Hypoxia in the glioblastoma microenvironment: shaping the phenotype of cancer stem-like cells

Abstract

Glioblastoma is the most common and aggressive malignant primary brain tumor. Cellular heterogeneity is a characteristic feature of the disease and contributes to the difficulty in formulating effective therapies. Glioma stem-like cells (GSCs) have been identified as a subpopulation of tumor cells that are thought to be largely responsible for resistance to treatment. Intratumoral hypoxia contributes to maintenance of the GSCs by supporting the critical stem cell traits of multipotency, self-renewal, and tumorigenicity. This review highlights the interaction of GSCs with the hypoxic tumor microenvironment, exploring the mechanisms underlying the contribution of GSCs to tumor vessel dynamics, immune modulation, and metabolic alteration.

The past several decades of cancer research have identified a list of purported mechanisms of tumorigenesis. Complexity of cancer is illustrated by various descriptive models, a disease of the “genome,” a dysregulated “developmental disorder,” and a “wound” that fails to heal. As reflected by the evolving conceptual models, the research field has benefited from expanding beyond exclusively cell autonomous or tumor-centric investigations to include potential targets in the surrounding microenvironment. Rather than dissociating tumor tissues and studying cells in isolation, a better understanding of the oncogenic impact of local niche factors requires identification of relevant in vivo conditions to recapitulate this complex biology. Technical barriers had hindered progress, but with increasing availability of refined research tools and advances in the understanding of tumor pathophysiology, studies evaluating the heterogeneity of the in vivo glioblastoma microenvironment are now possible.

Although the concept of cancer stem cells has generated considerable interest over the past decade, as early as the mid-1800s, Rudolf Virchow suggested that neoplastic diseases may arise from undifferentiated embryonic-like tissues.¹ Virchow’s hypothesis was likely founded upon astute observations of the histologic similarities between specimens obtained during early human development and the undifferentiated appearance of many tumors. A detailed understanding of normal embryologic processes helped to establish a definitive link between cancer pathogenesis and the mechanisms signaling the differentiation of developing tissues. Advances in the field of developmental biology and laboratory tools such as monoclonal antibodies to target specific subpopulations of cells led to the delineation of the hematopoietic cellular ontogeny in the mouse.² Identification of the biological events driving the differentiation of the normal hematopoietic stem cell allowed for another important discovery: the molecular signature of a leukemic stem cell capable of recapitulating the disease in an immunodeficient host.³ The notion of the “cancer stem cell” was introduced with much controversy, a dispute that extended to glioblastoma.^4,5 Depending on the context of discussion, it has been suggested that “cancer stem cell” be replaced with “cancer-initiating cell,” “cancer stem-like cell,” or even “cancer recapitulating cell” in an attempt to pin down a term that faithfully reflects the underlying biology. For the purposes of this review, the use of “cancer stem cell” does not imply that cancers originate from stem cells. Here we use this term in name only to refer to the presence of an inherently heterogeneous subpopulation of tumor cells that feature some properties of stemness, including the enhanced ability to form tumors in immunodeficient hosts.

There are specific aspects of tumor biology that the glioma stem cell (GSC) model endeavors to explain. Although there is considerable controversy over an exact definition, a functional, working understanding of the cancer stem cell model highlights the following features: a relatively undifferentiated state, the ability to self-renew, and reliable tumorigenicity when implanted in an immunodeficient host—importantly, the specific ability to recapitulate the bulk tumor heterogeneity seen in human disease.^6,7 A better understanding of GSC biology may help explain the failure of even multimodality treatment: surgical resection followed by temozolomide and concurrent radiotherapy.^8,9 However, “stemness” has been shown to be a mutable cellular quality that represents a transient, reversible state that most cells can adopt given the appropriate environmental cues.¹⁰ One of the most prominent non-cell autonomous factors influencing the glioma stem cell phenotype is the presence of hypoxia in the tumor microenvironment.

