Poor patient adherence to CPAP for the treatment of OSA underscores the need to develop better treatment options [4]. Combined with increasing prevalence of OSA and its associated comorbidities [1]–[3], tolerable and easily applicable treatments for OSA, like pharmacotherapies, must be developed. However, decades of unsuccessful efforts to identify effective drug treatments for OSA [5] highlight the need for novel and innovative approaches.

Dronabinol, a non-selective CB agonist, has been shown to decrease sleep apnea in conscious rats [13] and in humans with OSA [14]. In an acute rat model of reflex apnea, 5-HT-induced apnea was attenuated or blocked and phasic EMG increased in rats receiving nodose ganglia injections of dronabinol [16], suggesting that modulating vagal afferent activity has an important role in respiratory control [6]. Here, we show that dronabinol modulates vagal afferents through both CB₁ and CB₂ receptors. By blocking these receptors through IP pre-treatment with CB antagonists, we reversed the attenuation of reflex apnea produced by nodose ganglia injections of dronabinol.

The concept of pharmacologically modulating vagal afferent activity to treat sleep apnea has been employed previously. Yoshioka et al. first showed that modulation of vagal afferents by antagonism of 5-HT₃ receptors in acute experiments in rats attenuated reflex apnea [15]. This observation was then extended to spontaneous sleep-related apnea by Radulovacki et al. who showed in a chronically-instrumented conscious rats that 5-HT₃ antagonism decreased sleep apnea frequency [9]. Fenik and colleagues suggested that peripheral antagonism of 5-HT₃ receptors on vagus nerves augments inspiratory drive to hypoglossal nerves in anesthetized, paralyzed, and mechanically ventilated rats [8]. These three studies together provided a rationale for testing 5-HT antagonists in a model of obstructive sleep apnea. Indeed, Veasey et al. demonstrated that 5-HT₃ antagonism reduced the respiratory disturbance index during sleep in the English bulldog [10]. In humans, a serotonin antagonist combined with a selective serotonin reuptake inhibitor decreased apnea-hypopnea index during sleep, but the therapeutic window for this effect appears to be narrow [5], [31]. Thus, modulating vagal afferent activity to treat sleep apnea has potential if the correct pharmacotherapy can be applied.

CBs have been used to modulate vagal afferent activity via the inhibitory G protein-coupled (G_i/o) CB₁ and CB₂ receptors located on the nodose ganglia [22]. Vagal afferent modulation via CB receptors has been suggested as a treatment not just for sleep apnea [13], [14], but also for emesis [17]–[19] and for chronic cough [20], [21]. In the vagally-mediated behavior of emesis, CB₁ receptors are implicated. In contrast, in the vagally-mediated behavior of chronic cough, CB₂ receptors are involved. We hypothesized that dronabinol’s attenuation of reflex apnea occurred via CB₁ receptors, and not CB₂ receptors, because CB₁ receptors are more abundant in the nervous system than CB₂ receptors [23]. Moreover, CB₁ receptor knock-out mice display more sleep apneas than wild-type controls, suggesting CB tone is important in respiratory stability [24]. However, in our study, when either CB receptor subtype was antagonized, or when both were antagonized, by IP pre-treatment, reflex apnea attenuation by local injection of dronabinol was reversed. CB₁ and CB₂ receptors are positively coupled to A-type potassium channels [32], which are expressed on nodose ganglion cells and are known to decrease nodose ganglion/vagus nerve excitability [33]–[35]. It is possible that activating only one of the two CB receptor subtypes is insufficient to initiate a G_i/o cascade that overcomes 5-HT-induced excitability of the vagus nerve. In contrast, activation of either CB₁ or CB₂ receptor centrally had an antiemetic effect in brainstem-induced emesis [36]. The difference in effects of CB antagonism on vagal afferents seen in reflex apnea compared to vagally-mediated emesis [17]–[19] or chronic cough [20], [21] could be due to differential receptor expression on A- and/or C-fibers of the vagus nerves/nodose ganglia. In other words, there might be differences in CB receptor expression profiles between the vagally-mediated behaviors of apnea, emesis, and cough. Less than 60% of nodose ganglia cells exhibit CB₁ expression [37], and though CB₂ receptors are located on vagal afferents [22], no studies have been conducted to quantify to CB₂ receptors expression; however, there is differential expression of other receptor types on cell bodies in the nodose ganglia [7]. It is possible that the vagal afferents responsible for emesis and chronic cough express only one of the two subtypes of CB receptors, while vagal afferents responsible for sleep apnea express both CB receptor subtypes. Studies to look at CB receptor expression in the nodose ganglia would be of interest.

