Restricting Glutamine or Glutamine-Dependent Purine and Pyrimidine Syntheses Promotes Human T Cells with High FOXP3 Expression and Regulatory Properties

Regulatory T cells (Tregs) expressing the transcription factor FOXP3 are pivotal for balanced immunity, as revealed by the severe immune dysregulation and autoimmune pathology caused by FOXP3 loss-of-function mutations (1). Evidence of Treg-mediated protection in numerous preclinical models of autoimmunity and transplantation has sparked efforts to exploit their therapeutic potential for human therapy (2, 3). FOXP3⁺ Tregs may be of thymic origin or develop from mature peripheral T cells (4). Clinical Treg applications have largely relied on cellular therapies using in vitro–expanded pre-existing Tregs rather than induced Tregs due to the superior stability and more clearly defined functionality of the former (5, 6). If Tregs were promoted in vivo rather than supplemented through cellular therapy, it might be possible to tap the potential of multiple types of Tregs through targeting pathways that selectively favor them over T effector cells.

T cell subsets, including Tregs and T effectors, differentially use metabolic pathways to meet specific nutritional needs according to their function and activation state (7, 8). Insight into T cell subtype-specific metabolic requirements might be harnessed, therefore, to selectively promote Tregs over T effectors. Tregs are also preferentially induced or selectively advantaged in the presence of oxidative stress, in situations of general low energy as recapitulated by loss of signals from the metabolic hormone leptin, and thanks to their superior capacity to use preformed fatty acids when de novo synthesis is blocked (9–11). Collectively, these findings appear to indicate that Tregs may also be more resilient under metabolic stress.

Besides glucose, Gln is a critical and versatile metabolite in multiple pathways that sustain cellular integrity and the synthesis of macromolecules required for cell growth and division, as well as antiapoptotic mechanisms under metabolic stress (12). As Gln requirements often exceed cellular capacities for endogenous synthesis, Gln has been termed a conditionally essential amino acid. Accordingly, Gln is by far the most highly concentrated amino acid in serum (∼0.6 mM) and standard cell culture media (2 mM). Although the importance of Gln for activated T cells is well established, it has not been defined whether T cell subsets differ in their Gln requirements (13–15).

In the present study we have analyzed the consequences of Gln restriction for the development of human FOXP3^hi and FOXP3^lo/−CD4 T cells during activation in vitro. Limiting Gln favored FOXP3^hi cells by diminishing FOXP3^lo/−CD4 T cell numbers and proliferation, and through an increase in FOXP3^hi cell numbers. Low Gln-induced constraints on proliferation and cell numbers were more closely mimicked by targeting the purine and pyrimidine synthesis pathways, respectively, and the absolute increase in FOXP3^hiCD4 T cell numbers between 0.2 and 0.4 mM Gln was abrogated by inhibiting Gln synthetase (GS), suggesting a superior capacity of activated FOXP3^hi cells to produce endogenous Gln under limiting extracellular Gln.

CD25^hiCD4 T cells enriched in FOXP3^hi cells through activation under low Gln or purine/pyrimidine pathway inhibition impaired the proliferation of primary T cells, but they also produced inflammatory cytokines, including IL-17A. IL-17A producers were predominantly CTLA-4^hi cells. Although IL-17A production coincided with weaker suppression by FOXP3^hi-enriched CTLA-4^hi over CTLA-4^lo/− subpopulations, neutralizing IL-17A with the former and adding IL-17A to the latter did not change their inhibitory capacities with respect to in vitro proliferation. Taken together, these results indicate that restricting Gln or Gln-dependent nucleotide synthesis promotes functionally suppressive human FOXP3^hi T cells, yet it also spares IL-17A–producing CD4 T cells.

PBMC were isolated from buffy coats (blood donation foundation SRK Aargau-Solothurn Kantonsspital Aarau AG and blood donation service SRK Bern AG) by density gradient using Ficoll-Paque plus (GE Healthcare) in Leucosep tubes (Greiner Bio-One). Residual RBC were removed using RBC lysis buffer (Amimed). CD4⁺CD25⁻ T cells were enriched from freshly isolated PBMC using EasySep CD4⁺ T cell enrichment and CD25 depletion kits (Stemcell Technologies). The purity of the enriched cells was >95%.

