Acute oxygen sensing: diverse but convergent mechanisms in airway and arterial chemoreceptors

Aerobic metabolism requires an adequate supply of O₂, and rapid adaptation to changes in the partial pressures of inspired atmospheric gases is crucial to survival. During episodes of compromised O₂ availability, numerous chemosensory systems, acting in concert, rapidly modulate pulmonary ventilation and perfusion to optimize the supply of O₂ from alveolus to metabolising tissues. This review focuses on two key systems involved in this homeostatic response: the carotid bodies (CBs) and neuroepithelial bodies (NEBs), representative chemoreceptors of the arterial circulation and the airway, respectively [1,2]. So far, CBs and NEBs, together with pulmonary smooth muscle (which will not be examined in great depth here), have been the most extensively studied of O₂-sensitive tissues, and recent investigations have provided major new insights into the expression and interactions of molecular components that link a decreased partial pressure of oxygen (p_O2) to appropriate cellular responses in the circulation and respiratory systems.

CBs are highly vascularised organs, located at the bifurcations of the common carotid arteries, that rapidly initiate increased activity in afferent chemosensory fibres of the carotid sinus nerve in response to systemic hypoxaemia. There is widespread agreement that the sensory elements of the CB are the type I (glomus) cells, which contain numerous transmitters and lie in synaptic contact with afferent sensory neurones [1,3]. Type I cells release catecholamines, acetylcholine and ATP in response to hypoxia to initiate afferent discharge [4]. Commonly located at airway bifurcations are NEBs, tight clusters of neurone-derived, transmitter-containing cells that synapse with branches of both afferent and efferent neurones. They evoke appropriate responses to airway hypoxia (as opposed to hypoxaemia) by initiating afferent information to the respiratory centres [5] and releasing peptides and amine modulators [particularly 5-hydroxytryptamine (serotonin)] [6] into the local pulmonary circulation [2]. The prominence of NEBs in neonatal lungs and the association of pathological conditions, such as apnoea of prematurity and sudden infant death syndrome, with NEB cell hyperplasia strongly suggest that NEBs are involved in both the initiation of breathing at birth and cardiorespiratory control postnatally [7].

Although the specific details of the signal transduction mechanisms that link a decreased p_O2 to transmitter release in CBs and NEBs exhibit significant differences, the unifying response elements in both are p_O2-sensitive K⁺ channels [8]. Thus, decreasing p_O2 causes, sequentially, K⁺ channel inhibition [9,10], membrane depolarization [11,12], activation of voltage-gated Ca²⁺ channels and Ca²⁺-dependent transmitter release [13]. This is not generally agreed to be so in pulmonary arterioles; there is still controversy about the relative roles of capacitative/voltage-independent Ca²⁺ entry [14] and O₂-sensitive K⁺ channels in hypoxic pulmonary vasoconstriction (HPV) [15,16].

Investigations into the nature of O₂ sensing in CBs and NEBs, from sensor to effector, have had surprisingly similar aetiologies. As more detailed dissection of the signal transduction pathways was required, the use of isolated, cultured and cellular models of CBs and NEBs emerged. Thus, the precise mechanistic perspectives that are now available have been derived from the whole gamut of techniques ranging from human studies through intact CB/sinus nerve and lung slice preparations to cellular and molecular studies in PC12 cells (a rat phaeochromocytoma cell line, a model for CBs), H146 cells (a human small cell carcinoma of the lung cell line, a model for NEBs) and, most recently, knockout and recombinant experiments.

It has been clear for some time that putative O₂ sensors would be drawn from a pool of proteins that naturally underwent oxido-reductive transitions. Candidates included plasma membrane bound enzymes, cytosolic enzymes and mitochondrial complexes that contained, as key elements in the proposed redox mechanism, one or more transition metals. Thus, iron-containing haem proteins, including cytochromes and NADPH oxidases, were proposed some time ago as potential O₂ sensors in a variety of cellular systems. In NEBs, a number of lines of evidence point towards a significant, if not exclusive, involvement of NADPH oxidase in airway O₂ sensing [17,18,19]. The NADPH oxidase model for O₂ chemoreception suggests that, under normoxic conditions, the oxidase tonically generates superoxide (O₂^{- .}) from O₂ which is rapidly converted to H₂O₂ by several enzymes including superoxide dismutase and catalase. This H₂O₂ is believed to promote channel activity. Thus, native, isolated and cultured NEB cells express a number of important proteins that together constitute the multimeric functional NADPH oxidase enzyme complex, including gp91 ^phox and p22 ^phox [17]. Hypoxia caused decreased fluorescence of rhodamine 123 (indicative of decreased free radical formation) and K⁺ channel inhibition, effects that were suppressed by the relatively non-selective NADPH oxidase inhibitor, diphenylene iodonium ('DPI') [17]. Furthermore, H₂O₂ (a product of the oxidase activity) was able to stimulate K⁺ channels [17].

The suggestion that NADPH oxidase acted as a O₂ sensor and transduced the signal via changes in the intracellular redox potential was tested in the human NEB model, H146 cells [12], by exploiting the fact that NADPH oxidase activity can be regulated by the protein kinase C (PKC)-dependent phosphorylation of two components of the complex, p67 ^phox and p47 ^phox [20]. H146 cells express these proteins, hypoxia suppresses H₂O₂ levels, H₂O₂ activates 4-aminopyridine-insensitive K⁺ currents, and hypoxic K⁺ channel inhibition is suppressed by PKC activation [19]. These results provide direct functional evidence to support a role for NADPH oxidase in this important process and also suggest that PKC might modulate chemoreception by altering the affinity of the oxidase for O₂. Recently, the involvement of this oxidase has received further reinforcement by the observation that NEB cell K⁺ currents recorded from gp91 ^phox knockout mouse lung slices were acutely insensitive to acute hypoxia [18].

