Inhibitory role of the transcription repressor Gfi1 in the generation of thymus-derived regulatory T cells

Significance

Thymic development of regulatory T cells (T_reg) is a crucial event for immune homeostasis. Despite extensive descriptions of molecular pathways functioning within T_reg cells, little is known how T_reg development is shaped by the control of extrinsic factors. Here, we show that the transcription factor growth factor independent 1 (Gfi1) restrains thymic generation of T_reg cells. Loss of Gfi1 led to aberrant production of the cytokine IL-2. Elevated IL-2 then drove the expansion of T_reg cells. Accordingly, deficiency of Gfi1 dampened antitumor immunity. Our results establish that Gfi1 actively controls thymocyte IL-2 production, thereby shaping thymic generation of T_reg cells.

Abstract

Foxp3⁺ regulatory T (T_reg) cells are essential for the maintenance of self-tolerance and immune homeostasis. The majority of T_reg cells is generated in the thymus as a specific subset of CD4⁺ T cells, known as thymus-derived or natural T_reg (nT_reg) cells, in response to signals from T-cell receptors, costimulatory molecules, and cytokines. Recent studies have identified intracellular signaling and transcriptional pathways that link these signals to Foxp3 induction, but how the production of these extrinsic factors is controlled remains poorly understood. Here, we report that the transcription repressor growth factor independent 1 (Gfi1) has a key inhibitory role in the generation of nT_reg cells by a noncell-autonomous mechanism. T cell-specific deletion of Gfi1 results in aberrant expansion of thymic nT_reg cells and increased production of cytokines. In particular, IL-2 overproduction plays an important role in driving the expansion of nT_reg cells. In contrast, although Gfi1 deficiency elevated thymocyte apoptosis, Gfi1 repressed nT_reg generation independently of its prosurvival effect. Consistent with an inhibitory role of Gfi1 in this process, loss of Gfi1 dampens antitumor immunity. These data point to a previously unrecognized extrinsic control mechanism that negatively shapes thymic generation of nT_reg cells.

Normal development of Foxp3⁺ regulatory T (T_reg) cells is critical for maintaining self-tolerance and preventing exuberant immune responses (1). T_reg cells are produced mainly in the thymus, known as thymus-derived or natural T_reg (nT_reg) cells, and they require expression of the transcription factor Foxp3. T-cell receptor (TCR) specificity to self-antigens seems to be a primary determinant for nT_reg lineage commitment in the thymus, with c-Rel being an important factor that links TCR engagement and Foxp3 expression (2, 3). Costimulatory factors (such as CD28) and cytokines, predominantly IL-2, also play crucial roles for the induction of Foxp3 and thymic development of nT_reg cells (2, 3). In a two-step model of nT_reg development, TCR engagement leads to the expression of the high-affinity IL-2Rα that subsequently responds to IL-2 stimulation for the induction of Foxp3 expression and nT_reg lineage commitment (4, 5). However, the cellular source of IL-2 is unclear (6). Moreover, whereas much emphasis has been placed on T cell-intrinsic control of nT_reg development, how the production of these extrinsic factors is controlled to shape the nT_reg pool remains poorly understood.

Growth factor independent 1 (Gfi1), a transcription repressor, has emerged as an important regulator of hematopoietic and immune system cells. Gfi1 is required for the normal development and homeostasis of hematopoietic stem cells and both myeloid and lymphoid progenitors (7, 8). Specifically, loss of Gfi1 impairs the development of neutrophils and B cells while expanding the monocyte and myeloid populations (9–11). In the T-cell lineage, Gfi1 expression is dynamically regulated (12), and its deficiency diminishes double-negative (DN) cell generation but increases the differentiation of CD8⁺ T cells in the thymus (13). In the periphery, Gfi1 has been implicated in the differentiation and in vivo function of CD4⁺ effector and regulatory T-cell subsets (14–18), but it is dispensable for CD8⁺ T cell-mediated immune responses in vivo (16). These results indicate an important but cell context-dependent function for Gfi1 in the immune system.

