This study tested the hypothesis that early functional alterations in neuronal synchrony in the partially deafferented cortex may lead to spontaneously occurring electrographic seizures. In vivo experiments with partial deafferentation of cat suprasylvian gyrus after extensive undercut of the white matter were conducted using multi-site EEG, extracellular unit and intracellular recordings. The amplitudes of EEG waves were much higher in the areas surrounding deafferented cortical fields as compared with control and with undercut cortex. In 40% of animals with undercut cortex, paroxysmal activity occurred 2–3 h after the undercut and was initiated in the relatively intact cortex, adjacent to the more disconnected one. The seizures that followed the undercut consisted of spike-wave/polyspike-wave complexes and fast runs, resembling the electrographic patterns of some clinical epileptic syndromes. An increased local synchrony in the relatively intact cortex evolved into paroxysmal activity that ultimately spread to the deafferented cortex. The electrographic seizures were found only in animals that showed a propagation of the slow sleep-like oscillation in control conditions. The increase of long-range synchrony within a given seizure was associated with seizure termination. These results indicate that alterations in neuronal synchrony following neuronal trauma can be a critical factor triggering electrographic seizures.
Cerebral cortical trauma may lead to paroxysmal activities. Within 24 h following head injury with penetrating wounds, up to 80% of patients display clinical seizures (Kollevold, 1976; Dinner, 1993). Immediately after trauma, the level of extracellular K+ increases and this may increase the propensity to seizures (Moody et al., 1974; Traynelis and Dingledine, 1988; McNamara, 1994). Trauma also elicits a partial deafferentation and, consequently, a decrease in input signals that can result in enhanced intrinsic and synaptic excitability of individual neurons (Turrigiano et al., 1998; Desai et al., 1999). Chronic neuronal hyperexcitability and epileptogenesis in experimental animals have been demonstrated in isolated neocortical islands with intact pial circulation in vivo (Burns, 1951; Sharpless, 1969; Halpern, 1972) and in neocortical in vitro slices after chronic cortical injury (Prince and Tseng, 1993; Hoffman et al., 1994; Prince et al., 1997; Li and Prince, 2002). A computational model of posttraumatic epileptogenesis in isolated cortical islands concluded that paroxysmal discharges are due to changes in intrinsic properties of pyramidal cells and enhanced NMDA synaptic conductances, without altered or even increased inhibition (Bush et al., 1999).
However, there is little information about the spontaneous development of paroxysmal activity immediately after neocortical injury. In view of earlier data from human cortical slabs isolated at the time of surgery, showing that the deafferented cortical tissue can display paroxysmal, high-voltage activity shortly after isolation (Echlin et al., 1952; Henry and Scoville, 1952), we hypothesized that early functional modifications in the deafferented cortex can develop neuronal hyperexcitability, due to anatomical and functional changes in synaptic efficacy and cortico-cortical connectivity, which may lead to spontaneously occurring paroxysmal activity.
In the present study, we attempted to answer the following questions concerning the mechanisms that lead to the initiation, propagation and cessation of electrographic paroxysmal activities in the deafferented cortex. (i) What are the spatio-temporal properties of the slow sleep-like oscillation (~0.5–1 Hz) and spontaneously occurring paroxysmal activity in the partially deafferented cortex? It was shown that the slow sleep oscillation may develop, without discontinuity, into electrographic seizures (Steriade et al., 1998a). As slow sleep oscillation is generated intracortically (Steriade et al., 1993a,b; Timofeev et al., 2000), we conducted experiments in the partially deafferented cortex. (ii) What are the alterations in neuronal behavior within the partially deafferented and the adjacent, relatively intact neocortex during spontaneously developing paroxysmal activity? As a completely isolated cortex is mainly silent (Burns, 1951; Burns and Webb, 1979), we used a model of partial cortical deafferentation with relatively intact cortico-cortical connections, which led to initiation of spontaneous paroxysmal activity. Some of these results have been presented in abstract form (Topolnik et al., 2001).
