INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In patients with obstructive sleep apnea (OSA), activity of the upper airway (UA) dilator muscles such as the genioglossus is increased compared with normal subjects during wakefulness. This augmented motor tone presumably represents a neuromuscular compensatory mechanism for maintaining pharyngeal patency (1). Activity of the genioglossus declines during sleep (2, 3). The sleep-related loss of this motor activity in the presence of abnormal pharyngeal anatomy contributes to airway collapse in patients with OSA. Several lines of evidence implicate serotonergic mechanisms in the state-related changes in upper airway muscle activity. First, activity of the hypoglossal nerve, which innervates the genioglossus, is increased with application of serotonergic agonists directly on the hypoglossal motor nucleus (4). Second, serotonergic neurons located in the caudal raphe of the brainstem send projections to the hypoglossal nucleus (5). Finally, activity of these raphe neurons is highest during wakefulness and declines progressively during sleep (6). Thus, withdrawal of excitatory serotonergic input to the hypoglossal nucleus originating in the caudal raphe may predispose to pharyngeal instability during sleep. Several clinical trials of serotonergic drugs in patients with OSA have demonstrated improvement in sleep-disordered breathing (7, 8). These medications may act directly on brainstem regions modulating UA muscle activity.

The genioglossus is also influenced by mechanoreceptor stimulation occurring locally in the pharynx (9). Negative pressure applied to the isolated upper airway causes reflex activation of the genioglossus, whereas positive airway pressure reduces genioglossus electromyographic (EMGgg) activity (1, 8, 9, 12, 15, 16). This pressure-elicited mechanoreflex is also attenuated in sleep, further contributing to state-related changes in the stability of the pharynx.

The mechanoreflex response of the genioglossus muscle, however, does not appear to be modulated by serotonergic inputs. Douse and White found that serotonin applied to the hypoglossal motor nucleus increased tonic hypoglossal nucleus activity but did not significantly change the hypoglossal whole nerve reflex response to UA negative pressure in decerebrated and vagotomized cats (17). We therefore hypothesized that augmentation in genioglossus activity brought about by serotonergic agents should not be affected by the application of nasal continuous positive airway pressure (CPAP), which abolishes the mechanoreflex-mediated increase in EMGgg.

To test this hypothesis we studied a selective serotonin reuptake inhibitor (SSRI), paroxetine hydrochloride, for its effects on EMGgg in normal humans during quiet breathing and under steady-state levels of hypercapnia (HYP), with and without nasal CPAP. Elevated CO₂ was used to increase baseline EMGgg activity.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Studies were performed on 11 healthy male volunteers between 23 and 44 yr of age (mean = 36.72 ± SEM 2.24 yr) having body mass index (BMI) between 22.2 and 31.1 kg/m² (mean = 27.94 ± SEM 1.0 kg/ m²). None complained of frequent or heavy snoring. A twelfth subject was recruited but dropped out due to nausea while on paroxetine.

Study Design

A double-blind, placebo-controlled, crossover design was used. Subjects were given either the SRI paroxetine hydrochloride at a dose of 30 mg daily for 5 d or an identical-looking placebo. After a minimum washout period of 7 d, the subjects were crossed over to the other agent. Treatment order was randomized. Measurements were made at the end of each 5-d exposure period. The study protocol was approved by the institutional review board of the University of Medicine and Dentistry of New Jersey-Robert Wood Johnson Medical School and informed consent was obtained prior to testing.

