Foxp3 and Toll-like receptor signaling balance T_reg cell anabolic metabolism for suppression

Abstract

CD4⁺ effector T cells (T_eff cells) and regulatory T cells (T_reg cells) undergo metabolic reprogramming to support proliferation and immunological function. Although signaling via the lipid kinase PI(3)K (phosphatidylinositol-3-OH kinase), the serine-threonine kinase Akt and the metabolic checkpoint kinase complex mTORC1 induces both expression of the glucose transporter Glut1 and aerobic glycolysis for T_eff cell proliferation and inflammatory function, the mechanisms that regulate T_reg cell metabolism and function remain unclear. We found that Toll-like receptor (TLR) signals that promote T_reg cell proliferation increased PI(3)K-Akt-mTORC1 signaling, glycolysis and expression of Glut1. However, TLR-induced mTORC1 signaling also impaired T_reg cell suppressive capacity. Conversely, the transcription factor Foxp3 opposed PI(3)K-Akt-mTORC1 signaling to diminish glycolysis and anabolic metabolism while increasing oxidative and catabolic metabolism. Notably, Glut1 expression was sufficient to increase the number of T_reg cells, but it reduced their suppressive capacity and Foxp3 expression. Thus, inflammatory signals and Foxp3 balance mTORC1 signaling and glucose metabolism to control the proliferation and suppressive function of T_reg cells.

The activation of CD4⁺ T cells is accompanied by increased growth, proliferation and specification into functional effector T cell (T_eff cell) or regulatory T cell (T_reg cell) subsets that promote or suppress immunity and subsequent inflammatory resolution. To support this transition from quiescent naive T cells to T_eff cells and T_reg cells, activated CD4⁺ T cells reprogram their cellular metabolism¹. Resting T cells utilize a catabolic oxidative metabolism of glucose, lipids and amino acids. After being activated via co-stimulation, however, T cells upregulate the expression of glucose and amino acid transporters and increase their metabolic rate. Although both mitochondrial oxidative phosphorylation and glycolytic metabolism increase, glycolytic flux is elevated to a greater degree. In T_eff cells, this leads to a predominantly glycolytic phenotype that is reminiscent of the proliferative metabolism of cancer cells, aerobic glycolysis, that is thought to provide biosynthetic intermediates for anabolic cell growth². CD4⁺ T_eff cells are dependent on the glucose transporter Glut1 and aerobic glycolysis for proliferation and inflammatory functions, as inhibition of glycolysis or deletion of Glut1 impairs T_eff cell function in vivo^3,4. In contrast, induced T_reg cells have been shown to utilize mainly a distinct metabolic program based on mitochondrial oxidation of lipid and pyruvate^5–8, and T_reg cells can act independently of Glut1 in vivo⁴. This oxidative metabolism provides inefficient support for biosynthesis for cell growth. However, T_reg cells can be highly proliferative and glycolytic in vivo^9,10. Although T_reg cell are important in autoimmunity⁹ and inflammatory resolution and healing^11–13, the mechanisms that regulate T_reg cell metabolism and how these metabolic pathways affect T_reg cell proliferation and function remain poorly understood.

Signals through the transcription factors Myc and HIF1α, or the PI(3)K-Akt-mTORC1 pathway¹ and transcription factors of the CD4⁺ T cell subset^14,15, facilitate the induction of Glut1 expression and aerobic glycolysis in T_eff cells. The PI(3)K-Akt-mTOR pathway has also been suggested to be a regulator of T_reg cell metabolism. However, T_reg cells do not require mTOR kinase¹⁶. Furthermore, although PI(3)K-Akt-mTOR signaling can be activated in T_reg cells^17,18, selective loss of mTORC1 (ref. 18) or constitutive activation of mTORC1 as a result of deficiency in the lipid phosphatase PTEN or the tumor suppressor TSC2 leads to increased numbers of T_reg cells, but those T_reg cells have impaired capacity for immunosuppression^19–21. Although mTORC1 regulates protein translation and other aspects of cell physiology beyond anabolic metabolism, T_reg cell metabolism might be heterogeneous or oscillatory¹⁷, and these changes could affect T_reg cell function and proliferation.

Another signaling mechanism that modulates T_reg cell proliferation and function is signaling through TLRs. T_reg cells express several TLRs, including TLR1, TLR2, TLR4, TLR5 and TLR8 (refs. 22–25). In each case, ligation of the TLR leads to increased T_reg cell proliferation but reduced suppressive function^23–26. Pattern recognition via TLR signaling appears to be critical for T_reg cell homeostasis and function, as microbiota can promote the loss of Foxp3 expression on intestinal T_reg cells²⁷ and T_reg cells deficient in the TLR adaptor-transducer MyD88 are reduced in number in the gut, which leads to the onset of inflammatory bowel disease (IBD)²⁸. Similarly, microbiota that provide TLR ligands have been shown to be important for driving T_reg cell proliferation in the skin²⁹. TLR ligation has also been shown to activate PI(3)K-Akt-mTORC1 signaling through modulation of PI(3)Kγ in macrophages³⁰ and to induce glycolysis in B cells³¹ and dendritic cells³². However, the role of TLR signals as modulators of T_reg cell metabolism and how metabolic pathways balance the anabolic growth and proliferation of T_reg cells with suppression remain unknown.

We investigated mechanisms that regulate T_reg cell metabolism and how changes in glucose metabolism affect T_reg cell proliferation and function. Although most T_reg cell used a catabolic metabolism with low Glut1 expression, proliferating T_reg cell showed elevated Glut1 levels and mTOR activity. Ligation of TLR1 and TLR2 on activated T_reg cells further increased glycolysis and proliferation, but the cells’ suppressive capacity was reduced in an mTORC1-dependent manner. Conversely, Foxp3 opposed PI(3)K-Akt-mTOR signaling, reducing glycolytic and anabolic metabolism while increasing the activity of oxidative and catabolic pathways. These metabolic changes directly modified T_reg cell function, as transgenic expression of Glut1 in T cells³³ reduced T_reg cell suppressive capacity and led to the downregulation of Foxp3. Thus, TLR signals and Foxp3 counter-regulate T_reg cell metabolism to balance proliferation and suppressive function.

