Sleep EEG synchronization mechanisms and activation of interictal epileptic spikes

Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

www.elsevier.com/locate/clinph

Sleep EEG synchronization mechanisms and activation of interictal epileptic spikes Franco Ferrillo a,*, Manolo Beelke a, Lino Nobili a,b a

Center of Sleep Medicine, Chair of Neurophysiopathology, DISMR, University of Genoa, Genoa, Italy b Child Neuropsychiatry, DSN, Gaslini Institute, University of Genoa, Genoa, Italy

Abstract Objective: The temporal course of sleep interictal epileptic discharges (IEDs) has been studied focusing their relationship with the temporal course of the main sleep-EEG frequency bands that is thought to re¯ect the action of different synchronization neural mechanisms. The existence of a mutually exclusive mechanism between spindles and delta waves should be re¯ected in a mutually exclusive facilitation of IEDs activation by slow wave activity (SWA) and sigma activity (SA) during synchronized NREM sleep. Methods: We reanalyzed data from 19 children and 15 adult patients affected by different partial epileptic syndromes. The temporal series of SWA, SA and theta band (TB), derived from spectral analysis, were obtained from a spike-free and pathologic alteration-free derivation, controlateral to the most active lead, where the IEDs count was performed. Relationships between SA, SWA and TB and time series of IEDs were tested by means of correlation techniques after data normalization. Results: A positive correlation of spike distribution with SWA time course has been found in the majority of adults. Only a few adult patients showed IEDs that were correlated with SA or TB. Conversely SA was shown to be positively correlated with spiking in many different epileptic syndromes of childhood. Moreover, in the contest of the NREM sleep cycle an inverse relationship between the SWA and SA mode of spike activation has been detected. Conclusions: Overall results give evidence that 3 main rhythmic spectral components that characterize sleep EEG can exert positive in¯uences on IEDs production. Our studies demonstrate that within NREM sleep the facilitating in¯uences on IEDs production exerted separately by either spindle activity or delta synchronization mechanisms can be detected. Moreover, a mutually exclusive mechanism between SA and SWA oscillations is detectable in the opposite relationship of the correlation between IEDs and the two bands in the central part of the NREM cycle. q 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Epilepsy; Interictal epileptic activity; Slow wave activity; Sigma activity; Spindles; Delta waves

1. Introduction The activation of interictal epileptic discharges (IEDs) by sleep is a well known feature of many epileptic syndromes both in children and in adults. Studies on the relationship between sleep and the promotion of IEDs generation in human pathology are generally based on data obtained from conventional sleep scoring in stages. Conventional scoring rules individuate rather rigid states and their stepwise temporal course does not allow the evaluation of the dynamics between the different synchronization and desynchronization processes. All night spectral analysis reveals quantitative and dynamic aspects that can not be recognized on the basis of traditional scoring rules. The time course of * Corresponding author. Cattedra di Neuro®siopatologia, Largo R. Benzi 10, 16132 Genova, Italy. Tel.: 139-010-353-74-65; fax: 139-010-353-7699. E-mail address: ®®@dism.unige.it (F. Ferrillo).

the major synchronization mechanism is expressed by the values of delta power (slow wave activity (SWA), 0.5±4.0 Hz), the dynamics of which show a progressive decline across the consecutive NREM cycles; this temporal course describes a damped sinusoid waxing and waning in a series of peaks and troughs corresponding to NREM sleep cycles and to REM sleep episodes, respectively (Lubin et al., 1973). Another mechanism of cortical synchronization related to the production of sleep spindles is represented by sigma activity (SA, 12.0±16.0 Hz). Within the NREM sleep cycle SA oscillates inversely with SWA showing a parallel course during REM sleep. The time course of SA does not show a progressively declining pattern. The SA values reach peaks at the beginning and at the end of each NREM cycle and its values are relatively stable overnight. The desynchronized EEG pattern occurring during REM is represented by the time course of relative values of slow alpha and theta frequencies that oscillate reciprocally with SWA. Quantitative analysis techniques have been used to

1388-2457/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S 1388-245 7(00)00404-1

