Main

Mounting a robust and effective adaptive immune response in vertebrates is metabolically costly and requires proper allocation of essential yet limited energy and carbon resources. Metabolism must be tightly controlled at the cellular level to coordinate rapid expansion followed by a fine-tuned differentiation process in T cells. Beyond acting as bioenergetic substrates and biosynthetic precursors, metabolites can directly control cellular signalling responses through influencing DNA, RNA and protein modifications, signalling receptors’ activities and the production of reactive oxygen species1,2,3,4,5. As such, metabolism is fundamental to fine-tuning carbon and nitrogen allocation and optimizing immune response, which is at the centre of many diseases. Previous studies have used systemic approaches to comprehensively characterize the transcriptome, the abundance of intracellular metabolites and the overall catabolic activities of T cells at the different stages during the T cell life cycle6,7. These studies have generated critical temporal snapshots of the metabolic landscapes, which help establish a conceptual foundation for understanding T cell metabolic reprogramming. However, most of these studies have centred mainly on intracellular metabolites and activities of the central carbon metabolism. The overall metabolic landscape of T cells can also be delineated by monitoring the metabolites consumed from and secreted into the growth medium. The extracellular metabolome represents the ultimate outcome of metabolic input, processing and output. Extracellular metabolome profiling (also called metabolic footprinting) has been applied as a standard technique to optimize microbial bioprocesses by analysing substrates consumed from and metabolites secreted into a microorganism’s culture medium8,9. Here, we took a similar approach (Fig. 1a) to compare the extracellular metabolome profiles of naive T (Tnai) cells and different subsets of effector T (Teff) cells, including T helper (TH0, TH1, TH17) cells and induced regulator T (iTreg) cells.

Fig. 1: GAB is an abundant metabolite produced in TH17 and iTreg cells.
figure 1

a, Experimental scheme of T cell extracellular metabolome profiles (LC–MS). b, Extracellular metabolites associated with amino acid metabolism in the indicated T cell subsets were profiled by LC–MS. The value for each metabolite represents n = 3 biologically independent samples. The heatmap represents the value of the relative amount (see colour scale). The complete metabolomic profile is provided as source data. c,d,j,k, Indicated metabolites were quantified by GC–MS. The value for each metabolite represents n = 3 biologically independent samples. Cysteinylglycine disulfide*, (2R)-2-amino-3-{[2-amino-2-(carboxymethylcarbamoyl)ethyl]disulfanyl}propanoic acid; alpha-ketoglutaramate*, 2-keto-glutaramate; 2-oxoarginine*, 5-[(diaminomethylidene)amino]-2-oxopentanoic acid; 2,3-dihydroxy-5-methylthio-4-pentenoate (DMTPA)*, (2R,3R,4E)-2,3-dihydroxy-5-(methylsulfanyl)pent-4-enoic acid. The heatmaps (c,k) represent the log value (medium) or the absolute value (pellet) of the indicated metabolite quantity (see colour scale). The complete data are provided as source data. The volcano plots (d,j) show changes in metabolites in the cell medium and cell pellet. Heatmaps and volcano plots are representative of n = 3 independent biological samples from n = 2 independent biological experiments. e, Schematic of the pathway of GABA metabolism. f, RNA was isolated from the indicated T cell subsets (n = 3 biologically independent samples) and used for qPCR analyses of the indicated metabolic genes. mRNA levels of Tnai cells were set to 1. The heatmap represents the log value of the relative mRNA expression level (see colour scale). Values and s.d. are provided as source data. g,h, GAB in the indicated groups was determined and quantified by NMR (n = 3 biologically independent samples). i, As illustrated by the experimental scheme (left), GAB production in iTreg and tTreg cells was quantified by a GAB bioassay kit (n = 4 biologically independent samples). Statistical analysis was performed by R (b) or unpaired two-tailed Student’s t-test (c,d,i–k). GAD, glutamate decarboxylase; SSADH, succinic semialdehyde dehydrogenase; SSAR, succinic semialdehyde reductase; GHB-R, γ-hydroxybutyrate receptor; GATs, GABA transporters; VGAT, vesicular GABA transporter; GABAA-R, GABA type A receptor; GABAB-R, GABA type B receptors; DBI, diazepam binding inhibitor.

Source data

Full size image

Results

GAB is an abundant metabolite produced by effector T cells

The control (blank) medium and the spent medium of different subsets of Teff cells (Extended Data Fig. 1a) were profiled on a semi-quantitative untargeted global metabolomics platform based on liquid chromatography–mass spectrometry (LC–MS), with broad coverage of up to 1,000–2,500 compounds, including amino acids, energy metabolites, nucleotides and lipids. Using this approach, we have classified metabolites as having changes in production or consumption according to whether the fold change compared with control was positive or negative, respectively. Hierarchical clustering analysis, the pairwise comparison and the principal-component analysis revealed that T cell subsets were characterized by distinct extracellular metabolome profiles (Fig. 1b and Extended Data Fig. 1b–g). Consistent with the role of central carbon metabolism in supporting cell growth, the hyper-proliferative Teff groups consumed more carbohydrates and produced more lactate than the Tnai group (Extended Data Fig. 1e). Additionally, the TH17 group was characterized by the highest production of polyamines (Fig. 1b), in line with the recent finding of a critical role for polyamine in determining TH17 differentiation10,11,12. Intriguingly, iTreg cells produced high levels of γ-aminobutyrate (GAB) and its derivatives (Fig. 1b). Next, we applied gas chromatography–MS (GC–MS)-based targeted metabolomics and nuclear magnetic resonance (NMR) to validate and quantify intracellular and extracellular GAB production. We confirmed that iTreg cells produced much higher levels of GAB than TH17 cells (Fig. 1c–h). Unexpectedly, GAB was the most abundant intracellular metabolite and among the top three extracellular metabolites in iTreg cells (Fig. 1c,d). However, neither prolonged culture nor restimulation would significantly changed GAB excretion (Extended Data Fig. 1h,i). Following activation, thymus-derived Treg (tTreg) cells could also excrete a comparable amount of GAB to the medium as iTreg cells (Fig. 1i). Notably, the intracellular level of GAB was even higher than that of glutamate (Glu) in iTreg cells, which is one of the most abundant intracellular metabolites in various organisms13. GAB is produced by catabolizing glutamine (Gln) through the (GABA) γ-aminobutyric acid shunt and elicits GABAergic response through GABA receptors (GABA-Rs) in neurons. To better understand the molecular nature that determines GAB production and function in T cells, we examined the expression of a panel of GABA-related metabolic and receptor genes by qPCR (Fig. 1e,f). Consistent with the previous findings on the GABA-R expression profile in immune cells14. Teff cells expressed a selected group of GABA-R subunits. However, only iTreg and TH17 cells expressed high levels of glutamate decarboxylase (GAD), the enzyme that catalyzes the decarboxylation of Glu to GAB. Unexpectedly, the TH17 group exhibits a higher level of GAD than the iTreg group and was the only group that expressed a high level of the GAB-catabolizing enzyme 4-aminobutyrate aminotransferase (ABAT), indicating increased GAB catabolism in TH17 cells but not in iTreg cells (Fig. 1f). Collectively, these findings suggest that extracellular metabolome profiling is a robust approach to revealing T cell metabolic characteristics in vitro. Using this approach, we have found that GAB is an abundant metabolite produced by T cells.

