what an epigenome remembers

Prospects & Overviews

What an epigenome remembers

Ulrike C. Lange and Robert Schneider�

During mammalian development, maintenance of cell

fate through mitotic divisions require faithful replication

not only of the DNA but also of a particular epigenetic

state. Germline cells have the capacity of erasing this

epigenetic memory at crucial times during development,

thereby resetting their epigenome. Certain marks, how-

ever, appear to escape this reprogramming, which

allows their transmission to the offspring and potentially

guarantees transgenerational epigenetic inheritance.

Here we discuss the molecular requirements for faithful

transmission of epigenetic information and our current

knowledge about the transmission of epigenetic infor-

mation through generations.

Keywords:.epigenetics; histone modification; transmission of marks

Introduction

In the nineteenth century, the work of Gregor Mendel inau-gurated modern genetics – the science of inheritance. A cen-tury later, DNA as the molecular carrier of genetic informationwas discovered. The dogma was born that all inheritance mustbe achieved by transmitting DNA from mother to daughtercell, from parent to offspring. However, a small number ofscientists defended an idea that phenotypic information mayalso be coded beyond the DNA sequence. It was C. H.Waddington who first coined the term ‘epigenetic’ in 1942for means by which genotype brings about phenotype.Today, the terminology of epigenetics is still controversiallydiscussed [1]. Here we define epigenetics for simplicity reasonsas the study of inherited changes in phenotypic traits orgenome function without changes in the underlying DNAsequence – hence epi (greek: "pi – over, above)-genetics.

Epigenetic marking occurs through numerous mechan-isms. It encompasses covalent DNA and histone modifications,histone variants, nuclear RNA and means of organisingnuclear structure. A unifying feature of epigenetic marks istheir ability to regulate transcription. In eukaryotic cells, DNAis packaged into chromatin, while the degree of compactionregulates gene accessibility. First, the DNA helix is wrappedaround the nucleosomal core particle, a histone octamer con-sisting of two of each histones H2A, H2B, H3 and H4 [2]. Withthe help of linker histone H1, DNA is organised into higher-order chromatin structures, eventually resulting in hetero-chromatic (densely packed, transcriptionally repressed) andeuchromatic (open, accessible chromatin, containing mosttranscribed genes) regions [3].

Covalent modifications of chromatin components canregulate chromatin function. The best studied example of suchmodifications is the addition of methyl groups to DNA, whichin mammals occurs mostly symmetrically at cytosine-guano-sine dinucleotides (CpG sites) [4]. Methylation convertscytosine to 5-methylcytosine, which pairs up correctly witha complementary guanosine. CpG methylation frequentlyoccurs at repeated sequences in the genome and is involvedin the suppression of transcriptional activity and immobilityof transposable elements, such as intracisternal A-particles(IAPs) [5]. It also plays a key role in regulating monoallelicgene expression, as seen, for example at imprinted geneloci [6] Recently, CpG methylation has been mapped genome-wide in human cells [7]. In mammals, DNA methylation

DOI 10.1002/bies.201000030

Max Planck Institute for Immunobiology, Stubeweg 51, 79108 Freiburg,Germany

*Corresponding author:R. SchneiderE-mail: [email protected]

Abbreviations:PCNA, proliferating cell nuclear antigen; Dnmt, DNA methyltransferase; NP95,nuclear zinc finger protein 95; HP1, heterochromatin protein 1; Suv39,suppressor of position effect variegation 39; PRC, polycomb repressivecomplex; MBD1, methyl-CpG binding protein 1; IAP, intracisternal A-particles; LTR, long terminal repeat; SETDB1, SET domain bifurcated 1;HDAC, histone deacetylase.

Bioessays 32: 659–668,� 2010 WILEY Periodicals, Inc. www.bioessays-journal.com 659

Review

essays

patterns are established by a rather complex interplay ofthree DNA methyltransferases called Dnmt1, Dnmt3A andDnmt3B. The loss of any of these methyltransferases is lethalin mice [8].

Histones can also be abundantly modified. Of these modi-fications, histone acetylation, methylation and phosphoryl-ation have been studied most extensively [9]. Thesemodifications can change the affinity of DNA to the octamer,resulting, for example in chromatin loosening. They also serveas recruitment platforms for ‘reading complexes’, whichregulate downstream events and/or influence chromatin struc-ture directly [9] – a feature termed the ‘histone code’ [10, 11]. Ithas become evident that certain histone modification patternscan be correlated with specific activity states of the genome[12]. In addition to covalent modifications, RNA molecules canalso mediate epigenetic changes in gene expression patterns.Examples are para-mutations in plants, which refers to thetransfer of an acquired epigenetic state to an unlinked hom-ologous locus [13]. In this phenomenon, small RNAs are used tosense homology and RNA-bound complexes can trigger epi-genetic modifications resulting in gene silencing.

Two aspects of epigenetic information are particularlyremarkable. First, epigenetic information is in its entiretyan analogue form of information – a network of interactingmarkings with many different identities. This contrasts thedigital representation of the genetic four-base code. Wedo not yet fully understand how epigenetic complexity isbrought about, i.e. in which ways epigenetic marks influenceeach other and whether a hierarchy of epigenetic modifi-cation exists. Second, epigenetic information is principallyreversible and as such is dynamic. Covalent modificationscan be enzymatically removed, histones and methylatednucleotides can be exchanged, RNAs can be degraded.There appear to be differences in the stability of marks: whilemany enzymes removing histone modifications have beenidentified [9], it is still unclear to what extent CpG dinucleo-tides are actively demethylated [14,15]. Nevertheless, thisdynamicity allows for a much broader environmental influ-ence on epigenetic information as compared with the DNAcode.

The epigenetic network impacts globally as well as at thesingle-gene level on basic cellular events such as nuclearstructure and transcription and more complex processes suchas differentiation and development. It has now become evi-dent that the epigenetic state partly determines cell functionand developmental potential and as such, faithful trans-mission of epigenetic information from one cell to anotheris vital. However, there are periods in mammalian develop-ment when the genome-wide epigenetic profile is reset. Thesephases of epigenetic instability or the so-called ‘epigeneticreprogramming’ are, as far as we know, unique to cells of thegermline and mechanistically not yet fully understood [16].The resetting of epigenetic information at these time points isradical and near global. Only at certain genomic regions,epigenetic marking appears to escape erasure [16]. Thisescape allows for persistence of epigenetic states throughgermline development and ultimately to the next generation.In these instances, epigenetic information, just as the geneticcode, is subject to primary inheritance, i.e. independent anddirect transmission from F0 to F1.

