INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A TRIPLE-RISK MODEL HAS BEEN proposed for the sudden infant death syndrome (SIDS) in which a sequence of events leading to death is triggered during sleep when a vulnerable infant is exposed to an exogenous stressor during a critical period of development (12). Among many possible exogenous stressors identified in epidemiological and physiological studies (17, 19, 42), we have been interested in the laryngeal chemoreflex (LCR). The LCR is elicited when fluid stimulates laryngeal mucosal receptors. The reflex response may involve swallowing, coughing, apnea, bradycardia, hypertension, redistribution of blood flow, and arousal from sleep (15, 16, 20). The manifestations of the LCR seem to evolve over the course of development: swallowing and apnea are prominent in preterm infants, swallowing remains in full-term neonates, but the duration of apnea declines during this period, and cough may emerge as a more prominent element of the response in adult animals (47). Arousal from sleep is common, but not universal, when the reflex is elicited; arousals tend to be less frequent during active sleep (36, 43). The respiratory disruption and apnea associated with the LCR have led to speculation that the LCR may contribute to the pathogenesis of the SIDS (10, 17, 41). Furthermore, prone positioning of infants reduced swallowing and enhanced the respiratory disruption associated with the LCR during active sleep (19).

A number of neurotransmitter receptor deficits have been identified in certain nuclei in the brain stems of infants who died of the SIDS (22, 38, 39). To understand the function of the neuroanatomic abnormalities identified in babies dying of SIDS, we turned to studies of neonatal piglets. We focused on the rostral ventral medulla (RVM) in these animals. We defined the RVM as a region within 1.2 mm of the ventral surface that extends rostrocaudally the length of the facial nucleus and mediolaterally ~1.5-4.0 mm from the midline. This area includes the retrotrapezoid nucleus, the parapyramidal region, and the juxtafacial parts of the nucleus paragigantocellularis lateralis. Thus many of the nuclei in which decreased neurotransmitter receptor binding was found in babies dying of the SIDS are included in the RVM, and the RVM in piglets may be homologous with parts of the arcuate nucleus in humans. Inhibition of neurons within the RVM after dialysis of 10 mM muscimol, a GABA_A agonist, reduced the ventilatory response to 5% inhaled CO₂ in neonatal piglets during wakefulness and non-rapid eye movement (NREM) sleep (5) and disrupted the architecture of sleep in these animals (8). In decerebrate piglets, muscimol dialysis in the RVM enhanced the respiratory inhibition associated with activation of baroreceptors (6), and cooling of the ventral medullary surface enhanced respiratory inhibition after superior laryngeal nerve stimulation in anesthetized piglets (26). The foregoing studies suggest that functional deficits in the RVM may reduce respiratory stability in that a response to a ventilatory stimulant (CO₂) was blunted, but a response to ventilatory inhibition [elevated blood pressure (BP)] was enhanced. In the present study, we tested the hypothesis that muscimol dialysis in the RVM would enhance the inhibitory action of the LCR on respiration in neonatal piglets. We elicited the LCR by instilling either 0.9% saline or water into the larynx during wakefulness, NREM sleep, and rapid eye movement (REM) sleep before and after muscimol dialysis in the RVM.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experiments were performed on 28 piglets ranging in age from 3 to 16 days (8.3 ± 0.7 days; means ± SE) with an average weight of 2.4 ± 0.1 kg (means ± SE). The Institutional Animal Care and Use Committee of Dartmouth College approved all surgery and experimental protocols.

Surgical instrumentation. Anesthesia was induced with isoflurane in O₂, and surgery was performed under sterile conditions. A femoral venous catheter was inserted, and each piglet was treated with intravenous antibiotics (20 mg/kg iv, cefazolin). A dual-lumen umbilical catheter (4-Fr, 40 cm; Vygon, Ecguon, France) was inserted into the femoral artery and advanced until the tip was in the abdominal aorta. A thermistor was placed subcutaneously lateral to the abdominal midline. Two wire electromyographic (EMG) electrodes were sewn into the diaphragm through a subcostal incision in the right upper quadrant of the abdomen to measure diaphragmatic EMG (EMGdia) activity. A second set of EMG wires was inserted into the genioglossus through a submental incision to measure genioglossal EMG (EMGgg) activity. A 2.7-mm-diameter catheter was placed in the trachea just below the cricoid cartilage (feeding tube, 8-Fr, 16 in., Kendall, Mansfield, MA) to record endotracheal pressure. This catheter was also used for tracheal fluid injections in a subset of animals. The wires and catheters were tunneled subcutaneously and exited the skin at the top of the skull.

