Heritable rather than age-related environmental
and stochastic factors dominate variation in
DNA methylation of the human IGF2/H19 locus
Bastiaan T. Heijmans1,*, Dennis Kremer1, Elmar W. Tobi1, Dorret I. Boomsma2
and P. Eline Slagboom1
1Molecular Epidemiology Section, Leiden University Medical Centre, Leiden 2333 ZC, The Netherlands and
2Biological Psychology, Vrije Universiteit, Amsterdam 1081 BT, The Netherlands
Received December 7, 2006; Revised and Accepted January 26, 2007
Epigenetic variation may significantly contribute to the risk of common disease. Currently, little is known
about the extent and causes of epigenetic variation. Here, we investigated the contribution of heritable influ-
ences and the combined effect of environmental and stochastic factors to variation in DNA methylation of the
IGF2/H19 locus. Moreover, we tested whether this locus was subject to age-related degeneration of epige-
netic patterns as was previously suggested for global methylation. We measured methylation of the H19
and IGF2 differentially methylated regions (DMRs) in 196 adolescent and 176 middle-aged twins using a
recently developed mass spectrometry-based method. We observed substantial variation in DNA methylation
across individuals, underscoring that DNA methylation is a quantitative trait. Analysis of data in monozygotic
and dizygotic twins revealed that a significant part of this variation could be attributed to heritable factors.
The heritability of methylation of individual CpG sites varied between 20 and 74% for the H19 DMR and
was even higher, between 57 and 97%, for the IGF2 DMR. Remarkably, the combined influence of environ-
mental and stochastic factors on DNA methylation was not greater in middle-age than in adolescence,
suggesting a limited role for age-related degeneration of methylation patterns at this locus. Single nucleotide
polymorphisms in the IGF2/H19 locus were significantly associated with DNA methylation of the IGF2 DMR
(P 5 0.004). A preliminary analysis suggested an association between H19 DMR methylation and body size
(P < 0.05). Our study shows that variation in DNA methylation of the IGF2/H19 locus is mainly determined
by heritable factors and single nucleotide polymorphisms (SNPs) in cis, rather than the cumulative effect
of environmental and stochastic factors occurring with age.
Epigenetic mechanisms are intimately involved in phenomena
that rely on a stable control of gene expression such as cell
differentiation, X chromosome inactivation and genetic
imprinting (1). In mammals, the molecular basis of epigenetic
mechanisms includes a range of histone modifications, inclu-
ding acetylation, and the methylation of DNA, particularly
methylation of cytosines of CpG dinucleotide sequences.
A genomic region with deacetylated histones and methylated
DNA is not accessible for the transcriptional machinery and
gene expression is silenced (2). Although generally stable,
the epigenetic state of a genomic region may undergo changes
that result from transient environmental circumstances or
from stochastic events related to inaccuracies of the machinery
maintaining epigenetic patterns (3). Such changes may even-
tually have significant phenotypic consequences as the epige-
netic state is heritable during mitosis and thus an increasing
number of cells may display altered gene expression. Recent
proof of this concept came from experiments with Agouti
Avymice (4). Mice exposed to prenatal methyl-donor deficiency
showed a reduced methylation of the agouti gene across
tissues which was associated with increased expression of
the gene product, which results in a yellow coat color. This
epigenetically induced phenotype was stable over the lifetime.
# The Author 2007. Published by Oxford University Press. All rights reserved.
For Permissions, please email: email@example.com
*To whom correspondence should be addressed at: Molecular Epidemiology Section, Leiden University Medical Centre, Postal Zone S-05-P, PO Box
9600, 2300 RC Leiden, The Netherlands. Tel: þ31 715269785; Fax: þ31 715268280; Email: firstname.lastname@example.org
Human Molecular Genetics, 2007, Vol. 16, No. 5
Advance Access published on March 5, 2007
Human studies commonly investigate the role of epigenetic
mechanisms in cancer where an epigenetic event in a single cell
contributes to tumor formation (2). Studies on more widespread
epigenetic changes that may affect tissue function and contrib-
ute to other diseases are, however, scarce and generally focus
on the influence of diet and age. Intervention studies indicated
an effect of diet on average genomic (i.e. global) DNA methyla-
tion. Several weeks of moderate depletion of the methyl-donor
folate resulted in significant global hypomethylation of genomic
DNA from lymphocytes in postmenopausal women (5,6).
Global hypomethylation could afterwards be corrected by a
diet containing sufficient folate (5). More recently, monozygotic
(MZ) twins were comprehensively investigated for epigenetic
differences. With increasing age, MZ twins were found to
display significantly more epigenetic differences particularly
at the level of global DNA methylation, methylation of repeti-
tive sequences and global histone acetylation (7). The phenoty-
pic consequences of such global epigenetic changes, however,
are as yet unclear.
