The Pro-tumorigenic Effects of Metabolic Alterations in Glioblastoma Including Brain Tumor Initiating Cells

Abstract

De-regulated cellular energetics is an emerging hallmark of cancer with alterations to glycolysis, oxidative phosphorylation, the pentose phosphate pathway, lipid oxidation and synthesis and amino acid metabolism. Understanding and targeting of metabolic reprogramming in cancers may yield new treatment options, but metabolic heterogeneity and plasticity complicate this strategy. One highly heterogeneous cancer for which current treatments ultimately fail is the deadly brain tumor glioblastoma. Therapeutic resistance, within glioblastoma and other solid tumors, is thought to be linked to subsets of tumor initiating cells, also known as cancer stem cells. Recent profiling of glioblastoma and brain tumor initiating cells reveals changes in metabolism, as compiled here, that may be more broadly applicable. We will summarize the profound role for metabolism in tumor progression and therapeutic resistance and discuss current approaches to target glioma metabolism to improve standard of care.

1. Introduction

One of the emerging hallmarks of cancer is the deregulation of cellular energetics [1]. Due to their rapid growth, cancer cells display an increased requirement for nutrients; these metabolic shifts are commonly known as the Warburg effect [1-4]. Warburg recognized that tumor cells shift their metabolism toward the utilization of glycolysis rather than oxidative phosphorylation (OXPHOS) to generate ATP [1, 4]. Glycolysis also provides additional metabolic intermediates that can be used for multiple cellular processes necessary for tumor growth even under restrictive conditions [3, 5, 6, 7]. Thus, many metabolic pathway components are altered in cancer; not only those involved in glycolysis and OXPHOS but also the pentose phosphate pathway (PPP), lipid oxidation and synthesis, and amino acid metabolism (Figure 1).

Figure 1. Overview of Altered Metabolic Pathways in Glioblastoma.

To permit the dysregulated cell growth in cancer, multiple changes to metabolism occur. These include changes to glycolysis, oxidative phosphorylation, the pentose phosphate pathway, lipid oxidation and synthesis, and amino acid metabolism. Major pathways and enzymes or proteins discussed in this review are shown.

Renewed interest in the metabolism of the lethal brain tumor glioblastoma (GBM; grade IV astrocytoma) can be partially attributed to genomic studies that aimed to identify mutations critical for disease initiation and prognosis. Many of the hits generated from these studies were in pathways regulating metabolism or metabolic enzymes [3, 8, 9]. The metabolic profile of GBM is, therefore, being explored as a means to further understand tumor initiation, progression, and therapeutic resistance, as well as develop novel imaging methods. Improved knowledge of the metabolism of brain tumors is vital as the prognosis of GBM patients remains poor. One factor contributing to poor patient outcome is the profound degree of heterogeneity within these tumors, including different cellular subsets and microenvironments that can alter metabolism. GBM can be separated into multiple subtypes based on DNA methylation, genetic mutation, and/or mRNA expression profiles[10, 11]. GBM molecular subtypes are IDH mutant or IDH wildtype proneural, classical, and mesenchymal, with genetic links to alterations in platelet derived growth factor receptor (PDGFR), epidermal growth factor receptor (EGFR), and Neurofibromin (NF1) respectively [10, 11]. Although subtypes are typically considered in the context of intertumoral heterogeneity, single cell RNA-seq experiments demonstrate that multiple subtypes are present within individual GBMs [12]. Differences in GBM molecular subtypes are not yet widely used to distinguish between treatment strategies; however, clinical trials are increasingly incorporating evaluation of tumor genetic or epigenetic (particularly with regard to MGMT promoter methylation) status into study designs. Imaging of tumor metabolites is likely to similarly increase in clinical trials as our imaging capabilities and understanding of tumor metabolism improves.

Additional contributors to intratumoral heterogeneity are brain tumor initiating cells (BTICs), which are a meaningful component of this review. BTICs are hypothesized to be at the top of the tumor hierarchy and, consequently, can produce the bulk of the tumor [13-15]. BTICs express stem cell markers such as CD133 and Sox2, can self-renew, and can recapitulate the original tumor in xenograft models [13-15]. These cells are also highly resistant to standard therapies and contribute to tumor recurrence [14-17]. While there are several proposed BTIC niches, BTICs are often enriched in areas near necrosis with low oxygen and nutrient levels, indicating they likely have altered metabolic profiles [6, 13-19]. In order to study BTICs, a combination of selection markers (i.e. CD133, CD44), functional markers (sphere formation, tumor formation), and/or culture conditions (without serum and containing EGF/FGF with a serum free supplement) are utilized to select for and analyze these cells specifically. There are controversies regarding the optimal markers to use for prospective BTIC isolation as well as methods of tumor dissociation, which may influence marker percentages. Opinions also differ as to whether prospective isolation of BTICs from xenografts is absolutely required or if selection of sphere forming cultures is sufficient. However, the hallmark capacity to enrich for tumor formation and produce a secondary tumor with characteristics of the primary tumor must be determined. Deviations from these standard practices will be discussed in this review when relevant. Unfortunately, current therapies for GBM are largely inefficient at targeting BTICs with most patients experiencing rapid disease recurrence. There is hope that through targeting enzymes linked with metabolism, the growth of GBM and maintenance of BTICs may be decreased, as we will discuss here [7, 20, 21].

2. Glucose Metabolism

A large portion of the studies on metabolic alterations in GBM and BTICs focus on glucose metabolism. In the normal brain, astrocytes are primarily glycolytic, while neurons rely on OXPHOS [3, 7]. Tumor cells often have elevated glycolytic rates compared to normal cells, even in the presence of oxygen: this aerobic glycolysis is termed the Warburg effect after Otto Heinrich Warburg, who first described it in the 1920s. The Warburg effect is the switch in cancer cells to glycolysis as the major energy source due to insufficient OXPHOS, which eventually becomes irreversible [22]. Aerobic glycolysis is less efficient than OXPOHS in producing ATP (approximately 4 versus 36 mol ATP/ mol glucose), leading to the question: why would a tumor cell evolve to use this method of energy production? While many reasons potentially explain this paradox, a prominent hypothesis is that aerobic glycolysis allows tumor cells to produce precursors for other biosynthetic pathways necessary for growth and proliferation [3, 7]. Supporting this notion, studies have demonstrated that the Warburg effect allows cells to shuttle carbon from glucose into biosynthetic pathways for fatty acids, nucleic acids, and, to a lesser extent, proteins [3].

Multiple experiments using [¹³C]glucose nutrient infusions in orthotopic GBM models established that tumors utilize glycolysis, as indicated by an accumulation of labeled lactate in the tumors.[23, 24] Additionally, these studies identified that GBMs also utilized OXPHOS, with labeling of components of the citric acid cycle including acetyl-CoA and oxaloacetate. The authors determined that glucose supplied additional biosynthetic pathways. Accumulation of labeled glutamine within the tumors was attributed to elevated expression glutamine synthetase in GBMs and glutamine was not significantly catabolized when investigated using a [U-¹³C] glutamine infusion. The rate of glycolysis was also demonstrated to be greater than that of the PPP [23]. Mashimo et al. reported that GBM obtains less than 50% of its carbon for acetyl-coA from glucose. Infusion of ¹³C-acetate and ¹³C-glucose showed a large throughput of ¹³C-acetate in the TCA cycle in orthotopic tumor models indicating that acetate, not glucose, largely supplies the TCA cycle. This should not be wholly unexpected if the carbons from glucose are being shuttled to other biosynthetic pathways and the authors postulated that this may allow rapidly proliferating tumor cells to meet bioenergetic demands while largely shuttling glucose to lactate [24].

