The world-wide incidence of malignant brain tumors may be increasing in both children and the elderly 1234. Regardless of these ominous findings, the standard therapies for malignant gliomas (surgical resection and radiation) are basically the same today as they have been for over five decades 567. While these therapies may retard glioma growth over the short term, they can facilitate glioma recurrence and enhance growth rate over the longer term through alterations in morphogenetic fields 89. Chemotherapy has little long-term benefit on most malignant gliomas and is often associated with adverse effects that diminish the length or quality of life 710. The therapeutic targeting of brain tumor-associated mutations may also be problematic as most tumor mutations arise as epiphenomena of tissue disorganization and their relationship to causality is uncertain 81112. Despite modest gains in survival with temozolomide chemotherapy, few things are more certain in the brain tumor field than the impotence of most current therapies 2713. Hence, new approaches are needed that can provide long-term management of malignant brain tumors while permitting a decent quality of life.

Metabolic control analysis evaluates the degree of flux in metabolic pathways and can be used to analyze and treat complex diseases 141516. The approach is based on findings that compensatory genetic and biochemical pathways regulate the bioenergetic potential of cells and ultimately the phenotype 141517. As rate-controlling enzymatic steps in biochemical pathways are dependent on the metabolic environment of the physiological system, the management of disease phenotype depends more on the flux of the entire system than on the expression of any specific gene or enzyme alone 16171819. In other words, complex disease phenotypes can be managed through self-organizing networks that display system wide dynamics involving glycolysis and respiration. Global manipulations of these metabolic networks can restore orderly adaptive behavior to widely disordered states involving complex gene-environmental interactions 15172021.

As abnormal energy metabolism and biological chaos are characteristics of brain tumors 8222324, the general principles of metabolic control analysis can be effective for brain cancer management. This hypothesis is based on the known differences in energy metabolism between normal and neoplastic brain cells. As long as brain tumors are provided a physiological environment conducive for their glycolytic energy needs, they will survive; when this environment is restricted or abruptly changed, they will either growth arrest or perish. Here we describe how new therapeutic approaches, which lower circulating glucose and elevate ketone bodies (acetoacetate and β-hydroxybutyrate), target brain tumors while enhancing the metabolic efficiency of normal neurons and glia.

To manage brain cancer through metabolic targeting it is necessary to consider energy metabolism in the normal orthotopic tissue. Figure 1 illustrates some of the metabolic pathways discussed here. Under normal physiological conditions, the mature brain derives almost all of its energy from the aerobic oxidation of glucose 152526. The glucose transporter, GLUT-1, is enriched in the brain capillary endothelial cells and mediates the facilitated diffusion of glucose through the blood brain barrier. Most of the glucose is metabolized to pyruvate, which enters the mitochondria of neurons and glia and is converted to acetyl-CoA before entering the TCA cycle. Only about 13% of glycolytic pyruvate is converted to lactate under normal conditions 26. Fatty acids are attached to lipoproteins and do not pass the blood brain barrier as fuel substrates, though octanoate may be an exception 262728. It is also unlikely that lactate is used for energy metabolism in adult brain, but this remains somewhat controversial 26293031. Hence glucose is the primary, if not exclusive, brain metabolic fuel under normal physiological conditions.

While glucose is the preferred energy substrate, neurons and glia will metabolize ketones for energy under fasting-induced reductions of blood glucose. This is a conserved physiological adaptation to prolonged food restriction and evolved to enhance survival and maintain adequate brain function while sparing proteins 1527323334. Ketone bodies, consisting of acetoacetate, and β-hydroxybutyrate (β-OHB) are derived from fat catabolism in the liver and their concentration in blood is inversely related to that of glucose 21353637. Ketone bodies are transported into the brain through the blood-brain barrier monocarboxylic transporters (MCTs), whose expression is regulated in part by circulating ketone and glucose levels 30313438394041. β-OHB is the predominant blood ketone body and is rapidly oxidized to acetyl-CoA in the mitochondria through an enzymatic series involving 3-hydroxybutyrate dehydrogenase, SCOT (succinyl-CoA-acetoacetate-CoA transferase), and mitochondrial acetyl-CoA thiolase 33344243. Acetone is a non-enzymatic byproduct of ketone body synthesis and is largely excreted in the urine or exhaled from the lungs 3444.

