Content uploaded by Bas Heijmans
Author content
All content in this area was uploaded by Bas Heijmans on Oct 06, 2014
Content may be subject to copyright.
The FASEB Journal •Research Communication
Variation, patterns, and temporal stability of DNA
methylation: considerations for epigenetic
epidemiology
Rudolf P. Talens,* Dorret I. Boomsma,
§
Elmar W. Tobi,* Dennis Kremer,*
J. Wouter Jukema,
†
Gonneke Willemsen,
§
Hein Putter,
‡
P. Eline Slagboom,*
,㥋
and
Bastiaan T. Heijmans*
,㥋,1
*Department of Molecular Epidemiology,
†
Department of Cardiology, and
‡
Department of Medical
Statistics, Leiden University Medical Center, Leiden, The Netherlands;
§
Department of Biological
Psychology, VU University Amsterdam, Amsterdam, The Netherlands; and
㛳
Netherlands Consortium
for Healthy Ageing, Leiden, The Nethelands
ABSTRACT The prospect of finding epigenetic risk
factors for complex diseases would be greatly en-
hanced if DNA from existing biobanks, which is gener-
ally extracted from whole blood, could be used to
perform epigenetic association studies. We character-
ized features of DNA methylation at 16 candidate loci,
8 of which were imprinted, in DNA samples from the
Netherlands Twin Register biobank. Except for un-
methylated or fully methylated sites, CpG methylation
varied considerably in a sample of 30 unrelated indi-
viduals. This variation remained after accounting for
the cellular heterogeneity of blood. Methylation of
CpG sites was correlated within loci and, for 4 im-
printed loci, across chromosomes. In 34 additional
individuals, we investigated the DNA methylation of 8
representative loci in 2 longitudinal blood and 2 longi-
tudinal buccal cell samples (follow-up 11–20 and 2–8
yr, respectively). Five of 8 loci were stable over time
(>0.75) in both tissues, indicating that prospective
epigenetic studies may be possible. For 4 loci, the DNA
methylation in blood (mesoderm) correlated with that
in the buccal cells (ectoderm) (>0.75). Our data suggest
that epigenetic studies on complex diseases may be feasi-
ble for a proportion of genomic loci provided that they
are carefully designed.—Talens, R. P., Boomsma, D. I.,
Tobi, E. W., Kremer, D., Jukema, J. W., Willemsen, G.,
Putter, H., Slagboom, P. E., Heijmans, B. T. Variation,
patterns, and temporal stability of DNA methylation:
considerations for epigenetic epidemiology. FASEB J. 24,
3135–3144 (2010). www.fasebj.org
Key Words: epigenome 䡠biobank 䡠human studies 䡠complex
disease
Epigenetics refers to heritable differences in gene
expression potential that are not caused by variation in
the DNA sequence (1, 2). Its molecular basis is the
chemical modification of either the DNA itself (cyto-
sine methylation in CpG dinculeotides) or the histones
that package the chromatin (e.g., methylation, acetyla-
tion, phosphorylation.) (3–5). It has frequently been
proposed that changes in these epigenetic marks sig-
nificantly contribute to the risk of complex diseases,
including cancer, cardiovascular, and metabolic disease
(1, 6–10). However, with the exception of studies on
cancer, empirical data from epidemiological studies
supporting these hypotheses are largely absent, mainly
due to technical and methodological limitations.
Many of the technical limitations have been resolved,
in particular, with respect to the high-throughput mea-
surement of DNA methylation (11, 12). DNA methyl-
ation is correlated with other layers of epigenetic
marks, particularly histone modifications (13). DNA
methylation may be the most suitable epigenetic mark for
large-scale epidemiological studies, since methyl groups
are covalently bound to CpG dinucleotides and are not
lost during routine DNA extraction, unlike histone mod-
ifications. This opens the possibility of exploiting existing
DNA biobanks for research purposes, to discover epige-
netic risk factors for complex disease.
Epigenetic studies will require the development of
data resources analogous to those that facilitated ge-
netic association studies. The resources should include
epigenome maps charting DNA methylation marks
(14), the description of interindividual variation in
DNA methylation (cf. single nucleotide polymorphisms
and copy number variants) (15), and data on the
patterns within this variation (cf. linkage disequilib-
rium) (16). To guide the development of such epi-
genome-wide resources, candidate loci may be studied.
In this respect, differentially methylated regions influ-
encing imprinting (17), transposon-derived sequences
(17), CpG island shores (18), and recognition se-
quences for methylation-dependent transcription fac-
tors (19) are of particular interest.
In addition, several issues potentially limiting the use
of existing biobanks for epigenetic epidemiology need
to be addressed. First, DNA in biobanks is mostly
extracted from whole blood, which, like any tissue,
1
Correspondence: Molecular Epidemiology, Leiden Uni-
versity Medical Center, Postal Zone S-05-P, PO box 9600,
2300RC, Leiden, NL. E-mail: b.t.heijmans@lumc.nl
doi: 10.1096/fj.09-150490
31350892-6638/10/0024-3135 © FASEB
consists of different cell types that may carry different
epigenetic marks and whose relative numbers may vary
between individuals (20). Second, the stability of DNA
methylation over time should be known before the asso-
ciation of DNA methylation with future disease risk can be
assessed. Global (or average) DNA methylation has been
reported to change over time (21, 22), but DNA methyl-
ation of specific loci may be more stable (16, 23). Finally,
it will be crucial to address the extent to which DNA
methylation measured in blood is a marker for less
accessible tissues that are directly involved in disease.
Despite scattered reports that this may be the case (24–
26), the issue remains largely unresolved.
We assessed whether genomic DNA stored in existing
biobanks would be suitable for epigenetic epidemiological
studies. To this end, we addressed the interindividual varia-
tion in DNA methylation of 16 candidate loci for cardiovas-
cular and metabolic disease, the influence of blood cell
heterogeneity on this variation, the stability of DNA methyl-
ation over time, and its correlation between whole blood
(mesoderm) and buccal cells (ectoderm) in individuals
from the Netherlands Twin Register (NTR) (27, 28).
MATERIALS AND METHODS
Study populations
The individuals investigated in this study were selected from
the Netherlands Twin Register (NTR) biobank (27, 28),
which includes DNA samples from Dutch twins and their
family members (parents, siblings, offspring, and spouses).
First, unrelated individuals were selected (n⫽30; Supplemen-
tal Table 3) to study interindividual variation in DNA meth-
ylation, the influence of cell heterogeneity, and patterns of
DNA methylation. These 30 individuals were selected from
the ongoing NTR biobank project for which 9560 individuals
were included. Random selection would result in a sample of
individuals with characteristics very close to the average in the
complete cohort. For this, we applied the D-optimality crite-
rion to the Fisher information matrix, which enabled us to
select 30 individuals representative of the whole range of phe-
notypic variation in age and metabolic parameters present in the
complete cohort. The age of the individuals selected ranged
from 21 to 73 yr; metabolic parameters of interest included waist
circumference, fasting blood glucose level, serum LDL, and
HDL cholesterol. Plasma and serum measurements and cell
counts of whole blood were obtained using the standardized
methods previously described (missing cell count information
for 2/30 individuals). Furthermore, we selected the proportion
of males to females and of those who had never been smokers,
to former and current smokers, so that it was equal to the
proportion in the complete cohort.
