Metabolic dialogues: regulators of chimeric antigen receptor T cell function in the tumor microenvironment

Tumor‐infiltrating lymphocytes (TILs) and chimeric antigen receptor (CAR) T cells have demonstrated remarkable success in the treatment of relapsed/refractory melanoma and hematological malignancies, respectively. These treatments have marked a pivotal shift in cancer management. However, as “living drugs,” their effectiveness is dependent on their ability to proliferate and persist in patients. Recent studies indicate that the mechanisms regulating these crucial functions, as well as the T cell's differentiation state, are conditioned by metabolic shifts and the distinct utilization of metabolic pathways. These metabolic shifts, conditioned by nutrient availability as well as cell surface expression of metabolite transporters, are coupled to signaling pathways and the epigenetic landscape of the cell, modulating transcriptional, translational, and post‐translational profiles. In this review, we discuss the processes underlying the metabolic remodeling of activated T cells, the impact of a tumor metabolic environment on T cell function, and potential metabolic‐based strategies to enhance T cell immunotherapy.


Introduction
Our ability to optimally respond to pathogens, toxins, and foreign substances is largely controlled by the activation of T lymphocytes.T cells have also emerged as a potent tool in the treatment of cancer, providing robust anti-tumor activity.However, the potential of a T cell to mount an immune response depends on an increase in its energetic state.Intracellular ATP is generated via the transport of a wide range of fuels from the extracellular space and these fuels are then catabolized through processes such as glycolysis, oxidative phosphorylation, and fatty acid oxidation, among others.Indeed, the energetic profile of the cell is intricately related to its ability to harness these resources, providing ATP for the cell [1][2][3][4].These fuels also provide biosynthetic intermediates supporting nucleic acid, amino acid, and phospholipid synthesis as well as the production of reducing equivalents to maintain the cell's redox state [1,5,6].Moreover, metabolic pathway utilization governs the cell's epigenetic landscape, both directly, through the production of intermediates like acetyl-CoA that contribute to histone acetylation, and indirectly, by generating metabolites that regulate enzymes such as demethylases [7,8].In this manner, fuel uptake and utilization govern the differentiation and function of a T lymphocyte, resulting in a wide range of immune responses.Here, we review the impact of metabolic pathways on T cell differentiation and focus on the role of metabolite shifts in the tumor microenvironment, conditioning the function and persistence of anti-tumor T cells.

Metabolic regulation of na€ ıve T cells
Naive T cells (T N ) are continuously patrolling through the peripheral blood and secondary lymphoid tissues to achieve immune surveillance.The size of the T N pool in the peripheral lymphoid compartment is tightly regulated and its maintenance is driven by homeostatic cytokines, in particular IL-7, as well as T cell receptor (TCR)-mediated tonic signals [9].The metabolism of T N is characterized by basal levels of ATP synthesis, necessary for the cytoskeletal remodeling associated with migration, and low levels of macromolecule biosynthesis that maintain cellular homeostasis.IL-7 and low-affinity TCR-major histocompatibility complex (MHC)-peptide interactions maintain metabolic activity in T N , at least in part through glucose uptake via low levels of the GLUT1/SCL2A1 transporter [10][11][12][13].Overall, T N metabolism relies mainly on mitochondrial oxidative phosphorylation (OXPHOS), fueled by glucose-derived pyruvate as well as fatty acid oxidation.However, metabolic pathways in T N can be quickly induced, due to the presence of a high level of untranslated mRNAs encoding metabolic enzymes and transporters as well as a rapid turnover of the transcription factors that regulate quiescence [14,15].Indeed, the high number of idling ribosomes in T N provides the capacity to rapidly translate these mRNAs and remodel T cell metabolism within minutes following the engagement of TCR and CD28 costimulatory pathways, enabling a rapid switch to an effector program (T EFF ).The activation of the PI3K-AKT-mammalian target of rapamycin (mTOR) axis upregulates cell surface nutrient transporter expression and subsequent nutrient uptake.This leads to a massive increase in both aerobic glycolysis and OXPHOS, albeit with a bias towards the former [16][17][18][19][20]. Indeed, increased OXPHOS is required for the exit of T N from quiescence and this process is regulated by mTORC1-induced mitochondrial biogenesis [21,22].Upon TCR-engagement, the mitochondrial mass polarizes at the immune synapse (IS), enabling efficient calcium buffering and local ATP production that sustain IS formation and migration [23,24].Moreover, mitochondrial reactive oxygen species (ROS) generated by the electron transport chain (ETC) complex III activate NFAT signaling and IL-2 induction after antigen encounter, contributing to T cell activation [25].The higher energetic state, resulting from a local increase of ATP production, sustains a massive increase in protein synthesis and a concurrent remodeling of the proteome through the non-redundant activities of mTORC1, the MYC transcription factor, and translation initial factors [15,[26][27][28][29]. Importantly though, there is an intricate crosstalk between OXPHOS and glycolysis.mTORC1/MYC signaling also increases glycolysis, promoting the production of biosynthetic intermediates that are critical for T cell proliferation and differentiation.

Metabolic programs associated with T cell differentiation
Following activation of CD4 + T N , the recruitment of different cytokine-induced signaling cascades and transcription networks results in the differentiation of T helper cells (T H ) with distinct phenotypes and properties.Of note, the polarization of CD4 + T cells to a T H effector versus suppressor fate is tightly coordinated with the establishment of different metabolic programs.T H 1, T H 2, and T H 17 effector cells are highly glycolytic while suppressive regulatory T cells (T REG ) rely mostly on lipid oxidation [30,31].In accord with the high level of IFNc produced by T H 1 effectors, regulated by the master T H 1 transcription factor, Tbet, both Tbet and IFNc expression are coupled to a glycolytic program [32,33].Glutaminolysis and the anaplerotic production of alpha-ketoglutarate (aKG), feeding the tricarboxylic acid cycle (TCA), is also critical for T cell polarization.The absence of glutamine-derived aKG inhibits T H 1 polarization and biases cells towards a T REG fate [34,35].Conversely, high levels of exogenous aKG inhibit T REG polarization, resulting in an inflammatory T H 1-like phenotype [36].Nonetheless, the implication of glutaminolysis in T cell differentiation is complex as glutaminase-1 (GLS1)-deficient CD4 + T cells initially exhibit elevated T H 1 function but these cells become exhausted over time [37].