T cells use both Gln and Arg to produce GAB

Given the higher expression of GAD and ABAT in TH17 cells relative to iTreg cells, we reasoned that both iTreg cells and TH17 cells could produce GAB. However, the fate of GAB depends on ABAT, that is, GAB is diverted into the tricarboxylic acid (TCA) cycle in the presence of ABAT in TH17 cells instead of being exported into the extracellular compartment in the absence of ABAT as in iTreg cells. To test this idea, we cultured TH17 cells with or without the potent ABAT inhibitor vigabatrin (Vig)15,16, for 6 h and then measured the levels of a panel of metabolites. Inhibiting ABAT activity by Vig led to the accumulation of intracellular GAB and GAB release into the medium (Fig. 1k,j). Notably, inhibiting ABAT activity rendered GAB one of the most abundant metabolites in the medium and cell pellet (Fig. 1k,j). Moreover, we have validated that ABAT was expressed in TH17 cells but not in iTreg cells using immunoblot (IB) analysis and intracellular staining (Fig. 2a,b). Interestingly, inhibiting ABAT activity reduced Gln consumption without changing Glu levels significantly but increased GAB levels over 100-fold (Fig. 2c,d). The reciprocal changes in Gln consumption versus GAB production raise the possibility of a Gln-independent GAB production route in TH17 cells. Gln catabolism via the GABA shunt is the canonical GAB biosynthesis pathway17. Alternatively, GAB could be formed from putrescine (Put), a metabolite mainly derived from arginine (Arg) (Fig. 2d)18,19. Indeed, the metabolic genes involved in converting Arg into GAB were highly expressed in TH17 cells (Fig. 2e). To determine to what extent Gln and Arg contribute to GAB biosynthesis, we cultured TH17 cells with Vig in the presence or absence of Gln, Arg or both. Then, we collected spent medium to measure the levels of various metabolites. While removing Gln or Arg reduced GAB production, the removal of both completely blocked GAB production (Fig. 2f). Next, we supplied [13C6]Arg, [13C5]Gln, [13C6]glucose (Glc) or [13C4]Put as metabolic tracers in the culture medium and followed 13C incorporation into individual metabolites by GC–MS. The presence of the 13C4 isotopologue of GAB and the corresponding 13C4 or 3C5 isotopologues of upstream metabolites further confirmed that Gln and Arg are carbon donors of GAB (Fig. 2g). However, only the 13C2 isotopologue of GAB was detected in samples with [13C6]Glc, suggesting that Glc can support Glu (and GAB) synthesis through the TCA cycle (Fig. 2h). Finally, we showed that Put can be converted to GAB via a diamine oxidase (DAO)-dependent reaction as its inhibitor aminoguanidine (AG) completely blocked the production of [13C4]GAB from [13C4]Put (Fig. 2i). In addition to a general requirement of both amino acids for protein synthesis, we envisioned that Gln and Arg might also support TH17 function and survival through supporting GAB biosynthesis. To test this idea, we cultured TH17 cells in Gln/Arg-replete medium or suboptimal medium (with low levels of Gln/Arg) in the absence or presence of high levels of GAB. Supporting our hypothesis, reducing the amount of either amino acid led to defects in the maintenance of viability and interleukin (IL)-17+ populations. Notably, GAB supplementation could correct both defects (Fig. 2j,k). We, therefore, conclude that TH17 cells can use both Gln-derived and Arg-derived carbon to synthesize GAB and support cell viability and function.

Fig. 2: Gln and Arg are the main carbon sources for GAB biosynthesis in Teff cells.
figure 2

a,b, ABAT protein levels were determined by IB (ABAT molecular weight, 56 kDa; full-scan images are provided as source data) (a) and cytometry (b). c, As illustrated by the experimental scheme (left), the indicated metabolites in TH17 cells were quantified by GC–MS (right); data are representative of n = 3 biologically independent samples. d, Scheme of the GAB metabolic pathways and a pharmacological inhibitor Vig of ABAT. e, Schematic diagram of GAB biosynthesis from Arg (left), with the expression of relevant genes determined by qPCR detection (right). mRNA levels for Tnai cells were set to 1. The heatmap represents the relative mRNA expression level (see colour scale). Values and s.d. are provided as source data. f. As illustrated by the experimental scheme (left), the indicated metabolites in TH17 cells were quantified by YSI (Gln and Glu) or by bioassay kits (Arg and GAB) (n = 3 biologically independent samples). g, Diagram of conversion of [13C5]Glu and [13C6]Arg to downstream metabolites (top). Indicated metabolites in TH17 cells were quantified by GC–MS (n = 3 biologically independent samples) (bottom). Numbers along the x axis represent those of 13C atoms in the given metabolites. h, Diagram of the conversion of [13C6]Glc to downstream metabolites. Indicated metabolites in TH17 cells were quantified by GC–MS (n = 3 biologically independent samples). Black dot, 12C; red dot, 13C derived from [13C6]Glc. i, Diagram of the conversion of [13C4]Put to GAB. Indicated metabolites in TH17 cells were quantified by GC–MS (n = 3 biologically independent samples). Unpaired two-tailed Student’s t-test. j, As illustrated by the experimental scheme (top), cytokine expression and cell viability were determined by flow cytometry (bottom). All experiments with bicuculline (5 μM); complete medium, 2 mM Gln + 0.1 mM Arg; Gln (low), 10 μM Gln + 0.1 mM Arg; Gln (low plus GABA), 10 μM Gln + 0.1 mM Arg + 1 mM GABA; Arg (low), 2 mM Gln + 10 μM Arg; Arg (low plus GABA) 2 mM Gln + 10 μM Arg + 1 mM GABA. k, Statistical analysis for j; data are representative of n = 3 biologically independent samples. Significance was calculated by one-way ANOVA with Tukey’s multiple-comparisons test. Gln, glutamine; Glu, glutamate; Arg, arginine; ODC, ornithine decarboxylase; DAO, diamine oxidase; PDH, pyrroline dehydrogenase; M, million.