CLINPH 1034

S66

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

assess relationships between the pattern of temporal distribution of IEDs and the above described synchronizing and desynchronizing mechanisms (Malow et al., 1997, 1998; Nobili et al., 1999a,b, 2000; Ferrillo et al., 2000; Beelke et al., 2000). They allowed a higher correlation between IEDs and SA to be highlighted in many childhood epileptic syndromes, while SWA correlated with IEDs better than SA in partial adult epilepsy. The aim of the present study is to review our data, focusing on the intracycle dynamics of SWA and SA. It will take into account their mutual relationship as well as the possibility of a mutually exclusive role of their underlying mechanisms, both in the production of spindles and delta waves, respectively, and in the promoting functions of IEDs. 2. Sleep structure Sleep structure is in¯uenced partly by the previous wakefulness duration, of which sleep is a compensatory homeostatic response, and partly by the circadian chronobiological system (Daan et al., 1984). However, internal mechanisms that regulate the cyclic alternation between NREM and REM sleep determine the intrasleep structure (Massaquoi and McCarley, 1992). NREM sleep consists of 4 stages in which the level of EEG synchronization progressively grows, reaches a plateau and declines in parallel with the depth of sleep. A sleep cycle consists of the progressive and rather orderly occurrence of NREM sleep stages and of the subsequent REM sleep period. Such a cycle, which represents the basic and repetitive model of sleep organization, recurs 4±5 times during the night. Each time the level of synchronization achieved is lower and it con®gures the classic macroscopic sleep structure where stages of deep sleep 3 and 4 are predominant in the ®rst part of the night and the duration of REM sleep episodes is progressively growing. EEG features during sleep depend largely on different functional states of the thalamocortical system and result from membrane properties and from oscillations in the membrane potentials (LlinaÂs and PareÂ, 1991). At sleep onset the ®ring rate of midbrain reticular formation and mesopontine cholinergic nuclei diminishes, thus removing an excitatory drive from cortical and thalamocortical elements and allowing the membrane potential to reach a higher level of hyperpolarization. A further hyperpolarization level of the membrane at the beginning of light NREM sleep leads to the appearance of oscillations in the frequency range of spindles in the reticular nucleus of the thalamus and in thalamocortical cells (Steriade et al., 1993a). As sleep deepens slow cortical oscillations (less than 1 Hz) begin organizing small territories by thereafter recruiting larger ones through coupling mechanisms. The onset of the depolarizing phase of cortical slow oscillations also triggers and further synchronizes thalamus-generated spindles (Amzica and Steriade, 1998). Approaching deep sleep and in parallel with the degree of hyperpolarization of the thalamocortical

membrane potential, spindle oscillations are progressively replaced by slow frequency oscillations in the delta range (Steriade et al., 1993a). During stage 4 relay cells reach the maximum level of hyperpolarization and a delta sequence can hence be triggered by direct corticothalamic volleys in a more synchronized manner (Amzica and Steriade, 1998). During the transition to REM sleep inputs coming from activating cholinergic pontine REM-on cells (McCarley and Hobson, 1975) lead to a relative depolarization of the thalamocortical cells which blocks thalamocortical oscillations both in the spindle and delta range. The major features of this organization at EEG level can be appreciated by observing the time course of temporal series extracted by spectral analysis of the EEG. SWA probably re¯ects the level of synchronization of cortical thalamocortical neurons during their higher hyperpolarized status. The analysis of the SWA temporal course shows the typically monotonic decline over successive NREM sleep episodes. By sleep deprivation or nap experiments (Borbely and Achermann, 1992) it has been demonstrated that the prevalence of SWA depends on the duration of prior wakefulness. These features suggest that the maximum level of synchronization achievable by the cortical thalamocortical system (the one that can give rise to delta waves in surface EEG) is ruled by a homeostatically-regulated and sleep±wake-dependent process. The spindle synchronization mode seems to be relatively independent of homeostatic in¯uences and more linked to mechanisms of synchronization build-up and decline. In contrast with the declining trend of SWA in the course of sleep, SA shows a rather stable or moderately increasing trend (Aeschbach and Borbely, 1993). Though SA cannot be equated with spindle activity, within this frequency range spindles are organized and it may be assumed that the SA temporal course gives information about spindle number, amplitude and duration. Typically within NREM sleep SA shows a bimodal pattern with an initial and a terminal peak which give rise to a U-shaped curve. In the middle part of the NREM sleep cycle SWA and SA oscillate inversely. This pattern of interrelationship between SWA and SA is particularly evident in the ®rst part of the night when homeostatic pressure for delta synchronization is at the highest level. In the later NREM sleep episodes, when delta synchronization is reaching its lowest levels, the inverse relationship between SWA and SA seems to be attenuated (Uchida et al., 1991; Merica and Fortune, 1997). During REM episodes both SWA and SA are reduced to minimum values. According to the reciprocal interaction model (McCarley and Hobson, 1975) the NREM±REM sleep cycle is generated by the interplay of discharges coming from two brainstem cell groups (REM-on and REM-off cells) regulated by a reciprocal self-inhibition and self-excitation mode. The desynchronizing system is the expression of the balance between REM-on REM-off neurones and its time course is re¯ected by the temporal

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

S67

Fig. 1. (A) Temporal series of SWA, SA and relative values of theta activity (upper) and the corresponding hypnogram (lower) in a single subject. (B) Temporal series of spikes/min plotted together with SWA in a single representative adult subject affected by partial epilepsy. (C) Temporal series of spikes/min plotted together with SA in a single representative subject affected by LKS. (D) Temporal series of spikes/min plotted together with relative values of theta activity in a single representative adult subject affected by partial epilepsy.

course of spectral bands either from theta, slow alpha relative values or beta activity. Summing up, sleep structure can be viewed as the result of a dynamic balance between one desynchronizing system acting as an ultradian oscillator and two systems of synchronization (Fig. 1A).

