Criteria for effective design, construction, and gene knockdown by shRNA vectors

Abstract

Background

RNAinterference(RNAi)technologyisapowerfulmethodologyrecentlydevelopedforthespecificknockdownoftargetedgenes.RNAiismostcommonlyachievedeithertransientlybytransfectionofsmallinterfering(si)RNAoligonucleotides,orstablyusingshorthairpin(sh)RNAexpressedfromaDNAvectororvirus.MuchcontroversyhassurroundedthedevelopmentofrulesforthedesignofeffectivesiRNAoligonucleotides;andwhethertheserulesapplytoshRNAisnotwellcharacterized.

Results

TodeterminewhetherpublishedalgorithmsforsiRNAoligonucleotidedesignapplytoshRNA,weconstructed27shRNAsfrom11humangenesexpressedstablyusingretroviralvectors.Wedemonstrateanefficientmethodforpreparingwild-typeandmutantcontrolshRNAvectorssimultaneouslyusingoligonucleotidehybrids.WeshowthatsequencingthroughshRNAvectorscanbeproblematicduetotheintrinsicsecondarystructureofthehairpin,andwedetermineastrategyforeffectivesequencingbyusingacombinationofmodifiedBigDyechemistriesandDNArelaxingagents.Theefficacyofknockdownforthe27shRNAvectorswasevaluatedagainstsixpublishedalgorithmsforsiRNAoligonucleotidedesign.OurresultsshowthatnoneofthescoringalgorithmscanexplainasignificantpercentageofvarianceinshRNAknockdownefficacyasassessedbylinearregressionanalysisorROCcurveanalysis.Applicationofamodificationbasedonthestabilityofthe6centralbasesofeachshRNAprovidesfair-to-goodpredictionsofknockdownefficacyforthreeofthealgorithms.Analysisofanindependentsetofdatafrom38shRNAspooledfrompreviouspublicationsconfirmsthesefindings.

Conclusion

TheuseofmixedoligonucleotidepairsprovidesatimeandcostefficientmethodofproducingwildtypeandmutantcontrolshRNAvectors.TheadditiontosequencingreactionsofacombinationofmixeddITP/dGTPchemistriesandDNArelaxingagentsenablesreadthroughtheintrinsicsecondarystructureofproblematicshRNAvectors.SixpublishedalgorithmsforsiRNAoligonucleotidedesignthatweretestedinthisstudyshowlittleornoefficacyatpredictingshRNAknockdownoutcome.However,applicationofamodificationbasedonthecentralshRNAstabilityshouldprovideausefulimprovementtothedesignofeffectiveshRNAvectors.

Background

RNAinterference(RNAi)isanaturallyoccurringphenomenonbywhichRNAduplexesknownasshortinterferingRNA(siRNA)canreducegeneexpressionthroughenzymaticcleavageofatargetmRNAmediatedbytheRNA-inducedsilencingcomplex(RISC).TheabilityofsyntheticsiRNAtoinhibittargetedgeneswithnearspecificitymakesitanextremelypowerfultoolforfunctionalgenomicsthathasdrawnconsiderableinterestrecently[1,2].RNAiiscommonlyachievedbyintroducingchemicallysynthesizedsiRNA19–22mersintocellsbytransfection.However,manycellsandcelllinesareeitherrefractorytooradverselyaffectedbytransfection,andthetransientnatureofthismethodologyrendersitunsuitableforthegenerationoflong-termcelllinesofthedesirablephenotype.TwoalternativestosyntheticsiRNAareDNA-vectormediatedRNAiproduction[3-5],andmostrecentlyviral-mediatedsiRNAsynthesis[6-10].Forthelattertechnologies,senseandantisensestrandscanbeexpressedfromdifferentpromoters[11].Alternatively,shorthairpin(sh)RNAs,expressedfromasinglepromoter,areprocessedintosiRNAsbyDicerorahomologousdoublestrandRNase[12].

OnecaveatofsiRNAdesignisthatnotall19–22baseRNAduplexeswillcleavetheirtargetwithefficacy,andmuchefforthasgonetowardsidentifyingasetofrulesforselectinganeffectivesiRNAtargetsitewithinagene.Recentfindings[13,14]offeredthefirstcluetowardsthedevelopmentofguidelinesforselectingansiRNAtargetsite.ThesestudiesshowedthattheRISCcomplexisasymmetricandfavorsthestrandofthesiRNAduplexwiththeleastthermodynamicallystable5terminus.Subsequently,Reynoldsetal.designedanalgorithmbasedonstatisticaldatashowingpatternsofefficacyforsiRNAoligonucleotidescontainingspecificresiduesatdefinedpositionswithinthe19-mer[15].Alimitationoftheirstudyisthatasmallnumberofgenesweretested.SeveraladditionalalgorithmsfordesigningeffectivesiRNAshavebeenpublishedsincethoseinitialreportswithsurprisinglydisparateresults,makingthedeterminationofwhichresiduesaregenerallyfavorableforsiRNAefficacyapointofcontroversy[16-20].Additionally,whetheranyofthealgorithmsdevelopedforsyntheticsiRNAoligonucleotidesapplytothedesignofshRNAexpressedstablyfromavectorhasnotbeenwellexplored.

Inthepresentreport,weconstructandanalyzeasetof27shRNAsfor11differenthumangenes.ToourknowledgethisisthelargestindividualsetofdatapublishedforshRNA19-mers.WedescribeamethodforsimultaneouslypreparingwildtypeandcontrolmutantshRNAvectorsthatistimeandcostefficient,andshowthatsequencingofshRNAplasmidscanbequiteproblematicduetotheintrinsicsecondarystructureofthehairpin.WeexamineseveraldifferentstrategiesforovercomingthisproblemincludingtheuseofmodifiedBigDyechemistriesandtheadditionofagentsknowntorelaxDNAstructure.Theknockdownefficacyforeachofthe27shRNAswasevaluatedagainstsixpublishedalgorithmsforsiRNAoligonucleotidedesignbylinearregressionandROCcurveanalyses.Wedescribeamodificationofthreeofthealgorithmsthatprovidesfair-to-goodpredictionofshRNAefficacy,andconfirmthesignificanceofthemodifiedalgorithmsusingapooledsetofshRNAsfrompreviouspublications.ThesefindingsshouldbeofgeneralapplicabilityinthedesignandconstructionofshRNAvectors.

Resultsanddiscussion

DesignandpreparationofshRNAplasmids

ToaddressthequestionofhowshRNAsequencecorrelateswithknockdownefficacy,27shRNAvectorsfrom11differentgenesweredesignedandconstructed(Table1).TargetsequenceswereselectedinthecodingregionofeachgeneandweredesignedtobroadlyconformtotheseminalstudiesofsequencefeaturesforsiRNAoligomerefficacy[13-15].Accordingly,sequencesarelowinrunsandhaveaG/Cratioofabout50%.TheshRNAsweredesignedtotargetsitesthataredevoidofsinglenucleotidepolymorphisms,andcorrespondtoallsplicevariantsamplifiedbyourrealtimePCRprimersets.