Source data

Full size image

ABAT confers GAB-dependent anaplerosis on TH17 cells

Next, we reasoned that the expression of ABAT may render TH17 cells capable of diverting GAB into the TCA cycle in a way that maximizes carbon allocation and oxidative phosphorylation (OXPHOS) in mitochondria. To test this idea, we added [13C4]GABA as a metabolic tracer into the culture medium and followed 13C incorporation into intermediate metabolites of the TCA cycle in iTreg cells and TH17 cells with or without Vig treatment (Fig. 3a,b). In line with the expression of ABAT in TH17 cells but not in iTreg cells, TH17 cells exhibited much higher levels of the 13C4 isotopologue of succinate and its downstream metabolites in the TCA cycle than iTreg cells (Fig. 3a). Inhibiting ABAT activity by Vig completely abolished the 13C4 isotopologue of succinate and its downstream metabolites in TH17 cells, supporting the idea that GAB is diverted to the TCA cycle via an ABAT-dependent reaction (Fig. 3b). Next, we sought to determine the temporal change in respiration following a sequential supplementation of GABA, Vig, oligomycin or carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) into the TH17 cell culture medium. Indeed, GABA supplementation enhanced oxygen consumption in an ABAT-dependent manner, while ATPase inhibitor oligomycin suppressed and FCCP maximized oxygen consumption as expected (Fig. 3c). We and others have recently shown that Arg-dependent polyamine biosynthesis is required to support T cell proliferation and TH17 cell differentiation10,11,12. We reasoned that ABAT expression in TH17 cells might allow Arg-derived carbons to be diverted into the TCA cycle through Put and GAB. Supporting this idea, [13C6]Arg-derived and [13C4]Put-derived 13C were incorporated into the 13C4 isotopologue of succinate and its downstream metabolites in an ABAT-dependent manner in TH17 cells (Fig. 3d, e). Finally, we sought to determine whether Gln-derived carbons could enter the TCA cycle via ABAT. Gln is a major carbon donor known to drive the TCA cycle and OXPHOS via glutamine transaminase and glutamate dehydrogenase (GDH) in Teff cells20,21,22. We found that a sequential supplementation with Vig and the GDH inhibitor R162 (ref. 23) suppressed oxygen consumption additively (Fig. 3f). Similarly, combining Vig and R162 suppressed [13C5]Gln-derived 13C incorporation into the TCA cycle metabolites more profoundly than single-agent treatment (Fig. 3g). Collectively, we have identified GAB as a conditional anaplerotic substrate in T cells, and its catabolism via the TCA cycle depends on the expression of ABAT.

Fig. 3: ABAT enables diversion of GAB into the TCA cycle in T cells.
figure 3

a,b,c,d,e,g,Diagrams of the conversion of [13C4]GABA (a,b, left), [13C6]Arg (d, left), [13C4]Put (e, left) and [13C5]Gln (g, left) to downstream metabolites. Indicated metabolites in TH17 cells were quantified by GC–MS (n = 3 biologically independent samples) (right). Black dot, 12C; blue dot, 13C derived from the indicated tracers. Numbers along the x axis represent those of 13C atoms in the given metabolites. Significance was calculated by unpaired two-tailed Student’s t-test (a, b, d, e and g); NS, no significant differences. c,f, OCR of TH17 cells with the indicated treatments was determined by Seahorse. Data are representative of n = 16 independent biological samples from n = 3 independent biological experiments. Two-way ANOVA with Sidak’s multiple-comparisons test (c) and (f). α-KG, α-ketoglutarate; GDH, glutamate dehydrogenase.

Full size image

GAB metabolism controls proliferation and differentiation

To further delineate the role of ABAT in T cells, we generated a T cell-specific Abat-knockout strain (Abat cKO) by crossing the Abatfl strain with the Cd4-Cre strain. qPCR, IB and intracellular staining analyses validated the deletion of ABAT (Fig. 4a, b). ABAT deletion did not result in T cell development defects in the thymus, the spleen or lymph nodes (Extended Data Fig. 2a–f). In addition, ABAT deletion did not affect cell viability, the expression of cell surface activation markers, the cell cycle progression from G0/G1 to the S phase, RNA, DNA or protein contents, cell size, or viability 24 h after activation in vitro (Extended Data Fig. 3a–d). However, ABAT deletion moderately suppressed overall T cell proliferation after activation in vitro (Fig. 4c and Extended Data Fig. 3e). Remarkably, both genetic and pharmacological ablation of ABAT activity inhibited pro-inflammatory TH17 cell differentiation while enhancing anti-inflammatory iTreg cell differentiation in vitro (Fig. 4c). Supporting these findings, the RNA-seq analysis of wild-type (WT) and Abat cKO T cells activated under the TH0 condition revealed enriched gene signatures associated with inflammation and T cell differentiation (Fig. 4d–f). Notably, the ABAT inhibitor (Vig) did not potentiate the effect of genetic deletion in suppressing TH17 cell differentiation, suggesting that Vig is a specific inhibitor of ABAT (Extended Data Fig. 4a). Moreover, overexpressing ABAT (ABAT-OE) suppressed iTreg differentiation and could synergize with IL-6 to increase the percentage of IL-17+ cells under the iTreg-polarizing condition in vitro (Extended Data Fig. 4b). Next, we examined the effect of ABAT inhibition on TH1 and TH2 differentiation. Genetic and pharmacological ablation of ABAT activity inhibited TH1 cell differentiation without significantly changing TH2 cell differentiation significantly in vitro (Extended Data Fig. 4c, d). Finally, we asked whether TH1 cells could divert GAB into the TCA cycle as TH17 cells did. We applied [13C4]GABA as a metabolic tracer and followed 13C incorporation into intermediate metabolites of the TCA cycle in TH1 and TH17 cells. Only TH17 cells exhibited high levels of the 13C4 isotopologue of succinate and its downstream metabolites in the TCA cycle (Extended Data Fig. 4e). Notably, genetic and pharmacological ablation of ABAT activity completely abolished the 13C4 isotopologues of metabolites in TH17 cells (Extended Data Fig. 4e).

Fig. 4: Genetic ablation or pharmacological inhibition of ABAT suppresses TH17 but enhances iTreg cell differentiation.
figure 4

a,b, ABAT mRNA and protein levels were determined by qPCR and IB analysis (a, full-scan images are provided in as source data) or by flow cytometry (b); unpaired two-tailed Student’s t-test for a (n = 3 biologically independent samples). c, Expression of the indicated cytokines and CFSE dilution in the indicated groups determined by flow cytometry (WT, n = 5 independent biological samples from n = 5 independent biological experiments; Abat cKO, n = 5 independent biological samples from n = 3 independent biological experiments; WT plus Vig, n = 3 independent biological samples from n = 3 independent biological experiments). Two-way ANOVA with Sidak’s multiple-comparisons test. d–f, Hierarchical clustering analysis, gene set enrichment analysis (GSEA) and Ingenuity Pathway Analysis (IPA) of a list of inflammatory genes performed by using RNA-seq data of CD4+ T cells that were cultured under TH0 culture conditions and collected at 36 h after activation (n = 3 independent biological samples). In d, heatmaps represent the log-transformed value of the quantity (see colour scale). In f, the orange dotted line along the x axis indicates the cut-off value (P = 0.05). The complete data are provided as source data. g, As illustrated by the experimental scheme (top), the expression of the indicated proteins in CNS-infiltrating CD4+ T cells isolated from EAE animals was determined by flow cytometry (n = 3 independent biological samples from n = 2 independent biological experiments). h,j, EAE clinical scores in the indicated groups evaluated daily from mice of the indicated genotypes (n = 6 independent biological samples from n = 2 independent biological experiments in h and n = 5 independent biological samples from n = 3 independent biological experiments in j); significance was calculated by unpaired two-tailed Student’s t-test. i,k, The expression of the indicated markers in CNS-infiltrating T cells determined by flow cytometry (n = 5 independent biological samples in i and n = 3 independent biological samples in k). PTX, paclitaxel; NES, normalized enrichment score; FDR, false discovery rate.

