AnatomyofaComparativeGeneExpressionStudy

ThegoalofcomparativecDNAhybridizationistocomparegenetranscriptionintwoormoredifferentkindsofcells.Wewilldescribesomeexperimentsofparticularinterest;however,thepossibilitiesforinformativecomparativetranscriptionstudiesarelimitedonlybytheinvestigator"simagination.

Tissue-specificGenes

Cellsfromtwodifferenttissues(say,cardiacmuscleandprostateepithelium)arespecializedforperformingdifferentfunctionsinanorganism.Althoughwecanrecognizecellsfromdifferenttissuesbytheirphenotypes,itisnotknownjustwhatmakesonecellfunctionassmoothmuscle,anotherasaneuron,andstillanotherasprostate.Ultimately,acell"sroleisdeterminedbytheproteinsitproduces,whichinturndependonitsexpressedgenes.Comparativehybridizationexperimentscanrevealgeneswhicharepreferentiallyexpressedinspecifictissues.Someofthesegenesimplementthebehaviorsthatdistinguishthecell"stissuetype,whileothercontrollinggenesmakesurethatthecellonlyperformsthefunctionsforitstype.

RegulatoryGeneDefectsinCancer

Geneticdiseaseisoftencausedbygeneswhichareinappropriatelytranscribed--eithertoomuchortoolittle--orwhicharemissingaltogether.Suchdefectsareespeciallycommonincancers,whichcanoccurwhenregulatorygenesaredeleted,inactivated,orbecomeconstitutivelyactive.Unlikesomegeneticdiseases(e.g.cysticfibrosis)inwhichasingledefectivegeneisalwaysresponsIBLe,cancerswhichappearclinicallysimilarcanbegeneticallyheterogeneous.Forexample,prostatecancer(prostaticadenocarcinoma)maybecausedbyseveraldifferent,independentregulatorygenedefectseveninasinglepatient.Inagroupofprostatecancerpatients,everyonemayhaveadifferentsetofmissingordamagedgenes,withdifferingimplicationsforprognosisandtreatmentofthedisease.

Comparativehybridizationcanservetwopurposesinstudyingcancer:itcanpinpointthetranscriptiondifferencesresponsibleforthechangefromnormaltocancerouscells,anditcandistinguishdifferentpatternsofabnormaltranscriptioninheterogeneouscancers.Understandingthediversebasisofacanceriscrucialforinventingtherapiestargetedtothedifferentvarietiesofthedisease,sothateachpatientreceivesthemostappropriateandeffectivetreatment.

Cancersarecommonexamplesofgeneticallyheterogeneousdiseases,buttheyarebynomeanstheonlyones.Diabetes,heartdisease,andmultiplesclerosisareamongthediseasesforwhichgeneticriskfactorsareknowntobeheterogeneous.

CellularResponsestotheEnvironment

Howdoesacelladapttochangesinitsenvironment?CellssurviveinthefaceofchangesintemperatureandpH,changingnutrientavailABIlity,andthepresenceofenvironmentaltoxinsandionizingrADIation.Usually,achangeinenvironmentrequiresthatexpressionofsomegenesbeturnedupordownsothattheorganismcanrespondappropriately.Forexample,commonyeasthasbeenextensivelystudiedtounderstandhowitswitchesbetweenmetabolizingsugarsintoethanolandethanol,inturn,intoaceticacid(thisiswhywinewithactiveyeasteventuallybecomesvinegar).Themovefromonemetabolicstatetotheother,calleddiauxicshift,involvesshuttingdowngenesforprocessingsugarsandactivatingothersforprocessingethanol,aswellasageneralstressresponseduetothegreaterdifficultyofderivingenergyfromethanol.

Comparativehybridizationexperimentscanpointoutgeneswhosetranscriptionchangesinresponsetoanenvironmentalstimulus.Inthesimplestexperiment,apopulationofcellsissubjectedtothestimulusandallowedtoreachasteadystateoftranscription.Transcriptionlevelsinthealteredcellscanthenbecomparedtothoseinacontrolpopulation.Amoreinformativeexperimentsubjectscellstoachange,thentakessamplesofthecellpopulationatsuccessivepointsintime.Inthisway,theexperimentercanwatchasthegenetranscriptionpatternschangefromtheoldtothenewsteadystate.Temporalstudiescanidentifynotonlygeneswhosetranscriptionchangesbutalsotheorderofthechanges,providingevidenceaboutwhichgenescontroltheresponsedirectlyandwhichareonlyindirectlyaffectedbyit.

