A R T I C L E
Genetics supersedes epigenetics in colon cancer phenotype
Kentaro Yamashita, Tomoko Dai, Yuichi Dai, Fumiichiro Yamamoto, and Manuel Perucho*
The Burnham Institute, La Jolla Cancer Center, Cancer Genetics and Epigenetics Program
10901 North Torrey Pines Road, La Jolla, California 92037
A CpG island DNA methylator phenotype has been postulated to explain silencing of the hMLH1 DNA mismatch repair
gene in cancer of the microsatellite mutator phenotype. To evaluate this model, we analyzed methylation in CpG islands
from six mutator and suppressor genes, and thirty random genomic sites, in a panel of colorectal cancers. Tumor-specific
somatichypermethylation wasa widespreadage-dependentprocess thatfollowed anormalGaussian distribution.Because
there was no discontinuity in methylation rate, our results challenge the methylator phenotype hypothesis and its hypotheti-
cal pathological underlying defect. We also show that the mutator phenotype dominates over the gradual accumulation
of DNA hypermethylation in determining the genotypic features that govern the phenotypic peculiarities of colon cancer
of the mutator pathway.
nonpolyposis colorectal cancers (HNPCC) and sporadic gastro-
intestinal tumors (Kolodner and Marsischky, 1999). Some of
these MSI-positive cancers undergo epigenetic silencing of the
DNA MMR gene hMLH1, a process that is accompanied by
hypermethylation of the gene’s promoter (Kane et al., 1997).
The existence of a CpG island methylator phenotype (CIMP)
has been postulated to explain the somatic hypermethylation
associated with silencing of the hMLH1 mutator gene and sev-
eral tumor suppressor genes (Ahuja et al., 1997; Toyota et al.,
The CIMP has been proposed to be responsible for the
manifestation of the microsatellite mutator phenotype (MMP)
characteristic of tumors with MSI (Toyota et al., 1999a; Toyota
and Issa, 2000). MSI-positive tumors display a mutator pheno-
type characterized by an over two orders of magnitude higher
mutation rate than normal cells (Ionov et al., 1993; Shibata et
al., 1994). The MMP leads to the accumulation of oncogenic
mutations in cancer genes(oncogenes and tumor suppressors),
ultimately leading to cancer (Perucho, 1996; Kinzler and Vo-
al., 1994; Perucho, 1996; Olschwang et al., 1997; Breivik et
al., 1997). Mutator genes are therefore more fundamental than
oncogenes and tumor suppressor genes, as the former cause
mutations that trigger the oncogenic potential of the latter. Can-
cer driven by mutator genes represents a “remote control”
mechanism for carcinogenesis, as mutator gene inactivation
The accumulation of genetic alterations during tumorigenesis
substantiates the mutational theory of cancer (Loeb, 1991), but
epigenetic alterations are also germane to carcinogenesis
(Jones and Laird, 1999). Evolution by natural selection has de-
veloped epigenetics as a means to efficiently achieve the regu-
tion and development. This process has been called the
(Jenuwein and Allis, 2001; Turner, 2002). In a highly interrelated
and complex process, posttranslational modifications in his-
tones and other chromatin proteins, together with changes in
DNA methylation at CpG sequences, lead to changes in gene
expression and permanent silencing (Bird, 2001; Baylin and
Somatic hypomethylation and hypermethylation have been
associated with tumorigenesis. Global hypomethylation occurs
in human tumors (Feinberg and Vogelstein, 1983) and can either
suppress (Laird et al., 1995) or induce (Eden et al., 2003) tumors
in mice. Hypermethylation has been linked to tumor suppressor
ple is the silencing of hMLH1 DNA mismatch repair (MMR) gene
in colon cancer.
About 13% of unselected colon tumors accumulate hun-
dreds of thousands of somatic mutations in microsatellite se-
quences (Ionov et al., 1993). MMR deficiency underlies this
genome-wide microsatellite instability (MSI) in some hereditary
S I G N I F I C A N C E
Silencing of hMLH1 in tumors of the microsatellite mutator phenotype illustrates the importance of epigenetics in cancer. The bimodal
distribution of ubiquitous microsatellite mutations defines the microsatellite mutator phenotype, a critical determinant of tumor cell
fate driving tumorigenesis through a specific pathway. The methylator phenotype hypothesis, as begetting the microsatellite mutator
phenotype, adds a more fundamental earlier step in carcinogenesis. However, our evidence for the absence of a bimodal distribution
for somatic hypermethylation is conclusive since it was obtained by an unbiased approach. The epigenetic origin of the mutator
phenotype, and cancer, remains fascinating and mysterious. However, it is not initiated by a punctual event underlying a pathogenic
methylator phenotype, but rather by a gradual age-dependent disintegration of the epigenetic code.
CANCER CELL : AUGUST 2003 · VOL. 4 · COPYRIGHT 2003 CELL PRESS 121
A R T I C L E
does not immediately lead to altered cell growth or survival
(Perucho et al., 1994; Perucho, 1996; Cahill et al., 1999).
The sequence of events in the MMP pathway for cancer
can be summarized as follows: inactivation of MMR (mutator)
genes causes a mutator phenotype, which causes oncogenic
mutations, which causecancer. The implicit sequenceof events
in the methylator phenotype “ultra remote control” pathway for
cancer is as follows: CIMP causes inactivation of MMR, which
causes the MMP, which causes oncogenic mutations, which
cause cancer (Toyota et al., 1999a). Methylator phenotype is
thuseven morefundamentalthan mutatorphenotype forcancer
pathogenesis, as it generates the MMP in tumors not carrying
germline and/or somatic MMR mutator mutations.
Despite its importance, the CIMP concept has not been
precisely defined. The current definition rests on the simultane-
ous tumor-specific hypermethylation of multiple CpG islands.
The CIMP conceptually parallels the MMP that was defined by
the presence of ubiquitous somatic mutations in mononucleo-
tide repeats (Ionov et al., 1993). But the MMP was defined
precisely because the majority of the tumors did not contain
these mutations. The bimodal distribution of microsatellite mu-
and tumors without these ubiquitous microsatellite mutations
(Ionov et al., 1993; Perucho et al., 1994).
The methylator phenotype hypothesis also necessarily im-
plies a clear-cut distinction between tumors with and tumors
without an enhanced pathological rate of somatic DNA hyper-
methylation. This segregation is necessary to postulate that
some tumors possess a methylator phenotype, similar to the
mutator phenotype possessed by tumors of the MMP. The
methylator phenotype concept is important because it implies
an underlying defect in the cellular machinery responsible for
the generation of hypermethylation events, similar to the MMR
deficiency underlying MSI and the MMP (Laird, 2003). We car-
ried out this study to test this hypothesis. We analyzed the
hypermethylation alterations occurring in CpG islands in some
cancer genes (mutators and suppressors) and random genomic
sites, in a panel of colorectal tumors with and without MSI. The
results show that tumor-specific somatic hypermethylation is a
widespread phenomenon shared by all colon tumors and that
the alterations are dispersed into a nearly perfect normal
Gaussian distribution when a sufficient number of loci are ana-
lyzed. We also show that the MSI phenotype is dominant over
the CpG island methylation phenotype.
which of the tumors induce methylation with rates higher than
normal cells and which ones do not (Figure 1C). However, be-
cause the number of loci analyzed was very small, it was possi-
ble that a distinctive group of tumors could contain none or
very few methylation alterations (Figures 1B and 1C).
