Acute oxygen sensing by the carotid body: a rattlebag of molecular mechanisms

Abstract

The molecular underpinnings of the oxygen sensitivity of the carotid body Type I cells are becoming better defined as research begins to identify potential interactions between previously separate theories. Nevertheless, the field of oxygen chemoreception still presents the general observer with a bewildering array of potential signalling pathways by which a fall in oxygen levels might initiate Type I cell activation. The purpose of this brief review is to address five of the current oxygen sensing hypotheses: the lactate–Olfr 78 hypothesis of oxygen chemotransduction; the role mitochondrial ATP and metabolism may have in chemotransduction; the AMP‐activated protein kinase hypothesis and its current role in oxygen sensing by the carotid body; reactive oxygen species as key transducers in the oxygen sensing cascade; and the mechanisms by which H₂S, reactive oxygen species and haem oxygenase may integrate to provide a rapid oxygen sensing transduction system. Over the previous 15 years several lines of research into acute hypoxic chemotransduction mechanisms have focused on the integration of mitochondrial and membrane signalling. This review places an emphasis on the subplasmalemmal–mitochondrial microenvironment in Type I cells and how theories of acute oxygen sensing are increasingly dependent on functional interaction within this microenvironment.

Introduction

It is a truth universally acknowledged that a carotid body Type I cell in possession of potassium channels must be in need of an oxygen sensing mechanism. Or two, or three. Or indeed five as are discussed in this article.

This concise review was written following numerous requests by students at the XXth conference of the International Society for Arterial Chemoreception (ISAC), Johns Hopkins University, Baltimore, USA. There was a desire to see the main theories of acute oxygen sensing not simply presented but also critiqued. However, with a word limit of 2000 it was extremely difficult to do this in any great detail. Consequently, this short article has focused on only five acute oxygen sensing mechanisms and has attempted to find cellular requirements that they share. Additionally, for each example, an attempt has been made to present some contradictory data or ideas that challenge the hypothesis. It is the authors’ hope that this method of presenting the data will spark discussion and it is our intent to use the review as an aid to team‐based learning in the classroom where groups of students will champion and defend individual hypotheses. An exercise like this will force students to read far beyond the papers cited here and will hopefully help with the development of their critical thinking.

When one compares the various recent theories of acute oxygen sensing in the carotid body, it quickly becomes apparent that interactions between mitochondria and the plasma membrane underpin the majority of these concepts. The main goal of the ISAC presentation was to highlight the fact that over the last 15 years acute oxygen sensing hypotheses have moved away from the polarizing ‘membrane’ versus ‘mitochondrial’ theories of the previous decades and have begun to embrace more integrated ‘mitochondria–membrane’ models.

To be concise, the response of a carotid body Type I cell to hypoxia involves the sensing of the change in oxygen being transduced into cellular depolarization and neurotransmitter release. The ‘membrane’ hypothesis of oxygen sensing in the carotid body posited that electrophysiological events in the Type I cell plasma membrane were critical for chemotransduction (Lopez‐Barneo et al. 1988). This required that oxygen sensors, transducers and effectors were all focused on evoking Type I cell plasma membrane depolarization. By contrast the ‘mitochondrial’ hypothesis of oxygen sensing stated that the mitochondria were the oxygen sensors and that calcium release from the mitochondria triggered neurosecretion (Duchen & Biscoe, 1992a,b).

Interestingly, several membrane‐delimited hypotheses (sensor, transducer and effector all in the plasma membrane) have emerged over the years (Acker et al. 1989; Cross et al. 1990; Williams et al. 2004; Lee et al. 2006) and some have been contradicted by knockout mouse data (Roy et al. 2000; He et al. 2002; Ortega‐Sáenz et al. 2006). However, one remaining conundrum exists within the membrane‐delimited hypothesis: it has been shown that the inhibition of potassium channels by hypoxia is critical for the Type I cellular depolarization to occur (Lopez‐Barneo et al. 1988), but are the channels intrinsically oxygen sensitive or are they dependent on a separate sensor? A couple of laboratories have indicated that proposed ‘oxygen‐sensitive’ channels are indeed intrinsically sensitive (Lewis et al. 2001; Riesco‐Fagundo et al. 2001) while others disagree (Wyatt & Peers 1995; Buckler et al. 2000; Lee et al. 2006; Kim, 2013). Most of the mechanisms discussed in this review focus on oxygen sensors that are not in the membrane but which ultimately engage with ion channels located in the membrane. If the ion channels themselves were intrinsically oxygen sensitive then none of these mechanisms would need to exist and the mystery of acute oxygen sensing by the carotid body would have been solved. It is for this reason the ISAC presentation focused on the concept of integrated ‘membrane–mitochondrial’ hypotheses.