In this review, we aim to address the concept of ‘epigenetictransmission’ and in particular the underlying molecularmechanisms, focussing on two features: first, the transmissionof epigenetic marks through mitosis, also termed ‘epigeneticmaintenance’ or ‘stability’, and second the transmission ofepigenetic information in between generations, also termed‘epigenetic inheritance’. It should be noted that many phrasesin the field are ill-defined and often used ambiguously, refer-ring to a variety of phenomena, and this will require furtherclarification in the future. This ambiguity concerns the term‘epigenetics’ but also the concept of ‘inheritance’. It is import-ant to distinguish between mitotic transmission of marks (i.e.through cell divisions) and the persistence of epigenetic marksthrough gametogenesis, which can lead to transgenerationalinheritance (i.e. the transmission of epigenetic marks throughgenerations via the gametes). Moreover, the re-establishmentof epigenetic marks postmeiosis in the gamete also needs to beconsidered. In many cases, the transgenerational inheritanceof epigenetic information via direct transmission through thegametes is not yet fully proven and still a controversial topic.Hence the use of the term ‘epigenetic transmission’ mightoften be more appropriate. For reasons of simplicity, thisreview will concentrate on the mammalian system.Transgenerational transmission of epigenetic marks in plants,invertebrates and non-mammalian organisms has been wellstudied and is excellently reviewed elsewhere [17].

Epigenetic transmission during mitosis

Mammalian development requires countless somatic cell div-isions. Every single such division demands duplication of notonly the genetic, but also the epigenetic information – aremarkable undertaking considering that DNA replicationrequires complete disruption of the chromatin structure. Wewill now discuss our current perception of the molecularmechanisms governing the transmission of epigenetic modi-fications from mother to daughter cell [18,19].

Transmission of DNA methylation during replication

To ensure correct duplication of the genetic information, DNAis asymmetrically replicated in a semi-conservative manner byspecialised DNA polymerases [20,21]. These DNA polymerasesare assisted by additional factors such as the loading clampand processivity factor PCNA (proliferating cell nuclear anti-gen), which links both strands and recruits important playersto the replication fork [20, 22] (Fig. 1).

Along with the replication of DNA, methylated cytosinebases are also semi-conservatively transmitted. This results inhemi-methylated DNA, i.e. DNA methylated on only one of thetwo strands. The DNA methyltransferase Dnmt1 binds to thePCNA complex [23] as well as hemi-methylated DNA [24] andcatalyses addition of a methyl group to the complementaryunmethylated CpG. Since Dnmt1 uses DNA as template it iscapable of reading DNA methylation regardless of the chro-matin environment. Recent studies showed an important rolefor Set- and Ring-associated (SRA)-domain containingproteins such as nuclear zinc finger protein 95 (NP95) formaintaining DNA methylation. NP95 binds preferentially to

U. C. Lange and R. Schneider Prospects & Overviews....

660 Bioessays 32: 659–668,� 2010 WILEY Periodicals, Inc.

Review

essays

hemi-methylated DNA, interacts with Dnmt1, and NP95deletion results in severe DNA methylation defects [25]. Inaddition, chromatin remodelling factors are required to pro-vide access for the cytosine methylation complex at thereplication fork [26]. These pathways ensure that DNA meth-ylation patterns are faithfully propagated over many cellgenerations, hence fulfilling the requirements of a ‘true’ epi-genetic mark.

Transmission of histone modifications

The transmission of histone marks through DNA replication ismore complex. During the passage of the replication fork,parental nucleosomes, including the modifications they carry,are disrupted and disassembled. Once the fork has passed,nucleosomes are reassembled on both daughter strands usingparental as well as newly synthesised histones (Fig. 1). Thecorrect histone modification patterns then have to be restored;however, in contrast to DNA methylation, there is no template.According to the current model, newly synthesised histonesand parental histones are incorporated on both daughterstrands in a random fashion [27]. To avoid a stepwise dilutionof modifications, it has been proposed that neighbouringnucleosomes on non-replicated DNA could instruct modifi-cations on nucleosomes of the newly synthesised strands[22]. Marks on parental histones would be recognised andbound by specific ‘readers’, which then recruit ‘writers’that modify the new histones accordingly [28,29] (Fig. 1).

There are now good examples that this might indeed be thecase. The heterochromatin protein 1 (HP1) specifically recog-nises the heterochromatic mark H3K9me3 via its chromodomain [30–32]. HP1 also interacts with Suv39, the enzymetrimethylating H3K9 [33], thereby recruiting further SUV39 totrimethylate H3K9 and to establish a self-sustaining loopensuring a high density of this mark. As a result, heterochro-matic regions spread across newly incorporated unmodifiedhistones and re-establishment of correct marks is ensured.

A further example are the polycomb group proteins, majorplayers in maintaining a repressive chromatin state and insilencing inappropriate gene expression particularly duringembryonic development [34]. These proteins act within twomain complexes, polycomb repressive complex 1 and 2(PRC1/2). PRC2 mediates trimethylation of H3K27 and againspecifically recognises the chromatin mark it sets [35,36].Interestingly this binding to H3K27 stimulates the PRC2 meth-yltransferase activity [35]. Thus the presence of H3K27me3, viathe mechanisms of PRC2 recruitment and stimulation of itsactivity, leads to an expansion of the mark onto neighbouringchromatin. Notably, at least in vitro, PRC1 complexes are notdisrupted during replication, but stay bound to chromatin orDNA [37]. Both, the binding of PRC2 to H3K27me3 and theretention of PRC1 on chromatin, ensure that upon cell division,repressed polycomb target genes stay repressed in daughtercells and active genes avoid polycomb-mediated repression.

Whereas this concept of ‘readers’ recruiting ‘writers’ is agood model to explain the maintenance of large chromatindomains, it falls short of explaining the maintenance of modi-fications specific to only one or two nucleosomes, e.g. thesharp peaks of activating modifications at transcription startsites or at boundary regions [38]. Furthermore, our currentknowledge is mostly limited to the perpetuation of therepressed state, whereas insights into the maintenance ofactive chromatin domains are more scarce.

Transmission of histone variants

Different histone variants have been found to distinguishdistinct chromatin states [39]. The histone variant H3.3, forexample marks active chromatin [40] and, in contrast to thecanonical H3, is deposited not only during S phase, but

Figure 1. Molecular mechanisms underlying mitotic epigenetictransmission. The schematic shows replication of DNA (dark blue,parental DNA; light blue, newly synthesised DNA) at the replicationfork (orange arrow). During replication, parental nucleosomes (grey)are disrupted and re-assembled on the daughter strands usingparental (grey) and newly synthesised histones (blue). Histone marksare shown as green squares and mechanisms responsible for theirtransmission are delineated in green. DNA methylation is depictedas red circles, and mechanisms responsible for methylationmaintenance in red. Mitotic transmission of epigenetic marks can bereplication-dependent, e.g. via coupling with the replication fork, orindependent of the replication machinery. Currently, little is knownabout how active chromatin domains are maintained (light greendelineation). See text for details.