Materials and Methods
Acute experiments were carried out on 35 adult cats of either sex anesthetized with ketamine and xylazine (10–15 and 2–3 mg/kg i.m., respectively) and on seven cats anesthetized with sodium pentobarbital (35 mg/kg). All pressure points and the tissues to be incised were infiltrated with lidocaine. The animals were paralyzed with gallamine triethiodide and artificially ventilated while monitoring the end-tidal CO2 concentration at 3.5–3.8%. A permanent sleep-like state, as ascertained by continuous recording of the EEG, was maintained throughout the experiments by administering additional doses of the same anesthetic. Body temperature was maintained at 37–39°C. Heart rate was continuously monitored (90–110 beats/s). The stability of intracellular recordings was ensured by cisternal drainage, bilateral pneumothorax, and by filling the hole made for recordings with a solution of 4% agar.
The undercut cortex (Fig. 1A) was produced by large white matter transections below the suprasylvian gyrus (13–15 mm postero-anteriorly, 3–4 mm medio-laterally and 3–4 mm deep). A knife was inserted into the posterior part of the suprasylvian gyrus, perpendicularly to its surface at a depth of 3–4 mm, then rotated 90° and advanced rostrally along the gyrus parallel to its surface for a total distance of 13–15 mm, then moved back, rotated 90° and removed from the same place where it entered the cortex. Thus, the white matter below the posterior part of the gyrus was transected, creating conditions of partial cortical deafferentation. In two experiments the undercut was produced in the same way but in the opposite direction: the knife was inserted into the anterior part of the suprasylvian gyrus and the undercut was made from the anterior to the posterior part of the gyrus.
(i) Field potential (EEG) recordings were obtained with an array of seven coaxial electrodes, ~1.5 mm apart, placed with the outer pole at the cortical surface and the inner pole at ~0.8–1 mm in the cortical depth, along the suprasylvian gyrus (Fig. 1A). Those electrodes were inserted in the same place before (for control EEG recordings) and after the undercut (see sites EEG1 to EEG7 in Figs 1–3 and 6) under microscope guidance. In some experiments, another array of EEG electrodes was inserted into the postcruciate and anterior part of suprasylvian gyri (see sites EEG1.1 to EEG7.1 in Fig. 1A). (ii) Extracellular unit recordings were performed with tungsten microelectrodes (impedance: 8–12 MΩ). (iii) Single and dual intracellular recordings were obtained using glass micropipettes filled with 3 MΩ potassium acetate (DC resistance, 30–70 MΩ). We recorded neuronal activity from the anterior (area 5), middle (area 7) and posterior (area 21) parts of the undercut suprasylvian gyrus. A high-impedance amplifier (bandpass, 10 kHz) with an active bridge circuitry was used to record and inject currents into the cells. All electrical signals were sampled at 20 kHz and digitally stored on Vision (Nicolet, WI) for offline computer analysis.
At the end of experiments, the cats were given a lethal dose of intravenous sodium pentobarbital (50 mg/kg). The brain was removed and the location of undercut and recording electrodes was verified on 80 μm sections stained with thionine (see Fig. 1A).
Wave-triggered averages (WTAs) were calculated as follows. The first depth-negative peak for one of EEG traces was detected, and equal windows around that point (500 ms before and 800 ms after) were extracted from all channels. All segments belonging to a given channel were averaged.
Auto- and cross-correlograms of different EEG channels were computed from periods of 2–3 min of stable activity. The most posterior electrode was taken for the time reference. In experiments in which the undercut was made postero-anteriorly, we calculated the auto- and cross-correlations between the most anterior electrode (EEG 7) and other recording sites (see Fig. 5).
The firing rate analyses were performed using extracellular recordings from 100 s periods for each studied neuron. Values in the text are given as means ± SE. The first depth-negative peak for each EEG trace was chosen as a zero and equal windows around that point (from −1500 ms to 1000 ms, bin width 10 ms) were taken to create the peri-event histograms.
For spike-triggered-averages (STAs) the first action potential that occurred after the period of hyperpolarization was selected as zero-time and all EEG traces were extracted with equal windows around this point (from −500 to 500 ms).
The histograms of membrane potential distribution were created for successive periods of 10 s. The peaks of distribution that corresponded to the most probable mode of membrane potential were taken as the level of membrane potential.