Hypoxia

Glioblastomas are characterized by extensive tissue hypoxia. Even nonneoplastic tissues exhibit lower oxygen concentration compared with ambient air. Physiologically normoxic conditions for adult human tissues range from 2% to 9% O₂, much lower than inspired 20.8% O₂ atmospheric air.¹¹ Low oxygen environment is often a necessary condition to prompt development and maintenance of normal somatic stem cells ranging from hematopoietic to neural compartments.^12,13 Tissue oxygen tension (pO₂) values vary greatly within the normal CNS depending on region, ranging from as low as 0.5% O₂ in the midbrain to 8% O₂ in the pia.¹⁴ A 2002 study recorded intra- and peritumoral oxygenation in patients with glioblastoma in addition to external factors affecting tissue pO₂ levels. Unsurprisingly, intratumoral tissue was consistently less oxygenated (1.25% O₂) than peritumoral areas (2.5% O₂), a trend that worsened when patients were asleep (1.0% vs 1.8% O₂). Inhalation of 100% O₂ did not improve oxygen levels in either tissue, but osmotic diuretics did increase peritumoral pO₂, suggesting that increased intracranial or intratumoral pressure impacts oxygen delivery. However, hyperbaric oxygen was able to improve oxygenation of tissues both within and surrounding the tumor.¹⁵

The development of intratumoral hypoxia occurs over time in 3-dimensional space, and the degree of tissue oxygenation is likely a dynamic process that varies continuously. Recent studies with single cell RNA and genome sequencing analyses have demonstrated striking intratumoral heterogeneity for profiles that correspond to both hypoxia and stemness.^16,17 Each individual tumor contained a gradient of hypoxia markers, indicating that the population of tumor cells respond to a range of oxygen tension in the surrounding microenvironment. Furthermore, this evidence suggests that the “stemness” property of the GSC populations is not fixed, but is rather a transient phenotype that can vary along a broad spectrum. Hypoxia has been described as an essential environmental cue for the maintenance of GSCs, the cell population believed to be responsible for tumor resistance to chemotherapy and radiation.^6,18–19 Because GSCs likely contribute to tumor recurrence, strategies aimed at eliminating this subpopulation of tumor cells or inducing terminal differentiation are being actively investigated.²⁰ In addition to enhancing maintenance of GSCs, hypoxia contributes to resistance to irradiation, the most effective therapeutic modality for patients with glioblastoma.⁹ Presence of low oxygen content in tumor tissues impedes radiation-induced generation of free radicals that assert antitumor effect. Development of imaging methods to quantify oxygen content in tumors, such as the oxygen-enhanced MRI platform, may facilitate identification of regional hypoxia in order to boost radiation dose appropriately.²¹ Hypoxia may further contribute to tumor pathogenesis by promoting genomic instability. In certain mammalian cells, low oxygen environment can attenuate expression of DNA mismatch repair genes, which may then promote acquisition of additional gene mutations.²²

Driven by these mechanisms, hypoxia is associated with a more aggressive tumor phenotype.²² The immediate adaptive cellular response to low levels of oxygen is the stabilization of transcription factors known as hypoxia inducible factors (HIFs). These proteins are involved in the maintenance of the stem cell–like phenotype of certain cancers.²³ This hypoxia-dependent mechanism has also been shown in glioblastoma cell lines, where differences in oxygen tension alter the expansion and maintenance of GSCs and promote their tumorigenic capacity and expansion in a HIF-1α–dependent mechanism.^18,24 HIF-1α has also been linked with increased invasiveness and angiogenesis, further contributing to aggressive behavior. While both GSCs and non-GSCs within glioblastoma exhibit increased HIF-1α expression, the GSCs also demonstrate increased levels of HIF-2α.²⁵

An important downstream activation of HIF signaling cascade is angiogenesis. Not surprisingly, a histologic hallmark of glioblastoma is vascular proliferation. Despite the apparent abundance of angiogenic activity, a characteristic feature of glioblastoma is abnormally functioning blood vessels. Vessels lack the structural support of pericytes and are characteristically tortuous and leaky, impeding blood flow and decreasing perfusion.²⁶ At the tumor tissue level, microvascular thromboses occlude vessels, further promoting intratumoral hypoxia.²⁶ Because the volume of blood flow influences regional tissue oxygenation, understanding the crosstalk between GSCs and the neighboring blood vessels may provide novel therapeutic opportunities.

