At a molecular level, hypoxia induces the stabilization and activation of the α-subunit of an α/β heterodimeric transcription factor, appropriately termed HIF (hypoxia-inducible factor). Hypoxia is encountered, in particular, in tumour tissues, as a result of an insufficient and defective vasculature present in a highly proliferative tumour mass. In this context the active HIF heterodimer binds to and induces a panel of genes that lead to modification in a vast range of cellular functions that allow cancer cells to not only survive but to continue to proliferate and metastasize. Therefore HIF plays a key role in tumorigenesis, tumour development and metastasis, and its expression in solid tumours is associated with a poor patient outcome. Among the many genes induced by HIF are genes responsible for glucose transport and glucose metabolism. The products of these genes allow cells to adapt to cycles of hypoxic stress by maintaining a level of ATP sufficient for survival and proliferation. Whereas normal cells metabolize glucose through a cytoplasmic- and mitochondrial-dependent pathway, cancer cells preferentially use a cytoplasmic, glycolytic pathway that leads to an increased acid load due, in part, to the high level of production of lactic acid. This metabolic predilection of cancer cells is primarily dependent directly on the HIF activity but also indirectly through changes in the activity of tumour suppressors and oncogenes. A better understanding of HIF-dependent metabolism and pH regulation in cancer cells should lead to further development of diagnostic tools and novel therapeutics that will bring benefit to cancer patients.
The term hypoxia is used to describe a ‘low’ PO2 (partial pressure of oxygen), whereas normoxia refers to a ‘normal’ PO2. However, in tissues what is hypoxic to one tissue may be normoxic to another and vice versa, since different tissues have different normoxic levels of oxygen, which vary according to the degree of vascularization  and degree of oxygen consumption. In an in vitro cellular situation HIF (hypoxia-inducible factor)-α becomes progressively more stable as the PO2 decreases from 40 mmHg (around 5% oxygen) to anoxia. The tumour PO2 (1–10 mmHg) lies well within this range  as does that for certain normal tissues such as in the brain, retina and skin. The stability of HIF-α in hypoxia is explained by the inhibition of its destruction by the proteasomal machinery. In the presence of oxygen, HIF-α is highly unstable due to successive post-translational hydroxylation and ubiquitination that earmarks it for proteasomal capture. Stable HIF-α translocates to the nucleus where it dimerizes with HIF-β and binds to site-specific DNA sequences on target genes. Over 70 genes, involved in a vast array of cellular functions, are induced or repressed in a HIF-dependent manner [3–5]. In tumours, the induction of genes such as the vegf (vascular endothelial growth factor) and ang-2 (angiopoietin-2) reflects an attempt by tumour cells to re-establish vascular delivery of nutrients and oxygen through the promotion of the growth of new blood vessels from an existing network, a process referred to as angiogenesis; thereby continued cell proliferation is assured. The recognition of this angiogenic, adaptive response and its inhibition has lead to the development of effective drug treatments for several types of cancer . Targeting of other HIF-dependent gene products and/or HIF-induced modifications to the tumour micro-environment should also lead to the development of efficacious drug treatments, but a full understanding of the hypoxic signalling cascade in tumours is essential .
The HIF family
HIF is composed of an α- and β-subunit (Figure 1). These proteins belong to a large family of proteins termed bHLH (basic-helix-loop-helix) proteins containing a PAS (Per-Arnt-Sim) domain that plays diverse functions, including in the response to hypoxia . Three α-subunit isoforms, coded by different genes, together with a number of splice variants have been described. Research efforts have focused primarily on HIF-1α and more recently on HIF-2α whereas HIF-3α, the lesser-studied isoform, exists as six different splice variants , one of which plays a dominant-negative regulator role . Comparison of the sequence identity between HIF-1α and -2α shows 48% overall identity, with the N-terminal region showing the highest similarity. HIF-3α also shows considerable similarity in the bHLH-PAS region compared with the other isoforms. Both HIF-1α and -2α are expressed in a broad range of tissues but do show a degree of tissue specificity, in particular in leucocytes, endothelial cells and smooth muscle cells . HIF-1β is also expressed in a broad range of tissues, whereas the lesser-studied HIF-2β is highly expressed in developing mouse brain . The β-subunit can associate with other α-class bHLH-PAS proteins involved in a number of cellular functions, including detoxification, tracheal development and neurogenesis . This multi-association, leading to multi-functions, suggests that the β-subunit can be limiting under certain circumstances. The α-subunits are highly labile in the presence of oxygen while the β-subunits are not regulated by oxygen. The inherent instability of HIF-α results from successive post-translational hydroxylation and ubiquitination leading to proteasomal degradation.
