Frontal cortex leads other brain structures in generalised spike-and-wave spindles and seizure spikes induced by picrotoxin

ELSEVIER

Electroencephalographyand clinical Neurophysiology98 (1996) 157- 166

Frontal cortex leads other brain structures in generalised spike-and-wave spindles and seizure spikes induced by picrotoxin ’ A. Medvedev, L. Mackenzie, J.J. Hiscock, J.O. Willoughby

*

Centre for Neuroscience and Department of Medicine, Flinders University and Flinders Medical Centre, P.O. Box 2100, Adeluide, SA 5001, Australia

Acceptedfor publication: 25 September 1995

Abstract

Generalised spike-and-wave (SW) spindles (5-7 Hz) associated with myoclonic jerks precede the occurrence of regular spikes (2-3 Hz) associated with convulsive seizure induced by picrotoxin. SW spindles occur spontaneously in rodent and cat under some experimental conditions and are considered to be models of human generalised epilepsy. These spindles have been proposed as being led by a thalamic pacemaker. To examine this possibility in picrotoxin-induced SW spindles and seizure spikes, we recorded EEG using chronically implanted unipolar electrodes during intravenous picrotoxin infusion in freely behaving rat. The 6 EEG signals were digitally sampled at 1000 Hz. Linear correlation, spectral, coherence and phase analyses were undertaken to determine time differences (TDs) between EEG channels and the brain structure leading seizure activity. One frontal cortex led all other structures during SW spindles. TJI between SW spindles in the leading frontal cortex (Frl) and the contralateral Frl was 3.6 f 0.5 msec. All ipsilateral structures (hippocampus, thalamus, amygdala, caudate nucleus and occipital cortex) were delayed by more than 3 msec from Frl (intralaminar thalamic nuclei - by 6.3 k 0.9 msec). TDs of SW spindles between subcortical regions were less than 1.5 msec. Similar relationships with slightly smaller TDs were found with spikes during convulsive seizure except TDs between frontal cortices did not significantly differ from zero. We suggest that seizure activity induced by picrotoxin is led by one Frl during SW spindles and by both frontal cortices working as one system during convulsive seizure.

Keywords:

General&d epilepsy; EEG; Spike-and-wave;Seizure propagation; Time differences

1. Introduction Among different epileptic disorders, primary generalised epilepsy (PGE) remains least studied. A common

genetic basis of different lkinds of PGE accompanied by its electrographical correlate, spontaneous generalised spikeand-wave (SW) discharge, has been proposed (Gloor et al., 1982). There are several experimental models of PGE. One of them is spontaneously occurring episodes of behavioural quiescence accompanied by cortical, bilaterally synchronous, high amplitude bursts in different strains of

Corresponding author. Centre for Neuroscience and Dept. of Medicine, Flinders Medical Centre, Bedford Park, SA 5042, Australia. ’Suowrted bv grants from the National Health and Medical Research Council.. _ ??

0013~4694/96/$15.00 0 1996 Elsevier Science Ireland Ltd. All rights reserved SSDI 0013-4694(95)002251

laboratory rats reminiscent of absence episodes in human (labelled high-voltage spindles by Buzs&i et al. (19881, SW activity by Robinson and Gilmore (1980) and Van Luijtelaar and Coenen (1986)). Another model of PGE is feline generalised penicillin epilepsy induced in anaesthetized cats by systemic administration of penicillin, a blocker of gamma-aminobutyric acid (GABA)-mediated central inhibition (Prince and Farrell, 1969; Gloor and Testa, 1974). Feline generalised penicillin epilepsy is also characterised by cortical bilaterally synchronous SWs originating from normal thalamocortical spindles (Kostopoulos et al., 1981a, b). Steriade et al. (1985) have revealed a pacemaking role of the reticular thalamic CRT) nucleus in the generation of normal spindles as a result of its rhythmic inhibitory bursts of 6 or 10 Hz targeting the thalamo-cortical relay cells (see also Steriade

EEG 95052

158

A. Meduedeu et al./ Electroencephalography and ciinicul Neurophysiology 98 11996) 157-166

et al., 1990). A similar role has been confirmed in the rat absence model where SW spindles were severely affected after lesion of the RT (Buzsaki et al., 1988). These data suggest that the thalamus might have a pacemaking role in different types of SW spindles. Another agent, the GABA-receptor linked chloride channel antagonist picrotoxin, causes generalised SW activity accompanied by myoclonic jerks which may lead to generalised convulsions (Kaplan and Williamson, 1978; King, 1979; Massotti, 1985). In spite of the well-known pharmacological properties of picrotoxin, neurophysiological mechanisms of seizure activity due to its action remain to be established. Using comprehensive methods of analysis of electroencephalographic (EEG) activity it is possible to measure small time differences (TDs) between two EEG channels and, as a result, to define the brain region leading others in a particular type of electrical activity. These methods are based on the Fast Fourier Transformation (FFT) of multiple EEG signals and cross-spectral phase/coherence analyses (Brazier, 1972; Gotman, 198 1, 1983) as well as on linear (r*) or non-linear (/z’> correlation or regression analyses (Cohn and Leader, 1967; Guilford and Fruchter, 1985; Femandes de Lima et al., 1990; Pijn et al., 1990). PGE models have not been studied carefully by these methods and it has only been shown that TDs between seizure discharges in homologous areas of different hemispheres in patients with PGE are less than 5 msec (Gotman, 1983). In this study, inter- and intrahemispheric time differences between electrographic epileptiform events in generalised seizure activity caused by systemic administration of picrotoxin were measured with phase and linear correlation ( r2) methods. We used these methods to determine if there was a leading structure in SW spindles and seizure spikes. Specifically, we tested if thalamus led the epileptic activity and showed that, in this model, frontal cortex led.

