Central and peripheral factors contributing to obstructive sleep apneas

Apnea, the cessation of breathing, is a common physiological and pathophysiological phenomenon. Among the different forms of apnea, obstructive sleep apnea (OSA) is clinically the most prominent manifestation. OSA is characterized by repetitive airway occlusions that are typically associated with peripheral airway obstructions. However, it would be an oversimplification to conclude that OSA is caused by peripheral obstructions. OSA is the result of a dynamic interplay between chemo- and mechanosensory reflexes, neuromodulation, behavioral state and the differential activation of the central respiratory network and its motor outputs. This interplay has numerous neuronal and cardiovascular consequences that are initially adaptive but in the long-term become major contributors to morbidity and mortality. Not only OSA, but also central apneas (CA) have multiple, and partly overlapping mechanisms. In OSA and CA the underlying mechanisms are neither "exclusively peripheral" nor "exclusively central" in origin. This review discusses the complex interplay of peripheral and central nervous components that characterizes the cessation of breathing.


Introduction
Obstructive Sleep Apnea (OSA) is a major health issue worldwide affecting 3-7% of adult men and 2-5% of adult women (Young et al., 2002) with the incidence increasing because of the dramatic rise in obesity (Bhattacharjee et al., 2012). Weight change predicts the incidence of OSA, and a 10% increase in weight is associated with a 32% increase in the apnea/hypopnea index (Peppard et al., 2000a). Furthermore, OSA is an important contributor to the morbidity and mortality associated with obesity (Gozal and Kheirandish-Gozal, 2009;Tuomilehto et al., 2012). OSA is defined as the cessation of breathing caused by the repetitive, episodic collapse of the pharyngeal airway due to an obstruction or increased airway resistance. The first modern description of OSA was by Burwell and colleagues (1956) but was documented much earlier (Bickelmann et al., 1956;Bray, 1994;Lavie, 1984). OSA is distinguished from central apnea (CA), which is primarily caused by the cessation of the central respiratory network. CA is highly prevalent in congestive heart failure but is also present in normal subjects (Eckert et al., 2009a).
The distinction between each form of apnea, however, is not straightforward. OSA (Fig. 1) as well as CA is the result of complex interactions between the peripheral and central nervous system (Eckert et al., 2009a). These interactions lead to short-term and long-term changes that contribute to the evolution of OSA and CA. Consequences of these disorders include excessive daytime somnolence, neurocognitive impairment, and increased risk for accidents related to sleep deprivation Jordan and White, 2008;Kim et al., 1997;Young et al., 1997). In addition to these neurological consequences, the number of apneas that patients experience is positively correlated with an increased risk of hypertension (Peppard et al., 2000b). Other serious cardiovascular morbidities include increased risk for stroke, coronary artery disease, and heart failure (Phillips, 2005).
Mechanistically, increased sympathetic activity, endothelial dysfunction, and systemic inflammation as well as oxidative stress are all contributors to myocardial damage and hypertension (Baguet et al., 2012). Thus, the airway obstruction in OSA as well as CA is the beginning of a complex series of events that affect numerous central and peripheral neuronal and cardiovascular mechanisms (Eckert et al., 2009a;Gozal et al., 2013;Jordan and White, 2008;Leung and Bradley, 2001;Meier and Andreas, 2012;Susarla et al., 2010). Some of the long-term consequences of Upper trace: integrated extracellularly recorded hypoglossus activity obtained simultaneously with an intracellular recording from an inspiratory neuron located within the pre-Bötzinger complex (lower trace). The recordings were obtained in an isolated transverse slice preparation shown in the schematic. The preparation contains the functionally active pre-Bötzinger complex (blue), the nucleus tractus solitaries (NTS, pink), which in the intact animal would receive reflex inputs. Contained within the slice is also the hypoglossal nucleus (XII, green), which in the intact animal would innervate the genioglossus muscle. Note the XII activity is not always phase-locked with the pre-Bötzinger complex resulting in XII apneas that are uncoupled from the respiratory rhythm generator located within the pre-Bötzinger complex. (Figure modified from Ramirez et al., 1996). The actual data shown on the left panel inspired the middle panel, which schematically illustrates how the rhythmically active pre-Bötzinger complex (blue trace) could continue to activate the phrenic activity (black trace) resulting in diaphragmatic activity, while activity in the XII becomes uncoupled resulting in a cessation of XII activity (green trace, apnea). The right panel illustrates the anatomical components contributing to an airway occlusion: A decreased activity in the genioglossus (tongue in the schematic) and continued activity in the diaphragm (schematically drawn beneath the lung) will lead to the negative pressure that results in a pharyngeal collapse. The bilaterally organized pre-Bötzinger complex isolated in the transverse slice preparation, which was obtained from the lower brainstem (medulla -schematically shown in blue). The airway occlusion is the result of sleep state, neuromodulation, central respiratory drive, reflexes and the airway obstruction. These contributors are altered by long-term consequences of the intermittent hypoxia that is caused by repetitive pharyngeal collapses. For more details see text.
