Sequential ChIP-bisulfite sequencing enables direct
genome-scale investigation of chromatin and DNA
Arie B. Brinkman,1Hongcang Gu,2,4Stefanie J.J. Bartels,1,4Yingying Zhang,2,3,4
Filomena Matarese,1Femke Simmer,1Hendrik Marks,1Christoph Bock,2,3
Andreas Gnirke,2Alexander Meissner,2,3and Hendrik G. Stunnenberg1,5
1Radboud University, Nijmegen Center for Molecular Life Sciences, Department of Molecular Biology, 6500 HB Nijmegen,
The Netherlands;2Broad Institute, Cambridge, Massachusetts 02138, USA;3Department of Stem Cell and Regenerative
Biology, Harvard University, Cambridge, Massachusetts 02138, USA
Cross-talk between DNA methylation and histone modifications drives the establishment of composite epigenetic sig-
natures and is traditionally studied using correlative rather than direct approaches. Here, we present sequential ChIP-
bisulfite-sequencing (ChIP-BS-seq) as an approach to quantitatively assess DNA methylation patterns associated with
chromatin modifications or chromatin-associated factors directly. A chromatin-immunoprecipitation (ChIP)-capturing
step is used to obtain a restricted representation of the genome occupied by the epigenetic feature of interest, for which
a single-base resolution DNA methylation map is then generated. When applied to H3 lysine 27 trimethylation
(H3K27me3), we found that H3K27me3 and DNA methylation are compatible throughout most of the genome, except
for CpG islands, where these two marks are mutually exclusive. Further ChIP-BS-seq-based analysis in Dnmttriple-knockout
(TKO) embryonic stem cells revealed that total loss of CpG methylation is associated with alteration of H3K27me3 levels
throughout the genome: H3K27me3 in localized peaks is decreased while broad local enrichments (BLOCs) of H3K27me3
are formed. At an even broader scale, these BLOCs correspond to regions of high DNA methylation in wild-type ES cells,
suggesting that DNA methylation prevents H3K27me3 deposition locally and at a megabase scale. Our strategy provides
a unique way of investigating global interdependencies between DNA methylation and other chromatin features.
[Supplemental material is available for this article.]
Epigenetic regulation, involving DNA methylation and histone
modifications, is fundamental to a multitude of biological pro-
cesses such as transcription, DNA replication, and repair. The dif-
ferent modifications do not act independently of each other.
Instead, cross-talk between different modifications plays an im-
portant role in establishment of chromatin diversity within the
genome. Interdependent deposition and mutual exclusion of var-
ious marks result in complex modification patterns with different
functional outcomes (Fischle 2008; Cedar and Bergman 2009; Lee
et al. 2010). Classically, such patterns are determined by parallel
genomic mapping of the various modifications within the same
samples, using chromatin immunoprecipitation and deep se-
quencing (ChIP-seq) (Barski et al. 2007; Mikkelsen et al. 2007).
However, the analysis of cross-talk through independent profiling
experiments is complicated by cell population heterogeneity, cell
cycle effects, and allele-specific marking of chromatin such as in
imprinting or X-inactivation.
Here, we present a method for the integrated analysis of his-
tone modification or transcription factor deposition patterns and
the underlying DNA methylation. In our approach, termed ChIP-
BS-seq, ChIP capturing is followed by bisulfite conversion and
deep sequencing to directly assess DNA methylation levels in
captured chromatin fragments. While the use of whole-genome
bisulfite shotgun sequencing is limited by the cost of the required
ChIP capturing allows one to reach adequate coverage at routine-
scale sequencing, providing increased quantitative accuracy of
DNA methylation measurements within captured regions.
We used ChIP-BS-seq to study the global cross-talk between
H3K27me3 and DNA methylation, which are both linked to
repression. Polycomb Repressive Complex 2 (PRC2) catalyzes
H3K27me3 methylation via the SET domain of EZH2, while Poly-
comb Repressive Complex 1 (PRC1) is recruited to the H3K27me3
The DNA methyltransferases DNMT3A/B and DNMT1 are respon-
sible for the establishment and maintenance of the DNA methyl-
ation mark, respectively (Cedar and Bergman 2009). The interplay
between DNA methylation and H3K27me3/Polycomb has been
subject to extensive studies, and different phenomena have been
described (Cedar and Bergman 2009). Direct interactions between
Polycomb and the DNA methylation machinery have been
reported, suggesting that H3K27me3 and DNA methylation co-
occur (Vire ´ et al. 2006). Co-occurrence was further supported by
ChIP experiments analyzing DNA hypermethylated promoters
in cancer cells (McGarvey et al. 2006; Schlesinger et al. 2007).
Whereas H3K27me3 has been shown to ‘‘prime’’ gene promoters
for later DNA methylation (Ohm et al. 2007; Schlesinger et al.
2007; Widschwendter et al. 2007; Gal-Yam et al. 2008; Mohn et al.
2008), several reports have shown antagonism or mutual exclu-
siveness between H3K27me3 and DNA methylation (Kondo et al.
4These authors contributed equally to this work.
Article published online before print. Article, supplemental material, and publi-
cation date are at http://www.genome.org/cgi/doi/10.1101/gr.133728.111.
1128 Genome Research
22:1128–1138 ? 2012, Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/12; www.genome.org
2008; Lindroth et al. 2008; Bartke et al. 2010; Wu et al. 2010). Al-
though the different observations are not necessarily incom-
patible, the co-occurrence of both marks is still a subject of debate.
Using ChIP-BS-seq, we addressed the overlap of H3K27me3 and
DNA methylation directly on a genome-wide scale. Our results
unequivocally show that DNA methylation and H3K27me3 gen-
erally co-occur but are mutually exclusive in CpG-dense regions.
This mutual exclusivity is found in both a cancercell line as well as
in mouse ES cells. Loss of DNA methylation in ES cells is associated
with the formation of H3K27me3 patterns previously described as
broad local enrichments (BLOCs), and, at an even larger scale,
high DNA-methylation in wild-type cells.
Establishment of the strategy
An outline of our strategy, which we named ChIP-BS-seq, is shown
in Figure 1A. ChIP is used to capture a genomic subfraction asso-
ciated with a specific histone modification or transcription factor.
Similarly, such subfraction can be obtained by capture of methyl-
ated DNA using a methyl-CpG binding domain (MBD). Captured
DNA fragments are subjected to end-repair, adapter ligation using
methylated adapters, bisulfite conversion, PCR-amplification, and
deep sequencing (Fig. 1A). In this way, the levels of DNA methyl-
ation can be quantitatively assessed at base-resolution within the
genomic subfraction of interest.
To provide technical proof-of-principle of our strategy, we
used capturing of methylated DNA by MethylCap (Bock et al.
