INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

RAPID-EYE-MOVEMENT (REM) sleep is a behavioral state identified by its tonic and phasic events. The tonic phenomena include postural muscle atonia and desynchronized electroencephalogram (EEG). In the cat, one of the striking phasic events is the pontogeniculooccipital (PGO) wave that is recorded with macroelectrodes from the lateral geniculate nuclei (LGN) during REM sleep and the transition from non-REM (NREM) sleep to REM sleep (6). Spontaneous PGO waves occur singly during the transition from NREM sleep to REM sleep but singly and in clusters during REM sleep. Another phasic event observed during REM sleep is the fractionation: a pause in diaphragm activity that lasts 20-100 ms (20). These phasic interruptions of respiration often also occur in clusters and can lead to prolonged apneas in the bulldog, an animal model of sleep apnea (15). Other phasic events of REM sleep include rapid eye movements and muscle twitches. PGO waves are closely correlated with eye movements in REM sleep (8), and muscle twitches and highly fractionated breaths often occur during bursts of PGO waves in REM sleep (11, 12). The common co-occurrence of phasic events may suggest that there is a common pacemaker for phasic events in REM sleep or that the mechanism underlying the PGO wave drives the occurrence of other phasic events, because it is the most frequent phasic event observed in REM sleep. We investigated this possibility by characterizing the relationship between elicited PGO waves and fractionations across states.

Besides occurring spontaneously in REM sleep, PGO waves and fractionations share several characteristics that suggest they may be generated by the same region of the brain. Lesions that eliminate PGO waves encompass the dorsolateral pontine tegmentum, where neurons projecting to the LGN fire in short bursts 5-10 ms before the appearance of a PGO wave in the LGN (19, 24, 25, 27). Other lesions in the dorsolateral pontine tegmentum abolish spontaneous fractionations in REM sleep (14). Studies have shown that auditory stimuli presented in different behavioral states can elicit fractionations and PGO waves, with characteristics suggesting great similarity to spontaneous waves, although these studies examined either PGO waves or fractionations alone (2, 5, 16, 28). In REM sleep, however, analysis of spontaneous PGO waves and fractionations has shown that the fractionation usually occurs 10-40 ms after onset of the PGO wave, based on cumulative histograms of diaphragm activity centered on PGO waves (20). It has been hypothesized that both PGO waves and fractionations involve activation of a startle reflex (5, 16); however, more recent work indicates significant differences between the classically defined acoustic startle reflex (ASR; involving a short-latency whole body jerk and increased muscle activity recorded from a variety of skeletal muscles in response to a 100- to 120-dB auditory stimulus) and PGO waves, which can be elicited by lower intensity stimuli that do not result in increased muscle activity or behavioral arousal (1, 10, 23, 28). PGO waves may more accurately be described as a central marker of alerting, important in preparing the brain for incoming sensory information, rather than as a reflexive response to a startling stimulus (3).

These recent observations would suggest that, although fractionations (16) and PGO waves (24) can both be produced by tonal stimuli, they may not be as tightly linked as previous studies would suggest. We therefore conducted experiments in which we applied auditory stimuli in different states of waking and sleep in cats and examined the data for the occurrence of both PGO waves and fractionations. We decided to study elicited waveforms because we could control the characteristics of the generating stimulus and we could measure precise latencies relative to the stimulus, as opposed to spontaneous waveforms, which are loosely associated with variable latencies relative to each other. We observed that tones could produce two types of fractionations: short-latency fractionation responses (SFRs), which occur within 10-60 ms of the stimulus, and long-latency fractionation responses (LFRs), which occur 60-120 ms after the tone. Our results do not support the existence of a common phasic-event generator, because PGO waves and both types of fractionations could be elicited independently in all states and response rate data suggest that they are independent events that may co-occur under some conditions.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Animal preparation. The four adult female cats (2.1-3.5 kg) used in this study were preanesthetized with acepromazine (0.1 mg/kg im) for sedation. Sterile surgery was performed under general anesthesia (halothane, 1-4% to effect). The cats were implanted with standard sleep recording electrodes, including electrooculogram (EOG), EEG, and nuchal electromyogram (EMG). Stainless steel screws were placed in the back of each bony orbit via the frontal sinus to record the EOG. The same type screws were positioned over the frontal cortex through the posterior wall of the frontal sinus to record the EEG. In addition, tripolar electrodes (Formvar-coated stainless steel wire, 0.25-mm diameter, 1-mm tip separation) were stereotaxically implanted in both LGN to record PGO waves (stereotaxic coordinates: anteroposterior +6.0, mediolateral ±10.0, dorsoventral +2.0; from Ref. 4). Electrodes for recording the nuchal EMG, diaphragm EMG, and electrocardiogram (ECG) were constructed of flexible fluorocarbon-coated stainless steel wires (Cooner Wire, Chatsworth, CA). A pair of these electrodes was sutured into the dorsal cervical neck muscles through a midline neck incision for recording the nuchal EMG. Three wire electrodes were sutured into both the right and left costal diaphragm through a small midline abdominal incision, and two separate wire electrodes were placed subcutaneously along the left chest wall to record the ECG. All of the wire electrodes were led subcutaneously to the headcap. All leads were soldered to a 26-pin connector, which was cemented to the skull with dental acrylic that covered all exposed wires.

