Mitochondrial respiration is responsible for more than 90% of oxygen consumption in humans. Cells utilize oxygen as the final electron acceptor in the aerobic metabolism of glucose to generate ATP which fuels most active cellular processes. Consequently, a drop in tissue oxygen levels to the point where oxygen demand exceeds supply (termed hypoxia) leads rapidly to metabolic crisis and represents a severe threat to ongoing physiological function and ultimately, viability. Because of the central role of oxygen in metabolism, it is perhaps not surprising that we have evolved an efficient and rapid molecular response system which senses hypoxia in cells, leading to the induction of an array of adaptive genes which facilitate increased oxygen supply and support anaerobic ATP generation. This response is governed by HIF (hypoxia-inducible factor). The oxygen sensitivity of this pathway is conferred by a family of hydroxylases which repress HIF activity in normoxia allowing its rapid activation in hypoxia. Because of its importance in a diverse range of disease states, the mechanism by which cells sense hypoxia and transduce a signal to the HIF pathway is an area of intense investigation. Inhibition of mitochondrial function reverses hypoxia-induced HIF leading to speculation of a role for mitochondria in cellular oxygen sensing. However, the nature of the signal between mitochondria and oxygen-sensing hydroxylase enzymes has remained controversial. In the present review, two models of the role for mitochondria in oxygen sensing will be discussed and recent evidence will be presented which raises the possibility that these two models which implicate ROS (reactive oxygen species) and oxygen redistribution respectively may complement each other and facilitate rapid and dynamic activation of the HIF pathway in hypoxia.
THE JOURNEY OF OXYGEN FROM THE ATMOSPHERE TO CELLS
The oxygen level in the atmosphere has played a critical role in shaping evolution since it was introduced as a by-product of photosynthesizing cyanobacteria ∼2.3 billion years ago [1–3]. Since then, the earth's atmosphere has changed from being virtually anoxic to containing approx. 15% oxygen 1.8 billion years ago to current conditions where it contains ∼ 21% oxygen . The introduction of atmospheric oxygen with the accompanying increase in oxidizing chemistry had a catastrophic effect on the vast majority of unicellular anaerobic life forms which were rapidly made extinct . However, despite its highly reactive and potentially toxic oxidizing chemical properties, eukaryotic cells evolved to utilize oxygen as the final electron acceptor in the aerobic metabolism of glucose as the primary source of metabolic energy for most active cellular processes . The evolution of the utilization of oxygen in such a manner, which greatly increases the efficiency of ATP generation, is linked to the development of multicellular complex organisms. Thus our evolution has been and continues to be intrinsically linked to the levels of oxygen in the delicate thin terrestrial layer of the earth's atmosphere.
An efficient system of oxygen delivery which involves the combined activity of the respiratory and circulatory systems facilitates the absorption of oxygen into the bloodstream and the delivery of this oxygen to individual cells for metabolic purposes . This ‘oxygen trail’ from atmosphere to individual cells involves a gradient of pO2 values across the tissues of the body ranging from approx. 150 mmHg in the upper airway to levels as low as 5 mmHg in tissues such as the retina . Furthermore, oxygen gradients exist within tissues depending on the distance of cells from the closest oxygen-supplying blood vessel and the cellular metabolic demand. Finally, when oxygen arrives at its cellular destination, there also exist oxygen gradients within individual cells towards sites of high oxygen consumption. Although accurate measurements of intracellular oxygen gradients have remained elusive due to technical difficulties, it is highly likely that sites of high oxygen consumption such as mitochondria have lower local pO2 values than other compartments such as the cytoplasm. Thus, in normal physiological conditions (normoxia), there is a gradient of oxygen across the body ranging from 150 mmHg in the upper airway to less than 5 mmHg at the level of individual mitochondria (Figure 1). It is likely that these oxygen gradients play an important role in the control of oxygen delivery to the sites where it is most needed.