Environmentalchangesofinterestalsoincludetheintroductionofsignalingmolecules,suchashormones,interleukins,andinterferons,aswellastheactionsofdrugs.Allthesemoleculesstimulateachangeinacell"sbehavior(includingpossiblyitsdeath).Whilesomeofthechangesmaybemediatedpurelyattheproteinlevel,othersrequirenewtranscriptionwhichcanbedetectedbycomparativehybridization.

CellCycleVariations

Eveninastableenvironment,cellsundergoDNAreplication,mitosis,andeventuallydeath.Theseactivitiesrequirequitedifferentgeneproducts,suchasDNApolymerasesforgenomereplicationormicrotubulespindleproteinsformitosis.Acell"sgenesencodethe"programs"fortheseactivities,andgenetranscriptionisrequiredtoexecutethoseprograms.Comparativehybridizationcanbeusedtodistingishgenesthatareexpressedatdifferenttimesinthecellcycle.Inthisway,thepathwaysresponsibleforcontrollingbasiclifeprocessescanbeuncovered.

GeneswhichcodeforproteinaretranscribedintomessengerRNA"s(mRNA"s)inthecellnucleus.ThemRNA"sinturnaretranslatedintoproteinsbyribosomesinthecytoplasm.ThetranscriptionlevelofageneistakentobetheamountofitscorrespondingmRNApresentinthecell.ComparativehybridizationexperimentscomparetheamountsofmanydifferentmRNA"sintwocellpopulations.

TopreparemRNAforuseinamicroarrayassay,itmustbepurifiedfromtotalcellularcontents.mRNAaccountsforonlyabout3%ofallRNAinacell[1],soisolatingitinsufficientquantityforanexperiment(1-2micrograms)canbeachallenge.CommonmRNAisolationmethodstakeadvantageofthefactthatmostmRNA"shaveapoly-adenine(poly(A))tail.Thesepoly(A)⁺mRNA"scanbepurifiedbycapturingthemusingcomplementaryoligodeoxythymidine(oligo(dT))moleculesboundtoasolidsupport,suchasachromatographiccolumnoracollectionofmagneticbeads.

CapturedmRNA"sarestilldifficulttoworkwithbecausetheyarepronetobeingdestroyed.TheenvironmentisfullofRNA-digestingenzymes(therearesomeonyourfingers,yourkeyboard,yourmouse,andeveryotherexposedsurfacearoundyourightnow),sofreeRNAisquicklydegraded.Topreventtheexperimentalsamplesfrombeinglost,theyarereverse-transcribedbackintomorestableDNAform.TheproductsofthisreactionarecalledcomplementaryDNA"s(cDNA"s)becausetheirsequencesarethecomplementsoftheoriginalmRNAsequences.Thereversetranscriptionreactionusuallystartsfromthepoly(A)tailofthemRNAandmovestowarditshead;suchareactioniscalledoligo(dT)-primed.

AproblemwithcDNAproductionisthatnotallmRNA"sarereverse-transcribedwiththesameefficiency.Thisfactleadstoreversetranscriptionbias,whichcanchangetherelativeamountsofdifferentcDNA"smeasuredbythemicroarrayassay.ReversetranscriptionbiasisnotaproblemwhencomparingthesamemRNAacrosstwocellpopulationsunlessitcausesthemRNAnottobetranscribedatall.However,thebiasdoesprohibitquantitativecomparisonbetweendifferentmRNA"sononearray.AnotherproblemcausedbybiasisthatsomemRNA"smaybereverse-transcribedforonlypartoftheirlengths,makingthemlesslikelytobindandstayboundtotheircomplementsonthearray.OnewayofgettingaroundthisproblemistoprimereversetranscriptionfromrandomstartingpositionsonthemRNA"s,ratherthanalwaysstartingfromtheirtails.Thismethodcanreducebias,butitalsomakesuselesscDNAfromanyremainingribosomalandtransferRNA"sinthesample.

InordertodetectcDNA"sboundtothemicroarray,wemustlabelthemwithareportermoleculethatidentifiestheirpresence.Thereporterscurrentlyusedincomparativehybridizationtomicroarraysarefluorescentdyes(fluors),representedbytheredandgreencirclesattachedtothecDNA"sinthediagram[2].Adifferently-coloredfluorisusedforeachsamplesothatwecantellthetwosamplesapartonthearray.ThelabeledcDNAsamplesarecalledprobesbecausetheyareusedtoprobethecollectionofspotsonthearray.

