Carotid body chemoreceptors: physiology, pathology, and implications for health and disease

Abstract

The carotid body (CB) is the main peripheral chemoreceptor for arterial respiratory gases O₂ and CO₂ and pH, eliciting reflex ventilatory, cardiovascular, and humoral responses to maintain homeostasis. This review examines the fundamental biology underlying CB chemoreceptor function, its contribution to integrated physiological responses, and its role in maintaining health and potentiating disease. Emphasis is placed on 1) transduction mechanisms in chemoreceptor (type I) cells, highlighting the role played by the hypoxic inhibition of O₂-dependent K⁺ channels and mitochondrial oxidative metabolism, and their modification by intracellular molecules and other ion channels; 2) synaptic mechanisms linking type I cells and petrosal nerve terminals, focusing on the role played by the main proposed transmitters and modulatory gases, and the participation of glial cells in regulation of the chemosensory process; 3) integrated reflex responses to CB activation, emphasizing that the responses differ dramatically depending on the nature of the physiological, pathological, or environmental challenges, and the interactions of the chemoreceptor reflex with other reflexes in optimizing oxygen delivery to the tissues; and 4) the contribution of enhanced CB chemosensory discharge to autonomic and cardiorespiratory pathophysiology in obstructive sleep apnea, congestive heart failure, resistant hypertension, and metabolic diseases and how modulation of enhanced CB reactivity in disease conditions may attenuate pathophysiology.

CLINICAL HIGHLIGHTS.

Over recent years, the carotid body (CB) has been implicated in pathological consequences associated with obstructive sleep apnea, congestive heart failure, resistant hypertension, and metabolic diseases. An enhanced CB chemosensory activity has been linked with sympathetic hyperactivity, a feature common to these disease conditions. Resection of the CB normalizes sympathetic activity and cardiorespiratory alterations in preclinical models, highlighting the potential role played by the CB in the progression of sympathetic-related diseases. These findings support the concept of modulating the CB to regulate CB chemosensory discharge as a useful strategy for attenuation of pathophysiological consequences of these diseases.

1. INTRODUCTION

The carotid and aortic chemoreceptors are unique organs capable of sensing the partial pressures of O₂ (Po₂) and CO₂ (Pco₂) and pH in the arterial blood, initiating rapid systemic responses to overcome severe hypoxemia, with a temporal course compatible with enhancing survival. The carotid body (CB) located in the bifurcation of the common carotid artery is the main peripheral chemoreceptor in terms of its contribution to the cardiorespiratory, autonomic, and humoral responses to hypoxemia (1–4). The CB (also called ganglion minutum, carotid corpuscule, carotid ganglion, glomus caroticum, and carotid gland) was first described by Taube and Von Haller in the eighteenth century, although its physiological role remained largely unidentified until the twentieth century. Heinrich Hering in 1925 (5) reported that mechanical or electric stimulation of the carotid sinus, a dilated area at the base of the internal carotid artery in the carotid bifurcation, evoked hypotension and bradycardia, suggesting the reflexogenic barosensory function of this region. Fernando De Castro, a disciple of Santiago Ramón y Cajal in Spain, based upon his elegant histological studies, proposed that the CB was able to detect chemical changes in arterial blood (see Ref. 6 for historical review). At the same time, Jean-François Heymans and his son Corneille Heymans (7), working at the University of Ghent in Belgium, proposed that the carotid sinus region holds two sensory modalities: the carotid sinus, which responds to mechanical deformation and the CB, which is stimulated by changes in arterial blood chemical composition. From 1929 to 1933, Corneille Heymans and Fernando De Castro had the opportunity to meet several times. In 1938 Corneille Heymans was awarded the Nobel Prize in Physiology or Medicine “for the discovery of the role played by the sinus and aortic mechanisms in the regulation of respiration,” which revealed the physiological role of the CB, proposed in part by De Castro.

Acute hypoxemic or hypercapnic-acidic stimulation of the CB increases chemosensory discharge, triggering reflex responses including hyperventilation, increased vagal cardiac activity (eliciting bradycardia), and heightened sympathetic activity to the muscle, splanchnic, and renal vascular beds [eliciting acute increases in blood pressure (BP)]. In normoxia, the CB tonically stimulates ventilation, since the inhibition of chemosensory discharge with 100% O₂ (Dejours test) reduces baseline ventilation (8). The structural unit of the CB is the glomoid formed by chemoreceptor (type I or glomus) cells, which are organized around the capillaries and are surrounded by cells of glial origin (type II cells). Type I cells are innervated by axons of primary sensory neurons with somata in the petrosal ganglia (PG) that project to the nucleus of the solitary tract (NTS). The most widely accepted hypothesis for stimulus transduction holds that O₂ and CO₂-H⁺ sensing occurs in the type I cells and is coupled to Ca²⁺-dependent secretion of one or more excitatory transmitters that increase discharge of sensory neurons. Several hypotheses have been advanced to explain the O₂ sensing process. However, despite major advances in understanding the O₂ sensing process in the CB over the last 50 yr, the precise nature of the molecular mechanisms remains incompletely understood. Furthermore, the identity of the excitatory transmitter(s) between type I cells and nerve endings of the petrosal neurons is still debated.

Over recent years, the CB has been implicated in autonomic and cardiovascular responses, pathophysiology, and consequences associated with obstructive sleep apnea (OSA), congestive heart failure (CHF), resistant hypertension, and metabolic diseases. A potentiated CB chemosensory response to hypoxemia has been linked to sympathetic hyperactivity, a feature common to each of these diseases. Ablation of the CBs normalizes sympathetic activity and cardiorespiratory alterations in preclinical models, highlighting the potential role played by the CB in the progression of sympathetic-related diseases. These findings support the concept that abolishment or modulation of CB chemosensory hyperactivity may be useful in restoring normal function in humans. Accordingly, the main goal of this review is to provide a balanced discussion of the arguments for and against the proposed hypotheses of O₂ transduction, review synaptic mechanisms in the CB, and discuss emerging data on the contributions of nonrespiratory functions of the CB to sympathetic-related human diseases.

We discuss in detail 1) the O₂, CO₂-H⁺ transduction mechanisms, underscoring the role of hypoxic inhibition of O₂-dependent K⁺ channels, and the relationship with mitochondrial oxidative metabolism and associated reactive oxygen species (ROS), as well as the role played by activation of other channels in type I cells (for recent reviews see Refs. 9–19); 2) the role played by the main proposed transmitters [ATP, acetylcholine (ACh), dopamine (DA)] and the contribution of gaseous molecules [nitric oxide (NO), carbon monoxide (CO), and H₂S] as modulators of chemoreception (for recent reviews see Refs. 11, 20–25); and 3) recent advances in understanding the role played by the CB in health and disease, discussing the integrated reflex responses to CB chemosensory excitation, emphasizing that the responses differ dramatically depending on the nature of the physiological, pathological, or environmental challenge and the interactions of the chemoreceptor reflex with other reflexes in optimizing oxygen delivery; 4) evidence showing that an abnormally enhanced CB chemosensory discharge increases central sympathetic outflow, thereby contributing to cardiovascular pathophysiology in obstructive sleep apnea, congestive heart failure, resistant hypertension, and metabolic disease; and finally 5) the molecular mechanisms underlying enhanced CB chemosensory discharge, such as oxidative stress, inflammation, increased levels of endothelin 1 (ET-1) and angiotensin II (ANG II), along with reduced NO levels in the CB (for recent reviews see Refs. 1, 4, 26–30) and their relationship to disease conditions (20, 31–39).

2. CAROTID BODY MORPHOLOGY AND SYNAPTIC ORGANIZATION

The fundamental morphological unit in the CB is the glomoid, which consists of clusters of type I cells synaptically apposed to the terminals of petrosal ganglion (PG) neurons and surrounded by sustentacular (type II) cells of glial origin (FIGURE 1). The type I cells are considered the primary transduction loci for O₂ and CO₂-H⁺, whereas the petrosal ganglia neurons convey the chemosensory discharge to the NTS in the brain stem. In the petrosal ganglion are located the somata of the chemosensory neurons that innervate the CB and the gustatory neurons that innervate the posterior third of the tongue through the glossopharyngeal nerve. In addition, in the petrosal ganglion are located the somata of the barosensory neurons that innervate the carotid sinus through the carotid sinus nerve (CSN) and the mechanosensory neurons that innervate the tongue, pharynx, and trachea through the glossopharyngeal nerve (40). Several studies of the morphology of human and animal CB are available (41–44). In humans the mean dimensions of the CB are ∼2–3 mm and a mean weight of ∼18–20 mg (41, 44). In laboratory mammals, the CB is smaller, measuring ∼1 mm, weighing 1–3 mg, and containing ∼12,000 chemoreceptor cells in the rat (45) and 60,000 in the cat (46). Despite the relatively small size, the CB has the highest blood flow (1.0–2.0 L/min/100 g) reported for any organ (47, 48). The blood flow to the CB is supplied by one or two arteries, which generate an intricate vascular network comprised of small arteries, fenestrated capillaries, and venules. The blood vessels receive sympathetic innervation from the superior cervical ganglion through the ganglioglomerular nerves. In addition, parasympathetic neurons located within the CB and in the glossopharyngeal nerve also innervate the blood vessels. Most of the postganglionic sympathetic fibers innervate the glomus capillaries, although some fibers innervate type I cells (49).

FIGURE 1.

Interactions between cellular elements in the carotid body. A: carotid bifurcation and location of the carotid body. B: schematic representation of a carotid body glomerulus. Chemosensory and chemoproliferative synapses are indicated. C: schematic representation of the tripartite synapse. AR, adenosine receptor; DR, dopamine receptor; Panx-1, pannexin-1 channel; P2XR, purinergic 2X ionotropic receptor; P2YR, purinergic 2Y metabotropic receptor. (Reprinted from Ref. 3, with permission from Annual Review of Physiology.)

The current model for CB oxygen chemoreception proposes that hypoxia induces inhibition of voltage-gated and voltage-independent K⁺ currents, leading to type I cell depolarization, entry of Ca²⁺ through L-type Ca²⁺ channels, and the subsequent release of one or more excitatory transmitters, which increases the discharge of nerve endings of the petrosal neurons (1–4, 40, 50). Among several molecules present in type I cells, acetylcholine and adenosine triphosphate fulfill most of the criteria to be considered the excitatory transmitters between the chemoreceptor cells and petrosal nerve endings (23, 40). However, other molecules such as dopamine, histamine, nitric oxide (NO), and endothelin-1 modulate the chemosensory process by producing tonic actions on CB blood vessels or direct effects on chemoreceptor cells (27, 51, 52). Type I cells, derived from the neural crest, contain several molecules such as catecholamines, acetylcholine, adenosine nucleotides, and peptides, which are candidates to be the excitatory transmitters in the junctions between type I cells and the nerve terminals of petrosal neurons (23, 27, 40). A high degree of colocalization for amine-synthesizing enzymes [tyrosine hydroxylase (TH), dopamine-β-hydroxylase, choline acetyltransferase (ChAT)] and substance P and met-encephalin has been found in type I cells (2, 27). Thus, it is likely that type I cells store and release more than one excitatory transmitter in response to natural stimuli.

In the CB, the synaptic relationships between type I and type II cells is complex, resembling a tripartite synapse. Nurse and colleagues (23, 53) proposed that hypoxic stimulation of type I cells induces the release of ATP via pannexin-1 channels in type II cells, contributing to synaptic integration and cross talk among type I cells, type II cells, and the PG nerve endings, like in a tripartite synapse. Type II cells express purinergic P2Y₂ receptors and ATP-permeable pannexin channels, as well as muscarinic and angiotensin type 1 (AT₁) receptors (23, 53–57). Opening of pannexin-1 channels in rat type II cells promotes further release of ATP and amplification of the ATP signal via the mechanism of ATP-induced ATP release. Type II cell-derived ATP can be broken down to adenosine by 5′-ectonucleotidases present in the CB (58), which may activate A2A receptors, leading to a rise in intracellular Ca²⁺. Reciprocal cross talk between type I and type II cells may contribute to chemosensory processes in the CB via purinergic signaling pathways. Synaptic plasticity mechanisms during intermittent hypoxia or chronic sustained hypoxia might contribute to modifications in neurotransmitter signaling at the tripartite synapse (FIGURE 1).

For a long time, the type II cells were considered as unexcitable supportive cells that engulf type I cells. However, an elegant series of experiments performed by López-Barneo and colleagues revealed the crucial role for type II cells in CB plasticity induced by sustained hypoxia. They found in adult rodents that the CB is a neurogenic niche, which contains a population of multipotent adult neural crest-derived stem cells (3, 59, 60). These stem cells are quiescent in normoxia but in response to sustained hypoxia proliferate and differentiate into new type I cells as well as smooth muscle and endothelial cells. Their pioneering experiments showed that CB stem cells were sustentacular type II cells that lose their glial phenotype [glial fibrillary acidic protein (GFAP)+] and are transformed into nestin+ proliferating progenitors (59–61). The CB stem cells are not directly activated by hypoxia, and their activation depends on the type I cells. The term “chemoproliferative synapse” was proposed by López-Barneo and colleagues to designate the frequent synapses established between the type I cells and glial stem cells in the CB. They found that endothelin-1 (ET-1) release from type I cells in response to sustained hypoxia induces proliferation of type II cells, which express ET receptors (Ref. 60; FIGURE 1).

3. STIMULUS TRANSDUCTION

The participation of CB type I cells in the generation of afferent chemosensory activity has been widely discussed since it was initially postulated (62, 63). Although several lines of evidence strongly support their role in the generation of chemosensory activity, it was not until the last quarter of the last century that the properties of type I cells were clearly established. The electrophysiological properties of type I cells were first studied with sharp conventional intracellular electrodes that produced uncertain data on the electrical properties of type I cells, owing principally to the small size of these cells (64–67). However, use of the whole cell patch-clamp recording technique allowed a clearer characterization of their electrophysiological properties and the responses of type I cells to carotid body natural stimuli (FIGURE 2).

FIGURE 2.

Excitability and transduction mechanism components in type I cells. Type I cells are endowed with several voltage-dependent Na⁺ (VDNC), Ca²⁺ (VDCC), and K⁺ channels that render these cells excitable and capable of firing action potentials. Several K⁺ channels [oxygen sensitive (${\mathrm{K}}_{{\mathrm{O}}_2}$), large conductance (BK), TASK] have been shown to be directly or indirectly modulated by chemosensory stimuli (Po₂, pH, Pco₂). Thus, ${\mathrm{K}}_{{\mathrm{O}}_2}$ and BK channels appear to be directly modulated by Po₂, as indicated by the reduction of channel activity by hypoxia in excised patches. On the other hand, TASK channel activity is reduced directly by extracellular pH and indirectly by hypoxia, through reductions of intracellular ATP or the activation of AMP-activated protein kinase (AMPK) by the reduction of mitochondrial ATP production or [ATP]-to-[ADP] [P_i] ratio. This reduction of K⁺ currents leads to cell depolarization, recruitment of voltage-dependent Na⁺ and Ca²⁺ channels that further depolarize the cell, and increased intracellular Ca²⁺ concentration ([Ca²⁺]_i), a key element for the exocytotic release of the neurotransmitter(s) (NT). The latter one(s) will generate or modify afferent activity by activation of nerve terminals but also could modify type I cell properties acting on presynaptic receptors. Modification of type I cells by transmitter receptors may increase activity of phospholipase C (PLC) and protein kinase C (PKC) that can modify the activity of both transduction and voltage-dependent channels, thus enabling temporal modulation of sensory responses. Extracellular pH can also act on acid-sensing ion channels (ASICs) present in type I cells, depolarizing them and contributing to the sensory response. Finally, both transient receptor potential (TRP) channels and nonspecific cationic channels may also participate in the responses to hypoxia, both directly or indirectly through reactive oxygen species (ROS) produced during hypoxic challenges. ACh, acetylcholine; mAChR, muscarinic cholinergic receptor.

3.1. Electrophysiological Properties of Type I and Type II Cells

The first demonstrations that rabbit and rat carotid body type I cells are excitable were published toward the end of the last century, taking advantage of the whole cell patch-clamp recording technique. These recordings showed that isolated type I cells present both inwardly directed Na⁺ and Ca²⁺ currents and a slowly activating outwardly directed K⁺ current evoked from negative holding potentials (44, 68–72). Na⁺ current is fast activating and inactivating, activated at potentials above −50 mV, peaks near 0 mV, reverses at positive holding potentials, and is blocked by tetrodotoxin (TTX) (68, 72), but is absent in embryonic type I cells (69). Ca²⁺ currents are evoked by depolarizations over −40 mV, peak near +10 mV, reverse at positive holding potentials, and are slowly inactivating and blocked by cobalt (Co²⁺), methoxyverapamil (D600), nickel (Ni²⁺), and high concentrations of magnesium (Mg²⁺) (68, 69, 72, 73). Further electrophysiological characterization indicated that both dihydropyridine-sensitive L-type (74, 75) and ω-conotoxin GVIA-sensitive N-type (75, 76) voltage-dependent Ca²⁺ channels contribute to the whole cell Ca²⁺ current in type I cells. Although conflicting results have been reported on the presence and function of T-type Ca²⁺ channels in type I cells (72, 75, 77, 78), the expression of mRNA coding for T-type channels has been reported in the rat (77) and mouse (12, 79) CB. The K⁺ current is outwardly rectifying and maximal near +20 mV (68). The outward current is reduced in the presence of extracellular Ca²⁺ (68, 69), suggesting the existence of a Ca²⁺-dependent component in the K⁺ current. Experiments in isolated membrane patches further showed that the outward K⁺ current comprises at least three different K⁺ channels (80). Thus, a large-conductance (∼210 pS) Ca²⁺-dependent K⁺ channel (BK, MaxiK), a small-conductance (∼16 pS) channel that is active in the absence of Ca²⁺ in the medium, and a medium-conductance (∼42 pS) Ca²⁺-independent K⁺ channel (80) are present in type I carotid body cells. Additionally, an ATP-sensitive K⁺ current has also been described in rat type I cells (81). Type I cells are recognized as excitable and capable of firing action potentials. Action potential currents can be recorded under resting conditions (68), and type I cells can fire action potentials induced by depolarizing current pulses (68, 82). Moreover, in whole cell patch-clamp recordings, switching from voltage to current clamp can evoke trains of action potentials in type I cells (71, 72). On the other hand, type II cells are devoid of inwardly directed currents, showing only a slowly activating outwardly directed current (68, 72, 83). This current is blocked in the presence of tetraethylammonium (TEA) in the medium or the replacement of K⁺ by cesium (Cs⁺) in the pipette (68), indicating that sustentacular cells are not excitable and present mainly K⁺ currents, as do most glial cells (84–87).

3.2. O₂-Dependent K⁺ Channels

In isolated rabbit type I cells, the inwardly directed Na⁺ and Ca²⁺ voltage-dependent currents are unaffected by hypoxia. Conversely, the slowly activating K⁺ current that is blocked by TEA is reversibly reduced during a hypoxic challenge (69–71), with a maximal reduction of ∼50% near 80 mmHg in the 10–150 mmHg Po₂ range (71). Similarly, recordings showed that the delay to the generation of the first spike is reduced in a hypoxic environment, suggesting that firing frequency can be increased during hypoxia (71). Recordings from outside-out patches of rabbit type I cells showed the activity of single channels that are opened by depolarizing pulses. The ensemble current obtained during multiple equivalent depolarizations resembles the whole cell current (80, 88) and is blocked by TEA (88), and reduced by hypoxia (80, 88,89). The current-voltage (I-V) curve of these single inside-out channel recordings showed that the unitary conductance in asymmetric K⁺ concentrations is ∼20 pS and that this value remains largely unmodified during hypoxia (80, 88). The open probability of this O₂-sensitive channel, termed ${\mathrm{K}}_{{\mathrm{O}}_2}$, is voltage dependent (80, 89) and reduced by hypoxia in a dose-dependent manner (88). These data indicate that ${\mathrm{K}}_{{\mathrm{O}}_2}$ can be modulated by Po₂ outside the cellular context, indicating that the channel open probability per se is modulated by O₂ or by a mechanism ligated tightly enough to the channel to be active in cell membrane patches (80, 88). However, it has been shown that cellular components, such as mitochondria, can also remain attached to the membrane in excised patches (90) and modulate K⁺ channel activity (90) and are suggested to modulate CB type I cell responses to hypoxia (3, 91). Thus, the O₂ dependence of the ${\mathrm{K}}_{{\mathrm{O}}_2}$ channel in excised patches does not rule out the possible involvement of organelles or molecules located near the channel.

In addition to the responses to O₂, whole cell outward K⁺ currents recorded from type 1 cells were also reduced when glucose was omitted from the medium (92). The reduction in current was seen over the entire voltage range tested and was independent of intracellular ATP (4 mM ATP in the pipette), the ATP-regulated K⁺ channel (not blocked by glibenclamide), and the large-conductance Ca²⁺-activated MaxiK channel [not blocked by iberiotoxin (Ibtx)]. The reduction of glucose was not accompanied by a reduction of input resistance, indicating that background potassium channels appear not to be involved in the response (92). The low-glucose challenges were accompanied by type I cell membrane depolarization (44, 93), increases in intracellular Ca²⁺ (44, 92), and increased neurotransmitter release (44, 92–95). Thus, the K⁺ current reduced by low glucose has characteristics similar to those described for ${\mathrm{K}}_{{\mathrm{O}}_2}$, suggesting that both stimuli act on the same K⁺ current and depolarize type I cells. However, the secretory response induced by low glucose could still be evoked in the presence of TEA and iberiotoxin, and the depolarization was abolished in the absence of extracellular Na⁺ (95), indicating that an increase in inward Na⁺ current is also involved in the type I cell response to low extracellular glucose (95). Thus, although the K⁺ current reduced by low glucose has characteristics similar to those described for ${\mathrm{K}}_{{\mathrm{O}}_2}$, glucose-induced depolarization of type I cells may result from interaction of several membrane currents.

