Content uploaded by Carrie Shawber
Author content
All content in this area was uploaded by Carrie Shawber on Apr 10, 2014
Content may be subject to copyright.
Multiplexed profiling of candidate genes for CpG
island methylation status using a flexible
PCR/LDR/Universal Array assay
Yu-Wei Cheng,
1
Carrie Shawber,
2
Dan Notterman,
3
Philip Paty,
4
and Francis Barany
1,5
1
Department of Microbiology and Immunology, Weill Medical College of Cornell University, New York, New York 10021, USA;
2
Department of OB/GYN, Columbia University Medical Center, New York, New York 10032, USA;
3
Departments of Pediatrics
and Molecular Genetics, University of Medicine and Dentistry of New Jersey (UMDNJ)-Robert Wood Johnson Medical School,
New Brunswick, New Jersey 08901, USA;
4
Department of Surgery, Colorectal Surgery Service, Memorial Sloan-Kettering Cancer
Center, New York, New York 10021, USA
DNA methylation in CpG islands is associated with transcriptional silencing. Accurate determination of cytosine
methylation status in promoter CpG dinucleotides may provide diagnostic and prognostic value for human cancers.
We have developed a quantitative PCR/LDR/Universal Array assay that allows parallel evaluation of methylation
status of 75 CpG dinucleotides in the promoter regions of 15 tumor suppressor genes (CDKN2B,CDKN2A,CDKN2D,
CDKN1A,CDKN1B,TP53,BRCA1,TIMP3,APC,RASSF1,CDH1,MGMT,DAPK1,GSTP1, and RARB). When compared with an
independent pyrosequencing method at a single promoter, the two approaches gave good correlation. In a study
using 15 promoter regions and seven blinded tumor cell lines, our technology was capable of distinguishing
methylation profiles that identified cancer cell lines derived from the same origins. Preliminary studies using 96
colorectal tumor samples and 73 matched normal tissues indicated CpG methylation is a gene-specific and
nonrandom event in colon cancer. This new approach is suitable for clinical applications where sample quantity and
purity can be limiting factors.
[Supplemental material is available online at www.genome.org.]
Aberrant methylation of CpG dinucleotides in the 5⬘regulatory
region of genes often results in transcriptional inactivation and
has been implicated in aging, heart and neurodegenerative dis-
eases, as well as in the pathogenesis of various types of cancers
(Feinberg and Vogelstein 1983; Gardiner-Garden and Frommer
1987; Post et al. 1999; Baylin and Herman 2000; Robertson and
Wolffe 2000; Warnecke and Bestor 2000; Feinberg 2001; Jones
and Baylin 2002; Cui et al. 2003). There is a growing interest in
understanding the correlation between aberrant DNA methyl-
ation and tumorigenesis (Huang et al. 1999; Toyota et al. 1999;
Costello et al. 2000; Yamashita et al. 2003) in order to facilitate
disease marker discovery, diagnostic tool development, and the
study of chemotherapeutic response (Laird 2003).
Current methods for detecting 5-methylcytosine can be di-
vided into three major approaches (Laird 2003): (1) profiling
methylation globally, (2) identifying methylation patterns at a
cluster of CpG sites, and (3) determining methylation levels at
individual CpG dinucleotides. Each category offers a different
perspective for studying DNA methylation. In general, global
screening methods rely on methylation-sensitive restriction en-
zyme digestion and provide opportunities for new epigenetic
marker discovery (Huang et al. 1999; Toyota et al. 1999; Costello
et al. 2000; Yamashita et al. 2003). Methylation-specific PCR
(MSP) (Herman et al. 1996) and variations of this procedure (Cot-
trell and Laird 2003; Zeschnigk et al. 2004) were introduced to
study the methylation pattern of a few closely neighboring CpG
sites. Since DNA methylation is believed to be an early event
during carcinogenesis (Laird 1997), the high sensitivity of MSP is
suitable for use as an early detection tool on known epigenetic
markers (Hoque et al. 2004). For quantitative assessment of in-
dividual CpG dinucleotide methylation status, the commonly
used methods, including bisulfite treatment, were followed by
sequencing (e.g., bisulfite sequencing and pyrosequencing)
(Frommer et al. 1992; Uhlmann et al. 2002; Dupont et al. 2004;
Yang et al. 2004), primer extension (e.g., SNuPE) (Gonzalgo and
Jones 1997), restriction enzyme digestion (e.g., COBRA) (Xiong
and Laird 1997), or real-time PCR (Zeschnigk et al. 2004). These
assays provide quantitative profiling or detailed analysis of
5-methylcytosine distribution. The quantitative information
generated is currently being used for correlating disease-specific
methylation markers to clinical outcomes and facilitating the
discovery of anti-tumorigenic drugs (Cheng et al. 2004; Issa
2004). However, the current methods analyze CpG methylation
status one gene at a time and have limited multiplexing capabil-
ity. Bisulfite sequencing provides the most comprehensive data
on methylation status at every CpG but requires subcloning and
sequence analysis of 10–20 individual clones. Higher throughput
has been achieved by combining bisulfite-PCR with microarray
technology utilizing oligonucleotide probes designed to form a
perfect match with either methylated or unmethylated alleles
within the target sequences (Adorjâan et al. 2002; Balog et al.
2002; Gitan et al. 2002). This allows parallel evaluation of CpG
methylation status at numerous CpG sites across multiple ge-
nomic regions of interest. However, bisulfite treatment renders
genomic DNA into AT-rich sequences, which exacerbates non-
specific and mismatch hybridizations due to differences in an-
nealing temperatures between different probe sequences. In ad-
5
Corresponding author.
E-mail barany@med.cornell.edu; fax (212) 746-8104.
Article published online ahead of print. Article and publication date are at
http://www.genome.org/cgi/doi/10.1101/gr.4181406.
Methods
282 Genome Research
www.genome.org
16:282–289 ©2006 by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/06; www.genome.org
dition, probes containing two or more CpG dinucleotides may
lack the sensitivity to distinguish partially methylated sequences
from those that are fully methylated in heterogeneous clinical
samples.
We seek to develop a robust assay for clinical application
that provides quantitative methylation levels for multiple CpG
dinucleotides in a given genomic region, as well as allowing spe-
cific evaluation of many genes in parallel. Such an assay can
provide a representational CpG methylation profile of candidate
genomic regions, and this profile information may be useful for
disease stratification or as predictors of therapeutic response. This
work presents a new method that aims to substantially improve
quantitative microarray-based methylation detection to meet
these needs. As illustrated in Figure 1, combining PCR, ligase
detection reaction (LDR), and universal Array (where zip-code
sequences appended to LDR primers, guide products to zip-code
complements on an array) (Gerry et al. 1999; Favis et al. 2000)
allows multiplexing and provides high specificity and accuracy.