These data highlight the importance of metabolic shifts in controlling the balance between T H 1 and T REG differentiation, but it is also important to note that metabolic programs regulate the lineage fate of T H 17 cells as compared to T REG .De novo fatty acid synthesis is required for T H 17 while T REG acquire the ability to take up exogenous fatty acids that are used in fatty acid oxidation (FAO) [38].T H 17 and T REG also diverge with regards to their utilization of other nutrients and their dependence on metabolic programs.Specifically, T REG cells can flourish in low glucose environments and indeed, expression of the T REG transcription factor Foxp3 inhibits glycolysis [39].In contrast, the hypoxia-inducible factor HIF1a promotes T H 17 differentiation over T REG , at least in part via an enhanced glycolytic program mediated by the upregulation of GLUT1 and other glycolytic enzymes [40,41].Arginine-based polyamine metabolism varies between T H 17 and T REG .T H 17 cells are characterized by significantly higher levels of polyamines than T REG cells.Furthermore, inhibition of polyamine generation, leading to a remodeling of the transcriptome and epigenome, skews pathogenic T H 17 cells to a T REG -like fate [42,43].Higher levels of glutamate transamination also occur in T H 17 compared to T REG , leading to the accumulation of aKG and 2-hydroxyglutarate (2-HG) as well as methylation of the Foxp3 locus.In line with these observations, inhibition of the transaminase GOT1 is associated with a block in T H 17 differentiation and the reprogramming of these cells to a T REG fate [44].Thus, the crosstalk between metabolic pathways, epigenetic modifications, and transcription factor expression plays a crucial role in regulating Th cell plasticity.

Metabolic specificities of memory T cells
When compared to T EFF cells, memory T cells (T MEM ) display a more quiescent but metabolically primed metabolism, dependent on fatty acid oxidation, OXPHOS, and autophagy [45].Mitochondrial remodeling participates in this process.The acquisition of elongated-fused mitochondria with a tight cristae organization is associated with a higher mitochondrial reserve capacity and optimized efficiency of the ETC, creating a bioenergetic advantage for rapid recall [46][47][48].Also, T MEM maintains a higher level of translational activity compared to T N cells with a two fold higher turnover of ribosomal and glycolytic enzymes, promoting a state of readiness that optimizes the recall response [15].This metabolic divergence between memory and effector subsets is established within the early phases of T cell activation through an asymmetric sorting of the transcription factor MYC and amino acid transportersbefore the first cell division.T cells with high MYC/high levels of transporters are capable of mounting a robust primary response while conversely, cells with low MYC/low levels of transporters support the generation of a memory response [49,50].Importantly, mTORC1 activity is also asymmetrically inherited after the first cell division through a mechanism notably dependent on amino acid influx.Notably, increasing this metabolism-coupled asymmetric cell division in aged mice restores their potential to develop memory T cells.Thus, metabolic remodeling occurs throughout all stages of CD4 + /CD8 + T cell activation and differentiation.

Metabolic rewiring in exhausted T cells
Finally, the acquisition of an exhausted T cell phenotype is linked to metabolic rewiring.Importantly, repeat stimulations of the TCR within a given T cell, such as that occurring in the context of a chronic viral infection or tumor, lead to a dysfunctional T cell state commonly referred to as exhaustion [51][52][53].In these contexts of chronic stimulation, T cells upregulate inhibitory molecules such as PD-1 and CTLA-4, and their engagement, either on anti-tumor T cells or T REG , results in changes in FAO and glycolysis [54][55][56][57].Moreover, high levels of lactic acid in glycolytic tumors result in the upregulation of PD-1 on T REG [55,[58][59][60].Furthermore, mitochondrial biogenesis is negatively impacted through a reduction in the expression of the transcriptional coactivator PGC1a [58,61].Indeed, mitochondrial dysfunction is a hallmark of TILs; Intra-tumoral T cells display a reduction in mitophagy and accumulate depolarized mitochondria and mitochondrial ROS, thereby accelerating their differentiation through a terminally exhausted phenotype [62][63][64][65].Thus, overexpression of PGC1a may represent a therapeutic strategy via which CAR T cell exhaustion can be attenuated [61].More generally, modulating the expression of candidate genes that impact metabolic pathways presents an area of research that holds much promise for improving the fitness of CAR T cells.Some of the genes that are being targeted in CAR T cells are presented in Section 5.2.

Crosstalk between T cell metabolic pathways and function: Signaling, epigenetic modifications, and post-transcriptional regulation
Over the last decade, it has been increasingly recognized that T cell metabolism is not only a downstream event of TCR and cytokine signaling pathways but also impacts T cell function, at levels beyond energy production and building blocks for biosynthesis.Indeed, increases in intracellular metabolites, via nutrient transporter upregulation as well as de novo biosynthesis, regulate signaling changes that are intricately connected to metabolic programs that differ across T cell subsets (Fig. 1).One critical pathway that integrates extracellular signals with the metabolic environment of the cell is mTOR, a serine-threonine kinase.mTOR signaling is mediated via mTORC1 and mTORC2 complexes, tightly regulated by the intracellular concentration of amino acids.mTORC1 activity is activated by a multi-step approach.Firstly, amino acids inhibit negative regulators of mTOR, including Sestrins, CASTOR, and SAMTOR.Secondly, amino acids activate mTORC1 on lysosomes via Rag and Rheb GTPases [66][67][68][69].mTORC2 is also activated by growth factors but the specific pathways are less welldefined [70].mTORC1 activity is required for both T H 1 and T H 17 differentiation while mTORC2 activity is a prerequisite for T H 2 differentiation [4,71,72].Furthermore, mTORC1 activation inhibits T REG differentiation and function [73][74][75][76][77][78].Consistent with the critical role of glucose and amino acids in positively regulating mTORC1 activity, conditions that limit these metabolite levels block effector T cell generation but not T REG differentiation [35,[79][80][81][82].