Finally, the presentation touched on the unique subcellular microenvironment that exists in carotid body Type I cells. Others have stressed the importance of this microenvironment (Varas et al. 2007; Gao et al. 2017b) and when one considers the potential interactions between mitochondria and plasma membrane, the physical arrangement of cellular organelles may be critical for a cell to transduce a hypoxic signal. Indeed as can be seen in Fig. 1A in Wyatt et al. (2007), Type I cells have large nuclei that will restrict mitochondria to a thin ribbon of cytoplasm just beneath the plasma membrane, an ideal location for them to interact with ion channels.

With all these things in mind we should address the following five theories of acute oxygen sensing. They are presented in no particular order or ranking, although the most recent hypothesis is presented first.

Lactate–Olfr 78 hypothesis

The work of Chang et al. (2015) brought to light a novel mechanism by which hypoxia might excite carotid body Type I cells. Their hypothesis rests on one of the cornerstones of hypoxic chemotransduction theories, that the Type I cell mitochondria are uniquely sensitive to hypoxia and that hypoxia inhibits electron transport. The unusual sensitivity of Type I cell mitochondria to hypoxia is key to most of these oxygen sensing hypotheses and has been recently and elegantly demonstrated by Buckler & Turner (2015) and Holmes et al. (2016).

Chang's work demonstrated the presence of an olfactory receptor gene, Olfr78, in mouse Type I cells and showed that Olfr78 knockout mice lose their acute hypoxic ventilatory response but not their response to hypercapnia. Further experiments indicated that the carotid bodies of these knockout animals appeared normal but had reduced responses to hypoxia. Postulating that during hypoxia lactate is known to accumulate and will activate Olfr 78, their hypoxic chemotransduction hypothesis is as follows. Hypoxia inhibits electron transport in Type I cell mitochondria ultimately causing glucose to form lactate. Lactate accumulates and activates the G‐protein‐coupled receptor Olfr 78, which by an as yet undetermined mechanism causes the Type I cells to depolarize and release neurotransmitter.

There are several issues with this hypothesis that need to be addressed. For example, can the lactate hypothesis explain why isolated, perfused Type I cells respond in milliseconds to a hypoxic challenge (Buckler, 1997)? In these in vitro preparations, any lactate that might accumulate extracellularly will be swept away before it can activate its receptor. Additionally can a mechanism dependent on a G‐protein‐coupled receptor be fast enough to produce the cellular hypoxic response? However, there are many extremely fast G‐protein‐coupled reactions (Hardie & Raghu, 2001) and it would be useful to know the kinetics of the Olf 78 receptor; indeed Jang & Hyeon (2017) have already begun to address this. Furthermore, if we turn our attention to the ligand for the receptor, there are now questions regarding the sensitivity of Olfr 78 to lactate. At least two laboratories have reported that Olfr 78 was only sensitive to lactate outside the physiological range (Aisenberg et al. 2016; Zhou et al. 2016). And finally, Gao et al. (2017a) have demonstrated high expression of Olfr 78 in non‐oxygen sensing tissue. Can this be an argument against it playing a role in oxygen sensing? We don't believe so. For Olfr 78 to be involved in oxygen sensing it needs to be partnered with mitochondria that are unusually sensitive to hypoxia and placed in a microenvironment where it can rapidly signal to appropriate ion channels. If Type I cells can provide all these requirements, then Olfr 78 might be involved in oxygen sensing, but if any of these requirements are missing, Olfr 78 could be present in the tissue but uncoupled from cellular excitation, over the appropriate range of hypoxia.