These data may indicate the widespread occurrence of diet
and age-dependent degeneration of epigenetic patterns, which,
in turn, might contribute to the phenotypic changes that arise
during aging. An alternative interpretation, however, is that the
epigenetic changes uncovered using global assessments may
be enriched for well-tolerated alterations with limited pheno-
typic consequences. One approach to differentiate between
these interpretations is the investigation of specific loci that are
under epigenetic control and produce phenotypic consequences
when this control is lost. The maternally imprinted insulin-like
growth factor II (IGF2) gene on chromosome 11p15.5 is
one of the best-characterized epigenetically regulated loci. It
involves both the IGF2 region itself, of which the maternal
allele is methylated, and the H19 region located ?150 kb
p-ter, of which the paternal allele is methylated. Moreover, the
loss of epigenetic control, specifically demethylation of the
IGF2 differentially methylated region (DMR), leads to IGF2
overexpression and is associated with a ?5-fold increased risk
of colorectal neoplasia (8,9). IGF2 imprinting defects also
underlie the Beckwith–Wiedeman syndrome (10). Furthermore,
IGF2 is a major fetal growth factor and is implicated in body
composition (11,12) and atherosclerosis (13).
The aim of the current study is 3-fold: first to evaluate the
inter-individual variation in DNA methylation of the H19
(14) and IGF2 (8) DMRs in the general population, second
to assess the influence of heritable versus environmental and sto-
chastic factors on this variation and, third to examine whether
the effect of these factors depends on age. To this end, DNA
methylation was assessed in monozygotic (MZ) twins, who are
identical at the DNA sequence level, and in dizygotic (DZ)
twins, who on average share half of their genome identical-
by-descent. In total, 372 twins were studied, of whom 196
were adolescent (mean age 17 years) and 176 middle-aged
(mean age 45 years). DNA methylation was measured using
a recently developed method by employing mass spectrometry
that allows quantifying the methylation of individual CpG
sites (15). In addition to the measurement of DNA methyla-
tion, SNPs were genotyped in the IGF2/H19 region (in cis)
and the MTHFR Ala222Val variant (in trans), which is
associated with disturbed methyl-donor metabolism and
global DNA hypomethylation (16,17), to test their association
with methylation of the H19 and IGF2 DMRs. Finally, the
association of methylation of the H19 and IGF2 DMRs with
body size was considered in a preliminary analysis.
Characteristics of twins and IGF2/H19 DMRs
DNA methylation of the IGF2/H19 locus was analyzed in 196
adolescent and 176 middle-aged twins (in total, 372 indivi-
duals). The adolescent twins had a mean age of 16.7 years
(SD, 2.0) and consisted of 54 MZ and 44 DZ twin pairs.
The middle-aged twins had a mean age of 44.8 years (SD,
6.8) and consisted of 48 MZ and 40 DZ twin pairs.
DNA methylation was assessed at two locations: at the 50
region of the H19 transcript overlapping a CpG island,
which is part of the H19 DMR, and in exon 3 of the
IGF2AS transcript, which is part of the IGF2 DMR (Fig. 1).
In the H19 DMR, the methylation of 12 CpG sites was
measured: six individually and the other six in pairs because
they were directly adjacent and could not be resolved with
the technology used. In the IGF2 DMR, six CpG sites were
assessed: four individually and two as a pair. Figure 2
shows the mean methylation of the various CpG sites sepa-
rately for adolescents and middle-aged twins. The mean
methylation of CpGs in the H19 DMR varied between 23
and 30%. The mean methylation of the IGF2 DMR was
higher and varied around 50% (43–62%). For both regions,
the average methylation was similar in adolescent and
middle-aged individuals and independent of sex and age.
Inter-individual variation in DNA methylation
Considerable inter-individual variation in methylation was
observed for each CpG site (Fig. 3). Methylation for the
majority of sites was normally distributed, others were bimod-
ally distributed, which could be explained by the presence of
an SNP at the CpG site itself. For example, CpG site
located at 382 bp in the H19 DMR was non-methylated in a
subset of individuals (Fig. 3).
Correlation in DNA methylation
DNA methylation of different CpG sites was correlated
(Fig. 4). Particularly within the H19 region, methylation was
highly correlated (up to 0.92), although the two q-ter CpG
sites showed lower correlations with the sites p-ter (0.42–
0.73). Within the IGF2 DMR, the bimodally distributed CpG
site at 296 bp (Fig. 3) was uncorrelated with other sites. Of
interest, there was a lower correlation up to 0.25 (P ¼ 1025)
between sites in the H19 and the IGF2 DMRs, which are
located ?134 kb apart.
Principal components analysis (PCA) was used to reduce
the dimensionality in the data by obtaining component
scores for methylation. PCA extracts components from a cor-
related set of phenotypes and is not affected by factors merely
influencing a single CpG site, such as an SNP abolishing a
CpG site. Summarizing the correlated data into a smaller
number of components also reduces the testing burden.