T cell fate is modulated by both "top-down" and "bottom-up" metabolic signaling cascades, as proposed by Shyer et al. [83] TCR and extracellular signals are responsible for metabolic adaptations within the cell but metabolic changes also govern the potential of the T cell to respond to these engaged signaling networks.Moreover, metabolic signals act directly at the level of the TCR itself.MGAT5, an enzyme in the N-glycosylation pathway, restricts TCR recruitment while conversely, its absence decreases the TCR activating threshold resulting in autoimmunity [84,85].In the context of "bottom-up" signaling, metabolic intermediates and enzymes play roles outside of their canonical metabolic pathways.Indeed, Ho et al.
elegantly showed that TCR-induced increases in the glycolytic intermediate phosphoenolpyruvate (PEP) are not just important for glycolysis.PEP also sustains calcium release from the endoplasmic reticulum by directly blocking SERCA ATPase, thereby promoting NFAT signaling [86].
With regards to enzymes involved in glycolysis, glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and lactate dehydrogenase A (LDHA) exhibit important moonlighting activities in T cells, binding to the 3 0 UTR mRNA of IFNc and repressing its translation.The repressive RNA-binding protein activity of GAPDH and LDHA is released when these enzymes are engaged in glycolysis, promoting the translation of IFNc in a manner that couples glycolytic activity with cytokine release [32,87].Moreover, the polyaminederived amino acid hypusine, whose generation is generally dependent on arginine catabolism, regulates key effector functions in T cells at the translational level.Hypusination of eIF5A promotes translation of the transcription factor Tbet as well as IFNc and IFNa among others [27,[88][89][90].
decreases histone acetylation, thereby attenuating the T H 17-associated transcriptional program [91,92].Conversely, the attenuation of mitochondrial pyruvate carrier 1 (MPC1) activity leads to an inhibition of pyruvate entry into the mitochondria.This results in an increased intracellular pool of acetyl-CoA and subsequent histone acetylation, driving differentiation to a memory-like phenotype [93].Moreover, histone methylation, requiring the transfer of 3 methyl groups onto arginine and lysine histone residues, plays a critical role in T cell fate [94,95].The main source of these methyl groups is S-adenosylmethionine (SAM), derived from methionine and one-carbon metabolism.In this regard, it is interesting to note that the high expression of the methionine transporter SLC43A2 in certain tumors reduces methionine availability for T cells, thereby attenuating T cell cytotoxicity [96].
Loss of a mitochondrial enzyme, isocitrate dehydrogenase 2 (IDH2), results in a decrease in the reductive carboxylation of glutamine, which in turn attenuates the activity of the KDM5 histone demethylase and promotes T cell memory formation [97].Indeed, Jumonji C-domain lysine demethylases (JmjC-KDMs), ten-eleven translocation (TET) DNA cytosineoxidizing enzymes, and prolyl hydroxylases (PHDs) are all regulated by a large array of metabolism-related factorsincluding a-ketoglutarate and succinate TCA cycle metabolites, 2-HG, iron, ascorbate, and oxygen levelsunderscoring the intricate regulation of T cell fate by its metabolic environment [98].This multi-level crosstalk, together with the differential utilization of metabolic pathways in T cell subsets, highlights the large potential for metabolic interventions to improve T cell immunotherapy.However, it also exposes the potential complexity of these types of strategies.

Nutrient uptake through metabolite transporters governs the metabolic fitness of T lymphocytes
Metabolite transporters mediate the uptake of key nutrients including amino acids, sugars, nucleosides, iron, and fatty acids, thereby playing a critical role at the interface of the metabolic environment and intracellular metabolic pathways (Fig. 1).In the context of cytotoxic T cells, proteomic studies have shown that the most quantitatively prominent proteins in cytotoxic T cells include metabolic regulators, with at least 72 metabolite transporters [99,100].These transporters belong to the SLC (solute carrier) superfamily of multi-membrane spanning proteins, containing more than 440 recognized members that mediate the intracellular entry of a large number of substrates, with variable levels of specificity [100].Importantly though, the range and density of metabolite transporters that are expressed on T cells vary substantially as a function of the cell's activation state [99].
The study of metabolite transporters has been restricted by the lack of reliable antibodies against their extracellular domains, likely because of their low immunogenicity and their high conservation during evolution [101].An innovative tool set has been developed to overcome this limitation, taking advantage of the discovery that gamma-like envelopes of retroviruses and endogenous retroviral sequences use metabolite transporters as receptors, [12,[101][102][103][104].Using receptor binding domain (RBD) fusion proteins derived from retroviral envelopes, the cell surface expression of GLUT1/SLC2A1 and ASCT2/SLC1A5 glucose and glutamine transporters, respectively, have been shown to be rapidly increased at the surface of CD4 + T cells, within 2 h of TCR stimulation [19].Furthermore, these tools have revealed the potential of T cells to engage compensatory mechanisms.Both GLUT1 and ASCT2 surface expression are significantly upregulated in response to glucose deprivation on na€ ıve T cells, potentially explaining recent data showing that transient glucose deprivation can prime T cells and increase their anti-tumor activity following adoptive transfer into tumor-bearing mice [19,105].
