Epigenetic modifications in an imprinting clusterare controlled by a hierarchy of DMRs suggestinglong-range chromatin interactionsSusana Lopes1,{, Annabelle Lewis1,{, Petra Hajkova2, Wendy Dean1, Joachim Oswald2,

Thierry Forne3, Adele Murrell1, Miguel Constancia1, Marisa Bartolomei4, Jorn Walter2 and

Wolf Reik1,*

1Laboratory of Developmental Genetics and Imprinting, Developmental Genetics Programme, The Babraham

Institute, Cambridge CB2 4AT, UK, 2Universitat des Saarlandes, Fr 8.2 Genetik, 66041 Saarbrucken, Germany,3Institut de Genetique Moleculaire de Montpellier, Montpellier, France and 4Howard Hughes Medical Institute and

Department of Cell and Developmental Biology, University of Pennsylvania School of Medicine, Philadelphia,

Pennsylvania 19104, USA

Received September 20, 2002; Revised and Accepted November 22, 2002

Imprinted genes and their control elements occur in clusters in the mammalian genome and carry epigeneticmodifications. Observations from imprinting disorders suggest that epigenetic modifications throughout theclusters could be under regional control. However, neither the elements that are responsible for regionalcontrol, nor its developmental timing, particularly whether it occurs in the germline or postzygotically, areknown. Here we examine regional control of DNA methylation in the imprinted Igf2-H19 region in the mouse.Paternal germline specific methylation was reprogrammed after fertilization in two differentially methylatedregions (DMRs) in Igf2, and was reestablished after implantation. Using a number of knockout strains in theregion, we found that the DMRs themselves are involved in regional coordination in a hierarchical fashion.Thus the H19 DMR was needed on the maternal allele to protect the Igf2 DMRs 1 and 2 from methylation, andIgf2 DMR1 was needed to protect DMR2 from methylation. This regional coordination occurred exclusivelyafter fertilization during somatic development, and did not involve linear spreading of DNA methylation,suggesting a model in which long-range chromatin interactions are involved in regional epigeneticcoordination. These observations are likely to be relevant to other gene clusters in which epigeneticregulation plays a role, and in pathological situations in which epigenetic regulation is disrupted.

INTRODUCTION

Imprinted genes in mammals are those genes that areexpressed from predominantly one of the parental alleles.These genes have important functions in mammalian deve-lopment particularly in fetal growth, placental developmentand function, and in certain behaviours after birth (1–4).There are approximately 50 imprinted genes currentlyidentified in the mouse genome, with over 100 expected,and most are conserved in the human genome (www.mgu.har.mrc.ac.uk/imprinting/imprin-viewdatagenes.html). One ofthe interesting features of organization of imprinted genes isthat most of them occur in clusters in the genome. Theimprinting cluster on distal chromosome 7 in mouse andproximal chromosome 11p15.5 in human, for example,contains 14 imprinted genes (2).

Imprinted genes are characterized by regions of parent-specific epigenetic modifications including DNA methylation,histone acetylation and histone methylation (1–6). Genetically,the maintenance of DNA methylation has been shown to becrucial for maintaining imprinting, with either biallelicexpression or biallelic silencing of imprinted genes in aknockout of the Dnmt1 gene (7). Recently, a number ofelements have been characterized whose parent-specificmethylation regulates expression or repression of imprintedgenes. These include promoters, promoters of antisense RNAs,chromatin boundaries, silencers and activators. These elementsare often involved in regulating more than one gene in animprinting cluster. Some of these differentially methylatedregions (DMRs) receive their methylation imprints in theparental germ cells, and these are then maintained throughall developmental stages, whereas others are significantly

*To whom correspondence should be addressed. Tel: þ44 1223496338; Fax: þ44 1223496015; Email: wolf.reik@bbsrc.ac.uk{The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors

Human Molecular Genetics, 2003, Vol. 12, No. 3 295–305DOI: 10.1093/hmg/ddg022

Human Molecular Genetics, Vol. 12, No. 3 # Oxford University Press 2003; all rights reserved

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

reprogrammed during development. The signals for inheritanceor reprogramming of DNA methylation imprints are unknown.

The clustering of imprinted genes and their sharing ofdifferentially methylated control sequences suggest thatepigenetic modifications throughout the cluster might be underregional control. Indeed, earlier observations on patients withthe classical imprinting disorders Prader–Willi/Angelmansyndromes (PWS/AS) or Beckwith–Wiedemann syndromehad suggested that defects in putative imprinting centres (IC)in the clusters could lead to altered epigenetic modificationsthroughout the cluster, and consequently altered expression ofimprinted genes (8–10). In the PWS/AS cluster it was proposedthat this regional coordination of the epigenotype occurred inthe germline, and was defective in PWS or AS patients withIC mutations (8). Subsequent work, however, showed thatIC mutations can also alter the epigenotype in the clusterduring somatic development (11). Thus, whether coordinationoccurs in the germline or during somatic development (or both)is not known.

Within the imprinting cluster on distal chromosome 7 inmouse there are likely to be two subdomains which areregulated separately, although a higher level of coordinationbetween the two domains cannot be ruled out (2). One of thesedomains includes the Cdkn1c and Kcnq1 genes and others, and

the antisense RNA gene Kcnq1ot1. The other domain containsthe paternally expressed insulin 2, insulin-like growth factor 2,and maternally expressed untranslated RNA gene H19. Withinthis domain there are four DMRs. The paternally methylatedH19 DMR contains a methylation sensitive chromatinboundary and silencer element, which restricts, whenunmethylated, the access to Igf2 of enhancers located distalto H19 (12–17). DMR1 is located upstream of the fetal pro-moters of Igf2, is paternally hypermethylated, and contains amethylation-sensitive silencer (18,19). DMR2 is located in thelast exon of Igf2 and contains a methylation-sensitive activator(20,21). The function of the maternally methylated DMR0which is located in the promoter of the placenta specifictranscript of Igf2 is currently unknown (22).

We have previously shown that maternal deletion of theH19 DMR and gene in the mouse leads to methylation ofthe maternal DMR1 and 2 regions, suggesting regionalcoordination of the epigenotype (23). Here we examinesystematically methylation patterns in the DMRs duringembryogenesis, and the control of these patterns in cis byusing various knockouts in the cluster. We show thatcoordination requires a hierarchy of the DMRs themselves,occurs during somatic development, and may involve long-range chromatin interactions.

