Behavioral/Systems/Cognitive
HighlySelectiveReceptiveFieldsinMouseVisualCortex
CristopherM.NiellandMichaelP.Stryker
W.M.KeckFoundationCenterforIntegrativeNeuroscience,DepartmentofPhysiology,UniversityofCalifornia,SanFrancisco,SanFrancisco,California94143-0444
Geneticmethodsavailableinmicearelikelytobepowerfultoolsindissectingcorticalcircuits.However,thevisualcortex,inwhichsensorycodinghasbeenmostthoroughlystudiedinotherspecies,hasessentiallybeenneglectedinmiceperhapsbecauseoftheirpoorspatialacuityandthelackofcolumnarorganizationsuchasorientationmaps.WehavenowappliedquantitativemethodstocharacterizevisualreceptivefieldsinmouseprimaryvisualcortexV1bymakingextracellularrecordingswithsiliconelectrodearraysinanesthetizedmice.Weusedcurrentsourcedensityanalysistodeterminelaminarlocationandspikewaveformstodiscriminateputativeexcitatoryandinhibitoryunits.Wefindthat,althoughthespatialscaleofmousereceptivefieldsisuptooneortwoordersofmagnitudelarger,neuronsshowselectivityforstimulusparameterssuchasorientationandspatialfrequencythatisneartothatfoundinotherspecies.Further-more,typicalresponsepropertiessuchaslinearversusnonlinearspatialsummation(i.e.,simpleandcomplexcells)andcontrast-invarianttuningarealsopresentinmouseV1andcorrelatewithlaminarpositionandcelltype.Interestingly,wefindthatputativeinhibitoryneuronsgenerallyhavelessselective,andnonlinear,responses.Thisquantitativedescriptionofreceptivefieldpropertiesshouldfacilitatetheuseofmousevisualcortexasasystemtoaddresslongstandingquestionsofvisualneuroscienceandcorticalprocessing.
Keywords:visualcortex;receptivefield;mouse;orientation;spatialfrequency;contrast-invarianttuning
Introduction
Overthepastnearlyhalfcenturysincevisualresponseswerefirstdescribedinthemammalianvisualcortex(HubelandWiesel,1962),therehasbeenintensiveresearchintotheneuralcircuitanddevelopmentalmechanismsthatgiverisetoselectiverecep-tivefield(RF)properties.However,althoughthedescriptionofvisualencodinghasbecomeincreasinglymorequantitative(Ringach,2004;Carandinietal.,2005),ashastheanatomyofneuronalsubtypes(Gilbert,1983;DouglasandMartin,2004),ithasbeendifficulttolinkthesefunctionalandanatomicalfindingsintoalocalcircuitmodelofcorticalprocessing.Evenmoreques-tionsremainabouthowsuchacircuitmightbeassembledduringdevelopment,despiteadvancesinunderstandingthemolecularmechanismsinvolved(Waitesetal.,2005;RashandGrove,2006;Polleuxetal.,2007)andtheconsequencesofalteredsensoryinput(Hensch,2005;Hoferetal.,2006).
Therecentproliferationofgenetictechnologyinmicemayprovidetoolstoanswermanyofthesequestions(Callaway,2005).Targetedgenedisruptionandtransgeneexpressioncanresultinmuchmorespecificmanipulationsthanhavebeenpos-sibleviapharmacologyorsensoryalterationandcanallowper-ReceivedFeb.11,2008;revisedApril23,2008;acceptedJune12,2008.
ThisworkwassupportedbyNationalInstitutesofHealthGrantEY02874(M.P.S.)andaHelenHayWhitneyFoundationfellowship(C.M.N.).WethankDrs.JonathanHorton,SteveLisberger,andLindaWilbrechtandmembersoftheStrykerlaboratoryforcommentsonthismanuscriptandhelpfuldiscussions.WealsothankDrs.MichaelLewickiandEizaburoDoiforprovidingroutinestofitGaborfunctionsandDr.DarioRingachfordataonGaborfitsandcircularvariancefrommacaque.
CorrespondenceshouldbeaddressedtoMichaelP.Stryker,DepartmentofPhysiology,513ParnassusAvenue,RoomHSE-802,UniversityofCalifornia,SanFrancisco,SanFrancisco,CA94143-0444.E-mail:stryker@phy.ucsf.edu.DOI:10.1523/JNEUROSCI.0623-08.2008
Copyright©2008SocietyforNeuroscience0270-6474/08/287520-17$15.00/0turbationofcellularsignaling(Huangetal.,1999;Karpovaetal.,2005),synapticplasticity(Zengetal.,2001;Sawtelletal.,2003),andfiringpatterns(Tanetal.,2006;Zhangetal.,2007a),evenatthesingle-celllevel(Brechtetal.,2004).Furthermore,fluorescentproteinlabelingprovidespreciseanatomicaltechniquestoyieldinformationaboutcelltype(Fengetal.,2000;Tamamakietal.,2003)orevensynapticconnectivity(Wickershametal.,2007)intheintactbrain,bridgingthegapbetweencircuitstructureandfunction.Indeed,studieshavealreadybeguntotakeadvantageofgeneticmethodstostudyvisualsystemdevelopment(Fagiolinietal.,2003;Cangetal.,2005),plasticity(Fagiolinietal.,2004;Sykenetal.,2006),andfunction(Sohyaetal.,2007).
Studiesofcorticalvisualprocessinghavetypicallyusedcarni-voresorprimates,whichareconsideredtohaveamorerefinedvisualsystem,includingamuchlargercorticalregionforvisualprocessing,higheracuity,extensivevisualbehaviors,andorien-tation,oculardominance,andspatialfrequencycolumns(Issaetal.,2000;OhkiandReid,2007;VanHooser,2007).Understand-ingvisualprocessinginsuchasimplesystemasthemousecortex,whichlacksbothfine-scalespatialacuityandmapssuchasori-entationcolumns,shouldprovideinsightintotheminimalmechanismsnecessaryforreceptivefielddevelopmentandfunc-tion.However,althoughrecentstudiesinratandsquirrelhavedescribedreceptivefieldpropertiesintheabsenceoforientationmaps(Fagiolinietal.,1994;Girmanetal.,1999;Heimeletal.,2005;Ohkietal.,2005),muchlessisknownaboutvisualencod-ingandprocessinginmice.Severalearlyelectrophysiologystud-ies(Dra¨ger,1975;ManginiandPearlman,1980;Metinetal.,1988)(forreview,seeHubener,2003)indicatedthat,althoughmousecorticalneuronscanbeclassifiedintocategoriessimilartothosedescribedinotherspecies,theiroveralllevelofstimulus
NiellandStryker•ReceptiveFieldsinMouseVisualCortexselectivitymaybesignificantlylessthanthatofotherspecies.However,thesestudiesaredifficulttoevaluatebecausetheywereperformedbeforemanyoftherecentquantitativetechniquesforreceptivefieldcharacterizationweredeveloped.Arecentstudy,usingtwo-photonimagingofcalciumsignals,measureddiffer-encesinorientationselectivityininhibitoryandexcitatoryneu-ronsofmouseprimaryvisualcortexV1(Sohyaetal.,2007)butdidnotperformathoroughcharacterizationofothervisualre-sponsesandselectivityandwasrestrictedtoimagingthesuperfi-ciallayers.
WethereforeundertookaquantitativesurveyofreceptivefieldpropertiesinV1oftheanesthetizedmouse.Inadditiontodeterminingthetypesofstimuliandrangeofstimulusparame-tersthatareappropriateforprobingvisioninmice,wesoughttoconfirmthatthebasicpropertiesofvisualprocessingthathavebeenstudiedinotherspeciesarepresentinmice.Theseincludeorientationandspatialfrequencytuning,thepresenceofbothsimpleandcomplexresponsetypes,andhigher-orderpropertiessuchascontrast-invarianttuning.Furthermore,torelatethesefunctionalpropertiestocomputationinthecorticalcircuit,weanalyzedresponsesbytheirlaminarlocationwithincortexandbyputativeidentityasinhibitoryversusexcitatoryneuronsbasedonspikewaveform.
MaterialsandMethods
Invivophysiology.RecordingsweremadefromadultC57BL/6mice,2–6monthsofage.TheanimalsweremaintainedintheanimalfacilityatUniversityofCalifornia,SanFrancisco(UCSF)andusedinaccordancewithprotocolsapprovedbytheUCSFInstitutionalAnimalCareandUseCommittee.Forsurgery,miceweresedatedwithanintraperitonealin-jectionofchlorprothixene(5mg/kg)andthenanesthetizedwithure-thane(0.5–1.0g/kg,i.p.,at10%w/vinsaline).Administrationofchlor-prothixeneseveralminutesbeforeurethanegreatlyreducedthedosageofurethanenecessarytoinducesurgicalanesthesia.Additionally,atropine(0.3mg/kg)anddexamethasone(2mg/kg)wereadministeredsubcuta-neouslytoreducesecretionsandedema,respectively.Theanimalwasmaintainedat37.5°Cbyafeedback-controlledheatingpad.Atracheot-omywasperformed,andasmallglasscapillarytubewasinsertedtomaintainafreeairway.Afterretractingthescalp,weperformedasmallcraniotomy,ϳ1mmindiameter,andnickedaslitintheduratoallowinsertionofthemultisiteelectrode.Theexposedcorticalsurfacewascoveredwith2.5%agaroseinextracellularsaline(inmM:125NaCl,5KCl,10glucose,10HEPES,and2CaClmechanicalsupport.Theelectrode2,pH7.4)topreventdryingandprovidewasloweredintothebrainthroughtheagarosetoanappropriatedepthandwasallowedtosettlefor30–45minbeforethebeginningofrecording.Theelectrodewasplacedwithoutregardforthepresenceofvisuallyresponsiveunits,andallunitsstablyisolatedovertherecordingperiodwereincluded.Forsuperficialrecordings,theelectrodewasofteninsertedatanangle,toincreasethedistancebetweentheinsertionandrecordingsites.Theeyeswerecoveredwithophthalmiclubricantointmentuntilrecording,atwhichtimetheeyeswererinsedwithsalineandathinlayerofsiliconeoil(30,000cen-tistokes)wasappliedtopreventdryingwhileallowingclearopticaltrans-mission.Wedidnotinducecycloplegia,andtherestingpupildiameterwasϳ1mm.Werecordedonlyatsiteswithreceptivefieldslocatedatleast20°lateraltothevisualmeridian,toavoidconfoundingeffectsat-tributabletothebinocularzoneofvision.
Attheendofrecording,theanimalwaskilledbyoverdoseofbarbitu-rates.Forhistology,theanimalwasintracardiallyperfusedwith4%para-formaldehydeinPBS,andthebrainwassectionedcoronallyat100mwithavibratome(Lancer).Sectionswereincubatedforseveralhoursin0.1g/ml4Ј,6-diamidino-2-phenylindole(DAPI;Sigma-Aldrich)tostainnucleiandimagedonanepifluorescencemicroscope.
RecordingsweremadewithsiliconmicroprobesfromNeuroNexusTechnologies.Twoconfigurationswereused:alinearprobewith16sitesspacedat50mintervals(modela1x16-3mm50-177),whichcouldbe
J.Neurosci.,July23,2008•28(30):7520–7536•7521
usedtospanacrossmultiplelayersofcortex;andatetrodeconfiguration,withfourtetrodeclusters,eachconsistingoffoursitesseparatedby25monaside(modela2x2-tet-3mm-150-121),whichwasusedprimarilytoprovidebetterisolationofunitsinlayers2/3and4.Approximatelytwo-thirdsofallunitswererecordedwiththetetrodeconfiguration.Theshanksoftheprobeswere15mthickand3mmlong,withamaximumwidthatthetopoftheshankof94m(tetrode)or123m(linear.)Forexperimentsfollowedbyhistologytoreconstructpenetrations,theelec-trodewascoatedwithasmallamountofthelipophilicvitaldyeDiI(Invitrogen).SignalswereacquiredusingaSystem3workstation(Tucker-DavisTechnologies)andanalyzedwithcustomsoftwareinMatlab(MathWorks).Forlocalfieldpotential(LFP)recording,theex-tracellularsignalwasfilteredfrom1to300Hzandsampledat1.5kHz.Currentsourcedensity(CSD)wascomputedfromtheaverageLFPbytakingthediscretesecondderivativeacrosstheelectrodesites,attwo-sitespacingtoreducenoise,andinterpolatedtoproduceasmoothCSDmap.Forsingle-unitrecording,theextracellularsignalwasfilteredfrom0.7to7kHzandsampledat25kHz.Spikingeventsweredetectedon-linebyvoltagethresholdcrossing,anda1mswaveformsamplewasacquiredaroundthetimeofthresholdcrossing.Toimproveisolationofsingleunits,recordingsfromgroupsoffourneighboringsiteswerelinked,sothatawaveformwasacquiredonallfoursitesinresponsetoathresholdcrossingonanyofthefour.Thisprocedurewasusedforboththetetrodeandlinearconfigurationelectrodes;inthelattercase,sites1–4,5–8,9–12,and13–16(inorderalongtheshank)weregroupedtogether.Inbothcases,this“virtualtetrode”acquisitionhadtwoprimarybenefits:improveddiscriminabilitywhenawaveformappearedonmorethanonesite,andcommon-modenoiserejectionofsignalssharedonallfoursites.Whereasthelargeramplitudespikesoflayer5andlayer6unitsweresometimesrecordedsimultaneouslyonadjacentsitesatthe50mspac-ingofthelinearelectrode,layer2/3neuronsoftenappearedpredomi-nantlyononesite,evenatthe25mspacingofthetetrodeconfigura-tion.Inbothcases,manyunitshadsignalsonnondominantsitesthatwerebelowthevoltagetriggerthreshold;however,thesimultaneousac-quisitionallowedthislow-amplitudeinformationtobeintegratedtoimprovediscriminability.
