Familial aggregation of abnormal methylation of parental alleles at the IGF2/H19 and IGF2R...

Familial aggregation of abnormal methylationof parental alleles at the IGF2/H19 and IGF2Rdifferentially methylated regions

Ionel Sandovici1, Mark Leppert2, Patricia Red Hawk3, Alexander Suarez4, Yendi Linares5 and

Carmen Sapienza1,6,*

1Fels Institute for Cancer Research and Molecular Biology, Temple University School of Medicine, 3307 North Broad

Street, Philadelphia, PA 19140, USA, 2Eccles Institute of Human Genetics, University of Utah, 15 N 2030 E, Salt Lake

City, UT 84112, USA, 3Temple University School of Medicine, 3400 North Broad Street, Philadelphia, PA 19140, USA,4Bucknell University, 701 Moore Avenue, Lewisburg, PA 17837, USA, 5Phillips Academy, 180 Main Street, Andover,

MA 01810, USA and 6Department of Pathology and Laboratory Medicine, Temple University School of Medicine, 3307

North Broad Street, Philadelphia, PA 19140, USA

Received March 5, 2003; Revised and Accepted April 28, 2003

Loss of imprinting (LOI) has been observed in many types of human tumors and may be a predisposing eventin some colon cancers. LOI is strongly associated with alteration of normal DNA methylation patterns indifferentially methylated regions (DMRs) of affected loci but it is not known whether LOI is caused bystochastic, environmental or genetic factors. We have developed a simple, quantitative assay formeasurement of allelic methylation ratios based on methylation-sensitive restriction endonuclease digestionof genomic DNA and ‘hot-stop’ PCR. We examined allelic methylation ratios at DMRs within the IGF2/H19- andIGF2R-loci in a panel of 48 three-generation families. We observed familial clustering of individuals withabnormal methylation ratios at the IGF2/H19 DMR, as well as stability of this trait over a period of nearly twodecades, consistent with the possibility that constitutional LOI at this locus is due largely to genetic factors.At the IGF2R DMR, we observed more variability in the allelic methylation ratios over time but also observedfamilial clustering of abnormal methylation ratios. Overall, our observations at IGF2R suggest that sharedgenetic factors are responsible for a major fraction of inter-individual variability in parental origin-dependentepigenetic modifications. However, temporal changes also occur in isolated cases, as well as within multipleindividuals in the same family, indicating that environmental factors may also play a role.

INTRODUCTION

Genomic imprinting is a differential epigenetic modification ofchromosomes that may result in the parent-of-origin specifictranscription of some genes. DNA methylation (reviewed in 1),as well as various histone modifications (methylation, acetyla-tion, phosphorylation and ubiquitination) (reviewed in 2,3), arelikely to be important components of the somatically heritableimprint.

The great majority of transcriptionally imprinted genes showallelic differences in DNA methylation at so-called ‘differen-tially methylated regions’ (DMRs). ‘Loss of imprinting’ (LOI),or relaxation of imprinting (4,5), in which a gene transcribednormally from only one parent’s allele is transcribed from both

alleles, has been observed in both pediatric and adult cancers(reviewed in 6), as well as normal tissues (7). There is a strongassociation between LOI and altered DNA methylation patternsat DMRs.

A number of studies suggest that there is inter-individualvariability in allele-specific expression of some imprintedgenes. For example, the IGF2R gene, although differentiallymethylated on maternal and paternal alleles, is transcribed fromboth alleles in the vast majority of the population (8). However,a subset of individuals express predominantly or only thematernal allele (9). The WT1 gene has been reported to betranscribed from both alleles in the majority of placentae butplacentae from some individuals express only the maternalallele (10). Further evidence for inter-individual variability in

*To whom correspondence should be addressed. Tel: þ1 2157077373; Fax: þ1 2157071454; Email: [email protected]

Human Molecular Genetics, 2003, Vol. 12, No. 13 1569–1578DOI: 10.1093/hmg/ddg167

Human Molecular Genetics, Vol. 12, No. 13 # Oxford University Press 2003; all rights reserved

imprinting comes from the observation of significant biallelicexpression of the insulin-like growth factor II gene (IGF2) inabout 10% of normal individuals (11). LOI at IGF2 also occursin tumor tissue of 31% (12) to 44% (7) of sporadic coloncancer cases. Surprisingly, LOI is not thought to be a tumor-specific event in these cases, but is likely to be constitutionalbecause both alleles of IGF2 are expressed in the normalcolonic mucosa of 90% of these patients and also in thelymphocytes of one-third of these same individuals (7).

In the present study, we have attempted to determinewhether stochastic, environmental or genetic factors areresponsible for the potentially large population of ‘normal’individuals (11) who have altered genome imprints. We havedeveloped an assay based on methylation-sensitive restrictionendonuclease digestion of genomic DNA and ‘hot-stop’ PCR.We used this assay to examine inter-individual variability inmethylation at CpG sites within the IGF2/H19 DMR and theIGF2R DMR among a population of 680 individuals from 48three-generation families. We also examined intra-individualvariability in this phenotype by comparing peripheral bloodDNA samples taken nearly two decades apart from asubpopulation of the same individuals as a way of detectingthe influence of environmental or stochastic factors onparental origin-dependent epigenetic modifications over thistime period.

RESULTS

Familial aggregation of individuals with abnormalmethylation ratio at the IGF2/H19-DMR

Because we wished to screen a population of moderate size(680 individuals) for LOI using archival material, we developedan assay based on the correlation between parent of origin-specific transcription and methylation of CpG sites withinDMRs (Fig. 1). Briefly, CpG sites within DMRs on the paternalallele at the IGF2/H19 locus are normally methylated, whilethose on the maternal allele are normally unmethylated (13).There are several DMRs at this locus but the most consistentobservations indicating a role in the control of transcription ofthe IGF2 and H19 genes involve a CpG island located in a 5 kbregion centromeric to the H19 gene (14). This is also the regionthat contains seven different binding sites for the CTCF protein(15) and the methylation status of the sixth binding site wasfound to be most consistently associated with the transcrip-tional status of both IGF2 and H19 (16). We used a singlenucleotide polymorphism (a C/T polymorphism within a CfoIsite, Fig. 1A) and pedigree analysis to identify maternal andpaternal alleles of heterozygous individuals and a methylation-sensitive restriction endonuclease (MluI and, in selected cases,MaeII) to determine the methylation status of specific CpGsites within the DMR. If all paternal alleles are methylated andall maternal alleles are unmethylated at these sites in a sampleof genomic DNA, then all maternal alleles should be cleavedby MluI while all paternal alleles will remain uncleaved.Amplification of the region by PCR using primers that flank theMluI site should amplify only paternal alleles (identified bypost-PCR cleavage with CfoI). Amplification of maternalalleles indicates resistance to cleavage by MluI. This may

occur as a result of methylation of the CpG site within the MluIrecognition sequence (the principle upon which the assay isbased), mutation of the MluI site or technical artifact. The lattertwo possibilities may be distinguished from the first by DNAsequencing, assay reproducibility and use of additionalmethylation-sensitive restriction endonucleases.