For experiments on Gln restriction, culture medium consisted of Gln-free RPMI 1640 (Lonza, BioWhittaker) supplemented with varying concentrations of l-glutamine (Sigma-Aldrich), low TGF-β1 (0.2 ng/ml, R&D Systems), and 5 ng/ml IL-2 (R&D Systems), a concentration determined in pilot studies to be suboptimal for promoting FOXP3^hi T cells in the absence of additional supporting conditions. For all other assay conditions, RPMI 1640 with GlutaMAX (Life Technologies) was used. Treg suppressor assays were not supplemented with TGF-β and IL-2. All media were supplemented with 10% heat-inactivated FCS (PAA Laboratories) that was dialyzed through 3.5-kDa size-exclusion membranes. This reduced the Gln in FCS from 782 to 46 μM, as determined by liquid chromatography–mass spectrometry, thereby adding another 4.6 μM Gln to the indicated Gln concentrations. Methionine sulfoximide (MSO), dimethyl α-ketoglutarate, glutamate, and mycophenolic acid were from Sigma-Aldrich. N-phosphonacetyl-l-aspartate (PALA) was from the Novartis compound archive. Abs included purified anti-human CD28 (clone CD28.2) and fluorochrome-labeled CD4 (clone RPA-T4) and CD25 (clone M-A251 and clone 2A3), which were from BD Biosciences. Fluorochrome-labeled anti-human FOXP3 (clone PCH101) and CTLA-4 (clone L3D10) were from eBioscience; fluorochrome-labeled anti–IL-17A (clone eBIO64DEC17) was from BioLegend. Purified anti-human CD3 (clone OKT3) was provided by Novartis. Cell tracer dye CFSE and anti-CD3/CD28–coated Dynabeads (type 11129D) were from Life Technologies.

PBMC or CD4⁺CD25⁻ T cells were nucleofected with 300 nM small interfering RNA (siRNA) using P3 Primary Cell 4D-Nucleofector kit with an Amaxa 4D-Nucleofector device according to the manufacturer’s protocol (Lonza). siRNAs were from GE Dharmacon and included human carbamoyl phosphate synthetase/aspartate transcarbamylase/dihydroorotase (CAD; E-009471-00-0005), glutamine phosphoribosylpyrophosphate amidotransferase (PPAT; E-006003-00-0005), and nontargeting SMARTpool siRNA (D-001910-01-05, no. 1). Nucleofected cells were rested for 2 h at 37°C in culture medium. Expression knockdown was confirmed by real-time quantitative PCR (RT-qPCR).

Freshly harvested cells were lysed and their RNA was extracted according to the manufacturer’s protocols using RNeasy Mini or Micro kits (Qiagen). cDNAs were generated from 200 ng isolated RNA using the high-capacity cDNA reverse transcription kit from Life Technologies. To analyze gene expression, RT-qPCR was performed in triplicate according to the TaqMan Master Mix protocol and run on the QuantStudio 12K Flex (Life Technologies). TaqMan primers and probes were from Life Technologies and included CAD (assay ID Hs00983188_m1), PPAT (assay ID Hs00601264_m1), and 18S rRNA (assay ID 4319413E). The amounts of mRNA expression were normalized to 18S rRNA and calculated according to the comparative cycle threshold (Ct) method (16).

For Gln restriction experiments, freshly isolated PBMC were used without further treatment or stained with 1 μM CFSE according to the manufacturer’s protocol. Two to 2.4 × 10⁶ cells per well were plated in 24-well flat-bottom plates in medium containing specified l-glutamine concentrations. In experiments with small molecule inhibitors, cells were preincubated with inhibitors at 37°C for 30 min before T cell activation was started with soluble 0.03 μg/ml anti-CD3 and 0.3 μg/ml anti-CD28. siRNA-transfected cells were activated in full 2 mM Gln-containing medium under the same culture conditions. After 4 d, cells were harvested and analyzed by flow cytometry for CD4, CD25, and FOXP3 expression. When cells were expanded to be tested as Tregs in functional assays, they were split after 3 d and grown for another 3 d.