Whereas a role for Gfi1 in early thymocytes and peripheral T cells has been described, its function in the development of nT_reg cells is unclear. We have previously found that thymic development of nT_reg cells is orchestrated by S1P₁ (19), which is under the control of Klf2 (20) that can be further regulated by Gfi1 (13), but the roles of Gfi1 in nT_reg cells are poorly understood. Therefore, we generated T cell-specific Gfi1-deficient mice and had a surprising finding that Gfi1 deletion enhanced nT_reg development through a noncell-autonomous mechanism. Additional analysis revealed an exuberant production of IL-2 by Gfi1-deficient thymocytes as the main mechanism, thereby highlighting a previously unrecognized mechanism in which IL-2 produced by conventional T cells shapes thymic microenvironment to direct nT_reg development. Furthermore, Gfi1 function in T cells was required for optimal antitumor immunity, consistent with its effects at inhibiting nT_reg generation and function. Finally, although Gfi1 deficiency increased thymocyte apoptosis, Gfi1 repressed generation of nT_reg cells independently of its prosurvival effect. These data point to an extrinsic control mechanism that negatively shapes thymic generation of nT_reg cells.

Results

Gfi1 Deficiency Promotes the Generation of nT_reg Cells.

To investigate the function of Gfi1 in T-cell development, we first analyzed mice with germ-line deletion of Gfi1 (Gfi1^−/− mice) (12). Loss of Gfi1 altered thymic populations, most notably a higher frequency of CD8 single-positive (CD8SP) cells, consistent with a previous report using Gfi1^−/− mice (13). Detailed analysis of T-cell subsets indicated that the Gfi1^−/− CD4 single-positive (CD4SP) compartment contained a markedly increased frequency of thymic Foxp3⁺ nT_reg population (Fig. S1A). Given a prominent role of Gfi1 in the development of multiple hematopoietic lineages (8, 21), it remained possible that the expansion of Gfi1^−/− nT_reg cells was secondary to defective early development and/or influences by nonlymphoid cells. Hence, we crossed loxP-flanked Gfi1 alleles (Gfi1^fl/fl) with two Cre lines, the hCD2-iCre and CD4-Cre transgenic mice, in which Cre expression was initiated at the early and late DN stages, respectively (called Gfi1^hCD2 and Gfi1^CD4) (22, 23). Gfi1 was efficiently deleted in double-positive (DP) and single-positive (SP) cells from both lines (Fig. S1B). More efficient recombination was achieved in Gfi1^hCD2 mice, possibly because of the codon improvement of the Cre recombinase in hCD2-iCre mice (22). As expected, deletion of Gfi1 in both systems resulted in an elevated ratio of CD8SP/CD4SP T cells (Fig. 1 A and B). More importantly, loss of Gfi1 considerably up-regulated the frequency of nT_reg cells among CD4SP T cells (Fig. 1 A and B), although the absolute number of nT_reg cells remained largely unaltered because of the reduction of total thymocytes and especially, CD4SP cells (Fig. S1C). Greater proportions of Foxp3⁺ cells among CD4 T cells were also observed in the spleen (Fig. 1C and Fig. S1D), liver, blood, and even DP thymocytes from Gfi1^CD4 mice (Fig. S1E). Comparable results derived from these different deletion systems indicate a specific role for Gfi1 in restraining nT_reg generation. For the purpose of clarity and consistency, we mainly used Gfi1^CD4 mice in our analyses.

Fig. 1.

Gfi1 deficiency promotes the generation of nT_reg cells. (A and B) Flow cytometry analysis of thymic populations in (A) WT controls and Gfi1^hCD2 or (B) Gfi1^CD4 mice for expression of CD4 and CD8 and frequency of Foxp3⁺ cells among gated CD4⁺TCR⁺ thymocytes (CD4SP). (Right) Ratio of CD8SP/CD4SP T cells (Upper) and frequency of Foxp3⁺ cells (Lower). (C) Frequency of Foxp3⁺ cells in gated CD4⁺ splenocytes of WT and Gfi1^CD4 mice and ratio of CD8⁺/CD4⁺ cells (Upper Right). (D) BrdU staining in thymocytes of WT and Gfi1^CD4 mice at 18 h after injection with BrdU. (E) Annexin V staining (Left) and caspase activity (Right; by FITC-VAD-FMK (carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone) binding) in CD4SP thymocytes. Data are representative of at least five independent experiments. Error bars indicate SEM. *P < 0.05; **P < 0.01.

