10.1101/gad.947102 Access the most recent version at doi: 2002 16: 6-21 Genes Dev.Adrian BirdDNA methylation patterns and epigenetic memoryReferenceshttp://genesdev.cshlp.org/content/16/1/6.full.html#related-urlsArticle cited in: http://genesdev.cshlp.org/content/16/1/6.full.html#ref-list-1This article cites 197 articles, 84 of which can be accessed free at:serviceEmail alertingclick here top right corner of the article orReceive free email alerts when new articles cite this article - sign up in the box at theCollectionsTopic (137 articles) Chromatin and Gene Expression Articles on similar topics can be found in the following collectionshttp://genesdev.cshlp.org/subscriptions go to:Genes & Development To subscribe to Cold Spring Harbor Laboratory PressCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from REVIEWDNA methylation patternsand epigenetic memoryAdrian Bird1Wellcome Trust Centre for Cell Biology, University of Edinburgh, Edinburgh EH9 3JR, UKThe character of a cell is defined by its constituent pro-teins, whicharetheresultofspecificpatternsofgeneexpression. Crucial determinants of gene expression pat-terns are DNA-binding transcription factors that choosegenes for transcriptional activation or repression by rec-ognizingthesequenceofDNAbasesintheirpromoterregions. Interactionofthesefactorswiththeircognatesequences triggers a chainof events, ofteninvolvingchanges in the structure of chromatin, that leads to theassembly of an active transcription complex (e.g., Cosmaet al. 1999). But the types of transcription factors presentin a cell are not alone sufficient to define its spectrum ofgeneactivity, asthetranscriptional potential of age-nomecanbecomerestrictedinastablemannerduringdevelopment. The constraints imposed by developmen-tal history probably account for the very low efficiencyof cloning animals from the nuclei of differentiated cells(Rideout et al. 2001; Wakayama and Yanagimachi 2001).A transcription factors only model would predict thatthegeneexpressionpatternofadifferentiatednucleuswould be completely reversible upon exposure to a newspectrumoffactors. Althoughmanyaspectsofexpres-sioncanbereprogrammedinthisway(Gurdon1999),some marks of differentiation are evidently so stable thatimmersion in an alien cytoplasmcannot erase thememory.The genomic sequence of a differentiated cell isthought to be identical in most cases to that of the zy-gotefromwhichitisdescended(mammalianBandTcellsbeinganobviousexception).Thismeansthatthemarks of developmental history are unlikely to becaused by widespread somatic mutation. Processes lessirrevocable than mutation fall under the umbrella termepigenetic mechanisms. A current definition of epige-neticsis: Thestudyofmitoticallyand/ormeioticallyheritablechangesingenefunctionthat cannot beex-plained by changes inDNAsequence (Russo et al.1996). There are two epigenetic systems that affect ani-mal development and fulfill the criterion of heritability:DNAmethylationand the Polycomb-trithorax group(Pc-G/trx) protein complexes. (Histone modification hassome attributes of an epigenetic process, but the issue ofheritability has yet to be resolved.) This review concernsDNAmethylation, focusingonthegeneration, inheri-tance, andbiological significanceof genomicmethyl-ationpatternsinthedevelopment of mammals. Datawill be discussed favoring the notion that DNA methyl-ation may only affect genes that are already silenced byother mechanisms in the embryo. Embryonic transcrip-tion, on the other hand, may cause the exclusion of theDNA methylation machinery. The heritability of meth-ylation states and the secondary nature of the decision toinvite or exclude methylation support the idea that DNAmethylationisadaptedfor aspecificcellular memoryfunction in development. Indeed, the possibility will bediscussed that DNA methylation and Pc-G/trx may rep-resent alternativesystems of epigeneticmemorythathave been interchanged over evolutionary time. AnimalDNA methylation has been the subject of several recentreviews (Bird and Wolffe 1999; Bestor 2000; Hsieh 2000;Costello and Plass 2001; Jones and Takai 2001). For re-cent reviews of plant and fungal DNA methylation, seeFinnegan et al. (2000), Martienssen and Colot (2001), andMatzke et al. (2001).Variable patterns of DNA methylation in animalsAprerequisiteforunderstandingthefunctionofDNAmethylationisknowledgeofitsdistributioninthege-nome. Inanimals, thespectrumof methylationlevelsandpatternsisverybroad. Atthelowextremeisthenematode worm Caenorhabditis elegans, whose genomelacks detectablem5Canddoes not encodeaconven-tional DNAmethyltransferase. Another invertebrate,the insect Drosophila melanogaster, long thought to bedevoidof methylation, hasaDNAmethyltransferase-like gene (Hung et al. 1999; Tweedie et al. 1999) and isreported to contain very low m5C levels (Gowher et al.2000; Lyko et al. 2000), although mostly in the CpT di-nucleotide rather than in CpG, which is the major targetfor methylation in animals. Most other invertebrate ge-nomes have moderately high levels of methyl-CpG con-centrated in large domains of methylated DNA separatedby equivalent domains of unmethylated DNA (Bird et al.1979; Tweedie et al. 1997). This mosaic methylation pat-tern has been confirmed at higher resolution in the sea1E-MAIL A.Bird@ed.ac.uk; FAX 0131-650-5379.Article and publication are at http://www.genesdev.org/cgi/doi/10.1101/gad.947102.6 GENES & DEVELOPMENT 16:621 2002 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/02 $5.00; www.genesdev.orgCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from squirt, Cionaintestinalis(Simmenetal. 1999). Attheopposite extreme from C. elegans are the vertebrate ge-nomes,whichhavethehighestlevelsofm5Cfoundinthe animal kingdom. Vertebrate methylation is dis-persed over much of the genome, a pattern referred to asglobal methylation. The variety of animal DNA methyl-ationpatterns highlights thepossibilitythat differentdistributions reflect different functions for the DNAmethylation system (Colot and Rossignol 1999).Mammalian DNA methylation patterns vary in timeand spaceIn human somatic cells, m5C accounts for 1% of totalDNAbasesandthereforeaffects70%80%ofallCpGdinucleotides in the genome (Ehrlich 1982). This averagepatternconcealsintriguingtemporal andspatial varia-tion. Duringadiscretephaseof earlymousedevelop-ment,methylationlevelsinthemousedeclinesharplyto 30% of the typical somatic level (Monk et al. 1987;Kafri et al. 1992). De novo methylation restores normallevels by the time of implantation. A much more limiteddropinmethylationoccursinthefrogXenopuslaevis(Stancheva and Meehan 2000), and no drop is seen in thezebrafish, Danio rerio (MacLeod et al. 1999). Even withinvertebrates, therefore, interspecies variation is seen thatcouldreflect differencesinthepreciseroleplayedbymethylation in these organisms. For mice and probablyother mammals, however, the cycle of early embryonicdemethylation followed by de novo methylation is criti-cal in determining somatic DNA methylation patterns.A genome-wide reduction in methylation is also seen inprimordial germ cells (Tada et al. 1997; Reik et al. 2001)during the proliferative oogonial and spermatogonialstages.The most striking feature of vertebrate DNA methyl-ationpatternsisthepresenceof CpGislands, thatis,unmethylated GC-rich regions that possess high relativedensitiesof CpGandarepositionedat the5 endsofmany human genes (for review, see Bird 1987). Compu-tational analysisof thehumangenomesequencepre-dicts 29,000 CpG islands (Lander et al. 2001; Venter et al.2001). Earlier studies estimatedthat 60%of humangenes areassociatedwithCpGislands, of whichthegreat majority are unmethylated at all stages of develop-ment and in all tissue types (Antequera and Bird 1993).Because many CpG islands are located at genes that havea tissue-restricted expressionpattern, it follows thatCpGislands canremainmethylation-free evenwhentheirassociatedgeneissilent.Forexample,thetissue-specifically expressed human-globin (Bird et al. 1987)and 2(1) collagen (McKeon et al. 1982) genes have CpGislandsthatremainunmethylatedinalltestedtissues,regardless of expression.A small but significant proportion of all CpG islandsbecome methylated during development, and when thishappens the associated promoter is stably silent. Devel-opmentally programmed CpG-island methylation of thiskind is involved in genomic imprinting and X chromo-some inactivation (see below). The de novo methylationevents occur in germ cells or the early embryo (Jaenischet al. 1982), suggesting that de novo methylation is par-ticularly active at these stages. There is evidence, how-ever,thatdenovomethylationcanalsooccurinadultsomaticcells.AsignificantfractionofallhumanCpGislandsarepronetoprogressivemethylationincertaintissues during aging (for review, see Issa 2000), or in ab-normal cells such as cancers (for review, see Baylin andHerman 2000) and permanent cell lines (Harris 1982; An-tequera et al. 1990; Jones et al. 1990). The rate of accu-mulation of methylated CpGs in somatic cells appears tobeveryslow. For example, denovomethylationof aprovirus inmurine erythroleukemia cells tookmanyweekstocomplete(Lorinczetal. 2000). Similarly, therecoveryof global DNAmethylationlevels followingchronic treatment of mouse cells with the DNA meth-ylationinhibitor5-azacytidinerequiredmonths(Flatauet al. 1984).Howdo patterns of methylated and unmethylatedmammalianDNAarise indevelopment andhowarethey maintained? Why are CpG islands usually, but notalways, methylation-free?Whatcausesmethylationofbulknon-CpG-islandDNA? These burning questionscannot be answered definitively at present, but there aredistinct hypotheses that have been addressed experimen-tally. The available data will be conveniently consideredinthree parts: (1) mechanisms for maintaining DNAmethylation patterns; (2) mechanisms and consequencesof methylationgain; and(3) mechanisms andconse-quences of methylation loss.Maintenance methylationnot so simpleMaintenancemethylationdescribestheprocessesthatreproduce DNA methylation patterns between cell gen-erations. The simplest conceivable mechanism for main-tenance depends on semiconservative copying of the pa-rental-strand methylation pattern onto the progenyDNAstrand(HollidayandPugh1975; Riggs1975). Inkeeping with the model, the methylating enzymeDNMT1preferstomethylatethosenewCpGswhosepartnersontheparentalstrandalreadycarryamethylgroup (Bestor 1992; Pradhan et al. 1999). Thus a patternofmethylatedandnonmethylatedCpGsalongaDNAstrandtendstobecopied, andthisprovidesawayofpassing epigenetic information between cell generations.TheideathatmammalianDNAmethylationpatternsareestablishedinearlydevelopmentbydenovometh-yltransferases DNMT3AandDNMT3B(Okanoet al.1998a, 1999; Hsieh1999b) andthencopiedtosomaticcells by the maintenance DNA methyltransferaseDNMT1 is elegant and simple, but, as discussed below,maynot fullyexplainpersistenceof methylationpat-terns during cell proliferation.Experimentsthatfirstshowedreplicationofmethyl-ationpatternsonartificiallymethylatedDNAalsore-vealed a relatively low fidelity for the process (Pollack etal. 1980; Wigler et al. 1981). After many cell generations,methylation of the introduced DNA was retained, but atamuchlowerlevel thaninthestartingplasmid. TheDNA methylation and epigenetic memoryGENES & DEVELOPMENT 7Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from failureof maintenancewasestimatedtooccurwithafrequency of 5% per CpG site per cell division. Quan-titative studies of an endogenous CpGsite broadlyagreed with this figure (Riggs et al. 1998). Cell clones inwhich this site was initially unmethylated acquiredmethylationandcloneswhereit wasmethylatedlostmethylation. The rate of change was estimated at 4%percellgeneration.Errorratesofthismagnitudemeanthat a detailed methylation pattern would eventually be-come indistinct as cells proliferate. Indeed, dynamicchanges in detailed methylation patterns have been ob-served in monoclonal lyomyomas (Silva et al. 1993) andat the methylated FMR1 gene (Stger et al. 1997). Thesestudiesestablishedthat clonal populationsof cellsdonot havethehomogeneous methylationpatterns thatwould be predicted by the replication model of mainte-nance methylation. Not only does DNA methyltransfer-ase fail to complete half-methylated sites at a significantrate, but also significant de novo methylation occurs atunmethylated sites.At first sight, these findings appear to undermine theconcept of maintenance methylation, but this does notfollow. Although detailed methylation patterns may notbemaintainedatthelevelofasingleCpGnucleotide,themethylationstatusofDNAdomainsappearstobefaithfullypropagatedduringdevelopment(Pfeiferetal.1990).CpGislands,forexample,keeptheiroverallun-methylated state (or methylated state) extremely stablythrough multiple cell generations. DNMT1 is partly re-sponsible for this stability, but there is likely to be an-otherasyetunknowncomponenttothemaintenanceprocess. Dramatic evidence for this alternative mainte-nance mechanism comes from the finding that CpG-is-land methylation is stably maintained even in the appar-ent absence of the only known maintenance DNA meth-yltransferase, DNMT1 (Rhee et al. 2000). Asimilarphenomenon may account for the maintenance of allele-specific DNAmethylationimprints under conditionswhere the concentration of DNMT1 is severely limiting(Jaenisch 1997).De novo DNA methylation by default?The origin of DNA methylation patterns is a long-stand-ing mystery in the field. The de novo methyltransferasesDNMT3A and DNMT3B (Okano et al. 1998a, 1999) arehighly expressed in early embryonic cells, and it is at thisstage that most programmed de novo methylation eventsoccur. What determineswhichregionsof thegenomeshould be methylated? An extreme possibility is that denovoDNAmethylationinearlymammaliandevelop-mentisanindiscriminateprocesspotentiallyaffectingall CpGs. Compatible with the default model is the ap-parent absence of intrinsically unmethylatable DNA se-quences inmammaliangenomes. EvenCpGislands,most of which are unmethylated at all times in normalcells, canacquiremethylationunder special develop-mentalcircumstancesorinabnormalcells(permanentcell lines or cancer cells). It is clear, however, that not allregionsof thegenomeareequallyaccessibletoDNAmethyltransferases. DNMT3B in particular is known tobe required for de novo methylation of specific genomicregions, as mice or human patients with DNMT3B mu-tationsaredeficientinmethylationofpericentromericrepetitiveDNAsequencesandat CpGislandsontheinactive X chromosome (Miniou et al. 1994; Okano et al.1998b; Hansen et al. 2000; Kondo et al. 2000). DNMT3Bmay therefore be adapted to methylate regions of silentchromatin.Evidence that accessory factors are also needed to en-sure appropriate methylation came initially from plants,wheretheSNF2-likeproteinDDM1wasshowntobeessential for full methylation of the Arabidopsisthalianagenome(Jeddelohetal. 1999). Anequivalentdependence is seen in animals, as mutations in humanATRX (Gibbons et al. 2000) and mouse Lsh2 genes (Den-nisetal. 2001), bothof whichencoderelativesof thechromatin-remodeling protein SNF2, have significant ef-fects on global DNA methylation patterns. Loss of LSH2protein, inparticular, matches the phenotype of theDDM1 mutation in Arabidopsis, for both mutants losemethylation of highly repetitive DNA sequences, but re-tainsomemethylationelsewhereinthegenome. Per-haps efficient global methylation of the genome requiresperturbation of chromatin structure by these chromatin-remodeling proteins so that DNMTs can gain access totheDNA. CollaborationbetweenDNMTsandfactorsthatallowthemaccesstospecializedchromosomalre-gions may be particularly important in regions that areheterochromatic and inaccessible. Although the net re-sult of these processes is apparently global genomicmethylation, the evidence for selectivity means that theword default is probably not appropriate.Targeting de novo methylation to preferredDNA sequencesAnother hypothesis to explain global methylation is thattheDNAmethylationmachineryis preferentiallyat-tractedbycertainDNAsequencesinthemammaliangenome(Turker 1999). Thepresenceof highlevelsofmethylation in DNA outside such a DNA methylationcenter couldbe explainedbyspreading intothe sur-rounding DNA. Barriers to spreading would lead to theformationof CpGislands. Ahypothetical trigger forDNAmethylationisDNAsequencerepetition, whichcan promote de novo methylation in filamentous fungiandplants under certaincircumstances (Selker 1999;Martienssen and Colot 2001). The most suggestive evi-dence in mammals concerns manipulation of transgenecopy number at a single locus in the mouse genome us-ing cre-lox technology (Garrick et al. 1998). High levelsoftransgenerepetitionwerefoundtocausesignificanttransgenesilencingandconcomitantmethylation. Theefficiencyofexpressionincreasedascopynumberwasreducedatthelocus, andthelevel of methylationde-creased. Whether repetition caused methylation directly,or indirectly as a consequence of some other event (e.g.,transcriptional silencing; see below), is not known.The clearest definition of a DNA methylation centerBird8 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from comesfromthefungusNeurospora, whereshortTpA-rich segments of DNA were found to induce methylation(Miao et al. 2000). Identification of a mammalian DNAmethylationcenterlocatedupstreamofthemousead-enine phophoribosyltransferase (APRT) gene has been re-ported (Mummaneni et al. 1993; Yates et al. 1999). Theregion contains B1 repetitive elements and attracts highlevels of de novo methylation upon transfection into em-bryonic cells, althoughthe effect is relative, becausemanyDNAsequencesaresubjecttodenovomethyl-ationinthesecells.TheAPRTmethylationcenterbe-comesmethylatedinDNMT1-deficient EScells, sup-porting the idea that it corresponds to a region that is afavorable substrate for de novo methylation (Yates et al.1999).Because the evidence suggests that replication ofmethylation patterns by DNMT1 is only partly respon-sible for maintenance methylation (see above), an attrac-tive possibility is that the features of a DNA domain thathelpmaintainitsmethylatedstatusarethesamefea-tures that promote its de novo methylation. Imprintingboxes, for example, whose differential methylation is as-sociated with genomic imprinting (Tremblay et al. 1997;Birger et al. 1999; Shemer et al. 2000), tend to retain theirmethylation levels tenaciously even when the amount ofthemaintenanceenzymeDNMT1isreduced(Beardetal. 1995). The de novo methylases DNMT3A andDNMT3B(Okanoetal. 1998a, 1999) maybeattracteddisproportionatelytothesesequences, andthisattrac-tion may also underlie the decision to methylate the boxinthefirstplace.Inotherwords,denovomethylationmay not occur once at a discrete and perhaps rather in-accessible stage of germ-cell development, but may hap-pen repeatedly (assisted by DNMT1) as embryonic cellsdivide.Unusual DNA structures and RNAi as triggersfor de novo methylationStudiesof purifiedDNMT1revealedthat theenzymepreferstomethylateunusual DNAstructuresinvitro(Smith et al. 1991; Laayoun and Smith 1995). This led tothe idea that such structures might be generated duringrecombinationbetweenrepetitive elements or duringtransposition events and directly trigger de novo meth-ylation(BestorandTycko1996). Subsequentevidence,however, does not support a role for DNMT1 in de novomethylationinvivo(Lykoet al. 1999; Howell et al.2001), andtherefore the biological significance of itspredilectionfor deformedDNAisuncertain. Thereisevidence for transfer of methylation from one copy of asequencetoasecondpreviouslyunmethylatedcopyofthe same sequence in the fungus Ascobolus (Colot et al.1996). Theprocessmightusemechanismsinvolvedinhomologous DNA recombination and may therefore in-volve deformationof DNA. Howidentical sequencessenseoneanotherandtransferepigeneticinformationremains unknown, however.Exciting recent developments inthe DNAmethyl-ation field have arisen through molecular genetic studiesofposttranscriptionalgenesilencinginplants.Double-stranded RNA directs the destruction of transcripts con-taining the same sequence, but there is compelling evi-dence that it can also direct de novo methylation of ho-mologous genomic DNA (Wassenegger et al. 1994;Bender 2001; Matzke et al. 2001). Posttranscriptionalgenesilencingbydouble-strandedRNAisprobablyanancient genome defence systembecause it occurs infungi, plants, and animals; but DNA methylation is notan obligatory accompaniment, as silencing is efficient inC. elegansinthecompleteabsenceof genomicm5C.Even in the fungus Neurospora, where transgene arraysare often methylated, DNA methylation is not requiredforposttranscriptional genesilencing(orquelling; Co-goni et al. 1996). There are also specific features of RNA-directedDNAmethylationthatmaynotoccurinani-mals; notably the occurrence of methylation at multiplenon-CpG cytosines in an affected DNA sequence tract.Although there is evidence for non-CpG methylation inES cells, most probably owing to DNMT3A, whichstrongly methylates CpA as well as CpG (Ramsahoye etal. 2000; GowherandJeltsch2001), non-CpGmethyl-ation is barely detectable in adult cells (Ramsahoye et al.2000). PlantshaveaCpGmethylationsystem, but itdoesnotappeartobeessential forRNA-directedgenesilencing (for reviews, see Wassenegger et al. 1994;Bender 2001; Matzke et al. 2001). Optimism that RNA-directeddenovomethylationwillalsoapplyinmam-mals is tempered by this sequence disparity, and by theabsence so far of a clear demonstration that mammaliandouble-stranded RNA leads to DNA methylation-medi-ated gene silencing.Transcriptionally silent chromatin as a de novomethylation targetSeveral lines of evidence suggest that DNA methylationdoesnotintervenetosilenceactivepromoters, butaf-fects genes that are already silent. It was reported manyyears ago that retroviral transcription is repressed in em-bryoniccellsat2dafterinfection, whereasdenovomethylation is delayed until 15 d (Gautsch and Wilson1983; Niwa et al. 1983). De novo methylation of proviralsequencesinembryocellsdependsonDNMT3AandDNMT3B (Okano et al. 1999), but initial retroviral shut-downoccursasusual