Cells were surface stained with fluorochrome-labeled Abs using standard procedures (30 min at 4°C in PBS, 2% FCS, 2 mM EDTA, 0.1% NaN₃). For intracellular staining of FOXP3, CTLA-4, and cytokines, buffers and reagents from eBioscience were used according to the manufacturer’s instructions. Data were acquired on an LSR II flow cytometer and analyzed using FlowJo analysis software (Tree Star). Total cell numbers were determined from acquired cell numbers within the live forward/side scatter gate and fixed numbers of beads (CompBeads, BD Biosciences) added to each sample prior to FACS acquisition. To enrich FOXP3^hi cells for functional assays, cells were surface stained for CD4, CD25, and CTLA-4 and sorted on a FACSAria Fusion (BD Biosciences) using a 70-μm ceramic nozzle with a sheath pressure of 70 psi and an acquisition rate of 12,000 events/s. Gated CD4⁺ cells were sorted based on CD25^hi or CD25^hiCTLA-4^{hi or lo/−} subgates as depicted in the figures.

CD4⁺CD25⁻ CFSE-labeled autologous T responder cells (Tresp; 3 × 10⁴) were cocultured with sorted cells obtained from activation cultures under FOXP3^hi T cell–promoting conditions and activated with Dynabeads at a 1:2 bead/total cell ratio in round-bottom 96-well plates. After 4–5 d, supernatants and cells were harvested. Tresp were analyzed by flow cytometry for CFSE staining intensity, and in some experiments also for intracellular cytokines. To detect intracellular cytokines, cells were restimulated for 4 h with PMA and ionomycin in the presence of brefeldin A (Leukocyte Activation Cocktail, BD Biosciences). Cytokines in supernatants were measured by a multiplex electrochemoluminescence assay (Meso Scale Discovery) or ELISA (eBioscience) kit. To test the role of IL-17A during the suppressor assay, a neutralizing anti–IL-17A Ab (clone MAB317; R&D Systems) or isotype control (MAB004; R&D Systems) was added at the beginning of the assay at 1.5 μg/ml. Recombinant human IL-17A (Novartis Institutes for BioMedical Research) was added at 40 ng/ml.

Limiting Gln during the activation of human PBMC with anti-CD3/CD28 consistently enriched high FOXP3-expressing CD4 T cells (Fig. 1A). Activated (CD25⁺) CD4 T cells from all donors gradually increased in FOXP3^hi cells as Gln concentrations were reduced (Fig. 1B), although exact dose responses could vary between donors (Fig. 1C).

We subsequently asked what Gln-consuming pathways could account for the different Gln requirements of FOXP3^lo/−CD25⁺ and FOXP3^hi cells (Fig. 2). Gln, upon conversion to α-ketoglutarate via glutamate, is a key substrate to replenish TCA cycle metabolites and support NADPH generation for macromolecular synthesis during cell growth and proliferation (17, 18). If these pathways contributed to FOXP3^hi cell enrichment through Gln restriction, then providing extra α-ketoglutarate or glutamate should at least partially compensate for Gln to reverse this effect. This was not the case, however. Supplementing media with α -ketoglutarate (in its cell-permeable form dimethyl-α-ketoglutarate) did not compromise FOXP3^hi cell enrichment under limiting Gln (Fig. 3A). At concentrations >2 mM, excess dimethyl-α-ketoglutarate appeared to be generally cytoreductive or toxic in these cell cultures (not shown). Extra glutamate did not change overall FOXP3^hi cell proportions (Fig. 3B). In some experiments, glutamate supplements moderately reduced total cell numbers of both FOXP3^hi and FOXP3^lo cells without changing their proportions (Fig. 3B). Taken together, these data suggest that Gln availability for these pathways does not substantially influence the balance between FOXP3^hi and FOXP3^lo/−CD25⁺ cells upon T cell activation in this model.

