Intermittent hypercapnic hypoxia during sleep does not induce ventilatory long-term facilitation in healthy males

Transcription

1 J Appl Physiol 123: , First published June 15, 2017; doi: /japplphysiol RESEARCH ARTICLE Intermittent hypercapnic hypoxia during sleep does not induce ventilatory long-term facilitation in healthy males Naomi L. Deacon, 1,2 R. Doug McEvoy, 1,2,3 Daniel L. Stadler, 2 and X Peter G. Catcheside 1,2,3 1 Discipline of Physiology, School of Medical Sciences, University of Adelaide, Adelaide, South Australia, Australia; 2 Adelaide Institute for Sleep Health: A Flinders Centre of Research Excellence, Repatriation General Hospital, Daw Park, South Australia, Australia; and 3 School of Medicine, Flinders University, Bedford Park, South Australia, Australia Submitted 8 November 2016; accepted in final form 13 June 2017 Deacon NL, McEvoy RD, Stadler DL, Catcheside PG. Intermittent hypercapnic hypoxia during sleep does not induce ventilatory long-term facilitation in healthy males. J Appl Physiol 123: , First published June 15, 2017; doi: /japplphysiol Intermittent hypoxia-induced ventilatory neuroplasticity is likely important in obstructive sleep apnea pathophysiology. Although concomitant CO 2 levels and arousal state critically influence neuroplastic effects of intermittent hypoxia, no studies have investigated intermittent hypercapnic hypoxia effects during sleep in humans. Thus the purpose of this study was to investigate if intermittent hypercapnic hypoxia during sleep induces neuroplasticity (ventilatory long-term facilitation and increased chemoreflex responsiveness) in humans. Twelve healthy males were exposed to intermittent hypercapnic hypoxia (24 30 s episodes of 3% CO 2 and % O 2 ) and intermittent medical air during sleep after 2 wk washout period in a randomized crossover study design. Minute ventilation, end-tidal CO 2, O 2 saturation, breath timing, upper airway resistance, and genioglossal and diaphragm electromyograms were examined during 10 min of stable stage 2 sleep preceding gas exposure, during gas and intervening room air periods, and throughout 1hofroom air recovery. There were no significant differences between conditions across time to indicate long-term facilitation of ventilation, genioglossal or diaphragm electromyogram activity, and no change in ventilatory response from the first to last gas exposure to suggest any change in chemoreflex responsiveness. These findings contrast with previous intermittent hypoxia studies without intermittent hypercapnia and suggest that the more relevant gas disturbance stimulus of concomitant intermittent hypercapnia frequently occurring in sleep apnea influences acute neuroplastic effects of intermittent hypoxia. These findings highlight the need for further studies of intermittent hypercapnic hypoxia during sleep to clarify the role of ventilatory neuroplasticity in the pathophysiology of sleep apnea. NEW & NOTEWORTHY Both arousal state and concomitant CO 2 levels are known modulators of the effects of intermittent hypoxia on ventilatory neuroplasticity. This is the first study to investigate the effects of combined intermittent hypercapnic hypoxia during sleep in humans. The lack of neuroplastic effects suggests a need for further studies more closely replicating obstructive sleep apnea to determine the pathophysiological relevance of intermittent hypoxia-induced ventilatory neuroplasticity. intermittent hypoxia; long-term facilitation; sleep apnea; neuroplasticity; ventilation Address for reprint requests and other correspondence: N. L. Deacon, 9300 Campus Point Dr., no. 7381, La Jolla, CA ( OBSTRUCTIVE SLEEP APNEA (OSA) is characterized by repeated episodes of upper airway obstruction causing concomitant hypercapnia and hypoxia. Experimental intermittent hypoxia (IH) can induce neuroplastic changes in ventilatory neural control (14, 37) suggesting that IH induced ventilatory neuroplasticity could play a role in OSA pathophysiology. However, both arousal state and the level of concomitant CO 2 utilized during IH are also important determinants of whether neuroplasticity is induced (14, 18, 31, 35). Very few studies have examined the effects of combined intermittent hypercapnic hypoxia (IHH) on ventilatory neuroplasticity in animals or humans (9, 48), and none during sleep. Thus further studies are needed to determine if IHH, comparable to that occurring in OSA, induces ventilatory neuroplasticity during sleep in humans. Long-term facilitation (LTF) is a form of neuroplasticity which causes a sustained increase in baseline neural activity following an inducing stimulus (29). Acute IH (AIH; ranging from 3 episodes to an entire day) has been shown to induce LTF in ventilatory motor neurons in several animal species (6, 22, 30, 35, 47). In rats phrenic LTF (pltf) manifests as an increase in minute ventilation (V I) (24) and hypoglossal LTF (hltf) as an increase in genioglossal electromyogram (EMG gg ) phasic burst amplitude and reduced upper airway resistance (R UA ) (7). In intact animals these phenomena are referred to as ventilatory LTF (vltf) and upper airway LTF (UALTF), respectively. In rats, chronic IH (CIH; 3 days of 8hofIH)induces an increase in carotid body hypoxic response (36). In humans, vltf has been elicited during wakefulness and sleep (8, 14), and in healthy individuals and OSA patients (19). Genioglossal LTF has also been elicited during wakefulness and sleep in healthy participants (7, 14). Unlike animals that require CIH, humans experience an increase in the hypoxic ventilatory response following acute IH during wakefulness (19), although this has not been reported following AIH during sleep. Results of IH studies in humans have been highly variable and have often failed to induce neuroplasticity (1a, 9, 16, 23), possibly due to differences in experimental protocols. Animal studies show that factors such as intensity of hypoxia (34), duration, and frequency/pattern of hypoxic episodes (28, 48, 49), level of concomitant CO 2 (18, 35), and arousal from sleep (25) are all important determinants of LTF induction. Although IH protocols delivered on a background of sustained CO 2 levels [isocapnic during sleep and hypercapnic during wake /17 Copyright 2017 the American Physiological Society

