The Complex Integration of T-cell Metabolism and Immunotherapy

Immunotherapy has revolutionized clinical oncology and now provides a distinct arm of cancer treatment to empower the immune system to eliminate cancer cells. Following on early efforts to vaccinate against cancer that led to disappointing results, immunotherapy through adoptive cell therapy (ACT) or immune checkpoint blockade (ICB) now can lead to promising and durable remissions (1, 2). Although many methods for cancer immunotherapy are under study that include a variety of cell mechanisms and targets, a common approach in ACT and ICB is to engage or remove inhibition of T cells to eliminate target cancer cells. Despite early successes, however, immunotherapy has not been sufficiently effective across cancer types or for all patients, as the majority of patients fail to respond, and strategies to improve immunotherapy outcomes are needed. The basic biology of T cells and how T cells interact with and are suppressed by the tumor microenvironment (TME) may contribute to these failures of ACT and ICB. In each case, cellular metabolism offers an avenue to understand the limitations of immunotherapy and may provide rational approaches to improve antitumor T-cell activity by improving the metabolic fitness of antitumor T cells.

An essential component of an antitumor T-cell response is antigen-driven T-cell activation. In ACT, transferred cancer-specific T cells or T cells expressing a chimeric antigen receptor (CAR) are stimulated upon interaction with target cells, whereas ICB blocks inhibitory signals that prevent the activation of endogenous anticancer T cells. In each case, antigen and costimulatory signals must be present for optimal activation, proliferation, and effector function. These costimulatory signals are provided in culture for in vitro expansion and as integral components of CARs in ACT and costimulatory signals are the direct targets for ICB, as PD-1 and CTLA4 are both inhibitory receptors of the CD28 costimulatory family of receptors (3, 4). A key function of costimulatory signals is to enhance T-cell metabolism to support energetic requirements, biosynthesis, and cell-signaling processes for effector activities. CD28 promotes T-cell nutrient uptake, glycolysis, and mitochondrial fitness (5, 6), whereas PD-1 and CTLA4 each can suppress this pathway to instead reprogram T-cell metabolism away from anabolic and effector pathways (7). Here, we review how these metabolic changes influence T cell–based immunotherapy and how the TME may further suppress the metabolism of T cells. These challenges, however, also offer opportunities to improve ACT and ICB.

Resting T cells utilize an oxidative catabolic metabolism that undergoes a dramatic shift upon activation. Tonic signals from the T-cell receptor (TCR) through RasGRP and the cytokine IL7 signal to maintain a low basal activity of mTORC1 in naïve T cells. Together with sphingosine-1 phosphate that promotes mitochondrial metabolism and oxidative phosphorylation, these pathways generate ATP essential for cell survival and immune surveillance (8–10). Upon antigen encounter, the TCR activates calcium signaling that can result in tolerance or anergy if costimulatory signals through receptors such as CD28 are not received to enhance TCR signals and robustly activate the PI3K/AKT/mTORC1 and MYC signaling pathways (4, 11). Importantly, TCR signals alone in the absence of costimulation do not efficiently trigger reprogramming to support anabolic metabolism essential for cell growth and effector differentiation (5, 12), and T cells that fail or are unable to undergo metabolic reprogramming do not adopt effector functions. Conversely, T cells with excessive metabolic reprogramming can promote inflammation and autoimmunity (13, 14). An appropriate balance of costimulatory signals, therefore, is essential to regulate T-cell metabolism for efficient yet not excessive responses.

Metabolic programming for T-cell activation involves transition from catabolism to the induction of a wide range of anabolic processes that resemble the metabolic transitions long known to occur in cancer (Fig. 1). Oncogenic signals through PI3K, MYC, and others can lead to increased glucose uptake and metabolism with the production of lactate even in the presence of abundant oxygen (15). This process, termed aerobic glycolysis, is similarly induced in T cells following TCR activation with costimulation and is limiting to establish the extent of T-cell inflammatory function and proliferation (13). Serine and one carbon metabolism, as well as glutamine uptake and oxidation, are also increased to support redox balance, nucleotide synthesis, and mitochondrial metabolism (16, 17). To support the increased flux required for these pathways, T cells upregulate nutrient transporters and uptake of glucose (18) and amino acids, including glutamine and essential amino acids including methionine (19, 20). Uptake and metabolism of additional amino acids are also important and can play key metabolic and immunomodulatory roles, with arginine and tryptophan in particular playing key roles to shape inflammation and antitumor immunity (21, 22). Even uptake of the nonessential amino acid alanine is necessary to satisfy the high protein synthesis demand accompanying T-cell activation (23). Importantly, as T cells activate and are guided by inflammatory and cytokine cues to gain distinct sets of effector functions, they also develop and require specific metabolic programs (24). If these fail to establish or are blocked, T cells are unable to elicit effector functions and may instead develop into immune suppressive regulatory T cells [induced Treg (iTreg)]. Although iTreg can often rely on fatty acid oxidation (FAO), activated Treg in vivo may also utilize glycolysis during proliferation and for migration into tissues (25, 26). Treg-flexible utilization of glycolysis and fatty acid metabolism may contribute to the overall immunosuppressive nature of the TME and grant a selective advantage compared with effector T cells (27).