Table1

ShRNAvectorspreparedforthisstudy

SincesiRNAscanhaveoff-targeteffects,itisimportantforfunctionalassaystomakeaspecificmutantwithoneormorebasemismatchwithinthetargetrecognitionsiteasacontrol[21].Toconservetimeandcost,wehavedevelopedamethodofmakingwild-typeandmutantshRNAvectorssimultaneously(detailedinMethodsandFigureFigure1).1).Geneknockdownresultsforfourwild-type/mutantshRNApairsareshowninFigureFigure2.2.TheseresultsdemonstratetheutilityofthismethodinprovidingapointmutantshRNAvectorthatcanserveasaloss-of-functioncontrolforgeneknockdownbywildtypeshRNAs.ThoughdetailedprotocolshavebeenpublishedforconstructionofshRNAvectors[22],thisisthefirstprotocolforproducingwild-typeandmutantvectorssimultaneouslyandshouldfacilitatetheimplementationofhighlycontrolledsystemforshRNA.

Figure1

Designforproducingwild-typeandmutantshRNAvectorssimultaneously.Aforwardstrandofthewild-typehairpin(blue)issynthesizedtogetherwithareversestrandcontainingaonebpmutationwithinboththesenseandantisensecopyofthetarget(more...)

Figure2

Geneexpressionanalysisforwild-typeandmutantshRNAvectorspreparedsimultaneouslyusingwild-type/mutantdoublestrandedhybrids.(A)Sequencesofthetargetsitesforfourwild-typeandmutantshRNAvectorsthatwerepreparedsimultaneouslyas(more...)

Strategyforaccuratesequencingthroughhairpinstructures

VerifyingthesequenceofanshRNAhairpinisessentialsincemismatchofevenonenucleotidewithinthetargetsequencecanablateknockdown(Figure(Figure22and[5,23].)AnissuethatisfrequentlyencounteredinthepreparationofshRNAvectorsisthatmanyaredifficulttosequenceduetotheintrinsicsecondarystructureofthehairpin.Onestrategyrecentlyproposedtoovercomethisissueinvolvesengineeringarestrictionsitewithintheloop/stemregionofthehairpintophysicallyseparatetheinvertedrepeatsbydigestion,andthenpiecingtogethersequenceusingsenseandantisenseprimers[24].However,theabilitytoachievesequencingofshRNAconstructswithoutmodifyingstem/loopsequencewouldbeofclearadvantage.Toaddressthispossibility,weevaluatedmodifiedsequencingreactionsforimprovementintheread-throughofthehairpinsecondarystructureinthreeshRNAhairpins.ModificationsincludeaddingagentsknowntorelaxDNAstructureincludingDMSO,Betaine,PCRxEnhancerandThermoFidelaseI;andaddingincreasingamountsofdGTPBigDyeterminator(dGTP)chemistrytothestandardBigDyev1.1(BD)chemistrywhichcontainsdITPratherthandGTP.

SequencingresultsforeachofthethreeDNAconstructsaresummarizedinTable2.Read-throughofthehairpinstructurewasmeasuredastheratioofthepeakheightabout300basesafterthehairpinstructuretothesignalabout50basesbeforethehairpinstructure.Aratioof1indicatesnolossinsignaland0indicatescompletelossofread-though.IntheabsenceofanyadditivetoBDchemistry,thehairpincausedareductioninpeakheightratioforourlesstightlystructuredhairpin,pHSPG-shmutTLR4,to0.4,andacompletelossinreadthroughfortheothertwoplasmids.ThiscanbevisualizedasanabruptstopinthesequencepeakprofileforpHSPG-shTLR4(Figure(Figure3A3A).

Table2

EvaluationofsequencingresultsofthreeDNAhairpinconstructs.Averageratioofpeakheightaftertobeforethehairpinregionwasdeterminedasameasureofhowwellthesequencereadthroughthehairpinstructure.Thegreaterthepeakheightratio,(more...)

Figure3

DNAsequencingofpHSPG-shTLR4usingmodifiedreactionconditions.DNAsequencingpeaksareshowninafullscaleviewwherebasepositionsareindicatedbytherowofnumbersineachpanelandtheYaxisisthesignalintensity.Sequencingreactionconditions(more...)

AmongtheDNArelaxingagents,5%DMSO,0.83MBetaineand1×PCRxEnhancereachimprovedthesequencereadsignificantlyforsomeconstructs.However,theadditionof0.83MBetaineplus1×PCRxEnhancertoBDchemistrywasfoundtosequencemostconsistently,withpeakheightratiosof0.5–0.9(Table2andFigureFigure3B).3B).Theadditionof10:1BD:dGTPchemistriesalonealsoimprovedreadthroughsomewhat,withpeakheightratiosof0.5–0.6(Table2andFigureFigure3C).3C).Thesub-optimalpeakheightratiofor10:1BD:dGTPcanbeattributedtoavisiblestepinthesequencepeakprofileafterthesecondarystructureregionwherethesignalisreduced(Figure(Figure3C,3C,arrow).IncreasingthedGTPchemistrycontentto5:1and3:1BD:dGTPorusingstraightdGTPchemistryincreasedthepeakheightratioandreducedthestepsomewhat(0.6to0.8ratio).However,themixedincorporationofdITPanddGTPresultedinworsepeakbroadeningastheamountofdGTPusedincreased[seeAdditionalfile1],anddGTPonlychemistrycausedseveresequencecompressions(datanotshown).ThebestoverallresultswereobservedbycombiningBetaineplusPCRxand10:1BD:dGTPmixedchemistriestogether.Thiscombinationreducedthestepwithlesspeakbroadeningandincreasedpeakheightratiosto0.9–1.0(Table2andFigureFigure3D).3D).ThermoFidelaseI,aDNAdestabililizingenzymethatisfrequentlyusedtoimprovesequencingofgenomicDNA[25,26],didnotimprovesequencingofanyofthethreehairpinsinstraightBDchemistry(datanotshown),andactuallyreducedthepeakheightratiosignificantlyin10:1BD:dGTPchemistriesforallthreeshRNAconstructs,causingthereappearanceofastopatthehairpinstructure(Table2andFigureFigure3E3E).

Insummary,thecombinationof10:1BD:GTPchemistries,0.83MBetaine,and1×PCRxEnhancerprovidedoptimalsequencing,andmixedBD:dGTPchemistries,Betaine,PCRxEnhancer,andDMSOeachhadsomepositiveeffectsontheirown.ThermoFidelaseI,however,probablyshouldbeavoidedforshRNAvectorswithdifficultintrinsicsecondarystructure.

