Aerobic glycolysis promotes T helper 1 cell differentiation through an epigenetic mechanism

See allHide authors and affiliations

Science  28 Oct 2016:
Vol. 354, Issue 6311, pp. 481-484
DOI: 10.1126/science.aaf6284

Metabolic support for T cell functions

For immunological T cells, responding to infections is energetically demanding. T cells rewire their metabolism so that they rely more heavily on aerobic glycolysis. This helps them to support important effector functions such as secreting the cytokine interferon γ (IFNγ). Peng et al. now provide insight into how aerobic glycolysis promotes T cell effector function. Activated T cells express the aerobic glycolysis–supporting enzyme lactate dehydrogenase A (LDHA), allowing these cells to maintain high amounts of acetyl–coenzyme A, which in turn promotes histone acetylation and transcription of cytokines such as IFNγ. Engineered mice whose T cells lacked LDHA were protected from IFNγ-dependent pathologies that often characterize autoinflammatory diseases.

Science, this issue p. 481


Aerobic glycolysis (the Warburg effect) is a metabolic hallmark of activated T cells and has been implicated in augmenting effector T cell responses, including expression of the proinflammatory cytokine interferon-γ (IFN-γ), via 3′ untranslated region (3′UTR)–mediated mechanisms. Here, we show that lactate dehydrogenase A (LDHA) is induced in activated T cells to support aerobic glycolysis but promotes IFN-γ expression independently of its 3′UTR. Instead, LDHA maintains high concentrations of acetyl–coenzyme A to enhance histone acetylation and transcription of Ifng. Ablation of LDHA in T cells protects mice from immunopathology triggered by excessive IFN-γ expression or deficiency of regulatory T cells. These findings reveal an epigenetic mechanism by which aerobic glycolysis promotes effector T cell differentiation and suggest that LDHA may be targeted therapeutically in autoinflammatory diseases.

T cell activation and differentiation are associated with metabolic rewiring (14). A metabolic hallmark of activated T cells is aerobic glycolysis (5), the conversion of glucose to lactate in the presence of oxygen, but its physiopathological functions remain incompletely understood (68). As the major carbon source, glucose plays important roles in T cell development, proliferation, and function (915). However, the specific contribution of aerobic glycolysis to T cell responses has not been well defined. Using galactose as a sugar source, aerobic glycolysis was proposed to support interferon-γ (IFN-γ) expression through 3′UTR-mediated mechanisms (12). Although galactose is metabolized at a slower rate than glucose via the Leloir pathway, both sugars are converted to lactate (16), rendering the galactose system unable to model aerobic glycolysis deficiency in a definitive manner.

By converting pyruvate to lactate with regeneration of nicotinamide adenine dinucleotide (NAD+) (17), lactate dehydrogenase (LDH) defines the biochemical reaction of aerobic glycolysis. LDHA and LDHB form five tetrameric LDH isoenzymes (A4B0, A3B1, A2B2, A1B3, and A0B4) with distinct kinetic properties (17). In a zymography assay, activated CD4+ T cells manifested LDH activity predominantly in the form of A4B0, similar to that of muscle tissues (fig. S1A). Consistently, LDHA, but not LDHB, was induced upon T cell activation (fig. S1B), likely due to hypoxia-inducible factor 1α (HIF-1α)– and c-Myc–induced transcription of Ldha (10, 11).

To study the definitive function of aerobic glycolysis, we deleted LDHA specifically in T cells (CD4CreLdhafl/fl, designated as knockout, KO) (fig. S1B) (18). Compared to activated wild-type (WT) CD4+ T cells, KO T cells barely produced lactate (Fig. 1A). Furthermore, glucose consumption in KO T cells was reduced to ~30% of WT amounts (Fig. 1B), in line with a critical role for LDHA in sustaining aerobic glycolysis through regeneration of NAD+ consumed at the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) step of glycolysis (17). Consistent with low Glut1 (13) and LDHA expression (fig. S1B), naïve CD4+ T cells showed little activity of glucose metabolism, with negligible glucose-induced extracellular acidification rate (ECAR) or oxygen consumption rate (OCR) (Fig. 1, C and D). Upon activation, ECAR was increased in WT CD4+ T cells, which was largely diminished in KO T cells (Fig. 1, C, E, and F). In contrast, both basal and maximal OCR were elevated in KO T cells (Fig. 1, D, G, and H).

