Corticothalamic 5–9 Hz oscillations are more pro-epileptogenic than sleep spindles in rats

Abstract

Absence-related spike-and-wave discharges (SWDs) occur in the thalamocortical system during quiet wakefulness or drowsiness. In feline generalized penicillin epilepsy, SWDs develop from sleep spindles. In contrast, in genetic absence epilepsy rats from Strasbourg (GAERS), SWDs develop from wake-related 5–9 Hz oscillations, which are distinct from spindle oscillations (7–15 Hz). Since these two oscillation types share common frequency bands and may contribute to SWD genesis, it is important to compare their thalamic cellular mechanisms. Under neuroleptic analgesia, in GAERS and control non-epileptic rats barbiturates abolished both SWDs and 5–9 Hz oscillations but increased the incidence of spindle-like oscillations. Within the thalamocortical circuit 5–9 Hz oscillations occurred more coherently than spindle-like oscillations. Intracellular events associated with 5–9 Hz and spindle-like oscillations were distinctively different in both thalamic relay and reticular neurons. In both cell types, SWDs and 5–9 Hz oscillations emerged from a significantly more depolarized membrane potential than spindle-like oscillations. In relay neurons, 5–9 Hz oscillations were mainly characterized by a rhythmic depolarization, which occurred during a tonic hyperpolarization and which could trigger an apparent low-threshold Ca²⁺ potential, whereas spindle-like oscillations were characterized by a rhythmic hyperpolarization. In reticular cells, SWDs and 5–9 Hz oscillations occurred during a tonic hyperpolarization, whereas spindle-like oscillations occurred during a long-lasting depolarizing envelope. The difference in the intracellular events between 5–9 Hz and spindle-like oscillations and similarities between 5–9 Hz oscillations and SWDs indicate that in GAERS, 5–9 Hz oscillations are more pro-epileptogenic than spindle-like oscillations. In conclusion, the present study strongly supports the hypothesis that SWDs in GAERS are generated by a wake-related corticothalamic resonance, and not by sleep-related, hypersynchronous, spindle-like activity originating in the thalamus.

Absence epilepsy is a neurological disorder characterized by recurrent losses of consciousness and bilateral spike-and-wave discharges (SWDs) in the thalamocortical (TC) system (Williams, 1950; Gloor & Fariello, 1988). The genetically determined aetiology of absence-related epileptic discharges remains unknown. Therefore, experimental models are necessary to understand the pathophysiology of absence seizures and the network mechanisms underlying these chronic states. Since the middle of the 20th century, intensive studies have been conducted in two (electrophysiological and pharmacological) feline models of absence seizures: cats displaying bilateral SWDs following either midline thalamic stimulation (Jasper & Drooglever-Fortuijn, 1946), or after systemic injection of penicillin (Prince & Farell, 1969). In feline generalized penicillin epilepsy, SWDs were recorded as a gradual transformation of sleep spindles (Kostopoulos et al. 1981; Kostopoulos, 2000). These SWDs were correlated with a hypersynchronization of rhythmic cellular discharges in related cortical and thalamic regions (Avoli & Gloor, 1982; Avoli et al. 1983). Furthermore, a comprehensive study conducted in ferret thalamic slices has revealed that spindle-like oscillations can be transformed into paroxysmal oscillations following blockade of GABA_A receptors (von Krosigk et al. 1993). Taken together, these in vivo and in vitro neurophysiological studies have yielded data supporting Gloor's corticoreticular hypothesis (Gloor, 1968; Kostopoulos, 2000; McCormick & Contreras, 2001), which claims that a normal thalamic activity reaching a hyperexcitable cortex triggers SWDs (Gloor & Fariello, 1988).

However, typical well-organized absence-related SWDs occur especially during immobile, inattentive wakefulness or drowsiness (Niedermeyer, 1965; Mirsky et al. 1986; Halasz, 1991; Loiseau, 1992), whereas sleep spindles characterize the early phase of sleep (Steriade et al. 1993). Thus, it is not clear in which behavioural state sleep spindles could initiate paroxysmal oscillations. In genetic absence epilepsy rats from Strasbourg (GAERS), SWDs develop during the immobile state in the TC system from wake-related normal 5–9 Hz oscillations (Pinault et al. 2001). In those rats, SWDs (6–8 spike-and-wave complexes s⁻¹) correspond to hypersynchronous 5–9 Hz oscillations, which are launched by corticothalamic (CT) neurons (Pinault, 2003). Physiological 5–9 Hz oscillations occur during awake immobility and are distinguishable from sleep-related spindle (7–15 Hz) oscillations in electrocorticographic recordings (Pinault et al. 2001). Since both oscillation types share a common frequency band, one could be the derivative of the other. To clarify the relationship of spindle-like (7–15 Hz) and 5–9 Hz oscillations, and their similarity or difference to SWDs, it is essential to systematically compare the thalamic cellular mechanisms of these two types of oscillations.

In this study we conducted extracellular and current-clamp intracellular recordings in the thalamus of GAERS and control non-epileptic (NE) rats (free of spontaneous SWDs) under neuroleptic analgesia in combination with the EEG of the frontoparietal cortex (primary motor and somatosensory cortices). Because barbiturates are well known for inducing spindle-like oscillations in the TC system (Gandolfo et al. 1985; Contreras et al. 1997; Mackenzie et al. 2004), the incidence of spindle-like oscillations was increased either following intravenous injection of pentobarbital at subanaesthetic doses in rats under neuroleptic analgesia (in GAERS and in control NE rats) or under pentobarbital–fentanyl anaesthesia. Our findings support the hypothesis that, in GAERS, SWDs are generated by a wake-related CT resonance, and not by sleep-related TC oscillations.

Methods

Animals

Experiments were conducted on inbred, adult male Wistar rats (35 GAERS and 58 control NE rats, 280–350 g), complying with our institutionally recommended procedures for animal use and care (Comité Régional d'Ethique en Matière d'Expérimentation Animale, Strasbourg).

Anaesthesia and surgery

All surgical procedures were done under deep general anaesthesia (pentobarbital: 40 mg kg⁻¹, i.p., Sanofi, Libourne, France; and ketamine: 50 mg kg⁻¹, i.m., Merial, Lyon, France). Tracheotomy and catheterization of the penile vein were performed, and the animal was placed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA). A new stabilizing craniotomy–duratomy technique (Pinault, 2005) was systematically performed to improve the success rate of single-cell electrophysiology experiments, to increase the precision of the stereotaxical approach to single neurons in the target region, and to considerably reduce undesirable non-neuronal rhythms (heart beat and breathing) during intracellular recordings. Before the end of this general pentobarbital–ketamine anaesthesia, the rat was subjected either to neuroleptic analgesia (35 GAERS and 52 control NE rats) or barbiturate–fentanyl anaesthesia (6 control NE rats) (see below).

