ATPIF1 inactivation promotes antitumor immunity through metabolic reprogramming of CD8+ T cells

Induction of CD8+ T cell activity is a promising strategy in the cancer immunotherapy. In this study, we identified ATP synthase inhibitory factor 1 (ATPIF1) as a potential target in the induction of CD8+ T cell immunity against tumor. Inactivation of ATPIF1 gene in mice promoted the antitumor activity of CD8+ T cells leading to suppression of tumor growth of B16 melanoma and Lewis lung cancer. The phenotype was abolished by deletion of CD8+ T cells in the ATPIF1-KO mice. The tumor infiltrating CD8+ T cells exhibited strong activities in the proliferation, effector and memory as revealed by the single cell RNA sequencing results of CD45+ tumor infiltrating lymphocytes (TILs) isolated from the tumors. The CD8+ T cells expressed more antitumor makers in the tumor microenvironment and in coculture with the tumor cells. The cells had a higher level of glycolysis after the T cell receptor-mediated activation as revealed by the targeted metabolomics assay. The cells exhibited an extra activity of oxidative phosphorylation before the activation as indicated by the oxygen consumption rate. The cells gained capacities in the proliferation, apoptosis resistance and mitophagy in the glucose-limiting environment. These data suggest that inhibition of ATPIF1 activity by gene inactivation rewired the energy metabolism of CD8+ T cells to enhance their immune activities to the tumors. ATPIF1 is a potential molecular target in the induction of antitumor immunity through metabolic reprogramming of CD8+ T cells for the cancer immunotherapy.


Introduction
The application of PD-1/PD-L1 antibody revives cancer immunotherapy for advanced or drug-resistant cancers in the clinic practice. The success shed light on the strategy for metabolic reprogramming in the induction of immune activities of T cells 1 , such as durability, longevity, and functionality 2 . Energy supply is a primary factor in the control of those activities of T cells, especially in the glucose-limiting condition of the tumor microenvironment, in which T cells suffer energy deficiency due to lack of glucose and other nutrients 3 . Glucose provides energy (ATP) and building materials to the T cells in support of cell survival and function 4 . The energy shortage is intensified by the exhaustion proteins (PD-L1) of tumors 2 . Metabolic reprogramming is a potential strategy to overcome the energy shortage in T cells by maintaining the energy homeostasis 5 . However, the strategy is limited by a lack of the molecular targets in the metabolic reprogramming 4,6 .
ATPIF1 is an inhibitory protein of F 1 F o -ATP synthase (ATP synthase, Complex V) in mitochondria, which catalyzes ATP production through phosphorylation of ADP at expense of mitochondrial potential in the energy substrate-enriched environment. The ATP synthase hydrolyzes ATP in the energy substrate-deficient conditions to protect cells from apoptosis by maintaining the potential. Both activities of ATP synthase are regulated by ATPIF1 7,8 , which involves a protein-protein interaction 9 . ATPIF1 is highly expressed in mitochondrion-enriched cells to match the ATP synthase activity. An increase in ATPIF1 activity in the ATPIF1 overexpression mice, the ATP synthase activity is decreased leading to energy metabolism reprogramming for down-regulation of oxidative phosphorylation (OXPHOS) and up-regulation of glycolysis in mice 7 . The reprogram increases the risk of cell apoptosis under conditions of respiration inhibition and energy deficiency from the mitochondrial potential collapse 10 . ATPIF1 activity has been investigated in the regulation of metabolism in the neuronal cells 11 , red blood cells 12 , epithelial cells 13 , hepatocytes 14 and tumor cells 15 . However, its role remains unclear in lymphocytes.
In this study, we observed that ATPIF1 inactivation promoted the CD8 + T cells activities in proliferation, survival and effector memory as indicated by single cell RNA sequencing assay of CD45 + leukocytes isolated from the B16 tumor. The impact was a consequence of metabolic reprogramming in CD8 + T cells following an elevation of the ATP synthase activity.
These results suggest that APTIF1 is a potential molecular target in the induction of anti-tumor activity of CD8 + T cells by metabolic reprogramming.