Measurements. The animals were first studied ~24 h after surgery. Respiration was measured by using a barometric plethysmograph modified to allow continuous gas flow (40). A reference chamber with a slow leak was connected to the plethysmograph to minimize pressure fluctuations associated with changes in room pressure (2). Air flowing through the plethysmograph was heated (~38°C) and fully humidified. The average box temperature was between 25 and 26°C. Respiratory activity was derived from the pressure fluctuations in the box, which were measured with a model DP103 transducer (Validyne, Northridge, CA). All catheters and recording wires were passed through a sealed port in the side of the plethysmograph.

Protocol. Animals were studied for 1-4 days after surgery. One and one-half hours before the experiment, the plethysmograph was sealed to allow the temperature and humidity to stabilize. The plethysmograph was calibrated by sequential triplicate injections of 1, 2, 3, and 5 ml of air. During the calibrations, the piglet was brought into the laboratory setting and fed to increase the likelihood of sleep during the experiment. The piglet was placed prone in a sling in the box and connected to the monitoring equipment. The stylet of the microdialysis guide tube was removed and replaced with a microdialysis probe. Artificial cerebrospinal fluid (aCSF) was passed through the dialysis probe at 8.5 µl/min (this dialysate flow rate caused no volume transudation of fluid across the dialysis membrane). Before each experiment, one side of the piglet's nose was anesthetized with 1 ml of 2% lidocaine jelly. A double-lumen nasal catheter (4-Fr double-lumen PICC, Braun Medical) was passed into the larynx to the appropriate depth determined during surgery, and the catheter was tightly attached to the nose ring. Once the nasal catheter was in place, the piglet was placed in the box and acclimated to the box for ~1 h.

Data analysis: definition of the LCR. We measured the duration of the LCR as our primary index of the strength of the reflex. The duration of the LCR was determined from the respiratory tracings of the plethysmograph EMG, EMGgg, and EMGdia activity and fluctuations in endotracheal pressure. The onset of the LCR was taken from the onset of inspiratory activity of the breath during which the saline or water injection was made. The mixture of events that constitute the LCR may include swallowing, apnea, coughing, a startle, and arousal from sleep and movement of the entire head, and most of these events disrupt the regularity of the respiratory pattern. We defined the end of the LCR as the time at which a regular pattern of breathing was restored, and the onset of a regular respiratory pattern was defined by the occurrence of five uninterrupted breaths. This definition of LCR duration encompasses both apnea and irregular and obstructed breathing caused by swallows or coughs, but these events were not explicitly included in our definition of the LCR, except to the extent that these events prevented the restoration of regular breathing. Apneas were defined as the cessation of respiratory activity for a time greater than the duration of the two breaths that preceded the breath during which fluid was injected into the larynx.

Cardiorespiratory and sleep state scoring variables. Breath-to-breath tidal volume, respiratory frequency, and minute ventilation (E) were calculated from plethysmographic pressure fluctuations. EEG signals were resampled at 100 Hz and filtered with a bandwidth of 0.1-30 Hz. The EEG record was divided into 5-s epochs, and absolute spectral power density was computed over a frequency range of 0-50 Hz, as described previously (8). The power for each epoch was averaged over the delta (0.1-4.0 Hz), theta (5.0-9.0 Hz), or sigma (10.0-14.0 Hz) frequency bands. Periods of wakefulness, NREM sleep, and REM sleep were defined by comparing continuous plots of tidal volume, mean arterial pressure, heart rate, delta, theta, and sigma EEG power density, EOG movements, and eye openings. Wakefulness was defined by the presence of low-amplitude fast EEG activity, low delta power, and the presence of nuchal EMG activity. The piglets also had intermittent eye opening and gross body movements during wakefulness. NREM sleep was defined by periods of high delta power and high EEG amplitude. During NREM sleep, the piglet often showed signs of shivering, which completely disappeared during REM sleep. REM sleep periods were characterized by the presence of REM, nose and ear twitching, an irregular pattern of breathing, a drop of mean arterial BP, low delta power, a high-frequency EEG signal, and loss of nuchal EMG activity. Arousals were defined by marked sleep state transitions (NREM to wakefulness and REM to wakefulness) lasting 10 s. We did not define microarousals lasting <10 s because we saw them so infrequently; when arousal was part of the response to fluid injection, the arousal tended to be long lasting.