Earlier studies had demonstrated a nonredundant role of Gln for T cell proliferation that could not be replaced by glutamate or other amino acids (13, 15). We next asked how Gln restriction affected the proliferation of activated FOXP3^lo/− and FOXP3^hiCD4 T cells. The CD25⁺FOXP3^lo/− population progressively lost cell numbers and cell cycles as Gln concentrations declined (Fig. 4, left). FOXP3^hi cells usually underwent one cell cycle less than did FOXP3^lo/− cells at high Gln; they were not constrained by Gln reductions down to 0.4 mM Gln and only weakly at 0.2 mM. Furthermore, FOXP3^hi cell numbers were consistently higher at intermediate than at high Gln concentrations (Fig. 4, right).

Gln is the main nitrogen source for de novo nucleotide synthesis in animals (Fig. 2). If skewing toward FOXP3^hi cells under low Gln depended on impaired pyrimidine and/or purine de novo synthesis, then inhibiting these pathways should promote FOXP3^hi cells similarly to Gln reduction. We initially tested this possibility with small molecule inhibitors.

PALA selectively blocks the trifunctional Gln-consuming enzyme CAD that catalyzes the first three reactions of the de novo pyrimidine synthesis (19, 20). The first committed Gln-consuming rate-limiting step of the purine de novo synthesis is catalyzed by PPAT. Because there is no selective PPAT inhibitor we initially used mycophenolic acid (MPA), a selective inhibitor of inositol monophosphate dehydrogenases 1 and 2 (IMPDH1/2). Its short-lived product, XMP, is the substrate for, and determines the rate of, Gln-dependent GMP synthesis (Fig. 2).

PALA or MPA added to full 2 mM Gln during T cell activation increased FOXP3^hi cells to levels otherwise achieved through Gln restriction (Fig. 5). The consequences for cell divisions and cell numbers, however, could differ between pyrimidine and purine inhibition, as well as between FOXP3^lo/−CD25⁺ and FOXP3^hi cells. In the experiment depicted in Fig. 5A, 100 nM MPA enriched FOXP3^hi cells similar to 0.2–0.4 mM Gln; 100 nM MPA recapitulated FOXP3^lo/−CD25⁺ cell numbers and divisions at 0.4 mM, whereas it more closely resembled FOXP3^hi cell divisions at 0.2 mM and FOXP3^hi cell numbers at 0.4 mM Gln. FOXP3^hi cell enrichment with 100 μM PALA was comparable to 0.2 mM Gln, but proliferation was less strongly reduced with PALA for both FOXP3^lo/−CD25⁺ and FOXP3^hi cells.

As T cells from different donors could vary in their dose responses to Gln, MPA, and PALA, and because inhibition of each purine and pyrimidine synthesis appeared to better reflect Gln restriction–induced inhibition of proliferation and constraints on FOXP3^lo/−CD25⁺ cell numbers, respectively, further experiments included titrations of MPA and PALA alone and in combination. For T cells from two different donors, the bar graphs of Fig. 5B depict concentrations and combinations of MPA and PALA that induced FOXP3^hi cell enrichment closest to 0.4 and 0.2 mM Gln, the Gln range that consistently resulted in an absolute increase or stable FOXP3^hi cell numbers compared with standard 2 mM Gln.

For FOXP3^hi cells from both donors, the increase at 0.4 mM Gln was best mimicked by MPA (Fig. 5Bb, 5Bd). For donor 2 FOXP3^lo/−CD25⁺ T cells, inhibition of proliferation and numbers at 0.4 mM Gln were similar to MPA or PALA (Fig. 5Bc). An MPA/PALA combination closer to 0.4 mM Gln was not identified in this experiment. In contrast, donor 1 FOXP3^lo/−CD25⁺ cell numbers and proliferation at 0.4 mM Gln were most closely recapitulated by combined MPA and PALA (Fig. 5Ba).