RESULTS

Proliferative T_reg cells activate mTORC1 and glucose uptake

To assess the regulation and role of T_reg cell metabolism in proliferation and function, we first investigated the metabolic programs of proliferative and non-proliferative T_reg cell³⁴. Foxp3⁺ thymus-derived T_reg cells (tT_reg cells) were more proliferative than Foxp3⁻ CD4⁺ T cells were under homeostatic conditions, as determined by staining for the proliferation marker Ki67 (Supplementary Fig. 1a). The tT_reg cells with high Ki67 expression (Ki67^hi), however, were phenotypically distinct from those with low Ki67 expression (Ki67^lo). Consistent with glycolysis being an anabolic metabolic program that supports proliferation, Ki67^hi tT_reg cells had increased Glut1 expression (Fig. 1a). mTORC1 activity was also elevated in Ki67^hi tT_reg cells, as measured by phosphorylated ribosomal protein S6 (Fig. 1b). Thus, tT_reg cells were not metabolically uniform, and those cells that had undergone proliferation showed increased expression of markers of glycolysis.

Figure 1.

T_reg cell proliferation is associated with increased glycolysis and anabolic signaling. (a,b) Flow cytometry analyzing the expression of Glut1 (a) and phosphorylated S6 (p-S6) (b) in Ki67_high and Ki67_low populations of CD4⁺Foxp3⁺ tT_reg cells from the spleens of wild-type mice (gating strategy, Supplementary Fig. 1a). MFI, mean fluorescent intensity. *P < 0.05 (two-tailed Student’s t test). Data in a and b are each representative of three independent experiments (left) or are pooled and normalized to Ki67^low from three independent experiments (right; mean + s.d.).

TLR1, TLR2 promote iT_reg glycolysis but impair suppression

We next assessed the effect of T_reg cell activation on glucose metabolism. Although proliferative tT_reg cells had higher surface expression of Glut1 protein than that of non-proliferative tT_reg cells, Glut1 expression was even higher in activated induced T_reg cells (iT_reg cells) (Fig. 2a). T_reg cells express TLR1 and TLR2 at high levels²² and can respond to TLR ligation with increased proliferation²⁴. Treatment of activated iT_reg cells with the TLR1 and TLR2 agonist Pam₃CSK₄ further increased the expression of Glut1 and hexokinase 2 (HK2), an enzyme that phosphorylates intracellular glucose (Fig. 2a,b). This was accompanied by a sharp increase in T_reg cell size (Supplementary Fig. 1b), glycolytic lactate production (Fig. 2c and Supplementary Fig. 1c), activation of the mTORC1 pathway, and proliferation, as determined by Ki67 expression (Fig. 2d). To determine how TLR1- and TLR2-induced changes in T_reg cell signaling and metabolism affect T_reg cell function, we treated iT_reg cells overnight with vehicle or Pam₃CSK₄ and then re-purified the cells. TLR-ligand-treated iT_reg cells showed decreased capacity to suppress effector CD8⁺ T cells and instead increased T_eff cell proliferation (Fig. 2e). Activation of mTORC1 signaling by Pam₃CSK₄ was critical for the subsequent decrease in T_reg cell suppressive capacity, as treatment with the mTORC1 inhibitor rapamycin together with ligation of TLR1 and TLR2 before re-purification restored the ability of iT_reg cells to suppress the proliferation of T_eff cells (Fig. 2e).

Figure 2.

Signaling via TLR1 and TLR2 drives T_reg cell glycolysis and proliferation, yet reduces suppressive capacity. (a) Glut1 expression (as in Fig. 1) of Ki67^high and Ki67^low tT_reg cells derived from the spleens of wild-type mice an of iT_reg cells derived from CD4⁺CD25⁻ T cells isolated from the spleens of wild-type mice and polarized for 5 d under T_reg cell–skewing conditions and treated with H₂O vehicle (Veh) or Pam₃CSK₄ (PAM; 5 μg/ml) for the final 24 h. (b) Immunoblot analysis of HK2, Glut1 and actin (loading control) in iT_reg cells as in a. (c,d) ECAR of vehicle or Pam₃CSK₄ treated iT_reg (c), and Ki67 expression and phosphorylated S6, determined by flow cytometry (d), in iT_reg cells as in a. (e) Inhibition of the proliferation of T_eff cells by T_reg cells treated for 24 h with DMSO (vehicle (−)) or 20 nM rapamycin (Rap) along with Pam₃CSK₄ (5 μg/ml) (left margin), then repurified by magnetic selection and functionally assessed an in vitro suppression assay with various ratios (above plots) of T_reg cells to T_eff cells. *P < 0.05 (Two-tailed Student’s t-test). Data are representative of three independent experiments (a,b,d,e) or two independent experiments with four technical replicates per group (c; mean + s.d.).

Foxp3 promotes oxidative metabolic gene expression

Genetic deletion of Foxp3 leads to a loss of T_reg cell suppressive function and a broad inflammatory autoimmune disorder³⁵. Foxp3 expression has also been reported to diminish Glut1 expression and Akt activity³⁶. We therefore examined published data sets to determine whether deletion of Foxp3 also altered T cell metabolism and markers of mTORC1 signaling. Following deletion of Foxp3, tT_reg cells were reported to induce widespread changes in gene expression that included not only genes encoding immunological products but a large number of genes encoding products that regulate or participate in varied metabolic pathways (Supplementary Fig. 1d). Alterations in the expression of genes that regulate primary metabolic process included protein products involved in PI(3)K-Akt-mTORC1 signaling as well as glucose metabolism, such as PTEN, phospho-fructokinase, enolase 2, transketolase, triosephosphate isomerase and others (Supplementary Table 1). Consistent with the inverse relationship of Foxp3 and PI(3)K signaling, it has also been reported that tT_reg cell activation in vivo correlates with a diminished Akt-related gene-expression signature³⁷.