Oxygen gradients in the steady state
The delivery of oxygen from the atmosphere to individual cells is clearly of critical importance and involves a number of processes, including the convection of oxygen (in air) down the airway to the alveoli, passive diffusion into pulmonary capillaries, binding to haemoglobin, convective vascular delivery to tissues and diffusion out of the microcirculation, across the interstitium and cell membranes and ultimately to the mitochondria . Although traditionally it has been assumed that the oxygen freely diffuses across cell membranes, recent results have suggested that this may be facilitated by the existence of membrane oxygen channels . A critical issue in determining the rate of diffusion of oxygen are the pO2 gradients outlined above. This journey of oxygen from the atmosphere to individual cells has been expertly and extensively discussed elsewhere . In the present review, I will concentrate on recent developments in our understanding of the role of oxygen in signalling at the cellular level.
HYPOXIA AND CELLULAR METABOLIC CRISIS
In the steady state, cellular energy homoeostasis is a dynamically regulated process. In health, the pulmonary and cardiovascular systems combine to deliver sufficient oxygen to cells to satisfy demand for physiological processes to occur and for normal function to be maintained. However, physiological cellular demand for oxygen can vary depending on tissue requirements at a given moment. For example, an exercising muscle has significantly higher oxygen demand than the same tissue in a relaxed state. As a consequence, the rate of oxygen delivery to tissues must be dynamic. This rapid response to changes in physiological oxygen demand is largely orchestrated through the co-ordinated regulation of tissue perfusion via altered release of vasoactive substances such as NO and an acute oxygen-sensing system in small organs known as the carotid bodies. The carotid bodies are specialized sensory organs situated at the bifurcation of the carotid artery in such a way that they are in a position to detect changes in arterial oxygen . Inhibition of potassium channels in the carotid bodies results in increased calcium influx leading to the initiation of a neuronal signal to the CNS (central nervous system) ultimately signalling an increased rate and depth of breathing. Such alterations in respiratory rate and tissue perfusion give dynamic range to vascular oxygen supply in response to changes in physiological oxygen demand. The oxygen-sensing mechanisms in carotid bodies is still an area of continued investigation. However, it appears that such a mechanism, at least in part, involves a role for haem-containing proteins which interact with ion channels to initiate neuronal signalling. This acute, non-transcriptional response to physiological changes in oxygen demand has been recently expertly reviewed elsewhere [9,10]. Interestingly, it has been recently postulated that inappropriate activation of this pathway in patients suffering from OSAS (obstructive sleep apnoea syndrome), who experience dramatic nocturnal intermittent hypoxia, may play an important role in the cardiovascular pathologies (such as hypertension) commonly observed in OSAS patients .
Importantly, under steady-state conditions, the dynamic oxygen delivery to tissues is sufficient to satisfy demand with a significant excess of available oxygen. This excess of oxygen allows organisms to initially cope somewhat with falls in oxygen delivery and to facilitate the functioning of non-mitochondrial dioxygenases . However, when oxygen demand exceeds supply (hypoxia), cells rapidly enter a state of metabolic crisis which requires a fundamental shift in the cellular metabolic strategy to facilitate the creation of an adaptive state which ultimately supports tissue survival.
Despite the fact that we have evolved dynamic respiratory and metabolic systems allowing increased oxygen delivery, as outlined above, it is necessary to have protective cellular mechanisms in place which will prevent energy depletion and increase tissue perfusion in times of metabolic crisis. Two of the main cellular pathways for dealing with hypoxic stress are the AMPK (AMP-activated protein kinase) pathway and the HIF (hypoxia-inducible factor) pathway.
