RESEARCHOpenAccess - PDF document

RESEARCHOpenAccess NormalizingtoGADPHjeopardisescorrect quantificationofgeneexpressioninovarian tumours – IPO8andRPL4arereliablereference genes ZuzanaKolkova 1* ,ArsenArakelyan 2 ,BertilCasslén 1 ,StefanHansson 1 andEvaKriegova 3 Abstract Background: Toensureacorrectinterpretationofresultsobtainedwithquantitativereal-timereversetranscription- polymerasechainreaction(RT-qPCR),itiscriticaltonormalizetoareferencegenewithstablemRNAexpressionin thetissueofinterest.GADPHiswidelyusedasareferencegeneinovariantumourstudies,althoughlacking tissue-specificstability.TheaimofthisstudywastoidentifyalternativesuitablereferencegenesforRT-qPCRstudies onbenign,borderline,andmalignantovariantumours. Methods: WeassayedmRNAlevelsfor13potentialreferencegenes – ABL1,ACTB,CDKN1A,GADPH,GUSB,HPRT1, HSP90AB,IPO8,PPIA,RPL30,RPL4,RPLPO,andTBP – withRT-qPCRin42primaryovariantumours,using commerciallypre-designedRT-qPCRprobes.Expressionstabilitywassubsequentlyanalysedwithfourdifferent statisticalprograms(GeNorm,NormFinder,BestKeeper,andtheEquivalencetest). ExpressionofIPO8,RPL4,TBP,RPLPO,andACTBhadtheleastvariationinexpressionacrossthetumour samplesaccordingtoGeNorm,NormFinder,andBestKeeper.TheEquivalencetestfoundvariationinexpression withina3-foldexpressionchangebetweentumourgroupsfor:IPO8,RPL40,RPL30,GUSB,TBP,RPLPO,ACTB,ABL1, andCDKN1A.However,onlyIPO8satisfiedata2-foldchangeasacut-off.Overall,IPO8andRPL4hadthehighest, whereasGADPHandHPRT1thelowestexpressionstability.Employmentofsuitablereferencegenes(IPO8,RPL4) incomparisonwithunsuitableones(GADPH,HPRT1),demonstrateddivergentinfluenceonthemRNAexpression patternofourtargetgenes  GPERanduPAR. Conclusions: WefoundIPO8andRPL4tobesuitablereferencegenesfornormalizationoftargetgeneexpression inbenign,borderline,andmalignantovariantumours.Moreover,IPO8canberecommendedasasinglereference gene.NeitherGADPHnorHPRT1shouldbeusedasreferencegenesinstudiesonovariantumourtissue. Background Mostcasesofovariancancerarediagnosedatanad- vancedstage,withpoorprognosisforthepatients.Early stagesofovariancancerare,ontheotherhand,more accessibletotreatmentandhavemuchbetterprognosis. Thereisanongoingsearchforbiomarkerswithcapacity todetectinparticularearlystagesofthediseasein screeningprograms,sincethiswouldbethesinglemost importantsteptowardsimprovingtheprognosis.A selectivebiomarkermight,furthermore,behelpfulinthe preoperativeassessmentofovarianlesionsinorderto employoptimalsurgery. Analysisofgeneexpressionbyquantitativereal-time reversetranscription-polymerasechainreaction(RT- isafrequentapproachforthebiomarkerdiscoveryin tumourtissue.However,inordertoobtainreliablere- sultsbyRT-qPCRinheterogeneousclinicalsamples,the expressionofatargetgeneneedstobenormalizedtoa stablyexpressedreferencegene(RG)tominimizethe influenceofvariationsin,e.g.extractionyield,reverse- transcriptionyield,andamplificationefficiency[1]. *Correspondence: Zuzana.Kolkova@med.lu.se 1 DepartmentofObstetrics&Gynaecology,LundUniversity,SkåneUniversity HospitalLund,Lund,SE22185,Sweden Fulllistofauthorinformationisavailableattheendofthearticle ©2013Kolkovaetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycited. Kolkova etal.JournalofOvarianResearch 2013, 6 :60 http://www.ovarianresearch.com/content/6/1/60 Stabilityofsuchreferencegeneshastobevalidatedinbenignandmalignanttissuesfromthespecificorganstudied.Useofanunstablereferencegenewillinevi-tablyproduceerroneousresults.Needlesstosay,thisrequirementappliesalsoforovariantumourswithdif-ferentdifferentiationgradesandhistologicaltypes.Thetraditionallyusedhouse-keepinggene,glyce-raldehyde-3-phosphatedehydrogenase(GADPH),wasreportedtodisplaymanydiverseactivitiesunrelatedtoitsglycolyticfunction(e.g.apoptosisandDNAreplica-tion)[2],andtobeup-regulatedinprostatecanceralreadyinthe1990s[3].ThemostcommonlyusedRT-qPCRreferencegenesusedforovarianstudieshasbeenGADPH(~40%),-actin(ACTB)(~20%),ribosomalRNA(18Sand28SrRNA)(~10%)andhypoxanthinephosphoribosyltransferase1(HPRT1)()[4].MorerecentstudyhasadvisedagainsttheuseofGADPHandACTBasRGs,duetotheirnumerouspseudogenespresentinthehumangenome[5].Uptonow,onlytwostudieshavefocusedonfindingareliableRGinnormalovariantissue,andbenignandmalignantserousovariantumours.Theobtainedresults,however,differ;Lial.recommendedcombinationofGUSB,PPIA,andTBP[4],whereasFual.concludedthatcombinationofRPL4,RPLPO,andHSP90AB1(HSPCB)aremoresuitable[6].Bothstudieswereper-formedonChinesepopulations,didnotincludebor-derlinetumours,andusedSYBRGreenRT-qPCRtechnique.ThepresentstudywasperformedonaScandinavianpopulation,includedborderlinetumours,usedprede-signedcommercialRT-qPCRprobes,andappliedfourdifferentstatisticalsoftwareprograms.Inadditiontotheabovementionedtraditionallyusedandearlierre-commendedRGsforovariantissue,wealsoselectedfourgenesfromacommerciallyprintedarray(ABL1,CDKN1A,IPO8,andRPL30).Thus,altogether13genesweincludedinthestudy.Finally,twotargetgeneswerechosentodemonstratethedivergentresults,whichmaybeobtainedbynormalizingtheirmRNAstosuitablevs.unsuitableRGs:Gprotein-coupledestrogenreceptor(GPER),whichhasnodifferencesinexpressionbetweenbenignandmalignantovariantumoursandurokinaseplasminogenactivatorreceptor(uPAR),whichisup-regulatedinmalignanttumours.OvariantumourtissueTissuesamples(n=42)wereobtainedfromprimaryovariantumoursduringsurgeryattheDepartmentofObstetricsandGynaecology,LundUniversityHospital,during20012007.Noneofthepatientshadreceivedchemotherapypriortotheoperation.Thesampleswerecutin5×5×5mmcubes,quickfrozenondryice,andstoredat80°Cuntilused.Inadditiontotheroutinehisto-pathologicalexamination,eachspecimenwasre-evaluatedbyasecondpathologist.Histologicaldifferen-tiationwasclassifiedasbenign(n=9),borderline(n=11),andmalignant(n=22);thehistologicaltypeswereserous(n=21),mucinous(n=13),andendometrioid(n=8)(Table1).Themeanageofincludedpatientswas59years(range2280)inthebenigngroup,55years86)intheborderlinegroup,and62years(4385)inthemalignantgroup.TheEthicalReviewBoardatLundUniversityHospitalapprovedthestudydesignandin-formedconsentwasobtainedfromeachpatient.ExtractionoftotalRNATotalRNAwasextractedfromabout125mgfrozenova-riantumourtissue.ThetissuewashomogenizedinTrizol50mg/mL(Invitrogen,Carlsbad,CA)usingrotating-knives(Polytron).AllRNAsampleswerecheckedforcon-centrationandpuritybyNanoDropSpectrophotometerND-1000(SaveenWerner,Limhamn,Sweden)having260/280andA260/230~2.RNAqualityandintegritywasverifiedbyAgilent2100BioAnalyzer(AgilentTechnolo-gies,PaloAlto,CA),i.e.allsampleshadRNAIntegrityNumber%-4;d.2;馘砀7.7.cDNAsynthesisGeneAmp®RNAPCRkit(AppliedBiosystems,FosterCity,CA)wasusedforreversetranscriptionoftotalRNA(0.2g)tocDNA.ThefinalconcentrationofcDNAwas1L(+/7%)andA260/280ratio~1.8asassessedbyNanoDrop.ThecDNAsampleswerestored20°Cuntilfurtheruse.QuantitativeRT-qPCRamplificationRT-qPCRwasperformedusingaStepOnePluscycler(AppliedBiosystems)understandardthermalcyclingconditions(activationofcontaminationpreventingen-zymeat50°Cfor2min,enzymeactivationat95°Cfor10min,40cyclesofdenaturationat95°Cfor15s,andannealingat60°Cfor1min).PCRreactionswereruninduplicatesandnegativecontrolswereincludedineachamplificationset.Foreachgeneanalysed,pre-manufacturedreal-timeqPCRassayswereused(Ap- Table1DistributionoftheprimaryovariantumoursaccordingtohistopathologySerousMucinousEndometrioidTotal4596511Grade1628Grade2134Grade355102113842etal.JournalofOvarianResearchPage2of10http://www.ovarianresearch.com/content/6/1/60 pliedBiosystemsorIntegratedDNAtechnologies,Inc.,Coralville,IA,USA)(Table2),withprobesspanningexonjunctionsandnotdetectinggenomicDNA.UsingonemalignanttumoursampleandauniversalhumanreferenceRNA(Stratagene,LaJolla,CA,USA),quanti-ficationexperimentswereperformedusingtwostan-dardcurvesfrom10-foldserialdilutionsofthecDNA0.08ng).IdentificationofnewpotentialreferencegenesInordertoidentifynewcandidatereferencegenesinova-riantumourtissue,weemployedacommercialarray(TaqMan®ExpressEndogenousControlPlate,catno4396840,AppliedBiosystems)consistingof32potentialRGs(18S,GADPH,HPRT1,GUSB,ACTB,B2M,HMBS,IPO8,PGK1,RPLPO,TBP,TFRC,UBC,YWHAZ,PPIA,POLR1A,CASC3,CDKN1A,CDKN1B,GADD45A,PUM1,PSMC4,EIF2B1,PES1,ABL1,ELF1,MT-AT6,MRPL19,POP4,RPL37A,RPL30,RPS17).Weanalysedonebenignandonemalignantsampleofovariantumour,whichwereselectedbasedonthegreatestdifferenceinexpressionoftraditionallyusedRGs(ACTB,GADPH,andHPRT1),asmeasuredbyRT-qPCR.Thedifferencebetweenthethresholdcycles(ofthetwosampleswasthencalculatedforeachofthe32genesinthearray.Fourgeneswiththelowestwereselectedforinclusioninourmainstudy.StatisticalanalysisDescriptivestatistics,F-testforCvarianceequalityandKolmogorov-Smirnovtestfornormalityoflog-trans-formedrelativeexpressionvalueswerecalculatedbysoftwareSPSS19.0(SPSSInc,Chicago,IL).TheEquiva-lencetest[7-9]andstatisticalappletsBestKeeper[10],geNorm[11],andNormFinder[12]wereusedforana-lysisofgenesexpressionstability.GeNormcalculatesagene-stabilitymeasure,M-value,astheaveragepair-wisevariationofaparticulargenetoallothercandidaterefe-rencegenes[11].Ontheotherhand,thestabilityvaluecalculatedwithNormFindercombinesestimatedbothintra-groupandinter-groupvariations[12].GeneswiththelowestM-valueshavethemoststableexpression(leastvariability).RelativeexpressionvaluesfortargetgeneswereanalysedbyKruskal-WallisandMannWhitneytests,andthelog-transformedvaluesbyone-wayANOVA.P0.05wasconsideredsignificant.SelectionofbestRGsfromthecommercialgenearrayInordertoselectoptimalcandidateRGsforthisstudyonovariantumours,betweenonebenignandone Table2Referencegenes,targetgenesandassaysusedGenenamesynonymsFunctionNCBIGenereferenceAssayIDC-abloncogene1,non-receptortyrosinekinaseCelldifferentiation,division,adhesionandstressresponse.NM_005157.3,NM_007313.2Hs00245445_m1Actin,betaCellmotility,structure,integrityNM_001101.3Hs99999903_m1Cyclin-dependentkinaseinhibitor1A(p21,Cip1)RegulationofcellcycleprogressionatG1.NM_004064.3Hs00355782_m1Glyceraldehyde-3-phosphatedehydrogenaseCatalysationofanimportantenergy-yieldingstepincarbohydratemetabolism.NM_002046.3Hs99999905_m1Glucuronidase,betaDegradationofglycosaminoglycansNM_000181.2Hs99999908_m1Hypoxanthinephosphoribosyltransferase1Generationofpurinenucleotidesthroughthepurinesalvagepathway.NM_000194.2Hs99999909_m1Heatshockprotein90Proteinfolding,responsetostress.NM_007355Hs.PT.49a.20846338Importin8Nucleartransport.NM_001190995.1NM_006390.3Hs00183533_m1PeptidylprolylisomeraseA(cyclophilinA)Proteinfolding,ligandforCyclosporinA.NM_021130.3Hs99999904_m1Hs.PT.39a.22214851RibosomalproteinL30Componentof60Ssubunit.Catalysationofproteinsynthesis.NM_000989.2Hs00265497_m1RibosomalproteinL4Componentof60Ssubunit.NM_000968Hs.PT.49a.20266660Ribosomalprotein,large,POComponentof60Ssubunit.NM_053275.3,NM_001002.3Hs99999902_m1TATAboxbindingproteinInitiationoftranscriptionofRNApolymerases.M34960.1M55654.1Hs99999910_m1Gprotein-coupledestrogenRapidestrogensignalling.NM_001505.2Hs00173506_m1UrokinaseplasminogenactivatorCellinvasion,migration,signallingviaERK1/2.NM_001005376.2NM_001005377.2NM_002659.3Hs00182181_m1etal.JournalofOvarianResearchPage3of10http://www.ovarianresearch.com/content/6/1/60 malignantovariantumoursamplewiththegreatestdif-ferenceinexpressionofthetraditionallyusedRGs(ACTB,GADPH,andHPRT1),wasmeasuredbyRT-qPCRandcalculatedforall32genesincludedinthear-rays.Thelowest,i.e.theleastvariation,wasfoundforCDKN1A(0.47),ABL1(0.76),RPL30(0.83),RPS17(1.09),MT-ATP6(1.42),andIPO8(1.71),whereasPOP4(6.11),GADPH(5.04),HPRT1(4.91),POLR2A(4.41),CASC3(3.48)hadthehighest.Themostabundantgeneswere18S(meanC±SD:12.11±1.85)andMT-ATP6(21.64±1.00),thegeneswithlowestex-pressionwereYWHAZ(31.42±2.14)andTBP(31.37±2.06).CDKN1A,ABL1,RPL30andIPO8werechosentobeincludedinourpanelofpotentialreferencegenes.ExpressionofselectedcandidatereferenceandtargetgenesinprimaryovariantumoursWeanalysedaltogether13candidatereferencegenes(ABL1,ACTB,CDKN1A,GADPH,GUSB,HPRT1,HSP90AB,IPO8,PPIA,RPL30,RPL4,RPLPO,andTBP)andtwotargetgenes(GPERanduPAR)byRT-qPCR.ExpressionlevelsandvariabilityofCvaluesareshownfortheRGs(Table3).Ofallgenes,PPIAhadthehighest(meanC±SD:22.12±0.82)andGUSBthelowest(31.20±0.99)levelofmRNA(Figure1).Theamplifica-tionefficienciesoftheTaqMan-basedRT-qPCRassayswereintherange8599%forallRGs,exceptABL1andHPRT1,whichhad82%efficiency.Thelinearregressioncoefficient(r)ofthestandardcurvesforallgenesrangedbetween0.998and1.GeneexpressionstabilitycalculatedbyGeNormExpressionstabilityofthe13candidateRGswasfirstassessedbyGeNorminthewholesetoftumoursam-ples.Theexpressionstabilityvalue(M-value)wascalcu-latedbasedontheaveragepair-wisevariationbetweenallgenestested(Table4).ThegeneswiththelowestM-valuehavethemoststableexpressionandwererankedasfollows:themoststable-IPO8�RPL4�TBP�RPLPO�ACTB�PPIA�HSP90�HPRT1�GADPH�ABL1�CDKN1A�GUSB�RPL30.GeneexpressionstabilitycalculatedbyNormFinderM-valueswerecalculatedforindividualRGsusingNormFinderthatassessedtheexpressionstabilitybycombiningestimatedinter-andintra-groupvariation(Table4).Thegeneswererankedaccordingtoexpres-sionstabilityasfollows:themoststable-TBP�RPLPO�IPO8�ACTB�RPL4�PPIA�HSP90�GADPH�HPRT1�CDKN1A�RPL30�GUSB�ABL1.Thefivebest-rankedgenesTBP,RPLPO,IPO8,ACTB,andturnedouttobethesamefivemoststablegenesfoundbyGeNorm.Moreover,NormFinderallowedstabilityanalysisbe-tweensubgroups:1)benign,2)borderline,3)malignant,4)serousbenignandborderlinetumours5)mucinous,benignandborderlinetumours,6)serousmalignanttumours,and7)endometrioidmalignanttumours(Table5).Combiningthetwomoststablegenesfur-therimprovedtheM-valueingroup-wisecomparison.Inallobtainedcombinations,IPO8followedbyRPL4cameoutasthemoststablegenes.AnalysisofexpressionstabilitybyBestKeeperandequivalencetestInthenextstep,candidateRGswereevaluatedbyBestKeeperandtheEquivalencetestforvariationsinex-pressioninthewholedatasetandbetweentumoursgroupsasdescribedabove.IPO8hadtheloweststan-darddeviation(SD)oftheCvalueacrossthegroups(meanC±SD:29.10±0.65).Thebest-rankedgenesbyGeNormandNormFinderIPO8,ACTB,TBP,RPL4,andRPLPOfulfilledtheBestKeepercriteriaforsta-bilityvariationoftheCvaluewithSD1(Table3). Table3DescriptiveandcorrelationanalysisofthecandidateRGsobtainedbyBestKeeperABL1ACTBCDKN1AGADPHGUSBHPRT1HSP90IPO8PPIARPL30RPL4RPLPOTBP41424241424142424242424242gM[C28.0523.7328.5425.3931.2029.0226.8129.1022.1228.7825.8824.8628.70aM[C28.0723.7528.5725.4231.2329.0426.8429.1122.1528.8125.9024.8828.71min[C25.9021.8026.4323.0227.7526.6324.3027.4819.9126.3423.7922.9127.28max[C30.3925.8731.2327.8034.0631.9129.5530.6424.5331.0627.9826.6631.55SD[±C0.870.731.051.050.990.910.860.650.821.090.770.810.75CV[%C3.103,073.694.113.173.133.192.223.713.782.983.272.62min[x-fold]max[x-fold]4.043.855.854.416.785.646.622.614.734.113.823.066.15SD[±x-fold]1.681.551.881.871.811.721.671.471.631.961.611.661.59GeometricmeanofCCCt]),arithmeticmean(aM[]),minimumandmaximumvaluesofCCCt],max[Ct]),standarddeviationofCCCt]),coefficientofvarianceexpressedasapercentageontheClevel((Ct]),extremevaluesofexpressionlevelsexpressedasanabsolutex-foldover-orunder-regulationcoefficient((x-fold],max[x-fold]),andstandarddeviationoftheabsoluteregulationcoefficients([±x-etal.JournalofOvarianResearchPage4of10http://www.ovarianresearch.com/content/6/1/60 GADPHhadSD�1andhencedidnotmeetthestability criteria. Further,weappliedtheEquivalencetestincludingboth cut-offsof2-foldand3-foldexpressionchangetoiden- tifythebestcandidatesaccordingequivalentexpression ingroup-wisecomparison(Figure2)[8].TheEquiva- lencetestcriteriaat3-foldexpressionchangewereful- filledforIPO8,RPL4,RPL30,GUSB,TBP,RPLPO, ACTB,ABL1,andCDKN1Ainallsubgroups(Table6). GAPDHwasstablyexpressedonlyintwooutofthefive subgroups,followedbyHPRT1,HSB90AB1,andPPIA thatwereequivalentlyexpressedinthreesubgroups usingcut-offof3.However,IPO8wastheonlygene withequivalentexpressionwithin2-foldchangeinall subgroups. Interpretationoftargetgenesexpression InordertoshowtheeffectoftheunstableRGsonthe finalexpressionoftargetgenes,GPERanduPAR mRNAswererelatedtoeitherIPO8andRPL4,or GADPHandHPRT1mRNA.Thechoiceoftargetgenes wasbasedonourpreviousobservationsthatGPER mRNAexpressiondidnotshowanyvariationbetween benign,borderline,andmalignantovariantumoursam- ples[13],whereasuPARmRNAwashigherinborderline andmalignantthanbenignovariantumoursamples[14]. Inaccordancewithourpreviouslypublishedresults, thetissuecontentofGPERmRNAnormalizedtoIPO8 orRPL4mRNAshowednosignificantdifferencesbe- tweenbenign,borderline,andmalignanttumoursam- ples.Incontrast,GPERmRNAnormalizedtoGADPH orHPRT1mRNAwashigherinbenignandborderline tumoursthaninmalignanttumours(Figure3).uPAR mRNAnormalizedtoIPO8orRPL4wassignificantly up-regulatedinborderlineandmalignanttumoursas comparedtobenigntumours,whereaswhenitwas