Recording and signal processing procedures. The cats were allowed to recover for at least 1 wk after surgery before being studied. During the recording sessions, the cats were placed in a sound-attenuated chamber (1 m³) with food, water, a litter box, and comfortable bedding. The animals were connected to a lightweight, counterbalanced cable that has a built-in commutator to allow the cat to turn and move freely around the chamber. A Grass polygraph equipped with 75P11 amplifiers recorded the EOG, EEG, left and right LGN activity (all filtered to be within a 1- to 100-Hz band), and the three EMG signals (the neck and left and right diaphragm; band-pass filtered between 30 and 1,500 Hz) on paper. At the same time, a seven-channel digital recorder (Vetter, Rebersburg, PA; pulse code modulation recording adapter model 3000A) stored (for data archiving) the processed EEG, neck EMG, left and right LGN, left and right diaphragm EMG, and ECG (band-pass filtered at 30-1,500 Hz and amplified) on magnetic tape at a sampling rate of 88.2 kHz. A video camera connected to a videocassette recorder recorded the cat's behavior through a Plexiglas window on the front of the recording chamber.

Auditory stimulation protocol. Pure-tone stimuli (2,000 Hz, 90 dB, 20-ms duration, 0.1-ms rise-fall time) were generated by a frequency generator (model 200CD, Hewlett-Packard), routed through a Coulbourn-shaped rise-fall gate (model S84-04) to an amplifier (model PM645, Harman/Kardon) for volume control, and presented to the cat via two built-in speakers situated on either side of the chamber. A transistor- transistor logic (TTL) pulse produced by a customized DataWave Systems software routine was used to gate the tone and control its duration. On "sham" trials, no tones were delivered. All of the data were collected for a sham trial as for a tone trial, except that a tone signal was not gated to the speakers. Both the auditory stimuli and the sham signals were triggered during midinspiration as determined by the experimenter from the moving average of the diaphragm activity. The time of the tone or the sham signal was simultaneously marked on magnetic tape with a TTL pulse, on the paper record through a Grass direct-current driver amplifier, and on the computer record as a separate analog-to-digital channel (different channels were used to mark sham trials and tone trials on all records). The computer also collected, online, the raw diaphragm EMG and LGN activity, as well as signal markers for the tone and sham trials, for a period of 20 ms before the tone onset and 180 ms after the tone onset. All five channels (right and left LGN and raw diaphragm EMG and either a tone- or sham-trial marker) for each 200-ms trial were stored in a computer file for later analysis. Tones were alternated with sham trials at a minimum interval of 7.5 s so that the interval between tones was at least 15 s, thus minimizing habituation to the stimulus. Studies were done on separate days at least 7 days apart for each state: waking, NREM sleep, and REM sleep. The order of states tested was randomized for each cat. The stimuli were presented by the experimenter after the cat entered the appropriate state for that day (waking, NREM sleep, or REM sleep as determined by standard criteria in Ref. 17) and then stopped if the animal changed state. Stimuli were presented during all episodes of the appropriate state until a total of 50-150 tones were given. The tone trials were presented over a 4- to 8-h recording session, alternating with an equal number of sham trials.