Antitumor immunity is enhanced in ATPIF1-KO mice in dependence on CD8 + T cells
The ATPIF1-KO (ATPIF1 -/-) mice was generated using the CRISPR-Cas9 strategy, and the inactivation was verified with a decrease in ATPIF1 protein in the spleen lymphocytes by Western blotting (Fig. 1a). The knockout mice exhibited no difference from the wild type (WT) mice in growth and reproduction. The impact of ATPIF1 inactivation in the immune function remained unknown. We addressed this issue by analyzing the percentage of T cells, B cells, Treg cells, dendritic cells, and neutrophil cells in the cell population of ATPIF1-KO mice spleen. No significant alteration was detected in those cells in the normal condition ( Fig.   S1), but the percentage of CD11c + dendritic cells was increased in the population analysis (Fig. S1c).
To investigate the immune system further, we examined the antitumor activity of KO mice in the two tumor-bearing models of B16 melanoma and Lewis lung cancer (LLC). The B16 tumor growth was significantly reduced after engraftment into the KO mice with a smaller tumor size at two weeks. A 56.6% reduction in the tumor weight was observed (p = 0.0017, Fig. 1b). A similar reduction in tumor size was found in LLC tumor in the KO mice for 41% decrease in tumor wight relatively to that of WT mice (p = 0.0045, Fig.1c). To elucidate the T cell activity in the phenotype, the CD8 + T cells were depleted in the mice using a CD8 antibody during the B16 tumor growth (Fig. S2). The depletion led to a robust acceleration in tumor growth in the WT and KO mice (Fig. 1d), which abolished the difference between the two groups of mice in tumor growth. These data suggest that ATPIF1 inactivation enhanced the tumor immunity in the KO mice to inhibit growth of engrafted tumors, which was dependent on CD8 + T cells.

ATPIF1 inactivation favors CD8 + T cell expansion in TILs
Above data suggest that the ATPIF1 inactivation may generate an impact in CD8 + T cells to enhance the tumor immunity. To test the possibility, the single-cell RNA sequencing (scRNA-seq) technology was used in the analysis of TILs heterogenicity. The scRNA-seq technology has been reported in the landscape analysis of TILs of hepatocellular carcinoma 16 , breast cancer 17 and colorectal cancer 18 . The assay was conducted using CD45 + cells isolated from the tumor tissues with 8408 cells from the KO mice and 10059 cells from the WT mice ( Fig. S3). The sequencing data were obtained from 6007 leukocytes in the KO group and 8010 leukocytes in the WT group based on the 10x genomics method ( Table S1). The violin chart and the principal component analysis (PCA) indicated the well quality control during the scRNA-seq (Fig. S4a and S4b), also, the T-distributed random neighborhood embedding (tSNE) and uniform manifold approximation and projection (UMAP) graph showed the similar landscape between the WT and KO sample ( Fig.S4c and S4d).

ATPIF1 inactivation promoted maturation of naïve T cells and proliferation of T cells to
increase the percentage of CD8 + T effector (T eff ) and T memory (T mem ) cells in TILs. Among the CD45 + leukocytes, 1353 CD3 + T cells in the KO sample and 907 CD3 + T cells in the WT sample were finally sequenced. Eight clusters of CD3 + T cells were identified in TILs by scRNA-seq (Fig. 2a) and the marker genes in each cluster was summarized (Fig.S5). In the KO mice, the clusters of naïve T cells (KO:WT=39.4%:57.8%) and CD4 + T reg (KO:WT=5.54% 7.72%) were decreased. The clusters of proliferative T cells (KO:WT=9.02%:3.31%), CD4 + T cells (KO:WT= 21.7%:12.4%) and CD8 + T eff and T mem cells (KO:WT=8.80%:3.64%) were increased (Fig. 2b). The proliferation was mainly found in the CD8 + T cells by the markers of MKI67 and Stmn1 (Fig. 2c). IFN-γ mRNA was increased in T cells of KO mice (Fig. 2d).
These indicate an enhancement in the proliferation and effector (IFN-γ secretion) capacities in T cells of the KO mice. The STARTRAC (single T cell analysis by RNA-seq and TCR tracking) program 18 was employed to track the dynamic relationships among T cell subsets in TILs. STARTRAC is a novel analytical framework in investigation of the clonal expansion, migration and state transition of TILs, which provide critical insights into the T cell lineages 18,19 . In the KO mice, the CD8 + T eff and T mem cells exhibited an enhanced clonal expansion, which was followed by expansion of T-proliferative cells (Fig. 2, e and f). In contrast, the expansion of CD4 + Treg was predominant in the WT mice (Fig. 2g), which was associated with expansion of CD4 + T cells (Fig. 2h). CD4 + Treg mediates the immunosuppression activity in tumor 6 . The pseudotime trajectory analysis was used to characterize the lineage differentiation of TILs. The developmental trajectory from the T proliferative cells to CD8 + T eff was enhanced in the KO mice (Fig. 2i). To verify the result, we analyzed the CD8 + T eff and T mem in the peripheral blood mononuclear cells (PBMC) and TILs using flowcytometry.

