INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

OBSTRUCTIVE SLEEP APNEA is characterized by occlusions of upper airways (UAWs) during sleep. The activity of the UAW dilator muscles plays an important role in the patency of the UAWs. Among those muscles, the genioglossus (GG), which is innervated by the medial branch of the hypoglossal (HG) nerve, has been given particular attention, because the main function of the GG is to protrude the tongue. Anesthesia (13, 21) and sedation (4) can affect the HG activity level. Thus several chronic animal models have been developed to study the GG response to the loading of the UAWs in unsedated animals (13, 15, 19, 23). In these models, the airways are terminated with an elastic, resistive, or infinite load (total occlusion) to simulate the effects of the occlusions in patients with obstructive sleep apnea. In some other chronic animal models, the UAW is closed remotely with a computer-controlled valve (17) or an inflatable balloon placed in the trachea (22) to study the physiological consequences of the occlusions. In this study, we developed a new dog model of UAW obstruction whereby an external force is applied directly on the submental region to mechanically narrow the pharynx and, therefore, partially occlude the airways. Using this model, we studied the response of the HG nerve to loading of the pharynx during sleep.

Direct recordings of the HG nerve activity have not been reported in unsedated animals. Alternatively, in conscious animals and humans, electromyogram (EMG) recordings with wire electrodes have been utilized to study the activity of the muscles innervated by the HG nerve, especially the GG muscle. However, EMG recordings with wire electrodes have several problems: the signals favor the local activity (26) and contain movement artifacts, and the signal amplitudes are not reproducible from implant to implant. Another method of choice is to place a cuff electrode directly on the HG nerve. Although nerve cuff electrodes are not entirely free from the problems similar to those encountered with the EMG electrodes, the mechanical interface, and thus the reproducibility of recordings, is much more improved. The spiral-cuff-electrode design (20), which was previously shown to be able to record the HG and phrenic nerve discharges under different conditions in acute preparations (24, 25), was used. In this study, we also addressed the issues regarding chronic electrode implantations on the HG nerve, such as the nerve insult and the signal quality of the recordings.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Two healthy beagles (young adult, 10-12 kg) with normal UAW anatomy were chronically implanted with cuff electrodes for recordings of HG activity and with electroencephalogram (EEG) and electrooculogram (EOG) electrodes for sleep staging. The cuff electrodes were implanted bilaterally in one animal (beagle 1, male) for durations of 17 mo (cuff 1) and 7 mo (cuff 2) and unilaterally in the other animal (beagle 2, female) for 6 mo (cuff 3). All the surgical and experimental procedures were designed according to the Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee of Case Western Reserve University, Cleveland, Ohio.

Surgical procedures. The animals were initially anesthetized with a short-acting barbiturate (thiopental, 25 mg/kg iv), intubated, and ventilated with a gaseous mixture of halothane, nitrous oxide, and oxygen. Incisions (5 cm long) were made on both sides of the upper neck area. Superficial muscles of the region were retracted, and an ~3-cm length of the HG nerve was separated from its surrounding tissue. A self-coiling spiral-cuff electrode (Fig. 1A) was implanted on the main trunk of the HG nerve near the bifurcation point of lateral and medial branches. A couple of loops were made in the electrode wires and sutured to a nearby muscle to avoid direct pulling on the electrode. Electrode wires were tunneled subcutaneously to an exit site between the scapulae. The incision was closed, and the animal was turned over for implantation of EEG and EOG electrodes.

View larger version (27K): [in this window] [in a new window]   Fig. 1.   A: self-coiling spiral-cuff electrode. Three platinum contacts were embedded in the innermost layer and exposed through windows cut out from silicone cuff on inside. B: tripolar connection of electrode contacts to amplifier with use of a transformer. HG, hypoglossal; BW, bandwidth.