AMPK acts as a cellular energy sensor and its activity is sensitive to alterations in the cellular AMP:ATP ratio. As such it acts as a cellular energy gauge. When ATP levels fall, AMPK becomes activated and promotes catabolic processes while inhibiting anabolic processes. The mechanisms underlying activation of the AMPK pathway have been recently expertly reviewed elsewhere . In an AMP-dependent manner, AMPK phosphorylates and inhibits ACC (acetyl-CoA carboxylase), the rate-limiting enzyme in fatty acid synthesis; ACC catalyses the formation of malonyl-CoA, a potent inhibitor of fatty acid oxidation. Accordingly, AMPK acts to elevate fat oxidation and reduce lipogenesis. AMPK also catalyses the AMP-dependent phosphorylation and inhibition of HMG-CoA (3-hydroxy-3-methylglutaryl-CoA) reductase, the rate-limiting enzyme in cholesterol biosynthesis, thus reducing cholesterol formation. In addition, AMPK activation suppresses the expression of several lipogenic genes and activates phosphofructokinase-1, thereby suppressing glucose oxidation and enhancing glycolysis. AMPK is activated in exercise, where it triggers glucose uptake by stimulating translocation of GLUT4 (glucose transporter 4) to the skeletal muscle. The importance of AMPK in the regulation of glucose metabolism and metabolic control has recently been expertly reviewed .
A second critical pathway in determining the cellular response to hypoxia involves the stabilization and activation of HIF, a master transcriptional regulator of hypoxia-dependent gene expression. HIF-α subunits are oxygen sensitive by virtue of the fact that they are substrates of a family of proline and asparagine hydroxylases which utilize dioxygen, ferrous iron (Fe2+) and 2-oxoglutarate to catalyse the hydroxylation of specific residues on HIF-α subunits. Upon hydroxylation of Pro402 and Pro564, the human HIF-1α subunit becomes a target for the E3 ubiquitin ligase known as the VHL (von Hippel–Lindau) protein which facilitates its ubiquitination and degradation . A further hydroxylase known as FIH (factor inhibiting HIF) hydroxylates an asparagine residue (Asp803) in human HIF-1α, which prevents transcriptional activity through inhibiting the interaction with the transcriptional co-activator CBP/p300 [where CBP is CREB (cAMP-response-element-binding protein)-binding protein]. Thus hydroxylases confer a two-pronged oxygen-dependent repression of the HIF pathway (Figure 2) . In hypoxia, this repression is removed and the HIF pathway becomes rapidly activated. HIF-dependent genes include angiogenic factors such as VEGF (vascular endothelial growth factor), vasodilatatory factors such as iNOS (inducible NO synthase), and other factors which support an increase in tissue oxygenation such as EPO (erythropoeitin). Interestingly, a number of genes involved in the regulation of metabolism are also HIF-dependent, including a number of glycolytic enzymes and genes involved in actively decreasing the basal respiratory rate [15,16]. Because of its central role in the adaptive response to hypoxia, the HIF pathway has become an area of intense investigation as a therapeutic target in ischaemic disease and cancer where hypoxia plays a significant role in disease development [17,18].
The HIF pathway
HIF-DEPENDENT REGULATION OF METABOLISM
As outlined above, cells largely derive the metabolic energy required for active processes from the hydrolysis of the high-energy phosphate bond of ATP. The most efficient metabolic pathway for the generation of ATP is through the oxidative metabolism of glucose. This process, which depends upon the free availability of both glucose and molecular oxygen, efficiently generates ATP and involves three distinct phases: glycolysis, the TCA (tricarboxylic acid) cycle (also known as the Krebs cycle or the citric acid cycle) and oxidative phosphorylation. Although the components of these metabolic pathways are well understood and are described in many basic biochemistry textbooks, the regulation of cellular metabolic processes is an area which has been recently revisited with the aid of new genetic models. Technological advances such as the development of genetic knockout animals and siRNA (small interfering RNA) technology are allowing researchers to ask fundamental questions regarding the mechanisms underlying the regulation of cellular metabolism, how this impacts on cellular signalling and how this system may be targeted for therapeutic use in a range of disorders. In the present review I will outline the processes of glycolysis, the TCA cycle and oxidative phosphorylation and how the activity of these metabolic pathways may be under the control of products of HIF-dependent gene expression in hypoxia.
The first stage of respiration is glycolysis which occurs in the cell cytoplasm and involves the co-ordinated activity of ten enzymes in the conversion of glucose into pyruvate. The net ATP gain of glycolysis is two molecules of ATP per molecule of glucose metabolized. As well as generating this small yield of ATP, glycolysis also generates two electron-carrying NADH equivalents which facilitate oxidative phosphorylation.
