LETTER
DNAmethylationdynamicsofthehumanpreimplantationembryo
doi:10.1038/nature13581
ZacharyD.Smith1,2,3,4*,MichelleM.Chan1,5*,KathrynC.Humm3,6,7,8,9*,RahulKarnik1,2,3,ShilaMekhoubad3,4,AvivRegev1,9,10,KevinEggan1,2,3,4,11&AlexanderMeissner1,2,3Inmammals,cytosinemethylationispredominantlyrestrictedtoCpGdinucleotidesandstablydistributedacrossthegenome,withlocal,cell-type-specificregulationdirectedbyDNAbindingfactors1–3.Thiscomparativelystaticlandscapeisinmarkedcontrastwiththeeventsoffertilization,duringwhichthepaternalgenomeisgloballyreprogrammed.PaternalgenomedemethylationincludesthemajorityofCpGs,althoughmethylationremainsdetectableatseveralnota-blefeatures4–7.Thesedynamicshavebeenextensivelycharacterizedinthemouse,withonlylimitedobservationsavailableinothermam-mals,anddirectmeasurementsarerequiredtounderstandtheextenttowhichearlyembryoniclandscapesareconserved8–10.Wepresentgenome-scaleDNAmethylationmapsofhumanpreimplantationdevelopmentandembryonicstemcellderivation,confirmingatran-sientstateofglobalhypomethylationthatincludesmostCpGs,whilesitesofresidualmaintenanceareprimarilyrestrictedtogenebodies.Althoughmostfeaturessharesimilardynamicstothoseinmouse,maternallycontributedmethylationisdivergentlytargetedtospecies-specificsetsofCpGislandpromotersthatextendbeyondknownimprintcontrolregions.Retrotransposonregulationisalsohighlydiverse,andtransitionsfrommaternallytoembryonicallyexpressedelements.Together,ourdataconfirmthatpaternalgenomedemeth-ylationisageneralattributeofearlymammaliandevelopmentthatischaracterizedbydistinctmodesofepigeneticregulation.
Wegeneratedgenome-scalemethylationmapsofhumanpreimplan-tationdevelopmentusingreducedrepresentationbisulphitesequenc-ing(RRBS)toaccommodateminimalDNAinputs7.Wethawedandscreenedmorphologicallynormalcleavagestageembryosandblasto-cyststorepresentearlyandlatepreimplantation,tworeplicatesofpooled,matchedinnercellmass(ICM)andtrophectoderm,motilespermfromfourunrelatedhealthydonors,andfetaltissues(ExtendedDataFig.1).Toestimatethetime,extentandtargetsofglobalremethylation,wegeneratedderivationtimeseriesofthreehumanembryonicstem(ES)celllines,collectingtheprimaryoutgrowth,firstandfifthpassageperline.Onaverage,replicatesshowedhighreproducibilityandcaptured1,753,958CpGsofmethylationdataat103coverage(ExtendedDataFig.2).
WenotedtwodistinguishablearchitecturesforDNAmethylationacrossthistimeseries:somatic-likeCpG-density-dependentbimodalityinsperm,EScellsandfetaltissues,andextensiveCpG-density-independenthypomethylationinpreimplantationembryos(Fig.1a).Thesubstan-tialintermediatemethylationinspermreflectsdisparaterepetitiveele-mentregulation,thoughnon-repetitivesequencesstillfitthesomaticparadigm11(Fig.1b).AlmostnohypermethylatedCpGspersistintocleavage,withresidualmethylationdiminishingfurtherintotheblas-tocyst,indicatingthattheembryoniclandscapeisrapidlyestablishedbeforethethirdembryonicdivision(Fig.1bandExtendedDataFig.2f).
1Withoutaccesstohumanepiblasts,invivocharacterizationofglobalremethylationisunavailable,butcross-speciescomparisonbetweenmouseepiblastandhumanEScellssuggestthattheyareareasonablyproxyforpostimplantationpluripotency(ExtendedDataFig.3).Notably,withinprimaryEScelloutgrowths,globalremethylationisnearlycomplete,includingforintermediatelymethylatedrepetitiveelementsapparentinsperm(Fig.1b).