EMGgg was measured using a bipolar intraoral surface electrode similar in design to that described by Doble and colleagues (18). Briefly, a cast of the inside of the lower mouth of each subject was formed using a bite tray (Premiere Dental Products, Norristown, PA) filled with fast-setting dental impression material (Jeneric/Entron Inc., Wallingford, CT). Two lengths of Teflon-coated wire bared at both ends (diameter 0.010 inch, bare; 0.013 inch, coated; A-M Systems, Everett, WA) were inserted into the cast such that, with the cast inserted in the mouth, they ran along the floor of the mouth for approximately 1.5 cm at a distance of 0.5 cm from the midline. The free end of each wire was connected to a biologically isolated AC amplifier (Grass Instruments, Quincy, MA) amplified by 1,000 and filtered between 10 and 1,000 Hz. The amplifier output was sent to a computer for on-line signal viewing and storage (CODAS; DATAQ Instruments Inc., Akron, OH). The EMGgg signal was subsequently rectified and a moving time average obtained using a 100-ms window. Using these techniques EMGgg measurements during quiet breathing were reproducible within subjects on two different days (n = 6, p = 0.75) (Table 1).

End-tidal CO₂ was sampled (Model 223 CO₂ analyzer; Datex Engstrom, Helsinki, Finland) from a cannula positioned at the subject's nares. The cannula was inserted through a tight-fitting nasal CPAP mask. Mask pressure was obtained using a physiologic pressure transducer (Model C15; Validyne, Northridge, CA). Outputs from both the CO₂ analyzer and pressure transducer, along with the amplified electromyographic (EMG) signal, were recorded continuously by desktop computer for subsequent analysis.

Protocol

Studies were performed with each subject awake and seated in a dental chair with the head elevated and supported at a 45° angle. EMGgg was measured under three conditions in each subject: (1) eupnea or room air breathing (RA); (2) breathing a gas mixture enriched with CO₂ which resulted in fractional end-tidal CO₂ of 7% (Fet_CO2, HYP); and (3) breathing CO₂-enriched gas while on 10 cm H₂O nasal CPAP (HYP + CPAP), again at a Fet_CO2 of 7%. Each condition was maintained for 5 min and ordering of the conditions was randomized. HYP was induced by blending CO₂ with room air and delivering the mixture into the intake end of a nasal CPAP machine (Respironics, Monroeville, PA) which was modified as shown in Figure 1. The subject wore a nasal CPAP mask and tubing. During RA, the CPAP mask was disconnected from the tubing and the CPAP machine, thereby eliminating dead space. During HYP, the tubing and mask were connected to the machine without a CPAP valve in line and with the machine delivering air enriched with CO₂. During HYP + CPAP, a CPAP valve was placed in line between the machine and the subject, and the valve was adjusted to deliver 10 cm H₂O pressure. Fet_CO2 was maintained at 7% during HYP and HYP + CPAP by intermittent adjustments to the blender setting.

View larger version (15K): [in this window] [in a new window]   Figure 1.   Schematic representation of the experimental setup used to produce a constant FetCO2 with subjects off and on CPAP during the HYP and HYP + CPAP conditions, respectively.

Data Analysis

Three minutes of steady-state EMGgg activity in each of the conditions were analyzed. Swallows, sighs, and nonrespiratory tongue activity were excluded from analysis. In one subject, only 1 min of data could be analyzed owing to artifact in the level of baseline EMGgg activity. Because phasic activity was variably present in the subjects during CO₂ conditions only, EMGgg was analyzed as mean activity without regard to respiratory phase.

Statistical Analysis

EMGgg was compared across treatments and experimental conditions by two-way repeated measures analysis of variance (ANOVA). Significant differences between conditions were analyzed by a Student-Newman-Keuls multiple comparison test. All analyses of EMGgg were performed after natural log transformation (log_n) of the data to normalize the distribution. Statistical significance was assumed at p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Figure 2 illustrates EMGgg activity in one representative subject under conditions of RA, HYP, and HYP + CPAP during placebo and drug exposures. Panel A represents responses to placebo and panel B to paroxetine. After placebo administration, baseline EMGgg was minimal in each experimental condition. In the presence of paroxetine, an increase in mean tonic EMGgg activity is seen compared with placebo. Of note in this subject, an increase in phasic EMGgg activity during HYP while on paroxetine is seen. Although in most subjects this augmentation of phasic activity did occur, it was not seen consistently throughout the study. Application of nasal CPAP in the presence of hypercapnia (HYP + CPAP) resulted in no decrement in the drug-augmented EMGgg response.