Second, 34 individuals were selected for assessing the
correlation of DNA methylation across time and tissue and for
validating the findings on the group of 30 individuals de-
scribed above. This group consisted of participants in the
NTR biobank project, who also took part in previous NTR
projects. This allowed for recent DNA samples from whole
blood (with information of cell counts) and buccal cells, as
well as previous DNA samples from whole blood drawn 11 to
20 yr earlier, and from buccal swabs taken 2 to 8 yr earlier
(Supplemental Table 4). The age at first sampling ranged
from 14 to 62 yr. Among the 34 individuals, 17 were male, 26
were unrelated individuals, and 8 were included as 4 monozy-
gotic twin pairs. DNA from all samples was extracted from
whole blood and buccal swabs using standard methods.
Third, the results on within-individual correlation between
CpG units were validated using 60 controls (28 males, mean
age 57 yr) from the Dutch Hunger Winter Families Study
(29). DNA methylation was measured at the same loci using
the same methods as used in the current study (30, 31).
DNA methylation
Loci were selected on the basis of their potential involvement
in cardiovascular and metabolic disease through the role of
the adjacent candidate gene in growth, lipid metabolism,
energy metabolism, inflammation, or stress response. Assay
design focused on the regions of these loci that contained
features with a potential for epigenetic regulation as observed
in human, animal, or cell culture experiments (19, 24,
32–44). The loci selected included promoter elements, CpG
TABLE 1. Characteristics of methylation assays
Locus Chromosome Megabase Gene function CpG sites
a
Single CpG sites
b
Imprinted CpG-island
IL10
e
01q32.1 205.01 Anti-inflammation 4 2
NR3C1 05q31.3 142.67 Stress response 20 4 ⫹
TNF 06p21.33 31.65 Proinflammatory 8 5
IGF2R
e
06q25.3 160.35 Growth/apoptosis 10 0 ?⫹
GRB10 07p12.2 50.82 IIS inhibitor 16 5 ⫹⫹
LEP
e
07q32.1 127.67 Metabolism 10 5 ⫹
CRH
e
08q13.1 67.25 Stress response 5 5
ABCA1 09q31.1 106.73 Cholesterol transport 16 3 ⫹
IGF2
e
11p15.5 2.13 Early growth 5 3 ⫹
INSIGF
e
11p15.5 2.14 Embryonic growth 4 4 ?
KCNQ1OT1
e
11p15.5 2.68 Imprinting control region 14 7 ⫹⫹
MEG3 14q32.2 100.36 Growth suppressor 7 3 ⫹⫹
FTO 16q12.2 52.38 Development 10 4
APOC1
e
19q13.32 50.11 Metabolism 6 6
GNASAS 20q13.32 56.86 Growth/lypolytic signal 17 3
GNAS A/B 20q13.32 56.90 Growth/lypolytic signal 12 3 ⫹⫹
a
CpG sites that met the quality criteria described in Materials and Methods.
b
CpG sites of which the methylation proportion was measured
individually.
c
Methylation-sensitive transcription factor-binding sites.
d
CpG methylation previously reported to associate with gene expression.
e
Stability across time and correlation between tissues were also investigated.
3136 Vol. 24 September 2010 TALENS ET AL.The FASEB Journal 䡠www.fasebj.org
islands, transposon-derived sequences, methyl-sensitive tran-
scription factor binding sites (mTFBS), imprinted differentially
methylated regions (DMR), and regions reported to regulate
transcription through DNA methylation. Methylation assays
were designed using the methprimer tool (45) on sections of
sequence downloaded from the University of California, Santa
Cruz genome browser (46). Fifty-eight assays were tested for the
reliability of the methylation measurement. Forty assays gave a
reliable measurement, and on the basis of the priority given to
the associated candidate gene and epigenetic properties (Table 1),
16 of these were selected to cover the whole range of possible
average methylation levels (0–100%).
One microgram of genomic DNA was bisulfite-converted
using the EZ 96-DNA methylation (Zymo Research, Orange,
CA, USA). DNA of the 30 individuals, in whom variation in
DNA methylation was investigated, was converted on a single
96-well plate. DNA methylation of all 16 loci was measured
using the same bisulfite-converted sample. The 4 samples
(blood, buccal, recent, and old) from individuals selected for
testing the correlation over time and across tissues were
bisulfite-treated on the same 96-well plate. For this substudy,
two 96-well plates were used to process the 136 samples, each
plate with an equal number of individuals. Methylation of the
8 loci was measured using a single bisulfite-converted DNA
sample. Primers used to amplify the region of each assay are
given in Supplemental Table 5A. DNA methylation was mea-
sured using a mass spectrometry-based method (Epityper;
Sequenom, San Diego, CA, USA) (47), whose quantitative
accuracy (R
2
duplicate measurementsⱖ0.98) and concor-
dance with clonal PCR bisulfite sequencing was reported
previously (48, 49). All measurements were done in triplicate.
Quality control consisted of several steps. CpG site-containing
fragments that had equal or overlapping mass, making them
irresolvable by mass spectrometry, and CpG sites containing
fragments whose measurement was confounded by single-
nucleotide polymorphisms (16) according to dbSNP build
128 were discarded (Supplemental Table 6). Next, at least 2
of the 3 replicate measurements had to be successful, and the
sd of the replicate measurements had to be ⱕ0.10. Only CpG
sites with a success rate ⬎75% for the latter two criteria were
considered fit for further analysis, and the average was
calculated for the replicate measurements. With these criteria
applied, DNA methylation of 164 CpG sites, distributed over
104 CpG site-containing fragments (CpG units; ref. 47), could
be measured in the first sample of 30 individuals (Supplemental
Table 5B); 62 CpG units contained 1 CpG site, 28 CpG units
contained 2 CpG sites, 10 CpG units contained 3 CpG sites, and
4 CpG units contained 4 CpG sites. The methylation of multiple
CpG sites occurring on one fragment (CpG unit) cannot be
resolved individually. Average CpG methylation for these CpG
units was calculated using the RSeqMeth module (48). The
average success rate for the 104 CpG units assessed was 97%. In
the second sample of 34 individuals, 41 CpG units containing 55
CpG sites could be measured, applying the same criteria only to
the recent blood samples and 38 CpG units containing 52 CpG
sites to all 4 samples (Supplemental Table 5C).
