Excessive Th1 responses due to the absence of TGF-β signaling cause autoimmune diabetes and dysregulated Treg cell homeostasis

Abstract

TGF-β signaling in T cells is critical for peripheral T-cell tolerance by regulating effector CD4⁺ T helper (Th) cell differentiation. However, it is still controversial to what extent TGF-β signaling in Foxp3⁺ regulatory T (Treg) cells contributes to immune homeostasis. Here we showed that abrogation of TGF-β signaling in thymic T cells led to rapid type 1 diabetes (T1D) development in NOD mice transgenic for the BDC2.5 T-cell receptor. Disease development in these mice was associated with increased peripheral Th1 cells, whereas Th17 cells and Foxp3⁺ Treg cells were reduced. Blocking of IFN-γ signaling alone completely suppressed diabetes development in these mice, indicating a critical role of Th1 cells in this model. Furthermore, deletion of TGF-β signaling in peripheral effector CD4⁺ T cells, but not Treg cells, also resulted in rapid T1D development, suggesting that conventional CD4⁺ T cells are the main targets of TGF-β to suppress T1D. TGF-β signaling was dispensable for Treg cell function, development, and maintenance, but excessive IFN-γ production due to the absence of TGF-β signaling in naive CD4⁺ T cells indirectly caused dysregulated Treg cell homeostasis. We further showed that T cell–derived TGF-β1 was critical for suppression of Th1 cell differentiation and T1D development. These results indicate that autocrine/paracrine TGF-β signaling in diabetogenic CD4⁺ T cells, but not Treg cells, is essential for controlling T1D development.

On antigenic stimulation, naive CD4⁺ T cells differentiate into distinct functional T-cell subsets, including T-helper 1 (Th1), Th2, Th9, Th17, and follicular helper T cells, characterized by production of different cytokines (1, 2). CD4⁺Foxp3⁺ regulatory T (Treg) cells are another distinct subset of Th cells and play a critical role in immune homeostasis (3, 4). An imbalance of pathogenic T cells and Treg cells is a hallmark feature of a variety of inflammatory diseases, including type 1 diabetes (T1D), an autoimmune disease characterized by destruction of the insulin-producing β cells of the pancreatic islets by autoreactive CD4⁺ and CD8⁺ T cells (5). Although both Th1- and Th17-related cytokines, such as IFN-γ, IL-17, IL-21, and IL-22, have been suggested to play a role in diabetogenesis in nonobese diabetic (NOD) mice and humans (5, 6), their cellular sources, targets, and precise mechanisms of action are complex and remain to be defined.

The differentiation, function, and homeostasis of effector CD4⁺ T cells and Treg cells are thought to be regulated by TGF-β1 (7). Deletion of the TGF-β receptor (TGFβRII) specifically in T cells during early thymic T-cell development by the CD4Cre line (CD4Cre-Tgfbr2^f/f) results in a lethal inflammatory disease (8, 9). In the absence of TGF-β signaling in T cells, naive CD4⁺ T cells differentiate into Th1 and Th2 cells, whereas peripheral Treg cells are markedly reduced (8, 9). Furthermore, a significant but transient reduction in thymic Treg cells in the early neonatal period is also observed (10, 11). TGF-β signaling–deficient T cells are insensitive to Treg cell–mediated suppression. This finding was demonstrated in experimental systems where TGF-β signaling–deficient effector T cells were transferred together with WT Treg cells into lymphopenic hosts (12), or mixed bone marrow chimeras were generated by transferring bone marrow cells from CD4Cre-Tgfbr2^f/f mice and WT mice (8, 9). Although these results imply the T-cell intrinsic requirement of TGF-β to control their self-reactivity, a recent study has reported that ablation of TGF-β signaling in peripheral T cells in adult mice using a distal LckCre line (dLckCre-Tgfbr2^f/f) does not lead to development of autoimmune disease (13). This study suggests that TGF-β signaling prevents lethal autoimmune diseases by controlling the homeostatic proliferation and effector function of self-reactive T cells in the neonatal period that is associated with the lymphopenic state (13). Nonetheless, as an autoimmune disease observed in CD4Cre-Tgfbr2^f/f mice is associated with dysregulated peripheral Treg cell homeostasis (10, 11), and dLckCre-Tgfbr2^f/f Treg cells retain TGFβRII expression (13), it is still controversial to what extent the immunopathology and exacerbated effector T-cell activation in CD4Cre-Tgfbr2^f/f mice is due to the failure of reduced numbers of functional Treg cells.