Source data

Full size image

The expansion and balance between pro-inflammatory CD4+ Teff cells and anti-inflammatory CD4+ Treg cells determine the pathogenic development of experimental autoimmune encephalomyelitis (EAE), a mouse model of multiple sclerosis (MS), which is an inflammatory demyelinating disease of the central nervous system (CNS). Consistent with the expression profile of ABAT in vitro, the IL-17+CD4+ T cell group expressed the highest level of ABAT among all the CD4+ T subsets with infiltration into the CNS in animals with EAE (Fig. 4g). Notably, the genetic deletion of Abat in T cells or the systemic delivery of Vig conferred significant protection against EAE pathogenic progression, associated with more infiltrated FoxP3+CD4+ T cells and fewer infiltrated inflammatory CD4+ T cells, reciprocally (Fig. 4h,i and Extended Data Fig. 5a,b). However, Vig treatment resulted in better protection against EAE and a broader impact on periphery CD4+ T cells in the periphery than T cell-specific deletion of Abat, indicating that the systemic inhibition of ABAT might affect inflammation through both T cell-intrinsic and T cell-extrinsic mechanisms (Fig. 4j,k, and Extended Data Fig. 5c,d). We also used a competitive antigen-specific, T cell receptor (TCR)-dependent proliferation assay (OT-II) and a competitive homeostatic proliferation assay to assess T cell proliferation and differentiation in vivo. Notably, the ratio between WT and Abat cKO CD4+ T cells, CFSE dilution patterns and the percentage of IL-17+CD4+ or interferon-γ (IFNγ+)CD4+ T cells in various tissues suggested that the loss of ABAT dampens T cell proliferation and TH1 and TH17 differentiation in vivo (Extended Data Fig. 6a–e). Collectively, our results indicate that ABAT status determines the fate of intracellular GAB and, hence, pro-inflammatory TH17 and anti-inflammatory iTreg cell differentiation in vitro and in vivo.

GAB regulates T cell differentiation through the GABAA receptor

In line with earlier studies14, we have found that Teff cells express various subunits of the GABAA receptor (GABAA-R) (Fig. 1f). Additionally, T cells can produce and secrete a large amount of GAB into the extracellular compartment, which may elicit a context-dependent autocrine signalling response to regulate T cell differentiation (Fig. 5a). Supporting this idea, a low level of GAB supplementation could reduce TH17 but enhance iTreg differentiation without significantly affecting T cell activation and proliferation in vitro (Fig. 5b,c and Extended Data Fig. 7a,b). Conversely, GABAA-R antagonists with distinct antagonistic mechanisms enhanced TH17 cell differentiation but reduced iTreg cell differentiation without affecting T cell activation and proliferation in vitro (Fig. 5c–e and Extended Data Fig. 7c,d). The β-subunit is a core component of GABAA-R, and the β3 subunit (encoded by Gabrb3) was highly expressed in all Teff subsets (Fig. 1f). We generated a T cell-specific Gabrb3-knockout strain (Gabrb3 cKO) by crossing the Gabrb3fl strain with the CD4-Cre strain. Gabrb3 deletion did not result in T cell development defects in the thymus, spleen and lymph nodes (Extended Data Fig. 8a–f). In addition, cell viability, the expression of cell surface activation markers and cell proliferation were comparable in both WT and Gabrb3 cKO T cells after activation in vitro (Extended Data Fig. 9a,b). However, genetic ablation of Gabrb3 promoted pro-inflammatory TH17 cell differentiation while reducing anti-inflammatory iTreg cell differentiation in vitro (Fig. 5c,f). Notably, GABA supplementation only affected WT but not Gabrb3 cKO T cell differentiation in vitro (Fig. 5c,f). Finally, the T cell-specific Gabrb3 deletion let to significantly deteriorated EAE pathogenic progression, associated with increased inflammatory CD4+ T cells and decreased FoxP3+CD4+ T cells in the CNS and periphery (Fig. 5g,h and Extended Data Fig. 9c,d).

Fig. 5: Receptor-mediated GAB autocrine signalling response reciprocally suppresses TH17 and enhances iTreg differentiation.
figure 5

a,d, Schematic diagram of GAB metabolism and the GABAA-R-mediated autocrine signalling response in T cells (a) and schematic diagram of the binding sites for GABAA-R antagonists (d). b,e,f, Cytokine expression of the indicated groups determined by flow cytometry. c, Combination statistical analysis of b, e and f; data are shown as mean ± s.e.m. (n = 4 independent biological samples). Significance was calculated by two-way ANOVA with Sidak’s multiple-comparisons test. g, EAE clinical scores in the indicated groups evaluated daily (n = 5 independent biological samples from n = 3 independent biological experiments); significance was calculated by unpaired two-tailed Student’s t-test. h, Expression of the indicated markers in CNS-infiltrating T cells were determined by flow cytometry (n = 3 independent biological samples). Bicl, bicuculline; PicroT, picrotoxin; Flu, flumazenil; PTX, paclitaxel.

Full size image

GAB regulates T cells through a bimodal mechanism of action

Next, we sought to dissect the effect of ABAT-dependent mitochondrial anaplerosis and GABAA-R-mediated signalling on T cell differentiation and function (Extended Data Fig. 10a). We envisioned that the ABAT-dependent anaplerotic reaction might support TH17 differentiation by providing succinate to fuel mitochondrial OXPHOS (Fig. 6a). Indeed, inhibiting ABAT activity by Vig suppressed oxygen consumption, which was reversed by adding a cell-permeable succinate analogue NV118 (Fig. 6a,b)24. In line with the effect of NV118 on oxygen consumption, the NV118 supplementation could partially reverse the inhibition of TH17 differentiation resulting from genetic or pharmacological inhibition of ABAT (Fig. 6c,d). Next, we asked whether ABAT-dependent mitochondrial anaplerosis could impact transcription factors critical for TH17 lineage differentiation, such as RORγt and STATs25. To test this idea, we reduced the medium’s Gln concentration and added a high concentration of GAB (1 mM) with a GABAA-R antagonist. We reasoned that reducing Gln levels would force cells to use GAB as a mitochondrial fuel and adding the receptor antagonist would eliminate the effects of receptor signaling. Indeed, GAB supplementation significantly enhanced the levels of RORγt and phosphorylated STAT3 (pSTAT3) but reduced the levels of phosphorylated STAT5 (pSTAT5) (Extended Data Fig. 10b).