Figure1 Expressionlevelsof13candidatereferencegenesinbenign(BE),borderline(BO),andmalignant(MA)primaryovarian tumours. Valuesaregivenasthecyclethreshold(C t )andareinverselyproportionaltotheamountoftemplate.Expressionlevelsofthegenes studiedareshownaswhiskersboxplots. Table4Rankingof13candidateRGsaccordingtotheir expressionstabilitybyGeNormandNormFinder GeNormNormFinder GeneM-valueGeneM-value IPO80.55TBP0.225 RPL40.55RPLPO0.251 TBP0.58IPO80.253 RPLO0.60ACTB0.264 ACTB0.62RPL40.272 PPIA0.65PPIA0.339 HSP900.67HSP900.357 HPRT10.72GADPH0.373 GADPH0.77HPRT10.396 ABL10.86CDKN1A0.433 CDKN1A0.93RPL300.441 GUSB1.00GUSB0.444 RPL301.10ABL10.515 Kolkova etal.JournalofOvarianResearch 2013, 6 :60 Page5of10 http://www.ovarianresearch.com/content/6/1/60 normalizedtoGADPHorHPRT1mRNAtherewerenodifferencesbetweenthetumourgroups(Figure4).DiscussionAlthoughRT-qPCRisthemostcommonlyusedmethodforassessinggeneexpression,in-depthstudiesofpoten-tialreferencegenesandtheirexpressionpatterninovar-iantumourtissueareinsufficient.TheaimofthisstudywastoidentifythemoststablyexpressedRGs,whichcanberecommendedfornormalizationofRT-qPCRre-sultsinbenign,borderlineandmalignantovariantu-moursamples.WeanalysedthetraditionallyusedRGs,thosereportedasbeingsuitableforovariantissue,andthefourmostpromisinggenesfromacommercialRGarray.Alto-gether13potentialreferencegenesweretestedforsta-bilityacrossgroupsofbenign,borderline,andmalignantprimaryovariantumoursofdifferenthistologicalsub-types.Ofthegenesstudied,IPO8,RPL4,TBP,RPLPO,andACTBwerefoundtobethemoststableaccordingtothestatisticalappletsGeNorm,NormFinderandBestKeeper.OurfindingsonRPL4,RPLPO,andTBPinaScandinavianpopulationareinaccordancewithprevi-ousreportsinAsianpopulations[4,6].Incontrast,ourresultsdidnotsupportPPIAassuitableRG,whichhasbeenobservedpreviously[4].Withregardtothehetero-geneityofovariantumourmaterialsanddifferentrank-ingresultsproducedbythecommonlyusedstatisticalapproaches,wedecidedtofurtheremploytheEquiva-lencetestinouranalysis.ByapplyingstrictcriteriaintheEquivalencetest,i.e.onlyallowinga2-foldchangeofexpression,wecouldidentifyIPO8expressionasthemoststableofallcandidategenestested.WeincludedIPO8inourstudybecauseitshowedlowvariationinexpressionbetweenthebenignandthema-lignantsampleinthecommercialarray.Thisgenewasequivalentlyexpressedacrossthetumoursubgroupsofdifferentmalignantpotentialandhistology.IPO8isaRan-bindingproteinmediatingnuclearimport[15]andhasbeenalreadyreportedstablyexpressedinlungtis-sues[16],gliomas[17],andcoloncancer[18].ThesecondbestRGforgroup-wisecomparison,RPL4,encodesaproteinthatisacomponentofthe60Sribosomesubunit[19].Apartfromovariantissue,ithaspreviouslybeenrecommendedasRGincombinationwithPGK1forexfoliatedcervicalcells[20].RPLPO,an-othergenefromtheribosomalproteinfamily,hadstableexpressioninHPV-positiveasinHPV-negativecervicalsamples[21]andintamoxifenorestrogentreatedbreastcancercells[22].TBP,akeyregulatorofgeneexpres-sion,haspreviouslybeenidentifiedasasuitableRGforexpressionstudiesonhumanhepatitisBvirus-relatedhepatocellularcarcinoma[23],humanrenalcellcarci-noma[24],andglioblastomas[17].RPLPOandTBPalsobelongedtooneofthemoststablyexpressedgenesinbreastcarcinomas[25].TwoothercandidatesthathavenotpreviouslybeentestedasRGsinovariantumourtissue,ABL1andCDKN1A,wereselectedfromthecommercialgenearray.BothgenessatisfiedtheEquivalencetestat3-foldexpressionchange.ABL1,originallyidentifiedasaho-mologueofthetransforminggeneoftheAbelsonmur-ineleukemiavirus,isaproto-oncogene,whichhasbeenimplicatedinmitogenesis,regulationofgenetranscrip-tion,andinhibitionofapoptosis[26].Nucleotidepoly-morphismintheABL1genehasbeenassociatedwith Table5NormFinderrankingof13candidateRGsandcombinationsofthetwobestingroup-wisecomparisonGenenameBE×BO×MABE+BO×MABE×BO+MABE×MASer×Muc(BE+BO)Ser×End(MA)ALB11313131399ACTB246574CDKN1A1212881211GADPH8911111010GUSB7712121112HPRT11087785HSP906115628IPO8532112PPIA9109943RPL3011610101313RPL4113237RPLPO424366TBP351451BestcombinationRPL4/ACTBRPL4/RPLPOIPO8/TBPIPO8/RPL4IPO8/HSP90IPO8/TBPM-value0.1040.0880.0600.0790.0600.073etal.JournalofOvarianResearchPage6of10http://www.ovarianresearch.com/content/6/1/60 