OSA, such as hypertension, often persist even after obstructions are eliminated or prevented through surgery or continuous positive airway pressure (CPAP) (Alchanatis et al., 2001;Vanderveken et al., 2011). Moreover, after surgical removal of the anatomical obstructions, or after treatment with CPAP, patients often remain refractory and shift toward the generation of central apneas (Boyd, 2009;Eckert et al., 2009b;Susarla et al., 2010).
In this review we use OSA as a template to discuss the complex interactions between factors that contribute to apnea pathogenesis. The first key concept we hope to convey is that OSA results from the convergence of multiple peripheral and central nervous system factors, not a single factor in isolation. The second concept is that many of the peripheral and central nervous system changes associated with OSA are initially reversible, and possibly even adaptive, but they become detrimental and irreversible during disease progression.

Airway obstruction and the importance of hypoglossal activity
Various anatomical abnormalities can contribute to the airway obstructions associated with OSA. Thus surgical procedures to remove these obstructions need to be adapted to the individual pattern and type of airway obstruction (Bhattacharjee et al., 2010;Sher et al., 1996). Obstructions can include macroglossia, adenotonsillar hypertrophy, increased nasal resistance, pharyngeal edema, and craniofacial abnormalities such as micrognathia and retrognathia (Bhattacharjee et al., 2010;Enoz, 2007;Lam et al., 2010;Prabhat et al., 2012;Shott and Cunningham, 1992;Verbraecken and De Backer, 2009;White, 2005;Won et al., 2008). Craniofacial factors are particularly important for pediatric OSA (Gozal, 2000).
However, alone none of these anatomical determinants is sufficient to cause an airway occlusion.
Under normal conditions airflow is facilitated by a central respiratory drive to the upper airways ( Fig. 1). Of critical importance are the hypoglossal (XII) motoneurons that innervate the genioglossus muscle via the medial branch of the hypoglossal nerve. The genioglossus muscle is the largest extrinsic muscle of the human tongue (Abd-El-Malek, 1938;Saboisky et al., 2007;Takemoto, 2001). Innervation of the genioglossus is very complex and many types of phasic and tonic motor units originating in the hypoglossal motor nucleus contribute to the genioglossus contraction (Haxhiu et al., 1992;Hwang et al., 1983;Saboisky et al., 2007). The XII motoneurons phasically activate the genioglossus muscle during each inspiration (Fig. 1), and some activity is maintained during expiration (Akahoshi et al., 2001;Fogel et al., 2001;Otsuka et al., 2000;Saboisky et al., 2010;Sauerland and Harper, 1976).
Overall, however, respiratory drive increases genioglossus muscle tone preferentially during inhalation, resulting in a contraction that pulls the tongue forward (Brouillette and Thach, 1979) and enlarges the upper airways (Bailey and Fregosi, 2004;Fuller et al., 1999;Mann et al., 2002;Oliven et al., 2001;Sokoloff, 2000). This mechanism largely prevents airway collapse during wakefulness. Indeed during wakefulness, electromyography (EMG) activity of the genioglossus is enhanced in OSA patients when compared to controls (Fogel et al., 2001;Mezzanotte et al., 1992), an adaptation that seems to compensate for the increased upper airway resistance and compliance that characterizes OSA patients (Malhotra and White, 2002;Randerath, 2007;Saboisky et al., 2007). However, during sleep or while anesthetized, the central respiratory drive to the genioglossus muscle weakens, and, as a consequence, anatomical obstructions can occlude the airway during inhalation  ( Eastwood et al., 2002;Remmers et al., 1978;Sauerland and Harper, 1976). Because a decreased central drive during sleep is necessary for the occlusion to occur during inhalation, OSA must be considered as a neuronal issue.
