A DNA repeat, NBL2, is hypermethylated in some cancers but hypomethylated in others.

Abstract

Hypermethylation at certain CpG-rich promoters and hypomethylation at repeated DNA sequences are very frequently found in cancers. We provide the first report that a DNA sequence (NBL2) can be either extensively hypermethylated or hypomethylated in cancer. Previously, it was shown that NBL2, a complex tandem DNA repeat in the acrocentric chromosomes, is hypomethylated at NotI sites in >70% of neuroblastomas and hepatocellular carcinomas and in cells from ICF syndrome (DNMT3B-deficiency) patients. Unexpectedly, by Southern blot analysis of 18 ovarian carcinomas, 51 Wilms tumors, and various somatic control tissues, we found that >70% of the cancers exhibited large increases in methylation at HhaI sites in NBL2 compared with all the controls. In contrast, 17% of the carcinomas showed major decreases in methylation at HhaI and NotI sites. The intermediate levels of methylation at HhaI sites in somatic controls enabled this discovery of cancer-linked hypermethylation and hypomethylation in NBL2. In a comparison of ovarian epithelial carcinomas, low malignant potential tumors, and cystadenomas, NBL2 hypermethylation at HhaI sites was significantly related to the degree of malignancy, and hypomethylation was seen only in the carcinomas. By RT-PCR, we found NBL2 transcripts at low levels in a few cancers and undetectable in various normal tissues. In the tumors there was no association of NBL2 hypomethylation and transcription, but this may reflect NBL2's lack of identifiable promoter elements and our evidence for run-through transcription from adjacent sequences into NBL2. The propensity of NBL2 sequences to become either hypermethylated or hypomethylated in cancer suggests that these opposite epigenetic changes share an early step during carcinogenesis and that cancer-linked hypermethylation might be spontaneously reversible.

Full text

Page 1

©2005 LANDES BIOSCIENCE. DO NOT DISTRIBUTE.
This research was supported in part by grant
R01-CA81506 to M.E. from the NIH and by a
research fellowship to R.N. (10681) from the Japan
Society for the Promotion of Science by Young
Scientists.
tumors, and a wide variety of normal tissues. NBL2, a CpG-rich repeat that is specific to
humans and some other primates,29was hypomethylated at NotI sites in 83% of neurob-
lastomas and 75% of hepatocellular carcinomas26,29when compared with leukocyte DNA
[Cancer Biology & Therapy 4:4, 440-448, April 2005]; ©2005 Landes Bioscience
440
Cancer Biology & Therapy
2005; Vol. 4 Issue 4
Rie Nishiyama1,†
Lixin Qi2
Koji Tsumagari1
Karen Weissbecker1
Louis Dubeau3
Martin Champagne4
Suresh Sikka5
Hisaki Nagai6
Melanie Ehrlich1,2,*
1Human Genetics Program; 2Department of Biochemistry and Tulane Cancer
Center; 5Department of Urology; Tulane Medical School, New Orleans, Louisiana
USA
3Department of Pathology; University of Southern California; Norris
Comprehensive Cancer Center; Los Angeles, California USA
4Service D’hematologie Oncologie; Hospital Sainte-Justine; Montreal, Quebec,
Canada
6Institute for Molecular and Cellular Biology; Osaka University; Osaka, Japan
†Present address: Department of Integrated Biosciences; Graduate School of
Frontier Sciences; University of Tokyo; Kashiwanoha 5-1-5 Kashiwa, Chiba 277-8562
Japan
*Correspondence to: Melanie Ehrlich; Human Genetics Program SL31; Tulane
Medical School; 1430 Tulane Ave.; New Orleans, Louisiana 70112 USA; Tel.:
504.988.2449; Fax: 504.988.1763; Email: ehrlich@tulane.edu
Received 01/20/05; Accepted 02/21/05
Previously published online as a Cancer Biology & Therapy E-publication:
http://www.landesbioscience.com/journals/cbt/abstract.php?id=1622
KEY WORDS
cancer, DNA hypomethylation, DNA hyperme-
thylation, DNA repeats
ABBREVIATIONS
Chr1
Sat2
SCH
FISH fluorescence in situ hybridization
LMPlow malignant potential
chromosome 1
satellite 2 DNA
somatic cell hybrid
ACKNOWLEDGEMENTS
Research Paper
A DNA Repeat, NBL2, Is Hypermethylated in Some Cancers
but Hypomethylated in Others
ABSTRACT
Hypermethylation at certain CpG-rich promoters and hypomethylation at repeated
DNA sequences are very frequently found in cancers. We provide the first report that a
DNA sequence (NBL2) can be either extensively hypermethylated or hypomethylated in
cancer. Previously, it was shown that NBL2, a complex tandem DNA repeat in the acro-
centric chromosomes, is hypomethylated at NotI sites in >70% of neuroblastomas and
hepatocellular carcinomas and in cells from ICF syndrome (DNMT3B-deficiency) patients.
Unexpectedly, by Southern blot analysis of 18 ovarian carcinomas, 51 Wilms tumors,
and various somatic control tissues, we found that >70% of the cancers exhibited large
increases in methylation at HhaI sites in NBL2 compared with all the controls. In contrast,
17% of the carcinomas showed major decreases in methylation at HhaI and NotI sites.
The intermediate levels of methylation at HhaI sites in somatic controls enabled this dis-
covery of cancer-linked hypermethylation and hypomethylation in NBL2. In a comparison
of ovarian epithelial carcinomas, low malignant potential tumors, and cystadenomas,
NBL2 hypermethylation at HhaI sites was significantly related to the degree of malignan-
cy, and hypomethylation was seen only in the carcinomas. By RT-PCR, we found NBL2
transcripts at low levels in a few cancers and undetectable in various normal tissues. In
the tumors there was no association of NBL2 hypomethylation and transcription, but this
may reflect NBL2’s lack of identifiable promoter elements and our evidence for
run-through transcription from adjacent sequences into NBL2. The propensity of NBL2
sequences to become either hypermethylated or hypomethylated in cancer suggests that
these opposite epigenetic changes share an early step during carcinogenesis and that
cancer-linked hypermethylation might be spontaneously reversible.
INTRODUCTION
The first type of epigenetic change reported in human cancer was DNA hypomethyla-
tion.1-3Subsequently, the opposite type of change, cancer-linked hypermethylation of
CpG island-promoters of tumor suppressor genes, was shown to be crucial to downregu-
lating expression of many alleles not inactivated by mutation.4,5Although the functional
importance of cancer-linked DNA hypomethylation is less well understood than that of
hypermethylation,6loss of DNA methylation is frequently observed in various cancers.7-14
Downregulating DNA methylation in some model systems increases tumor formation,
while upregulating it in others does the same.15-19
We found hypomethylation in centromeric satellite αDNA or juxtacentromeric (cen-
tromere-adjacent) satellite 2 DNA (Sat2) in most examined breast adenocarcinomas,
Wilms tumors, and ovarian epithelial carcinomas.20-25Satellite hypomethylation was sig-
nificantly associated with global DNA hypomethylation. For ovarian epithelial tumors, we
also observed significant associations of hypomethylation of Sat2 with malignant poten-
tial,20tumor grade and stage, and lower patient survival rates.25Not only are the tandem
repeats of satellite DNA often hypomethylated in cancer, but also nonsatellite tandem
repeats in the vicinity of the centromere or at interstitial locations can be frequently
hypomethylated in certain types of cancer.26-29These repeats include NBL2 (DMHD-1;
CNIC; Y10752) and NBL1 (DMHA-1; HTRS; L09552), which have 1.4- and 13-kb
monomer sizes, respectively. Itano et al.26found that hypomethylation of these repeats in
hepatocellular carcinomas is significantly correlated with a poor prognosis, independently
of other markers.
In this study, we analyzed NBL2 methylation in ovarian epithelial tumors, Wilms