We determined the cellular mechanisms by which Gfi1 restrains nT_reg development. We reasoned that the increased frequency of nT_reg cells in Gfi1-deficient mice could be caused by differential regulation of cell expansion or survival between Foxp3⁺ and Foxp3⁻ CD4SP T cells. To assess the effects of Gfi1 deficiency on cell proliferation, we first used BrdU incorporation assays, which label cells that have synthesized new DNA and progressed into the S phase of the cell cycle. Whereas Foxp3^– CD4SP T cells from wild-type (WT) and Gfi1^CD4 mice incorporated BrdU to a comparable extent, a greater percentage of Foxp3⁺ T cells from Gfi1^CD4 mice incorporated BrdU (Fig. 1D). Similar results were obtained from splenic Foxp3⁺ T cells from Gfi1^CD4 mice (Fig. S1F). Moreover, K_i-67 staining, which denotes proliferating cells at all active phases of the cell cycle, also revealed an increased frequency of proliferating Foxp3⁺ cells in the thymus and spleen of Gfi1^CD4 mice (Fig. S1G). We next analyzed apoptosis of different populations of CD4SP cells. Freshly isolated CD4SP T cells from WT and Gfi1^CD4 mice showed a similar degree of Annexin V binding and caspase activity (two hallmarks of apoptotic cell death) in both the CD25⁻ and CD25⁺ compartments (Fig. 1E). Therefore, the increased nT_reg population in Gfi1^CD4 mice is associated with an enhanced rate of cell proliferation but not survival.

Gfi1 Deficiency Enhances nT_reg Function and Impairs Antitumor Immunity.

Having shown a role of Gfi1 in nT_reg development, we next examined whether Gfi1 affected the immunosuppressive function of nT_reg cells. Gfi1-deficient nT_reg cells exhibited a stronger inhibitory capacity against the proliferation of conventional T cells in standard in vitro T_reg suppression assays based on carboxyfluorescein succinimidyl ester (CFSE) dilution and thymidine incorporation (Fig. 2A and Fig. S2A). We then examined the expression of the integrin molecule CD103 and the ectonucleotidase CD73, effector molecules associated with nT_reg cell suppressive activity (24, 25). Whereas Gfi1 has been shown to regulate these molecules in discrete contexts (17, 26), the extent to which this regulation is operative in nT_reg cells is poorly defined. Compared with the WT counterparts, deletion of Gfi1 resulted in prominent up-regulation of CD103 (Fig. 2B) and CD73 (Fig. 2C) on nT_reg cells from both the thymus and periphery. Thus, Gfi1 negatively controls nT_reg function, which is associated with its effects to down-modulate expression of selective suppressive molecules (24, 25).

Fig. 2.

Gfi1 restrains nT_reg function and is required for antitumor immunity. (A) CFSE levels in CFSE-labeled effector T (T_eff) cells activated in the presence of nT_reg cells for 3 d. (B and C) Expression of (B) CD103 or (C) CD73 among Foxp3⁺ cells. (D–G) WT and Gfi1^CD4 mice were inoculated with B16 melanoma cells, and (D) tumor size was measured. Tumor-infiltrating lymphocytes were isolated at day 16 and analyzed for Foxp3 staining; shown are (E) Foxp3⁺ numbers, (F) the ratio of Foxp3⁺/Foxp3^– among CD4⁺ cells, and (G) mean fluorescent intensity (MFI) of CD103 and CD73 among Foxp3⁺ tumor-infiltrating lymphocytes. Data are representative of (A and D–G) two and (B and C) five independent experiments. Error bars indicate SEM. *P < 0.05.

Immune tolerance mediated by nT_reg cells is a major component that prevents productive immune responses against tumors (27–29). Given the inhibitory effects of Gfi1 on the nT_reg pool and suppressive activity, we tested whether lack of Gfi1 affected antitumor immunity using a well-established B16 melanoma model. Gfi1^CD4 mice failed to restrain the late phase of tumor growth (Fig. 2D), despite slightly increased proportions of IFN-γ⁺ cells among tumor-infiltrating CD4⁺ and CD8⁺ cells (Fig. S2B). Associated with impaired antitumor immunity in Gfi1^CD4 mice was an increased number of Foxp3⁺ cells in tumor-infiltrating cells (Fig. 2E) and a greater Foxp3⁺/Foxp3^– ratio (Fig. 2F). Furthermore, CD103 and CD73 expression levels on tumor-infiltrating nT_reg cells were elevated in Gfi1^CD4 mice (Fig. 2G). These results indicate that Gfi1 is required for optimal antitumor immunity, likely through its inhibitory effects on T_reg generation and function.