As indicated above, nutrient uptake in T lymphocytes is mediated by several transporters with a certain level of redundancy.Nonetheless, it is important to note that the deletion of a single transporter can lead to important functional consequences and these consequences can differ between T cell subsets.Ablation of the glucose transporter Glut1 in murine T cells has been found to attenuate CD4 + effector T cell function as well as CD4 + -mediated inflammatory responses.However, it does not appear to be required for CD8 + or T REG function, likely due to a compensatory activity of the Glut3 glucose transporter [79].Indeed, Glut3 ablation in murine T cells results in a different phenotype, negatively impacting T H 17 effector function and T H 17-mediated autoimmunity through a reduction in glucose-derived acetyl-CoA and disturbed epigenetic regulation of inflammatory genes [92].Moreover, deficiency of the glutamine transporter Asct2/Slc5a1 impairs T H 1 and T H 17 differentiation and decreases the generation of CD4 + memory populations in older mice, but notably, T REG generation and function are not impacted [82].Consistent with these data, glutamine deprivation significantly attenuates T H 1 but not T REG differentiation [34,35].In contrast, inhibition of the large neutral amino acid transporter Slc7a5/Lat1 has a more marked effect.Namely, it abrogates antigen-induced effector T cell proliferation and function, most likely due to the loss of leucine-mediated mTORC1 complex activation and Myc expression [81].
These examples highlight the importance of metabolite transporters in T cell effector generation and responsiveness to antigens.Nonetheless, recent work has unveiled the potential for transporter downregulation to improve the potential for T cell differentiation and function.An in vivo CRISPR screen, performed in the context of lymphocytic choriomeningitis mammarenavirus (LCMV) infection in mice, found that the Slc7a1/Cat1 cationic amino acid transporter and the Slc38a2/Snat2 neutral amino acid transporter both negatively regulate CD8 + memory T cell generation.Indeed, deletion of either of these transporters modulates mTORC1 signaling, improving the persistence and in vivo killing capacity of CD8 + memory T cells [106].Moreover, deletion of the lactate transporter Slc16a1/Mct1 in T REG attenuates their function, thereby decreasing tumor growth and enhancing the responsiveness of effector T cells to immune checkpoint blockade [107].Finally, strategies that increase surface levels of metabolite transporters on cytotoxic T cells point to complex and intricate connections that result in the fine-tuning of effector functions.For example, high levels of GLUT1 expression are associated with enhanced IFNc and IL-17 production but may decrease long-term persistence of T cells engineered to express an anti-tumor chimeric antigen receptor (CAR) [108].Together, these studies show that multiple metabolite transporters serve as key drivers of T cell metabolic fitness, making them promising targets in the immunotherapy arsenal.

Immunosuppressive metabolic environments
The tumor microenvironment (TME) is characterized by distinctive metabolic features such as changes in the concentrations of specific nutrients and adjustments in metabolic scavenging systems.Within necrotic tumor core tissue, poorly vascularized regions, and areas of high tumor growth, T cells compete for limited oxygen and nutrient supplies [109].Moreover, the metabolic parameters of the TME display a high degree of heterogeneity and compartmentalization, increasing the complexity for immune cells to adapt within these different metabolic niches.Globally though, the TME presents a modified metabolic environment that results in compromised immunosurveillance [110].As discussed in Section 2, an effective anti-tumor T cell response is dependent on the T cell's ability to increase nutrient uptake but within a tumor, the conditions resulting in nutrient deprivation are likely to disrupt effector T cell differentiation and function.Indeed, T cells isolated from patients with renal cell carcinoma as well as chronic lymphocytic leukemia (CLL) were found to exhibit suboptimal metabolic profiles with attenuated glycolytic activity and mitochondrial dysregulation [111][112][113].This metabolic dysfunction is intimately associated with T cell exhaustion and impairment of memory formation, both of which represent significant barriers to successful T cell immunotherapy (Fig. 2).
Several murine models have been harnessed to dissect the mechanisms via which the competition for nutrient resources impacts anti-tumor immunity.Many studies have focused on glucose availability and indeed, in a mouse sarcoma model, limited glucose resources has been found to be a major factor associated with impaired tumor-infiltrating T cell responses [54].Glucose availability can also be limited by nontumor cells.In a murine colorectal cancer model, tumor-infiltrating myeloid cells were found to consume the highest amounts of glucose, potentially limiting glucose uptake by T cells [114].In this regard, it is interesting to note that strategies that increase glucose uptake by T cells may have clinical relevance.Overexpression of Glut1 was found to restore glycolytic activity in T cells in a murine leukemia model [113].Moreover, checkpoint blockade antibodies against CTLA-4, PD-1, and PD-L1 appear to improve T cell cytotoxicity and IFNc production, at least in part through increases in glycolysis in syngeneic leukemia and solid tumor models [54].
As described in Section 2.2, glycolysis-derived metabolites and enzymes play a crucial role in antitumor T cell cytotoxicity.Notably though, the impaired glycolytic activity in tumor-infiltrating T cells is not necessarily the result of low glucose availability but rather, it can also be related to intrinsic alterations in the T cell's metabolic machinery.For example, decreased levels of the PEP glycolytic intermediate, required for calcium/NFAT signaling (Fig. 1), can be due to glucose deprivation or to a reduced activity of enolase 1, the rate-limiting step in PEP production [86].Indeed, tumor-infiltrating T cells in a mouse model of melanoma exhibited attenuated enolase 1 activity, indicating yet another glycolytic step that can be negatively impacted in the tumor microenvironment [115].
Beyond the critical role of glucose, decreased amino acid availability has been shown to significantly impair anti-tumor responses.In this regard, it is important to note that several critical amino acids are highly depleted within the TME of both murine and human tumors.These include glutamine, arginine, aspartate, asparagine, and serine [116][117][118][119].As described in Section 2, these amino acids contribute to optimal T cell signaling, translation, and epigenetic remodeling (Figs 1 and 2A) [35,[120][121][122][123][124].The finding that pharmacological inhibition of glutamine transport in a mouse triplenegative breast cancer model increases the effector function of tumor-infiltrating T cells [125] strongly supports the hypothesis that the tumor cell-T cell competition for glutamine in the TME negatively impacts the latter.Arginine also directly impacts T cell function within the TME [109,120,121].The high levels of arginine consumption by cancer cells together with the expression of arginine-metabolizing enzymes [arginase (ARG)] by tumor-associated macrophages (TAMs) and myeloid-derived suppressor cells (MDSCs) contribute to the formation of an arginine-depleted hostile environment [126].Indeed, high levels of ARG1/2 within the TME have been correlated with poor prognosis in patients with ovarian carcinoma, pancreatic cancer, and head and neck squamous cell carcinoma [120,[127][128][129].