3.3. Background K⁺ Channels (TASK)

In isolated neonatal and adult rat and rabbit type I cells, hypoxia decreased the activity of channels recorded in cell-attached patches (96–99). Acidification of the medium produced a similar reduction of the cell-attached patch activity, in a concentration-dependent manner (98), and a reduction of the whole cell current recorded in cell-attached mode (96). The cell-attached mode whole cell current showed almost no outward rectification and no time-dependent activation or inactivation and was dose-dependently reduced by hypoxia (96, 100). The lack of time-dependent activation and inactivation of the current is like that observed in mouse background TASK channels expressed in Xenopus oocytes (101). Similarly, methanandamide, a TASK-1 (K2P3.1) and TASK-3 (K2P9.1) channel blocker (102), reduced the whole cell current as well as the O₂-sensitive current (103). This O₂-sensitive current was only slightly modified by TEA (10 mM) and 4-aminopyridine (4-AP; 5 mM) and was absent when K⁺ was omitted from the medium (100, 104). The current-voltage relationship constructed under different K⁺ concentrations indicates that the main ion contributing to this O₂-sensitive current was K⁺ (98, 100, 104–106). This O₂-sensitive current was largely voltage independent, partially blocked by barium (Ba²⁺), zinc (Zn²⁺), bupivacaine (200 µM), and quinidine (1 mM), and enhanced by halothane (1.5%) (96, 100), pharmacological hallmarks of TASK-1 and TASK-3 channels (101, 107, 108).

The expression of channel protein and mRNA coding for TASK-1 channels has been demonstrated in mice (12, 79, 109) and neonatal (96, 110) and adult (99, 111, 112) rats, although TASK-3 has also been described in the latter (99, 111, 112), in mice (12, 109), and in rabbits (97). The expression reported for both TASK-1 and TASK-3 mRNAs is one of the highest in CB cells compared with hypoxia-nonsensitive cells (12, 79). TASK-1 channels expressed in human embryonic kidney 293 (HEK293) cells present whole cell currents that were inhibited by both acidification and hypoxia in a dose-dependent manner (113). Recording of type I channel activity in cell-attached mode showed at least two conductance levels, of approximately 14 pS (96) and 32 pS (98), and a main conductance level of 33 pS (105), similar to the currents recorded in HeLa cells transfected with TASK-1 or TASK-3 cDNA (98). These data suggest that the main entity present in type I cells is TASK-1,3 heteromers (98, 105). TASK-1- and TASK-1,3-deficient mice present a reduced hypoxia-induced and hypercapnia-induced increase in the carotid sinus nerve (CSN) afferent discharge (114). Moreover, TASK-1-deficient mice present an increased CSN discharge frequency in normoxia (114). In TASK-1,3-deficient mice there is a significant reduction in type I cell membrane conductance and a more depolarized resting membrane potential, whereas no changes were observed in the electrical properties of TASK-1-deficient type I cells (109).

Although TASK channels have been demonstrated to be directly modulated by extracellular pH (101, 115) and indirectly modulated by Po₂ (104, 116, 117), intracellular signaling molecules may also modulate K⁺ background channel activity (118–120). Activation of isolated type I cell muscarinic cholinergic receptors (mAchRs) by muscarine (50 μM) or methacholine (100 μM) reduces background channel activity and increases intracellular Ca²⁺ concentration ([Ca²⁺]_i) (121), effects that are blocked by atropine (1 μM). Modification of intracellular Ca²⁺ with thapsigargin (1 μM) or BAPTA (100 μM) has no effect on the channel activity (121), suggesting the absence of direct or secondary actions of Ca²⁺ on the channel activity. A protein kinase C (PKC) activator (1-oleoyl-2-acetylglycerol, 20 μM) decreases channel activity, whereas inhibition of PKC activity with calphostin C or chelerythrine increases the activity of the channels (121), suggesting that increased phospholipase C (PLC) activity and diacylglycerol levels mediate the responses to activation of mAChRs (121). Expression of TASK-1 channels in Xenopus oocytes results in a time- and voltage-independent background K⁺ current that is reduced by the application of GTPγS or by the activation of concomitantly expressed ANG II receptors or M1 type mAChRs (122). The reduction of background K⁺ current is partially blocked by a PLC inhibitor (U-73122, 2.5 mM) and is not affected by inositol 1,4,5-trisphosphate (InsP₃), Ca²⁺, or a PKC inhibitor (122), suggesting that in this case the activation of M1 mAChR acts by a different intracellular mediator.

Early studies suggested that mitochondria and the AMP-to-ATP ratio are important in the transduction process in type I cells. In recent years it has been shown that subtle modifications of this ratio impact the activity of AMP-activated protein kinase (AMPK) that leads to protein phosphorylation (123). AMPK α1-subunit has been localized in type I cell membranes (124, 125). In isolated postnatal rat type I cells the application of AMPK activators 5-aminoimidazole-4-carboxamide riboside (AICAR; 1 mM) and phenformin (10 mM) increases intracellular Ca²⁺, an effect that is partially blocked by Cd²⁺, is absent in zero extracellular Ca²⁺ (124, 125), and is largely attenuated by an AMPK inhibitor (compound C; 40 μM). Thus, AMPK activation results in the activation of voltage-dependent Ca²⁺ channels and not in Ca²⁺ release from intracellular stores, as was observed in pulmonary arterial smooth muscle (124). AICAR application to isolated type I cells produces depolarization and a 33% reduction of a Ba²⁺-sensitive, voltage-independent background current (125). However, in postnatal rat type I cells a lack of effect of either AICAR (1 mM) or A-769662 (100 μM) on background channel activity, resting membrane potential, or intracellular Ca²⁺ has been reported (126). Moreover, in heterologously expressed rat TASK-1, TASK-3, and TASK-1,3 channels, conductance is not modified by AICAR (127). However, in the same type of preparation both two-pore-domain background potassium channel 1 and 2 (TREK-1 and TREK-2) activity were inhibited by AICAR (127). Although TREK channels are not directly sensitive to hypoxia (128, 129) and TASK-1,3 heterotrimeric channels appear to be the most important component of the background K⁺ current (98, 105), their inhibition by AMPK may also participate in the generation of O₂ sensitivity in type I cells. It is noteworthy that AICAR increased the afferent activity recorded from isolated carotid bodies in vitro, an effect that was reversibly blocked by the elimination of Ca²⁺ from the medium (124, 125). However, this effect cannot be directly attributed to type I cells because of the complex relationship between PG nerve terminals and the carotid body parenchymal cells as well as the functional contacts between carotid body parenchymal cells (130). Despite this evidence, the lack of selectivity and specificity of the pharmacological tools used to assess AMPK participation in the transduction process has been pointed out (131). Moreover, conditional deletion of AMPK-α1 and AMPK-α2 genes in mouse TH-synthesizing cells, that comprise type I cells, reduced the hypoxic ventilatory response with respect to control animals but without affecting the CSN responses to hypoxia (132). Thus, although AMPK activity can modulate the type I cell response to hypoxia, it is not essential for generation of the afferent activity.

TASK-1, TASK-3, and TASK-1,3 heterodimeric channels expressed in HeLa cells are activated by high concentrations of hydrogen peroxide (H₂O₂) (>20–30 mM) when recorded in inside-out patches but had no effect on outside-out patches, suggesting an intracellular effect (133). The effect of H₂O₂ in inside-out patches is not affected by glutathione (1 mM) or dithiothreitol (DTT; 1 mM), suggesting that oxidation of –SH groups does not underlie increased channel activity (133). On the other hand, increased superoxide production, with xanthine/xanthine oxidase (XO) in the medium, has no effect on the activity of the excised patches, regardless of the membrane orientation (133). A similar effect was observed in inside-out patches obtained from type I cells, suggesting that reactive oxygen species (ROS) are not direct mediators of the reduced activity of background channels brought about by hypoxic challenges.

3.4. Large-Conductance Ca²⁺- and Voltage-Activated K⁺ Channels

Rat carotid body type I whole cell voltage-dependent outwardly directed K⁺ currents were activated above −20 mV and presented a local maximal near +20 mV (68, 69, 134, 135). The magnitude of the K⁺ outward current was enhanced when extracellular Ca²⁺ was increased (10 µM) or in the presence of the Ca²⁺-channel agonist Bay K 8644 (136), suggesting the existence of a Ca²⁺-dependent component in the K⁺ current (I_KCa). A part of the whole cell K⁺ current was inhibited by hypoxia in a reversible manner, an effect that was completely abolished in the presence of Cd²⁺ (134, 137) or high Mg²⁺ and low Ca²⁺ (134, 135). Moreover, nifedipine and D600 (134–137) reduced the magnitude of the outward current and completely obliterated the hypoxia-induced reduction in the outward current, indicating that I_KCa underlies O₂ sensitivity in type I cells. Charybdotoxin (ChTX), a large-conductance Ca²⁺-dependent K⁺ channel (BK; K_Ca1.1) blocker that reduced the magnitude of the outward current and completely obliterated the hypoxia-induced response (138), also depolarized and reduced the activity of channels recorded from type I cells in a manner similar to hypoxia (139). Thus, depolarization by reduced activity is the mechanism by which BK channels contribute to the activation of type I cells. Cell-attached single-channel recordings from rabbit type I cells have shown that hypoxia reversibly reduces the activity of recorded patches, and that the reduced current reverses at 0 mV in symmetric K⁺, with a core conductance of 137 pS (140). Large-conductance (≈187 pS in symmetric K⁺) channel activity, recorded from excised patches of rat type I cells (139, 141), was reduced by hypoxia in a Ca²⁺-dependent manner (141), displacing the voltage dependence to more positive potentials by almost 27 mV. The large channel conductance and its voltage and Ca²⁺ dependence (139) indicate that BK channels underlie the hypoxia-induced reduction of I_KCa in type I cells. Similarly, the acidification-induced reduction of I_KCa was blocked by D600 (136). It was previously shown that the inwardly directed Ca²⁺ current was unaffected by hypoxia (70, 137, 140), suggesting that I_KCa carried through BK channels is reduced by both hypoxia and acidification, two parameters that depolarize type I cells and increase afferent chemosensory activity and ventilation. AMPK activation selectively reduces hypoxia-induced reduction in K⁺ currents, including BK (83, 124, 142). Outwardly directed K⁺ currents were inhibited by AMPK activator A-769662 and by low-Ca²⁺ (high Mg²⁺) medium, with no additive effect of low Ca²⁺ over AMPK activation, suggesting that AMPK inhibits BK channel activity (142). Similarly, intracellular dialysis of type I cells with a recombinant thiophosphorylated AMPK reduced the outwardly directed K⁺ currents, whereas a similar application of an inactive form of AMPK had no effect (142). These data suggest that AMPK that is expressed in the rat CB cells (142) inhibits the activity of BK channels and could lead to type I cell depolarization.

It has been suggested that type I cell O₂ sensitivity may depend on the cellular redox state. However, the hypoxia-induced reduction of I_KCa was not mimicked by a reducing agent such as dithiothreitol (DTT). Conversely, DTT reversibly increased the channel activity and the open probability whereas the oxidizing agent 2,2′-dithiopyridine (DTDP) had no effect on channel activity (141). Cysteine residue modification has been shown to modify Ca²⁺ binding and the coupling of voltage sensor activation and Ca²⁺ binding to channel opening (143) as well as the rundown of BK channels (144). Thus, redox modification of cysteine residues of the channel COOH terminus appears not to be involved in the hypoxia-induced modification of BK channel activity in type I cells. An O₂ dependence has been reported for BK channels present in mitochondria of astrocytes and a glioma cell line (LN 229), but in these latter cases the channel open probability was increased by hypoxia and sodium dithionite (1 mM), without modification of the channel conductance (276 pS; Refs. 145, 146). Because the O₂-induced modification of BK channel activity can be recorded in excised patches, it has been suggested that the effect is restricted to the membrane domain (141). Chronic systemic CO hyperpolarizes rat pulmonary artery myocytes and increases the depolarizing effect of acute charybdotoxin, without affecting channel conductance (147), suggesting that CO increases BK activity. Similarly, in heterologously expressed BK channels, composed of either the pore-forming subunit alone or the pore-forming and auxiliary subunits, CO reversibly increased open probability in excised patches, even in the absence of Ca²⁺, without modifying unitary current amplitude (148, 149). Mutation of intracellular BK channel histidine residues decreased or even abolished the activating effect of CO (148), sites that are necessary for heme group binding and inhibition of BK activity by heme groups (150, 151). Thus, CO appears to increase BK activity by removing heme inhibition of channel activity. In type 1 cells, CO blocked the inhibitory effects of hypoxia on BK channels, increasing the channel open probability above basal levels (141). Heme oxygenase-2 (HO-2), which cleaves heme and finally releases CO, colocalizes and coimmunoprecipitates with BK channels in HEK cells expressing BK, and hypoxia reduced the channel activity whereas a CO donor increased it (149). The knockdown of HO-2 in these transfected cells abolished the hypoxia-induced reduction of BK activity but not the activation produced by CO (149). Thus, the effect of hypoxia on BK channel activity could be understood as an increased inhibitory action of a heme moiety that is released by O₂ reduction and that is experimentally sequestered by CO, eliminating the inhibition and even increasing channel activity over normoxia levels by a larger reduction of free heme groups. It has been reported that mouse type I cell hypoxia-induced catecholamine secretory response is unaffected in HO-2-deficient animals, although BK channels with normal electrical properties but reduced density were present in the cells (152). Thus, BK channels may participate in the transduction of hypoxia, but they appear not to be central to the hypoxic response.

In turtle cerebral neurons it has been reported that anoxia decreased an I_KCa and reduced the open probability of single channels in cell-attached mode, an effect that was lost in excised patches (153). The current was activated at about −46 mV, was outwardly directed, depended on internal [Ca²⁺], had a conductance of 223 pS, and was blocked by either TEA or iberiotoxin, indicating that BK channels carried the recorded I_KCa. The PKC inhibitor chelerythrine completely blocked the effects of anoxia, whereas the PKC activator phorbol 12-myristate 13-acetate (PMA) produced a reduction of the channel activity with a similar time course and maximal effect than anoxia (153). A similar action of PMA on BK channel activity has been reported in smooth muscle cells of guinea pigs (154) and rats (155). Thus, although the effect of hypoxia on BK channel activity appears not to be dependent on the cellular redox state and to be directly linked to a membrane-related mechanism, it appears that modifications of PKC activity may also modulate BK activity and the transduction process.

3.5. Acid-Sensing Ion Channels

Although most of the proposed transduction mechanisms in type I cell chemoreceptors rely on the modification of one or several K⁺ currents, the participation of an inward Na⁺ current has also been suggested to be involved in the generation of the acid-induced responses of the carotid body. Patch-clamp current recordings of rat isolated type I cells showed that extracellular acidification (pH range: 7–4) produced an inward, rapidly activated, and desensitizing current in a population (40–50%) of type I cells (156, 157). Similarly, voltage recordings in the same condition showed transient depolarizations in about half of the recorded cells and sustained depolarizations in all the recorded cells (156, 157). The amplitude of the current was pH dependent, with a threshold pH around 7.0 and a half-maximal response at pH 6.3. This transient current was reduced by almost 90% in the absence of extracellular Na⁺ and in a dose-dependent reversible manner by amiloride, an acid-sensing ion channel (ASIC) blocker, to a similar extent. Similarly, the transient depolarization was also blocked by amiloride, leaving the sustained depolarization largely unaffected (156, 157). Moreover, increasing extracellular lactate, which decreases extracellular Ca²⁺ concentration ([Ca²⁺]_e) (557), or decreasing [Ca²⁺]_e, which increases ASIC activation (159–161), increases both the magnitude of the inward current and the percentage of responding cells at the same pH (156). This transient inward current was not affected by a transient receptor potential vanilloid (TRPV)1, an ASIC1a, or a BK channel blocker. Nevertheless, the sustained inward current was still present after these two blocking maneuvers. Blocking BK, or voltage-dependent K channels, had no effect on the sustained depolarization, whereas low intracellular K⁺ and a TASK-1 channel blocker partially inhibited the acid-induced depolarization. All the preceding data suggest that ASICs participate in the generation of the transient current and depolarization and at least in part in the acid-induced response of type I cells (156). Acid stimulation induces increases in intracellular calcium concentration ([Ca²⁺]_i) in the majority (∼80%) of isolated rat type I cells, although the magnitude of the response varied between cells (162). In mice, [Ca²⁺]_i increases induced by acidification were reduced, and increased in ASIC3-null (KO) and transgenic animals, in comparison to wild-type (WT) animals. Increases in [Ca²⁺]_i induced by metabolic inhibition [sodium cyanide (NaCN)] were reduced in transgenic animals with respect to KO animals, with no significant differences with WT mice (162). Although the population of type I cells that increased their [Ca²⁺]_i in response to pH 6 stimulation was significantly reduced, the percentage of cells increasing their [Ca²⁺]_i in response to NaCN was preserved in ASIC3 KO mice. Similarly, [Ca²⁺]_i increases evoked by hypoxia were preserved in ASIC3 KO mice (162). Evaluation of rat type I cell mRNA for ASIC subunits showed that several subunits were expressed, with an ASIC3/ASIC1b/ASIC2a relation of 1/0.6/0.1; ASIC1a and ASIC2b mRNA subunits were scarcely expressed in the carotid body (156, 157). Confocal immunofluorescence showed high staining of ASIC1 and ASIC3 subunits, with low ASIC2a staining in

most type I cell profiles. ASIC1 and ASIC3 subunits were colocalized in type I cell profiles presenting tyrosine hydroxylase (TH) immunofluorescence (157), with little colocalization of ASIC1 with ASIC2a (156). Thus, ASIC subunits are normally expressed in type I cells, with high colocalization of ASIC1 and ASIC3 subunits. Intercellular spaces were devoid of immunofluorescence, suggesting that CB sustentacular cells and petrosal ganglion nerve terminals are devoid of ASIC subunits. Hypercapnia-induced increases in ventilatory frequency, tidal volume, and minute ventilation in ASIC1−/−, ASIC2−/−, and ASIC3−/− mice are similar to WT animals, whereas isocapnic hypoxic responses are only slightly reduced in ASIC2−/− mice (163), although the responses from chemosensory afferents were not recorded. This result suggests that individual ASIC channel deletion does not affect acid-induced responses, although the exact location of the effect (peripheral vs. central) has yet to be determined. On the other hand, rat petrosal ganglion neurons express mRNA and protein for ASIC1a, ASIC1b, ASIC2a, ASIC2b, and ASIC3 subunits (164). These last results indicate that petrosal ganglion neurons may also participate in the acid-induced increases in chemosensory activity. It is noteworthy that chemosensory basal activity and responses to low Po₂ from rat carotid body in vitro preparations are insensitive to the ASIC antagonist A-317567, whereas hypoxic responses are reduced (∼50%) by the blocker after 10 days of chronic hypobaric hypoxia (164). Similarly, rat isolated nodose ganglion (NG) neurons in tissue culture are relatively insensitive to acid stimulation, whereas acid-induced responses can be recorded in cocultures of NG neurons and CB cells (165). Thus, neurons connecting with CB cells appear to be relatively insensitive to acidification, whereas inflammation induced by chronic hypoxia increases ASIC expression and its participation in the generation of chemosensory afferent activity even to hypoxic stimuli (164). The preceding data indicate that type I cells are sensitive to extracellular acidification through, at least in part, the activation of ASICs. This activation appears to be by a direct action of H⁺ on the channel extracellular domain (166). However, ASIC activity can also be modified by the oxidative status of the cells (166). Thus, in CHO-K1 rat cells transiently expressing ASIC1a, reducing agents increase the acid-induced current, and oxidizing agents have the opposite effect (166). In trigeminal and dorsal ganglion neurons, the reducing agents DTT and glutathione (GSH) increased acid-induced inward currents, without modifying the pH-response curve, whereas the oxidizing agent DTNB had the opposite effect (167). In vascular smooth muscle cells (VSMCs), DTT reversibly increased both peak and sustained acid-induced inward currents as well as the number of responsive VSMCs (168), effects blocked by the ASIC inhibitor amiloride. Application of the oxidizing agent DTNB reversed the potentiating effects of DTT in VSMCs (168). Inhibition of NADPH oxidase (NOX) with apocyanin potentiates acid-induced currents and the number of responsive VSMCs, whereas NOX activity inhibition with NSC23766 increases the potentiating effects of DTT (168). Similarly, inhibition of xanthine oxidase (XO) increases the acid-induced responses and the number of responsive VSMCs (168). These findings suggest that endogenous NOX and XO activities may suppress ASIC-like channel activity in cerebral VSMCs. All the preceding data indicate that ASICs participate in the generation of the fast component of the acid sensing response by type I cells (156, 157) and can be also involved in hypoxia-induced responses after chronic inflammation (164). Moreover, modification of the redox status of type I cells, as during intermittent hypoxia (169, 170), may modulate the response of ASICs (166–168) and the generation of chemosensory afferent activity.

