Chemoreceptor mechanisms regulating CO₂-induced arousal from sleep

Abstract

Arousal from sleep in response to CO₂ is a life-preserving reflex that enhances ventilatory drive and facilitates behavioural adaptations to restore eupnoeic breathing. Recurrent activation of the CO₂-arousal reflex is associated with sleep disruption in obstructive sleep apnoea. In this review we examine the role of chemoreceptors in the carotid bodies, the retrotrapezoid nucleus and serotonergic neurons in the dorsal raphe in the CO₂-arousal reflex. We also provide an overview of the supra-medullary structures that mediate CO₂-induced arousal. We propose a framework for the CO₂-arousal reflex in which the activity of the chemoreceptors converges in the parabrachial nucleus to trigger cortical arousal.

Graphical Abstract

Elevated arterial P_CO2 during sleep activates the carotid bodies and neurons in the retrotrapezoid nucleus in the brainstem that drive breathing to increase alveolar ventilation. If arterial P_CO2 continues to increase, the carotid bodies and retrotrapezoid nucleus stimulate ascending neural pathways that produce arousal from sleep.

Overview of CO₂-induced arousal

Arousal from sleep in response to CO₂ is a life-preserving reflex response to hypoventilation and asphyxia involving cortical arousal, identified by electroencephalogram (EEG) desynchronization, and a stereotyped activation of the respiratory and autonomic nervous system. This response enhances ventilatory drive and facilitates behavioural adaptations to restore eupnoeic breathing when respiratory homeostasis is compromised (Phillipson & Sullivan, 1978; Caples et al. 2005; Dempsey et al. 2010). Recurrent activation of the CO₂-arousal reflex causes sleep disruption in patients with obstructive sleep apnoea (OSA), which is associated with daytime sleepiness, cognitive impairment and metabolic syndrome (Dempsey et al. 2010; Javaheri & Javaheri, 2020). Importantly, the CO₂-arousal reflex is considered to be a modifiable physiological trait causing OSA, and sedatives that suppress EEG arousal but preserve ventilatory function have been tested for efficacy in OSA patients (Younes, 2008; Eckert & Younes, 2014). This review is focused on the mechanisms mediating CO₂-induced EEG desynchronization from/during non-rapid eye movement (NREM) sleep (abbreviated here as ‘the CO₂-arousal reflex’).

In theory, the CO₂-arousal reflex requires the activation of specialized cells, defined as chemoreceptors, that detect changes in pH, CO₂, and/or HCO₃⁻ and transduce this into a cellular response (Feldman et al. 2003; Putnam et al. 2004). Respiratory chemoreceptors contribute to the stimulation of breathing elicited by hypercapnia or metabolic acidosis (Guyenet et al. 2010). By analogy, a CO₂ chemoreceptor for arousal contributes to the neocortical or EEG activation in response to hypercapnia or metabolic acidosis. This review will examine role of chemoreceptors in the carotid bodies (CBs), retrotrapezoid nucleus (RTN) and dorsal raphe (DR) in the CO₂-arousal reflex, and discuss the integrative function of the parabrachial nucleus (PB) in this reflex.

The carotid bodies and the CO₂-arousal reflex

The CBs are polymodal chemoreceptors located in the carotid sinus that encode various blood-borne stimuli (for reviews see Zera et al. 2019; Ortega-Saenz & Lopez-Barneo, 2020). Evidence accumulated over the past 50 years demonstrates that the resting activity of the CBs contribute to blood gas homeostasis while breathing room air (Prabhakar & Semenza, 2015; Ortega-Saenz & Lopez-Barneo, 2020). The best characterized function of the CBs is the regulation of ventilation in relation to $P_{\text{a}}{}_{{\text{O}}_2}$ (Heeringa et al. 1979; Forster & Smith, 2010; Mouradian et al. 2012). Additionally, the CBs also regulate ventilation in relation to arterial P_CO2 and pH (Heeringa et al. 1979; Smith et al. 2015) and there is a synergistic interaction between the effect of CO₂ and O₂ on the CBs (Kumar & Prabhakar, 2012). However, there is a clear redundancy in the chemoreceptors responsible for respiratory homeostasis in rats because hypoventilation after CB ablation is relatively modest and temporary in this species (Mouradian et al. 2012; Souza et al. 2019).