evenwhenboththesedenovomethyltransferases are absent (Pannell et al. 2000).Clearly, de novo methylation is not required for silenc-ing in the first instance, reinforcing the view that meth-ylation is a secondary event.Methylationof genesthat arealreadysilent isalsoobserved during X chromosome inactivation in themammalianembryo. Kineticstudies showedthat thephosphoglyceratekinasegeneissilentonthemamma-lianinactiveXchromosomebeforemethylationof itsCpG-islandpromotersoccurs(Locketal.1987).Subse-quent studies of the mouse, in which the process is bestunderstood, haveestablishedthatexpressionofanon-codingchromosomal RNAfromtheXist geneontheinactive X chromosome triggers the inactivation processDNA methylation and epigenetic memoryGENES & DEVELOPMENT 9Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from in cis. Specifically, activation of the Xist gene and onsetof its late replication precede CpG-island methylation byseveral days(Keohaneet al. 1996; WutzandJaenisch2000). In other words, methylation affects the X chromo-someonwhichgenesarealreadyshutdownbyothermechanisms. Istranscriptional inertiaduringembryo-genesisthetriggerfordenovomethylation?Studiesoftheoriginof methylation-freeCpGislandsoffersomesupport for this idea. The coincidence between CpG is-lands andpromoters is striking(Bird1987), andfoot-printing shows that the 5 extremity of CpG islands of-ten corresponds to the region occupied by transcriptionfactorsinvivo(Cuadradoetal.2001).EvenwhenCpGislands areidentifiedinunusual locations, theyhaveturnedouttocorrespondtopromoters. Forexample, aCpGislandlocatedinintron2of theIgf2rgeneisanactive promoter (Wutz et al. 1997; Lyle et al. 2000), as isaCpGislandthatcoversexon2of theclassII majorhistocompatibilitygene(MacLeodetal.1998).Thepo-tential importance of promoter function in the genesis ofCpG islands is highlighted by studies in transgenic mice.CpG-island-containingtransgenesnormallyfaithfulre-producetheirmethylation-freecharacter,buttheirim-munitytomethylationislost if promoter functionisimpaired(Brandeis et al. 1994; MacLeodet al. 1994).Similarly, viral DNA integrated into ES cell genomes byhomologousrecombinationbecomesmethylatedwhenthe promoter is weakened by absence of an enhancer, butexcludes methylation when an enhancer is present(Hertz et al. 1999). A parsimonious interpretation of theresults is that failure to transcribe invites de novo meth-ylation(seeFig. 2below), althoughotherpotentialex-planations (Brandeis et al. 1994; Mummaneni et al. 1998)cannot be discounted.Thesignal for this putativegenesilence-relateddenovomethylationisunknown,butthepossibilitythatchromatin states inform the DNA methylation machin-ery is attractive (Selker 1990). The acetylation andmethylation state of nucleosomal histones is tightly cor-related with transcriptional activity (Jenuwein andAllis2001) andcouldbereadbythemethylationma-chinery, leading it to either methylate or fail to methyl-ate a particular domain. Indeed, recent work on Neuros-pora(TamaruandSelker2001) hasshownanintimatelinkbetweenhistone methylationandDNAmethyl-ationinthatfungus, asmutationofahistonemethyl-transferasethat methylatesLys9of histoneH3abol-ishedgenomicmethylation. Inmammalianandyeastsystems, histone H3Lys 9methylationis associatedwithtranscriptionallyrepressedheterochromatin(Ban-nisteretal. 2001; Nakayamaetal. 2001; Nomaetal.2001; ZhangandReinberg2001). If thedependenceofDNAmethylationonpriorhistonemethylationturnsout tobeapplicabletomammals, this wouldfurtherstrengthentheargumentthatDNAmethylationistar-geted to genes that are already silent. The nature of themolecular cues that trigger transfer of methyl groups tounmethylatedDNAshouldbeilluminatedbyongoingstudies of multiproteincomplexes that containDNAmethyltransferases (Fuks et al. 2000, 2001; Robertson etal. 2000; Bachman et al. 2001) and the identification ofgenes that modify DNA methylation patterns (Weng etal. 1995).Consequences of methylation gain: stabletranscriptional silencing of genesWhy methylate genes that are already silent? A plausibleanswer is: to silence themirrevocably. Methylationclearlycontributestothestabilityof inactivation, be-cause both X inactivation (Mohandas et al. 1981a;Graves 1982; Venolia et al. 1982) and retroviral silencing(Stewart et al. 1982; Jaenisch et al. 1985) can be relievedby treatment of somatic cells with demethylatingagents. Individuals wholackDNMT3Bshowreducedmethylation of some CpG islands on the inactive X chro-mosome and also silence X-linked genes imperfectly(Miniou et al. 1994; Hansen et al. 2000). The implicationthat irreversibilityinvolvesDNAmethylationissup-ported by the frequent reactivation of an X-linked trans-gene in mouse embryo cells and in cultured somatic cellswhen DNMT1 is absent or inhibited (Sado et al. 2000).This view is sustained by differences in the stability ofinactivity states pre- and postmethylation. For example,X inactivation caused by expression of an Xist transgeneinembryonicstemcellsisinitiallyreversedwhentheXistgeneisshutdown, butafter3d, inactivationbe-comesirreversibleandindependentof Xist(WutzandJaenisch2000).Irreversibilitymayreflectthearrivalofpromoter methylation.In artificial systems, DNA methylation represses tran-scription in a manner that depends on the location anddensityof themethyl-CpGs relativetothepromoter(BoyesandBird1992;Hsieh1994;Kassetal.1997a,b).But what genes are affected by DNA methylation-medi-ated gene silencing? Early studies relied on the use of thedemethylating drug 5-azacytidine (Jones and Taylor1980),whichwasshowntoactivategenesontheinac-tiveXinrodenthumancell hybrids(Mohandasetal.1981b; Graves1982). Morerecently, miceandmurinecell lines lacking DNMT1 (Li et al. 1992) have clarifiedtheeffectsofDNAmethylationongeneexpression.Inplacental mammals, repressionof X-linkedgenes fol-lows expression of Xist, which sets in train the inactiva-tion process, culminating in widespread methylation ofCpGislands. TheactiveXchromosome, ontheotherhand, must be protected from silencing, and this requiresrepressionof Xist andagaindepends onmethylation(PanningandJaenisch1996). Anintact DNAmethyl-ationsystemisalsoessential forgenomicimprinting,becausedeletionof Dnmt1leadstodisruptionof themonoallelic expression of several imprinted genes (Li etal. 1993).BothXinactivationandgenomicimprintinginvolvesilencing of one allele only, leaving the other unaffected.An unusual set of genes that are active in the germ line,most of which are X-linked, appears to use methylationforcompletesilencinginsomaticcells(DeSmetetal.1996, 1999). Several of thehumanandmurineMAGEgenes, for example, have CpG-island promoters that areBird10 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from methylation-freeingermcells, but aremethylatedinsomatic cells of the adult. The genes were discovered asnovelantigensintumors, wheregenomicmethylationlevelsareoftenlowandMAGE-geneCpGislandsareundermethylated.MAGEexpressioncanbeinducedbytreating nonexpressing cells with demethylating agents,supportingtheideathat methylationis animportantcomponentoftherepressionofthesegenesinsomaticcells.Transposable element silencing as a consequenceof DNA methylationAnotherwell-documentedconsequenceofDNAmeth-ylationdeficiencyistheactivationoftransposableele-ment-derived promoters. Like much of the mammaliangenome, transposable element-related sequences areheavilymethylatedandtranscriptionallysilent inso-matic cells. Mouse cells, for example, normally represstranscription of intracisternal A particle (IAP) elements,whichconstituteahomogeneousandtranspositionallyactive family of elements. In embryos lacking DNMT1,transcriptionofIAPelementsismassivelyinduced,ar-guing that methylation is normally responsible for theirrepression(Walshet al. 1998). Derepressionof LINE(Woodcocketal. 1997) andSINE(Liuetal. 1994) pro-motersinthehumangenomealsooccurswhenDNAmethylation is reduced. The most abundant SINE in thehuman genome is the Alu family, which consists of sev-eral hundred thousand elements (Smit 1999). Only a tinyminority of elements are capable of transposition (95% of animal species (Tweedie et al. 1997).Colonization of the genome by transposable elementscan only occur in the germ-cell lineage because somatictranspositioneventsleavenoheritabletrace.Paradoxi-cally, transposableelementsareoftentranscriptionallyactive and unmethylated in germ cells and totipotent EScells(forreview, seeBird1997). IAPelements, forex-ample, becomeunmethylatedduringthegonial prolif-erationphase, whenprimordial germcell number in-creases from 75 to 25,000 (Walsh et al. 1998). The fre-quent absence of DNA methylation in germ cells, whentranspositioncando long-termdamage (Maliket al.1999), contrasts with its repressive presence in somaticcells, where transposition would be an evolutionary deadend. It is too early to discount the possibility that trans-posonpromoters, mostof whichbelongtodegenerateelementsthat areincapableof transposition, must besilenced to suppress transcriptional noise.Mechanisms of DNA methylation-mediatedtranscriptional repressionWhydoes DNAmethylationinterferewithtranscrip-tion? Two modes of repression can be envisaged, and it islikely that both are biologically relevant. The first modeinvolves direct interference of the methyl group in bind-ingof aproteintoitscognateDNAsequence(Fig. 1).Manyfactors are knowntobindCpG-containing se-quences, and some of these fail to bind when the CpG ismethylated. Strong evidence for involvement of thismechanism in gene regulation comes from studies of therole of the CTCF protein in imprinting at the H19/Igf2locus in mice (Bell and Felsenfeld 2000; Hark et al. 2000;Szabo et al. 2000; Holmgren et al. 2001). CTCF is asso-ciated with transcriptional domain boundaries (Bell et al.1999) and can insulate a promoter from the influence ofremoteenhancers. ThematernallyderivedcopyoftheIgf2 gene is silent owing to the binding of CTCF betweenDNA methylation and epigenetic memoryGENES & DEVELOPMENT 11Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from its promoter and a downstream enhancer. At the pater-nal locus, however, these CpG-richbinding sites aremethylated, preventingCTCFbindingandtherebyal-lowing the downstream enhancer to activate Igf2 expres-sion. Although there is evidence that H19/Igf2 imprint-inginvolvesadditional processes(Ferguson-SmithandSurani 2001), theroleof CTCFrepresentsoneof theclearest examples of transcriptional regulation by DNAmethylation.The second mode of repression is opposite to the first,as it involves proteins that are attracted to, rather thanrepelled by, methyl-CpG (Fig. 1). A family of five methyl-CpG-bindingproteinshasbeencharacterizedthateachcontains aregioncloselyrelatedtothemethyl-CpG-binding domain (MBD) of MeCP2 (Nan et al. 1993, 1997;Cross et al. 1997; Hendrich and Bird 1998). Four of theseproteinsMBD1, MBD2, MBD3, and MeCP2havebeen implicated in methylation-dependent repression oftranscription (for review, see Bird and Wolffe 1999). Anunrelated protein Kaiso has also recently been shown tobind methylated DNA and bring about methylation-de-pendent repression in model systems (Prokhortchouk etal. 2001). In vitro, Kaiso requires a 5 m5CGm5CG motif,and binding is highly dependent on the presence of meth-ylation. Thepresenceofmultiplemethyl-CpG-bindingproteins withrepressiveproperties supports theargu-ment that these maybe important mediators of themethylationsignal, but their involvement inspecificprocesses that require transduction of the DNA methyl-ation signal has yet to be shown. Targeted mutation ofthe gene for MeCP2 is, however, associated with neuro-logical dysfunctioninhumans andmice(Amir et al.1999; Chen et al. 2001; Guy et al. 2001), and mutation ofthe mouse Mbd2 gene leads to a maternal behavior de-fect (Hendrich et al. 2001).Excluding DNA methylation by denying accessTheprecedingdiscussionhasconsideredsomemecha-nistic aspects of de novo DNA methylation and its bio-logical consequences. Although methylation affectsmost of the mammalian genome, it is conspicuously ab-sent fromcertainregions. Ways inwhichthesenon-methylated domains may arise will now be considered.Asimplemechanismforcreatinganonmethylateddo-main within an otherwise densely methylated genome isto mask a stretch of DNA by protein binding. The DNA-binding protein would accomplish this passive demeth-ylationby, for example, stericallyexcluding DNMTs(Bird 1986). The feasibility of this mechanism has beenverifiedusinganartificiallymethylatedepisomecon-taining EBNA1 or lac repressor binding sites (Hsieh1999a; Linetal. 