Cuff electrodes. The self-coiling spiral design was chosen for the cuff electrodes, the ability of which to record the activity of the HG nerve has been demonstrated in anesthetized preparations (24, 25). A detailed description of the electrode fabrication can be found elsewhere (20, 25). The cuff electrodes used in this study were 20 mm in length and 2.5 mm in diameter (inner diameter of the first layer) and had 2.25-2.75 turns (Fig. 1A), snugly fitting the nerves in their resting position. Cuff electrodes had three contacts (each 9 mm apart, size of the exposed area = 2 × 1.25 mm) made from platinum foil (purity = 99.95%, thickness = 25 µm; Goodfellow), which were spot welded to multistrand stainless steel (316 LVM) Teflon-coated wires (1 × 7 × 0.00135 in., Fort Wayne Metals) for connections. The electrode wires were twisted together before implantation to reduce the tension on the wires under bending forces. A longitudinal silicone piece from one end of the cuff to the other end was glued on the opposite side of the contacts on the first layer as a backbone to improve the longevity of the contacts. The electrode wires exited the cuff from one end on the first layer through a tapering tail that helped reduce the mechanical stress on the wires at the exit point. Nerve activity was recorded from the middle contact with respect to the end contacts that were shorted as shown in Fig. 1B.

Experimental setup. The dogs were trained to sleep lying on one side with their necks in a straight position in a one-side-open Faraday cage (52 × 70 × 165 cm) in the presence of the experimenter. The leads from the implanted electrodes were connected to the recording electronics before each session via a flat cable that was long enough to allow the animal to move freely inside the cage. A custom-designed apparatus with a pneumatic piston that could be advanced remotely (Fig. 2) was used to apply a perpendicular force on the submental region, ~2 cm rostral to the hyoid bone, thereby narrowing the pharyngeal portion of the UAWs (see the small dog head figure in Fig. 2B). The force applicator was held in place with the help of a thermoplastic mold that was worn around the animal's head. A small condenser microphone was mounted on the thermoplastic mold near the pharynx to record the snoring sounds. A custom-made cylindrical balloon was placed in the lower one-third of the thoracic portion of the esophagus (by having the animal swallow it with a bolus of soft food) before each sleep session for measurements of the esophageal pressure (Pes), as an estimate of the tracheal pressure. Respiratory abdominal movements were measured with Respitrace (Ambulatory Monitoring) by using an inductive band transducer worn around the belly. All the raw signals were continuously digitized (Digital Data Recorder, model VR-10B, Instrutech) and recorded on videotapes for later analysis.

View larger version (19K): [in this window] [in a new window]   Fig. 2.   Force applicator. A: the profile as worn by the animal. B: detailed presentation.

Experimental procedure. HG activity was recorded continuously in wakefulness (W) and sleep. Sleep experiments were held at night, usually before midnight. The animals were exercised 30-45 min before each session by being walked on a leash. The recording experiments (a total of 53 sessions) began at least 2 mo after surgery and were spread over time until animals were terminated. Each session lasted between 2 and 4 h and included multiple sleep cycles. A force was applied on the submental region to mechanically collapse and, therefore, load the UAWs internally during non-rapid-eye-movement (NREM) sleep. The submental force was increased in steps of 1 or 2 N, starting from zero up to a maximum value, waiting at least 10 breaths at each level. The maximum force was defined as the largest force value at which the animal was not aroused from sleep. The force transition from one level to the next took less than two breath cycles.

Sleep staging. EEG and EOG signals and the observed state of the animals were used to differentiate between W, NREM sleep, and rapid-eye-movement (REM) sleep stages. The NREM sleep stage was characterized by larger amplitudes and slower frequency components in the EEG signal relative to either the W or REM sleep stage. REM sleep was characterized by low amplitudes in EEG and often sharp edges in the EOG signal. The REM sleep stage was typically associated with twitches in the face and jerks in the legs.

Force applicator. A 5-ml glass syringe was mounted on a thermoplastic mold that was shaped to fit comfortably around the dog's head (Fig. 2). The outside end of the plunger was cut off, and a Plexiglas piece with a relatively larger surface area (2.75 cm²) was glued on the top by using fast-drying epoxy. The Plexiglas piece was shaped to conform to the anatomic structures in the submental area to minimize the disturbing effect of the force on the animal during sleep. A thin latex bag was placed around the exposed end of the plunger, and the bag was filled with water-soluble lubricating jelly for smooth movement of the plunger. A piece of rubber sheath cut into an appropriate shape was wrapped around the mold, and the ends were held together over the head with the help of Velcro attachments to further stabilize the apparatus around the animal's head. A 40-cm-long flexible tubing (ID = 2.4 mm, OD = 4 mm; Tygon, Fisher Scientific) was attached to the syringe and continued with a longer and stiffer polyethylene tubing (ID = 3.05 mm, length = 2 m; TFE 9 Standard Wall, Zeus Industrial Products, Orangeburg, SC) to transmit the pressure to a remote transducer (Deltran, Utah Medical Products, Midvale, UT). The pressure measurements inside the system were scaled with the cross-sectional area of the syringe to determine the value of the submental force. Another syringe was included into the system for the remote control of the submental force by adding or removing air. The system had a volume of 18 cm³, excluding the syringes.

