Competing mechanisms of plasticity impair compensatory responses to repetitive apnea

Abstract

Many forms of sleep apnea are characterized by recurrent reductions in respiratory neural activity, which leads to inadequate ventilation and arterial hypoxia. Both recurrent reductions in respiratory neural activity and hypoxia activate mechanisms of compensatory plasticity that augment inspiratory output and lower the threshold for apnea, inactivity-induced inspiratory motor facilitation (iMF) and long-term facilitation (LTF), respectively. However, despite frequent concurrence of reduced respiratory neural activity and hypoxia, mechanisms that induce and regulate iMF and LTF have only been studied separately. Here, we demonstrate that recurrent reductions in respiratory neural activity (“neural apnea”) accompanied by cessations in ventilation that result in moderate (but not mild) hypoxemia does not elicit increased inspiratory output, suggesting that concurrent induction of iMF and LTF occludes plasticity. A key role for NMDA receptor activation in impairing plasticity following concurrent neural apnea and hypoxia is indicated since recurrent hypoxic neural apneas triggered increased phrenic inspiratory output in rats in which spinal NR2B-containing NMDA receptors were inhibited. Spinal application of retinoic acid, a key molecule necessary for iMF, bypasses NMDA receptor-mediated constraints, thereby rescuing plasticity following hypoxic neural apneas. These studies raise the intriguing possibility that endogenous mechanisms of compensatory plasticity may be impaired in some individuals with sleep apnea, and that exogenously activating pathways giving rise to respiratory plasticity may be a novel pharmacological strategy to improve breathing.

Introduction

Sleep apnea is an insidious disorder estimated to affect 34% of men and 17% of women aged 30–70 years (Peppard et al., 2013). Untreated sleep apnea causes substantial deleterious health consequences, including hypertension, cardiovascular morbidities, cognitive deficits, metabolic dysfunction and premature death (Abbott and Videnovic, 2016; Chowdhuri et al., 2016; Marshall et al., 2008; Mehra and Redline, 2014; Shamsuzzaman et al., 2003; Somers et al., 2008; Yaffe et al., 2011; Young et al., 2008). Two diagnostic subclasses are often used to define sleep apnea: obstructive sleep apnea (OSA) and central sleep apnea (CSA). Whereas OSA is characterized by a narrowing of the upper airway that impedes airflow during attempted breaths, CSA is characterized by insufficient neural drive to respiratory pump muscles. Recent evidence demonstrates CSA and OSA often co-exist in the same individual and that insufficient inspiratory motor output may be a precipitating factor for OSA (Dempsey et al., 2014; Hoffman and Schulman, 2012; Westhoff et al., 2012), suggesting reduced respiratory neural activity may be a common etiology for CSA and OSA. Treatment options for sleep apnea are strikingly limited, and patient adherence to existing therapies is poor (Rotenberg et al., 2016).

Despite the prevalence of sleep apnea, the neural system controlling breathing exhibits a profound capacity to adapt in response to a perceived failure through compensatory plasticity (Fuller and Mitchell, 2017). Two well-studied forms of compensatory respiratory plasticity are inactivity-induced inspiratory motor facilitation (iMF) and hypoxia-induced long-term facilitation (LTF), respectively triggered by recurrent reductions in respiratory neural activity (neural apnea or hypopnea; (Mahamed et al., 2011; Streeter and Baker-Herman, 2014a) and intermittent hypoxemia (Bach and Mitchell, 1996; Fuller et al., 2000; Mahamed and Mitchell, 2008). Both iMF and LTF are reflected as a long-lasting increase in inspiratory motor output due to (distinct) mechanisms operating at the level of the inspiratory motor pool (Baker-Herman et al., 2004; Baker-Herman and Mitchell, 2002; Streeter and Baker-Herman, 2014a; Strey et al., 2012), and are associated with a decrease in the CO₂ threshold for apnea (Baertsch and Baker, 2017b; Mahamed and Mitchell, 2008). In the case of iMF, enhanced inspiratory motor output is proportional to the magnitude of activity deprivation (Braegelmann et al., 2017) and the reduction in the apneic threshold is proportional to the magnitude of iMF (Baertsch and Baker, 2017b). Most of our understanding regarding these distinct forms of respiratory plasticity derive from studies investigating iMF or LTF in phrenic motor neurons, which drive the diaphragm. Although the physiological significance for iMF and LTF has not yet been conclusively demonstrated, both are proposed to stabilize breathing and prevent future apneas/hypopneas by strengthening inspiratory motor output and lowering the propensity for apnea (Braegelmann et al., 2017; Mahamed and Mitchell, 2007; Mahamed and Mitchell, 2008; Mateika and Komnenov, 2017; Mateika et al., 2004; Mateika and Syed, 2013; Strey et al., 2013).

A central question, then, is: if the respiratory control system is endowed with an endogenous capacity to compensate following recurrent apnea/hypopnea, why don’t these mechanisms mitigate the severity of sleep apnea? To date, adaptive responses to reduced respiratory neural activity have only been studied without concomitant hypoxia and vice versa. Despite similar adaptations in inspiratory motor output, iMF and LTF elicit plasticity through mechanistically distinct signaling pathways (Fields and Mitchell, 2015; Strey et al., 2013; Strey et al., 2012); recent progress in our understanding of these different mechanisms has provided a platform for understanding how these processes may interact to undermine compensatory plasticity expression. One such example is a contrasting requirement for spinal NMDA receptor activation: whereas spinal NMDA receptor activation is necessary for induction and maintenance of LTF (McGuire et al., 2008; McGuire et al., 2005), spinal NMDA receptor activation constrains iMF expression (Streeter and Baker-Herman, 2014b). Thus, we hypothesize that a contrasting drive to enhance and reduce NMDA receptor activity undermines the capacity for either stimulus to elicit plasticity when presented together, potentially explaining the absence of compensatory breathing adaptions in individuals with sleep apnea.

Here, we demonstrate that compensatory plasticity is constrained when recurrent reductions in respiratory neural activity are accompanied by moderate (but not mild) hypoxemia, and we identify molecular targets that regulate these competing mechanisms. Specifically, we show that hypoxia-induced activation of spinal NR2B-containing NMDA receptors obstruct expression of iMF. Spinal application of all-trans retinoic acid, a molecule necessary for iMF expression (Baertsch and Baker, 2017a), overcomes NMDA receptor-induced constraints to enable compensatory respiratory plasticity. This is the first study to provide a possible mechanistic explanation for why some individuals with sleep apnea may lack compensatory respiratory behaviors that protect against recurrent breathing disruptions, an important first-step in identifying novel targets for pharmacologically treating breathing disorders characterized by insufficient inspiratory neural activity.

Methods

Experiments were performed on 2.5–3.5 month old male Sprague-Dawley rats from Harlan Laboratories, colony 217. Rats were housed 2 per cage with 12hr light/dark cycles and food/water ad libitum. The Animal Care and Use Committee at the University of Wisconsin, Madison approved all experimental protocols.

Surgical Preparation.

