Epilepsy Research (2010) 89, 17—26

journa l homepage: www.e lsev ier .com/ locate /ep i lepsyres

REVIEW

Thalamo-cortical mechanisms of sleep spindles andspike—wave discharges in rat model of absenceepilepsy (a review)

Evgenia Sitnikova ∗

Institute of Higher Nervous Activity and Neurophysiology, RAS, Butlerova str. 5A, 117485 Moscow, Russia

Received 1 July 2009; received in revised form 26 August 2009; accepted 7 September 2009Available online 13 October 2009

KEYWORDSWAG/Rij rats;EEG coherence;Thalamo-corticalnetwork associations;Time—frequency EEGanalysis;Seizure-precursoractivity;Neuronal modeling

Summary According to the most generally accepted cortico-reticular theory of absenceepilepsy, sleep spindles and spike—wave discharges (SWD, characteristic hallmarks of absenceepilepsy) are closely related. The present review critically evaluates this theory based on theoriginal data obtained in WAG/Rij rat model of absence epilepsy. It is demonstrated that (1)sleep spindles and spike—wave discharges are distinguished in time—frequency domain. (2)EEG waveforms of sleep spindles and SWD are underlain by different synaptic processes, asdetermined with the aid of computational neuronal model of cortical field potentials. (3) Sleepspindles do not precede SWD. EEG analysis of SWD-precursor activity provides us with a clue topossible prediction of absence epilepsy episodes. Furthermore, by studying Granger causalityand EEG coherence at the onset of SWD we gain more insight into the dynamics of the thalamo-cortical neuronal network associations that underlie occurrence of absence seizures. In general,spindle activity and SWD display different time—frequency characteristics as measured in cor-

tex and thalamus, they are accompanied by different neuronal processes and require different involvement of neurotransmitters. Sleep spindles and SWD are considered as autonomous EEGphenomena, and straightforward relationship between them is doubtful.© 2009 Elsevier B.V. All rights reserved.

Contents

Introduction............................................................................................................... 18Relationship between sleep spindles and spike—wave discharges .......................................................... 18Time—frequency EEG analysis ............................................................................................. 19Neuronal model of cortical field potential during sleep spindles and spike-and-wave discharges........................... 20

∗ Tel.: +7 495 334 70 61; fax: +7 495 338 85 00.E-mail address: jenia-s@mail.ru.

0920-1211/$ — see front matter © 2009 Elsevier B.V. All rights reserved.doi:10.1016/j.eplepsyres.2009.09.005

18 E. Sitnikova

EEG characteristics of genuine precursor activity of SWD.................................................................. 22Refinements on applicability of animal models ............................................................................ 24Conclusion remarks........................................................................................................ 25Acknowledgements........................................................................................................ 25

.....

I

Rs(hievIat(ci(Wao(aL

nf1dGtcFfifislDitl12tabPsa

aa

iSatAats

cebrSlrstc

Rs

Atbmtb

deBBwos1(ca(

as to the probability that SWD derive form spindle oscil-

References ................................................

ntroduction

odents with genetic predisposition to absence epilepsy,uch as Genetic Absence Epilepsy Rats from StrasbourgGAERS) and Wistar Albino Glaxo from Rijswijk (WAG/Rij),ave contributed greatly to the understanding of pathophys-ology and neuronal mechanisms of human epilepsy (Danobert al., 1998; Coenen and van Luijtelaar, 2003; Depaulis andan Luijtelaar, 2005; van Luijtelaar and Sitnikova, 2006).n all individuals of these rat strains, absence seizuresppear spontaneously and accompanied by characteris-ic spike—wave discharges (SWD) in electroencephalogramEEG). Behavioral expression of SWD in rats is similar to thelinical manifestation of absence seizures in humans, e.g.,mmobility, minimal facial myoclonic jerks and twitchesvan Luijtelaar and Coenen, 1986; Marescaux et al., 1992).AG/Rij rat model enables predictions about pharmacother-py of epileptic patients (i.e., this model fulfills to a criteriaf predictive validity) and it is based on theoretical groundsconstruct validity), therefore this strain is considered asvalid model of human absence epilepsy (Coenen and van

uijtelaar, 2003).In the late sixties of the last century, basic mecha-

isms of generalized epilepsy were extensively studied ineline generalized penicillin epilepsy (Prince and Farrell,969, see also, Time—frequency EEG analysis, section foretails). Based on experimental data obtained in this model,loor and his co-workers introduced a ‘cortico-reticular’

heory, which claims that sleep-related natural oscillationsan give rise to absence-related SWD (Gloor, 1968, 1969).urther investigations in other animal models have con-rmed that sleep spindles and SWD share many commoneatures. It is known that sleep spindles are character-stic EEG hallmarks of sleep stages I—II. Similar to sleeppindles, SWD preferentially occur during drowsiness andight slow-wave sleep (van Luijtelaar and Coenen, 1986;rinkenburg et al., 1991). Both oscillations are produced

n thalamo-cortical neuronal network, which is formed byhe neocortex, the specific thalamic and the reticular tha-amic nuclei (Steriade and Deschenes, 1984; Steriade et al.,993; Steriade, 2003; Avanzini et al., 1996; Kostopoulos,000; Destexhe and Sejnowski, 2001). Notwithstanding ishat sleep spindles are initiated in the thalamus (Steriadend Deschenes, 1984; Steriade et al., 1993; Steriade, 2003),

ut SWD are triggered by the cortex (Niedermeyer, 1996;olack et al., 2007),1 rapidly distributed over the othertructures throughout intracortical (ipsilateral and callosal)nd descending projections and quickly became general-

1 More specifically, in WAG/Rij rats SWD are triggered by the localrea of perioral projections in the somatosensory cortex (Meeren etl., 2002, 2005).

lca

w

i

........................................................... 25

zed (Steriade and Amzica, 1994; Lemieux and Blume, 1986;itnikova and van Luijtelaar, 2006). Cortex certainly playsprimary role in the pathogenesis of absence seizures, yet

halamic neurons are also implicated in this process (e.g.,vanzini et al., 1992, 2000; Blumenfeld, 2002; Buzsáki etl., 1988; Buzsáki, 1991). Functional aspects of cortico-halamic interactions that underlie development of sleeppindle oscillations and SWD are still not fully understood.

