Pharmaco-fUS: Quantification of pharmacologically-induced dynamic changes in brain perfusion and connectivity by functional ultrasound imaging in awake mice

HAL Id: inserm-02971488

https://www.hal.inserm.fr/inserm-02971488

Submitted on 19 Oct 2020

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Pharmaco-fUS: Quantification of

pharmacologically-induced dynamic changes in brain

perfusion and connectivity by functional ultrasound

imaging in awake mice

Claire Rabut, Jérémy Ferrier, Adrien Bertolo, Bruno Osmanski, Xavier

Mousset, Sophie Pezet, Thomas Deffieux, Zsolt Lenkei, Mickaël Tanter

To cite this version:

Claire Rabut, Jérémy Ferrier, Adrien Bertolo, Bruno Osmanski, Xavier Mousset, et al..

Pharmaco-fUS: Quantification of pharmacologically-induced dynamic changes in brain perfusion and

connectiv-ity by functional ultrasound imaging in awake mice. NeuroImage, Elsevier, 2020, 222, pp.117231.

�10.1016/j.neuroimage.2020.117231�. �inserm-02971488�

NeuroImage222(2020)117231

ContentslistsavailableatScienceDirect

NeuroImage

journalhomepage:www.elsevier.com/locate/neuroimage

Pharmaco-fUS:

Quantification

of

pharmacologically-induced

dynamic

changes

in

brain

perfusion

and

connectivity

by

functional

ultrasound

imaging

in

awake

mice

Claire

Rabut

_,

_Jérémy

_Ferrier

_,

_Adrien

_Bertolo

_,

_Bruno

_Osmanski

_,

_Xavier

_Mousset

_,

Sophie

Pezet

_,

_Thomas

_Deffieux

_,

_Zsolt

_Lenkei

_,

_Mickaël

_Tanter

a Physics for Medicine Paris, Inserm, ESPCI Paris, CNRS, PSL Research University - Paris, France

b Institute of Psychiatry and Neurosciences of Paris, University of Paris, INSERM U1266, 102 rue de la Santé, 75014 Paris, France c Iconeus, 6 Rue Calvin, 75005 Paris, France

a

r

t

i

c

l

e

i

n

f

o

Keywords:

Functional ultrasound imaging Awake imaging

Functional biomarker

Support Vector Machine classifier Drug effects monitoring

a

b

s

t

r

a

c

t

Thereisacriticalneedforreliablequantitativebiomarkerstoassessfunctionalbrainalterationsinmousemodels ofneuropsychiatricdiseases,butcurrentimagingmethodsmeasuringdrugeffectsthroughtheneurovascular coupling,faceissuesincludingpoorsensitivity,drug-inducedchangesinglobalbrainperfusionandtheeffectsof anesthesia.

Herewedemonstratetheproof-of-conceptofaminimally-invasivefUSimagingapproachtodetecttheacute cholinergicmodulatoryeffectsofScopolamine(ScoP)onfunctionalbrainconnectivityinawakeandbehaving mice,throughtheintactskull.Amachine-learningalgorithmconstructedanad-hocpharmacologicalscorefrom theScoP-inducedchangesinconnectivitypatternsoffivemice.Thediscriminationmodelshowsimportant ScoP-inducedincreaseofthehippocampo-corticalconnectivity.Thepharmacologicalscoreledtorobustdiscrimination ofScoPtreatmentfrombaselineinanindependentdatasetandshowed,inanotherindependentgroup, dose-dependentspecificeffectsofcentralcholinergicmodulationoffunctionalconnectivity,independentfromglobal brainperfusionchanges.Inconclusion,weintroducepharmaco-fUSasasimple,robust,specificandsensitive modalitytomonitordrugeffectsonperfusionandfunctionalconnectivityintheawakemousebrain.

1. Introduction

Developmentofnew drugsforbraindisorders ischallenged both by the lackof precise understanding of drug effects on brain func-tionandbythepaucityofvalidatedpreclinicalmodels.FunctionalMRI (fMRI),byprovidingdirectreadoutofbrainactivationand connectiv-ity patternsthrough the neurovascular coupling, hasbeen proposed asaninteresting toolboth in pre-clinical andclinical drug develop-ment(Carmichaeletal.,2018;Nathanetal.,2014;Wandschneiderand Koepp, 2016) a technique referred to as ‘pharmacological MRI’ or ‘pharmaco-fMRI’ (phMRI). However, pre-clinical use of phMRI is stronglylimitedbytherequirementforimmobilizationoranesthesia andbylowsensitivity,especiallyinmice,amajorpre-clinicalrodent model.Inaddition,thecurrentgoldstandardmethodofphMRI,BOLD (bloodoxygenation-leveldependent)imagingcannotmeasureanother importantconfoundingfactor-changesinglobalbrainperfusion. Nu-clearimagingtechniquessuchasPETofferthesensitivityrequiredto

∗_{Correspondingauthor.}

E-mailaddress:mickael.tanter@espci.fr(M.Tanter).

#_{Co-seniorauthors.}

monitor drugdistribution, pharmacokineticsandpharmacodynamics. PETisalsoalargelyusedtoolincentralnervoussystemdrug discov-eryanddevelopment(Halldinetal.,2001;Pieletal.,2014),partlydue toagrowinglistofneuroreceptor-specificPETtracersandthegrowing availabilityofsmall-animalPETcameras(Cherry,2001).Asignificant numberofsmall-moleculereceptorligandshavebeenlabelledwith11C and18F,andsomeoftheseligandsreadilycrosstheblood–brain bar-rierandbindtotheirintendedtargets.Thislabellingdoesnotaffectthe chemicalpropertiesofthestudiedcompound,allowingthemonitoring ofthedrugbiodistribution.Itisextremelysensitive(uptopicomoles concentration)butreliesonlimitedspatialandtemporalresolutions.

Based on ultrasensitive Doppler imaging, functional ultrasound (fUS)imagingprovidesanemergingmethodforimagingbrain activa-tion bydirectlymeasuring cerebralbloodvolume (CBV)changes in-duced by neurovascular coupling (Macé et al., 2011). It allows the studyofdynamicbrainactivationpatternswithalargefieldofview and has been used successfully in animal models, from awake rats

https://doi.org/10.1016/j.neuroimage.2020.117231

Received24May2020;Receivedinrevisedform24July2020;Accepted31July2020 Availableonline12August2020

1053-8119/© 2020INSERM,France.PublishedbyElsevierInc.ThisisanopenaccessarticleundertheCCBYlicense. (http://creativecommons.org/licenses/by/4.0/)

(Sieu et al., 2015) to behaving primates (Dizeux et al., 2019).fUS imaging also enables to measure functional connectivity in rodents (Osmanski et al., 2014). While we have recently demonstrated fUS imagingof functional brainactivationin freely movingawake mice (Tiranetal.,2017),currentlywedonotknowwhetherfUSwouldbe use-fulforrobust,specificandquantitative(i.e.dose-dependent)imagingof pharmacologically-inducedchangesinbrainactivationand connectiv-ityintheawakemousebrain,whichallowstoavoidanesthesia-induced bias.

Inthisproof-of-conceptstudyofPharmacologicalfunctional Ultra-sound(Pharmaco-fUS),weaimedtorevealtime-anddose-dependent functionalconnectivitymodulationsinawakemice,afterthe adminis-trationofscopolamine,awidelyusedpsychoactivecompoundthat mod-ifiescholinergicneurotransmission(Deutsch,1971),whichisthoughtto beinvolvedinthephysiopathologyofAlzheimer’sdisease.Afterthe es-tablishmentofanovelawake-fUSimagingsetup,wehavefocusedonthe developmentandvalidationofamachine-learningapproach,which,by usingtherichdatacontentprovidedbyfUSconnectivityimaging,would enabletoobtainareliable‘fingerprint’ofscopolamine-inducedchanges inbrainconnectivity.This‘scopolaminefingerprint’couldbeusedto betterunderstandneuronalmechanismsofcholinergicmodulationand, bycomparingittocontrol,todevelopandvalidatenovel pharmacolog-icalcompoundsoftherapeuticvalue.

2. Materials and methods 2.1. Animals

Micewerehousedfourpercageinacontrolledenvironmentand wereprovidedwithfoodandwateradlibitum.Tominimizestress dur-ingtheexperimentalprocedure,miceweregivena7-dayacclimation periodaftertheirarrival.Allanimalsreceivedhumancareincompliance withtheEuropeanCommunitiesCouncilDirectiveof2010.Thisstudy wasapprovedbythelocalcommitteeforanimalcare(Comité d’ethique enmatièred’expérimentationanimaleN°59)underAgreement #19847-201711172358582v4.ExperimentswerecarriedoutinC57Bl/6male mice(JanvierLabs),7-week-oldatthebeginningoftheexperiment’s procedure(startingwiththeimplantationofmetalframeontheskull). Alleffortsweremadetominimizeanimalsuffering.