2010; Brinkman et al. 2010), followed by bisulfite-deep sequenc-
ing (MethylCap-BS-seq). In MethylCap, an MBD domain is used
to capture methylated DNA (Brinkman et al. 2010). Genomic
DNA isolated from normal and tumor colon tissues was used for
MethylCap-BS-seq as well as for conventional MethylCap-seq ex-
periments. Sequence reads were mapped as described in Methods
(see also Supplemental Table S1). Read densities across the entire
genome of MethylCap and MethylCap-BS-seq experiments corre-
lated well (Pearson R = 0.833) (Supplemental Fig. S1A), which is
comparable to technical replicates of conventional MethylCap
samples (Pearson R # 0.85) (AB Brinkman and HG Stunnenberg,
unpubl.). This indicated that the same genomic subfraction was
captured and sequenced in both procedures. Within each se-
quencing read, cytosines within a CpG context were scored for
their methylation status by counting the percentage of bisulfite-
induced mutations. As a control for bisulfite conversion efficiency,
we assessed DNA methylation within mitochondrial DNA which
was present at low levels due to its cellular abundance. Mito-
chondrial DNA is known to be completely unmethylated, and
conversion was calculated to be 99.91%. In addition, efficient
conversion of 99.7% was found in genomic CHG/CHH context.
inspection of individual peaks showed that they were mostly
hypermethylated (Fig. 1B; Supplemental Fig. S1B). This was con-
firmed on a global scale (Fig. 1C); mean methylation was 87%
within MethylCap peaks. At boundaries, methylation showed
a sharp drop that continued further with increasing distance from
peaks. This was accompanied by decreases in read densities and,
thus, CpG coverage. When focused exclusively on regions inside
the MethylCap peaks, we found that 85% and 79% of the CpGs
were at least 80% methylated in normal and tumor, respectively
(Supplemental Fig. S1C).
Differential methylation between normal and tumor as
observed by conventional MethylCap-seq was confirmed by
MethylCap-BS-seq; differentially methylated regions showed cor-
responding alterations in CpG coverage and absolute CpG meth-
ylation (Fig. 1B, bottom panel; Supplemental Fig. S1B, middle
panel). The same was observed on a global scale: Regions that
gained DNA methylation in tumor compared to normal tissue
(color-coded in Fig. 1D) showed increased coverage and methyla-
for regions that lost methylation. Taken together, MethylCap-BS-
seq showed hypermethylation of MethylCap-captured DNA, and
differences observed between normal and tumor tissue could be
corroborated and extended using MethylCap-BS-seq. These experi-
ments demonstrate the successful integration of capturing experi-
ments with bisulfite-deep sequencing.
We next applied our approach to ChIP-captured DNA (ChIP-BS-
seq) for analysis of DNA methylation patterns associated with
specific chromatin modifications. H3K27me3 and DNA methyla-
tion are both involved in gene silencing, but their interplay is
ChIP-BS-seq on H3K27me3, using HCT116 colon carcinoma cells.
For comparison, we also generated conventional ChIP-seq profiles
for H3K27me3. Read densities of the conventional H3K27me3
ChIP-seq and ChIP-BS-seq experiments correlated well (Pearson
R = 0.854) (Supplemental Fig. S2A), showing that the bisulfite/
mapping procedure did not alter the H3K27me3 patterns. The
H3K27me3 genome-wide profile showed H3K27me3 enrichments
over broad regions, comprising genes and intergenic regions (Fig.
2A; Supplemental Figs. S2C, S3C). This pattern resembled the
H3K27me3 ‘‘BLOCs’’ profile in mouse and human fibroblast cells
patterns in mouse ES cells, where the mark is present in focal areas
at silent promoters (Mikkelsen et al. 2007; Pan et al. 2007; Zhao
et al. 2007; Marks et al. 2009) (see also below). As described before,
genes that were located within the H3K27me3 BLOCs were gen-
erally silent, whereas genes outside BLOCs had a significantly
higher average expression level (Supplemental Fig. S3A).
To rule out that crosslinking/decrosslinking of chromatin
interfered with bisulfite conversion, we performed bisulfite se-
quencing on 12 independent PCR fragments amplified from ge-
nomic DNA isolated directly or after crosslinking/decrosslinking
(Supplemental Fig. S2B). None of the fragments showed altered
DNA methylation patterns after crosslinking/decrosslinking, show-
ing that ChIP-derived DNA did not affect bisulfite conversion.
DNA hypomethylation in H3K27me3-enriched high CpG
We then interrogated the DNA methylation status of captured
H3K27me3-marked chromatin. H3K27me3-marked chromatin
coincided mostly with fully methylated CpGs, with smaller
patches of lower methylation occurring in between (Fig. 2A; Sup-
with H3K27me3 contained exclusively unmethylated CpGs. For
example, the SERTAD4, SMAD7, and OVOL2 genes were located
mostly DNA methylated, except for the CpG islands encompassing
their gene promoters; these were completely hypomethylated.
To analyze DNA methylation of the H3K27me3-enriched
fraction on a genome-wide scale, DNA methylation was de-
DNA methylation and H3K27me3 cross-talk
termined in 300-bp windows throughout H3K27me3 BLOCs, and
windows were subsequently categorized according to their geno-
mic function (intergenic, intron, exon, non-CpG island promoter,
CpG island promoter). Histograms displaying DNA methylation
levels are shown in Figure 2C. Within a window size of 300 bp, the
DNA methylation pattern was clearly bimodal, as shown pre-
viously (Meissner et al. 2008). Genes, intergenic regions, and non-
was prevalent. In contrast, CpG island promoters marked with
seq and MethylCap-BS-seq procedures. Capturing of genomic regions of interest is achieved by the MethylCap procedure or ChIP with an antibody of
interest. Captured DNA is processed as indicated. Shown intermediate products and final PCR fragments indicate the fate of unmethylated as well as
methylated cytosines throughout the procedure. (B) Examples of conventional MethylCap-seq and MethylCap-BS-seq data of normal (N) and tumor (T)
colon tissues. For each covered CpG, percentage methylation as derived from the MethylCap-BS-seq data is indicated by color (yellow, 0%; blue, 100%).
(Green) CpG islands and a CpG density profile (CpG/bp). (C) Average profiles of DNA methylation and coverage in MethylCap peaks of normal colon
tissue, as determined by MethylCap-BS-seq. (Blue) Percentage DNA methylation; (magenta) CpG coverage; (brown) read density. (D) MethylCap-BS-seq
analysis of differentially methylated regions from normal/tumor colon tissue. Regions that gain DNA methylation in tumor tissue show increased CpG
coverage (x-axis) and read-density (y-axis). Color-code depicts absolute changes in percent methylation of these regions, as determined by bisulfite
sequencing. (Blue) increase; (yellow) decrease.
Integration of capturing methods and bisulfite-deep sequencing: ChIP-BS-seq and MethylCap-BS-seq. (A) Schematic outline of the ChIP-BS-
Brinkman et al.