As the above studies indicate, the use of orthotopic model systems with metabolic profiling of the whole tumor has provided important insights and elegantly demonstrated the value of investigating the role of metabolism in GBM in vivo. One limitation of this approach, however, is that it does not allow for the identification and assessment of the metabolic profile(s) of subpopulations of cells (like BTICs): it is, therefore, not possible to determine the utilization of different metabolic processes, such as glycolysis or OXPHOS, in separate cellular subsets. Advances to comprehensively analyze the metabolism of individual cell populations in vivo would greatly enhance our ability to accurately investigate heterogeneity in tumor metabolism.

To date, the metabolic state of BTICs is not definitively understood. Several studies have demonstrated that BTICs display a more glycolytic phenotype; including higher expression of glucose transporters and a decrease in oxygen consumption [21, 25-27]. For example, Saga et al. profiled BTICs derived from H-Ras transduced mouse neural progenitor cells. For their studies, they utilized two clones which displayed very different metabolic signatures. One clone was highly glycolytic as indicated by extracellular acidification, lactic acid production, 2-NBDG uptake, and metabolomic analysis when compared to a second clone or untransformed neural progenitor cells. Both clones maintained stem cell-like properties as determined by sphere formation and ability to differentiate into multiple lineages, but the more glycolytic clone was more proliferative [27]. Another study indicated an influx of glucose in BTICs was utilized to maintain de novo purine synthesis, confirming that reliance on glucose metabolism could provide intermediates and a carbon source for anabolic processes [28]. However, Vlashi et al. suggested BTICs are less glycolytic than non-BTICs under normoxic conditions as determined by extracellular acidification rates, oxygen consumption, and lactate production [5]. The apparent dichotomy of findings when different research groups have investigated BTIC metabolic states may be explained by the recognition that BTICs have a great capacity for shifting their metabolic profile in order to adapt and survive. For example, oxidative BTICs can become glycolytic and glycolytic BTICs are capable of shifting to oxidative metabolism when treated with pharmacological agents [5, 21, 25, 26]. Differences may also be due to distinct methods for BTIC isolation and propagation that can impact experimental results. Vlashi et al. use proteasome activity to define BTICs with culture in BTIC media (bFGF, EGF, and serum free supplementation with B27 etc.) additionally supplemented with insulin. As it is more common to use marker expression (CD133) and/or media enrichment, these differences may contribute to differing metabolic states observed in distinct cell populations [5, 13, 25, 29-33]. Limitations are also presented by our inability to fully mimic in vivo tumor cell-cell interactions and microenvironments in cell culture conditions under which BTICs are often characterized.

2.1. Signaling pathways regulating glycolytic changes in GBM

To better understand the molecular mechanisms contributing to metabolic alterations in GBM and BTICs, we will next discuss some of the signaling pathways involved including the expression and activity of glycolytic pathway enzymes.

Hypoxia/HIFs

The presence of diverse microenvironments in GBM influences the tumor cells and promotes heterogeneity within the tumor. Regions of hypoxia are hallmarks of GBM, which leads to stabilized expression of hypoxia inducible factors 1 alpha and 2 alpha (HIF1A and HIF2A). HIFs bind to the promoters of many genes, including those regulating metabolism, as a way of reprogramming cells for survival under restricted oxygen. Metabolic genes induced by hypoxia and the stabilization of HIFs include the glucose transporters GLUT1 and GLUT3 which aid in increasing glucose uptake in GBM. Additional details regarding the role of GLUTs and other hypoxia target genes in GBM, as well as hypoxia regulation of glycolytic enzymes is provided in the sections below. However, it is important to note that the first enzyme in glycolysis, hexokinase (HK), is also hypoxia/HIF regulated. Furthermore, hypoxia induces aldolase, plasma membrane lactate transporters, and lactate dehydrogenase, which are critical for lactate shuttling. This is vital for the cellular response to amplified lactate production due to increased levels of glycolysis [3, 21]. Hypoxia also promotes the BTIC phenotype and deregulates differentiation, a more broadly applicable finding also observed in induced pluripotent stem cells [34-46].

A few studies have sought to determine effects of hypoxia on the GBM metabolome. Kucharzewska et al. determined that hypoxia increased glucose, glycolytic intermediates, and lactate in U87 GBM cells [47]. Hypoxia was also shown to support the activation of alternate glucose metabolism pathways; such as the polyol pathway to protect cells from anoxic cell death and the PPP to provide molecules necessary for the synthesis of nucleic acids and fatty acids. Paradoxically, the authors showed that, despite an increase in the PPP and intermediates from the pyrimidine synthesis pathways, U87 cells displayed lower levels of nucleotides indicating a decrease in their synthesis. There was also an increase in genes involved in glycoprotein and glycolipid production and modification including ST3 beta-galactoside alpha-2,3 sialyltransferase (ST3GAL6) [47]. While these data are compelling, it is essential to recognize that U87 cells are an established cell line grown with serum that behave differently than primary tumor or PDX models. U87 and other standard GBM cell lines are useful tools for investigating the molecular processes important in GBM, but there are limitations and validation of results with multiple models is crucial.

The above experiments indicated a profound role for hypoxia in metabolic reprogramming of GBM as a whole. When a BTIC model was used, hypoxia induced a switch between the PPP and glycolysis. Culture in hypoxia for up to one week resulted in increased expression of glycolysis related genes such as HK2 and phosphofructokinase phosphatase (as discussed below), and a decrease in PPP related genes like glucose-6-phosphate dehydrogenase (G6PD) and phosphogluconate dehydrogenase. In converse experiments, transferring BTICs cultured under hypoxia to normoxia resulted in upregulated expression of PPP enzymes such as transketolase-like protein 1 and G6PD. Protein expression in GBM patient specimens also suggested that glucose flux through the PPP is important in highly proliferative areas of tumors [48, 49]. Building on this concept of PPP repression in hypoxia, other studies indicated that PPP enzymes are initially reduced, only to be upregulated following long-term hypoxia [48, 50]. Overall, these data demonstrate that activation of glycolysis by hypoxia is critical for metabolic adaptation during GBM growth.

PI3K/AKT/mTOR

PI3K/AKT/mTOR pathway activation is common in cancer and is well known to promote cell growth and survival; however, the pathway also directly regulates tumor metabolism. In GBM, AKT is often activated through the loss of its negative regulator, PTEN, or constitutive activation of upstream signaling receptors such as epidermal growth factor receptor (EGFR) via mutation or amplification. AKT promotes aerobic glycolysis by increasing the expression of glucose transporters and regulating their translocation to the cell surface. AKT also increases the recruitment of HK2 to the mitochondria, which couples glycolysis and OXPHOS and prevents apoptosis. PI3K and AKT regulate the movement of carbon from glucose into other biosynthetic pathways such as fatty acid synthesis. Protein and lipid synthesis can also be regulated by mTOR, a downstream modulator of PI3K/AKT, by sensing nutrient levels within the cells, though the precise mechanism remains unknown. mTOR has also been shown to regulate glycolytic metabolism independently of AKT through the activation of Myc [3, 51-54].

While multiple growth factor receptors lead to downstream activation of PI3K/AKT signaling, the metabolic effects of PDGFR signaling has been specifically investigated. In BTICs, PDGFR activated AKT to regulate aerobic glycolysis independent of cellular proliferation [20]. In a glioma mouse model, PDGFR signaling increased growth, lactate production, and glucose consumption of BTICs relative to neural stem cells. Inhibition of PDGFR with the small molecule AG 1295 was able to decrease glucose consumption and lactate production in cell lines where PDGFR induced AKT activation. The effects were associated with decreased levels of GLUT1, hexokinases, lactate dehydrogenase A, and pyruvate kinase 1 [20]. These data highlight the importance of PI3K/AKT/mTOR signaling in cancer metabolism.