We identified 318 individuals (46.8%) who were hetero-zygous at the CfoI polymorphism (Fig. 1A) within the IGF2/H19 DMR and were able to determine unequivocally theparental origins of the alleles in 163 cases. Because previousstudies have indicated that LOI is a quantitative trait (7),i.e. expression of the maternal IGF2 allele is variable and maynot occur to the same extent in all individuals identified ashaving LOI, we calculated the ratio between the DNAmethylation levels on maternal and paternal (M/P) alleles asan indicator of imprinting status (see Materials and Methods).A ratio of zero corresponds to exclusive methylation of theMluI site on the paternal allele, while a ratio of one signifiesmethylation of this site on an equal number of maternal andpaternal alleles.

The distribution of individual M/P methylation ratios(Fig. 2A) shows that the great majority of individuals have amethylated paternal allele, with no detectable methylation ofthe maternal allele. (Because none of the 163 individuals forwhom we could determine parental origin of alleles by pedigreeanalysis had a maternal allele that was more methylated thanthe paternal allele, we have assumed that the less methylatedallele is maternal in the remainder of cases compiled in

Figure 1. (A) The IGF2/H19 DMR is located between the H19 gene and IGF2and is normally methylated only on the paternal allele (13). A C/T polymor-phism located at the sixth binding site for CTCF is recognized by CfoI.MaeII and MluI are methyl-sensitive restriction endonucleases. F3* is theprimer that is 50 end labeled for use in the ‘hot-stop’ PCR cycle. (B) TheIGF2R DMR is located in the second intron and is normally methylated onthe maternal allele. The C/T polymorphism is recognized by MspI. NotI andXhoI are methyl-sensitive restriction endonucleases. F4* is the primer that is50 end labeled in the ‘hot-stop’ PCR cycle.

1570 Human Molecular Genetics, 2003, Vol. 12, No. 13

Fig. 2A.) However, 21 individuals (6.6%) have M/P ratiosgreater than 0.1. We obtained similar ratios after digestion ofgenomic DNA with either MluI or MaeII, indicating thatmethylation of the maternal allele in these individuals is notrestricted to the MluI site. We excluded the possibility that asubpopulation of lymphocytes from these individuals carrieda mutation at the MluI or MaeII sites by DNA sequencingof the amplified maternal allele (Fig. 2C). An additional 28individuals exhibit lower, but reproducible, levels of methyla-tion on the maternal allele (M/P ratios between 0.05 and 0.1).While some of these individuals occur as isolated cases in somefamilies, a significant proportion are found within families inwhich other members also have methylation of maternal alleles.Nine out of the 21 individuals (42.8%) with M/P ratios over 0.1are found within two families (Fig. 3). In one of these families,six of 10 informative individuals have M/P ratios over 0.1 (forH0: six ‘abnormal individuals’ occur among 10 informativeindividuals by chance, given 6.6% population frequency, w2

with Yates correction¼ 20.2, P< 0.001). When a lower level ofmaternal allele methylation is considered (an M/P ratio over0.05), the familial aggregation of individuals is even morepronounced, with 34 out of 49 (69.4%) individuals having at

least one relative with this phenotype. Moreover, not everyindividual in these families is informative, and the true numberof cases in which only a single family member is affected maybe overestimated. We also note that 10 of the 15 individualswho have no relatives with methylation of the maternal alleleare in the first generation of their pedigrees. In these cases, nosibling, and only a single offspring (either the mother or thefather of the third generation), is available for analysis.

Familial aggregation of individuals with altereddifferential methylation at the IGF2R DMR

Although the human IGF2R gene is normally expressed fromboth alleles, differential methylation of maternal and paternalalleles is maintained in the human, as it is in the mouse (17),and some humans do appear to have transcriptional imprintingof IGF2R (9). We designed an assay (Fig. 1B) to examinerelative levels of methylation at CpG sites on maternal andpaternal alleles within the DMR of IGF2R (located in thesecond intron) using the same principles upon which the IGF2/H19 assay was based. We used a C/T polymorphism withinan MspI site to distinguish heterozygous individuals, pedigree

Figure 2. (A) The distribution of DNA methylation ratios at IGF2/H19 DMR for 318 informative individuals. M/P represents the DNA methylation ratio betweenmaternal and paternal alleles. (B) Lane 0 shows a control result obtained by a 1:1 mix of DNAs from individuals homozygous for the C allele and T allele (ratioT/C¼ 1). Lanes 1–3 depict informative individuals with exclusive paternal methylation, while lanes 4–6 show examples of informative individuals with methyla-tion at the MluI site on both paternal and maternal alleles. (C) DNA sequencing of an amplified maternal allele with abnormal methylation (postcleavage with MluIenzyme) that shows the absence of any point mutation at MluI site.

Human Molecular Genetics, 2003, Vol. 12, No. 13 1571

analysis to determine maternal and paternal alleles andcleavage of genomic DNA by a methylation-sensitive restrictionendonuclease followed by PCR using primers that flank thecleavage site to distinguish methylation of CpG sites onmaternal and paternal alleles.

We observed 165 individuals (24.3%) who were informativefor the MspI polymorphism found within the IGF2R DMR andwe identified maternal and paternal alleles for 112 individualsby pedigree analysis. Although maternal and paternal alleleswere distinguishable in almost all individuals on the basis ofexpected preferential methylation of the maternal allele, thedistribution of paternal/maternal (P/M) methylation ratios atIGF2R DMR shows that, unlike the IGF2/H19 DMR, mostindividuals have a measurable level of methylation at the CpGwithin the XhoI cleavage site on the paternal allele (P/M> 0.1,Fig. 4). P/M ratios were similar for both XhoI and NotIrestriction endonuclease digestion of genomic DNA (correla-tion coefficient for paired measurements r¼ 0.715) indicating,again, that the methylation of CpG sites within the paternalDMR is not restricted to a single site.

