ContentslistsavailableatScienceDirect
NeuroImage
journalhomepage:www.elsevier.com/locate/neuroimage
Focal lesions induce large-scale percolation of sleep-like intracerebral activity in awake humans
S. Russo
a,1, A. Pigorini
a,1, E. Mikulan
a, S. Sarasso
a, A. Rubino
b, F.M. Zauli
a, S. Parmigiani
a, P. d’Orio
b,c, A. Cattani
a,d, S. Francione
b, L. Tassi
b, C.L.A. Bassetti
f, G. Lo Russo
b, L. Nobili
e,i, I. Sartori
b, M. Massimini
a,g,h,∗aDepartment of Biomedical and Clinical Sciences “L. Sacco ”, University of Milan, Milan, Italy
b"C. Munari" Epilepsy Surgery Centre, Department of Neuroscience, Niguarda Hospital, Milan 20162, Italy
cInstitute of Neuroscience, CNR, via Volturno 39E, 43125 Parma, Italy
dDepartment of Psychiatry, University of Wisconsin-Madison, Madison, WI, 53719, USA
eChild Neuropsychiatry, IRCCS Istituto G. Gaslini, Genova 16147, Italy
fDepartment of Neurology, Inselspital, University of Bern, Switzerland
gIRCCS, Fondazione Don Carlo Gnocchi, Milan 20148, Italy
hAzrieli Program in Brain, Mind and Consciousness, Canadian Institute for Advanced Research, Toronto, Canada
iDepartment of Neurosciences, Rehabilitation, Ophthalmology, Genetics, Maternal and Child Health (DiNOGMI), University of Genoa, Genoa, Italy
a r t i c le i n f o
Keywords:
Bistability Effective connectivity Intracranial recording
Radio-frequency thermo-coagulation Stroke
a b s t r a ct
Focalcorticallesionsareknowntoresultinlarge-scalefunctionalalterationsinvolvingdistantareas;however, littleisknownabouttheelectrophysiologicalmechanismsunderlyingthesenetworkeffects.Here,weaddressed thisissuebyanalysingtheshortandlongdistanceintracranialeffectsofcontrolledstructurallesionsinhumans.
ThechangesinStereo-Electroencephalographic(SEEG)activityafterRadiofrequency-Thermocoagulation(RFTC) recordedin21epilepticsubjectswereassessedwithrespecttobaselinerestingwakefulnessandsleepactivity.In addition,Cortico-CorticalEvokedPotentials(CCEPs)recordedbeforethelesionwereemployedtointerpretthese changeswithrespecttoindividuallong-rangeconnectivitypatterns.Wefoundthatsmallstructuralablationslead tothegenerationandlarge-scalepropagationofsleep-likeslowwaveswithintheawakebrain.Theseslowwaves matchthoserecordedinthesamesubjectsduringsleep,areprevalentinperilesionalareas,butcanpercolateup todistancesof60mmthroughspecificlong-rangeconnections,aspredictedbyCCEPs.Giventheknownimpact ofslowwavesoninformationprocessingandcorticalplasticity,demonstratingtheirintrusionandpercolation withintheawakebrainaddkeyelementstoourunderstandingofnetworkdysfunctionaftercorticalinjuries.
1. Introduction
Focal cortical injuries are thought to disrupt neuronal activity across large-scale networks,extending wellbeyond thearea of neu- ronalloss (Falcon et al., 2016). The occurrence of functional alter- ationsin brainstructuresthat arenotdirectly affected bystructural damage, named‘diaschisis’ by Von Monakow (Von Monakov 1969; Feeney andBaron 1986) in 1914,has nowbecomea subject of ac- tiveempiricalinvestigation(CarreraandTononi2014; Fornitoetal., 2015;Baldassarreetal.,2016).Sofar,theneuralunderpinningsofdi- aschisishave beenmainlystudiedwithPET(Levasseur etal.,1992) andfMRI, showing hypo-perfusion (Yamakami andMcIntosh 1991), hypo-metabolism (Carmichael et al., 2004; Vagnozzi et al., 2010)
∗Correspondingauthorat:DepartmentofBiomedicalandClinicalSciences“L.Sacco”,UniversityofMilan,Milan,Italy.
E-mailaddress:marcello.massimini@unimi.it(M.Massimini).
1 Theseauthorsequallycontributedtothepresentwork.
(Carmichaeletal.,2004;Vagnozzietal.,2010),andalteredfunctional connectivity(Priceetal.,2001;Carteretal.,2012)extendingfarbeyond thesiteofinjury(Siegeletal.,2016).Theseneuroimagingpatternsof functionalnetworkdisruptionareclinicallyrelevantastheycanexplain behaviouraldeficitsandtheirrecovery(Saengeretal.,2018),however theirneuronalunderpinningsarestillelusive.
Electrophysiological evidence derived from animal models (Gloor et al., 1977; Leemburg et al., 2018) and non-invasive clini- cal recordings in stroke and traumatic braininjured subjects report a relativeslowingof electroencephalographic(EEG)andmagnetoen- cephalographic (MEG)rhythmsthatis prominentin areas ipsilateral to the lesion (Nuwer et al., 1987a; Butz et al., 2004), may ex- tend to the contralateral hemisphere (Buchkremer-Ratzmann et al.,
https://doi.org/10.1016/j.neuroimage.2021.117964
Received6November2020;Receivedinrevisedform15February2021;Accepted8March2021 Availableonline23March2021
1053-8119/© 2021TheAuthor(s).PublishedbyElsevierInc.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense
1996; Rorden andKarnath 2004),and persist in the chronic phase (Poryazovaetal.,2015;Cassidyetal.,2020).Furthermore,recentwork employing direct cortical perturbations with transcranial magnetic stimulation (TMS) revealed the occurrence of slow EEG responses in the perilesionalareaassociated with disruption of local informa- tion processing (Sarasso et al., 2020) and behavioural impairment (Tscherpeletal.,2020).Overall,thesemacroscopicrecordingssuggest theinterestingpossibilitythattheintrusionoflowfrequencyneuronal activityinanatomicallysparedcorticaltissuerepresentsanimportant elementoffunctionalnetworkdisruptionafteracquiredbraininjury.
Testingthishypothesiswouldrequirerevealingtheprecisenature ofpost-lesionelectrophysiologicalalterationswithrespecttopre-lesion activity,theirlocalspatialextentand,mostimportant,theirpotential forpropagatingwithinlarge-scalenetworks.Sofar,however,suchsys- tematicexplorationhasbeenhamperedbyfundamentalchallengessuch as(i)thespuriousandvariablenature(e.g.ischaemic,haemorrhagic, andtraumatic)ofaccidentalbrainlesions,(ii)thelowspatialresolution ofnon-invasiveelectrophysiologicalrecordingtechniquesand(iii)the lackofbaselinerecordingsacquiredbeforetheinjurywithinthesame subject.