Figure 1. Map of the Igf2-H19 locus and knockouts within the region. All exons of the Igf2 gene and promoters of both genes are represented. DMRs are shownunderneath together with their allele-specific methylation status (filled, hypermethylated). DMR0 methylation pattern refers only to placenta. DMRs 1 and 2 areexpanded to show maps of restriction enzymes used for methylation analysis by Southern blotting [R, B, D, E, N, F and h represent EcoRI, BamHI, DraI, EcoNI,FspI and HpaII, respectively; (m) or (s) indicate that the restriction enzyme is specific to M.m.domesticus or to M.spretus, respectively] and further expandedto show regions analysed by bisulphite sequencing. CpGs are represented by circles and numbered. s is a polymorphism specific to the M.spretus strain.Underneath is a map of the knockout strains used.

296 Human Molecular Genetics, 2003, Vol. 12, No. 3

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

Figure 2. Methylation levels in the Igf2-H19 locus at early developmental stages. (A) Methylation levels of DMRs in gametes and preimplantation embryosdetermined by bisulphite sequencing. DNA samples were treated with sodium bisulphite, amplified by PCR (see Fig. 1 for amplified region), cloned and sequenced.Dot diagrams show the methylation profile of DMR1 in oocytes (n¼ 12), sperm (n¼ 13), one-cell embryos (n¼ 20) and morulae (n¼ 16). Each line represents asingle DNA template molecule (where n ¼ the total number of sequences analysed). A representative selection of sequences is shown. Filled circles representmethylated CpGs. A polymorphism was used to distinguish the parental origin of the sequences in one-cell embryos and morulae. The graphs underneath sum-marize the methylation levels (overall percentage of methylated CpGs out of total CpGs in analyzed region) in gametes, zygotes and morulae in Igf2 DMR1, DMR2and H19 DMR. Note the reprogramming of Igf2 DMR2 and DMR1 gametic methylation differences, and the maintenance of the germline H19 DMR methylationlevels. The Igf2 DMR2 and H19 DMR methylation levels are taken from the indicated references. (B) Methylation levels of Igf2 DMR1 and DMR2 in postim-plantation embryos. Whole embryo E9 DNA and liver and kidney E15 DNA samples were treated with sodium bisulphite, amplified by PCR (see Fig. 1 for ampli-fied region), cloned and sequenced. Dot diagrams show the methylation profile of E9 whole embryos at DMR1 (n¼ 19) and DMR2 (n¼ 18), E15 kidney at DMR1(n¼ 33) and DMR2 (n¼ 20) and E15 liver at DMR1 (n¼ 21) and DMR2 (n¼ 20). Each line represents a single DNA template molecule (where n ¼ the totalnumber of sequences analysed). A representative selection of sequences is shown. Filled circles show methylated CpGs. Polymorphisms were used to distinguishbetween sequences of maternal and paternal origin. The graphs underneath summarize the methylation levels (overall percentage of methylated CpGs out of totalCpGs in analysed region). Note that the adult methylation pattern of DMR1 and DMR2 is established by E15 except for DMR2 in the kidney, which is still hypo-methylated.

Human Molecular Genetics, 2003, Vol. 12, No. 3 297

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

298 Human Molecular Genetics, 2003, Vol. 12, No. 3

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

RESULTS

Reprogramming of germline methylation in Igf2 DMRs

The Igf2-H19 region has four DMRs, three of which(H19 DMR, Igf2 DMRs 1 and 2) are paternally methylated(Fig. 1, this figure also summarizes the regions in DMR1 and 2analysed by Southern blotting and bisulphite sequencing, andthe knockout alleles that were used). The H19 DMR ismethylated in sperm but not oocytes and this differentialmethylation is maintained during all stages of development andall tissues, except in germ cells where the pattern is erased andre-established (24,25). Figure 2 summarizes the methylationpattern of the H19 DMR during preimplantation development(55). The methylation patterns of the Igf2 DMRs wereestablished by bisulphite sequencing (Fig. 2A). The gameticpattern of DMR2 had previously been determined (26)(Fig. 2A). DMR1, just as DMR2, was highly methylated insperm, and not methylated in oocytes. In striking contrasthowever to the H19 DMR, methylation was lost from thepaternal copy already in the zygote (DMR2 more pronouncedthan DMR1) and essentially all differential methylation waserased at the morula stage (Fig. 2). The diagrams in Figure 2show the methylation patterns of a representative selection oftemplate sequences from each developmental stage. Thecorresponding graphs summarize the methylation levels fromall analysed sequences. The total number of sequences analysedfor each region and time point is listed in the legend to Figure 2.

Paternal methylation patterns became re-established atpostimplantation stages (Fig. 2B). For both DMR1 and 2,there was some increase in methylation on the paternal allele inwhole embryonic day 9 (E9) fetuses (to 43% in DMR1 and20% in DMR2). Further increases occurred up to E15 and laterin individual tissues. Interestingly, in largely mesodermaltissues such as the kidney, an increase in DMR1 methylationpreceded the increase in DMR2 methylation temporally. Thus,while DMR1 was differentially methylated in E15 kidney,DMR2 differential methylation was not yet established on E15

(Fig. 2B), but was in place at late fetal stages (data not shown).A small increase in methylation of both DMRs was alsoobserved on the maternal allele (Fig. 2B).

DMRs protect from regional methylation in ahierarchical fashion

In order to see whether we could identify cis acting elementswithin the cluster that were important either for coordinatingmethylation imprints in the germline (all three DMRs aremethylated in sperm but not oocyte) or postzygotically whendifferential methylation is re-established, we tested threedifferent types of sequence with a total of six knockouts(Fig. 1). The Igf2 and H19 genes themselves, the knownendoderm enhancers, which are located 30 of H19 and arerequired for expression of both genes, and the three DMRsthemselves were tested. We measured differential methylationin the three DMRs, Igf2 promoters 1 and 3, and three differentlocations between the Igf2 and H19 genes, following maternalor paternal transmission of the knockouts. The overall resultsare summarised in Figure 5.