Function of CD8 + T cells is enhanced in KO mice under the glucose-limiting condition
Above data suggest that the CD8 + T cells exhibited an enhanced proliferative capability and effector memory activity in the KO mice. To verify the results, CD8 + T cells were investigated in proliferation and IFN-γ secretion in the cell culture in vitro. The cells were isolated and activated with T cell receptor signal mimics (CD3 plus CD28) in the culture. The test was conducted with three concentrations of glucose as glucose is required for T cell function and IFN-γ production 20 . Interestingly, the KO cells had a delayed proliferation under the high (HG) and normal (NG) levels of glucose (Fig. 3a). However, the phenomenon was reversed under the low glucose (LG) condition, which resembles the glucose-limiting environment in tumor. Consistently, IFN-γ secretion was increased in the KO cells under the low glucose condition (Fig. 3b). Glucose uptake was examined in T cells using a fluorescent glucose analog (2-NBDG) to access the glucose metabolism 21 . In the KO cells, the uptake was significantly increased in the LG condition (Fig. 3, c and d), which was observed with more mRNA expression of glucose transporter 1 (Glut1, slc2a1) (Fig. 3e). mRNA of pyruvate dehydrogenase kinase 1 (PDK1) was upregulated in the KO cells in the same condition ( Fig.   3f), suggesting an inhibition of the oxygen-dependent glucose metabolism. The data suggest that T cell proliferation, IFN-γ production and glucose uptake were all enhanced in the KO cells under the low glucose condition.

Glycolysis is enhanced in KO cells upon activation by CD3 plus CD28 signals
Glucose is the major fuel in T cells in the production of ATP, which involves two steps: glycolysis and oxidative phosphorylation (OXPHOS). The two steps were examined in CD8 + T cells using the Seahorse equipment for the oxygen consumption rate (OCR) of OXPHOS, and the extracellular acidification rate (ECAR) for aerobic glycolysis, respectively. In the KO CD8 + T cells, OXPHOS was more active in the resting state (or naive state) (Fig. 4a, p<0.05), but less active in the activated state (Fig. 4b, p<0.05). The spare respiratory capacity (SRC) was increased in both resting and activated conditions (Fig. 4, a and b). Glycolysis was less active in the resting state as indicated by the low value of ECAR (Fig. 4c, p<0.05), which became more active after cell activation for the rise in ECAR value (Fig. 4d). The glycolytic reserve (GR) was significant higher in the KO cells in both resting and activated states (Fig. 4, c and d, p<0.01). The increase in SRC is a character of energy metabolism in the memory antitumor T cells, and GR increase is a character of the effector antitumor T cells 31,32 . The data suggest that glucose metabolism is reprogrammed in CD8 + T cells of KO mice in favor of an energy adaptation to the antitumor activity.