In the present study, we overcome these issues by adopting an approachwherebytheeffectsof controlledlesionsareassessedwith spatially resolved electrophysiological recordings and compared to prelesional measurements within the same individual. To this end, wecontrastedspontaneousrestStereo-Electroencephalography(SEEG) recordedin21subjectsbothbeforeandaftertheinductionofsmallfo- calcorticallesionsbySEEG-guidedRadiofrequency-Thermocoagulation (RFTC) performed as a therapeutic option for drug-resistant focal epilepsy(Cossuetal., 2015a;Bourdillon etal.,2017;Dimova etal., 2017).BaselinemeasurementsincludedrestingactivityduringNREM sleepand during wakefulness aswell ascortico-cortical evoked po- tentials elicited by single-pulse electrical stimulation to assess indi- vidual patterns of long-range connectivity (Matsumoto et al., 2017; Trebauletal.,2018).Wefoundthatpost-lesionintracerebralactivity duringwakefulness was characterizedby slow waves closelymatch- ingthoserecordedduringbaselineNREMsleepinthesameindividu- als.Crucially,thesesleep-likeslowwaveswereprominentintheper- ilesional areas but could also percolate through a network of con- nectedareasaspredictedbyindividualpatternsoflong-rangeeffective connectivity.
Thepresentresultssuggestthat thegenerationof focalsleep-like slowwavesandtheirlong-rangepropagationwithintheawakebrain mayrepresentakeyelectrophysiologicalcomponentofdiaschisisand, moreingeneral,acrucialelementforunderstandingthefunctionalcon- sequencesoffocalandmulti-focalinjury.
2. Materialsandmethods
2.1. Subjectsandimplantationprocedures
Datawerecollectedduringthepresurgicalevaluationof21subjects withahistoryof drug-resistant,focalepilepsy(Supplementary Table 1).SEEGwasindicatedwhenatailorednon-invasivepresurgicaleval- uationfailedtoclearlydefinethesubject’sepileptogeniczone(EZ).In- tracerebralrecordingswereperformedtodefinethecerebralstructures involvedintheonsetandpropagationofseizures,basedontheresults ofnon-invasiveevaluations.Theplacementofintracerebralelectrodes was determined solelybythe clinical necessityforsearching for EZ (Cossu etal.,2015a; Cardinaleetal.,2019).Eachsubjectunderwent brainMRI(Achieva1.5T,PhilipsHealthcare)andCT(O-arm1000sys- tem[Medtronic])toacquireappropriatesequencesforSEEGplanning.
Theinvestigatedhemisphere/s(9Left;6Right;6Bilateral),theloca- tionandthenumberof exploredsitesweredeterminedbasedonthe non-invasiveanatomo-electro-clinicalassessment.ThedurationofSEEG investigationwasbasedonlyonclinicalneeds.Placementofintracere- bralelectrodeswasperformedundergeneralanaesthesiabymeansof
arobotizedpassivetool-holder(Neuromate,RenishawMayfieldSA).A variablenumberofplatinum–iridiumsemiflexiblemulti-contactintrac- erebralelectrodes,withadiameterof0.8mm,acontactlengthof2mm, aninter-contactdistanceof1.5mmandamaximumof18contactsper electrode(Microdeepintracerebralelectrodes,D08[DixiMedical],or DepthElectrodesRange2069[Alcis])wereplacedandfixed.Afterim- plantation,afinecone-beamCTdatasetwasacquiredbyusingtheO- armandcoregisteredwiththeT1-weighted3DMRimagetoverifythe actualpositionoftheelectrodes(Cardinaleetal.,2019).
2.2. LocalisationofSEEGcontacts
Thepreciselocalisationofeachcontactwasobtainedbyemploying theSEEGAtool(Narizzanoetal.,2017)andcheckedbytrainedneuro- physiologists.Anycontactshowinganinaccuratepositionwascorrectly placedatthecentroidofitscorrespondingCTartefact.Theanatomical areaofeachcontactwasextractedfromFreesurfer’s(Fischl2012)par- cellationusingtheDesikan/Killianyatlas.Thepositionofeachcontact wasnormalisedtotheMNI152usingtheANTstoolboxafterbrainex- traction(Avantsetal.,2011)andplottedona3Drendering(Fig.1A)and onaflatmap(Fig.2A,SupplementaryFig.1)usingpycortex(Gaoetal., 2015).Thedistancebetweeneachrecordingcontactandthelesionin- ducedbyRFTCwasoperationalizedastheminimumEuclideandistance fromtherecordingsiteandeachRFTCsite.
2.3. Dataacquisitionandsinglepulseelectricalstimulation(SPES)
Duringinvasivediagnosticevaluation,21subjectsunderwentindi- vidualinvestigationwithspontaneousactivityrecordingsaloneorcom- binedwithsinglepulseelectricalstimulation(SPES)duringeyesopen restingwakefulness.Inaddition,in19ofthe21subjects’spontaneous recordings werecollectedduringsleepbeforeRFTC(sleeprecordings werenotavailableforsubjects6and13).SEEGsignalswererecorded usinga192-channelrecordingsystem(Nihon-KohdenNeurofax-1200) withasamplingrateof1000Hz.SEEGdatawererecordedandexported inEEGNihon-Kohdenformat.Recordingswerereferencedtoacontact locatedentirelyinwhitematter.Besidesrecordingsofspontaneousac- tivitytodetecttheoriginoftheictaldischarge,singlepulseelectrical stimulation(SPES)wasperformedtoidentifyeloquentareasandeffec- tivenetworksconnectedwiththeepileptogeniczone(Valentínetal., 2002;Kelleretal.,2011;Matsumotoetal.,2017).SPESwasdelivered througheachpairofadjacentcontacts,withat3–5mAcurrentinten- sity,singlepulseof0.5ms(biphasicrectangularstimuliofalternating polarity),at1Hzfrequency,for15s.Inallsubjects,87.26%oftheim- plantedcontactswerestimulated,thusobtainingatotalamountof2864 stimulationsessions(average±std;143±21persubject).Stimulation, recordinganddatatreatmentprocedureswereapprovedbythelocal EthicalCommitteeandinlinewithGDPR(ID939–2.12.2013,Milano AREACNiguardaHospital,Milan,Italy).Allsubjectsprovidedwritten informedconsent.