Pes measurements. A 5-cm-long cylindrical silicone tubing (2.5 mm diameter, Dow Corning) with very thin walls served as a sensor in the design of the esophageal balloon (Fig. 3A). The end was closed with a small ball made of silicone curing agent (MDX4-4210, Dow Corning). The open end of the cylindrical sensor was glued with the curing agent to a 14-cm-long silicone tube with a smaller diameter (ID = 0.64 mm, OD = 1.19 mm; Dow Corning), continued with another silicone tubing of 15-cm length (ID = 0.30 mm, OD = 0.64 mm; SM-7755A Durometer, Sil-Med), and terminated with a plastic socket. The balloon was positioned inside the esophagus such that the pressure readings were maximum during the inspiratory phases. The first piece of tubing connected to the pressure sensor traveled along the esophagus to the mouth cavity. The thinner piece ran from behind the right teeth out through the corner of the mouth. The plastic socket was connected to a pressure transducer (Deltran, Utah Medical Products) during experiments. The balloon was deflated in the beginning of each session to avoid saturation in the mechanical transfer function of the balloon at large negative Pes values. The steady-state response of the balloon and the pressure transducer combination was nearly linear for the range of negative pressures measured in the esophagus (Fig. 3B). The response time (rise time from 10 to 90% of maximum) of the Pes measurement system was estimated as 170 ms by putting the balloon in a vacuum and quickly raising the pressure to the ambient pressure.

View larger version (20K): [in this window] [in a new window]   Fig. 3.   A: esophageal balloon design. A 5-cm-long silicone tubing with very thin walls served as a pressure sensor. D, diameter; L, length, I.D., inner diameter. B: transfer function of esophageal balloon for steady negative pressures.

Signal conditioning. HG nerve recordings were first amplified with a step-up audio transformer (turn ratio = 1:5; part #24500, PICO Electronics) and then further amplified and filtered between 300 Hz and 10 kHz (P5 series, Grass Medical Instruments) as shown in Fig. 2. The HG signal was then digitized at a rate of 47.2 ksamples/s and converted to an appropriate format for storing the data on videotapes (Digital Data Recorder, model VR-10B, Instrutech). EEG and EOG signals were also amplified, band-pass filtered from 1 to 30 Hz, and digitized at 60 samples/s.

Data analysis. The MATLAB programming tool was used for breath-by-breath analysis. The inspiratory (TI) and expiratory times and the breathing rate were measured from the Pes signal. The area under the phasic HG signal during the inspiratory period above the baseline was calculated as a measure of total HG output (AreaHG). The height of the phasic Pes (PeakPes) and the line integral during the inspiratory phase (AreaPes) were computed. A Student's t-test with an assumption of unequal variances was used for all measurements of statistical significance.