To directly test how Foxp3 affects the expression of genes encoding products involved in metabolism, we stimulated CD4⁺ T cells in vitro under neutral conditions and transduced them to express Foxp3. Pathway analysis revealed that the group of genes with significantly altered expression in Foxp3-expressing cells showed enrichment for those encoding products involved in metabolic pathways (Fig. 3a). In particular, genes encoding products involved in lipid and peptide hormone metabolism pathways were upregulated, whereas those encoding products involved in glucose and nucleotide-metabolism pathways were downregulated. A comparison of genes altered by increased Foxp3 and those reported to change following genetic deletion of Foxp3 (ref. 35) revealed an overlap of numerous genes encoding products involved in metabolism (Supplementary Table 2). Both metabolic and immune regulatory pathways were broadly altered (Supplementary Table 3). Although the regulatory mechanisms were unclear, analysis of Foxp3-bound loci by chromatin immunoprecipitation and sequencing showed that Foxp3 associated with genes encoding the pyruvate dehydrogenase kinase PDK3 and the catalytic γ-subunit of PI(3)K (PI(3)Kcγ) (Supplementary Fig. 1e), which correlated with decreased expression (Supplementary Figs. 1f and 2a). Together these data indicated that Foxp3 regulated metabolic gene expression both directly and indirectly through modulation of the PI(3)K-Akt-mTORC1 pathway.

Figure 3.

Foxp3 reduces aerobic glycolysis and promotes mitochondrial oxidative metabolism. (a) Gene-expression microarray analysis and pathway analysis (with the DAVID bioinformatics database) of RNA from primary mouse CD4⁺CD25⁻ T cells stimulated and transduced with control or Foxp3-expressing retrovirus, presented as enrichment for selected pathways (left and right margins) among genes upregulated (top) or downregulated (bottom) by Foxp3. (b–g) Immunoblot analysis (b,c), glucose uptake (d), glycolysis (as ECAR after glucose injection) (e) basal OCR (f) and basal OCR/ECAR (g) of primary mouse CD4⁺CD25⁻ T cells stimulated and transduced with control (NGFR) or Foxp3-expressing retrovirus. Numbers above lanes (b,c) indicate band intensity relative to that of actin. CPM, counts per minute. *P < 0.05 (Two-tailed Student’s t-test). Data are representative of an experiment with biological triplicates for each group (a), three independent experiments (b,c) or two independent experiments with technical triplicates (d–g; mean + s.d.).

Foxp3 inhibits PI(3)K-Akt-mTORC1 signaling and glycolysis

We next assessed the effect of Foxp3 expression on metabolic proteins and pathways. Consistent with reduced PI(3)Kcγ expression and PI(3)K activity, Foxp3 expression led to diminished phosphorylation of Akt, S6 kinase and S6 relative to that of T cells transduced with a control vector (Fig. 3b). In addition, Foxp3 expression was sufficient to reduce the expression of Glut1 and HK2 and to increase the level of HK1 protein (Fig. 3c), similar to the expression pattern of T_reg cells⁶. Foxp3 expression in T cells led to lower glucose uptake and glycolysis, but the oxygen consumption rate (OCR) and the ratio of oxygen consumption to lactate production (OCR/extracellular media acidification rate (ECAR)) increased (Fig. 3d–g). We next transduced the mouse pro-B cell line FL5.12 to express a conditionally active Foxp3; activated Foxp3 led to reduced phosphorylation of S6 kinase, Akt and mTOR, and we observed lower amounts of PI(3)Kcγ and HIF1α protein in this setting (Supplementary Fig. 2a). Conversely, PTEN protein was modestly increased by activation of Foxp3 (Supplementary Fig. 2a). Expression of the mitochondrial proteins cytochrome C and CPT1a increased, whereas the expression of Glut1, HK2 and the majority of glycolytic enzymes decreased (Supplementary Fig. 2b,c and Supplementary Table 4). Activation of Foxp3 in FL5.12 cells was also sufficient to promote an oxidative shift, with reprogramming of the metabolome to increase catabolic metabolites and decrease many anabolic metabolites (Supplementary Fig. 2d and Supplementary Table 5). Altered metabolites included increased succinyl-carnitine, which is indicative of mitochondrial metabolism, and decreased glucoseamine-6-phosphate, glyceraldehyde-3-phosphate and N-carbamoyl-aspartate, which are indicative of glycosylation, glycolysis and pyrimidine synthesis, respectively (Supplementary Fig. 2e). Foxp3 activation led to decreased glucose uptake, glycolysis, lactate production and pentose-phosphate-pathway flux (Fig. 4a–d). Conversely, oxygen consumption, the OCR/ECAR ratio, pyruvate oxidation and palmitate oxidation all increased (Fig. 4e–h). Not all pathways were altered, however, as glutamine oxidation remained unchanged (Fig. 4i). Consistent with diminished anabolic metabolism, Foxp3 slowed the rate at which cells accumulated in vitro (Fig. 4j). Together these data indicated that TLR signals promoted T_reg cell proliferation, PI(3)K-Akt-mTORC1 activation and glycolysis, whereas Foxp3 promoted oxidative metabolism and slowed anabolic metabolism and growth.

Figure 4.

Foxp3 expression is sufficient to suppress glycolysis and promote oxidative metabolism. (a–i) Glucose uptake (a), glycolytic rate (assessed by a radiolabeled-glucose-tracer assay) (b), ECAR (c), flux of glucose into the pentose phosphate pathway (PPP) (d), OCR (e), OCR/ECAR (f), pyruvate oxidation (g), fatty-acid oxidation of palmitate (h) and glutamine oxidation (i) of vector-control-transduced FL5.12 pro-B cells or FL5.12 pro-B cells transduced with tamoxifen-inducible expression of Foxp3, treated with 4-hydroxytamoxifen. (j) Quantification of live cells as in a–i to assess cell accumulation over time. *P < 0.05 (two-tailed Student’s t test). Data are representative of three (a,b,j) or two (c–i) independent experiments with three independent clones per group (mean + s.d. in a–i or (mean ± s.d. in j).