2. Methods and materials The rats and experimental procedure were described in our previous paper (Willoughby et al., 1995) and will be reviewed briefly here. We used 10 male inbred SpragueDawley rats (400-500 g), known not to spontaneously express spike and wave discharges (Willoughby and Mackenzie, 1992).

2.1. Electroencephalography Animals were prepared with a jugular venous catheter for picrotoxin infusion and with surface and intracerebral EEG electrodes under anaesthesia (pentobarbitone 60 mg/kg intraperitoneally). Surface electrodes consisted of stainless steel screws inserted through the skull but not

penetrating the dura and positioned as follows: an indifferent electrode anteriorly over the frontal sinus, 2 frontal electrodes symmetrically 2.5 mm from the midline and 2 mm anterior to bregma (frontal cortex (Frl) left and right) and one electrode 6 mm posterior to bregma (occipital cortex) and an earth electrode in the occipital bone overlying the cerebellum. Intracerebral electrodes (20% iridium80% platinum 25.4 pm wire insulated except at the tip in 30-gauge stainless steel tubes) were stereotaxically implanted. Posterior (P) and lateral (L) positions relative to bregma and ventral (V) to the brain surface were: thalamus P = 1.8, L = 1.2, V = 5.2; caudate-putamen P = 1.3, L = 4.4, V = 6; hippocampus P = 3.3, L = 2, V = 3; amygdala P = 3.3, L = 4.5, V = 9 (mm). Fine wires micro-welded to the steel screws and nichrome wires were soldered to a female integrated circuit socket, all of which were embedded in dental cement. Animals were allowed to recover for at least 7-10 days and then placed in cages inside isolation recording chambers to acclimatise for 3 days. The venous line and EEG lead were connected the day before infusion so that animals were studied in isolation, without handling or restraint. Video-monitoring was also undertaken. Monopolar recordings were made via an FET amplifier (Buzs&i et al., 1988). Computer-digitized 6-channel EEG recording (MacLab, AD Instruments, Australia) was made throughout the experimental procedure (20-40 min) starting 3 min before the first picrotoxin infusion. Half-amplitude cut-off frequencies were at 1 and 100 Hz. All electrodes positions were verified after experiments by Nissl staining of 100 (cLrnbrain sections. The thalamic electrodes were mainly in the paracentral region (intralaminar thalamic group). 2.2. Infusions Intravenous infusions were made using a Model 341B Sage Pump delivering 0.1 ml/min. Picrotoxin 1 mg/ml (Sigma, St. Louis, MO, USA) was dissolved in 10% dimethylsulphoxide (Ajax Chemicals, Sydney, Australia) in 0.9% saline and heparin (David Bull Labs, Mulgrave, Australia) 30 units/ml. The infusate was administered intermittently for 1 min every 3 min and was continued until a bilateral tonic-clonic convulsion (usually lasting 15-30 set) was induced. All infusions were commenced between 08.30 and 09.30 h. 2.3. Data analysis Programs for off-line EEG analysis were written using Igor software (WaveMetrics, USA) and implemented on a Macintosh Quadra 950 computer. All episodes of seizure activity free from artifacts and interference were analysed. EEG potentials were measured relative to the average amplitudes of background activity before picrotoxin infusion. Linear correlation, spectral, coherence and phase analyses of EEG signals were undertaken pairwise to

A. Medvedev et al./ Electroencephalography

determine time differences between EEG channels and the brain structure leading seizure activity. As an example, the results of TD measurement between the thalamus (L) and

and clinical Neurophysiology

1.59

98 (19%) 157-166

frontal cortex (L) are presented in Figs. 1 and 2. Statistical tests for normality were done with Minitab software (Minitab Inc., USA).

Frl (L).

AmygU-1

Hip (L)