Transient TCR signaling from acutely cleared antigens leads to a distinct metabolic outcome compared with chronic stimulation from persistent antigens. Following acute antigen exposure and clearance, T cells that fail to reduce aerobic glycolysis undergo apoptosis, whereas those that can revert back to catabolic metabolism using lipid oxidation as a primary fuel to support mitochondrial respiratory capacity can survive as memory cells (28). The morphology of mitochondria can play a key role in electron transport efficiency, underlying mitochondrial capacity and fitness. Upon T-cell activation, mitochondria fragment and cristae remodel leading to relatively reduced oxidative phosphorylation, although overall metabolic flux is elevated compared with a nonactivated T cell (29). Transient inhibition of fragmentation during activation resulted in fused mitochondria that maintained mitochondrial oxygen consumption and supported long-term T-cell survival and memory. Importantly, this treatment increased the ability of T cells to survive and eliminate tumors upon subsequent adoptive transfer (29). A key characteristic of memory T cells is the ability to mount fast responses upon secondary antigen encounter. Metabolically, memory T cells remain poised through endoplasmic reticulum/mitochondrial associations and metabolic complexes to rapidly reactivate aerobic glycolysis to promote robust effector function and proliferation (30). Importantly, memory responses may be enhanced if anabolic cell metabolism is transiently suppressed during initial activation. Although mechanisms are unclear, it is possible that mild metabolic stress, such as low-dose 2-deoxyglucose to inhibit glycolysis, can lead to mitochondrial adaptation to prime for future respiratory capacity and survival as memory cells (31).

If antigens are not rapidly cleared, chronic antigen exposure can lead to T-cell dysfunction, often termed T-cell exhaustion. In contrast to memory T cells that are primed to quickly respond to restimulation, exhausted T cells do not produce the full array of inflammatory cytokines upon ex vivo stimulation, have reduced proliferative potential, express characteristic inhibitory coreceptors, and have epigenetic changes suggestive of terminal differentiation (32). Exhausted T cells can also have marked metabolic dysfunction. Glucose uptake and mitochondrial oxidative metabolism are decreased, yet mitochondrial mass is increased (33). Depolarized mitochondria accumulate in exhausted T cells, and pharmacologic enhancement of depolarization promotes epigenetic changes, suggesting a causal link between metabolism and the transcriptional program of exhaustion (34). Indeed, clearance of damaged and nonfunctional mitochondria to maintain a high level of mitochondrial quality control may be a key event to reinvigorate exhausted T cells and allow the generation of long-lived memory T cells (33, 35). The combination of continuous TCR signaling with hypoxia induces an exhausted program in T cells by restricting mitochondrial regulation of reactive oxygen species (36). Furthermore, the ability of exogenous pyruvate or acetate to rescue exhausted T-cell functions supports a central role for metabolic impairment in T-cell exhaustion (37, 38).

These normal biological processes are critical to maintain T-cell fitness and allow rapid and robust immunity. Because each T-cell subset employs a distinct metabolic program, however, metabolic requirements are closely tailored and linked to T-cell function and fate. Glutamine metabolism provides a striking example of selectivity and regulation of T-cell responses. Although all effector T cells require increased glutamine uptake upon activation and can instead differentiate into Treg when glutamine itself is limiting (17, 19), the further metabolism of glutamine differs between T-cell subsets. Metabolism of glutamine through glutaminase (GLS) to generate glutamate is required for CD4 Th17 cells, and inhibition or genetic deletion of GLS is sufficient to provide protection in models of airway inflammation (17), inflammatory bowel disease (17), psoriasis (39), and systemic lupus erythematosus (17, 40). Conversely, glutamine metabolism appears to restrain the differentiation and function of CD4 Th1 and CD8 cytotoxic lymphocytes (CTL), as both cell types show increased markers of activation and effector function when glutamine metabolism is inhibited (17, 41). Increased differentiation upon GLS inhibition may come at the cost of excessive or terminal fates, however, as GLS-deficient Th1 cells also upregulated inhibitory receptors and appeared to lose function over time. It remains unclear if this ultimate decrease of overall function was due to reversion to a state similar to chronic antigen stimulation or terminal differentiation and increased susceptibility to apoptosis.