Theindividualwaveformsampleswerealignedbytheirmostnegativetimepoint.Toidentifysingleunits,thespikewaveformsfromthefoursitestogetherwereparameterizedby6–10independentcomponentsus-ingtheFastICApackageforMatlab(http://www.cis.hut.fi/projects/ica/fastica/)andclusteredbyamixture-of-GaussiansmodelusingKlusta-Kwik(Harrisetal.,2000).Usingindependentcomponents,ratherthanprincipalcomponents,allowedustotakeadvantageofcommon-modenoiserejectionfromwaveformssimultaneouslyacquiredacrossmultiplesitesinthevirtualtetrodeconfiguration.Whereasindependentcompo-nentsanalysis(ICA)hasbeenusedpreviouslytofilterthecontinuousdataacrosschannels(Snellingsetal.,2006),weinsteadperformedICAontheindividualwaveforms,tointegratewaveformparameterizationandnoiserejection.QualityofseparationwasdeterminedbasedontheMahalanobisdistanceandL-ratio(Schmitzer-Torbertetal.,2005)andthepresenceofaclearrefractoryperiod.
Unitswerethenclassifiedasnarroworbroadspikingbasedonprop-ertiesoftheiraveragewaveforms,attheelectrodesitewithlargestampli-tude.Threeparameterswereusedfordiscrimination(Fig.1F,G):theheightofthepositivepeakrelativetotheinitialnegativetrough,thetimefromtheminimumoftheinitialtroughtomaximumofthefollowingpeak,andtheslopeofthewaveform0.5msaftertheinitialtrough.Thisthirdmeasureprovidedaproxyforthetotaldurationoftheslowerpositivepeak,becauseourwaveformsamplingwasnotofsufficientdu-rationtomeasuretheentirereturntobaseline.Twolinearlyseparableclusterswerefound,correspondingtonarrow-spiking(putativeinhibi-tory)andbroad-spiking(putativeexcitatory)neurons.Theseclusterswereseparatedidenticallybybothk-meansandlinkageclustering.Unitclassificationwasstableanddidnotchangewhenonlytwoofthethreemeasurementswereused.Thewidthoftheinitialtrough,asusedprevi-ously(BrunoandSimons,2002),didnotgivegoodseparation,whichmaybeattributabletofilteringbytheelectrodesoracquisitionsystembecausethisistheshortesttimescaleinthewaveform,ortoitssensitivitytothealignmentofindividualwaveformsincomputingtheaverage
7522•J.Neurosci.,July23,2008•28(30):7520–7536NiellandStryker•ReceptiveFieldsinMouseVisualCortex
waveform.Becauseofthesmallnumbersofnarrow-spikingunitsineachlayer,incasesinwhichnosignificantdifferencewasseenacrosstheindividuallayers,wepooltogetherdatafornarrow-spikingunitsfromalllayersintoonecategoryforpresentation.
Visualstimuli.Achallengeinvolvedinourapproachofunbiasedrecordingfromanumberofneuronssimultaneouslyisthatstimulican-notbetailoredtoindividualneurons.Forex-ample,itismuchfastertomeasurespatialfre-quencytuningofasingleunitbyfindingtheoptimalorientationandthenvaryingthespatialfrequencyonlyatthisorientationthanbypre-sentingallorientationsandallspatialfrequen-cies,asinthisstudy.Thisalsolimitedourabilitytovarymultipleparametersinasinglestimulusset:forexample,tomeasurecontrast-invarianttuning,wechoseonespatialfrequencyandthenvariedorientationandcontrast,becausethecurseofdimensionalitymakesitimpracticaltosamplefinelyacrossallthreeparameters.Thus,theresponsewaslimitedtothoseneuronstunedtothespatialfrequencywechose.Thiswillbecomeevenmoreofalimitationinnewtechniques,suchastwo-photoncalciumimag-ing,whichcansamplelargepopulationsbuthavepoorertemporalresolution.Stimulisimi-lartothecontrast-modulatednoisemoviesde-scribedbelow,whichcanproviderapidmea-suresofresponsivenessacrossawidelytunedpopulation,mayhelptoaddressthesechallenges.
StimuliweregeneratedinMatlabusingthePsychophysicsToolboxextensions(Brainard,1997;Pelli,1997)anddisplayedwithgammacorrectiononamonitor(NanaoFlexscan,30ϫ40cm,60Hzrefreshrate,32cd/m2meanlumi-nance)placed25cmfromthemouse,subtend-ingϳ60ϫ75°ofvisualspace.Episodicstimuli
wererepeatedfivetoseventimes,withstimulusFigure1.Laminarlocationandspikewaveformclassification.A,Schematicoflinearmultisiteprobe.B,AverageLFPresponsesconditionsrandomlyinterleaved,andagrayfor16sitesthroughthedepthofcortex.Arrowsshowconsecutivepeaksofthehigh-frequencyoscillation.C,CSDanalysisoftracesblankcondition(meanluminance)wasin-inBdemonstratessegregationofresponsesbylayer.Bluerepresentscurrentsinks,andredrepresentscurrentsources.Becausecludedinallstimulussetstoestimatethespon-theCSDisbasedonasecondderivativeattwo-sitespacing,theCSDcannotbecomputedforthetoptwoandbottomtwosites,so
itspans550mratherthan750m.UnitsforCSDarenormalizedfromϪ1to1.D,Averagespikewaveformsforallunitstaneousfiringrate.
Episodicstimuliincludeddriftingsinusoidalanalyzed,alignedtominimumandnormalizedbytroughdepth,demonstratingnarrow-spiking(blue;nϭ45)andbroad-spikinggratings[1.5sduration,temporalfrequencyof(green;nϭ186)units.E,Averageofallwaveformsfornarrow-spikingandbroad-spikingunits.F,G,Scatterplotofspike2Hz,12directions,spatialfrequencyof0.01,waveformparametersforallunits.0.02,0.04,0.08,0.16,0.32,and0cycles/°(cpd),
1997),inthatbothexplicitlyrestricttheregionoffrequencyspacethatisi.e.,full-fieldflicker],full-lengthdriftingbars(widthof5°,velocityof
sampled.Toprovidecontrastmodulation,thismoviewasmultipliedby30°/s,16directions),driftingshortbars(4ϫ8°,30°/s,fourdirections),
asinusoidallyvaryingcontrast.Moviesweregeneratedat60ϫ60pixelscontrast-reversing(counterphase)sinusoidalgratings(0.04cpd,sinu-andthensmoothlyinterpolatedto480ϫ480pixelsbythevideocardtosoidallyreversingattemporalfrequencies1,2,4,and8Hz,eightorien-tations,2sduration),andcontrast-reversingsquarecheckerboard(0.04appearat60ϫ60°onthemonitorandplayedat30framespersecond.
cpd,square-wavereversingat0.5Hz).EpisodicstimuliwereshownatEachmoviewas5minlongandwasrepeatedtwotothreetimes,for100%contrastwiththebackgroundatmeanluminance,exceptina10–15mintotalpresentation.Abriefclipofthecontrast-modulatesubsetofexperimentsoncontrast-invarianttuning,inwhichdriftingmovieisavailableassupplementaldata(availableatwww.jneurosci.orgsinusoidalgratingswerepresentedasdescribedabovebutatfixedspatialassupplementalmaterial).frequencyof0.04cpdandcontrast6.25,12.5,25,50,and100%.Dataanalysis.Theaveragespontaneousrateforeachunitwascalcu-Gaussiannoisemovieswerecreatedbyfirstgeneratingarandomspa-latedbyaveragingtherateoverallblankconditionpresentations.FortiotemporalfrequencyspectrumintheFourierdomainwithdefineddriftinggratings,responsesateachorientationandspatialfrequencyspectralcharacteristics.Todriveasmanysimultaneouslyrecordedunitswerecalculatedbyaveragingthespikerateduringthe1.5spresentationaspossible,weusedaspatialfrequencyspectrumthatdroppedoffasandsubtractingthespontaneousrate.Thepreferredorientationwasde-A(f)ϳ1/(fϩfc),withfcϭ0.05cpd,andasharpcutoffat0.12cpd,toterminedbyaveragingtheresponseacrossallspatialfrequenciesandapproximatelymatchthestimulusenergytothedistributionofspatial
calculatinghalfthecomplexphaseofthevalue
frequencypreferences.Thetemporalfrequencyspectrumwasflatwithasharplow-passcutoffat4Hz.Thisthree-dimensional(x,y,t)spec-¥F͑͒e2itrumwastheninvertedtogenerateaspatiotemporalmovie.Thisstimu-Sϭ.¥F͑͒lusisrelatedtothesubspacereversecorrelationmethod(Ringachetal.,
NiellandStryker•ReceptiveFieldsinMouseVisualCortexTheorientationtuningcurvewasconstructedforthespatialfrequencythatgavepeakresponseatthisorientation.Giventhisfixedpreferredorientationpref,thetuningcurvewasfittedasthesumoftwoGaussianscenteredonprefandwidth,withaconstantprefϩ,ofdifferentamplitudesAbaselineB.Fromthisfit,wecalculated1andA2butequaltwometrics:anorientationselectivityindex(OSI)representingtheratioofthetunedversusuntunedcomponentoftheresponse,andthewidthofthetunedcomponent.OSIwascalculatedasthedepthofmodulationfromthepreferredorientationtoitsorthogonalorientationϩ/2,as(RorthoϭprefprefϪRortho)/(RtheprefϩRfitaboveortho).Tuningwidthwasthehalf-widthathalf-maximumofthebaseline,Rcalculatedfromthefittedfunctionortho.Inaddition,directionselectivitywasas(RRprefϪRopposite)/(WeusedprefϩRthesemeasuresopposite).
ofselectivity,ratherthanthecircularvari-ance,becausethecircularvariance,whichisasingleglobalmeasureofthetuningcurve,combinesaspectsofbothdepthofmodulationandtuningwidthintoonevalueandthusdoesnotgiveasintuitiveadescriptionofthetuningcurve(butseeRingachetal.,2002forathoroughexposition).Furthermore,becauseithasnotyetgainedcommonusage,thereislesspreviousdataforcomparison.However,measuringorientationselectiv-ityas(1Ϫcircularvariance),i.e.,theabsolutevalueofSasdefinedabove,gavesimilarresults(supplementalFig.S2,availableatwww.jneurosci.orgassupplementalmaterial)andshouldbeusefulinthefuture,particularlyincasesinwhichasinglemetricfororientationselectivityisdesirable.Thespatialfrequencytuningcurvewasdeterminedfromdriftinggrat-ingsfromtheresponseatorientationprefdescribedaboveandwasfittoadifferenceoftwoGaussians(HawkenandParker,1987).Bandwidthwascalculatedfromthefitastheratioofthespatialfrequenciesthatgavehalf-maximalresponse.Unitswereclassifiedashavinglow-passspatialfrequencytuningiftheresponseto0cpd,thefull-fieldflicker,wasatleast50%ofthepeakresponse.
Linearityofresponsewascalculatedfromdriftinggratings,attheori-entationandspatialfrequencythatgavepeakresponse.First,webinnedthe1.5spresentationintoaspikehistogramat100msintervalsandsubtractedthespontaneousrate.WethenappliedthediscreteFouriertransformandcomputedF1/F0,theratioofthefirstharmonic(responseatthedriftfrequency)tothe0thharmonic(meanresponse).
Responsestodriftingfull-fieldbarswereanalyzedbycomputingperi-stimulustimehistogramswithbinsize100ms.Thespontaneousratewassubtracted,andthepeakfiringrateforeachorientationwasusedtogenerateanorientationtuningcurve,whichwasfittothesumoftwoGaussiansandanalyzedasdescribedabove.
