cells. IL-6Rα was also found to be highest in Th17 cells (Fig. S1 A). 3 d after infection with Gfi-1–GFP-RV, although expression of RORγt, TGF-βRI, and TGF-βRII in infected cells was not affected (Fig. 4 B and Fig. S1 B), expression of IL-6Rα was suppressed (Fig. S1 B). Consistent with the intracellular staining results shown in Fig. 4 A, both IL-17a and IL-17f mRNA levels were reduced by enforced expression of Gfi-1 (Fig. 4 B). Enforced Gfi-1 expression in Gfi1 cKO Th17 cells also suppressed IL-17a but not RORγt expression (Fig. S2). Suppression of IL-17 requires the Gfi-1 suppressive domain, because Gfi-1 bearing a single mutation in this domain (Gfi-1(P2A)) failed to suppress IL-17 production (Fig. 4 C).

The failure of Gfi-1 to suppress RORγt transcription in this setting is apparently at odds with the genomic data shown in Fig. 1, in which the Rorc locus is not H3K4 trimethylated in wild-type Th2 cells but is trimethylated in Gfi1 cKO Th2 cells. However, this discrepancy may be explained by the fact that retroviral-mediated expression of Gfi-1 does not occur until 24–48 h after infection, whereas RORγt up-regulation by TGF-β occurred within 24 h (unpublished data). Alternatively, Gfi-1 may require another Th2 factor as a cosuppressor of RORγt transcription. Nevertheless, the data indicate that Gfi-1 suppresses IL-17 even when RORγt is normally expressed.

To test the direct binding of Gfi-1 to its putative target genes, ChIP using anti–Gfi-1 antibody was performed. The Gfi-1 DNA binding sites contain AATC, and sequences containing AAATCNNNG are represented at the highest frequency in a random oligonucleotide selection assay, suggesting that such sequences may have particularly high binding affinity for Gfi-1 (51). Eight putative Gfi-1 binding sites containing AAATCNNNG or its complement CNNNGATTT, including the Rorc promoter (Rorcp) from the Rorc locus and nine sites from the Il17a/Il17f loci along with Gfi1 promoter (Gfi1p) as a positive control (52), were tested for Gfi-1 binding. Only the Gfi1p region near the GAAAATCGAAGTA sequence and an intergenic region of Il17a/Il17f near the AGAAATCTCAGGG sequence but not the other sites tested were enriched by the anti–Gfi-1 antibody in this assay (Fig. 4 D, left). Consistent with Gfi-1 binding, this Il17a/Il17f intergenic region also recruited LSD1 in wild-type cells, and the binding of LSD1 to this site was not found in the absence of Gfi-1 (Fig. 4 D, right).

When CD4 T cells from Gfi1 cKO mice were primed under Th17 conditions, a higher percentage of the Gfi1 cKO cells than of wild-type cells produced IL-17 (28.4 vs. 14.5%, Fig. 4 E, left). In a series of nine independent experiments, there was a highly statistically significant increase in IL-17–producing cells among Gfi1 cKO Th17 cells compared with wild-type Th17 cells (Fig. 4 E, right). These data indicate that endogenous expression of Gfi-1 in Th17 cells, even at low levels, is sufficient to modulate IL-17 production. Indeed, because Gfi-1 can be induced transiently by TCR stimulation even in Th17 cells (unpublished data), such induction may serve as a negative regulator of IL-17 production.

Because Gfi-1 is also underexpressed in iTreg cells, we tested the effect of enforcing Gfi-1 expression during iTreg cell differentiation. Cells primed under iTreg cell conditions were infected with Gfi-1–GFP-RV or control GFP-RV. 4 d after infection, cells were costained for Foxp3 and IL-2 to identify bona fide Treg cells. In the group infected with control GFP-RV, the majority of both the GFP⁻ and GFP⁺ cells were Foxp3⁺IL-2⁻ (53.7 and 61.7%; Fig. 5 A). Although 69.2% of GFP⁻ cells from the Gfi-1–GFP-RV–infected group were Foxp3⁺IL-2⁻, only 34.9% of GFP⁺ cells were Foxp3⁺IL-2⁻. Thus, enforced expression of Gfi-1 significantly reduced the generation of iTreg cells. Gfi-1 had a minimal effect on the frequency of Foxp3⁺ cells induced by TGF-β, but among the Foxp3⁺ cells, a higher percentage of cells also produced IL-2. In a separate experiment, enforced Gfi-1 expression during iTreg cell differentiation enhanced the proportion of IL-2–producing cells from ∼20 to ∼35%, whereas Gfi-1(P2A) had no effect (Fig. 5 B), suggesting that the repressive domain of the Gfi-1 is required for its function in regulating iTreg cell generation.