Fig. 1 LDHA dictates aerobic glycolysis in activated CD4+ T cells.

(A and B) Naïve CD4+ T cells isolated from wild-type (WT) or CD4CreLdhafl/fl (KO) mice were stimulated with anti-CD3 and anti-CD28 in the presence of IL-2 for 2 days. Cells were replenished with fresh medium, which was harvested 24 hours later. Lactate production (A) and glucose consumption (B) were determined with triplicates. (C to I) Naïve WT or KO CD4+ T cells were stimulated as in (A). Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were measured with a glycolysis stress test kit [(C), (E), and (F)] or a Mito stress test kit [(D) and (G) to (I)]. Baseline ECAR (E) and stressed ECAR (F) of activated CD4+ T cells were calculated according to (C). Baseline OCR (G), stressed OCR (H) and baseline OCR/ECAR (I) were calculated according to (D). Statistics (E) to (I) were from one of two independent experiments each with eight biological replicates (n = 8); data represent mean ± SD, two-tailed unpaired t test, ***P ≤ 0.001.

13C-isotope–labeled glucose (13C6-glucose) tracing experiments showed that glycolysis was slowed down at the GAPDH step in KO T cells (fig. S2). However, 13C-labeled citrate was increased in the absence of LDHA (fig. S2). In particular, doubly (m+2) and quadruply (m+4) 13C-labeled citrate, readouts of tricarboxylic acid (TCA) cycle activity through the first and second turns, respectively, were much increased in KO T cells (fig. S2), in line with enhanced OCR (Fig. 1, D, G, and H). Together, these findings reveal that glucose metabolism in activated CD4+ T cells is shifted from aerobic glycolysis toward oxidative phosphorylation in the absence of LDHA (Fig. 1I).

LDHA deficiency did not affect thymic development of conventional or regulatory T (Treg) cells (fig. S3), or T cell homeostasis in peripheral (fig. S4). LDHA deficiency also did not alter expression of activation markers, size, or survival of activated CD4+ T cells (fig. S5, A to D), whereas proliferation was slightly delayed (fig. S5E). These phenotypes were distinct from those of Glut1-deficient T cells (13), suggesting that general glucose metabolism, but not aerobic glycolysis, is required for T cell development and homeostasis.

Glycolysis promotes expression of effector molecules including the type 1 cytokine IFN-γ (9, 1215). Indeed, LDHA deficiency led to diminished IFN-γ expression in T cells differentiated under T helper 1 (TH1) conditions (Fig. 2, A and B). A recent study proposed that aerobic glycolysis enhanced IFN-γ production by sequestering GAPDH away from binding to the Ifng 3′UTR, thereby enhancing IFN-γ translation (12). To determine whether such regulation accounted for reduced IFN-γ expression in KO T cells, we used a green florescent protein (GFP) reporter in which the GFP open reading frame was fused to the 3′UTR of Ifng or Gapdh, with the latter not known to repress mRNA translation (19). As previously reported (19), GFP expression controlled by the Ifng 3′UTR showed decreased mean fluorescence intensity compared to that regulated by the Gapdh 3′UTR in WT T cells (Fig. 2, C to E). GFP expression under the control of Ifng 3′UTR was diminished to a similar extent in KO T cells (Fig. 2, C to E), suggesting that Ifng 3′UTR is insufficient to mediate aerobic glycolysis regulation of IFN-γ expression.

Fig. 2 LDHA regulates IFN-γ expression independent of its 3′UTR.

