TGF-β Signalling Is Required for CD4+ T Cell Homeostasis But Dispensable for Regulatory T Cell Function

Signalling by the cytokine TGF-β regulates mature CD4+ T cell populations but is not involved in the survival and function of regulatory T cells.


Introduction
Transforming growth factor b (TGF-b1) is a cytokine that is expressed throughout the hematopoietic system. It was initially described to suppress T cell proliferation [1], but later found to also support differentiation of T cells into specialized subsets. Mice deficient in TGF-b1 show hyperactivation and uncontrolled expansion of T cells leading to a lethal multi-organ autoimmune disorder [2]. Likewise, deficiency of either subunit of the heterodimeric TGF-b receptor (TR) is lethal [3,4]. T-cell-specific expression of a dominant-negative receptor mutant (DN-TR2) under the CD4 promoter results in a lymphoproliferative disease with infiltrates to multiple organs and development of severe inflammatory bowel disease [5]. Similar mice that express a DN-TR2 construct under human CD2 promoter presented with disease symptoms at the age of 3-4 mo [6]. The lymphoproliferative disease in these mice is caused primarily by expansion of CD8 + T cells that even progress to leukaemia [7,8]. Genetic ablation of TR2 in hematopoietic or, more specifically, in T cells leads to lymphadenopathy, splenomegaly, and inflammation of various organs followed by death at 3 wk of age [9][10][11]. Thus, all these models of interrupted TR function suggest an essential, nonredundant role for TGF-b1 in the maintenance of T cell tolerance. The observed autoimmune syndrome was initially thought to be caused by the apparent loss of regulatory T (T reg ) cells. Particularly thymic T reg development and survival seem to require TGF-b signals [10][11][12]. Yet neither reconstitution with WT T reg cells nor mixed bone marrow chimera experiments resulted in an amelioration of the autoimmune phenotype in these models [10,11]. Notwithstanding that cell type and reason for T-cell-mediated autoimmunity in the absence of TGF-b signalling remain unclear, in all these models thymic development was found to be altered [11][12][13][14]. A recent study shows that deletion of TR2 in mature T cells by dLck-Cre does not cause a fatal autoimmune syndrome [15]. In this model, however, the role of TGF-b signalling in mature T reg cells could not be addressed, since dLck-Cre did not lead to deletion of TR2 in Foxp3 + CD4 + T cells. Thus, our current, unbiased knowledge about the in vivo role of TGF-b for peripheral T, especially T reg , cells appears to be incomplete. To overcome this and analyze TGF-b function in T helper and T reg cells independent of developmental defects as well as systemic autoimmunity, we inducibly abrogated TGF-b signalling in peripheral CD4 + T cells. Surprisingly, loss of TR2 function in mature T cells, including T reg cells, did not lead to the spontaneous development of autoimmunity. Adoptive transfer of TR2-deficient CD4 + T cells into lymphopenic hosts led only to colitis but not systemic disease. However, the induced TR2 deletion in thymocytes of lymphopenic mice resulted in a rapidly developing lethal auto-inflammatory disorder. When TR2 ablation was restricted to postthymic T cells, we observed that not only T em (CD62L lo CD44 hi ) cells but also T reg cells exhibited hyperproliferation resulting from increased sensitivity to TCR signalling. TR2-deficient T reg cells retained their suppressive capacity both in vitro and in vivo. For mature CD4 + T cells in the adult mouse, TGFb seems therefore to solely be a regulator of T reg as well as T em cell expansion but is not required for the maintenance of tolerance.

Induced Ablation of TR2 From Peripheral CD4 + T Cells
Previous genetic studies of the function of TGF-b1 in T cells have either relied on germline gene deficiencies or involved celltype-specific mutagenesis. Consequently, these models were biased by strong alterations in thymic development [10][11][12][13]. To specifically address the role of TGF-b signalling in mature CD4 + T cells, we circumvented any developmental impact through the generation and use of a new tamoxifen-inducible [16] CD4-CreER t2 knock-in strain ( Figure S1A,B). We first tested the tamoxifen-induced recombination of this strain with the help of two fluorescent reporter mouse strains (RAGE and ROSA-EYFP) and found efficient and specific recombination in CD4 + T cells (unpublished data). As anticipated, recombination also took place in a small fraction of CD4 + splenic dendritic cells (DCs) and innate lymphoid cells (unpublished data). Since removal of the neomycin resistance gene (Neo(R)) by FLP-mediated deletion [17] reduced recombination efficiency without gain of specificity, a phenomenon observed before [18], we used animals bearing the Neo(R)containing allele for all further experiments. To investigate the role of TGF-b signalling in peripheral T helper cells, we crossed the CD4-CreER t2 with a conditional TR2 allele (TR2 f ) [9] obtaining CD4-CreER t2 /TR2 f/f (hereafter called iCD4TR2) experimental mice as well as CD4-CreER t2 /TR2 f/+ and TR2 f/f control mice. First, we tested ablation of TR2 in CD4 + T cells 14 d after commencing 5 d of tamoxifen treatment of iCD4TR2 mice (hereafter called tam-iCD4TR2 mice). mRNA levels within sorted naïve, central memory, effector memory/effector (T em ) and regulatory (T reg ) CD4 + T cells from tam-iCD4TR2 mice indicated a recombination frequency of 90%-95% in mature T cells ( Figure 1A), placing it at similar efficiency as many constitutive Cre strains [10,11]. Equally high deletion efficiencies were observed in CD4 + T cells from mesenteric LN (mLN) and the lamina propria (LP) ( Figure 1B). Surface expression analysis of TR2 on blood and spleen CD4 + T cells supported this observation ( Figure 1C, Figure S2A, and unpublished data) and also showed that expression of TR2 by other cell types including CD8 + T cells was unaffected ( Figure 1C and Figure S2A). As expected, TGF-binduced phosphorylation of Smad2 [19] was virtually absent in CD4 + T cells from tam-iCD4TR2 mice 2 wk p.a. ( Figure 1D). We thus conclude that target allele recombination in our inducible system is specific and efficient.
Because CD4-Cre-mediated TR2-ablation resulted in a strong thymic phenotype including reduction of single positive CD8 + T cells, abrogation of NK T cell development, and enhanced negative selection [11,12], we investigated the extent and consequences of induced TR2 ablation in the thymus. Two weeks p.a. the surface expression of TR2 was reduced only slightly on CD4 + thymocytes but not on any of the other major thymocyte compartments ( Figure 1E). Reduction of TR2 mRNA levels was found not only in CD4 + but also in a small fraction of CD8 + SP thymocytes only 2 wk p.a., indicating low-level and transient recombination at the DP stage ( Figure S2B). Also, expression of CD4, CD8, CD5, CD24, CD62L, and CD69 were unchanged in the thymus ( Figure 1F and unpublished data) as were number and phenotype of thymic T reg cells ( Figure 1F). Thus, our model allows the study of TGF-b signalling in mature CD4 + T cells by pulsechase experiments without significantly disturbing thymic T cell development.
Peripheral Abrogation of TR2 Signalling in CD4 + T Cells Does Not Lead to Autoimmunity Constitutive ablation of TGF-b signalling during thymic development of abT cells (CD4-Cre or Lck-Cre-mediated) but not during peripheral life (dLck-Cre-mediated) invariably results in a generalised and rapidly lethal autoimmune disorder [10,11,14,15]. When we followed tam-iCD4TR2 mice after induction of receptor deletion, they appeared healthy and showed no weight loss 2 and 4 wk after 5 d of tamoxifen treatment and up to 3 mo after feeding with tamoxifen citrate for 2 mo (Figure 2A). Because mild autoimmunity may not present clinically, we performed histopathological analysis of liver, kidney, pancreas, heart, colon, and thyroid gland but could not detect cellular infiltrates in any of these organs (unpublished data). We also tested for secondary dysregulation of B cell tolerance manifested by autoantibody production, as previously described upon thymic deletion of TR2 expression [11]. Yet sera from tam-iCD4TR2 mice at 6 wk or even 5 mo p.a. did not contain significant levels of