Fig. 6: GAB exerts both bioenergetic and receptor signalling-mediated control of T cell differentiation.
figure 6

a, Schematic diagram of GAB catabolism through the TCA cycle. b, OCR of TH17 cells with the indicated treatments determined by Seahorse (n = 16 independent biological samples from n = 3 independent biological experiments); significance was calculated by two-way ANOVA with Sidak’s multiple-comparisons test. c,d, Expression of IL-17A in the indicated groups determined by flow cytometry (n = 3 independent biological samples); significance was determined by one-way ANOVA with Tukey’s multiple-comparisons test. e, Left, experimental diagram of the competitive Treg suppression assay. Right, CFSE dilution was determined by flow cytometry (n = 3 independent biological samples); significance was determined by two-way ANOVA with Sidak’s multiple-comparisons test. f, Schematic conceptual model in which GAB generated by Gln and Arg can regulate T cell differentiation by entering the TCA cycle or being exporting into the extracellular environment and acting on GABAA-R. Tconv, conventional T cells.

Full size image

Next, we sought to determine whether modulating GABAA-R affects key signalling molecules involved in regulating TH17 and iTreg differentiation. We treated T cells with a low dose of GAB (10 μM) in the presence of a GABAA-R antagonist. We reasoned that the low dose of GAB could engage the receptor-mediated signalling response without significantly fuelling mitochondrial metabolism. We assessed the levels of phosphorylated STAT proteins and the phosphorylation of a canonical mTORC1 substrate (pS6) because mTORC1 is critical for determining TH17 and iTreg differentiation26,27. Treating TH17 and iTreg cells with a low dose of GAB suppressed pSTAT3 and mTORC1 substrate phosphorylation (pS6) but increased pSTAT5 (Extended Data Fig. 10c). Notably, the effects of GAB on these signalling molecules could be reversed by a GABAA-R antagonist (Extended Data Fig. 10c). We showed that iTreg cells can excrete GAB into the extracellular compartment (Extended Data Fig. 1g–i). Finally, we sought to determine whether GAB contributes to Treg-dependent immune suppression. We performed a competitive Treg suppression assay by co-culturing iTreg cells with WT and Gabrb3 cKO CD4+ T cells that carried different isogenic markers (Fig. 6e). Indeed, Gabrb3 cKO CD4+ T cells proliferated better than the WT group, indicating that genetic ablation of GABAA-R could partially alleviate iTreg-mediated suppression (Fig. 6e). Together, these results suggest that GAB is an abundant metabolite produced by T cells and exerts both bioenergetic control and receptor-mediated signalling control of T cell differentiation (Fig. 6f).

Discussion

The vertebrate immune and nervous systems are intimately connected with each other developmentally, anatomically and physiologically. Interaction between the two systems coordinates their sensory functions to ensure organismal homeostasis and survival28,29,30,31. Immune cells and neurons can communicate with each other through a group of shared ligand molecules and receptors, including the neurotransmitter GABA and its receptors14,32. Beyond mediating intersystem communication between the immune and nervous systems, growing evidence suggests that GABA can also act as a paracrine signalling molecule mediating intrasystem communication to regulate immune response33. One recent study has found that B cells can produce GABA and suppress anti-tumour immunity through paracrine modulation of intratumoural macrophages and CD8+ T cells34. Additionally, GABA in macrophages has been implicated as an intracellular metabolite with a pro-inflammatory function14. Here, we show that GAB (the biochemical form of GABA at physiological pH) is one of the most abundant metabolites in T cells and promotes inflammation through modulating T cell proliferation and differentiation. Depending on the status of its catabolizing enzyme ABAT, GAB can act as a conditional anaplerotic substrate to promote TH17 cell differentiation or an autocrine signalling metabolite to enhance iTreg cell differentiation. In addition to its role in mediating intercellular communications, GAB also serves as a metabolic and signalling gatekeeper to regulate inflammation in a T cell-autonomous manner.

Teff cells consume Gln and Arg at high rates35,36. Beyond a general requirement for protein synthesis, Gln and Arg support T cell proliferation and function through their catabolic products. Gln is a primary carbon source to sustain the TCA cycle, which generates energy through OXPHOS and allocates carbon to produce biosynthetic precursors to support T cell growth21,36,37. Similarly, Arg catabolism is coupled with the urea cycle to produce bioactive metabolites such as polyamines to support T cell proliferation and differentiation10,11,12. Our results show that both Arg catabolism (via Put) and Gln catabolism (via Glu) are coupled with GAB biosynthesis in TH17 cells, implicating GAB as a crucial metabolic node and a branch point in amino acid catabolism. GAB can be consumed through the TCA cycle to enhance bioenergetic and biosynthetic capacities or be secreted as an autocrine signalling metabolite depending on the status of ABAT. We have further revealed that Gln can replenish the TCA cycle intermediate metabolites through either Glu or GAB anaplerosis. Glu increases the levels of α-ketoglutarate (α-KG), while GAB increases the levels of succinate. Therefore, it is conceivable that the carbon input from Glu or GAB may change the intracellular α-KG to succinate ratio reciprocally. Hence, the GABA shunt in T cells may impact the hypoxia signalling response and/or DNA/histone methylation patterns by modulating the enzymatic activities of the α-KG-dependent dioxygenase family37,38,39,40, and Glu and Put are highly abundant intracellular metabolites that can be secreted to the extracellular environment by TH17 cells (Fig. 1a)10,41. The GAB-catabolizing enzyme ABAT may provide a sensitive and precise regulation of the three interconnected and highly abundant metabolites: GAB, Glu and Put, permitting rapid metabolic and signalling responses to control inflammation.

The high and dynamic metabolic demands of T cells during inflammatory and autoimmune responses require fine-tuned regulation of central carbon and ancillary metabolic pathways. Hence, metabolic pathways have been therapeutically exploited to target inflammatory and autoimmune diseases42,43. Disruption of central carbon catabolism can affect many cellular processes and cell types. However, targeting ancillary metabolic pathways engaged in a small group of specialized immune cells under physio-pathological conditions may result in less toxicity but maximal clinical benefits6. Gene and protein expression profiling studies have suggested that human autoimmune diseases, including MS, type 1 diabetes and rheumatoid arthritis, are associated with the dysregulation of GABA-related metabolic and signaling genes44,45,46,47. Interestingly, cortical GAB levels are lower in patients with relapsing–remitting multiple sclerosis MS than in healthy controls48,49. In addition, one recent study based on genome-scale metabolic modelling and in silico simulations for drug response indicated that GAB metabolism and signalling pathway not only are involved in the disease process but also are potential drug targets in human autoimmune diseases50. Consistent with clinical profiling and in silico studies, pharmacological modulation of GAB metabolism and receptor-mediated signalling response could ameliorate pathological phenotypes in several preclinical models of autoimmune diseases51,52,53,54,55. Our results further elucidate a previously unrecognized aspect of the T cell-intrinsic effects conferred by GAB catabolism and receptor-mediated signalling. Collectively, GAB-modulating strategies via blockade of GAB catabolism, activation of receptor-mediated response, or both may present a promising therapy for treating inflammatory and autoimmune diseases.

