mTORC1 couples immune signals and metabolic programming to establish T_reg cell function

Abstract

The mechanistic target of rapamycin (mTOR) pathway integrates diverse environmental inputs, including immune signals and metabolic cues, to direct T cell fate decisions¹. Activation of mTOR, comprised of mTORC1 and mTORC2 complexes, delivers an obligatory signal for proper activation and differentiation of effector CD4⁺ T cells^2,3, whereas in the regulatory T cell (T_reg) compartment, the Akt-mTOR axis is widely acknowledged as a crucial negative regulator of T_reg de novo differentiation^4–8 and population expansion⁹. However, whether mTOR signaling affects the homeostasis and function of T_regs remains largely unexplored. Here we show that mTORC1 signaling is a pivotal positive determinant of T_reg function. T_regs have elevated steady-state mTORC1 activity compared to naïve T cells. Signals via T cell receptor (TCR) and IL-2 provide major inputs for mTORC1 activation, which in turn programs suppressive function of T_regs. Disruption of mTORC1 through T_reg-specific deletion of the essential component Raptor leads to a profound loss of T_reg suppressive activity in vivo and development of a fatal early-onset inflammatory disorder. Mechanistically, Raptor/mTORC1 signaling in T_regs promotes cholesterol/lipid metabolism, with the mevalonate pathway particularly important for coordinating T_reg proliferation and upregulation of suppressive molecules CTLA-4 and ICOS to establish T_reg functional competency. In contrast, mTORC1 does not directly impact the expression of Foxp3 or anti- and pro-inflammatory cytokines in T_regs, suggesting a non-conventional mechanism for T_reg functional regulation. Lastly, we provide evidence that mTORC1 maintains T_reg function partly through inhibiting the mTORC2 pathway. Our results demonstrate that mTORC1 acts as a fundamental ‘rheostat’ in T_regs to link immunological signals from TCR and IL-2 to lipogenic pathways and functional fitness, and highlight a central role of metabolic programming of T_reg suppressive activity in immune homeostasis and tolerance.

The evolutionarily conserved mTOR signaling couples cell growth and proliferation to nutrient availability and metabolic cues¹⁰. To investigate the function of mTORC1 in naturally occurring T_regs, we compared mTORC1 activity between T_regs and naïve T cells at steady state. T_regs had elevated phosphorylation of S6 and 4E-BP1, two major mTORC1 downstream targets (Fig. 1a and Supplementary Fig. 1a–c), whereas STAT5 phosphorylation was similar between these cells (Supplementary Fig. 1d). This finding is consistent with a recent study describing elevated S6 phosphorylation in T_regs compared to non-T_regs¹¹. T_regs also contained higher abundance of CD71 (the transferrin receptor) and to a lesser extent, CD98 (a subunit of L-amino acid transporter), key nutrient receptors that depend upon mTORC1 activity for expression¹², (Fig. 1b). Previous studies have demonstrated a requirement of mTORC1 for mitochondrial metabolism¹³, the dysregulation of which could affect homeostasis of memory T cells¹⁴ and hematopoietic stem cells¹⁵. T_regs had reduced mitochondrial membrane potential, whereas their mitochondrial mass and reactive oxygen species (ROS) production were largely comparable to naïve cells (Supplementary Fig. 1e). Thus, T_regs exhibit distinct regulation of mTORC1 activity and metabolism under steady state.

Figure 1.

mTORC1 signaling is constitutively active in T_regs and its disruption results in a fatal early-onset inflammatory disorder. a, Comparison of phosphorylation of S6 and 4E-BP1 between naïve T cells (CD4⁺CD44^loFoxp3⁻) and T_regs (CD4⁺Foxp3⁺; from Foxp3^YFP-Cre mice). b, Comparison of CD71 and CD98 expression between naïve T cells and T_regs. c, T_regs from C57BL/6J mice were activated with anti-CD3 or anti-CD3 and IL-2 for 3 days, washed and then used for in vitro suppression assays at multiple T_reg vs T_eff ratios; freshly isolated T_regs were included for comparison (shown only for 1:4 and 1:8 ratios). Error bars represent s.d. (n=3). d, Comparison of phosphorylation of S6 and 4E-BP1 and expression of CTLA-4 and ICOS between freshly isolated and pre-activated T_regs. e, Images of 42-day-old wild-type (WT) and Foxp3^creRaptor^fl/fl mice. Arrows indicate the scaly tail and ulceration of the body. f, Hematoxylin and eosin staining of colon, lung, liver and skin from 21-day-old wild-type and Foxp3^creRaptor^fl/fl mice. g, IFN-γ, IL-17 and IL-4 production in CD4⁺ and IFN-γ production in CD8⁺ cells from wild-type and Foxp3^creRaptor^fl/fl mice. P values are determined by Mann Whitney test. Mean fluorescent intensity (MFI) is presented above the plots of the relevant staining (a, b, d). Results represent 5 (a, g), 3 (b) and 2 (c-f) independent experiments.