-4

i

0.3 mV 1 set

0

10

20

30

40Hz

-6&m

0

10

0

10

20

Freqtency

30

40Hz

30

40Hz

-40

20

0 zo Time shift

40

so

F

7s

Pdsitio2n of &o&

in Zegm6ent

Fig. I. Picrotoxin-induced generalised spike-and-wave (SW) activity. Measurement of time difference (ID) using phase (C-D) and linear correlation (E-F) techniques. A: 3 SW spindles (first two are separated by background activity, indicated by vertical dashed lines) are combined in one EEG segment to calculate the sample spectra and the sample time differences between all EEG channels (one measurement). Frl (L) and Frl (R), frontal cortex, left and right; Amyg, amygdala; Hip, hippocampus; Thal, thalamus. B-F: the result of ID measurement between thalamus (L) and frontal cortex (L). B: power spectra (solid line, frontal cortex; thin line, thalamus), linear scale, relative units. C: phase spectrum between thalamus (L) and frontal cortex (L) (thick line), coherence between thalamus (L) and frontal cortex (L) (dashed line) and the lower bound of the 99% confidence interval of the coherence (thin line). D: phase estimates with significant coherence (taken from C, markers) and their regression tit (solid line). E: correlation function between thalamus (L) and frontal cortex (L) for a single 1 set epoch commencing at 3.5 set from the beginning of the EEG segment in A. Peak value of correlation occurs at lag of 11 msec (point A). F: plot of ‘TDs obtained for all half-overlapping epochs in EEG segment in A (solid line) and the corresponding peak values of correlation (dashed line). Epochs with peak values of correlation below 0.4 are excluded from analysis (shown as zero points in TD graph). Point A corresponds to the point A in E.

A. Medvedev et al./ Electroencephalography and clinical Neurophysiology 98 (1996) 157-166

160

2.3.1. Phase analysis

dashed lines) is shown. After each EEG epoch in a particular segment had been weighted by a Hamming spectral window and subjected to FIT, the smoothed spectrum of this segment was calculated as an average of all epochs. Two to 6 segments of SW activity were analysed in each animal. The duration of regular spike runs during tonicclonic convulsions varied from 15 to 30 set and therefore they were long enough to allow analysis of l-3 segments in each animal. For each EEG segment, the smoothed power spectra (Fig. IB), coherence (Fig. lC> and phase spectrum (Fig. lC> were calculated for each pair of EEG

To get statistically reliable results in spectral analysis, it is necessary to average spectra from a number of epochs (Jenkins and Watts, 1968). Therefore, we took EEG segments 6-12 set long and divided them into half-overlapping 1 set epochs (1 l-23 epochs in each segment). The duration of one spike-and-wave spindle varied from 2 to 6 set and, therefore, 2-3 separate consecutive SW spindles were usually combined in the segment for one measurement. In Fig. IA such a segment consisting of 2 SW episodes (3 set and 5 set long, separated by vertical

A Frl (L)

Thal (

1 set

0

10

20

30

-

40Hz

i -6Oms

-40

-20

I 0

20

40

60

Time shift 0

10

20

30

40Hz

-

m5 IO -

E

P 0

30

10

40Hz

generalised

-10 1

Posit

Freqvency Fig. 2. Regular spikes during picrotoxin-induced

-5

convulsive

iZon of 3epfA

in z&me,“;

seizure. Figure structure and all notations are the same as in Fig.

I.

A. M’edvedev et al./ Electroencephalography

channels in the range of l-40 Hz with a frequency resolution of 1 Hz. The lower 99% confidence limit of coherence was calculated (Jenkins and Watts, 1968) and only those points on the phase spectrum were selected which were in the range of significant coherence (P < 0.01; Fig. 1D). If there were 8 or more of these points, a weighted leastsquares regression line (Knapp and Carter, 1976; Ktonas and Mallart, 1991) was fitted to the phase estimates (Fig. 1D) and a time difference was calculated from the slope of this line (Gotman, 1981, 1983). The error of the time delay was based on the error of the regression line slope. 2.3.2. Linear correlation analysis For each epoch, the linear correlation function of the two EEG signals was calculated in a range of time shifts from - 60 to + 60 msec with time resolution 1 msec (Fig. 1E). To exclude cases with low linear relationship between channels, only those epochs where the absolute extreme value (maximum or minimum) of the correlation function was above 0.4 were taken into consideration (similar to Allen et al. (1992), who took epochs with peak correlation above 0.5). The abscissa of the correlation peak gives the TD between EEG channels (point A on Fig. lE), the negative value means that the second channel (in this case, left frontal cortex) leads. The TDs obtained for each epoch in a particular EEG segment were plotted on a graph for all consecutive epochs. As an illustration, the TD for one epoch commencing at 3.5 rjec in EEG segment in Fig. 1A corresponds to the point A. in Fig. 1E and it is plotted as point A in Fig. 1F. The mean value of TDs over all epochs of a particular EEG segment was calculated and this was considered as a single measurement. Hereafter, when considering a TD for a particular pair of EEG channels, we will designate it as “between the first signal and the second signal.” A positive TD would mean that the first signal leads the second and vice versa. We used the two estimates of TD obtained by both methods as two independent estimates in each animal. The mean values of TDs obtained by both methods and averaged over all animals were compared to zero using Student’s t test.