Alterations of Slow Oscillation by Cortical Undercut and Development to Seizure Activity
Under ketamine–xylazine anesthesia, the EEG from the cortical depth demonstrated a spontaneous slow oscillation (<1 Hz) (Fig. 1B, left panel—CONTROL). This type of anesthesia best mimics the slow sleep oscillation (Contreras and Steriade, 1995), with the same features as it appears during natural slow-wave sleep in experimental animals (Steriade et al., 1996a,b, 2001; Timofeev et al., 2001) and natural human sleep (Achermann and Borbély, 1997; Amzica and Steriade, 1997). Immediately after the undercut of suprasylvian gyrus (earliest measurements were obtained 5 min following the undercut), the amplitude of EEG waves was reduced, especially in the posterior and middle parts of the suprasylvian gyrus where the degree of deafferentation was maximal (Fig. 1B,C, middle panel—UNDERCUT). The EEG activity in this partially deafferented cortex remained markedly decreased up to 1–2 h. In a few experiments, surviving spindles (7–14 Hz) could be observed in the anterior part (electrodes 6 and 7) of the undercut cortex (see Fig. 3—UNDERCUT), probably because some thalamocortical connections remained intact in those areas. Between 2 and 3 h after the undercut, the activity in the undercut cortex reorganized, and increased amplitude of EEG waves in the middle and anterior parts of the suprasylvian gyrus could be detected (Fig. 1B,C, panels indicating 3 HOURS AFTER; a, b and c, three different animals).
In 60% of animals, EEG activity recovered to the pattern of normal slow oscillation, with absence of spindles in deafferented sites. This type of activity could be maintained up to 8 h during several experiments. The remaining 40% of animals demonstrated paroxysmal-like activity or clear-cut electrographic seizures. By paroxysmal-like activity we mean high-amplitude slow waves with the morphological features of interictal spikes. Their amplitude was at least twice as high as during the normal slow oscillation (Fig. 1B,C, 3 HOURS AFTER; see panels b and c). In ~50% of cases electrographic seizures developed progressively from the slow oscillation, while in the other half of cases seizures started suddenly. The paroxysmal activities were variable in their electrographic patterns. In different experiments we observed interictal spikes, sharp waves (Fig. 1B,C—upper right panel), spike-wave (SW) and polyspike-wave (PSW) complexes at 2–3 Hz and fast runs at 10–15 Hz (Fig. 2). The lower frequencies of SW/PSW complexes in conjunction with fast runs are typical EEG pattern in patients with Lennox–Gastaut syndrome and may also be found in posttraumatic epilepsy (Niedermeyer, 1999). The electrographic seizures lasted from 5 to 50 s, with an average of 17.0 ± 3.3 s, and such paroxysmal activity could last from 1 to 5 h within an experiment.
The amplitude of EEG waves during seizures was significantly higher in the anterior, more intact, part of the undercut suprasylvian gyrus, compared to the posterior part (Fig. 1B,C, upper right panels; Fig. 2, middle panel) (n = 14, P < 0.001). Recording of paroxysmal activity from the intact neighboring areas (between the somatosensory cortex and anterior suprasylvian gyrus) revealed that electrographic seizures was initiated in the intact postcruciate gyrus (Fig. 2, bottom panel; see location of electrode 7.1 in Fig. 1A) and only ~7 s later paroxysmal events appeared in the undercut cortex. In that case too, the amplitude of seizures was relatively higher in the anterior part of the undercut suprasylvian gyrus (Fig. 3), pointing to higher local synchrony in this part of deafferented cortex.