One of the hallmarks of T_regs is that they are anergic in vitro upon TCR stimulation alone¹⁶, but are highly proliferative in vivo^17,18. Whereas reversal of anergy has been suggested to impair T_reg function¹⁶, T_regs stimulated by TCR together with IL-2, or under lymphopenic conditions, possess enhanced suppressive activity^18,19. Indeed, prior stimulation of T_regs with anti-CD3 upregulated their suppressive activity, and this was further enhanced by concomitant treatment with IL-2 (Fig. 1c). Such functional enhancements were associated with increased mTORC1 activity and expression of CTLA-4 and ICOS, important effector molecules mediating T_reg function^20,21 (Fig. 1d). Compared with IL-2, CD28-mediated co-stimulation had more modest enhancing effects on these events (Supplementary Fig. 1f,g). Thus, TCR and IL-2 represent predominant signals at promoting mTORC1 activity and suppressive function of T_regs.

Although these results indicate that mTORC1 is positively correlated with T_reg suppressive activity, a number of recent studies have revealed a negative role of mTORC1 in T_regs, including suppression of T_reg differentiation^2,4–8 and population expansion⁹. To establish the physiological relevance of mTORC1 signaling in T_regs, we deleted Raptor, an obligatory component of mTORC1 complex¹⁰, in T_regs by crossing Raptor^fl/fl mice with Foxp3^YFP-Cre mice²² (designated Foxp3^creRaptor^fl/fl). Raptor expression and S6 and 4E-BP1 phosphorylation were abolished in Raptor-deficient T_regs, but were normal in Foxp3⁻ populations (Supplementary Fig. 2a–c). Additionally, a slight increase of phosphorylation of Akt Ser473 was observed in mutant T_regs, whereas phosphorylation of Akt Thr308 and Erk in T_regs was not affected (Supplementary Fig. 2d). Thus, deletion of Raptor specifically abrogated mTORC1 activity.

Surprisingly, Foxp3^creRaptor^fl/fl mice developed profound inflammatory diseases, indicated by a reduced body size and hunched posture, crusting of ears, eyelids and tail, and skin ulceration particularly on the head and upper back (Fig. 1e). Foxp3^creRaptor^fl/fl mice showed extensive lymphadenopathy (Supplementary Fig. 3a) and infiltrations of lymphocytes and myeloid cells in colon mucosa, lung, liver sinusoids, skin (Fig. 1f), among other organs, before they died at a young age (Supplementary Fig. 3b). Foxp3^creRaptor^fl/fl mice had increased CD8⁺ cell percentage and altered CD4⁺/CD8⁺ ratio, highly elevated memory/effector phenotype (CD44^hiCD62L^lo) T cells, and expansion of CD11b⁺ granulocytes and macrophages (Supplementary Fig. 3c–e). Further, T cells from Foxp3^creRaptor^fl/fl mice showed increased IFN-γ, IL-4 and IL-17 producing CD4⁺ cells and IFN-γ producing CD8⁺ cells (Fig. 1g). These phenotypes are reminiscent of those observed in mice carrying scurfy mutation²³, indicating a loss of T_reg function.

Despite severe autoimmune diseases, Foxp3^creRaptor^fl/fl mice had increased Foxp3⁺ T_regs (Supplementary Fig. 4a). Lineage tracing experiments indicated that the increase of T_regs was mainly ascribed to a more abundant population of canonical T_regs, as indicated by concomitant expression of YFP-Foxp3 and Cre activity (Fig. 2a). Raptor-deficient T_regs contained normal Foxp3 expression and either increased or normal expression of many T_reg signature molecules examined, whereas CD62L expression was reduced (Supplementary Fig. 4b). Thus, in settings of immune activation, Raptor-deficient T_regs acquired activated phenotypes. Additional analysis revealed highly reduced mitochondrial membrane potential and mass and ROS production in mutant T_regs, whereas in vitro suppressive activity appeared largely undisturbed (Supplementary Fig. 4b,c). To test function of Raptor-deficient T_regs in vivo, we used a model of colitis induced by the transfer of T_eff (CD4⁺CD45RB^hiCD25⁻) cells into Rag1^−/− hosts, which could be prevented by cotransfer with wild-type T_regs. However, cotransfer with Raptor-deficient T_regs failed to inhibit colitis (Fig. 2b), or the expansion and IFN-γ production of T_eff cells (Supplementary Fig. 4d). Therefore, Raptor is required for suppressive function of T_regs in vivo.