Noncell-Autonomous Effects of Gfi1 Deficiency on nT_reg Development.

We next investigated whether the increased thymic nT_reg population in Gfi1-deficient mice was a cell-autonomous defect. To this end, we generated mixed bone marrow (BM) chimeras by reconstituting alymphoid Rag1^−/− mice with a 1:1 mixture of WT or Gfi1^CD4 (CD45.2⁺; donor) and WT (CD45.1⁺; spike) BM cells. As a comparison, we also generated chimeras that contained WT or Gfi1^CD4 BM cells exclusively and observed the expected increase of the thymic and splenic nT_reg population in reconstituted chimeras derived from Gfi1-deficient BM cells (Fig. S3A). In the mixed chimeras, Gfi1^CD4-derived thymocytes retained the higher CD8SP/CD4SP ratio compared with WT- or spike-derived cells in the mixed chimeras (Fig. 3A), indicative of a cell-autonomous effect of Gfi1 deficiency on CD8⁺ differentiation.

Fig. 3.

Gfi1 deficiency affects nT_reg generation by a noncell-autonomous mechanism. (A and B) Mixed BM chimeras were generated by reconstituting sublethally irradiated Rag1^−/− mice with a 1:1 mixture of spike (CD45.1⁺) and WT or Gfi1^CD4 (CD45.2⁺) BM cells and analyzed later for (A) CD8/CD4 ratios among thymocytes and (B) frequency of Foxp3⁺ cells among CD4SP thymocytes. (C) Linear regression analyses of fold increases of Foxp3⁺ frequency among CD4SP thymocytes in Gfi1^CD4 mice relative to the ages of experimental mice. Each dot represents one independent experiment. (D) Gfi1 deletion efficiency in Gfi1^Foxp3 thymic or splenic nT_reg cells. (E) Frequency of Foxp3⁺ cells among CD4SP thymocytes or CD4⁺ splenocytes from WT and Gfi1^Foxp3 mice. Data are representative of at least (A–C) five and (D and E) three independent experiments. Error bars indicate SEM. **P < 0.01; NS, not significant.

In contrast, the frequency of thymic and splenic Foxp3⁺ cells derived from Gfi1^CD4 donor cells in the mixed BM chimeras was indistinguishable from the frequency of Foxp3⁺ cells from WT donors or the cotransferred CD45.1⁺ spike cells (Fig. 3B and Fig. S3B). Thus, the increased nT_reg population in the unmanipulated Gfi1^CD4 mice seemed to be a noncell-autonomous or bystander defect. However, the up-regulation of CD103 expression among nT_reg cells persisted in Gfi1-deficient donor cells (Fig. S3C), and this cell-intrinsic effect on CD103 expression is in agreement with previous studies identifying CD103 as a direct Gfi1 target gene (17). To further investigate the possibility of a bystander effect on the nT_reg population caused by Gfi1-deficient cells, we set up mixed BM chimeras, in which WT or Gfi1^CD4 (CD45.2⁺) progenitors were a majority and WT spike (CD45.1.2⁺) cells were a minority (at a 20:1 ratio). As expected, the Foxp3⁺ population was increased in Gfi1-deficient cells, but more strikingly, the nT_reg percentage derived from spike thymocytes was also elevated in the predominant Gfi1-deficient thymus, even with a greater degree that probably reflected a better survival fitness of these cells (see below) (Fig. S3D). Similar observations were obtained in the spleen (Fig. S3D). The up-regulation of Foxp3 in WT spike cells indicates that Gfi1 deficiency in T cells exerts a cell-extrinsic effect on the nT_reg population.

We noted that an analogous bystander up-regulation of thymic chemokine receptor expression in Klf2-deficient mice results in an age-related accumulative effect (30). We, therefore, examined whether Gfi1 deficiency affected the kinetics of nT_reg generation by analyzing thymic nT_reg cells in mice of different ages. Linear regression analysis indicated that the enhancing effect of Gfi1 deficiency on nT_reg cells had a significant positive correlation (P = 0.0001) with the age of mice (Fig. 3C). The age-dependent accumulation of nTreg cells likely reflects an extended exposure to nTreg-stimulating factors that are differentially produced by Gfi1-deficient T cells.

