Coalescence of Sleep Rhythms and Their Chronology in Corticothalamic Networks

Transcription

1 Sleep Research Online 1(1): 1-10, Printed in the USA. All rights reserved X 1998 WebSciences Coalescence of Sleep Rhythms and Their Chronology in Corticothalamic Networks Mircea Steriade and Florin Amzica Laboratoire de Neurophysiologie, Faculté de Médicine, Université Laval, Quebec, Canada The cellular substrates of sleep oscillations have recently been investigated by means of multi-site, intracellular and extracellular recordings under anesthesia, and these data have been validated during natural sleep in cats and humans. Although various rhythms occurring during the state of resting sleep (spindle, 7-14 Hz; delta, 1-4 Hz; and slow oscillation, <1 Hz) are conventionally described by using their different frequencies, they are coalesced within complex wave-sequences due to the synchronizing power of the cortically generated slow oscillation (main peak around 0.7 Hz). In intracellular recordings from anesthetized animals, the slow oscillation is characterized by a biphasic sequence consisting of a prolonged hyperpolarization and depolarization. Basically similar patterns are observed by means of extracellular discharges and/or field potentials in naturally sleeping animals and humans. The depolarizing component of the slow oscillation is transferred to the thalamus where it contributes to the synchronization of spindles over widespread territories. The association between the depolarizing component of the slow oscillation and the subsequent sequence of spindle waves forms what is termed the K- complex. The slow oscillation also groups cortically generated delta waves. At variance with previous assumptions that the brain lies for the most part in the dark and a global inhibition occurs in resting sleep, cortical cells are quite active in this behavioral state. This unexpectedly rich activity raises the possibility that, during sleep, the brain is occupied to specify/reorganize circuits and to consolidate memory traces acquired during wakefulness. CURRENT CLAIM: During quiescent sleep, low- and high-frequency thalamic and cortical rhythms are grouped into complex wave-sequences due to the depolarizing component of a cortically generated slow oscillation. This paper is an attempt at revising the current thinking on the generation and synchronization of various oscillations that define the state of resting (non-rem) sleep at the EEG level. We present three main points. (A) A novel slow oscillation, described in intracellular recordings from cortical and thalamic neurons (Steriade et al., 1993a, 1993c, 1993d), has the virtue of grouping other sleep rhythms, spindles and delta, into complex wave-sequences. The slow oscillation has a frequency of about 0.6 to 1 Hz in ketamine-anesthetized as well as naturally sleeping animals (Steriade et al., 1996a, 1996b) and humans (Steriade et al., 1993c; Achermann and Borbély, 1997; Amzica and Steriade, 1997a). Instead of considering different sleep oscillatory types as generated within isolated networks, we envision the cerebral cortex and thalamus as a unified oscillatory machine in which the depolarizing component of the cortically generated slow oscillation drives thalamic reticular (RE) and thalamocortical (TC) cells to produce spindles (7-14 Hz) and a clock-like component of delta waves (1-4 Hz). The generation of slow oscillation within the neocortex has been demonstrated by its survival in cortex after thalamectomy (Steriade et al., 1993d), its disruption by disconnection of intracortical synaptic linkages (Amzica and Steriade, 1995b), and its absence in the thalamus of decorticated animals (Timofeev and Steriade, 1996). However, in intact animals the slow oscillation is reflected in the thalamus by both RE and TC cells. This contributes to the grouping of thalamically generated oscillations (spindles and clock-like delta). (B) The combination of the excitatory component of the slow oscillation with spindles leads to the appearance of sleep K-complexes in both cats and humans (Amzica and Steriade, 1997a, 1997b). (C) The orderly appearance of various rhythms throughout the state of resting sleep, under the umbrella of the slow oscillation, is associated with a progressive increase in the corticothalamic coherence of sleep rhythms. Why is it important to study the neuronal substrates underlying spontaneous brain rhythms even if their functional significance is far from being elucidated? Although different EEG oscillatory types were described by British and Eastern European investigators more than a century ago, their cellular bases have only been revealed during the past 10 to 15 years. Since the mid-1980s, the apparent chaos of EEG waves has been reduced to a few basic cellular operations that shed light on the origin and mechanisms generating various EEG graphoelements. Our reductionistic attempt provides explanations of the mechanisms underlying brain rhythms and may ultimately shed light on the functional role played by these oscillations. To give just one example: some still describe different types of EEG sleep spindles, with lower or higher frequencies, while intracellular data allow us to understand that these supposedly different types are attributable to a single event, namely, the duration of hyperpolarization-rebound sequence in TC neurons. The hyperpolarization de-inactivates a low-threshold Ca 2+ -mediated current, underlying postinihibitory rebound spike-bursts that are transferred to cortical areas (Steriade and Llinás, 1988). If the hyperpolarization is of about 70 msec, the Correspondence: Professor M. Steriade, M.D., D.Sc., Laboratoire de Neurophysiologie, Faculté de Médicine, Université Laval, Quebec, Canada GIK 7P4, Tel: , Fax: , mircea.steriade@phs.ulaval.ca.