(A and B) Naïve wild-type (WT) or CD4CreLdhafl/fl (KO) CD4+ T cells were differentiated under TH1 conditions for 3 days, and restimulated with phorbol myristate acetate (PMA) and ionomycin for 4 hours. IFN-γ expression was determined by intracellular staining. Representative plots (A) and mean fluorescence intensity (MFI) (B) are shown. (C to E) Naïve WT or KO CD4+ T cells were activated for 2 days and transduced with green fluorescent protein (GFP) constructs fused with the 3′ untranslated region (3′UTR) of the Ifng or Gapdh gene. GFP expression was measured 48 hours after transduction. Representative plots (C) and MFI of GFP in GFP+ cells are shown (D). (E) MFI of GFP from IFN-γ 3′UTR was normalized to that of GAPDH 3′UTR control. (F) A diagram of IFN-γ mRNA expressed from the WT Ifng or Yeti allele. (G and H) Naïve CD4+ T cells from Yeti/Yeti or KOYeti/Yeti mice were differentiated as in (A). The expression of IFN-γ and eYFP were determined by flow cytometry. Representative plots and respective statistical analysis are shown in (G) and (H). Statistics (B), (E), and (H) were from one of two independent experiments each with three biological replicates (n = 3); data represent mean ± SD, two-tailed unpaired t test, *P ≤ 0.05; ***P ≤ 0.001; ns, not significant.

To explore the definitive function of 3′UTR in LDHA control of IFN-γ expression, we used an Ifng reporter allele Yeti (yellow-enhanced transcript for IFN-γ) (20), in which the Ifng 3′UTR was replaced by 3′UTR of the bovine growth hormone (BGH) gene (Fig. 2F). We found that expression of both IFN-γ and yellow fluorescent protein (YFP) driven by the internal ribosome entry site (IRES) element in the reporter allele were proportionally reduced in the absence of LDHA (Fig. 2, G and H), demonstrating that Ifng 3′UTR is not required for LDHA control of IFN-γ expression.

To determine whether reduced IFN-γ production in KO T cells was caused by diminished transcription, we performed RNA sequencing experiments. A total of 363 transcripts were differentially expressed between WT and KO TH1 cells, among which 220 transcripts, including Ifng, were down-regulated in KO T cells (figs. S6A and S7A and table S1). IFN-γ transcription in TH1 cells is induced by T-bet (21). However, T-bet expression was not affected in KO T cells (fig. S7, B and C), suggesting that LDHA promotes IFN-γ expression via T-bet–independent mechanisms.

Glucose metabolism is implicated in the control of gene expression through epigenetic mechanisms that include histone acetylation (7, 22). Chromatin immunoprecipitation–sequencing (ChIP-seq) analysis of histone H3 acetylation at the lysine 9 residue (H3K9Ac), a histone mark associated with active transcription, showed that the differentially expressed genes between WT and KO TH1 cells had varying amounts of H3K9Ac (fig. S6B and tables S2 and S3), which encoded proteins involved in signal transduction, transcription, metabolism, and effector functions (fig. S6C). Notably, 86% of down-regulated transcripts that included Ifng had decreased H3K9Ac in KO T cells (fig. S6B and table S3), suggesting that LDHA may promote IFN-γ expression by modulating histone acetylation.

Compared to the constitutively active Cd3e locus, diminished H3K9Ac was observed in Ifng promoter, gene body, and the conserved noncoding sequence 22 kilobase pairs upstream of Ifng (CNS-22) in KO T cells (Fig. 3, A and B). Acetylation of histone H3 at lysine 27 (H3K27Ac) was also reduced, whereas total histone H3 was comparable (fig. S8, A and B). Notably, diminished histone acetylation was associated with reduced RNA polymerase II (Pol II) recruitment to the Ifng locus in KO T cells (fig. S8C). Histone acetylation requires acetyl–coenzyme A (acetyl-CoA) as a substrate, with glucose being a critical source. Considering enhanced TCA cycle activity in KO T cells (Fig. 1 and fig. S2), we hypothesized that in the absence of LDHA, less citrate would be exported from the mitochondria for acetyl-CoA regeneration. Indeed, cytosolic acetyl-CoA was decreased in LDHA-deficient TH1 cells (Fig. 3C).