2.4. RadiofrequencyThermocoagulation(RFTC)
RadiofrequencyThermocoagulation(RFTC)isasurgicaltechnique thatcanbeexecutedwithoutanaesthesia/sedationattheendofSEEG investigationwiththeaimtoreducethenumberandintensityofseizures beforehavingtoresorttomoreinvasivesurgery(Guénotetal.,2004; Cossu etal., 2015a;Schollyetal.,2019).RFTCofeachtargetsiteis performedbetweenapairofcontiguouscontactspertainingtothesame electrode(Fig.1A)andselectedaccordingtooneofthecriteriadefined inCossuetal.(2015b),thatare:(1)intralesionallocation(seeSupple- mentaryTable1),(2)habitualictalclinicalphenomenainducedwhen electricalstimulationswereperformedonthosecontiguouscontacts,or (3)thecontactsarepartofthenetworkinvolvedintheictaldischarge.
Inallcases,theymustbesilentatfunctionalmapping(e.g.,movement, speech,vision)anddistant>2mmfromvascularstructures(Cossuetal.,
Fig.1.Recordingsfromarepresentativesubject.PanelA:T1MRI(axialsectionofthelefthemisphere)and3Dsurfacereconstruction(oftherighthemisphere) fromarepresentativesubjectshowingoneRFTCsite(blackovoid,T’6–7,leftSuperiorTemporalGyrus)andfourbipolarrecordingcontacts(whitepoints:T’1–2and T’2–3intheleftInsula,T’3–4andT’4–5inleftSuperiorTemporalGyrus)fromthesameelectrodeandprogressivelydistancingfromtheRFTC.PanelB:Spontaneous activityrecordedbeforeRFTC(inblack)andafterRFTC(ingrey)fromeachofthebipolarcontactshowninPanelA.PanelC:forthesamecontactsandconditions (samecolour-coding),highfrequencynormalizedPowerSpectralDensityisshown.Thegreyshaded,verticalbandhighlightsthefrequencybandselectedfordelta (0.5-4Hz)quantifications.PanelD:forthesamerepresentativecontacts,percentagedeltavariationofpost-RFTCpowerwithrespecttopre-RFTCpowerisshown.
Fig.2. Groupleveldeltapowerincrease.PanelA:SpatialdistributionofallSEEGcontactslocations(blackdots)fromallsubjects,depictedonatemplateflatmap (MNI152).PanelB:relationshipbetweenspectralvariationsinthedelta(0.5–4Hz)poweranddistancefromtheclosestRFTCsites(Power-DistanceRelationship, PDR)performedatthegrouplevel.EachdotcorrespondstoacontactdepictedinPanela.Dashedverticallineistheuppervalueoftheoverlappingareaidentified byhierarchicalclustering(28.74mm,seeresults).Greysolidlineindicatesexponentialdecayfunctionfit.
2015a).Thetwoselectedcontiguouscontactsarethenconnectedtoan RFlesion-generatorequipment(NeuroN50andNeuroN100withDixi MedicalandAlciselectrodes,respectively)andRFTCisperformedbyin- creasingelectricalpowerprogressivelyfrom1.5to8.32Wwithin60s or10s(dependingon anatomicalcaveats– e.g.closenesstovessels) withinalltheselectedepileptogenicsites.Foreachsite,thisprocedure producesanellipsoidallesionwithamajorradiusof3mmandaminor radiusof1.5mm,whosefocicorrespondtothetwocontactsinwhich thecurrentisdelivered(Bourdillonetal.,2016).Beforeandafterthe thermocoagulationprocedure,werecordedonaverage12min(3to23) ofSEEGspontaneousactivityduringwakefulness.
2.5. Pre-processingofspontaneousrecordingsandsleepscoring
Datawereimportedfrom EEGNihon Kohdenformat intoMatlab (TheMathWorksInc.)andconvertedusingacustomizedMatlab-based script.Dataunderwentlineardetrendingandhigh-passfiltering(0.5Hz, third-orderButterworthfilter,zero-phaseshift).Bipolarmontageswere
calculatedbysubtractingthesignalsfromadjacentcontactsofthesame depth-electrode(166±18bipolarcontactspersubject)tominimisevol- umeconductionandtomaximisespatialresolution (Lietal., 2018).
Datawerevisuallyinspectedbytrainedneurophysiologists,andcon- tactsexhibitingsustainedartefactualactivityorcontinuousepileptiform SEEG activitywereexcluded from furtheranalysistoavoid interfer- enceofnon-biologicalandepileptiformdischargesonthesubsequent quantifications(seeControlsfortheroleofepilepticactivity).Inaddi- tion,contactsselectedfordeliveringRFTCwereexcludedfromfurther analyses(yellowcontactsinSupplementaryFig.1).Theremainingcon- tacts(138±27 persubject – percentage:82.90± 11.18%)usedfor thephysiologicalinvestigationunderwentfurthervisualinspectionin ordertomarkandremoveelectricalartefactsandpossible,rareparox- ysmalwaveforms(spikes,spike-wave,sharp-waves,slowepilepticde- flectionsandparoxysmalactivity,SupplementaryFig.2A).Thesame pre-processingwasappliedonbothwakefulnessandsleeprecordings.
Onthelatter,off-linesleepscoringwasappliedusingone scalpEEG derivation,together withonebipolarelectrooculographic(EOG) and
oneelectromyographic(EMG)derivation toconfirmthat allsessions wererecordedduringNREMN3(Silberetal.,2007).
2.6. Spectralanalysis
Foreachsubjectandcontact,wequantifiedthepowerspectralden- sity(PSD,Fig.1C)oftheSEEGactivityrecordedbeforeandafterRFTC byaveragingthePSDofallartefact-freeepochs(5slength,pre-RFTC numberofepochs=115±50,post-RFTCnumberofepochs=96±56).
EachPSDfrequencybinwasnormalisedfortheaverageacrosshighfre- quencies(30–40Hz)(Nelsonetal.,2013)tominimisetheeffectsoflocal tissueresistancemodificationspossiblyinducedbyRFTC(Gaetz2004) – seealsoSupplementaryFig.3.Toquantifydeltapower,weaveraged PSDbinsbetween0.5and4Hz.Thenwecomputedthepercentagevari- ationofdeltapowerduringpost-withrespecttopre-RFTC(i.e.[post- RFTC– pre-RFTC]/[pre-RFTC]∗100),asshownforexampleinFig.1D.
Weappliedthesameprocedurestocalculatethepercentagevariationof deltapowerduringNREMsleepwithrespecttopre-RFTCactivity.The relationshipbetweendeltapowervariationandthedistancefromthe RFTCsiteswasrepresentedbyascatterplot,defined Power-Distance Relationship(PDR),whichcouldbefittedwithanegativeexponential fit(Fig.2,SupplementaryFig.4).Finally,basedonthePDR,weop- erationallydefinedtheextentoftheperilesionalarea(E-perilesion)by applyingahierarchicalclusteringalgorithm(Fig.2andSupplementary Fig.5,MatlabStatisticandmachinelearningtoolboxwithautomatic numberofclustersselectionusingtheEuclideanandWarddistances).