Isoflurane anesthesia was induced in a closed container and continued on a nose cone with 2.5% isoflurane in 30/70 O₂/N₂ balance. A tail vein catheter was placed for urethane delivery and fluid infusion (10/5/1 lactated rings solution/hetastarch/8% sodium bicarbonate solution) in order to maintain anesthesia and blood pressure/pH homeostasis, respectively, throughout experimental protocol. The trachea was cannulated, and mechanical ventilation was begun (Model 683, Harvard Apparatus, Holliston Massachusetts; ~70 breaths/min, 2.5–3ml; 21% O₂). To facilitate apneic threshold measurements and rapid induction of neural apnea during the protocol (see below), rats were slightly hyperventilated and a small amount of CO₂ was added in the inspired line to maintain an end-tidal CO₂ (ETCO₂) near ~45mmHg (assessed through an expired airflow; Respironics). The vagus nerve was isolated and cut bilaterally at the cervical level to prevent entrainment of spontaneous respiratory frequency to the ventilator. The right femoral artery was isolated and catheterized to monitor arterial blood pressure and draw blood samples (0.3ml per sample) for pH and blood-gas analysis (ABL800; Radiometer, Copenhagen, Denmark). The C2 spinous process was exposed and a C2 laminectomy was performed over the spinal midline. A small hole was cut in the dura of the exposed spinal segment and a silicone catheter (2 French; Access Technologies) connected to a 50-μl Hamilton syringe was inserted into the intrathecal space and carefully advanced caudally (5mm) so that the tip of the catheter laid on the dorsal surface of the C4 spinal segment. The left phrenic nerve was dissected by dorsal approach, cut distally, de-sheathed, and placed on a bipolar silver electrode to record inspiratory motor output. Rats were then converted to urethane anesthesia (0.180g/100g rat i.v.) as isoflurane was gradually withdrawn over a period of ~15min. Following confirmation of adequate anesthetic depth, pancuronium bromide was infused through the tail vein (3mg/kg, i.v.) to induce neuromuscular paralysis. Periodically throughout the protocol, pressor and phrenic nerve amplitude responses to toe pinch were assessed to ensure adequate anesthesia depth. Body temperature was monitored by rectal thermometer and maintained near 37.5°C (± 1°C) with a custom heated surgery table.

Experimental Protocols.

A minimum of one hour following termination of isoflurane anesthesia, the apneic threshold (measured by ETCO₂) was identified by slowly reducing inspired CO₂ and/or ventilator settings until phrenic bursting ceased for at least 20sec. Inspired CO₂ was then slowly increased until nerve activity returned (i.e. recruitment threshold; RT; measured by ETCO₂) and raised an additional 2–3mmHg above RT for “baseline” nerve recordings. Baseline nerve activity was recorded for 15min with 2 arterial blood samples drawn 5min apart to obtain baseline PaCO₂, PaO₂, and pH measurements (temperature corrected) prior to initiation of neural/ventilator apnea challenges (described below). Blood gases were taken 15, 30, and 60min after the final neural/ventilator apnea challenge to ensure baseline parameters were maintained within ±1.5mmHg for PaCO₂ (of eupneic baseline), ±10mmHg for PaO₂ (baseline 100mmHg PaO₂), ±1.5 for pH, and ±1.0°C for temperature. Adjustments to inspired gases and/or fluids were made after each blood test if necessary. In a subset of rats, the apneic threshold was retested after the final blood gas measurement (i.e., 60 min). At the end of each protocol, rats were exposed to a maximum respiratory challenge consisting of hypercapnia (90mmHg < ETCO₂ < 100mmHg) with hypoxia (12% inspired O₂) to ensure observed results were not due to deterioration of the preparation or a lack of dynamic range in respiratory motor output. At the end of the protocol, rats were euthanized by an overdose of urethane (3–5 ml bolus of 1.6–1.8 g/kg) followed by turning off the ventilator. Heart rate and phrenic burst activity was monitored until all activity ceased for 3–5 min. Urethane overdosing was then immediately followed by a secondary method of euthanasia, such as exsanguination.

To test the hypothesis that co-induction of iMF and LTF occludes plasticity, we exposed rats to recurrent reductions in respiratory neural activity (“neural apnea” to induce iMF) that were accompanied by cessations in ventilation that resulted in moderate hypoxemia (“25sec ventilator apnea” to induce LTF). Control groups were exposed to either stimulus alone, or exposed to reductions in respiratory neural activity with brief cessations in ventilation (“6sec ventilator apnea”) that result in only mild hypoxemia and, as such, do not elicit LTF. Rats were exposed to one of six protocols: 1) 5, ~1min episodes of absent respiratory neural activity while continuing mechanical ventilation (neural apnea), 2) 5, 25sec episodes in which mechanical ventilation was temporarily stopped while respiratory neural activity remained intact (“25sec ventilator apnea”), 3) 5, 6sec episodes in which mechanical ventilation was temporarily stopped while respiratory neural activity remained intact (“6sec ventilator apnea”), 4) 5, 25sec ventilator apneas during concurrent neural apnea, 5) 5, 6sec ventilator apneas during concurrent neural apnea, and 6) a “time control” group that received the same surgical procedure without neural or ventilator apnea experiences. Sample size for each group is listed in the results section.

Neural apnea.

To induce a reduction in respiratory neural activity without concomitant hypoxia, inspired CO₂ was quickly lowered until rhythmic phrenic activity ceased, while maintaining ventilator settings (volume and frequency) at the same levels used during baseline recordings. Neural apnea was confirmed for 10sec before inspired CO₂ was returned to baseline levels, though nerve activity did not return until ~ 1min after ETCO₂ reached baseline levels. This delay in spontaneous respiratory activity is consistent with previous work from our lab (Baertsch and Baker-Herman, 2013; Baertsch and Baker-Herman, 2015). Importantly, rats were continuously ventilated during the neural apnea, and did not experience hypoxia. 5 episodes of neural apnea were given with each episode separated by 5min of baseline inspired CO₂.

Ventilator apnea.

To simulate a cessation in breathing without a reduction in respiratory neural activity, a “ventilator apnea” was induced by turning off the ventilator for either 6sec or 25sec, depending on the experimental group. This was repeated 5 times with 5min separating each ventilator apneic episode. Ventilator apneas resulted in a consistent decrease in arterial O2 (in results section), as measured by pulse oximeter (8600V, Nonin Medical Inc., Plymouth, MN, USA). Following cessation of the ventilator apnea, the ventilator was turned back on at baseline settings. Each animal naturally corrected arterial blood gases between ventilator apnea episodes without any adjustments being made to the ventilator or inspired gases.

Neural apnea with ventilator apnea.

To induce a reduction in respiratory neural activity followed by a rapid hypoxemia, a neural apnea was induced as described above; following a 10sec period of absent phrenic firing activity, the ventilator was turned off for either 6sec or 25 sec. The acute hypoxic episode resulted in a natural rescue from neural apnea ~4sec after turning the ventilator off, though the ventilator was kept off for the full 6sec or 25sec duration to maintain a consistent drop in arterial oxygen (confirmed with pulse oximetry). Since a ventilator apnea also resulted in mild hypercapnia (+ 2–5 mmHg ETCO2 above baseline), we confirmed our findings in one rat in which we induced 5 episodes of neural apnea with concomitant hypoxia (11% O2) delivered via the inspired gas line (not shown); consistent with findings presented here, plasticity was not expressed.

Pharmacological Treatments.

Rats received intrathecal injections of one of the following pharmacological compounds 15–20 min prior to initiation of the protocol: 1) 4-Diethylaminobenzaldehyde (DEAB; Sigma-Aldrich St. Louis, MO), an inhibitor of retinaldehyde dehydrogenase, 2) all-trans retinoic acid (RA; Sigma-Aldrich St. Louis, MO), an agonist of retinoic acid receptors, 3) amino-5-phosphonovaleric acid (APV; Sigma-Aldrich St. Louis, MO), an NMDA receptor antagonist, 4) Co 101244 (Tocris Bristol, UK), an antagonist of NMDA receptors containing the subunit NR2B, 5) eliprodil (Tocris Bristol, UK), an antagonist of NR2B containing NMDA receptors, or 6) vehicle. Stock solutions were dissolved in DMSO and stored at −20°C for up to 1 week. Prior to injecting, stock solutions were diluted with artificial CSF (aCSF; 120 NaCl, 3 KCl, 2 CaCl, 2 MgCl, 23 NaHCO₃, 10 glucose bubbled with 95% O₂/5% CO₂) to a concentration that was less than 20% of stock. Solutions were delivered intrathecally over spinal segments containing the phrenic motor nucleus over a 2min period. Vehicle treated rats received 10–12μl of a 20% DMSO in aCSF solution. Subgroups of time control rats were treated with each drug solution (except RA).