Because the distribution in Figure 4A is two-tailed (unlikethe distribution of P/M methylation ratios at IGF2/H19 DMR),we examined the distribution of individuals from each tailamong the 48 families. Because the vast majority of individualshave some methylation of the paternal allele, this modificationseems unlikely to inhibit transcription of the normallytranscribed paternal allele. However, it is possible thatindividuals who have little or no methylation of the paternalallele are those individuals in which the maternal allele ispredominately or exclusively transcribed, as reported by Xuet al. (9). We observed 19 individuals (11.5%) with P/M ratiosunder 0.1. Fourteen of these 19 individuals are distributedamong only five families, with five individuals being found in asingle family (Fig. 5) (for H0: five ‘abnormal individuals’ occuramong seven informative by chance, given 11.5% populationfrequency, w2 with Yates correction¼ 10.5, P< 0.01).

We also identified five individuals in which the CpG sitebeing examined was methylated on approximately equalnumbers of maternal and paternal alleles (P/M> 0.8). Two ofthese individuals are siblings within the same family (pedigreenot shown).

Familial aggregation of individuals with abnormalmethylation of IGF2/H19 and IGF2R DMRs;genetics or environment?

The familial clustering of individuals with abnormal allelicmethylation ratios may be explained by the action of sharedenvironmental factors or shared genetic factors. We haveattempted to distinguish between these possibilities by geneticlinkage analysis in those families in which multiple membersare affected and by examining the temporal stability of theallelic methylation ratio over a period of nearly two decades.

Genetic linkage analysis of the abnormal methylation ratiophenotype. The methylation of CpG sites within DMRs isone component of the overall chromatin structure of imprintedregions of chromosomes. Abnormal methylation of normallyunmethylated CpG sites could result from some aspect of thestructure of the particular chromosome in which it is found(an effect in cis) or result from the action of modifying geneticfactors that do not reside in the vicinity of the DMR (an effectin trans). We may gain supporting evidence for the involvementof cis-acting factors by determining whether family memberswho share abnormal allelic methylation ratios also share thesame chromosome haplotype in the vicinity of the DMR.

The assays used in this study identify abnormal methylationof the maternal IGF2/H19 DMR and abnormal methylation ofthe paternal IGF2R DMR. Therefore, we determined whichmaternal chromosome 11p15.5 was inherited by individuals infamilies in which more than one individual had an abnormalmethylation ratio at the IGF2/H19 DMR and which paternalchromosome 6q26 was inherited by individuals in families inwhich more than one individual had an abnormal methylationratio at the IGF2R DMR (see Materials and Methods).

We identified only one family (data not shown) in which allindividuals with abnormal IGF2/H19 DMR methylation ratiosshared the same maternal chromosome 11p15.5 haplotype.Cis-acting genetic factors may be involved in the abnormalmethylation of maternal IGF2/H19 alleles in this family.However, some ‘normal’ individuals in this family inherited thesame maternal chromosome 11p15.5 as did those individualswith abnormal methylation ratio so that there were no families

Figure 3. Familial aggregation of individuals with a significant level of methylation on the maternal allele at IGF2/H19 DMR. The numbers depicted under eachinformative individual represent the value of M/P ratio. (A) A family with both parents homozygous (CC or TT) and uninformative for the DNA methylation assayand all the siblings heterozygous (C/T) and informative. Only some of the siblings exhibit an abnormal methylation pattern. (B) Family with three cases of abnor-mal methylation ratios (maternal grandfather, maternal grandmother and mother). None of the informative siblings exhibit abnormal methylation on the maternalallele.


in which there was perfect concordance between methylationratio phenotype and chromosome 11p15.5 haplotype. Therewere no families in which all individuals with abnormalmethylation ratios at the IGF2R DMR shared the same paternalchromosome 6q26 haplotype.

We have not attempted to perform linkage analysis ofcandidate trans-acting genetic factors because the numberof potential candidate genes is large and the likelihood ofdetecting spurious associations within any one of the families ishigh. However, the chromosome haplotype analyses we haveperformed indicate that if familial clustering of the abnormalmethylation ratio trait is due to genetic factors, these factorsmust often act in trans.

Temporal stability of allelic methylation ratio. In the absenceof a hypothesis on the specific identity of an environmentalagent that might influence abnormal methylation of DMRs(and associated hypotheses on how and when such an agent

might act), we sought to obtain evidence on whether any envir-onmental factor might be involved in the generation of abnor-mal methylation ratios by comparing the allelic methylationratios of the same individual at two different times. Thisapproach is based on the premise that, if an environmental fac-tor causes an alteration of the allelic methylation ratio, then onepredicts a change in this phenotype between the time of estab-lishment of the original, ‘normal’ ratio and the time at whichthe environmental factor exerts its effect.

To test the stability of the allelic methylation ratio over time,two sets of peripheral blood DNA samples taken from the sameindividuals (the first taken between 1982 and 1985 and thesecond in 2001) were analyzed. Both ‘old’ and ‘new’ sampleswere available for 133 individuals informative for the CfoIpolymorphism at IGF2/H19 DMR and for 75 individualsinformative for the MspI polymorphism at IGF2R DMR.

At the IGF2/H19 DMR, only three individuals sampled atboth times (individual nos 12, 100 and 125 in Fig. 6A) had M/Pallelic methylation ratios of greater than 0.1 and theseindividuals had a ratio of greater than 0.1 at both samplingtimes. In fact, there was no case in which the methylation ratioat the IGF2/H19 DMR changed substantially between the twosampling times (Fig. 6A). Because the level of methylation onthe maternal allele at the IGF2/H19 DMR locus was very closeto zero for the majority of individuals in both ‘old’ and ‘new’DNA samples, we did not attempt to assign statisticalsignificance to small differences between individual pairedresults. For the data set, as a whole, a paired t-test indicates thatthere is no significant difference between the two time points(P¼ 0.1051).