To exclude the influence of DNA sequence variation not
present in dbSNP on higher correlations observed between
DNA methylation measured on a recent and an old sample
and on a blood and a buccal swab sample, the evaluateSNPs()
function of the R-module MassArray was used (50). The
Epityper method for DNA methylation measurements is
based on a protocol to resequence genomic DNA using mass
spectrometry (MassCleave; Sequenom) (51). By comparing
the mass spectrum observed with the one expected according
to a reference sequence, data points can be identified that are
suspected of being confounded by sequence variation. This
interference can be direct when the sequence variant affects
a fragment containing a CpG site or indirect when the
sequence variant changes the mass of a non-CpG fragment so
that it overlaps with a CpG-containing fragment. It is note-
worthy that sequencing genomic DNA would deal with direct
interference only. Data points suspected to be affected by
unknown sequence variation were excluded, and correlations
were recalculated to examine their influence. CpG measure-
ments were removed for APOC1, CpG 1 [4 individuals ex-
cluded for all DNA samples (recent, old, blood, buccal)],
CpG 10 (8 exclusions) and CpG 11 (11 exclusions); LEP, CpG
8 (1 exclusion); IGF2, CpG 6–7 (21 exclusions); IGF2R, CpG
4–5 (1 exclusion), and CpG 11–13 (4 exclusions); for CRH,
no measurements were excluded.
Statistical analyses
Variation in DNA methylation
CpG sites with average methylation levels close to 0 or 100%
by definition have a truncated variance. To circumvent this
TABLE 1. (continued)
Promoter Intragenic Intergenic Putative mTFBS
c
Confirmed mTFBS
c
Transposon Literature
d
⫹36⫹⫹(32)
⫹32 ⫹(19)
⫹2⫹(34)
⫹⫹ 1⫹(35)
⫹42 ⫹(36)
⫹3⫹(37)
⫹11 ⫹(38)
⫹44 ⫾(39)
⫹⫹ ⫹(24)
2⫹(40)
⫹ ⫹(41)
⫹11 ⫹(42)
⫹ ⫹
⫹3⫾(43)
⫹1⫹(44)
⫹ ⫹(33)
3137EPIGENETIC HUMAN POPULATION STUDIES
problem, the following variance-stabilizing transformation
was applied (52): Transformed value ⫽Arc tan {[methyl-
ation/(1 ⫺methylation)]
2
}. Using the transformed values, we
tested the equality of the variance of CpG sites with Levene’s
test.
Accounting for cellular heterogeneity
To test whether variation in DNA methylation was con-
founded by cellular heterogeneity, nested linear mixed mod-
els (53) were applied to the transformed data. The basic
model was created as a baseline, to be subtracted from the
nested models. It included the CpG site as a fixed effect. The
3 nested models each had one percentage of a major white
blood cell subclass, namely neutrophils, lymphocytes, or
monocytes, added to the basic model as an extra fixed effect
to test whether variation in this percentage could explain part
of the variation in DNA methylation. The actual amount of
variation in DNA methylation that could be explained by the
percentage of the white blood cell type was calculated as
100% minus the percentage of the residual variance of the
nested model with respect to the residual variance of the basic
model. The linear mixed model accounts for correlated
methylation within individuals and deals with methylation
data missing at random without imputation (30). It may be
seen as an extension of the paired ttest: the model will reduce
to a paired ttest with identical results if between-group
methylation differences are assessed for a single CpG site and
if data are complete and all other factors are omitted.
Correlation and patterns of CpG methylation
Bivariate Pearson correlation coefficients between CpG sites
were calculated after adjusting the transformed methylation
levels for neutrophil percentage. Nonsignificant correlations
were treated as if there were no correlations (value of
correlation set to 0). Patterns in the correlation matrix were
visualized with a heat map after unsupervised complete
linkage clustering, which utilizes the Euclidean distance (the
difference between two points in the matrix squared).
Stability over time
The difference between DNA methylation at two time points
was calculated per individual for each CpG unit as methyl-
ation of the old sample minus methylation of the new sample.
Missing values were excluded pairwise. Spearman’s rank
correlation coefficient () was used to calculate the correla-
tion between the two time points.
Correlation between tissues
The difference between DNA methylation in blood and
buccal cells was calculated per individual for each CpG unit as
methylation of the recent blood sample minus methylation of
the recent buccal cell sample. Missing values were excluded
pairwise. Spearman’s was used to calculate the correlation
between the two tissues.
Male and female data were analyzed separately in every
test. However, in view of the study size, the outcomes of these
analyses should be considered as purely explorative. All P
values are 2-sided, and statistical analyses were performed
using SPSS 16.0 (SPSS Inc., Chicago, IL, USA).
RESULTS
Interindividual variation in DNA methylation
We quantitatively measured the DNA methylation of 16
candidate loci using DNA samples from the NTR
biobank, to estimate the variation in the general pop-
ulation. The DNA was extracted from whole blood. The
DNA samples were from 30 unrelated individuals, who
were selected to represent the broad range in age and
metabolic parameters of the whole biobank. After
removing CpGs, for which local genetic polymorphisms
could interfere with DNA methylation measurements,
and CpGs that did not meet the quality criteria, we
recorded methylation of 164 CpG sites, distributed over
104 CpG units, of which 62 contained a single CpG site
(Table 1). The average methylation of the loci studied
ranged from 0 to 98% (Fig. 1). Within the majority of
loci, the average methylation of CpG sites was similar.
Exceptions were LEP, with a 34% methylation differ-
ence between CpG sites located 18 bases apart, and
ABCA1, with a 31% difference between sites 26 bases
apart. Previous studies reported methylation differ-
ences between men and women (31, 54). We could not
detect such differences in an explorative analysis. This
may be related to the study size and the number of tests
performed.
A considerable interindividual variation in CpG
methylation was observed. The variation approximated
a normal distribution except for CpG units that showed
no or very little variation (average DNA methylation
close to 0 or 100%). The extent of this variation varied
per CpG unit (0⬍sd⬍15%). To exclude the possibility
that this difference in variation might merely be due to
the fact that DNA methylation is truncated at 0 and
100%, a variance stabilizing transformation was applied
(52). The variation remained significantly different
between CpG units (P⫽10
⫺16
). To validate these find-
ings, we measured the DNA methylation of 8 of 16 loci
(55 CpG sites, 31 CpG units with 1 CpG site) in 34
additional individuals from the NTR biobank, and this
yielded similar results (Supplemental Fig. 1).
Cellular heterogeneity
DNA methylation was measured on genomic DNA
extracted from whole blood. As whole blood consists of
different cell types, which may display differences in
DNA methylation, we tested the extent to which the
interindividual variation observed was influenced by
cellular heterogeneity, as assessed by counting the
major cell subclasses. The largest cellular fraction con-
sisted of neutrophils (average proportion 53.7⫾8.8%),
thus contributing the majority of DNA molecules on
which CpG methylation was analyzed. For 10 of 16 loci,
the variation in DNA methylation was not associated
with this measure of cellular heterogeneity (Table 2).
For the remaining loci, the variation in DNA methyl-
ation that could be explained by variation in cellular
heterogeneity was generally small, and associations
were of borderline significance. IL10, which is highly
expressed in leukocytes, was a notable exception: 50%
3138 Vol. 24 September 2010 TALENS ET AL.The FASEB Journal 䡠www.fasebj.org
of its variation in DNA methylation could be attributed
to the neutrophil cell count. The analysis was repeated
using the lymphocyte percentage (average proportion
34.4⫾8.4%), which was highly correlated with the neu-
trophil percentage (r⫽⫺0.95), and similar results were
observed (data not shown). No influence of monocyte
percentage was observed on DNA methylation (average
8.0⫾1.8%; correlation with the neutrophil proportion:
r⫽⫺0.48). The influence of cellular heterogeneity was
not affected by gender (data not shown).