When naive Gfi1 cKO CD4 cells were primed under iTreg cell conditions, 25.6% of the cells became Foxp3⁺CD25^hi cells, similar to wild-type cells primed under same conditions (22%; Fig. 5 C, top). However, 22.5% of the cells from the Gfi1 cKO culture were also Foxp3⁺CD103⁺, whereas only 12.3% of the wild-type cells displayed such a phenotype (Fig. 5 C, bottom). In addition, 93.3% of the Foxp3⁺ cells from the wild-type culture were CD62L^hi, whereas only 63.5% of the Foxp3⁺ Gfi1 cKO cells expressed CD62L (unpublished data). The more abundant presence of CD103⁺CD62L^lo cells in the Gfi1 cKO culture suggests that endogenous Gfi-1 limits differentiation of a population of highly activated Treg cells. Indeed, ex vivo staining of lymph node cells from wild-type and Gfi1 cKO mice showed an apparently different phenotype of Treg cells (Fig. 5 D). Although the total percentage of Foxp3⁺ cells did not differ from the wild type to the KO, there was a substantial difference in Foxp3⁺CD103⁺ cells (6.8% in the Gfi1 cKO and 2% in the wild type; Fig. 5 D, top). A more careful analysis showed that ∼20% of the Foxp3⁺ cells in wild-type mice were CD103⁺, whereas in Gfi1 cKO mice, such CD103⁺ cells represented ∼50% of the Foxp3⁺ Treg cells (Fig. 5 D, bottom). Three putative Gfi-1 binding sites containing AAATCACAG from the Cd103 locus were tested for Gfi-1 binding. The site from Cd103 intron 1 (near the GGAAATCACAGAG sequence), but not the other two sites, was enriched by Gfi-1 antibody by 15–20-fold (Fig. 5 E, left). LSD1 was also recruited to this site in wild-type but not Gfi-1 cKO cells (Fig. 5 E, right). These data suggest that CD103 is a direct target of Gfi-1.

To address the physiological effects of Gfi-1 on TGF-β–mediated responses in vivo, we studied the responses of the cKO mice and normal controls to MOG-induced experimental allergic encephalitis (EAE). As expected, 13–14 d after MOG immunization, wild-type mice began to develop clinical symptoms, with a peak in intensity at 3 wk. The onset of disease in Gfi1 cKO mice was significantly delayed, by approximately 1 wk (Fig. 6 A). IL-17 has been reported as a key player in the induction of the disease (53). IL-17a mRNA in CD4 T cells was measured by real-time PCR before or after various times of immunization without in vitro restimulation. There was an approximately twofold increase of IL-17a mRNA in Gfi1 cKO mice at all time points after MOG immunization (Fig. 6 B). Neither IFN-γ nor IL-4 mRNA was found to be consistently different from the levels in wild-type mice, possibly because of the modest Th1 or Th2 responses of these mice to immunization. CD103 mRNA is higher in Gfi1 cKO CD4 cells before immunization, consistent with the staining data shown in Fig. 5 D, and was further increased after MOG immunization. Ex vivo restimulation of CD4 T cells 1 wk after immunization followed by anti–IL-17 intracellular staining confirmed the enhancement of IL-17 production in Gfi1 cKO mice (Fig. 6 C). However, intracellular staining of IL-17–producing cells 3 wk after immunization revealed no difference between wild-type and Gfi1 cKO CD4 T cells (unpublished data). The discrepancy between IL-17 mRNA and IL-17 protein at 3 wk after immunization may result from the very strong stimulus used for restimulation. Nevertheless, the data certainly indicate that there was no defect in the generation and expansion of Th17 cells in Gfi1 cKO mice. Flow cytometry analysis showed a continuous increase of Foxp3⁺ cells in Gfi1 cKO mice after immunization, which peaked at 3 wk (39.1 and 38.5% in Gfi1 cKO mice compared with 22.8 and 20.5% in wild-type mice; Fig. 6 D). Although CD103⁺ Treg cells were already high in Gfi1 cKO mice before immunization, there was a substantial increase of such cells after immunization (26.1 and 28.9% in Gfi1 cKO mice vs. 6% and 3.7% in wild-type mice), suggesting that Gfi-1 normally plays a negative role in the expansion of Treg cells in vivo. No apparent lymphocyte infiltration into the central nervous system (CNS) was found in healthy mice from either group after immunization. However, among the sick mice, Foxp3⁺ cells were more abundant in the CNS of the Gfi1 cKO mice than of the wild-type mice (Fig. S3 A, available). Most Foxp3⁺ cells in the CNS were also CD103⁺ (Fig. S3 B). Thus, Foxp3⁺CD103⁺ Treg cells appear to have a dominant effect over the IL-17–producing cells in Gfi1 cKO mice resulting in the delay of disease onset.