The current paper consists of four parts. The first partritically evaluates the cortico-reticular theory of absencepilepsy (Gloor, 1968, 1969) in respect to the relationshipetween sleep spindles and SWD. The second summarizesesults of time—frequency analysis of sleep spindles andWD. The third part describes neuronal processes under-ying occurrence of sleep spindles and SWD that had beenetrieved from cortical neuronal model. The fourth part con-iders EEG features of SWD-precursor activity as recorded inhe cortex and in the thalamus and evaluates dynamics ofortico-thalamo-cortical associations at seizure onset.

elationship between sleep spindles andpike—wave discharges

ccording to the ‘cortico-reticular’ theory, ‘SWD develops inhe same circuit, which normally generate sleep spindles,y an initially cortical transformation of one every two orore spindle waves to a ‘spike’ component of SWD, while

he next spindle wave or waves are eliminated and replacedy a slow negative wave’ (cited by Kostopoulos, 2000).

‘Cortico-reticular’ theory has gained support from theata obtained in vivo and in vitro in genetic rat mod-ls of absence epilepsy (Avanzini et al., 1992, 1993, 2000;lumenfeld, 2002; Drinkenburg et al., 1993; Kandel anduzsaki, 1997; van Luijtelaar, 1997). In WAG/Rij rats, itas shown that both sleep spindles and SWD preferablyccur during low vigilance states such as drowsiness, lightlow-wave sleep, passive wakefulness (Drinkenburg et al.,991), and characterized by similar EEG characteristicsDrinkenburg et al., 1993). Also distribution of extracellularurrents throughout the cortical depth during sleep spindlesnd during spike—wave discharges2 is known to be the same3

Kandel and Buzsaki, 1997).More recent studies in WAG/Rij rats cast some doubts

ations and, therefore, ‘cortico-reticular’ theory has beenhallenged. In particular, it was found that absence seizuresre triggered by local area in the somatosensory cortex

2 The authors used the term ‘high voltage spindles’ (HVS) for whate call SWD.3 It follows from the analysis of simultaneous EEG and unit record-

ngs in WAG/Rij rats.

at m

Sftbsf2

atnaiaoepd

Thalamo-cortical mechanisms of sleep spindles and SWD in r

(Meeren et al., 2002). These findings lead to a new ‘cor-tical’ theory, which claims that (1) SWD have a local sourcein the neocortex and (2) the thalamus plays secondary rolein absence seizures. We obtained experimental evidencesin favor to ‘cortical’ theory: temporal deactivation of thesomatosensory cortex reversibly reduced the incidence ofSWD (Sitnikova and van Luijtelaar, 2004). In general, ‘cor-tical’ theory argues against generic relationship betweensleep spindles SWD. In order to highlight similarities and dis-tinctions between sleep spindles and SWD, we performedtime—frequency EEG analysis. The next section providessome pro- and counter arguments to the assumption thatSWD are transformed from sleep spindles (Gloor, 1969;Kostopoulos, 2000).

Time—frequency EEG analysis

Sleep spindles are distinguished from SWD in time and infrequency domain (Sitnikova and van Luijtelaar, 2003):

• As compared to sleep spindles, SWD lasted ten timeslonger, their amplitude was nearly two times higher andtheir mean frequency was about 2 Hz lower (Fig. 1A and

B).

• In EEG power spectrum, as measured in frontal EEG, SWDwere distinguished from sleep spindles by having higherpower in beta frequencies and lower power in delta andtheta bands (Fig. 1C); but in the occipital EEG, SWD and

wToeb

Figure 1 Sleep spindles (B) and SWD (A) as recorded in frontal andDistribution of EEG power over the following frequency bands: 1—4(beta), 32—59.5 Hz (gamma1) and 60—100 Hz (gamma2). Two way ANanalysis of EEG power (n = 9 rats). *Significant differences according

odel 19

sleep spindles did not differ much (Fig. 1D) (Sitnikovaand van Luijtelaar, 2003; Sitnikova, 2008; Sitnikova et al.,2009).

Amplitude—frequency parameters of sleep spindles andWD were also examined using continuous wavelet trans-orm (complex Morlet wavelet basis ω0 = 2�, Fig. 2). Inhe wavelet spectrum (Fig. 2B), SWD were characterizedy a vast increase of energy in frequencies >10 Hz. Eachpike component in SWD corresponded to the isolated high-requency burst in wavelet spectrum (Sitnikova et al.,009).

In order to retrieve statistical identities of sleep spindlesnd SWD, we have developed a wavelet-based algorithm forhe automatic identification of spindle events and SWD inon-segmented full-length EEG (Sitnikova et al., 2009). Thislgorithm was tested in 5—7 h EEGs recorded in freely mov-ng animals. In order to achieve the best performance of theutomatic recognition system, it was necessary to select theptimal wavelet template (Morlet, Mexican hat, Harr, Gabor,tc.). Almost 100% of SWD were identified in EEG with highrobability using standard Morlet wavelet, but sleep spin-les were poorly recognized with the aid of the standard

avelet templates (maximum ∼50—60% of true detections).

wo types of adoptive ‘spindle wavelets’ were designed inrder to capture sleep spindles in EEG of our rats. In gen-ral, wavelet analysis has revealed the following distinctionsetween sleep spindles and SWD (Sitnikova et al., 2009):

occipital cortex in 12-month-old male WAG/Rij rat. (C and D).5 Hz (delta), 5—8.5 Hz (theta), 9—13.5 Hz (alpha), 14—31.5 HzOVA (factors ‘band’ and ‘spindles/SWD’ was used for statisticalto post hoc Fisher test.