2 fulness (14, 38)] have successfully induced vltf and increased hypoxic sensitivity in humans, this does not mimic the gas disturbances experienced in OSA. Consequently the relevance of these findings to OSA pathophysiology is uncertain. Only one previous study in humans appears to have mimicked the hypercapnic hypoxic blood gas disturbances that occur during airway obstruction in OSA and this study failed to exhibit neuroplasticity (9). However, responses were only examined during wakefulness where wake ventilatory drive inputs could mask ventilatory chemoreflex responses which dominate ventilatory control during sleep. Therefore, although IH appears to be capable of inducing vltf and increased hypoxic sensitivity in humans, the aim of this study was to examine if a protocol designed to mimic the blood gas perturbations experienced in moderate to severe OSA would induce ventilatory neuroplasticity. We hypothesized that IHH would induce vltf and increase chemosensitivity during sleep in healthy males. IHH during Sleep Does Not Induce LTF Deacon NL et al. 535 MATERIALS AND METHODS Ethical Approval Twelve healthy male volunteers between the ages of yr gave written informed consent and participated in the study, which was approved by the University of Adelaide and the Southern Adelaide Clinical Human Research and Ethics Committees. Participants All participants were nonsmokers, demonstrated normal lung function [ 80% predicted for both forced expiratory volume in the first second (FEV 1) and forced vital capacity (FVC); JLab software version 4.53; Compactlab, Jäger, Würzburg, Germany], were not born or previously lived at high altitude, took no regular medications, were self-reported nonsnorers, and had no history of any other symptom suggestive of a sleep disorder. Participants were asked to abstain from alcohol and caffeine for the whole day and to avoid strenuous exercise during the latter half of the day before each study. Participants were also asked to eat their evening meal at the same time of night and a minimum of 4 h before commencement of each study night. Study Design This protocol comprised three study nights, each separated by a minimum of 2 wk (Fig. 1). During the first 3honthefirst night, participants underwent a standard polysomnography study to rule out those with sleep disorders and those with difficulty sleeping in the laboratory environment. The second half of the first night was used to determine the level of O 2 required during a 30 s exposure to achieve arterial desaturations to 80 85%. The base gas mixture contained 3% CO 2 (determination of O 2 content detailed in Preliminary night), based on findings reported by Younes et al. (50) that a 30-s exposure to hypoxia (15% or 11% O 2) with 3% CO 2 was the highest CO 2 content that all participants were able to sleep through which also achieved both hypercapnia and hypoxia. This subject-specific experimental gas composition was used in subsequent experiments. The 2nd and 3rd experimental nights comprised, in random order 2 wk apart, IHH gas exposure on one night, and intermittent medical air (control) on another night with participant blinding to gas condition. If there was insufficient sleep to complete the protocol on any night, the subject returned a minimum of 2 wk later for a fourth night to repeat the protocol. Participants slept in a sound-insulated room and were studied by two research staff from an adjacent room via continuous monitoring of respiration, sleep, and a video display from infrared camera overlying the bed. Fig. 1. Protocol flow diagram. Equipment and Measurements Sleep was measured by two channels of electroencephalogram (C4/M1, C3/M2), left and right electro-occulogram, submental electromyogram (EMG) and electrocardiogram. Respiration was monitored via measurement of nasal pressure (PTAF2, Pro-Tech Services, Woodinville, WA), oronasal thermistor, arterial oxyhemoglobin saturation via ear pulse oximetry (POET II model 602-3; Criticare Systems, Waukesha, WI), and chest and abdominal wall motion (piezo-electric bands). A tracheal microphone was used to monitor snoring. Body position and bilateral anterior tibialis electromyograms were also continuously recorded. Data were acquired on a Compumedics data acquisition system (E-series, Compumedics, Melbourne, Australia). For the second half of the preliminary night, all equipment from the PSG remained in place except the nasal cannula and thermistor were removed, the participant s mouth was taped to prevent mouth breathing, and a nasal mask (Gel mask, Respironics, Murrysville, PA) was fitted with a two-way non-rebreathing valve attached (series 2600, Hans Rudolph, Kansas City, MO). CO 2 was continuously sampled at the mask (Capstar-100, CWE). A pneumotachograph (PT16, Jaeger, Germany) connected on the inspiratory side of the breathing valve measured flow immediately downstream from a Gatlin-shaped valve system (series 2440C, Hans Rudolph) for delivery of inspiratory gases. The Gatlin-shaped valve common outlet port was attached to the pneumotachograph. Of four inlet ports, three were open to room air while the fourth was connected to a foil bag (40 L, Scholle

3 536 IHH during Sleep Does Not Induce LTF Deacon NL et al. Industries, Adelaide, Australia) containing a hypercapnic hypoxic (HH) gas mixture. Each port could be rapidly occluded or opened with a solenoid and pneumatically driven balloon valve. Only one inlet port was open at a time and all changes between ports were conducted during expiration. On the main experimental nights, participants were instrumented as above but using a 300-liter foil bag in place of a 40-liter bag used on the preliminary night. Genioglossal EMG (EMG gg) was also recorded using two intramuscular electrodes (316SS3T wire, Medwire, Mt. Vernon, NY) inserted 4 mm either side of the frenulum to a depth of cm after surface anesthesia with 4% lignocaine (11). Surface diaphragmatic EMG (EMG dia) was also recorded from a pair of surface EMG electrodes placed in the right sixth, seventh, or eighth intercostal spaces adjacent to the costal margin from the electrode pair that produced the highest inspiratory phasic activity. Epiglottic pressure (P epi) was measured with a thin air-perfused nasal catheter (see 15 for further detail) inserted through the most patent nostril following nasal decongestion (xylometazoline hydrochloride nasal spray, Novartis