HOT Models in Flux: Mitochondrial Glucose Oxidation Fuels Glioblastoma Growth

Abstract

Cancer cells in culture obtain ATP and biosynthetic precursors primarily by aerobic glycolysis, not by mitochondrial glucose oxidation. In this issue of Cell Metabolism, Marin-Valencia et al. demonstrate that glioblastoma, an aggressive and, in culture, highly glycolytic cancer, primarily uses glucose oxidation to meet energetic and biosynthetic demands in vivo.

Cancer cells preferentially metabolize glucose to lactate, despite sufficient levels of oxygen to support mitochondrial glucose oxidation (i.e. The Warburg effect). This provides less ATP per molecular of glucose, but cancer cells derive significant gain from the metabolic short cut. Aerobic glycolysis yields enough ATP to meet energetic demand while providing NADPH and nutrient precursors for lipid biogenesis and macromolecular synthesis (Vander Heiden et al., 2009). It has been inferred that human cancers in vivo also rely primarily on aerobic glycolysis, but do they? Unlike established cell line models that have been used to provide insight into cancer metabolic reprogramming, human tumors in situ present a much more complex and heterogeneous microenvironment. A more complete understanding of the metabolic behavior of human tumors requires models that more accurately reflect this complexity. In this issue of Cell Metabolism, Marin-Valencia et al. perform metabolic flux analysis of ¹³C-labelled nutrients in human orthotopic tumor models (HOTs), in which cells isolated from clinical tumor samples are implanted directly into the brain of mice (Marin-Valencia et al., 2012). They provide compelling evidence to suggest that glioblastomas utilize mitochondrial glucose oxidation in vivo to support aggressive tumor growth (Figure 1).

Figure 1. Glioblastomas in their native microenvironment primarily use mitochondrial glucose oxidation to meet energetic and biosynthetic demands.

(A) When cells from a patient's brain tumor are grown in culture, glucose is primarily metabolized to lactate with minimal mitochondrial oxidation. (B) Here, Marin-Valencia et al., show that when cells from a patient's brain tumor are implanted into the brain of immunodeficient mice to simulate their native microenvironment, glioblastomas primarily utilize mitochondrial glucose oxidation.

Aerobic glycolysis is not exclusive to cancer. Proliferative normal cells can also use this mechanism, but with a salient difference. Aerobic glycolysis in non-neoplastic cells remains responsive to environmental levels of nutrients and growth factors, a process that is orchestrated through tightly regulated feed-forward and feedback signaling loops (Koppenol et al., 2011). In contrast, cancer cells appear to be “addicted” to aerobic glycolysis because their regulatory circuits are overwhelmed by persistent oncogenic driver mutations. This difference could potentially be therapeutically exploited. The use of dichloroacetate (DCA), a pharmacological inhibitor of pyruvate dehydrogenase kinase (PDK), to target glioblastoma, a highly glycolytic, highly aggressive form of primary brain cancer, provides one example (Michelakis et al., 2010). Suppression of PDK by DCA derepressees the pyruvate dehydrogenase complex (PDH), enabling flux through the citric acid cycle. This targeting strategy assumes that glioblastomas rely almost exclusively on aerobic glycolysis and that forcing tumor cells to use mitochondrial glucose oxidation could be a lethal insult. There is some empirical evidence to support this hypothesis (Michelakis et al., 2010). However, a recent study of brain tumors from patients infused with ¹³C -labelled glucose does not support the underlying hypothesis, instead demonstrating extensive glucose mitochondrial oxidation (Maher et al., 2012).