T cell activation as well as IL-4 stimulation transiently induces Gfi-1 expression, resulting in optimal Th2 cell expansion (37). In the absence of Gfi-1, Th2 expansion is affected both in vitro and in vivo (41). Th1 cells from Gfi1 cKO mice expand normally but make more IFN-γ than WT Th1 cells. In addition, Gfi1 cKO Th2 cells display a different pattern of chromatin histone modification than wild-type Th2 cells. These observations led us to reassess the function of Gfi-1 during T cell differentiation, particularly because two new Th lineages, Th17 and iTreg cells, have been recently defined. In this paper, we showed that Gfi-1 is down-regulated by TGF-β stimulation during Th17 and iTreg cell–inducing conditions, and enforced expression of Gfi-1 inhibits both Th17 and iTreg cell differentiation. Furthermore, in the absence of Gfi-1, the percentage of Foxp3⁺ cells that express CD103 was substantially higher than in wild-type mice, and this subset expanded in vivo upon immunization.

Gfi-1 is a transcriptional repressor (51). Gfi-1 and its congener Gfi-1b have been recently reported to associate with the histone demethylase LSD1 and CoREST to form a repressive complex (46). Gfi-1 directs the complex to the promoter of target genes where LSD1 can demethylate H3K4 residues and thus repress transcription. Knockdown of LSD1 in erythroid cells leads to increased H3K4 dimethylation and trimethylation of Gfi-1b targets. In this paper, we showed that Il17a/Il17f and Cd103 are the targets for Gfi-1 in T cells. Furthermore, LSD1 is recruited to those sites in a Gfi-1–dependent manner.

The Rorc and Il23r promoter regions are heavily methylated on H3K4 residues in Gfi1 cKO Th2 cells but not in WT Th2 cells; methylation levels are comparable to those found in WT Th17 cells. However, we were not able to locate Gfi-1 binding sites within the Rorc or Il23r loci, suggesting that they may not be the direct targets of Gfi-1. A future goal will be to understand the mechanism through which Gfi-1 regulates the chromatin modification at the Rorc and Il23r loci. RORγt is also expressed in the cells cultured under iTreg cell–inducing conditions, correlating with the H3K4 methylation in such cells, although iTreg cells fail to express IL-17 (unpublished data). These results indicate that RORγt can be up-regulated by TGF-β alone (28) and that down-regulation of Gfi-1 by TGF-β is critical for RORγt expression during Th17 cell differentiation.

In a model of EAE involving immunization with MOG, the percentage of IL-17–producing cells found in Gfi1 cKO mice was modestly higher than in wild-type controls. In addition, Toxoplasma gondii infection or Schistosoma mansoni egg injection induced higher levels of IL-17 production in Gfi1 cKO than in wild-type mice (unpublished data). Thus, similar to our in vitro results, the optimal appearance of IL-17–producing cells induced by TGF-β in vivo requires down-regulation of Gfi-1 expression, which is transiently induced by T cell activation and/or IL-4 stimulation.

Gfi-1 expression is also down-regulated by TGF-β during iTreg cell differentiation, and enforced expression of Gfi-1 partially suppresses the induction of Foxp3⁺IL-2⁻ Treg cells. By down-regulating Gfi-1, Treg cells become more active and also display a tissue-seeking phenotype. In Gfi1 cKO mice, the proportion of Foxp3⁺CD103⁺ effector/memory Treg cells is already high, but this population is further increased after MOG immunization. In addition, Foxp3⁺CD103⁺ are also expanded during T. gondii infection or in response to S. mansoni egg injection (Fig. S4, available). We have not yet determined whether the increased percentage of Foxp3⁺CD103⁺ cells in the Gfi1 cKO mice and the expansion of this subset after immunization reflects a unique role for Gfi-1 in the development of Foxp3⁺CD103⁺ thymic-derived Foxp3⁺ T cells, or whether the higher percentage of Foxp3⁺CD103⁺ T cells is secondary to an increased contribution of peripherally generated Foxp3⁺CD103⁺ in the cKO mice that is further potentiated after immunization. In normal mice, the Foxp3⁺CD103⁺ cell population is actively cycling and contains a high proportion of apoptotic cells (7), so it remains possible that Gfi-1 may play a role in the homeostasis of this subset.