2.2. Implantationofmetalframe

Eight toten days before the fUS imaging acquisition session on awakefreelymovingmice,aflatmetalframewasfixedonthemouse skulltoenablethemagneticfixationoftheultrasonicprobe(Fig.1a-c), followingthesameprocedurethanpreviouslydescribed(Tiranetal., 2017).Forthesurgicalimplantationofthemetalframe,micewere anes-thetizedusingamixtureofketamine(100mg/kg,intraperitoneal(ip)) andmedetomidine(1mg/kg,ip)andplacedonastereotaxicframe. Af-terverifyingthelossofpedalwithdrawalreflex,theskinandperiosteum wereremovedtoexposetheskull.Themetalframe,aimedtostabilizea miniaturizedultrasonicprobeinamagnetizedprobeholder(Fig.1a.b) duringtheacquisitionsessions,was fixedon theskullwithtwo sur-gicalanchoringscrews(Fig.1c).Thestabilityofthemetalframewas furtherensuredusingdentalcement(DentalAdhesiveResinCement -SuperbondC&B).Theframewindow(rangingroughlyfromtheskull landmarksBregmatoLambda)wascoveredwithSiliconElastomerKwik Cast○R _{(WorldPrecisionInstruments)toensureprotectionandmaintain}

integrityofthebone.Anesthesiawasthenreversedwithasubcutaneous injectionofatipamezole(1mg/kg,Antisedan).Allmicereceiveda pro-phylacticadministrationofantibiotics(Borgal24%,0.6ml/kg/day, sub-cutaneous(s.c.))andmeloxicam(Metacam5mg/kg/day,s.c.)to pre-ventpost-operativepain.Themicewereleftfortwodaystorecover fromthisprocedure.Micecarrychronicallyonlythemetalframeand twoscrews,correspondingtoatotalweightof1.4g.Thenormalactivity ofthemicewasnotnotablydisturbed.

2.3. Functionalultrasoundimaging

Imagingprocedure,acquisitionswereperformedusinga15MHz ul-trasonicultralightprobeprototype(15MHz,64elements,0.110mm pitch,Vermon,Tours,France)(Fig.1a)connectedtoaprototype ultra-fastresearchultrasoundscannerwithreal-timeneuro-imagingsoftware (Neuroscan, Iconeus,Paris,France).Theprototypeimplemented real-timetranscranialfunctionalultrasound(Maceetal.,2013).Eachimage wasobtainedfrom200compoundedframesacquiredat500Hzframe rate,using11tiltedplanewavesseparatedby° 2(-10°,-8°,-6°,-4°,-2°,0°, 2°,4°,6°,8°,10°)acquiredata5500Hzpulserepetitionfrequency(PRF). Imagingsessionscouldlastaslongasneeded(typicallyonehourand half)andwereperformedbyreal-timecontinuousacquisitionsof succes-siveblocksof400msofcompoundedplanewaveimaging.Eachblock wasprocessedusingaSVDclutterfilter(Demené etal.,2015)to sepa-ratetissuesignalfrombloodsignaltoobtainapowerDopplerimage. 2.4. Imagingproceduresonawakemice

Mainpreclinicalimagingmodalitiesrequiretheimmobilityof the subjectwhich necessitatestherestraining orthesedation ofanimals. Bothanesthesiaandconstriction-inducedstressinevitablyaffect cere-bralactivityandputatively interferewithdatainterpretation.To en-abletranscranialwhole-depthbrainimaginginawakebehavingmice, weusedaminiaturizedprototypeultrasonicprobe(Fig.1a),placedinan adjustablemagneticprobeholder(Fig.1b),which,afterbeingclipped onametalframechronicallyfixedontheintactmouseskullby mini-mallyinvasivesurgery(Fig.1c),allowsthechoiceofthetranscranial imagingplaneforeachimagingsession.ThereforefUSallowstostudy behavinganimals,howeverfurtherstepsofdataandsignalfilteringare necessarytoremovemotionartefactsrecordedintheawakemouse(see Section2.6.1.Framesremoval,filteringandmotionregression).

Mice were placed in a home-made box freely rolling aroundon beads, whichpreventedexcessivetwistingofthetransducercableby allowingthecagetofreelyrotateandallowedrelativelyfreemotion oftheanimalsduringfUSimagingsessions(Fig.1d).Thishybrid con-figurationbetweenfreely-movingandhead-fixedsetupsgavethebest compromiseintermsofmotionandheadrestraints.Afteraresting pe-riodandacoupleoftrainingsessions(Supplementary Fig.1a ),mice werehabituatedtoremainupto3hinthesetup,withtheprobefixed ontheirhead.

Inordertoselecttheadequateimagingplane,eachmousewasfirst rapidlyanesthetizedwithisoflurane(1.5%).TheKwik-Cast○R _layer

pro-tectingtheskullwasremovedandtheskullwascleanedwithsaline. Aftermagneticallyclippingtheprobe-holder,itspositionwasmanually adjustedwithsmallscrewsusingthereal-timeimageofthetranscranial Dopplerimages.Forallthisstudy,weacquiredimagesatthecoronal planeat Bregma-2.3mm.Thechoiceof thiscoronalplanewas mo-tivatedbytheresults frompreviousstudiesbasedon histology anal-ysis (J-C.Leeetal.,2018;J-SLeeetal., 2016; Parketal.,2016) or fMRI (Sperlingetal., 2002; Wink etal., 2006) showing that scopo-lamineinduced-effectsandrecovery-effectsinmiceweresignificantly noticeableinhippocampus,notablybecauseofthememoryimpairment causedbyscopolamine.Whenrequired,andbeforestartingacquisition, themousewaslefttorecoverfromtheisofluraneforatleast30min in-sidethemobile-boxwiththeprobeattached.Acquisitionswerestarted afterthisinitialrecoveryperiod,assoonasthemousewasquietand behavednormally.

Inordertostudythefunctionalconnectivityatgenuineresting-state periods, weanalyzedtimeperiodswherethemice werenotactively moving(seeSection2.6.2Extractingquiettimeperiods).

2.5.Pharmacologicalstudies

Longitudinal study: Wefirstinvestigatedthepharmacodynamicsof functionalconnectivitychangesinducedby3mg/kgscopolamineina

C. Rabut, J. Ferrier and A. Bertolo et al. NeuroImage 222 (2020) 117231

Fig.1.AwakefUSimagingsetupandthemethodforthelimitationofmotionartifacts.a)Miniaturizedultrasonicprobeb)Adjustablemagneticprobeholder.c) Metalframechronicallyfixedonthemouse’sskull.ScrewswerepositionedinfrontoftheBregmaandbehindtheLambdapoints.Thefieldofviewisapproximately 5mmwide.d)Imageofamousefreelymovingintherolling-box.(e-h): Exampleofaselectionofaquiettimeperiode)Threerepresentativeperiodsofthemouse activityduringrecording.WecanseparatethetissuemotionfromtheDopplersignal.Thetissuemotionadmitshighvaluesalloverthebrainduringthemouseactive timewhereasthismotionishardlydetectableintheDopplersignal.f)NormalizedUltrasonicCerebralSignal(alsocalled‘IQsignal’),withoutfiltering,representing allthebackscatteredechoesreceivedfromthebrain.g)NormalizedpPowerDopplersignal(afterSVDfilteringtoremovethefirstmodesofenergyinanew spatio-temporalorthogonalbase(Demené etal.,2015),showingthetemporalcourseoftheCBVsignal.Theblackrectanglefocusesona10speriodwherethestandard deviationofbothnormalizedsignalsdoesn’texceed10%,definingitasaquiettemporalperiod.h)Confirmationofthetimeperiodselectionbyanalysisofthetissue signal:whennormalized,thetimecourseofthetissuesignalduringtheselectedtimeperiodshouldalsoadmitastandarddeviationunder10%.Thetimeslotsthat formedasuccessiveuninterruptedperiodof80seconds(toavoidfragmentedorsmallindividualperiods)wereusedforthefunctionalconnectivityanalysis.

groupof5mice.Forthisstudy,wefirstperformeda30mincontinuous baselineacquisitionofthecerebralperfusionofawakemice.Wethen injectedsubcutaneously300μLofscopolamine(3mg/kg)totheawake mouseandacquiredhemodynamicactivityduringoneh,starting15min afterscopolamineinjection(Supplementary Fig.1b ).

Dose-response Study: Wenextinvestigatedthedose-dependence, as well as the effects of post-scopolamine treatment with the mus-carinic acetylcholine receptor agonist milameline and theeffects of aperipherically-restrictedscopolaminederivative:methyl-scopolamine (Andrewsetal.,1994).Thisstudywasperformedonsixdifferentgroups ofmice.Fourdosesofscopolamineweretested(0.1mg/kg;0.2mg/kg; 0.5mg/kg:3mg/kg;n=5pergroup).Onegroupreceivedaninjection of 300μLof saline(group0) andalast group(group 5)receiveda doseofmethyl-scopolamine(1mg/kg,n=4micepergroup).Allthose groupsofmiceweresubjectedtothefollowingprotocol(Supplemen- tary Fig.1.c ):afirst30minbaselinepharmacologicalsetoffUSimages wereacquired,afterwhichsaline,scopolamine(0.1mg/kg;0.2mg/kg; 0.5mg/kg;3mg/kg)ormethyl-scopolamine(3mg/kg)(allinjected vol-umes:0.3ml)wassubcutaneously(s.c.)administered.After15min an-other30minoffUSimageswasacquired.Then,themicewereinjected withmilameline(1mg/kg,s.c.)exceptforgroup0,andalast30min fUS-acquisitionwas performed15minafterthemilameline injection (allinjectedvolumes0.2ml).

Methyl-scopolamineis amethylatedderivativeof scopolamine. It presentsthesamereceptorbindingcharacteristicsasscopolaminebut doesnotcrosstheblood–brain–barrier.Methyl-scopolaminecanbeused asacontroldrugatthesamedosethanscopolamine, forbehavioral studiesusingscopolamine(Shahetal.,2015).