Thisanalysisempiricallyidentifiedtwoclusters(SupplementaryFig.5) clearlyseparatedonthexaxis(i.e.distancefromRFTC)at28.74mm.
Foreachsubject,singlecontactspertainingtoeachclusteraredepicted inSupplementaryFig.1.
2.7. Pre-processingofcortico-corticalpotentialsevokedbySPESand effectiveconnectivityevaluation
DuringthepresurgicalevaluationandbeforetheRFTCprocedure, anextensiveeffectiveconnectivityevaluationwasperformedforeach and every subject by means of Single Pulse Electrical Stimulation (SPES)andCortico-CorticalEvokedPotentials(Matsumotoetal.,2017; Trebauletal.,2018)– seeforexampleFig.3.CCEPswerepre-processed usingtheautomaticpipelinedescribedbelow.Theelectricalstimulation artefactwasremovedbyapplyingaTukey-windowedmedianfilter.Sig- nalswerethenfiltered(0.5Hzhigh-pass3rdorderButterworthfilter) andsingletrialsweresplitbasedontheinterstimulusinterval(−330ms, +666ms).Eachtrialwasbaseline-corrected,consideringbaselinefrom
−300 ms to−20 ms before the stimulation, toavoid possible stim- ulation artefact residuals.SPES delivered with alternatemonophasic pulseswereseparatedinpositiveandnegativeandanalysedindepen- dently.Automatictrialsrejectiontoexcludeepileptiformabnormalities andelectricalartefactsfromfurtheranalysiswasperformedasshown inSupplementaryFig.6,followingthesesteps:(1)foreachtimesam- pleandsingletrial, theaverageof allthetrialswas subtractedfrom each trial;(2) todefinethepre-stimulusactivity,a nulldistribution wasobtainedbymergingthebaseline(−300msto−20ms)ofalltri- als(Usamietal.,2015);(3)trialsexceedingthenulldistributionwere identifiedaccordingtoChebyshev’sinequalityand(4)ifthemaximum voltagevalueacrosstrialsexceeded7𝜎ofthenulldistribution(corre- spondingto𝛼=0.05),thecorrespondingtrialwasrejected;(5)thispro- cesswasperformediteratively,excludingateachsteptherejectedtrials (alsofromthenulldistribution)untilthemaximumvaluedidn’texceed theChebyshev’sinequalitythreshold(LeVanQuyenetal.2001).After thisautomatictrialrejection,wequantifiedthepowerofthesignificant (>3𝜎ofthebaseline)CCEPs(Matsumotoetal.,2017;Trebauletal., 2018)elicitedbySPESofRFTCcontacts.Inordertoassesstheoverall strengthoftheeffectiveconnectivitybetween eachrecordingcontact andtheRFCTsiteswecomputedtheaverageofthepowerofCCEPs fromalltheRFTCsites.
2.8. Slowwavedetectionandhigh-frequencymodulation
Wethenassessedthesimilarities betweenpost-RFTCwakefulness andnaturalsleepinthe19subjectswithsleeprecordings.Wefirstiden- tifiedthosecontactsshowingapost-RFTCdeltapowerincreasecom- parabletotherelativedeltapowerofNREMsleepwithrespecttopre- RFCT(bluepointsinFig.4B)– i.e.withintheinter-quartilerangeof theNREMsleepdeltapowerdistribution(Fig.4A,rightdistribution).
Next,weappliedastandardslow-wavedetectionalgorithmbasedon zero-crossingsandperiodcriteria(Riedneretal.,2007a)onthesame contacts.TothisendtheanalysisemployedinRiedneretal.wasmod- ifiedtoaccommodatethefactthatthetypicalnegativepolarityofthe slowwavesrecordedatthescalpEEGlevel,couldbeeitherpositiveor negativeinthecaseofintracerebralbipolarmontage(Valderramaetal., 2012).Then,weusedthemedianamplitudeoftheeventsidentifieddur- ingNREMtosetanamplitudethresholdfordetectingtheoccurrenceof slowwaves inbothpost-RFTCandpre-RFTCwakefulnessrecordings.
Tofurtherconfirmthatthedetectedeventswereassociatedwithneu- ronaleventssimilartothoseofsleep-likeslowwaves,weperformed time-frequencyanalysistoassessthepresenceof significantsuppres- sionsofhighfrequencyactivity.Asrevealedbypreviousstudiesinan- imalandhumans,thesefrequencymodulationsoftheintracranialEEG signalareareliableextracellularproxyoftheperiodsof neuronalsi- lence(orOFF-periods)characterizingthecorticalslowoscillationdur- ingsleep(Mukovskietal.,2007;Cashetal.,2009;Csercsaetal.,2010).
Todemonstratetheirco-occurrencewithslowwaveswefirstperformed a time-frequency analysis at the single contactlevel in both NREM sleepandpost-RFTCconditions(Fig.4E-F,left),andthenweperformed a wave-triggeredaverage of the envelope of gammaactivity(above 60Hz),basedontheslowwavepreviouslydetected(Rayetal.,2008).
2.9. Statisticalanalysis
Group-levelstatisticalanalyseswereperformedinRusingamulti- levelapproach,duetotheclusterednatureofthedata– i.e.channels withinsubjects(Aartsetal.,2014).Mixedeffectsmodelswereemployed totest(i)theexponentialdecay(self-startingnon-linearasymptoticre- gression),(ii)theperilesionalnormalizeddeltapowerincrease(linear mixed effectsmodel),(iii)therelationshipbetweennormalizeddelta power,connectivity,anddistance(mixed-effectslinearregression),and (vi) thenumber of slowwaves across conditions (generalized linear mixed effects).Instead,themodulationofgammaactivitywithindif- ferentconditionswastestedusingahierarchicalbootstraptestasthe mixedeffectsapproachfailedtoconverge.Adetaileddescriptionofthe statisticalanalysesisavailableintheSupplementaryMaterials.
2.10. Dataavailability
Alldataneededtoevaluatethepaper’sconclusionsarepresentin thepaper.Additionaldatarelatedtothispapermayberequestedfrom thecorrespondingauthor.Rawdataareavailableonrequest.
3. Results
3.1. SEEGmeasurementsperformedbeforeandafterfocalbrainlesions inducedbyRFTC
The21epilepticpatientsincludedinthepresentworkunderwent RFTCaspreliminarytherapeutictreatment.RFTCinducesafocallesion (Fig.1A)atspecificcorticalsitesidentifiedasepileptogenicorpresum- ablyinvolvedintheepileptogenicnetwork(seemethodsfordetails).All contactsconsideredfreefromepilepticactivity(seemethods)wereused tocompareSEEGactivitybeforeandafterRFTC(Fig.1B).Inaddition, severalanalyses wereperformedtocontrolforpossibleinfluences of epilepticactivityindeterminingtheSEEGchangesobservedafterRFTC (seeControlsfortheroleofepilepticactivityandSupplementaryFig.