Statistical analyses.

Phrenic burst activity was amplified (x10k), band-pass filtered (0.3–10kHz; AM Systems), integrated (time constant 50ms), and rectified. The resulting signal was digitized and analyzed with PowerLab (AD Instruments; Lab Chart 7.0 software). 60-breath bins were taken immediately prior to blood samples at baseline, 15, 30 and 60min post final neural and/or ventilator apnea. For surgical time control groups, blood samples were taken at baseline, and 15, 30 and 60min after a mock recurrent neural/ventilator apnea protocol. Nerve burst amplitude was expressed as a percent change from baseline, and burst frequency was expressed in absolute values. There were no significant differences in phrenic burst amplitude or frequency between individual time control groups (vehicle n=7; APV n=5; DEAB n=5; Co 101244 n=3; eliprodil n=3); for analysis of the effect of NR2B-NMDA receptor inhibition on plasticity, time controls receiving Co 101244 or eliprodil were combined. Statistical differences between groups were determined using two-way repeated measures ANOVA and Tukey’s post-hoc tests. Similar results were obtained at 15, 30 and 60 min; thus, only 60 min data are graphically presented for clarity. iMF and/or LTF was considered to be expressed if phrenic inspiratory output was significantly greater than corresponding time points in surgical time control rats not experiencing neural and/or ventilator apnea.

Changes in the apneic threshold were determined by subtracting the difference between the ETCO2 level at which phrenic burst activity ceased during the first neural apnea versus the final neural apnea. A one sample t-test was used to determine if the change in the apneic threshold measurement was significantly different than zero, and a one way ANOVA and Tukey’s post-hoc tests were used to determine statistical differences between groups. Blood gas variables (e.g., PaO2 and PaCO2) were analyzed using a two-way repeated measures ANOVA, and individual time point comparisons were determined using Tukey’s post-hoc test. Groups were considered significantly different if p-values were < 0.05. Data are shown as mean ± SE.

Results

Hypoxic neural apneas do not elicit plasticity in phrenic inspiratory output

To determine the impact of co-induction of iMF and LTF on plasticity expression, we first confirmed that intermittent reductions in respiratory neural activity (neural apnea) and intermittent cessations in breathing were independently sufficient to elicit iMF and LTF, respectively. As we have previously shown (Strey et al., 2012), 5 brief episodes of reduced respiratory neural activity in ventilated rats (thus, did not experience hypoxia/hypercapnia) elicited a significant increase in phrenic inspiratory burst amplitude (54.3 ± 7.8% baseline; n=8; p<0.001; Figure 1A and B), indicating iMF. Similarly, as previously reported (Mahamed and Mitchell, 2008), 5, 25sec ventilator apneas resulted in arterial hypoxemia (71 ± 0.9% SaO₂; estimated PaO₂: 38 ± 0.7 mmHg), and elicited a significant increase in phrenic inspiratory burst amplitude (59.9 ± 4.1% baseline; n=7; p<0.001; Figure 1A and B), indicating LTF. To determine the impact of co-induction of iMF and LTF, a neural apnea was induced, then the ventilator was immediately turned off for 25sec (“hypoxic neural apnea”); this process was repeated 5 times. In contrast to when each stimulus was presented alone, intermittent hypoxic neural apnea did not elicit a significant change in phrenic inspiratory output (7.7 ± 6.1% baseline; n=7; p>0.05; Figure 1A and B), indicating that co-presentation of reduced respiratory neural activity with hypoxia occludes plasticity.

Figure 1: Hypoxic neural apneas do not elicit plasticity in phrenic inspiratory output.

A) Representative phrenic neurograms from rats exposed to intermittent reductions in respiratory neural activity (“neural apnea”; top), intermittent 25 sec cessations in ventilation (“hypoxia”; middle), and intermittent neural apnea superimposed with 25 sec ventilator apnea (“hypoxic neural apnea”; bottom). B) Average change in phrenic inspiratory burst amplitude 60 min following intermittent neural apnea or 25 sec ventilator apnea, the combination, or at equivalent point in time control rats subjected to the same surgical preparation, but no apnea. Both intermittent neural and ventilator apnea elicit long-lasting enhancements in phrenic motor output when presented separately (n=8 and n=7, respectively; p<0.001), indicating iMF and LTF, respectively. By contrast, superimposed intermittent neural and ventilator apnea does not elicit long-lasting changes in phrenic inspiratory motor output (n=7; p>0.05), indicating co-activation of iMF and LTF occludes plasticity. C) Representative phrenic neurograms from rats exposed to intermittent 6 sec cessations in ventilation (”mild hypoxia”) or intermittent neural apnea superimposed with 6 sec cessations in ventilation (“mild hypoxic neural apnea”). D) Average change in phrenic inspiratory burst amplitude 60 min following intermittent neural apnea, 6 sec ventilator apnea, the combination, or at equivalent points in time control rats subjected to the same surgical preparation, but no apnea. Intermittent 6 sec ventilator apnea neither elicit LTF (n=8; p>0.05) nor obstructed expression of iMF triggered by reduced respiratory neural activity (n=7; p<0.001). Data represent mean values ± SEM at 60min post challenge. *significantly different than surgical time controls and rats exposed to hypoxic neural apnea (p<0.05).

Superimposed neural and ventilator apneas were met with a rapid return of inspiratory neural activity. As such, the neural apnea duration was significantly less than rats that experienced neural apnea alone (82.4 ± 7.2sec vs 14.3 ± 0.6sec; p<0.001). To confirm that hypoxia’s constraint of iMF was not an artifact of the reduced apnea duration, we investigated a “mild” hypoxic experience consisting of 6sec ventilator apneas. First, we demonstrated that 5 episodes of 6sec ventilator apnea resulted in oxygen desaturation (87.5 ± 0.4% SaO₂; estimated PaO₂: 56 ± 0.3 mmHg) that was not sufficient to elicit LTF (−0.9 ± 2.8% baseline; p>0.05; n=8; Figure 1C and D). This is consistent with previous reports demonstrating moderate hypoxia (35–45mmHg PaO₂) is necessary to initiate the signaling processes for hypoxia-induced respiratory plasticity (Fuller et a., 2000). Further, while neural apnea with concurrent mild hypoxia (6sec; 87.5 ± 0.4% SaO₂) and moderate hypoxia (25sec; 71 ± 0.9% SaO₂) similarly reduced the duration of the neural apneas (13.7 ± 0.4sec vs 14.3 ± 0.6sec respectively; p>0.05), mild hypoxia did not disrupt expression of iMF since co-presentation of neural apnea with a 6sec cessation in ventilation elicited a compensatory enhancement of phrenic inspiratory output (59.2 ± 7.8% baseline; p<0.001; n=7) that was not significantly different than that observed with neural apnea alone (p>0.05; Figure 1C and D). Collectively, these data suggest that co-induction of iMF and LTF during hypoxic neural apneas impairs plasticity expression.