Histology. At the end of the study, animals were deeply anesthetized with pentobarbital (50 mg/kg iv) and perfused with saline followed by 4% paraformaldehyde. The main trunks and the branches of the HG nerves were dissected bilaterally from the surrounding tissue and cut 1-2 cm proximal and distal to the implanted cuff electrodes. The contralateral HG nerve that was not implanted in beagle 2 was used as a control. Explanted nerves were placed in 3.5% glutaraldehyde. Histological sections were made at longitudinal locations of 5 and 10 mm away from each end of the cuff, at the edges of the cuff, and at locations corresponding to the center of all three contacts inside the cuff.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Dog model of UAW obstruction. During sleep sessions, the application of a small submental force initially produced a light snoring sound, and the sound level became louder with increasing force amplitude. At near-obstruction-force levels, the snoring sounds either became deeper and smoother or turned into louder intermittent vibratory sounds, the latter of which were usually associated with arousals. The near-obstruction state could be achieved only in deep-sleep stages. The two dogs used in this study were different in terms of the forces required to collapse their UAWs. It took larger submental forces to bring beagle 1 to a near-obstruction state than beagle 2 (6-8 vs. 3-4 N). Also, similar amounts of subobstructive levels of internal loading, judged by AreaPes parameter, could be achieved in beagle 2 with one-half of the force level required in beagle 1. In general, the force levels that were needed to cause severe breathing difficulty in REM sleep were much lower than those required in NREM sleep in both dogs (0-3 vs. 6-8 N in beagle 1 and 0-1 vs. 3-4 N in beagle 2). In some cases, the presence of the thermoplastic mold around the head was sufficient to cause the UAWs to collapse on a transition from NREM to REM sleep and an immediate arousal from sleep.

View larger version (20K): [in this window] [in a new window]   Fig. 4.   Relative changes in various physiological parameters when submental force was increased from 0 (control) to a maximum level in non-rapid-eye-movement (NREM) sleep. TI, inspiratory time; fb, breathing rate; ABD, abdominal movement; PeakPes, peak esophageal pressure (Pes); AreaPes, area of Pes during inspiration. Bar plots show mean + SD of relative increases in multiple trials (21 trials in beagle 1 and 5 trials in beagle 2). Control values correspond to 100%.

HG activity in NREM sleep. During NREM sleep without loading, phasic HG activity was low and sometimes not observable above the baseline level. The signal-to-noise ratio of the recordings, defined as the peak HG signal divided by the baseline level, had a value of 1.59 ± 0.43 (n = 24).

View larger version (40K): [in this window] [in a new window]   Fig. 5.   Force transition maneuver in NREM sleep in beagle 1. Traces from top to bottom: submental force, Pes, rectified-integrated HG activity, ABD, and electroencephalogram (EEG) signal.

View larger version (21K): [in this window] [in a new window]   Fig. 6.   Breath-by-breath measurements of area under phasic HG signal (AreaHG) vs. AreaPes in NREM sleep in beagle 1 (A) and beagle 2 (B). Data were obtained from 9 different force trials in each animal.

Temporal responses to UAW loading in NREM sleep. A typical force transition maneuver during NREM sleep is shown in Fig. 7A. The Pes swings and the phasic HG activity are increased in the next breath after a step increase of 2 N in the submental force. Both responses persist as long as the submental force is held at the elevated level.

View larger version (2723K): [in this window] [in a new window]   Fig. 7.   Temporal responses to loading in NREM sleep. A: force transition maneuver in NREM sleep. Traces from top to bottom: submental force, Pes, rectified-averaged (100-ms) HG activity, ABD, and EEG signal. Relative increases in AreaHG (B), AreaPes (C), and TI parameters (D) during force transition maneuvers. Force transition takes place within 6th and 7th breaths. Bars show average of corresponding breaths from 23 different trials. All measurements within a trial were normalized before corresponding breaths were averaged from all trials such that mean of the 5 breaths before force transition was 0% and mean of 5 breaths after force transition was 100%.

HG activity in REM sleep. The HG nerve usually became more active after a transition from the NREM to REM sleep stage without applying the submental force while the force applicator was in place (Fig. 8A). The HG nerve was even more active when a small submental force was applied after the onset of REM sleep (Fig. 8B). Although the activity level was very variable in the REM sleep stage, the largest activity levels were usually associated with irregular abdominal movements and sometimes intermittent snoring sounds, which were suggestive of partial occlusions (Fig. 8B).

View larger version (3141K): [in this window] [in a new window]   Fig. 8.   HG activity recorded in beagle 1. A: during a transition from NREM to rapid-eye-movement (REM) sleep without application of submental force; B: while a submental force of 3 N is applied after onset of REM sleep. Note that HG nerve is very active during REM sleep, especially when there is a submental force. EOG, electrooculogram.