2 2 STERIADE AND AMZICA spindle frequency is about Hz; if the hyperpolarization is longer because its progenitors, GABAergic RE neurons, fire longer spike-bursts, the frequency is lower. Understanding that the spindle oscillation is due to rhythmic inhibitory postsynaptic potentials (IPSPs) generated by RE neurons (Steriade et al., 1985; Bal et al., 1995) may explain one of their functional roles, which is the blockade of synaptic transmission through the thalamus (Steriade et al., 1969; Timofeev et al., 1996), thus deafferenting the cortex from the outside world and allowing a peaceful sleep. Similar examples may be taken from the study of other brain oscillations defining resting sleep. The aim of the present paper is to reveal the cellular substrates of sleep oscillations and to propose some avenues to understand their functions. METHODS Data reported in this paper result from intracellular recordings (in some cases dual simultaneous impalements of cortical neurons or cortical and thalamic neurons) in conjunction with multi-site extracellular recordings in acutely prepared cats under different types of anesthesia (mainly ketamine-xylazine) or from multi-site extracellular recordings during natural wake and sleep states in chronically implanted cats (see technical details in Steriade et al., 1996a, 1996b). The intrinsic properties and input-output organization of different cortical and thalamic neuronal types were defined by standard electrophysiological procedures (depolarizing and hyperpolarizing current pulses at different levels of membrane potential - V m, antidromic and orthodromic responses) and the morphological features of recorded neurons were known by intracellular staining with Lucifer yellow or Neurobiotin (see Steriade et al., 1993c; Contreras and Steriade, 1995). The results from intracellular recordings are based on neurons with resting V m more negative than -55 mv and overshooting action potentials. Different analyses used to assess the synchronization processes were cross-correlograms, measures of synchrony coefficient, wave-triggered excitatory postsynaptic potentials (EPSPs), spike-triggered-averages, first-spike-analysis, and sequential field correlations as visualized by three-dimensional surfaces and contour maps (see Steriade and Amzica, 1994; Amzica and Steriade, 1995a, 1995b). RESULTS Coalescence of three types of sleep rhythms grouped by the cortical slow oscillation For didactic purposes, three types of oscillations are usually described as characterizing the state of resting sleep: spindles (7-14 Hz), delta (1-4 Hz) and slow oscillation (below 1 Hz, usually 0.6 to 1 Hz). However, in brain-intact animals and humans, the sleep oscillations are not seen in isolation but they are grouped by the recently discovered slow oscillation. Figure 1 shows that the intracellularly recorded slow oscillation ( Hz) progressively develops in conjunction with the increased synchronization of EEG field potentials. This Figure 1. Transformation of slow oscillation patterns with progressive increase in the depth of anesthesia. Intracellular recording of area 5 (suprasylvian) neuron, together with field potentials from the cortical depth (about 1 mm) in the same area. Membrane potential is indicated (-80 mv). The two top traces are separated by a non-depicted period of 8 s. The bottom traces are averages (10 sweeps from the second period) of field and intracellular potentials. Each sweep is extracted around the point of maximum or minimum slope of the intracellular activity. This point (at the vertical dotted line) is obtained by calculating the first derivative of the whole trace and by detecting its positive and negative peaks coinciding with the onset of the depolarization (above) or hyperpolarization (below). Two situations were considered: 1, with more sluggish onset and lower values of the slope (during incipient synchronization), and 2, with more abrupt onset (during a fully synchronized epoch). development is associated with steeper slopes of both depolarizing and hyperpolarizing components of the slow oscillation (compare 1 to 2 in bottom panels). The depolarizing components of the slow oscillation, reflected as a sharp depthnegative field potential (Figs. 1 and 2A), give rise to corticothalamic volleys that, by driving thalamic neurons, induce brief sequences of spindle waves consisting of rhythmic IPSPs in TC cells. The amplitudes of spindle-related IPSPs in TC cells, which increase under steady depolarization (Fig. 2B), are induced by spike-bursts in GABAergic RE cells that, in turn, are driven by the depolarizing component of the cortical slow oscillation (Steriade et al., 1993a; Contreras and Steriade, 1995). Distinctly from the waxing-and-waning pattern of spindle oscillations in the decorticated animals or under barbiturate anesthesia when cortical neurons display reduced spontaneous activities, the spindles triggered by the