To further dissect the noncell-autonomous effects of Gfi1 on nT_reg development, we crossed Gfi1^fl/fl mice with Foxp3^YFP-Cre mice, in which the Cre recombinase was knocked into the endogenous Foxp3 locus (31), to delete Gfi1 selectively in nT_reg cells (called Gfi1^Foxp3 mice). Gfi1 expression was considerably diminished in Foxp3⁺ nT_reg cells in the thymus, and complete deletion was achieved in the periphery (Fig. 3D), which was in line with the timing of Cre expression in Foxp3^YFP-Cre mice (31). Consistent with a direct role of Gfi1 in CD103 expression, CD103 was up-regulated in Gfi1-deficient nT_reg cells, especially in splenic nT_reg cells that had more efficient Gfi1 deletion (Fig. S3E). Accordingly, nT_reg cells from Gfi1^Foxp3 mice had a small increase of suppressive activity (Fig. S3F). However, the specific deletion of Gfi1 in nT_reg cells did not affect the frequency of the Foxp3⁺ nT_reg population in the thymus or spleen (Fig. 3E). Taken together, we conclude that the nT_reg cell expansion induced by Gfi1 deficiency was caused by a noncell-autonomous mechanism. This expansion was a unique effect, because altered CD8SP/CD4SP ratios and CD103 regulation in nT_reg cells were intrinsic to the loss of Gfi1 function. These findings further highlight that deletion of Gfi1 in total T cells altered the thymic environment important for nT_reg generation, and the underlying mechanism was further explored in the following studies.

Gfi1 Affects nT_reg Development by Modulating IL-2 Production.

Thymic nT_reg development is dependent on TCR, costimulation, and cytokine signaling (2, 3). A noncell-autonomous effect mediated by Gfi1 on nT_reg development prompted us to test whether Gfi1 deficiency altered cytokine production. We first focused on IL-2, a key cytokine capable of driving nT_reg development and expansion (32–34). However, the cellular source responsible for IL-2 production in this process remains poorly defined (6), because IL-2 can be produced by dendritic cells and additional cells other than T cells (35). Compared with WT cells, CD4SP and CD8SP T cells from Gfi1^CD4 mice expressed a higher level of Il2 mRNA under resting conditions as well as after phorbol 12-myristate 13-acetate (PMA) and ionomycin stimulation (Fig. 4A). The increased Il2 was restricted to the non-T_reg population of Gfi1-deficient CD4SP cells, because IL-2 expression in the mutant nT_reg cells was unaltered (Fig. S4A). Also, splenic CD4⁺ cells or thymic dendritic cells of Gfi1^CD4 mice showed normal expression of IL-2 (Fig. S4 B and C). Consistent with mRNA analysis, intracellular cytokine staining revealed an increased proportion of Gfi1-deficient CD4SP T cells capable of producing IL-2 (Fig. 4B). A similar trend of increase was noticed in Gfi1-deficient CD8SP cells, although this increase did not reach statistical significance in the limited number of mice analyzed (Fig. 4B).

Fig. 4.

Enhanced IL-2 production by Gfi1^CD4 thymocytes contributes to the expansion of nT_reg cells. (A) Il2 mRNA expression in resting WT and Gfi1^CD4 CD4SP or CD8SP thymocytes or those cells stimulated with PMA and ionomycin (P+I) for 4–5 h. (B) Intracellular IL-2 production by thymic CD4SP or CD8SP cells of WT and Gfi1^CD4 mice after PMA and ionomycin stimulation for 4–5 h in the presence of monensin. (C) Frequency of Foxp3⁺ cells among CD4SP thymocytes from WT, Gfi1^CD4, Il2^−/−, and Gfi1^CD4 Il2^−/− mice (6 wk). (D) Analysis of Foxp3⁺ cells after in vivo neutralization of IL-2 signaling. WT and Gfi1^CD4 mice were injected i.v. with neutralizing anti–IL-2 (S4B6) or anti–IL-2R (PC61) antibody and analyzed 2 wk later. (E and F) Expression of (E) Tgfb1 and (F) Il10 in WT and Gfi1^CD4 thymocyte populations with or without P+I stimulation (Il10 was nondetectable in CD8SP cells). Data are representative of (A, B, and D–F) four and (C) three independent experiments. Error bars indicate SEM. *P < 0.05; **P < 0.01; NS, not significant.