Author Summary
TGF-b is a cytokine thought to be critical for the maintenance and function of tolerance in the immune system. In many studies the disruption of TGF-b signalling in CD4 + T cells (a type of white blood cell that coordinates immune responses) has resulted in autoimmune syndromes. We show here that the induced removal of this cytokine's receptor from these specialised blood cells results in an astonishingly mild outcome. Contrary to expectations, the number of regulatory T cells is actually increased, and we find that these cells are not dependent on TGF-b signalling. We also show that removal of the receptor from mature CD4 + T cells does not lead to lethal autoinflammation; only when we removed the receptor during development of the cells did we see the characteristic lethal multi-organ inflammation reported previously in constitutive models of TGF-b receptor ablation. In summary, our findings indicate that although TGF-b regulates maintenance of mature CD4 + T cells, its signals are dispensable for immune tolerance within this cell population.
autoantibodies as shown by ELISA against dsDNA ( Figure 2B). Thus, removal of TR2 from mature CD4 + T cells does not result in tolerance loss or autoimmunity. To exclude that recent thymic emigrants could dilute out cells lacking TR2, we thymectomized mice 1 wk before an 8-wk tamoxifen citrate treatment. Again, we failed to observe any signs of autoimmune disease by clinical appearance and weight loss up to 5 mo p.a. ( Figure 2C) even though the frequency of TR2-deficient cells remained constant until the end of the experiment ( Figure 2D). The absence of elevated serum antibody levels, autoantibodies, and organ infiltrates again excluded subclinical tolerance loss ( Figure 2E,F and unpublished data). Furthermore, CD8 low or CD4 + NK1.1 + T cells, proposed to carry autoreactive activity in the constitutive TR2-deficient models [10], were not detected 2 and 4 wk as well as 5 mo p.a. and after thymectomy ( Figure 2G and unpublished data). In conclusion, in contrast to models of constitutively impaired TGF-b signalling in T cells [5,6,10,11], induced TR2 ablation from peripheral CD4 + T cells does not result in the loss of self-tolerance.