3.6. Transient Receptor Potential Channels

Although their relevance to the generation of chemosensory activity has not yet been established, canonical transient receptor potential (TRPC) channel proteins have been immunohistochemically identified both in rat petrosal ganglion (PG) neurons projecting to the carotid body and in type I cells (171) as well as their mRNA in type I cells (12). TRPC3, TRPC4, TRPC5, TRPC6, and TRPC7 were present in all PG neurons; TRPC1 colocalized with TH immunolabeling (171), thus accounting for all neurons projecting to the carotid body (172). Similarly, in situ hybridization and mRNA determination have shown that PG neurons express TRPM3 (melastin), TRMP6, TRPM7, TRPM8, TRPA1 (ankyrin), and TRPV1 (vanilloid) transcripts (173, 174), with no coexpression of TRPM8 and TRPA1 in the same neurons, whereas some neurons coexpress TRPA1 and TRPV1 transcripts (173). On the other hand, TRPC1, TRPC3, TRPC4, TRPC5, TRPC6, and TRPC7 transcripts were present in type I cells, with TRPC1 and TRPC7 localized intracellularly whereas the rest of the channels were also localized to the membrane; no colocalization studies were performed (171). Similarly, single-cell RNA sequencing (RNA-seq) indicates that Trpc (Trpc3, Trpc5, Trpc6, Trpc7) and Trpv (Trpv2, Trpv4) genes are differentially expressed in mice CB cells (12, 79), with TRPC5 being the most highly expressed (12). TRPM3, TRPM4, TRPM6, and TRPM7 transcripts have also been described in the carotid body (174, 175). Although there is no evidence of participation of TRPC channels in the type I cell transduction mechanism (176), activation of these channels is brought about by intracellular cascade mechanisms, including PLC and PKC activation (177). Thus, activation of these cascades through receptor/growth factor receptors may modulate these TRPC channels and regulate type I cell basal membrane conductance, modifying the receptor currents and cellular response. Moreover, TRPC channels may operate as Ca²⁺ entry points in store-operated calcium entry (SOCE), in conjunction with STIM1 and Orai1 proteins, participating in the regulation of [Ca²⁺]_i (178). Leptin increases the CSN frequency discharge response to hypoxia, an effect inhibited by TRP channel blockers [SKF96365 or 2-aminoethoxydiphenylborane (2-APB)] (174). Additionally, application of a broad-spectrum TRP

channel blocker (2-APB) significantly reduced isolated type I cell [Ca²⁺]_i increases induced by hypoxia, anoxia, NaHS, NaCN, and carbonyl cyanide p-trifluormethoxyphenylhydrazone (FCCP) (176). However, [Ca²⁺]_i increases induced by high K⁺, as well as the NADH-to-NAD⁺ ratio and the mitochondrial membrane potential changes induced by anoxia, were not modified by 2-APB, suggesting a direct involvement of TRP channels in [Ca²⁺]_i increases brought about by the transduction mechanism in type I cells. Ventilatory responses in mice to mild hypoxia (13%) were reduced in a dose-dependent manner by a selective TRPA1 antagonist (HC-030031), whereas hyperventilation in response to severe hypoxia (7%) was only reduced by the largest dose; hyperoxia (100% O₂)- and hypercapnia(5% CO₂)-induced ventilatory responses were largely unaffected by TRPA1 antagonism (179). Hypoxic and hyperoxic media, acting on TRPA1 channels, modify [Ca²⁺]_i in nodose and dorsal root ganglion neurons and transfected HEK 293 cells, with minimal [Ca²⁺]_i value at normoxia and increasing both in hyperoxic and hypoxic conditions (180). Although O₂ appears to act directly on TRPA1 channels in hyperoxia, hypoxia-induced increases of TRPA1 currents were absent in excised inside-out patches, suggesting that hypoxic activation of the channel is mediated by diffusible intracellular components (180). In Trpa1-knockout mice, hypoxia-induced vagal afferent activity increases and the number of hypoxia-responsive neurons is reduced with respect to control (wild type) values (180). These data suggest that vagal afferents may also be involved in the ventilatory responses to mild hypoxia through the activation of TRPA1 channels. Similarly, anandamide (an endocannabinoid, endovanilloid, and TASK channel blocker) applied to the rat carotid body increases ventilation and the CSN frequency discharge and responses to hypoxia, effects blocked by the TRPV1 antagonists AMG-9810 and casazepine; the effects of anandamide were completely absent in TRPV1-null mice (181). Moreover, increases in afferent CSN activity induced by increases in temperature from 37°C to 39°C were enhanced by anandamide and restored to basal values when AMG-9810 was also present in the medium. Immunohistochemical and RT-PCR determinations showed that TRPV1 appeared not to be expressed in the carotid body parenchymal cells but are present in PG neuron terminals in the carotid body (181). Thus, activation of PG terminals through TRPV1 receptors may also modulate the responses of type I cells through synaptic contacts established between these two elements.

3.7. Sodium and Nonselective Cation Channels

Although most of the proposed transduction mechanisms in type I cells implicate the direct or indirect inhibition of one or more K⁺ channels, it has been proposed that activation of nonselective cation channels and a Na⁺-carrying depolarizing current may also participate in the generation of the receptor current (175, 182). During hypoxic stimulation and after reduction of K⁺ channel activity, a 20-pS conductance could be recorded in cell-attached patches, the activity of which was inversely related to the hypoxic level (175, 182). An L-type Ca²⁺ channel agonist reversibly activated the 12-pS channel in cell-attached patches, an effect that was dependent on the presence of extracellular Ca²⁺. Moreover, increasing Ca²⁺ in inside-out patches increased channel activity, whereas increasing intracellular Ca²⁺ buffer capability (applying BAPTA intracellularly) abolished activation of this channel (175), indicating that the channel activation depended on Ca²⁺ entry through voltage-dependent Ca²⁺ channels (VDCCs), although release from intracellular stores also increased channel activity (182). Neither hypoxia nor REDOX state regulators were capable of modifying the 20-pS channel activity (175), suggesting that this channel is not directly involved in the transduction process. The 20-pS channel appeared to be largely voltage insensitive in the ±100 mV range, and both Na⁺ and K⁺ ions permeate through the channel (175). Thus, although not sensitive to hypoxia, this 20-pS channel could modulate the receptor potential after the depolarization and activation of VDCCs.

4. GENERATION OF AFFERENT ACTIVITY: SYNAPTIC MECHANISMS

The CB is a compound or secondary receptor, where the CB sensory (type I) cells are connected by the nerve terminals of PG neurons that are in contact with them. Even though electrical synapses have been postulated to participate in the communication (130), chemical synapses are the most commonly accepted mechanism involved in the activation of afferent fibers in the CSN to generate afferent activity. There is a multitude of transmitter molecules present in the CB that may act at both the presynaptic and postsynaptic levels, including the sustentacular cells. Significant progress has been made in the comprehension of which transmitter molecules are relevant for the generation of afferent activity, although a clear understanding of the peculiarities that can be brought about by different physiological and pathophysiological conditions has not been reached. An additional difficulty arises from the different experimental models and approaches that make it difficult to reconcile all the experimental data. We focus here on the evidence of the major transmitter molecules present in the CB that are recognized in most species to participate in the generation of afferent activity.

4.1. Acetylcholine and Its Receptors

Acetylcholine (ACh) was one of the first transmitters to be postulated to act in the CB. Applications of ACh or cholinomimetics increased ventilation (183) and the afferent activity of the CSN (184, 185), whereas nicotinic antagonists had the opposite effect. Moreover, ACh is released from the cat (94, 186, 187), dog (188), and human (189) CBs during hypoxic stimulation (94, 186–189), suggesting that communication between type I cells and PG nerve terminals is at least partially dependent on ACh release. ACh can be detected in the CB, and its content is not modified by the removal of sensory or vasomotor (superior cervical ganglion) innervation (190). The enzyme that synthetizes ACh, choline acetyltransferase (ChAT), has been immunolocalized in rat (191, 192), cat (193, 194), and rabbit (194) CB cells, and the vesicular ACh transporter (VAChT) has been immunodetected in rat type I cells (192, 195). A high-affinity, sodium-dependent choline uptake mechanism has also been reported in the cat CB (194). Similarly, ChAT and VAChT have been also immunodetected in the aortic bodies of rats (196, 197), suggesting that cholinergic machinery is present in all type I cells independently of anatomical location. Despite these previous results, in situ hybridization immunohistochemistry has failed to reveal the presence of ChAT (198, 199) and VAChT (198–200) mRNA in the adult rat CB.

In isolated rat and cat CB-sinus nerve preparations in vitro, nicotinic ACh receptor (nAChR) blockers largely reduced the hypoxia-induced chemosensory discharge (40, 201, 202), as well as the responses induced by exogenously applied ACh (202, 203). Several subunits that comprise nAChRs have been described to be present in type I cells (204, 205). Cultured cat and rat type I cells respond to exogenously applied ACh and cholinomimetics with a depolarization (192) and an inwardly directed current (204–206) that is completely abolished in the presence of mecamylamine or hexamethonium (204–206) and was not mimicked by muscarine (206). Additionally, in rat isolated type I cells, ACh induces a dose-dependent increase in [Ca²⁺]_i, which can also be induced by nicotine and muscarine, responses that were completely blocked by mecamylamine and atropine, respectively (207). The preceding data indicate that ACh released from type I cells may result in a positive feedback loop activating type I cell nAChRs, depolarizing type I cells, and increasing [Ca²⁺]_i and probably transmitter release during natural stimulation. Additionally, and consistent with a postsynaptic effect, responses to ACh can be directly elicited in vitro from the perikaryon of PG neurons projecting to the carotid body (208–211) or in primary tissue culture (192, 212–215), activating one or several different nAChRs present in the same neuron (212, 215). The α4 and α7 subunits of the nAChR have been immunodetected in the perikaryon and fibers of cat PG neurons (213, 216) and α7 in PG neuron terminals surrounding type I cells in the rat (217). It is noteworthy that cat PG neurons in culture are highly sensitive to an homomeric α7 nAChR agonist (212). These data indicate that functional nAChRs are present on PG neurons (212, 215) and particularly in identified neurons projecting to the cat CB (40, 211). Thus, release of ACh would result in both post- and presynaptic activation of nAChRs, underlying, at least partly, the generation of afferent activity. Similarly, in PG neurons cocultured with CB cells, the hypoxia- and hypercapnia-induced responses recorded in the neurons are partially blocked by nAChR antagonists, but still leaving a smaller remnant response (195); cultures devoid of CB cells are nonresponsive to hypoxia and hypercapnia (195) but still responsive to ACh (192, 212, 214). The inability of nAChR blockade to eliminate afferent activity suggests the participation of at least an additional transmitter or receptors involved in the generation of chemosensory discharges (195, 211, 218, 219).

Activation of muscarinic ACh receptors (mAChRs) has no measurable effect on in vitro basal PG neuron activity in the rabbit, but their blockade with atropine enhances ACh-induced PG neuron activity, suggesting that the activation of mAChRs results in an inhibitory action on rabbit PG neurons (210). On the other hand, activation of mAChRs in cultured cat CB cells increases an outwardly directed current, an effect that was reversed by a muscarinic antagonist (220). Muscarine and its agonists increased [Ca²⁺]_i in isolated rat type I cells, an effect that was partially reduced in low-Ca²⁺-containing medium, suggesting the involvement of Ca²⁺ release from intracellular stores in the generation of this response (207, 221). M1 and M2 mAChRs have been detected by RT-PCR and immunohistochemistry in cat PG neurons and CB cells (222). Thus, activation of mAChRs appears to have dual effects, increasing [Ca²⁺]_i and probably synaptic communication between type I cells and PG terminals and directly inhibiting PG nerve terminals.

4.2. ATP and Ionotropic (P2X) Receptors

Adenine nucleotides have been localized in mouse type I cells by fluorescence microscopy, stored within granules (223). ATP is released from the whole CB, CB slices, and CB cells in culture in response to hypoxia (224, 225) and high extracellular K⁺ (224), release that is completely blocked by low extracellular Ca²⁺ (225) and addition of extracellular cadmium (Cd²⁺; 50 µM) or nifedipine (50 µM) (224), suggesting a Ca²⁺-dependent exocytotic process. Moreover, ATP release from the rat CB superfused in vitro is exponentially related to Po₂ of the solution (225, 226). Intracarotid ATP injections (227, 228) as well as adenosine (227, 229) produce a dose-dependent increase in cat CSN discharge, but adenine, guanosine, cytidine, inosine, and uridine were devoid of any appreciable effect on CSN discharge (227). Similarly, in the cat perfused-superfused CB preparation, in vitro ATP and adenosine 5′-[γ-thio]triphosphate (ATP-γ-S), but not adenosine, increased CSN frequency discharge in a dose-dependent manner (230). In the cat superfused CB preparation in vitro, exogenously applied ATP increased the CSN frequency discharge in a dose-dependent manner, a response that was specifically blocked by the simultaneous application of mecamylamine (2 µM) and suramin (50 µM) (202). In a similar rat CB preparation, suramin (25–100 µM) reduced both basal discharge and hypoxia-induced responses (219), and in the CSN responses induced by hypoxia and hypercapnia were largely reduced by suramin (50 µM) (225, 226). Intracarotid ATP injections in the cat produced a dose-dependent increase in tidal volume and CSN frequency discharge, effects that were blocked by a prior treatment of the animal with suramin (1 mg) (228). Similar blockade of afferent activity and ATP-induced responses has been recorded in mice by action of the P2 receptor antagonist pyridoxal-5′-phosphate-6-azophenyl-2′,4′-disulfonic acid (PPADS, 100 µM) (231). Moreover, ventilatory responses induced by reduced carotid body blood flow were significantly attenuated by suramin (14 µmol/kg) in the anesthetized rat, without modifying basal ventilation (226).

The CSN frequency discharge recorded from isolated cat and rabbit PG-carotid nerve preparations in vitro is increased by ATP in a dose-dependent manner, with minimal responses only induced by large doses of AMP (210, 232). Cat PG neurons in culture responded to exogenously applied ATP with a dose-dependent depolarization and a sustained inwardly directed current at resting membrane potential that was blocked by suramin (50 µM) (212, 219). The current time course and the fact that the inward current could also be induced by α,β-methylene ATP (212, 219) suggest that the receptors involved in the response are P2X_2/3 heterotrimers (233). Immunostaining of rat PG and CB indicate that both P2X₂ (218, 219) and P2X₃ (218) subunits are expressed in the ganglion, and that a vast majority of the neurons (>65%) express both subunits, but they always colocalize in the PG terminals within the CB parenchyma (218). In a chemosensory preparation reconstituted in vitro with rat CB cells and PG neurons, the basal activity recorded from PG neurons apposed to type I cells was reduced by suramin (25–50 µM). In the same preparation, responses to hypoxia (218, 219) and hypercapnia or acidification (195, 218) could be partially reduced by application of the P2X receptor blocker suramin (25–100 µM). These responses were dependent on exocytotic release of a transmitter from type I cells because they were blocked in the presence of high Mg²⁺ and low Ca²⁺ in the medium (195, 218). These data indicate that ATP released by type I cells and acting on P2X_2/3 receptors located in PG terminals in the CB is important for the generation of afferent activity. In mice, ventilatory responses and afferent activity are largely reduced in P2X₂-deficient, but not P2X₃-deficient, animals but are reduced to a larger extent in the double-deficient animals (231). In a preparation in vitro where the CB remained functionally connected to the PG, the discharge of neurons identified to be connected with the CB increased in response to acidification or stop flow, an effect that was partially blocked by suramin (211). These responses were similar to those recorded in the reconstituted chemosensory system (195, 218, 219) in the sense that a remaining response was still observed after the blockade of P2X receptors (195, 211, 218, 219) and indicate that other transmitter molecules are likely also involved in the transmission.

4.3. Combined Cholinergic-Nucleotidergic Transmission

It has been proposed that type I cells may release ATP and ACh in response to natural stimuli (195, 219, 234). In both the in vitro reconstituted chemosensory system (219) and the CB functionally connected to the PG (211), the responses elicited by stimulation were only partly reduced by the blockade of P2X or nAChRs (195, 211, 218, 219). However, the combined blockade of both receptor types (mecamylamine or hexamethonium plus suramin) eliminated both the basal activity (235) and the stimulus-induced activity (195, 211, 218, 219). Thus, in these preparations the combined elimination of nAChRs and P2X receptor activation was sufficient to eliminate chemosensory activity from innervating neurons, suggesting that ACh and ATP are the key transmitter molecules in the generation of chemosensory activity. It is noteworthy that a similar double blockade of nAChRs and P2X receptors in the cat in situ and in an in vitro superfused preparation obliterated the responses to exogenously applied ACh and ATP but only reduced the ventilatory and CSN responses to hypoxia (202, 228). This last result indicates that at least an additional communication mechanism must be responsible for the maintenance, at least in part, of the responses.

Nurse and colleagues (23, 53) proposed that hypoxic stimulation of type I cells may induce the release of “gliotransmitters” such as ATP via pannexin-1 channels from type II cells, contributing to the synaptic integration and cross talk among type I cells, type II cells, and the PG nerve endings, like in a tripartite synapse. Type II cells express purinergic P2Y₂ receptors and ATP-permeable pannexin channels (23, 55, 57), as well as muscarinic and angiotensin type 1 receptors (23). Opening of pannexin-1 channels in rat type II cells promotes further release of ATP and amplification of the ATP signal via the mechanism of ATP-induced ATP release. Type II cell-derived ATP, as well as that released from type I cells, can be broken down by 5′-ectonucleotidases to adenosine, which in turn may activate A2A receptors, leading to a rise in intracellular Ca²⁺ in both type I (236) and type II (55, 83) cells, although A2B receptors may also participate in increasing stimulus-induced catecholamine secretion from type I cells (237). Reciprocal cross talk between type I and type II cells may contribute to chemosensory processes in the CB via purinergic signaling pathways. Synaptic plasticity mechanisms during intermittent hypoxia or chronic sustained hypoxia might contribute to modifications in neurotransmitter signaling at the tripartite synapse (21, 23, 53). Recently, it has been shown that neuroblasts expressing tyrosine hydroxylase are the possible source of the hypoxia-induced increase in number of cells (hyperplasia) in the CB (238). It is worth noting that incubation of these isolated neuroblasts in media containing ATP or ACh significantly increased their maturation into O₂-sensitive cells, suggesting that in addition to their function as synaptic mediators, both transmitters may also participate in functional changes in the CB brought about by increased synaptic activity resulting from environmental modifications (238).

4.4. Dopamine

Type I cells contain dense-cored vesicles like those found in the adrenal medulla chromaffin cells (239). Dopamine (DA) is released from the rabbit CB and isolated type I cells (240–246), as well as from CBs of rats (92, 247–250), cats (251, 252), and humans (44). However, electrochemical determinations of DA show a lack of temporal correlation between extracellular DA levels and CSN discharge during increased chemosensory activity (247, 252), suggesting that DA could not participate as an excitatory transmitter in the communication between the CB cells and the PG terminals. In the rabbit, DA exogenously applied to the CB increases CSN discharge, but in most species tested DA reduces or completely blocks chemosensory activity. In the cat, DA injections evoke a transient reduction of CSN discharge, and continuous intravenous infusion of DA reduces the afferent response to both hypoxia and hypercapnia (253–255). The inhibitory effects of DA are blocked by a DA type 2 receptor (D2) antagonist (256), which also increases basal CSN discharge and enhances responses to hypoxia (253). It is noteworthy that exogenously applied DA has no effect on isolated rabbit type I cell Na⁺ or K⁺ currents while reversibly reducing the inwardly directed Ca²⁺ current (257). Thus, DA release from type I cells may regulate [Ca²⁺]_i increases, induced by the transduction mechanism, and transmitter release through a negative feedback mechanism.

PG neurons express D2 receptors in the nerve terminals located within the CB (258, 259), although D2 receptors appear to also be present in CB cells (260). In the isolated cat PG-carotid nerve preparation, in vitro application of DA to the ganglion produces no effect on CSN basal activity (261, 262) but increases CSN activity in a dose-dependent manner in the same rabbit preparation (230). Nevertheless, if DA is applied before ACh or ATP, it produces a dose-dependent modulation of the ACh (262)- and ATP(261)-induced responses in the cat, an effect that was reversibly and largely reduced by blocking D2 receptors (262). For any constant dose of ACh, the lower doses of DA potentiated the responses induced by ACh, whereas larger ones inhibited the responses elicited by both ACh (262) and ATP (261). Thus, although DA is released from CBs of many species, it appears to behave more as a modulator than a transmitter involved in the communication between type I cells and PG terminals and the generation of afferent activity. Nevertheless, postsynaptic actions on PG neurons are quite different between species and suggest caution in the extrapolation of synaptic mechanisms between species. Additionally, PG neurons projecting to the rat carotid body (172) and cat PG neurons in culture (263) express catecholaminergic traits and release DA in response to stimulation (263). Thus, DA released from PG neurons may modulate synaptic transmission and cellular activity by acting on both type I glomus cells and PG neurons.

4.5. NO, CO, and H₂S

Gas molecules are involved in many regulatory roles including neurotransmission, transcription, vascular resistance, and metabolism. NO and CO work as gas transmitters in the nervous system. Their small size and neutral charge enable gases to permeate through the cell membrane into the cell, allowing gases to contact rapidly with various functional groups of different molecules. Gas molecules have an ability to coordinate with metal centers of prosthetic groups of proteins (e.g., heme); NO and H₂S also exert their actions by covalently modifying the sulfhydryl group of cysteines in target proteins, processes called S-nitrosylation by NO and S-sulfhydration by H₂S. Thus, gas actions are pleiotropic in nature. NO and CO are also synthesized in the CB. Furthermore, NO from neuronal nitric oxide synthase (nNOS), expressed in the nerve endings that innervate type I cells, mediates efferent inhibition of the CB chemosensory activity (22, 23, 264). At physiological concentrations NO acts as a tonic inhibitory modulator of chemosensory discharges (264). NO modulates the chemoreception process by several mechanisms, by controlling vascular tone and oxygen delivery and by modulating the excitability of chemoreceptor cells and petrosal neurons (22). At high concentration, NO has a dual action depending on the oxygen level. In hypoxia, NO is predominantly an inhibitory modulator of chemoreception, whereas in normoxia, high NO levels produce chemoexcitation (265).