Page 2

or DNA from apparently normal liver adjacent to the carcino-
mas.27,29,30Surprisingly, in ovarian carcinomas and Wilms tumors
vs. many somatic control tissues, we found that hypermethylation at
HhaI sites in NBL2 predominated over hypomethylation.
MATERIALS AND METHODS
Patients and DNA samples. Primary tumor samples were obtained from
surgery patients prior to chemotherapy or radiation therapy, with the excep-
tion of two of the 51 Wilms tumors. These two were recurrent tumors in
patients treated with chemotherapy. With IRB approval, informed consent
from all patients was given or unlinked samples were used. The ICF cell line
was described (ICF B31). Control somatic tissues were autopsy specimens of
trauma victims and normal sperm. DNA was purified by standard methods32
with a 0.04 M-dithiothreitol treatment during the proteinase K incubation
for semen samples.
Southern blot hybridization. DNA (1.5 µg) was digested overnight with
15–30 units of restriction endonuclease under standard conditions (New
England Biolabs, Beverly, MA). Immediately after mixing, 10 µl from the
60-µl solution was added to 2 µl containing 0.1 µg of phage λDNA,
p13E-11,33or pDMHD-127for incubation in parallel to the test sample.
Complete digestion was obtained for all internal controls. The NBL2 probe,
a 1.4-kb pDMHD-127NotI fragment, was labeled with [α-32P]dCTP by
random priming. After UV-nicking of the ethidium bromide-stained 1%
agarose gel, blot-hybridization (Zeta-GT membrane, Bio-Rad, Richmond
CA) was done at 42˚C in 120 mM sodium phosphate, pH 7.2, 250 mM
NaCl, 7% SDS, 50% formamide, and 100 µg of heat-denatured salmon
sperm DNA/ml. Washing was at 42˚C with 2 x SSC (0.3 M sodium chlo-
ride, 0.03 M sodium citrate), 1% SDS, 30% formamide, and then at 42˚C
with 0.2 x SSC, 1% SDS and 30% formamide. By phosphorimaging, we
quantitated the hybridization signal from HhaI fragments of the following
size ranges: a, >4 kb; b, 2.8 kb; c, 1.4 kb; d, 0.9-1.3 kb; and e, 0.3–0.6 kb.
The signal in (d + e)/(b + c) or a/(b + c) for the test sample was divided by
the analogous average ratio from normal somatic control tissues in the same
blot to give a normalized ratio for hypomethylation and hypermethylation
scores, respectively. Scores for methylation (in boldface) were assigned as
follows: for hypomethylation, normalized (d + e)/(b + c) ratio >4.5, -3;
3.5–4.5, -2; 2.5–3.4, -1; for hypermethylation, normalized a/(b + c) ratio
>10.0, +3; 5.0–10.0, +2; 2.5–4.9, +1; both normalized ratios <2.5, 0.
Scoring of hypomethylation at NotI sites was estimated from band patterns.
Southern blot analysis and quantitation of satellite DNA hypomethylation
was as previously described25except that we used a hypomethylation scale
of 0 to -3.
RT-PCR. Total RNA from ~10 mg of frozen tissue or 5 x 106cultured
cells was isolated (Trizol, Invitrogen, Carlsbad, CA; or RNeasy kit, Qiagen,
Valencia CA). RNA (3 µg) was treated with three units of DNase I
(Amplification Grade, Invitrogen) for 45 min at room temperature. After
enzyme inactivation, cDNA was synthesized with M-MLV reverse tran-
scriptase (Invitrogen) using random hexamers or oligo(dT) in a 40-µl
mixture. For semi-quantitative RT-PCR of NBL2 sequences, 1 µl of cDNA
www.landesbioscience.com
Cancer Biology & Therapy
441
Hyper- and Hypomethylation of a DNA Repeat in Cancer
Figure 1. Map and chromosomal distribution of NBL2. (A) NBL2 restriction map showing sites present in >60% of the copies of the full-length repeats in
BAC clone AC0128692, Y10752, and U59100. (B) Electrophoresed PCR products from a panel of monochromosomal human-rodent SCH DNAs and
human or murine genomic DNA using NBL2 primers F2 and R1. (C) Diagram of PCR products using the SCH DNA panel with primers from NBL2 and adjacent
sequences to the NBL2 array in BAC clone AC0128692. The PCR products are shown mapped to this BAC clone, which contains 20 full-length and two
partial copies of NBL2 with >90% homology to one another and to Y10752 and U59100.
A
B
C