Methods

Mice

C57BL/6 (WT), Flippase (B6.129S4Gt(ROSA)26Sortm1(FLP1)Dym/RainJ), OT-II (B6.Cg-Tg(TcraTcrb)425Cbn/J), CD45.1+ (B6.SJL-PtprcaPepcb/BoyJ), Rag1−/− (B6.129S7-Rag1tm1Mom/J), IL17A-IRES-GFP-KI (C57BL/6-Il17atm1Bcgen/J), FoxP3GFP+ (C57BL/6-Tg(Foxp3-GFP)90Pkraj/J) and Gabrb3fl (B6;129-Gabrb3tm2.1Geh/J) mice were obtained from the Jackson Laboratory (JAX, Bar Harbor, ME). Mice with one targeted allele of Abat on the C57BL/6 background (Abattm1a(EUCOMM)Hmgu) were generated by the European Conditional Mouse Mutagenesis Program (EUCOMM)56. The mice were first crossed with a transgenic Flippase strain (B6.129S4Gt(ROSA)26Sortm1(FLP1)Dym/RainJ) to remove the lacZ-reporter allele and then crossed with the Cd4-Cre strain to generate the T cell-specific Abat knockout strain (Abat cKO). OT-II mice were crossed with Cd4-Cre Abat cKO mice to generate the OT-II Cd4-Cre Abat cKO mice. OT-II mice were crossed with Thy1.1+ mice (B6.PL-Thy1a/CyJ) to generate the OT-II Thy1.1 mice. Gabrb3fl mice were crossed with the Cd4-Cre strain to generate T cell-specific Gabrb3-knockout strain (Gabrb3 cKO). For one independent experiment, we used male and female mice from the same strain that were both sex and age matched (6–12 weeks old), such as two males and two females for WT mice, as well as for KO mice. All mice were bred and kept in specific pathogen-free conditions at the Animal Center of the Abigail Wexner Research Institute at Nationwide Children’s Hospital. A low-fat diet was provided (Envigo 2920, the irradiated form of 2020X; https://insights.envigo.com/hubfs/resources/data-sheets/2020x-datasheet-0915.pdf). Animals were killed by carbon dioxide asphyxiation followed by cervical dislocation under protocols approved by the Institutional Animal Care and Use Committee of the Abigail Wexner Research Institute at Nationwide Children’s Hospital (IACUC; protocol number AR13-00055).

Murine T cell isolation and culture

Naive CD4+ T cells were enriched from mouse spleen and lymph nodes by negative selection using the MojoSort™ Mouse CD4+ Naive T Cell Isolation Kit (MojoSort, BioLegend) according to the manufacturer’s instructions. For the activation assay, freshly isolated CD4+ T cells were either maintained in a culture medium with 5 ng/ml-1 IL-7 or activated with 5 ng/ml-1 IL-2 and plate-bound anti-mouse CD3 and anti-mouse CD28. The culture plates were precoated with 2 μg/ml-1 anti-mouse CD3 and 2 μg/ml-1 anti-mouse CD28 antibodies overnight at 4 °C. Naive tTreg cells were enriched from mouse spleen and lymph nodes by positive selection using the MojoSort™ Mouse CD4+CD25+ Regulatory T Cell Isolation Kit (MojoSort, BioLegend) according to the manufacturer’s instructions. For the activation assay, freshly isolated CD4+CD25+ regulatory T cells were either maintained in a culture medium with 5 ng/ml-1 IL-2 or activated with 5 ng/ml-1 IL-2 and anti-mouse CD3/CD28 beads according to the manufacturer’s instructions (Gibco, Thermo Fisher Scientific). Unless indicated separately, the cells were seeded in the RPMI-1640 medium (Corning) supplemented with 10% FBS, or heat-inactivated dialysed FBS (DFBS), 2 mM l-glutamine, 1% sodium pyruvate (Sigma-Aldrich), 100 units/ml-1 penicillin, 100 μg/ml-1 streptomycin and 0.05 mM 2-mercaptoethanol (Sigma-Aldrich) at 37 °C and 5% CO2.

For CD4+ T cell differentiation, 48-well culture plates were precoated with 2 μg/ml-1 (iTreg differentiation), 5 μg/ml-1 (TH1/ TH2 differentiation) or 10 μg/ml-1 (TH17 differentiation) anti-mouse CD3 and anti-mouse CD28 antibodies overnight at 4 °C. Freshly isolated naive CD4+ T cells (0.5 × 106 cells per ml) were activated with plate-bound antibodies and with mouse IL-2 (3 ng/ml-1) and human TGF-β1 (10 ng/ml-1) for iTreg differentiation, with mouse IL-2 (10 ng/ml-1) and mouse IL-12 (20 ng/ml-1) for TH1 differentiation, with mouse IL-2 (2 ng/ml-1), mouse IL-4 (50 ng/ml-1) and anti-mouse IFN-γ (10 μg/ml-1) for TH2 differentiation, or with mouse IL-6 (50 ng/ml-1), human TGF-β1 (20 ng/ml-1), anti-mouse IL-2 (8 μg/ml-1), anti-mouse IL-4 (8 μg/ml-1), and anti-mouse IFN-γ (8 μg/ml-1) for TH17 differentiation. In some experiments, Vig (1 mM), GABA (0.1 μM~1 mM), NV118 (25 μM), GABAA-R antagonists including bicuculline (Bicl, 5 or 50 μM), picrotoxin (PicroT, 5 or 50 μM) and flumazenil (10 or 1 μM), R162 (20 μM), oligomycin (1.5 μM), FCCP (1 μM), or AG (0.2 mM) was added to cell culture medium. Additional information on cytokines, antibodies and chemicals is listed in Supplementary Table 1.

Flow cytometry

For analysing surface markers, cells were stained in phosphate-buffered saline (PBS) containing 2% (wt/vol) BSA and the appropriate antibodies from BioLegend. For analysing the intracellular cytokines IFN-γ and IL-17A, T cells were stimulated for 4 hrs with eBioscience™ Cell Stimulation Cocktail (eBioscience) before being stained with cell surface antibodies. Cells were then fixed and permeabilized using FoxP3 Fixation/Permeabilization solution according to the manufacturer’s instructions (eBioscience). Cell proliferation was assessed using CFSE staining according to the manufacturer’s instructions (Invitrogen). Cell viability was evaluated by 7AAD staining according to the manufacturer’s instructions (BioLegend). For analysing DNA/RNA content, cells were collected and stained for surface markers before being fixed with 4% paraformaldehyde for 30 min at 4 °C, followed by a permeabilization step with FoxP3 permeabilization solution (eBioscience). Cells were stained with 7AAD for 5 min and then stained with pyronin-Y (4 μg/ml-1; PE) for 30 min before being analysed using flow cytometer with the PerCP channel for 7AAD (DNA) and PE channel for pyronin-Y (RNA). A protein synthesis assay kit (Item No.601100, Cayman) was used for analysing protein content. Briefly, cells were incubated with O-propargyl-puromycin (OPP) for 1 hr and then they were fixed and stained with 5 FAM-azide staining solutions before being analysed using a flow cytometer with the FITC channel. For analysing the cell cycle profile, cells were incubated with 10 μg/ml-1 BrdU for 1 hr, followed by cell surface staining, fixation and permeabilization based on the Phase-Flow Alexa Fluor 647 BrdU Kit (BioLegend). Flow cytometry data were acquired on Novocyte (ACEA Biosciences) and were analysed with FlowJo software (TreeStar). Additional information on flow cytometry antibodies is listed in Supplementary Table 2.