3 COALESCENCE OF RHYTHMS IN CORTICOTHALAMIC NETWORKS 3 Figure 2. The depolarizing component of the cortical slow oscillation triggers thalamic spindles. Dual simultaneous intracellular recordings of cortical and TC cells (A) and intracellular recording of TC cell from VL nucleus (B) together with field potentials from the depth of cortical area 4. In this and following figures, polarity of field potentials is as for intracellular recordings (positivity up). The slow oscillation (0.8 Hz in A, 0.5 Hz in B) is best seen in field potential recordings. In A, both cortical and VL cells were under depolarizing current (indicated). Note that, after the initial excitation in VL cell (closely related to the excitation of cortical cell), a few IPSPs developed in VL cell building up a brief spindle sequence (arrow). In B, the depth-negative component of cortical field potential in area 4 was followed by a hyperpolarizing spindle sequence in intracellularly recorded VL neuron. Spindle increased in amplitude by depolarizing the cell with +0.5 na, as compared to the resting V m (-66 mv). Unpublished data by M. Steriade and D. Contreras (A) and I. Timofeev and M. Steriade (B). corticothalamic volley of the slow oscillation under ketaminexylazine anesthesia are shorter and display an exclusively waning pattern (Fig. 2A-B). This is due to the fact that the synchronous excitation of corticothalamic neurons during the slow oscillation entrain, right from the start, a great population of neurons implicated in the genesis of spindles within a thalamic territory, thus explaining the absence of an initial waxing process (Contreras and Steriade, 1996). The coalescence of the slow and spindle oscillations is especially visible during light sleep. The evolution of sleep rhythms, with their progressively increased amplitudes from the end of waking state toward the end of deep resting sleep, is illustrated by means of multi-site recordings in Fig. 3 which shows that, in naturally sleeping animals, the slow oscillation dominates brain electrical activity throughout the state of resting sleep. During light sleep, every cycle of the slow oscillation generally leads to a sequence of spindle waves, on one, another, or all cortical leads. As shown above (Fig. 2), this is due to the synaptic engagement of thalamic neurons implicated in spindle genesis. Notably, though deep sleep displays less spindles, toward the end of deep sleep, just before EEG activation occurs in REM sleep, spindles recover their power, as during the initial stages of resting sleep (Fig. 3). This can be explained by the voltage-dependency of sleep rhythms in TC cells. Indeed, at the single-cell level, spindles occur at the resting V m of TC neurons, whereas, at more hyperpolarized levels, spindles are progressively replaced by intrinsically generated, clock-like delta potentials (Steriade et al., 1991;

4 4 STERIADE AND AMZICA Figure 3. Chronology of sleep rhythms in chronically implanted, behaving cat. Multi-site recordings of field potentials from the cortical depth (about 1 mm in areas 4, 18, 17, 5, 7; see brain figurine) and thalamic centrolateral (CL) rostral intralaminar nucleus. Below the long-term recording of a full wake-sleep cycle (440 s), three panels illustrate expanded recordings during light sleep, deep sleep, and the end of deep sleep before entering REM sleep. Note rhythmicity of PGO waves (at about 0.5 Hz) in area 17 during REM sleep. During light sleep, one cycle of the slow oscillation followed by a spindle sequence is depicted (top trace is a filtered trace to display spindles). Two cycles of slow oscillation (about 0.7 Hz) are depicted during deep sleep. Note, at the end of deep sleep, more pronounced spindles than during deep sleep (see text for comments). Nuñez et al., 1992). These intracellular data from anesthetized preparations found support in results obtained in naturally sleeping animals, showing that thalamic spindles are maximal at sleep onset and decrease thereafter, whereas thalamic delta waves increase gradually during resting sleep (Lancel et al., 1992). Thus, with increasing hyperpolarization of TC cells during resting sleep, due to the progressive diminished firing rates of cholinergic and other types of brainstem-thalamic activating neurons (reviewed in Steriade and McCarley, 1990), the incidence and amplitude of spindles are largely diminished during deep sleep stages. On the other hand, the reappearance of spindles toward the end of resting sleep (see Fig. 3) is attributable to a relative depolarization of TC cells, due to the increased firing rates of brainstem-thalamic reticular neurons that display precursor-increased rates of spontaneous firing, 30 to 60 s before the onset of REM sleep (Steriade et al., 1990). Spindling is not the only sleep rhythm that is modulated and grouped by the cortical slow oscillation. (A) The intrinsically generated delta rhythm of TC cells is influenced by the slow oscillation because the rhythmic depolarizing corticothalamic drives increase the membrane conductance of TC cells and prevent the interplay between a hyperpolarization-activated