Page 3

solution was amplified (Hotstart Taq Mix; Qiagen) with the following
NBL2 primers (0.5 µM; all primers written 5' to 3'): F4, GGCTGATGGAT-
GAGTGAA, and R4, GCTGTGTTCTCTGGACTCC. GAPDH reference
primers were TTCCAGGAGCGAGATCCCT and CACCCATGACGAA-
CATGGG. PCR conditions were 95˚C for 15 min; 30 or 35 cycles of 95˚C
for 15 sec, 60˚C for 15 sec, 72˚C for 30 sec; and then 72˚C for 5 min.
Real-time PCR (iCycler; Bio-Rad, Richmond, CA; SYBR green PCR Master
Mix, Applied Biosystems, Foster City, CA) was done similarly but with
melting curve determination and a standard curve for relative quantitiation
as described previously.34Semi-quantitative RT-PCR of NBL2 and neigh-
boring sequences with various other primers was done as above except that
40 cycles with the 72˚C incubation for 1–1.5 min were used and the
following oligonucleotides primed the reaction from within NBL2 (boldface
name) or from sequences adjacent to NBL2 from BAC clone AC018692
(Fig. 1C) with the approximate sizes of the products in kb given in paren-
thesis: F5, CTTGGGTATCGAGGCTCTGAG, and R1,35TCCACTCCT-
GACAGATAGGCTG (1.28); F6, GTCGTTCTCCTCCAGAGCAGA,
and R1 (1.02); F2, GTTTGTACCAGCCGAAGTTGC,35and R5,
CCTCGATACCCAAGCACTGC (0.99); F2 and R6, TGCTCTGGAG-
GAGAACGACC (1.25); F10 in Region A, CCTTCTCTACCTCCTCTC-
CAATG, and R9, TGAGTGAGATAGGATGGGCTC (0.51); F12,
GCTAAGGCTGTGGTGGGAAA, and R11, TTGGAGAGGAGGTAGA-
GAAGGT (both in Region A, 0.67); F6 and R8 in Region B (0.40), GAT-
GAGATTTCACTTGCTCCG; F17, ATCCGCTTTGGGTGTGTGA
and R17, CATCTTTGTTGAAGGCATTCGG (both in Region B, 0.47);
or F12 and R9, followed by F11 in Region A, GAGTAAACTGAAA-
GCAGGCAGA, and R9 for semi-nested PCR.
PCR with somatic cell hybrids (SCHs). A panel of monochromosomal
human-rodent SCH DNAs (Coriell Institute Panel 2, v. 3; 15 ng each) was
amplified with the following primers to give products with the following
approximate sizes in kb given in parenthesis: F2 and R1 (0.85); F9 in Region A,
CCTTACTAGCTTCTGTGGGCA and R9 (0.48); F11 and R11 (both in
Region A, (0.62); or F12 in Region A and R9 (1.16); F6 and R8 in Region B
(0.40); or F8, GATGACAGCAGTCGCAAAG, and R17 in Region B (1.00).
Statistical analysis. The Kendall tau, rank-order statistic (SPSS program,
version 10.0) was used to compare differences in the extent of hypermethy-
lation across the three ovarian tumor types, using the hypermethylation
scores (0, +1, +2 or +3) and excluding the three hypomethylated carcinomas.
The Spearman rho correlation was computed to assess the association between
Table 1
NBL2 methylation changes vs. transcription
Samples Meth. change relative to
somatic controlsa
NBL2 HhaI
sites
NBL2 NotI
sites
NBL2
transcriptionc
Normal controls
Kidney, lung, liver, thyroid,
heart, and spleenb
Small intestine
Lymphocytes
Placenta
Sperm
Ovarian carcinomas
A
B
D
E
F
G
J
Q
S
Wilms tumors
4
9
16
21
30
35
37
40
50
54
67
00N
N
N
N
ND
ND
0
-1
-3
ND
0
-1
-3
+1
+1
-2
-1
+3
-2
+1
+3
+2
0
0
-3
-2
0
-3
0
0
0
N
N
N
Y
N
N
N
N
N
00
0
0
0
0
0
0
0
0
0
-2
N
N
N
N
Y
N
N
Y
Y
N
N
+1
+3
+3
+2
+2
+2
+1
+2
+3
+1
aMethylation was scored from Southern blots as described in Materials and Methods. ND, not determined.
bThese control tissues were used as Southern blot standards for comparison to tumors and have similar
methylation patterns. cNBL2 transcription results were concordant by semi-quantitative and real-time
RT-PCR for the tumors, lung, placenta, intestine, and lymphocytes. For the other normal tissues, only semi-
quantiative RT-PCR was done. N, no detectable transcripts; Y, RT-dependent transcripts.
442
Cancer Biology & Therapy
2005; Vol. 4 Issue 4
Hyper- and Hypomethylation of a DNA Repeat in Cancer
Figure 2. Southern blot analyses with an NBL2 probe and tumor or control DNAs digested with CpG methylation-sensitive NotI (A) or CpG methylation-insen-
sitive SphI (B). Partial digestion with SphI reveals the ladder of fragments indicative of tandem arrays of NBL2.
AB