To validate these findings, we performed the same
test on the 8 loci in an additional 34 individuals
(Supplemental Table 1). The loci previously not show-
ing an association were again not associated with the
neutrophil percentage (IGF2R,IGF2,INSGF, and
KCNQ1OT1). Two of the associations of borderline
significance were not found (APOC1 and CRH), but the
modest association of LEP with the neutrophil percent-
age was replicated (P⫽1.0⫻10
⫺4
). Again, a substantial
proportion of the variation in IL10 methylation could
be attributed to the neutrophil percentage (27.9%,
P⫽8.0⫻10
⫺8
).
Correlations and patterns of CpG methylation
To investigate patterns of DNA methylation further
within and across loci, correlations between the meth-
ylation of CpG sites were computed and visualized
using a heat map after unsupervised clustering (Fig. 2).
CpG methylation was particularly correlated within loci
(r
max
⫽0.95) but also across loci (r
max
⫽0.68). The cluster
of loci correlating irrespectively of chromosomal location
included paternally imprinted loci (MEG3 and GNASAS)
and maternally imprinted loci (GRB10,KCNQ1OT1, and
GNAS A/B). These observations were unaffected by vari-
ance-stabilizing transformation or adjustment for cell het-
erogeneity prior to analysis (data not shown).
The correlation was studied for the same 16 loci in 60
controls of the Dutch Hunger Winter Families Study that
we measured previously (30, 31). The heat map of the
correlations revealed similar patterns (Supplemental Fig.
2). Again, a strong correlation was observed within loci.
Across loci, significant correlations were observed for
GNASAS,GNAS A/B,GRB10, and MEG3. In addition, a
correlation between these loci and IGF2 was observed,
Figure 1. Interindividual variation in DNA methylation. Methylation percentage, yaxis, at every CpG unit, xaxis, for each of the
30 individuals, colored dots. Order of the loci is based on their chromosomal location, starting at the lowest designation. CpG
unit numbers are counted from the forward primer onwards. Name of each locus is given below the xaxis. Bold horizontal bar
gives the median; thin horizontal bars show the interquartile range for each CpG unit. Vertical lines across the plot separate the
loci. Corresponding CpG sites of each CpG unit are given in Supplemental Table 5B. Individually measured CpG sites are
marked with an S below the unit number.
3139EPIGENETIC HUMAN POPULATION STUDIES
whereas the correlation with KCNQ1OT1 was not repro-
duced. The correlations were similar for both sexes (data
not shown).
Stability over time
To study the stability of DNA methylation over time, we
selected 34 additional individuals from the NTR for
whom two blood samples were taken 11–20 yr apart.
The methylation of 8 loci that were representative of
the set of 16 loci was measured (Table 1). Overall, DNA
methylation was similar at the two time points (Fig. 3A),
and only minor differences were observed (Table 3).
Similar average methylation levels between the time
points do not indicate stability per se since methylation
may increase in some and decrease in other individuals
over time. The variation around the average difference
was greatest for IL10, which also showed the greatest
average difference (⫺2.8⫾9.1%; Table 3). It was lowest
for KCNQ1OT1 (sd⫽2.8%), indicating relative stability
over time. An alternative way to express stability, which
takes into account the differences in interindividual
variation of the loci, is to compute correlation coeffi-
cients (Table 3). For 5 of the 8 loci, the correlation ⬎
0.75 indicated substantial stability between the time
points. These loci included IGF2R (⫽0.88) and APOC1
(⫽0.96). Note that the correlation was low for
KCNQ1OT1 (⫽0.31), which can be attributed to the
very low level of interindividual variation. Temporal
stability was similar in both sexes (data not shown).
From the same 34 individuals, DNA samples from
buccal swabs taken 2–8 yr apart were available and
showed similar results (Supplemental Table 2).
To exclude the possibility that the higher correla-
tions observed were due to sequence variation not
present in dbSNP, we used the mass spectra to identify
CpG methylation measurements that were suspected to
have been influenced by sequence variation (50). This
was the case for 1 or more individuals for 7/41 CpG units.
Removal of these CpG measurements did not affect the
correlations ⬎0.75 (IGF2R,⫽0.87; LEP,⫽0.90; IGF2,
⫽0.92; CRH,⫽0.94; and APOC1,⫽0.95).
Correlation of CpG methylation between tissues
To test whether DNA methylation in blood could mark
that in other tissues, we studied DNA methylation of the
8 loci in the recent blood (mesoderm) and buccal swab
(ectoderm) samples of the individuals in whom stability
over time was tested. The average level of DNA meth-
ylation was generally different between the two tissues
and the extent of the difference depended on the locus
(Fig. 3B,Table 4). The variation around the average
TABLE 2. Association of neutrophil proportion with DNA
methylation
Locus Variance explained (%) Pvalue of effect
IL10 50.1 3.9 ⫻10
⫺06
NR3C1 0.2 0.555
TNF 8.0 0.037
IGF2R 5.0 0.208
GRB10 0.7 0.625
LEP 7.4 0.019
CRH 4.0 0.022
ABCA1 7.3 0.021
IGF2 3.5 0.185
INSIGF 0.3 0.674
KCNQ1OT1 0.4 0.733
MEG3 4.1 0.165
FTO 0.1 0.714
APOC1 6.2 0.026
GNASAS 0.8 0.590
GNAS A/B 0.0 0.889
Figure 2. Correlation between CpG sites within
and across loci. Heat map depicting correla-
tions between the methylation levels of all CpG
units of the 16 loci measured in the first sample
of 30 individuals. For reference, the CpG units
are annotated by a color, based on the locus
(key at right). Diagonal axis running from
bottom left to top right corner is the line of
symmetry where each CpG unit hypothetically
correlates with itself. Full correlation (⫽1) is
plotted as the brightest red shade; full inverse
correlation (⫽⫺1) is plotted as the brightest
green shade; no correlation (⫽0) is plotted as
black (key at top left). Nonsignificant correla-
tions are depicted as no correlation. Complete
clustering is based on the Euclidean distance.
3140 Vol. 24 September 2010 TALENS ET AL.The FASEB Journal 䡠www.fasebj.org
difference also varied per locus. Again, IL10 showed the
highest variation (sd⫽17.1), and KCNQ1OT1 showed
the lowest (sd⫽3.0). For all loci, the sd of the average
difference between the tissues was higher than that of
the average difference between the time points. For 4
of the 8 loci, the correlation of DNA methylation
between the two tissues was ⬎0.75. These loci included
IGF2R (⫽0.83) and APOC1 (⫽0.82). Again, the cor-
relations were similar in both sexes (data not shown).
After removing CpGs measurements suspected to be
influenced by sequence variation not present in dbSNP,
similar correlations were found (IGF2R,⫽0.82; LEP,
⫽0.80; CRH,⫽0.90; and APOC1,⫽0.81).
DISCUSSION
Epigenetic risk factors are thought to contribute to the
development of common diseases such as cardiovascu-
lar and metabolic disease (6, 9, 10). Here, we investi-
gated whether genomic DNA from existing biobanks is
suitable for the identification of these risk factors in
epidemiological studies (1, 7).
