Fueling Immunity: Insights into Metabolism and Lymphocyte Function

Lymphocyte Metabolism Is Dynamically Regulated

Maintenance of cellular bioenergetics is an essential function of all living cells, and lymphocytes are no exception. In T lymphocytes, glucose is a critical substrate for adenosine triphosphate (ATP) production (4). During glycolysis, glucose is broken down into two molecules of pyruvate. This process, which does not require oxygen, yields two reduced nicotinamide adenine dinucleotide (NADH) molecules and two net ATP molecules per molecule of glucose. Pyruvate has two alternate fates. Most terminally differentiated, nonproliferating cells can fully oxidize pyruvate in the tricarboxylic acid (TCA) cycle. This process generates NADH and reduced flavin adenine dinucleotide (FADH₂), which the cell can use to fuel OXPHOS, an oxygen-dependent process that produces up to 36 molecules of ATP per glucose molecule. Alternatively, pyruvate can be transformed (or fermented) into lactate, regenerating NAD⁺ for subsequent use in glycolysis (5). From a bioenergetic perspective, engaging OXPHOS maximizes the amount of ATP that can be derived from glucose.

Bioenergetic profiling of T cells has revealed that T cell metabolism changes dynamically with activation state (Fig. 1). Upon antigen encounter, T cells become activated, undergo extensive proliferation, and differentiate into effector T cells (T_EFF); upon pathogen clearance, most T_EFF cells die, leaving behind a small population of long-lived antigen-specific memory T cells (T_M). Consistent with the metabolism of other nonproliferating cells, resting naïve T cells (T cells that have not yet encountered antigen) maintain low rates of glycolysis and predominantly oxidize glucose-derived pyruvate via OXPHOS or engage fatty acid oxidation (FAO) to make ATP. Upon activation, T cells switch to a program of anabolic growth and biomass accumulation to generate daughter cells, which by definition dictates increased demand for ATP and metabolic resources. In this state, T cells are considered to be metabolically activated (Fig. 1). T cell receptor (TCR) signaling directs the metabolic reprogramming of naïve T cells. TCR ligation promotes the coordinated up-regulation of glucose and amino acid transporters (6–8), facilitating nutrient uptake and T cell blastogenesis. TCR-mediated up-regulation of the transcription factors c-Myc (9) and estrogen-related receptor α (ERRα) (10) enhances the expression of genes involved in intermediary metabolism. In addition, catabolic pathways of ATP generation such as fatty acid β-oxidation are actively suppressed (9). The predominant metabolic phenotype of activated T cells is a shift to aerobic glycolysis [reviewed in (11)]. Both CD4⁺ and CD8⁺ T_EFF cells engage aerobic glycolysis, which is marked by the conversion of glucose-derived pyruvate to lactate despite the availability of oxygen for complete glucose oxidation. This process, also known as the Warburg effect from earlier work in cancer biology, is a common trait of actively proliferating cells (5). It is important to note that OXPHOS is still engaged in T_EFF cells (9); however, the production of lactate from pyruvate by aerobic glycolysis is the dominant pathway of glucose metabolism in T_EFF cells. Regulating energy metabolism may provide a way for T cells to reversibly switch between quiescent and highly proliferative states (12).

As a quiescent T cell population, T_M cells adopt a metabolic profile similar to that of naïve T cells—a catabolic metabolism characterized by increased reliance on OXPHOS and lower rates of nutrient uptake and biosynthesis relative to T_EFF cells (Fig. 1). However, T_M cells also display a characteristic increase in mitochondrial mass, which translates into greater mitochondrial spare respiratory capacity (SRC) relative to naïve or T_EFF populations (13). SRC can be viewed as the maximal respiratory capacity available to a cell, much like the maximum speed that can be achieved by a car engine. Under increased workload, stress, or nutrient limitation, cells engage this reserve capacity to generate more energy and promote cell viability (14, 15). We have recently shown that increased mitochondrial mass and SRC of T_M cells allows for rapid mitochondrial ATP production upon TCR engagement, conferring a bioenergetic advantage to T_M cells upon secondary exposure to antigen (16). From this vantage, T_M cells may be viewed as being metabolically primed, with mitochondrial metabolism fueling the rapid recall response to reinfection. The memory T cell–promoting cytokine interleukin (IL)–15 plays a key role in this catabolic switch by promoting mitochondrial biogenesis (13).