Figure 2.

Raptor deletion abrogates T_reg suppressive activity. a, Foxp3-YFP and GFP expression in CD4⁺ T cells from 25-day-old Foxp3^creRaptor^+/+ROSA26^GFP and Foxp3^creRaptor^fl/flROSA26^GFP mice. b, Representative colon histology in Rag1^−/− mice given T_eff cells alone or in combination with wild-type or Foxp3^creRaptor^fl/fl T_regs. c, Percentages of T_regs in mixed chimeras reconstituted with bone marrow cells from CD45.1⁺ mice and wild-type or Foxp3^creRaptor^fl/fl mice (n=3). d, Expression of T_reg signature molecules and mitochondrial membrane potential (tetramethylrhodamine methyl ester, TMRM) and mass (Mitotracker) in the reconstituted chimeras. e, Expression of T_reg signature molecules in 6-week-old wild-type and Cd4^creRaptor^fl/fl mice. f, In vitro suppression assays mediated by T_regs from wild-type and Cd4^creRaptor^fl/fl mice. Error bars represent s.d. (n=3). g, Representative colon histology in Rag1^−/− mice given T_eff cells alone or in combination with wild-type or Cd4^creRaptor^fl/fl T_regs. P values are determined by Mann Whitney test (c) or t-test (f). Results represent 3 (a, d, f), 2 (b, c, g) and 5 (e) independent experiments.

We speculated that the increased number and heightened activation phenotypes of T_regs in Foxp3^creRaptor^fl/fl mice could be due to a compensatory response to the ongoing inflammation²⁴. To test this, we generated mixed bone marrow chimeras by reconstituting sub-lethally irradiated Rag1^−/− mice with bone marrow cells from CD45.1⁺ mice mixed with those from either wild-type or Foxp3^creRaptor^fl/fl mice. Raptor-deficient T_regs were under-represented in the chimeras (Fig. 2c), with impaired expression of CTLA-4 and ICOS but largely normal CD25 levels and mitochondrial parameters (Fig. 2d). These findings were recapitulated in 10-day-old Foxp3^creRaptor^fl/fl mice that exhibited minimal immune activation (Supplementary Fig. 4e). Thus, in a disease-free environment, loss of Raptor in T_regs renders them a competitive disadvantage and compromises expression of CTLA-4 and ICOS. In contrast, the pronounced mitochondrial defects of T_regs from Foxp3^creRaptor^fl/fl mice (Supplementary Fig. 4b) were not intrinsic due to Raptor deficiency, but were likely a compensatory response, as observed in hematopoietic stem cells lacking the metabolic sensor LKB1¹⁵.

To facilitate mechanistic studies of Raptor-deficient T_regs without the influence of ongoing inflammation, we generated Cd4-Cre;Raptor^fl/fl (Cd4^creRaptor^fl/fl) mice, in which Raptor was deleted in all αβ T cells. Cd4^creRaptor^fl/fl mice showed a small reduction of T_reg numbers in peripheral lymph nodes but not the spleen (Supplementary Fig. 5a), and immunoblots confirmed the loss of S6 and 4E-BP1 phosphorylation in these T_regs (Supplementary Fig. 5b). Expression of CTLA-4, ICOS and CD71 was impaired in T_regs from Cd4^creRaptor^fl/fl mice, whereas the levels of Foxp3, CD25 and other markers were normal (Fig. 2e and Supplementary Fig. 5c). In vitro assays revealed loss of suppressive activity of T_regs from Cd4^creRaptor^fl/fl mice (Fig. 2f). These T_regs also exhibited severely impaired in vivo suppressive activity, because they failed to suppress colitis or IFN-γ production mediated by T_eff cells (Fig. 2g and Supplementary Fig. 5d). Analysis of mixed chimeras comprised of CD45.1⁺ and Cd4^creRaptor^fl/fl cells confirmed that Raptor-deficient T_regs were reduced in percentage with impaired CTLA-4 and ICOS expression (Supplementary Fig. 5e). Thus, T_regs deficient in mTORC1 have an intrinsic defect in suppressive activity. Nonetheless, Cd4^creRaptor^fl/fl mice exhibited normal health status, which was likely due to impaired activation of conventional T cells (to be described elsewhere). We found a similar requirement of Raptor in T_reg suppressive activity following Cre-ER^T2-mediated acute deletion in adult mice (Supplementary Fig. 6).