Fig. 3. Relationship between delta increase andconnectivitywithRFTCareas.PanelA: T1MRI(axialandcoronalsectionsoftheleft hemisphere)and3Dsurfacereconstruction(of the right hemisphere) froma representative subject.Dashedgreenarrowsshowthedistance fromoneRFTCsite(blackovoid,T’6–7,leftSu- periorTemporalgyrus)tofourbipolarrecord- ingcontacts(whitepoints:B’12–13intheleft MiddleTemporalgyrus;U’9–10intheleftSu- periorTemporalgyrus;X’4–5intheleftMedial Orbitofrontalcortex;H’10–11intheleftRos- tralMiddleFrontalgyrus).Thefourrepresen- tativecontactsarealllocatedinthelefthemi- sphereand3Dreconstructionoftherighthemi- sphereisshowasareference.Shadedbluearea indicatesE-perilesion.PanelB:Forthesame contactsdepictedinPanela,thecorrespond- ingCCEPsareshownsortedbydistance(from toptobottom). PanelC: forthe same four contactsreportedinPanelb,thePSDofpre- RFTC(black) andpost-RFTC (grey)activity.
PanelD:ontheleft,relationshipbetweendelta power(0.5–4Hz)variationsanddistancefrom thenearestRFTCsiteforeachoftherecord- ingsitespertainingtoE-perilesion.Greenline depictsthelinearfitusingamixedeffectlin- earregression.Ontheright,relationshipbe- tweendeltapower(0.5–4Hz)variationsaver- agemagnitudeof CCEPscollectedfromeach oftherecordingsiteswithinE-perilesionwhen stimulatingRFTCsites.Redlinedepictsthelin- earfitusingamixedeffectlinearregression.
PanelE:sameasPaneldbutforcontactfar fromRFTCsite(i.e.outofE-perilesion).(For interpretationofthereferencestocolourinthis figurelegend,thereaderisreferredtotheweb versionofthisarticle.)
2).AfterRFTC,thefastfrequencyactivitytypicalofwakefulnesswasre- placedbyslowSEEGactivitythatwasmostprominentclosetothether- mocoagulatedareaandgraduallydiminishedwithdistance(Fig.1B).
Toquantifythesechanges,whichcouldbealreadyappreciatedbyvi- sualinspection,wecalculatedthePSD(Fig.1C)ofSEEGactivityonall artefact-freeepochs(seemethodsforadescriptionofthepre-processing andmanualdatacleaningprocedures)collectedbothbeforeandafter theRFTCprocedure(averagedurationofrecordings:12.97±4.69min and12.95±10.65min,respectively).Computingthepercentagevari- ationofdelta(0.5–4Hz)powerofpost-RFTCwithrespecttopre-RFTC activityconfirmedagradientdecreasingwithdistancefromtheRFTC site(Fig.1D).
3.2. ElectrophysiologicaleffectsofRFTCextendbeyondtheareaof structuralinjury
Thepresenceofaspatialgradientofdeltapoweraroundthelesion wasfurther characterizedbya regressionanalysis(non-linearmixed
model)basedonallartefact-free contactsfromallsubjects(Fig.2A).
Specifically, thepercentagevariationofnormaliseddelta powerwas representedasafunctionofdistance fromtheclosestRFTCsiteina scatterplotthatwehenceforthcallPower-DistanceRelationship(PDR).
Here,wefoundthatPDRcouldbefitwithanexponentialdecayfunction atthegrouplevel(Fig.2B,SupplementaryTable2)andbyanegative exponentialregressionatthesinglesubjectlevel(SupplementaryFig.
4).Althoughtheincreaseinnormalizeddeltapowerwasnotlimitedto thecontactsclosetotheRFTCandwasalsovisibleatdistantrecord- ingsites(upto6cm,seePDRinFig.2),wefirstaimedatdefiningthe extentoftheperilesionalareacharacterisedbymaximaldeltaincrease.
Thisboundarywasoperationallyidentifiedbyemployinghierarchical cluster analysis.Thisblind,data-drivenprocedure alloweddemarcat- ingaclusterofcontactswithinavolumewitharadiusof28.74mm from theRFTCsites,(dashedverticallineinFig.2B)henceforthde- finedastheelectrophysiologicalperilesionalarea(E-perilesion).Inthis area,thedifferencebetweenpre-andpost-RFTCdeltapowerwasstatis- ticallysignificantasassessedbyalinearmixedeffectsmodel(𝛽=0.45,
Fig.4.Sleep-likefeaturesofslowwavesde- tectedpost-RFTC.PanelA:Ontheleft,group level PDR obtainedby NREM sleep record- ings.BlackXrepresents,foreachsinglecon- tact,therelativedeltapowerofNREMsleep withrespectto pre-RFTC.On the right,the correspondingprobabilitydensity withinter- quartile-range(IQR)highlightedinblue.Panel B:GrouplevelPDR.Bluedotsindicate con- tactsshowinga deltapowerincreasewithin theIQRofNREMsleep.Greydotsareallthe others. PanelC:PSD of SEEGactivity from onerepresentativecontactduringNREMsleep (black),pre-RFTC(cyantrace)andpost-RFTC (bluetrace).PanelD:ontheleft,representa- tivespontaneousactivity(5s)recordeddur- ingNREMsleep(blacktrace),pre-RFTC(cyan trace) and post-RFTC (blue trace) in wake- fulness.Asterisks indicate the detected slow waves. On the right, comparison of the to- talnumberofoscillations perminute(inthe deltarange)detectedintheSEEGactivitydur- ingwakefulnessbefore RFTC(cyan boxplot) andafterRFTC(blue)andduringNREMsleep (black boxplot) is shown. Asterisks indicate significance values as reportedin Table S5.
PanelE:ontheleft,average(solidblackline, top panel) and time-frequency power spec- trum (bottom panel) of the slowwaves de- tected during NREM sleep from one repre- sentativecontact;ontheright,median(solid line,IQRshaded)ofalltheslowwaves(top panel)detectedfromallsubjectsduringNREM sleepandthecorrespondingmedian(solidline, IQRshaded)ofthegammaactivitymodulation (bottompanel).PanelF:sameasPanelEbut forpost-RFTCcondition.(Forinterpretationof thereferencestocolourinthisfigurelegend, thereaderisreferredtothewebversionofthis article.)
p<0.001,R2c=0.32;seeSupplementaryTable3foracompletereport ofthemodel’sestimatesandstatisticalresults).Notably,theE-perilesion (radius28.74mm)identifiedbydata-drivenclusteringexceededbyal- most10times thesizeofthetypicalstructurallesion(anovoidwith radius~ 3mm)describedbypreviousstudiesemployingRFTCwith thesameparameters usedin the presentwork(Cossuet al., 2015a; Bourdillonetal.,2016,2017).