....Prospects & Overviews U. C. Lange and R. Schneider

Bioessays 32: 659–668,� 2010 WILEY Periodicals, Inc. 661

Review

essays

throughout the cell cycle [41]. How could such an H3.3-containing chromatin domain be maintained throughout celldivision? One model suggests that during replication-coupleddeposition of H3, the incorporation of parental H3.3 creates anactive and/or open chromatin region. Although quantitativelydiluted, compared to the parental state, this H3.3 is sufficientto restart transcription at the replicated site. Once initiated,transcription could then trigger transcription-coupled replace-ment of canonical H3 by H3.3, resulting in a full restorationof the H3.3 domain (Fig. 1). Such transcription-coupledmechanisms could likewise ensure the localised transmissionof other variants and perhaps also the transmission of activechromatin marks.

How is mitotic epigenetic transmission coordinated?

Contrary to genetic information, remarkably coded in thecombination of only four different bases along the helicalDNA backbone, the epigenetic framework of a cell is muchmore complex due to the sheer number of players involved(covalent modifications, proteins and RNAs). It is intuitive thatreliably transmitting such network through multiple rounds ofcell division requires coordinated mechanisms, acting along aset hierarchy to re-establish the parental epigenetic state.What does this hierarchy look like?

As explained above, the reproduction of DNA methylationoccurs faithfully after DNA replication. Hence DNA methylationcould act as cornerstone to rebuild an epigenetic framework. Itcould recruit histone modifiers, which in turn set anchors forchromatin organisers that orchestrate the higher-order nuclearstructure of the daughter cell. Evidence for such coupling is themethyl-CpG-binding protein 1 (MBD1), which binds specificallyto methylated cytosines and forms a complex with theH3K9 methyltransferase SET domain bifurcated 1 (SETDB1)[42] at replication forks. It thereby targets the repressiveH3K9 methyl mark to DNA-methylated (i.e. repressed) genomicregions and restores heterochromatin. Additionally, histonedeacetylases (HDAC) [43,44] and the lysine methyltransferaseG9a/Glp1 [45, 46] interact with Dnmt1, again linkinghistone marks to DNA methylation (Fig. 1). DNA methylationcould hence serve as a central scaffold for epigenetic stability.

Recently also the reverse direction of DNA methylation byhistone modifications has been reported [47]. The de novo DNAmethyltransferases Dnmt3A/B bind to nucleosomes, which donot carry the active mark H3K4me3. At these regions they meth-ylate sites which were missed by Dnmt1 during replication – aback-up mechanism for correct maintenance of DNA methyl-ation postreplication (Fig. 1). Therefore, efficient transmission ofDNA methylation marks depends on the cooperation of differentDnmts and continuous feedback to histone modifications.

It is noteworthy that in contrast to the replication of thegenetic code, which is tightly controlled by specific proof-read-ing mechanisms [48], we have so far not been able to elucidatesuch molecular proof reading activity for the transmission ofepigenetic information. In fact, it may be due to the principallydynamic and reversible nature of epigenetic marks that strictproof reading is not required. Instead delayed feedback mech-anisms, as described above, can fortify and correct an initially‘weak’ transmission immediately after replication. These mech-anisms are possibly better suited for maintainance of epigenetic

marks since contrary to genetic information, not only qualitybut also local quantity of marks impacts on epigenetic coding.

Epigenetic stability and reprogramming inthe germline

Epigenetic stability is an important feature of somatic differ-entiation. It guarantees maintenance of specific gene expres-sion programs, locking cell lineages to developmental fates. Akey feature of the mammalian germline, however, is a lack ofsuch lineage commitment, replaced instead by a unique devel-opmental pluri- or even totipotency. This potency guaranteesdevelopment and reproduction. In order to achieve suchplasticity in cell fate, epigenetic reprogramming occurs attwo crucial time points in the germline: during early pre-implantation development and during germ cell development[16] (Fig. 2). These events are characterised by extensivechanges in DNA methylation status and chromatin modifi-cations. Nevertheless, certain epigenetic marks are main-tained through these near-global rearrangements. Thisallows for gametic transmission of epigenetic informationand can lead to transgenerational epigenetic inheritance –an aspect observed in health as well as disease.

Germline epigenetic reprogramming

Fertilisation brings together two parental genomes with verydifferent epigenetic states: whilst the maternal genome, in theprocess of completing meiosis, is packaged in abundantlymodified histones, the paternal genome is wrapped mostlyin histone substitutes called protamines [49]. In the oocyte,protamines are rapidly replaced by hyperacetylated histonesfrom the cytoplasmatic pool. The paternal DNA undergoesfast, near genome-wide DNA demethylation, sparing onlyfew genomic regions such as pericentromeric heterochroma-tin, IAP retrotransposons and paternally methylatedimprinted genes [50–52]. Once the pronuclei have unifiedand the zygote starts to divide, a second round of globalDNA demethylation on maternal DNA occurs [53, 54]. Againit exempts certain genomic sites such as imprints, and isaccompanied by dynamic changes in particular histone modi-fication marks [55–58]. This also is the time of the first lineagedifferentiation, characterised by different epigenetic signa-tures [49, 55, 58]. De novo genome-wide DNA methylationnow occurs in cells committed to the inner cell mass, whichupon implantation of the blastocyst begin to exert their fulldifferentiation potential [59].

It is around three days later in development (i.e. six daysafter fertilisation) that germ cell development begins in themouse [60] (Fig. 2). Originally arisen from the inner cell mass,cells specified to the germ cell lineage act as carriers of thegermline totipotent potential for future generations. This isonce more mirrored in a particular epigenetic state: germ cellsacquire histone modifications that set them apart from theirsomatic neighbours, such as enhanced levels of active marks(e.g. H3K9ac, H3K4me2/3) and reduction of repressive marks(e.g. H3K9me2, H3K27me3 and H4/H2AR3me2s) [61, 62]. Inaddition to repression of the somatic gene expression pro-gramme and induction of pluripotency-specific genes, this



Review

essays

setting likely provides the platform for the subsequent drasticepigenetic reprogramming in gonadal germ cells: transientglobal changes in the nuclear architecture, erasure of histonemodifications and exchange of histone variants accompanyrapid genome-wide DNA demethylation [62–64]. This reprog-ramming event is thought to be essential in order to maintaingermline DNA in a state of ‘underlying totipotency’ and toguarantee the proper development of the next generation. Itallows for imprints to be sex-specifically reset and X-chromo-some reactivation to occur [49]. Now exposed to the cell-to-cellsignalling influences, germ cells will shortly thereafter enterthe sex-specific program of gametogenesis.