Cytokines produced by T_regs represent an important immunosuppressive mechanism²⁵. The anti-inflammatory cytokines TGF-β1 and IL-10 were expressed normally in Raptor-deficient T_regs (Supplementary Fig. 7a,b). Although T_regs from Foxp3^creRaptor^fl/fl mice expressed increased IFN-γ that may compromise T_reg function²⁶, T_regs from Cd4^creRaptor^fl/fl did not (data not shown), thereby excluding such phenotypes as cell-intrinsic. Moreover, ablation of IFN-γ did not ameliorate the immune activation phenotypes of Foxp3^creRaptor^fl/fl mice (Supplementary Fig. 7c). Thus, the severe autoimmune phenotypes of Foxp3^creRaptor^fl/fl mice likely occur independently of cytokine dysregulation.

We therefore focused on the effects of Raptor on T_reg surface effector molecules especially CTLA-4 and ICOS, which were highly reduced in T_regs from multiple Raptor-deficient genetic models in an intrinsic manner (Fig, 2d,e and Supplementary Fig. 4e,5e). Moreover, CTLA-4 and to a lesser extent, ICOS, were further increased upon in vitro stimulation, and this upregulation correlated with T_reg proliferation. Such upregulation was blunted in T_regs from Cd4^creRaptor^fl/fl mice, associated with impaired proliferation (Fig. 3a and Supplementary Fig. 8a). Moreover, when wild-type T_regs were transferred into unmanipulated CD45.1⁺ mice, a proportion of them divided and upregulated CTLA-4 (Fig. 3b). However, T_regs from Cd4^creRaptor^fl/fl or Foxp3^creRaptor^fl/fl mice failed to proliferate or upregulate CTLA-4 upon transfer into CD45.1⁺ mice (Fig. 3b and Supplementary Fig. 8b). Markedly diminished homeostatic proliferation and CTLA-4 upregulation were also observed when Raptor-deficient T_regs were transferred into Rag1^−/− mice (Supplementary Fig. 8c). Altogether, loss of Raptor inhibits T_reg proliferation in vitro and in vivo. Furthermore, the proliferative defect was associated specifically with impaired CTLA-4 and ICOS expression, whereas apoptosis or other activation-induced events such as upregulation of GATA-3 and CD103 were independent of Raptor (data not shown). Thus, mTORC1 is important for coordinating T_reg proliferation and CTLA-4 and ICOS expression that may collectively contribute to T_reg suppressive function.

Figure 3.

Raptor coordinates T_reg proliferation and effector molecule expression by orchestrating cholesterol/lipid biosynthetic metabolism, especially the mevalonate pathway. a, CFSE-labeled T_regs from wild-type and CD4^creRaptor^fl/fl mice were stimulated with anti-CD3, anti-CD28 and IL-2 for 3 days, followed by analysis for CFSE dilution and CTLA-4 and ICOS expression (MFI presented above the plots). b, Celltrace violet-labeled T_regs from wild-type and Foxp3^creRaptor^fl/fl mice were transferred into unmanipulated CD45.1⁺, and Celltrace violet dilution and CTLA-4 expression were analyzed 7 days later. c, Gene-set enrichment analysis reveals the under-representation of cholesterol biosynthesis genes in Raptor-deficient T_regs. The lower part of the plot shows the distribution of the genes in the cholesterol biosynthesis signature gene set (‘Hits’) against the ranked list of genes, and the list on the right shows the top hit genes. d, Real-time PCR quantification of lipogenic gene expression in wild-type and Cd4^creRaptor^fl/fl freshly isolated T_regs or those stimulated with anti-CD3, anti-CD28 and IL-2 for 8 hours. e, De novo lipogenesis of T_regs from wild-type and Cd4^creRaptor^fl/fl mice after 24 hours of anti-CD3, anti-CD28 and IL-2 stimulation, with the final 4 hours labeled with ¹⁴C-glucose. f, Total cholesterol level was measured using lysates of activated T_regs from wild-type and Cd4^creRaptor^fl/fl mice. g, In vitro suppression assays mediated by T_regs previously activated by anti-CD3, anti-CD28 and IL-2 for 3 days in the presence of vehicle, simvastatin (2 μM), simvastatin and mevalonate (100 μM) or mevalonate alone. Error bars represent s.d. (n=3). Results represent 2 independent experiments except for c (n=4 from one experiment).