CorrelationbetweenshRNAknockdownefficiencyandpublishedalgorithmsforsiRNAdesign

TodeterminewhethertheefficacyofknockdownbyshRNAvectorscorrelateswithpublishedrulesforthedesignofeffectivesiRNAoligonucleotides,shRNAswereevaluatedfortheirabilitytoknockdowngeneexpression.TheshRNAsweretransducedstablyintoeitherTHP1orJurkathumancelllinesasdetailedinTable3,firsttwoColumns.TheaverageknockdownwasdeterminedfromRNAcollectedonthreeormoredifferentdaysandislistedforeachshRNA(Column3).Knockdownwasshowntobereproducibleforcelllinesthatwereindependentlytransducedandsorted,suggestingthatknockdownisafunctionoftheshRNAtargetsequenceratherthanfeaturesoftheviraltransduction[seeAdditionalfile2].MorethanonethirdoftheshRNAvectorsconstructedwereunabletosuppresstranscription(<10%inColumn3),despitecomparablegrowthratesandlongtermexpressionoftheGFPMarkerathighlevelsinthesecelllines.Furthermore,greatvariationsinknockdownefficacyforseveralshRNAsmadeagainstmanyofthesamegenes(i.e.,CLR16.2,CLR19.3andTLR4)argueagainstanysimplebiologicalreasonsfordifferencesinefficacyforthesegenes.ManyoftheineffectiveshRNAshavenegative5ΔΔGvaluesandhighReynoldsscoring,eachwhichhavebeenhypothesizedtocorrelatewithsiRNAknockdownefficacy(Table3,Columns4and5)[13-15].Conversely,amongtheshRNAsthatwereabletoconfergeneknockdown,severalhadeitherpositive5ΔΔGvaluesorlowReynoldsscores.Thesefindingsindicatethat5ΔΔGandReynoldsscoringalgorithmforsiRNAmaynotprovidepositivecorrelativecriteriaforshRNAdesign.

Table3

ComparisonofknockdownefficacyandsiRNAdesignalgorithm.Averageknockdownwasmeasuredbyreal-timePCRoftriplicatesamples.Allaveragesareaccuratewithin10%SEM.AsterisksindicatehighTakasakietal.algorithmscoresthathavepoorcorresponding(more...)

TodeterminewhetherotherpublishedalgorithmsforsiRNAoligonucleotidedesigncanbeappliedtoshRNAvectors,eachoftheshRNAtargetsiteswasevaluatedbyfouradditionalalgorithms,andscoreswereplottedagainstthepercentknockdownforeachshRNA(Table3,Columns6–9andFig.Fig.4).4).ForeachalgorithmplotabestfitlinewasdrawnandtheR²valuecalculatedasanindicationofwhetherthevarianceinknockdownefficacycanbeexplainedbythealgorithmscoring.ResultsconfirmapoorassociationbetweenshRNAefficacyandeither5ΔΔG(freeenergydifferential)considerations[13]ortheReynoldsetal.algorithm[15],andalsodemonstrateapoorassociationwiththeHsiehetal.algorithm[19],witheachinfactshowingaweakreversecorrelationwiththedata.ThealgorithmsofAmarguizouietal.[20],Ui-Teietal.[18],andTakasakietal.[17],correlatedirectlywithshRNAefficacy.However,noneofthealgorithmscoresexplainasignificantpercentageofthevarianceinknockdownefficacy.Amongthealgorithmstested,theTakasakietal.scoringsystemshowsthehighestassociation,withanR²valueof0.0251.

Figure4

CorrelationbetweenshRNAknockdownefficacyandscoringforsixpublishedalgorithmsforsiRNA.AlgorithmscoresforeachshRNAtargetsitefromTable2areplottedagainstobservedknockdownefficiencyfortheHsiehetal.(A),5ΔΔG(more...)

BecausetheseresultssuggestthatalinearrelationshipdoesnotstronglyapplytoshRNAknockdownforanyofthesixalgorithms,weevaluatedeachofthealgorithmsbyROCcurveanalysistodeterminewhetheranyalgorithmissuperiortotheothersatidentifyingeffectiveshRNAs.TheROCcurveisaplotofsensitivity(thetruepositivefraction,TPF)versus1minusthespecificity(thefalsepositivefraction,FPF)thatisgeneratedbyvaryingthedecisionthresholdbetweentheminimumandmaximumalgorithmscore.ThediagonaloftheROCplotrepresentstheROCcurveforanalgorithmthatisnobetteratdiscriminationthanrandomselection.AlgorithmsthatarepoordiscriminatorshaveROCcurvesthattrackalongthediagonalandhaveanareaundertheROCcurve(AUC)thatisnotsignificantlydifferentfromtheAUCofthediagonal(0.5).AlgorithmsthataregooddiscriminatorshaveROCcurveswithstrongconvexdeviationfromthediagonalandAUCsthatapproach1andaresignificantlydifferentfromtheAUCofthediagonal.

TheHsiehetal.algorithmhadaconcaveROCcurve(Fig.(Fig.5A)5A)indicatingunacceptablesensitivityandspecificyindiscriminatingeffectivefromineffectiveshRNAs.TheROCcurvesforallotheralgorithms(Figs.5B–F)trackednearthediagonaloftheROCplotandhadAUCsthatwerenotsignificantlydifferentfromtheAUCofthediagonal(Figs5B–F).Thus,noneofthealgorithmsshowedastatisticallysignificantabilitytodiscriminatebetweeneffectiveandineffectiveshRNAs.

Figure5

ROCcurveanalysisofsiRNAscoringalgorithms.Thetruepositivefractionwasplottedagainstthefalsepositivefractionasthedecisionthresholdvariedfromminimumtomaximumscores(seeMaterialsandMethodsfordetails)fortheHsiehetal.(A),(more...)

TheTakasakietal.algorithm(Fig.(Fig.5F)5F)showedthemostpromiseasadiscriminatorofeffectivefromineffectiveshRNAs.However,thisalgorithmsufferedfromarelativelyhighfalsepositivefractionfordecisionthresholdsnearthemaximumscoreasindicatedbytheweak,erraticdeviationfromthediagonalneartheoriginoftheROCcurve(Fig.(Fig.5F).5F).ThisindicatedthatthealgorithmassignedahighscoretoanumberofineffectiveshRNAs.Inspectionofthedatarevealedthattwoofthethreehigh-scoringineffectiveshRNAstargetedgeneswhoseexpressionwassuccessfullyknocked-downbyothershRNAs(Table3,asterisks).ThusitisunlikelythattheinefficacyoftheshRNAsisaconsequenceofselectivepressureagainstthestablesuppressionofgeneexpression.ItismorelikelythattheTakasakietal.algorithmdoesnotaccountforacriticalfeatureofeffectiveshRNAs.

,Applicat,ionofanalgorithmmodificationbasedonthestabilityofthe6centralbasesofeachshRNA

InspectionofthephysicalpropertiesofthehighscoringineffectiveshRNAsrevealedthattheaveragestabilityoftheduplexformedbythe6centralbasesoftheshRNAs(bases6–11ofthesensestrandhybridizedtobases9–14oftheantisensestrand)wasgreaterthantheaveragestabilityofhighscoringeffectiveshRNAs(ΔG=-13.1±0.1versus-11.1±1kcal/molrespectively).Basedonthisobservation,theTakasakietal.algorithmwasmodifiedsuchthatshRNAswithacentralduplexΔGequaltoorlessthan-12.9kcal/molwereassignedaminimumscore(Table4).ThismodificationassignedminimumscorestofiveshRNAs,fourwhichwereineffective,thusincreasingthespecificityofthealgorithmwithoutasignificantlossinsensitivity.AminimumscoreassignedtooneeffectiveshRNA(71%knockdown),indicatesthatotherpropertiesinadditiontocentralduplexstabilityinfluenceefficacy.Nevertheless,theadditionofthismodificationeliminatedtheweakerraticdeviationoftheROCcurvefromthediagonalforhighdecisionthresholdsandincreasedtheAUCto0.79(Fig.(Fig.5I).5I).SimilarmodificationoftheAmarzguiouietal.andUi-Teietal.algorithmsalsoraisedtheAUCsoftheirROCcurves(Figs.(Figs.5G5Gandand5H).5H).Withthismodification,theAUCsoftheROCcurvesforallthreemodifiedalgorithmsweresignificantlydifferentfromtheAUCofthediagonal(Figs.5G–I),indicatingstatisticallysignificantpredictivecapability.DifferencesbetweenAUCsoftheROCcurvesforthemodifiedalgorithmswerenotsignificant,soonstatisticalgroundsallthreeofthemodifiedalgorithmswereofequalutility.The5ΔΔG,Reynoldsetal,andtheHsiehetal.algorithmswerenotimprovedtoastatisticallysignificantpredictivecapabilitybyapplyingthecentralduplexΔGmodification(datanotshown).