Targeted metabolomics reveals an elevation in glycolysis and pentose phosphate pathway in the KO T cells
Targeted metabolomics was used to investigate the metabolic reprogramming in the KO cells. CD8 + T cells (10 7 ) were purified from the spleen of tumor-bearing (B16-OVA) mice through FACS-mediated sorting. Metabolites in the energy metabolism pathways were analyzed using the liquid chromatography-mass spectroscopy (LC-MS). The metabolites include those in the tricarboxylic acid cycle (TCA), glycolytic pathway, OXPHOS and pentose phosphate pathway. The metabolite profile revealed a higher activity in the glycolysis and the pentose phosphate pathways for the increase of D-Glucose 1-phosphate, D-Glucose 6-phosphate, α -D-Ribose 5-phosphate and coenzyme A (Fig. 5, a and b, raw data are summarized in Table S2). The TCA cycle activity was reduced for the decrease in acetoacetyl-CoA and acetyl-CoA. The changes in metabolite profiles favor the production of antitumor cytokines, such as IFN-γ in T cells 29 . The data suggest that CD8 + T cells of KO mice were more active in adaptation to aerobic glycolysis from OXPHOS in the fight against tumor.

CD8 + T cells exhibit an enhanced capacity in the maintenance of mitochondrial potential (Δψm) in the KO mice
The F 1 F o -ATP synthase synthesizes as well as hydrolyzes ATP, which will rise in the absence of ATPIF1 according to studies in other cell types. However, the activity alteration remains unknown in T cells of KO mice. To address the issue, the CD8 + T cells were examined with several parameters including the intracellular ATP level, mitochondrial potential (Δψm), ROS, and MitoSOX. The tests were conducted in the cells under different glucose concentrations to understand the alteration in ATP synthase activities. The intracellular ATP content was increased in the KO CD8 + T cells at the high (HG) and normal (NG) glucose concentrations to support an increase in the synthetic activity of ATP synthase in the energy sufficient conditions (Fig. 6a). The ATP was decreased at the low glucose (LG) concentration to reflect the rise in hydrolytic activity of ATP synthase in the energy deficient condition (Fig. 6a). The hydrolytic activity is required for the maintenance of Δ ψ m was examined in the cells using JC-1 and the level was higher in the KO cells over the WT cells at the LG condition (Fig. 6b). The difference was also observed in the HG condition. The mitochondrial mass was determined using MitoTracker staining. The mass was decreased in the KO CD8 + T cells in the normal (NG) and low (LG) glucose conditions (Fig.   6c). These data demonstrate that the ATP synthase activity was reprogrammed in CD8 + T cells of ATPIF1-KO mice for an improved adaptation to the energy supply.

Immune activity of CD8 + T cells is enhanced in KO mice in vitro and in vivo
The antitumor function of CD8 + T cells was tested further for IFN-γ expression in vitro and in vivo. The study was conducted with OT-I mice, in which the CD8 + T cells are engineered to express a T cell receptor (TCR) motif for recognition of ovalbumin antigen of SIINFEKL peptide 34 . ATPIF1 was inactivated in the OT-I mice by crossing the KO mice with OT-I mice. The CD8 + T cells was isolated from the hydride mice (ATPIF1 -/-OT-I) and challenged with the ovalbumin-expressing tumor (B16-OVA) cells to induce IFN-γ expression in vitro. Intracellular IFN-γ was quantified in the T cells using flow cytometry.

IFN-γ was elevated in the cells by HG and decreased in the cells by
LG in the WT mice ( Fig.   6d). However, the LG response were significantly attenuated in the KO cells ( Fig. 6d) for much more IFN-γ production in the low glucose condition. The KO cells were resistant to apoptosis under the LG condition (Fig. 6e, p<0.001). The data suggest that CD8 + T cells of ATPIF1-KO mice exhibited a much better activity in the INF-γ production and apoptosis resistance under the glucose-limiting condition. To confirm the result in vivo, the CD8 + T were isolated from the OT-1 mice or ATPIF1 -/-OT-1 mice using the Dynabeads TM of CD8 + positive isolation kit and then transferred into the B16-OVA xenografted Rag1 mice via the intratumor injection. The tumor inhibition was significantly improved in the ATPIF1 -/group over the WT group (p=0.042) (Fig. 6f), supporting the enhanced antitumor efficacy of CD8 + T cells of ATPIF1-KO mice.