2.6.1. Framesremoval,filteringandmotionregression

Whileusinghead-fixedprobesfUSimagingismuchlesssensitiveto motionartifactsthanfMRIandalthoughtheuseoftherollingbox fur-thersmoothenstheeffectofsuddenmovements,motionartifactsmay stillremainintheDopplersignal.Astheyappearinseveralbrain struc-turesatthesametime,thoseartifactscanleadtoartificiallyhigh cor-relationsbetween structureswhicharenotdue toneuralsignalsbut tomotion.Tomitigatethoseeffectsinourresults,severalstepswere necessary.First,weusedanadvancedclutterfilterbasedonsingular valuedecompositiontoseparatebloodmotionfromtissuemotionusing spatio-temporalcoherence(Fig.1e).Previously,thisfilterwasshown tobe arobusttooltoreducemovementartifactsinseveralscenarios (Demené et al.,2015).Secondly,weidentifiedframestoberemoved from eachacquisition byusing acombination of thresholds-a first thresholdonDopplersignalschosenat +/-10% ofthemedianvalue (Fig.1g)andasecondthresholdusingatissuevelocityestimatorbased onbeamformedIQdata(Kasaietal.,n.d.)andchosenat+10%above thebaseline(Fig.1.h).Thirdly,ofallremainingframesweonlykept framesthatformedasuccessiveuninterruptedperiodof80s,toavoid fragmentedorsmallindividualperiods.

Wethenappliedalow-passfilterat0.2Hztoremovehighfrequency signalswhilepreservingtheresting-statefrequencyband.Thesignalwas thendetrendedwithapolynomialfitoforder4toremovelowfrequency fluctuations.Finally,singularvaluedecompositionofthedatawas per-formed.Sincethefirstspatialeigenvectoriscoupledtothewholebrain tissue,itwassystematicallyremoved.Althoughthisformofglobal sig-nalregressioncanleadtoanti-correlationvaluesbetweenstructuresthat maybedifficulttointerpret,itwaspreferredtoavoidanyremaining globalmotionsignalthatcanaffectthewholebraintissueandinduce artificiallystrongcorrelations.

2.6.2. Extractingquiettimeperiods

Longitudinalstudy:foreachofthe5mice,wewereabletoextract fromtheirlongitudinalacquisitionssevendifferentblocksofuseful sig-nallastingfrom80to120s(asdefinedinthepreviousparagraphFrames removal,filteringandmotionregression)andcorrespondingtoseven dis-tinctperiodsoftheacquisition:twoperiodspriorscopolamineinjection

Period Acquisition Range

T1 Baseline T = 2min - > 6min

T2 Baseline T = 22min - > 26min

T3 After Scopolamine T = 1min - > 5min T4 After Scopolamine T = 13min - > 17min T5 After Scopolamine T = 28min - > 31min T6 After Scopolamine T = 44min - > 47min T7 After Scopolamine T = 55min - > 60min

T1andT2,andfoursuccessiveperiodspostscopolamineinjectionfrom T3toT7.ThetimerangesofthoseperiodsaredetailedinTable1and presentedinFig.2.

Dose-response study :foreachgroup,weextractedfromtheir longi-tudinalacquisitionsoneblockofatimeperiodrangingfrom80sto200s ineachpharmacologicalstate(i.e.baselinestate,afterscopolamine in-jection,aftermilamelineinjection,…).

2.6.3. Seed-basedanalysis

Seed-basedmapswerealsocalculatedestimatedfromthecorrelation overtimeofthetimecourseoftheCBVsignalextractedfromasmall selectedROI(theseedregionasdefinedbytheoperator),withsignals fromalltheotherpixelsofthebrain.Twodifferentseedswerechosen: thelefthippocampus(seedS_h)andtheleftfrontalcortex(seedS_c)both atthelevelofBregma-2.3mm.

2.6.4. Functionalconnectivitymatrices

Inordertobuildfunctionalconnectivitymatrices[M],weusedthe timecourseoftheCBVsignalextractedfrom10differentROIsdefined fromthePaxinosAtlas(KetihFranlkin,2011)(SeeFig.2cforROIs loca-tion).WethendeterminedthePearsoncorrelationbetweeneachfiltered ROIs’signalandreportthecorrelationcoefficientsona10×10matrix. Attheintersectionoftheith_lineandjth_{column,thevalueofthecell} M_i,jofthematrixM isthecorrelationcoefficientbetweentheregionn°i andtheregionn°j.

2.6.5. Hippocampo-corticalcorrelationscore

Thehippocampo-corticalcorrelationscoreevaluatesthecorrelation betweentheregionsofinterest(ROIs)locatedinthecortex(regions1to 6)andtheregionsofthehippocampus(regions7and8).Itcorresponds tothemeanvalueofthecellsfromtheconnectivitymatrices express-ingthehippocampo-corticalcorrelation,i.e.cellsofcoordinates:(7,1); (7,2);(7,3);(7,4);(7,5),(7,6);(8,1);(8,2);(8,3);(8,4);(8,5);(8,6) 2.6.6. PharmacologicalfUSstatistics

Paired T-tests were used to compare baseline, post-scopolamine andpost-milameline injectionscores forthedifferent studiedgroups (Fig.3.d&Fig.5.b).Similarly,pairedT-testswereusedtocomparethe baseline,scopolamineandmilamelinehippocampo-corticalscoresfor thedifferent studiedgroups(Supplementary Fig.3 ).Forthe statisti-calthresholdsignificanceforpvalues,weapplythefollowing conven-tion(followedbytheAPA(AmericanPsychologicalAssociation)andthe NEJM(NewEnglandJournalofMedicine)):

p>0.05:non-significant (ns) p<0.05:significant (∗₎ p<0.01:verysignificant (∗∗₎ p<0.001:extremelysignificant (∗∗∗₎

Significance values however, were not corrected for multiple-comparisons.

2.6.7. StatisticalanalysisusingaSupportVectorMachinemodel

We trained a linear regression model (Support Vector Machine (SVM)type)torecognizeinanunbiasedwaytheconnectivitychanges duetoscopolamineinjectioninafirstdataset.ThegoalofanSVM clas-sifieristolearnhowtodetermineaborderbetweentwotraining cat-egories.InanN-dimensionalspace,twosetsof(training)pointsofN

C. Rabut, J. Ferrier and A. Bertolo et al. NeuroImage 222 (2020) 117231

Fig.2.Longitudinaleffectsofasingle-dosescopolamineinjectiononcerebralconnectivity.AfterabaselineimagingacquisitionattheBregma-2.3mmcoronalplane during30min,300μlofscopolaminewassubcutaneouslyinjectedata3mg/kgdose,followedbycontinuousimagingacquisitionofonehour.Sevenmotion-freetime periods(twobaselineT1-T2,and5post-scopolamineT3-T7)wereselectedtoanalyzethedynamicevolutionofconnectivity.a-b)Longitudinalanalysisofconnectivity ofthehippocampalseed(Sh)positionedin(a)thelefthippocampusandofthecorticalseed(Sc)positionedin(b)theleftcortex.c-d)Meanseed-basedanalysis correlationmaps(averagedover5mice):longitudinalanalysisofconnectivityofthecorticalseed(Sc)positionedintheleftplacedintheleftlateralparietalassociatif (L-PtA)cortex.e)10regionsofinterest(ROIs)areselectedtorepresenttheconnectivitypatternsusingconnectivitymatrices:1-6:lateralparietalassociationcortex (LPtA);2-5:medialparietalassociationcortex(MPtA);3-4:retrosplenialgranularanddysgranularcortex(RS);7-8:hippocampus(CA);9-10:posteriorthalamic nucleargroup(Po).f)Meanconnectivitymatricesateachtimeperiodshowingpair-wisetemporalcorrelationsbetweenthesignalofthe10selectedROIS.

coordinatesbelongingtotwodifferentcategorieswillbeseparatedby linearregressionbyan“optimal” hyper-line.Thisline(ofNdimensions) correspondstotheborderfarthestfromallthepoints:itissaidtohave the"bestgeneralizationcapacity".

Wetrainedourmodelwiththecorrelationmatricesofsize10×10 ofthefivemiceusedduringthepharmacodynamicstudytofindthis optimalborderbetweenthetwopharmacologicalstates.Intotal10 ma-trices:fivematricesatbaselinestateandfivematricesfor scopolamine-treatmentwereused.Weassignedascore=0tothebaselinematrices andascore=1tothematricesofanimalstreatedwithscopolamine (Fig.3a).

Knowingthescore (0 or1 )ofeachconnectivitymatrixC ofthe train-inggroups(baselinematricesorscopolaminematrices),theSVM algo-rithmcalculatedtheoptimalhyper-line A oftheequation score =f(C ) (Fig.3b).

Theresultinghyper-lineisa10×10matrix A sothat:

𝐬𝐜𝐨𝐫𝐞=𝐂.𝐀+𝐛

(wherebisthey-interceptofthehyper-lineCandb=0.1575) EachcellA_i,jof A representstheoptimalpointmaximizingthe dis-tancebetweenthecellsBi,jfromthematricesofthebaselinegroupand

thecellsSi,jfromthematricesofthescopolaminegroup.

Then,tovalidatethescoreofthesecondcohortofmice(whichis independentofthetrainingcohort),weconsideredtheinverseproblem andcalculatedforeachconnectivitymatrixC’ toevaluatetheestimated scoreaccordingtotheSVMmodel(Fig.3.c.d).