3. Spindles and IEDs EEG spindles constitute the ®rst level of brain electrical synchronization during NREM sleep. They originate from the synaptic interaction of the reticular thalamic nucleus with thalamocortical cells and cortical pyramidal neurons. At the cellular level, during light sleep bursts of the reticular thalamic nucleus inhibit thalamocortical neurons through a GABAergic mechanism. This inhibition leads to the appearance of rhythmic inhibitory postsynaptic potentials (IPSPs) in thalamocortical neurons. These are mainly GABAb receptor-mediated IPSPs and induce burst ®ring of the thalamocortical cells by activating low threshold Ca 21 potentials. The burst pause ®ring of thalamocortical neurons converges into the reticular thalamic neurons and self-facilitates their rhythmic oscillations. At the same time IPSPs induce an excitatory postsynaptic potential (EPSPs) in cortical pyramidal cells. This event allows sleep spindle waves to appear on surface EEG (Steriade et al., 1993a). A relationship between spindle activity and spikes and

waves generation has been experimentally proved by Gloor (1979) in the feline generalized epilepsy model. It has been observed that when the excitability of the cortex is increased, the thalamocortical oscillations that normally induce spindles as well as the evoked thalamocortical or corticothalamic responses mimicking spindles on the surface EEG may develop into paroxysmal synchronization and induce spike and wave bursts (Gloor et al., 1990; Steriade and Contreras, 1995). This mechanism seems to be mainly related to the GABAergic receptor properties of thalamocortical circuitry since GABAb receptor activity can increase the incidence of spike and wave discharges (Liu et al., 1992; Hosford et al., 1992). According to Kellaway et al. (1990) a spindle-related mechanism of spike activation could explain the relationship between spindling and the distribution of spikes and waves in human generalized epilepsy. Indirect evidence of the transition from spindles to spikes and waves is derived from the observation that the progressive increase of serum anti-epileptic drugs reduces the number of spikes and waves and increases the rate of spindles (Kellaway et al., 1990). Polygraphic data of children affected by absence seizures show that generalized IEDs are mainly facilitated by stage 2 (Ross et al., 1966; Sato et al., 1973). A preferential activation of focal spikes has been reported in partial epilepsy during the phase of sleep electroencephalographically characterized by spindles both in scalp (Perria et al., 1966) and depth EEG ®ndings (Gentilomo et al., 1975;

S68

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

Montplaisir et al., 1980). As for spectral analysis techniques, our group has found a higher correlation between spike distribution and temporal series of SA with respect to SWA in partial epilepsy of childhood. This ®nding characterizes sleep of benign epilepsy with rolandic spikes (BERS) (Nobili et al., 1999a), symptomatic or cryptogenetic epilepsy of childhood strongly activated by sleep (Nobili et al., 1999b), benign epilepsy with occipital paroxysms (BEOP) (Beelke et al., 2000), and Landau±Kleffner syndrome (LKS) (Nobili et al., 2000) (Fig. 1C). This higher correlation with SA seems to be more frequent in childhood; nevertheless, a pattern where spike distribution correlates more with SA than with SWA has been found in 3 young adult patients out of 15 suffering from partial or cryptogenic epilepsy activated by NREM sleep (Ferrillo et al., 2000). In a similar study Malow et al. (1998) reported a good correlation between spikes and SWA in sleep of adult epileptic patients affected by partial epilepsy; however, they reported that in some subjects IEDs were more likely to occur on the ascending limb of SWA, i.e. at sleep onset and at the beginning of each NREM cycle. In this temporal span spindles are largely represented and both SA and SWA show a common ascending time course.

interconnected bursting neurons, reverberating volleys could eventually lead to epileptic-like activity. Several clinical studies seem to con®rm a positive relationship between delta production and spike activation chie¯y in partial epilepsy. Studies on both scalp (Sammaritano et al., 1991) and depth electrode recordings (Lieb et al., 1980; Rossi, 1984) have demonstrated that in most patients IEDs are activated preferentially by deep NREM sleep (stages 3 and 4). To de®ne more precisely the relationship between the intensity of delta production and the occurrence of IEDs, studies applying quanti®ed power techniques seem to be useful. Malow et al. (1997, 1998), exploring only the time series of absolute log delta power as a continuous measure of sleep depth, found a positive relationship between spike distribution over time and sleep depth during the whole night in a population of adults affected by partial epilepsy. Recently, Ferrillo et al. (2000) explored the correlation between spike distribution over time during the whole night and temporal series of the values of SWA and SA. They proved the existence of a higher correlation of IEDs with SWA with respect to SA in 12 subjects out of 18 (Fig. 1B).

4. Delta sleep and IEDs

5. REM sleep and IEDs

At the cellular level delta potentials occurring during the late stage of NREM sleep are intrinsically generated by thalamocortical neurons at a higher level of membrane hyperpolarization (Steriade et al., 1993a; Amzica and Steriade, 1998). The transition from light sleep to deep sleep is therefore accompanied by a progressive hyperpolarization of thalamocortical neurons which is probably determined by the decrease of ®ring rates of brainstem cholinergic and monoaminergic neurons that reach their minimum level of activity during stage 4 of sleep. Such a mechanism of hyperpolarization seems to be a GABAa receptor process, since GABAa agonists play a positive role in NREM sleep maintenance and in the synchronization of slow EEG waves and a negative role in spindle-related processes (Faulhaber et al., 1997). The relationship between delta synchronization mechanisms and focal epileptogenesis is not completely known yet. It is already well established that a link between hypersynchronization and epileptogenesis does exist. If too many neurons ®re at the same time this ampli®cation may go awry and lead to an epileptic seizure (Steriade et al., 1994). Intrinsically bursting slow oscillating cortical cells have the propensity to transform their evoked response into self-sustained paroxysmal barrages of action potentials modulated by a slow rhythm whose frequency is similar to that of slow oscillations. As stated by Steriade et al. (1994) the fact that this evolution might also take place in animals with extensive lesions of thalamic nuclei projecting to the recorded area suggests that in cortical networks of