3. Results

and clinical Neurophysiology 98 (19%) 157-166

161

previously (Willoughby et al., 1995). All electrical seizures were accompanied by behavioural seizures. In Fig. 1 two separate spike-and-wave spindles and the result of TD measurement between the thalamus (L) and the frontal cortex (L) are presented. The duration of spindles varied from 2 to 6 sec. Each spindle consisted of relatively slow waves (150-200 msec) and high amplitude (up to 1 mV> relatively short (40-60 msec) spikes. Spikes were positive in subcortical structures and predominantly negative or negative-positive in the neocortex (Fig. 1A). The power spectrum of spindles usually had two or three clear peaks, the main peak was at the fundamental frequency (5-7 Hz) and the others were the first (lo-14 Hz) and the second (20-28 Hz) harmonics (Fig. 1B). During spike-and-wave activity, coherence was significant in a wide range of frequencies with peak values at the fundamental frequency and at harmonic frequencies (Fig. lC>. Because of significant coherence between spindles in different brain areas (usually above 0.5 at peak values), it was possible to get a number of phase estimates in each phase spectrum sufficient for reliable measurements of time differences (Fig. 1D). The negative slope of the regression line of the phase spectrum in Fig. 1D meant that the second structure, in this case the left frontal cortex, led the first structure, i.e., the thalamus. The regression line would cross the ordinate axis at approximately 180”, meaning that spindles in cortex and in thalamus were opposite in polarity. The average time difference calculated from the correlation function was also negative. In Fig. 2, the segment of EEG activity during a seizure in the same animal and similar analyses between the thalamus (L) and the frontal cortex (L) are presented. Usually, spike-and-wave spindles occurred more frequently as more picrotoxin was infused. The main electrographic change manifesting a seizure was the abrupt disappearance of the spindle pattern and its replacement by another pattern - a relatively long (15-30 set) run of regular spikes without significant amplitude modulation (Fig. 2A). These seizure spikes were similar to the spindle spikes but they were separated by “slower” slow waves (300-400 msec) or by flat periods, so that there was a decrease in the fundamental frequency from 5-7 Hz to 2.5-3 Hz (Fig. 2B). The coherence, phase and correlation characteristics of the seizure activity were similar to that of the spindle activity.

3.1. General 3.2. Interhemispheric time differences After 15-40 min, picrotoxin induced repeatedly occurring short runs of high-voltage spike-and-wave spindles which were generalised from the start (Fig. 1A) as previously described for GABA antagonists (Kaplan and Williamson, 1978; King, 1979; Ben-Ari et al., 1981; Massotti, 1985). SW spindles were accompanied by myoclonic jerks, followed finally by tonic-clonic convulsions confined to the head and upper limbs, without loss of posture control, so called “restricted seizures” as described by us

3.2.1. SW spindles

All time differences measured by correlation and phase methods during SW spindles, i.e., between the right frontal cortex and the left frontal cortex, are plotted in Fig. 3A. The TD distributions for the SW spindles appeared to be bimodal: there were two peaks, to the right and to the left, from the zero point of the histogram separated by a trough around the zero point for both methods (Fig. 3A). The

162

A. Meduedeu

et al./

Electroencephalography

and clinical

Neurophysiology

(A) Frl Right-Left Time Differences

98 (1996) 157-166

Frl Leading-Trailing Time Differences

12 E 10 % %!8 5 : 6

15

p2 -10

-5

0

5

10

15

) byCorrelation

m ) by Phase

g E ; 8 10 0 %

$4 0 -15


I

9 5

(B) Comparison to Normal Distribution

Bins(rrus)

lb

1'2

14

Fig. 4. Distribution histograms of time differences between leading and trailing frontal cortices during picrotoxin-induced spike-and-wave activity.

-is

-io

_j

5

Binsqms)

m m

lb

1.5

positive or negative away from zero, that is, one frontal cortex (right or left) always led the other. The “positive” peak of the histogram had a larger amplitude than the “negative” one: in 29 cases the right cortex led and in 12 cases the left cortex led. To measure by how much the leading hemisphere (regardless right or left) was ahead, we plotted the distribution of absolute values of TDs and calculated the means, 3.5 f 0.5 msec by correlation and 3.7 + 0.5 msec by phase (Fig. 4). Both values significantly differed from zero (P < 0.05). The regression coefficient between time differences calculated by two methods was equal to 0.7 k 0.1, showing a significant relationship (P < 0.001).

by Correlation (~~0.1) byPhase(pc0.05)

Fig. 3. Interhemispheric time differences during picrotoxin-induced spikeand-wave activity. A: distribution histograms of all TDs between frontal cortices measured by phase and correlation methods. B: the graph of differences between the cumulative histograms of the observed TDs and the corresponding normal distributions with the same means and standard deviations for both methods used. The positive deviations around - 10 msec and +5 msec mean that such TDs were observed more often than the normal distribution predicted, the negative peak around the zero point means that very small or zero TDs were observed more rarely than the normal distribution predicted.

Ryan-Joiner Normality Test confirmed that these distributions differed from the corresponding normal distributions (P < 0.05 for the phase method and P < 0.1 for the correlation method). TDs around - 10 msec and + 6 msec were observed more often and TDs around zero point were observed more rarely than it would be prescribed by the normal distribution (Fig. 3B). Thus, interhemispheric time difference during SW activity had a strong tendency to be

3.2.2. Seizures The distribution of time differences between the right and the left cortex during seizures had only one peak around positive value of 1 msec (Fig. 5). This distribution did not significantly differ from normal one (P > 0.1). The averaged time differences (0.7 f 1.2 msec by correlation and 0.4 f 0.8 msec by phase) were not significantly different from zero.