Hypoxia-induced NMDA receptor activation constrains phrenic inspiratory plasticity during hypoxic neural apneas

To determine if signaling mechanisms giving rise to LTF occlude iMF, we sought to selectively block LTF during hypoxic neural apneas. We focused on spinal NMDA receptors since iMF and LTF have been reported to have contrasting requirements for spinal NMDA receptor activation. Specifically, spinal NMDA receptor activation is necessary for LTF (McGuire et al., 2008; McGuire et al., 2005), whereas spinal NMDA receptor activation constrains at least some forms of iMF (Streeter and Baker-Herman, 2014b). To test the hypothesis that NMDA receptor activation occludes plasticity following recurrent hypoxic neural apneas, APV (10μL x 100μM), a selective NMDA receptor antagonist, was injected intrathecally over C3-C6 (cervical spinal region containing phrenic motor neurons) prior to initiation of the protocol. In surgical time control rats not experiencing apnea, intrathecal APV did not alter phrenic burst amplitude (7.3 ± 7.2% baseline; n=5; p>0.05), indicating that inhibition of basal NMDA receptor activation does not impact phrenic inspiratory output. Consistent with previous work, pretreatment with APV prior to intermittent 25 sec ventilator apnea abolished LTF (2.1 ± 3.1% baseline; n=8; p>0.05), a response that was significantly different than vehicle treated rats exposed to intermittent 25 sec ventilator apneas (p<0.001; Figure 2A and B). Conversely, intrathecal APV prior to recurrent neural apnea did not impair iMF expression (73.0 ± 7.6% baseline; n=8; p<0.001; Figure 2A and B), although there was a slight trend for an enhancement relative to vehicle treated recurrent neural apnea.

Figure 2: NMDA receptor activation by hypoxia constrains phrenic inspiratory plasticity.

A) Representative phrenic neurograms from rats pretreated with intrathecal APV prior to intermittent reductions in respiratory neural activity (“APV + neural apnea”), 25 sec ventilator apnea (“APV + hypoxia”), or intermittent neural apnea superimposed with 25 sec ventilator apnea (“APV + hypoxic neural apnea”). B) Average change in phrenic inspiratory burst amplitude 60 min following apnea in rats receiving intrathecal APV or vehicle over the phrenic motor pool prior to intermittent neural apnea, 25 sec ventilator apnea, the combination, or at an equivalent point in surgical time controls (not receiving apnea). In rats pretreated with intrathecal APV, intermittent neural apnea triggered significant increases in phrenic inspiratory output (n=8; p<0.001), indicating NMDA receptor activation is not necessary for iMF. No significant differences in phrenic inspiratory output were observed between APV and vehicle treated groups exposed to neural apnea (p>0.05). Rats pretreated with APV prior to intermittent 25 sec ventilator apnea did not express an enhancement in phrenic motor output (n=8; p>0.05), indicating that NMDA receptor activation is necessary for LTF. Pretreatment with APV restored the capacity for hypoxic neural apnea to elicit increased phrenic inspiratory output (n=7, p<0.001), indicating that hypoxia-induced activation of NMDA receptors occludes plasticity triggered by intermittent neural apnea. C) Average change in phrenic inspiratory burst amplitude 60 min following apnea in rats receiving intrathecal NR2B inhibitors (CO 101244 or eliprodil) or vehicle prior to intermittent neural apnea, 25 sec ventilator apnea, the combination (hypoxic neural apnea), or at an equivalent point in surgical time control rats. Pretreatment with selective NR2B inhibitors CO101244 (n=5; p<0.001) and eliprodil (n=5; p<0.001) enabled hypoxic neural apneas to elicit increased phrenic inspiratory output, which was significantly different than vehicle treated rats exposed to hypoxic neural apnea (p<0.001). There was no significant difference between CO101244 and eliprodil hypoxic neural apnea groups (p>0.05). Data represent mean values ± SEM at 60min post challenge. In B, (*) indicates significant differences from surgical time controls receiving APV and rats pretreated with APV prior to 25sec ventilator apneas (p<0.05). In C, (*) indicates significant differences from surgical time controls receiving Co 101244 or eliprodil, and rats treated with vehicle prior to hypoxic neural apnea (p<0.05).

While basal activity of spinal NMDA receptors does not appear to constrain iMF, hypoxia is known to enhance NMDA receptor signaling in the phrenic motor nucleus (McGuire et al., 2008). Therefore, we hypothesized that hypoxia-induced increases in spinal NMDA receptor activation conditionally constrains plasticity in phrenic inspiratory output following hypoxic neural apneas. To test this hypothesis, rats were pretreated with intrathecal APV prior to hypoxic neural apneas. Whereas vehicle treated rats did not show phrenic inspiratory plasticity in response to hypoxic neural apneas, plasticity was evident in rats pretreated with APV prior to hypoxic neural apneas (60.9 ± 8.0% baseline; n=7; p<0.001; Figure 2B and C), a response that was significantly different than that observed in vehicle-treated rats exposed to hypoxic neural apneas (p<0.001). Collectively, these data indicate that pharmacologically blocking LTF by impairing spinal NMDA receptor activation releases constraints on plasticity imposed by hypoxic neural apneas.

Although mechanisms by which hypoxia enhances spinal NMDA receptor signaling for LTF expression have not yet been determined, previous reports suggest that hypoxia results in phosphorylation of NMDA receptor NR2B subunits (Takagi et al., 2003), which has been shown to increase calcium influx through NMDA receptors (Paul and Connor, 2010; Takasu et al., 2002) and enable some forms of plasticity (Xu et al., 2006). To assess if NR2B-NMDA receptor activation obstructs compensatory respiratory plasticity during hypoxic neural apneas, CO101244 (10μL x 1.0mM) and eliprodil (10μL x 100μM), two selective inhibitors of NR2B containing NMDA receptors (Gill et al., 2002; Whittemore et al., 1997; Zhou et al., 1999), were delivered in subsets of rats prior to hypoxic neural apneas. In surgical time controls, neither CO101244 nor eliprodil alone (without neural or ventilator apnea) altered phrenic inspiratory output (−0.2 ± 2 and 2.7 ± 3.2% baseline, respectively; n=3, each; p>0.05; Figure 2C), indicating that inhibiting NR2B containing NMDA receptors does not alter basal phrenic neural activity. However, pretreatment with either intrathecal CO101244 or eliprodil enabled hypoxic neural apneas to trigger phrenic inspiratory plasticity (61.6 ± 5.01% and 63.6 ± 11.2% baseline, n=5 each, p<0.001), responses that were significantly different than vehicle treated rats exposed to hypoxic neural apnea (p<0.001; Figure 2C). Collectively, these data suggest that activation of NR2B containing NMDA receptors in the region of the phrenic motor pool impairs compensatory increases in phrenic inspiratory output in response to hypoxic neural apneas.

Exogenous retinoic acid rescues phrenic plasticity following hypoxic neural apneas

We next sought to determine if plasticity could be rescued during hypoxic neural apneas by pharmacologically activating pathways giving rise to iMF. Retinoic acid synthesis is necessary for several well-studied models of inactivity-induced plasticity (Aoto et al., 2008; Wang et al., 2011), including iMF (Baertsch and Baker, 2017a). To date, it is unknown whether retinoic acid synthesis is necessary for LTF. To confirm a mechanistic distinction between plasticity induced by reductions in respiratory neural activity (i.e., iMF) and hypoxia (i.e., LTF), DEAB (10μL x 1.0mM), a retinaldehyde dehydrogenase (RALDH) inhibitor, was injected intrathecally at C3-C6 prior to neural or 25sec ventilator apneas. Consistent with previous reports (Baertsch and Baker, 2017a), DEAB pretreatment impaired neural apnea-induced increases in phrenic inspiratory output (5.0 ± 3.7% baseline; p>0.05 relative to time controls; p<0.001 relative to vehicle treated neural apnea group; n=6; Figure 3A and B), indicating retinoic acid synthesis is necessary for iMF. Conversely, DEAB pretreatment did not attenuate increases in phrenic inspiratory output elicited by 25sec ventilator apneas (60.1 ± 7.9% baseline; p<0.001 relative to time controls; p>0.05 relative to vehicle treated 25sec ventilator apnea rats; n=7; Figure 3A and B), indicating that LTF does not require retinoic acid synthesis. Collectively, these data demonstrate that spinal retinoic acid synthesis is necessary for inactivity-induced, but not hypoxia-induced, respiratory plasticity.

