Cellular interactions in the rat somatosensory thalamocortical system during normal and epileptic 5–9 Hz oscillations

Abstract

In Genetic Absence Epilepsy Rats from Strasbourg (GAERS), generalized spike-and-wave (SW) discharges (5–9 SW s⁻¹) develop during quiet immobile wakefulness from a natural, medium-voltage, 5–9 Hz rhythm. This study examines the spatio-temporal dynamics of cellular interactions in the somatosensory thalamocortical system underlying the generation of normal and epileptic 5–9 Hz oscillations. Paired single-unit and multi-unit recordings between the principal elements of this circuit and intracellular recordings of thalamic, relay and reticular, neurones were conducted in neuroleptanalgesied GAERS and control, non-epileptic, rats. The identity of the recorded neurones was established following juxtacellular or intracellular marking. At least six major findings have emerged from this study. (1) In GAERS, generalized spike-and-wave discharges were correlated with synchronous rhythmic firings in related thalamic relay and reticular neurones. (2) Usually, corticothalamic discharges phase-led related relay and reticular firings. (3) A depolarizing wave emerging from a barrage of EPSPs was the cause of both relay and reticular discharges. (4) In some relay cells, which had a relatively high membrane input resistance, the depolarizing wave had the shape of a ramp, which could trigger a low-threshold Ca²⁺ spike. (5) In reticular cells, the EPSP barrage could further trigger voltage-dependent depolarizations. (6) The epilepsy-related thalamic, relay and reticular, intracellular activities were similar to the normal-related thalamic activities. Overall, these findings strongly suggest that, during absence seizures, corticothalamic neurones play a primary role in the synchronized excitation of thalamic relay and reticular neurones. The present study further suggests that absence-related spike-and-wave discharges correspond to hypersynchronous wake-related physiological oscillations.

The fundamental mechanisms that are responsible for the generation of generalized thalamocortical (TC) spike-and-wave (SW) discharges (SWD) in epileptic patients are not fully understood. Studies using a variety of experimental in vivo and in vitro preparations (Kostopoulos et al. 1981; Avoli & Gloor, 1982; von Krosigk et al. 1993; Steriade & Contreras, 1995; Blumenfeld & McCormick, 2000) have demonstrated that the cortex and the thalamus are two essential structures in the generation of SWD. These studies support the notion that sleep-related oscillations can give rise to SWD (see reviews by Gloor & Fariello, 1988; Steriade et al. 1994; Kostopoulos, 2000; McCormick and Contreras, 2001).

In contrast to previous observations, Genetic Absence Epilepsy Rats from Strasbourg (GAERS), a well characterized genetic model of human absence epilepsy, show spontaneous TC SWD (6–8 SW s⁻¹) that develop during quiet immobile wakefulness from a natural medium-voltage 5–9 Hz oscillation (Pinault et al. 2001). This TC rhythm is also recorded in non-epileptic (NE) rats (Pinault et al. 2001), the inbred control strain that is free of any spontaneous SWD (Danober et al. 1998). Transitions from medium- to high-voltage 7–11 Hz oscillations have likewise been recorded in other rodent strains (see Fig. 5 and Fig. 6 in Semba & Komisaruk, 1984). The cellular and network mechanisms responsible for the generation of generalized SWD in genetic models of absence epilepsy are not well understood.

Figure 5. Electrophysiological features of two thalamocortical neurones, A and B, of the somatosensory thalamus.

These two cells were recorded in the control non-epileptic strain. They had a membrane input resistance of 31.0 ± 8.7 and of 12.4 ± 3.3 MΩ, respectively. A1 and B1, superimposed traces of responses of the thalamocortical neurone to a square pulse of constant hyperpolarizing current of increasing intensity. In B1, note the spontaneous occurrence of identified lemniscal EPSPs. A2 and B2, rhythmic activity associated with spontaneously occurring 5–9 Hz oscillations in the related surface electrocorticogram (ECoG). In A3, the curved arrows indicate individual depolarizing potentials. The grey areas are expanded in A3 and B3. C, plot of the amplitude of the decaying voltage responses occurring during hyperpolarizing current pulses (V_sag, see A1) vs. either the initial membrane potential (MP, see A1; from 20 and 21 cells in control NE rats (open symbols) and in GAERS (filled symbols), respectively), or the peak membrane input resistance at rest (PIR; sag values measured at a MP between −80 and −90 mV).

Figure 6. Thalamocortical intracellular activities associated with epileptic and normal 5–9 Hz oscillations.

Thalamocortical intracellular activity associated with spontaneously occurring spike-and-wave discharges (A1 and A2) or normal 5–9 Hz oscillations (B1 and B2). The grey area in B2 is expanded in D. C, a rhythmic depolarizing wave-hyperpolarizing wave sequence occurring in parallel with the development of a tonic hyperpolarization associated with normal 5–9 Hz ECoG oscillations. Note that, in C, the hyperpolarizing wave progressively decreases in amplitude during the development of the tonic hyperpolarization. In its depth, a depolarizing wave triggers an apparent low-threshold Ca²⁺ spike (asterisk, expanded in Fig. 7E). E and F, temporal relationship between the thalamocortical depolarizing wave-hyperpolarizing wave sequence with the SW complex (5 superimposed sequences). The relay cell in F displays a hyperpolarization-activated depolarizing sag to square pulses of constant hyperpolarizing current whereas the cell in E does not. Every curved arrow in A!, B1, B2, C, and D indicates an EPSP barrage. The horizontal dotted line indicates the AP threshold (−58 mV) at rest. The APs are clipped in B1, C, D, E and F.

Thalamocortical neurones are relay units conveying information to the cerebral cortex (Jones, 1985). They receive two main excitatory inputs from prethalamic and layer VI corticothalamic (CT) neurones and one major inhibitory input originating from the GABAergic reticular thalamic nucleus (RTN). This latter nucleus receives excitatory inputs from TC and CT axon collaterals (Sherman & Guillery, 2001). Layer VI CT neurones innervate both RTN and TC neurones (Bourassa et al. 1995). In addition, there are no or very few intrinsic GABAergic interneurones in the somatosensory thalamus of the rat (Barbaresi et al. 1986; Harris and Hendrickson, 1987).

In neuroleptanalgesied GAERS, the patterns of extracellular TC and RTN discharges of action potentials (APs) associated with SWD resembles those observed during normal 5–9 Hz oscillations (Pinault et al. 2001). Intracellular recordings revealed that small-amplitude depolarizing potentials trigger the SW-related AP discharges in these two cellular types (Pinault et al. 1998; Slaght et al. 2002). From these data, at least three fundamental questions have emerged. (1) What is (are) the excitatory input(s) responsible for the triggering of the depolarizing potentials in thalamic, relay and reticular neurones during the generation of SWD? (2) What is the role of CT cells in the generation of the corresponding thalamic discharges? (3) Are TC and RTN discharges associated with SWD dictated by intracellular mechanisms similar to those associated with the natural 5–9 Hz rhythm? Answers to these questions are crucial for understanding the spatio-temporal cellular dynamics occurring in the TC system during the generation of normal and epileptic 5–9 Hz oscillations in rodents.

This study characterizes in vivo, in GAERS and NE rats, the triggering events of the TC and RTN extracellular discharges and the underlying membrane mechanisms associated with epileptic and normal 5–9 Hz oscillations. Because SWD mainly occur in the lateral part of the dorsal thalamus (Vergnes et al. 1990), the objectives of the present study were achieved in the somatosensory system by employing four, histologically controlled, approaches in anaesthetized GAERS and NE rats exhibiting spontaneous epileptic and normal 5–9 Hz oscillations in the bilateral frontoparietal electrocorticogram (ECoG): (1) paired extracellular single-unit recordings between the three principal elements of the TC system, (2) dual multi-unit recordings in the corresponding cortical and thalamic regions, (3) simultaneous extracellular field potential recordings of related thalamic and cortical sites, and (4) intracellular recordings of TC and RTN cells of the somatosensory system, sometimes with QX-314-filled micropipettes to further characterize threshold oscillations. Parts of this study have previously been published in abstract form (Pinault, 2001,2002).

METHODS

Animals

Experiments were conducted in adult male Wistar rats (73 GAERS, from 39th to 55th generations, and 55 NE rats, from 31st to 47th generations, 280–350 g), complying with the institutionally recommended procedures for animal use and care (Comité Régional d'Ethique pour l'Expérimentation Animale, Strasbourg). The GAERS and NE strains were born, raised and selected under standard conditions in our research unit (Faculté de Médecine, Université Louis Pasteur, Strasbourg, France), and all efforts were made to avoid animal suffering and to use the minimum number of animals necessary to produce reliable data.