Table 1 Intrahemispheric (ipsilateral) time differences between frontal cortex and other structures during picrotoxin-induced spike-and-wave spindles. parentheses, the number of rats and the total number of measurements are shown. The values significantly different from zero are shown by asterisks Trailing side

Leading side Frontal cortex Time difference Correlation Caudate n. Thalamus Amygdala Hippocampus Occipital cortex

3.9* 1.4(3, 3.6f0.8(7, 3.4* 1.3 (4, 3.8& 1.1 (4, 5.1 * 2.2 (4,

Frontal cortex Time difference

(msec) by

12) 18) 12) 8) 12)

* *

* * *

7.8 rto.7 (3, 6.3*0.9(7, 7.0fl.0(4, 5.5 + 1.2 (4, 10.0* 1.0(4,

(msec) by

Correlation

Phase 12) 18) 12) 91 131

* * * * *

-2.4* 1.7 (3, l.O& 1.2(5, 2.0* 1.6 (4, 3.6* 1.0 (5, 4.2*1.7(5,

Phase 10) 18) 13) 19) * 11) *

- 0.3 f 2.4 (3, 3.1 f 1.5 (5, 4.4 * 1.9(4, 5.2*1.2(5, 3.9f 1.0(5,

1I ) 18) * 13) 19) * 14)

??

??

In

A. Medvedev et al./ Electroencephalography

3.3. Intrahemispheric time diferences

Intrahemispheric

frontal cortex and all other ipsilateral structures, were calculated separately for the leading and trailing hemispheres according to the leadership of the frontal cortex, that is, if the right frontal cortex led the left frontal cortex during SWs, we considered the right hemisphere as leading and vice versa. The data are summarised in Table 1. Again, both methods were in agreement that frontal cortex led other structures. In the leading side, all time differences were referred to the frontal cortex and we:re significant and positive, thus the frontal cortex led all other structures including the thalamus. The occipital coatex was most delayed from the frontal cortex (Table 1). The average TDs between all subcortical regions in the leading side were less than 1 msec by correlation and less than 1.5 msec by phase. In total, the time differences between subcortical regions were smaller than 1.5 msec in 89 of 98 measurements (91 %‘o>. In the trailing side intrahemispheric relationships were slightly different. The time differences for the caudate nucleus were negative but not significantly different from zero, whereas they were significantly positive in the leading side. For all other structures, time differences in the trailing side were positive and tended to be smaller than the corresponding values in the leading side though significantly only in occipital cortex by the phase method (3.9 k 1.0 msec vs. 10.0 * 1.0 msec). 3.3.2. Seizures Because in seizure activity we did not demonstrate a leading side, intrahemispheric measurements were performed without separating them into leading and trailing sides. The results were similar: the frontal cortex led all

10

-I’0

Fig. 5.

Time Differences


1

0 Interhemispheric

0.7+1.2 ms ( m 0.4+0.8 ms ( m

Bins (ms)

time differences

) by Correlation ) by Phase

10 during picrotoxin-induced

15 con-

vulsive seizure are shown as distribution histograms of all TDs measured between frontal cortices by com:lation

163

98 (1996) 157-166

Table 2

3.3.1. SW spindles The intrahemispheric time differences, i.e., between the

Frl Right-Left

and clinical Neurophysiology

and phase methods.

(ipsilateral)

time differences

between frontal cortex and

other structures

during picrotoxin-induced convulsive seizures. In parentheses, the number of rats and the total number of measurements are shown. The values significantly different from zero ate shown by asterisks Frontal cortex Time difference (msec) by Phase

Correlation Caudate n. Thalamus Amygdala Hippocampus Occipital cortex

0.3 f 2.6 (4, 4.1rtl.6(6, 3.6 f 2.8 (5, 5.4k 1.6 (5, 4.6 f 2.0 (6,

7) 16) 9) 12) * 11) *

??

2.4 f 4.4f 5.5 f 4.7f 5.6rt

2.7 (4, 1.5 (6, 2.5 (5, 1.1 (5, 1.5 (6,

8) 16) 9) 13) * 11)

??

??

other structures by approximately the same times as in spike-and-wave spindles although TDs did not significantly differ from zero for the caudate nucleus and the amygdala. The thalamic spikes were significantly delayed by more than 4 msec from the frontal spikes (Table 2). Again, the time differences between subcortical regions were smaller than 1.5 msec in 30 of 38 measurements (79%).

4. Discussion Since Brazier (1972) first proposed an approach and subsequently Gotman (1981, 1983) improved a computerised method of phase/coherence analysis of the brain potentials, such methods have been used to study tbe propagation of brain electrical activity in normal and pathological conditions. The analysis is usually based on the consideration of multi-channel EEG records with the assumption of linear or non-linear statistical relationship between signals. Linear approaches use phase spectrum of two signals fitted by regression line (Gotman, 1981, 1983; Boeijinga and Lopes da Silva, 1989; Allen et al., 1992; Kobayashi et al., 1994) or linear correlation function (r2) (Cohn and Leader, 1967; Femandes de Lima et al., 1990). The non-linear methods use the average amount of mutual information (Mars and Van Arragon, 1982) or non-linear correlation ratio (h’) (Guilford and Fruchter, 1985; Pijn et al., 1990). The non-linear methods are supposed to be more robust because they do not assume a simple linear relationship between signals (Femandes de Lima et al., 1990). However, a comparison study did not reveal a significant advantage of non-linear correlation ( h2) compared to the linear analog (r2> (Allen et al., 1992). In our experimental model we found that in most cases of seizure activity correlation and coherence between EEG channels were high enough (> 0.5) to apply linear methods of analysis. We used two linear approaches based on correlation (r2> and phase spectrum. Both methods were consistent and gave us similar results.