Although TGF-β1 is produced by multiple cell types, T cells are an essential source of the TGF-β1 required to maintain peripheral tolerance (14, 15). TGF-β1 produced by conventional T cells, but not by Treg cells, is required for downregulating Th1 cell differentiation and colitis development under homeostatic conditions (14). In contrast, during experimental autoimmune encephalomyelitis, TGF-β1 promotes the generation of Th17 cells in the presence of IL-6, which leads to disease development in an autocrine manner (14), indicating that T cell–produced TGF-β1 differentially regulates Th1 and Th17 cell–mediated autoimmune diseases. However, the regulation of CD4⁺ T-cell differentiation by TGF-β1 is context dependent, as TGF-β1 has been shown to suppress IL-22 and GM-CSF production from Th17 cells (16, 17), whereas it promotes Th9 cell differentiation in the presence of IL-4 (18, 19). Thus, the contribution of T cell–produced TGF-β1 in diabetogenic CD4⁺ T-cell differentiation and spontaneous T1D development is still unknown.

In this study, we generated TGFβRII or TGF-β1 conditional KO mice on the NOD genetic background and used the BDC2.5 transgenic mouse model to investigate which T-cell population needs to receive TGF-β signals or whether T cell–produced TGF-β1 is required to control T1D development.

Results

In the Absence of TGF-β Signaling in Thymic T Cells, BDC2.5 NOD Mice Develop Accelerated Diabetes.

To address the role of TGF-β signaling in CD4⁺ T cells during diabetes development, we generated T cell–specific TGFβRII-deficient NOD mice by backcrossing the TGFβRII floxed allele onto the NOD/Lt background and further crossing with CD4Cre NOD mice, which is used to drive Cre expression in CD4⁺CD8⁺ thymic T cells. CD4Cre-Tgfbr2^f/f NOD mice developed severe inflammatory infiltration to multiple organs, and they died at 3–4 wk of age before a hyperglycemic phenotype could be observed in the NOD background, similar to the massive autoimmunity seen on the C57BL/6 genetic background (8, 9). Because the BDC2.5 T-cell receptor (TCR) transgene, which recognize the pancreatic islet antigen chromogranin A, has been shown to be able to rescue the autoimmune phenotype manifested as generalized tissue infiltration (20), we crossed CD4Cre-Tgfbr2^f/f NOD mice onto BDC2.5 NOD mice. As expected, CD4Cre-Tgfbr2^f/f NOD mice carrying the BDC2.5 transgene (BDC-CD4Cre-Tgfbr2^f/f) survived beyond 3–4 wk of age, and all BDC-CD4Cre-Tgfbr2^f/f mice became diabetic between 14 and 21 d of age regardless of sex (Fig. 1A). In contrast, their littermate controls (BDC-CD4Cre-Tgfbr2^f/+ or BDC-Tgfbr2^f/f) showed no signs of diabetes, as is customary for this transgene on the NOD background (20). Massive infiltration of leukocytes into pancreatic islets was also observed in BDC-CD4Cre-Tgfbr2^f/f mice before the onset of diabetes (10–15 d old; Fig. S1A). These results indicate that TGF-β signaling in T cells is essential for suppressing T1D development in BDC2.5 NOD mice.

Fig. 1.

Accelerated diabetes development in the absence of TGF-β signaling in T cells is associated with reduced Treg cell number. (A) Diabetes incidence of BDC-CD4Cre-Tgfbr2^f/f and control littermates (n = 9). (B) Number of total lymphocytes in the spleen and PLNs of BDC-CD4Cre-Tgfbr2^f/f mice or littermate controls at 3 wk old. (C) Percentage and number of CD4⁺TCRVβ4⁺Foxp3⁺ T cells in the spleen and PLNs of the same mice as in B. Each dot represents one mouse. Data are representative of two independent experiments. (D) Expression of Foxp3, CD44, CXCR3, and T-bet on splenic CD4⁺TCRVβ4⁺ cells (Left). (E) Histogram of CD44, CXCR3, and T-bet expression on CD4⁺TCRVβ4⁺Foxp3⁺ cells (Right). Data represent three experiments with two to four mice. *P < 0.05, ***P < 0.001 vs. control. Data are representative of two independent experiments.

Diabetes Development in the Absence of TGF-β Signaling in Thymic T Cells Is Associated with an Altered Treg Cell Phenotype and Increased Th1 Differentiation.

TGF-β1 has been shown to play an important role in the function and maintenance of peripheral Foxp3⁺ Treg cells in vivo (8, 9, 21), and Treg cells are essential for suppression of T1D in BDC2.5 NOD mice (20). The total cellularity of the spleen and pancreatic lymph node (PLN) was reduced in 3-wk-old diabetic BDC-CD4Cre-Tgfbr2^f/f mice compared with control littermates (Fig. 1B). The frequencies and numbers of CD4⁺TCRVβ4⁺ (BDC) Treg cells were significantly reduced in the spleens (Fig. 1C) and pancreas (Fig. S1 B and C) of BDC-CD4Cre-Tgfbr2^f/f mice compared with control mice, although the percentage of Treg cells in the PLN was increased (Fig. 1C). Most of the Foxp3⁻ BDC T cells in the spleens of BDC-CD4Cre-Tgfbr2^f/f mice showed a CD44^hiCXCR3^hiT-bet^hi activated phenotype (Fig. 1D). Interestingly, TGFβRII-deficient Foxp3⁺ Treg cells also expressed higher levels of T-bet and CXCR3 (Fig. 1 D and E), which are mainly expressed in Th1 cells but are also induced in Treg cells under inflammatory conditions (22).