All rats were artificially ventilated (SAR-830; CWE, Ardmore, PA, USA) in the pressure mode (8–12 cmH₂O; 60–65 breaths min⁻¹) using an O₂-enriched gas mixture (70–50% air and 30–50% O₂). The rat's rectal temperature was maintained at its physiological level (37–38.3°C) using a thermoregulated blanket (Fine Science Tools Inc., Heidelberg, Germany). The EEG, which spontaneously displayed synchronized slow oscillations (see Results), and the heart rate were also under continuous monitoring to maintain a steady depth of anaesthesia by adjusting the injection rate of the anaesthetic solution.

Neuroleptic analgesia was initiated and then maintained by a continuous intravenous injection (0.5 ml h⁻¹) of the following mixture: d-tubocurarine chloride (0.4 mg; Sigma-Aldrich, Saint-Quentin Fallavier, France), fentanyl (1 μg; Janssen, Boulogne-Billancourt, France), Haldol (100 μg; Janssen), and glucose (25 mg). (The adequacy of the neuroleptic analgesia was established in the absence of neuromuscular blockade.)

Barbiturate–fentanyl anaesthesia was initiated and maintained using a continuous intravenous injection (0.5 ml h⁻¹) of the following mixture (quantity given per hour for a rat of 300 g): d-tubocurarine chloride (0.4 mg), fentanyl (1 μg), pentobarbital (3.5–8.2 mg) and glucose (25 mg).

Electrophysiology

Glass micropipettes (30–70 MΩ) were filled with a solution containing 1.5% N-(2-aminoethyl)biotinamide hydrochloride (Neurobiotin; Vector Laboratories, Burlingame, CA, USA) dissolved in either 1 m CH₃COOK, or in 3 m KCl. The pipette was lowered with a stepping microdriver (Burleigh, Fishers, NY, USA) into the somatosensory or motor thalamus to reach a single TC or thalamic reticular nucleus (TRN) neuron, which was extracellularly and/or intracellularly recorded simultaneously along with the EEG of the primary motor and somatosensory cortices. Dual extracellular single-unit TC–TRN recordings were simultaneously performed with the EEG.

Multi-unit recordings were done in the thalamus using glass micropipettes (tip diameter: 2–3.5 μm) in conjunction with the cortical EEG. The location of the recording sites was identified histologically following extracellular application (500–600 nA, 200 ms on/200 ms off, 5–10 min) of Neurobiotin.

Signal conditioning

Electrophysiological data were processed with band passes of 0.1–1200 Hz for the EEG, of 0–6 kHz for cellular activity, and of 0.3–6 kHz for multi-unit recordings (Cyber-Amp 380; Axon Instruments, Union City, CA, USA). Signals were digitized at a sampling rate >18 kHz. During the intracellular recording session, a current pulse in the range of −0.2 to −0.5 nA was applied every 2 s to keep the Wheatstone bridge well balanced. Using square-wave current pulses (range of ± 3 nA), the input membrane resistance and intrinsic firing patterns of thalamic, relay and reticular neurons could be assessed.

Histology

At the end of the recording session, the neurons were individually labelled using the juxtacellular (Pinault, 1996) or intracellular tracer microiontophoresis technique for standard histological identification. For the paired single-unit recording experiments, the juxtacellular filling procedure was applied only in the two neurons of the last pair that had been recorded. After a survival period of at least 30 min, the animal was killed with an intravenous overdose of pentobarbital. Then it was transcardially perfused with a fixative containing 4% paraformaldehyde and 0.25% glutaraldehyde in 10 mm phosphate-buffered saline, and the brain tissue was processed using standard histological techniques for retrieving the tracer-filled neurons.

Data analysis

Electrophysiological recordings were analysed with Axon software (Clampex, v7; Axon Instruments), and the tracer-filled neurons were examined with a light microscope (E600; Nikon France, Champigny-sur-Marne, France). Some of the neurons were reconstructed using the Neurolucida system (Microbrightfield, Colchester, VT, USA). The location of any marked cell was ascertained by referring to a stereotaxic atlas (Paxinos & Watson, 1986).

Fast Fourier transformations (FFT) were computed using DataWave softwares (SciWorks, v4; DataWave Technologies, Berthoud, CO, USA). Fourier Transform analysis was based on 1.6 s epochs, with a resolution of 0.6 Hz, and was applied on segments of EEG or extracellular field potential signals (re-digitized at a sampling rate of 2.5 kHz) of at least 2 min. Autocorrelograms of action potential (AP) trains were also performed with DataWave software.

The background noise of the membrane potential of the recorded thalamic neurons, which contained spontaneously occurring intrinsic and synaptic oscillations, was quantified using spectral analysis. A series of 25–35 successive FFTs (epochs of 0.3 s) of the membrane potential, which oscillated between −80 mV and −100 mV (current-clamp mode), was computed. The values of the total power were extracted for 4 frequency bands: slow (1–15 Hz), β (16–30 Hz), γ1 (31–49 Hz), and γ2 (51–100 Hz). The 50 Hz values were discarded to avoid contamination from possible AC noise.

Data are presented as means ± s.e.m. They were evaluated for statistical significance using Student's t test, the significance level being set to 0.05.

Results

In this study the term ‘sleep spindles’ refers to spindle oscillations in the 7–15 Hz range, which have been recorded in freely moving animals, especially during delta waves of the early phase of sleep (Steriade & Deschênes, 1984; Contreras et al. 1997; Pinault et al. 2001; Mackenzie et al. 2004). The term ‘spindle-like oscillations’ refers to spindle oscillations in the 7–15 Hz range, which have been recorded during anaesthesia, especially during delta waves (Contreras et al. 1997; Pinault et al. 2001).

Data base

The data reported in this paper are based on EEG recordings of five GAERS and five control NE rats, 32 simultaneous extracellular recordings of TC and TRN neurons and on individual intracellular recordings of 30 TC and 23 TRN neurons (30 GAERS, 58 control NE rats). All intracellular recordings had to fulfil the following three criteria: (1) a stable resting membrane potential without holding hyperpolarizing current (Table 1); (2) a firing pattern similar to that recorded extracellularly in the same or in other neurons of the same category; and (3) an overshooting of the APs. Following an intravenous injection of barbiturate, the resting membrane potential of the neurons being recorded became significantly more hyperpolarized and the membrane input resistance significantly increased (Table 1).

Table 1. Membrane properties of thalamic neurons recorded under different experimental conditions.

	n	Conditions	Row no.	Resting Vm (mV)	Peak Rin (MΩ)
Control NE	11	Neurolept	1	−60.5 ± 1.2	20.3 ± 2.2
TC cells		(before 5–9 Hz)		1/3: ns	1/3: ns
		+ Pentobarbital	2	−64.5 ± 0.5	25.1 ± 3.3
		(before 7–15 Hz)		1/2: P < 0.05	1/2: P < 0.05
GAERS	12	Neurolept	3	−59.6 ± 0.9	21.0 ± 1.7
TC cells		(before SWD)		2/3: P < 0.05	2/3: P < 0.05
Control NE	8	Neurolept	4	−61.6 ± 1.2	31.3 ± 6.1
TRN cells		(before 5–9 Hz)		4/6: ns	4/6: ns
		+ Pentobarbital	5	−68.3 ± 1.3	47.4 ± 6.8
		(before 7–15 Hz)		4/5: P < 0.05	4/5: P < 0.05
GAERS	12	Neurolept	6	−60.9 ± 1.4	26.7 ± 2.7
TRN cells		(before SWD)		5/6: P < 0.05	5/6: P < 0.05

Values are mean ± s.e.m.; n represents number of neurons. In control non-epileptic (NE) rats, the measurements were done under neuroleptic analgesia (Neurolept), before the occurrence of 5–9 Hz oscillations, then 5–15 min after intravenous injection of pentobarbital at a subanaesthetic dose, before the occurrence of 7–15 Hz oscillations. In GAERS, the measurements were done before the occurrence of spike-and-wave discharges (SWDs). Student's t test was used to compare the means. V_m, membrane potential; R_in, input resistance; TC, thalamocortical; ns, not significant.