2.6.8. Cross-validation

Toevaluatethestabilityofthepresentedresults,wesubjectedour SVM-modeltocross-validation.Todoso,werandomlypicked100 com-binationsof5trainingmatricesand8validationmatricesamongthe13 correlationmatrices(5fromthemiceofthetrainingcohort,8fromthe miceofthevalidationcohort)usedfortoestablishandvalidatetheSVM modelatfirst(Fig.4.a).Thesamelinearregressionmodelwasusedto calculate100hyper-linesC optimizingtheborderbetweenthebaseline

statesandscopolaminestatesofthe5randomlypickedtraining matri-ces.ThemeanSVM-coefficientmatrixisrepresentedinFig.4b.

Theperformanceofthecross-validationmodelisrepresentedwith themeanReceiverOperatingCharacteristic(ROC)curveofFig.4c.The ROCcurvesofeachrandomsamplingrepresentthediagnosticabilityof ourclassifiertodiscriminatethebaselinestatesfromthepathological states.WecreateherethemeanROCcurvebycalculatingforeach sam-pling(andthenaveragingbythenumberofsampling)thetruepositive ratefunctionofthefalsepositiverateatvariousthresholdsettings.Four indicesforeachsamplingevaluatetheadequacyofthemodel:

AreaUnderCurve(AUC):areaundertheROCcurve,linearly corre-latedtotheperformanceofthemodel.

Accuracy:(numberofcorrectlyestimatedscopolaminematricesand correctlyestimatedbaselinematrices)/(totalnumberofmatrices) Sensitivity:(numberofcorrectlyestimatedscopolaminematrices)/

(totalnumberofscopolaminematrices).

Specificity:(numberofcorrectlyestimatedbaselinematrices)/(total numberofbaselinematrices)

ThemeanAUC,Accuracy,SensitivityandSpecificityaveragedover the100trialsaredisplayedinFig.4c.

2.7. Codeaccessibilityanddataavailability

The data that support the findings of this study, including pre-processing andanalysis scripts, areavailable in an openrepository: https://sandbox.zenodo.org/record/558582.

3. Results

3.1. Longitudinaleffectsofasingle-dosescopolamineinjectiononcerebral connectivity

UsingtheawakefUSimagingsetupdescribedin2.4,wefirst inves-tigatedthepharmacodynamicsoffunctionalconnectivitychanges in-ducedby3mg/kgscopolamineinagroupof5mice.Thescopolamine

Fig.3.Automaticdetectionofpharmacologically-inducedconnectivitychangesa)WetrainedtheSVMlinearmodelwith10setsofdata:5baselineconnectivity matricesassignedwithascore0(“baselinestate”)and5scopolaminematricesassignedwithascoreof1(“pathologicalstate”)b)SVM-coefficientmatrixassociated tothetrainedmodel:valuesofindividualcellsindicatetheirimportanceinthediscriminationofthetwotraininggroups.Notably,mostcellsofthe hippocampo-corticalcorrelationdisplayhighvalues.ROIs:1-6:lateralparietalassociationcortex(LPtA);2-5:medialparietalassociationcortex(MPtA);3-4:retrosplenialgranular anddysgranularcortex(RS);7-8:hippocampus(CA);9-10:posteriorthalamicnucleargroup(Po)(l=left;r=right)c) Validationofthepredictedmodelonasecond independentcohortofmicewhichhaveundergonethesameexperimentalprotocol.d)Predictionofthe“scopolaminescore” ofthe16validationmatrices,byusing thetrainedmodel.Atbaselinestate:𝐬𝐜𝐨𝐫𝐞𝐒𝐕𝐌= −0.20.Afterscopolamineinjection:𝐬𝐜𝐨𝐫𝐞𝐒𝐕𝐌= 0.69(𝐩=0.0004).Graph:mean±singlescore.

doseusedforthisexperimentisinthetypicalrangeforrodent mod-els(Dengetal.,2019;Guoetal.,2016;Haideretal.,2016;Linetal., 2016;Muhammadetal.,2019;Shahetal.,2015;Xuetal.,2016).At baselinestate,weselectedtwomotion-freeperiods,thefirstfromthe 0-10minutesperiod(labeledT1onFig.2)andthesecondfromthe 20-30minutesperiod(T2).Aftersubcutaneousscopolamineinjection,we selectedmotion-freeanalysisperiodsfromevery10minutesrangesof acquisition(T3-T7,Supplementary.Fig.1 ).

InordertocomparefUSimagingwithconventionalpharmacological studiesbasedonthevariationofbaselineCBVorBOLDsignal,wefirst studiedtheglobalcerebralflowbaselinevariationsafterscopolamine injectionintwocontrolgroupsofN=5mice.Althoughwerecognizethe lowsamplingsizeofthesetwogroups,wedidnotmeasuresignificant changeofglobalCBVbaselineafterscopolamineinjection(Supplemen- tary Fig.2a ).

Weusedseed-basedanalysistoevaluatethecorrelationoftheCBV signalbetweenallpixelsoftheimagingplaneandtwoseedregions: ahippocampalseed (Sh),placed intheleft dorsal hippocampusand acorticalseed (Sc),placedin theleft lateralparietalassociation (L-PtA)cortex(Fig.2a,c).Beforethescopolamineinjection(time-points

T1 and T2), the two hippocampi were the only regions correlating with Sh(Fig.2a-b),andthecorticalregions wereonly strongly cor-relatingwithSc(Fig.2c-d),illustratingthepowerfulinterhemispheric connectionsbetweenthesestructures.Rightafterthescopolamine in-jection(timeT3,correspondingtothe0-10minutespost-injection pe-riod),theseed-basedmapsdidnotshownotabledifferenceascompared tobaseline(T1 andT2).However,increasebothin interhemispheric andhippocampo-corticalcorrelationscanbeobservedfromT4:the cor-ticalregions emergeamongtheregions stronglycorrelatingwithSh. Similarly,Sc andbothhippocampi presentedelevatedcorrelation.At timeT5,T6andT7,thishippocampo-corticalcorrelationismaintained, suggesting aloss of theintra-regional distinctiveness of the connec-tions(cortico-corticalconnectionsandhippocampo-hippocampal con-nections).

Next,weextendedtheanalysisofscopolamine-inducedchangesin functionalconnectivitytomultipleregionsofinterest(ROIs)atthe in-vestigatedcoronalplanebyusing correlationmatrices(Fig.2e-f).At baselinetime-points(T1-T2),thecorticalregions(regions1to6),the hippocampalregions(regions7and8)andthethalamicregions(regions 9and10)showhighintra-regionalcorrelations,andweakinter-regional

C. Rabut, J. Ferrier and A. Bertolo et al. NeuroImage 222 (2020) 117231

Fig.4. Cross-validationoftheSVMclassificationmodela)Illustrativedesignofthecross-validationmodelusing100iterations.Ateachtrial,adifferentcombination of5trainingmice/8validationmiceisrandomlyevaluated.b)Crossvalidationcoefficientmatrixaverageoverthe100trials.c)Cross-validationmeanROCcurve. Theerrorbarrepresentsthestandarddeviationoverthe100trials.d)CircularGraphtoillustratetheconnectivityconnectionsinanetwork.

correlations.StartingfromT4,similarlytotheseed-basedanalysis,we observedasubstantialincreaseofthecorrelationcoefficientsbetween theROI’s1to6(corticalregions)andtheROI’s:7and8(hippocampal regions),indicatingelevatedinter-regionalconnectivity.

These observationssuggest thatscopolamine induces an increase of the hippocampo-cortical functional connectivity in mice. While showingcleartendencies,which reachedsignificantlevelsin several ROIcombinations(datanotshown),therelativelylownumberof an-imals (n=5) in these pilot groupsdid not provide sufficient statisti-calpowertocompensateforthefalse-positive-resultinducingeffectof multiplecomparisons. Thetraditionalresponsetothisclassical prob-lemof under-samplingis toelevatethenumberofexperimental ani-mals,which,byraisingethical,financialandlogisticchallenges, ulti-matelyhinderstheadvanceofpre-clinicalresearch.Wehypothesized thattherichdatacontentoffUS-detectedconnectivitychangesinduced byscopolamineat asinglecoronalsectionmayallowtoconstructa unique‘score’,thatcouldreliably classifyanimalsintocontrol (base-line)andtreatedgroups.Suchascorewouldbeusefulfor pharmaceu-ticalresearch,throughtestingofdrug candidatesbycomparingtheir capacitytorescuepharmacologically-inducedpathological connectiv-itypatterns,suchasscopolamine-inducedhippocampal

dysconnectiv-ity,inareducednumberofexperimentalanimals.Indeed,arecent clin-icalstudyusingmachine-learningimprovedanalysisofEEGbiomarker datainhumanshasshownhighperformancetoclassify scopolamine-treatedvs.non-treatedindividuals(Simpragaetal.,2017).Toconfirm thishypothesis,weimplementedandthenpharmacologicallytesteda supportvectormachine(SVM)basedautomaticclassifieralgorithmto constructanad-hoc“scopolaminescore”.Importantly,thematrixscore wouldindicateandquantify,inacompletelyunbiasedway,the con-nectivitychangesinducedbypharmacologicalcompoundsthatmodify brainfunction.