Since both bisulphite sequencing and Southern blotting wereused for the analysis of differential methylation (see below), weneeded to establish whether the two techniques resultedin comparable results. Hence, an experiment was carried outin which liver DNA was examined for DMR1 and 2 methyl-ation at a single restriction enzyme site by Southern blottingand at the same CpG by bisulphite sequencing (Fig. 3A). Thequantitative results were remarkably similar, so we decided thatwe could compare results obtained by the two differentmethods.

Deletion of the H19 DMR in the H19D13 knockout (27) (Fig. 1)had a striking effect on methylation of both Igf2 DMRs in bothendodermal and mesodermal tissues (Fig. 3B). Thus maternaltransmission of the H19D13 knockout resulted in a substantialincrease in methylation of the maternal DMR1 and 2 regions[from 38 to 63% in kidney (P< 0.01) and from 24 to 44% in liver(P< 0.01) in DMR1, and from 8 to 35% in kidney (P< 0.0001)

Figure 3. Effect of maternal deletion of the H19 DMR and of Igf2 DMR1 on methylation levels in the Igf2 DMRs. (A) Comparison of methylation levels of CpG 5in DMR1 and of CpG 21 in DMR2 in wild-type liver using Southern Blot analysis and bisulphite sequencing. M.m.domesticus females were mated with SD7 malesand E15 offspring analysed. For Southern blot analysis of DMR1 genomic DNAs (n¼ 4) were digested with XbaI, DraI and NaeI and probed with a XbaI–NaeIprobe. The methylation-sensitive enzyme NaeI restriction site corresponds to CpG 5 analysed by bisulphite sequencing. For Southern blot analysis of DMR2genomic DNAs (n¼ 7) were digested with BamHI, FspI and EcoNI and probed with a FspI–BamHI probe. The methylation-sensitive enzyme FspI correspondsto CpG 21 analysed by bisulphite sequencing. Graphs show comparison of methylation levels in paternal and maternal allele in liver. Following Southern blotanalysis methylation levels were calculated as the intensity of each allele-specific band normalized to the total amount of DNA loaded in the gel. Methylationlevels for bisulphite sequencing are shown as percentage of methylated CpGs. (B) Maternal transmission of the H19D13 knockout results in increased maternalmethylation levels in the DMR1 and DMR2 regions of Igf2. Liver and kidney were collected from E15 H19D13�/þ and H19D13þ/þ mice. DNA was extracted, treatedwith sodium bisulphite, amplified by PCR (see Fig. 1 for amplified region), cloned and sequenced. Maternal and paternal alleles were distinguished using poly-morphisms (Fig. 1). The graphs summarize the methylation levels (overall percentage of methylated CpGs out of total CpGs in analysed region) of the paternallyand maternally inherited alleles of wild-type and H19D13maternal transmission samples (n� 17 for all samples). The levels of maternal methylation are stronglyincreased in both DMR1 and DMR2 following maternal transmission of the H19D13allele. (C) Maternal transmission of H19DDMD results in increased methylationin Igf2 DMR1 and DMR2. Southern blot analysis of methylation levels of DMR1 and DMR2 in kidney and liver of post-natal day 1 H19DDMD�/þ (n¼ 1) comparedwith H19DDMDþ/þ (n¼ 1) mice. For DMR1 methylation analysis genomic DNAs were digested with EcoRI and HpaII and probed with an EcoRI–XbaI probe. ForDMR2 methylation analysis genomic DNAs were digested with BamHI and HpaII and probed with a KpnI–BamHI probe. Graphs represent the methylation levelscalculated as a ratio between density of higher molecular weight bands (for DMR1, h5, h6 and h7; for DMR2, h1, hm and h3) and lower molecular weight bands(for DMR1, h2, h3 and h4; for DMR2, h3 and h4). Note the increase in methylation in kidney for DMR1 and in kidney and liver for DMR2. (D) Maternal trans-mission of Igf2DDMR1-U2 results in methylation changes in the Igf2 DMR2. Southern blot analysis of methylation levels of DMR2 in kidney and liver of E18Igf2DDMR1-U2�/þ (n¼ 2) compared with Igf2DDMR1-U2þ/þ (n¼ 2) mice. Genomic DNAs were digested with BamHI, FspI and EcoNI and probed with aKpnI–BamHI probe. Graphs represent the methylation levels of maternal and paternal alleles in wild-type and mutant offspring, calculated as the density of eachallele-specific band normalized to the total amount of DNA loaded in the gel. Note the significant increase in methylation of the maternal allele in mutants, in bothkidney and liver.

Human Molecular Genetics, 2003, Vol. 12, No. 3 299

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

and from 21 to 58% in liver (P< 0.0001) in DMR2], whereaspaternal transmission of the same allele had little effect on DMR1and 2 (data not shown). Significantly, the maternal increase couldbe directly attributed to the H19 DMR through analysis ofsamples from the H19DDMD knockout (28), which qualitativelygave the same result as the H19D13 allele (Fig. 3C).

The effect of the DMR1 knockout, Igf2DDMR1-U2 (19)(Fig. 1), on methylation of the other regions was similarlytested (Fig. 3D). Maternal transmission of this knockoutresulted in a very substantial increase in methylation in DMR2.A representative CpG (29) showed an increase in both kidney(from 10 to 89%) and liver (from 18 to 65%) tissues, whereaspaternal transmission again had little effect (data not shown).However, DMR1 deletion did not have any effect on themethylation of the H19 DMR (data not shown, Fig. 5). Finally,two different DMR2 deletions (Igf2laczDMR2, Igf2DDMR2) (20)(Fig. 1) did not have any effect on either DMR1 or H19 DMRmethylation, or methylation of the Igf2 promoters or theintergenic regions (data not shown, Fig. 5).

Disruption of the Igf2 gene (Igf2lacz) (19) (Fig. 1) did nothave any significant effect on methylation of the DMRs, theIgf2 promoters, or the intergenic regions (data not shown).Deletion of the endoderm enhancers (30) (Fig. 1) did not haveany effect on methylation of all three DMRs (data not shown),suggesting that methylation and expression can be uncoupled.This conclusion is further supported by the observationmentioned above that the Igf2DDMR2 knockout, in which Igf2transcription is repressed up to 5-fold (20), had no effect onDNA methylation.