The immunosuppressive nature of the TME is also mediated by the secretion of metabolites (Fig. 2B).The high consumption of glucose by tumor cells results in the secretion of large amounts of lactate, blunting NFAT signaling in T cells [130] and skewing their differentiation.Furthermore, unlike conventional T cells, intra-tumoral Foxp3 + T REG cells display a metabolic program that is compatible with proliferation and sustained immunosuppressive activity that is sustained in glucose-low/ lactate-rich tumor environments [31,105].Conversely, inhibition of lactate dehydrogenase (LDH), coupled with IL-21, may promote T cell stemness [131], but this may depend on the model system.For example, sodium lactate itself appears to promote T cell stemness in a murine colorectal model [132] and lactate oxidation sustains the anti-tumor activity of CD8 + T cells in a murine melanoma TME [93].
Kynurenine and its metabolites are generated from tryptophan by cells in the TME expressing Indoleamine 2,3-dioxygenase (IDO); these include tumor cells, TAMs, and MDSCs.Mechanistically, kynurenine, 3hydroxykynurenine, and 3-hydroxyanthranilic acid can suppress T cell proliferation [133,134].Furthermore, the interaction of kynurenine and derived metabolites with the aryl hydrocarbon receptor (AHR) results in T REG generation and tolerogenic myeloid cell differentiation [135].Signaling through the AHR pathway is selectively active in IDO/TDO-overexpressing tumors and is associated with resistance to immune checkpoint inhibitors [135].There is also an extensive kynurenine-mediated crosstalk between tumor-repopulating cells (TRCs) and tumor-infiltrating CD8 + T cells.IFNc, produced by activated CD8 + T cells, triggers kynurenine secretion by TRCs which then feeds back on the intra-tumoral T cells, upregulating PD-1 expression and inducing T cell dysfunction [136,137].
The metabolic state of a tumor can also result in the abnormal accumulation of metabolites.Mutations in fumarate hydratase (FH) lead to hereditary as well as sporadic forms of cancer [138,139] and these malignancies are characterized by fumarate accumulation.Notably, recent work has shown that in a murine melanoma model, intra-tumoral fumarate suppresses tumor-infiltrating CD8 + T cells via succination of the ZAP-70 protein tyrosine kinase [140].Similarly, mutations in isocitrate dehydrogenase 1 (IDH1), occurring in malignant gliomas, result in the production of R-2hydroxyglutarate (R-2HG).This oncometabolite impairs CD8 + T effector function via inhibition of the glycolytic enzyme LDH [141,142].Thus, tumor cells secrete metabolites that negatively affect anti-tumor T cells via an extraordinarily broad range of mechanisms (Fig. 2B).
In addition to metabolites secreted by live tumor cells, the necrosis of tumor cells results in the liberation of immunosuppressive components.Extracellular potassium, released by necrotic tumor cells, has been identified as an inhibitory chemical element limiting T cell effector function.The ionic imbalance inhibits metabolic signaling through the AKT and mTOR kinases, suppressing T cell effector function.Mechanistically, high levels of extracellular potassium limit nutrient uptake into T cells, inducing autophagy and reducing histone acetylation at effector and exhaustion loci [143][144][145].
Non-tumoral cells also secrete metabolites that participate in the creation of a hostile TME.One example is the production of itaconate [product of immuneresponsive gene 1 (Irg1)] by MDSCs.Itaconate uptake by CD8 + T cells inhibits the biosynthesis of aspartate and serine/glycine, leading to decreased proliferation, cytokine production, and cytotoxic activity [146].In this regard, it is interesting to note that the knockdown of Irg1 reduces melanoma growth, inhibits the immune-suppressive activities of MDSCs, promotes anti-tumor immunity of CD8 + T cells, and enhances the efficacy of anti-PD-1 treatment [146].
Furthermore, ATP, actively released in response to cellular stress, is another key component of the TME [147,148].The CD39 and CD73 ecto-nucleotidases catabolize ATP into extracellular adenosine (eADO) [149] and in turn, eADO can inhibit anti-tumor responses through its binding to purinergic receptors expressed on T cells [148,150].In line with these results, the absence of CD39 has been shown to be associated with stem cell memory (T SCM )-like TILs which exhibit long-term persistence in patients responding to TIL therapies [151].Conversely, expression of CD39 on tumor-infiltrating T cells is sufficient to limit anti-tumor immunity [152], highlighting the important role of adenosine metabolism in regulating T cell function within the TME.
Oxygen tension is a critical factor modulating the TME.The high consumption of oxygen by tumor cells and the abnormal tumor blood vasculature generates hypoxic conditions that significantly impact T cell function.Furthermore, several studies have reported that the combination of intra-tumoral hypoxic conditions and chronic antigen exposure provokes T cell exhaustion, with the hypoxic environment impairing responsiveness to checkpoint blockade [152][153][154].
While many nutrients and metabolic intermediates are depleted in the tumor microenvironment, this is not the case for lipid-based metabolites.Cholesterol is notably enriched in tumor tissues [155] and its accumulation in tumor-infiltrating CD8 + T cells results in the upregulation of T cell exhaustion markers and an ER-stress-XBP1-dependent inhibition of anti-tumor function [156].Acyl-coenzyme A:cholesterol acyltrans-ferase1 (ACAT1)-driven cholesterol esterification inhibits TCR signaling by binding to the transmembrane region of the TCRb chain, disrupting TCR clustering.Interestingly, it appears that this pathway can be targeted for therapy.For instance, ACAT1 inhibition by pharmacological and gene-editing approaches potentiates the anti-tumor activity of T cells [157].Furthermore, cholesterol-driven liver X receptor (LXR) activation can result in differential effects as a function of the targeted T cell subset.Its activation has been found to inhibit IL-9-producing cytotoxic CD8 + T cell differentiation and associated anti-tumor cytotoxicity [158].In this context, it is notable that pharmacological modulators of lipid metabolism are being evaluated as part of the arsenal of anti-tumor T cell therapies [159].