Treg cell suppression assay

For the iTreg suppression assay, naive CD4+ T cells isolated from CD45.1 mice using the naive CD4+ mouse T cell isolation kit (BioLegend) were differentiated for 3 d to generate iTreg cells. Naive CD4+ T cells isolated from CD45.2/Thy1.1 WT donor mice and CD45.2/Thy1.2 Gabrb3 cKO donor mice were mixed at a 1:1 ratio (as Tconv cells) and labelled with CFSE. Then, approximately 5 × 104 Tconv cells were mixed with iTreg cells (with indicated ratios) and cultured with 3 ng/ml-1 IL-2 and anti-mouse CD3/CD28 beads. Cells were collected 4 d later and processed to assess proliferation by flow cytometry analysis.

Retrovirus production and transduction

Phoenix Eco cells that were cultured in fresh DMEM media (Corning) supplemented with 10% heat-inactivated FBS and 0.5% penicillin-streptomycin were transfected with the control plasmid (pMIC, MSCV-IRES-mCherry) or pMIC-ABAT (Supplementary Table 3). Viral-Boost Reagent (ALSTEM) was added to the culture medium at a 1:600 dilution 6 h after transfection. Cell medium was collected at 48 h after transfection, centrifuged at 300g for 10 min, and then filtered through a 0.45-μm filter unit (GVS Filter Technology). Retrovirus Precipitation Solution (ALSTEM) was added to retrovirus-containing supernatant at 1:4 dilution and incubated overnight at 4 °C, followed by centrifugation at 1,500g for 30 min at 4 °C to concentrate the virus. Then, approximately 0.3 × 106 activated CD4+ T cells (1 d after activation) were resuspended in 1 ml of retroviral supernatant containing 8 µl/ml-1 Lipofectamine (Invitrogen) and cultured under iTreg differentiation for 4 d.

Adoptive cell transfer assays

For homeostatic proliferation in lymphopenic Rag−/− mice, naive CD4+ T cells isolated from donor mice using a naive CD4+ mouse T cell isolation kit (BioLegend) were labelled with CFSE. Approximately 1 × 107 cells (mix of WT and KO cells at a 1:1 ratio) in 150 μl of PBS were transferred via caudal venous injection into 6- to 8-week-old sex-matched host mice. Mice were killed between 4–7 d after cell transfer. Lymph nodes and spleen were collected and processed to assess cell ratio and proliferation by flow cytometry analysis.

For antigen-driven proliferation using OT-II mice, naive CD4+ T cells isolated from OT-II/CD45.2 TCR-transgenic donor mice using the naive CD4+ mouse T cell isolation kit (BioLegend) were labelled with CFSE. Approximately 1 × 107 cells (mix of WT and KO cells at a 1:1 ratio) in 150 μl of PBS were transferred via caudal venous injection into 6- to 8-week-old sex-matched CD45.1 host mice. Host mice were immunized subcutaneously in the hock area (50 μl each site) in both legs with 1 mg/ml-1 ovalbumin (OVA)323–339 peptide (InvivoGen) emulsified with complete Freund adjuvant (CFA; InvivoGen). The mice were then killed 8 d after immunization. Lymph nodes were collected and processed to assess cell ratio, proliferation and protein expression by flow cytometry analysis.

EAE

Mice were immunized subcutaneously with 100 μg of myelin oligodendrocyte glycoprotein (MOG)35–55 peptide emulsified in CFA, which was made from IFA (Difco) plus Mycobacterium tuberculosis (Difco). Mice were injected intraperitoneally with 200 ng of pertussis toxin (PTX, List Biological Laboratories) on the day of immunization and 2 d later. In the experiments shown in Fig. 4j,k and Extended Data Fig. 5c,d, the mice were injected intraperitoneally with 250 mg/kg-1 of Vig in 100 μl PBS daily from day 3 after immunization throughout the end of the experiment. In the experiments shown in Fig. 5g,h and Extended Data Fig. 9c,d, the animals were injected with PTX only once on the day of immunization for a suboptimal EAE induction. All mice were observed daily for clinical signs and scored as described previously10. In some experiments, the mice were killed when the control mice reached the onset of symptoms. The CNS (brain and spinal cord), spleen and peripheral lymph nodes were collected and mashed to generate the single-cell suspension. The cell suspension was centrifuged on a 30%/70% Percoll gradient at 500g for 30 min to isolate mononuclear cells from the CNS, followed by cell surface and intracellular staining and flow cytometry analysis described above.

Stable isotope labelling experiments

[13C5]Gln, [13C6]Arg and [13C6]Glc labelling of TH17 cells

Naive CD4+ T cells isolated from WT mice were polarized for 72 h under TH17 culture conditions before being collected and reseeded at 2 × 106 cells per ml in a conditional medium (RPMI-1640) containing 4 mM [13C5]Gln, 1 mM [13C6]Arg or 10 mM [13C6]Glc. After 12 h of culture, around 1 × 107 cells for each sample were collected and washed three times with PBS before being snap-frozen.

[13C6]Arg labelling of TH17 cells

TH17 cells (as described above) were pretreated with vehicle or Vig (1 mM) for 1 h before being collected and reseeded at a density of 2 × 106 cells per ml in a conditional medium (RPMI-1640) containing 4 mM 1 mM [13C6]Arg with vehicle or Vig (1 mM). After 6 h of culture, around 1 × 107 cells for each sample were collected and washed three times with PBS before being snap-frozen.

[13C4]Put labelling of TH17 cells

TH17 cells (as described above) were pretreated with vehicle, Vig (1 mM), or AG (0.2 mM) for 1 h and then collected and reseeded at a density of 2 × 106 cells per ml in a conditional medium (RPMI-1640) containing 0.1 mM [13C4]Put and 10 µM Arg and with vehicle, Vig (1 mM), or AG (0.2 mM) treatment. After 6 h of culture, around 1 × 107 cells for each sample were collected and washed three times with PBS before being snap-frozen.