Electrical stimulation of the carotid sinus nerve in anaesthetized cats evokes a defense response (reviewed in Marshall, 1994), which suggests the CBs command neural pathways that promote cortical arousal. Notably, CB ablation blunts hypoxia-induced arousal from sleep in dogs (Bowes et al. 1981), lambs (Fewell et al. 1989) and rats (Souza et al. 2019). This effect is consistent with an important role for the CBs in the physiological response to hypoxia. On the other hand, the contribution of the CBs to the CO₂-arousal reflex was untested until recently. In our study, we demonstrated that CB ablation in rats impairs the CO₂-arousal reflex (Fig. 1A) (Souza et al. 2019). The deficit in the CO₂-arousal reflex after CB ablation persists for at least 3 weeks, whereas the hypercapnia ventilatory reflex (HCVR) is unaffected by CB ablation. The contribution of the CBs to the CO₂-arousal reflex may therefore be somewhat independent of their influence on breathing. Our study suggests that the deficit in the CO₂-arousal reflex following CB ablation is modest; however, our results are likely to underestimate the contribution of the CBs to a mixed hypoxia-hypercapnia stimulus, as in sleep apnoea.

Figure 1. RTN neurons contribute to the CO₂-arousal reflex: evidence from gain- and loss-of-function experiments.

A, RTN and CB ablation blunt the CO₂-arousal reflex. Sleep survival curves (mean ± SEM) for rats exposed to CO₂ in a plethysmography chamber are shown. Three groups of rats (sham operated n = 10, RTN ablation n = 9; CB ablation n = 6) were subjected to air-to-15% CO₂ or air-to-air gas changes. $F_{\text{I,}}{}_{{\text{CO}}_2}$ during CO₂ challenges is shown at the bottom of panel A. Ablation of either RTN or CBs reduces the CO₂ arousal reflex but the loss of the RTN has a much more pronounced effect. Data from Souza et al. (2019). Significant differences were determined by a Bonferroni test following a two-way repeated measures ANOVA. *P < 0.05, **P < 0.01 for CB ablation vs. sham. *P < 0.05, **P < 0.01***P < 0.001 for RTN ablation vs. sham. Data from Souza et al. 2019. B, optogenetic RTN stimulation causes arousal from sleep. Sleep survival curves (mean ± SEM) during selective stimulation of RTN (n = 8) with increasing frequencies of stimulation are shown. Arousal requires stimulation frequencies of greater than 6–9 Hz, a discharge rate matched by RTN neurons when recording in anaesthetized rats with an end tidal CO₂ of 7% (Guyenet et al. 2005). Data from Souza et al. 2020. Significant differences were determined by a Bonferroni test following a two-way repeated measures ANOVA. *P < 0.05 RTN stim vs. no stim.

The retrotrapezoid nucleus and the CO₂-arousal reflex

Here we consider retrotrapezoid nucleus (RTN) neurons as a population of perifacial CO₂-activated neurons that rely on the transcription factors PHOX2B and ATOH1 during embryogenesis (Goridis et al. 2010; Ruffault et al. 2015; Guyenet et al. 2019). RTN neurons are glutamatergic (Stornetta et al. 2006; Holloway et al. 2015; Ruffault et al. 2015), hence excitatory, but they also express several neuropeptides (galanin, neuromedin B and pituitary adenylate cyclase-activating polypeptide) whose contribution to synaptic transmission is only partially understood (Li et al. 2016; Shi et al. 2017, Shi et al. 2020). RTN neurons retain their pH sensitivity after acute isolation, indicating that they possess a degree of intrinsic sensitivity to pH (Wang et al. 2013b). This property is currently attributed to the expression of two pH sensitive membrane proteins, GPR4 and TASK2 (Wang et al. 2013a; Kumar et al. 2015; Guyenet et al. 2019). However, the pH response of RTN neurons in vivo is markedly greater than in vitro (Mulkey et al. 2004), suggesting that the pH sensitivity of these neurons may also rely on additional mechanisms that include paracrine effects of CO₂ on RTN neurons mediated by local blood vessels and astrocytes, and synaptic inputs from CO₂/pH-activated neurons located elsewhere in the brain (Gourine et al. 2010; Guyenet et al. 2019; Wu et al. 2019; Cleary et al. 2020). The downstream connectivity of the RTN has not been elucidated at the cellular level, but these neurons innervate brain areas linked with the brainstem respiratory network including the entire ventral respiratory column, Kölliker-Fuse and lateral PB, and nucleus of the solitary tract (Abbott et al. 2011; Bochorishvili et al. 2012; Souza et al. 2020). Selective stimulation of RTN neurons produces a pattern of breathing indistinguishable from the HCVR, including active expiration (Abbott et al. 2011; Souza et al. 2020). Conversely, selective inhibition of RTN neurons produces hypopnea/apnoea during quiet breathing (Marina et al. 2010; Basting et al. 2015; Burke et al. 2015a) and RTN ablation reduces the HCVR by at least 70% (Souza et al. 2018). Collectively, the evidence indicates that the central respiratory chemoreflex requires the activation of RTN neurons but mechanisms responsible for the activation of RTN neurons by CO₂ are not fully understood.