Induced Ablation of TR2 During Thymic Development Combined with Lymphopenia Results in Lethal Autoimmunity
Abrogation of TGF-b signalling through CD4-Cre takes place during thymic development and affects all abT cells including T reg cells. This thymic TR deletion results in lethal autoimmunity, in contrast to peripheral TR deletion through CD4-CreER t2 . To understand the underlying cause of the different outcomes of TR2 ablation in T cells, we assessed whether they were caused by deletion frequency, by ablation during thymic development, or by lymphopenia. To mimic the environment found in neonates with constitutive T-cell-specific TR2 ablation [10,11], we induced recombination during thymic development in a lymphopenic environment: We initiated treatment with tamoxifen of Rag-1 2/2 mice 3 d before irradiation and reconstitution with bone marrow of iCD4TR2 or control mice (experimental scheme outlined in Figure S3A). Tamoxifen treatment led to deletion of TR2 in the thymus of mice reconstituted with iCD4TR2 bone marrow: both CD4 + and CD8 + T cells lacked TR2 expression ( Figure 3A). The mice developed signs of sickness 28 d after bone marrow reconstitution (apathy, runted appearance, weight loss) and started dying after 30 d ( Figure 3B,C). Mice that received control bone marrow (CD4-CreER t2 /TR2 f/+ ) did not shown any signs of disease ( Figure 3B). Sick bone marrow chimeras presented with T cell infiltrates only in lung and liver tissue but not the colon ( Figure 3D and unpublished data). The frequency of CD62L hi CD44 2 naïve T (T n ) cells was severely reduced and of CD62L lo CD44 + T em cells highly increased in both CD4 + and CD8 + T cell compartments ( Figure 3E). As has been reported Figure 2. Absence of autoimmunity after TR2 ablation in mature CD4 + T cells. (A) iCD4TR2 and control mice were treated with tamoxifen citrate for 2 mo followed by 3 mo on normal diet and body weight was monitored (mean 6 SEM, 5 mice per group, representative data of two independent experiments). (B) ELISA for anti-dsDNA antibodies in sera from tam-iCD4TR2 and control mice 2, 4, 6 wk, and 5 month p.a.; as positive control a serum from a fas-deficient mouse was used. (C) iCD4TR2 and control mice were thymectomised 1 wk before the beginning of tamoxifen treatment and treated with tamoxifen citrate for 2 mo followed by 3 mo on normal diet and body weight was monitored (mean 6 SEM, 5 mice per group, representative data of two independent experiments). (D) Flow cytometric analysis of TR2 expression by CD4 + and CD8 + T cells from peripheral blood after 5 mo postthymectomy and tamoxifen treatment. before in constitutive CD4-Cre-mediated deletion [10,11], we also observed a significantly decreased number of peripheral T reg cells ( Figure 3F). CD4 + CD25 high T cells showed similar levels of Foxp3 expression while CTLA-4 was significantly up-regulated ( Figure 3G). When tamoxifen treatment started 5 wk after bone marrow transfer, we observed no sign of autoimmunity, tissue infiltration, and only slight T em expansion ( Figure 3H,I, Figure  S3B, and unpublished data), thus excluding the possibility that bone marrow chimerism per se could predispose to development of the disease.
The treatment of bone marrow chimeras with tamoxifen from the onset on led to deletion of TR2 also in CD8 + T cells, indicating efficient recombination already during the thymic DP stage. To exclude that these TR2-negative CD8 + T cells were causing disease, we repeated the experiment described in Figure 3A-G with deletion of CD8 + T cells ( Figure S3C). Experimental chimeras treated with anti-CD8 antibody developed the lethal autoimmune syndrome only slightly later than those treated with isotype control ( Figure 3J). Lungs and livers isolated from both groups of experimental chimeras presented with infiltrates ( Figure 3K and unpublished data). We found no difference in weight loss, survival, or tissue pathology between control mice treated with anti-CD8 antibody or isotype control. Also, TR2 deletion was equally efficient in both groups (unpublished data). Thus, CD4 + T cells are the main effectors of disease upon TR2 ablation during thymic development in a lymphopenic situation.
Taken together, the fact that the TR2 deletion frequency in CD4 + T cells is similar in all these bone marrow chimerism experiments suggests that the cause of the autoimmune syndrome is either the absence of TR2 during thymic development or during repopulation of a lymphopenic environment, thus mimicking the observations made in constitutive models of TR2 deletion (i.e., CD4-Cre/TR2 f/f ).

Transfer of TR2-Deficient CD4 + T Cells Into Lymphopenic Hosts Does Not Lead to Multi-Organ Inflammation But to Colitis
To distinguish between altered thymic development and lymphopenia as cause of the autoinflammation, we assessed the influence that different lymphopenic conditions had on the behaviour of TR2-deficient CD4 + T cells. To achieve CD4 + Tcell-restricted lymphopenia, we treated thymectomized animals with a low dose of anti-CD4 antibody. This treatment led to a ,90% depletion of blood CD4 + T cells ( Figure 4A, Figure S4A, and unpublished data) followed by acquisition of an activated phenotype by the remaining CD4 + T cells in both control and experimental mice ( Figure 4B). To obtain severe general lymphopenia we used sublethal irradiation ( Figure 4C and Figure  S4B). In this setting both CD4 + and CD8 + T cells acquired the activated phenotype (unpublished data). Even though in both acute lymphopenic conditions we observed pronounced lymphopenia-driven T cell expansion, none of the animals presented with signs of autoimmune disease by appearance, anti-dsDNA autoantibody titres, and histology until week 20 (unpublished data). Again, TR2 ablation was similarly efficient throughout these experiments ( Figure S4A). Only when we adoptively transferred TR2-deficient cells into a completely lymphopenic environment (RAG-1 deficiency), strong colitis developed within 30 d ( Figure 4D,E and unpublished data), which was accompanied by a massive infiltration of CD4 + T cells to mesenteric lymph nodes ( Figure S4C). Interestingly, the mice presented with neither apathy nor runted appearance, and histology did not reveal any inflammation of lungs or livers, which is a characteristic feature in mice in which TGF-b signalling is abrogated in thymocytes ( Figure 4E and unpublished data). Thus, adoptive transfermediated colitis can be induced by TR2-deficient T cells even in the presence of T reg cells, but a general autoinflammatory syndrome is not observed. In vivo TR2 deletion in CD4 + T cells combined with acute lymphopenia, however, does not lead to loss of tolerance.

Dysregulated Effector CD4 + T Cell Homeostasis in Absence of TGF-b Signalling
To better understand the role of TGF-b signalling in mature CD4 + T cells, we analysed T effector homeostasis after TR2 removal in a longitudinal manner. We found slightly reduced CD4 + T cell numbers in spleen and LNs 2 and 4 wk p.a. ( Figure 5A and unpublished data) while the total number of CD8 + and of central memory CD4 + T (CD62l hi CD44 + ) cells remained unchanged (unpublished data). In addition, we observed a modest but significant expansion of T em cells. This phenotype was transient as cell numbers and the frequency of T em cells returned to normal 6 wk p.a. ( Figure 5B and unpublished data). In support of this observation, BrdU incorporation revealed increased proliferation of T em but not of T n and central memory CD4 + T cells 2 wk p.a. ( Figure 5C and unpublished data). To test whether the increase of T em cells was transient due to replacement by new TR2-expressing T cells, we thymectomized mice prior to TR2 ablation. In the absence of thymic emigration we observed that the elevated numbers of T em cells persisted ( Figure 5D).
To investigate whether the activation and proliferation of T em cells upon TR2 ablation was a cell-intrinsic property or driven in trans by cell extrinsic factors, we generated bone marrow chimeras by mixing WT CD45.1 + and either iCD4TR2 CD45.2 + or control TR2 f/f CD45.2 + bone marrow (scheme depicted in Figure 5E). In chimeras containing iCD4TR2 bone marrow, the frequency of mutant CD4 + T cells was increased significantly at 4 wk p.a. (unpublished data). Two weeks p.a. activation of CD4 + T cells and T em cell proliferation were restricted to cells lacking TR2 ( Figure 5F). The TR2-deficient T n cell compartment was diminished while the mutant central memory compartment was unchanged (unpublished data). Analysis of control chimeras showed no differences between the CD45.1 + and CD45.2 + populations. These data thus suggest that TR2 regulates the homeostasis of mature T em and T n cells. While in models of constitutive TR2 ablation a large fraction of CD4 + T cells developed into IFN-c-producing Th1 cells [11], we found only slightly increased IFN-c production but no difference in T-bet levels after peripheral deletion of TR2 ( Figure 5G). Production of Th2 cytokines was hardly detectable ( Figure 5G) and the expression of the chemokine receptors CCR4, CCR5, CCR6, and CCR7 was unchanged (unpublished data).
Thus, hyperactivation, increased proliferation of T em cells, and the reduction of the T n cell compartment are cell-intrinsic consequences of TR2 ablation. The postthymic abrogation of TGF-b signalling does not lead to spontaneous acquisition of a Th1 phenotype.