20 E. Sitnikova

Figure 2 Continuous wavelet transform (complex Morlet wavelet ω0 = 2�) of spike—wave discharges (SWD) and sleep spindles(indicated by arrows). Repetitive spikes in SWD train correspond to the isolated high-frequency bursts in wavelet spectrum. In spiteo EG af uen

(

(

(

eo1ismiSwc

Ndd

CiDscoti

sdsct

smTtraaastcaStifra

fie5tinhibition (Fig. 3). Pyramidal cells received excitatory termi-nals from the thalamus via AMPA- and NMDA-ergic synapseslocated on their apical dendrites. Feed-forward inhibitioninterneurons received AMPA-ergic thalamic input and, in

f the fact that SWD and sleep spindles are well recognized in Erom sleep spindles by having extremely high power in high freq

1) Identification of SWD and spindles required differentwavelet templates, suggesting a fundamental differ-ence between EEG patterns of sleep spindles and SWD.

2) EEG features of SWD are rather uniform with a very low(if any) variability between subjects and also within asubject. EEG waveform of sleep spindles is very incon-sistent and shows large individual differences. Sleepspindles comprise two families: normal spindles ‘type1’ and abnormal spindles ‘type 2’.

3) Sleep spindles and SWD are characterized by signifi-cant increase of wavelet energy in different frequencywindows: sleep spindles—–between 7 and 14 Hz, andSWD—–30 and 50 Hz.

As compared to SWD, spindle oscillations are less gen-ralized. Local sleep spindles are found in frontal andccipital cortical areas in rats (Terrier and Gottesmann,978; Gandolfo et al., 1985; van Luijtelaar, 1997), and alson humans (Jobert et al., 1992). Noteworthy is that sleeppindles and SWD are controlled by different age-relatedechanisms. In WAG/Rij rats, the incidence of SWD highly

ncreases with age, yet non-epileptic rats do not developWD (or just sporadic ones). In non-epileptic rat strain, asell as in WAG/Rij rats, density of sleep spindles does nothange with age (van Luijtelaar and Bikbaev, 2007).

euronal model of cortical field potentialuring sleep spindles and spike-and-waveischarges

omputational models have been used for decades tonvestigate mechanisms of epileptogenesis. For instance,estexhe et al. (2001) presented a model of cortical

eizures, in which inhibition-rebound interactions betweenortical cells accounted for seizure activity. Another modelf cortical seizures (Timofeev et al., 2002) included exci-atory connections of pyramidal neurons in addition tonhibitory inputs from interneurons. Although both models

aa

s 8—12 Hz oscillations, SWD in wavelet space are distinguishedcies (>10 Hz).

imulated some kind of spike-and-wave activity, this activityiffered from the typical absence seizures. The frequency ofpike-and-wave oscillation was low, 1—2 Hz; and the ‘spike’omponent was usually less pronounced as compared to theypical absence patterns.

Computational models give better approximation to realpike—wave seizures when the thalamus is included in theodel (Destexhe, 1998; Destexhe and Sejnowski, 2001).halamo-cortical model4 of SWD (Destexhe, 1998) includeswo types of thalamic cells: thalamo-cortical and thalamiceticular cells, and two types of cortical neurons: pyramidsnd interneurons. Synchronous oscillations (sleep spindlesnd SWD) can be simulated by changing membrane currentsctivated by GABAB receptor-mediated responses. Sleeppindle appear when the cortex stays under strict control ofhe fast GABAA receptor-mediated inhibition. An increase ofortical excitability strengthens cortico-thalamic feedbacknd thalamic circuits can enter slow oscillatory mode (∼3 HzWD). Such a schema can assume a progressive transforma-ion from spindle oscillations to SWD as cortical excitabilitys increased. SWD of higher frequency (∼10 Hz characteristicor rodent models) can also be simulated by the same neu-onal model, based on a different balance between GABAA

nd GABAB receptors (Destexhe, 1998).Recently we have constructed a cortical neuronal model

or simulating local field potentials during SWD and dur-ng sleep spindles (Sargsyan et al., 2007). In our model,lectrical activity was generated by a cortical cylinder of00 �m diameter filled up with pyramidal cells and with twoypes of interneurons provided feed-forward and recurrent

4 See also Destexhe, A., 2007. Spike-and-wave oscillations. Schol-rpedia 2 (2), 1402. http://www.scholarpedia.org/article/Spike-nd-wave oscillations.

Thalamo-cortical mechanisms of sleep spindles and SWD in rat model 21

Figure 3 Schematic representation of cortical neuronal model that was developed for simulating local field potentials of sleepyram

d syninhibcaus

tfct

aaieaow

tmAaGionRribs

tb

spindles and spike—wave discharges. The schema shows two pfeed-forward inhibition and another for recurrent inhibition) anexcitatory synaptic contacts, AMPA and NMDA; and two types ofby the constant frequency of 10 Hz (this frequency is chosen, be

turn, sent inhibitory terminals (GABAa- and GABAb-ergic) toapical dendrites of pyramidal cells. Pyramidal cells had lat-eral excitatory connections (via AMPA and NMDA receptors)with other pyramidal cells and AMPA-ergic connections withrecurrent-inhibition interneurons. These interneurons madeinhibitory contacts of GABAa type on somata of pyramidalcells.

Electrical field potentials were calculated under assump-tion that a large number of identical and similarly orientedpyramidal cells uniformly occupy a cylindrical slab of givenradius. Since both spindles and SWD in rats appear with afrequency of 8—12 Hz, we used 10 Hz frequency to simulatethalamic input.

