S1P₁-mTOR axis directs the reciprocal differentiation of T_H1 and regulatory T cells

Abstract

Naïve CD4⁺ T cells differentiate into diverse effector and regulatory lineages to orchestrate immunity and tolerance. The differentiation of pro-inflammatory T_H1 and anti-inflammatory Foxp3+ regulatory T cells (Treg) was reciprocally regulated by S1P₁, a receptor for the bioactive lipid sphingosine-1-phosphate. S1P₁ inhibited extrathymic and natural Treg generation while driving T_H1 cell development in a reciprocal manner and disrupted immune homeostasis. S1P₁ signaled through mTOR and antagonized TGF-β function mainly by attenuating sustained Smad3 activity. S1P₁ function was dependent upon endogenous sphingosine kinase activity. Remarkably, two seemingly unrelated immunosuppressants FTY720 and rapamycin targeted the same S1P₁ and mTOR pathway to regulate the dichotomy between T_H1 and Treg cells. Our studies establish an S1P₁-mTOR axis that controls T cell lineage specification.

INTRODUCTION

CD4⁺ helper T cells are central regulators of adaptive immune responses. In response to antigen stimulation, naïve CD4⁺ T cells proliferate and differentiate into T helper type 1 (T_H1) cells, T_H2 cells and T_H17 cells to exert specific effector functions. These effector T cells are characterized by distinct patterns of cytokine secretion: interferon-γ (IFN-γ), interleukin 4 (IL-4) and IL-17 are the signature cytokines for T_H1, T_H2 and T_H17 cells, respectively¹. Naïve precursors can also develop into antigen-specific Foxp3⁺ T_reg cells, known as induced T_reg cells (iT_reg), which act in synergy with naturally occurring T_reg cells (nT_reg) to establish immune tolerance and counter-balance effector T cell functions^2,3. Induction of iT_reg cells in the peripheral immune compartment is closely related to the generation of T_H17 cells, as the differentiation of both lineages is dependent upon the pleiotropic cytokine transforming growth factor β (TGF-β; http://www.signaling-gateway.org/molecule/query?afcsid=A002271)⁴. Despite extensive progress in this area, it remains poorly understood how the differentiation of naïve precursors into opposing regulatory and effector lineages is regulated.

T cell differentiation is programmed by polarizing cytokines, which signal through cell surface cytokine receptors and intracellular pathways, ultimately leading to the expression of lineage-specific transcription factors and effector cytokines. IL-12, IL-4 and TGF-β drive the differentiation of T_H1, T_H2 and iT_reg cells, respectively, while stimulation by TGF-β together with IL-6 or IL-21 promotes T_H17 development¹. Differentiation of T cells is further shaped by non-cytokine receptors, especially those in the nuclear receptor superfamily that recognize endogenous metabolites, environmental toxins and other immune modulators. Retinoic acid, a vitamin A metabolite, directs reciprocal T_H17 and iT_reg differentiation by activating the retinoic acid receptor^5–7. Aryl hydrocarbon receptor, a transcription factor activated by environmental toxins, regulates T_reg and T_H17 cell differentiation in a ligand-specific fashion^8,9. Aside from the cytokine and nuclear receptors, whether other classes of receptors directly control T cell lineage choices has not been well established.

Sphingosine 1-phosphate (S1P) is a bioactive lysophospholipid enriched in blood¹⁰. S1P signals through five known G protein-coupled receptors (GPCR): S1P₁-S1P₅. FTY720, a new immunosuppressant, induces T cell sequestration in lymphoid organs by acting on four of the five S1P receptors. FTY720 is effective in relapsing multiple sclerosis and has become the first oral therapy for the debilitating autoimmune disease¹¹. However, mechanism of action for FTY270 is complex, and it remains controversial whether FTY720 acts as an agonist or a functional antagonist or both to regulate lymphocyte trafficking. Genetic approaches to alter the function of S1P₁ [http://www.signaling-gateway.org/molecule/query?afcsid=A000813] indicate that S1P₁ is the main S1P receptor that facilitates the egress of T cells from lymphoid organs^12,13. Using gain- and loss-of-function genetic systems, we recently identified a novel role of S1P₁ in the negative control of thymic generation and suppressive activity of nT_reg cells in a process dependent upon the downstream Akt-mTOR [http://www.signaling-gateway.org/molecule/query?afcsid=A000094] pathway¹⁴.

To investigate the function of S1P₁ in the lineage determination of peripheral T cells, we used a combination of genetic systems and pharmacological approaches. We found that S1P₁ inhibited differentiation of Foxp3⁺ T_reg cells while promoting the development of T_H1 cells in a reciprocal manner. S1P₁ antagonized TGF-β receptor function, through an inhibitory effect on Smad3 activity, to control the dichotomy between these two T cell lineages. Moreover, this regulatory function was dependent upon the effect of S1P₁ to sustain mTOR activation in T cells. Our studies demonstrate that differentiation of pro-inflammatory T_H1 and anti-inflammatory T_reg cells is reciprocally regulated by S1P₁-mTOR and the opposing TGF-β-Smad3 signaling.