Anaesthesia and surgery

All surgical procedures were made 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). The rat's rectal temperature was maintained at 37°C with a thermoregulated blanket (Fine Science Tools Inc., Heidelberg, Germany).

A tracheotomy and a catheterization of the penile vein were performed, and the animal was placed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA). A stabilizing craniotomy-duratomy was devised to improve the success rate of single-cell electrophysiology experiments. It consisted of performing an individual minute opening (diameter < 0.8 mm) of the cranium for the insertion of each micropipette and did not require cisternal drainage. Such a surgical procedure increased the precision of reaching single neurones in a target region stereotaxically, especially in deep brain structures, and considerably reduced undesirable non-neuronal rhythms (heart and respiratory beats) during intracellular recordings. The craniotomies- duratomies were carried out on the right hemisphere above the dorsal thalamus and frontoparietal cortex (Fig. 1B). To prevent dessication of the exposed cortical areas, surgical sponges were placed on the cortex; these were impregnated with warm artificial cerebrospinal fluid containing (mm): NaCl 137, KCl 3, CaCl₂ 1.3, NaH₂PO₄-H₂O 1.13, MgCl₂.6 H₂O 1, glucose 11, NaHCO₃ 2.4, and adjusted to a final pH of 7.4. Four holes were made in the skull for positioning the Ag-AgCl ECoG electrodes above the dura (Fig. 1B): two active electrodes over the somatosensory cortex, and two reference electrodes in the two parietal-occipital crests. The bilateral ECoG electrodes were connected to preamplifiers (AI 402, × 50; Axon Instruments, Union City, CA, USA).

Figure 1. Single-cell anatomo-electrophysiology of somatosensory-related corticothalamic, thalamocortical, and reticular thalamic neurones in whole animal preparation.

A1 and A2, juxtacellular staining of simultaneously recorded CT and TC neurones, respectively. The corresponding recording traces are shown in Fig. 3A. A3, partial reconstruction of the CT neurone of A1. B, schema illustrating the location of the recording micropipettes in the somatosensory thalamus (ventral posteromedial thalamic nucleus and the corresponding RTN sector) and of the Ag–AgCl ECoG electrode in the related frontoparietal cortex. The stereotaxic plates (in mm from bregma) are drawn from the Paxinos & Watson's atlas (1986). Note that minimal craniotomies were made for the insertion of the micropipettes, and that the frontoparietal cortex was not directly injured by the corresponding craniotomy. C1, antidromic activation of a recorded TC neurone of the somatosensory thalamus following layer IV electrical stimulation, which is located below the ECoG recording site. When the antidromic AP is triggered immediately after a spontaneously occurring orthodromic AP, it cannot be recorded because of collision (curved arrow). C2, orthodromic activation of a RTN cell of the somatosensory sector following layer IV electrical stimulation. Cx stim, cortical stimulation. D1 and D2, juxtacellular marking of simultaneously recorded TC and RTN neurones, respectively. CL, central lateral; CM, central medial; CPu, caudate putamen; ep, entopedoncular; ic, internal capsule; LD, lateral dorsal; MD, medial dorsal; Po, posterior thalamic nuclear group; VL, ventral lateral; VM, ventral medial; VPl, ventral posterolateral; VPm, ventral posteromedial; WM, white matter; ZI, zona incerta.

The neuroleptanalgesia was initiated before the end of the general pentobarbital-ketamine anaesthesia by an intravenous injection (0.25 ml during 10 min) of the following mixture: d-tubocurarine chloride (0.2 mg; Sigma-Aldrich, Saint-Quentin Fallavier, France), fentanyl (0.22 μg; Janssen, Boulogne-Billancourt, France), haldol (25 μg; Janssen), and glucose (12.5 mg). It was then maintained by continuous administration (0.5 ml h⁻¹) of the same mixture (quantity given per hour for a rat of 300 g): d-tubocurarine chloride (0.4 mg), fentanyl (0.425 μg), haldol (50 μg), and glucose (25 mg). (The adequacy of the neuroleptic anaesthesia was established in the absence of neuromuscular blockade). The rat was artificially ventilated (SAR-830, CWE, Ardmore, PA, USA) in the pressure mode (8–12 cmH₂O; 60–65 breaths min⁻¹). In order to maintain the values of arterial P_O2 and P_CO2 in the physiological range (90–110 and 35–40 mmHg, respectively), it was necessary to proceed during the experiment with an O₂-enriched gas mixture (70 % air-30 % O₂). The rat's rectal temperature was progressively increased to and maintained at its physiological level (38–38.3°C). The ECoG, the temperature, and the heart rate (maximum accep table 350 beats min⁻¹) were under continuous monitoring to maintain the depth of the neuroleptanalgesia steady (Flecknell, 1996). The interaural pressure was relaxed (about 0.2 mm each side). Recording sessions started about 2 h after the end of all surgical procedures. Subcutaneous infiltrations of xylocaine (2 %, Astra, Rueil-Malmaison, France) were applied every three hours at all surgical sites.

Electrophysiology

Every glass micropipette (30–70 MΩ) was filled with a solution containing 1.5 % N-(2 amino ethyl) biotin amide hydrochloride (Neurobiotin; Vector Labs, Burlingame, CA, USA) dissolved in either 1–3 m CH₃COOK, 1–3 m CH₃COOK plus 0.1 m QX-314 (Sigma-Aldrich), or in 3 m KCl. It was then lowered with a stepping micro-driver (Burleigh, Fishers, NY, USA) to reach a single somatosensory-related TC, RTN, or CT neurone, which was extracellularly and/or intracellularly recorded simultaneously with the bilateral surface ECoG of the somatosensory cortex (Fig. 1). The location of the ECoG electrodes was verified by evoked potentials (two subcutaneous needles; about 3 mA, 0.1 ms, 0.2 Hz) to forepaw electrical stimulation (Pinault et al. 2001).

Once hyperpolarizing current was no longer needed to maintain stable intracellular recordings with QX-314-containing micropipettes, a depolarizing constant current in the range +0.5 to +1.5 nA was injected continuously for at least 2 min, till the total disappearance of full-sized sodium APs. Afterwards, QX-314 continued to diffuse passively into the cell's dendritic tree and axon.

Dual extracellular single-unit recordings of somatosensory-related cells (either TC and RTN neurones, TC and cortical (layer VI) neurones, or RTN and layer VI cells) were simultaneously performed with a bilateral ECoG of the frontoparietal cortex in 21 GAERS and 24 NE rats (Fig. 1A1-A3 and D1-D2).

Paired multi-unit recordings were performed simultaneously in the somatosensory cortex (layers IV, V and VI) and related thalamic region with glass micropipettes (tip diameter: 2–3.5 μm) in conjonction with the bilateral ECoG (4 GAERS and 2 NE rats). The locations of the recording sites were identified histologically following extracellular application (500–600 nA, 200 ms on/200 ms off, 5–10 min) of Neurobiotin.

Triple extracellular field potential recordings of somatosensory-related thalamic, RTN and layer VI sites were carried out with large glass micropipette tips (diameter: 3.5–5.5 μm) in GAERS (n = 2) and in NE (n = 3) rats simultaneously with the corresponding bilateral ECoG (Fig. 1). The locations of the recording sites were identified electrophysiologically (i.e. following electrical stimulation of the peripheral receptive field), and histologically (i.e. following extracellular application (500–600 nA, 200 ms on/200 ms off, 20 min) of dextran biotin amine, lysine fixable (molecular weight = 3000, Molecular Probes, Leiden, The Netherlands) (see Fig. 4B3)).

Figure 4. Dual multi-unit (A) and triple extracellular field potential (B) recordings of somatosensory-related thalamic and cortical sites along with spontaneously occurring spike-and-wave discharges (A and B1) or normal 5–9 Hz oscillations (B2) in the ECoG.

A1–4, a typical example of paired multi-unit (bandpass: 0.3–6 kHz), thalamic and cortical, recordings. The thalamic recording is performed in the somatosensory thalamus (staining not shown). The intracortical recordings are performed at different depths below the ECoG recording site (A2). The last intracortical recording corresponds to the deeper site, which is marked following extracellular micro-iontophoretic application of Neurobiotin (A1). A2, the drawing indicates two sites recorded from in layer V and layer VI (asterisks). The corresponding records are shown in A3 and in A4, respectively. A5, mounting of three time-matching superimpositions of layer V, layer VI, and thalamic multi-unit recordings from another experiment. B3, evoked potentials following electrical stimulation of the contralateral forepaw. The photomicrographs (up = dorsal; right = lateral) show large extracellular applications (epicentre indicated by a white asterisk) of dextran biotin amine where extracellular field potential recordings were carried out simultaneously in the somatosensory system, that is, from top to bottom, in layer VI, in the thalamus, and in the reticular thalamic nucleus (RTN). In B1 and B2, the bandpass was 0.1 Hz–6kHz). CPu, caudate putamen; ic, internal capsule; VPl, ventral posterolateral thalamic nucleus; VPm, ventral posteromedial thalamic nucleus; WM, white matter.

