Content uploaded by Eugenia Migliavacca
Author content
All content in this area was uploaded by Eugenia Migliavacca
Content may be subject to copyright.
Research
DNA methylation profiles of human active
and inactive X chromosomes
Andrew J. Sharp,
1,4,5
Elisavet Stathaki,
1
Eugenia Migliavacca,
1,2
Manisha Brahmachary,
3
Stephen B. Montgomery,
1
Yann Dupre,
1
and Stylianos E. Antonarakis
1
1
Department of Genetic Medicine and Development, University of Geneva, 1211 Geneva 4, Switzerland;
2
Swiss Institute
of Bioinformatics, University of Lausanne, 1015 Lausanne, Switzerland;
3
Department of Genetics and Genomic Sciences,
Mount Sinai School of Medicine, New York, New York 10029, USA
X-chromosome inactivation (XCI) is a dosage compensation mechanism that silences the majority of genes on one X
chromosome in each female cell. To characterize epigenetic changes that accompany this process, we measured DNA
methylation levels in 45,X patients carrying a single active X chromosome (X
a
), and in normal females, who carry one X
a
and one inactive X (X
i
). Methylated DNA was immunoprecipitated and hybridized to high-density oligonucleotide arrays
covering the X chromosome, generating epigenetic profiles of active and inactive X chromosomes. We observed that XCI is
accompanied by changes in DNA methylation specifically at CpG islands (CGIs). While the majority of CGIs show in-
creased methylation levels on the X
i
, XCI actually results in significant reductions in methylation at 7% of CGIs. Both intra-
and inter-genic CGIs undergo epigenetic modification, with the biggest increase in methylation occurring at the promoters
of genes silenced by XCI. In contrast, genes escaping XCI generally have low levels of promoter methylation, while genes
that show inter-individual variation in silencing show intermediate increases in methylation. Thus, promoter methylation
and susceptibility to XCI are correlated. We also observed a global correlation between CGI methylation and the evo-
lutionary age of X-chromosome strata, and that genes escaping XCI show increased methylation within gene bodies. We
used our epigenetic map to predict 26 novel genes escaping XCI, and searched for parent-of-origin-specific methylation
differences, but found no evidence to support imprinting on the human X chromosome. Our study provides a detailed
analysis of the epigenetic profile of active and inactive X chromosomes.
[Supplemental material is available for this article.]
X-chromosome inactivation (XCI) is a mechanism of dosage com-
pensation that equalizes the expression of sex-linked genes be-
tween 46,XY males and 46,XX females (Lyon 1961). This results in
the silencing of the majority of genes on one of the two X chro-
mosomes in each somatic cell of females (Carrel and Willard 2005)
and a transition to a heterochromatic state. XCI is accompanied by
several epigenetic modifications to the inactive X (X
i
), such as an
accumulation of variant histones and transcripts from the XIST
gene (Clemson et al. 1996; Costanzi and Pehrson 1998), a delay in
replication timing (Willard and Latt 1976), and relocalization to
the nuclear periphery (Barr and Bertram 1949).
Several studies have shown that DNA methylation of the X
i
plays an important role in the maintenance of its inactive state.
While the active X (X
a
) and X
i
have very similar global levels of
methylation (Bernadino et al. 1996), studies have shown that CpG
islands (CGIs) have a tendency to be methylated on the X
i
and
unmethylated on the X
a
(Tribioli et al. 1992; Hellman and Chess
2007). In contrast, the CGIs of genes escaping XCI often remain
unmethylated on both the X
i
and X
a
(Weber et al. 2007). Mouse
knockouts for the DNA methyltransferase enzyme Dnmt1 show
defects in X inactivation (Panning and Jaenisch 1996). Further-
more, treatment of cells in vitro with the demethylating agent
5-azacytidine has been shown to cause the X
i
to decondense (Haaf
1995), replicate earlier in the cell cycle ( Jablonka et al. 1985), and in
several studies using somatic cell hybrids has led to the reactivation
of previously silenced genes (Hansen et al. 1996). However, despite
its clear importance in the XCI process, to date only a single study
has attempted to perform chromosome-wide studies of the distri-
bution of DNA methylation on the X chromosome (Yasukochi et al.
2010).
DNA methylation also plays an important role in imprint-
ing, a phenomenon in which the expression of a gene is de-
pendent on its parent of origin. Because of thei r role in a variety of
human phenotypes, there is considerable interest in identifying
imprinted loci, and to date, about 70 human imprinted tran-
scripts are known (http://www.geneimprint.com/). Although no
imprinted genes have been identified on the human X chromo-
some, differences in both cognitive function and the frequency
of autism between Turner syndrome patients with a maternally
versus paternally derived X suggest the presence of X-linked
imprinted gene(s) (Skuse et al. 1997), and in mouse a cluster of
X-linked imprinted genes has been identified, although these
apparently lack human homologs (Davies et al. 2005; Raefski and
O’Neill 2005). Furthermore, in several non-primate mammals,
the entire XCI process is subject to imprinting, with the paternal
X chromosome being preferentially silenced in extra-embryonic
tissues (Takagi and Sasaki 1975).
To characterize the epigenetic state of X
a
and X
i
chromo-
somes, we have used methylated DNA immunoprecipitation
(MeDIP) followed by hybridization to high-density oligonucleo-
tide arrays covering the entire X chromosome to profile DNA
methylation patterns in both 46,XX females and Turner syndrome
patients with a 45,X karyotype. As 45,X individuals carry a single
4
Present address: Department of Genetics and Genomic Sciences,
Mount Sinai School of Medicine, New York, NY 10029, USA.
5
Corresponding author.
E-mail andrew.sharp@mssm.edu.
Article published online before print. Article, supplemental material, and pub-
lication date are at http://www.genome.org/cgi/doi/10.1101/gr.112680.110.
1592 Genome Research
www.genome.org
21:1592–1600 Ó2011 by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/11; www.genome.org
X
a
, while 46,XX individuals carry one X
a
and one X
i
, the difference
in methylation between these two groups allows measurement of
methylation changes that occur with XCI. As our study included
multiple Turner syndrome patients with a single X chromosome of
either maternal (X
MAT
) or paternal (X
PAT
) origin, we also used our
epigenetic map to search for parent-of-origin-specific methylation
differences on the X chromosome that might indicate imprinting.
Our study provides an epigenetic profile of X inactivation giving
novel insights into the phenomenon of dosage compensation.