To directly address this possibility, we analyzed the global
methylation pattern in 32 colorectal cancer specimens by meth-
ylation-sensitive amplified fragment length polymorphism (MS-
the genome (Yamamoto et al., 2001). In these particular experi-
ments, we scanned the status of the two CpG sites contained
in the NotI restriction endonuclease methylation-sensitive rec-
ognition site (GCGGCCGC). In MS-AFLP fingerprinting, tumor-
specific somatic hyper- and hypomethylation are recognized as
differences of band intensity between PCR products derived
from normal and tumor DNA (Figure 2).
About 100 bands appeared in each experiment. Of these,
about 75 were universally amplified from normal tissue DNA
(Figure 2). Nine of eleven such bands, characterized by cloning
and sequencing, matched human sequences derived from the
Human Genome Project through BLAST search. Only one of
a well-defined CpG island according to the currently accepted
criterion (Gardiner-Garden and Frommer, 1987).
Hypermethylation and hypomethylation in colon cancer
exhibited methylation alterations. Hyper- and hypomethylation
were each observed in about 30 of the 75 bands (?40%) with
a 20% overlapping (six bands showing both hyper- and hypo-
methylation in different tumors). The average level of hyper-
methylation in the 32 tumors analyzed was 21.5% (203 hyper-
methylated bands of 944 total analyzed bands) while the
bands). The average number of altered bands per tumor was
6.34 for hypermethylation and 3.3 for hypomethylation. There
was no correlation between hypermethylation and hypomethyl-
ation. Although tumors with levels of hypermethylation higher
than average had a higher level of hypomethylation, and vice
versa, the differences were not significant (data not shown).
In this report, we focus on hypermethylation and we will not
further elaborate on hypomethylation. The methylation status
of 30 bands showing tumor-specific hypermethylation is sum-
marized in Figure 3A. Some bands without methylation changes
(positions 5 and 17 of Figure 3A top panel and the last two
they exhibited alterations in a parallel analysis of 20 gastric
cancers. Bands with no methylation alterations in colon and
gastric cancers were excluded. Some bands exhibited more
alterations than others (top panel), but the alterations were dis-
persed gradually among the colon tumors (bottom panel). The
results with the gastric cancers analyzed were similar (data not
Hypermethylation of cancer genes and anonymous
We examined by methylation-specific PCR (MSP) (Herman et
al., 1996) the promoters of six genes in a panel of 207 colorectal
cancers and the corresponding paired normal tissue. These
included the MMR mutator gene hMLH1, the O6MGMT DNA
repair gene, and the tumor suppressor genes p16INK4A, p14ARF,
APC, and CDH1 (E-cadherin) (Figure 1A). All of these genes
undergo hypermethylation in colon cancer (Toyota et al., 1999a;
Esteller et al., 2001; Shen et al., 2003). A diagram of the tumor
arrangement based on the number of methylated loci per tumor
formed a gradual pattern rather than a bimodal distribution (Fig-
ure 1B).These resultsare not compatiblewith theCIMP hypoth-
esis since it is not possible to decide with a sound criterion
Distribution of genetic and epigenetic somatic
Figure 3B shows the effect of increasing the number of loci
analyzed in the shape of the distribution of methylation alter-
ations in the tumors. Analysis of only 5 loci was ambiguous,
witha gradualdiminishing oftumors containingmultiple methyl-
CANCER CELL : AUGUST 2003
A R T I C L E
Figure 1. Gradual distribution of CpG island DNA methylation alterations in colon cancer
A: Representative MSP experiments for methylation analysis of hMLH1 and p16 genes. PCR products amplified with unmethylated (U) and methylated (M)
sequence-specific primers. Untreated DNA is a negative control in which the sodium bisulfite treatment was omitted. MW: DNA fragment ladder used as
molecular weight marker.
B: Summary of CpG hypermethylation. Upper: Methylation status of six CpG islands from known suppressor/mutator genes as well as an anonymous CpG
island (ACG) identified through MS-AFLP (see Figure 2) in 207 colorectal cancers (filled: methylated, blank: unmethylated). Microsatellite instability (MSI) is
indicated at the top row (filled, MSI-positive, blank, MSI-negative). The cut off points for the grouping of tumors into two groups with relatively high and
low methylation alterations are indicated (see text).
C: The methylated loci per tumor from the above data show a nonbimodal distribution. MSI-positive tumors were also evenly distributed.
ated loci (Figure 3B, top), similar to the curve of the functional
loci (Figure 1C). However, when more loci were analyzed, the
distribution profile acquired a distinctive Gaussian shape, re-
flecting anormal distribution ofrandom events (Figure3B, lower
panels). No single tumor remained without methylation alter-
ations after all 35 loci were considered. Therefore, tumor-spe-
cific somatic hypermethylation is a widespread and gradual
tions was distinctively bimodal, regardless of the number of loci
analyzed (Figure 3B).
Thus, there is a fundamental difference between the tumor-
ually, without a defined boundary (Figure 3A). Mutations in
microsatellite sequences exhibited on the other hand a discon-
tinuous distribution of (Figure 3C), with a sharp border splitting
Methylator versus mutator genotype-phenotype
We next compared the genotypic features of colon tumors with
respect to their mutator and ‘methylator’ phenotypes. We di-
vided the tumors according to the methylation status of the
CpG islands analyzed by MSP (Figure 1B), into a group with
two or more methylated CpG islands and another with one or
none (criterion 1). This is a division at approximately the median
valueformethylation (1.5methylatedlociper tumor).Theresults
showed a significant association between methylation and right
side location (Figure 4A). Comparison of tumors with no methyl-
2), that resembles the comparison of lower and higher quartiles
CANCER CELL : AUGUST 2003123
A R T I C L E
prototypical in colorectal tumorigenesis, K-ras oncogene and
APC and p53 tumor suppressors (Figure 4B).
Comparison of the genetic and epigenetic alterations in re-
gard to tumor cell phenotype revealed that the mutator pheno-
type and the genetic alterations (MSI) were dominant over the
“methylator phenotype” and the epigenetic alterations (CpG
island methylation). The differences in genotype and phenotype
observed between MSI-positive and-negative tumors were in-
dependent on CpG island methylation status (Figure 4D). On
the other hand, filtering out the MSI-positive cases from the
population of tumors diminishedthe asymmetries between rela-
tively high and low CpG island methylation in stage and grade,
and K-ras and p53 mutations (Figure 4C). Therefore, while tu-
mors with and without the MMP exhibited marked differences
in genotype and phenotype, tumors with relatively high and low
CpG island DNA methylation were essentially indistinguishable.
The preferential location in the proximal colon was shared
by DNA hypermethylation and MSI, with the latter showing an
even more pronounced tendency (Figure 5). The only significant
feature of DNA hypermethylation that was not shared with MSI
was its associationwith older age. The methylationof each CpG
island in particular, and the average of all the CpG islands,
increased with the patients’ age (Figure 6).