A detailed, quantitative methylation profile of essentially any set
of CpG dinucleotides can be determined by using this assay.
Fifteen tumor suppressor genes commonly linked to transcrip-
tional silencing in various human cancers were chosen and the
methylation status of their promoter regions evaluated (http://
www.mdanderson.org/departments/methylation). Up to six
CpG dinucleotides per promoter regions
were investigated and a total of 75 CpG
dinucleotides queried per sample.
Results
Determining assay specificity and
quantitative accuracy of
bisulfite-PCR/LDR/Universal Array
The general design of the assay is illus-
trated in Figure 1. Genomic DNAs were
treated with sodium bisulfite to convert
unmethylated, but not methylated, cy-
tosines into uracils. We have modified
the standard bisulfite protocol to ensure
a thorough deamination of unmethyl-
ated cytosines and increase DNA recov-
ery (Boyd and Zon 2004). Gene-specific
PCR primers bearing 5⬘universal tails
were designed to flank each promoter re-
gion. A second, universal PCR step al-
lows approximately equal fragment am-
plification of all sequences amplified in
the primary PCR. Since PCR is not the
final readout in this assay, primer design
is flexible and less constrained by se-
quence context and is independent of
CpG dinucleotide methylation status.
Three LDR primers were used to deter-
mine the methylation status of each
CpG dinucleotide. LDR primers were de-
signed to tolerate mismatched base pairs
at internal CpG sites and allow hybrid-
ization to fully and partially methylated
sequences, as well as unmethylated se-
quences. A high fidelity Tth ligase (Luo
et al. 1996), only ligates the upstream
(discriminating) and downstream (com-
mon) primers when the 3⬘discriminat-
ing nucleotide at the junction is comple-
mentary to the DNA template. This
feature allows accurate, quantitative de-
tection of targeted CpG dinucleotides re-
gardless of the presence of internal CpG
dinucleotides within the primer se-
quences. For example (Fig. 1), at the
methylated CpG site 1, only Cy3-C–
labeled ligation products are formed,
whereas only Cy5-T–labeled ligation
Figure 1. Schematic diagram of the assay. Two hypothetical CpG dinucleotide sites 1 and 2 are
designated as methylated and unmethylated, respectively. Sodium bisulfite converts unmethylated,
but not methylated, cytosines into uracils. This conversion renders the genomic DNAs into two asym-
metrical, noncomplementary strands, and only one designated bisulfite-modified strand (highlighted
in orange) is amplified and analyzed. In the initial amplification, PCR primers are designed with a
gene-specific 3⬘portion and an upstream universal sequence (highlighted in black). This universal
sequence is used as a PCR primer in the subsequent PCR to simultaneously amplify all the primary
amplicons (for ease of illustration, only one amplicon is shown). LDR is performed in a multiplex fashion
with three primers (two discriminating and one common primers) interrogating each of the selected
CpG sites. The discriminating primers contain a 5⬘fluorescent label and a 3⬘discriminating nucleotide
to determine either methylated (with 5⬘Cy3 and 3⬘C) or unmethylated (with 5⬘Cy5 and 3⬘T)
cytosines. The common primers bear a 5⬘phosphate and a 3⬘unique zip-code complement sequence
(e.g., cZip1 and cZip2). Ligation occurs only if the nucleotides at the ligation junction are perfectly
base-paired with a complementary template and the ligation products are captured onto a Universal
microarray with prespotted zip-codes (addresses). For example, address Zip1 identifies methylated
cytosine in methylation site 1, and address Zip2 identifies unmethylated cytosine in methylation site 2.
Three Universal microarray addresses are assigned for each promoter region, and each address is
double-spotted to ensure the quality of array fabrication and oligonucleotide hybridization efficiency.
Multiplex detection of CpG methylation in tumors
Genome Research 283
www.genome.org
products are formed at the unmethylated CpG site 2. Unique
complementary zip-code sequences on the 3⬘ends of common
primers guide LDR products to their corresponding zip-codes on
a Universal Array (Gerry et al. 1999; Favis et al. 2000). The zip-
codes are unique sequences designed with a constant T
m
and
have no homology to either the target sequence or to other se-
quences in the genome. This design eliminates false signals due
to nonspecific binding and mismatch hybridizations.
The assay was validated on genomic DNAs extracted from
two commonly used colorectal (HCT15) and prostate cancer
(LNCaP) cell lines. Promoter regions chosen in this study have
included 15 tumor suppressors (CDKN2B [formerly known as
p15
INK4b
], CDKN2A [formerly known as p16
INK4a
], CDKN2D [for-
merly known as p14
ARF
], CDKN1A [formerly known as p21
CIP
],
CDKN1B [formerly known as p27
KIP
], TP53 [formerly known as
p53], BRCA1,TIMP3,APC,RASSF1,CDH1 [formerly known as
ECAD], MGMT,DAPK1 [formerly known as DAPK], GSTP1, and
RARB [formerly known as RAR

]) and one hemi-methylated im-
printed gene (SNRPN, as an internal control). The methylation
profiles of the candidate promoter regions were determined by
bisulfite sequencing, which revealed CDKN2B,CDKN2A,
CDKN2D,CDKN1A,CDKN1B,TP53,BRCA1,DAPK1,CDH1,
MGMT, and TIMP3 were unmethylated in LNCaP, while
CDKN2A,CDKN2D,MGMT,RARB, and RASSF1 were methylated
in HCT15 among 15 tumor suppressor genes. The initial LDR/
Universal Array assay was designed to evaluate methylation sta-
tus of three CpG sites per promoter region. LDR primers detect-
ing methylated and unmethylated cytosines were validated by
using in vitro methylated (SssI methylase) and untreated normal
human lymphocyte genomic DNAs, respectively (data not
shown). Following bisulfite treatment, genomic DNA of each cell
line sample was multiplex PCR amplified, and the pooled PCR
products were subjected to LDR/Universal Array analysis (Fig.
2A). We tested the assay specificity by using LNCaP DNA and
subsets of LDR primers that detect only unmethylated cytosines
(Fig. 2B; data not shown). The capture of Cy5 fluorescence signals
only at the designated zip-code addresses for each LDR primer set
indicated that LDR/Universal Array did not generate nonspecific
ligation products and that mismatch hybridization was absent.