3.3. Whole-braindeltapowerincreasecorrelateswitheffectiveconnectivity fromRFTCsites
WhiletheclusteranalysisallowedoperationalizingE-perilesionasa volumeofstrongelectrophysiologicalalterationsextendingaroundthe RFTCstructurallesion,thechangesindeltapowercouldoftendeviate
from thefitofthepower-distancerelationship. Indeed,asevidentin Fig.2B,anumberofremotecontacts(locatedupto6cmawayfromthe centroidofthelesion)showedamarkedincreaseindeltapower.Wethus testedthehypothesisthatthisincreasemayreflectpreferentialconnec- tionsfromtheRFTCregion,aboveandbeyonddistanceperse(Fig.3A).
Tothisend,weretrospectivelypre-processedandanalysedthepre-RFTC Cortico-Cortical EvokedPotentials(CCEPs)obtained from20 subject (CCEPsofsubjectn.11notavailable)bymeansofSingle-PulseElectrical Stimulation(SPES;Fig.3B)performedateachRFTCsite(Supplementary Fig.6).Briefly,thepoweroftheevokedresponseateachrecordingsite wasaveragedacrossRFTCsitestoobtainthemeanconnectivityvalue with respecttotheensemble of allthermocoagulatedareas foreach recordingcontact(seemethodsfordetails).Therelationshipbetween distancefromRFTCsite,connectivity,anddeltapowerincreaseisexem-
plifiedinFig.3.Throughthisanalysis,wefoundthat,whilewithinthe E-perilesionmostcontactsfeaturedprominentincreasesindeltapower thatcorrelatedwithbothdistanceandCCEPsmagnitude,thiswasnot thecaseforcontactsfarfromtheRFTCsite;outoftheE-perilesion,the increaseindeltapowerdependedonthemagnitudeofCCEPs,rather thanongeometricdistanceperse.Theseobservationswerequantified atthegrouplevelbyemploying,foreachcluster(i.e.E-perilesionand far),amixedeffectslinearregressionanalysisusingdeltapowerincrease asdependantvariablesanddistanceandconnectivityaspredictors(see methods).Thisanalysisshowedthat,whilebothconnectivityanddis- tanceweresignificantpredictorsofdeltapowerincreasewithintheE- perilesion(𝛽 connectivity=2.04,p<0.05;𝛽 distance =−6.16,p<
0.001;R2m=0.20,R2c=0.38;Fig.3D,TableS4,SupplementaryFig.7), onlyconnectivitymaintaineditspredictivepowerfarfromtheRFTCsite (𝛽connectivity=2.25,p<0.01;𝛽distance=−1.06,p=0.06;R2m= 0.11,R2c=0.52;Fig.3E,TableS5,SupplementaryFig.8).Thisiscon- sistentwiththegeneralnotionthattheconnectomeischaracterisedby preferentialpatternsoflong–rangeconnectivityandwiththeviewof diaschisisasanetwork-specificphenomenon(Fornitoetal.,2015).
3.4. Deltapowerincreaseisassociatedwithsleep-likeslowwavesand OFF-periods
Wefinallyaskedwhethertheobservedsignificantincreaseindelta power reflected the occurrence of local sleep-like dynamics during wakefulness.Totestthishypothesis,wecompared theSEEGactivity recordedbothpre-andpost-RFTCwiththatcollectedduringperiodsof sleeprecordedinthesamesubjectsandcontacts.Thecomparisonbe- tweenpost-RFTCandNREMsleepdeltapowerrevealedthepresenceof contacts(n=250,~10%)inwhichpost-RFTCwakefulnessdeltapower increasedtolevelscomparabletothose of NREMsleep– i.e. within theInter-QuartileRange(IQR)of thedeltapowerincrease ofNREM sleep(Fig.4A-C – seemethodsfordetails).Wethenaskedwhether, forthesecontacts,post-RFTCdeltapowerincreasereflectedtheoccur- renceofslowwaves(Fig.4D),accordingtocriteriadefinedbyprevious sleepstudies(Mukovskietal.,2007;Cashetal.,2009).First,weap- pliedaperiod-amplitudedetectionalgorithm(seeMethods)toidentify sleepslowwavesinNREMsleeprecordings.Then,weusedthemedian amplitudeofthese eventstosetathresholdfortheidentification of sleep-likeslow-wavesinbothpost-andpre-RFTCwakefulnessrecord- ings.Asshown inFig.4D,wefoundthattheslow-wavedensity(os- cillationsper minute)in post-RFTCcondition(13.5± 13.5)wassig- nificantlyhigherthanpre-RFTCcondition(7.58±8.46;p<0.001)al- beitlowerthanNREMsleep(23.2±2.89;p<0.01),asconfirmedbya zero-inflatedPoissongeneralizedlinearmixedeffectsmodel(R2m=0.32, R2c=0.72;seeSupplementaryTable6).
Finally,weassessedwhetherthepost-RFTCsignificantincreasein bothdeltapowerandnumberofslowwaveswasassociatedwiththe occurrenceofneuronalOFF-periods,thehallmarkofsleepslowwaves.
Awave-triggeredaverageanalysison theslowwavesdetectedin thepost-RFTCperilesionalcontactsrevealedasignificant(p<0.001,hi- erarchicalbootstraptest,SupplementaryTable7)modulationofhigh- frequency(>60Hz)activity(Steriadeetal.,1993;AmzicaandSteri- ade1998;Mukovskietal.,2007;Cashetal.,2009)similartooneob- servedduringNREMsleep(p<0.001,hierarchicalbootstraptest,Table S8),suggestingtheoccurrenceofneuronalOFF-periods,associatedwith post-RFTCslowwaves(Fig.4F,seeMethodsfordetails)(Mukovskietal., 2007;Cashetal.,2009).
3.5. Controlsfortheroleofepilepticactivity
Weappliedmanualrejectionofallepochsandcontactscontaining pathologicalactivity– suchasspikes,spike-and-waves,paroxysmaldis- charges,andepilepticslowdeflections.(seeSupplementaryFig.2A)-in orderminimizetheimpactofSEEGabnormalitiesonourobservations.