Heme oxygenase (HO) catalyzes the formation of endogenous carbon monoxide (CO). The isoform HO-2 is localized in type I cells, and HO inhibitors stimulate the CB chemosensory discharge activity. CO is an inhibitory modulator of CB chemosensory activity at low levels but at high concentration produces chemosensory excitation (266). As NOS and HO require molecular O₂ as a cofactor for the enzymatic process and inhibition of NOS and HO mimics hypoxia effects, it has been proposed that hypoxia stimulates CB by reducing the generation of CO and NO, two inhibitory influences of CB sensory activity. Type I cells also generate H₂S, another modulator. Cystathionine γ-lyase (CSE) and cystathionine β-synthase (CBS) are the two major enzymes that catalyze the formation of endogenous H₂S. CSE immunoreactivity has been demonstrated in rat type I cells by colocalization with tyrosine hydroxylase (267) and CBS (268), whereas mRNA for both enzymes (CBS and CSE) has been detected in rat (268), mouse, and cat CBs.

Like NO and CO, H₂S is a potent vasodilator and has been implicated in hypoxia-induced vasorelaxation. Peng et al. (267, 269) demonstrated that hypoxia increases H₂S levels in mouse and rat CB. Interestingly, H₂S generation by hypoxia was absent in the CB from CSE−/− mice and in rats pretreated with an inhibitor of CSE. Remarkably, CSE−/− mice displayed a reduced hypoxic ventilatory response, but ventilatory responses to hypercapnia were unaffected in CSE-knockout mice. Prabhakar and colleagues proposed that H₂S participates in the oxygen sensing process in the CB. In normoxia CO inhibits CSE-mediated H₂S production, whereas in hypoxia HO-2 produces less CO, which in turn increases H₂S production, increasing CB chemosensory discharge (267, 269). However, the role played by H₂S in CB oxygen sensing is controversial. Buckler (270) found that both single TASK K⁺ and whole currents were reduced by H₂S. In addition, cyanide, a mitochondrial blocker, caused a similar inhibition of TASK K⁺ current and rise in [Ca²⁺]_i, and the effects of H₂S and cyanide on the TASK K⁺ current were not additive, implying that the block by sulfide of the TASK channel was secondary to its effect on the mitochondria. Thus, the consequence of H₂S on O₂ chemoreception seems to be produced by the inhibition of TASK K⁺ channel activity due to a primary alteration of mitochondrial function.

5. CHEMOTRANSDUCTION IN THE CB

The emergence of photosynthetic organisms ∼3 billion years ago increased the Po₂ levels in the atmosphere and allowed the evolution of organisms that use glucose and O₂ to produce ATP through mitochondrial electron transport and oxidative phosphorylation. All cells respond and adapt to hypoxia (i.e., HIF transcription factor-mediated gene expression), but the peripheral chemoreceptors are able to sense arterial hypoxia, initiating a signaling cascade that produces fast and functional reflex responses (i.e., hyperventilation, increased cardiac output, etc.) that overcome tissue hypoxia. In mammals, O₂ sensing mechanisms have been studied in detail in erythropoietin-producing cells, neonatal chromaffin and PC12 cells derived from pheochromocytoma, bulbar and cortical neurons, pulmonary neuroepithelial cells, smooth muscle cells of pulmonary arteries, and CB type I cells (235, 271). Several molecules have been proposed as possible oxygen sensors in the CB, highlighting among them hemelike proteins (272–274), molecules associated with K⁺ channels (70, 100), and gases (17, 269, 275). However, despite great efforts over the last 50 years, the exact nature of the oxygen sensor still remains elusive (for recent reviews, see Refs. 9, 10, 12, 14, 16, 19).

Several hypotheses have been proposed to explain O₂ sensing in the CB. The “metabolic hypothesis” argues that the O₂ sensor is linked to oxidative phosphorylation metabolism in type I cells. The “membrane hypothesis” proposes that the sensors are molecules closely associated with K⁺ channels present in the type I cell membrane. Studies for and against both hypotheses are inconsistent. This could be attributed to the fact that experiments have been carried out with different species, stages of development, and experimental conditions. Critical analysis of these hypotheses makes it clear that neither one alone fully explains hypoxic chemoreception in the CB or can account for recent physiological and pharmacological findings. However, both mechanisms are indeed potential detectors of O₂ changes, and new findings support the idea that they may interact in the generation of the CB chemosensory response to hypoxia. In addition, it has been proposed that H₂S and CO (17, 269, 275), NO (276), or the olfactory Olfr78 receptor activated by lactate (277) is involved in CB oxygen chemoreception.

5.1. Relationship between Po₂ in the CB and Chemosensory Discharge

The CB chemoreceptors respond to changes in Po₂ rather than a drop in oxygen saturation (2, 278). The content of O₂ in the blood is essentially determined by the concentration and saturation of hemoglobin, and to a much lesser extent by the dissolved O₂. The available evidence indicates that CB chemosensory discharge increases in response to hypoxia produced by a reduction in arterial Po₂ ($\mathrm{P}{\mathrm{a}}_{{\mathrm{O}}_2}$) in vivo or in the saline Po₂ in in vitro preparations (278, 279) but not by a reduction in hemoglobin concentration or O₂ saturation (278, 280, 281).

The CB chemosensory discharge is inversely related to arterial Po₂ levels. A decrease in arterial Po₂ produces an exponential increase of frequency of CB chemosensory discharge (2, 4, 282, 283). Therefore, it is expected that the affinity for O₂ of the potential sensor molecules should be in the range of oxygen detection in the CB. Thus, it was important to determine the levels of Po₂ in the CB tissue and how they change in response to hypoxia. Microelectrode recordings of Po₂ in tissue in vivo and in situ preparations of the CB produced conflicting results, with Po₂ values ranging from 10–20 mmHg (284) to 50–70 mmHg (285), despite similar arterial Po₂ values (∼100 mmHg). Rumsey et al. (286) and Lahiri et al. (287) measured microvascular Po₂ in cat CB preparations in vitro and in situ with an optical method based on O₂-dependent quenching of the phosphorescence emitted by Pd-meso-tetra-(4-carboxyphenyl)porphine. Using this method, Lahiri et al. (287) found that the cat CB microvascular Po₂ in situ was ∼50 mmHg when the arterial Po₂ was ∼110 mmHg. The exponential increase of the chemosensory discharge begins at 40 mmHg in the CB vessels (corresponding to an arterial Po₂ of 90 mmHg) and reaches its maximum at values of 10–20 mmHg (arterial Po₂ ∼40 mmHg). Since the chromoporphine molecules do not leave the capillaries because they are associated with albumin, the Po₂ measured in the CB tissue should be lower. Indeed, Wilson et al. (288) measured intracellular Po₂ in the rat CB with nitroimidazole, which accumulates in cells depending on Po₂. During the incubation of the CB with a balanced solution to Po₂ ∼125 mmHg, they found that the Po₂ in the glomus tissue was 30 mmHg. These values are similar to those found in the brain and other tissues with microelectrodes (289).

5.2. Metabolic Hypothesis of CB Chemoreception

The original metabolic hypothesis stated that hypoxia reduces oxidative phosphorylation in type I cells, decreasing ATP, or cytoplasmic [ATP]-to-[ADP] [P_i] ratio, which in turn depolarizes mitochondria, releases Ca²⁺ from intracellular organelles, and triggers transmitter release (274, 290–294). The fact that metabolic inhibitors such as sodium cyanide (NaCN) and antimycin A applied to CB preparations in situ or in vitro produce chemosensory excitation in a dose-dependent manner strongly supports this hypothesis. Another argument in favor comes from Lahiri and colleagues (274, 293, 295–297), who found that CB chemosensory responses to hypoxia and NaCN disappeared after the administration of oligomycin, an ATP synthesis blocker, whereas the responses to hypercapnia and nicotine persisted, showing different sensing mechanisms. Several molecules or processes related to oxidative metabolism have been proposed as potential O₂ metabolic sensors (see TABLE 1 for historical review), such as a low-affinity cytochrome oxidase for O₂ (273, 299), a cytochrome b linked to NADPH (272), and cytochrome oxidase a₃ (294). Similar to hypoxia, oxidative phosphorylation uncouplers 2,4-dinitrophenol (2,4-DNP), carbonyl cyanide p-trifluormethoxyphenylhydrazone (FCCP), NaCN, oligomycin, and rotenone increase CB chemosensory discharge in a dose-dependent manner. These metabolic poison agents decrease ATP content, regardless of effects on the rate of O₂ consumption and mitochondrial potential. In the isolated rabbit CB, hypoxia reduces NADPH, depolarizes mitochondria, and increases [Ca²⁺]_i, an effect partially reduced by the removal of extracellular Ca²⁺ (292). Contrarily, oligomycin hyperpolarizes the mitochondria but also increases [Ca²⁺]_i (305), an effect that cannot be attributed to the release of Ca²⁺ from the mitochondria because hyperpolarized mitochondria retain Ca²⁺. Accordingly, Duchen and Biscoe (292) proposed that the increase in [Ca²⁺]_i could be produced by release from the intracellular compartment, sensitive to ATP or the [ATP]-to-[ADP] [P_i] ratio.

Table 1.

Historical proposed putative mechanisms for O₂ and CO₂-H⁺ sensing in the carotid body

Main Mechanisms	Main Players	References
Metabolic O2 sensing hypotheses
Hypoxia inhibits oxidative phosphorylation, decreasing ATP, which in turn depolarizes  mitochondria, releasing Ca2+ from intracellular organelles and transmitter release.	Respiratory chain	290, 298
	Mitochondrial oxidative phosphorylation	274
	Low-affinity cytochrome oxidase for O2	273, 299
	Cytochrome b linked to NADPH	272
	Cytochrome oxidase a3	294
Hypoxia inhibits respiratory chain, decreasing ATP and increasing AMP, which activates AMPK  to phosphorylate K+ channels, producing membrane depolarization, Ca2+ entry, and  transmitter release.	AMP-activated protein kinase (AMPK)	123, 125
Membrane O2 sensing
Hypoxia closes K+ channels, producing membrane depolarization, entry of Ca2+, and  transmitter release	K+ O2-dependent channels	70
	K+ voltage-gated channels	134, 135
	K+ TASK channels	100
Metabolic-membrane O2 sensing
Hypoxia reduces mitochondrial ATP production, which in turn closes K+ TASK channels	Mitochondria—ATP production	300
	Modulation by ATP of K+ TASK channel	106, 117
Hypoxia increases mitochondrial produced ROS and NADH, which closes K+ channels	Mitochondrial complex I electron transport chain	152, 301
	Modulation by ROS-NADH of K+ O2-dependent channels	12, 91
Other O2 sensing hypotheses
Complex relationship between CO and hydrogen sulfide (H2S). Hypoxia inhibits CO  production, leading to a decrease in protein kinase G-dependent phosphorylation  cystathionine-γ-lyase, which increases H2S production, causing K+ channel inhibition.	Hypoxia inhibits CO production, which in turn enhances H2S production	267, 269
Hypoxia inhibits electron transport lactate accumulation that activates G protein-coupled  receptor Olfr78, which by an unknown mechanism releases transmitter.	Lactate production—olfactory receptors	277
CO2-H+ sensing
Nerve endings are sensitive to acidity, and extracellular pH is held stable by an active  mechanism that becomes less effective in hypoxia	Common mechanism of acidification mediated by O2 and CO2-H+	302
[H+]i	Key role for carbonic anhydrase	302–304
High [H+]i closes K+ channels, leading to cell depolarization, entry of Ca2+ and transmitter  release.	K+ TASK1/3 channels	100
	Acid-sensing ion channels (ASICs)	156

AMPK, AMP-activated protein kinase; [H⁺]_i, intracellular H⁺ concentration; ROS, reactive oxygen species.

Mills and Jöbsis (273, 299) postulated the existence of a low-affinity cytochrome oxidase for O₂, probably located in type II cells in the CB. They studied the absorption spectrum resulting from hypoxia and subsequent reduction of respiratory pigments with sodium dithionite in a perfused preparation of cat CB. However, no study has confirmed the existence of a cytochrome oxidase with such an affinity for O₂ (∼80 mmHg) in any cell. Mills and Jöbsis measured Po₂ in the perfusion medium and not in the CB, which in these conditions is expected to be lower (286), and it is possible that their spectrum was contaminated with myoglobin. Cytochrome oxidase a₃ is another plausible candidate to be the oxygen sensor in the CB (266, 294, 306). Cytochrome oxidase a₃ oxidizes at very low levels of Po₂ (K_m ∼1–2 mmHg) in isolated mitochondria. However, the K_m of cytochrome oxidase a₃ for O₂ is higher in tissues, because of the existence of important Po₂ gradients. Erecińska and Wilson (289) found that the oxidized cytochrome a₃ of many cells was reduced from a Po₂ of 10–20 mmHg, values that are compatible with Po₂ in the CB during moderate hypoxia (266, 286). Wilson et al. (294) reported that the CO-induced increase of CB chemosensory discharge was completely reversed by bright light. At submaximal light intensities, the reduction of CO-induced chemosensory excitation was maximal for light of 432 ± 2 and 590 ± 2 nm, with a ratio 432/590 of ∼6. CO is a structural analog of O₂ and specifically binds the site for O₂ in cytochrome a₃. The action spectrum for light reversal of the CO effect in the CB corresponds to cytochrome-c oxidase of the mitochondrial respiratory chain (294).

5.3. Arguments against the Metabolic Hypothesis of Chemoreception

One important argument against the original metabolic hypothesis is the possibility that some effects of oligomycin on the chemosensory responses to hypoxia and NaCN may take place not only in the mitochondria. It is plausible that the ATP depletion produced by oligomycin affects the function of ion channels and membrane transporters, modifying the membrane potential. On the other hand, high doses of oligomycin (>1 µg/mL) block the Na⁺-K⁺-ATPase pump (307). An alternative explanation for why oligomycin inhibits the response to hypoxia but not to nicotinic agonists is that the depletion of energy can interfere with the storage of transmitters in vesicles (ACh and ATP) and with their subsequent release. Measurements of ATP in the CB in response to hypoxia showed contradictory results, because it was not possible to measure ATP levels accurately. Acker and Starlinger (308) found that ATP in the cat CB was not changed by hypoxia, but Obeso et al. (309), using the same preparation, found that ATP decreased in response to hypoxia. However, the same group found later that although NaCN reduces ATP by 45%, 2,4-DNP did not modify its levels (310). Verna et al. (311) studied the effects of NaCN, antimycin A, and hypoxia on the nucleotide content in rabbit CB and found that metabolic inhibitors reduced ATP but not hypoxia. These results are inconclusive because the CB is not homogeneous, and type I cells are only a part of the organ. Thus, the biochemical determinations of ATP in CB may contain large errors. One of the most important arguments against the metabolic hypothesis was that several groups proposed that Ca²⁺ should be released from intracellular organelles (274, 278, 298). This is in contradiction to the data suggesting that membrane depolarization and extracellular Ca²⁺ influx from the extracellular space are absolutely necessary to induce transmitter release from type I cells (70, 135, 300).

Evans and colleagues proposed the AMP-activated protein kinase (AMPK) hypothesis of oxygen sensing, which is dependent on mitochondrial oxidative metabolism. They suggested that hypoxia reduces the mitochondrial electron transport, triggering a reduction in ATP production, which in turn activates the enzymatic conversion of ADP to ATP and AMP, mediated by adenylate kinase. The increased levels of AMP activate AMPK, which phosphorylate K⁺ channels in the membrane of type I cells, eliciting depolarization, Ca²⁺ entry, and transmitter release. This construct was based on the use of AMPK activators mimicking the effects of hypoxia and antagonists that reversed the effects of hypoxia (124, 125). However, the hypothesis was questioned because of the poor selectivity of AMPK activators and blockers (312, 313) and the fact that other studies were unable to repeat the initial observations (126). To address these contradictory results, Evans and colleagues used a TH promoter to drive deletion of AMPK-α1 and -α2 genes in all catecholaminergic cells, including type I cells of the CB. They found that the AMPK produced marked reduction of the ventilatory response to hypoxia, provoking hypoventilation rather than hyperventilation and frequent prolonged apneas (11, 132). The AMPK deletion in catecholaminergic cells may produce alteration in the CB as well as in the brain stem centers that control ventilation. However, AMPK deletion did not block the CB chemosensory response to hypoxia and hypercapnia. Accordingly, Evans (131) revised the original hypothesis and proposed that “AMPK integrates local hypoxic stress at defined loci within the brain stem respiratory network with an index of peripheral hypoxic status, namely afferent chemosensory inputs.” Thus, AMPK crucially participates in ventilatory regulation, maintaining the O₂ and ATP supply to the organism.

5.4. Membrane Hypothesis of CB Chemoreception

The membrane hypothesis states that the O₂ sensors are molecules closely associated with ion channels in the type I cell membrane. Accordingly, O₂ sensing in the type I cells is coupled to membrane depolarization and Ca²⁺-dependent secretion of one or more excitatory transmitters that increase discharge of sensory neurons. As mentioned in sects. 2.2 and 2.3, López-Barneo et al. (70) reported the existence of a voltage-dependent current of K⁺ whose conductance decreases reversibly with hypoxia in isolated rabbit type I cells. This seminal work was the first report of the action of Po₂ on a membrane current. Currently, the number and different types of channels whose conductance depends on Po₂ has markedly increased (100, 314–317). The presence of O₂-sensitive K⁺ voltage-dependent currents has been confirmed in rat, rabbit, and cat type I cells (99, 134, 140, 318, 319). The K⁺ current described by López-Barneo et al. (70) in rabbit type I cells is type A, depends on the membrane potential, is activated at −30 to −40 mV, presents inactivation, and does not depend on Ca²⁺. In neonatal rat type I cells, Peers (134) described a high-conductance O₂-dependent K⁺ current that is also Ca²⁺ dependent (type BK-Ca²⁺, 190 pS). In adult rat (71) and cat (318) type I cells, the O₂-dependent K⁺ currents are voltage dependent and independent of Ca²⁺ and did not show inactivation. In addition, Buckler (100) found in neonatal rat type I cells that hypoxia blocked a “background” K⁺ current, which is voltage independent and corresponds to a TASK type (96).

5.5. Arguments against the Membrane Hypothesis of Chemoreception

The ${\mathrm{K}}_{{\mathrm{O}}_2}$ current and the opening probability of apertures of single channels from rabbit type I cells are reduced by hypoxia in the range from 150 to 80 mmHg (K_m > 90 mmHg). At 80 mmHg the inhibition is maximal and the frequency of channel opening and conductance are not further reduced by lower Po₂ (71, 80). In the cat CB, Shirahata and Sham (320) reported that maximal inhibition of O₂-sensitive K⁺ current occurs at 40 mmHg, with a K_m > 60 mmHg. This contrasts with the measurements of Po₂ showing that the chemosensory discharge increases when Po₂ falls below 40 mmHg in the CB tissue in vivo and in vitro (266, 286). The K⁺ current whose affinity for Po₂ is compatible with the Po₂ values measured in the CB is the K⁺ TASK channel described by Buckler (100) in type I cells from neonatal rats, with a K_m of ∼15 mmHg for Po₂. Thus, voltage-dependent O₂-dependent currents in type I cells of different species are inhibited by hypoxia with low affinity for O₂, whereas the K⁺ TASK channel showed an affinity for Po₂ compatible with the threshold values for chemosensory activation found in the CB tissue.

López-Barneo et al. (70) and Stea et al. (321) found that rabbit and rat type I cells generated action potentials when membrane potential was previously maintained at −90 mV. Even in these conditions, hypoxia discretely increased the frequency of action potentials (72). These observations contrast with the results obtained by other researchers who only found tonic depolarization in response to hypoxia (99, 140, 318). A possible explanation could be the fact that the normal resting membrane potential of type I cells is on the order of −40 mV (−20 to −50 mV). At this potential, most voltage-dependent Na⁺ channels are inactivated, and type I cells cannot produce action potentials. Only at highly hyperpolarized membrane potentials (−90 mV) are action potentials produced. If type I cells do not generate action potentials, the contribution of the voltage-dependent ${\mathrm{K}}_{{\mathrm{O}}_2}$ channels to the hypoxic response should be small, since their maximum inhibition occurs at positive voltages in the range of +10 to +40 mV. Therefore, the K⁺ TASK channel is the best candidate to mediate the hypoxic response to low Po₂, since its current is voltage independent and insensitive to ASD, 4-aminopyridine (4-AP), and charybdotoxin (ChTX) and contributes to basal conductance (100). Recently, Wang and Kim (322) proposed differential roles for background and voltage-dependent K⁺ channels in hypoxia. They measured [Ca²⁺]_i in rat type I cells to study the effects of K⁺ channel blockers alone or in combination with hypoxia and membrane depolarization with high [K⁺]_i. They tested the effects of iberiotoxin (Ibtx), a specific blocker of MaxiK, and TEA + 4-AP, a combination that blocks most K⁺ voltage-gated channels, and found that Ibtx or TEA + 4-AP did not increase [Ca²⁺]_i. Interestingly, TEA + 4-AP increased [Ca²⁺]_i when type I cells were previously depolarized with 15 mM K⁺. In addition, TEA + 4-AP enhanced the hypoxia-induced increase in [Ca²⁺]_i. Wang and Kim (322) also found that MaxiK activity was not affected by hypoxia. Accordingly, they proposed that MaxiK and K⁺ O₂-dependent channels did not contribute to depolarization induced by hypoxia in rat type I cells. They proposed that K⁺ TASK1/3 channels, which are the predominant conductance channels in resting conditions, are responsible for initiating the depolarizing response to hypoxia in rat type I cells, whereas K⁺ voltage-gated channels work as a feedback to prevent further membrane depolarization. However, Ortega-Sáenz et al. (109) found that the secretory DA responses to hypoxia and hypercapnia in CB preparations from TASK1^−/− or double TASK1/3^−/− mice were normal and were maintained when MaxiK was blocked. The TASK1/3-null type I cells showed a significantly higher membrane resistance and presented a more hyperpolarized resting potential. The TASK1/3-null cells showed diminished responses to pH 6.8.