To further demonstrate the assay’s accuracy, LDR primers that
detected only methylated cytosines were used to investigate a
total of 48 CpG sites simultaneously for each cell line (Fig. 2C).
Our data are consistent with the bisulfite sequencing results, in-
dicating an accurate methylation profile was obtained. Different
levels of fluorescence intensity were observed at several zip-code
addresses. These variations suggested that the targeted CpG di-
nucleotides may have different methylation levels within the
same promoter regions (e.g., RASSF1 and TIMP3).
To determine if the assay could be quantitative, genomic
DNA from HCT15 (carrying methylated CpG dinucleotides) was
mixed with normal human lymphocytes (carrying unmethylated
alleles), such that the test samples contained 0%, 20%, 40%,
60%, 80%, and 100% of HCT15 DNA. These mixtures were sub-
jected to bisulfite-PCR/LDR/Universal Array analysis (Fig. 3A;
data not shown). The average fluorescence intensity representing
either methylated (Cy3) or unmethylated (Cy5) alleles from each
double-spotted zip-code address was used to calculate the meth-
ylation ratio of Cy3/(Cy3 + Cy5). Each experiment was repeated
at least twice and produced consistent results. Most of the CpG
dinucleotides we evaluated have R
2
values between 0.98 and
0.89. Those CpG sites that gave lower R
2
values are likely due to
inefficient competition between LDR primers targeted toward
Figure 2. Representative bisulfite-PCR/LDR/Universal Array analysis of
16 promoter regions of cell lines HCT15 and LNCaP. (A) For the ease of
demonstration, either five or six promoter regions were amplified in one
PCR, and a total of 16 genes were simultaneously analyzed. The gene
names and the corresponded PCR fragments are as follows: (lane 1)
CDKN2B (317 bp), CDKN2A (363 bp), CDKN1A (391 bp), CDKN1B (426
bp), SNRPN (442 bp), and BRCA1 (459 bp); (lane 2)CDKN2D (346 bp),
TIMP3 (404 bp), APC (433 bp), RASSF1 (474 bp), and CDH1 (513 bp); and
(lane 3)MGMT (362 bp), TP53 (418 bp), DAPK1 (434 bp), GSTP1 (507
bp), and RARB (522 bp). (B) LDR/Universal Array analysis of the unmeth-
ylated cytosines in LNCaP amplicons. All PCR products were pooled as
LDR templates, but only selected LDR primers were used in each reaction
(LDR set1: SNRPN,CDKN2B,CDKN1A; LDR set2: CDKN2A,TP53,BRCA1).
The subset of promoter regions that were interrogated in each LDR are
depicted in the diagram (green circles) under each array image. The
Cy5-labeled LDR products (false color green, designed for unmethylated
cytosines) were captured on Universal Arrays. (C) All PCR products of each
sample were pooled and subjected to LDR/Universal Array assay. Only
Cy3-labeled LDR primers (false color red) were used in this assay to detect
methylated cytosines. The diagram under each array image depicts the
correlated zip-codes (circles) that were assigned to represent the CpG
methylation status in each of the 16 promoter regions. Each zip-code was
double-spotted on the array to ensure fabrication quality. Red and empty
circles represent methylated and unmethylated CpG sites, respectively.
Pink circles represent those CpG dinucleotides that have lower level of
methylation. The PCR and LDR primer sequences and their concentra-
tions used in these experiments were listed in the Supplemental Tables 1,
2, and 3.
Cheng et al.
284 Genome Research
www.genome.org
unmethylated and methylated alleles and can be resolved by re-
designing the LDR primers to have a higher melting temperature.
Our analysis confirmed the different percentage of methylation
at each CpG dinucleotide and suggested that methylation level is
not 100% at each CpG site in tumor cell line DNA. For example,
for HCT15, a medium level of methylation was observed at the
first CpG dinucleotide and <10% methylation at the other two
CpG dinucleotides in TIMP3. By comparing the ratio of (meth-
ylated): (methylated + unmethylated) DNA in different cell lines,
we could extrapolate the CpG methylation level at a given posi-
tion. Alternatively, SssI could be used to methylate DNA in vitro
to completion to generate standard curves for calibration (data
not shown). We tested assay sensitivity by mixing tumor cell line
DNA with normal human lymphocyte DNAs. A preliminary
study suggested that the unbiased PCR primer design was suffi-
cient to detect the presence of methylated alleles, even when
diluted down to 1% in unmethylated alleles (Supplemental Fig.
1). Nevertheless, in the colorectal cancer study, we currently use
10%–15% as a cut-off for our scoring criteria. Overall, our data
demonstrate that bisulfite-PCR/LDR/Universal Array approach is
a quantitative and sensitive method for the measurement of DNA
methylation.
Comparison of the method with a pyrosequencing assay
using clinical tumor samples
To further evaluate the quantitative accuracy and clinical utility,
we analyzed MGMT methylation level at three CpG dinucleotides
per sample on a total of 15 colorectal tumors, using both our
assay as well as by pyrosequencing. Genomic DNAs of these tu-
mors were bisulfite treated, amplified by using multiplex PCR,
and analyzed by LDR/Universal Array. Alternatively, MGMT was
uniplex amplified from the same tumor DNAs for pyrosequenc-
ing. Three sequencing primers were designed to investigate the
methylation level of the same CpG dinucleotides that were stud-
ied by LDR/Universal Array. The comparison revealed a high cor-
relation between these two methods (Fig. 3B). The few samples
where results varied may be due to variations in array fabrication,
differences in efficiency of multiplex versus uniplex PCR ampli-
fication, and, most likely, the pyrosequencing primer design.
Two of the sequencing primers contained two and three CpG
sites, respectively, and generated a biphasic plot while establish-
ing standard curves. This may have led to bias in the quantifica-
tion seen by pyrosequencing. Nevertheless, this highly signifi-
cant correlation suggests bisulfite/PCR/LDR/Universal Array is an
accurate method for quantitative analysis of heterogeneous clini-
cal samples.