3.2. Automaticdetectionofpharmacologically-inducedconnectivity changes

Inordertoconstructaclassifierwhichwouldallowthe discrimina-tionofscopolamine-inducedconnectivitypatternsfromthebaselinein awakemice,wetrainedasupportvectormachine(SVM)linear regres-sionmodelalgorithm,whichaimstomaximizethegeometricmargin betweenthefunctionandthedata.Weusedtheabovepresented con-nectivitymatricesfromtimepointT1forbaseline(assignedscore0)and timeT5forscopolamine-treatment(score1)aspredictordata(intotal 10matrices,n=5)(Fig.3.a).Aftertrainingofthemodel,weobtained

Fig.5. Pharmacologicalvalidationfortheautomaticdetectionofscopolamine-inducedconnectivitychanges.a)Schemaofthetreatmentprotocol.b)Pharmacological scoresobtainedinthe5treatmentgroups.Groupseitherreceivedsaline(group0)orincreasingdosesofscopolamine(groups1-4)orasingledoseofthe peripherally-restrictedmethyl-scopolamine(group5),followedbyasingledoseoftheantagonistmilameline.PairedT-testswereusedtocomparebaseline,post-scopolamine andpost-milamelineinjectionscoresforthedifferentstudiedgroups(∗_p<0.05;∗∗_{p<0.01).Bargraph:mean±SEM.}

thebestvaluesmaximizingdistancesoftheinterceptbandoftheslope

A inthelinearregressioncalculatedupontheequation: score=𝐂.𝐀+b

where:

- C isaconnectivitymatrix(variable)

- A isa matrix(samesizeas C )which cell’svalues (notedAi,j for

thevalueofthecellofcoordinate(i,j)inthematrix)representthe SVM-coefficientsofeachcellinthediscriminationofthegroups.A isboundedbetween-0.098and0.11

- b=0.1575

Notably, the highest values of A were provided by the cells of cortico-hippocampal correlation, in particular: 𝐴1,7 =0.088, 𝐴2,7 = 0.089, 𝐴3,7 =0.097, 𝐴6,7 =0.11, 𝐴2,8 =0.080, 𝐴3,8 =0.11, 𝐴6,8 = 0.093(Fig.3b). Themeanvalueofthecellsbelongingtothe cortico-hippocampalcorrelationis 0.70whereas themeanvalueof the rest of the cells of A is -0.0091. This difference is highly significative (p=0.0004).Thisresultconfirms,inanunbiasedway,that connectiv-itychangesinthehippocampo-corticalcorrelationareamajor conse-quenceofscopolaminetreatmentandraisethepossibilitythattheycan serveasarelevant discriminatoryfactorforthereliabledetection of scopolamine-inducedpharmacologicalstates(i.e.theensembleof func-tionalconnectivityalterations)inareduced numberof experimental animals.

Wethenvalidatedthismodelinevaluatingthe“scopolaminescore” oftheconnectivitymatricesbefore(baselinestate)andafterinjection ofscopolamineofasecondcohortofmice,independentfromthefirst one,thathadfollowedthesamepharmacologicalprotocol.Remarkably, theSVMlinearmodelwasstillabletoreliablydifferentiate(p=0.0004) thetwopharmacologicalstates,yieldingascoreof-0.2forthebaseline matricesand0.69forthescopolaminematrices(Fig.3c.d).

Next, in order to testthe robustnessof theclassification perfor-mance,weusedcross-validationwith100iterations(Fig.4).Notably, the mean SVM-coefficientsmatrix over those 100 iterations yielded againthehighestvaluesforthecellsofthecortico-hippocampal cor-relation(𝐴2,7 =0.079, 𝐴3,7 =0.076,𝐴4,7 =0.079,𝐴6,7 =0.044,𝐴2,8 =

0.080,𝐴3,8 =0.11,𝐴6,8 =0.093).Theexcellentperformanceofthe cross-validationtest(AUC=0.97 ± 0.03,Accuracy=95% ± 4%, Sensi-tivity=94% ± 8%,Specificity=95% ± 7%)furthervalidatedthe robustnessoftheSVMclassificationapproach.

TheseresultsdemonstratetheabilityandrobustnessoftheSVM re-gression modeltoautomaticallydeterminethepharmacologicalstate ofmice(i.e.baselinevs.scopolamine-treated),byusingsparsetraining data.

3.3. Pharmacologicalvalidationfortheautomaticdetectionof scopolamine-inducedconnectivitychanges

Forthepharmacologicalvalidationofthescore-basedautomatic de-tectionmethodofscopolamine-inducedconnectivitychanges,described above,wenextinvestigatedthedose-dependencyofscorevaluesonsix different awakemicecohortsinjected atdifferentscopolamine doses (seeMethods.)Wealsoevaluatedtheeffectsofpost-scopolamine treat-mentwiththethemuscarinicacetylcholinereceptoragonist milame-lineandtheeffectsofaperipherically-restricted scopolamine deriva-tive,methyl-scopolamine(Andrewsetal.,1994).Weusedtheregression modelpresentedabovetoevaluatethepharmacologicalscoreforeach group(Fig.5.b).

As expected, the score of baseline time periods was low (𝑠𝑐𝑜𝑟𝑒𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒_𝑔𝑟𝑜𝑢𝑝0= −0.15; 𝑠𝑐𝑜𝑟𝑒𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒_𝑔𝑟𝑜𝑢𝑝1= −0.066; 𝑠𝑐𝑜𝑟𝑒𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒_𝑔𝑟𝑜𝑢𝑝2= −0.12; 𝑠𝑐𝑜𝑟𝑒𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒_𝑔𝑟𝑜𝑢𝑝3= −0.097; 𝑠𝑐𝑜𝑟𝑒𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒_𝑔𝑟𝑜𝑢𝑝4= −0.26; 𝑠𝑐𝑜𝑟𝑒𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒_𝑔𝑟𝑜𝑢𝑝5= −0.091). Nosignificantdifferenceofthescorewasnoticedaftertheinjectionof salineingroup0(𝑠𝑐𝑜𝑟𝑒𝑠𝑎𝑙𝑖𝑛𝑒= −0.17),orafterinjectionofscopolamine

forgroup1(𝑠𝑐𝑜𝑟𝑒𝑠𝑐𝑜𝑝0.1= 0.35)andgroup2(𝑠𝑐𝑜𝑟𝑒𝑠𝑐𝑜𝑝0.2= 0.13) de-spitetheincreaseofthescoresforbothofthosetwolow-doseinjected groups.Remarkably,fromgroup3 thescopolaminetreatedgroups ex-hibitsignificantlyhigherscores:group3(𝑠𝑐𝑜𝑟𝑒𝑠𝑐𝑜𝑝0.5= 0.56,significant increase comparetobaseline:𝑝=0.014);group 4(𝑠𝑐𝑜𝑟𝑒_{𝑠𝑐𝑜𝑝}3= 0.61, significantincreasecomparetobaseline:𝑝=0.003).Nonetheless,these resultsindicatethatthepharmacologicalscoreshowsdose-dependent elevations,indicatingpharmacologicalrelevance.Importantly,changes infunctionalconnectivitywerenotcorrelatedwithchangesinglobal

C. Rabut, J. Ferrier and A. Bertolo et al. NeuroImage 222 (2020) 117231

cerebralbloodflow(Suppl. Fig.2 ),asabove.Furthermore,thescoreof group5,treatedbytheperipherallyrestrictedmethyl-scopolaminewas notsignificantlydifferentfromthescoreofbaselinegroup,indicating thatthe scored brainconnectivity changesrequire that scopolamine traversetheblood-brainbarrier.Finally,treatmentwiththemuscarinic acetylcholine receptoragonist milameline reverted the scopolamine-induced pharmacological score in each group near to baseline values:𝑠𝑐𝑜𝑟𝑒_{𝑚𝑖𝑙𝑎}_𝑔𝑟𝑜𝑢𝑝1= 0.17;𝑠𝑐𝑜𝑟𝑒𝑚𝑖𝑙𝑎_𝑔𝑟𝑜𝑢𝑝2= 0.23;𝑠𝑐𝑜𝑟𝑒𝑚𝑖𝑙𝑎_𝑔𝑟𝑜𝑢𝑝3= 0.15; 𝑠𝑐𝑜𝑟𝑒𝑚𝑖𝑙𝑎_𝑔𝑟𝑜𝑢𝑝4= 0.19; 𝑠𝑐𝑜𝑟𝑒𝑚𝑖𝑙𝑎_𝑔𝑟𝑜𝑢𝑝5= 0.16 further confirm-ing pharmacologicalrelevance.Noneof thosescoresis significantly differentfromtheirbaselinereference.

Together,theseresultsshowthatthefunctionalconnectivitydata ac-quiredwithfUSimagingatasinglecoronalplaneallowsthe establish-mentofapharmacologicallyrelevantscorewhichisusefulforefficient classificationoftreatmentconditions,withoutanyinitialassumption,in alownumberofanimals.

4. Discussion

Inthisstudywehavefirstdevelopedanexperimentalprotocolfor transcranialfUSimaging of functionalconnectivity in freely-moving awake-mice, which prevents confounding effects of anesthesia. Ulti-mately,thisapproachwhichusesalightultrasonicprobe,maybe ex-tendedtofunctionalimagingofbrainactivationandconnectivityduring behavioralessays.