An impairment in the development of RTN neurons probably contributes to the respiratory phenotype of congenital central hypoventilation syndrome (CCHS), a condition caused most often by polyalanine expansion mutations of Phox2b (Amiel et al. 2003; Dubreuil et al. 2008; Weese-Mayer et al. 2017). Individuals with CCHS have a very weak HCVR, exhibit varying degrees of hypopnea and apnoea during sleep (Matera et al. 2004; Weese-Mayer et al. 2010), and deficits in the CO₂-arousal reflex and air hunger (Shea et al. 1993). Studies in mice suggest that the absence or hypoplasia of the RTN at birth largely accounts for the loss of the HCVR in CCHS (Dubreuil et al. 2008; Ramanantsoa et al. 2011). However, the CO₂-arousal reflex of mutant mice designed to model CCHS has not been tested. In order to test the contribution of the RTN to the CO₂-arousal reflex in adult rats we ablated the RTN using local injections of the toxin substance P-saporin; these lesions blunted the CO₂-arousal reflex by ~40% after 40 s of CO₂ exposure (Souza et al. 2019) without altering hypoxia-induced arousal or basal sleep-wake patterns. Thus, the deficits in the arousal reflex are specific to CO₂ and not related to changes sleep-wake behaviour or generalized arousal. Based on these observations RTN activation contributes to CO₂-induced arousal. Consistent with this interpretation, selective RTN stimulation promotes arousal from NREM sleep (Burke et al. 2015b; Souza et al. 2020). In Souza et al. (2020), we also demonstrate that arousal from NREM sleep occurs in response to RTN stimulation at frequencies close to the in vivo firing rate of RTN neurons in vagally intact ventilated anaesthetized rats with an end tidal CO₂ of 7%, which is roughly the midpoint of the dynamic range of RTN neurons in those conditions (Guyenet et al. 2005). Notably, the stimulation frequency required to produce sleep-wake transitions was considerably above that required to increase ventilation, suggesting that the activity of the RTN at discharge frequencies associated with mild hypercapnia stimulates breathing without causing arousal from sleep (Souza et al. 2020). The effects of RTN activity on the spectral composition of EEG during NREM sleep were not examined. Hence, the available evidence suggests that RTN neurons may be a target for therapies aimed at stabilizing alveolar ventilation during sleep without deleterious effects on sleep quality.

Interactions between CBs and RTN in the CO₂-arousal reflex

CB activation potentiates the gain of the central ventilatory response to CO₂ in intact mammals to varying degrees (Forster & Smith, 2010; Smith et al. 2010; Dempsey & Smith, 2019). In dogs, for example, selective hyperoxygenation of vascularly isolated CBs markedly blunts the HCVR (Blain et al. 2010). A neural framework for the interaction between the CBs and central chemoreceptors has been proposed (Smith et al. 2010) and may involve interactions between neurons in the nucleus of the solitary tract (NTS) and RTN (Takakura et al. 2006). The activity of the CBs may modulate the CO₂-arousal reflex. Consistent with this idea, mild hypoxia in humans (SaO₂ 87%) modestly facilitates the CO₂-arousal reflex (Gothe et al. 1986). However, administering low oxygen therapy during sleep $S_{\text{a}}{}_{{\text{O}}_2}$ in OSA patients impairs arousal from apnoea (Edwards et al. 2014), consistent with the well-known depressant effect of hypoxia on sensory processing (e.g. Eckert et al. 2005), and the opposite of what is expected if hypoxia and hypercapnia exerted additive effects on arousal. In mice, the probability of arousal from sleep in response to mixed hypoxia-hypercapnia (10% CO₂ + 10% O₂) is not different from the hypercapnia alone (10% CO₂ + 21% O₂) (Kaur et al. 2013), suggesting a limited interaction, though it is possible 10% CO₂ alone is a saturating stimulus (i.e. produces the highest probability of arousal possible). Hence, the existence of a facilitatory interaction between the effects of hypoxia and hypercapnia on EEG arousal is plausible but presently lacks convincing evidence.