In the past decade, exciting progress has been made in ourunderstanding of the molecular mechanisms responsible forthese two reprogramming events. Their distinguishing featureis the near global epigenetic rearrangement, involving alllevels of epigenetic information [16, 49, 65]. In order toachieve this rearrangement, passive dilution as well as activeerasure of DNA methylation, histone modifications andvariants appear to be employed [62]. Likely there are consider-able mechanistic differences between the two events,which furthermore also appear to differ in the extent of epi-genetic erasure. This is demonstrated, for example by the

maintenance of imprinted marks during pre-implantationdevelopment as opposed to their erasure in gonadal germcells [64, 66]. Together, however, both reprogramming eventsassure that the new generation is freed from any unnecessaryand potentially harmful parental epigenetic burden – an ulti-mate proof reading mechanism for epigenetic marks [62, 63].

Escaping reprogramming – the persistence of marks andepigenetic inheritance

If germline reprogramming in mammals were complete, trans-generational epigenetic inheritance would be non-existent.However, it appears that this might not be the case (see also[17, 67] and the references therein). The most prominent sup-port for transgenerational inheritance in mammals comesfrom experiments focusing on the agouti (A) locus in mice,a locus that controls coat colour. In mice which carry an Avy

allele, an IAP retrotransposon has inserted upstream of theagouti locus, causing ectopic expression of the Agouti protein,which results in yellow fur, obesity, diabetes and increasedsusceptibility to tumours [68]. When the cryptic promoter inthe long terminal repeat (LTR) of the IAP is silenced, the coatcolour reverts to agouti. Crucially, it was found that the coat

Figure 2. Reprogramming in the mousegermline. Epigenetic reprogrammingcharacterised by the resetting of histonemodification patterns and DNA methylationoccurs twice in the life cycle of a mouse, duringpre-implantation development and duringdevelopment of germ cells (green). Via yetunknown mechanisms, distinct marks canescape the reprogramming and persist (redline). On the left, different stages of mousedevelopment are shown together on a timelineof embryonic days postfertilisation. Germlinecells are schematically highlighted in blue. Onthe right, changes in epigenetic marks areexplained. See text for details. ICM, inner cellmass; TE, trophoectoderm.



Review

essays

colour of the mother shifts the proportion of phenotypes in theoffspring toward the maternal colour, i.e. yellow mothers aremore likely to produce yellow offspring and agouti mothersmore likely to give rise to agouti offspring [69–71]. This inher-itance is not the result of a maternally contributed environ-ment, but instead arises from the incomplete erasure of anepigenetic mark in the female germline [68].

So how could this be possible? It appears that the abovedescribed genome-wide reprogramming in germline cells is infact not as ‘global’ as the word might imply. Indeed, it hasbeen found using bisulfite sequencing that the methylatedstate of certain IAP elements is protected from demethylationduring reprogramming [52]. How this exemption is achievedand specifically targeted to a subset of IAP elements is stillunclear. There are, however, many examples of IAP elementsinfluencing gene expression on neighbouring loci via theirstate of DNA methylation [72]. One of them is expression ofthe agouti gene on the Avy allele [68]. Hence, since the meth-ylation status of IAPs can be resistant to reprogramming,transcriptional activity of neighbouring genes can be inheritedover multiple generations – a mechanism for transgenera-tional epigenetic inheritance in mice.

The murine Axin-fused (AxinFu) allele is a further example,where such mechanism likely plays a role [73,74]. In the AxinFu

allele, an IAP retrotransposon has inserted into the axin locus,which controls embryonic axis formation in vertebrates. Theinsertion results in aberrant axin transcripts, causing kinkingof the tail [74]. The molecular basis for this lies in a differentialDNA methylation at the IAP LTR. This phenotype is variablyexpressed among AxinFu carriers and transmitted throughboth sexes to the next generation.

Both Avy and AxinFu alleles are examples of epigenetictransmission linked to the substantial epigenetic stability ofIAP elements. These examples are signposts to the fact thatIAP elements largely escape germline reprogramming. IAPsbelong to the group of parasitic sequence elements acquiredby the mammalian genome, whose transcriptional activationcauses dysregulation of neighbouring gene loci and inducesnew mutations via replicative transposition [75, 76]. Amongthese parasitic sequences, the 1,000–2,000 IAP provirusesappear to be the most aggressive in the mouse genome andare hence kept silent by CpG methylation [77, 78]. A reactiva-tion during germline reprogramming would destabilise thegenome and leave the subsequent generation with a highburden of novel mutations.

However, these examples also demonstrate how much weyet have to learn about epigenetic control, reprogramming andcertainly transgenerational epigenetic transmission. There isevidence that, as opposed to most IAP sequences, the LTR ofthe IAP in the Avy allele is demethylated during pre-implan-tation development and hence does not always escape reprog-ramming [79]. This occurs both upon paternal and maternaltransmission. Hence in case of the Avy locus, cytosine meth-ylation of the IAP does not appear to be the sole epigeneticmark responsible for the described inheritance through thematernal germline [68, 79]. This of course raises the questionas to which epigenetic mark could instead be inherited.Specific histone modifications, such as H4K20me3, associatedwith IAP LTRs in embryonic stem cells, could be plausiblecandidates for such primary mark [80]. If this were the case,

we would likely have to query again our current concept ofhierarchy, interplay and stability of epigenetic markings.

Generally it is noteworthy that epigenetic transgenera-tional inheritance can occur through both the maternal andpaternal germline [81]. During spermatogenesis, a small per-centage of histones escape the histone to protamine replace-ment and recent studies have shown that these transmittednucleosomes are enriched at developmentally important geneloci and their modification status could impact on embryo-genesis [82,83]. The paternal genome hence contributes moresubstantially toward the epigenetic information of the zygotethan previously thought.

Epigenetic transmission and disease

While evidence for epigenetic inheritance in mice is sparse, inhumans our understanding is even more limited. It is undis-puted that epigenetic marks influence human phenotype(Fig. 3). The high degree of phenotypic discordance betweenmonozygotic twins might be taken as evidence [84]. It has alsobeen proposed for decades that the so-called ‘epimutations’,i.e. abnormal epigenetic states, can contribute to human dis-ease [85, 86]. However, it is uncertain whether and to whatextent such abnormal epigenetic states can be transmitted tothe next generation.