We next determined the expression levels of several cytokines in the pancreas-infiltrating BDC T cells. An increase in Il2, Il21, Ifng, and Csf2 mRNA expression was observed in BDC T cells isolated from the pancreases of diabetic BDC-CD4Cre-Tgfbr2^f/f mice, whereas the expression levels of Il9, Il17a, and Il22 were reduced (Fig. 2A), consistent with the known role of TGF-β signaling in inhibiting Th1 responses while promoting Th9 and Th17 cell differentiation. On the other hand, expression of mRNA for Th2-associated cytokines Il4 and Il13 did not differ between different genotypes (Fig. 2A). Furthermore, the percentages of CD4⁺ T cells secreting IFN-γ and GM-CSF in the spleens and pancreas of diabetic BDC-CD4Cre-Tgfbr2^f/f mice were markedly greater than those of control littermates that did not develop diabetes, whereas IL-17 production was reduced (Fig. 2 B and C; Fig. S1 D and E). Although diabetic BDC-CD4Cre-Tgfbr2^f/f mice had reduced cellularity of the spleen (Fig. 1B), the absolute number of Th1 cells was comparable between different groups due to increased percentage of IFN-γ–producing CD4⁺TCRVβ4⁺ cells (Fig. 2D). Taken together, these results suggest that the decreased Treg cell number and the enhanced differentiation of naive BDC T cells into pathogenic effector Th1 cells, but not Th17 cells, in BDC-CD4Cre-Tgfbr2^f/f mice may contribute to the rapid development of diabetes.

Fig. 2.

Th1 cell–, but not Th17 cell–, associated cytokines are highly expressed in TGF-β signaling deficient diabetogenic CD4⁺ T cells. (A) Pancreas-infiltrating CD4⁺Vβ4TCR⁺ cells were stimulated in vitro with phorbol 12-myristate 13-acetate (PMA) and ionomycin, and mRNA expression of cytokines was determined with real-time PCR. The RNA sample was pooled from four mice for each group. The mRNA expression level in the control cells is defined as 1. Data represent means ± SEM. (B and C) Splenocytes isolated from the indicated mice at 3 wk old were stimulated in vitro with PMA and ionomycin, and expression of IFN-γ, IL-17, or GM-CSF was measured by flow cytometry. Representative plots of the IFN-γ, IL-17, or GM-CSF expression are shown in C. (D) Percentage of CD4⁺TCRVβ4⁺ cells (Left) and number of CD4⁺TCRVβ4⁺IFN-γ⁺ T cells (Right) in the spleen of the same mice as in B. Each dot represents one mouse. **P < 0.01, ***P < 0.001 vs. control. All data are representative of three independent experiments.

IFN-γ, but Not IL-21, IL-22, or GM-CSF, Is Required for Diabetes Development in BDC-CD4Cre-Tgfbr2^f/f Mice.

To examine the role of Th1- and Th17-related cytokines in autoimmune diabetes, we crossed BDC-CD4Cre-Tgfbr2^f/f mice with mice lacking IFN-γ, IL-21R, or IL-22. Similarly to the BDC-CD4Cre-Tgfbr2^f/f mice, all IL-21R– or IL-22–deficient BDC-CD4Cre-Tgfbr2^f/f mice developed diabetes by 21 d of age (Fig. 3A). However, BDC-CD4Cre-Tgfbr2^f/f × Ifng^−/− mice did not develop diabetes at all, and they survived longer than BDC-CD4Cre-Tgfbr2^f/f × Ifng^+/+ mice. When BDC-CD4Cre-Tgfbr2^f/f mice were injected with anti–IFN-γ mAbs, the onset of diabetes was delayed compared with isotype treatment (Fig. 3B). In addition, although BDC T cells isolated from the spleen of BDC-CD4Cre-Tgfbr2^f/f × Ifng^+/+ mice showed an increased percentage of GM-CSF expression, IFN-γ deficiency further increased GM-CSF production but did not affect IL-17 expression (Fig. 3C). Collectively, these results suggest that IFN-γ, but not IL-21, IL-22, or GM-CSF, is responsible for the development of diabetes in BDC-CD4Cre-Tgfbr2^f/f mice.

Fig. 3.