We then determined whether augmented IL-2 production from Gfi1-deficient thymocytes contributed to the nT_reg development phenotypes. Specifically, we used two approaches to block IL-2 function. First, we crossed Gfi1^CD4 mice with Il2^−/− mice to generate Gfi1^CD4 Il2^−/− double KO mice. To avoid the confounding effects caused by an autoimmune syndrome manifested in adult Il2^−/− mice (34), we analyzed these mice at a young age (5–6 wk). Deletion of IL-2 blocked the higher frequency of Foxp3⁺ cells in the thymus of Gfi1^CD4 mice (Fig. 4C). Second, we treated WT and Gfi1^CD4 mice with blocking antibodies for IL-2 or IL-2R (36). After administration of either neutralizing antibody, the frequency of Foxp3⁺ cells in the thymus between WT and Gfi1^CD4 mice became comparable, whereas the difference remained evident in PBS-treated control groups (Fig. 4D). Thus, an important mechanism by which Gfi1 deficiency promotes nT_reg development is to induce excessive IL-2 production.

Aside from IL-2, the immunosuppressive cytokines TGF-β1 and IL-10 have been implicated in promoting Foxp3⁺ cell generation under different conditions (37–40). Whereas TGF-β1 expression was comparable between WT and Gfi1^CD4 cells (Fig. 4E), increased IL-10 expression was observed in Gfi1^CD4 CD4SP cells (Fig. 4F). To test the contribution of the elevated IL-10 expression to the phenotypic alteration of nT_reg cells, we generated Gfi1^CD4 Il10^−/− mice. Deletion of IL-10 had minimal effects on the altered Foxp3⁺ percentage or CD103 expression among nT_reg cells in Gfi1^CD4 mice (Fig. S4 D and E), indicating that the enhanced IL-10 expression by Gfi1-deficient T cells does not play an important role in nT_reg cell generation. Altogether, our results highlight an IL-2–dependent, cell-extrinsic mechanism responsible for the expansion of nT_reg cells in Gfi1^CD4 mice.

Survival-Independent Functions of Gfi1 in Thymic nT_reg Generation.

Gfi1 mutation is frequently associated with elevated apoptosis of multiple cell types (8, 21). Undoubtedly, regulation of apoptosis is a major component of thymocyte developmental processes, including nT_reg generation (2, 3). Although Gfi1 germ-line deletion promoted apoptosis of early DN thymocytes (13), Gfi1 did not seem to be essential for the survival of DP and SP thymocytes under steady state conditions (Fig. 1E). However, because apoptotic cells can be rapidly cleared in vivo, we performed additional analysis to closely examine the effects of Gfi1 loss on thymocyte survival. When cultured for 2 d in vitro, Gfi1^CD4 CD4SP thymocytes showed modestly impaired survival (Fig. 5A). Moreover, in a competitive environment introduced by the mixed chimeras, Gfi1^CD4-derived donor cells had impaired capacity to populate the thymic CD4SP compartment (Fig. 5B). These results reveal a requirement of Gfi1 for the survival fitness of mature thymocytes.

Fig. 5.

Survival defect of Gfi1-deficient thymocytes does not contribute to the altered nT_reg development. (A) CD4SP thymocytes from WT or Gfi1^CD4 mice were cultured for 2 d followed by Annexin V and 7-aminoactinomycin D (7-AAD) staining. (B) Competitive fitness of CD4SP thymocytes in vivo. Mixed BM chimeras were set up as described in Fig. 3A, and the contributions of WT or Gfi1^CD4 donor (CD45.2⁺) and spike (CD45.1⁺) cells were analyzed. (Right) Ratios of CD45.2/CD45.1 (after normalization to nonrelevant B220⁺ cells). (C) CD4SP thymocytes from WT, Gfi1^CD4, Bcl2-TG, or Gfi1^CD4 Bcl2-TG mice were cultured for 3 d followed by Annexin V and 7-AAD staining. (D) Frequency of Foxp3⁺ cells among gated CD4SP cells from WT, Gfi1^CD4, Bcl2-TG, or Gfi1^CD4 Bcl2-TG mice. Data are representative of (A–C) five and (D) three independent experiments. Error bars indicate SEM. **P < 0.01.