Since the methylator concept was proposed in colon cancer
(Ahuja et al., 1997; Toyota et al., 1999a), this hypothesis has
received considerable attention, and the CIMP is currently re-
esis pathway (Toyota et al., 1999a; Toyota and Issa, 2000; Elsa-
leh et al., 2000; Peltomaki et al., 2000; Baylin and Bestor, 2002;
van Rijnsoever et al., 2002; Iacopetta, 2003). The CIMP has also
been assigned to various types of tumors other than colorectal
cancer. So far, the CIMP has been reported in neoplasms of
stomach (Toyota et al., 1999b), pancreas (Ueki et al., 2000),
ovary (Strathdee et al., 2001), hepatocellular carcinoma (Shen
et al., 2002), and adenoma (Rashid et al., 2001), as well as
hyperplastic polyp-polyposis (Chan et al., 2002) of the large
intestine. Two groups have recently reported the absence of
evidence supporting the CIMP model. One group interpreted
their data as not supporting the CIMP hypothesis in esophageal
adenocarcinoma (Eads et al., 2001) and another group reported
data inconsistent with the CIMP in colorectal cancer, although
the results were not explicitly interpreted as contradicting the
CIMP hypothesis (Hawkins et al., 2002).
The definition of the CIMP has been in constant evolution
since its inception (Toyota et al., 1999a, 1999b, 2000; Ueki et
al., 2000; Shen et al., 2002, Rashid et al., 2001; Chan et al.,
2002). Initiallythe conceptwas definedby the“frequent concor-
dant methylation of the type C clones examined” (Toyota et al.,
1999a). The CIMP-positive group was classified as having a
high level of type C methylation (three or more loci) (Toyota et
al., 1999a). Subsequently, the CIMP has been diversely charac-
terized from a methylation tendency (Peltomaki et al., 2000) to
a variable pattern of hypermethylation of CpG islands in tumor
suppressor genes (Shen et al., 2002). The current common
definition of simultaneous methylation of multiple CpG islands
is not very precise. The issue is further complicated because
tumors were classified into two distinct categories (CIMP? and
CIMP?) only in the first report of colorectal tumors (Toyota et
Figure 2. DNA hypermethylationin coloncancer detectedby MS-AFLPDNA
Autoradiogram of a MS-AFLP fingerprint of colon tumors with and without
MSI. Two different amounts of template DNA (10 and 15 ng) were amplified
for each sample. Left empty arrowheads represent hypomethylation,
whereas right solid arrowheads indicate hypermethylation. Asterisks denote
bands exhibitingrecurrent hypo-or hypermethylationin severaltumors, with
the most common alterations highlighted by double asterisks. N: normal, T:
tumor DNA. Additional altered smaller bands were detected in the same
experiment by shorter electrophoresis.
(Figure 1B), yielded similar results (data not shown). No signifi-
cant associationswere observedwith therest ofthe parameters
analyzed (gender, progression, differentiation, K-ras or p53 mu-
differences compared to MSI negative tumors, in accord with
MSI positive tumors were predominantly found in the proximal
colon, were less advanced in tumor progression, and many
also displayed a low mutational incidence in the cancer genes
CANCER CELL : AUGUST 2003
A R T I C L E
Figure 3. DNA hypermethylation follows a normal distribution in contrast with the bimodal distribution of MSI
A: DNA hypermethylation is gradual in colon cancers. The methylation status of 30 NotI CpG sites analyzed by MS-AFLP is shown. The five loci analyzed by
MSP are also included. In the top panel, the 35 CpGs (each lane) are ordered in an unbiased manner according to the size of the MS-AFLP fingerprint
band. The five MSP loci are evenly distributed after every five loci. In the bottom panel, the 35 CpG loci and the 32 colorectal cancers are sorted out by
decreasing methylation frequency.
B: The distribution of CpG epigenetic alterations is dependent on the number of loci analyzed, in contrast with the independency of MSI-genetic alterations.
The tumors are distributed by percentage of methylated and mutated loci analyzing 5 (top) to 35 (bottom) CpGs. The microsatellite markers are also
distributed in groups of five. The DNA methylation data is derived from the top panel of Figure 3A and the mutation data from Figure 3C. The first five loci
in this panel correspond to the first 5 CpGs and the first 5 microsatellite markers at the left in panel A top and panel C, respectively. Scale on the X axis:
0 ? tumors with no hypermethylation or mutation (0%); 0.2 ? tumors with one or more loci altered up to 20%; 0.4 ? tumors with more than 20% up to 40%
loci altered, and so on. P values were obtained from the statistical comparison of observed values with those predicted for a normal distribution.
C: MSI is discontinuous in colon cancer. The extent of MSI was analyzed using 25 microsatellite markers (5 mononucleotide and 20 dinucleotide markers)
in 61 colorectal cancer specimens, which are sorted out by mutation frequency.
al., 1999a). In almost all subsequent publications (Toyota et al.,
1999b; Ueki et al., 2000; Shen et al., 2002; Rashid et al., 2001;
Chan et al., 2002), additional groups such as CIMP-I (intermedi-
ate)or CIMP-L(low) wereusedto embracetumors withinterme-
diate levels of methylation. Despite these difficulties, the classi-
fication of CIMP-positive tumors is usually done by selecting
tumors with 2–3 altered genes of the commonly used 3–6 loci.
This is arbitrary, as recognized by the same authors of these
less, the CIMP model remains generally supported even when
the experimental data appears inconsistent with the concept
(Ueki et al., 2000; Rashid et al., 2001; Shen et al., 2002, 2003;
van Rijnsoever et al., 2002; Chan et al., 2002).
cancer, as well as the data from others (Eads et al., 2001;
Hawkins et al., 2002; Shen et al., 2002, 2003), show that it is
impossible to draw a precise borderline between CIMP-positive
and -negative tumor groups due to the gradual distribution pat-
tern of CpG island somatic hypermethylation, which is far from
the bimodal distribution reported in the original paper (Toyota
et al., 1999a). The reason for this discrepancy might partly rest
on the criterion that originated the CIMP in the initial report.
CANCER CELL : AUGUST 2003125
A R T I C L E
Figure 4. Genetics, but not epigenetics, underlies colon cancer clinicopathological features
A: CpG island methylation associates to proximal location but not to any of the other phenotype parameters. Tumors with weak methylation (CIMP?)
were compared with those with intense methylation (CIMP?) using the MSP data from the 207 cases. Numbers on bars represent the actual number of
cases, and p values less than 0.05 are highlighted. The borderline between CIMP? and CIMP? was derived by criterion 1 elaborated in Figure 1. Criterion
2 also yielded similar results, with only the site showing significant right side predominance (p ? 0.00001). The MSI/CIMP relationship also remained significant
(p ? 0.004).
B: Mutator phenotype exhibits strong associations with many parameters, including right side preference. The panel demonstrates the results from the 207
cases used for the MSP analysis. The distribution of MSI-positive/negative tumors regarding tumor progression was 3/11 (21%); 12/57 (17%); 11/63 (15%); 1/35
(2.7%); and 0/5 (0%) for Dukes’ A:B:C:D: liver metastases, respectively (p ? 0.024?2test for trends). Results obtained from a larger tumor sample for which
the MSI status has been previously determined (Malkhosyan et al., 2000) showed that the significant associations of MSI-positive tumors extended to gender
(female predominance, p ? 0.0017); to Dukes’ stage (nonmetastatic predominance, p ? 0.003); and K-ras (wild-type predominance, p ? 0.0001). APC
suppressor gene mutations were also analyzed in a subset of tumors with a significantly lower mutational incidence in MSI-positive cancers (14 APC mutant
and 36 wild-type in MSI-positive tumors versus 36 APC mutant and 37 wild-type in MSI-negative tumors, p ? 0.017).