Methylation of the H19 DMR could be described by a single
548Human Molecular Genetics, 2007, Vol. 16, No. 5
main principal component describing 73% of the total varia-
tion brought about by the 12 CpG sites measured in this
DMR. For the IGF2 DMR (six CpG sites), 53% of the varia-
tion was captured with a single principal component.
Heritability of DNA methylation
The component scores obtained using PCA were used to esti-
mate the genetic and environmental influences on DNA
methylation in a twin design. The heritability of H19 methyla-
tion was significant and virtually identical in the adolescent
(0.33; 95% CI, 0.11–0.52) and middle-aged twins (0.35;
95% CI, 0.04–0.59), thus suggesting that the genetic control
on DNA methylation at this locus was not compromised in
the age range tested. There was no evidence for the influence
of common environmental (e.g. familial) factors (P ¼ 0.87).
The heritability of the component score describing the IGF2
DMR methylation was higher (and thus the environmental
influence smaller): 0.80 (95% CI, 0.68–0.87) in adolescent
and 0.75 (95% CI, 0.59–0.84) in middle-aged twins. Again,
the heritability was similar in both groups (P ¼ 0.82), indica-
ting that the influence of environmental factors did not
increase with age. As for the H19 DMR, an influence of
common environment was absent (P ¼ 0.91).
Figure 5 shows the heritability of all CpG sites individually for
both adolescent and middle-aged twins separately as well as
combined. As for the component scores, the heritabilities were
substantially higher for the IGF2 DMR than for the H19 DMR
and the heritabilities in adolescent and middle-aged twins
were not statistically different with one exception (at 382 bp).
Particularly, high heritabilities were observed for the three non-
normally distributed CpG sites (H19 at 382 kb IGF2 at 202
and 296 bp). This high heritability is compatible with the effect
of an SNP allele abolishing the CpG site and thus DNA methyl-
ation. When the analysis was repeated for these sites after exclud-
ing individuals with a low methylation, the heritability decreased
toward that of other CpG sites in the two DMRs.
SNPs and DNA methylation
Subsequently, we investigated whether the heritability could
be attributed to genetic variation at the IGF2/H19 locus itself
and the Ala222Val variant of the MTHFR gene (rs1801133),
which is known to influence global DNA methylation
(16,17). We genotyped three SNPs in the H19 region and
five SNPs in the IGF2/IGF2AS region (Fig. 1). The three
H19 SNPs together accounted for 1.4% of the total variance
in the component score describing methylation of the H19
region, which was not significant (P ¼ 0.40, df ¼ 3). Neverthe-
less, SNP rs2839701 was associated with four individual H19
CpG sites (Fig. 6). The association of rs217727 with the non-
normally distributed CpG site H19 at 382 bp (P ¼ 0.0002)
was lost when individuals not methylated at this site were re-
moved (P ¼ 0.37). This is presumably explained by linkage
disequilibrium (LD) between the SNP genotyped and an
unknown, rare SNP abolishing the CpG site at 382 bp.
The associations of IGF2/IGF2AS SNPs with methylation
of the IGF2 DMR were more striking. The five SNPs com-
bined explained 6.5% of the variance in the component
score (P ¼ 0.004, df ¼ 5). A similar percentage of the var-
iance of CpGs at 41 bp and 57,60 bp could be attributed to
the SNPs (Table 1). Individual SNPs showing the strongest
association with methylation of these two sites were rs680
(P ¼ 5 ? 1024
(P ¼ 5 ? 1024and 4 ? 1024, respectively; Fig. 6). The two
SNPs were in modest LD (D0¼ 0.92, r2¼ 0.32) and statistical
tests could not exclude an independent association of the
As expected, the strongest association was observed for
CpG site at 296 whose bimodal distribution was explained
by rs1004446 (P ¼ 7 ? 10213), which is in perfect LD
(r2¼ 1) with rs3741209 that abolishes the CpG site.
In contrast to SNPs in cis, the MTHFR SNP was not
associated with either H19 or IGF2 DMR methylation (Fig. 4).
and1023, respectively) andrs1003483
Figure 2. Mean DNA methylation for individual CpG sites in the H19 and
IGF2 DMRs as observed in 196 adolescent (mean age 16.7) and 176
middle-aged twins (mean age 44.8).
Figure 1. Overview of the H19 and IGF2/IGF2AS genomic regions with, in black, the H19 and IGF2 DMRs assayed for methylation and the SNPs measured.
Human Molecular Genetics, 2007, Vol. 16, No. 5549
Methylation and body size
Measures of body size were available for the middle-aged twins.
Higher methylation of the H19 DMR as estimated using the
component score was associated with a smaller waist circumfer-
ence (P ¼ 0.046) and a lower waist–hip ratio (WHR)
(P ¼ 0.023), but not with body mass index (BMI). Methylation
of the IGF2 DMR was not associated with body size.
We observed substantial variation in DNA methylation of the
IGF2/H19 locus in both adolescent and middle-aged twins
underscoring that DNA methylation is a quantitative trait (18).