Page 4

methylation changes at NBL2 HhaI sites and Sat2 hypomethylation for each
tumor type.
RESULTS
Distribution of NBL2 in human DNA. Only a small percentage of
NBL2 copies have been sequenced, probably because most are in constitu-
tive heterochromatin. GenBank has two 1.4-kb NotI fragments of NBL2
(Y10752 and U59100)27,29and a tandem NBL2 array in a BAC clone
(AC018692) bordered by sequences that we call Regions A and B (Fig. 1A
and C). Southern blot analysis of control DNAs digested only partially with
the CpG methylation-insensitive enzyme SphI, which has an average of
about one site per sequenced NBL2 monomer (Fig. 1A), gave a ladder of
fragment sizes that were integral multiples of 1.4 kb (Fig. 2B). Therefore,
most copies of NBL2 are arranged in tandem.
Using PCR primers from within NBL2 (F2 and R1) with monochro-
mosomal human-rodent SCH DNAs, human chromosomes 13, 14, 15, and
21 were strongly amplified and 9 and Y were weakly amplified to give the
expected size product (Fig. 1B). An in silico search revealed 2-5 clustered
truncated copies of NBL2 with various orientations in 9q21 or 9p11 con-
tigs, NT078064, NT078066, NT078077, NT078051, NT078053, and
NT086759. Primer-pairs from NBL2 and Region A or B gave the expected
PCR products from all NBL2-positive chromosomes (diagrammed in Fig.
1C). This suggests that many of the NBL2 arrays have similar border
sequences. The copy number of NBL2 in two individuals was about
200–400 per haploid genome based upon analyses of serial dilutions of
genomic DNA and NBL2 plasmid.
Infrequent hypomethylation of NBL2 at NotI sites in ovarian cancers
and Wilms tumors. Although NotI hypomethylation was previously found
in the majority of hepatocellular carcinomas and neuroblastomas,26,29only
3 of 18 of ovarian epithelial carcinomas showed hypomethylation at NotI
sites when compared to various normal postnatal somatic tissues (Fig. 2A).
NotI cleaves 5'-GCGGCCGC-3' if neither CpG is methylated. Sperm
DNA, the hypomethylated standard, gave the expected 1.4-kb fragments
(Fig. 2A). Term placenta exhibited some hypomethylation at NotI sites (data
not shown; also see Fig. 3B), but not as much as sperm, in parallel to results
for satellite DNA.22,36None of the analyzed 13 low malignant potential
(LMP) ovarian tumors or 4 ovarian cystadenomas and only one of 51 Wilms
tumors (a pediatric kidney cancer) displayed hypomethylation at NotI sites.
This Wilms tumor (WT67) was one of the two recurrent tumors (also
WT25) in this study. WT67 and WT25 were also the only tumors from
patients treated with chemotherapy before surgery. The pre-chemotherapy,
primary tumor from the WT67 patient (WT27) did not exhibit this
hypomethylation.
Infrequent hypomethylation and frequent hypermethylation at NBL2
HhaI sites in ovarian carcinomas and Wilms tumors. We next examined
NBL2 methylation with HhaI, which cleaves 5'-GCGC-3' if the CpG is
unmethylated. Methylation at HhaI sites was frequently increased, rather
than decreased, in ovarian cancers and Wilms tumors compared to various
normal postnatal somatic standards (Figs. 3 and 4A). This cancer-linked
hypermethylation could be detected because HhaI digests of the standard
DNAs showed NBL2 fragments of intermediate molecular-weight (Fig. 3B).
The size of the high-molecular-weight fragments in the tumor DNAs often
exceeded 20 kb as seen in 0.5% agarose gels. With internal controls of phage
or plasmid DNA for each tumor and standard DNA tested (e.g., Fig. 4B),
we ruled out the presence of inhibitors of HhaI. In addition, there were no
inhibitors of HhaI as seen in tumor DNA samples admixed with sperm
DNA, digested with increased amounts of enzyme (40 U of enzyme per µg
of DNA), and examined by Southern blotting. Also, a high degree of
sequence variation in tumors cannot explain our results because complete
digestion of various tumor DNAs with the CpG methylation-insensitive
SphI gave similar NBL2 fragment profiles (data not shown).
We found hypomethylation at HhaI sites in NBL2 in only 3 of the 18
ovarian carcinomas (D, E, and G; Fig. 4A). These three tumors were also
hypomethylated at NotI sites (Table 1). None of the 51 Wilms tumors
showed this hypomethylation.
In cancers that displayed the most hypermethylation (score +3, Figs.
3 and 4), about three-fourths of the hybridization signal was in hyperme-
thylated fragments. Conversely, the two cancers that were most hypomethy-
lated at HhaI sites (ovarian carcinomas D and G, score -2; Fig. 4A) had
almost half of the hybridization signal in hypomethylated fragments.
Therefore, HhaI site hypomethylation in some cancers and hypermethyla-
tion in others affected a large fraction of the NBL2 repeats. Nonetheless,
ovarian carcinoma D and, to a lesser extent, carcinoma E gave evidence of a
www.landesbioscience.com
Cancer Biology & Therapy
443
Hyper- and Hypomethylation of a DNA Repeat in Cancer
Figure 3. Hypermethylation at HhaI sites in NBL2 in ovarian carcinomas and
Wilms tumors relative to various normal standards. DNA from uncultured
lymphocytes and normal kidney and brain from 3–5 individuals gave similar
results (not shown) as from other postnatal somatic controls. See Materials
and Methods for size definitions of the indicated regions of the blot used for
quantitation of methylation changes by phosphorimaging and calculation of
methylation scores relative to the standards.
A
B