C: CIMP does not differentiate among colon cancers. MSI-positive cases were eliminated from the analysis.
D: MSI is dominant over CIMP. MSI-positive and MSI-negative tumors were compared using only cases with intense methylation (CIMP?) according to
criterion 1 (Figure 1).
Analysis of individual CpG islands (not shown) showed that methylation of hMLH1 correlated with right side (p ? 0.0004) and female predominance (p ?
0.06),oldage (p?0.003),poorlydifferentiatedhistological phenotype(p?0.0002),andlow mutationalincidenceatK-rasandP53. Similarly,p16methylation
correlated with right side, poorly differentiated phenotype, and methylated p14 and MGMT exhibited right side predominance. These characteristics
overlapped with the features of MSI-positive colon cancer. Except right side location, these differences lost their statistical significance in multiple regression
analysis and univariate analysis after MSI-positive tumors were eliminated (not shown).
Mutation analysis for K-ras and p53 was performed as previously described (Ionov et al., 1993; Yamamoto et al., 1997, 1999). Somatic APC mutations (from
codons 865 to 1590, comprising the mutation cluster region in exon 15) were identified and characterized by SSCP and sequencing of normal-tumor DNA
as described (Yamamoto et al., 1997, 1999). p values were calculated by Fisher’s exact test or Chi square with Yates correction.
The CIMP concept was reached after selecting particular
types of CpG islands showing methylation, those denominated
classC orcancer specific.ClassA, orage-specific, CpGislands
the distinction of class A from class C CpG islands depended
on whether methylation could be detected in normal tissues,
while the semiquantitative method utilized for detection was not
very sensitive (Toyota et al., 1999a; Shen et al., 2002, 2003). A
CpG island showing a faint band in the MSP experiments from
the normal tissue DNA was classified as belonging to class A,
CANCER CELL : AUGUST 2003
A R T I C L E
Figure 5. Gradient of CpG island DNA methyla-
tion and MSI in colon cancers in the large in-
The tumors are classified for CIMP according to
criterion 1 (Figure 1).
but another CpG island failing to reveal a band was classified
as class C. However, the CpG islands initially classified as type
C have been subsequently shown to be methylated in normal
tissues when the sensitivity of the detection assays was in-
creased (Kuismanen et al., 1999; Shen et al., 2003). Moreover,
these class C loci also have been found to have a tight connec-
tion with aging, which was a peculiarity of the class A CpG
islands (Kuismanen et al., 1999; Malkhosyan et al., 2000; Shen
et al., 2003; this work). In other words, a class C CpG island is
no more than a disguised class A CpG island.
There is no clear boundary between class A and class C
CpG islands, and they follow a gradual distribution regarding
their presence and detectability in normal tissues. Furthermore,
a CpG island that is classified as class C in one type of tumor
is classified as class A in another type of tumor (Eads et al.,
2001; Ueki et al., 2000). But if there is hypermethylation of a
particular CpG island in any normal tissue, this means that it is
not cancer specific. The restriction of some class C CpG islands
to a particular tumor type would lead to the enigmatic concept
of a tissue-specific and tumor-specific methylator phenotype.
untenable concept. The contradictory nature of what CIMP has
Figure 6. Age is the only parameter where DNA hypermethylation is independent of MSI status
A: Average ages of each patient group of the 207 tumor cohort (Figure 4) were compared by Student’s t test.
B: The 207 patients were divided into seven groups according to age with a 10 year interval. The average of methylated loci per tumor (percentage of
the seven loci) was plotted for each age group.
CANCER CELL : AUGUST 2003 127
A R T I C L E
become can be exposed by the statement that the cancer-
specific hypermethylator phenotype can also occur in normal
tissue (Peltomaki et al., 2000).
The data from Figure 3 imply that up to 40% of all genomic
CpG islands undergo hypermethylation in some colon cancers.
This extrapolation is based on the unbiased nature of the NotI
sites analyzed by MS-AFLP DNA fingerprinting (Yamamoto et
al., 2001) and the calculation that the vast majority (?80%) of
NotI sites are located in CpG islands (Kutsenko et al., 2002).
The last premise for such an extrapolation is that there are no
structural or functional differences between CpG islands with
and without NotI sites. In this line, other different NotI sites
amplified with different primer/adaptor combinations also have
similar levels of hypermethylation (unpublished observations).
factors required for the protection of CpG islands from en-
acting factors would be required to cover a significant fraction
of all the genomic CpG islands. Therefore, it is more likely that
resent frequent and accumulative stochastic fluctuations (Laird,
2003) occurring prior to and/or during tumorigenesis.
Many of these individual random CpG sequences were hy-
permethylated in a majority of the colon and gastric tumors (six
different MS-AFLP bands were each hypermethylated in over
50% of the tumors, Figure 3A and data not shown). The results
also imply that widespread CpG island hypermethylation often
affects the same genomic loci in tumorigenesis. Therefore, if
different defects were to lead to hypermethylation of distinct
groups of structurally similar CpG islands (Laird 2003), this
would require the concomitant existence in each tumor of sev-
eral defective methylator genes. The results obtained by unsu-
pervised clustering that show groups of tumors with high and
low levels of concordant methylation (Yan et al., 2002) can be
better explained by arbitrary ordering of the clustered tumors
and arbitrary selection of subclusters of CpG islands.
Nevertheless, even though no CIMP underlies the common
somatic methylation alterations observed in many tumors, there
is an unequal distribution of methylation alterations following
an increasing gradient from the distal to proximal colon (Figure
5). Another critical variable is the age. Methylation of all CpG
islands analyzed in colorectal cancer demonstrated an inclina-
tion toward older age (Figure 6). Together with other reports
(Kuismanen et al., 1999; Wiencke et al., 1999; Malkhosyan et
al., 2000; van Rijnsoever et al., 2002; Shen et al., 2002), these
results show that hypermethylation of most CpG islands is a
process strongly age dependent in gastrointestinal tumorigene-
sis. Thus, proximal colon cancers in old patients have more
methylation than distal cancers in young patients (Figure 5).
