Evaluation of PCR conditions for characterizing bacterial ...

... and V4 hypervariable regions of the 16S rRNA genes using primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′). Skiptomaincontent Thankyouforvisitingnature.com.YouareusingabrowserversionwithlimitedsupportforCSS.Toobtain thebestexperience,werecommendyouuseamoreuptodatebrowser(orturnoffcompatibilitymodein InternetExplorer).Inthemeantime,toensurecontinuedsupport,wearedisplayingthesitewithoutstyles andJavaScript. Advertisement nature scientificreports articles article EvaluationofPCRconditionsforcharacterizingbacterialcommunitieswithfull-length16SrRNAgenesusingaportablenanoporesequencer DownloadPDF DownloadPDF Subjects MicrobialecologyMicrobiome AbstractMinION(OxfordNanoporeTechnologies),aportablenanoporesequencer,wasintroducedin2014asanewDNAsequencingtechnology.MinIONisnowwidelyusedbecauseofitslowinitialstart-upcostsrelativetoexistingDNAsequencers,goodportability,easy-handling,real-timeanalysisandlong-readoutput.However,differencesintheexperimentalconditionsusedfor16SrRNA-basedPCRcanbiasbacterialcommunityassessmentsinsamples.Therefore,basicknowledgeaboutreliableexperimentalconditionsisneededtoensuretheappropriateuseofthistechnology.Ourstudyconcernsthereliabilityoftechniquesforobtainingaccurateandquantitativefull-length16SrRNAampliconsequencingdataforbacterialcommunitystructureassessmentusingMinION.WecomparedfivePCRconditionsusingthreeindependentmockmicrobialcommunitystandardDNAsandestablishedappropriate,standardized,betterPCRconditionsamongthetrials.WethensequencedtwomockcommunitiesandsixenvironmentalsamplesusingIlluminaMiSeqforcomparison.ModifyingthePCRconditionsimprovedthesequencingquality;theoptimizedconditionswere35cyclesof95 °Cfor1 min,60 °Cfor1 minand68 °Cfor3 min.OurresultsprovideimportantinformationforresearcherstodeterminebacterialcommunityusingMinIONaccurately. IntroductionMostmicrobesinthenaturalenvironmenthavenotyetbeencultured,butrecentmoleculartechnologicaladvancesmakeitpossibletostudythemwithoutcultivation.Newtechnologieshaveallowedbreakthroughstobemadeintheelucidationofrolesofmicrobesinthenaturalenvironmentandinthefieldsofhumanhealth;forexample,ininvestigationsofthehumangutmicrobiome1,2andinbio-engineeringforagriculture,bioremediationandindustry3.Moleculartechniqueshaveprovidedresearcherswithvariousanalyticalproceduresforunderstandingmicrobialcommunitiesusingclonelibraries4,5,T-RFLP(terminalrestrictionfragmentlengthpolymorphism)analysis6,7,andDGGE(denaturinggradientgelelectrophoresis)techniques8,9.Full-lengthbacterial16SrRNAgeneshavehistoricallybeensequencedusingconventionalmolecularcloningandSangersequencing,butthisapproachistime-consuming,expensive,andhaslowthroughput10.Currently,MiSeqsequencing(Illumina,SanDiego,CA,USA)isthemostwidelyusedplatformfor16SrRNAgeneampliconsequencingformicrobialcommunityanalysis.PCRisconductedonthevariableregions(V2,V3,andV4)inbacteria,withtheprimersfocusingontheconservedregionof16SrRNA11.MiSeq,whichhasbecomepopularforitshighprecision(99.9%),enablesthePCRampliconsequencedeterminationbymergingoverlappedregionofpaired300ntfacingreads12.However,intermsoftaxonomicresolution,comparativeanalyseshaverevealedtheimportanceofthetargetregionandthechoiceoftheprimerpair,asrevealedbythefollowingstudiesinthisarea.Caietal.13reportedtheeffectsofthe16SrRNAgeneprimersetsandrecommendedtheuseofV3andV4primerpairsforseveralenvironmentalsampletypes.Ontheotherhand,Guoetal.14proposedtheuseoftheV1andV2regionsforanalysingthefunctionalbacterialgroupsinasludgesample.However,Wangetal.15recommendedusingtheV5,V6,andV7regionsforascertainingthebacterialcommunitystructureinagingflue-curedtobaccosbecausechloroplastandmitochondrialgeneshavelowerco-amplificationlevels.Kindworthetal.16showedthat,basedonthecomparisonofmicrobialcommunityobtainedusingthemultipleuniversalprimersets,eachuniversalprimersetgeneratesignificantdifferencesintaxonomicspectrum.Theshort-readlengths(100–300-bp)inherentinthesingleuniversalprimersettechniquesalsopreventspecies-levelanalysesinmicrobialecology17. In2012,introductionofthehigh-throughputPacificBiosciences(PacBio,MenloPark,CA,USA)sequencerfacilitatedstructuralanalysisofmicrobialcommunities18.ThePacBioplatformcanobtainfull-length16SrRNAgenesequences,whichincreasestaxonomicresolutionbysequencingthenumberoftheinformativesites.Itsprimarylimitationslieinitslackofversatilityandexemplifiedbytedioussamplepreparation19.In2014,theNanoporeMinIONsequencer(OxfordNanoporeTechnologies,London,UK),nowregardedasbreakthroughinDNAsequencing,wasdeveloped.Itcontainsseveralintriguingfeaturesthatenablereal-time,on-siteanalysesofanygeneticmaterial.Thedevicehasbeenusedindiversewaysinvariousfields,includingdrug-resistancegeneanalysesandassessmentoftherapidgaininreptileandamphibianbiodiversityinrainforests20,21.MinIONstartstobeusedmoreoftenanditssequencingqualityhasbeenimprovingwithhighersequencingreadaccuracyin1Dsequencing(94%).Recently,anincreasingnumberofstudieshavereportedtheirconcernsabouton-siteandreal-timemeasurementsusingMinION22,23,24.TheseMinION-basedgenesequencingtechniqueshaveprovidednewinsightintomicrobialcommunitystructuresmuchmorerapidlyandeasilythaneverbefore.Theoptimization,establishmentandstandardizationofmethodsforthequantitativeevaluationofmicrobialcompositionintheenvironmentareinevitable.Thiswouldallowscientiststoaccuratelyanswerthefundamentalquestionofmicrobialecology:whatkindofandhowmanymicroorganismsarepresentintheenvironment.PCR-based16SrRNAanalysisofbacterialcommunitystructureissubjecttobiasesfromthePCR-relatedconditions.Theseincludethetemplateconcentration,DNApolymerasechoice,numberofcyclesused,amplificationreactiontime,andthereactiontemperature25,26,27,28,29.Bacterialcommunitiescanalsobereconstructedbyonlycollectingthe16SrRNAsequencesobtainedfrommetagenomes,therebyavoidingPCRbias;however,PCR-freelibrariesrequirerelativelylargeamountsofinputDNA,andareimpracticalformanysampletypes30.Therefore,cost-effectivemarkergeneampliconsequencingisoftenpreferredovermetagenomicsequencingformicrobialcommunityanalysisbecauseitenablestheassessmentofuncultivableorganisms.Withthisbackground,theaimofthisstudywastoevaluateMinIONPCRconditionsthroughthreeapproaches