2000). TheideathatCpGislandsareentirely attributable to exclusion of this kind is in doubt,however, as in vivo footprinting and nuclease accessibil-itystudiesshowCpGislandstobemoreaccessibletoproteins(nucleases) thanbulkgenomicDNA, notless(TaziandBird1990).Ofcourse,itispossiblethatpro-tection is only present at the transient embryonic stagewhen mammalian de novo methylation occurs and hastherefore escaped detection. A protein that is reported tobind unmethylated CpGs might be a candidate CpG-is-land protector (Voo et al. 2000).Immunity to DNA methylation causedby transcriptionally active chromatin:the origin of unmethylated CpG islandsManyoftheknownbiologicaleffectsofDNAmethyl-ation are associated with CpG islands. It has been arguedabove that their methylation in the early embryo followsFigure1. Mechanismsoftranscriptionalrepression by DNA methylation. A stretchof nucleosomal DNAis shownwithallCpGsmethylated(redcircles). Belowthediagram is a transcription factor that is un-abletobinditsrecognitionsitewhenamethylatedCpGiswithinit.Manytran-scriptionfactorsarerepelledbymethyl-ation, including the boundary elementprotein CTCF (see text). Above the line areprotein complexes that can be attracted bymethylation, includingthemethyl-CpG-binding proteinMeCP2(plus the Sin3Ahistone deacetylase complex), the MeCP1complex comprising MBD2 plus theNuRDcorepressorcomplex, andtheun-characterized MBD1 and Kaiso complexes.MeCP2 and MBD1 are chromosome-boundproteins, whereasMeCP1maybeless tightly bound. Kaiso has not yet beenshown to associate with methylated sitesin vivo.Bird12 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from silencing events that are likely to be DNA methylation-independent. If transcriptional silence indeedtriggersDNAmethylation, thenthecorollaryisthatpromoteractivityearlyindevelopmentshouldcreateamethyl-ation-free CpG island (Fig. 2). In other words, unmethyl-ated CpG islands might be footprints of embryonic pro-moteractivity. AnobviouspredictionofthismodelisthatallunmethylatedCpGislands, includingthoseatpromotersofhighlytissue-specificallyexpressedgenes,should contain promoters that function during early de-velopment whenthe methylationmemorysystemismost active. Although very limited, the data so far favorthis theory, because a CpG-island promoter whose prod-uctRNAisnotexpectedtooccurintheearlyembryo(-globin) is nevertheless expressed, whereas transcriptsfroma CpG-deficient promoter (-globin) are not de-tected (Daniels et al. 1997). Similarly, expression of the68kneurofilamentgene, whichhasaCpG-islandpro-moter, wasdetectedinEScells, butopsinandcaseingenes, whichareCpG-deficient genes, appearedtobesilent (MacLeod et al. 1998).Whyshouldactivepromoterregionsescapedenovomethylation? CpG islands often colocalize with originsof DNA replication (Delgado et al. 1998), and, accordingto one speculation, an early replication intermediate cre-ates the DNA methylation-free footprint (Antequera andBird1999). Amoredirect(butnotmutuallyexclusive)mechanismwould involve the sensing of chromatinstatesbythedenovomethylationsystemasdiscussedabove. WhereashistoneH3tailsmodifiedbymethyl-ationonLys9mightrecruitDNAmethyltransferases(Tamaru and Selker 2001), modifications associated withactivechromatin, suchasacetylationof H3or H4ormethylation of Lys 4 of histone H3, may actively excludethese enzymes. Biochemical evidence addressing this is-sue is eagerly awaited.Active demethylation of DNAProtectionagainstdenovomethylationbyboundpro-teinsorchromatincanensurethatDNAmethylationneverreachesaDNAsequencedomain.UnmethylatedFigure 2. A hypothetical scenario relating embryonic transcriptional activity to DNA methylation status in mammals. Starting froma notional transcription ground state, embryonic demethylation leads to substitution of methylated sites (red circles) by nonmethyl-ated sites (yellow circles). Two alternative fates are then envisaged: either transcription persists leading to restoration of the unmeth-ylated CpG island (bracket) flanked by methylated non-island-flanking DNA (pink arrows); or transcription is extinguished by othermechanisms in the embryo and this invites de novo methylation of the CpG island and its flanks. In this way the activity of embryonicpromoters is imprinted for the duration of that somatic lifetime.DNA methylation and epigenetic memoryGENES & DEVELOPMENT 13Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from domains could also arise by actively removing the modi-fication from DNA. This so-called active demethylationcould be accomplished either by the thermodynamicallyunfavorablebreakageof thecarboncarbonbondthatlinks the pyrimidine to its methyl group, or by a repair-likeprocessthat excisesthem5Cbaseor nucleoside,leadingtoitsreplacement withC(Kresset al. 2001).Several laboratories have striven to isolate demethylaseenzymes(forreview,seeWolffeetal.1999).Themostimpressivecatalyticactivitywasshownbyafractionderived from human cells (Ramchandani et al. 1999) thatwassubsequentlyidentifiedasMBD2(Bhattacharyaetal. 1999). Theexpressedproteinreportedlyshowedro-bust demethylation in vitro in the absence of added co-factors and released methanol as a by-product. AttemptstoobservethispropertyofMBD2inotherlaboratorieshave not been successful.Acellextractshowingdemethylaseactivitywasde-tectedinratmyoblastcells(Weissetal. 1996). Initialindications that the reaction was RNA-dependent werenotsustaineduponfurtherenrichmentof theactivity(Swisher et al. 1998). An RNA-containing demethylatingcomplex was, however, reported in chicken cells (Jost etal. 1997, 1999). These investigators searched for proteinswithm5C-DNAglycosylaseactivityandidentifiedthepreviously known thymine DNA glycosylase TDG,which can remove the pyrimidine base from T:G or U:Gmismatches (Zhuet al. 2000b). MBD4, anunrelatedDNA glycosylase with similar properties, was also foundtobeactiveagainstm5C:Gpairs(Zhuetal.2000a).Astheefficiencyofthesereactionswasmuchlowerthanthat seenwiththecognatemismatchedsubstrates, itmightbearguedthatthem5Cglycosylaseactivityrep-resentsaminor sidereactionof littleinvivosignifi-cance. Set against this is evidence that stable expressionofachickenTDGresultsinsignificantactivationandconcomitant demethylation of a reporter gene driven bya methylated ecdysone-retinoic acid-responsive pro-moter (Zhuet al. 2001). Thenormallysilent reportercould also be activated by demethylation with 5-azacyti-dine, but generalized demethylation of the genome wasnot observed in TDG transfected cells. Previous studiesshowedanassociationbetweenretinoidreceptorsandTDG, and implicated TDG in transcriptional activation(Umet al. 1998). Timewill tell if thestimulationofretinoid-responsivepromotersbyTDGdependsonitsdemethylating activity.Theneedtoisolatedemethylatingenzymeshasbe-come more acute with the finding that the paternal ge-nomeissubjecttoactivedemethylationsoonafterfer-tilization (Mayer et al. 2000; Oswald et al. 2000). Similarprocesses have been reported in pig and bovine embryos(Bourchis et al. 2001; Kang et al. 2001a,b). This dramaticillustration of methylation loss in the absence of DNAreplicationraisesquestionsabouttheprevalenceofde-methylationbythismechanism.Interestingly,thema-ternal genome, whichalsodemethylates duringearlymouse development, does so by a different mechanism:passivefailuretomethylateprogenystands(Rougieretal. 1998). Whyshouldmaternal andpaternal genomeschoose such different routes to the same end? An intrigu-ing possibility is that the parental struggle over maternalresources for the embryo that is thought to underlie ge-nomicimprinting(MooreandHaig1991) maybein-volved. The oocyte may be equipped to directly disarmthespermgenomeofmethylationimprintsthatmightoverexploit maternal resources (Reik and Walter 2001). Itisevenpossiblethat thepaternal genome, indelayedretaliation, may organize a campaign of interferencewiththemaintenancemethylation(e.g., byexportingmaternal DNMTs to the cytoplasm). The extraordinaryneed for an oocyte variant of DNMT1 to translocate intothenucleusfor onlyonecleavagecycle(thedoublingfrom8to16cells; Howelletal. 2001) couldrepresentmaternal measures to compensate for interference of thiskind.Consequences of methylation loss: gene activationduring developmentInterest in DNA methylation has long been fueled by thenotion that strategic loss of methyl groups during devel-opment could lead to activation of specific genes in theappropriate lineage. As has been emphasized (Walsh andBestor1999),muchoftheevidenceforthisscenarioisinconclusive, but recent studies have revived the idea. Inthe frog, gene expression is suppressed from fertilizationuntil the mid-blastula stage (5000 cells), at which timetranscription is activated. Inhibition of DNMT1 using anantisense strategy caused reduced methylation and pre-matureactivationof certaingenes, suggestingadirectrole for DNA methylation in maintaining their early si-lence prior to the blastula stage (Stancheva and Meehan2000). Deletion of the Dnmt1 gene in cultured somaticcellsofthemousealsocausedwidespreadgeneactiva-tion (Jackson-Grusby et al. 2001). About 10% of all genesdetectedusing microarraytechnologywere activated,whereas only 1%2% were down-regulated. Some of theup-regulated genes are normally only expressed in termi-nally differentiated cells. These findings raise the possi-bility that DNA methylation contributes to silencing oftissue-specific genes innonexpressing cells, andtheyconfirm DNA methylation as a global repressor of geneexpression. The scenario has been modeled using an ar-tificial construct that containedaDNAsequenceca-pableofexcludingmethylationlocallyduringearlyde-velopment (Siegfriedet al. 1999). Whenthissequencewas present, the reporter gene stayed unmethylated dur-ingdevelopment, andwidespreadexpressionoccurred.Deletion of the element in situ in the early embryo led tomethylationof thereporter geneandconcomitant si-lencing in several adult tissues.Asubtlepotential rolefor lossof methylationat aspecific gene has been reported for the rat tyrosine ami-notransferase gene (Thomassin et al. 2001). When amethylatedformofthisgeneisinducedbyglucocorti-coids,delayeddemethylationoccursatspecificsitesinanenhancerandadditionalDNA-associatedfactorsaresubsequentlyrecruited.Demethylation(whetheractiveor passive is not known) persists after the wave of TATBird14 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from expressionhassubsided, andreinductionof thesilentgene bya further hormone treatment is significantlystronger as a result. This system provides a model for aDNAmethylation-mediatedmemoryof thefirst hor-moneinduction(Kresset al. 2001). Itssignificanceissomewhat less certain in normal development, however,becausedemethylationofthesesitesoccursbeforethegene becomes hormone-inducible.Thereissuggestiveevidencethat programmedrear-rangementof theimmunoglobulingenesduringB-celldevelopment mayinvolve DNAmethylation(Mosto-slavskyetal. 