In the model experiments, it was shown that the shapeof population field potentials determined by the degreeof synchronization in the population of pyramidal cells.Synchronous firing of pyramidal cells led to the occur-rence of sharp high-voltage ‘spike’ component in simulatedfield potentials that resembled the real ‘spike’ of SWD.A decrease of synchronization between pyramidal cellsresulted in gradual smoothing of the ‘spike’, and the fieldpotential transformed into a sine wave, i.e., a typicalwaveform of sleep spindles. In order to investigate themechanism, by which synaptic currents could influence thewaveform of field potential, we changed the strengths ofsynaptic connections.

The effect of AMPA-ergic input ‘thalamus’ → ‘pyramidalcell’ distinguished from the effect of AMPA-ergic synapticcontacts between pyramidal cells (Fig. 4). In particu-

lar, thalamic input to pyramidal cells via AMPA synapsesaffected negative component of simulated field potential.This implies that thalamic excitation is primarily contributedto the ‘spike’ component of SWD. AMPA-ergic contactsbetween pyramidal cells affected the shape and the ampli-

Smdtt

idal cells, two types of interneurons (one is responsible foraptic interactions between those cells. There are two types ofitory contacts, GABAa and GABAb. Thalamic input is simulatede sleep spindles and SWD in rats have a frequency of 8—12 Hz).

ude of positive component (‘positive transient’ that wasollowed by the ‘spike’), suggesting that AMPA-ergic intra-ortical mechanisms are involved in shaping of ‘positiveransient’ in SWD (Table 1).

The present model also helped to assess the role of AMPA-nd GABA-ergic mechanisms in waveform of sleep spindlesnd SWD (Table 1). In both SWD and spindle modes, anncrease or decrease of the synaptic strength of GABAa-rgic contacts between recurrent-inhibition interneuronsnd pyramidal cells resulted to, correspondingly, a positiver negative shift of the entire field potential, but the overallaveform of field potential did not change.

Changes in the strength of AMPA-ergic synaptic con-acts between pyramidal cells caused the most pronouncedodification of the structure of field potentials, whereasMPA-ergic inputs from thalamus to pyramidal cells hardlyffected the waveform of both SWD and sleep spindles.ABAa,b-ergic synaptic inputs from feed-forward inhibition

nterneurons to pyramidal cells influenced the amplitudef both SWD and sleep spindles in a straightforward man-er: the larger the inhibition, the larger the amplitude.einforcement or weakening of GABAa-ergic inputs fromecurrent-inhibition interneurons to pyramidal cells resultedn positive/negative shift of field potential against baseline,ut did not influence the waveform of both SWD and sleeppindles.

In general, it was found that the shape of field poten-ial substantially depended on the degree of synchronizationetween pyramidal cells. In case of sleep spindles and

WD, the assumption about the invariability of trans-embrane currents is a significant simplification. Theifference between SWD and sleep spindles in respect toheir waveforms may either be explained by differences inhe composition of trans-membrane currents, characteriz-

22 E. Sitnikova

Figure 4 Results of neuronal model simulations, indicating that the waveform of SWD depends on the strength of synaptici plexf . (C)i

ious

Ea

Is

52smba

nteractions. (A) Four elements of human spike-and-wave comrontal SWD of WAG/Rij rat (Sitnikova and van Luijtelaar, 2007)ncrease of synaptic interactions.

ng two different states of network activity; or by the effectf spatiotemporal summation of trans-membrane currentsnder the conditions of high (in case of SWD) and low (sleeppindle) synchronization of cortical neuronal activity.

EG characteristics of genuine precursor

ctivity of SWD

nvestigations in GAERS clearly demonstrate thatpike—wave seizures are preceded by medium-voltage

tseis

Table 1 The role of AMPA- and GABA-ergic synaptic inputs in sha

Synaptic input Change

SWD m

AMPAThalamus → pyramidalcells

Reduction The sppositiv

Increase No sign

AMPAPyramidalcell ↔ pyramidal cell

Reduction Negati

Increase The poprofou

GABAa,bFeed-forward inhibitioninterneurons → pyramidalcells

Reduction The waspike d

Increase The wapositiv(the sp

GABAaRecurrent-inhibitioninterneurons → pyramidalcells

Reduction Positivpotent

Increase Negatifield p

as determined by Weir (1965). (B) Weir’s components in theModification of the waveform of SWD caused by a decrease or

—9 Hz oscillations (Pinault et al., 2001; Pinault et al.,006). The presence of 5—9 Hz precursors of absenceeizures has never been confirmed neither in other animalodels, nor in human patients. Considering the close resem-lance between GAERS and WAG/Rij rat strains (Danober etl., 1998; Depaulis and van Luijtelaar, 2005), we expected

hat SWD in WAG/Rij rat would also be preceded by theame 5—9 Hz oscillations. In order to test this hypothesis, wexamined time—frequency EEG parameters of 1 s episodesmmediately before the onset of SWD (so-called precur-or epochs of SWD, or preSWD epochs) in WAG/Rij rats

ping the waveforms of sleep spindles and SWD.

s in EEG waveform

ode Spindle mode

ike is decreased thee transient is increased

Very little increase of thepositive deflection

ificant changes No significant changes

ve shift of field potential Negative deflection isincreased (amplitude of fieldpotential is increased)

sitive transient isndly increased

The negative deflection isdisappeared; the positivedeflection is increased (a phaseshift of field potential)

ve is increased (theid not change)

The positive deflection isdecreased (amplitude of fieldpotential is decreased)

ve is decreased; thee transient is increasedike did not change)

The positive deflection isincreased (the amplitude offield potential is increased)

e shift of the entire fieldial

Positive shift of the entire fieldpotential

ve shift of the entireotential

Negative shift of the entirefield potential

at m

phptpmynt

Thalamo-cortical mechanisms of sleep spindles and SWD in r

(Sitnikova and van Luijtelaar, 2009). Our data suggest thatthere is a remarkable diversity in the waveform and fre-quency profile of preSWD epochs. This somehow disagreeswith Pinault et al.’s (2001) findings in GAERS that medium-voltage 5—9 Hz rhythm constantly precedes the onset ofSWD.