One major challenge is to identify which cases of trans-generational epigenetic effects, often termed ‘soft inheri-tance’, do arise from ‘true’ gametic epigenetic inheritance,as opposed to a number of different mechanisms in whichenvironmental influences can also cause heritable non-geneticdetermination of phenotype [17, 87–89]. In cases of truegametic transmission, the phenotype is passed on to F1 andsubsequent generations, independent of the environmentalinfluences that might have initiated the original phenotypicchange in the parental generation. In contrast, development ofthe offspring can also be affected in an environment-depend-ent manner by physiological and behavioural processes, suchas maternal nutrition, postfertilisation transfer of a viral infec-tion or poor maternal care in the context of stress response orcompromised maternal health [90, 91]. In these cases perpetu-ation of the phenotype to subsequent generations is due tocontinued or repeated exposure to a particular influenceand not due to transmission via the germline. Notably suchcases can easily mimic gametic transmission, since multiplegenerations, if timed appropriately, can be affected by asingle exposure: the pregnant mother, the F1 embryo/fetusand finally its developing germline and hence the future F2generation [17].

As a possible example for non-gametic transgenerationaltransmission it has been found that rats experiencing reducedpostnatal maternal care in the context of increased environ-mental stress ‘memorise’ this postnatal experience. Onceadult, they display the same behaviour of reduced caretowards their own offspring [92]. At a molecular level,the memory of postnatal mothering style can be explainedby alterations in the hippocampal serotonergic tone. Thisleads to changes in the expression of nerve growth factor-inducible protein A (NGFI-A) and eventually altered DNAmethylation and histone acetylation at the glucocorticoidreceptor locus. The number of glucocorticoid receptors in



Review

essays

the hippocampal region correlates with the fearfulness of therat and hence impacts on their maternal behaviour [91].Importantly, cross-fostering of pups from one mother typeto the other resulted in the pups to demonstrate the maternalcare style of their adoptive mothers.

Studies in humans are immensely complicated by ourenormous genetic heterogeneity. While the Avy or AxinFu allelecould be studied in inbred and hence genetically identicalmice, it will be very challenging to provide the formal proofthat variable expressivity in familial clustering is due togametic epigenetic inheritance in humans.

One example where inheritance of epimutations has beensuggested as underlying disease pathology is a subgroup ofpatients with Prader-Willi and Angelman syndromes [93, 94].Both syndromes are disorders caused by the loss of function ofoppositely imprinted genes on the human chromosome 15q[95]. In most cases, this loss of function is due to chromosomaldeletion at the 15q region, uniparental disomy, microdeletionor other genetic mutation. However, in up to 3% of describedcases, no genetic defect attributing to the phenotype has beenfound. Instead loss of gene function in these cases could bedue to aberrant imprinting, as a result of failure to erasea grandmaternal imprint in the paternal germline orvice versa [94].

Another debated example of transgenerational epigenetictransmission in humans comes from emerging evidence thatsome cases of familial colorectal cancer are caused by inher-ited epimutations. Hereditary non-polyposis colorectal cancer(HNPCC) syndrome is linked to germline loss of functionmutations of DNA mismatch repair genes, such as MLH1

and MSH2 [96]. Apart from genetic mutation, somatic hyper-methylation of the MLH1 gene promoter leading to its tran-scriptional silencing has previously been described insporadic colorectal cancer with mismatch repair deficiency[97, 98]. In recent years, a small number of families have beenidentified which carry a heritable epimutation at the MLH1 orMSH2 locus, causing a familial tumour syndrome reminiscentof HNPCC [99–102]. It is currently controversially debatedwhether these cases are prototype examples of gametic trans-generational inheritance in humans. The presence of bothepimutations in all three germ layers would support this idea[99, 100]. However, aberrant methylation patterns could alsobe secondary to a germline genetic aberration such as a crucialelement for transcriptional control, a possibility that isparticularly difficult to exclude, since enhancer elementsare capable of acting over a considerable distance to theirtarget locus [86, 103, 104].

The notion that epimutations in tumour suppressor genescould be passed on through gametic inheritance is a tantalis-ing one, despite the lack of formal proof. Many tumour syn-dromes show familial associations, which we have so far failedto understand. Germline transmission of an epimutationwould be one plausible cause, escaping detection using cur-rent sequencing technologies. In fact, not only in cancer, butfor a number of complex human diseases, the concept oftransgenerational epigenetic inheritance has been suggestedas an explanation for familial clustering [105, 106] (Fig. 3).While genome-wide association studies have identifiednumerous genetic variants associated with complex diseasessuch as type II diabetes, age-related macular degeneration orsystemic lupus erythematosus, these variants often onlyexplain small increments in risk and a small proportion offamilial cases. This discrepancy has been termed ‘missingheritability’ [105, 106].

Since epimutations, if transmitted through the germline,would show a non-mendelian inheritance, they would bedifficult to detect in affected pedigrees, even if technically

Figure 3. Influence of epigenetic and genetic effects on mammalianphenotype. The environment (blue) can impact on the epigenome(left) and the genome (right), both dictating the phenotype (red). Themolecular mechanisms of gametic epigenetic inheritance are not yetunderstood. Complex diseases are likely due to a combination ofgenetic, epigenetic and environmental effects.



Review

essays

appropriate analysis were done. In addition, due to the ulti-mately dynamic nature of epigenetic information, reversion tothe normal epigenetic state either spontaneously or uponpassing the germline would be expected in a certain percent-age of cases [107]. Both ‘metastability’, i.e. labile nature [107],and non-mendelian inheritance patterns of epimutationscould make them a potential source for missing heritability.Notably though, theoretical modelling suggests that althoughepigenetic inheritance could explain missing causality of com-plex traits, it is specifically the metastability of epigeneticmarks that will decrease their contribution to recurrence riskand hence the problem of missing heritability [108].

Conclusions

Epigenetic transmission is a field of research in its early days.We have just begun to grasp an idea of the variety of epigeneticmarks impacting on genomic function. However, their persist-ence and the mechanisms of transmission to daughter cells oroffspring still largely escape our understanding. While it isclear that mitotic epigenetic transmission is an aspect of everycell division, the question remains of whether a set hierarchyof epigenetic marks exists to govern this transmission andwhether the same hierarchy applies to different cell types.Determining the developmental potential of a cell is in fact acore function of epigenetics: with the one individual geneticcode that we all are born with, how else could we achieve thevariety in cell fates, if not via epigenetic means? In otherwords, since the genetic code is fixed, mammalian develop-ment is a matter of epigenetics and faithful epigenetic trans-mission is a prerequisite for its success.

The questions around the concept of transgenerationalepigenetic inheritance in mammals are numerous and havelargely not reached the level of experimentally addressing theunderlying molecular mechanisms. Even the existence of thisphenomenon is still controversially discussed [17]. From anevolutionary point of view, the reason for gametic inheritanceis indeed not obvious. Epigenetic marks appear too dynamicand ‘instable’ for providing a reliable evolutionary tool.Rather, they seem an ideal means of allowing short-termadaptation to environmental challenges.