While in the steady state (normoxia), the primary metabolic function of glycolysis is to feed pyruvate into the TCA cycle. However, eukaryotic cells retain the ability to shift metabolic strategy such that glycolysis may itself become the primary source of metabolic energy in cases of metabolic crisis such as hypoxia. In such cases there is an increase in the rate of flux through the glycolytic pathway resulting in increased glycolytic ATP production, through HIF-dependent transcriptional up-regulation of the genes involved in glycolysis, a phenomenon known as the Pasteur effect . Interestingly, tumour cells demonstrate increased glycolysis even in the presence of sufficient oxygen, an effect known as the Warburg's effect . Indeed recent work has suggested that targeting this unique metabolic feature of cancer cells may represent a new and exciting therapeutic possibility in the chemotherapeutic targeting of tumours [21,22]. Notably, as outlined above, another regulator of the glycolytic pathway is AMPK, which can positively regulate glycolysis and further promote anaerobic ATP generation .
Thus glycolysis is the first step of aerobic glucose metabolism which may be utilized as the primary source of ATP under conditions of metabolic crisis. Whether increased HIF-dependent expression of glycolytic enzymes alone is sufficient to account for the dramatic increase in glycolytic ATP production observed in hypoxia or whether other events such as the spatial reorganization of glycolytic enzymes into glycolytic complexes are important remains to be determined. However, it is probable that the dramatically enhanced efficiency of the glycolytic pathway observed in hypoxia involves the co-ordinated events of increased expression and subcellular spatial re-organization of the enzymes involved.
Each molecule of glucose that is metabolized through glycolysis results in the generation of two molecules of pyruvate. In the steady state the majority of pyruvate generated is oxidized by pyruvate dehydrogenase to form acetyl CoA, which combines with oxaloacetic acid to form citric acid. This citrate then enters the TCA cycle which is housed within the mitochondria where a series of eight enzymatic reactions result in the net generation of one molecule of ATP, three molecules of NADH and one molecule of FADH2 per molecule of pyruvate metabolized. While the ATP generated during the TCA cycle is utilized to provide the energy necessary to carry out active cellular processes, the NADH and FADH2 will function to carry electrons in the process of oxidative phosphorylation.
The entry of pyruvate into the TCA cycle can be a regulated process and is also under transcriptional control by the HIF pathway . Briefly, in hypoxia, HIF-dependent up-regulation of PDK1 (pyruvate dehydrogenase kinase 1), which regulates the activity of pyruvate dehydrogenase (the enzyme responsible for converting pyruvate into acetyl-CoA), causes a decrease in the pyruvate available for entry into the TCA cycle. Secondly, LDHA (lactate dehydrogenase A) is also increased in hypoxia in a HIF-dependent manner leading to increased conversion of pyruvate into lactate. In combination, the HIF-dependent expression of PDK1 and LDHA in hypoxia leads to inhibition of the TCA cycle.
Oxidative phosphorylation is the process by which cells utilize over 20 electron-carrying proteins located within the mitochondrial electron transport chain and arranged in four polypeptide complexes to generate cellular energy necessary to generate ATP. This process utilizes the NADH and FADH2 generated during glycolysis and the TCA cycle. The mechanism underlying this process involves the transport of high-energy electrons from NADH and FADH2 through a series of carrier molecules which maintain the electrons at gradually lower energy levels, resulting in the release of energy allowing the generation of a proton gradient across the inner mitochondrial membrane which fuels the activity of ATP synthase which phosphorylates ADP to form ATP. Molecular oxygen acts as the final electron acceptor at complex IV of the electron transport chain (cytochrome oxidase). As outlined above, this enzyme has a uniquely low Km value for molecular oxygen (<1 μM) and is responsible for over 90% of the body's oxygen consumption . Like glycolysis and the TCA cycle, oxidative phosphorylation through the electron transport chain can be regulated by gene products of the HIF pathway in conditions of hypoxia. This occurs through transcriptional mechanisms leading to alteration in the cytochrome oxidase subunits expressed to optimize energy consumption in conditions of low oxygen tension .