The variation could be attributed to heritable factors, parti-
cularly in case of the IGF2 DMR for which the heritability
was 75–80%. With 35%, the heritable influences on variation
in H19 methylation were smaller. One of the most striking
Figure 3. Distribution methylation of four representative CpG sites. SNPs
underlie the non-normal distributions observed. IGF2 DMR at 296 bp:
rs3741209 at 297 bp; IGF2 DMR at 202 bp: rs4929963 at 202 bp (data not
shown); H19 at 382 bp unknown SNP.
Figure 5. Heritability estimates for CpG sites and component scores based on
PCA separately for adolescent and middle-aged twins as well as combined.
For one CpG site (H19 at 382 bp), the heritabilities between the two groups
were significantly different (P ¼ 0.01) and for another CpG site (IGF2
DMR at 202 bp), there was evidence for a significant contribution of
common environment (P ¼ 0.02). PC denotes the methylation score descri-
bing the methylation per DMR obtained using PCA.
Figure 4. Graphical representation of the strength of the correlation between
550 Human Molecular Genetics, 2007, Vol. 16, No. 5
findings of the current study was that the combined influence of
environmental and stochastic factors did not increase from ado-
lescence to middle age at the expense of the heritability. This
suggests that age-related degeneration of methylation patterns
plays a minor role at this locus. A moderate but highly signifi-
cant part of the heritability of DNA methylation was accounted
for by SNPs in the IGF2/H19 locus itself. In contrast, the
MTHFR Ala222Val variant, which is known to be associated
with global DNA hypomethylation (16,17), was not associated.
It has previously been speculated that variation in DNA
methylation and other epigenetic characteristics may be partly
heritable (3,19). This is the first study to show a substantial
heritability for epigenetic variation at the IGF2/H19 locus
using a large series of twins. Previous data already hinted at
the involvement of genetic factors in IGF2/H19 methylation.
Indications for familial aggregation were observed for H19
methylation assayed using methylation-sensitive restriction
endonucleases (20) and loss-of-imprinting of the IGF2 DMR
was not associated with a series of specific environmental
factors including diet (9). We observed the highest heritability
for the IGF2 DMR whose methylation status was previously
shown to be directly correlated with IGF2 expression, parti-
cularly CpG sites at 41 and 57,60 bp (8).
In the current study, we reported on the methylation of an
imprinted locus. Future studies are needed to establish
whether our findings extend to histone modifications and non-
imprinted loci. In addition, our study was limited to individ-
uals with a maximum age of 62 years. It will be valuable to
examine the occurrence of age-related degeneration of epige-
netic patterns in very old age.
Methylation of the IGF2/H19 locus was associated with
SNPs in cis. For three CpG sites, the association was simply
explained by a rare allele of an SNP abolishing the CpG site
and thus its potential to become methylated. The other SNPs
found to be associated, however, did not affect a CpG site in
the DMRs and are of particular interest. Their further investi-
gation may be a promising starting point for elucidating more
sequence motifs guiding DNA methylation (21,22). The
association of specific SNPs with DNA methylation suggests
that its heritability has a genetic and not an epigenetic origin.
This would be in line with the thought that DNA methylation
of imprinted loci is erased during gametogenesis after which
parent-of-origin specific patterns are established (23).
Our study indicated that the influence of environmental
factors on IGF2/H19 methylation was not greater in middle
age than in adolescence. This implies that age-related degene-
ration is of limited relevance to this locus, even though its
epigenetic state was hypothesized to be especially susceptible
to environmental dysregulation (24). Fraga et al. (7) observed
a greater discordance in global epigenetic measures among
MZ twin pairs of increasing age. This greater discordance
was observed in MZ twins from the age of 28 years and
onwards (range 28–70 years), which is comparable with the
age of the middle-aged twins we studied (range 34–62
years). Our study suggests that some loci may be relatively
resistant to age-related changes in methylation. Fraga et al.
(7) found age-related differences particularly in global DNA
methylation, global histone modifications and repetitive
sequences. Previous studies indicated that such global
changes can be readily induced as well as reversed by diet
(5,6) and their phenotypic consequences are currently unclear.
It might be hypothesized that epigenetic patterns with
greater stability during aging are more likely to be associated
with phenotypic consequences if changed. The putative
phenotypic consequences of quantitative variation in methyla-
tion of the IGF2/H19 locus as reported here have yet to be
characterized. Imprinting defects of the IGF2/H19 locus,
however, are associated with severe epigenetic disease (25)
and an increased risk of colorectal adenoma (8,9).
A limitation of both our and the Fraga et al. study is that they
relied on genomic DNA extracted from whole blood so that
heterogeneity in cell populations may have contributed to
the outcomes (26). Our study likely is less sensitive to such
Figure 6. Association of SNPs in the H19 and IGF2/IGF2AS genomic regions
and the MTHFR gene with DNA methylation of H19 and IGF2 DMRs.