Identification of functionally connected cortical and thalamic regions

All cortical and thalamic recordings were performed in the primary somatosensory system, as ascertained following identification of the receptive field and/or by histological controls (see Fig. 4A1 and B3). The bilateral ECoG electrodes were placed over the forepaw representation, which could be verified by evoked potentials (two subcutaneous needles; about 3 mA, 0.1 ms, 0.2 Hz) to forepaw electrical stimulation (Fig. 4B3). The thalamic recording sites were positioned in a zone (2.8 ± 0.2 mm behind the bregma) where thalamic neurones could be activated antidromically and orthodromically with bipolar stimulating semi-microelectrodes (SNE-200, Rhodes Medical Instruments, Inc.) positioned (layer IV) just below the ECoG electrode (4 rats; Fig. 1C1 and 2) and where anterograde staining was observed following extracellular application of Neurobiotin in layer VI, perpendicularly below the ECoG electrode (2 rats, Fig. 4A1). All intracortical recording sites were located perpendicularly below (horizontal range: ± 500 μm) the ECoG electrode (Fig. 1B and Fig. 4A1 and 2).

Histology

At the end of the recording session, the units were individually labelled using the juxtacellular (Pinault, 1996) or intracellular tracer microiontophoresis technique for standard histological identification (Fig. 1A1, A2, D1 and D2). In any dual single-unit recording experiment, the juxtacellular filling procedure was applied only in the two neurones of the last pair recorded from. After a survival period of more than 1 h, the animal was humanely killed with an overdose of pentobarbital applied intravenously. Then it was transcardially perfused with a fixative containing 4 % paraformaldehyde and 0.5 % glutaraldehyde in 10 mm phosphate buffer saline, and the brain tissue was processed using standard histological techniques for retrieving the tracer-filled neurones.

Signal conditioning

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

Data analysis

Electrophysiological recordings were analysed with the pClamp software, and the tracer-filled neurones were examined with a light microscope (E600, Nikon France, Champigny-sur-Marne, France). Some of the neurones were reconstructed using the Neurolucida system (Microbrigthfield, Colchester, VT, USA) (Fig. 1A3). The location of any marked cell was ascertained by consulting a stereotaxic atlas (Paxinos & Watson, 1986).

Cross-correlograms of paired AP discharges were performed during at least three distinct spontaneous episodes of normal or epileptic 5–9 Hz ECoG oscillations. To assess eventual spontaneous fluctuations of the time relationships between the firings of the two recorded neurones during a given ECoG oscillation, it was necessary to perform two types of analysis: one lengthy (up to 5 s) and the other sequential (each analysis segment < 2 s). They were derived with 2-ms resolution using the Common Processing Analysis Software (DataWave Technologies, Longmont, CO, USA). This program was also used for performing fast Fourier transformations and auto-correlograms of 1–5 s of ECoG or intracellular segments (redigitized at a sampling rate of 150–500 Hz). In the results section data are presented as mean ±s.d.

RESULTS

In both GAERS and NE rats, the fast low-voltage bilateral background ECoG activity was spontaneously interrupted by short-lasting (< 10 s) episodes of medium-voltage (< 0.5 mV) 5–9 Hz oscillations (Fig. 2A1). This rhythm was recorded during quiet immobile wakefulness. While in GAERS and in NE rats normal oscillations slowly waxed and waned, in GAERS they usually led to the occurrence of SWD (about 5–9 SW s⁻¹; Fig. 2B1) accompanied with whisker twitching (Vergnes et al. 1982). Medium-voltage 5–9 Hz oscillations could resume at the end of the seizures (Fig. 3A1 and B1). Detailed characteristics of this natural rhythm are given by Pinault et al. (2001).

Figure 2. Simultaneous thalamocortical and reticular thalamic extracellular activities associated with spontaneous medium-voltage (A) or high-voltage (B) 5–9 Hz ECoG oscillations.

Recordings in A and B are from the same TC–RTN pair in a GAERS. The framed traces (in grey in A1 and B1) are expanded in A2 and B2. C and D, superimposition of four representative cross-correlograms (2 ms resolution) computed from four paired recordings obtained during normal (C) or epileptic (D) 5–9 Hz oscillations. ipsi, ipsilateral; contra, contralateral.

Figure 3. Dual recordings of a corticothalamic neurone with either a thalamocortical (A1–A3 and D), or a reticular thalamic (B1–B3 and E) cell during spontaneous spike-and-wave discharges.

A2 and B2, five superimposed successive recordings of SW-related AP discharges. A3, B3, cross-correlograms (2 ms resolution) computed from pair recordings. C, means and standard deviations of the time relationship between the CT, TC, and RTN AP discharges and the SW complex (CT: −19.4 ± 8.8 ms, 88 SW complexes from 4 rats; RTN: −11.9 ± 7.3 ms, 108 SW complexes from 6 rats; TC: −12.6 ± 8.2 ms, 133 SW complexes, 7 rats. In D, the asterisks indicate the beginning of the CT rhythmic discharge, which starts before both the TC rhythmic firing and the occurrence of the seizure.

Rhythmic TC and RTN discharges occur in synchrony during generalized SWD

Understanding the functional links of the three principal structures (thalamus, RTN, and cortex) of the somatosensory TC system during the generation of normal and epileptic 5–9 Hz oscillations required simultaneous recordings of the corresponding neurones. A first series of experiments aimed at recording pairs of TC and RTN neurones in both GAERS (16 pairs) and NE rats (17 pairs).

During the interseizure episodes in GAERS, as in NE rats, the recorded spontaneous extracellular firings of TC and RTN cells commonly shared irregular patterns. Relay discharges were composed of single APs (up to 25 APs s⁻¹), occasionally replaced by high-frequency (up to 500 Hz) bursts of a few APs (0.1–8 bursts s⁻¹). Reticular discharges were also composed of irregularly occurring single APs (up to 35 APs s⁻¹), which were on occasion replaced by high-frequency (up to 500 Hz) bursts of APs (up to 15 APs; 0.1–16 bursts s⁻¹).

During the generation of natural 5–9 Hz ECoG oscillations (Fig. 2A1) or SWD (Fig. 2B1), the TC and RTN ongoing firings slowed down and exhibited a rhythmic pattern of single APs (in TC cells) or high-frequency bursts (200–500 Hz; up to 4 and up to 13 APs in TC and in RTN cells, respectively). The probability of occurrence of an AP discharge during every cycle of the TC or RTN extracellular field potential oscillation that was associated with SWD was usually greater than that accompanying the natural 5–9 Hz rhythm (compare Fig. 2A with Fig. 2B; see also Pinault et al. 2001). During SWD, the patterns of TC and RTN discharges were more stereotyped than those observed during normal 5–9 Hz oscillations (compare Fig. 2A and C with Fig. 2B and D).

Cross-correlation analyses revealed that TC and RTN discharges had the propensity to occur in much more synchronous and phase-locked manners during epileptic than during natural 5–9 Hz oscillations (compare Fig. 2C with Fig. 2D). On average, rhythmic discharges occurred in TC and RTN cells about 12 ms before the spike component of the SW complex (TC: 12.3 ± 5.8 ms; RTN: 12.0 ± 4.9 ms; n = 337 SW complexes from 16 pairs). The corresponding anatomical data established that these higher-degree synchronizations were not related to an apparent direct synaptic connection (TC → RTN) between the recorded neurones (not shown), suggesting that they were driven by a common excitatory input.

CT discharges phase-lead related TC and RTN discharges during SWD

Because lamina layer VI CT cells innervate both the relay and the reticular neurones, it was also important to obtain dual single-unit recordings in the somatosensory system between layer VI and thalamic, relay or reticular, neurones during normal and epileptic 5–9 Hz ECoG oscillations. Paired recordings carried out included five TC-layer VI and six RTN-layer VI pairs in GAERS, and five TC-layer VI and five RTN-layer VI pairs in NE rats. The thalamic-layer VI data concerning histologically identified pyramidal neurones (see Fig. 1A1-A3) included six pairs (3 TC-VI and 3 RTN-VI) in GAERS and two (1 TC-VI and 1 RTN-VI) in NE rats. From these eight pyramidal cells, five (from 4 GAERS and from 1 NE rat) were identified as CT neurones on the basis of morphological criteria (Zhang and Deschênes, 1997), which included an axon running into the striatum.