Herewehaveusedthissetupfortheimagingoffunctional connectiv-itychangesinducedbytreatmentwithscopolamine,amuscarinic acetyl-cholinereceptorantagonistandwidelyusedreferencedrugfor induc-ingcognitivedeficitsinhealthyhumansandanimals(Klinkenbergand Blokland,2010).Weselectedthefunctionalconnectivitytime-domains where the ultrasonic cerebral signal resulting either from tissue signalsor from bloodflow signals can be discriminated.Qualitative evaluationindicatedtime-dependentchangesofhippocampaland corti-calconnectivityfollowingscopolamineapplicationatthestudied coro-nal slice.This was confirmedbyunbiased selectionof themost im-portantconnectivitychangesbyamachine-learningalgorithm,which alsoprovidedapharmacologicalscoreofscopolaminetreatmentafter trainingin5experimentalanimals.Analysisoftheconnectivity matri-cesinanindependentexperimentalcohortindicatedthatthis “scopo-laminescore” ishighlyefficient for theclassificationof baselinevs. post-treatmentbrainstates. Finally,pharmacologicalvalidationofthe “scopolaminescore” showedefficientmeasureofdose-dependent con-nectivity changes,followed byreversal of thescore tobaseline val-uesbythemuscarinicacetylcholinereceptoragonistmilameline,and therequirementforthepenetrationofscopolaminethroughthe blood-brainbarrier,asdemonstratedbythelackofasignificantscoreafter administrationoftheperipherally-restricted methyl-scopolamine. To-gether,theseresultsshowthatthefunctionalconnectivitydataacquired withfUSimagingatasinglecoronalplane,non-invasivelyandinawake freely-movingmicewithoutthebiasofanesthesia,allowsthe establish-mentofapharmacologicallyrelevantscore,whichisusefulforefficient classificationoftreatmentconditions,withoutanyinitialassumption,in alownumberofanimals.

While we have previously demonstrated that fUS imaging is able measure the functional connectivity in anesthetized rats (Osmanskietal.,2014),subsequentexperimentsindicatedthatinmice thealreadysignificantlyweakerfunctionalconnectivityishighly sen-sitivetostandardanesthesiaprotocols(Ferrieretal.,2020),possibly explainingtherelativepaucityoffunctionalconnectivitystudiesinthis majorpre-clinicalanimalmodel.OurresultsindicatethatfUSimagingin awakemiceisabletoreliablydetectbothfunctionalconnectivityatthe resting-state,suchasthecharacteristicbi-lateralconnectivitypatternsin thehippocampusandinthecortexinbaselineacquisitions(Chuangand Nasrallah,2017),andthepharmacologically-inducedchangesofthese patterns.

Ourresultsindicatethatscopolamine-treatmentleadstoenhanced functionalconnectivitybetweenthehippocampusandcorticalregions, asindicatedfirstqualitativelybothbyseed-basedandROI-based con-nectivityanalysisandconfirmedbytheSVMlinearmodel.According to(Mash andPotter, 1986)and(Schwabetal., 1992),M1receptors arehighlyexpressedintheisocortexandhippocampuswhilehavinga lowexpressionlevelinthethalamus.Althoughscopolamineand mil-amelinearenotconsideredtobe specificof anymuscarinicreceptor subtypes(Falsafi etal.,2012),ourresultsseemtoindicateacorrelation betweenM1receptorsdistributionandchangeinfunctional connectiv-ity.Thisenhancedconnectivitywasvisibleafter10minutesof subcu-taneousscopolamineinjection,wasdose-dependent,couldberescued tobaselinevaluesbythemuscarinicacetylcholinereceptoragonist mil-ameline,andwasabsentaftertreatmentwiththeperipherallyrestricted scopolamineanalogue,methyl-scopolamine.Contrarytoprevious stud-ies (Méndez-Lópezetal.,2011),wedidnotfindsignificantthalamic changesintheleft-rightconnectivityafterscopolamineinjections.This differencecouldbe duetothesmallnumberofanimalsused forthe presentstudyortotheselectedcoronalplane.

Our resultsareincontradictionwiththeonlystudywhich inves-tigatedscopolamine-induced changesin rodentbrainfunctional con-nectivity,byusingBOLD-basedfMRI(Shahetal.,2015).Inthisstudy a generaldecrease offunctionalconnectivitywas reportedfollowing scopolaminetreatment.WhilebothfMRIandfUSdetectfunctional con-nectivitythroughtheneurovascularcoupling,importantexperimental differencesmayunderliethesecontradictoryfindings.First,whileBOLD fMRImeasureshemoglobinoxygenationasanindirectmeasureofblood flowchanges,fUSdirectlymeasurescerebralbloodvolumewithhigher spatio-temporalresolution, leadingtohighersensitivity(Boidoetal., 2019),asdiscussedpreviously(Osmanskietal.,2014). Second,Shah etalusedanaesthesia,whichwasshowntoaffectnegativelyfunctional connectivitystrength(ChuangandNasrallah,2017).Combined,these differences may leadtoa criticalloss of detectionsensitivity in the mousebrain,which,ascomparedtothelargerratbrain,presentsa chal-lengeforfMRIimaging.Arecentauthoritativereview(Chuangand Nas-rallah,2017)proposedthatinordertobeabletocomparephysiological conditionanddataqualityacrosslaboratories,theconsistencyin detect-ingthestereotypicalconnectivitypatterns,suchasthebi-lateral connec-tivityinthecortexorinthehippocampus,shouldbeassessed similarly totheapproachusedinhumanstudiestoevaluatedenoisingmethodsfor rsfMRI(Pruimetal.,2015).Notably,Shahetal.donotreport stereotyp-icalbi-lateralconnectivitypatterns,confirmingthepossibilityof insuffi-cientsensitivity.Incontrast,fUS-basedconnectivitydetectioninawake miceconsistentlydetectsbi-lateralconnectivityinthecortexorinthe hippocampus asshown byTiran etal.(ChuangandNasrallah,2017; Mechlingetal.,2014;Osmanskietal.,2014)andinthepresentstudy (Fig.2and Supplementary Fig.3 ),similarlytorecenthigh-sensitivity fMRIstudies(ChuangandNasrallah,2017;Mechlingetal.,2014).

What isthefunctionalrelevanceofourresults? Incorresponding humancorticalregions,earlyreportsusingpositronemission tomogra-phy(PET)indicatedhyperactivationfollowingscopolamineinjection, as indicated by elevated glucose consumption, aresult that the au-thorsconsideredasunexpected(Blinetal.,1995;Molchanetal.,1994). Concerningchangesinconnectivity,whereastopologicalsimilarity be-tweenhumanandmouseconnectivitynetworksdoesnotnecessary im-plyfunctionalsimilarity,itisnotable thatwhile humanfMRI analy-sisofscopolamine-inducedchangesinfunctionalconnectivityin corti-calregionsalsoshowsageneraltendencyforhigherconnectivity,with markeddecreaseinasubsetofregions(Chhatwaletal.,2019),inthe hippocampusscopolaminesignificantlyenhancedcortico-hippocampal low-frequency coherence,similar to ourresults (Wink etal., 2006). InadditiontofMRI-basedapproaches,humanelectroencephalography (EGG)studieshavealsoreportedscopolamine-inducedincreasesinthe power of low-frequency (delta andtheta) activity, while alpha and beta-frequency activity were reduced (Alvarez‐Jimenez et al., 2016; Ebertetal.,1998;Kikuchietal.,1999;Liem‐Moolenaaretal.,2011;

Sannitaetal.,1987).Notably,low-frequencydeltaandthetaEEG activ-itywasfoundtocorrelatewellwiththeresting-statefunctionalMRI signal (Marawar etal., 2017), further suggesting pathophysiological relevancefor our results,which show scopolamine-induced increase incortico-hippocampalfunctionalconnectivitypatterns.Notably,DFA (Detrended Fluctuation Analysis) of EEGsignals in humans hasalso shownincreasedlong-rangecorrelations,inallfrequencybands,after scopolamineinjection(Simpragaetal.,2017). Suchsimilarityin func-tionaldysconnectivity inresponse topharmacologicalmanipulations indicateimportanttranslationalpotential,enablingtheuseof animal modelsashighlyefficienttoolstoempiricallytesthypotheseson connec-tomeinvolvementinbraindysfunction(MpandO,2019).Importantly, thescopolaminescorecouldberevertedtonormalvaluesby milame-line,providingapharmacologicalrescueofapathologicalindex.Inthis context,thescopolaminescorevalidatedinourstudymaybeusedasa sensitivebiomarkertodetectdose-dependenteffectsofscopolamineon thebrain,indicatingimportantpotentialbothforexperimental pharma-cologyanddrugdevelopment.

Ifthisstudyprovestherelevancyoffunctionalultrasoundfor phar-macologicalstudies,weacknowledgethatthelimitationoftheimaging modalityintwodimensionscanbebottleneckfortheinvestigationof brainmechanismswithoutaprioriinformationonthestudiedpathways. Werecentlyintroducedaproof-of-conceptof3DfUSneuroimagingin therodentbrainbasedona1024channelsresearchscannerandamatrix arrayprobe(Rabutetal.,2019).Unfortunately,asoftoday,thismethod comeswithasignificantcomputationalcostandtypicallyyieldreduced sensitivityorframeratecomparedto2DfUS.However,webelievethat thedevelopmentofalwaysmoreperformantprocessorunitswillsoon enablehighlysensitive3DfUSimagingtomonitorbrainmechanisms inthethreedimensions.Theadventof3DfUSopensthepossibilityto recordneurovasculardynamicsofcompletebrainregionvolumes.