riskofovariancancer[27].CDKN1A(alsoknownas p21)wasinitiallydescribedasaninhibitorofcancercell proliferation[27].However,recentstudiessuggestthatit hasdualfunctionssinceitalsomaypromotetumour progression[28]andbeassociatedwithcisplatinresis- tanceinovariancancer[29]. AccordingtoBestKeeperandEquivalencetestcriteria, wefoundthatGADPHhadtheworstexpressionstability inoursetofovariantumoursamples.Similarunfavou- rableresultswereobtainedforHPRT1.Theseobserva- tionsareinlinewithpreviousstudiesonothertissue typesthathavediscourageduseofGADPHandHPRT1 asRGsforclinicallungspecimens[16]andrenalcell cancer[24].Mostrecently,amicroarraystudyidentified agroupofgeneshighlycorrelatedtoGADPHup- regulationinvarioussolidtumours,whichwereand proportionallyassociatedwithadvancedstages[30].Pre- viousreportsonGADPHinovariantissuehaveeither pointedouthigherexpressioninmalignantthanin benigntumoursandnormaltissue[6],ornotmeeting theGeNormstabilitycriteria[4].Wefurtherdemon- stratedthatemploymentofGADPHorHPRT1for Figure2 Variationinexpressionof13candidatereferencegenesanalysedbyEquivalencetestbetweentumourgroups. Differencesof themeans(
)andmatchingsymmetricalconfidenceintervals(-)areshownforthelog2-transformedrelativegeneexpression.Y-axisrepresents thefoldchangeinexpressionamongsubgroups.Thedeviationarea[-l;l]forafoldchange  2lieswithinthedashedlines;thedeviationarea [-2;2]forafoldchange  3lieswithinthesolidlines.Thegeneisconsideredtobeequivalentlyexpressed,ifthesymmetricalconfidenceinterval isapartofthedeviationareaandcontains0init.Thevariationinexpressionofthe13referencegeneswascomparedbetweenbenignvs. borderlineandmalignanttumours (A) ,benignandborderlinevs.malignanttumours (B) ,benignvs.malignanttumours (C) ,mucinousvs.serous benignandborderlinetumours (D) ,andserousvs.endometrioidmalignanttumours (E) . Kolkova etal.JournalofOvarianResearch 2013, 6 :60 Page7of10 http://www.ovarianresearch.com/content/6/1/60 Table6ExpressionstabilityofthecandidateRGsanalysedbyequivalencetest BE×BO+MABE+BO×MABE×MASer×Muc(BE+BO)Ser×End(MA)Totalpasses2-fold/3-fold ABL10/10/10/11/10/11/5 ACTB*0/10/10/11/10/11/5 CDKN1A0/11/10/10/10/11/5 GADPH0/00/00/00/10/10/2 GUSB0/10/11/11/10/12/5 HPRT10/10/00/00/10/10/3 HSP900/10/00/00/10/10/3 IPO8*1/11/11/11/11/15/5 PPIA0/10/00/01/10/11/3 RPL301/10/10/10/11/12/5 RPL4*1/10/10/10/11/12/5 RPLPO*0/10/10/10/11/11/5 TBP*1/10/10/10/11/12/5 Theexpressionwithin(1)oroutside(0)2-fold/3-foldexpressionchangecut-offandthetotalnumberofmeetingthecut-offcriteriainthefivesubg roups. *Genesbest-rankedbyGeNorm,NormFinderandBestKeeper. Figure3 GPERmRNAassayedandnormalizedtoIPO8,RPL4,GADPH,andHPRT1mRNA. Ovariantumoursweresub-groupedaccordingto thehistologicalmalignantpotentialasbenign(BE,n=9),borderline(BO,n=11)andmalignant(MA,n=22).NormalizationtoIPO8andRPL4 showednosignificantvariationoftheGPERmRNAcontentbetweenBE,BOandMAtumours (A , B) .Incontrast,GPERmRNAwashigherin BE/BOcomparedtoMAwhennormalizedtoGADPH(p=0.002)orHPRT1(p=0.008) (C , D) . Kolkova etal.JournalofOvarianResearch 2013, 6 :60 Page8of10 http://www.ovarianresearch.com/content/6/1/60 normalizationresultedinerroneousconclusionson expressionoftargetgenes. Toourknowledge,thisisthefirstreportonRGsin ovariantumoursthatincludeborderlinetumoursin additiontobenignandmalignanttumours.Sincethey areconsideredanon-invasivepre-stageofmolecular typeIovariancancer,itisimportanttoincludethemin anystudyonbiomarkerdiscovery[31]. Ovariancancercomprisestumoursofdifferentmorph- ologyandpathogenesis,whichmayhavedifferentgene expressionprofiles[32].Thereforewewishedtosee whetherthehistologyofovariantumoursinfluencesthe stabilityofRGs.Thus,incontrasttotheprevious studiesconductedexclusivelyonserousmalignanttu- mours,ourstudyalsoincludedmucinousandendo- metrioidtumours.However,smallnumberofsamples insomegroupslimitedthecomparisonsthatcould beperformed. Conclusions Inconclusion,thoroughstatisticalevaluationofour13 candidateRGsidentifiedIPO8followedbyRPL4asthe mostsuitableforthenormalizationofgeneexpression datainbenign,borderline,andmalignantovariantu- mours.Forthefirsttime,IPO8ispresentedasthebest normaliserforgeneexpressionstudiesonovariantumour