A selective effect of Gfi-1 on Treg cells with a modest effect on Th17 cell differentiation in vivo could be explained by the differential requirement of TGF-β in the induction of these two lineages. A low concentration of TGF-β is sufficient for Th17 cell differentiation, whereas iTreg cell differentiation requires a higher concentration of TGF-β (54). In the absence of Gfi-1, Treg cell numbers may increase in response to levels of TGF-β that normally only induce Th17 cell differentiation in wild-type mice after MOG immunization.

The selective expansion of the Foxp3⁺CD103⁺ cells correlates with delayed EAE onset. Treg cells have been suggested to suppress autoimmune diseases by several mechanisms: by suppressing the activation and expansion of the effector cells (6); by suppressing functions of effector cells, such as cytokine production (55); and by limiting recruitment of effector cells to sites of inflammation (56). In our model, the disease onset was delayed in the Gfi1 cKO mice, yet the number of effector cells that are capable of making IL-17 was not only not reduced but was modestly increased. Thus, it is most likely that the Foxp3⁺CD103⁺ cells suppress disease by affecting recruitment and/or activation of the effector cells at inflammation sites consistent with the tissue-seeking phenotype of the Foxp3⁺CD103⁺ T cells that are increased in frequency in Gfi1 cKO mice. The majority of Treg cells found in the CNS are Foxp3⁺CD103⁺, even in wild-type mice. In addition, Treg cells are a much larger proportion of CNS CD4 T cells in Gfi1 cKO than in wild-type mice.

Given the dual effects of Gfi-1 on effector and Treg cells, modulation of its expression and function could have opposite effects depending on the timing, location, and cell type in which it is expressed, just as is the case with the administration of TGF-β or anti–TGF-β (57). However, blockade of Gfi-1 function during early stages of autoimmune diseases may be beneficial through the resultant expansion of Treg cells. Likewise, the specific targeting of Gfi-1 in Treg cells may enhance their migration to sites of inflammation as well as increase their suppressive activity. Thus, the predominant effect of Gfi-1 on Treg cells in vivo suggests Gfi-1 and its pathway of action could be considered as a target for modulating Treg cells.

C57BL/6 mice were obtained from Taconic or the Jackson Laboratory. Gfi1^fl/fl-CD4Cre (41) mice have been backcrossed to C57BL/6 mice for nine generations. G4 mice have been previously described (48). Gfi1^fl/+-CD4Cre-G/4 and Gfi1^fl/fl-CD4Cre-G/4 mice were generated by crossing Gfi1^fl/fl-CD4Cre to G4 mice. All mice were bred and maintained in a National Institute for Allergy and Infectious Diseases (NIAID) specific pathogen-free animal facility, and the experiments were performed when mice were 8–16 wk of age under protocols approved by the NIAID Animal Care and Use Committee.

Lymph node CD4 or CD8 T cells were prepared by MACS purification using CD4 or CD8 microbeads (Miltenyi Biotec), and purity was usually 90–95%. In some experiments, lymph node nonregulatory naive CD4 T cells were prepared by depleting CD8, B220, IA^b, CD24, NK1.1, CD16/CD32, CD11b, and CD25-positive cells using FITC-labeled antibodies (BD), followed by anti-FITC microbeads and autoMACS purification (Miltenyi Biotec). T cell–depleted APCs were prepared by incubating spleen cells with anti-Thy1.2 and rabbit complement (Cedarlane Laboratories Limited) at 37°C for 45 min, and then irradiating them at 30 Gy (3,000 rad). T cells were co-cultured with APCs at a 1:5 ratio in the presence of 1 µg/ml anti-CD3 and 3 µg/ml anti-CD28, together with different combinations of antibodies and cytokines: for Th1 conditions, 10 µg/ml anti–IL-4, 10 ng/ml IL-12, and 100 U/ml IL-2; for Th2 conditions, 10 µg/ml anti–IFN-γ and 10 µg/ml anti–IL-12, 5,000 U/ml IL-4, and 100 U/ml IL-2; for Th17 conditions, 10 µg/ml anti–IL-4, 10 µg/ml anti–IFN-γ, 10 µg/ml anti–IL-12, 5 ng/ml TGF-β, 10 ng/ml IL-6, and 10 ng/ml IL-1β; for iTreg cell conditions, 10 µg/ml anti–IL-4, 10 µg/ml anti–IFN-γ, 10 µg/ml anti–IL-12, 5 ng/ml TGF-β, and 100 U/ml IL-2; for ThN (nonpolarization) conditions, 10 µg/ml anti–IL-4, 10 µg/ml anti–IFN-γ, 10 µg/ml anti–IL-12, and 100 U/ml IL-2; and for neutral conditions, no exogenous cytokines or antibodies were added.

