Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
REVIEWS WHY DO CANCERS HAVE HIGH AEROBIC GLYCOLYSIS? Robert A. Gatenby* and Robert J. Gillies ‡ Abstract | If carcinogenesis occurs by somatic evolution, then common components of the cancer phenotype result from active selection and must, therefore, confer a significant growth advantage. A near-universal property of primary and metastatic cancers is upregulation of glycolysis, resulting in increased glucose consumption, which can be observed with clinical tumour imaging. We propose that persistent metabolism of glucose to lactate even in aerobic conditions is an adaptation to intermittent hypoxia in pre-malignant lesions. However, upregulation of glycolysis leads to microenvironmental acidosis requiring evolution to phenotypes resistant to acid-induced cell toxicity. Subsequent cell populations with upregulated glycolysis and acid resistance have a powerful growth advantage, which promotes unconstrained proliferation and invasion. *Departments of Radiology and Applied Mathematics, University of Arizona, Tucson, Arizona 85721, USA. ‡ Departments of Radiology and Biochemistry and Molecular Biophysics, University of Arizona, Tucson, Arizona 85721, USA. Correspondence to R.A.G. e-mail: rgatenby@ radiology.arizona.edu doi:10.1038/nrc1478 NATURE REVIEWS | C ANCER The multistep process of carcinogenesis is often described as occuring by somatic evolution, because it seems formally analogous to Darwinian processes, wherein phenotypic properties are retained or lost depending on their contribution to individual fitness. According to this model, traits that are found in invasive cancers must arise as adaptive mechanisms to environmental proliferative constraints during carcinogenesis1. Conversely, the common appearance of a phenotypic property in cancer populations is presumptive evidence that it must confer a selective growth advantage. A curious, but common, property of invasive cancers is altered glucose metabolism. Glycolysis — literally lysis of glucose — first requires the conversion of glucose to pyruvate (FIG. 1) and then to the waste product lactic acid. In most mammalian cells, glycolysis is inhibited by the presence of oxygen, which allows mitochondria to oxidize pyruvate to CO2 and H2O. This inhibition is termed the ‘Pasteur effect’, after Louis Pasteur, who first demonstrated that glucose flux was reduced by the presence of oxygen2. This metabolic versatility of mammalian cells is essential for maintenance of energy production throughout a range of oxygen concentrations. Conversion of glucose to lactic acid in the presence of oxygen is known as aerobic glycolysis or the ‘Warburg effect’. Increased aerobic glycolysis is uniquely observed in cancers. This phenomenon was first reported by Warburg in the 1920s3, leading him to the hypothesis that cancer results from impaired mitochondrial metabolism. Although the ‘Warburg hypothesis’ has proven incorrect, the experimental observations of increased glycolysis in tumours even in the presence of oxygen have been repeatedly verified4. Following Warburg’s initial observation, interest in the metabolic property of cancers has varied over time. Intense investigation in the 1960s was followed by a steep decline concomitant with the widespread application of newer molecular techniques. The atmosphere of the day was summarized by Sidney Weinhouse, who said “Since our perspectives have broadened over the years, the burning issues of glycolysis and respiration in cancer now flicker only dimly”5. However, interest in tumour metabolism has been rekindled, mainly because of the widespread clinical application of the imaging technique positronemission tomography (PET) using the glucose analogue tracer 18fluorodeoxyglucose (FdG)6–8. FdG PET imaging of thousands of oncology patients has unequivocally shown that most primary and metastatic human cancers show significantly increased glucose uptake (FIG. 2). For many cancers, the specificity and sensitivity of FdG PET to identify primary and metastatic lesions is near 90%9. Sensitivity is lowered because FdG PET has difficulty resolving lesions less VOLUME 4 | NOVEMBER 2004 | 8 9 1 REVIEWS HEXOKINASES Summary Enzymes that catalyse the transfer of phosphate from ATP to glucose to form glucose-6phosphate. This is the first reaction in the metabolism of glucose and prevents efflux of glucose from the cell. HYPOXIA Refers to a low oxygen level. This means different levels to different investigators, but for radiation biologists hypoxia occurs at levels less than 0.1% oxygen in the gas phase. Normoxia refers to normal levels of oxygen (>10%) and anoxia refers to no oxygen. • Widespread clinical use of 18fluorodeoxyglucose positron-emission tomography has demonstrated that the glycolytic phenotype is observed in most human cancers. • The concept of carcinogenesis as a process that occurs by somatic evolution clearly implies that common traits of the malignant phenotype, such as upregulation of glycolysis, are the result of active selection processes and must confer a significant, identifiable growth advantage. • Constitutive upregulation of glycolysis is likely to be an adaptation to hypoxia that develops as pre-malignant lesions grow progressively further from their blood supply. At this stage, the blood supply remains physically separated from the growing cells by an intact basement membrane. • Increased acid production from upregulation of glycolysis results in microenvironmental acidosis and requires further adaptation through somatic evolution to phenotypes resistant to acid-induced toxicity. • Cell populations that emerge from this evolutionary sequence have a powerful growth advantage, as they alter their environment through increased glycolysis in a way that is toxic to other phenotypes, but harmless to themselves. The environmental acidosis also facilitates invasion through destruction of adjacent normal populations, degradation of the extracellular matrix and promotion of angiogenesis. • We propose that the glycolytic phenotype, by conferring a powerful growth advantage, is necessary for evolution of invasive human cancers. than 0.8 cm3, and specificity is lowered because other tissues, notably immune cells, also avidly trap FdG. When these limitations are accounted for, it can be reasonably surmised that virtually all invasive cancers avidly trap FdG. The increased glucose uptake imaged with FdG PET is largely dependent on the rate of glycolysis. FdG uptake and trapping occurs because of upregulation of glucose transporters (notably GLUT1 and GLUT3) and 10,11 HEXOKINASES I and II Although metabolic control over glycolytic rate can be applied at many steps in the glycolytic pathway12,13, most studies in cancer support the hypothesis that control over glycolytic flux primarily Blood vessel Glucose HbO2 Glucose O2 Anion exchanger HCO3– H+ Glucose transporter 36 ATP Lactate Mitochondrion Lactate Glucose Monocarboxylate H+ transporter 2 ATP Hexokinase Glucose-6phosphate Pyruvate H+ Sodium–hydrogen exchanger Figure 1 | Glucose metabolism in mammalian cells. Afferent blood delivers glucose and oxygen (on haemoglobin) to tissues, where it reaches cells by diffusion. Glucose is taken up by specific transporters, where it is converted first to glucose-6-phosphate by hexokinase and then to pyruvate, generating 2 ATP per glucose. In the presence of oxygen, pyruvate is oxidized to HCO3, generating 36 additional ATP per glucose. In the absence of oxygen, pyruvate is reduced to lactate, which is exported from the cell. Note that both processes produce hydrogen ions (H+), which cause acidification of the extracellular space. HbO2, oxygenated haemoglobin. 