Inoue-2017-Maternal H3K27me3 controls DNA meth.pdf

2 7 J u ly 2 0 1 7|V O l 5 4 7|N A T u R E|4 1 9ARTiclEdoi:10.1038/nature23262Maternal H3K27me3 controls DNA methylation-independent imprintingAzusa inoue1,2,3*,lan Jiang1,2,3*,Falong lu1,2,3*,Tsukasa Suzuki1,2,3&yi Zhang1,2,3,4,5Sperm and oocytes are generated from primordial germ cells through distinct processes.Consequently,their genomes are packaged differ-ently with distinct epigenetic landscapes1.After fertilization,paternal chromatin releases protamines and is repackaged with maternally stored histones that are devoid of most histone modifications,while maternal chromatin harbours various histone modifications inher-ited from oocytes2,3.The different processes of parental chromatin formation result in parental epigenetic asymmetry in zygotes1,which becomes largely equalized during subsequent development with the exception of certain genomic loci,including imprinting control regions(ICRs)4.Transcriptional regulatory elements,such as promoters and enhancers,can be mapped by DNase I hyper-sensitivity assay5,6.By using a low-input DNase I-sequencing(liDNase-seq)technique,we recently mapped the transcriptional regulatory landscape of preim-plantation embryos;single nucleotide polymorphism(SNP)-based analysis revealed that chromatin accessibility of the two parental alleles is,overall,comparable,except for imprinted gene promoters7.A similar conclusion was also reached using an assay for transposase-accessible chromatin with high-throughput sequencing(ATAC-seq)8.However,the mechanisms underlying parent-of-origin specific chromatin acces-sibility are unknown.Allelic DNase I hypersensitive sites in zygotesTo comprehensively profile parental allele-specific DNase I hyper-sensitive sites(DHSs)in zygotes,we isolated paternal and maternal pronuclei from pronucleus stage 5(PN5)zygotes and performed liD-Nase-seq(Fig.1a and Extended Data Fig.1a).Using stringent criteria(Extended Data Fig.1b)and excluding data for sex chromosomes,we identified 3,462,687,and 169 bi-allelic DHSs,paternal allele-specific DHSs(Ps-DHSs),and maternal allele-specific DHSs(Ms-DHSs),respectively(Fig.1b,Extended Data Fig.1c and Supplementary Table 1).The genomic location of allelic DHSs was heavily biased to non-promoter elements compared with bi-allelic DHSs enriched in promoters and CpG islands(Extended Data Fig.1d,e).Similar to previous findings7,Ps-DHSs included ICRs of known imprinted genes(Extended Data Fig.1f).Interestingly,both Ps-and Ms-DHSs also included promoters of genes not previously known to be imprinted(Extended Data Fig.1g,h).Since promoter DHSs can prime gene expression at the next developmental stage7,we asked whether allelic DHSs in zygotes can prime allelic gene expression at zygotic genome activation(ZGA).RNA sequencing(RNA-seq)analysis of two-cell stage androgenetic and gynogenetic embryos,using -amanitin treatment as a negative control,identified 107 androgenetic-and 14 gynogenetic-specific dif-ferentially expressed genes(DEGs),including 8 known imprinted genes(Extended Data Fig.2ad and Supplementary Table 2).Integrated analysis of allelic ZGA and allelic promoter DHSs in zygotes revealed that most(59%and 79%)of the androgenetic-and gynogenetic-specific DEGs were associated with paternal and mater-nal allele-biased chromatin accessibility,respectively(Extended Data Fig.2e,f).Genes showing such a correlation included not only known imprinted genes but also those not known to be imprinted(Fig.1c).These results suggest that allelic DHSs in zygotes can mark promoters that are primed for allelic ZGA.DNA methylation and allelic DHSsTo understand how allelic DHSs in zygotes are specified,we first exam-ined whether they were inherited from gametes.We profiled DHSs