The mechanisms governing the transition of T cells from effector to memory states are still poorly understood, but recent work hints that changes in metabolism may influence this process. We previously demonstrated that mitochondrial FAO stimulated downstream of TNF (tumor necrosis factor) receptor–associated factor 6 (TRAF6) is required for memory CD8⁺ T cell development (17). Oxidation of free fatty acids (FFAs) generates acetyl–coenzyme A (CoA), which can be metabolized further in the TCA cycle, as well as FADH2 and NADH, which can be used directly by the electron transport chain (ETC) to make ATP. FFAs are energy-dense molecules, and FAO may be a preferred fuel source for T_M cells as they rely on OXPHOS-dependent metabolic program. Administration of metformin, a metabolic stressor that activates the energy sensor adenosine monophosphate–activated protein kinase (AMPK), enhances the generation of CD8⁺ T cell memory (17). One consequence of AMPK activation is the suppression of mammalian target of rapamycin complex 1 (mTORC1) activity in response to energetic stress (18). Consistent with this, the drug rapamycin, which also inhibits mTORC1, enhances the generation of CD8 T_M cells (17, 19, 20). These observations suggest that manipulating the metabolism of antigen-specific cells during contraction can influence the development of T_M cells. Given these observations, T_M formation may be influenced by a number of enzymes and transporters involved in fatty acid synthesis, desaturation, and oxidation, as well as the availability of FFAs to memory precursor cells. Some important players to consider in this regard include acetyl-CoA carboxylase (ACC2) (21), the mitochondrial lipid transporter CPT1A (13, 22), and metabolites such as acetyl-CoA and malonyl-CoA (23). AMPK activation and mTOR inhibition are also both potent activators of autophagy, a catabolic process induced during starvation that promotes the degradation and recycling of cellular components [reviewed in detail in (24)]. Proper induction of autophagy has been shown to be important for the maintenance of cellular bioenergetics and sustained T cell viability after activation (25, 26). It will be interesting to determine whether autophagy, by coupling catabolic fuel supply to mitochondrial metabolism, is important for T_M formation after infection.

Mitochondrial OXPHOS and T Cell Activation

Although much focus has been placed on the shift toward glycolysis that accompanies T cell activation, evidence suggests that mitochondrial OXPHOS is also important for T cell activation. Oligomycin, a specific inhibitor of mitochondrial ATP synthase, can block the expression of early activation markers after TCR ligation and blunts subsequent T cell proliferation (27), suggesting that the naïve-to-effector transition requires either de novo production of ATP by mitochondria or specific signals generated during mitochondrial ATP production. Mitochondrial-derived reactive oxygen species (ROS) may function as such a “bioenergetic” second messenger. There has long been evidence that ROS can play critical roles in shaping T cell responses (28–30). However, recent work suggests that mitochondrial ROS produced during OXPHOS is essential for T cell activation. T cells deficient for ubiquinol-cytochrome c reductase (Uqcrfs1), a component of complex III of the ETC, display impaired TCR-dependent ROS production and defects in antigen-specific proliferation (31). Intracellular calcium (Ca²⁺) flux, an early event in TCR signal transduction, may provide the functional link between TCR ligation, mitochondrial OXPHOS, and cell proliferation. Uptake of Ca²⁺ by mitochondria stimulates Ca²⁺-dependent dehydrogenases of the TCA cycle, driving mitochondrial NADH production and ATP production by OXPHOS during early T cell activation (32). T cells lacking the apoptosis regulators Bax and Bak, which display defects in intracellular Ca²⁺ homeostasis, exhibit reduced Ca²⁺-dependent mitochondrial ROS production and T cell proliferation after TCR stimulation (33). Restoring Ca²⁺ signals in Bax/Bak-null T cells restores mitochondrial ROS production and T cell proliferation (33). Thus, although toxic in many biological settings, mitochondrial-dependent ROS may prime T cells and license full T cell activation.

Metabolic Signatures Vary with Differentiation State

Although the paradigm of T cell metabolism as summarized in Fig. 1 holds true with respect to activated versus quiescent states, the metabolic signature of T cells can also vary depending on differentiation state. This was first demonstrated by Michalek et al. (34), who determined that proinflammatory CD4⁺ T helper (T_H) cells (T_H1, T_H2, and T_H17 lineages) displayed a strong bias toward glycolysis over mitochondrial metabolism, whereas induced CD4⁺ T regulatory (T_reg) lineage cells displayed a mixed metabolism involving glycolysis, lipid oxidation, and OXPHOS. In particular, T_H17 cells display increased reliance on glycolysis for their development and maintenance. T_H17 cell development is promoted by hypoxia inducible factor–1α (HIF-1α) (35, 36), an oxygen-sensitive transcription factor that regulates glycolytic gene expression in T_H17 cells. Blocking glycolysis during T_H17 cell differentiation reduced the development of T_H17 cells and favored the formation of T_regs (35). Added to this are recent results indicating that extracellular salt (NaCl) (37, 38) and short-chain fatty acids (39) can influence T_H17 and T_reg homeostasis, respectively. This raises the intriguing possibility that the metabolic microenvironment (i.e., nutrient and oxygen availability) can influence T cell polarization (to be discussed later). Determining whether the metabolic signature of differentiated T cells is simply a consequence of lineage-specific cytokine signaling or is instructive for T cell function (i.e., essential for regulating T cell plasticity and/or effector function) remains a question for the field. Examining the influence of key metabolic regulators such as HIF-1α, mTOR, and AMPK on T cell differentiation and plasticity will help in resolving these issues.