In both GAERS and NE rats, spontaneous extracellular CT discharges that were associated with episodes of desynchronized ECoG oscillations reflected irregular emissions of APs (from about 0.1 to 35 APs s⁻¹; Fig. 3A1 and B1). Spike-and-wave discharges and normal 5–9 Hz ECoG oscillations were associated with a slow-down of the ongoing firing of the recorded CT neurones, during which rhythmic discharges of 1–5 APs (intra-burst frequency 50–260 Hz) occurred (Fig. 3A2 and B2).

In all instances during SWD in GAERS, SW-related CT discharges had a clear propensity to occur a few milliseconds earlier than TC or RTN discharges (Fig. 3B2 and 3). Corticothalamic discharges appeared on average 19.4 ± 8.8 ms (from + 5 to −35 ms, n = 88 from 4 cells, 4 rats) before the spike component of the SW complex whereas TC and RTN discharges occurred simultaneously about 12 ms (during the same recording sessions TC: 12.6 ± 8.2 ms, n = 133 from 7 cells; RTN: 11.9 ± 7.3 ms, n = 108 from 6 cells) before the spike component (Fig. 3C). Again, all the other recorded (local-circuit and projecting non-CT) cells likewise phase-led related TC or RTN discharges (20.8 ± 13.2 ms before the spike component of the SW complex, n = 224 from 13 units).

The temporal relationship between the discharges of CT and TC neurones could not be determined during natural 5–9 Hz oscillations because of the low firing rates. However, cross-correlation analyses of CT-RTN recordings revealed a clear trend toward synchrony (±50 ms), with little variation of the middle peak position from one episode to another (data not shown).

Furthermore, during single-unit extracellular recordings of CT-TC or CT-RTN pairs, 5–9 Hz oscillations could first be recorded in both the surface ECoG and the layer VI extracellular field potential (Fig. 3B1 and D). Moreover, in five out of eight recorded neuronal pairs (2 CT-TC and 3 CT-RTN), rhythmic AP discharges could be recorded in CT neurones before the occurrence of rhythmic AP discharges in related TC or RTN neurones (Fig. 3D). In addition, CT cells could even display rhythmic AP discharges a few hundreds of milliseconds to a few seconds before rhythmic TC and RTN discharges (Fig. 3D and E). The frequency of the CT rhythmic AP discharges occurring before onset of SWD was sometimes slightly higher (6–12 Hz) than that occurring during SWD (Fig. 3E). In addition, the AP frequency in the ongoing firing of CT cells could slow down before the arrival of 5–9 Hz oscillations in the surface ECoG and before the associated TC and RTN slow downs (data not shown). Short-lasting (< 1 s) 5–9 Hz oscillations in the surface ECoG could even be accompanied by a significant inhibition of the ongoing firing of CT neurones while corresponding TC or RTN firing was apparently not affected (data not shown).

Because the recorded CT cells did not fire on each cycle of the rhythmic activity and had a more variable time relationship with the spike component of the SW complex than did thalamic neurones (Fig. 3), it was necessary to sample cortical activities at the multi-unit level. In so doing, we confirmed in all experiments (n = 4) that SW-related firing usually started earlier in layer VI cells than in related thalamic neurones and that layer IV/V cells had a clear tendency to discharge after those thalamic neurones (Fig. 4A). Contrary to layer VI cells, layer IV/V neurones almost always fired on each cycle of the SWD. On the other hand, in control NE rats these latter neurones did not always fire during each cycle of the natural 5–9 Hz oscillation (data not shown). Again, when recording the extracellular field potentials of electrophysiologically and histologically identified (Fig. 4B3), somatosensory-related thalamic, RTN and layer VI sites, 5–9 Hz extracellular field potential oscillations were seen to start first in layer VI, sometimes even several seconds before their occurrence in the corresponding thalamus (Fig. 4B1 and 2).

EPSP barrages trigger TC and RTN discharges

During the development of SWD, CT discharges usually preceded related TC and RTN discharges. Therefore, it was crucial to seek in thalamic units the intracellular correlates associated with spontaneously occurring generalized SWD (in GAERS) or medium-voltage 5–9 Hz oscillations (in GAERS and NE rats).

Sixty-one RTN cells (41 in GAERS and 20 in NE rats) and 85 TC neurones (46 in GAERS and 39 in NE rats) were recorded intracellularly. The RTN data reported herein are based on the recording of 28 (from GAERS) and 11 (from NE rats) neurones, which had a stable resting membrane potential (−60 ± 3 mV and −60 ± 5 mV, respectively) and AP amplitude up to or above 50 mV. In these reticular cells, an apparent low-threshold Ca²⁺ spike could be evoked (see Fig. 10C). They had a membrane input resistance of 22 ± 10 MΩ in GAERS and of 26 ± 8 MΩ in NE rats.

Figure 10. Voltage-dependent features of RTN depolarizing wave-hyperpolarizing wave sequence.

A and B1-B4, in a typical experiment, the intracellular current (top trace in A) was progressively increased from −0.41 to +1 nA (100 ms current pulses of −0.4 nA every 2 s). The record is truncated (area in grey) for clarity. Four depolarizing waves topped or not by action potentials, indicated by asterisks in A, are expanded below (B1-B4). The second one (B2) is overlaid at the same temporal scale with two barrages of EPSPs, one (•) that occurs at the same membrane potential and does not reach the triggering threshold of a depolarizing hump crowned by action potentials, the second (curved arrow) that is the EPSP barrage recorded in Fig. 11A3 with a QX-314-filled micropipette. C, triggering of a low-threshold spike topped by a high-frequency AP burst at the offset of a square pulse of constant hyperpolarizing current. D1 and D2, five superimposed extracellular and intracellular traces of the same RTN neurone, respectively, which occur in phase with a spontaneous SW complex.

The intracellular TC data are based on the recording of 21 (GAERS) and 20 (NE rats) neurones, which had a stable resting membrane potential (−58 ± 3 mV and −59 ± 1 mV, respectively) and AP amplitude above 50 mV. All these recorded cells could exhibit a low-threshold spike (see Fig. 5A1 and B1), which was similar in waveform to an apparent low-threshold Ca²⁺ spike (Deschênes et al. 1984). They had an average input resistance of 25 ± 8 MΩ in GAERS and of 23 ± 7 MΩ in NE rats.

Two main types of TC neurones could be distinguished on the basis of their ability to generate a slow inward rectification in the depth of hyperpolarizations induced by square pulses of constant current (Fig. 5A1 and B1). Furthermore, the TC cells showing the hyperpolarization-activated depolarizing sag usually had a higher input resistance than those having no apparent sag (Fig. 5C). The amplitude of the depolarizing sag (9.0 ± 3.4 mV with an initial membrane potential within the range −80 to −90 mV, n = 23) increased with increasing hyperpolarization (Fig. 5C). Most of the TC units having the sag displayed, after an apparent low-threshold Ca²⁺ spike, an after-depolarization that was more important in duration than that of units having no sag (not shown). For each type, no significant difference was observed between both rat strains.

TC neurones

The occurrence of epileptic or normal 5–9 Hz oscillations in TC neurones was correlated with the sudden or progressive replacement of the ongoing low-voltage fast membrane potential oscillations by a rhythmic (5–9 Hz) activity, which was superimposed on a steady or tonic hyperpolarization (Fig. 6). This long-lasting hyperpolarization was usually deeper in TC cells that showed the hyperpolarization-activated depolarizing sag than in those that did not (ΔV up to about −20 mV vs.−10 mV; compare Fig. 5A2 with Fig. 5B2). The epilepsy-related cellular oscillations were longer in duration and more stereotyped than the normal-related oscillations. Whatever the cellular type, the rhythmic activity mainly consisted of the recurrence of a depolarizing wave- hyperpolarizing wave sequence, the former wave triggering 0–4 APs. The time course of this sequence was, however, variable from one TC neurone to another, to all appearances depending on the membrane input resistance, which was highly correlated with the amplitude of the hyperpolarization-activated depolarizing sag (Fig. 5C). For instance, the neurones in Figs 5A, 6A2 and 6B2 had a relatively high input resistance (25–40 MΩ) and exhibited a hyperpolarization-activated inward rectification during hyperpolarizing current pulses whereas those in Figs 5B, 6A1 and 6B1 had a lower input resistance (< 25 MΩ) and did not display an apparent sag.