López-Barneo (16) acknowledged the crucial contribution of background K⁺ channels in type I cells to maintain the resting membrane potential, because they are the predominant resting ionic conductance, but the contribution of MaxiK or voltage-gated K⁺ (K_v) channels to the type I cell electrophysiological properties should not be underestimated. López-Barneo (16) states that the resting membrane potential in type I cells, which are electrically coupled and modulated by autocrine, paracrine, and efferent mechanisms, may be altered toward values at which a fraction of K_v or MaxiK channels are open. This idea is supported by the findings that in isolated type I cells, which are slightly depolarized with high external K⁺, 4‐AP + TEA or even Ibtx increased [Ca²⁺]_i (322). In addition, TEA induces the release of catecholamines from rat or mouse type I cells in the CB slice preparation (109). Undoubtedly, further studies are needed to understand the precise role played by background and voltage-gated K⁺ channels in CB chemoreception.

Another major argument against the proposal that that K⁺ channels alone are the primary O₂ sensors is the time course required for the depolarization of type I cell membrane and the mitochondria. Indeed, Buckler and Vaughan-Jones (300) reported that FCCP and 2,4-DNP depolarized the mitochondria 3–4 s before membrane depolarization began. These experiments were done with rhodamine 123, which responds slowly to the mitochondrial potential change. Thus, the existence of some signal between the mitochondrial potential, which determines the production of ATP, and the membrane potential of the type I cells is feasible. It is likely that this signal may link mitochondrial oxidative metabolism with cell membrane excitability (117).

5.6. Interaction between Oxidative Metabolism and Membrane Currents

In many systems that critically depend on O₂, such as coronary blood flow, signals that relate oxidative metabolism to physiological functions have been sought. One of these signals is ROS, which depending on their concentration may participate as intercellular signals in the transduction processes or can cause cellular damage (323–325). There is evidence that ROS may inhibit the activity of ionic channels and transporters in cell membranes (323), including the ${\mathrm{K}}_{{\mathrm{O}}_2}$ channels present in type I cells and in smooth muscle cells (316). Gulbis et al. (326) found that the β-subunit of voltage-dependent K⁺ channels is an oxidoreductase with an active site for NADH, whose activity can be modulated by the redox state of the cells. This subunit could be part of the coupling mechanism between metabolism and cell excitability. During hypoxia, a decrease in the production of ROS generated by a NADPH oxidase that converts O₂ to superoxide ions can occur. In mouse CB, based on studies of absorption spectra, Acker et al. (327) proposed that a cytochrome b linked to NADPH oxidase was the O₂ detector. However, the participation of this system has been questioned for various reasons. First, knockout mice without the gp91 subunit of NADPH have normal ventilatory responses to hypoxia. In addition, cytochrome b linked to NADPH was found in macrophages and not in type I cells (328). In addition, NADPH oxidase inhibitors did not increase DA release from rabbit CB (329) or chemosensory discharge in the rat (330). Another possibility is that hypoxia actually increases ROS and that an increase, and not a decrease, works as a signal to inhibit the ${\mathrm{K}}_{{\mathrm{O}}_2}$ channels. In fact, the electron transport chain and cytochrome-c oxidase have an important role in the generation of ROS (325, 331). During hypoxia, cytochrome c is reduced, its maximum speed decreases, and the normal transport of electrons is reduced; as a consequence, electrons are transferred to the ubiquinone complex, generating superoxide (331, 332).

Sato (333) found that FCCP increases [Ca²⁺]_i in rabbit type I cells. This increase was reduced by 80% with Ni²⁺ and by 53% with verapamil. The removal of extracellular Ca²⁺ eliminated the FCCP-induced increase of [Ca²⁺]_i in 71% of type I cells studied, but in the remaining 29% FCCP reduced it by just 68%. These results showed that in most type I cells FCCP increases [Ca²⁺]_I, probably through voltage-L-type channels, but in some type I cells Ca²⁺ may be released from intracellular organelles. Buckler and Vaughan-Jones (300) reported that FCCP and 2,4-DNP produced an increase in [Ca²⁺]_i in rat neonatal type I cells. This increase was not due to intracellular acidification and was reduced by 80% with Ni²⁺ and by the removal of Ca²⁺ and extracellular Na⁺ but only by 50% with verapamil. Buckler and Vaughan-Jones (300) found that FCCP and 2,4-DNP depolarize mitochondria 3–4 s before starting the depolarization of the cell membrane, suggesting the existence of a signal between the mitochondria and the membrane. Similarly to hypoxia, the oxidative phosphorylation uncouplers 2,4-DNP and FCCP and the inhibitors NaCN, myxothiazol, oligomycin, and rotenone reduced the activity of the K⁺ TASK channels in cell-attached patches (104, 105, 117) and mimicked whole cell current hypoxia-induced responses (106). Moreover, in the presence of NaCN, rotenone, or FCCP, the effects of hypoxia on this current were completely obliterated (106), suggesting that the mechanisms of action of O₂ and the uncouplers/inhibitors converge at an end point, i.e., the K⁺ current. A similar effect on this K⁺ current was brought about by application of H₂S to the extracellular medium, reducing whole cell current in attached type I cells (270), apparently by its action as an inhibitor of cytochrome-c oxidase activity, oxidative phosphorylation, and ATP production (270, 334, 335). Additionally, the effects of H₂S were not additive with the effects of cyanide, suggesting a common pathway in the reduction of K⁺ current. It is noteworthy that the activity of the channels was largely reduced in inside-out excised patches (96, 104, 116, 117) and that in this last condition the sensitivity to O₂ was completely absent (96). These results show that hypoxia-induced reduction in K⁺ TASK current is dependent on the integrity of the cell, suggesting that an intracellular mediator and not oxygen per se modulates the activity of the channel. Moreover, the activity of the recorded excised patches could be partially restored by the application of ATP to the intracellular side of the patches (102, 268, 324), with a half-maximal effect at ∼2 mM (116, 117). The effect was not antagonized by ADP or AMP (96) but was mimicked by a nonhydrolyzable ATP analog (AMP-PCP, 10 mM) (116) as well as purine (GTP, 5 mM) and pyrimidine (UTP, 5 mM) triphosphate nucleotides (96). These results suggest that these K⁺ channels appear to be directly modulated by ATP and that this molecule could be the link between O₂ availability and channel activity.

López-Barneo and colleagues (3, 91, 301, 336) have proposed that O₂ sensing in type I cells depends critically on ROS production in the mitochondria I complex. They suggested that hypoxic inhibition of the mitochondrial electron transport chain increased the production of ROS and reduced pyridine nucleotides from complex I of the type I cell mitochondria (3, 12, 91, 337). Indeed, Ortega-Sáenz and López-Barneo (3) put forward a unified metabolic-membrane hypothesis entitled “Mitochondria to membrane signaling model of acute oxygen sensing.” They state that acute hypoxia increases the reduced state of mitochondrial complex IV (MCIV), causing the accumulation of electrons along the electron transport chain. The accumulation of electrons increases reduced ubiquinone (CoQH₂) levels, which in turn slows down (or even reverses) mitochondrial complex I (MCI), leading to increases in ROS and NADH. Both ROS and NADH may change the redox status of the K⁺ channels in the membrane of type I cells, acting on the microdomains between membrane channels and mitochondria (FIGURE 3).

FIGURE 3.

Model for acute O₂ sensing by type I cells. Schematic representation of the mitochondrial-to-membrane signaling model of phototransduction in carotid body (CB) type I cells. Mitochondrial complex (MC)IV is the oxygen sensor, MCI is the effector, and NADH and reactive oxygen species (ROS) are signaling molecules that modulate ion channels in the plasma membrane. The size of MCIV relative to others is enlarged to facilitate explanation of the model. COX412, cytochrome-c oxidase subunit isoform 2; COX8B, cytochrome oxidase subunit VIIIbIMS, intermembrane space; QH2, reduced ubiquinone; FMN, flavin mononucleotide. (Reprinted from Ref. 338, with permission from Science Signaling.)

López-Barneo and colleagues proposed that a rotenone-binding molecule is involved in oxygen sensing in type I cells, because rotenone, a blocker of MCI, reduced the responses of type I cells to hypoxia but not the responses to hypoglycemia (95, 301, 337). Fernández-Agüera et al. (91) used a transgenic mouse without the Ndufs2 gene, which encodes for the 49-kDa subunit of the ubiquinone (CoQ) binding site for rotenone in the catalytic core of MCI. They found that mice without Ndufs2 restricted to tyrosine hydroxylase (TH)-positive cells showed a complete abolition of the hypoxic ventilatory response but a normal response to hypercapnia (91). In addition, hypoxia did not increase catecholaminergic secretion in type I cells from Ndufs2 null, but the cell responded normally to hypercapnia, hypoglycemia, and high extracellular K⁺. Gao et al. (12) compared gene expression profiles from three catecholaminergic cell types (type I, adrenal chromaffin cell, and superior cervical ganglion neurons) with different oxygen sensitivity but common lineage. They reported the expression in type I cells of a gene profile featured by high levels of pyruvate carboxylase and the presence of three atypical mitochondrial subunits (Ndufa4l2, Cox4i2, and Cox8b) as well as the downregulation of Phd3 and upregulation of Hif2α and high levels of biotin. COX412 is a nuclear-encoded isoform of cytochrome-c oxidase that is encoded by the gene Cox4i2. This enzyme couples the transfer of electrons from cytochrome c to molecular O₂ and contributes to produce the proton electrochemical gradient across the inner mitochondrial membrane. The presence of this atypical mitochondrial subunit COX4I2 in the type I cell may make cytochrome-c oxidase more sensitive to low Po₂, causing a backup of electrons in the electron transport chain and a further increase in the CoQH₂-to-CoQ ratio, leading to increased ROS and NADH production in mitochondrial complex I. Finally, the increase of ROS and NADH modulates the opening and closing features of O₂-dependent K⁺ ion channels. The idea that ROS are part of oxygen sensing is partially contradictory; oxidizing and/or reducing agents and ROS scavengers did not mimic or inhibit the rapid effects of hypoxia on K⁺ TASK channels (106, 133). Moreover, whether ROS are involved in O₂ sensing in the CB is debatable, because H₂O₂ did not increase chemosensory discharge in rat (339) or cat (340) CB.

More recently, Moreno-Domínguez et al. (338) provided further evidence for the participation of HIF2a and the genes that codify for the mitochondrial subunit COX4I2. They proposed that the gene expression profile and O₂-sensing properties of type I cells depend on Epas1, the gene that encodes HIF2a. They found that type I cells from adult mice with conditional Epas1 and Cox4i2 gene deletion showed a marked reduction of the intracellular [Ca²⁺]_i increase in response to hypoxia, but not to hypercapnia, as well as a marked reduction of NADPH autofluorescence during hypoxia. In addition, adult mice with conditional Epas1 and Cox4i2 gene deletion showed reduced hypoxic ventilatory responses. Thus, their results support an interesting relationship between the expression of the atypical mitochondrial COX4I2 subunit and Epas1 in type I cells. Moreover, they found in mice studied at an interval of ∼2 mo after Epas1 inactivation a marked reduction in the expression of Cox4i2, Ndufa4l2, and Cox8b, suggesting that HIF2a induces the expression of CB genes needed for the atypical pattern of mitochondrial subunit expression, which may underlie the relatively high affinity of the mitochondria and type I cells for hypoxia. Another interesting finding is that the NADH and ROS increases in the intermembrane space that normally occur in hypoxia were practically abolished in cox4i2-knockout mice, suggesting that the ROS increase occurs from an accumulation of electrons proximal to MCIV, which modulates the opening of ${\mathrm{K}}_{{\mathrm{O}}_2}$ channels (FIGURE 3). Further experiments will be necessary to understand the exact nature of the oxygen sensing mechanisms.

5.7. Gases as Modulators of Oxygen Sensing

Prabhakar and colleagues proposed that complex interactions among CO, NO, and H₂S with the O₂-dependent K⁺ channels and/or mitochondrial electron transport chain in type I cells are necessary for oxygen sensing in the CB, suggesting that H₂S takes part in the oxygen sensing process in the CB. In normoxia, CO inhibits CSE enzymatic H₂S production, activating the protein kinase G-dependent (PKG), which in turn inhibits H₂S generation by CSE. In hypoxia, heme oxygenase-2 (HO-2) produces less CO, enhancing the H₂S production that increases CB chemosensory discharge (267, 269, 341). The concept of gases as the primary mediators of sensing of O₂ in the CB is controversial. First, HO-2-knockout mice show normal responses to hypoxia (152, 182). Second, gases are amply distributed and may have several effects at different parts of the chemosensory process. For instance, the application of H₂S to the extracellular medium reduced the K⁺ TASK whole cell current in attached type I cells (270), apparently by its action as an inhibitor of cytochrome-c oxidase activity, oxidative phosphorylation, and ATP production (270, 334, 335). Additionally, the effects of H₂S were not additive with the effects of cyanide, suggesting a common pathway in the reduction of K⁺ current. Therefore, if hypoxia causes an increase of H₂S it is possible that a high concentration of H₂S may impair electron transport and oxidative phosphorylation. On the other hand, H₂S did not modify the K⁺ TASK channel activity (176), and knockout of CSE did not affect oxygen sensing in the mouse CB (342).

At high concentrations CO and NO compete with O₂ in cytochrome oxidase a₃. This effect is photosensitive and reverted maximally with wavelengths of 432 ± 2 and 590 ± 2 nm, which correspond to the photo-spectrum of action of the CO-cytochrome oxidase a₃ complex (294). At low values, CO reduced chemosensory excitation in response to hypoxia (287) and reversed the inhibition of hypoxia on the K⁺ O₂ current (343). This dual behavior has been attributed to the action of CO on heme groups in cytochrome a₃ and in a protein (subunit B) that would be part of or associated with the O₂-sensitive K⁺ channel. NO has the same behavior as CO; at low concentrations it produces inhibition of the chemosensory hypoxic response, and at high concentrations it increases carotid discharge (264, 265). NO binds to cytochrome oxidase a₃ (344), and inhibits the production of ATP (345). Oligomycin blocked the responses to high NO as well as to hypoxia (296), suggesting a common pathway.

5.8. Olfactory Receptors Activated by Lactate

Chang et al. (277) proposed that the olfactory Olfr78 receptor activated by lactate is the oxygen sensor in the CB. They found that OLFr78 is highly expressed in type I cells and is activated by lactate at low millimolar concentrations. In addition, they reported that Olfr78^−/− mutant mice showed a reduced ventilatory response to acute hypoxia. The proposal that the olfactory Olfr78 receptor is involved in O₂ sensing is interesting, but these results were not reproducible in other laboratories. Indeed, Torres-Torrelo et al. (346) were unable to replicate these findings. They found that Olfr78^−/− mutant mice from the Frankfurt colony have a normal ventilatory response to hypoxia and the catecholamine release evoked by hypoxia and lactate from type I cells was similar in Olfr78^−/− and wild-type mice. Chang et al. (347) recognized that the major difference between their original observation (277) and the findings of Torres-Torrelo et al. (346) was the reduced hypoxic response that they found in Olfr78^−/− mutant mice. To address the differences between the results obtained by Torres-Torrelo et al. (346) and their previously published study, Chang et al. (347) tested female progeny of newly cryo-recovered Olfr78^+/− mutant mice from JAX as well as from the Frankfurt colony, with a protocol similar to that used by Torres-Torrelo et al. (346). “Surprisingly, we did not observe a consistent defect in the hypoxic ventilatory response of either current strain under these conditions” (347). Although it is not possible to preclude effects due to differences in conditions (i.e., the animal facility and equipment used in their original experiments) Chang et al. (347) recognized that “…the original cohort had an unusual genetic background that rendered a defect in this response more apparent in Olfr78^−/− mutants.” It is worth noting that Chang et al. (347) in the new experimental series found a marked variance in the ventilatory frequency in hypoxia compared to WT. Thus, they suggest the existence of variable compensatory responses induced by Olfr78 loss, that may involve the CB or central command of ventilation.

5.9. Chemotransduction for CO₂-H⁺

The CB senses Pco₂ and H⁺ in arterial blood, contributing to 20–30% of the reflex regulation of CO₂-H⁺ levels in response to hypercapnic acidosis, while the central chemoreceptors are responsible for the other 70–80%. The transduction in type I cells involves the action of carbonic anhydrase (302, 304, 348), K⁺ TASK channels (349, 350), and acid-sensing ion channels (156, 157). The most widely accepted hypothesis is that high [H⁺]_i inhibits K⁺ TASK channels, which initiates cell depolarization, entry of Ca²⁺, and the release of transmitters (96). However, the CB chemosensory discharge also increased in response to isohydric hypercapnia (304). In the absence of external acidosis, the intracellular pH is reduced by the CO₂ flux mediated by the presence of carbonic anhydrase. Moreover, in the presence of CO₂–$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$, the intracellular pH is acid compared to the extracellular pH (96, 279, 349, 350). The chemosensory discharge depends on the functional presence of carbonic anhydrase and CO₂–$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ (279, 304, 351). Indeed, extracellular CO₂–$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ improved CB chemosensory responses to hypoxia (304). Most of the studies agree that the transduction of CO₂ in the CB results from the decrease in intracellular pH. An increase in arterial Pco₂ causes a parallel rise in intracellular Pco₂ very rapidly as the gas diffuses in seconds from the capillaries into the type I cells in the presence of carbonic anhydrase (304).

5.10. Conclusions and New Directions in O₂-CO₂ H⁺ Chemotransduction

Despite >80 yr of physiological, molecular, and genetic studies concerning the O₂ and CO₂-H⁺ sensing mechanisms in the type I cells of the CB, the precise nature of the mechanisms remains elusive. The mechanisms that underlie O₂ sensing in type I cells have been a matter of debate and discussion. Several hypotheses and putative candidates for the O₂ sensor have been proposed (see TABLE 1). Some of these hypotheses have been recently questioned by conclusions drawn from the same proposers, i.e., hypoxia-response coupling by AMPK (131) and olfactory receptors (347). The CB shows an exquisite sensitivity to Po₂ in the physiological range (arterial Po₂ < 80 mmHg). Similarly, the administration of metabolic poisons, which mimic most of the effects of hypoxia such as type I cell depolarization, increases of cytosolic [Ca²⁺]_i, transmitter release, and chemosensory excitation, supports a main role for mitochondrial oxidative metabolism in the chemosensory process. On the other hand, the seminal electrophysiological findings of López-Barneo et al. (70) and Buckler (100) supporting a major contribution of membrane voltage-gated and background K⁺ channels to the hypoxic response unlocked a new era in CB research. Nevertheless, as discussed above, both hypotheses for CB chemoreception failed to explain the whole process of O₂ chemotransduction. In recent years a combined hypothesis, integrating the participation of mitochondrial oxidative metabolism and the membrane currents, has been proposed (see TABLE 2). Accordingly, hypoxia alters mitochondrial function, modifying the levels of molecules (i.e., ROS, NADH, ATP) produced within the mitochondria, which in turn modulate the opening of ${\mathrm{K}}_{{\mathrm{O}}_2}$ and TASK K⁺ channels, leading to cell membrane depolarization, entry of extracellular Ca²⁺, and subsequent transmitter release. Indeed, Fernández-Agüera et al. (91) have shown that intracellular dialysis with H₂O₂ at millimolar concentration increases input resistance in type I cells, an effect that is compatible with inhibition of resting K⁺ conductance by mitochondrion-derived ROS. The coupling of mitochondrial oxidative metabolism and K⁺ channel function is made possible by the close association between the mitochondria and the ion channels in the cell membrane microdomains in type I cells. A growing body of electrophysiological evidence supports the idea that hypoxia reduces the mitochondrial oxidative phosphorylation, decreasing the ATP production, which in turn mediates the closing of TASK channels (106, 117, 300). On the other hand, as shown in FIGURE 3, López-Barneo and colleagues, based on electrophysiological studies performed in transgenic mice, have proposed that hypoxia increases the reduced state of MCIV, leading to an accumulation of electrons along the electron transport chain increasing the levels of CoQH₂, which results in slowdown (or even reversal) of MCI with production of ROS and accumulation of NADH. The increase of ROS and NADP modulates ${\mathrm{K}}_{{\mathrm{O}}_2}$ channel opening (3, 12, 91, 152, 301, 338).

Table 2.

Current models for O₂ and CO₂-H⁺ sensing in the carotid body

Hypothesis	Mechanisms	References
Unified oxidative metabolism and membrane currents for O2 sensing Low Po2 modifies the mitochondrial function, altering the levels of molecules that modulate K+ channels, leading to cell membrane depolarization, entry of Ca2+, and transmitter release. In type I cells, mitochondria are close to ion channels in the cell-membrane forming O2 sensing microdomains.	Hypoxia reduces the mitochondrial oxidative phosphorylation, which decreases the ATP production.	106, 300
	ATP of K+ TASK channel	117
	Hypoxia increases the reduced state of MCIV, causing an accumulation of electrons along the electron transport chain, which enhances QH2.The enhanced QH2 slows down, or reverses, the production of ROS and accumulation of NADH in MCI. ROS and NADH modulates K+ O2-dependent channels	3, 12, 91, 152, 301, 338
High [H+]i closes K+ channels, leading to cell depolarization, entry of Ca2+, and transmitter release. Carbonic anhydrase mediates the fast intracellular pH response to hypercapnia-acidosis.	Key role for carbonic anhydrase CO2/$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ and [H+]i	251, 302,303, 350, 352
	K+ TASK1/3 channels	100
	Acid-sensing ion channels	156

[H⁺]_i, intracellular H⁺ concentration; MCI and MCIV, mitochondrial complexes I and IV; QH₂, reduced ubiquinone; ROS, reactive oxygen species.