As such, gametic transmission to multiple successivegenerations might not be necessary or even advantageous.If, for example a stressful environment impacts negatively onmaternal behaviour and this impact is mediated by epigeneticchanges, it would not be desirable to transmit these changes inthe long term to successive generations [91]. On the otherhand, due to the relative instability of epigenetic marks anepigenetic change resulting in a beneficial phenotype couldeasily be lost. Transgenerational inheritance might in fact alsobe dangerous. Epimutations that escape germline reprogram-ming could predispose entire families to disease. Neoplastictransformations in familial clusterings are just one example ofhow such inheritance might manifest itself [99–101, 109].Having said this, research focus around gametic epigeneticinheritance is for practical reasons likely biased towardspathological aspects. Certainly inheritance of epigeneticallycoded beneficial traits does exist. It will just be exceedinglymore difficult to detect such traits.

It is likely that in the future, after an era of focus on geneticmutation, medical understanding will be shaped more byinsights into epigenetic regulation and its somatic and germ-line persistence and inheritance. Potentially this may yield anexciting prospect: since epigenetic changes are of reversiblenature, diseases involving epimutations are in theory causallytreatable – familial traits could be annihilated. Will our under-standing of the epigenome ever be sufficient to reach this goal?Time will tell.

AcknowledgmentsThe authors wish to apologise to the authors of all the studiesthat we did not have space to include. We would like to thankN. Youngson for critical and constructive comments on themanuscript. Work in the R. S. Laboratory is supported bythe Max Planck Society, the DFG (through SFB 746), the EU(the Epigenome) and an ERC starting grant.

References

1. Ho DH, Burggren WW. 2010. Epigenetics and transgenerational trans-fer: a physiological perspective. J Exp Biol 213: 3–16.

2. Luger K, Mader AW, Richmond RK, et al. 1997. Crystal structure of thenucleosome core particle at 2.8 A resolution. Nature 389: 251–60.

3. Clapier CR, Cairns BR. 2009. The biology of chromatin remodelingcomplexes. Annu Rev Biochem 78: 273–304.

4. Doerfler W. 2008. In pursuit of the first recognized epigenetic signal-DNA methylation: a 1976 to 2008 synopsis. Epigenetics 3: 125–33.

5. Slotkin RK, Martienssen R. 2007. Transposable elements and theepigenetic regulation of the genome. Nat Rev Genet 8: 272–85.

6. Keverne B. 2009. Monoallelic gene expression and mammalian evol-ution. Bioessays 31: 1318–26.

7. Lister R, Pelizzola M, Dowen RH, et al. 2009. Human DNA methylomesat base resolution show widespread epigenomic differences. Nature462: 315–22.

8. Li E, Bestor TH, Jaenisch R. 1992. Targeted mutation of the DNAmethyltransferase gene results in embryonic lethality. Cell 69: 915–26.

9. Kouzarides T. 2007. Chromatin modifications and their function. Cell128: 693–705.

10. Strahl BD, Allis CD. 2000. The language of covalent histone modifi-cations. Nature 403: 41–5.

11. Turner BM. 1993. Decoding the nucleosome. Cell 75: 5–8.12. Barski A, Cuddapah S, Cui K, et al. 2007. High-resolution profiling of

histone methylations in the human genome. Cell 129: 823–37.13. Suter CM, Martin DI. 2010. Paramutation: the tip of an epigenetic

iceberg? Trends Genet 26: 9–14.14. Tahiliani M, Koh KP, Shen Y, et al. 2009. Conversion of 5-methylcy-

tosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partnerTET1. Science 324: 930–5.

15. Popp C, Dean W, Feng S, et al. 2010. Genome-wide erasure of DNAmethylation in mouse primordial germ cells is affected by AID deficiency.Nature 463: 1101–5.

16. Reik W. 2007. Stability and flexibility of epigenetic gene regulation inmammalian development. Nature 447: 425–32.

17. Youngson NA, Whitelaw E. 2008. Transgenerational epigenetic effects.Annu Rev Genomics Hum Genet 9: 233–57.

18. Probst AV, Dunleavy E, Almouzni G. 2009. Epigenetic inheritanceduring the cell cycle. Nat Rev Mol Cell Biol 10: 192–206.

19. Ng RK, Gurdon JB. 2008. Epigenetic inheritance of cell differentiationstatus. Cell Cycle 7: 1173–7.

20. Kunkel TA, Burgers PM. 2008. Dividing the workload at a eukaryoticreplication fork. Trends Cell Biol 18: 521–7.

21. Watson JD, Crick FH. 1953. The structure of DNA. Cold Spring HarbSymp Quant Biol 18: 123–31.

22. Corpet A, Almouzni G. 2009. Making copies of chromatin: the challengeof nucleosomal organization and epigenetic information. Trends Cell Biol19: 29–41.

23. Chuang LS, Ian HI, Koh TW, et al. 1997. Human DNA-(cytosine-5)methyltransferase-PCNA complex as a target for p21WAF1. Science277: 1996–2000.



Review

essays

24. Hermann A, Goyal R, Jeltsch A. 2004. The Dnmt1 DNA-(cytosine-C5)-methyltransferase methylates DNA processively with high preference forhemimethylated target sites. J Biol Chem 279: 48350–9.

25. Sharif J, Muto M, Takebayashi S, et al. 2007. The SRA protein Np95mediates epigenetic inheritance by recruiting Dnmt1 to methylated DNA.Nature 450: 908–12.

26. Jeddeloh JA, Stokes TL, Richards EJ. 1999. Maintenance of genomicmethylation requires a SWI2/SNF2-like protein. Nat Genet 22: 94–7.

27. Annunziato AT. 2005. Split decision: what happens to nucleosomesduring DNA replication? J Biol Chem 280: 12065–8.

28. Taverna SD, Li H, Ruthenburg AJ, et al. 2007. How chromatin-bindingmodules interpret histone modifications: lessons from professionalpocket pickers. Nat Struct Mol Biol 14: 1025–40.

29. Groth A, Corpet A, Cook AJ, et al. 2007. Regulation of replication forkprogression through histone supply and demand.Science 318: 1928–31.

30. Bannister AJ, Zegerman P, Partridge JF, et al. 2001. Selective recog-nition of methylated lysine 9 on histone H3 by the HP1 chromo domain.Nature 410: 120–4.

31. Lachner M, O’Carroll D, Rea S, et al. 2001. Methylation of histone H3lysine 9 creates a binding site for HP1 proteins. Nature 410: 116–20.

32. Lachner M, O’Sullivan RJ, Jenuwein T. 2003. An epigenetic road mapfor histone lysine methylation. J Cell Sci 116: 2117–24.