Table 1. Total variance of DNA methylation of the H19 and IGF2 DMRs
explained by SNPs in cis
H19 DMR methylation and three SNPs in cis
At 26 bp
At 177 bp
At 195,197 bp
At 216 bp
At 253,255 bp
At 262 bp
At 306,308 bp
At 341 bp
At 382 bp
IGF2 DMR methylation and five SNPs in cis
At 41 bp
At 57,60 bp
At 202 bp
At 251 bp
At 296 bp
Three SNPs were genotyped in the H19 DMR (test with df ¼ 3) and five
in the IGF2/IGF2AS region (df ¼ 5; see Fig. 1).
Human Molecular Genetics, 2007, Vol. 16, No. 5 551
heterogeneity because the epigenetic state of imprinted loci is
less dependent on cell differentiation and, importantly, a previous
study showed that when demethylation of the IGF2 DMR was
observed in peripheral blood lymphocytes of an individual,
this was also found in colon tissue (8), which has a distinct
embryologic origin (endoderm and mesoderm, respectively).
In pigs, IGF2 promotes muscle growth at the expense of fat
deposition (12). In a preliminary analysis, we observed an
association of higher H19 DMR methylation with a smaller
waist circumference and a lower WHR in middle-aged twins.
Although the notion that common epigenetic variation might
influence a phenotype in the general population is intriguing,
the result should be interpreted with care. The sample size
was relatively small and an explanation for the finding is specu-
lative. It is currently unknown whether modestly increased H19
DMR methylation upregulates IGF2 expression. Also, it has not
been established whether such quantitative differences mea-
sured in lymphocytes mark a soma-wide phenomenon as was
suggested for loss-of-imprinting of the IGF2 DMR (8).
We show the presence of substantial epigenetic variation at
the IGF2/H19 locus in the general population. Once estab-
lished, DNA methylation of this locus is mainly genetically
determined and appears to be relatively stable during aging.
Our data highlight the potential of studying genetic and epige-
netic variation in concert as a means to identify novel genomic
determinants of common disease.
MATERIALS AND METHODS
The adolescent and adult twins studied are part of The Nether-
lands Twin Register. Detailed characteristics of these samples
have been described previously (27,28). The adolescent twins
(age range 13–22 years) were recruited between 1985 and
1988 and all lived in the same household with their parents;
the middle-aged twins (age range 34–62 years) were recruited
between 1992 and 1995 and generally lived apart with their
own families. Zygosity of the twins was initially determined by
standardized questionnaires, typing of blood group markers and
was confirmed with short tandem repeat markers measured as
part of a genome-wide linkage scan (29). In the middle-aged
twins, measures of body size were assessed during a visit to
the research centre. Measurements included weight and height
for calculation of BMI and waist and hip circumference for
calculation of WHR. Informed consent was obtained from all
participants and the study was approved by the Medical Ethics
Committee of the Vrije Universiteit Amsterdam.
DNA methylation measurement
DNA methylation measurements were performed on genomic
DNA extracted from whole blood. Genomic DNA was treated
with sodium bisulfite using the EZ methylation kit (Zymo-
Research). Sodium bisulfite converts unmethylated cytosines
to uracils, whereas it does not affect methylated cytosines.
One microgram of genomic DNA was bisulfite-treated.
Methylation of individual CpG sites was assessed using the
MALDI-TOF mass spectrometry-based method (Epityper,
Sequenom) (15). Briefly, 10 ng of bisulfite-treated DNA is
PCR-amplified with primers containing a T7-promoter. After
PCR, the reverse strand of the product is transcribed using
T7-polymerase and the resulting RNA is cleaved with
RNase A. RNase A cleaves 30of every uracil thus producing
different sized fragments, part of which will contain one or
more original CpG sequences. Note that the reverse strand
of the bisulfite-treated DNA is transcribed so that original
CpG sequences are either present as CpA if unmethylated or
CpG if methylated. As fragments originating from a methyl-
ated CpG sequence contain a G instead of an A-base, they
will have a 16 Da higher molecular weight. This mass differ-
ence can be readily detected using a MALDI-TOF mass spec-
trometer (Autoflex, Bruker Daltonics). The spectra produced
by the mass spectrometer were analyzed using the software
QMA 1.1 (Sequenom). This software uses the genomic
sequence of the amplicon to predict the fragments that occur
after RNase A fragmentation and matches the expected and
observed peaks taking into account that fragments are 16 Da
heavier for every methylated CpG sequence they contain.
The percentage methylation of individual CpG sites is calcu-
lated using the area under the peak of the signal from the
unmethylated and methylated fragments. As DNA was
extracted from whole blood, the percentage methylation
reflects the average methylation in a population of cells.