At the IGF2R DMR, we observed eight cases in which achange in methylation ratio [only changes greater than threetimes the level of assay reproducibility (�0.067, see Materialsand Methods) were considered as significant for this purpose]moved an individual from a normal to an abnormal category orvice versa. One individual (individual no. 43 in Fig. 6B) had aP/M methylation ratio of 0.2 in the first sample and this ratiowas nearly one in the second sample. Two unrelated individuals(nos 32 and 38 in Fig. 6B) had P/M methylation ratios of lessthan 0.1 in the first sample and ratios of 0.26 and 0.37,respectively, in the second sample. Five individuals (individualnos 52, 54, 55, 57, and 61 in Fig. 6B) exhibited potentiallysignificant changes in the opposite direction, i.e. attaining a

Figure 4. (A) The distribution of DNA methylation ratios at IGF2R DMR for165 informative individuals. P/M represents the DNA methylation ratiobetween paternal and maternal alleles. (B) Lane 0 represents a control obtainedfrom a 1:1 mixture of DNAs from individuals homozygous for the C allele andT allele respectively (ratio T/C¼ 1). Lanes 1–3 depict informative individualswith nearly exclusive maternal methylation, while lanes 4–6 are examples ofinformative individuals with a significant level of methylation on both maternaland paternal alleles.

Figure 5. Familial aggregation of individuals with abnormal allelic methylationratios at the IGF2R DMR. There are both homozygous individuals (CC or TT),uninformative for the DNA methylation assay and heterozygous individuals(C/T), informative for the assay. All the siblings that are informative exhibita low level of methylation on the paternal allele, while informative individualsfrom the first and the second generations exhibit ‘normal’ P/M ratios.


P/M methylation ratio of less than 0.1 in the second sample.Three of these individuals are from the same family (a motherand two offspring, pedigree not shown), suggesting a commonenvironmental effect on this phenotype or a genetic predis-position to the observed change.

Overall, we observed very similar results for the two setsof measurements (correlation coefficient for paired measure-ments r¼ 0.7821) and the slope of the linear regression line(P/M methylation ratio at time 1 versus P/M methylation ratioat time 2) is 0.848� 0.079, near the expected value of 1.0 if nochange in methylation ratio occurred in the interval betweencollection of the two samples. These data indicate that,although changes in allelic methylation ratio can be observed,this phenotype is temporally stable in the majority of thepopulation.

Bisulfite sequencing analysis of individuals withabnormal pattern of DNA methylationat IGF2/H19-DMR

DNA sequence analysis of bisulfite treated DNA (18,19)has become the standard technique for determination of themethylation status of CpG sites in imprinted regions. Thetechnique has the advantage that all CpG sites are amenable toanalysis but has the disadvantage that it is labor-intensive andrelatively expensive to use as a screening tool. In addition, thenumber of individual DNA molecules that must be sequencedto demonstrate a significant difference between individualswith a M/P allelic ratio of zero and individuals with a M/Pallelic ratio of 0.1, for example, is much greater than the 10–20molecules normally assayed in this procedure. These con-siderations led us to attempt this method only in those familiesin which we observed abnormal methylation ratios using themethylation-sensitive restriction endonuclease-based method.

As expected, we observed DNA molecules with an abnormal‘mixed’ methylation pattern for many of these individuals(Fig. 7). However, we also analyzed a series of three singlenucleotide polymorphisms present in the sequenced region ofthe individual shown in Fig. 7, as well as additional abnormaland informative individuals, in order to confirm the parentalorigin of the abnormal alleles. We found that the majorityof these mixed methylation pattern molecules result frommaternal/paternal heteroduplex products that are created duringthe PCR step of the assay and then repaired, randomly, by thebacterial mismatch repair system during the cloning procedure.We confirmed the potential for artifactual origin of moleculeswith a mixed methylation pattern by mixing the DNAs ofindividuals who are homozygous for different SNP haplo-types and detecting individual molecules with both mixedSNP haplotypes and mixed maternal/paternal methylationpatterns. Because we were unable to establish conditions inwhich such heteroduplex molecules could be completelyeliminated [including techniques designed to maintain favor-able primer: template ratios (20) and techniques for eliminatingheteroduplexes using enzymatic digestion with T7 endo-nuclease I (21); data not shown], we did not pursue thismethod of analysis further.

DISCUSSION

Characterization of variability in imprinting phenotypeusing allele-specific methylation assays

We have examined the relative level of methylation of CpGsites on maternal and paternal alleles within differentiallymethylated regions of the IGF2/H19 and IGF2R loci as asurrogate measure of imprinting. The distributions of thesequantitative measures among informative individuals indicates

Figure 6. (A) Longitudinal study of M/P methylation ratios at IGF2/H19 DMR for 133 informative individuals and (B) longitudinal study of P/M methylationratios at IGF2R DMR for 75 informative individuals. Each vertical lane corresponds to an individual. Represented with circles are the results for samples obtainedbetween 1982 and 1985 and with squares are the results for samples obtained in 2001.


that approximately 6.5% of the population has a significantfraction (>10%) of peripheral blood lymphocytes that havemethylation of CpG sites on the maternal IGF2/H19 allele.We note that this estimate is similar to the fraction of thepopulation designated as having LOI (10%) after measuringtranscription of maternal and paternal IGF2 alleles (11).The distribution of P/M allelic methylation at IGF2R is morevariable than that at IGF2/H19. Approximately 11% of thepopulation may be described as having abnormally low levelsof methylation on the paternal allele (P/M allele ratios of <0.1)while a smaller fraction of individuals (about 3%) have P/Mallelic methylation ratios greater than 0.8. Strong familialclustering of individuals from the tails of both distributionsis observed (Figs 3 and 5), indicating that shared geneticfactors or shared environmental factors are responsible for themajority of abnormal allelic methylation ratios (data summar-ized in Table 1).

We attempted to obtain evidence that allelic methylationratios may be affected by environmental factors by examiningsamples taken from the same individuals nearly 20 years apart.We did not detect significant changes at the IGF2/H19 DMR inany of the 133 individuals examined. Although we cannoteliminate the possibility that shared environmental factors mayhave acted before the first peripheral blood lymphocyte samplewas taken, we suggest that familial clustering of abnormalmethylation ratios at the IGF2/H19 DMR is most likely due toshared genetic factors. At the IGF2R DMR, significant changesin methylation ratio did occur in eight of 75 informative

individuals. Three of these ‘changed’ individuals are from thesame family, indicating that this acquired phenotype can alsocluster in families and such familial clustering may be dueto shared environmental factors or a genetic predisposition totemporal or environmentally induced changes.