The reason underlying this distortion remains mysterious,
but it implies a disintegration of the epigenetic code during
aging in tissues with high cell turnover. However, to assign a
methylator phenotype to the increased gradient of methylation
in the proximal colon of old cancer patients (from whatever
unknown reasons) would be equally inadequate as to assign a
mutator phenotype to a concentration gradient of carcinogens
DNA hypermethylation at multiple loci correlates with MSI-
positive colon cancer (Figure 4A). This is because many tumors
the determinant genotypic feature that propels the phenotypic
differences of colorectal cancer is MSI and mutator phenotype
(Figure 4). Our results show that differences in genotype and
phenotype (K-ras and p53 mutations, differentiation and pro-
gression) are independent of the status of CpG island methyla-
tion (Figure 4D). On the other hand, if MSI-positive cancers are
eliminated, the remaining tumors with higher DNA methylation
were indistinguishable from the group of tumors with lower
methylation, regardless of the criteria for their segregation (Fig-
The distinctive features that MMP tumors display compared
with tumors without MSI is due to the particular spectrum of
mutated cancer genes in MSI-positive tumors (Markowitz et
al., 1995; Rampino et al., 1997; Lindblom, 2001). MMP tumors
display a low incidence of mutations in the APC and P53 tumor
suppressor genes and K-ras oncogene, prototypical for colon
cancer (Ionov et al., 1993; Perucho et al., 1994; Kim et al., 1994;
Olschwang et al., 1997; Breivik et al., 1997). Instead, MSI-posi-
tive colon tumors carry a plethora of different mutated genes,
tumors (Markowitz et al., 1995; Rampino et al., 1997; Woerner
et al., 2003). This is because in a MMR deficiency background,
mutations occur preferentially in genes with simple repeats in
their coding or regulatory sequences (Perucho, 1996; Suzuki et
al., 2002). The differences in genotype can explain the differ-
ences in phenotype displayed by MMP tumors, such as poorly
differentiated histological features, a less advanced stage of
tumor progression, and a better survival (Ionov et al., 1993;
Thibodeau et al., 1993; Boland et al., 1998; Elsaleh et al., 2000).
Theanatomical distributionof MSItumors overlapsthe pref-
erential localization of hypermethylation in the proximal colon
(Figures 4 and 6). The reason for the exacerbated asymmetry
in anatomical location for both mutator and methylator pheno-
types is intriguing. HNPCC tumors also show a preferential
localization in the proximal region of the large bowel, although
the disparity (about 70%) (Kuismanen et al., 2000) is not so
remarkable as in sporadic tumors (about 90%) (Kuismanen et
al., 2000; Figure 4).
tumors and CpG island methylation tumors could be due to an
intrinsic asymmetry in some critical cellular processes between
proximal and distal colon, such as stem cell renewal and mitotic
activity (Lipkin et al., 1962; Potten et al., 1992). A higher mitotic
activity may be sufficient to increase the probability of occur-
rence of both mutator mutations (somatic structural alterations
inactivating MMR genes) and epigenetic silencing of hMLH1,
and more importantly, the necessary additional cell replications
before neoplastic transformation can eventually occur (Ionov et
al., 1993; Perucho et al., 1994; Perucho, 1996; Tsao et al., 2000).
The age dependence of MSI-positive tumors with hMLH1 meth-
ylation is also consistent with this hypothesis. The difference in
incidence of MSI-positive proximal tumors between hereditary
and sporadic cases could be explained by the lower depen-
dence of the hereditary cancers on mitogenesis, as these can-
This hypothesis predicts that MSI-positive tumors with low
regardless of their hereditary or nonhereditary origins. Our data
CANCER CELL : AUGUST 2003
A R T I C L E
microsatellite repeats, including mono- and dinucleotide repeats (the list of
the loci and PCR primer information available upon request). Up to 30% of
the tumors shifted from MMP negative to the category of MSI-L (as defined
by the Bethesda Criteria of tumors with at least one and no more than 40%
mutated microsatellite markers) after analysis of 75 dinucleotide markers.
We extrapolated that all tumors would fall into the MSI-L category after
analysis of 200–250 dinucleotide repeats. Therefore, MSI-L cancers were
grouped together with the MMP-negative.
are consistent with this hypothesis although corroboration with
imal versus 1 MSI-positive/64 MSI-negative distal (p ? 0.020)
in the CIMP-negative group and 18/36 versus 1/22 in the CIMP-
positive group (p ? 0.007; Figure 4D).
ylation of hMLH1 remainsto beelucidated. However,the appar-
ent lack of association of MSI with aging is explained by the
heterogeneous mixture of familial and sporadic tumors in unse-
lected colorectal cancer series. Familial cancers occur in
younger patients while silencing and methylation in older tumor
patients, and the two extremes neutralize each other (Malkho-
syan et al., 2000).
independent, the genotype and phenotype features of colon
cancers with MSI supersede those displayed by the high hyper-
methylation tumor group. There is no clear explanation for the
association of MSI and CpG methylation with gender (Breivik
et al., 1997; Malkhosyan et al., 2000; Elsaleh et al., 2000), and
environmental and genomic factors may play roles. However,
anatomical location, gender, and age preempt cancer develop-
ment and fall beyond cancer cell genotype-phenotype relation-
In conclusion, after its manifestation, the microsatellite mu-
tator phenotype appears to determine the fate of the tumor
cell and to be dominant over the age-dependent epigenetic
there are two distinct groups of colon cancers defined by the
presence or absence of MSI and the underlying MMP. Whether
the MSI-negative tumors can be subdivided into two classes
depending upon their degree of epigenetic alterations remains
tion of the association of hypermethylation with aging (Issa et
of a phenomenon that seems to reach vast and deep ramifica-
tions. The task ahead is to find the mechanistic links between
Notwithstanding the importance of the disruption of the epige-
netic code for cancer development, our studies show that, re-
garding the neoplastic phenotype, genetics transcends epige-
Methylation-specific PCR (MSP)
MSP was carried out based on the original method developed by Herman
et al. (1996) with minor modification. Five hundred nanograms to one micro-
gram of genomic DNA were subjected to sodium bisulfite treatment, then
After bisulfite treatment and subsequent purification, DNA was amplified
separately using specific primers for unmethylated and methylated genomic
sequences. We used the published MSP primers and identical annealing
temperatures for hMLH1 (Yamamoto et al., 2001), p16 (Herman et al., 1996),
p14 (Shen et al., 2003), MGMT (Esteller et al., 1999), promoter 1A of APC
(Tsuchiya et al., 2000), and CDH1 (Graff et al., 1997, Island 3). Primers for
an anonymous CpG island (ACG) identified by MS-AFLP were designed
based on the BLAST search data (GenBank AC008425). The sequences for
unmethylated-specific primers were 5?-GGGTTTGGGTAAATTTGTTGTTT-3?
(nt 7733–7755) and 5?-AATCAAACACATCTCACA-3? (nt 7879–7862), which
amplify a 147 bp PCR product. For the methylated-specific reaction, the
5?-ATCAAACGCATCTCGCGA-3? (nt 7878–7861), which amplify a 141 bp
product. The annealing temperature for the ACG primers was set at 58?C. A
total of 10 ?l PCR mixture consisted of 1 ?l bisulfite-treated DNA, 1? PCR
buffer (16.6 mM of ammonium sulfate, 67.0 mM of Tris-HCl, 6.7 mM of
magnesium chloride, and 10 mM of 2-mercaptoethanol), 0.4 mM dNTP, 0.5
?M each primer, and 0.25 unit of Platinum Taq polymerase (Gibco BRL,
Rockville, Maryland). The amplification was started at 95?C for 5 min, followed
by 35 cycles of 95?C for 30 s, various annealing temperature (58?C –65?C) for
30 s, and 72?C for 30 s, and finished with 8 min of final extension at 72?C.
The PCR products were loaded on a 2.0% agarose gel and visualized under
ultraviolet illumination with ethidium bromide.