The quantitative nature of the assay was shown previously
(15). To achieve optimal quantification within the limitations
of the technology, all samples were measured in triplicate. If
the triplicate measurements had a standard deviation equal
to or greater than 0.10, all data for the sample involved were
discarded (removing 3.5% of measurements). In addition,
DNA samples for which methylation could be established
for ,60% of the CpG sites were removed prior to analyses
(removing 4.0% of measurements). Finally, CpG sites with a
success rate lower than 80% after applying the previous cri-
teria were not analyzed (discussed subsequently).
DNA methylation was assessed in two regions (Fig. 1): in the
50region of the H19 transcript overlapping a CpG island, which
is part of the H19 DMR (14), and in exon 3 of the IGF2AS tran-
script, which is part of the IGF2 DMR (8). The H19 amplicon
was amplified using the primers (tags in lower case) aggaaga-
CTACCAAC (reverse) and encompassed 413 bp (NCBI build
34, chr11: 1 975 948–1 976 360). All primers were designed
on the reverse complement strand and anneal to sequences
devoid of CpG dinucleotides. CpG sites are numbered according
onwards. The H19 DMR amplicon harbored 25 CpG sites, 12 of
which were successfully measured (three fragments contained
two adjacent CpG sites). Six CpG sites on four fragments
could not be measured independently because the fragments
trum; six CpG sites could not be assessed because of a success
rate ,80%; and one site could not be measured because of
an adjacentSNP (rs12292822)compromising the correct predic-
tion of fragments by the software. The IGF2 DMR amplicon
was amplified using the primers aggaagagag-TGGATAGGA-
GATTGAGGAGAAA (forward) and cagtaatacgactcactata-
gggagaaggct-AAACCCCAACAAAAACCACT (reverse) and
encompassed 338 bp (NCBI build 34, chr11: 2 126 035–
552Human Molecular Genetics, 2007, Vol. 16, No. 5
2 126 372; note that base 2 126 154 is a T instead of a C in
.99% of individuals of European descent, thus abolishing a
CpG site). The amplicon harbored seven CpG sites, six of
which could be measured successfully (one fragment contained
two CpG sites). CpG sites at 41 bp and 57,60 bp coincided
with those studied by Cui et al. (8) and were highly correlated
with IGF2 expression. Measurement of the remaining CpG
sites was compromised by an adjacent SNP (rs17883338). The
meansuccess ratefor the 18successfullyassayedCpG sites(dis-
tributed over 14 fragments) was 92.5%.
Using the HapMap phase II data, SNPs were selected in the H19
and IGF2/IGF2AS regions using a pairwise tagging strategy
(r2. 0.8; Fig. 1). Two SNPs commonly genotyped in other
studies were also selected [rs217727 and rs680 (11,30)]. In
addition, the Ala222Val (¼677C/T¼ rs1801133) variant of the
MTHFR gene was genotyped. Genotyping of the nine SNPs
was done using the mass spectrometry-based hME assay (Seque-
nom) in two multiplex reactions. Average genotyping success
rate was 99.1%; for 13 of 378 samples with DNA methylation
data, there was insufficient DNA for genotyping, resulting in
an effective average success rate of 95.7%. The observed minor
allele frequencies were 0.18 (rs217727), 0.47 (rs2839701), 0.27
(rs2251375), 0.31 (rs680), 0.20 (rs3213223), 0.41 (rs3213221),
0.46 (rs1003483), 0.38 (rs1004446) and 0.36 (rs1801133).
Statistical analyses were carried out with four goals (1): to
reduce the high dimensional data set of correlated CpG
methylation data into a smaller number of independent
factors using PCA (2), to estimate the heritability of variation
in methylation of the individual sites as well as the component
scores obtained using PCA (3), to assess the association
between SNPs and methylation and (4) to test for association
between DNA methylation and body size. Prior to PCA, cor-
relations between methylation of different CpG sites were
visualized using GOLD (31). PCA was performed on corre-
lated methylation data of individual CpG sites separately for
the H19 and IGF2 DMRs using SPSS 11. PCA indicated
one factor with an eigenvalue exceeding 1 for either DMR.