33. Aagaard L, Laible G, Selenko P, et al. 1999. Functional mammalianhomologues of the Drosophila PEV-modifier Su(var) 3-9 encode cen-tromere-associated proteins which complex with the heterochromatincomponent M31. EMBO J 18: 1923–38.

34. Schuettengruber B, Chourrout D, Vervoort M, et al. 2007. Genomeregulation by polycomb and trithorax proteins. Cell 128: 735–45.

35. Margueron R, Justin N, Ohno K, et al. 2009. Role of the polycombprotein EED in the propagation of repressive histone marks. Nature 461:762–7.

36. Hansen KH, Bracken AP, Pasini D, et al. 2008. A model for trans-mission of the H3K27me3 epigenetic mark. Nat Cell Biol 10: 1291–300.

37. Francis NJ, Follmer NE, Simon MD, et al. 2009. Polycomb proteinsremain bound to chromatin and DNA during DNA replication in vitro. Cell137: 110–22.

38. Mikkelsen TS, Ku M, Jaffe DB, et al. 2007. Genome-wide maps ofchromatin state in pluripotent and lineage-committed cells. Nature 448:553–60.

39. Henikoff S, Furuyama T, Ahmad K. 2004. Histone variants, nucleo-some assembly and epigenetic inheritance. Trends Genet 20: 320–6.

40. Ahmad K, Henikoff S. 2002. The histone variant H3.3 marks activechromatin by replication-independent nucleosome assembly.Mol Cell 9:1191–200.

41. Loyola A, Almouzni G. 2007. Marking histone H3 variants: how, whenand why? Trends Biochem Sci 32: 425–33.

42. Sarraf SA, Stancheva I. 2004. Methyl-CpG binding protein MBD1couples histone H3 methylation at lysine 9 by SETDB1 to DNA replicationand chromatin assembly. Mol Cell 15: 595–605.

43. Fuks F, Burgers WA, Brehm A, et al. 2000. DNA methyltransferaseDnmt1 associates with histone deacetylase activity. Nat Genet 24: 88–91.

44. Rountree MR, Bachman KE, Baylin SB. 2000. DNMT1 binds HDAC2and a new co-repressor, DMAP1, to form a complex at replication foci.Nat Genet 25: 269–77.

45. Feng YQ, Desprat R, Fu H, et al. 2006. DNA methylation supportsintrinsic epigenetic memory in mammalian cells. PLoS Genet 2: e65.

46. Esteve PO, Chin HG, Smallwood A, et al. 2006. Direct interactionbetween DNMT1 and G9a coordinates DNA and histone methylationduring replication. Genes Dev 20: 3089–103.

47. Jones PA, Liang G. 2009. Rethinking how DNA methylation patterns aremaintained. Nat Rev Genet. 10: 805–11.

48. Reha-Krantz LJ. 2009. DNA polymerase proofreading: multiple rolesmaintain genome stability. Biochim Biophys Acta 1804: 1049–63.

49. Morgan HD, Santos F, Green K, et al. 2005. Epigenetic reprogrammingin mammals. Hum Mol Genet 14: R47–58.

50. Olek A, Walter J. 1997. The pre-implantation ontogeny of the H19methylation imprint. Nat Genet 17: 275–6.

51. Rougier N, Bourc’his D, Gomes DM, et al. 1998. Chromosome meth-ylation patterns during mammalian preimplantation development.GenesDev 12: 2108–13.

52. Lane N, Dean W, Erhardt S, et al. 2003. Resistance of IAPs to meth-ylation reprogramming may provide a mechanism for epigenetic inher-itance in the mouse. Genesis 35: 88–93.

53. Monk M, Boubelik M, Lehnert S. 1987. Temporal and regional changesin DNA methylation in the embryonic, extraembryonic and germ celllineages during mouse embryo development. Development 99: 371–82.

54. Howlett SK. 1991. Genomic imprinting and nuclear totipotency duringembryonic development. Int Rev Cytol 127: 175–92.

55. Erhardt S, Su IH, Schneider R, et al. 2003. Consequences of thedepletion of zygotic and embryonic enhancer of zeste 2 during preim-plantation mouse development. Development 130: 4235–48.

56. Torres-Padilla ME, Parfitt DE, Kouzarides T, et al. 2007. Histonearginine methylation regulates pluripotency in the early mouse embryo.Nature 445: 214–8.

57. Daujat S, Weiss T, Mohn F, et al. 2009. H3K64 trimethylation marksheterochromatin and is dynamically remodeled during developmentalreprogramming. Nat Struct Mol Biol 16: 777–81.

58. Chapman V, Forrester L, Sanford J, et al. 1984. Cell lineage-specificundermethylation of mouse repetitive DNA. Nature 307: 284–6.

59. Santos F, Hendrich B, Reik W, et al. 2002. Dynamic reprogramming ofDNA methylation in the early mouse embryo. Dev Biol 241: 172–82.

60. Saitou M. 2009. Specification of the germ cell lineage in mice. FrontBiosci 14: 1068–87.

61. Ancelin K, Lange UC, Hajkova P, et al. 2006. Blimp1 associates withPrmt5 and directs histone arginine methylation in mouse germ cells. NatCell Biol 8: 623–30.

62. Hajkova P, Ancelin K, Waldmann T, et al. 2008. Chromatin dynamicsduring epigenetic reprogramming in the mouse germ line. Nature 452:877–81.

63. Hajkova P, Erhardt S, Lane N, et al. 2002. Epigenetic reprogramming inmouse primordial germ cells. Mech Dev 117: 15–23.

64. Sasaki H, Matsui Y. 2008. Epigenetic events in mammalian germ-celldevelopment: reprogramming and beyond. Nat Rev Genet 9: 129–40.

65. Li E. 2002. Chromatin modification and epigenetic reprogramming inmammalian development. Nat Rev Genet 3: 662–73.

66. Bartolomei MS. 2009. Genomic imprinting: employing and avoidingepigenetic processes. Genes Dev 23: 2124–33.

67. Jablonka E, Raz G. 2009. Transgenerational epigenetic inheritance:prevalence, mechanisms, and implications for the study of heredityand evolution. Q Rev Biol 84: 131–76.

68. Morgan HD, Sutherland HG, Martin DI, et al. 1999. Epigenetic inher-itance at the agouti locus in the mouse. Nat Genet 23: 314–8.

69. Wolff GL, Galbraith DB, Domon OE, et al. 1978. Phaeomelanin syn-thesis and obesity in mice. Interaction of the viable yellow (Avy) andsombre (eso) mutations. J Hered 69: 295–8.

70. Wolff GL. 1978. Influence of maternal phenotype on metabolic differ-entiation of agouti locus mutants in the mouse. Genetics 88: 529–39.