Previous studies showed that 25% of cats anesthetized with ketamine–xylazine demonstrated spontaneous seizures (Steriade et al., 1998a, Timofeev et al., 2002b). As mentioned above, 40% of cats exhibited paroxysmal activities in the present study. To rule out a major role of ketamine–xylazine anesthesia in the development of paroxysmal activities, we performed experiments with undercut cortex on seven cats under barbiturate anesthesia. Barbiturates enhance GABAergic inhibition via prolongation of time of opening of Cl– channels (Twyman et al., 1989). Thus, we did not expect to see the development of full-blown seizures in those experiments, but we expected to see an enhanced activity in areas surrounding the undercut cortex. This was indeed the case (see Fig. 1 of supplementary material). In control conditions, we recorded spindle and sharp waves in all seven experiments. According to the normal distribution of spindles their maximal amplitude was found in anterior parts of suprasylvian gyrus and lower amplitudes were observed in the posterior parts of this gyrus [see (Morrison and Dempsey, 1942)]. Following the undercut, the amplitude of all EEG waves was decreased for 1 h. Thereafter, the spindle activity was significantly enhanced in the relatively intact anterior part of suprasylvian gyrus, whereas it was absent in the posterior part of suprasylvian gyrus because of the destruction of thalamocortical connections. Similarly to recordings in ketamine–xylazine anesthetized cats, 3–8 h after the undercut the activity in the anterior parts of suprasylvian gyrus was characterized by the sharp waves that were higher in amplitude, compared to the activity in control periods. These data show that, although clear-cut seizures similar to those shown in Figures 1–2 were not seen in barbiturate-anesthetized cats, the decrease in the amplitude of EEG waves in the cortical areas surrounding the undercut during first 1–2 h followed by their enhancement afterwards is a general rule, which applies to both ketamine–xylazine (Fig. 1B,C) and barbiturate (Fig. 1 of supplementary material) anesthesia.
Propagation of the Slow Oscillation and Paroxysmal Activities in the Deafferented Cortex
The propagation of normal and paroxysmal activities was studied in 23 experiments. The horizontal propagation of the slow oscillation mainly depended on the direction the undercut was performed. In all cases the slow oscillation started near the electrode from the relatively intact part of the partially deafferented cortex and propagated toward more deafferented sites.
The changes in EEG synchrony, with or without paroxysmal development, are shown at Figure 3. (i) In those cases that did not reveal paroxysmal activities within 8–10 h after the undercut, the cross-correlation analysis of EEG waves as well as wave-triggered averages (WTAs) showed strong synchrony between different recording sites before the undercut (time lag between the first and seventh EEG electrodes was 1.8 ± 0.8 ms) (CONTROL, circles in bottom panel). The undercut led to a slightly increased time lag in the propagation from anterior to posterior sites, 4.3 ± 1.7 ms (UNDERCUT). Recovery in EEG patterns was seen 2–3 h after the undercut, but the time lag increased to 7.8 ± 6.0 ms (RECOVERY). Thus, the slow oscillation occurred highly synchronized within a distance of ~9 mm. Partial cortical deafferentation decreased the velocity of propagation of slow oscillation from 5.0 m/s to 1.1 m/s. (ii) In cases with paroxysmal development, the comparison between the EEG waves’ propagation before the undercut and seizures (triangles in bottom plots) revealed that the time lag between area 5 and area 21 was 20.8 ± 11.8 ms before the undercut, and was significantly increased 2–3 h after the undercut, during paroxysmal activity (bottom plots). In all experiments, a significant slowing-down in EEG propagation started from a site between EEG electrodes 4 and 5, likely representing the region with significantly decreased neuronal excitability. The velocity of propagation of paroxysmal events in the partially deafferented cortex was between 0.3 and 0.5 m/s.
The above data show the following tendencies: (i) if animals demonstrated a higher long-distance synchrony of EEG activity before the undercut, they would not display paroxysmal activity after undercut; (ii) if, however, EEG activity was less synchronous between the anterior and posterior suprasylvian areas before the undercut, cortical deafferentation led to an increase of asynchrony and those animals later displayed paroxysmal activity. The differences in time lag of EEG activity propagation were significant (P < 0.01) in the two groups of animals. As shown below, the lower synchrony between distant sites in cases with seizure development stood in contrast with enhanced local synchrony in the relatively intact cortex, where seizures were initiated.
In two experiments we performed cortical undercut in the opposite direction (from anterior to posterior sites in the suprasylvian gyrus) and found similar patterns of propagation of EEG activity, namely from more intact to more deafferented sites (data not shown).