Indeed, airway obstructions are promoted by multiple central and peripheral nervous systems factors. These factors include sleep state-dependent pathologies and respiratory instabilities that are caused by loop gain changes as has been discussed in great detail (Thomas et al., 2004;White, 2005). Yet, whether and how an obstruction causes the cessation of breathing, i.e. the actual apnea, are not trivial questions. It is safe to conclude that the mechanisms and events leading to apneas are not fully understood, and that multiple factors must come together. In the following section we will discuss some of the potential mechanisms that contribute to the apnea.

The pharyngeal collapse
Cessation of airflow with continued respiratory effort is the hallmark of OSA (Praud et al., 1988;Remmers et al., 1978;Zucconi et al., 1996). Fig. 2 illustrates two example traces from OSA patients (A from, Praud et al., 1988;B from, Remmers et al., 1978). In both examples oro-nasal flow is blocked, while respiratory efforts continue in the abdomen. From a biomechanical perspective, continued respiratory effort in the thorax/abdomen increases thoracic volume and decreases pressure at the level of the pharynx, which would normally enable air to flow into the lungs. But, due to the increased oro-nasal resistance, the decreased pharyngeal pressure results in negative intraluminal pressure within the upper airways and leads to pharyngeal collapse (Fig. 1). Considering the mismatch between negative intraluminal pressure and the decreased airflow arriving through the upper airways, OSA may not only result from an upper airway obstruction, but it could also be caused by an imbalance in lung volume compared to upper airway size. Thus, various anatomical causes together with decreased XII activation are important contributors to the pharyngeal collapse and thus to the airway occlusion in OSA ( Figs. 1 and 2).

Neuromodulators and transmitters
Multiple neuronal mechanisms contribute to a sleep-related decrease in XII activation as both neurotransmitter and neuromodulatory systems undergo drastic state dependent changes. As demonstrated in intracellular recordings, glutamatergic and GABAergic mechanisms (Chase et al., 1989;Funk et al., 1997;Soja et al., 1987Soja et al., , 1991 as well as a powerful glycinergic premotor inhibitory system likely contribute to the REM specific decrease in XII motoneuron activity (Yamuy et al., 1999). However, the degree of inhibition may only be detectable in intracellular recordings, while active inhibition is difficult to demonstrate in EMG recordings (Funk et al., 2011). This difficulty may partly explain why the relative importance of fast neurotransmission remains a matter of discussion (Chan et al., 2006;Morrison et al., 2003aMorrison et al., , 2003b. In addition to increased active inhibition by fast synaptic transmitters, there is also a pronounced sleep related decrease in the activity of noradrenergic (Aston-Jones and Bloom, 1981) and serotonergic neurons (Jacobs and Fornal, 1991;Leung and Mason, 1999) suggesting that the loss of noradrenergic and serotonergic neuromodulatory inputs play critical roles (Fenik et al., 2005a;Funk et al., 2011;Horner, 2008Horner, , 2009Kubin et al., 1998;Ladewig et al., 2004). This hypothesis is consistent across various manipulations in unrestrained animals (Chan et al., 2006;Morrison et al., 2003a;Sood et al., 2005Sood et al., , 2007, slice preparations (Funk et al., 1994;Viemari and Ramirez, 2006), and with research in the so-called carbachol model for rapid eye movement (REM) sleep (Fenik et al., 2004(Fenik et al., , 2005a(Fenik et al., ,b,c, 2008. The noradrenergic neurons from the A5 and A7 regions converge at the level of the XII motoneurons (Aldes et al., 1992) and seem to have their effect through ␣1 adrenergic receptor activation (Parkis et al., 1995;Selvaratnam et al., 1998;Volgin et al., 2001). Interestingly, the pre-Bötzinger complex (preBötC), an area critical for breathing also receives noradrenergic and serotonergic inputs and is activated by a variety of serotonergic and adrenergic receptors (Doi and Ramirez, , 2010Lalley et al., 1995;Pena and Ramirez, 2002;Ptak et al., 2009;Tryba et al., 2006;Viemari et al., 2011;Viemari and Ramirez, 2006). These neuromodulators are particularly important during hypoxic conditions since endogenous noradrenergic and serotonergic drive within the ventrolateral medulla is essential for the activation of bursting neurons that depend on the persistent sodium current (Pena and Ramirez, 2002;Viemari et al., 2011). It is likely that a decrease in noradrenergic and serotonergic drive during sleep will weaken respiratory network activity and thus may contribute to or exaggerate the instabilities associated with OSA. Thus, noradrenergic and serotonergic excitatory inputs may play a role in the modulatory effects on both the central respiratory network and the XII motor output.