Page 5

small portion of NBL2 repeats hypermethylated at HhaI sites (Fig. 4A, com-
pare high-molecular-weight DNA regions in carcinomas D and E and
somatic controls).
Association of NBL2 methylation changes with the degree of malig-
nancy and satellite DNA hypomethylation. We compared methylation at
NBL2 HhaI sites in carcinomas, LMP tumors, and cystadenomas, all of
which are derived from a similar37ovarian epithelial cell type. The percentages
of samples displaying appreciable hypermethylation relative to various
normal standards were 72, 54 and 22, respectively. This increase in the extent
of HhaI site hypermethylation with the degree of malignancy was statistically
significant (P = 0.001; Figs. 4, 5 and 6A). Hypomethylation at HhaI sites
was observed in three of the 18 carcinomas, but none of 13 LMP tumors
and 9 cystadenomas.
Because it had been proposed that DNA hypermethylation and
hypomethylation in cancer are linked, we looked for an association of NBL2
hypermethylation at HhaI sites with the frequent cancer-related
hypomethylation of another tandem repeat, namely, Sat2 in the juxtacen-
tromeric region of Chr1 at BstBI sites.6,38Sat2 hypomethylation is signifi-
cantly associated with global DNA hypomethylation.20,21,24,39Sat2
hypomethylation in ovarian carcinomas was significantly associated with
lower methylation scores for NBL2 when both sets of scores were compared
(P = 0.005) or when the carcinomas were dichotomized into those that
displayed Sat2 hypomethylation (11 with scores -1, -2, or -3) and those that
did not (7 with scores of 0; p = 0.016). The 51 Wilms tumors showed no
such association (Fig. 6B). This difference between the ovarian carcinomas
and Wilms tumors could be due to the ten-
dency of ovarian carcinomas to show con-
siderably more extensive hypomethylation
in Chr1 Sat2 than the Wilms tumors (aver-
age Sat2 hypomethylation score among Sat2
hypomethylation-positive tumors of -2.4
and -1.1, respectively).
Transcription of NBL2 in some cancers
irrespective of NBL2 methylation status.
Previously, we found by RT-PCR that NBL2
was transcribed in several ICF lymphoblas-
toid cell lines (including ICF B, Fig. 7A),
which display hypomethylation at the over-
lapping EagI/NotI site.35By RT-PCR with
primers F4 and R4, we tested for NBL2
transcripts in malignant and control samples.
Although we did not analyze the normal
counterpart from which the cancers are
derived (embryonic kidney nephrogenic
rests for the Wilms tumors and a still con-
troversial, tiny ovarian epithelial subcom-
partment40for the ovarian carcinomas), we
compared the cancers to a wide variety of
normal somatic tissues, as for the Southern
blots (Table 1). If the somatic controls are
similar to one another and samples from
disparate types of cancer are markedly different
from the somatic controls, such differences
are likely to be cancer-specific.
Only one ovarian carcinoma, three Wilms
tumors, and the ICF LCL gave RT-dependent
PCR products (Fig. 7A, Table 1).
Transcripts were observed both from
oligo(dT)- and random hexamer-primed
cDNA and by semi-quantitative and
real-time PCR. The average level of NBL2
transcripts was low, as seen in the absence of
Northern blot-detectable transcripts from
10 µg of ICF cell RNA. Moreover, 35 cycles
were necessary to visualize the NBL2
RT-PCR product from NBL2 RNA-positive tumors upon gel electrophore-
sis while only 30 were needed to observe a strong GAPDH RT-PCR band in
all samples. Furthermore, real-time RT-PCR gave only about 2–20% as
much GAPDH-normalized NBL2 signal from these NBL2 RNA-positive
tumors as from ICF cDNA. The level of GAPDH-normalized NBL2 prod-
uct from normal lung, small intestine, term placenta, or lymphocytes was
not appreciably above background, namely about 0.01-0.1% that from the
ICF cDNA standard in real-time RT-PCR. Upon semi-quantitative
RT-PCR, the nontumor tissues gave no specific band with NBL2 primers
(Table 1). In the tumors, there was no correlation between NBL2transcription
and methylation status.
Using RNA from an NBL2 transcript-positive cancer or the ICF cells,
specific transcripts were detected with various NBL2 primer-pairs in addition
to F4 and R4 (Fig. 7B). This suggests that all of the sequence can be
transcribed in vivo. Moreover, cDNA from these two sources was amplified
with primers that spanned NBL2 and adjacent Region A or B (Fig. 7C),
which indicates cotranscription.
DISCUSSION
Previously, reports of DNA methylation changes in cancer have
described either hypermethylation or hypomethylation of a given
DNA sequence. Here, we demonstrate for the first time that a DNA
sequence, the NBL2 tandem repeat, is hypermethylated in many
444
Cancer Biology & Therapy
2005; Vol. 4 Issue 4
Hyper- and Hypomethylation of a DNA Repeat in Cancer
Figure 4. Hypermethylation and hypomethylation at HhaI sites in NBL2 in ovarian carcinomas. (A) Southern
blot with methylation scores (negative for hypomethylation and positive for hypermethylation at HhaI sites)
relative to the standards (Stds). A few different exposures of the same blot are combined. (B) Ethidium bro-
mide-induced fluorescence from an electrophoresis gel containing internal controls for complete digestion
(Materials and Methods). Because of the high overall level of methylation of HhaI sites throughout the
genome, a high-molecular-weight smear is seen from human DNA. RE, plasmid digested with restriction
endonuclease (HhaI) in the absence of human DNA; no RE, only undigested plasmid.
A
B

Page 6

tumors while it is hypomethylated in others of the same type.
Seventy- two, 17, and 11% of the ovarian epithelial carcinomas dis-
played hypermethylation, hypomethylation, or no change in methy-
lation at HhaI sites. In many of these cancers, hypermethylation at
HhaI sites extended over >20 kb even though there are about 5 HhaI
sites per 1.4-kb monomer and very little high-molecular-weight
hybridizing material was observed in HhaI digests of various control
DNAs. Conversely, in three of the 18 carcinoma DNAs, there was con-
siderable hypomethylation (Fig. 4A).
Studies of NBL2 methylation in cancer in the past were essentially
confined to the NotI site in NBL2 and demonstrated that most
hepatocellular carcinomas and neuroblastomas were hypomethylated
at this site.26,29In the present study, only three out of 18 ovarian
carcinomas and one recurrent tumor out of 51 Wilms tumors (49
primary and two recurrent) showed hypomethylation at NotI sites
in NBL2 arrays. The same ovarian cancers that were hypomethylated
at NotI sites were hypomethylated at HhaI sites. Like the ovarian
carcinomas, most of the Wilms tumors (84%) showed hypermethy-
lation at HhaI sites. Because of the very infrequent cleavage of NBL2
arrays by NotI in somatic control DNAs, as compared to the
www.landesbioscience.com
Cancer Biology & Therapy
445
Hyper- and Hypomethylation of a DNA Repeat in Cancer
Figure 5. Hypermethylation at HhaI sites in NBL2 in ovarian LMP tumors and
cystadenomas. Note the lower frequency of hypermethylation of these ovar-
ian tumors than for ovarian carcinomas in Figure 4.
Figure 6. Association of NBL2 methylation changes with the degree of malig-
nancy and with Sat2 hypomethylation. (A) Hypermethylation (+1, +2 or +3)
and hypomethylation (-1 or -2) scores of 18 ovarian epithelial carcinomas,
13 LMP tumors, and 9 cystadenomas as well as 51 Wilms tumors at NBL2
HhaI sites; 0, no appreciable change from the standards. Hypermethylation
scores were significantly related to the malignant potential of the ovarian
tumors (p = 0.001) (B) Average methylation score for NBL2 at HhaI sites in
cancers with no hypomethylation of Chr1 Sat2 vs. that for cancers with
hypomethylation of Chr1 Sat2. The difference between the two groups of
ovarian carcinomas was significant (p = 0.016), and the same trend was
seen (although not statistically significant) even when the three carcinomas
with NBL2 hypomethylation at HhaI sites were excluded.
A
B