RESULTS

S1P₁ blocks differentiation of iT_reg cells

To gain insight into mechanisms of iT_reg differentiation, we determined the role of S1P₁ in this process. First, we crossed T cell-specific S1pr1-transgenic mice (hereafter called S1P₁-Tg), which expressed 2–3 fold more S1P₁ than wild-type mice (data not shown);¹³, with OT-II transgenic mice (CD4⁺ T cell receptors (TCR) specific for ovalbumin (Ova)) on a Rag1^−/− background. Placing a TCR transgene on the Rag1^−/− background results in all peripheral T cells having single antigen specificity without detectable Foxp3 expression, obviating the confounding effects of S1P₁ function in the nT_reg compartment¹⁴. In a model of oral antigen-induced iT_reg generation^6,7, Foxp3 induction in the mesenteric lymph nodes (MLN) and spleen of S1P₁-Tg;OT-II;Rag1^−/− mice was markedly reduced as compared with the control OT-II;Rag1^−/− mice (Fig. 1a). Expression of T_reg cell specific factors LAG-3 and Gpr83 was also decreased in S1P₁-Tg;OT-II;Rag1^−/− T cells (Fig. 1b). More pronounced effects were observed in the lamina propria, in which an extensive Foxp3⁺ population was induced in Ova-fed OT-II;Rag1^−/− mice but not in S1P₁-Tg;OT-II;Rag1⁻/⁻ mice, despite similar numbers and distribution of total CD3⁺ cells in these mice (Fig. 1c). To formally exclude the contribution of the residual thymic derived nT_reg cells, we crossed S1P₁-Tg;OT-II mice with Foxp3^gfp knockin mice that marked Foxp3 expression with GFP to distinguish T_reg and conventional T cells¹⁴. We purified naive Foxp3⁻ T cells from these mice and transferred them into wild-type immunocompetent mice. Administration of oral antigen to the recipients induced a sizable Foxp3⁺ population from wild-type donors, while the Foxp3⁺ population from S1P₁-Tg donors was decreased by 70% (Supplementary Fig. 1). Taken together, S1P₁ inhibits de novo induction of Foxp3-expressing iT_reg cells from the peripheral T cell pool.

Figure 1. S1P₁ inhibits de novo generation of Foxp3⁺ iT_reg cells.

(a–c) Induction of Foxp3⁺ iT_reg cells after feeding of OT-II;Rag1^−/− and S1P₁-Tg;OT-II;Rag1^−/−mice with Ova in the drinking water for 5 days. (a) Foxp3 and CD25 expression in CD4 T cells and proportions and absolute numbers of Foxp3⁺ CD4 T cells. MLN, mesenteric lymph nodes. (b) mRNA expression of Foxp3, LAG-3 and Gpr83 in CD4 T cells isolated from MLN (n=3). (c) Distribution of Foxp3⁺ and CD3⁺ cells in the lamina propria. (d) Analysis of Foxp3 expression after naïve T cells from OT-II;Foxp3^gfp and S1P₁-Tg;OT-II;Foxp3^gfp mice were stimulated with MLN-derived CD103⁺ DCs and the Ova peptide or anti-CD3 for 5 days. (e) Analysis of Foxp3 expression after wild-type (WT) and S1P₁-Tg naïve T cells were activated with anti-CD3 and anti-CD28 in the presence of TGF-β (5 ng/ml). The right panels show proportions and absolute numbers of Foxp3⁺ and Foxp3⁻ cells (n=3). (f) Kinetics of Foxp3 mRNA expression after wild-type and S1P₁-Tg naïve cells were activated with anti-CD3 and anti-CD28 in the presence of TGF-β (5 ng/ml) for various times. Data represent three to four independent experiments.

A specialized subset of dendritic cells (DCs) in the mucosa expressing CD103 is tolerogenic by inducing Foxp3⁺ iT_reg cells^6,7. We purified CD103⁺ DCs from MLN and cultured them with naïve T cells from OT-II;Foxp3^gfp mice (CD62L^hiCD44^loFoxp3⁻) in the presence of the cognate antigen or anti-CD3. Under these conditions, a substantial population of T cells from OT-II;Foxp3^gfp mice was induced to express Foxp3, while S1P₁-Tg;OT-II;Foxp3^gfp T cells exhibited profound defects in Foxp3 induction (Fig. 1d). One key mechanism for CD103⁺ DC-mediated T_reg cell differentiation involves the production of TGF-β^6,7. Indeed, neutralizing TGF-β abrogated the function of these DCs to induce Foxp3 expression from wild-type and S1P₁-Tg cells (data not shown). We conclude that S1P₁ inhibits DC-induced iT_reg generation.

Foxp3⁺ cells can be generated directly from naïve precursors by antigen stimulation in the presence of TGF-β¹⁵. When naïve cells were activated in the presence of TGF-β in an antigen-presenting cell (APC)-free condition, those expressing the S1P₁ transgene were considerably impaired to differentiate into Foxp3⁺ iT_reg cells (Fig. 1e). Wild-type and S1P₁-Tg cells proliferated to a similar degree and comparably expressed apoptotic markers and Bcl2 (Fig. 1e and data not shown). Furthermore, crossing S1P₁-Tg mice with Bcl2-transgenic animals failed to rectify the diminished iT_reg differentiation (data not shown). Thus, the impaired iT_reg generation was not due to altered T cell proliferation or survival, indicating that S1P₁ inhibits TGF-β-mediated iT_reg differentiation. We next explored the kinetics involved in S1P₁-dependent inhibition of Foxp3 induction. In wild-type cells, Foxp3 mRNA was strongly induced at 24 h after TGF-β stimulation and continued to increase at 36–48 h. S1P₁-Tg cells showed greatly reduced Foxp3 expression at 24–48 h (Fig. 1f). In addition, transduction of wild-type cells with S1P₁-expressing retrovirus, followed by TGF-β treatment, resulted in diminished Foxp3 induction. Moreover, S1P₁ exerted its inhibitory effect on Foxp3 induction even when transduction occurred 20–24 h after TGF-β stimulation (Supplementary Fig. 2), suggesting that S1P₁ may interfere with sustained TGF-β signaling.

