The relationship between behavioral state, discharge pattern, and discharge rate was investigated in 26 lateral geniculate nucleus (LGN) units recorded in cats in the dark during waking (W), synchronized sleep (S), and desynchronized sleep (D). A distinctive state-dependent discharge pattern was the presence of stereotyped bursts of 2-7 spikes that occurred in 63% of the units. These bursts were most frequent in S, much less frequent in D, and rarely occurred in W. Lack of association with discharge rate changes between states showed the bursting to be a true state-dependent phenomenon. A burst consisted of 2-7 spikes, with each successive interspike interval being longer than the preceding one; in the 200 ms prior to burst occurrence, discharge probability decreased markedly. This structure of burst organization suggested a model of generation wherein each burst was caused by a unitary event of varying intensity, perhaps a rebound following a hyperpolarization. Spectral and autocorrelational analyses showed bursts occurred rhythmically in three cells at a frequency of 3-4 Hz and in two cells at a frequency of 10-12 Hz, indicating a possible linkage with slow-wave generators. While the number of bursts in the various behavioral states was a state-dependent phenomena, other aspects of discharge pattern were shown to be rate dependent. To evaluate discharge pattern apart from the occurrence of bursts, a "primary event spike train" was formed; this consisted of individual spikes and the first spike of each burst. This analysis showed that, within S, the probability of burst occurrence was highest when the primary spike rate was low. Quantitative analyses showed that first-order pattern measures (the form of the interspike interval histogram, IH) were dependent on the mean interspike interval (ISI, the inverse of mean rate). This association explained 83-89% of the variance in a power series approximation of IH form. Joint interval histograms (JIH) were used to evaluate the signature of bursts and of the form of the primary spike train. As with interval histograms, the main features of the form of the primary spike JIH were dependent on the primary spike rate. Thus, we concluded that first- and second-order discharge patterns of primary events were rate dependent and not state dependent. Our data are compatible with a model where in the absence of retinal input, the frequency of LGN primary spikes over behavioral state changes is largely determined by brain stem reticular formation input.(ABSTRACT TRUNCATED AT 400 WORDS)

Pontogeniculooccipital (PGO) waves appeared almost simultaneously in both lateral geniculate nuclei (LGB), but in each case on had a larger amplitude and preceded the other by a few milliseconds. The larger, earlier wave is called the primary wave. Primary waves were found to appear with equal frequency in each LGB. During rapid eye movement sleep (REM sleep), LGB primary waves were ipsilateral to the direction of rapid eye movements. During REM sleep a group of cat midbrain neurons, which we call PGO burst cells, fired in stereotyped bursts at fixed latencies before ipsilateral primary waves, but they almost never fired bursts when the primary waves were contralateral. PGO burst neuron discharge also correlated with the direction of rapid eye movements during REM sleep. In wakefulness, PGO burst cells fired single spikes, not bursts, which had some correlation with LGB waves when averaged by computer. The results suggest that PGO burst cells are output elements in the PGO wave-generation system ad that PGO waves convey eye movement information to the sensory visual system in REM sleep. They also may have a role in the production of saccade-related waves in the visual system during wakefulness.

In vivo microdialysis was used to analyze the role of dorsal raphe nucleus (DRN) neurons in regulating the sleep-waking cycle. Measurements of extracellular serotonin (5-HT) were made in the DRN of freely moving adult cats before and during microdialysis perfusion of 8-hydroxy-2-(di-n-propylamino)tetralin (8-OH-DPAT), a selective 5-HT1A receptor agonist, in artificial CSF. Behavioral state alterations were measured by simultaneous polygraphic recordings. During waking and artificial CSF perfusion of probes histologically localized to the DRN, extracellular 5-HT was 4 fmol/7.5 micro L dialysate sample. With the addition of 8-OH-DPAT (10 microM in artificial CSF) to the perfusate, 5-HT levels in the same state decreased 50%, to 2 fmol/sample (p < 0.01), presumably through 5-HT1A autoreceptor-mediated inhibition of serotonergic neural activity. Concomitantly, this 8-OH-DPAT perfusion produced a short latency, threefold increase in rapid eye movement (REM) sleep, from 10 to 30% of the total recorded time (p < 0.05), whereas waking was not significantly affected. In contrast, and suggesting DRN specificity, 8-OH-DPAT delivery through a probe in the aqueduct did not increase REM sleep but rather tended to increase waking and decrease slow wave sleep. The data on REM sleep provide the first biochemically validated and direct evidence that suppression of DRN serotonergic activity increases REM sleep, and furnish a key complement to our laboratory's in vitro data indicating that mesopontine cholinergic neurons, a target of DRN projections, are inhibited by 5-HT. The 8-OH-DPAT-induced reduction of DRN 5-HT is consistent with the hypothesis that the concomitant REM sleep disinhibition is mediated by DRN serotonergic projections to mesopontine cholinergic neurons, which other data implicate in REM sleep production.

No abstract

A limit cycle mathematical model of the rapid-eye-movement (REM) sleep oscillator system has been developed from a structural model of interaction of populations of REM-on and REM-off neurons. The marked differences in latency, amplitude, and duration of the first REM sleep period seen with circadian variation and depressive pathology are modeled by beginning the REM oscillation at different initial points relative to the final position in the limit cycle. Beginning from a point that is graphically interior to the limit cycle produces a long-latency, short-duration, and less intense first REM period. Beginning from a point graphically exterior to the limit cycle produces a short-latency, long-duration, and more intense first REM period. In the model the determinant of whether the oscillation begins exterior or interior to the limit cycle is the time course of decay of the REM-off population discharge activity at sleep onset. When this time course is made to depend on circadian phase, the model produces a very close match to the empirically observed large shifts between the first and second REM periods in duration (often a 50% change) and intensity and also closely mimics the empirically observed shifts in REM latency as human sleep begins at different circadian phases. Although this variation in limit cycle entry accounts for the major changes in REM sleep over the night, the model also postulates a continuous but small circadian variation (of the order of +/- 5% change in REM parameters) acting throughout the course of a night's sleep. Because the model is derived from actual physiological data, rather than being a purely ad hoc or phenomenological construct, it offers the possibility of direct tests of its postulates through neurobiological studies in animals, by circadian phase-related manipulations of the sleep cycle, and through perturbations of the system in humans by the use of drugs. Indeed, an explicit phase-response curve of the system to cholinergic agonists has been developed; this will permit experimental tests of the model in both animals and humans.