Methylation-sensitive amplified fragment length
MS-AFLP was performed as described previously (Yamamoto et al., 2001)
with slight modifications. Briefly, 1 ?g of genomic DNA was digested over-
night with 5 units of methylation-sensitive restriction endonuclease NotI
(Roche, Indianapolis, Indiana) and 2 units of methylation-insensitive MseI
(NE Biolabs, Beverly, Massachusetts) at 37?C. Two pairs of oligonucleotides
were annealed overnight at 37?C to generate NotI (5?-CTCGTAGACTGCG
TAGG-3? and 5?-GGCCCCTACGCAGTCTAC-3?) and MseI (5?-GACGAT
GAGTCCTGAG-3? and 5?-TACTCAGGACTCAT-3?) specific adaptors. The
digested DNA was ligated to 1.25 ?l each of 5 pmol/?l NotI and 50 pmol/
?l MseI adaptor using 1 unit of T4 DNA ligase (Roche) overnight at 16?C. A
primer complementary to the NotI adaptor (NotI primer, 5?-GACTGCG
TAGGGGCCGCG-3?) was labeled at the 5? end using32P-?ATP (NEN) and T4
polynucleotide kinase (Promega, Madison, Wisconsin). The adaptor-ligated
template DNA was amplified by PCR using the32P-labeled NotI primer and
not-labeled MseI primer (5?-GATGAGTCCTGAGTAAC-3?). A total of 20 ?l
PCR mixture contained 6 ng of32P-labeled NotI primer, 30 ng of MseI primer,
0.4 mMdNTP, and1 unitof AmpliTaqDNA polymerase (Perkin-Elmer,Foster
City, California) and 5 to 15 ng of template DNA. The PCR started at 72?C
for 30 s, 94?C for 30 s, then followed by 35 cycles of 94?C for 30 s, 52?C
for 30 s, and 72?C for 2 min. The final extension was performed for 10
min at 72?C. Each PCR sample was electrophoresed on a denaturing gel
(Sequagel-6, National Diagnostics, Atlanta, Georgia) after heat denaturing.
The gel was dried on a gel drier and exposed to an X-ray film.
DNA preparation and MSI screening
Surgically removed frozen tissues of colorectal cancers and paired adjacent
noncancerous tissue were obtained from the Cooperative Human Tissue
Network. The tumors analyzed represent a random subgroup of 207 tumors
from a consecutive collection of over 700 unselected colorectal cancers.
Genomic DNA was prepared by standard phenol-chloroform extraction and
ethanol precipitation. Radioactive PCR using32P-?dCTP (NEN-life Science
Products, Boston, Massachusetts) was carried out to analyze MSI as de-
scribed previously (Yamamoto et al., 1999). Microsatellite status was deter-
mined using two mononucleotide markers (BAT-26 and AP?3) and one
dinucleotide marker (D1S158). Tumors with deletion of multiple repeat units
in any of the two mononucleotide loci were defined as MMP positive, al-
though the vast majority of MMP-positive cancers exhibited mutation in both
the two mononucleotide markers as well as the dinucleotide marker (see
Figure 3C). This analysis classified the tumors into 28 (14%) MMP-positive,
and 154 (75%) MMP-negative. MMP-positive tumors in our classification
(Rampino et al., 1997; Perucho, 1999) correspond to MSI-H tumors in the
Bethesda classification (Boland et al., 1998). In addition, 22 (11%) displayed
sporadic alterations (deletion or insertion of one or two repeat units) only in
Scoring of methylation alteration in tumors by MS-AFLP
The sensitivity of MS-AFLP is highly dependent on the degree of normal
tissue contamination in the tumor specimens. A considerable amount of
normal cells in the tumor sample may obscure the intensity differences. We
were able to approximately estimate the contamination level in MSI-positive
samples from the autoradiographs of BAT-26 and other mononucleotide
CANCER CELL : AUGUST 2003 129
A R T I C L E
Chan,A.O.O.,Issa, J.P.J.,Morris,J.S.,Hamilton,S.R.,and Rashid,A.(2002).
Concordant CpG island methylation in hyperplastic polyposis. Am. J. Pathol.
ination level in MSI-negative cases, LOH analysis using 5–7 polymorphic
dinucleotide markers was conducted prior to MS-AFLP. Samples showing
a LOH pattern with the band corresponding to the lost allele with an intensity
less than 25% of the band corresponding to the retained allele at any marker
were regarded as adequate for MS-AFLP analysis.
of methylation alterations. Scoring of quantitative changes between normal
and tumor DNA were made by visual inspection as described previously
(Yamamoto et al., 2001). Scoring of signal intensity changes was done by
two observers independently. Only commonly detected alterations by the
two independent investigators were considered for the analysis. Further-
more, only changes commonly detected in at least two independent finger-
prints with duplicates of DNA from normal and tumor tissue were included.
Eads, C.A., Lord, R.V., Wichramasinghe, K., Long, T.I., Kurumboor, S.K.,
Bernstein, L., Peters, J.H., DeMeester, S.R., DeMeester, T.R., Skinner, K.A.,
and Laird, P.W. (2001). Epigenetic patterns in the progression of esophageal
adenocarcinoma. Cancer Res. 61, 3410–3418.
Eden, A., Gaudet, F., Waghmare, A., and Jaenisch, R. (2003). Chromosomal
instability and tumorspromoted by DNA hypomethylation.Science 300, 455.
Elsaleh, H., Joseph, D., Grieu, F., Zeps, N., Spry, N., and Iacopetta, B.
(2000). Association of tumour site and sex with survival benefit from adjuvant
chemotherapy in colorectal cancer. Lancet 355, 1745–1750.
Esteller, M., Hamilton, S.R., Burger, P.C., Baylin, S.B., and Herman, J.G.
(1999). Inactivation of the DNA repair gene O6-methylguanine-DNA methyl-
transferase by promoter hypermethylation is a common event in primary
human neoplasia. Cancer Res. 59, 793–797.
Bands showing intensity alterations in MS-AFLP fingerprints were excised
from the gels and subjected to a second round PCR using the same primer
pair. After column-purification (QIAquick PCR purification kit, Qiagen, Valen-
cia, California), the PCR products were directly sequenced using ABI Prism
BigDye Terminator Cycle Sequence Ready Reaction Kit and a DNA se-
quencer (Perkin-Elmer). The sequences were subjected to a BLAST search
to identify their origins in the human genome.
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.
Feinberg, A.P., and Vogelstein, B. (1983). Hypomethylation distinguishes
genes of some human cancers from their normal counterparts. Nature 301,
Comparison of MS-AFLP with MS-PCR
To contrast the sensitivity and specificity of the two methods, one of the
MS-AFLP bands showing methylation alterationswas isolated, and the char-
acterized CpG island containing the NotI site (ACG) was further analyzed
by MS-PCR in parallel to the six loci examined in the panel of 206 tumors.
The results showed an 85% concordance between MS-AFLP and MS-PCR.
The same tumors that exhibited tumor-specific hypermethylation of the ACG
sequence by MS-AFLP also showed hypermethylation in tumor tissue by
Gardiner-Garden, M., and Frommer, M. (1987). CpG islands in vertebrate
genomes. J. Mol. Biol. 196, 261–282.
Graff, J.R., Herman, J.G., Myohanen, S., Baylin, S.B., and Vertino, P.M.