Matching cell numbers and divisions at 0.2 mM Gln required combinations of MPA and PALA for FOXP3^lo/−CD25⁺ T cells from both donors (Fig. 5Be, 5Bg) and for FOXP3^hi cells from donor 2 (Fig. 5Bh). FOXP3^hi T cell divisions and numbers from donor 1at 0.2 mM were more closely matched by PALA between no and four divisions and by combined PALA and MPA between four and six divisions (Fig. 5Bf).

To validate and compare directly the importance of the rate-limiting Gln-consuming steps of de novo purine and pyrimidine synthesis for the FOXP3^hi/FOXP3^lo/−CD25⁺ cell balance, we specifically knocked down CAD and PPAT using stabilized selective siRNA pools. Upon T cell activation, CAD and PPAT were transiently upregulated with similar kinetics, and siRNA targeting reduced their expression by >80% in CD4 T cells or PBMC (Fig. 6A). In addition to CAD knockdown, PPAT knockdown also resulted in increased FOXP3^hi cells, thereby confirming the relevance of de novo purine synthesis for FOXP3^hi cell enrichment through inhibiting the first and rate-limiting Gln-dependent step (Fig. 6B).

Cells may generate Gln through GS, a tightly controlled enzyme that is positively regulated by low Gln and nucleotide starvation (Fig. 2) (21, 22). It seemed conceivable that the resilience of activated FOXP3^hi cells to Gln restriction was due to a superior capacity to generate endogenous Gln. This possibility was tested with the selective GS inhibitior MSO (23, 24). As with dose responses to Gln and other inibitiors, the sensitivity to MSO at different Gln concentrations could vary between donors. Two general consistent observations with all donors were, however, that MSO at full 2 mM Gln already substantially reduced FOXP3^lo/−CD25⁺ but not FOXP3^hi cell numbers, and that MSO always abolished the characteristic increase and maintenance of FOXP3^hi cells at intermediate–low Gln. With activated T cells from donor 1, MSO reversed the increase of FOXP3^hi cells between 2 and 0.2 mM Gln and caused a decline comparable to FOXP3^lo/−CD25⁺ cells in the presence of MSO (Fig. 7A, left). The analysis of cell numbers per cell division revealed that MSO caused an overall drop in cell numbers per cell cycle without reducing cell divisions (Fig. 7B, left). Donor 2 FOXP3^hi T cells peaked at 0.4 mM Gln, and this was reduced by MSO to FOXP3^hi baseline levels at 2 mM Gln (Fig. 7A, right and Fig. 7B, upper right).

We next aimed to test the suppressor functions of FOXP3^hi cells that emerged under low Gln and through inhibiting nucleotide synthesis. Because we did not identify surface markers or marker combinations that would clearly distinguish between CD25⁺FOXP3^lo/− and FOXP3^hi cells (e.g., CD30, GITR, GARP, CD39, CD73, CD127; data not shown), we initially enriched FOXP3^hi cells further by sorting the 10% highest CD25-expressing CD4 T cells. This resulted in ∼60–70% FOXP3^hi-enriched CD25^hi cells that strongly suppressed the proliferation of CD4 T cells (Tresp) when added in equal numbers and more weakly at <1:1 ratios (Fig. 8A). FOXP3^hi-enriched T cells alone or cultured together with Tresp did not consistently produce elevated IL-10 (not shown) but did reproducibly secrete effector cytokines, including IL-17A, IL-4, and IFN-γ (Fig. 8B).

We next asked whether effector cytokine production could be separated from suppressor functions. The pivotal checkpoint inhibitor CTLA-4 was shown to be highly expressed by human Th17 cells and murine Th2 cells (25–27). Although CTLA-4 is more prominently known as a marker and functional mediator of Tregs, CTLA-4–mediated regulation does not depend on its expression by FOXP3⁺ Tregs, and CTLA-4–independent Treg-mediated suppression has been reported for in vitro models (28). When we sorted CD25^hi cells as separate CTLA-4^lo/− and CTLA-4^hi subpopulations, the latter was more strongly enriched in FOXP3^hi cells (Fig. 9A). CTLA-4^hiCD25^hi cells, despite larger proportions of FOXP3^hi cells, inhibited more weakly than did CTLA-4^lo/−FOXP3^hi–enriched T cells, and not at all with cells from one donor (Fig. 9B). CTLA-4^lo/−FOXP3^hi–enriched T cells from CAD siRNA nucleofected cultures (siCAD) and PPAT siRNA nucleofected cultures (siPPAT) consistently shifted Tresp proliferation toward fewer cell cycles (Fig. 9B).