Thus, the unmethylated maternal H19 DMR seems to protectDMRs 1 and 2 from methylation, whereas the maternal DMR1protects DMR2 from methylation.

Regional control acts postzygotically

It was important to ask whether maternal methylation in DMRs1 and 2 as a result of H19 DMR deletion arose in germ cellsduring oogenesis, or was a postzygotic event (or both). Thus,oocytes carrying the H19D13 deletion were analysed for DMR1and DMR2 methylation by bisulphite sequencing (Fig. 4A).The results clearly show that DMRs 1 and 2 were unmethylatedin the knockout oocytes, just as in wild-type ones (comparewith Fig. 2). In addition, morulae containing the maternallytransmitted H19D13 allele were unmethylated at DMR1(summarized in Fig. 4B). However, DMR1 and in particularDMR2 show a significant increase in methylation of thematernal H19D13allele by E9 (summarized in Fig. 4B). Maternalmethylation of DMRs in the knockout is therefore a postzygoticevent, which appears to parallel the de novo methylation eventsoccurring on the wild-type paternal allele (Fig. 4B).

Does coordination occur by linear spreading?

The only mechanistic model for coordination of DNAmethylation patterns proposes linear spreading along theDNA (31). We therefore analysed DNA methylation in theIgf2 promoters and in the intergenic regions between the twogenes in all knockouts. The regions were chosen so thatthey contained both relatively methylated areas (Igf2-H19intergenic region, probes A11 and AE31) (32) and relatively

undermethylated areas (Igf2 promoter region and Igf2-H19intergenic region, probe A4) (32) (Fig. 5). No changes wereobserved in any of the knockouts; in particular no changes wereobserved in the Igf2 promoters with maternal transmission ofIgf2DDMR1-U2, which led to increased methylation in DMR2,nor in the Igf2 promoters or the intergenic regions withmaternal transmission of H19D13(data not shown). Theseresults, summarized in Figure 5, show that the methylationchanges observed in the absence of the maternal DMRs werehighly specific to DMRs 1 and 2 and did therefore not arisefrom linear methylation spreading between the two genes, orbetween the two DMRs in Igf2 (Fig. 5).

DISCUSSION

We have carried out the first systematic analysis of coordinationof epigenetic modifications in an imprinting cluster in themouse. The most significant result of this analysis is thatmethylation in three DMRs in the Igf2-H19 region iscoordinately regulated, with the three DMRs acting in ahierarchical fashion to achieve this coordination. Epigeneticcoordination does not occur in the germ line, but happens duringpostimplantation development in a tissue-specific fashion andmay thus be linked to cellular differentiation. Finally, theobservation that coordinated methylation patterns are highlyspecific to DMRs, and do not involve linear spreading along theDNA, suggests that secondary structures in chromatin allowremote DMRs to interact with each other. We propose a model ofhow this might occur. These observations explain how inimprinting clusters the monoallelic expression of many genescan be regulated by local differential methylation, when onlysome genes in the cluster carry methylation imprints that areinherited from the gametes unchanged (2). Our observationsmay also apply to other epigenetically regulated gene clusters(e.g. olfactory receptor and interleukin genes) (33), and topathological situations in which epigenetic regulation isdisrupted by mutations or epimutations.

The three DMRs investigated (DMRs 1 and 2 in Igf2 and theH19 DMR) are fully methylated in sperm, and unmethylated inoocytes. We therefore considered the possibility that all threeare germ line methylation imprints, which are maintainedunchanged in the embryo after fertilization. While this is truefor the H19 DMR, the DMRs in Igf2 lose paternal methylationsoon after fertilization. Previous work on DMR2 has shownthat there is massive loss of paternal methylation in the zygotewhich probably occurs by an active demethylation mechanism(26). Loss of methylation in DMR1 is less dramatic, but thereare clearly some CpG sites that are fully methylated in spermand completely demethylated in zygotes, again indicatingactive removal. In addition, there is more delayed loss as wellover the subsequent cleavage divisions, indicating perhaps arole of passive demethylation. A small amount of de novomethylation occurs on the maternal allele after fertilization, aspreviously observed for DMR2 (26). Thus, by the morula andblastocyst stage all paternal methylation in the Igf2 DMRs iserased, consistent with the proposal that most paternal germline methylation imprints are erased by reprogramming in theearly embryo (34). How the H19 DMR is protected againstdemethylation remains unknown.

300 Human Molecular Genetics, 2003, Vol. 12, No. 3

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

Following implantation and early postimplantation develop-ment, there is de novo methylation of the paternal alleles ofDMR1 and 2. This de novo methylation occurs relatively slowly,and there are lineage specific differences, with earlier methyla-tion of DMR1 than DMR2 in some tissues like the kidney.De novo methylation is likely to be caused by Dnmt3a and b,since DMR2 methylation in ES cells was shown to be dependenton the two de novo methylase enzymes (35). There is also asmall but consistent amount of de novo methylation on thematernal allele, which we interpret as the same de novomethylation process acting on both parental alleles, but thematernal allele being relatively more protected against thismethylation (see below). It is interesting to note that expressionof Igf2 in blastocysts is from both parental alleles, but becomesincreasingly monoallelically paternal after embryo implantation.This parallels a dramatic increase in levels of Igf2 mRNA (36).Thus, a progressive increase of Igf2 expression during earlydevelopment is accompanied by increased DNA methylation inthe paternal copies of DMR1 and 2, which is consistent withtheir functions as silencer and activator, respectively (18–20).

Maternal deletion of the H19 DMR and of DMR1 in Igf2, butof no other cis acting element in the cluster, leads to de novo

methylation of the maternal DMR1 and 2, and of DMR2,respectively. This de novo methylation of the maternal alleleoccurs precisely at the times in development at which thepaternal allele becomes methylated normally (summarized inFig. 4B). Therefore, maternal deletion of the H19 DMR hasremoved the protection against de novo methylation of thematernal DMR1 and 2, and maternal deletion of DMR1 hasremoved protection against methylation of DMR2.