Tumor Vessel Dynamics

The application of anti–vascular endothelial growth factor (VEGF) therapy for glioblastoma was reasonable considering that angiogenesis is a defining “hallmark of cancer” and a prominent feature of glioblastoma.²⁷ This approach may have the additional benefit of targeting the GSC perivascular niche, the goal being to cut off the nutrient supply to this resistant cell population. Although treatment with bevacizumab revealed a reduction in the size of the enhancing mass, at times subjects relapsed with adoption of an invasive phenotype.²⁸ A xenograft mouse model of U87 glioma cells, characteristically described as non-invasive, was found to display “satellitosis” or “secondary structures” representing migratory tumor cell aggregates in perivascular regions and in Virchow–Robin spaces.²⁹ A study of glioblastoma patients undergoing anti-VEGF therapy revealed a sizable cohort of patients (30%) who developed multifocal or diffuse patterns of recurrence. Comparison of tissue samples with primary tumor revealed increased expression of genes associated with cell invasion and/or motility.³⁰ A potential mechanism to explain such observation could be that VEGF normally suppresses cell invasiveness by blocking hepatocyte growth factor–mediated activation of the MET/VEGF receptor 2 heterocomplex. When VEGF is inhibited, a T-cadherin to N-cadherin switch is seen along with increased migratory behavior of the cell, reminiscent of epithelial-to-mesenchymal transition.³¹ It was thought that the “vascular normalization” expected to occur as a result of anti-VEGF therapy would reduce the amount of intratumoral hypoxia; however, with long-term treatment the opposite was found to occur.²⁹ Furthermore, the GSC response to certain therapies may induce new intratumoral vessel formation. Repeated exposure to radiation led to increased secretion of insulin-like growth factor 1 by GSCs, resulting in radioresistance and enhanced secretion of VEGF with accompanying angiogenesis.³²

While less resilient cell types may cope poorly in hypoxia, the GSC population not only persists but may activate a variety of adaptive mechanisms of neovascularization, including angiogenesis, vasculogenesis, vessel co-option,³³ intussusception,³⁴ and vascular mimicry.³⁵ Hypoxia-driven regulation of factors modulating the angiogenic response such as VEGF and angiopoietin, both products of the HIF-activated transcriptional cascade, is robust in glioblastoma.^36,37 Specifically, GSCs create a perivascular gradient of VEGF that selects for an endothelial “tip cell” induced to undergo tube formation, ultimately giving rise to a new capillary. This mechanism of VEGF-mediated angiogenesis has been attributed to the GSC population in particular.³⁸ The symbiotic relationship between endothelial cells and GSCs is perhaps best characterized by the concept of the perivascular niche. Excess levels of VEGF not only promote proliferation and survival of endothelial cells, but also confer survival advantages to GSC. Indeed, a VEGF autocrine loop was found to maintain a stem-like phenotype via binding to VEGF receptor 2 on the surface of GSCs.³⁹ The process of angiogenesis further supports the GSC population via several endothelial cell–mediated mechanisms. Notch signaling between endothelial cells and GSCs assists in maintaining the stem cell phenotype. Endothelial upregulation of Delta-like-4 (DLL-4) and JAGGED1 result in tip-cell formation and these factors may also simultaneously engage the Notch receptor on perivascular GSCs.^40,41 Furthermore, the endothelial release of nitric oxide diffuses into GSCs and acts via Notch signaling in a similar manner.⁴² Hypoxia not only stimulates angiogenesis but also promotes maintenance of GSC stemness; HIF-1α physically interacts with the Notch intracellular domain (NICD) to stabilize downstream signaling, thereby preventing cell differentiation.⁴³

Vasculogenesis, an alternative mechanism leading to blood vessel formation, is dependent on the recruitment of cells originating in the bone marrow. Bone marrow–derived endothelial progenitor cells (EPCs) exit systemic circulation along a chemokine gradient and function to support endothelial survival and proliferation.⁴⁴ These recruited bone marrow–derived cells can even have a restorative influence on radiation-induced vascular damage.⁴⁵ Interestingly, anti-VEGF therapy has been shown to facilitate increased homing of these bone marrow–derived cells to the tumor microenvironment in glioblastoma, a potential compensatory mechanism that becomes upregulated upon inhibition of a related pathway.⁴⁶ The chemoattraction of bone marrow–derived cells into the periphery has been largely attributed to the stromal cell–derived factor 1 (SDF-1)/C-X-C chemokine receptor type 4 (CXCR-4) axis. GSCs are responsible for the secretion of SDF-1, a soluble factor that engages CXCR4 on circulating myeloid cells.⁴⁴ This process is mediated by HIF-1–dependent signaling that augments secretion of SDF-1 by GSCs and upregulates CXCR4 on myeloid cells.^18,47 A study of a mouse model found that inhibition of vasculogenesis using AMD3100, a small molecule inhibitor of the SDF-1/CXCR4 axis, prevented development of functional vasculature, presumably by interrupting influx of bone marrow–derived cells, and subsequently led to recurrence of glioblastoma following radiation therapy; however, such an effect was not observed if only angiogenesis, not vasculogenesis, was disrupted.⁴⁸ Mechanisms of tumor-associated blood vessel formation is illustrated in Figure 1.