In any recorded TC cell the depolarizing wave apparently included a barrage of EPSPs (duration up to 85 ms). Indeed, such a barrage seemed to be the summation of successive unitary synaptic depolarizing events (Fig. 7B-F) whose amplitude increased with hyperpolarization and whose frequency of occurrence did not seem to depend on the polarization level of the recorded neurone. Furthermore, the depolarizing wave amplitude increased with hyperpolarization and decreased with depolarization, and its frequency of occurrence was not affected by injection of sustained hyperpolarizing or depolarizing currents (Fig. 7A and C). To further characterize the EPSP barrages, intracellular recordings with QX-314-filled micropipettes were conducted in seven TC neurones. The amplitude of the depolarizing wave was abolished at a membrane potential above 0 mV (4 cells; typical example in Fig. 7C). In GAERS, the EPSP barrage usually developed at the apparent onset of or during the spike component of the ECoG SW complex (Fig. 6E and F).

Figure 7. The thalamocortical depolarizing wave includes a barrage of EPSPs.

A, response of a TC neurone to a hyperpolarizing ramp current applied just before the end of a spontaneous SWD. The rhythmic hyperpolarizing wave reverses at −69 mV. B, recorded at two different membrane potentials, two successive depolarizing wave-hyperpolarizing wave sequences associated with normal 5–9 Hz ECoG oscillations. Individual depolarizing potentials (straight arrows) are revealed in the depth of hyperpolarizations. This TC cell does not display an apparent depolarizing sag during the application of a hyperpolarizing current pulse. C, three SWD-related traces recorded at different membrane potentials with a QX-314-filled micropipette. In B and C, every curved arrow indicates either an EPSP barrage (e), an IPSP barrage (i), or an EPSP-IPSP barrage (e + i). D1, three successive depolarizing waves, the ascending slope of one of these occurring during the injection of a 100 ms square pulse constant hyperpolarizing current (−0.2 nA). D2, three successive depolarizing waves (from neurone of Fig. 6B2) recorded while injecting a sustained hyperpolarizing current (−0.5 nA). E, initial part of a depolarizing wave (the one indicated by an asterisk in Fig. 6C), which apparently includes individual synaptic depolarizing potentials (straight arrows) and a low-threshold Ca²⁺ spike. F, a typical depolarizing wave recorded with a QX-filled micropipette. G, typical TC intracellular events (action potentials clipped for clarity) occurring during mechanical stimulation of the receptive field. On the rat is shown the location of the receptive field (in black) gently activated by the experimenter's hand.

The recurrent character of the depolarizing wave was visible from the earliest stage of the ECoG 5–9 Hz oscillations, often before the full development of the associated steady hyperpolarization (Fig. 6). When this latter was sufficiently deep, the depolarizing wave was capable of triggering an apparent low-threshold Ca²⁺ spike (Fig. 5A3 and Fig. 6C and D). It seemed that the strength of such spike was variable from one to another cell (compare Fig. 5A1 with Fig. 5B1). In TC cells displaying an apparent depolarizing sag during a hyperpolarizing pulse of constant current, the wave coincided with a ramp-shaped depolarization, which could trigger an apparent Ca²⁺ spike (Figs 6B2 and D and 7D1 and 2). Although it was difficult to determine whether the onset of the rhythmic activity was either a depolarizing or hyperpolarizing wave, the former wave was very often the first in a majority of the recorded TC neurones whatever the cellular type (15/21 in GAERS, and 13/20 in NE rats).

Were the barrage EPSPs different from prethalamic EPSPs? Yes. Indeed, spontaneously occurring TC activities were sometimes accompanied by lemniscal EPSPs, which were identifiable following mechanical stimulation of the receptive field (Fig. 7G). Such prethalamic EPSPs most often remained subthreshold during the rhythmic activity without altering its time course, at the very most smoothing it (not shown). Their frequency of occurrence could be increased following activation of the receptive field without changing the internal frequency of the epilepsy-related TC rhythmic activity (not shown). In contrast to the unitary EPSPs that were included in the depolarizing wave, lemniscal EPSPs were identifiable near the AP triggering threshold (Fig. 7G).

RTN cells

Spike-and-wave discharges or natural 5–9 Hz oscillations in RTN cells were associated with a long lasting hyperpolarization from which a series of prominent rhythmic depolarizing waves were generated, which were often crowned by AP bursts (2–13 APs at 200–500 Hz; Fig. 9). The long-lasting hyperpolarization could also be seen as the result of a recurrent hyperpolarizing wave, which followed the depolarizing wave (see below). During the natural oscillations, this sequence of events was similar but much less stereotyped than that observed during SWD. Also, the recurring depolarizing wave was more variable in duration than the epilepsy-related recurrent depolarization (Fig. 9); in addition, it could become subthreshold while the rhythmic activity was fully developed (Fig. 9B1). The normal-related depolarizing wave was on average longer than the epilepsy-related wave (up to 180 ms vs.110 ms; compare Fig. 9A2 with B2, and Fig. 11C1 with C2). In GAERS, the recurring depolarizing wave occurred in phase with the SW complex (Fig. 10D1 and 2), beginning usually either before (5–30 ms) its apparent onset, or during its spike component, and lasting just as long (80–110 ms).

Figure 9. Reticular intracellular activity associated with spontaneously occurring epileptic (A) or normal (B) 5–9 Hz oscillations.

The curved arrows indicate barrages of EPSPs. In A1, a current square pulse of −0.6 nA was delivered every 2 s (top record), and the grey area is expanded in A2. In B1, a current square pulse of −0.2 nA was delivered every 2 s. The recording traces in B1 and B2 are from two RTN cells. In A2 and B2, the arrowhead indicates the sudden occurrence of a prominent hyperpolarization that precedes a threshold large depolarizing wave. A3, B3, when relatively more hyperpolarized, the EPSP barrages are more evident. The horizontal dotted line indicates the AP threshold (−58 mV) at rest. The APs are clipped in A2 and 3, and in B2 and 3.

Figure 11. Intrinsic and synaptic features of the reticular depolarizing wave.

A1-A3, SW-related RTN depolarizing events recorded about 9 min after impalement with a QX-314-filled micropipette: a barrage of EPSPs (left depolarizing events in A1, A2 and A3), an EPSP barrage triggering, in an all-or-none fashion, a high-threshold (right depolarization in A2) or low-threshold (right depolarization in A3) depolarizing spike. B1 and B2, typical individual reticular EPSPs (B1, curved arrow) and a barrage of similar EPSPs B2, curved arrow), which were recorded with a QX-314-filled micropipette before and during a SWD, respectively. Both records are at the same membrane potential (−53 mV, 0 nA). These depolarizing events and the records in A1-A3 are from the same neurone. B3, recorded in a TC cell with a QX-314-filled micropipette, a SW-related barrage of EPSPs triggers an apparent low-threshold Ca²⁺ spike (membrane potential: −76 mV). Curved arrows indicate individual EPSPs, which resemble individual EPSPs recorded in RTN cells (B1 and B2). C1 and C2, comparison of two typical recurring threshold depolarizing waves, which were recorded during high-voltage (C1) or medium-voltage (C2) 5–9 Hz ECoG oscillations with a KAc- (upper traces) or QX-314-filled (lower traces) micropipette. The asterisks in A2, A3, and in C1 indicate a late hyperpolarization that probably results from a Ca²⁺-activated K⁺ current. The APs are truncated in C1 and in C2 for clarity.

The depolarizing wave seemingly included a barrage of EPSPs. Its ascending slope had an indented time course and often ended with a depolarizing hump and tail (Fig. 10B1–4 and D2). Its amplitude (up to 35 mV) increased during the development of the rhythmic activity (Fig. 9A3 and B3), was reduced with depolarization, and its frequency of occurrence was not modified when injecting sustained hyperpolarizing or depolarizing currents (Fig. 10A). Such a barrage was more evident during the earliest stage of the rhythmic activity because it appeared as the summation of unitary depolarizing events and its amplitude increased in parallel with the development of the long-lasting hyperpolarization (Fig. 9A2, A3 and B3). At the apparent onset of the rhythmic activity, the first threshold depolarizing wave triggered an AP burst that was often far less robust than those crowning the following prominent depolarizations (Fig. 9A2, A3, B2 and B3). These prolonged depolarizations reached maximal amplitude after a few cycles of the rhythmic activity. Furthermore, the depolarizing wave could trigger a low-threshold potential in the hyperpolarization trough (Fig. 10B2). This potential resembled a depolarizing spike triggered in an all-or-none fashion at the offset of a hyperpolarizing pulse (Fig. 10C), like a low-threshold Ca²⁺ spike (Mulle et al. 1986; Avanzini et al. 1989).