IFN-γ, but not IL-21 or IL-22, is required for the development of diabetes in BDC-CD4Cre-Tgfbr2^f/f mice. (A) Incidence of diabetes in BDC-CD4Cre-Tgfbr2^f/f mice deficient in indicated cytokines at 3 wk of age. (B) Incidence of diabetes in BDC-CD4Cre-Tgfbr2^f/f mice treated with anti–IFN-γ or rat IgG every other day from 12 d old. Data show pooled results from two independent experiments (n = 10–12/group). (C) Proportion of intracellular IFN-γ⁺, IL-17⁺, and GM-CSF⁺ cells among gated splenic CD4⁺Vβ4TCR⁺ cells stimulated with PMA and ionomycin. Data are representative from three independent experiments.

Ablation of the TGFβRII in Effector CD4⁺ T Cells, but Not Foxp3⁺ Treg Cells Alone, Results in Enhanced Th1 Cell Responses Predominantly Within the Pancreas, Leading to Diabetes Development.

To examine to what extent the function of TGF-β is mediated through Treg cell–dependent or –independent cellular responses, we specifically deleted the Tgfbr2 gene in effector T cells and Treg cells or only in Treg cells by crossing BDC-Tgfbr2^f/f mice with OX40Cre (23) or Foxp3Cre mice (24), respectively (Fig. 4 A–C). Interestingly, ablation of TGFβRII in both effector T cells and Treg cells by OX40Cre resulted in diabetes in BDC2.5 NOD mice, although disease onset was slightly delayed compared with BDC-CD4Cre-Tgfbr2^f/f mice (Figs. 1A and 4D). Similarly, 34% of BDC-negative OX40Cre-Tgfbr2^f/f NOD mice also developed spontaneous diabetes by 3 mo of age, and the remaining mice developed wasting colitis regardless of sex, whereas their littermate controls did not (Fig. S2). In contrast, TGFβRII ablation only in Treg cells did not result in diabetes development on the BDC2.5 NOD background (Fig. S3A). Interestingly, although BDC-OX40Cre-Tgfbr2^f/f mice developed rapid diabetes, the frequency of IFN-γ–producing Th1 cells in the spleen was not increased compared with that of control mice (Fig. 4E), which is different from the results observed in BDC-CD4Cre-Tgfbr2^f/f mice (Fig. 2C). Similarly, CD4⁺ T cells from BDC-negative OX40Cre-Tgfbr2^f/f NOD mice produced minimal IFN-γ compared with those of CD4Cre-Tgfbr2^f/f NOD mice (Fig. 4 F and G). However, pancreas-infiltrating cells showed a significantly higher frequency of Th1 cells in BDC-OX40Cre-Tgfbr2^f/f mice compared with control mice (Fig. 4E). These results indicate that TGFβRII deficiency in diabetogenic effector T cells is sufficient to induce T1D, and TGFβRII ablation only in Treg cells does not have a significant effect on effector CD4⁺ T-cell functions.

Fig. 4.

Ablation of the TGFβRII in effector CD4⁺ cells, but not Treg cells, results in dysregulation of Th1 cell responses. (A) TGFβRII expression on CD44^lo naive and CD44^hi memory CD4⁺ T cells isolated from spleens of OX40Cre-Tgfbr2^f/f or Tgfbr2^f/f mice. (B and C) TGFβRII expression on Foxp3⁺ and Foxp3⁻ CD4⁺ T cells isolated from spleens of (B) OX40Cre-Tgfbr2^f/f or (C) Foxp3Cre-Tgfbr2^f/f mice. (D) Diabetes incidence of BDC-OX40Cre-Tgfbr2^f/f and BDC-OX40Cre-Tgfbr2^f/+ or BDC-Tgfbr2^f/f control littermates (n = 16/group). (E) The percentage of IFN-γ⁺CD4⁺TCRVβ4⁺ cells in indicated tissues from 3-wk-old BDC-OX40Cre-Tgfbr2^f/f and control mice stimulated with PMA and ionomycin. (F and G) Percentage of IFN-γ⁺CD4⁺ cells in spleens from (F) OX40Cre-Tgfbr2^f/f and control mice at 2–3 mo of age or (G) CD4Cre-Tgfbr2^f/f and control mice at 3 wk of age. (H) CD4⁺TCRVβ4⁺ effector T cells were sorted from BDC-WT mice. CD4⁺TCRVβ4⁺Foxp3-RFP⁺ Treg cells were isolated from BDC-CD4Cre-Tgfbr2^f/f mice or control littermates. CD4⁺TCRVβ4⁺ effector T cells (2 × 10⁵ cells) were injected into Scid/NOD mice alone or together with CD4⁺TCRVβ4⁺Foxp3-RFP⁺ Treg cells at 2:1 ratio, and diabetes incidence was assessed over time. Data show pooled results from two independent experiments (n = 14–15/group). (E–G) Each dot represents one mouse. ***P < 0.001 vs. control. Data are representative of two (A–C) or three (E and F) independent experiments.