1130 Genome Research
H3K27me3 contained only hypomethylated DNA. Since this exclu-
sively applied to CpG island containing promoters, we also catego-
rized windows according to CpG density (Supplemental Fig. S2E).
While low-CpG-dense windows (<0.06 CpGs/bp) contained mostly
DNA hypermethylation, high-CpG-density windows (>0.06 CpGs/
on H3K27me3-captured CpG islands. Median DNA methylation
from 5 kb outside to 0.5kb within these CpG islands was plotted
along with CpG-density and H3K27me3 levels (Fig. 2D). A clear and
resulting in an almost perfect inverse correlation between DNA
methylation and CpG density. Inside CpG islands, H3K27me3 was
consolidated. Taken together, our results show that H3K27me3 and
DNA methylation generally co-occur in low- CpG-density regions,
i.e., the bulk of the human genome. Within H3K27me3-marked re-
gions of high-CpG-density, such as CpG islands, DNA is exclusively
hypomethylated. We conclude that H3K27me3 and DNA methyla-
tion are mutually exclusive in CpG-dense regions.
Hypermethylated CpG islands show local H3K27me3
depletion within H3K27me3 BLOCs
The results obtained by H3K27me3 ChIP-BS-seq clearly demon-
strate the existence of mutual exclusiveness of H3K27me3 with
DNA methylation in CpG islands. To further confirm the rele-
vance of this observation, we analyzed genome-wide profiles for
H3K27me3 ChIP-BS-seq profiles are shown, as well as the derived DNA methylation data per covered CpG and per 200-bp window. Percentage meth-
ylation is color-coded as in Figure 1B. (C) Histograms showing the distribution of mean methylation in 300-bp windows throughout H3K27me3-enriched
regions, as derived from H3K27me3 ChIP-BS-seq results. Windows were categorized according to functional genomic elements (intergenic, intron, exon,
flanking H3K27me3-enriched CpG islands, as determined from H3K27me3 ChIP-BS-seq data.
H3K27me3-enriched CpG island promoters are devoid of DNA methylation. (A,B) Examples of H3K27me3 ChIP-BS-seq data of HCT116 cells.
DNA methylation and H3K27me3 cross-talk
DNA methylation also excludes H3K27me3. Obviously, the latter
could not be directly shown by ChIP-BS-seq because regions lacking
H3K27me3 were not captured and, thus, not available for DNA
To test our hypothesis, we examined DNA-methylated CpG
islands within H3K27me3 BLOCs. We detected 6456 H3K27me3
BLOCs with a median length of 77 kb (Supplemental Fig. S3B).
Thirty-one percent, or 8781 out of the total28,226 annotated CpG
islands (http://genome.ucsc.edu/) were located within H3K27me3
BLOCs. Of these CpG islands, 6473 contained MethylCap peaks.
Visual inspection of such methylated CpG islands within
H3K27me3 BLOCs revealed local depletions of H3K27me3 (Fig.
3A; Supplemental Fig. S3C). To assess this at a genome-wide scale,
we generated density maps representing
ChIP around BLOCs-contained methylated
CpG islands (Fig. 3B). A clear local de-
pletion of H3K27me3 was evident over
methylated CpG islands. To exclude that
this local depletion of H3K27me3 was
we used publicly available DNaseI hy-
persensitivity data from HCT116 cells
(ENCODE). MethylCap peaks were not
enriched for DNaseI hypersensitivity and
are, thus, unlikely to represent nucleo-
some-depleted regions (Supplemental Fig.
S3D). Taken together, our data show that
DNA-hypermethylated CpG islands pres-
ent within H3K27me3 BLOCs are locally
exclusiveness of DNA methylation and
Mutual exclusiveness of H3K27me3
and DNA methylation in mouse
To extend our observations to non-
ChIP-BS-seq strategy to mouse embryonic
stem (mES) cells. Regions of H3K27me3
enrichment were smaller than in HCT116,
representing the typical, more peak-like
mES pattern described before (Fig. 4A;
Supplemental Fig. S4A; Mikkelsen et al.
2007; Pan et al. 2007; Zhao et al. 2007;
or transcription start sites encompassing
CpG islands (32%) (Supplemental Fig.
these CpG-rich regions contained exclu-
sively unmethylated CpGs. For example,
the Htra4/Plekha2 and Lmx1b genes were
mostly DNA methylated, except for the
CpG islands underneath the H3K27me3
peaks (Fig. 4A). We generated average
profiles for H3K27me3 peaks over CpG
islands and accompanying DNA methylation from H3K27me3-
ChIP-BS-seq (Fig. 4B). Whereas DNA methylation on the flanks of
CpG islands was 80%, this dropped to zero in CpG islands. This
decrease appeared to be instigated at least 0.5 kb away from CpG
islands, where CpG density is elevated but, by far, not maximal. To
further relate DNA methylation–H3K27me3 mutual exclusiveness
to high CpG density, we applied a 300-bp sliding window approach
over all H3K27me3 peaks, categorized these windows according to
their CpG density, and inferred their DNA methylation status from
H3K27me3 and DNA methylation co-occurred within the same
windows. This CpG density corresponds to that encountered at the
windows of higher CpG density (>0.05 CpG/bp), DNA methyl-
ation was virtually absent, confirming the antagonism between
nome-wide H3K27me3 ChIP-seq and MethylCap-seq data, demonstrating hypermethylated CpG is-
lands within H3K27me3 BLOCs, and concomitant local depletion of H3K27me3. H3K27me3-enriched
BLOCs and MethylCap peaks are indicated as red and blue rectangles, respectively. (B) Density maps of
MethylCap-seq and H3K27me3 ChIP-seq read densities in 20-kb regions surrounding MethylCap peaks
that reside in H3K27me3 BLOCs and overlap with CpG islands.
H3K27me3 is locally depleted at hypermethylated CpG islands. (A) Examples of the ge-
Brinkman et al.
1132 Genome Research
percentage methylation as derived from the H3K27me3-BS-seq data is indicated in color. (B,C) Average profiles of DNA methylation (blue), H3K27me3
(red), and CpG density (green) in H3K27me3 peaks over CpG islands, as determined from H3K27me3-ChIP-BS-seq. (D) Histograms showing the dis-
CpG density. (E) H3K27me3 changes in CpG islands that were either hypermethylated (>90%, left) or hypomethylated (<10%, right) in wild-type mES
cells. Read counts shown are from conventional H3K27me3 ChIP.