892 | NOVEMBER 2004 | VOLUME 4 resides at the transport and phosphorylation steps14–16. FdG PET imaging also allows quantitation of glucose uptake. These studies have consistently correlated poor prognosis and increased tumour aggressiveness with increased glucose uptake 17,18. In addition, hypoxic tumours, which require increased glycolysis to survive, are often19–22, but not always23, more invasive and metastatic than those with normal oxygen levels. These results demonstrate the clinical importance of glucose metabolism and have moved the glycolytic phenotype from a laboratory oddity to the mainstream of clinical oncology. Cells derived from tumours typically maintain their metabolic phenotypes in culture under normoxic conditions, indicating that aerobic glycolysis is constitutively upregulated through stable genetic or epigenetic changes. Consistent with the FdG PET results, the glycolytic rate in cultured cell lines seems to correlate with tumour aggressiveness. For example, non-invasive MCF-7 breast cancer cells have much lower aerobic glucose consumption rates compared with the highly invasive MDA-mb-231 breast cancer cell line (FIG. 3). These observations indicate that altered metabolism of glucose by tumours is more than a simple adaptation to HYPOXIA. We suggest that the nearuniversal observation of aerobic glycolysis in invasive human cancers, its persistence even under normoxic conditions and its correlation with tumour aggressiveness indicate that the glycolytic phenotype confers a significant proliferative advantage during somatic evolution of cancer and must, therefore, be a crucial component of the malignant phenotype. At first glance, this hypothesis seems at odds with an evolutionary model of carcinogenesis, because the proliferative advantage of the glycolytic phenotype is not immediately apparent. First, anaerobic metabolism of glucose is inefficient — it produces only 2 ATP per glucose, whereas complete oxidation produces 38 ATP per glucose (FIG. 1). Second, the metabolic products of glycolysis, such as hydrogen ions (H+), cause a spatially heterogeneous but consistent acidification of www.nature.com/reviews/cancer REVIEWS the extracellular space, which might result in cellular toxicity24–26 (FIG. 4a,b). Intuitively, it would seem that the Darwinian forces prevailing during the somatic evolution of invasive cancers would select against a metabolic phenotype that is more than an order of magnitude less efficient than its competitors and that is environmentally poisonous. In other words, the accepted tenet of ‘survival of the fittest’ would seem to generally favour populations with more efficient and sophisticated substrate metabolism. So, why do tumour populations consistently evolve to the inefficient and potentially toxic glycolytic phenotype? We propose that the remarkable prevalence of upregulated glycolysis in clinical cancers is neither random nor accidental. Rather, it represents an evolved solution to common environmental growth constraints during carcinogenesis, and its persistence in primary and metastatic malignancy indicates that it continues to confer a proliferative advantage even to fully transformed cells. So, we suggest that increased glycolysis is an essential component of the malignant phenotype and, therefore, a hallmark of invasive cancers. Herein we explore its causes and consequences. The microenvironment in pre-malignant lesions WINDOW CHAMBER A metal chamber with a glass window that is placed on the dorsal skin of an animal. This allows in vivo tumour growth to be continuously observed microscopically. NATURE REVIEWS | C ANCER Figure 2 | Positron-emission tomography imaging with 18 fluorodeoxyglucose of a patient with lymphoma. The mediastinal nodes (purple arrow) and supraclavicular nodes (green arrows) show high uptake of 18fluorodeoxyglucose (FdG), showing that tumours in these nodes have high levels of FdG uptake. The bladder (yellow arrow) also has high activity, because of excretion of the radionuclide. 60 Normoxia Hypoxia 50 Glucose consumption rate (nmol min–1 mg protein–1) Although pre-malignant lesions are often characterized as highly vascularized, this is true only in a macroscopic sense. That is, although a pre-malignant lesion such as a polyp or carcinoma in situ might have a vascular stroma, the hyperplastic epithelia are physically separated from their blood supply by a basement membrane. This is illustrated in FIG. 5, as the hyperplastic epithelium of a carcinoma in situ is clearly delimited from the stroma by a thin basement membrane. Blood vessels are confined to the stromal compartment and, therefore, early carcinogenesis and development of the malignant phenotype actually occur in an avascular environment. As a result, substrates, such as oxygen and glucose, must diffuse from the vessels across the basement membrane and through layers of tumour cells, where they are metabolized. This process of diffusion and consumption was modelled by Krogh as early as 1919 through reaction–diffusion equations that showed that oxygen concentrations decreased with distance from a capillary such that oxygenated cells were limited to a distance of less than 150 µm from a blood vessel27. In the 1950s, empirical studies by Thomlinson and Gray showed that viable tumour cells were not observed at distances greater than 160 µm from blood vessels, consistent with Krogh’s calculations28. Subsequent experimental studies in WINDOW CHAMBERS in animal models have demonstrated that near-zero partial pressures of oxygen (pO2) are observed at distances of only 100 µm from a vessel29,30. Therefore, pre-malignant lesions, provided their basement membranes remain intact, will inevitably develop hypoxic regions near the oxygen diffusion limit, as persistent proliferation leads to a thickening of the epithelial layer, pushing cells ever more distant from their blood supply, which remains on the other side of the basement membrane. At this penumbral layer, microenvironmental selection forces will favour 40 30 20 10 W P 0 MCF-7 MDA-MB-231 Figure 3 | Pasteur and Warburg effects in non-invasive and metastatic breast cancer cell lines. In both cell lines, glucose consumption is reduced in the presence of oxygen — the Pasteur effect (P). However, the more aggressive cell line, MDA-MB-231, has much higher glucose consumption in the presence of oxygen than the MCF-7 cells with a non-invasive phenotype — the Warburg effect (W). This is consistent with positron-emission tomography scans with 18 fluorodeoxyglucose, which show that higher glucose uptake correlates with more aggressive phenotypes and poorer clinical outcomes. VOLUME 4 | NOVEMBER 2004 | 8 9 3 REVIEWS 7.4 14 pH pO2 7.2 12 10 pH 8 7.0 6 4 6.8 pO2 mm Hg a 2 0 6.6 0 100 200 300 400 Distance (mm) 7.0 Extracellular pH b MDA-MB-435 6.4 Figure 4 | Hyperacidity of tumours. These figures illustrate the micro- and macro-heterogeneity of pH. a | Tumour interstitial pH and partial pressure of oxygen (pO2) are shown with distance from a vessel wall. These were measured in vivo in MCF-7 breast cancer cells using fluorescent ratio imaging. b | The extracellular pH of a MDA-MB-435 breast tumour in mice was imaged with the pH indicator IEPA and measured by 1H magneticresonance spectroscopy. Part a reproduced with permission from REF. 30 © (1997) Nature Publishing Group. Part b reproduced with permission from REF. 26 © (2002) Wiley. phenotypes that adapt to harsh environments (through resistance to hypoxia and acid-induced cell toxicity) and successfully compete for scarce resources, such as oxygen and glucose31,32. Emergence of the glycolytic phenotype HAEMATOCRIT A measure of the concentration of red cells in the blood. A reduced haematocrit decreases the oxygen-carrying capacity of the blood. VASOMOTION Rhythmic oscillations in vascular tone caused by local changes in smooth muscle. VASCULAR REMODELLING The active process of altering structure and arrangement in blood vessels through cell growth, cell death, cell migration and production or degradation of the extracellular matrix. 