Increased TCR-Dependent Activation in Absence of TR2
Since the up-regulation of CD69 upon TR2 ablation ( Figure 6A) supported the idea of a TCR-dependent effect [20] of TR2 signalling on proliferation rate of mature CD4 + T cells in vivo, we hypothesized that TR2 alters either the sensitivity to TCR ligands or the signalling through homeostatic cytokine receptors. First we assessed whether the absence of TR2 would result in increased sensitivity to TCR triggering. We stimulated CD4 + T cells in the presence of titrated amounts of anti-CD3 antibody and found an increased response to TCR stimulation through suboptimal anti-CD3 concentrations ( Figure 6B), indicating an increased sensitivity to TCR stimulation after deletion of TR2. To exclude the possibility that this was the result of a larger fraction of experienced CD44 hi cells within the pool of stimulated TR2deficient cells, we activated naïve and effector memory cells separately. Again the TR2-deficient cells, naive and experienced, presented with higher sensitivity to TCR stimulation ( Figure  S5A).
Indo-1-based ratiometric analysis revealed a small but significant increase in steady-state cytoplasmic calcium concentrations ( Figure 6C). Upon stimulation the calcium influx was accelerated, suggesting an increased sensitivity of TR2-deficient T cells ( Figure 6C). We did not find, however, differences in phosphorylation of CD3f, SLP76, lck, ZAP70, or ERK upon TCR stimulation (unpublished data). While no differences in CD25 expression or IL-2 production were found ( Figure 6D), a slight increase of CD122 (IL-2Rb) on CD4 + tam-iCD4TR2 T cells was detected 2 wk p.a. (Figure 6D). This could potentially have led to altered sensitivities to IL-15 or IL-2. However, the hyperproliferation of TR2-deficient T cells stimulated by suboptimal TCR activation could not be corrected by blockade of the IL-2R ( Figure 6E and Figure S5B). Likewise, dose responses to IL-2 and IL-15 were equal in TR2 + and TR2 2 CD4 + T cells ( Figure 6F and Figure S5B). Furthermore, reactivity to IL-7 was unchanged in this in vitro assay (unpublished data). Thus, the hyperproliferative activity of TR2-deficient CD4 + T cells appears to be the result of increased TCR sensitivity but not of cytokine-mediated signals. To rule out that increased survival of TR2-deficient cells was contributing to the observed expansion, we removed the receptor from splenic CD4 + T cells in vitro and assessed survival. The result was an increased incidence of apoptosis by cells lacking TR2 as shown by Annexin V staining after 20 and 45 h ( Figure S5C), thus excluding this possibility.

Dysregulated Homeostasis of Regulatory T Cells After Ablation of TR2
T reg cell development and maintenance is generally thought to be critically dependent on TGF-b signals [21,22] as thymic deletion of TGF-b signalling was shown to result in a strong reduction of T reg cell number in the periphery [10,11,14]. When we examined T reg cell homeostasis after peripheral ablation of TR2 from CD4 + T cells, we made the surprising observation of increased frequency and number of T reg cells 2 and 4 wk p.a. as a result of hyperproliferation ( Figure 7A and unpublished data). This increase in the T reg population size was also observed in various LNs, the spleen, the lung, Peyer's patches, but not in the intestinal lamina propria ( Figure 7B and unpublished data) and was stable in thymectomized animals ( Figure 7C and unpublished data). Mixed bone marrow chimeras showed that similar to T em cells, increased proliferation is an intrinsic property of the TGF-b unresponsive T reg cells ( Figure 7D,E and Figure S6A,B). The majority of the expanded splenic and LN Foxp3 + Treg cells in tam-iCD4TR2 mice expressed Nrp-1 and Helios ( Figure 7F, Figure S6C, and unpublished data), which suggests a thymic origin [23]. Proliferation of TR-2-deficient Nrp-1 + regulatory T cells was increased compared to WT T reg cells ( Figure 7G). In thymic T reg precursors lacking TGF-b signalling, a decrease in the levels of the key transcription factor FoxP3 [24][25][26] had been reported [27]. We did not observe such reduced Foxp3 levels in CD4 + CD25 + cells in short-and long-term experiments (unpublished data) and in the mixed bone marrow chimeras ( Figure 7H). Also, the expression of key surface molecules by TR2-deficient T reg cells-namely CTLA-4, GITR (Tnfrsf18), [28,29] and ICOS [30]-remained unchanged ( Figure 7H) in contrast to our observations in the bone marrow chimera model presenting with a lethal autoinflammatory syndrome. We found, however, that as described for TR2-negative T em also T reg cells presented with an up-regulation of CD69 ( Figure S6D). Therefore, to investigate TCR hyperresponsiveness, splenocytes isolated from tam-iCD4TR2 and control mice 2 wk p.a. were CFSE-labelled and stimulated with different concentrations of anti-CD3 antibody. Increased proliferation of TR2negative T reg cells was observed after 3 d of culture with   Figure S6E). Similarly, sorted T reg cells showed hyperproliferation to anti-CD3-mediated stimulation when cultured alone or in presence of IL-2 (unpublished data). Taken together, TGF-b signalling suppresses overt T reg cell proliferation but does not seem to be required for maintenance of the T reg cell phenotype.