Other neuromodulators will also play important roles. Acetylcholine could be a key modulator involved in modulating respiratory activity (Shao and Feldman, 2009;Tryba et al., 2008) and suppressing genioglossus activity during REM sleep (Bellingham and Berger, 1996; Bellingham and Funk, 2000;Grace et al., 2013;Liu et al., 2005;Robinson et al., 2002). The recent study by Grace et al. (2013) demonstrated that REM specific suppression can be overcome by injecting muscarinic antagonists into the XII motoneuron pool (Grace et al., 2013). At the cellular level, this inhibitory effect appears to involve the activation of G protein-coupled inward rectifying potassium (GIRK) channels. These modulatory mechanisms appear to suppress XII motor activity by acting on the motoneurons themselves (Grace et al., 2013). This cholinergic drive could come from XII premotor neurons, a subpopulation of which is cholinergic (Volgin et al., 2008).
It is important to note, that the neuromodulatory mechanisms contributing to OSA and CA are likely very different. The number of apneas significantly increases during REM sleep in OSA patients, and some patients show apneas exclusively during REM sleep (Eckert et al., 2009b;Findley et al., 1985;Kass et al., 1996). By contrast, the number of central apneas is lowest during REM sleep (Eckert et al., 2007a). Thus, further research will need to explain how the modulatory and activity characteristics associated with the different sleep states relate to the different forms of apnea. REM sleep is characterized by decreased firing of noradrenergic and serotonergic neurons, which could lead to decreased activation of respiratory neurons within the preBötC (Funk et al., 2011;Pena and Ramirez, 2002;Viemari et al., 2011). Such a decreased activation could contribute to a weakened central drive to the hypoglossal nucleus that could suffice to predispose the upper airways to a pharyngeal collapse. However, it is more difficult to understand why the incidence of CA should decrease under these conditions. One possibility is that CAs occur less often during REM sleep because excitatory cholinergic inputs are capable of compensating for decreased levels of norepinephrine and serotonin. This compensatory mechanism is not sufficient to prevent pharyngeal collapse but is sufficient to prevent CAs in the absence of an airway obstruction.
With regard to OSA the modulatory mechanisms during REM sleep could indeed explain not only decreased activation of the respiratory network but also a decrease in airway tone (Remmers et al., 1978;Sauerland and Harper, 1976). Yet either mechanism can only partly explain how decreased XII motoneuronal activation predisposes the upper airways to a pharyngeal collapse. Thus, it remains uncertain how the apneas themselves are generated.
Indeed, the possibility that modulators are causing the decreased tone but not the apnea itself is consistent with the wellknown inefficiency of aminergic therapies that have largely failed to alleviate OSA (Dempsey et al., 2010;Funk et al., 2011). Moreover, noradrenergic and serotonergic innervation is strengthened following exposure to chronic intermittent hypoxia which may oppose the decreased muscle tone during sleep (Rukhadze et al., 2010), and OSA patients show a variety of neurogenic changes in the upper airways that could potentially compensate for decreased muscle tone. These adaptations include increased activation, earlier firing, and increased sprouting of the XII motoneurons (Saboisky et al., 2007(Saboisky et al., , 2012. As illustrated in Fig. 2 there is not a general suppression of the upper airways, but instead the airflow is "suddenly" disturbed for a few cycles and then the oral-nasal flow reappears and re-synchronizes with the respiratory abdominal muscles. Thus, while a persistently decreased drive to the XII motoneurons may predispose the pharynx to sudden collapse, the sudden failure in XII motor activity cannot be entirely explained by altered modulatory tone generating persistent atonia during a specific sleep state.
As illustrated in Fig. 2, genioglossus EMG activity is specifically weakened and less phasic during the airway occlusion but not before or after the occlusion. Thus, in addition to a neuromodulatorand transmitter-driven decrease in muscle tone, one needs to consider additional central nervous and reflex mechanisms that contribute to the disconnect between the ongoing phasic respiratory activity that drives the diaphragmatic activity and the decrease in phasic respiratory drive to the XII motoneurons which is specifically associated with the airway occlusion.