tissuewithheterogeneoushistologywhenusedasasingle RG.NeitherGADPHnorHPRT1shouldbeusedasRGs forovariantissuestudies,becauseofpoorexpression stability.Normalizingtothesegenesmayerroneously influencethequantificationofthetargetgene(s)and hencereducethereliabilityoftheRT-qPCRresults. Abbreviations RT-qPCR: Quantitativereal-timereversetranscription-polymerasechain reaction;RG:Referencegene;IPO8:Importin8;RPL4:Ribosomalprotein4; GADPH:Glyceraldehyde-3-phosphatedehydrogenase;HPRT1:Hypoxanthine phosphoribosyltransferase1. Figure4 UPARmRNAassayedandnormalizedtoIPO8,RPL4,GADPH,andHPRT1mRNA. Ovariantumoursweresub-groupedaccordingto thehistologicalmalignantpotentialasbenign(BE,n=9),borderline(BO,n=11)andmalignant(MA,n=21).uPARmRNAcontentwashigherin BO/MAthaninBEwhenrelatedtoIPO8(p=0.003)andRPL4(p=0.001) (A , B) .NosignificantdifferenceswerefoundintheamountofuPAR mRNAwhenitwasnormalizedtoGADPHorHPRT1mRNA (C , D) . Kolkova etal.JournalofOvarianResearch 2013, 6 :60 Page9of10 http://www.ovarianresearch.com/content/6/1/60 Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. Authors ’ contributions ZKcarriedoutthegeneexpressionexperimentsanddraftedthemanuscript. AAperformedthestatisticalanaly sis.BCdraftedthemanuscript. SHcontributedmethodologicalknow-how.EKparticipatedinthestudy designanddraftedthemanuscript.Allauthorsreadandapprovedthe finalmanuscript. Acknowledgements ThisstudywassupportedbytheSwedishCancersociety,SkåneUniversity HospitalandRegionSkåne. Authordetails 1 DepartmentofObstetrics&Gynaecology,LundUniversity,SkåneUniversity HospitalLund,Lund,SE22185,Sweden. 2 InstituteofMolecularBiology,NAS RA7HasratyanSt,Yerevan0014,Armenia. 3 DepartmentofImmunology, FacultyofMedicineandDentistry,PalackyUniversity,Olomouc, CzechRepublic. Received:10May2013Accepted:18August2013 Published:30August2013 References 1.BustinSA,BenesV,GarsonJA,HellemansJ,HuggettJ,KubistaM,MuellerR, NolanT,PfafflMW,ShipleyGL,VandesompeleJ,WittwerCT: TheMIQE guidelines:minimuminformationforpublicationofquantitativereal- timePCRexperiments. ClinChem 2009, 55 (4):611 – 622. 2.SiroverMA: Newinsightsintoanoldprotein:thefunctionaldiversityof mammalianglyceraldehyde-3-phosphatedehydrogenase. Biochimicaet biophysicaacta 1999, 1432 (2):159 – 184. 3.ChangTJ,JuanCC,YinPH,ChiCW,TsayHJ: Up-regulationofbeta- actin,cyclophilinandGAPDHinN1S1rathepatoma. OncolRep 1998, 5 (2):469 – 471. 4.LiYL,YeF,HuY,LuWG,XieX: Identificationofsuitablereferencegenes forgeneexpressionstudiesofhumanserousovariancancerbyreal- timepolymerasechainreaction. AnalBiochem 2009, 394 (1):110 – 116. 5.SunY,LiY,LuoD,LiaoDJ: PseudogenesasweaknessesofACTB(Actb) andGAPDH(Gapdh)usedasreferencegenesinreversetranscription andpolymerasechainreactions. PloSOne 2012, 7 (8):e41659. 6.FuJ,BianL,ZhaoL,DongZ,GaoX,LuanH,SunY,SongH: Identification ofgenesfornormalizationofquantitativereal-timePCRdatainovarian tissues. ActabiochimicaetbiophysicaSinica 2010, 42 (8):568 – 574. 7.StefanW: TestingStatisticalHypothesesofEquivalenceand Noninferiority ;2003. 8.HallerF,KulleB,SchwagerS,GunawanB,vonHeydebreckA,SultmannH, FuzesiL: Equivalencetestinquantitativereversetranscription polymerasechainreaction:confirmationofreferencegenessuitablefor normalization. AnalBiochem 2004, 335 (1):1 – 9. 9.KriegovaE,ArakelyanA,FillerovaR,ZatloukalJ,MrazekF,NavratilovaZ, KolekV,duBoisRM,PetrekM: PSMB2andRPL32aresuitable denominatorstonormalizegeneexpressionprofilesinbronchoalveolar cells. BMCMolBiol 2008, 9: 69. 10.PfafflMW,TichopadA,PrgometC,NeuviansTP: Determinationofstable housekeepinggenes,differentiallyregulatedtargetgenesandsample integrity:BestKeeper – excel-basedtoolusingpair-wisecorrelations. BiotechnolLett 2004, 26 (6):509 – 515. 11.VandesompeleJ,DePreterK,PattynF,PoppeB,VanRoyN,DePaepeA, SpelemanF: Accuratenormalizationofreal-timequantitativeRT-PCR databygeometricaveragingofmultipleinternalcontrolgenes. Genome Biol 2002, 3 (7).RESEARCH0034. 12.AndersenCL,JensenJL,OrntoftTF: Normalizationofreal-time quantitativereversetranscription-PCRdata:amodel-basedvariance estimationapproachtoidentifygenessuitedfornormalization,applied tobladderandcoloncancerdatasets. CancerRes 2004, 64 (15):5245 – 5250. 