To further characterize the depolarizing wave, we injected intracellularly QX-314 molecules (> 8 min; see Methods for details) into eight RTN cells (5 in GAERS and 3 in NE rats). This way it was possible to decompose the depolarization into at least three components: an initial barrage of EPSPs (Fig. 10B2 and Fig. 11A1–3), whose amplitude could nearly be abolished at membrane potentials more positive than −20 mV (Fig. 11A1), followed by QX-resistant low-threshold spike (Fig. 11A3) and/or a high-threshold spike (Fig. 11A2), and by a QX-sensitive depolarization that was selectively cancelled by the lidocaine derivative (Fig. 11C1 and 2). With QX-filled micropipettes, the duration of the rhythmic depolarization was reduced about one-third to a half from that recorded with CH₃COOK-filled micropipettes (Fig. 11C1 and 2). The initial EPSP barrage (Fig. 11B2) was composed of several individual EPSPs (n > 2 at 60–400 Hz), which were similar in amplitude and waveform to individual EPSPs that contributed to the fast membrane potential oscillations (Fig. 11B1). Also, these EPSPs resembled EPSPs recorded with QX-314 filled micropipettes in four identified TC neurones which, either occasionally (in a spontaneous fashion) or when applying sustained hyperpolarizing constant current, could trigger an apparent low-threshold Ca²⁺ spike during the occurrence of a SW complex (Fig. 11B3).

Relay and reticular discharges are followed by a hyperpolarizing wave

In both TC and RTN neurones, the epilepsy- or normal-related rhythmic activity consisted mainly of recurrent AP discharges underlain by a depolarizing wave, which was followed by a hyperpolarizing wave. To further characterize this rhythmic hyperpolarization, intracellular recordings with either KAc-, KCl- or QX-filled micropipettes were performed at different membrane potentials.

TC neurones

As with the SWD (see also Pinault et al. 1998), the normal-related hyperpolarizing wave in TC neurones was probably mediated by activation of GABA_A receptors by RTN cells. Indeed, its polarity was reversed at about −70 mV (−68 to −73 mV; Fig. 7A) and became depolarizing when recording with KCl-filled micropipettes (not shown). This hyperpolarizing wave was apparently a barrage of GABA_A IPSPs, which were very likely to have been generated by a presynaptic RTN burst of APs (Fig. 8). Moreover, such successive unitary IPSPs (up to 10) occurred at 200–500 Hz with an accelerating- decelerating pattern like that observed within the high-frequency AP burst discharges of RTN cells (Mulle et al. 1986; Avanzini et al. 1989). The decay of the hyperpolarizing wave had a time course that was variable from one cell to another, probably because of the efficacy of the recorded TC neurones in generating a hyperpolarization-activated inward rectification and/or a low-threshold Ca²⁺ spike (see Fig. 5 and Fig. 6).

Figure 8. High-frequency AP bursts in RTN cells are responsible for the occurrence of barrage of IPSPs (arrows) in TC neurones.

From top to bottom are shown: a typical individual threshold depolarizing wave-hyperpolarizing wave sequence recorded in a TC neurone (APs cut for clarity), and a typical extracellular RTN high-frequency burst of APs. Note that the acceleration-deceleration patterns of the IPSPs barrage and of the RTN burst are similar. The time scale bar is valid for both traces.

RTN neurones

During both the SWD and the normal 5–9 Hz oscillations, in RTN neurones the hyperpolarizing wave could be virtually abolished at a membrane potential close to the AP threshold with KAc-filled micropipettes (Fig. 10), and fully abolished at more positive membrane potential, with a QX-filled micropipette (not shown). During SWD, QX experiments further revealed a relatively short-lasting (about 50 ms) hyperpolarization that followed the QX-resistant depolarization (> 10 min following onset of impalement), which was less prominent at a membrane potential below −88 mV (asterisks in Fig. 11A2 and 3). This hyperpolarization resembled a Ca²⁺-dependent K⁺ outward current, which had been recorded in the cat with QX-containing micropipettes (Mulle et al. 1986) and characterized in vitro (Avanzini et al. 1989).

Such QX-resistant short-lasting hyperpolarization could not be convincingly revealed during normal 5–9 Hz oscillations. This was seemingly due to the spontaneous occurrence of numerous, small-amplitude, membrane depolarizing potentials. Moreover, in contrast to the epilepsy-related recurring depolarizing wave-hyperpolarizing wave sequence, that associated with the normal 5–9 Hz rhythm was notably mixed with numerous depolarizing potentials (compare Fig. 9A1 and 3 with Fig. 9B1 and 3). These potentials were presumably synaptic events because they were more evident during deep hyperpolarizations. It was conspicuous that they made the time course of the normal-related sequence with more synaptic noise than the epilepsy-related sequence, probably reflecting the low- and high-degree synchronizations of presynaptic inputs, respectively.

DISCUSSION

By the use of modern in vivo single-cell anatomo-electrophysiological methods, this study has unravelled novel thalamic cellular and network mechanisms underlying the generation of SWD in a well recognized genetic model of absence epilepsy. The main observations are as follows. First, in the somatosensory system, rhythmic TC and RTN AP discharges occur in synchrony during generalized SWD. Second, layer VI CT rhythmic firing usually phase-leads related thalamic, relay and reticular, discharges. Third, both discharges are a consequence of a depolarizing wave beginning by a barrage of EPSPs, which is very likely induced by layer VI CT neurones. Fourth, in TC cells displaying a hyperpolarization-activated ramp-shaped depolarization during the application of hyperpolarizing current pulses, the EPSP barrage coincides with a post-inhibitory ramp-shaped depolarization, which can trigger an apparent low-threshold Ca²⁺ spike. Fifth, in RTN cells the EPSP barrage can further trigger apparent voltage-dependent depolarizations. Sixth, TC and RTN neurones in both GAERS and NE rats have similar passive membrane properties, and epilepsy-related intracellular events are basically similar to those correlated with the natural 5–9 Hz rhythm.

Technical considerations

Using in vivo paired single-cell recordings to define the time relationships between somatosensory-related cortical and thalamic regions is a challenging problem, for at least two reasons: (1) whether the two neurones of any recorded pair belong to functionally connected cortical and thalamic zones, and (2) the electrical behaviour of any single neurone is not always representative of the activity of the cellular population to which it belongs.

The absence of cisternal drainage and the performance of minute craniotomies-duratomies allow for reliable single-cell anatomo-electrophysiological investigation of living intact brain networks (see Methods). It prevents significant outflow of the cerebrospinal fluid, keeping the brain volume constant in the cranial cavity, preventing tissue slump, and avoiding swelling and oedema. These surgical conditions permit great precision recordings from intracerebral stereotaxic targets (±200 μm) and the obtaining of homogeneous cellular samples.

Although control experiments strongly indicate that my paired recordings were conducted in functionally connected thalamic and cortical regions (see Methods), the corresponding neurones were always located approximately in the same region of the somatosensory system (centred at the forepaw area) and were not systematically in register in a point-to-point manner. Moreover, the electrophysiological (spike-triggered averaging) and anatomical cellular data provide no evidence that the two neurones of any pair were in direct synaptic connection, even when they had approximately the same receptive field. In spite of this, during paired cortical and thalamic recordings, the thalamic micropipette could be moved up or down, within a range of ±350 μm, while keeping the location of the intracortical micropipette stable, without revealing significant variations in the temporal relationships between cortical and thalamic firings. However, corticothalamic cells have a more variable temporal relation with the ECoG SW complex than that shown by related thalamic neurones. That is why multi-unit recordings were used to corroborate in compelling fashion that the rhythmic firing of layer VI cells indeed precedes related thalamic AP discharges.

Relay cellular events during normal and epileptic 5–9 Hz oscillations

The intracellular recordings of the present study reveal that any TC depolarizing wave always includes an EPSP barrage, which can concur with apparent intrinsic depolarizations. The recurrent EPSP barrage might primarily be generated by CT cellular discharges. Indeed, the paired TC-CT recordings of the present study show that epilepsy-related CT firing usually phase-leads at the millisecond scale the corresponding thalamic discharges. Under the experimental conditions of this study, spontaneous or induced prethalamic (especially lemniscal) EPSPs can occur without altering the time course of the TC rhythmic activity. Furthermore, rhythmic EPSP barrages obviously occur during the earliest stage of the normal or epileptic oscillation and precede the occurrence of apparent intrinsic activities (see below). The recordings further reveal that, in the midst of the steady hyperpolarization, the intrinsic Ca²⁺ spike and the EPSP barrage can concur at some cycles but as a rule the barrage occurs systematically earlier than the low-threshold Ca²⁺ spike.