We note that the restriction endonuclease/PCR-based assayused in our experiments provides a ratio of the number ofmaternal/paternal alleles at which a particular CpG site ismethylated rather than the absolute number of DNA moleculeson which a CpG site is methylated/unmethylated. For example,a sample in which neither the maternal nor the paternalIGF2/H19 DMR was methylated in 90% of cells but in whichthe maternal IGF2/H19 DMR was methylated in 10% of theremaining 10% of cells would give the same ratio as a samplein which the paternal DMR was methylated in all cells andthe maternal DMR was methylated in 10% of all cells. Thesetwo possibilities may be distinguished by bisulfite sequencingbecause the vast majority of DNA molecules, be they paternalor maternal, will have the unmethylated pattern in the first case.We did not encounter such a situation (or the reverse) in any ofthe 10 ‘abnormal’ or four ‘normal’ samples that we examinedby bisulfite sequencing. Nearly equal numbers of moleculeswith methylated and unmethylated patterns were found(disregarding molecules arising from repair of heteroduplexes).Corroborating evidence in favor of the second possibility wasobtained by real-time PCR (see Materials and Methods) onseveral additional individuals, comparing methylation sensitiverestriction endonuclease cleaved DNA with uncleaved DNA(data not shown).

Bisulfite sequencing assay versus methyl-sensitiverestriction endonuclease assay

Bisulfite sequencing has become the standard method forsingle-molecule and single-base resolution analysis of DNAmethylation (18,19). The main advantage of this method isthe ability to examine the methylation of many CpG sites ona single DNA molecule. However, this method has somelimitations (reviewed in 22,23), especially in situations inwhich inter-cellular or inter-individual variability itself is ofinterest. The most obvious disadvantage is the practicallimitation on the number of chromosomes that can be analyzed.Sequencing of 20 DNA molecules from each individual, forexample, yields an estimate of epigenetic variability based on asample of only 10 cells.

An additional problem is the generation of heteroduplexartifacts during the amplification of mixed templates usinguniversal primers. After bisulfite treatment a high diversity oftemplate molecules will occur as the result of different genomicDNA methylation patterns between the parental chromosomesas well as differences in methylation of specific sites betweencells. The presence of native single-nucleotide polymorphisms(SNPs) and any errors induced by the DNA polymerase willincrease further the potential for heteroduplexes. When such aheteroduplex molecule is cloned, the host’s mismatch repairsystem can convert it into a single hybrid sequence (reviewed in24). As the repair enzymes cannot identify a parent strand,either strand is chosen as a template for the synthesis of the‘repaired’ strand. The repaired sequence will be a ‘mosaic’,i.e. composed of portions of the two parent heterologous

Figure 7. Bisulfite sequencing results for a 350 bp fragment of the IGF2/H19-DMR containing 22 CpG dinucleotides. Filled circles represent methylatedcytosines, normally found on the paternal chromosomes (P) while open circlesrepresent unmethylated cytosines, normally found on the maternal chromo-somes (M). The MluI and MaeII sites are indicated and the CTCF-bindingregion is boxed. There are three SNPs that permit differentiation of the parentalorigin of the alleles: a C/T polymorphism—at the CfoI site, utilized for themethylation assay (nt. 7966, GenBank AF125183), a C/A polymorphism (nt8008) and a G/A polymorphism (nt 8097). By pedigree analysis the haplotypesof the two parental alleles were established to be: C-C-A (for paternal chromo-somes) and T-A-G (for maternal chromosomes). After bisulfite treatment, theC/T polymorphism (which is located within a CpG dinucleotide) is no longerobligately informative because an unmethylated C will be converted in T.The figure exhibits a significant number of alleles with a ‘mosaic’ pattern ofmethylation (*). The two chromosomes identified by arrows proved to berecombinants (C-A-G haplotype for the first chromosome and T-A-A haplotypefor the second chromosome) as a consequence of an in vivo repair process ofheteroduplexes during the subcloning step. For the other two chromosomeswith a mosaic pattern of methylation, the haplotypes are not informative todistinguish between an abnormal methylation pattern and a bacterial recombi-nation artifact.


sequences. Several protocols to reduce the incidence of mosaicmolecules have been described (20,21). Such protocols rely onmaintaining a primer : template ratio that favors synthesis ofnew molecules over hybridization of ‘old’ strands, or post-synthesis enzymatic cleavage of heteroduplexes. However, it isdifficult to eliminate completely the cloning of heteroduplexmolecules and the utility of such data depends on thesignificance of these molecules to the interpretation of results.A mosaic pattern may be misinterpreted as abnormalhypermethylation or hypomethylation of some portion of aparental chromosome (Fig. 7). In our case, we could nottolerate the generation of heteroduplex artifacts because theyoccurred at levels greater than or equal to the fraction ofendogenous molecules targeted for assay (those paternal IGF/H19 alleles and maternal IGF2R alleles that becomemethyalted at a particular site, see Figs 2 and 4).

The main limitation of methyl-sensitive restriction endo-nuclease assays is the ability to analyze only one or a few CpGdinucleotides in a region. On the other hand, this method islikely to provide a better quantitative estimate of the ratio ofmethylation between the two parental alleles because a muchlarger number of chromosomes are analyzed. In a sample of100 ng DNA, the methylation status of the selected CpG siteis assayed on over 15 000 homologous pairs of a particularchromosome, compared with the larger number of CpG sites(but generally less than 50) that are analyzed on the five to 10homologous chromosome pairs usually assessed by bisulfite-sequencing. The incorporation of the ‘hot-stop’ PCR techniquecircumvents the difficulty of analysis of heteroduplex mole-cules because only molecules synthesized during the lastamplification cycle are radio-labeled (25).