S1P₁ inhibits iT_reg generation and maintenance

We then tested whether S1P₁ is required to suppress iT_reg differentiation. Given the marked reduction of peripheral T cells in S1pr1^flox/flox mice crossed with Cd4-Cre transgenic mice¹⁴, we bred these mice onto the Foxp3^gfp background and purified Foxp3⁻ CD4 single-positive (CD4SP) thymocytes. When transferred into Rag1^−/− mice, a subset of donor cells became Foxp3⁺. Deficiency of S1P₁ resulted in an augmented Foxp3⁺ population, indicating increased iT_reg generation in vivo (Fig. 2a). To more directly address the requirement of S1P₁ in iT_reg development from peripheral T cells, we crossed S1pr1^flox/flox mice with Rosa26-Cre-ER^T2 (called ‘S1pr1^CreER mice’ here). After 4-hydroxytamoxifen (4-OHT) treatment, the floxed S1pr1 gene was deleted in naïve T cells (Supplementary Fig. 3a). When these cells were differentiated toward iT_reg cells by exogenous TGF-β, they exhibited a higher frequency of the Foxp3⁺ population (Fig. 2b), associated with increased Foxp3 mRNA abundance (Supplementary Fig. 3b) and normal cell survival (data not shown). Moreover, CD103⁺ DC-induced Foxp3⁺ population was also enhanced in the absence of S1P₁ (Fig. 2c). Therefore, S1P₁ deficiency directly potentiates iT_reg differentiation.

Figure 2. S1P₁ is required to restrain the generation and maintenance of Foxp3⁺ iT_reg cells.

(a) Analysis of Foxp3 expression 4 weeks after Foxp3⁻ CD4SP thymocytes from wild-type and S1pr1^flox/flox;Cd4-Cre mice were transferred into Rag1^−/− mice. PLN, peripheral lymph nodes. (b) Analysis of Foxp3 expression after naïve T cells from wild-type and S1pr1^CreER mice were treated with 4-OHT and then activated in the presence of TGF-β for iT_reg differentiation. The right panels show proportions and absolute numbers of Foxp3⁺ and Foxp3⁻ T cells. (c) Analysis of Foxp3 expression after naïve T cells from wild-type and S1pr1^CreER mice were treated with 4- OHT and then activated by CD103⁺ DCs and anti-CD3, without any exogenous cytokines. (d) Analysis of Foxp3 expression in mature iT_reg cells upon acute deletion of S1P₁. Naïve T cells from wild-type and S1pr1^CreER mice were differentiated into iT_reg cells. Foxp3⁺ (GFP⁺) cells were sorted, treated with 4-OHT and cultured with IL-2. Foxp3 expression was analyzed 4–5 days later. Data represent three independent experiments.

Continued cells, Foxp3 Foxp3 expression is necessary for T_reg function. As compared with nT_reg expression in iT_reg cells is less stable and depends upon continuous TGF-β signaling¹⁶. To test the involvement of S1P₁ in the maintenance of Foxp3 expression, we generated and sorted Foxp3⁺ (GFP⁺) iT_reg cells from S1pr1^CreER mice, and then treated them with 4-OHT to induce S1pr1 deletion in the mature iT_reg cells. Upon TGF-β withdrawal, control iT_reg cells that retained S1P₁ expression readily lost Foxp3 expression, whereas iT_reg cells deficient in S1P₁ contained a much greater Foxp3⁺ population with increased Foxp3 and CD25 expression (Fig. 2d). Conversely, Foxp3⁺ iT_reg cells generated from S1P₁-Tg mice were unable to sustain Foxp3 expression in the absence of TGF-β (Supplementary Fig. 4). Therefore, S1P₁ negatively regulates the maintenance of Foxp3 expression in iT_reg cells.

S1P₁ drives T_H1 differentiation

The inhibitory effects of S1P₁ on iT_reg induction and maintenance prompted us to investigate whether S1P₁ diverges naïve T cells into alternative lineages using three in vivo systems. First, in the oral antigen model of OT-II;Rag1^−/− mice, only a small fraction of IFN-γ⁺ cells was induced in OT-II;Rag1^−/− mice, as would be expected for T cell responses under the tolerizing conditions. In contrast, S1P₁-Tg;OT-II;Rag1^−/− mice contained a markedly increased population (5–10 fold more abundant) of IFN-γ⁺ CD4 T cells in various lymphoid organs examined (Fig. 3a). Accordingly, quantitative RT-PCR identified elevated expression of IFN-γ and the T_H1 transcriptional factor T-bet in S1P₁-Tg cells following oral antigen exposure (Fig. 3b). IL-4 and IL-17 protein and RNA abundance were low and comparable between the two types of mice (Fig. 3a and data not shown), indicating selective differentiation of S1P₁-Tg cells into the T_H1 lineage.

Figure 3. S1P₁ drives differentiation of T_H1 cells.

(a,b) Analysis of T_H1 differentiation after OT-II;Rag1^−/− and S1P₁-Tg;OT-II;Rag1^−/− mice were fed with Ova in the drinking water for 5 days. (a) Expression of IFN-γ and IL-17 and proportions of IFN-γ+ CD4 T cells. (b) Quantitative RT-PCR analysis of T-bet and IFN-γ expression in CD4 T cells (n=3). (c) Proportions of IFN-γ⁺ CD4 T cells in gated donor cells from OT-II;Foxp3^gfp and S1P₁-Tg;OT-II;Foxp3^gfp mice that were transferred to naïve wild-type mice and immunized with Ova intravenously. (d) IFN-γ expression in total CD4 and CD8 T cells from MLN of wild-type and S1P₁-Tg mice, detected by intracellular staining 5 h post PMA/ionomycin stimulation (left), and quantitative RT-PCR after 24 h stimulation with anti-CD3/CD28 (right) (n=3–4). Data represent four independent experiments.

Second, we adoptively transferred T cells from OT-II or S1P₁-Tg;OT-II mice (Thy1.1+) into naïve recipient and immunized recipients with Ova intravenously. Five days later, a substantially increased IFN-γ⁺ population was detected among S1P₁-Tg donor-derived cells, as compared with wild-type donors (Fig. 3c). Also, secretion of IFN-γ, but not of IL-4 or IL-17, was enhanced in the S1P₁-Tg cell transfer group after ex vivo peptide stimulation (data not shown).