Mitophagy is enhanced in CD8 + T cells of KO mice under the glucose-limiting condition
In the glucose-limiting condition, the survival ability was enhanced in the CD8 + T cells of ATPIF1-KO mice to support the immune function. These suggest that the KO cells may gain extra ability to obtain energy beyond glucose. Autophagy was examined to test the possibility by examining the representative markers. In the KO cells, mitophagy (PINK1 and Parkin) and autophagy (ATG5) markers were significantly increased ( Fig. 7, a and b). The increases were observed under the LG as well as NG condition. The increases occurred regardless of activation status in the CD8 + T cells. Consistently, more autophagic vacuoles were found in the cytoplasm of KO cells under the electronic microscope (Fig. 7c).

Mitophagy was induced in the normal (WT) T cells following activation by CD3+CD28
under the LG condition, and the response was further enhanced by the mitochondrial uncoupler CCCP (an mitophagy agonist) (Fig. 7d). Similar responses were observed in the KO cells in the presence of higher level of basal mitophagy (Fig. 7d). The enhanced autophagy was associated with more IFN-γ expression in the KO cells after activation by CD3/CD28 (Fig. 7e). However, the CCCF-induced IFN-γ response was attenuated in the KO cells (Fig. 7e). These data suggest that mitophagy was increased in the KO cells to provide energy in the glucose-limiting conditions.

Macrophage and dendritic cell clusters in TILs support the enhanced tumor immunity in KO mice
The alteration in macrophage and dendritic cell (DC) clusters of TILs supports the enhanced tumor immunity in the KO mice. In the scRNA-seq assay, four major clusters and 18 subclusters were identified in TILs based on the gene expression profiles (Fig. 8a). The major clusters included T cells, macrophages, B cells, and dendritic cells by the canonical marker genes (Fig. S6), indicating a well coverage of the cluster spectrum. Five macrophage clusters were identified in TILs (Fig. 8a). The macrophage_S100a8 cluster was decreased dramatically in the KO mice by 75% (Fig. 8b). Macrophage_S100a8 is an indicator of poor clinical outcome in the breast cancer for tumor invasion and migration 22 . The immune suppression-related genes including Socs3, Plaur (urokinase plasminogen activator receptor, uPAR), and Arginase 2 (Arg2) were significant decreased in the cluster in the volcano plot ( Fig. 8c), supporting that the immunosuppressive activities of macrophages were reduced in the KO mice 23,24,25 . In the other four clusters, three of them such as macrophage_Lars2, macrophage_Apoe, and macrophage_Dct were increased, and one of them, Macrophage_CellCycle, was not altered in the KO mice (Fig. 8b). Functions of those four clusters are not known yet in the tumor immunity.
The enhanced immunity is further supported by the alteration of dendritic cells (DCs) (Fig. 8, a and (Fig. S7). Clec9a is an important marker of DCs in the recognition of necrotic cells and cross-presented antigens 26,27 . In this cluster DCs, the genes including Socs3 and IкBα (Nfkbia) were decreased significantly together with Gm42031, mt-Atp8, and Ler3 (Fig. 8d). Clec10a is an endocytic receptor for maturation and initiation of immune response in DCs 28 . Plasmacytoid DCs (pDCs) produce type I interferons in the antiviral immunity, and its function is multifaced in the tumor immunity 29 . A 2-fold decrease (KO:WT=0.33%:1.10%) was found in the cluster of pDCs. The DC cluster profiles further support the enhanced tumor immunity in TILs of the KO mice.