2.2. Alterations in glycolytic pathway components

Glucose transporters

The solute carrier family 2 or facilitative glucose transporter (GLUT) family is of increasing interest for cancer biology. Glucose uptake into the cell is the first rate-limiting step for glycolysis. GLUTs have well established roles in glucose transport, though they also transport other substrates including fructose, mannose, galactose and/or glucosamine. Several GLUTs, including GLUTs 1, 2, 3, 6, 8, 10, and 13 are known to be expressed in the brain, with GLUT1 and GLUT3 being the most notable and widely expressed.

GLUT1 is critical for the transport of glucose across the blood-brain barrier [55-57]. GLUT1 is, in part, regulated via hypoxia/HIF1-α, allowing cells within hypoxic regions to survive through uptake of an increased level of glucose [6]. High GLUT1 mRNA expression did not correlate with decreased patient survival in GBM, but has been associated with poor prognosis in other solid tumor types [25, 58]. BTICs displayed a limited increase in GLUT1 expression at the mRNA and protein level compared to non-BTICs in vitro, indicating that this may not be the most critical driver of BTIC glycolytic phenotypes [25]. Targeting GLUT1 suggested a role in tumor initiation rather than progression [51].

GLUT3 is the neuronal glucose transporter and has an increased affinity for glucose compared to GLUT1 Regulators of GLUT3 expression include HIF and NF-κB, both of which are expressed in GBM and BTICs [6, 59]. High GLUT3 expression was associated with worse patient prognosis in GBM and many other solid cancers [6, 25], and correlated with Oct4 expression, an indicator of a more stem cell-like state [25, 60]. Our prior study demonstrated a significant increase in GLUT3 expression in BTICs over non-BTICs in multiple patient-derived xenolines [25]. Knockdown of GLUT3 in a BTIC model decreased GBM growth in vitro and in vivo [25]. One group has also reported on a subset of classical and proneural GBM tumors that are addicted to GLUT3 through aberrant signaling involving integrin ανβ3 [61]. As discussed in additional detail below, GLUT3 was also recently implicated in promoting resistance to bevacizumab, an antiangiogenic therapeutic [62]. These data indicate that GLUT3 may be of potential therapeutic interest for the future.

Glycolytic Enzymes

Directly after glucose enters the cell, it undergoes glycolysis, a pathway that utilizes at least 10 enzymes to metabolize glucose to pyruvate [50]. Many groups have seen an increase in glycolysis and lactate production due to hypoxia, either innate in the tumor microenvironment or a result of antiangiogenic therapy [47, 50, 63, 64]. Tumors also display increased glycolytic activity, regardless of oxygen availability, as seen in the Warburg effect.

The effect of hypoxia on the expression of glycolytic enzymes in patient derived BTICs and established GBM cell lines grown in the absence and presence of serum, respectively, was recently investigated [50]. Following exposure to hypoxia for up to seven days, glycolytic pathway enzymes were induced in all cells analyzed. Using an in vivo screen, cells expressing 55 shRNAs were pooled with the assumption that shRNAs against genes conferring a survival advantage would be depleted in the resulting tumor. Depletion of seven of the eleven shRNAs included against glycolytic enzymes indicated a requirement for tumor growth. Critical glycolytic enzymes identified included but were not limited to hexokinase 2 (HK2), aldolase A (ALDOA), and pyruvate dehydrogenase kinase 1(PDK1). PDK1 was also identified as an important survival factor for BTICs in a separate siRNA screen [65]. Knockdown of each of these genes inhibited GBM cell growth in both BTICs and established adherent GBM lines: ALDOA and PDK1 knockdown of ALDOA and PDK1 increased cell death under hypoxia, but had a minimal effect on cell death when cells were grown under aerobic conditions. Therefore, glycolytic enzymes are important for GBM cell survival under hypoxia and for promoting tumor growth in vivo, but effects are likely to be cell and enzyme dependent.

The glycolytic enzyme HK2 phosphorylates glucose to generate glucose-6-phosphate, a key metabolite for both glycolysis and the PPP. HK2 can be recruited to the mitochondrial membrane to inhibit apoptosis by preventing BAX, a proapoptotic protein, from binding to voltage-dependent anion channels and increasing cytochrome c release [3, 51]. In established GBM cell lines cultured in the presence of serum, knockdown of HK2 inhibited aerobic glycolysis and promoted OXPHOS in GBM [66]. HK2 expression also increased with glioma grade and correlated with poor survival for GBM patients [66]. These studies indicate the likely importance of HK2 in the bulk of the GBM tumor. While additional experiments are still needed, we can infer a potential role for HK2 in BTICs based on other available data. HK2 is negatively regulated by miR-143, which is often downregulated in BTICs and functions to decrease aerobic glycolysis [67]. miR-143 also promoted the differentiation of BTICs in vitro and reduced BTIC initiated tumor growth in subcutaneous models in vivo [67]. Together, these data suggest that loss of miR-143 increases HK2 to increase glycolysis and support the maintenance of BTIC phenotypes.

An RNAi screen also identified 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 (PFKFB4) as an essential mediator of BTIC maintenance [65]. As suggested by its name, PFKFB4 regulates the formation and hydrolysis of fructose-2,6-biphosphatase. This metabolite, in turn, regulates phosphofructokinase activity to control glycolysis. Targeting of PFKFB4 induced apoptosis as determined via measurement of caspase activity. Analysis of publicly available databases showed high expression of PFKFB4 correlated with poor prognosis in patients with GBM, once again suggesting the importance of glycolysis related metabolism for patient outcomes [65].

Lactate Dehydrogenase

Lactate dehydrogenase (LDH) catalyzes the interconversion between pyruvate and lactate, LDH-A preferentially in the direction of lactate and LDH-B in the direction of pyruvate. High levels of LDH have been reported within GBMs, and elevated LDH correlates with invasion and poor survival in patients treated with radiation [68]. In comparison to neuroblastoma cells, U87 GBM cells had increased LDH-A and LDH-B along with an increase in extracellular acidification, indicating a preference for aerobic glycolysis [68]. This is a metabolic phenotype more typical of astrocytes, especially when compared to neurons, providing further insights into the metabolic characteristics of GBM [68]. Silencing of LDH-A by shRNA reversed the Warburg effect as indicated by decreased tumor progression, increased OXPHOS and reactivation of mitochondria-mediated apoptosis [51, 69]. Daniele et al. also reported that inhibition of LDH-A by the small molecule inhibitors NHI-1 and NHI-2 resulted in BTIC differentiation and death. However, this effect was not BTIC specific as the inhibitors caused decreased proliferation and increased death in non-BTICs as well. The data suggest that LDH-A activity has importance for GBM growth across cellular differentiation states [70].

Aldehyde Dehydrogenases

Aldehyde dehydrogenases (ALDHs) catalyze the oxidation of aldehyde to carboxylic acid. Elevated ALDH activity has been used to enrich for TIC phenotypes, including from brain tumors [71]. According to Mao et al., ALDH activity was increased in and important for regulating glycolytic metabolism in BTICs of the mesenchymal subtype. Among ALDH family members, data suggested an important role for ALDH1A3. Microarray analysis demonstrated that ALDH1A3 mRNA was up-regulated in mesenchymal BTICs when compared to mesenchymal non-BTICs and proneural BTICs. Inhibition of ALDH1A3 by a small molecule inhibitor (diethylaminobenzaldehyde) or shRNA-mediated knockdown decreased the growth of mesenchymal, but not proneural, BTICs [72]. These data indicate that activity of enzymes regulating metabolism can differ between GBM cells of distinct subtypes and again highlight the importance of using well-characterized model systems to be able to appropriately interpret results.