Importantly, the process of oxidative phosphorylation is not 100% efficient and approx. 20% of protons undergo regulated proton leak, which gives rise to the generation of mitochondrial ROS (reactive oxygen species) and possibly thermogenic processes . Thus a basal level of ROS generation occurs where oxidative phosphorylation is uncoupled from ATP synthesis.
In total, the combined activity of glycolysis, the TCA cycle and oxidative phosphorylation is very efficient and generates a net yield of 38 molecules of ATP per molecule of glucose metabolized. Critically, the regulation of each of the three phases of aerobic respiration, glycolysis, the TCA cycle and oxidative phosphorylation, are under regulatory control determined by transcriptional events which are governed by HIF (Figure 3). Thus HIF plays a central role in determining the cellular metabolic strategy under conditions of hypoxia.
Regulation of metabolism by HIF
HIF-INDEPENDENT REGULATION OF MITOCHONDRIAL METABOLISM
As detailed above, basal metabolic processes in hypoxia can be regulated at the transcriptional level through HIF-dependent events. This is critical, as cellular metabolic demands can change depending on physiological function at any given time. Thus it is critical that the rate of steady-state mitochondrial metabolism can be regulated. A significant question remains as to what factors besides glucose and oxygen availability and the HIF-dependent transcriptional pathways outlined above are responsible for the control of basal metabolism. Previously, it has been discovered that the consumption of oxygen by the electron transport chain, may be regulated by the endogenous gas NO . NO acts as a competitive inhibitor of cytochrome oxidase by competing for binding with molecular oxygen. Thus an increase in cellular NO results in an increase in the apparent Km of cytochrome oxidase for oxygen. This implicates a critical role for NO in the regulation of cellular metabolism via the control of oxidative phosphorylation. Thus elevated local release of NO will have the effect of inhibiting cellular respiration and determining the tonic activity of the electron transport chain.
A second mechanism by which mitochondrial activity may be regulated in the steady state is through the regulation of mitochondrial biogenesis. It was recently found that NO can also modulate changes in mitochondrial biogenesis . Furthermore, transcriptional regulators of mitochondrial proteins and components of the electron transport chain, including nuclear receptors such as NRFs (nuclear respiratory factors), ERRs (oestrogen receptor-related proteins), PPARs (peroxisome-proliferator-activated receptors) and the transcriptional co-activator PGC-1α (PPARγ co-activator 1α), may also be important regulators of the basal metabolic rate through the control of mitochondrial biogenesis . Thus a number of regulated pathways leading to changes in the cellular mitochondrial mass may also impact upon the basal level of metabolic activity.
Finally, although it is an area of limited understanding and extensive investigation, it is probable that changes in mitochondrial ultrastructure including both mitochondrial fission and fusion may be important events in adjusting the dynamic steady-state metabolic level [29–31]. Exciting recent studies in this area have uncovered a central role for the dynamic mitochondrial ultrastructure in the regulation of basal cell metabolism [29–31]. In summary, the steady-state rate of mitochondrial metabolism and ATP generation is a dynamic process which can be altered by changes in mitochondrial enzyme activity, mitochondrial biogenesis and mitochondrial fusion/fission.
MITOCHONDRIA AND OXYGEN SENSING IN THE HIF PATHWAY
Clearly the HIF pathway plays an important role in the regulation of metabolism under conditions of hypoxia. A critical key question which remains an area of intense investigation and controversy involves the determination of the nature of the signalling events linking a fall in environmental pO2 to inhibition of hydroxylase activity and activation of the HIF pathway in cells. In 1998, Chandel, Schumacker and co-workers  proposed that the mitochondria, as the primary oxygen consuming organelles of a cell, play an important role in signalling to the activation of HIF in hypoxia. Initial studies by this group demonstrated the necessity for a functioning electron transport chain to stabilize HIF-1α in hypoxia. This idea has been supported by a significant number of publications which demonstrate that either pharmacological or genetic inhibition/knockout of the mitochondrial electron transport chain leads to a loss of hypoxia-induced HIF [32–35]. This response has proven consistent for Rho zero cells (which lack mitochondria), cells with genetic deficiencies in key respiratory proteins (cytochrome c, Rieske-iron sulfur protein), cells in which the respiratory chain has been pharmacologically inhibited (by rotenone, myxathiozol, antimycin etc.) or physiologically inhibited (by NO) [32–35]. Critically, under conditions of anoxia, the ability of these cells to stabilize and activate HIF is retained.