Data analysis. The data for each 200-ms trial were individually examined on the computer screen for the occurrence of PGO waves and fractionations of diaphragm activity, as shown in Fig. 1. Any trials in which ECG artifact occurred during the data-acquisition window were excluded from analysis. A fractionation was defined as a pause in diaphragm activity (a decrease in amplitude equal to expiratory levels) with a duration of at least 10 ms. Peak amplitudes and both onset and peak latencies from tone onset for PGO waves were measured with cursors as were fractionation onset latencies and durations (for definitions, see Fig. 1). Two varieties of fractionation were observed: a SFR and a LFR. After individual PGO waves and fractionations were scored, the raw diaphragm data files from each trial (divided into tone trials or sham trials) were full-wave rectified, smoothed by using a 3.5-ms sliding window, and averaged for each state in each animal. The LGN recordings were also averaged without rectification or smoothing. To control for possible differences in recording quality as well as baseline and amplitude levels of the diaphragm EMG across different recording days, the averaged diaphragm data for each day were normalized in the following manner. The amplitude of the diaphragm activity at the onset of the tone was set to 100%, and 0.0-mV activity was considered to be 0% (Fig. 2). This data manipulation allowed the comparison of different animals and different recording days as percent change in diaphragm activity. The variables extracted from the averaged trials are shown in Fig. 2. We measured the latencies from tone onset to the first nadir (nadir 1), the highest point in the average (peak), and the second nadir (nadir 2). Amplitudes and durations also were measured for nadir 1, peak, and nadir 2. 

View larger version (48K): [in this window] [in a new window]   Fig. 1.   Individual example of data collected for 1 trial and variables measured. Top trace: left lateral geniculate nuclei (LGN) recording, from which pontogeniculoccipital (PGO) wave peak and onset latencies and PGO wave amplitude were obtained. Bottom trace: unrectified, left diaphragm EMG recording, from which short-latency fractionation response (SFR) and long-latency fractionation response (LFR) latencies and durations (DUR) were obtained. This example was taken from cat JB15 in rapid-eye-movement (REM) sleep. Mean values for the individual trials across cats are shown in Tables 1, 2, and 3.

View larger version (26K): [in this window] [in a new window]   Fig. 2.   Averaged activity for 97 trials in REM sleep with an elicited PGO wave and SFR (and/or LFR) for cat JB15 showing variables measured. Top trace: averaged left LGN recording (not rectified or smoothed). Bottom trace: averaged, rectified, smoothed (by using a 3.5-ms sliding window) left diaphragm EMG recording. Amplitude of diaphragm activity was normalized to a percent scale by setting amplitude at the time of tone onset to 100% (i.e., all amplitude values for the average were divided by amplitude at tone onset and multiplied by 100) and by setting 0 mV to 0%. Latencies (LAT), durations, and amplitudes (AMP) of nadir 1, peak, and nadir 2 measurements are indicated by interval arrows. Variables for each state were separately measured for each cat. PK, peak. Mean values across cats are shown in Table 4.

Histology. After completion of all studies, each animal was given an overdose of pentobarbital sodium and perfused intracardially with warm saline followed by 10% Formalin. Relevant portions of the brain were embedded in celloidin, cut on a microtome in 40-µm sections, and stained with cresyl violet. The slides then were examined for gliosis, indicating the location of the tripolar recording electrode in the LGN.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Responses. Several different responses to auditory stimuli were observed. A tone could elicit both a PGO wave and a fractionation, a PGO wave alone, a fractionation alone, or neither a PGO wave nor a fractionation (no response). If a fractionation was elicited, it could be a SFR with a 10- to 60-ms latency or a LFR having a >60-ms latency. SFRs frequently occurred without LFRs, but LFRs rarely occurred without SFRs. On occasion, particularly in REM sleep, tones produced both a SFR and a LFR. Examples of the different types of responses are shown in Fig. 3.

View larger version (57K): [in this window] [in a new window]   Fig. 3.   Individual examples of types of responses observed in tone trials. In A-F, top trace is left LGN recording and bottom trace is raw, unrectified, left diaphragm EMG recording. Movement in LGN baseline observed from 0 to 20 ms in A-E is stimulus artifact. Examples shown are no response (A), a PGO wave alone (B), a SFR alone (C), a SFR and a LFR without a PGO wave (D), a PGO wave with a SFR (E), and PGO wave with a SFR and a LFR (F).

Characteristics of fractionations (Table 2). SFRs were elicited with a mean latency of 23-25 ms (range across states) and a mean duration of 21-26 ms (range across states). These values did not vary significantly across states or response types (i.e., if elicited with or without a PGO wave). Elicited LFRs were also very consistent across states and response types, with a mean latency of 76-89 ms (range across states) and a mean duration of 18-27 ms (range across states). Again, these values were not significantly different across states or response types.

Characteristics of PGO waves (see Table 3). Elicited PGO waves had a significantly longer mean peak latency in REM sleep (onset 62.5 ms, peak 84.1 ms) than in NREM sleep (onset 40.6 ms, peak 65.1 ms, t = 8.73, df = 3, P < 0.05) or waking (onset 37.0 ms, peak 57.8 ms, t = 12.08, df = 3, P < 0.05). The mean peak latency in NREM sleep (65.1 ms) was also significantly longer than in waking (57.8 ms, t = 3.35, df = 3, P < 0.05). There were no significant differences across response types (i.e., whether or not a fractionation was elicited concurrently). Elicited PGO wave amplitudes did not vary significantly across states or response types.