TR2 Expression by T reg Cells Is Not Required for Their Suppressive Capacity
Since peripheral TR2 ablation resulted in loss of cell cycle control and thus T reg cell expansion, we tested whether these mutant T reg cells retained functional characteristics identical to WT T reg cells. In an in vitro suppression assay TR2-deficient T reg cells and WT T reg cells inhibited the proliferation of conventional T cells to a similar extent ( Figure 8A,B). This observation was confirmed in vivo by the capacity of TR2-deficient T cells to suppress the development of colitis. TR2-deficient or control T reg cells were co-transferred with CD4 + T n cells into lymphopenic RAG1-deficient recipients. While transfer of T n cells alone resulted in severe weight loss, indicative of colitis, the disease was similarly suppressed by co-transfer of T reg cells from tam-iCD4TR2 or control animals ( Figure 8C). Colon histopathology did not reveal any difference between mice that received TR2deficient or WT T reg cells ( Figure 8D).
Given that T-cell-mediated autoimmunity upon thymic ablation of TGF-b signalling could not be suppressed by WT T reg cells [11], we tested whether WT and TR2-deficient T reg could suppress mutant responder T cells. In the in vitro as well as the in vivo suppression assay, TR2-deficient T cells remained susceptible to suppression by T reg cells (WT and mutant) ( Figure 8E,F). Taken together, TGFb is not necessary for the functional capacity of T regs to suppress immune responses.