In the case of the H19 DMR, this single factor explained
73% of the variance in methylation, whereas it explained
53% of the variance in IGF2 DMR methylation. To assess
the heritability, estimates of additive genetic V(A), common
environmental V(C) and unique environmental and stochastic
V(E) variance in DNA methylation were obtained using
variance components analysis implemented in Mx 1.61
(http://www.vcu.edu/mx/). The analysis makes use of the
differential genetic resemblance of monozygotic and dizygotic
twins to estimate the effects of genes and environment on
phenotypic variation. Differences in variance components
and in heritability between adolescent and middle-aged
twins were assessed by testing whether the model with the
same absolute estimates for V(A), V(C) and V(E) in adoles-
cent and middle-aged twins had a significantly worse fit to
the data than a model with separate estimates for the two
age groups. The association of SNPs with DNA methylation
was tested as additive genotypic effects within the same var-
iance components setting, thereby taking into account the
dependency in the observations from twins. Pairwise linkage
disequilibria between SNPs were estimated using HaploView
(32). Testing for association between DNA methylation and
body size was also done in the variance components setting
with body size as outcome measure, the component scores
for the two DMRs as explaining variables and age and sex
as confounders. Prior to statistical analysis, methylation data
were adjusted for bisulfite batch to exclude potential con-
We wish to thank Gabor Abbas for excellent technical assis-
tance and Drs Matthias Ehrich and Dirk van den Boom from
Sequenom, San Diego CA, USA for making available soft-
ware for DNA methylation measurements. Financial support
from The Netherlands Heart Foundation (grants 86.083,
88.042 and 90.313), The Netherlands Organization for Scien-
tific Research NWO (grant 911-03-016) and the Centre for
Medical Systems Biology (CMSB), a centre of excellence
approved by the Netherlands Genomics Initiative/Netherlands
Organization for Scientific Research (NWO), is gratefully
Conflict of Interest statement. No conflicts of interests
1. Jaenisch, R. and Bird, A. (2003) Epigenetic regulation of gene expression:
how the genome integrates intrinsic and environmental signals. Nat.
Genet., 33 (suppl.), 245–254.
2. Jones, P.A. and Baylin, S.B. (2002) The fundamental role of epigenetic
events in cancer. Nat. Rev. Genet., 3, 415–428.
3. Petronis, A. (2006) Epigenetics and twins: three variations on the theme.
Trends Genet., 22, 347–350.
4. Waterland, R.A. and Jirtle, R.L. (2003) Transposable elements: targets for
early nutritional effects on epigenetic gene regulation. Mol. Cell Biol., 23,
5. Jacob, R.A., Gretz, D.M., Taylor, P.C., James, S.J., Pogribny, I.P., Miller,
B.J., Henning, S.M. and Swendseid, M.E. (1998) Moderate folate
depletion increases plasma homocysteine and decreases lymphocyte DNA
methylation in postmenopausal women. J. Nutr., 128, 1204–1212.
(2000) Genomic DNA methylation decreases in response to moderate folate
depletion in elderly women. Am. J. Clin. Nutr., 72, 998–1003.
7. Fraga, M.F., Ballestar, E., Paz, M.F., Ropero, S., Setien, F., Ballestar,
M.L., Heine-Suner, D., Cigudosa, J.C., Urioste, M., Benitez, J. et al.
(2005) Epigenetic differences arise during the lifetime of monozygotic
twins. Proc. Natl Acad. Sci. USA, 102, 10604–10609.
8. Cui, H., Cruz-Correa, M., Giardiello, F.M., Hutcheon, D.F., Kafonek,
D.R., Brandenburg, S., Wu, Y., He, X., Powe, N.R. and Feinberg, A.P.
(2003) Loss of IGF2 imprinting: a potential marker of colorectal cancer
risk. Science, 299, 1753–1755.
9. Cruz-Correa, M., Cui, H., Giardiello, F.M., Powe, N.R., Hylind, L.,
Robinson, A., Hutcheon, D.F., Kafonek, D.R., Brandenburg, S., Wu,
Y. et al. (2004) Loss of imprinting of insulin growth factor II gene: a
potential heritable biomarker for colon neoplasia predisposition.
Gastroenterology, 126, 964–970.
10. Robertson, K.D. (2005) DNA methylation and human disease. Nat. Rev.
Genet., 6, 597–610.
11. O’Dell, S.D., Miller, G.J., Cooper, J.A., Hindmarsh, P.C., Pringle, P.J.,
Ford, H., Humphries, S.E. and Day, I.N. (1997) Apal polymorphism in
Human Molecular Genetics, 2007, Vol. 16, No. 5553
insulin-like growth factor II (IGF2) gene and weight in middle-aged
males. Int. J. Obes. Relat. Metab. Disord., 21, 822–825.
12. Van Laere, A.S., Nguyen, M., Braunschweig, M., Nezer, C., Collette, C.,
Moreau, L., Archibald, A.L., Haley, C.S., Buys, N., Tally, M. et al. (2003)
A regulatory mutation in IGF2 causes a major QTL effect on muscle
growth in the pig. Nature, 425, 832–836.
13. Zaina, S., Pettersson, L., Ahren, B., Branen, L., Hassan, A.B., Lindholm,
M., Mattsson, R., Thyberg, J. and Nilsson, J. (2002) Insulin-like growth
factor II plays a central role in atherosclerosis in a mouse model. J. Biol.
Chem., 277, 4505–4511.
14. Vu, T.H., Li, T., Nguyen, D., Nguyen, B.T., Yao, X.M., Hu, J.F. and
Hoffman, A.R. (2000) Symmetric and asymmetric DNA methylation in
the human IGF2-H19 imprinted region. Genomics, 64, 132–143.