Regulation of the stability of HIF-α through post-translational modification
The major mechanism for regulation of HIF-α stability involves coupled hydroxylation and ubiquitination, but other forms of post-translational modification, including acetylation and sumoylation, have been reported to influence either stability or activity [12,13] however, this awaits further confirmation. The mechanism of post-translational regulation of stability is similar for the different a-subunits and is controlled by oxygen-requiring dioxygenases . PHD (prolyl hydroxylase domain) dioxygenases hydroxylate proline residues in the oxygen-dependent degradation domain of HIF-α (Pro402 and Pro564 in human HIF-1α). This modification signals recruitment to HIF-α of an E3 ubiquitin ligase complex containing the VHL (von Hippel-Lindau tumour-suppressor) protein (Figure 2). VHL, together with its elongin B, elongin C, Cul 2 and Rbx-1 ubiquitinates HIF-α earmarking it for proteasomal destruction. Mutations in the vhl gene leading to loss-of-function are associated with RCC (renal cell carcinoma) and VHL disease, a familial syndrome . HIF-α is therefore stable in these cancers and HIF target genes are activated, as reflected by the very high vascularization of these tumours. These characteristics also provide a strong link between HIF and angiogenesis and tumorigenesis.
HIF-dependent gene induction
Post-translational hydroxylation of HIF-α regulates not only stability but also transcriptional activity. A dioxygenase termed FIH (factor inhibiting HIF) controls the transcriptional activity of the α/β heterodimer through hydroxylation of an asparagine residue in the C-terminal region of HIF-1α and -2α . Hydroxylation by FIH abrogates interaction with a histone acetyltransferase co-activator p300/CBP (cAMP-response-element-binding protein-binding protein) (Figure 2). FIH is predominantly located in the cytoplasm of cells and its expression is not regulated by the oxygen concentration. Thus under low PO2 HIF-α is stable and translocates to the nucleus where it associates with HIF-β through their respective N-terminal bHLH-PAS domains. The heterodimer then targets genes containing site-specific HREs (hypoxia-response elements), present on a large number of genes, and transcriptionally activates or represses their expression [3–5] (Figure 3).
The transcriptional activity of HIF is determined by two TADs (transcriptional activation domains), the NAD (N-terminal TAD) and CAD (C-terminal TAD), located in the C-terminal region of HIF-1α and -2α; where only the CAD is negatively regulated by FIH. Investigation into the significance of the existence of two TADs revealed that different genes are more or less dependent on FIH inhibition, thus on one or the other, or both, TAD  (Figure 4). In vitro assays indicate that PHD and FIH are active under different ranges of PO2, whith PHD requiring a higher PO2 for activity. Since an oxygen gradient exists in tumours, a situation where HIF-α is stable but inactive in its CAD, due to FIH hydroxylation, is therefore hypothetically valid as illustrated by differential dependence of different genes on FIH . Thus this double-headed property in transcriptional control represents a gene-selective mechanism for fine-tuning of the hypoxic response in tumours.
HIF isoform selectivity represents another gene-selective mechanism. This is illustrated by the finding that the HIF-dependent gene ca9 (carbonic anhydrase 9) is predominantly induced by HIF-1 whereas HIF phd3 (prolyl hydroxylase 3) is induced by HIF-1 and -2 .