Page 7

appreciable level of cleavage by HhaI, only HhaI can reveal NBL2
hypermethylation in the tumors vs. the somatic controls. The low
percentage of tumors displaying hypomethylation at NBL2 HhaI or
NotI sites in the cancers in this study compared to the high percentage
of hepatocellular carcinomas and neuroblastomas with hypomethyla-
tion at the NBL2 NotI site in previous studies26,29probably represents
tumor-type specificity in epigenetic changes.5,41
While hypermethylation of CpG islands in cancer has been
found in numerous studies,4,5the only repeated DNA sequence that
was reported as hypermethylated in human cancer is the ribosomal
RNA gene unit (rDNA).42-45Although most NBL2 repeats are in
four of the five rDNA-rich acrocentric chromosomes, Thoraval and
colleagues29localized the NBL2 FISH signal to the centromeric region
rather than to the rDNA-containing short arms. This centromeric
region location of NBL2 implies that it is closer to sequences prone
to hypomethylation in cancer, namely centromeric satellite
DNA,24,25,39than to the tandem rDNA units.
Like promoters within CpG islands that can become hyperme-
thylated in cancer, NBL2 has a high (C + G) content and a high ratio
of (observed CpG)/(expected CpG)46(60% and 0.67, respectively,
for Y10752). However, the tandem arrays of these 1.4-kb repeats are
much longer than standard CpG islands. Most CpG islands that
have been described as hypermethylated in cancer were studied by
methods that select for extensive methylation of a previously
unmethylated DNA sequence. In contrast, NBL2 has considerable,
but incomplete, methylation in normal somatic tissues, which allows
the detection of both hyper- and hypomethylation in cancer. Our
comparison of HhaI digests of DNA from ovarian carcinomas,
Wilms tumors, and nine types of normal postnatal somatic controls
raises the following questions. How are methyla-
tion patterns maintained with strong similarity
in diverse normal tissues in regions of the
genome having intermediate levels of methyla-
tion? How do these patterns become deranged
during tumorigenesis in sequences like NBL2
that can undergo either increases or decreases in
methylation? These issues will be addressed by
bisulfite-based genomic sequencing.
An in silico search of the NBL2 array in the
BAC clone shown in Figure 1C revealed multi-
ple small open reading frames in each repeat
although no protein product has been demon-
strated. We detected NBL2 transcripts in several
tumors (Table 1) and ICF cells.35However, the
NBL2 transcript level in these tumors was low
compared to that of GAPDH even though there
are about 200–400 more copies of the NBL2
monomer in the genome. Because NBL2 and
adjacent heterologous sequences were cotran-
scribed and no candidate promoter sequences
(Proscan) or polyadenylation signals in NBL2
were detected, polyadenylated NBL2 RNA prob-
ably resulted from run-through transcription
originating in a non-NBL2 promoter adjacent to
one or more of the NBL2 repeats. This may
explain why there was no correlation of NBL2
hypomethylation and transcription among the
tumors and why there was only a low level of
NBL2 transcripts. About 30 reported ESTs or
cDNAs have a very high degree of homology to
NBL2 over ~200–600 bp but many of these come from cancers or a
few normal tissue cDNA libraries, including from kidney (Blastn
search of human EST database). These molecular clones, whose
preparation usually involves NotI digestion, might be due to con-
tamination with NotI site-containing genomic NBL2.47This would
be consistent with our lack of detection of NBL2 transcripts by
RT-PCR from various normal tissues, including kidney. Evidence
from Susan Clarke and colleagues48indicated that inhibition of
transcription of a previously unmethylated CpG-island promoter led
to its hypermethylation in concert with sporadic and random
methylation of individual CpG sites. However, given the lack of
detectable NBL2 transcripts in normal tissues, such inhibition prob-
ably could not account for NBL2’s very extensive hypermethylation
in Wilms tumors and ovarian carcinomas unless it involved a tran-
script too transient, small, or tissue-restricted to be detected by our
RT-PCR assays. Therefore, we propose that there is a cancer-hyper-
methylation pathway that alters DNA methylation without inhibi-
tion of transcription.
We previously demonstrated that there is not a significant positive
association between hypermethylation of CpG islands in Wilms
tumors and hypomethylation of satellite DNA.21However, both
DNA hypomethylation (satellite DNA, LINE1 repeats, NBL2, and
global DNA) and hypermethylation (CpG islands, rDNA, and
NBL2) are cancer-linked phenomena. Because NBL2 sequences are
predisposed to both hypermethylation and hypomethylation in
cancer, these opposite epigenetic changes may share a common step
during carcinogenesis. That step might affect chromatin structure
with independent subsequent steps leading specifically to either
hypomethylation or hypermethylation of cytosine residues.
446
Cancer Biology & Therapy
2005; Vol. 4 Issue 4
Hyper- and Hypomethylation of a DNA Repeat in Cancer
Figure 7. RT-PCR of NBL2 and adjacent sequences. (A) Electrophoresis of NBL2 and internal stan-
dard (GAPDH) RT-PCR products obtained with primers R4 and F4 from oligo(dT)-primed ICF LCL,
ovarian carcinoma, and Wilms tumor cDNAs. Controls without reverse transcriptase (RT, -) were
included for each sample. (B) Diagram showing the RT-PCR products from cDNA of NBL2 RNA-pos-
itive WT50 amplified with various NBL2 primer-pairs. The same products were obtained with
cDNA from ICF cells. (C) Diagram of RT-PCR products obtained from both WT50 and ICF cDNAs
using an NBL2 primer and a primer from adjacent Region A or B (Fig. 1C) or just two Region A
primers. The double-sided arrow labeled 1.6-kb shows the longest overlap of the PCR products
and, importantly, is more than the monomer length of 1.4 kb.
A
B
C