Changes in cell metabolism are now known to affect epigenetic marks and histone acetylation so as to influence gene expression and cell differentiation that can last long after T cells are removed from metabolic interventions (42). This long-lasting impact of cell metabolism may affect both T cells that have left TMEs as well as adoptively transferred T cells that can be epigenetically programmed in vitro through metabolic manipulation for specific differentiation fates and to withstand metabolically hostile microenvironments. Acetylation of histones depends on acetyl donor molecule acetyl coenzyme A (acetyl-CoA) and the enzymatic activity of histone acetyl transferases. Cytoplasmic and nuclear acetyl-CoA is derived from fatty acid metabolism, tricarboxylic acid (TCA) cycle citrate exported to the cytosol and converted to acetyl-CoA by ACLY, and acetate though the enzyme ACSS2. Enhanced nucleocytosolic acetyl-CoA enhances epigenetic acetylation of effector molecule loci in T cells. Genetic ablation of lactate dehydrogenase reduced aerobic glycolysis and reduced IFNγ production and histone acetylation in Th1 cells by sequestering citrate in the TCA cycle to maintain cellular metabolism (43). Acetate metabolism promotes CD8 effector function at sites of infection, but ACSS1/2 are downregulated in memory CD8 T cells potentially to prevent further epigenetic remodeling in these stem-like cells (44). Universal methyl donor S-adenosyl-L-methionine, necessary for histone and DNA methylation, is largely derived from methionine in activated T cells. Methionine restriction reduced Th17 cell effector molecule production and attenuated experimental autoimmune encephalitis pathology (45). Conversely, inhibition of glutamine metabolism may promote hypermethylated states, as many demethylation reactions rely on the downstream glutamine metabolite α-ketoglutarate (17, 41). These epigenetic mechanisms now offer a straightforward in vitro metabolic approach to in vivo immunotherapy, as the epigenetic differentiation of effector immune cell therapy may be directly altered metabolically during ex vivo production.

In the case of how cancer immunotherapy intersects with basic T-cell immunometabolism, it is ultimately necessary to balance the generation of highly functional effector T cells that may exhaust or die with the need for long-lasting memory responses that may continue to battle within the TME. Further, those T cells may be required to enter the TME that are metabolically hostile due to poor vasculature and metabolite exchange that may lead to reduced availability of essential nutrients and accumulation of toxic waste products. Tumors can be hypoxic, high in lactate, low in essential nutrients, and poorly vascularized, all of which can impair effector T cells and favor immune-suppressive Treg (46–49). Likewise, many nutrients may be altered in spatially heterogeneous patterns across tumors in ways that affect T cells that we have yet to understand. Shared features of cancer and effector T-cell metabolism (15) can further complicate efforts to exploit T-cell metabolism as an approach to enhance immunotherapy, as any strategy to increase effector T-cell function may simultaneously benefit cancer cells. Each of these challenges can differentially affect ACT and ICB, which we explore below.

ACT includes a variety of treatments in which immune cells are infused into patients to eliminate cancer cells. These approaches include the transfer of multiple different cell types that can be stimulated, differentiated, and engineered ex vivo. T cells have represented the major application for ACT and can include expanded tumor-infiltrating lymphocytes (TIL), engineered T cells modified to express TCRs specific for tumor-specific antigen peptide/MHC complexes, or antibody-based CARs specific for cell surface tumor antigens (50). There are numerous advantages and disadvantages to different ACT strategies, which are reviewed elsewhere (50). ACT has had promising success in some tumor types, particularly with CD19-specific CAR T cells used against B-cell acute lymphoblastic leukemia and B-cell lymphomas, where more than half of patients can achieve durable remission (51). Responses in other settings, however, remain poor. To improve and broaden these responses, T-cell metabolism must be optimized to best support both robust initial and durable T-cell responses for the specific activation stimuli, target cell, and TME.