1998). Demethylationof oneof thetwoparentally derived alleles of the kappa light chain gene isobserved in small pre-B cells, and there is evidence thatthisearlylossof methylationpredisposestheaffectedallele to rearrangement. By precluding rearrangement ofoneallele, differential DNAmethylationmayhelptoexplainallelicexclusionatthekappachainlocus.Itisnot certainwhether transcriptional regulationper seplaysarole, althoughtheprocessisdependentontheintronic and 3 kappa gene enhancers.Loss of genome integrity as a consequence of DNAmethylation loss?Early studies with the DNA-methylation inhibitor5-azacytidinerevealedbizarrechromosomal rearrange-ments intreatedculturedcells (Viegas-Pequignot andDutrillaux1976).Althoughthesefindingsmightbeat-tributed to the effects of reduced DNA methylation, theycouldalsobearesultofthechemicalreactivityoftheincorporated base analog, inparticular, its ability tocross-link proteins to DNA (Juttermann et al. 1994). Theformer possibility is supported somewhat by the findingthat mitogen-stimulated lymphocytes frompatientswith mutations in DNMT3B show very similar chromo-somerearrangements, involvingcoalescenceofcentro-meric regions that contain methylation-deficient repeti-tivesequences(Jeanpierreetal. 1993; Xuetal. 1999).Oddly, therearrangementsarenotseenincellsofthepatients, despitesimilar hypomethylationof thesere-gions.Itseemsthatlossofgenomicintegrityisnotanobligatory consequence of hypomethylation of juxtacen-tromeric repeat elements.Atafinerlevel, twolaboratorieshaveexaminedtheeffectsof greatlyreducedDNAmethylationlevelsonmutationrates inmouseembryonicstemcells, withsomewhatdifferingresults.Inonestudy,themutationrate at two endogenous loci was found to have increased10-foldcomparedtothesameloci inwild-typecells(Chenetal.1998),suggestingthatlackofmethylationpredisposed to aberrant recombination events. A secondstudyexaminedtransgenesofexogenousoriginusingaselection system to detect mutations (Chan et al. 2001).This allowed screening of large numbers of mutations attwoindependentloci, butneitherpointmutationsnorgenomic rearrangements were increasedunder condi-tionsoflimitingDNAmethylation.Infact,mutationsappeared to be suppressed by genomic hypomethylation.These inconsistencies raise questions about the pro-posed relationship between genome integrity and DNAmethylationthatwill needtobeaddressedbyfurtherresearch.Developmental memory: DNA methylationand Polycomb/trithorax complexesas interchangeable systemsThe foregoing discussion has highlighted features of theDNAmethylationsysteminmammalsthat resembleanotherestablishedsystemof cellularmemory: Pc-G/trx. The final section of the review will compare the twosystems. The credentials of Pc-G/trx protein complexesas an epigenetic system in development are compelling(Paroet al. 1998; Pirrotta1999; FrancisandKingston2001). This multiprotein assembly is targeted to specificregionsof thegenomewhereiteffectivelyfreezestheembryonicexpressionstatusof agene, beitactiveorinactive, and propagates that state stably through devel-opment. Elegant experiments with model gene con-structs have shown that brief activation (or inactivation)of a promoter during early Drosophila development leadstostableactivity(orinactivity) thereafter(CavalliandParo1998, 1999; Pouxet al. 2001). Attemptstoalterexpression at most other stages of development were un-successful, indicating that there is a window of time dur-ingwhichtranscriptionpatternscanbecommittedtodevelopmental memory. The Pc-G/trx system is reactiverather than proactive, as the setting up of segment-spe-cificpatternsofactivegenesisnotdisruptedbymuta-tions in Pc-G group genes. Only the capacity to sustainthe patterns is lost in the mutants. This ability to copyand propagate the expression patterns without influenc-ing or perturbing them makes this a subtle and flexiblememorysystem. Littleisknown, however, about themechanisms responsible for the heritability of Pc-G/trx.What do Pc-G/trx and DNA methylation in mammalshave in common? First, both systems are able to represstranscription in a heritable manner. Second, both appearto be reactive in that they lock in expression states thatthey played no part in setting up (e.g., DNA methylationinviralgenomesilencingandCpG-islandmethylationon the X chromosome). Third, both are activated prima-rilyduringadiscretewindowoftimeinearlydevelop-ment. Thus, likePc-G/trx, DNAmethylationhastheproperties of a developmental memory.What is memorized by DNA methylation? Arguably,its major role is to stably demarkate by its absence a setof embryonically active promoters, namely, CpGis-lands, so that they remain potentially active throughoutdevelopment and adulthood. At the same time, regionsdevoid of promoter activity in the embryo become meth-ylatedandcarrythis repressive influence withthemthroughdevelopment.Thedegreeofrepressionmaybeweak or strong depending on the density of methylation(BoyesandBird1992; Hsieh1994). Thus, CpGislandsthataresilencedbyothermechanismsduringembryo-genesiswouldacquiredensemethylationleadingtoir-reversible silencing. When, however, the density ofmethylated CpGs is low, as it is in most of the genome,DNA methylation and epigenetic memoryGENES & DEVELOPMENT 15Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from repression is likely to be weak and may be overcome bythe presence of strong activators. Weak repression of tis-sue-specificgenes(e.g., -globin) thatareembeddedinregionsof low-densitymethylationmaycontributetotheir silence in inappropriate tissues.It is proposed here that DNA methylation and Pc-G/trxarealternativesystemsofcellularmemorythatareinterchangeableover evolutionarytime. InC. elegansand Drosophila, for example, Pc-G group proteins(Birchleretal.2000)havebeenimplicatedinsilencingof repetitive-element transcription in somatic cells,whereas DNA methylation may play this role in mam-mals (Yoder et al. 1997). The involvement of DNA meth-ylation in genome defence may, therefore, be to memo-rize the silent state of elements imposed by primary ge-nomedefencesystems.Themoststrikingevidenceforinterchangeability is the finding that X chromosome in-activationinextraembryonictissuesof themousede-pendsonthepolycombgroupproteinEed. Lossoftheeed gene leads to reactivation of the inactive X in extra-embryonic tissue, but has no effect in somatic cell types(Wangetal.2001).Incontrast,Dnmt1mutationsreac-tivate the inactive X of the embryo proper, but not theextraembryonic inactive X (Sado et al. 2000). The findingthat certain CpG islands on the inactive X chromosomeare methylated in somatic cells but not in extraembry-onictissues (Iidaet al. 1994) fits withtheviewthatmethylation replaces Pc-G in somatic tissues. Therefore,evenwithinasinglespecies, it appearsthat differenttissuesemployPc-G/trxandDNAmethylationinter-changeably. From an evolutionary perspective, it is pos-siblethatvaryingdegreesoffunctionalsubstitutionbyPc-G (or vice versa) can explain the dramatic differencesbetween DNA methylation levels across animal species.Concluding remarksOur understanding of the relationshipbetweenDNAmethylation and transcriptional control is growing fast,but is still far fromcomplete. Ongoing biochemicalanalysisof thegrowingnumberof componentsof theDNAmethylationsystem(andtheirpartners),coupledwithgeneticapproaches, will strengthenthelinksbe-tween DNAmethylation and mainstreamtranscrip-tional mechanisms. Regulationof gene expressioniscomplex (Lemon and Tjian 2000), and the emerging evi-dence hints that the roles of DNA methylation will betoo. It maybeunrealistictoexpect that anyunifiedtheory will encompass all the biological consequences ofDNA methylation.Least understood are the mechanisms by which meth-ylationpatternsaregenerated.Followingconsiderationof the criteria for attracting and repelling DNA methyl-ation, this review has entertained the possibility that aprimaryfunctionof denovoDNAmethylationis tomemorize patterns of embryonic gene activity, creatingCpG islands that are competent for transcriptionthroughout development, or their antithesis, regionsthataremethylatedandtranscriptionallyincompetent.The idea depends on evidence that methylation does notintervene to silence genes that are actively transcribed,but only affects genes that have already been shut downby other means. There is reason to believe that transcrip-tional activity may somehow imprint the methylation-freestatusof CpGislands. Theinvolvement of DNAmethylationininactivationof transposable elementscould likewise be due to its capacity for stabilizing thetranscriptional shutdownorganizedbyother systems.Parallels betweenthese emerging attributes of DNAmethylation and the Pc-G system in Drosophila suggestthatbotharemechanismsforsensingandpropagatingcellular memory.AcknowledgmentsI amgrateful toEricSelker, BernardRamsahoye, HelleJr-gensen, Brian Hendrich, and Catherine Millar for comments onthe manuscript. Research by A.B. is supported by The WellcomeTrust.ReferencesAmir, R.E., Van den Veyver, I.B., Wan, M., Tran, C.Q., Francke,U., and Zoghbi, H.Y. 1999. Rett syndrome is caused by mu-tationsinX-linkedMECP2,encodingmethyl-CpG-bindingprotein 2. Nat. Genet. 