5 COALESCENCE OF RHYTHMS IN CORTICOTHALAMIC NETWORKS 5 Figure 4. Intracortical and corticothalamic synchronization of slow oscillation. Sequential field correlation analyses (see technical details in Amzica and Steriade, 1995) of synchronization between cortical areas 5 and 7, areas 4 and 18, and between area 7 and thalamic CL nucleus. Brain figurine indicates the location of cortical electrodes. The contour maps (white indicates maximally synchronized activity) depict timewindows of 4 s (from -2 s to +2 s) during a full wake-sleep cycle lasting for 250 s (to be read from bottom to top). Maps derived from wavelets (see text). The maps show increased synchronization during deep sleep, compared to both light sleep and brain-activated states of waking and REM sleep. Higher coherence of slow oscillation appears among closely located areas 5 and 7 than among more distant ones (areas 4 and 18). Note also clear-cut synchronization of the slow oscillation between cortical area 7 and reciprocally related CL thalamic nucleus. cation current (I h ) and a calcium current de-inactivated by membrane hyperpolarization (I t ), thus periodically dampening the slow oscillation (see Fig. 10A in Steriade et al., 1993a; and Box 1 in Steriade et al., 1994). However, as corticothalamic volleys also drive GABAergic RE neurons, singly deltaoscillating TC cells may be synchronized because RE cells set their V m at the adequate level where delta rhythm is generated (Steriade et al., 1991). (B) The other component of delta waves, that is generated intracortically after thalamectomy (see above), has not yet been systematically studied at the intracellular level to shed light on its neuronal substrate(s). One possibility is that the frequency band of 1-4 Hz in the power spectrum during late stages of resting sleep results, at least partially, from the shape of the depth-negative (depolarizing) component of the slow oscillation ( s), which represents the K-complex (Amzica and Steriade, 1997a,

6 6 STERIADE AND AMZICA Figure 5. Similarity of slow oscillation patterns in intracellular recordings under ketamine-xylazine anesthesia and extracellular recordings in chronically implanted, naturally sleeping cats. (A) The slow oscillation (about 0.9 Hz) in dual simultaneous intracellular recordings from regularspiking cell in cortical area 4 and TC cell in the ventrolateral (VL) nucleus. Cat under ketamine-xylazine anesthesia. Arrow points to a lowthreshold spike-burst. An expanded cycle is shown at right. Note: (a) depth-positive (upward) EEG waves are associated with hyperpolarization of cortical and thalamic cells, whereas the sharp depth-negativies are associated with depolarization and action potentials in cortical cell, while the thalamic neuron display a rebound spike-burst with a delay of ms; (b) brief sequence of EEG spindles after the depth-negative sharp deflection (third cycle of slow oscillation); and (c) fast depolarizing waves (40-50 Hz) in cortical neuron during the sustained depolarization. (B) Chronically implanted, naturally sleeping cat. Six traces represent: depth-eeg from motor (precruciate) area 4; depth-eeg from visual area 17; unit discharges and slow focal potentials from association area 5 in the anterior suprasylvian gyrus; and similar recording from an adjacent focus (2 mm apart) in area 5; electrooculogram (EOG); and electromyogram (EMG). Right part in SWS panel shows reduction, up to disappearance, of fast rhythms (filtered Hz) during the prolonged depth-positive wave of the slow oscillation that, in intracellular recordings, is associated with hyperpolarization of cortical and thalamic neurons. Unpublished data from experiments by M. Steriade and D. Contreras (A) and by M. Steriade and F. Amzica (B); inset in (B) is modified from Steriade et al. (1996a). 1997b). Anyway, typical delta waves, at a frequency of 2-4 Hz, generated by both regular-spiking and intrinsically bursting cortical neurons, are grouped within sequences recurring with the slow rhythm (see Fig. 3 in Steriade et al., 1993d). And, in human sleep EEG, sequential mean amplitudes of delta waves show their periodic recurrence with the rhythm of slow oscillation (Steriade et al., 1993c). That delta and slow oscillation represent two distinct phenomena was recently demonstrated by Achermann and Borbély (1997) who showed differences in the dynamics between the slow and the delta oscillations, as the latter declines in activity from the first to the second non-rem sleep episode, whereas the former does not.