Table4

ModificationofalgorithmscoresbaseduponshRNAcentralduplexΔG.ThepercentknockdowndatarepresentstheaverageknockdownasshowninTable3.shRNAswithacentralΔGequaltoorlessthan-12.9kcal/molareunderlined.Thesewere(more...)

ToaddressthepossibilitythattheimprovementachievedbythemodificationoftheAmarzguiouietal,Ui-Teietal,andTakasakietal.algorithmsisaconsequenceofoverfittingoursetofshRNAs,anindependentsetof38shRNAspooledfrompreviouspublications([18,27-33];Table5)weresubjectedtoanalysis.WhilenoneoftheROCcurvesforthethreeunmodifiedalgorithmshadanAUCsignificantlydifferentfromthatofthediagonal(Amarzguiouietal.,p=0.174;Ui-Teietal.p=0.09;Takasakietal.,p=0.26),allofthemodifiedalgorithmsyieldedROCcurveswithAUCssignificantlydifferentfromtheAUCofthediagonal(p=0.0001–0.009;Figs.5J–L).Onstatisticalgrounds,allthreeofthemodifiedalgorithmswereofequalutilityastheAUCsoftheROCcurvesforthemodifiedalgorithmswereallsignificantlydifferentfromtheAUCofthediagonal,butnotsignificantlydifferentfromeachother.ThisanalysisofanindependentsetofshRNAssuggeststhatthemodificationofthealgorithmsisofgeneralvalidity.

Table5

PreviouslypublishedshRNAsequencesanalyzedinthisstudy

BecauseminimizingthefalsepositiverateistheprimaryconcerninshRNAdesign,werecommendusingthemodifiedUi-Teietal.algorithm,whichhadthelowesthighfalsepositivefractionatdecisionthresholdsnearthemaximumscoreasindicatedbythestrongdeviationfromthediagonalneartheoriginoftheROCcurve(Figs.(Figs.5H5Handand5K).5K).Usingadecisionthresholdof3limitsselectionofshRNAstoaregionoftheROCcurvewherethesensitivitywasacceptable(0.28–.33),whilethespecificitywasverygood(1.0).Bysettingthisdecisionthreshold,thefalsepositivefractionwasminimized,while28–33%oftheeffectiveshRNAswereidentifiedfromourshRNAsandthepublishedsetofshRNAsrespectively.Shouldthesensitivityneedtobeincreased,werecommendusingadecisionthresholdof2.Thisthresholdhadasensitivityof0.54–0.55andaspecificityof0.88–0.9.Ifthedecisionthresholdwasfurtherrelaxedto0,thesensitivityincreasedto0.86–0.9,butthespecificityfellto0.55–0.54.Werecommendusingthehighestofthesedecisionthresholdspossible.

Thoughstatisticallysmall,thisstudyhastheadvantagetoourknowledgeofbeingthelargestpublishedsetof19-merbasedshRNAstodate.Inaddition,unlikeothershRNAstudiesthatarenecessarilyskewedtowardeffectiveshRNAs,ourstudyincludesbothfunctionalandnon-functionalshRNAs.WehaveshownthatmodifiedUi-Teietal.,Amarzguiouietal.andTakasakietal.algorithmsarefairtogoodpredictivetoolsthatdistinguisheffectivefromineffectiveshRNAs.However,significantshortcomingsstillexistinthemodifiedalgorithms.AdirectassessmentofthealgorithmmodificationsusingshRNAsdesignedaccordingtoeachoriginalandmodifiedalgorithmwouldlendsupporttothesefindings.ThesealgorithmsaremeanttoreducethenumberoffalsepositiveshRNAsselected,notcompletelyeliminatethemaltogether,andthusthiswouldrequirealargenumberofshRNAstoobtainastatisticallysignificantdifferenceinfalsepositiverate.TheavailabilityoflargershRNAdatasetsshouldsupportthedevelopmentofalgorithmswithimprovedsensitivityandspecificity.Additionally,severalsoftwareapplicationsforsiRNAoligonucleotidedesignthatwerenotconsideredinthisstudymaybeofuseinthedesignofshRNAs[16,34-36].CriteriafordesigningfunctionalsiRNAoligonucleotidesremaincontroversialasevidencedbythelargenumberofstudiesstillbeingdevisedforsiRNAdesign,andsincewedidnottestthesesequencesassiRNAsitcannotbeestablishedwhetherthemodificationofthesealgorithmsalsoappliesinthecontextofsiRNAoligonucleotides.shRNAhasanaddedlayerofcomplexityoversiRNAoligonucleotidessincethehairpinneedstobeprocessedwithinthecellbeforeenteringtheRISCcomplex.Moreover,selectivepressureagainstthestableexpressionofshRNAsthataredeleterioustocellgrowthwouldbeexpectedtolendanadditionalconstrainttothestableexpressionofcertainshRNAs.Despitethesecomplexities,ourfindingsbegintobringinsightintotheabilitytoapplysiRNAalgorithmsfordesignoffunctionalshRNAs.

Conclusion

WehaveprovidedseveralimportantstrategiesthatshouldfacilitatethegenerationofeffectiveshRNAvectorsforgeneknockdowninmammaliancells.Theabilitytoproducewild-typeandmutantshRNAvectorssimultaneouslyusingmixedoligonucleotidepairsprovidesanefficientmethodtogenerateaspecificcontrolvectorwithlittleaddedtimeorcost.Thisstrategyshouldbeparticularlyusefulingeneratingspecificcontrolsinhighthroughputapplications.DifficultyinsequencingthroughthehighintrinsicsecondarystructureofsomehairpinvectorsalsohaspresentedamajorconstraintintheconstructionofshRNAvectors,andtheknowledgethatsequencingissuescanberesolvedbymodifyingBigDyechemistriesandaddingBetaineandotherDNArelaxingagentsshouldbevaluableregardlessofthemethodofshRNAdesignandconstruction.Usingdatafrom27shRNAsthatwehaveconstructedwehaveperformedananalysisoftheabilityofpublishedalgorithmsforsiRNAoligonucleotidetargetselectiontopredictknockdownefficacy.OurresultsshowthatshRNAefficacycannotstrictlybeexplainedbyanyofthesixalgorithmstested.Weprovideamodification,however,thatgreatlyimprovesthepredictabilityoftheUi-Teietal.,Amarzguiouietal.andTakasakietalalgorithms.Resultswereconfirmedusingdatafrom38previouslypublishedshRNAs.ThesefindingsshouldbeofsignificantapplicabilityinthedesignandpreparationoffunctionalshRNAs.