Fig. 3 LDHA promotes IFN-γ expression through an epigenetic mechanism.

(A) Representative H3K9 acetylation (H3K9Ac) peaks at the Cd3e or Ifng promoter and enhancer (CNS22) regions from one of two ChIP-Seq experiments are shown. (B) H3K9Ac at the Ifng promoter and CNS22 enhancer regions in WT or KO TH1 cells were assessed by ChIP-qPCR (quantitative polymerase chain reaction). Enrichment was normalized to H3K9Ac at the Cd3e promoter region. (C) Naïve WT or KO CD4+ T cells were differentiated under TH1 conditions, and acetyl-CoA concentrations were measured by an acetyl-CoA assay kit. (D and E) Naïve WT or KO CD4+ T cells were differentiated under TH1 conditions for 3 days and either left untreated or supplemented with 20 mM sodium acetate for another 24 hours. Cells were subsequently restimulated with PMA and ionomycin for 4 hours. The expression of IFN-γ protein (D) and mRNA (E) was determined by flow cytometry and qPCR, respectively. mRNA level of IFN-γ was normalized to that of β-actin. MFIs of IFN-γ are shown in (D). (F) T cells were cultured as in (D), and H3K9Ac at the Ifng promoter and CNS22 enhancer regions were assessed by ChIP-qPCR. Enrichment was normalized to that of H3K9Ac at the Cd3e promoter region. Statistics (B), (C), (E), and (F) were from one of three independent experiments each with three biological replicates (n = 3); data represent mean ± SD, two-tailed unpaired t test, *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns, not significant.

To determine whether reduced cytosolic acetyl-CoA was sufficient to repress IFN-γ expression, we inhibited adenosine 5′-triphosphate (ATP)–citrate lyase (ACL), the enzyme converting citrate to acetyl-CoA, and found that IFN-γ expression was diminished in WT T cells (fig. S9, A to D). Acetyl-CoA can be generated from acetate by acetyl-CoA synthetase independent of citrate (23). Acetate supplementation augmented acetyl-CoA production (fig. S9E) and corrected IFN-γ expression in KO T cells (Fig. 3, D and E) without affecting T-bet expression (fig. S9F). Instead, enhanced IFN-γ production was associated with normalization of H3K9Ac in Ifng promoter and enhancer regions (Fig. 3F). Histone acetylation is a dynamic process controlled by competing activities of histone acetyltransferases (HATs) and histone deacetylases (HDACs) (24). TSA, a HDAC inhibitor, rectified IFN-γ expression in KO T cells without affecting T-bet expression (fig. S9, G and H), which was associated with increased H3K9Ac in the Ifng locus (fig. S9I). Collectively, these findings demonstrate that LDHA promotes IFN-γ expression through an epigenetic mechanism of histone acetylation.

In the absence of endogenous 3′UTR, the Ifng transcript in Yeti/Yeti mice is stable, resulting in sustained IFN-γ production and a lethal autoinflammatory phenotype (25). Indeed, all Yeti/Yeti mice succumbed to death by 3 weeks of age (Fig. 4A), which was associated with severe liver damage (Fig. 4B and fig. S10A). Deletion of Ldha in T cells corrected the liver immunopathology and conferred long-term survival of Yeti/Yeti mice (Fig. 4, A and B, and fig. S10A). LDHA deficiency did not substantially affect the cellularity or activation of Yeti/Yeti CD4+ T cells (fig. S10, B to E). However, IFN-γ expression in Yeti/Yeti CD4+ T cells was diminished in the absence of LDHA, whereas its expression in LDHA-sufficient Yeti/Yeti natural killer (NK) cells was not affected (Fig. 4, C and D). The correction of a lethal autoinflammatory disease in Yeti/Yeti mice by T cell–specific deletion of Ldha supports the notion that aerobic glycolysis promotes IFN-γ expression in T cells through an epigenetic mechanism of Ifng transcription but not via a 3′UTR-dependent mechanism of translation in vivo.