:(1)sequencingthefull-lengthbacterial16SrRNAgenefromasinglebacterialspeciestoexamineourbioinformaticspipeline;(2)sequencingtheampliconoffull-lengthbacterial16SrRNAgenefromthreedifferenttypesofbacterialmockcommunityDNAsunderfivedifferentPCRconditions;and(3)sequencingtheampliconoffull-length16SrRNAgenesfromsixenvironmentalsamplestocomparetheresultswiththoseofbacterial16SrRNAV3–V4regionssequencedusingMiSeq.ResultsMinIONdatafilteringbylengthWeinitiallyusedtheRibosomalDatabaseProject(RDP)classifierversion2.11(https://rdp.cme.msu.edu/)31,andtheRDPclassifier16StrainingsetNo:16asdatabase(https://sourceforge.net/projects/rdp-classifier/files/RDP_Classifier_TrainingData/RDPClassifier_16S_trainsetNo16_rawtrainingdata.zip/download).However,thistoolerroneouslyassignedVibrioasAllomonas.Analysiswithanothertool(mothur32)requiredanexcessivelylongruntime.WeeventuallychosetouseBurrows-WheelerAligner(BWA-MEM,v.15.0.7orv.0.7.17)33withadatabasederivedfromtheRDP34asdescribedinMethods.TheMinIONsequencelengthdistributionandspeciesidentificationaccuracywereinvestigatedusingasinglebacterialspecies,Vibriocholerae.Acommerciallyavailablekit(16SRapidSequencingKitand16SBarcodingKit;OxfordNanoporeTechnologies)withprimersforfull-length16SrRNAampliconsequencingontheMinIONplatformwasused.Thedistributionofsequencingreadlengthsshowedthehighestfrequencyataround1,500-basereads.Bothshorter(5-base)andlonger(200,000-base)readsalsoappeared.Three-stepfilteringrangingfrom1,000–2,000bases,1,200–1,800bases,and1,400–1,600baseswasusedtoincludethehighestfrequencylength(1,500-base)ineachstep.AsshowninFig. 1, thehitratio(V.cholerae/totalreads)increasedfrom75%(2,998/3,994reads)withoutfilteringto86%(1,489/1,735reads)afterfilteringwith1,400–1,600bases.Hence,wefilteredthereadsbylength,usingthoseinthe1,400–1,600baserangethereafter.Figure1TheeffectoflengthtrimmingonVibriocholeraedata.Fullsizeimage PCRconditionsbasedonmockcommunitiesTheresultsobtainedfromtheZymoBIOMICSmockcommunitywerecomparedforthefivePCRconditionsshowninFig. 2,withdetailedinformationforeachconditionbeingdescribedinTable1.TheGoodscoveragevaluesweregreaterthan99%forallsamples(Table2).Bray–Curtisdissimilarity35wasusedasameasureforassessingthedifferencebetweentheobservedandtheoreticalcommunitiesstructuresforeachPCRcondition.Bray–Curtisdissimilarityisboundedbetween0and1,where0meansthatthetwocomparedsampleshavethesamecomposition,and1meansthetwositesdonotshareanyspecies36.ThedissimilarityvalueforPCRconditionT0was0.28,whereasthoseforT1,T2,T3andT4were0.40,0.24,0.25and0.24,respectively(indetailandspeciesleveldata,seeSupplementalTableS2).TheinitialT0andpreferedT4conditionswerealsocomparedusingtwoothermockcommunities(thatis,anevenmixof10strains;ATCC10,andanevenmixof20strains;ATCC20)(Fig.3).ForATCC10,thedissimilarityvaluesbetweentheobservedandtheoreticalvaluesforT0andT4werealmostthesame,at0.253and0.257,respectively.However,inthecaseofATCC20,thedissimilarityvaluesforT0andT4were0.338and0.233,respectively.Thistendencywasmorepronouncedatthespecies-leveldata(SupplementalTableS3).WeobtainedabacterialcommunitycompositionsimilartothetheoreticaloneunderT4conditions,comparedwiththatobtainedundertheT0conditions.AlmostallthegeneraexceptforBifidobacteriumweredetectedunderbothconditions.WhentheATCC10andATCC20mockcommunitieswereanalysedusingMiSeq,thedissimilarityvalueswere0.184and0.216,respectively.ThesevaluesaresmallerthanthoseforMinIONunderT0andT4conditions.Figure2ComparisonoffivePCRconditionsusingamockcommunity.Thetheoreticalvaluewasevaluatedbyshot-gunsequencingusingtheIlluminaMiseq(2 × 150 bp).Fullsizeimage Table1PCRconditionsusedinthisstudy.Fullsizetable Table2Sequencereadsandtheproportionsof“Unassigned”generatedpersample.Fullsizetable Figure3Comparisonofbacterialcommunitycompositionofmock10and20samplesusingMinIONundertheT0andT4conditionsversusMiSeq.ThetheoreticalvaluesofmockATCC10and20wereevaluatedbywholegenomeshot-gunsequencingusingtheIlluminaplatform(https://www.atcc.org/~/media/PDFs/Presentations/2017/Microbiome%20Reference%20Standards.ashx).Fullsizeimage PCRconditionsusingenvironmentalsamplesandMiSeqsequencingWeappliedouroptimizedPCRcondition(T4)toenvironmentalsamplescomprisingbathtubinletbiofilms,showerheadfeedwaterandshowerheadbiofilmsfromabathroom(n = 6samples).EachextractedDNApreparedfromthesesampleswasusedseparatelyasaPCRtemplate,and16SrRNAgeneampliconlibrariesweresequencedonbothMinIONandMiSeqplatforms.TheGoodscoveragevaluesweregreaterthan96%forallsamples(Table2).Figure 4showsthe15mostprevalentgenerainthesamples.UnderMinIONT4conditions,thegenusdistributionwassimilartothoseunderMiSeq,butonlyfortheB6sample(Fig. 4B).TheremainingfivesamplesfromMinIONwithT4outputdataresemblethosegeneratedunderMinIONT0conditions.Table2liststhereadnumbersfromMiSeq,MinIONT0andMinIONT4.Regardingtheproportionofthe“Unassigned”category,thevaluesfromMiSeq,MinIONT0andMinIONT4were11.9%(SD11.7),5.5%(SD6.3)and0.0%(SD0.0),respectively.ThevaluesdidnotdiffersignificantlyforMinIONT0toMinIONT4(pvalue = 0.45 > 0.05),orMinIONT0toMiSeq(p-value = 0.36 > 0.05).Conversely,theMinIONT4valuewassignificantlylowerthanthatofMiSeq(pvalue = 0.044).Atthegenuslevel,thebacterialcompositionsfromMinIONT4showedgreatertaxonomicresolutionthanthosefromMiSeq.Figure4Comparisonofthebacterialcommunitycompositionofenvironmentally-sourcedsamplesusingMinIONunderT0andT4conditionsversusMiSeq.Thetop15mostabundantgeneraarelisted.Fullsizeimage DiscussionWeinvestigatedwhetherfractionalchangesintaxonomicassignmentandbacterialcommunitycompositionexistinthecomparisonofPCRconditionsusingtheMinIONsequencer(OxfordNanoporeTechnologies)withmockcommunityandenvironmentalsamples.WealsocomparedtheresultsfromthebacterialcommunitysamplesfromMinIONwiththosefromtheMiSeqsequencer(Illumina).Thesearchfornewanalyticaltoolswithshorterruntimeshasprogressedconsiderablywiththethird-generationMinIONsequencingplatform,becauseofitsrapidandeasyhandling37.Theincreasedinformationcontentinherentfromlongerreadlengthshelpresearcherswithalignment-basedtaxonomyassignment17.Withtheabilitytogeneratelongerreadlengths,MinIONanalysiscantargettheentire16SrRNAgenecodingregiontoofferhighlyaccurate,sensitiveandrapidpathogendetection20,38.Ourgoalwastodeterminethebetterconditionsunderwhichaccuratebacterialcommunitystructuringdatacouldbeobtainedusingananoporesequencer.DNAamplificationwasperformedfor35cyclesinallourPCRprotocols.SeveralstudieshaveshownthatlargerPCRcyclenumberscausechimeragenerationandinterferewithbacterialcommunitystructureanalysis39,40.Hence,minimizingthenumberofPCRcyclesbyoptimizingthestartingtemplateconditionsandconcentrationsisimportant26.However,inoursituation,reducingthenumberofPCRcyclestolessthan35decreasedthenumberofenvironmentalDNAsamplesthatwereamplified.Tapwaterhasrelativelylowermicrobialdensitythanthatinotherenvironmentalsamplessuchasseawaterandsoil41.Therefore,theproceduresusedinthisstudymayalsobeappliedtosampleswithlowmicrobialcelldensities,suchasatmospheric(~ 104cellsm−3)42samples,too.Thesequencingdatafromthestrictestfiltrationrange(1,400–1,600bases)provided86%matchingtoV.choleraeinourdatabase(Fig. 1).Thisindicatesthathigh-resolutionanalysisatthespecieslevelispossiblewithMinIONbyeliminatingextraneousreaddata.Jethroetal.11statedthatbyusingfull-lengthsequencesitispossibletoclassifynearlyallenvironmentalsequencesintocorrectspecies.Thereadnumber(1,735reads)afterthistreatment(43%of3,994readsintotal)wasusedforsubsequentanalyses.Decreasingthereadnumbercausednoprobleminthisexperimentbecausetheprecisemappingofonlyonespecies(V.cholera)wasthemainobjective.Mitsuhashietal.23andNakagawaetal.43reportedthata5-minand3-minrunningtimeonMinION,respectively,wereenoughfordetectingspecificbacteria.However,deepersequencingisrequiredtoobtainbetterestimatesofbacterialcommunitystructureandhigherGoodscoveragevalues44,45.Weconducteda48-hMinIONoperationforthemockcommunitiesandenvironmentalsamplestoprovidesufficientreadnumbersinourstudy(Table2).OptimalPCRconditionsneedtobeestablishedtoobtainaccuratebacterialcommunitystructureanalysesusingMinION.WethereforecomparedfivedifferentPCRconditionsusingmockcommunitiesinpreliminaryexperiments(Table1).ThedissimilarityvalueswithinthecommunitiesweresmallerwiththeT2condition(longerannealingtimethanT0),theT3condition(longertimeforbothannealingandextensionthanT0)andtheT4condition(longertimeforallstagesthanT0),thanthatoftheT0condition(Fig. 2).Theseresultssuggestthatthepolymeraseextensiontimedoesnotaffectthebacterialcommunitystructureanalysis.Conversely,ashorterannealingtime,asintheT3condition,resultedinrelativelyhigherdissimilarityvaluescomparedwiththosefromtheothercases(T2andT4).ConsideringthehigherdissimilarityachievedunderT1togetherwiththeresultsunderT3,alongerannealingtimewasdeemednecessaryfortheproperassessmentofbacterialcommunitystructureusingfull-length16SrRNAPCRanalysis.AsshowninFig. 3,attheATCC10,thedifferencebetweenT0andT4conditionisnotsignificant.WhereasattheATCC20,thebacterialcompositionobtainedfromT4conditionwasclosertothetheoreticalvaluesthanthoseobtainedfromtheT0condition.Thisismorepreciseatthespecies-level(SupplementalTableS3).TheseresultssuggestthattheT4PCRconditionswithlongerreactiontimesprovidebetterresultsthantheT0conditionwhenthesamplediversityishigh.However,BifidobacteriumwasnotdetectedbyMinIONanalysisusingeithertheT0orT4conditions