Resting or naive T cells predominantly oxidize glucose-derived pyruvate, in addition to lipids and amino acids, via OXPHOS, to maintain quiescence and immune surveillance (39). Upon T cell activation, lipid oxidation is sharply reduced, and the cells rely instead on increased aerobic glycolysis to support extensive proliferation and Teff differentiation and function (14,40). At the end of an immune response, a small population of antigen-specific T cells survives to become long-lived memory T cells, which revert back to a metabolic program comparable with that of resting T cells (14,18,41). However, memory T cells exhibit an increased capacity for efficient energy generation, characterized by an increase in mitochondrial mass and, consequently, maximal mitochondrial spare respiratory capacity, which allows for the rapid and vigorous production of ATP upon repeat encounter with antigens (42,43). In addition to Teffs, activated T cells can differentiate into regulatory T cells (Tregs), which serve a critical role in self-tolerance and immunosuppression (3,44,45). Unlike Teffs, Tregs primarily use glucose-derived pyruvate and fatty acids to efficiently produce ATP through the tricarboxylic acid cycle and lipid β-oxidation (46). Distinct metabolic programs are required for functionally different T-cell lineage differentiation and commitment (47,48). A comprehensive study by Michalek et al demonstrated that Teffs, including T helper (Th)1, Th2 and Th17 cells, were selectively increased in glucose transporter (GLUT)1 transgenic mice, and were dependent on a highly glycolytic metabolism (48). Tregs, by contrast, expressed a low level of GLUT1 and relied on high rates of lipid oxidation (48). Gene array analysis on CD8⁺ cytotoxic T cells under conditions of glucose deprivation or incubated in the presence or absence of 2-deoxy-D-glucose, which inhibits glycolysis, demonstrated that multiple key gene expression events and effector functions were selectively inhibited, including the production of interferon-γ (IFN-γ), cell cycle progression and cytolytic activity (49). Consistently, impaired T-cell metabolism directly contributed to T-cell dysfunction and exhaustion in leukemia, whereas the genetically increased expression of GLUT1 and hexokinase 2 (HK2) may partially restore T-cell function (50). The upregulation of glycolysis by the transgenic overexpression of GLUT1 or glycolytic genes was sufficient to augment T-cell activation, ultimately resulting in lymphadenopathy and a systemic lupus erythematosus-like autoimmunity in aging mice (17).

The identification of transcription factors potentially responsible for the metabolic reprogramming upon T-cell activation revealed c-Myc and HIF-1α as two of the top-ranked candidates, as both were found to be induced at the mRNA and protein levels upon T-cell stimulation (21). c-Myc specifically upregulates the expression of all glycolytic genes, including GLUT1, lactate dehydrogenase type A (LDHA), HK2 and pyruvate kinase muscle isoform 2 (PKM2). Subsequently, the acute genetic deletion of c-Myc markedly inhibits the upregulated glycolytic activity. In addition, an HIF-1α-mediated glycolytic switch regulates the balance of Th17/Treg differentiation (22,51). Th17- but not Treg-polarizing conditions elicited a HIF-1α-dependent acceleration of glycolysis via upregulation of glycolytic enzyme expression. By contrast, the inhibition of glycolytic metabolism resulted in the inhibition of Th17 differentiation and promotion of Treg development. Upon investigation of the underlying molecular mechanism, HIF-1α was found to be selectively induced in Th17 differentiation through the mammalian target of rapamycin (mTOR) signaling pathway, whereas the deficiency of HIF-1α led to decreased Th17 commitment but enhanced generation of Treg, which protected mice from experimental autoimmune encephalomyelitis (22).

PI3K/Akt is activated by various stimuli in T lymphocytes, including T cell antigen receptor, costimulatory molecules, cytokine receptors and chemokine receptors (23,24,52), and PI3K/Akt signaling serves a fundamental role in T-cell activity. For example, the trafficking of GLUTI to the cell surface and prevention of internalization in T cells are promoted by Akt (53). Of note, mTOR, as a downstream target of Akt, is activated by Akt and serves a key role in linking the activation of PI3K to Th-cell differentiation (54,55). mTOR is a catalytic unit of two distinct multi-protein assemblies, referred to as mTOR complex (mTORC)1 and mTORC2. The activation of mTORC1 can initiate a signaling cascade, which leads to metabolic reprogramming characterized by increased aerobic glycolysis. Of note, mTOR differentially regulates Teff and Treg lineage commitments through the activation of specific signal transducer and activator of transcription pathways and, consequently, the induction of lineage-specific transcription factors (54). By contrast, rapamycin treatment, which targets mTORC1, has been shown to exhibit an inhibitory effect on glycolytic switching upon T-cell activation (56). AMP-activated protein kinase (AMPK), as a well-known evolutionarily conserved energy sensor, is activated by an increased AMP/ATP ratio and acts in opposition to mTORC1 to maximize energy production via promoting mitochondrial phosphorylation (57). AMPKα1^−/− T cells exhibit an impaired ability to transit from an anabolic and glycolytic metabolism to a catabolic and lipid oxidative state under metabolic stress (58).