Figure 3: Retinoic acid synthesis is necessary for plasticity triggered by intermittent neural apnea, but not intermittent hypoxia.

A) Representative phrenic neurograms from rats pretreated with intrathecal DEAB prior to intermittent reductions in respiratory neural activity (”DEAB + neural apnea”) or intermittent 25 sec ventilator apnea (“DEAB + hypoxia”). B) Average change in phrenic inspiratory burst amplitude 60 min following apnea in rats receiving intrathecal DEAB or vehicle over the phrenic motor pool prior to intermittent neural apnea, 25 sec ventilator apnea, the combination, or at an equivalent point in surgical time control rats. DEAB pretreatment abolished increases in phrenic inspiratory output triggered by intermittent reductions in respiratory neural activity (n=6; p>0.05), indicating retinoic acid synthesis is necessary for iMF. DEAB pretreated neural apnea group was statistically different from vehicle treated neural apnea group (n=8; p<0.001). Conversely, DEAB pretreatment did not affect hypoxia induced LTF (n=7; p<0.001), which was not significantly different than rats pretreated with vehicle (n=7; p>0.05). Data represent mean values ± SEM. *significantly different than surgical time controls receiving DEAB, and rats pretreated with DEAB prior to neural apnea (p<0.05).

Calcium inhibits RALDH to constrain inactivity-induced plasticity in the hippocampus (Wang et al., 2011). Since NR2B phosphorylation increases calcium influx through NMDA receptors (Strack and Colbran, 1998; Viviani et al., 2003), and NR2B-NMDA receptors constrain plasticity during hypoxic neural apneas (Figure 2C), hypoxia may undermine iMF by impairing calcium sensitive RALDH. Thus, we hypothesized that exogenous retinoic acid would circumvent RALDH inhibition and rescue plasticity induced by hypoxic neural apneas. Intrathecal injections of all-trans retinoic acid (10μL x 50μM) over the phrenic motor pool (without neural or ventilator apneas) elicited a slowly-progressing, modest increase in phrenic inspiratory burst amplitude (41.6 ± 5.9% baseline, n=7; p<0.001; Figure 4), suggesting that retinoic acid alone is sufficient to trigger plasticity. Consistent with our hypothesis, pretreatment with retinoic acid enabled hypoxic neural apneas to trigger phrenic inspiratory plasticity (81.3 ± 14.4% baseline; n=7; p<0.001), a response that was significantly different than vehicle treated rats exposed to hypoxic neural apnea (p<0.001; Figure 3A and B). Further, the magnitude of phrenic inspiratory plasticity observed in rats receiving retinoic acid pretreatment prior to hypoxic neural apneas was significantly greater than that triggered by retinoic acid treatment alone (p=0.02). Collectively, these data indicate that exogenous retinoic acid overcomes mechanistic constraints to enable hypoxic neural apneas to elicit respiratory plasticity.

Figure 4: Intrathecal retinoic acid rescues increased phrenic inspiratory output triggered by hypoxic neural apneas.

A) Representative phrenic neurograms from rats receiving intermittent neural apnea superimposed with 25 sec ventilator apnea (“hypoxic neural apnea”), intrathecal retinoic acid and no apnea, and intrathecal retinoic acid prior to hypoxic neural apneas. B) Average change in phrenic inspiratory burst amplitude 60 min following hypoxic neural apnea (or equivalent in surgical time controls) in rats pretreated with intrathecal retinoic acid or vehicle over the phrenic motor pool. Intrathecal retinoic acid in time controls not receiving apnea elicits increases in phrenic motor output (n=7; p<0.001). Whereas hypoxic neural apneas alone do not elicit increased phrenic inspiratory output, intrathecal retinoic acid prior to hypoxic neural apnea triggers increased phrenic inspiratory output (n=6; p<0.001), which was significantly greater than the response in retinoic acid-treated time controls (n=7; p=0.002). Data represent mean values ± SEM. *significantly different than rats pretreated with vehicle prior to hypoxic neural apneas (p<0.05).

Regulation of physiological variables

Table 1 lists average mean arterial pressure (MAP), PaCO₂, PaO₂ and breathing frequency at baseline and 60min after treatments in all groups. No consistent changes in blood gases or breathing frequency were detected across groups or time. However, two minor differences were observed: rats receiving eliprodil or retinoic acid prior to hypoxic neural apnea had a slightly higher PaO₂ at the end of the protocol than rats experiencing ventilator apnea alone. The differences were small (~10 mmHg) and not considered to be of physiological significance.

Table 1:

Physiological variables.

		Vehicle Control (n = 7)	Neural Apnea (n = 8)	25sec Ventilator Apnea (n = 7)	25sec Ventilator + Neural Apnea (n = 7)	6sec Ventilator Apnea (n = 8)	6sec Ventilator + Neural Apnea (n = 7)	DEAB Control (n = 5)	DEAB + Neural Apnea (n = 6)	DEAB + 25sec Ventilator Apnea (n = 7)
MAP (mmHg)	Baseline	116 ± 5	124 ± 5	113 ± 4	113 ± 4	112 ± 7	118 ± 5	109 ± 13	120 ± 8	89 ± 14
	Post 60min	90 ± 11	101 ± 7	103 ± 5	104 ± 6	101 ± 5	96 ± 3	103 ± 9	100 ± 7	82 ± 9
pO2 (mmHg)	Baseline	108.4 ± 2.6	107.7 ± 1.2	100.9 ± 1.7	104.3 ± 2.1	103.0 ± 2.2	105.8 ± 1.2	101.3 ± 2.9	105.2 ± 1.2	106.3 ± 1.5
	Post 60min	106.2 ± 2.7	105.4 ± 1.6	96.9 ± 1.9	103.9 ± 1.8	104.3 ± 2.0	106.5 ± 1.8	101.6 ± 3.8	104.1 ± 1.6	103.1 ± 1.5
pCO2 (mmHg)	Baseline	44.7 ± 1.6	47.2 ± 1.6	45.7 ± 1.4	47.7 ± 1.2	44.9 ± 1.2	47.4 ± 1.6	44.2 ± 0.6	45.3 ± 1.7	44.5 ± 2.0
	Post 60min	45.0 ± 1.6	47.7 ± 1.6	46.3 ± 1.5	47.7 ± 1.2	45.0 ± 1.1	47.7 ± 1.6	44.7 ± 0.8	44.6 ± 1.7	44.7 ± 2.0
Frequency (breaths/min)	Baseline	50 ± 1	48 ± 2	49 ± 1	50 ± 2	52 ± 3	48 ± 3	46 ± 2	51 ± 3	52 ± 3
	Post 60min	49 ± 1	48 ± 2	50 ± 2	48 ± 2	48 ± 2	50 ± 3	42 ± 2	51 ± 3	52 ± 3
		APV Control (n = 5)	APV + Neural Apnea (n = 8)	APV + 25sec Ventilator Apnea (n = 8)	APV + 25sec Ventilator + Neural Apnea (n = 7)	CO101244 Control (n = 3)	CO101244 + 25sec Ventilator + Neural Apnea (n = 5)	Eliprodil Control (n = 3)	Eliprodil + 25sec Ventilator + Neural Apnea (n = 5)	Retinoic Acid (n = 7)	Retinoic Acid + 25sec Ventilator + Neural Apnea (n = 6)
MAP (mmHg)	Baseline	117 ± 7	112 ± 4	103 ± 11	97 ± 12	108 ± 16	118 ± 11	129 ± 5	133 ± 7	108 ± 6	101 ± 12
	Post 60min	94 ± 6	87 ± 6	86 ± 8	89 ± 9	96 ± 10	121 ± 11	118 ± 6	125 ± 9	97 ± 4	86 ± 5
pO2 (mmHg)	Baseline	103.3 ± 2.4	100.6 ± 1.9	99.1 ± 2.6	100.0 ± 3.1	102.7 ± 1.2	104.7 ± 2.1	107.0 ± 1.7	106.9 ± 2.7	101.6 ± 2.4	108.4 ± 2.7
	Post 60min	100.9 ± 1.6	102.2 ± 2.2	101.5 ± 2.7	101.9 ± 0.9	103.7 ± 1.7	103.4 ± 2.2	106.7 ± 4.4	*108.9 ± 2.9	106.0 ± 2.7	*108.0 ± 2.4
pCO2 (mmHg)	Baseline	45.1 ± 1.9	46.2 ± 1.0	43.2 ± 1.0	48.0 ± 1.2	47.7 ± 0.4	47.8 ± 1.4	47.4 ± 1.3	46.8 ± 0.7	44.3 ± 1.0	47.4 ± 1.3
	Post 60min	45.1 ± 2.0	46.7 ± 1.0	43.2 ± 1.2	48.1 ± 1.2	47.6 ± 0.3	47.5 ± 1.2	47.2 ± 1.0	47.2 ± 0.9	44.2 ± 1.2	46.8 ± 1.4
Frequency (breaths/min)	Baseline	52 ± 3	50 ± 2	51 ± 2	49 ± 1	46 ± 4	48 ± 1	52 ± 3	52 ± 2	51 ± 2	52 ± 2
	Post 60min	47 ± 2	52 ± 2	46 ± 2	51 ± 2	42 ± 4	52 ± 3	51 ± 2	55 ± 1	53 ± 2	54 ± 1