The Warburg Effect: More than ATP Synthesis

Rapid glucose processing promoted by the Warburg effect allows proliferating T cells to generate ATP quickly; glycolysis also generates metabolic intermediates important for cell growth and proliferation (Fig. 2). Metabolism of glucose through the oxidative or nonoxidative arms of the pentose phosphate pathway (PPP) generates ribose-5-phosphate (Rib-5P), a key intermediate in nucleotide biosynthesis. The oxidative arm of the PPP also produces NADPH, the key metabolic currency for nucleotide and fatty acid biosynthesis. T cell activation promotes a rapid increase in glucose flux through the oxidative PPP (9). Dihydroxyacetone-phosphate (DHAP) is used to generate the glycerol backbone for glycerophos-pholipids, and 3-phosphoglycerate (3PG) is a key intermediate in both amino acid and nucleotide biosynthesis (discussed below). Pyruvate that is not converted to lactate can enter into the mitochondria and be converted into acetyl-CoA by the pyruvate dehydrogenase (PDH) complex. In proliferating cells, mitochondria adopt an additional role as a biosynthetic hub, converting pyruvate and other metabolites into metabolic intermediates involved in protein and fatty acid biosynthesis (44).

One of the key metabolic intermediates for biosynthesis is acetyl-CoA. Acetyl-CoA has a central role in membrane biogenesis because it provides two-carbon units for fatty acid and isoprenoid biosynthesis, as well as in other diverse processes such as protein prenylation and N-glycosylation (45). The flow of glucose to the cytosolic acetyl-CoA pool is regulated by using TCA cycle intermediates and a truncated TCA cycle (46). In this model, mitochondrial citrate is formed from condensation of oxaloacetate and acetyl-CoA, after which citrate is exported from the mitochondrion to the cytosol and converted back to acetyl-CoA by ATP citrate lyase (ACL) (47). Despite the availability of extracellular lipids for membrane biosynthesis, FFAs are generated de novo from glucose in proliferating cells using this pathway (46). CD8+ T cells unable to engage acetyl-CoA–dependent lipid biosynthetic pathways display defects in antigen-driven blastogenesis and clonal expansion in response to pathogens (48). Acetyl-CoA can also influence metabolic flux through acetylation of metabolic enzymes (49, 50), reinforcing metabolic pathways such as glycolysis when carbon availability is high. Glucose availability and acetyl-CoA production can also influence epigenetics by regulating the cytosolic acetyl-CoA pools available for histone acetylation reactions (51), raising the prospect that glucose-dependent metabolic flux may help drive or reinforce T cell differentiation programs.

One consequence of using glucose-derived mitochondrial citrate for lipid biosynthesis is the potential depletion of TCA cycle intermediates. Oxaloacetate (OAA) is a rate-limiting substrate for acetyl-CoA entry into the TCA cycle. OAA generated from ACL-mediated cleavage of cytosolic citrate can potentially cycle back into the mitochondria to maintain the TCA cycle. Inefficient cycling of this pathway would lead to cumulative depletion of mitochondrial OAA, leading to collapse of the TCA cycle and disruption of mitochondrial function. One way tumor cells counter this is by engaging glutaminolysis, a metabolic shunt that converts glutamine into α-ketoglutarate (α-KG, also known as 2-oxoglutarate) for use in the TCA cycle (Fig. 2). Glutamine has long been known as a key metabolite for supporting T cell function (52). Recent evidence suggests that glutamine metabolism is as dynamically regulated in T cells as glucose metabolism. Glutamine transporters (SNAT1 and SNAT2) as well as key glutaminolysis enzymes (GLS, GPT, GOT, and GLUD) are up-regulated early after T cell activation similar to glycolysis genes (7, 9); several groups have correlated these changes in gene expression to enhanced glutaminolytic flux in lymphocytes (9, 53). Recently, it was found that engagement of the TCR leads to the expression of Slc7a5, an amino acid transporter that mediates the import of large neutral amino acids, such as leucine (8). Amino acid influx via this transporter is required for the activation of mTOR and expression of c-Myc and as such coordinates activation-induced metabolic reprogramming and differentiation of T cells. Thus, amino acids such as glutamine and leucine appear to play additional roles in T cell function beyond protein biosynthesis and may directly influence T cell activation by regulating metabolic reprogramming.

Control of Glycolytic Flux by Pyruvate Kinase M2 (PKM2)

Pyruvate kinase (PK) is a key enzyme of the glycolytic pathway. It catalyzes the terminal reaction of glycolysis by promoting the conversion of phosphoenolpyruvate (PEP) to pyruvate (Fig. 2), and is one of two ATP-generating steps of glycolysis (the second is mediated by phosphoglycerate kinase). The muscle version of PK exists as one of two isoforms, M1 or M2, generated from differential splicing of the PKM primary mRNA transcript, with the PKM2 splice variant expressed in embryonic tissues, proliferating cells, and tumor cells (54). Naïve T cells express both M1 and M2 isoforms at rest; mitogen-dependent activation promotes the rapid accumulation of the M2 isoform, which becomes the dominant isoform expressed in T_EFF cells (55). PKM2 exists as either an inactive dimer or an active tetramer, and oscillation between these two states influences the ability of cells to maintain anabolic metabolism (56).

