TGF-β and PI3K Signals Modulate Distinct Metabolism of Treg Subsets

Abstract

Murine Foxp3⁺ regulatory T cells differentiated in vitro (iTregs) in the presence of anti-inflammatory cytokine TGF-β rely predominantly upon lipid oxidation to fuel mitochondrial oxidative phosphorylation. Foxp3 expression underlies this metabolic preference as it suppresses glycolysis and drives oxidative phosphorylation. Here we show that in contrast to iTregs, thymic-derived Tregs (tTregs), engage in glycolysis and glutaminolysis at levels comparable to effector T cells despite maintained Foxp3 expression. Interestingly, exposure of tTregs to the anti-inflammatory cytokine TGF-β represses PI3K mediated mTOR signaling, inhibits glucose transporter and Hk2 expression, and reprograms their metabolism to favor oxidative phosphorylation. Conversely, replicating the effects of inflammation via elevation of PI3K signaling has minimal effects on tTregs, but dramatically enhances the glycolysis of normally oxidative iTregs resulting in reduction of Foxp3 expression. Collectively, these findings suggest both extrinsic and intrinsic factors govern the unique metabolic signature of Treg subsets.

Introduction

Metabolic disorders such as obesity and type-2 diabetes are associated with immune dysregulation, and immune-pathological conditions themselves have been linked to metabolic maladaptation of immune cells (1). Foxp3+ regulatory T cells (Tregs) are a key component in this bidirectional link between immunity and metabolism (2). Although once considered to be a single subset, Tregs can be further subdivided into two categories. The first, thymic Tregs (tTregs) arise during T cell ontogeny in the thymus with TCRs that have a relatively high avidity for peptide+self-MHC, albeit not so high as to trigger negative selection (3). The second category, peripheral Tregs (pTregs), develop from “conventional” CD4 T cells when they undergo activation under non-inflammatory conditions (e.g., low doses of antigen, presence of TGF-β, absence of IL-6 etc.) (4–6). Such cells, when generated in vitro, are referred to as induced Tregs (iTregs). Current models suggest that tTregs restrain immune responses against self-antigens, while pTregs play a predominant role in maintaining immune tolerance/dampening inflammation at mucosal surfaces and at the maternal-fetal interface (7–9).

Immunometabolic studies have demonstrated that the metabolic preference of immune cells is not only a result of cell differentiation but can be causal as well. For instance, stimulating CD4 T cells in a manner that fosters glycolytic metabolism (i.e., strong phosphatidylinositol-3 kinase (PI3K), myc and mTOR signals) promotes differentiation into effector T lineage (Th1, Th17 etc), while prevention of glycolysis (via induction of AMPK rather than mTOR signals) and/or forced utilization of oxidative metabolism leads to differentiation into iTregs (10, 11). This may be a self-reinforcing feature, as Foxp3 itself can drive lipid oxidation via suppression of myc (a master regulator of glycolysis), and upregulation of components of the electron transport chain required for mitochondrial oxidative phosphorylation (12–14). In contrast, available data on tTregs suggest that they can engage in glycolytic metabolism to a certain degree, which is critical for their proliferation, upregulation of suppressive molecules such as CTLA4 and ICOS and their migration (15–17). This indicates that although Foxp3 may be a key driver of oxidative metabolism in Tregs induced from naïve T cells, the same is not necessarily true in pre-existing Tregs.

Here we perform a detailed metabolic comparison of tTregs and iTregs and show that in contrast to iTregs, tTregs can engage aerobic glycolysis comparable to differentiated effector T cells. Strikingly, exposing tTregs to iTreg-inducing conditions (i.e., TGF-β) reduces mTOR dependent glycolytic metabolism. Interestingly, elevation of PI3K/mTOR activity via ablation of the lipid phosphatase PTEN reconfigures the metabolism of tTregs and iTregs distinctly. While PTEN deficiency in tTregs promotes glycolytic-lipogenesis and nucleotide biosynthesis without affecting Foxp3 expression, its loss in iTregs enhances aerobic glycolysis and is associated with a significant decrease in Foxp3 expression. Together, these findings demonstrate inherent metabolic differences between Treg subsets and show that adaptations occurring due to context dependent cues modulate Treg metabolism.