164

A. Medvedev et (11./Elecrroencephalography

The results obtained from detailed EEG analysis have been useful in defining the focus of epileptic activity in different experimental and clinical situations (Gotman, 1983; Mars and Lopes da Silva, 1983; Femandes de Lima et al., 1990) and in differentiation of apparently bilateral synchronous epileptic activity between patients with PGE and patients with secondary generalised discharges (Gotman, 1981; Allen et al., 1992; Kobayashi et al., 1994). The main assumption used in this differentiation is that PGE is defined by TDs smaller than 5 msec as first proposed by Gotman (1981). Further analysis of seizure activity in PGE at higher time resolution has not been done. We used high speed digital EEG sampling (1000 samples/set) and in some cases revealed statistically significant TDs smaller than 1 msec. The picrotoxin-induced epilepsy model has been studied mainly with the aim of analysing the role of GABA and GABA-mediated drugs in the genesis and control of epileptic activity, but without detailed analysis of the underlying neurophysiological mechanisms (Kaplan and Williamson, 1978; King, 1979; Massotti, 1985). In our experimental model we observed two types of generalised epileptic EEG activity caused by systemic infusion of picrotoxin. Spike-and-wave spindles (5-7 Hz) accompanied myoclonic jerks, while regular spikes (2-3 Hz) separated by low voltage slow waves or rather flat electrical periods accompanied tonic-clonic convulsions of head and forelimbs. The EEG patterns of these two types of epileptic activity were different in frequency, spike shape and duration of the burst (Figs. 1 and 2). Such differences are probably due to different cellular mechanisms underlying the two types of epileptic activity. We previously observed two behavioural seizure patterns accompanying the 2-3 Hz activity, both bilateral involving head and forelimbs, but differentiated by the additional involvement of hind limbs and, sometimes, loss of postural control (Willoughby et al., 1995). All of the seizures in this study were of the milder type. None of the more extensive type were observed in the present group of animals, the only difference in experimental design being the presence of intracerebral electrodes, a finding that requires further study. In our model, epileptiform activity appeared to be generalised from the start, the very first spike-and-wave complex being observed in all structures studied. This observation is somewhat different to what was reported in other models. For example, by visual inspection of EEG in feline generalised penicillin epilepsy, Fisher and Prince (1977) observed the earliest appearance of SW activity in the cortex while in other regions, including the thalamus, such activity was not yet well developed. In our model, SW activity did vary in amplitude in different structures and sometimes the amplitude of SW complexes was lower in the cortex than in subcortical structures. Despite occasional smaller amplitudes of cortical SWs, the method we used, being independent of EEG potential amplitudes, still

and clinical Neurophysiology 98 (1996) 157-166

revealed leadership by frontal cortex. Therefore, our results are in accord with those of Fisher and Prince (1977) of a significant role of cortex in these different experimental models of SW activity. While we have not demonstrated a unique site of origin of SW spindles and seizure discharges without their presence also in other structures, leadership by the frontal cortex does imply that this structure more rapidly or readily expresses hyperexcitability in response to input from a possible leading structure, or in the absence of one, the frontal cortex might play a causal role. The time differences between SW discharges in contralateral frontal cortices (2-5 msec) found in this study are consistent with the time delays of hippocampal contralateral afterdischarges caused by unilateral electrical stimulation in rat (3-7 msec, Femandes de Lima et al., 1990). Similarly, we assume that our TDs may be accounted for by a propagation of seizure discharges through commissural pathways. During convulsive seizure, the TDs between two frontal cortices are smaller and not significantly different from zero, that is, bilateral synchrony becomes stronger. Such hypersynchrony may be caused by a powerful drive from a third structure strongly connected to both frontal cortices. Another explanation is that the excitability of the trailing hemisphere has increased as a consequence of pre-ictal processes so that impulses from the leading side, propagating through callosal pathways, evoke faster ( < 2 msec) excitation. A third explanation is that commissural connections are antidromically activated (Schwartzkroin et al., 1975), thereby resulting in very rapid excitation of the contralateral cortex. If these mechanisms result in contralateral excitation so fast that it compares to the time of ipsilateral involvement, then it becomes impossible to distinguish which side, left or right, leads the other and, therefore, both cortices can be regarded as a single system. Some intrahemispheric time differences during epileptic discharges may also be explained by propagation through normal neuronal connections. However, the time differences of the cortico-caudate and the cortico-thalamic relationships on the trailing side during SW activity do not differ significantly from zero (Table 1). Such small TDs might be because these structures are directly led by the contralateral frontal cortex as a common pacemaker. Direct connections between cortical areas and contralateral caudate putamen and thalamus have been demonstrated (Veening et al., 1985; Bentivoglio et al., 1991). Thus, we propose that there may be different seizure propagation routes from the leading frontal cortex to the contralateral hemisphere, via contralateral cortex or directly to contralateral subcortical areas. Through which connections the cortical seizure discharge propagates may depend on functional states of the appropriate pathways and the current excitability level of the recipient structures. During both types of epileptic activity, the TDs between subcortical structures were smaller than 1.5 msec in most