CBs have been reported to allosterically modulate many different types of ionotropic receptors [28], including 5-HT receptors on nodose ganglion cells. Anandamide, a cannabimimetic, attenuated 5-HT-induced currents in cultured nodose ganglion cells independent of G protein coupled signaling [25]. In cultured HEK 293 cells and Xenopus oocytes, cells that lack CB receptors, Δ9-THC and anandamide inhibited 5-HT-induced currents [26], [27]. Barann and colleagues also showed that 9-THC reduced 5-HT-induced current in dissociated rat nodose ganglion cells [26]. However, no CB receptor antagonists were used in these electrophysiological experiments, so determination of whether the current reduction was attributable to CB receptor dependent or independent mechanisms is unknown. However, in the present work, CB antagonism was sufficient to reverse dronabinol’s effect on reflex apnea, arguing against dronabinol acting via CB receptor-independent mechanisms. However, Yang and colleagues showed that allosteric inhibition of the 5-HT₃ receptor via cannabidiol, a cannabimimetic, is dependent on expression levels 5-HT₃ receptor [27]. It could be that the high expression levels of 5-HT₃ on nodose ganglion cells [7] and the dose of dronabinol used to inject into the nodose ganglia is insufficient to allosterically inhibit 5-HT₃ receptors, thus leaving the activation of CB receptors as the main inhibitory signal of vagal afferents. Patch clamping electrophysiology of nodose ganglion cells using CB receptor antagonisms need to be conducted to resolve this issue.

In contrast to CB antagonism reversing dronabinol’s attenuation of reflex apnea, CB antagonism did not reverse dronabinol-induced increases in phasic or tonic genioglossus EMG. Three explanations can account for this observation. The first possibility is that brainstem respiratory control of the hypoglossal motoneurons responsible for genioglossus activity is distinct from that of the phrenic motoneurons responsible for controlling the diaphragm [38]–[40]. Vagal afferents from pulmonary and upper airway tissues synapse in the nucleus of the solitary tract (NTS) [41], [42], which is part of the respiratory circuitry including the ponto-medullary pattern generator that projects to the phrenic and hypoglossal motor nuclei [43]. Different areas of the NTS stimulated with L-glutamate lead to different patterns of respiratory responses [42], and CB receptors are located presynaptically on first-order glutamatergic vagal afferent neurons and on second-order GABAergic neurons in NTS [44], [45]. It is feasible that systemic pre-treatment with CB receptor antagonists can increase genioglossus activity through modulation of first- and/or second-order NTS neurons that project to hypoglossal motoneurons, but does not affect phrenic motoneurons. A second possibility is that antagonism of CB receptors on hypoglossal motoneurons leads to postsynaptic potentiation [46]. It is possible that systemic pre-treatment with CB receptor antagonists can increase genioglossus activity by directly increasing hypoglossal nerve activity. It remains to be seen how central antagonism of CB receptors differentially modulates genioglossus and diaphragmatic activity.

A final possible explanation of the observation that CB antagonism did not reverse dronabinol-induced increases in EMGgg is that vagal afferent neurons that project to circuitry responsible for upper airway activity might not contain CB receptors [37], and therefore are not affected by CB receptor antagonism. These vagal afferents are then only inhibited or attenuated by allosteric modulation of the 5-HT₃ receptor [26], [27]. Fenik and colleagues have shown that antagonizing 5-HT₃ receptors peripherally with ondansetron increased inspiratory modulation of hypoglossal nerve activity [8]. This could explain the increases in phasic/tonic EMGgg seen in this study.

In summary, we show that CB₁ and CB₂ receptor antagonists reversed the suppression of 5-HT-induced apnea by locally-injected dronabinol into the nodose ganglia, but did not reverse increases phasic activation of the genioglossus. These data support the view that systemic dronabinol stabilizes respiratory pattern, in part, by acting via CB₁ and CB₂ receptors at the nodose ganglia. These findings underscore the therapeutic potential of dronabinol in the treatment of OSA and implicate participation of both cannabinoid receptors in dronabinol’s apnea suppression effect.