71. Wolff GL, Kodell RL, Moore SR, et al. 1998. Maternal epigenetics andmethyl supplements affect agouti gene expression in Avy/a mice.FASEB J 12: 949–57.

72. Druker R, Bruxner TJ, Lehrbach NJ, et al. 2004. Complex patterns oftranscription at the insertion site of a retrotransposon in the mouse.Nucleic Acids Res 32: 5800–8.

73. Reed SC. 1937. The inheritance and expression of fused, a newmutation in the house mouse. Genetics 22: 1–13.

74. Rakyan VK, Chong S, Champ ME, et al. 2003. Transgenerationalinheritance of epigenetic states at the murine Axin(Fu) allele occurs aftermaternal and paternal transmission. Proc Natl Acad Sci USA 100:2538–43.

75. Yoder JA, Walsh CP, Bestor TH. 1997. Cytosine methylation and theecology of intragenomic parasites. Trends Genet 13: 335–40.

76. Stocking C, Kozak CA. 2008. Murine endogenous retroviruses.Cell MolLife Sci 65: 3383–98.

77. Kazazian HH, Jr., Moran JV. 1998. The impact of L1 retrotransposonson the human genome. Nat Genet 19: 19–24.

78. Walsh CP, Chaillet JR, Bestor TH. 1998. Transcription of IAP endogen-ous retroviruses is constrained by cytosine methylation. Nat Genet 20:116–7.

79. Blewitt ME, Vickaryous NK, Paldi A, et al. 2006. Dynamic reprogram-ming of DNA methylation at an epigenetically sensitive allele in mice.PLoS Genet 2: e49.

80. Martens JH, O’Sullivan RJ, Braunschweig U, et al. 2005. The profile ofrepeat-associated histone lysine methylation states in the mouse epi-genome. EMBO J 24: 800–12.

81. Chong S, Vickaryous N, Ashe A, et al. 2007. Modifiers of epigeneticreprogramming show paternal effects in the mouse. Nat Genet 39: 614–22.

82. Carrell DT, Hammoud SS. 2010. The human sperm epigenome andits potential role in embryonic development. Mol Hum Reprod 16:37–47.

83. Hammoud SS, Nix DA, Zhang H, et al. 2009. Distinctive chromatin inhuman sperm packages genes for embryo development. Nature 460:473–8.



Review

essays

84. Machin GA. 1996. Some causes of genotypic and phenotypic discor-dance in monozygotic twin pairs. Am J Med Genet 61: 216–28.

85. Holliday R. 1987. The inheritance of epigenetic defects. Science 238:163–70.

86. Whitelaw NC, Whitelaw E. 2008. Transgenerational epigenetic inher-itance in health and disease. Curr Opin Genet Dev 18: 273–9.

87. Waterland RA. 2009. Is epigenetics an important link between early lifeevents and adult disease? Horm Res 71: 13–6.

88. Skinner MK, Manikkam M, Guerrero-Bosagna C. 2010. Epigenetictransgenerational actions of environmental factors in disease etiology.Trends Endocrinol Metab 21: 214–22.

89. Dolinoy DC, Jirtle RL. 2008. Environmental epigenomics in humanhealth and disease. Environ Mol Mutagen 49: 4–8.

90. Morse HC, III, Yetter RA, Stimpfling JH, et al. 1985. Greying with agein mice: relation to expression of murine leukemia viruses. Cell 41: 439–48.

91. Weaver IC, Cervoni N, Champagne FA, et al. 2004. Epigenetic pro-gramming by maternal behavior. Nat Neurosci 7: 847–54.

92. Francis D, Diorio J, Liu D, et al. 1999. Nongenomic transmission acrossgenerations of maternal behavior and stress responses in the rat.Science 286: 1155–8.

93. Gurrieri F, Accadia M. 2009. Genetic imprinting: the paradigm ofPrader-Willi and Angelman syndromes. Endocr Dev 14: 20–8.

94. Buiting K, Gross S, Lich C, et al. 2003. Epimutations in Prader-Willi andAngelman syndromes: a molecular study of 136 patients with an imprint-ing defect. Am J Hum Genet 72: 571–7.

95. Lalande M. 1996. Parental imprinting and human disease. Annu RevGenet 30: 173–95.

96. Lynch HT, de la Chapelle A. 2003. Hereditary colorectal cancer. N EnglJ Med 348: 919–32.

97. Kane MF, Loda M, Gaida GM, et al. 1997. Methylation of the hMLH1promoter correlates with lack of expression of hMLH1 in sporadic colon

tumors and mismatch repair-defective human tumor cell lines. CancerRes 57: 808–11.

98. Herman JG, Umar A, Polyak K, et al. 1998. Incidence and functionalconsequences of hMLH1 promoter hypermethylation in colorectal car-cinoma. Proc Natl Acad Sci USA 95: 6870–5.

99. Suter CM, Martin DI, Ward RL. 2004. Germline epimutation of MLH1 inindividuals with multiple cancers. Nat Genet 36: 497–501.

100. Chan TL, Yuen ST, Kong CK, et al. 2006. Heritable germline epimuta-tion of MSH2 in a family with hereditary nonpolyposis colorectal cancer.Nat Genet 38: 1178–83.

101. Hitchins MP, Ward RL. 2007. Erasure of MLH1 methylation in sperma-tozoa-implications for epigenetic inheritance. Nat Genet 39: 1289.

102. Hitchins MP, Wong JJ, Suthers G, et al. 2007. Inheritance of a cancer-associated MLH1 germ-line epimutation. N Engl J Med 356: 697–705.

103. Simonis M, Klous P, Splinter E, et al. 2006. Nuclear organization ofactive and inactive chromatin domains uncovered by chromosomeconformation capture-on-chip (4C). Nat Genet 38: 1348–54.

104. Zhao Z, Tavoosidana G, Sjolinder M, et al. 2006. Circular chromosomeconformation capture (4C) uncovers extensive networks of epigeneti-cally regulated intra- and interchromosomal interactions. Nat Genet 38:1341–7.

105. Manolio TA. 2009. Cohort studies and the genetics of complex disease.Nat Genet 41: 5–6.

106. Manolio TA, Collins FS, Cox NJ, et al. 2009. Finding the missingheritability of complex diseases. Nature 461: 747–53.

107. Rakyan VK, Blewitt ME, Druker R, et al. 2002. Metastable epialleles inmammals. Trends Genet 18: 348–51.

108. Slatkin M. 2009. Epigenetic inheritance and the missing heritabilityproblem. Genetics 182: 845–50.

109. Hitchins MP, Lin VA, Buckle A, et al. 2007. Epigenetic inactivation of acluster of genes flanking MLH1 in microsatellite-unstable colorectalcancer. Cancer Res 67: 9107–16.



Review

essays

what an epigenome remembers

Documents