(1997).Mapping patternsof CpGislandmethylation innormal andneoplastic
cells implicates both upstream and downstream regions in de novo methyla-
tion. J. Biol. Chem. 272, 22322–22329.
Hawkins, N., Norrie, M., Cheong, K., Mokany, E., Ku, S.L., Meagher, A.,
O’Connor, T., and Ward, R. (2002). CpG island methylation in sporadic
colorectal cancers and its relationship to microsatellite instability. Gastroen-
terology 122, 1376–1387.
This work was sponsored by NIH grants R01 CA38579 and R37 CA63585.
We thank Drs. Hiroyuki Yamamoto and Zhe Piao for some dinucleotide
microsatellite instability analyses and Lloyd Slivka for help with statistical
Methylation-specific PCR: a novel PCR assay for methylation status of CpG
islands. Proc. Natl. Acad. Sci. USA 93, 9821–9826.
Iacopetta, B. (2003). Aberrant DNA methylation: have we entered the era of
more than one type of colorectal cancer? Am. J. Pathol. 162, 1043–1045.
Ionov, Y., Peinado, M.A., Malkhosyan, S., Shibata, D., and Perucho, M.
(1993). Ubiquitous somatic mutations in simple repeated sequences reveal
a new mechanism for colonic carcinogenesis. Nature 363, 558–561.
Received: May 21, 2003
Revised: July 15, 2003
Published: August 25, 2003
Issa, J.P., Ottaviano, Y.L., Celano, P., Hamilton, S.R., Davidson, N.E., and
Baylin, S.B. (1994). Methylation of the oestrogen receptor CpG island links
ageing and neoplasia in human colon. Nat. Genet. 7, 536–540.
Ahuja, N., Mohan, A.L., Li, Q., Stolker, J.M., Herman, J.G., Hamilton, S.R.,
Baylin,S.B.,and Issa,J.P.J.(1997).Associationbetween CpGislandmethyl-
ation and microsatellite instability in colorectal cancer. Cancer Res. 57,
Jenuwein, T., and Allis, C.D. (2001). Translating the histone code. Science
Jones, P.A., and Laird, P.W. (1999). Cancer epigenetics comes of age. Nat.
Genet. 21, 163–167.
Baylin, S., and Bestor, T.H. (2002). Altered methylation patterns in cancer
cell genomes: cause or consequence? Cancer Cell 1, 299–305.
Jones, P.A., and Baylin, S.B. (2002). The fundamental role of epigenetic
events in cancer. Nat. Rev. Genet. 3, 415–428.
Bird, A. (2001). Molecular biology. Methylation talk between histones and
DNA. Science 294, 2113–2115.
Kane, M.F., Loda, M., Gaida, G.M., Lipman, J., Mishra, L., Goldman, H.,
Jessup, J.M., and Kolodner, R. (1997). Methylation of the hHMLH1 promoter
correlates with lack of expression of hHMLH1 in sporadic colon tumors and
mismatch repair-defective human tumor cell lines. Cancer Res. 57, 808–811.
Boland, C.R., Thibodeau, S.N., Hamilton, S.R., Sidransky, D., Eshleman,
J.R., Burt, R.W., Meltzer, S.J., Rodriguez-Bigasu, M.A., Fodde, R., Ranzani,
G.N., and Srivastava, S. (1998). A national cancer institute workshop on
microsatellite instability for cancer detection and familial predisposition: De-
velopment of international criteria for the determination of microsatellite
instability in colorectal cancer. Cancer Res. 58, 5248–5257.
Kim, H., Jen, J., Vogelstein, B., and Hamilton, S.R. (1994). Clinical and
pathological characteristics of sporadic colorectal carcinomas with DNA
replication errors in microsatellite instability. Am. J. Pathol. 145, 148–156.
Breivik,J.,and Gaudernack,G.(1999).Carcinogenesis andnaturalselection:
a new perspective to the genetics and epigenetics of colorectal cancer. Adv.
Cancer Res. 76, 187–212.
Kinzler, K.W., and Vogelstein, B. (1996). Lessons from hereditary colorectal
cancer. Cell 87, 159–170.
Breivik, J., Lothe, R.A., Meling, G.I., Rognum, T.O., Borresen-Dale, A.L., and
Gaudernack, G. (1997). Different genetic pathways to proximal and distal colo-
rectal cancer influenced by sex-related factors. Int. J. Cancer 74, 664–669.
Kolodner, R.D., and Marsischky, G.T. (1999). Eukaryotic DNA mismatch
repair. Curr. Opin. Genet. Dev. 9, 89–96.
Kuismanen, S.A., Holmberg, M.T., Salovaara, R., Schweizer, P., Aaltonen,
L.A., de la Chapelle, A., Nystrom-Lahti, M., and Peltomaki, P. (1999). Epige-
Cahill, D.P., Kinzler, K.W., Vogelstein, B., and Lengauer, C. (1999). Genetic
instability and Darwinian selection in tumors. Trends in Cell Biol. M57–M60.
CANCER CELL : AUGUST 2003
A R T I C L E
netic phenotypes distinguish microsatellite-stable and –unstable colorectal
cancers. Proc. Natl. Acad. Sci. USA 96, 12661–12666.
Strathdee, G., Appleton, K., Illand, M., Millian, D.W.M., Sargent, J., Paul, J.,
and Brown, R. (2001). Primary ovarian carcinomas display multiple methyl-
ator phenotypes involving known tumor suppressor genes. Am. J. Pathol.
Kuismanen, S.A., Holmberg, M.T., Salovaara, R., de la Chapelle, A., and
Peltomaki, P. (2000). Genetic and epigenetic modification of HMLH1 ac-
counts for a major share of microsatellite-unstable colorectal cancers. Am.
J. Pathol. 156, 1773–1779.
Suzuki, K., Dai, T., Suzuki, I., Dai, Y., Yamashita, K., and Perucho, M. (2002).
Low mutation incidence in polymorphic noncoding short mononucleotide
repeats in gastrointestinal cancer of the microsatellite mutator phenotype
pathway. Cancer Res. 62, 1961–1965.
Kutsenko, A.S., Gizatullin, R.Z., Al-Amin, A.N., Wang, F., Kvasha, S.M.,
sky, V.G., et al. (2002). NotI flanking sequences: a tool for gene discovery
and verification of the human genome. Nucleic Acids Res. 30, 3163–3170.
Thibodeau, S.N., Bren, G., and Schaid, D.J. (1993). Microsatellite instability
in cancer of the proximal colon. Science 260, 816–819.
Toyota, M.,and Issa,J.P. (2000).The role ofDNA hypermethylationin human
neoplasia. Electrophoresis 21, 329–333.
Laird, P.W. (2003). The power and the promise of DNA methylation markers.
Nat. Rev. Cancer 3, 253–266.
Toyota, M., Ahuja, N., Ohe-Toyota, M., Herman, J.G., Baylin, S.B., and Issa,
J.P.J. (1999a). CpG island methylator phenotype in colorectal cancer. Proc.
Natl. Acad. Sci. USA 96, 8681–8686.
Laird, P.W., Jackson-Grusby, L., Fazeli, A., Dickinson, S.L., Jung, W.E.,
Li, E., Weinberg, R.A., and Jaenisch, R. (1995). Suppression of intestinal
neoplasia by DNA hypomethylation. Cell 81, 197–205.