A. Afedvedev et al./ Elecrroencephalography and clinical Neurophysiology 98 (1996) 157-166

cases, indicating a high degree of synchronisation. Such synchronisation may be induced by a powerful drive from the leading and/or pacemaking structure. Our main finding is that the frontal cortex leads other brain regions including the thalamus in both types of picrotoxin-induced epileptic activity. In other experimental models of generalised SW activity such as rat absence epilepsy and feline generalised penicillin epilepsy a thalamo-cortical synchronising mechanism has been proposed (Kostopoulos et al., 1981b; Buzs&i et al., 1988). The ability of RT neurones to generate rhythmic bursts of frequency 6 or 10 Hz independently and their pacemakiqg role in generalised spindle-like activity has been clearly revealed in lesion experiments (Steriade et al., 1985; Buzs&i et al., 1988). Because picrotoxin-induced SW pattern is very similar to those observed in other models, it is possible that the RT oscillating system also participates in the formation of this pattern, particularly in determining the period of oscillations. However, the results presented here reveal an important role of the frontal cortex as a structure leading in epileptiform activity. Cortical spikes may interact with the reticula-thalamic system ;in a complex way. Therefore, besides any possible role of thalamo-cortical influences, we suggest an important role for cortico-thalamic projections in the development of SW spindles. As shown in feline generalised penicillin epilepsy, an increased impulse activity underlies cortical spikes (Kostopoulos et al., 198 1b) which may propagate back to the thalamus (Avoli et al., 1983) and other structures and cause a development of secondary spikes. Therefore, although we have no direct evidence, picrotoxin-induoed SW spindles may result from cortico-thalamo-cortical reverberation of neural impulses and the role of RT nucleus may be in phasing these discharges and creating oscillations of a particular frequency (6 Hz). The role of the frontal cortex seems to become even more important during convulsive seizure when the spindle pattern disappears and another rhythmicity of 2-3 Hz appears. Our findings emphasize the role of the cortex, especially, the frontal cortex, in epilepsy induced by generalised impairment of GABA-mediated inhibitory mechanisms.

Acknowledgements Supported by grants from the National Health and Medical Research Council.

References Allen, P.J., Smith, S.J.M. and Scott, C.A. Measurement of interhemispheric time differences in generalised spike-and-wave. Electroenceph. clin. Neurophysiol., 1992,82: 81-84. Avoli, M., Gloor, P., Kostopoulos, G. and Gotman, J. An analysis of

165

penicillin-induced generalised spike and wave discharges using simultaneous recordings of cortical and thalamic single units. J. Neurophysiol., 1983, 50: 819-837. Ben-Ari, Y., Tremblay, E., Riche, D., Ghilmi, G. and Naquet, R. Electrographic, clinical and pathological alterations following systemic administration of kainic acid, bicuculline or pentetnuole: metabolic mapping using the deoxyglucose method with special reference to the pathology of epilepsy. Neuroscience, 1981,6: 1361- 1391. Bentivoglio, M., Balercia, G. and Kruger, L. The specificity of the nonspecific thalamus: the midline nuclei. Progr. Brain Res., 1991, 87: 53-80. Boeijinga, P.H. and Lopes da Silva, F.H. A new method to estimate time delays between EEG signals applied to beta activity of the olfactory cortical areas. Electroenceph. clin. Neurophysiol., 1989.73: 198-205. Brazier, M.A.B. Spread of seizure discharges in epilepsy: anatomical and electrophysiological considerations. Exp. Neurol., 1972, 36: 263-272. Buzst%ki,G., Bickford, R.G., Ponomareff, G., Dial, L.J., Mandel, R. and Gage, F.H. Nucleus basalis and thalamic control of neocortical activity in the freely moving rat. J. Neurosci., 1988, 8: 4007-4026. Cohn, R. and Leader, H.S. Synchronization characteristics of paroxysmal EEG activity. Electroenceph. clin. Neurophysiol., 1967, 22: 421-428. Femandes de Lima, V.M., Pijn, J.P., Nunes Filipe, C. and Lopes da Silva, F.H. The role of hippocampal commissures in the interhemispheric transfer of epileptiform afterdischarges in the rat: a study using linear and non-linear regression analysis. Electroenceph. clin. Neurophysiol., 1990.76: 520-539. Fisher, R.S. and Prince, D.A. Spike-wave rhythms in cat cortex induced by parenteral penicillin. I. Electroencephalographic features. Electroenceph. clin. Neurophysiol., 1977, 42: 608-624. Gloor, P. and Testa, G. General&d penicillin epilepsy in the cat: effects of intracarotid and intravertebral pentylenetetrazol and amobarbital injections. Electroenceph. clin. Neurophysiol., 1974, 36: 499-515. Gloor, P., Metrakos, J., Metrakos, K., Andermamr, E. and Van Gelder, N.M. Neurophysiological, genetic and biochemical nature of the epileptic diathesis. In: R.J. Broughton (Ed.), Henri Gastaut and the Marseilles School’s Contribution to the Neurosciences. (EEG Suppl. No. 35). Elsevier Biomedical Press, Amsterdam, 1982: 45-56. Gotman. J. Interhemispheric relations during bilateral spike-and-wave activity. Epilepsia, 1981, 22: 453-466. Gotman, J. Measurement of small time differences between EEG channels: method and application to epileptic seizure propagation. Electroenceph. clin. Neurophysiol., 1983, 56: 501-514. Guilford, J.P. and Fruchter, B. Fundamental Statistics in Psychology and Education. McGraw-Hill, New York, 1985. Jenkins, M.G. and Watts, D.G. Spectral Analysis and its Applications. Holden Day, Oakland, CA, 1968: 525 pp. Kaplan, B.J. and Williamson, P.D. Electroencephalogram and somatosensory evoked potential changes after administration of six convulsant drugs. Exp. Neurol., 1978, 59 124-136. King, G.A. Effects of systemically applied GABA agonists and antagonists on wave-spike ECoG activity in rat. Neurophannacology, 1979, 18: 47-55. Knapp, C.H. and Carter, G.C. The generalized correlation method for estimation of time delay. IEEE Trans. Acoust. Speech Signal Proc., 1976.24: 320-327. Kobayashi, K., Nishibayashi, N., Ohtsuka, Y., Oka, E. and Ohtahara, S. Epilepsy with electrical status epilepticus during synchrony. Epilepsia, 1994, 35: 1097-l 103. Kostopoulos, G., Gloor, P., Pellegrini, A. and Siatitsas, I. A study of the transition from spindles to spike and wave discharge in feline generalized penicillin epilepsy: EEG features. Exp. Neural., 1981a, 73: 43-54. Kostopoulos, G.. Gloor, P., Pellegrini, A. and Gotman, J. A study of the transition from spindles to spike and wave discharge in feline generalized penicillin epilepsy: microphysiological features. Exp. Neurol., 1981b. 73: 55-77.