In the case of CO₂-H⁺ sensing, most investigators agree that the type I cells sense an increase in [H⁺]_i, which inhibits K⁺ TASK channels, causing membrane cell depolarization, entry of Ca²⁺, and the release of the transmitters (96, 349, 350). It is also possible that acid-sensing ion channels may contribute to the process by decreasing the intracellular pH (156, 157). The fast transduction of extracellular CO₂-H⁺ stimuli depends on the enzymatic action of carbonic anhydrase (302, 304, 348).

6. CAROTID BODY CHEMORECEPTOR REFLEXES—INTEGRATED PHYSIOLOGY

As described in sect. 3, transmitters released from CB type I cells depolarize chemoreceptor afferent nerve terminals that signal second-order neurons in the NTS via the glossopharyngeal (IX) nerves (2, 282). Cell bodies of the sensory neurons are located in the petrosal ganglia (282). Synaptic projections ultimately reach the respiratory pattern generator, presympathetic neurons in rostral ventrolateral medulla (RVLM) and cardiovagal preganglionic neurons in nucleus ambiguus (NA) and the dorsal motor nucleus of the vagus (DMX) that regulate respiration, sympathetic nerve activity, and parasympathetic nerve activity, respectively (26). The central nervous system circuitry and the integration of peripheral and central chemoreceptor reflexes are described in recent reviews (11, 26, 131, 353–355).

The primary physiological functions of CB chemoreceptors are to regulate CO₂ and pH in arterial blood, defend against decreases in arterial O₂ levels (hypoxemia), and optimize O₂ delivery to tissues during diverse physiological, pathological, and environmental challenges. Activation of chemoreceptors induces reflex increases in the frequency and depth of breathing, cardiovagal (parasympathetic) activity, and peripheral sympathetic activity, along with a host of cardiovascular, endocrine, gastrointestinal, renal, and metabolic effects (2, 26, 30, 282, 356, 357) (FIGURE 4). Direct vascular actions of hypoxemia decrease vascular resistance, thereby opposing the sympathetic-mediated vasoconstriction (282, 358). Hypoxia-induced increases in ventilation and cardiac output and vasodilation are adaptive in that they facilitate O₂ delivery to tissues. In addition, metabolic demand in tissues is decreased, thereby reducing O₂ consumption (359). The latter effect is positively correlated with resting metabolic rate of the species and therefore prominent in small mammals such as mice but small in larger mammals including humans (359).

FIGURE 4.

Illustration of the widespread, multiorgan system effects mediated by stimulation of peripheral arterial chemoreceptors: schematic representation of most of the reflex responses that can be evoked by stimulating the carotid body chemoreceptors in a number of mammalian species. These include hematological, renal, gastrointestinal, endocrine, metabolic, and behavioral effects. BAT, brown adipose tissue; GFR, glomerular filtration rate; GI, gastrointestinal. (Reprinted from Ref. 30, with permission from Physiology).

The direction and magnitude of the blood pressure response depend on multiple factors, including the severity and duration of the hypoxemia, the balance between sympathetic-mediated vasoconstriction and the direct hypoxemia-induced vasodilation, the magnitude of the ventilatory response (see sect. 6.1), and the species studied. Although blood pressure typically increases slightly during systemic hypoxia in larger mammals including humans, it decreases significantly in small mammals like mice (360). Interestingly, the magnitude of the decrease in blood pressure varies between inbred strains of mice (360). The mechanisms underlying the variable change in blood pressure between species and strains of mice are not known. A combination of genetic and physiological factors is likely involved. The ventilatory response to hypoxia also varies between different inbred strains of mice (361) and could play a role. Differences in the strength of hypoxia-induced vasodilation, reflex sympathetic-mediated vasoconstriction, and baroreflex sensitivity might also contribute to the different blood pressure responses. These knowledge gaps require further investigation.

It is important to appreciate that the reflex responses to chemoreceptor activation strongly depend on the underlying cause and nature of the blood gas abnormalities. For example, hypoxemia may result from low levels of atmospheric O₂ (as occurs at high altitude), decreased respiratory drive and hypoventilation, ventilation-perfusion mismatch, or cessation of breathing (apnea). Causes of the latter three conditions include respiratory disease, trauma, and drug overdose. The various conditions require very different reflex responses to optimize organ system function and survival, which is achieved in large part via chemoreflex interactions with other reflexes.

6.1. Ventilation Alters Reflex Responses to Chemoreceptor Activation

Voluntary apnea preceded by breathing a hypoxic gas mixture rapidly reverses tachycardia to bradycardia and markedly increases muscle sympathetic nerve activity in humans (357, 362). These findings convincingly demonstrate the powerful influence respiration has on the autonomic and cardiovascular responses to hypoxia. The mechanisms underlying these results have been investigated extensively in animal models [see Eyzaguirre et al, (363) and Abboud and Thames (364) for reviews of earlier work]. Selective stimulation of CB chemoreceptors with localized exposure to hypoxic blood in anesthetized, mechanically ventilated dogs (in the absence of an increase in ventilation) evokes a powerful reflex, parasympathetic-mediated bradycardia accompanied by increases in peripheral sympathetic nerve activity and vasoconstriction in muscle and viscera that maintains or increases arterial blood pressure (365–367) (FIGURE 5, left). Similar results have been demonstrated in other mammals. The responses are adaptive in that the decreased heart rate reduces cardiac work (and hence cardiac oxygen need) and the vasoconstriction redistributes blood flow to the heart and brain. In contrast, in spontaneously breathing animals, selective activation of CB chemoreceptors and systemic hypoxemia each increase heart rate, an effect attributed to lung inflation-induced reflex inhibition of parasympathetic activity to the heart (365, 366) (FIGURE 5, right). The increased ventilation also attenuates the chemoreflex-induced increases in sympathetic nerve activity and vascular resistance (367) (FIGURE 5). The importance of increased ventilation in mediating hypoxia-induced tachycardia in humans has been challenged (368–370). Recent studies have provided evidence that the tachycardia may be independent of the ventilatory response to hypoxia, is mediated by parasympathetic inhibition, and is unlikely to be baroreflex mediated (371, 372) (FIGURE 6).

FIGURE 5.

Influence of ventilation on autonomic and cardiovascular responses to chemoreceptor stimulation. Left: under conditions when breathing (ventilation) is decreased or absent, activation of chemoreceptors evokes powerful parasympathetic-mediated bradycardia accompanied by sympathetic-mediated vasoconstriction. Right: in contrast, during exposure to hypoxia (e.g., at high altitude), the large increase in ventilation reflexively inhibits both parasympathetic and sympathetic activity. Neurogenic-induced vasoconstriction is opposed by direct vasodilator effects of hypoxemia. Art, arterial; SNA, sympathetic nerve activity; TPVR, total peripheral vascular resistance.

FIGURE 6.

Hypoxia-induced vagal withdrawal and tachycardia are independent of the hypoxic ventilatory response in young healthy men. Exposure to hypoxia (10.5% O₂) increased heart rate (HR) despite volitional suppression of the ventilatory [minute ventilation ($\dot{\text{V}}\text{E}$)] response. Hypoxia decreased total peripheral resistance (TPR) and did not change mean arterial pressure (MAP). Sympathetic-mediated increases in HR were blocked by prior iv administration of the β-adrenergic receptor blocker propranolol. Reprinted from Ref. 372, with permission from Journal of Applied Physiology.

Whereas the ventilatory response to hypoxia is mediated primarily by the CB chemoreceptor reflex, autonomic and cardiovascular responses are driven by both CB and aortic body chemoreflexes (282, 363, 373–375). Arterial chemoreceptors are activated by a variety of chemical factors including metabolites produced endogenously in glomus cells during hypoxia. Intravascular administration of one such factor, adenosine, has been shown to activate both CB and aortic chemoreceptor afferents in animals (376–378) and to induce chemoreceptor reflex responses in rats (376) and humans (379–382). Interestingly, bolus injections of adenosine into the common carotid artery of conscious humans caused dose-dependent increases in ventilation and blood pressure that were accompanied by decreases in heart rate (381). In the same subjects, transient systemic hypoxemia induced similar increases in ventilation and blood pressure but increased heart rate (381). These results demonstrate that selective chemical activation of CB chemoreceptors decreases heart rate in humans even in the presence of an increased ventilation and suggest that activation of aortic chemoreceptors may cause or contribute to hypoxia-induced tachycardia (381). The results are consistent with an earlier study showing that hypoxia-induced tachycardia measured in patients who underwent bilateral CB resection ∼25 yr previously for treatment of asthma was enhanced compared with control subjects matched for age and pulmonary function (383). Moreover, the tachycardic response to hypoxia was preserved 1 mo after CB resection in patients with heart failure, whereas reflex increases in ventilation and systolic blood pressure were strongly attenuated (384). These results point to activation of aortic chemoreceptors as the mechanism of the hypoxia-induced increase in heart rate in these CB-resected subjects.

Studies in anesthetized dogs, cats, and other medium- to large-sized mammals have provided abundant evidence that selective stimulation of aortic chemoreceptors elicits reflex increases in heart rate, cardiac contractility, peripheral vascular resistance, and blood pressure (367, 373–375, 385, 386). The methods used to stimulate aortic chemoreceptors in these studies included perfusion of isolated aortic arch with hypoxic blood, injection of chemical stimuli (e.g., cyanide), and electrical stimulation of the aortic depressor nerve. Head-to-head comparisons of responses to CB versus aortic chemoreceptor stimulation were made in many of the studies (373–375, 385). Unlike responses to CB chemoreceptor stimulation, responses to aortic chemoreceptor stimulation were not modulated significantly by respiration (367, 373, 385). Taken together, the evidence points to two major mechanisms underlying hypoxia-induced tachycardia: 1) selective inhibition of CB chemoreceptor-induced cardiac parasympathetic activity by the lung inflation reflex and 2) an aortic chemoreceptor-induced increase in cardiac sympathetic nerve activity and/or inhibition of parasympathetic activity. The relative contribution of each mechanism in various species under various conditions and the influence of disease are poorly understood—a substantial and important knowledge gap that should be addressed in future studies.

Aortic and CB chemoreceptors and the reflex responses to their activation differ in many respects. For example, CBs located bilaterally near the carotid artery bifurcations are highly perfused and sense $\mathrm{P}{\mathrm{a}}_{{\mathrm{O}}_2}$ (282). In contrast, aortic bodies are diffusely located, receive much less blood flow per gram of tissue, and are thought to functionally sense primarily arterial O₂ content (56, 197, 278, 387). Activation of CB chemoreceptors markedly increases ventilation, whereas the influence of aortic chemoreceptors is minimal (282). As discussed above, during apnea or in the absence of changes in ventilation, CB chemoreceptor activation decreases heart rate whereas aortic chemoreceptor activation increases heart rate and myocardial contractility.

The differential effects of CB and aortic body chemoreflexes on ventilation, cardiac function, and vascular resistance raise the question as to how these reflexes are integrated when simultaneously activated in hypoxic states. Presumably, fine-tuning of the reflex responses to best defend specific threats will optimize organ system function and survival. For example, CB chemoreceptors dominate the ventilatory response to hypoxia, which can be lifesaving by increasing O₂ uptake and O₂ delivery to tissues. But if breathing is prevented or compromised by trauma or disease, the priority shifts to maintaining blood pressure and diverting blood flow (and O₂ delivery) to brain and heart, which is achieved by CB and aortic chemoreflex-induced sympathetic-mediated vasoconstriction in muscle and viscera. CB chemoreflex-induced bradycardia is protective in that it decreases myocardial O₂ demand. In addition to arterial oxygenation, it is important to maintain O₂ delivery to tissues (arterial O₂ content × blood flow). The unique ability of aortic chemoreceptors to sense decreases in arterial O₂ content (56, 278) and evoke reflex increases in heart rate and myocardial contractility may provide a means to restore O₂ delivery to peripheral tissues. Carboxyhemoglobinemia and anemia, conditions in which arterial O₂ content is reduced (and arterial Po₂ is approximately normal), have been shown to preferentially activate aortic body chemoreceptors (278, 387). Although it is established that activation of aortic body chemoreceptors evokes reflex changes in cardiovascular function with little or no effect on ventilation, the mechanisms integrating CB and aortic chemoreceptor, lung inflation, and baroreceptor inputs during physiological and pathophysiological challenges are poorly understood and in need of further investigation.

6.2. Reflex Interactions Modulate the Chemoreflex in Health and Disease

The chemoreceptor reflex is acutely modulated by the activity of many types of sensory nerves responding to signals initiated in diverse tissues/organs. By rapidly taking into account multiple inputs, reflex interactions may optimize functional responses to physiological challenges. For example, when submerged in water, stimulation of trigeminal afferents evokes a “diving reflex” resulting in apnea, bradycardia, and sympathetic-mediated vasoconstriction (364, 388). Coactivation of trigeminal and chemoreceptor afferents (the latter by the changing blood gases) enhances the bradycardia and vasoconstriction, whereas the trigeminal-induced apnea overrides the chemoreceptor drive to breathe (364, 388).

The CB chemoreceptor reflex and other sensory inputs/reflexes known to modulate the chemoreflex are illustrated in FIGURE 7. Interactions between the chemoreflex and arterial and cardiopulmonary baroreflexes have been studied extensively (364, 389–391). When arterial blood pressure is low, decreased arterial baroreceptor activity enhances the chemoreceptor reflex, thereby facilitating sympathetic-mediated restoration of arterial blood pressure. Conversely, increased baroreceptor activity attenuates the chemoreceptor reflex. Autonomic and cardiorespiratory responses to exercise involve multiple reflex interactions that vary depending on the type, intensity, and duration of the exercise (392, 393). The primary drivers of the autonomic and cardiovascular responses arise from “central command” and afferent input from the contracting skeletal muscle (392). Afferent input from arterial and cardiopulmonary baroreceptors partially restrains exercise-induced increases in sympathetic activity and blood pressure, although this restraint may be abrogated at the initiation of exercise because of rapid, central resetting of the baroreflex.

FIGURE 7.

Chemoreflex interactions with arterial baroreceptor, cardiopulmonary vagal, cardiac sympathetic spinal, exercise pressor, and chemosensitive renal afferent reflexes are illustrated. Afferent pathways in blue and green are sympathoinhibitory and sympathoexcitatory, respectively. See text for description of the interactions. DRG, dorsal root ganglion; HR, heart rate; SNA, sympathetic nerve activity.

Using acute hyperoxia or intravenous administration of DA to inhibit chemoreceptor activity in healthy humans, Stickland and colleagues (394–396) provided evidence that activation of the chemoreflex contributes to both sympathetic-mediated vasoconstriction during exercise and reflex increases in muscle sympathetic nerve activity and heart rate during activation of muscle metaboreceptors during postexercise muscle ischemia.

In hypertension and heart failure, arterial and cardiopulmonary baroreflexes are impaired, whereas sympathoexcitatory reflexes are often exaggerated, including those elicited by activation of arterial chemoreceptors, cardiac sympathetic afferents, skeletal muscle afferents, and chemosensitive renal afferents (397–399, 558) (FIGURE 7). Importantly, these afferents are sensitized and become tonically active in pathological states. Furthermore, activation of one type of sympathoexcitatory afferent enhances the central neurotransmission and reflex responses to activation of other sympathoexcitatory afferents (397, 399–401). It will be important to understand the mechanisms underlying sensitization and tonic activity of sympathoexcitatory sensory neurons and to delineate the neural circuits that enhance positive-feedback interactions between sensory nerves innervating CB, heart, skeletal muscle, and kidney. Substantial gaps in knowledge remain.

6.3. Chemoreflex-Neurohumoral Interactions in CIH-Induced Hypertension and Metabolic Disease

A potentiated CB chemoreceptor reflex and associated increase in sympathetic nerve activity play major roles in the pathogenesis of sleep apnea-induced hypertension in patients and in animal models subjected to chronic intermittent hypoxia (CIH). Intermittent hypoxia impacts other tissues/organs besides the chemoreceptors and is known to increase circulating levels or induce central nervous system actions of several neurohormones including ANG II, ET-1, vasopressin, and serotonin that contribute to neuro-cardiovascular responses (402–410). CIH also leads to neuroplastic changes within the NTS, lamina terminalis, median preoptic nucleus, RVLM, and paraventricular nucleus of the hypothalamus that enhance the sympathetic activation and hypertension (411–413).

We briefly summarize recent studies demonstrating relevant actions of ANG II produced during exposure of rats to intermittent hypoxia. The peripheral renin-angiotensin system (RAS) is activated during CIH (414), and pretreatment with the ANG II AT₁ receptor blocker losartan prevents the hypertension (415). CB cells express AT₁ receptors, and the afferent nerves are activated by ANG II (416). Losartan (icv) and siRNA knockdown of AT₁ receptors in the subfornical organ (SFO) each decrease blood pressure in CIH-treated rats (412, 417). The effects of ANG II and AT₁ receptor blockade or knockdown on sympathetic nerve activity were not assessed in the above studies. Kim et al. (408) evaluated the contributions of the CB versus the SFO to increases in sympathetic nerve activity after acute intermittent hypoxia (AIH) and intermittent injections of ANG II in anesthetized rats (408). AIH increased sympathetic nerve activity in control rats, an effect that was blocked by losartan (FIGURE 8), attenuated by either pharmacological blockade of SFO or bilateral CB denervation (FIGURE 9), and abolished by the combination of SFO inhibition + CB denervation (FIGURE 9). Intermittent systemic injections of ANG II increased sympathetic nerve activity, thereby mimicking the response to AIH (408). Thus, ANG II acting through CB chemoreflex- and SFO-dependent mechanisms is a major contributor to AIH-induced increases in sympathetic activity. Although studies of AIH in anesthetized animals enable mechanisms driving sympathoexcitation to be refined, additional factors need to be considered in studies of CIH and associated hypertension and end-organ damage in conscious animals.

FIGURE 8.

A and B: arterial pressure (AP) and sympathetic nerve activity (SNA) responses to acute intermittent hypoxia (AIH) in saline-treated (A) and losartan-treated (B) rats. C: group data for the percent change in SNA measured 60 min post-AIH minus SNA measured just before implementing AIH. D and E: acute hypoxia (10% O₂ for 45 s) induced increases in SNA measured pre-AIH and 60 min post-AIH in a saline-pretreated rat and a losartan-pretreated rat (D), with group data shown in E. The hypoxia-induced increase in SNA [area under the curve (AUC)] was normalized to the baseline response. ****P < 0.0001 vs. time control. ††††P < 0.0001 vs. AIH + saline. Reprinted from Ref. 408, with permission from Journal of Physiology.

FIGURE 9.

A: sympathetic nerve activity (SNA) responses to acute intermittent hypoxia (AIH) in control rats (vehicle injected), subfornical organ (SFO)-inhibited rats (SFOiso), carotid body-denervated (CBD) rats, and rats subjected to both CBD and SFOiso. B and C: group data for changes in SNA 60 min post-AIH in sham-operated and CBD rats, and rats treated with vehicle or isoguvacine (iso) injected into SFO, and CBD rats injected with isoguvacine. ***P < 0.001 vs. AIH + sham. ****P < 0.0001. †††P < 0.001. ‡‡P < 0.01. D and E: acute hypoxia induced increases in SNA measured pre- and post-AIH in rats previously injected with vehicle or isoguvacine into SFO (D), with group data shown in E. Reprinted from Ref. 408, with permission from Journal of Physiology.

Interactions between the CB chemoreflex and neurohumoral regulation are also important in metabolic diseases such as diabetes and obesity that are associated with hyperglycemia and insulin resistance (discussed further in sect. 7.4). Neurohumoral systems regulate metabolic as well as cardiovascular functions. For example, hypoglycemia induces increases in sympathetic activity and release of counterregulatory hormones (e.g., epinephrine, norepinephrine, glucagon, cortisol) into the circulation that function to return blood glucose levels back toward normal. Studies in animals and humans using hyperinsulinemic, euglycemic, and hypoglycemic clamps to control and change blood glucose concentration provide evidence that the CB chemoreflex plays an important role in mediating the counterregulatory hormonal responses to hypoglycemia (418–421). Functionally inhibiting chemoreceptor activity in healthy adult human subjects by having them breathe 100% O₂ attenuated the hypoglycemia-induced increases in counterregulatory hormones by ∼50% (420). Unexpectedly, counterregulatory responses to hypoglycemia measured in subjects who had undergone bilateral CB resection for type I tumors years earlier were not significantly impaired (421). Hyperoxia did not attenuate the counterregulatory responses in these subjects, suggesting that redundant mechanisms may compensate for the chronic loss of CB chemoreceptors. The relative contribution of hypoglycemia versus the counterregulatory hormones to activation of CB chemoreceptors remains an active area of investigation. In anesthetized male Wistar rats, insulin-induced hypoglycemia increased basal ventilation and the ventilatory response to CO₂, effects that were abrogated by either beta-adrenergic receptor blockade or adrenalectomy and mimicked by infusion of epinephrine (422). Epinephrine-induced increases in ventilation and CO₂ sensitivity were attenuated by either carotid sinus denervation or beta receptor blockade (422).

Insulin has also been implicated in activation of CB chemoreceptors. Rats fed a high-calorie diet exhibited insulin resistance, increased basal ventilation (measured under anesthesia), and increased ventilatory responses to bilateral carotid occlusion, effects that were abrogated by prior chronic section of the CSN (423). CB responses to hypoxia (in vitro) were also enhanced (423). These investigators also provided evidence that insulin directly activates CB chemoreceptors in vitro and in vivo (423). More recently, increasing blood insulin concentration by hyperinsulinemic-euglycemic clamp increased ventilation in healthy humans, independent of changes in blood glucose (424). Interestingly, brief inhalation of 100% oxygen to inhibit chemoreceptor activity (Dejours test) decreased respiratory frequency in patients with prediabetes, an effect that correlated with insulin resistance (425). Intermittent hypoxia has been shown to increase blood glucose concentration in humans (426) and mice (427), with the effect in mice attenuated by prior CB denervation (427). Additional discussion of neurohumoral mechanisms in hypertension, heart failure, and metabolic disease is provided in sect. 7, along with review of the cellular and molecular mechanisms underlying the excessive CB chemoreceptor afferent activity in these pathological states.