To study the mechanisms responsible for the slowing-down of propagation of the slow oscillation from the relatively intact to the partially deafferented cortex, we performed paired intracellular recordings (n = 7) from neurons located in those two parts of the suprasylvian gyrus. In such experiments we analyzed the temporal relations between intracellularly recorded neurons during the slow oscillation. Figure 4 shows that both neurons were depolarized during the depth-negative EEG waves and hyperpolarized during the depth-positive EEG phases of the slow oscillation. The depolarized states consisted of postsynaptic potentials that occasionally triggered action potentials, which were also seen in simultaneous multiunit recordings. The hyperpolarized states in both neurons are due to a global disfacilitation resulting from the temporal absence of synaptic activity in corticothalamic networks and are associated with an increase in the apparent input resistance (Contreras et al., 1996; Timofeev et al., 1996, 2002b; Neckelmann et al., 2000). We found different durations of the disfacilitation period in the two simultaneously recorded neurons depicted in Figure 4, depending on the site of recording. The periods of hyperpolarization started earlier and lasted longer in the deafferented cortex (Fig. 4, cell 2, −64 mV, blue) than in the relatively intact cortex, and the onset of the subsequent depolarizing phase was delayed in the deafferented site, compared with the adjacent intact areas (bottom right panel). Thus, at the time when neurons in the relatively intact area reached firing threshold and started to trigger action potentials, neurons in the deafferented site were still hyperpolarized. As deafferented neurons needed more time to reach firing threshold, the time lag in the propagation of activity between intact and undercut areas would be increased.
Correlation Between Local Neuronal Firing in the Deafferented Cortex
Since the alterations in the amplitude and spatio-temporal properties of cortically generated slow oscillation are produced by changes in synchronous firing of cortical neurons, we studied the mean firing rates of neurons recorded from areas 5 and 21, before and after undercut (n = 8, see Fig. 2 of supplementary material). In four of these experiments, the slow oscillation developed into paroxysmal episodes. In all recorded sites (n = 55), neurons fired during the depth-negative phase of the slow oscillation and were silent during the depth-positive phase. The mean duration of silent periods was 230 ± 20 ms. The mean firing rate before the undercut was 25.1 ± 12.1 Hz (n = 24). Immediately after the undercut, neuronal firing rate dramatically decreased to 2.0 ± 1.0 Hz (n = 11) in both sites, namely, the deafferented area 21 and the relatively intact area 5, thus suggesting that neuronal activity was significantly affected by the undercut. The duration of silent periods (due to disfacilitation) was longer (365 ± 125 ms). During the next 2–3 h, we observed a partial recovery of EEG that occurred in parallel with increased neuronal firing: 16.9 ± 8.8 Hz (n = 17); however, the periods of disfacilitation become dramatically longer in the deafferented cortex (405 ± 15 ms), compared with control periods. In other instances, the undercut led to development of electrographic paroxysmal discharges and the mean firing rates of neurons in the anterior (relatively intact) part of the undercut cortex (area 5) slightly increased to 27.1 ± 8.3 Hz, whereas the firing rate of neurons in the posterior part (area 21) remained low (1.5 ± 0.3) Hz.
Thus, the reduction in the amplitude and velocity of propagation of slow oscillation mentioned above was associated with a reduction in mean local firing rate of partially deafferented neurons and by an enhancement in the duration of periods of disfacilitation. By contrast, the development of paroxysmal activity was initiated from the relatively intact areas of the undercut cortex, and was accompanied by an increase in the local neuronal firing rate in those areas, compared with the deafferented cortex.
Spatiotemporal Patterns of Seizure Generation in Single Neurons of Undercut Cortex
Intracellular recordings (n = 67) in conjunction with field potentials from the undercut suprasylvian gyrus were used to analyze the patterns of synchrony during electrographic seizures and during different periods of the same seizure. Periods of paroxysmal activity were analyzed in 11 regular-spiking (RS), 12 intrinsically bursting (IB), four fast-rhythmic-bursting (FRB), and three fast-spiking (FS) neurons [see (Connors and Gutnick, 1990; Gray and McCormick, 1996; Steriade et al., 1998b)]. We found that, prior to seizures, FRB and FS neurons displayed increased firing rates compared with other neuronal types. The relative increased discharge frequencies of FS neurons during ketamine–xylazine anesthesia are congruent with the higher firing rates of the same neuronal type during natural sleep states (Steriade et al., 2001).