Multiple lines of evidence point to a role for PKM2 in coordinating glycolytic flux and cell proliferation. Tumor cells engineered to exclusively express the M2 isoform display increased lactate production characteristic of the Warburg effect and gain a growth advantage in vivo over M1-expressing tumors (57). Interestingly, the M2 isoform is actually less efficient at converting PEP to pyruvate than the M1 isoform. Moreover, growth-factor–stimulated tyrosine phosphorylation of PKM2 further decreases its activity (58, 59). This prompts the question: Why would proliferating cells, including T cells, promote the expression of a pyruvate kinase isoform that is less efficient at generating ATP? The answer to this question may lie with the second function of Warburg metabolism, namely, supporting anabolic growth. Buildup of PEP because of reduced PKM2 activity promotes the accumulation of glycolytic intermediates, which can then be shunted into upstream biosynthetic pathways to support amino acid, triglyceride, and nucleotide biosynthesis. Activating PKM2 by using small-molecule agonists increases PEP-to-pyruvate conversion, reducing the flux of glycolytic intermediates toward anabolic pathways and slowing tumor cell growth (56, 60). Thus, the ability of PKM2 to support cell proliferation may have less to do with ATP production and more with supporting biosynthetic pathways required for tumor cell growth.

PKM2 may also exert some of its effects on cell proliferation through nonmetabolic functions, including transcriptional and epigenetic regulation (61, 62). Of particular interest to immunologists is that PKM2 can phosphorylate signal transducer and activator of transcription 3 (STAT3) at Tyr⁷⁰⁵, promoting increased STAT3-dependent transcription (63). The phosphorylation of STAT3 by PKM2 demonstrates that PKM2 possesses both protein and pyruvate kinase activities, the former using PEP as a phosphate donor rather than ATP. The protein kinase activity of PKM2 (and any subsequent effects on STAT3 activity) would be predicted to be sensitive to metabolic flux, favored under high-glycolysis PEP conditions and antagonized by low-energy high-ADP concentrations. The development of mouse models to study the impact of PKM2 activity in vivo will be important for elucidating the role(s) of PKM2 in immune function.

The Serine Biosynthesis Pathway

Another glycolytic intermediate that can double as an anabolic precursor is 3PG. 3PG is the starting point for the glucose-dependent biosynthesis of serine and glycine via the serine biosynthesis pathway (SBP) (Fig. 3A). Key enzymes of the SBP are phosphoglycerate dehydrogenase (PHGDH), the rate-limiting step for serine biosynthesis, and serine hydroxymethyltransferase (SHMT), which uses serine as a methyl donor to convert tetrahydrofolate (THF) to methylene-THF, generating glycine in the process. Methylene-THF is a key intermediate in folate-mediated one-carbon metabolism that fuels nucleotide biosynthesis and methylation reactions. Serine is also an allosteric activator of PKM2 (64) and thus provides feedback to the glycolytic pathway to regulate 3PG levels and serine biosynthesis.

In mammals, serine and glycine are non-essential amino acids and are widely abundant in serum (and tissue culture medium). However, glucose-dependent serine biosynthesis is actively engaged in some tumors regardless of serine abundance. Amplification of PHGDH has been observed in breast cancer and melanoma (65), and increased PHGDH expression can promote both enhanced serine biosynthesis and the proliferation of cancer cells (65, 66). Enhanced flux through the SBP may confer a growth advantage to tumor cells beyond providing increased serine and glycine for biosynthetic reactions. First, PHGDH produces NADH, which can be used to maintain cytosolic redox balance or fuel mitochondrial OXPHOS to make ATP. The second step of the pathway, the conversion of 3-phosphohydroxypyruvate to 3-phosphoserine by phosphoserine aminotransferase (PSAT), requires glutamate and produces α-KG (Fig. 3A) (66). Thus, the SBP may promote an alternate pathway of α-KG production for mitochondrial metabolism or promote the activity of α-KG– dependent enzymes (to be discussed later). Last, serine and glycine are both intermediates in the production of reduced glutathione (GSH), a key cellular antioxidant. Recent evidence indicates that cancer cells actively produce GSH from glucose via the SBP as a buffer against oxidative damage (67). The expression of SBP enzymes is up-regulated in Myc-driven lymphomas (68); thus, we hypothesize that T_EFF cells may actively engage the SBP upon activation, even in the presence of exogenous serine. Future investigation will be needed to determine how the SBP influences T cell biology.

Reductive Carboxylation of α-KG

As mentioned previously, it is now appreciated that in proliferating cells the TCA cycle functions as a source of biosynthetic precursors in addition to its role in ATP production (46). Carbon enters the TCA cycle primarily at one of two entry points: (i) the condensation of glucose-derived acetyl-CoA with OAA to generate citrate and (ii) conversion of glutamine to α-KG via glutaminolysis (Fig. 3B). Carbon-tracing experiments using proliferating glioblastoma cells have established that both glucose and glutamine contribute to mitochondrial citrate pools and subsequent lipid synthesis (69). The oxidative decarboxylation of isocitrate to α-KG by isocitrate dehydrogenase (IDH) is highly favored thermodynamically, such that this reaction is believed to be irreversible and the reason for the “clockwise” flow of metabolic intermediates through the TCA cycle.