HG activity in W. The ENG signal was at the baseline level when the animals did not use their tongue, as shown during the first 10 s in Fig. 9A. There were no noticeable variations in the baseline level that could clearly be attributed to the position of the head or the posture of the animals when the dogs rested quietly. The small changes observed could result from physical deformation of the electrodes or other sources of noise. Small phasic variations were superimposed on the tonic activity during panting (Fig. 9A). The peak HG activity values were on the order of several microvolts above a baseline level of ~0.25 µV during various types of voluntary tongue movements, such as swallowing and licking water (Fig. 9, B and C).

View larger version (26K): [in this window] [in a new window]   Fig. 9.   Rectified-averaged HG nerve activity recorded during voluntary tongue movements: panting (A), swallowing (B), and drinking water (C).

ENG baseline signal. The combined impedance of nerve and cuff electrode were measured after the tissue encapsulation process was complete to estimate the thermal noise generated in the signal source (nerve/cuff electrode) by using the Boltzmann equation (Fig. 10). Measurements were made by using the tripolar connection of the contacts (as in Fig. 1B) at various frequencies from 100 up to 10,000 Hz. The impedance values measured at 2 kHz were 3.3, 2.7, and 4.6 k for cuffs 1, 2, and 3, respectively. The calculated thermal noise levels due to each one of these electrode impedances were in agreement with the measured baseline levels in the HG recordings, indicating that most of the baseline signal was due to thermal noise. There was never a noticeable increase in the baseline level because of the submental force in beagle 1 (Fig. 5), with neither of the cuffs implanted and only a slight tonic response in beagle 2 (not shown).

View larger version (19K): [in this window] [in a new window]   Fig. 10.   Cuff/nerve impedances measured in tripolar connection (see Fig. 1B) at various frequencies. Plot for each cuff shows mean ± SD of 3-4 sets of measurements taken every month after first 2 mo of surgery.

Nerve insult. In the study reported here, the instrumented dogs fully recovered and returned to their presurgical eating habits within 1 wk after surgery. Neither of the dogs had any observable functional loss in the use of their tongues for the duration of observation. There was no evidence in the behavioral pattern of the animals to suggest that the presence of the cuff electrodes on the HG nerves caused disturbance.

View larger version (167K): [in this window] [in a new window]   Fig. 11.   Histology sections from both HG nerves in beagle 2. One side was implanted for 6 mo, and the other was the control side. Shown are sections proximal to (A), in the middle of (B), and distal to the cuff (C), and the control side (D).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Chronic recordings of HG nerve. The results of this study demonstrated the feasibility of chronic recordings of the HG nerve with spiral-cuff electrodes. The cuff electrodes of this study were twice as long as the ones used for the initial demonstration of the HG recordings in the anesthetized animals (25). Longer cuffs were chosen because the amplitudes of ENG signals increase with increasing cuff lengths (27). The cuff-electrode implants in these dogs did not cause any observable functional impairment in the tongue function for implantation periods as long as 17 mo. The sample size is rather small in this study (3 electrodes). However, these data provide valuable evidence for the feasibility of the cuff-electrode implants on the HG nerve when the implantation times and the size of the electrodes (20 mm long) are considered. In a recent study in which six adult dogs were implanted bilaterally with half-cuff electrodes (9 mm long) for 3 mo, no significant histological lesions were noted in the HG nerve, endoneurium, or perineurium (6).