IL-17A and IL-4 were predominantly produced by CTLA-4^hi FOXP3^hi–enriched T cells cultured alone or in cocultures with Tresp, whereas IFN-γ and other cytokines such as TNF-α and IL-13 were more variably secreted by both CTLA-4^lo/− and CTLA-4^hi FOXP3^hi-enriched T cells (Fig. 9C, Supplemental Fig. 1).

Cocultures with Tresp mostly contained higher cytokine concentrations than did CTLA-4^hiFOXP3^hi−–enriched T cells alone, even though cytokine production by Tresp alone is barely visible on the y-axis scales in Fig. 9C.

To identify cytokine producers in cocultures, intracellular cytokine staining was performed. This revealed prominent IL-17A⁺ populations among T cells originating from CTLA-4^hiFOXP3^hi–enriched T cells but few or none among Tresp from either single cultures or cocultures (Fig. 9D). Taken together, these data suggest that enhanced cytokine production in Tresp/FOXP3^hi-enriched T cell cocultures over FOXP3^hi-enriched T cells alone was attributable to Tresp-aided increased cytokine production by FOXP3^hi-enriched T cells. IL-17A–producing cells were detected among FOXP3^lo/− cells, but also among the small proportions of sorted FOXP3^hi-enriched T cells that remained FOXP3^hi at the end of the suppressor assay (Fig. 9D).

Given the correlation between high IL-17A production and reduced suppressor capacity of CTLA-4^hi over CTLA-4^lo/− FOXP3^hi-enriched T cells, we interrogated the functional role of IL-17A for inhibition of Tresp proliferation in these assays. A neutralizing anti–IL-17A Ab was added to cocultures of Tresp with CTLA-4^hiCD25^hiFOXP3^hi–enriched cells and rIL-17A was included at 40 ng/ml in cocultures of Tresp with CTLA-4^lo/−CD25^hiFOXP3^hi–enriched cells (Fig. 9E). However, these reagents, although validated for their biological activity in other functional assays (data not shown), had no influence on the inhibition of Tresp proliferation (Fig. 9E).

Gln is an abundant, versatile metabolic substrate in high demand by activated T cells. In this study, we show that limiting Gln during T cell activation induces a profound shift toward high FOXP3–expressing CD4 T cells. We have investigated the underlying cellular changes, identified corresponding Gln-dependent metabolic pathways, and demonstrated regulatory properties of FOXP3^hi-enriched CD4 T cells emerging from T cell activation under Gln restriction and through pathway inhibition.

Cellular activation may enhance the flux of Gln through reactions that generate energy, metabolites, and NADPH for the biosynthesis of macromolecules required for cell growth and proliferation. These Gln functions may be replaced by glutamate or α-ketoglutarate (17, 18). In a recently published study using murine T cell cultures, a proportional increase in Foxp3⁺ T cells through Gln deprivation was reversed by α-ketoglutarate supplements (29). We had generated similar data with murine T cells (our unpublished data), but never with human T cells. It remains to be investigated whether these discrepancies reflect real differences in the T cell biology of mice and humans, or whether the different T cell sources, that is, spleen and lymph nodes versus blood, could account for these differences.