Methods

Celllinesandcellculture

THP1monocyticcellandJurkatTcelllineswereculturedinRPMI,10%FCS.Culturesweremaintainedbetween2and8×10⁵cells/mlandstandardizedtoequivalentdensitiesbeforeassessingknockdownefficiencies.

Plasmiddesignandconstruction

RetroviralvectorsforshRNAexpressionhaveapHSPGbackbone[37]withaninsertedH1RNApromoterdrivingshRNAexpression.ThepHSPGvectoralsohasagreenfluorescentprotein(GFP)genedrivenbyaphosphoglyceratekinasepromoterasamarker.TheH1promoterandshRNAexpressioncassettewereinsertedintothepHSPGvectorbyoneoftwomethods.Inthefirstmethod,adoublestrandedoligomerissynthesizedwithBglIIandXhoIhalfsitesontheends.Thisispreparedaseitheramatchedpairorawild-type/mutanthybrid(Fig.(Fig.1).1).Topreparewild-typeandmutantshRNAvectorssimultaneously,aforwardstrandoligomerissynthesizedthatcontainsthewild-typehairpin.Inparallel,amutantreversestrandwithaonebpmismatchwithinthetargetsequenceisalsosynthesized.Despitethemismatchesbetweentheforwardwild-typeandreversemutantstrands,annealingcanstilloccurefficientlyunderoptimizedconditions.Thedsoligonucleoltideisannealedbycombining1000pmolofeacholigomerstrandin50μlofannealingbuffer(100mMpotassiumacetate,30mMHEPES-KOH,pH7.4,2mMMg-acetate).Themixtureisboiledforfiveminutesandthencooledslowlyto4°C.TheannealeddoublestrandedoligomerisligatedintoBglIIandXhoIhalfsites3oftheH1promoterthatisinsertedintothe3longterminalrepeat(LTR)ofpHSPGgeneratingaself-inactivatingLTR.Thedoublestrandedhybridisligatedintothevector5ofapolIIIpromoterandistransformedintocompetentbacteria.Sincereplicationissemi-conservative,thedaughterbacteriawillbeoftwodifferentpopulationsthatcarryeitheradouble-strandedwild-typeoradouble-strandedmutantvector.Bacteriacarryingeitherwild-typeormutantvectorscanthenbeisolatedfromindividualcoloniesandsequenced.Oligosusedforthismethodhadthesequence:GATCCCC-N19-TTCAAGAGA-rN19-TTTTTGGAAA;andTCGATTTCCAAAAA-N19-TCTCTTGAA-rN19-GGG(whereN19isthesenseofthetargetsequenceandrN19istheantisense).WehaveroutinelyusedDH5αtopreparewild-typeandmutantshRNAvectorswithapproximatelyequalyieldsofeachtypeofvector;however,arepair-deficientE.colimutantcouldtheoreticallyimprovetheefficiencyofsimultaneousconstruction.

AseconddesigninvolvesPCRusingaprimercomplementarytothe5endoftheH1promotertogetherwithanshRNA-specificlong-primerwhose3endiscomplementarytothe3endoftheH1promoter.PCRisperformedusingPfxpolymerasewithPCRxenhancer(thiscombinationhasprovedessentialforreducingthenumberofmutationsintroducedwithintheamplifiedregion).Oligosusedforthismethodwere:GCGGCCGCGATATCGAACGCTGACGTCATCAACCC(universaloligo);andTGCTCTAGAAAAA-N19-TCTCTTGAA-rN19-GGGAAAGAGTGGTCTCATACAGAACTTATAAGATTCC,whereN19isthesenseofthetargetsequenceandrN19istheantisense.SequencescomplimentarytotheH1promoterareunderlined.PCRfragmentsweredigestedwithEcoRVandXbaIandligatedintothe3LTRofpHSPG.Allconstructswereverifiedbysequencing.

SequencingofshRNAvectors

DNAsequencingwasdoneattheUNC-CHGenomeAnalysisFacility.Sequencingreactionswere12.5uLtotalvolumecontaining1×BigDyeTerminatorv1.1CycleSequencingReadyReactionMix(AppliedBiosystems),0.26ugofDNAand3.75pmoleofprimer.LTRaprimer(sequenceCGCGAACAGAAGCGAGAA)thatbindstheHSPGvectorapproximately120bpdownstreamfromtheinsertedhairpinwasusedinallsequencingreactions.TheshRNAvectorsusedtoassesssequencingefficacywereconstructedasstemloophairpinsasdescribedaboveandcontainthefollowingtargetsequences:pHSPG-shTLR4,AGGTGATTGTTGTGGTGTC;pHSPG-shmutTLR4,AGGTGATTCTTGTGGTGTC;pHSPG-shmCNN3,AGGAATGAGCGTGTATGGG;andpHSPG-shTLR2,GTATGAACTGGACTTCTCC.ModifiedsequencingreactionssubstitutedpartoralloftheBigDyev1.1chemistrywithABIPrismdGTPBigDyeTerminatorReadyReactionMix(AppliedBiosystems).Ratiosof20:1,10:1,5:1and3:1BD:dGTPchemistriesandstraightdGTPchemistrywereused.Additivesevaluatedinsequencingreactionswere:0.83MBetaine(Sigmapart#B-0300),5%DMSO(Sigmapart#D-2650),1×PCRxEnhancer(inInvitrogenkitpart#11495-017),1×(1uLThermofidelase/20uLsequencingreaction)ThermoFidelaseI(FidelitySystems)and10×primerconcentration.Thethermalcyclerprotocolusedforcyclesequencingwas:95Cfor3minutes(or5minuteswhenusingThermoFidelaseI)followedby25cyclesof98Cfor40seconds(1^stcycle)or10seconds(subsequentcycles),50Cfor5secondsand60Cfor4minutes.SequencingreactionswerepurifiedusingCentri-Sep96wellspinplates(PrincetonSeparations),andthepurifiedreactionproductswererunona3730DNAAnalyzer(AppliedBiosystems)witha50cmarrayusingtheLongReadprotocol.Asameasureofreadthroughefficacypeakheightratiosweredeterminedabout300basesafterand50basesbeforethehairpin.

Viruspreparation,transductionandcellsorting

Topreparevirus,pHSPG-shRNAplasmidswereco-transfectedinto293Tcellswithgag/polandVSVgvectorsbythecalciumphosphatemethod.Viralsupernatantswerecollected24and48hoursfollowingtransfectionandusedtotransduceTHP1orJurkatcellsbyspinoculation.THP1cellsweretransducedwithvirusontwoconsecutivedaystoincreasetransductionlevels.Followingapproximatelyoneweekofculture,stablytransducedcellswereisolatedbysortingforGFP.FACSanalysisstudiessuggestthatGFPexpressionis95%stableforatleasttwomonthsfollowingsorting(notshown).