Using genomic DNA from the Netherlands Twin Reg-
ister biobank (27, 28), we first assessed the interindividual
variation in DNA methylation for 16 candidate loci, since
the human epigenome map is still in development (14),
and epigenome-wide resources on variation (i.e., the
epivariome) are lacking. We observed considerable varia-
tion in CpG methylation between individuals, except for
loci that are either not methylated or fully methylated.
The extent of this variation varied between CpG sites.
Earlier reports frequently characterized CpG methylation
as hypomethylation, isomethylation, or hypermethylation
(24, 55). Our data support our own previous work (16)
and that of others (15) in which DNA methylation was
more accurately described as a quantitative trait.
Second, we addressed the possibility that the varia-
tion in DNA methylation could simply be attributed to
cellular heterogeneity in leukocytes between individu-
als (20). Blood, like any tissue, consists of a mixture of
different cell types that all may have a cell-specific
epigenome (3). Our results show that for the large
majority of candidate loci, interindividual differences
in the cellular composition of the blood sample did not
contribute to the variation observed in DNA methyl-
ation or explained only a minor proportion of this
variation. One notable exception was the IL10 locus,
for which cellular heterogeneity explained up to half of
the total variation in DNA methylation. If cell counts
are available for whole blood samples stored in a
biobank, the potentially confounding influence of cel-
Figure 3. Temporal stability and comparison between blood
and buccal cell DNA methylation. Scatterplots for individual
comparison of CpG methylation between the DNA samples.
CpG units of each individual are annotated by coloring based
on the locus. Diagonal x⫽yline is plotted in black for
reference. A) CpG methylation in the first blood sample (x
axis) is plotted against methylation in the second, more
recent, blood sample (yaxis). B) CpG methylation in the
recent blood DNA sample (xaxis) is plotted against methyl-
ation in the recent buccal swab DNA sample (yaxis). Each dot
represents 1 CpG unit of 1 individual in both DNA samples.
TABLE 3. Comparison of DNA methylation in blood samples of the two time points
Locus
Methylation (%)
Difference (%) Spearman’s Old blood New blood
IL10 22.4 ⫾9.0 25.2 ⫾6.6 ⫺2.8 ⫾9.1 0.422
IGF2R 65.8 ⫾16.8 67.6 ⫾17.7 ⫺1.8 ⫾8.1 0.883
LEP 20.0 ⫾11.5 21.8 ⫾13.0 ⫺1.8 ⫾6.1 0.895
CRH 63.4 ⫾22.1 63.8 ⫾21.1 ⫺0.4 ⫾6.7 0.942
IGF2 49.4 ⫾11.8 49.0 ⫾11.3 0.4 ⫾4.7 0.924
INSIGF 86.4 ⫾4.1 85.5 ⫾4.3 0.9 ⫾3.7 0.649
KCNQ1OT1 30.8 ⫾2.2 31.8 ⫾2.6 ⫺1.0 ⫾2.8 0.307
APOC1 19.7 ⫾11.4 19.6 ⫾11.7 0.1 ⫾3.4 0.956
Values are means ⫾sd.
3141EPIGENETIC HUMAN POPULATION STUDIES
lular heterogeneity can be monitored using standard
statistical methods. If no data on cellular heterogeneity
are available, it may be necessary to exclude the associ-
ation of cellular heterogeneity either with the outcome
of interest or with methylation of the locus studied. The
latter can be addressed, for example, in a substudy for
which data on leukocyte populations are available. Our
study suggests that no such relationship will be ob-
served for many loci, in which case biobanks without
data on cellular heterogeneity may still be useful.
Third, we investigated patterns in CpG methylation
within and across loci. We found that within the locus,
CpG methylation is highly correlated, except for candi-
date loci that were not methylated or fully methylated,
which corroborates recent findings (16, 56). This obser-
vation suggests that assessing the methylation of a subset
of CpGs is sufficient to cover the variation in DNA
methylation at a locus. This is analogous to genetic
association studies in which a small number of tagging
SNPs can cover all genetic variation at a locus due to
linkage disequilibrium (57). Moreover, our results provide
the first indication that methylation of CpG sites can also be
correlated irrespective of their chromosomal location. This
was observed for a subset of mainly imprinted loci, which
may be related to the mechanisms responsible for establish-
ing methylation marks at DMRs (58, 59).
Fourth, since DNA methylation is a reversible process
(60), it may not be stable over time. If so, this would
preclude conclusions about causality in epidemiological
studies, since DNA methylation may change during a
follow-up period. The majority of loci tested were stable
over time in DNA from blood and buccal cells despite
possible changes in cellular composition during the fol-
low-up period. The fact that we investigated DNA samples
that were taken 11 to 20 yr apart, implies that these DNA
methylation marks may be investigated in most prospec-
tive cohort studies in which participants are followed for
the development of disease for similar or shorter fol-
low-up periods. However, for a minority of loci, we found
that although, on average, there was no difference in
DNA methylation between the time points, the correla-
tion was lower, indicating relaxed maintenance of these
DNA methylation marks. This data resemble recent re-
sults on global DNA methylation studying similar fol-
low-up periods (22). The age of the individuals in our
study was limited to young and middle ages (14 to 62 yr
old). Therefore, we cannot exclude instability over very
long periods of time, nor can we exclude the occurrence
of greater changes in old age. Indeed, instability of the
DNA methylation marks in old age has been reported for
both locus-specific (23, 30) and global (61) DNA methyl-
ation. Moreover, our study did not address the possible
occurrence of changes in DNA methylation as a conse-
quence of disease or processes preceding its clinical
manifestation (62).
DNA from existing biobanks generally is extracted from
easily accessible tissues such as blood. Future studies may
reveal DNA methylation patterns in such tissues that mark
the risk of disease. As a first step toward establishing a
possible causal role, it will be necessary to determine that
DNA methylation measured in peripheral tissues is asso-
ciated with that in tissues directly involved in the disease of
interest. Although DNA methylation is thought to be a
mechanism driving cell differentiation leading to tissue-
specific differentially methylated regions (18), initial re-
ports indicated that DNA methylation measured in blood
may be informative. For example, IGF2 and ER-␣methyl-
ation in blood marked that of colon tissue (24, 25). Also,
an autopsy study of 6 subjects and 11 tissues, which did not
include blood, suggested that the hypomethylation and
hypermethylation status of loci is commonly preserved
across tissues (26). Comparing the methylation of candi-
date loci in blood and buccal cells, we found that for half of
the loci tested, DNA methylation measured in blood was a
marker for that in buccal cells. These results are promising,
since blood and buccal cells stem from different germ layers
(mesoderm and ectoderm, respectively) and warrant the
investigation of correlations with other tissues involved in
disease. Genome-scale studies in particular, will be informa-
tive for defining the (sequence) characteristics of loci show-
ing correlations across tissues. Such studies will be required
to interpret the results of epidemiological studies on DNA
methylation in blood in a meaningful way.