Accumulating evidence has confirmed that the MYC oncogene, the PI3Ks/Akt/mTOR pathway and HIF-1 (62,63), in addition to tumor suppressor p53, are implicated in the metabolic reprogramming of tumor cells (64–66) (Fig. 1). The MYC oncogene encodes a transcription factor, c-Myc, which links altered cellular metabolism to tumorigenesis (67). Generally, c-Myc directly and/or indirectly regulates the expression of genes involved in glucose, glutamine and nucleotide metabolism. For example, glycolytic genes, including LDHA are directly upregulated by c-Myc (68); however, c-Myc can repress the expression of microRNA-23a/b to indirectly promote the protein expression of glutaminase and metabolism of glutamine (69). The depletion of c-Myc has been shown to result in the repression of several genes encoding enzymes rate-limiting for deoxyribonucleoside triphosphates (dNTPs) metabolism, including thymidylate synthase, inosine monophosphate dehydrogenase 2 and phosphoribosyl pyrophosphate synthetase 2. The depletion of c-Myc also leads to a decrease in dNTPs and inhibited cell proliferation (70). A number of glycolytic genes have been documented to be directly regulated by c-Myc in screens for c-Myc target genes, including GLUT1, HK2 and muscle phosphofructokinase (71,72). In addition, c-Myc may cooperatively serve a pivotal role in hypoxic adaptation with HIF-1 through upregulating pyruvate dehydrogenase kinase 1 under non-normoxic conditions, thereby accelerating glycolytic metabolism by favoring the conversion of pyruvate to lactate and suppressing mitochondrial oxidative respiration (72–74).

The PI3K/Akt/mTOR signaling pathway has been found to be activated at a high level and contribute to the metabolic transformation of tumors (75,76) (Fig. 1). Akt, a serine/threonine kinase, has been shown to be constitutively activated in tumor cells through the amplification of PI3K, which phosphorylates membrane-associated phosphatidylinositol 4,5-bisphosphate (PIP2) to generate phosphatidylinositol 3,4,5-trisphosphate as an upstream activator of Akt (77). Human glioblastoma cells with constitutive Akt activity exhibit high rates of aerobic glycolysis through the direct effect of Akt on glucose metabolism, including upregulating the expression and/or localization of glucose transporters and glycolytic enzymes, including GLUT1 and HK2 (75). Akt also activates mTOR, which also contributes to the glucose metabolic reprogramming of tumor cells (78,79). mTOR is also an upstream activator of HIF-1 and c-Myc in tumor cells, and high levels of Akt and mTOR activity lead to high HIF-1 activity and adaption to hypoxia (78,79).

Hypoxia is a common feature of various malignant tumor cells (59), and the adaption of tumor cells to hypoxia is predominantly mediated by HIF-1, a key transcription factor that regulates nine of 10 enzymes involved in glycolytic energy metabolism (80). In addition, an accumulation of pyruvate and lactate derived from high rates of aerobic glycolysis may promote hypoxia-inducible gene expression independently of hypoxia via stimulating the induction and stability of HIF-1α (81). Similarly, p53, which is mostly known for its tumor suppressor properties, is implicated in the metabolic adaption of tumors through decelerating glycolysis and accelerating mitochondrial oxidative respiration (82,83). As a ubiquitin ligase, mouse double minute 2 homolog (Mdm2) may be phosphorylated by Akt and mediates the ubiquitylation and proteasome-dependent degradation of p53 (84).

In conclusion, relevant transporters and receptors on tumor cells integrate signals from growth factors, cytokines and nutrient availability in the tumor microenvironment to activate the PI3K/Akt/mTOR signaling pathway, which regulates the expression of various transcription factors, including c-Myc, HIF-1 and p53, leading to the reprogramming of glucose metabolism in tumors. Of note, a reciprocal interaction exists between molecular signaling pathways regulating c-Myc, HIF-1 and p53, forming a complicated and intricate regulatory network controlling the metabolic switch in tumors (85).