Comparison of PGO waves and fractionations from trials in which both were elicited. The onset latency of the elicited SFR was consistently shorter than the onset or peak latency of the elicited PGO wave across states (Fig. 4). The difference between onset latencies for SFR and PGO waves was significant in NREM sleep (t = 3.69, df = 3, P < 0.05) and REM sleep (t = 9.50, df = 3, P < 0.05) but not during waking. In addition, the LFR latency was significantly longer than the PGO peak latency in NREM sleep (t = 5.07, df = 3, P < 0.05) and waking (t = 4.62, df = 3, P < 0.05).

View larger version (33K): [in this window] [in a new window]   Fig. 4.   Mean latencies to events in waking, non-REM (NREM) sleep, and REM sleep. Note consistency of SFR latency across states. Also note that PGO wave onset latency was consistently longer than SFR latency. PGO wave latencies increased significantly in REM sleep (2, 3). Increased PGO wave latency in REM sleep was not accompanied by any significant change in SFR or LFR latency (trend toward decreased LFR latency in REM sleep was not significant). LFR latency is consistently longer than PGO onset latency. Error bars, SD. * Significantly longer than SFR latency, P < 0.05. § Significantly longer than PGO peak latency, P < 0.05.

Positional analysis. Position of the animal had no significant effect on fractionation latency, duration, or proportion of trials with an elicited fractionation, as determined by using a mixed-model ANOVA to construct statistical tests. Because position had no significant effect on diaphragm measurements, all relevant trials were included in the calculations regardless of position.

Averaged trials. All trials with an elicited fractionation and PGO wave were included in the calculation of averages (Fig. 5). We analyzed the data in this manner to detect possible decrements in diaphragm activity that did not decrease to expiratory levels and thus would not be scored as SFRs or LFRs. This method also allowed us to characterize the postdepression rebound (peak) observed in the averaged data but not scored separately in the individual trial data. Nadir 1 represents the averaged SFRs, and nadir 2 represents the averaged LFRs (see Fig. 5). The mean latencies of nadir 1, peak, and nadir 2 were not significantly different across states, although there was a trend toward longer latencies in NREM sleep and REM sleep compared with waking (Table 4). The mean duration of nadir 1 in REM sleep (38.9 ms) was significantly longer than in waking (31.0 ms, t = 4.08, df = 3, P < 0.05). This probably reflects the trend toward increased mean SFR duration in REM sleep (25.8 ms) compared with waking (21.7 ms). The mean durations of peak and nadir 2 varied across states. There was a consistently larger mean change in amplitude in REM sleep than in NREM sleep or waking for nadir 1, peak, and nadir 2, but this trend was not significant. In general, there was no significant additional decrement in diaphragm activity that was not already reflected in the individual SFR and LFR scores.

View larger version (30K): [in this window] [in a new window]   Fig. 5.   Averaged left LGN (top traces) and rectified, smoothed diaphragm EMG (bottom traces) activity for cat JB15 in waking (top graphs), NREM sleep (middle graphs), and REM sleep (bottom graphs). Tone trial averages (left) include all trials that elicited a PGO wave and a SFR (and/or LFR). Sham trials (right) include all trials regardless of response (vast majority of sham trials result in no response). In tone trials, there is an initial poststimulus decline in diaphragm activity (nadir 1), followed by a rebound in activity above baseline (peak), followed by a second, smaller dip in diaphragm activity (nadir 2). All cats had a measurable nadir 1 in all states, but cat JB18 did not have a peak in REM sleep and 2 cats did not have a measurable nadir 2 in 2 states (cat JB18 in waking and REM sleep, cat JB7 in waking and NREM sleep). Mean values across cats are shown in Table 4. Number of trials that are included in each average is in top right of each graph.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The aim of this study was to characterize the relationship between two phasic events that occur spontaneously in REM sleep, specifically PGO waves and diaphragm fractionations. We tested the hypothesis that there is a common phasic-event generator in REM sleep that produces both PGO waves and fractionations or that the PGO wave generator drives the generation of other phasic events in REM sleep. We studied elicited events to determine the precise latencies of the events relative to a known stimulus and to see whether the correlation between elicited events was stronger in REM sleep than in waking or NREM sleep. We found that a 90-dB tone was effective in eliciting PGO waves and fractionations across states and seldom woke the animal. Two varieties of fractionations were observed: a SFR that was elicited at a high response rate across states (76-77% of trials) and a LFR that occurred significantly more often in REM sleep (32% of trials) than in the other states (<4% of trials). PGO waves and fractionations frequently occurred together, but each could also be elicited alone. SFRs were remarkably consistent in latency and duration across states, whereas PGO waves had a significantly increased latency in REM sleep. LFR latency and duration varied across states and displayed a trend toward decreased latency in REM sleep. Despite the decreased latency in REM sleep, on average the LFR onset was always later than PGO wave onset, and SFR onset was always earlier than PGO wave onset. The accentuation of phasic diaphragm suppression observed in REM sleep was demonstrated by the significantly greater reduction in averaged diaphragm activity as well as the significantly increased occurrence of LFRs compared with NREM sleep and waking. One of the animals did show a significant association between PGO waves and SFRs in waking and REM sleep, as well as between PGO waves and LFRs in REM sleep. Overall, however, our findings argue against the existence of a common REM sleep phasic-event generator and suggest that the PGO wave generator does not drive the production of fractionations.