Materials and Methods

Mice

Foxp3-YFP-Cre (WT) and Pten^fl/fl Foxp3-YFP-Cre (PTEN ΔTreg) mice (18) were maintained in specific pathogen-free conditions at Massachusetts General Hospital (MGH), Boston. Healthy mice aged between 6–14 weeks were used for all experiments. All animal experiments were performed in compliance with the institutional guidelines approved by the MGH Animal Care and Use Committee (IACUC).

Cell sorting and flow cytometry

CD4⁺ T cells from spleens and lymph nodes (inguinal, axiliary and cervical) were enriched using the CD4⁺ T cell negative selection kit (ebioscience) and sorted to over 95% purity on a FACSAria II (BD) for YFP+ CD4+ Tregs and YFP- CD4+ CD62L^hi CD44^lo CD25 –ve naïve T cells. PhosphoFlow, intracellular staining for PTEN and Glut1 was performed using the BD Phosflow™ protocol III with antibodies for PTEN (BD, Clone A2B1), phospho Akt 473 (Cell signaling), phospho S6 (Cell Signaling) and Glut1 (abcam). The affinity pure (F-(ab’)2) fragment donkey anti-rabbit IgG (Jackson Immunoresearch) was used as secondary antibody for staining pAkt473 and pS6. The following antibodies to ICOS (Clone 7E.17G9) and CTLA-4 (Clone UC10–4B9) were also used.

In vitro cell culture and suppression assay

Sorted cells were re-suspended in complete clicks media (Irvine Scientific) along with 10% FBS, 1% Pen/Strep, 1% L-glut (2mM), 55μM βMe. 48 well flat bottom plates were coated with plate bound αCD3 (2μg/ml, Clone 145–2C11) and αCD28 (2ug/ml, Clone 37.51) in 1X PBS overnight. For proliferation and suppression assays, cells were labeled with cell-trace violet (Life Technologies) as indicated. The in vitro differentiation assay conditions used for 3 days were: tTreg and Th0- h.rIL-2 (RnD systems) 100IU/ml; iTreg- IL-2 100 IU/ml, h rTGF-β 10ng/ml (Peprotech), α-IL4 and α-IFNγ 10μg/ml; Th1- h.r.IL-2 100IU/ml, mouse IL-12 5ng/ml (Peprotech) and α-IL4 10ug/ml. For differentiated cells analyzed at day 5, cells were split 1:2 on day 3 and rested with IL-2 100IU/ml for 48 hours before analysis as described previously (10). For suppression assays, Treg cells at day 5 were cocultured with labeled CD45.1⁺ total T cells in the presence of soluble αCD3 at 0.25ug/ml and irradiated APCs (5×10⁴ cells) at the indicated Treg:T cell ratios for 3 days and proliferation of T responders was analyzed on day 8 (19).

Metabolic assays

The extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were measured using the seahorse bioflux analyzer (18). Glucose derived lipogenesis assay was measured using radiolabeled 14C-U-glucose as described previously (15). Samples for the untargeted metabolite profiling were processed as described previously (20). The peak intensities were analyzed using Metaboanalyst software. All the data were subject to column-wise normalization method. The media supernatants were used to analyze the changes in glucose, glutamine, and lactate and glutamate levels using the YSI 2950 metabolite analyzer.

Real-time PCR and immunoblotting

QPCR was performed on cDNA using the SYBR green ROX PCR mastermix (Qiagen) as per the manufacturer’s protocols. The relative gene expression of β-actin, Glut1, Glut3, Hk2 and Cpt1a was calculated using the ΔΔCt method and normalized to β-actin. Immunoblotting for glucose transporters and β-actin was performed as previously described (10, 21).

Statistics

Groups of more than three were compared using ordinary one-way ANOVA analysis with Tukey post-test analysis using GraphPad Prism software.