Localized peaks of H3K27me3 in wild-type mES cells over CpG islands. (A) Screenshots of localized H3K27me3 peaks (red). Per covered CpG,
DNA methylation and H3K27me3 cross-talk
H3K27me3 and DNA methylation. Together, these data clearly con-
firm and extend our findings on mutual exclusiveness of DNA
methylation and H3K27me3 in regions of high CpG density to mES
As an extension of our observations with H3K27me3, we
classically been linked to DNA hypermethylation. Indeed, exam-
ples of H3K9me3 peaks, such as in the imprinting control region
(ICR) upstream of H19 (Bell and Felsenfeld 2000; Hark et al. 2000;
Kanduri et al. 2000), showed that this mark was associated with
hypermethylated DNA in mES cells (Supplemental Fig. S5A). Of all
detected H3K9me3 peaks, over 90% of them were associated with
hypermethylated DNA (Supplemental Fig. S5B). This is in sharp
contrast to H3K27me3 peaks, under which there is a more equal
subdivision of hypomethylated and hypermethylated DNA (Sup-
plemental Fig. S5B), corresponding to high-CpG-density and low-
CpG-density sequences, respectively (see Fig. 4D). These data
show that histone marks other than H3K27me3 may display
different DNA methylation properties, which extends and con-
firms the validity of the ChIP-BS-seq strategy.
H3K27me3 changes upon loss of DNA methylation
We next addressed the question of what happens to the
H3K27me3 distribution upon removal of DNA methylation. There-
fore, we performed H3K27me3-ChIP-BS-seq on Dnmt[1kd,3a?/?,
3b?/?] (TKO) mES cells (Meissner et al. 2005). DNA methylation
in virtually all of the captured/sequenced CpGs had disappeared
(Supplemental Fig. S5B).
We first focused on CpG islands and created average profiles
for H3K27me3 peaks over CpGislands (Fig. 4C). DNA methylation
had disappeared, and the H3K27me3 signal over CpG islands that
was generally high in wild-type cells had also decreased in TKO
type mES cells are scarceand not captured by H3K27me3 ChIP due
to the observed antagonism, so their contribution to these average
profiles is almost zero. To enable analysis of H3K27me3 changes in
hypermethylated CpG islands, we subselected such CpG islands
using publicly available data (Stadler et al. 2011) and plotted the
changes in H3K27me3 from conventional ChIP data. In case of
antagonism between DNA methylation and H3K27me3, we
in TKO cells, which was, indeed, the case (Fig. 4E). The opposite
effect, although to a lesser magnitude, occurred at hypomethylated
CpG islands. These CpG islands—containing high H3K27me3—
displayed a loss in H3K27me3 in TKO cells (Fig. 4E). Thus, antago-
in hypermethylated CpG islands but appeared to be absent in
unmethylated CpG islands, where loss of DNA methylation caused
a concomitant decrease of H3K27me3.
The H3K27me3 changes described above concern mainly the
sharp and localized peaks of H3K27me3 typical of mES cells. An
additional and striking observation in TKO cells was the appear-
ance of large regions of H3K27me3 enrichment, resembling the
typical BLOCs (Fig. 5A). To analyze the co-occurring changes of
peaks and BLOCs in more detail, we plotted the fold-change of
H3K27me3 BLOCs against the fold-change of the localized peaks
within these BLOCs (Fig. 5B). In 43% of the cases, the increase
in H3K27me3 in BLOCs was accompanied by a decrease of
H3K27me3 in localized peaks within the same BLOCs (Fig. 5B,
quadrant II). In another 18% of the cases, BLOCs appeared, but
the peaks of H3K27me3 in these BLOCs were maintained (Fig. 5B,
quadrant I). The observed changes of H3K27me3 in TKO cells—
decrease in peaks and accumulation in BLOCs—could be clearly
confirmed using targeted ChIP-qPCR. Ten out of the 12 tested
peaks showed a decrease, and eight out of the nine tested BLOCs
showed an increase (Supplemental Fig. S6). These results not only
validated our genome-wide analyses but also excluded the possi-
bility that a decrease of H3K27me3 peaks was a technical arti-
fact due to a higher complexity of the TKO sequencing libraries
by accumulation of H3K27me3 throughout a larger part of the
A possible explanation for accumulation of H3K27me3 BLOCs
could be a compensatory repressive effect instigated by the loss of
DNA methylation. If this were the case, H3K27me3 BLOCs ele-
vated in TKO cells are expected to represent genomic regions with
high DNA methylation in wild-type cells. We made use of our
in TKO-BLOC regions in wild-type cells (Fig. 5C). Indeed, in cases
where BLOCs became more prominent (quadrants I & II) (Fig.
5C), DNA methylation in wild-type mES cells was significantly
higher compared to regions where H3K27me3 was lost (quad-
rants I vs. IV, P = 2.8 3 10?68; quadrants II vs. III, P = 1.5 3 10?137,
Mann–Whitney U-test) (Fig. 5C). These data suggest that large
chromosomal regions with high DNA methylation become
more susceptible for accumulation of H3K27me3 upon removal
of DNA methylation. To analyze H3K27me3 changes instigated
by loss of DNA methylation at an even larger scale and in-
dependent of BLOC-like patterns, we generated a MethylCap
DNA methylation profile for wild-type mES cells. A sliding
window of 0.5 Mb was applied (Supplemental Fig. S7A). We
found that H3K27me3 in TKO cells resembled the MethylCap
profile of wild type (Supplemental Fig. S7B). Correlation be-
tween H3K27me3 and MethylCap profiles increased from 0.41
in wild type to 0.71 in TKO (Pearson correlation) (Supplemental
To analyze the consequence—if any—of H3K27me3 accu-
mulation in BLOCs at the level of gene expression, we catego-
rized RefSeq genes according to their position within or outside
BLOCs and plotted their expression levels in wild-type and TKO
cells (RNA-seq data from Karimi et al. 2011). Genes located out-
side BLOCs were expressed at higher levels than genes located
within BLOCs (Supplemental Fig. S8A), which is in line with
observations made by others (Pauler et al. 2009) and by us in
HCT116 cells (see Supplemental Fig. S3A). Strikingly, in wild-type
mES cells, these differences were already evident, even though
BLOCs are less prominent than in TKO cells. As shown before
(Karimi et al. 2011), loss of DNA methylation did not cause mas-
sive deregulation, and the expression levels of most genes were
stably maintained. Only 190 transcripts were found to be de-
regulated (FDR of 0.05, minimal twofold change) (Supplemen-
tal Fig. S8B).
A closer inspection of H3K27me3 patterns in wild-type mES
cells revealed that BLOCs of TKO cells could already be distin-
guished in wild-type cells, although signals were much weaker
(Fig. 5A). This was confirmed by plotting density maps of all
BLOC transition regions in both wild-type and TKO cells (Fig.
5D): the same BLOC boundaries were present in wild-type cells as
in TKO cells. Taken together, total removal of DNA methylation
caused an accumulation of H3K27me3 signal in BLOC-like pat-
terns with boundaries that had been set in wild-type cells. This
suggests thatbesidesthelocal antagonism betweenthetwomarks
at high-CpG-dense regions, there is also antagonism between the
two marks at a much larger scale in the genome.