Despitepredominanthypomethylation,erasureisnotthedefactofateofallloci.Non-repetitive100-bpgenomictileswereclusteredusingk-meansinto10dynamicpatterns.Ofspermhypermethylatedtiles,45%retainsomemethylationoverpreimplantation,and23%displayhighenoughlevelsincleavageembryostobebiparentallyinheritedandatleastpartiallymaintained(Fig.1c).Localmaintenanceissignificantlyweightedtogenebodies:only31%of40,486spermhypermethylatedintergenictilesare$0.2methylatedincleavageembryos,comparedto57%or59%ofexonsorintrons,respectively(Fig.1d,e).Frequently,genebodymethylationextendsfortenstohundredsofkilobaseswithinasinglegene(Fig.1f).SitesofretainedembryonicmethylationsuggestresidualDNAmethyltransferaseactivitywithinaphasewheremain-tenanceappearsotherwiseimpeded12.
WeincorporatedrecentlypublishedRNA-sequencing(RNA-seq)datatointerprettherelationshipbetweenDNAmethylationandexpression13(SupplementaryTable1).Despiteglobalhypomethylation,thecanonicalnegativecorrelationbetweenpromotermethylationandgeneexpres-sionextendstopreimplantation,althoughtheoverallrangeinpromotermethylationiscontracted(ExtendedDataFig.4a).Fewdemethylatedpromotersaretranscribed,suggestingthatpromoterdemethylationlargelyreflectstheglobaltrend(ExtendedDataFig.4b).However,deme-thylatedpromotersaremorefrequentlyinducedthanrepressedandincludePOU5F1(alsoknownasOCT4),whoseembryonicinductionisessen-tialfordevelopment14(ExtendedDataFig.4c,d).Thus,ourpreimplan-tationdatasupportmodelswheredistinctmechanismsmayregulateglobalversustargetedreprogramming15.
TheglobalDNAmethylationdynamicsofthehumanembryocloselymirrorthoseofthemouse,withsharptransitionsbothintoandoutofpreimplantation7,16(Fig.2a).Weinvestigatedthebehaviourofortho-logousexonswithinthemousepreimplantationtimeline7,predicatedontheirhumandynamics.Surprisingly,simplysortingmouseexonsaccordingtothedynamicsoftheirhumanorthologuesrecapitulatedsimilartrendsinspermandoverpreimplantation(Fig.2b).Inmouse,demethylatedexonsareerasedearly,withmoderatepassivedepletionovercleavage(Fig.2c,d).Alternatively,exonsthatmaintainmethyla-tionbehavesimilarlyinmouseandarehypermethylatedinbothgam-etes(Fig.2c,d).Introndynamicsrevealedcomparabletrends,butonlyafterrepetitiveelementswereremovedfrommethylationestimates(ExtendedDataFig.5).Thus,bothspeciespassthroughanequivalent
BroadInstituteofMITandHarvard,Cambridge,Massachusetts02142,USA.2HarvardStemCellInstitute,Cambridge,Massachusetts02138,USA.3DepartmentofStemCellandRegenerativeBiology,HarvardUniversity,Cambridge,Massachusetts02138,USA.4DepartmentofMolecularandCellularBiology,HarvardUniversity,Cambridge,Massachusetts02138,USA.5ComputationalandSystemsBiologyProgram,MassachusettsInstituteofTechnology,Cambridge,Massachusetts02142,USA.6DivisionofReproductiveEndocrinology&Infertility,DepartmentofObstetrics&Gynecology,BethIsraelDeaconessMedicalCenter,Boston,Massachusetts02215,USA.7Obstetrics,Gynecology,andReproductiveBiology,HarvardMedicalSchool,Boston,Massachusetts02215,USA.8BostonIVF,Waltham,Massachusetts02451,USA.9HowardHughesMedicalInstitute,MassachusettsInstituteofTechnology,Cambridge,Massachusetts02142,USA.10MassachusettsInstituteofTechnology,Cambridge,Massachusetts02142,USA.11HowardHughesMedicalInstitute,HarvardUniversity,Cambridge,Massachusetts02138,USA.*Theseauthorscontributedequallytothiswork.00MONTH2014|VOL000|NATURE|1
?2014Macmillan Publishers Limited. All rights reservedRESEARCHLETTER
a
Count (×105)420>105<1
Sperm
n = 1,023,928
Cleavage
n = 981,622
Blastocyst
n = 936,354
ICMTEBlast.
ES cell
n = 793,094
p0p1p5
Somatic
n = 674,198
BrainHeartLung
00.5100.5
100.510Methylation
So0.5100.51
ceSollm1.00.8Fraction0.60.40.20.0
ESSplBlCmbc
≥0.80.2–0.8<0.2
d
DemethylationLowMethylation0
0.5
1Sp
–Log10 P val0Cl
800
Bl
ES cell
Methylationn = 245,6670
0.5
1
Som
e
MethylationFigure1|Humanpreimplantationembryosaregloballyhypomethylated.a,Top,DNAmethylationacross100-bptilesforhumansperm,
preimplantationembryos,includingtheICMandtrophectoderm(TE),
EScellderivationfromoutgrowthtofifthpassage(p5),andsomaticfetaltissuesrepresentingallgermlayers.Greyhighlightstheaverage.nequalstheaveragenumberoftilescapturedforagivenstage.Theyaxisshowsthenumberoftilesineachmethylationbin.Bottom,boxplotsofmethylationatdifferentlocalCpGdensities(yaxis).Bullseyesindicatethemedian,boxesandlinesthe25thand75th,and2.5thand97.5thpercentiles,respectively.b,Barplotsof100-bptilessegregatedbynon-repetitive(unique)orrepetitivedesignationandbinnedbymethylationstatus.Bl,blastocyst;Cl,cleavage;Som,somatic;Sp,sperm.‘EScell’and‘Som’showtheaverageofthesetimepoints.
c,Non-repetitive100-bptilesareclusteredviak-meansinto10dynamics.Spermhypermethylatedsequencesfollowthreegeneraltrajectories:persistentmaintenance,incompleteorcompletedemethylation.Otherdynamicsincludespermspecifichypermethylationandhypomethylationsharedbysperm
andtheearlyembryofollowedbydenovomethylationinEScells.Finally,3,586tilesarehypomethylatedinspermandEScellsbutmethylatedinembryos,representingtransientimprint-likesignatures.nequalsthenumberoftilessharedbetweenthetimepointsandusedforclustering.d,Dynamicsforspermhypermethylated,non-repetitivetilesasclusteredinc.Leftheatmap,per-clusteraverageoftiles.Rightheatmap,–log10Pvalueofhypergeometricenrichmentforeachclusterforintergenic,exonic,intronic,CGIorTSS
annotationsusingspermhypermethylatedregionsasthebackground.e,Violinplotforspermhypermethylatedintergenic(Inter),exonicandintronicfeatures.f,TheOBSCNgeneexhibitshighinter-andintra-genicmethylationandanunmethylatedpromoterinspermandEScells.Incleavageembryos,a130-kbregion,highlightedinblue,remainsspecificallymethylatedwhiletheperipheryisdemethylated.EachdotreferstoaCpGcapturedbyRRBS.TheyaxisforDNAmethylationrepresentsthefrequencyinwhichcapturedCpGsaremethylatedandrangesfrom0to1.
CpG Densityell cHigh1.00.80.60.40.20.0
Sperm methylation levelSpCl/ESBl cInellteExroInntroCnGTSISUReniqpeuetiUtivReniqepeuetUitivReniqepeuetUitivReniqepeuetUitivReniqepeuetitiveESSpClBlIntExeroInntroInntExeroInntroInntExeroInntroInntExeroInntroInntExeroInntronnearlyallofwhichweremonoallelicallymethylated(Fig.3bandExtendedDataFig.7).
GiventhatmaternalDMRsinmousearealsoenrichedforCGIs,wenextexaminedtheconservationoftargetedlocibetweenspecies(Fig.3c).Wefoundthat795and293CGIsbehaveastransientDMRsinhuman
50 kb
8 CICMEpif
1
Sperm
0
1CleavageMethylationCount (×105)0
1Blastocyst0
1p0 ES cell0
Chr 1: 228,370,831–228,573,181
OBSCN
CpG islandsSpn = 867,594n = 559,9741210210–101–101Bl to ES cell/Epin = 832,109n = 507,724Cl to Bln = 833,594n = 512,594Exon dynamics0n = 11,055
0
ClBl2ES cSpella
HumanMouse
Sp to Clb
HumanMouse
Δ Methylation
Methylation
0.5
11
Difference
>0<
1
Δ Methylation
globalreprogrammingwithcongruentkinetics,whilemanyorthologousregionsmaintainmethylationthatdecayspassively.
Transient,maternallyinheritedmonoallelicmethylationhasbeenpreviouslyobservedinmouse5–7.Toidentifycandidatelociinhumanwithoutaccesstooocytes,wesearchedforregionsthataresignificantlymoremethylatedinpreimplantationembryosthaninsperm,asthismeth-ylationwouldbelikelytobeofmaternalorigin(ExtendedDataFig.6aandMethods).Usingourcriteria,weidentify5,265100-bpcandidatematernaldifferentiallymethylatedregions(DMRs),includingmostcanon-icalICRs.WeclusteredtheseregionsbytheirresolutioninEScellsandfoundthatthemajorityarepreimplantation-specificandeitherhyper-orhypomethylatedinsomatictissue(Fig.3a).ThelocationandCpGdensityofDMRtilesdependsontheirresolution,withsomaticallyhypo-methylatedDMRssubstantiallyenrichedforCpGisland(CGI)-containingpromoters,whereassomaticallyhypermethylatedDMRsaremorelikelytobeintragenicanddistributedfurtherdownstream(Fig.3aandExtendedDataFig.6b,c).Toconfirmthatthesesignaturesrepresenttrueimprint-likemaintenance,wegeneratedRRBSlibrariesoftwounrelated,singleblastocystsandidentifiedCpGsthatcouldbeassignedtoeachallele,
2|NATURE|VOL000|00MONTH2014
c
1.0Exonmethylation0.50.0HumanMouseMaintainedDemethylatedl8 CMIC2 CFigure2|Humanpreimplantationdynamicsaregloballysimilartomouse.a,Histogramsofmethylationchanges(Dmethylation)for100-bptilesacrosshumanandmousepreimplantationfromfertilization(SptoCl)throughpreimplantation(CltoBl)toglobalremethylationatimplantation,asmeasuredfromblastocysttoEScellinhumanandICMtoembryonicday6.5(E6.5)epiblastinmouse(BltoEScell/Epi).nshowsthenumberoftilescontributingtoeachcomparison.b,Exonsclusteredbydynamicsinhumanwithequivalentmethylationvaluesfororthologoussequencesinmouse.TheDmethylationheatmapdisplaysthedifferenceinmethylationformatchedtimepoints.
nshowsthenumberofexonsavailableforcrossspeciescomparison.c,Violinplotsoforthologoushumanspermhypermethylatedexonsclassifiedasmaintainedversusdemethylated(toptwoclustersinb)andmeasuredoverhumanandmousepreimplantation.
?2014Macmillan Publishers Limited. All rights reservedES c4 EpiSpellBlSpZyoCOCLETTERRESEARCH
ES ceBrllainHeartLungCGITSSExInontroICnRa
SplCBlb
1.0Methylation0.80.60.40.20.0Replicate 1Replicate 2
CGICpG (n = 74)CGICpG (n = 103)c1Sperm0500 bp1Sperm01Oocyte01Cleavage01Blastocyst01Somatic0500 bp1Cleavage01Blastocyst01Somatic0Chr 13 : 100,547,333–100,549,211Chr 14 : 122,804,906–122,806,794CLYBL 3′ UTR
Methylation0
0.5
1
n = 5,265
CpG islands
CpG
Single blastocyst CpG
GM5089CpG islandsAllele 1Allele 2
de
ES cBrellainHearLutngoSplBlCODMR resolutionHypoInt/varHyperMethylation0
0.5
1
n = 11
f
15913227MethylationResolve to Hypo1.00.80.60.40.20.0SpermPreImpES cellOocyteSpermPreImpEpiFigure3|TransientmaternalDMRstargetadivergentsetofCpGislandpromoters.a,Heatmapof5,265100-bptilesconsistentwithmaternallycontributedmonoallelicmethylation(Methods).Tilesarepartitioned
accordingtotheirhypomethylated(#0.2),intermediateorvariable(.0.2,,0.8)orhypermethylated($0.8)resolutioninEScells.Featureannotationsareincludedasseparateheatmaps.b,BoxplotsofCGIDMRmethylationfortwoindependentsingleblastocysts,withheterozygousSNP-linkedCpGshighlighted.Withineachreplicate,31and33CGIDMRscontainCpGsthatcouldbeassignedtoparentalloci.Ineachcase,DNAmethylationisrestrictedtoonlyoneofthetwoalleles.Untrackedreferstotheinferredmethylationstatusbeforehaplotypesegregation.Redlinesignifiesthemedian,boxesandwhiskersthe25thand75th,and2.5thand97.5thpercentiles.c,SingleCpGresolutionmethylationvaluesofaconservedpreimplantation-specificDMRinhumanandmouse.Humanblastocystdataincludesinformationfromthepooledsampleaswellasforasingleblastocystreplicate(purple)with
allele-trackedmethylationfor10CpGshighlightedinpinkandblue.AnnotatedCGIsareincludedbelow.d,ResolutionofCGIsthatbehaveasmaternalDMRsinhumanandmouse.e,HeatmapoforthologousICRsoverhumanandmousepreimplantationdevelopment.f,Orthologoushypomethylation-resolvingCGIDMRsinhumanandmouseshareonly13equivalentlyregulatedregions.WhenmethylationvaluesofmouseorhumanspecificDMRsaretrackedinthealternatespecies,theyareconstitutivelyhypomethylated,indicatingthat
oogenesistargetsequivalentgenomicfeaturesbutatspecies-specificsequences.PreImpreferstotheaveragevalueforcleavageandblastocystinhumanor8cellandICMinmouse.
andmouse,respectively,withsubstantiallymoreresolvingtohyper-methylationinhuman(Fig.3dandSupplementaryTable2).Notably,humanDMRsresolvingtohypomethylationaremorelikelytobeanno-tatedasCGIsinmousethanthoseresolvingtohypermethylation(ExtendedDataFig.6d,e).WerestrictedourcomparisontoDMRsthatshareCGIstatusandsomatichypomethylationinbothspeciesandfoundthatmaternallycontributed,preimplantation-specificDMRsarestrikinglydivergent,withonly7.5%foundinhumanequivalentlyregulatedinmouse.Noobvioustrenddistinguishedsharedfromspecies-specificsignatures,
thoughseveral,suchasthesomaticpromoterofDNMT1,indicatecon-servedregulatoryutility17(ExtendedDataFig.8a).ThedisparityofmaternalmethylationtargetingcontraststrueICRs,whicharegenerallyconserved18(Fig.3eandExtendedDataFig.8b).Moreover,hypomethylation-resolvingDMRsspecifictoonespeciesareconstitutivelyhypomethylatedintheother,suggestingthatthesesignaturesarefrequentlyanddiverselytar-geted(Fig.3f).
Wenextinvestigatedspecies-specificrepetitiveelements,incorporatingRNA-seqdatatointerpretDNAmethylation’sroleintheirregulation13.Inhumansperm,repetitiveelementsarefrequentlyincompletelymeth-ylatedorhypomethylated11.Long-terminal-repeat-containingelements(LTRs)areunexpectedlybimodal,withonlyafractionhypermethylatedandmostdisplayinggameticescapethatpersistsoverpreimplantation(ExtendedDataFig.9a).Longinterspersednuclearelements(LINEs)aregenerallyhighlymethylatedinsperm,demethylatedintheearlyembryo,withsomepartialremethylationinhumanEScellsandcompletehyper-methylationinsomaticcells(ExtendedDataFig.9b).Finally,shortinter-spersednuclearelements(SINEs),whichrepresentthemajorityofhumangenomerepetitivecontent,exhibitauniformglobalbehaviour(Sup-plementaryTable3).Althoughintermediatelymethylatedinsperm,theirdynamicsoverpreimplantationwereotherwisesimilartointer-genicsequencesingeneral(ExtendedDataFig.9c).Alternatively,diverseLTRandLINEsubfamiliesaredynamic,providingseveralexampleswhereDNAmethylationappearstoparticipateinspecificregulatorytransitions(SupplementaryTables4and5).
WefoundthatthebimodalityofLTRmethylationisexplainedbyspecies-specificendogenousretrovirus1(ERV1)familyelements.Alter-natively,theERVKfamilygenerallymaintainshighmethylationlevels,similartoobservationsinmouseandsuggestingconserved,constitutivetargeting4(Fig.4a,ExtendedDataFig.9d-g).Afterfertilization,expres-sionsharplytransitionedfromaMalR-dominatedearlycleavagestateresemblingtheoocytetoonecomposedofERV1andERVKelementsintheblastocystandEScells(Fig.4b).Unexpectedly,someERV1sseemtobeinducedlater,includingafterglobalremethylationinEScells,indi-catingatransitioninthespecificsubfamiliesthatareexpressed.Tran-scriptspresentearlyinpreimplantationaregenerallyfromgameticallyhypomethylatedERV1elementsthataredownregulatedbeforedenovomethylation(Fig.4c,d).Fortheseelements,methylationandexpressionarenegativelycorrelated,indicatingdiscriminatorytargetingforevenextremelyrelatedsequences(ExtendedDataFig.9h).Incontrast,theLTR7subfamilyishypermethylatedinsperm,rapidlydemethylated,andupregulatedintheblastocystandhumanEScells(Fig.4c,d).LTRsarerarelydynamicoutsideofthisearlyversuslatepreimplantationaxis:theyareeitheralreadyexpressedintheoocyteandsilencedlaterorareinducedfollowingdemethylationandremainexpressedinEScells.Nota-bly,thislatterdynamicincludesalimitednumberofrecentlyemergent,unrelatedsubfamilies19–21.
ComparedtoLTRs,LINEsmaintainhighermethylationlevelsandonlytheprimate-specificL1PAphylogenyisdynamicallyexpressed(Fig.5aandExtendedDataFig.9c).Astheonlyactivelytransposinglineageinhumans,L1PAsubfamiliesemergedasalinearphylogeny22.Wefoundthatthehuman-specificL1HSanditstwoclosestancestors,L1PA2andL1PA3,aredemethylatedearly,whereasolderelementsmaintainhigherembryonicmethylation(Fig.5b).Correspondingly,nearlyallembryonictranscriptioncouldbeattributedtothesethreeyoungestsubfamilies(Fig.5candExtendedDataFig.10a).Giventhehomologybetweensub-families,wesearchedforsequencecompositionchangesthatmayexplainthepreimplantation-specificescapeofyoungerelements.Wealigned59untranslatedregions(UTRs)offull-lengthL1PA7toL1HSandcomparedsequencesthatdemarcateddemethylatedL1PA3-descendedelementsfromconstitutivelytargetedancestors.Thelargestdiscretedifferencecorrespondstoan,130-bpdeletionfoundwithintheL1PA3lineageitselfthatseparatesolderelementsfromtheL1HSprogenitorL1PA3a23(Fig.5dandExtendedDataFig.10b,c).Intriguingly,thepres-enceorabsenceofthisregionisolatestwodisparatelyregulatedsubpo-pulationsandmarksthetransitiontoembryonicexpression(Fig.5e,f
00MONTH2014|VOL000|NATURE|3
MethylationICRs?2014Macmillan Publishers Limited. All rights reserved2 C4 C8 CICME6.5E7.5BrainHearLitverSpZyAll SN(612UP ()nt3ra1)ckAledlelAle 1lele 2All (SN606UP ()nt33ra)ckAledlelAle 1lele 2RESEARCHLETTER
a
Methylation1.0ERV10.80.60.40.20.0
p0p5SomSpClBlERVKERVLERVL-MalR
a
Total RNA contribution
8%
17r%
12%
15%
41%
50%9%
45%
45%8%
L1PA
Other LINE1LINE2Other
22%
27%
63V%
p0p5SomlBlp0p5Somp5SomSplBlSpSplBlp0CCCES cellES cell
Total RNA contribution
8%
ES cell4%
82%
7%
ES cell
3%
7%ERV1
ERVKERVL
ERVL-MalROther
Oocyte2 Cell8 CellBlastocystp10 ES cell
b
29%
56%
12%2%
Oocyte
2 Cell67%
20% 9%
b
Methylation1.00.80.60.40.2
Cleavage methylation20%
61%
3%
8 Cell
10%9%
78%
9%
L1HSL1PA2L1PA3L1PA4
L1PA5L1PA6L1PA7
500 bpL1HS
d
10.50CpGAlign
10.50CpGAlign
10.50CpGAlign
5′ UTR
Blastocystp10 ES cell
HeaExpression (Log2 FPKM)1.0Methylation0.80.60.40.20.0
lBlp0p1p5BraHineaLurtngSpCExpression (Log2 FPKM)ERV1: LTR12CERV1: HERV9-INTERVL-MalR: MLT1H2ERV1: LTR7ERVK: LTR5Hs50
ERV1: LTR7ERVK: LTR5HsERV1: LTR12CERVL-MalR: MLT1H2ERV1: HERV9-INTp0p10ClBlc
5
ES cell
BrLucd
0.0
p0p5nSpp1rtngClBlL1PA3aL1PA3b
ai–5
oZy2 C8 CO0
L1PA4
–5
ES cellES cell
ORF1L1PA alignmentConsensusNonconsensusGap
L1PA3a3b
Oocyte8 CellBlastocyst
2 C8 Cp0ZyFigure4|LTRsubfamilydynamicsaredividedintoearlyandlate
preimplantationphases.a,ViolinplotsforthefourLTRfamiliespresentinhumanoverearlydevelopmentandEScellderivation.b,PiechartsofLTRfamilyexpressioncalculatedasthenumberoffragmentspermillion(FPM)thataligntoelementswithinthefamily.c,MeanmethylationofnotableERV1,ERVKandMalRsubfamilies.ThreeERV1subfamiliesareincludedtorepresentdiscretedynamics:sharedgameteandearlyembryonichypomethylation(LTR12c),constitutivemethylation(HERV9-INT)andrapiddemethylation(LTR7).TheERVKsubfamilyLTR5Hsisalsodemethylated.d,Expressiondynamicsforthesamesubfamiliesinc.LTR12cisexpressedearlyand
downregulatedintheblastocyst.Alternatively,LTR7isexpressedthroughout,butupregulatedintheblastocystandmaintainedinEScells,whereitaccountsforthemajorityofERV1transcripts.LikeLTR7,LTR5HsisonlyintermediatelymethylatedduringEScellderivationandisembryonicallyinduced.
Alternatively,theERV1HERV9-INTremainsrepressed.MLT1H2istheprevailingMalRtranscribedintheoocyteandislostafterfertilization.Expressionisthefragmentspermillionthataligntosubfamilyelements,dividedbythekbannotatedasthesubfamilyinthegenome(FPKM).
p10OCpG frequency00.5
1
ClBloES cell
Cleavage methylatione
1.00.80.60.40.20.0
f
Expression(read count ×10–4)1.0
0.5
n De= let1eL1107dPA3n De=eL1 19tedPA03n Pres= en96tPren s= en38t60.0
L1PA3aL1PA3b
5′ UTR0 kb
7 kb
andExtendedDataFig.10d).ThisadaptationmayrepresentaspecificmomentintheevolutionaryprogressionoftheL1PAswhenemergingelementsevadedaseeminglysequence-directed,repressivemechanism.Whetheroldersubfamiliesretainexpressionortranspositionpotentialwithoutactivesilencing,orifthissignaturereflectsavestigial,hostgen-omeadaptationthatisnolongerrequired,remainstobeinvestigated.WepresentbasepairresolutionmapsofDNAmethylationasitisdynamicallyreconfiguredduringhumanearlydevelopment.Thesedataidentifyasetoftransient,maternallycontributedmethylationatCGIpromoters,theresolutionofwhichsuggestindependentmodesofacqui-sition:malegermlinespecificprotectionagainstmethylationordenovotargetingintheoocyteofotherwisecanonicallyunmethylatedCGIs24.BotharecommonandshowpoorconservationcomparedtoclassicICRs,indicatingthatshort-lived,parent-specificsignaturesarelessevolutio-narilyconstrainedthanthosepersistingafterimplantation.Wefindthatrepetitiveelementregulationisnotablydiverseinhuman,moresothaninmouse,withgameticallyhypomethylatedLTRsubfamiliespre-sentintheoocyteandearlyembryoandotherssharplydemethylatedandinducedembryonically.ForLINEs,thestepwisephylogenywithintheprimate-specificL1PAlineagepinpointsaspecific,adaptivetrans-ition.AsL1HSelementsremaintranspositionallyactive,includingsomat-icallyinnumerouscancers,thetargetingofepigeneticsilencingmachineryduringpreimplantationmayberelevantinidentifyingtherootcauseoftheiraberrantinductionlater25.Understandingtheregulatoryprinci-plesinherenttotheearlyembryowillimprovecontinuingeffortstoeval-uatecomplextraitswithunclearmodesofepigeneticinheritance.Futureworktocharacterizethemechanismsthatimposethesediverselytar-getedembryonicmethylationpatternswillilluminatetheircontributiontonormalhumandevelopmentanddisease.