As an excess supply of the Gln-derived metabolites glutamate and α-ketoglutarate did not prevent the contraction of activated FOXP3^lo/−CD4 T cells and the shift toward FOXP3^hi cells under low Gln in our human T cell cultures, this implied a role for limiting Gln-derived amide rather than the Gln carbon scaffold. Furthermore, the tight correlation between gradually reduced Gln and CD25⁺FOXP3^lo/− cell divisions recapitulated earlier evidence of a nonredundant role of Gln for T cell proliferation, suggesting insufficient Gln-derived amide for nucleotide biosynthesis (13, 15). This was supported by the ability to reproduce the effects of Gln restriction by inhibiting Gln-dependent de novo purine and pyrimidine synthesis. The purine synthesis inhibitor MPA often more closely matched Gln deprivation–induced inhibition of proliferation while leaving higher cell numbers, whereas the pyrimidine inhibitor PALA more strongly reduced T cell numbers.

Lower numbers of activated FOXP3^lo/−CD4 T cells per cell division possibly resulted from apoptosis. Although not measured in our experiments, earlier studies identified apoptosis as a consequence of human T cell activation in the presence of pyrimidine inhibitors whereas purine inhibitors primarily impaired cell cycle progression (30). We found that the increase and resilience of FOXP3^hiCD4 T cells under Gln restriction was abolished by blocking GS. GS inhibition did not reduce cell cycles but selectively affected cell numbers, similar to inhibiting pyrimidine synthesis. GS was previously shown to selectively support pyrimidine and not purine synthesis, and pyrimidine rather than purine synthesis was required to protect cells from inflammation-induced apoptosis (31, 32). Inflammation and tissue injury are physiological conditions of Gln reduction, and GS is de-repressed under low Gln, nucleotide starvation, and by glucocorticoid hormones (22, 33–35). Collectively, these findings suggest that Gln synthesized by GS in stressful situations is primarily used for pyrimidine synthesis, and is more important for cell survival than cell proliferation.

The different consequences of GS inhibition at full or suboptimal Gln for activated FOXP3^lo/− and FOXP3^hiCD4 T cells raise questions about the molecular mediators behind FOXP3-associated differential “wiring” between extracellular Gln and GS activity. Although entirely speculative at present, these might include FOXO transcription factors previously shown to support GS expression in skeletal muscle and cell lines, as well as FOXP3 expression in T cells (36–38). The extensive regulation of FOXOs includes feedback loops to control their own transcription and activation as well as FOXP3 within the PTEN/PI3K/Akt/mTOR pathway (38, 39). Inhibition of mTOR favors Tregs, and factors that control mTOR activity include Gln import through the transporter SLC1A5 (40, 41). Although it seemed possible that SLC1A5 inhibition would recapitulate FOXP3^hi cell promotion through Gln restriction, we found that siRNA knockdown of SLC1A5 diminished activated T cells without favoring FOXP3^hi T cells, and it generally reproduced the functional profile previously published for SLC1A5 knockout mice but not the consequences of Gln restriction in our model (our unpublished data) (42). More studies are needed to identify nodes that connect Gln, GS, FOXP3, and possibly FOXO and/or other factors with T cell fate and function.

Gln concentrations in vivo may drop transiently during exercise and inflammation, thus raising questions about the stability and function of FOXP3^hiCD4 T cells emerging under such temporary metabolic stress. FOXP3^hi cells enriched through Gln restriction or pathway inhibition reduced the proliferation of primary T cells, although high FOXP3 expression was largely lost during the 4-d suppressor assay. This demonstrated that high FOXP3 expression was stable only for as long as these metabolic constraints persisted and implied that a limited period of suppression following coculture of FOXP3^hi-enriched T cells with Tresp was sufficient to reduce Tresp cycle numbers. This would be in line with previous studies demonstrating in a mouse model that the first 6–12 h of Treg/Tresp coculture were necessary and sufficient to inhibit Tresp proliferation (43).

Because CTLA-4^hiCD25^hiCD4 T cells were somewhat more strongly enriched in FOXP3^hi cells, we might have expected them to be stronger suppressors for this reason but not due to their higher CTLA-4 expression, as CTLA-4–mediated inhibition was shown to operate via cell-extrinsic mechanisms through the shared CD28/CTLA-4 ligands CD80/CD86 that were not specifically provided in our assays (44). Despite their moderately larger FOXP3^hi cell enrichment, CTLA-4^hiCD25^hi cells reduced Tresp cycles less strongly than did CTLA-4^lo/−CD25^hi cells. The former were the predominant producers of IL-17A and IL-4, whereas other cytokines, including IFN-γ, TNF-α, and IL-13, could be similarly produced by CTLA-4^lo/−CD25^hi cells.