Although it is now generally accepted that the level of mitochondrial activity impacts upon activation of the HIF pathway, two different interpretations of the nature of the signal between mitochondria and the prolyl hydroxylases regulating HIF have evolved. In the present review, I will refer to these two theories as ‘the ROS hypothesis’ and ‘the oxygen redistribution hypothesis’. While initially these two theories of the role of mitochondria in oxygen sensing were considered opposing and led to extensive and often heated debate, recent studies have raised the possibility that the two pathways may in fact represent different and possibly complementary arms of the signal transduction pathway linking altered mitochondrial activity to HIF activation in hypoxia. In the present review, I will outline these two theories of the role of mitochondria in cellular oxygen sensing.
THE ROS HYPOTHESIS
The first proposed mechanism underlying the inhibition of hypoxic HIF-1 activation by mitochondrial inhibitors was proposed by Chandel et al.  in 1998. In this model, the paradoxical generation of mitochondrial ROS in hypoxia was proposed as a link between the mitochondria and oxygen sensing by the HIF pathway. Since this proposal, a significant amount of supportive data has been generated to indicate a role for mitochondrial ROS in the regulation of the HIF pathway in hypoxia-dependent and -independent activation of the HIF pathway .
First, a number of groups have reported that ROS levels are increased in hypoxia [36–38]. Conversely, a number of other groups have reported that ROS are decreased in hypoxia [36–38]. The most likely explanation for this controversy is that the altered production of ROS by cells in hypoxia is dependent upon basal cell metabolism and the degree and duration of hypoxia experienced and is thus dependent to some degree on the model adopted. Of interest is that the transcription factor Nrf-2 (nuclear factor-erythroid 2 p45 subunit-related factor 2) which senses alterations in oxidative stress to induce an increase in the expression of the cells antioxidant capacity is not activated in hypoxia but is activated upon re-oxygenation , indicating that the levels of ROS generated in hypoxia are in line with a signalling response rather than a pathophysiological response and that different ROS-dependent transcription factors demonstrate differential responsiveness depending on the degree of oxidative stress experienced by a cell.
Secondly, it has been demonstrated that the addition or expression of antioxidants is sufficient in some systems to reverse hypoxia-induced HIF activation. Studies using non-targeted pharmacological antioxidants gave contradictory results in determining the antioxidant sensitivity of the hypoxia-induced HIF-pathway [33,40,41]. A recent study utilizing the mitochondrial-targeted antioxidant Mito-Q has given further strength to the possibility of a role for ROS in mitochondrial signalling to the HIF pathway in hypoxia . Furthermore, recent anti-tumorigenic effects of antioxidants have been attributed to inhibition of HIF-dependent events . Recent studies also demonstrate the Qo site of mitochondrial complex III as being required for the generation of mitochondrial ROS in hypoxia .
Thirdly, the addition of exogenous pro-oxidants such as hydrogen peroxide is sufficient to induce HIF activity in normoxia [34,35]. Although these experiments are difficult to relate to physiological or pathophysiological states where intracellular hydrogen peroxide levels are difficult to predict, they indicate that an increase in oxidative stress can indeed lead to activation of the HIF pathway. This idea is supported by an elegant study by Gerald et al.  which demonstrated that genetic removal of the JunD-dependent antioxidant pathway leads to increased HIF activation through decreased activity of the HIF prolyl hydroxylases [36,43]. Of note, this study offered, for the first time, a potential mechanistic link between oxidative stress and hydroxylase inhibition via altered cellular Fenton chemistry leading to an accumulation of iron in the ferric (Fe3+) state which cannot be utilized by hydroxylases which require iron in the ferrous (Fe2+) state as a cofactor.