TGFβRII Deficiency in Foxp3⁺ Treg Cells Does Not Impair Treg Cell Function.

BDC-CD4Cre-Tgfbr2^f/f mice had reduced numbers of peripheral Treg cells that also acquired Th1 cell properties (Fig. 1E). To examine whether this phenotype interferes with their suppressive capacity and if accelerated diabetes is also due, at least in part, to functional defects in Treg cells, we performed a Treg cell transfer experiment using Foxp3-red fluorescent protein (RFP) NOD mice (25). When effector BDC T cells were adoptively transferred into Scid/NOD mice, the animals developed rapid diabetes, whereas cotransfer of BDC-CD4Cre-Tgfbr2^f/f Foxp3-RFP⁺ Treg cells with effector BDC T cells prevented disease as efficiently as control BDC Foxp3-RFP⁺ Treg cells (Fig. 4H). Similar results were also obtained when we used Treg cells from BDC-Foxp3Cre-Tgfbr2^f/f mice (Fig. S3B), indicating that TGFβRII deficiency does not impair the suppressive function of Treg cells. Next, we examined the capacity of TGFβRII-deficient effector BDC T cells to induce diabetes. BDC T cells were isolated from BDC-CD4Cre-Tgfbr2^f/f or control mice and were transferred into Scid/NOD mice. By 13 d after transfer into Scid/NOD animals, all of the recipients had developed diabetes regardless of the genotype (Fig. S3C), suggesting that TGF-β signaling is dispensable for suppression of diabetogenic T-cell function in this lymphopenic environment. We next tested the sensitivity of TGFβRII-deficient effector BDC T cells to CD25⁺ BDC Treg cell–mediated suppression. Although CD25⁺ BDC Treg cells were sufficient to inhibit diabetes caused by control effector BDC T cells, they failed to prevent diabetes induced by TGFβRII-deficient effector BDC T cells (Fig. S3C). Collectively, these findings suggest that TGFβRII-deficient effector T cells lose the sensitivity to Treg cell–mediated suppression, whereas TGFβRII-deficient Treg cells are functionally normal.

TGFβRII-Deficient Th1 Cells Indirectly Impair Treg Cell Homeostasis.

Because BDC-CD4Cre-Tgfbr2^f/f mice show enhanced Th1 cell–mediated inflammatory responses, we decided to determine whether ablation of TGF-β signaling impairs peripheral Treg cell homeostasis under limited inflammatory conditions. We found that BDC-FoxpCre-Tgfbr2^f/f mice showed normal Treg cell numbers in the PLN and spleen (Fig. S4A). Similarly, the absolute number of Treg cells in the spleen and PLN of 3-wk-old BDC-OX40Cre-Tgfbr2^f/f mice was also normal (Fig. S4B), which was different from that observed in BDC-CD4Cre-Tgfbr2^f/f mice that showed reduced numbers of peripheral Treg cells (Fig. 1C). Interestingly, the percentage and the number of Treg cells were not reduced but rather increased in the PLN and spleen of BDC-negative OX40Cre-Tgfbr2^f/f or Foxp3Cre-Tgfbr2^f/f NOD mice (Fig. S4 C and D). These expanded Treg cells expressed higher levels of a natural Treg cell marker, Helios (Fig. S4E). Furthermore, depleting TGFβRII via a Foxp3Cre-mediated recombination also led to increased rather than decreased numbers of thymic Treg cells in NOD mice in both a BDC-positive (Fig. S5A) and -negative (Fig. S5B) background. These results suggest that the reduced number of peripheral Treg cells observed in BDC-CD4Cre-Tgfbr2^f/f mice is not due to the abrogation of TGF-β signaling, but may be due to indirect effects of enhanced Th1 cell differentiation and the general systemic inflammation.

To determine the effects of blockade of Th1 responses in peripheral Treg cell homeostasis, we analyzed Treg cell development in BDC-CD4Cre-Tgfbr2^f/f × Ifng^−/− mice. Although the frequency of Foxp3⁺ BDC T cells in the spleen and PLN of 3- to 4-wk-old BDC-CD4Cre-Tgfbr2^f/f × Ifng^−/− mice was the same as that of BDC-CD4Cre-Tgfbr2^f/f mice (Figs. 1C and 5 A and B), the absolute cell number of total lymhpocytes and Treg cells in the spleen of BDC-CD4Cre-Tgfbr2^f/f × Ifng^−/− mice was not reduced, but rather increased in the PLN compared with control mice (Fig. 5B). Furthermore, the increased expression of T-bet and CXCR3 in splenic Treg cells of BDC-CD4Cre-Tgfbr2^f/f × Ifng^+/+ mice was corrected by the deficiency of IFN-γ, whereas the expression in Foxp3⁻CD44^hi cells was not affected (Fig. 5C), indicating that increased IFN-γ production caused by deficiency of TGFβRII is responsible for the reduced peripheral Treg cells and their acquisition of a Th1-like phenotype in BDC-CD4Cre-Tgfbr2^f/f mice.