EEG oscillations in the frontoparietal cortex in GAERS and control NE rats

In GAERS under neuroleptic analgesia, high-voltage (>0.5 mV) SWDs (6–8 spike-and-wave complexes s⁻¹) occurred in alternation with small-voltage fast oscillations (< 0.2 mV, >15 Hz; Fig. 1A). Spike-and-wave discharges lasted a few seconds to a couple of minutes as described before (Pinault et al. 2001). In control NE rats under neuroleptic analgesia, the surface EEG of the frontoparietal cortex alternated between small-voltage fast oscillations and medium-voltage slower oscillations (< 0.5 mV, < 15 Hz). The latter occurred at 0.5–4 oscillations min⁻¹, lasted on average 3.4 ± 0.9 s (0.5–20 s), and included 1–4 Hz and 5–9 Hz rhythmic waves, which could wax and wane in amplitude (Fig. 1B and Ca).

Figure 1. Pentobarbital abolishes spike-and-wave discharges (SWDs) and induces spindle-like oscillations.

A: top row, surface EEG recordings (truncated) of the primary somatosensory cortex in a GAERS (genetic absence epilepsy rat from Strasbourg) under neuroleptic analgesia, which received a subanaesthetic dose of pentobarbital; bottom row, expanded EEG traces. Before pentobarbital injection and during the recovery period, the EEG displays recurrent high-voltage SWDs of variable duration. Note that pentobarbital induces the occurrence of spindle-like episodes (asterisks) in a GAERS (A) and in a control non-epileptic (NE) rat (B). B, in a control NE rat, 5–9 Hz oscillations spontaneously occur before and at least 20–30 min after pentobarbital injection (recovery). Ca and b, expanded traces of a typical 5–9 Hz oscillation recorded in a control NE rat (Ca) and a 5–9 Hz oscillation that gives rise to a SWD in a GAERS (Cb). Da–c, three types of spindle-like oscillations recorded in GAERS and control NE rats under barbiturate influence: the first type contains only sinusoid-like waves (Da), the second type especially biphasic spikes (Db), and the third type mainly monophasic spikes (Dc).

In the surface EEG of GAERS under neuroleptic analgesia, 5–9 Hz oscillations gave rise to SWDs (Fig. 1Cb), whereas spindle-like oscillations were associated with delta waves (1–4 Hz) but not with the development of SWDs as described before (Pinault et al. 2001). To increase the incidence of spindle-like oscillations we injected pentobarbital at subanaesthetic doses (2.5–4 mg kg⁻¹, i.v.) in GAERS and control NE rats under neuroleptic analgesia. In all GAERS under neuroleptic analgesia (n = 5), barbiturates consistently abolished SWDs and 5–9 Hz oscillations, and induced medium-voltage slower waves and spindle-like oscillations (one episode every 2–5 s) in the surface EEG of the frontoparietal cortex for at least 10–15 min (Fig. 1A). In control NE rats (n = 5), barbiturates consistently abolished 5–9 Hz oscillations and increased the incidence of spindle-like oscillations (Fig. 1B). In GAERS and control NE rats that were under barbiturate influence, spindle-like oscillations consisted of variable waveforms, including sinusoid, monophasic and/or biphasic medium-voltage spikes (Fig. 1Da–c). In GAERS, these spindle-like oscillations were never followed by SWDs (Fig. 1A).

Additional experiments were performed to check whether the spindle-like oscillations induced by a single subanaesthetic barbiturate injection under neuroleptic analgesia were similar to those recorded under continuous barbiturate–fentanyl anaesthesia (6 control NE rats). The cortical EEG and thalamic relay and reticular firings resembled those recorded in rats under neuroleptic analgesia that had received an intravenous injection of pentobarbital at subanaesthetic doses (data not shown).

Oscillations at 5–9 Hz occur in a more coherent manner in the TC system than spindle-like oscillations

Since 5–9 Hz oscillations gave rise to highly coherent SWDs in GAERS, whereas spindle-like oscillations did not, it was important to determine the degree of coherence in the TC system of both oscillation types. Oscillations at 5–9 Hz were obviously more prominent under neuroleptic analgesia, whereas short-lasting (<2 s) spindle-like oscillations were more prominent in the presence of barbiturates (Fig. 1). To assess the degree of coherence in the cerebral cortex and in the TC system of these two oscillation types, simultaneous EEG recordings were performed in the primary motor and somatosensory cortices and in the somatosensory thalamus in control NE rats, which were under neuroleptic analgesia or barbiturate influence. Spectral analysis was computed on periods of recording of at least 2 min. The values of total power were extracted and normalized for the 5–9 Hz (under neuroleptic analgesia) and 7–15 Hz (under barbiturate influence) bands and were plotted against time (Fig. 2Aa, b and Ba, b, upper graphs). These charts allowed the determination of the coincidence in time of a given oscillation in two regions. The relationship between the corresponding normalized FFT values was assessed using a parametric test (linear regression), which made it possible to compare the coherence tendencies of both types of oscillations using the correlation coefficient (R) (Fig. 2Aa, b and Ba, b, lower graphs). Thus, oscillations at 5–9 Hz recorded under neuroleptic analgesia consistently occurred in a more correlated manner in the frontoparietal cortex and in the related thalamus than barbiturate-induced spindle-like oscillations. This finding was consistent in all cases using different anaesthetic conditions (3 rats under neuroleptic analgesia, 2 rats under barbiturate–fentanyl anaesthesia, and 2 rats under neuroleptic analgesia that had received a subanaesthetic dose of pentobarbital).

Figure 2. Oscillations of 5–9 Hz occur in a more coherent manner in the frontoparietal cortex and in the related thalamus than spindle-like oscillations.

Aa and b, normalized fast Fourier transform (FFT; total power) of 5–9 Hz oscillations occurring during a 2 min episode under neuroleptic analgesia in primary motor cortex (M1) and primary somatosensory cortex (S1) (Aa, upper graph) or in S1 and somatosensory thalamus (Th) (Ab, upper graph). Relationships between the corresponding FFT values of the M1 and S1 5–9 Hz oscillations (Aa, lower graph) and of the S1 and Th 5–9 Hz oscillations (Ab, lower graph). Ba and b, normalized FFT (total power) of 7–15 Hz oscillations under barbiturate–fentanyl anaesthesia occurring during a 2 min episode in M1 and S1 (Ba, upper graph) or in S1 and Th (Bb, upper graph). Relationships between the corresponding FFT values of the M1 and S1 7–15 Hz oscillations (Ba, lower graph) and of the S1 and Th 7–15 Hz oscillations (Bb, lower graph). Note more correlated values under neuroleptic analgesia especially between the cortical areas (Aa).