894 Evolutionary game theory is a mathematical approach that analyses strategy dynamics in adaptation to environmental growth — winners in this game proliferate, whereas losers become extinct. Recently, this method has been applied to somatic evolution of the malignant phenotype33. This analysis showed that proliferation of normal cells is controlled by their interactions with other cells and the extracellular matrix (ECM), and by the levels of growth factors. Importantly, cell proliferation and survival in normal tissue is not constrained by substrate availability, except under pathological conditions such as acute vascular occlusion (for example, caused by strokes and myocardial infarcts) or chronic occlusion (as seen in diabetic ulcers). It follows, therefore, that the earliest steps in carcinogenesis require alterations in cellular sensitivity to these normal tissue constraints. So, proliferation will follow genetic alterations that reduce sensitivity to growth constraints generated by other cells, the ECM | NOVEMBER 2004 | VOLUME 4 and/or growth factors33. For example, in many tissues, pre-malignant lesions are initiated by mutations in HRAS or KRAS genes, which alter cellular responses to growth factors34. From this, it follows that mutations affecting substrate use cannot be early events in carcinogenesis because they would not confer a selective growth advantage in an environment in which proliferation is not limited by substrate availability. The evolutionary models show, however, that clonal expansion of pre-malignant tumour populations is eventually limited by substrate availability33, as cell proliferation, unconstrained by normal tissue interactions, carries the population increasingly far from its blood supply (see above). In FIG. 5, note the distances between blood vessels and the necrotic zone of late-stage carcinoma in situ. Low oxygen concentrations seem to be the first substrate limitation confronting neoplastic cell populations, as reaction–diffusion models and empirical studies have shown that pO2 decline more rapidly with distance from blood vessels than do glucose levels25,30,35,36. Although the presence of hypoxia in pre-malignant in situ lesions has not been measured directly, it can be inferred from the frequent observation of necrosis in these lesions and by demonstration of hypoxia-inducible enzymes such as carbonic anhydrases IX and XII in latestage ductal carcinoma in situ, particularly adjacent to areas of necrosis37. We suggest that hypoxia in the penumbral region of pre-malignant tumours produces an adaptive landscape that favours a switch to anaerobic metabolism, which allows maintenance of metabolic activities in the absence of oxygen. A key factor in this adaptive landscape seems to be the exposure of cells near the oxygen diffusion limit to an unstable environment due to fluctuations in the haemodynamics of distant blood vessels. Oxic–hypoxic cycles in tumours have been measured to occur with periodicities of minutes38, hours39 or days40. For instance, Gallez’s group has recently imaged tumour xenografts using a magnetic-resonance imaging (MRI) technique that is sensitive to oxygenation status41. Analyses showed that fluctuations in signal intensity (oxygenation) occurred with discrete periodicities of 1 and 20 cycles per hour. By contrast, Dewhirst and colleagues used microelectrodes to show periodicities of about 1–2 cycles per minute42. However, it should be noted that MRI, although imaging the whole tumour, is insensitive to rapid fluctuations, and microelectrode instabilities render these electrodes insensitive to slower changes. Nonetheless, all of these studies show that oxygen delivery to tumours is inconsistent. These temporal cycles are probably due to a range of physiological mechanisms. Relatively rapid oxic–anoxic cycles can occur because of fluctuations in HAEMATOCRIT43 and VASOMOTION44. Variations occurring over days probably involve VASCULAR 45,46 REMODELLING or cycles of neoangiogenesis and regression due to hypoxia-induced expression of secreted vascular endothelial growth factor (VEGF), which is an induction and survival factor for new blood vessels40. From a bioenergetic standpoint, periodic hypoxia will select for cells in which anaerobic glucose metabolism is constitutively upregulated, as they would better survive www.nature.com/reviews/cancer REVIEWS 0.16 mm N T B S Figure 5 | Late-stage ductal carcinoma in situ. A 5µm-thick biopsy sample was stained with haematoxylin and eosin, and digitized with the DMetrix camera system (see online links box) with a resolution of 0.45 µm/pixel. Blood vessels (blue) are seen in the stroma (S) surrounding the tumour (T), but the tumour itself — within the ducts and surrounded by the basement membrane (B) — is avascular. The centre of the tumour is necrotic (N). VMAX and KM Terms from the Michaelis–Menten model. Applied to transport, Vmax is the maximum possible rate of uptake of a specific substrate. Km is the substrate concentration at which the substrate uptake is half of Vmax. Cell populations with low Km are better adapted to maintaining substrate uptake in conditions in which substrate concentrations are low. NATURE REVIEWS | C ANCER the anoxic episodes. The mechanisms underlying this upregulation are discussed in the next section. Although the upregulation of glycolysis is a successful adaptation to hypoxia/anoxia, it also has significant negative consequences because of increased acid production, which causes significant decreases in local extracellular pH. Prolonged exposure of normal cells to an acidic microenvironment typically results in necrosis or apoptosis through p53- and caspase-3-dependent mechanisms47,48. The physiological trigger for apoptosis might be collapse of the transmembrane H+ gradient that occurs with intracellular acidosis, but other factors might have a role49. So, constitutive upregulation of glycolysis requires additional adaptation to the negative effects of extracellular acidosis through resistance to apoptosis or upregulation of membrane transporters to maintain normal intracellular pH. Intracellular pH is maintained by multiple families of H+ transporters, which are co-expressed and redundant50,51. Na+–H+ exchange51,52 and vacuolar H+-ATPases53 have both been observed to be upregulated in cancers, and vacuolar H+-ATPase might confer resistance to apoptosis54. Additional adaptations might also be required as the increased glucose consumption rates further decrease glucose concentrations. Cellular competition for this increasingly limited resource will therefore increase and favour phenotypes with greater numbers of either high V (for example, GLUT1) or low K (for example, GLUT3) glucose transporters. Such