Neuroanatomy. At the conclusion of experiments, each piglet was killed with an injection of 1.5 ml/kg pentobarbital sodium followed by 5-10 ml of saturated potassium chloride administered intravenously. Microinjections of 20-50 µl of 1% potassium permanganate were made into the RVM through a broken microdialysis probe passed through the microdialysis guide tube to mark the location of the tip of the microdialysis probe and site of dialysis (the distribution of the permanganate was not used in any way to judge the anatomic distribution of muscimol dialysis). The permanganate formed an insoluble tar, which was not removed during subsequent tissue processing (44). The ventral surface of the brain was photographed to provide a permanent record of the site of muscimol microdialysis with respect to external brain stem landmarks. The external landmarks were correlated with the location in computer reconstructions and further analyses of cut sections. The brain stem was removed from the animal, placed in cryoembedding medium (Tissue-Tek OCT 458, Sakura Finetek, Torrance, CA), and frozen in isopentane at 70°C. Brain stems were sectioned (50 µm) in a cryostat at 18°C, and sections were mounted on gelatinized glass slides. Sections were fixed over night in 10% formalin in phosphate-buffered saline (pH 7.0) and stained with cresyl violet (1, 28).

Analysis and statistics. We studied 19 piglets during muscimol dialysis. We tried to study each piglet on multiple days, and we succeeded in studying four piglets on 3 days, seven piglets on 2 days, and the remainder on 1 day only. We stimulated the LCR >1,500 times in these animals. We treated each study day as a separate observation, but the responses of each piglet within each day were averaged within each treatment variable (sleep state and type of fluid injected before and after muscimol treatment). Thus we had 34 study days or observations on 19 piglets (we obtained identical statistical results when we averaged multiple days from each piglet and treated each piglet as an observation). The individual average values that made up an observation day were entered into a three-way repeated-measures ANOVA (Systat 9.0, Evanston, IL) in which the type of fluid (water or saline), the muscimol dose (0, 1, or 10 mM), and sleep state (wakefulness, NREM sleep, or REM sleep) were within-subject factors. When the ANOVA indicated that significant differences existed among treatment conditions, we performed multiple preplanned comparisons using the Bonferroni adjustment of P values. A similar statistical approach was used to analyze control studies in which the LCR was tested at the same times as in the muscimol studies, but no muscimol was added to the aCSF dialysate. We analyzed the effect of muscimol, the type of fluid used to stimulate the LCR, and sleep state (REM or NREM sleep) on the likelihood of arousal from sleep after stimulation of the LCR using a series of ² tests. We also compared the effect of injection of fluid above and below the larynx in three piglets. We used an ANOVA in which route of injection, volume injected, and the presence or absence of muscimol in the dialysate were between-subject variables. Finally, we tried to correlate the location of the dialysis within the brain stem with the magnitude of the change in LCR duration after muscimol dialysis. We did this using cluster analysis (K-means procedure, Systat 9.0) in which the percent change in the LCR was associated with particular sets of rostrocaudal, dorsoventral, and mediolateral coordinates that described the location of the dialysis probe within the brain stem (6). Values are expressed as means ± SE, and the criterion for statistical significance was set at P  0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Characterization of the LCR. An example of the LCR, which demonstrates the multiple facets of the reflex response, is shown in Fig. 1. Swallowing and coughing occurred frequently, and apneas, when they occurred, often followed a period of swallowing or coughing rather than being at the onset of the LCR. We did not see consistent bradycardia during the apneas after laryngeal stimulation of the LCR. Although not shown well in Fig. 1, the LCR often had a stuttering appearance in which a few normal-appearing respiratory efforts occurred before further swallowing, coughing, or apnea occurred. For this reason, we required five regular breaths to define the termination of the LCR. The example also demonstrates the typical pattern of EEG arousal. We did not see coughing without prior arousal, just as others have reported previously (43).