Results
Characterization of methylation differences between active
and inactive X chromosomes
As 45,X Turner syndrome individuals carry a single X
a
, while
46,XX females carry one X
a
and one X
i
, the difference in methyl-
ation between these two groups provides a measure of methylation
specifically on the X
i
. Initial visual comparison of methylation
profiles revealed frequent large differences in methylation levels
between 45,X and 46,XX individuals overlapping gene promoter
regions and CGIs (Fig. 1; Supplemental Table 1). To gain a global
view of methylation changes on the X chromosome with XCI, we
generated composite plots of mean methylation levels within 271
RefSeq genes that are either subject to or escape XCI (Fig. 2), and
3060 CGIs (Fig. 3; Bock et al. 2007). We observed that for genes
subject to XCI, a general increase in methylation on the X
i
occurs
at gene promoters. In comparison, genes escaping XCI show
higher levels of methylation within gene bodies on both the X
a
and X
i
, but reduced promoter methylation only on the X
i
. Addi-
tionally, for genes escaping XCI, a slight but consistent increase in
gene body methylation is evident on the X
i
. For CGIs, we observed
that on average these show significantly higher methylation on
the X
i
, consistent with previous studies (Supplemental Table 2;
Tribioli et al. 1992; Weber et al. 2007; Yasukochi et al. 2010).
Next, we examined methylation levels at individual CGIs in
relation to their physical position on the X
i
. We observed wide
variation in the change in CGI methylation with XCI. Although
the majority of CGIs show increased methylation on the X
i
(68%
have higher methylation in 46,XX vs. 45,X cases, p<0.01), we
observed that 7% of CGIs showed significantly lower levels of
methylation on the X
i
(n=159 had lower methylation in 46,XX
compared to 45,X cases, p<0.01), indicating that XCI results in
reduced methylation at many loci (Supplemental Table1). Many of
these sites of reduced methylation cluster together, and their lo-
cation is strongly correlated with the position of genes that escape
XCI (Fig. 4; Supplemental Figs. 2, 3). Intersecting CGIs with 410
RefSeq genes of known X-inactivation status (Carrel and Willard
2005) yields correlation r=0.81 between mean difference in CGI
methylation between 45,X and 46,XX cases and XCI score,
showing a strong trend that CGIs overlapping genes that escape
XCI have lower methylation than the CGIs overlapping genes that
are subject to XCI.
As previous epigenetic studies of XCI have reported only in-
creased methylation on the X
i
, we sought to confirm this obser-
vation using an independent data set. We analyzed published data
of methylation levels at individual CpG dinucleotides on the X
chromosome measured in 600 brain samples produced by hy-
bridization of bisulfite-converted DNA to Illumina bead arrays
(Gibbs et al. 2010). Using this alternative technology, we observed
a similar distribution of changes in methylation with XCI as with
MeDIP, with 35% of the 1084 probes targeted to X-linked gene
promoters showing lower methylation levels in females than males.
Further analysis of these data indicated that the vast majority of
gender differences in methylation result from probes mapping to
the X chromosome, indicating XCI as the underlying mecha-
nism, and that as observed by MeDIP, XCI is accompanied by
both gains and losses of methylation at different sites (Supple-
mental Figs. 1, 2). We also validated our results using PCR and
sequencing of bisulfite-converted DNA in other individuals with
Figure 1. Comparative methylation profiling of the X chromosome in 45,X and 46,XX individuals shows that X inactivation results in increased
methylation of CpG islands at gene promoters. Regions containing high densities of CpG dinucleotides, corresponding with the transcription start sites
of genes subject to X inactivation, show raised methylation levels on the X
i
. The image shows the screenshot of a 250-kb region of Xp11.3
(chrX:46,475,000–46,725,000, hg18), with methylation data uploaded as custom tracks in the UCSC Genome Browser.
Genome Research 1593
www.genome.org
X-chromosome methylation
45,X and 46,XX karyotypes that had not been tested by array. At
sites indicated by MeDIP array analysis to have lower methylation
on the inactive X, we observed that while 45,X individuals showed
complete methylation, 46,XX females showed a mix of both
methylated and unmethylated DNA, indicating either methyla-
tion specific to the X
a
or a mix of methylation on both the X
a
and
X
i
(Supplemental Fig. 4).
BasedonthecorrelationbetweenCGImethylationandthe
location of genes escaping XCI, we investigated the relationship
between the X-inactivation status of a gene and its promoter
methylation more closely. We analyzed CGIs located within 1 kb of
the transcription start site(TSS) of genes consistentlyscored as either
subject to or escaping XCI (score =0 or 9) (Fig. 5; Supplemental
Tables 1, 3; Carrel and Willard 2005). CGIs at promoters of ge nes
subject to XCI show statistically significant increases in methyla-
tion on the X
i
compared to the X
a
.In
contrast, the majority of CGIs at promoters
of genes escaping XCI show methylation
levels on the X
i
similar to those on the X
a
.
We also investigated genes that
exhibit variable or unstable XCI, being
expressed from the X
i
in some individuals
but silent in others. We observed an in-
verse correlation between the frequency
with which a gene is expressed from the
X
i
and the methylation level of CGIs
located near the TSS on the X
i
(R
2
=0.18,
p<0.001) (Fig. 6). As we had measured
methylation in multiple individuals, we
investigated whether genes that show
polymorphic XCI also undergo poly-
morphic methylation, but could find no
evidence to support this hypothesis.
Genes that exhibit variable XCI (scores
1–8) (Carrel and Willard 2005) did not
show higher inter-individual variance in
CGI methylation at their promoters than
genes that are always active or inactive
(score 0 or 9) (Supplemental Table 1).
Finally, we observed a significant
trend between CGI methylation and
chromosomal position specifically on the
X
i
(linear regression, R
2
=0.102, p=7.58 310
16
in 46,XX in-
dividuals; R
2
=0.001, p=0.2 in 45,X individuals) (Supplemental
Fig. 6). In general, CGIs on the X
i
show smaller increases in
methylation levels in the distal short arm compared to long arm,
mirroring the known evolutionary ages of different portions of
the human X chromosome (Lahn and Page 1999).