5. Conclusion

In conclusion, our study introduces fUS imaging, coupled to machine-learningbasedclassification,asahighlysensitiveandrobust tooltodetectpharmacologically-inducedchangesinbrain connectiv-itypatterns,withoutanesthesiabias.Potentiallyconfoundingvascular effectsarecontrolledbyreal-timebuilt-inmonitoringof globalbrain perfusion.Thepharmacologicalrelevanceofthepharmaco-fUSscoreis shownbyitsdose-dependencyandsensitivitytoantagonist pharmaco-logicalaction.Ourproof-of-conceptdemonstrationofrescuinga patho-logicalconnectivitypatternbyapharmacologicalagentopenstheway ofusingpharmaco-fUSasaphysiologicallyrelevantandclinically trans-latableread-out,bothforbasicneuroscienceandfordrugdiscovery.

Declaration of Competing Interest

T.D.,M.T.andZ.Lareco-foundersandshareholdersofIconeus com-panycommercializingultrasoundneuroimagingscanners.

CRediT authorship contribution statement

Claire Rabut: Datacuration,Software,Methodology,Writing- orig-inaldraft,Writing-review&editing. Jérémy Ferrier: Datacuration, Writing-review& editing,Methodology. Adrien Bertolo: Software, Writing-review&editing. Bruno Osmanski: Software,Writing- re-view&editing.Xavier Mousset: Datacuration,Software,Writing- re-view&editing. Sophie Pezet: Datacuration,Methodology,Software, Writing-review&editing. Thomas Deffieux: Methodology,Software, Writing-review&editing. Zsolt Lenkei: Supervision, Conceptualiza-tion,Methodology,Writing-review&editing. Mickaël Tanter: Super-vision,Conceptualization,Methodology,Writing-review&editing.

Acknowledgments

Theresearchleadingtotheseresultshasreceivedfundingfromthe EuropeanResearchCouncilundertheEuropeanUnion’sSeventh Frame-workProgramme(FP7/2007-2013)/ERCGrandAgreementNo. 339244-FUSIMAGINE, theLABEX WIFI (Laboratory of Excellencewithin the FrenchProgram‘InvestmentsfortheFuture’)underreference ANR-10-IDEX-0001-02PSL,theNationalAgencyforResearchundertheprogram ‘futureinvestments’withthereferenceANR-10-EQPX-15andthe Euro-pean programANR-15-HBPR-0004-02 FUSIMICEoftheHumanBrain Project.

Supplementary materials

Supplementarymaterialassociatedwiththisarticlecanbefound,in theonlineversion,atdoi:10.1016/j.neuroimage.2020.117231.

References

Alvarez ‐Jimenez, R., Groeneveld, G.J., Gerven, J.M.A.van, Goulooze, S.C., Baakman, A.C., Hay, J.L., Stevens, J., 2016. Model-based exposure-response analysis to quantify age related differences in the response to scopolamine in healthy subjects. Br. J. Clin. Pharmacol. 82, 1011–1021. https://doi.org/10.1111/bcp.13031 .

Andrews, J.S., Jansen, J.H.M., Linders, S., Princen, A., 1994. Effects of disrupting the cholinergic system on short-term spatial memory in rats. Psychopharmacology (Berl.) 115, 485–494. https://doi.org/10.1007/BF02245572 .

Blin, J. , Chase, T.N. , Piercey, M.F. , 1995. Do the effects of muscarinic receptor blockade on brain glucose consumption mimic the cortical and subcortical metabolic pattern of Alzheimer’s disease in normal volunteers ? In: Comar, D. (Ed.) PET for Drug Devel- opment and Evaluation, Developments in Nuclear Medicine. Springer, Netherlands, Dordrecht, pp. 123–132 .

Boido, D., Rungta, R.L., Osmanski, B.-F., Roche, M., Tsurugizawa, T., Bi- han, D.L., Ciobanu, L., Charpak, S., 2019. Mesoscopic and microscopic imaging of sensory responses in the same animal. Nat. Commun. 10, 1–13.

https://doi.org/10.1038/s41467-019-09082-4 .

Carmichael, O., Schwarz, A.J., Chatham, C.H., Scott, D., Turner, J.A., Upadhyay, J., Coim- bra, A., Goodman, J.A., Baumgartner, R., English, B.A., Apolzan, J.W., Shankapal, P., Hawkins, K.R., 2018. The role of fMRI in drug development. Drug Discov. Today 23, 333–348. https://doi.org/10.1016/j.drudis.2017.11.012 .

Cherry, S.R., 2001. Fundamentals of positron emission tomography and appli- cations in preclinical drug development. J. Clin. Pharmacol. 41, 482–491.

https://doi.org/10.1177/00912700122010357 .

Chhatwal, J.P., Schultz, A.P., Hedden, T., Boot, B.P., Wigman, S., Rentz, D., John- son, K.A., Sperling, R.A., 2019. Anticholinergic amnesia is mediated by alter- ations in human network connectivity architecture. Cereb. Cortex 29, 3445–3456.

https://doi.org/10.1093/cercor/bhy214 .

Chuang, K.-H., Nasrallah, F.A., 2017. Functional networks and network perturba- tions in rodents. NeuroImage 163, 419–436. https://doi.org/10.1016/j.neuroimage. 2017.09.038 .

Demené, C., Deffieux, T., Pernot, M., Osmanski, B.-F., Biran, V., Gennisson, J.-L., Sieu, L.-A., Bergel, A., Franqui, S., Correas, J.-M., Cohen, I., Baud, O., Tanter, M., 2015. Spatiotemporal clutter filtering of ultrafast ultrasound data highly increases doppler and fultrasound sensitivity. IEEE Trans. Med. Imaging 34, 2271–2285.

https://doi.org/10.1109/TMI.2015.2428634 .

Deng, G., Wu, C., Rong, X., Li, S., Ju, Z., Wang, Y., Ma, C., Ding, W., Guan, H., Cheng, X., Liu, W., Wang, C., 2019. Ameliorative effect of deoxyvasicine on scopolamine-induced cognitive dysfunction by restoration of cholinergic function in mice. Phytomedicine 63, 153007. https://doi.org/10.1016/j.phymed.2019.153007 .

Deutsch, J.A., 1971. The cholinergic synapse and the site of memory. Science 174, 788– 794. https://doi.org/10.1126/science.174.4011.788 .

Dizeux, A., Gesnik, M., Ahnine, H., Blaize, K., Arcizet, F., Picaud, S., Sahel, J.-A., Deffieux, T., Pouget, P., Tanter, M., 2019. Functional ultrasound imaging of the brain reveals propagation of task-related brain activity in behaving primates. Nat. Commun. 10, 1–9. https://doi.org/10.1038/s41467-019-09349-w .

Ebert, U., Siepmann, M., Oertel, R., Wesnes, K.A., Kirch, W., 1998. Pharmacokinetics and pharmacodynamics of scopolamine after subcutaneous administration. J. Clin. Phar- macol. 38, 720–726. https://doi.org/10.1002/j.1552-4604.1998.tb04812.x .

Falsafi, S.K., Deli, A., Höger, H., Pollak, A., Lubec, G., 2012. Scopolamine adminis- tration modulates muscarinic, nicotinic and nmda receptor systems. PLoS ONE 7.

https://doi.org/10.1371/journal.pone.0032082 .

Ferrier, J., Tiran, E., Deffieux, T., Tanter, M., Lenkei, Z., 2020. Functional imag- ing evidence for task-induced deactivation and disconnection of a major de- fault mode network hub in the mouse brain. Proc. Natl. Acad. Sci.. 201920475

https://doi.org/10.1073/pnas.1920475117 .

Guo, C., Shen, J., Meng, Z., Yang, X., Li, F., 2016. Neuroprotective effects of polygalacic acid on scopolamine-induced memory deficits in mice. Phytomedicine 23, 149–155.

https://doi.org/10.1016/j.phymed.2015.12.009 .

Haider, S., Tabassum, S., Perveen, T., 2016. Scopolamine-induced greater alterations in neurochemical profile and increased oxidative stress demonstrated a bet- ter model of dementia: a comparative study. Brain Res. Bull. 127, 234–247.

C. Rabut, J. Ferrier and A. Bertolo et al. NeuroImage 222 (2020) 117231 Halldin, C., Gulyas, B., Farde, L., 2001. PET studies with carbon-11 radioligands in

neuropsychopharmacological drug development. Curr. Pharm. Des. 7, 1907–1929.

https://doi.org/10.2174/1381612013396871 .

Ketih Franlkin, G.P. , 2011. Paxinos and Franklin’s the Mouse Brain in Stereotaxic Coordi- nates. Academic Press .

Kikuchi, M., Wada, Y., Nanbu, Y., Nakajima, A., Tachibana, H., Takeda, T., Hashimoto, T., 1999. EEG changes following scopolamine administration in healthy subjects. Neu- ropsychobiology 39, 219–226. https://doi.org/10.1159/000026588 .

Klinkenberg, I., Blokland, A., 2010. The validity of scopolamine as a pharmacological model for cognitive impairment: A review of animal behavioral studies. Neurosci. Biobehav. Rev. 34, 1307–1350. https://doi.org/10.1016/j.neubiorev.2010.04.001 .