View larger version (42K): [in this window] [in a new window]   Fig. 1.   An example of the laryngeal chemoreflex (LCR) elicited by injecting 0.05 ml of water (first vertical dashed line) during non-rapid eye movement (NREM) sleep after 1 mM muscimol dialysis. Respiration (Respir; obtained from the plethysmograph), genioglossal electromyographic (EMG; EMGgg), diaphragmatic EMG (EMGdia), tracheal pressure (Ptrach), blood pressure (BP), EEG, electrooculographic (EOG), and nuchal EMG (EMGnu) activity are shown in the tracings. Note the occurrence of a swallow, which was associated with a characteristic negative Ptrach deflection and a burst of EMGgg activity that interrupted the EMGdia. Coughing was detected by the massive increase in EMGdia activity that preceded forceful expiratory activity, which was indicated by the increase in Ptrach. Apneas, defined as periods without breathing that lasted longer than the last 2 breaths that preceded the injection of fluid to stimulate the LCR, were also associated with cessation of EMGdia activity and respiratory pressure fluctuations in the Ptrach record. Arousals were identified from the occurrence of body movements, opening of the eyes, increased EMGnu activity, and increased fast activity in the EEG signal, in this case from high amplitude and low frequency to low amplitude and fast frequency (from NREM sleep to wakefulness). The LCR duration was defined as the time from the beginning of the first breath on injection of fluid until 5 regular consecutive breaths occurred. The 2 vertical dashed lines mark the length of the LCR.

Effect of muscimol on LCR duration during wakefulness and sleep. The effects of 1 mM muscimol dialysis and 10 mM muscimol dialysis on the LCR duration were not different, and the results from these two muscimol doses were combined. We were unable to obtain data during REM sleep during control or test conditions on 12 observation days, and the repeated-measures ANOVA is intolerant of missing data. Therefore, we excluded REM sleep in our initial analysis and compared wakefulness and NREM sleep. We had complete data for this analysis from 30 study days. Muscimol prolonged the LCR (P < 0.05), and the prolongation was significantly greater during NREM sleep than during wakefulness (P < 0.05). The LCR was longer when elicited by water compared with saline (P < 0.001), and this effect was also significantly greater during NREM sleep compared with wakefulness (P < 0.001). We studied piglets ranging in age from 3 to 16 days, and we included age as a covariate in the ANOVA. We found no evidence that the age of the piglet modified the response of the LCR to muscimol, to the type of fluid injected, or to the sleep state. Thus inhibition of small regions within the RVM prolonged the LCR during NREM sleep, and water was a more effective stimulus than saline.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   LCR durations after either saline (A) or water injections (B) into the larynx during wakefulness, NREM sleep, and rapid eye movement (REM) sleep before () and after muscimol (+) are shown. The statistical results were taken from a separate analysis of the response to saline and the response to water infusion. Values are means ± SE. None of the comparisons was significant among the test conditions after saline injection into the larynx. * LCR was significantly longer during REM sleep compared with wakefulness, P < 0.05. ** LCR was prolonged after water injection after muscimol dialysis, P < 0.001.

Anatomic locations of dialysis probes. The anatomic locations of the tips of the dialysis probes used to administer muscimol are shown in Fig. 3 (triangles, 1 mM muscimol; circles, 10 mM muscimol dialysis; whether the symbols are open or solid reflects the magnitude of the LCR, see below). On the left side of Fig. 3, the location of each dialysis probe was projected onto the ventral surface of a piglet's brain stem. The grid reflects the approximate position of the facial nucleus. On the right side of Fig. 3, probe locations were placed on hemisections of the medulla at the appropriate mediolateral, dorsoventral, and rostrocaudal location. In previous studies using 10 mM fluorescein to mimic the distribution of muscimol, the region affected by dialysis for 10-20 min was ellipsoidal and had a volume of ~5.8-6.3 µl (4, 33). The volume of distribution of 1 mM muscimol would be substantially smaller. We had hoped that, in some probe locations, the LCR would be prolonged, but we expected that some of the probe locations would fall outside the RVM and not affect the LCR duration. However, most of the probe locations were within the RVM. Nonetheless, some of the piglets demonstrated much greater prolongation of the LCR than others. To determine whether the location within the RVM correlated with the magnitude of LCR prolongation, we calculated the average percent change in the LCR for each piglet across all sleep states (wake and NREM and REM sleep) after stimulation with distilled water. In 15 piglets, the LCR was prolonged, and the average prolongation was 53 ± 6% (means ± SE; range 13-88%; open symbols), and, in four piglets, the LCR was shortened by 10 ± 5% (range 1-22%; solid symbols). We performed a cluster analysis to determine whether the magnitude of the LCR response was correlated with the location of the dialysis probe, and we found no evidence that any region within the volume of the brain stem that we studied was either especially effective or ineffective in modifying the LCR duration after muscimol dialysis.