Page 8

20. Qu G, Dubeau L, Narayan A, Yu M, Ehrlich M. Satellite DNA hypomethylation vs. over-
all genomic hypomethylation in ovarian epithelial tumors of different malignant potential.
Mut Res 1999; 423:91-101.
21. Ehrlich M, Jiang G, Fiala ES, Dome JS, Yu MS, Long TI, Youn B, Sohn O-S,
Widschwendter M, Tomlinson GE, Chintagumpala M, Champagne M, Parham DM,
Liang G, Malik K, Laird PW. Hypomethylation and hypermethylation of DNA in Wilms
tumors. Oncogene 2002; 21:6694-702.
22. Narayan A, Ji W, Zhang X-Y, Marrogi A, Graff JR, Baylin SB, Ehrlich M.
Hypomethylation of pericentromeric DNA in breast adenocarcinomas. Int J Cancer 1998;
77:833-8.
23. Qu G, Grundy PE, Narayan A, Ehrlich M. Frequent hypomethylation in Wilms tumors of
pericentromeric DNA in chromosomes 1 and 16. Cancer Genet Cytogenet 1999; 109:34-9.
24. Ehrlich M, Hopkins N, Jiang G, Dome JS, Yu MS, Woods CB, Tomlinson GE,
Chintagumpala M, Champagne M, Diller L, Parham DM, Sawyer J. Satellite hypomethy-
lation in karyotyped Wilms tumors. Cancer Genet Cytogenet 2003; 141:97-105.
25. Widschwendter M, Jiang G, Woods C, Muller HM, Fiegl H, Goebel G, Marth C, Holzner
EM, Zeimet AG, Laird PW, Ehrlich M. DNA hypomethylation and ovarian cancer biolo-
gy. Cancer Res 2004; 64:4472-80.
26. Itano O, Ueda M, Kikuchi K, Hashimoto O, Hayatsu S, Kawaguchi M, Seki H, Aiura K,
Kitajima M. Correlation of postoperative recurrence in hepatocellular carcinoma with
demethylation of repetitive sequences. Oncogene 2002; 21:789-97.
27. Nagai H, Kim YS, Yasuda T, Ohmachi Y, Yokouchi H, Monden M, Emi M, Konishi N,
Nogami M, Okumura K, Matsubara K. A novel sperm-specific hypomethylation sequence
is a demethylation hotspot in human hepatocellular carcinomas. Gene 1999; 237:15-20.
28. Thoraval D, Asakawa J, Kodaira M, Chang C, Radany E, Kuick R, Lamb B, Richardson
B, Neel JV, Glover T, Hanash S. A methylated human 9-kb repetitive sequence on acro-
centric chromosomes is homologous to a subtelomeric repeat in chimpanzees. Proc Natl
Acad Sci USA 1996; 93:4442-7.
29. Thoraval D, Asakawa J, Wimmer K, Kuick R, Lamb B, Richardson B, Ambros P, Glover
T, Hanash S. Demethylation of repetitive DNA sequences in neuroblastoma. Genes
Chromosomes Cancer 1996; 17:234-44.
30. Nagai H, Ponglikitmongkol M, Mita E, Ohmachi Y, Yoshikawa H, Saeki R, Yumoto Y,
Nakanishi T, Matsubara K. Aberration of genomic DNA in association with human hepa-
tocellular carcinomas detected by 2-dimensional gel analysis. Cancer Res 1994;
54:1545-50.
31. Ehrlich M, Buchanan K, Tsien F, Jiang G, Sun B, Uicker W, Weemaes C, Smeets D,
Sperling K, Belohradsky B, Tommerup N, Misek D, Rouillard J-M, Kuick R, Hanash S.
DNA methyltransferase 3B mutations linked to the ICF syndrome cause dysregulation of
lymphocyte migration, activation, and survival genes. Hum Mol Genet 2001; 10:2917-31.
32. Ehrlich M, Gama-Sosa M, Huang L-H, Midgett RM, Kuo KC, McCune RA, Gehrke C.
Amount and distribution of 5-methylcytosine in human DNA from different types of tis-
sues or cells. Nucleic Acids Res 1982; 10:2709-21.
33. Lemmers RJL, de Kievit P, van Geel M, van der Wielen MJ, Bakker E, Padberg GW, Frants
RR, van der Maarel SM. Complete allele information in the diagnosis of facioscapulo-
humeral muscular dystrophy by triple DNA analysis. Ann Neurol 2001; 50:816-9.
34. Jiang G, Yang F, van Overveld PG, Vedanarayanan V, van der Maarel S, Ehrlich M. Testing
the position-effect variegation hypothesis for facioscapulohumeral muscular dystrophy by
analysis of histone modification and gene expression in subtelomeric 4q. Hum Mol Genet
2003; 12:2909-21.
35. Kondo T, Comenge Y, Bobek MP, Kuick R, Lamb B, Zhu X, Narayan A, Bourc’his D,
Viegas-Pequinot E, Ehrlich M, Hanash S. Whole-genome methylation scan in ICF syn-
drome: Hypomethylation of nonsatellite DNA repeats D4Z4 and NBL2. Hum Mol Gen
2000; 9:597-604.
36. Ehrlich M, Tsien F, Herrera D, Blackman V, Roggenbuck J, Tuck-Muller CM. High fre-
quencies of ICF syndrome-like pericentromeric heterochromatin decondensation and
breakage in chromosome 1 in a chorionic villus sample. J Med Genet 2001; 38:882-884.
37. Zheng J, Benedict WF, Xu HJ, Hu SX, Kim TM, Velicescu M, Wan M, Cofer KF, Dubeau
L. Genetic disparity between morphologically benign cysts contiguous to ovarian carcino-
mas and solitary cystadenomas. J Natl Cancer Inst 1995; 87:1146-53.
38. Saito Y, Kanai Y, Sakamoto M, Saito H, Ishii H, Hirohashi S. Expression of mRNA for
DNA methyltransferases and methyl-CpG-binding proteins and DNA methylation status
on CpG islands and pericentromeric satellite regions during human hepatocarcinogenesis.
Hepatology 2001; 33:561-8.
39. Jackson K, Yu M, Arakawa K, Fiala E, Youn B, Fiegl H, Muller-Holzner E, Widschwendter
M, Ehrlich M. DNA hypomethylation is prevalent even in low-grade breast cancers.
Cancer Biol Ther 2004; 3:1225-31.
40. Dubeau L. The cell of origin of ovarian epithelial tumors and the ovarian surface epitheli-
um dogma: Does the emperor have no clothes? Gynecol Oncol 1999; 72:437-42.
41. Florl AR, Lower R, Schmitz-Drager BJ, Schulz WA. DNA methylation and expression of
LINE-1 and HERV-K provirus sequences in urothelial and renal cell carcinomas. Br J
Cancer 1999; 80:1312-21.
42. Toyota M, Ho C, Ahuja N, Jair KW, Li Q, Ohe-Toyota M, Baylin SB, Issa JP. Identification
of differentially methylated sequences in colorectal cancer by methylated CpG island
amplification. Cancer Res 1999; 59:2307-12.
43. Yan PS, Rodriguez FJ, Laux DE, Perry MR, Standiford SB, Huang TH. Hypermethylation
of ribosomal DNA in human breast carcinoma. Br J Cancer 2000; 82:514-7.
44. Kim YS, Yoo HS, Lee KT, Goh SH, Jung JS, Oh SW, Baba M, Yasuda T, Matsubara K,
Nagai H. Detection of genetic alterations in the human gastric cancer cell lines by
two-dimensional analysis of genomic DNA. Int J Oncol 2000; 17:297-308.