Application of the method to cancer cell lines
and heterogeneous clinical samples
We performed a blinded study to determine methylation profiles
of several cancer cell lines (five colorectal, one breast, and one
prostate) with this approach (Fig. 4). Three CpG sites per pro-
moter region were evaluated. Methylation was not observed in
the promoter regions of CDKN2B,CDKN1A,CDKN1B,TP53, and
BRCA1 among all the tested cell line DNAs. These results sug-
gested that promoter methylation is not a random event. All cell
lines under the blinded study have very distinct methylation
profiles among the 15 tumor suppression genes except for two
pairs of cell lines: HCT15/DLD-1 and HT29/WiDr, which share
almost identical promoter methylation patterns. Each pair of cell
lines HCT15/DLD-1 and HT29/WiDr was established from the
same colon carcinomas (additional cell line information is de-
scribed in American Type Culture Collection [ATCC] Web site
http://www.atcc.org) (Chen et al. 1995). This observation sug-
gested that colorectal cancer cell lines derived from the same
tumor have essentially identical methylation profiles that are
Figure 3. The quantification curves of the assay. Genomic DNAs of
HCT15 and normal human lymphocytes were mixed in 0%, 20%, 40%,
60%, 80%, and 100% ratios and subjected to bisulfite-PCR/LDR/
Universal Array analysis. (A) Representative array images are shown
scanned in both Cy3 and Cy5 channels. False color red (Cy3) and green
(Cy5) represent the methylated and unmethylated alleles of CpG di-
nucleotides, respectively. Color composites of the two channels reflect
the methylation levels. Each zip-code was double-spotted on the array to
ensure fabrication quality. MGMT and TIMP3 were used as examples to
show the assay linearity measured at individual CpG dinucleotides. The
plotted value at y-axis represents the fluorescence intensity Cy3/
(Cy3 + Cy5) ratio. The value at x-axis represents the percentage of HCT15
mixed with normal human lymphocyte genomic DNAs. The R
2
and P-
values of each linear regression line were calculated, although the lines
were omitted in the plots for visual clarity. Nearly no methylation was
observed at two of the CpG sites of TIMP3 resulting in poor statistical
correlation (circles: R
2
= 0.81, P= 0.02; crosses: R
2
= 0.08, P= 0.01). The
experiments were repeated three times with different sample prepara-
tions and array hybridizations. The PCR and LDR primer sequences and
their concentrations used in these experiments are listed in Supplemental
Tables 3 and 4. (B) Comparison of the MGMT methylation level from 15
colorectal carcinomas using pyrosequencing technology and bisulfite/
PCR/LDR/Universal Array. Three CpG sites were evaluated per DNA
sample. The plotted value at y-axis represents the percentage of meth-
ylated cytosines in the tumor samples as obtained from pyrosequencing.
The value at x-axis represents the ratio of fluorescence intensities Cy3/
(Cy3 + Cy5). The mixed genomic DNAs of HCT15 and normal human
lymphocytes shown in Awere included as controls.
Multiplex detection of CpG methylation in tumors
Genome Research 285
www.genome.org
distinct from cell lines derived from a different patient. To ensure
the accuracy of scoring a candidate gene promoter as hypermeth-
ylated, the methylation status of three additional CpG sites per
promoter was examined (examples are shown in Supplemental
Fig. 2) for the 10 genes that were found to be methylated in the
cell line study (CDKN2A,CDKN2D,TIMP3,APC,RASSF1,CDH1,
MGMT,DAPK1,GSTP1, and RARB). Within each promoter, the
six CpG sites examined were dispersed evenly throughout the
amplified genomic sequence. Some of the CpG sites investigated
were within 50 bases and resulted in the design of overlapping
LDR primers. Nevertheless, we have found that LDR efficiency
was not altered by the overlapping primer design (data not
shown). Thus, the cytosines of all 75 CpG sites for a given sample
were interrogated independently at the same time.
In a pilot study, we have profiled methylation status of 96
colorectal tumor samples and 73 matched normal tissues (se-
lected data are shown in Fig 4; Supplemental Fig. 2). Our prelimi-
nary analysis shows a rich variety of methylation profiles. Seven
promoter regions (CDKN2A,CDKN2D,APC,MGMT,RARB,
RASSF1, and TIMP3) showed statistically significant increased
methylation in tumor compared to normal tissues. Promoter re-
gions of CDKN1A,CDKN1B,TP53, and BRCA1 revealed little or
no methylation. Thus, methylation profiles of clinical samples
were similar to those observed in the colorectal cancer cell lines.
One of the CpG dinucleotides of CDKN2B was frequently meth-
ylated in the tumor samples. The biological significance of this
methylation remains to be investigated by examining additional
CpG dinucleotides to determine CDKN2B promoter hypermeth-
ylation status. Nevertheless, these results reaffirm that methyl-
ation in colorectal cancer is gene specific and nonrandom. Ad-
ditional tumor samples will be examined to provide sufficient
power to determine the correlation between promoter methyl-
ation and the tissue pathological-clinical information.
Discussion
We have presented an accurate and quantitative assay that pro-
vides a representational CpG methylation profile from colorectal
cancer samples, where stromal cell infiltration is often seen. This
assay simultaneously determines the DNA methylation status of
multiple gene promoters, querying a total of 75 CpG dinucleo-
tide sites per sample. Genomic DNAs isolated from seven cancer
cell lines and 169 colon tumor samples were tested. Cell lines
derived from the same tumor have essentially identical methyl-
ation profiles. The percentage of CpG dinucleotide methylation
of MGMT promoter in clinical samples was compared by our
assay with those derived from pyrosequencing and resulted in a
high correlation.
The candidate promoter regions, involving genes in cell
cycle regulation, DNA repair, and tumor metastasis and invasion,
were previously reported in the literature as associated with ab-
normal gene silencing in tumors or cancer cell lines. Although
the percentage of methylated promoters in a cohort may vary
due to the sample source and the assays used in determining
DNA methylation status, our preliminary analysis of the colorec-
tal tumor methylation profile gave consistent results as those
published previously (Esteller et al. 2001). For example, in agree-
ment with our own findings, it has been shown that there is
Figure 4. Selected examples of DNA methylation profiles of cancer cell lines and clinical samples. Five colorectal (HCT15, DLD-1, HT29, WiDr, and
SW620), one breast (MCF7), one prostate (LNCaP) cancer cell lines, 20 primary colorectal cancer (T series), and 10 adjacent normal tissues (N series)
were analyzed. Six CpG sites per promoter region were analyzed in each sample except CDKN2B,CDKN1A,CDKN1B,TP53, and BRCA1. Standard curves
were not established for the genes mentioned above since hypermethylation is not reported in the literature or observed in our clinical sample study.