To explore Raptor/mTORC1-dependent molecular mechanisms, we used functional genomics and compared gene expression profiles of T_regs from Cd4^creRaptor^fl/fl and wild-type mice. Gene-set enrichment analysis revealed that cholesterol biosynthesis was the most significantly altered pathway in a Raptor-dependent manner: Expression of many genes in this pathway was downregulated after Raptor deletion (Fig. 3c and Supplementary Fig. 9a), including those encoding 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), squalene epoxidase (SQLE) and isopentenyl-diphosphate delta isomerase 1 (IDI-1), which were verified by real-time PCR analysis (Fig. 3d). Further, upon TCR stimulation, Raptor-deficient T_regs failed to efficiently synthesize lipids from glucose (Fig. 3e), and had reduced cellular cholesterol (Fig. 3f). These results establish a role of mTORC1 in promoting cholesterol/lipid biosynthesis. Given the interconnected relationship between the biosynthetic and bioenergetic programs²⁷, we determined the role of mTORC1 in T_reg bioenergetics by measuring extracellular acidification rate (ECAR) and oxygen consumption rate (OCR), which denoted glycolysis and mitochondrial respiration, respectively. Raptor-deficient T_regs showed decreased ECAR and OCR (Supplementary Fig. 9b,c), indicating a requirement for mTORC1 in bioenergetics. We therefore determined de novo lipid synthesis rate by measuring the incorporation of ¹⁴C-acetate into cellular lipids, thereby bypassing the requirement of mTORC1 in glycolysis or mitochondrial activity. Indeed, Raptor deficiency diminished synthesis of lipids from ¹⁴C-acetate in T_regs (Supplementary Fig. 9d). These results demonstrated a specific role of mTORC1 in orchestrating the lipogenic program.

To test functional significance of lipid metabolism, we activated T_regs in the presence of 25-hydroxycholesterol, and this general lipid synthesis inhibitor potently blocked T_reg suppressive activity (Supplementary Fig. 10a). Direct inhibition of HMGCR, the rate-limiting enzyme in the synthesis of cholesterol and isoprenoid lipids, by simvastatin also impaired T_reg suppressive activity. Importantly, simvastatin-induced inhibition was completely reversed by the simultaneous addition of mevalonate, the metabolite downstream of HMGCR (Fig. 3g). Similar effects were observed after treatments with atorvastatin and lovastatin (Supplementary Fig. 10b). The inhibitory effects of these agents on T_reg function were associated with impaired T_reg proliferation and effector molecule upregulation (Supplementary Fig. 10c), in a mevalonate-dependent manner (Supplementary Fig. 10d). Further, proliferation and CTLA-4 and ICOS upregulation in T_regs transferred into congenic CD45.1⁺ mice were diminished by statin treatment in vivo (Supplementary Fig. 11). Altogether, Raptor/mTORC1 signaling promotes the lipogenic program, with the mevalonate pathway particularly important for mediating T_reg proliferation, CTLA-4 and ICOS upregulation, and functional fitness.

Moreover, lipid metabolism-dependent regulation of CTLA-4 and ICOS is also operative in conventional T cells, because naïve T cells treated with inhibitors of this pathway or deficient in Raptor failed to effectively upregulate these molecules following TCR stimulation (Supplementary Fig. 12). We speculate this constitutes a feedback mechanism to account for the immunosuppressive activity of activated non-T_reg cells previously described^28,29, a notion that awaits further investigation.

Crosstalk between mTORC1 and mTORC2 signaling pathways has been reported¹⁰, although molecular details and functional significance are unclear. Notably, mTORC2 contributes to Akt activation that in turn signals to Foxo1²⁶, and acts in synergy with mTORC1 to negatively control generation of induced T_regs². We therefore examined the role of mTORC2 and its interaction with mTORC1 in naturally occurring T_reg cells. T_reg-specific deletion of Rictor, the defining component of mTORC2, resulted in a small reduction of T_regs in the periphery (Fig. 4a). However, loss of Rictor in T_regs did not perturb T cell homeostasis, cellularity or cytokine expression at steady state (Fig. 4b and Supplementary Fig. 13a,b), or the health status of the mice. Also, Rictor-deficient T_regs had normal expression of T_reg signature molecules, mitochondrial activity and in vitro suppressive activity (Supplementary Fig. 13c,d). Thus, in contrast to a crucial requirement of mTORC1 in programming T_reg activity, mTORC2 likely contributes to T_reg maintenance but is dispensable for T_reg function.

Figure 4.