23: 185188.Antequera, F. andBird, A. 1993. NumberofCpGislandsandgenes in human and mouse. Proc. Natl. Acad. Sci.90: 1199511999..1999.CpGislandsasgenomicfootprintsofpromotersthat are associated with replication origins. Curr. Biol.9: R661R667.Antequera, F., Boyes, J., and Bird, A. 1990. High levels of de novomethylation and altered chromatin structure at CpG islandsin cell lines. Cell 62: 503514.Bachman, K.E., Rountree, M.R., and Baylin, S.B. 2001. Dnmt3aand Dnmt3b are transcriptional repressors that exhibituniquelocalizationpropertiestoheterochromatin. J. Biol.Chem. 276: 3228232287.Bannister, A.J., Zegerman, P., Partridge, J.F., Miska, E.A., Thom-as, J.O., Allshire, R.C., andKouzarides, T. 2001. Selectiverecognition of methylated lysine 9 on histone H3 by the HP1chromo domain. Nature 410: 120124.Baylin, S.B. and Herman, J.G. 2000. DNA hypermethylation intumorigenesis: Epigenetics joins genetics. Trends Genet.16: 168174.Beard, C., Li, E., andJaenisch, R. 1995. Lossof methylationactivates Xist in somatic but not in embryonic cells. Genes& Dev. 9: 23252334.Bell, A.C. and Felsenfeld, G. 2000. Methylation of a CTCF-de-pendent boundary controls imprinted expression of the Igf2gene. Nature 405: 482485.Bell, A.C., West, A.G., andFelsenfeld, G. 1999. TheproteinCTCF is required for the enhancer blocking activity of ver-tebrate insulators. Cell 98: 387396.Bender, J. 2001. A vicious cycle. RNA silencing and DNA meth-ylation in plants. Cell 106: 129132.Bestor, T.H. 1992. Activationof mammalianDNAmethyl-transferasebycleavageofaZnbindingregulatorydomain.EMBO J. 11: 26112617.. 2000. The DNA methyltransferases of mammals. Hum.Mol. Genet. 9: 23952402.Bestor, T. and Tycko, B. 1996. Creation of genomic methylationBird16 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from patterns. Nat. Genet. 12: 363367.Bhattacharya, S.K., Ramchandani, S., Cervoni, N., and Szyf, M.1999.Amammalianproteinwithspecificdemethylaseac-tivity for mCpG DNA. Nature 397: 579583.Birchler, J.A., Bhadra, M.P., and Bhadra, U. 2000. Making noiseabout silence: Repression of repeated genes in animals. Curr.Opin. Genet. Dev. 10: 211216.Bird, A.P. 1986. CpG-richislands andthefunctionof DNAmethylation. Nature 321: 209213.. 1987. CpGislandsasgenemarkersinthevertebratenucleus. Trends Genet. 3: 342347.. 1995. Gene number, noise reductionand biologicalcomplexity. Trends Genet. 11: 94100.. 1997. Does DNA methylation control transposition ofselfish elements in the germline. Trends Genet. 13: 469470.Bird, A. andWolffe, A.P. 1999. Methylation-inducedrepres-sionBelts, braces and chromatin. Cell 99: 451454.Bird, A.P., Taggart, M.H., and Smith, B.A. 1979. Methylated andunmethylatedDNAcompartments inthe sea urchinge-nome. Cell 17: 889901.Bird, A.P., Taggart, M.H., Nicholls, R.D., and Higgs, D.R. 1987.Non-methylated CpG-rich islands at the human -globin lo-cus: Implications for evolution of the-globin pseudogene.EMBO J. 6: 9991004.Birger, Y., Shemer, R., Perk, J., and Razin, A. 1999. The imprint-ing box of the mouse Igf2r gene. Nature 397: 8488.Bourchis,D.,LeBourhis,D.,Patin,D.,Niveleau,A.,Comiz-zoli,P.,Renard,J.,andViegas-Pequignot,E.2001.Delayedand incomplete reprogramming of chromosome methylationpatterns in bovine cloned embryos. Curr. Biol.11: 15421546.Boyes, J. and Bird, A. 1992. Repression of genes by DNA meth-ylation depends on CpG density and promoter strength: Evi-dencefor involvement of amethyl-CpGbindingprotein.EMBO J. 11: 327333.Brandeis, M., Frank, D., Keshet, I., Siegried, Z., Mendelsohn, M.,Nemes, A., Temper, V., Razin, A., and Cedar, H. 1994. Sp1elementsprotectaCpGislandfromdenovomethylation.Nature 371: 435438.Cavalli, G. andParo, R. 1998. TheDrosophilaFab-7chromo-somal element conveys epigenetic inheritance during mito-sis and meiosis. Cell 93: 505518..1999.Epigeneticinheritanceofactivechromatinafterremoval of the main transactivator. Science 286: 955958.Chan,M.F.,vanAmerongen,R.,Nijjar,T.,Cuppen,E.,Jones,P.A., and Laird, P.W. 2001. Reduced rates of gene loss, genesilencing, and gene mutation in dnmt1-deficient embryonicstem cells. Mol. Cell. Biol. 21: 75877600.Chen,R.Z.,Pettersson,U.,Beard,C.,Jackson-Grusby,L.,andJaenisch, R. 1998. DNA hypomethylation leads to elevatedmutation rates. Nature 395: 8993.Chen, R.Z., Akbarian, S., Tudor, M., and Jaenisch, R. 2001. De-ficiencyofmethyl-CpGbindingprotein-2inCNSneuronsresults in a Rett-like phenotype in mice. Nat. Genet.27: 327331.Chu, W.M., Ballard, R., Carpick, B.W., Williams, B.R., andSchmid, C.W. 1998. Potential Alufunction: Regulationofthe activity of double-stranded RNA-activated kinase PKR.Mol. Cell. Biol. 18: 5868.Cogoni, C., Irelan, J.T., Schumacher, M., Schmidhauser, T.J.,Selker, E.U., andMacino, G. 1996. Transgenesilencingofthe al-1 gene in vegetative cells of Neurospora is mediated bya cytoplasmic effector and does not depend on DNADNAinteractions or DNA methylation. EMBO J. 15: 31533163.Colot, V. andRossignol, J.L. 1999. EukaryoticDNAmethyl-ation as an evolutionary device. BioEssays 21: 402411.Colot, V., Maloisel, I.L., and Rossignol, J.-L. 1996. Interchromo-somal transfer of epigenetic states in Ascobolus: Transfer ofDNA methylation is mechanistically related to homologousrecombination. Cell 86: 855864.Cosma,M.P.,Tanaka,T.,andNasmyth,K.1999.Orderedre-cruitment of transcription and chromatin remodeling factorstoa cell cycle- anddevelopmentally-regulatedpromoter.Cell 97: 299311.Costello, J.F. and Plass, C. 2001. Methylation matters. J. Med.Genet. 38: 285303.Cross, S.H., Meehan, R.R., Nan, X., and Bird, A. 1997. A com-ponent of the transcriptional repressor MeCP1 is related tomammalian DNA methyltransferase and trithorax-like pro-tein. Nat. Genet. 16: 256259.Cuadrado,M.,Sacristan,M.,andAntequera,F.2001.Species-specificorganizationofCpGislandpromotersatmamma-lian homologous genes. EMBO Rep. 2: 586592.Daniels,R.,Lowell,S.,Bolton,V.,andMonk,M.1997.Tran-scription of tissue-specific genes in human preimplantationembryos. Hum. Reprod. 12: 22512256.Delgado, S., Gomez, M., Bird, A., and Antequera, F. 1998. Ini-tiationofDNAreplicationatCpGislandsinmammalianchromosomes. EMBO J. 17: 24262435.Dennis, K., Fan, T., Geiman, T., Yan, Q., and Muegge, K. 2001.Lsh, a member of the SNF2 family, is required for genome-wide methylation. Genes & Dev. 15: 29402944.De Smet, C., De Backer, O., Faraoni, I., Lurquin, C., Brasseur, F.,and Boon, T. 1996. The activation of human gene MAGE-1intumor cellsiscorrelatedwithgenome-widedemethyl-ation. Proc. Natl. Acad. Sci. 93: 71497153.De Smet, C., Lurquin, C., Lethe, B., Martelange, V., and Boon, T.1999. DNAmethylation is the primary silencing mechanismfor a set of germ line- and tumor-specific genes with a CpG-rich promoter. Mol. Cell. Biol. 11: 73277335.Ehrlich, M. 1982. Amount and distribution of 5-methycytosineinhumanDNAfromdifferent types of tissues or cells.Nucleic Acids Res. 10: 27092721.Ferguson-Smith, A.C. andSurani, M.A. 2001. Imprintingandtheepigeneticasymmetrybetweenparentalgenomes. Sci-ence 293: 10861089.Finnegan, E.J., Peacock, W.J., and Dennis, E.S. 2000. DNAmeth-ylation, a key regulator of plant development and other pro-cesses. Curr. Opin. Genet. Dev. 10: 217223.Flatau,E.,Gonzales,F.A.,Michalowsky,L.A.,andJones,P.A.1984. DNA methylation in 5-Aza-2-deoxycytidine-resistantvariants of C3H 10T1/2 C18 cells. Mol. Cell. Biol. 4: 20982102.Francis, N.J. and Kingston, R.E. 2001. Mechanisms of transcrip-tional memory. Nat. Rev. Mol. Cell. Biol. 2: 409421.Fuks, F., Burgers, W.A., Brehm, A., Hughes-Davies, L., andKouzarides, T. 2000. DNA methyltransferase Dnmt1 associ-ateswithhistonedeacetylaseactivity.Nat.Genet.24: 8891.Fuks, F., Burgers, W.A., Godin, N., Kasai, M., and Kouzarides, T.2001. Dnmt3abindsdeacetylasesandisrecruitedbyase-quence-specificrepressortosilencetranscription.EMBOJ.20: 25362544.Garrick, D., Fiering, S., Martin, D.I., andWhitelaw, E. 1998.Repeat-inducedgene silencing inmammals. Nat. Genet.18: 5659.Gautsch, J.W. and Wilson, M.C. 1983. Delayed de novo meth-ylationinteratocarcinomacellssuggestsadditionaltissue-specific mechanisms for controlling gene expression. Nature301: 3237.Gibbons, R.J., McDowell, T.L., Raman, S., ORourke, D.M.,Garrick, D., Ayyub, H., and Higgs, D.R. 2000. Mutations inDNA methylation and epigenetic memoryGENES & DEVELOPMENT 17Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from ATRX, encoding a SWI/SNF-like protein, cause diversechangesinthepatternof DNAmethylation. Nat. Genet.24: 368371.Gowher, H. andJeltsch, A. 2001. Enzymaticpropertiesofre-combinant Dnmt3aDNAmethyltransferasefrommouse:The enzyme modifies DNA in a non-processive manner andalsomethylatesnon-CpG[correctionof non-CpA] sites. J.Mol. Biol. 309: 12011208.Gowher, H., Leismann, O., and Jeltsch, A. 2000. DNA of Dro-sophilamelanogastercontains5-methylcytosine. EMBOJ.19: 69186923.Graves, J.A. 1982. 5-Azacytidine-induced re-expression of alle-lesontheinactiveXchromosomeinahybridmousecellline. Exp. Cell Res. 141: 99105.Gurdon, J.B. 1999. Geneticreprogrammingfollowingnucleartransplantation in Amphibia. Semin. Cell Dev. Biol.10: 239243.Guy, J., Hendrich, B., Holmes, M., Martin, J.E., andBird, A.2001. AmouseMecp2-null mutationcauses neurologicalsymptoms that mimic Rett syndrome. Nat. Genet. (in press).Hansen, R.S., Stoger, R., Wijmenga, C., Stanek, A.M., Canfield,T.K., Luo, P., Matarazzo, M.R., DEsposito, M., Feil, R.,Gimelli,G.,etal.2000.EscapefromgenesilencinginICFsyndrome: Evidence for advanced replication time as a majordeterminant. Hum. Mol. Genet. 9: 25752587.Hark, A.T., Schoenherr, C.J., Katz, D.J., Ingram, R.S., Levorse,J.M., andTilghman, S.M. 2000. CTCFmediates methyl-ation-sensitiveenhancer-blockingactivityat theH19/Igf2locus. Nature 405: 486489.Harris, M. 1982. Induction of thymidine kinase in enzyme-de-ficient Chinese hamster cells. Cell 29: 483492.Hendrich, B. and Bird, A. 1998. Identification and characteriza-tionof afamilyof mammalianmethyl-CpGbindingpro-teins. Mol. Cell. Biol. 18: 65386547.Hendrich, B., Guy, J., Ramsahoye, B., Wilson, V.A., and Bird, A.2001. Closely related proteins Mbd2 and Mbd3 play distinc-tivebutinteractingrolesinmousedevelopment.Genes&Dev. 15: 16131618.Hertz, J.M., Schell, G., and Doerfler, W. 1999. Factors affectingdenovomethylationofforeignDNAinmouseembryonicstem cells. J. Biol. Chem. 274: 2423224240.Holliday, R. andPugh, J.E. 1975. DNAmodificationmecha-nisms and gene activity during development. Science186: 226232.Holmgren, C., Kanduri, C., Dell, G., Ward, A., Mukhopadhya,R., Kanduri, M., Lobanenkov, V., and Ohlsson, R. 2001. CpGmethylationregulates the Igf2/H19insulator. Curr. Biol.11: 11281130.Howell, C.Y., Bestor, T.H., Ding, F., Latham, K.E., Mertineit, C.,Trasler, J.M., andChaillet, J.R. 2001. Genomicimprintingdisrupted by a maternal effect mutation in the Dnmt1 gene.Cell 104: 829838.Hsieh, C.-L. 1994. Dependence of transcriptional repression onCpG methylation density. Mol. Cell. Biol. 14: 54875494..1999a.EvidencethatproteinbindingspecifiessitesofDNA demethylation. Mol. Cell. Biol. 19: 4656.. 1999b. In vivo activity of murine de novo methyltrans-ferases, Dnmt3a and Dnmt3b. Mol. Cell. Biol.19: 82118218.. 2000. Dynamics of DNAmethylationpattern. Curr.Opin. Genet. Dev. 10: 224228.Hung, M.S., Karthikeyan, N., Huang, B., Koo, H.C., Kiger, J., andShen, C.J. 1999. DrosophilaproteinsrelatedtovertebrateDNA (5-cytosine) methyltransferases. Proc. Natl. Acad. Sci.96: 1194011945.Iida, T., Nakahori, Y., Tsutsumi, O., Taketani, Y., andNak-agome, Y. 1994. The CpG island of the FMR-1 gene is meth-ylated differently among embryonic tissues: Implication forprenatal diagnosis. Hum. Reprod. 9: 14711473.Issa, J.P. 2000. CpG-islandmethylationinagingandcancer.Curr. Top. Microbiol. Immunol. 249: 101118.Jackson-Grusby, L., Beard, C., Possemat, R., Tudor, M., Fam-brough,D.,Csankovszki,G.,Dausman,J.,Lee,P.,Wilson,C., Lander, E., et al. 2001. Loss of genomicmethylationcauses p53-dependent apoptosis and epigenetic deregulation.Nat. Genet. 