On the existence of a central respiratory oxygen sensor

Abstract

A commonly held view that dominates both the scientific and educational literature is that in terrestrial mammals the central nervous system lacks a physiological hypoxia sensor capable of triggering increases in lung ventilation in response to decreases in Po₂ of the brain parenchyma. Indeed, a normocapnic hypoxic ventilatory response has never been observed in humans following bilateral resection of the carotid bodies. In contrast, almost complete or partial recovery of the hypoxic ventilatory response after denervation/removal of the peripheral respiratory oxygen chemoreceptors has been demonstrated in many experimental animals when assessed in an awake state. In this essay we review the experimental evidence obtained using in vitro and in vivo animal models, results of human studies, and discuss potential mechanisms underlying the effects of CNS hypoxia on breathing. We consider experimental limitations and discuss potential reasons why the recovery of the hypoxic ventilatory response has not been observed in humans. We review recent experimental evidence suggesting that the lower brain stem contains functional oxygen sensitive elements capable of stimulating respiratory activity independently of peripheral chemoreceptor input.

the high metabolic rate of the brain associated with the activities of millions of nerve cells processing information requires constant, optimal nutrient and oxygen supply, as well as effective removal of carbon dioxide and other metabolic waste products. Adequate oxygenation of the arterial blood supplying the brain is monitored by specialized respiratory oxygen sensors located in the carotid bifurcation (carotid bodies) and, in some species, the aortic arch (aortic bodies). These peripheral chemoreceptors detect decreases in the arterial Po₂ and transmit chemosensory information to the brain stem respiratory centers, triggering adaptive changes in breathing (34, 48). This simple “textbook” view on chemosensory control of breathing implies that detection of the arterial partial pressure of O₂ (Po₂) at the level of the peripheral chemoreceptor is sufficient to ensure appropriate oxygenation of all regions of the brain. However, significant gradients of brain tissue oxygen levels have been demonstrated at normal arterial Po₂ (15, 31), supporting the contentious idea that many central neurons may operate in a low-oxygen environment (19, 44). Being located “upstream” from the central nervous system, arterial respiratory chemoreceptors are obviously not able to detect and respond to significant regional differences in brain oxygenation or local brain hypoxia. Moreover, all species of terrestrial mammals studied so far survive surgical denervation or removal of the peripheral oxygen chemoreceptors with no major adverse physiological consequences. Although hypoxic stress at sea level is rare, peripherally chemodenervated experimental animals and humans can tolerate hypoxia that might be common during sleep in a host of disparate disease states.

A commonly held view that dominates both the scientific and educational literature is that the central nervous system lacks a physiological oxygen sensor capable of stimulating the brain stem respiratory network and lung ventilation in response to decreases in the Po₂ of the brain parenchyma. Indeed, no recovery of the normocapnic hypoxic ventilatory response has ever been reported in humans following bilateral resection of the carotid bodies (63). In contrast, significant evidence from experimental animal studies demonstrates almost complete or partial recovery of the hypoxic ventilatory response after surgical denervation/removal of the peripheral respiratory oxygen chemoreceptors (1, 2, 12–14, 18, 40, 42, 46, 51, 53, 54).

In this short review article we discuss the experimental evidence obtained in studies of the hypoxic ventilatory response using animal models (in vitro and in vivo) as well as human subjects with denervated peripheral respiratory oxygen sensors. We consider potential reasons why the recovery of the hypoxic ventilatory response has not been observed in human studies. We also discuss recent experimental evidence suggesting that the lower brain stem contains functional oxygen-sensitive elements capable of stimulating breathing independently of peripheral chemoreceptor input.

Animal Studies

In mammals, acute moderate hypoxia generally induces a biphasic hypoxic ventilatory response, which consists of an initial strong increase in the respiratory effort (within the first minute of exposure to hypoxia) (phase I) followed by a secondary reduction in respiratory activity (phase II, roll-off or hypoxic ventilatory decline) over the next several minutes reaching a new steady-state level of ~25–40% above the baseline (43, 60). The initial increase in the respiratory activity is believed to be primarily triggered by activation of the peripheral respiratory oxygen chemoreceptors whereas the secondary reduction of ventilation is traditionally attributed to the hypoxia-induced depression of the brain stem respiratory circuits [although there is evidence that in the awake state hypoxic ventilatory decline is also dependent on the carotid body input (e.g., see 33)].