7 COALESCENCE OF RHYTHMS IN CORTICOTHALAMIC NETWORKS 7 Figure 6. The K-complex in human sleep. Fast Fourier transform (FFT) from six cortical leads (see head figurine) in a healthy adult subject. FFT shows a peak at about 0.8 Hz (dotted line) reflecting the slow oscillation during stage 2. Note also an increased power spectrum at about 15 Hz (arrow) reflecting spindle sequences that follow the slow oscillation (see also Figs. 1 to 3 from animal experiments). Bottom panels depict K complexes consisting of the depth-negative (depolarizing) component of the slow oscillation followed (left) or not (right) by spindles in various cortical leads (same arrangements as in the top panel). The current confusion in the literature between delta oscillation and delta waves is probably due to the fact that the presence of a peak in power spectrum may result from an oscillation with the frequency of the peak and/or a frequent occurrence of waves with a duration and shape that would contribute to that particular peak. We investigated the synchronization of slow oscillation in naturally sleeping animals by using the wavelet procedure that detects waveforms similar in shape with a preset pattern. Thus, the original EEG trace is digitally filtered and tranformed into a new time series (signal) that conserves from the original only the relevant wavelets (Amzica and Steriade, 1997a). In order to match K-complexes (see below), we used Daubechies' wavelets (Daubechies, 1988). Figure 4 shows that the intracortical as well as corticothalamic synchronization of the slow oscillation is most obvious during deep sleep and is best expressed among areas across the same gyrus (areas 5 and 7) or among cortical areas and related thalamic nuclei (area 7 and rostral intralaminar nuclei). However, synchronization is also seen between morphologically distant and functionally different cortical fields (motor area 4 and visual area 18). Finally, because the slow oscillation was first described intracellularly under different anesthetics, we had to validate the similarity of these cellular patterns to those observed extracellularly in chronically implanted, unanesthetized animals. Under both ketamine-xylazine anesthesia and natural sleep, the slow oscillation has a frequency of 0.6 to 1 Hz (Fig. 5A-B). Under anesthesia, the field potentials associated with this oscillation are prolonged depth-positive (hyperpolarizing) and depth-negative (depolarizing) components (Fig. 5A). Similar aspects are observed in natural resting sleep, when the depth-positive waves are accompanied by silenced firing, while depth-negative sharp deflections are associated with brisk firing (Fig. 5B). Surprisingly, because fast oscillations (generally Hz) are conventionally associated with brainactivated states, similar fast oscillations also appear during resting sleep but are selectively obliterated during the depthpositive component of the slow oscillation (see inset in Fig.

8 8 STERIADE AND AMZICA slow oscillation during natural sleep (see Fig. 2C in Amzica and Steriade, 1997a). (B) The power spectrum reveals a major peak around 1 Hz, that becomes evident from stage 2 and continues throughout resting sleep (top panel in Fig. 7). The slow oscillation is particularly abundant in fronto-parietal leads (bottom panel in Fig. 7). These data invite human sleep researchers to consider the two types of oscillatory activities below 4 Hz (delta, 1-4 Hz; slow, <1 Hz) and, accordingly, to analyze their results by taking into account the distinctness of these two oscillations, as demonstrated by Achermann and Borbély (1997; see above). Figure 7. Power spectrum (FFT) of human sleep during various stages of slow-wave sleep (top panel) and in recordings from different cortical leads (for C3, P3 and O1, see head figurine in Fig. 6; T3 is left temporal; and Fp1 is left fronto-parietal). Note two peaks of the slow oscillation (0.4 and 1 Hz) during stages 2 and 3-4. Also note prevalence of the slow oscillation in fronto-parietal lead. Top panel from C3; bottom panel from sleep stages 3-4. Data are from two subjects, one for each panel. 