While the methylation state of Igf2 DMRs 1 and 2 isimportant for expression of the gene, as mentioned above, it islikely that the methylation changes observed here are notdirectly caused by changes in expression. As far as the paternalallele is concerned, transcriptional repression of Igf2 byenhancer deletion (30) or DMR2 deletion (20) leaves methyla-tion in DMRs unchanged. On the maternal allele, Igf2DDMR1-U2

causes methylation in DMR2 in all tissues, but expression ofthe maternal Igf2 allele is limited to mesodermal tissues (19).Similarly, maternal transmission of the minute mutation leads tomethylation of DMR1 in all tissues, but the maternal Igf2 alleleis only expressed in endodermal organs (37).

Coordination of methylation in the cluster is highly specific toDMRs, and there are no changes anywhere else between or in the

Figure 4. Developmental timing of methylation changes at Igf2 DMR1 and 2 following maternal transmission of the H19D13 knockout. (A) Bisulphite analysisshows no change of methylation levels in oocytes from H19D13�/� females at Igf2 DMR1 and DMR2. DNA was extracted from oocytes, treated with sodiumbisulphite, amplified by PCR (see Fig. 1 for amplified region), cloned and sequenced. Note low levels of methylation in both samples not significantly differentfrom wild-type ones (compare with Fig. 2). (B) Summary of the effect of maternal transmission of the H19D13allele on methylation at Igf2 DMR1 and DMR2.Graphs show overall percentages of methylation in wild-type (solid lines) and H19D13�/þ (dashed lines) liver samples shown in Figures 2 and 3. Methylationprofiles in H19D13�/þ embryos are shown only when clearly different from wild-type ones. (DMR2 methylation was not analysed in H19D13�/þ embryos atthe morulae stage but we have assumed from the DMR2 oocyte data and DMR1 morulae data that there is no change.) Both DMR1 and DMR2 show differentialmethylation at the E9 and E15 stages in wild-type samples. This differential methylation is lost in H19D13�/þ embryos due to an increase in maternal levels by theE9 stage. This maternal increase occurs at the same time as the establishment of adult differential methylation patterns in wild-type embryos.

Human Molecular Genetics, 2003, Vol. 12, No. 3 301

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

genes, ruling out that it occurs by linear spreading of methylationalong the chromosome. How does coordination occur? It hasbeen previously proposed that there are higher-order chromatininteractions between the H19 and Igf2 genes (38). Thus in ourmodel (Fig. 6) the maternal H19 DMR together with its proteinfactors, including CTCF, may protect DMR1 (and secondarilyDMR2) from de novo methylation. It is interesting to note that notonly maternal deletion of the H19 DMR, but also its maternalmethylation, leads to methylation of DMR1 (K.Davies, unpub-lished data). This model also suggests how the DMR of H19(acting as a chromatin boundary) and the DMR1 in Igf2 mightcooperate in silencing of the maternal Igf2 allele, and thus howboth deletion of the DMR or deletion of DMR1 could disrupt thiscooperation and lead to expression of the maternal Igf2 allele(Fig. 6). Deletion of DMR2, by contrast, does not reactivate thesilent maternal Igf2 (20). The proposal of long-range chromatininteractions is supported by recent findings in the beta globinlocus in which local interactions between the beta major gene andthe LCR were directly demonstrated (39). Long-range chromatinrearrangements in the nucleus are generally thought to underliechanges in gene expression during differentiation, and are oftentissue-specific (40–42), as observed here.

Our results also show that regional coordination by protectionfrom de novo methylation occurs exclusively during postzygoticembryonic development, and does not take place duringoogenesis where deletion of DMRs did not have any effect onregional methylation. Thus the factors that protect from de novomethylation (if any) during oogenesis remain unknown. Therecent observation that deletion of Dnmt3l in oocytes results ininability of a number of maternal methylation imprints to beestablished (43,44) may suggest that demethylation of DMRs inoocytes is a default state, rather than requiring specificprotection. Similarly, the factors leading to methylation of thethree DMRs during spermatogenesis remain unknown. Thatregional coordination does not operate during gametogenesis issignificant in the light of the proposal that the imprinting centrein the Prader–Willi/Angelman syndrome imprinting cluster onchromosome 15q11–13 is required for regional coordination ofmethylation in the germline (8). However, more recently it has

been shown that the IC may operate postzygotically (11), asshown for the Igf2-H19 region here. Thus whether the PWS/ASIC can operate in the germline as proposed is still unknown.

Our work appears to reveal a general principle which isapplicable to other imprinting clusters. Thus, in the PWS/ASorthologous cluster in the mouse, the Snrpn DMR1 (part of theIC) carries a maternal germline imprint, and a number of linkedimprinted genes are maternally methylated and paternallyexpress (45,46). When the DMR1 region is deleted on thepaternal chromosome, DMRs in the linked genes becomemethylated. One of these linked genes, Ndn, has a maternalmethylation imprint in oocytes, which is reprogrammed afterfertilization and re-established during postimplantation devel-opment (47). Interestingly, a higher-order chromatin structureinvolving matrix attachment regions (MARs) has beenproposed to be involved in the regional coordination in thiscluster (48). Although less well characterized, similar situationsare likely to exist in other imprinting clusters (1,49,50) andperhaps also in other gene clusters that involve epigeneticregulation (33). The challenge now is to work out how long-range chromatin interactions are involved in these situations

MATERIALS AND METHODS

Mouse crosses

Wild-type crosses. M.m.domesticus females were mated to SD7males (SD7 is a C57Bl/6 X CBA M.m.domesticus strain contain-ing the distal portion of M.spretus chromosome 7). One-cellzygotes and morulae were collected from super-ovulated females(51). Whole embryos and tissue samples were collected on E9 andE15, respectively. Oocytes were collected from super-ovulatedhomozygous M.m.domesticus females and epididymal spermwas collected from homozygous M.m.domesticus males.

Maternal transmission of H19D13. Homozygous H19D13

females were mated to SD7 males. Tissue samples werecollected from embryos on E15. Oocytes were collected fromsuper-ovulated homozygous H19D13 females.

Figure 5. Summary of methylation analysis in the Igf2-H19 region with maternal transmission of the H19D13, Igf2DDMR1-U2 and Igf2lacZDMR2 alleles. The regionsanalysed are shown; from left to right, DMR1, Igf2 promoter1, Igf2 promoter 3 (3 and 4 CpGs tested, respectively), DMR2, probes A4, A11 and AE31 (4,2 and 7CpGs tested, respectively) (32) and H19 DMR. A ratio of 1 indicates there was no change. Altered methylation patterns are specific to DMR1 and 2.