PGO waves and fractionations as individual events. The PGO waves elicited in this study had similar characteristics to elicited and spontaneous PGO waves in previous studies (2, 5). Prior experiments had also demonstrated the increased PGO wave latency observed in REM sleep (2). The fractionations elicited in this study were similar to those elicited by Kline et al. (16), except that our cats had less variable responses to the auditory stimulus. All of our cats showed suppression of diaphragm activity in response to the tone across states, whereas in the study of Kline et al. some of the cats showed augmentation of diaphragm activity in NREM sleep and waking. The augmentation was probably due to the louder stimuli (100 vs. 90 dB) used in that study, because Wu et al. (29) have shown that, as decibel level increases, the EMG response is more likely to be augmented rather than suppressed. LFRs were not reported in the study of Kline et al. (16), but there is a visible dip in their figure (see Fig. 1 of Ref. 16) of the averaged diaphragm activity in REM sleep at ~80 ms after tone onset, which corresponds to the latency of LFRs in this study.

Relationship between PGO waves and SFRs. PGO waves and fractionations were often elicited together but could also be elicited independently. The fact that each event can occur alone argues for independent generating mechanisms rather than a common pacemaker. They may simply be generated by different neuronal populations that have similar auditory input and thus could co-occur independently in response to stimuli. A complicated and, we believe, less likely explanation is that PGO waves and fractionations are generated by the same neurons and then modulated differently in their independent output pathways.

Relationship between PGO waves and LFRs. LFRs, although different from SFRs in latency and response rates across states, also showed no significant association with PGO waves in most of the animals. The same animal that showed a significant association between PGO waves and SFRs also had a significant association between PGO waves and LFRs in REM sleep. When taken as a whole, however, the response-rate data support the conclusion that PGO waves and fractionations are independently generated in all states. The LFR, which is elicited primarily in REM sleep with a longer latency than the PGO wave, appears to match the spontaneous fractionation that occurs normally in REM sleep, because the spontaneous fractionation is also more common in REM sleep than in other states and occurs more frequently after the PGO wave than before it (20).

Possible mechanisms for the production of fractionations. Kohlmeier et al. (17) have found that auditory stimuli produce hyperpolarizing potentials in masseter motoneurons in carbachol-produced REM sleep-like atonia but not in the preceding anesthetized state. A similar mechanism may explain the augmentation of diaphragm inhibition in REM sleep demonstrated by the increased occurrence of LFRs. Another possible mechanism for the production of both short- and long-latency fractionations from a single stimulus is the presence of both fast- and slow-conducting axons in the pathway. Kohyama et al. (18) have shown that both fast- and slow-conducting reticulospinal neurons are involved in producing hindlimb muscle atonia elicited by electrical stimulation of the pons and medulla in decerebrate cats. Thus they found both an early and late component of hindlimb muscle tone suppression, the latencies of which are similar to those of the short- and long-latency fractionations observed in this study. We know that the phrenic motoneurons are receiving an auditory-induced short-latency inhibition in all states as well as a longer-latency inhibition primarily in REM sleep, but we cannot know the source(s) of that inhibition without further investigation of the pathways involved. Perhaps the inhibition of postural muscle tone present in REM sleep is able to suppress the phrenic motoneurons briefly when stimulated by an auditory tone. Perhaps the SFR is produced by a minor activation of the ASR pathway, not enough to cause awakening or a whole body response, which is the usual measure of startle. We can only speculate on the exact anatomic pathways involved in the generation of elicited PGO waves and fractionations as seen in this study, because we did not record or manipulate the areas that may be involved.