Taken together these data indicate that the HIF pathway can be regulated by alterations in cellular oxidative stress and that this probably occurs through altered cellular Fe2+ levels leading to altered hydroxylase activity. Changes in mitochondrial ROS production occur in hypoxia although the direction and magnitude of this change probably differs depending on the basal metabolism of the cell type and the degree and duration of hypoxia experienced. In summary, the HIF pathway demonstrates responsiveness to alterations in mitochondrial ROS production probably through altering the availability of ferric iron for utilization as cofactors by the PHDs (prolyl hydroxylases)/FIH.
THE OXYGEN REDISTRIBUTION HYPOTHESIS
The second proposed mechanism underlying the inhibition of hypoxic HIF-1 activation by mitochondrial inhibitors involves the oxygen redistribution hypothesis . According to this hypothesis, under conditions of hypoxia, the low Km of cytochrome oxidase (<1 μM) dictates that the vast majority of available oxygen is consumed by the mitochondria leaving little or no ‘spare’ cytosolic oxygen to facilitate the activity of dioxygenases which have significantly higher Km values for oxygen (e.g. prolyl and asparaginyl hydroxylases). In vitro estimates of the hydroxylase Km values are in the region of 250 μM ; however, these values are probably much lower in vivo. Thus HIF hydroxylation is repressed and HIF becomes stabilized and transactivated. According to this theory, the inhibition of mitochondrial function in hypoxia increases the apparent Km of cytochrome oxidase for molecular oxygen and results in decreased oxygen consumption leading to derepression of hydroxylase activity via the redistribution of oxygen away from the mitochondria. This hypothesis was initially developed from studies investigating the inhibition of mitochondrial respiration by NO [26,40]. In the study which proposed this model , the reversal of HIF stabilization is detectable under atmospheric oxygen concentrations as low as 1%. This indicates that even full mitochondrial inhibition will only result in an increase in intracellular pO2 to a maximum of approx. 7–8 mmHg. These results indicate that the actual in vivo Km values of the hydroxylases for oxygen are significantly lower than those in vitro measurements made to date . Indeed, it has been extensively demonstrated that HIF is stabilized only at atmospheric concentrations below 5% atmospheric oxygen. Under such conditions, intracellular oxygen concentrations are probably much lower as oxygen consumption occurs and probably exceeds the diffusion gradient of oxygen across the cell culture medium. In an elegant demonstration of this, Doege et al.  demonstrated that although mitochondrial inhibition is effective in reversing hypoxia-induced HIF activation in cells grown on oxygen-impermeable plastic, this effect is lost when cells are grown on oxygen-permeable membranes where the oxygen gradient across the medium is removed. In summary, the oxygen redistribution theory proposes that the activation of HIF in hypoxia is dependent upon the availability (or lack thereof) of oxygen not consumed by the cytochrome oxidase of the electron transport chain.
A COMMON THEORY OF THE ROLE OF MITOCHONDRIA IN OXYGEN SENSING
The two theories on the role of mitochondria in cellular oxygen sensing in the HIF pathway outlined above implicate separate roles for altered levels of ROS and molecular oxygen in the hydroxylase-dependent regulation of HIF activity. In the present review, I would contest that these theories are not necessarily contradictory. As has been proposed by Gerald et al. , elevated ROS may mediate their effects in hypoxia on HIF through altered Fenton chemistry, leading to increased accumulation of Fe3+ and decreased Fe2+ availability. This would probably profoundly affect the activity of hydroxylases as they require Fe2+ as a cofactor. Inhibition of mitochondrial ROS production via inhibition of the respiratory chain would, in this case, reverse hypoxia-induced HIF activation by increasing Fe2+ availability, thus facilitating hydroxylase activity and reversing the activation of the HIF pathway.