We propose an empirical classification of preSWD basedon significant difference of EEG power as measured intraditional frequency bands, namely, preSWD�, preSWD�,

preSWD� and preSWDn (Fig. 5, Sitnikova and van Luijtelaar,2009). These classes are significantly different in respectto EEG power in characteristic delta—theta—alpha bandsas measured in the frontal EEG. More precise, preSWD�

epochs are characterized by the highest power delta band,

paS

d

Figure 5 Normalized power spectra of preSWD epochs and subsequthe frontal cortex is distinguished from that in the thalamus. The mopreSWDn) show peaks in theta frequencies (∼7 Hz as indicated byabsent in the frontal cortex, but it is present in the thalamus. Durinclosely coincide with each other, however, peak frequencies of SWDshift (∼10 Hz versus ∼9 Hz with approximately 0.5—1.5 Hz difference

odel 23

reSWD�—–the highest power in theta band, preSWD�—–theighest power in alpha band and preSWDn showed the leastower in all investigated frequency bands. It is notewor-hy that 73% of SWD (preSWD� + preSWD�) exhibit a wellronounced theta component (5—9 Hz), suggesting that theajority of SWD in WAG/Rij is preceded by substantial theta,

et this just a theta-frequency component in EEG that isot shaped into a clear rhythm (likewise in GAERS). Besidesheta, almost all preSWD comprised a substantial delta com-

onent, suggesting that co-existence of delta and thetactivities in the cortex is favorable for the occurrence ofWD.

Different classes of preSWD activity showed remarkableifferences in the waveform of cortical and thalamic power

ent SWD. The EEG power frequency profile of preSWD epochs inst remarkable distinction is that all classes of preSWD (exceptarrows). In preSWD� epoch, theta-band component is nearlyg SWD, power spectra in the frontal cortex and the thalamusin the frontal cortex and in the thalamus showed a consistent).

2

sSeeswiasc

naVo(sa

eshmStritat3acq

nabpIcpoa2

tfdiutS

anf

csGG

psmatadwwwicaogr

ssrc

wcplsifopuwmatau

R

Aeen1aOt(aoserIp

4

pectra. However, within a single second after the onset ofWD, cortical and thalamic power spectra closely resembleach other (Fig. 5). It seems that the similarities betweenlectroencephalographic features of cortical and thalamicignals increase with seizure onset. The occurrence of SWDas associated with a sharp ∼10 Hz frequency component

n the frontal cortex and in the thalamus. However, theverage peak frequency in the thalamus (8.5—9.5 Hz) wasignificantly lower than the mean frequency in the frontalortex (10 Hz, p < 0.05).

Next, we examined dynamic changes of network synchro-ization between the cortex (frontal and occipital areas)nd the thalamus (ventroposteromedial and reticular nuclei,PM and RTN correspondingly) during the pre-ictal stagef absence seizures by measuring the ordinary coherenceSitnikova and van Luijtelaar, 2009). EEG coherence analy-is was performed in continuous EEG data, including preSWDnd subsequent SWD.

Intrathalamic coherence. The highest values of coher-nce were found between RTN and VPM irrespective of theubtypes of preSWD epochs, suggesting that these nuclei areighly coupled even before the onset of seizure. Intrathala-ic coherence just slightly changed with the onset of SWD.

ignificant changes were found in transitions from preSWDno SWDn and from preSWD� to SWD�, while coherenceemained unchanged in other cases. In SWDn, coherencencreased in the narrow frequencies, 8—9.5 Hz, as comparedo preSWDn. Transitions from preSWD� to SWD� were char-cterized by a broader increase of coherence, i.e., from 9.5o 17 Hz. Only SWD� showed a decrease of coherence in.5—5 Hz and in 26—31.5 Hz (preSWD� versus SWD�), whilell other seizure subtypes showed only a tendency of thetaoherence (4—9.5 Hz) to be reduced as compared to subse-uent SWD.

Intracortical fronto-occipital coherence. The most sig-ificant increase of coherence was found between 17.5nd 21.5 Hz. In 75% of seizures, coherence in alphaand increased with seizure onset, except transitionreSWD� → SWD�, when alpha coherence did not change.n all investigated transitions, we found an increase ofoherence in harmonic beta frequencies, 17—23 Hz. InreSWD� → SWD� and in preSWD� → SWD�, the increasef coherence in beta frequencies was particularly strongnd widespread (i.e., two- to three-fold increase in1.5—31.5 Hz and in 17—27.5 Hz correspondingly).

Thalamo-cortical coherence. The onset of all seizure sub-ypes correlated with an increase of coherence betweenrontal cortex and the VPM. The most robust increase wasetected in alpha frequencies (8—14 Hz) and a moderatencrease was found in beta frequencies (>14 Hz). A partic-larly strong increase in beta coherence was found at theransition from preSWD� to SWD� and from preSWD� toWD�.

In general, different classes of preSWD epochs were char-cterized by slightly different synchronization pattern, yeto features of EEG coherence can be considered as uniqueor any class of preSWD epoch.

In order to investigate directionality of thalamo-ortical network associations in transition ‘normaltate ↔ paroxysmal state ↔ normal state’, we appliedranger causality concept (Sitnikova et al., 2008). Inranger causality, the knowledge about the immediate

tatfi

E. Sitnikova

ast of one signal is used to predict the future of anotherignal. Linear model of Granger causality was used foreasuring bidirectional coupling strength between cortex

nd thalamus and for studying directionality of informationransfer between cortex and thalamus before, during andfter SWD. It was found that (1) linear Granger causalitieso not change before the onset of SWD; coupling strengthas constant and very low, yet influence thalamus → cortexas stronger than vice versa influence. (2) The onset of SWDas characterized by a rapid increase of coupling strength

n both directions. (3) The strength of thalamus → cortexoupling remained continuously high during the seizuresnd did not return to the initial level even after cessationf SWD. (4) The strength of cortex → thalamus couplingradually diminished shortly after the onset of SWD andeturned to the preSWD level, when SWD stopped.