Discussion
The pleiotropic nature of the TGF-b family members has made it extremely challenging to unravel their function in vivo. All models of constitutive ablation of TGF-b signalling in abT cells during thymic development have invariably revealed autoimmune phenotypes [5,6,[9][10][11][31][32][33][34][35]. In most of these models [10][11][12]14] the T reg cell population collapsed, resulting in almost complete loss of T reg -mediated peripheral suppression. These observations led to the dogma that TGF-b1 is required for establishment and maintenance of T cell tolerance. Yet because gene ablation in all these systems took place prior to or during thymic development, it could not be excluded that the observed immune dysregulation was a consequence of T cell development in the absence of TGF-b signals. In fact, while Doisne and colleagues showed that NKT cells depend critically on TGF-b for their development [13], Ouyang and colleagues reported that even conventional T cell development is modified by TGF-b signalling [12]. Therefore, to study the importance of TGF-b for peripheral CD4 + T lymphocytes, we circumvented such thymus-related abnormalities cells (right panel) within the LN CD4 + T cells of the indicated CD45.1 + or CD45.2 + bone marrow-derived cells (experiment described in Figure 5E, mean 6 SEM, 10 mice per group, analysed in three independent experiments). (E) The percentage of BrdU + T reg cells isolated from LN (experiment described in Figure 5E [60]. (E) Criss-cross in vitro suppression assay: sorted conventional tam-iCD4TR2 and WT T cells were cocultured with sorted tam-iCD4TR2 and wt T reg cells at various ratios. Analysis was performed after 96 h (representative data of two independent experiments). (F) Development of colitis in Rag1 2/2 mice after adoptive transfer of conventional tam-iCD4TR2 and wt T cells alone or in combination with tam-iCD4TR2 T reg cells. Change in body weight after 8 wk posttransfer (mean 6 SEM, 3 mice per group, representative data of two independent experiments). doi:10.1371/journal.pbio.1001674.g008 through induced deletion of TR2 in mature T helper cells only by use of a novel CD4-CreER t2 system. It allowed induced recombination of the TR2 target allele within up to 95% of postthymic CD4 + T cells. Recombination during thymic development remained small after short-term tamoxifen application, and we did not find a contribution of TR2-deficient CD8 + T cells to the peripheral T cell pool in this setup. Even long-term treatment with tamoxifen resulted in recombination in only a small fraction of CD8 + T cells. Only in one experimental setup, when thymic development was initiated in the presence of tamoxifen, did we observe a higher recombination frequency in CD8 + T cells. Therefore, our CD4-CreER t2 system proved to be useful for the study of TR2 function in peripheral CD4 + T cells. The reason for the small recombination rate in DP thymocytes may lie in the fact that the cells reside only shortly at this stage. Also, during this stage expression of CreER t2 is initiated and protein has to first accumulate. Therefore only a small number of DP cells is receptive and exposed to tamoxifen at the same time. Further, only a minor fraction of T cells is generated newly during the 5-d treatment setup and can thus contribute to the peripheral pool. Another factor influencing efficiency of Cre-mediated recombination is the accessibility of the target allele. Thus the extent of recombination in the thymus may depend on the target used as well as duration and route of tamoxifen application.
Taken together by using the tam-iCD4TR2 mouse for the analysis of TGF-b's role for CD4 + T cells, we report (i) the absence of clinical or immunological manifestations of autoimmunity, even when removal of TR2 from CD4 + T cells was maintained for months; (ii) the development of a lethal autoimmune syndrome only when TR2 ablation is induced during thymic development; (iii) an increase in T em and T reg proliferation as a result of increased TCR signal sensitivity in the absence of TR2; and (iv) that TGF-b signalling is not required for the maintenance of the suppressive capacity of mature T reg cells in vitro and in vivo. We therefore propose that the widely held notion of TGF-b signalling being a major contributor to peripheral T cell tolerance and T reg cell maintenance results from observations of developmental aberrations due to the use of complete gene ablation during thymic development under lymphopenic conditions.
Since the autoimmune syndrome in the most drastic cases of TGF-b signal manipulation results in lethality at a very young age, we hypothesized that neonatal lymphopenia was a contributing factor. To distinguish between effects through mere lymphopenia and aberration in thymic development, we performed TR2 ablation in three different lymphopenic setups, but all of them failed to drive systemic autoinflammation. One condition, adoptive transfer of TR2-deficient T cells into a lymphopenic host, resulted solely in colitis, similar to a recently published study of dLck-Cre-driven TR2 ablation [15]. We could show that the difference between development of colitis (CD4-CreER t2 /TR2 f/f , dLck-Cre/TR2 f/f ) and a lethal systemic autoimmune syndrome (our thymic deletion model, CD4-Cre/TR2 f/f , Lck-Cre/TR1 f/f ) lies in whether TGF-b signalling is disrupted later than or at the DP thymic stage, respectively. By use of the acute lymphopenia models of sublethal irradiation and anti-CD4 depleting antibody treatment, we focused on systems which were shown to drive T cell proliferation by increased availability of IL-7, a setting which amplifies signals from weak TCR/self-pMHC contacts [36,37]. We again observed hyperproliferation and basically complete conversion to a (post)activated phenotype (unpublished data), yet neither organ-specific nor systemic autoimmunity developed. A difference in recombination frequencies could have been another cause of the difference in development of autoimmunity between the inducible and cell-type-specific systems, but we and others [10,11] provided evidence to rule out this possibility. The decrease of T reg cell numbers or of their suppressive activity in the constitutive models are also an unlikely cause for development of autoimmunity since activated TR2-deficient T cells in these models were shown to be resistant to control by even WT T reg cells [10,11,27]. Taken together, minor differences in gene deletion efficiency as well as the presence/absence of T reg cells cannot account for the divergent autoimmune phenotypes of the constitutive and our inducible model. Also, impaired thymic negative selection was ruled out by a study that found actually a slight increase of negative selection in absence of TR2 [12]. Another possibility was that TR2 was not ablated in the T cell subset required for development of autoimmunity, but several studies indicated that both CD4 + and CD8 + T cells are responsible for the autoimmune phenotype in absence of TGF-b signalling [10,38,39]. We excluded this possibility by depletion of CD8 + T cells in the setup. This resulted also in lethal autoimmunity, albeit slightly delayed, and did not prevent weight loss and infiltration into the lung and liver. Further, a recent study showed that TR2deficient CD4 + T cells alone can induce autoimmune colitis under lymphopenic conditions [15], an observation that we confirmed in our model. One report placed the pathogenic activity of TR2deficient T cells in unconventional NK T cells [10], but we show that NK1.1 expression is not a mandatory characteristic of such autoreactive T cells. Taken together, the only common denominator for development of severe lethal autoimmunity upon removal of TGF-b signalling in T cells is that recombination must occur in developing thymocytes in a severely lymphopenic animal. While a recent study reported that also the abrogation of TGF-b signalling in CD11c + cells leads to the development of autoimmune disease [40], this seems unlikely in our system since recombination in CD4 + CD11c + cells was very rare.
When we used our model to analyze the role of TGF-b signalling for survival and homeostasis of peripheral CD4 + T cells, it revealed a modest increase in size of the T em and T reg compartments with both populations being activated (CD69 + ) and cycling (BrdU + ). For T em cells such an effect has also been reported when TGF-b signalling was abrogated by CD4-Cre or through a CD4 promoter-driven dominant-negative TR2 transgene, but the extent was far stronger than in our model [5,10,11]. In all such models the hyperproliferation and activation in CD8 + T cells was considerably larger than in CD4 + T cell compartment [5,10,11].
As reported before [8,10,11] this hyperproliferation of TGF-b unresponsive T cells depends on specific peptide-MHC recognition. The introduction of a TCR transgene, whose antigen is not recognized in the periphery [41], into our model resulted in amelioration of the proliferative phenotype (unpublished data). This is supported by data from dLck-Cre-mediated TR2 ablation, acting in late thymic development. In this model combination with OT-I TCR transgene showed increased TCR sensitivity when peptides of different affinities were used [15]. Similarly, we also found increased sensitivity to TCR stimulation in vitro in the absence of TR2, which may be a result of increased steady-state intracytoplasmic calcium concentrations and faster calcium response. Taken together, in all models of abrogation of TGF-b signalling in T cells the threshold of productive TCR signalling is decreased.
Even though we and Zhang and Bevan showed that increased sensitivity of TCR signalling is a major factor in the increased proliferative phenotype of TR2-deficient T cells, also direct control of cell cycle by TGF-b signalling has been reported [42][43][44][45]. Such cell-cycle control appears to be connected to the establishment of T cell tolerance (anergy) in the DO11.10 model, thus again highlighting the importance of TGF-b in maintaining T cell homeostasis [45].
As an additional cause of hyperproliferation and activation of TGF-b unresponsive T cells, dysregulated sensitivity to cc cytokines has been reported. The homeostasis especially of CD8 + memory T cells was shown to be under IL-15 control, possibly due to higher expression of CD122 in comparison to CD4 + memory T cells (for rev: [37,46]). In agreement with this, a CD2 promoter-driven dominant-negative TR2 transgene was found to present with a IL-15-dependent expansion of memorylike CD8 + T cells [8]. The T-bet-mediated up-regulation of CD122 in CD4-Cre TR2 model, however, was observed in both CD4 + and CD8 + T cells and was considered to be central for maintenance and expansion of TGF-b-unresponsive memory T cells [11]. In our inducible model we found neither evidence of increased IL-2 production by activated CD4 + T cells nor increased expression of CD25 (IL-2 Ra). We observed, however, a slight increase in CD122 expression (IL2Rb chain) upon induced ablation of TR2 in CD4 + T cells. Yet this did not lead to increased sensitivity to common gamma chain cytokines, making them an unlikely cause of the hyperproliferation seen in T em and T reg cells. Thus, the impairment of TGF-b signalling in peripheral T cells leads to hyperproliferation independent of cytokine signalling, while constitutive ablation during thymic development results in dysregulated cc cytokine production/signalling and massive autoimmunity.
We therefore conclude that TGF-b controls postthymic homeostasis of both naïve and memory CD4 + T cells. Yet the extent of hyperproliferation in induced TR2-deficient T cells is significantly lower than in constitutive TR2 mutants.
Previously, the almost complete absence of peripheral T reg cells in CD4-Cre/TR2 f/f animals was taken as evidence for a prominent role of TGF-b in T reg cell maintenance [11]. In this model T reg cells and precursors also showed a hyperproliferative phenotype [11], but in the periphery T reg cells died prematurely [12]. Also, the drastic expansion of thymic T reg cells in TR1deficient thymocytes [14] resulted in their failure to survive in the periphery. In contrast to these observations, upon peripheral TR2 ablation T reg cells show increased proliferation without associated collapse through massive apoptosis. This expansion of TR2negative T reg cells in peripheral lymphoid organ in our model was restricted to Nrp-1 + Foxp3 + T reg cells, most likely cells of thymic origin [23]. A specific increase of nT reg cells upon TR2 ablation is also supported by the increase in number of Helios + T reg cells [47]. Conversely and similar to our observation, TGF-b1 gene ablation from activated peripheral T cells resulted in an expansion of T reg cells [48], possibly a consequence of decreased local TGF-b availability. Expansion of T reg cells could have been the result of IL-2 production by the increased number of activated T cells in tam-iCD4TR2 animals, but our data excluded this possibility. Also, only mutant T reg cells showed increased cycling in mixed bone marrow chimeras, thus ruling out any effects in trans. The up-regulation of CD69 by T reg cells and in vitro stimulation with suboptimal anti-CD3 concentrations suggest increased TCR sensitivity, similar to conventional T cells.
We are to our knowledge the first to study T reg cell physiology after induced peripheral abrogation of TGF-b signalling in vivo. Our finding of the TGF-b independence of the mature, postthymic T reg phenotype and function does agree with epigenetic imprinting taking place at the Foxp3 locus and also many other loci relevant for T reg cell physiology [49]. Upon induced TR2 ablation we observed neither reduction of Foxp3, CTLA-4, and GITR levels nor any decrease in the suppressive capacity of TR2-deficient T reg cells. Tone and colleagues [50] showed that TGF-b-induced Smad-mediated activation of the Foxp3 locus through interaction with a conserved Smad-NFAT response element (CNS1) in thymocytes is essential for nT reg cell generation. Once the T reg phenotype is established, expression of FoxP3 in nT reg cells is largely independent of the promoter region CNS1 [51,52], which is supported by our observation of unchanged FoxP3 levels. Further, Miyao and colleagues recently showed that T reg cells represent a stable lineage, which robustly maintains its committed state once the Treg cell-specific demethylation region (TSDR) has been demethylated [53].
TGF-b receptor signalling is, however, required for the induction of Foxp3 among peripheral naïve CD4 + T cells (iT regs ) [54,55]. In agreement with this, we found that in the lamina propria, where iT reg cells contribute substantially to the T reg pool, no T reg expansion can be observed upon induced TR2 ablation.
Finally, we could show that the proliferation of TR2-deficient effector cells can be inhibited by TR2-deficient T reg cells, a question that has been raised recently [56]. Thus, we propose that ablation of TGF-b signalling during thymic development leads to intrathymic hyperproliferation of T reg cells, which then cannot survive in the periphery. Conversely, when TR2 is removed from already established peripheral T reg cells, these cells keep their T reg cell characteristics and undergo increased proliferation.
Taken together, our study suggests that several misconceptions about TGF-b function for mature T cells are the result of gene ablation during T cell development. By restricting the genetic defect (TR2-deficiency) to mature CD4 + T cells, we show that TGF-b signalling is not essential for the suppression of autoimmunity. Furthermore, it is not obligatory for the maintenance of a functional T reg cell pool. Instead it is required for the homeostasis of T reg and T em cells by curbing TCR signalling and therefore overt proliferative activity. In contrast, ablation of TGF-b signalling during thymic development as well as during lymphopenia may predispose for development of autoimmunity. Thus, TGF-b1 remains a cytokine with critical function in the regulation of T cells, yet its role in peripheral tolerance has been overestimated.