Our study on the suitability of DNA from existing
biobanks for epigenetic studies provides leads for setting
up new biobanks specifically aimed at epigenetic epide-
miology. Since the correlation between DNA methylation,
as measured in DNA from blood and other tissues (di-
rectly involved in disease), appears to be complex and
locus-dependent, such initiatives should ideally include
efforts to sample tissues others than blood (mesoderm),
at least for a subgroup representing the cohort. Tissues
representing the three germ layers for which collection is
feasible include the mesoderm: biopsies of skeletal mus-
TABLE 4. Comparison of DNA methylation in recent blood and buccal cell samples
Locus
Methylation (%)
Difference (%) Spearman’s Blood Buccal cells
IL10 24.8 ⫾7.1 64.9 ⫾18.8 ⫺40.1 ⫾17.1 0.442
IGF2R 68.3 ⫾17.4 81.6 ⫾12.4 ⫺13.3 ⫾10.6 0.827
LEP 21.7 ⫾13.7 11.8 ⫾8.8 9.9 ⫾9.9 0.798
CRH 63.8 ⫾21.9 58.4 ⫾21.4 5.4 ⫾7.9 0.905
IGF2 48.8 ⫾11.5 32.4 ⫾11.0 16.4 ⫾10.8 0.557
INSIGF 85.4 ⫾4.3 84.4 ⫾4.7 1.0 ⫾4.6 0.371
KCNQ1OT1 32.0 ⫾2.6 34.7 ⫾3.1 ⫺2.7 ⫾3.0 0.481
APOC1 19.9 ⫾11.7 10.7 ⫾9.6 9.2 ⫾7.6 0.822
Values are means ⫾sd.
3142 Vol. 24 September 2010 TALENS ET AL.The FASEB Journal 䡠www.fasebj.org
cle, subcutaneous fat and the dermal layer of a skin punch
biopsy (fibroblasts); the ectoderm: the epidermal layer of
a skin punch biopsy (keratinocytes) and buccal cells; and
the endoderm: a urine sample (bladder lining) and a
stool sample (colonic mucosa). In addition, follow-up
sampling of the various tissues should be included (at
least for a subgroup) to assess temporal stability and
changes in DNA methylation as a consequence of pathol-
ogy. To account for the cellular heterogeneity of blood
samples, blood cell populations should be counted if
whole blood is biobanked. This can easily be done using
cheap, routine methods. An alternative approach to re-
duce the cellular heterogeneity is to store peripheral
blood mononuclear cells (PBMCs) instead of whole
blood. PBMCs include lymphocytes (T and B cells) and
monocytes, while the granulocytes (mainly neutrophils)
are lost. To completely remove cellular heterogeneity,
cells can be separated using magnetic-activated cell sort-
ing (MACS). However, this is exceedingly costly and will
not be feasible for larger numbers in most projects.
Taken together, our results indicate that there are good
prospects for the use of existing biobanks for epigenetic
studies. Loci that are suitable for testing in epigenetic studies
demonstrate interindividual variation in DNA methylation,
stability of this variation in DNA methylation over time, and
a correlation between DNA methylation as measured in
blood and the tissue of interest. Our data shows that meeting
these criteria is locus-dependent. Therefore, it may be nec-
essary to address these aspects for each combination of locus,
tissue, and disease in new studies.
The authors thank Dr. L. H. Lumey (Mailman School of
Public Health, Columbia University, New York, NY, USA), for
access to the Dutch Hunger Winter Family Study samples
used in our analysis. This work was supported by grants from
the Netherlands Heart Foundation (NHS2006B083); the
Netherlands Organization for Scientific Research (NWO);
the twin-family database for behavior genomics studies (480-
04-004, 911-03-016); Spinozapremie (SPI 56-464-14192); the
Centre for Medical Systems Biology (CMSB) in the framework of
the Netherlands Genomics Initiative (NGI) and the Centre for
Neurogenomics and Cognitive Research (CNCR-VU); genome-
wide analyses of European twin and population cohorts (EU/
QLRT-2001-01254); the European Union-funded Network of
Excellence LifeSpan (FP6 036894); and the NGI/NWO-funded
Netherlands Consortium for Healthy Ageing (050 60 810).
REFERENCES
1. Waterland, R. A., and Michels, K. B. (2007) Epigenetic Epide-
miology of the developmental origins hypothesis. Annu. Rev.