5B). This demonstrates the voltage(depolarization)- dependency of fast oscillations. The slow oscillation and K-complexes in human sleep After preliminary data showing the presence of slow oscillation during natural sleep in humans (Steriade et al., 1993c), the human slow oscillation (<1 Hz) was recently reported in parallel studies from two laboratories (Achermann and Borbély, 1997; Amzica and Steriade, 1997a). Here, we document different aspects of the human slow sleep oscillation. (A) During stage 2, scalp recordings show a prevalent peak (0.8 Hz) within the frequency range of the slow oscillation as well as a minor mode around 15 Hz reflecting spindle waves (Fig. 6). The depth-negative components of the slow oscillation, followed or not by spindles, represent the K-complexes (bottom panels in Fig. 6). The frequency of K-complexes (peaks at 0.5 Hz in stage 2, 0.7 Hz in stages 3-4 of human sleep) is very similar, up to identity, to the frequency of the DISCUSSION Cellular mechanisms of sleep oscillations Sequences of spindle waves, recurring with a slow rhythm at about Hz, occur during early sleep stages and are generated by interactions between RE and TC neurons (Steriade et al., 1993b; Bal et al., 1995). The pacemaking role of RE neurons was shown by abolition of spindles in TC systems after disconnection from RE nucleus (Steriade et al., 1985) and by preservation of spindles in the deafferented rostral pole of the RE nucleus (Steriade et al., 1987). Although spindles are generated in the thalamus after decortication (Morison and Bassett, 1945), corticothalamic volleys are important in triggering and synchronizing spindles throughout TC systems (Steriade et al., 1972; Contreras et al., 1996a, 1997). Delta waves, usually regarded as a single type of EEG waves, consist of two components. The cortical one survives thalamectomy (Villablanca, 1974; Steriade et al., 1993d). The thalamic-generated delta oscillation is present after decortication (Curró Dossi et al., 1992), is stereotyped and clock-like, and its intrinsic-cell nature is due to the interplay of two hyperpolarization-activated currents of TC neurons, I h and I t (McCormick and Pape, 1990; Soltesz et al., 1991). The basic features of the recently described slow oscillation (see Introduction) consist of a prolonged hyperpolarization (up to 1 s), associated with a depth-positive (surface-negative) EEG wave, followed by a long-lasting depolarization (up to 0.8 s), associated with a depth-negative (surface-positive) field potential. These sequences recur periodically, with a rhythm of 0.6 to 1 Hz. The long-lasting depolarization consists of EPSPs, fast prepotentials (FPPs) and fast IPSPs reflecting the action of synaptically coupled GABAergic local-circuit cells; in addition, the depolarizing component is made up of both NMDA-mediated synaptic excitatory events and a voltagedependent persistent Na + current, as the depolarizing envelope is shortened by adminstration of ketamine, an NMDA blocker, or intracellular injection of QX-314, a blocker of Na + currents (Steriade et al., 1993c). The prolonged hyperpolarization results from a series of factors; among them, disfacilitation in cortical networks is probably the most important (Contreras et al., 1996b). Possible significance of sleep rhythms The frenzied activity of cortical neurons during the slow oscillation, occurring in natural sleep or deep anesthesia (see

9 COALESCENCE OF RHYTHMS IN CORTICOTHALAMIC NETWORKS Figs. 1 and 5) during which consciousness is conventionally thought to be annihilated, prompts us to consider different roles played by the rhythmic bombardment of thalamic and cortical neurons upon their targets. Indeed, the deafferentation of thalamocortical networks produced by the spindle-related IPSPs in TC cells, with the consequence of obliterating incoming messages and thus disconnecting the brain from the outside world, is probably not the only effect of sleep oscillations. To begin with, whereas TC cells do not transmit ascending afferent signals to cortex during the hyperpolarizing phase of the slow oscillation because the EPSPs do not reach firing threshold, the internal dialogue of the brain (as tested by corticocortical and corticothalamic volleys) is not disrupted during this hyperpolarizing phase (Timofeev et al., 1996). The preservation, during resting sleep, of this form of internal communication is reminiscent of earlier data showing that callosally evoked discharges in precentral neurons of behaving monkeys are not diminished from waking to resting sleep and may even be enhanced (Steriade et al., 1974). The state of resting sleep may subserve even more noble functions during different oscillatory activities of thalamic and cortical neurons. Rhythmic activation of cortical neurons, produced by repetitive spike-bursts of TC cells during sleep spindles, are hypothesized to reinforce and/or specify the circuitry, to stimulate dendrites to grow more spines, and to contribute to the consolidation of memory traces acquired during wakefulness (Steriade et al., 1993a). The idea of such plastic changes may be tested by using procedures of cellular conditioning and by lesioning different structures implicated in the production of sleep oscillations. For example, the effects of abolishing spindles in a hemisphere, after chemical lesioning of the RE nucleus (Steriade et al., 1985), may be investigated upon the time required for establishing conditioned responses in the ipsilateral