Brinkman et al.
1134 Genome Research
Various methods for the generation of
genome-wide DNA methylation maps
exist. An extensive comparison of the
most frequently used bisulfite- and en-
richment-based technologies has re-
cently been performed (Bock et al. 2010;
Harris et al. 2010). In our strategy, we in-
tegrated MethylCap and ChIP-capturing
methylation analysis. Although similar
strategies have been used to interrogate
DNA methylation within a limited num-
ber of preselected regions (Kagey et al.
2010; Thomson et al. 2010), our strategy
allows us to obtain base-resolution DNA
methylation information within all frag-
ments obtained by a capturing step. Any
subset of the genome occupied by a spe-
cific feature can be directly assessed for
DNA methylation, provided that the fea-
ture of interest can be enriched for or
captured. Not only does this open up the
possibility to assess cross-talk between
DNA methylation and other chromatin
features, it also allows for the analysis of
DNA methylation within binding sites of
transcription factors that are dependent
on or excluded by DNA methylation. In
addition, it allows for detection of allele-
specific marking of chromatin such as in
imprinting or X-inactivation.
of-principle for our strategy and showed
that DNA fragments obtained by Meth-
ylCap represent a highly methylated
fraction of the genome. In addition, dif-
ferences in methylation between normal
and tumor tissue as detected by conven-
tional MethylCap-seq were corroborated
and extended by MethylCap-BS-seq.
ChIP-BS-seq was successfully established
using H3K27me3 and H3K9me3 as the
CpG density, H3K27me3 and DNA meth-
ylation co-occur or are mutual exclusive,
whereas H3K9me3 and DNA methylation
Upon loss of DNA methylation in
TKO cells, we observed two notable ef-
fects: (1) accumulation of H3K27me3 in
BLOC patterns; and (2) a decrease (flat-
tening) of sharp localized H3K27me3
peaks. Thus, depending on whether one
looks at BLOCs or smaller peaks, the ef-
fects of DNA methylation loss appear to
be different. Our data strongly suggest
that the increase in H3K27me3 BLOCs is
H3K27me3 accumulates in BLOC patterns
(TKO) mES cells. (A) Examples of H3K27me3 BLOCs appearing in TKO cells, and concomitant loss of
H3K27me3 in localized peaks. Per covered CpG, percentage methylation as derived from H3K27me3-
ChIP-BS-seq is indicated in color. (B) Scatterplot of H3K27me3 changes in peaks (x-axis) vs. BLOCs
(y-axis). H3K27me3 peaks were matched against BLOCs in which they reside. I–IV indicate the four
quadrants of the plot, as determined by log2fold changes deviating from zero. (C) Histograms of
in wild-type mES cells. (D) Density maps of H3K27me3 through BLOC transition regions detected in
TKO mES cells. Average profiles are shown on top.
Changes in H3K27me3 patterns upon loss of DNA methylation in Dnmt triple-knockout
DNA methylation and H3K27me3 cross-talk
have elevated methylation already in wild-type mES cells. A de-
crease of H3K27me3 in peaks occurs mostly in unmethylated CpG
islands; likely, surrounding DNA methylation normally constrains
H3K27me3 to unmethylated CpG islands in wild-type cells,
a constraint which is lost in TKO cells. This constraint probably
relates to the same antagonism observed elsewhere in the ge-
nome and, as such, limits the spread of H3K27me3 into neigh-
boring chromatin. Thus, although the H3K27me3 changes ob-
served in BLOCs and peaks are different, they may result from
We have ruled out that the decrease in peaks is a technical artifact
caused by increased complexity of sequencing libraries because
a larger part of the genome is enriched for H3K27me3 (Supple-
mental Fig. S6).
How the observed mutual exclusiveness of H3K27me3 and
DNA methylation within regions of high CpG density is achieved
mechanistically is not yet clear, but several studies have described
an antagonism between H3K27me3 and DNA methylation. In
SILAC nucleosome affinity purifications, DNA methylation of
nucleosome positioning sequences impeded PRC2 binding to
H3K27me3-modified nucleosomes (Bartke et al. 2010). The CpG
density of the nucleosome positioning sequences (‘‘601’’ and
‘‘603’’) was 0.09 CpG/bp. At this density, we observed mutual ex-
Therefore, one explanation for our observations may be obstruc-
tion of PRC2 recruitment at DNA methylated CpG-dense regions.
Such obstruction may also be involved in epigenetic switching as
described for prostate cancer cells (Gal-Yam et al. 2008). Genes
initially silenced by Polycomb in normal prostate cells acquired
DNA methylation and lost H3K27me3 in the PC3 cancer cells.
The appearance of H3K27me3 BLOCs in TKO cells suggests
that DNA methylation more globally antagonizes accumulation of
H3K27me3, which is alleviated in TKO cells. Still, the resulting
BLOCs occur in specific genomic regions that, importantly, are
preset in wild-type mES cells, indicating that there are positional
restraints on the deposition of H3K27me3. It is possible that these
regions and their boundaries are characterized by other (epi)ge-
netic features such as overrepresentation of genomic elements like
domains(Guelenetal.2008),or transcriptional factorbinding.We
anticipate that the two-dimensional information obtained using
ChIP-BS-seq will provide new insights into the composition of
different types of chromatin and their biological roles.
ChIP-seq, MethylCap-seq, and RNA-seq
HCT116 cells were cultured in McCoy’s 5A medium supplemented
with 10% fetal bovine serum and 1% penicillin/streptomycin (all
Gibco/Invitrogen) at 37°C in 5% CO2atmosphere. Mouse ES cells
were cultured as described in Meissner et al. 2005). Chromatin
harvesting and ChIPs were performed as described (Denissov et al.
2007). The following antibodies were used: anti-H3K27me3 (07-
MethylCap-seqwas performedas described(Brinkmanetal.2010).