Importantly, we have identified a cellular type having the ability to generate a post-hyperpolarization ramp-shaped depolarization. It increases in amplitude with increasing hyperpolarization, resembling the depolarizing sag underlain by a H-current (McCormick & Pape, 1990; Soltesz et al. 1991). In agreement with previous in vitro (Turner et al. 1997) and in vivo (Pinault et al. 1998) observations, the present findings thus strongly suggest that the somatosensory thalamus contains two distinct TC types. Furthermore, the one that displays the sag has an average input resistance higher than the one that does not. However, because in in vivo conditions both the activities of the synaptic inputs and the content of the extracellular space are not under the experimenter's control, one may not exclude the existence of a continuum between these two apparently distinct types. Indeed, whether recorded TC cells that do not apparently exhibit a depolarizing sag in the trough of a current-induced hyperpolarization do not have functional H-channels is a question that merits further investigation. The hyperpolarization-activated cation conductance is well known to contribute at least to the determination of the resting membrane potential and to the generation of intrinsic pacemaker activities (see reviews by Pape, 1996 and Lüthi & McCormick, 1998).

The present work further demonstrates that both cellular types operate during normal and epileptic 5–9 Hz oscillations. The TC cells showing an apparent depolarizing sag have a relatively high membrane input resistance, making them more excitable than those that do not display the sag. It is shown that the post-hyperpolarization ramp-shaped depolarization usually develops shortly after the occurrence of the first few cycles of rhythmic EPSP barrages. Furthermore, this rebound depolarization coincides with such a barrage and can seemingly trigger or facilitate the triggering of an AP discharge. Assuming that the depolarizing sag reflects an H-current, it is expected to contribute to the TC oscillations by enhancing the rate of repolarization of the membrane during the summation of CT-induced EPSPs. Of course, the contribution of such intrinsic rebound depolarization depends in particular on the depth of the hyperpolarization. Knowing that the recurrent EPSP barrage does not systematically reach the AP threshold (see also Pinault et al. 1998), the hyperpolarization-activated ramp-shaped depolarization should increase the probability of reaching the firing threshold on each cycle of the rhythmic activity, probably by facilitating the generation of a low-threshold Ca²⁺ spike. Thereby, the post-hyperpolarization rebound depolarization ending with a low-threshold Ca²⁺ spike topped by an AP discharge might emerge like a resonant phenomenon.

The present study also reveals that, during the natural and epileptic 5–9 Hz oscillations, the hyperpolarizing wave that follows the depolarizing wave mainly includes a barrage of GABA_A IPSPs triggered by RTN cells. This means that during such oscillations RTN cells, which are well known to shape TC activities (Kim et al. 1997), should effectively curtail the depolarizing waves through the activation of GABA_A receptors. Such a powerful inhibitory mechanism would explain why TC cells do not systematically exhibit a high-frequency AP burst on each cycle of the SWD (see also Pinault et al. 1998). Thereby, RTN cells might further prevent the development in TC neurones of a NMDA-dependent depolarization similar to that identified in vivo in cat TC neurones after lesion of the RTN (Deschênes & Hu, 1990). On the other hand, RTN cells might strongly activate H-channels in their target TC neurones (Lüthi & McCormick, 1998).

The TC intracellular recordings of the present study are somehow different from those obtained in the ketamine- xylazine anaesthetized cat during SWD (Steriade & Contreras, 1995). Indeed, these latter recordings revealed two subpopulations of TC neurones, one with units discharging synchronous robust AP bursts (60 %), the other with tonically inhibited cells (40 %). In the present recording sample, most TC cells occasionally emit a robust AP burst during every cycle of the rhythmic activity (see also Pinault et al. 1998). This indicates that in the rat, large populations of TC cells fire in synchrony during every cycle of absence-related SWD. The difference observed between feline and rodent SWD-related TC activities may simply be accounted for by the fact that, in contrast to what is observed in the rat somatosensory thalamus (see Introduction), a significant number of interneurones exist in the feline thalamus (Jones, 1985). Also, it should be highlighted that feline SWD develop during ketamine- xylazine anaesthesia from sleep-related oscillations (Steriade et al. 1994; Steriade & Contreras, 1995) whereas rodent absence-related SWD principally emerge from wake-related oscillations (Semba & Komisaruk, 1984; Danober et al. 1998; Pinault et al. 2001), and similar absence seizures likewise occur in a spontaneous manner under neuroleptanalgesia (Inoue et al. 1993; Seidenbecher et al. 1998; Pinault et al. 2001).

In conclusion, during normal and epileptic 5–9 Hz oscillations most if not all of the TC neurones recorded from in this study exhibit at least a recurrent EPSP barrage-IPSP barrage sequence, which can lead to an AP discharge (Fig. 12). In relay cells that display a presumed H-current, the corresponding depolarizing wave is amplified by a post-hyperpolarization rebound depolarization, which probably includes a low-threshold Ca²⁺ spike. In those that do not, the depolarizing wave can trigger only such an intrinsic spike, very probably in the midst of the 5–9 Hz oscillations where cellular synchronization and hyperpolarization are maximal.

Figure 12. Summary of the reported data.

Schematic diagram showing likely spatio-temporal cellular interactions within the TC system occurring during natural medium voltage 5–9 Hz oscillations in control non-epileptic rats (A) and during SWD in GAERS (B). In both NE rats and GAERS, at least two types of TC neurones coexist, of which one (TC2) is endowed with a presumed H-current. Note that thalamic, relay and reticular, discharges occur in much more synchronous and phase-locked manners during SWD than during natural 5–9 Hz oscillations. C, a mounting of SW-related extracellular CT and intracellular RTN and TC activities. From top to bottom: a SW complex (ECoG), an extracellular CT discharge, an intracellular RTN discharge, and two typical intracellular TC discharges. The second TC cell (TC2) exhibits a presumed H-current, coinciding with an EPSP barrage. The ramp-shaped depolarization, which includes a presumed I_h, can trigger a low-threshold Ca²⁺ spike (LTS). In RTN cells, the EPSP barrage can trigger voltage-dependent components (V-components). D, schematic drawing of the anatomical relationships between the three main elements that make up the TC system.

Reticular cellular events during normal and epileptic 5–9 Hz oscillations

The intracellular recordings presented reveal that in RTN cells the recurrent, threshold or subthreshold, depolarizing wave includes an initial EPSP barrage, which can trigger apparent intrinsic depolarizing events. Again, the initial EPSP barrage might be generated by CT inputs. Indeed, individual EPSPs have a time course similar to individual EPSPs recorded in TC neurones during the same SW-related period. Furthermore, the paired recordings carried out show that epilepsy-related CT discharges usually occur a few milliseconds before the corresponding reticular discharges. It is worth mentioning that the recurrent EPSP barrage is often recorded during the earliest stage of the 5–9 Hz oscillations. Although the dual single-unit recordings reveal that TC and RTN rhythmic discharges occur in synchrony, especially during SWD, TC neurones may contribute to the triggering of RTN EPSP barrages, as demonstrated in a thalamic slice preparation (Bal et al. 1995).

A recent in vivo intracellular study has suggested that the recurrent depolarizing wave is mainly accounted for by a low-threshold Ca²⁺ potential brought about by the summation of depolarizing potentials (Slaght et al. 2002). The present intracellular recordings with QX-filled micropipettes strongly suggest that the prominent depolarization may include other components. On the basis of the known intrinsic properties of RTN cells (e.g. Mulle et al. 1986; Avanzini et al. 1989) and the plural effects of intracellularly injected QX-314 molecules (Andrade, 1991; Perkins & Wong, 1995; Talbot and Sayer, 1996), the low-threshold and high-threshold QX-resistant depolarizing components are presumably mediated in part by Ca²⁺ currents. The high-threshold QX-resistant depolarization may involve intrinsic and/or NMDA-dependent entries of Ca²⁺. Indeed, in in vitro experiments, NMDA induces membrane depolarization with sustained firing (Spreafico et al. 1988; de Curtis et al. 1989). The QX-sensitive depolarizing component may be mediated in part by a non-inactivating inward current carried by Na⁺, similar to the one recorded in cat RTN cells (Mulle et al. 1986).

The present study thus strongly indicates that in RTN cells, spontaneous SWD or natural 5–9 Hz oscillations are associated with rhythmic, high-frequency burst discharges, which are initiated by CT-induced EPSP barrages. These can further trigger diverse voltage-dependent conductances (Fig. 12), which might operate individually or in concert, strengthening the burst discharges and thus the GABA-mediated hyperpolarizations in target TC neurones.