Akt promotes T_reg cell growth but inhibits suppressive capacity

To directly test how the metabolic transitions noted above affect T_reg cell phenotype and function, we modified the Akt-mTORC1 signaling and metabolism of T_reg cells through the use of mice with transgenic T cell–specific expression of constitutively active Akt (Akt-tg mice) or Glut1 (Glut1-tg mice)⁷. tT_reg cells isolated from Akt-tg mice had increased glucose uptake and diminished reactive oxygen species (ROS) than that of tT_reg cells from wild-type mice (Supplementary Fig. 3a,b), and were more abundant in Akt-tg mice than in wild-type mice (Supplementary Fig. 3c,d). Akt-tg T_reg cell were also enlarged and had altered cell surface markers that were consistent with activation (Supplementary Fig. 3e–i). Similar to TLR1- and TLR2-stimulated T_reg cells and consistent with poor suppression by T_reg cells deficient in the tumor suppressors PTEN or TSC2 (refs. 19–21), Akt-tg T_reg cells were less able to suppress T_eff cell proliferation than were control wild-type T_reg cells (Supplementary Fig. 3j).

Glut1 enhances T_reg cell growth, but reduces function

Foxp3 and TLR signals might, in principle, modify T_reg cell metabolism and function through control of Glut1 expression and glycolysis. To directly assess the role of Glut1 in the regulation in T_reg cell function, we studied tT_reg cells from Glut1-tg mice ex vivo³³. Glut1-tg CD4⁺ T cells showed increased Glut1 expression and glucose uptake, whereas ROS were diminished (Supplementary Fig. 4a–c), relative to that of wild-type cells. Glut1-tg T_reg cells also had higher ECAR and OCR than that of wild-type control T_reg cells (Supplementary Fig. 4d). Glut1-tg tT_reg cells accumulated in terms of both numbers and frequency (Fig. 5a,b) and were modestly enlarged (Fig. 5c). However, Glut1-tg mice develop autoimmunity with age⁷, which suggests that their T_reg cell function is impaired. Consistent with altered function, tT_reg cells from Glut1-tg mice had a greater proportion of CD25^lo Foxp3⁺ cells than did those from wild-type mice (Supplementary Fig. 4e). We also observed a modestly increased frequency of Helios⁻ Foxp3⁺ CD4⁺ T cells in Glut1-tg mice (Supplementary Fig. 4f), which might indicate a greater frequency of peripherally induced T_reg cells or lower T_reg cell suppressive capacity.

Figure 5.

Glut1 expression increases the number and size of T_reg cells but reduces their suppressive capacity. (a–c) Number (a), frequency (b) and size (by forward scatter (FSC)) (c) of Foxp3⁺ tT_reg cells from the spleen of control and Glut1-tg mice, assessed by flow cytometry. (d) RNA-sequencing analysis of T_reg cells from age-matched control and Glut1-tg mice (two independent cohorts), presented as normalized data. (e,f) Flow cytometry analyzing the expression of CD25, Foxp3, ID3 and EZH2 (e) or IFN-γ (f) by CD4⁺CD25⁻ T cells isolated from the spleen of control and Glut1-tg mice and polarized under T_reg cell–skewing conditions; results presented as the geometric mean fluorescent intensity (MFI) in e, right. Numbers adjacent to outlined areas (f) indicate percent IFN-γ⁺ cells. (g) Inhibition of T_eff cell proliferation by control and Glut1-tg T_reg cells as in e,f (presented as in Fig. 2e). *P < 0.05 (two-tailed Student’s t test). Data are representative of five independent experiments (a,b; mean + s.d.), four independent experiments (c; mean + s.d.), an experiment with two independent cohorts (two control mice with three Glut1-tg mice, and three control mice with one Glut1-tg mouse) (d), an experiment with four biological replicates per genotype (e,f; mean + s.d. in e) or three independent experiments (g).

To better assess the effect of Glut1 on T_reg cells, we activated and induced wild-type control and Glut1-tg T cells to generate Glut1-tg iT_reg cells. Similar to tT_reg cells, Glut1-tg iT_reg cells were generated at an increased frequency, were enlarged and had reduced expression of the cytokine receptor CD25 (IL-2Rα), relative to that of wild-type control iT_reg cells (Supplementary Fig. 5a–c). Unlike tT_reg cells, Glut1-tg iT_reg cells also showed reduced expression of the co-stimulatory receptor ICOS, relative to that of control iT_reg cells (Supplementary Fig. 5d). In two cohorts of control and Glut1-tg iT_reg cells, we found, by RNA sequencing, that Glut1 expression altered the expression of genes encoding products involved in both metabolism and immunological function (Fig. 5d). These changes included upregulation of Ifng and Tbx21 (which encodes the transcription factor T-bet), and downregulation of Foxp3, Il2ra (CD25), Foxo1 and Bcl6. Gene-ontology analysis of significantly altered genes revealed enrichment for the expression of genes encoding products involved in T cell activation, cytokine signaling and regulation of epigenetic mechanism pathways (Supplementary Table 6). Similarly, quantitative-PCR array analysis of wild-type control and Glut1-tg iT_reg cells also showed numerous differences in gene expression, including downregulation of Foxp3, Icos, Pdcd1 (PD-1), Gata3 and Tnfrs8 (CD30) (Supplementary Fig. 5e and Supplementary Table 6). We assessed expression of CD25, Foxp3, the transcription factor ID3 and the histone methyltransferase EZH2 by flow cytometry and found that expression of all of them was reduced in Glut1-tg iT_reg cells (Fig. 5e and Supplementary Fig. 4e). A portion of Glut1-tg T_reg cells also showed altered cytokine expression, with decreased Il-10 (Supplementary Fig. 5e) and an increase in the population of IFN-γ-producing cells and a decrease in the population of IL-2-producing cells (Fig. 5f and Supplementary Fig. 5f). Consistent with reduced expression of functional T_reg cell markers, induced Glut1-tg T_reg cells had reduced in vitro suppressive capacity relative to that of control iT_reg cells (Fig. 5g and Supplementary Fig. 6a).