Australasia, Rowville, Victoria, Australia) and anesthesia (2% lignocaine spray), then advanced 1 cm below the tongue base under direct visualization, taped at the nose, and connected to a pressure transducer (MP45; Validyne Engineering, Northridge, CA). Mask pressure (P mask) and end-tidal CO 2 (PET CO2 ) were measured continually from the mask (P mask; MP45, Validyne Engineering, Northridge, CA, PET CO2 ; Capstar-100, CWE). Pneumotachograph airflow, EMG gg, EMG dia, P epi, P mask, and CO 2 signals beyond conventional channels recorded on the clinical sleep acquisition system (Compumedics) were recorded on a separate 32- channel Windaq (DI-720 DATAQ Instruments) acquisition system at sampling rates of 1 khz for EMG and time matching signals and 200 Hz for all other channels. Time matching between recording systems was achieved through the use of a computer generated timing signal simultaneously recorded on both systems and sampled at the highest rate available on the Compumedics system (512 Hz). Protocol Preliminary night. Recordings began between 2200 and 0000 and finished a minimum of 3 h later after at least one complete sleep cycle (Fig. 1). Participants were instructed to sleep in the supine position as much as possible. All studies were scored by one of two technicians using arousal and respiratory event scoring according to standard criteria (Chicago) (1), and using a total sleep apnea-hypopnoea index (AHI) cut-off of 15 events/h for the diagnosis of OSA (39). Participants showing evidence of sleep or breathing disorders were excluded from further participation in the study and referred for separate clinical follow-up. During the second half of the preliminary night participants slept in the lateral position to help achieve a more stable airway to prevent flow limitation upon gas delivery necessary to achieve consistent blood gas disturbances. Participants chose either their left or right side and remained in that position for the duration of the protocol. A 40-liter foil bag was filled with a HH gas mixture starting with 6% O 2, 3% CO 2, and the balance N 2. During stable stage 2 sleep the breathing circuit was switched from room air to the bag for 30 s (Fig. 1). The bag was then refilled with progressively lower O 2 content until consistent oxygen desaturations of 80 85% were achieved. This same gas mixture was then used for the subject s subsequent experimental study night. Main protocol. Participants slept in the same lateral position as for the second half of the preliminary night. Following 10 min of accumulated stage 2 sleep, participants were exposed to twenty-four 30-s episodes of medical air or the subject-specific HH gas determined from the preliminary night. Each gas exposure was initiated during sleep and separated by 2 min breathing room air (Fig. 1). If the subject aroused at any stage during HH or medical air gas exposure, the subject was immediately returned to room air breathing and the HH delivery protocol was paused until a minimum of 2 min stable sleep had resumed. The initial episode was always given in stage 2 sleep; however thereafter the protocol was continued irrespective of sleep stage. Recording continued while participants breathed room air for 60 min of recovery following the last HH episode. Data Analysis The inspiratory flow signal (pneumotachograph) was digitally integrated using custom software used previously (10) to derive breathby-breath inspiratory tidal volume (V T), minute ventilation (V I), breath timing, inspiratory (T I), expiratory (T E), total breath (T TOT) times, breathing frequency (F B), peak inspiratory flow (PIF), epiglottic pressure (P epi) and upper airway resistance (R UA). Breath-bybreath measurements of expiratory (end-tidal) partial pressure of CO 2 (PET CO2 ) were determined from the nadir and peak in mask CO 2 after adjusting for gas sampling delay. Both EMG gg and EMG dia were band-pass filtered (0.3 1 khz), rectified, and moving time averaged with a time constant of 100 ms. For each breath, the end-expiratory tonic and peak inspiratory phasic EMG activities were determined from the moving time averaged signal. Cross-correlation between breath-by-breath measures of V I and Sa O2, and V I and PET CO2 signals were used to adjust for signal delays associated with circuit dead-space, lung-ear Sa O2 circulation time delays. This allowed temporal alignment of changes in V I to the corresponding changes in Sa O2 and PET CO2 following the onset and offset of gas delivery. Data from periods with poor signal quality or clear EMG, P epi, Sa O2,orPET CO2 artifacts (e.g., wire dislodgement, blocked epiglottic pressure catheter, probe off) were excluded. Data with arousals and 30 s before the arousal were also excluded from analysis. Remaining data were then combined for analysis of all sleep stages together as well as for stage 2 sleep alone to allow more direct comparisons with baseline stage 2 sleep data. Following adjustment for Sa O2 and PET CO2 signal delays, breath-bybreath ventilatory measurements were averaged over the 10-min baseline period, each 30 s of gas exposure, the last 30 s of each intervening room air period, and each 5 min over the hour recovery following the last gas exposure period. To examine potential changes in chemoreflex responses from the first gas exposure to the last, averaged breath-by-breath measurements were determined from the 5 breaths preceding gas onset, the first 3 and last 3 breaths during gas exposure, and all breaths during the 70 s immediately after gas offset and return to room air breathing. Statistical Analysis Based on the work of Pierchala et al. (38), which used a similar study design, we anticipated a within-subject standard deviation of minute ventilation in the order of l/min and a moderate to large effect of IHH on the change in ventilation pre vs. post gas exposure. Thus we anticipated that 12 participants would be sufficient to detect a rise in minute ventilation in the order of 0.5 l/min in the post-gas exposure recovery period with 80% power and a two-sided significance level of Linear mixed model analysis was used to examine day (IHH vs. intermittent medical air control), time and day by time interaction effects in each variable [V I, V T,T I,T E,T TOT, F B, PIF, P epi, R UA, inspiratory (phasic) and expiratory (tonic) EMG for both genioglossus and diaphragm (Insp and Exp EMG gg and EMG dia, respectively)], expressed as absolute values and as a percent change from baseline. Three separate analyses were conducted for each variable to examine gas periods alone, all room air periods throughout the protocol (baseline, between gas episodes and recovery) and room air periods only during baseline and recovery to remove potential confounding effects in ventilatory measures which may not have returned to baseline between gas exposures. Time and day (IHH vs. control) were examined as repeated factors within subjects, using an auto-regressive