Engineered CAR T cells offer a particularly promising approach to modulate T-cell metabolism. By building specific signaling domains and properties into CARs, adoptively transferred T cells might be tailored for effector activity or long-lasting memory. Although many variations are being rapidly generated, the two FDA-approved CARs consist of an extracellular domain to target the B-cell marker CD19 that is coupled with an intracellular portion that includes signaling domains from CD3ζ and the costimulatory molecules CD28 or 4-1BB. By linking TCR signaling directly with costimulatory signaling domains, CARs can be activated to elicit effector functions without additional need for inflammatory or costimulatory cues. Downstream signaling through CD28 or 4-1BB domains is thought to mediate the primary differences between the two CAR T-cell treatments. Indeed, each signaling domain can elicit different metabolic effects (52). CAR T cells that express the CD28 signaling domain CAR can have increased glycolysis and effector responses but can be short-lived. Conversely, CAR T cells that express the 4-1BB signaling domain CAR have increased mitochondrial respiration and lipid oxidation and can be longer-lived, associated with a more central memory immunophenotype. These phenotypes are not surprising given the normal physiologic functions of these costimulatory molecules. CD28 costimulation normally promotes glycolysis and effector differentiation through activation of the PI3K/AKT/mTORC1 pathway and regulation of mitochondrial morphology and function (5, 6, 29), whereas 4-1BB may instead activate AMP-activated protein kinase and stimulate lipid oxidation and mitochondrial biogenesis and metabolism (53). 4-1BB also stimulates noncanonical NFκB signaling to promote cell survival (54). These metabolic phenotypes of CD28 and 4-1BB CARs are reminiscent of short-lived effector T cells and long-lived memory T cells, respectively. Given this and detailed understanding of the mechanisms of CD28 and 4-1BB signaling, a promising approach to increase CAR T-cell effector function or survival has been to introduce signaling mutations. Mutation of the YMNM signaling motif of CD28, for example, was recently shown to increase CAR T-cell survival and reduce signs of T-cell exhaustion to enhance durable tumor control in preclinical models (55). The metabolic implications of this mutation are unclear, but AKT phosphorylation was maintained while MAPK and calcium signaling were reduced, suggesting a shift in metabolic and mitochondrial regulation.

Other costimulatory molecules modulate T-cell metabolism and may be incorporated into CAR constructs or targeted with agonist antibodies. ICOS, a member of the CD28 family, similarly promotes glycolysis and mTORC1 activity in T follicular helper cells (56). GITR agonism increases CD8 TIL metabolism (57). OX40 was expressed more highly in Treg TIL than in CD4 Tconv TIL and was associated with enrichment of glycolysis and lipid metabolism transcripts. Agonism of OX40 enhanced lipid uptake in these Tregs (27).

In addition to CAR, T cells harvested from tumors and in vitro expanded or transduced with engineered TCR can also be subject to metabolic manipulations to optimize function. These approaches are generally directed toward solid tumors, take advantage of natural or synthetically derived specificity against tumors, and have both advantages and challenges. Although tumor specific, T cells obtained from patient tumors are likely already epigenetically programmed, and any metabolic manipulation will need to overcome this barrier. Engineered TCR can start from a naïve T cell similar to CAR but must differentiate to gain sufficient effector functions and longevity to eliminate tumors similar to CAR. The in vitro stimulation period for each does offer, however, the opportunity to adapt nutrient conditions or T-cell metabolism to favor appropriate effort and longevity states. In particular, T cells with low rates of glycolysis or mitochondrial potential may be generated or selected to have greater longevity while retaining effector functions (58). Similarly, targeting other pathways, such as glutamine metabolism, may also promote increased T-cell function in these settings (41).

A significant advantage ACT has over ICB in metabolic approaches to immunotherapy is that treatments can be carried out during the in vitro stimulation and T-cell engineering period rather than in situ. This approach allows the opportunity to modify T-cell metabolism and mitochondria independent of effects on other cells and tissues and thus prevents any potential benefit of a metabolic therapy meant to boost T cells from also promoting the metabolism of cancer cells. This approach also eliminates concern for potential toxic side effects of targeting metabolism in other tissues. The benefit, however, is predicated on the capacity of transient metabolic modifications during an in vitro period to lead to lasting effects on T-cell differentiation, function, and fate in vivo to allow tumor elimination. This cell-intrinsic stability is inherent in genetically engineered T cells, but it is also now clear that transient pharmacologic approaches may shift T-cell metabolism to favor long-lived anticancer effector T cells for ACT.