Reflex regulation
In OSA, airway occlusion also involves reflex mechanisms ( Fig. 1) that are characterized by pathological gain changes in the mechano-and chemosensory reflex loops regulating ventilation. These reflex pathways become specifically dysregulated during sleep and could therefore destabilize the respiratory response to an airway obstruction resulting in pharyngeal collapse during sleep and not wakefulness (Douglas et al., 1982;White, 2005).

Mechano-sensory reflexes
To prevent pharyngeal collapse, mechanoreceptors located within the pharyngeal walls specifically regulate the XII motoneurons (Fig. 1). These receptors are activated by negative intraluminal pressure generated during inspiration. They transmit this afferent information via the superior branch of the internal laryngeal nerve, and genioglossus premotoneurons located near the obex mediate the reflex (Chamberlin et al., 2007). This is an important reflex, as activation of the hypoglossal muscles caused by a pressure drop should counteract a pharyngeal collapse (Eckert et al., 2007b;Horner et al., 1991;Malhotra et al., 2000). Under physiological conditions this mechano-sensory pathway, as well as central nervous system components that are not involved in the reflex, contribute to the phasic genioglossus contraction during inspiration (Chamberlin et al., 2007;Fogel et al., 2001;Horner, 2000;Susarla et al., 2010;van Lunteren, 1993). Importantly, the reflex activation of the genioglossus during these pressure drops is dramatically reduced or even suppressed during sleep, a finding that is of great significance in understanding OSA because a reduced activation could promote a pharyngeal collapse (Wheatley et al., 1993).

Arterial chemosensory mechanisms
Hypoxia and hypercapnia initiated chemoreflexes are known to contribute to the regulation of ventilation (Fig. 1), and a high gain in any of these chemosensory loops could contribute to breathing instabilities (White, 2005). The following lines of evidence suggest that the arterial chemoreflex is augmented in OSA subjects: (a) brief hyperoxic exposure, which inhibits chemoreceptor activity, reduces blood pressure in OSA patients but not in control subjects (Narkiewicz et al., 1998), (b) the hypoxic ventilatory response, a hallmark response of the chemoreflex, is augmented in OSA subjects compared to controls (Hedner et al., 1992), and (c) activation of muscle sympathetic nerve activity by apneas is more pronounced in OSA subjects compared to controls (Smith et al., 1996). Development of altered chemosensory reflexes in OSA is further supported by studies using intermittent hypoxia (IH), the hallmark manifestation of recurrent apnea. Rodents exposed to chronic IH showed: (a) enhanced carotid body sensitivity to hypoxia, and (b) a progressive increase in baseline carotid body sensory activity, a phenomenon termed sensory long-term facilitation (sLTF) (Pawar et al., 2008;Peng et al., 2003Peng et al., , 2006Peng et al., , 2009Peng and Prabhakar, 2004;Rey et al., 2004). The subnuclei of the nucleus tractus solitarius (NTS, Fig. 1), especially the commissural part of the NTS (cNTS), receive inputs from the carotid body (Chitravanshi and Sapru, 1995;Zhang and Mifflin, 1993). Neuronal activity in cNTS is regulated by various neurotransmitters, including glutamate, an excitatory amino acid transmitter, and dopamine, an inhibitory biogenic amine. Chronic IH up regulates GluR2/3 glutamate receptor subunit expression in cNTS (Costa-Silva et al., 2012) and down regulates tyrosine hydroxylase (TH) expression, the rate-limiting enzyme in dopamine (DA) synthesis (Gozal et al., 2005;Kline et al., 2002). It is likely that chronic IH, by down regulating the synthesis of DA, enhances glutamatergic excitatory transmission in NTS (Chen et al., 1999) resulting in the enhanced hypoxic ventilatory response (HVR). Collectively, these studies indicate that the augmented chemoreflex by chronic IH involves reconfiguration of neurotransmitter profiles in the central nervous system.