Regulation of the metabolic balance in cancer cells by HIF
Histological examination of solid tumours generally shows a central necrotic region surrounded by a ring of viable cells that are poorly vascularized. The progression towards necrosis in a tumour mass is associated with a decreasing gradient in the level of oxygen and nutrients, and an increasing gradient toward a lower pH. To compensate for a low oxygen and nutrient supply, which limits the production of ATP required for survival and proliferation, tumour cells respond by increasing their capacity to take up and metabolize glucose rapidly. HIF directly up-regulates the expression of glucose transporters (GLUT1 and GLUT3) and of glycolytic enzymes (Figure 5) , therefore allowing for more efficient glucose capture and rapid glucose conversion into pyruvate. In addition, cancer cells bypass the oxidative steps for subsequent pyruvate metabolism by converting it into lactic acid . Conversion into lactic acid results in the rapid regeneration of the NAD+ that is mandatory for glycolysis. Thus, by increasing glucose uptake and the rate of metabolism to pyruvate coupled to diversion to lactic acid, cancer cells can maintain ATP levels under low glucose and low PO2 conditions. This predilection of cancer cells for conversion of pyruvate into lactic acid contrasts with normal cells that, under normoxic conditions, first metabolize glucose to pyruvate in the cytoplasm and then metabolize pyruvate through the TCA (tricarboxylic acid) cycle and oxidative phosphorylation in mitochondria. In contrast with the truncated cancer pathway, this pathway is oxygen-dependent and 19-fold more efficient in ATP production. However, paradoxically cancer cells, even in the presence of oxygen, prefer to use the truncated pyruvate to lactate pathway, a phenomenon termed the Warburg effect. The mechanism to explain this is not totally clear-cut but probably lies in modification in the activity of oncogenes and tumour suppressors that directly or indirectly affect HIF and HIF-target genes. This might occur through a diverse number of stimuli that implicate signalling pathways including growth factors, nitrogen oxide and ROS (reactive oxygen species), which are known to induce the translation and/or stabilization/activation of HIF, even under non-hypoxic conditions. In particular, the TCA intermediate, oxaloacetate has been reported to inhibit PHDs by binding to the 2-OG (2-oxoglutarate)-binding site  and the PHD activity might also be regulated by the oxidative status of the iron bound . ROS produced by mitochondria and released into the cytoplasm of cells exposed to hypoxia promote oxidation of Fe2+ to Fe3+. Thus hypoxia would decrease the availability of Fe2+ required for PHD activity and might be an explanation for ROS-mediated stabilization of HIF-1α . The PI3K (phosphoinositide 3-kinase)/Akt/mTOR (mammalian target of rapomycin) pathway plays a key role in nutrient utilization through a dual action on glucose metabolism. The Akt protein kinase, when fully active in cells deficient for the tumour suppressor PTEN (phosphatase and tensin homologue), can stimulate glycolysis via enhanced glucose transport and hexokinase mobilization , however mTOR can also stimulate respiration . HIF not only directly shunts pyruvate metabolism toward lactate but also inhibits pyruvate metabolism through the mitochondrial pathway. The former occurs through the induction by HIF of LDH-A (lactate dehydrogenase A) expression, the enzyme that converts pyruvate into lactic acid, while the latter occurs though the induction of PDK1 (pyruvate dehydrogenase kinase 1) that inhibits PDH (pyruvate dehydrogenase) [25,26], the enzyme that converts pyruvate into acetyl-CoA. The formation of lactic acid is important for tumorigenesis, as illustrated by the silencing of LDH-A in neu-initiated tumour cells that switches pyruvate metabolism from the cytoplasmic to the mitochondrial pathway, with the consequence of reduced tumour growth . These observations suggest that it is possible to harness cancer glucose metabolism in a therapeutic perspective. This is already the case in diagnosis and management where the high rate of uptake of glucose is visualized in tumours by PET (positron emission tomography) detection after injection of a non-metabolizable radio-analogue of glucose .