Indeed, the majority of central neurons respond to hypoxia with a reduction in excitability. However, in the early 1990’s Sun and Reis (57, 58) demonstrated that some CNS neurons, in particular presympathetic neurons of the brain stem, increase their discharge in low-Po₂ conditions and trigger generalized increases in sympathetic nerve activity during brain hypoxia. Subsequent experimental studies performed using in vitro, in situ, and anesthetized animal models demonstrated that in addition to presympathetic neurons, brain stem respiratory control circuits are sensitive to, and can be activated by, decreases in local parenchymal Po₂ or cytotoxic hypoxia (10, 38, 45, 47, 49, 52, 55, 56, 61). These studies demonstrated biphasic responses of the isolated (in an in vitro brain stem slice) central respiratory network to hypoxia, quantified Po₂ sensitivity, described differential neuronal responses, and suggested some plausible mechanisms of neuronal oxygen sensing [e.g., heme oxygenase-dependent hypoxia-induced depolarization of cultured brain stem neurons (10)].

These experimental studies conducted in vitro support a central stimulatory component of the hypoxic ventilatory response; however, whether these mechanisms (activated in vitro) are relevant to the physiological responses that are involved in the homeostatic regulation of brain oxygenation by the respiratory network remains a contentious issue. Few investigators working in intact preparations (e.g., in vivo) consider that the responses of the respiratory network to hypoxia/anoxia in vitro are physiologically relevant. This is a valid criticism because baseline conditions of the in vitro preparations are characterized by extreme hyperoxia and the strength of the hypoxic stimulus varies with tissue depth since O₂ delivery is via diffusion (discussed in detail in 17).

In anesthetized animal models hypoxia triggers a prototypical biphasic hypoxic ventilatory response, which relies on the integrity of the peripheral respiratory oxygen sensors for its full expression. Acute denervation of the carotid bodies abolishes the initial stimulatory effect of hypoxia and the hypoxic ventilatory response often manifests as a depression of central respiratory drive (e.g., see 23). In contrast, unanesthetized experimental animals (e.g., Figure 2 in Ref. 9) and humans with (chronically) denervated carotid bodies (e.g., Figure 3 in Ref. 59) increase and/or maintain the respiratory activity during the acute hypoxic challenge above and/or at the normoxic level and do not exhibit depression of ventilation in response to CNS hypoxia.

Earlier experimental studies involving chronic denervation of the peripheral oxygen chemoreceptors and assessment of respiratory activity demonstrated almost complete or partial recovery of the hypoxic ventilatory response in awake animals, including dogs (13), cats (18, 42), ponies (2, 3), goats (12), and rats (40, 46, 51). More recent studies in rodents also reported (but did not comment on this particular aspect) robust respiratory responses to hypoxia in awake mice and rats with chronically denervated carotid bodies (14, 54). The latest studies of the mechanisms underlying peripheral oxygen sensitivity using transgenic animal models reported that during the early postnatal period in mice, chemosensitive carotid body glomus cells do not express critical components of the hypoxia-sensitive signaling pathway (e.g., olfactory receptor Olfr78 which is activated by lactate produced during hypoxia; in the absence of Olfr78 isolated carotid bodies are reported to be insensitive to hypoxia) (6). Yet, neonatal mice display robust and sustained hypoxic ventilatory responses (4, 50).

A number of potential mechanisms may account for the development of the hypoxic ventilatory response in the absence of carotid body chemoreceptor input. In some species subsidiary peripheral chemoreceptors (e.g., aortic bodies) may take over and compensate for the loss of the carotid body afferent activity (for a review see 24). Recent data in rodents provide strong evidence that the hypoxic ventilatory response that develops in conditions of peripheral chemoreceptor denervation is centrally mediated and involves activation of oxygen sensitive glial cells leading to the enhanced activity of the brain stem respiratory network (1).