Deletion of Rictor does not alter T_reg function but partially rescues inflammation in Foxp3^creRaptor^fl/fl mice. a, T_reg percentage in 2–3 month old wild-type and Foxp3^creRictor^fl/fl mice. b, Expression of CD62L and CD44 on splenic T cells from wild-type and Foxp3^creRictor^fl/fl mice. c, T_regs were activated by anti-CD3 and IL-2 for 4 hours followed by immunoblots. d, Expression of CD62L and CD44 expression on splenic CD4⁺ T cells from 3–4 week old wild-type, Foxp3^creRaptor^fl/fl and Foxp3^creRaptor^fl/flRictor^fl/fl mice. Right, percentage of CD62L^hiCD44^lo naïve CD4⁺ cells in the spleen. e, IFN-γ production in splenic CD4⁺ T cells from wild-type, Foxp3^creRaptor^fl/fl and Foxp3^creRaptor^fl/flRictor^fl/fl mice. f, Numbers of total TCRβ⁺ cells in peripheral lymph nodes from wild-type, Foxp3^creRaptor^fl/fl and Foxp3^creRaptor^fl/flRictor^fl/fl mice. g, In vitro suppression assays mediated by T_regs from 3–4 week old wild-type, CD4^creRaptor^fl/fl and CD4^creRaptor^fl/flRictor^fl/fl mice. Error bars represent s.d. (n=3). h, CTLA-4 expression in T_regs from the spleen of wild-type, CD4^creRaptor^fl/fl and CD4^creRaptor^fl/flRictor^fl/fl mice. P values are determined by Mann Whitney test (a) and ANOVA (d, f, g). Results represent 3 (a-c, e-g) and 2 (d, h) independent experiments.

To examine the crosstalk between the two mTOR complexes, we analyzed mTORC2 activity in Raptor-deficient T_regs. Phosphorylation of well-established mTORC2 targets, Akt Ser473 and Foxo1/3, was elevated in Raptor-deficient T_regs after anti-CD3/IL-2 (Fig. 4c) or anti-CD3/CD28 stimulation (Supplementary Fig. 14), consistent with the observation in freshly isolated Raptor-deficient T_regs (Supplementary Fig. 2d). In contrast, Akt Ser473 phosphorylation was nearly abolished after Rictor deletion (Fig. 4c and Supplementary Fig. 14a). Thus, mTORC1 institutes a negative feedback on mTORC2 in T_regs. To determine the contribution of mTORC2 activity to Foxp3^creRaptor^fl/fl phenotypes, we generated Foxp3^creRaptor^fl/flRictor^fl/fl mice. Compared with T cells in Foxp3^creRaptor^fl/fl mice, those in Foxp3^creRaptor^fl/flRictor^fl/fl mice contained an increased percentage of naïve phenotypes (Fig. 4d) and showed modestly reduced IFN-γ production (Fig. 4e). Also, the lymphadenopathy phenotype observed in Foxp3^creRaptor^fl/fl mice was ameliorated in Foxp3^creRaptor^fl/flRictor^fl/fl mice (Fig. 4f). To test the intrinsic suppressive activity of Raptor/Rictor-deficient T_regs, we generated Cd4^creRaptor^fl/flRictor^fl/fl mice that eliminated both mTORC1 and mTORC2 activities (Supplementary Fig. 15a,b). In vitro suppression assay revealed an intermediate phenotype of Raptor/Rictor-deficient T_regs compared with Raptor-deficient and wild-type cells (Fig. 4g). However, impaired CTLA-4 expression and TCR-induced proliferation upon Raptor deletion were not restored by the concomitant loss of Rictor (Fig. 4h and Supplementary Fig. 15c). Further, Foxp3^creRaptor^fl/flRictor^fl/fl mice still succumbed to an inflammatory disorder, albeit with a small extension of lifespan and less inflammation as compared with Foxp3^creRaptor^fl/fl mice (Supplementary Fig. 16). Therefore, Raptor/mTORC1 signaling promotes T_reg function in part through inhibiting mTORC2, although the predominant effect of mTORC1 is to coordinate metabolic programs and T_reg suppressive molecules.

Much emphasis has been placed on the transcriptional mechanisms that orchestrate T_reg suppressive activity, but how immunological signals are sensed and integrated by T_regs for their functional activation remains obscure. Here we describe that mTORC1-dependent metabolic programing is a central mechanism to couple immune signals including TCR and IL-2 and T_reg suppressive function. Raptor/mTORC1 promotes the lipogenic program, with the mevalonate pathway particularly important for coordinating T_reg proliferation and optimal induction of effector molecules CTLA-4 and ICOS (Supplementary Fig. 17). Thus, mTORC1-dependent lipid metabolism provides a novel link between two crucial T_reg regulators, IL-2 and CTLA-4, the loss of which causes fulminant autoimmune diseases^21,30. An additional mechanism is via mTORC1-dependent inhibition of mTORC2, although the modest rescue effect of Rictor deletion in Foxp3^creRaptor^fl/fl mice suggests a relatively minor contribution of this feedback pathway. Notably, mTORC1-dependent metabolic programming operates both under steady state and immune stimulation, and this provides important mechanistic insight into the apparently conflicting observations between T_regs’ in vitro anergy and in vivo antigen-primed state^16–18. We propose that T_regs adopt the evolutionarily ancient mTORC1 signaling to link immunological inputs to metabolic activity and functional fitness, therefore implicating mTORC1 as a fundamental rheostat to program T_reg suppressive activity via a non-conventional mechanism.