Even if the lactate hypothesis cannot underpin the rapid response to hypoxia, it might sustain any response to hypoxia throughout the time course of the hypoxic exposure. Future experiments addressing how Olfr 78 might couple to excitation and whether its kinetics are adequate for producing the hypoxic response are likely already being conducted and this reviewer, for one, eagerly awaits the results.

Mitochondrial ATP production or metabolic hypothesis

This hypothesis is perhaps the most straightforward of the acute oxygen sensing hypotheses presented here. It is dependent on the unique sensitivity of the Type I cell mitochondria to hypoxia and builds on the pioneering work of Duchen & Biscoe (1992a,b) amongst many others. Research conducted in Dr Keith Buckler's laboratory has served to indicate the importance of mitochondria in the sensing of hypoxia and how hypoxic inhibition of mitochondrial function might couple to potassium channel inhibition and cellular excitation (Williams & Buckler, 2004; Varas et al. 2007). Furthermore, his collaborative research with Dr Prem Kumar is now beginning to tease apart why the Type I cell mitochondria might be so sensitive to changes in oxygen levels (Holmes et al. 2016).

It has been demonstrated that potassium channels that set the resting membrane potential of Type I cells can be inhibited by mitochondrial inhibition (Wyatt & Buckler, 2004) leading to excitation and hence neurotransmitter release. Inhibition of electron transport is likely to impact ATP production, and Varas et al. (2007) demonstrated that the Twik‐related acid sensing K⁺ (TASK) channels are strongly activated by MgATP. Thus, under normoxic conditions with high levels of MgATP, the TASK channels are activated and membrane potential is hyperpolarized. When hypoxia inhibits ATP production, the TASK channels close and the cell depolarizes, leading to neurotransmitter release. A highly organized subcellular microenvironment where mitochondria are close to the plasma membrane will likely facilitate this mechanism. Critically, the missing piece for this oxygen sensing hypothesis seems to be the current absence of a Mg‐nucleotide sensor that could bind MgATP and couple to TASK channel regulation. However, is this the only missing piece for this theory? Before we move to discuss our next hypothesis, it is important to consider one final set of data. Ortega‐Saenz et al. (2010) demonstrated that when both TASK1 and TASK3 channels were knocked out, mice carotid bodies retained the ability to respond to hypoxia. This suggested that either there are other channels regulated by ATP in Type I cells, in addition to TASK channels, or that plasticity had occurred in the cells following knockout and a different oxygen sensing mechanism had become dominant under these conditions. To date what had happened to oxygen sensing following TASK channel knockout in these Type I cells has not been defined.

AMP‐activated protein kinase hypothesis

Conceived by Dr A. Mark Evans in the early 2000s (Evans, 2004) the AMP‐activated protein kinase (AMPK) hypothesis of oxygen sensing is also dependent on mitochondria that are uniquely sensitive to hypoxia and also a tightly organized subcellular microenvironment. In brief, the premise of the hypothesis is that hypoxia will inhibit mitochondrial electron transport causing a fall in ATP production. This fall in ATP production will be mitigated by adenylate kinase, which converts ADP to ATP and AMP, and consequently AMP levels will rise in the cell. This rise in AMP levels will activate AMPK, which will phosphorylate membrane ion channels causing depolarization and neurotransmitter release.

Initial pharmacological experiments looked promising with AMPK activators mimicking the effects of hypoxia while an AMPK antagonist reversed the effects of hypoxia (Evans et al. 2005; Wyatt et al. 2007). Indeed, further studies demonstrated that AMPK could phosphorylate potassium channels known to be regulated by hypoxia in Type I cells (Ross et al. 2011). However, some studies cast doubt on the selectivity of the pharmacological agents used (Emerling et al. 2007; Guigas et al. 2007) and others failed to repeat the observed effects of the compounds (Kim et al. 2014). To address these contradictory data, a range of approaches using knockout mice were used to test the AMPK hypothesis. Eventually, using a careful and complex knockout strategy, Dr Evans's laboratory successfully deleted both catalytic subunits of AMPK in tissue containing tyrosine hydroxylase (critically including the carotid body). Then, using unrestrained plethysmography techniques and electrophysiology, they and their collaborators demonstrated that this knockout did indeed alter the acute hypoxic ventilatory response but that it was not via an effect on the oxygen sensing ability of the carotid body (Evans et al. 2016; Mahmoud et al. 2016).