It is hypothesized that propagation and maintenance ofeizure activity could be facilitated under the strong andustained influence of the thalamus to the cortex, whileapid decrease of cortex → thalamus influence during SWDould promote seizure cessation.

In general, Granger causality appears to be an effectiveay to gain more insight into the dynamics of the thalamo-ortical neuronal network mechanism underlying dynamicalroperties of SWD. We have to admit that even though theinear estimation of Granger causality appeared suitable fortudying thalamo-cortical mechanisms, which are involvedn maintenance and stopping of SWD, it was not sufficientor predicting episodes with absence epilepsy. Applicationf non-linear Granger estimations may help to solve thisroblem. On the other hand, absence seizures might benpredictable by nature (e.g., they might have no signs thatould allow any kind of prediction). At least, computationalodel of thalamo-cortical neuronal network (Suffczynski et

l., 2004) explains the unpredictability of SWD simply byhe fact that fluctuations of control parameters (synapticnd cellular properties of thalamic and cortical cells) arenpredictable by definition.

efinements on applicability of animal models

nimal models of human primary (idiopathic) absencepilepsy are widely used to study various aspects of absencepilepsy, including neuronal, genetic and cellular mecha-isms, as well as treatment and diagnosis (Danober et al.,998; Marescaux et al., 1992; Vadász et al., 1995; Coenennd van Luijtelaar, 2003; Depaulis and van Luijtelaar, 2005).ne of the first animal models of epilepsy was experimen-al model of generalized epilepsy, ‘feline penicillin model’Prince and Farrell, 1969). In this model, penicillin wasdministered in high doses systemically or locally in cortexr thalamus. This suppressed GABA-ergic neurotransmis-ion (Avoli and Gloor, 1982; Kostopoulos, 2000), increasedxcitability of cortical neurons and, finally, led to the occur-ence of generalized bilaterally synchronous SWD in EEG.t should be taken into account that basic mechanisms ofharmacologically induced seizures remarkably differ from

hat in spontaneous epilepsies. Pharmacological treatmentffects neuronal excitability and alters oscillatory activity ofhalamo-cortical network, resulting in an ‘induced’ trans-ormation of sleep spindles to SWD that does not happenn normal conditions. For example, pharmaco-EEG study in

at m

R

A

A

A

A

A

B

B

B

C

D

D

D

D

D

D

D

G

G

G

Thalamo-cortical mechanisms of sleep spindles and SWD in r

genetic WAG/Rij rats, similarly to feline penicillin model,revealed a reciprocal relationship between the number ofanterior sleep spindles and the number SWD5 (van Luijtelaar,1997).

Spontaneous absence epilepsy in rodent genetic mod-els, unlike cats (and, probably, pharmacologically inducedepilepsies in humans), are not obviously correlated withsleep spindle oscillations because of interspecies dif-ferences. In humans and highly developed mammals,GABA-ergic inhibitory interneurons are present throughoutthe entire thalamus, including relay nuclei, in contrast torodents, whose inhibitory interneurons are absent in tha-lamus, except the lateral geniculate nucleus and reticularthalamic nucleus (Jones, 1985). This means that the vastmajority of thalamic nuclei in rats receive only externalinhibitory inputs from the RTN, while intrinsic inhibitionis absent in the largest part of the thalamus. Therefore,thalamo-cortical oscillations in rodents might be governedby slightly different membrane mechanisms, as comparedto other mammals and humans.

Conclusion remarks

According to our analysis, there is no straightforward rela-tionship between sleep spindles and SWD. The geneticWAG/Rij rat model does not demonstrate a direct transi-tion from sleep spindles to SWD (see, Neuronal model ofcortical field potential during sleep spindles and spike-and-wave discharges, section, precursors of SWD), in contrastto what have been convincingly demonstrated in the phar-macological feline penicillin model (i.e., Gloor, 1968, 1969;Kostopoulos, 2000). SWD and sleep spindles appear in EEGas hallmarks of two different states of thalamo-cortical net-work activity (SWD appear when synchronization betweenpyramidal cells in the cortex is high, and sleep spindles—–when synchronization is low). Spontaneous SWD do notemerge from sleep spindles. EEG precursors of SWD arecharacterized by a remarkable diversity in the waveformand frequency profile. In any case, seizure precursors com-prise substantial delta and theta components, suggestingthat co-existence of delta and theta activities in the cortexis favorable for the occurrence of SWD. Granger causalitytest shows that seizure activity sustains under the stronginfluence thalamus → cortex, and rapid attenuation ofinfluence cortex → thalamus during SWD promotes seizurecessation.

Acknowledgements

I gratefully acknowledged the help and scientific supervisionof Dr. Gilles van Luijtelaar. I highly appreciate collaborationwith Prof. Alexander Hramov and Prof. Alexey Koronovsky

and their help with EEG analysis. The study was financiallysupported Russian Foundation for Basic Research (projectNo. 09-04-01302).

5 Phenobarbital and flunitrazepam suppressed the incidence ofSWD, but enhanced anterior sleep spindles; in opposite, clonidineincreased the number of SWD and reduced the number of anteriorsleep spindles.