upregulation of glucose transporters has been observed during carcinogenesis in oesophageal, gastric, breast and colon cancers55–57. In summary, we suggest that the glycolytic phenotype initially arises as an adaptation to local hypoxia (FIG. 6). Persistent or cyclical hypoxia subsequently exerts selection pressures that lead to constitutive upregulation of glycolysis, even in the presence of MAX M oxygen. This constitutive upregulation might occur through mutations or epigenetic changes such as alteration in the methylation patterns of promoters. The consequences of increased glycolysis require further adaptation to environments with high acid and low glucose concentrations. We propose that this is a crucial evolutionary sequence in the development of invasive cancer. First, it results in a phenotype with a powerful proliferative advantage, in that, through persistent aerobic glycolysis, it is able to alter the local microenvironment in a way that is harmless to itself, but fatal to competing populations. Second, acidification of the microenvironment facilitates tumour invasion both through destruction of adjacent normal populations and through acid-induced degradation of the ECM and promotion of angiogenesis. The underlying molecular, cellular and environmental dynamics are discussed next. Molecular mechanisms The molecular mechanisms leading to constitutive upregulation of aerobic glycolysis are not well defined. As mentioned above, it is commonly assumed that glucose transporters and hexokinases are the key molecules regulating glycolytic flux. It must be noted that a corollary of the current hypothesis is that the selective advantage conferred by the glycolytic phenotype is insensitive to the exact mechanism of glycolytic induction. A key regulator of the glycolytic response is the transcription factor hypoxia-inducible factor-1α (HIF1α)58. This factor mediates a pleiotropic response to hypoxic stress by inducing survival genes, including glucose transporters; angiogenic growth factors (for example, VEGF); hexokinase II59; and haematopoeitic factors (for example, transferrin and erythropoietin)60. In some systems, constitutively increased HIF1α levels are associated with constitutively high glucose consumption rates. This is the case in the renal-cell carcinoma cell line RCC4, which has constitutively high HIF1α because of a mutation in the von Hippel–Lindau (VHL) ubiquitin ligase. (The wild-type enzyme targets HIF1α for degradation.) Re-inserting VHL as a transgene in these cells restores normal HIF1α levels and greatly reduces aerobic glucose consumption rates61. Although HIF1α strongly links aerobic glycolysis to carcinogenesis62, it would be premature to conclude that the glycolytic phenotype in cancer is invariably due to dysregulation of the HIF system. Although it is termed the hypoxia-inducible factor, HIF1α levels can in fact be stabilized by a range of factors, including cyclooxygenase-2 activity, insulin-like growth factor 2, ERBB2, epidermal growth factor receptor, phosphatidylinositol 3-kinase, heat-shock protein 90, microtubule status, thioredoxin and histone deacetylase, to name a few 63–65. Additionally, stabilization of HIF1α in tumours can result from hypoxia-reoxygenation injury 66, which indicates that its constitutive upregulation might result from the periodic oxic–hypoxic cycles that occur in pre-malignant tumours. Consistent with our somatic-evolution model, lack of HIF1α decreases survival in response to hypoxia67, leading to selection of cells with upregulated HIF1α. VOLUME 4 | NOVEMBER 2004 | 8 9 5 REVIEWS Tumour stage Normal epithelium Physiological state Process Interstitial neoplasia Initiation Intermittant hypoxia Proliferation Carcinoma in situ HIF1α stabilization Selection Invasive carcinoma Glycolytic phenotype Induction Acidosis Metabolism Motility Metastatic disease Degradation of basement membrane and vascularization Selection VEGF Model Glucose diffusion limit O2 diffusion limit Basement membrane Blood vessel Stroma Figure 6 | Model for cell–environment interactions in carcinogenesis. Early carcinogenesis proceeds from normal tissues through initiation to a hyperplastic state to interstitial neoplasia, progressing to carcinoma in situ. Until this stage, epithelial cancers are avascular, as shown by histopathology (FIG. 5). Following breakdown of the basement membrane, cells gain access to existing and newly formed blood and lymphatic vascular routes for metastasis. The stages of tumour growth and their associated physiological states are diagrammed, showing that progression from one stage to the next is governed by state processes. Normal epithelial cells (grey) become hyperproliferative (pink) following induction. As they reach the oxygen diffusion limit, they become hypoxic (blue), which can either lead to cell death (apoptotic cells shown with blebbing) or adaptation of a glycolytic phenotype (green), which allows cells to survive. As a consequence of glycolysis, lesions become acidotic, which selects for motile cells (yellow) that eventually breach the basement membrane. As cancer progression proceeds, the mutations in cells increase (nuclei shown as light orange for one mutation and darker oranges for more mutations). HIF1α, hypoxia-inducible factor-1α; VEGF, vascular endothelial growth factor. Multiple cellular pathways might lead to the glycolytic phenotype, so that altered glucose metabolism might even result in cells with normal HIF levels. For example, upregulation of glycolytic enzymes can be coordinated in response to oxidation–reduction changes by the Sp1 transcription-factor complex68. GLUT1 can be upregulated directly by MYC13,69 or indirectly by KRAS70. Interestingly, in this latter study, KRAS activation was only associated with a subset of GLUT1-positive colon cancers, indicating that it is one of several mechanisms to activate glycolysis in this system. RAS activation of GLUT1 transcription seems to be mediated through HIF1α transactivation71. Hexokinase II can be transcriptionally activated by mutant p53 (REF. 72) or through demethylation of its promoter73. It is also intriguing to note that transfection of fibroblasts with H+-ATPase or Na+–H+ exchange raises the intracellular pH, makes them tumorigenic and leads to marked increases in glycolysis74,75. These alternative systems for upregulating glycolysis are consistent with our basic proposal that the mechanism of induction is not as important as the induction itself. That is, the glycolytic phenotype is not a secondary phenomenon that results from induction of some other pathway during carcinogenesis. Rather, it is directly selected because it provides a growth advantage and acquisition of the glycolytic phenotype might be 896 | NOVEMBER 2004 | VOLUME 4 achieved through multiple mechanisms, including oncogene activation or stabilization of transcription factors such as HIF1α. Angiogenesis We suggest that the glycolytic phenotype evolves in a microenvironment that is avascular; that is, the evolving tumour cells remain physically separated from their blood supply by a basement membrane, as occurs in in situ tumours. This invokes the diffusion of substrates from the vascularized stroma to the proliferating tumour epithelium. Therefore, even though late-stage carcinoma in situ can be characterized as ‘angiogenic’, the tumour does not become vascularized until the basement membrane is breached by an invasive cell. In fact, there is emerging evidence that the ‘glycolytic switch’ occurs before the ‘angiogenic switch’; lactic acid has been observed in regions of invasive gliomas76,77 that lack vessel permeability, as shown by the absence of contrast enhancement with MRI78. We do not wish to indicate that angiogenesis does not have a role in this process. In fact, it is likely that angiogenic factors, such as VEGF, are produced by the tumour and that this will promote increased vascularity within the stroma (FIG. 6). However, these new vessels remain physically separated from the tumour cells by the basement membrane (see figure 2 www.nature.com/reviews/cancer REVIEWS CLASTOGENIC Describing any substance or processes that increases alterations in the structure of chromosomes. GAP JUNCTIONS Linked channels through contiguous cell membranes that interconnect the cytoplasm of adjacent cells and allow direct exchange of ions and small molecules. in REF. 79). This will not necessarily result in increased substrate delivery through diffusion, as substrate concentrations in the reaction–diffusion equation are unaltered. So, a hallmark of these early cancers is a failure of angiogenesis to relieve hypoxia because of the physical separation between vessels and the cells they feed. This can result in a futile cycle of hyperproliferating blood vessels. Once the basement membrane is breached, the tumour will become vascular both by co-opting the pre-existing vessels within the stroma and by promoting new vessel growth directly into the tumour mass. Acidosis and invasion Although the glycolytic phenotype seems to be the result of adaptation to environmental constraints in pre-malignant lesions, its persistence in primary and metastatic cancers even in conditions of normoxia indicates that it continues to provide a strong selective growth advantage following malignant progression. We suggest, in fact, that acquisition of the glycolytic phenotype is required for invasive tumour growth. A constitutive and persistent increase in glycolysis results in acute and chronic acidification of the local environment. Indeed, numerous studies have shown that the extracellular pH of human and animal tumours is consistently acidic and can reach pH values approaching 6.0 (REFS 80,81) (FIG. 4). We have demonstrated both mathematically and empirically that intratumoral acidosis results in diffusion of H+ ions along concentration gradients into peritumoral normal tissue32. Normal cells, which lack a mechanism to adapt to extracellular acidosis (such as a p53 mutation) are unable to survive under such conditions, whereas the tumour populations continue to proliferate. In addition, acidosis itself can be mutagenic and CLASTOGENIC82, possibly through inhibition of DNA repair (for a review, see REF. 80) and can lead to both inhibition of GAP-JUNCTION conductance and to spontaneous transformation of normal diploid fibroblasts83. The resulting phenotypic diversity enhances the evolutionary potential of the tumour population, which accelerates malignant progression and adaptation to therapeutic strategies34 (BOX 1). Finally, under some (but not all) conditions, low pH stimulates in vitro invasion84 and in vivo metastasis85. The mechanisms of such induction Box 1 | Consequences of hypoxia and acidosis As tumours evolve and become first hypoxic and then acidic, malignant progression is accelerated and resistance to therapeutic strategies occurs. For further information, see REF. 36. Hypoxia Acidosis Radioresistance Increased radioresistance Drug resistance Resistance to anthracyclines Metastasis and invasion Increased metastases Increased mutation rate Increased migration and invasion Gene expression induced by hypoxiainducible factor Mutagenesis/clastogenesis Apoptosis Apoptosis NATURE REVIEWS | C ANCER are not known, but might involve the metalloproteinases and/or cathepsins, which promote the degradation of the ECM and basement membranes86,87. Metastasis So far, we have focused on the role of upregulated glycolysis and resistance to extracellular acidosis in adaptation to conditions in early pre-malignant lesions and in the evolution of invasive primary cancers. However, we note that this phenotype might also be crucial in the maturation of metastases as well. Upregulated glycolysis, evidenced by increased intratumoral lactate concentrations, is associated with increased incidence of metastasis in cervical and head and neck cancers88,89. Furthermore, a correlation between GLUT1 expression levels and metalloproteinase expression has also been reported in metastatic cancers90. During the process of metastasis, migratory cells invade the stromal tissue and move to distant sites, lodging in pre-capillary arterioles and capillaries91,92. These cells probably also experience periodic hypoxic or anoxic episodes as they proliferate and occlude the intravascular space. Therefore, the end stage of the metastasis sequence will also favour cells that are glycolytic and resistant to hypoxia- or acid-induced apoptosis. Any selective advantage is important, as the success rate of metastasis is low. For example, in a typical lung-colonization assay, as many as 105 lung cancer cells are injected into mouse tail veins, but fewer than 100 cells generally survive to form colonies. Cells pre-treated with hypoxia for 24 hours are four times more likely to survive than their normoxic counterparts93. Although there are other possible interpretations of these data, we suggest that they support the hypothesis that the glycolytic phenotype contributes to the efficiency of metastasis by allowing cells to survive transient hypoxia. Summary and future directions In summary, we suggest that upregulation of glycolytic metabolic pathways in the vast majority of invasive cancers is the result of adaptation to consistent environmental pressures in pre-malignant lesions, when diffusion limitations result in gradients of hypoxia and acidosis. Cellular traits selected by these conditions include constitutive upregulation of glycolysis and resistance to acidinduced apoptosis. Mathematical models and empirical observation indicate that the advantages conferred by this combination of phenotypic traits are both sufficient and necessary to promote unconstrained tumour proliferation. Furthermore, both mathematical models and empirical evidence indicate that diffusion of acid from the tumour into peritumoral normal tissue provides a specific mechanism promoting tumour invasion32. The crucial importance of the glycolytic phenotype is emphasized by studies demonstrating that increased glucose uptake is observed to coincide with the transition from pre-malignant lesions to invasive cancer94,95. These evolutionary advantages explain the remarkable prevalence of the glycolytic phenotype in human cancers and the otherwise puzzling observation that malignant cells remain glycolytic even in the presence of normoxia. VOLUME 4 | NOVEMBER 2004 | 8 9 7 REVIEWS The molecular basis for evolution of the glycolytic phenotype has been clarified by recent advances in understanding the HIF system, but much additional work will be required to fully understand the complex pathways involved in hypoxic response, metabolic controls and adaptation to acidosis in cancer progression. Despite gaps in our knowledge, the glycolytic phenotype could be exploited for treatment at several levels. As this phenotype emerges early in carcinogenesis, it might represent a possible target in cancer prevention. At later stages, a more complete understanding of the molecular and physiological consequences might lead to targeted therapies. Finally, this model of carcinogenesis indicates new avenues of investigation. For example, what is the 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 898 Bernards, R. & Weinberg, R. A. A progression puzzle. Nature 418, 823 (2002). A compelling opinion piece, in which the authors convincingly argue that the molecular phenotypes of metastatic cancers arose early during carcinogenesis. Although somatic evolution is implied in this work, the environmental nature of the selection pressures are not discussed. Racker, E. History of the Pasteur effect and its pathobiology. Mol. Cell. Biochem. 