2.3. Glucose metabolism and epigenetics

Metabolic enzymes and specific metabolites are beginning to emerge as important factors in directly regulating tumor epigenetics, specifically isocitrate dehydrogenase (IDH) mutations and pyruvate kinase muscle isozyme 2 (PKM2) [73, 74].

IDH mutations

Gliomas with isocitrate dehydrogenase 1 and 2 (IDH1/2) mutations are a distinct tumor subtype that deserve a more comprehensive review than could be presented here and have been previously well-covered by Clark et al. and Zhang et al. [75, 76]. Briefly, IDH1 mutations frequently occur in low-grade gliomas and are therefore very common in secondary rather than primary GBM. IDH mutation is associated with increased survival and a CpG island hypermethylator phenotype (G-CIMP) [73, 74]. The G-CIMP phenotype is promoted by the oncometabolite 2-hydroxyglutarate (2-HG), a new product of mutant IDH1 and 2. 2-HG has been shown to inhibit some demethylases leading to hypermethylation of histones [3, 53]. In cells with an IDH mutation, glutamate and lactate levels were lower when compared to wild-type, but no change in growth was observed [77]. LDH-A and other essential glycolytic enzymes are also decreased in IDH mutant glioma indicating a lower glycolytic capacity that may contribute to differences in prognosis [3, 51, 53, 78]. Currently, studies of IDH mutant tumors are hampered by the difficulty of in vitro passage with IDH mutation maintenance along with the limited availability of patient derived xenograft or animal models but this is an active area of investigation.

PKM2

The final step in glycolysis is the production of pyruvate and ATP from phosphoenolpyruvate (PEP) and ADP by pyruvate kinase. There are two pyruvate kinase isoforms that differ in both function and expression. Pyruvate kinase muscle isozyme 1 (PKM1) and pyruvate kinase muscle isozyme 2 (PKM2) differ in only 23 amino acids due to alternative splicing. PKM1 is the predominant isoform in the adult brain, and PKM2 is associated with the developing brain and stem-like cells [3, 53, 59, 79]. PKM2 is expressed in GBMs and other cancers, and overexpression of PKM2 in LN229 GBM cells exposed to ¹³C-glucose resulted in increased labeled pyruvate [80]. PKM2 is endogenously present in tumors as a dimer that has less affinity for PEP, decreasing the rate of glycolysis and allowing metabolites upstream of pyruvate to enter other biosynthetic pathways critical for tumor growth [3, 53]. PKM2 can be phosphorylated by EGFR or interact with Jumonji or prolyl hydroxylase family members [79, 81-83]. Phosphorylation of PKM2 was shown to mediate its translocation to the nucleus where it phosphorylated histone 3 or promoted a hypoxia-like state, respectively by interacting with β-catenin or HIF1α [3, 51, 53, 79-83]. EGFR activation also upregulates PKM2 expression in a NF-κB dependent manner [59, 79]. Translocation of PKM2 to the nucleus can be blocked via ERK inhibition. The docking of ERK phosphorylates PKM2, which provides a binding site for prolyl hydroxylase family members in order to mediate a pseudo-hypoxic response [79, 81, 82]. There is, however, some controversy regarding the kinase activity of PKM2. When Hosios et al. used [32P]-PEP to label proteins, there was no evidence that PKM2 directly phosphorylates any substrates in cancer cells including U87 parental or EGFR-VIII cells [84]. These data underscore the importance of understanding differences in multiple model systems and further investigation of the role of PKM2 in GBM metabolic and nonmetabolic effects.

2.4. Mitochondria in glioma

Mitochondria are the major site for energy production of the cell via cellular respiration. A multitude of mitochondrial dysfunctions have been reported in gliomas, including structural abnormalities, altered energy metabolism, impaired mitochondria-dependent apoptosis, and mitochondrial DNA mutations [85]. However, several groups have identified glioma cell lines that rely on mitochondrial OXPHOS, highlighting again the large degree of heterogeneity in GBM metabolism [86-88]. In a landmark study, Prabhu et al. determined that reduced mitochondrial reserve capacity caused by Ras-mediated regulation of PDH activity via pyruvate dehydrogenase phosphatase catalytic subunit 1 (PDP1) repression was a common metabolic abnormality in a variety of GBM lines [89]. This data suggests that activating PDH in GBM could provide a therapeutic benefit.

Mutations in mitochondrial DNA, encoding subunits of complex I of the electron transport chain (ETC) were identified in GBM and may contribute to the switch to glycolysis by reducing the function of the ETC [90]. Lloyd et al. found over 200 mutations in mtDNA of GBM cells and tissues, including 9 non-synonymous mutations in complex III and IV predicted to affect OXPHOS [91]. Analysis of patient samples showed that 43% of tumors contained at least one mutation [91]. Another study confirmed these findings, though the roles of the identified mtDNA mutations still require further investigation to understand their impact of these mtDNA mutations in GBM [92].

Mitochondria and BTIC maintenance

Mitochondrial dysfunction plays an important role in BTIC maintenance. We reported that the hypoxia-induced upregulation of the BTIC marker CD133 could be mimicked via inhibitor-induced mitochondrial dysfunction. Furthermore, mitochondrial DNA-depleted GBM cells, which possess an uncoupled glycolytic phenotype, showed a significant and stable upregulation of stem and progenitor cell marker expression and increased self-renewal [30]. BTICs treated with the glycolysis inhibitor and glucose analog 2-deoxyglucose (2-DG) were pushed toward mitochondria-dependent metabolism. This shift resulted in increased differentiation via multiple mechanisms including destabilization of HIF1α and decreased ROS signaling [93]. While BTICs could readily switch between glycolysis and OXPHOS as a means of adapting to environmental changes (see above discussions in the Glucose Metabolism section), a study on radioresistant BTICs suggested that OXPHOS was their main source of energy production [5]. This result is supported by studies demonstrating that inhibition of OXPHOS in GBM BTICs via downregulation of insulin-like growth factor 2 mRNA-binding protein 2 (IMP2) decreased BTIC maintenance. IMP2 facilitated mRNA delivery to mitochondria as well as regulated the assembly of respiratory complexes I and IV [94]. Janiszewska et al. demonstrated a strong correlation between IMP2 expression and GBM cell self-renewal in vitro as well as worse survival in animal models [94]. These contrasting studies highlight the important role of OXPHOS in BTICs as well as their high degree of metabolic plasticity.

Additionally, Cytochrome c oxidase (CcO), the terminal enzyme of the mitochondrial ETC, has been shown to mediate an important role in GBM proliferation and high expression correlated with worse patient survival [95]. Subunit 4 (COX4), a nuclear gene-encoded subunit of CcO, and specifically the COX4-1 isoform, enables CcO to couple the rate of ATP production to energetic requirements (OXPHOS). We reported that COX4-1 expression promoted GBM proliferation through BMI1 expression and attenuating ROS production [96]. COX4-1 expression increased the self-renewal of BTICs and stem marker expression on U251 cells cultured as neurospheres; however, COX4-2 inhibited neurosphere formation [96].

The generation of ROS from active mitochondrial respiration is important for BTIC maintenance, but the extent of ROS changes can dictate whether responses are pro or anti-tumorigenic. ROS production can be elevated in BTICs, as well as other cancer types, as a result of transformation, which may cause intracellular damage. To keep the basal ROS levels in check, cells upregulate antioxidant enzymes like glutathione peroxidases [97-99]. Antioxidant mediated reduction of oxidative stress from ROS is critical for tumor cell survival and chemoresistance. Targeting antioxidant enzymes increased ROS-induced GBM cell death [99]. High levels of ROS production were also associated with increased toxicity of proton beam irradiation compared to x-ray photon radiation in BTICs [100]. Yuan et al. suggested that in the presence of serum, BTIC mitochondrial respiration produced ROS that decreased BTIC marker expression via NFκB signaling. In their model, the serum treated BTICs showed increased tumorigenic potential. The authors attributed this effect, which is distinct from the typical less tumorigenic phenotype of serum cultured non-BTICs, to the progression toward progenitor cells as well as the pro-tumorigenic role of the generated ROS in redox signaling during tumorigenesis [101]. Together, data indicate that the strength, duration, and differentiation state of ROS exposed cells can impact their survival, therapeutic resistance, and tumorigenic potential.