It is likely that intense nutrient competition exists between tumor cells and TILs in the tumor microenvironment, as tumor cells may deplete nutrients due to their dependence on enhanced aerobic glycolysis (10). This tumor-imposed glucose restriction may lead to TIL dysfunction due to reduced glucose uptake and metabolic reprogramming. In an established mouse model of progressing and regressing tumors, the progressing tumors exhibited higher rates of glycolytic activity compared with regressing tumors, suggesting that progressing tumors consume more glucose. Consistently, T cells in progressing tumors exhibited decreased phosphorylation of eukaryotic translation initiation factor 4E-binding protein 1 (4EBP1) and S6 kinase compared with that in regressing tumors. However, Tregs and M2 macrophages, neither of which depend on enhanced aerobic glycolysis, but rather on fatty acid oxidation, are unaffected by the glucose-depleted tumor microenvironment, and may suppress the antitumor immune response.

In addition to tumor-imposed nutrient limitation for tumor-infiltrating immune cells, metabolites, including the enhanced lactate production by tumors due to increased dependence on glycolysis (59,62,89), are suggested to be key metabolic components in the communication between tumor cells and tumor-infiltrating immune cells (34,90). It is important that the excess cellular lactate produced by tumors is exported, mainly by MCTs (MCT1/MCT4), in order to prevent acidosis in tumor cells, which leads to the accumulation of lactate in the tumor milieu (91). Lactate has been shown to promote cancer cell stemness (92) and metastasis (93) through the increased production of several tumor progression-promoting factors, including transforming growth factor (TGF)-β (94,95), hyaluronic acid and CD44 (96). In addition to its direct effect on tumor cells, lactate act as an immunomodulator, mediating the immune evasion of tumor cells (97). The exogenous lactate treatment of natural killer (NK) cells inhibited their cytotoxicity directly and indirectly by increasing the number of myeloid-derived suppressor cells (MDSCs), which may repress NK cell function (12,98). By contrast, LDHA-deleted pancreatic cancer cell xenografts, with a defect in lactate production, exhibited improved cytolytic function of NK cells in C57BL/6 mice, with higher expression of perforin and granzyme and a decreased number of MDSCs in the spleen (12,98). Furthermore, the immunomodulatory effects of lactic acid have been demonstrated not only for dendritic cells, but also for T cells (13). However, the export of excess lactic acid by activated T cells is inhibited due to the lactic acid gradient between the cytoplasm and extracellular milieu, due to the accumulation of lactic acid secreted by surrounding tumor cells with high rates of glycolysis. To conclude, the accumulation of tumor-derived lactate in the extracellular milieu may lead to a metabolic obstruction in cytotoxic T lymphocytes (CTLs); subsequently, CTLs become hyporesponsive, exhibiting decreased proliferation, cytokine secretion and cytotoxic activity (13,99,100). Tregs and M2 macrophages are not affected by the presence of high levels of lactate, as their distinct metabolic program relies mainly on fatty acid oxidation rather than aerobic glycolysis (48).

As MCT1 and MCT4 are H⁺/lactate symporters, lactate efflux via MCT1/MCT4 is accompanied by H⁺ transport, which eventually creates an acidic microenvironment (97). Extracellular acidosis is characteristic of the tumor micromilieu, and the local acidification allows tumor cells to be aggressive via increased extracellular matrix degradation and enhanced survival and metastasis (101,102). In addition to the promoting effect of extracellular acidosis on tumor progression, the acidic microenvironment is important in the immune evasion of tumor cells mediated by immune cell dysfunction. Interferon (IFN)-γ, as a critical mediator of the differentiation of Th1 and Th2 cells, promotes Th1 polarization and inhibits Th2 differentiation (103). However, IFN-γ is acid-labile and likely to be denatured in an acidic milieu, which diverts Th-cell differentiation from the antitumor Th1 lymphocytes towards the tumor-promoting Th2 lymphocytes (104). HIF-1, a key transcription factor predominantly mediating the adaptation of tumor cells to hypoxia, may not be degraded under sufficient oxygen supply (105), thereby contributing to HIF-dependent glycolytic reprogramming in tumor cells and, consequently, decreased antitumor immunity.