Clearly, AMPK plays a key role in the generation of the acute hypoxic ventilatory response, but does it play any role in the oxygen sensitivity of the carotid body? The answer to that question depends on whether one views knockout experiments as definitive proof of the role of a protein. Compensation for the knockout can occur, and in a system like the carotid body where there appears to be a fistful of alternative oxygen sensing mechanisms that are primed to step in when one fails, compensation seems likely. Further studies where the inactivation/deletion of AMPK occurs more rapidly than seen in conventional targeted knockout may reveal its role in normal carotid body physiology.

Mitochondrial complex I signalling hypothesis

The breadth and depth of literature available on reactive oxygen species and their role in oxygen sensing by the carotid body is daunting for any student of acute oxygen sensing (Gonzalez et al. 2010). However, within the past two decades research from the laboratory of Dr Jose Lopez‐Barneo has worked towards a comprehensive mechanism of acute oxygen sensing that critically relies on the production of reactive oxygen species by Type I cell mitochondria, and has focused specifically on complex I of the electron transport chain (Ortega‐Saenz et al. 2003). Recently these mechanisms were described in detail and indicated that hypoxic inhibition of mitochondrial electron transport resulted in increased production of reactive oxygen species and reduced pyridine nucleotides from complex I of the mitochondria (Fernandez‐Aguera et al. 2015; Gao et al. 2017b). This increase in reactive oxygen species production was hypothesized to change the redox status of membrane ion channels and thus initiate excitation. Furthermore, this hypothesis placed much emphasis on there being subplasmalemmal microdomains between mitochondria and membrane ion channels that would facilitate modulation of carotid body ion channel activity. Thus, this theory of acute oxygen sensing fits well in this review where a focus has been placed on integrating mitochondrial oxygen sensing and signal transduction in subcellular microenvironments. Reactive oxygen species as signalling molecules also fulfil another requirement for chemotransduction, namely they are promiscuous and will affect a range of targets. This is critical as ion channels that are inhibited by hypoxia vary across species, and evidence suggests that when one oxygen‐sensitive channel is removed others become oxygen sensitive (Ortega‐Sáenz et al. 2010); reactive oxygen species would allow for this to occur.

As compelling as this hypothesis is, there are data that contradict the role of reactive oxygen species in acute oxygen sensing. However, much of this work has made use of oxidizing/reducing agents and reactive oxygen scavenging agents (Wyatt & Buckler, 2004; Gonzalez et al. 2010; Papreck et al. 2012) and it could be argued that the gross extracellular application of these compounds may not accurately mimic/inhibit the rapid signalling that occurs in subcellular microenvironments. Controversially, although the direct application of redox modifying agents to the intracellular face of an inside‐out membrane patch should recapitulate a microenvironment and allow for ROS modified ion channel activity, there are data indicating that this is not the case (Papreck et al. 2012).

Hydrogen sulphide hypothesis

The concept that hydrogen sulphide (H₂S) could play a key role in oxygen sensing was originally proposed and demonstrated in the laboratory of Dr Ken Olson, initially in vascular tissue (Olson et al. 2006; Olson, 2008) and then in trout gill chemoreceptors (Olson et al. 2008). Building upon and extending these studies, Dr Nanduri Prabhakar's group then knocked out the enzymes responsible for generating H₂S in mice and observed that this interfered with oxygen sensing in the carotid bodies (Peng et al. 2010). By integrating the research of several groups (Williams et al. 2004), the hydrogen sulphide hypothesis of acute oxygen sensing has developed considerably over the past decade. The current model proposed by Dr Prabhakar is as follows.