Characterization of sequence features in relation
to methylation
We investigated the relationship between CpG density and
methylation changes on the X
i
and observed a positive correlation
between CpG density and the difference in log
2
methylation ratios
between 46,XX and 45,X individuals (Supplemental Fig. 7). Given
this potentially confounding relationship, we investigated whether
Figure 2. Methylation patterns of genes subject to and escaping X inactivation on active and inactive
X chromosomes. (A) For genes subject to XCI, methylation differences between 46,XX and 45,X in-
dividuals occur specifically at gene promoters, with 46,XX individuals showing increased methylation
compared to 45,X cases. As 83% of RefSeq transcripts listed on the X chromosome have a CGI within 1
kb of their TSS and CGIs are heavily methylated on the X
i
versus the X
a
, this enrichment of methylation at
gene promoters on the X
i
is a correlate of CGI methylation. (B) In contrast, for genes escaping XCI, there
is a generalized increase in methylation in the bodies of genes in both 45,X and 46,XX individuals
compared to genes subject to XCI, with the biggest difference between 46,XX and 45,X occurring at the
promoter region. Additionally, a slight but consistent increase in methylation in the bodies of genes
escaping XCI is evident in 46,XX compared to 45,X individuals. Two hundred sixteen RefSeq genes listed
on chrX scored as subject to XCI (score =0) and 52 RefSeq genes scored as escaping XCI (score =8or9)
by Carrel and Willard (2005) were analyzed. Where multiple splice forms were listed for a single gene,
we chose the maximal boundary of all isoforms to define the transcribed region. Probes were then
assigned into one of 40 bins depending on their relative position along the gene body, defined by the
maximal transcription start and end coordinates. Similar data were observed when considering only
those genes of intermediate length (10–100 kb) (Supplemental Figs. 11, 12).
Figure 3. Differential methylation between active and inactive X chromosomes occurs specifically at CpG islands. Each panel shows a composite plot of
mean methylation levels within and flanking 3060 CGIs on the X chro mosome in (A) 46,XX females; (B) 45,X Turner syndrome cases, corresponding to the
X
a
;and(C) the difference between these two groups, corresponding to the X
i
. Three thousand sixty CGIs defined using epigenetic criteria (Bock et al.
2007) were analyzed. Those separated by <500 bp (n=726) were merged into single regions, and the mean methylation level both within CGIs, and +5kb
and 5 kb was calculated. The apparent increase in methylation in flanking regions in 46,XX individuals is likely due to the clustering of CGIs on the X
chromosome, which results in neighboring CGIs being sampled in the 5-kb flanking regions (10% of CGIs are separated by <1 kb, 36% by <5 kb). (Gray
dashes) Borders of the CGIs. Mean methylation within CGIs was calculated by assigning probes into 10 windows proportional to the length of each CGI.
Mean methylation in flanking regions was plotted in 250-bp windows. Underlying mean log
2
values with variance and associated P-values are shown in
Supplemental Table 2.
Sharp et al.
1594 Genome Research
www.genome.org
this phenomenon might account for some of our prior observa-
tions. However, we observed no apparent relationship between the
density of CpG dinucleotides within CpG islands at gene pro-
moters or genes and their X-inactivation status, or between CpG
density within CpG islands and physical position on the X chro-
mosome (Supplemental Figs. 7, 8, 9). We therefore conclude that
variations in promoter CpG density do not explain susceptibility
of gene to XCI, and that the correlation we observed between
methylation and evolutionary age of X chromatin does not result
from variations in the CpG density of CpG islands. Investigation of
repeat content of genes subject to and escaping XCI confirmed
previous observations of reduced LINE content in genes escaping
XCI relative to genes that are subject to XCI (Supplemental Fig. 11;
Wang et al. 2006).
We hypothesized that regions susceptible to XCI might be
enriched for sequence features associated with X inactivation, and
we could therefore use our methylation map to search for motifs
that might play a role in the XCI process. Although MEME analysis
identified three high-frequency motifs
10–20 bp in size that showed potential
enrichment in a set of 115 CGIs that
showed the largest increase in methyla-
tion on the X
i
(data not shown), these
motifs were highly degenerate and of low
complexity in nature. Subsequent analy-
sis by FIMO suggested that these motifs
were also enriched in a set of 115 CGIs
from the opposite end of the distribution
(those showing the largest decrease in
methylation on the X
i
) and in a set of 115
randomly chosen CGIs from the X chro-
mosome, suggesting that any association
of these motifs with XCI methylation was
non-specific. Thus, we do not consider
there to be any specific association of these
motifs with XCI. DRIM analysis did not
identify any motifs that were signifi-
cantly correlated with CGI methylation.
Prediction of novel genes escaping
X inactivation based on promoter
methylation and validation
by RNA-seq
We investigated the utility of our meth-
ylation map of the X
i
and X
a
to classify
genes based on their X-inactivation sta-
tus. For genes known to escape XCI (score
8 or 9), the mean difference in TSS CGI
methylation between 45,X and 46,XX
individuals is generally much lower than
at genes subject to XCI, with 50% of
genes escaping XCI showing a difference
in log
2
<0.39. Using this as a metric to
score the genes tested by Carrel and
Willard (2005), 14 of the 29 (48%) genes
escaping XCI are below this threshold,
while only three of 135 (2%) genes sub-
ject to XCI satisfy this criterion. Further-
more, in 46,XX individuals, the absolute
methylation at the CGI of genes consis-
tently subject to XCI (score =0) is gener-
ally high, with the bottom fifth percentile of this distribution be-
ing log
2
=0.65. Using this as a metric to classify genes previously
tested (Carrel and Willard 2005), 20 of 31 (65%) genes escaping
XCI have methylation levels below this threshold, while only
seven of 135 (5%) genes subject to XCI satisfy this criterion. Using
a stringent combination of both criteria together (absolute meth-
ylation in 46,XX <0.65, difference in methylation between 46,XX
and 45,X <0.44) identifies 12 of the 29 (41%) genes previously
classed as escaping XCI, and only one of the 135 (0.7%) RefSeq
genes scored as inactive. Thus, both the absolute and relative
methylation levels of gene promoters in 45,X and 46,XX individuals
strongly correlates with their expression from the X
i
, and promoter
CGI methylation is predictive of XCI status.
We hypothesized that we could therefore use our methylation
map of the X chromosome to prospectively identify the majority of
genes that escape XCI on the entire X chromosome based on their
promoter methylation state. Classifying all known genes on the
X chromosome by promoter methylation, as described above,
Figure 4. X inactivation results in highly variable changes in methylation of CpG islands that correlate
with the location of genes escaping X inactivation. While the majority of CGIs (61%) show increased
methylation on the X
i
(yellow circles), 7% of CGIs have significantly lower levels of methylation in 46,XX
compared to 45,X individuals (blue squares; p<0.01), contradicting the notion that XCI is always
associated with increased methylation. The location of these sites of reduced methylation highly cor-
relates with the physical position of genes known to escape XCI (r=0.81). (Gray diamonds) The
remaining 32% of CGIs showed no significant difference in methylation (p>0.01). Each point repre-
sents the change in mean methylation between 45,X and 46,XX individuals at a CGI (Bock et al. 2007),
with the black line showing the moving average of methylation at 10 CGIs. The heat map shows genes
expressed from the X
i
(blue), silent genes (yellow), pseudoautosomal genes (purple), and untested
genes (white) in each of nine cell lines (adapted from Carrel and Willard 2005 and reprinted with
permission from Nature Publishing Group Ó2005).