Prostaglandin E2 (PGE2) has also been detected in the TME, impairing the survival and anti-tumor function of both CD4 + and CD8 + T cells [160,161].Similarly to other metabolites that inhibit the function of conventional but not regulatory T cells, PGE2 induces FOXP3 expression and enhances suppressive T cell function [162].PGE2 also plays an indirect role in regulating T cell function.PGE2 downregulates IRF8 expression in intra-tumoral type 1 dendritic cells (DCs) and the subsequent DC dysfunction results in an impaired CD8 + T cell infiltration within the tumor [163].
In light of the myriad factors and conditions that suppress T cell function in the tumor microenvironment, it is remarkable that anti-tumor T cells can sometimes succeed in controlling tumor growth.In the next section, recent approaches that facilitate the ability of anti-tumor T cells to overcome the immune hostile TME are discussed.

Metabolic interventions to optimize adoptive T cell immunotherapy
Adoptive T cell therapy represents a cutting-edge strategy, designed to treat cancer patients through the transfer of anti-tumor T cells.T cells engineered to express a CAR against a tumor antigen have demonstrated remarkable success in the treatment of human hematological malignancies [164][165][166].Nonetheless, despite this remarkable success, CAR T-treated patients do relapse, at least in part due to poor in vivo persistence and function of transferred CAR T cells [167].Furthermore, the efficiency of CAR T cell approaches in the treatment of solid tumors remains limited [168].The lack of a memory T cell phenotype in the transferred T cells and a suboptimal metabolic fitness of the CAR T cells may account for their constrained function, especially in the solid tumor microenvironment.
To overcome these limitations, much research has focused on the adoptive transfer of less differentiated memory T cells.Indeed, administration of T cells with a na€ ıve and/or T SCM phenotype has demonstrated robust anti-tumor responses and long-term remission in preclinical murine, macaque, and human models [169][170][171].Importantly, mechanistic studies showing differences in the metabolism of T cell subsets (see Section 1) correlate with the distinct phenotypes of adoptively transferred CAR T cells in patients.The transcriptomic profiles of CAR T cells in complete-responding patients with chronic lymphocytic leukemia (CLL) correlated with memory T cell formation and low glycolytic activity [172].Indeed, pharmacological inhibition of glycolysis was found to result in the generation of increased frequencies of human central memory CAR T cells [172].Interestingly, low glycolytic activity was previously found to enhance murine CD8 + T cell memory [173].Thus, generating CAR T cells with a metabolic program that promotes an "early memory" T cell phenotype is likely to improve intra-tumoral CAR T cell persistence and function, thereby promoting remission in patients (Fig. 3).Nonetheless, in the context of T cell-based immunotherapies for solid tumors, it is critical that the anti-tumor T cells have the potential to traffic and home to the tumor.Subsequently, they must navigate the tumor stroma network and mount an effective anti-tumor response within the hostile tumor microenvironment (see Section 4).To this end, chemokine receptors and their ligands are attractive targets, modulating T cell recruitment/function and depleting MDSCs/TAMs [174].Effector T cells may have an advantage as they express chemokine receptors such as CXCR3, that can promote T cell recruitment into solid tumors [175,176].Interestingly though, the upregulation of CXCR3 on anti-tumor T cells within the tumor microenvironment may be even more important than its role outside the tumor.This upregulation appears to restore T cell responsiveness after anti-PD1 checkpoint blockade by guiding intra-tumoral CXCR3 + T cells to CXCL9-producing CD103 + DCs in the context of several different tumors [177,178].More generally, much research has been invested in equipping CAR T cells with enhanced tumor targeting through the transgenic expression of chemokine receptors such as CXCR1/2, CCR4, CCR2b, and CXCR6 [174].

Pharmacological targeting of CAR T cells
To generate CAR T cells, T cells isolated from patients are stimulated through their TCR, transduced with a retroviral or lentiviral vector harboring the CAR transgene, and then are expanded in vitro.During this latter step, ectopic cytokines are added.While IL-2 has been classically added, other cytokines such as IL-7, IL-15, and IL-21 have also been found to support the generation of human CAR T cells [179][180][181].Importantly, each cytokine differentially impacts T cell metabolism.Cytokines such as IL-15 and IL-21 have been reported to decrease glycolysis, increase fatty acid oxidation, and increase mitochondrial activity, thereby promoting the generation of CAR T cells that may present with superior metabolic fitness for in vivo persistence [182][183][184][185].This is also the goal behind the approach of rapidly generating patient CAR T cells following a rapid 24-h stimulation with IL-7 and IL-15 alone [186].Moreover, IL-2 partial agonists were engineered in order to develop a molecule that allows T cell expansion but does not induce a metabolic program resulting in T cell exhaustion.One such molecule, H9T, promoted the generation of murine CAR T cells with an optimized mitochondrial fitness and these lymphocytes exhibited enhanced anti-tumor efficacy in leukemia and melanoma models [187].
Pharmacological inhibitors can also be used to directly or indirectly target metabolic pathways.The utilization of duvelisib to inhibit PI3Kd/c during the manufacturing of CAR T cells from patients with CLL increased mitochondrial biomass, a metabolic change that was associated with an enhanced number of T SCM and greater in vivo persistence and cytotoxicity [188].Other PI3K inhibitors, such as idelalisib and LY294002, also favor early memory T cell generation and improve human CAR T cell function [189,190].The AKT and mTOR kinases are downstream of PI3K (Fig. 3A) and their pharmacological targeting by agents such as the AKT inhibitor VIII (AKTi), ipatasertib, and rapamycin have been found to result in similar effects in murine and human models [191][192][193].The subsequent metabolic reprogramming modulates a FOXO1-driven transcription program with upregulation of early memory T cell phenotypes, resulting in the induction of CXCR4 on human CAR T cells and increased infiltration into the bone marrow [192,194].In addition, other pathways have also been targeted such as activation of WNT using the GSK3b inhibitor TWS-119 and activation of NOTCH via the DLL1 NOTCH ligand.Activation of both these pathways results in altered metabolic programs, and the hypothesis is that decreased glycolysis facilitates the generation of T SCM and robust anti-tumor activity [195][196][197][198][199]. Urolithin A (UA), an inducer of mitophagy that is a metabolite of pomegranate extract, was found to drive WNT signaling and promote the generation of murine T SCM cells with high anti-tumor cytotoxicity against a murine colorectal adenocarcinoma [200], highlighting the multiple interconnections between these pathways in T lymphocytes.