View larger version (83K): [in this window] [in a new window]   Fig. 3.   Left: anatomic locations of 26 microdialysis probes are indicated on a grid, reflecting the location of the facial nucleus, projected on to a photograph of the ventral surface of a piglet brain stem. Right: hemisections (A, most rostral, to E, most caudal) are shown with the appropriate locations of the dialysis probes. Circles, dialysis sites with 10 mM muscimol; triangles, dialysis sites with 1 mM muscimol; solid symbols, animal was a "responder"; open symbols, animal was a "nonresponder"; diamonds, locations of the probes in piglets studied in time control experiments and of these, 3 piglets, indicated by X mark, also received injections above and below the larynx; 7N, facial nerve; SO, superior olive; TB, trapezoid body; VII, facial nucleus; RP, raphé pallidus; X, vagal motor nucleus; IO, inferior olive; XII, hypoglossal motor nucleus; VIII, auditory nucleus; NTS, nucleus tractus solitarius.

LCR response during time control studies. Muscimol is a long-acting drug, and we added muscimol consistently as the second treatment in the foregoing experiments. To make sure that the passage of time alone was not responsible for the prolongation of the LCR that we found after muscimol treatment, we conducted control studies in nine piglets. The timing of these control studies was identical to the muscimol studies, but, at the time muscimol would have been added to the dialysate, we simply continued the dialysis with aCSF. A three-way ANOVA identical to the analysis of test data was used, and the results of this analysis are shown in Fig. 4. There was no effect of time in these studies. As in the previous set of studies, water was a more effective stimulus than saline (P = 0.05), and there was a significant main effect of sleep (P < 0.001). The LCR was longer during REM sleep than NREM or wakefulness, but none of these specific comparisons among sleep states was significant once the P values were adjusted for multiple comparisons. Thus there was no evidence that time alone altered the LCR response over the duration of our experiments. The locations of the dialysis probes used in these control experiments are also shown in Fig. 3 (diamonds and X marks). The probe locations were, in general, more caudal than some of the probes used in the muscimol studies, but the control locations were still within the RVM.

View larger version (16K): [in this window] [in a new window]   Fig. 4.   Time control studies of the LCR are shown. The LCR duration was measured after either saline (A) or water (B) injection into the larynx during wakefulness, NREM sleep, and REM sleep. Artificial cerebrospinal fluid was dialyzed continuously, and measurements were made during the usual control period (time 1) and at the time muscimol would have been given (artificial cerebrospinal fluid; time 2). Values are means ± SE. Water injection prolonged the LCR significantly more than saline injection (P = 0.05 comparing pooled data from A with pooled data from B). There was a significant effect of sleep state indicating that the LCR got progressively longer as the piglets moved from wakefulness to NREM sleep to REM sleep, but the specific comparisons were not significant once the P values were adjusted for multiple comparisons.