View larger version (18K): [in this window] [in a new window]   Figure 2.   EMGgg activity in one representative subject under conditions of RA, HYP, and HYP + CPAP. Panel A shows response to placebo and panel B response to paroxetine. Note the increase in EMGgg activity in the presence of paroxetine. FetCO2 was maintained at 7% during HYP and HYP + CPAP conditions and CPAP (Pmask) was maintained at 10 cm H2O pressure during both placebo and paroxetine treatments.

Figure 3 shows log transformed EMGgg results for all 11 subjects. The 2-way repeated measures ANOVA revealed a drug and condition effect, with no significant interaction between the two. Independent of condition, paroxetine hydrochloride caused a significant increase in EMGgg activity as compared with placebo (p = 0.02). The Student-Newman- Keuls multiple comparisons test showed that independent of treatment, EMGgg during the conditions of HYP and HYP + CPAP was significantly greater than during RA (p = 0.006), and that HYP and HYP + CPAP were not significantly different from each other. The effects of paroxetine hydrochloride on EMGgg during RA, HYP, and HYP + CPAP for individual subjects are illustrated in Figure 4 and Table 2. Nine of the 11 subjects studied showed an increase (mean = 67.4% for all subjects) in EMGgg activity after paroxetine during RA. Two subjects did not show any increase in EMGgg activity with paroxetine during RA and HYP (Subjects 7 and 10). Overall paroxetine caused a 49.1% increase in EMGgg activity during HYP (Figure 4B) and a 23.6% increase during HYP + CPAP (Figure 4C).

View larger version (8K): [in this window] [in a new window]   Figure 3.   Natural log (logn) of EMGgg activity in 11 subjects expressed as mean ± SEM while on placebo (circles) and paroxetine (triangles). Independent of condition, paroxetine caused a significant increase in logn EMGgg activity (*p = 0.02, placebo versus paroxetine). Logn EMGgg during the conditions HYP and HYP + CPAP, independent of treatment, was significantly greater than during RA (p = 0.006).

View larger version (16K): [in this window] [in a new window]   Figure 4.   Group means (± SEM) and individual responses in EMGgg after placebo and paroxetine exposures during RA (A), HYP (B), and HYP + CPAP (C ) conditions. Paroxetine caused a significant increase in EMGgg in all conditions (*p = 0.02).

Figure 5 compares the individual and mean responses in EMGgg during HYP and HYP + CPAP conditions in the presence of paroxetine. Although small increases and decreases were seen in individual responses, augmentation of EMGgg by paroxetine persisted during the application of nasal CPAP (183% versus 182.2% of placebo, respectively), in keeping with our hypothesis.

View larger version (14K): [in this window] [in a new window]   Figure 5.   Individual and mean (± SEM) responses in EMGgg (expressed as percent of RA placebo) during exposure to CO2 on and off CPAP and in the presence of paroxetine. CPAP had no effect on the paroxetine-augmented EMGgg response.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The findings from this double-blind, placebo-controlled crossover study indicate that serotonergic drugs such as paroxetine hydrochloride are able to augment genioglossal muscle activity in normal subjects while awake. Paroxetine may mediate this response by increasing extracellular concentrations of serotonin, thereby activating postsynaptic receptors on the hypoglossal motor neuron pool (19). Kubin and coworkers (4) showed that in decerebrate, paralyzed, vagotomized and artificially ventilated cats, serotonin and its analogues microinjected into the hypoglossal motor nucleus increased tonic activity of the genioglossal branch of the hypoglossal nerve by over 200%. This increase in tonic activity was accompanied by a reduction in phasic responsiveness. This is consistent with the predominantly tonic EMGgg responses to paroxetine seen in our subjects.