Another major research axis has focused specifically on modulating glucose availability as well as glycolysis, both during CAR T cell generation and in the final product [201].Inhibition of the lactate transporter MCT1 results in a bias towards oxidative phosphorylation over glycolysis and while this intervention had no impact on the phenotype of human CD19 CAR T cells, their anti-leukemia cytotoxicity was enhanced [202].Furthermore, attenuation of pyruvate entry into the mitochondria, through inhibition of the SLC54A1/MPC1 transporter, increased the pool of acetyl-CoA and associated histone acetylation.This led to a memory-like phenotype of both murine and human CD19 CAR T cells with enhanced anti-tumor activity [93].The antidiabetic drug metforminan inhibitor of mitochondrial respiratory chain complex I and activator of AMPK [203] appears to promote memory T cell generation, improving the function of HER2 CAR T cells against human lung adenocarcinoma [204].Moreover, the presence of metformin within the tumor microenvironment may be advantageous.The inclusion of a metformin-containing hydrogel scaffold in human gastric carcinoma tissue improved CAR T cell infiltration and proliferation, likely by upregulating oxidative phosphorylation [205].This summary highlights the large array of metabolicfocused CAR T cell interventions that have been successfully tested in the lab (Fig. 3A).Therefore it will be of much interest to determine the potential efficacy of these interventions in the good manufacturing practices (GMP) manufacturing of CAR T cells for clinical trials.

Genetic manipulation of CAR T cells
The interventions described above are all dependent on extracellular cues.Notably, gene-editing strategies, have been very successful, at least in preclinical models, by directly targeting metabolic pathways va overexpression or knockdown of specific genes.Indeed, overexpression of PGC1a, a transcription factor associated with mitochondrial biogenesis, rescues EGFR CAR T cells from mitochondrial dysfunction in the TME, improving their fitness and cytotoxicity against human lung cancer cells (Fig. 3B) [61,206].However, metabolic interventions are likely to differ as a function of the CAR target and the tumor model.Additionally, the impact of these interventions may vary as a function of the costimulatory domain incorporated into the CAR.CAR constructs generally include either a CD28 or 4-1BB costimulatory domain and the downstream signaling moieties that are activated following engagement of CARs with these different domains are not equivalent [207][208][209][210][211]. Activation of distinct signaling cascades will, in turn, alter the cell's metabolic program, but interestingly, metabolic analyses comparing CD28 and 4-1BB CARs have only been reported in non-activated human T cells electroporated for transient CAR expression [212].It will therefore be critical to review the role of the CAR costimulatory domain in driving CAR T cell metabolism in the context of the different therapeutic strategies discussed below.The metabolism of CAR T cells can be directly altered by inhibiting or overexpressing metabolic enzymes and transporters.Arginine metabolism, discussed in Sections 2 and 3, plays a critical role in T cell proliferation and survival, and the TME is relatively depleted in arginine.To overcome arginine depletion, Mussai, De Santo and colleagues ectopically expressed enzymes involved in either arginine catabolism (ARG1/ARG2) or arginine resynthesis [arginosuccinate synthase (ASS) and ornithine transcarbamylase (OTC)] in CD33-as well as GD2-targeted CAR T cells.They found that these modified CAR T cells exhibited enhanced in vivo anti-tumor responses against human AML and neuroblastoma [213,214].Similarly, our group and others have modulated the expression of glucose, glutamine, and arginine transporters, among others, in order to enhance CAR T cell function in nutrient-deprived tumor microenvironments (Fig. 3B) [108,214].
Gene-editing strategies have also been harnessed in an attempt to protect CAR T cells from the potentially harmful effects of tumor metabolites in the tumor microenvironment.To attenuate the negative effects of tumor-generated fumarate (see Section 4), mouse and human CD19 CAR T cells were engineered to overexpress fumarate hydratase, thereby promoting their catabolism of intracellular fumarate.Notably, these CAR T cells demonstrated an enhanced anti-tumor activity [140].Several other groups took related approaches in an attempt to abrogate the negative effects of adenosine.CAR T cells engineered with a deletion in the adenosine receptor A 2A R gene as well as CAR T cells overexpressing adenosine deaminase 1, catabolizing adenosine into inosine, exhibited greater anti-tumor activity in several models, including mouse mammary cancer and human ovarian cancer.
CRISPR screening approaches in CD4 + as well as CD8 + T cells have been undertaken in an attempt to discover genes and gene networks that regulate T cell immunity in hematological malignancies as well as solid tumors [215][216][217][218][219][220].Given the importance of metabolism in T cell immunity, it is not surprising that many identified genes regulate metabolic pathways.One gain of function CRISPR screen identified proline dehydroge-nase2 (Prodh2) as the top hit, increasing proline, as well as arginine metabolism.Most notably, PRODH2 overexpression in human CD22-, BCMA-, as well as HER2directed CAR T cells significantly increased anti-tumor activity [217].Another CRISPR screen found that deletion of Regnase 1, a native regulator of the basic leucine zipper ATF-Like transcription factor (Batf), enhanced mitochondrial function as well as CAR T cell responsiveness [216].Moreover, BATF itself has been found to promote T cell survival and anti-tumor responsiveness of CAR T cells [221][222][223].Nonetheless, the specific experimental conditions are critical because inverse effects have also been reported, knocking out BATF in human CAR T cells under exhaustion conditions was reported to improve cytotoxicity, and high BATF3 expression in TET2-edited CAR T cells resulted in reduced cytotoxicity [224,225].While these discrepant results are likely due to underlying differences in T cell metabolism in leukemia and solid tumor models being evaluated, they hinder extrapolation to clinical trials in patients.Together, these data highlight the hurdles in advancing these targeted approaches to clinical trials but they also point to the extraordinary potential of harnessing metabolic pathways to promote anti-tumor T cell responses.