RESULTS AND DISCUSSION

TGF-β inhibits glycolysis of tTregs

For the studies below, absent an agreed upon surface marker for tTregs, we have considered Tregs isolated directly ex vivo from spleen and peripheral non-mesenteric lymph nodes as being of thymic origin as these cells predominantly express neuropilin-1, a marker previously associated with thymically derived Tregs (22). To determine the metabolic signature of tTregs, we activated YFP⁺CD4⁺ (tTregs) in the presence of IL-2 for 3 days while simultaneously activating YFP- Tconv cells under Th0 conditions (IL-2 alone) or Th1-differentiating conditions (in the presence of IL-2 and IL-12) and iTreg inducing conditions (in the presence of IL-2 and the anti-inflammatory cytokine TGF-β) (Fig. 1A). By Seahorse bioenergetic flux analysis, we found tTregs exhibited essentially equivalent glycolytic function (ECAR) as Th1 cells (Fig. 1B). In contrast, iTregs had a relatively lower ECAR as well as basal OCR compared to tTregs (Fig. 1B).

Figure 1: TGF-β inhibits glycolytic metabolism in tTregs.

(A). Schematic of the 3 day activation protocol of tTregs, T effector and iTreg differentiation, and culture of tTregs in iTreg (i.e., TGF-β). (B). The ECAR (n=4), OCR (n=3) in all the cell types tested (C). Histograms of Foxp3 expression (% and MFI) in day 3 activated cell types as indicated respectively (n=4). (D).Protein expression of Glut1, Glut3 in the indicated cell types.Result is representative of 4 independent experiments. (E). Relative mRNA expression of Hk2. Each point is indicative of a single experiment. All the groups in were compared using one way-ANOVA analysis with Tukey post-test analysis. P values *, **, *** and **** indicate p<0.05, p<0.01, p<0.001 and p<0.0001 respectively. Error bars represent means ± SD.

TGF-β, an anti-inflammatory cytokine, which facilitates the development of iTregs, functions in a complimentary role to Foxp3 in promoting mitochondrial oxidative metabolism (13, 14). Since a key difference in the culture conditions for tTregs and iTregs is the use of TGF-β in iTreg differentiation, we stimulated tTregs in iTreg conditions (i.e., including TGF-β) to determine the effect of TGF-β on tTreg metabolism (Fig. 1A). Although, inclusion of TGF-β in the cultures had no adverse effect on tTreg survival (data not shown) or Foxp3 expression (% and MFI) (Fig. 1C), it dramatically lowered glycolytic function of tTregs (Fig. 1B). Furthermore, TGF-β also blunted glutamine consumption, a key anaplerotic event that accompanies glycolysis in activated cells (data not shown).

A key checkpoint for glycolysis is glucose uptake, which is controlled by the glucose transporter family of proteins. Among the 13 members, only four (Glut1, Glut3, Glut6 and Glut8) are detected in activated T cells with Glut6 and Glut8 exhibiting lower copy numbers than Glut1 and Glut3. Interestingly, neither tTregs nor iTregs require Glut1 for their development or function in vivo, suggesting that other members of the Glut family may be playing a non-redundant role in these subsets (21). In accordance, tTregs in fact expressed much higher levels of Glut1 and Glut3 protein, like Th1 cells, while iTregs exhibited either low or undetectable levels of this protein (Fig. 1D). Most strikingly, TGF-β abrogated both mRNA (data not shown) and protein expression of Glut1 and Glut3 in tTregs (Fig. 1D). TGF-β exposure also dramatically inhibited tTreg expression of Hk2, a key enzyme in proximal glycolysis, (Fig. 1E) suggesting a metabolic reprogramming of tTregs in the presence of TGF-β.