Arousal responses before and after muscimol dialysis. Arousal forms an important part of the LCR and may have some survival advantage. Therefore, we used a series of ² tests to assess the effect of type of fluid injected, sleep state (NREM vs. REM sleep; data from wakefulness were excluded), and the presence or absence of muscimol dialysis on the likelihood of arousal after stimulation of the LCR. Because dialysis of 10 mM muscimol into the RVM disrupts normal sleep architecture and reduces the occurrence of REM sleep (8), we performed this arousal analysis only on data obtained before or after dialysis with 1 mM muscimol. We performed a series of stratified multiway comparisons using the Mantel-Haenszel test, but only the main effects of the treatments that we studied revealed significant differences. As a result, the simpler ² tests on the pooled data from all of the arousals elicited are presented in Table 3. Not surprisingly, given the apparent greater effectiveness of water compared with saline as a stimulus for the LCR, arousal was significantly more likely after injecting water into the larynx than saline (Table 3; P < 0.001). Arousal was more likely from NREM sleep than REM sleep (Table 3; P < 0.05). However, 1 mM muscimol had no effect on the likelihood of arousal (Table 3).

Response to sublaryngeal stimulation of the LCR. Apnea is a common feature of the LCR in studies of anesthetized animals (23, 25), but uncommon in unanesthetized sleeping piglets (36). The increase in the occurrence and duration of apnea may be attributed in part to anesthesia (37) but the route of administration also differs among these studies. To test the hypothesis that a sublaryngeal route of administration of fluid might enhance the likelihood or duration of apneas, we performed further studies in three of the piglets used in the time control studies. We administered saline or distilled water, either through the nasal catheter or through the catheter used to measure tracheal pressure, which was 2-3 cm below the larynx. An example of the response to sublaryngeal administration of 0.15 ml of distilled water during NREM sleep in the control period before muscimol was administered is shown in Fig. 5. Note the immediate onset of apnea and bradycardia late in the first apnea. The respiratory disruption was prolonged, and repetitive apneas were punctuated by coughing, swallowing, and brief bursts of respiratory effort.

View larger version (43K): [in this window] [in a new window]   Fig. 5.   An example of the LCR response to tracheal injection of 0.15 ml of water during NREM sleep is shown. Note the abrupt onset of apnea (indicated by A) and the prolonged episodes of coughing and swallowing interspersed with brief respiratory efforts. Bradycardia was apparent in the first apnea, but only late in the apnea and not in the subsequent shorter apneas. There was no recording of Ptrach because the tracheal catheter was used to administer water or saline into the trachea. The 2 vertical dashed lines mark the length of the LCR, and the arrow marks the onset of arousal.

View larger version (16K): [in this window] [in a new window]   Fig. 6.   Effects on the LCR duration of injecting water into the larynx or trachea are shown. Small-volume injections (0.05-0.075 ml) and large-volume injections (0.10-0.15 ml) were made in both locations before (M) and after (M+) 1 mM muscimol dialysis. Tracheal injections prolonged the LCR more effectively than laryngeal injections. This effect was most apparent when large-volume injections were made into both locations before muscimol was given and when small volumes were given after muscimol dialysis (* P < 0.05 both). The volume injected into the larynx, over the range that we tested, did not affect the LCR duration. Values are means ± SE.

View larger version (17K): [in this window] [in a new window]   Fig. 7.   LCR responses were expressed in terms of the length of the first breath (in s) immediately after the water injection. Values are means ± SE. The horizontal solid line reflects the average length of 2 breaths, i.e., the apnea threshold, and the horizontal dashed lines represent the 95% confidence intervals. When the length of the first breath exceeded the dashed line, an apnea occurred by our definition. Injections given into the larynx rarely caused an initial apnea, regardless of the volume injected, whereas the majority of injections through the trachea caused apnea as the first manifestation of the LCR. This effect was most apparent when large-volume injections were made before muscimol was given and when small volumes were given after muscimol dialysis (* P < 0.05 both).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We tested the hypothesis that inhibition of small regions of the RVM after muscimol dialysis would prolong the LCR. In general, the results confirmed this hypothesis. In addition, we confirmed previous studies in which water was a more effective stimulus of the LCR than saline, and the potency of water as a stimulus was greater in NREM sleep than in wakefulness and greater still in REM sleep. Apnea was a more prominent feature of the LCR after tracheal injection than after laryngeal injection. The modulatory effect of inhibition of neurons within the RVM is a novel finding that is probably relevant to the pathogenesis of the SIDS.