166

A. Medvedev et al./ Electroencephnlogruphy and clinical Neurophysiology 98 (1996) 157-166

Ktonas, P.Y. and Mallart, R. Estimation of time delay between EEG signals for epileptic focus localization: statistical error considerations. Electroenceph. chn. Neurophysiol., 1991, 78: 105-I 10. Mars, N.J.I. and Lopes da Silva, F.H. Propagation of seizure activity in kindled dogs. Electroenceph. clin. Neurophysiol., 1983, 56: 194-209. Mars, N.J.I. and Van Arragon, G.W. Time delay estimation in non-linear systems using Average Amount of Mutual Information Analysis. Signal Proc., 1982, 4: 139-153. Massotti, M. Electroencephalographic investigations in rabbits of drugs acting at GABA-benzodiazepine-barbitnrate/picrotoxin receptors complex. Pharmacoi. B&hem. Behav., 1985, 23: 661-670. Pijn, J.P.M., Vijn, P.C.N., Lopes da Silva, F.H., Van Ende Boas, W. and Blanes, W. Localization of epileptogenic foci using a new analytical approach. Neurophysiol. Clin., 1990, 20: l-l 1. Prince, D. and Farrell, D. “Centtencephalic” spike-wave discharges following parenteral penicillin injection in the cat. Neurology, 1969, 19: 309-310. Robinson, P.F. and Gilmore, S.A. Spontaneous generalised spike-wave discharges in the electrocorticograms of albino rats. Brain Res., 1980, 201: 452-458. Schwartzkroin, P.A., Futamachi, K.J., Noebels, J.L. and Prince, D.A.

Transcallosal effects of a cortical epileptiform focus. Brain Res., 1975, 99: 59-68. Steriade, M., Deschenes, M., Domich, L. and Mulle, C. Abolition of spindle oscillations in thalamic neurons disconnected from nucleus reticularis thalami. J. Neurophysiol., 1985, 54: 1473-1497. Steriade, M., Gloor, P., Llinls, R.R., Lopes da Silva, F.H. and Mesulam, M.-M. Basic mechanisms of cerebral rhythmic activities. Electroenceph. clin. Neurophysiol., 1990, 76: 481-508. Van Luijtelaar, E.L.J.M. and Coenen, A.M.L. Two types of electrocortical paroxysms in an inbred strain of rats. Neurosci. Lett., 1986, 70: 393-397. Veening, J.G., Comelissen, F.M. and Lieven, P.A.J.M. The topical organization of the afferents to the caudatoputamen of the rat. A horseradish peroxidase study. Neuroscience, 1985, 5: 1253- 1268. Willoughby, J.O. and Mackenzie, L. Nonconvulsive electrocorticographic paroxysms (absence epilepsy) in rat strains. Lab. Anim. Sci., 1992, 42: 55 l-554. Willoughby, J.O., Mackenzie, L., Medvedev, A. and Hiscock, J.J. Distribution of Fos-positive neurons in cortical and subcortical structures after picrotoxin-induced convulsions varies with seizure type. Brain Res., 1995, 683: 73-87.