Abstract
Solute carriers of the glucose transporter (GLUT) family mediate the first step for cellular glucose usage. The upregulation of GLUTs has been reported in numerous cancer types as a result of perturbation of gene expression or protein relocalization or stabilization. Because they enable to sustain the energy demand required by tumor cells for various biochemical programs, they are promising targets for the development of anticancer strategies.
Recently, important biological insights have come from the fine crystal structure determination of several GLUTs; these advances will likely catalyze the development of new selective inhibitory compounds. Furthermore, deregulated glucose metabolism of nontumor cells in the tumor mass is beginning to be appreciated and could have major implications for our understanding of how glucose uptake by specific cell types influences the behavior of neighboring cells in the same microenvironment.
In this review, we discuss some of the deregulation mechanisms of glucose transporters, their genetic and pharmacological targeting in cancer, and new functions they may have in nontumor cells of the tumor environment or beyond glucose uptake for glycolysis.
Abbreviations
- 18F‐FDG
- 2‐Deoxy‐2‐[fluorine‐18] fluoro‐D‐glucose
- CAFs
- cancer‐associated fibroblasts
- CRC
- colorectal cancer
- DERL3
- Derlin‐3
- DHA
- dehydroascorbic acid
- EGFR
- epidermal growth factor receptor
- EMT
- epithelial‐mesenchymal transition
- GFPT
- glutamine‐fructose‐6‐phosphate‐aminotransferase
- GLUT
- glucose transporter
- HBP
- hexosamine biosynthesis pathway
- IDH
- isocitrate dehydrogenase
- LUAD
- lung adenocarcinoma
- mTOR
- mammalian target of rapamycin
- NSCLC
- nonsmall cell lung cancer
- PET
- positron emission tomography
- PF4
- platelet factor 4
- PHD2
- prolyl hydroxylase domain‐containing protein 2
- PI3K
- phosphatidyl‐inositol 3‐kinase
- PON2
- paraoxonase 2
- SGLT
- Na+/glucose cotransporter
- SLC
- solute carrier
- TXNIP
- thioredoxin‐interacting protein
- VEGF
- vascular endothelial growth factor
- α‐KG
- α‐ketoglutarate
Introduction
Solute carriers (SLC) are integral membrane transport proteins consisting of > 400 SLC family members, which carry nutrients, ions or other metabolites across cell or organelle membranes [1]. The vast majority of its individual members are still poorly characterized. Hence, a profound understanding of their structure, regulation, and function in cancer could lead to the discovery of drugs that efficiently interfere with their activity, to counteract malignant progression [2].
Within the SLC superfamily, glucose transporters enable the entry of a major nutrient source, glucose, across the hydrophobic cell membrane. These solute carriers form essentially two families, the secondary active Na+/glucose cotransporter (SGLT, gene family name solute carrier SLC5A) and the facilitative sugar transporter (GLUT, gene family name SLC2A). Additionally, another transporter, SWEET (SLC50A1) has been reported, however, its role in human cells remains enigmatic; in contrast, SWEET homologues are better characterized in plants, where they mediate glucose efflux in connection to symbionts and pathogens [3]. The SGLT family can be further divided into sugar transporters (SGLT1 and SGLT2) or sensor (SGLT3) [4]. The physiology and tissue expression of these transporters have been recently reviewed [5].
In this review, we focus on the facilitative glucose transporter family, which comprises fourteen members in humans. The 14 GLUT proteins can be categorized into three classes according to their sequence similarity: Class 1 (GLUTs 1–4, 14); Class 2 (GLUTs 5, 7, 9, and 11); and Class 3 (GLUTs 6, 8, 10, 12, and 13/HMIT). The amino acid sequences show 39–65% identity between the well‐characterized human GLUT1‐5, and at least 28% identity between GLUT1 and all other GLUTs [6]. The different GLUTs play a key role during embryo development, and have different affinities for glucose or other sugars such as fructose or mannose [6]. Their distribution across tissues reveals a tissue‐dependent GLUT expression. For example, the brain that strongly relies on glucose to support metabolic demands mainly expresses two high affinity glucose transporters, GLUT1 and GLUT3 [7, 8]. Certain tissues express multiple GLUTs, exemplified by the muscle, where the expression of GLUT3‐5 and [10, 11] has been reported [9-12].
In the context of cancer, tumor cells often increase their glucose consumption and lactate production even in the presence of physiological oxygen concentrations and functional mitochondria; this aerobic glycolysis is known as the Warburg effect [13]. In recent years, an increasing number of studies are identifying GLUT1 and GLUT3 as the preeminent actors in the accelerated metabolism. Although glucose intermediary metabolism has been the subject of intense investigations, glucose uptake and the underlying mechanisms of GLUT regulation and function remain less known.
In this review, we concentrate on the four following aspects, which we believe constitute exciting current and future research explorations: (a) mechanisms of regulation, focusing on GLUT1 and GLUT3 whose recent research provided new insights into their importance in cancer, (b) glucose transporter targeting, (c) the role of glucose transporters in the tumor microenvironment, and (d) possible functions for glucose carriers beyond glucose uptake for glycolysis.
Mechanisms of GLUT1 and GLUT3 regulation
Tumor cells are known to have accelerated metabolic rates and high glucose demand in a nutrient‐poor environment. The combination of these factors may result in a metabolic dependence on a continuous energy and nutrient supply for cells within the tumor mass. This is often associated with deregulated expression of glucose carriers, particularly GLUT1 and GLUT3. These two transporters share 64% amino acid sequence identity and display high affinity and maximum turnover number.