It is now believed that there is a complex relationship between two gaseous transmitters in the Type I cells. During normoxia the enzyme haem oxygenase‐2 generates carbon monoxide (CO), and hypoxia has been demonstrated to inhibit the production of this gasotransmitter. The decrease in CO production leads to a decrease in protein kinase G‐dependent phosphorylation of an enzyme (cystothionine‐γ‐lyase), which leads to an increased production of H₂S (Yuan et al. 2015) and subsequent potassium channel inhibition (Li et al. 2010; Telezhkin et al. 2010; Buckler, 2012), cellular excitation and neurotransmitter release. Interestingly, however, it has been demonstrated that haem oxygenase‐2 knockout mice have normal carotid body responsiveness to hypoxia (Ortega‐Saenz et al. 2006). This again suggests that either plasticity has occurred following knockout or that haem oxygenase is not involved in the sensing of acute changes in oxygen by the carotid body.

Upon first inspection this hypothesis does not appear to integrate mitochondria and plasma membrane at all. However, it is extremely important to note that H₂S can inhibit mitochondrial electron transport (Buckler, 2012). Thus, if the production of H₂S is localized in cellular subdomains by mitochondria, or even within mitochondria, it is conceivable that hypoxia will cause a rise in H₂S of a large enough magnitude that electron transport will be inhibited and potentially all the other signalling pathways described in this review could be activated. To the general reader this must seem extremely exciting. However, as with the majority of hypotheses of acute oxygen sensing there are now conflicting data to temper any enthusiasm. Recent studies have indicated that an increase in H₂S does not modulate ion channels (Kim et al. 2015) and knockout of the enzymes that produce H₂S has no effect on acute oxygen sensing in the carotid body (Wang et al. 2017). These contradictory data will need to be addressed before we can fully understand the role of H₂S in acute oxygen sensing.

Conclusion

Because of the limits on space, this review has focused primarily on acute oxygen sensing mechanisms that integrate mitochondrial oxygen sensing with transduction mechanisms which will interact with membrane ion channels (see Abstract Figure). However, this does mean that other theories of oxygen sensing were not addressed and we would therefore like to encourage readers to familiarize themselves with the carbon monoxide work that has come out of Dr Paul Kemp's laboratory (Williams et al. 2004, 2008; Kemp & Telezhkin, 2014) and also the NOX‐4 research of Dr J. W. Park's research group (Lee et al. 2006; Park et al. 2009). Sadly, these are not the most egregious omissions in this review. The reasons why the carotid body mitochondria are so sensitive to reductions of oxygen have barely been addressed at all and would make an excellent stand‐alone article. Research tackling this issue will be presented at the ISAC meeting in Lisbon, 2020 hopefully, but in the meantime, readers might enjoy two articles which address this unique oxygen sensitivity, Holmes et al. (2016) and Gao et al. (2017a).

Which theory of acute oxygen sensing is closest to what is happening physiologically? As yet it is too difficult to say. Will a unified theory be discovered or confirmed within the next decade or so? That is a possibility but I think it wise that we consider the words of the great science philosopher Karl Popper before pinning our hopes on one theory of acute oxygen sensing: ‘whenever a theory appears to you as the only possible one, take this as a sign that you have neither understood the theory nor the problem which it was intended to solve.’ Novel hypotheses that address how hypoxia is sensed and transduced by carotid body Type I cells are constantly being tested and modified. This is a healthy situation and training the next generation of scientists to think critically about all these theories and to develop their own is how the mystery of acute oxygen sensing in the carotid body will ultimately be resolved.

Additional information

Competing interests

Both authors declare no competing interests.

Author contributions

Both authors wrote sections of the manuscript. All authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

No funding supported this article.

Biographies

Ryan Rakoczy is a PhD candidate studying in the Wyatt laboratory and is one of the most recent recipients of a Heymans, DeCastro, Neil Prize, awarded at the XXth meeting of The International Society of Arterial Chemoreception, Johns Hopkins University, Baltimore, USA.

Christopher Wyatt is an Associate Professor in the Neuroscience, Cell Biology and Physiology Department and Associate Dean of The College of Science and Mathematics at Wright State University, Dayton, Ohio. He began his research into hypoxic chemotransduction by carotid body Type I cells in 1991 while working in the laboratory of Chris Peers. Subsequent postdoctoral studies on the carotid body were conducted in Keith Buckler's laboratory in Oxford and A. Mark Evans's laboratory in St Andrews, Scotland.