.Previouspublicationshaveshownthattheuniversalprimerscommonlyusedformetagenomicanalyses(suchasthe27Fprimer)possesslimitationsrelatedtoamplificationbias.The27Fforwardprimerusedinthe16SBarcodingKit(SQK-RAB204,OxfordNanoporeTechnologies)containsthreebase-pairmismatchesagainstBifidobacterium(27Fprimer:5′-AGAGTTTGATCMTGGCTCAG-3′);thatis,thesequenceoftheB.adolescentisprimersiteis5′-AGGGTTCGATTCTGGCTCA-3′(themismatchedbasesareunderlined)46.Conversely,Huetal.47,forexample,detectedBifidobacteriumspeciesbysequencingwithuniversalprimers(384Fand806R)inMiSeq.Thus,primersequencemodificationsarerequiredtoavoidpreferentialdetectionofparticulartaxa,sothatabroadrangeofbacteriaspeciesiscovered,aswasthecaseherewithourB.adolescentis.MinIONhaslowerreadaccuracybutcangeneratemuchlongerreadlengthsthanthosefromMiSeq.Nygaardetal.48analysedbuilding-dustmicrobiomesusingMinIONandMiSeqandshowedthat,atthegenusandspecieslevels,MinIONreportedgreatertaxonomicresolutionthanMiSeq.Longreadshelpalignment-basedassignmentoftaxonomyaswell,becauseoftheirincreasingtaxonomicalinformationcontent.Inthisstudy,undertheMinIONT4condition,alltheenvironmentalsamplesshowedbettertaxonomicresolutionatthegenuslevelthanthatunderMiSeq,thesameaspreviouslyreported17,48.ManypapershavebeenpublishedonsoftwaredevelopmentsandshorterrunningtimeswithMinION.Forexample,Kaietal.38reportedonthepossibilityofdecreasingthesequencingtimeofMinIONbydirectPCRapproachesandfoundthata3-minsequencingrungeneratedasufficientnumberofreadsfortaxonomicassignmentandlessthantwohourswasrequiredforidentifyingappropriatebacterialspecies.Characterizationofenvironmentalbacterialcommunitiesrequiresbothqualitativeandquantitativeinformationthroughappropriatesequencereadfilteringaswellasexperimentalprocedures.Here,wehavedemonstrated,forthefirsttime,thattheaccuratedataonbacterialcommunitiesusingMinIONcanbegeneratedbycomparingandchoosingappropriatePCRconditions.ThereactionconditioninthisstudyarethelongestamongPCRconditionscomparedtopreviousstudiesonbacterialcommunitystructureanalysiswith16SrRNAwiththeMinIONsystem(seeSupplementalTableS4);however,usingthisconditionwewereabletoobtainbacterialcommunitystructuresthatwerecomparableinqualitywithMiSeq.MethodsSampleandDNApreparationThefull-lengthbacterial16SrRNAgenefromV.choleraeDNA,obtainedthroughthecourtesyofDr.TaichiroTakemura(NagasakiUniversity,Japan),wasusedtoexaminethebioinformaticspipeline.PCRconditionsevaluationwasinitiallyperformedusingareferencegenomicDNA(ZymoResearchCorp.,Irvine,CA,USA;https://www.zymoresearch.com).TheZymoBIOMICSmicrobialcommunityDNAstandard(ZymoBIOMICScatalog#D6305)containsamixtureofgenomicDNAsisolatedfromthepureculturesofeightbacterialandtwofungalstrains,andanequalmolarquantityof16SrDNAfromeachorganismisprovided.PCRconditionswereexaminedforthemockcommunityDNAsamples(10StrainEvenMixGenomicMaterial(MSA-1000)and20StrainEvenMixGenomicMaterial(MSA-1002);AmericanTypeCultureCollection(ATCC),Manassas,VA,USA)aswellasenvironmentalbiofilmandwatersamples(thatis,theinsidesofshowerheads,bathtubinletsandshowerheadfeedwater)inJapan.Thebiofilmsampleswerecollectedasdescribedpreviouslywithaswab49.Twolitterofshowerheadfeedwaterwasfilteredon-siteusinga50-mLsyringe(Terumocorporation,Tokyo,Japan)and0.2-µmfiltercartridge(Sterivex,Millipore,MA,USA).Thesampleswereimmediatelyputinacoolbox,carriedtothelabandkeptat− 