X-chromosome methylation
Genome Research 1595
www.genome.org
allowed us to predict 31 novel genes (21 RefSeq genes and 10 ad-
ditional transcripts listed as UCSC genes) that likely escape XCI
with high confidence based on two criteria: (1) Their mean pro-
moter CGI methylation in 46,XX individuals is <0.65; and (2) the
difference in mean promoter methylation between 46,XX and
45,X is <0.39 (Supplemental Table 3).
We validated a subset of these predictions using transcribed
SNPs in RNA-seq data from 46,XX lymphoblastoid cell lines that
showed highly skewed X-chromosome inactivation. In a female in
which every cell has silenced the same parental X chromosome,
escape from XCI would manifest as biallelic expression of that
gene, while monoallelic expression would indicate that the gene
was subject to XCI. Of the 31 novel genes we predicted to escape
XCI, eight contained a heterozygous SNP covered at sufficient read
depth to be assayed in females with skewed XCI. Six of these eight
genes showed biallelic expression in all individuals tested, thus
confirming their predicted status as escaping XCI (DDX3X,
EIF1AX,EIF2S3,TXLNG,RPS4X,HDHD1) (Supplemental Table
4). One further gene (UBA1) was biallelically expressed in one
individual, but monoallelically expressed in a second in-
dividual, suggesting that it undergoes polymorphic XCI. Only
one of the eight genes tested (MAGED1) showed monoallelic
expression in all cases analyzed, suggesting that this gene is
subject to XCI and that our prediction of escape from XCI was
incorrect. Although we were only able to test a minority of the
31 predicted genes that escape XCI, largely due to the fact that
many genes had low or absent expression in lymphoblasts
combined with the limited availability of transcribed poly-
morphisms, these analyses demonstrate that our predictions of
genes escaping XCI are largely correct.
Imprinting analysis of the X chromosome using patients
with Turner syndrome
Previous phenotype studies have suggested the presence of one
or more imprinted loci on the human X chromosome that in-
fluence cognitive function (Skuse et al. 1997). We reasoned that
as most imprinted genes are accompanied by parent-of-origin-
specific epigenetic marks, a search for differentially methylated
regions between patients with 45,X
MAT
(n=4) versus those with 45,X
PAT
(n=3)
might uncover imprinted loci on the X
chromosome. Previously, we used an
identical MeDIP and array hybridization
protocol in patients with uniparental
disomy of chromosome 15 to successfully
detect novel DMRs associated with im-
printed loci on chromosome 15 (Sharp
et al. 2010). After applying the same
analysis to compare methylation patterns
in 45,X
MAT
versus 45,X
PAT
samples, none
of the >1 million probes covering the X
chromosome showed a difference that
was even suggestive of parent-of-origin-
specific X-chromosome methylation (all
probes had FDR-adjusted p>0.9999).
Despite this, we further analyzed our data
set in an attempt to identify weak signals
of differential methylation that did not
survive multiple testing correction. We
searched for clusters of probes with
nominal significance and identified 11 X-
chromosome loci comprising three or more probes separated by
<500 bp, each with an unadjusted P-value <0.01 (Supplemental
Table 5). However, of these 11 loci, 10 showed characteristics that
made them highly unlikely to be genuine DMRs and were excluded
from further analysis: Three had a very low CpG density (<1 CpG
per 250 bp), three were composed solely of common repetitive
elements embedded in duplicated portions of the genome, and
four showed differences in methylation intensity, but not di-
rection, that resembled false-positive signals we had observed in
previous studies of chromosome 15 (Sharp et al. 2010). We
designed bisulfite PCR primers to amplify the one remaining locus
and performed bisulfite sequencing in two 45,X
MAT
, two 45,X
PAT
samples, and two normal controls, but observed no methylation
differences between maternally and paternally derived X chro-
mosomes at this locus. Thus, we could find no evidence of signif-
icant parent-of-origin-specific methylation on the X chromosome.
Discussion
We have performed a detailed analysis of DNA methylation pat-
terns on the active and inactive X chromosomes. Consistent with
previous studies, we observe that XCI is accompanied by gains in
methylation at the majority of CGIs of genes silenced on the X
i
.
However, our data show that methylation increases are not limited
to gene promoter regions, and we demonstrate that the majority of
intergenic CGIs also show increased methylation on the X
i
, al-
though to a lesser extent than that seen at gene promoters. Our
study also shows that, in contrast to current thinking, XCI actually
results in reduced methylation at a significant proportion of CGIs.
These CGIs that show higher methylation on the X
a
tend to occur
outside of gene promoters. We also observed a relative increase in
methylation within the transcribed region of genes escaping XCI
compared to those that are subject to XCI, consistent with pre-
vious reports of both decreased methylation in the bodies of epi-
genetically silent genes (Hellman and Chess 2007) and increased
methylation in expressed genes versus inactive genes (Li et al.
2010). However, as both 45,X and 46,XX individuals have an X
a
and the MeDIP assay does not provide allelic methylation in-
formation, we are unable to distinguish if there is increased meth-
Figure 5. Methylation status of CpG islands at gene promoters varies depending on gene inactivation
status. (A) CGIs at promoters of genes subject to XCI show much higher methylation levels in 46,XX
compared to 45,X individuals (mean 46,XX log
2
methylation =0.99, mean log
2
methylation differ-
ence =1.04, p=1.3 310
23
). (B) In contrast, the majority of CGIs at promoters of genes escaping XCI
show much lower methylation levels in 46,XX individuals, similar to those seen in 45,X individuals
(mean 46,XX log
2
methylation =0.54, mean log
2
methylation difference =0.34, p=0.00013). Each
point represents the mean log
2
methylation of a CGI located within 1 kb of the TSS of a RefSeq gene
scored by Carrel and Willard (2005) as being (A) silenced on the X
i
(expressed in 0 of 9 hybrids con-
taining an inactive X, n=135), or (B) expressed from the X
i
(expressed in $8 of 9 hybrids containing an
inactive X, n=29). Colored lines show the moving average of methylation at 10 and 3 CGIs for genes
subject to and escaping XCI, respectively. Data for CGIs that lie outside of RefSeq gene promoters are
shown in Supplemental Figure 14.
Sharp et al.
1596 Genome Research
www.genome.org
ylation specific to the active X. An alternative possibility that we
cannot exclude is that the reduced methylation observed at some
sites in 46,XX versus 45,X individuals represents a mosaic mix of
methylation on both the X
a
and X
i
. However, in favor of the for-
mer hypothesis, allele-specific studies in a 46,XX female with
completely skewed XCI showed monoallelic, rather than biallelic,
methylation (Supplemental Fig. 5).