In the transition from NREM to REM sleep activating peducolopontine cholinergic or locus coeruleus noradrenergic nuclei block slow rhythms and promote an activated EEG pattern. Slow sleep oscillations are replaced by fast rhythms (20±40 Hz activities) which are presumably synchronized since they are recordable as EEG potentials (Steriade et al., 1993b) overimposed to theta and slow alpha EEG activities. This fast cortical activity was also observed during natural states of wake and dreaming sleep in humans (LlinaÂs and PareÂ, 1991). In clinical studies REM sleep shows antagonistic behaviour on epileptic events of surface EEG usually accompanied by a marked drop in the number of IEDs. In depth EEG recordings spiking may persist but is less frequent and is usually limited to the `hard core' of the epileptogenic focus (Rossi, 1984; Wieser, 1991). This effect seems basically due to the enhanced inhibition of the aspeci®c thalamocortical synchronization with consecutive inhibition of the spread of epileptic events (Pompeiano, 1969). However, several reports emphasize a positive connection between REM sleep and the occurrence of ictal and interictal discharges, especially in mesiobasal limbic structures (Passouant and Cadillhac, 1970). The activation of IEDs during REM sleep can be explained by the reciprocal behaviour of the mesiobasal limbic structures and neocortex during sleep in the different phases of sleep. The mesiobasal limbic structure seems to be synchronized at the level of theta rhythm when the neocortex is aroused (Kaytor et al., 1957; Lopes da Silva and Arnolds, 1978). The occurrence of hypersynchro-

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

nized theta activity in the hippocampus of epileptic patients during REM sleep has been described (Wieser, 1991; Liu et al., 1992). Data derived from spectral analysis (Ferrillo et al., 2000) demonstrated a high correlation between theta band relative values and spike occurrence during stage 1 and REM sleep in 3 out of 18 patients affected by partial epilepsy, while an inverse correlation with SA and SWA was observed (Fig. 1D).

S69

performing a correlation analysis limited to the part of the ®rst NREM episode. Data obtained from the ®rst cycle, in which a negative correlation between SA and SWA is clearly evident, are coherent with data obtained from successive NREM cycles and the con¯icting role of the two bands is highlighted. We analyzed subjects that showed a higher correlation of IEDs with SA (group 1: 9 BERS, 7 children affected by symptomatic or cryptogenetic epilepsy, 3 LKS, 3 adults

6. Diverging effects of SWA and SA on IEDs activation The above data are indicative of a mutually exclusive role of the spindle and delta oscillating modes in the synchronization processes. At the cellular level intracellular recordings con®rmed an incompatibility between spindle and delta oscillations in thalamocortical neurons of the cat (NunÄez et al., 1992). Spindles appeared at a membrane potential between 255 and 265 mV, whereas delta oscillations occur when the membrane potential is between 268 and 290 mV. A number of reports, both experimental (Lancel et al., 1992) and on human sleep (Uchida et al., 1991; Aeschbach and Borbely, 1993; Merica and Fortune, 1997), have con®rmed the existence of a negative correlation between SA and SWA at least in the central part of NREM sleep episodes. Such features are well described in the transition probability model proposed by Merica and Fortune (1997). Within NREM sleep episodes the sigma curve peaks early at the beginning of the episode while delta peaks just after the centre. A strong negative SA and SWA correlation occurs in the central zone after the sigma maximum where SWA is rising and SA is falling. After reaching its maximum delta begins to fall and sigma reaches a second maximum. At the end of the NREM episode SA and SWA fall together. The existence of a mutually exclusive mechanism between SA and SWA oscillations should be re¯ected in a mutually exclusive facilitation of IEDs activation by SWA and SA during synchronized NREM sleep. The studies of our group show that both SWA and SA were positively correlated when considering the whole night pro®le. In some epileptic syndromes of childhood SA showed a better correlation with IEDs than SWA, while in adult partial epilepsy most patients showed a higher correlation between SWA and IEDs with respect to SA. This feature could be easily explained considering that both bands show a parallel rise at the beginning of the NREM cycles and a parallel decline at the end of the NREM cycles and during REM episodes. When the analysis was limited to NREM sleep a prevalent correlation of one band with the other was highlighted. For a better evaluation of the relationship between IEDs distribution and the time course of spectral bands, we reanalyzed data from our subjects, both adults and children, by

Fig. 2. (A) Temporal series of SWA and SA during the ®rst NREM sleep cycle in a single representative subject. (B) Temporal series of spikes/min plotted together with SWA in the ®rst NREM cycle in a single representative adult subject affected by partial epilepsy. (C) Temporal series of spikes/ min plotted together with SA in the ®rst NREM cycle in a single representative subject affected by LKS.