RNAexpressionanalyses

TotalRNAwasisolatedwithanRNeasyisolationkit(Qiagen)usingtherecommendedprotocol.Toincreasespecificity,cDNAwasreversetranscribedusingoligodTprimerandSuperscriptIIIRT(GibcoBRL).Real-timePCRexperimentswereperformedusinganABPrism7700instrument(AppliedBiosystems)with57°Cannealingtemperature.For18s,CLR19.6/NALP11,CLR19.3/NALP12,MYD88,TLR2,TLR4,andTRAF6,real-timePCRwasperformedusingAbsoluteQPCRMix(ABgene)mixandeitherTETorFAMlabeledprobes.Thefollowingarethesequencesoftheoligonucleotidesused,listedas[forward;reverse;probe]:18s-[CGGCTACCACATCCAAGG;GCTGCTGGCACCAGACTT;Tet-CAAATTACCCACTCCCGACCCG-Tamra];CLR19.6/NALP11-[TCAATGATGCGTAAGGAAAGA;ACTTTCCCATTGCAGCATGA;Fam-CTTTGCATGCCTCCTGATTGCGGT-Tamra];CLR19.3/NALP12-[AGAGGACCTGGTGAGGGATAC;CTTCCAGAAGGCATGTTGAC;Fam-CCCGTCCTCACTTGGGAACCA-Tamra];MYD88-[CTCTGTAGGCCGACTGC;CTGCTGCTGCTTCAAGATA;Fam-TGGCAATCCTCCTCAATGCTGGGTC-Tamra];TLR2-[GGTCATCATCAGCCTCTCCA;GAGCTGCCCTTGCAGATAC;Fam-CCTCCAATCAGGCTTCTCTGTCTTGTGACC-Tamra];TLR4:[AGAGCCTAAGCCACCTCT;CTAGAGATGCTAGATTTGTCTCCA;Fam-AGCCACCAGCTTCTGTAAACTTGATAGTCCAGA-Tamra];TRAF6:[CCATGCGGCCATAGGTT;TTTCCAGCAGTATTTCATTGTCA;Fam-TGGACATTTGTGACCTGCATCCCTTATTGAT-Tamra].ForASC/PYCARD,CLR16.2,MAL/TIRAP,TRAM/TICAM2,andTRIF/ICAM1,realtimePCRwasperformedusingABsoluteSYBRgreenmix(ABgene)andthefollowingprimers,listedas[forward;reverse]:ASC/PYCARD1-[AACCCAAGCAAGATGCGGAAG;TTAGGGCCTGGAGGAGCAAG];CLR16.2-[TCAACACAGCCCTCACTGCTCTCTATCTC;AGCCACCCCAATGGCATTTCCTCTTAAGTC];MAL/TIRAP-[GGACTCATCTCCTGCCTAAC;CATGGTGAGGCCTGCAATCT];TRAM/TICAM2-[GGCACAGTGTGGATACAAGT;ACATCTCTTCCACGCTCTGA];TRIF/TICAM1-[CAGGAGCCTGAGGAGATGAG;GGGTAGTTGGTGCTGGTTTC].Primersweredesignedtospanexon/intronjunctionswherepossible.AllRNAexpressionanalysesweredoneatleastintriplicateforRNAisolatedondifferentdaysandknockdownswereverifiedwithatleastonecontrolhairpin.ValuesrepresentaverageobservedknockdownforRNAfromdifferentdaysofcellcultureandwerestandardizedto18srRNAexpression.

Implimentationofalgorithms

Thefreeenergy(ΔG)ofRNAduplexformationforthe5basesatthe5endofthesenseandantisensestrandswasdeterminedusingthethermodynamicparametersandexpandednearest-neighbormodelofXiaetal.[38].The5ΔΔG(differentialfreeenergy)wascalculatedbysubtractingtheΔGoftheantisensestrandfromthatofthesensestrand.DeterminationofscoresfortheReynoldsetal.,Amarzgiuouietal.,andTakasakietal.algorithmswasasdescribed[15,17,20].TheHsiehetal.scorerepresentstheinterpretationoftheHsiehetal.designcriteriaaspublishedbySaetromandSnove[16,19].FortheUi-TeialgorithmsequenceswithaCorGonthe5endscored1point,whereasthosewithanAorTscored-1point.SequenceswithanAorTonthe3endscored1point,whereasthosewithaCorGscored-1point.Sequenceswith5ormoreAorTbasesintheseven3basesscored2points,whereasthosewith4AorTbasesscored1point.Sequencescanbeclassifiedbyscoreasfollows:4–classIa,3–classIb,2,1or0–classIIand-1or-2–classIII.Allknockdownsof<10%aregraphedas0.

ModificationsoftheAmarzgiuouietal.,Ui-Teietal.,andTakasakietal.algorithmswereappliedasfollows.ThefreeenergyofRNAduplexformationfor6centralbasesofeachshRNA(bases6–11ofthesensestrandhybridizedtobases9–14oftheantisensestrand)wascalculated.shRNAswithcentralduplexΔGsequaltoorlessthan-12.9kcal/molwereassignedaminimumscore(-4fortheAmarzgiuouietal.algorithm,-2fortheUi-Teietal.algorithmand-13.26fortheTakasakietal.algorithm).ThescoresforshRNAswithcentralduplexΔGsgreaterthan-12.9kcal/molwereleftunchanged.Thecutoffvalueof-12.9kcal/molwasselectedempiricallybasedupontherangeofcentralduplexΔGsforallshRNAs(seeTable4).

ROCcurveanalysis

ROCcurveswereconstructedasdescribed[39].ROCanalysisrequiresthateachshRNAisclassifiedaseithereffectiveorineffective.Forouranalyses,ashRNAwasclassifiedaseffectiveifitreducedmRNAexpressionby50%ormore.AROCcurvewasgeneratedforeachalgorithmasfollows.ThedecisionthresholdwassettooneunitbelowthelowestshRNAscore.BydefinitionshRNAswithscoresgreaterthanorequaltothedecisionthresholdwerepredictedtobeeffective,whilethosewithscoreslessthanthedecisionthresholdwerepredictedtobeineffective.TheneachshRNAwasclassifiedasatruepositive(effectivepredictedtobeeffective),afalsenegative(effectivepredictedtobeineffective),atruenegative(ineffectivepredictedtobeineffective)orafalsepositive(ineffectivepredictedtobeeffective).Thetruepositivefraction(TPF)forthedecisionthresholdwascalculatedasthenumberoftruepositivesdividedbythesumofthetruepositivesandfalsenegatives.Thefalsepositivefraction(FPF)wascalculatedasthenumberoffalsepositivesdividedbythesumofthefalsepositivesandtruenegatives.ThedecisionthresholdwasincreasedbyoneunitandtheTPFandFPFcalculatedagain.ThisprocesswasrepeateduntilthedecisionthresholdwasoneunitgreaterthanthehighestscoringshRNA.ROCcurveswereconstructedbyplottingTPFversustheFPFforalldecisionthresholds.TheareaundertheROCcurvewasestimatedbyintegrationusingthetrapezoidrule.