[13C4]GABA labelling of TH17, iTreg and TH1 cells

Naive CD4+ T cells isolated from WT mice were polarized for 72 h under TH17, iTreg or TH1 culture conditions before being collected and re-seeded at a density of 2 × 106 cells per ml in a conditional medium (RPMI-1640) containing 0.5 mM [13C4]GABA, 0.1 mM Gln and the GABAA-R antagonist bicuculline (5 µM). After 12 h of culture, around 1 × 107 cells for each sample were collected and washed three times with PBS before being snap-frozen.

[13C4]GABA labelling of TH17 cells with Vig

TH17 cells (as described above) generated from WT or Abat cKO mice were pretreated with vehicle or Vig (1 mM) for 1 h before being collected and reseeded at a density of 2 × 106 cells per ml in the conditional medium containing 0.5 mM [13C4]GABA, 0.1 mM Gln and the GABAA-R antagonist bicuculline (5 µM) and with vehicle or Vig (1 mM) treatment. After 12 h of culture, around 1 × 107 cells for each sample were collected and washed three times with PBS before being snap-frozen.

[13C5]Gln labelling of TH17 cells with multiple inhibitors

TH17 cells (as described above) were pretreated with vehicle, Vig (1 mM) or R162 (20 μM) for 1 h and then collected and reseeded at a density of 2 × 106 cells per ml in a conditional medium (RPMI-1640) containing 4 mM [13C5]Gln and with vehicle, Vig (1 mM), R162 (20 μM) or the combination of Vig and R162 treatment. After 6 h of culture, around 1 × 107 cells for each sample were collected and washed three times with PBS before being snap-frozen. Additional information on stable isotope labelling is listed in Supplementary Table 4.

Gas chromatography–mass spectrometry sample preparation and analysis

GC–MS was performed as previously described57, and cell pellets were resuspended in 0.45 ml of −20 °C methanol/water (1:1 v/v) containing 20 µM l-norvaline as internal standard. Further extraction was performed by adding 0.225 ml of chloroform followed by vortexing and centrifugation at 15,000g for 5 min at 4 °C. The upper aqueous phase was evaporated under vacuum using a Speedvac centrifugal evaporator. Separate tubes containing varying amounts of standards were evaporated. Dried samples and standards were dissolved in 30 μl of 20 mg/ml-1 isobutylhydroxylamine hydrochloride (TCI #I0387) in pyridine and incubated for 20 min at 80 °C. An equal volume of N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA) (Soltec Ventures) was added and incubated for 60 min at 80 °C. After derivatization, samples and standards were analysed by GC–MS using an Rxi-5ms column (15 m × 0.25 internal diameter × 0.25 μm, Restek) installed in a Shimadzu QP-2010 Plus GC–MS instrument. GC–MS was programmed with an injection temperature of 250 °C, injection volume of 1.0 µl and a split ratio of 1/10. The GC oven temperature was initially 130 °C for 4 min, rising to 250 °C at 6 °C min-1 and to 280 °C at 60 °C min-1 with a final hold at this temperature for 2 min. GC flow rate, with helium as the carrier gas, was 50 cm s-1. GC–MS interface temperature was 300 °C, and the (electron impact) ion source temperature was 200 °C, with an ionization voltage of 70 eV. Fractional labelling from 13C-labelled substrates and mass isotopomer distributions were calculated as described previously57. Data from standards were used to construct standard curves in MetaQuant58, from which metabolite amounts in samples were calculated. Metabolite amounts were corrected for the recovery of the internal standard and for 13C labelling to yield total (labelled and unlabelled) quantities in nanomoles per sample and then adjusted by cell number.

Liquid chromatography–mass spectrometry sample preparation and analysis

Naive CD4+ T cells were polarized under TH0, TH1, TH17 and iTreg culture conditions or cultured with IL-7 (Tnai condition) for 72 h. Then, the cells were collected, washed with PBS and reseeded at a density of 5 × 106 cells per ml in fresh medium. After 6 h of culture, the cell medium was collected and snap-frozen. Sample preparation and analysis were carried out as described previously at Metabolon59. In brief, sample preparation involved protein precipitation and removal with methanol, shaking and centrifugation. The resulting extracts were profiled on an accurate mass global metabolomics platform consisting of multiple arms differing by chromatography methods and MS ionization modes to achieve broad coverage of compounds differing by physiochemical properties such as mass, charge, chromatography separation and ionization behaviour. Metabolites were identified by automated comparison of the ion features in the experimental samples to a reference library of chemical standard entries that included retention time, molecular weight (m/z), preferred adducts and in-source fragments as well as associated MS spectra and were curated by visual inspection for quality control using a software developed at Metabolon.

Metabolite quantification

In some experiments, TH17 cells were suspended at a density of 5 × 106 cells per ml with medium containing vehicle or Vig (1 mM). After 6 h of culture, blank medium (without cells) and spent medium were collected. The levels of Gln and Glu were measured using the Bioanalyzer (YSI 2900). Following the manufacturer’s instructions, Arg and GAB quantities were determined by l-Arginine Assay Kit (BioVision) and GABA Research ELISA Kit (LDN). Consumption or production of each metabolite was determined by calculating the difference between blank and spent media.

OCR

Following the manufacturer’s instructions, the OCR was determined using the Seahorse XFe96 Analyzer (Agilent Technologies). Briefly, approximately 1 × 105 TH17 cells were suspended in a 50 µl assay medium (Seahorse XF RPMI Assay Medium, pH 7.4, Agilent Technologies) containing 10 mM Glc, 2 mM Glu and 1 mM pyruvate and were seeded in an XF96 Cell Culture Microplates (Seahorse, Agilent Technologies) precoated with poly(d-lysine) (50 µg ml-1; Millipore). The cells were centrifuged at 200g for 2 min on a zero-braking setting to immobilize the cells before they were supplied with an additional 130 µl of assay medium and kept in a non-CO2 incubator for 30 min. Data analysis was performed using the Seahorse Wave Software (Seahorse, Agilent Technologies). In some experiments, the GABAA-R antagonist bicuculline (5 µM) was added along with GABA to prevent the activation of GABAA-R. Various compounds were injected into each well sequentially to achieve the following final concentrations: 0.5 mM GABA, 1 mM Vig, 20 µM R162, 1.5 µM oligomycin, and 1 µM FCCP.

Western blot analysis, RNA extraction, qPCR, and RNA-seq and NMR analysis of medium

Details are provided in the Supplementary Information.

Statistical analysis

Statistical analysis was conducted using the GraphPad Prism software (GraphPad Software; v 8.0.1). To determine the statistical significance, different tests including unpaired two-tailed Student’s t-test, one-way ANOVA with Tukey’s multiple-comparisons test and two-way ANOVA with Sidak’s multiple-comparisons test were used as indicated in the figure legends. The number of experimental repeats is indicated in the figure legends. R software (v 4.2.1) was used for Metabolon and RNA-seq data analysis. P values that were considered significant are shown in the corresponding figures.

Reporting summary

Further information on the research design is available in the Nature Research Reporting Summary linked to this article.