Although the levels of glucose and ketones in brain are proportional to their levels in blood 3445, the adult brain does not usually metabolize ketone bodies for energy unless blood glucose levels are reduced 36. Therapeutic efficacy of the ketogenic diet is best when coupled to dietary energy restriction under which conditions circulating glucose levels are gradually reduced in conjunction with elevations of ketone bodies 1521. The emphasis here is on the term "gradual", as ketone bodies cannot be used for energy following acute hypoglycemia 46. The situation is different, however, for in vitro preparations where neuronal glial interactions are disrupted and the blood brain barrier is absent 4748. The gradual transition from glucose to ketone bodies for energy in vivo requires a flexible genome for the coordinated integration of multiple metabolic networks according to principles of metabolic control analysis 1536.

Ketone bodies are more energetically efficient than either pyruvate or fatty acids because they are more reduced (greater hydrogen/carbon ratio) than pyruvate and do not uncouple the mitochondrial proton gradient as occurs with fatty acid metabolism 14. In contrast to glucose, ketone bodies by-pass cytoplasmic glycolysis and directly enter the mitochondria where they are oxidized to acetyl-CoA 4449. The amount of acetyl-CoA formed from ketone body metabolism is also greater than that formed from glucose metabolism 50. This increases TCA cycle metabolites (from citrate to α-ketoglutarate) while reducing the mitochondrial NAD couple, [NAD⁺]/[NADH], and increasing the mitochondrial Q couple [Q]/[QH₂] 1450. The difference between these couples increases the redox span between the NADH dehydrogenase complex (site I), and the CoQH₂-cytochrome C reductase (site III) thus enhancing the mitochondrial proton gradient 14. This enhances the energy available from the hydrolysis of ATP, ΔG'ATP, the cell's key energy reserve generated through the mitochondrial Fl ATPase 1416. Remarkably, the ketone body-induced increase in the ΔG'ATP is also accomplished using less oxygen 4850. These and other findings led Veech to designate ketone bodies as a "super fuel" 14.

In addition to increasing ATP production while reducing oxygen consumption, ketone body metabolism can also reduce production of damaging free radicals 141648. The semiquinone of Q, the half reduced form, spontaneously reacts with oxygen and is the major source of mitochondrial free radical generation 1451. Oxidation of the Q couple reduces the amount of the semiquinone form thus decreasing superoxide production 14. Since the cytosolic free NADP⁺/NADPH concentration couple is in near equilibrium with the glutathione couple, ketone body metabolism will increase the reduced form of glutathione thus facilitating destruction of hydrogen peroxide 14. The reduction of free radicals through ketone body metabolism will also reduce tissue inflammation provoked by reactive oxygen species. Thus, ketone bodies are not only a more efficient metabolic fuel than glucose, but also possess anti-inflammatory potential.

In contrast to normal brain that oxidizes glucose as well as ketone bodies for energy, malignant brain tumors from either humans or animal models lack metabolic flexibility and are largely dependent on glucose for energy 212352535455565758. Enhanced glycolysis produces excess lactic acid that can return to the tumor as glucose through the Cori cycle 59 (Figure 1). Although some neural tumors metabolize ketone bodies, this metabolism is largely for lipid synthesis rather than for energy production 6061. Many tumors also have reduced activity of SCOT, the rate-controlling step for utilizing β-OHB as a respiratory fuel 426263. Consistent with these observations, we recently found that β-OHB could rescue normal mouse astrocytes under low glucose conditions, but could not rescue mouse astrocytoma cells 20. Although glutamine may provide energy to some non-neural tumors, glutamine stimulates glycolysis in C6 rat glioma cells and may not serve as a direct respiratory fuel 64. Hence brain tumors, like most malignant tumors, depend heavily on glucose and glycolysis for their metabolic energy and are generally unable to metabolize β-OHB for energy.