Our demonstration of alternative epigenetic fates for the same
DNA sequence in cancer indicates that tumorigenesis-linked DNA
methylation changes are much more flexible than commonly real-
ized. We hypothesize that CpG-island promoters that become
hypermethylated in earlier stages of cancer may subsequently under-
go demethylation during waves of tumor progression-linked DNA
demethylation. These waves of global DNA demethylation often
involve satellite DNA.20,21,24,39That demethylation of parts of the
genome could include a naturally occurring reversal of some
cancer-linked hypermethylation is consistent with the significant
association that we found in breast and ovarian cancers between
more hypermethylation of a 5' CDH13 CpG island and less
hypomethylation of Sat2.25Similarly, in this study, higher NBL2
methylation scores were significantly associated with less Sat2
hypomethylation in ovarian carcinomas. Spontaneous demethylation
of a CpG island promoter in a tumor after its hypermethylation
might result in gene reactivation that could sometimes be advanta-
geous to tumor progression.49
References
1. Feinberg AP, Vogelstein B. Hypomethylation distinguishes genes of some human cancers
from their normal counterparts. Nature 1983; 301:89-92.
2. Feinberg AP, Vogelstein B. Hypomethylation of ras oncogenes in primary human cancers.
Biochem Biophys Res Commun 1983; 111:47-54.
3. Gama-Sosa MA, Slagel VA, Trewyn RW, Oxenhandler R, Kuo KC, Gehrke CW, Ehrlich
M. The 5-methylcytosine content of DNA from human tumors. Nucleic Acids Res 1983;
11:6883-94.
4. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nat Rev Genet
2002; 3:415-28.
5. Costello JF, Fruhwald MC, Smiraglia DJ, Rush LJ, Robertson GP, Gao X, Wright FA,
Feramisco JD, Peltomaki P, Lang JC, Schuller DE, Yu L, Bloomfield CD, Caligiuri MA,
Yates A, Nishikawa R, Su Huang H, Petrelli NJ, Zhang X, O’Dorisio MS, Held WA,
Cavenee WK, Plass C. Aberrant CpG-island methylation has nonrandom and
tumour-type-specific patterns. Nat Genet 2000; 24:132-8.
6. Ehrlich M. DNA methylation in cancer: Too much, but also too little. Oncogene 2002;
21:5400-13.
7. Wong N, Lam WC, Lai PB, Pang E, Lau WY, Johnson PJ. Hypomethylation of chromo-
some 1 heterochromatin DNA correlates with q-arm copy gain in human hepatocellular
carcinoma. Am J Pathol 2001; 159:465-71.
8. Saito Y, Kanai Y, Sakamoto M, Saito H, Ishii H, Hirohashi S. Overexpression of a splice
variant of DNA methyltransferase 3b, DNMT3b4, associated with DNA hypomethylation
on pericentromeric satellite regions during human hepatocarcinogenesis. Proc Natl Acad
Sci USA 2002; 99:10060-5.
9. Yamamoto F, Yamamoto M, Soto JL, Kojima E, Wang EN, Perucho M, Sekiya T,
Yamanaka H. Notl-Msell methylation-sensitive amplied fragment length polymorhism for
DNA methylation analysis of human cancers. Electrophoresis 2001; 22:1946-56.
10. Tsuda H, Takarabe T, Kanai Y, Fukutomi T, Hirohashi S. Correlation of DNA hypomethy-
lation at pericentromeric heterochromatin regions of chromosomes 16 and 1 with histo-
logical features and chromosomal abnormalities of human breast carcinomas. Am J Pathol
2002; 161:859-66.
11. Cheng P, Schmutte C, Cofer KF, Felix JC, Yu MC, Dubeau L. Alterations in DNA methy-
lation are early, but not initial, events in ovarian tumorigenesis. Br J Cancer 1997;
75:396-402.
12. Lin CH, Hsieh SY, Sheen IS, Lee WC, Chen TC, Shyu WC, Liaw YF. Genome-wide
hypomethylation in hepatocellular carcinogenesis. Cancer Res 2001; 61:4238-4243.
13. Kim Y-I, Giuliano A, Hatch KD, Schneider A, Nour MA, Dallal GE, Selhub J, Mason JB.
Global DNA hypomethylation increases progressively in cervical dysplasia and carcinoma.
Cancer 1994; 74:893-9.
14. Soares J, Pinto AE, Cunha CV, Andre S, Barao I, Sousa JM, Cravo M. Global DNA
hypomethylation in breast carcinoma: Correlation with prognostic factors and tumor pro-
gression. Cancer 1999; 85:112-8.
15. Eads CA, Nickel AE, Laird PW. Complete genetic suppression of polyp formation and
reduction of CpG-island hypermethylation in Apc(Min/+) Dnmt1-hypomorphic mice.
Cancer Res 2002; 62:1296-9.
16. Gaudet F, Hodgson JG, Eden A, Jackson-Grusby L, Dausman J, Gray JW, Leonhardt H,
Jaenisch R. Induction of tumors in mice by genomic hypomethylation. Science 2003;
300:489-92.
17. Carr BI, Reilly G, Smith SS, Winberg C, Riggs A. The tumorigenicity of 5-azacytidine in
the male Fisher rat. Carcinogenesis 1984; 5:1583-90.
18. Denda A, Rao PM, Rajalakshmi S, Sarma DS. 5-azacytidine potentiates initiation induced
by carcinogens in rat liver. Carcinogenesis 1985; 6:145-146.
19. Thomas GA, Williams ED. Production of thyroid tumours in mice by demethylating
agents. Carcinogenesis 1992; 13:1039-42.
www.landesbioscience.com
Cancer Biology & Therapy
447
Hyper- and Hypomethylation of a DNA Repeat in Cancer

Page 9

448
Cancer Biology & Therapy
2005; Vol. 4 Issue 4
Hyper- and Hypomethylation of a DNA Repeat in Cancer
45. Powell MA, Mutch DG, Rader JS, Herzog TJ, Huang TH, Goodfellow PJ. Ribosomal
DNA methylation in patients with endometrial carcinoma: An independent prognostic
marker. Cancer 2002; 94:2941-52.
46. Takai D, Jones PA. Comprehensive analysis of CpG islands in human chromosomes 21
and 22. Proc Natl Acad Sci USA 2002; 99:3740-5.
47. Sorek R, Safer HM. A novel algorithm for computational identification of contaminated
EST libraries. Nucleic Acids Res 2003; 31:1067-74.
48. Stirzaker C, Song JZ, Davidson B, Clark SJ. Transcriptional gene silencing promotes DNA
hypermethylation through a sequential change in chromatin modifications in cancer cells.
Cancer Res 2004; 64:3871-7.
49. Cheng CW, Wu PE, Yu JC, Huang CS, Yue CT, Wu CW, Shen CY. Mechanisms of inac-
tivation of E-cadherin in breast carcinoma: Modification of the two-hit hypothesis of
tumor suppressor gene. Oncogene 2001; 20:3814-23.