Temporal HG response to loading. This study showed that phasic HG activity has a rapidly increasing and persistent response to the internal loadings of the UAWs during NREM sleep. Contrary to some other reports, a progressive response spanning five or more breaths was not observable in our experiments (15, 16, 18). This progressive HG or GG response, which follows an increasing respiratory drive, is mediated through the chemoreflex mechanisms, whereas an immediate response probably involves the reflexes that are elicited by stimulation of the UAW mechanoreceptors. In sleeping dogs and humans, phasic GG activity increased on the first occluded breath and continued to increase progressively over the next two to three breaths (15, 16, 18). In another human study, the immediate response of the GG activity was completely absent despite a progressive response when a continuous negative airway pressure was applied to the UAWs during NREM sleep (1). However, a number of studies have shown the effectiveness of using oscillatory pressures as a mechanical stimulus in the UAWs to generate an immediate GG response. A study in dogs showed that high-frequency oscillations applied to the isolated UAWs at frequencies similar to those observed during snoring but at much less amplitude can induce immediate and sustained augmentation of GG activity in W and sleep (23). This finding was confirmed in normal subjects during sleep and in patients with sleep apnea (10). The immediate activation of the GG muscle by a brief stimulus (<1 s) of negative pressure was demonstrated in sleeping humans, although the response was much reduced and delayed compared with W (11, 30). These findings suggest that a changing pressure in the UAWs is a much more efficient stimulus than a continuous negative pressure in terms of eliciting the immediate GG response. This can explain the phasic HG response observed in this study, which increased rapidly after the force and reached its steady-state value as soon as the force transition was complete (the force was not applied abruptly so as not to arouse the animal from sleep). The external force was applied to the submental region such that it would narrow the pharyngeal passage (see the small head figure in Fig. 2B). Thus the submental force used in this study was more likely to produce oscillatory pressure changes (e.g., snoring) in the pharynx than a continuous airway pressure applied to the airways or the total occlusion of the airways. Moreover, the submental force is less likely to generate a progressive HG response because it should not alter the blood gases as much as would total occlusions.

Temporal TI response to UAW loading. The loading method used in this study also caused a fast increase in the TI. During resistive and elastic loadings in NREM sleep, there is no discernible immediate first- or second-breath increase in the diaphragm activity and overall inspiratory drive in humans (2, 5, 12, 14, 31). Moreover, in dogs that were breathing through an endotracheal tube, the TI did not increase significantly at the first breath as a response to resistive and elastic loadings during sleep (5). On the contrary, a negative airway pressure applied to either the nasopharynx or the larynx (but not the mouth) in anesthetized dogs produced a prolongation of the inspiratory period in the first breath, and it gradually returned to the control values within several breaths (29). Thus stimulation of airway mechanoreceptors may be essential for eliciting an immediate response in the respiratory drive. Unlike the quickly increasing and decaying or progressively increasing responses reported in these other studies, in this study the TI increased when it loaded quickly and persisted as long as the submental force was applied. This discrepancy suggests a fundamental difference between the loading scheme used in this study and that used in the other studies in the way that they elicited the respiratory drive response.

HG activity in REM sleep. The irregular pattern of respiratory drive in REM sleep makes it difficult to compare the HG response to loading in REM sleep with that in NREM sleep. However, we observed that the HG nerve was more active in REM sleep, especially when submental force was applied, than in the preceding NREM sleep episodes. This observation seems to contradict some previous reports of GG recordings (15, 26). The discrepancy may be due to the nature of the loading scheme used in this study. A smaller force was able to collapse the UAWs in REM sleep compared with in NREM sleep. This could be attributed to the overall reduction in the tonic innervation of the UAW muscles in REM sleep. As a result, the submental force was probably able to stimulate the UAW mechanoreceptors much more efficiently in REM sleep than in NREM sleep by being able to force the airways to a smaller size.

Tonic HG activity. The measured HG baseline signal levels were close to the estimated thermal noise levels and never increased as a response to the submental force with either of the cuffs in beagle 1. This suggests that the baseline signal mostly consisted of the thermal noise that was generated in the resistive component of the nerve/cuff impedance and that the tonic activity recorded with this implant was small compared with the thermal noise level. Although there was a small tonic response in beagle 2, the change in the HG baseline was much smaller than that of the phasic component and also that of the baseline variations between experiments. The variation in the HG baseline observed between experiments can be attributed to changes in the cuff-electrode impedance because of the reshaping of the cuff that is caused by the forces created by the surrounding muscles. In this study, a small size was chosen for the cuff-electrode contacts (2 × 1.25 mm) to increase the mechanical durability. This probably increased the contact impedances and thereby the thermal noise levels in the recordings. Otherwise, the presence of tonic HG activity has been demonstrated in human (26) and animal (8, 9, 15) studies with GG muscle recordings during sleep and W. Because the thermal noise level was higher than the HG tonic activity level, the decrease in the HG tonic activity at the onset of REM sleep, as reported by other groups (15, 26), was not observed in our recordings. This can also explain why the phasic component and the changes in the baseline signal in correlation with the head position, which were demonstrated with GG recordings in cats (3), were not detectable during quiet W in our dogs.