LOI, parent of origin dependent methylation and cancer

The presence of LOI at IGF2 in both tumor cells and normaltissues raises the possibility that constitutional LOI representsa risk factor for cancer (7). Although use of the term ‘LOI’implies a qualitative difference between individuals in whomthere is parental-origin-dependent transcription of only oneallele and individuals in whom there is transcription ofboth alleles, such differences are generally quantitative (7).Individuals characterized as having LOI exhibit a relaxationof allele-specific methylation with partial methylation onthe normally unmethylated allele. This methylation ‘defect’at IGF2/H19 is similar to what we found in the majority of

individuals at the IGF2R DMR and one might argue that thiscircumstance may explain the normal biallelic expression ofthis gene. However, we also identified a number of individualswith nearly exclusive methylation of the maternal IGF2Rallele. In these cases, one might hypothesize maternal allele-specific transcription, as described by Xu et al. (9). In thisregard, the imprinting status of IGF2R is a potentially interes-ting variable in cancer susceptibility. A number of reportssupport its activity as a tumor suppressor gene in some types oftissues, including liver (26), breast (27) and kidney (28) andIGF2R is expressed from only the maternal allele in 50% ofWilms’ tumors (28), consistent with the hypothesis that thetranscriptional imprinting of this gene is associated withtumorigenesis.

Assays of allele-specific methylation exploit the fact thatdifferential DNA methylation of maternal and paternal alleles isone of the hallmarks of genome imprinting (reviewed in 29).Many imprinted loci show allele-specific DNA methylation,and in the absence of the maintenance methyltransferase, or inthe presence of methyltransferase inhibitors, the imprinting ofseveral genes is disrupted (30,31). A major advantage of allele-specific methylation assays is that the methylation phenotype ofan individual can be scored in a DNA sample. A similarsurrogate measure of the transcriptional status of the androgenreceptor gene has been used extensively as an indicator ofX-chromosome inactivation in the cells of human females(32,33). A disadvantage of surrogate measures of imprintingis the perception that methylation differences, in and ofthemselves, do not reflect ‘real’ imprinting unless they are alsoaccompanied by parental allele-specific transcription (29).However, such differences must be at the root of the imprintingprocess, regardless of whether or not the process results intranscriptional differences between maternal and paternalalleles (34). The identification of inter-individual variabilityin the establishment or maintenance of epigenetic marks, anddetermining the causes of this variability, is an important stepin determining the role of imprinting in evolution, developmentand cancer.

MATERIALS AND METHODS

Subjects

DNA samples obtained from unfractionated nucleated peri-pheral blood cells (not lymphoblast cell lines) from the Salt

Table 1. Summary of abnormal allelic methylation ratios at IGF2/H19 DMR and IGF2R DMR

Year Informative cases Percentage offamilial cases

Percentage offamilial cases

Significantacquired changes

IGF2/H19 DMR M/P ratios �0.1 M/P ratios �0.051982–1985 318/680 (46.76%) 21/318 (6.6%) 9/21 (42.8%) 49/318 (15.4%) 34/49 (69.4%) N.A.2001 133/288 (46.18%) 3/133 (2.25%) — 11/133 (8.27%) 8/11 (72.7%) None

IGF2R DMR P/M ratios � 0.1 P/M ratios � 0.8

1982–1985 165/680 (24.26%) 19/165 (11.5%) 14/19 (73.7%) 5/165 (3.03%) 2/5 (40%) N.A.2001 75/288 (26.04%) 12/75 (14.67%) 10/12 (83.3%) 3/75 (4%) — 8 individuals


Lake City collection of CEPH/Utah pedigrees between 1982and 1984 (680 individuals from 48 three-generation families)were studied. For 25 of these families (288 individuals) asecond set of DNA samples was obtained during 2001. Allsubjects gave informed consent under University of Utah I.R.B.approved protocol number 6090-96.

Methylation analysis using methyl-sensitive restrictionendonucleases MluI, MaeII, NotI and XhoI

The upstream H19 sequence used in this study is available fromGenBank (accession number AF125183), as is the sequenceof IGF2R (GenBank accession number AF069333). In orderto analyze allele-specific methylation we screened the DNAsamples for two SNPs: a C/T polymorphism (Fig. 1A)recognized by CfoI at the IGF2/H19-DMR (close to the sixthbinding site for CTCF) (16) and a C/T polymorphismrecognized by MspI at the IGF2R-DMR (Fig. 1B). Parentalorigin of alleles of heterozygous individuals was determined bypedigree analysis.

A 100 ng sample of genomic DNA from informativeindividuals was digested overnight at 37�C with an excess ofa methyl-sensitive restriction endonuclease: MluI or MaeII forthe IGF2/H19-DMR and XhoI or NotI for the IGF2R-DMRrespectively. In each experiment we used as control individualswho were homozygous for C alleles, homozygous for T alleles,as well as a mix of 50 ng genomic DNA from an individualhomozygous for the C allele and 50 ng genomic DNA from anindividual homozygous for the T allele. Digested DNA wasthen ethanol precipitated and amplified in a ‘hot-stop’ PCRassay using the following primers: F3, 50-GAGATGGGAGGAGATACTAGG-30 and R3, 50-GTCAGTTCAGTAAAAGGCTGG-30 for the IGF2/H19-DMR, and F4, 50-GACAGACCAGGATTAGCGTG-30 and R4, 50-GGACATGCACTTCTCCCGTG-30

for IGF2R-DMR. Briefly, ‘hot-stop’ PCR involves the addition,after 30 cycles, of a primer that has been radiolabeled at the 50

position and allowing synthesis to take place for a final cycle(25). The labeled primer for the last cycle was one of the sameprimers used in the first 30 cycles. However, competitionbetween labeled and unlabeled primer did not appear to be asignificant problem because the concentration of the unlabeledprimer was substantially reduced during the first 30 cycles andthe labeled primer was added at the original concentration ofthe unlabelled primer (0.4 mM).

PCR products were digested with the enzyme used foridentification of the parental origin of the alleles, the fragmentswere separated on 5% polyacrylamide gels and the intensity ofthe bands (alleles) was quantified using a Fuji BAS 2000phosphorimager (33). In the case of controls, no uncleaved Calleles were detected in any C/C homozygous individuals(indicating that the CfoI and MspI enzymes cleaved the PCRproducts with >99% efficiency) and the mixture of Chomozygote/T homozygote resulted in a 1:1 ratio of C/Tintensity (data not shown).

All the assays were performed three times, and the averagevalues were used. At the IGF2R-DMR, all samples wereassayed using both XhoI and NotI, while at the IGF2/H19-DMR, MaeII was used only for those samples with abnormalpatterns of methylation after MluI digestion and for a limitednumber of samples with a normal pattern of methylation.