Inadditiontothisstandardprocedure,weperformedasetofspecific
controlstoverifythattheoccurrenceofslowwavesandtheirpropa- gationafterRFTCdidnotdependonongoingpathologicalactivitynor onpre-existingnetworkalterations.First,weverifiedthattheincrease indeltapowerdidnotcorrelatewiththefrequencyofepilepticactiv- itydetectedafterRFTCatthelevelofeachrecordingsite(Supplemen- taryFig.2B),thusexcludingaroleSEEGabnormalitiesindetermining theobservedspectralchanges.Second,weperformedthesameanalysis withrespecttothespatialdistributionofepilepticactivitydetectedbe- foreRFTCtocontrolforthepossibleinfluenceofpre-existingnetwork alterations(SupplementaryFig.2C).Third,weverifiedthatthedelta increasedidnotdependonwhetherRFTCwasperformedonacontact pertainingtotheepileptogeniczone(EZ,asdefinedbyspontaneousic- talrecordingsand/orclinicaloutcomes)oronacontactpertainingtoan early-propagationzone(EPZ;definedastheareaofearlypropagation ofictaldischarges,SupplementaryFig.2D)(Bartolomeietal.,2017).
Fourth,weverifiedthatourresultswerenotrelatedtothepresenceof specificunderlyingaetiology-suchasperiventricularnodularhetero- topia(Battagliaetal.,2006)andtemporallobeepilepsywithorwithout Hippocampussclerosis(Wengetal.,2020)-thatareknowntobeasso- ciatedwithalteredthalamocorticalconnectivity(seeMRIanddiagnosis inSupplementaryTable1).Inlinewiththis,sinceepilepticdisorders associatedtostructurallesionsareknowntoinducenetworkmodifi- cations,wealsoverifiedthatallourresultscouldbeconfirmedwhen includingonlyMRInegativesubjects(i.e.nostructurallesions)inthe statisticalanalyses.
Fifth,weverifiedthatcontactsshowingsleep-likeactivityafterRFTC didnotshow apriorihighnarrow-band3–4Hzpowerpossiblyindi- catingpathologicalactivity(SupplementaryFig.2E-F,linearmixedre- gressionandlinearmixedmodelanalysis).Sixth,alongthesameline wecomparedthepre-RFTCspectralprofilebetweencontactsshowing sleep-likedeltaincreaseafterRFTCandcontactsthatdidnotshowedan increaseandwefoundnodifferences(SupplementaryFig.2G,Wilcoxon ranksumtest,FDRcorrected).Finally,weverifiedthatthesleep-like eventsdetectedafterRFTCshowedasimilarmorphologytophysiolog- ical sleepslowwaves andthat bothshowed adifferent morphology with respect toepilepticslowdeflections (Nir,2011).Indeed, based on theknowledge thatsleep-slow-waves tend tobemore symmetric whileepilepticeventsaretypicallyasymmetric(Riedneretal.,2007b; Niretal.,2011)wecomparedtheslopesofthedetectedeventsacross conditions(SupplementaryFig.2H-J).Withthisanalysis,wefoundthat Post-RFTC eventsandsleepeventswerenot different betweenthem (p=0.83;p=0.72),butbothdifferedfromtheepilepticbenchmark (p<0.01).
4. Discussion
Byprovidingahigh-resolutionintracerebralassessmentof thelo- calandnetwork-levelconsequencesof controlledcorticallesions,the presentworkprovidesdirectevidenceforelectrophysiologicaldiaschisis inhumans.Asrevealedbythecomparisonbetweenpre-andpost-RFTC corticalactivity,brainlesionsinducedanincreaseintheamplitudeand number ofslow waves,whichwas unrelatedtoepilepticalterations.
Crucially,theseslowwavesintrudingwithinwakefulnesswerecompa- rabletothoseoccurringduringNREMsleep,wereprominentwithinthe perilesionalarea,andcouldpercolatetodistantsitesthroughspecific connectivitypatterns,aboveandbeyondtheanatomicaldistanceperse. 4.1. Generationofperilesionalsleep-likeslowwaves
Microinvasivesurgery,suchasRFTC,resultsinfocalbrainlesions thatarenotassociatedwithsignificantvascularalterationsandwhose localisationandextentcanbepreciselycontrolled(Cossuetal.,2015a).
Sofar,theeffectsofRFTChavebeenquantifiedbymeansofmorphologi- calobservations(Bourdillonetal.,2016;Garbellietal.,2016).Typically (Bourdillonetal.,2016),theextensionofthestructuraldamageinduced byaRFTCthroughapairofcontactsisonaverageanovoidwithradii
of1.5mmand3mm.Thefirstimportantfindingofthepresentstudy isasignificantincreaseofdeltapowerinanareasurroundingtheRFTC site.Thiseffect,whichsteeplydecayedfromthecentroidofthelesion (Fig.1,Fig.2andSupplementaryFig.1),identifiedaperilesionalarea extendingtoaradiusof28mm– anorderofmagnitudelargerthanthat ofthestructurallesion.
Various factors may explain this perilesional effect. First, local oedema,asobservedinanimalmodels(Clasenetal.,1958;Schauletal., 1976)canleadtoafailureofsodium-potassiumpumpsandionicun- balancesthatfacilitatetheexpressionofdeltaactivity(Rabilleretal., 2015).This mechanism plays a relevantrole in theacute phases of braindamage,butinthecaseofRFTCitshouldbelimitedtoathin layer(<1mm)ofbraintissuesurroundingtheRFTCarea,assuggested byinflammation-andoedema-relatedhistologicalsigns(Garbellietal., 2016).Asecond factorthatcanenabletheexpressionof slowwaves intheE-perilesionisthelackofinputfromascendingactivatingsys- temsduetodamageofwhitematterfibres(McCormicketal.,1991).
This mechanismhasbeenthoroughly documentedby animalexperi- ments(Nitaetal.,2007)inwhichthelesioninvolvestotheunderlying whitematter,suchasinthecaseofRFTC.Athird,importantmecha- nismleadingtotheexpressionofneuronalOFF-periodsandslowoscil- lationsisanunbalancebetweenexcitationandinhibition(Funketal., 2017),duetothedisruptionoflateralcortico-corticalexcitatorycon- nections(Welikyetal.,1995;Timofeev2000;Boucseinetal.,2011).In allcases,thepresenthigh-resolutionintracranialexploration,demon- stratingtheappearance ofsleep-like slowwaves inalarge areasur- roundingthelesion,explainstheearlymacroscaleobservationofEEG slowing(Nuweretal.,1987a;Butzetal.,2004;Poryazovaetal.,2015; Cassidyetal.,2020)aswellasrecentTMS-EEGevidenceofalteredre- activity(Sarasso etal.,2020; Tscherpeletal., 2020) inthelesioned hemisphereofstrokepatients.