Does an augmented chemoreflex contribute to pathogenesis of apnea? It was proposed that the increased carotid body sensitivity to hypoxia can lead to a greater magnitude of hyperventilation during each episode of apnea, thus driving the respiratory controller below the apneic threshold for CO 2 , leading to greater number of apneas (Prabhakar, 2001). In other words, the heightened hypoxic sensitivity of the carotid body might act as a "positive feedback," thereby exacerbating the occurrence of apneas. Supporting such a possibility is the finding that chronic IH exposed rats with intact carotid bodies exhibit greater incidence of spontaneous apneas. This effect was absent in carotid body sectioned rats exposed to chronic IH (Prabhakar, 2013). Since peripheral chemoreceptors regulate hypoglossal motoneuron activity (Bruce et al., 1982), it remains to be established whether the chemoreflex directly or indirectly contributes to the hypoglossal motoneuron dysfunction leading to OSA.

Airway occlusion and the central respiratory network
Chemo-and mechanosensory afferents and modulatory inputs converge via the NTS on the XII motoneurons where they closely interact with the central respiratory drive acting on the XII motoneurons. As shown in Fig. 1, an apnea generated at the level of the XII motoneurons could involve a temporary drop-out of central XII activity while respiratory rhythmic activity continues to be generated within the central respiratory network. Neuronal mechanisms that could lead to such a drop out could occur locally within the medulla. Located within the same transverse plane as the XII nucleus is the pre-Bötzinger complex (preBötC; Fig. 1). The preBötC is a well-defined neuronal network that is essential for breathing. Selective lesion of the preBötC in intact animals abolishes breathing (Gray et al., 2010;Ramirez et al., 1998;Tan et al., 2008). Moreover, isolated in medullary slice preparations that encompass the preBötC (Fig. 1, preBötC, blue), this neuronal network continues to spontaneously generate inspiratory rhythmic activity (Fig. 1, left  panel). Inspiratory activity generated within the preBötC is transmitted to the XII nucleus and leads to the phasic activation of an inspiratory population burst within the hypoglossal nucleus (Fig. 1,  left panel). Located within this slice preparation are premotor neurons that transmit the respiratory signal from the preBötC to the XII motoneurons (Chamberlin et al., 2007;Dobbins and Feldman, 1995;Luo et al., 2006;Peever et al., 2002;Sebe and Berger, 2008). These premotor neurons are not simple followers, but they seem to play a major role in generating synchronous oscillations within the XII motor nucleus (Sebe and Berger, 2008). Interestingly, not every burst in the preBötC is transmitted to the XII and in some slice preparations the XII can fail to burst in phase with the respiratory cycle generated within the preBötC ( Fig. 1; Ramirez et al., 1996). It is conceivable that such an activation failure could provide a mechanistic explanation for XII inactivity during continued inspiratory respiratory rhythm generation from the preBötC. The inspiratory rhythm generated in the preBötC would then continue to be transmitted to the phrenic nucleus (Fig. 1). Continued activation of the diaphragm is an important aspect of OSA, as it is the activated diaphragm that produces the expansion of the thorax which together with a lack of genioglossus activation creates negative pressure and pharyngeal collapse. At this point, we do not know how the respiratory drive from the preBötC is transmitted under conditions that mimic sleep states or conditions that mimic sleep apnea. During hypoxia, however, transmission failure from the pre-BötC to the XII motoneurons is increased . Thus, an important avenue for future research will be to understand how chronic intermittent hypoxia or certain neuromodulatory conditions associated with sleep can cause such transmission failures between the respiratory rhythm generator and the XII motor output. Investigating this issue could provide important and much needed clues into the pathology of OSA.