Groundbreaking new work suggests that metabolite flow through biochemical pathways does not always conform to conventional dogma. Although oxidation of glutamine-derived α-KG in the TCA cycle serves as a minor source of lipogenic acetyl-CoA under normal growth conditions, α-KG can be converted to citrate through reductive carboxylation under conditions of stress such as hypoxia or mitochondrial dysfunction (70–72). In this reaction, glutamine-derived α-KG is carboxylated, rather than decarboxylated, by IDH1 (in the cytosol) or IDH2 (in mitochondria) to generate citrate (Fig. 3B). This switch to reductive versus oxidative metabolism of α-KG is regulated in part by HIF-1α (71, 72), although the exact mechanism by which HIF-1α does so remains unclear. Engaging reductive carboxylation of α-KG in essence bypasses the conventional TCA cycle by using glutamine to generate the acetyl-CoA required for fatty acid synthesis (Fig. 3B). Under such metabolic reprogramming, cancer cells continue to use glycolysis for ATP production but switch from glucose to glutamine as the major lipogenic precursor.

One implication of this work is that tumor cells display a high degree of metabolic plasticity and can adapt their metabolism to support proliferation and viability under fluctuating environmental conditions. Whether T cells display similar metabolic flexibility in response to environmental cues requires further investigation. The work from Metallo and colleagues (71) suggests that T cells can use reductive glutamine metabolism for fatty acid biosynthesis under hypoxic conditions. Physiologic oxygen tension varies significantly between tissues, and hypoxic regions can be detected in the bone marrow, thymus, and spleen (73). Moreover, T cells are likely to experience hypoxia at sites of tissue inflammation (74). Stabilization of HIF-1α and metabolic reprogramming to favor reductive carboxylation may help T cells maintain proliferation and/or effector function at hypoxic inflammatory sites. HIF-1α protein expression has also been observed early after T cell activation (75), so it is unclear whether reductive carboxylation may also be engaged as part of the normal program of T cell expansion. As mentioned, HIF-1α has been implicated in the differentiation of both proinflammatory T_H17 cells (35, 36) and CD4⁺FoxP3⁺ T_reg cells (76). It is tempting to speculate that this alternate pathway of glutamine metabolism may influence the expansion or phenotypic stability of these T cell subsets.

Regulation of LKB1-AMPK Signaling by Adenylates

ATP is the primary carrier of chemical energy in the cell and essential for life. Thus, adenylates [ATP, adeonsine diphosphate (ADP), and AMP] are important units of cellular metabolic currency. In mammalian cells, fluctuations in cellular energy are monitored by the heterotrimeric AMPK complex [reviewed in (83)]. ATP, ADP, and AMP compete for nucleotide-binding sites of the γ regulatory subunit of AMPK, with AMP (low energy) promoting and ATP (high energy) antagonizing AMPK activation. As such, AMPK functions as a sensor of the cellular adenylate energy charge (84, 85). Elevation of the cellular AMP:ATP ratio leads to increased phosphorylation of AMPK at Thr¹⁷² of its activation loop by the kinase LKB1 (Fig. 4A). AMPK can also be activated by Ca²⁺ (via CamKKβ) and the cytokine transforming growth factor–β (TGF-β) (via TAK1), although LKB1 appears to be the sole kinase that couples adenylate binding to AMPK activation.

Together LKB1 and AMPK function as part of an evolutionarily conserved energy-sensing pathway that maintains cellular energy balance by promoting catabolic pathways of ATP production and limiting processes that consume ATP. Protein synthesis is one of the most energy-consuming processes in the cell, accounting for ∼20% of cellular ATP consumption (86). As mentioned, AMPK antagonizes mRNA translation through negative regulation of mTORC1 (Fig. 4A) (18). By regulating AMPK activity, adenylates directly influence pathways of energy usage in the cell. As ATP levels drop, AMP binds to AMPK, and AMPK is switched on to promote ATP production and block its consumption; once ATP homeostasis has been reestablished, increased binding of ATP to AMPK shuts off the kinase.

Recent work indicates that LKB1-AMPK signaling can influence T cell metabolism and function. Lymphocytes exclusively express the a1 catalytic subunit of AMPK (87), and TCR stimulation promotes LKB1-dependent AMPK activation in lymphocytes (87, 88). Glycolysis is enhanced in resting AMPKα-deficient T cells (88), consistent with observations that silencing AMPK in tumor cells promotes the Warburg effect (89). Loss of LKB1-AMPK signaling promotes increased mTORC1 activity in both naïve and T_EFF cells (88, 90), which in turn facilitates production of the T_H1 cytokine interferon-γ (IFN-γ) by T_EFF cells (88). Thus, LKB1 and AMPK act in concert to limit the anabolic growth of T cells by suppressing glycolysis and mTOR-dependent biosynthesis. Perhaps not surprisingly, deletion of either LKB1 or AMPKα1 disrupts normal lymphocyte homeostasis, resulting in an accumulation of activated (CD44^hi CD62L^lo) CD8⁺ T cells in animals (88). These results suggest that cellular adenylate levels and AMPK may help regulate lymphocyte pools in the whole organism.