302 Human Molecular Genetics, 2003, Vol. 12, No. 3

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

Maternal transmission of H19DDMD. Heterozygous femaleswere mated to M.m.domesticus males. Tissue samples werecollected from neonates on the day of birth.

Maternal transmission of Igf2DDMR1-U2. Heterozygousfemales were mated to SD7 males. Tissue samples werecollected from embryos on E18.

Nuclease sensitivity assays. Igf2DDMR1-U2 heterozygous weremated with SD7 males. Tissue samples were collected frommice on post-natal day 21.

For developmental staging purposes, day of conception isconsidered to be day 1 of pregnancy.

Bisulphite sequencing

Genomic DNA was isolated from E15 tissues, E9 wholeembryos, 50–60 morulae and epididymal sperm. The sampleswere digested with Proteinase K (50 mg/ml) for 0.5–3 h.The DNA was then purified by extraction with phenol : chloro-form : isoamyl alcohol, precipitated with 100% ethanol andresuspended in H2O. Yeast tRNA (5 mg) carrier was used in theextraction of DNA from morulae. b-Mercaptoethanol (70 mM)was added to the proteinase K solution to increase yield ofsperm DNA. The DNA was then digested with EcoRI for1–16 h and denatured at 100�C for 5–10 min. NaOH was addedto a final concentration of 0.35 M.

Denatured DNA, 1–25 ng, or whole cells (50–100 oocytes orone-cell zygotes) were embedded in LMP agarose (2%),digested with EcoRI and treated with sodium bisulphitesolution as previously described (52). An H2O-only agarosebead was also treated in each reaction to control forcontamination.

PCR amplification was carried out on single agarose beadswith primers specific for bisulphite-treated DNA. Both DMR1and DMR2 regions of Igf2 were amplified using a nested primerapproach. DMR1 amplifications were performed in a 100 ml

reaction using the primers 50-GGTTAGGTGAAGGTTTT-GTGGGTAGTTATA-30 and 50-ATATTCCCCTTTCAAATTC-CAATCTACATC-30 (94�C for 60 s, 50�C for 120 s, 72�C for180 s for five cycles, then 94�C for 30 s, 50�C for 120 s, 72�Cfor 90 s (þ5 s each cycle) for 25 cycles followed by 72�C for6 min. Two microlitres of the first amplification products wereused in a second reaction with the nested primers 50-GGT-GGTTTTTTAATGGATATTTTAAGGTGA-30 and 50-CCAA-CCTCTATCCCTAACTTTTCTAACCTC using the conditionsabove. Amplification of Igf2 DMR2 was carried out using theprimers and conditions previously described (52).

The resulting PCR products were gel-purified (Qiaquick,Qiagen) and ligated into pCR2.1 (Topo-TA cloning kit,Invitrogen). Individual clones were sequenced using an ABIsequencer. Clones were only accepted with �95% cytosineconversion. The non-converted cytosine residues were usedto ensure each accepted clone originated from a differenttemplate DNA. Two separate sodium bisulphite treatmentswere carried out for the samples shown to verify theresults.

Statistical analysis of bisulphite sequence data

Statistical analysis was carried out in collaboration withE. Walter, Cambridge. The methylation percentages wereobtained for each individual clone within a sample (numberof methylated CpGs per clone divided by the total numberof CpGs per clone). These were then used to calculate theoverall methylation level, standard deviation and standarderror of the mean of each sample. The graphs show theoverall methylation level of each sample. Where error barsare shown, these represent the standard error of the mean.A logistical regression test from the Genstat statisticalpackage was used to test for differences between samples.The samples are considered significantly different whenP< 0.01.

Figure 6. Model for methylation protection of the maternal allele of the Igf2 gene. In the wild-type situation a specific chromatin structure promotes physicalproximity between the H19 DMR and the Igf2 DMR1. The two regions interact, potentially via a complex of proteins (orange oval) that may involve CTCFor GCF2 or both, and DMR1 is thus protected from de novo methylation. As a consequence the unmethylated DMR1 protects DMR2 from de novo methylation,also possibly via the binding of protein complexes. When the H19 DMR is deleted the protection is disrupted and DMR1 and consequently DMR2 become de novomethylated. When DMR1 is deleted the methylation protection over DMR2 is disrupted and this region becomes de novo methylated. The model also shows howthe two knockouts can relieve repression of the maternal allele of Igf2.

Human Molecular Genetics, 2003, Vol. 12, No. 3 303

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

Southern blot analysis

Genomic DNA was extracted from tissues as previouslydescribed (53) and phenol-chloroformed. Twenty microgramsof DNA were digested with 65 units of each restriction enzyme,loaded on a 1% agarose gel, blotted onto Nytran-plusmembrane (Schleicher and Schuell) in alkaline conditions andhybridized with a specific probe. Hybridizations and washeswere performed as described by Church and Gilbert (54).

Quantifications of allelic methylation were performed with aPhosphorimager (Fujifilm FLA-3000 and AIDA software).

ACKNOWLEDGEMENTS

We thank S. Tilghman for making the endoderm enhancer andH19D13 knockouts available to us. We thank G. Kelsey forcomments on the manuscript, K. Davies for communicatingresults prior to publication, and E. Walters for help withstatistical analysis. This work was supported by BBSRC, CRCand MRC. S.L. was in receipt of a Praxis XXI/FCT Scholarship(Portugal).

REFERENCES

1. Ferguson-Smith, A.C. and Surani, M.A. (2001) Imprinting and theepigenetic asymmetry between parental genomes. Science, 293,1086–1089.

2. Reik, W. and Walter, J. (2001) Genomic imprinting: parental influence onthe genome. Nat. Rev. Genet., 2, 21–32.

3. Sleutels, F., Barlow, D.P. and Lyle, R. (2000) The uniqueness of theimprinting mechanism. Curr. Opin. Genet. Devl., 10, 229–233.

4. Tilghman, S.M. (1999) The sins of the fathers and mothers: genomicimprinting in mammalian development. Cell, 96, 185–193.