4 Table 1. Time (min) spent awake and in each sleep stage covariance structure, and with subjects entered as a random effect, each with a separate intercept. Significant main and interaction effects were examined using post hoc pairwise contrasts conducted within each linear mixed model incorporating Bonferroni correction for multiple comparisons. All values are reported as means SE. P values of 0.05 were considered statistically significant. RESULTS Control Night Experimental Night P Value Stage 1 time, min Stage 2 time, min SWS time, min REM time, min Wake time, min Total time, min Data are means SE. Time (min) spent awake and in each sleep stage on the control (C) and experimental (E) night. SWS, slow wave sleep; REM, rapid eye movement. Subjects The physical characteristics of the 12 participants were age, yr; height, cm; weight, kg; body mass index, kg/m 2 ; FEV 1, % predicted, FVC, % predicted and AHI /h. The final O 2 composition used during the main experimental night for all participants was %. Sleep Architecture There was no significant difference between the control or experimental night in the amount of time spent awake or in IHH during Sleep Does Not Induce LTF Deacon NL et al. each sleep stage (Table 1), except in the percent of total sleep time spent in stage 2, which was significantly lower on the experimental night (E % vs. C %, P 0.003). Gas Periods 537 An example of raw data is presented in Fig. 2. During gas exposures on the experimental night V I,PET CO2,V T,F B, and PIF were all higher while Sa O2 was lower, T I,T E, and T TOT were shorter, and P epi was more negative compared with the medical air control night [during gas exposure periods on experimental (E) and control (C) nights: PIF, E vs. C l/min, P 0.001; T I,E vs. C s, P 0.001; T E,E vs. C s, P 0.001; T TOT,E vs. C s, P 0.001; P epi,e vs. C cmh 2 O, P 0.002; other variables shown in Fig. 3]. Thus hypercapnic hypoxia was achieved during the experimental gas exposures. Inspiratory (phasic) and expiratory (tonic) EMG dia, with data expressed as percent change from baseline, were significantly higher during gas exposures on the experimental night [during gas exposures on the experimental (E) and control (C) nights: inspiratory EMG dia,e % baseline vs. C % baseline, P and expiratory EMG dia,e % baseline vs. C % baseline, P 0.016]. There was no significant difference in either phasic or tonic EMG gg during gas exposures between control and experimental nights, with data expressed as absolute values or percent change from baseline. EMG gg and EMG dia presented as absolute values are shown in Fig. 4. Differences between ventilatory and EMG dia Fig. 2. Example of raw data during IHH. Example of raw data from one participant during two consecutive HH exposures showing inspiratory flow (Flow), electroencephelogram (EEG), diaphragm electromyogram (EMG dia), genioglossal electromyogram (EMG gg), mask CO 2 (PET CO2 ), oxyhemoglobin saturation of oxygen (Sa O2 ), and epiglottic and mask pressure (P epi and P mask, respectively). Thirty-second gas exposures are marked with vertical dashed lines. The horizontal dashed line shows mild hypercapnia (~2 3 mmhg) was achieved during gas exposure.

5 538 IHH during Sleep Does Not Induce LTF Deacon NL et al. both with all sleep stages combined and with analysis restricted to stage 2 sleep (night time interaction effect for all comparisons, P 0.05). When F B data were grouped for the first 5 and last 5 gas episodes and compared between conditions, there was no longer a significant difference from the start to the end of the gas exposures on the experimental day (condition time P 0.238). There was also no significant difference between nights across the 24 gas exposures in any other variable to suggest any change in the ventilatory response from the first to last HH episode (Fig. 3). Similarly, when comparing Fig. 3. Ventilatory measures during the intermittent gas protocol and recovery. Group data for ventilatory measures at baseline (B), throughout the intermittent gas protocol (G1 to G24) and 60 min of recovery (marked at 30 and 60 min, R30 and R60, respectively) with all sleep stages combined. Panels show minute ventilation (V I), end-tidal partial pressure of CO 2 (PET CO2 ), oxyhemoglobin saturation ( SaO2 ), tidal volume (V T), and breathing frequency (F B). Filled circles represent experimental data (HH) and open circles represent control data (intermittent medical air). Data expressed as means SE; n 12. Symbols mark significant difference between experimental conditions. Across all gas exposure periods for experimental (E) and control (C): V I,E vs. C l/min, P 0.001; PET CO2,E vs. C mmhg, P 0.001; Sa O2,E % vs. C %, P 0.001; V T, E vs. C l/min, P 0.001; F B,E vs. C breaths/min, P During all room air periods: V I, E vs. C l/min, P 0.001; F B,E vs. C breaths/min, P Room air breathing measures during baseline and recovery only were V I, E vs. C l/min, P 0.003; PET CO2,E vs. C mmhg, P responses remained when data analyses were restricted to stage 2 sleep. During HH exposures there was a significant increase in F B during the last several gas exposures when data were analyzed as absolute values (Fig. 3) and as percent change from baseline, Fig. 4. EMG measures during the intermittent gas protocol and recovery. Group data for phasic (Insp) and tonic (Exp) EMG gg and EMG dia measures at baseline (B), throughout the intermittent gas protocol (G1 to G24), and 60 min of recovery (marked at 30 and 60 min, R30 and R60, respectively) with all sleep stages combined presented as absolute values; area under the curve per second (au/s). Filled circles represent experimental data (HH) and open circles represent control data (intermittent medical air). Data expressed as means SE; n 12.