In contrast, the idea that NADPH oxidase provides the upstream signal for K⁺ channel inhibition has been thoroughly investigated and largely discounted by most investigators in the CB field; the haem hypothesis has gained greater credence since the observation that hypoxic inhibition of K⁺ channels can be completely reversed upon the application of carbon monoxide [21]. Similarly, the involvement of NADPH oxidase as an O₂ sensor in the pulmonary circulation has essentially been discounted by the recent report that HPV is maintained in pulmonary arterioles isolated from gp91 ^phox knockout mice [22].

The generation of reactive oxygen species (ROS) from mitochondria, as demonstrated in a number of cell types, has been suggested as one mechanism by which hypoxia can induce a cellular response [23]. However, results from most of these studies are inconsistent with mitochondrial ROS production being the major mechanism for rapid O₂ sensing, such as that seen in CBs and NEBs, because ROS are not significantly elevated during the first 10 min of the hypoxic challenge and do not become maximal for up to 2 h [24]. Mitochondrial ROS production is therefore more likely to underlie responses to chronic hypoxia, which exerts effects at the level of the gene. This does not in itself discount mitochondrial involvement in rapid O₂ sensing, because specific inhibitors of mitochondrial complexes mimic the actions of hypoxia in isolated type I CB cells [25], suggesting a potential interaction of different ROS-generating systems acting on different timescales.

An interesting parallel has arisen in CB and NEB studies relating to the specific identity of the K⁺ channels involved in the hypoxic response downstream of the sensor. In both tissues, voltage-dependent and voltage-independent channels have been implicated, and controversy still exists about the physiological contribution of each in the overall cellular response to hypoxia. Studies on CB have been further complicated by genuine species variation [26] (a factor that has not yet been thoroughly investigated for NEBs). In the rat CB, iberiotoxin-sensitive, high-conductance, Ca²⁺-activated K⁺ (maxi-K) channels were first proposed as being the O₂-sensitive channel [27], but several years later this was brought into question with the identification of a low-conductance, acid-sensitive background K⁺ channel that was proposed to be TWIK-related, acid-sensitive K_2P channel-1 (TASK1; TWIK refers to 'tandem of P-domains, weakly inward rectifying K_2P channel') — a member of the newly emerging gene family of voltage-insensitive tandem P-domain K⁺ (K_2P) channels [28].

The importance of maxi-K in transducing hypoxic stimuli into CB transmitter release had been contested until the recent observation that iberiotoxin (the selective maxi-K channel inhibitor) could, like acute hypoxia, evoke catecholamine secretion from type I cells in a novel thin slice preparation of CBs [29]. However, the contribution of TASK1 to the overall hypoxic response cannot be discounted, and awaits clarification in a preparation in situ. Similarly, a number of K⁺ channels have been implicated in HPV but recent recombinant studies point toward a voltage-activated shaw K⁺ channel (KCNC1), Kv3.1b, as the primary pulmonary arteriolar effector [16].

In NEBs, a similar controversy has arisen, in part owing to the vexed nature of consistently isolating native NEB cells from airway. At present, hypoxic inhibition of both Ca²⁺-sensitive and Ca²⁺-insensitive K⁺currents has been demonstrated in NEBs, both isolated [10] and in situ [30], but there has been a paucity of further information on the channels that underlie these currents, because of the unsuitability of primary cultured cells and lung slices for detailed molecular characterisation. A recent approach to this problem has been to establish the H146 cell as an appropriate model in which to study O₂ sensing in human NEB-derived cells [12,19,31]. Employing this model, it has been possible to verify that O₂-sensitive channels are insensitive to Ca²⁺ [12], but the contribution of Ca²⁺-activated channels still remains to be investigated robustly in native human cells and lung slices. Notwithstanding that H146 cells and native cells show some differences, it is clear that the Ca²⁺-insensitive components in the two species are almost certainly identical because their pharmacologies and biophysical natures are essentially indistinguishable. On the basis of these observations, debate still rages about the molecular identification of the Ca²⁺-insensitive K⁺ channel: a voltage-activated shaw K⁺ channel (KCNC3), Kv3.3, is proposed in native NEBs [17] and a TASK-like conductance is suggested in H146 cells [31].

Screening, by reverse-transcriptase-mediated polymerase chain reaction, for all the known human homologues of the K_2P gene family has indicated that only TWIK1 and TWIK-related, arachidonic acid-sensitive K_2P channel ('TRAAK') are not expressed in H146 cells [32]. Importantly, however, in situ hybridisation and immunohistochemical studies have now exclusively localised TASK to mouse NEB cells in lung, and recent antisense knock-down experiments in the H146 cell model have shown a high correlation between quantitative TASK expression and functional hypoxic sensitivity [33]. This antisense approach could not distinguish between TASK1 and TASK3 because they share such high identity in their open reading frame sequences; of considerable import, however, is the recent demonstration that recombinant TASK1 and TASK3 are exquisitely sensitive to decreased p_O2 when expressed in HEK 293 cells [34]. Further pharmacological dissection (using Zn²⁺ as a discriminating blocker) has now lent support to the notion that the O₂-sensitive channel is TASK3, although heterodimerism in H146 cells cannot at present be excluded (PJ Kemp, GJ Searle and C Peers, unpublished data).