Third, we determined the proportions of T cells capable of producing IFN-γ in S1P₁-Tg mice by intracellular cytokine staining. Compared with MLN cells from wild-type mice, which had few IFN-γ producing CD4⁺ and CD8⁺ T cells, a higher percentage of S1P₁-Tg cells produced IFN-γ (Fig. 3d). Quantitative RT-PCR revealed 7–8 fold more IFN-γ in S1P₁-Tg CD4⁺ and CD8⁺ T cells as compared with the wild-type counterparts (Fig. 3d). Taken together, S1P₁ promotes IFN-γ production and T_H1 differentiation in vivo.

To establish whether S1P₁ regulates intrinsic T cell differentiation, we sorted naïve T cells from wild-type and S1P₁-Tg mice and cultured them under nonpolarizing conditions (T_H0). A larger proportion of S1P₁-Tg cells spontaneously differentiated into IFN-γ⁺ cells associated with increased T-bet expression (Supplementary Fig. 5a and data not shown). Higher IFN-γ expression was also observed in S1P₁-Tg;OT-II cells (Supplementary Fig. 5a) and wild-type cells transduced with S1P₁-expressing retrovirus (data not shown). Conversely, deletion of S1P₁ diminished IFN-γ expression (Supplementary Fig. 5b). Therefore, S1P₁ promotes IFN-γ expression and T_H1 differentiation.

S1P₁ directs reciprocal T_H1 and iT_reg differentiation

We next determined the mechanisms for the altered iT_reg and T_H1 differentiation in S1P₁-Tg and deficient cells. TGF-β is a pleiotropic cytokine with pronounced effects on cell fate determination of multiple T cell lineages ¹⁷, but whether TGF-β directly coordinates development of T_H1 and iT_reg cells from naïve precursors is not well understood. We therefore differentiated wild-type naïve cells under T_H0 and T_H1 conditions in the presence of TGF-β and measured Foxp3 and IFN-γ expression simultaneously. As expected, a subset of cells expressed IFN-γ but few Foxp3⁺ cells were detected under nonpolarizing conditions. Addition of TGF-β induced Foxp3 and terminated IFN-γ expression. Under T_H1 conditions, most T cells became IFN-γ positive, yet addition of TGF-β decreased IFN-γ⁺ population and simultaneously induced Foxp3⁺ population, resulting in the co-existence of T_H1 and iT_reg cells (Fig. 4a). The effects of TGF-β to promote Foxp3 over IFN-γ expression were profoundly lost in S1P₁-Tg cells (Fig. 4b). In particular, when TGF-β was added to differentiating T_H1 cells to foster development of both T_H1 and iT_reg cells, the ratio of IFN-γ⁺: Foxp3⁺ cells was reversed in S1P₁-Tg cells (Fig. 4c). This reciprocal change was also observed in antigen-specific S1P₁-Tg;OT-II T cells (Fig. 4c). Conversely, deletion of S1P₁ resulted in the expansion of iT_reg cells at the expense of T_H1 cells (Fig. 4d). Therefore, S1P₁ mediates reciprocal T_H1 and iT_reg differentiation in vitro.

Figure 4. S1P₁ regulates reciprocal T_H1 and iT_reg differentiation and immune homeostasis in vivo.

(a,b) Analysis of IFN-γ and Foxp3 expression in wild-type (a) and S1P₁-Tg (b) cells differentiated under the specified conditions. (c,d) Ratios of IFN-γ⁺ and Foxp3⁺ populations in wild-type and S1P₁-Tg cells (c) and 4-OHT treated wild-type and S1pr1^CreER cells (d) that were differentiated under T_H1 conditions in the presence of TGF-β. (e,f) Analysis of IFN-γ⁺ and Foxp3⁺ populations in wild-type and S1P₁-Tg cells (e) and 4-OHT treated wild-type and S1pr1^CreER cells (f) that were activated by CD103⁺ DCs and Ova or anti-CD3, without any exogenous cytokines. (g-i) Analysis of T cell-dependent colitis and in vivo differentiation. Wild-type or S1P₁-Tg naive T cells were transferred in combination with wild-type Treg cells (CD45.1⁺) into Rag1^−/− mice. (g) Changes in body weight. (h) Representative intestine histology. (i) Proportions of Foxp3⁺ and IFN-γ⁺ CD4 T cells derived from naïve T cell donors. Data represent three independent experiments.

As an independent system of T cell differentiation, we primed T cells with tolerogenic CD103⁺ DCs, without any exogenous cytokines^6,7. Under these conditions, distinct populations of Foxp3⁺ and IFN-γ⁺ cells were generated from wild-type cells. T cells with increased S1P₁ signaling produced IFN-γ instead of Foxp3 (Fig. 4e), whereas those deficient in S1P₁ preferentially developed into Foxp3⁺ T_reg cells instead of IFN-γ⁺ T_H1 cells (Fig. 4f). These results further indicate an intrinsic role of S1P₁ in directing T_H1 and iT_reg lineage commitment.

Previous studies showed that T_H1 cells and IFN-γ are pathogenic in colitis¹⁸, whereas iT_reg cells act in synergy with nT_reg cells to establish mucosal tolerance¹⁹. To determine the in vivo relevance of S1P₁-mediated T cell differentiation, we transferred naïve T cells from wild-type or S1P₁-Tg (CD45.2⁺) mice in conjunction with wild-type congenic (CD45.1⁺) T_reg cells into Rag1^−/− mice. In this system, in situ developed iT_reg cells from naïve donors, together with nT_reg donor cells, are required to control colitis¹⁹; wild-type nT_reg cells were used here to circumvent the confounding effects in S1P₁-Tg nT_reg cells¹⁴. While there was minimal weight loss and inflammation in the wild-type co-transfer group, transfer of S1P₁-Tg naïve cells and wild-type T_reg cells resulted in significant weight loss associated with severe colitis and leukocyte infiltration (Fig. 4g,h). To dissect the underlying mechanisms, we measured expression of Foxp3 and IFN-γ in cells derived from naïve cell donors (CD45.2⁺). As compared with wild-type donor cells, we observed fewer Foxp3⁺ iT_reg cells from S1P₁-Tg mice and a significant increase of IFN-γ-expressing T cells (Fig. 4i). Therefore, S1P₁ controls the reciprocal relationship between iT_reg and T_H1 cells in vivo.