The standard curves should be established when applying this assay to other tumor types such as breast cancer. Around 10%–15% established CpG
methylation standard curves gave lower R
2
values (between low 80s and 70s), including CDKN2A (CpG㛭3 and CpG㛭6), CDKN2D (CpG㛭1), GSTP1
(CpG㛭3), DAPK1 (CpG㛭3), RASSF1 (CpG㛭4), and TIMP3 (CpG㛭2). Only four CpG sites of RARB were shown due to defects on zip-code addresses during
array fabrication. The color scale represents the percentage of methylation levels determined from the standard curves at each CpG dinucleotide. Notice
that for the cell lines HCT15/DLD-1 and HT29/WiDr, each pair is derived from a same tissue origin reflected in their identical methylation patterns. The
PCR and LDR primer sequences and their concentrations used in these experiments are listed in the Supplemental Tables 3, 4, and 5.
Cheng et al.
286 Genome Research
www.genome.org
essentially no hypermethylation in CDKN2B,TP53,BRCA1, and
GSTP1 promoter regions, while CDKN2D,CDKN2A,MGMT, and
APC were methylated to a higher level among all the samples
tested (Esteller et al., 2001).
Our assay allows virtually any CpG site in the promoter and
first intron regions for more than a dozen genes to be analyzed
simultaneously. There is increasing evidence that genes such as
MLH1 and RASSF1A exhibit an increasing gradient of methyl-
ation from the promoter proximal region to the first exon (Deng
et al. 1999; Yan et al. 2003). To avoid the bias of scoring a hyper
(or hypo)-methylated promoter and linking it to its disease state,
multiple CpG sites across a larger window of the genomic region
should be investigated in each assay. Studies done with MSP- and
restriction enzyme–based methods only reveal the methylation
pattern of small sequence regions; additional sequence contexts
may be needed to sufficiently determine the promoter methyl-
ation status. Our approach interrogated multiple CpG dinucleo-
tides that were evenly distributed over 300–500 bases. For paraf-
fin-embedded tissue, two or more adjacent shorter PCR ampli-
cons should be designed to overcome the poor amplification
typically observed in these types of samples (Supplemental Fig. 3).
Moreover, the LDR efficiency was not affected, even when some of
the LDR primers were designed with overlapping sequences. This
technique provides a detailed mapping of the methylation pro-
file in each promoter CpG island locus that may correlate with
transcriptional silencing during disease progression.
The bisulfite-PCR/LDR/Universal Array approach provides
several advantages over existing methods for the analysis of DNA
methylation patterns. First, there are two levels of specificity fa-
cilitated by gene-specific primers, initially during PCR and sub-
sequently during LDR. Given the numerous duplications anno-
tated and suspected in the human genome, this approach en-
hances the ability to reliably target the CpG islands in the locus
of interest. Second, LDR allows accurate identification of low
abundance nucleotide alterations with a remarkable accuracy
due to the high fidelity of Tth ligase (Luo et al. 1996). This unique
feature eliminates the concern of mismatch hybridization due to
partial methylation of internal CpGs of the LDR primer se-
quences and allows LDR primers to be placed at essentially any
CpG dinucleotide of interest. Third, the unique zip-code se-
quences were designed with similar T
m
across the platform and
have no sequence homology in the human genome. This feature
allows only ligated LDR products to be captured, thus avoiding
background signals. As the Universal Array can be easily ex-
panded (Gerry et al. 1999), additional CpG sites or genes can be
evaluated in a single assay. Fourth, this assay has the potential to
detect low abundance methylated alleles. By redesigning the
multiplex gene-specific PCR primers to be methyl specific, a de-
tection of at least 0.1% was achieved, albeit with reduced quan-
titative dynamic range (Supplemental Fig. 1). An inherent prob-
lem with many DNA amplification techniques is that greater de-
tection sensitivity comes at the cost of increased false positives.
Although methylation assays relying solely on PCR as a readout
tool offer sensitive detection, they are prone to false positives
resulting from the AT richness or incomplete deamination of
bisulfite-treated DNAs. Consequently, such assays would be lim-
ited in their multiplex capability. Our assay confirmed promoter
methylation status via six CpG sites within a PCR fragment; this
approach offers high specificity and accuracy while avoiding
false positives. Moreover, each module of the PCR/LDR/Universal
microarray approach is ideal for multiplexing and can be auto-
mated by using a liquid handing system to increase throughput.
A recent publication has reported the detection of dozens to hun-
dreds of possible mutations by using multiplex PCR/LDR in a
single-tube format (Favis et al. 2004). The capture of multiplex
LDR products onto an array format provides an efficient “modu-
lar”readout and substantially increases assay throughput. Uni-
versal microarray experiments in our laboratory are now per-
formed in an array-of-arrays format, where 64 array hybridiza-
tions are carried out simultaneously (data not shown). This array-
of-arrays approach drastically reduces the cost of array
fabrication, minimizes the variation during hybridization, and
increases throughput.
In summary, we present a robust and accurate method that
determines cytosine methylation at any selected set of CpG di-
nucleotides in the genome. Importantly, this new method allows
the evaluation of methylation level at individual CpG sites.
Quantitative values for this parameter may facilitate stratifica-
tion of tumors, since based on the degree of methylation, it may
be possible to estimate disease progression. In addition, the abil-
ity to quantify methylation at specific sites in advanced tumors
may provide information on tumor heterogeneity. These data
will enable clinical decisions related to individualized treatment
strategies. Our goal is to expand this prototype assay into a fo-
cused array platform that investigates 30–50 frequently methyl-
ated tumor suppressor promoters observed in several different
human carcinomas. Around 10–15 individual CpG dinucleotides
evenly distributed over a larger window of each promoter region
will be interrogated. We anticipate that such a focused platform
will facilitate the development of DNA-based molecular markers
for disease diagnosis and prognosis and will be suitable for rou-
tine clinical use.
Methods
Cell line culture, tumor samples, and DNA extraction
Normal human lymphocyte genomic DNA was purchased from
Roche. Colorectal, breast, and prostate cancer cell lines were ob-
tained from American Type Culture Collection and cultured un-
der the ATCC-recommended media conditions. Fresh frozen pri-
mary colorectal adenocarcinomas were obtained from Memorial
Sloan Kettering Cancer Center under Institutional Review Board
(IRB)-approved protocols. Genomic DNAs were extracted by us-
ing the DNeasy Tissue Kit (Qiagen) according to the manufactur-
er’s guidelines.