Listofabbreviations

siRNA,smallinterferingRNA;shRNA,shorthairpinRNA;RNAi,RNAinterference;RISC,RNA-inducedsilencingcomplex;BDchemistry,BigDyeTerminatorv1.1TerminatorCycleSequencingChemistry;dGTPchemistry,ABIPrismdGTPBigDyeTerminatorCycleSequencingChemistry;ROCanalysis,receiveroperatingcharacteristicanalysis,AUC,areaunderthecurve;TPF,truepositivefraction;FPF,falsepositivefraction.

Authorscontributions

DJTdesigned,preparedandtestedshRNAvectors,conceivedmethodsforconstructingandassayingshRNA,performedanalysis,interpretation,andpresentationofdata,anddraftedthemanuscript.LLperformedexperimentstodetermineoptimalsequencingconditions,andassistedintheinterpretationandpresentationofsequencingdata.JZ,BJC,HI,KLW,andJLeachassistedintheconstructionofshRNAvectorsandacquisitionofdataonknockdownefficiencies.JTandWRassistedintheconception,design,developmentandcoordinationofthestudy,andaddedtotheintellectualcontentofthemanuscript.WRalsoperformedstatisticalanalysesofshRNAefficacy.

SupplementaryMaterial

AdditionalFile1

EffectofmixedBD:dGTPchemistriesonpeakresolution.Sequencingfromthe500baseregionofpHSPG-shTLR2-2271,containingahairpinstructurewhichsequencedwithoutprobleminstraightBDchemistry,isshown.SequencingchemistriesusedwereBDchemistry(A),20:1BD:dGTPchemistries(B),10:1BD:dGTPchemistries(C),5:1BD:dGTPchemistries(D)and3:1BD:dGTPchemistries(E).PeakresolutiondecreasedastheamountofdGTPusedincreased(seeboxedAAAAregionatpostion475).

Clickhereforfile^(2.5M,eps)

AdditionalFile2

Geneknockdownissimilarforcelllinesderivedfromdifferentroundsoftransductionandsorting.Realtimedataisshownforcelllinesderivedindependentlyusingthesameviralvectors.ViruswaspreparedandusedtotransduceTHP1cellsindependentlyforeachround.Valuesarepresentedasaverage+SEMforatleastthreeassaysruninduplicate,withtheexceptionofshclr19.3-1504,secondtransduction(singlevalue).

Clickhereforfile^(701K,eps)

Acknowledgements

ThisresearchwassupportedbyNationalInstitutesofHealthgrantsDK38108andAI57175(toJ.T.);andU.S.EnvironmentalProtectionAgencyCooperativeAgreementCR829522.WewouldliketothankDrs.SusanSilva,ChrisMoore,BeckleyDavis,CaseyClements,andHankVanDeventerforhelpfulsuggestions.