Recently, several studies took advantage of publicly available gene expression datasets, such as The Cancer Genome Atlas (TCGA) or Oncomine, to highlight an increased expression of GLUT family members in different types of cancer. In hepatocellular carcinoma, GLUT2 expression is more elevated than that of other GLUTs and a strong expression correlates with poor overall survival [14]; in bladder cancer, GLUT3 is overexpressed in muscle‐invasive compared to noninvasive tumors [15]; in glioblastoma, GLUT3 expression is elevated in aggressive compared to lower grade lesions, and is associated with poorer overall survival [16]. More generally, high expression of GLUT1 and/or GLUT3 is associated with poor survival in most cancer types interrogated, including colorectal carcinoma, breast carcinoma, lung adenocarcinoma, squamous cell carcinoma, ovarian carcinoma, and glioblastoma [16, 17]. As we discuss below, deregulated GLUT1 and GLUT3 expression occurs via a direct interaction with gene regulatory elements, or by controlling the cellular trafficking of the transporters to the membranes.
GLUT1 regulation
Although GLUT1 and GLUT3 share some mechanisms of regulation, for example in response to HIF‐1α or p53, they often respond to many distinct stimuli (Fig. 1). Indeed, GLUT1 was found to be activated as a response to HIF‐1α in glioblastoma stem cells [18]. Additionally, other hypoxia‐associated factors, such as VEGF receptor and calcium channel transactivation, were described to upregulate GLUT1 synthesis and trafficking to the cellular membrane [19].
At the transcriptional level, GLUT1 (SLC2A1) expression is controlled by a number of different mechanisms. For example, the transcription factors c‐Myc (encoded by the well‐known MYConcogene) and sine oculis homeobox 1 (SIX1) stimulate glycolysis by direct transactivation of GLUT1 and other glycolysis‐related genes [20, 21]. GLUT1 expression was found to be increased in Burkitt's lymphoma cell lines, which are characterized by chromosomal translocations of MYC [20]. In nonsmall cell lung cancer (NSCLC), GLUT1 expression was significantly positively correlated with mutation in each oncogene EGFR and KRAS [22]. Interestingly, oncogenic mutant Kras copy number is a critical determinant of a metabolic shift, as KrasG12D/G12D cells are more glycolytic than KrasG12D/WT or KrasWT/WT cells [23]. Additionally, Glut1 was upregulated upon oncogenic Kras activation in bronchoalveolar stem cells, and lung tumor development was accelerated in chronic hyperglycemia induced by subtoxic doses of streptozotocin [24]. In cells from colorectal cancer (CRC), GLUT1 expression was higher in KRAS or BRAF mutant compared to isogenic cell lines expressing the wild‐type proto‐oncogenes, conferring them a survival advantage in low glucose conditions. In contrast, when cells with wild‐type KRAS were exposed to low glucose, the few surviving cells had upregulated GLUT1 expression, and a minority of them had even acquired KRASmutations, highlighting the selection pressure imposed by glucose deprivation [25].
Although only few factors have been described to impact GLUT1 expression negatively, a striking example is that of thioredoxin‐interacting protein (TXNIP). This multifunctional protein was shown to repress GLUT1 function by two ways: through binding to it, leading to a clathrin‐dependent endocytosis, and through reducing GLUT1 mRNA expression, in HepG2 liver cancer cells. Upon energy stress, or in the presence of insulin, AMPK [26] and Akt [27]phosphorylate TXNIP, respectively, resulting in its degradation thereby elevating GLUT1‐dependent glucose uptake and restoring energy homeostasis. Another negative regulator of GLUT expression is tumor suppressor p53, which was shown to inhibit each of GLUT1 and GLUT3 transcription [28, 29]. Furthermore, in cells from pancreatic ductal adenocarcinoma, p53 was shown to repress the transcription of paraoxonase 2 (PON2). This leads to a reduction of PON2 protein, augmenting the interaction between GLUT1 and the membrane protein stomatin, which results in a decreased glucose uptake [30, 31].
Epigenetic studies have contributed a better understanding of GLUT1 regulation. In the mouse brain, it was observed that fasting‐induced production of the ketone body β‐hydroxybutyrate enhances expression of Glut1. This was correlated with an increase of H3K9 acetylation at a critical cis‐regulatory region of the Glut1 gene. This was further explored through CRISPR/Cas9‐mediated disruption of Hdac2, which increased Glut1 expression [32].
A few microRNAs have been shown to regulate GLUT1: the decreased expression, in tumors compared to healthy tissue, of miR‐144 and miR‐132, in lung and prostate cancer, respectively, resulted in an increased glucose uptake and glycolysis, which was attributed to GLUT1 de‐repression [33, 34]. In CD46 costimulated CD4+ T cells, a downregulation of miR‐150 and GLUT1 upregulation were observed, identifying GLUT1 as a target of this microRNA in T cells [35].
DNA methylation was also found to regulate GLUT1, albeit indirectly: transcriptional silencing of the putative tumor suppressor DERL3 (Derlin‐3) was demonstrated through its promoter CpG island hypermethylation, which correlated with shorter relapse‐free survival in CRC and resulted in a Warburg effect in HCT‐116 colon cancer cells. In the same cellular system, stable isotopic amino acid labeling identified GLUT1 as a target of degradation mediated by DERL3, offering a molecular explanation for the metabolic consequences of DERL3 loss, and the increased GLUT1 expression in tumors with DERL3 hypermethylation [36].
