munityhaveresultedindirectfunctionalanalysisofonlyabout1000genestodate(5),thismayseemlikeatallorder.However,itappearslikelythattheefficiencygainedby“reversegenetics”rgecollectionsofinsertionmutantsareavailableforArabidopsis,maize,petunia,andsnapdragon,andcollectionsofinsertionmutantswillprobablybecreatedinseveralotherspeciesincludingrice.Thesecollectionscanbescreenedforaninsertionalinactivationofanygenebyusingthepoly-merasechainreaction(PCR)primedwitholigonucleotidesbasedonthesequencesofthetargetgeneandtheinsertionalmutagen(3,17).ThepresenceofaninsertioninthetargetgeneisindicatedbythepresenceofaPCRproduct.BymultiplexingDNAsam-ples,hundredsofthousandsoflinescanbescreenedandthecorrespondingmutantplantscanbeidentifiedwithrelativelysmalleffort.Inaddition,severalgroupsareembarkingonsequencingthegenomicDNAflankingalargenumberofinsertionssothataninsertioninvirtuallyanygenecanbeidentifiedbyacomputersearch(2,18).Analysisofthephe-notypeandotherpropertiesofthecorre-spondingmutantwillfrequentlyprovideaninsightintothefunctionofthegene.

AmajorlimitationtotheanalysisofgenefunctionbymutationisthatahighdegreeofgeneduplicationisapparentinArabidopsis(16)andis,therefore,probablyacommonfeatureofplantgenomes.BecausemanyofthegeneduplicationsinArabidopsisareverytightlylinked,itusuallywillnotbefeasibletoproducedoublemutantsbygeneticrecom-bination.Apossiblesolutionmaybetousehomologousrecombinationtoeliminatetan-demgenessimultaneouslybygenereplace-ment(19).Alternatively,amethodforpro-ducingpointmutationsbyusingRNA-DNAhybridsmaybeuseful(20).Bytargetingmutagenicsequencesthatintroducestopco-donstoregionsthatareconservedamongallduplicatesitmaybepossibletogenerateallcombinationsofnullmutationsinallmem-bersofamultigenefamilyfromoneexperi-ment.Althoughthesemethodsareamenabletoagene-by-geneapproach,theyarenotwellsuitedtoahighthroughputapproachbecauseoflowefficiencyorbecauseofthenecessityofregeneratingplantsfromsingleculturedcells.BecauseoftheeasewithwhichlargenumbersoftransgenicArabidopsisplantscanbegeneratedbyinfectingflowerswithAgrobacteriumtumefacienscontaininganin-sertionalmutagen(21),amethodofgenesilencingbasedonproducingdouble-strand-edRNAfrombidirectionaltranscriptionofgenesintransgenicplantsmaybebroadlyusefulforhigh-throughputgeneinactivation(22).Thismethodcould,inprinciple,bede-signedtousepromotersthatareexpressedinonlyafewcelltypesorataparticulardevel-opmentalstageorinresponsetoanexternalstimulus.Thiscouldsignificantlyobviateproblemsassociatedwiththelethalityofsomemutations.Althoughthemechanismisnotyetunderstood,itbearssomeresem-blancetodouble-strandedRNA–mediatedgenesilencinginnematodes(23).

Formanyapplications,particularlyinspe-ciesotherthanArabidopsiswhereproductionoftensofthousandsoftransformantsisslowandtime-consuming,virus-inducedgenesi-lencingmaybethemostfacilemethodforsuppressinggenefunction(24).Thismethodexploitsthefactthatsomeorallplantshaveasurveillancesystemthatcanspecificallyrecognizeviralnucleicacidsandmountasequence-specificsuppressionofviralRNAaccumulation.Byinoculatingplantswitharecombinantviruscontainingpartofaplantgene,itispossibletorapidlysilencetheendogenousplantgene(25).

Itisexpectedthatapplicationoftheseandrelatedmethodswillleadtotheassignmentofsomedegreeofgenefunctiontoallgenesinthebasicangiospermgenomewithinthenextdecade.Inaddition,parallelstudiesofthefunctionofgenesinother,nonplant,organ-ismswillcontributeagreatdealtounder-standinggenefunction.Thiscomprehensiveapproachtounderstandinggenefunctionwillgreatlyfacilitatethecreationofplantim-provementsthatarebasedonknowledgeoftheentiresysteminsteadofonagene-by-genebasis.

rolesofthegeneaffectedbythemutation(27).