The recovery from airway occlusion
In addition to the onset and maintenance of airway occlusion, recovery from an airway obstruction has been the subject of intense discussions (Fig. 2). One notion is that reflex recruitment of pharyngeal dilator muscles is insufficient to open the airway once it is occluded and that arousal is required for the termination. This is an important consideration, since breathing instabilities that promote OSA likely involve pathological changes in arousal threshold (Younes, 2004). Arousal is stimulated by increased negative pharyngeal pressure and increasingly hypoxic and hypercapnic conditions that in turn increase respiratory drive (Berry and Gleeson, 1997;Gleeson et al., 1990;Kimoff et al., 1994). The stimulation of arousal then activates dilator activity and opens the airways (Remmers et al., 1978). Yet, arousal is not required for apnea termination (Younes et al., 2012). Fig. 2 illustrates the recovery from an airway occlusion which is typically abrupt and associated with a sudden burst of genioglossus EMG (Berry and Gleeson, 1997;Rees et al., 1995;Remmers et al., 1978;Wulbrand et al., 1998Wulbrand et al., , 2008. The abrupt increase in genioglossus muscle activity seems to be due to the recruitment of phasic inspiratory motor units (Wilkinson et al., 2010). Wulbrand and co-workers proposed that this burst is synchronized with the generation of sigh or sigh-like neuronal mechanisms . The induction of sighs by airway occlusion has also been reported by Alvarez et al. (1993). Sighs are large amplitude inspiratory efforts that are mechanistically different from normal "eupneic" breaths ( Fig. 3) (Lieske and Ramirez, 2006a,b;Lieske et al., 2000). Although, it was initially believed that sighs are exclusively dependent on lung stretch receptor stimulation (Bartlett, 1971;Reynolds, 1962;Wulbrand et al., 2008); there is ample evidence to the contraryi.e. sighs are generated within the central nervous system and do not require afferent input (Orem and Trotter, 1993). For example sighs can be generated following deafferentation in vivo (Cherniack et al., 1981), and humans continue to generate sighs following lung transplantation (Shea et al., 1988).
Moreover, sighs are even generated in the in situ working heart preparation, a fully deafferented brainstem preparation, (Ramirez and Viemari, 2005), as well as transverse slice preparations that contain the preBötC (Fig. 3) (Hill et al., 2011;Lieske et al., 2000;Pena et al., 2004). Important for the discussion of OSA, this centrally generated mechanism is specifically facilitated under hypoxic conditions (Bartlett, 1971;Bell et al., 2009;Bell and Haouzi, 2010;Cherniack et al., 1981;Hill et al., 2011;Lieske et al., 2000;Schwenke and Cragg, 2000). Although peripheral chemoreceptors certainly play a facilitatory role (Cherniack et al., 1981;Glogowska et al., 1972;Matsumoto et al., 1997), even in the absence of peripheral chemoreceptors, hypoxic conditions within the preBötC are sufficient to centrally activate the generation of sighs (Hill et al., 2011;Koch et al., 2013;Pena et al., 2004;Telgkamp et al., 2002). Thus, the hypoxic conditions associated with OSA will likely play a role in activating sighs. As characterized in infants, sighs triggered by an airway occlusion are coordinated with a sleep startle, that marks the beginning of arousal (Figs. 3 and 4), and accompanying changes in electroencephalogram (EEG) and EMG activity (Wulbrand et al.,  Fig. 3. Mechanisms of sigh and post-sigh apnea. Left panel: Upper trace: integrated extracellularly recorded pre-Bötzinger complex population activity obtained simultaneously with an intracellular recordings from an inspiratory neuron located within the pre-Bötzinger complex (middle trace). The lower trace represents a recording from a human subject. While the lower trace was obtained using induction plethysmographic recording (for more details see Weese-Mayer et al., 2006), the upper and middle traces were obtained in an isolated transverse slice preparation shown in the schematic. The preparation contains the functionally active pre-Bötzinger complex (blue), the nucleus tractus solitaries (NTS, pink), which in the intact animal would receive reflex inputs. Contained within the slice is also the hypoglossal nucleus (XII, green), which in the intact animal would innervate the genioglossus muscle. The actual data shown on the left panel inspired the middle panel, which schematically illustrates how the rhythmically active pre-Bötzinger complex (blue trace) could activate a sigh and apnea in the phrenic nucleus (black trace) resulting in diaphragmatic activity, and sigh activity and apnea in the XII (green trace, apnea). The right panel illustrates the anatomical components contributing to the central apnea associated with the sigh, as well as the illustration of arousal that is associated with the generation of the sigh. The bilaterally organized pre-Bötzinger complex isolated in the transverse slice preparation was obtained from the lower brainstem (medulla -schematically shown in blue). The sigh is activated by changes in the state, oxygenation and CO2 levels. For more details see text.
2008). Although cortical arousal is not always observed, sighs consistently coincide with a sudden rise in limb EMG activity and a distinct neck extension, an adaptive response that can contribute to the termination of an airway occlusion (Wulbrand et al., 2008). Not only in infants, but also in adults, sighs are linked to EMG activation and EEG changes (Perez-Padilla et al., 1983).