Fig. 5.

TGFβRII-deficient Th1 cells indirectly impair Treg cell homeostasis. (A) Number of total lymphocytes in the spleen and PLNs of BDC-CD4Cre-Tgfbr2^f/f × Ifng^+/−, BDC-CD4Cre-Tgfbr2^f/f × Ifng^−/−, and control (BDC-CD4Cre-Tgfbr2^f/+ or BDC-Tgfbr2^f/f) mice at 3–4 wk of age. (B) Percentage and absolute number of CD4⁺TCRVβ4⁺Foxp3⁺ cells in the PLNs and spleens of the same mice as in A. (C) Expression of T-bet and CXCR3 on splenic CD4⁺TCRVβ4⁺Foxp3⁺ cells (Right) or CD4⁺TCRVβ4⁺Foxp3⁻CD44^hi cells (Left) from the same mice as in B. Each dot represents one mouse. *P < 0.05, **P < 0.01, ***P < 0.001 vs. BDC control. Data are representative from two (A and B) and three (C) independent experiments.

T Cell–Produced TGF-β1 Is Critical for Protection Against Diabetes Development in BDC2.5 NOD Mice.

Because T cell–derived TGF-β1 has been shown to be essential for controlling T-cell tolerance and differentiation (14, 15), we examined whether T cell–produced TGF-β1 plays an important role in regulating diabetes development. As the mouse Tgfb1 gene is located at 6.5 cM of chromosome 7, only 2.5 cM away from the Idd7 locus (4.0 cM), we generated Tgfb1 conditional KO mice by targeting the NOD chromosome of NOD/129 F₁ ES cells (Fig. S6) (26), and backcrossed these mice to NOD for 10 generations. BDC-CD4Cre-Tgfb1^f/f, but not BDC-Tgfb1^f/f, mice, developed rapid diabetes, although they had a relatively low incidence and slow kinetics of disease progression compared with those of BDC-CD4Cre-Tgfbr2^f/f or BDC-OX40Cre-Tgfbr2^f/f mice (Figs. 1A, 4D, and 6A). The frequency of Th1 cells in the spleen and PLN of BDC-CD4Cre-Tgfb1^f/f mice was significantly higher than that of BDC-Tgfb1^f/f mice (Fig. 6B). Furthermore, the deficiency of T cell–produced TGF-β1 led to a significant increase in the frequency of Foxp3⁺ Treg cells in the thymus, PLNs and spleen and the number of Treg cells in PLN (Fig. 6C). In addition, BDC-negative CD4Cre-Tgfb1^f/f NOD mice had a significantly increased frequency of Th1 cells in the PLNs and spleen, and expansion of thymic and peripheral Foxp3⁺ Treg cells was also observed compared with Tgfb1^f/f NOD mice (Fig. S7). These observations indicate that T cell–produced TGF-β1 is critical for suppression of Th1 cell–mediated diabetes development and inhibiting Treg cell expansion in an autocrine/paracrine manner.

Fig. 6.

T cell–produced TGF-β1 is required for the development of diabetes in BDC2.5 mice. (A) Diabetes incidence of BDC-CD4Cre-Tgfb1^f/f and BDC-Tgfb1^f/f control littermates (n = 20–22). (B) The proportion of intracellular IFN-γ–producing cells among gated CD4⁺Vβ4TCR⁺ T cells was analyzed by flow cytometry. Each dot represents one mouse. (C) Number of total lymphocytes in the thymus, PLNs, and spleen of BDC-CD4Cre-Tgfb1^f/f and BDC-Tgfb1^f/f littermate controls at 2–3 mo of age. (D) Percentage and absolute number of CD4⁺TCRVβ4⁺Foxp3⁺ cells in the thymus, PLNs, and spleen of the same mice as in C. Each dot represents one mouse. (B–D) Data are representative from two independent experiments.

Discussion

Our results demonstrate that TGF-β signaling in conventional CD4⁺ T cells, but not Treg cells, is essential for the prevention of spontaneous T1D in BDC2.5 NOD mice through the blockade of Th1 cell differentiation in an autocrine/paracrine manner. TGF-β signaling is not required for peripheral Treg cell function and maintenance. However, a highly polarized Th1 environment, due to the absence of TGF-β signaling in naive CD4⁺ T cells, indirectly disrupts Treg cell homeostasis.