To directly address whether this prosurvival effect of Gfi1 is a contributing factor to nT_reg development, we crossed Gfi1^CD4 mice with lymphocyte-specific transgenic mice for Bcl-2 that potently inhibits thymocyte death (41). As expected, the introduction of the Bcl-2 transgene nearly completely rectified the survival defects of Gfi1-deficient thymocytes in vitro (Fig. 5C). However, Gfi1^CD4 Bcl2-TG mice still contained an increased thymic nT_reg population (Fig. 5D). Similarly, deletion of Bim (encoded by Bcl2l11), a proapoptotic molecule important for thymic-negative selection (42), did not affect the increased nT_reg cells in Gfi1^CD4 mice (Fig. S5). Collectively, Gfi1 controls thymic generation of nT_reg cells through a mechanism largely independently of cell survival.

Discussion

Thymic development of nT_reg cells is orchestrated by signals transduced from TCRs, costimulation, and cytokines (2, 3). Although the involvement of nT_reg-intrinsic pathways has been described extensively, how nT_reg development is shaped by extrinsic factors remains poorly understood (43). Additionally, although a role for IL-2 to promote thymic nT_reg development has been appreciated for some time (32–34), the source of IL-2–producing cells remains undefined (6). In this study, we have identified a noncell-autonomous role of Gfi1 in inhibiting nT_reg generation by down-regulating IL-2 expression from conventional thymocytes. Our conclusions are derived from the analysis of mixed BM chimeras and Gfi1^Foxp3 mice, the exuberant production of IL-2 from Gfi1-deficient thymocytes, and the functional consequences of IL-2 blocking. This model is further supported by age-dependent accumulation of thymic nT_reg cells observed in Gfi1^CD4 mice and the increased BrdU incorporation among thymic nT_reg cells lacking Gfi1. Altogether, these data establish a previously unrecognized mechanism for the active control of IL-2 production in the thymic microenvironment to shape nT_reg development.

Our identification of a noncell-autonomous mode of nT_reg cell generation echoes the recent findings describing bystander development of innate CD8⁺ T cells, another nonconventional T-cell subset derived from the thymus. Specifically, thymocytes deficient in transcription factors such as Klf2, Itk, or Id3 aberrantly produce IL-4 that mediates the generation of such cells (30, 44–46). These studies collectively highlight the importance and complexity of transcription factor-dependent shaping of thymic cytokine levels to ensure normal development of various thymus-derived populations.

Using a B16 melanoma tumor model, we showed that Gfi1 function is required for efficient antitumor immunity. The uncontrolled tumor growth in Gfi1^CD4 mice was associated with increased nT_reg cells infiltrating the tumor as well as elevated expression of CD103 and CD73 that endows nT_reg cells with strong suppressive activity (24, 25). Furthermore, nT_reg cells deficient in Gfi1 had enhanced suppressive activity in vitro. These results highlight the important roles of Gfi1 in negatively affecting both nT_reg generation and suppressive activity. Despite the impaired antitumor immunity in Gfi1^CD4 mice, we observed increased IFN-γ production from effector cells lacking Gfi1. These results indicate that the increased frequency and function of Gfi1-deficient nT_reg cells are important factors for the defective in vivo responses, although we could not exclude additional contribution from conventional T cells.

Consistent with a prosurvival role of Gfi1 observed in other cell types (8, 21), deficiency of Gfi1 modestly elevated apoptosis of CD4SP thymocytes and rendered them a competitive disadvantage. However, forced expression of the Bcl2 transgene or deletion of Bim did not affect the generation of nT_reg cells in Gfi1^CD4 mice. In contrast, in hematopoietic stem cells deficient in Gfi1, the survival defect is a central mechanism for impaired stem cell function, because ectopic Bcl-2 expression allows Gfi1-deficient stem cells to self-renew and initiate multilineage differentiation (47). These results reveal that Gfi1-mediated prosurvival function exerts context-specific physiological effects.