Fig. 1.

Tumor vessel dynamics. Mechanisms of tumor-associated blood vessel formation. HIF-mediated VEGF is secreted by GSCs during angiogenesis, resulting in endothelial tip cell formation and sprouting of vessels. HIF promotes Notch signaling by stabilizing the liberated intracellular domain (ICD). Bone marrow–derived endothelial progenitor cells (EPCs) are recruited by GSC-secreted SDF-1. CSL = CBF1/suppresor of Hairless/Lag1; NO = nitric oxide.

Additional mechanisms by which gliomas enrich for vascular supply are vessel co-option, intussusceptive microvascular growth, and vascular mimicry. Tumor growth may initially begin by co-opting host vasculature rather than relying on angiogenesis.^33,49 Once the neoplastic mass reaches a critical size, angiogenesis is necessary to support further growth. How GSCs may regulate vessel co-option is unclear. Intussusceptive microvascular growth is a form of vascular remodeling in which an existing vessel is split in two by establishing transluminal pillars or transendothelial bridges.³⁴ Disturbingly, this form of “splitting” or “non-sprouting” angiogenesis is stimulated upon exposure to anti-angiogenic agents as a compensatory response.⁵⁰ Another mechanism of intratumoral vessel formation is vascular mimicry, in which functional vascular channels constructed by tumor cells and unlined by endothelial cells, yet capable of transporting red blood cells, emerge.³⁵ Such formation of tumor cell–constituted vascular channels is observed more frequently in higher-grade tumors.⁵¹

Immunomodulation

The persistence of an immunosuppressive microenvironment is a defining biological feature of glioblastoma, with numerous interconnected obstacles that preclude an effective antitumor response. Hypoxic and inflammatory responses overlap in the GSC niche to shape the immune microenvironment by acting upon the GSC as well as infiltrating immune cells. Lymphocytes routinely encounter hypoxic environments during development and migration through peripheral tissues.⁵² Engagement of the T-cell receptor induces synthesis of HIF-1α, and the protein is stabilized in hypoxic conditions.⁵³ In vitro studies have demonstrated that hypoxia enhances the lytic capability of CD8+ T lymphocytes⁵⁴ and interferon-gamma secretion by CD4+ T cells.⁵⁵ Yet, oxygen restriction also impairs cytotoxic T lymphocyte (CTL) development, proliferation, and expression of inflammatory cytokines.⁵⁴ GSCs, in contrast to differentiated tumor cells, preferentially inhibit the proliferation of activated T cells, although the precise mechanisms are not fully delineated.⁵⁶ One potential mechanism involves secretion of galectin-3, a molecule that binds mature tumor-specific T cells within the tumor microenvironment and induces T-cell apoptosis.⁵⁷ Galectin-3 knockout mice demonstrate enhanced tumor-free survival and increased effector T-cell functions.⁵⁸ Although the source of galectin-3 (tumor vs the microenvironment) is unclear, galectin-3 expression has been associated with hypoxic regions within the tumor. Interactions between the GSCs and immune cells is summarized in Figure 2.

Fig. 2.

Immune modulation. Effects of HIF stabilization are cell type specific, including increased TAM expression of PD-L1 and increased CD4+ T-cell secretion of interferon-gamma (IFN-γ). NANOG-mediated secretion of TGF-β by GSCs supports local immunosuppressive cells and promotes stem cell phenotype in an autocrine fashion.