Nutr. 27, 363–388
2. Jaenisch, R., and Bird, A. (2003) Epigenetic regulation of gene
expression: how the genome integrates intrinsic and environ-
mental signals. Nat. Genet. 33(Suppl.), 245–254
3. Suzuki, M. M., and Bird, A. (2008) DNA methylation landscapes:
provocative insights from epigenomics. Nat. Rev. Genet. 9, 465–476
4. Kouzarides, T. (2007) Chromatin modifications and their func-
tion. Cell 128, 693–705
5. Kouzarides, T. (2007) SnapShot: histone-modifying enzymes.
Cell 128, 802
6. Petronis, A. (2001) Human morbid genetics revisited: relevance
of epigenetics. Trends Genet. 17, 142–146
7. Foley, D. L., Craig, J. M., Morley, R., Olsson, C. J., Dwyer, T.,
Smith, K., and Saffery, R. (2009) Prospects for epigenetic
epidemiology. Am. J. Epidemiol. 169, 389– 400
8. Feinberg, A. P. (2008) Epigenetics at the epicenter of modern
medicine. JAMA 299, 1345–1350
9. Pons, D., de Vries, F. R., van den Elsen, P. J., Heijmans, B. T.,
Quax, P. H., and Jukema, J. W. (2009) Epigenetic histone
acetylation modifiers in vascular remodelling: new targets for
therapy in cardiovascular disease. Eur. Heart. J. 30, 266–277
10. Turunen, M. P., Aavik, E., and Yla-Herttuala, S. (2009) Epige-
netics and atherosclerosis. Biochim. Biophys. Acta 1790, 886– 891
11. Shen, L., and Waterland, R. A. (2007) Methods of DNA meth-
ylation analysis. Curr. Opin. Clin. Nutr. Metab. Care 10, 576–581
12. Smith, Z. D., Gu, H., Bock, C., Gnirke, A., and Meissner, A.
(2009) High-throughput bisulfite sequencing in mammalian
genomes. Methods 48, 226–232
13. Bernstein, B. E., Meissner, A., and Lander, E. S. (2007) The
mammalian epigenome. Cell 128, 669– 681
14. Lister, R., Pelizzola, M., Dowen, R. H., Hawkins, R. D., Hon, G.,
Tonti-Filippini, J., Nery, J. R., Lee, L., Ye, Z., Ngo, Q. M., Edsall,
L., ntosiewicz-Bourget, J., Stewart, R., Ruotti, V., Millar, A. H.,
Thomson, J. A., Ren, B., and Ecker, J. R. (2009) Human DNA
methylomes at base resolution show widespread epigenomic
differences. Nature 462, 315–322
15. Rakyan, V. K., Hildmann, T., Novik, K. L., Lewin, J., Tost, J., Cox, A. V.,
Andrews, T. D., Howe, K. L., Otto, T., Olek, A., Fischer, J., Gut, I. G.,
Berlin, K., and Beck, S. (2004) DNA methylation profiling of the
human major histocompatibility complex: a pilot study for the human
epigenome project. PLoS Biol. 2, e405
16. Heijmans, B. T., Kremer, D., Tobi, E. W., Boomsma, D. I., and
Slagboom, P. E. (2007) Heritable rather than age-related environ-
mental and stochastic factors dominate variation in DNA methyl-
ation of the human IGF2/H19 locus. Hum. Mol. Genet. 16, 547–554
17. Waterland, R. A., and Jirtle, R. L. (2003) Transposable elements:
targets for early nutritional effects on epigenetic gene regula-
tion. Mol. Cell. Biol. 23, 5293–5300
18. Irizarry, R. A., Ladd-Acosta, C., Wen, B., Wu, Z., Montano, C.,
Onyango, P., Cui, H., Gabo, K., Rongione, M., Webster, M., Ji, H.,
Potash, J. B., Sabunciyan, S., and Feinberg, A. P. (2009) The human
colon cancer methylome shows similar hypo- and hypermethylation at
conserved tissue-specific CpG island shores. Nat. Genet. 41, 178–186
19. Weaver, I. C., Cervoni, N., Champagne, F. A., D’Alessio, A. C.,
Sharma, S., Seckl, J. R., Dymov, S., Szyf, M., and Meaney, M. J.
(2004) Epigenetic programming by maternal behavior. Nat.
Neurosci. 7, 847–854
20. Martin, G. M. (2005) Epigenetic drift in aging identical twins.
Proc. Natl. Acad. Sci. U. S. A. 102, 10413–10414
21. 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., Boix-Chornet, M., Sanchez-Aguilera, A., Ling, C.,
Carlsson, E., Poulsen, P., Vaag, A., Stephan, Z., Spector, T. D.,
Wu, Y. Z., Plass, C., and Esteller, M. (2005) Epigenetic differ-
ences arise during the lifetime of monozygotic twins. Proc. Natl.
Acad. Sci. U. S. A. 102, 10604 –10609
22. Bjornsson, H. T., Sigurdsson, M. I., Fallin, M. D., Irizarry, R. A.,
Aspelund, T., Cui, H., Yu, W., Rongione, M. A., Ekstrom, T. J.,
Harris, T. B., Launer, L. J., Eiriksdottir, G., Leppert, M. F.,
Sapienza, C., Gudnason, V., and Feinberg, A. P. (2008) Intra-
individual change over time in DNA methylation with familial
clustering. JAMA 299, 2877–2883
23. Ito, Y., Koessler, T., Ibrahim, A. E., Rai, S., Vowler, S. L.,
bu-Amero, S., Silva, A. L., Maia, A. T., Huddleston, J. E.,
Uribe-Lewis, S., Woodfine, K., Jagodic, M., Nativio, R., Dunning,
A., Moore, G., Klenova, E., Bingham, S., Pharoah, P. D.,
Brenton, J. D., Beck, S., Sandhu, M. S., and Murrell, A. (2008)
Somatically acquired hypomethylation of IGF2 in breast and
colorectal cancer. Hum. Mol. Genet. 17, 2633–2643
24. 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
25. Ally, M. S., Al-Ghnaniem, R., and Pufulete, M. (2009) The
relationship between gene-specific DNA methylation in leuko-
cytes and normal colorectal mucosa in subjects with and without
colorectal tumors. Cancer Epidemiol. Biomarkers Prev. 18, 922–928
26. Byun, H. M., Siegmund, K. D., Pan, F., Weisenberger, D. J.,
Kanel, G., Laird, P. W., and Yang, A. S. (2009) Epigenetic
profiling of somatic tissues from human autopsy specimens
identifies tissue- and individual-specific DNA methylation pat-
terns. Hum. Mol. Genet. 18, 4808– 4817
3143EPIGENETIC HUMAN POPULATION STUDIES
27. Boomsma, D. I., de Geus, E. J., Vink, J. M., Stubbe, J. H., Distel,
M. A., Hottenga, J. J., Posthuma, D., van Beijsterveldt, T. C.,
Hudziak, J. J., Bartels, M., and Willemsen, G. (2006) Nether-
lands Twin Register: from twins to twin families. Twin Res. Hum.
Genet. 9, 849– 857
28. Boomsma, D. I., Willemsen, G., Sullivan, P. F., Heutink, P.,
Meijer, P., Sondervan, D., Kluft, C., Smit, G., Nolen, W. A.,
Zitman, F. G., Smit, J. H., Hoogendijk, W. J., van, D. R., de Geus,
E. J., and Penninx, B. W. (2008) Genome-wide association of
major depression: description of samples for the GAIN Major
Depressive Disorder Study: NTR and NESDA biobank projects.
Eur. J. Hum. Genet. 16, 335–342
29. Lumey, L., Stein, A. D., Kahn, H. S., van der Pal-de Bruin, K. M.,
Blauw, G., Zybert, P. A., and Susser, E. S. (2007) Cohort profile: the
Dutch Hunger Winter Families Study. Int. J. Epidemiol. 36, 1196–1204
30. Heijmans, B. T., Tobi, E. W., Stein, A. D., Putter, H., Blauw, G. J.,
Susser, E. S., Slagboom, P. E., and Lumey, L. H. (2008) Persistent
epigenetic differences associated with prenatal exposure to famine in
humans. Proc. Natl. Acad. Sci. U. S. A. 105, 17046 –17049
31. Tobi, E. W., Lumey, L. H., Talens, R. P., Kremer, D., Putter, H., Stein,
A. D., Slagboom, P. E., and Heijmans, B. T. (2009) DNA methylation
differences after exposure to prenatal famine are common and
timing- and sex-specific. Hum. Mol. Genet. 18, 4046– 4053
32. Dong, J., Ivascu, C., Chang, H. D., Wu, P., Angeli, R., Maggi, L.,
Eckhardt, F., Tykocinski, L., Haefliger, C., Mowes, B., Sieper, J.,
Radbruch, A., Annunziato, F., and Thiel, A. (2007) IL-10 is
excluded from the functional cytokine memory of human
CD4⫹memory T lymphocytes. J. Immunol. 179, 2389–2396
33. Hayward, B. E., Kamiya, M., Strain, L., Moran, V., Campbell, R.,
Hayashizaki, Y., and Bonthron, D. T. (1998) The human GNAS1 gene
is imprinted and encodes distinct paternally and biallelically expressed
G proteins. Proc. Natl. Acad. Sci. U. S. A. 95, 10038 –10043
34. Sullivan, K. E., Reddy, A. B., Dietzmann, K., Suriano, A. R., Kocieda,
V. P., Stewart, M., and Bhatia, M. (2007) Epigenetic regulation of
tumor necrosis factor alpha. Mol. Cell. Biol. 27, 5147–5160
35. 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
36. Arnaud, P., Monk, D., Hitchins, M., Gordon, E., Dean, W.,
Beechey, C. V., Peters, J., Craigen, W., Preece, M., Stanier, P.,
Moore, G. E., and Kelsey, G. (2003) Conserved methylation
imprints in the human and mouse GRB10 genes with divergent
allelic expression suggests differential reading of the same
mark. Hum. Mol. Genet. 12, 1005–1019
37. Melzner, I., Scott, V., Dorsch, K., Fischer, P., Wabitsch, M.,
Bruderlein, S., Hasel, C., and Moller, P. (2002) Leptin gene
expression in human preadipocytes is switched on by matura-
tion-induced demethylation of distinct CpGs in its proximal
promoter. J. Biol. Chem. 277, 45420– 45427
38. McGill, B. E., Bundle, S. F., Yaylaoglu, M. B., Carson, J. P.,
Thaller, C., and Zoghbi, H. Y. (2006) Enhanced anxiety and
stress-induced corticosterone release are associated with in-
creased Crh expression in a mouse model of Rett syndrome.