Excitatory serotonergic neurons in the caudal raphe that project to UA motor nuclei are sleep state dependent, being highest during waking, with a progressive reduction in spontaneous activity occurring from non-rapid eye movement (non-REM) to rapid eye movement (REM) sleep (6). Serotonergic agents, such as paroxetine, might therefore maintain excitatory serotonergic inputs to the UA muscles during sleep.

Microinjections of serotonin into the hypoglossal motor nucleus resulted in attenuation of the carbachol-induced REM sleep state suppression of hypoglossal nerve activity in decerebrate cats (20). Maintaining increased concentrations of serotonin should attenuate the reduction in hypoglossal nerve activity during REM sleep and prevent REM-induced apneas. However, in clinical studies, use of SSRIs such as paroxetine and fluoxetine in the treatment of OSA results in a decrease in the number of apneas during non-REM sleep and not during REM sleep (7, 8). This observation may be due to a dependence of such agents on presynaptic serotonin release (21), which is essentially absent during REM sleep. Thus, specific receptor agonists that substitute for decreased endogenous serotonin might more effectively eliminate pharyngeal collapse, particularly during REM sleep.

The receptor subtypes that mediate the excitatory effect of serotonin continue to be studied; however, there is strong evidence that type 2 receptors are involved (4, 22, 23). Stimulation of type 2 receptors in the brainstem has been shown to facilitate hypoglossal nerve activity and thus UA function in adult cats (24), and messenger RNA (mRNA) coding for both 2A and 2C receptors have been detected within the hypoglossal motor nucleus (25).

The effect of paroxetine on genioglossal EMG activity may have resulted in part from a global increase in central respiratory output. This is unlikely because Veasey and her associates showed that systemic administration of serotonin receptor antagonists in an animal model of OSA had no effect on diaphragmatic EMG activity but caused a marked reduction in activity of another upper airway dilator muscle, the geniohyoid (26). In preliminary studies we noted an increase in EMGgg compared with placebo in the absence of similar increases in tidal volume or minute ventilation during CO₂ rebreathing in normal subjects given another serotonergic drug, fenfluramine (unpublished data).

In the present study we also examined the effect of positive airway pressure on the drug-induced activation of the muscle. CPAP application resulted in no overall decrement in paroxetine-augmented genioglossal activity, suggesting that serotonergic stimulation is independent of UA mechanoreflexes and may persist during sleep, when UA mechanoreflexes are diminished (27, 28). Our results are consistent with the findings of Douse and White who showed that serotonin had no effect on the pressure-elicited reflex activation of hypoglossal motor nerve activity despite causing an increase in tonic hypoglossal output (17).

There are some limitations in our study that need to be addressed. Our study was performed in awake normal volunteers and not in sleeping patients. Further studies are needed to demonstrate whether drug-induced augmentation of EMGgg occurs in patients with sleep apnea and whether it persists during sleep. We used an intraoral surface electrode to measure EMGgg rather than intramuscular wires, which are traditionally employed. Doble and colleagues (18) obtained EMGgg simultaneously from surface and intramuscular electrodes during quiet breathing and CO₂ rebreathing and showed similar patterns of muscle activity, using spectral analysis. They concluded that the surface electrode satisfactorily reflects bioelectrical activity of the genioglossus. Also, because of its fixed relationship to the tongue, the surface electrode allowed for comparisons between experimental sessions within a given subject without the need to normalize activity against a maximal value. Normalizing EMG activity in this way has recently been called into question (29). Our methods were validated by measuring EMG in the same subjects during quiet breathing on different days, which resulted in highly reproducible findings (Table 1). We are confident, therefore, that the increases in activity that we recorded after paroxetine exposure reflect accurately the genioglossus response to the drug.

In conclusion, this study demonstrates that serotonergic stimulation of EMGgg can be obtained in normal volunteers during wakefulness and that effects on EMGgg are independent of UA mechanoreflex activity. Because UA reflexes are diminished during sleep, serotonergic drugs may aid in the treatment of OSA through direct stimulation of UA muscles.