6 IHH during Sleep Does Not Induce LTF Deacon NL et al. 539 Fig. 5. Breath-by-breath comparison of first vs. last HH gas episode. Breath-by-breath changes over the first HH gas episode (solid circles) to the last HH (open circles) for minute ventilation (V I), end-tidal partial pressure of CO 2 (PET CO2 ), and oxyhemoglobin saturation (Sa O2 ). Left panel shows breaths preceding onset of gas exposure (marked by the bar at time 0 s) and breaths during the 30-s gas exposure. Right panel shows breaths during 70 s immediately after gas exposure and return to room air breathing. Data are absolute values expressed as means SE; n 12. only the first and last gas episodes, there was no significant difference in V I, PET CO2,orSa O2 from first to last gas exposure to suggest a change in chemoreflex responsiveness following 24 HH exposures (Fig. 5). There was also no significant difference between nights across the 24 gas exposures in R UA, phasic or tonic EMG gg,oremg dia (Fig. 4). Room Air Periods Although there were no significant night by time interaction effects for any variable during the room air breathing periods (baseline, between gas episodes and recovery) to indicate that IHH had induced vltf, ggltf, dialtf, or UALTF, there were significant differences between nights, where V I, F B, and PIF were higher, and T I and T TOT were shorter on the experimental night vs. control [during room air periods on experimental (E) and control (C) nights: T I,E vs. C s, P 0.001; T TOT,E vs. C s, P 0.004; PIF, E vs. C l/min, P Values for V I and F B are shown in Fig. 3]. Similar differences between nights, suggesting a persistent elevation of ventilatory drive throughout the whole experimental night, were evident with all sleep stages combined and with data restricted to stage 2 sleep, with data expressed as absolute values and as percent change from baseline. There was no difference between nights in the absolute measures of R UA, phasic or tonic EMG gg or EMG dia (Fig. 4), with all sleep stages combined or with analyses restricted to stage 2 sleep. However, there was a significant difference between nights in the increase from baseline in both tonic and phasic EMG gg activity (during room air periods on experimental (E) and control (C) nights: phasic EMG gg,e % vs. C %, P 0.005, tonic EMG gg,e % vs. C %, P 0.001). Baseline and Recovery To exclude the possibility of confounding by ventilatory measures which may not have reached baseline between gas exposures, room air periods between gas exposures were excluded, and data from only the baseline and recovery period were analyzed. When comparing all sleep stages combined as absolute values from only baseline and recovery periods, V I, PET CO2, and PIF were higher and T I was shorter on the experimental night compared with the control night [during baseline and recovery periods on the experimental (E) and control (C) nights: PIF, E vs. C l/min, P 0.006; T I,E vs. C s, P Values for V I and PET CO2 are shown in Fig. 3]. However, with all sleep stages combined and with data restricted to stage 2 sleep, expressed as absolute values, there remained no significant day by time interaction effect for any ventilatory variable, phasic or tonic EMG or R UA to support the presence of LTF during the hour long recovery period following IHH (Figs. 3, 4, and 6). When analyzed as the percent change from baseline, there was a significant difference between conditions in breath times and EMG gg, with all data combined and with data restricted to stage 2 sleep [during baseline and recovery periods on the Fig. 6. Minute ventilation during only stage 2 sleep at baseline and 15 min intervals of recovery. Minute ventilation (V I) during only stage 2 sleep in each 5 min block during baseline and at 15, 30, 45, and 60 min of recovery during control (intermittent medical air) and the experimental day (IHH). Data are absolute values expressed as means SE; n 12.

7 540 IHH during Sleep Does Not Induce LTF Deacon NL et al. experimental (E) and control (C) nights with all sleep stages combined: T E,E % vs. C %, P 0.001, T TOT,E % vs. C %, P 0.002, phasic EMG gg,e % vs. C %, P 0.016, tonic EMG gg,e % vs. C %, P 0.017]. DISCUSSION Both vltf and UALTF have previously been reported during sleep in humans without OSA following AIH on a background of sustained isocapnia (38, 43). However, the current study showed no evidence to support that acute IHH, which more closely reflects blood gas disturbances experienced in OSA, induces vltf or UALTF. There was also no change in the ventilatory response during successive HH exposures to suggest sensitization or depression of chemoreflex responsiveness. In combination, these findings support other literature suggesting that the concomitant level of CO 2 importantly modulates neuroplastic effects of IH (14, 18, 35, 48). The absence of significant effects in this study could reflect type II error. Significant increases in ventilation and reductions in R UA have been reported during sleep in healthy nonsnorers following a very similar protocol with similar sample sizes (38), suggesting that similar effects should have been detected in this study. During sleep, Pierchala and colleagues exposed 12 participants to 15 isocapnic hypoxic exposures of ~1 min in duration, reducing Sa O2 to % (38), comparable to the % achieved with the current protocol. Participants were also returned to room air breathing between gas exposures for a similar length of time ( s) compared with the 2 min room air breathing between gas exposures in the current protocol, and recovery was also monitored during room air breathing. Pierchala and colleagues reported that during recovery, V I and V T were increased to % and % of control, respectively, and R UA was decreased to % of control following IH (38). Using the withinsubject standard deviation in V I during baseline on both days of 1.05 l/min, we estimate that the current study had sufficient power to detect a 1.34 l/min (15.4%) or greater increase in V I, with 80% power and 2-tailed significance level of Thus an effect size similar or smaller than that reported by Pierchala and colleagues could certainly have been missed due to Type II error. Lack of vltf and