Discrete mechanisms in iT_reg and T_H1 differentiation

We next determined whether control of T_H1 and iT_reg differentiation by S1P₁ is interdependent. IFN-γ can inhibit iT_reg generation²⁰, although conflicting conclusions also exist²¹. We tested whether inhibition of iT_reg differentiation by S1P₁ is a secondary consequence of increased IFN-γ production. Retroviral transduction of S1P₁ into either Ifng^−/− or wild-type cells resulted in diminished Foxp3⁺ induction under iT_reg conditions (Fig. 5a). Also, the S1P₁ transgene downregulated iT_reg generation in both Ifng^+/+ and Ifng^−/− backgrounds (Fig. 5b). Thus, S1P₁ regulates iT_reg differentiation independent of its role in facilitating IFN-γ expression.

Figure 5. S1P₁ mediates iT_reg and T_H1 differentiation through discrete mechanisms.

(a) Analysis of Foxp3 expression after wild-type and Ifng^−/− naïve T cells were transduced with control (RV) and S1P₁-expressing retrovirus (S1P₁-RV) and activated in the presence of TGF-β differentiation. (b) Analysis of Foxp3 expression after naïve T cells from Ifng ^−/− and for iT_regS1P₁-Tg;Ifng^−/− mice were differentiated in the presence of different doses of TGF-β. (c) Analysis of IFN-γ expression in wild-type and S1P₁-Tg cells activated under TH0 conditions and transduced with control (RV) and Foxp3-expressing retrovirus (Foxp3-RV) linked with a GFP reporter; gated GFP⁺ cells are shown in the FACS plots. (d) Analysis of IFN-γ expression in Foxp3-deficient cells. Wild-type (CD45.2⁺) and Scurfy (CD45.1⁺.2⁺) bone marrow cells were mixed at 1:1, and transferred into sublethally irradiated Rag1^−/− mice. At 6 weeks after reconstitution, naïve T cells from the two donor populations were purified, activated under T_H0 conditions, and transduced with control (RV) and S1P₁-expressing retrovirus (S1P₁-RV) linked with a Thy1.1 reporter; gated Thy1.1⁺ cells are shown in the FACS plots. (e) Analysis of IFN-γ expression in Foxp3-deficient cells after naïve T cells from OT-II and S1P1-Tg;OT-II mice bred onto the Scurfy background were activated by Ova and irradiated splenic APC. Data represent three independent experiments.

Conversely, we tested whether the reduced Foxp3 expression in S1P₁-Tg cells is responsible for their enhanced IFN-γ expression. To this end, we transduced the Foxp3 gene into wild-type and S1P₁-Tg cells, and found that forced Foxp3 expression had no effects at reducing the expression of IFN-γ in S1P₁-Tg cells (Fig. 5c). To further assess the function of Foxp3 in S1P₁-mediated T_H1 differentiation, we examined whether Foxp3 deficiency affects the ability of S1P₁ to drive T_H1 differentiation. To prevent the lymphoproliferative autoimmune phenotype due to Foxp3 deficiency, we constructed mixed bone marrow chimeras by co-transferring bone marrow cells from Foxp3-deficient Scurfy mice (CD45.1⁺.2⁺) and wild-type mice (CD45.2⁺) into Rag1^−/− recipients; the presence of functional T_reg cells derived from the wild-type donors prevented autoimmune activation of Scurfy T cells¹⁴. Naïve T cells from these two donor populations were purified and transduced with S1P₁-expressing retrovirus. S1P₁ expression was equally effective to promote IFN-γ expression in wild-type and Foxp3-deficient cells (Fig. 5d). As an independent approach, we bred OT-II and S1P₁-Tg;OT-II mice with Scurfy mice, in which the TCR transgene ameliorated the confounding effects of autoimmune inflammation²². Naïve T cells from these mice were stimulated with Ova peptide, and increased T_H1 differentiation was observed in those expressing the S1P₁ transgene (Fig. 5e). Therefore, either ectopic expression or deficiency of Foxp3 did not prevent S1P₁ from driving T_H1 differentiation, suggesting that S1P₁ mediates T_H1 differentiation independent of its effects at inhibiting Foxp3 expression.

Aged S1P₁-Tg mice developed high titers of autoantibodies, concomitant with spontaneous T cell activation¹⁴. Since IFN-γ production is associated with the pathogenesis of systemic autoimmune disease, we tested whether increased IFN-γ in S1P₁-Tg cells contributes to the breakdown of immune tolerance. As compared with S1P₁-Tg mice, S1P₁-Tg;Ifng^−/− mice contained significantly reduced titers of anti-double stranded (ds) DNA antibody (data not shown). In contrast, enhanced expression of CD44 on S1P₁-Tg cells, indicative of their spontaneous activation in vivo¹⁴, was observed irrespective of the IFN-γ background (data not shown), consistent with our previous observation that spontaneous T cell activation was mainly due to defective T_reg cells in S1P₁-Tg mice14. Therefore, S1P₁ regulates immune homeostasis through both T_reg-dependent and independent mechanisms.

S1P₁ antagonizes TGF-β-Smad3 signaling

We noticed that the exacerbated colitis following S1P₁-Tg cell transfer phenocopied the abnormalities in T cells with transgenic expression of a dominant negative TGF-β receptor (CD4-dnTGFβRII) in a similar model^23,24. Moreover, mice with impaired TGF-β receptor signaling in T cells spontaneously differentiate into T_H1 cells in vivo^25,26. We hypothesize that S1P₁ mediates reciprocal T_H1 and iT_reg differentiation by antagonizing TGF-β receptor signaling. To test this hypothesis, we first compared the response of S1P₁-Tg cells with that of CD4-dnTGFβRII cells. In response to CD103⁺ DCs, both S1P₁-Tg and CD4-dnTGFβRII T cells were skewed into IFN-γ⁺ T_H1 cells rather than Foxp3+ T_reg cells (Fig. 6a). Similarly, in response to exogenous TGF-β, CD4-dnTGFβRII and S1P₁-Tg cells had similar defects in upregulating Foxp3; instead these cells expressed IFN-γ (Supplementary Fig. 6). Therefore, CD4-dnTGFβRII mice phenocopy the abnormalities seen in S1P₁-Tg mice, suggesting that S1P₁ likely affects TGF-β receptor signaling.