Stratification of the LCR response. The manifestations of the LCR vary as a function of animal age (47) and appear to vary in neonates, depending on the type and location of the stimulus and on the sleep state of the newborn. Some aspects of the LCR are conservative, in the original sense of the word. Apnea, bradycardia, and redistribution of blood flow conserve limited oxygen reserves without removing or reversing the inciting stimulus. Other aspects of the LCR seem more concerned with removing the offending stimulus and restoring airway patency and function. Swallowing and coughing, for example, clear the airway, and arousal may be necessary to permit coughing (43). However, the conservative and restorative aspects of the reflex are, to some extent, mutually exclusive; apnea precludes coughing, for example. Therefore, the respiratory control system seems to perform a triage in which restorative or conservative aspects of the reflex function sequentially, or one aspect is emphasized more than the other. The choice of restorative vs. conservative responses seems to vary as a function of the strength of the stimulus. We found that saline was relatively ineffective as a stimulus. Swallowing and coughing were elicited, but the respiratory disruption was mild, and apnea was less common after saline stimulation. Water in the larynx was more potent, but most potent was water injected below the larynx. We saw significant apnea and bradycardia after tracheal instillation, although saline was still less effective than water. This pattern suggests to us that restorative aspects of the LCR are the first line of defense, but, to the extent that they are ineffective, conservative aspects of the reflex become more prominent. The graded response of the LCR may arise from the persistence of the stimulus as restorative mechanisms fail and conservative responses develop. This stratification of responses is consistent with the observation that apneas often followed swallowing when injections were made in the larynx. The early onset of prolonged apneas after tracheal stimulation indicates that conservative responses need not follow failed restorative mechanisms. The immediate onset of apneas raises the possibility of two classes of receptors in which laryngeal stimulation elicits the restorative elements of the LCR, but tracheal stimulation fosters the occurrence of apnea. To the extent that restorative airway clearance mechanisms fail, laryngeal stimulation might lead to tracheal stimulation as liquid leaks from the larynx into the trachea. The failure of even large volumes of water injected above the larynx to elicit significant apnea suggests that a large area of stimulation alone may not initiate apnea and favors the hypothesis that tracheal receptors may have some specificity for the apneic response. Finally, we saw no consistent bradycardia accompanying the apneas after laryngeal stimulation. Bradycardia was apparent only during apneas and only when the apnea duration was prolonged after tracheal stimulation of the LCR. A similar lack of consistent bradycardia was reported in normoxic infants and preterm infants who, nonetheless, developed significant apnea after water was infused into the hypopharynx (41, 51). The bradycardia may not be part of the LCR, but it may develop during longer apneas as a result of hypoxia and carotid body stimulation (51).

Effect of sleep state on the LCR and arousals. Sleep modified the duration of the LCR. The LCR was generally longest during REM sleep, especially when water was used to stimulate the LCR. The LCR was of intermediate duration during NREM sleep and shortest during wakefulness. E was also least in REM sleep, intermediate in NREM sleep, and greatest during wakefulness. Thus some of the prolongation of the LCR may have been due to the sleep state-dependent reduction in ventilatory drive. In general, there is an inverse relationship between the duration of the LCR and respiratory drive in unanesthetized animals. For example, hyperoxia prolongs the LCR in awake and sleeping piglets (49). Others have also noted the decline in respiratory drive associated with different sleep states and the greater severity of the LCR as respiratory drive diminishes in newborn lambs (29). Stimulation of muscle afferents, which stimulate ventilation, also shortens the LCR in lambs (14). The effect of hypoxia on the LCR is less clear cut. Hypoxia may prolong the LCR in anesthetized piglets and exacerbates the apnea and bradycardia associated with the reflex (23), but other studies, also in piglets, suggest that hypoxia shortened apnea duration (52). Hypoxia shortens the apneic response to electrical stimulation of the superior laryngeal nerve in decerebrate piglets (7), but it prolongs the apnea associated with laryngeal infusion of water in infants (51). Thus, whereas it is appealingly simple to suggest that the duration of the LCR is inversely related to the level of respiratory drive (and some data, particularly during sleep, support this concept), the reality is more complicated, especially when one considerers the effects of hypoxia.