Receptivefieldsizewascalculatedfrom4ϫ8°lightbars.Barsweresweptacrossthevisualfieldateightlocationsalongtheaxisperpendic-ulartothedirectionofmotion,e.g.,forhorizontallymovingbars,eachpresentationsweptabaracrossatadifferentverticalposition.There-sponsesfromtheeightsweepswerebinnedat100msandusedtocon-structfiringrateasafunctionofbarposition.Thiswasfittedwithatwo-dimensionalGaussian,withindependentwidthsxandy,andRFradiuswascalculatedbyaveragingthehalf-widthathalf-maximumofthetwoaxesoftheGaussianfit(equivalenttothesemi-majorandsemi-minoraxesoftheellipsegeneratedbythehalf-maximumcontour).Thisprocesswasrepeatedforfourdifferentdirectionsandaveragedacrossallconditionsthatgaveasufficientresponse.Thebarlength,8°,waschosentoelicitstrongresponsesfromasmanyunitsaspossible.However,itputsalowerlimitonourmeasurementofRFsize,ϳ4°,sowemayhaveoverestimatedthesizeofthesmallestreceptivefields.
Temporalfrequencywascalculatedfromtheresponsetocontrast-reversingsinusoidalgratingsatafixedspatialfrequencyof0.04cpd.Theaveragefiringrateovereach2spresentationwascalculated,andthespontaneousratewassubtracted.Fortheorientationthatgavethelargesttotalresponse,atemporalfrequencytuningcurvewasfitusingadiffer-enceoftwoGaussians.
Toanalyzetheresponsetocontrast-modulatedwhitenoisemovies,webinnedthenumberofspikesinresponsetoeachframeofthemovie.AresponsivenessmetricwascalculatedbytakingthediscreteFouriertrans-formatthemodulationfrequency,normalizedbytheaveragefiringrate.Additionally,thespike-triggeredaverage(STA)ofcontrast-modulatedmovieresponseswascomputedbythemeanoftheframesprecedingeach
J.Neurosci.,July23,2008•28(30):7520–7536•7523
spike.Becauseweuseda1/fpowerspectrumforthestimulusset,therawSTAisbroadenedbythecorrelationsinthestimulusset.However,be-causethestimulusisGaussianandthereforeonlycontainssecond-ordercorrelations,wewereabletocorrecttheSTAexactlybynormalizingitsFouriertransformbythepowerspectrumofthestimulusset(Theunissenetal.,2001;Sharpeeetal.,2004).Ingeneral,thestrongestSTAsprecededspikesbyalagoftwomovieframes(66ms).PreferredorientationandspatialfrequencywerecalculatedbyfindingthepeakofthespatialFou-riertransformoftheSTAat66mslag.STAreceptivefieldswerealsofittoGaborfunctions(asinusoidwithaGaussianenvelope),describedby
F͑xЈ,yЈ͒ϭAexpͩϪxЈ2yЈ2
22Ϫ2ͪcos͑2fxЈϩ͒,
x2y
wherexЈandyЈarerotated,andtranslatedcoordinatesaredefinedbyxЈϭcos(xϪxperformedc)ϩsin(yϪyinMatlab,c)andyЈϭϪcos(xϪxbasedonroutinesprovidedc)ϩcos(yϪybyMichaelc).FitswereLewickiandcolleagues.
Contrast–responsecurvesweregeneratedfromtheresponsetodrift-inggratingswithspontaneousratesubtracted,forthepreferredorienta-tiondeterminedasabove.Becausestimuliwerepresentedatapreselectedspatialfrequency(0.04cpd)ratherthanthepreferredspatialfrequencyofeachunit,thepeakfiringrateisnotindicativeofthemaximalrespon-sivenessoftheunit,andthecurveswerethusnormalizedtomaximumresponseof1.NormalizedcurveswerefittotheNaka–Rushtonequation(NakaandRushton,1966):R(C)ϭg/(1ϩ(C/Cthegain,C50)b),whereCiscontrast,gis50isthemidsaturationcontrast,andbisafittingexponentthatdescribestheshapeofthecurve.
StatisticalsignificancewasdeterminedbyMann–WhitneyUtest,ex-ceptwhenotherwisestated.Inthefigures,*pϽ0.05,**pϽ0.01,and***pϽ0.001.Forfiguresrepresentingthemedianofdata,errorbarsshowstandarderrorofthemedianascalculatedbyabootstrap.Inothercases,errorbarsrepresentSEM.
LaminarResults
andcell-typeidentity
Werecorded235isolatedsingleunitsfrom27adultmice.Indi-vidualrecordingsessionsconsistedofsimultaneousrecordingsofϳtrode4–12array.isolatedOfthesesingleunits,units87%across(204theof235)16siteswereofresponsivethemultielec-toatleastoneepisodicvisualstimulus.Nonresponsiveunitswerenotincludedinanalysisoftuningproperties.Wegenerallyper-formedonlyoneelectrodeinsertionperanimaltoavoiddamagetocortexfrommultiplepenetrationsandtomaintainastableanestheticstatewithoutneedingtoredose.
Toverifythelaminarlocationofeachrecordingsiteofthemultielectrodearray,wemeasuredinadditiontoelectrodedepththeLFPresponsetosquare-wavecontrastreversalofachecker-board.AsshowninFigure1B,theaveragedresponsesshowalargedeflectionstartingϳ40msafterthecontrastreversal,whichvariesinbothamplitudeandwaveformacrosstherecordingsites.CSDanalysisprovidesamethodtotransformthesetofLFPrecordingsintothelocationsofcurrentsourcesandsinks,withcurrentsinksgenerallycorrespondingtositesofsynapticconduc-tances(Mitzdorf,1985;Swadlowetal.,2002).Thistransformationrevealedalaminardistributionofactivationinresponsetochecker-boardreversal(Fig.1C),withacurrentsinkbeginninginlayer4,inwhichsensoryinputfirstarrives,spreadinguptolayer2/3,andfinallyaweaksustainedsinkinlayer5.Retrospectivehistology(supplementalFig.S1,availableatwww.jneurosci.orgassupple-mentalmaterial)confirmedthecorrespondencebetweenCSDandlaminaridentity.Furthermore,layer4correspondedtothemaximumnegativityofinitialdipintheLFP(Fig.1B,C)(sup-plementalFig.S1,availableatwww.jneurosci.orgassupplemen-talmaterial),whichallowedustodeducethelaminarlocationevenforrecordingsthatdidnotspantheentiredepthofcortex
7524•J.Neurosci.,July23,2008•28(30):7520–7536andthereforecouldnotgenerateafullCSDmapofthelayers.Asshownbelow,thisassignmentoflayerscorrelateswithdifferentpropertiesofthevisualresponse.Furthermore,duringrecording,itwasgenerallypossibletorecognizetheselaminarpositionsqualitativelybythenatureofongoingactivity:layer2/3hasverysparsebackgroundactivity;inlayer4,thespontaneousratein-creasesandmoreunitsarepresentoneachrecordingsite;layer5hasdramaticallyhigherspontaneousratesandmuchlargerspikeamplitudes;and,inlayer6,spontaneousratesandspikeampli-tudesdecreaseandevokedresponsesaremoresparse.
Interestingly,weoftensawafastoscillation,atϳ50–55Hz,superimposedontheLFPandCSD(Fig.1B,arrows).Thisoscil-lationwasstimulusevoked,withcoherentactivitycommencingwiththeonsetoftheevokedLFP.Furthermore,itsfrequencyspectrumwasdistinctfromtypical60Hznoiseanddidnotchangeaswevariedtherefreshrateofthedisplay.Wethuspre-sumethatthiscorrespondstogammaoscillationsthathavebeendescribedinmanyspecies(Buzsa´kiandDraguhn,2004),includ-ingmice(Naseetal.,2003).However,wedidnotinvestigatethisoscillationfurther.
Asanadditionalmeansofcorrelatingvisualresponseswithcorticalcircuitelements,weclassifiedunitsbasedontheirspikewaveform.Figure1Dshowstheaveragewaveformsfromalltheunitsrecorded,exceptforfourunitsthathadunusualwaveformsandwereexcludedfromclassification.Thewaveformscanbeseentofallintotwogeneralclasses,narrowspiking(blue)andbroadspiking(green),withaveragesacrossthetwoclassesshowninFigure1E.Previousresultshavesuggestedthatnarrow-spikingcellscorrespondtoinhibitory,predominantlyfast-spiking,inter-neurons(McCormicketal.,1985;ConnorsandKriegstein,1986;Bartho´etal.,2004),andrecentstudieshaveusedthisextracellularsignatureasameansofdistinguishingexcitatoryandinhibitoryneurons(Swadlow,2003;Andermannetal.,2004;Leeetal.,2007;Mitchelletal.,2007;AtencioandSchreiner,2008).Avarietyofwaveformparametershavebeenusedtoseparatethesetwoclasses,includingtrough-to-peaktime(Bartho´etal.,2004;Mitchelletal.,2007),theratiooftrough-to-peakamplitude(An-dermannetal.,2004;Hasenstaubetal.,2005),andthedurationofthelonger,positivepeak(BrunoandSimons,2002;AtencioandSchreiner,2008).Becauseouracquisitiondidnotrecordtheentiredurationofthepositivepeak,weusedtheslope0.5msaftertheinitialtrough(i.e.,howquicklythevoltageisreturningtobaseline)asaproxyforthisparameter.Thesethreeparametersprovidedgoodseparabilityanddistinguishedthesametwogroupsofwaveforms,asshowninFigure1,FandG.Nineteenpercentofthetotalpopulationofrecordedcells(45units)fellintothenarrow-spikingcategory,andtherewasnotasignifi-cantlygreaterfractionofnarrow-spikingunitsinanyparticularlayer(pϾ0.1,2test).ThisproportionmatchesthefractionofGAD-positiveneuronspreviouslyidentifiedhistochemically(Tamamakietal.,2003).Forsimplicity,werefertothesetwogroupsas“putativeinhibitory”and“putativeexcitatory,”butthisshouldnotimplythatthereisaperfectcorrespondencebe-tweenthespikewaveformandcelltypeassignment(seeDiscussion).
Selectiveandnonselectiveresponses
Acharacteristicofvisualcortexresponsesisthetransformationfromprimarilycenter-surround,nonorientedreceptivefieldsobservedinretinaandlateralgeniculatenucleusofthalamus(LGN),toresponsesinV1thatareoptimalforbarsandedgesofaparticularorientation.Tomeasurethedegreeoforientationandotherresponseselectivityinmousevisualcortex,wepre-
NiellandStryker•ReceptiveFieldsinMouseVisualCortex
sentedfull-length5°widebarsofvaryingorientationsdriftingat30°/sanddriftingsinusoidalgratingsmovingat2Hzofvaryingorientationandspatialfrequency.Wefoundarangeofresponsetypes,includingmanyunitsthatwerehighlyselectiveforstimu-lusorientation,otherunitsthatrespondedtoawiderangeofstimuli,andasmallfractionofunitsthatdidnotsignificantlychangetheirfiringinresponsetovisualstimuli.ExamplesofahighlyselectiveandapoorlyselectiveresponseareshowninFig-ures2and3,respectively.
TheunitshowninFigure2wasabroad-spiking,putativeexcitatoryneuronlocatedinlayer4.Figure2Ashowsspikeras-tersforrepeatedpresentationsofabardriftingin16differentdirections.Itgivesastrongresponsetobarsoveranarrowrangeoforientationsandhasashortdurationofpeakresponse,corre-spondingtoϽ10°ofvisualspace,asthebarcrossesitsreceptivefield.ThetuningcurveshowninFigure2B,measuringpeakre-sponseforeachorientation,demonstratesthattheorientationtuningwidthislessthanϮ30°andthatithasϳ2:1preferenceforonedirectionofmotion.Figure2Cshowsspikerastersfromthesameunitinresponsetofull-fieldsinusoidaldriftinggratings,whichdemonstrateasimilarorientationtuningandselectivity(Fig.2D).Furthermore,itrespondstogratingsonlyoverapar-ticularrangeofspatialfrequencies(Fig.2E),withmaximalre-sponseϳ0.08cpd,correspondingtogratingbarsspaced12°apart.Finally,theunitrespondstothedriftinggratingwithaperiodicallymodulatedresponse.Thisischaracteristicofalinear,orsimplecell,response,suchthattheunitrespondsmaximallywhenanorientedstimulusisinphase,i.e.,aligned,withONandOFFsubregionsofitsreceptivefield,andrespondsminimallyorreducesitsresponsewhenthestimulusisoutofphasewiththereceptivefield.Theamplitudeofthismodulationrelativetotheaveragefiringrate,theF1/F0ratio,hasbecomeastandardquan-titativemetricforthequalitativelydefinedsimpleandcomplexcelldistinction(Skottunetal.,1991).