Lee, J.-C., Park, J.H., Ahn, J.H., Park, J., Kim, I.H., Cho, J.H., Shin, B.N., Lee, T.-K., Kim, H., Song, M., Cho, G.-S., Kim, D.W., Kang, I.J., Kim, Y.-M., Won, M.-H., Choi, S.Y., 2018. Effects of chronic scopolamine treatment on cognitive impairment and neu- rofilament expression in the mouse hippocampus. Mol. Med. Rep. 17, 1625–1632.

https://doi.org/10.3892/mmr.2017.8082 .

Lee, J.-S., Hong, S.-S., Kim, H.-G., Lee, H.-W., Kim, W.-Y., Lee, S.-K., Son, C.-G., 2016. Gongjin-Dan enhances hippocampal memory in a mouse model of scopolamine- induced amnesia. PLoS ONE 11. https://doi.org/10.1371/journal.pone.0159823 .

Liem ‐Moolenaar, M., Boer, P.de, Timmers, M., Schoemaker, R.C., Hasselt, J.G.C.van, Schmidt, S., Gerven, J.M.A.van, 2011. Pharmacokinetic–pharmacodynamic relation- ships of central nervous system effects of scopolamine in healthy subjects. Br. J. Clin. Pharmacol. 71, 886–898. https://doi.org/10.1111/j.1365-2125.2011.03936.x .

Lin, J., Huang, L., Yu, J., Xiang, S., Wang, J., Zhang, J., Yan, X., Cui, W., He, S., Wang, Q., 2016. Fucoxanthin, a marine carotenoid, reverses scopolamine-induced cognitive im- pairments in mice and inhibits acetylcholinesterase in vitro. Mar. Drugs 14, 67.

https://doi.org/10.3390/md14040067 .

Macé, E., Montaldo, G., Cohen, I., Baulac, M., Fink, M., Tanter, M., 2011. Functional ultrasound imaging of the brain. Nat. Methods 8, 662–664.

https://doi.org/10.1038/nmeth.1641 .

Mace, E., Montaldo, G., Osmanski, B.-F., Cohen, I., Fink, M., Tanter, M., 2013. Functional ultrasound imaging of the brain: theory and basic principles. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 60, 492–506. https://doi.org/10.1109/TUFFC.2013.2592 .

Marawar, R.A., Yeh, H.J., Carnabatu, C.J., Stern, J.M., 2017. Functional MRI correlates of resting-state temporal theta and delta eeg rhythms. J. Clin. Neurophysiol. 34, 69–76.

https://doi.org/10.1097/WNP.0000000000000309 .

Mash, D.C., Potter, L.T., 1986. Autoradiographic localization of M1 and M2 muscarine receptors in the rat brain. Neuroscience 19, 551–564.

https://doi.org/10.1016/0306-4522(86)90280-0 .

Mechling, A.E., Hübner, N.S., Lee, H.-L., Hennig, J., von Elverfeldt, D., Harsan, L.-A., 2014. Fine-grained mapping of mouse brain functional connectivity with resting-state fMRI. NeuroImage 96, 203–215. https://doi.org/10.1016/j.neuroimage.2014.03.078 .

Méndez-López, M., Méndez, M., López, L., Arias, J.L., 2011. Memory per- formance and scopolamine: hypoactivity of the thalamus revealed by cytochrome oxidase histochemistry. Acta Histochem 113, 465–471.

https://doi.org/10.1016/j.acthis.2010.04.004 .

Molchan, S.E., Matochik, J.A., Zametkin, A.J., Szymanski, H.V., Cantillon, M., Cohen, R.M., Sunderland, T., 1994. A double FDG/PET study of the ef- fects of scopolamine in older adults. Neuropsychopharmacology 10, 191–198.

https://doi.org/10.1038/npp.1994.21 .

Mp, van den, H., O, S., 2019. A cross-disorder connectome landscape of brain dysconnectivity. Nat. Rev. Neurosci. 20, 435–446. https://doi.org/10.1038/ s41583-019-0177-6 .

Muhammad, T., Ali, T., Ikram, M., Khan, A., Alam, S.I., Kim, M.O., 2019. Melatonin res- cue oxidative stress-mediated neuroinflammation/ neurodegeneration and memory impairment in scopolamine-induced amnesia mice model. J. Neuroimmune Pharma- col. 14, 278–294. https://doi.org/10.1007/s11481-018-9824-3 .

Nathan, P.J., Phan, K.L., Harmer, C.J., Mehta, M.A., Bullmore, E.T., 2014. Increasing phar- macological knowledge about human neurological and psychiatric disorders through functional neuroimaging and its application in drug discovery. Curr. Opin. Pharma- col., Neurosciences 14, 54–61. https://doi.org/10.1016/j.coph.2013.11.009 .

Osmanski, B.-F., Pezet, S., Ricobaraza, A., Lenkei, Z., Tanter, M., 2014. Functional ultra- sound imaging of intrinsic connectivity in the living rat brain with high spatiotemporal resolution. Nat. Commun. 5, 1–14. https://doi.org/10.1038/ncomms6023 .

Park, H.R., Lee, H., Park, H., Cho, W.-K., Ma, J.Y., 2016. Fermented sipjeondaebo-tang alle- viates memory deficits and loss of hippocampal neurogenesis in scopolamine-induced amnesia in mice. Sci. Rep. 6, 22405. https://doi.org/10.1038/srep22405 .

Piel, M., Vernaleken, I., Rösch, F., 2014. Positron emission tomography in CNS drug discovery and drug monitoring. J. Med. Chem. 57, 9232–9258.

https://doi.org/10.1021/jm5001858 .

Pruim, R.H.R., Mennes, M., Buitelaar, J.K., Beckmann, C.F., 2015. Evaluation of ICA- AROMA and alternative strategies for motion artifact removal in resting state fMRI. NeuroImage 112, 278–287. https://doi.org/10.1016/j.neuroimage.2015.02.063 .

Rabut, C., Correia, M., Finel, V., Pezet, S., Pernot, M., Deffieux, T., Tanter, M., 2019. 4D functional ultrasound imaging of whole-brain activity in rodents. Nat. Methods 16, 994–997. https://doi.org/10.1038/s41592-019-0572-y .

Sannita, W.G., Maggi, L., Rosadini, G., 1987. Effects of scopolamine (0.25–0.75 mg i.m.) on the quantitative EEG and the neuropsychological status of healthy volunteers. Neu- ropsychobiology 17, 199–205. https://doi.org/10.1159/000118365 .

Schwab, C., Brückner, G., Rothe, T., Castellano, C., Oliverio, A., 1992. Au- toradiography of muscarinic cholinergic receptors in cortical and subcortical brain regions of C57BL/6 and DBA/2 mice. Neurochem. Res. 17, 1057–1062.

https://doi.org/10.1007/BF00967281 .

Shah, D., Blockx, I., Guns, P.-J., De Deyn, P.P., Van Dam, D., Jonckers, E., Delgado y Pala- cios, R., Verhoye, M., Van der Linden, A., 2015. Acute modulation of the cholinergic system in the mouse brain detected by pharmacological resting-state functional MRI. NeuroImage 109, 151–159. https://doi.org/10.1016/j.neuroimage.2015.01.009 .

Sieu, L.-A., Bergel, A., Tiran, E., Deffieux, T., Pernot, M., Gennisson, J.-L., Tanter, M., Cohen, I., 2015. EEG and functional ultrasound imaging in mobile rats. Nat. Methods 12, 831–834. https://doi.org/10.1038/nmeth.3506 .

Simpraga, S., Alvarez-Jimenez, R., Mansvelder, H.D., Gerven, J.M.A.van, Groeneveld, G.J., Poil, S.-S., Linkenkaer-Hansen, K., 2017. EEG machine learning for accurate de- tection of cholinergic intervention and Alzheimer’s disease. Sci. Rep. 7, 1–11.

https://doi.org/10.1038/s41598-017-06165-4 .

Sperling, R., Greve, D., Dale, A., Killiany, R., Holmes, J., Rosas, H.D., Cocchiarella, A., Firth, P., Rosen, B., Lake, S., Lange, N., Routledge, C., Albert, M., 2002. Functional MRI detection of pharmacologically induced memory impairment. Proc. Natl. Acad. Sci. 99, 455–460. https://doi.org/10.1073/pnas.012467899 .

Tiran, E., Ferrier, J., Deffieux, T., Gennisson, J.-L., Pezet, S., Lenkei, Z., Tanter, M., 2017. Transcranial functional ultrasound imaging in freely moving awake mice and anes- thetized young rats without contrast agent. Ultrasound Med. Biol. 43, 1679–1689.

https://doi.org/10.1016/j.ultrasmedbio.2017.03.011 .

Wandschneider, B., Koepp, M.J., 2016. Pharmaco fMRI: Determining the func- tional anatomy of the effects of medication. NeuroImage Clin 12, 691–697.

https://doi.org/10.1016/j.nicl.2016.10.002 .

Wink, A.M., Bernard, F., Salvador, R., Bullmore, E., Suckling, J., 2006. Age and cholinergic effects on hemodynamics and functional coherence of human hippocampus. Neuro- biol. Aging 27, 1395–1404. https://doi.org/10.1016/j.neurobiolaging.2005.08.011 .

Xu, H., You, Z., Wu, Z., Zhou, L., Shen, J., Gu, Z., 2016. WY14643 Attenuates the Scopolamine-Induced Memory Impairments in Mice. Neurochem. Res. 41, 2868– 2879. https://doi.org/10.1007/s11064-016-2002-1 .