Extracellular recording of TC neurons

Under neuroleptic analgesia, the extracellular thalamic field potential oscillated during the occurrence of 5–9 Hz oscillations in the related cortex (Fig. 3A). Single APs or a burst of 2–4 APs at 200–500 Hz could occur on some of the cycles of the oscillation. When the recorded neurons were not strongly inhibited, a modulation at 5–9 Hz could be revealed using autocorrelation analysis (Fig. 3B).

Figure 3. Discharge patterns of thalamocortical neurons during 5–9 Hz oscillations under neuroleptic analgesia.

The recordings are from control non-epileptic rats. A, recording of a thalamocortical (TC) neuron of the somatosensory thalamus simultaneously with the EEG of the related primary somatosensory cortex (S1Cx). The EEG displays medium-voltage slow oscillations, which include 5–9 Hz oscillations, in alternation with faster low-voltage oscillations. The extracellular field potential of the thalamic recording also displays oscillations in the same frequency range. Note the occurrence of two high-frequency bursts of action potentials (a burst is expanded in the inset) during the oscillation and that an action potential (AP) discharge does not occur on every cycle of the oscillation of the extracellular field potential. B, another TC neuron demonstrating rhythmic modulation of its firing in the 6 Hz range (see the autocorrelograms with a 2 ms bin width, in the top row; the asterisks indicate the rhythmicity) during the occurrence of 5–9 Hz oscillations in the related S1Cx.

During barbiturate-induced EEG oscillations, multiunit recordings of TC cells revealed short-lasting (< 2 s) episodes of rhythmic, synchronized AP discharges at spindle frequencies (7–15 Hz in 9 out of 11 cases; Fig. 4Aa and b). Extracellular recordings of isolated single units for at least 15 min showed that all recorded TC neurons displayed an apparent irregular firing pattern, which included single APs and high-frequency bursts of APs (Fig. 4B). However, the computation of autocorrelograms of such AP trains could reveal 7–15 Hz rhythmicity embedded in the apparent irregular activity (Fig. 4B). In some of these neurons, short-lasting episodes of rhythmic AP discharges at 5–15 Hz, which could include at least one high-frequency burst of APs (200–500 Hz; up to 4 APs), were apparent (Fig. 4C).

Figure 4. Discharge patterns of thalamocortical neurons under barbiturate influence.

The recordings are from control non-epileptic rats. Aa and b, multi-unit recording of TC neurons in the somatosensory thalamus. Note the occurrence of at least three successive 7–15 Hz episodes (asterisks) during a time period of 8 s (Aa) and three successive peaks of synchronized TC firings within a 7–15 Hz episode (Ab). The grey area in Aa is enlarged in Ab, lower trace. The upper trace in Ab is an enlargement of a synchronized AP discharge. B, extracellular single unit recording of a TC cell showing an apparent irregular firing pattern composed of single APs and of high-frequency bursts of APs (a burst is expanded in the inset) during barbiturate–fentanyl anaesthesia. Note that the autocorrelogram (resolution of 1 ms), computed from successive inter-AP intervals, reveals the 7–15 Hz rhythmicity. C, a more rhythmic TC cell is shown, which exhibits 5 successive spindle-like episodes composed of single APs and of high-frequency bursts of APs (a burst is expanded in the inset) during barbiturate–fentanyl anaesthesia. The autocorrelogram reveals the 7–15 Hz rhythmicity.

Extracellular recording of TRN neurons

Under neuroleptic analgesia, all recorded TRN neurons displayed a rhythmic discharge pattern during the occurrence of 5–9 Hz oscillations in the related cortex (Fig. 5A). The rhythmic pattern is principally characterized by the occurrence of high-frequency bursts of APs, which did not systematically occur on every cycle of the oscillation. This was ascertained by the autocorrelogram (Fig. 5A). Under barbiturate influence, TRN cells principally fired recurrent (0.15 ± 0.06 Hz) short-lasting (1.1 ± 0.4 s) trains of high-frequency bursts of APs (Fig. 5B). The bursts in the trains occurred at a variable frequency (5–15 Hz; also see autocorrelogram in Fig. 5B). The trains were sometimes associated with spindle-like oscillations in the EEG of the related cortex (Fig. 5B). The intraburst firing characteristics of TRN cells were appreciably different during the 7–15 Hz-related versus the 5–9 Hz-related high-frequency bursts (Fig. 5D and E). The AP discharge frequency within a burst was on average significantly higher during spindle-like than during 5–9 Hz oscillations (361.1 ± 0.9 Hz versus 242.4 ± 1.0 Hz, Student's t test, P < 0.001). This was mainly caused by the shortening of the inter-AP intervals during the beginning of the bursts. Thus, the instantaneous frequency of two successive APs in a burst was significantly higher for the first nine APs during spindle-like oscillations than during 5–9 Hz oscillations (Student's t test, P < 0.001; Fig. 5E). Furthermore, the AP discharge in the 5–9 Hz bursts presented a less pronounced acceleration–deceleration pattern than that in spindle bursts.

Figure 5. Characteristics of thalamic reticular nucleus (TRN) high-frequency bursts during 5–9 Hz and spindle-like oscillations, and during SWDs.

A, extracellular recording of a TRN cell simultaneously with the surface EEG of the related primary somatosensory cortex (S1Cx) during spontaneously occurring 5–9 Hz oscillations under neuroleptic analgesia. The autocorrelogram reveals the 5–9 Hz rhythmicity. B, 7–15 Hz rhythmic burst firing is recorded in the same TRN cell following (14 min) pentobarbital injection (3.3 mg kg⁻¹, i.v.). The autocorrelogram reveals the 7–15 Hz rhythmicity. C, rhythmic burst firing in a TRN cell of the somatosensory system during a typical SWD in the S1 cortex of a GAERS. D, five successive (from top to bottom) bursts taken during 5–9 Hz and 7–15 Hz oscillations, and during a SWD. E, the instantaneous frequency of two successive APs in bursts relating to 5–9 Hz and 7–15 Hz oscillations, and SWD, are compared with the use of Student's t test (5 TRN cells, each one with twenty 5–9 Hz bursts (neuroleptic analgesia) and twenty 7–15 Hz bursts (neuroleptic analgesia + barbiturate influence); and 5 TRN cells from GAERS, each one with 20 SWD bursts (neuroleptic analgesia); *P < 0.05; ns, not significant). The error bars (2.5 APs s⁻¹ < s.e.m. < 20.0 APs s⁻¹) cannot be shown in the chart at this scale. Note slower intraburst frequencies during normal 5–9 Hz oscillations in TRN cells.