Total RNA was isolated using TRIzol (Invitrogen). First-strand cDNAs were made using the SuperScript Preamplification System (Invitrogen). Quantitative real-time PCR was performed on a sequence detection system (model 7900HT; Applied Biosystems). The sequences of the primers and minor groove binder (MGB) probe for Gfi-1 are 5′-TGGGCGGCGGCTCCTACAAAT-3′, 5′-TGCGGTGTGGAGAACACCTGA-3′, and FAM-5′-TGCAGCAAGGTGTT-3′-MGB, respectively. Inventoried primer/probe sets for detecting IFN-γ (Mm99999071_m1), IL-4 (4312482), IL-17a (Mm00439619_m1), IL-17f (Mm00521423_m1), Foxp3 (Mm00475156_m1), Bcl-2 (Mm00477631_m1), Bim (Mm00437795_m1), SOCS-1 (Mm00782550_s1), CD103 (Mm00434443_m1), CD4 (Mm00442754_m1), TGF-βRI (Mm00436971_m1), TGF-βRII (Mm00436978_m1), IL-6Rα (Mm00439653_m1), GAPDH (4308313), and RORγt (Mm01261022_m1) were purchased from Applied Biosystems.

Gfi-1–GFP-RV and Gfi-1(P2A)–GFP-RV were previously constructed (37). Retroviruses were packaged in the Phoenix-Eco packaging cell line, as previously described (36). Infection was performed at 24 h after the initiation of the culture in which CD4 T cells were activated with anti-CD3/anti-CD28 under neutral conditions in the presence of APCs.

The expression of GFP in retrovirally infected cells was measured by flow cytometry. Cell-surface staining was performed in PBS containing 0.1% PBS with different combination of antibodies. Anti–IFN-γ–FITC, anti–IL-4–PE, anti–IL-17a–PE, anti-CD25–PE, anti-CD25–PerCP–Cy5.5, anti-CD4–PE–Cy5, anti–IFN-γ–allophycocyanin, and anti-CD44–allophycocyanin were purchased from BD. Anti-CD103–FITC (2E7), anti-CD103–PE (2E7), anti-Foxp3–PE (FJK-16s), anti-Foxp3–allophycocyanin (FJK-16s), anti–IL-17a–Alexa Fluor 647 (eBio17B7), anti–IL-2–Pacific blue (JES6-5H4), anti-CD62L–Pacific blue (MEL-14), anti–IFN-γ–PerCP–Cy5.5 (XMG1.2), anti-CD44–allophycocyanin–Alexa Fluor 750 (IM7), and anti-CD4–Alexa Fluor 700 (GK1.5) were purchased from eBioscience. For intracellular cytokine staining, cells were restimulated with 10 ng/ml PMA and 1 µM ionomycin for 4 h or with platebound anti-CD3 and anti-CD28 (3 µg/ml each) for 6 h in the presence of 2 µM monensin during last 3 h. Harvested samples were fixed with 4% formaldehyde, washed, and permeabilized in 0.5% Triton X-100–0.1% BSA in PBS before staining for cytokine production. Intracellular staining of Foxp3 was performed according to the instructions of the Foxp3 staining kit (eBiosciences). When GFP is present, Foxp3 was stained in Triton X-100 buffer, which is used for staining cytokines.