27: 3139.Jaenisch, R. 1997. DNAmethylation and imprinting: Whybother? Trends Genet. 13: 323329.Jaenisch, R., Harbers, K., Jahner, D., Stewart, C., andStuhl-mann, H. 1982. DNAmethylation, retroviruses, andem-bryogenesis. J. Cell Biochem. 20: 331336.Jaenisch, R., Schnieke, A., and Harbers, K. 1985. Treatment ofmice with 5-azacytidine efficiently activates silent retroviralgenomes in different tissues. Proc. Natl. Acad. Sci. 82: 14511455.Jeanpierre,M.,Turleau,C.,Aurias,A.,Prieur,M.,Ledeist,F.,Fischer, A., andViegas-Pequignot, E. 1993. Anembryonic-likemethylationpatternof classical satelliteDNAisob-served in ICF syndrome. Hum. Mol. Genet. 2: 731735.Jeddeloh, J.A., Stokes, T.L., andRichards, E.J. 1999. Mainte-nanceofgenomicmethylationrequiresaSW12/SNF2-likeprotein. Nat. Genet. 22: 9497.Jenuwein, T. and Allis, C.D. 2001. Translating the histone code.Science 293: 10741080.Jones, P.A. and Takai, D. 2001. The role of DNA methylation inmammalian epigenetics. Science 293: 10681070.Jones, P.A. and Taylor, S.M. 1980. Cellular differentiation, cy-tidine analogues and DNA methylation. Cell 20: 8593.Jones, P.A., Wolkowicz, M.J., Rideout, W.M.I., Gonzales, F.A.,Marziasz, C.M., Coetzee, G.A., and Tapscott, S.J. 1990. Denovo methylation of the MyoD1 CpG island during the es-tablishmentof immortal cell lines. Proc. Natl. Acad. Sci.87: 61176121.Jost, J.P., Fremont, M., Siegmann, M., and Hofsteenge, J. 1997.TheRNAmoietyofchickembryo5-methylcytosine-DNAglycosylase targets DNA demethylation. Nucleic Acids Res.25: 45454550.Jost, J.P., Siegmann, M., Thiry, S., Jost, Y.C., Benjamin, D., andSchwarz, S. 1999. A re-investigation of the ribonuclease sen-sitivity of a DNA demethylation reaction in chicken embryoand G8 mouse myoblasts. FEBS Lett. 449: 251254.Juttermann, R., Li, E., and Jaenisch, R. 1994. Toxicity of 5-aza-2-deoxycytidine to mammalian cells is mediated primarilyby covalent trapping of DNA methyltransferase rather thanDNA demethylation. Proc. Natl. Acad. Sci.91: 1179711801.Kafri, T., Ariel, M., Brandeis, M., Shemer, R., Urven, L., McCar-rey, J., Cedar, H., and Razin, A. 1992. Developmental patternof gene-specific DNA methylation in the mouse embryo andgerm line. Genes & Dev. 6: 705714.Kang, Y.K., Koo, D.B., Park, J.S., Choi, Y.H., Chung, A.S., Lee,K.K., and Han, Y.M. 2001a. Aberrant methylation of donorgenome in cloned bovine embryos. Nat. Genet. 28: 173177.Kang, Y.K., Koo, D.B., Park, J.S., Choi, Y.H., Kim, H.N., Chang,W.K., Lee, K.K., andHan, Y.M. 2001b. Typical demethyl-ation events in cloned pig embryos. Clues on species-specificdifferences in epigenetic reprogramming of cloned donor ge-nome. J. Biol. Chem. 276: 3998039984.Kass, S.U., Landsberger, N., and Wolffe, A.P. 1997a. DNA meth-ylation directs a time-dependent repression of transcriptioninitiation. Curr. Biol. 7: 157165.Kass, S.U., Pruss, D., and Wolffe, A.P. 1997b. How does DNABird18 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from methylationrepresstranscription?TrendsGenet. 13: 444449.Keohane, A.M., ONeill, L.P., Belyaev, N.D., Lavender, J.S., andTurner, B.M. 1996. X-inactivationandhistoneH4acetyla-tion in embryonic stem cells. Dev. Biol. 180: 618630.Kondo, T., Bobek, M.P., Kuick, R., Lamb, B., Zhu, X., Narayan,A., Bourchis, D., Viegas-Pequignot, E., Ehrlich, M., et al.2000. Whole-genomemethylationscaninICFsyndrome:Hypomethylationof non-satelliteDNArepeatsD4Z4andNBL2. Hum. Mol. Genet. 9: 597604.Kress, C., Thomassin, H., and Grange, T. 2001. Local DNA de-methylation in vertebrates: How could it be performed andtargeted? FEBS Lett. 494: 135140.Laayoun,A.andSmith,S.S.1995.Methylationofslippeddu-plexes, snapbacks and cruciforms by human DNA(cytosine-5)methyltransferase. Nucleic Acids Res. 23: 15841589.Lander, E.S., Linton, L.M., Birren, B., Nusbaum, C., Zody, M.C.,Baldwin, J., Devon, K., Dewar, K., Doyle, M., FitzHugh, W.,et al. 2001. Initial sequencingandanalysisof thehumangenome. Nature 409: 860921.Lemon, B. andTjian, R. 2000. Orchestratedresponse: Asym-phonyof transcriptionfactors for genecontrol. Genes &Dev. 14: 25512569.Li, E., Bestor, T.H., and Jaenisch, R. 1992. Targeted mutation oftheDNAmethyltransferasegeneresultsinembryonicle-thality. Cell 69: 915926.Li, E., Beard, C., and Jaenisch, R. 1993. Role for DNA methyl-ation in genomic imprinting. Nature 366: 362365.Lin, I.G., Tomzynski, T.J., Ou, Q., and Hsieh, C.L. 2000. Modu-lation of DNA binding protein affinity directly affects targetsite demethylation. Mol. Cell. Biol. 20: 23432349.Liu, W.M., Maraia, R.J., Rubin, C.M., and Schmid, C.W. 1994.Alutranscrips: CytoplasmiclocalisationandregulationbyDNA methylation. Nucleic Acids Res. 22: 10871095.Liu, W.M., Chu, W.M., Choudary, P.V., and Schmid, C.W. 1995.Cellstressandtranslationalinhibitorstransientlyincreasethe abundance of mammalian SINE transcripts. Nucleic Ac-ids Res. 23: 17581765.Lock,L.F.,Takagi,N.,andMartin,G.R.1987.MethylationoftheHprtgeneontheinactiveXoccursafterchromosomeinactivation. Cell 48: 3946.Lorincz,M.C.,Schubeler,D.,Goeke,S.C.,Walters,M.,Grou-dine, M., and Martin, D.I.K. 2000. Dynamic analysis of pro-viral induction and de novo methylation: Implications for ahistone deacetylase-independent, methylation density-de-pendent mechanism of transcriptional repression. Mol. Cell.Biol. 20: 842850.Lyko, F., Ramsahoye, B.H., Kashevsky, H., Tudor, M., Mastran-gelo, M.-A., Orr-Weaver, T.L., and Jaenisch, R. 1999. Mam-malian (cytosine-5) methyltransferases cause genomic DNAmethylation and lethality in Drosophila. Nat. Genet.23: 363366.Lyko, F., Ramsahoye, B.H., and Jaenisch, R. 2000. DNA meth-ylation in Drosophila melanogaster. Nature 408: 538540.Lyle,R.,Watanabe,D.,teVruchte,D.,Lerchner,W.,Smrzka,O.W., Wutz, A., Schageman, J., Hahner, L., Davies, C., andBarlow, D.P. 2000. The imprinted antisense RNA at the Igf2rlocus overlaps but does not imprint Mas1. Nat. Genet.25: 1921.MacLeod, D., Charlton, J., Mullins, J., and Bird, A.P. 1994. Sp1sites in the mouse aprt gene promoter are required to preventmethylation of the CpG island. Genes & Dev. 8: 22822292.MacLeod, D., Ali, R.R., and Bird, A.P. 1998. An alternative pro-moter in the mouse major histocompatibility cmplex class III-Abgene:ImplicationsfortheoriginofCpGislands.Mol.Cell. Biol. 18: 44334443.MacLeod, D., Clark, V., and Bird, A. 1999. Absence of genome-widechangesinDNAmethylationduringdevelopmentofthe zebrafish (Danio rerio). Nat. Genet. 23: 139140.Malik, H.S., Burke, W.D., and Eickbush, T.H. 1999. The age andevolution of non-LTR retrotransposable elements. Mol. Biol.Evol. 16: 793805.Martienssen, R.A. andColot, V. 2001. DNAmethylationandepigenetic inheritance in plants and filamentous fungi. Sci-ence 293: 10701074.Matzke, M., Matzke, A.J., and Kooter, J.M. 2001. RNA: Guidinggene silencing. Science 293: 10801083.Mayer, W., Niveleau, A., Walter, J., Fundele, R., andHaaf, T.2000. Demethylationof thezygoticpaternal genome. Na-ture 403: 501502.McKeon, C., Ohkubo, H., Pastan, I., anddeCrombrugghe, B.1982. Unusual methylationpatternof the2(1) collagengene. Cell 29: 203210.Miao,V.P.,Freitag,M.,andSelker,E.U.2000.ShortTpA-richsegments of the region induce DNA methylation in Neu-rospora crassa. J. Mol. Biol. 300: 249273.Miniou, P., Jeanpierre, M., Blanquet, V., Sibella, V., Bonneau,D., Herbelin, C., Fischer, A., Niveleau, A., and Viegas-Pequignot,E.1994.Abnormalmethylationpatternincon-stitutiveandfacultative(Xinactivechromosome) hetero-chromatin of ICF patients. Hum. Mol. Genet. 3: 20932102.Mohandas, T., Sparkes, R.S., and Shapiro, L.J. 1981a. Reactiva-tionofaninactivehumanXchromosome: EvidenceforXinactivation by DNA methylation. Science 211: 393396.. 1981b. Reactivationof aninactivehumanX-chromo-some: Evidence for X-inactivation by DNA methylation. Sci-ence 211: 393396.Monk, M., Boubelik, M., andLehnert, S. 1987. Temporalandregional changes in DNA methylation in the embryonic, ex-traembryonic and germ cell lineages during mouse embryodevelopment. Development 99: 371382.Moore, T. and Haig, D. 1991. Genomic imprinting in mamma-lian development: A parental tug-of-war. TIG 7: 14.Mostoslavsky, R., Singh, N., Kirillov, A., Pelanda, R., Cedar, H.,Chess, A., andBergman, Y. 1998. chainmonoallelicde-methylation and the establishment of allelic exclusion.Genes & Dev. 12: 18011811.Mummaneni, P., Bishop, P.L., andTurker, M.S. 1993. Acis-acting element accounts for a conserved methylation patternupstreamof themouseadeninephosphoribosyltransferasegene. J. Biol. Chem. 268: 552558.Mummaneni, P., Yates, P., Simpson, J., Rose, J., andTurker,M.S. 1998. The primary function of a redundant Sp1 bindingsite in the mouse aprt gene promoter is to block epigeneticgene inactivation. Nucleic Acids Res. 26: 51635169.Nakayama, J., Rice, J.C., Strahl, B.D., Allis, C.D., and Grewal,S.I. 2001. Role of histone H3 lysine 9 methylation in epige-netic control of heterochromatin assembly. Science292: 110113.Nan, X., Meehan, R.R., andBird, A. 1993. Dissectionof themethyl-CpG binding domain from the chromosomal proteinMeCP2. Nucleic Acids Res. 21: 48864892.Nan, X., Campoy, J., and Bird, A. 1997. MeCP2 is a transcrip-tional repressor withabundant binding sites ingenomicchromatin. Cell 88: 471481.Niwa,O.,Yokota,Y.,Ishida,H.,andSugahara,T.1983.Inde-pendent mechanisms involved in suppression of theMoloneyleukaemiavirusgenomeduringdifferentiationofmurine teratocarinoma cells. Cell 32: 11051113.Noma, K., Allis, C.D., andGrewal, S.I. 2001. Transitionsindistinct histone H3 methylation patterns at the heterochro-matin domain boundaries. Science 293: 11501155.DNA methylation and epigenetic memoryGENES & DEVELOPMENT 19Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from Okano, M., Xie, S., and Li, E. 1998a. Cloning and characteriza-tionof a family of novel mammalianDNA(cytosine-5)methyltransferases. Nat. Genet. 19: 219220.. 1998b. Dnmt2 is not required for de novo and mainte-nancemethylationofviralDNAinembryonicstemcells.Nucleic Acids Res. 26: 25362540.Okano, M., Bell, D.W., Haber, D.A., and Li, E. 1999. DNA meth-yltransferases Dnmt3a and Dnmt3b are essential for de novomethylation and mammalian development. Cell99: 247257.Oswald, J., Engemann, S., Lane, N., Mayer, W., Olek, A., Fun-dele, R., Dean, W., Reik, W., andWalter, J. 2000. Activedemethylation of the paternal genome in the mouse zygote.Curr. Biol. 10: 475478.Pannell, D., Osborne, C.S., Yao, S., Sukonnik, T., Pasceri, P.,Karaiskakis, A., Okano, M., Li, E., Lipshitz, H.D., and Ellis,J. 2000. Retrovirusvector silencingisdenovomethylaseindependent and marked by a repressive histone code.EMBO J. 19: 58845894.Panning, B. and Jaenisch, R. 1996. DNA hypomethylation canactivate Xist expression and silence X-linked genes. Genes &Dev. 10: 19912002.Paro, R., Strutt, H., and Cavalli, G. 1998. Heritable chromatinstates induced by the Polycomb and trithorax group genes.Novartis Found. Symp. 214: 5161.Pfeifer, G.P., Steigerwald, S.D., Hansen, R.S., Gartler, S.M., andRiggs, A.D. 1990. Polymerase chain reaction-aided genomicsequencing of an X chromosome-linked CpG island: Meth-ylationpatterns suggest clonal inheritance, CpGsiteau-tonomy, and an explanation of activity state stability. Proc.Nat. Acad. Sci. 87: 82528256.Pirrotta,V.1999.Polycombsilencingandthemaintenanceofstable chromatin states. Results Probl. Cell. Differ. 