The signal transducer and activator of transcription 3 (STAT3) signaling pathway is also activated under hypoxic conditions within solid tumors, including glioblastoma. STAT3 has been identified as a key molecular signaling node for GSCs. Extracellular signals such as interleukin-6 and growth factors are transduced through receptor tyrosine kinases such as epidermal growth factor receptor and platelet derived growth factor receptor β, activating STAT3 and resulting in GSC survival, treatment resistance, and immunosuppression.⁵⁹ This inflammatory pathway can mediate transcription of HIF-1α target genes, but direct regulation of HIF-1α protein by STAT3 appears to be cell specific.⁶⁰ Targeted inhibition of STAT3 or HIF-1α depletion increases tumor susceptibility to CTL-mediated lysis, further suggesting that hypoxia-induced signaling pathways in tumor cells can impede CTL activity.⁶¹

HIF-1α–dependent expression of the transcription factor NANOG is induced by hypoxia, further promoting the stemness signature of GSCs.⁶² Small interfering RNA–mediated NANOG knockdown sensitized breast cancer cells to lysis by tumor-specific T cells, linking hypoxia and the stem cell phenotype to immune evasion.⁶³ NANOG binds to transforming growth factor (TGF)-β1 promoter, increasing production of the cytokine. TGF-β appears to support the maintenance of GSC phenotype as it engages an autocrine feedback loop that maintains stemness via increased expression of sex determining region Y box (Sox)2 and Sox4, a mechanism that is enhanced in hypoxic conditions.⁶⁴ TGF-β decreases major histocompatibility complex II expression on tumor cells, thereby impairing antigen presentation.⁶⁵ Experimental targeting of NANOG results in the depletion of tumor associated macrophages (TAMs) and regulatory T cells (Tregs), replacing the immune infiltrate with antitumor effector T cells.⁶⁶ These findings support the feasibility of triggering an antitumor immune response via targeted inhibition of signaling pathways consistently upregulated in a hypoxic environment.

The hypoxic tumor microenvironment also recruits immunosuppressive cells, including TAMs, myeloid-derived suppressor cells (MDSCs), and Tregs.⁶⁷ These cells dampen the immune response through secreted factors and expression of surface molecules that engage inhibitory molecules on effector immune cells. As mentioned previously, GSCs have been commonly found in the tumor hypoxic niche, placing them in close proximity to immune suppressive cells also localized to that region. Investigations into the molecular crosstalk in this location revealed that GSCs secrete a chemoattractant, periostin, that specifically recruits immunosuppressive M2-polarized TAMs.⁶⁸ GSCs were separately found to secrete macrophage migratory inhibitory factor that upregulates the immunosuppressive function of MDSCs, specifically their synthesis of arginase-1.⁶⁹ The T-cell inhibitory programmed cell death ligand 1 (PD-L1), contains a hypoxia response element in its promoter that can directly bind HIF-1α in MDSCs, macrophages, dendritic cells, and tumor cells.⁷⁰ Upregulated PD-L1 in hypoxic settings can interact with PD-1 on activated T cells to provide inhibitory signals to T cells, demonstrating an immunosuppressive role of the tumor microenvironment via direct cell-to-cell contact.

TAMs are the predominant immune cell infiltrating glioblastoma tumor tissues, accounting for 0.5%–30% of the entire cell population, and correlate with the grade of malignancy.⁷¹ The location of TAMs suggests crosstalk with GSCs, as populations of TAMs are overrepresented near tumor microvasculature as well as in regions of hypoxia.^71,72 GSCs are more likely than non-GSCs to recruit TAMs, and this behavior increases during hypoxia.⁷² The GSC-secreted soluble factors mediating this chemoattraction include VEGF, matrix metalloproteinase 2, TGF-β, and SDF-1.⁷² Although not explicitly dependent on hypoxia, TAM polarization toward an immunosuppressive M2 phenotype can be induced by colony stimulating factor 1, TGF-β1, and macrophage inhibitory cytokine 1, all secreted factors of GSCs.⁷³ TAMs with an M2 phenotype have multiple effects on the glioma microenvironment, including increased angiogenesis, remodeled extracellular matrix, and enhanced tumor cell invasion.⁷⁴ GSCs impact basic TAM functions, resulting in decreased phagocytosis and increased production of immunosuppressive cytokines by TAMs.⁷³ HIF-related transcriptional activity overlaps with the nuclear factor-kappaB (NF-κB) pathway in cells of the innate immune system, regulating pro-inflammatory cytokines such as interleukin-1 and TNF-α, in addition to the immunomodulating enzymes cyclo-oxygenase 2 and inducible nitric oxide synthase. The regulatory functions shared between NF-kB and HIF are evident by the extensive crosstalk between the 2 signaling cascades.⁷⁵