J

JK

odel 25

eferences

vanzini, G., de Curtis, M., Marescaux, C., Panzica, F., Spreafico,R., Vergnes, M., 1992. Role of the thalamic reticular nucleus inthe generation of rhythmic thalamo-cortical activities subserv-ing spike and waves. J. Neural Transm. 35 (Suppl.), 85—95.

vanzini, G., Vergnes, M., Spreafico, R., Marescaux, C., 1993.Calcium-dependent regulation of genetically determined spikeand waves by the RTN of rats. Epilepsia 34, 1—7.

vanzini, G., de Curtis, M., Franceschetti, S., Sancini, G., Spreafico,R., 1996. Cortical versus thalamic mechanisms underlying spikeand wave discharges in GAERS. Epilepsy Res. 26, 37—44.

vanzini, G., Panzica, F., de Curtis, M., 2000. The role of thethalamus in vigilance and epileptogenic mechanisms. Clin. Neu-rophysiol. 2 (Suppl.), S19—S26.

voli, M., Gloor, P., 1982. Role of the thalamus in generalized peni-cillin epilepsy: observations on decorticated cats. Exp. Neurol.77, 386—402.

lumenfeld, H., 2002. The thalamus and seizures. Arch. Neurol. 59,135—137.

uzsáki, G., Bickford, R.G., Ponomareff, G., Thal, L.J., Mandel,R., Gage, F.H., 1988. Nucleus basalis and thalamic control ofneocortical activity in the freely moving rat. J. Neurosci. 8,4007—4026.

uzsáki, G., 1991. The thalamic clock: emergent network proper-ties. Neuroscience 41, 351—364.

oenen, A.M.L., van Luijtelaar, E.L.J.M., 2003. Genetic animal mod-els for absence epilepsy: a review of the WAG/Rij strain of rats.Behav. Genet. 33, 635—655.

anober, L., Deransart, C., Depaulis, A., Vergnes, M., Marescaux,C., 1998. Pathophysiological mechanisms of genetic absenceepilepsy in the rat. Prog. Neurobiol. 55, 27—57; Pathophysio-logical mechanisms of genetic absence epilepsy in the rat. Prog.Neurobiol. 55, 27—57.

epaulis, A., van Luijtelaar, E.L.J.M., 2005. Genetic models ofabsence epilepsy in the rat. In: Pitkänen, A., Schwartzkroin,P.A., Moshé, S.L. (Eds.), Models of Seizures and Epilepsy. ElsevierAcademic Press, Amsterdam, pp. 233—248.

estexhe, A., 1998. Spike-and-wave oscillations based on the prop-erties of GABAB receptors. J. Neurosci. 18, 9099—9111.

estexhe, A., Contreras, D., Steriade, M., 2001. LTS cells in cerebralcortex and their role in generating spike-and-wave oscillations.Neurocomputing 38, 555—563.

estexhe, A., Sejnowski, T.J., 2001. Thalamocortical Assemblies.Oxford University Press, Oxford.

rinkenburg, W.H., Coenen, A.M., Vossen, J.M., van Luijtelaar, E.L.,1991. Spike—wave discharges and sleep—wake states in rats withabsence epilepsy. Epilepsy Res. 9, 218—224.

rinkenburg, W.H., van Luijtelaar, E.L., van Schaijk, W.J., Coenen,A.M., 1993. Aberrant transients in the EEG of epileptic rats: aspectral analytical approach. Physiol. Behav. 54, 779—783.

andolfo, G., Glin, L., Gottesmann, C., 1985. Study of sleep spin-dles in the rat: a new improvement. Acta Neurobiol. Exp. (Wars)45, 151—162.

loor, P., 1968. Generalized cortico-reticular epilepsies: some con-siderations on the pathophysiology of generalized bilaterallysynchronous spike and wave discharge. Epilepsia 9, 249—263.

loor, P., 1969. Neurophysiological bases of generalized seizurestermed centrencephalic. In: Gaustaut, H., Jasper, H.H., Ban-caud, J., Waltregny, A. (Eds.), The Physiopathogenesis of theEpilepsies. Charles C. Thomas, Springfield, IL, pp. 209—236.

obert, M., Poiseau, E., Jähnig, P., Schulz, H., Kubicki, S., 1992.

Topographical analysis of sleep spindle activity. Neuropsychobi-ology 26, 210—217.

ones, E.G., 1985. The Thalamus. Plenum Press, New York.andel, A., Buzsaki, G., 1997. Cellular-synaptic generation of sleep

spindles, spike-and-wave discharges, and evoked thalamocor-

2

K

L

M

M

M

N

P

P

P

P

S

S

S

S

S

S

S

S

S

S

S

S

S

S

T

T

V

v

v

v

6

tical responses in the neocortex of the rat. J. Neurosci. 17,6783—6797.

ostopoulos, G.K., 2000. Spike-and-wave discharges of absenceseizures as a transformation of sleep spindles: the continuingdevelopment of a hypothesis. Clin. Neurophysiol. (Suppl.) 2,S27—S38.

emieux, J.F., Blume, W.T., 1986. Topographical evolution ofspike—wave complexes. Brain Res. 373, 275—287.

arescaux, C., Vergnes, M., Depaulis, A., 1992. Genetic absenceepilepsy in rats from Strasbourg: a review. J. Neural Transm.(Suppl.) 35, 37—69.

eeren, H.K., Pijn, J.P., van Luijtelaar, E.L., Coenen, A.M., Lopesda Silva, F.H., 2002. Cortical focus drives widespread corticotha-lamic networks during spontaneous absence seizures in rats. J.Neurosci. 22, 1480—1495.

eeren, H., van Luijtelaar, G., Lopes da Silva, F., Coenen, A., 2005.Evolving concepts on the pathophysiology of absence seizures:the cortical focus theory. Arch. Neurol. 62, 371—376.

iedermeyer, E., 1996. Primary (idiopathic) generalized epilepsyand underlying mechanisms. Clin. Electroencephalogr. 27, 1—21.