Given the possible associations between inferior Treg functions during some autoimmune conditions, including those with IL-17A components, and in view of the high IL-17A concentrations produced by CTLA-4^hiCD25^hi T cells during the suppressor assays, we speculated that IL-17A might be associated with inferior inhibition of proliferation in these assays. This was not the case, however. A neutralizing anti–IL-17A Ab did not improve inhibition by CTLA-4^hiCD25^hi cells, nor did addition of IL-17A affect suppression by CTLA-4^lo/−CD25^hi cells. Although the latter could be explained by a failure to respond to IL-17A, the former would have been expected to remove a potential causal link between the presence of IL-17A and inferior inhibition of proliferation by CTLA-4^hiCD25^hi cells. Generally, our data seem to be in line with previous studies showing full suppression by a subset of IL-17A–producing human ex vivo isolated Tregs, although IL-17A concentrations in these studies were, at ∼0.5 ng/ml, ∼20- to 100-fold lower than those measured in our assays (45, 46). IL-17A producers in our suppression assays could have included FOXP3^hi cells, former FOXP3^hi cells that lost FOXP3 expression, or differentiated Th17 cells, with all possibilities having been demonstrated previously, including human and clinical settings (25, 47–50).

Unexpectedly, cytokine production was often higher in cocultures of CTLA-4^hiCD25^hiCD4 T cells with Tresp than in CTLA-4^hiCD25^hiCD4 single cultures, regardless of whether Tresp proliferation was inhibited, and even though Tresp did not substantially contribute to cytokine production. This suggests that primary and resting CD4 T cells enhanced cytokine production by preactivated T effector/memory cells, an interesting phenomenon that demands further analysis and mechanistic understanding to evaluate its physiological relevance.

Because Gln can be limiting under some physiological conditions, as outlined above, our studies may reflect aspects of immune modulation in vivo. A shift toward FOXP3^hiCD4 T cells capable of curbing the proliferation of formerly naive or resting T cells may help to restrain productive (destructive) immunity in the absence of detrimental inflammation, help to avoid prolongation and spreading of an immune response, and prevent an acute immune response from turning chronic.

It is not entirely clear at present how (and if) CTLA-4^hi IL-17A–producing T cells could interfere with Treg functions in vivo. However, observations of seemingly therapeutically ineffective Foxp3⁺CD4 T cells accumulating in inflamed tissues, also in association with IL-17A, suggest that further insight into the effects of IL-17A on Treg functions might help to harness the therapeutic potential of Tregs in IL-17A–driven diseases (49, 51).

FOXP3^hiCD4 T cells maintained under or emerging from activation under low Gln or during impaired de novo purine/pyrimidine syntheses may presumably be suppressive in vivo, both under desirable and undesirable circumstances. Our data might help to understand, for example, why functional Tregs are well maintained in mycophenolate mofetil–treated transplant patients and might further hint at Treg-compatible immunosuppressive treatments (52). On the opposite side of the coin, cancer drugs targeting these pathways could possibly have a suboptimal therapeutic effect in some cases (53, 54). Furthermore, our results might help to explain why Tregs may thrive well within tumors if these are addicted to Gln or, similar to activated FOXP3^hiCD4 T cells in our studies, are also resistant to Gln deprivation through utilizing GS (55).

We thank Franco Di Padova, Friedrich Raulf, and Eva D’Hennezel for critically commenting on the manuscript. We are also grateful to Juan Zhang for analyzing FCS, to Marie-Jo Duriatti for advice on IL-17A assay reagents, and to Friedrich Raulf’s group and Eva D’Hennezel for technical advice on the siRNA technology.

All authors are employees of Novartis, Switzerland.