Metabolic intermediates can also modulate the stability and activity of HIF through the regulation of the 2-OG-requiring PHD and FIH dioxygenase activity. The level of 2-OG, or α-ketoglutarate, an intermediate in the TCA cycle, may be limiting in hypoxic conditions (where the TCA cycle is slowed); though it can be produced from metabolism of amino acids. In addition, accumulation of the TCA cycle intermediates succinate or fumarate, which occur in the case of mutations leading to loss of function of SDH (succinate dehydrogenase) and FH (fumarate hydratase) respectively, can act in feedback inhibition [29,30]. In fact SDH, also a component of complex II of the electron transport chain involved in oxidative phosphorylation, and FH like VHL are tumour-suppressors. This information provides yet a further link between HIF and tumorigenesis.
The singularities of cancer cell metabolism also arise from the action of other tumour-suppressors, such as p53, that not only slow glycolysis but also tip the balance of glucose metabolism toward normal mitochondrial respiration. Regulation of glycolysis by p53 can occur though the induction by p53 of the TIGAR (TP53-induced glycolysis and apoptosis regulator) gene that lowers the level of fructose-2,6-bisphosphate leading to inhibition of glycolysis  and through the negative regulation of the glycolytic enzyme PGM (phosphoglycerate mutase) . Regulation of respiration by p53 occurs via the induction of synthesis of SCO2 (cytochrome c oxidase 2) , a component required in the formation of the mitochondrial cytochrome c oxidase complex that is involved in the electron transport chain. Oncogenes promoting cell survival such as Ras , Myc  and the protein kinase Akt [23,36], are also responsible for the cancer cell predilection for glucose metabolism to lactic acid, in part, through affects on HIF and HIF-target genes. HIF-1 and nutrient deprivation also negatively regulate ATP-consuming protein translation by inhibition of the mTOR pathway [7,37], thus promoting cell survival through energy conservation. On the other hand the activation of mTOR drives glucose metabolism through the mitochondrial pathway and stimulates mitochondrial oxygen consumption and oxidative capacity . It is interesting to note that mTOR is inhibited by the upstream regulator and tumour suppressor TSC1–TSC2 (tuberous sclerosis complex), a finding that provides yet another link between metabolism and tumour suppressor function.
HIF controls tumour pH
The pHe (extracellular pH) of tumours is generally lower than that of normal tissues (6.2–6.8 compared with 7.2–7.4) [38,39]. This is a consequence of the overload in lactic acid and its poor vascular removal but also in production of carbonic acid. Under hypoxic conditions lactate secretion from cells is accelerated through the up-regulation of certain isoforms of the H+/lactate MCT (monocarboxylate transporter), in particular MCT4 , which is located on the plasma membrane (Figure 6). In addition, the carbon dioxide produced diffuses out of cells across the plasma membrane and is converted into carbonic acid by the membrane-bound, HIF-induced ectoenzyme CA (carbonic anhydrase) IX or XII. Thus hypoxic induction of CA IX and CA XII has been suggested to contribute to acidification of the extracellular space (low pHe) . Subsequent uptake of HCO3−, a weak base, by members of the Na+-dependent and -independent HCO3− transporters allows for an increase in the pHi (intracellular pH). It is interesting to note that the AE2 (anion exchanger isoform 2) was more highly expressed in VHL rescued, VHL-deficient RCCs compared with VHL-deficient cells, however overall Cl−/HCO3− exchange activity was reduced . Whether these findings point to HIF repression of AE2 expression in RCC remains to be determined. Hypoxia also brings into play one of the most important membrane bound regulators of pHi, the growth factor-activatable and amiloride-sensitive NHE1 (Na+/H+ exchanger) , the expression and activity of which is increased during hypoxia . CA IX and XII isoform expression is highly induced by HIF and histochemical studies on tumour sections have shown that these isoforms are also highly expressed to the extent that detection has been proposed as a diagnostic marker. In addition to interacting with a number of membrane exchangers and cotransporters the CA II isoform has been shown to interact with MCT1 and to enhance the rate of H+ flux mediated by MCT1 via a mechanism that does not involve its enzyme activity . It would be of interest to know whether interaction leading to inter-regulation of activity also exists between CA IX and other transporters and exchangers in cells exposed to hypoxia.