Proc. Natl. Acad. Sci. U. S. A. 103, 18267–18272
39. Probst, M. C., Thumann, H., Aslanidis, C., Langmann, T.,
Buechler, C., Patsch, W., Baralle, F. E., Dallinga-Thie, G. M.,
Geisel, J., Keller, C., Menys, V. C., and Schmitz, G. (2004)
Screening for functional sequence variations and mutations in
ABCA1. Atherosclerosis 175, 269–279
40. Adkins, R. M., Fain, J. N., Krushkal, J., Klauser, C. K., Magann,
E. F., and Morrison, J. C. (2007) Association between paternally
inherited haplotypes upstream of the insulin gene and umbilical
cord IGF-II levels. Pediatr. Res. 62, 451–455
41. Mitsuya, K., Meguro, M., Lee, M. P., Katoh, M., Schulz, T. C.,
Kugoh, H., Yoshida, M. A., Niikawa, N., Feinberg, A. P., and
Oshimura, M. (1999) LIT1, an imprinted antisense RNA in the
human KvLQT1 locus identified by screening for differentially
expressed transcripts using monochromosomal hybrids. Hum.
Mol. Genet. 8, 1209–1217
42. Rosa, A. L., Wu, Y. Q., Kwabi-Addo, B., Coveler, K. J., Reid, S. V.,
and Shaffer, L. G. (2005) Allele-specific methylation of a
functional CTCF binding site upstream of MEG3 in the human
imprinted domain of 14q32. Chromosome Res. 13, 809– 818
43. Medstrand, P., Landry, J. R., and Mager, D. L. (2001) Long
terminal repeats are used as alternative promoters for the
endothelin B receptor and apolipoprotein C-I genes in humans.
J. Biol. Chem. 276, 1896–1903
44. Hayward, B. E., and Bonthron, D. T. (2000) An imprinted
antisense transcript at the human GNAS1 locus. Hum. Mol.
Genet. 9, 835–841
45. Li, L. C., and Dahiya, R. (2002) MethPrimer: designing primers
for methylation PCRs. Bioinformatics 18, 1427–1431
46. Kent, W. J., Sugnet, C. W., Furey, T. S., Roskin, K. M., Pringle,
T. H., Zahler, A. M., and Haussler, D. (2002) The human
genome browser at UCSC. Genome Res. 12, 996–1006
47. Ehrich, M., Nelson, M. R., Stanssens, P., Zabeau, M., Liloglou,
T., Xinarianos, G., Cantor, C. R., Field, J. K., and van den Boom,
D. (2005) Quantitative high-throughput analysis of DNA meth-
ylation patterns by base-specific cleavage and mass spectrometry.
Proc. Natl. Acad. Sci. U. S. A. 102, 15785–15790
48. Coolen, M. W., Statham, A. L., Gardiner-Garden, M., and Clark, S. J.
(2007) Genomic profiling of CpG methylation and allelic specificity
using quantitative high-throughput mass spectrometry: critical evalu-
ation and improvements [Online]. Nucleic Acids Res. 35, e119
49. Ehrich, M., Turner, J., Gibbs, P., Lipton, L., Giovanneti, M., Cantor, C.,
and van den Boom, D. (2008) Cytosine methylation profiling of
cancer cell lines. Proc. Natl. Acad. Sci. U. S. A. 105, 4844 –4849
50. Thompson, R. F., Suzuki, M., Lau, K. W., and Greally, J. M.
(2009) A pipeline for the quantitative analysis of CG dinucle-
otide methylation using mass spectrometry. Bioinformatics 25,
2164–2170
51. Stanssens, P., Zabeau, M., Meersseman, G., Remes, G., Ganse-
mans, Y., Storm, N., Hartmer, R., Honisch, C., Rodi, C. P.,
Bocker, S., and van den Boom, D. (2004) High-throughput
MALDI-TOF discovery of genomic sequence polymorphisms.
Genome Res. 14, 126–133
52. Altman, D. G. (1991) Practical Statistics for Medical Research,
Chapman and Hall, London
53. West, B., Welch, K., and Galecki, A. (2006) Linear Mixed Models:
A Practical Guide Using Statistical Software, Chapman Hall/CRC
Press, Boca Raton, FL, USA
54. El-Maarri, O., Becker, T., Junen, J., Manzoor, S. S., az-Lacava, A.,
Schwaab, R., Wienker, T., and Oldenburg, J. (2007) Gender-
specific differences in levels of DNA methylation at selected loci
from human total blood: a tendency toward higher methylation
levels in males. Hum. Genet. 122, 505–514
55. Herman, J. G., Graff, J. R., Myohanen, S., Nelkin, B. D., and
Baylin, S. B. (1996) Methylation-specific PCR: a novel PCR assay
for methylation status of CpG islands. Proc. Natl. Acad. Sci.
U. S. A. 93, 9821–9826
56. Bock, C., Walter, J., Paulsen, M., and Lengauer, T. (2008) Inter-
individual variation of DNA methylation and its implications for
large-scale epigenome mapping [Online]. Nucleic Acids Res. 36, e55
57. The International HapMap Consortium (2003) The Interna-
tional HapMap Project. Nature 426, 789–796
58. 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
59. Ooi, S. K., Qiu, C., Bernstein, E., Li, K., Jia, D., Yang, Z.,
Erdjument-Bromage, H., Tempst, P., Lin, S. P., Allis, C. D.,
Cheng, X., and Bestor, T. H. (2007) DNMT3L connects un-
methylated lysine 4 of histone H3 to de novo methylation of
DNA. Nature 448, 714–717
60. Ramchandani, S., Bhattacharya, S. K., Cervoni, N., and Szyf, M.
(1999) DNA methylation is a reversible biological signal. Proc.
Natl. Acad. Sci. U. S. A. 96, 6107–6112
61. Bollati, V., Schwartz, J., Wright, R., Litonjua, A., Tarantini, L.,
Suh, H., Sparrow, D., Vokonas, P., and Baccarelli, A. (2009)
Decline in genomic DNA methylation through aging in a cohort
of elderly subjects. Mech. Ageing Dev. 130, 234–239
62. Heijmans, B. T., Tobi, E. W., Lumey, L. H., and Slagboom, P. E.
(2009) The epigenome: Archive of the prenatal environment.
Epigenetics 4526–531
Received for publication November 24, 2009.
Accepted for publication March 18, 2010.
3144 Vol. 24 September 2010 TALENS ET AL.The FASEB Journal 䡠www.fasebj.org