As we observed that CGI methylation directly correlates with
the probability of a gene being subject to XCI, we used our
methylation map to predict genes that escape XCI based on their
absolute and relative promoter methylation on the X
i
and X
a
.By
first calibrating suitable thresholds based on the inactivation status
of 362 RefSeq genes that have a CGI within 1 kb of their TSS
studied by Carrel and Willard (2005), we were able to identify
a high-confidence set of 31 additional genes that likely also escape
XCI. Multiple lines of evidence show that these predictions based
on epigenetic profiling are largely accurate. Using RNA-seq data
from 46,XX cell lines that showed completely skewed XCI, we were
able to validate seven out of eight of these predicted genes as being
expressed biallelically from both t he X
i
and X
a
, confirming their
escape from XCI. In addition, we note that this set of 31 predicted
genes that escape XCI includes CD99,KDM5C,SMC1A,TBL1X,
and KDM6A, all of which have already been shown to escape XCI
by other studies (Dracopoli et al. 1985;
Wu et al. 1994; Brown et al. 1995; de
Conciliis et al. 1998; Greenfield et al.
1998). Of the other 26 genes, NLGN4X,
PCDH11X,TBL1X,andTXLNG have func-
tional Y-homologs, STS and KAL1 have
pseudogene Y-homologs, and PLCXD1 is
located in the Xp pseudo-autosomal re-
gion, all of which are features shared
by many other genes that are known to
escape XCI. In addition, 14 of these 31
genes were highlighted as potentially es-
caping XCI based on their consistently
higher expression in females versus males
( Johnston et al. 2008), and two were also
previously identified as likely escaping
XCI due to a lack of promoter methyla-
tion in females (Yasukochi et al. 2010).
We conclude that the use of comparative
methylation profiling provides a power-
ful method for the classification of genes
escaping XCI that is complementary to
methods based on expression profiling.
We also tested the hypothesis that
variable XCI results from polymorphic
promoter methylation but could find no
evidence to support this. In 46,XX fe-
males, there was no significant difference
in the inter-individual methylation vari-
ance at genes that were scored as showing
variable inactivation XCI compared to
those that are always active or always in-
active (Carrel and Willard 2005). Thus,
our results do not support the notion that
variable XCI results from inter-individual
epigenetic variation. Instead, our obser-
vations favor the alternative scenario that
variability in XCI is caused by lower pro-
moter methylation levels, likely resulting
in less stable maintenance of XCI at these genes (Fig. 5; Lock et al.
1987; Keohane et al. 1998; Lingenfelter et al. 1998).
Our study also reveals a correlation between CGI methylation
and chromosomal position on the X
i
that corresponds with the
evolutionary age of different X-chromosome strata (Supplemental
Fig. 6). Previous studies have shown that the human X chromo-
some was formed from an ancestral X, corresponding to Xq, with
multiple translocation events progressively adding additional au-
tosomal chromatin onto Xp (Lahn and Page 1999). We observed
a significant trend in which CGIs on the X
i
show smaller increases
in methylation levels in the younger evolutionary strata (distal Xp)
compared to the ancestral portions (Xq). These data suggest that
the chromatin regions recently added to the X chromosome are
less heavily influenced by the XCI signal than the ancestral por-
tions of the X, consistent with the higher density of genes escaping
XCI in these regions.
Multiple lines of evidence suggest that our methodology
provides accurate measurement of methylation on both the X
i
and
X
a
. The patterns of CGI methylation we observed at numerous
individual genes agree closely with those reported by previous
studies that have used different methodologies, such as methyla-
tion-sensitive restriction enzyme digestion (Tribioli et al. 1992;
Hellman and Chess 2007) or hybridization of bisulfite-converted
Figure 6. Difference in methylation levels between 46,XX and 45,X individuals at transcription start
sites is inversely correlated with X-inactivation status. Boxplots show the mean methylation of CGIs
located within 1 kb of the TSS of 363 RefSeq genes scored by Carrel and Willard (2005) for their ex-
pression status on the X
i
(Supplemental Table 3). Genes are divided based on their expression status on
the X
i
, with a score from 0 to 9 corresponding to the number of somatic cell hybrids containing an
inactive X chromosome that express that gene. Genes with score 0 are always subject to XCI, genes with
score 9 always escape XCI, and genes with intermediate scores show polymorphic or unstable in-
activation. The XIST gene is unique in being transcribed only from the X
i
, and shows lower methylation
in 46,XX versus 45,X individuals. Linear regression analysis of XCI score with methylation difference
between 46,XX and 45,X yields R
2
=0.18, p<0.001. ANOVA followed by Tukey analysis showed that
genes with scores 4–9 showed statistically significant methylation differences compared to genes with
score 0 (Bonferroni-adjusted p<0.001, indicated by an asterisk within each box). Boxplots define lower
and upper quartiles of each distribution, the internal band shows the median, while whiskers correspond
to top and bottom deciles. Categories containing less than 15 genes were combined to yield sufficient
sample sizes for meaningful comparison.
X-chromosome methylation
Genome Research 1597
www.genome.org
DNA to bead arrays (Gibbs et al. 2010). Furthermore, based solely
on measurements of CGI methylation, we are able to accurately
predict and validate genes that are either known to or have been
hypothesized to escape XCI. Our conclusions also largely mirror
those from a previous study of gender differences in X-chromo-
some methylation using sequencing of HpaII and MspI restriction
fragments (Yasukochi et al. 2010). As we observed, this study also
identified a high correlation between low promoter methylation
and escape from XCI, and that such genes tend to occur in evo-
lutionarily younger strata.
We chose to study Turner syndrome patients with a 45,X
karyotypeasthisprovidesacleaner system that allows more
accurate determination of methylation patterns on the active X
chromosome. This is because the large areas of homology between
the X and Y chromosomes would cause cross-hybridization arti-
facts on the microarray if we had studied 46,XY males, introducing
significant noise to our data. It should be noted that >90% of 45,X
concepti die in utero, and it is therefore possible that this might
indicate additional genome abnormalities in Turner syndrome
that could bias our results. However, as one of the common fea-
tures of imprinted genes is the presence of differentially methyl-
ated regions (DMRs), our study design also enabled us to compare
data from Turner syndrome patients with a maternally derived X
against those with a paternally derived X as a method to search for
imprinted loci on the X chromosome. While our previous use of
this same methodology to study patients with UPD15 has suc-
cessfully identified many novel DMRs on chromosome 15 (Sharp
et al. 2010), our analysis of maternal and paternal X chromosomes
failed to identify any sites of differential methylation, suggesting
a lack of imprinting on the human X chromosome. We note,
however, that the use of MeDIP and array hybridization has some
limitations. Firstly, its resolution is limited, being only able to de-
tect methylation changes at clusters of multiple CpGs. Thus,
MeDIP is likely to miss subtle differences comprising only a few
CpGs, or those occurring at high-copy sequences not represented
on our microarray design. More specifically for the study of im-
printing, some DMRs associated with imprinted genes are only
differentially methylated in specific tissues, while other imprinted
genes have been identified that apparently lack nearby DMRs.
