The CAR incorporated at the surface of T cells has a T cell receptor (TCR)-like structure. The extracellular part consists of a single variable chain fragment (scFv) of an antibody that recognizes the antigen present on the cancer cell surface. The transmembrane and intracellular parts contain co-stimulatory domains that permit the amplification of the signaling and, thus, the response of the T cell following binding to the tumor antigen. In the first-generation CAR T cell design, CD3ζ was the only signaling domain used, but in the following generations, CD28 and/or 4-1BB (CD137) co-stimulatory domains were added to this structure. The choice of these domains is also important for the metabolic status and the survival of the CAR T cells. It was observed in patients that the use of 4-1BB permitted a persistence of CAR T cells in time, without exhaustion of those CAR T cells, via the stimulation of the noncanonical nuclear factor kappa B (NF-κB) pathway ^[1][69]. In contrast, the CD28 costimulatory domain does not permit the cells to survive longer than 30 days in the patients ^[2][3][4][5][70,71,72,73]. This phenomenon can be explained by the fact that the 4-1BB promotes improved mitochondrial function so that T cells can rely on mitochondrial respiration as their energy source, which promotes the survival of central memory T cells by the activation of adenosine monophosphate kinase (AMPK), as demonstrated in mice ^[6][74]. When the CD28 co-stimulatory domain was included in the CAR, patient T cells were relying more on glycolysis (via activation of the PI3K/AKT/mTOR axis), which pushed their differentiation towards effector memory T cells ^[7][8][75,76]. Following antigen recognition by the CAR, CD28 stimulated GLUT1 via the PI3K/AKT pathway, linked to the mTOR/Myc pathways, which are also implicated in amino acid/lipid metabolism ^[9][10][17,19]. These data underline the importance of the choice of the co-stimulatory domain to shape CAR T cell metabolism and permit short-term or long-term anti-tumor efficacy. As for CAR T cells, TIL performance also depends strongly on the metabolic status of the malignancy and the TME.

Cytokines also play a role in the metabolic shape of the CAR T cells. Indeed, the fourth CAR T cell generation, also known as “TRUCKS” (T-cell redirected for unrestricted cytokine-mediated killing), has a transgenic cytokine expression system that improves their expansion and persistence in vivo via metabolic changes ^[11][77]. The cytokines employed in this strategy are interleukin 12 (IL12), IL15 and IL18. IL15 can, for example, lower the glycolysis level in human CAR T cells via a decrease in mTOR complex 1 (mTORC1) activity, whilst OXPHOS levels/respiratory capacity and expression of fatty acid oxidation-related genes are increased in these cells. This metabolic phenotype allows the human CAR T cells to have a stem cell memory behavior with higher cell proliferation and in vivo longevity ^[12][78]. More recent fifth-generation CAR T cells express a sub-unit of the IL2 receptor (IL2RB) between the stimulatory domains (CD3ζ and CD28/4-1BB). Since IL2RB presents a binding site for STAT3, the cells can activate the JAK-STAT pathway, following the antigen recognition, which leads to a higher proliferation rate. Moreover, these human CAR T cells seem to have a stronger anti-cancer activity against leukemic cells ^[13][79] (Figure 12).

As they are immune cells and similar to TILs, CAR T cells are highly sensitive to metabolic changes in their environment, which can influence their survival and their functionality. When cultured ex vivo, CAR T cells are often cultured in the presence of supra-physiological nutrient levels in order to generate enough cells for reinfusion. This also means that the cytotoxic efficiency is compromised when changing from an overdosed medium to a nutrient restrictive microenvironment. Thus, the composition of expansion media plays an important role in the process of CAR T cell generation and needs to be optimized in order to facilitate the transition between the ex vivo expansion and in vivo adaptation. For example, carnosine, an amine poorly present in sera, when added to the medium, facilitates human T cell transduction and their in vivo engraftment by switching the main metabolic state from glycolysis to OXPHOS, which allowed a better anti-tumoral response ^[14][80]. Currently, expansion media rely essentially either on fetal bovine serum (FBS) or human serum (HS), which does not take into account some blood fractions, which might highly compromise human CAR T cell fitness ^[15][81].

The effect of cytokines on T cell metabolism is only beginning to be explored in the field. Survival cytokines, IL2 and IL7, promote glycolysis in T cells ^[16][17][21,82], while IL-15 promotes their mitochondrial biogenesis ^[18][36]. TGFβ seems to suppress both glycolysis and mitochondrial respiration. Therefore, the choice of cytokines or growth factors in the media for CAR T cell expansion might be crucial for their in vivo adaptation.

It is therefore important to be aware that the ex vivo expansion of these CAR T cells is a tricky step in the process and can be limiting and compromise treatment efficacy. Several research teams are currently trying to improve CAR T cell therapy by circumventing the ex vivo expansion step and directly generating CAR T cells in vivo. In recent studies, lentiviral vectors specifically recognizing the CD3⁺ or even the CD8⁺ human T cells were used to generate efficiently anti-CD19 CAR T cells directly in vivo and were able to wipe out the targeted healthy and malignant B cells in humanized mouse models ^{[19][20][21][22][23]}[83,84,85,86,87]. Although this still needs to be evaluated, one can speculate that these in vivo generated CAR T cells might expand in their proper environment and maintain their metabolic fitness in vivo.

More recently, an elegant strategy to overcome in vivo CAR T cell exhaustion has been reported with better outcomes than PD-1/PD-L1 blockade, which can improve T cell functions but do not change the epigenetic reprogramming associated with T cell exhaustion. The approach is based on inducing a transient rest in CAR signaling and, therefore, temporally blocking the sustained CAR T cell activation that leads to exhaustion ^[24][88]. This approach made use of temporal downregulation of the surface expression of the CAR by inducing its degradation by pharmacological treatments. The tunable control of CAR expression levels not only improved CAR T cell anti-tumoral functions in vivo in terms of intensity and duration but also led to non-exhausted T cells with a memory-like phenotype in xenograft mouse models of leukemia and osteosarcoma. Interestingly, the already established exhaustion of CAR T cells can even be reverted after a transient rest of the CAR signaling ^[24][88].