METHODS

Mice

C57BL/6J, CD45.1⁺, Raptor^fl, Rag1^−/−, Ifng^−/−, and ROSA26^GFP (a loxP site-flanked STOP cassette followed by the GFP-encoding sequence was inserted into the ROSA26 locus) mice were purchased from the Jackson Laboratory. Rictor^fl/fl, Cd4-Cre and ROSA26-Cre-ER^T2 mice have been described previously³¹. Foxp3^YFP-Cre mice were a generous gift from Dr. Alexander Rudensky²². Foxp3^creRaptor^fl/fl mice were used at 3–4 weeks old unless otherwise noted, with the age and gender-matched wild-type mice containing the Foxp3^cre allele as controls. Other mice were used at 8–10 weeks old unless otherwise noted. Bone marrow chimeras were generated by transferring 1 × 10⁷ T cell-depleted bone marrow cells into sublethally irradiated (5 Gy) Rag1^−/− mice, followed by reconstitution for at least 2 months. For treatment with tamoxifen, mice were injected intraperitoneally with tamoxifen (1 mg per mouse) in corn oil daily for 4 consecutive days and then analyzed 7 days after the last injection. All mice were kept in specific pathogen free facility in Animal Resource Center at St. Jude Children’s Research Hospital. Animal protocols were approved by the Institutional Animal Care and Use Committee of St. Jude Children’s Research Hospital.

Flow cytometry

For analysis of surface markers, cells were stained in PBS containing 2% (wt/vol) fetal bovine serum (FBS) with antibodies from eBioscience, unless otherwise noted. Foxp3 staining was performed per manufacturer’s instruction (eBioscience). Intracellular staining of CTLA-4 was performed together with Foxp3 using anti-CTLA-4 antibody (UC10-4F10-11; BD Biosciences). BrdU and caspase-3 staining was performed per manufacturer’s instruction (BD Biosciences). For detection of phosphorylated signaling proteins, lymphocytes were rested in complete medium for 1 hour. They were fixed with Phosflow Lyse/Fix buffer, followed by permeabilization with Phosflow Perm buffer III (BD Biosciences) and staining with antibodies to S6 phosphorylated at Ser235 and Ser236 (D57.2.2E; Cell Signaling Technology), 4E-BP1 phosphorylated at Thr37 and Thr46 (236B4; Cell Signaling Technology), Akt phosphorylated at Ser473 (M89-61; BD Biosciences) and Thr308 (J1-223.371; BD Biosciences), and Erk phosphorylated at Thr202 and Tyr204 (20A; BD Biosciences). For staining mitochondria, lymphocytes were incubated for 30 minutes at 37°C with 10 nM MitoTracker Deep Red (Life Technologies) or 20 nM TMRM (tetramethylrhodamine, methyl ester; ImmunoChemistry Technologies) after staining surface markers. ROS were measured by incubation with 10 μM CM-H2DCFDA (5-(and-6)-chloromethyl-2’,7’-dichlorodihydrofluorescein diacetate, acetyl ester; Life Technologies) or 2.5 μM CellROX Deep Red (Life Technologies) after staining surface markers. Flow cytometry data were acquired on LSRII or LSR Fortessa (BD Biosciences) and analyzed using Flowjo software (Tree Star).

Cell purification and culture

Lymphocytes were isolated from lymphoid organs (spleen and peripheral lymph nodes that included inguinal, auxiliary and cervical lymph nodes) and naïve and T_reg cells were sorted on a MoFlow (Beckman-Coulter) or Reflection (i-Cyt). Sorted naïve (CD4⁺CD62L⁺CD44^loCD25⁻) and T_reg cells were used for in vitro culture in Click’s medium (plus β-mercaptoethanol) supplemented with 10% (vol/vol) FBS and 1% (vol/vol) penicillin-streptomycin. T_regs were activated with plate-coated anti-CD3 (5 μg/ml; 2C11; Bio X Cell), anti-CD28 (5 μg/ml; 37.51; Bio X Cell) and human IL-2 (200 U/ml), in the presence or absence of simvastatin (2 μM; EMD Millipore), or 25-Hydroxycholesterol (0.5 μg/ml; Sigma). Thymidine incorporation assays were performed by adding 0.2 × 10⁵ T cells and 1 × 10⁵ irradiated splenocytes together with soluble anti-CD3 (0.2 μg/ml) in the presence or absence of human IL-2 (100 U/ml) for 72 hours, followed by pulsing with [³H]-thymidine at 1 µCi per well for the last 12–16 hours of culture. For labeling with CFSE and Celltrace violet (both from Life Technologies), T cells were incubated in Click’s medium with 5% FBS and 4 μM of either dye at 37°C for 25 minutes, followed by extensive washes. In vitro T_reg suppression assays were performed as described⁷. Briefly, freshly sorted T_eff (CD4⁺CD45RB^hiCD25⁻; 5 × 10⁴) and T_reg cells (at different ratios with T_eff) were cultured in 96-well plates along with anti-CD3 antibody and irradiated splenocytes for 72 hours, followed by Thymidine incorporation assays. For T_reg suppression assays using previously activated T_regs, T_regs were sorted from C57BL/6J mice and activated with anti-CD3, anti-CD28 and IL-2 for 3 days in the absence or presence of simvastatin (2 μM), 25-hydroxycholesterol (0.5 μg/ml), 22(R)-hydroxycholesterol (5 μM; Sigma), atorvastatin (5 μM; EMD Millipore), lovastatin (5 μM; EMD Millipore), mevalonate (100 μM; Sigma), or certain combinations as specified; vehicle control did not have measurable effects (not shown). Viable T_regs were purified using Ficoll and incubated with freshly isolated T_eff cells for additional analysis as described above.