Given these caveats, while we cannot exclude the presence of
cryptic DMRs or imprinting not detected by our methodology, our
analysis was unable to find any evidence of imprinted genes on the
human X chromosome. Finally, we also attempted to identify mo-
tifs that were enriched in sequences that underwent large changes
in methylation on the X
i
. However, although motifs were identified
in our initialanalysis, these sequences were highly redundantand of
low complexity in nature, and further testing of these motifs yielded
apparent enrichments in many different sequence sets, suggesting
that these statistical associations do not represent genuine enrich-
ments. On balance, we concludethat we were unable to identify any
motifs that could be unambiguously linked to the XCI process.
In summary, our study provides a comprehensive profile of
methylation patterns on human active and inactive X chromo-
somes. These data provide novel insights into the X-inactivation
process and demonstrate the utility of epigenetic profiling as
a method to study mechanisms of genome regulation.
Methods
Genomic DNA was extracted from peripheral blood samples from
(1) four unrelated Turner syndrome patients with a single mater-
nally derived X chromosome (45,X
MAT
), (2) three unrelated Turner
syndrome patients with a single paternally derived X chromosome
(45,X
PAT
), and (3) three unrelated control females with a normal
46,XX karyotype. The ages of each sample were unknown. For each
individual with Turner syndrome, cytogenetic analysis of 30
stimulated T-lymphocyte nuclei showed the presence of a 45,X
karyotype in all cells examined. Parental origin analysis of the X
chromosome was determined using a panel of microsatellite
markers mapping to the X chromosome in each Turner syndrome
patient and their parents. Neither cytogenetic nor molecular
analysis showed any signs of mosaicism in any of the Turner
syndrome cases analyzed.
Methylated DNA was immunoprecipitated using monoclonal
antibodies that recognize methylated cytosine (Weber et al. 2007).
Briefly, 15 mg of DNA was sonicated to generate fragments 200–800
bp in size (Branson 450D Sonifier), incubated with 10 mg of anti
5-methyl cytidine (Diagenode), immunoprecipitated using Protein
A Sepharose beads (Life Technologies), and purified by phenol:
chloroform extraction. Immunoprecipitated and input DNA
from each case were labeled by random priming using Cy5 and
Cy3-conjugated random nonamers (TriLink BioTechnologies), and
hybridized to tiling oligonucleotide arrays.
We used arrays composed of 2.1 million 50-mer to 75-mer
oligonucleotides covering chromosomes 20, 21, 22, X, and Y at
a median probe density of 1 per 100 bp (Roche NimbleGen, catalog
#05543142001). Of these, 1,148,358 probes map to the 152-Mb
sequenced portion of chromosome X. DNA labeling, array hy-
bridizations, and washes were performed according to the manu-
facturer’s recommendations, and slides were scanned using a
G2565 scanner at 5-mm resolution (Agilent Technologies). Array
images were analyzed using NimbleScan v2.5 software (Roche
NimbleGen) with default parameters incorporating spatial cor-
rection, and the resulting files of probe log
2
ratios were used for
subsequent analysis. The log
2
value is the Cy5:Cy3 fluorescence
ratio (methylated DNA recovered by IP:total input DNA) for each
probe, converted to a log
2
scale, and represents a relative measure
of the amount methylated DNA at each locus. We applied quantile
normalization (Bolstad et al. 2003) to the raw data and filtered
outlier probes to remove low-quality data points (Sharp et al.
2010). Technical replicates of the MeDIP and array hybridizations
were performed for four of the seven individuals in this study, and
correlation coefficients between these technical replicates were
high (mean r=0.93 after quantile normalization and outlier re-
placement). Subsequent data analyses were performed in Galaxy
(http://main.g2.bx.psu.edu/). For all plots of differences between
45,X and 46,XX cases, we calculated the mean log
2
methylation in
multiple individuals with the same karyotype. However, in order
to account for potential inter-individual variability in methylation
patterns or noise resulting from probe variability, for all statistical
analyses reported we used log
2
values for individual probes in each
individual. Data from each probe in each 45,X and 46,XX in-
dividual tested were then combined into two separate groups for
the purposes of statistical comparison.
We used data sets of CGIs identified by epigenetic criteria
(Bock et al. 2007) and downloaded additional information on
RefSeq genes and UCSC genes from build36/hg18 of the UCSC
Genome Browser (http://genome.ucsc.edu/). Data on the inacti-
vation status of X-linked genes were taken from a previous study
that scored expression for 624 loci in a panel of nine somatic cell
hybrids containing a single inactive X chromosome (Carrel and
Willard 2005). However, when mapped to hg18, we found that
a significant number of these loci did not match well with current
gene annotations. We therefore decided to be conservative in our
approach and included only those loci tested by Carrel and Willard
that we could map unambiguously to RefSeq annotations (n=
410). We conducted analysis of G+C nucleotide content, CpG
Sharp et al.
1598 Genome Research
www.genome.org
dinucleotide content, and common repeat content of genes and
CGIs to investigate sequence features that might influence our
methylation results.
We used a custom analysis pipeline to detect regions of dif-
ferential methylation between Turner syndrome individuals with
X chromosomes of maternal and paternal origin (Sharp et al.
2010). Since invariant probes are less likely to contribute to the
distinctions between subgroups, probes with a standard deviation
across all samples <0.2 were removed from further analyses, and
a moderated t-test (Smyth 2004) was conducted to identify probes
that showed significantly different methylation between 45,X
MAT
and 45,X
PAT
samples usinga 5% false discovery rate (FDR) correction
(Benjamini and Hochberg 1995). All statistical analyses were per-
formed using software from the Bioconductor project (Gentleman
et al. 2004).
To assess X-chromosome methylation patterns assayed using
a different technique, we analyzed a published data set of meth-
ylation at >27,000 CpG sites measured in 600 brain samples de-
rived from 150 individuals by hybridization of bisulfite-converted
DNA to Illumina bead arrays (Gibbs et al. 2010). Of these, 1084
probes mapped to the X chromosome. Samples were divided by
gender, and analysis performed both for individual tissues, and
by combining all four brain regions together, producing globally
very similar results in each case.