Modulation of GLUT1 trafficking is another important way to control glucose uptake and has been reported in cancer cells. Although wild‐type p53 represses GLUT1 expression (see above), tumor‐associated p53 mutation stimulates the Warburg effect by promoting GLUT1 protein translocation to the plasma membrane. This depends on actin polymerization mediated by the small GTPase RhoA and its downstream kinase ROCK [37]. Interestingly, a gain‐of‐function of mutant p53 was linked to its stimulation of the Warburg effect, as GLUT1 knockdown inhibited mutant p53‐mediated anchorage‐independent cell growth and xenograft tumor development [37].
Other factors were also linked to GLUT1 cellular trafficking. In hematopoietic cells, growth factor interleukin‐3 helps maintain GLUT1 at the cell surface, aided by a small GTPase Rab11a‐dependent recycling of intracellular GLUT1 [38]. Furthermore, the phosphatidyl‐inositol 3‐kinase (PI3K) effector Akt kinase activity augments upon interleukin‐3 stimulation, and enhances both GLUT1 gene transcription and GLUT1 protein trafficking to the cell membrane [38-41]. Interestingly, it was found in T cells that transgenic expression of Glut1 increased T cell activation, and led to the accumulation of activated memory‐phenotype T cells with signs of autoimmunity in aged mice [42]. Conversely, tumor suppressor PTEN, which removes the phosphate on position 3 in the inositol ring of phosphatidyl‐inositol(3,4,5) tri‐phosphate – catalyzed by PI3K – interacts with SNX27 and thereby prevents binding between SNX27 and VPS26, impairing the formation of a functional retromer complex, ultimately diminishing GLUT1 recycling to the plasma membrane. Surprisingly, this effect was independent on its phosphatase activity [39]. Altogether, these studies highlight a tight control of GLUT1 expression and localization by growth factor‐mediated PI3K‐Akt signaling, in diverse cell types including hematopoietic and cancer cells.
GLUT3 regulation
Similar to GLUT1, different factors have been identified to regulate GLUT3 in the context of cancer. The use of actinomycin‐D and cloning of the human GLUT3 promoter enabled to determine that the second messenger cAMP synthetic analogue 8‐br‐cAMP regulates transcriptionally GLUT3 in a breast cancer cell line [43]. In colon cancer cell lines, caveolin‐1 (CAV1) triggers a nuclear localization of HMGA1, which stimulates GLUT3 transcription by binding to a DNA sequence located within the GLUT3 promoter [44]. Increasingly, cancer‐related signaling pathways are shown to impact glucose uptake via GLUT3 modulation: in Tsc1‐ or Tsc2‐null mouse embryonic fibroblasts and human cancer cell lines, the mammalian target of rapamycin (mTOR) pathway upregulates GLUT3. Specifically, the rapamycin‐sensitive mTOR complex 1 (mTORC1) increases GLUT3 expression through NF‐κB pathway stimulation [45]. Additionally, increased NF‐κB signaling, resulting from p53 deletion in mouse embryonic fibroblasts, stimulates Glut3 transcription by binding to a specific NF‐κB binding element located within the first intron of the mouse Glut3 gene [29]. In 2014, a link between oncogenic EGFR signaling and GLUT3 has also been demonstrated in lung adenocarcinoma (LUAD) cells, whereby tumor‐associated EGFR mutations, which occur in 15–30% LUAD, confer a gain‐of‐function of the encoded mutant receptor tyrosine kinase leading to an increased glucose uptake and glycolytic behavior. In drug‐sensitive lung tumor cells, EGFR tyrosine kinase inhibitors attenuated glucose consumption and lactate production, and inhibited GLUT3 but not GLUT1 expression [46]. Activation of the PI3K‐Akt‐mTOR signaling pathway by IGF‐I leads to GLUT3 induction through HIF‐1α in neuronal PC12 cells, involving a hypoxia‐response element (HRE) in the GLUT3 promoter [47]. GLUT3 expression has also been correlated with OCT4, a protein exhibiting cooperative interactions with HIF‐1α and acting as a pluripotency marker [48]. This is interesting, as the GLUT3genomic region is located in the same 200‐kb gene cluster as the embryonic stem cell gene NANOG, and interacts in cis with the Nanog promoter in pluripotent cells [49]. In germ cell tumors, NANOG, GLUT3 and other genes of this cluster are coregulated, and their levels are higher in less differentiated tumors [50]. These studies suggest a particular mechanism whereby GLUT3 is regulated concomitantly with genes linked to the embryonic stem cell program, which could have implications in cancer stem cells or tumor cell dedifferentiation.
In cancer cells, in addition to promoter activation, an enhancer located in the second intron of human GLUT3 is critical for gene induction in response to the activity of different transcription factors. Indeed, we identified that the epithelial‐mesenchymal transition (EMT) transcription factor ZEB1 directly binds and activates this intronic enhancer in a panel of human NSCLC and hepatocellular carcinoma cell lines. Consequently, inhibiting GLUT3 expression reduces glucose uptake and the proliferation of mesenchymal lung tumor cells, while its ectopic expression in epithelial cells sustains their proliferation in low glucose [51]. The importance of the same GLUT3 intronic enhancer was further highlighted in another study using tumor cells from a pediatric liver cancer, hepatoblastoma, whose principal molecular alteration is a frequent gain‐of‐function mutation of CTNNB1, coding for β‐catenin. We demonstrated a direct binding of TCF4/β‐catenin to a DNA sequence in intron‐2, resulting in strong expression of GLUT3 in this tumor type. GLUT3 was particularly elevated and important for glucose consumption and glycolysis in an aggressive subtype of this malignancy, called embryonal hepatoblastoma [52]. Contrasting with these data reporting GLUT3 induction, the ZBTB7A transcription factor binds to the GLUT3 promoter, and to that of other glycolytic genes, PFKP and PKM, and represses their expression, thus acting as a glycolytic inhibitor and a tumor suppressor. In multiple human tumors, ZBTB7A is frequently lost at late disease stages, and the expression of each of GLUT3, PFKP, and PKM is negatively correlated with ZBTB7A, at least in CRC [53].
Although cancer is a genetic disease, chromatin and epigenetic aberrations play important roles in tumor initiation through to progression [54]. So far only a few studies have begun to interrogate epigenetic events associated with GLUT3 expression in cancer. After treatment with the epigenetic targeting agents histone deacetylase inhibitors, GLUT3 expression was induced in NSCLC cell lines, with minimal effect in non‐transformed immortalized human bronchial epithelial cells [55]. In bladder cancer cells, miR‐195‐5p targets the GLUT3 3′‐untranslated region (UTR). Small interfering RNA (siRNA)‐ and miR‐195‐5p‐mediated GLUT3 knockdown experiments revealed that miR‐195‐5p decreases glucose uptake, inhibits cell growth and promotes apoptosis by suppressing GLUT3 [56].
GLUT3 is also controlled at the protein level through membrane relocalization, which occurs during immune cell activation and could be relevant for its role in the tumor microenvironment (see below). For example, in resting immune cells, insulin triggers a translocation of GLUT3 (and GLUT4) to the plasma membrane in B‐lymphocytes and in monocytes, without affecting its localization in neutrophils or T‐lymphocytes; in contrast, upon immune cell activation with a combination of phorbol myristate acetate and lipopolysaccharide, GLUT3 expression increases at the plasma membrane in B‐ and T‐lymphocytes, as well as monocytes and neutrophils [57-59]. Molecular mechanisms involved in GLUT3 re‐localization have been reviewed [60].
Glucose transporters in cancer therapies
The detection of early stage disease, advanced primary or metastatic tumors is essential to determine the beginning of a medication, to follow disease evolution, treatment efficacy or resistance. Several non‐invasive techniques, such as X‐ray imaging for mammograms, computed tomography scan (CT scan) or positron emission tomography (PET) scan, have been developed to provide useful information about tumor volumes, anatomy, location, function, or metabolism. The metabolic imaging PET scan relies on the higher glucose uptake and rate of glycolysis of tumors. It uses radiolabeled glucose analog 2‐Deoxy‐2‐[fluorine‐18] fluoro‐D‐glucose (18F‐FDG) [61, 62]. This radiotracer crosses cell membranes via GLUTs, is then phosphorylated by hexokinases into FDG‐6‐phosphate but cannot be further metabolized. As a consequence, it accumulates into the cytoplasm revealing an increased cell tracer activity [63]. Although PET is a central clinical methodology to monitor tumor metabolism, the sensitivity of this technique varies depending on the cancer types [62, 14, 64]. This heterogeneity has been particularly associated with GLUT1 or GLUT3 tumor expression since 18F‐FDG has a high affinity for glucose transporters, and especially for these two uniporters [14]. A positive correlation between GLUT1 expression and 18F‐FDG uptake has indeed been shown in cervical cancer [65], thymic epithelial tumors [66], primary gastric lymphoma (PGL) [67], as well as metastatic pulmonary tumors [68]. GLUT3 expression and 18F‐FDG accumulation have been linked in primary central nervous system lymphoma (PCNSL) [69]. In contrast, in colorectal adenocarcinoma GLUT1 expression does not correlate with 18F‐FDG [70]. These limitations may also be explained by the alternative carbon sources that can be used other than glucose. For example, in vivo experiments have suggested that lactate can be used as a predominant carbon fuel in lung tumors, or that glutamine considerably contributes to the Krebs cycle in pancreatic cancer [71, 72].
Although many cancer types are initially sensitive to chemotherapy, resistance inevitably occurs through a large variety of mechanisms, including metabolic changes that promote drug inhibition, degradation or cellular export [73]. Among the different possibilities that enable cancer cells to resist therapy, glucose transporters are reported to be involved. As an example, the antiangiogenic VEGF‐neutralizing antibody called bevacizumab, used in the treatment of glioblastoma, has a limited response duration. A recent study comparing bevacizumab‐responsive to resistant patient‐derived tumor xenografts identified higher glucose uptake, glycolysis, and survival in low glucose conditions in drug‐resistant tumors, phenotypes that were recapitulated upon GLUT3 overexpression. Accordingly, GLUT3 was upregulated in bevacizumab‐resistant versus sensitive tumors [74]. Another link between glucose transporter expression and tumor angiogenesis was established for ovarian carcinoma, which revealed, using immunohistochemistry, that the expression of GLUT1, GLUT3, and GLUT4 correlated positively with that of VEGF [75].
In an analogous manner, the role of GLUT1 in response to radiotherapy and clinical outcomes in advanced cervical squamous cell carcinoma was investigated in patients who received radiation therapy. Whereas overexpression of GLUT1 was observed in 53% of the radiation‐sensitive group, it reached almost 90% in the radiation‐resistant one. Moreover, patients with high GLUT1 expression in tumors showed a more pronounced resistance to radiotherapy and shorter progression‐free survival than those with low GLUT1 [76]. GLUT1 was also found to be a marker of radioresistance in oral squamous cell carcinoma by immunohistochemistry from 40 pretreated tumor biopsies [77]. Finally, GLUT1 staining highlighted a link with chemoresistance in the NCI‐60 panel of human cancer cell lines [78]. The underlying mechanisms by which GLUTs are involved in chemo‐ or radio‐resistance remain largely unknown. However, a few studies suggest a connection between glycolysis and the anti‐apoptotic gene MCL1, whereby glycolysis inhibition blocks its translation [79, 80]. Importantly, glycolysis inhibition used in combination with ABT‐737, a pro‐apoptotic BH3‐mimetic compound, increases the sensitivity of lymphoma cells to this drug [79].
In future investigations, it will be important to monitor glucose transporters in tumors treated with immunotherapies, to determine if their expression in tumor cells or cells of the tumor environment changes upon treatment, and if their activity is linked to the outcome of therapy.
Glut gene targeting to interrogate their function in tumors
The consequences of constitutive gene deletion of individual glucose transporters was extensively studied in gametogenesis, embryogenesis, or specific cell types including pancreatic β‐cells or neurons [81-85]. For Glut1 and Glut3, their knockout results in embryonic lethality [82, 86]. To overcome this problem and be able to generate mice with spatiotemporal control of Glut deletion, conditional mouse models have been generated. Although all Glut genes from class 1 have been engineered to enable Cre/LoxP‐mediated recombination (Glut1Flox/Flox, Glut2Flox/Flox, Glut3Flox/Flox and Glut4Flox/Flox mice [87-90]), only Glut1 conditional gene deletion has been studied in the context of tumor development, in breast cancer. First, isolated mammary cells from Glut1Flox/Flox mice were transformed in vitrowith polyomavirus middle T antigen (PyMT), followed by recombination using adenovirus‐Cre (Ad‐Cre) or Ad‐GFP as control. This resulted in a reduced cellular capacity to consume glucose, synthesize lipids and proliferate in vitro. Upon implantation into the mammary fat pads of immunodeficient mice, Glut1‐deficient cells formed tumors that grew slower and had a reduced cell proliferation compared to control tumors [87]. However, Glut1 deletion was achieved in vitro, prior to cell transplantation, precluding longitudinal analyses of the same tumors in vivo before and after recombination. In a subsequent study, Glut1Flox/Floxmice were crossed to a transgenic mouse model of mammary tumorigenesis based on an oncogenic variant of Neu (ErbB2/Her2) and also expressing Cre recombinase, both under control of the mouse mammary tumor virus promoter (MMTV‐Neu‐IRES‐Cre, simplified MMTV‐NIC). This led to an elegant model of breast cancer, where immunocompetent mice express or not Glut1 (NIC‐Glut1+/+, NIC‐Glut1Flox/+, NIC‐Glut1Flox/Flox) in the mammary epithelial compartment and the tumor epithelial cells [91]. At 200 days of life, all NIC‐Glut1+/+mice had tumors, whereas almost all NIC‐Glut1Flox/Flox mice remained tumor‐free even at > 500 days. Strikingly, NIC‐Glut1Flox/+ heterozygous mice behaved like NIC‐Glut1Flox/Flox mice, with most animals having no tumor in the long‐term, indicating that the loss of even a single Glut1 allele is sufficient to impose a strong break to breast tumor development in this model [91]. Epithelial cells from mammary glands proliferated less upon loss of one or both Glut1alleles, but the mammary ductal tree had an overall normal appearance. Nevertheless, in Glut1Flox/+ and Glut1Flox/Flox mice, there was a lack of strong Cre recombinase expression detected in the nucleus of mammary epithelial cells, suggesting that Glut1 deletion (on one or both alleles) leads to the elimination of mammary cells with a tumorigenic potential, potentially explaining the lack of tumor formation. Undoubtedly, additional investigations from this and other mouse models of cancer with Glut conditional deletion will continue to provide important biological information on their role in cancer development.
Structure‐aided glucose transporter targeting
One strategy to perturb tumor cells that depend on the expression of a specific glucose transporter is to use their transport capacity for the entry of defined molecules, which are structurally similar to the physiological substrates, or can even be the normal substrates. An interesting example comes from the capacity of glucose transporters GLUT1 and GLUT3 to transport, in addition to glucose, the oxidized form of ascorbic acid, dehydroascorbic acid (DHA) [92]. High‐dose of DHA impairs the growth of mutant KRAS or BRAF colon cancer cells in xenograft models; increased import of DHA via GLUT1 into cancer cells leads to reactive oxygen species (ROS) accumulation, which in turn inhibits glycolysis and ultimately results in cell death [93].
The use of glucose transporter inhibitors for cancer therapy aims to reduce cancer cell growth by triggering an acute metabolic stress. However, the majority of these inhibitors have been studied in silico or in vitro [94-97], with only few tested in actual tumor models in vivo. WZB117 inhibits GLUT1 and has been demonstrated to prevent tumor initiation by cancer stem cells, or to suppress tumor growth, in immunodeficient mice transplanted with human tumor cell lines [98, 99]. STF‐31 directly binds to GLUT1 and disrupts glucose uptake and utilization in xenografts from renal cell carcinoma [100]. However, the selectivity of such pharmacological inhibitors has been challenged [101-103].
To increase GLUT inhibitor selectivity, one can capitalize from the recent glucose transporter crystal structures. Solving the fine structure of a transporter is indeed key toward a better understanding of conformational changes, solute transport mechanisms, disease‐related mutations, or to identify potential drugs (Fig. 2). Hence, a lot of effort has been placed toward the elucidation of the structure of glucose transporters in recent years, with successes for the structure determination for some of them. XylE, a sugar transporter, is a distant homologous bacterial protein of GLUT1, GLUT2, GLUT3, and GLUT4. Its three different ligand‐bound outward‐facing crystal structures allowed to build homology‐based modeling of class 1 GLUTs [104]. These models, while helping to increase our fundamental knowledge on glucose transporters, need to be complemented by the actual crystal structure of the mammalian transporters.
The atomic structure of human GLUT1 has first been elucidated at a resolution of 3.2 Å on a mutated GLUT1 (N45T, E329Q) in an inward‐open conformation [105]. This structure, which provides the putative glucose moiety‐binding site, might greatly help to develop new inhibitory compounds. The inward‐open cocomplex of human wild‐type GLUT1 at 2.9–3.0 Å with structurally different inhibitors (cytochalasin B and two phenylalanine amides) have been crystallized subsequently, showing the binding modes of molecules with different chemical backbones, which overlap with the binding site for glucose and help explain their inhibitory mechanism of action [106]. Together, these findings and comparisons with the structures of other GLUTs or structure‐based homology modeling will undoubtedly contribute the development of more selective therapeutic compounds. Along these research lines, a high‐throughput small molecule screen complemented by structure‐activity studies and preliminary in vivo pharmacokinetics analyses identified the BAY‐876 molecule as promising selective GLUT1 inhibitor, with an IC50 at least 100 × superior to GLUT2‐4 inhibition [101].
For human GLUT3, an experimental 3D structure of this transporter bound to D‐glucose in the outward‐occluded conformation was recently obtained at high‐resolution (Fig. 2) [107]. This structure not only provides insights regarding the substrate access cycle of the GLUT family, but it also paves the way for computer‐aided structure‐based design of GLUT3 inhibitors, potentially allowing the discovery of GLUT3 ligands through virtual screening. Several docking softwares can be used to perform structure‐based virtual screening, like GOLD [108], Glide [109], DOCK [110] and Autodock Vina [111] to cite a few. Since GLUTs are transmembrane proteins, force‐field based docking software programs like EADock [112] or Attracting Cavities [113], able to adapt their scoring functions to take the influence of the membrane into account, could be of particular use for this family. Finally, the X‐ray crystal structure of GLUT5 has been resolved recently [114]. Although we do not focus on this fructose carrier in this review, its structure determination could be important to develop inhibitors having anticancer properties. Indeed, in acute myeloid leukemia, increased GLUT5‐dependent fructose uptake sustains tumor cell proliferation and malignant behavior [115]. This study nicely illustrates the possibility that the transport of specific nutrients by specific GLUTs be tumor type‐dependent.
Glucose transport in noncancerous cells of the tumor microenvironment
Most studies on GLUTs in cancer are focused on their role and regulation in the tumor cells. However, solid tumors are composed of multiple cell types from various origins, which communicate to each other forming a dynamic and complex network that orchestrates tumor development. Hence, when considering various mechanisms of tumor development, including paracrine communication, or the response to conventional or new therapeutic agents, one should consider all the cell types that compose the tumor and not merely the tumor cells. Glucose uptake in tumors is no exception. Accordingly, the function of T lymphocytes in tumors has gained a profound interest in recent years, since they constitute a primary cellular target for immunotherapies including adoptive T cell therapy and immune checkpoint blockade. In the mouse, T lymphocyte subtypes express varied levels of the glucose transporters Glut1, ‐3, ‐6 and ‐8, while other Gluts are not detected. Furthermore, Glut1 expression increases during T cell activation, and CD4 T cell‐restricted Glut1 gene deletion using Glut1Flox/Flox mice combined to CD4‐Cre transgenic animals resulted in reduced effector T cell activation. This demonstrated the critical role for Glut1 in activation‐induced glycolytic reprograming enabling survival, growth and expansion of effector T cells [116]. Recent in vivo cancer studies have established an interesting concept of competition between intratumor T cells and tumor cells for the limited availability of glucose. In a mouse model of sarcoma, glucose consumption by tumor cells was shown to alter that of tumor‐infiltrating T cells, inhibiting mTOR activation and production of IFN‐γ, an important cytokine for effector T cell function [117]. Using a mouse model of melanoma, increased glucose consumption by tumor cells was shown to perturb glycolysis in T cells, leading to a decreased expression of the cytoplasmic glycolysis intermediary metabolite, phosphoenolpyruvate, perturbing Ca++‐dependent NFAT signaling and thereby T cell effector functions [118]. Interestingly, combined hypoxic and hypoglycemic conditions in melanoma led to a metabolic switch toward fatty acid catabolism in tumor‐infiltrating CD8 T cells, enabling to preserve their effector function [119]. These studies begin to illuminate the metabolic relationship between cells of the immune microenvironment and tumor cells. Undoubtedly, nutrient consumption within tumors involves and impacts multiple cell types of the tumor microenvironment – not only tumor cells and T cells but also different immune cells, cancer‐associated fibroblasts and endothelial cells [120]. Indeed, other immune cells have been shown to promote tumor progression in a way that could be directly linked to their glucose metabolism. An interesting example is that of platelets, which are produced from megakaryocytes and are a major driver of thrombus formation. In LUAD, platelets can be recruited to tumors via an endocrine communication mediated by platelet factor 4 (PF4) [121]. Platelets have recognized protumor functions, at least partly because they release TGF‐β upon activation‐dependent degranulation, which can trigger Smad and NF‐κB responses in tumor cells, promoting EMT and metastasis formation [122]. Platelets are known for decades to rewire their metabolism upon activation, notably increasing glycolysis [123]. GLUT1 and GLUT3 are expressed in platelets, with GLUT3 being translocated from α‐granules to the plasma membrane upon degranulation, which probably explains the increased glucose uptake and glycolysis [124]. A recent study elegantly elucidated the contribution of Glut1 and Glut3 in platelet function in vivo, using mice deficient for both glucose transporters specifically in platelets (Glut1Flox/Flox; Glut3Flox/Flox; Pf4‐Cre). Glut1/3double knockout abrogated glucose entry and lactate production in platelets, led to impaired activation and even to a decreased platelet production from megakaryocytes [89]. Hence, platelet activation in vivo relies on glucose entry mediated by Glut1 and Glut3, which suggests that Glut1/3 loss or blockade will compromise their protumor activities, a hypothesis that remains to be tested in future investigations.
Cancer‐associated fibroblasts (CAFs) are known to promote tumor growth, invasion, chemoresistance and angiogenesis [125]. GLUT1 induction was observed in human fibroblasts placed in contact with prostate cancer cells [126]. Moreover, after stimulation with TGF‐β or PDGF, a switch from oxidative to glycolytic metabolism with higher glucose uptake was observed in CAFs. Isocitrate dehydrogenase (IDH)‐3α, an enzyme of the Krebs cycle, plays a pivotal role in this metabolic reprogramming [127]. Indeed, the loss of IDH3α decreases α‐ketoglutarate (α‐KG) levels, leading to stabilization of HIF‐1α via the inhibition of prolyl hydroxylase domain‐containing protein 2 (PHD2). HIF‐1α, in turn, promotes glycolysis by increasing glucose uptake. However, the consequences of the glycolytic switch in CAFs on the tumor cells remain to be determined [128, 129].
In conclusion, the requirements for glucose entry and usage by cancer‐supporting or cancer‐antagonizing cells add on to the complexity of metabolic rewiring in cancer; the dynamic metabolic interdependencies within tumors promise to constitute exciting new research territories.
Role for GLUTs beyond glucose uptake for glycolysis
Important questions that remain mostly unanswered are whether and to which extent there is more than simply glucose uptake to enhance glycolysis for the function of glucose carriers in cancer. We make a few considerations, which might of course not be restricted to glucose carriers but also be true for other members within the SLC superfamily.
Protein–protein interaction
Glucose transporters could physically bind to other membrane‐associated proteins, which would perturb the activity of the glucose transporter and/or that of the interacting partner.
A physical association was demonstrated between the secondary active Na+/glucose cotransporter SGLT1 and the plasma membrane receptor tyrosine kinase EGFR, whose gain‐of‐function activity, often resulting from gene mutation or amplification, drives cancer development in a subset of different tumor types including glioblastoma, LUAD, and CRC. Although EGFR kinase activity is critical for cancer progression and is the target of small molecule receptor tyrosine kinase inhibitors, a kinase‐independent prosurvival function was suggested, in which EGFR prevents autophagy‐associated cell death in cancer cells through sustained glucose uptake. Mechanistically, the authors showed that EGFR binds and stabilizes SGLT1. EGFR and SGLT1 coexpression also conferred a survival advantage in low glucose concentration in vitro [130]. This interaction might be dynamically regulated, as it was reported to be increased in cancer cells treated with radiation therapy, and to contribute to the recovery from radiation therapy‐mediated cellular ATP depletion. However, in that situation SGLT1 was phosphorylated and stabilized in an EGFR tyrosine kinase activity‐dependent manner [131]. Although the role of the transporter in EGFR signaling was not tested, this interesting interaction between EGFR and SGLT1 suggests the possibility that other glucose transporters also interact with membrane‐associated proteins that have signaling properties (Fig. 3, left). Immunoprecipitation experiments followed by mass spectrometry analyses might help to uncover physically and functionally linked GLUT interacting proteins.
Glucose‐dependent signaling
Glucose‐derived intermediary metabolites could act as signaling molecules.
Glucose metabolism generates multiple metabolites that can serve as substrates for enzymes involved in chromatin remodeling with potential implications in cell signaling. For example, ten‐eleven translocation (TET) enzymes are involved in DNA demethylation [132]and require α‐KG as a cosubstrate [133]. Interestingly, supra‐physiological administration of glucose in mice leads to a rapid increase of α‐KG that is correlated with a higher global hydroxymethylation in various tissues as assessed by GC‐MS analysis. In addition, the transcription of 732 genes was altered after this treatment with 89 of them also exhibiting a higher hydroxymethylation level [134]. Adding to this complexity, the oncometabolite 2‐hydroxyglutarate competes with α‐KG for dioxygenase (including TET) enzyme regulation [135].
This metabolic control of the epigenome was also observed at the histone level. Several studies suggested a link between glucose metabolism and histone acetylation with a central role of acetyl‐CoA. The acetylation of the lysine residues in histones promotes the relaxing of the chromatin, necessary for gene transcription [136]. By the use of glycolysis inhibitors, it was observed that the global level of histone acetylation is altered and correlates with the levels of acetyl‐CoA [137]. This link was further demonstrated in a study on embryonic stem cell differentiation, where the authors found that a large proportion of acetyl‐CoA produced by pluripotent cells comes from glycolysis [138]. In addition, functional assays revealed that acetyl‐CoA and histone acetylation levels are directly linked to cell differentiation. Such studies highlight the importance and complexity of glucose uptake for genome‐wide transcriptional changes and cell fate.
Gene coregulation
Glucose transporter gene regulation might be coordinated with that of other genes whose encoded proteins serve similar functions.
Cellular glucose uptake does not necessarily imply its full oxidation into CO2 for ATP production, or its use for glycolysis. In fact, a recent study with analysis of metabolic fluxes performed in vivo challenged the notion that tissues directly transform glucose into CO2, but rather shows a production and excretion of lactate, which then reenters cells for conversion into pyruvate to enter the Krebs cycle [71, 72]. Other intracellular usage of glucose includes the pentose phosphate shunt and the hexosamine biosynthesis pathway (HBP); these might be particularly important for cancer cells, to support nucleotide biosynthesis and antioxidant capacity, or to enable protein and lipid glycosylation respectively [139, 140]. When considering GLUT3, during preimplantation embryogenesis its upregulation at compaction coincides with the competence of the embryo to use glucose. In the absence of glucose, embryos fail to form blastocysts, whereas even a short exposure to glucose during the cleavage state rescues blastocyst formation and is needed to trigger GLUT3 transcription. Interestingly, glucosamine can substitute for glucose to induce GLUT3. Two enzymes, glutamine‐fructose‐6‐phosphate‐aminotransferase (GFPT)1 and ‐2, are required for the first step of the HBP, producing glucosamine‐6‐phosphate from fructose‐6‐phosphate and glutamine [141]. Their inhibition by azaserine prevents glucose pulse‐mediated GLUT3 induction, but does not impact glucosamine‐induced GLUT3, glucosamine acting downstream of GFPT. This indicates a tight connection between the nutrient‐sensing HBP and GLUT3 regulation and function in early embryo development [142]. We propose that this connection exists in cancer cells, too. Indeed, in human tumor samples from serous‐type ovarian cancer, both GLUT3 and GFPT2 were part of an interleukin‐6‐correlated 40‐gene signature list, suggesting their coregulation in this malignancy [143]. Additionally, interrogating three large and distinct public gene expression datasets of human LUAD, we invariably find a strong positive correlation between GLUT3 and GFPT2 (unpublished data). Hence, it is tempting to speculate that, in tumor cells, coinduction of both genes will foster glucose uptake channeled into the HBP (Fig. 3, right). This metabolic bifurcation could have major implications for tumor progression, as the importance of this pathway is increasingly recognized in cancer [140, 144]. It was for example described that glucose metabolism and the HBP can be coupled to regulate interleukin‐3 receptor α, and potentially other growth factor receptors, which can promote cancer development [145, 146].
Perspectives
GLUTs have been extensively studied for their role in glucose transport in normal adult tissues as well as during embryogenesis. Their functions in cancer progression are the current subject of multiple investigations. These research efforts are and will continue to provide fundamental knowledge on their regulation and deregulation in tumors and highlight the importance to target them in anticancer strategies. For anticancer treatment, we need to take into consideration the potential undesirable impact of blocking GLUTs in cells or tissues that express them and need them for physiological glucose homeostasis. Will the increased GLUT expression in cancerous tissues be sufficient to provide a therapeutic window for treatment? Or will we need to engineer systems for restricted delivery of GLUT inhibitory drugs at the tumor site?
Although many results have come from in vitro experiments with cancer cell lines, it becomes critical to investigate GLUT regulation and function in vivo, ideally studying the development of complex autochthonous tumors in immunocompetent mouse models, to take into consideration the metabolic adaptation of the tumor microenvironment [147]. Currently, the impact of GLUT proteins outside of the cancer cell remains largely to be explored. Genetic approaches enabling Cre recombinase expression in specific cell types, combined with recent Glut conditional knockout mice and mouse models of cancer may illuminate the importance of glucose carriers in distinct cell types of the tumor mass.
Another issue that needs to be addressed is the chemical inhibition of GLUTs. Despite recent studies using chemical approaches in vivo, specific and potent GLUT inhibitors are still missing. Rational design approaches combining crystal structures and homology‐based models generate precious information for chemical screens. The discovery of GLUT inhibitory compounds might prove useful for their functional characterization, but also for the development of drugs that could be used, probably in combination with conventional chemotherapies or new immunotherapies, to combat cancer.
Acknowledgments
We thank Prof. Ping‐Chih Ho (University of Lausanne) for critical reading of the manuscript, and Prof. Vincent Zoete (University of Lausanne and SIB Swiss Institute of Bioinformatics) as well as the Protein Modeling Facility of the University of Lausanne for support in Molecular Modeling. This work was supported by the Swiss Cancer Research Foundation (KFS‐3681‐08‐2015‐R), by the Emma Muschamp Foundation and by FORCE, a Foundation for Children Cancer Research.