S70

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

correlation can be detected. Its duration ranges between 27 and 51% of the total NREM cycle. In group 1 the distribution of IEDs over time parallels the time course of SA while an inverse dynamic with respect to SWA is noticed in the middle part of the cycle (Fig. 2C). In Table 1 the nominal signi®cance level of the single correlation coef®cients between temporal series of IEDs and both SA and SWA is reported. As for the whole NREM cycle, all correlation coef®cients between SA and counts of IEDs are high and statistically signi®cant at a 0.0001 nominal level. In the same context SWA values do not correlate signi®cantly with counts of IEDs in 14 out of 21 cases. The mean value of the correlation of counts of SA-IEDs for the whole group is signi®cantly greater than zero (mean 0.68, P , 0:0001), while the counts of SWA-IEDs do not differ signi®cantly from zero. As for the segment with diverging time course of sigma± delta activity, correlation coef®cients between SA and IEDs counts are still highly positive and highly signi®cant, while correlation between SWA and IEDs counts are always signi®cantly negative. Mean values are signi®cantly higher than zero for SA (mean 0.72, P , 0:0001) and signi®cantly lower than zero for SWA (mean 20.64, P , 0:0001). In group 2 the distribution of IEDs over time shows an

affected by symptomatic or cryptogenetic epilepsy) and 12 adult patients affected by partial epilepsy that showed a higher correlation between SWA and IEDs (group 2). For details on recording techniques and data analysis see Nobili et al. (1999a). In the present study correlation analysis techniques have been separately applied to the ®rst NREM episode considered as a whole and to the part of the ®rst NREM episode in which a negative correlation between SA and SWA was evident. The limits of this segment were de®ned for each subject by computing the correlation coef®cient between SA and SWA for a 30 min moving window shifted step by step during NREM episodes. The presence of a single continuous sequence of negative coef®cients allowed the de®nition of the segment taking the centre of the window as a reference point. The temporal series of SA and SWA show their typical features (Fig. 2A). At the beginning of NREM, SA and SWA build up in parallel; later on they show an inverse temporal course where the increase of SWA is paralleled by a decreasing trend of SA. Approaching the beginning of REM sleep both SA and SWA decay together. In all subjects (except one in group 1 and two in group 2) a long central segment characterized by a strong sigma±delta negative

Table 1 Pearson correlation coef®cients (r) for both delta and sigma bands in group 1 (subjects with BERS (A), LKS (B), children with partial symptomatic or cryptogenetic epilepsy (C) and adults with partial symptomatic or cryptogenetic epilepsy (D)) a Subject

NREM

Diverging zone

SWA

P value

SA

P value

SWA

P value

SA

P value

A

1 2 3 4 5 6 7 8

0.02 20.04 0.03 0.55 0.38 20.13 20.10 0.49

NS NS NS * * NS NS *

0.64 0.90 0.70 0.82 0.84 0.62 0.74 0.78

* * * * * * * *

20.93 20.69 20.42 20.20 20.68 20.83 20.87 20.87

* * * ** * * * *

0.91 0.87 0.73 0.60 0.88 0.55 0.91 0.85

* * * * * * * *

B

9 10 11

0.27 20.14 20.27

** NS **

0.84 0.77 0.96

* * *

20.81 20.67 20.90

* * *

0.93 0.93 0.97

* * *

C

12 13 14 15 16 17 18

0.19 20.13 20.09 0.31 0.11 20.15 0.01

NS NS NS ** NS NS NS

0.52 0.66 0.42 0.76 0.75 0.45 0.52

* * * * * * *

20.53 20.63 20.41 20.57 20.45 20.39 20.53

* * * * * * *

0.54 0.56 0.53 0.61 0.72 0.49 0.53

* * * * * * *

D

19 20 21

20.37 20.05 0.03

* NS NS

0.53 0.73 0.41

* * *

20.72 20.93 20.43

* * *

0.54 0.88 0.62

* * *

0.04

NS

0.68

***

20.64

***

0.72

***

Mean a

The correlation is computed for the whole NREM cycle and for the central segment in which the delta and sigma activity negatively correlate (diverging zone). Symbols represent the nominal signi®cance levels of correlation coef®cients: *P , 0:0001; **P , 0:01; NS, not signi®cant. Means are computed for z values and then anti-transformed to obtain the estimated r values; *** indicates a statistically signi®cant difference from zero mean (P , 0:0001).

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

S71

Table 2 Pearson correlation coef®cients (r) for both delta and sigma bands in group 2 (adults affected by partial symptomatic or cryptogenetic epilepsy) a Subject

NREM

Diverging zone

SWA

P value

SA

P value

SWA

P value

SA

P value

1 2 3 4 5 6 7 8 9 10

0.76 0.83 0.59 0.92 0.83 0.87 0.79 0.85 0.86 0.84

* * * * * * * * * *

0.18 0.21 20.35 20.62 20.24 20.01 0.46 20.41 20.57 0.18

NS ** * * ** NS * * * NS

0.87 0.67 0.93 0.94 0.72 0.93 0.93 0.96 0.91 0.63

* * * * * * * * * *

20.65 20.66 20.74 20.91 20.53 20.76 20.76 20.95 20.89 20.51

* * * * * * * * * *

Mean

0.81

***

20.12

NS

0.85

***

20.74

***

a The correlation is computed for the whole NREM cycle and for the central segment in which the delta and sigma activity negatively correlate (diverging zone). Symbols represent the nominal signi®cance levels of correlation coef®cients: *P , 0:0001; **P , 0:01; NS, not signi®cant. Means are computed for z values and then anti-transformed to obtain the estimated r values; *** indicates a statistically signi®cant difference from zero mean (P , 0:0001).