Published studies have shown that TGF-β signaling is critical for peripheral Treg cell function and homeostasis (8, 9). Although a cell-intrinsic mechanism of TGF-β–mediated control of T-cell reactivity has been suggested (8, 9, 12), a decreased proportion of peripheral Treg cells is also thought to contribute to the pathogenesis of autoimmune disease in TGF-β signaling–deficient mice (11, 27, 28). We have shown here that abrogation of TGF-β signaling in conventional CD4⁺ T cells leads to Th1 cell–dependent rapid T1D development in BDC2.5 NOD mice, whereas TGF-β signaling deficiency does not impair the development, maintenance, and function of Treg cells, indicating that conventional CD4⁺ T cells are the main targets of TGF-β to suppress T1D development. The differences in the frequency and phenotypes of Treg cells observed in CD4Cre, OX40Cre, and Foxp3Cre systems indicate that the magnitude of Th1 cell–mediated inflammatory responses can indirectly influence peripheral Treg cell homeostasis. Indeed, the reduced number of peripheral Treg cells was observed only BDC-CD4Cre-Tgfbr2^f/f mice, which had the highest systemic Th1-mediated inflammation. The number of Treg cells was restored, and their acquisition of a Th1 cell phenotype in BDC-CD4Cre-Tgfbr2^f/f mice was reversed by IFN-γ deficiency. These data suggest that dysregulated peripheral Treg cell homeostasis previously observed in TGF-β1 or T cell–specific TGFβRII KO mice is due, at least in part, to indirect effects of enhanced Th1 cell differentiation and inflammation. In support of this notion, reduced peripheral Treg cell numbers and their acquisition of a Th1 cell phenotype were observed during a highly Th1 cell polarized inflammatory response (22, 29), which resemble the autoimmune diseases observed in BDC-CD4Cre-Tgfbr2^f/f mice in this study. Recent studies have also suggested that Foxp3⁺Treg cells lose Foxp3 expression and differentiate into effector Th cells under inflammatory conditions (30, 31). These exTreg cells can be detected in BDC2.5 NOD and regular NOD mice under steady-state conditions and may contribute to T1D development (31). However, ablation of TGF-β signaling in Foxp3⁺Treg cells and possibly exTreg cells by using Foxp3Cre did not result in T1D development in BDC2.5 NOD mice. This result suggests that TGF-β is not required for the inhibition of Treg to exTreg cell conversion or that TGF-β does not suppress exTregs effector functions in this model.

T cell–specific deletion of TGF-β1 results in expansion of both thymic and peripheral Treg cells (14, 15). The dispensable role of T cell–derived TGF-β1 in Treg cell maintenance suggests that other sources of TGF-β1 may be critical for their homeostasis. However, we found that, in a limited inflammatory condition, ablation of TGF-β signaling in Treg cells also led to expansion of both thymic and peripheral Treg cells, as demonstrated by using OX40Cre-Tgfbr2^f/f and Foxp3Cre-Tgfbr2^f/f mice. These results indicate that, although TGF-β signaling is not required for both thymic and peripheral Treg cell maintenance, the autocrine action of TGF-β1 is critical for limiting Treg cell expansion. Therefore, our results help to reconcile conflicting data regarding the unexpected role of TGF-β1 in controlling Treg cell expansion under steady-state conditions.

We found that BDC2.5 T cells from CD4Cre-Tgfbr2^f/f mice, in which thymic T cells lacked TGFβRII had increased expression of IFN-γ, IL-21, and GM-CSF. The contribution of Th cell subsets to diabetes development seems to be dependent on the disease models studied. In the BDC2.5 T-cell transfer model, IFN-γ, but not IL-17, has been shown to promote disease progression (32–34), although the requirement for these cytokines in mediating spontaneous diabetes development in BDC2.5 NOD mice is not well known. In contrast, the development of spontaneous diabetes in NOD mice is not significantly influenced by the deficiency of IFN-γ (35), IL-17 (36), or IL-22 (Fig. S8), whereas abrogation of IL-21 signaling almost completely suppressed diabetes development (37–39), suggesting that IL-21 is critical mediator to induce spontaneous diabetes in NOD mice. Several Th cell subsets including Th1 and follicular helper T cells produce IL-21 in vivo (40). However, abrogation of IL-21 signaling did not affect disease progression in BDC-CD4Cre-Tgfbr2^f/f mice. Similarly, deletion of IL-22 did not suppress diabetes progression. Instead, we demonstrated that IFN-γ–producing Th1 cells are responsible for diabetes development in these mice and that anti–IFN-γ treatment alone was sufficient to delay disease onset. Indeed, the deficiency of IFN-γ completely suppressed disease development, although we cannot exclude the possibility that GM-CSF may still contribute to the development of T1D in these mice in conjunction with IFN-γ. We also found that deletion of TGF-β1 from diabetogenic T cells results in increased Th1 cell differentiation, leading to accelerated diabetes development in BDC2.5 NOD mice. However, diabetes development in BDC-CD4Cre-Tgfb1^f/f mice was much milder than what we observed in BDC-CD4Cre-Tgfbr2^f/f mice, suggesting that although T cell–derived TGF-β1 is critical for regulating diabetogenic Th1 cell tolerance in a T-cell autonomous manner, other TGF-β1–producing cells also contribute to disease suppression.