4.2. Percolationofslowwavesthroughlarge-scalenetworks
Thereappearanceofslowwavesatdistantsiteswellbeyondtheper- ilesionalareacannotbeexplainedbylocalphenomenasuchasoedema ordirectdisruptionoflocalconnectionsbythelesion.Thepresentre- sultssuggestthattheseinstancescanbepredictedbyspecificpatterns oflong-rangeeffectiveconnectivityasassessedinindividualsubjects fromacausalperspectivebymeansofCCEPs(Matsumotoetal.,2017; Trebauletal.,2018).Indeed,whileclosetothelesionbothgeometrical distanceandconnectivityweresignificantpredictorsofdeltaincrease, witharelativehigherweightfordistance,outsideof E-perilesionef- fectiveconnectivityclearlyoutperformedgeometricaldistance(Fig.3).
Thisresultprovidesthefirstdirectelectrophysiologicalevidenceofa selectivenetwork-levelalterationofneuralactivityaftercorticallesion, inlinewiththeclassicdefinitionofdiaschisis.Hence,theintrusionand long-rangepercolationofsleep-likeslowwavesatspecificnodesofthe networkreportedhereprovidesdirectempiricalsupporttotheoriginal hypothesisofaremotefunctionaldisruptioninregionsconnectedtosite offocallesion(FeeneyandBaron1986)throughcortico-corticaland/or cortico-subcorticalconnections(McCormickandBal1997;Crunelliand Hughes2010).
Whetherdistantslowwavesoccurbecauseofthepropagationofac- tivityfromtheperilesionalareaorbecauseoflackofexcitatoryinput fromthelesionedsiteremains aninterestingopenquestion.Inprin- ciplebothmechanismsareplausible.Indeed,physiologicalsleepslow waveshavebeenshowntopropagatefromageneratortodistantsites throughexistinganatomicalconnections(Volgushevetal.,2006),po- tentiallyfollowingpreviouslyexistinganatomicalandfunctionalpath- ways(Zhangetal.,2018;Silversteinetal.,2020).Ontheotherhand, deafferentationcanleadtobothenhancedslowwavesandareduction ofhighfrequencyactivity(Timofeev2000).Whilethedeltapowerin- creaseoccurredalsointheabsolute spectrum(SupplementaryFigure 3),itwouldbeinterestingforfuturestudiestosystematicallyexplore
thepresenceofaconcomitantreductionofhighfrequencies,whichwas precludedherebythepresentnormalizationstrategy.
4.3. Implicationsoflocalsleepafterfocalbraininjury
Intracerebral studiesperformedin sleep-deprivedrodentsandhu- manshaveshownthatslowwavesandneuronalOFF-periodscanoccur locallyandatspecificcorticalsitesinthecontextofanawakebrain (Nobilietal.,2012)leadingtoselectivemotorimpairmentorcognitive lapsesdependingontheregionofthecortexinvolved(Vyazovskiyetal., 2011;Niretal.,2017).SubjectsundergoingRFTCcanbenefitfromthe procedureandrarelyshowneuropsychologicalorneurologicaldeficits, possibly due tothe fact that the lesions are small and deliberately performed in areas that are identified as functionally non-primary (Cossuetal.,2015a;Bourdillonetal.,2017).Withintheconstraintsim- posedbytheclinicalprocedure,itwillbeimportantforfuturestudies toassesssubjectsbehaviourallyinbothbaselineconditionsandpost- RFTCwithafine-grainedbatteryofmotor/cognitivetests,tailoredon thespecificbrainareasandnetworksinvolvedbytheRFTC.Thepresent demonstrationthatevensmalllesionscanleadtotheintrusionoflocal sleep-likeslowwavesandOFF-periodsatthenetworklevelbearsimpor- tantimplications.Forexample,itisknownthattheconsequencesofmi- croinfarctionsinhumansmayrangefromcognitivesparingtodementia butthereasonforthislargediscrepancyisunclear;aninterestingpossi- bilityisthattheinductionofpathologicalslowwavesinintactportions ofthecortexfollowingmicroinfarctionsmayhaveunpredictable,mag- nifyingeffectsatthenetworklevelinthebrainofsomesubjectswith multiplemicrovascularlesions.Allthemoreso,theelectrophysiological mechanismdescribedhereisexpectedtoplayamajorroleinsubjects sufferingfromlargerlesions,suchasstrokesubjects.Inthesepatients, TMS-EEGstudieshaverevealedsleep-likestereotypicalresponsesassoci- atedwithimpairmentofinformationprocessingintheperilesionalarea (Sarassoetal.,2020;Tscherpeletal.,2020)andEEGrecordingsoften detectslowingextendingtothecontralateralhemisphere(Nuweretal., 1987b).Linkingtheintrusionofsleep-likeeventsintheawakebrainto focalinjuryisveryimportant,giventhecomplexnatureoftheeffects thatslowwavescanhaveoncorticalcircuits.Forexample,theoccur- renceofslowwavesandtheassociatedsilentOFF-periodsareknown toaffectinformationprocessing,impairingcognitiveandmotorfunc- tions(Vyazovskiyetal.,2011;Niretal.,2017).Further,theycanin- terferewiththepropagationofhigher-frequencytravellingwaves,po- tentiallyimpairingtherelatedcognitiveabilities(Zhangetal.,2018).
Ontheotherhand,slowwavesandOFFperiodscanalsoexertbenefi- cial,protectiveeffects(Paceetal.,2015;Cassidyetal.,2020)andare knowntoplayaroleinshapingandremodellingcorticalconnections (Buzsáki1998;TononiandCirelli2014).Toaddalayerofcomplexity, thisremodellingeffectscanpositivelyornegativelyimpactthecognitive functionsdependingonwhatkindofslowwavesareoccurring;forex- ample,Kimandcolleagueshaverecentlyshownthatsingleslowwaves andsustaineddeltaactivityhavecompetingrolesinmemoryconsolida- tionprocesses(Kimetal.,2019).
Inthiscontext,thepresentfindingstriggeranumberofinteresting questions.Whatisthetruenatureofslowwaveactivityfollowingbrain injury?Areslowwavesjustadetrimentalsideeffectalteringnetwork activityaboveandbeyondthestructuraldamage?Aretheyinsteadthe pricetopayforenergysavingand/orcircuitsremodellingafterbrain injury?Oneinterestingpossibilityisthatslowwavesmightbeprotec- tiveandbeneficialintheacuterecoveryphaseandthenbecomedetri- mentaliftheypersistindefinitelyintothechronicphase.Importantly, acute andchronicslow wavesmaybeoriginated bydifferentpatho- geneticmechanisms;whileinflammatorycytokinesarelikelytoplaya roleintheacutephase(KruegerandMajde1995),chronicslowwaves aremorelikelytobe caused,amongstotherfactors,byanalteration of theexcitation/inhibitionbalance(TakeuchiandIzumi2012).The presentmesoscaleexplorationmayhelpaddressingthesequestionsby bridgingmicroscale animalexperimentstomacroscaleEEGandneu-