For Gfi-1 and LSD1 ChIP, 5 × 10⁶ CD4⁺ T cells that have been cultured under Th2 or Th17 conditions for 4–5 d were cross-linked with formaldehyde treatment, and the chromatin was sonicated into small fragments from ∼200 to 1,000 bp. The fragmented chromatin was immunoprecipitated with the specific antibody. Anti–Gfi-1 (N-20) was purchased from Santa Cruz Biotechnology, Inc. Anti-LSD1 was purchased from Abcam. Precipitated DNA fragments were reverse cross-linked, and the enrichment of specific sites was tested by real-time PCR using SYBR green and the relative enrichment was normalized to a negative site from Gata3-intron3. The primers are as follows: Rorcp, 5′-ATTCCATGAGGGCTTGCCT-3′ and 5′-ACCTGTCATCAGCCTCCCATA-3′; Gfi1p, 5′-CGAAGTATTCTTGGCTCCTGGA-3′ and 5′-GCCCAAACAGTCGACAGGATT-3′; Il17a/Il17f, 5′-TGCTGGGCAGGAAGATACTC-3′ and 5′-TCCCCACTCTGTCTTTCCAG-3′; Cd103-5′, 5′-CCCACCCGCTGTAGACATGT-3′ and 5′-CCCAGCCATACCCGAAGAT-3′; Cd103-intron1, 5′-CCCACCCAACACGTCCTCTAA-3′ and 5′-CTCATATCTTAATCTAAGCACTCAGCGTT-3′; Cd103-3′, 5′-CCCTGCCTGGACTACAAAGTG-3′ and 5′-CTGCCTCCACCTCCCAAGT-3′; and Gata3-intron3, 5′-TCAGGGCACTAAGGGTTGTTAACTT-3′ and 5′-GAATTCCATCCATGAGACACACAA-3′.

For ChIP-Seq experiments (47), 10⁷ CD4⁺ T cells that have been cultured under Th2 or Th17 conditions for 4 d were cross-linked with formaldehyde treatment and digested with MNase to generate mainly mononucleosomes with a minor fraction of dinucleosomes. ChIP was performed by using anti-H3K4me3 (Abcam). The ChIP DNA ends were repaired using polynucleotide kinase and Klenow enzyme, followed by treatment with Taq polymerase to generate a protruding 3′ A base used for adaptor ligation. After ligation of a pair of Solexa adaptors to the repaired ends, the ChIP DNA was amplified using the adaptor primers for 17 cycles, and the fragments around 220 bp (mononucleosome plus adaptors) were isolated from agarose gel. The purified DNA was used directly for cluster generation and sequencing analysis using the Illumina/Solexa 1G Genome Analyzer according to the manufacturer's protocols. Sequence reads were obtained and mapped to the mouse genome with the Illumina GA Pipeline. Read positions were adjusted 75 bp to be centered on the sequenced chromatin fragments, and the number of adjusted reads originating in 200-bp windows were summed and displayed as custom tracks on the University of California Santa Cruz Genome Browser.

EAE was induced via subcutaneous immunization into the hind flank with 200 µl of an emulsion containing 400 µg of MOG35-55 peptide (MEVGWYRSPFSRVVHLYRNGK) and 400 µg of Mycobacterium tuberculosis strain H37Ra in complete Freund's adjuvant (Difco). On days 0 and 2, the mice received an intraperitoneal injection of 200 ng pertussis toxin (EMD) dissolved in PBS. Clinical symptoms were assessed by the following criteria: 0, no signs of disease; 1, complete tail paralysis; 2, hind limb paresis; 3, complete hind limb paralysis; 4, unilateral forelimb paralysis; and 5, moribund/death. Mice observed in a moribund state were killed. Data presented are the mean clinical scores of five mice per group.

Fig. S1 shows the expression of RORγt, TGF-βRI, TGF-RβRII, and IL-6Rα in Th1, Th2, and Th17 cells and in Gfi-1–RV–infected Th17 cells. Fig. S2 shows the suppression of IL-17 expression by Gfi-1 in Gfi1 cKO Th17 cells. Fig. S3 shows the frequency and the phenotype of Treg cells found in the CNS of the WT and Gfi1 cKO mice after MOG immunization. Fig. S4 shows the expansion of Foxp3⁺CD103⁺ Treg cells in Gfi1 cKO mice during T. gondii infection or in response to S. mansoni egg injection.

We thank J. Hu-Li for her technical support and S. Tanksley for cell sorting. We thank Dr. T.-Y. Roh for assistance with the ChIP-Seq analysis. We thank Dr. M.O. Li for helpful discussions.

This work is supported by the Division of Intramural Research, NIAID and the Division of Intramural Research, National Heart, Lung, and Blood Institute, National Institutes of Health.

The authors have no conflicting financial interests.