cortex devoid of spindles. That spindles and their artificial model, augmenting responses evoked by lowfrequency (10 Hz) repetitive volleys, are able to produce shortterm plasticity was demonstrated even in the thalamus of decorticated animals (Steriade and Timofeev, 1997). In these experiments, intrathalamic stimulation at 10 Hz produced a progressive and persistent increase in slow depolarizing responses of TC cells, as well as to a persistent and prolonged decrease in the amplitudes of the IPSPs. Even more pronounced plastic changes are expected to occur with augmenting responses in an animal with intact cortex (Morison and Dempsey, 1943; Morin and Steriade, 1981; Castro- Alamancos and Connors, 1996) since augmenting does not lead to prolonged paroxysmal developments in decorticated animals but such transformations can occur in the presence of intact thalamocorticothalamic loops (unpublished data). After a series of repetitive responses in bursting thalamic neurons, evoked by cortical volleys at 10 Hz, the neurons spontaneously produced spike-bursts very similar to the shape and frequencies of those evoked stimuli (see Fig. 7 in Steriade, 1991). The "memory" of the circuit, presumably due to resonant frequencies in the thalamus and neocortex, may eventually lead to paroxysmal events, consisting of spike-wave seizures at 2-4 Hz (Steriade et al., 1976). Thus, synapses within intracortical and thalamocortical circuits may be thought of as dynamically stabilized by internally generated, apparently "non-utilitarian" excitations during sleep oscillations (see Kavanau, 1994). The "rehearsal", in resting sleep, of information acquired during active behavior (Buzsáki, 1989; Wilson and McNaughton, 1994) is also revealed by the persistence, during periods of subsequent sleep, of intracortical and corticothalamic synchrony of fast (gamma) oscillations acquired during conditioning sessions (Amzica et al., 1997). All these data suggest that the reexpression of information during sleep may be related to memory consolidation. ACKNOWLEDGMENTS This work was supported by grants from the Medical Research Council of Canada and Human Frontier Science Program. F. Amzica is a postdoctoral fellow, partially supported by Fonds de Recherche en Santé du Québec. We thank D. Contreras and I. Timofeev for their collaboration in some unpublished experiments. REFERENCES 1. Achermann P, Borbély AA. Low-frequency (<1 Hz) oscillations in the human sleep EEG. Neuroscience 1997; 81: Amzica F, Steriade M. Short- and long-range synchronization of the slow (<1 Hz) cortical oscillation. J Neurophysiol 1995a; 73: Amzica F, Steriade M. Disconnection of intracortical synaptic linkages disrupts synchronization of a slow oscillation. J Neurosci 1995b; 15: Amzica F, Steriade M. The K-complex: its slow (<1 Hz) rhythmicity and relation with delta waves. Neurology 1997a; 49: Amzica F, Steriade M. Cellular substrates and laminar profile of sleep K-complex. Neuroscience 1997b; 82: Amzica F, Neckelmann D, Steriade M. Instrumental conditioning of fast (20- to 50-Hz) oscillations in corticothalamic networks. Proc Natl Acad Sci USA 1997; 94: Bal T, von Krosigk M, McCormick DA. Role of the ferret perigeniculate nucleus in the generation of synchronized oscillations in vitro. J Physiol Lond 1995; 483: Buzsáki G. Two-stage model of memory trace formation: a role for "noisy" brain states. Neuroscience 1989; 31: Castro-Alamancos MA, Connors BW. Cellular mechanisms of the augmenting response: short-term plasticity in a thalamocortical pathway. J Neurosci 1996; 16: Contreras D, Steriade M. Cellular basis of EEG slow rhythms: a study of dynamic corticothalamic relationships. J Neurosci 1995; 15: Contreras D, Steriade M. Spindle oscillations in cats: the role of corticothalamic feedback in a thalamically generated rhythm. J Physiol Lond 1996; 490: Contreras D, Destexhe A, Sejnowski TJ, Steriade M. Control of spatiotemporal coherence of a thalamic oscillation by corticothalamic feedback. Science 1996a; 274: Contreras D, Timofeev I, Steriade M. Mechanisms of longlasting hyperpolarizations underlying slow sleep 9