Figure 6. S1P1 attenuates TGF-β-Smad3 signaling.

(a) Analysis of Foxp3 expression in wild-type, CD4-dnTGFβRII, and S1P₁-Tg cells activated by CD103⁺ DCs and anti-CD3. (b) Analysis of Smad3 nuclear translocation after 2 d stimulation of wild-type and S1P₁-Tg cells. The numbers below each panel show the quantification of band intensity relative to the loading controls. (c) Analysis of Foxp3 expression in wild-type and S1P₁-Tg cells transduced with control (RV) and constitutively active Smad3 retrovirus (Smad3CA-RV) and activated in the presence of TGF-β for iT_reg differentiation. (d) Analysis of Foxp3 expression in 4-OHT treated wild-type and S1pr1^CreER cells that were activated in the presence of TGF-β and a Smad3 inhibitor SIS3 (5 μM).

We next investigated the molecular mechanisms by which S1P₁ and TGF-β receptor interact. Smad3 is an important transcription factor to mediate TGF-β effects in iT_reg generation²⁷. TGF-β stimulation resulted in prolonged activation of Smad3 in wild-type T cells (Supplementary Fig. 7a). Immediately after TGF-β stimulation, Smad3 phosphorylation was comparable between wild-type and S1P₁-Tg cells (data not shown). However, S1P₁-Tg cells were unable to maintain Smad3 phosphorylation after 16 h of TGF-β treatment, and more pronounced reduction was observed at 24–48 h of stimulation (Supplementary Fig. 7a). Accordingly, nuclear translocation of total and phosphorylated Smad3 was profoundly reduced in S1P₁-Tg cells after 2 days of stimulation (Fig. 6b). To understand the importance of sustained Smad3 activation, we transduced a dominant negative Smad3 molecule into wild-type cells at 24 h after initiation of iT_reg differentiation, which was effective at inhibiting Foxp3 induction (Supplementary Fig. 7b). Moreover, increased p-Smad3 was observed in S1P₁-deficient cells at and after 24 h of stimulation (Supplementary Fig. 8). To directly determine whether Smad3 mediates S1P₁ function, we introduced constitutively active Smad3 into S1P₁-Tg cells, which restored Foxp3 expression in S1P₁-Tg cells by up to 60% (Fig. 6c). Conversely, treatment of S1P₁-deficient T cells with the Smad3 inhibitor reduced iT_reg differentiation (Fig. 6d). Therefore, S1P₁ antagonizes TGF-β-dependent effects on T cell differentiation mainly by attenuating Smad3 signaling.

The S1P₁-mTOR axis is targeted by immunosuppressants

We further explored the signaling pathway utilized by S1P₁ to antagonize TGF-β-Smad3 signaling. The Akt-mTOR pathway has recently been shown to restrain iT_reg generation^28–31. However, S1P₁ activates mTOR in nT_reg cells but not naïve T cells immediately after TCR stimulation¹⁴. Given the effects of S1P₁ on sustained but not early Smad3 activation, we examined mTOR activity at later time points by assessing the phosphorylation of the ribosomal protein S6, a well-established target of mTOR. S1P₁-Tg cells exhibited increased amounts of p-S6 after 16 h of TCR stimulation, and more prominent increase was observed at later time points, indicating an important role for S1P₁ to sustain mTOR activation (Fig. 7a).

Figure 7. S1P₁-mTOR axis is targeted by FTY720 and rapamycin.

(a) Kinetics of mTOR activation in wild-type and S1P₁-Tg cells stimulated for various times. (b) Analysis of Smad3 and S6 activation in wild-type and S1P₁-Tg cells treated with FTY720 and rapamycin and activated for 2 days. (c,d) Analysis of Foxp3 (c) and IFN-γ expression (d) in OT-II;Rag1^−/− and S1P₁-Tg;OT-II;Rag1^−/− mice fed with Ova in the drinking water for 5 days, accompanied by daily FTY720 (1 mg/kg) or rapamycin (3 mg/kg) treatment. (e) Analysis of Foxp3 expression in thymic CD4SP cells from wild-type and S1P₁-Tg mice treated daily with FTY720 and rapamycin for a total of 5 days. Data represent three independent experiments. (f) Induction of Foxp3 in thymic T_reg precursors from wild-type and S1pr1^flox/flox;Cd4-Cre mice stimulated with IL-2 (50 U/ml) in the presence of SIS3.

This newly revealed role of S1P₁-mediated reciprocal differentiation of iT_reg and T_H1 cells led us to examine possible therapeutic implications. Both rapamycin (Sirolimus), an inhibitor of mTOR, and FTY720 have been shown to modulate iT_reg cells^28–30,32. To address whether FTY720 and rapamycin share an immunosuppressive mechanism, we first examined mTOR activation in S1P₁-Tg cells treated with FTY720 or rapamycin. FTY720 and rapamycin both decreased p-S6 (Supplementary Fig. 9). Moreover, both treatments restored p-Smad3 activity in S1P₁-Tg cells to that observed in wild-type cells (Fig. 7b). These results identified an S1P₁-mTOR axis that acts to interfere with Smad3 signaling and is targeted by immunosuppressive drugs.