No consistent differences in mean arterial pressure (MAP), PaCO₂, PaO₂, or phrenic burst frequency was observed in any individual group or among different groups. A small, but significant difference was observed in PaO2 at the end of the protocol in rats receiving eliprodil or retinoic acid prior to hypoxic neural apnea versus rats experiencing ventilator apnea alone.

^*significantly different than 25sec ventilator apnea, p<0.05. Values expressed as mean ± SEM.

In subsets of rats, the CO₂ level that triggered an apnea (so called “apneic threshold”) was measured. Both intermittent neural apnea and intermittent hypoxia have been reported to reduce the apneic threshold (Baertsch and Baker, 2017b; Mahamed and Mitchell, 2008), an effect hypothesized to decrease susceptibility to future apneas through signaling mechanisms similar to phrenic inspiratory plasticity (Braegelmann et al., 2017). Consistent with these reports, both recurrent neural apnea and recurrent 25sec ventilator apneas elicited a significant reduction in the apneic threshold (−3.2 ± 0.5mmHg CO₂ and −4.0 ± 0.7mmHg CO₂, respectively; n=4, each; p<0.01; Figure 5A). By contrast, intermittent hypoxic neural apneas did not lead to a reduction in the apneic threshold (−0.2±0.4mmHg CO₂; n=5; p>0.05; Figure 5A), consistent with our observations that hypoxic neural apneas also do not result in compensatory enhancements in phrenic inspiratory output. Strikingly, spinal retinoic acid pretreatment revealed the ability for hypoxic neural apneas to elicit a significant reduction in the apneic threshold (−7.8 ± 1.9mmHg PaCO₂; n=6; p<0.01; Figure 5B); this response was significantly different than that observed in vehicle treated rats exposed to hypoxic neural apneas or retinoic acid treated time controls (p<0.01). Thus, exogenous spinal retinoic acid bypasses signaling constraints to enable a robust apneic threshold reduction following hypoxic neural apneas.

Figure 5: Intrathecal retinoic acid rescues apneic threshold decreases triggered by hypoxic neural apneas.

A) Both reduced respiratory neural activity (“neural apnea”; n=4; p<0.01) and 25 sec ventilator apneas (“hypoxia”; n=4; p<0.01) were independently sufficient to reduce the CO₂ threshold that triggers an apnea. By contrast, concomitant reductions in respiratory neural activity with breathing cessation (“hypoxic neural apnea”) did not result in a lowering of the apneic threshold (n=5; p>0.05). B) Intrathecal retinoic acid alone does not modulate apneic threshold (n=5; p>0.05); however, intrathecal retinoic acid pretreatment prior to hypoxic neural apneas resulted in a lowering of the apneic threshold (n=6; p<0.01). Data represent mean values ± SEM. In A, (*) indicates significant differences from surgical time controls and rats exposed to hypoxic neural apnea (p<0.05). In B, (*) indicates significant differences from rats receiving retinoic acid (and no apnea) and rats pretreated with vehicle prior to hypoxic neural apnea (p<0.05).

Discussion

Regular, rhythmic breathing is a necessity for life, and even brief pauses in breathing can be profoundly detrimental. Rhythmic firing of the neural network controlling breathing is supported by intrinsic mechanisms of plasticity that respond to reductions in respiratory neural activity by proportionally enhancing inspiratory motor output (i.e., iMF) and lowering the threshold for apnea (Baertsch and Baker, 2017b; Braegelmann et al., 2017; Strey et al., 2013). Similarly, intermittent hypoxia, such as would occur during cessations in breathing, elicits distinct mechanisms of plasticity that also strengthen inspiratory motor output (i.e., LTF) and lower the threshold for apnea (Katayama et al., 2007; Mahamed and Mitchell, 2008). Despite these safeguards, breathing disorders exist. Previous investigations have advanced our understanding of the signaling processes enabling compensatory respiratory plasticity, but have not mimicked the hallmark pattern of reductions in respiratory neural activity with hypoxia that characterizes many breathing disorders. An incomplete understanding of cellular events obstructing plasticity during clinically relevant breathing dysfunction has hindered development of targeted pharmacological therapies.

Here, we show that co-induction of iMF and LTF, such as would be expected to occur in forms of sleep apnea characterized by reduced inspiratory neural output, occludes compensatory plasticity during hypoxic neural apneas by cross-talk inhibition of their respective signaling cascades. These data are consistent with findings suggesting that concomitant mild hypercapnia is necessary for LTF in humans (Griffin et al., 2012; Harris et al., 2006), presumably to prevent arterial CO₂ from falling below the apneic threshold (and hence, reducing respiratory neural activity)(Mateika et al., 2018). In our studies, we chose to use a “ventilator apnea” (i.e., temporarily cease mechanical ventilation) to induce LTF rather than deliver hypoxia in the inspired gas mix for several reasons: 1) to better simulate apneic respiratory mechanics (cessation of lung and chest wall expansion), 2) to better simulate respiratory gas exchange (stagnant air in lungs), and 3) to induce a rapid hypoxemia followed by a rapid re-oxygenation, clinically relevant oxygen dynamics. Careful monitoring of expired CO₂ ensured ventilator apneas did not cause significant hypercapnia (only ~2–5 mmHg ETCO₂ above baseline). Previous work has demonstrated that mild hypercapnia does not elicit or occlude respiratory plasticity (Bach and Mitchell, 1996; Baker et al., 2001; Mahamed and Mitchell, 2008; Mateika et al., 2018); thus, we conclude that antagonistic signaling convergence from co-induction of iMF and LTF by reduced respiratory neural activity and hypoxia, respectively, obstructs plasticity during hypoxic neural apneas, potentially explaining the absence of adaptive breathing behaviors in some individuals with sleep apnea.