Hypoxia-Induced Metabolic Alterations

Knowledge of the altered metabolism of cancer cells dates back to Otto Warburg’s observation nearly a hundred years ago that cancerous cells preferentially ferment glucose for energy generation.⁷⁶ Aerobic glycolysis has been confirmed in many cancer types, including glioblastoma, where even in the presence of abundant O₂ concentrations, pyruvate is converted to lactate rather than entering the tricarboxylic acid (TCA) cycle.^77,78 It appears that this “Warburg effect” is at least in part due to activation of oncogenes and loss of tumor suppressors that switch cancer cell metabolic activities so as to decrease the use of oxidative phosphorylation in preference for an increase in glycolytic activity and lactate production.⁷⁹ Hypoxia, by stabilizing HIF-1α, further contributes to aerobic glycolysis by amplifying glucose uptake by tumor cells and promoting conversion of pyruvate to lactate rather than acetyl-CoA (Figure 3). Stabilized HIF-1α activates the transcription of glucose transporters, glycolytic enzymes phosphofructokinase 1 (PFK1), aldolase, hexokinase 2 (HK2), and lactate dehydrogenase A (LDHA).⁸⁰ Glycolytic intermediates generated by incomplete oxidation are used for anabolic reactions to synthesize key intermediates such as nucleotides, phospholipids, and amino acids, which provide the building blocks required for tumor cell growth and proliferation.

Fig. 3.

Metabolic alterations. Cells respond to hypoxic environment by upregulating glucose transporter (GLUT)1/3 transporters, which facilitates glucose influx. HIF activation upregulates glycolytic enzymes and inhibits oxidative phosphorylation. HIF decreases pyruvate entry into the TCA cycle via upregulation of PDK1 and PDK1-mediated deactivation of PDH. Increased glycolysis leads to lactate accumulation; however, MCT facilitates its extracellular transport. HK = hexokinase; PDK1 = pyruvate dehydrogenase kinase 1; DHAP = dihydroxyacetone phosphate.

One well-described mechanism by which HIF-1α attenuates the flux through the TCA cycle involves upregulation of pyruvate dehydrogenase (PDH) kinase 1, which leads to the inactivation of the catalytic subunit of PDH via phosphorylation. Inactivation of PDH prevents formation of acetyl-CoA and therefore inhibits entry of pyruvate into the TCA cycle. In addition, HIF-1α activates the transcription of the LDHA gene, which encodes for LDHA enzyme.⁸⁰ Increasing expression of LDHA allows conversion of the otherwise excess pyruvate into lactate. LDH is the mediator for the shift in metabolism from oxidation in the mitochondria to glycolysis in the cytoplasm, and LDHA is largely found in muscle and tumor tissue. Recently, it was shown that chemical inhibition of LDHA in GSCs can successfully induce their differentiation and/or apoptosis.⁸¹ The combined effect of LDHA overexpression and PDH inhibition is an increase in aerobic glycolytic activity which a variety of macromolecular intermediates feed into, decreased TCA cycle, and accumulation of lactate. Since aerobic glycolysis leads to formation of only 2 molecules of ATP per glucose, to keep up with the energetic demands, the GSCs rely on HIF to activate the transcription of genes that encode for glucose transporters to facilitate glucose uptake.¹⁶ Therefore, GSCs are more efficient at nutrient uptake than their non-GSC neighbors.⁸² Lactate accumulation leads to acidification, and tumor cells cope by releasing it into the extracellular space with the help of HIF upregulated plasma membrane monocarboxylate transporter isoforms 1 and 4.^80,83,84 Hypoxic tumor subpopulations may be creating a symbiotic microenvironment where metabolic substrates are utilized in a complementary fashion such that surrounding tumor cells benefit. In fact, metabolic codependence in other forms of cancer has been demonstrated in which hypoxic cells produce lactate that can then be utilized by normoxic tumor cells as fuel.^85–87