Quantification of the methylation levelusing real-time PCR

Samples of 100 ng of genomic DNA were digested under thesame conditions as mentioned above using MluI, XhoI or NotIrestriction endonucleases. As controls, samples of 100 nggenomic DNA were digested using DraI which cleaves outsidethe region of interest. For both the IGF2/H19-DMR and IGF2RDMR we selected representative samples with normal andabnormal methylation pattern in the previous method. Afterethanol precipitation, PCR reactions were set up using 10 ng ofDNA as template in the LightCycler DNA Master SYBR GreenI kit (Roche), according to the manufacturer’s instructions. Theefficiency of the methyl-sensitive restriction enzymes wasquantified as the ratio of fluorescence between the PCRproducts from DNA samples digested with MluI, XhoI or NotIenzymes and PCR products from the same samples digestedwith DraI restriction enzyme. The results were similar usingthe two methods and suggest that abnormal ratios are the resultof increased methylation of the normally unmethylated alleles(data not shown).

Bisulfite-sequencing analysis at IGF2/H19 DMR

A 2.5 mg aliquot of genomic DNA was denatured at 37�C for20 min using freshly prepared 3 M sodium hydroxide (finalconcentration of 0.3 M). Denatured DNA was subsequentlyincubated at 55�C with monosulfite solution (2 M sodiummetabisulfite and 1 mM hydroquinone) in the dark, overnight(19). After purification with the QIAquick PCR purification kit(Qiagen), the converted DNA was used as a template for anested PCR. The first set of primers was F1, 50-GTGTTTGATTTA-TTTTAGGGTGTATTG-30 and R1, 50-CCAACAACACAAAAATCCTAAACCA-AAA-30 and the second was F2,50-GGAATAATGAGGTGTTTTAGTTTTA-30 and R2, 50-CTAACCACTTAAAACTAAAAAAATC-30. The PCR productswere separated in 1.5% agarose gels, purified using the QIAEXII gel extraction kit (Qiagen) and subsequently subcloned intoa TA Cloning vector (Invitrogen) according to the manufac-turer’s instructions. The DNA from 15–20 individual cloneswas sequenced for each individual.

Haplotype analysis

In order to analyze the concordance between a shared haplotypeand familial aggregation of individuals with abnormal methyla-tion ratios, we determined the genotypes of these individualsfor several SNPs adjacent to the IGF/H19 and IGF2R loci.For chromosome 11 we analyzed seven polymorphisms, inthe following physical order: D11S2071 (0.29 Mb from thetelomere), the HRAS1 minisatellite, the A/C polymorphism atexon 5 of the H19 gene (recognized by RsaI), the A/Gpolymorphism at exon 5 of H19 gene (recognized by AluI),the C/T polymorphism at IGF2/H19-DMR (recognized byCfoI), the A/G polymorphism at exon 9 of IGF2 gene(recognized by ApaI enzyme), and the VNTR minisatellite atthe 50 flanking region of insulin gene (2.27 Mb from thetelomere). For chromosome 6 we analyzed the C/T polymor-phism in intron 2 of IGF2R (recognized by MspI) and the C/Tpolymorphism at intron 35 of IGF2R gene (recognized by


MspI). The genotypes of some individuals at the insulin VNTRand HRAS1 minisatellite loci were obtained from the CEPHdatabase. Haplotypes were established by pedigree analysis.

ACKNOWLEDGEMENTS

We are grateful to Andrew Feinberg and Hengmi Cui fordiscussion and communication of results prior to publication,to Laurie Mecham, Tina Varvil and Joe Vaughan foradministrative and technical assistance. We would like toextend our sincere thanks to all family members whoparticipated in the Utah Genetic Reference Project. Thanksalso to Andreas P. Peiffer, MD, PhD, UGRP Medical Director,and Melissa M. Dixon, UGRP Study Coordinator. Thisresearch was supported by a grant from the NationalInstitutes of Health (NIH R21ES/CA11607) (to C.S.) and bya Public Health Services research grant to the HuntsmanGeneral Clinical Research Center at the University of Utah,grant number M01-RR00064 from the National Center forResearch Resources. It was also supported by generous giftsfrom the W.M. Keck Foundation and from the George S. andDelores Dore Eccles Foundation.

REFERENCES

1. Bird, A. (2002) DNA methylation patterns and epigenetic memory.Genes Dev., 16, 6–21.

2. Li, E. (2002) Chromatin modification and epigenetic reprogramming inmammalian development. Nat. Rev. Genet., 3, 662–673.

3. Jenuwein, T. and Allis, C.D. (2001) Translating the histone code.Science, 293, 1074–1780.

4. Rainier, S., Johnson, L.A., Dobry, C.J., Ping, A.P., Grundy, P.E. andFeinberg, A.P. (1993) Relaxation of imprinted genes in human cancer.Nature, 362, 747–749.

5. Ogawa, O., Eccles, M.R., Szeto, J., McNoe, L.A., Yun, K., Maw, M.A.,Smith, P.J. and Reeve, A.E. (1993) Relaxation of insulin-like growth factorII gene imprinting implicated in Wilms’ tumor. Nature, 362, 749–751.

6. Falls, J.G., Pulford, D.J., Wylie, A.A. and Jirtle, R.L. (1999) Genomicimprinting: implications for human disease. Am. J. Pathol., 154, 635–647.

7. Cui, H., Horon, I.L., Ohlsson, R., Hamilton, S.R. and Feinberg, A.P. (1998)Loss of imprinting in normal tissue of colorectal cancer patients withmicrosatellite instability. Nat. Med., 4, 1276–1280.

8. Kalscheuer, V.M., Mariman, E.C., Schepens, M.T., Rehder, H. andRopers, H.H. (1993) The insulin-like growth factor type-2 receptor gene isimprinted in the mouse but not in humans. Nat. Genet., 5, 74–78.

9. Xu, Y., Goodyer, C.G., Deal, C. and Polychronakos, C. (1993) Functionalpolymorphism in the parental imprinting of the human IGF2R gene.Biochem. Biophys. Res. Commun., 197, 747–754.

10. Nishiwaki, K., Niikawa, N. and Ishikawa, M. (1997) Polymorphic andtissue-specific imprinting of the human Wilms tumor gene, WT1. Jpn.J. Hum. Genet., 42, 205–211.