The immunopathology of T cell–specific TGFβRII KO mice appears to be dependent on the timing of loss of TGFβRII expression, as revealed by the different phenotype of CD4Cre-Tgfbr2^f/f and dLckCre-Tgfbr2^f/f C57BL/6 mice (8, 9, 13). Deletion of TGFβRII in mature peripheral T cells using the dLckCre line, in which Cre-mediated deletion starts at the single-positive stage and continues in peripheral T cells, does not lead to systemic autoimmunity; instead, neonatal lymphopenic environment is thought to be required for hyperproliferation of TGFβRII-deficient T cells (13). Thus, the neonatal lymphopenia may also contribute to the rapid diabetes development in BDC-CD4Cre-Tgfbr2^f/f mice. However, we found that OX40Cre-mediated deletion of TGFβRII in mature peripheral CD4⁺ T cells after initial activation was sufficient to induce local Th1 responses and accelerated diabetes development in BDC2.5 NOD and regular NOD mice, despite the fact that almost all of the CD4⁺Foxp3⁻ peripheral naive CD4⁺ T cells retained TGFβRII expression. In this regard, another study using adult Mx1Cre-Tgfbr2^f/f mice, in which TGFβRII is systemically deleted after treatment with polyI:polyC, demonstrated that adult mice developed lethal autoimmune diseases by 2–3 wk after polyI:polyC treatment (41), suggesting that TGFβRII-deficient mice are prone to developing autoimmunity in adulthood and that neonatal lymphopenia-induced proliferation is not the only cause of autoimmunity in TGFβRII-deficient mice. Thus far, it is unclear why deletion of Tgfbr2 in thymic or peripheral T cells using different Cre lines results in the distinct outcomes. Although BDC-CD4Cre-Tgfbr2^f/f and BDC-OX40Cre-Tgfbr2^f/f mice develop T1D with almost the same kinetics, only a small fraction of memory BDC-OX40Cre-Tgfbr2^f/f T cells produce IFN-γ after in vitro stimulation compared with BDC-CD4Cre-Tgfbr2^f/f T cells. Because OX40⁺ cells are detected in less than 1% of naive CD4⁺ thymocytes that strongly respond to thymic self-ligands (23), the peripheral memory T-cell pool of OX40Cre-Tgfbr2^f/f mice may consist of TGFβRII-deficient cells in which Tgfbr2 is thymically and peripherally deleted. As all T cells that emerge from the thymus in CD4Cre- and Mx1Cre-Tgfbr2^f/f mice have already deleted the Tgfbr2 gene, whereas almost all dLckCre-Tgfbr2^f/f thymic T cells retained TGFβRII expression, it is possible that the abrogation of TGF-β signaling in thymic T cells results in the selection of an altered TCR repertoire. Further studies are required to address the relative contribution of TGF-β signaling depletion in thymic immature and peripheral mature T cells in immune homeostasis, as well as the involvement of lymphopenia in this process.

In conclusion, we showed that TGF-β1 derived from T cells acts on diabetogenic CD4⁺ T cells, but not Foxp3⁺ Treg cells, to control Th1 cell differentiation and spontaneous T1D development in BDC2.5 mice. Our study demonstrates the relationship between TGF-β1– and Treg cell–mediated maintenance of peripheral T-cell tolerance and extends the understanding of the cellular mechanisms involved in the control of autoimmune diseases and as such may aid in development of therapeutics to prevent the onset of autoimmunity, including T1D.

Materials and Methods

CD4⁺ T cells were enriched from spleen and LN cells by positive selection with anti-CD4 microbeads (Miltenyi Biotec), according to the manufacturer’s protocol. Effector CD4⁺ T cells were further purified from enriched cells with a cell sorter (Becton Dickenson) by gating on CD4⁺Vβ4TCR⁺ or CD4⁺Vβ4TCR⁺Foxp3RFP⁻ cells. Treg cells were FACS sorted by gating on CD4⁺Vβ4TCR⁺CD25⁺ or CD4⁺Vβ4TCR⁺Foxp3RFP⁺ cells. A total of 2 × 10⁵ effector T cells and 1 × 10⁵ Treg cells were transferred i.v. alone or in combination to Scid/NOD mice. After T-cell reconstitution, mice were monitored for diabetes development. Additional information is available in SI Materials and Methods.