The acidic nature of the micro-environment of tumours has also been suggested to promote metastasis through the modification of the activity of HIF-induced metalloproteases responsible for extracellular matrix turnover  and through enhanced cell migration [47,48]. Of particular interest is the finding that the HIF-induced glycolytic enzyme PGI (phosphoglucose isomerase) also termed AMF (autocrine motility factor) possesses cytokine activity . AMF/PGI is up-regulated in a number of cancers and the cytokine function has been implicated in cell migration and metastasis. HIF also activates the proto-oncogene c-met and the chemokine CXCR4 that are implicated in metastasis [50,51] and represses the epithelial adhesion, tumour suppressor molecule E-cadherin presumably via activation of the transcriptional repressors Snail and SIP1 thereby further promoting the invasive potential of cells [7,52].
HIF in cell survival compared with cell death
Is HIF a pro-survival or a pro-death factor? The answer is certainly both, but with an inclination toward survival, dictated by the stress conditions within the tumour micro-environment at a given time in development of the tumour mass. Specific PO2 levels, nutrient availability and the pH in the tumour environment might allow cells to adapt and survive through HIF-mediated modulation of angiogenesis, metabolism and pH. Autophagy, a process involving vacuolar trapping of cellular components (ribosomes, mitochondria, protein aggregates) for lysosomal degradation and recycling, might also represent a hypoxia-induced cell survival strategy. However, under more severe stress conditions a cell death response may prevail, though the type of cell death, be it apoptotic or necrotic, needs to be further defined. HIF does induce the expression of several proteins reported to be pro-apoptotic . These include: mitochondrial HGTD-P, Noxa, and the protein BNIP3 (Bcl-2/adenovirus EIB 19 kD-interacting protein 3) and its homologue NIP3L (Nip-3 like) or NIX (protein X). The structural similarity of BNIP3 to other BH3-only pro-apoptotic proteins suggests that it is implicated in cell death. However, BNIP3 seems to require specific nutrient and pH conditions for it to reveal cell death and its precise role in the hypoxic response of tumour cells needs further investigation . Overall, the mechanisms of HIF-mediated apoptosis and implication in events such as autophagic nutritional rescue of cells awaits further clarification.
The recognition that HIF plays a fundamental role in tumour development has stimulated detailed investigation into the mechanisms through which it functions and ways to inhibit its activity as a potential therapeutic approach for cancer . The question as to whether it is best to abrogate HIF activity directly or rather to target specific HIF genes is open for discussion. Indeed, the targeting of the product of the HIF-inducible gene vegf, which results in diminished angiogenesis, has shown substantial promise as a therapeutic approach. Exploiting the metabolic predilection of cancer cells may represent an alternative approach in forcing cells to go down the pathway of cell death. However, a better understanding of the role of HIF in cell fate is essential to the development of such an approach.
• Hypoxia is a common feature of solid tumours and is associated with resistance to radio- and chemo-therapy, and thus to poor patient prognosis.
• HIF is induced under hypoxic conditions and transcriptionally regulates a myriad of genes involved in diverse cellular functions, in particular in cell survival via switching metabolism to glycolysis.
• A better understanding of the HIF-dependent metabolic changes that result in extracellular acidosis in cancer cells should lead to the development of additional diagnostic tools and therapeutic approaches that will prove beneficial to the management of cancer.
The laboratory is funded by grants from the Ligue Nationale Contre le Cancer (Equipe labellisée), the Centre A. Lacassagne, the CNRS (Centre National de la Recherche Scientifique), the Ministère de l’Education, de la Recherche et de la Technologie, and INSERM (Institut National de la Santé et de la Recherche Médicale). We apologize to the many research groups whose work was cited indirectly by reference to review articles.
- © The Authors Journal compilation © 2007 Biochemical Society