Reduced respiratory neural activity and hypoxia elicit distinct mechanisms of plasticity

We’ve recently come to appreciate that the respiratory control system monitors respiratory neural activity in inspiratory motor neurons driving the respiratory muscles--even in the absence of a change in blood gases—and when respiratory neural activity falls below a target level, mechanisms local to the inspiratory motor pool initiate a proportional enhancement in inspiratory motor output (Streeter and Baker-Herman, 2014a), a phenomenon known as inactivity-induced inspiratory motor facilitation (iMF). Importantly, complete “inactivity” need not be achieved to activate iMF since a “neuronal hypopnea” (i.e., reduced, but not absent respiratory neural activity) (Braegelmann et al., 2017) or a local, partial blockade of synaptic inputs to the motor neuron (Streeter and Baker-Herman, 2014a) are sufficient to elicit iMF. Although the precise role for iMF in the control of breathing is unknown, we have recently proposed that iMF represents one of possibly many components in a continuum of “homeostatic plasticity” mechanisms (Turrigiano, 2008) that assure appropriate respiratory motor output throughout life (Braegelmann et al., 2017; Strey et al., 2013).

Although much of our knowledge regarding mechanisms and regulatory constraints of iMF have been investigated in phrenic motor neurons, iMF is also robustly expressed in other inspiratory motor pools, including the hypoglossal (Baertsch and Baker-Herman, 2013) and inspiratory intercostals (Strey et al., 2013). Different induction mechanisms are activated to trigger compensatory enhancements in inspiratory motor output depending on the pattern of reduced respiratory neural activity: intermittent reductions in respiratory neural activity trigger retinoic acid synthesis (Baertsch and Baker, 2017a), whereas prolonged reductions in respiratory neural activity trigger TNFα release from, presumably, neighboring glia (Broytman et al., 2013). Ultimately, both the retinoic acid- and TNFα-dependent pathways converge on atypical PKC, and lead to the association of an atypical PKC isoform (PKCζ or PKCι) with the scaffolding molecule p62/SQSTM1 to stabilize iMF (Baertsch and Baker-Herman, 2015; Strey et al., 2012). NMDA receptor activation constrains this stabilization of iMF through unknown mechanism (Streeter and Baker-Herman, 2014b).

Considerable evidence suggests that mechanisms distinct from iMF regulate expression of LTF (Strey et al., 2012). The working model suggests that moderate intermittent hypoxia induces the release of serotonin in the phrenic motor pool (Baker-Herman and Mitchell, 2002), which activates 5-HT2 receptors on phrenic dendrites (Fuller et al., 2001); necessary downstream signaling includes ERK MAP kinase signaling (Hoffman et al., 2012), de novo synthesis of brain-derived neurotrophic factor (BDNF) (Baker-Herman et al., 2004), activation of the high-affinity BDNF receptor tropomyosin-related kinase B (TrkB)(Dale et al., 2017), activation of protein kinase C theta (Devinney et al., 2015) and reactive oxygen species formation (MacFarlane and Mitchell, 2008; MacFarlane et al., 2009; MacFarlane et al., 2008). In contrast to iMF, LTF does not require atypical PKC activation (Strey et al., 2012) or retinoic acid synthesis (Figure 3), and requires NMDA receptor activation (Figure 2) for both induction and maintenance (McGuire et al., 2008; McGuire et al., 2005). Similar to iMF, LTF is expressed in multiple inspiratory motor pools, including the hypoglossal (Bach and Mitchell, 1996; Baker-Herman and Strey, 2011; Wilkerson et al., 2017) and inspiratory intercostals (Fregosi and Mitchell, 1994; Lovett-Barr et al., 2006; Navarrete-Opazo et al., 2014).

Although we are making headway in understanding mechanisms by which reduced respiratory neural activity and hypoxia elicit enhanced phrenic inspiratory output, mechanisms underlying the reduction in the apneic threshold are unknown (Baertsch and Baker, 2017b). The traditional view is that the apneic threshold is determined by peripheral and central chemoreceptor interactions with brainstem rhythm generating neurons. However, available evidence indicates that the amount that the apneic threshold is changed following reduced respiratory neural activity is proportional to iMF magnitude (Baertsch and Baker, 2017b); a striking observation since mechanisms local to the inspiratory motor pool give rise to iMF (Streeter and Baker-Herman, 2014a; Strey et al., 2012). There are several possible explanations for how changes in the apneic threshold and iMF/LTF may be linked: 1) Multiple forms of plasticity, both spinally and supra-spinally, may be elicited in response to reduced respiratory neural activity and intermittent hypoxia. Thus, it is possible that mechanisms occurring elsewhere in the respiratory circuit (e.g., respiratory rhythm generator) that communicate with the inspiratory motor pools give rise to plasticity in the apneic threshold. 2) Reductions in the apneic threshold may be the result of iMF/LTF. Indeed, it is possible that some brainstem inspiratory neurons remain rhythmically active near the apneic threshold, but fail to transmit this activity to inspiratory motor pools (Batsel, 1967; Ezure et al., 2003; Garcia et al., 2016; Kam et al., 2013; St. John, 1998); plasticity within the motor neuron may augment inspiratory motor neuron excitability sufficiently to enable expression of this otherwise sub-threshold descending drive. Indeed, the apneic threshold for inspiratory activity is not necessarily the same in all respiratory motor pools (Fregosi and Mitchell, 1994). Finally, 3) Plasticity in inspiratory motor neurons could induce medullary plasticity indirectly via spinal interneurons with spinobulbar projections (Fuller et al. 2013, Golder et al., 2001; Lane et al., 2009).

Our data suggest that cross-talk inhibition between distinct signaling cascades activated by reduced respiratory neural activity and hypoxia occludes plasticity when these stimuli are presented concurrently. Although we do not yet know mechanisms by which reduced respiratory neural activity occludes expression of hypoxia-induced plasticity (i.e., iMF occlusion of LTF), we identify a key role for hypoxia-induced NMDA receptor activation, or it’s downstream signaling cascade, in the occlusion of plasticity elicited by reduced respiratory neural activity (i.e., LTF occlusion of iMF). These data are consistent with reports suggesting that NMDA receptor activation constrains some forms of rapid homeostatic plasticity induced by reductions in neural activity in the hippocampus (Chen et al., 2014; Gonzalez-Islas et al., 2018; Sutton et al., 2006). Specifically, we show that moderate (but not mild) hypoxia undermines iMF by enhancing NR2B containing NMDA receptor signaling in the phrenic motor pool (Figure 2). We hypothesize that hypoxia-enhanced NR2B-NMDA receptors constrain iMF by inhibiting calcium sensitive RALDH (Wang et al., 2011), thereby inhibiting the synthesis of retinoic acid. Consistent with this hypothesis, the capacity for hypoxic neural apneas to trigger iMF can be rescued by either inhibiting spinal NR2B-NMDA receptors or bypassing NMDA receptor-induced constraints with exogenous retinoic acid delivered locally to the phrenic motor pool (Figures 2 and 4). Intriguingly, although intrathecal retinoic acid alone was sufficient to elicit (modest) increases in phrenic inspiratory output in rats that did not experience ventilator or neural apnea, it was not sufficient to elicit decreases in the apneic threshold (Figure 5). Nevertheless, intrathecal retinoic acid rescued the ability for hypoxic neural apneas to lower the apneic threshold. Mechanisms whereby intrathecal retinoic acid conditionally gates apneic threshold changes associated with hypoxic neural apneas are unknown.