TGF-β inhibits glucose metabolism and suppressive capability via inhibition of PI3K mediated mTOR dependent signaling in tTregs

TGF-β-mediated SMAD3 activation limits CD4 T cell growth and proliferation by attenuation of CD28 mediated costimulation and PI3k/Akt/mTOR pathway activity (23). We therefore examined phosphorylation of S6kinase and Akt, two major downstream targets in this pathway. We found that TGF-β decreased phosphorylation of S6 (an mTORC1 target) and Akt (ser473) (an mTORC2 target) in tTregs (Fig. 2A). Consequently, the overall growth (data not shown) and division of tTregs stimulated in the presence of TGF-β was also diminished significantly suggesting that TGF-β reduces Treg proliferation by inhibiting glycolytic metabolism independent of Foxp3 expression (Fig. 2B). Additionally, the mTOR dependent glycolysis-lipogeneic pathway is critical for the suppressive capability of Tregs (15). We observed that tTregs cultured in the presence of TGF-β exhibited decreased suppressive function in vitro compared to tTregs (Fig. 2C). However, they still exhibited higher levels of suppression than iTregs, suggesting that glycolysis partly contributes to the suppressive function of tTregs. Concomitantly, we also observed a significant reduction in the expression (% and MFI) of key Treg molecules such as ICOS, and CTLA4 to levels resembling that of iTregs (Fig. 2D).

Figure 2: TGF-β diminishes mTOR dependent signaling that is associated with decrease in proliferation and suppressive capability in tTregs.

(A). Histogram plots with average percentages +/− SD of phosphorylated S6 kinase (pS6) (n=5) and phosphorylated Akt473 (pAkt473) (n=3). (B). Histogram plots of dilution of cell trace violet (cell proliferation) in the 4 cell types respectively. Results are representative of 3 independent experiments. (C). Histograms (1:12 ratio) and % inhibition of CD4 T cell proliferation on day 8 of Treg suppression assay. Data is representative of two independent experiments (D). Histogram plots of phenotypic markers such as ICOS and CTLA4. The average % and MFI of each marker within each cell type is indicated. Results represent 4 independent experiments. P values for all the groups were calculated using one-way ANOVA analysis with Tukey post-test. P values *, **, *** and **** indicate p<0.05, p<0.01, p<0.001 and p<0.0001 respectively. Error bars represent means ± SD.

Collectively, these data suggest that TGF-β reprograms tTreg metabolism via blockade of PI3K–mediated mTOR signaling and glucose metabolism. Moreover, given the interplay between metabolic regulation and cellular function, this suggests that TGF-β may selectively modulate the metabolism of tTregs and iTregs in a context-dependent manner, i.e., TGF-β rich environments can normalize the glycolytic differences between tTregs and iTregs, and metabolic differences observed between these subsets in vitro may be a result of their culture conditions rather than a fixed metabolic program.

Effects of manipulating PI3K activity on Treg subsets

Previously, others and we have shown that control of PI3K activity is critical to maintain Treg stability and function (18, 24). Conditions of inflammation such as TLR activation and defective autophagy activate PI3K-Akt mediated mTOR signaling and enhance glycolysis in Tregs, leading to impaired lineage stability and function (25, 26). In T cells, the lipid phosphatase PTEN is the primary negative regulator of PI3K activity. While PTEN is constitutively expressed in all resting T cells, it is downregulated in activated T conv cells, which facilitates their optimal expansion (27). Tregs, on the other hand, have been reported to retain PTEN expression post-activation as a means of avoiding high PI3K activity (27), and we confirmed that this is true in both tTregs and iTregs (Supplemental Fig. 1A, 1B and 1C).

Given that tTreg and iTreg exhibit differential requirements for glycolysis and mTOR signaling, we hypothesized that over activation of PI3K signaling via PTEN loss would have distinct effects on glycolysis in these subsets. To establish a system to study PTEN deficient tTregs we used Tregs isolated from PTEN^fl/fl x Foxp3-YFP-Cre mice, also termed PTEN-ΔTreg mice (18). To obtain PTEN deficient iTregs, we stimulated CD4 T conv cells from either Foxp3-YFP-Cre or PTEN^fl/fl x Foxp3 YFP-Cre mice in iTreg inducing conditions (PTENΔ−iTreg),and confirmed that PTEN expression is decreased following induction of Foxp3 (Supplemental Fig. 1D, 1E and 1F).