opposite picture paralleling the time course of SWA and showing inverse dynamics with respect to SA (Fig. 2B). In Table 2 the nominal signi®cance level of the single correlation coef®cients between temporal series of IEDs and both SA and SWA is reported. As for the whole NREM cycle, all correlation coef®cients between SWA and IEDs counts are high and statistically signi®cant at a 0.0001 nominal level. In the same context SA values do not correlate signi®cantly with counts of IEDs in 3 out of 10 cases. The mean value of the correlation of counts of SWA-IEDs for the whole group is signi®cantly greater than zero (mean 0.81, P , 0:0001), while the counts of SA-IEDs do not differ signi®cantly from zero. As for the segment with diverging time course of sigma±delta activity, correlation coef®cients between SWA and IEDs counts are still highly positive and highly signi®cant, while correlation coef®cients between SA and IEDs counts are always signi®cantly negative. Mean values are signi®cantly higher than zero for SWA (mean 0.85, P , 0:0001) and signi®cantly lower than zero for SA (mean 20.74, P , 0:0001). 7. Conclusions Spectral analysis techniques have been shown to be an optimal tool for the exploration of modulation exerted by sleep on IEDs production. Sleep stages are rather rough oversimpli®cations of EEG features during sleep. The same sleep stage can be characterized by largely variable amounts of power pertaining to different frequency bands; hence, different samples of EEG scored as stage 2 may contain different amounts of SA and SWA depending on their position in the course of the night. In our studies the distribution of IEDs is directly correlated with the level of the power expressed in the EEG by each band; for example, the positive relationship between SWA and spikes can be

highlighted and studied as a continuous variable in the course of the night. Our data con®rm that the activating properties of sleep are not due to the stage per se but depend largely on the level of activity of synchronizing mechanisms as stated by Shouse et al. (1996). Overall, the results con®rm that 3 main rhythmic spectral components that characterize sleep EEG can exert positive in¯uences on IEDs production. This ®nding con®rms early reports of our group (Ferrillo et al., 1987). Our data also con®rm that the activation of spikes during REM sleep is the less frequent mode of activation that seems to be related to the presence of synchronized theta activity in the EEG. The activation of IEDs both during REM sleep and during stage 1 could re¯ect the similarities between EEG features of these stages (Borbely, 1982) and could also be explained by the occurrence of highly synchronized theta activity in deep temporal structures (Yu et al., 1997). Our studies demonstrate that within NREM sleep the facilitating in¯uences on IEDs production exerted separately by either spindle activity or delta synchronization mechanisms can be detected. Moreover, a mutually exclusive mechanism between SA and SWA oscillations is detectable in the opposite relationship of the correlation between IEDs and the two bands in the central part of the NREM cycle. The SWA mode of activation is the most frequent feature in adult partial epilepsy patients. Its temporal course re¯ects the link between spike production and the intensity of sleep homeostatically related to the quantity of prior wakefulness; this mode of activation seems to be linked to the hyperpolarization-activated spike bursts that occur in isolated pools of thalamocortical neurons that generate delta wave oscillations. In principle, topographical factors could affect the distribution over time of IEDs and their relationship with temporal distribution of EEG frequency bands. Our data

S72

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73

are derived from a population composed of mostly temporal lobe epilepsy subjects (14), a minority of frontal lobe epilepsy subjects (3) and one single subject with occipital lobe epilepsy. The distribution of subjects within the 3 correlation groups is not indicative of a close link to topographical factors. Temporal lobe epilepsy is represented in all 3 groups; two subjects over 3 with frontal lobe epilepsy belong to the group highly correlated with theta. However, the size of the studied population does not allow any meaningful consideration. The build up, maintenance and destruction of EEG synchronization is a ¯uctuating and relatively unstable process. In visual EEG scoring these oscillatory features have been interpreted by Terzano and his group in the framework of the cyclic alternating pattern phenomenon whose role in spike promotion has been well elucidated (Terzano et al., 1991; Parrino and Terzano, 2000). Less than 1 Hz rhythm (Amzica and Steriade, 1998) and additional slower rhythms with a periodicity of 20±40 s have also been identi®ed with differently quanti®ed analysis of sleep EEG (Ferrillo et al., 1997; Achermann and Borbely, 1997). Their possible pivotal role in the processes of synchronization and IEDs modulation should be assessed by proper techniques with higher time resolution. The SA mode of activation has been found to be prevalent in epilepsies of childhood characterized by a strong NREM sleep spike activation. Such an activation is grounded in the spindle producing neural mechanisms. The cellular mechanisms that generate spindles and the ones that allow their transformation in generalized spikes and waves have been depicted above. Studies in progress of our group show a direct correlation between SA time series and the distribution of generalized spike and wave complexes in childhood absence epilepsy. In our opinion the promotion of spikes by the SA mode in focal epilepsy could be explained as an exaggerated response to the physiological mechanisms that evoke spindles; in this case this response should be considered as con®ned within a regional thalamocortical circuitry and involving only circumscribed pools of hyperexcitable neurons in the cortex. References Achermann P, Borbely AA. Low-frequency (,1 Hz) oscillations in the human sleep electroencephalogram. Neuroscience 1997;81:213±222. Aeschbach D, Borbely A. All-night dynamics of the human sleep EEG. J Sleep Res 1993;2:70±81. Amzica F, Steriade M. Electrophysiological correlates of sleep delta waves. Electroenceph clin Neurophysiol 1998;107:69±83. Beelke M, Nobili L, Baglietto MG, De Carli F, Robert A, De Negri E, Ferrillo F. Relationship of sigma activity to sleep interictal epileptic discharges: a study in children affected by benign epilepsy with occipital paroxysms. Epilepsy Res 2000;40:179±186. Borbely AA. A two process model of sleep regulation. Hum Neurobiol 1982;1:195±204. Borbely AA, Achermann P. Concepts and models of sleep regulation: an overview. J Sleep Res 1992;1:63±79. Daan S, Beersma DGM, Borbely A. Timing of human sleep: recovery