Increased Chemoreflex Responsiveness In the current study V I increased during experimental HH through increased V T and F B. When all gas periods were included in analysis, there was a significant time condition effect in F B which could indicate LTF of F B during HH gas exposures. However, this effect was lost when the first 5 gas episodes were grouped and compared with the last 5 episodes, suggesting that more variable F B during the middle portion of the HH exposures may reflect incomplete F B recovery between acute gas exposures, rather than a systematic LTF effect from the beginning to the end of the protocol. With data analyzed as absolute values and as percent change from baseline, and with all sleep stages combined or restricted to stage 2 sleep, increased ventilation was evident during room air breathing periods on the experimental night. To help determine if this was due to incomplete recovery of baseline breathing between gas exposures, room air breathing periods between gas exposures were excluded and only baseline and recovery room air breathing periods were analyzed to more specifically look for LTF after gas delivery. With all sleep stages combined and with data restricted to only stage 2 sleep, breathing remained elevated across the whole protocol on the experimental night, including during the baseline period before gas exposures. Although statistically significant, these differences were small, with V I only ~0.5 l/min higher compared with the control night. Given the randomized crossed-over design, this difference remains difficult to explain. Despite some baseline differences, when analyzed as absolute values and also as percent change to control for baseline differences, there were no changes from baseline in any ventilatory measure, R UA or EMG measurement during the 1 h recovery following IHH to indicate the presence of LTF. Some other small percent changes from baseline in breath timing and a small reduction in EMG gg activity on experimental nights compared with a small rise on control nights remain difficult to explain, but are not consistent with LTF. These findings could reflect Type I error, or real effects on ventilatory and upper airway control that warrant further investigation. Neuroplastic Mechanisms of CO 2 We elected to use 3% CO 2 based on the work by Younes and colleagues who found that the combination of 3% CO 2 with hypoxia (15% or 11% O 2 ) for ~30 s was the highest CO 2 level that participants remained asleep during hypercapnic hypoxia (50). Although not equivalent to severe OSA, this represents an AHI around 24/h with oxyhemoglobin desaturations of 80 85% more common in moderate OSA, and the highest CO 2 concentration felt likely to allow healthy participants to maintain sleep. The main difference in the protocol used by Pierchala and colleagues and that of the current study was the level of concomitant CO 2 during hypoxic exposures. Pierchala et al. supplemented CO 2 only during hypoxia to maintain isocapnia (38), compared with supplemental CO 2 throughout hypoxia to induce concomitant hypercapnia in the current study. However, it is possible that concomitant intermittent hypercapnia resulted in the lack of LTF in the current study; a possibility already speculated in some detail by Kinkead and colleagues in their review (18). When it was first discovered that IH induced serotonin-dependent pltf and hltf (3), the same authors tested the effects of other respiratory stimuli and found intermittent hypercapnia had an opposing effect, inducing longterm depression characterized by a sustained reduction in peak neural burst amplitude and frequency despite maintenance of PET CO2 (2). Pretreatment with 2 -adrenergic antagonists blocked long-term depression (2), suggesting that intermittent hypercapnia induces ventilatory long-term depression via activation of locus ceruleus neurons, as opposed to the induction of LTF following IH via activation of serotonergic raphe neurons (18). Interestingly, following pretreatment with 2 -adrenergic antagonists (both yohimbine and RX ) some rats expressed pltf (2). Chemoreflex afferent stimulation by either hypercapnia or hypoxia activates caudal raphe neurons, which provide serotonergic innervation of respiratory motorneurons such as the phrenic and hypoglossal nerves (18). This led the authors to propose that blockade of 2 -adrenergic receptors may have

8 allowed facilitation via hypercapnic carotid chemoreflex afferent activated serotonergic pathways (2). Thus it appears that hypercapnia and hypoxia may act in a push-pull fashion, via apposing adrenergic and serotonergic pathways (18). Although intermittent hypercapnia-induced long-term depression has only been demonstrated in rats following severe hypercapnic exposures [5 min exposures to 10% inspired CO 2, raising PET CO2 to mmhg (2) and 3 min exposures to 15% CO 2,PET CO2 not reported by Stipica et al. (45)] and not following moderate hypercapnia (3 5% CO 2, raising PET CO2 to 60 mmhg) similar to that used in the current study (2), it is possible that the activation of both serotonergic and adrenergic pathways during the combination of IHH may have had counteracting effects (18), resulting in the lack of either facilitation or depression in this study. This is supported by a recent study of phrenic neural activity in rats in which intermittent hypoxia induced LTF of phrenic amplitude and intermittent hypercapnia induced long-term depression of phrenic frequency, whereas combined IHH induced no changes in either frequency or amplitude at 60 min after the last stimulus (45). The only other study we are aware of that has examined the effects of IHH on ventilatory neuroplasticity in humans found that fifteen 30-s episodes of breathing 6% O 2 and 5% CO 2 separated by 90 s of breathing air during wakefulness also failed to induce vltf (9). However, hypercapnic adrenergic pathways clearly do not always negate facilitatory serotonergic pathways, given that simulated apneas in rats inducing IHH has been shown to induce serotonin-dependent LTF of both phrenic and hypoglossal nerve activity to a similar magnitude as IH alone (21). In addition, a study of IHH-induced ventilatory neuroplasticity in