20 °C.DNAwasextractedfromthesesamplesusingtheDNeasyPowerBiofilmKit(QIAGEN,Germantown,MD,USA)withslightmodification50.TheextractedDNAwaskeptat− 20 °CafterpurificationandprecipitationusingDr.GenTLEprecipitationcarrier(TakaraBioInc.).TheconcentrationsandpuritiesoftheextractedDNApreparationsweredeterminedusingtheSpectro/Fluorometer(DS-11FX+,DeNovix,Wilmington,USA)andQuantiFluordsDNASystem(Promega,Madison,WI,USA).PCRconditionsPolymeraseamplificationefficiencywasinitiallycheckedusing13biofilmandwatersamplescollectedfrombathrooms.FivedifferentDNApolymeraseenzymesweretestedandMightyAmpDNApolymerasev.2(TakaraBioInc.)providedthehighestamplificationefficiencyamongotherpolymerasesusedinourevaluation,asreportedpreviouslyelsewhere51.ThePCRswereconductedusingaprimerpair(27Fand1492R)specificforthe16SrRNAgene-targetingsequencecontainedinthelibrarypreparationkit(SQK-RAS201orSQK-RAB204,OxfordNanoporeTechnologies).Somesampleswerebarcodedusingarapidampliconbarcodingkit(SQK-RAB201,OxfordNanoporeTechnologies)accordingtothemanufacturer’sprotocol(seeSupplementalTableS1fordetails).ThefirstPCRcondition(T0)involvedapre-heatingstepat98 °Cfor2 min,35cyclesat98 °Cfor10 s,60 °Cfor15 sand68 °Cfor2 min.AlternativePCRconditions(T1–T4)forthedurationofeachsteparelistedinTable1.ThePCRonV.choleraeDNAwasperformedusingtheT0condition.Table1showsthefourdifferentPCRconditions(T1–T4)thatwereusedwiththeZymoBIOMICSmockcommunitysample.TheATCCmockcommunityandtheenvironmentally-sourcedsampleswereamplifiedusingtwodifferentPCRconditions(T0andT4).Theamplifiedfragmentswereseparatedon2%agarosegels,stainedwithSafelookLoad-Green(Wako,Osaka,Japan),andvisualizedontheFASNanoGelDocumentSystem(NipponGenetics,Tokyo,Japan).NanoporesequencinglibraryconstructionAfterpurifyingthePCRproducts(50 μleach)with30 μlofAgencourtAMPureXPbeads(BeckmanCoulter,Tokyo,Japan),theamountandpurityofDNAelutedwith10 μlofbuffersolution(10 mMTris-HClpH8.0,with50 mMNaCl)wasdeterminedusingaSpectro/Fluorometer(DS-11FX+,DeNovix)andQuantiFluordsDNAsystem(Promega).PurifiedampliconDNA(100or50fmol)wasusedasinputDNAfortheMinION-compatiblelibraries.Theampliconswereaddedto1 μlofrapidadapter(OxfordNanoporeTechnologies)andincubatedatroomtemperaturefortherequiredtime.Nanoporesequencingandbase-callingThenanoporesequencinglibrarieswereseparatelyrunonFLO-MIN106R9.4flowcells(OxfordNanoporeTechnologies)afterperformingplatformqualitycontrolanalysis.Theampliconlibrary(11 μl)wasdilutedwithrunningbuffer(35 μl)containing3.5 μlofnuclease-freewaterand25.5 μlofloadingbeads.A48-hsequencingprotocolwasinitiatedusingtheMinIONcontrolsoftware,MinKNOWfromv.1.6.1-1.10.23(SupplementalTableS1containsthedetailedinformation).MinIONsequencereads(thatis,fast5data)wereconvertedintofastqfilesusingAlbacorev.1.2.1orv.2.1.3software(OxfordNanoporeTechnologies)andfilteredwiththethresholdofameanqualityscoreover7.