In our studies, RTN or CB ablation alone attenuated the CO₂-arousal reflex, but neither prevented it. This suggests that the CO₂-arousal reflex relies, at least in part, on independent actions of CO₂ at the RTN and the CBs. Optogenetic stimulation of RTN produced arousal less reliably than CO₂, which could be interpreted as evidence that CO₂-arousal is mediated by the activation of chemoreceptors besides the RTN, for example, the carotid bodies or serotonergic neurons. However, based on histological counts, we stimulated only 30% of the bilateral population of RTN neurons in the rat and it remains possible that activating a greater proportion of RTN neurons, as would be the case during CO₂ exposure, would promote EEG arousal more reliably. In any case, our results support the notion of an interaction between RTN and CB activity in the CO₂-arousal reflex, though its nature (additive or hyperadditive) is currently undefined.

Role of respiratory stimulation in the CO₂-arousal reflex

Respiratory chemoreceptors like the CBs and RTN neurons are essential for the neural drive to breathe and therefore may promote arousal through activation of elements in the central respiratory network in the brainstem that interact with ascending arousal systems (e.g. Yackle et al. 2017), or via sensory feedback related to breathing itself (Berry & Gleeson, 1997; Guyenet & Abbott, 2013). Interestingly, the aversive sensation of elevated $P_{{\text{CO}}_2}$ in humans is counteracted by feedback inhibition from stretch receptors in the lungs (Manning et al. 1992; Banzett et al. 1996; O’Donnell et al. 2007). This suggests that the degree of matching between neural ventilatory drive (i.e. chemoreceptor activity) and lung ventilation (i.e. sensory feedback) determines whether a given level of CO₂ is perceived. The integration of chemoreceptor activity with respiratory-related sensory feedback is likely to involve the NTS (Fig. 2), for example, GABAergic NTS neurons activated by slowly adapting pulmonary stretch receptors innervate and inhibit RTN neurons during lung inflation (Guyenet et al. 2005; Moreira et al. 2007).

Figure 2. Central organization of CO₂-arousal reflex: a hypothesis.

The initial trigger to CO₂-induced arousal may be the sudden acidification of the CBs followed by that of a small number of lower brainstem structures, especially the RTN. This acidification increases the excitatory drive from the RTN and CBs to respiratory centres in the ventral respiratory column (VRC) and nucleus of the solitary tract (NTS) and pontine respiratory group (not shown). The integrated output of the respiratory centres is communicated to the peripheral respiratory system through premotor and motor neurons located in the medulla and spinal cord. Increased ventilatory motor output also activates sensory afferents from the lungs, chest and airways that provide feedback via cranial and spinal nerves. Serotoninergic neurons in the medullary raphe regulate the activity of the entire central respiratory network, including the RTN. Arousal probably results from an abrupt rise in the activity of the external lateral parabrachial nucleus (PBel), a structure that probably integrates direct or polysynaptic inputs from a variety of sources related to breathing such as the carotid bodies, RTN, VRC, NTS and sensory feedback for the lungs chest and airways. The PBel promotes EEG arousal through a diffuse forebrain network that includes the basal forebrain (BF), lateral hypothalamic area (LHA), and central nucleus of the amygdala (CeA). PBel neurons are regulated by serotonergic inputs from the dorsal raphe. The neurotransmitter phenotype of each neural compartment is indicated; Glu, glutamate; GABA, gamma-aminobutyric acid; Gly, glycine; ACh, acetylcholine; 5HT, serotonin.