olack, P.O., Guillemain, I., Hu, E., Deransart, C., Depaulis, A.,Charpier, S., 2007. Deep layer somatosensory cortical neuronsinitiate spike-and-wave discharges in a genetic model of absenceseizures. J. Neurosci. 27, 6590—6599.

inault, D., Vergnes, M., Marescaux, C., 2001. Medium-voltage5—9 Hz oscillations give rise to spike-and-wave discharges in agenetic model of absence epilepsy: in vivo dual extracellularrecording of thalamic relay and reticular neurons. Neuroscience105, 181—201.

inault, D., Slezia, A., Acsady, L., 2006. Corticothalamic 5—9 Hzoscillations are more pro-epileptogenic than sleep spindles inrats. J. Physiol. 574 (Pt 1), 209—227.

rince, D.A., Farrell, D., 1969. ‘‘Centrencephalic’’ spike—wavedischarges following parenteral penicillin injection in the cat.Neurology 19, 309—310.

argsyan, A., Sitnikova, E., Melkonyan, A., Mkrtchian, H., van Lui-jtelaar, G., 2007. Simulation of sleep spindles and spike andwave discharges using a novel method for the calculation of fieldpotentials in rats. J. Neurosci. Methods 164, 161—178.

itnikova, E.Yu., van Luijtelaar, G., 2003. The relation between twotypes of spike—wave discharges and sleep spindles: a spectralanalytical approach. Abstracts of the 25th International EpilepsyCongress, Lisbon, Portugal, October 2003. Epilepsia 44 (Suppl. 8(72)), P147.

itnikova, E., van Luijtelaar, G., 2004. Cortical control of gen-eralized absence seizures: effect of lidocaine applied tothe somatosensory cortex in WAG/Rij rats. Brain Res. 1012,127—137.

itnikova, E., van Luijtelaar, G., 2006. Cortical and thalamic coher-

ence during spike—wave seizures in WAG/Rij rats. Epilepsy Res.71, 159—180.

itnikova, E., van Luijtelaar, G., 2007. Electroencephalographiccharacterization of spike—wave discharges in cortex and tha-lamus in WAG/Rij rats. Epilepsia 48, 2296—2311.

v

W

E. Sitnikova

itnikova, E., Dikanev, T., Smirnov, D., Bezruchko, B.P., vanLuijtelaar, G., 2008. Granger causality: cortico-thalamic inter-dependencies during absence seizures in WAG/Rij rats. J.Neurosci. Methods 170 (2), 245—254.

itnikova, E., 2008. Thalamo-cortical mechanisms of sleep spindlesand spike—wave discharges in WAG/Rij rats: disclosing informa-tion hidden in the electroencephalogram. Ph.D. Thesis (Proef-schrift), RadboudUniversiteit Nijmegen: Ipskamp Partners;http://webdoc.ubn.ru.nl/mono/s/sitnikova e/thalmeofs.pdf.

itnikova, E., Hramov, A.E., Koronovskii, A.A., van Luijtelaar, G.,2009. Sleep spindles and spike—wave discharges in EEG: theirgeneric features, similarities and distinctions disclosed withFourier transform and continuous wavelet analysis. J. Neurosci.Methods 180, 304—316.

itnikova, E., van Luijtelaar, G., 2009. Electroencephalographicprecursors of spike—wave discharges in a genetic rat model ofabsence epilepsy: power spectrum and coherence EEG analyses.Epilepsy Res. 84, 159—171.

teriade, M., Amzica, F., 1994. Dynamic coupling among neocorti-cal neurons during evoked and spontaneous spike—wave seizureactivity. J. Neurophysiol. 72, 2051—2069.

teriade, M., Deschenes, M., 1984. The thalamus as a neuronaloscillator. Brain Res. Rev. 8, 1—63.

teriade, M., McCormick, D.A., Sejnowski, T.J., 1993. Thalamocor-tical oscillation in the sleeping and aroused brain. Science 262,679—685.

teriade, M., 2003. Neuronal Substrates of Sleep and Epilepsy. Cam-bridge University Press, Cambridge.

uffczynski, P., Kalitzin, S., Lopes Da Silva, F.H., 2004. Dynamicsof non-convulsive epileptic phenomena modeled by a bistableneuronal network. Neuroscience 126, 467—484.

errier, G., Gottesmann, C.L., 1978. Study of cortical spindles dur-ing sleep in the rat. Brain Res. Bull. 3, 701—706.

imofeev, I., Bazhenov, M., Sejnowski, T.J., Steriade, M., 2002.Cortical hyperpolarization-activated depolarizing current takespart in the generation of focal paroxysmal activities. Proc. Natl.Acad. Sci. U.S.A. 99, 9533—9537.

adász, C., Carpi, D., Jando, G., Kandel, A., Urioste, R., Horvath,Z., Pierre, E., Vadi, D., Fleischer, A., Buzsáki, G., 1995. Geneticthreshold model of rodent spike-and-wave epilepsy. Am. J. Med.Genet. 60, 55—63.

an Luijtelaar, E.L.J.M., Coenen, A.M.L., 1986. Two types of elec-trocortical paroxysms in an inbred strain of rats. Neurosci. Lett.70, 393—397.

an Luijtelaar, E.L.J.M., 1997. Spike—wave discharges and sleepspindles in rats. Acta Neurobiol. Exp. (Warsz) 57, 113— 121.

an Luijtelaar, G., Bikbaev, A., 2007. Midfrequency cortico-thalamicoscillations and the sleep cycle: genetic, time of day and ageeffects. Epilepsy Res. 73, 259—265.

an Luijtelaar, G., Sitnikova, E., 2006. Global and focal aspects ofabsence epilepsy: the contribution of genetic model. Neurosci.Biobehav. Rev. 30, 983—1003.

eir, B., 1965. The morphology of the spike-wave complex. Elec-troencephalogr. Clin. Neurophysiol. 19, 284—290.