piglets found that cycle duration of gas exposures was a critical determinant of whether depression, facilitation, or no net change in ventilatory output was induced. Waters et al. (48) found that continuous exposure for 24 min, and 24 min of 2 min cycles did not induce changes to ventilation post gas exposure. However, 8 min cycles induced vltf and 4 min cycles induced ventilatory long-term depression (48). Therefore, although the IHH protocol utilized in the current study during sleep and that of Diep et al. (9) in awake humans both failed to elicit vltf, these findings do not preclude that more variable blood gas disturbances experienced in OSA may nevertheless induce ventilatory neuroplasticity. Position, Arousal, and Sleep Stage Effects on LTF In addition to both hypoxia and hypercapnia, other respiratory stimuli experienced during apneas, such as negative upper airway pressure (41), intermittent vagal feedback (51) and periodic withdrawal of neuromuscular activity (4), have also been found to induce LTF in various motor neurons and ventilatory muscles. Negative pressure pulses initiate ggltf in rats of similar magnitude as IH, either alone or in combination with IH (41). In humans, early studies suggested that vltf was more easily induced in snorers vs. nonsnorers (1a). During IH in the supine position where flow limitation and upper airway resistance are typically increased, higher ventilatory drive and negative upper airway pressures would tend to be greater in snorers with more compliant airways. Therefore, negative upper airway pressure during IH may be an important stimulus for inducing vltf in humans. Participants in this study slept laterally specifically to help stabilize the airway to reduce the IHH during Sleep Does Not Induce LTF Deacon NL et al. 541 potential for more variable flow limitation to confound IHH induced changes. It is possible that reduced negative upper airway pressure stimuli associated with lateral vs. supine positioning contributed to a lack of LTF in the current study. Given the role of negative upper airway pressure stimuli for inducing vltf in humans has not been specifically studied, this possibility requires further evaluation. In rats sleep fragmentation inhibits both the induction of vltf and sensitization of the hypoxic ventilatory response via adenosine A 1 receptors (25). Sleep stage is also thought to alter neuroplastic responses to IH, as in both animals and humans LTF has been difficult to induce during wake as opposed to sleep or anesthesia (16, 31) and LTF is more easily induced with progressively deeper sleep stages in Lewis rats (31). However, a recent study in humans found the magnitude of vltf is enhanced during wake vs. sleep when IH is conducted during mild hypercapnia (46). Therefore, the inhibitory effect of wake on the induction of LTF may be secondary to lower PET CO2. Despite the uncertainty of arousal state on neuroplasticity, it is possible that arousals and varying sleep stages during the protocol could also have impeded neuroplasticity in this study. Pierchala et al. successfully induced vltf during sleep in healthy males, using gas exposures conducted only while subjects were in stable stage 2 or stage 3 sleep, with arousals and wakefulness only occurring for min (out of 15 min) during IH, and only for min (out of 20 min) of the recovery period (38). This equates to arousals and wakefulness occurring during ~15% of the total protocol time and is comparable to the sleep quality achieved in the current study. In this study gas exposures were also only initiated during stable sleep, and on the experimental night participants spent ~25 50% of the night in stage 2 or slow wave sleep and were awake for min of ~160 min (Table 1; ~14%). Importantly, the amount of time spent in slow wave sleep and wakefulness was not significantly different between nights. Given comparable sleep quality to that reported by Pierchala et al., poor sleep appears unlikely to explain the lack of vltf in the current study. Furthermore, arousals, sleep fragmentation, and reduced deep sleep are major features of OSA. Consequently, if sleep fragmentation and reduced deep sleep did impair the development of LTF in humans, LTF may be of limited relevance in OSA patients. On the other hand OSA patients do consistently exhibit treatment reversible abnormalities in chemoreflex control (20, 33, 42, 44, 50), increased sympathetic neural activity, hypertension (32) and state-dependent changes in genioglossal activity (17, 26, 27, 40) suggestive of IH-induced neuroplasticity. Thus there remains strong evidence of ventilatory neuroplasticity in OSA despite frequent arousals and sleep fragmentation. Conclusion It is possible that the findings of this study and that of Diep et al. (9) reflect a true absence of IHH-mediated ventilatory neuroplasticity in humans, unlike IHH-induced LTF already established in rats and piglets (21, 48). However, IH-induced LTF has been demonstrated in multiple animal species including rats (35), cats (22), dogs (6), goats (47) and even phylogenetically distant avian species (30), suggestive of a highly conserved evolutionary feature. Several laboratories have suc-

9 542 IHH during Sleep Does Not Induce LTF Deacon NL et al. cessfully induced LTF and sensitized hypoxic ventilatory response in humans using IH (5, 13, 14), supporting that ventilatory neuroplasticity is also conserved in humans. OSA patients consistently exhibit treatment reversible abnormalities in cardiorespiratory control measures reflective of changes experimentally induced by IH (20, 32, 33, 42). This suggests IHH during naturally occurring OSA does induce neuroplasticity in humans, and that in OSA these forms of neuroplasticity can be pathogenic. Therefore, it appears more likely that either the particular gas composition or the cycle duration of gas exposures resulted in the lack of LTF and chemosensitization in the current study, rather than a lack of IHH inducible effects per se. More work is needed to further clarify potential effects of IHH, as occurs in OSA, on ventilatory, chemoreflex and upper airway control to better guide the development of future potential treatments for correcting unstable