Sodium bisulfite treatment of genomic DNAs
Typically, 1 µg genomic DNA was denatured in 40 µL of 0.2 N
NaOH by incubating for 10 min at 37°C before addition of 30 µL
of freshly prepared 10 mM hydroquinone and 520 µL of 3 M
sodium bisulfite. The reaction was incubated for 20 min at 50°C,
15 sec at 85°C for 48 cycles (16 h). The DNA clean-up procedure
was as follows: (1) the total reaction volume (∼600 µL) was trans-
ferred to a Microcon NCO30 filter (Millipore) and centrifuged at
13000gfor 16 min; (2) 500 µL deionized H
2
O were added to the
upper chamber, centrifuged at 13,000gfor 7 min, the filtrate
discarded, and the wash repeated twice; (3) 500 µL of 0.3M NaOH
were added to the upper chamber, incubated for 5 min at room
temperature and then centrifuged at 13,000gfor 8 min; (4) 500
µL deionized H
2
O were added to the upper chamber, centrifuged
at 13,000gfor 8 min, the filtrate discarded, and the wash re-
peated; and (5) the filter was inverted to collect the bisulfite-
converted DNA. An appropriate volume of water (if needed) was
Multiplex detection of CpG methylation in tumors
Genome Research 287
www.genome.org
used to rinse the upper chamber to recover DNA in a final volume
of 20 µL.
Multiplex PCR amplification
The multiplex PCR consists of two stages. PCR stage I (12.5 µL)
contained 1.5 µL bisulfite-modified DNA, 400 µM of each dNTP,
1⳯AmpliTaq Gold PCR buffer, 4 mM MgCl
2
, and 1.25 U
AmpliTaq Gold polymerase (Applied Biosystems). Mineral oil was
added prior to thermal cycling. The PCR stage I conditions were
as follows: 10 min at 95°C; 15 cycles of 30 sec at 94°C, 1 min at
60°C, and 1 min at 72 °C; followed by a final extension step of 5
min at 72°C. PCR stage II (12.5 µL) contained 400 µM of each
dNTP, 1⳯AmpliTaq Gold PCR buffer, 4 mM MgCl
2
, 12.5 pmol
universal primer (UniB2, see Supplemental Table 1), and 1.25 U
AmpliTaq Gold polymerase. The 12.5 µL reaction mixture was
added through the mineral oil to the completed stage I PCR. The
PCR stage II conditions were as follows: 10 min at 95°C; 30 cycles
of 30 sec at 94°C, 1 min at 55°C, 1 min at 72°C; followed by a
final extension step of 5 min at 72°C. Taq DNA polymerase was
inactivated by adding 1.25 µL Proteinase K (20 mg/mL, Qiagen)
to the completed stage II PCR, incubating for 10 min at 70°C and
15 min at 90°C. Before pooling the PCR products for LDR assay,
the presence of amplicons was confirmed by electrophoresis on a
3% agarose gel.
LDR, Universal Array hybridization, and data analyses
A typical LDR (20 µL) contained 20 mM Tris-HCl (pH 7.6), 10
mM MgCl
2
, 100 mM KCl, 10 mM DTT, 1 mM NAD, 25 fmol
wild-type Tth ligase (Zirvi et al. 1999), 500 fmol of each LDR
primer, and 5–10 ng of each PCR amplicon. The LDR conditions
were as follows: 3 min at 95°C; 25 cycles of 30 sec at 95°C and 4
min at 60°C. The LDR reaction was diluted with an equal volume
of 2⳯hybridization buffer (600 mM MES at pH 6.0, 20 mM
MgCl
2
, 0.2% SDS), denatured for 3 min at 95°C and plunged on
ice. The Universal Arrays were pre-equilibrated with 1⳯hybrid-
ization buffer at room temperature for at least 15 min. Coverwells
(Grace Bio-Labs) were attached to arrays and filled with 40 µL
denatured LDR reactions. The assembled arrays were incubated
in a rotating hybridization oven for 60 min at 65°C. After hy-
bridization, the arrays were washed in 300 mM Bicine (pH 8.0),
0.1% SDS for 10 min at 60°C. An updated version with 384 ad-
dresses will accommodate all the LDR products. Each array was
scanned by using a Perkin Elmer ProScanArray under the same
laser power and PMT within the linear dynamic range. The Cy3
and Cy5 dye bias was determined by measuring the fluorescence
intensity of an equal mole of Cy3- and Cy5-labeled LDR primers
manually deposited on a slide surface. This fluorescence intensity
ratio (W = I
Cy3
/I
Cy5
) was used to normalize the label bias when
calculating the methylation ratio Cy3/(Cy3 + Cy5). MetaMorph
Imaging System (Universal Imaging) was used to create images
depicting the Cy3 (red) and the Cy5 (green).
Oligonucleotide design and synthesis
Oligonucleotides were obtained from IDT or synthesized in-
house on an ABI 394 DNA Synthesizer (PE Biosystems) using
standard phosphoramidite chemistry (Khanna et al. 1999).
Spacer phosphoramidite C18, 3⬘-amino-modifier C3 CPG, C3
spacer, and Cy3, Cy5, and standard phosphoramidites were pur-
chased from Glen Research. All other reagents were purchased
from PE Biosystems. The zip-code oligonucleotides were synthe-
sized on a 3⬘amino modifier C3 column with a spacer C18 in-
serted before the first base. The common LDR primers were syn-
thesized with 5⬘phosphates and 3⬘C3 spacers as blocking
groups. Oligonucleotides with cyanine labels were cleaved from
the CPG supports and deprotected according to manufacturer’s
recommendations. Both labeled and unlabeled LDR oligonucleo-
tides were purified and desalted on SuperPure columns (Bio-
search Technologies) according to the manufacturer’s instruc-
tions, then spin-dried (Speed-Vac) and stored at ⳮ20°C. For
those primers that inevitably covered CpG dinucleotides in the
body of their sequences, the nucleotides that base paired with
cytosines in CpG dinucleotide were synthesized in two ways.
One was to use nucleotide analogs dK or dP in the primers’syn-
theses. The pyrimidine derivative dP base pairs with either A or
G, while the purine derivative dK base pairs with either C or T at
similar efficiency. Alternatively, to reduce the cost of primer syn-
thesis, those nucleotide positions with analogs dK or dP incor-
porated were substituted by nucleotides dG or dC, respectively.
For example, the substituted nucleotide dG in a PCR primer
formed either Watson-Crick base pair with C (methylated) or
wobble base pair with U (unmethylated) on the bisulfite-
modified DNA template.