Why would a signaling pathway that normally monitors cellular energy levels regulate T cell function? Similar to tumor cells (91), AMPK may regulate a metabolic checkpoint in T cells, acting as a brake on lymphocyte expansion when energy conditions are poor. T_EFF cells with defective AMPK signaling would be freed from these metabolic checkpoints and continue to proliferate and produce cytokines as if cellular bioenergetics were suitable to support T cell function. Additionally, AMPK may regulate the metabolic plasticity of lymphocytes, coordinating metabolic changes in response to nutrient fluctuation and allowing Tcells to survive changes in their environment. As evidence for this, the AMPK agonist metformin, which promotes FAO in activated T cells, enhances the production of CD8⁺ memory T cells in vivo (17). Furthermore, it was recently shown that AMPK-deficient T cells are defective in their ability to generate CD8⁺ memory T cells during infection (92). There is also a growing body of evidence implicating AMPK in the control of inflammation (3). Future work will focus on the role of AMPK in regulating the metabolic fitness of lymphocytes, dissecting LKB1-and AMPK-specific effects on immune function and investigating the role(s) of LKB1 and AMPK in regulating inflammation in vivo.

Regulation of α-Ketoglutarate-Dependent Enzymes by TCA Cycle Metabolites

Although TCA cycle metabolites play central roles in energy metabolism, many function as chemical intermediates in other biological reactions. For example, fumarate can be used to chemically modify cysteine residues of proteins, resulting in the formation of S-(2-succinyl)cysteine or succinylation of these residues (93). Cancer cells harboring mutations in fumarate hydratase (FH), the TCA cycle enzyme that converts fumarate to malate, accumulate intracellular fumarate and display increased amounts of succinylated proteins (94). Increased protein succinylation has been associated with renal carcinoma and mechanistically can influence signal transduction pathways similar to protein phosphorylation or acetylation (95).

α-KG is another key metabolite involved in non-TCA cycle biochemical reactions. α-KG is essential for the activity of a class of enzymes known as α-KG–dependent dioxygenases (Fig. 4B). These enzymes use α-KG along with molecular O₂, iron, and ascorbate to modify target proteins (96). The most extensively characterized enzymes of this family are the HIF prolyl hydroxylases (PHDs). PHDs use O₂ to hydroxylate HIF-1α on two conserved proline residues, facilitating their recognition by the E3 ubiquitin ligase VHL and promoting HIF-1α degradation by the proteosome. When O₂ availability is low, HIF-1α protein is stabilized because of reduced PHD activity, resulting in increased HIF-1α-dependent transcription (97). This regulatory circuit allows HIF-1α to promote glycolytic ATP production when O₂ cannot support mitochondrial OXPHOS, an example of metabolic adaptation in response to environmental conditions. Other α-KG–dependent enzymes using this chemistry are the TET family of DNA hydroxylases, which hydroxylate 5-methylcytosine residues to promote DNA demethylation, and the Jumonji C (JmjC) class of histone demethylases (Fig. 4B). A detailed summary of these processes has recently been reviewed (98).

The activity of α-KG–dependent enzymes is directly affected by the abundance of TCA cycle intermediates. For example, 2-HG acts as a competitive inhibitor of α-KG, and its production by mutant IDH1 consumes α-KG (81, 99), leading to reduced TET2 and JmjC activity in tumor cells (79, 80). High levels of succinate and fumarate can inhibit PHD2 activity through product-mediated inhibition of PHD function, leading to HIF-1α protein stabilization under normoxic (20% O₂) conditions (100, 101). Additionally, total abundance of α-KG is low in most cell types. The implication of these findings is that metabolic flux through the TCA cycle can affect gene transcription and/or epigenetic programs. It was recently shown that succinate plays a central role in production of the cytokine IL-1 β by lipopolysaccharide (LPS)–stimulated macrophages (102). LPS induces the de novo production of succinate from glutamine, leading to PHD inhibition, stabilization of HIF-1α, and increased HIF-1α-dependent IL-1β production. Mutant IDH2 [Arg¹⁴⁰→Gln¹⁴⁰ (R140Q)], which promotes α-KG depletion/2-HG production, promotes DNA hypermethylation in hematopoietic cells and can inhibit myeloid differentiation when overexpressed in bone marrow stem cells (79). Differentiation of T cells to specific T_H lineages is driven by specific transcription factors but reinforced by epigenetic modifications on histone tails (H3K4 and H3K27 trimethylation) and DNA methylation of CpG dinucleotides (103). Differential TCA cycle flux or buildup of intermediates such as succinate or fumarate may influence the activity of α-KG-dependent enzymes that regulate T cell epigenetics. This may be one way in which the environment (O₂, nutrient levels) can influence the plasticity of T_H lineages at sites of infection or inflammation in vivo. Thus, much like the role of oncometabolites in tumorigenesis, studying the metabolism of T cell responses may reveal the existence of “immunometabolites” that influence T cell responses and inflammation.