11. Sakatani, T., Wei, M., Katoh, M., Okita, C., Wada, D., Mitsuya, K.,Meguro, M., Ikeguchi, M., Ito, H., Tycko, B. and Oshimura, M. (2001)Epigenetic heterogeneity at imprinted loci in normal populations. Biochem.Biophys. Res. Commun., 283, 1124–1130.

12. Nakagawa, H., Chadwick, R.B., Peltomaki, P., Plass, C., Nakamura, Y.and de La Chapelle, A. (2001) Loss of imprinting of the insulin-likegrowth factor II gene occurs by biallelic methylation in a core region ofH19-associated CTCF-binding sites in colorectal cancer. Proc. Natl Acad.Sci. USA, 98, 591–596.

13. Vu, T.H., Li, T., Nguyen, D., Nguyen, B.T., Yao, X.M., Hu, J.F. andHoffman, A.R. (2000) Symmetric and asymmetric DNA methylation inthe human IGF2-H19 imprinted region. Genomics, 64, 132–143.

14. Srivastava, M., Hsieh, S., Grinberg, A., Williams-Simons, L., Huang, S.P.and Pfeifer, K. (2000) H19 and Igf2 monoallelic expression is regulatedin two distinct ways by a shared cis acting regulatory region upstreamH19. Genes Dev., 14, 1186–1195.

15. Hark, A.T., Schroenherr, C.J., Katz, D.J., Ingram, R.S., Levorse, J.M. andTilghman, S.M. (2000) CTCF mediates methylation-sensitive enhancer-blocking activity at the H19/Igf2 locus. Nature, 405, 486–489.

16. Takai, D., Gonzales, F.A., Tsai, Y.C., Thayer, M.J. and Jones, P.A. (2001)Large scale mapping of methylcytosines in CTCF-binding sites in thehuman H19 promoter and aberrant hypomethylation in human bladdercancer. Hum. Mol. Genet., 10, 2619–2626.

17. Smrzka, O.W., Fae, I., Stoger, R., Kurzbauer, R., Fisher, G.F., Henn, T.,Weith, A. and Barlow, D.P. (1995) Conservation of a maternal-specificmethylation signal at the human IGF2R locus. Hum. Mol. Genet., 4,1945–1952.

18. Frommer, M., McDonald, L.E., Millar, D.S., Collis, C.M., Watt, F.,Grigg, G.W., Molloy, P.L. and Paul, C.L. (1992) A genomic sequencingprotocol that yields a positive display of 5-methylcytosine residues inindividual DNA strands. Proc. Natl Acad. Sci. USA, 89, 1827–1831.

19. Clark, S.J., Harrison, J., Paul, C.L. and Frommer, M. (1994) Highsensitivity mapping of methylated cytosines. Nucl. Acids Res., 22,2990–2997.

20. Thompson, J.R., Marcelino, L.A. and Polz, M.F. (2002) Heteroduplexesin mixed-template amplifications: formation, consequence and eliminationby ‘reconditioning PCR’. Nucl. Acids Res., 30, 2083–2088.

21. Lowell, J.L. and Klein, D.A. (2000) Heteroduplex resolution using T7endonuclease I in microbial community analyses. Biotechniques, 28,676–678, 680, 681.

22. Grunau, C., Clark, S.J. and Rosenthal, A. (2001) Bisulfite genomicsequencing: systematic investigation of critical experimental parameters.Nucl. Acids Res., 29, e65.

23. Warnecke, P.M., Stirzaker, C., Song, J., Grunau, C., Melki, J.R. andClark, S.J. (2002) Identification and resolution of artifacts in bisulfitesequencing. Methods, 27, 101–107.

24. Grilley, M., Holmes, J., Yashar, B. and Modrich, P. (1990) Mechanismsof DNA-mismatch correction. Mutat. Res., 236, 253–267.

25. Uejima, H., Lee, M.P., Cui, H. and Feinberg, A.P. (2000) Hot-stop PCR:a simple and general assay for linear quantification of allele ratios. Nat.Genet., 25, 375–376.

26. De Souza, A.T., Hankins, G.R., Washington, M.K., Orton, T.C. andJirtle, R.L. (1995) M6P/IGF2R gene is mutated in human hepatocellularcarcinomas with loss of heterozigosity. Nat. Genet., 11, 447–449.

27. Hankins, G.R., De Souza, A.T., Bentley, R.C., Patel, M.R., Marks, J.R.,Iglehart, J.D. and Jirtle, R.L. (1996) M6P/IGF2R receptor: a candidatebreast tumor suppressor gene. Oncogene, 12, 2003–2009.

28. Xu, Y.Q., Grundy, P. and Polychronakos, C. (1997) Aberrant imprintingof the insulin-like growth factor II receptor gene in Wilms’ tumor.Oncogene, 14, 1041–1046.

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

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

31. Dao, D., Walsh, C.P., Yuan, L., Gorelov, D., Feng, L., Hensle, T., Nisen, P.,Yamashiro, D.J., Bestor, T.H. and Tycko, B. (1999) Multipoint analysisof human chromosome 11p15/mouse distal chromosome 7: Inclusion ofH19/IGF2 in the minimal WT2 region, gene specificity of H19 silencingin Wilms’ tumorigenesis and methylation hyper-dependence of H19imprinting. Hum. Mol. Genet., 8, 1337–1352.

32. Allen, R.C., Zoghbi, H.Y., Moseley, A.B., Rosenblatt, H.M. andBelmont, J.W. (1992) Methylation of HpaII and HhaI sites near thepolymorphic CAG repeat in the human androgen-receptor genecorrelates with X chromosome inactivation. Am. J. Hum. Genet., 51,1229–1239.

33. Naumova, A.K., Bird, L.M., Slamka, C., Fonseca, M., Verner, A.E.,Wang, M., Leppert, M., Morgan, K. and Sapienza, C. (1995) Sex-specifictransmission-ratio distorsion of X-chromosome alleles in the offspring offemales with nonrandom X-inactivation, Dev. Genet., 17, 198–205.

34. Pardo-Manuel de Villena, F., de la Casa Esperon, E. and Sapienza, C.(2000) Natural selection and the function of genome imprinting: beyondthe silenced minority. Trends Genet., 16, 573–579.


Familial aggregation of abnormal methylation of parental alleles at the IGF2/H19 and IGF2R...

Documents

Transcript of Familial aggregation of abnormal methylation of parental alleles at the IGF2/H19 and IGF2R...