Discussion
Current study suggests that ATPIF1 is a new candidate of molecular target in the induction of antitumor activity of CD8 + T cells. The antitumor activities of CD8 + T cells were enhanced in the ATPIF1-KO mice, which represents a novel observation. ATPIF1 activity has been studied in several types of cells, but not yet in lymphocytes. In the knockout studies, ATPIF1 inactivation enhanced pro-survival activity of hepatocytes by inhibition of apoptosis in the respiratory chain toxification models 10 , and attenuated the pressure-induced cardiomyocyte hypertrophy 30 . The inactivation led to an increase in ATP production by OXPHOS in macrophages 31 and promotion of adipocyte differentiation 32 . In the overexpression studies, ATPIF1 activation reduces mitochondrial ATP production in the neuronal cells 11 , intestinal epithelial cells 13 and hepatocytes 14 . The activation leads to more ROS production by mitochondria for a high risk of pro-inflammatory response in the brain 11 and gut 13 . The activation raises the glycolytic activity in favor of hepatoma growth 14 . In the red blood cells, ATPIF1 is required for haem synthesis during the erythroblast differentiation in the hematopoietic models 12 . However, an impact of ATPIF1-deficiency or activation in the immune system has not been reported in the literature. Current study provides evidence that ATPIF1 inactivation led to an increase in the tumor immunity in the tumor-baring mice.
Deletion of CD8 + T cells abolished the antitumor immunity in the ATPIF1-KO mice. The data demonstrates a role of ATPIF1 in the regulation of CD8 T cells.
The enhanced activities of CD8 + T cells are supported by the cluster analysis of TILs with scRNA-seq and STARTRAC method, which reveal the subset landscape, the developmental trajectory, and TCRs clonal lineage in TILs. In this study, scRNA-seq was employed to determine the CD8 + T cell activity in TILs, in which eight major subsets were identified. In the KO mice, the T-proliferative and CD8 + T eff /T mem clusters were increased, while the naïve T cell cluster was decreased. The CD8 + T clusters were most active in 1 proliferation among the T cell subsets. To delineate the dynamic relationships of different T cell subsets, the STARTRAC indices were established in TILs to measure clone expansion.
The CD8 + T eff /T mem cluster was most active in the clone expansion. These suggest that activities of CD8 + T cells were enhanced in the proliferation and differentiation of T eff /T mem in the KO mice. These data support a role of ATPIF1 in the control of proliferation and development of CD8 + T cell subsets.
Our data reveals that energy metabolism was reprogrammed in CD8 + T cells of In contrast, ATPIF1 inactivation was reported to enhance tumor growth by inhibition of apoptosis in another study 10 . The exact reason for the different conclusions remains unknown in the two studies. The experiment conditions, such as glucose concentrations and model systems, may hold a promise to answer the question.
The glucose metabolism was reprogrammed in CD8 + T cells in the ATPIF1-KO mice.
The reprogramming was observed with an elevation in glucose uptake and glycolysis.
GLUT1 is the major glucose transporter in T cells for glucose uptake from the microenvironment 4 . The increased mRNA of GLUT1 led to an active glucose uptake in the ATPIF1-deficient cells. Glycolysis is required for T cell proliferation and production of cytokines including IFN-γ 20 . In current study, the enhanced glycolysis was identified in the KO cells by the Seahorse assay and the results were confirmed by the data of targeted metabolomics analysis. Glycolysis remains to be established in the differentiation of memory CD8 + T cells although OXPHOS is essential for the maintenance of CD8 + T mem cells 34 .
Constitutive glycolysis supports the differentiation of CD8 + T eff and enhances the antitumor efficacy of CD8 + T cells 35 . OXPHOS was also enhanced in CD8 + T cells of the ATPIF1-KO mice in the glucose metabolism. The reprogramming of glucose metabolism in both glycolysis and OXPHOS provides a basis for the activities of CD8 + T eff /T mem cells in the KO mice.
Mitophagy provides an alternative source of energy to T cells beyond glucose.
Mitophagy, a mitochondrion-specific autophagy, is induced by factors including ATP depletion to gain nutrients (fatty acids, amino acids, etc.) from the damaged mitochondria 36 (Fig. 9). The elevation in ATP synthase activity contributes to the metabolic reprogramming for glycolysis, OXPHOS and mitophagy in the glucose-limiting conditions. These data suggest that ATPIF1 protein inhibits the CD8 + cell activities in the physiological conditions. ATPIF1 represents a potential molecular target in the induction of CD8 + T cell activities for the cancer immunotherapy.