The gradual involvement of a RS neuron in the generation of electrographic seizure is illustrated in Figure 5. RS neurons constitute the majority of neocortical cells. Most neurons recorded within the focus of paroxysmal activity (where EEG amplitude was maximal) demonstrated a gradual transformation from normal slow oscillatory behavior to large-amplitude (20–40 mV) depolarizations with features of paroxysmal depolarizing shifts (PDSs) (Matsumoto and Ajmon-Marsan, 1964a,b). During periods of slow oscillation (Fig. 5, middle left), all recorded neurons revealed depolarized states associated with action potentials, followed by repolarization or hyperpolarized states. The membrane potential during the slow oscillation displayed a bimodal distribution with peaks at −66.0 ± 10.2 mV and −57.6 ± 6.9 mV, representing the down and up states, respectively (Fig. 5). The development of electrographic paroxysmal activity was associated with a gradual shift of neuronal membrane potential in both de- and hyperpolarizing direction (Fig. 5, bottom right panel). In many instances the strong neuronal depolarization led to partial or full spike inactivation (Fig. 5, middle right). The end of electrographic seizures was usually followed by postictal depression, reflected as a prolonged period of neuronal hyperpolarization (Fig. 5, upper panel).
The alterations in synchrony between single neurons and field potentials as well as the propagation of normal and paroxysmal activities were studied in 14 intracellularly recorded neurons. In some cases, at the onset of electrographic seizures, the paroxysmal activity was clearly seen in all EEG electrodes, but the intracellularly recorded neuron (which was located in the relatively intact cortex, between EEG6 and EEG7; see Fig. 1A) still displayed ‘normal’ slow oscillatory pattern of activity (Fig. 6, middle left panel). At the same time, the propagation of EEG activity from the anterior to the posterior site of recording occurred relatively slowly, with a time lag of ~70 ms (Fig. 6, bottom left STA). Within the next 10–20 s, EEG synchrony became higher and the time lag between anterior and posterior parts of field potential recording decreased to 24 ms and, further, down to 12 ms (Fig. 6, bottom middle and right STA). As a result, the recorded neuron was progressively depolarized and demonstrated PDSs, and soon thereafter the seizure terminated. Thus, the long-range synchrony increased during seizure, promoted the recruitment of neurons in the generation of paroxysmal activity, but immediately after the whole neuronal pool was generating a highly synchronous paroxysmal activity, the seizure stopped.
Synchronized periods of fast runs at 10–15 Hz were often present during electrographic seizures (Fig. 3 of supplementary material). In intracellularly recorded neurons, they appeared as stereotyped and rhythmic depolarizing potentials, ~2–20 mV in amplitude, often crowned by full-blown action potentials. These components demonstrated variable features in different electrographic seizures. In field potential recordings, they were highly synchronized locally, and diminished progressively with the distance from the seizure focus. During one period of fast runs, the cell located rostral to EEG7 (see Fig. 1A) fired immediately after the peak of depth-EEG negativity, thus demonstrating synchrony between the neuron and neighboring EEG activity (Fig. 3 of supplementary material, bottom left panel). However, 10 s later, the cell fired before the peak of depth-EEG negativity (Fig. 3 of supplementary material, bottom right panel). Thus, the time relation between this neuron and the closest field potentials shifted in phase. During both periods of fast runs, the same propagation pattern was detected from the anterior to posterior parts of the undercut cortex, with time lags between these sites of ~40–50 ms. The velocity of fast runs propagation along deafferented cortex was ~0.25 m/s. Although they could propagate, fading was so high that fast runs were hardly distinguishable in distant sites of deafferented cortex. We conclude that fast runs occurred as highly synchronous local paroxysmal events at the initiation site of electrographic seizures, but became variable and weak during propagation to remote sites.