Applying the oxygen redistribution hypothesis in normoxia, sufficient available non-mitochondrial oxygen is free to facilitate hydroxylase activity in the steady state, maintaining the HIF pathway in a repressed state. Under conditions of hypoxia, the low Km of cytochrome oxidase dictates that virtually all oxygen is consumed by the mitochondria leaving no cytosolic oxygen for HIF hydroxylase activity removing the repression of the pathway leading to stabilization and transactivation of HIF. In this case, mitochondrial inhibition in hypoxia leads to a redistribution of oxygen to the cytoplasm and other cellular compartments where hydroxylases are present, thus reactivating the hydroxylases and causing the cell to fail to sense hypoxia.
The following model is proposed (Figure 4). Under conditions of normoxia, ATP production is sufficient to allow physiological function and sufficient excess oxygen is available for non-mitochondrial purposes such as facilitating the activity of hydroxylases, thus maintaining a tonic repression of the HIF pathway. In this state a low level of ROS are produced through the proton leak. As oxygen levels fall and the tissue becomes hypoxic, cytochrome oxidase (as a result of a high affinity for oxygen) consumes virtually all available oxygen rendering non-mitochondrial compartments essentially anoxic leading to decreased hydroxylase activity and activation of the HIF pathway. In tandem with this decrease in oxygen availability, ROS levels in cells increase leading to a shift in Fenton chemistry and a chemical environment which favours the existence of ferric iron (Fe3+). In this case, hydroxylase activity is further inhibited due to the lack of availability of Fe2+. In this case, alterations in available oxygen and Fe2+ may combine to rapidly and effectively inhibit hydroxylase activity thus facilitating the rapid activation of the HIF pathway. Under conditions of hypoxia where mitochondria are inhibited, increased availability of oxygen to non-mitochondrial compartments, along with increased availability of ferrous iron will lead to an efficient re-activation of hydroxylases and repression of HIF activity (Figure 4). In summary, the combined impact of hypoxia on oxygen availability to non-mitochondrial targets (where hydroxylases reside) and ferrous iron availability, both of which critically depend on the level of mitochondrial activity, may be critical events in the signalling pathway linking environmental hypoxia with activation of the HIF pathway.
A model of oxygen sensing in hypoxia
The discovery of the HIF pathway and the elucidation of the role of hydroxylases were seminal discoveries in our understanding of oxygen sensing and the cellular response to hypoxia. Because of the importance of this pathway in a range of disease processes, much work has been invested in understanding the signalling mechanisms involved the perception of hypoxia by a cell. A clear understanding of this event will lead to the development of therapeutics that target the HIF pathway for either repression or activation. In ischaemic disease such as stroke, it can be envisioned that activation of the HIF pathway would be beneficial and promote tissue survival. Conversely, the inappropriate activation of HIF in a hypoxic tumour, which promotes tumour angiogenesis and progression, represents an attractive target in cancer therapy.
Future work in developing our understanding of the hydroxylase/HIF pathway will expand to determine the role of other hydroxylase cofactors as potential therapeutic targets. For example, recent work identifying a deficiency in the TCA cycle enzyme fumarate hydratase, which leads to decreased levels of the hydroxylase cofactor 2-oxoglutarate, as a causative mutation in oncogenesis indicates that regulation of TCA cycle intermediates may be another attractive therapeutic approach for manipulating the HIF pathway [46,47]. Furthermore, as outlined above, iron availability and Fenton chemistry probably play important roles in the regulation of hydroxylases. The regulation of such events will probably evolve as a further mechanism for the therapeutic regulation of the HIF pathway. A recent study has indicated that altering mitochondrial activity can have an impact upon the effectiveness of hypoxia-sensitive chemotherapeutic agents .
Since its discovery by Semenza and Wang  just over 15 years ago, the HIF pathway has allowed the development of a greater understanding of how cells sense oxygen and respond accordingly. Understanding the signalling mechanisms involved in this will allow the manipulation of this pathway in a broad new range of therapeutic approaches in diseases where hypoxia plays a role in the development of disease or recovery.
The author's work is funded by Science Foundation Ireland.
AMP-activated protein kinase
CREB (cAMP-response-element-binding protein)-binding protein
factor inhibiting HIF
lactate dehydrogenase A
obstructive sleep apnoea syndrome
pyruvate dehydrogenase kinase 1
reactive oxygen species