To improve the therapeutic efficacy of CAR T cells, especially with regards to their application in the treatment of solid tumors, metabolic interventions by pharmacological and gene-editing technologies present promising avenues.In this regard, it is notable that multidimensional omics data from both tumor and anti-tumor immune cells can now be leveraged to identify novel metabolic targets [226].Moreover, spatial organizations of metabolic programs have been revealed through the profiling of human tissues and cells at a single-cell level using recently described techniques such as scMEP (single-cell metabolic regulome profiling), SCENITH (Single Cell ENergetIc metabolism by profiling Translation inHibition), and scSpa-Met (Single Cell Spatially resolved Metabolic), incorporating CyTOF, bulk metabolic assays, protein synthesis assays, and multiplex Imaging Mass Cytometry [227][228][229].An understanding of the metabolic genes whose expression profiles are altered in tumors may promote the design of CAR T cells that are remodeled to better function within the metabolic complexity of TME [230,231].Indeed, single-cell analyses of human hepatocellular carcinoma and human pancreatic ductal adenocarcinoma have recently revealed distinctive metabolic changes in methionine recycling, glutamine metabolism, and lipid accumulation [232][233][234][235].However, deducing the activity of a particular metabolic pathway from scRNAseq data can be challenging due to weak correlations between mRNA levels and protein levels.Additionally, there remains an inadequate understanding of the activity of metabolic enzymes and transporters within the highly interconnected metabolic networks [14,15,236].System-based computational modeling approaches that integrate mRNA levels of metabolic genes with the prior knowledge of these networks may help to overcome these limitations, promoting a more comprehensive view of singlecell metabolic profiles [42,236,237].The hope is that by improving the fitness of CAR T cells within tumorspecific microenvironments, the outcomes of patients receiving CAR T cell therapy will be significantly improved.
6. Perspectives and conclusions: Mice and humans. . .As highlighted 20 years ago in a review from Mestas and Hughes entitled "Of mice and not men: Differences between Mouse and Human Immunology," 65 million years of evolution have resulted in significant differences between mice and humans.Genes impacting both innate and adaptive immunity differ in their expression profiles and functions, including critical T cell signaling molecules such as ZAP-70 and the common gamma chain [238].Moreover, some pathways are species-specific.For example, while the mouse CD46 complement regulator does not contain any known signaling motifs, human CD46 is a critical regulator of macrophage and T cell function [239,240].These divergences are compounded by the particularities of mice strains and the controlled environments in which mice are maintained.
Mice are housed in specific pathogen-free conditions resulting in the maintenance of a na€ ıve immune environment, an anomalous state as compared to humans.The immune cell compartment of laboratory mice can be altered by exposing them to the microbiota of their wild counterpartsmarkedly modifying their responses to antigens [241][242][243][244][245] but to date, this is not common in immune studies.Indeed, these disparities have led to conclusions regarding disease processes and treatments that are often not universally applicable, not just to humans but to other strains of mice and diverse conditions [246][247][248][249]. Furthermore, while humanized mouse models are essential for preclinical studies of novel anti-tumor receptors on human T cells within a human tumor setting, the absence of an adaptive immune system in these mice frequently complicates interpretation of the data.Nonetheless, while these examples highlight the myriad difficulties in extrapolating mouse experimental data into clinical trials for patients, syngeneic and humanized mouse models are powerful tools.Their optimization will be invaluable in the quest for safer and more efficient anti-tumor immunotherapies.

Fig. 2 .
Fig. 2.An immune hostile tumor microenvironment: Impact on T cell function.(A) The high uptake of diverse nutrients by tumor cells frequently results in the selective depletion of glutamine, asparagine, aspartate, serine, and arginine within the TME, leading to their decreased availability for anti-tumor T lymphocytes.Moreover, arginase-1, produced by tumor-infiltrating TAM and MDSC, catabolizes arginine to ornithine, further limiting the utilization of arginine by tumor-infiltrating cytotoxic T cells.The TME is often further compromised by suboptimal gas exchange, resulting in low oxygen concentrations (left).A limited accessibility to nutrients results in decreased mTOR signaling in tumor-infiltrating T cells which in turn results in lower levels of cell surface metabolite transporters such as GLUT1.Decreases in glycolytic enzymes such as enolase result in the reduced generation of PEP, attenuating Ca 2+ -NFAT activity via SERCA activity (right).(B) An abundance of tumor-related metabolites suppresses T cell function.These include cholesterol, fumarate, lactate, low pH (secondary to high lactate secretion by tumor cells), and kynurenineconverted from tryptophan by IDO.PGE2 inhibits cDC function, thereby impairing T cell immune responses.Furthermore, ATP and ADP, released from dying cells, are converted to AMP and eADO by CD39 and CD73, respectively, on the surface of exhausted T cells.eADO as well as potassium from necrotic tumor tissue suppress T cell activity in the TME.ADO, aldehyde deformylating oxygenase; ADP, adenosine diphosphate; AMP, adenosine monophosphate; ATP, adenosine triphosphage; CD39, cluster of differentiation 39; CD73, cluster of differentiation 73; cDC, conventional dendritic cells; eADO, extracellular adenosine; ER, endoplasmic reticulum; GLUT1, glucose transporter; IDO, indoleamine 2, 3-dioxygenase; MDSC, myeloid-derived suppressor cell; mTOR, mammalian target of rapamycin; NFAT, nuclear factor of activated T cells; PEP, phosphoenolpyruvate; PGE2, prostaglandin E2; SERCA, sarcoendoplasmic reticulum calcium ATPase; TAM, tumor-associated macrophage; TME, tumor microenvironment.