Figure3showsamuchlessselectiveresponse,fromanarrow-spikingputativeinhibitoryunitalsoinlayer4.Thisunitrespondstobarsofallorientation(Fig.3A),withonlyaslightlyelevatedresponsetocertainorientations(Fig.3B).Italsorespondsoveramuchlargerregionofvisualspace,upto40°across.Theresponsetodriftinggratings(Fig.3C)issimilarlyuntunedfororientation(Fig.3D),anditextendsoverawiderrangeofspatialfrequencies(Fig.3E),evengivingasmallresponseto0cpd,i.e.,afull-fieldflicker.ThisunitalsoshowsarelativelyconstantrateoffiringasthegratingdriftsratherthantheperiodicallymodulatedresponseseeninFigure2C.Thiswouldthusbeclassifiedasanonlinear,orphase-invariant,response,becauseitsresponsedoesnotdependontheprecisealignmentofthegratingwithinthereceptivefield.Responsepropertiesacrossthepopulation
Theresponsesofthesetwounitsarerepresentativeofthetwoendsoftheselectivityspectrum.Weusedtwomeasurestode-scribetheselectivityoftheorientationtuningcurvesinresponsetodriftinggratings.First,theOSImeasurestherelativeresponseatthepreferredorientationversustheorthogonalorientation,inotherwords,thetunedversusuntunedresponse.Figure4AshowstheOSIacrosstheentirepopulationofunitsthatre-spondedtodriftinggratings(nϭ182).TheOSIwasdefinedas(RinpreftheϪpreferredRortho)/(Rorientation,prefϩRortho),whereRandRprefisthepeakresponseorthoistheresponseintheorthogonaldirection.Bythiscriterion,perfectorientationselec-tivitywouldgiveOSIof1,anequalresponsetoalldirectionswouldhaveOSIϭ0,and3:1selectivitycorrespondstoOSIϭ0.5.Alargefractionofunitsthusshoworientationselectivity,with
NiellandStryker•ReceptiveFieldsinMouseVisualCortexJ.Neurosci.,July23,2008•28(30):7520–7536•7525
Figure2.Responsepropertiesofatypicaloriented,linearunit.A,Spikerastersinresponsetodriftingbarsin16directions.B,OrientationtuningcurvefromrastersinAdemonstratessharptuning.DashedcurveisfittoasumofGaussians.C,Spikerastersinresponsetodriftinggratingsof12directionsandsixspatialfrequencies.Periodicresponseisindicativeofalinearresponsetype.D,OrientationtuningcurvefromrastersinC.E,Spatialfrequencytuningcurveshowsbandpassselectivity.DashedcurveisfittodifferenceofGaussians.
74%(135of182)havingOSIϾ0.5,andmanyshowingnearlycompleteselectivity,suchastheunitinFigure2.Figure4Bshowsthepreferredangleforunitsthatwereorientationselectiveandrespondedstronglytobothgratingsandbars(nϭ87),demon-stratingthatthistuningwasconsistentbetweenstimuli(r2ϭ0.85),andshowedlittlesystematicpreferenceforcertainorien-tations.Fororientation-selectiveunits(OSIϾ0.5),wemeasuredthewidthofthetunedresponse,asthehalf-widthathalf-maximumabovethebaseline.Figure4Cshowsthatmostoftheorientedunitsshowrelativelynarrowtuning,similartothatseenintheunitofFigure2,withmediantuningwidthϮ28.0°.Itshouldbenotedthatsomestudieshavemeasuredtuningwidthashalf-widthathalf-maximumincludingtheuntunedresponseratherthanabovetheuntunedbaseline.Becausethispredomi-nantlyaffectsthepoorlyselectiveunits,includingtheuntunedbaselineonlyshiftedthemedianwidthofunitswithOSIϾ0.5to28.9°.
Thedegreeoforientationselectivityvarieddramaticallyacrosslayersandbetweenputativeexcitatoryandinhibitoryunits,asshowninFigure4D,ascatterplotoforientationselec-tivityforallunitsthatrespondedtodriftinggratings,andFigure4E,whichshowsthemeanorientationselectivityforeachofthesegroups.Inlayers2/3,4,and6,thereisasharpdistinctionbetweencelltypes,becausealmostallputativeexcitatoryunitswerehighlyorientationselective,whereasmostputativeinhibitoryunitswereuntuned.Althoughtherewasnotagreatdifferenceintheorien-tationselectivityindex(tunedvsuntunedresponse)betweenex-citatoryunitsacrosstheselayers,wedidobserveaslightsharpen-inginthewidthoforientationtuninginupperlayer2/3,asshowninFigure4F.Inlayer5,putativeexcitatoryunitsshowedarange
7526•J.Neurosci.,July23,2008•28(30):7520–7536NiellandStryker•ReceptiveFieldsinMouseVisualCortex
Figure3.Responseofapoorlyselective,nonlinearputativeinhibitoryunit.A,Responsetodriftingbarsofvariousorientationsdemonstratesminimalorientationselectivity.B,OrientationtuningcurvefromrastersinA,withfittosumofGaussians(dashedline).C,Responsetodriftinggratingsshowscontinuous,ratherthanperiodic,response,withminimalorientationtuning.D,Orientationtuningcurvefordriftinggratings.E,Spatialfrequencytuningcurve,withfittodifferenceofGaussians(dashedline).
oforientationselectivity,but,overall,theselectivitywasmuchless(pϽ0.001)andveryfewshowedthehighorientationselec-tivitythatwastypicalofthemoresuperficiallayers.Furthermore,evenamongorientation-selectiveunits,thewidthoftuninginlayer5wasmuchbroader(pϽ0.001),demonstratingthatboththewidthofthetunedresponseandthemagnitudeoftheun-tunedresponseareincreased.Asecondmeasureoforientationselectivity,basedonthecircularvarianceofresponseacrossori-entations,gavesimilarresultsforbothindividualunitsandpop-ulationcomparisons(supplementalFig.S2,availableatwww.jneurosci.orgassupplementalmaterial).Althoughmanycellsshowedpreferentialresponsesforapar-ticularorientation,itwaslesscommonforthemtobeselectiveforthedirectionofmotionatthatorientation.Figure5Ashowstheresponsetodriftingbarsforonesuchdirection-selectiveunit,whichrespondedstronglytobarsmovingat202°butnegligiblytobarsmovingintheoppositedirection,22.5°(whichcorrespondstothesamebarorientation).Figure5Bshowstheprevalenceofdirectionselectivityinresponsetodriftinggratingsacrossthepopulation,demonstratingthat,althoughasmallfractionshowquitestrongselectivity,most(77%,140of182)aresubstantiallylessselective,withindicesϽ0.5,whichcorrespondstoa3:1bias
NiellandStryker•ReceptiveFieldsinMouseVisualCortexJ.Neurosci.,July23,2008•28(30):7520–7536•7527
Figure4.Orientationselectivity.A,HistogramofOSIforallresponsiveunits(nϭ182),withgrayandblackrepresentingproportionofputativeexcitatory(exc)andinhibitory(inh)units.ArrowsshowvaluesforunitsinFigures2and3.B,Comparisonofpreferredorientationangle(degree)asmeasuredwithbarsandgratingsdemonstratesconsistency.C,Histogramofmeantuningwidthforallorientation-selectiveunits(nϭ135).ArrowshowsvalueforunitinFigure2.D,Orientationselectivitybylayerandcelltype.E,MeanOSIforeachlayerandcelltype.F,Medianwidthoforientationtuningforallorientedunitsbylayer.
Figure5.Directionselectivity.A,Spikerastersforresponseofarepresentativedirection-selectiveunittobarsdriftingin16directions.B,Histogramofdirectionselectivityfromdriftinggratingsacrossthepopulationofrecordedunits(nϭ182),withgrayandblackrepresentingrelativeproportionofputativeexcitatory(exc)andinhibitory(inh)units.ArrowsshowvaluesfromunitsshowninFigures3,2,and5A,respectively.C,Distributionofdirectionselectivitybylayerandcelltypeshowsthatmosthighlydirection-selectivecellsareputativeexcitatoryunitsinlayers2/3and4.
forpreferredversusoppositedirection.AsFigure5Cshows,al-mostalldirection-selectiveunitswereputativeexcitatoryunitsinlayers2/3and4.
AsshowninFigures2and3,unitsinmouseV1respondedpreferentiallytoparticularspatialfrequenciesofsinusoidalgrat-ing.Figure6Apresentsahistogramofthespatialfrequencythatelicitedthemaximumresponseattheoptimalorientation,whichgenerallyrangedfrom0.02to0.08cpd(medianof0.036cpd),althoughseveralunitsrespondedoptimallyatspatialfrequenciesupto0.32cpd,thehighesttested.Asmallfractionrespondedbestto0cpd,full-fieldcontrastreversal.Thepreferredspatialfre-quencydidnotshowanysystematicvariationacrosslayers(Fig.6B),exceptforlayer6,whichrespondedtosignificantlylowerspatialfrequencies,asdidputativeinhibitoryunits.Thepreferredspatialfrequencydidnotshowanysignificantvariationwithre-ceptivefieldeccentricity,whichrangedfrom20to90°inazimuth(r2ϭ0.009;pϽ0.37),consistentwiththelackofafovea/areacentralis,therelativelyconstantphotoreceptorandretinalgan-glioncellspacingacrossthemouseretina(Jeonetal.,1998),andtheuniformmagnificationofthecorticalrepresentation(KalatskyandStryker,2003).Mostneuronsshowedbandpassspatialfrequencytuning(Fig.6C).Themedianbandwidthofspatialfrequencytuningforbandpassunitswas2.46octaves,althoughsomeneuronsrespondedtoonlyoneofthespatialfre-
7528•J.Neurosci.,July23,2008•28(30):7520–7536NiellandStryker•ReceptiveFieldsinMouseVisualCortex
quenciespresented,andasmallfractionshowedlow-passtuning(Fig.6C).Thewidthofspatialfrequencytuningwasgreaterforbothputativeinhibitoryandlayer5putativeexcitatoryunits(Fig.6D).Acommonclassificationofvisualre-sponsesisintosimplecells,whosere-sponsetostimulicanbepredictedbythelinearsumofresponsestospotsatindivid-uallocations,andcomplexcells,whichdemonstratenonlinearspatialsumma-tion.Correspondingly,simplecellsarephasedependent,inthattheygenerallyre-spondtoanedgeofparticularorientationataspecificlocation,whereascomplexcellsarephaseinvariant,inthattheyre-spondtotheedgeatanylocationwithinthereceptivefield.Thisdistinctionmaybemeasuredusingthemodulationofre-sponsetodriftinggratings(Skottunetal.,1991).AhighF1/F0ratioindicatesthatthefiringoftheunitismodulatedatthetem-poralfrequencyofthegrating,whereasalowF1/F0indicatesthattheunitfiresrela-tivelyconstantlythroughoutthepresenta-tionofthegrating.TheunitsinFigures2and3hadF1/F0of1.43and0.36,respec-tively.Figure7Ashowsthebimodaldistri-Figure6.Spatialfrequencytuning.A,Histogramofpeakspatialfrequencyresponse.B,MedianpeakspatialfrequencyacrossbutionoftheF1/F0ratioforallunitsthatlayersdemonstrateslowerpreferredspatialfrequencyinlayer6andputativeinhibitory(inh)units.C,Widthofspatialfrequencyrespondedtogratings(nϭ182),empha-tuning.ArrowsinAandCshowvaluesfromunitspresentedinFigures2and3.LP,Low-passtuning.D,Medianspatialfrequencysizingthedistinctionoflinearandnonlin-tuningwidthshowsbroadertuninginlayer5andputativeinhibitoryunits.nϭ182unitstotal;nϭ14,40,31,34,24,and39byearunitsbythismetric.Mostputativecelltype.inhibitoryunitsshowednonlinearre-theywerenotconsideredresponsivetothegratings.Figure8Fsponses,asdidlayer5broad-spikingunits(Fig.7B,C).Inlayers
demonstratesthattheevokedrateisrelativelyconstantacross2/3,4,and6,themajorityofputativeexcitatoryunitswerelinear,
layers,althoughgreaterinputativeinhibitoryneurons.Theme-althoughasizeablefractionwerenonlinear.