Careful scrutiny of the literature reveals that the use of anesthesia is a potential confounding factor. If there is a central stimulatory component of the hypoxic ventilatory response, it appears to be very sensitive to suppression by general anesthesia. Indeed, a majority of studies that reported recovery of a hypoxic ventilatory response after carotid body denervation were conducted in unanesthetized animals. A recent study measured changes in ventilation in the same cohort of peripherally chemodenervated rats (10 wk after carotid body ablation) in awake state and under general anesthesia (1). In these rats, the hypoxia-induced arousal response (5, 42) was markedly reduced following carotid body denervation indicating that the afferent input from the peripheral oxygen chemoreceptors was absent/impaired. Yet, when awake these animals displayed a robust hypoxic ventilatory response that was dramatically reduced under general anesthesia (urethane) (1), suggesting that the signaling mechanisms underlying central respiratory oxygen sensitivity are inhibited by anesthetic agents.

Two studies specifically investigated the effect of CNS hypoxia on the respiratory activity in unanesthetized sleeping dogs (9) and awake goats (12). Both studies reported significant (~30% in dogs and ~75% in goats) increases in ventilation during CNS hypoxia, despite the fact that the carotid body respiratory chemoreceptors were maintained normocapnic/normoxic by means of vascular isolation and separate perfusion. In both studies the hypoxic ventilatory response induced by CNS hypoxia was mediated entirely by increased respiratory frequency, strongly suggesting an effect of low Po₂ on the respiratory rhythm-generating circuits. In dogs, however, once the carotid bodies were acutely denervated, CNS hypoxia failed to trigger increases in ventilation suggesting that tonic (permissive) peripheral chemoreceptor input (provided by normoxic/normocapnic carotid bodies) is required for central oxygen-sensitive mechanisms to stimulate breathing (9). In goats, a significant proportion (~40%) of the ventilatory response to CNS hypoxia was preserved following carotid body denervation (12).

Human Studies

In humans, denervation of the carotid body chemoreceptors abolishes the hypoxic ventilatory response (for comprehensive reviews of the literature see 60, 63). Most of these studies were conducted in small numbers of patients following bilateral resection of healthy carotid bodies to treat bronchial asthma or chronic obstructive pulmonary disease (25, 26, 37, 62, 66, 69). Following denervation, these patients also failed to show the characteristic rapid decline in ventilation that occurs in response to rapid administration of 100% oxygen (the Dejours test) (37). Ventilatory responses to CO₂ were also reduced, typically by ~20–30% (for a review see 63). However, more extreme respiratory deficits were reported shortly after the denervation surgery, including 75% reductions in ventilatory CO₂ sensitivity that recovered gradually over 2–3 years (11).

A key point is that the majority of human studies involving carotid body denervation have been conducted in patients suffering from chronic lung disease, which may (by itself) significantly impact chemoreflex function. Two studies (16, 62) recruited individuals who had undergone bilateral carotid body tumor resection, but were otherwise free of pulmonary disease. The hypoxic ventilatory response was abolished in seven bilaterally carotid body-resected subjects studied by Fatemian and colleagues (16). Similarly, eight patients recruited by Timmers and colleagues (62) failed to mount a significant ventilatory response to hypoxia under normocapnic conditions. However, two of these eight subjects responded to hypoxia with an increase in ventilation when the stimulus was applied in hypercapnic conditions (62), confirming similar observations in carotid body-resected patients with a history of bronchial asthma (59).