Dorsal raphe serotonin neurons and the CO₂ arousal reflex

Central serotonin (5HT) neurons have long been implicated in CO₂ homeostasis (Nattie, 2011; Teran et al. 2014) and there is substantial evidence that firmly establishes these neurons, and specifically the dorsal raphe (DR) 5HT neurons, in the CO₂-arousal reflex (Buchanan & Richerson, 2010; Smith et al. 2018; Buchanan, 2019; Kaur et al. 2020). DR 5HT neurons heavily innervate parts of the forebrain that regulate arousal including the cortex, thalamus, hypothalamus and midbrain (Donovan et al. 2019). DR 5HT innervation of the brainstem is sparse by comparison, but does include the rostral ventrolateral medulla (Vertes & Kocsis, 1994). Genetically targeted acute inhibition of DR 5HT neurons in adult mice, but not DR 5HT neurons in the medullary raphe, attenuates the CO₂-arousal reflex without affecting the HCVR (Smith et al. 2018; Kaur et al. 2020). Hypoxia-induced arousal is also intact in mice lacking 5HT neurons from birth (Lmx1b^f/f/p mice) (Buchanan & Richerson, 2010). Collectively, this indicates DR 5HT neurons have a somewhat specialized role in mediating the arousal-promoting effects of CO₂.

DR 5HT neurons regulate the CO₂-arousal reflex primarily through 5HT_2a receptors (Buchanan et al. 2015; Kaur et al. 2020), which are expressed throughout the CNS. The CO₂-arousal reflex is inhibited by 5HT_2a antagonists (Buchanan et al. 2015). Activating 5HT_2a receptors with a systemic agonist restores CO₂-induced arousal in Lmx1b^f/f/p mice (Buchanan et al. 2015), and prevents the effects of optogenetic inhibition of DR inputs to the PB (Kaur et al. 2020). These observations demonstrate an important role for DR 5HT neurons and 5HT signalling in the CO₂-arousal reflex, but also that the CO₂-arousal reflex can be triggered by other chemoreceptors.

A subset of DR 5HT neurons are pH sensitive owing in part to a pH-sensitive TWIK-related acid sensitive K⁺ channel (Richerson, 2004; Mulkey et al. 2007). In awake freely behaving cats, Veasey et al. (1997) reported that 8 of 36 dorsal raphe neurons, characterized as serotonergic by electrophysiological properties, increased their activity during CO₂ exposure. This remains the only study to directly assess the activity of DR 5HT neurons during CO₂ exposure in intact conditions. Still, the results are consistent with the idea that a subset of DR 5HT neurons is activated during CO₂ in vivo through intrinsic pH sensing. Further supporting this idea, acidification of the DR by microdialysis in mice markedly reduces the latency to arousal during NREM sleep (Smith et al. 2018). Notably, acidification of the DR does not cause arousal in Lmx1b^f/f/p mice following treatment with a 5HT_2a receptor agonist (Smith et al. 2018), which indicates that arousal from sleep during DR acidification is due to activation of 5HT neurons. This stands in contrast to the observation that systemic 5HT_2a agonists restore the CO₂-arousal reflex in Lmx1b^f/f/p mice. Again, this supports the view that the CO₂-arousal reflex involves the transduction of pH/CO₂ by chemoreceptors other than 5HT neurons.

Whether stimulating DR 5HT neurons is a stimulus for arousal/wakefulness per se has been addressed using optogenetic and chemogenetic stimulation of DR 5HT neurons. Tonic optogenetic stimulation of DR 5HT neurons in mice at 3 Hz promotes sleep (Oikonomou et al. 2019), and chemogenetic stimulation promotes sleep (Venner et al. 2020) or has no effect on sleep-wake patterns (Kaur et al. 2020). In contrast, optogenetic stimulation at high frequency (>20 Hz) promotes wakefulness (Ito et al. 2013; Oikonomou et al. 2019). Oikonomou and colleagues propose that these results reflect two modes of DR 5HT function, with tonic low frequency activity promoting quiescence and/or sleep and high tonic activity or burst firing facilitating arousal and wakefulness. While caution is warranted in predicting the role of 5HT DR neurons in the CO₂-arousal reflex from the effects of opto- and chemogenetic stimulation, this evidence suggests that DR 5HT neurons need to be strongly activated to trigger arousal from sleep, at least in isolation of other drives. In Veasey et al. (1997), the activity of CO₂-activated DR neurons increased from on average 3 Hz to 5 Hz during exposure to 8–9% CO₂; however, it is not possible to draw a strong conclusion from these data regarding the effect of CO₂ exposure on DR 5HT neurons given the limited sample. To resolve this issue, more comprehensive recordings of the activity of DR 5HT neurons during CO₂ exposure in unanaesthetized mammals are required. These recordings would ideally take into account the diversity of DR 5HT neurons identified by single-cell transcriptional profiling of these neurons (Ren et al. 2018; Okaty et al. 2020).