Thesedatabasesofgeneexpressioninfor-mationwillprovideinsightsintothe“path-ways”ofgenesthatcontrolcomplexresponsesandwillbeafirststeptowardanecologyofthegenomeinwhichthegenomeisviewedasawholeandtherelationshipsofgeneproductstoeachotherwillbeconsideredfromatleastoneperspective(relativelevelofexpression).Per-hapsthetypesofmodelsthatecologistscurrent-lyuseforunderstandingtheinteractionsineco-systemswillproveuseful(28).Indeed,becausemicroarrayscanbemadeforanyorganismforwhichcomplementaryDNAscanbeisolateditseemslikelythatecologicalapplicationswillbefound.ItisnotnecessarytoknowthesequenceofthegenesonaDNAmicroarraybefore-hand—thiscanbedeterminedafterthearrayshavebeenusedtoidentifygenesthatmaybeofinterestbysomecriterion.

TheaccumulationofDNAmicroarrayorgenechipdatafrommanydifferentexperi-mentswillcreateapotentiallyverypowerfulopportunitytoassignfunctionalinformationtogenesofotherwiseunknownfunction.Thecon-ceptualbasisoftheapproachisthatgenesthatcontributetothesamebiologicalprocesswillexhibitsimilarpatternsofexpression.Thus,byclusteringgenesbasedonthesimilarityoftheirrelativelevelsofexpressioninresponsetodi-versestimuliordevelopmentalorenvironmen-talconditions,itshouldbepossibletoassignhypotheticalfunctionstomanygenesbasedontheknownfunctionofothergenesinthecluster(29).Workwithplantmicroarraysisjustbegin-ningbutthereiseveryreasontobelievethatthisapproachwillsoonbeastandardcomponentoftherepertoireofplantbiologists(30).Theprin-cipalchallenge,atpresent,istodevelopmeth-odsfordatabasingandinterrogatingthemas-siveamountsofdatathatresultfromthistypeofexperiment.

Oneofthemostimportantadvancesinplantimprovementwasthediscoveryofhybridvigorandtheexploitationofthisphenomenoninmodernbreedingprograms.Inspiteofexten-sivespeculationaboutthemechanisticbasisforhybridvigor,itispoorlyunderstood(31).Itwillbeveryinterestingtocomparewholegenomemicroarraysofinbredparentallineswiththeheterotichybrids.Wespeculatethatthehybridswillexhibitsignificantdifferencesintheex-pressionofclustersoffunctionallyrelatedgenesandthatdifferenthybridswillhavedif-ferentpatternsofexpression.Ifthisprovestobethecase,itmaybepossibletoprogresstowardmorepredictivedevelopmentofheterotichy-bridsbybreedingforcertainpatternsofgeneexpression.Itmayalsoprovideamuchneededlinkagebetweenthebreedingofdifferentplants—thatis,ifitisfoundthatvariationintheexpressionofcertainpathwaysorprocessesisassociatedwithenhancedyieldorqualityinonespecies,thismayprovidetestabletargetsfor

ImpactofGeneChipsandMicroarrays

Oneofthemostimportantexperimentalapproachesfordiscoveringthefunctionofgenespromisestobegenechipsandmi-croarrays.Inprinciple,DNAsequencesrepresentingallthegenesinanorganismcanbeplacedonminiaturesolidsupportsandusedashybridizationsubstratestoquantitatetheexpressionofallthegenesrepresentedinacomplexmRNAsample(26).Thus,wemayexpecttohaveexten-sivedatabasesofquantitativeinformationaboutthedegreetowhicheachgenere-spondstopathogens,pests,drought,cold,salt,photoperiod,andotherenvironmentalvariation.Similarly,wewillhaveextensiveinformationaboutwhichgenesrespondtochangesindevelopmentalprocessessuchasgerminationandflowering.Inaddition,wewillsoonknowwhichgenesrespondtothephytohormones,growthregulators,safeners,herbicides,andrelatedagrichemi-cals.Perhapslessobviously,wemayex-pecttohavesimilarinformationformanymutantsornaturalaccessionsthatdifferinsomewaythatcannotbereadilyassignedtogeneticvariationbyothercriteria.Knowl-edgeofwhichgenesexhibitchangesinexpressionofamutantofinterestwillbeusefulforformulatinghypothesesaboutthe