Because both 5–9 Hz-related and SWD-related TRN high-frequency bursts were triggered by CT excitatory postsynaptic potentials and voltage-dependent depolarizations (Pinault, 2003), it was important to compare their intraburst firing characteristics. Almost all of the recorded TRN cells of the somatosensory system discharged a high-frequency burst of APs on each cycle of the SWD (Fig. 5C). This rhythmic pattern could even start before the apparent onset of the SWD (see also Pinault et al. 2001). The AP discharge frequency within an epilepsy-related burst was the highest (434.4 ± 0.4 Hz; Student's t test, P < 0.001 for comparison with the frequencies of the 5–9 Hz-related and of the 7–15 Hz-related bursts) and the first inter-AP interval of the SWD-related bursts was significantly the shortest (Student's t test, P < 0.001; Fig. 5E) when compared to TRN bursts during 5–9 Hz or spindle-like oscillations. In addition, the AP discharge in the epilepsy-related bursts had a significantly more pronounced acceleration–deceleration pattern, at least for the first eight APs (Student's t test, P < 0.001; Fig. 5E).

Intracellular recording of TC neurons

In TC neurons, the membrane potential events underlying the 5–9 Hz and spindle-like oscillations were readily distinguishable (Fig. 6). Oscillations at 5–9 Hz recorded under neuroleptic analgesia started from a significantly more depolarized membrane potential (−60.5 ± 1.2 mV; Student's t test, P < 0.05; Table 1; Fig. 6A) than spindle-like oscillations recorded under barbiturate influence (−64.5 ± 0.5 mV; Table 1; Fig. 6B). In GAERS, SWD-related TC oscillations recorded under neuroleptic analgesia started from a membrane potential statistically similar to normal 5–9 Hz oscillations (−59.6 ± 0.9 mV; Table 1).

Figure 6. Distinct intracellular events associated with spontaneously occurring 5–9 Hz and spindle-like oscillations in thalamic relay neurons.

The recordings are from control non-epileptic rats. A and B, spontaneous 5–9 Hz oscillations (A) or barbiturate-induced 7–15 Hz oscillations (asterisks, B). Note that individual synaptic potentials of presumably lemniscal origin (arrows) occur during and in between the 7–15 Hz episodes (B). C, a spontaneous 5–9 Hz oscillation recorded in a TC neuron during the application of a holding hyperpolarizing current under neuroleptic analgesia. The steady hyperpolarization persists. Da–c, spindle-like oscillation recorded at different membrane potentials in a TC neuron under barbiturate influence. Both the rhythmic hyperpolarizing waves and the envelope disappear at −76 mV, the reversal potential (Db). Arrows indicate spontaneous postsynaptic potentials of presumed lemniscal origin. Note that these postsynaptic potentials do not trigger a low-threshold Ca²⁺ potential. Dd, superimposed responses of the TC neuron to a square pulse of constant current of three different intensities. Note the rebound burst at the offset of hyperpolarizing current. E and F, superimposition of three successive rhythmic depolarizations, which occur during 5–9 Hz (E) and during 7–15 Hz oscillations (F) in a TC cell. Note the occurrence of small synaptic and/or intrinsic unitary events at the beginning of the depolarizing waves in E. In E, one of the 3 depolarizations triggers an apparent low-threshold Ca²⁺ spike topped by a high-frequency burst of APs, whereas in F only a single action potential was evoked during one of the events. The action potentials are truncated in Da, E and F. S1Cx, primary somatosensory cortex.

The membrane input resistance of the recorded TC neurons was significantly lower before the occurrence of 5–9 Hz oscillations or SWDs (20.3 ± 2.2 or 21.0 ± 1.7 MΩ, respectively; Student's t test, P < 0.05; Table 1), i.e. under neuroleptic analgesia, than before the occurrence of spindle-like oscillations (25.1 ± 3.3 MΩ; Table 1), i.e. under barbiturate influence.

The intracellular 5–9 Hz oscillations of TC cells consisted of the rhythmic occurrence of a threshold/subthreshold depolarizing wave–hyperpolarizing wave sequence (Pinault, 2003; Fig. 6A). In contrast, spindle-like oscillations were characterized by rhythmic hyperpolarizations (Fig. 6B and Da) with an occasional AP discharge between two successive hyperpolarizations (Fig. 6Da). Following intravenous injections of pentobarbital at subanaesthetic doses, these spindle-like episodes regularly occurred every 2–5 s (Fig. 6B).

For both the 5–9 Hz and the spindle-like oscillations the rhythmic waves occurred in the trough of a long-lasting hyperpolarization (Fig. 6A and B). This hyperpolarizing envelope lasted as long as the membrane potential was oscillating. The hyperpolarizing envelope of the 5–9 Hz oscillations could not be abolished by applying sustained hyperpolarizing currents even at −92 mV (Fig. 6C). It resembled the SWD-related long-lasting hyperpolarization (Pinault et al. 1998). In contrast, the long-lasting hyperpolarization of spindle-like oscillations was reversed in polarity at a membrane potential ranging from −68 to −78 mV (Fig. 6Da–c). It became depolarizing either when the membrane potential was deeply hyperpolarized (Fig. 6Dc), or when recorded with KCl-filled micropipettes (n = 3 neurons, not shown).

The 5–9 Hz oscillation-related recurrent depolarization started with the summation of synaptic unitary events and could trigger an apparent low-threshold Ca²⁺ potential that launched a high-frequency burst of APs (Fig. 6E). On the other hand, during spindle-like oscillations, no TC cells displayed high-frequency bursts of APs induced by an apparent low-threshold Ca²⁺ potential in our sample (Fig. 6F), even when applying a holding hyperpolarizing current (Fig. 6Db and c). However, single APs could occur following the rhythmic hyperpolarization (Fig. 6F). All TC neurons tested could generate an apparent low-threshold Ca²⁺ potential spontaneously at the offset of a hyperpolarization evoked by a square pulse of cathodal current (Fig. 6Dd). Thus the lack of spontaneous rebound bursts was not the consequence of the suboptimal quality of our recordings. It should also be mentioned that the membrane potential displayed more powerful fast rhythmic activities when recorded under neuroleptic analgesia (Fig. 6C) than when recorded under barbiturate influence (Fig. 6Db) (see below).

Intracellular recording of TRN neurons

Intracellular recordings of TRN neurons revealed that the membrane events underlying the two rhythmic activities at 5–9 Hz and 7–15 Hz were readily distinguishable (Fig. 7). Similarly to TC cells, in TRN cells 5–9 Hz oscillations recorded under neuroleptic analgesia started from a significantly more depolarized membrane potential (−61.6 ± 1.2 mV; Student's t test, P < 0.05; Table 1; Fig. 7A) compared to the spindle-like oscillations recorded under barbiturate influence (−68.3 ± 1.3 mV; Table 1; Fig. 7B). In GAERS, SWD-related TRN oscillations recorded under neuroleptic analgesia started from a membrane potential statistically similar to normal 5–9 Hz oscillations (−60.9 ± 1.4 mV; Table 1).

Figure 7. Distinct intracellular events associated with spontaneously occurring 5–9 Hz and spindle-like oscillations in thalamic reticular neurons.