For RNA-seq, total RNA was isolated using Trizol (Invitrogen)
according to the manufacturer’s recommendations. 100 mg total
RNA was subjected to two rounds of poly(A) selection (Oligotex
mRNA Mini Kit, QIAGEN), followed by DNaseI treatment
(QIAGEN). 100 ng mRNA was fragmented by hydrolysis (53 frag-
mentationbuffer:200 mM Tris acetate,pH 8.2, 500 mM potassium
acetate, and 150 mM magnesium acetate) at 94°C for 90 sec and
purified (RNeasy MinElute Cleanup Kit, QIAGEN). cDNA was
Transcriptase (Invitrogen). ds-cDNA synthesis was performed in
second strand buffer (Invitrogen) according to the manufacturer’s
recommendations and purified (MinElute Reaction Cleanup Kit,
QIAGEN). ChIP and MethylCap DNA and ds-cDNA were prepared
for Illumina sequencing according to the manufacturer’s protocols
ChIP-BS-seq and MethylCap-BS-seq
100 ng ChIP DNA or 10 ng MethylCap DNA from the HIGH frac-
tion (Brinkman et al. 2010) was prepared for bisulfite-deep se-
quencing. DNAwas first subjected to end-repair in a 30-ml reaction
containing 6 units T4 DNA polymerase, 2.5 units DNA Polymerase
I (Large Klenow Fragment), 20 units T4 Polynucleotide Kinase (all
New England Biolabs), dATP, dCTP, dGTP, and dTTP (0.125 mM
each), and 13 T4 Ligase buffer with ATP for 30 min at 20°C. Illu-
mina sequencing generates sequences corresponding to the 59
ends of the input DNA fragments (see also Figure 1A). Therefore,
fill-in of 59 overhangs did not alter sequenced DNA methylation
patterns. Purification was performed using a standard phenol:
chloroform:isoamyl alcohol (25:24:1) protocol and ethanol pre-
cipitation as described previously (Smith et al. 2009). DNA was
then adenylated in a 20-ml reaction containing 10 units Klenow
Fragment (39!59 exo-) (New England Biolabs), 0.5 mM dATP and
13 NEB buffer 2 for 30 min at 37°C. After phenol extraction and
ethanol precipitation, Illumina genomic DNA adapters containing
5-methylcytosine instead of cytosine (ATDBio), preventing de-
amination during bisulfite conversion, were ligated. In a 20-ml reac-
tion, DNA was incubated with 1.5 mM preannealed adapters, 2000
with ATP, for 16–20 h at 16°C.
Adapter-ligated DNA fragments were subsequently purified
by phenol extraction and ethanol precipitation and size-selected
on gel. 50 ng sheared and dephosphorylated Escherichia coli K12
genomic DNA was added to adapter-ligated DNA as carrier during
size-selection and bisulfite conversion. DNA was run on 2.5%
Nusieve 3:1 Agarose (Lonza) gels. Lanes containing marker (50 bp
ladder; New England Biolabs) were stained with SYBR Green (Invi-
trogen), and size regions to be excised were marked with toothpicks.
adapter-ligated DNA fragments from 200–400 and 400–550 bp,
respectively, were excised. The Illumina adapters cause the frag-
mentstorunslower, presumablyduetotheforkedstructure,as has
been described before (Smith et al. 2009). Note that, after PCR,
dsDNA libraries appear at 140–340 bp and 340–490 bp, in accor-
dance with exact sizes. DNA was isolated from gel using the
MinElute Gel Extraction kit (QIAGEN). The low and high libraries
were kept separate in subsequent steps.
Analytical PCRs were performed to check ligation efficiency
and sizes of the libraries. Amplifications were performed in 10-ml
reactions containing 0.3 ml template DNA (from 20 ml eluted after
size-selection), 0.5 units Pfu Turbo Cx Hotstart DNA Polymerase
(Stratagene), Illumina primers LPX 1.1 and 2.1 (0.3 mM each),
dNTPs (0.25 mM each), and 13 Turbo Cx buffer under the fol-
lowingthermocyclerconditions: 94°C for 5 min, n cycles (94°C for
30 sec, 65°C for 30 sec, 72°C for 1 min), and 72°C for 7 min. We
tested three different cycle numbers (n)—10, 15, and 20—and
analyzed PCR products on 4%–20% TBE Criterion precast gels
(BioRad) using SYBR Green staining.
Adapter-ligated and size-selected DNA was subjected to two
subsequent 5-h bisulfite treatments using the EpiTect Bisulfite kit
(QIAGEN) following the manufacturer’s protocol for DNA isolated
Brinkman et al.
1136 Genome Research
from FFPE tissue samples. After bisulfite conversion, analytical
PCRs were performed as before to determine the minimum num-
ber of cycles required in the final amplification step for each
sample. In this case, tested cycle numbers were several cycles
higher than before bisulfite conversion: 15, 19, and 22 PCR cycles
were performed. The minimum cycle number for final large-scale
amplification was determined as the lowest cycle number that
generated enough PCR product of the desired size range to be vi-
sualized on analytical gels as above. Large-scale amplification was
performed in eight reactions of 25 ml, each containing 3 ml DNA
(from 40 ml bisulfite-converted DNA; the remainder was stored at
?80°C as back-up), 1.25 units Pfu Turbo Cx Hotstart DNA Poly-
merase (Stratagene), primer LPX 1.1 and 2.1 (0.3 mM each), dNTPs
(0.25 mM each), 13 Turbo Cx buffer, and thermocycler condi-
tions as above. Amplified libraries were purified with the
MinElute PCR Purification kit (QIAGEN) and subsequently pu-
rified from gel essentially as described above; whole gels were
stained with SYBR Green, and no carrier DNA was added. Final
libraries were analyzed on analytical 4%–20% TBE Criterion
precast gels (BioRad), and measured by Quant-iT dsDNA HS As-
says (Invitrogen). The protocol for preparation of captured DNA
for bisulfite-deep sequencing was adapted from Smith et al.
(2009)and Guetal. (2010, 2011). Sequence reads were generated
on the Illumina Genome Analyzer IIx or the HiSeq2000 using
a standard 36-base protocol. After sequencing, low and high li-
brary reads were combined.
Bisulfite sequencing of PCR fragments
Cultured cells (SKNO-1) were either untreated or crosslinked di-
rectly in culture medium by the addition of 1% formaldehyde for
15 min. at room temperature. Chromatin harvesting, decross-
linking, and DNA isolation were done as described previously
(Denissov et al. 2007). Bisulfite conversion was performed using
the EpiTect Bisulfite kit (QIAGEN) following the manufacturer’s
standard protocol. For each sample, PCR amplicons of the appro-
priate size were excised from agarose gel, pooled, and subjected to
end-repair as above. Pooled fragments were subsequently con-
catamerized by ligation in the presence of 17% PEG3350 and
sonicated using a Bioruptor (Diagenode) at high power for 30 min.
in a final volume of 300 ml. The obtained DNA fragments were
subjected to library preparation according to the standard pro-
cedure (Illumina). Index sequences were introduced by using in-
house generated single-read adapters that contained a six-base
barcode directly after the sequence primer binding site. The two
samples (untreated and crosslinked/decrosslinked) were pooled
and sequenced on the Illumina Genome Analyzer IIx using a
standard, 36-base protocol.
ChIP-seq and MethylCap-seq data analysis
MethylCap peaks in HCT116 and H3K27me3 peaks in mES cells
were called by MACS (Zhang et al. 2008), with mfold = 4 and
P-value = 1310?06and 1 3 10?10, respectively. MethylCap LOW,
MEDIUM, and HIGH fractions (Brinkman et al. 2010) were used
individually for peak calling, after which peaks were merged.
Identification of H3K27me3-enriched regions (BLOCs) was per-
formed using the RSEG algorithm (Song and Smith 2011), which
models the read counts with a negative binomial distribution after
correcting for the effect of genomic dead zones. Subsequently, it
uses a two-state HMM for segmentation of the genome into fore-
ground domains and background domains. We used a bin size of
500 bp in combination with default RSEG settings. These include
that the posterior probability of each bin obtained by HMM
decoding is larger than 0.95 and that the mean of read counts
within a region is above the top 90th percentile of foreground
emission distribution. BLOCs within 20-kb proximity were merged,
and BLOCs smaller than 20 kb were discarded.
ChIP-BS-seq and MethylCap-BS-seq data analysis
Initial data processing and base-calling was done using the Illu-
mina pipeline software. Mapping of bisulfite-converted sequence
reads was done using a custom-made pipeline using a strategy
similar to that in Lister et al. (2009). To reduce PCR artifacts,
a maximum of three identical sequence reads was allowed. To
perform mapping independently of DNA methylation status, se-
quence reads were in silico bisulfite-converted (C to T) and sub-
sequently mapped to two different in silico converted hg18 ge-
nome sequences; one C to T converted genome and one G to A
converted genome. Reads mapping to both genomes were dis-
carded, which typically represented a very minor fraction of all
reads. Mapping was done using the Burrows-Wheeler Aligner
mismatch. Percentages of uniquely mapped reads ranged from
78% to 50% for the 80- to 280-bp and 280- to 430-bp libraries,
respectively (see Supplemental Table S1 for details). The obtained
mapping positions were used to align unconverted sequence reads
with their corresponding unconverted genomic sequence and to
subsequently determine the methylation status of each sequenced
cytosine within a CpG context, both on the forward strand as well
as on the reverse strand. The mapping and CpG methylation
scoring procedure was driven by a custom-generated Perl script.
Further data analysis was done using in-house generated scripts
written in LINUX shell, Python, Perl, and R. Gene annotations
were based on RefSeq (hg18); CpG islands annotations were based
on UCSC (http://genome.ucsc.edu/)
The data generated for this work have been deposited in the NCBI
Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/
geo/) and are accessible through GEO Series accession number
We thank E.M. Janssen-Megens, Y. Tan, and K.-J. Francoijs for
technical support. This work was supported by the Dutch Cancer
the CancerDIP EU Collaborative project HEALTH-F2-2007-200620,
the Harvard Stem Cell Institute, and the U.S. National Institutes of
Health Roadmap Initiative on Epigenomics (U01ES017155).
BarskiA,Cuddapah S,CuiK,RohT-Y,SchonesDE,WangZ, WeiG,Chepelev
human genome. Cell 129: 823–837.
Bartke T, Vermeulen M, Xhemalce B, Robson SC, Mann M, Kouzarides T.
2010. Nucleosome-interacting proteins regulated by DNA and histone
methylation. Cell 143: 470–484.
Bell AC, Felsenfeld G. 2000. Methylation of a CTCF-dependent
boundary controls imprinted expression of the Igf2 gene. Nature 405:
Bock C, Tomazou EM, Brinkman AB, Mu ¨ller F, Simmer F, Gu H, Ja ¨ger N,
Gnirke A, Stunnenberg HG, Meissner A. 2010. Quantitative comparison
of genome-wide DNA methylation mapping technologies. Nat
Biotechnol 28: 1106–1114.
Brinkman AB, Simmer F, Ma K, Kaan A, Zhu J, Stunnenberg HG. 2010.
Whole-genome DNA methylation profiling using MethylCap-seq.
Methods 52: 232–236.
Cedar H, Bergman Y. 2009. Linking DNA methylation and histone
modification: Patterns and paradigms. Nat Rev Genet 10: 295–304.
Denissov S, van Driel M, Voit R, Hekkelman M, Hulsen T, Hernandez N,
Grummt I, Wehrens R, Stunnenberg H. 2007. Identification of novel
DNA methylation and H3K27me3 cross-talk
functional TBP-binding sites and general factor repertoires. EMBO J 26:
Fischle W. 2008. Talk is cheap–cross-talk in establishment, maintenance,
and readout of chromatin modifications. Genes Dev 22: 3375–3382.
Gal-Yam EN, Egger G, Iniguez L, Holster H, Einarsson S, Zhang X, Lin JC,
Liang G, Jones PA, Tanay A. 2008. Frequent switching of Polycomb
repressive marks and DNA hypermethylation in the PC3 prostate cancer
cell line. Proc Natl Acad Sci 105: 12979–12984.
Gu H, Bock C, Mikkelsen TS, Ja ¨ger N, Smith ZD, Tomazou E, Gnirke A,
Lander ES, Meissner A. 2010. Genome-scale DNA methylation mapping
of clinical samples at single-nucleotide resolution. Nat Methods 7: 133–
Gu H, Smith ZD, Bock C, Boyle P,Gnirke A, Meissner A. 2011. Preparation of
reduced representation bisulfite sequencing libraries for genome-scale
DNA methylation profiling. Nat Protoc 6: 468–481.
Guelen L, Pagie L, Brasset E, Meuleman W, Faza MB, Talhout W, Eussen BH,
de Klein A, Wessels L, de Laat W, et al. 2008. Domain organization of
human chromosomes revealed by mapping of nuclear lamina
interactions. Nature 453: 948–951.
Hark AT, Schoenherr CJ, Katz DJ, Ingram RS, Levorse JM, Tilghman SM.
2000. CTCF mediates methylation-sensitive enhancer-blocking activity
at the H19/Igf2 locus. Nature 405: 486–489.
Harris RA, Wang T, Coarfa C, Nagarajan RP, Hong C, Downey SL, Johnson
BE,Fouse SD,DelaneyA,ZhaoY,etal. 2010. Comparisonofsequencing-
based methods to profile DNA methylation and identification of
monoallelic epigenetic modifications. Nat Biotechnol 28: 1097–1105.
Hawkins RD, Hon GC, Lee LK, Ngo Q, Lister R, Pelizzola M, Edsall LE, Kuan
S, Luu Y, Klugman S, et al. 2010. Distinct epigenomic landscapes of
Kagey JD, Kapoor-Vazirani P, McCabe MT, Powell DR, Vertino PM. 2010.
Long-term stability of demethylation after transient exposure to 5-aza-
29-deoxycytidine correlates with sustained RNA polymerase II
occupancy. Mol Cancer Res 8: 1048–1059.
Kanduri C, Pant V, Loukinov D, Pugacheva E, Qi CF, Wolffe A, Ohlsson R,
Lobanenkov VV. 2000. Functional association of CTCF with the
insulator upstream of the H19 gene is parent of origin-specific and
methylation-sensitive. Curr Biol 10: 853–856.