of fully grown oocytes(Extended Data Fig.3a)and analysed sperm DHSs7.Although sperm had only 34 reproducible DHSs(Supplementary Table 3),some of them contributed to Ps-DHSs(Extended Data Mammalian sperm and oocytes have different epigenetic landscapes and are organized in different fashions.After fertilization,the initially distinct parental epigenomes become largely equalized with the exception of certain loci,including imprinting control regions.How parental chromatin becomes equalized and how imprinting control regions escape from this reprogramming is largely unknown.Here we profile parental allele-specific DNase I hypersensitive sites in mouse zygotes and morula embryos,and investigate the epigenetic mechanisms underlying these allelic sites.Integrated analyses of DNA methylome and tri-methylation at lysine 27 of histone H3(H3K27me3)chromatin immunoprecipitation followed by sequencing identify 76 genes with paternal allele-specific DNase I hypersensitive sites that are devoid of DNA methylation but harbour maternal allele-specific H3K27me3.Interestingly,these genes are paternally expressed in preimplantation embryos,and ectopic removal of H3K27me3 induces maternal allele expression.H3K27me3-dependent imprinting is largely lost in the embryonic cell lineage,but at least five genes maintain their imprinted expression in the extra-embryonic cell lineage.The five genes include all paternally expressed autosomal imprinted genes previously demonstrated to be independent of oocyte DNA methylation.Thus,our study identifies maternal H3K27me3 as a DNA methylation-independent imprinting mechanism.1Howard Hughes Medical Institute,Boston Childrens Hospital,Boston,Massachusetts 02115,USA.2Program in Cellular and Molecular Medicine,Boston Childrens Hospital,Boston,Massachusetts 02115,USA.3Division of Hematology/Oncology,Department of Pediatrics,Boston Childrens Hospital,Boston,Massachusetts 02115,USA.4Department of Genetics,Harvard Medical School,Boston,Massachusetts 02115,USA.5Harvard Stem Cell Institute,Boston,Massachusetts 02115,USA.Present address:State Key Laboratory of Molecular Developmental Biology,Institute of Genetics and Developmental Biology,Chinese Academy of Sciences,Beijing 100101,China.*These authors contributed equally to this work.2017 Macmillan Publishers Limited,part of Springer Nature.All rights reserved.articlereSearcH4 2 0|N A T u R E|V O l 5 4 7|2 7 J u ly 2 0 1 7Fig.3b).However,most Ps-DHSs were absent in sperm and oocytes,indicating that they were generated after fertilization(Extended Data Fig.3c,d).In contrast,most Ms-DHSs and bi-allelic DHSs were already present in oocytes(Extended Data Fig.3eh),indicating that most maternal DHSs are inherited from oocytes.To determine how the maternal allele at Ps-DHSs remains inacces-sible,we first hypothesized that maternal DNA methylation prevents DHS formation.Analysis of a public whole-genome bisulfite sequencing (WGBS)data set of oocytes and sperm9 revealed that only 17%of Ps-DHSs overlapped with oocyte germline differentially methylated regions(gDMRs)(Extended Data Fig.4a and Supplementary Table 4).Despite extending to a 100kilobase(kb)region flanking Ps-DHSs,only an additional 21%were found to be associated with oocyte gDMRs(Extended Data Fig.4a and Supplementary Table 4).Even when the oocyte DNA methylation level alone was considered,48%of Ps-DHSs were devoid of oocyte DNA methylation(Extended Data Fig.4b),indicating the existence of a DNA methylation-independent mecha-nism that prevents accessibility of the maternal allele.Maternal allelic protection by H3K27me3The fact that Polycomb-mediated H3K27me3 can mediate silencing of DNA hypomethylated promoters10 prompted us to postulate that H3K27me3 might be responsible for maternal