Thecolorsofthefluorsinthediagramarejustforillusrtration.Theactualfluorsdonotshowtheircolorsunlessstimulatedwithaspecificfrequencyoflightbyalaser.Eventhen,thecolorsarenotdirectlyobserved;rather,thewavelengthoftheemittedlightisusedtotuneadetectorwhichmeasuresthefluorescence.Thechoiceofredandgreencolorsisreminiscentoftheemissionwavelengthsofcommonly-usedfluors,suchasrhodamineandfluoresceinorCy3andCy5.

ThenumberoffluormoleculeswhichlabeleachcDNAdependsonitslengthandpossiblyitssequencecomposition,bothofwhichareoftenunknown.ThisisonemorereasonthatfluorescentintensitiesfordifferentcDNA"scannotbequantitativelycompared.However,identicalcDNA"sfromthetwoprobesarestillcomparableaslongasthesamenumberoflabelmoleculesareaddedtothesameDNAsequenceineachprobe.

ToequalizethetotalconcentrationsofthetwocDNAprobesbeforeapplyingthemtothearray,theprobesolutionsaredilutedtohavethesameoverallfluorescentintensity.Thisproceduremakestwopossiblyunjustifiedassumptions:first,thatthetotalamountofmRNAineachcelltypebeingtestedisidentical;andsecond,thateachfluoremitsthesameamountoflightrelativetoitsconcentration.Thesecondassumptioncanbeeliminatedbysuitablecalibration,butthefirstoneisdifficulttocheck.ItmaythereforebethatcellswithmoreabundantmRNAaremadeintoaprobewithartificallylowmRNAconcentrations.

ThetwocDNAprobesaretestedbyhybridizingthemtoaDNAmicroarray.Thearrayholdshundredsorthousandsofspots,eachofwhichcontainsadifferentDNAsequence.IfaprobecontainsacDNAwhosesequenceiscomplementarytotheDNAonagivenspot,thatcDNAwillhybridizetothespot,whereitwillbedetectablebyitsfluorescence.Inthisway,everyspotonanarrayisanindependentassayforthepresenceofadifferentcDNA.ThereisenoughDNAoneachspotthatbothprobescanhybridizetoitatoncewithoutinterference.

MicroarraysaremadefromacollectionofpurifiedDNA"s.AdropofeachtypeofDNAinsolutionisplacedontoaspecially-preparedglassmicroscopeslidebyanarrayingmachine.Thearrayingmachinecanquicklyproducearegulargridofthousandsofspotsinasquareabout2cmonaside,smallenoughtofitunderastandardslidecoverslip.TheDNAinthespotsisbondedtotheglasstokeepitfromwashingoffduringthehybridizationreaction.

ThechoiceofDNA"stobeusedinthespotsonamicroarraydetermineswhichgenescanbedetectedinacomparativehybridizationassay.Fororganismswhosegenomeshavebeencompletelysequenced,includingseveralbacteriaandthetheyeastSaccharomycescerevisciae,itispossibletoarraygenomicDNAfromeveryknowngeneorsUSPectedopenreadingframe(ORF)intheorganism.EachgeneorORFisamplifiedfromtotalgenomicDNAbyPCR,producingenoughDNAtomakeunlimitednumbersofarrays.ThePatBrownLabatStanfordUniversityhasarrayedallknownorsuspectedgenesofS.cerevisciae(roughly6100)onasinglemicroarray.

Becausethehumangenomehasnotbeencompletelysequenced,wecannotyetproduceacomprehensivearrayforallitsgenes.Moreover,thenumberofhumangeneshasbeenestimatedatsomewherebetween10,000and100,000,soseveralarrayswillprobablyberequiredtoholdthemall.Despitetheselimitations,severalstrategiescanbeusedtodaytomakearraysforstudyinghumangenes.Wedoknowthelocationandsequenceofquiteafewhumangenesnow,sothesamemethodusedtoarrayyeastgeneswillproduceatleastapartialhumangenomearray.TherearetwootherwaystoproducearrayableDNAevenforunknowngenes:amplifycloneinsertsfromhumancDNAlibraries,orsynthesizeoligonucleotidesdirectlyfromknownexpressedsequenceinformationsuchasEST"s.WhileneitherofthesemethodswillproduceDNA"sforeveryhumangene,bothcanyieldenoughdifferentexpressedsequencestomakesubstantialarrays.BothtypesofDNAhavebeenusedbeforeinarray-likeapplications:cDNAlibrarieswereusedforcomparativehybridizationbeforetheadventoffluorescentmicroarrays,whileoligonucleotidearraysareavailablecommerciallytodayfromAffymetrixCorporationforrapidresequencingofafewgenesimportanttoAIDSandsomecancers.