10 10 STERIADE AND AMZICA oscillations in cat thalamocortical networks. J Physiol Lond 1996b; 494: Contreras D, Destexhe A, Sejnowski TJ, Steriade M. Spatiotemporal patterns of spindle oscillations in cortex and thalamus. J Neurosci 1997; 17: Curró Dossi R, Nuñez A, Steriade M. Electrophysiology of a slow (0.5-4 Hz) intrinsic oscillation of cat thalamocortical neurones in vivo. J Physiol Lond 1992; 447: Daubechies I. Orthonormal bases of compactly supported wavelets. Commun Pure Appl Math 1988; 41: Kavanau JL. Sleep and dynamic stabilization of neural circuitry: a review and synthesis. Behav Brain Res 1994; 63: Lancel M, van Riezen H, Glatt A. The time course of sigma activity and slow wave activity during NREMs in cortical and thalamic EEG of the cat during baseline and after 12 hours of wakefulness. Brain Res 1992; 596: McCormick DA, Pape HC. Properties of a hyperpolarizationactivated cation current and its role in rhythmic oscillation in thalamic relay neurones. J Physiol Lond 1990; 431: Morin D, Steriade M. Development from primary to augmenting responses in primary somatosensory cortex. Brain Res 1981; 205: Nuñez A, Curró Dossi R, Contreras D, Steriade M. Intracellular evidence for incompatibility between spindle and delta oscillations in thalamocortical neurons of cat. Neuroscience 1992; 48: Soltesz I, Lightowler S, Leresche N, Jassik-Gerschenfeld D, Pollard CE, Crunelli V. Two inward currents and the transformation of low-frequency oscillations of rat and cat thalamocortical cells. J Physiol Lond 1991; 441: Steriade M. Alertness, quiet sleep, dreaming. In: A Peters, EG Jones, eds. Cerebral Cortex, vol. 9, Normal and Altered States of Function. New York: Plenum Press, 1991, pp Steriade M, Amzica F. Dynamic coupling among neocortical neurons during evoked and spontaneous spike-wave seizure activity. J Neurophysiol 1994; 72: Steriade M, Llinás RR. The functional states of the thalamus and the associated neuronal interplay. Physiol Rev 1988; 68: Steriade M, McCarley RW. Brainstem Control of Wakefulness and Sleep. New York: Plenum Press, Steriade M, Timofeev I. Short-term plasticity during intrathalamic augmenting responses in decorticated cats. J Neurosci 1997; 17: Steriade M, Iosif G, Apostol V. Responsiveness of thalamic and cortical motor relays during arousal and various stages of sleep. J Neurophysiol 1969; 32: Steriade M, Wyzinski P, Apostol V. Corticofugal projections governing rhythmic thalamic activity. In: TL Frigyesi, E Rinvik, MD Yahr, eds. Corticothalamic Projections and Sensorimotor Activities. New York: Raven Press, 1972, pp Steriade M, Deschênes M, Wyzinski P, Hallé JY. Input-output organization of the motor cortex and its alterations during sleep and waking. In: O Petre-Quadens, JD Schlag, eds. Basic Sleep Mechanisms. New York: Academic Press, 1974, pp Steriade M, Oakson G, Diallo A. Cortically elicited spikewave afterdischarges in thalamic neurons. Electroencephalogr Clin Neurophysiol 1976; 41: Steriade M, Deschênes M, Domich L, Mulle C. Abolition of spindle oscillations in thalamic neurons disconnected from nucleus reticularis thalami. J Neurophysiol 1985; 54: Steriade M, Domich L, Oakson G, Deschênes M. The deafferented reticularis thalamic nucleus generates spindle rhythmicity. J Neurophysiol 1987; 57: Steriade M, Datta S, Paré D, Oakson G, Curró Dossi R. Neuronal activities in brain-stem cholinergic nuclei related to tonic activation processes in thalamocortical systems. J Neurosci 1990; 10: Steriade M, Curró Dossi R, Nuñez A. Network modulation of a slow intrinsic oscillation of cat thalamocortical neurons implicated in sleep delta waves: cortically induced synchronization and brainstem cholinergic suppression. J Neurosci 1991; 11: Steriade M, Contreras D, Curró Dossi R, Nuñez A. The slow (<1 Hz) oscillation in reticular thalamic and thalamocortical neurons: scenario of sleep rhythm generation in interacting thalamic and neocortical networks. J Neurosci 1993a; 13: Steriade M, McCormick DA, Sejnowski TJ. Thalamocortical oscillations in the sleeping and aroused brain. Science 1993b; 262: Steriade M, Nuñez A, Amzica F. A novel slow (<1 Hz) oscillation of neocortical neurons in vivo: depolarizing and hyperpolarizing components. J Neurosci 1993c; 13: Steriade M, Nuñez A, Amzica F. Intracellular analysis of relations between the slow (<1 Hz) neocortical oscillation and other sleep rhythms of the electroencephalogram. J Neurosci 1993d; 13: Steriade M, Contreras D, Amzica F. Synchronized sleep oscillations and their paroxysmal developments. Trends Neurosci 1994; 17: Steriade M, Amzica F, Contreras D. Synchronization of fast (30-40 Hz) spontaneous cortical rhythms during brain activation. J Neurosci 1996a; 16: Steriade M, Contreras D, Amzica F, Timofeev I. Synchronization of fast (30-40 Hz) spontaneous oscillations in intrathalamic and thalamocortical networks. J Neurosci 1996b; 16: Timofeev I, Steriade M. Low-frequency rhythms in the thalamus of intact-cortex and decorticated cats. J Neurophysiol 1996; 76: Timofeev I, Contreras D, Steriade M. Synaptic responsiveness of cortical and thalamic neurones during various phases of slow sleep oscillation in cat. J Physiol Lond 1996; 494: Villablanca J. Role of the thalamus in sleep control: sleepwakefulness studies in chronic diencephalic and athalamic cats. In: O Petre-Quadens, JD Schlag, eds. Basic Sleep Mechanisms. New York: Academic Press, 1974, pp Wilson MA, McNaughton BL. Reactivation of hippocampal ensemble memories during sleep. Science 1994; 265: