MolecularSystemsBiology9;Articlenumber632;doi:10.1038/msb.2012.65 Citation:MolecularSystemsBiology9:632 &2013EMBOandMacmillanPublishersLimited Allrightsreserved1744-4292/13 www.molecularsystemsbiology.com An Oct4-Sall4-Nanog network controls developmental progression in the pre-implantation mouse embryo MengHowTan1,4,7,KinFaiAu2,7,DeniseELeong1,KiraFoygel1,5,WingHWong2,3,*andMyleneWMYao1,6,* 1 DepartmentofObstetricsandGynecology,StanfordUniversitySchoolofMedicine,Stanford,CA,USA,2 DepartmentofStatistics,SchoolofHumanitiesand Sciences,StanfordUniversity,Stanford,CA,USAand3 DepartmentofHealthResearchandPolicy,StanfordUniversitySchoolofMedicine,Stanford,CA,USA 4Presentaddress:DepartmentofGenetics,StanfordUniversitySchoolofMedicine,Stanford,CA94305,USA 5Presentaddress:DepartmentofRadiology,MolecularImagingProgram,StanfordUniversitySchoolofMedicine,Stanford,CA94305,USA 6Presentaddress:UnivfyInc.,LosAltos,CA94022,USA 7Theseauthorscontributedequallytothiswork * Correspondingauthors.WHWong,DepartmentofStatistics,SchoolofHumanitiesandSciences,StanfordUniversity,390SerraMall,Stanford,CA94305,USA. Tel.:+16507252915;Fax:+16507258977;E-mail:[email protected],DepartmentofObstetricsandGynecology,StanfordUniversitySchoolof Medicine,Stanford,CA94305,USA.Tel.:+16507998003;Fax:+16509611377;E-mail:[email protected] Received28.7.11;accepted30.11.12 Landmark events occur in a coordinated manner during pre-implantation development of the mammalian embryo, yet the regulatory network that orchestrates these events remains largely unknown.Here, wepresent the first systematic investigation of the network in pre-implantation mouseembryosusingmorpholino-mediated gene knockdowns ofkeyembryonicstemcell(ESC) factors followed by detailed transcriptome analysis of pooled embryos, single embryos, and individualblastomeres.WedelineatedtheregulonsofOct4,Sall4,andNanogandidentifiedasetof metabolism- and transport-related genes that were controlled by these transcription factors in embryosbutnotinESCs.Strikingly,theknockdownembryosarrestedatarangeofdevelopmental stages.WeprovidedevidencethattheDNAmethyltransferaseDnmt3bhasaroleindeterminingthe extenttowhichaknockdownembryocandevelop.Wefurthershowedthatthefeed-forwardloop comprisingDnmt3b,thepluripotencyfactors,andthemiR-290-295clusterexemplifiesanetwork motifthatbuffersembryosagainstgeneexpressionnoise.OurfindingsindicatethatOct4,Sall4,and Nanogformarobustandintegratednetworktogovernmammalianpre-implantationdevelopment. MolecularSystemsBiology9:632;publishedonline8January2013;doi:10.1038/msb.2012.65 SubjectCategories:simulationanddataanalysis;development Keywords: pluripotencyfactors;pre-implantationdevelopment;transcriptionalnetworks Introduction indefinitely in culture. The transcription factor Oct4, also known as Pou5f1, has been shown to have a key role in the Manycriticalcellulareventsoccurwithinafewdaysduringthe maintenance ofan undifferentiatedstate in bothhuman and pre-implantationdevelopmentofamammalianembryo.First, mouseESCs.Chromatinimmunoprecipitation(chIP)assaysin the paternal and maternal genomes are reprogrammed to a ESCsrevealthatOct4bindstomorethanathousandlociinthe totipotent state. Second, the embryonic genome is activated genome, suggesting that it may regulate many genes as a afteraperiodofrelianceonmaternalfactors.Third,polarityis guardian of pluripotency (Boyer et al, 2005; Loh et al, 2006; established when the inner and outer cells of a multicell Chenetal,2008a).Inaddition,otherkeytranscriptionfactors embryo are biased toward different cell fates. Finally, two that control the self-renewing and pluripotent properties of importantcellfatedecisionsoccurbeforeimplantation.Inthe ESCsincludeSall4andNanog.Previousstudiesindicatethat first cell fate decision, the outside cells develop into extra- Oct4,Sall4,andNanogshareaclosefunctional relationship. embryonic trophectoderm, while cells that are inside the embryoretainpluripotencyandformtheinnercellmass(ICM). Co-immunoprecipitationexperimentsshowthattheyinteract Inthesecondcellfatedecision,someoftheICMcellsfurther with one another in vivo (Wu et al, 2006; Liang et al, 2008; differentiateintotheprimitiveendoderm(Zernicka-Goetzetal, Yangetal,2008).Furthermore,thethreetranscriptionfactors 2009). Given their complexities, all of these early cellular co-occupy the promoters of many target genes (Boyer et al, eventshavetobecarefullyorchestratedbya generegulatory 2005; Wu et al, 2006; Yang et al, 2008), indicating that they networktoensuretheproperdevelopmentoftheembryo. formanintegratednetworkinESCs. Thegeneregulatorynetworkofembryonicstemcells(ESCs) Compared with ESCs, much less is known about the gene hasbeenextensivelyinvestigatedinrecentyears.ESCs,which regulatorynetworkinearlymammalianembryoslargelydue are derived from the ICM of mammalian embryos, are to technical challenges. In particular, limited number of pluripotentlikecellsintheICMbuttheycanalsoself-renew embryos or cells has so far precluded extensive studies of &2013EMBOandMacmillanPublishersLimited MolecularSystemsBiology2013 1 AnOct4-Sall4-Nanognetwork MHTanetal any regulatory network operating during pre-implantation multiple one-cell embryos (Supplementary Figure S1). Hence, development. For example, chIP experiments to map tran- wechosetofocusonthethreetranscriptionfactorsinthisstudy. scriptionfactorbindingsitesinpre-implantationembryosare TodetermineifOct4,Sall4,orNanog playedanimportant currentlyunfeasible.Fortunately,advancesingeneexpression roleduringpre-implantationdevelopmentofthemammalian profilingarenowallowingstudiesintosmallnumberofcellsor embryo, we depleted each of the proteins by injecting even single cells. Several laboratories used microarrays to translation-blocking morpholinos (MOs) (Supplementary profile the transcriptome during pre-implantation develop- Figure S2), which targeted both maternal and embryonic ment of mouse embryos (Hamatani et al, 2004; Wang et al, transcripts, into one-cell fertilized zygotes and then cultured 2004;Zengetal,2004;Xieetal,2010).Expressionanalysisof theinjectedembryostogetherwithcontrolembryosinserum 48 genes in individual cells from the one-cell zygote to the supplementedHumanTubalFluid(SAGEIn-VitroFertilization blastocyst identified markers of outer and inner cells (Guo Inc.). For each MO injection, we confirmed that the protein etal,2010).Inaddition,next-generationsequencingtechnol- wasabsentbyimmunocytochemistry(Figure1A).After4days ogy was used to investigate the pre-implantation transcrip- ofinvitroculture,97.7%ofuninjectedembryosand85.1%of tome in single mouse blastomeres (Tang et al, 2011). While control MO-injected embryos developed at least until the thesepreviousstudieswereimportantforunderstandingearly morula stage, but surprisingly, o25% of embryos that were mammaliandevelopment,theyreliedprimarilyontranscrip- depletedofanyofthethreepluripotencyfactorscompacted.In tome profiling of wild-type embryos and did not perform addition, 40–60% of embryos injected with a MO targeting Oct4,Sall4,orNanogarrestedbythe4-cellstageafter4daysof perturbation experiments to dissect interactions between the differentgenes. in vitro culture, compared with o10% for the uninjected or Here,weshowthatthekeypluripotencyfactorsinESCsalso control MO-injected embryos (Po0.05, Student’s t-test). To verifythatourfindingswerenotduetooff-targeteffects,we form critical nodes in the regulatory network governing pre- performed rescue experiments for each pluripotency factor implantation development of the mammalian embryo. We (Figure1BandC).Thepercentageofembryosthatarrestedby used gene-specific knockdowns followed by genome-wide the4-cellstagedecreasedby25.3%whenweco-injectedOct4 microarray analysis as well as extensive single-embryo and MOwithOct4mRNA,whilethepercentagedecreaseforSall4 single-cellexpressionprofilingto deconstruct theOct4-Sall4- Nanog transcriptional network in pre-implantation embryos. orNanogwas18.3or24.7%respectively(Po0.05,Student’s t-test).Theseandotheradditionalrescuedata(Supplementary Our study yielded novel and unexpected insights into the ResultsandSupplementaryFiguresS3andS4)indicatedthat regulatory mechanisms that control the earliest stages of theMOswerespecificfortheirintendedtargetsandthatour mammaliandevelopment. observedphenotypeswereadirectconsequenceofknocking down the pluripotency factors. Taken together, our experi- Results ments indicate that Oct4, Sall4, and Nanog are essential for pre-implantation development. Interestingly, these results Oct4, Sall4, and Nanog are essential for pre- contradicted published knockout mouse models, where implantation development of the mouse embryo embryosthatwerehomozygousnullforOct4,Sall4,orNanog Since a network of transcription factors has been shown to appearedtodevelopinvivoatleastuntiltheblastocyststage, regulate pluripotency in ESCs (Zhou et al, 2007; Chen et al, althoughtheywerestillembryoniclethal(Nicholsetal,1998; 2008a), we hypothesize that some of these factors may also Mitsuietal,2003;Sakaki-Yumotoetal,2006).Alikelyreason haveanintegralroleincontrollingdevelopmentalprogression for the apparent discrepancy is that maternal transcripts are of the pre-implantation embryo. We previously found that noteliminatedinknockoutmousemodels(seeSupplementary Oct4 may play an important role at the maternal-zygotic informationformorediscussion). transition(Foygeletal,2008).Here,wesoughttoundertakea morecomprehensivestudyofthefunctionsofkeyESCfactors in pre-implantation embryos. We first measured the abun- Global transcriptome analysis identifies genes dance of several pluripotency factors in individual one-cell dependent on Oct4, Sall4, and Nanog zygotesusing amicrofluidicreal-time PCRsystem known as Biomark (Fluidigm). While the Kru¨ppel-like factor Klf5 was Tounderstandthemolecularmechanismsthatmightcausethe absent,wedetectedthetranscriptsofOct4,Sall4,andNanogin embryostoarrestduringknockdownofapluripotencyfactor, Figure1 MO-mediatedknockdownofOct4,Sall4,orNanoginpre-implantationmouseembryoswaseffectiveandspecificandresultedindevelopmentalarrest. (A)Confocalmicroscopyimagesofembryosstainedwitha-Oct4(leftpanel),a-Sall4(middlepanel),ora-Nanog(rightpanel)antibodies.Theembryoswerestainedfor theprotein-of-interest2daysaftermicroinjection.Oct4andNanogproteinswerecompletelyundetectablebyimmunostainingin100%oftheembryosinjectedwiththe correspondingMO,whiletheuninjectedembryosandthecontrolMO-injectedembryoswerestronglystained.Inaddition,whileasmallamountofSall4proteinwas detectedinembryosinjectedwithaSall4MO,theproteinlevelintheseembryoswassignificantlylessthanthatinallthecontrolembryos.(Scalebar¼45mm). (B)Representativeimagesofpre-implantationembryosundervariousexperimentalconditions.Foreverysetofrescueexperiments,therewerefourconditions:(i)no injection,(ii)injectionwithacontrolMO,(iii)injectionwiththeexperimentalMOalone,and(iv)co-injectionwiththeexperimentalMOandthecorrespondingmRNAthat wasinvitrotranscribed.(C)Percentagesofembryosthatarrestedbythe4-cellstageduring4daysofinvitroculture.Leftpanel(greenbars):comparedwithembryos injectedwiththeOct4MOonly,weobservedthatco-injecting0.6mMOct4MOwith0.20–39ng/mlOct4mRNAresultedinasmallerpercentageofembryosarrestingby the4-cellstage(9independentexperiments)(Po0.05,Student’st-test).Middlepanel(bluebars):comparedwithembryosinjectedwiththeSall4MOalone,wefound thatco-injecting0.6mMSall4MOwith7–146ng/mlSall4mRNAloweredtherateofdevelopmentalarrestbythe4-cellstage(8independentexperiments)(Po0.05, Student’st-test).Rightpanel(redbars):comparedwithembryosinjectedwiththeNanogMOalone,wediscoveredthatco-injecting0.6mMNanogMOwith5ng/ml NanogmRNAalsoresultedinpartialrescueofthephenotype(7independentexperiments)(Po0.05,Student’st-test). 2 MolecularSystemsBiology2013 &2013EMBOandMacmillanPublishersLimited AnOct4-Sall4-Nanognetwork MHTanetal we investigated transcriptome changes in pooled embryos theuninjectedembryosareatthemulticellstage.Comparison using Affymetrix microarrays at B20 and 44h after MO oftheearliertimepointwiththelatertimepointrevealedthat injection (Supplementary Files S1–S5). At 20h, uninjected depletionofapluripotencyfactorreducedtheextentofzygotic controlembryosareatthe2-cellstage,whileat44h,almostall genome activation (see Supplementary information and A DNA Oct4 Overlay DNA Sall4 Overlay DNA Nanog Overlay d d d e e e ct ct ct e e e nj nj nj ni ni ni U U U O O O M M M ol ol ol ntr ntr ntr o o o C C C Oct4 MO Sall4 MO Nanog MO B Uninjected Control MO Oct4 MO Oct4 MO+Oct4 mRNA Uninjected Control MO Sall4 MO Sall4 MO+Sall4 mRNA Uninjected Control MO Nanog MO Nanog MO+Nanog mRNA C 100 100 100 90 90 90 e g 80 80 80 a st 70 70 70 cell 60 60 60 y 4- 50 50 50 st b 40 40 40 e 30 30 30 Arr 20 20 20 % 10 10 10 0 0 0 d O O +A d O O +A d O O +A Uninjecte Control M Oct4 M Oct4 MOOct4 mRN Uninjecte Control M Sall4 M Sall4 MOSall4 mRN Uninjecte Control M Nanog M Nanog MOanog mRN N &2013EMBOandMacmillanPublishersLimited MolecularSystemsBiology2013 3 AnOct4-Sall4-Nanognetwork MHTanetal Supplementary Figure S5). We also found that MO-mediated that were downregulated during Sall4 knockdown were also knockdownofOct4hadthemostprominenteffect,compared downregulated during Oct4 knockdown and 133 out of 245 withknockdownsofSall4andNanog.Atthe20-htimepoint, genes that were downregulated during Nanog knockdown a total of 634 genes were differentially expressed in Oct4 werealsodownregulatedduringOct4knockdown,indicating MO-injected embryos, while o10 genes were differentially that the three transcription factors shared a close functional expressedinSall4MO-andNanogMO-injectedembryos(false relationshipduringpre-implantationembryonicdevelopment discovery rate (FDR)o0.1) (Figure 2A) (see Materials and (Po10(cid:2)100,chi-squaretest). methods fordetails). However, at 44hpost injection, a clear We asked which of the differentially expressed genes effect on global expression levels was observed for all three are directly controlled by Oct4, Sall4, or Nanog. Since transcriptionfactors(Figure2B).Notably,505outof799genes chIP experiments could not be readily performed on A B Sall4 Nanog Sall4 Nanog Sall4 Nanog Sall4 Nanog 2 0 0 1 0 4 286 8 104 200 20 244 0 0 54 53 0 0 0 1 451 79 230 130 456 177 1476 1125 Oct4 Oct4 Oct4 Oct4 Downregulated Upregulated Downregulated Upregulated C D 2 2 Vesicle-mediated transport 148 (76) Endocytosis 78 (38) 1 1 Lysosomal transport 13 (2) Exocytosis 55 (25) 0 0 5 10 15 Phosphate metabolism 34 (15) RNA localization 16 (5) EDownregulated (embryos and ESCs): Localization 20 (7) 2 2 Lipid metabolism 116 (80) Coenzyme metabolism 18 (7) 1 1 Carbohydrate metabolism 90 (66) 0 0 Nitrogen compound metabolism 12 (5) 5 10 Mitochondrion organization 7 (2) Neurotransmitter secretion 35 (21) Upregulated (embryos): 2 2 Oxygen species metabolism 11 (4) Peroxisomal transport 6 (2) 1 1 0 2 4 6 8 10 12 14 0 0 –log10 (P-value) 5 10 Ion transport 62 (35) Upregulated (ESCs): Cation transport 52 (29) 2 2 Carbohydrate metabolism 74 (47) 1 1 Meiosis 27 (13) Vesicle-mediated transport 81 (54) 0 0 Phosphate metabolism 20 (11) 5 10 0 2 4 6 8 10 12 14 Core motif: –log (P-value) 10 Figure2 Microarrayanalysisofpooledembryosidentifiedtheprotein-codingregulonsofOct4,Sall4,andNanog.(A)Venndiagramshowingthenumberofgenesthat weredifferentiallyexpressed20hafterMOinjectiontoknockdownoneofthethreepluripotencyfactors.Atthisearliertimepoint,onlyknockdownofOct4exerteda strongeffectonglobalgeneexpressionlevels.(B)Venndiagramshowingthenumberofgenesthatwasdifferentiallyexpressed44hafterMOinjection.Atthislater timepoint,knockdownofanyofthethreetranscriptionfactorscausedhundredsofgenestobesignificantlydownregulatedorupregulated(FDRo0.1).(C)GOanalysis ofgenesthatweredifferentiallyexpressedduringknockdownofOct4,Sall4,orNanoginpre-implantationembryosbutshowednoevidenceofbeingcontrolledbythe pluripotencyfactorsinESCs.Thesegenesmaybenoveldirecttargetsofthethreetranscriptionfactorsintheembryosortheymayalsobeindirectlyregulated.Upper panel:GOanalysisoftheembryo-specificgenesthatweredownregulated.Lowerpanel:GOanalysisoftheembryo-specificgenesthatwereupregulated.Allthe biologicalprocessesthatareuniquetotheembryo(Po0.01)areplotted.Theactualnumbersofgenesforeachprocess(withtheexpectednumbersgiveninbrackets) areshownattheendofthebars.Manyofthefunctionalcategoriesarerelatedtoeithertransport(inred)ormetabolism(inblue).(D)A15-bpmotifwithinOct4-bound regulatoryregionsofgenesthatwerealsodifferentiallyexpressedduringknockdownofOct4inpre-implantationembryosorESCs.(E)A7-bpcoremotifwasfoundtobe enrichedinNanog-boundregulatoryregionsofgenesthatwerealsodifferentiallyexpressedduringknockdownofNanoginpre-implantationembryosorESCs.Thecore motifwasextendedatthe50endforgenesthatweredownregulatedduringNanogknockdownanditwasextendedatthe30endforgenesthatwereupregulatedduring Nanogknockdown. 4 MolecularSystemsBiology2013 &2013EMBOandMacmillanPublishersLimited AnOct4-Sall4-Nanognetwork MHTanetal pre-implantation embryos, we compared our microarray majorrolesinpre-implantationdevelopmentalprocesses.We resultswithpubliclyavailablechIPdatasetsinESCsinstead thereforequantifiedmiRNAexpressionfromthe1-celltothe (Boyeretal,2005;Lohetal,2006;Limetal,2008;Yangetal, blastocyst stage in uninjected mouse embryos and detected 2008; Chen et al, 2008a; Ouyang et al, 2009). 20.4% of the 297 distinct miRNAs in at least one of these stages downregulated genes and 18.8% of the upregulated genes (SupplementaryFileS10).K-meansclusteringfurtherrevealed have promoters or enhancers that are bound by the corre- that many of the miRNAs were temporally regulated over sponding transcription factor (Supplementary Figure S6). development(Figure3A). GeneOntology(GO)analysisusingthePantherclassification WehypothesizethatsomeofthehighlyexpressedmiRNAs system(Thomaset al,2003)furtherrevealed thatthe down- are integral components of the Oct4-Sall4-Nanog gene regulated genes are involved in the establishment or main- regulatory network in the pre-implantation embryo. To tenance of chromatin architecture, while the upregulated determine which miRNAs are regulated by Oct4, Sall4, or genes are involved in stress response and differentiation or Nanog,weknockeddowneachtranscriptionfactorseparately developmental processes. Manyof the directtargetsarealso andmeasuredmiRNAexpressionlevels46haftermicroinjec- involvedinthecellcycleorarecomponentsofkeysignaling tionwhentheknockdownphenotypeswerereadilyobserva- pathways(SupplementaryFigureS7;SupplementaryFileS6). ble. More miRNAs were downregulated than upregulated SinceESCsdonotexistinanaturaldevelopmentalcontext, (Figure 3B; Supplementary File S11), indicating that Oct4, thepluripotencyfactorsmaycontroladditionalgenesinpre- Sall4, and Nanog served primarily as positive regulators of implantation embryos but not in ESCs. To determine the miRNAexpressioninpre-implantationembryos.Inaddition, embryo-specific genes, we subtracted away from our micro- mostofthedownregulatedmiRNAsweresignificant(Po0.05) arrayresultsgenesthatwerepresentintheESCchIPdatasets inatleasttwodifferentknockdownconditions,suggestingthat as well as genes that responded to Oct4, Sall4, or Nanog the expression of many miRNAs was under combinatorial depletion in ESCs (Supplementary Figure S6). GO analysis control by the transcription factors(Figure 3B). In total, 120 revealed that many of these embryo-specific genes were distinct miRNAs were differentially expressed. 40 of them associated with metabolism- and transport-related functions (33.3%) have promoters or enhancers that are occupied by (Figure2C;SupplementaryFileS7),whichreflectedtheunique Oct4,Sall4,orNanoginESCs(Limetal,2008;Marsonetal, embryonicconditionsthatareabsentfromESCs. 2008). The remaining 80 miRNA genes may be regulated Toidentifycis-actingelementsthatmediatethedirecttargeting indirectly by the three transcription factors or they may of Oct4, Sall4, or Nanog to the genome in pre-implantation represent novel direct targets of Oct4, Sall4, or Nanog in the embryos,wesearchedtheboundlociforknownmotifsandalso pre-implantationembryo. performed de novo motif discovery using CisGenome (Ji et al, Next,wesoughttounderstandhowthepluripotencyfactors 2008). The Oct4-Sox2 consensus sequence was present in the and their protein-coding and non-coding regulons are inter- promoterorenhancerregionsof83downregulatedgenes(outof connected. 61.1% of the protein-coding genes (2646 out of 284 genes) and 35 upregulated genes (out of 153 genes) 4329 genes) thatweredifferentiallyexpressedduringknock- (Figure 2D; Supplementary File S8). In addition, we obtained downofOct4,Sall4,orNanogwerepredictedbythemiRanda withhighconfidenceacoremotif,TGACCTT,thatwasenrichedin algorithm (www.microrna.org) (Betel et al, 2008) to be the Nanog-bound promoter or enhancer regions of 25 down- targetedbyatleastonemiRNAthatwasalsoregulateddirectly regulatedgenes(outof40genes)and31upregulatedgenes(outof orindirectlybythethreetranscriptionfactors(Supplementary 46 genes) (Supplementary File S9). Notably, the motif for FileS12).Similarresultswereobtainedwhenwerestrictedour downregulatedgenesisextendedatthe50endbytwobasepairs, target predictions to protein-coding genes and miRNAs that whilethemotifforupregulatedgenesisextendedatthe30 end werepresentintheESCchIPdatasets(62.7%;622outof992 by one to three basepairs (Figure 2E). A de novo search for genes) (Supplementary File S13). In comparison, when we cis-regulatory elements using a different program, SCOPE, also randomlyselected10000distinctsetsofprotein-codinggenes returnedthe same coremotif (Carlsonet al, 2007; Chakravarty andmiRNAs,wefoundthatonly52.8±0.7%oftheprotein- etal,2007;SupplementaryFigureS8).Takentogether,ourresults codinggeneswerepredictedtobemiRNAtargets.Hence,our suggestthatdifferentcofactorsmaymediateNanog’sfunctionasa analysis indicates that Oct4, Sall4, and Nanog and their transcriptionalactivatororrepressor.Furthermore,thecoremotif protein-coding and non-coding regulons are more intercon- is similar to previously identified motifs for the retinoic acid nectedthanrandom(Po10(cid:2)6,Z-test).Toincreaseconfidence receptorRaraaswellastheestrogen-relatedreceptorsEsrraand inthemiRNA-targetpredictions,wefurtherdeterminedwhich Esrrb(SupplementaryFigureS9),whichsuggeststhatNanogmay ofthemiRNAsoutputtedbymiRandawerealsopredictedbyat co-regulatemanygenestogetherwithRaraandEsrra/Esrrb. least one other algorithm (see Materials and methods and Supplementary Files S14 and S15). An example of a subnet- workisillustratedinFigure3C. MicroRNAs form an integral component of the Oct4-Sall4-Nanog regulatory network in pre-implantation embryos Knockdown of Oct4, Sall4, or Nanog causes gene expression changes even in embryos that appear MicroRNAs (miRNAs) are an important class of post-tran- to develop normally scriptional regulators that have diverse and critical roles duringanimaldevelopmentandregeneration(YangandWu, Asourmicroarraydataweregeneratedfrompoolsofembryos 2007;StefaniandSlack,2008).Itislikelythattheyalsohave and the ensemble averages may have masked interesting &2013EMBOandMacmillanPublishersLimited MolecularSystemsBiology2013 5 AnOct4-Sall4-Nanognetwork MHTanetal A B 1-cell2-cell4-cellMulticellBlastocyst I Sall4 Nanog Sall4 Nanog II 9 3 15 14 3 5 22 0 III 10 8 2 2 IV 32 3 Oct4 Oct4 V Downregulated Upregulated Repression Induction –1 0 1 C Klf5 Klf2 miR-20a Pou2f3 miR-92a Pou5f1 miR-191 Dnmt3a miR-709 Dnmt3b miR-720 Jarid2 miR-805 Smarcd1 Oct4 miR-291a-3p Sall4 Fgfr2 Nanog miR-292-3p Fzd4 miR-293 Fzd7 miR-294 Spry4 miR-295 Dazl Sf3b2 Lhx2 Sf1 Meis2 Neurog2 Atg2a Atg7 Figure3 Oct4,Sall4,andNanogcontroltheexpressionofnumerousdevelopmentallyregulatedmiRNAsinthepre-implantationembryo.(A)K-meansclustered expressionprofilesfor235miRNAsthatwereclearlyexpressedinuninjected(normal)pre-implantationembryos.Genesareinrows,withtemporalprogressionfromleft toright.Redindicateshighexpressionlevels,whilegreenindicateslowexpressionlevels.ThemajorityofthesemiRNAsshowageneralincreaseinexpressionlevels (clusterIandclusterV)fromthe1-celltotheblastocyststage.(B)VenndiagramshowingthenumberofmiRNAsthatweredifferentiallyexpressedduringknockdownof Oct4,Sall4,orNanog(Po0.05).MiRNAsinthemiR-290-295clusterweredifferentiallyexpressedunderallthreeknockdownconditions.(C)Anexampleofan integratednetworkcontrolledbyOct4,Sall4,andNanoginpre-implantationembryos.Theprotein-codinggenesaredepictedinvariouscolorstosignifydifferent functionalcategories.Darkbrown:pluripotencytranscriptionfactors;lightblue:epigeneticregulators;lightbrown:signalingmolecules;pink:differentiationfactors; purple:embryo-specificfactors.ThelistedmiRNAsarewithinthetop20mostabundantmiRNAsintheembryos.X-Ymeans‘XregulatesY.’Forthearrowsemanating fromthecenter,asolidarrowindicatesconfirmeddirectregulation,whileadottedarrowindicateseitherindirectregulationorpotentialdirectregulation. embryo-to-embryo variability or important expression (Fluidigm). We focused our single embryo analysis on 48 dynamics at the single embryo level, we sought to monitor protein-coding genes, 43 of which were found from our gene expression levels in individual embryos using BioMark microarray experiments to be differentially expressed during 6 MolecularSystemsBiology2013 &2013EMBOandMacmillanPublishersLimited AnOct4-Sall4-Nanognetwork MHTanetal Oct4,Sall4,orNanogknockdown.The43geneswerechosen Hence, the expression of these seven genes (highlighted in because they perform essential functions in ESCs, are key green in Figure 4C and Table I) is low in 2-cell and 3-cell epigenetic regulators or signaling molecules, have important knockdownembryos,highin4-cellknockdownembryos,and roles during differentiation events, or may have embryo- evenhigherinmulticellknockdownembryos. specific functions. The five remaining genes had unchanged Since promoter DNA methylation contributes to ESC gene expressionlevelsandservedasnegativecontrols. regulation and methylation-deficient ESCs are restricted in In our first set of analyses, we compared between the theirdevelopmental potential (Fouseet al, 2008),we decided various experimental conditions, namely (1) no injection, to investigate the DNA methyltransferase Dnmt3b in greater (2)injectionwithacontrolMO,and(3)injectionwithaMO depth.KnockdownofDnmt3busingatranslation-blockingMO targeting Oct4, Sall4, or Nanog. For each condition, we revealedamoreseverephenotypethanthatforanyofthethree collected between 80 and 120 embryos at 44h after MO pluripotency factors. Specifically, 66.3±21.6% of embryos injection. When all the embryos were considered in the injectedwiththeDnmt3bMOarrestedbythe2-cellstage,even analysis, we found that our BioMark and microarray results though the percentages for Oct4, Sall4,and Nanog were only agreed reasonably well with each other (Supplementary between5.5and19.0%(Po0.001,Student’st-test)(Figure5A). Information, Supplementary Figure S10, and Supplementary Rescue experiments with in vitro transcribed Dnmt3b mRNA File S16). However, since depletion of a pluripotency factor confirmed the specificity of the Dnmt3b MO (Figure 5B–D). caused embryos to arrest at different developmental stages These results suggest that Dnmt3b is essential for pre- (Supplementary Figure S11), some of the transcript changes implantation development and that it may have a role in observed between experimental conditions could be due to determining the extenttowhich an embryo depleted ofOct4, expression changes between the various stages. Hence, we Sall4,orNanogcandevelop.Inaddition,althoughDnmt3b(cid:2)/(cid:2) next compared 4-cell and multicell control embryos with embryos had been reported to develop normally before knockdownembryosthatwerealsoatthe4-cellandmulticell E9.5 (Okano et al, 1999b; Ueda et al, 2006), it is likely that stages (red boxes in Figure 4A). Intriguingly, we found that maternalDnmt3b,whichwedetectedin1-cellmousezygotes manygeneswerestillsignificantlydownregulatedorupregu- (Supplementary Figure S1), may be driving the earliest lated (Figure 4B). In particular, at least four genes, namely developmentalstagesintheknockoutmousemodel. Jarid2,Aqp3,Atg2a,andSf3b2,weredifferentiallyexpressed inallthreeknockdownconditions. Hence, although someof the knockdown embryos appeared to be morphologically A coherent feed-forward loop controls the normal,numerousmolecularchangeshadalreadyoccurredin expression of Dnmt3b them. In ESCs, chIP experiments showed that the pluripotency factors can control the transcription of Dnmt3b directly. Interestingly,thefactorscanalsoregulateDnmt3bexpression Expression of Dnmt3b distinguishes early- indirectly via the ESC-specific miRNAs in the miR-290-295 arrested embryos from late-arrested embryos cluster. Specifically, Oct4 and Nanog directly control the during Oct4, Sall4, or Nanog knockdown expressionofthesemiRNAs,whichshowsageneralupward Strikingly, we observed that, while uninjected and control- trend from the 1-cell to the blastocyst stage (Supplementary injected embryos were able to develop in a stereotypical Figure S12). The miRNAs then target the 30UTR of the manner, embryos that lacked any of the three pluripotency retinoblastoma-like 2 (Rbl2) mRNA, whose protein product factorsarrestedoverarangeofdevelopmentalstagesinstead functions as a transcriptional repressor of Dnmt3b (Benetti (Supplementary Figure S11). At 44h after MO injection, the et al, 2008; Sinkkonen et al, 2008). In the pre-implantation knockdownembryosweredistributedmainlyoverthe2-cell, embryo,wefoundfromourmicroarrayanalysisthatalthough 3-cell, 4-cell, and more advanced multicell stages. Hence, in Rbl2didnotpassourcutoffcriterionforsignificantgenes,its our second set of analyses, we asked if our BioMark expression is still upregulated 1.6- to 2.9-fold during knock- experiments could identify genes that would differentiate downof Oct4,Sall4, orNanog.Our microarrayanalysis also early-arrestedembryos(2-celland3-cellembryos)fromlate- revealedthatDnmt3bexpressionwasdownregulatedby3-to arrested embryos (4-cell and multicell embryos) during 19-foldwhenOct4,Sall4,orNanogwasdepleted(FDRo0.1; knockdownofapluripotencyfactor(blueboxesinFigure4A). Supplementary Files S3 and S5). Hence, a coherent feed- Interestingly,theexpressionofahandfulofgenescorrelates forward loop (FFL) consisting of the pluripotency factors, with the extent to which a knockdown embryo develops. the miR-290-295 cluster, Rbl2, and Dnmt3b regulates Figure 4C shows a heatmap of the top 10 genes that are DNA methylation in the pre-specified embryo and in ESCs differentially expressed between early-arrested embryos and (Figure8B). late-arrested embryos. Most of them encode DNA-binding Toobtainaquantitativeunderstandingofthefeed-forward proteins,suchasthetranscriptionfactorTbx3,whichshares architecture, we constructed a model of non-linear ordinary manycommontargetswithOct4andNanog(Hanetal,2010). differential equations (ODEs) based on Michaelis-Menten Wefurthercompared4-cellknockdownembryoswithmulti- kinetics with cooperativity (Figure 6A) and performed the cellknockdownembryos(TableI)andfoundthat7ofthe10 simulation using a Runge-Kutta algorithm. We chose the genes that were differentially expressed between early- parameters so that the model would reproduce known arrested embryos and late-arrested embryos were also more experimental observations. In uninjected (wild-type) highlyexpressedinmulticellembryosthanin4-cellembryos. embryos, the pluripotency factors are present at high levels &2013EMBOandMacmillanPublishersLimited MolecularSystemsBiology2013 7 AnOct4-Sall4-Nanognetwork MHTanetal and activate the expression of the miR-290-295 cluster. The approximately bigger than 5, the hypersensitive response is miRNAsthendownregulatetheproductionofRbl2.Whenthe lost and the FFL essentially behaves like the simple two level of Rbl2 is low, Oct4, Sall4, and Nanog can then fully componentsystem.However,whentheratiobecomessmaller, activate the expression of Dnmt3b (Figure 6B, left panel). the switch-like behavior becomes more pronounced However, the converse is true during knockdown of the (Figure 6D; Supplementary Figure S14). In contrast, the pluripotencyfactors(Figure6B,rightpanel).Furthermore,in steepest slope of the response curve is unaffected by the Dicer-null ESCs, both the transcript and protein levels of parameter b (see Supplementary information and c Dnmt3bweresignificantlylowerthanthatinwild-typeESCs SupplementaryFigureS15).Hence,ouranalysisshowedthat (Benettietal,2008;Sinkkonenetal,2008).Inagreementwith inordertoachieveahypersensitiveresponse,theregulationof theseobservations,whenwesettheexpressionofthemiRNAs Dnmt3b transcription cannot be dominated by the pluripo- to zero in our model, we found that the peak Dnmt3b tencyfactors,sothatRbl2canexertitseffect. expression level dropped by 1.7-fold for our chosen set of A prediction from our model is that at high levels of parameters(SupplementaryFileS17). pluripotency factors, for example in pre-specified embryos WenextinvestigatedtheadvantagesthattheFFLconferson and in ESCs, the concentration of Dnmt3b becomes more the pluripotency factors’ regulation of DNA methylation. susceptibletointrinsicgeneexpressionnoisewhentheFFLis Superficially, the pathway through the miRNAs and Rbl2 broken.To test the prediction, wedetermined the cell-to-cell appears to be unnecessary, as the pluripotency factors can variabilityofDnmt3btranscriptlevelforwild-typeESCsand simply activate the expression of Dnmt3b. To identify for Dicer(cid:2)/(cid:2) ESCs (Calabrese et al, 2007). Specifically, differencesbetweenthisstraightforwarddesignandtheFFL, we performed quantitative real-time PCR (RT-qPCR) experi- wealsosimulatedthetwo-componentmodelconsistingofthe ments on fourdifferent sets of cells: (1) Dicer(cid:2)/(cid:2) ESCs,(2) pluripotency factors and Dnmt3b only (Supplementary File Dicer(cid:2)/(cid:2) ESCstransfectedwithacontrolmiRNAmimicthat S18). When we varied the concentration of the pluripotency hasminimal sequence identitywithmiRNAsin humansand factors from low (0) to high (1) and examined the output mice(Dharmacon),(3)Dicer(cid:2)/(cid:2)ESCstransfectedwithamiR- (normalized Dnmt3b expression level) of both models, we 295 mimic (Dharmacon), and (4) wild-type ESCs. To verify foundthatthesimpletwo-componentmodelproducedamore thatthetransfectionofmiR-295wassuccessful,wemeasured gradedoutput,whiletheFFLgaveamoreswitch-likeresponse thelevelsofitstargetgenesRbl2,p21,andCasp2(Benettietal, (Figure 6C). Hence, the miRNA-Rbl2 pathway served to 2008; Sinkkonen et al, 2008; Wang et al, 2008; Zheng et al, sharpen the system output so that it has a more switch-like 2011) and confirmed that they were all downregulated behavior. (SupplementaryFigureS16).Wealsotransfectedeachofthe Sincetheperformanceofanon-lineardynamicalsystemis four cell populations with a plasmid encoding green fluor- dependent on the precise values of the parameters, we escentprotein(GFP),sothatwecouldsortforsinglecellsby wondered how our results may change with different flowcytometry, and then determined the transcript levels of parameter choices. We first found that the feed-forward Dnmt3b, Oct4, Sall4, and Nanog in individual ESCs using network motif is robust with respect to the synthesis rate of Biomark(Fluidigm).Strikingly,Dnmt3bexpressionexhibited the Rbl2 transcriptional repressor (see Supplementary the widest spread of Ct values in Dicer-null ESCs. The informationandSupplementaryFigureS13).Next,westudied interquartilerangeforDicer-nullESCsandforcellstransfected the behavior of the dynamical system when we varied the with the control miRNA was 8.94 and 3.33 respectively, synthesisrateofDnmt3b.TheexpressionofDnmt3bisunder whiletherangeforDicer-nullESCstransfectedwithmiR-295 combinatorialregulationbythepluripotencyfactorsandRbl2. and for wild-type ESCs was 2.87 and 2.79 respectively InourmathematicaldescriptionoftheFFL,wehavemodeled (Figure 6E, upperpanel). The standarddeviation of Dnmt3b the influences of each transcription factor as both mutually expressionlevelwasalsolargerinDicer-nullESCsandcontrol- independent(secondandthirdtermsontheright-handsideof transfected cells than in miR-295-transfected cells and wild- thethirdODE)andcooperative(fourthtermofthethirdODE). type ESCs (Supplementary Figure S17A). Furthermore, these However, the relative contributions of each regulator to the resultsarenotduetotheexpressionofthepluripotencyfactors transcription of Dnmt3b are unknown. Thus, we varied the exhibitingdifferentdegreesofvariabilitybetweenthefoursets ratioofthetwoparameters,b andb ,andobservedhowthe of cells (Figure 6E, lower panel, and Supplementary Figure a b concentrationofDnmt3brespondstochangesinthelevelof S17B). When corrected for the standard deviation of the the pluripotency factors. We found that when b /b is pluripotencyfactors,theexpressionofDnmt3bwasstillmore a b Figure4 Geneexpressionanalysisofindividualembryosrevealshiddenmolecularchangesandtranscriptdifferencesbetweenearly-andlate-arrestedembryos. (A)Aschematicdiagramshowingthedevelopmentalprogressionofuninjectedandinjectedembryos.Embryoswerecollectedat44hpost-injectionforexpression analysis.Twomainanalyseswereperformed.First,wecomparedtheexpressionofcontrolembryosthatwereatthe4-cellandmulticellstageswiththeexpressionof knockdownembryosthatwerealsoatthe4-cellandmulticellstages(redboxes).Second,wecomparedtheexpressionofknockdownembryosthatwereatthe2-celland 3-cellstages(early-arrested)withtheexpressionofknockdownembryosthatwereatthe4-cellandmulticellstages(late-arrested)(blueboxes).(B)Thefirstcomparison. Althoughalltheembryosinthisanalysisweremorphologicallysimilar(4-cellandmulticellstages),manygeneswerestilldifferentiallyexpressedintheknockdown embryos.TheindividualCtvaluesforthetop10rankinggenesofeachknockdownconditionareplottedasheatmaps.P-valuesaregiveninbrackets.Toppanel:Oct4 knockdown;middlepanel:Sall4knockdown;bottompanel:Nanogknockdown.Mostofthegenesshownaredownregulateduponknockdownofapluripotencyfactor. Thegenesthatareupregulatedaremarkedbyanasterisk(*).Inaddition,genesthatarehighlightedinpinkarecommonamongallthethreeknockdownconditionsand theyparticipateincriticalfunctionssuchasepigeneticgenesilencing(Pengetal,2009;Lietal,2010;Pasinietal,2010),autophagy(Tsukamotoetal,2008),and alternativesplicing(Tangetal,2010).(C)Thesecondcomparison.TheindividualCtvaluesforthetop10rankinggenesareplottedasaheatmap,withtheP-values giveninbrackets.Theexpressionlevelsofthegeneshighlightedingreenarealsosignificantlydifferentbetweenthe4-cellstageandthemulticellstage. 8 MolecularSystemsBiology2013 &2013EMBOandMacmillanPublishersLimited AnOct4-Sall4-Nanognetwork MHTanetal variable in Dicer-null cells and control-transfected cells that the FFL serves to buffer Dnmt3b expression against than in miR-295-transfected cells and wild-type ESCs intrinsic variabilityin the concentrations of the pluripotency (Supplementary Figure S17C). Hence, our results indicate factors. A 44 h later Uninjected Injected with 44 h later control MO Injected with 44 h later Oct4/Sall4/Nanog MO B Uninjected Control MO Oct4 MO Hat1 (P = 4.46×10–56) Jarid2 (P = 1.35×10–55) Aqp3 (P = 1.81×10–54) Atg2a (P = 7.82×10–54) Klf2 (P = 8.72×10–50) Sf3b2 (P = 3.84×10–48) Vps72 (P = 5.80×10–48) Sf1 (P = 1.74×10–46) Dnmt3b (P = 4.51×10–46) Trim28 (P = 3.35×10–42) Uninjected Control MO Sall4 MO Atg2a (P = 6.47×10–46) Sf1 (P = 2.20×10–44) Jarid2 (P = 1.47×10–43) Sf3b2 (P = 1.52×10–43) Trim28 (P = 1.46×10–42) Dnmt3a (P = 2.07×10–41) Dppa5a (P = 3.28×10–41) Aqp3 (P = 5.61×10–41) Pou2f3 (P = 1.73×10–37) *Spry4 (P = 3.93×10–35) Uninjected Control MO Nanog MO Sf3b2 (P = 8.23×10–25) Smarcd1 (P = 1.10×10–20) Dnmt3b (P = 2.15×10–20) Jarid2 (P = 4.30×10–20) Atg2a (P = 3.27×10–19) Aqp3 (P = 3.41x10–18) Hmgb1 (P = 3.84×10–18) Klf2 (P = 1.58×10–17) *Fzd4 (P = 1.70×10–17) Tcfap2c (P = 1×10–17) ≤–3–2 –1 0 1 2 ≥3 Ct relative to mean C 2-cell and 3-cell 4-cell Multicell Hmgb1 (P = 1.16×10–33) Sf3b2 (P = 1.86×10–22) Dnmt3b (P = 1.29×10–20) Klf2 (P = 2.76×10–17) Klf5 (P = 1.10×10–16) Trim24 (P = 1.07×10–13) Pou5f1 (P = 1.50×10–13) Wwtr1 (P = 2.84×10–13) Atg7 (P = 2.21×10–12) Tbx3 (P = 1.28×10–11) &2013EMBOandMacmillanPublishersLimited MolecularSystemsBiology2013 9 AnOct4-Sall4-Nanognetwork MHTanetal Single-cell analysis reveals an increase in gene development,whileotherembryosmightexpressalowerlevel expression noise during knockdown of Oct4 or of Dnmt3b and arrest earlier in development. In uninjected Sall4 embryos, the pluripotency factors are present at high levels andfluctuationsintheirconcentrationsduetointrinsicnoise Our simulations suggest an intriguing explanation for why do not affect the level of Dnmt3b significantly (right side of during knockdown of a pluripotency factor, some embryos bluegraphinFigure6C).However,whenapluripotencyfactor mightexpressahigherlevelofDnmt3bandprogressfurtherin has been knocked down to a low level, the gradient of the responsecurveincreasesdramaticallyandthesystemexistsin theswitchregimeinstead(leftsideofbluegraphinFigure6C). Table I The top 10 genes that were differentially expressed between 4-cell In this case, even small fluctuations in the concentrations of knockdownembryosandmulticellknockdownembryos the pluripotency factors can affect the level of Dnmt3b Gene P-value Foldchange appreciably. To test the hypothesis that the expression of Dnmt3b and Fgfr2 1.80(cid:3)10(cid:2)12 2.44 possiblymanyothergenesbecomenoisierduringknockdown Hmgb1 2.13(cid:3)10(cid:2)12 2.20 Pou5f1 8.16(cid:3)10(cid:2)12 2.46 of a pluripotency factor, we performed single-cell real-time Dnmt3b 1.47(cid:3)10(cid:2)11 1.91 PCRexperiments using BioMark (Fluidigm). In these experi- Tbx3 1.52(cid:3)10(cid:2)11 2.45 ments, we focused on Oct4 and Sall4, since knockdown of Sf3b2 6.92(cid:3)10(cid:2)8 1.48 Atg7 4.53(cid:3)10(cid:2)7 2.02 Nanog gave a milder phenotype that is consistent with the Racgap1 6.17(cid:3)10(cid:2)7 1.49 expression of Nanog being significantly lower than that of Klf5 8.43(cid:3)10(cid:2)7 1.64 Oct4 or Sall4 in the earliest developmental stages. For each Dnmt3a 2.31(cid:3)10(cid:2)6 1.46 experimental condition (no injection, injection with control MO,injectionwithOct4MO,andinjectionwithSall4MO),we Thegeneshighlightedingreenwerealsoexpressedatsignificantlylowerlevels in2-celland3-cellembryoswhencomparedwithlaterstageembryos. collectedembryosthatweremorphologicallyatthe4-cellstage A C D 100 100 110 90 90 100 80 80 90 age 70 age 70 ent 80 Arrest by 2-cell st 34560000 Arrest by 2-cell st 34560000 Morula developm 45670000 % % % 30 20 20 20 10 10 10 0 0 0 O O O O d O O +A d O O +A Oct4 M Sall4 M Nanog M Dnmt3b M Uninjecte Control M Dnmt3b M nmt3b MOmt3b mRN Uninjecte Control M Dnmt3b M nmt3b MOmt3b mRN Dn Dn D D B Uninjected Control MO Dnmt3b MO Dnmt3b MO+Dnmt3b mRNA Figure5 TheDNAmethyltransferaseDnmt3bisessentialforpre-implantationdevelopment.(A)Percentagesofembryosthatarrestedbythe2-cellstageforfour differentknockdownconditions.Unexpectedly,weobservedthatthemajorityofembryosinjectedwithaMOtargetingDnmt3barrestedaroundthematernal-zygotic transition(Po0.001,Student’st-test,whencomparingDnmt3bwithanyofthepluripotencyfactors).(B)Representativemicroscopyimagesofpre-implantationmouse embryosundervariousexperimentalconditions.(C)Quantificationofobservedphenotypesafter4daysofinvitroculture.Comparedwithembryosthatwereinjected with0.6mMDnmt3bMOonly,co-injectionof0.6mMDnmt3bMOwith1–10ng/mlDnmt3bmRNAreducedthefractionofembryosthatarrestedbythe2-cellstage (7independentexperiments)(Po0.05,Student’st-test).(D)Percentagesofembryosthatdevelopeduntilatleastthemorulastageafter4daysofculture.Compared withembryosthatwereinjectedwithDnmt3bMOonly,co-injectionofDnmt3bMOwithinvitrotranscribedDnmt3bmRNAincreasedthepercentageofembryosthat reachedthemorulastage(Po0.05,Student’st-test). 10 MolecularSystemsBiology2013 &2013EMBOandMacmillanPublishersLimited