Importantly, patients recruited in both of these studies (16, 62) underwent carotid body resection to remove tumors caused by a mutation of the gene encoding for succinate dehydrogenase complex subunit D (SDHD, part of Cytochrome b₅₈₈ of the mitochondrial respiratory chain complex II). Healthy age- and sex-matched volunteers were used as controls. However, the hypoxic ventilatory response in healthy humans may vary by a factor of 20 due to various genetic factors (60, 67). Tumor-free carriers of the SDHD mutation (with intact carotid bodies) display a reduced isocapnic hypoxic ventilatory response at the very low end of the “normal” range (60). Could the reduction of the hypoxic ventilatory response in conditions of SDHD mutation reflect impaired central respiratory oxygen sensitivity in these individuals, a possibility consistent with the proposed key role of the mitochondrial hypoxia-sensitive mechanism (discussed below)?

Thus studies of the hypoxic ventilatory response following bilateral resection of the carotid bodies in humans are somewhat hampered by either the underlying chronic lung disease or genetic factors which may alter chemoreflex function. Therefore, it will remain unknown whether the hypoxic ventilatory response in young and healthy humans would recover over time following bilateral carotid body denervation in a manner similar to that documented in many experimental animals.

Central Respiratory Oxygen Sensor

Growing evidence from in vitro and in vivo experimental animal models suggests that the central stimulatory effect of hypoxia on breathing is mediated by a mechanism that operates within the brain stem regions harboring neuronal circuits responsible for the generation of respiratory rhythm and pattern. Recent data suggest that this oxygen-sensitive mechanism is not neuronal in nature (1, 38).

The hypoxic ventilatory response that remains in rats following denervation of peripheral oxygen chemoreceptors is abolished when the ventral regions of the medulla oblongata are transduced to express a potent ectonucleotidase, transmembrane prostatic acid phosphatase (TMPAP) (1). Catalytic activity of TMPAP effectively blocks purinergic signaling mechanisms by preventing vesicular accumulation of ATP and by promoting rapid degradation of extracellular purines (39, 68, 71). Mammalian brain cells that are not electrically excitable but show calcium excitability, glial cells (astrocytes and microglia), release ATP as the major signaling molecule through which they communicate with neighboring glia, neurons, and other brain cell types. Thus glial cells were hypothesized to be responsible for sensing brain hypoxia.

Consistent with this hypothesis, the hypoxic ventilatory response of both carotid body intact and peripherally chemodenervated rats was significantly reduced when brain stem astrocytes were selectively targeted to express the light chain of tetanus toxin (TeLC) (1). TeLC cleaves certain SNARE proteins required for vesicular docking and fusion, effectively blocking vesicular release of gliotransmitters (including ATP) by astrocytes (1, 8). Interestingly, bilateral TeLC expression in brain stem astrocytes (in carotid body intact rats) and bilateral carotid body ablation (in rats expressing control transgene) resulted in quantitatively similar reductions in the magnitude of the hypoxic ventilatory response (see Figure 7 in Ref. 1).

Subsequent studies demonstrated that astrocytes are able to sense physiologically-relevant decreases in brain parenchymal Po₂ a few mmHg below normal brain oxygenation level (1). The astroglial signaling cascade triggered by hypoxia involves inhibition of mitochondrial respiration, increased formation of free radicals, activation of phospholipase C, IP₃ receptors, release of Ca²⁺ from the intracellular stores, and enhanced vesicular release of ATP (Fig. 1). Real-time biosensor measurements in vivo and in vitro (23) demonstrated hypoxia-induced release of ATP within and in close proximity to the brain stem respiratory networks, including the rhythm-generating circuits of the pre-Bötzinger complex that are sensitive to, and potently excited by, ATP (20, 21, 23, 29, 30, 35, 36).

Fig. 1.

Hypothesized cellular mechanisms underlying central oxygen sensitivity. The astroglial signaling cascade triggered by hypoxia involves inhibition of mitochondrial respiration, increased formation of reactive oxygen species (ROS), lipid peroxidation, activation of phospholipase C (PLC), IP₃ receptors, release of Ca²⁺ from the intracellular stores and enhanced vesicular release of ATP. Hypoxia may also alter opening probability of connexin (Cx) hemichannels permeable to ATP and lactate. Released ATP acts in an autocrine and paracrine manner, spreads astroglial Ca²⁺ signals within the neuropil, and enhances respiratory and sympathetic activities via excitation of the respiratory rhythm generating circuits of the pre-Bötzinger complex (preBötC) and sympathoexcitatory (presympathetic) neurons of the ventrolateral medulla oblongata.