5, 17–23 (1974). Warburg, O. Ueber den stoffwechsel der tumoren. (Constable, London, 1930). Semenza, G. L. et al. ’The metabolism of tumours’: 70 years later. Novartis Found. Symp. 240, 251–260 (2001). In this timely review, Semenza describes the relation between HIF1α and the regulation of glycolysis. Weinhouse, S. The Warburg hypothesis fifty years later. Z. Krebsforsch. Klin. Onkol. Cancer Res. Clin. Oncol. 87, 115–126 (1976). Hawkins, R. A. & Phelps, M. E. PET in clinical oncology. Cancer Metastasis Rev. 7, 119–142 (1988). Weber, W. A., Avril, N. & Schwaiger, M. Relevance of positron emission tomography (PET) in oncology. Strahlenther. Onkol. 175, 356–373 (1999). Gambhir, S. S. Molecular imaging of cancer with positron emission tomography. Nature Rev. Cancer 2, 683–693 (2002). A well written review on FdG PET imaging. Czernin, J. & Phelps, M. E. Positron emission tomography scanning: current and future applications. Annu. Rev. Med. 53, 89–112 (2002). A comprehensive review of extant literature. The authors convincingly document the very high sensitivity and specificity of FdG PET in diagnosing and staging diverse types of metastatic cancers. Bos, R. et al. Biologic correlates of 18fluorodeoxyglucose uptake in human breast cancer measured by positron emission tomography. J.Clin.Oncol. 20, 379–387 (2002). This well-conducted study quantitatively analysed the molecular phenotypes of tumours that had either high or low rates of FdG trapping. Burt, B. M. et al. Using positron emission tomography with [18F]FDG to predict tumor behavior in experimental colorectal cancer. Neoplasia (New York) 3, 189–195 (2001). Schilling, C. H., Schuster, S., Palsson, B. O. & Heinrich, R. Metabolic pathway analysis: basic concepts and scientific applications in the post-genomic era. Biotechnol. Prog. 15, 296–303 (1999). Dang, C. V., Lewis, B. C., Dolde, C., Dang, G. & Shim, H. Oncogenes in tumor metabolism, tumorigenesis, and apoptosis. J. Bioenerg. Biomembr. 29, 345–354 (1997). One of many papers in this issue of the Journal of Bioenergetics and Biomembranes that dealt with the molecular controls of glucose metabolism. In this review, primary data were presented to support the importance and molecular controls of the glucose transporter and its regulation by MYC. Rivenzon-Segal, D., Boldin-Adamsky, S., Seger, D., Seger, R. & Degani, H. Glycolysis and glucose transporter 1 as markers of response to hormonal therapy in breast cancer. Int. J. Cancer 107, 177–182 (2003). One of many papers that demonstrates the important role of the glucose transporter in regulating glycolytic flux. | NOVEMBER 2004 | VOLUME 4 relationship and timing between the ‘angiogenic switch’ and the ‘glycolytic switch’? Can pharmacological agents be developed to inhibit emergence of the glycolytic/acidic phenotype and, therefore, retard the progression in early lesions? Would alteration in systemic pH perturb tumour growth dynamics and confer relative resistance to tumour development? Finally, would local or systemic alteration of buffering capacity and balance of extracellular tumour pH reverse the aggressive tumour phenotype in the absence of any other change? On this last point, we have tantalizing evidence that mild renal failure — which is typically accompanied by systemic acidosis — is associated with improved prognosis in patients with metastatic renal cancer following nephrectomy 96. 15. Artemov, D., Bhujwalla, Z. M., Pilatus, U. & Glickson, J. D. Two-compartment model for determination of glycolytic rates of solid tumors by in vivo 13C NMR spectroscopy. NMR Biomed. 11, 395–404 (1998). 16. Mathupala, S. P., Rempel, A. & Pedersen, P. L. Aberrant glycolytic metabolism of cancer cells: a remarkable coordination of genetic, transcriptional, post-translational, and mutational events that lead to a critical role for type II hexokinase. J. Bioenerg. Biomembr. 29, 339–343 (1997). Provides a cogent argument for the role of hexokinase in regulating glycolytic flux and its regulation by oncogenes and subcellular localization. 17. Kunkel, M. et al. Overexpression of Glut-1 and increased glucose metabolism in tumors are associated with a poor prognosis in patients with oral squamous cell carcinoma. Cancer 97, 1015–1024 (2003). This careful study is one of many that document the diagnostic importance of GLUT1 and glycolysis in carcinomas. 18. Mochiki, E. et al. Evaluation of 18F-2-deoxy-2-fluoro-Dglucose positron emission tomography for gastric cancer. World J. Surg. 28, 247–253 (2004). 19. Postovit, L. M., Adams, M. A., Lash, G. E., Heaton, J. P. & Graham, C. H. Oxygen-mediated regulation of tumor cell invasiveness. Involvement of a nitric oxide signaling pathway. J. Biol. Chem. 277, 35730–35737 (2002). 20. He, X. et al. Hypoxia increases heparanase-dependent tumor cell invasion, which can be inhibited by antiheparanase antibodies. Cancer Res. 64, 3928–3933 (2004). 21. Buchler, P. et al. Hypoxia-inducible factor 1 regulates vascular endothelial growth factor expression in human pancreatic cancer. Pancreas 26, 56–64 (2003). 22. Postovit, L. M., Adams, M. A., Lash, G. E., Heaton, J. P. & Graham, C. H. Nitric oxide-mediated regulation of hypoxiainduced B16F10 melanoma metastasis. Int. J. Cancer 108, 47–53 (2004). 23. Krtolica, A. & Ludlow, J. W. Hypoxia arrests ovarian carcinoma cell cycle progression, but invasion is unaffected. Cancer Res. 56, 1168–1173 (1996). 24. Schornack, P. A. & Gillies, R. J. Contributions of cell metabolism and H+ diffusion to the acidic pH of tumors. Neoplasia (New York) 5, 135–145 (2003). Determined proton production rates in breast cancer lines with low and high metastatic capability, and related these to glycolytic rate. These rates were used in a reaction–diffusion model to predict steady-state tumour pH values. 25. Griffiths, J. R., McIntyre, D. J., Howe, F. A. & Stubbs, M. Why are cancers acidic? A carrier-mediated diffusion model for H+ transport in the interstitial fluid. Novartis Found. Symp. 240, 46–62 (2001). 26. Bhujwalla, Z. M. et al. Combined vascular and extracellular pH imaging of solid tumors. NMR Biomed. 15, 114–119 (2002). Used spectroscopic imaging to measure the spatial variations in tumour pH, and these were related to vascular perfusion measures in the same tumours. 27. Krogh, A. The number and distribution of capillaries in muscles with calculations of the oxygen pressure head necessary for supplying the tissue. J. Physiol. 52, 409–415 (1919). 