Clinical implications for sleep apnea

While aberrant firing of neurons in the brainstem that generate respiratory rhythm and pattern is thought to be the primary cause of many breathing disorders characterized by insufficient respiratory motor output (McKay and Feldman, 2008), poor neural transmission of brainstem inspiratory signals to motor neurons driving the inspiratory muscles is emerging as a potentially important contributor (Garcia et al., 2016; Kam et al., 2013; Ramirez and Baertsch, 2018). For example, prolonged exposure to intermittent hypoxia reduces transmission fidelity of brainstem-generated inspiratory signals to respiratory motor nuclei (Garcia et al., 2016). Such inspiratory transmission failure from the brainstem to motor neurons controlling upper airway caliber or spinal motor neurons innervating respiratory pump muscles could lead to OSA and CSA, respectively (Garcia et al., 2016; Kam et al., 2013; Wheatley et al., 1993). Therapies improving transmission fidelity of the respiratory control network (i.e. respiratory plasticity) may be one method for treating some forms of OSA, CSA, and related breathing disorders.

Both iMF and LTF strengthen inspiratory motor output at the level of the inspiratory motor neuron, including phrenic, hypoglossal and inspiratory intercostal motor neurons, and have been proposed to stabilize breathing and prevent future apneas/hypopneas by enhancing inspiratory motor output (Braegelmann et al., 2017; Mahamed and Mitchell, 2007; Mahamed and Mitchell, 2008; Mateika and Komnenov, 2017; Mateika et al., 2004; Mateika and Syed, 2013; Strey et al., 2013) and lowering the threshold for transmission failure (Baertsch and Baker, 2017b; Braegelmann et al., 2017). Individuals with sleep apnea are capable of expressing LTF following intermittent hypoxia (without concomitant reductions in respiratory neural activity)(Aboubakr et al., 2001; Lee et al., 2009), and reports suggest that intermittent hypoxia therapy can improve upper airway patency in individuals with sleep apnea (El-Chami et al., 2017), presumably by LTF-like mechanisms. One central question, then, is: why don’t individuals with sleep apnea self-correct apneas/hypopneas due to endogenous activation of iMF or LTF mechanisms? Indeed, a number of studies suggest that apnea/hypopnea severity actually worsens during the course of the night, despite repeated self-dosing with reduced respiratory neural activity (Javaheri et al., 2018) and/or intermittent hypoxia (Cala et al., 1996; Charbonneau et al., 1994; Lavie et al., 1981; Oksenberg et al., 2001; Sforza et al., 1998). Although OSA is the most common form of sleep apnea, OSA and CSA often co-exist in the same individual and reductions in upper airway motor neuron activity is frequently a precipitating factor for OSA (Dempsey et al., 2014; Hoffman and Schulman, 2012; Westhoff et al., 2012), suggesting that many individuals with sleep apnea experience concomitant reductions in inspiratory motor neuron activity and hypoxemia. While there are likely many possible explanations for a lack of compensatory plasticity in individuals with sleep apnea (Mateika and Komnenov, 2017; Mateika et al., 2018), our findings suggest that in some individuals, concurrent induction of iMF and LTF may occlude compensatory respiratory behaviors via cross talk inhibition of their distinct signaling pathways. Consistent with this hypothesis, evidence suggests that PaCO₂ has to be increased by a small amount (3–4 mmHg) during intermittent hypoxic exposures in order to reliably elicit LTF in humans (El-Chami et al., 2017; Harris et al., 2006; Lee et al., 2009; Mateika et al., 2018). The reasons for this are not well understood; however, it is possible that mild hypercapnia prevents hypoxia-induced hyperventilation from lowering PaCO₂ below/near the apneic threshold, thereby creating a central apnea/hypopnea (Mateika et al., 2018) and thus, occluding plasticity by the concurrent exposure to both LTF and iMF-inducing stimuli.

These experimental observations are of central significance in understanding several important clinical observations. First, individuals with mixed (obstructive and central) sleep apnea receiving CPAP or tracheostomy to alleviate upper airway obstruction will continue to have central apneas initially upon treatment. However, in many of these individuals, central events will spontaneously resolve through mechanisms that are not understood (Arzt et al., 2009; Deacon and Catcheside, 2015; Salloum et al., 2010). Secondly, many individuals with CSA experience a reduction in central events following treatment with inspired oxygen (Chowdhuri et al., 2012), suggesting hypoxia as a potential precipitator of CSA. Finally, periodic breathing is common in full-term and premature infants who experience transient episodes of hypoxic neural apneas. While some infants receive supplemental oxygen, most infants with mild hypoxic events naturally correct their breathing pattern within 6 months of life without intervention (Kelly et al., 1985). Collectively, these data support the existence of an endogenous form of compensatory plasticity that “self-corrects” low inspiratory motor output, and suggest that it may be hindered by concurrent moderate hypoxia (35–45 mmHg PaO₂). When hypoxia is resolved through medical intervention (or is mild in intensity) these endogenous mechanisms emerge to restore breathing stability.

Two studies have investigated the efficacy of NMDA receptor antagonists in treating sleep apnea (Hedner et al., 1996; Torvaldsson et al., 2005). Hedner and colleagues reported a dose-dependent reduction in oxygen desaturation in individuals with OSA following a two-week treatment with an NMDA receptor antagonist, suggesting an improvement in apnea occurrence or duration. Similarly, while Torvaldsson and colleagues reported that a single dose of an NMDA receptor antagonist did not improve average AHI in 8 individuals with OSA, a trend for a reduction in the average central apnea index was observed (p=0.105). Together these studies provide early evidence of the potential for NMDA receptor antagonists in treating at least some forms of sleep apnea; however, it’s therapeutic usefulness is likely to be limited by psychotropic side effects that are characteristic of NMDA receptor modulation (Torvaldsson et al., 2005). Thus, alternative therapies that bypass NMDA receptor constraint to plasticity, such as retinoic acid, may ultimately be of more benefit.

Although Hedner and Torvaldsson hypothesized that hypoxia-induced release of glutamate in the CNS during apneic episodes led to repetitive airway collapse via ventilatory oscillations caused by recurrent hyper-activation followed by hypo-activation of brainstem respiratory neurons (via chemoreflex feedback), we suggest an alternative hypothesis. Specifically, we suggest that in some cases, breathing instability (CSA, OSA and otherwise) may arise from obstruction of innate compensatory plasticity mechanisms. Indeed, data presented here suggest the intriguing possibility that in clinically relevant apneas, co-induction of distinct forms of plasticity induced by reductions in respiratory neural activity (i.e., iMF) and hypoxemia (i.e., LTF) may occlude adaptive neural responses to recurrent apnea. By identifying a mechanistic phenomenon, our novel hypothesis readily identifies discrete alternative therapeutic targets to pharmacologically stabilize pathologic breathing patterns. This study provides important mechanistic insight into understanding how clinically relevant apneas may undermine compensatory forms of respiratory plasticity, and indicates that retinoic acid may be a novel pharmacological approach to stabilize breathing in individuals with sleep apnea and related breathing disorders.

Key points.

• Intermittent reductions in respiratory neural activity, a characteristic of many ventilatory disorders, leads to inadequate ventilation and arterial hypoxia. Both intermittent reductions in respiratory neural activity and intermittent hypoxia trigger compensatory enhancements in inspiratory output when experienced separately, forms of plasticity called inactivity-induced inspiratory motor facilitation (iMF) and long-term facilitation (LTF), respectively.

• Reductions in respiratory neural activity that lead to moderate, but not mild, arterial hypoxia occludes plasticity expression, indicating that concurrent induction of iMF and LTF impairs plasticity through cross-talk inhibition of their respective signaling pathways.

• Moderate hypoxia undermines iMF by enhancing NR2B-containing NMDA receptor signaling, which can be rescued by exogenous retinoic acid, a molecule necessary for iMF.

• These data suggest that in ventilatory disorders characterized by reduced inspiratory motor output, such as sleep apnea, endogenous mechanisms of compensatory plasticity may be impaired, and that exogenously activating respiratory plasticity may be a novel strategy to improve breathing.