process gated by a circadian pacemaker. Am J Physiol 1984;246:R161±R178. Faulhaber J, Steiger A, Lancel M. The GABAA agonist THIP produces slow wave sleep and reduces spindling activity in NREM sleep in humans. Psychopharmacology 1997;130:285±291. Ferrillo F, Balestra V, De Carli F, Nuvoli GF, Romano CM, Rosadini G, Sannita WG. Overnight distribution of paroxysmal activities in epileptic patients. Adv Epileptol 1987;16:677±680. Ferrillo F, Gabarra M, Nobili L, Parrino L, Schiavi G, Stubinsky B, Terzano MG. Comparison between visual scoring of cyclic alternating pattern (CAP) and computerized assessment of slow EEG oscillations in the transition from light to deep non-REM sleep. J Clin Neurophysiol 1997;14:210±216. Ferrillo F, Beelke M, De Carli F, Cossu M, Munari C, Rosadini G, Nobili L. Sleep-EEG modulation of interictal epileptiform discharges in adult partial epilepsy: a spectral analysis study. Clin Neurophysiol 2000;111:916±923. Gentilomo A, Colicchio G, Pola P, Rossi GR, Scerrati M. Brain depth recording of interictal epileptic potentials during sleep in man. In: Levin P, Koella WP, editors. Sleep 1974, Basel: Karger, 1975. pp. 444±446. Gloor P. Generalized epilepsy with spike-and-wave discharge: a reinterpretation of its electrographic and clinical manifestations. Epilepsia 1979;20:571±588. Gloor P, Avoli M, Kostopoulus G. Thalamocortical relationships in generalized epilepsy with bilaterally synchronous spike-and-wave discharges. In: Avoli M, Gloor P, Kostopoulos G, Naquet R, editors. Generalized epilepsy, Boston, MA: BirkaÈuser, 1990. pp. 190±212. Hosford DA, Clark S, Cao Z, Wilson Jr WA, Lin FH, Morrisett RA, Huin A. The role of GABAB receptor activation in absence seizure of lethargic (lh/lh) mice. Science 1992;257:398±401. Kaytor F, Hullay J, Farago L, Haberland K. Effect of barbiturate sleep on the electrical activity of the hippocampus of patients with temporal lobe epilepsy. Electroenceph clin Neurophysiol 1957;9:441±451. Kellaway P, Frost JD, Crawley JW. The relationship between sleep spindles and spike and wave bursts in human epilepsy. In: Avoli M, Gloor P, Kostopoulos G, Naquet R, editors. Generalized epilepsy, Boston, MA: BirkaÈuser, 1990. pp. 36±48. Lancel M, van Riezen H, Glatt A. The time course of sigma activity and slow wave activity during NREMS in cortical and thalamic EEG of the cat during baseline and after 12-hours of wakefulness. Brain Res 1992;596:285±295. Lieb JP, Joseph JP, Engel J, Walker J, Crandall PH. Sleep state and seizure foci related to depth spike activity in patients with temporal lobe epilepsy. Electroenceph clin Neurophysiol 1980;49:538±557. Liu Z, Vergnes M, Depaulis A, Marescaux C. Involvement of intrathalamic GABAB neurotransmission in the control of absence seizures in the rat. Neuroscience 1992;48:87±93. LlinaÂs RR, Pare D. Of dreaming and wakefulness. Neuroscience 1991;44:521±535. Lopes da Silva FH, Arnolds DE. Physiology of the hippocampus and related structures. Annu Rev Physiol 1978;40:185±216. Lubin A, Nute C, Naitoh P, Martin WB. EEG delta activity during sleep as a damped ultradian rhythm. Psychophysiology 1973;10:27±35. Malow BA, Kushwaha R, Lin X, Morton KJ, Aldrich MS. Relationship of interictal epileptiform discharges to sleep depth in partial epilepsy. Electroenceph clin Neurophysiol 1997;102:20±26. Malow BA, Lin X, Kushwaha R, Aldrich MS. Interictal spiking increase with sleep depth in temporal lobe epilepsy. Epilepsia 1998;39:1309± 1316. Massaquoi SG, McCarley RW. Extension of the limit cycle reciprocal interaction model of REM cycle control. An integrated sleep control model. J Sleep Res 1992;1:138±143. McCarley RW, Hobson JA. Neuronal excitability modulation over the sleep cycle: a structural and mathematical model. Science 1975;189:58±60. Merica H, Fortune RD. A neuronal transition probability model for the