ATPIF1 -/mice
The ATPIF1 knockout mouse was generated with the CRISPR/Cas9 gene editing method with sgRNA PAM sequences targeting ATPIF1: SgRNA-1 GCAGTCGGATAGCATGGATACGG and SgRNA-2 GGCTCCACCAGCTTCTCGGATGG. A similar method was described in our published study 40 . For identification of ATPIF1 -/mice, PCR was conducted to amplify the ATPIF1 gene using the Forward primer CATCAGCCTTGGAATTCTGC and the Reverse primer CTTCGTCTCGGACTCGGTAG. The agarose electrophoresis was performed to determine the genotype, and the amplified PCR product was used in gene sequencing with ABI-3730XL to confirm the genotype. The ATPIF1 -/mice were maintained in the pathogen-free animal facility with free access to food and water, temperature of 20±2 °C,

Tumor growth and tumor-infiltrating T cells
Mice were implanted subcutaneously with 1×10 6 /mouse of B16-F0 melanoma Beijing Analytical Biosciences Technology Co., Ltd, which was described in detail in two studies 16,41 .

Glucose uptake
The CD3 + T cells were collected and cultured in the high glucose (HG

Targeted metabolomics
The CD8 + T cells were isolated from the spleen of tumor-bearing mice, which were inoculated with B16-OVA tumor for 2 weeks (n = 6), and frozen immediately with dry ice.
The energy metabolites were quantified using the LC-MS method. The analysis included 32 major metabolites of the tricarboxylic acid cycle (TCA), glycolytic pathway, pentose phosphate pathway and oxidative phosphorylation pathway. The hierarchical clustering and quantity of the metabolites are presented to show the change in lymphocytes.

MitoTracker assay
T cells were cultured in the high glucose (HG, 4.5g/L) medium, the low glucose (LG, 0.2g/L) medium and normal glucose (NG, 2g/L) medium. The mediums were prepared in glucose-free RMPI-1640 medium supplemented with glucose, 10% FBS, 1% Penicillin-Streptomycin solution. After culture for 24 h, the mitochondrial mass MitoTracker (Cat. M7514, Thermo Fisher Scientific) were determined according to the manufacturer's protocol with flow cytometry (BD FACS Calibur).

Coculture of OT-I CD8 + T cells with B16-OVA tumor
The ATPIF1 -/-OT-I CD8 + T cell and OT-I CD8 + T cells were isolated from the spleen of Then the IFN-γ-APC was stained for detection of intracellular IFN-γ expression. For T CM detection, the CD3, CD8, CD44 and CD62L antibody were used to stain the collected T cells, then the CD62L + CD44 + CD8 + T cells were indicated as T CM cells and CD62L -CD44 + CD8 + T cells were indicated as T EM cells. In the coculture process, the apoptosis was also analyzed using the Annexin V and 7-AAD. Firstly, the collected T cells were labeled with CD3-APC, CD8-FITC, Annexin V-PE and 7-AAD for detection of apoptosis according to the protocol, then loaded onto the BD Calibur FACS for analysis.

Data availability
All the raw data of scRNA-seq and the TCR sequencing were uploaded to the GEO with the accession number of GSE158278. (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE158278)

Statistical analysis
The quantitative data are presented as the mean ± SD. Unpaired two-tailed t-test was used in the data analysis for a difference between the WT and KO groups, and p value < 0.05 was considered statistically significant. FlowJo 10.0 was used to analyze the flow cytometry results.       (f) Transfer of ATPIF1 -/-CD8 + T cells to the B16-OVA xenografted Rag1 mice. Rag1 mice were inoculated with 2x10 6 B16-OVA melanoma cells to establish the tumor-baring model, 9 days later, 2x10 6 of ATPIF1 -/-OT-I or OT-I CD8 + T cells were intratumor injection to test the antitumor immunity. Nine days later, the mice were sacrificed for tumor size analysis (n=5). Data represents mean ± SD. *p<0.05, **p<0.01, ***p<0.001.     WT mice with CD8 antibody administration. Data indicated the CD8 T cells were effectively deleted after the administration of CD8 antibody. Figure S3. Sorting of CD45 + leucocytes from the tumor single cell suspension. The tumors excised from the WT and KO mice were digested to obtain the single cell suspension and then subject to sorting by the flow cytometry. The BV421 was used to sort the live and dead cells, BV421 negative cells was gated for the CD45 + sorting as indicated in WT and KO sample.     Tables  Table S1. Sample statistics after quality control during the 10×genomics single cell sequencing. Table S2. Summarized data in Excel format of the targeted metabolomics of the WT and KO CD8 + T cells.