KarimiMM, Goyal P, Maksakova IA, Bilenky M,Leung D,Tang JX, Shinkai Y,
Mager DL, Jones S, Hirst M, et al. 2011. DNA methylation and SETDB1/
H3K9me3 regulate predominantly distinct sets of genes, retroelements,
and chimeric transcripts in mESCs. Cell Stem Cell 8: 676–687.
Kondo Y, Shen L, Cheng AS, Ahmed S, Boumber Y, Charo C, Yamochi T,
Urano T, Furukawa K, Kwabi-Addo B, et al. 2008. Gene silencing in
cancer by histone H3 lysine 27 trimethylation independent of promoter
DNA methylation. Nat Genet 40: 741–750.
Lee J-S, Smith E, Shilatifard A. 2010. The language of histone crosstalk. Cell
Li H, Durbin R. 2010. Fast and accurate long-read alignment with Burrows-
Wheeler transform. Bioinformatics 26: 589–595.
Lindroth AM, Park YJ, McLean CM, Dokshin GA, Persson JM, Herman H,
Pasini D, Miro ´ X, Donohoe ME, Lee JT, et al. 2008. Antagonism between
DNA and H3K27 methylation at the imprinted Rasgrf1 locus. PLoS Genet
4: e1000145. doi: 10.1371/journal.pgen.1000145.
Lister R, Pelizzola M, Dowen RH, Hawkins RD, Hon G, Tonti-Filippini J, Nery
JR, Lee L, Ye Z, Ngo Q-M, et al. 2009. Human DNA methylomes at base
Marks H, Chow JC, Denissov S, Franc xoijs K-J, Brockdorff N, Heard E,
Stunnenberg HG. 2009. High-resolution analysis of epigenetic changes
associated with X inactivation. Genome Res 19: 1361–1373.
McGarvey KM, Fahrner JA, Greene E, Martens J, Jenuwein T, Baylin SB. 2006.
return to a fully euchromatic chromatin state. Cancer Res 66: 3541–3549.
Meissner A, Gnirke A, Bell GW, Ramsahoye B, Lander ES, Jaenisch R. 2005.
Reduced representation bisulfite sequencing for comparative high-
resolution DNA methylation analysis. Nucleic Acids Res 33: 5868–5877.
Meissner A, Mikkelsen TS, Gu H, Wernig M, Hanna J, Sivachenko A, Zhang
X, Bernstein BE, Nusbaum C, Jaffe DB, et al. 2008. Genome-scale DNA
methylation maps of pluripotent and differentiated cells. Nature 454:
Brockman W, Kim T-K, Koche RP, et al. 2007. Genome-wide maps of
chromatinstate in pluripotentand lineage-committedcells. Nature448:
Mohn F, Weber M, Rebhan M, Roloff TC, Richter J, Stadler MB, Bibel M,
Schu ¨beler D. 2008. Lineage-specific polycomb targets and de novo DNA
methylation define restriction and potential of neuronal progenitors.
Mol Cell 30: 755–766.
Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L, Mohammad
HP, Chen W, Daniel VC, Yu W, et al. 2007. A stem cell-like chromatin
and heritable silencing. Nat Genet 39: 237–242.
Pan G, Tian S, Nie J, Yang C, Ruotti V, Wei H, Jonsdottir GA, Stewart R,
Thomson JA. 2007. Whole-genome analysis of histone H3 lysine 4 and
lysine 27 methylation in human embryonic stem cells. Cell Stem Cell 1:
Pauler FM, Sloane MA, Huang R, Regha K, Koerner MV, Tamir I, Sommer A,
Aszodi A, Jenuwein T, Barlow DP. 2009. H3K27me3 forms BLOCs over
silent genes and intergenic regions and specifies a histone banding
pattern on a mouse autosomal chromosome. Genome Res 19: 221–
Peters AHFM, Kubicek S, Mechtler K, O’Sullivan RJ, Derijck AAHA, Perez-
Burgos L, Kohlmaier A, Opravil S, Tachibana M, Shinkai Y, et al. 2003.
Partitioning and plasticity of repressive histone methylation states in
mammalian chromatin. Mol Cell 12: 1577–1589.
Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Zimmerman J,
Eden E, Yakhini Z, Ben-Shushan E, Reubinoff BE, et al. 2007. Polycomb-
mediated methylation on Lys27 of histone H3 pre-marks genes for de
novo methylation in cancer. Nat Genet 39: 232–236.
Simon JA, Kingston RE. 2009. Mechanisms of polycomb gene silencing:
Knowns and unknowns. Nat Rev Mol Cell Biol 10: 697–708.
Smith ZD, Gu H, Bock C, Gnirke A, Meissner A. 2009. High-throughput
bisulfite sequencing in mammalian genomes. Methods 48: 226–232.
Song Q, Smith AD. 2011. Identifying dispersed epigenomic domains from
ChIP-Seq data. Bioinformatics 27: 870–871.
Stadler MB, Murr R, Burger L, Ivanek R, Lienert F, Scho ¨ler A, Wirbelauer C,
Oakeley EJ, Gaidatzis D, Tiwari VK, et al. 2011. DNA-binding factors
shape the mouse methylome at distal regulatory regions. Nature 480:
Thomson JP, Skene PJ, Selfridge J, Clouaire T, Guy J, Webb S, Kerr ARW,
Deaton A, Andrews R, James KD, et al. 2010. CpG islands influence
chromatin structure via the CpG-binding protein Cfp1. Nature 464:
Vire ´ E, Brenner C, Deplus R, Blanchon L, Fraga M, Didelot C, Morey L, Van
Eynde A, Bernard D, Vanderwinden J-M, et al. 2006. The Polycomb
group protein EZH2 directly controls DNA methylation. Nature 439:
Widschwendter M, Fiegl H, Egle D, Mueller-Holzner E, Spizzo G, Marth C,
Weisenberger DJ, Campan M, Young J, Jacobs I, et al. 2007. Epigenetic
stem cell signature in cancer. Nat Genet 39: 157–158.
Wu H, Coskun V, Tao J, Xie W, Ge W, Yoshikawa K, Li E, Zhang Y, Sun YE.
2010. Dnmt3a-dependent nonpromoter DNA methylation facilitates
transcription of neurogenic genes. Science 329: 444–448.
Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, Nusbaum
C, Myers RM, Brown M, Li W, et al. 2008. Model-based analysis of ChIP-
Seq (MACS). Genome Biol 9: R137. doi: 10.1186/gb-2008-9-9-r137.
Zhao XD, Han X, Chew JL, Liu J, Chiu KP, Choo A, Orlov YL, Sung W-K,
Shahab A, Kuznetsov VA, et al. 2007. Whole-genome mapping of
histone H3 Lys4 and 27 trimethylations reveals distinct genomic
compartments in human embryonic stem cells. Cell Stem Cell 1: 286–
Received October 21, 2011; accepted in revised form March 28, 2012.
Brinkman et al.
1138 Genome Research