To further investigate the effects of the immunosuppressants on immune responses, we administered oral Ova antigen to OT-II;Rag1^−/− mice and treated them daily with FTY720 or rapamycin. Such treatments increased iT_reg cells and reduced IFN-γ⁺ T_H1 cells (Supplementary Fig. 10). Treatment of S1P₁-Tg;OT-II;Rag1^−/− mice with FTY720 or rapamycin reversed the altered differentiation of iT_reg and T_H1 cells, to similar to untreated wild-type control mice (Fig. 7c,d). Thus, FTY720 and rapamycin modulate reciprocal T_H1 and iT_reg differentiation in vivo. To ascertain whether this reflects intrinsic effects, we treated T cells with these drugs in vitro. Treatment of S1P₁-Tg cells with FTY720 or rapamycin increased iT_reg and reduced T_H1 differentiation (Supplementary Fig. 11). Taken together, FTY720 and rapamycin directly modulate reciprocal differentiation of iT_reg and T_H1 cells by targeting the S1P₁-mTOR axis.

We next asked whether FTY720 and rapamycin also target thymic nT_reg cells. We treated wild-type and S1P₁-Tg mice with FTY720 or rapamycin for 5 days, and assessed the frequency and number of Foxp3⁺ CD4SP thymocytes. As previously reported, S1P₁-Tg mice contained decreased thymic nT_reg cells¹⁴. After FTY720 or rapamycin treatment, nT_reg cells in these mice were similar to wild-type mice (Fig. 7e). Accordingly, the diminished in vitro differentiation of S1P₁-Tg nT_reg precursors (CD25⁺Foxp3^–) into mature Foxp3⁺ cells was rescued by FTY720 and rapamycin (Supplementary Fig. 12). Therefore, FTY720 and rapamycin affect thymic development of nT_reg cells by targeting the S1P₁-mTOR axis. To further explore whether Smad3 mediates S1P₁ signaling in thymic nT_reg cell differentiation, we treated nT_reg precursors from wild-type and S1pr1^flox/flox;Cd4-Cre mice with the Smad3 inhibitor SIS3. Blocking Smad3 activity ablated the increased nT_reg differentiation observed in precursors lacking S1P₁ (Fig. 7f). Although TGF-β is required for T_reg survival in the thymus³³, SIS3 treatment reduced Foxp3 induction even in Bcl2-Tg mice (data not shown), indicating that the effect of Smad3 in nT_reg differentiation is independent of cell survival. Taken together, the drug-sensitive S1P₁-mTOR axis interferes with Smad3 signaling to control development of both nT_reg and iT_reg subsets.

Sphingosine kinases control T cell differentiation

Given an essential role of S1P₁ in T_reg cell differentiation, we explored mechanisms of S1P₁ activation. Supplement of exogenous S1P to naïve T cells did not alter their differentiation into iT_reg cells (Fig. 8a). We thus tested whether the endogenously produced S1P is involved. S1P is synthesized by one of the two sphingosine kinases (SphKs) on the substrate sphingosine³⁴. In differentiating T cells, SphK1 was strongly upregulated by the treatment of TGF-β (Fig. 8b), suggesting the likely involvement of intrinsic SphK activity in iT_reg cell generation. To test this hypothesis, we treated naïve T cells with N,N,-dimethylsphingosine (DMS) or sphingosine kinase inhibitor (SKI), two widely used inhibitors of SphK activity. Blocking SphK activity strongly upregulated Foxp3 induction (Fig. 8c). Treatment of S1P₁-Tg cells with DMS or SKI restored their abilities to differentiate into iT_reg cells (Fig. 8d). Similarly, blocking SphK activity considerably rectified the defects of S1P₁-Tg cells in the reciprocal T_H1 and T_reg cell differentiation (Supplementary Fig. 13a). Thus, T cell differentiation mediated by S1P₁ is dependent upon intrinsic SphKs, whose expression is upregulated by TGF-β as a feedback mechanism to limit TGF-β responses (Supplementary Fig. 13b).

Figure 8. SphK activity regulates T cell differentiation.

(a) Analysis fo Foxp3 expression in wild-type cells treated with various concentrations of S1P in a serum-free medium. (b) Analysis of SphK1 and SphK2 mRNA expression in cells activated in the absence or presence of TGF-β. Levels in naïve T cells were set to 1. (c) Analysis of Foxp3 expression after naïve T cells were pre-treated with DMS or SKI, and activated under iT_reg conditions for 5 days. (d) Analysis of Foxp3 and CD25 expression in wild-type and S1P1-Tg naïve T cells were pre-treated with DMS (0.25 μM) or SKI (2.5 μM), and activated under iT_reg conditions for 5 days. Data represent three independent experiments.

DISCUSSION

Our results demonstrate that differentiation of T_H1 and T_reg cells is reciprocally regulated. S1P₁ is a switch factor that drives the development of T_H1 cells at the expense of T_reg generation. S1P1 function is dependent upon endogenous SphK activity, and it signals through mTOR and intersects with TGF-β signaling mainly by attenuating the sustained Smad3 activity. Although TGF-β and mTOR have pleiotropic functions in regulating multiple CD4 T cell lineages^17,30, the interplay between S1P₁-mTOR and TGF-β-Smad3 signaling with dynamic kinetics selectively controls the reciprocal differentiation of T_H1 and T_reg cells. We propose that this novel lineage commitment process controls the balance between immunity and tolerance and contributes to mechanisms of action for immunosuppressive therapy.

The dichotomy of T_H1 and T_reg cell fate determination is reminiscent of the reciprocal differentiation of iT_reg and T_H17 cells ^1,4. Moreover, the effect on T cell differentiation by S1P₁, a G protein-coupled receptor for the endogenous bioactive lipid, is analogous to the control of reciprocal T_reg and T_H17 differentiation mediated by non-cytokine receptors such as retinoid acid receptor and aryl hydrocarbon receptor that recognize certain endogenous metabolites and environmental toxins^5–9. These results collectively establish a new paradigm of T cell lineage specification controlled by non-cytokine immune modulators. Unlike inhibition of T_H17 cell differentiation by Foxp3^22,35,36, Foxp3 ectopic expression or complete ablation did not prevent S1P₁ from driving T_H1 differentiation. Moreover, S1P₁ blocked iT_reg generation independent of its effects to drive T_H1 differentiation. Thus, S1P₁ regulates T_H1 and iT_reg differentiation through discrete mechanisms, further supporting the reciprocal nature of these two lineages.