Fig. 4 LDHA deficiency in T cells protects Yeti/Yeti mice from a lethal autoinflammatory disease.

(A) The survival curve of Yeti/Yeti (n = 5) and CD4CreLdhafl/flYeti/Yeti (KO Yeti/Yeti) mice (n = 6). (B) Representative hematoxylin and eosin staining of liver sections from wild-type (WT), CD4CreLdhafl/fl (KO), Yeti/Yeti, and KO Yeti/Yeti mice. Arrows indicate the necrotic regions. (C and D) Splenocytes from mice of the indicated genotypes were stimulated with PMA and ionomycin for 4 hours. IFN-γ and YFP expression in CD4+ T cells and NK1.1+TCR NK cells was determined by flow cytometry. Representative plots (C) and statistics (D) are shown; data represent mean ± SD, n = 4 to 5 mice per genotype, two-tailed paired t test, ***P ≤ 0.001; ns, not significant.

To further explore aerobic glycolysis in control of effector T cell responses, we used Scurfy mice with a mutation in the Treg cell lineage Foxp3 gene (Foxp3sf) (26). LDHA deficiency in T cells corrected the Scurfy phenotype and extended life span of Foxp3sf mice (fig. S11, A and B), which was associated with reduced inflammation in multiple organs without affecting lymphadenopathy or splenomegaly (fig. S11, C to F). T cell activation and expansion were comparable between Foxp3sf and KO Foxp3sf mice (fig. S11, G and H), which was in line with our in vitro findings (fig. S5, A to E). Nevertheless, CD4+ T cells produced less IFN-γ in the absence of LDHA (fig. S11, I and J), demonstrating a critical role for LDHA-mediated aerobic glycolysis in promoting autoreactive TH1 cell responses.

Our findings do not support a translational mechanism of aerobic glycolysis in IFN-γ production (12), as LDHA promotes IFN-γ expression independent of its 3′UTR. Previous studies have shown that utilization of glycolytic intermediates for biosynthesis accounts for a small fraction (~7%) of the glucose consumed in activated T cells (27), whereas aerobic glycolysis produces the majority (~60%) of ATP (28). Hence, LDHA-mediated aerobic glycolysis may primarily relieve the burden of mitochondria as an energy house to “burn” carbons to generate ATP. As a result, more citrate can be exported out of mitochondria to generate acetyl-CoA and promote histone acetylation in selected gene loci (fig. S12). LDHA inhibitors are being developed to target tumor cell metabolism (29). Our data suggest that LDHA inhibitors could be immunosuppressive, and thus may complicate their applications in cancer. Instead, LDHA inhibitors, as well as inhibitors targeting other enzymes involved in acetyl-CoA metabolism, such as ACL, may be useful in treatment of autoimmune diseases and transplant rejection.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 to S3

References (3037)

References and Notes

  1. See materials and methods in supplementary materials on Science Online.
  2. Acknowledgments: We thank R. Locksley for the Yeti mice, A. Abbas and A. Villarino for the 3′UTR GFP constructs, and J. Cross and J. Ye for help with Seahorse experiments. We also thank the Li laboratory for discussions. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. MSKCC has filed a provisional patent application with the U.S. Patent and Trademark Office (application no. 62/394,859). The application is directed toward inhibition of acetyl-CoA metabolism for treatment and prevention of immune disease and disorders. M.P., N.Y., and M.O.L. are listed as inventors on this patent application. This work was supported by the Ludwig Center for Cancer Immunology (M.O.L.), the Functional Genomics Initiative (M.O.L.), a Faculty Scholar grant from the Howard Hughes Medical Institute (M.O.L.), and the Memorial Sloan Kettering Cancer Center Support Grant/Core Grant (P30 CA008748). RNA-seq and ChIP-Seq data are deposited in the Genome Expression Omnibus under accession number GSE86188.
View Abstract

Stay Connected to Science

Navigate This Article