Generation of the CD4-CreER t2 Allele
Targeting of the CreER T2 fusion gene to the CD4 locus was achieved by replacing part of exon 2 including the start codon with the targeting vector pBluescript CD4-Cre19ER T2 by homologous recombination in the C57BL/6 derived ES cell line Bruce4.

Animal Maintenance and Experiments
Tgfbr2 fl/fl [57] were kindly provided by U. Malipiero, and ROSA-EYFP mice [58] were kindly provided by A. Diefenbach, C57BL/6J (B6) mice were purchased from Charles River and congenic C57BL/6-CD45.1 bred in-house. CD4-CreER t2 , CD4-Cre, RAGE, TGFbRII fl/fl , B6, B6-CD45.1, Rag1 2/2 , and ROSA-EYFP mice were maintained in barrier and specific pathogen-free facilities at the University of Zurich and Technical University of Munich and handled in accordance with approved protocols under permits of the cantonal veterinary office. Organs and sera of fas-deficient animals were provided by Bojan Polic (Rijeka, Croatia).
For genotyping of CD4-CreER t2 mice, the following primers were used: GCC AGC TCA TTC CTC CCA CTC, CAT GGG ACT TTG GGC TTC TAG G and CCC AAC CAA CAA GAG AGC TCA AG, amplifying wt (440 bp) and transgene (720 bp). Genotyping of TR2 was performed according to [57].
If not stated otherwise all mice used for the experiment were at the age of 6 to 10 wk.
For tamoxifen (Sigma) application, tamoxifen was dissolved in 100% ethanol to 1 g/ml, vortexed, and mixed with olive oil to a final concentration of 100 mg/ml. The suspension was incubated at 56uC for 15 min and sonicated for 20 min. iCD4TR2 mice were force-fed with 5 mg tamoxifen per day for 5 consecutive days. Day 7 after start of application is denoted 1 wk p.a. For longterm treatment, mice were fed with tamoxifen citrate (Harlan) for 8 or 12 wk ad libitum.
For induction of lymphopenia, mice were irradiated with 550 rad or injected i.p. once with 10 mg GK1.5 anti-CD4.
For depletion of CD8 + T cells, mice were injected at day 15 and 17 post-bone marrow transplantation with 250 mg of anti-CD8 antibody (YTS 169.4 clone, BioXcell) followed by 100 mg every week. Controls received the same amount of isotype control antibody LTF.2 (rat IgG2b, BioXcell).
For in vivo proliferation analysis BrdU (80 mg/100 ml) was added to drinking water and changed every second day for 7 d. Intracellular staining with anti-BrdU antibody (eBioscience) and Foxp3 Staining Buffer Set (eBioscience) followed by DNAse (Invitrogen) treatment for 1 h in 37uC was performed. Samples were analyzed by using FACS Canto II.
For the in vivo suppression assay, Rag1 2/2 mice were injected intraperitonelly with 4610 5 conventional T cells (CD45.1 + ) alone or in combination with 2610 5 regulatory T cells (CD45.2 + ). Mice were weighed and assessed for clinical sings of colitis weekly and were killed 9 wk after transfer. Colons were fixed in 4% formalin, paraffin-cut, and stained with hematoxylin and eosin according to standard procedures.
For adoptive transfer experiments, Rag 2/2 mice were injected intraperitonelly with 2610 6 purified T cells (T cell isolation kit, Milteny). Mice were weighed, assessed for clinical symptoms of colitis weekly, and killed 7 wk after transfer.
To enrich CD4 + T cells, population magnetic sorting was performed (CD4 T cell isolation kit, Milteney Biotech) according to the manufacturer's protocol.
T reg cells and T conv cells were sorted on a FACS Aria (BD Biosciences) on the basis of being CD4 + CD45RB lo CD25 + and CD4 + CD45RB hi CD25 2 , respectively. Naïve CD4 T cells were sorted on the basis of being CD4 + CD44 lo CD62l hi , effector memory/experienced CD4 + CD44 hi CD62l lo , and central memory CD4 + CD44 hi CD62l hi . For proliferation assays naïve and memory CD4 + CD25 2 T cells were sorted.