F. Ferrillo et al. / Clinical Neurophysiology 111, Suppl. 2 (2000) S65±S73 evolution of power in the sigma and delta frequency bands of sleep EEG. Physiol Behav 1997;62:585±589. Montplaisir J, Saint Hilaire JM, LaverdieÁre M, Walsh J, Bouvier G. Contribution of all-night polygraphic recording to the localization of primary epileptic foci. In: Canger R, Angeleri F, Penry JK, editors. Advances in epileptology, New York: Raven Press, 1980. pp. 135±138. Nobili L, Ferrillo F, Baglietto MG, Beelke M, De Carli F, De Negri E, Schiavi G, Rosadini G, De Negri M. Relationship of sleep interictal epileptiform discharges to sigma activity (12-16 Hz) in benign epilepsy of childhood with rolandic spikes. Clin Neurophysiol 1999a;110:39±46. Nobili L, Ferrillo F, Baglietto MG, Beelke M, De Carli F, De Negri E, Rosadini G, De Negri M. Modulation sleep interictal epileptiform discharges in partial epilepsy of childhood. Clin Neurophysiol 1999b;110:839±845. Nobili L, Baglietto MG, Beelke M, De Carli F, De Negri E, Tortorelli S, Ferrillo F. Spindles inducing mechanism modulates sleep interictal epileptic discharges activation in the Landau-Kleffner syndrome. Epilepsia 2000;41:201±206. NunÄez A, CurroÁ Dossi R, Contreras D, Steriade M. Intracellular evidence for incompatibility between spindle and delta oscillations in thalamocortical neurons of cat. Neuroscience 1992;48:75±85. Parrino L, Terzano MG. Epilepsy, cyclic alternating pattern and mechanisms of sleep regulation. In: Dinner DS, LuÈders HO, editors. Epilepsy and sleep, San Diego, CA: Academic Press, 2000 in press. Passouant P, Cadillhac J. Decharges epileptiques et sommeil. In: Niedermayer E, editor. Epilepsy, Basel: Karger, 1970. pp. 87±104. Perria L, Rosadini G, Rossi GF, Gentilomo A. Neurosurgical aspects of epilepsy: physiological sleep as a means for focalizing EEG epileptic discharges. Acta Neurochir 1966;14:1±9. Pompeiano O. Sleep mechanisms. In: Jasper HH, Ward AA, Pope A, editors. Basic mechanism of epilepsies, Boston, MA: Little Brown, 1969. pp. 453±473. Ross JJ, Johnson LC, Water R. Spike and wave discharges during stages of sleep. Arch Neurol 1966;14:399±407.

S73

Rossi G. Interictal epileptic activity during sleep: a stereo-EEG study in patients with partial epilepsy. Electroenceph clin Neurophysiol 1984;58:97±106. Sammaritano M, Gigli GL, Gotman J. Interictal spiking during wakefulness and sleep and the localization of foci in temporal lobe epilepsy. Neurology 1991;41:290±297. Sato S, Dreifuss FE, Penry JK. The effect of sleep on spike-wave discharges in absence seizures. Neurology 1973;23:1335±1345. Shouse M, Martins da Silva A, Sammaritano M. Circadian rhythm, sleep, and epilepsy. J Clin Neurophysiol 1996;13:32±50. Steriade M, Contreras D. Relations between cortical and thalamic cellular events during transition from sleep patterns to paroxysmal activity. J Neurosci 1995;15:623±642. Steriade M, McCormick DA, Sejnowski TJ. Thalamocortical oscillations in the sleeping and aroused brain. Science 1993a;262:679±685. Steriade M, CurroÁ Dossi R, Contreras D. Electrophysiological properties of intralaminar thalamocortical cells discharging rhythmic (approximately 40 HZ) spike-bursts at approximately 1000 HZ during waking and rapid eye movement sleep. Neuroscience 1993b;56:1±9. Steriade M, Contreras D, Amzica F. Synchronized sleep oscillations and their paroxysmal developments. Trends Neurosci 1994;17:199±208. Terzano MG, Parrino L, Spiaggiari MC, Barusi R, Simeoni S. Discriminatory effect of cyclic-alternating pattern in focal lesional and benign rolandic interictal spikes during sleep. Epilepsia 1991;32:616±628. Uchida S, Maloney T, March JD, Azari R, Feinberg I. Sigma (12-15 Hz) and delta (0.3-3 Hz) EEG oscillate reciprocally within NREM sleep. Brain Res Bull 1991;27:93±96. Wieser HG. Temporal lobe epilepsy, sleep and arousal: stereo-EEG ®ndings. In: Degen R, Rodin EA, editors. Epilepsy, sleep and sleep deprivation, Amsterdam: Elsevier, 1991. pp. 97±119. Yu Q, Rayport M, Farison JB, Dennis MJ, Choi YS. Computer analysis of human depth EEG in different sleep stages. Biomed Sci Instrum 1997;33:184±190.