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. Demonstrates the annulus of tissues that can be oxygenated by a single capillary. Thomlinson, R. H. & Gray, L. H. The histological structure of some human lung cancers and the possible implications for radiotherapy. Br. J. Cancer 9, 539–549 (1955). Documents that necrosis in tumours occurs at distances from blood vessels and that this was consistent with the oxygen diffusion distances. Dewhirst, M. W., Secomb, T. W., Ong, E. T., Hsu, R. & Gross, J. F. Determination of local oxygen consumption rates in tumors. Cancer Res. 54, 3333–3336 (1994). Helmlinger, G., Yuan, F., Dellian, M. & Jain, R. K. Interstitial pH and pO2 gradients in solid tumors in vivo: high-resolution measurements reveal a lack of correlation. Nature Med. 3, 177–182 (1997). Despite its title, this very well conducted study documents the correlation between pH and oxygenation as they decrease with distances from feeding capillaries. Graeber, T. G. et al. Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours. Nature 379, 88–91 (1996). Documents the somatic evolutionary pressure mediated by hypoxia. Gatenby, R. A. & Gawlinski, E. T. A reaction-diffusion model of cancer invasion. Cancer Res. 56, 5745–5753 (1996). Mathematical methods and empirical evidence were used to demonstrate the acid-induced tumourinvasion model for the first time. Gatenby, R. A. & Vincent, T. L. An evolutionary model of carcinogenesis. Cancer Res. 63, 6212–6220 (2003). The formal mathematical development of evolutionary game theory in carcinogenesis. Fearon, E. R. & Vogelstein, B. A genetic model for colorectal tumorigenesis. Cell 61, 759–767 (1990). Introduced the molecular genetic changes that occur during gastrointestinal carcinogenesis and discussed the concept of clonal outgrowth in this context. There was no discussion of environmental selection pressures. Chresand, T. J., Gillies, R. J. & Dale, B. E. Optimum fiber spacing in a hollow fiber bioreactor. Biotechnol. Bioeng. 32, 983–992 (1988). Secomb, T. W. et al. Theoretical simulation of oxygen transport to tumors by three-dimensional networks of microvessels. Adv. Exp. Med. Biol. 454, 629–634 (1998). Wykoff, C. C. et al. Expression of the hypoxia-inducible and tumor-associated carbonic anhydrases in ductal carcinoma in situ of the breast. Am. J. Pathol. 158, 1011–1019 (2001). This work shows, with histopathology, the expression of CA IX and CA XII in carcinoma in situ lesions. These two carbonic anhydrases are sensitive to hypoxia and these data are consistent with significant hypoxia in in situ lesions. Kimura, H. et al. Fluctuations in red cell flux in tumor microvessels can lead to transient hypoxia and reoxygenation in tumor parenchyma. Cancer Res. 56, 5522–5528 (1996). Hill, R. P., De Jaeger, K., Jang, A. & Cairns, R. pH, hypoxia and metastasis. Novartis Found. Symp. 240, 154–165 (2001). Gilead, A. & Neeman, M. Dynamic remodeling of the vascular bed precedes tumor growth: MLS ovarian www.nature.com/reviews/cancer REVIEWS 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. carcinoma spheroids implanted in nude mice. Neoplasia (New York) 1, 226–230 (1999). Baudelet, C. et al. Physiological noise in murine solid tumors using T2*-weighted gradient echo imaging: a marker for tumor acute hypoxia? Phys. Med. Biol. 49, 3389–3411 (2004). Braun, R. D., Lanzen, J. L. & Dewhirst, M. W. Fourier analysis of fluctuations of oxygen tension and blood flow in R3230Ac tumors and muscle in rats. Am. J. Physiol. 277, H551–H568 (1999). Dewhirst, M. W. et al. Microvascular studies on the origins of perfusion-limited hypoxia. Br. J. Cancer Suppl. 27, S247–S251 (1996). Sonveaux, P. et al. Endothelin-1 is a critical mediator of myogenic tone in tumor arterioles: implications for cancer treatment. Cancer Res. 64, 3209–3214 (2004). Patan, S. et al. Vascular morphogenesis and remodeling in a human tumor xenograft: blood vessel formation and growth after ovariectomy and tumor implantation. Circ. Res. 89, 732–739 (2001). Kiani, M. F., Pries, A. R., Hsu, L. L., Sarelius, I. H. & Cokelet, G. R. Fluctuations in microvascular blood flow parameters caused by hemodynamic mechanisms. Am. J. Physiol. 266, H1822–H1828 (1994). References 38–46 document the periodic nature of tumour oxygenation. Park, H. J., Lyons, J. C., Ohtsubo, T. & Song, C. W. Acidic environment causes apoptosis by increasing caspase activity. Br. J. Cancer 80, 1892–1897 (1999). Williams, A. C., Collard, T. J. & Paraskeva, C. An acidic environment leads to p53 dependent induction of apoptosis in human adenoma and carcinoma cell lines: implications for clonal selection during colorectal carcinogenesis. Oncogene 18, 3199–3204 (1999). Shrode, L. D., Tapper, H. & Grinstein, S. Role of intracellular pH in proliferation, transformation, and apoptosis. J. Bioenerg. Biomembr. 29, 393–399 (1997). References 47–49 deal with pH-induced apoptosis. Grinstein’s review concludes that cytoplasmic acidification is unlikely to be part of the apoptosis paradigm, but that externally lowered pH might promote apoptotic cell death. Lee, A. H. & Tannock, I. F. Heterogeneity of intracellular pH and of mechanisms that regulate intracellular pH in populations of cultured cells. Cancer Res. 58, 1901–1908 (1998). Ober, S. S. & Pardee, A. B. Intracellular pH is increased after transformation of Chinese hamster embryo fibroblasts. Proc. Natl Acad. Sci. USA 84, 2766–2770 (1987). McLean, L. A., Roscoe, J., Jorgensen, N. K., Gorin, F. A. & Cala, P. M. Malignant gliomas display altered pH regulation by NHE1 compared with nontransformed astrocytes. Am. J. Physiol. 278, C676–C688 (2000). Martinez-Zaguilan, R., Lynch, R. M., Martinez, G. M. & Gillies, R. J. Vacuolar type proton ATPases are functionally expressed in the plasma membranes of human tumor cells. Am. J. Physiol. 265, c1015–c1029 (1993). References 50–53 describe mechanisms of pH regulation that are documented to be upregulated in cancers. Gottlieb, R. A., Giesing, H. A., Zhu, J. Y., Engler, R. L. & Babior, B. M. Cell acidification in apoptosis: granulocyte colony-stimulating factor delays programmed cell death in neutrophils by up-regulating the vacuolar H+-ATPase. Proc. Natl Acad. Sci. USA 92, 5965–5968 (1995). Demonstrates that vacuolar H+-ATPase activity is anti-apoptotic. Younes, M., Ertan, A., Lechago, L. V., Somoano, J. & Lechago, J. Human erythrocyte glucose transporter (Glut1) is immunohistochemically detected as a late event during malignant progression in Barrett’s metaplasia. Cancer Epidemiol. Biomarkers Prev. 6, 303–305 (1997). Sakashita, M. et al. Glut1 expression in T1 and T2 stage colorectal carcinomas: its relationship to clinicopathological features. Eur. J. Cancer 37, 204–209 (2001). Grover-McKay, M., Walsh, S. A., Seftor, E. A., Thomas, P. A. & Hendrix, M. J. Role for glucose transporter 1 protein in human breast cancer. Pathol. Oncol. Res. 4, 115–120 (1998). Semenza, G. L. Hypoxia-inducible factor 1: master regulator of O2 homeostasis. Curr. Opin. Genet. Dev. 8, 588–594 (1998). Yasuda, S. et al. Hexokinase II and VEGF expression in liver tumors: correlation with hypoxia-inducible factor 1α and its significance. J. Hepatol. 40, 117–123 (2004). NATURE REVIEWS | C ANCER 60. Carmeliet, P. et al. Role of HIF-1α in hypoxia-mediated apoptosis, cell proliferation and tumour angiogenesis. Nature 394, 485–490 (1998). 61. Robey, I., Lien, A., Welsh, S., Baggett, B. & Gillies, R. J. HIF-1α and the glycolytic phenotype in tumors. Neoplasia (in the press). 62. Lu, H., Forbes, R. A. & Verma, A. Hypoxia-inducible factor 1 activation by aerobic glycolysis implicates the Warburg effect in carcinogenesis. J. Biol. Chem. 277, 23111–23115 (2002). 63. Semenza, G. Targeting HIF-1 for cancer therapy. Nature Rev. Cancer 3, 1–13 (2003). References 58–63 describe the role of HIF1α in regulating aerobic and anaerobic glycolysis. 64. Semenza, G. Signal transduction to hypoxia-inducible factor 1. Biochem. Pharmacol. 