5. Brannan, C.I. and Bartolomei, M.S. (1999) Mechanisms of genomicimprinting. Curr. Opin. Genet. Devl., 9, 164–170.

6. Xin, Z., Allis, C.D. and Wagstaff, J. (2001) Parent-specific complementarypatterns of histone H3 lysine 9 and H3 lysine 4 methylation at thePrader–Willi syndrome imprinting center. Am. J. Hum. Genet., 69,1389–1394.

7. Li, E., Beard, C. and Jaenisch, R. (1993) Role for DNA methylation ingenomic imprinting. Nature, 366, 362–365.

8. Buiting, K., Saitoh, S., Gross, S., Dittrich, B., Schwartz, S., Nicholls, R.D.and Horsthemke, B. (1995) Inherited microdeletions in the Angelmanand Prader–Willi syndromes define an imprinting centre on humanchromosome 15. Nat. Genet., 9, 395–400.

9. Sutcliffe, J.S., Nakao, M., Christian, S., Orstavik, K.H., Tommerup, N.,Ledbetter, D.H. and Beaudet, A.L. (1994) Deletions of a differentiallymethylated CpG island at the SNRPN gene define a putative imprintingcontrol region. Nat. Genet., 8, 5–7.

10. Reik, W., Brown, K.W., Schneid, H., Le Bouc, Y., Bickmore, W. andMaher, E.R. (1995) Imprinting mutations in the Beckwith-Wiedemannsyndrome suggested by altered imprinting pattern in the IGF2-H19 domain.Hum. Mol. Genet., 4, 2379–2385.

11. Bielinska, B., Blaydes, S.M., Buiting, K., Yang, T., Krajewska-Walasek, M.,Horsthemke, B. and Brannan, C.I. (2000) De novo deletions of SNRPNexon 1 in early human and mouse embryos result in a paternal to maternalimprint switch. Nat. Genet., 25, 74–78.

12. Hark, A.T., Schoenherr, C.J., Katz, D.J., Ingram, R.S., Levorse, J.M.and Tilghman, S.M. (2000) CTCF mediates methylation-sensitiveenhancer-blocking activity at the H19/Igf2 locus. Nature, 405, 486–489.

13. Kanduri, C., Pant, V., Loukinov, D., Pugacheva, E., Qi, C.F., Wolffe, A.,Ohlsson, R. and Lobanenkov, V.V. (2000) Functional association of CTCFwith the insulator upstream of the H19 gene is parent of origin-specific andmethylation-sensitive. Curr. Biol., 10, 853–856.

14. Kaffer, C.R., Srivastava, M., Park, K.Y., Ives, E., Hsieh, S., Batlle, J.,Grinberg, A., Huang, S.P. and Pfeifer, K. (2000) A transcriptional insulatorat the imprinted H19/Igf2 locus. Genes Devl., 14, 1908–1919.

15. Szabo, P., Tang, S.H., Rentsendorj, A., Pfeifer, G.P. and Mann, J.R. (2000)Maternal-specific footprints at putative CTCF sites in the H19 imprintingcontrol region give evidence for insulator function. Curr. Biol., 10,607–610.

16. Bell, A.C. and Felsenfeld, G. (2000) Methylation of a CTCF-dependentboundary controls imprinted expression of the Igf2 gene. Nature, 405,482–485.

17. Drewell, R.A., Brenton, J.D., Ainscough, J.F., Barton, S.C., Hilton, K.J.,Arney, K.L., Dandolo, L. and Surani, M.A. (2000) Deletion of a silencerelement disrupts H19 imprinting independently of a DNA methylationepigenetic switch. Development, 127, 3419–3428.

18. Eden, S., Constancia, M., Hashimshony, T., Dean, W., Goldstein, B.,Johnson, A.C., Keshet, I., Reik, W. and Cedar, H. (2001) An upstreamrepressor element plays a role in Igf2 imprinting. EMBO J., 20,3518–3525.

19. Constancia, M., Dean, W., Lopes, S., Moore, T., Kelsey, G. and Reik, W.(2000) Deletion of a silencer element in Igf2 results in loss of imprintingindependent of H19. Nat. Genet., 26, 203–206.

20. Murrell, A., Heeson, S., Bowden, L., Constancia, M., Dean, W.,Kelsey, G. and Reik, W. (2001) An intragenic methylated region in theimprinted Igf2 gene augments transcription. EMBO Rep.,2, 1101–1106.

21. Feil, R., Walter, J., Allen, N.D. and Reik, W. (1994) Developmental controlof allelic methylation in the imprinted mouse Igf2 and H19 genes.Development, 120, 2933–2943.

22. Moore, T., Constancia, M., Zubair, M., Bailleul, B., Feil, R., Sasaki, H.and Reik, W. (1997) Multiple imprinted sense and antisense transcripts,differential methylation and tandem repeats in a putative imprinting controlregion upstream of mouse Igf2. Proc. Natl Acad. Sci. USA, 94,12509–12514.

23. Forne, T., Oswald, J., Dean, W., Saam, J.R., Bailleul, B., Dandolo, L.,Tilghman, S.M., Walter, J. and Reik, W. (1997) Loss of the maternal H19gene induces changes in Igf2 methylation in both cis and trans. Proc. NatlAcad. Sci. USA, 94, 10243–10248.

24. Olek, A. and Walter, J. (1997) The pre-implantation ontogeny of the H19methylation imprint. Nat. Genet., 17, 275–276.

25. Tremblay, K.D., Duran, K.L. and Bartolomei, M.S. (1997) A 50

2-kilobase-pair region of the imprinted mouse H19 gene exhibitsexclusive paternal methylation throughout development. Mol. Cell. Biol.,17, 4322–4329.

26. Oswald, J., Engemann, S., Lane, N., Mayer, W., Olek, A., Fundele, R.,Dean, W., Reik, W. and Walter, J. (2000) Active demethylation of thepaternal genome in the mouse zygote. Curr. Biol., 10, 475–478.

27. Leighton, P.A., Ingram, R.S., Eggenschwiler, J., Efstratiadis, A. andTilghman, S.M. (1995) Disruption of imprinting caused by deletion ofthe H19 gene region in mice. Nature, 375, 34–39.