allele inaccessibility.Analyses of public ChIPseq(chromatin immunoprecipitation followed by sequencing)data sets11 revealed that the H3K27me3 level in oocytes was much higher than that of sperm at DNA hypomethylated Ps-DHSs,while it was reversed at DNA hypermethylated Ps-DHSs(Extended Data Fig.4c,left).SNP tracking analysis revealed that the hypometh-ylated Ps-DHSs maintained maternal allele-specific H3K27me3 in zygotes(Extended Data Fig.4c,right),indicating that H3K27me3 might be responsible for maternal allele inaccessibility at DNA hypomethylated regions.To test this possibility,we injected messenger RNA(mRNA)encod-ing an H3K27me3-specific demethylase Kdm6b(Kdm6bWT)with its catalytic mutant(H1390A)(Kdm6bMUT)as a control12(Fig.2a).Similarly,we prepared zygotes injected with an H3K9me3-specific demethylase Kdm4d or its catalytic mutant(H189A)13.Both wild-type and mutant Kdm6b and Kdm4d were expressed at a similar level(Extended Data Fig.4d),and Kdm6bWT and Kdm4dWT,but not their mutants,significantly reduced H3K27me3 and H3K9me3 levels,respectively(Extended Data Fig.4e,f).liDNase-seq of isolated pronuclei(Extended Data Fig.4g,h)revealed that 78 and 150 of the 431 most reliable Ps-DHSs became bi-allelic in Kdm6bWT-and Kdm4dWT-injected zygotes,respectively,while their catalytic mutants had little effect(Fig.2b,c,Extended Data Fig.4i and Supplementary Table 5).This result indicates that both maternal H3K27me3 and H3K9me3 are involved in maternal allele inaccessibility.Importantly,Kdm6b-affected Ps-DHSs were largely devoid of oocyte DNA methylation,which was markedly different from Kdm4d-affected Ps-DHSs that located at DNA hypermethylated regions(Fig.2d).Consistently,Ps-DHSs specifically affected by Kdm6b,but not Kdm4d,overlapped maternal allele-specific H3K27me3(Extended Data Fig.4j).These results suggest that maternal H3K27me3 and H3K9me3 restrict maternal allele accessibility at regions with hypomethylated and hyper-methylated DNA,respectively.H3K27me3-dependent imprintingTo understand to what extent allelic DHSs exist at a later embryonic stage,we generated androgenetic and gynogenetic morula embryos(Fig.3a)and performed liDNase-seq(Extended Data Fig.5a).Using Mov10l1 Akap1 Isl2 Pmaip1 Zfp871 SgceMcts2 10acBi-allelic DHSs(n=3,462)Ps-DHSs(n=687)Ms-DHSs(n=169)Paternal PNMaternal PN3.60RPKM Maternal PNPaternal PNIVF12 h bFPKMPaternal12Maternal12Pmaip1 1502.5 kbZfp871 02.5 kb10Two-cell androgenesis-specifc DEGs Undefned genes Sgce2005 kbMcts2 02.5 kbKnown imprinted genes 0 50 100 150 200 AG GG Ama 0 2 4 6 AG GG Ama 0 10 20 30 40 AG GG Ama 0 5 10 15 AG GG Ama Mov10l1 Akap1 1505 kb1505 kbIsl2 02.5 kb25Undefned genes Two-cell gynogenesis-specifc DEGs0 2 4 6 8 10 12 AG GG Ama 0 5 10 15 AG GG Ama 0 10 20 30 AG GG Ama 5555 kbFigure 1|Allelic DHSs in zygotes mark allelic gene expression at ZGA.a,Schematic for identifying parental allele-specific DHSs in zygotes.IVF,in vitro fertilization.PN,pronucleus.b,Heat map showing bi-allelic,Ps-,and Ms-DHSs in zygotes.Each row represents liDNase-seq signal intensity at a DHS 5 kb.c,Representative androgenesis-and gynogenesis-specific DEGs harbouring allelic promoter DHSs in zygotes.Top,genome browser views of DHSs in paternal and maternal pronuclei with biological duplicates.The DHS signal intensity and the genomic length of each view(in kilobases)are indicated at the top left and the bottom of each panel,respectively.Bottom,gene expression levels in androgenetic(AG),gynogenetic(GG),and -amanitin-treated(Ama)two-cell embryos.Error bar,s.d.of biological duplicates.Note that gynogenetic-specific expression is evident after subtraction of maternal pool transcripts.bKdm6b-affected n=78(18%)Kdm4d-affected n=150(35%)Kdm6bKdm4dMost reliable Ps-DHSs(n=431)aIVFKdm6b or Kdm4d