In summary, DR 5HT neurons are important for the CO₂-arousal reflex though recent evidence (Kaur et al. 2020) has raised questions as to whether CO₂/pH-dependent changes in the activity of DR 5HT neurons contributes to the CO₂-arousal reflex.

Parabrachial nucleus – a hub for viscerosensory signalling that promotes arousal

The PB is a convergence point for viscerosensory signalling from the brainstem and spinal cord and serves an integrative function in autonomic and respiratory control (Craig, 2002; Saper, 2002). The PB is composed of subnuclei defined by genetics, cytoarchitecture and connectivity (Fulwiler & Saper, 1984; Huang et al. 2020), with the external lateral subnucleus (PBel) playing a key function in the CO₂-arousal reflex (Kaur et al. 2013, 2017; Kaur & Saper, 2019). The PBel is composed of excitatory glutamatergic neurons that project to forebrain structures involved in arousal (Kaur & Saper, 2019). In rodents, neurons in the PBel robustly express cFOS, a marker of neuronal activation, following CO₂ exposure (Teppema et al. 1997; Berquin et al. 2000; Yokota et al. 2015), though PBel neurons are not known to be intrinsically sensitive to CO₂ or pH. This suggests that activation of the PBel during CO₂ is synaptically mediated. CO₂-activated inputs to the PBel originate from the lower brainstem respiratory network (Kaur & Saper, 2019), including possible monosynaptic inputs from the RTN (Souza et al. 2020). The PBel is also regulated by DR 5HT neurons, which has been shown to be important in the CO₂-arousal reflex (Kaur et al. 2020). It also receives inputs from the dorsal horn of the spinal cord and subnuclei of the NTS that process sensory feedback from the airways and CB activity (Kaur & Saper, 2019). Thus, the connectivity of the PBel indicates that these neurons receive respiratory-related feedback and convey this information to forebrain regions that mediate the CO₂-arousal reflex (specifically EEG arousal).

PBel neurons express calcitonin gene related peptide (CGRP), which has been used to conduct genetically targeted manipulations of this nucleus (Palmiter, 2018). Acute optogenetic inhibition of PBel CGRP neurons or genetically eliminating glutamate release from PBel neurons attenuates the CO₂-arousal reflex in mice (Kaur et al. 2013, 2017). PBel CGRP neurons broadly collateralize in subcortical forebrain regions (Bowen et al. 2020) and deficits in the CO₂-arousal reflex were observed when the projections of PBel neurons in the basal forebrain (BF), lateral hypothalamic area (LHA) and central amygdala (CeA) were inhibited (Kaur et al. 2017). This suggests that a diffuse projection from the PBel to the forebrain mediates the CO₂-arousal reflex. Chemogenetic and optogenetic stimulation of PBel CGRP neurons promotes wakefulness, but disrupting PBel CGRP neurons does not affect basal sleep wake behaviour, the HCVR, or the arousal response to audio stimuli or gentle physical shaking (Kaur et al. 2017; Kaur & Saper, 2019), suggesting a relatively restricted role in the arousal responses to CO₂. However, PBel CGRP neurons contribute to innate behavioural and physiological responses to a range of visceral sensory modalities (Palmiter, 2018), and it is unclear if and how CO₂ signalling is integrated by the PBel. In the context of CO₂ arousal, it will be important to determine if and how neurons in the PBel encode and integrate stimuli that are related, for example, to CO₂ and respiratory sensory feedback.

Role of subcortical arousal-promoting structures in the CO₂ arousal reflex

Sleep-wake patterns are governed by an interconnected neural network organized across layers in the CNS (Lee & Dan, 2012; Saper & Fuller, 2017; Chen et al. 2020). Our proposal is that arousal from sleep in response to CO₂, as experienced in sleep apnoea, relies on the activation of a select few chemoreceptors in the brainstem and periphery, principally the CBs and RTN. The aggregate neural activity stimulated by these chemoreceptors, including central respiratory activity and feedback from the periphery, activates PBel neurons. PBel neurons subsequently stimulate subcortical structures including, but not limited to, the BF, LHA, and CeA of the amygdala to produce EEG arousal.