Within the paroxysmal focus, different neurons were involved in the generation of the same electrographic seizure with various degrees, and the same neuron could be differently involved in successive electrographic seizures (see Fig. 4 of supplementary material). The recorded neuron was in the area of undercut between the third and fourth EEG electrodes. The degree to which it was involved in the generation of paroxysmal activity was analyzed during 24 consecutive seizures. In eight out of those paroxysms, the neuron revealed oscillatory behavior (SW/PSW complexes at ~3 Hz), synchronously with EEG paroxysmal activity. In other 12 seizures (with the same position of EEG electrodes), we did not observe paroxysmal activity at the neuronal level but detected such activity at the surrounding EEG. And, in the remaining four seizures, the neuron revealed seizure-like behavior, whereas no paroxysmal activity was detected at the EEG level. Such diverse relations between intracellular and field potential activities were observed in 20% of experiments, thus pointing to the focal nature of seizures associated with partial cortical deafferentation.
Our data show that (i) partial cortical deafferentation by undercut modified the spatio-temporal characteristics of the slow sleep-like oscillation; (ii) local cortical synchrony was increased in areas surrounding the undercut cortex where paroxysmal activity was initiated; (iii) following undercut the animals that revealed propagation of slow oscillation in control conditions also showed a high propensity for development of electrographic seizures; (iv) during electrographic seizures triggered by undercut, the long-distance synchrony increased progressively towards the seizure end; and (v) as soon as the electrographic paroxysm became highly synchronous, the probability that the seizure terminates greatly increased.
Some Remarks on Anesthesia
Our experiments were performed under ketamine–xylazine anesthesia, which best mimics natural patterns of slow-wave sleep (see first section in Results). The action of ketamine, an antagonist of N-methyl-d-aspartate (NMDA) receptors (Thomson, 1986), is species- and dose-dependent (Celesia et al., 1975; Velisek et al., 1993; Bloms-Funke et al., 1999). The effects mediated by xylazine, an α2 adrenoreceptor agonist, are inhibitory, acting by increasing a K+ conductance (Nicoll et al., 1990). We do not exclude the possibility that some animals maintained under ketamine–xylazine, an anesthetic that mimics natural patterns of slow-wave sleep (Feinberg and Campbell, 1993; Mahon et al., 2001), display seizures with SW/PSW complexes and fast runs. Such cases were reported in a minority of experimental animals (Steriade and Contreras, 1995; Steriade et al., 1998a). In the present series of experiments, data were only collected from animals in which long-lasting control periods of recording (up to 2–3 h) showed that no paroxysmal activity occurred before the cortical undercut. In addition, our experiments with barbiturate-anesthetized animals showed an increase in the amplitude of sharp waves after the undercut, thus pointing to an increased local synchrony in areas surrounding undercut cortex in the absence of ketamine–xylazine.
Alterations in Normal and Paroxysmal Patterns after Cortical Deafferentation
The slow oscillation (generally 0.5–1 Hz) has been shown to be cortical in origin because it survives extensive ipsilateral thalamectomy (Steriade et al., 1993b) and cannot be recorded in the thalamus of decorticated animals (Timofeev and Steriade, 1996). The intracortical synchronization of the slow oscillation is disrupted after reversible interruption or transections separating the anterior and posterior parts of the suprasylvian gyrus (Amzica and Steriade, 1995b). The slow oscillation reveals a high level of short- and long-range synchronization, consistent with a propagation rate of ~1 m/s in vivo (Amzica and Steriade, 1995a) and 0.1 m/s in cortical slices maintained in vitro (Sanchez-Vives and McCormick, 2000).
In the present experiments, we found the slow oscillation in the partially deafferented cortex. Slow waves decreased in amplitude immediately after the undercut, which was associated with the decreased coherence in neuronal firing, but 2–3 h later the slow oscillation recovered to almost its initial pattern. The time lag between the onset of the depolarizing (EEG depth-negative) phase of this sleep rhythm in areas 5 and 21 varied from 2 to 70 ms. Relatively short time-lags imply propagation through mono-, oligo- or multisynaptic excitatory connections, while time-lags exceeding ~50 ms would either involve inhibition-rebound cycles within corticothalamocortical loops (Amzica and Steriade, 1995a; Neckelmann et al., 1998) or be still intracortical with excitatory drives activating intrinsic neuronal currents that induce a firing of postsynaptic neurons with significant delays (Timofeev et al., 2000). Admittedly, the partial cortical deafferentation in our experiments excluded the presence of a great bulk of thalamocortical connections, thus implicating intracortical excitatory pathways and not thalamocortical postinhibitory rebound excitations.