Lower stability of Glut1-expressing T_reg cells in vivo in IBD

The phenotypic changes noted above suggested that increased glucose metabolism altered or reduced the suppressive function or stability of T_reg cells in vivo. To test whether the Glut1-tg T_reg cell were functionally impaired in vivo, we depleted naive T cell populations of T_reg cells and adoptively transferred the resultant cells into lymphopenic mice to induce IBD. After disease onset, we gave some of the host mice injection of CD4⁺CD25⁺CD45Rb^lo tT_reg cells sorted from wild-type control or Glut1-tg mice (Supplementary Fig. 6b). Despite potential alterations from transgenic Glut1 expression, we found no substantial difference between the CD4⁺CD25⁺CD45Rb^lo tT_reg cell populations in their frequency of Foxp3-expressing cells (Supplementary Fig. 6c). As expected, control tT_reg cells reversed disease progression, and the mice began to gain weight (Fig. 6a). Glut1-tg tT_reg cells, however, slowed disease progression only partially (Fig. 6a). At the termination of the experiment, we found that Glut1-tg tT_reg cells had restrained T_eff cell accumulation less effectively than control tT_reg cells had (Fig. 6b). Despite the transfer of similar initial numbers of tT_reg cells, CD4⁺ Foxp3⁺ T cells from Glut1-tg donors were also significantly less frequent than were control T_reg cells (Fig. 6c). Foxp3 expression was significantly reduced (P = 0.003) in transferred Glut1-tg T_reg cells in both mesenteric lymph nodes (Fig. 6d) and spleen (Supplementary Fig. 6d). Reduced Foxp3 expression was confirmed by transfer of congenically marked control and Glut1-tg T_reg cells; this revealed that increased T_reg cell glycolysis in IBD led to T_reg cell downregulation of Foxp3 expression (Fig. 6e and Supplementary Fig. 6e). Together these data showed that increased glucose uptake altered T_reg cell phenotype and Foxp3 expression in vivo. Foxp3 and TLR signals therefore act to balance T_reg cell metabolism and regulate the proliferation, function and potential stability of T_reg cells (Supplementary Fig. 6f).

Figure 6.

Increased glucose uptake impairs the suppressive ability of T_reg cells. (a) Body weight of T cell– and B cell–deficient Rag1^−/− host mice given injection of sorted naive T_eff cells (CD4⁺CD25⁻CD45RB^hi) for the induction of colitis, then (after weight loss indicated active disease), given injection of no additional cells (T_eff alone) or control or Glut1-Tg CD4⁺CD25⁺CD45RB^lo T_reg cells (downward arrow), assessed by weighing three times per week and presented as change from initial weight. (b–e) Frequency of CD4⁺Foxp3⁺ T cells (b), number of CD4⁺ T cells (c) in each T_reg cell recipient group, and expression of Foxp3 and the MFI of Foxp3 expression of CD4⁺Foxp3⁺ T cells in the mesenteric lymph nodes (d,e) of mice as in a (with naive effector cells expressing Thy1.1 and T_reg cells expressing Thy1.2 in e, to allow distinction of cells by these congenic markers). *P < 0.05 (b–d; two-tailed Student’s t test, a; two-way ANOVA followed by Tukey’s post hoc test). Data are representative of at least three independent experiments with over four mice per group in each (a; mean ± s.e.m.), two independent experiments with n = 6 mice (T_eff cells alone), n = 9 mice (control T_reg cells) or n = 8 mice (Glut1-Tg T_reg cells) (b,c; mean + s.d.), two experiments with at least four mice per group (c; mean + s.d.), or one experiment with n = 4 mice (T_eff cells alone), n = 4 mice (control T_reg cells) or n = 5 mice (Glut1-Tg T_reg cells) (e).

DISCUSSION

T_reg cells have been shown to rely mainly on the mitochondrial oxidative metabolic pathways of lipid and pyruvate^5–8. However, this mode of metabolism supports anabolic growth pathways only inefficiently, and T_reg cell can be highly proliferative in vivo⁹. T_reg cell have also been found to be capable of inducing glycolysis^17,18, but the mechanisms by which this occurs remain unclear. We found that T_reg cell metabolism was dynamically regulated by TLR signals and Foxp3 to balance the anabolic metabolism and proliferation of T_reg cells with suppressive capacity. T_reg cells responded to activation and TLR ligation by increasing Glut1 expression and glycolysis. Conversely, Foxp3 decreased Glut1 expression and glycolytic and anabolic metabolism to instead favor mitochondrial oxidative pathways. Similar inhibition of glycolysis has also been described for Foxo1 (ref. 38), which suggests that these two T_reg cell transcription factors might induce a barrier to aerobic glycolysis that favors suppression. In contrast, inflammatory signals via TLR1 and TLR2 ligation stimulated glucose uptake and downregulated Foxp3 expression. This model suggests that metabolic changes in acute infection or inflammation provide signals that promote T_reg cell glycolysis and proliferation while limiting suppressive capacity to allow T_eff cells to eliminate the invading pathogen. As inflammatory signals decrease, Foxp3 levels might increase and tilt the metabolic balance away from glycolysis to favor mitochondrial oxidative pathways instead. Although this would decrease proliferation, this shift might increase suppressive capacity and promote wound healing and inflammatory resolution.

The metabolic heterogeneity of T_reg cell that we observed might reflect the breadth of T_reg cell phenotypes observed in vivo. T_reg cells derived from the thymus or periphery, resting or activated by antigen or distinct stimuli, or circulating or tissue resident have each been shown to have distinct immunological and gene-expression phenotypes^9,39. Increased glucose metabolism resulting from Glut1 expression leads to an accumulation of T_reg cells that is reminiscent of effector T_reg cells⁴⁰, but with several key differences. In particular, T_reg cells with elevated Glut1 downregulated Foxp3 expression and suppressive capacity and seemed to have potentially reduced stability. Thus, although glucose metabolism contributes to T_reg cell phenotypes and heterogeneity, T_reg cell function and stability also depend on other factors and stimuli.

The direct regulation of metabolism by the central T_reg cell transcription factor Foxp3 ensures that metabolic pathways are matched to functional demands. Similar regulation of T cell metabolism by key subset transcription factors has been described for the transcription factor RORγt in the T_H17 subset of helper T cells¹⁴ and the transcription factor Bcl6 in follicular helper T cells¹⁵. Foxp3 can suppress T_reg cell glycolysis through several mechanisms, including both direct metabolic targets and inhibition of PI(3)K-Akt-mTORC1 signaling. Indeed, Foxp3 has been shown to lead to reduced activation of Akt³⁶, and Foxp3 activation leads to a gene-expression signature of diminished Akt function³⁷. In addition to potential regulation of the Akt phosphatase PHLPP³⁶, Foxp3 might repress this pathway by suppressing PI(3)K signaling and PI(3)Kcγ expression. We also found that Foxp3 associated with the locus encoding Pdk3, which led to reduced expression of Pdk3; this suggested that Foxp3 might repress expression of this kinase to allow increased mitochondrial oxidation of pyruvate in direct suppression of glycolysis. Consistent with a potential role for Pdk3 in regulating T_reg function, inhibition of Pdk1 has been shown to lead to increased generation and suppressive activity of T_reg cells in vivo, which improves protection from asthma and experimental autoimmune encephalomyelitis^6,41.