One metabolic strategy that has proven promising to modify T-cell differentiation is to selectively target the high rates of glycolysis that accompany effector T cells and that can impair the establishment of long-lasting memory. Glycolysis can be targeted directly with pharmacologic inhibitors, such as 2-deoxyglucose. This treatment of T cells stimulated in vitro prior to ACT in preclinical models reduced glycolysis and proliferation, but also led to a more memory-like phenotype and greater cell survival in vivo (31). As a result, 2-deoxyglucose–treated T cells were better able to control tumors and extend animal life. Restricted glucose during T-cell activation can also prime subsequent glucose uptake and shunting into the pentose phosphate pathway for nucleotide synthesis and altered redox balance, resulting in superior tumor control (59). The PI3K/AKT/mTORC1 pathway is a critical mechanism by which T cells induce aerobic glycolysis downstream of CD28 costimulation, and, similar to 2-deoxyglucose, inhibition of AKT reduced glucose uptake and glycolysis and led to a more central memory phenotype. Importantly, this also improved the ability of in vitro inhibitor-treated CAR T cells to eliminate B-cell leukemia following in vivo transfer (60).

Glutamine metabolism is generally coupled to aerobic glycolysis to provide anaplerotic substrates to maintain the TCA cycle. In CD4 Th1 and CD8 CTL, however, glucose and glutamine metabolism can become uncoupled, and inhibition of glutamine metabolism can lead to an adaptive increase in glycolysis that drives T-cell proliferation and effector differentiation (17, 41). Both GLS and broader glutamine metabolism inhibitors, therefore, can enhance the functionality of CD4 Th1 and CD8 CTL cells. In a CAR T-cell model, this led to increased proliferation, ability to kill target cells, and rapid cell accumulation in vivo. A potential challenge to this approach to modulate glucose metabolism by targeting glutamine is that enhanced effector differentiation may be self-limiting. Similar to the fate of normal effector T cells, GLS-deficient T cells also induced inhibitory receptors and markers of terminal differentiation over time. Ultimately, permanent genetic loss of GLS led to a sharp loss of cytotoxic function, and B-cell targets eventually accumulated to normal numbers (17). Nevertheless, transient in vitro inhibition was sufficient to induce epigenetic alterations that led to long-lasting changes in tumor-specific T cells, and timed or transient inhibition of glutamine metabolism may provide a means to promote effector function while not driving T cells to exhaustion or terminal fates. Inhibition of glutamine metabolism may also be promising to reinvigorate TIL in situ, as tumor CD8 T cells upregulated acetate metabolism to support anabolism (38, 44), in contrast to cancer cells which did not alter acetate metabolism and became catabolic.

Ultimately, memory or stem-like phenotypes are characterized by mitochondrial oxidative metabolism, and strategies to mimic this state may best support long-term T-cell survival and ability to eliminate cancer targets in ACT. This may be a key mechanism by which 4-1BB domain–containing CARs support ACT (52), and approaches to directly modify mitochondria may also prove valuable. The morphology of mitochondria can play a key role in electron transport efficiency. Upon T-cell activation, mitochondria fragment and cristae remodel, leading to reduced oxidative phosphorylation (29). Transient inhibition of this fragmentation during activation, however, resulted in fused mitochondria that maintained mitochondrial oxygen consumption and supported long-term T-cell survival and memory. Importantly, this treatment increased the ability of T cells to survive and eliminate tumors upon subsequent adoptive transfer (29). CRISPR-based gene knockout screen is a powerful tool to assess the role of metabolic genes in T cells in different contexts of inflammation and immunity. REGNASE-1 was discovered using this approach as a negative regulator of BATF, mediating mitochondrial function and morphology (61). Another strategy to rewire mitochondria in ACT is to directly promote mitochondrial lipid metabolism characteristic of memory T cells. Expression of the transcriptional coactivator PGC1α or treatment with the PPAR agonists bezafibrate and fenofibrate both increased lipid oxidation and increased T-cell functionality and ability to eliminate tumors in ACT (62–64).

