colocalization-荧光共定位分析

MMccMMaasstteerrBBiioopphhoottoonniiccssFFaacciilliittyy▪▪DDeepptt..BBiioocchheemmiissttrryyaannddBBiioommeeddiiccaallSScciieenncceess▪▪HHSSCC44HH1144,,11220000MMaaiinnSSttrreeeettMMccMMaasstteerrUUnniivveerrssiittyy▪▪HHaammiillttoonn,,OONN▪▪▪▪▪▪ttccoolllliinnss@@mmaaccbbiioopphhooppttoonniiccss..ccaa2TableofcontentsIntroduction3Qualitative3Threshold-basedanalysis4Confounds4Photonshot-noise4Backgroundcorrection4Bleed-througha.k.a.Cross-talk4QualitativeColocalisation5DyeOverlay5Threshold-overlap5QuantitativeColocalisation6Scatterplots6Pearson’sCorrelationcoefficient-Rr7OverlapCoefficient-R7OverlapCoefficientsforredandgreen–kredandkgreen7ColocalisationCoefficients–Mred,Mgreen7IntensityCorrelationQuotient–ICQ8Confoundsforcolocalisationanalysis9Bleed-through(a.k.a.cross-talk)9Backgroundcorrection9Poorchoiceofpartners9Examples10No-colocalisation14Highred-greencolocalisation15Littlegreenstaining,allcolocalisedwithred16Green50%colocalisedwithred;RedhardlycolocalisedwithGreen17Randomlocalisation(duplicaterotated90°-right)18Random,partiallocalisation19References20MMccMMaasstteerrBBiioopphhoottoonniiccssFFaacciilliittyy▪▪▪▪ttccoolllliinnss@@mmaaccbbiioopphhooppttoonniiccss..ccaa3IntroductionColocalisationanalysisinvolvescomparisonofthespatiallocalisationoftwoproteins.Thesetwoproteinshaveeitherbeenrenderedintrinsicallyfluorescentbymolecular-engineeringorindirectlyfluorescentusingfluorescenceimmunocytochemistry.Weareultimatelyinterestedinwhetherthesetwoproteinsoccupythesamevolumeofinterest.Forsomeexperiments,thevolumeofinterestmaybethesamecell–arethetwoproteinsexpressedinthesamecell?Ifthisisthecasethenthereisnobetterreplacementformanualcountingof“double-positivecells”.Fortissues,stereologicalapproachestodeterminenumberswouldbeadvised.Inothercaseswemaywanttoknowifthetwoproteinsoccurinthesameprotein-complexorsubcellularcompartment.Inthiscasewearelimitedbytheresolutionofthelightmicroscopeand,inabestcasescenario,canonlysaywhetherthetwoproteinsoccupythesame200×200×400nmvolume–orvoxel.Ifwewanthigherresolution,wewouldhavetousethetechniqueofFRETwhichcanresolveinteractionsofproteins7nmapart.Ascanbeseenfromthissimplecalculation,evenifwecanprovetheproteinsoccupythesamevoxel,thisdoesnotnecessarilymeantheyareinteracting;theymaybeupto400nmapart!Therearemanytechniquesusedtocalculateimagecorrelation.TheycanbroadlybedividedintoQualitative;Threshold-basedandIntensity-basedapproaches.Threshold-basedapproachesuseonlypixelsaboveacertainthresholdintheircalculationswhileintensitybasedapproachesuseallthepixelsregardlessoftheirintensity.Oftenthethreshold-basedanalysesaremoreintuitive,notrequiringexpertknowledgeofaparticularcoefficientandoftenreportedasaneasilycomprehended“percentagecolocalisation”value.However,sincetheuserhastodefinethethresholdtheanalysesmaybesubjecttobias.Therearealgorithmswhichwillautomaticallycalculatetheappropriatethresholdforanimagebasedontheimage’sintensityhistogram.However,theimage’shistogrammayvarytestandcontrolsamples.Intensity-basedanalysisremovesuserbiasbyanalysingallthepixelsbasedonoftheirintensity,however,itmustbenotedthatsomeauthorsconsiderthisadrawbackratherthananadvantageduetotheintrinsicuncertaintyofpixelintensity(Lachmanovichetal.,2003).IntensitybasedapproachesarealsosusceptibletothepotentialerrorsdescribedinsectionConfounds.QualitativeThisisoftensimplythepresentationofanimagewheretheredandgreenchannelshavebeenmergedresultinginsomeyellowpixelswhichitisclaimedtodemonstratecolocalisation.Oneissuewiththistypeofimageisthatitcannotbeseenby10%ofthepopulationwithred-greencolourblindness.Usingmagentainsteadofredimprovesthisandcombiningtheimagesascyan-magentausinga“difference”algorithmcanalsomakethecolocalisationmoreapparent(Demandolxetal.,1997).Thisapproachhastheadvantageofbeingeasilycomprehendedandproviding,ataglance,somespatialinformationofthecolocalisation.However,theamountof“yellow”isextremelysusceptibletoanyimageprocessingthatmodifiestheintensityhistogramoftheimages.Thissortofmergedimageshouldalwaysbeconsideredanaccompanimenttoaqualitativeanalysisofthetwochannels.MMccMMaasstteerrBBiioopphhoottoonniiccssFFaacciilliittyy▪▪▪▪ttccoolllliinnss@@mmaaccbbiioopphhooppttoonniiccss..ccaa4Threshold-basedanalysisKeytothistechniqueidtheuserdefinitionofthethreshold.Pixelsabovethisvalueareconsideredtobepositive,pixelsbelowthisvalueareconsideredtobenegative.Simplesegmentation(Fayetal.,1997)threshold=meanBG+2xstdevBGComplexTophatfilter/watershed/(Lachmanovichetal.,2003)Thissortofapproachseemstobebestsuitedtotwoimageswithdiscretepunctuatestaining.Oncethresholdedthetwochannelscanbeanalysedinanumberofways.Eithersimplythepercentageofredchannelabovethresholdthatcolocalisewithagreenpixelabovethreshold.Morecomplexanalysiscanalsobedone.Lachmanocivhcalculatedthecentreofmassforeachpunctaandconsid