T_reg cell are stimulated by antigens, cytokines and inflammatory signals. We found that T_reg cells increased their PI(3)K-Akt-mTORC1 signaling when proliferating under homeostatic conditions and after activation. Ligation of TLR1 and TLR2 on activated T_reg cells led to the greatest upregulation of Glut1 expression and glycolysis. Those findings are consistent with the proposal of a role for PTEN in suppressing PI(3)K-Akt-mTORC1 signaling in T_reg cells following stimulation with antigen and cytokines^19,20. The mechanism by which ligation of TLR1 and TLR2 led to activation of the PI(3)K-Akt-mTORC1 pathway in T_reg cells is not clear but might involve activation through PI(3)Kcγ³⁰, which we identified as a potential target of Foxp3 repression. In addition to activating PI(3)K-Akt-mTORC1, TLR signaling also substantially activates the transcription factor NF-κB⁴², which might also modulate metabolism and contribute to the regulation of T_reg cell proliferation and function. Consistent with a role for alternative pathways, TLR activation in dendritic cells leads to Akt activation and glycolysis, in part through activation of the kinase TBK1–kinase IKKε pathway in a mechanism independent of PI(3)K^32,43.

Increased glucose uptake on T_reg cells involved in an IBD response resulted in a sharp reduction in Foxp3 expression. Although these cells did not widely lose Foxp3 expression and convert to T_eff cells, as formerly T_reg cells, an increased portion of Glut1-tg T_reg cells were found to secrete IFN-γ, which suggested that T_reg cell stability was decreased. Similarly, Glut1-tg T_reg cells had downregulated expression of CD25, which is also a marker of reduced T_reg cell stability¹⁹. Prior to IBD, however, resting Glut1-tg T_reg cells showed only modestly lower Foxp3 expression, which indicated that activation and inflammatory signals were needed to more substantially downregulate Foxp3. Our finding that Glut1-tg T_reg cells had normal or modestly enhanced mitochondrial function but much greater glycolysis suggested that increased glycolysis itself might promote lower Foxp3 expression. The elevated glycolysis that occurs in PTEN- or TSC1-deficient T_reg cells^19–21 might therefore directly lead to the downregulation of Foxp3 mRNA and protein.

Together our data have described a dynamic regulation of T_reg cell metabolism with a glycolytic switch influencing anabolic metabolism for growth and proliferation or oxidative metabolism for maximal suppressive capacity. Metabolic separation of T_reg cell proliferation and T_reg cell suppression might allow kinetic resolution of T cell function. T_reg cells could accumulate in damaged or inflamed tissues without substantially impairing the protective immune response and are involved in inflammatory resolution and healing^11–13. Although multiple signals also strongly influence T_reg cell proliferation and function, metabolic regulation is a key contributor. The role of TLR signals in increasing T_reg cell glycolysis and proliferation while reducing suppression is further demonstrated in the interactions of T_reg cells with microbiota^23–28,39. Microbe-derived metabolites can drive T_reg cells to maintain homeostasis and tolerance, but physical contact and TLR ligation that indicate invading pathogenic bacteria instead allow proliferation and inflammatory function. Similarly, T_reg cells often accumulate in but fail to suppress inflammation in chronic autoimmune diseases, such as systemic lupus erythematosus and rheumatoid arthritis^44,45. Conversely, these pathways might promote the suppressive function of T_reg cells in tumors, in which glycolysis might be limited by reduced glucose or elevated lactate concentrations. Understanding the metabolic signaling and roles in T_reg cells might define new regulatory mechanisms and potential targets to modulate immunity.

ONLINE METHODS

Mice

6–8-week-old C57BL/6J mice from Jackson Laboratory were used for all experiments unless otherwise indicated, and in these cases animals were littermates and control and test animals were co-housed. Glut1 and MyrAkt transgenic mice were previously described^7,33,46. All procedures were performed under protocols approved by the Duke University Medical Center Institutional Animal Care and Use Committee.

Cell culture

Primary murine CD4⁺ and CD8⁺ T cells were isolated by magnetic bead negative (Miltenyi #130-104-454 or #130-104-075) and cultured in RPMI 1640 media supplemented with 10% FBS, 2 mM glutamine, 10 mM HEPES and 55 μM β-mercaptoethanol. Foxp3ER-NGFR and vector control-NGFR expressing FL5.12 cells were cultured in RPMI 1640 media lacking phenol red supplemented with 10% FBS, 2 mM glutamine, 10 mM HEPES and 55 μM β-mercaptoethanol.

Drug treatments

Foxp3ER and vector control-NGFR expressing cells were treated with 100 nM 4-hydroxytamoxifen (Sigma H6278) dissolved in 100% ethanol. Pam₃CSK₄ (Invivogen #tlrl-pms) was dissolved in DI H₂O and was used at a final concentration of 5 μg/ml. Rapamycin was dissolved in DMSO and was used at a final concentration of 20 nM.

iT_reg differentiation

CD4⁺CD25⁻ T cells were cultured on irradiated splenic feeder cells (30 Gy) with 2.5 μg/ml of anti-CD3 antibody at a ratio of 5:1 in RPMI supplemented with 10% FBS, sodium pyruvate, penicillin/streptomycin, HEPES and beta-mercaptoethanol. 3 ng/ml TGFβ was added to induce Foxp3⁺ iT_reg. On day 3 post-stimulation, cells were split 1:2 and re-plated with IL-2 alone for an additional 2 d. In some cases, cells were treated with 5 μg/ml Pam₃CSK₄ for 24 h beginning on day 4 post-stimulation.

T_reg suppression assay

T_reg cells were differentiated as described above and cultured at ratios of 1:1, 1:2 and 1:4 with CellTrace Violet (CTV) labeled CD8⁺ T cells along with 1 μg/mL of anti-CD3 antibody and irradiated feeder splenocytes (30Gy). T_reg suppression of effector T cell proliferation was determined 72 h post stimulation by CTV dilution of target population using flow cytometry.

Foxp3-ER cell lines

FL5.12 cells were infected with control or Foxp3-ER expressing lentivirus and selected with NGFR. The Foxp3-ER-NGFR construct was generously provided by M. Levings (University of British Columbia) and has been described previously⁴⁷. Three clones of each were selected based on equivalent NGFR expression and were Mycoplasma negative. To activate Foxp3-ER, 100 nM 4-hydroxytamoxifen (4-OHT) was added 36 h before experimental procedures.

Retroviral packaging

Retroviral constructs were packaged in Plat-E cells. Plat-E cells were maintained in DMEM supplemented with 10% FBS, with 1 μg/mL puromycin and 10 μg/mL blasticidin. To produce virus, Plat-E cells at 70% confluency in a six-well dish were transfected with construct plasmid along with Lipofectamine 2000 Transfection Reagent (Life Technologies). Media supernatant was removed 24 h post transfection, and fresh virus was collected at 72 and 96 h post transfection.

Retroviral Foxp3 expression

Primary CD4⁺CD25⁻ T cells were isolated from mice and stimulated with 3 ng/ml PMA and 1 μM ionomycin overnight. Cells were then treated with MSCV-Foxp3-NGFR or MSCV-NGFR control retroviral supernatant containing 8 μg/ml polybrene and centrifuged for 90 min at 2,500 rpm. Cells were rested for 5 h and then plated with fresh media and 10 U/ml interleukin-2 for 72 h. Foxp3-NGFR or NGFR control cells were positively selected with PE magnetic beads following NGFR-PE labeling (≥90% purity, Miltenyi Biotec).

Microarray gene expression analysis

Total RNA from activated transduced cells was isolated using TRIzol reagent (Invitrogen), and dissolved in RNase-free water. Total RNA treated with RNase-free DNase and cleaned up using Qiagen RNeasy columns. Biotinylated antisense cRNA was prepared by two rounds of amplification using the BioArray RNA Amplification and Labeling system according to the protocol for 10–1,000 ng of input RNA. cDNA libraries were amplified and hybridized to Affymetrix GeneChip Mouse Gene 1.0 ST array at the Nucleic Acid and Protein core facility of the Children’s Hospital of Philadelphia Research Institute. Array data were normalized and differential expression was determined using Utils, Stats, Samr, RankProd, topGO and Bioconductor. Significance was assessed using significance analysis of microarrays (SAM). Data are deposited in ArrayExpress (accession E-MTAB-4561). Genes differing in intensity by at least 1.5-fold (up or down) with P < 0.001 were considered Foxp3-regulated. Functional annotation of the top 200 FOXP3-regulated genes was achieved using DAVID software⁴⁸, and network analyses of the same gene sets were generated using Ingenuity software. Williams and Rudensky³⁵ gene expression array data from wild-type and Foxp3-deficient Treg were analyzed using Geo Data set Browser Data Analysis Tools to select genes with 1.5 fold expression change (P < 0.1) for gene ontology with PANTHER⁴⁹. Genes overlapping between Foxp3 expression or deficient T_reg were analyzed for gene ontology and pathway enrichment with DAVID.

ChIP-seq

Foxp3-chromatin complexes were immunoprecipitated from Foxp3- or empty-vector-transduced CD4⁺ T cells as described previously⁵⁰. Genomic DNA libraries were constructed from these samples using NEBNext Ultra DNA Library Prep kits (NEB) and sequenced using the HiSeq platform (Illumina). Sequence reads were mapped to mm9 using Bowtie, and HOMER software was used to identify read densities (peaks) enriched significantly (>5-fold) in ChIP samples over input controls. The top 30% of peaks were highly enriched for the Foxp3 consensus binding sequence (>85%), and were used as the set of high confidence, genome-wide Foxp3 binding sites.

RNA-seq of iT_reg

Glut1-tg (n = 4) and wild-type (control; n = 5) mice were used as naive T cell donors to generate iT_reg in two age-matched groups. Group 1 included control 1, 3 and Glut1-tg 2, 3, and 5 and Group 2 included control 2, 4, 5 and Glut1-tg 1. RNA was isolated from Glut1-transgenic and control Treg cells using the RNeasy Plus Mini kit (Qiagen) following the manufacturers’ instructions. Libraries were prepared with Illumina TruSeq Stranded mRNA sample kits using indexed adaptors (Illumina). Pooled libraries were subjected to 75-bp paired-end sequencing according to the manufacturer’s protocol (Illumina HiSeq3000). Targeted sequencing depth was 30 million paired-end reads per sample. Bcl2fastq2 Conversion Software (Illumina) was used to generate de-multiplexed Fastq files. The RNA reads obtained from sequencing were mapped with the TopHat 2 aligner using the default parameters to the UCSC mm10 mouse reference genome. Reads were examined for quality control and quantified using FeatureCounts. Subsequently, raw counts were normalized taking into account library sizes, filtering out genes with near zero expression, and the variance-stabilizing transformation was applied to the remaining genes (n = 13700) using the DESeq2 software package. Data are deposited in Geo Database (accession GSE84919). Statistical analyses were performed in the R Statistical Computing Environment. Differential gene expression was assessed between Glut1-tg and wild-type iT_reg samples, while at the same time adjusting for animal age groups. Benjamini-Hochberg false discovery rate (FDR) correction was subsequently used to account for multiple testing, and genes were considered to be differentially expressed (DE) if the corrected P < 0.05. To visualize gene expression differences between experimental groups, a heatmap was generated using the top 1000 most variably expressed genes. The mean expression values Glut1-tg and wildtype iT_reg blocked by age groups were used as heatmap input. Functional analysis was performed using Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics resources⁴⁸ and PANTHER⁴⁹. For this purpose, all differentially expressed genes (corrected p-values < 0.05) between Glut1-tg and wild-type iT_reg cell groups were split into two groups according to whether genes were downregulated or upregulated in the Glut1-tg experimental groups relative to the wild-type groups.

Flow cytometry analyses

Flow cytometry was performed on a Miltenyi Macsquant Analyzer. Antibodies used included anti-CD4 eFluor450 (eBioscience 48-0041-82), anti-CD25 PE (BD Biosciences 553866), anti-CD62L PE (eBioscience 12-0621-82), anti-CD69 PE (BD Biosciences 553237), anti-EZH2 PE (BD 562478), anti-FoxP3-APC (eBioscience 17-5773-80), anti-FoxP3-PE (eBioscience 12-5773-80), anti-Helios PE (eBioscience 12-9883-41), anti-ICOS PE (eBioscience 12-9942-81), anti-ID3 PE (BD Biosciences 564564), anti-mouse IFN gamma APC (eBioscience 17-7311-81), anti-mouse IL2 PE (BD Bioscience 561061), anti-KI67 eFluor450 (eBioscience 48-5698-80). Flow cytometry antibodies were used at a 1:100 dilution except for anti-EZH2 PE, anti-FoxP3-APC, anti-FoxP3-PE, anti-Helios PE, anti-ICOS PE, anti-ID3 PE, anti-mouse IFN gamma APC, anti-mouse IL2 PE, anti-KI67 eFluor450, which were all used at a 1:50 dilution. Cell trace violet (Invitrogen #C34557) was used following manufacturer’s specifications. Flow cytometry data analysis was performed using FlowJo.

PCR arrays

RNA was extracted as above and 1 μg total RNA was subjected to single-strand cDNA synthesis using the RT² first strand kit (Qiagen). The cDNA was used in the Glucose Metabolism and T cell Anergy and Immune Tolerance SuperArray RT² Profiler PCR arrays according to the manufacturer’s instructions and assayed on a ViiA 7 (Applied Biosystems). Data were analyzed using the RT² Profiler program supplied by Qiagen and normalized to the geometric mean of the housekeeping genes beta-actin, beta-2 microglobulin, beta-glucuronidase and heat shock protein 90 kDa alpha.

Immunoblotting

Immunoblotting was performed as described previously⁶. Primary antibodies were followed by mouse- or rabbit-conjugated horseradish peroxidase (HRP). HRP-conjugated antibodies (anti-mouse or anti-rabbit IgG HRP conjugate, Promega) were detected by enhanced chemiluminescence detection (Thermofisher). This included the following antibodies: Akt (9272, Cell Signaling), Cpt1a (15184-1-AP, Proteintech group), Hif-1α (10006421, Cayman Chemical Company), phosphor-Akt (4060, Cell Signaling), PIK3cγ (ab140310, abcam), phosphor-mTOR (5536, Cell Signaling), PTEN (9552, Cell Signaling), phosphor-p70 S6K (9204, Cell Signaling). Alternatively, primary antibodies were followed by fluorescently labeled anti-mouse or rabbit antibodies (LiCor) and imaged using the Odyssey infrared imaging system (LiCor). This included the following antibodies: Glut1 (ab652, Abcam), hexokinase 2 (2867, Cell Signaling), hexokinase 1 (ab104835, Abcam), cytochrome C (556432, BD Biosciences), β-actin (A5441, Sigma) and phosphor-S6 (4858, Cell Signaling). All primary antibodies were used at a 1:1,000 dilution, with secondary HRP antibodies used at 1:8000. Secondary LiCor antibodies were used at 1:10000 (anti-mouse) or 1:8000 (anti-rabbit).

Metabolic assays

Glycolysis and glucose uptake assays using ³H-glucose or ³H-2-deoxyglucose have been described previously⁶. Glutamine oxidation, pentose phosphate pathway (PPP) flux and glucose oxidation were determined by the rate of ¹⁴CO₂ release from U-¹⁴C-glutamine, 1-¹⁴C-glucose and 6-¹⁴C-glucose as described⁶. All values were normalized to cell number. OCR and ECAR were measured with an XF24 extracellular flux analyzer (Seahorse Bioscience) as described³¹. Suspension cells were attached to culture plates using Cell-Tak (BD Bioscience). OCR and ECAR were measured in unbuffered DMEM (Sigma-Aldrich) supplemented with 10 mM D-glucose (Sigma-Aldrich) and 10 mM L-glutamine, as indicated. OCR and ECAR values were normalized to cell number. For certain experiments, ECAR was measured over time following injection of 10mM D-glucose, oligomycin and 2-deoxy glucose (2DG). Glycolytic capacity is defined as the difference between the ECAR following the injection of oligomycin and ECAR following glucose injection. Nontargeted metabolomic analyses were performed as described using LC Q Exactive Mass Spectrometer (LC-QE-MS) (Thermo Scientific)⁶.

T cell transfer model of colitis

Splenic CD4 T cells were isolated as described above, and naive effector (CD4⁺CD25⁻CD45RB^hi) T cells were sorted (FACSVantage; BD Bioscience). Naive effector T cells were injected intraperitoneally into 6–8-week-old C57BL/6 RAG1^−/− recipients (10⁶ cells per mouse). Recipient animals were purchased from Jackson Labs and were assigned to groups randomly and co-housed in blinded fashion. Thy1.1 naive effector T cells were used in select experiments. Because the mice were H. pylori negative, and colitis does not occur spontaneously in this setting, disease was initiated 2 weeks following T cell injection with 200 p.p.m. piroxicam (Sigma-Aldrich) in powdered rodent chow for 3 d to enhance mucosal exposure to enteric bacteria and induce colitis⁶. 21 d following piroxicam treatment, Glut1 transgenic or littermate control T_reg cells (sorted CD4⁺CD25⁺CD45RB^lo, 2 × 10⁵ cells per mouse) were injected as indicated. Mouse weights were monitored tri-weekly. At the end of the experiment, total CD4⁺ and CD4⁺Foxp3⁺ T_reg in the spleen and mesenteric lymph nodes were analyzed by flow cytometry.

Statistical analysis

Sample sizes were determined by prior experience and estimated power calculations. Statistical analyses were performed with Prism software (GraphPad). Data were analyzed using a two-tailed Student’s t-test, and P < 0.05 was considered significant. Longitudinal data was analyzed by two-way ANOVA followed by Tukey’s post hoc test.