The recordings are from control non-epileptic rats. A and B, spontaneous 5–9 Hz oscillations under neuroleptic analgesia (A) or barbiturate-induced 7–15 Hz oscillations (B). C, a spontaneous 5–9 Hz oscillation recorded in a TRN cell during the application of a holding hyperpolarizing current under neuroleptic analgesia. The steady hyperpolarization cannot be abolished. Da–c, a spindle-like oscillation recorded at different membrane potentials in a TRN cell under barbiturate influence. The asterisks indicate the rhythmicity, which is similar at all membrane potentials. E and F, superimposition of three successive rhythmic depolarizations, which occur during 5–9 Hz (E) and 7–15 Hz oscillations (F) in a TRN cell. Note the occurrence of synaptic and/or intrinsic unitary events at the beginning of the depolarizing waves in E and F. The action potentials are truncated in Da, Db, E and F.

The membrane input resistance of the recorded TRN neurons was significantly lower before the occurrence of 5–9 Hz oscillations or SWDs (31.3 ± 6.1 or 26.7 ± 2.7 MΩ, respectively; Student's t test, P < 0.05; Table 1), i.e. under neuroleptic analgesia, than before the occurrence of spindle-like oscillations (47.4 ± 6.8 MΩ; Table 1), i.e. under barbiturate influence.

Both 5–9 Hz and spindle-like oscillation-related recurrent depolarizations regularly gave rise to a high-frequency burst of APs (200–500 Hz; up to 13 APs) in TRN neurons but with distinctive intraburst features (see above).

Fully developed 5–9 Hz oscillations occurred in the trough of a long-lasting hyperpolarization (Fig. 7A) similarly to SWD (Fig. 8). In contrast, spindle-like oscillations systematically emerged in parallel with the generation of a slow depolarizing envelope (Fig. 7B). Neither the 5–9 Hz-related nor the SWD-related steady hyperpolarization could be reversed by the application of sustained hyperpolarizing current even at −96 mV (Figs 7C and 8D). The long-lasting hyperpolarization, however, could be almost completely abolished at a membrane potential close to AP threshold (Fig. 8A). The spindle-like oscillation-related depolarizing envelope was also abolished at a membrane potential close to AP threshold (Fig. 7Da–cc). The recurrent depolarization of normal and absence-related 5–9 Hz oscillations and of spindle-like oscillations was significantly reduced in amplitude when the membrane potential was held close to AP threshold. Both the 5–9 Hz and the spindle-like oscillation-related recurrent depolarizations began by the summation of synaptic and/or intrinsic unitary events (Fig. 7E and F, respectively). It is worth mentioning that the SWD-related recurrent depolarization also began by the summation of depolarizing unitary events (Pinault, 2003). The membrane potential of the recorded TRN cells displayed more powerful fast rhythmic activities when recorded under neuroleptic analgesia (Figs 7C and 8D) than when recorded under barbiturate influence (Fig. 7Dc).

Figure 8. The SWD-related TRN steady hyperpolarization is almost completely abolished at a membrane potential close to AP threshold.

A–D, the SWD-related rhythmic cellular activity is recorded simultaneously with the EEG of the related primary somatosensory cortex (S1Cx) and is truncated (grey bar). The intracellular activity is recorded at different membrane potentials. Note that the long-lasting hyperpolarization still persists at a membrane potential of −92 mV.

Intracellular, fast rhythmic activities preceding spindle-like oscillations and normal or epileptic 5–9 Hz oscillations

As described above our intracellular recordings have revealed that, both in TC and TRN neurons, normal and SWD-related 5–9 Hz oscillations started from a significantly more depolarized membrane potential than spindle-like oscillations. Therefore, we performed a spectral analysis of the fast, synaptic and intrinsic oscillations of the membrane potential, especially in the β (16–30 Hz) and γ (31–100 Hz) frequency bands of TC and TRN neurons. The membrane potential of thalamic neurons displayed more powerful fast rhythmic activities before the occurrence of normal or epileptic 5–9 Hz oscillations than before the occurrence of spindle-like oscillations (Fig. 9A). The power of the β and γ oscillations was significantly higher in between normal or SWD-related 5–9 Hz oscillations than in between spindle-like oscillations (Fig. 9B).

Figure 9. The membrane potential of TRN and TC neurons displays more powerful fast rhythmic activities in between 5–9 Hz oscillations or SWDs than in between spindle-like oscillations.

A, intracellular recordings of spontaneously occurring membrane potential oscillations in TRN and TC neurons, which occur before the occurrence of 5–9 Hz oscillations (under neuroleptic analgesia), the occurrence of SWDs (under neuroleptic analgesia), or the occurrence of spindle-like oscillations (under barbiturate influence). These recordings were obtained while applying DC hyperpolarizing current of at least −1 nA. B, comparison of the normalized FFT values (total power) of the slow (1–15 Hz), β (16–30 Hz), γ1 (31–49 Hz), and γ2 (51–100 Hz) oscillations, which were recorded in TRN and TC neurons a few seconds before the occurrence of 5–9 Hz oscillations, of SWDs, or of 7–15 Hz oscillations. The 50 Hz values were discarded to avoid contamination from possible AC noise. Each datum point (mean ± s.e.m.) was computed from the values of 25–35 successive FFTs (epochs of 0.3 s). The data are compared using Student's t test (*P < 0.05; ns, not significant). Note that the membrane potential oscillations of the TRN and TC neurons are dominated by slow oscillations (1–15 Hz).

Discussion

The present in vivo study has demonstrated that in the rat's somatosensory TC system the two spontaneously occurring 5–9 Hz and spindle-like oscillations are based on distinct cellular mechanisms. The thalamic intracellular events occurring during spindle-like oscillations and SWDs are significantly different, which together with our earlier data (Pinault, 2003) strongly suggest that, in GAERS, CT 5–9 Hz oscillations are more pro-epileptogenic than spindle-like oscillations.

Oscillations at 5–9 Hz but not sleep spindles give rise to SWDs in GAERS

It is important to remember that, in humans, typical absence seizures are related more to arousal than to sleep (see Introduction). In undrugged GAERS, both SWDs and 5–9 Hz oscillations emerge from a relatively desynchronized EEG when compared to spindle oscillations (Pinault et al. 2001). From an EEG point of view, the neuroleptic analgesia maintains the rats in a state equivalent to inattentive and immobile wakefulness or drowsiness (Pinault et al. 2001). These experimental conditions are adequate for recording SWDs at a similar frequency of occurrence and duration to those recorded in freely moving animals (Pinault et al. 1998, 2001; Seidenbecher et al. 1998). Our data indicate that the cellular and network mechanisms underlying 5–9 Hz and spindle-like oscillations described in this study are similar to those of the undrugged intact brain.

The present study has shown that spindle-like oscillations and 5–9 Hz oscillations differ at the cellular level and show different pharmacological profiles. In rats under neuroleptic analgesia, 5–9 Hz oscillations occur in the cortex and in the thalamus in a more coherent manner than spindle-like oscillations. Pentobarbital at subanaesthetic doses abolished both SWDs and 5–9 Hz oscillations, whereas it increased the incidence of spindle-like oscillations. We have also demonstrated that the thalamic cellular mechanisms underlying 5–9 Hz and spindle-like oscillations are quite distinct, ruling out the possibility that one rhythm is a derivative of the other. However, the intracellular events underlying 5–9 Hz oscillations and SWDs have similar electrophysiological features (Pinault, 2003), suggesting that SWDs in GAERS more likely emerge from 5–9 Hz oscillations than from sleep spindles.