In summary, we report here a nonconventional mechanism for the regulation of nT_reg development. The transcription repressor Gfi1 restrains production of IL-2 from conventional T cells, thereby shaping the development of nT_reg cells by a noncell-autonomous mechanism. Loss of Gfi1 results in expansion of nT_reg cells and enhanced suppressive activity that contribute to impaired antitumor immunity. Our studies highlight that T_reg cells are regulated by more than T_reg-intrinsic pathways that sense and integrate signals from TCR and cytokines. Rather, thymic generation of these cells is further shaped by active control of cytokine production from conventional T cells.

Materials and Methods

Mice and in Vivo Studies.

C57BL/6, CD45.1, Rag1^−/−, Il2^−/−, Bcl2l11^−/−, and Il10^−/− mice were purchased from the Jackson Laboratory. CD4-Cre, hCD2-iCre, and Foxp3^Cre-YFP mice were as described (22, 23, 31). Details for the generation of Gfi1^fl/fl mice will be described elsewhere. The LoxP flanked region included coding exons for zinc fingers 1–4, which were essential for DNA binding; homozygosity for the excised allele reproduced the phenotype of the previously published conventional Gfi1-deficient allele (48). All mice have been backcrossed to the C57BL/6 background extensively. Mixed BM chimeras were generated by reconstituting sublethally irradiated (4 Gy) alymphoid Rag1^−/− mice with ∼5 × 10⁶ BM cells depleted of T cells as previously described (49), with the specified ratios of donor cells from various sources distinguished by congenic markers. To label proliferating cells in vivo, mice were injected with 0.8 mg BrdU/mouse i.p., and BrdU incorporation was analyzed 18 h later using the BrdU Staining Kit (BD Biosciences). To block IL-2 signaling, mice were injected i.v. with anti–IL-2 (clone S4B6) or anti–IL-2R (clone PC61) antibody (1 mg/mouse). Animal protocols were approved by the Institutional Animal Care and Use Committee of St. Jude Children's Research Hospital.

Flow Cytometry.

For flow cytometric analyses of surface markers, cells were stained with antibodies (eBioscience) in PBS containing 2% (wt/vol) BSA. Foxp3 and K_i-67 staining was performed per manufacturer’s instructions (eBioscience). For intracellular cytokine staining, T cells were stimulated for 4–5 h with PMA and ionomycin in the presence of monensin before staining per manufacturer’s instructions (BD Biosciences).

Cell Culture.

Lymphocytes were isolated from the thymuses, spleens, and lymph nodes of mice and sorted on a MoFlo (Beckman-Coulter) or Reflection (i-Cyt) cell sorter as described previously (50). SP thymocytes (CD4⁺CD8⁻TCRβ⁺ and CD4^–CD8⁺TCRβ⁺) and peripheral effector T cells (CD4⁺CD45Rb^hiCD25^–) were sorted based on the defined surface markers. T cells were cultured in Click's medium supplemented with 10% (vol/vol) FBS and 1% (vol/vol) penicillin-streptomycin. For nT_reg suppression assays, effector T cells from CD45.1 mice were labeled with CFSE and cocultured with nT_reg cells in the presence of anti-CD3 (145-2C11) for 72 h for the measurement of CFSE dilution of the CD45.1⁺ cells. Alternatively, unlabeled effector T cells were used, and cell proliferation was assayed by ³H-thymidine incorporation assay.

RNA Analysis.

RNA was extracted with an RNeasy Kit (QIAGEN), and cDNA was synthesized with SuperScript III/II Reverse Transcription (Invitrogen). Real-time PCR was performed as described previously using primer and probe sets purchased from Applied Biosystems (50).

B16 Tumor Challenge Model.

Inoculation with B16 melanoma was carried out as previously described with minor modifications (51). Briefly, mice were injected with 1.25 × 10⁵ to 5.0 × 10⁵ B16 cells i.d. in the back. Tumor diameter was measured every 2–3 d with an electronic caliper and reported as area (millimeters squared). To isolate tumor-infiltrating lymphocytes, solid tumors were excised around 15–16 d, and single-cell suspensions were prepared by mechanical dissociation followed by density gradient centrifugation on an 80%/40% (vol/vol) Percoll (GE Healthcare) gradient. The purified cells were then subjected to Foxp3 staining to enumerate T_reg and non-T_reg cells.

Statistical Analysis.

P values were calculated with Student t test or ANOVA. P values < 0.05 were considered significant.