7. CAROTID BODY IN DISEASE PATHOPHYSIOLOGY

Chemodectomas are slow-growing and highly vascular tumors originating from the CB cells. Normally, CB tumors are benign, and treatment is surgical resection (428–431). Other pathological conditions related to the CB include sudden infant death syndrome (199, 432–434) and congenital central hypoventilation syndrome (435–437). However, in the last decade the CB has attracted much clinical interest because new experimental evidence strongly supports a role for a sensitized CB chemoreflex in increasing central sympathetic outflow, an important mechanism underlying several disease conditions (4, 32, 33, 438–444). Indeed, emerging evidence suggests that heightened CB chemosensory discharge in normoxia and in response to low O₂ elicits sympathetic excitation, a hallmark of congestive heart failure (CHF), resistant hypertension, obstructive sleep apnea (OSA), and metabolic disease (FIGURE 10). Oxidative stress, inflammation, ET-1, ANG II, H₂S, and NO have been proposed as possible mediators of CB chemosensory potentiation (32, 169, 410, 444–447).

FIGURE 10.

Contribution of the carotid body to the pathological effects mediated by sympathetic hyperactivation in human diseases. Diagram showing the pivotal role played by the carotid body in the pathological effects mediated by sympathetic hyperactivation in heart failure, obstructive sleep apnea, hypertension, and cardiometabolic diseases. Oxidative stress and inflammation are associated with carotid body chemosensory potentiation, leading to an increase in sympathetic nervous system activity, which promotes autonomic dysfunction, ventilatory instability, baroreflex alterations, hypertension, fibrosis, and insulin resistance. ET-1, endothelin 1; NO, nitric oxide; NTS, nucleus of the solitary tract; PVN, paraventricular nucleus; RVLM, rostral ventrolateral medulla. (Reprinted from Ref. 32, with permission from Journal of Physiology.)

7.1. Obstructive Sleep Apnea and Intermittent Hypoxemia

Obstructive sleep apnea (OSA) is the leading sleep breathing disorder, affects 10–20% of the adult population worldwide, and is associated with daytime somnolence, sleep fragmentation, and cognitive dysfunction (448–450). Nevertheless, OSA is accepted as an independent risk factor for diurnal hypertension and is associated with pulmonary hypertension, stroke, and coronary artery disease (448, 449, 451–453). In addition, OSA patients often present concomitant comorbidities such as obesity and glucose intolerance that increase cardiovascular risk. Indeed, several studies have found that metabolic syndrome is highly prevalent in OSA patients (see for review Refs. 454–456). Indeed, 40–50% of OSA patients show hypertension, with a direct correlation between the apnea-hypopnea index and the prevalence of resistant hypertension (457, 458). During sleep, patients undergo recurrent episodes of complete or partial upper airway obstruction, resulting in hypoxemia and hypercapnia, which in turn stimulate the peripheral chemoreceptors, triggering acute autonomic and cardiorespiratory responses (449, 451, 452, 459, 460). The stimulation of the CB chemoreceptors increases respiratory muscle activity, producing inspiratory efforts, negative intrathoracic pressure, micro-arousals, and restoration of airflow. Chronic intermittent hypoxemia (CIH) is likely the main cause for progression of hypertension (449, 451, 461, 462). Furthermore, CIH is sufficient to enhance hypoxic ventilatory responses and elicit autonomic alterations characterized by sympathetic overflow and reduction of baroreflex sensitivity (BRS) and hypertension in animal models of OSA (169, 440, 446, 463–466). Similarly, OSA patients display augmented sympathetic nerve activity to muscle (MSNA), ventilatory, and blood pressure (BP) responses to hypoxemia, which were attributed to an enhanced peripheral chemoreflex (449, 459, 467). In addition, Fletcher et al. (463) found that bilateral carotid sinus denervation before the onset of intermittent hypoxemia exposure abolished the increase in BP in rats subjected to CIH for 35 days, suggesting that afferent chemosensory inputs from the CB are necessary to raise diurnal BP in the rat CIH model. This concept was strongly supported by recordings of CB neural discharge showing that CIH selectively enhances CB basal chemosensory discharges in normoxia and augments the hypoxic responses in rodents and cats (169, 310, 468). In addition, Peng et al. (447) reported that CIH induces carotid sensory long-term facilitation (sLTF) in rats. Indeed, they found that carotid chemosensitivity increased when the CB superfused in vitro was stimulated by acute intermittent hypoxic episodes in rats previously exposed to CIH. Rey et al. (466) found a linear correlation (r = 0.97) between the low-frequency band of heart rate variability (HRV) and baseline CB chemosensory discharges in cats exposed to CIH for 4 days. Accordingly, these results suggest that potentiation of CB chemosensory discharge is essential for autonomic changes and progression of CIH-induced hypertension. More recently, Del Rio et al. (469) studied the causal correlation between autonomic alterations and enhanced CB chemosensory activity by eliminating the CB chemosensory input to the brain stem in rats exposed to CIH. They performed ablation of both CBs at day 21 of CIH, and rats were kept 7 more days in CIH. CB ablation immediately normalized the elevated BP, restored BRS, and normalized HRV. Accordingly, the hypertension and autonomic alterations induced by CIH depended upon the enhanced CB chemosensory discharge.

The mechanisms underlying the potentiation of CB chemosensory discharge generated by CIH are not completely known. Ortiz et al. (99) studied the consequences of CIH on hypoxia-induced type I cell depolarization and the inhibition of K⁺ TASK channels in adult rats. They found that CIH sped up and increased the amplitude of inhibition of the opening probability of K⁺ TASK channels elicited by hypoxia and increased by threefold the amplitude of the depolarization in the presence of voltage-gated K⁺ channel blockers TEA and 4-AP. Several studies have found that oxidative stress is necessary for the potentiation of carotid chemosensory discharge during CIH (4, 169, 339, 446, 470, 471). Modulators of the CB chemosensory process linked to oxidative stress signaling pathways such as ET-1 (52, 410, 445), ANG II (31, 472, 473), and proinflammatory cytokines (50, 170, 472, 474) participate in chemosensory potentiation. Several studies in OSA patients and animals subjected to CIH confirm that hypoxia-reoxygenation cycles, the main feature of OSA, produce systemic and local oxidative and nitro-oxidative stress (50, 169, 446, 474–479). The molecular response to hypoxia is triggered by hypoxia-inducible transcription factors HIF-1α and HIF-2α (480). In the CB, CIH modifies the balance of the HIF-1α-to-HIF-2α ratio, promoting the upregulation of NADPH and downregulation of manganese superoxide dismutase (MnSOD) and leading to oxidative stress (479). Peng and Prabhakar (446) found evidence that the superoxide radical is crucial for the CB chemosensory potentiation induced by CIH. They administered the SOD mimetic manganese(III)tetrakis(1-methyl-4-pyridyl)porphyrin pentachloride (MnTMPyP) to rats 10 days before and during 10 days of CIH exposure and found that the antioxidant treatment prevented CB chemosensory potentiation. Peng et al. (447) submitted wild-type (WT) and heterozygous mice partially deficient in HIF-1α to CIH for 10 days and found that acute hypoxia enhanced the CB chemosensory responses and acute intermittent hypoxia induced sLTF in the CIH-treated WT mice. MnTMPyP blocked the CIH-induced ROS increase and the upregulation of HIF-1α in WT mice, implying that the activation of HIF-1 was crucial for CB sensory potentiation. Peng et al. (339) proposed that NADPH oxidase also contributes to increased ROS production following CIH. In addition, Schulz et al. (481) reported that knockout of NADPH oxidase 2 (NOX2) in mice reduced the elevated BP following CIH. Furthermore, the antioxidant ascorbic acid applied during CIH prevented systemic and local oxidative stress in the CB, as well as the potentiation of carotid chemosensory discharge, the enhanced hypoxic ventilatory response, and the elevated BP in rats exposed to CIH for 21 days (169). More recently, Moya et al. (482) tested the effects of ebselen, a peroxynitrite scavenger (483), on 3-nitrotyrosine immunoreactivity (3-NT-ir) accumulation in the CB and on CB chemosensory discharge and hypertension in CIH rats. Ebselen administered by osmotic pump prevented the increase of 3-NT-ir in the CB and the enhanced CB chemosensory discharges and was able to reduce the elevated BP in rats subjected to CIH for 2 wk. Therefore, both CIH-induced hypertension and CB chemosensory potentiation were dependent on peroxynitrite formation. Recently, Nanduri et al. (484) found that hypertension, irregular breathing, and enhanced peripheral chemoreflexes were normalized after 30 days of recovery from short-term CIH (10 days). By contrast, all these alterations persisted after recovery if the animals were subjected to CIH for >30 days. Nanduri et al. (485) found evidence that cardiorespiratory alterations were associated with oxidative stress in the CB and the adrenal medulla due to DNA hypermethylation‐dependent suppression of genes encoding for antioxidant enzymes.

It has been proposed that ET-1 mediates CB potentiation elicited by CIH (52, 410, 445, 486). ET-1 is a vasoconstrictor peptide that produces chemosensory excitation in the perfused CB, but not in a superfused preparation, showing a prominent vascular effect. CIH augmented ET-1 immunoreactivity in CBs from cats subjected to CIH for 4 days, whereas bosentan, an unspecific AT₁ receptor blocker, diminished the excitatory response elicited by hypoxia in the CB from CIH-treated cats (410). Similarly, CIH enhanced CB chemosensory responses, augmented ET-1 release from the CB, and upregulated the endothelin A (ET-A) receptor in neonatal rats, effects all attributed to oxidative stress since they were reduced by MnTMPyP (445). More recently, Peng et al. (487) found that CIH increased ET-1 in the CB, resulting from oxidative stress induction of the endothelin‐converting enzyme. Therefore, Peng et al. (487) found that a ROS-dependent increase of ET-1 acting through ET-A receptor contributes to the CB chemosensory potentiation elicited by CIH in neonatal rats. Recently, Li et al. (486) found evidence that PLC, PKC, and p38 MAPK participate in the ET-1-induced CB chemosensory excitation in rats subjected to CIH. Nevertheless, the role played by ET-1 in the long-term CB sensory excitation is unclear. It is likely that the contribution of ET-1 to CB potentiation in adult rats takes place in early stages of CIH, because ET-1 immunoreactive levels in the adult rat CB increased in the first week of CIH but then returned to normal levels during the second week of CIH (473, 474).

Similar to the CB chemosensory potentiation induced by CIH, it is well known that chronic sustained hypoxia (i.e., high altitude) progressively increases carotid baseline chemosensory discharges and responses to hypoxia (488, 489). In addition, long-term exposure to sustained hypoxia produces hypertrophy and/or hyperplasia of type I cells and enlarges the CB (490–493). ET-1 and its receptors have been involved in CB enlargement induced by sustained hypoxia (52, 488, 494). Indeed, ET-1 stimulates cellular proliferation in primary cultures of CB cells, activating the ET-1 receptor type A (495). Hypoxia for 2 wk increases ET-1 mRNA and the expression of the ET receptor type A in rat type I cells (496). More recent, Platero-Luengo et al. (60) revealed the molecular mechanism by which ET-1 produces proliferation in the CB. They found that ET-1, released from type I cells in response to sustained hypoxia, acts on stem glial cells, which are endowed by ET receptors that control their growth. The inhibition of type I cell transmitter release or their selective destruction markedly diminishes CB cell growth during sustained hypoxia. It is interesting to note that exposure to intermittent hypoxia, which increased ET-1 immunoreactive levels and its release from type I cells (474, 487), failed to enlarge the CB. Indeed, Peng and Prabhakar (446) reported that the volume and number of type I cells in the adult rat CB were unaffected by 10 days of CIH, and Del Rio et al. (497) found that exposure to CIH for 21 days did not modify the number of TH-positive cells or the volume of the rat CB. Interestingly, Welch et al. (498) found that CB size is not increased in patients with sleep apnea.

CIH progressively augments TNF-α, IL-1β, and inducible nitric oxide synthase (iNOS) levels in the CB, without macrophage infiltration (497). Contrarily, Lam et al. (472) reported that increased IL-1β, TNF-α, and IL-6 in the CB from CIH-rats was associated with macrophage infiltration, which was reduced with dexamethasone or ibuprofen. Del Rio et al. (470) examined whether CIH may induce an increment of cytokine levels in the CB and whether these cytokines contribute to enhance CB chemosensory discharges. They found that ibuprofen prevented overexpression of TNF-α and IL-1β but failed to moderate the enhanced CB chemosensory responses to hypoxia. Interestingly, ascorbic acid reduced both the augmented CB chemosensory responses and the TNF-α and IL-1β levels in the CB, suggesting that the enhanced CB chemosensory responsiveness to hypoxia depended on the oxidative stress (499). In addition, Del Rio et al. (470) found that the enhanced ventilatory response to hypoxia and the hypertension induced by CIH were reduced by ibuprofen, suggesting that multiple mechanisms involving cytokines may contribute to the cardiorespiratory alterations induced by CIH. Indeed, ibuprofen reduced the number of c-fos-positive neurons in the caudal NTS from rats subjected to CIH for 21 days, suggesting that neuroinflammation may be involved in the central ventilatory plasticity induced by CIH (170).

The renin-angiotensin system (RAS) has been also involved in the enhanced CB chemosensory reactivity and local inflammation acting on functional AT₁ receptors. Indeed, ANG II stimulated functional AT₁ receptors and increased intracellular Ca²⁺ in rat type I cells (31), whereas CIH increased angiotensinogen and AT₁ receptor expression and potentiated [Ca²⁺]_i in response to exogenous ANG II in chemoreceptors (500). Losartan, an AT₁ blocker, reduced the increased Ca²⁺ in response to ANG II, as well as gp91 expression and macrophage infiltration in the CB from CIH-treated rats. It is known that the RAS plays a pivotal role in the hypertension induced by CIH. Indeed, CIH increased plasma renin activity by fourfold compared with CIH rats that underwent bilateral renal artery sympathectomy (501). Thus, it is likely that the increase in BP is mediated in part by the activation of the renal sympathetic system that increases circulating ANG II in CIH.

Another molecule that may impact CB chemosensory potentiation is NO, production of which is reduced by CIH (22, 474, 502). At low physiological concentrations NO is an inhibitory tonic modulator of CB chemosensory discharges (265, 503–505). Exposure of rats to CIH for 7 days decreased CB endothelial nitric oxide synthase (eNOS) and nNOS immunoreactivity (474). In addition, CIH reduced NO production (22) in the CB of rats subjected to CIH. NO reacts with the superoxide radical to produce peroxynitrite, increasing levels of 3-NT-ir in the CB of CIH-rats (169, 474). High levels of peroxynitrite and NO, which inhibit oxidative phosphorylation (344), may account for the increased CB chemosensory discharge. Del Rio et al. (169) noted increased levels of 3-NT in the CB of rats subjected to CIH that paralleled the CB potentiation, suggesting that nitrosative stress plays a critical role in CB chemosensory potentiation in CIH (462).

Yuan et al. (341) proposed that CO produced by heme oxygenase-2 (HO-2) in the CB may inhibit H₂S production catalyzed by cystathionine-γ-lyase (CSE), reducing CB chemosensory activity. Oxidative stress inhibits CO generation in the CB through a mechanism that requires Cys in the heme regulatory motif of HO-2. ROS induced by CIH inhibited CO production and increased H₂S levels in the CB, increasing chemosensory discharge. Blockade of H₂S synthesis by CSE with pharmacological or genetic methodologies in mice inhibited the CB potentiation and the hypertension induced by CIH (341). Peng et al. (275) examined whether H₂S contributes to enhancing CB chemosensory activity induced by CIH. Both CIH-induced CB chemosensory potentiation and the sLTF were abolished by l-propargylglycine (L-PAG), an inhibitor of CSE. Interestingly, the hypertension, the sympathetic overflow, and the elevated plasma catecholamine levels induced by CIH were absent in L-PAG-treated rats and CSE-null mice, suggesting that oxidative stress-dependent activation of CSE-derived H₂S synthesis may contribute to the CB chemosensory response potentiation and to the hypertension.

OSA is associated with several comorbidities including impaired glucose metabolism. Shin et al. (427) proposed that the CB is involved in glucose intolerance and insulin resistance induced by chronic CIH. Accordingly, they performed carotid sinus denervation in mice that were exposed to CIH for 4–6 wk. CIH in the control mice increased fasting blood glucose, baseline hepatic glucose output (HGO), and the levels of the rate-limiting hepatic enzyme of gluconeogenesis phosphoenolpyruvate carboxykinase (PEPCK). Carotid sinus denervation prevented the CIH-induced fasting hyperglycemia and increases of HGO and liver PEPCK levels. In addition, carotid sinus neurotomy reduced the enhanced serum epinephrine levels and liver sympathetic innervation. Thus, the increased baseline HGO levels and the sympathetic overflow induced by CIH depend on the integrity of the CB input.

7.2. Congestive Heart Failure

Congestive heart failure (CHF) with reduced ejection fraction has been associated with sympathetic excitation and parasympathetic withdrawal, increased sympathetic-respiratory coupling, and irregular breathing patterns. These alterations have been attributed to an enhanced peripheral chemoreflex responsiveness (35, 61, 398, 444, 506, 558). Schultz and colleagues (507–512) reported increased CB chemosensory discharge in normoxia and enhanced responses to acute hypoxia in rabbit and rat heart failure models. CB ablation or CSN neurotomy reduced sympathetic-respiratory coupling, renal sympathetic neural discharge, apnea/hypopnea incidence, and ventilatory instability (507, 510). In addition, bilateral CB ablation strikingly increased survival time in young CHF rats (507). More recently, Fujii et al. (513) found that bilateral carotid sinus neurotomy in a hypertensive heart failure-Dahl salt-sensitive rat significantly reduced 24-h urinary norepinephrine and attenuated the elevated BP. Thus, in this hypertensive model, the elimination of CB inputs attenuated myocardial hypertrophy and improved rat survival rate. Similar to intermittent hypoxia, oxidative stress, ANG II, and NO play an important role in CB potentiation in CHF rat and rabbit models. Indeed, oxidative stress produced by CHF in rabbits inhibited voltage-gated K⁺ channels, which in turn depolarize type I cells, enhancing CB excitability (514). CB chemosensory potentiation in CHF models has been attributed primarily to a reduction of CB blood flow produced by the low cardiac output. Carotid blood flow reduction evoked oxidative stress and decreased NO production, by reduction of nNOS levels in the CB. The genetic transference of nNOS to CB cells with an adenoviral vector augmented NO levels in the CB of CHF rabbits and normalized the elevated chemosensory discharge (508). The transcription factor Krüppel-like factor 2 (KLF2), which inhibits angiotensin-converting enzyme (ACE) and induces the expression of eNOS in endothelial cells, was diminished in the CB of CHF rabbits (25). Remarkably, KLF2 genetic transference to the CB with an adenovirus reduces respiratory instability and normalizes the enhanced hypoxic ventilatory response in CHF rabbits (25). In addition, genetic transference of nNOS reduced inhibition of voltage-dependent K⁺ currents in chemoreceptor cells from CHF rabbits. Therefore, the available evidence underscores the importance of the role played by enhanced CB chemosensory discharge and heightened chemoreflex sensitivity in the pathophysiology and progression of autonomic and cardiorespiratory alterations in CHF.

7.3. Hypertension

The idea that the CB contributes to neurogenic hypertension is not new. Indeed, spontaneously hypertensive rats (SHRs) show enlarged CB (515) and display higher ventilatory responses to hypoxia (516). Similarly, mildly hypertensive young subjects manifest potentiated ventilatory and cardiovascular responses to hypoxia (517, 518). Experiments performed in untreated hypertensive patients have shown that breathing 100% O₂ (Dejours test) reduced the elevated resting sympathetic nerve activity to muscle (MSNA) and BP, suggesting an enhanced peripheral chemoreflex (519). Moreover, recordings of CB chemosensory discharge showed that SHRs display an augmented CB chemosensitivity to hypoxia, whereas the chemosensory responses to hypercapnia were similar to the Wistar-Kyoto control rats (520). In contrast, Tan et al. (157) reported that low pH elicited a high CB chemoreceptor cell depolarization in SHRs compared with Wistar-Kyoto control rats. In addition, they noted that the increase in sympathetic thoracic nerve activity in response to CB stimulation with NaCN in the rat working heart-brain stem preparation was enhanced in normotensive young SHRs compared with control rats (157). These observations suggest that before the onset of hypertension the CB displayed enhanced responses to hypoxia, NaCN, or acid stimuli. More recent evidence indicates that the CB contributes to increasing sympathetic activity and BP in preclinical models of hypertension. Indeed, bilateral CSN neurotomy in SHRs reduced the elevated BP, attenuated the enhanced renal sympathetic discharge, and improved BRS (441, 521). Abdala et al. (521) examined whether CSN chemosensory discharge was responsible for the elevated sympathetic activity and the progression of hypertension in young SHRs. They denervated the CSN in 4- and 12-wk-old SHRs and found that bilateral CSN denervation delayed the onset of hypertension in young SHRs and lowered the elevated BP in adult SHRs. Interestingly, unilateral CSN denervation in adult SHRs failed to reduce BP (521). These data are consistent with a significant contribution of CSN activity to increased arterial BP.

Pijacka et al. (522) found that bilateral carotid sinus denervation reduced BP in Goldblatt hypertensive (two kidney one clip: 2K1C) rats. Carotid sinus denervation performed after 5 wk from the renal artery clipping reduced the elevated BP in 2K1C rats. Three weeks after the surgery, BP continued to increase in the sham 2K1C group, whereas in 2K1C rats that underwent elimination of CB afferents mean arterial pressure decreased by 6 mmHg. Additionally, carotid sinus neurotomy improved BRS and cardiac autonomic balance, normalized proteinuria and albuminuria, and decreased [Ca²⁺]_i-evoked responses in sympathetic stellate ganglia neurons. Thus, normalization of the enhanced peripheral chemoreflex reactivity may be a useful method for controlling renovascular hypertension.