Now, Marin-Valencia et al., present a rigorous analysis of the metabolic fate of labeled nutrients in glioblastomas grown in their native microenvironment (Marcin-Valencia et al., 2012). They establish and genomically characterize a set of HOT models derived from tumor cells isolated during surgical resection of human glioblastomas and implanted directly into the brain of immunodeficient. These models are never placed in culture, preventing selection for tumor cells best adapted to the artificial environment of cell culture and maintaining the heterogeneous microenvironment of the native tumor. Following establishment and molecular characterization to ensure that the models are representative of tumors from which they are derived, Marin-Valencia et al., infused the mice with ¹³C -labelled nutrients in which the carbons are labeled at various positions. By using ¹³C NMR spectroscopy of tumor tissue and normal brain extracts, the fate of the nutrients could by traced by analyzing the position of the labelled carbons in the metabolites, and by the spectra themselves (singlets, doublets, etc). Regardless of genotype, all glioblastomas characterized demonstrated glycolysis and mitochondrial glucose oxidation via PDH and the citric acid cycle. Further, the analyses demonstrated that glucose supplied the metabolic intermediates for a number of biosynthetic activities. The HOT models also contained elevated levels of glutamine, but lacked glutamine catabolism and were not dependence on exogenous glutamine for growth, a result that may potentially be explained by de novo glutamine synthesis (Marcin-Valencia et al., 2012).

This study raises a number of intriguing questions. Cancer cells demonstrate a striking ability to rewire their circuitry to maintain flux to critical downstream signaling nodes (Cloughesy and Mischel, 2011). Does the metabolic circuitry of cancer show similar plasticity? Can tumor cells utilize either mitochondrial glucose oxidation or aerobic glycolysis if one or the other route is blocked? It has already been demonstrated that cancer cells that are prevented from using glucose-derived carbons for lipogenesis, due to either genetic or pharmacologic perturbations limiting glucose flux into and through the citric acid cycle, can use glutamine as a source for lipogenesis by reductive carboxylation with Isocitrate Dehydrogenase 1 (IDH1) (Metallo et al., 2012; Mullen et al., 2012). Metabolic flux analysis of HOT models provides a promising experimental platform for dissecting the plasticity of cancer metabolic reprogramming, an issue that may prove to be critical for preventing or reversing resistance to metabolically targeted therapies. A second major theme relates to the heterogeneity of the microenvironment. Solid organ cancers, including glioblastoma are heterogeneous; individual tumor cells may differ in functional and/or molecular attributes (Gupta et al., 2011). These cells are also influenced by their location within the tumor, i.e. near to or distant from a blood vessel, and by neighboring cells through the local chemical milieu. The hypoxic response provides a possible example. Surprisingly, few of the HOTs showed evidence for robust HIFα expression on western blot. However, it is highly likely that subsets of tumor cells, particularly those near necrotic regions or distant from blood vessels may experience HIFα-dependent metabolic reprogramming and thus may differ in how they metabolize glucose and glutamine. The technology used, by necessity, assumed that the tumor cells are metabolically homogeneous. However, cell-to-cell differences in metabolic circuitry could potentially influence clonal selection, if it provides tumor cell subsets with an adaptive advantage. Future studies examining metabolic heterogeneity in these models may prove to be of enormous value. A third interesting, and potentially surprising result, is the relative uniformity of the metabolic fate of glucose in these HOT models that did not appear to depend on genotype. Future studies applying genetic and pharmacologic modulations of key signaling networks in these HOT models may yield important information as to how genetic context can influence they way tumors utilize their nutrients in the native tumor microenvironment.

This intriguing paper by Marin-Valencia et al. may have important implications for the field of cancer metabolism, and may provide a valuable blueprint for dissecting the metabolic circuitry of cancer across various tumor types. It suggests that dissection of the metabolic circuitry of cancer may require analysis in models that recapitulate the native tumor microenvironment; a theme that should resonate with investigators interested in diverse cancer types. It also demonstrates, not surprisingly, that cancers can also use mitochondrial glucose oxidation, particularly in models systems simulating their native microenvironment in vivo. Future studies will be needed to better understand how cancer cells balance aerobic glycolysis with mitochondrial glucose oxidation and whether either, or both, are required to meet the metabolic demands imposed by aggressive tumor growth.