A fundamental difference between ACT and ICB is that although the challenge in ACT is to direct immunity against tumors using cells delivered as a pharmaceutical, the goal of ICB is to overcome inhibitory signals to activate endogenous antitumor specific T cells. ICB is also most commonly directed against solid tumors, where the TME can exert a strong influence on T-cell metabolism. Beyond the initial barrier that T cells must first infiltrate into tumors, ICB must surmount several key challenges. TILs may have an exhausted differentiation state, with high expression of inhibitory receptors and altered epigenetic marks that impede effector function (65). These changes are accompanied by diverse metabolic adaptations that lead to multiple deficits. Translocation of GLUT1 to the cell surface to support glucose uptake can be impaired, and the glycolytic enzymes GAPDH and Enolase can be downregulated or suppressed (37, 66). Mitochondria can also be dysregulated, with fragmented morphology, high production of reactive oxygen species, and poor capacity for respiration (37, 62). Rescue of TIL metabolism, however, with expression of PCK to promote gluconeogenesis and the replacement of intracellular glycolytic intermediates or by treatment with pyruvate or acetate to directly support mitochondrial metabolism can improve T-cell inflammatory function (37, 38, 48). These findings demonstrate that the lack of metabolic flux directly contributes to T-cell impairments in the TME.

One mechanism by which ICB may act to restore T-cell function is to enhance T-cell metabolic capacity to rescue T cells from metabolic defects imposed by the TME. Another is to promote differentiation of PD-1⁺ TCF1⁺ stem-like progenitor exhausted T cells into effector T cells, which may occur in secondary lymphoid tissue, tertiary lymphoid structures, or the tumor itself (67, 68). In either context, disrupting coinhibitory signals promotes the generation of effector cells. In addition to chronic TCR signaling, the TME can express high levels of immunosuppressive ligands, including PD-L1 or PD-L2. A normal function of both PD-1 and CTLA4 is to suppress PI3K and AKT signaling and thus oppose a key signaling mechanism of CD28 family–mediated costimulation (3). PD-1 can recruit SHP1/2 phosphatase to prevent PI3K activation, whereas CTLA4 can both inhibit the interaction of CD28 with the ligands B7-1 and B7-2 as well as recruit the phosphatase PP2a. It is not surprising, therefore, that PD-1 and CTLA4 inhibit CD28-mediated metabolic reprogramming and the induction of aerobic glycolysis that otherwise characterizes effector T cells (5, 7).

Distinguishing PD-1 and CTLA4, CTLA4 is thought to be more important for initial phases of T-cell activation and thereby promote a naïve metabolic program. PD-1, in contrast, is expressed after TCR engagement. PD-1 signals restrain glucose uptake and glycolysis in T-cell activation and instead promote lipid oxidation that is characteristic of long-lived quiescent cells. Intriguingly, both anti-CTLA4 and anti–PD-1 can promote the expansion of new antitumor clones. Overall, inhibitory signals can prevent inappropriate T-cell activation or excessive inflammation. In support of this model, stimulated PD-1–deficient T cells can maintain greater metabolic activity in chronic infection models (33, 69). By blocking the PD-1– and CTLA4-mediated physiologic signals that dampen effector T-cell metabolism, ICB can allow T-cell stimulation and increased metabolism. ICB treatments that inhibit ligand interactions of PD-1 or CTLA4, therefore, relieve inhibition of PI3K/AKT/mTORC1 signaling to allow T cells to adopt an effector-like metabolism. In addition, as a metabolic shift may lead to epigenetic modifications, successful ICB may metabolically reprogram epigenetic marks to favor effector functions or cell longevity, although these changes remain to be fully understood.

In addition to PD-1 and CTLA4, which have been most extensively pursued as targets for ICB, other coinhibitory molecules also modulate T-cell metabolism. Indeed, of the various ICB targets that provide or inhibit costimulatory signals to T cells, many have been shown to also influence cell metabolism, and this may be a general mechanism by which costimulation promotes or suppresses T-cell function. TIM3 negatively regulates glycolysis and GLUT1 expression, and may also inhibit glutaminolysis (70). LAG3 negatively regulates mitochondrial metabolism (71). Likewise, targeting 4-1BB and OX40 may enhance T-cell mitochondrial metabolism to promote effector function and long-term survival (53, 72). Each of these metabolic regulatory functions may both directly affect biosynthetic processes and influence epigenetic marks for long-lasting influence over T-cell differentiation and fate. Other family members and how these various T cell–modulating receptors may ultimately affect antitumor T-cell metabolism and functions, however, remain to be established.