Connecting Metabolism and Gene Regulation

A transformative conceptual change in the way we consider metabolism within cells is that changes in metabolism can be linked to changes in gene expression through posttranscriptional regulatory networks involving RNA, metabolites, and metabolic enzymes “moonlighting” as RNA binding proteins and regulating specific target mRNAs (104) (Fig. 5). Many enzymes, including those connected with the TCA cycle, glycolysis, PPP, fatty acid metabolism, and other pathways, have been shown to bind RNA in vitro and in cultured cells (105, 106). In addition, the RNA binding function of enzymes can be influenced by interactions with their metabolites or cofactors, illustrating how the metabolic state of the cell can control an enzyme’s RNA binding function.

The metabolic enzyme perhaps best characterized for its physiological role as an RNA binding protein is cytosolic aconitase (107, 108), a key regulator of cellular iron metabolism (109). In the early 1990s, it was shown that cytosolic aconitase and the RNA binding protein IRP-1 represent the same polypeptide and that the availability of iron triggers insertion or removal of an iron sulfur cluster—switching the protein’s function between RNA binding activity (iron low, IRP-1) and metabolic enzyme activity (iron high, aconitase) (110). Remarkably, IRP-1 binds to and regulates mRNAs encoding proteins that function in iron homeostasis. Work from this group and others has led to the idea that bifunctional enzymes and RNA binding proteins may represent a general mechanism of how metabolism and gene expression are coordinated through RNA/enzyme/metabolite (REM) networks [proposed by Hentze and Preiss in (104)].

Lending weight to the REM network hypothesis is a recent study showing that the enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH), by engaging or disengaging the glycolysis pathway and through fluctuations in its expression, regulates the posttranscriptional production of IFN-γ by T cells (27). GAPDH is a multifunctional enzyme that can bind a range of RNAs, including AU-rich regions in the 3’ untranslated region (UTR) of cytokine mRNAs, including IFN-γ and IL-2 (111). Activated T cells can use either OXPHOS or glycolysis to generate energy to support proliferation and survival. When T cells switch between these ATP-generating programs, as can occur with changes in nutrient availability, or costimulatory or growth factor signals, GAPDH switches from its function as a metabolic enzyme (glycolysis) to its function as an RNA binding protein controlling expression of immunomodulatory factors. When GAPDH is not engaged in glycolysis and OXPHOS is used for ATP production, it binds the 3′UTR of cytokine mRNAs, and translation of these mRNAs is dampened. Thus although OXPHOS can support T cell survival and proliferation, only aerobic glycolysis can facilitate full effector status. This regulatory mechanism provides a checkpoint to allow for uncoupling of T cell proliferation and survival from cytokine production. This is a desirable control mechanism over effector function, because T cells are required to undergo both homeostatic proliferation, when IFN-γ production is neither required nor desirable, and antigen-driven proliferation during an immune response, when effector cytokine production is essential.

The activity of GAPDH is not only controlled by pathway engagement as dictated by substrate availability but is also heavily influenced by redox balance within the cell. For example, GAPDH requires NAD⁺ for its enzymatic function, but NAD⁺ also interferes with mRNA binding, at least in vitro (112). Thus, NAD⁺ controls both enzyme activity and RNA binding in a mutually exclusive fashion. Metabolic enzymes that regulate NAD⁺/NADH balance, including LDHA and PHGDH, may also influence this process. Redox changes may also affect posttranslational modifications of GAPDH, altering its binding to mRNA, metabolites, as well as its localization within the cell (113, 114). This level of regulation between redox balance and RNA binding would be expected to occur with other RNA binding enzymes that share similar dinucleotide binding motifs with GAPDH. Although much work needs to be done to fully understand the biological importance of the interactions between enzymes, RNA, and metabolites, these observations clearly demonstrate how cofactors and substrates generally considered for their direct effects on metabolism may also coordinate metabolism with gene expression.

Technical Challenges in Studying Lymphocyte Metabolism

Although interest in studying lymphocyte metabolism and technological advances in metabolomics have increased over the past several years, the field faces many challenges going forward. For instance, although equipment for studying basic metabolism (i.e., oxygen consumption and proton production) is becoming more commonplace, specialized equipment for metabolite measurements [i.e., mass spectrometry (MS), nuclear magnetic resonance (NMR) spectroscopy] and analytical expertise is not routinely accessible to investigators. Another limitation is that, unlike microarray or sequencing technologies, metabolomic analyses do not adhere to a single global platform. For example, gas chromatography coupled to MS (GC-MS) is effective at quantifying organic acids such as TCA cycle intermediates but not most glycolytic intermediates. Thus, multiple platforms must be used to generate complete metabolite data sets. Current extraction methods do not allow for the measurement of subcellular metabolite pools, so information on metabolite localization or channeling between organelles is also lost. One caution of measuring steady-state metabolite levels is that these data provide no measurement of metabolite flux, that is, the speed or direction of metabolite flow through a given pathway (115). Metabolic flux analysis using isotopically labeled metabolites such as ¹³C-glucose or ¹³C-glutamine will be essential for delineating pathways of metabolite flow in T cells.