dianevokedfiringrateacrossthepopulationwas6.7spikes/s,Thesizeofthereceptivefieldswasmeasuredbysweeping
whichislowerthanstudiesinV1ofotherspecies(Girmanetal.,shortbars(4°wideϫ8°long)ofdifferentorientationsacrossthe
1999;Heimeletal.,2005)buthigherorconsistentwiththatseenvisualfield.Figure8Ashowtheradiusofthereceptivefield(half-inotherregionsofrodentcortex(Brechtetal.,2003;DeWeeseetwidthathalf-maximum)for108unitsthatgavesufficiently
al.,2003;Satoetal.,2007).Itispossiblethatanesthesialevelstrongresponsetothisstimulus.Theradiusofthereceptivefield
contributedtotheselowerratesorthatourmethodofisolatingwassmallestamonglayer2/3and4broad-spikingunitsandwas
unitswithoutregardtoresponsivenessmayallowustodetectgenerallyϳ5–7°(Fig.8B).Putativeinhibitoryandlayer6units
unitswithlowerfiringrates.Itshouldalsobenotedthatthesetendedtobesomewhatlarger,andlayer5unitswerenearlytwice
firingratesareaveragesover1.5s.Thepeakinstantaneousfiringaslarge(pϽ0.001),withsomereceptivefieldsupto15°radius
ratecanbesignificantlyhigher(supplementalFig.S3,availableatbythismeasure.
www.jneurosci.orgassupplementalmaterial),particularlyforThespontaneousfiringratealsovarieddramaticallyacross
simplecells,whichmodulatetheirfiringrateperiodicallyinre-layersandcelltype(Fig.8C,D).Putativeexcitatoryunitsinthe
sponsetogratings.Additionaldetailsofresponsemagnitude,superficiallayershadextremelylowspontaneousrates,oftenfir-spontaneousrate,andvariabilityarepresentedinsupplementalinglessthanonceevery10s,whereaslayer4broad-spikingunits
FigureS4(availableatwww.jneurosci.orgassupplementalwerealmosttwiceasactive(pϽ0.05).Putativeinhibitoryunits
material).hadawiderangeofspontaneousratesbuttendedtofireatrates
Wemeasuredtemporalfrequencyresponsebypresentingalmostanorderofmagnitudehigher.Nearlyallunitsinlayer5
contrast-reversingsinusoidalgratingsatafixedspatialfrequencyhadhighspontaneousrates:thedramaticincreaseinfiringrate
(0.04cpd)andvaryingtemporalfrequency(Fig.8G).Mostunitsgenerallyprovidedaclearindicationofthelayer5boundary.Itis
respondedoptimallyatϳ2Hz(medianof1.68Hz)althoughweinterestingtonotethatcelltypeswithhighspontaneousrate
sawasignificantincreaseinpeaktemporalfrequencyinlayer4.(layer5andputativeinhibitory)havetheloweststimulus
ThisisshowninFigure8H,whichpresentstheaveragetemporalselectivity.
frequencytuningcurvesforlayers2/3and4,demonstratinganThemagnitudeoftheevokedfiringratedidnotvaryasdra-increasedresponsebylayer4unitstostimuliat4and8Hz(pϽmaticallyacrosslayersasthespontaneousrate.Figure8Eshows
0.05).theaveragefiringrateover1.5sinresponsetotheoptimalgrating
Therelativeproportionofdifferentfunctionalresponsecate-stimulus(withspontaneousratesubtracted),forallunits(nϭ204)thatrespondedtoatleastoneoftheepisodicstimuli,evenifgoriesacrossthelayersandcelltypesisillustratedinFigure9,
NiellandStryker•ReceptiveFieldsinMouseVisualCortexJ.Neurosci.,July23,2008•28(30):7520–7536•7529
Figure7.Linearityofresponsetodriftinggratings.A,HistogramofF1/F0ratioacrossentirepopulation(nϭ182),withblackandgrayshowingrelativeproportionofputativeinhibitory(inh)andexcitatory(exc)units.ArrowsshowunitsfromFigures3and2(leftandright,respectively).B,Distributionoflinearitybylayerandcelltype.C,Fractionofunitswithlinearresponses(F1/F0Ͼ1).
usingOSIϭ0.5andF1/F0ϭ1asthresholdsfororientationselectivityandlinearity.Abinaryclassification,whichdependsonanarbitrarythreshold,isdifficultwhentheparameterisnotbimodallydistributed.Evenwhenaparameterdoesshowabi-modaldistribution,suchastheF1/F0ratio,thismaynotrepresentatruedistinctionineithertaxonomyormechanism(MechlerandRingach,2002).Ourcategorizationisthusmeantonlytoprovideasummaryoverviewofresponsetypesforcomparisonwithotherstudies.
Acrossthepopulation,13%ofunitswerenotresponsivetoanyofthestimuliweused,and9%wereleftunclassifiedbecausetheyfailedtogivesufficientresponsetodriftinggratingstoesti-mateorientationselectivityandlinearity.Inlayers2/3and4,broad-spikingputativeexcitatoryunitswerenearlyalwayssimpleandorientationselective,withafractionofnonlinearorientedunitsandasmallnumberofsimplenonorientedunitsinlayer4.Layer5,incontrast,showedmostlynonlinearresponses,withalargefractionofclassiccomplexorientedunitsandasmallernumberofnonlinearnonorientedunits.Layer6wassimilartolayers2/3and4,althoughwithagreaternumberofnonlinearorientedunitsandthehighestproportionofnonresponsiveunits.Finally,three-quartersofputativeinhibitoryunitswerenonlinear,ofwhichthevastmajoritywerenonoriented.Responsestonoisemovies
Tomeasureresponsestomorecomplete,yetstillwellparameter-ized,stimuli,wegeneratedmoviesofstochasticnoisewithde-finedspatialandtemporalfrequencyspectra(Fig.10A)(supple-mentalmovie,availableatwww.jneurosci.orgassupplementalmaterial).Weusedthesemoviestoprobetheoverallvisualre-sponsivenessofunitsbyperiodicallymodulatingthecontrastsothateachmovietransitionedsinusoidallyfromagrayback-groundtofullcontrastmovieandbacktograyagain,witha10speriod.Thisgenerallyresultedinaperiodicmodulationoffiring,asdemonstratedinFigure10B.Modulatingthecontrastalsoservedtomaintainhighfiringratesthroughoutthepresentation,becauseunitsoftenhabituatedandfiringratesdecreasedduringlongmovieswithoutvaryingcontrast.
Bymeasuringtheratiooftheresponseamplitudeatthefre-quencyofthemoviemodulationtotheaveragefiringratethroughoutthemovie(Fig.10C),weobtainedameasureofover-allvisualdrive.Thisvalueis0ifthefiringrateisconstantduringthemovie(noresponse)and1foraperfectsinusoidalmodula-tionwithnobaselinefiring.Itshouldbenotedthat,becausethisismeasuringthenetfiringinresponsetoarichstimulus,ittherebycombinesbothpeakresponsivenessandbroadnessoftuninginasinglemetric.Figure10Ddemonstratesthatmostunitswereresponsivetothisvisualstimulus,includingmanyoftheunits(55%,17of31)forwhichwecouldnotelicitaresponseandmeasurereceptivefieldpropertiesusingtheepisodicstimulidescribedpreviously.ThephaseoftheFouriercomponentatthecontrast-modulationfrequencydescribesthetimeofoptimalre-sponse,whichwasgenerallyslightlybeforethecontrastmaxi-mumat180°.Interestingly,afewunitsshowedtheoppositere-sponseandactuallydecreasedtheirfiringrateinresponsetothemovies,asindicatedbyaphasenear0°,oppositetotherestofthepopulation.Figure10Eshowstheresponsivenessbylayerandcelltype,demonstratingthatlayer5and6unitstendedtomodulatetheirfiringratelessinresponsetothewhitenoisemovies(pϾ0.001),whereasupperlayer2/3wasthemostresponsive(pϽ0.01).Modulatingthecontrastinthismannerandplottingtheaveragedresponseoverallcyclesalsopermittedaformofcon-trast–responsecurve,aswellasaroughmeasureofcontrastad-aptationoverperiodsϽ10s(supplementalFig.S5,availableatwww.jneurosci.orgassupplementalmaterial).
Noisemovieshavebeenusedextensivelytomeasurelinearspatiotemporalreceptivefieldstructuredirectlybycalculatingthespike-triggeredaverage(JonesandPalmer,1987a;Chichilni-sky,2001).Foreachspikethataneuronfires,wecollectedtheframethatprecededthespikebyatime.TheaverageofalltheseframesisknownastheSTAand,foraGaussianwhitenoisestimulus,representsthelinearkernelofthespatiotemporalre-sponseofaneuron.Inthiscase,wecorrectedtheSTAtoaccountforthenon-whiteGaussianspectrumofthestimuli,vianormal-izationbythepowerspectrumofthestimulusset(Theunissenetal.,2001;Sharpeeetal.,2004).
Forunitsthatrespondedlinearlytodriftinggratings,weweregenerallyabletorecoveraSTAreceptivefieldfromresponsestothecontrast-modulatednoisemovies.Approximatelytwo-thirds(64%,63of98)ofsimplecellsproducedareceptivefieldbySTAfora10–15minmoviepresentation.Ingeneral,theprimarylimitationingeneratinganSTAreceptivefieldforlinearunitswasthenumberofspikesobserved;forsimplecellsthatgeneratedatleast400spikes,Ͼ90%producedanSTAwithsufficientsignal-to-noise,suggestingthatlongermoviepresentationscouldrevealanSTAformoreunits.Notethat,becausethestimulussetwasspatiallyfrequencylimited,thenoiseintheSTAreceptivefieldsisalsofrequencylimited,leadingtothebeadedorrippledappear-anceoftheresidualnoiseintheestimatedRFs.MonteCarlosimulationsofGaborreceptivefieldswithourmoviestimulishowsimilarrippledappearanceatlownumbersofspikes,whichaverageoutforlongermoviepresentationsorhigherspikecount.
7530•J.Neurosci.,July23,2008•28(30):7520–7536NiellandStryker•ReceptiveFieldsinMouseVisualCortex
Fornonlinearunits,withalowF1/F0ratio,theSTAgenerallyshowednostructure,correspondingtothelackofalinearker-nel.Becauseoftherelativelysmallnumberofspikescollectedduringthesemoviepre-sentations,wedidnotattempttousespike-triggeredcovarianceorothermeth-odstoextractnonlinearreceptivefields.Figure11A–CshowsexamplesofSTAreceptivefields,includingtheunit(Fig.11B1)whoseresponsetogratingsandbarswasillustratedinFigure2.ManyRFshadbothanONandOFFsubregion(55%),oroftenthreesubregions(27%),whereas18%ofunitsonlyhadonesubregion(Fig.11A–C,respectively).TheorientationoftheSTAreceptivefieldsgenerallyagreedquitewellwiththepreferredorientationdeterminedfromdriftinggratings(Fig.11D)(r2ϭ0.91).Thecorrespondencebe-tweenthespatialfrequencyoftheSTAandthepeakresponsetogratingswasalsosig-nificant(Fig.11E)(r2ϭ0.63)butonlyforunitsthathadapeakspatialfrequencyϾ1⁄60cpdasmeasuredbydriftinggratings(bluecircles).Itisnotsurprisingthattheunitswiththelowestspatialfrequencies(Ͻ1⁄60cpd;graycircles)gaveSTAswithin-accuratespatialfrequencies,becausethetotalsizeofthemoviewas60°(corre-spondingto1⁄60cpd),preventingusfrommatchinglowerspatialfrequenciesthanthis.Interestingly,themoresubunitsfoundintheSTA,thenarrowertheorien-tationtuningwidthasdeterminedfromgratings(Fig.11F),consistentwithlinearmodelsoforientationselectivity.Simi-larly,spatialfrequencybandwidthnar-rowedwithincreasingnumberofsubunits(Fig.11G).