For validation of array results and putative DMRs, we de-
signed primers to amplify bisulfite-converted DNA using Methyl
Primer Express v1.0 (Life Technologies). Two micrograms of ge-
nomic DNA from 45,X and 46,XX individuals was converted and
purified using Epitect Bisulfite Kits (QIAGEN). Bisulfite-treated
DNA was amplified using JumpStart REDTaq DNA Polymerase
(Sigma-Aldrich), unincorporated primers and nucleotides were
removed by incubation with Exonuclease I and Shrimp Alkaline
Phosphatase (New England Biolabs), and the products were sub-
jected to Sanger sequencing.
For validation of genes predicted to escape XCI, we performed
analysis of published RNA-seq data from 30 female lymphoblastoid
cell lines from the CEU HapMap population (Montgomery et al.
2010). SNP genotypes from the 1000 Genomes Project Pilot 1 were
downloaded to annotate transcribed polymorphisms in X-linked
RefSeq genes in each individual. Two hundred seventy-one RefSeq
genes that had previously been shown to be consistently subject to
XCI (Carrel and Willard 2005) were initially used to identify which
of the 30 female cell lines showed completely skewed patterns of
XCI. Female samples were scored as having skewed XCI based on
the presence of monoallelic expression of at least three different
highly expressed genes per individual that were known to be
subject to XCI. Skewed XCI was defined by the observation in
mRNA of only one allele of a transcribed SNP that had $20 over-
lapping reads. To allow some tolerance for sequencing errors,
where $40 overlapping reads were present, $97.5% of reads were
required to be derived from the same allele. By this criterion, six of
the 30 females analyzed showed highly skewed patterns of XCI.
Although this clonality rate of 20% is quite high, it agrees well with
previous measurements in EBV-transformed lymphoblasts (Plagnol
et al. 2008). For these six females that showed almost exclusive
inactivation of one X chromosome, the allelic expression status of
the 31 genes that had been predicted to escape XCI was measured.
Biallelic expression of these genes would be indicative of escape
from XCI, while monoallelic expression would indicate that the
gene was subject to XCI. To enable testing of a useful number of
these 31 genes in the six individuals while still maintaining sta-
tistical rigor, we chose a lower threshold corresponding to $10
reads overlapping a transcribed heterozygous SNP. At this thresh-
old, eight of the 31 predicted genes could be tested. Under the null
hypothesis that reads are sampled at random from either allele,
this corresponds to p<0.001 that a biallelically expressed gene
would be incorrectly scored as showing monoallelic expression.
MEME (Bailey and Elkan1994) and DRIM (Eden et al. 2007)
were used to search for sequence motifs enriched in CGIs that
showed extreme gains or losses of methylation. For MEME analy-
sis, all CGIs on the X chromosome were ranked based on their
mean probe log
2
ratio, and the top and bottom 5% tails of this
distribution were tested for enriched motifs 5–20 bp in size. FIMO
(http://meme.sdsc.edu/meme/fimo-intro.html) was then used to
test the frequency and enrichment of the identified motifs in CGIs
on the X chromosome. For DRIM analysis, motifs 6–8 bp were
tested.
Data access
Array data have been deposited in the NCBI Gene Expression
Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo) under acces-
sion number GSE22551.
Acknowledgments
We thank Professor Patricia Jacobs for supplying DNA samples
from 45,X patients of known parental origin, and Dr. Mauro
Delorenzi for useful discussions. The research leading to these re-
sults has received funding from the Fondation Jerome LeJeune, the
European Commission Seventh Framework Program under grant
agreement 219250 to A.J.S., the European Commission Sixth
Framework Program under grant agreement LSH-2005-1.1.5-1
(anEUploidy), and a Swiss National Science Foundation grant to
S.E.A.
References
Bailey TL, Elkan C. 1994. Fitting a mixture model by expectation
maximization to discover motifs in biopolymers. Proceedings of the
Second International Conference on Intelligent Systems for Molecular Biology,
pp. 28–36. AAAI Press, Menlo Park, CA.
Barr ML, Bertram EG. 1949. A morphological distinction between neurones
of male and female, and the behaviour of the nucleolar satellite during
accelerated nucleoprotein synthesis. Nature 163: 676–677.
Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate:
a practical and powerful approach to multiple testing. J R Stat Soc Ser B
Methodol 57: 289–300.
Bernardino J, Lamoliatte E, Lombard M, Niveleau A, Malfoy B, Dutrillaux B,
Bourgeois CA. 1996. DNA methylation of the X chromosomes of the
human female: an in situ semi-quantitative analysis. Chromosoma 104:
528–535.
Bock C, Walter J, Paulsen M, Lengauer T. 2007. CpG island mapping by
epigenome prediction. PLoS Comput Biol 3: e110. doi: 10.1371/
journal.pcbi.0030110.
Bolstad BM, Irizarry RA, Astrand M, Speed TP. 2003. A comparison of
normalization methods for high density oligonucleotide array data
based on variance and bias. Bioinformatics 19: 185–193.
Brown CJ, Miller AP, Carrel L, Rupert JL, Davies KE, Willard HF. 1995. The
DXS423E gene in Xp11.21 escapes X chromosome inactivation. Hum
Mol Genet 4: 251–255.
Carrel L, Willard HF. 2005. X-inactivation profile reveals extensive
variability in X-linked gene expression in females. Nature 434: 400–404.
Clemson CM, McNeil JA, Willard HF, Lawrence JB. 1996. XIST RNA paints
the inactive X-chromosome at interphase: evidence for a novel RNA
involved in nuclear/chromosome structure. J Cell Biol 132: 259–275.
Costanzi C, Pehrson JR. 1998. Histone macroH2A1 is concentrated in the
inactive X chromosome of female mammals. Nature 393: 599–601.
Davies W, Isles A, Smith R, Karunadasa D, Burrmann D, Humby T, Ojarikre
O, Biggin C, Skuse D, Burgoyne P, et al. 2005. Xlr3b is a new imprinted
candidate for X-linked parent-of-origin effects on cognitive function in
mice. Nat Genet 37: 625–629.
de Conciliis L, Marchitiello A, Wapenaar MC, Borsani G, Giglio S, Mariani
M, Consalez GG, Zuffardi O, Franco B, Ballabio A, et al. 1998.
Characterization of Cxorf5 (71-7A), a novel human cDNA mapping to
Xp22 and encoding a protein containing coiled-coil alpha-helical
domains. Genomics 51: 243–250.
X-chromosome methylation
Genome Research 1599
www.genome.org
Dracopoli NC, Rettig WJ, Albino AP, Esposito D, Archidiacono N, Rocchi M,
Siniscalco M, Old LJ. 1985. Genes controlling gp25/30 cell-surface
molecules map to chromosomes X and Yand escape X-inactivation. Am
J Hum Genet 37: 199–207.
Eden E, Lipson D, Yogev S, Yakhini Z. 2007. Discovering motifs in ranked
lists of DNA sequences. PLoS Comput Biol 3: e39. doi: 10.1371/
journal.pcbi.0030039.
Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B,
Gautier L, Ge Y, Gentry J, et al. 2004. Bioconductor: open software
development for computational biology and bioinformatics. Genome
Biol 5: R80. doi: 10.1186/gb-2004-5-10-r80.
Gibbs JR, van der Brug MP, Hernandez DG, Traynor BJ, Nalls MA, Lai SL,
Arepalli S, Dillman A, Rafferty IP, Troncoso J, et al. 2010. Abundant
quantitative trait loci exist for DNA methylation and gene expression
in human brain. PLoS Genet 6: e1000952. doi: 10.1371/journal.pgen.
1000952.
Greenfield A, Carrel L, Pennisi D, Philippe C, Quaderi N, Siggers P, Steiner K,
Tam PP, Monaco AP, Willard HF, et al. 1998. The UTX gene escapes X
inactivation in mice and humans. Hum Mol Genet 7: 737–742.
Haaf T. 1995. The effects of 5-azacytidine and 5-azadeoxycytidine on
chromosome structure and function: Implications for methylation-
associated cellular processes. Pharmacol Ther 65: 19–46.
Hansen RS, Cranfield TK, Fjeld AD, Gartler SM. 1996. Role of late replication
timing in the silencing of X-linked genes. Hum Mol Genet 5: 1345–1353.
Hellman A, Chess A. 2007. Gene body-specific methylation on the active X
chromosome. Science 315: 1141–1143.
Jablonka E, Goitein R, Marcus M, Cedar H. 1985. DNA hypomethylation
causes an increase in DNase-I sensitivity and an advance in the time
of replication of the entire inactive X chromosome. Chromosoma 93:
152–156.
Johnston CM, Lovell FL, Leongamornlert DA, Stranger BE, Dermitzakis ET,
Ross MT. 2008. Large-scale population study of human cell lines
indicates that dosage compensation is virtually complete. PLoS Genet 4:
e9. doi: 10.1371/journal.pgen.0040009.
Keohane AM, Lavender JS, O’Neill LP, Turner BM. 1998. Histone acetylation
and X-inactivation. Dev Genet 22: 65–73.
Lahn BT, Page DC. 1999. Four evolutionary strata on the human X
chromosome. Science 286: 964–967.
Li Y, Zhu J, Tian G, Li N, Li Q, Ye M, Zheng H, Yu J, Wu H, Sun J, et al. 2010.
The DNA methylome of human peripheral blood mononuclear cells.
PLoS Biol 8: e1000533. doi: 10.1371/journal.pbio.1000533.
Lingenfelter PA, Adler DA, Poslinski D, Thomas S, Elliott RW, Chapman VM,
Disteche CM. 1998. Escape from X inactivation of Smcx is preceded by
silencing during mouse development. Nat Genet 18: 212–213.
Lock LF, Takagi N, Martin GR. 1987. Methylation of the Hprt gene on the
inactive X occurs after X chromosome inactivation. Cell 48: 39–46.
Lyon MF. 1961. Gene action in the X-chromosome of the mouse (Mus
musculus L.). Nature 190: 372–373.
Montgomery SB, Sammeth M, Gutierrez-Arcelus M, Lach RP, Ingle C,
Nisbett J, Guigo R, Dermitzakis ET. 2010. Transcriptome genetics using
second generation sequencing in a Caucasian population. Nature 464:
773–777.
Panning B, Jaenisch R. 1996. DNA hypomethylation can activate Xist
expression and silence X-linked genes. Genes Dev 10: 1991–2002.
Plagnol V, Uz E, Wallace C, Stevens H, Clayton D, Ozcelik T, Todd JA. 2008.
Extreme clonality in lymphoblastoid cell lines with implications for
allele specific expression analyses. PLoS ONE 3: e2966. doi: 10.1371/
journal.pone.0002966.
Raefski AS, O’Neill MJ. 2005. Identification of a cluster of X-linked
imprinted genes in mice. Nat Genet 37: 620–624.
Sharp AJ, Migliavacca E, Dupre Y, Stathaki E, Sailani MR, Baumer A, Schinzel
A, Mackay DJ, Robinson DO, Cobellis G, et al. 2010. Methylation
profiling in individuals with uniparental disomy identifies novel
differentially methylated regions on chromosome 15. Genome Res 20:
1271–1278.
Skuse DH, James RS, Bishop DV, Coppin B, Dalton P, Aamodt-Leeper G,
Bacarese-Hamilton M, Creswell C, McGurk R, Jacobs PA. 1997. Evidence
from Turner’s syndrome of an imprinted X-linked locus affecting
cognitive function. Nature 387: 705–708.
Smyth GK. 2004. Linear models and empirical Bayes methods for assessing
differential expression in microarray experiments. Stat Appl Genet Mol
Biol 3: Article 3. doi: 10.2202/1544-6115.1027.
Takagi N, Sasaki M. 1975. Preferential inactivation of the paternally derived
X chromosome in the extraembryonic membranes of the mouse. Nature
256: 640–642.
Tribioli C, Tamanini F, Patrosso C, Minalesi L, Villa A. 1992. Methylation
and sequence analysis around EagI sites: identification of 28 new CpG
islands in Xq24-q28. Nucleic Acids Res 20: 727–733.
Wang Z, Willard HF, Mukherjee S, Furey TS. 2006. Evidence of influence of
genomic DNA sequence on human X chromosome inactivation. PLoS
Comput Biol 2: e113. doi: 10.1371/journal.pcbi.0020113.
Weber M, Hellmann I, Stadler MB, Ramos L, Pa
¨a
¨bo S, Rebhan M, Schu
¨beler
D. 2007. Distribution, silencing potential and evolutionary impact of
promoter DNA methylation in the human genome. Nat Genet 39: 457–
466.
Willard HF, Latt SA. 1976. Analysis of deoxyribonucleic acid replication in
human X chromosomes by fluorescence microscopy. Am J Hum Genet
28: 213–227.
Wu J, Ellison J, Salido E, Yen P, Mohandas T, Shapiro LJ. 1994. Isolation
and characterization of XE169, a novel human gene that escapes
X-inactivation. Hum Mol Genet 3: 153–160.
Yasukochi Y, Maruyama O, Mahajan MC, Padden C, Euskirchen GM, Schulz
V, Hirakawa H, Kuhara S, Pan XH, Newburger PE, et al. 2010. X
chromosome-wide analyses of genomic DNA methylation states and
gene expression in male and female neutrophils. Proc Natl Acad Sci 107:
3704–3709.
Received July 9, 2010; accepted in revised form July 7, 2011.
Sharp et al.
1600 Genome Research
www.genome.org