Perhaps the most relevant challenge for immunologists is the amount of material required for metabolomic analysis. Because of advances in flow cytometry and the identification of new biomarkers to define T cell subsets, immunology has entered an era of cellular subspecialization, where rare cell populations are readily characterized. In comparison, a typical metabolic flux experiment requires millions of cells. This raises the issue of having limited material to analyze the metabolism of cells grown in vitro, let alone in vivo. The development of instrumentation with increased sensitivity will help reduce this gap, but better sample preparation and techniques to amplify metabolite signals are badly needed. The development of genetically encoded fluorescent biosensors for metabolic activity, such as those recently developed for NADH (116) and cellular energy charge (117), will be particularly powerful for studying T cell metabolism at the single-cell level.

Microenvironmental Effects on Metabolism: Are Our Model Systems Correct?

One of the ongoing questions regarding current experimental models in immunology is how closely cell culture models recapitulate immune responses in vivo. From a metabolic perspective, one can conclude that the two systems are worlds apart. Activated T cells cultured in standard medium (Iscove’s modified Dulbecco medium plus 10% serum) experience 25 mM glucose (five times standard blood glucose of 5 mM), 4 mM glutamine (eight times the standard concentration of 0.5 mM in blood), and 20% O₂ [two to four times the oxygen tension in blood, which varies depending on tissue type (73)]. Most in vitro assays to assess T cell function are performed at nutrient and O₂ levels much higher than seen in normal physiology; these conditions model a hyperglycemic and hyperoxic environment never seen in vivo.

As T_EFF cells move from a nutrient-replete environment in the lymph node or spleen to distant sites of infection, they are likely to experience more restrictive metabolic environments (Fig. 6). Some tissue sites, such as the thymus, bone marrow, and the gastrointestinal epithelium, are naturally hypoxic, whereas inflammation can promote local hypoxic regions in tissues (73). Unlike O₂, which can freely diffuse into tissues, nutrients such as glucose move through space by Brownian motion and require transport into cells, and thus they are likely to have a much more restricted distribution in tissues. Local metabolic activity of immune cells at the site of infection can rapidly consume available O₂ and nutrients, resulting in metabolic stress for infiltrating T cell populations (Fig. 6). There is evidence that T_EFF cells in the tumor microenvironment compete with tumor cells for available glucose, and this competition model of nutrient restriction limits the ability of T_EFF cells to produce effector cytokines such as IFN-γ (27). Thus, metabolic and environmental influences on T cell function in vivo may elicit very different responses and may account for experimental variance between T cell responses in a petri dish versus what is observed in animal models. Although these points provide sobering food for thought, the development of in vitro methods that control metabolic parameters (e.g., hypoxia incubators, perfusion systems for culture medium) may help to reconcile these differences. Studying metabolite flux of T cells in vivo by using isotopomer labeling techniques will further our understanding of metabolic pathways relevant for T cell function. These techniques are currently being developed in the cancer biology field (118, 119) and should be readily transferable to immunology research.

If activated T cells frequently transition between nutrient-replete states (lymph nodes) and nutrient-deficient states (sites of infection), then management of metabolic resources is an important consideration for lymphocytes in order to ensure maintenance of proliferation and/or effector function. Metabolic insufficiency may be a fundamental mechanism by which environmental context regulates T cell function, potentially influencing T cell tolerance and anergy. Metabolic interference mechanisms, such as indoleamine 2,3-dioxygenase (IDO)–1–dependent degradation of tryptophan by antigen presenting cells (APCs), may represent key regulatory mechanisms at sites of infection or inflammation (120). Tumors may similarly restrict antitumor immunity by influencing T cell metabolism. Competition between tumor cells and tumor-infiltrating T cells for available glucose can impose nutrient deprivation on T_EFF cells that limits their ability to produce effector cytokines (27). Tumor-derived lactate can also suppress CTL function directly by blocking lactate export by T cells, thus disrupting their ability to maintain glycolysis (121).

It remains to be determined whether T cells deal with nutrient restriction by displaying metabolic plasticity similar to tumor cells. In this context, AMPK and HIF-1α, which monitor cellular energy and O₂ levels, respectively, are likely to have a profound effect on the adaptive immune response. Loss of HIF-1α impairs T_H17 expansion and induction of experimental autoimmune encephalitis (35), consistent with a requirement for HIF-1α in promoting T_H17 function in vivo. Overall, we know very little regarding pathways that regulate metabolism and adaptation to metabolic stress in lymphocytes, particularly in vivo; this area of immunology is poised for important discoveries in the years to come.