The BF is a key nodal point connecting the ascending arousal systems to the cortex (Jones, 2017). This brain region receives a significant input from the PBel and accordingly mediates the CO₂-arousal reflex (Kaur et al. 2017; McKenna et al. 2020). BF inhibition impairs arousal not only to CO₂, but also to an auditory stimulus, which is consistent with the regulation of cortical excitability in general (McKenna et al. 2020). The LHA is a heterogenous region containing intermingled neurons, some of which have a wake promoting or sleep promoting function. If it is assumed that PBel neurons generate the CO₂-arousal reflex by stimulating arousal-promoting neurons, then there are at least two populations of neurons in the LHA that may mediated the CO₂-arousal reflex. The best characterized of the two are hypocretin/orexin neurons (Bonnavion et al. 2016), which mediate arousal-related modifications in autonomic and respiratory function (Nattie & Li 2012). Thus, hypocretin neurons may be an effector through which CO₂-exposure produces EEG arousal and the cardiorespiratory stimulation associated with sleep-wake transitions. However, the CO₂-arousal reflex in mice is unaffected by systemic administration of the hypocretin antagonist TCS 1102 (Buchanan et al. 2015). Another possible target for PBel neurons in the LHA is inhibitory neurons, which promote arousal through inhibition of sleep-promoting neurons in the preoptic area (Venner et al. 2019) and GABAergic neurons in the thalamoreticular nucleus (Herrera et al. 2016). However, the role of inhibitory neurons in LHA in the CO₂-arousal reflex has not been tested. The CeA is a hub for emotional control (Phelps & LeDoux, 2005) that is required for the behavioural responses of mice to CO₂ (Taugher et al. 2020). These behaviours necessarily require forebrain arousal, which may be driven in part by central amygdala projections to the basal forebrain. And finally, the CO₂-arousal reflex is also likely to involve ascending projections from the brainstem and pons that promote forebrain arousal in parallel to the PBel circuit. The existence of these pathways is supported by the presence of the CO₂-arousal reflex following loss-of-function manipulations of the PBel (Kaur et al. 2017). An example of such a pathway is the noradrenergic neurons of the locus coeruleus, which have widespread projections throughout the CNS and promote arousal (Guyenet & Abbott, 2013; Poe et al. 2020).

Hypocretin/orexin neurons in the LHA (Burdakov et al. 2013), neurons in the amygdala (Ziemann et al. 2009) and noradrenergic neurons in the LC (Putnam et al. 2004) are also reported to be intrinsically sensitive to changes in CO₂ and/or pH. In the case of amygdala neurons (Ziemann et al. 2009) and possibly orexin neurons (Song et al. 2012), pH sensing is thought to be related to the activity of the ASIC1a channel, an amiloride-sensitive sodium channel expressed throughout the CNS. ASIC1a has a pH_0.5of 6.5 (Chen et al. 1998) so is unlikely to contribute to the homeostatic regulation of ventilation, a view supported by the observation that ASIC KO mice have a normal HCVR (Ziemann et al. 2009; Detweiler et al. 2018). On the other hand, global and neuron-selective KO of ASIC1a in mice reduces the behavioural response to 10% CO₂, but not 5%. This has been attributed to ASIC1a channels expressed in the amygdala and bed nucleus of the striatum (Ziemann et al. 2009; Taugher et al. 2014, 2017), but may involve other targets (Vollmer et al. 2016). In summary, pH sensing in the forebrain modifies behaviour and therefore arousal, but there is presently no evidence to confirm or refute the role of chemoreceptors in the forebrain in the CO₂-arousal reflex. One possibility is that pH sensing in the forebrain regions is important at CO₂ levels that are higher than necessary to elicit CO₂-arousal reflex such as during prolonged apnoea, asphyxia, or ischaemia.