15. Ehrich, M., Nelson, M.R., Stanssens, P., Zabeau, M., Liloglou, T.,
Xinarianos, G., Cantor, C.R., Field, J.K. and van den, B.D. (2005)
Quantitative high-throughput analysis of DNA methylation patterns by
base-specific cleavage and mass spectrometry. Proc. Natl Acad. Sci. USA,
16. Friso, S., Choi, S.W., Girelli, D., Mason, J.B., Dolnikowski, G.G., Bagley,
P.J., Olivieri, O., Jacques, P.F., Rosenberg, I.H., Corrocher, R. et al.
(2002) A common mutation in the 5,10-methylenetetrahydrofolate
reductase gene affects genomic DNA methylation through an interaction
with folate status. Proc. Natl Acad. Sci. USA, 99, 5606–5611.
17. Castro, R., Rivera, I., Ravasco, P., Camilo, M.E., Jakobs, C., Blom, H.J.
and de Almeida, I.T. (2004) 5,10-methylenetetrahydrofolate reductase
(MTHFR) 677C–.T and 1298A–.C mutations are associated with
DNA hypomethylation. J. Med. Genet., 41, 454–458.
18. Sandovici, I., Kassovska-Bratinova, S., Loredo-Osti, J.C., Leppert, M.,
Suarez, A., Stewart, R., Bautista, F.D., Schiraldi, M. and Sapienza,
C. (2005) Interindividual variability and parent of origin DNA
methylation differences at specific human Alu elements. Hum. Mol.
Genet., 14, 2135–2143.
19. Bjornsson, H.T., Fallin, M.D. and Feinberg, A.P. (2004) An integrated
epigenetic and genetic approach to common human disease. Trends
Genet., 20, 350–358.
20. Sandovici, I., Leppert, M., Hawk, P.R., Suarez, A., Linares, Y. and
Sapienza, C. (2003) Familial aggregation of abnormal methylation of
parental alleles at the IGF2/H19 and IGF2R differentially methylated
regions. Hum. Mol. Genet., 12, 1569–1578.
21. Bock, C., Paulsen, M., Tierling, S., Mikeska, T., Lengauer, T. and Walter,
J. (2006) CpG island methylation in human lymphocytes is highly
correlated with DNA sequence, repeats, and predicted DNA structure.
PLoS Genet., 2, e26.
22. Feltus, F.A., Lee, E.K., Costello, J.F., Plass, C. and Vertino, P.M. (2006)
DNA motifs associated with aberrant CpG island methylation. Genomics,
23. Morgan, H.D., Santos, F., Green, K., Dean, W. and Reik, W. (2005)
Epigenetic reprogramming in mammals. Hum. Mol. Genet., 14
(Spec no. 1), R47–R58.
24. Waterland, R.A. and Jirtle, R.L. (2004) Early nutrition, epigenetic changes
at transposons and imprinted genes, and enhanced susceptibility to adult
chronic diseases. Nutrition, 20, 63–68.
25. Jiang, Y.H., Bressler, J. and Beaudet, A.L. (2004) Epigenetics and human
disease. Annu. Rev. Genomics Hum. Genet., 5, 479–510.
26. Martin, G.M. (2005) Epigenetic drift in aging identical twins. Proc. Natl
Acad. Sci. USA, 102, 10413–10414.
27. Boomsma, D.I., Kaptein, A., Kempen, H.J., Gevers Leuven, J.A. and
Princen, H.M. (1993) Lipoprotein(a): relation to other risk factors and
genetic heritability. Results from a Dutch parent-twin study.
Atherosclerosis, 99, 23–33.
28. Snieder, H., van Doornen, L.J. and Boomsma, D.I. (1997) The age
dependency of gene expression for plasma lipids, lipoproteins, and
apolipoproteins. Am. J. Hum. Genet., 60, 638–650.
29. Heijmans, B.T., Beekman, M., Putter, H., Lakenberg, N., van der Wijk,
H.J., Whitfield, J.B., Posthuma, D., Pedersen, N.L., Martin, N.G.,
Boomsma, D.I. et al. (2005) Meta-analysis of four new genome scans for
lipid parameters and analysis of positional candidates in positive linkage
regions. Eur. J. Hum. Genet., 13, 1143–1153.
30. Petry, C.J., Ong, K.K., Barratt, B.J., Wingate, D., Cordell, H.J., Ring,
S.M., Pembrey, M.E., Reik, W., Todd, J.A. and Dunger, D.B. (2005)
Common polymorphism in H19 associated with birthweight and cord
blood IGF-II levels in humans. BMC Genet., 6, 22.
31. Abecasis, G.R. and Cookson, W.O. (2000) GOLD—graphical overview of
linkage disequilibrium. Bioinformatics, 16, 182–183.
32. Barrett, J.C., Fry, B., Maller, J. and Daly, M.J. (2005) Haploview: analysis
and visualization of LD and haplotype maps. Bioinformatics, 21,
554 Human Molecular Genetics, 2007, Vol. 16, No. 5