(SupplementalTableS1containsthedetailedinformation).Figure 5showsthestudy’sworkflow.Figure5Overviewoftheexperimentalworkflow.Experimentswereperformedasshownintheflowchart.Fullsizeimage NanoporesequencingdataanalysisSequencelengthdistributionwasexaminedforeachbase-calledfastqfileusingFastQC(v0.11.2)(https://www.bioinformatics.babraham.ac.uk/projects/fastqc/)52.SeqKit0.8.0(https://bioinf.shenwei.me/seqkit/)53andoriginalrubyscriptwereusedtofilterthesequencedatabythreelengthsof1,000–2,000,1,200–1,800and1,400–1,600toincludethehighestreadlengthfrequency(1,500bases),withthefilteringeffectsexaminedbycalculatingthehitratiotoV.choleraeinalltheleads.AlldataexceptforV.choleraewerefilteredwithareadlengthof1,400–1,600.Afterfiltering,thesequencereadsweremappedusingBWA-MEM(v.15.0.7orv.0.7.17;https://github.com/lh3/bwa)33withtheMinIONanalysisoption(-xont2d)54toadatabasederivedfromtheRDP(RDPRelease11,Update5,Sept.30.2016;3,356,809alignedandannotated16SrRNAsequences)34andthetophitwasusedforthegenusandspeciesassignment.TheRDPhierarchybrowser(https://rdp.cme.msu.edu/hierarchy/hb_intro.jsp)wasusedwiththefollowingfilters:strain = “Type”;source = “isolates”;size“≥ 1,200”;quality = “Good”;taxonomy = “Nomenclatural”togenerateadownloadedsetof12,227sequences.IlluminasequencingThe16SrRNAsequencinglibrarywasconstructedaccordingtotheIllumina16SMetagenomicSequencingLibraryPreparationprotocol(Illumina)targetingtheV3andV4hypervariableregionsofthe16SrRNAgenesusingprimers341F(5′-CCTACGGGNGGCWGCAG-3′)and805R(5′-GACTACHVGGGTATCTAATCC-3′)16.MightyAmpDNAPolymerasev.2(TakaraBioInc.)wasusedforthePCRs.TheinitialPCRwasperformedusingregion-specificprimerstoensurecompatibilitywiththeIlluminaindexandsequencingmultiplexadapters.Theamplifiedfragmentswereseparatedon2%agarosegels,stainedwithSafelookLoad-Green(Wako),andvisualizedontheFASNanoGelDocumentSystem(NipponGenetics).TheamountofpurifiedDNArecoveredwasquantifiedusingaSpectro/Fluorometer(DS-11FX+,DeNovix).AnequimolarmixtureofallPCRproductswassenttoacommercialcompanyfor2 × 300 bppaired-endsequencingontheMiSeqplatformusingIlluminaMiSeqv3ReagentKit(Fasmac,Kanagawa,Japan).IlluminasequencingdataanalysisIllumina16SrRNAampliconsequencedataweredemultiplexed,andindexsequenceswereremovedusingMiSeqControlSoftware(MCS)v2.6.Pairedforwardandreversesequencesweremergedusing‘make.contings’withthedefaultparameterofmothur32(v.1.39.5).ThemergedsequencereadswereassignedtaxonomyusingBWA-MEM33againstRDP34,usingthesamedatabaseandparameterswithout‘ont2d’optionasournanoporesequencedata.DataanalysisAlldataanalysiswascarriedoutwithR(v.3.3.1)55.BacterialcommunitydissimilaritiesforthedifferentPCRconditionswerecalculatedbytheBray–Curtisindexwiththe‘vegan’package(v.2.5-5)35.Initially,MiSeqreadswererandomlysampledtoeliminatereadnumberdifferenceswhencomparingofunassignedpercentagesintheMiSeq,MinIONT0andMinIONT4runs.Afternormalizationandconfirmingtherewerenosignificantdifferencesbetweenthenumberofreads(pvalue = 0.6758 > 0.05),theunassignedratiowascomparedusingTukey’shonestsignificantdifferencetest,andthevariancesinthedatafromthethreegroupswerefoundnottobeequal(Fvalue