Given the profound effect of HIF-1 on gene regulation, T-cell differentiation is likely controlled by HIF-1 (15,22). Furthermore, TGF-β1 may stabilize HIF-1 through the inhibition of prolyl hydroxylase 2 under hypoxic conditions (106). A screening of key transcription factors for T-cell differentiation during inflammatory hypoxia of the mucosa revealed forkhead box (FOX)P3 as a direct target of HIF-1, and it has also been demonstrated that the hypoxic induction of FOXP3 and accumulation of Treg require both HIF-1 and TGF-β1 (107). HIF-1 has also been identified as a decisive factor in T-cell differentiation to Th17 or Tregs by promoting Th17-polarization and inhibiting Treg differentiation (22,51). TGF-β1 may be induced in hypoxia (108), and is also key role the differentiation of Th17 and Tregs (109). It is reasonable to hypothesize that the Th17/Treg balance is an integral outcome of HIF-1, TGF-β1 and inflammatory cytokine interplay in the local tumor milieu (22,51,107). The exposure of human breast and prostate cells, and mouse melanoma and mammary carcinoma cells to hypoxia resulted in the upregulation of programmed cell death ligand-1 (PD-L1), an important immunoinhibitory molecule, on the surface of tumor cells in an HIF-1-dependent manner, which eventually contributed to tumor cell evasion from antitumor immunity via the increased apoptosis of CTLs due to the enhanced PD-1/PD-L1 interaction (110). Although the mechanisms underlying increased the expression of PD-L1 on tumor cells remain to be fully elucidated, the hypoxia-induced upregulation of PD-L1 on tumor cells may represent a novel mechanism responsible for hypoxia-mediated immune evasion of tumor cells. In addition to the previously known inhibitory effects of TGF-β1 on T-cell differentiation and function, TGF-β1 has also been found to mediate T-cell hyporesponsiveness, in part through the enhanced expression of PD-1 on tumor-infiltrating antigen-specific T cells induced by mothers against decapentaplegic homolog 3 (Smad3)-dependent signaling (111,112).

In addition to non-metabolic molecular mechanisms responsible for the tumor progression associated with the upregulation of CD147, a metabolic molecular basis has become a focus of investigations. Blocking CD147 with a targeted monoclonal antibody or silencing CD147 by small interfering RNA has been shown to result in a marked decrease in glycolytic energy metabolism (117,118). Consistently, a study by Huang et al demonstrated that CD147 acts as an important regulator of cell proliferation through promoting glucose metabolic reprogramming by the post-transcriptional inhibition of p53 via the activation of PI3K/Akt/Mdm2 signaling promoted by MCT1-induced lactate export in HCC (119) (Fig. 1). CD147 is increasingly recognized as being implicated in glucose metabolism reprogramming in tumors through gain/loss-of-function studies (92,120).

According to the regulation of CD147, it is well established that the tumor micromilieu serves an important role in the overexpression of CD147 on the surface of tumor cells. As is known, TGF-β1 induces epithelial-to-mesenchymal transition (EMT) via Smad-dependent and -independent signaling pathways (121–123). A study by Wu et al demonstrated a positive correlation between the expression of CD147 and typical EMT markers, revealed that CD147 is a Slug target gene, and demonstrated that the upregulation of CD147 involves activation of the PI3K/Akt-GSK3β-Snail-Slug signaling pathway through the stimulation of TGF-β1 (124). A series of transcription factors have been found to be implicated in the fundamental metabolic adaptation of tumors to hypoxia, among which HIF-1 is critical to this process (125,126). HIF-1 is a heterodimer that consists of a constitutively expressed HIF-1β subunit and an oxygen-sensitive HIF-1α subunit (127). Under hypoxic conditions, HIF-1α binds to a conserved DNA consensus, referred to as hypoxia-responsive element (HRE), on the promoters of numerous hypoxia-responsive genes. There are two HIF-1-binding sites and three specificity protein 1 (SP1)-binding sites in the 3′ and 5′ flanking regions of the CD147 gene, respectively (128,129). Consistently, a genome-wide chromatin immunoprecipitation-on-chip assay and immunohistochemical staining identified CD147 as a novel hypoxia-responsive molecule. The identification of key molecules engaged in epithelial solid tumor glycolytic switch confirmed that the upregulation of CD147 was mainly mediated through the combined effect of HIF-1α and SP1 on activation of the CD147 promoter (120). A study by Kong et al also reported that upregulation of the expression of CD147 was mediated by promoter hypomethylation through increased SP1 binding in human HCC and lung cancer (127,130).

In conclusion, the signaling pathways responsible for the overexpression of CD147 on tumor cells mainly include TGF-β1 and HIF-1α, which are pivotal in tumor glycolysis and immunosuppression.