Universal Array fabrication
Polymer-coated slides were fabricated as previously described
(Gerry et al. 1999; Favis et al. 2000) or were purchased (CodeLink
slides) from Amersham Biosciences. Universal Arrays were spot-
ted by using a Pixsys5500 robot with a quill-type spotter in a
controlled humidity chamber (Cartesian Technologies). Zip-code
oligonucleotides each with a unique 24-mer sequence were pre-
pared by mixing 5 µL of 1000 µM stock oligonucleotides with 5
µL of 0.4 M K
2
HPO
4
/KH
2
PO
4
(pH 8.5) in 384 conical well spot-
ting plates. Arrays were printed under relative humidity 60%–
70%. To ensure that all the zip-codes were spotted without cross-
contamination during array fabrication, one out of 10 slides on
average was subjected to quality control by hybridizing fluores-
cein-labeled zip-code complements targeting a combination of
rows or columns of zip-code addresses. A batch of fabricated ar-
rays passed the quality control only when specific fluorescein
signals were present on all the targeted rows and columns with-
out extraneous signals on the adjacent, unexpected neighboring
addresses.
Pyrosequencing
A promoter sequence of MGMT was PCR amplified by using 1 µL
bisulfite-modified DNA, 400 µM of each dNTP, 1⳯AmpliTaq
Gold PCR buffer, 4 mM MgCl
2
, 0.2 µM PCR primers (5⬘-
GGTTTTAGGAGGGGAGAGATT-3⬘and 5⬘-CCTAACCCRA
ATAACCCTTC-3⬘), and 1.25 U AmpliTaq Gold polymerase (Ap-
plied Biosystems). The PCR condition was as follows: 15 min at
94°C; 45 cycles of 15 sec at 95°C, 30 sec at 58°C, and 15 sec at
72°C; followed by a final extension step for 5 min at 72°C. Three
sequencing primers (5⬘-GTAGTAGTTTAGAGTAGGAT-3⬘,5⬘-
TTTTAGAGAGTTTTTAGGAT-3⬘and 5⬘-AAATTAAGGTATA
GAGTTTT-3⬘) were designed to determine the CpG dinucleotide
methylation levels. The primers were designed and experiments
were performed by Biotage.
Acknowledgments
We thank the Neil Bander laboratory for providing prostate can-
cer cell lines for DNA extraction; Reyna Favis, Maneesh Pingle,
and Sarah Giardina for critical reading of the manuscript; and
Jurg Ott, Sandra Barral, and Robert Westphalen for statistic analy-
sis of the clinical sample data. We also thank Reyna Favis, Nor-
man Gerry, Jianmin Huang, Brian Kirk, Maneesh Pingle, Richard
Shattock, Hanna Pincas, Manny Bacolod, Kathy Granger, and Ali
Gure for insightful discussion and technical assistance. Work in
Cheng et al.
288 Genome Research
www.genome.org
the Barany laboratory is sponsored by the National Cancer Insti-
tute (P01-CA65930) and in part by a research grant from Applied
Biosystems, for which F.B. also serves as a consultant.
References
Adorjâan, P., Distler, J., Lipscher, E., Model, F., Mèuller, J., Pelet, C.,
Braun, A., Florl, A.R., Gèutig, D., Grabs, G., et al. 2002. Tumour class
prediction and discovery by microarray-based DNA methylation
analysis. Nucleic Acids Res. 30: e21.
Balog, R.P., de Souza, Y.E., Tang, H.M., DeMasellis, G.M., Gao, B., Avila,
A., Gaban, D.J., Mittelman, D., Minna, J.D., Luebke, K.J., et al. 2002.
Parallel assessment of CpG methylation by two-color hybridization
with oligonucleotide arrays. Anal. Biochem. 309: 301–310.
Baylin, S.B. and Herman, J.G. 2000. DNA hypermethylation in
tumorigenesis: Epigenetics joins genetics. Trends Genet. 16: 168–174.
Boyd, V.L. and Zon, G. 2004. Bisulfite conversion of genomic DNA for
methylation analysis: Protocol simplification with higher recovery
applicable to limited samples and increased throughput. Anal.
Biochem. 326: 278–280.
Chen, T.R., Dorotinsky, C.S., McGuire, L.J., Macy, M.L., and Hay, R.J.
1995. DLD-1 and HCT-15 cell lines derived separately from
colorectal carcinomas have totally different chromosome changes
but the same genetic origin. Cancer Genet. Cytogenet. 81: 103–108.
Cheng, J.C., Yoo, C.B., Weisenberger, D.J., Chuang, J., Wozniak, C.,
Liang, G., Marquez, V.E., Greer, S., Orntoft, T.F., Thykjaer, T., et al.
2004. Preferential response of cancer cells to zebularine. Cancer Cell
6: 151–158.
Costello, J.F., Frèuhwald, M.C., Smiraglia, D.J., Rush, L.J., Robertson,
G.P., Gao, X., Wright, F.A., Feramisco, J.D., Peltomèaki, P., Lang,
J.C., et al. 2000. Aberrant CpG-island methylation has non-random
and tumour-type-specific patterns. Nat. Genet. 24: 132–138.
Cottrell, S.E. and Laird, P.W. 2003. Sensitive detection of DNA
methylation. Ann. N.Y. Acad. Sci. 983: 120–130.
Cui, H., Cruz-Correa, M., Giardiello, F.M., Hutcheon, D.F., Kafonek,
D.R., Brandenburg, S., Wu, Y., He, X., Powe, N.R., and Feinberg, A.P.
2003. Loss of IGF2 imprinting: A potential marker of colorectal
cancer risk. Science 299: 1753–1755.
Deng, G., Chen, A., Hong, J., Chae, H.S., and Kim, Y.S. 1999.
Methylation of CpG in a small region of the hMLH1 promoter
invariably correlates with the absence of gene expression. Cancer Res.
59: 2029–2033.
Dupont, J.M., Tost, J., Jammes, H., and Gut, I.G. 2004. De novo
quantitative bisulfite sequencing using the pyrosequencing
technology. Anal. Biochem. 333: 119–127.
Esteller, M., Corn, P.G., Baylin, S.B., and Herman, J.G. 2001. A gene
hypermethylation profile of human cancer. Cancer Res.
61: 3225–3229.
Favis, R., Day, J.P., Gerry, N.P., Phelan, C., Narod, S., and Barany, F.
2000. Universal DNA array detection of small insertions and
deletions in BRCA1 and BRCA2. Nat. Biotechnol. 18: 561–564.