Model for CO₂-arousal reflex

The following model of the CO₂-arousal reflex (Fig. 3) is predicated on the PBel functioning as a gatekeeper for CO₂-related signalling to the forebrain. This organization is based on evidence the that silencing the PBel reduces the incidence of arousal during a 30 s period of 10% CO₂ by 50% (Kaur et al. 2017). During periods of eucapnia during NREM sleep, the activity of the PBel is low. Under these conditions, the PBel receives a tonic excitatory input from DR 5HT neurons mediated by 5HT_2a receptors and negligible inputs from the RTN and CBs pathways. The CBs and RTN pathways represent the aggregate activity of brainstem inputs to the PBel that are activated by those chemoreceptors (Fig. 2). During acute CO₂ exposure, PBel activity increases due to the activation of the CB and RTN pathways. The response of the PBel to CO₂ during NREM sleep also requires input from DR 5HT neurons, but this input is considered tonic (i.e. invariable with increasing CO₂) up to the point that arousal occurs. Increases in PBel activity cause widespread activation of the forebrain, and specifically the BF, which disinhibits the cortex. The medullary raphe may also contribute to the intensity of the CO₂ arousal reflex through actions on the RTN and respiratory network. The model includes a variable threshold for arousal (yellow range in Fig. 3) at which PBel activity triggers arousal from sleep, which is considered an all-or-none event. This threshold is set by factors that modulate the excitability of the forebrain network underlying EEG arousal. Whether this same circuitry is utilized for CO₂-responses beyond EEG arousal from sleep, such as fear responses, remains to be determined. However high CO₂ levels could recruit chemoreceptors not included in the model, and so may involve mechanisms not presented here. This model bears similarity to those describing the central processing of the sensations of breathing (O’Donnell et al. 2007) with added anatomical and genetic information.

Figure 3. CO₂-arousal reflex: schematic representation of the relative contribution the carotid bodies, central chemoreceptors, serotonergic system and basal forebrain.

In the intact brain, CO₂ exposure, the CBs and RTN directly or indirectly increase the activity of PBel. PBel activates the basal forebrain which causes EEG desynchronization and arousal when a particular threshold is reached (dashed line). Removal of either the CBs or RTN reduces CO₂-related input to the PBel thus delaying the activation of the basal forebrain during CO₂ exposure. Based on the findings in Souza et al. (2019), CB ablation causes a rightward shift in the PBel response to CO₂ exposure, indicating that the early detection of CO₂ is impaired. However, arousal invariably occurs because RTN inputs are intact. RTN ablation, on the other hand, produces a major deficit in the PBel response to CO₂ exposure, indicating that the RTN contributes to PBel activity over a broader range of CO₂ than the CBs. Hypothetically, removal of both the CBs and RTN may well abolish the CO₂ arousal reflex but the extreme hypoventilation resulting from by such a procedure may preclude such an experiment. Removal of the 5HT input to PBel causes a severe deficit in CO₂-induced arousal by reducing the excitability of the PBel, and its responsiveness to excitatory inputs from the RTN and CB. Inhibition of the basal forebrain, and other efferent targets of the PBel, raises arousal threshold so that only a very high PBel activity (i.e. a high CO₂ stimulus) is sufficient for arousal from sleep.

Conclusion

Understanding the physiological basis for CO₂-induced arousal function is an important challenge to address in order to understand the pathophysiology of human respiratory disease and disorder. Over the last decade, there has been meaningful progress in describing the neural processes that underpin the CO₂ arousal reflex owing in part to developments in the methods used to study these systems in rodents. There is still a great deal to discover. More work is needed to determine and differentiate, with certainty, the chemoreceptors responsible for the CO₂-induced arousal and those systems that modulate the sensitivity of this reflex. Moreover, it is important to untangle the relationship between respiratory CO₂ homeostasis and the CO₂-arousal reflex in order to target these mechanisms with approaches that are tractable in humans. Finally, it is of interest to determine the extent of overlap between the neural pathways mediating the CO₂ arousal reflex and other aversive signalling for the respiratory system, such as dyspnoea. If this is accomplished, it may be possible to develop new broadly applicable preventative treatments for sleep disordered breathing.

Biographies

Stephen B. G. Abbott is an Assistant Professor in the Department of Pharmacology and Neuroscience at the University of Virginia, Charlottesville, USA. Research in Dr. Abbott’s lab is focused on the neural control of breathing and blood pressure, and the mechanisms linking cardiorespiratory function with sleep-wake behaviour and forebrain arousal.

George M. P. R. Souza has a Ph.D. in physiology from the University of São Paulo, Brazil. He first came to University of Virginia as a postdoctoral fellow in the field of autonomic neuroscience with Professor Emeritus Patrice Guyenet, Ph.D. Currently, Dr. Souza is focused on investigating neural mechanisms underlying CO₂-induced arousal from sleep in Dr. Abbott’s lab.