4|NATURE|VOL000|00MONTH2014
Figure5|EmergentL1PAsubfamiliesescapeDNAmethylation-associatedrepressionduringpreimplantation.a,PiechartsoftheLINEexpressiondividedintotheL1PAsubfamily,otherLINE1andLINE2subfamilies.Totalexpressioniscalculatedasthenumberoffragmentspermillion(FPM)thataligntofamilyelements.b,MeanmethylationvaluesforthemostrecentL1PAsubfamilies.Incleavageembryos,L1HSthroughL1PA3aredemethylatedandmaintaintheselevelsthroughtheblastocyst.c,ExpressiondynamicsforthesamesubfamiliesincoverpreimplantationandinEScells.ThethreeyoungestL1PAsubfamiliesareinducedbythe8cellstage.Expressionisthenumberoffragmentspermillionthataligntosubfamilyelements,dividedthekbannotatedasthesubfamilyinthegenome(FPKM).Thecoloursincarethesameasinb.d,Compositeplotofcleavagestagemethylationvaluesacrossaligned59UTRsinL1PAsubfamilies.ThecompositeforL1PA3issplitbythepresence(red)orabsence(blue)ofa,130-bpsequencethatdistinguishesL1PA3bfromL1PA3aanddemarcatesmethylationvaluesbetweenolderandnewersubfamilies(highlightedinpink).Multiplesequencealignmentforeachsubfamilytotheassembledconsensusisbeloweachcomposite,withbluecorrespondingtoconservation,blacktodivergence,andwhitetogapsordeletions.ThexaxisrepresentspositionalongtheL1HS59UTRandaportionofORF1.CpGFrequencydescribesperCpGconservationwithinsingleelementstotheconsensus.TwooldersequencesspecifictoL1PA7are
highlightedingrey.e,BoxplotofL1PAmethylationincleavageembryos,sortedbythepresenceofthe,130bpsequenceforallelementsandL1PA3
specifically.Preimplantationmethylationishigherforelementsthatcontainthisinsert.Boldlinesignifiesthemedian,boxesandwhiskersthe25thand75th,and2.5thand97.5thpercentiles.f,Expressioncompositeoffull-lengthinsertdeletedL1PA3aandinsertcontainingL1PA3bsubfamiliesinoocyte,8cellandblastocyststageembryos.TranscriptionalinductionisnotapparentuntilafterfertilizationandisspecifictoL1PA3a.Readcountisthereadcoveragenormalizedbytotalreads(Methods).
OnlineContentMethods,alongwithanyadditionalExtendedDatadisplayitemsandSourceData,areavailableintheonlineversionofthepaper;referencesuniquetothesesectionsappearonlyintheonlinepaper.Received16September2013;accepted16June2014.Publishedonline23July2014.1.2.Suzuki,M.M.&Bird,A.DNAmethylationlandscapes:provocativeinsightsfromepigenomics.NatureRev.Genet.9,465–476(2008).Stadler,M.B.etal.DNA-bindingfactorsshapethemousemethylomeatdistalregulatoryregions.Nature480,490–495(2011).?2014Macmillan Publishers Limited. All rights reserved
相关推荐：
loading