TofurtheranalyzethestructureofSTAreceptivefields,wefitthemeasuredRFstoGaborfunctions,whicharesinusoidsmul-tipliedbyaGaussianenvelope.Figure11Hshowsthefractionofsignalvariancethatwasaccountedforbythefit(nϭ63),dem-onstratingthat,formanyunits,theGabor
providedagooddescriptionofthedata.Figure8.Receptivefieldsize,firingrates,andtemporalfrequencyresponse.A,ReceptivefieldsizemeasuredwithsmallBecauseourmeasuredRFshadrelativelysweepingbars,segregatedbylayerandcelltype.nϭ108units.B,Meanreceptivefieldsizeofeachgroup.C,Logarithmicplotoflargenoise(upto20%oftheSTAvari-spontaneousfiringrateforallunits,bylayerandcelltype.nϭ231units.D,Medianspontaneousfiringrateofeachgroup.E,ance)asaresultofthesmallnumberofFiringrateinresponsetooptimalgratingstimulus,forallvisuallyresponsiveunits(nϭ204),averagedover1.5spresentation.spikes,weexpectthatlongermoviepre-[InstantaneousfiringratesareshowninsupplementalFig.S3(availableatwww.jneurosci.orgassupplementalmaterial).]F,sentations,orfurtheroptimizationofMedianresponsetooptimalstimulusforeachgroup.G,Medianpeaktemporalfrequency,averagedforlayersandcelltype,movieparametersbasedonthetuningshowinghighertemporalfrequencyresponseinlayer4.nϭ96units.H,Meantemporalfrequencytuning,comparingresponsepropertiesdeterminedinthisstudy,wouldinlayers2/3and4.exc,Putativeexcitatoryunits;inh,putativeinhibitoryunits.produceSTAswithlowernoiseandbetter
thevaluesofnxandnyforunitsthatwerebestfitbytheGabor(fitfits.
errorϽ00.33;nϭ50)comparedwithsimilardatafrommacaqueTheshapeofaGaborcanbecharacterizedbytwonumbers,nx(Ringach,2002).ThemouseSTAreceptivefieldsoverlapwithϭxfandnyϭyf,whichcorrespondtowidthalongthesinusoid
mostofthemacaquedata,indicatingthattheyhavesimilarandperpendiculartothesinusoid,intermsofnumberofcyclesof
shapes.However,themousedoesnothaveunitswithverylargethesinusoid(Ringach,2002).Theratioofnxtonycanbethought
nx,consistentwiththefactthatwedidnotobserveSTAswithofastheaspectratio,whereasthevalueofnxisrelatedtothe
morethanthreestrongsubunits.AhistogramofthephaseofthenumberofsubunitsthatcanfitintheGaussian.Figure11Ishows
NiellandStryker•ReceptiveFieldsinMouseVisualCortexJ.Neurosci.,July23,2008•28(30):7520–7536•7531
spondedtothisspatialfrequency(nϭ22units)resultedinthemeantuningcurveobservedinFigure12C,whichshowsnoincreaseinwidthwithincreasedcontrast.Furthermore,acomparisonofallindivid-ualorientationtuningcurvesattwodiffer-entcontrastsshowsnosystematicchangeineitherorientationselectivity(Fig.12D)ortuningwidth(Fig.12E).
Inadditiontotestingcontrast-invarianttuning,thisstimulussetallowedustomeasuretheresponseasafunctionofcontrastforthoseunitsthatwererespon-sivetothisspatialfrequency.Figure12Fshowsthevalueofcontrastthatresultedinhalf-maximalresponse,whichshowedawiderange,withamedianvalueof19.8%.Theslopeofresponseovercontrastatthehalf-maximalresponseisshowninFigure12G;fornormalizedcontrast–responsecurves,aslopeof1indicatesarelativelylinearincreaseinresponsewithcontrast,whereasagreaterslopeindicatesasharptransitionfromlowtohighresponseatthe
Figure9.Distributionoffunctionalresponsecategoriesforlayersandcelltypes.exc,Putativeexcitatoryunits.midsaturationcontrast.Althoughmany
unitshadslopenear1,againtherewasa
sinusoid(Fig.11J),mappedontotherange0–90°toremovewiderange,withamedianvalueof2.4.symmetries(FieldandTolhurst,1986),showsanover-Discussionrepresentationofodd-symmetricreceptivefields.Estimatesof
Despitethepoorvisualacuity(Pruskyetal.,2000)andrelativelythespatialphasedistributionshowawiderangeinotherstudies,
smallregionofcortexdevotedtovisualprocessing,wefindthatincludinguniform,peakedat90°,orbimodalat0and90°(Field
mostneuronsinmouseV1respondtoatleastonetypeofvisualandTolhurst,1986;JonesandPalmer,1987b;DeAngelisetal.,
stimulus.Wehavedeterminedtherelevantrangeofstimulus1993;Ringach,2002).
parameters,suchasspatialandtemporalfrequency,andorienta-tionselectivity,whichareeffectiveindrivingthestrongestre-Contrast-invarianttuning
sponses.Thisbroadcharacterizationshouldbenefitfuturestud-Animportantaspectofcorticalvisualresponsesisthatorienta-iesbyguidingthedesignofappropriatevisualstimuli.Fortionselectivitydoesnotvarywithincreasingstimuluscontrast,a
instance,knowledgeoforientationandspatialfrequencytuningphenomenonknownascontrast-invarianttuning(Fersterand
widthcanhelpspecifytheresolutionatwhichtheseparametersMiller,2000;PriebeandFerster,2008).Thisconstancyconstrains
shouldbesampledtoproperlycharacterizeallcells.Theseresultsmodelsofcorticalprocessing,becausesimplefeedforwardmod-shouldalsofacilitatethedesignofoptimalstimuliforotherfunc-elssuchastheoriginalHubelandWieselproposal(Hubeland
tionalrecordingtechniquesthatlackthesingle-spiketemporalWiesel,1962)generallypredictanincreaseintuningwidthasthe
resolutionofelectrophysiology,suchasintrinsicsignal(Kalatskyvisualdrive(e.g.,contrast)isincreased(Fig.12A,right).How-andStryker,2003)orcalciumimaging(Stosieketal.,2003).Im-ever,itisgenerallyobservedthat,ascontrastincreases,thevisual
portantly,ourfindingthatresponsesvarygreatlyacrossthelayersresponseincreasesbutunitsdonotloseselectivity,resultingina
indicatesthattherecanbewidelydifferentresultsforsimplemultiplicativescalingofthetuningcurve(Fig.12A,left).Numer-measurements,suchasthedegreeoforientationselectivityorousfactorshavebeenidentifiedthatcouldexplainthisproperty,
receptivefieldsize,dependingonthedistributionofrecordingincludingrecurrentconnectivity,inhibition,membranepoten-sitesinagivenstudy.tialfluctuationsattributabletobackgroundactivity,andthespike
threshold,amongothers(SompolinskyandShapley,1997;Fer-TechnicalconsiderationssterandMiller,2000;Shapleyetal.,2003;Finnetal.,2007;Priebe
SeveraltechnicaldifferencesfrompreviousexperimentsmayandFerster,2008).
contributetotheresponsivenessandselectivityweobserved.InWesoughttodeterminewhetherthisconstancyispresentin
termsofsurgery,performingasmallcraniotomyandinsertingthemouseaswell,tobothassessthesimilarityincorticalprocess-theelectrodeatanangletothecorticalsurface,whichmayreduceingtootherspeciesandascertainwhethergeneticmanipulations
damagetosuperficialrecordingsites,aidedinisolatingunitsandinmicecouldprovidecausaltestsofproposedmechanismsof
evokingstrongresponsesfromthesuperficiallayers,inwhichtheinvarianttuning.Forasubsetofrecordings(nϭ6animals),we
mostselectivecellswerefound.Incontrasttotraditionalsingle-presenteddriftinggratingsofvaryingcontrastandorientation,at
electroderecording,whichinvolvesadvancinganelectrodeuntilfixedspatialfrequency(0.04cpd).Atypicalresponsefroman
arespondingunitisfound,weallowedourmultisiteelectrodetoorientedunitshownisshowninFigure12B,demonstrating
settleatonepositionandrecordedanyunitsthatcouldbede-contrast-invarianttuning,asthetuningcurveonlychangesin
tectedover1hormoreofrecordingtime.Thisenabledustoamplitude,notshape,ascontrastincreases.Aligningandaverag-isolateunitswithlowspontaneousratesandsparseresponses.ingthetuningcurvesofallorientation-selectiveunitsthatre-
7532•J.Neurosci.,July23,2008•28(30):7520–7536NiellandStryker•ReceptiveFieldsinMouseVisualCortex
Figure10.Responsetocontrast-modulatednoisemovies.A,Framesfromacontrast-modulatedmovie.B,Spikehistograminresponsetocontrast-modulatedmovie,withmoviecontrastingraybelow.C,Powerspectrumofresponse,showingpeakatfundamentalfrequencyofcontrastmodulation(0.1Hz).D,Polarplotofresponsemodulationandphaseatthefundamentalfrequency(nϭ188).E,Histogramofresponsemodulationwithcontrast,fordifferentlayersandcelltypes,showsdecreasedresponsivenessindeeperlayers.inh,Putativeinhibitoryunits.
Indeed,someunitswereisolatedthatproducedonlyafewhun-dredspikesover1hofrecording.Traditionalrecordingtech-niquesthatsearchforresponsesaremostlikelytodetectunitswithhighfiringratesorlargeextracellularactionpotentials,whichinmicetendtobethelessselectivelayer5andputativeinhibitoryunits.Furthermore,becauseisolatedunitsweregen-erallystableoverthisperiod,wewereabletopresentalargenumberofstimuli,samplingabroadparameterspace,toevokeresponsesfromunitswitheitherveryselectiveorpoorresponses.WeverifiedlaminaridentityusingtheLFPfromgeometricallydefinedrecordingsitesestablishedbythesiliconprobes,con-firmedinseveralcasesbyretrospectivehistology.Althoughthelayersinmousecortexarenotasclearlydefinedasinotherspe-cies,withlessdistinctcytoarchitecturalboundariesandamoreextensivethalamicprojection(FrostandCaviness,1980;Anto-ninietal.,1999),wefoundcleardifferencesinresponsetypesbetweentheselaminas.
Furthermore,spikewaveformswereusedtosegregateunitsintonarrow-spikingandbroad-spikingcategories,whichhavebeenusedfrequentlyasaroughclassificationofinhibitoryversusexcitatoryneurons(Raoetal.,1999;BrunoandSimons,2002;Andermannetal.,2004;Mitchelletal.,2007).Thereisstillsomeuncertaintyastotheexactcelltypesrepresentedbythisdivision,inparticularwhetherthenarrow-spikingclassincludesjusttheparvalbumin-positivefast-spikingunitsandwhethersomesmallfractionofnarrow-spikingunitsmaybeexcitatory(Swadlow,2003).However,theprofounddifferenceinfunctionalresponseswefindbetweenthesegroupssuggeststhat,althoughtheirexactcategorizationmaybeuncertain,narrow-spikingneuronsareprimarilyadistinctclassfrombroad-spikingneurons.Ourdes-ignationofnarrow-spikingunitsasputativeinhibitoryneuronsisfurthersupportedbytherecentfinding,usingtwo-photoncal-ciumimaginginmouse,thatinhibitoryGABAergicneuronsaremuchlessorientationselectivethannon-GABAergicneurons(Sohyaetal.,2007).Itwillbeinterestingtoseewhethersuchadichotomyinorientationselectivityexistsinotherspeciesthathaveasignificantnumberofnonorientedunits.Tworecentstud-iesincat,inwhichalmostallunitsareorientationselective,didshowdecreasedorientationselectivityamongpresumedinhibi-toryneuronsinlayer4(Cardinetal.,2007;Nowaketal.,2008).Comparisonwithpreviousstudies
Ourresultsgenerallyagreewithpreviousstudiesinmice,partic-ularlythefindingofgreaterorientationselectivityinsuperficiallayers(ManginiandPearlman,1980;Metinetal.,1988),andthelargenumberofpoorlyorientation-selectiveunitswithlargere-ceptivefieldsinlayer5,whichhavebeensuggestedpreviouslytobecorticotectalneurons(ManginiandPearlman,1980).Ourfindingsarealsoinagreementwithrecentstudiesinotherro-dentssuchasrat(Girmanetal.,1999;Ohkietal.,2005)andsquirrel(VanHooseretal.,2005),confirmingthathighorienta-tionselectivitycanexistintheabsenceoflarge-scaleorganizationsuchasanorientationmap.Wedidnotobservemanyofthesimple,nonorientedresponsesinlayer4thathavebeendescribedinpreviousmousestudies(ManginiandPearlman,1980;Metinetal.,1988).Itislikelythatpreviousrecordingswithsharptung-stenelectrodesmorereadilyisolatedafferentsofLGNneurons,whichgenerallyhavelinearnonorientedresponses.Itisalsopos-siblethatsinusoidalgratingselicitgreaterselectivitythansmallspotsorflashedstimuli,usedinsomepreviousstudies.
Ourresultsarealsoconsistentwithpreviousfindinginthemousedorsallateralgeniculatenucleus(dLGN)ofthalamus,whichprovidesdirectinputtoV1.GrubbandThompson(2003)foundamedianpeakspatialfrequencyresponseof0.027cpd,similartoourfindingof0.035cpd.Ourmeasurementofpeaktemporalfrequencyof1.7Hz,obtainedwithcontrast-reversingratherthandriftinggratings,showsadecreasefrom3.8Hzmea-suredindLGN,similartothedecreaseintemporalfrequencyresponsefromthalamustocortexfoundinotherspecies(Hawkenetal.,1996).Wedofindslightlyhighertemporalfre-quencytuninginlayer4,consistentwithitsreceivingdirectinputfromthalamus.