28. Thorvaldsen, J.L., Duran, K.L. and Bartolomei, M.S. (1998) Deletion ofthe H19 differentially methylated domain results in loss of imprintedexpression of H19 and Igf2. Genes Devl., 12, 3693–3702.

29. Weber, M., Milligan, L., Delalbre, A., Antoine, E., Brunel, C., Cathala, G.and Forne, T. (2001) Extensive tissue-specific variation of allelicmethylation in the Igf2 gene during mouse fetal development: relation toexpression and imprinting. Mech. Devl., 101, 133–141.

30. Leighton, P.A., Saam, J.R., Ingram, R.S., Stewart, C.L. and Tilghman, S.M.(1995) An enhancer deletion affects both H19 and Igf2 expression. GenesDevl., 9, 2079–2089.

31. Turker, M.S. and Bestor, T.H. (1997) Formation of methylation patterns inthe mammalian genome. Mutat. Res., 386, 119–130.

32. Koide, T., Ainscough, J., Wijgerde, M. and Surani, M.A. (1994)Comparative analysis of Igf-2/H19 imprinted domain: identification of ahighly conserved intergenic DNase I hypersensitive region. Genomics,24, 1–8.

33. Ohlsson, R., Paldi, A. and Graves, J.A. (2001) Did genomic imprintingand X chromosome inactivation arise from stochastic expression? TrendsGenet., 17, 136–141.

34. Reik, W. and Walter, J. (2001) Evolution of imprinting mechanisms: thebattle of the sexes begins in the zygote. Nat. Genet., 27, 255–256.

35. Okano, M., Bell, D.W., Haber, D.A. and Li, E. (1999) DNAmethyltransferases Dnmt3a and Dnmt3b are essential for de novomethylation and mammalian development. Cell, 99, 247–257.

36. Szabo, P.E. and Mann, J.R. (1995) Allele-specific expression and totalexpression levels of imprinted genes during early mouse development:implications for imprinting mechanisms. Genes Devl., 9, 3097–3108.

304 Human Molecular Genetics, 2003, Vol. 12, No. 3

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from

37. Davies, K., Bowden, L., Smith, P., Dean, W., Hill, D., Furuumi, H.,Sasaki, H., Cattanach, B. and Reik, W. (2002) Disruption of mesodermalenhancers for Igf2 in the minute mutant. Development,129, 1657–1668.

38. Banerjee, S. and Smallwood, A. (1995) A chromatin model of IGF2/H19imprinting. Nat. Genet., 11, 237–238.

39. Carter, D., Chakalova, L., Osborne, C.S., Dai, Y. and Fraser, P. (2002)Long-range chromatin regulatory interactions in vivo. Nat. Genet., 32, 623–626.

40. Fisher, A.G. and Merkenschlager, M. (2002) Gene silencing, cell fate andnuclear organisation. Curr. Opin. Genet. Devl., 12, 193–197.

41. Mahy, N.L., Perry, P.E., Gilchrist, S., Baldock, R.A. and Bickmore, W.A.(2002) Spatial organization of active and inactive genes and noncodingDNA within chromosome territories. J. Cell Biol., 157, 579–589.

42. Carmo-Fonseca, M. (2002) The contribution of nuclear compartmentali-zation to gene regulation. Cell, 108, 513–521.

43. Hata, K., Okano, M., Lei, H. and Li, E. (2002) Dnmt3L cooperates withthe Dnmt3 family of de novo DNA methyltransferases to establishmaternal imprints in mice. Development, 129, 1983–1993.

44. Bourc’his, D., Xu, G.L., Lin, C.S., Bollman, B. and Bestor, T.H. (2001)Dnmt3L and the establishment of maternal genomic imprints. Science,294, 2536–2539.

45. Nicholls, R.D., Saitoh, S. and Horsthemke, B. (1998) Imprinting inPrader–Willi and Angelman syndromes. Trends Genet., 14, 194–200.

46. Shemer, R., Birger, Y., Riggs, A.D. and Razin, A. (1997) Structure of theimprinted mouse Snrpn gene and establishment of its parental-specificmethylation pattern. Proc. Natl Acad. Sci. USA, 94, 10267–10272.

47. Hanel, M.L. and Wevrick, R. (2001) Establishment and maintenance ofDNA methylation patterns in mouse Ndn: implications for maintenanceof imprinting in target genes of the imprinting center. Mol. Cell. Biol.,21, 2384–2392.

48. Greally, J.M. (1999) Genomic imprinting and chromatin insulation inBeckwith–Wiedemann syndrome. Mol. Biotechnol., 11, 159–173.

49. Zwart, R., Sleutels, F., Wutz, A., Schinkel, A.H. and Barlow, D.P. (2001)Bidirectional action of the Igf2r imprint control element on upstream anddownstream imprinted genes. Genes Devl., 15, 2361–2366.

50. Liu, J., Litman, D., Rosenberg, M.J., Yu, S., Biesecker, L.G.and Weinstein, L.S. (2000) A GNAS1 imprinting defect inpseudohypoparathyroidism type IB. J. Clin. Invest., 106, 1167–1174.

51. Hogan, B., Beddington, R., Costantini, F. and Lacy, E. (1994) Manipulatingthe Mouse Embryo. A Laboratory Manual. Cold Spring Harbor LaboratoryPress, Cold Spring Harbor, NY.

52. Olek, A., Oswald, J. and Walter, J. (1996) A modified and improved methodfor bisulphite based cytosine methylation analysis. Nucleic Acids Res.,24, 5064–5066.

53. Laird, P.W., Zijderveld, A., Linders, K., Rudnicki, M.A., Jaenisch, R.and Berns, A. (1991) Simplified mammalian DNA isolation procedure.Nucl. Acids Res., 19, 4293.

54. Church, G.M. and Gilbert, W. (1984) Genomic sequencing. Proc. NatlAcad. Sci. USA, 81, 1991–1995.

55. Warnecke, P.M., Mann, J.R., Frommer, M. and Clark, S.J. (1998)Bisulfite sequencing in preimplantation embryos: DNA methylationprofile of the upstream region of the mouse imprinted H19 gene. Genomics,51, 182–190.

Human Molecular Genetics, 2003, Vol. 12, No. 3 305

by guest on June 9, 2013http://hm

g.oxfordjournals.org/D

ownloaded from