Why are spindle-like oscillations not pro-epileptogenic in GAERS, when they apparently contribute to SWD genesis in feline generalized penicillin epilepsy (Kostopoulos et al. 1981; Kostopoulos, 2000)? Using spectral analysis, the present study reveals that spindle-like oscillations occurred in a less coherent manner in the cortex and the thalamus in barbiturate-treated rats (i.e. in rats under neuroleptic analgesia that had received a subanaesthetic dose of pentobarbital, or in rats either lightly or deeply anaesthetized with pentobarbital and fentanyl) than 5–9 Hz oscillations (in rats under neuroleptic analgesia). The major factor that leads to synchronized spindle oscillations in the thalamus is the CT feedback in widespread thalamic regions in barbiturate-anaesthetized cats (Contreras et al. 1997). In our barbiturate-treated rats, the strength of CT connectivity was apparently not strong enough to generate synchronized discharges of high-frequency bursts in TRN cells and subsequently synchronized rebound responses in related TC neurons. This situation may be closer to the undrugged condition, since temporal synchrony in spindle-frequency oscillations between adjacent TRN neurons has not been recorded during normal sleep in rats (Marks & Roffwarg, 1993). Thus, the lack of a pro-epileptogenic effect of spindles might be the result of species-specific differences in the strength of CT synchrony between the rat and the cat. Alternatively CT synchrony during spindles may reach non-physiological levels in the penicillin model of cats.

Sleep-related TC spindle oscillation

The present data have demonstrated that the thalamic cellular and network mechanisms of spindle-like oscillations recorded in rats under neuroleptic analgesia are comparable to those of drowsiness- or sleep-related spindles recorded previously in the intact brain. Indeed, spindle-like membrane oscillations of TC and TRN neurons lasted as long as sleep spindles (< 2 s), had a periodicity similar to that of sleep spindles (2–5 s; Contreras et al. 1997), rarely occurred during a desynchronized EEG, and systematically occurred under the influence of barbiturate (Gandolfo et al. 1985; Contreras et al. 1997). In TRN cells, spindle-like oscillations were characterized by a rhythmic, threshold/subthreshold, depolarizing wave superimposed on a slow depolarizing envelope. The rhythmic depolarizing wave had electrophysiological features of EPSPs that could trigger an apparent low-threshold Ca²⁺ potential, especially at the beginning of spindle-like oscillations. These EPSPs were probably mediated by synchronized TC inputs (von Krosigk et al. 1993; Bal et al. 1995).

The depolarizing envelope, which disappears when holding the membrane potential close to AP threshold, has also been recorded in TRN neurons of barbiturate-anaesthetized cats (Mulle et al. 1986), but not in TRN cells in ferret thalamic slices (von Krosigk et al. 1993; Bal et al. 1995). It is worth mentioning that such a depolarizing envelope was not recorded in TRN cells during natural 5–9 Hz oscillations or SWDs (Pinault, 2003). The mechanism underlying this slow depolarization is unknown. One possible mechanism might involve the generation of a hyperpolarization-activated cation current, I_h, since genes coding for I_h channels are present in the TRN (Santoro et al. 2000). Another possible mechanism might involve a decrease of a K⁺ leak current similar to the one modulated by cholecystokinin (Cox et al. 1995).

In TC neurons, spindle-like oscillations were characterized by rhythmic hyperpolarizations, which appeared in parallel with a slow hyperpolarizing envelope. These intracellular events resemble those obtained earlier in the anaesthetized rat (Shosaku et al. 1989). Both the rhythmic hyperpolarization and the envelope reversed in polarity at a membrane potential close to −75 mV and also reversed to depolarizing events when recorded with chloride-filled micropipettes, suggesting that both intracellular events were mediated by the activation of GABA_A receptors. These TC hyperpolarizing events are basically similar to those recorded during spindle-like oscillations in ferret thalamic slices (von Krosigk et al. 1993; Bal et al. 1995). Since the rat's somatosensory thalamus is devoid of GABAergic interneurons (Barbaresi et al. 1986; Harris & Hendrickson, 1987), and since the TC rhythmic hyperpolarizations mirrored the TRN rhythmic threshold depolarizations, the spindle-like oscillation-related TC hyperpolarizing events were most probably mediated by the TRN-induced activation of GABA_A receptors.

Surprisingly, we never recorded a high-frequency rebound burst of APs at the end of the rhythmic hyperpolarization in our experimental conditions (in rats under neuroleptic analgesia that had received a subanaesthetic dose of pentobarbital or in rats lightly or deeply anaesthetized with pentobarbital and fentanyl). Instead, single APs could be recorded between two successive hyperpolarizations. However, all our intracellularly recorded TC neurons tested could generate an apparent low-threshold Ca²⁺ spike whatever the value of their membrane input resistance, indicating that the intracellular micropipette did not affect the biophysical properties of the recorded neurons. Since high-frequency bursts occurring at 7–15 Hz could sometimes be recorded in individual extracellular recordings of TC neurons, we think that our TC cellular sample of intracellular recordings is not large enough to include bursting TC cells. Nevertheless, our data indicate that, under our experimental conditions, the probability of generation of a low-threshold Ca²⁺ potential during spindle-like oscillations is lower than that observed during 5–9 Hz oscillations (Pinault, 2003). This finding merits further discussion because it is at variance with previous in vitro (von Krosigk et al. 1993) and in vivo (Deschênes et al. 1984; Steriade & Deschênes, 1984) recordings, which have demonstrated that, in TC neurons, the high-frequency AP discharge caused by a low-threshold Ca²⁺ potential occurs at least at some cycles during spindle-like oscillations. Two non-exclusive possibilities might explain the lack of this intrinsic potential in our intracellular recordings. It might be due to the lack of coherent oscillations between the cortex and the related thalamus, which might lead to the arrival of asynchronous reticular IPSPs in TC cells. Indeed, multi-unit recordings with glass micropipettes demonstrated asynchronous burst discharges (phase lag > 5 ms) in the TRN during spontaneously occurring spindle-like oscillations (D. Pinault, unpublished observation). The lack of synchrony in spindle-frequency burst discharges between nearby TRN neurons has also been observed in freely moving rats (Marks & Roffwarg, 1993). Knowing that adjacent TRN cells can innervate the same territories (Pinault & Deschênes, 1998), asynchronous arrival of IPSPs might have shunted or abolished the low-threshold Ca²⁺ current and subsequently the high-frequency burst in TC neurons (Ulrich & Huguenard, 1997).

The second possibility might rest on particular cellular and network properties that are specific to the somatosensory TC system of rodents. Indeed, spindle-like oscillations could be recorded in thalamic slices of the visual system of ferrets (von Krosigk et al. 1993) but not in thalamic slices of the somatosensory system of mice (Warren et al. 1994) and rats (Jacobsen et al. 2001). This difference may reflect interspecies variations and can be explained to some extent by the absence of interneurons in the somatosensory thalamic nuclei of rodents (Barbaresi et al. 1986; Harris & Hendrickson, 1987).