Surprisingly, while others and we have previously shown increased rates of aerobic glycolysis in PTEN-deficient tTregs shortly after activation (6–24 hours) (18, 24), we found that by 3 days after activation, ECAR and OCR rates of PTEN-deficient tTregs were indistinguishable from WT tTregs (Fig. 3A). This indicated that the ability of PTEN-deficient tTregs to undergo glycolysis is restored to WT levels during the late phase of activation while there is no major change in mitochondrial metabolism in tTregs in the absence of PTEN. This suggested that glucose might be redirected to other metabolic pathways other than lactate production in PTEN-deficient tTregs. One of the alternate fates of glucose other than aerobic glycolysis is entry into the pentose-phosphate pathway (PPP), which feeds directly into nucleotide biosynthesis. To better define this, we performed an unlabeled metabolomic analysis of freshly isolated WT and PTEN-ΔTregs. Metabolite enrichment analysis revealed enhanced nucleotide biosynthetic metabolites, including ADP, ATP, GTP and UTP in PTEN deficient tTregs compared to WT tTregs (Supplemental Fig.2A), These observations were consistent with our previous findings showing enhanced BrdU labeling of PTEN-ΔTregs compared with WT Tregs suggesting their rapid turnover in vivo (18). We also found that PTEN deficiency enhanced ¹⁴C glucose labeling of lipids in tTregs suggesting enhanced lipogenesis (Supplemental Fig. 2B).

Figure 3: Elevation of PI3K signaling specifically enhances glycolytic metabolism of iTreg subsets but not tTregs.

(A) and (B). Measure of ECAR and OCR of day 3 activated tTregs and iTregs from both WT and PTEN ΔTreg mice respectively. Results represent 4 independent experiments. (C) Foxp3 expression (% positive and MFI) in day 4 stimulated tTregs and iTregs that are either WT or PTEN deficient and Th1 cells (control) respectively. Results represent three mice per group. (D). tTregs and iTregs that are either WT or PTEN deficient were activated for 3 days in their respective conditions and rested for 2 days with IL-2. Total (both surface and intracellular) Glut1 expression (% positive and MFI) in these stimulated cells was analyzed on day 5. Results represent three mice per group. P values determined by one-way ANOVA analysis with Tukey post-test analysis. *, **, *** and **** indicate p<0.05, p<0.01, p<0.001 and p<0.0001 respectively. Error bars indicate means ± SD.

Interestingly, PTEN loss in iTregs predominantly upregulated aerobic glycolysis with no changes in oxidative metabolism (Fig. 3B). Notably, we observed a significant reduction in Foxp3 expression in PTEN deficient iTregs, but not in tTregs by day 4-post activation (Fig. 3C). Interestingly, by day 5 of activation post IL-2 culturing, PTEN deficient iTregs exhibit higher Glut1 protein expression compared to PTEN deficient tTregs (Fig. 3D). Taken together, this indicated that while enhanced PI3K activity in tTregs led to the diversion of glucose towards nucleotide biosynthesis and the glycolytic-lipogenesis pathway without substantively altering glycolysis, in iTregs it directly enhances glycolysis (Supplemental Fig. 2C). Collectively, these findings demonstrate that upregulation of PI3K signaling has different metabolic consequences in Treg subsets.

In conclusion, our study indicates that tTregs and iTregs are metabolically distinct, but subject to modulation by external and context dependent cues, such as inflammatory or anti-inflammatory stimuli (PI3K activation and TGF-β, respectively). The role of this metabolic plasticity is likely adaptation to the local environment, balancing cell division, cell survival and regulatory function to appropriately meet the needs of the organism.

Abbreviations:

PTEN

Phosphatase and tensin homolog

tTregs

Thymically derived Tregs

iTreg

Invitro induced Tregs

pTregs

peripherally induced Tregs