Apart from tumor cells themselves, manipulating the metabolic activity of immune cells to prevent immune cell hyporesponsiveness in tumors is currently considered a promising approach in cancer therapy (11,19,20). The fundamental principle of modulating the metabolism of T cells is to favor anabolic glycolysis rather than catabolic oxidative respiration. Specific antibodies against nutrient transporters have been confirmed as potential pharmacological agents targeting T-cell metabolism. For example, blockade of GLUT1 on T cells has been found to decrease glucose uptake, thus leading to T-cell dysfunction (17). The blocking of co-inhibitory receptors has been suggested as a promising immunotherapy option for enhancing antitumor immunity to eliminate tumor cells (131–134). Cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and PD-1, two well-known inhibitory receptors on T cells, are induced upon T-cell activation to control and moderate excessive immune responses, acting as checkpoints (131). However, CTLA-4 and PD-1 signaling have been shown to restrict T-cell activation and function by downregulating aerobic glycolysis (10,135). Therefore, it is hypothesized that checkpoint blocking may relieve the suppression of antitumor immunity, in part through remodeling T-cell metabolic programming to enhance nutrient uptake and glycolytic metabolism, consequently restoring their capacity to kill tumor cells (11). PD-L1 also has a PD-L1 and PD-1 interaction-independent metabolic function (111,136). PD-L1 on tumor cells is important for Akt/mTOR signaling, which in turn increases the rate of glycolysis through promoting the translation of glycolytic enzymes. Blocking PD-L1 may directly decrease glycolysis in tumors, increasing the nutrient availability in the extracellular tumor milieu for infiltrating lymphocytes (10,11).

In general, therapeutic drugs targeting the metabolic adaptation of tumor cells may be divided in two categories, namely direct and indirect. Indirect drugs target aberrant signaling pathways relevant to metabolic transformation in tumor cells, including the HIF-1α (137), c-Myc (67,138), PI3K/Akt/mTOR (139–141) and AMPK (142) pathways. For example, metformin, a drug originally designed to treat patients with type 2 diabetes (143), may activate the AMPK signaling pathway to oppose mTORC1, subsequently decreasing glycolytic metabolism and increasing OXPHOS in tumor cells to control tumor progression (144,145). Consistently, patients with type 2 diabetes who were treated with metformin were more likely to remain cancer-free over 8 years compared with those who received other treatments (146,147). Metformin is currently in phase I and II clinical trials for cancer therapy. Direct drugs comprise antagonists against multiple metabolic enzymes and several metabolites in glucose, amino acid, lipid and nucleotide metabolism (63). This review focuses on glycolytic metabolism. Almost all enzymes involved in every stage of glycolysis may represent potential targets, particularly tumor-specific enzyme isoforms and glycolytic metabolites, including PKM2 (129,148) and lactic acid (90,102). In terms of CD147, it has been reported in patients with HCC that targeted radioimmunotherapy with ¹³¹I-labeled HAb18 F(ab′)2 metuximab monoclonal antibody injection (licartin), which is a radiolabelled anti-CD147 monoclonal antibody, effectively prevented the recurrence and metastasis of HCC following hepatectomy and liver transplantation (149). Based on the evidence described above, it is reasonable, to a certain extent at least, to attribute the antineoplastic capacity of licartin to its ability to inhibit the glycolytic metabolism of HCC cells. Combination therapy of ¹³¹I-labeled metuximab and other metabolic transformation-targeting drugs may be more beneficial for antitumor treatment compared with monotherapy.

Regardless of the modulation of cellular metabolism in tumor cells or T cells, targeted delivery of specific drugs in the body is crucial for preventing off-target effects. Transporter-facilitated drug uptake (150,151), bi-specific antibodies (152,153) and nanoparticle-mediated delivery (154,155) have been developed to optimize drug efficacy. Optimal Teff function in the tumor microenvironment is necessary for successful adoptive T-cell immunotherapy. Combination treatment comprising metabolic intervention and adoptive T-cell immunotherapy appears promising for metabolic reprogramming of T cells to exert effective antitumor immunity (19,20). As reported previously, CD147 serves an important role in the reprogramming of glucose metabolism and cell proliferation in HCC cells (119), and a targeted radiolabeled anti-CD147 monoclonal antibody (licartin) effectively prevented the recurrence and metastasis of HCC following hepatectomy and liver transplantation (149). The blocking of CD147 inhibited the enhanced glycolysis of HCC cells and contribute to improved antitumor immunity in the tumor microenvironment, which is exactly what current endeavors are aiming to prove. Combination therapy comprising CD147 intervention and tumor immunotherapy is likely to lead to more marked antitumor effects than monotherapy. However, the timing of adoptive Teffs entering the local tumor milieu is an important issue requiring consideration. Using inhibitors of glycolysis prior to the adoptive transfer of T cells may assist in remodeling metabolic function in T cells in a hospitable tumor milieu with nutrient repletion (19,20).