CT cells play a primary role in the synchronization of relay and reticular neurones

The present recordings indicate that, during the generation of SWD, CT cells probably exert a leading role in the synchronization of their target TC and RTN neurones. During absence seizures, related TC and RTN cells display rhythmic discharges in a synchronous manner, which appear to be triggered by CT-induced EPSPs. Moreover, my paired recordings reveal that CT cells have a propensity to discharge on average 7 ms before thalamic relay and reticular neurones, whether or not the two neurones of any recording pair (CT-TC or CT-RTN) are in direct synaptic connection. In addition, as observed during single-unit and multi-unit recordings, CT cells can fire in a rhythmic manner a certain time before related TC and RTN neurones, and rhythmic EPSP barrages also occur in these neurones during the earliest stage of the normal as well as epileptic rhythmic activity. Thus, these findings raise an important question as to whether layer VI CT neurones behave in a synchronous fashion, as the synaptic pacemaker of thalamic 5–9 Hz oscillations. The answer to this question requires further studies to determine whether this pacemaker activity is either inherent to such CT cells, or the result of a network oscillation.

Corticothalamic neurones are glutamatergic, by far exceed in number their target TC and RTN neurones, and are very effective in generating larger excitatory synaptic conductances in RTN than in TC neurones (Golshani et al. 2001). Therefore, they could reliably synchronize large populations of relay neurones, directly through glutamatergic excitations and, indirectly and massively, through GABAergic inhibitions. Corticothalamic cells are thus expected to play the role of booster-drivers that excite related relay and reticular neurones in a rhythmic and synchronous fashion, thereby forcing the TC system to oscillate. The recordings of the present study indicate that this should occur in a much more powerful manner during SWD than during the natural 5–9 Hz rhythm, at least in the somatosensory system.

In addition, the present thalamic and cortical extracellular field potential recordings show that natural 5–9 Hz oscillations, which give rise to SWD (Pinault et al. 2001), can occur in layer VI several seconds before being propagated synchronously to related TC and RTN neurones. These results are in line with those obtained from multi-site cortical and thalamic field potential recordings in another well characterized genetic model of absence epilepsy, the WAG/Rij rats (Meeren et al. 2002). The hypothesis of a primary role of CT cells in the synchronization of thalamic neurones also finds support in the fact that in GAERS, ECoG signs of SWD precede the full development of electroencephalographic epileptic discharges in the thalamus (Seidenbecher et al. 1998). Furthermore, in vitro electrophysiological data suggest that CT cells might exert an essential control in the generation of paroxysmal rhythmic activities within the TC-RTN network (Bal et al. 2000; Blumenfeld & McCormick, 2000).

However, two previous studies conducted in neuroleptanalgesied GAERS (Seidenbecher et al. 1998) and WAG/Rij rats (Inoue et al. 1993), a genetic model of absence epilepsy similar to GAERS, showed that thalamic neurones fire well before cortical neurones, supporting the hypothesis of a leading role of the thalamus in the development of genetically determined SWD. The discrepancy between these results and those of the present study merits discussion. In the Inoue et al. (1993) study, the ECoG electrodes were located in the frontal cortex and there is no evidence that the thalamic unit activities were recorded at functionally related sites. In the present study, the ECoG electrodes were located in the frontoparietal (somatosensory) cortex and all thalamic unit activities were recorded in the related somatosensory thalamus. In the Seidenbecher et al. (1998) study, layer VI cells, which represent the principal cortical source of thalamic inputs (see Introduction), were not recorded from. On the other hand, it is noteworthy to emphasize that the multi-unit recordings of the present study reveal, in accordance with the results of Seidenbecher et al. (1998), that thalamic neurones have the propensity to fire a few milliseconds earlier than their target layer IV/V cells. Paired single-cell IV/V-VI recordings confirm these findings (D. Pinault, unpublished observations).

In summary, the present findings thus reveal the likely sequence of neuronal events occurring in the somatosensory system of GAERS during the development of SWD: layer VI CT neurones play a major role in the rhythmic, synchronized excitation of thalamic, relay and reticular, neurones (Fig. 12). Furthermore, to all appearances, TC neurones massively synchronize cortical neurones of middle layers.

Switch from natural 5–9 Hz oscillations to SWD in GAERS, a possible scenario

In neuroleptanalgesied rats the medium-voltage 5–9 Hz rhythm has an internal frequency that is lower (by 2–3 Hz) than that of the equivalent physiological rhythm recorded in motionless GAERS and NE rats (D. Pinault, unpublished observations). It is thus seemingly identical to an identified rodent sensorimotor 7–12 Hz rhythm, which also occurs during body immobility (Nicolelis et al. 1995; Fanselow et al. 2001). In normal rats, this rhythm begins in the cortex and then spreads to the thalamus and brainstem (Nicolelis et al. 1995). In GAERS, the natural medium-voltage oscillation gives rise to SWD (Pinault et al. 2001), which are always accompanied by whisker twitching (see also Vergnes et al. 1982; Semba & Komisaruk, 1984). This natural rhythm does not give rise to SWD in control NE rats, meaning that it is not itself sufficient for generating absence-related epileptic activity.

Understanding the factors that induce SWD from natural medium-voltage 5–9 Hz oscillations in GAERS is of crucial importance. The present study reveals that, contrary to TC and RTN neurones, CT cells do not usually discharge robust bursts of APs. In a previous study, we have shown that, during the development of SWD, RTN cells start to fire in the burst mode almost always before TC neurones (Pinault et al. 2001; see also Fig. 2). It is thus tempting to put forward that a CT induced resonance phenomenon may contribute to the switch of the natural medium-voltage 5–9 Hz oscillation into SWD in GAERS. On the basis of the present findings, the corresponding scenario might be as follows: (1) CT cells somehow launch and maintain the natural 5–9 Hz rhythm, assuring the phase-locking of threshold and subthreshold thalamic, relay and reticular, oscillations; (2) because RTN cells are endowed with powerful, synaptic and intrinsic, electroresponsive properties, they start to react – or to resonate – massively before TC neurones, exhibiting rhythmic robust high-frequency AP burst discharges (Pinault et al. 2001); (3) these RTN bursts generate GABA_A-dependent IPSP barrages in their target TC cells, (4) in those having an H-current, the RTN-induced hyperpolarizing wave triggers a rebound ramp-shaped depolarization leading to an AP discharge underlaid by a low-threshold Ca²⁺ potential (Lüthi & McCormick, 1998). This discharge in turn reinforces reciprocal cellular interactions between the thalamic relay and reticular nuclei (see also Bal & McCormick, 1993) through the recruitment of new active RTN units, thereby strengthening the GABA_A-dependent hyperpolarization in larger populations of TC neurones. These massively recruit new active units at the cortical level, in particular in layer IV/V. The cycle somehow starts again, and cellular hypersynchronization thereby would be generated.

Other possibilities for the switch of natural oscillations into SWD may exist. For instance, an in vitro study has demonstrated that layer V pyramidal neurones are able to generate synchronized oscillations in the same frequency range when NMDA receptors are activated in a tonic fashion (Silva et al. 1991). NMDA responses are considerably more important in middle and deep layers of the cerebral cortex in GAERS than in control rats (Pumain et al. 1992). Such excitatory responses, however, might result from the activity induced by hypersynchronous TC inputs. Also, recent in vivo and in vitro studies have shown that neocortical disinhibition generates synchronized 7–14 Hz oscillations, which are driven by deep cortical layers and which spread to the thalamus from layer VI (Castro-Alamancos, 2000; Castro-Alamancos & Rigas, 2002). Thus, the cause(s) switching medium-voltage 5–9 Hz oscillations into SWD in genetic models of absence epilepsy is a fundamental issue requiring additional investigation.

Conclusion

This study reveals novel spatio-temporal dynamics of the interactions between cortical and thalamic neurones occurring during the generation of rodent, normal and epileptic, 5–9 Hz oscillations. It strongly suggests that synchronized CT neurones represent a possible synaptic pacemaker of thalamic 5–9 Hz oscillations. Furthermore, in GAERS the TC and RTN rhythmic firings that are associated with SWD are provoked by intracellular, synaptic and intrinsic, mechanisms that are very similar to those associated with the natural medium-voltage 5–9 Hz rhythm. This suggests that, in GAERS, absence-related SWD correspond to a hypersynchronous wake-related physiological oscillation. The hypersynchronization mechanism would depend on a CT-induced resonance phenomenon.