Modulation of the LCR by dialysis of muscimol in the RVM. Dialysis of muscimol prolonged the LCR when water was used to stimulate the reflex, and the prolongation was greatest in REM sleep. These findings raise two issues: what is the specific mechanism whereby muscimol modifies the LCR, and, in the larger scheme of the structure and organization of the brain stem, what is the function of the RVM with respect to cardiovascular and respiratory reflex responses? The LCR could be prolonged because the clearance mechanisms are less effective. Muscimol might modify the muscular function of swallowing or coughing and reduce the efficiency of these processes. Our EMG and tracheal pressure recordings are too crude to address this issue; such an analysis would require something like the radiological assessment used by others (31). Furthermore, muscimol seemed to delay the onset of the clearance mechanisms. The onset of swallowing, apnea, coughing, and arousal were all delayed after muscimol in our subsidiary analysis of the elements of the LCR (Table 1). The reflex seemed to occur in slow motion. Either of these mechanisms, delayed onset or reduced clearance efficiency, might prolong the reflex by slowing clearance of the substances stimulating the LCR. Furthermore, tract tracing studies demonstrated that connections exist between the RVM and the motoneurons of the laryngeal musculature that might plausibly permit an interaction between the RVM and swallowing mechanisms (50). On the other hand, the progressive prolongation of the reflex from wakefulness to NREM sleep to REM sleep might reflect the state-related reduction in the respiratory drive to breathe. The sleep state-related changes in ventilation are also correlated with sleep state-related changes in CO₂ chemosensitivity (5): the ventilatory response to CO₂ is least during REM sleep, intermediate in NREM sleep, and slightly greater during wakefulness. Dialysis of 1 mM muscimol reduced the ventilatory response to CO₂ in a pattern identical to the effect of 10 mM muscimol (L. van der Velde, J. Roberts, A. K. Curran, R. A. Darnall, D. Bartlett, Jr., and J. C. Leiter, unpublished observations). Furthermore, the duration of the LCR was reduced in anesthetized humans (34) and anesthetized piglets (24) during exposure to hypercapnia, and hypercapnia blunted the inhibition of ventilation associated with superior laryngeal nerve stimulation in anesthetized piglets (26). Thus the mechanism of action of muscimol to prolong the LCR may be correlated with reduced sensitivity to CO₂. Finally, reduced efficiency of LCR clearance mechanisms and reduced ventilatory drive are not mutually exclusive, and both of these mechanisms might contribute to prolongation of the LCR.

Variation of the LCR among and within piglets. Among the 19 animals receiving muscimol, the LCR was prolonged in 15 "responders," but the LCR was slightly shorter in four piglets. One of the nonresponders received 10 µM muscimol, and three received 1 µM muscimol. The lack of a consistent dose response in either the responders or nonresponders was a little surprising, but it suggests to us that the proximity of the dialysis probe with respect to the neurons is more important than the drug dose, given the relatively small volume of distribution of the drug. This interpretation is also consistent with the cluster analysis of anatomic locations that we performed. Nonresponders and responders did not differ with respect to neuroanatomic coordinates. Thus we believe that neurons that modulate the LCR (and other reflexes) are not distributed homogenously but that they exist in small clusters spread heterogeneously throughout the RVM (6).

Implications for SIDS. The triple-risk model for the pathogenesis of SIDS states that three factors contribute to the death of each infant: an underlying vulnerability, a critical period of homeostatic control, and an exogenous stressor (12). We studied the LCR, an exogenous stressor that inhibits ventilation, in neonatal piglets with an experimentally induced vulnerability (inhibition of the RVM by muscimol dialysis). The results of the study are consistent with the triple-risk model of SIDS in that inhibition of the RVM prolonged the LCR and enhanced the disruption of a stable respiratory pattern. The LCR seems particularly relevant to SIDS (19, 45, 46). Esophageal reflux may stimulate the reflex, and the LCR may be less effective in the prone position (swallowing was reduced and respiratory disruption was enhanced when the LCR was elicited during active sleep in the prone sleeping position; Ref. 19), which may tie the epidemiology of sleeping position to this particular exogenous stressor. Furthermore, the results of the present study provide further evidence that neurons within the RVM sustain and stabilize respiratory activity, and to the extent that the neurons are absent or deficient in some function, the ventilatory responses to inhibitory stimuli will be enhanced, and responses to excitatory stimuli will be blunted.