64, 993–998 (2002). 65. Welsh, S. J., Bellamy, W. T., Briehl, M. M. & Powis, G. The redox protein thioredoxin-1 (Trx-1) increases hypoxiainducible factor 1α protein expression: Trx-1 overexpression results in increased vascular endothelial growth factor production and enhanced tumor angiogenesis. Cancer Res. 62, 5089–5095 (2002). 66. Moeller, B. J., Cao, Y., Li, C. Y. & Dewhirst, M. W. Radiation activates HIF-1 to regulate vascular radiosensitivity in tumors: role of reoxygenation, free radicals, and stress granules. Cancer Cell 5, 429–441 (2004). 67. Seagroves, T. et al. Transcription Factor HIF-1 is a necessary mediator of the Pasteur effect in mammalian cells. Mol. Cell. Biol. 21, 3436–3444 (2001). 68. Brand, K. Aerobic glycolysis by proliferating cells: protection against oxidative stress at the expense of energy yield. J. Bioenerg. Biomembr. 29, 355–364 (1997). 69. Osthus, R. C. et al. Deregulation of glucose transporter 1 and glycolytic gene expression by c-Myc. J. Biol. Chem. 275, 21797–21800 (2000). 70. Noguchi, Y. et al. Expression of facilitative glucose transporter 1 mRNA in colon cancer was not regulated by k-ras. Cancer Letters 154, 137–142 (2000). 71. Chen, C., Pore, N., Behrooz, A., Ismail-Beigi, F. & Maity, A. Regulation of glut1 mRNA by hypoxia-inducible factor-1. Interaction between H-ras and hypoxia. J. Biol. Chem. 276, 9519–9525 (2001). 72. Mathupala, S. P., Heese, C. & Pedersen, P. L. Glucose catabolism in cancer cells. The type II hexokinase promoter contains functionally active response elements for the tumor suppressor p53. J. Biol. Chem. 272, 22776–22780 (1997). 73. Goel, A., Mathupala, S. P. & Pedersen, P. L. Glucose metabolism in cancer. Evidence that demethylation events play a role in activating type II hexokinase gene expression. J. Biol. Chem. 278, 15333–15340 (2003). 74. Gillies, R. J., Martinez-Zaguilan, R., Martinez, G. M., Serrano, R. & Perona, R. Tumorigenic 3T3 cells maintain an alkaline intracellular pH under physiological conditions. Proc. Natl Acad. Sci. USA 87, 7414–7418 (1990). 75. Reshkin, S. J. et al. Na/H exchanger-dependent intracellular alkalinization is an early event in malignant transformation and play an essential role in the development of subsequent transformation-associated phenotypes. FASEB J. 14, 2185–2197 (2000). 76. Li, X. et al. Relationship of MR-derived lactate, mobile lipids and relative blood volume for in vivo gliomas. Am. J. Neuroradiol. (in the press). Describes the observation of increased lactate in nonenhancing grade III gliomas, indicating that metabolic upregulation might precede angiogenesis. 77. Nelson, S. J. Multivoxel magnetic resonance spectroscopy of brain tumors. Mol. Cancer Ther. 2, 497–507 (2003). 78. Dafni, H., Landstrom, L., Schechter, B., Kohen, F. & Neeman, M. MRI and fluorescence microscopy of the acute vascular response to VEGF165: vasodilation, hyperpermeability and lymphatic uptake, followed by rapid inactivation of the growth factor. NMR Biomed. 15, 120–131 (2002). 79. Hanahan, D. & Folkman, J. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 86, 353–364 (1996). 80. Raghunand, N., Gatenby, R. A. & Gillies, R. J. Microenvironmental and cellular consequences of altered blood flow in tumors. Br. J. Radiol. 77, S11–S22 (2004). 81. Gillies, R. J., Raghunand, N., Karczmar, G. & Bhujwalla, Z. MR Imaging of the tumor microenvironment. J. Magn. Reson. Imaging 16, 430–450 (2002). A comprehensive review describing MRI of clinical and experimental tumours. 82. Morita, T., Nagaki, T., Fukuda, I. & Okumura, K. Clastogenicity of low pH to various cultured mammalian cells. Mutat. Res. 268, 297–305 (1992). 83. Ruch, R. J., Klaunig, J. E., Kerckaert, G. A. & LeBoeuf, R. A. Modification of gap junctional intercellular communication by changes in extracellular pH in syrian hamster embryo cells. Carcinogenesis 11, 909–913 (1990). 84. Martinez-Zaguilan, R. et al. Acidic pH enhances the invasive behavior of human melanoma cells. Clin. Exp. Metastasis 14, 176–186 (1996). 85. Schlappack, O. K., Zimmermann, A. & Hill, R. P. Glucose starvation and acidosis: effect on experimental metastasic potential, DNA content and MTX resistance of murine tumour cells. Br. J. Cancer 64, 663–670 (1991). 86. Rozhin, J., Sameni, M., Ziegler, G. & Sloane, B. F. Pericellular pH affects distribution and secretion of cathepsin B in malignant cells. Cancer Res. 54, 6517–6525 (1994). 87. Montcourrier, P., Silver, I., Farnoud, R., Bird, I. & Rochefort, H. Breast cancer cells have a high capacity to acidify extracellular milieu by a dual mechanism. Clin. Exp. Metastasis 15, 382–392 (1997). 88. Brizel, D. M. et al. Elevated tumor lactate concentrations predict for an increased risk of metastases in head-and-neck cancer. Int. J. Radiat. Oncol. Biol. Phys. 51, 349–353 (2001). 89. Walenta, S. et al. High lactate levels predict likelihood of metastases, tumor recurrence, and restricted patient survival in human cervical cancers. Cancer Res. 60, 916–921 (2000). 90. Ito, S. et al. Coexpression of glucose transporter 1 and matrix metalloproteinase-2 in human cancers. J. Natl Cancer Instit. 94, 1080–1091 (2002). 91. Al Mehdi, A. B. et al. Intravascular origin of metastasis from the proliferation of endothelium-attached tumor cells: a new model for metastasis. Nature Med. 6, 100–102 (2000). A watershed paper describing the microenvironmental behaviour of lung metastases using a novel microscopy method. This paper challenges the paradigm that extravasation is a necessary component of the metastasis programme. 92. Wong, C. W. et al. Intravascular location of breast cancer cells after spontaneous metastasis to the lung. Am. J. Pathol. 161, 749–753 (2002). 93. Rofstad, E. K. & Danielsen, T. Hypoxia-induced metastasis of human melanoma cells: involvement of vascular endothelial growth factor-mediated angiogenesis. Br. J. Cancer 80, 1697–1707 (1999). Provides clear evidence that pretreatment with acute hypoxia can increase the efficiency of metastasis. 94. Younes, M., Lechago, L. V. & Lechago, J. Overexpression of the human erythrocyte glucose transporter occurs as a late event in human colorectal carcinogenesis and is associated with an increased incidence of lymph node metastases. Clin. Cancer Res. 2, 1151–1154 (1996). 95. Yasuda, S. et al. 18F-FDG PET detection of colonic adenomas. J. Nucl. Med. 42, 989–992 (2001). 96. Gatenby, R. A., Gawlinski, E. T., Tangen, C. M., Flanigan, R. C. & Crawford, E. D. The possible role of postoperative azotemia in enhanced survival of patients with metastatic renal cancer after cytoreductive nephrectomy. Cancer Res. 62, 5218–5222 (2002). Acknowledgements We wish to acknowledge the invaluable contributions of E. Gawlinski and T. Vincent for their efforts in the mathematical modelling that led to the insights presented here. We also thank E. Racker for stimulating this research by posing to us the question in the title. Competing interests statement The authors declare no competing financial interests. Online links DATABASES The following terms in this article are linked online to: National Cancer Institute: http://cancer.gov/ breast cancer | cervical cancer | colon cancer | gastric cancer | head and neck cancer | oesophageal cancer Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene caspase-3 | GLUT1 | GLUT3 | HRAS | HIF1α | KRAS | MYC | p53 | VEGF | VHL FURTHER INFORMATION DMetrix digital imaging program: www.dmetrix.com Access to this interactive links box is free online. VOLUME 4 | NOVEMBER 2004 | 8 9 9