loadingLETTERRESEARCH
ExtendedDataFigure2|Assemblyofagenome-scaleDNAmethylationtimeseriesthroughhumanearlydevelopmentandoverEScellderivation.a,SummaryofRRBSlibrariesgenerated,withnumberofbiologicalreplicates(n),numberCpGscapturedat13,53and103,meanandmedianmethylationvaluesfor100-bptilesestimatedfromCpGscoveredat$53,andmean
EuclideandistanceandPearsoncorrelationacrossbiologicalreplicatesforthesetiles.b,Pearsoncorrelationmatrixforsperm,earlyembryonicandfetaltissuesamples.c,Clusteringofgametic,somaticandpreimplantationmethylationprofilessegregateaccordingtotheirglobalDNAmethylationlandscape,withspermandfetaltissueformingasomaticmethylationclusterthatcontraststheuniqueepigeneticlandscapepresentinpreimplantationembryos.d,SummaryofRRBSlibrariesgeneratedforEScellderivation,withnumberofbiologicalreplicates(n),numberCpGscapturedat13,53and103,meanandmedianmethylationvaluesfor100-bptilesestimatedfromCpGscoveredat$53,andmeanEuclideandistanceandPearsoncorrelationacrossbiologicalreplicatesforthesetiles.‘EScellref’referstoareferencecollectionofpreviouslyassayedEScelllinesaspartoftheNIHRoadmapEpigenomicsProject(Methods).HumanICMandtrophectoderm(TE)wereisolatedthroughlaser-assistedmicrodissection.e,PearsoncorrelationmatrixforhumansamplesusedtomodelEScellderivation.Aconsistentsignatureisrapidlyacquiredbytheoutgrowthstage(p0)andstablymaintainedoveradditionalpassages.
f,Methylationhistogramsfor100-bptilesforhumanblastocystsanddissectedICMandTEtissueshowminimalglobaldifference,whichisalsoobservedwhencomparingpreviouslyassayed,immunosurgicallypurifiedmouseICMtomechanicallydissectedICMandTE.g,Boxplotsofthechangeinmethylation(Dmethylation)for100-bptilesfromcleavagetotheblastocyststageshowpassivedemethylationofDNAmethylation,particularlyforregionsthatexhibitthehighestmethylationlevelsatthisstage.Theredlinesignifiesthemedian,boxesandwhiskersthe25thand75th,and2.5thand97.5thpercentiles,respectively.
?2014Macmillan Publishers Limited. All rights reservedRESEARCHLETTER
ExtendedDataFigure3|ComparisonofhumanEScellderivationtoinvivomousepluripotenttissues.a,Globalmethylationhistogramsof100-bptilesforhumanICMandp5EScells(rows)comparedagainstmouse
preimplantationandpostimplantationembryos(ICM,E6.5Epi),aswellaswithextraembryonicectoderm(ExE)(columns),demonstratetherapidacquisitionofanepiblast-like,somaticmethylationpatternuponEScellderivationinhuman.b,RegionsthatdiscriminatemouseICMfromE6.5epiblastwereusedtoassignhumanICMandEScellstoanequivalentinvivopluripotentstatefor
orthologousfeatures.Thepositionalongtheaxisfrompreimplantation(ICM)topostimplantation(Epi)pluripotencyrepresentstheproportionofregionsinasetthatresembleonestateversustheother.Forallfeaturesets,humanEScellsrapidlyestablishanepiblastidentity,maintainingthissignaturefromtheoutgrowthstageoverensuingpassages.‘EScellref’referstoareferencecollectionofpreviouslyassayedEScelllinesaspartoftheNIHRoadmapEpigenomicsProject(Methods).
?2014Macmillan Publishers Limited. All rights reservedLETTERRESEARCH
ExtendedDataFigure4|Inversecorrelationbetweenexpressionandpromotermethylationisretainedduringhumanpreimplantation.
a,Scatterplotsofoocyte,preimplantationembryoandEScellderivationgeneexpressioncomparedtopromotermethylationdisplayacanonicalnegativecorrelation,evenduringpreimplantationwheretherangeofpromoter
methylationvaluesiscontractedbyglobalhypomethylation.b,Boxplotsofgeneexpressionvaluesforgenessignificantlyupregulatedby$twofoldfromoocyteto8cellcomparedtonon-dynamicgenesandcategorizedbypromotermethylationdynamics.Genesthatarebothdemethylatedandupregulatedareassociatedwithinductionfromasilencedstate,whilethosethatare
demethylatedbutnotupregulateddisplayonlybasalleveltranscriptionthatissignificantlylowerthanobservedinpromotersthatarenotdemethylated.Boldlinesignifiesthemedian,boxesandwhiskersthe25thand75th,and2.5thand97.5thpercentiles,respectively.c,Geneexpressiondynamicsfollowing
fertilizationforhypermethylatedspermpromotersdemethylated$0.5bythecleavagestagecomparedtotherestofpromoters(Other).123of541(22%)demethylatedpromotersdemonstratesignificantupregulation($twofold)comparedtoonly13.6%ofotherpromoters.Moreover,theratioofupregulatedtodownregulatedgenesinthedemethylatedsetsubstantiallyfavourszygoticactivation,whileotherpromotersincludeahigherproportionof
downregulatedmaternaltranscripts(oddsratio51.877,P51.34431028,hypergeometrictest).d,RNA-seqtrackofthepluripotencypromoting,
zygoticallyinducedgenePOU5F1,whosepromoterisdemethylatedfrom0.59inspermto0.02incleavage,concurrentwithitstranscriptionalinduction.
?2014Macmillan Publishers Limited. All rights reservedRESEARCHLETTER
ExtendedDataFigure5|LocalretentionofDNAmethylationissimilarforintronsoverhumanandmousepreimplantation.a,Intronsareclusteredaccordingtotheirdynamicsinhumanandthemethylationoforthologousregionsaretrackedinmouse.Divergenceispredominantlyrestrictedtointermediatelymethylatedfeaturesinhumanspermthataregenerally
hypermethylatedinmouse.TheDmethylationheatmapdisplaysthedifferenceinmethylationvaluesbetweenequivalentpreimplantationtimepoints,withEScellsinhumanservingasaproxyforcomparisontotheE6.5epiblastinmouse.Deviationismostapparentforintermediatelymethylatedhumansperm
introns,wheretheyarelessmethylatedthaninmousesperm.RMSKincluded,repeatmaskerannotatedregionsincluded.nshowsthenumberofintronsavailableforcrossspeciescomparison.b,Whenrepetitiveelementsare
removedfromthecalculationofintronmethylation,theapparentdivergencebetweenmouseandhumanvaluesinspermisdiminished.MethylationandDmethylationheatmapsareasina.Graydenotesmissingvalues(m.v.)whereestimatesforintronicmethylationwereexclusivelyderivedfromrepetitiveelements.RMSK-free:repeatmaskerannotatedregionsexcluded.c,Violinplotsofthetwomaindynamics(maintainedversusdemethylated,toptwoclustersinb)forspermhypermethylatedintronsoverhumanandmousepreimplantationafterrepetitiveelementsareremoved.Asobservedfororthologousexons,regionsthatretainhighmethylationthroughouthumanpreimplantationareconserved,hypermethylatedinbothmousegametes,anddisplaymaintainedregulationasearlyasthezygotestage.
?2014Macmillan Publishers Limited. All rights reserved