Itisalsointerestingtocomparethesefindingsinmousewiththoseof“higher”specieswhosevisualsystemshavebeenmorethoroughlystudied(foranextensivecomparisonoftuningprop-ertiesacrossspecies,seeVanHooser,2007).Themostreadilyapparentdifferenceisinspatialscale,withmedianspatialfre-quencyinmouseof0.04cpdcomparedwith0.9cpdincatandupto4cpdinprimates.Thetotalpercentageoforientedunits(74%ofresponsiveunits)isslightlylower,withmanyotherspecieshaving80%ormore,althoughϾ95%ofunitsincatarereportedtobeorientationselective(VanHooser,2007).Withinthepop-
NiellandStryker•ReceptiveFieldsinMouseVisualCortexJ.Neurosci.,July23,2008•28(30):7520–7536•7533
Figure11.Receptivefieldsbyspike-triggeredaverage.A–C,Examplesofspatialreceptivefieldswithtwo,three,andonesubfield,respectively,showingvaryingorientation,ON/OFFcenters,andspacingofsubfields.D,ComparisonofpreferredorientationfromSTAreceptivefieldsversusorientationmeasuredfromgratings.E,ComparisonofreceptivefieldspatialfrequencyfromSTAversusgratingsshowsgoodcorrespondenceforunitswithspatialfrequencygreaterthan0.017cpd(blue)andpoorcorrespondenceforunitswiththelowestspatialfrequencies(gray).F,MedianwidthoforientationtuningforunitswithdifferingnumbersofsubunitsintheSTA.G,Medianbandwidthofspatialfrequencytuningforvaryingnumberofsubunits.H,ResidualerrorinfitofSTAtoGaborfunction.I,GeometricdescriptionofreceptivefieldstructureforunitsaccuratelyfitbyGaborscomparedwithmacaquefromRingach(2002).J,DistributionofspatialphaseinGaborfits,inwhich0°representsevensymmetryand90°representsoddsymmetry.
ulationoforientedunits,thewidthoftuningwasclosetothatofothermodelspecies,withmedianorientationtuninghalf-widthof28–29°formousecomparedwith19–25°incatand24°inmacaque(VanHooser,2007).Thespatialfrequencybandwidthof2.5octavesinmouseissomewhatlargerthanthevalueof1.5octavesincatandmacaquebutclosetoowlmonkey,2.1octaves,andsquirrel,2.3octaves(VanHooser,2007).Thus,evenatthelow-resolutionlimitofthemousevisualsystem,neuronscanstillbenearlyasselectivefororientationandspatialfrequencyasinspecieswithmoredevelopedvisualsystems.Aspointedoutpre-viously(VanHooser,2007),thesetuningwidthsseemtobefairlyinvariantacrossspeciesandmayrepresentcommonconstraintsoncorticalprocessing.
Onestrikingdifferenceinmiceistherelativepaucityofcom-plexcellsinlayer2/3,inwhichwefoundthat75%ofresponsiveunitsweresimple,whereasinotherspeciesthemajorityofunitsoutsideoflayer4arecomplex(VanHooser,2007).Itisinterest-ingtospeculatethatthismayberelatedtotheabsenceoforien-tationcolumnsinmouse.Mostmodelsofcomplexresponsesinvolvepoolinginputsofsimilarorientation,whichcanresultfromnearest-neighborconnectivityinthepresenceoforienta-tioncolumns.However,withoutcolumnaranatomicalorganiza-tion,greaterconstraintsonsynapticspecificitywouldbeneededtopreserveorientationselectivitywhilepoolinginputs.Itmustbenoted,however,thatthesquirrelhasalargeproportionofcom-plexcellsbutlacksanorientationmap(VanHooseretal.,2005).Therelativelackoforientationandphaseselectivityinputa-tiveinhibitoryneuronshasinterestingimplicationsforcorticalprocessing.Theywouldbewellsituatedtoprovidesignalssuchasdivisivenormalization(Carandinietal.,1997),balancedinhibi-
7534•J.Neurosci.,July23,2008•28(30):7520–7536Figure12.Contrast-invarianttuningandcontrast–responsecharacteristics.A,Schematicofcontrast-invariant(left)versuscontrast-dependent(right)orientationtuning.B,Exampleoforientationtuninginresponsetogratingsofdifferentcontrast,forunitinFigure2.C,Orientationtuningatdifferentcontrastsaveragedacross22orientation-selectiveunits.Tun-ingcurvesofeachunitwerealignedtopreferredorientationandnormalizedtomaximumresponsebeforeaveraging.D,Comparisonoforientationselectivityat50and100%contrastshowsnosystematicchange(nϭ32).E,Widthoftuning,fororientation-selectiveunits(nϭ22),at50and100%contrast.F,Valueofcontrastthatgavehalf-maximalresponse(nϭ32).G,Slopeofthecontrastresponsefunctionathalf-maximalresponse.
tionandexcitation(Chanceetal.,2002),untunedinhibitionforcontrast-invarianttuning(LauritzenandMiller,2003),andforcouplingnetactivitytometabolismorbloodflow(Buzsa´kietal.,2007).Conversely,manymodelsoforientationselectivityrequireorientation-selectiveinhibition(FersterandMiller,2000),andthenetinhibitoryinputtocatcorticalcellshasinfactbeenshowntobeorientationtuned(Andersonetal.,2000).Onlyasmallfractionofputativeinhibitoryunitswerecordedhavesufficientorientationselectivitytoprovidesuchasignal.Itispossiblethatthisfractionoforientednarrow-spikingunits,orperhapsinhib-itoryneuronsamongthebroad-spikingunits,servesthisdistinctrole;alternativelytunedinhibitionmaynotbenecessaryforcontrast-invariantorientationselectivity(Finnetal.,2007;PriebeandFerster,2008).
NiellandStryker•ReceptiveFieldsinMouseVisualCortex
Futuredirections
Thepresenceofstrongselectivityanddifferentialresponsesbylayerandcelltypeopensthedoorforadditionalstudyofcorticaldevelopmentandfunction,takingadvantageofrecenttechnicaladvancesusingmousegeneticstostudyneuralcircuitry.Forex-ample,viral-basedlabelsforsynapticconnectivity(Wickershametal.,2007)couldbeusedtoinvestigatetherelativetuningprop-ertiesofconnectedneuronsand,incombinationwithcell-typespecificmarkers,couldelucidatethemicrocircuitthatgivesrisetoselectiveresponses.Theuseofquantitativevisualanalysiswithmanipulationsofgenesinvolvedingrowthandsynapticplasticitymayelucidatethedevelopmentalprocessesthatresultinspecificreceptivefieldproperties,suchasorganizedONandOFFsubre-gionsresultinginorientationselectivityandsimpleversuscom-plexcelltypes.Finally,theabilitytomakeprecisemanipulationsofneuralactivityindefinedcellpopulations(Boydenetal.,2005;Tanetal.,2006;Zhangetal.,2007b)shouldallowdirectcausaltestsofmodelsofcorticalprocessing.
References
AndermannML,RittJ,NeimarkMA,MooreCI(2004)Neuralcorrelatesofvibrissaresonance;band-passandsomatotopicrepresentationofhigh-frequencystimuli.Neuron42:451–463.
AndersonJS,CarandiniM,FersterD(2000)Orientationtuningofinputconductance,excitation,andinhibitionincatprimaryvisualcortex.JNeurophysiol84:909–926.
AntoniniA,FagioliniM,StrykerMP(1999)Anatomicalcorrelatesoffunc-tionalplasticityinmousevisualcortex.JNeurosci19:4388–4406.
AtencioCA,SchreinerCE(2008)Spectrotemporalprocessingdifferencesbetweenauditorycorticalfast-spikingandregular-spikingneurons.JNeurosci28:3897–3910.Bartho´P,HiraseH,MonconduitL,ZugaroM,HarrisKD,Buzsa´kiG(2004)
Characterizationofneocorticalprincipalcellsandinterneuronsbynet-workinteractionsandextracellularfeatures.JNeurophysiol92:600–608.BoydenES,ZhangF,BambergE,NagelG,DeisserothK(2005)Millisecond-timescale,geneticallytargetedopticalcontrolofneuralactivity.NatNeu-rosci8:1263–1268.
BrainardDH(1997)Thepsychophysicstoolbox.SpatVis10:433–436.BrechtM,RothA,SakmannB(2003)Dynamicreceptivefieldsofrecon-structedpyramidalcellsinlayers3and2ofratsomatosensorybarrelcortex.JPhysiol(Lond)553:243–265.
BrechtM,FeeMS,GaraschukO,HelmchenF,MargrieTW,SvobodaK,OstenP(2004)Novelapproachestomonitorandmanipulatesingleneuronsinvivo.JNeurosci24:9223–9227.
BrunoRM,SimonsDJ(2002)Feedforwardmechanismsofexcitatoryandinhibitorycorticalreceptivefields.JNeurosci22:10966–10975.Buzsa´kiG,DraguhnA(2004)Neuronaloscillationsincorticalnetworks.Science304:1926–1929.Buzsa´kiG,KailaK,RaichleM(2007)Inhibitionandbrainwork.Neuron56:771–783.
CallawayEM(2005)Amolecularandgeneticarsenalforsystemsneuro-science.TrendsNeurosci28:196–201.
CangJ,KanekoM,YamadaJ,WoodsG,StrykerMP,FeldheimDA(2005)Ephrin-asguidetheformationoffunctionalmapsinthevisualcortex.Neuron48:577–589.
CarandiniM,HeegerDJ,MovshonJA(1997)Linearityandnormalizationinsimplecellsofthemacaqueprimaryvisualcortex.JNeurosci17:8621–8644.
CarandiniM,DembJB,ManteV,TolhurstDJ,DanY,OlshausenBA,GallantJL,RustNC(2005)Doweknowwhattheearlyvisualsystemdoes?JNeurosci25:10577–10597.
CardinJA,PalmerLA,ContrerasD(2007)Stimulusfeatureselectivityinexcitatoryandinhibitoryneuronsinprimaryvisualcortex.JNeurosci27:10333–10344.
ChanceFS,AbbottLF,ReyesAD(2002)Gainmodulationfrombackgroundsynapticinput.Neuron35:773–782.
ChichilniskyEJ(2001)Asimplewhitenoiseanalysisofneuronallightre-sponses.Network12:199–213.
ConnorsBW,KriegsteinAR(1986)Cellularphysiologyoftheturtlevisual
NiellandStryker•ReceptiveFieldsinMouseVisualCortex
cortex:distinctivepropertiesofpyramidalandstellateneurons.JNeuro-sci6:164–177.
DeAngelisGC,OhzawaI,FreemanRD(1993)Spatiotemporalorganizationofsimple-cellreceptivefieldsinthecat’sstriatecortex.I.Generalcharac-teristicsandpostnataldevelopment.JNeurophysiol69:1091–1117.
DeWeeseMR,WehrM,ZadorAM(2003)Binaryspikinginauditorycor-tex.JNeurosci23:7940–7949.
DouglasRJ,MartinKA(2004)Neuronalcircuitsoftheneocortex.AnnuRevNeurosci27:419–451.Dra¨gerUC(1975)Receptivefieldsofsinglecellsandtopographyinmousevisualcortex.JCompNeurol160:269–290.
FagioliniM,PizzorussoT,BerardiN,DomeniciL,MaffeiL(1994)Func-tionalpostnataldevelopmentoftheratprimaryvisualcortexandtheroleofvisualexperience:darkrearingandmonoculardeprivation.VisionRes34:709–720.
FagioliniM,KatagiriH,MiyamotoH,MoriH,GrantSG,MishinaM,HenschTK(2003)SeparablefeaturesofvisualcorticalplasticityrevealedbyN-methyl-D-aspartatereceptor2Asignaling.ProcNatlAcadSciUSA100:2854–2859.
FagioliniM,FritschyJM,Lo¨wK,Mo¨hlerH,RudolphU,HenschTK(2004)
SpecificGABAAcircuitsforvisualcorticalplasticity.Science303:1681–1683.
FengG,MellorRH,BernsteinM,Keller-PeckC,NguyenQT,WallaceM,NerbonneJM,LichtmanJW,SanesJR(2000)ImagingneuronalsubsetsintransgenicmiceexpressingmultiplespectralvariantsofGFP.Neuron28:41–51.
FersterD,MillerKD(2000)Neuralmechanismsoforientationselectivityinthevisualcortex.AnnuRevNeurosci23:441–471.
FieldDJ,TolhurstDJ(1986)Thestructureandsymmetryofsimple-cellreceptive-fieldprofilesinthecat’svisualcortex.ProcRSocLondBBiolSci228:379–400.
FinnIM,PriebeNJ,FersterD(2007)Theemergenceofcontrast-invariantorientationtuninginsimplecellsofcatvisualcortex.Neuron54:137–152.FrostDO,CavinessVSJr(1980)Radialorganizationofthalamicprojectionstotheneocortexinthemouse.JCompNeurol194:369–393.