Colitis model

A total of 4 × 10⁵ T_eff cells (CD4⁺CD45RB^hiCD25⁻) from CD45.1⁺ mice were mixed with 2 × 10⁵ wild-type or Raptor-deficient T_regs (from Foxp3^creRaptor^fl/fl or Cd4^creRaptor^fl/fl mice), and were transferred intraperitoneally into Rag1^−/− mice. Mice were assessed for clinical signs of colitis weekly and were analyzed 8–10 weeks after transfer. Colons were fixed in 10% (vol/vol) neutral buffered formalin, and sectioned at 4 μm thickness. Colon pathology was assigned scores by an experienced pathologist (P. V.) as described⁷. Lymphocytes were isolated from spleen and mesenteric lymph nodes and analyzed by flow cytometry.

De novo lipogenesis, cholesterol measurement and bioenergetics assays

T_regs were activated with anti-CD3, anti-CD28 plus IL-2 for 20 hours, and [1-¹⁴C]-acetic acid (Perkin Elmer; 4 μCi/ml) was added for additional 4 hours of culture. For labeling with [U-¹⁴C]-glucose (American radiolabeled chemicals), T_regs activated for 20 hours were washed with PBS and incubated in glucose-free medium with dialyzed FBS and [U-¹⁴C]-glucose (4 μCi/ml) for additional 4 hours of culture. Following incubation, cells were collected, washed twice with PBS and lysed in 0.5% Triton X-100. Lipids were extracted by the addition of chloroform and methanol mixture (2:1 vol/vol) with vortexing, followed by the addition of water with vortexing. After centrifugation, the lipid-containing phase (at the bottom) was separated and ¹⁴C incorporation was measured using a Beckman LS6500 scintillation counter. Results are normalized to cell number. Cellular cholesterol level was measured using Amplex Red assay (Life Technologies). Briefly, cells were washed with PBS and lysed in the Amplex Red reaction buffer. After 15 minute of incubation, cell lysates were centrifuged at maximum speed for 5 minutes. Fifty μl of supernatants were pipetted into a 96-well tissue culture plate and a 50 ul aliquot of Amplex Red working solution was added to each well. The plate was incubated for 120 minutes at 37°C, protected from light. Fluorescence was subsequently measured on a fluorescence microplate reader (Molecular Devices). A cholesterol standard curve was determined for each plate using a cholesterol reference standard. The bioenergetic activities of the ECAR and OCR pathways were measured using the Seahorse XF24-3 extracellular flux analyzer per manufacturer’s instructions (Seahorse Bioscience).

Gene expression profiling and gene-set enrichment analysis

RNA samples from freshly isolated T_regs from wild-type and Cd4^creRaptor^fl/fl mice were analyzed using the Affymetrix HT_MG-430_PM GeneTitan peg array and expression signals were summarized using the RMA algorithm (Affymetrix Expression Console v1.1). Gene-set enrichment analysis within canonical pathways was performed as described³².

RNA and immunoblot analysis

Real-time PCR analysis was done as described with primers and probe sets from Applied Biosystems⁷, or using Power SYBR Green Master Mix from Life Technologies. Immunoblots were performed and quantified as described previously³¹, using the following antibodies: p-S6, p-Akt Ser473, p-4E-BP1, p-Foxo1/3, Raptor (all from Cell Signaling Technology) and β-actin (Sigma).

Statistical analysis

P values were calculated with Student’s t test, Mann Whitney test, or ANOVA (GraphPad Prism) as specified in figure legends, with proper post-test analysis performed. Statistical analysis of mouse survival and respective P values were determined using the Logrank test. *, P < 0.05; **, P <0.01; ***, P < 0.001; ns, not significant.