mRNA4 h+8 hdKdm6b-affectedPs-DHSs Kdm4d-affectedPs-DHSs 81%7%12%9%88%80100%2080%020%DNA methylation levelin oocytes c12Kdm6bMUTPatMat 1212PatMat 12Kdm6bWT25Slit32.5 kb0Trim92.5 kb25Hip12.5 kb0250250Dab2ip2.5 kbPaternalMaternallog2(fold change)(paternal/maternal)No injectionWTMutWTMut21012Figure 2|Oocyte-specific H3K27me3 prevents maternal chromatin accessibility at DNA hypomethylated regions.a,Schematic for studying the role of histone methylations in maternal chromatin inaccessibility.b,Heat map showing the allelic bias at Ps-DHSs in Kdm6b-or Kdm4d-injected zygotes.WT,wild-type.Mut,catalytic mutant.c,Genome browser view of representative Ps-DHSs affected by Kdm6bWT.Pat,paternal;Mat,maternal.d,Kdm6b-or Kdm4d-affected Ps-DHSs organized on the basis of their oocyte DNA methylation levels.2017 Macmillan Publishers Limited,part of Springer Nature.All rights reserved.article reSearcH2 7 J u ly 2 0 1 7|V O l 5 4 7|N A T u R E|4 2 1the same criteria for allelic DHSs as in zygotes and excluding data for sex chromosomes,we identified 36,569 common DHSs,247 androge-netic-specific DHSs(AG-DHSs),and 176 gynogenetic-specific DHSs(GG-DHSs)(Fig.3b and Supplementary Table 6).By SNP tracking analyses of a public DHS profile of hybrid morula embryos7,we confirmed that AG-DHSs,but not GG-DHSs,recapitulated the cor-responding parental allele-specific DHSs(Extended Data Fig.5b),indicating that AG-DHSs are physiological.Interestingly,AG-DHSs included almost all known maternally methylated ICRs(Extended Data Fig.5c).This finding raised the possibility that AG-DHSs could serve as indicators of genomic imprinting.Because both DNA methylation and H3K27me3 can contribute to maternal allele inaccessibility(Fig.2),we determined their respective contribution to the generation of the 247 AG-DHSs.Analyses of the oocyte DNA methylome9 identified 183(74%)AG-DHSs in DNA hypomethylated regions(Extended Data Fig.5d).Allelic H3K27me3 enrichment analysis revealed that 112 of the 183 were marked with maternal allele-biased H3K27me3 in inner cell mass(ICM)of blas-tocysts(Fig.3c).Of the 112 AG-DHSs,105 showed maternal allele-specific H3K27me3 enrichment in zygotes(reads per million 0.5,fold change(maternal/paternal)4),suggesting that the maternal allele-biased H3K27me3 is inherited from zygotic maternal chromatin.By associating the 105 H3K27me3-marked AG-DHSs with their near-est genes,we obtained 76 genes(Supplementary Table 7)as putative H3K27me3-dependent imprinted genes.To examine the allelic gene expression of the 76 genes,we performed RNA-seq analysis for androgenetic and gynogenetic morula embryos(Extended Data Fig.6a and Supplementary Table 8).After confirming androgenetic-or gynogenetic-specific expression of known imprinted genes(Extended Data Fig.6b),we calculated the relative androgenetic/gynogenetic expression levels for each candidate.Among the 76 genes,28 were expressed in either androgenetic or gynogenetic embryos(fragments per kilobase of exon per million mapped fragments(FPKM)0.5).Interestingly,27 of the 28 genes exhibited biased(fold change 2),and 23 genes exhibited highly biased(fold change 8),expression in androgenetic embryos(Fig.3d,left column).Using an RNA-seq data set of hybrid morula embryos14,we confirmed that all 13 SNP-trackable genes exhibit paternal allele-specific expression (Fig.3d,right columns).Importantly,the 28 genes include Sfmbt2,Gab1,Slc38a4,and Phf17(also known as Jade1)(Fig.3d),whose imprinted expression was suggested to be independent of oocyte DNA methylation1518.These non-canonical imprinted genes are coated with oocyte-specific H3K27me3 domains that are retained even in blastocysts(Extended Data Fig.6c),which is in contrast to DNA methylation-dependent canonical imprinted genes that are dev