Sighs are also associated with a heart rate increase followed by a heart rate decrease (Haupt et al., 2012;McNamara et al., 1998;Porges et al., 2000;Weese-Mayer et al., 2008;Wulbrand et al., 2008). The heart rate changes associated with the sigh are often altered in human diseases such as familial dysautonomia, sickle cell anemia, and SIDS (Franco et al., 2003;Sangkatumvong et al., 2011;Weese-Mayer et al., 2008). From the organismic perspective, the hypoxic sensitivity of the sigh becomes a very important central nervous system mechanism that allows the organism to detect and respond to hypoxic conditions. Hypoxia-induced sighs activate important muscles and can lead to subcortical and cortical arousal (Fig. 3). Once aroused, an organism can avoid the hypoxic condition by for example changing its sleeping position. The sigh may therefore link the hypoxic condition caused by OSA to arousal, which eventually results in sleep deprivation, one of the detrimental consequences of OSA. Interestingly sighs may also play an important role in the generation of periodic breathing as postulated by (Guntheroth, 2011).
This centrally generated mechanism is very sensitive to state changes (Orem and Trotter, 1993). The transition from sleep to wakefulness is often characterized by the activation of a sigh and Fig. 4. The generation of the sigh is associated with arousal. The example was obtained during the transition from sleep to wakefulness. Note, the sigh (lower trace) is associated with a prolonged central apnea that follows a decrease in the level of CO2, which was caused by the sigh, suggesting that the hypocapnia may contribute to the generation of the apnea that followed the sigh. arousal ( Fig. 4) (Eckert et al., 2007a). Note, in Fig. 4, the sigh seems to contribute to a decrease in CO 2 level. This decrease in CO 2 may be involved in the generation of the apnea that typically follows the sigh. Indeed, during an "augmented" breath simulated by a ventilator, a decreased CO 2 drive can generate a brief apnea as elegantly demonstrated by Remmers et al. (1978). However, these simulated augmented breaths evoked brief apneas only under certain conditions such as hypoxia. Moreover, we know that the post-sigh apnea can be generated centrally within the preBötC under conditions in which oxygen and CO 2 are not altered (Fig. 3). Thus, the post-sigh apnea is indeed a "central apnea" generated within the ventrolateral medulla. Interestingly, a "post-sigh-like apnea" can be simulated centrally, by maximally stimulating isolated medullary respiratory pacemaker neurons. This purely central electrical stimulation is followed by a prolonged pause in the rhythmic bursting of these respiratory neurons (Tryba et al., 2008).
The post-sigh apnea is an important manifestation of a central apnea (Eckert et al., 2007a;Radulovacki et al., 2001;Saponjic et al., 2007). Post-sigh apneas are very common in children (Haupt et al., 2012;O'Driscoll et al., 2009) but are also present in adults (Vlemincx et al., 2010). Post-sigh apneas can be exaggerated in neurological disorders such as Leigh Syndrome (Quintana et al., 2012;Saito, 2009;Yasaki et al., 2001), Familial Dysautonomia (Weese-Mayer et al., 2008, and Rett Syndrome (Voituron et al., 2010). Although it is clearly generated centrally, it must be emphasized that in the intact organism, additional chemosensory mechanisms will contribute and potentially exaggerate the post-sigh apnea because the post-sigh apnea is associated with significant changes in blood gases.

Conclusion
Apneas emerge through a complex interplay between peripheral and central nervous system factors that affect all levels of integration: from the molecular to the cellular and organismic level. This interplay affects many aspects of respiratory control making it difficult to clearly separate central versus peripheral contributions to the generation of the apnea. Indeed, several central and peripheral factors must come together to cause pharyngeal collapse and apneas in case of OSA. Moreover, OSA as well as central apneas have numerous long-term consequences that include changes in the neuromodulatory milieu, mechano-and chemosensory reflex loops, cardiorespiratory integration and neurotransmitter systems. These changes may be partly adaptive during wakefulness but they often fail to adequately adapt the organism during the night. Indeed, many of the consequences become critical contributors to the morbidity of the apneas. While traditionally, much emphasis has been placed on understanding the contributions of chemoand mechanosensory reflexes, the changes in blood gases, and the biomechanics of the apneas, we have only recently begun to understand how these contributors interact with the central respiratory network, an integration that still raises many unanswered questions. Future research will elucidate many of these questions and may inspire novel avenues for therapies that could target the most detrimental and persisting consequences of sleep apnea, a health issue that affects an increasing proportion of the pediatric and adult populations.