GilbertCD(1983)Microcircuitryofthevisualcortex.AnnuRevNeurosci6:217–247.GirmanSV,Sauve´Y,LundRD(1999)Receptivefieldpropertiesofsingleneuronsinratprimaryvisualcortex.JNeurophysiol82:301–311.
GrubbMS,ThompsonID(2003)Quantitativecharacterizationofvisualresponsepropertiesinthemousedorsallateralgeniculatenucleus.JNeu-rophysiol90:3594–3607.
HarrisKD,HenzeDA,CsicsvariJ,HiraseH,Buzsa´kiG(2000)Accuracyoftetrodespikeseparationasdeterminedbysimultaneousintracellularandextracellularmeasurements.JNeurophysiol84:401–414.
HasenstaubA,ShuY,HaiderB,KraushaarU,DuqueA,McCormickDA(2005)Inhibitorypostsynapticpotentialscarrysynchronizedfrequencyinformationinactivecorticalnetworks.Neuron47:423–435.
HawkenMJ,ParkerAJ(1987)Spatialpropertiesofneuronsinthemonkeystriatecortex.ProcRSocLondBBiolSci231:251–288.
HawkenMJ,ShapleyRM,GrosofDH(1996)Temporal-frequencyselectiv-ityinmonkeyvisualcortex.VisNeurosci13:477–492.
HeimelJA,VanHooserSD,NelsonSB(2005)Laminarorganizationofre-sponsepropertiesinprimaryvisualcortexofthegraysquirrel(Sciuruscarolinensis).JNeurophysiol94:3538–3554.
HenschTK(2005)Criticalperiodplasticityinlocalcorticalcircuits.NatRevNeurosci6:877–888.
HoferSB,Mrsic-FlogelTD,BonhoefferT,Hu¨benerM(2006)Lifelonglearning:oculardominanceplasticityinmousevisualcortex.CurrOpinNeurobiol16:451–459.
HuangZJ,KirkwoodA,PizzorussoT,PorciattiV,MoralesB,BearMF,MaffeiL,TonegawaS(1999)BDNFregulatesthematurationofinhibitionandthecriticalperiodofplasticityinmousevisualcortex.Cell98:739–755.HubelDH,WieselTN(1962)Receptivefields,binocularinteractionandfunctionalarchitectureinthecat’svisualcortex.JPhysiol(Lond)160:106–154.Hu¨benerM(2003)Mousevisualcortex.CurrOpinNeurobiol13:413–420.IssaNP,TrepelC,StrykerMP(2000)Spatialfrequencymapsincatvisualcortex.JNeurosci20:8504–8514.
JeonCJ,StrettoiE,MaslandRH(1998)Themajorcellpopulationsofthemouseretina.JNeurosci18:8936–8946.
J.Neurosci.,July23,2008•28(30):7520–7536•7535
JonesJP,PalmerLA(1987a)Thetwo-dimensionalspatialstructureofsim-plereceptivefieldsincatstriatecortex.JNeurophysiol58:1187–1211.JonesJP,PalmerLA(1987b)Anevaluationofthetwo-dimensionalGaborfiltermodelofsimplereceptivefieldsincatstriatecortex.JNeurophysiol58:1233–1258.
KalatskyVA,StrykerMP(2003)Newparadigmforopticalimaging:tempo-rallyencodedmapsofintrinsicsignal.Neuron38:529–545.
KarpovaAY,TervoDG,GrayNW,SvobodaK(2005)Rapidandreversiblechemicalinactivationofsynaptictransmissioningeneticallytargetedneurons.Neuron48:727–735.
LauritzenTZ,MillerKD(2003)Differentrolesforsimple-cellandcomplex-cellinhibitioninV1.JNeurosci23:10201–10213.
LeeSH,LandPW,SimonsDJ(2007)Layer-andcell-type-specificeffectsofneonatalwhisker-trimminginadultratbarrelcortex.JNeurophysiol97:4380–4385.
ManginiNJ,PearlmanAL(1980)Laminardistributionofreceptivefieldpropertiesintheprimaryvisualcortexofthemouse.JCompNeurol193:203–222.
McCormickDA,ConnorsBW,LighthallJW,PrinceDA(1985)Compara-tiveelectrophysiologyofpyramidalandsparselyspinystellateneuronsoftheneocortex.JNeurophysiol54:782–806.
MechlerF,RingachDL(2002)Ontheclassificationofsimpleandcomplexcells.VisionRes42:1017–1033.Me´tinC,GodementP,ImbertM(1988)Theprimaryvisualcortexinthemouse:receptivefieldpropertiesandfunctionalorganization.ExpBrainRes69:594–612.
MitchellJF,SundbergKA,ReynoldsJH(2007)Differentialattention-dependentresponsemodulationacrosscellclassesinmacaquevisualareaV4.Neuron55:131–141.
MitzdorfU(1985)Currentsource-densitymethodandapplicationincatcerebralcortex:investigationofevokedpotentialsandEEGphenomena.PhysiolRev65:37–100.
NakaKI,RushtonWA(1966)S-potentialsfromcolourunitsintheretinaoffish(Cyprinidae).JPhysiol(Lond)185:536–555.
NaseG,SingerW,MonyerH,EngelAK(2003)Featuresofneuronalsyn-chronyinmousevisualcortex.JNeurophysiol90:1115–1123.
NowakLG,Sanchez-VivesMV,McCormickDA(2008)Lackoforientationanddirectionselectivityinasubgroupoffast-spikinginhibitoryinterneu-rons:cellularandsynapticmechanismsandcomparisonwithotherelec-trophysiologicalcelltypes.CerebCortex18:1058–1078.
OhkiK,ReidRC(2007)Specificityandrandomnessinthevisualcortex.CurrOpinNeurobiol17:401–407.
OhkiK,ChungS,Ch’ngYH,KaraP,ReidRC(2005)Functionalimagingwithcellularresolutionrevealsprecisemicro-architectureinvisualcor-tex.Nature433:597–603.
PelliDG(1997)TheVideoToolboxsoftwareforvisualpsychophysics:transformingnumbersintomovies.SpatVis10:437–442.
PolleuxF,Ince-DunnG,GhoshA(2007)Transcriptionalregulationofver-tebrateaxonguidanceandsynapseformation.NatRevNeurosci8:331–340.
PriebeNJ,FersterD(2008)Inhibition,spikethreshold,andstimulusselec-tivityinprimaryvisualcortex.Neuron57:482–497.
PruskyGT,WestPW,DouglasRM(2000)Behavioralassessmentofvisualacuityinmiceandrats.VisionRes40:2201–2209.
RaoSG,WilliamsGV,Goldman-RakicPS(1999)Isodirectionaltuningofadjacentinterneuronsandpyramidalcellsduringworkingmemory:evi-denceformicrocolumnarorganizationinPFC.JNeurophysiol81:1903–1916.
RashBG,GroveEA(2006)Areaandlayerpatterninginthedevelopingce-rebralcortex.CurrOpinNeurobiol16:25–34.
RingachDL(2002)Spatialstructureandsymmetryofsimple-cellreceptivefieldsinmacaqueprimaryvisualcortex.JNeurophysiol88:455–463.RingachDL(2004)Mappingreceptivefieldsinprimaryvisualcortex.JPhysiol(Lond)558:717–728.
RingachDL,SapiroG,ShapleyR(1997)Asubspacereverse-correlationtechniqueforthestudyofvisualneurons.VisionRes37:2455–2464.RingachDL,ShapleyRM,HawkenMJ(2002)Orientationselectivityinma-caqueV1:diversityandlaminardependence.JNeurosci22:5639–5651.SatoTR,GrayNW,MainenZF,SvobodaK(2007)Thefunctionalmicroar-chitectureofthemousebarrelcortex.PLoSBiol5:e189.
SawtellNB,FrenkelMY,PhilpotBD,NakazawaK,TonegawaS,BearMF
7536•J.Neurosci.,July23,2008•28(30):7520–7536
(2003)NMDAreceptor-dependentoculardominanceplasticityinadultvisualcortex.Neuron38:977–985.
Schmitzer-TorbertN,JacksonJ,HenzeD,HarrisK,RedishAD(2005)Quantitativemeasuresofclusterqualityforuseinextracellularrecord-ings.Neuroscience131:1–11.
ShapleyR,HawkenM,RingachDL(2003)Dynamicsoforientationselec-tivityintheprimaryvisualcortexandtheimportanceofcorticalinhibi-tion.Neuron38:689–699.
SharpeeT,RustNC,BialekW(2004)Analyzingneuralresponsestonaturalsignals:maximallyinformativedimensions.NeuralComput16:223–250.SkottunBC,DeValoisRL,GrosofDH,MovshonJA,AlbrechtDG,BondsAB(1991)Classifyingsimpleandcomplexcellsonthebasisofresponsemodulation.VisionRes31:1079–1086.
SnellingsA,AndersonDJ,AldridgeJW(2006)Improvedsignalandreducednoiseinneuralrecordingsfromclose-spacedelectrodearraysusinginde-pendentcomponentanalysisasapreprocessor.JNeurosciMethods150:254–264.
SohyaK,KameyamaK,YanagawaY,ObataK,TsumotoT(2007)GABAer-gicneuronsarelessselectivetostimulusorientationthanexcitatoryneu-ronsinlayerII/IIIofvisualcortex,asrevealedbyinvivofunctionalCa2ϩimagingintransgenicmice.JNeurosci27:2145–2149.
SompolinskyH,ShapleyR(1997)Newperspectivesonthemechanismsfororientationselectivity.CurrOpinNeurobiol7:514–522.
StosiekC,GaraschukO,HolthoffK,KonnerthA(2003)Invivotwo-photoncalciumimagingofneuronalnetworks.ProcNatlAcadSciUSA100:7319–7324.
SwadlowHA(2003)Fast-spikeinterneuronsandfeedforwardinhibitioninawakesensoryneocortex.CerebCortex13:25–32.
SwadlowHA,GusevAG,BezdudnayaT(2002)Activationofacorticalcol-umnbyathalamocorticalimpulse.JNeurosci22:7766–7773.
SykenJ,GrandpreT,KanoldPO,ShatzCJ(2006)PirBrestrictsocular-dominanceplasticityinvisualcortex.Science313:1795–1800.
NiellandStryker•ReceptiveFieldsinMouseVisualCortex
TamamakiN,YanagawaY,TomiokaR,MiyazakiJ,ObataK,KanekoT(2003)Greenfluorescentproteinexpressionandcolocalizationwithcal-retinin,parvalbumin,andsomatostatinintheGAD67-GFPknock-inmouse.JCompNeurol467:60–79.
TanEM,YamaguchiY,HorwitzGD,GosgnachS,LeinES,GouldingM,AlbrightTD,CallawayEM(2006)Selectiveandquicklyreversibleinac-tivationofmammalianneuronsinvivousingtheDrosophilaallatostatinreceptor.Neuron51:157–170.
TheunissenFE,DavidSV,SinghNC,HsuA,VinjeWE,GallantJL(2001)Estimatingspatio-temporalreceptivefieldsofauditoryandvisualneu-ronsfromtheirresponsestonaturalstimuli.Network12:289–316.
VanHooserSD(2007)Similarityanddiversityinvisualcortex:isthereaunifyingtheoryofcorticalcomputation?TheNeuroscientist13:639–656.VanHooserSD,HeimelJA,ChungS,NelsonSB,TothLJ(2005)Orienta-tionselectivitywithoutorientationmapsinvisualcortexofahighlyvisualmammal.JNeurosci25:19–28.
WaitesCL,CraigAM,GarnerCC(2005)Mechanismsofvertebratesynap-togenesis.AnnuRevNeurosci28:251–274.
WickershamIR,LyonDC,BarnardRJ,MoriT,FinkeS,ConzelmannKK,YoungJA,CallawayEM(2007)Monosynapticrestrictionoftranssyn-aptictracingfromsingle,geneticallytargetedneurons.Neuron53:639–647.
ZengH,ChattarjiS,BarbarosieM,Rondi-ReigL,PhilpotBD,MiyakawaT,BearMF,TonegawaS(2001)Forebrain-specificcalcineurinknockoutselectivelyimpairsbidirectionalsynapticplasticityandworking/episodic-likememory.Cell107:617–629.
ZhangF,AravanisAM,AdamantidisA,deLeceaL,DeisserothK(2007a)Circuit-breakers:opticaltechnologiesforprobingneuralsignalsandsys-tems.NatRevNeurosci8:577–581.
ZhangF,WangLP,BraunerM,LiewaldJF,KayK,WatzkeN,WoodPG,BambergE,NagelG,GottschalkA,DeisserothK(2007b)Multimodalfastopticalinterrogationofneuralcircuitry.Nature446:633–639.
因篇幅问题不能全部显示，请点此查看更多更全内容