Favis, R., Huang, J., Gerry, N.P., Culliford, A., Paty, P., Soussi, T., and
Barany, F. 2004. Harmonized microarray/mutation scanning analysis
of TP53 mutations in undissected colorectal tumors. Hum. Mutat.
24: 63–75.
Feinberg, A.P. 2001. Methylation meets genomics. Nat. Genet. 27: 9–10.
Feinberg, A.P. and Vogelstein, B. 1983. Hypomethylation distinguishes
genes of some human cancers from their normal counterparts.
Nature 301: 89–92.
Frommer, M., McDonald, L.E., Millar, D.S., Collis, C.M., Watt, F., Grigg,
G.W., Molloy, P.L., and Paul, C.L. 1992. A genomic sequencing
protocol that yields a positive display of 5-methylcytosine residues
in individual DNA strands. Proc. Natl. Acad. Sci. 89: 1827–1831.
Gardiner-Garden, M. and Frommer, M. 1987. CpG islands in vertebrate
genomes. J. Mol. Biol. 196: 261–282.
Gerry, N.P., Witowski, N.E., Day, J., Hammer, R.P., Barany, G., and
Barany, F. 1999. Universal DNA microarray method for multiplex
detection of low abundance point mutations. J. Mol. Biol.
292: 251–262.
Gitan, R.S., Shi, H., Chen, C.M., Yan, P.S., and Huang, T.H. 2002.
Methylation-specific oligonucleotide microarray: A new potential for
high-throughput methylation analysis. Genome Res. 12: 158–164.
Gonzalgo, M.L. and Jones, P.A. 1997. Rapid quantitation of methylation
differences at specific sites using methylation-sensitive single
nucleotide primer extension (Ms-SNuPE). Nucleic Acids Res.
25: 2529–2531.
Herman, J.G., Graff, J.R., Myèohèanen, S., Nelkin, B.D., and Baylin, S.B.
1996. Methylation-specific PCR: A novel PCR assay for methylation
status of CpG islands. Proc. Natl. Acad. Sci. 93: 9821–9826.
Hoque, M.O., Begum, S., Topaloglu, O., Jeronimo, C., Mambo, E.,
Westra, W.H., Califano, J.A., and Sidransky, D. 2004. Quantitative
detection of promoter hypermethylation of multiple genes in the
tumor, urine, and serum DNA of patients with renal cancer. Cancer
Res. 64: 5511–5517.
Huang, T.H., Perry, M.R., and Laux, D.E. 1999. Methylation profiling of
CpG islands in human breast cancer cells. Hum. Mol. Genet.
8: 459–470.
Issa, J.P. 2004. CpG island methylator phenotype in cancer. Nat. Rev.
Cancer 4: 988–993.
Jones, P.A. and Baylin, S.B. 2002. The fundamental role of epigenetic
events in cancer. Nat. Rev. Genet. 3: 415–428.
Khanna, M., Park, P., Zirvi, M., Cao, W., Picon, A., Day, J., Paty, P., and
Barany, F. 1999. Multiplex PCR/LDR for detection of K-ras mutations
in primary colon tumors. Oncogene 18: 27–38.
Laird, P.W. 1997. Oncogenic mechanisms mediated by DNA
methylation. Mol. Med. Today 3: 223–229.
———. 2003. The power and the promise of DNA methylation markers.
Nat. Rev. Cancer 3: 253–266.
Luo, J., Bergstrom, D.E., and Barany, F. 1996. Improving the fidelity of
Thermus thermophilus DNA ligase. Nucleic Acids Res. 24: 3071–3078.
Post, W.S., Goldschmidt-Clermont, P.J., Wilhide, C.C., Heldman, A.W.,
Sussman, M.S., Ouyang, P., Milliken, E.E., and Issa, J.P. 1999.
Methylation of the estrogen receptor gene is associated with aging
and atherosclerosis in the cardiovascular system. Cardiovasc. Res.
43: 985–991.
Robertson, K.D. and Wolffe, A.P. 2000. DNA methylation in health and
disease. Nat. Rev. Genet. 1: 11–19.
Toyota, M., Ahuja, N., Ohe-Toyota, M., Herman, J.G., Baylin, S.B., and
Issa, J.P. 1999. CpG island methylator phenotype in colorectal
cancer. Proc. Natl. Acad. Sci. 96: 8681–8686.
Uhlmann, K., Brinckmann, A., Toliat, M.R., Ritter, H., and Nèurnberg, P.
2002. Evaluation of a potential epigenetic biomarker by quantitative
methyl-single nucleotide polymorphism analysis. Electrophoresis
23: 4072–4079.
Warnecke, P.M. and Bestor, T.H. 2000. Cytosine methylation and
human cancer. Curr. Opin. Oncol. 12: 68–73.
Xiong, Z. and Laird, P.W. 1997. COBRA: A sensitive and quantitative
DNA methylation assay. Nucleic Acids Res. 25: 2532–2534.
Yamashita, K., Dai, T., Dai, Y., Yamamoto, F., and Perucho, M. 2003.
Genetics supersedes epigenetics in colon cancer phenotype. Cancer
Cell 4: 121–131.
Yan, P.S., Shi, H., Rahmatpanah, F., Hsiau, T.H., Hsiau, A.H., Leu, Y.W.,
Liu, J.C., and Huang, T.H. 2003. Differential distribution of DNA
methylation within the RASSF1A CpG island in breast cancer. Cancer
Res. 63: 6178–6186.
Yang, A.S., Estâecio, M.R., Doshi, K., Kondo, Y., Tajara, E.H., and Issa,
J.P. 2004. A simple method for estimating global DNA methylation
using bisulfite PCR of repetitive DNA elements. Nucleic Acids Res.
32: e38.
Zeschnigk, M., Bèohringer, S., Price, E.A., Onadim, Z., Masshèofer, L.,
and Lohmann, D.R. 2004. A novel real-time PCR assay for
quantitative analysis of methylated alleles (QAMA): Analysis of the
retinoblastoma locus. Nucleic Acids Res. 32: e125.
Zirvi, M., Bergstrom, D.E., Saurage, A.S., Hammer, R.P., and Barany, F.
1999. Improved fidelity of thermostable ligases for detection of
microsatellite repeat sequences using nucleoside analogs. Nucleic
Acids Res. 27: e41.
Received May 25, 2005; accepted in revised form October 6, 2005.
Multiplex detection of CpG methylation in tumors
Genome Research 289
www.genome.org