Toyota, M., Ahuja, N., Suzuki, H., Itoh, F., Ohe-Toyota, M., Imai, K., Baylin,
S.B., and Issa, J.P.J. (1999b). Aberrant methylation in gastric cancer associ-
ated withthe CpGisland methylatorphenotype. CancerRes. 59,5438–5442.
Lipkin, M., Sherlock, P., and Bell, B. (1963). Cell proliferation kinetics in the
gastrointestinal tract of man. Gastroenterology 45,721–729.
Lindblom, A. (2001). Different mechanisms in the tumorigenesis of proximal
and distal colon cancers. Curr. Opin. Oncol. 13, 63–69.
Toyota, M., Ohe-Toyota, M., Ahuja, N., and Issa, J.P.J. (2000). Distinct
genetic profiles in colorectal tumors with or without the CpG island methyl-
ator phenotype. Proc. Natl. Acad. Sci. USA 97, 710–715. Loeb,L.A.(1991).Mutatorphenotypemayberequiredformultistagecarcino-
genesis. Cancer Res. 51, 3075–3079.
Tsao, J.L., Yatabe, Y., Salovaara, R., Jarvinen, H.J., Mecklin, J.P., Aaltonen,
L.A., Tavare, S., and Shibata, D. (2000). Genetic reconstruction of individual
colorectal tumor histories. Proc. Natl. Acad. Sci. USA 97, 1236–1241.
Malkhosyan, S.R., Yamamoto, H., Piao, Z., and Perucho, M. (2000). Late
onset and high incidence of colon cancer of the mutator phenotype with
hypermethylated hHMLH1 gene in women. Gastroenterology 119, 598.
Tsuchiya, T., Tamura, G., Sato, K., Endoh, Y., Sakata, K., Jin, Z., Motoyama,
T., Usuba, O., Kimura, W., Nishizuka, S., et al. (2000). Distinct methylation
lia. Oncogene 19, 3642–3646.
Markowitz, S.D., Wang, J., Myeroff, L., Parsons, R., Sun, L., Lutterbaugh,
J., Fan, R.S., Zborowska, E., Kinzler, K.W., Vogelstein, B., et al. (1995).
Inactivation of the type II TGF-beta receptor in colon cancer cells with
microsatellite instability. Science 268, 1336–1338.
Turner, B.M. (2002). Cellular memory and the histone code. Cell 111, 285–291.
Olschwang, S., Hamelin, R., Laurent-Puig, P., Thuille, B., De Rycke, Y., Li,
genetic pathways in colorectal carcinogenesis. Proc. Natl. Acad. Sci. USA
Ueki, T., Toyota, M., Sohn, T., Yeo, C.J., Issa, J.P.J., Hruban, R.H., and
carcinoma. Cancer Res. 60, 1835–1839.
van Rijnsoever, M., Grieu, F., Elsaleh, H., Joseph, D., and Iacopetta, B.
(2002). Characterisation of colorectal cancers showing hypermethylation at
multiple CpG islands. Gut 51, 797–802.
Peltomaki, P., de la Chapelle, A., Kuismanen, S.A., Holmberg, M.T., and
Salovaara, R. (2000). Reply to Perucho M., Genetic and epigenetic modifica-
tion of HMLH1. Am. J. Pathol. 157, 1052–1053.
M.R., Miike, R., Ballesta, A., and Trias, M. (1999). Aberrant methylation of
p16INK4ain anatomic and gender-specific subtypes of sporadic colorectal
cancer. Cancer Epidemiol. Biomarkers Prev. 8, 501–506.
Perucho, M. (1996). Cancer of the microsatellite mutator phenotype. Biol.
Chem. 377, 675–684.
Perucho, M. (1999). Correspondence re: C.R. Boland et al. Cancer Res. 59,
Woerner, S.M., Benner, A., Sutter, C., Schiller, M., Yuan, Y.P., Keller, G.,
Bork, P., Doeberitz, M.K., and Gebert, J.F. (2003). Pathogenesis of DNA
Target genes. Oncogene 22, 2226–2235.
Perucho, M., Peinado, M.A., Ionov, Y., Casares, S., Malkhosyan, S., and
Stanbridge, E. (1994). Defects in replication fidelity of simple repeated se-
quences reveal a new mutator mechanism for oncogenesis. Cold Spring
Harb. Symp. Quant. Biol. 59, 339–348.
Yamamoto, H., Sawai,H., and Perucho, M. (1997).Frameshift somatic muta-
tions in gastrointestinalcancer of the microsatellitemutator phenotype. Can-
cer Res. 57, 4420–4426.
Potten, S.C., Kellett, M., Roberts, S., Rew, D., and Wilson, G. (1992). The
in vivo measurement of proliferation in normal human colorectal mucosa
using bromodeoxyuridine. Gut 33, 71–78.
Yamamoto, H., Perez-Piteira, J., Yoshida, T., Terada, M., Itoh, F., Imai,
K., and Perucho, M. (1999). Gastric cancers of the microsatellite mutator
ogy 116, 1348–1357.
Rampino, N., Yamamoto, H., Ionov, Y., Li, Y., Sawai, H., Reed, J.C., and
Perucho, M. (1997). Somatic frameshift mutations in the BAX gene in colon
cancers of the microsatellite mutator phenotype. Science 275, 967–969.
Yamamoto, F., Yamamoto, M., Soto, J.L., Kojima, E., Wang, E.N., Perucho,
M., Sekiya, T., and Yamanaka, H. (2001). NotI-MseI methylation-sensitive
amplified fragment length polymorphism for DNA methylation analysis of
human cancers. Electrophoresis 22, 1946–1956.
island methylation in colorectal adenomas. Am. J. Pathol. 159, 1129–1135.
Shen, L., Ahuja, N., Shen, Y., Habib, N.A., Toyota, M., Rashid, A., and
Issa, J.P.J. (2002). DNA methylation and environmental exposure in human
hepatocellular carcinoma. J. Natl. Cancer Inst. 94, 755–761.
Yamamoto, H., Itoh, F., Nakamura, H., Fukushima, H., Sasaki, S., Perucho,
M., and Imai, K. (2001). Genetic and clinical features of human pancreatic
ductal adenocarcinomas with widespread microsatellite instability. Cancer
Res. 61, 3139–3144.
Shen, L., Kondo, Y., Hamilton, S.R., Rashid, A., and Issa, J.P.J. (2003).
p14 methylation in human colon cancer is associated with microsatellite
instability and wild-type p53. Gastroenterology 124, 626–633.
Yamamoto, H., Imai, K., and Perucho, M. (2002). Gastrointestinal cancer of the
microsatellite mutator phenotype pathway. J. Gastroenterol. 37, 153–163.
Shibata, D., Peinado, M.A., Ionov, Y., Malkhosyan, S., and Perucho, M.
(1994). Genomic instability in repeated sequences is an early somatic event
in colorectal tumorigenesis that persists after transformation. Nat. Genet. 6,
Yan, P.S., Efferth, T., Chen, H., Lin, J., Rodel, F., Fuzesi, L., and Huang, T.
(2002). Use of CpG island microarrays to identify colorectal tumors with a
high degree of concurrent methylation. Methods 27, 162–169.
CANCER CELL : AUGUST 2003 131