At least in ~40% of our experiments, partial cortical deafferentation led to development of electrographic seizures. Consistently, the initiation of paroxysmal activity took place in the relatively intact, anterior part of the deafferented gyrus, with only weak or absence of such activity in the posterior part. Initiation of paroxysmal activity in the area adjacent to the undercut cortex was related to an increase in local synchrony. Field potentials and multi-unit activity in relatively intact areas occurred synchronously (with a time lag of ~2 ms) showing an increased local synchrony. The time lag of propagation increased with transition to the more deafferented (posterior) areas, because of the relative hyperpolarization of deafferented neurons and their inability to reach firing threshold synchronously with more intact neurons. This may create conditions for a spatial localization of excitation and a permanent self re-excitation of neurons in more intact area through the recurrent network of excitatory connections (Traub and Wong, 1982) that remained functional. In turn, increased positive feedback in an excitatory system could lead to the initiation of electrographic paroxysms. Thus, we postulate that early (posttraumatic) alterations of spatio-temporal properties of the slow oscillation can predict the initiation of electrographic seizures.
The triggering factor in the cortical overexcitation may be the hyperpolarization-activated cation current, IH (Schwindt et al., 1988) whose depolarizing sag may lead to activation of the low-threshold transient Ca2+ current, IT (de la Peña and Geijo-Barrientos, 1996). It has previously been reported that, during seizures, there is an increase in the level of maximal hyperpolarization achieved by neurons between depolarizing events (Steriade et al., 1998a). Such an increased hyperpolarization would support an increase of the efficiency of IH. An increased [K+]o during seizures would shift the reversal potential for all K+-mediated currents and further involve IH in the generation of seizures (Timofeev et al., 2002a). Then, neurons located at the border between the undercut and the intact cortex would behave like triggers of paroxysmal activity under the conditions of some epileptogenic factors.
Many other trauma-activated factors can be regarded as epileptogenic in our experimental model. Neuronal and glial damage elicited by the undercut can increase, to a certain degree, the level of [K+]o, which is known to promote development of seizures (Zuckermann and Glaser, 1968; Moody et al., 1974). In acute conditions, brain lesions produce an immediate increase of [K+]o from intracellular pools, leading to cellular depolarization, increased excitability and decreased inhibition (Traynelis and Dingledine, 1988). It was shown that in conditions of increasing [K+]o the RS firing pattern of some neurons can converse into IB patterns, which would further increase the excitability of network (Jensen et al., 1994). Also, the accumulation of [K+]o can result in neuronal swelling and electrotonic coupling between neurons (Traynelis and Dingledine, 1988; McNamara, 1994). The last factor in turn can lead to the hyperexcitation (Carlen et al., 2000). Some consequences of traumatic insult can also play an epileptogenic role by resulting in excessive accumulation of extracellular glutamate (Lipton and Rosenberg, 1994) and in the raisinbg of intracellular Ca 2+ via transmitter- and voltage-gated channels. Increased intracellular Ca 2+ in turn enhances the response to glutamate (Yang and Benardo, 1997). Finally, a reduction in functional inhibition can underlie epileptiform activity, for example the depolarizing effect of GABA may result in a positive shift in Cl– reversal potential (from −70 to −44 mV) and in increasing the intracellular Ca2+ (Tasker and Dudek, 1991; van den Pol et al., 1996; Timofeev et al., 2002b).
This work was supported by grants from the Canadian Institutes for Health Research (MOP-37 862 and MT-3689), National Institute of Neurological Disorders and Stroke of the USA (1-RO1 NS-40 522), Fonds de la Recherché en Santé du Québec, and Savoy Foundation. We thank Y. Cissé for taking part in one experiment and P. Giguère and D. Drolet for technical assistance.
Address correspondence to I. Timofeev, Laboratoire de Neuro-physiologie, Faculté de Médecine, Université Laval, Québec, Canada G1K 7P4. Email: email@example.com.
Supplementary material can be found at: http://www.cercor.oupjournals.org