Because the formation of acetyl-CoA and its entry into the TCA cycle is further limited in hypoxia, tumor cells may switch to glutamine as the main carbon source for the production of lipids and maintenance of an anabolic state. In our experience, glioblastoma cells propagated in a hypoxic environment were found to depend on glutamine for the production of citrate via the reductive carboxylation of α-ketoglutarate catalyzed by isocitrate dehydrogenase (IDH) 2. Cells cultured in medium that lacked glutamine under hypoxic conditions showed decreased viability, suggesting the important role of glutamine for cell survival in reduced oxygen conditions. Other TCA cycle metabolites were also formed via this reductive carboxylation pathway. The purpose of these metabolic shifts seen in glioblastoma may be a compensatory response driven by the need to provide substrates so as to sustain the anabolic state of the tumor.^88,89 Nevertheless, studies have found that GSCs exhibit variable glucose dependence and mitochondrial oxygen consumption, likely suggesting the importance of native microenvironmental cues.⁸²

As discussed, altered tumor cell metabolism was largely attributed to loss of tumor suppressors and to the activation of oncogenes.⁷⁹ However, the discovery of tumors with a gain-of-function mutation in enzymes involved in routine metabolic pathways transforms the way we think about the reprogramming of metabolism in cancer. An example is provided by the recent identification of mutations involving the IDH genes. That a mutation of a constitutive enzyme of the TCA cycle may contribute to tumorigenesis was unexpected by most investigators, but may offer opportunities to further explore GSC biology within the context of dysfunctional metabolism. IDH mutations are known to promote histone methylation, leading to loss of demethylation-dependent differentiation into mature lineages.⁹⁰IDH mutant cells lose balance of redox potential through the disruption of the cytoplasmic and mitochondrial ratio of NADP/NADPH. Subsequent drastic depletion of NADPH results in toxic vulnerability to increased reactive oxygen species.^91,92 The IDH mutant phenotype leads to the accumulation of the metabolite D-2-hydroxyglutarate (D-2HG), a potent inhibitor of α-ketoglutarate–dependent enzymes, including histone demethylases.⁹³ In IDH mutant cells, HIF-1α was believed to be stabilized in normoxic conditions leading to angiogenesis and widespread histone methylation, promoting a state of “dedifferentiation” of glioblastoma multiforme cells into GSCs.⁹⁰ The IDH mutant state was hypothesized to induce a de facto “pseudohypoxic” state by stabilizing HIF expression and potentiating GSCs in normoxic conditions.⁹⁴ This upregulation of HIF signaling associated with IDH mutation was thought to serve as a compensatory mechanism to facilitate clearance of reactive oxygen species to promote cell survival. However, experimental data unexpectedly showed that in IDH mutant cells, HIF-1α was destabilized through a mechanism in which D-2HG increased the activity of prolyl hydroxylase 2, highlighting the complexity of the underlying biology.⁹⁶ Even in IDH wild-type tumor cells, hypoxia may force the cells to recapitulate a mirror-image metabolic response leading to similar epigenetic changes. Two recent reports demonstrate that hypoxia is capable of driving cells to produce the L-enantiomer form of 2-HG (IDH mutant cells generate D-2-HG).^95,96 The L-form of 2-HG is also capable of disrupting α-ketoglutarate–dependent enzymes, leading to a comparable repressive histone methylation pattern seen with IDH mutation. These observations confirm the remarkable ability of tumor cells to create a repressive epigenetic state either by a gene disruption leading to a pseudohypoxic state and generation of an oncometabolite or by exploiting a hypoxic environment that promotes production of a metabolite that facilitates an analogous outcome.

Conclusion

The concept of GSC underscores its transient state of existence. This inherent plasticity suggests that the GSC phenotype is an adaptive response, sensitive to external stress emanating from cues in the surrounding microenvironment. Given the strong association of GSCs with glioblastoma recurrence, improved understanding of the GSC population may provide key insights into the mechanisms of resistance. The hallmarks of cancer, such as neovascularization, immunomodulation, and metabolic dysfunction, are all tumor-supportive mechanisms in part interconnected by intratumoral hypoxia. Addressing the underlying common manifestation of disease such as hypoxia can provide a conceptual framework to enhance our understanding of the disease and a platform to identify more effective therapeutic interventions. The translation of laboratory research into hypothesis-based clinical trials is the foundation of the Precision Medicine Initiative. Successful extension of our increasing understanding of hypoxia in glioblastoma biology into effective therapies would be a seminal example of this integrated approach.

Funding

This work was supported by the Intramural Research Program, CCR, National Cancer Institute.

Conflict of interest statement. None.