References

MatzkeMA,BirchlerJA.RNAi-mediatedpathwaysinthenucleus.NatRevGenet.2005;6:24–35.doi:10.1038/nrg1500.[PubMed][CrossRef]
HuppiK,MartinSE,CaplenNJ.DefiningandassayingRNAiinmammaliancells.MolCell.2005;17:1–10.doi:10.1016/j.molcel.2004.12.017.[PubMed][CrossRef]
PaddisonPJ,CaudyAA,BernsteinE,HannonGJ,ConklinDS.ShorthairpinRNAs(shRNAs)inducesequence-specificsilencinginmammaliancells.GenesDev.2002;16:948–958.doi:10.1101/gad.981002.[PMCfreearticle][PubMed][CrossRef]
SuiG,SoohooC,AffarelB,GayF,ShiY,ForresterWC.ADNAvector-basedRNAitechnologytosuppressgeneexpressioninmammaliancells.ProcNatlAcadSciUSA.2002;99:5515–5520.doi:10.1073/pnas.082117599.[PMCfreearticle][PubMed][CrossRef]
BrummelkampTR,BernardsR,AgamiR.AsystemforstableexpressionofshortinterferingRNAsinmammaliancells.Science.2002;296:550–553.doi:10.1126/science.1068999.[PubMed][CrossRef]
PaddisonPJ,CaudyAA,SachidanandamR,HannonGJ.Shorthairpinactivatedgenesilencinginmammaliancells.MethodsMolBiol.2004;265:85–100.[PubMed]
WongAW,BrickeyWJ,TaxmanDJ,vanDeventerHW,ReedW,GaoJX,ZhengP,LiuY,LiP,BlumJS,McKinnonKP,TingJP.CIITA-regulatedplexin-A1affectsT-cell-dendriticcellinteractions.NatImmunol.2003;4:891–898.doi:10.1038/ni960.[PubMed][CrossRef]
,TomarRS,MattaH,ChaudharyPM.Useofadeno-associatedviralvectorfordeliveryofsmallinterferingRNA.Oncogene.2003;22:5712–5715.doi:10.1038/sj.onc.1206733.[PubMed][CrossRef]
RubinsonDA,DillonCP,KwiatkowskiAV,SieversC,YangL,KopinjaJ,RooneyDL,IhrigMM,McManusMT,GertlerFB,ScottML,VanParijsL.Alentivirus-basedsystemtofunctionallysilencegenesinprimarymammaliancells,stemcellsandtransgenicmicebyRNAinterference.NatGenet.2003;33:401–406.doi:10.1038/ng1117.[PubMed][CrossRef]
MooreMD,McGarveyMJ,RussellRA,CullenBR,McClureMO.StableinhibitionofhepatitisBvirusproteinsbysmallinterferingRNAexpressedfromviralvectors.JGeneMed.2005[PubMed]
TranN,CairnsMJ,DawesIW,ArndtGM.ExpressingfunctionalsiRNAsinmammaliancellsusingconvergenttranscription.BMCBiotechnol.2003;3:21.doi:10.1186/1472-6750-3-21.[PMCfreearticle][PubMed][CrossRef]
WadhwaR,KaulSC,MiyagishiM,TairaK.VectorsforRNAinterference.CurrOpinMolTher.2004;6:367–372.[PubMed]
KhvorovaA,ReynoldsA,JayasenaSD.FunctionalsiRNAsandmiRNAsexhibitstrandbias.Cell.2003;115:209–216.doi:10.1016/S0092-8674(03)00801-8.[PubMed][CrossRef]
SchwarzDS,HutvagnerG,DuT,XuZ,AroninN,ZamorePD.AsymmetryintheassemblyoftheRNAienzymecomplex.Cell.2003;115:199–208.doi:10.1016/S0092-8674(03)00759-1.[PubMed][CrossRef]
ReynoldsA,LeakeD,BoeseQ,ScaringeS,MarshallWS,KhvorovaA.RationalsiRNAdesignforRNAinterference.NatBiotechnol.2004;22:326–330.doi:10.1038/nbt936.[PubMed][CrossRef]
SaetromP,SnoveOJ.AcomparisonofsiRNAefficacypredictors.BiochemBiophysResCommun.2004;321:247–253.doi:10.1016/j.bbrc.2004.06.116.[PubMed][CrossRef]
TakasakiS,KotaniS,KonagayaA.AneffectivemethodforselectingsiRNAtargetsequencesinmammaliancells.CellCycle.2004;3:790–795.[PubMed]
Ui-TeiK,NaitoY,TakahashiF,HaraguchiT,Ohki-HamazakiH,JuniA,UedaR,SaigoK.GuidelinesfortheselectionofhighlyeffectivesiRNAsequencesformammalianandchickRNAinterference.NucleicAcidsRes.2004;32:936–948.doi:10.1093/nar/gkh247.[PMCfreearticle][PubMed][CrossRef]
HsiehAC,BoR,ManolaJ,VazquezF,BareO,KhvorovaA,ScaringeS,SellersWR.AlibraryofsiRNAduplexestargetingthephosphoinositide3-kinasepathway:determinantsofgenesilencingforuseincell-basedscreens.NucleicAcidsRes.2004;32:893–901.doi:10.1093/nar/gkh238.[PMCfreearticle][PubMed][CrossRef]
AmarzguiouiM,PrydzH.AnalgorithmforselectionoffunctionalsiRNAsequences.BiochemBiophysResCommun.2004;316:1050–1058.doi:10.1016/j.bbrc.2004.02.157.[PubMed][CrossRef]
JacksonAL,LinsleyPS.Noiseamidstthesilence:off-targeteffectsofsiRNAs?TrendsGenet.2004;20:521–524.doi:10.1016/j.tig.2004.08.006.[PubMed][CrossRef]
PaddisonPJ,ClearyM,SilvaJM,ChangK,ShethN,SachidanandamR,HannonGJ.CloningofshorthairpinRNAsforgeneknockdowninmammaliancells.NatMethods.2004;1:163–167.doi:10.1038/nmeth1104-163.[PubMed][CrossRef]
MillerVM,XiaH,MarrsGL,GouvionCM,LeeG,DavidsonBL,PaulsonHL.Allele-specificsilencingofdominantdiseasegenes.ProcNatlAcadSciUSA.2003;100:7195–7200.doi:10.1073/pnas.1231012100.[PMCfreearticle][PubMed][CrossRef]
DucatDC,HerreraFJ,TriezenbergSJ.OvercomingobstaclesinDNAsequencingofexpressionplasmidsforshortinterferingRNAs.Biotechniques.2003;34:1140–2,1144.[PMCfreearticle][PubMed]
MalykhA,MalykhO,PolushinN,KozyavkinS,SlesarevA.Finishing"workingdraft"BACprojectsbydirectedsequencingwithThermoFidelaseandFimers.MethodsMolBiol.2004;255:295–308.[PubMed]
SlesarevAI,MezhevayaKV,MakarovaKS,PolushinNN,ShcherbininaOV,ShakhovaVV,BelovaGI,AravindL,NataleDA,RogozinIB,TatusovRL,WolfYI,StetterKO,MalykhAG,KooninEV,KozyavkinSA.ThecompletegenomeofhyperthermophileMethanopyruskandleriAV19andmonophylyofarchaealmethanogens.ProcNatlAcadSciUSA.2002;99:4644–4649.doi:10.1073/pnas.032671499.[PMCfreearticle][PubMed][CrossRef]
TrushinSA,PenningtonKN,CarmonaEM,AsinS,SavoyDN,BilladeauDD,PayaCV.ProteinkinaseCalpha(PKCalpha)actsupstreamofPKCthetatoactivateIkappaBkinaseandNF-kappaBinTlymphocytes.MolCellBiol.2003;23:7068–7081.doi:10.1128/MCB.23.19.7068-7081.2003.[PMCfreearticle][PubMed][CrossRef]
ZhangXN,XiongW,WangJD,HuYW,XiangL,YuanZH.siRNA-mediatedinhibitionofHBVreplicationandexpression.WorldJGastroenterol.2004;10:2967–2971.[PubMed]
XiaoC,ShimJH,KluppelM,ZhangSS,DongC,FlavellRA,FuXY,WranaJL,HoganBL,GhoshS.EcsitisrequiredforBmpsignalingandmesodermformationduringmouseembryogenesis.GenesDev.2003;17:2933–2949.doi:10.1101/gad.1145603.[PMCfreearticle][PubMed][CrossRef]
ZhangL,YangN,Mohamed-HadleyA,RubinSC,CoukosG.Vector-basedRNAi,anoveltoolforisoform-specificknock-downofVEGFandanti-angiogenesisgenetherapyofcancer.BiochemBiophysResCommun.2003;303:1169–1178.doi:10.1016/S0006-291X(03)00495-9.[PubMed][CrossRef]
LiuXD,MaSM,LiuY,LiuSZ,SehonA.ShorthairpinRNAandretroviralvector-mediatedsilencingofp53inmammaliancells.BiochemBiophysResCommun.2004;324:1173–1178.doi:10.1016/j.bbrc.2004.09.190.[PubMed][CrossRef]
InmanCK,ShoreP.TheosteoblasttranscriptionfactorRunx2isexpressedinmammaryepithelialcellsandmediatesosteopontinexpression.JBiolChem.2003;278:48684–48689.doi:10.1074/jbc.M308001200.[PubMed][CrossRef]
KronkeJ,KittlerR,BuchholzF,WindischMP,PietschmannT,BartenschlagerR,FreseM.AlternativeapproachesforefficientinhibitionofhepatitisCvirusRNAreplicationbysmallinterferingRNAs.JVirol.2004;78:3436–3446.doi:10.1128/JVI.78.7.3436-3446.2004.[PMCfreearticle][PubMed][CrossRef]
ChalkAM,WahlestedtC,SonnhammerEL.ImprovedandautomatedpredictionofeffectivesiRNA.BiochemBiophysResCommun.2004;319:264–274.doi:10.1016/j.bbrc.2004.04.181.[PubMed][CrossRef]
YiuSM,WongPW,LamTW,MuiYC,KungHF,LinM,CheungYT.FilteringofineffectivesiRNAsandimprovedsiRNAdesigntool.Bioinformatics.2005;21:144–151.doi:10.1093/bioinformatics/bth498.[PubMed][CrossRef]
HueskenD,LangeJ,MickaninC,WeilerJ,AsselbergsF,WarnerJ,MeloonB,EngelS,RosenbergA,CohenD,LabowM,ReinhardtM,NattF,HallJ.Designofagenome-widesiRNAlibraryusinganartificialneuralnetwork.NatBiotechnol.2005;23:995–1001.doi:10.1038/nbt1118.[PubMed][CrossRef]
MeissnerEG,CoffieldVM,SuL.ThymicpathogenicityofanHIV-1envelopeisassociatedwithincreasedCXCR4bindingefficiencyandV5-gp41-dependentactivity,butnotV1/V2-associatedCD4bindingefficiencyandviralentry.Virology.2005;336:184–197.doi:10.1016/j.virol.2005.03.032.[PubMed][CrossRef]
XiaT,SantaLuciaJJ,BurkardME,KierzekR,SchroederSJ,JiaoX,CoxC,TurnerDH.Thermodynamicparametersforanexpandednearest-neighbormodelforformationofRNAduplexeswithWatson-Crickbasepairs.Biochemistry.1998;37:14719–14735.doi:10.1021/bi9809425.[PubMed][CrossRef]
RiffenburghRH.StatisticsinMedicine.SanDiego,CA,AcademicPress;1999.pp.248–251.

新闻动态

行业前沿

技术文章

对照品标准品，生物试剂盒，gibco胎牛血清，试剂采购网，中国试剂网，细胞因子试剂盒

Criteria for effective design, construction, and gene knockdown by shRNA vectors