In addition to glycolytic dependence, most tumors including brain tumors, express abnormalities in the number and function of their mitochondria 6566. Such abnormalities would prevent the bioenergetic utilization of ketone bodies, which require functional mitochondria for their oxidation 47. Warburg originally emphasized that the high glycolytic rate of tumors resulted from diminished or disturbed respiration 6768. While most cells die from damaged respiration, those cells that can enhance and modify their anaerobic glycolysis in response to respiratory damage will survive and form tumors 6768. Later studies in a variety of neural and non-neural tumor systems showed that these respiratory disturbances involve abnormalities in TCA cycle components, alterations in electron transport, and deficiencies in oxidative phosphorylation 2355697071. Mitochondrial DNA mutations, however, may not be involved 12. Structural defects of the inner mitochondrial membrane, that would reduce or dissipate the proton motive gradient, could also prevent normal ATP production despite the appearance of oxidative metabolism, i.e., oxygen consumption and CO₂ production 707273. Considered together, these findings indicate that brain tumors suffer from reduced respiratory capacity coupled to an increased glycolysis and lactic acid production.

Although persistent aerobic glycolysis (glycolysis in the presence of oxygen) or defects in the Pasteur effect (reduction of glycolysis in the presence of oxygen) are characteristics of many tumors, Warburg considered these phenomena too labile or too dependent on environmental conditions to be reliable indicators of tumor metabolism 68. Rather, he emphasized the importance of defects in the coordination of glycolysis with respiration. The latency between tumor initiation and progression was considered the period necessary to disconnect respiration from glycolysis 68. Recent studies in human glioblastoma cells suggest that this disconnect involves activation of the Akt oncogene rendering cancer cells dependent on aerobic glycolysis for continued growth and survival 74. This is consistent with the Warburg hypothesis that the increased glycolysis of tumor cells occurs gradually in order to compensate for respiratory failure 68. In contrast to normal brain cells, in which glycolysis and respiration are tightly coupled, tumor cells are defective in their ability to integrate energy metabolism between glycolysis and respiration 71. It is these defects that will render brain tumor cells vulnerable to metabolic targeting through metabolic control analysis. Support for this possibility comes from studies with the ketogenic diet and dietary energy restriction.

Based on the differences in energy metabolism between normal brain cells and brain tumor cells, we propose a radically different approach to brain cancer management that combines metabolic control analysis with the evolutionarily conserved capacity of normal cells to survive extreme shifts in physiological environment. The adaptation to environmental extremes is conserved within the genome according to the ecological instability theory of Potts 96. Consequently, only those cells with a flexible genome will be capable of surviving abrupt changes in metabolic landscape. Cells with genomic defects, which would limit flexibility, should be less adaptable to metabolic stress and therefore vulnerable to elimination through principles of metabolic control analysis. This strategy focuses more on the genetic capabilities of normal cells than on the genetic defects of tumor cells.

As a metabolic disorder involving the dysregulation of glycolysis and respiration, brain cancer is potentially manageable through abrupt changes in metabolic environment. Significant cellular energy is used to maintain the activity of transmembrane ion pumps (the Na⁺, K⁺-ATPase and Ca²⁺ and Mg²⁺ ATPases) 299798 (Figure 1). The amount of energy needed to maintain pump function is also greater than that needed for mitosis, which is largely dependent on glycolytic energy 7097. Despite differences in membrane potential, most cells have a constant ΔG' of about -57 kJ/mol 14. According to Veech this is "the still point in the turning world" and if cells cannot maintain this useable ATP they lose potassium, gain sodium and calcium, swell, and eventually die 16. In other words, regardless of whether the cell is a normal neuron, a glial cell, or even a transformed tumor cell, its survival depends on maintaining an adequate ΔG' of ATP hydrolysis. Tumor cells with limited genomic flexibility should therefore be less capable than normal cells in utilizing alternative energy substrates to maintain their ΔG' of ATP hydrolysis.

The energy used to maintain pump function and cell viability in normal brain cells comes from either glycolysis or aerobic respiration 2999. In the case of C6 glioma cells and most brain tumors for that mater, this energy is mostly derived from glycolysis 2999. This would then render brain tumor cells vulnerable to reductions in circulating glucose levels as these mutant cells would have difficulty oxidizing alternative fuels (ketone bodies) through respiration. While some brain tumor cells may survive through up-regulation of their glucose transporters 100101, most will either perish or reduce their growth. Direct support for this hypothesis comes from our recent findings in experimental mouse brain tumor models and from those of Nebeling and co-workers in human pediatric astrocytoma 20377581.