13.KolkovaZ,CasslenV,HenicE,AhmadiS,EhingerA,JirstromK,CasslenB: TheGprotein-coupledestrogenreceptor1(GPER/GPR30)doesnot predictsurvivalinpatientswithovariancancer. JOvarianRes 2012, 5: 9. 14.BorgfeldtC,HanssonSR,GustavssonB,MasbackA,CasslenB: Dedifferentiationofserousovariancancerfromcystictosolidtumorsis associatedwithincreasedexpressionofmRNAforurokinase plasminogenactivator(uPA),itsreceptor(uPAR)anditsinhibitor(PAI-1). IntJCancer 2001, 92 (4):497 – 502. 15.DeanKA,vonAhsenO,GorlichD,FriedHM: Signalrecognitionparticle protein19isimportedintothenucleusbyimportin8(RanBP8)and transportin. JCellSci 2001, 114 (Pt19):3479 – 3485. 16.NguewaPA,AgorretaJ,BlancoD,LozanoMD,Gomez-RomanJ,SanchezBA, VallesI,PajaresMJ,PioR,RodriguezMJ,MontuengaLM,CalvoA: Identification ofimportin8(IPO8)asthemostaccuratereferencegeneforthe clinicopathologicalanalysisoflungspecimens. BMCMolBiol 2008, 9: 103. 17.KrethS,HeynJ,GrauS,KretzschmarHA,EgenspergerR,KrethFW: Identificationofvalidendogenouscontrolgenesfordetermininggene expressioninhumanglioma. NeuroOncol 2010, 12 (6):570 – 579. 18.SorbyLA,AndersenSN,BukholmIR,JacobsenMB: Evaluationofsuitable referencegenesfornormalizationofreal-timereversetranscriptionPCR analysisincoloncancer. JExpClinCancerRes 2010, 29: 144. 19.KlingeS,Voigts-HoffmannF,LeibundgutM,ArpagausS,BanN: Crystal structureoftheeukaryotic60Sribosomalsubunitincomplexwith initiationfactor6. Science 2011, 334 (6058):941 – 948. 20.SteinauM,RajeevanMS,UngerER: DNAandRNAreferencesforqRT-PCR assaysinexfoliatedcervicalcells. JMD 2006, 8 (1):113 – 118. 21.DaudII,ScottME: Validationofreferencegenesincervicalcellsamples fromhumanpapillomavirus-infectedand-uninfectedwomenfor quantitativereversetranscription-PCRassays. CVI 2008, 15 (9):1369 – 1373. 22.ShahKN,FaridiJS: Estrogen,tamoxifen,andAktmodulateexpressionof putativehousekeepinggenesinbreastcancercells. JSteroidBiochemMol Biol 2011, 125 (3 – 5):219 – 225. 23.FuLY,JiaHL,DongQZ,WuJC,ZhaoY,ZhouHJ,RenN,YeQH,QinLX: Suitablereferencegenesforreal-timePCRinhumanHBV-related hepatocellularcarcinomawithdifferentclinicalprognoses. BMCCancer 2009, 9: 49. 24.JungM,RamankulovA,RoigasJ,JohannsenM,RingsdorfM,KristiansenG, JungK: Insearchofsuitablereferencegenesforgeneexpressionstudies ofhumanrenalcellcarcinomabyreal-timePCR. BMCMolBiol 2007, 8: 47. 25.LyngMB,LaenkholmAV,PallisgaardN,DitzelHJ: Identificationofgenesfor normalizationofreal-timeRT-PCRdatainbreastcarcinomas. BMCCancer 2008, 8: 20. 26.ColicelliJ: ABLtyrosinekinases:evolutionoffunction,regulation,and specificity. SciSignal 2010, 3 (139):re6. 27.CunninghamJM,VierkantRA,SellersTA,PhelanC,RiderDN,LiebowM, SchildkrautJ,BerchuckA,CouchFJ,WangX,FridleyBL,Gentry-MaharajA, MenonU,HogdallE,KjaerS,WhittemoreA,DiCioccioR,SongH,Gayther SA,RamusSJ,PharaohPD,GoodeEL: Cellcyclegenesandovariancancer susceptibility:atagSNPanalysis. BrJCancer 2009, 101 (8):1461 – 1468. 28.StivalaLA,CazzaliniO,ProsperiE: Thecyclin-dependentkinaseinhibitor p21CDKN1Aasatargetofanti-cancerdrugs. CurrCancerDrugTargets 2012, 12 (2):85 – 96. 29.XiaX,MaQ,LiX,JiT,ChenP,XuH,LiK,FangY,WengD,WengY,LiaoS, HanZ,LiuR,ZhuT,WangS,XuG,MengL,ZhouJ,MaD: Cytoplasmicp21 isapotentialpredictorforcisplatinsensitivityinovariancancer. BMCCancer 2011, 11: 399. 30.WangD,MoothartDR,LowyDR,QianX: Theexpressionof glyceraldehyde-3-phosphatedehydrogenaseassociatedcellcycle (GACC)genescorrelateswithcancerstageandpoorsurvivalinpatients withsolidtumors. PloSOne 2013, 8 (4):e61262. 31.RomeroI,BastRCJr: Minireview:humanovariancancer:biology,current management,andpathstopersonalizingtherapy. Endocrinology 2012, 153 (4):1593 – 1602. 32.Gomez-RaposoC,MendiolaM,BarriusoJ,HardissonD,RedondoA: Molecularcharacterizationofovariancancerbygene-expression profiling. GynecolOncol 2010, 118 (1):88 – 92. doi:10.1186/1757-2215-6-60 Citethisarticleas: Kolkova etal. : NormalizingtoGADPHjeopardises correctquantificationofgeneexpressioninovariantumours – IPO8 andRPL4arereliablereferencegenes. JournalofOvarianResearch 2013 6 :60. Kolkova etal.JournalofOvarianResearch 2013, 6 :60 Page10of10 http://www.ovarianresearch.com/content/6/1/60