OncethecDNAprobeshavebeenhybridizedtothearrayandanylooseprobehasbeenwashedoff,thearraymustbescannedtodeterminehowmuchofeachprobeisboundtoeachspot.Theprobesaretaggedwithfluorescentreportermoleculeswhichemitdetectablelightwhenstimulatedbyalaser.Theemittedlightiscapturedbyadetector,eitheracharge-coupleddevice(CCD)oraconfocalmicroscope,whichrecordsitsintensity.Spotswithmoreboundprobewillhavemorereportersandwillthereforefluorescemoreintensely.

Eachofthetwofluorescentreporters(fluors)usedhasacharacteristicexcitationwavelength;onlylightofthiswavelengthwillcausethemoleculetofluoresce.Theemittedlighthasacharacteristicemissionwavelengthwhichisdifferentfromtheexcitationwavelength.Thedetectorfortheemittedfluorescencefromthearrayissensitivetotheemissionwavelengthbutfiltersouttheexcitationwavelength;inthisway,thefluorescentlightofinterestcanbeseparatedfromthelaserlightscatteredofftheslide.

Agoodpairoffluorsforacomparativehybridizationexperimentshouldhaveverydifferentemissionorexcitationwavelengths.Iftheemissionwavelengthsaredifferent,lightemittedfromthetwofluorscanbeselectivelyfilteredtomeasuretheamountemittedbyeachfluorseparately.Iftheexcitationwavelengthsaredifferent,thetwofluorscanbestimulatedandscannedoneatatime.Ifoneoftheseconditionsisnotmet,thescannedintensitiescanbecontaminatedbycrosstalkbetweenthetwofluorescentchannels.

AlthoughthepurposeofthescanneristopickuplightemittedbyprobecDNA"sboundtotheircomplementaryspots,italsorecordslightfromafewmoleculesthathybridizedeithertothewrongspotornonspecificallytotheglassslide.Thisextralightbecomesthebackgroundofthescannedarrayimage.Oneadvantageofcurrentmicroarraytechnologyoveritspredecessorsisthatthebackgroundisextremelylow;consequently,thesignal-to-noiseratioofthescanneddatacanbequitehigh.

Theendproductofacomparativehybridizationexperimentisascannedarrayimage.Asmallpieceofsuchanimageisshownabove.Themeasuredintensitiesfromthetwofluorescentreportershavebeenfalse-coloredredandgreenandoverlaid.YellowspotshaveroughlyequalamountsofboundcDNAfromeachcellpopulationandsohaveequalintensityintheredandgreenchannels(red+green=yellow).SpotswhosemRNA"sarepresentatahigherlevelinoneortheothercellpopulationshowupaspredominantlyredorgreen.

Theintensitiesprovidedbythearrayimagecanbequantifiedbymeasuringtheaverageorintegratedintensitiesofthespots.TheratiooffluorescentintensitiesforaspotisinterpretedastheratioofconcentrationsforitscorrespondingmRNAinthetwocellpopulations.Schenaetal(1996)havedemonstratedtheabilitytodetectquantitativechangesofaslittleasafactoroftwo,withreasonableagreementbetweenexpressionratiosmeasuredonthearrayandratiosmeasuredbyanalternateformofRNAblotting.

Interpretingthedatafromamicroarrayexperimentcanbechallenging.Quantitationoftheintensitiesoneachspotissubjecttonoisefromirregularspots,dustontheslide,andnonspecifichybridization.Decidingtheintensitythresholdbetweenspotsandbackgroundcanbedifficult,especiallywhenthespotsfadegraduallyaroundtheiredges.Detectionefficiencymightnotbeuniformacrosstheslide,leadingtoexcessiveredintensityononesideofthearrayandexcessivegreenontheotherside.Evenafterovercomingdetectionandcalibrationproblems,themeasuredintensitiesforeachspotonlyrepresenttheratioofcDNA"sineachcellpopulation.LowlevelsofcDNAduetoreversetranscriptionbias,sampleloss,oraninherentlyraremRNAcancauselargeuncertaintiesintheseratios.

Numeroussoftwarepackages,bothfreeandcommercial,existforquantitatingmicroarraydata.Ihavedevelopedonesuchprogram,Dapple,toaddresssomeoftheabovementionedimagequalityissues,inthehopethatitsmethodsmightbeintegratedintootherarrayquantitationsystems.

Typically,theinterpretedarraydatawillhighlightarelativelysmallnumberofspotsrepresentingverydifferentially-expressedmRNA"swhosegenesdeservefurtherinvestigation.Alternatively,theoverallpatternofexpressioncanbeusedasa"fingerprint"tocharacterizespecificcelltypes(e.g.differentclassesoftumors),evenifnotallthedifferentially-expressedgenesonanarrayhavebeenidentified.