It is well known from in vivo and in vitro studies that the pacemaker of sleep spindles is located in the thalamus (Steriade et al. 1993). This is well supported here by the systematic occurrence of rhythmic high-frequency burst discharges in TRN cells. In addition, our single- and multi-unit extracellular recordings could reveal rhythmic peaks of synchronized firings in subsets of TC neurons during spontaneously occurring spindle-like oscillations. The generation of pacemaking activity also involves reciprocal interactions between TC and TRN neurons and high-frequency burst discharges in both cellular types (von Krosigk et al. 1993; Steriade et al. 1993).

Wake-related CT 5–9 Hz oscillation

Our previous study has revealed that, in the somatosensory system, layer VI CT neurons start to fire at 5–9 Hz a few milliseconds before related TC and TRN neurons (Pinault, 2003). Therefore the 5–9 Hz oscillation-related rhythmic depolarization recorded intracellularly in TC and TRN cells is most likely mediated by a CT EPSP barrage. This input can trigger a low-threshold Ca²⁺ spike in both cell types.

This intrinsic Ca²⁺ spike is an important factor that determines the internal pattern of the high-frequency burst. The acceleration–deceleration pattern of TRN 5–9 Hz bursts is less pronounced than the well-known acceleration–deceleration pattern of TRN spindle bursts, which is caused by the underlying low-threshold Ca²⁺ potential (Mulle et al. 1986; Spreafico et al. 1988; Avanzini et al. 1989; Bal & McCormick, 1993; Contreras et al. 1993). In addition, the instantaneous frequency within 5–9 Hz bursts is significantly lower than that in spindle bursts. On the other hand, our data demonstrate that the SWD-related TRN bursts have the highest internal frequency (when compared with the 5–9 Hz-related and spindle-related bursts), which suggest that the epilepsy-related TRN bursts are underlain by at least (Pinault, 2003) a low-threshold Ca²⁺ potential that is on average more powerful than that underlying the 5–9 Hz-related and 7–15 Hz-related TRN bursts. This might be the result of cellular hypersynchronization associated with absence-related SWDs.

In the case of TC neurons, it is well established that during spontaneously occurring 5–9 Hz oscillations, high-frequency bursts of APs underlain by a low-threshold Ca²⁺ potential do occur at times. They are more frequent in subsets of TC cells having a relatively high membrane input resistance and a presumed I_h current (Pinault, 2003). Together this pacemaker current and the recurrent burst are well known to be pro-epileptogenic (Pape, 1996; Lüthi & McCormick, 1998). Furthermore, our previous (Pinault, 2003) and present data have demonstrated that 5–9 Hz oscillations and SWDs are underlain by similar intracellular events in thalamic neurons, strongly suggesting that CT 5–9 Hz oscillations but not TC spindle-like oscillations are essential for the generation of SWDs in the somatosensory system of GAERS. It is worth mentioning that intralaminar thalamic nuclei might also contribute to the generation of absence-related SWDs but, apparently, not in a pacemaker role (Seidenbecher & Pape, 2001).

In this study we have demonstrated that the membrane potentials of TRN and TC neurons are more depolarized when they are just about to display normal or epilepsy-related 5–9 Hz oscillations (under neuroleptic analgesia) than when they are just about to display spindle-like oscillations (under barbiturate influence). Also, the membrane input resistance of these neurons is lower before the occurrence of 5–9 Hz oscillations or SWDs than before the occurrence of spindle-like oscillations. Furthermore, spectral analysis has revealed that the membrane potential oscillations of these thalamic neurons include β and γ rhythmic activities, which are more powerful in between 5 and 9 Hz oscillations or SWDs than those occurring in between spindle-like oscillations. Together, these findings suggest that TRN and TC neurons receive more incoming signals when their propensity is to oscillate at 5–9 Hz or to generate SWDs than to generate spindle-like oscillations. The majority of the excitatory inputs of these two thalamic cell types arise in layer VI of the neocortex. Thus, CT neurons might play a major role in determining the state of the membrane potential in thalamic neurons. This implies that CT cells are more active before the occurrence of 5–9 Hz oscillations or SWDs than before the occurrence of spindle-like oscillations. Indeed, tonic firing has been recorded in identified layer VI CT neurons in between 5–9 Hz oscillations or SWDs (Pinault, 2003). Our data thus indicate that desynchronized cortical activity (up state) may be a prerequisite for the initiation of 5–9 Hz oscillations and SWDs, but may not be necessary for the generation of sleep spindles.

The present study has further shown that, in TC and TRN neurons, a steady or tonic hyperpolarization develops in parallel with normal or absence-related 5–9 Hz oscillations (also see Pinault et al. 1998 and Slaght et al. 2002). At least four observations lead us to propose that this apparent steady hyperpolarization as well as normal-related or epilepsy-related 5–9 Hz oscillations are mainly mediated by layer VI CT neurons: (1) the steady hyperpolarization starts and ends with, respectively, the beginning and the end of 5–9 Hz oscillations; (2) both the steady hyperpolarization and normal-related or epilepsy-related 5–9 Hz oscillations are significantly reduced in amplitude when the membrane potential is close to AP threshold; (3) they cannot be abolished when applying strong hyperpolarizing DC current; and (4) the firing of layer VI cells becomes rhythmic with the development of 5–9 Hz oscillations (Pinault, 2003).

Therefore we propose that the apparent tonic hyperpolarization might well be a disfacilitation subsequent to an arrest of firing in excitatory inputs, especially from those originating in cortical layer VI. Indeed, layer VI CT inputs represent the greatest excitatory inputs in number and in density in the dorsal thalamus (Robertson & Rinvik, 1973; Singer, 1977), and they innervate simultaneously TC and TRN neurons (Bourassa et al. 1995). In addition, the normal or epilepsy-related 5–9 Hz TC and TRN rhythmic discharges are caused by the CT rhythmic firing (Pinault, 2003). This implies that the apparent tonic hyperpolarization in thalamic neurons would result in part from rhythmic synchronous inhibitions of layer VI CT cells. This mechanism does not exclude the contribution of potassium inwardly rectifying currents (Wilson, 1993; also see Pinault et al. 1998). Thus, a cortical up and down state sequence might play a large part in launching CT 5–9 Hz oscillations or absence-related SWDs in GAERS.

Conclusion

Despite the fact that spontaneously occurring 5–9 Hz and spindle-like oscillations share a common frequency band in the rat somatosensory TC system, the cellular mechanisms underlying these two rhythmic activities are different. More specifically, the intracellular events underlying spindle-like activity and 5–9 Hz oscillations are diametrically opposite to each other in both TC and TRN cells. Coherent 5–9 Hz oscillations in the TC system, which involve CT neurons as a leading device (Pinault, 2003), are required to trigger SWDs in GAERS, whereas when the network is in the state of spindle-like oscillations the conditions are not favourable for SWD genesis. Our findings demonstrate that for a given network, the raw EEG should be described not only in terms of frequency bands but also in terms of spatio-temporal dynamics of the cellular interactions in the system.