Although still an area of active investigation, changes in mitochondrial fission/fusion could also regulate BTIC phenotypes. In comparison to non-BTICs, BTICs were suggested to have an increased level of mitochondrial fragmentation, driven by fission. Xie et al. showed that mitochondrial fission was increased in BTICs through the activation of dynamin related protein 1 (DRP1) by cyclin-dependent kinase 5 (CDK5). DRP1 suppressed the activity of 5’ AMP-activated protein kinase (AMPK), which functions as a mediator for energy stress response [102]. However, effects of DRP1 have depended on cell density and alternative models suggested that loss of DRP1 increased stem cell markers, including SOX2 and ALDH1, and promoted mitochondrial respiration [103]. Overall, these findings indicate a key role for the mitochondria in promoting tumorigenesis and BTIC maintenance that need further investigation.

3. Lipid Metabolism and Biosynthesis

While a majority of research has focused on glucose metabolism alterations, progressively more research has indicated that lipid metabolism is also changed. Expression of enzymes required for fatty acid oxidation is altered in GBMs and inhibition of the pathway is anti-tumorigenic. Carnitine palmitoyltransferase 1A (CPT1A) and carnitine palmitoyltransferase 1C (CPT1C) are two regulators of long-chain fatty acid transport into the mitochondria and fatty acid beta-oxidation. CPT1A and CPT1C were altered in a group of GBM patients, with elevated CPT1A suggesting changes in lipid metabolism in high-grade glioma. Further correlation analyses revealed a connection between CPT1A and PTEN, a well-established tumor suppressor often lost in GBM [104]. Treatment of BTICs with etomoxir, an irreversible CPT1 inhibitor, decreased the percentage of proliferating cells as well as total cell number. Pharmacological inhibition of fatty acid oxidation with etomoxir in a syngeneic mouse glioma model also slowed tumor growth and increased survival. These results reinforce the observation that the oxidation of other bioenergetic substrates such as fatty acids is critical for GBM growth [105].

Many other fatty acid related molecules have also been implicated in GBM and BTIC growth. For example, fatty acid binding protein-7 (FABP7) is an established marker of neural stem cells that may also serve as a marker for BTICs. FABP-7 is a chaperone for various ω-3 fatty acids that was upregulated in BTICs at both mRNA and protein levels. Interestingly, FABP7⁺SOX2⁺ glioma cells were increased in GBMs in comparison to lower grade astrocytomas [106]. Expression of very long-chain acyl-CoA synthetase homolog 3 (ACSVL3), which adds coenzyme A to fatty acids to permit their use in additional metabolic processes, was increased in BTICs and played a key role in maintaining self-renewal [107, 108]. Loss of ACSVL3 decreased expression of BTIC markers such as CD133 and Sox2 and decreased tumor growth in vivo. Further analyses showed that the expression of ACSVL3 was regulated by signaling pathways such as EGFR, HCF/c-Met and AKT [107, 108]. These data provide a strong link between pro-tumorigenic signaling pathways, lipid metabolism, and BTIC maintenance.

Inhibition of the de novo fatty acid synthesis enzyme fatty acid synthase (FASN) is known to cause tumor growth inhibition. In gliomas, FASN expression correlated with WHO tumor grade and inhibition of FASN reduced GBM cell growth. For example, treatment with the FASN inhibitors cerulenin or orlistat, which is used in obesity treatment, significantly decreased viable GBM cells [109, 110]. FASN was also strongly upregulated in BTICs, leading to the observed higher lipogenesis in those cells compared to their more differentiated counterparts. Pharmacological inhibition of FASN with cerulenin significantly decreased proliferation and migration of BTICs, as well as downregulated the expression of BTIC markers Sox2 and FABP7 [111]. Thus, synthesis of fatty acids is important for BTIC maintenance.

GBM cells also appear more dependent on cholesterol for cell growth, making cholesterol related inhibitors a promising potential therapy. Many cells in the brain depend on cholesterol synthesized by astrocytes for their survival. GBM cells upregulated low-density lipoprotein receptors (LDLR) to increase their cholesterol uptake, while at the same time inhibited their own cholesterol synthesis [112]. This enabled GBM cells to escape from feedback mechanisms in which 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), the limiting enzyme of cholesterol synthesis, is inhibited and liver X receptors (LXRs) are activated by oxysterols to induce cholesterol efflux [112]. Recent evidence also suggests that HMGCR and other components of the mevalonate pathway required for cholesterol synthesis are elevated in BTICs [113]. Targeting of HMGCR with shRNA or the mevalonate pathway with simvastatin decreased GBM growth in vivo [113]. Together the data suggest the flux of lipids in BTICs is important for GBM maintenance and growth.

While hypoxia has been shown to reduce cholesterol synthesis through the impairment of HMGCR, total cholesterol was not decreased in GBM cells in hypoxia [24, 47]. However, hypoxia can alter lipid metabolism in GBM. Hypoxia increased the levels of palmitic acid and glycerol-3-phosphate, the foundation for glycerolipid synthesis. Choline, which is required for phosphatidylcholine and sphingomyelin, was also increased [47]. This is important as altered choline metabolism has correlated with glioma progression. Under hypoxic conditions, the level of pro-apoptotic ceramides was decreased, indicating that GBMs may decrease ceramide availability and increase conversion to anti-apoptotic glycosphingolipids [47]. The data presented in these studies indicate a substantial role of lipid metabolism in BTIC maintenance and GBM progression.

4. Amino Acid Metabolism

One final metabolic pathway important for glioma and BTIC maintenance to be reviewed involves amino acids. Some data suggested that GBM cells increase amino acid transport through amino acid transporters such as LAT1, which has been correlated with poor patient prognosis.[150] Under hypoxic conditions, data suggested that GBM increases its pool of free amino acids which supports the production of glutathione. Kucharzewska et al. also found evidence of increased protein catabolism and autophagy which is consistent with other findings of increased autophagy and protein degradation by chronic hypoxia. Hypoxic tumor cells rely on high levels of glutathione to maintain redox homeostasis in response to increased levels of redox stress, indicating an increase in glutathione synthesis in hypoxic cells [47, 114]. The data also lead to the conclusion that under hypoxic conditions, this ability to regulate glutathione may lead GBM cells to be more resistant to chemotherapy and radiation [47].

Most studies investigating amino acid metabolism in GBM focused specifically on glutamine. Glutamine is an important substrate for DNA and fatty acid synthesis in cells and stimulates glutathione synthesis to reduce oxidative stress and promote radiation and chemotherapy resistance. Additionally, glutamine aids in replenishing depleted liver glycogen stores, allowing for increased glucose to be present in the bloodstream to provide for the increased glycolytic rate of cancer cells [115]. In GBM, glutamine may be an important source of molecules for the TCA cycle, which can then be shuttled into other biosynthetic pathways to sustain tumor growth. Glutamine withdraw in GBM has been suggested to lead to a reversal of the Warburg effect and ultimately cell death, as there is not enough NADH₂/FADH₂ for sustained OXPHOS. This suggests that glutamine withdraw along with carbohydrate or glycolysis inhibition may be a promising therapeutic option [115]. However, other evidence suggested this strategy will not work well for all GBMs. A large portion of glutamate generated from glutamine was found to be secreted rather than enter the TCA cycle, and glutamine starvation did not result in GBM cell growth inhibition due to glutamine synthetase activity [116]. More recent evidence also suggested that the use of glutamine varies among GBMs and only a subset of tumors, perhaps of the mesenchymal subtype and lacking expression of the BTIC marker CD133, would be sensitive to targeting glutamine metabolism [117]. As several aspects of cancer metabolism are associated with glutamine and multiple combinatorial strategies in distinct subsets of tumors could be applied, additional research in this area is warranted [115].

Asparagine is another amino acid that is reported to confer resistance to apoptosis in cancer cells. Karpel-Massler et al. investigated the use of L-asparaginase, which is commonly used to treat acute lymphoblastic leukemia, in GBM. Using established BTIC and patient-derived xenografts treated with E. coli derived L-asparaginase, a subset of GBM models displayed decreased growth indicating cell type specificity. Dual therapy with L-asparaginase and ABT263, a compound that induces apoptosis, increased survival in vivo and even appeared to cause regression in addition to slowed tumor growth. Unfortunately, cells can acquire resistance to L-asparaginase through the upregulation of asparagine synthetase (ASNS). Identification of ASNS inhibitors and additional combinatorial approaches will be important to increase the efficacy of this treatment strategy [118].

The metabolism of lysine has also been implicated as a potential biomarker for cancer. Lysine catabolism proceeds through one of two pathways: saccharopine or pipecolate pathways [119]. The saccharopine pathway is predominately utilized in the developing brain, while the pipecolate pathway is typically utilized in the adult brain [119]. NMR demonstrated that the lysine catabolism intermediate, α-aminoadipate, was increased in BTICs indicating that lysine catabolism continued through the saccharopine pathway. High levels of α-aminoadipate correlated with poor patient survival when measured in the patients from whose tumors the BTIC cells were derived [119]. Ultimately, this indicates the importance of amino acid metabolism in the maintenance of the BTIC population and a preference for stem-like catabolism pathways in BTICs.

5. Metabolic Alterations and Invasion

Complete surgical resection of GBM is impossible due to tumor cell invasion of nonneoplastic brain which leads to rapid tumor recurrence [120]. BTICS have been shown to be highly invasive and metabolic alterations are beginning to be investigated for their ability to regulate the invasive potential of BTICs [121]. Bevacizumab is known to promote GBM cell invasion and bevacizumab resistant GBM cells are more glycolytic as further discussed below [62, 63]. Lower levels HIF1α and higher levels of mitochondrial complex I have been observed in invasive models versus angiogenic models, indicating that GBM cells adjust their metabolism based on the microenvironment [122]. In addition, following long-term hypoxia, there was an upregulation of glycolysis in invasive GBM cells that was not seen under short-term hypoxia. Unfortunately, these cells were not investigated further [122], but evaluating the role of increased glycolysis on invasion will likely prove interesting.

A few studies have evaluated the role of metabolic regulators on BTIC invasion. An increase in glucose-6-phosphatase (G6PC), an enzyme that hydrolyzes glucose-6-phosphate to yield glucose and orthophosphate, was identified in BTICs when compared to the normal brain. Elevated G6PC facilitated GBM cell resistance to glycolytic inhibition by 2-DG in vitro, and G6PC knockdown decreased migration, invasion, and proliferation in vitro and in vivo [123]. In support of this finding, glycolytic enzymes were upregulated in invasive GBM models in both bulk tumor cells and BTICs. Knockdown of one enzyme involved in glycolysis, aldolase C (ALDOC), decreased the invasion and migration of BTICs and established cells in vitro. However, knockdown of ALDOC alone was not sufficient to significantly alter tumor cell invasion in vivo, although knockdown cells were less proliferative. The discrepancy between these data is likely indicative of how tumor microenvironments and cell-cell interactions cannot be fully modeled in cell culture but have a large influence over pro-tumorigenic biologies in vivo. The ALDOC study concluded that expression of glycolytic enzymes was increased in pseudo-palisading cells, which may be tumor cells that are actively migrating away from regions of hypoxia [49]. Therefore, we can begin to identify a common pattern of increased glycolysis in invasive GBM cells and BTICs.

6. Therapeutic Resistance in Glioma

In addition to surgical resection, the current standard of care for GBM includes radiotherapy and chemotherapy with temozolomide (TMZ), a DNA alkylating agent. As mentioned above, the infiltrative nature of GBM hinders complete surgical resection and therapeutic resistance is common [16, 17, 124]. As such, GBM tumors recur quickly and remain incurable. Recently, the metabolome has been of interest for understanding changes that occur in the cell during and after treatment. Multi-drug resistant cancer cells have been shown to be more glycolytic than their sensitive counterparts, and we, therefore, need to better understand the metabolism of these cells to develop more effective treatment options.

6.1. Irradiation Induced Metabolic Shifts

To directly determine if metabolic shifts occur with radiotherapy, Wibom et al. used metabolomics to compare the profiles of extracellular fluid from GBM tumors and adjacent brain from patients before and during radiotherapy. A significant difference was identified between the metabolites of the tumor and adjacent brain, both before and after radiation. Changes post-treatment included decreased levels of galactose and tyrosine with increased glutamine, stearic acid, and alanine. While the patterns of metabolic changes observed can provide insight into new potential targets for therapy, the authors note that drawing conclusions regarding specific metabolites is extremely difficult due to multiple variables potentially influencing their amounts [125]. Another group has shown that following radiation cells rapidly switch to OXPHOS through inhibition of HK2 by mTOR [126]. Although this study uses several different cancer cell types, they have only performed these experiments in one GBM cell line, U87. Thus, more research needs to be done to better understand this process in other GBM cells and across subtypes [126]. The ability of cancer cells to quickly shift their metabolism accentuates the need to further understand the role metabolism plays in therapeutic resistance.

6.2. Changes in Glutamate, Glutamine and Mitochondria Post Treatment

Multiple groups have profiled aspects of the metabolism of TMZ resistant GBM cells and GBM cells following TMZ treatment. These studies indicated that mitochondria are important in modulating the response to TMZ, as well as altered glutamate metabolism [127-130]. We demonstrated changes in the activity of the enzymes of the electron transport chain in TMZ resistant cells as well as compositional changes to the mitochondria. Specifically, U251 GBM cells selected to be TMZ resistant in vitro displayed decreased levels of complex I and V with increased levels of complexes II/III and CcO. Similar changes were also observed in PDX and patient specimens. Targeting of CcO, both genetically and pharmacologically, in TMZ resistant cells sensitized GBM cells to chemotherapy [127, 131-133]. These data strongly suggest that mitochondrial metabolism plays a large role in the acquisition of TMZ resistance and targeting OXPHOS components may be a potential therapeutic avenue for chemoresistant GBM [132, 133]. Mitochondria have also been implicated in PI3K inhibition resistant GBM. Through a feedback loop, PI3K antagonists lead to the re-activation of AKT2 and mitochondrial support of cell growth and survival [134]. Unfortunately, PI3K inhibitors have shown only limited efficacy in the clinic, due to their inability to specifically target cancer cells [134]. New studies into dual PI3K/mTOR inhibitors have shown more therapeutic promise, but glutamine metabolism hindered the efficacy of mTOR inhibitor treatment, thus leading to the possibility of a reduction in efficacy of PI3K/mTOR inhibitors [135].

6.3. Changes in Metabolism Following Anti-VEGF Therapy

While great hope was once held for the anti-angiogenic bevacizumab, this VEGF antibody did not extend patient overall survival and only minorly improved progression free survival [64]. However, in the US bevacizumab continues to be used in GBM patients, particularly to control edema. This clinical use, along with the hope that bevacizumab-induced invasion could be inhibited makes understanding the metabolism of bevacizumab-treated tumors vital. The effect of bevacizumab administration on the metabolism of orthotopic models of GBM was determined through metabolic flux analysis using ¹³C-glucose infusion [64]. Bevacizumab-treated tumors displayed an increase in glucose influx along with an increase in lactate production and a decrease in TCA cycle metabolites [64]. This glycolytic phenotype was attributed to a more hypoxic microenvironment as inhibition of VEGF resulted in smaller and fewer blood vessels to the tumor [63]. A similar glycolytic shift was also reported by Kuang et al., where GLUT3 expression was significantly increased in bevacizumab-treated tumors and GLUT3 overexpression alone was able to mimic the bevacizumab-resistant cell phenotype. Ultimately, this pool of data indicates a role for cellular metabolism in the promotion of therapeutic resistance in GBM.

6.4. Metabolic Alterations in BTICs Related to Therapeutic Resistance

BTICs have been reported to be more resistant to therapeutic intervention than their non-BTIC counterparts, but the impact of metabolic differences on this phenotype remains to be fully determined, 130, 136]. Ye et al. described that therapy resistant BTICs, cultured both under BTIC enriching conditions and sorted for CD133⁺ expression, reduced their glucose dependence and increased fatty acid oxidation, indicating metabolic plasticity to respond to environmental stressors (Figure 2) [137]. Resistant cells were less proliferative than sensitive cells, suggesting involvement of a more quiescent stem cell population in the resistance phenotype. Many of the pathways activated in TMZ resistant lines are involved in maintenance and repair as well as suppressing growth [137]. Taken together, the current state of the research displays a major role for bioenergetics in the development of therapy resistance which still requires further investigation. The ability of GBM to adapt to multiple microenvironments highlights the difficulty that is faced when attempting to target metabolism as a means of anti-BTIC or GBM therapy.

Figure 2. Brain Tumor Initiating Cells Display Metabolic Plasticity.

Brain tumor initiating cells may adapt to tumor microenvironments and therapies by readily shifting their metabolic profiles, permitting improved survival.

7. Targeting Metabolic Changes in Glioma

There are very few therapies directly targeting BTICs, but there are many promising therapeutic avenues aimed at altering the metabolism of cancer cells. Most these potential therapeutics seek to alter glucose metabolism as is suggested by the larger body of research regarding glycolysis (Figure 3).

Figure 3. Strategies to Inhibit Brain Tumor Initiating Cell Metabolism.

Therapeutic options being explored in order to target the altered metabolism of brain tumor initiating cells. Treatments are unlikely to prove beneficial alone in glioblastoma patients, but combinatorial therapies may improve standard of care.

7.1. Targeting glucose metabolism

First and foremost, one therapeutic option to alter tumor cell metabolism is the direct inhibition of glucose uptake. The glucose analog 2-deoxy glucose (2-DG) inhibited glucose uptake and decreased tumor progression, but 2-DG effects are not tumor specific and may lead to undesirable side effects [6]. Azzalin et al. additionally reported that the GLUT inhibitors idinivar and ritonavir decrease GBM cell proliferation and glucose consumption. Ritonavir also displayed a synergistic response when combined with TMZ and carmustine [138]. A GLUT1 inhibitor, WZB117, prevented BTICs from initiating tumors when pretreated; however, it did not prevent progression of existing tumors [139, 140]. The more ubiquitous expression of GLUT1 also makes it a complex target [6, 55]. Further investigation into selective glucose uptake inhibitors may prove useful as a therapeutic option for targeting BTICs.

Inhibition of glycolysis is also possible further downstream. Dichloroacetate (DCA), a pyruvate dehydrogenase kinase (PDK) inhibitor already in use to treat lactic acidosis, has shown some growth inhibition both in vitro and in vivo [141]. In these studies PDK1 and HIF1α expression was significantly increased following radiation in U87 and primary GBM line, RN1. Radiation also increased glycolysis as evidenced by an increase in the extracellular acidification rate (ECAR) and a decrease in the oxygen consumption rate (OCR) [141]. The addition of DCA attenuated changes in OCR and ECAR, indicating a reversal of the glycolytic shift that occurs following radiotherapy [141]. This makes DCA an interesting compound for battling radioresistance in GBM [141].

Another glycolytic inhibitor, 3-bromo-2-oxopropionate-1-propyl ester (3-BrOP) showed promising results for targeting BTICs, particularly under hypoxic conditions in resistant BTICs. Combining 3-BrOP with carmustine, a chemotherapeutic agent, resulted in a significantly synergistic effect with decreased sphere formation in vitro and tumor formation in vivo by inhibiting glyceraldehyde-3-phosphate dehydrogenase [130]. Not only did this therapeutic option target BTICs, but also therapeutic resistance.

7.2. Targeting mitochondrial dysfunctions

Mitochondria-targeted therapeutics may include: modulation of mitochondrial energy metabolism, induction of mitochondria-related apoptosis, increasing sensitivity to ROS, and mitochondrial priming to enhance the efficacy of chemotherapies [85]. Hegazy et al. screened a drug library for small molecules with the ability to inhibit mitochondrial activity which ultimately decreased ATP production. One identified drug, nigericin, induced mTOR inactivation, increased autophagy, and decreased sphere formation and expression of BTIC markers in patient-derived xenolines [142]. In vivo, nigericin decreased tumor growth in association with decreased cell proliferation and regions of pseudo-palisading necrosis.

The diabetes drug metformin and its more potent analog, phenformin, have been shown to inhibit complex I. Metformin has previously been shown to inhibit the growth of TICs and prolong tumor remission [143]. Phenformin, shows promise for greater anti-tumor effect in multiple cancers [143], but clinical use is complicated by the possibility of lactic acidosis. In a study by Jiang et al., phenformin inhibited self-renewal and marker expression in BTICs as well as decreased tumor growth in vivo [143]. Phenformin was also able to be combined with TMZ and DCA, of which both increased the anti-tumor effect in vitro and prolonged survival of xenograft bearing mice [143].

Ivermectin, an anthelmintic drug shown to be anti-proliferative and antiangiogenic in GBM models, was reported to act via inhibition of mitochondrial respiration (affecting ETC complex I) and upregulation of mitochondrial superoxide [144]. Additional evidence for targeting ETC complexed is reported by Macchioni et al., where anti-glycolytic bromopyruvate effectively caused death in GL15 glioma cells with ROS elevation correlating with CcO degradation [145].

Moreover, Shen et al. developed an arsenic based mitochondrial toxin, 4-(N-(S-penicillaminylacetyl)amion) phenylarsonous acid (PENAO), that is was in a phase I clinical trial. PENAO triggered the mitochondrial apoptotic pathway and was shown to be efficacious in pancreatic cancer xenografts. In a 2015 study, Shen et al. combined DCA and PENAO to target GBM cells. The combination of the two drugs decreased the growth of GBM cells in vitro and induced G2/M cell cycle arrest. Through oxidative stress, DCA promoted the apoptotic effect of PENAO. This data was also replicated in vivo, indicating a potential for combinatorial trials in a clinical setting [146]. Targeting the mitochondria provides many opportunities to decrease tumor growth or induce apoptosis; however, further studies will be needed before clinical trials can begin.

7.3. The ketogenic diet

The ketogenic diet is a high fat, low carbohydrate diet adapted for use against brain tumors that had been previously used as an effective therapy for controlling seizures in children. [147]. The diet is designed to force the body to utilize ketone bodies instead of glucose for energy metabolism with the notion that the more glycolytic tumor cells will be selectively inhibited. In a small study on two patients, who had limited success following radiation and chemotherapy, the ketogenic diet increased long-term survival and decrease glucose utilization of the tumors as shown by positron emission tomography using fluoro-deoxy-glucose (FDG-PET). While the majority of ketogenic diet related clinical trials focused on safety, two studies in Germany have evaluated efficacy [147-149]. One reported some improvement without adverse side effects, but this study failed to look at the blood glucose or ketone levels of the treated patients. The second, the ERGO trial, saw a trend toward prolonged progression free survival (6 vs 3 weeks) in patients with stable ketosis as measured by urine ketone content [150]. This study also concluded that diet alone is unlikely to have a significant clinical outcome, but may improve patient survival when used in combination with other therapies [150].

Seyfried et al. utilized a similar method in mouse experiments, called dietary restriction, in which caloric intake is decreased with the main goal of decreasing glucose availability and increasing ketone bodies. Dietary restriction inhibited intracerebral CT-2A brain tumor growth in association with antiangiogenic and pro-apoptotic effects: IHC of intracranial U87-MG GBM tumors in mice determined reduced levels of biomarkers for angiogenesis (VEGF, IGF-1) and increased apoptosis markers (cleaved caspase-3, poly(ADP-ribose)polymerase cleavage) [148, 151]. In the mouse GBM models VM-M3 and GL261-luc2, mice treated with dietary restriction/ketogenic diet had more well-defined tumor borders, lower levels of Ki-67 staining and fewer blood vessels indicating a decrease in invasion, proliferation, and angiogenesis [148, 152]. Recently, Martuscello et al developed a less restrictive ketogenic like diet that provided similar effects that were associated with a reduced mTOR expression. The hope is that patients would be able to better adhere to this diet than the ketogenic diet[153]. Thus, these studies indicate that ketogenic diet may provide some therapeutic benefit for brain tumor patients.

In contrast to the studies above with larger numbers of subjects, Zucoli et al. performed a case study on the use of the ketogenic diet along with standard of care. Following incomplete tumor resection, the patient underwent a water-fast and then took part in the restricted calorie ketogenic diet. This diet supplied the patient with approximately 600 kcal/day and was supplemented with multivitamins. The patient had to change to a caloric restricted, non-ketogenic diet following hyperuricemia. During this time, the patient also received standard therapy of radiation and chemotherapy with TMZ. At a six-month MRI, the patient showed no definitive recurrent disease; however, the restricted diet was suspended and recurrence was seen approximately five months later. At the end of the ketogenic study, the patient had lost approximately 20% of their body weight but remained in good health [149]. Schwartz et al. reviewed 5 cases in which the ketogenic diet had been utilized in glioma patients in addition to studying the effects in 2 of their own patients. Five of the seven cases explored in detail had stable disease for between 6 weeks and 5 years following treatment with the ketogenic diet. They concluded that the ketogenic diet is safe and effective in some patients but only one case used the ketogenic diet alone so the potential as a monotherapy could not be assessed [154]. The heterogeneity of glioma may also dictate patient response to the ketogenic diet. In melanoma with the V600E BRAF mutation, the ketogenic diet may promote tumor growth indicating a similar outcome could be possible in BRAF mutant gliomas [155].

Currently, there is a lack of substantial data to support the implementation of dietary restriction/ketogenic diet which makes it difficult to broadly utilize this option in the clinic. Clinical trials would be necessary to develop a standard protocol of caloric restriction and to test the safety of severe restriction. Patient compliance would also be a factor that would need to be accounted for and tracked. As BTICs are typically more glycolytic than their non-BTIC counterparts, it would be interesting to more thoroughly investigate whether BTIC enrichment post-chemo or radiotherapy is inhibited when paired with dietary restriction/ketogenic diet. However, BTICs alter their metabolism to adapt to microenvironmental pressures and can better survive low glucose conditions. Thus, it is difficult to know without additional in vivo studies whether this approach could have benefit in targeting the BTIC fraction or preventing maintenance after the standard of care.

8. Conclusions

Current research indicates a substantial role for metabolic alterations in GBM and support of BTIC phenotypes. Not only is altered cellular metabolism an emerging hallmark of cancer, but it plays a role in many other cancer hallmarks including dysregulated proliferation, resistance to cell death, angiogenesis, and invasion [1]. While we are beginning to understand this extremely complicated phenomenon, there are still many unanswered questions, particularly in determining how cell type or microenvironment specific these alterations are. The development of models to more accurately study tumor microenvironments are necessary as they have a large effect on many cellular processes and likely contribute to the translational failure of pathway inhibitors investigated in vitro.

Emerging evidence suggests more major and critical shifts in cellular energetics and metabolism in TICs. The maintenance of TICs may require the alternation of different signaling pathways implicated in metabolic management. Further investigation of metabolic changes that support the TIC population is needed as there are only a few studies focused on them. It is also important to note that the culture conditions differ between TICs and non-TICs, which may influence metabolic differences seen in these cells. Unfortunately, the culture differences are necessary to support the maintenance of different cell types in vitro. Improved techniques for in vivo analysis of tumor cell subsets would greatly improve our ability to tease out these differences in metabolism. In terms of better therapies, targeting metabolic alterations poses a great opportunity but also a great challenge. While the bioenergetics of cancer cells are vastly different from those of normal cells, the components are often the same and the likelihood of off-target effects remains high. We need to develope options that can specifically target cancer cells and BTICs to most effectively slow tumor progression or prevent recurrence.

List of Abbreviations:

(ACSVL3)

acyl-coA synthetase VL3

(ALDH1)

aldehyde dehydrogenase

(ALDOA)

aldolase A

(ALDOC)

aldolase C

(AMPK)

5’ AMP-activated protein kinase

(ASNS)

asparagine synthetase

(BTICs)

brain tumor initiating cells

(3-BrOP)

3-bromo-2-oxopropionate-1-propyl ester

(CPT1A)

carnitine palmitoyltransferase 1A

(CPT1C)

carnitine palmitoyltransferase 1C

(G-CIMP)

CpG island hypermethylator phenotype

(CcO)

Cytochrome c Oxidase

(2-DG)

2 deoxy-glucose

(DCA)

dichloroacetate

(DRP1)

dynamin related protein 1

(EGFR)

epidermal growth factor receptor

(ETC)

Electron Transport Chain

(ECAR)

extracellular acidification rate

(FABP7)

fatty acid binding protein-7

(FASN)

fatty acid synthase

(GBM)

glioblastoma

(G6PD)

glucose-6-phosphate dehydrogenase

(GLUT1)

glucose transporter 1

(GLUT3)

glucose transporter 3

(HK)

hexokinase

(2-HG)

2-hydroxyglutarate

(HMGCR)

3-hydroxy-3-methylglutaryl-CoA reductase

(HIF1α)

hypoxia inducible factor 1-alpha

(HIF2α)

hypoxia inducible factor 2-alpha

(IMP2)

insulin-like growth factor 2 mRNA-binding protein 2

(IDH)

isocitrate dehydrogenase

(LDH)

lactate dehydrogenase

(LXRs)

liver X receptors

(LDLR)

low density lipoprotein receptor

(MGMT)

O-6-Methylguanine-DNA Methyltransferase

(NF1)

Neurofibromin

(NF-κB)

nuclear factor kappa beta

(OCR)

oxygen consumption rate

(OXPHOS)

oxidative phosphorylation

(PDGFR)

platelet derived growth factor receptor

(PENAO)

4-(N-(S-penicillaminylacetyl)amion) phenylarsonous acid

(PFKPB4)

6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4

(PEP)

phosphoenolpyruvate

(PPP)

pentose phosphate pathway

(FDG-PET)

positron emission tomography using fluoro-deoxy-glucose

(PDH)

pyruvate dehydrogenase

(PDK1)

pyruvate dehydrogenase kinase 1

(PDP1)

pyruvate dehydrogenase phosphatase catalytic subunit 1

(PKM2)

pyruvate kinase muscle isozyme 2

(ROS)

reactive oxygen species

(TCA cycle)

tricarboxylic acid cycle

(TMZ)

temozolomide