Elevated S1P₁ expression is associated with various autoimmune diseases³⁷. A direct role for S1P₁ in immune homeostasis is highlighted by the development of autoimmunity in aged S1P₁-Tg mice¹⁴, which we attributed to defects in both T_reg and naïve T cells. Also, transfer of either regulatory or naïve S1P₁-Tg cells into Rag1^−/− mice exacerbated the development of colitis¹⁴. These observations are reminiscent of TGF-β-dependent immune homeostasis via T_reg-dependent and independent mechanisms²⁵. Interestingly, premature egress of CD4⁺CD8⁺ thymocytes into peripheral lymphoid organs and tissues has been observed in another strain of S1P₁-transgenic mice that utilized an artificial promoter to drive the transgene expression in both T and B cells³⁸. We did not observe these alterations in multiple lines of founder mice expressing the S1P₁ transgene throuhg the T cell-selective CD2 promoter/enhancer¹³ (data not shown). The simplest interpretation for the discordance is the use of different promoters for the transgene expression. Importantly, diminished thymic nT_reg cells were apparent in both types of transgenic mice^14,38, highlighting the key function of S1P₁ in T_reg cell differentiation.

S1P₁ attenuates TGF-β-Smad3 function via signaling through mTOR. Even though Smad3 is activated within minutes after TGF-β stimulation, the sustained Smad3 activity, which is antagonized by the S1P₁-mTOR axis, is essential for iT_reg differentiation. This is in agreement with the observed 48 h delay in Foxp3 expression following TGF-β treatment¹⁶. The effect of S1P₁ on mTOR signaling in naive T cells is distinct from its function in nT_reg cells, in which it facilitates immediate mTOR activation¹⁴, suggesting cell context-specific regulation of mTOR signaling. Despite this, the S1P₁-mTOR axis negatively regulates the differentiation of both thymic nT_reg cells and extrathymic iT_reg cells, via antagonizing TGF-β-Smad3 signaling. Although the roles of TGF-β in nT_reg development remain controversial^33,39, a requirement for Smad3 signaling in this process is established by the significant reduction of thymic nT_reg cells in Smad3^−/− mice⁴⁰. This probably reflects the engagement of multiple and potentially opposing signaling pathways by TGF-β, as well as the pleiotropic effects of TGF-β on T cell development and homeostasis¹⁷. Taken together, while diverse mechanisms promote the generation of nT_reg and iT_reg cells³, a common pathway involving S1P₁-mTOR and the antagonistic interactions with TGF-β-Smad3 signaling negatively controls the differentiation of both T_reg subsets.

mTOR signaling is necessary for the differentiation of multiple effector T cells including T_H1, T_H2 and T_H17 cells while at the same time, inhibiting iT_reg generation³⁰. Here we found that the S1P₁-mTOR axis selectively controls T_H1 and T_reg differentiation, suggesting that other receptors may mediate mTOR activation for T_H2 and T_H17 differentiation. Although previous reports showed that a human S1P₁ transgene is capable of promoting T_H2 and T_H17 cell generation in vitro^41,42, we found no evidence of S1P₁ involvement in these processes upon antigen stimulation in vivo. Whether S1P₁ regulates these processes under more defined conditions requires further testing. Additional molecules involved in Akt-mTOR signaling, including Rictor (mTORC2), Cbl-b, Foxo1 and Foxo3, has recently been shown to modulate T cell responses^30,43,44. Given the potent effects of S1P₁ on T_reg cell induction and maintenance and T_H1 cell generation, S1P₁ may engage multiple pathways associated with Akt-mTOR. Furthermore, direct interactions between Smad3 and the Akt, mTOR and Foxo pathways have been observed outside the immune system^45,46, suggesting that Smad3 may integrate these diverse pathways to direct T cell differentiation.

S1P is maintained at high levels in the blood and lymph by the actions of SphKs but kept at low levels in lymphoid organs by S1P lyase. Separate sources appear to provide S1P to plasma and lymph⁴⁷, and a recent study indicates that neural crest-derived pericytes produce S1P to induce thymocyte egress^38. Combining pharmacological and genetic approaches, we found that S1P₁-mediated regulation of T cell differentiation requires endogenous SphK activity. Although exogenously added S1P has no apparent effects on T_reg cell generation, we cannot fully exclude the contribution of the exogenous source of S1P in this process, because partioning of S1P distribution might not be recapitulated in the isolated cell cultures. Nonetheless, our results clearly demonstrate a key role for the endogenously produced S1P in T cell differentiation in autocrine and/or paracrine manners. Moreover, SphK1 expression is strongly upregulated by TGF-β stimulation, which likely serves as a feedback mechanism to restrain prolonged TGF-β effects on Foxp3 expression. In growth factor-activated cells, S1P has been found to be produced and secreted out of the cell to activate S1P receptors in its vicinity in a model known as “inside-out” signaling of S1P³⁴. To our knowledge, our work represents the first time to identify the similar modes of action for S1P in T cell responses.

We further demonstrated that S1P₁ links fundamental processes of T cell lineage commitment to immunosuppressive therapy. FTY720 and rapamycin are thought to affect distinct molecular and cellular pathways in T cells by acting on S1P₁ to induce lymph node sequestration and on mTOR to block cell cycle entry, respectively^10,31. Here we found that these two potent immunosuppressants target the same S1P₁-mTOR axis to modulate the balance between pro-inflammatory T_H1 and anti-inflammatory T_reg cells in vivo. These effects are likely to be T cell intrinsic, as we were able to recapitulate their functions using in vitro culture systems. Therefore, regulation of T cell lineage choices by the S1P₁-mTOR axis can be further explored to develop novel therapeutics for autoimmunity and transplant rejection.