The mechanisms underlying the contribution of the CB to hypertension are not known but seem to be related to overactivation of P2X₃ receptors in petrosal identified chemosensory neurons (523, 524). Indeed, Pijacka et al. (524) found that mRNA expression of purinergic receptors was upregulated in chemosensory petrosal sensory neurons from SHRs. The specific blockade of P2X₃ receptors in petrosal neurons reduced BP and basal sympathetic activity in SHRs. Using the in situ arterially perfused preparation, they found that both enhanced carotid sinus nerve baseline discharge and responses evoked by NaCN were higher in SHRs compared with normotensive control rats. More recently, Moraes et al. (523), using an in vitro preparation of the CB and petrosal ganglion, found that electrical properties of petrosal neurons were different between identified chemoreceptive and baroreceptive neurons in SHRs and normotensive rats. Chemoreceptive neurons from SHRs were more depolarized and exhibited enhanced responses to NaCN, whereas baroreceptive petrosal neurons showed similar electrophysiological properties in relation to neurons from normotensive rats. Furthermore, they found that chemosensory neurons from SHRs showed enhanced firing rate responses to α,β-methylene ATP, which were blocked by AF-353, a selective P2X₃ receptor blocker. Thus, ATP responses mediated by P2X₃ receptors in petrosal chemoreceptive neurons play a critical role in SHRs, supporting the concept that targeting ATP-induced chemosensory potentiation may be useful to control hypertension.

7.4. Metabolic Diseases

A new line of research connecting the CB and the sympathetic nervous system with metabolic diseases arises from the observation that the CB participates in insulin resistance (423, 439, 525). Conde and colleagues (439, 455, 525) proposed that intermittent hyperinsulinemia caused by a hypercaloric diet potentiates CB chemosensitivity and activates the sympathetic-adrenal axis, which impairs glucose uptake by the liver and adipose tissue, increasing glucose and free fatty acids in plasma. Consequently, the pancreas secretes more insulin to counteract the hyperglycemia, enhancing the insulin-dependent CB chemosensory potentiation. Ribeiro et al. (423) reported that rats fed with hypercaloric diets showed increased basal ventilation and ischemic responses induced by bilateral common carotid artery occlusions. In addition, they noted that basal release of dopamine from the CB was not modified by hypercaloric diets, but the release induced by hypoxia (5% O₂) was increased in rats fed with hypercaloric diets. Conde et al. (439) further observed that rats fed a high-fat diet display enhanced CB chemosensory responses to hypoxia. In addition, they found that carotid sinus denervation prevented the development of insulin resistance and hypertension (423). Therefore, it is likely that insulin may induce CB chemosensory excitation contributing to sympathetic activation, creating a positive feedback that in turn promotes insulin resistance and elevates BP. Sacramento et al. (526) showed that bilateral carotid sinus neurotomy reversed the insulin resistance, reduced body weight, and attenuated autonomic dysfunction and hypertension in rats fed a high-fat diet. Surgical elimination of the CB afferents normalized glucose and insulin levels by improving glucose uptake. Thus, carotid sinus neurotomy restored normal metabolic function. These results suggest beneficial effects on insulin and glucose regulation following the elimination of CB chemosensory drive, indicating a putative therapeutic use for elimination of CB input in metabolic diseases. On a related note regarding blood pressure control, Nogueira et al. (527) studied the effect of CB ablation on ventilatory responses and progression of hypertension in offspring of malnourished rats. They fed rats with a low-protein (8% casein) diet during pregnancy and lactation. At 29 days of age, the progeny underwent CB ablation. In the long term (90 days), CB ablation reduced mean arterial blood pressure by ∼20 mmHg in offspring of malnourished rats and reduced the low-frequency component of arterial blood pressure variability.

The studies reviewed above demonstrate that CB chemoreceptor activity and/or sensitivity to hypoxia are enhanced in animal models fed high-calorie diets and that the CB chemoreceptors contribute in a major way to the subsequent development of insulin resistance, increased sympathetic activity, and hypertension. Recently, Cunha-Guimaraes et al. (425) studied the correlation between CB reflex responses with dysmetabolic clinical variables in human prediabetic patients. They found that insulin resistance was correlated with peripheral chemosensitivity, assessed by the Dejours test, and with abdominal circumference. Thus, the CB appears to be overactive in prediabetic subjects, and peripheral chemosensitivity correlates with fasting insulin and insulin resistance. These results point to a corresponding increase in ventilation that has important implications for patients with regard to acid-base balance and the comorbidities of sleep apnea and periodic breathing. Although increased ventilation is consistent with increased CB chemoreceptor activity, additional mechanisms must be considered. For example, additive and synergistic reflex interactions between CB chemoreceptor stimulation (with hypoxia or hypercapnia) and central chemoreceptor stimulation with hypercapnia enhance the increases in ventilation in healthy awake dogs (354, 528, 529). Moreover, central nervous system hypoxia can stimulate respiration in physiological preparations, with accumulating evidence supporting a role of ATP released from astrocytes in key brain stem sites (354, 355, 530–533). Finally, AMPK has been implicated in both CB chemoreceptor activation (123, 124) and hypoxia sensing in the brain stem respiratory network (11, 131, 132), with the latter hypothesis prevailing (see discussion in sect. 3.3). AMPK deficiency blocks the hypoxic ventilatory response but does not block CB chemoreceptor activation by hypoxia (131). Obesity is associated with decreased AMPK expression/activity, suggesting that hypoxia sensing in the brain stem may be blunted in obesity, thereby opposing the increased CB chemoreceptor drive to breathe. This imbalance might increase risk of periodic breathing or apnea. Studies of the central mechanisms integrating hypoxic stress discussed above are lacking in models of metabolic disease, and future studies are encouraged to fill this gap in knowledge.

Leptin, a pleiotropic peptide hormone mainly secreted by adipocytes, acts on the hypothalamus to regulate energy metabolism, food intake, and satiety (534). In addition, leptin is a potent breathing stimulant, acting mainly on the central nervous system but also at the level of the CB (for a recent review see Ref. 535). Groeben et al. (536) were the first to study the contribution of peripheral chemoreceptors to the ventilatory response induced by leptin by exposing leptin-deficient (ob/ob) and wild-type C57BL/6J mice to 100% fraction of inspired oxygen ($\mathrm{F}{\mathrm{I}}_{{\mathrm{O}}_2}$). They found that hyperoxia reduced the respiratory frequency in wild-type mice but not in ob/ob mice, an effect that was restored by administration of leptin. The proposal that leptin acts at the CB level was further supported by the findings that leptin and leptin receptor (Ob-R) isoforms are expressed in type I cells from human and rat CB (537). Yuan et al. (112) noted that Zucker mice lacking Ob-R showed reduced baseline and ventilatory responses to hypoxia compared with chow-fed lean or high-fat diet-fed littermates. In addition, they observed that phosphorylated Signal transducer and activator of transcription 3 (pSTAT3) factor and the TASK-1, TASK-3, and TASK-2 channels were significantly reduced in the CB of obese Zucker rats compared with the other two phenotype littermates. Moreover, they found that chronic administration of leptin in chow-fed lean Zucker rats did not modify basal ventilation but enhanced the hypoxic ventilatory response and the expression of pSTAT3. Furthermore, CB ablation abolished the enhanced ventilatory response to leptin. Systemic leptin administration was associated with enhanced expression of TASK channels in type I cells, suggesting that leptin may potentiate the hypoxic ventilatory responses, probably due to the increased expression of pSTAT3 and TASK channels in the CB. Caballero-Eraso et al. (538) provide further evidence for the contribution of the CB in the leptin-induced ventilatory response. They studied the effect of subcutaneous leptin infusion on baseline minute ventilation and the hypoxic ventilatory response to 10% O₂ in C57BL/6J mice before and after CB denervation. They found that leptin enhanced CB chemosensory baseline drive and responses to hypoxia, whereas this effect on baseline ventilation and hypoxic response was abolished by CB denervation. In addition, they reported in Ob-R-deficient obese mice that the adenovirus-mediated restoration of the Ob-R in both CBs increased both baseline ventilation and the hypoxic ventilatory response, without affecting food intake, circulating leptin levels, and body weight. Similarly, new evidence implicates the leptin signaling pathway in the CB in obesity-induced hypertension. Indeed, Shin et al. (539) found that overexpression of the leptin receptor in the CB of Ob-R-deficient obese mice induced hypertension. They reported that CB denervation reduced the increase of BP induced by leptin administration and that the Ob-R overexpression in the CB of Ob-R-deficient mice enhanced gene expression of the transient receptor potential melastatin 7 channel (TRPM7) and led to hypertension. These results suggest that leptin in the CB may contribute importantly to cardiovascular and respiratory function.

In summary, leptin plays a modulatory role in CB-mediated responses to hypoxia. Clearly, leptin increases basal ventilation and hypoxic responses, effects that were abolished by CB ablation or carotid sinus denervation. However, in a preclinical model of metabolic syndrome, the effects of leptin on CB chemosensory discharges and ventilatory regulation are blunted. Messenger and Ciriello (540) found that CIH underregulated the leptin receptors, although the expression of leptin within the rat CB is upregulated by CIH, suggesting a complex interaction between leptin and CIH in the CB. Ribeiro et al. (541) studied the effects of leptin on the expression of Ob-R, CB neurotransmitters involved in leptin signaling, CB chemosensory discharges, and ventilation in rats fed a high‐fat diet. They found that in the rat high-fat diet model the effect of leptin on ventilatory control was blunted compared with control rats. Indeed, rats fed with a high-fat diet (60% energy from fat) for 3 wk, despite having an enhanced basal ventilation, showed a large reduction in the excitatory effect of leptin on ventilation, which was unaltered by CSN neurotomy. In addition, they found that leptin did not modify intracellular Ca²⁺ in type I cells but increased the release of adenosine from the CB (541). Thus, these results suggest that leptin contributes to CB chemosensory potentiation in the early stages of metabolic dysfunction that led to that hyperleptinemia but that leptin resistance and blunted responses develop with chronic metabolic dysfunction.

8. CAROTID BODY MODULATION—THERAPEUTIC STRATEGIES

8.1. CB Ablation in Preclinical Models

The available evidence indicates that an enhanced CB chemoreceptor drive, which elicits autonomic dysfunction, is a cardinal feature of OSA, resistant hypertension, and congestive heart failure. In preclinical models of these diseases, the enhanced CB chemosensory neural activity contributes to sympathetic overactivation, impairs HRV and BRS, and causes respiratory instability and hypertension. Several preclinical studies have demonstrated that bilateral ablation of the CB or bilateral carotid sinus neurotomy ameliorated the enhanced sympathetic outflow, restored autonomic balance and BRS, lowered BP, reduced ventilatory instability, improved survival rate, and restored insulin tolerance in animal models of sympathetic-mediated diseases (FIGURE 11). These findings emphasize the importance of enhanced CB chemosensory drive in the progression of autonomic alterations and support the concept that removal of CB afferents may be useful in improving cardiovascular and endocrine dysfunction in sympathetic-related diseases.

FIGURE 11.

Effects of carotid body (CB) ablation or carotid sinus nerve denervation in preclinical models of sympathetic-related diseases. Spontaneous hypertension: Prevents increases in blood pressure (BP) in young spontaneously hypertensive rats (SHRs). Reduces increased BP and sympathetic activation and improves baroreflex sensitivity (BRS) in adult SHRs (521). Congestive heart failure: Normalizes autonomic and BRS alterations, reduces apnea and arrhythmias. Attenuates myocardial deterioration and increases rat survival rate (367). Reduces enhanced renal sympathetic discharges, reduces respiratory instability and arrhythmia incidence in rabbit (509). Hypertensive and natriuretic responses to sodium load: reduced enhanced BP response and natriuresis induced by hypertonic NaCl infusion (542). Renovascular hypertension: Reduces increased BP, improves BRS and autonomic balance in rats (522). Intermittent hypoxia [obstructive sleep apnea (OSA)]: Normalizes increased BP after 21 days of chronic intermittent hypoxia (CIH) in rats. Restores autonomic balance and BRS and reduces arrhythmia incidence (469). Prevents hyperglycemia and high hepatic glucose output in mice. Reduces increased levels of catecholamines in plasma (427). High-fat and -carbohydrate diets: Normalizes sympathetic overflow and increased BP. Decreases weight gain, normalizes plasma glucose and insulin levels. Improves endothelial function (423). Protein-malnourished diets: Normalizes increased BP in offspring of protein-restricted rats (527). Heart failure—Dahl salt-sensitive hypertension: Reduces increased BP and 24-h urinary norepinephrine levels, improves myocardial function and rat survival rate (513).

8.2. Resection of the CB in Patients with Resistant Hypertension or Systolic Heart Failure

Results from preclinical animal models reveal an important role of enhanced CB chemosensory drive in autonomic dysfunction, supporting the hypothesis that CB ablation or CSN denervation may improve autonomic dysfunction, ventilatory instability, and elevated BP (441, 443, 507, 543, 544). The proposition that surgical CB resection is a suitable method to treat consequences of human-related sympathetic diseases was recently tested in patients with resistant hypertension or systolic heart failure (FIGURE 12). Narkiewicz et al. (545) performed unilateral CB resection in patients with resistant hypertension (systolic and diastolic BP > 180 and 100 mmHg, respectively). Only 57% of the patients displayed transient hypotension lasting for 3–6 mo, but after a year the unilateral CB resection failed to maintain a stable reduced BP, and the pressure returned to the previous values. The responder patients had a reduction of MSNA, which paralleled the BP fall. Results from another study using an endovascular venous catheter to destroy the CB with microwaves was reported in abstract form (546). Unilateral CB destruction reduced systolic and diastolic BP by 7 and 5 mmHg, respectively. Niewinski et al. (35) assessed the effects of unilateral and bilateral CB ablation on sympathetic outflow and life quality of CHF patients. They resected one CB from six patients and both CBs in another four patients. Both unilateral and bilateral CB resection diminished the ventilatory response and MSNA response to hypoxia, but quality of life showed only a transient recovery. Importantly, patients with bilateral CB ablation developed nocturnal oxygen desaturation.

FIGURE 12.

Main effects of carotid body ablation (CBA) in humans with resistant hypertension or congestive heart failure. Resistant hypertension: First-in-man study to test the safety and feasibility of unilateral CBA in 15 patients with drug-resistant hypertension. No overall reduction in blood pressure (BP) was found. However, 8 patients showed transient reductions in mean BP for 6 mo and decreases in sympathetic nerve activity to muscle (MSNA) (545). Resistant hypertension: Preliminary data from 15 patients. Transvenous catheter-based unilateral right CBA. At 6 mo systolic and diastolic mean BP were reduced by 10 ± 15 and 4 ± 7 mmHg, respectively (546). Preliminary data from 10 patients. The 24-h ambulatory BP 152 ± 11/89 ± 12 mmHg (systolic/diastolic) was reduced by 9 ± 9/4 ± 6 mmHg at 1 mo after CBA. Ambulatory BP continued reduced at 6 mo by 10 ± 15/4 ± 7 mmHg (547). Congestive heart failure: First-in-man study to test the safety and feasibility of unilateral right-sided CBA in 4 patients and bilateral CBA (BCBA) in 6 patients with CHF. At 1 mo, both MSNA and ventilatory chemoreflexes to hypoxia were reduced. Quality of life and fatigue scores showed transient improvement at 1–2 mo (35).

8.3. CB Ablation or Pharmacological Modulation?

Clinical studies performed in hypertensive or CHF patients showed that unilateral CB ablation failed to produce a stable reduction of BP in resistant hypertension, although patients displayed a constant reduction of the hypoxic ventilatory response and baseline MSNA. Accordingly, results from clinical and preclinical studies suggest a likely contribution of the CB to progression of autonomic dysfunction in sympathetic-related diseases. Elimination of the chemosensory inputs from one CB is not sufficient to produce permanent normotension, and it seems to be necessary to eliminate both CBs.

The surgical CB resection developed by Nakayama was extensively used in Japan from 1945 to 1960 to relieve the perception of dyspnea associated with bronchial asthma in young subjects. Nakayama (548) reported the effects of unilateral or bilateral CB resections to relieve the sensation of air hunger in a large number of patients (>3,900) with bronchial asthma. He found that patients with severe asthma showed some benefit after the end of 2 yr. In addition, he reported that bilateral CB ablation did not change BP in 596 normotensive patients, whereas it reduced BP in 29 hypertensive patients and BP increased in 21 hypotensive patients. In the same year, Overholt (549) published the first study of CB resection in 69 patients in the United States. Following the contemporaneous publication of these papers, thousands of patients with severe asthma underwent CB resection. The results of this procedure were contradictory, and after a large controversy between proponents and opponents, reports from a Task Force of the Division of Lung Diseases, National Institutes of Health and the Executive Committee of the American Academy of Allergy and Immunology (550) concluded that even bilateral CB resection did not appear to be an effective treatment for asthma. After 20 yr, Honda had the opportunity to study the ventilatory hypoxic responses in Japanese patients who underwent CB resection and found that bilateral CB ablation blunted the reflex response to hypoxia and caused a severe depression of CO₂ chemosensitivity (158, 551). Thus, bilateral resection of the CBs in humans resulted in a permanent elimination of the hypoxic ventilatory responses in normocapnic conditions (552). Since the main role played by the CB chemoreceptor in humans is to control respiratory gases, several concerns regarding the potential deleterious effect of bilateral CB ablation have been raised (354, 451, 543, 553, 554). Rigorous and robust prospective longitudinal studies are required to assess the effects of CB ablation in diseases such as congestive heart failure and resistant hypertension, and perhaps also in OSA and metabolic syndrome. Pharmacological studies are also needed to target the mechanisms underlying enhanced CB chemosensory discharge, offering the opportunity for nonsurgical modulation of CB activity. TABLE 3 shows the proposed pharmacological agents and maneuvers for attenuating enhanced CB chemosensory activity in preclinical models of autonomic-related diseases. Therefore, reduction of enhanced CB chemosensory discharge seems to be useful in attenuating the pathological consequences of sympathetic-related diseases (FIGURE 13).

Table 3.

Proposed treatments/maneuvers for attenuating enhanced CB chemosensory discharge in preclinical models of autonomic-related diseases

Treatment	Mediator	Reference
Antioxidants	Oxidative stress	169, 339, 446, 482
ANG II type 1 receptor (AT1) blocker	ANG II via AT1 receptor	31, 408, 555
NO donors	NO	22, 508
Endothelin A (ET-A) receptor blocker	ET-A receptor—ROS dependent	410, 445, 487
Anti-inflammatory	Proinflammatory cytokines—ROS dependent	170, 499
Cystathionine γ-lyase inhibitor	H2S	267, 275
Purinergic receptor blocker	Purinergic receptors in petrosal neurons	522,523
Bioelectronic modulation	Carotid sinus nerve discharge block	556

CB, carotid body; NO, nitric oxide; ROS, reactive oxygen species.

FIGURE 13.

Modulation of carotid body (CB) chemosensory responsiveness to oxygen to restore autonomic, cardiorespiratory, and metabolic balance in sympathetic-related diseases. Nonsurgical modulation of enhanced CB chemosensory discharge is a possible alternative to CB ablation or carotid sinus neurotomy to restore autonomic balance and normalize chemoreflex and baroreflex function, arterial blood pressure, endothelial function, and metabolism (see Refs. 32, 423, 441, 443, 469, 507, 543, 544, 553).

9. CONCLUSIONS AND PERSPECTIVES

The CB chemoreceptors contribute importantly to maintaining homeostasis during physiological and pathological stress. This is facilitated by chemoreflex-induced modulation of ventilation and autonomic outflow, specifically sympathetic activation to peripheral blood vessels and vagal activation to the heart. These effects manifest as hyperventilation, peripheral vasoconstriction and blood pressure increases, and bradycardia. CB dysfunction has been noted in an increasingly broad range of disease conditions. Given the physiological effects of chemoreflex activation, these diseases include hypertension, heart failure, sleep apnea, and bradyarrhythmias. Interestingly, a role for the chemoreflex has also emerged in the pathophysiology of metabolic dysfunction.

All the above conditions are increasing in prevalence worldwide. There is hence great interest in chemoreflex modulation to mitigate these diseases, by ablation, resection, and pharmacological or bioelectronic interventions. These opportunities speak to the importance of comprehensive understanding of transmitters, channels, and their interactions in mediating CB sensing and responses.

Although there is a long history of studies of carotid body resection, most of the latter conceptual initiatives have only recently entered the arena of experimental therapeutics. Although preliminary findings are promising, there is some indication that effects may not necessarily be sustained over the long term. Adverse effects of these interventions are also potentially significant, and much work remains to be done in terms of establishing risk-benefit ratios before carotid body chemoreceptor therapeutics are ready to enter the mainstream of therapy of ventilatory, cardiovascular, and metabolic disorders.

GRANTS

R.I. is supported by Puente 010-2020 PUC and by Fondecyt 1211443 grants; J.A. was supported by FONDECYT 1130177 grant; M.W.C. was supported by NIH Grant P01 HL-14388; and V.K.S. is supported by NIH Grants HL-65176 and HL-134885.

DISCLOSURES

V.K.S. has served as a consultant for Respicardia, Roche, and Bayer. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

R.I., J.A., M.W.C., and V.K.S. prepared figures; R.I., J.A., M.W.C., and V.K.S. drafted manuscript; R.I., J.A., M.W.C., and V.K.S. edited and revised manuscript; R.I., J.A., M.W.C., and V.K.S. approved final version of manuscript.