The widely held notion that brain tumor cells are somehow hardy or tough and resistant to death (programmed or nonprogrammed) may be a misconception. How can brain tumor cells, or any tumor cell for that matter, that have multiple types and kinds of genetic mutations be more fit and hardy than normal cells that possess a flexible genome and can easily transition between glycolysis and respiration for energy maintenance? The notion that tumor cells are more versatile than nontumor cells is also illogical in the context of evolutionary biology. While knowledge of tumor-associated mutations and genomic instability is of considerable academic interest, this information has produced no new or effective clinical therapies for brain tumors. Regardless of when or how genomic defects become involved in the initiation or progression of brain tumors, these defects can be exploited for the metabolic destruction of the tumor.

Normal cells evolved to survive extremes in metabolic and physiological environment due largely to oscillatory changes in the physical environment 96. Moreover, the ability to adapt to extreme environmental stress is retained within the normal flexible genome 19102. It is this flexibility that allows normal brain cells to transition from glucose to β-OHB for energy under reduced energy availability. Due to accumulated nuclear genetic mutations and respiratory defects, brain tumor cells will be less adaptable than normal cells to abrupt changes in metabolic environment and can be either destroyed outright or isolated metabolically from normal cells. Hence, the genomic and metabolic flexibility of normal brain cells can be used to target indirectly the genetically defective and less metabolically flexible brain tumor cells.

According to this hypothesis, a novel strategy can be suggested for human brain tumor management that enhances the respiratory potential of normal brain cells while metabolically targeting the tumor cells. The approach would involve a sequential series of therapeutic steps and should be effective against any primary or secondary brain tumor regardless of cell of origin, anatomical location, or histological grade. Step one would lower circulating glucose levels and elevate circulating β-OHB levels through diet therapies or ketone body supplementation. Glucose ranges between 3.0–3.5 mM (55–65 mg/dl) and β-OHB ranges between 4–5 mM should be effective for tumor management in most patients. The conditions for these parameters have been described previously for children and adults and can be adjusted on a case-by-case basis 343575103. These values are also well within normal physiological ranges and will have antiangiogenic and proapoptotic effects causing metabolic isolation and significant tumor shrinkage that can be assessed from imaging analysis. Reduced glucose and elevated ketones could also antagonize tumor cachexia as previously mentioned 82.

Step two would involve surgical resection if necessary. Smaller brain tumors with reduced vascularity and clearly circumscribed boundaries should be easier to resect than larger brain tumors with poorly circumscribed boundaries and extensive vascularization 104. The diet therapy could also be adjusted following surgery to facilitate healing and to maintain metabolic pressure on any surviving tumor cells.

Finally, step three could involve the use of either conventional or novel targeted therapies. While these therapeutic approaches might have little if any long-term benefit on malignant brain tumor management if used initially, they could be highly effective following the step one strategy after the tumor cells are weakened and metabolically isolated from the physiologically strengthened normal brain cells. Moreover, glycolysis inhibitors that would adversely affect normal cells might also be more effective following the first two steps of the proposed therapeutic strategy 105106107108109. It is also possible that carefully executed weight cycling strategies could maintain metabolic pressure on surviving tumor cells and facilitate their eradication or growth retardation 110. In addition to global diet therapies, more specific amino acid restrictions could also be effective in eliminating surviving tumor cells 111.

The objective of this new brain tumor therapeutic approach is to consistently change the physiological and metabolic environment of the tumor and the host. Only those cells with a normal flexible genome, honed through millions of years of environmental forcing and variability selection, are expected to survive extreme shifts in metabolic environment. Indeed, extreme conditions of survival and fitness will test the limits of a cell population's persistence in any given location over time 96. While some brain tumor cells might survive under one restricted environment or another, it is unlikely that all tumor cells will survive all restricted environments. In other words, it is the theory of Potts applied with sustained pressure to the entire population of normal and neoplastic brain cells. We predict that this therapeutic approach will be more successful than current approaches because it is based on the principles of evolutionary biology and metabolic control analysis.

KD, ketogenic diet; SCOT, succinyl-CoA-acetoacetate-CoA transferase; β-OHB, β-hydroxybutyrate; TCA, tricarboxylic acid.