25: 205228.Pollack, Y., Stein, R., Razin, A., andCedar, H. 1980. Methyl-ationof foreignDNAsequencesineukaryoticcells. Proc.Natl. Acad. Sci. 77: 64636467.Poux, S., McCabe, D., andPirrotta, V. 2001. Recruitment ofcomponentsof PolycombGroupchromatincomplexesinDrosophila. Development 128: 7585.Pradhan, S., Bacolla, A., Wells, R.D., andRoberts, R.J. 1999.Recombinant human DNA (cytosine-5) methyltransferase. I.Expression, purification, andcomparisonof de novoandmaintenance methylation. J. Biol. Chem. 274: 3300233010.Prokhortchouk, A., Hendrich, B., Jorgensen, H., Ruzov, A.,Wilm, M., Georgiev, G., Bird, A., andProkhortchouk, E.2001. The p120 catenin partner Kaiso is a DNA methylation-dependent transcriptional repressor. Genes & Dev. 15: 16131618.Ramchandani, S., Bhattacharya, S.K., Cervoni, N., and Szyf, M.1999. DNAmethylationis a reversible biological signal.Proc. Natl. Acad. Sci. 96: 61076112.Ramsahoye, B.H., Biniszkiewicz, D., Lyko, F., Clark, V., Bird,A.P., and Jaenisch, R. 2000. Non-CpG methylation is preva-lent in embryonic stem cells and may be mediated by DNAmethyltransferase 3a. Proc. Natl. Acad. Sci. 97: 52375242.Reik,W.andWalter,J.2001.Evolutionofimprintingmecha-nisms: Thebattleof thesexesbeginsinthezygote. Nat.Genet. 27: 255256.Reik, W., Dean, W., and Walter, J. 2001. Epigenetic reprogram-ming in mammalian development. Science 293: 10891093.Rhee, I., Jair, K.W., Yen, R.W., Lengauer, C., Herman, J.G., Kin-zler, K.W., Vogelstein, B., Baylin, S.B., andSchuebel, K.E.2000. CpG methylation is maintained in human cancer cellslacking DNMT1. Nature 404: 10031007.Rideout, I.W., Eggan, K., and Jaenisch, R. 2001. Nuclear cloningand epigenetic reprogramming of the genome. Science293: 10931098.Riggs, A.D. 1975. X-inactivation, differentiation and DNAmethylation. Cytogenet. Cell. Genet. 14: 925.Riggs, A.D., Xiong, Z., Wang, L., and LeBon, J.M. 1998. Meth-ylationdynamics, epigenetic fidelityandXchromosomestructure. Novartis Found. Symp. 214: 214225.Robertson, K.D., Ait-Si-Ali, S., Yokochi, T., Wade, P.A., Jones,P.L., and Wolffe, A.P. 2000. DNMT1 forms a complex withRb, E2F1 and HDAC1 and represses transcription from E2F-responsive promoters. Nat. Genet. 25: 338342.Rougier, N., Bourchis, D., Gomes, D.M., Niveleau, A., Plachot,M.,Paldi,A.,andViegas-Pequignot,E.1998.Chromosomemethylationpatterns during mammalianpreimplantationdevelopment. Genes & Dev. 12: 21082113.Russo, V.E.A., Martienssen, R.A., and Riggs, A.D. 1996. Epige-neticmechanismsof generegulation. ColdSpringHarborLaboratory Press, Cold Spring Harbor, NY.Sado, T., Fenner, M.H., Tan, S.S., Tam, P., Shioda, T., and Li, E.2000. Xinactivationinthe mouse embryo deficient forDnmt1: Distinct effect of hypomethylationonimprintedand random X inactivation. Dev. Biol. 225: 294303.Selker, E.U. 1990. DNA methylation and chromatin structure:A view from below. Trends Biochem. Sci. 15: 103107.. 1999. Gene silencing: Repeats that count. Cell 97: 157160.Shemer, R., Hershko, A.Y., Perk, J., Mostoslavsky, R., Tsuberi,B., Cedar, H., Buiting, K., and Razin, A. 2000. The imprintingboxofthePrader-Willi/Angelmansyndromedomain. Nat.Genet. 26: 440443.Siegfried, Z., Eden, S., Mendelsohn, M., Feng, X., Tsuberi, B.Z.,andCedar,H.1999.DNAmethylationrepressestranscrip-tion in vivo. Nat. Genet. 22: 203206.Silva, A.J., Ward, K., and White, R. 1993. Mosaic methylation inclonal tissue. Dev. Biol. 156: 391398.Simmen, M.W., Leitgeb, S., Charlton, J., Jones, S.J.M., Harris,B.R., Clark, V.H., and Bird, A. 1999. Nonmethylated trans-posableelementsandmethylatedgenesinachordatege-nome. Science 283: 11641167.Smit, A.F. 1999. Interspersedrepeatsandothermementosoftransposable elements in mammalian genomes. Curr. Opin.Genet. Dev. 9: 657663.Smith, S.S., Kan, J.L.C., Baker, D.J., Kaplan, B.E., and Dembek,P. 1991. Recognition of unusual DNA structures by humanDNA(cytosine-5)methyltransferase. J. Mol. Biol. 217: 3951.Stancheva, I. andMeehan, R.R. 2000. Transient depletionofxDnmt1 leads to premature gene activation in Xenopus em-bryos. Genes & Dev. 14: 313327.Stewart, C.L., Stuhlmann, H., Jahner, D., and Jaenisch, R. 1982.Denovomethylationandinfectivityofretroviralgenomesintroduced into embryonal carcinoma cells. Proc. Natl.Acad. Sci. 79: 40984102.Stger, R., Kajimura, T.M., Brown, W.T., and Laird, C.D. 1997.EpigeneticvariationillustratedbyDNAmethylationpat-terns of the fragile-X gene FMR1. Hum. Mol. Gen. 6: 17911801.Swisher, J.F., Rand, E., Cedar, H., andMariePyle, A. 1998.Analysis of putative RNase sensitivity and protease insensi-tivityof demethylationactivityinextractsfromratmyo-blasts. Nucleic Acids Res. 26: 55735580.Szabo, P., Tang, S.H., Rentsendorj, A., Pfeifer, G.P., and Mann,J.R. 2000. Maternal-specific footprints at putative CTCFsites in the H19 imprinting control region give evidence forinsulator function. Curr. Biol. 10: 607610.Tada, M., Tada, T., Lefebvre, L., Barton, S.C., and Surani, M.A.1997. Embryonicgermcellsinduceepigeneticreprogram-Bird20 GENES & DEVELOPMENTCold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from ming of somatic nucleus in hybrid cells. EMBO J. 16: 65106520.Tamaru, H. and Selker, E.U. 2001. A histone H3 methyltrans-ferase controls DNAmethylation in Neurospora crassa.Nature 414: 277283.Tazi, J. andBird, A. 1990. AlternativechromatinstructureatCpG islands. Cell 60: 909920.Thomassin, H., Flavin, M., Espinas, M.L., and Grange, T. 2001.Glucocorticoid-induced DNA demethylation and genememory during development. EMBO J. 20: 19741983.Tremblay,K.D.,Duran,K.L.,andBartolomei,M.S.1997.A52-kilobase-pair region of the imprinted mouse H19 gene ex-hibitsexclusivepaternalmethylationthroughoutdevelop-ment. Mol. Cell. Biol. 17: 43224329.Turker, M.S. 1999. The establishment and maintenance ofDNA methylation patterns in mouse somatic cells. Semin.Cancer Biol. 9: 329337.Tweedie, S., Charlton, J., Clark, V., and Bird, A. 1997. Methyl-ationof genomesandgenesattheinvertebratevertebrateboundary. Mol. Cell. Biol. 17: 14691475.Tweedie, S., Ng, H.-H., Barlow, A.L., Turner, B.M., Hendrich, B.,and Bird, A. 1999. Vestiges of a DNA methylation system inDrosophila melanogaster. Nat. Genet. 23: 389390.Um, S., Harbers, M., Benecke, A., Pierrat, B., Losson, R., andChambon, P. 1998. Retinoicacidreceptorsinteractphysi-cally and functionally with the T:G mismatch-specific thy-mine-DNA glycosylase. J. Biol. Chem. 273: 2072820736.Venolia, L., Gartler, S.M., Wasserman, E.R., Yen, P., Mohandas,T., and Shapiro, L.J. 1982. Transformation with DNA from 5azacytidine-reactivatedXchromosomes. Proc. Natl. Acad.Sci. 79: 23522354.Venter,J.C.,Adams,M.D.,Myers,E.W.,Li,P.W.,Mural,R.J.,Sutton,G.G.,Smith,H.O.,Yandell,M.,Evans,C.A.,Holt,R.A., et al. 2001. The sequence of the human genome. Sci-ence 291: 13041351.Viegas-Pequignot, E. and Dutrillaux, B. 1976. Segmentation ofhuman chromosomes induced by 5-ACR(5-azacytidine).Hum. Genet. 34: 247254.Voo, K.S., Carlone, D.L., Jacobsen, B.M., Flodin, A., and Skalnik,D.G. 2000. Cloning of a mammalian transcriptional activa-tor that binds unmethylated CpG motifs and shares a CXXCdomain with DNA methyltransferase, human trithorax, andmethyl-CpGbinding domain protein 1. Mol. Cell. Biol.20: 21082121.Wakayama, T. and Yanagimachi, R. 2001. Mouse cloning withnucleusdonorcellsofdifferentageandtype.Mol.Reprod.Dev. 58: 376383.Walsh,C.P.andBestor,T.H.1999.Cytosinemethylationandmammalian development. Genes & Dev. 13: 2634.Walsh, C.P., Chaillet, J.R., and Bestor, T.H. 1998. Transcriptionof IAPendogenousretrovirusesisconstrainedbycytosinemethylation. Nat. Genet. 20: 116117.Wang, J., Mager, J., Chen, Y., Schneider, E., Cross, J.C., Nagy, A.,andMagnuson, T. 2001. ImprintedXinactivationmain-tained by a mouse Polycomb group gene. Nat. Genet.28: 371375.Wassenegger, M., Heimes, S., Riedel, L., and Sanger, H.L. 1994.RNA-directed de novo methylation of genomic sequences inplants. Cell 76: 567576.WaughONeill, R.J., ONeill, M.J., andMarshall Graves, J.A.1998. Undermethylation associated with retroelement acti-vation and chromosome remodelling in an interspecificmammalian hybrid. Nature 393: 6872.Weiss,A.,Keshet,I.,Razin,A.,andCedar,H.1996.DNAde-methylationinvitro: Involvement of RNA. Cell 86: 709718.Weng, A., Magnuson, T., andStorb, U. 1995. Strain-specifictransgenemethylationoccursearlyinmousedevelopmentand can be recapitulated in embryonic stem cells. Develop-ment 121: 28532859.Wigler, M., Levy, D., and Perucho, M. 1981. The somatic repli-cation of DNA methylation. Cell 24: 3340.Wolffe, A.P., Jones, P.L., and Wade, P.A. 1999. DNA demethyl-ation. Proc. Natl. Acad. Sci. 96: 58945896.Woodcock, D.M., Lawler, C.B., Linsenmeyer, M.E., Doherty,J.P., and Warren, W.D. 1997. Asymmetric methylation in thehypermethylated CpG promoter region of the human L1 ret-rotransposon. J. Biol. Chem 272: 78107816.Wutz, A. and Jaenisch, R. 2000. A shift from reversible to irre-versibleXinactivationistriggeredduringEScelldifferen-tiation. Mol. Cell 5: 695705.Wutz,A.,Smrzka,O.W.,Schweifer,N.,Schellander,K.,Wag-ner, E.F., and Barlow, D.P. 1997. Imprinted expression of theIgf2r gene depends on an intronic CpGisland. Nature389: 745749.Xu, G.-L., Bestor, T.H., Bourchis, D., Hsieh, C.-L., Tommerup,N., Bugge, M., Hulten, M., Qu, X., Russo, J.J., andViegas-Pequignot,E.1999.Chromosomeinstabilityandimmuno-deficiency syndrome caused by mutations in a DNA meth-yltransferase gene. Nature 402: 187191.Yates, P.A., Burman, R.W., Mummaneni, P., Krussel, S., andTurker, M.S. 1999. Tandem B1 elements located in a mousemethylation center provide a target for de novo DNA meth-ylation. J. Biol. Chem. 274: 3635736361.Yoder, J.A., Walsh, C.P., and Bestor, T.H. 1997. Cytosine meth-ylationandtheecologyof intragenomicparasites. TrendsGenet. 13: 335340.Zhang, Y. andReinberg, D. 2001. Transcriptionregulationbyhistonemethylation: Interplaybetweendifferent covalentmodifications of the core histone tails. Genes &Dev.15: 23432360.Zhu, B., Zheng, Y., Angliker, H., Schwarz, S., Thiry, S., Sieg-mann, M., and Jost, J.P. 2000a. 5-Methylcytosine DNA gly-cosylaseactivityisalsopresentinthehumanMBD4(G/Tmismatch glycosylase) and in a related avian sequence.Nucleic Acids Res. 28: 41574165.Zhu, B., Zheng, Y., Hess, D., Angliker, H., Schwarz, S., Sieg-mann, M., Thiry, S., and Jost, J.-P. 2000b. 5-Methylcytosine-DNAglycosylaseactivityispresentinaclonedG/Tmis-matchDNAglycosylaseassociatedwiththechickenem-bryoDNAdemethylationcomplex. Proc. Natl. Acad. Sci.97: 51355139.Zhu, B., Benjamin, D., Zheng, Y., Angliker, H., Thiry, S., Sieg-mann,M.,andJost,J.P.2001.Overexpressionof5-methyl-cytosine DNA glycosylase in human embryonic kidney cellsEcR293 demethylates the promoter of a hormone-regulatedreporter gene. Proc. Natl. Acad. Sci. 98: 50315036.DNA methylation and epigenetic memoryGENES & DEVELOPMENT 21Cold Spring Harbor Laboratory Press on February 14, 2012 - Published bygenesdev.cshlp.org Downloaded from