Histological Tissue Analyses
Mice were euthanized with CO 2 , perfused first with PBS, and then 4% paraformaldehyde in PBS. For histological analysis, kidney, liver, heart, colon, and small intestine were fixed in 4% paraformaldehyde in PBS, paraffin-embeded, cut into 30 mm thick sections, and stained with hematoxylin-eosin and anti-CD3 antibody according to standard procedures.

In Vitro Proliferation Assay
Cells from the spleen or lymph nodes were cultured in RPMI 1640 (Invitrogen) medium supplemented with 10% FCS, 1% penicillin-streptomycin, 0.5% 2-mercaptoethanol, stimulated for 72 h with different concentrations of anti-CD3 (2C11, BioXcell) alone or together with 5 mg/ml anti-CD28 (N37, BioXCell). For CFSE labelling, cells were first stained for 20 min in the dark at room temperature with CFSE (carboxyfluorescein diacetate succinimidyl diester, 5 mM) and washed in PBS. For thymidine incorporation assays, 1 mCi of thymidin per well was added for the last 24 h of cultures. For blocking of IL-2 signalling, the anti-CD25 antibody (PC61, eBioscience) at different concentrations was added to the culture together with anti-CD3 antibody. For assessment of the role of common cchain cytokine on proliferation, IL-2 and IL-15 (Peprotech) were added to the culture at different concentrations together with anti-CD3.

In Vitro Suppression Assay
The 5610 4 sorted conventional T cells and T reg cells (in ratios according to figure legend) were cultured for 96 h in roundbottomed plates along with anti-CD3 antibody (2C11, 2 mg/ml) and irradiated splenocytes in RPMI 1640 medium supplemented with 10% FCS, 1% penicillin-streptomycin, and 0.5% 2-mercaptoethanol. T cell proliferation was determined by thymidine incorporation and CFSE labelling of T conv cells as described above.

In Vitro Apoptosis Assays
Lymphocytes from LN and spleens were stained for CD4, CD44, CD62L, CD25, CD45RB, and sorted by using FACS Aria. The purity for each population was above 95%. Cells were cultured in AIM-V medium (Invitrogen) supplemented with 0.05% 2-mercaptoethanol (Sigma) without stimulation. After 20 and 40 h, staining for Annexin V (BD Bioscience) and Topro3 or propidium iodide (1 mg/ml) was performed.

Quantitative Real-Time PCR Analysis
Different subsets of CD4 T cells were isolated by FACS sorting and used for mRNA extraction (RNeasy mini kit, Qiagen). mRNA was transcribed with M-MLV reverse transcriptase (Invitrogen). Quantitative RT-PCR was performed with MyIQ cycler (Biorad) using SyberGreen (Invitrogen). The following primers were used: TR2: AAC GAC TTG ACC TGT TGC CTG T and CTT CCG GGG CCA TGT ATC TT; and T-bet: CAA CAA CCC CTT TGC CAA AG and TCC CCC AAG CAG TTG ACA GT. The expression of the genes was standardized to the relative quantity of RNA polymerase 2 detected by the primers CTG GTC CTT CGA ATC CGC ATC and GCT CGA TAC CCT GCA GGG TCA and normalized to the value of the respective control condition.

Enzyme-Linked Immunosorbant Analysis (ELISA)
For detection of anti-dsDNA antibodies from the serum, ELISA plates (BD Falcon) were pretreated with 0.1% poly-L-lysine (Sigma) for 2 h, coated with 100 mg/ml DNA (Sigma) overnight, and blocked with 2% BSA for 2 h. After washing with PBS, antimouse IgG c chain HPR-conjugated antibody (Sigma) was applied for 45 min and developed with stabilised chromogen (Invitrogen).

Immunoblotting
To analyze Smad2 phosphorylation MACS sorted CD4 + T cells from tam-iCD4TR2 and control mice were stimulated with 20 ng/ml TGF-b1 (Peprotech). Western blot analysis was performed after cell lysate separation in 10% SDS-PAGE gels with anti-pSmad2 (Cell Signaling Technology) and a donkey antirabbit Ig-G HRP secondary antibody.

Statistics
The p values were calculated with Student's t test using Prism software. The p values of less than 0.05 were considered significant. Where no p value is indicated through stars, no statistically significant difference was found.  Figure 3H. Flow cytometric analysis of TR2 expression by CD4 + and CD8 + T cells from peripheral blood after long-term tamoxifen citrate treatment. (C) The scheme of the experiment described in Figure 3J-K. Flow cytometric analysis of CD4 + and CD8 + T cell frequencies in the spleen of chimeric mice at day 55 following anti-CD8a (YTS 169.4) or isotype control treatment. (TIF) Figure S4 Schemes of experimental setups and FACS analysis of TR2 deletion in lymphopenic environment. (A) Scheme of the experiment described in Figure 4A and flow cytometric analysis of TR2 expression by CD4 + and CD8 + T cells from peripheral blood after long-term tamoxifen citrate treatment (day 90). (B) Scheme of the experiment described in Figure 4C