ventilatory control in OSA. ACKNOWLEDGMENTS We are very grateful to A. McKenna and M. Ahmer for sleep study scoring, Dr. J. Mercer for technical laboratory assistance, Dr. S. Johnston for O 2 analyzer loan, and to the volunteers who participated in these experiments. GRANTS N. L. Deacon was supported by an Australian Postgraduate Award and the Adelaide Institute for Sleep Health supplementary scholarship. P. G. Catcheside was supported by an Australian Research Council grant (FT ). Part of the study was funded by National Health and Medical Research Council (NHMRC) Grant DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the authors. AUTHOR CONTRIBUTIONS N.L.D., R.D.M., and P.G.C. conceived and designed research; N.L.D. and D.L.S. performed experiments; N.L.D. and P.G.C. analyzed data; N.L.D., R.D.M., D.L.S., and P.G.C. interpreted results of experiments; N.L.D. and P.G.C. prepared figures; N.L.D. drafted manuscript; N.L.D., R.D.M., D.L.S., and P.G.C. edited and revised manuscript; N.L.D., R.D.M., D.L.S., and P.G.C. approved final version of manuscript. REFERENCES 1. AASM Task Force. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. The Report of an American Academy of Sleep Medicine Task Force. Sleep 22: , doi: /sleep/ a.Babcock MA, Badr MS. Long-term facilitation of ventilation in humans during NREM sleep. Sleep 21: , Bach KB, Mitchell GS. Hypercapnia-induced long-term depression of respiratory activity requires alpha2-adrenergic receptors. J Appl Physiol (1985) 84: , Bach KB, Mitchell GS. Hypoxia-induced long-term facilitation of respiratory activity is serotonin dependent. Respir Physiol 104: , doi: / (96) Baertsch NA, Baker-Herman TL. Inactivity-induced phrenic and hypoglossal motor facilitation are differentially expressed following intermittent vs. sustained neural apnea. J Appl Physiol (1985) 114: , doi: /japplphysiol Brugniaux JV, Pialoux V, Foster GE, Duggan CT, Eliasziw M, Hanly PJ, Poulin MJ. Effects of intermittent hypoxia on erythropoietin, soluble erythropoietin receptor and ventilation in humans. Eur Respir J 37: , doi: / Cao KY, Zwillich CW, Berthon-Jones M, Sullivan CE. Increased normoxic ventilation induced by repetitive hypoxia in conscious dogs. J Appl Physiol (1985) 73: , Chowdhuri S, Pierchala L, Aboubakr SE, Shkoukani M, Badr MS. Long-term facilitation of genioglossus activity is present in normal humans during NREM sleep. Respir Physiol Neurobiol 160: 65 75, doi: /j.resp Chowdhuri S, Shanidze I, Pierchala L, Belen D, Mateika JH, Badr MS. Effect of episodic hypoxia on the susceptibility to hypocapnic central apnea during NREM sleep. J Appl Physiol (1985) 108: , doi: /japplphysiol Diep TT, Khan TR, Zhang R, Duffin J. Long-term facilitation of breathing is absent after episodes of hypercapnic hypoxia in awake humans. Respir Physiol Neurobiol 156: , doi: /j. resp Eckert DJ, Catcheside PG, Stadler DL, McDonald R, Hlavac MC, McEvoy RD. Acute sustained hypoxia suppresses the cough reflex in healthy subjects. Am J Respir Crit Care Med 173: , doi: /rccm oc. 11. Eckert DJ, McEvoy RD, George KE, Thomson KJ, Catcheside PG. Genioglossus reflex inhibition to upper-airway negative-pressure stimuli during wakefulness and sleep in healthy males. J Physiol 581: , doi: /jphysiol Garcia N, Hopkins SR, Powell FL. Effects of intermittent hypoxia on the isocapnic hypoxic ventilatory response and erythropoiesis in humans. Respir Physiol 123: 39 49, doi: /s (00) Harris DP, Balasubramaniam A, Badr MS, Mateika JH. Long-term facilitation of ventilation and genioglossus muscle activity is evident in the presence of elevated levels of carbon dioxide in awake humans. Am J Physiol Regul Integr Comp Physiol 291: R1111 R1119, doi: /ajpregu Hilditch CJ, McEvoy RD, George KE, Thompson CC, Ryan MK, Rischmueller M, Catcheside PG. Upper airway surface tension but not upper airway collapsibility is elevated in primary Sjögren s syndrome. Sleep 31: , doi: /sleep/ Jordan AS, Catcheside PG, O Donoghue FJ, McEvoy RD. Long-term facilitation of ventilation is not present during wakefulness in healthy men or women. J Appl Physiol (1985) 93: , doi: / japplphysiol Jordan AS, White DP, Lo YL, Wellman A, Eckert DJ, Yim-Yeh S, Eikermann M, Smith SA, Stevenson KE, Malhotra A. Airway dilator muscle activity and lung volume during stable breathing in obstructive sleep apnea. Sleep 32: , doi: /sleep/ Kinkead R, Bach KB, Johnson SM, Hodgeman BA, Mitchell GS. Plasticity in respiratory motor control: intermittent hypoxia and hypercapnia activate opposing serotonergic and noradrenergic modulatory systems. Comp Biochem Physiol A Mol Integr Physiol 130: , doi: /S (01) Lee DS, Badr MS, Mateika JH. Progressive augmentation and ventilatory long-term facilitation are enhanced in sleep apnoea patients and are mitigated by antioxidant administration. J Physiol 587: , doi: /jphysiol Loewen A, Ostrowski M, Laprairie J, Atkar R, Gnitecki J, Hanly P, Younes M. Determinants of ventilatory instability in obstructive sleep apnea: inherent or acquired? Sleep 32: , doi: / sleep/ Mahamed S, Mitchell GS. Simulated apnoeas induce serotonin-dependent respiratory long-term facilitation in rats. J Physiol 586: , doi: /jphysiol Mateika JH, Fregosi RF. Long-term facilitation of upper airway muscle activities in vagotomized and vagally intact cats. J Appl Physiol (1985) 82: , McEvoy RD, Popovic RM, Saunders NA, White DP. Effects of sustained and repetitive isocapnic hypoxia on ventilation and genioglossal and diaphragmatic EMGs. J Appl Physiol (1985) 81: , McGuire M, Liu C, Cao Y, Ling L. Formation and maintenance of ventilatory long-term facilitation require NMDA but not non-nmda receptors in awake rats. J Appl Physiol (1985) 105: , doi: /japplphysiol McGuire M, Tartar JL, Cao Y, McCarley RW, White DP, Strecker RE, Ling L. Sleep fragmentation impairs ventilatory long-term facilitation via adenosine A1 receptors. J Physiol 586: , doi: /jphysiol McSharry DG, Saboisky JP, Deyoung P, Matteis P, Jordan AS, Trinder J, Smales E, Hess L, Guo M, Malhotra A. A mechanism for upper airway stability during slow wave sleep. Sleep 36: , doi: /sleep Mezzanotte WS, Tangel DJ, White DP. Waking genioglossal electromyogram in sleep apnea patients versus normal controls (a neuromuscular