Regardless of the presence and function of costimulatory or inhibitory molecules, T cells in the TME must still access adequate essential nutrients to support metabolic demands. Various nutrients, however, may be limited or heterogeneous in abundance. In addition to limited oxygen in regions of hypoxia, glucose and other nutrients essential for efficient effector T-cell responses may be depleted in the TME (48, 49). Because cancer cells themselves can use an aerobic glycolysis metabolism that is similar to activated effector T cells and therefore can have similar metabolic requirements (15), competition may arise for these nutrients in settings where vascular exchange is limited. This potential for nutrient competition suggests that the metabolism of the TME itself may be directly immunosuppressive. Indeed, changes in tumor glucose metabolism can influence the ability of T cells to eliminate tumors, and increased tumor cell glycolysis can reduce antitumor immunity (49). The generality of such nutrient competition is uncertain, however, as measurements of metabolites in tumor interstitial fluid in some tumor settings have found that glucose and glutamine can be at normal or at only modestly decreased levels (37, 73). Nevertheless, T cells from these same tumors show metabolic impairments (37), suggesting that chronic antigen stimulation and immune inhibitory receptors, such as PD-1, contribute and may be the primary drive for many of these metabolic phenotypes. An unresolved question relevant to potential metabolic competition and tumors is to what extent T cells in the TME may be capable of reactivation with ICB, or do ICB therapies instead engage T cells that have yet to experience prolonged engagement with tumors and tumor antigens? This is an important question that will influence understanding of the metabolic implications of the TME on T cells as, despite these many barriers, ICB does successfully eliminate many tumors.

Optimizing T-cell metabolism is a promising avenue of synergy with cancer immunotherapy. Normal T-cell metabolic physiology can identify key transition and regulatory points that are altered in cancer and provide opportunities in ACT and ICB cancer immunotherapies. Metabolism underlies all cellular functions and may provide approaches to increase stemness and long-term memory, enhance effector function, and reduce exhaustion in T cells through diverse mechanisms. Organism-level metabolism is another important avenue of investigation, as obesity causes dysfunctional TIL partly through differential responses to lipids (74). The paradigm of T-cell stemness, supported by mitochondrial oxidative phosphorylation, has proved advantageous for ACT, whereas reversing any component of the global metabolic dysfunction of exhausted T cells in the TME may be a viable strategy. Interrogation of T-cell metabolism alone is insufficient for studying immunotherapy, as differences between T-cell and cancer cell metabolism may highlight new approaches to specifically target tumor cell subsets. It is important for future therapeutics targeting tumor metabolism to address their effect on cancer cells, T cells, and other tumor and immune cell subsets.

Despite and as a consequence of this progress, several questions have emerged for the field. The distinct microenvironments of tumor-specific T cells, either in the TME or in lymphoid tissue, do not adequately explain how ICB exerts activity despite potentially metabolic constraints in tumors. Terminal differentiation and T-cell dysfunction are intimately tied to epigenetic alterations. It remains an ongoing question if “terminal” differentiation and its associated epigenetics can be reversed, and whether cellular metabolism may be the lynchpin underlying effector function rescue. It also remains largely untested and uncertain which nutrients are actually available for T cells and other immune cells in the TME that may support cell functions and these epigenetic regulations. In this sense, our understanding of the TME is rudimentary at best, and a systems biology and spatial view will be essential. A fundamental question in immunotherapy for both ACT and ICB is what target is best suited to what indication? Ultimately, this will depend on gaining a better understanding of how each costimulatory or inhibitor molecule affects T-cell physiology and gene expression, and placing this in the context of the specific tumor setting and microenvironment, including available nutrients and heterogeneity of stromal, cancer, and other immune cells. As is the case for all exciting areas of research, one question begets many more as the field continues to advance and improve cancer therapy.

M.Z. Madden reports grants from NIH/NCI during the conduct of the study. J.C. Rathmell reports personal fees from Sitryx, Caribou, Nirogy, Merck, and Pfizer and grants from Incyte, Kadmon, Calithera, and Tempest outside the submitted work. No other disclosures were reported.

The authors thank members of the Rathmell lab for support. Particular thanks go to Brad Reinfeld and Jackie Bader. This work was supported by F30 CA239367 (to M.Z. Madden) and R01 CA217987 and R01 DK105550 (to J.C. Rathmell).