An important role of reactive oxygen species in this signaling pathway is supported by a number of earlier reports that demonstrated the stimulatory effect of CNS hyperoxia (which would be expected to increase free radical production) on ventilation in awake, decerebrate, or anesthetized experimental animals with denervated carotid bodies (see 12 and references therein). The bimodal shape of the relationship between lung ventilation and the arterial Po₂ (in conditions of isolated separately perfused normoxic carotid body) correlates very well with the relationship between reactive oxygen species formation and intracellular Po₂ (Fig. 2).

Fig. 2.

A: relationship between lung ventilation and the arterial Po₂ in 11 awake goats whose carotid bodies were isolated and separately perfused with normocapnic-normoxic blood (schematic adapted from the data reported in Ref. 12). B: bimodal distribution of reactive oxygen species formation as a function of intracellular Po₂. [Schematic adapted from Clanton (7) with permission.]

Thus, a significant body of evidence collectively suggests that the central stimulatory effect of hypoxia on breathing is mediated by the actions of ATP released by activated astrocytes intermingled with the neuronal networks responsible for the generation of respiratory activity. Hypoxia-induced excitation of brain stem presympathetic neurons also appears to be indirect, mediated by prior release and actions of ATP and lactate (38).

Summary

The existence of a brain hypoxia sensor capable of stimulating breathing is not universally accepted. Many investigators question the physiological relevance of the anoxia/hypoxia-evoked responses observed in the in vitro preparations of the neonatal rodent brain stem. In addition, human data suggest that functional respiratory oxygen sensitivity is an exclusive function of the carotid body chemoreceptors, as no recovery of the normocapnic hypoxic ventilatory response has been reported from studies conducted in individuals with bilaterally resected carotid bodies. Yet, significant recovery of the hypoxic ventilatory response after peripheral chemoreceptor denervation has been observed in all experimental animals studied so far (ponies, goats, dogs, cats, rats, and mice). That the respiratory responses in experimental animals with denervated carotid bodies are markedly suppressed by general anesthesia suggests that the operation of the physiological central respiratory oxygen-sensitive mechanism is readily inhibited by anesthetic agents.

Astrocytes are able to detect various sensory modalities (22, 27, 28, 32, 41, 64, 65) and appear to be functionally specialized as CNS oxygen sensors tuned to detect physiological decreases in brain oxygenation (1). When activated in hypoxic conditions, brain stem astrocytes release ATP, which stimulates the respiratory neuronal circuits leading to increases in lung ventilation. The variability in the degree to which the hypoxic ventilatory response recovers after peripheral chemodenervation may reflect species differences in the level of endogenous ectonucleotidase activity of the brain stem parenchyma (70). Future research may include studies of the ectonucleotidase activity in the human brain, and, in experimental animal models, analysis of the association between changes in the brain stem ectonucleotidase activity and recovery of the hypoxic ventilatory response over time after denervation of the carotid body chemoreceptors. It is plausible that in humans brain stem ectonucleotidase activity is relatively high, resulting in rapid degradation of ATP released during brain hypoxia and enhanced production of adenosine (which may contribute to the hypoxic ventilatory decline). We propose that during systemic hypoxia in human subjects with denervated carotid bodies, the central respiratory oxygen-sensitive mechanism effectively maintains lung ventilation by counteracting the hypoxia-induced depression of breathing, although in the absence of the peripheral chemoreceptor input it is not sufficiently potent to evoke increases in ventilation above the normoxic baseline.

GRANTS

Results of the authors’ experimental studies described in this review article were obtained with generous support of The Wellcome Trust and British Heart Foundation. A. V. Gourine is a Wellcome Trust Senior Research Fellow (Refs. 095064 and 200893).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.V.G. drafted manuscript; A.V.G. and G.D.F. edited and revised manuscript; A.V.G. and G.D.F. approved final version of manuscript.