Abstract

Cancer cells are characterized by a generalized disruption of the DNA methylation pattern involving an overall decrease in the level of 5-methylcytosine together with regional hypermethylation of particular CpG islands. The extent of both DNA hypomethylation and hypermethylation in the tumor cell is likely to reflect distinctive biological and clinical features, although no studies have addressed its concurrent analysis until now. DNA methylation profiles in sporadic colorectal carcinomas, synchronous adenoma–carcinoma pairs and their matching normal mucosa were analyzed by using the amplification of inter-methylated sites (AIMS) method. A total of 208 AIMS generated sequences were tagged and evaluated for differential methylation. Global indices of hypermethylation and hypomethylation were calculated. All tumors displayed altered patterns of DNA methylation in reference to normal tissue. On average, 24% of the tagged sequences were differentially methylated in the tumor in regard to the normal pair with an overall prevalence of hypomethylations to hypermethylations. Carcinomas exhibited higher levels of hypermethylation than did adenomas but similar levels of hypomethylation. Indices of hypomethylation and hypermethylation showed independent correlations with patient's sex, tumor staging and specific gene hypermethylation. Hierarchical cluster analysis revealed two main patterns of DNA methylation that were associated to particular mutational spectra in the K-ras and the p53 genes and alternative correlates of hypomethylation and hypermethylation with survival. We conclude that DNA hypermethylation and hypomethylation are independent processes and appear to play different roles in colorectal tumor progression. Subgroups of colorectal tumors show specific genetic and epigenetic signatures and display distinctive correlates with overall survival.

INTRODUCTION

Colorectal cancer is one of the best-studied systems of multistage human carcinogenesis. Epigenetic modification of DNA in the form of hypomethylation was included in early Vogelstein's tumor progression model together with a series of genetic alterations (1). DNA methylation is a post-replication modification predominantly found in cytosines of the dinucleotide CpG that is infrarepresented throughout the genome except at small regions named CpG islands (2). The pattern of DNA methylation in a given cell appears to be associated with the stability of gene expression states (3).

The biological significance of DNA hypomethylation, an early and common feature in colorectal cancer (4), is poorly understood (5). A relationship between global hypomethylation and genetic instability has been postulated (5,6). More recently, the attention of investigators has shifted to the study of cancer-associated regional hypermethylation at specific CpG islands and its association to transcriptional silencing (7,8) and loss of imprinting (9). Inspite of the large number of studies that have investigated cancer-associated hypermethylation in selected CpG islands, the obtention of global estimates of genome hypermethylation has been seldomly addressed (3,10,11).

Therefore, the roles of cumulated hypermethylation and hypomethylation in colorectal cancer progression and outcome are still unknown. By application of a methylome fingerprinting technique (amplification of inter-methylated sites, AIMS) (12), we have obtained information on the methylation status of more than 200 selected sequences in a series of colorectal carcinomas collected in a prospective design. We have also analyzed 11 adenomas synchronous to carcinomas of the former series. AIMS bands represent unique short DNA sequences (up to ∼1 kb long) flanked by two methylated SmaI sites (CCCGGG). Lack of methylation at either site will prevent amplification of the band. The AIMS generated sequences that can be tagged, isolated and individually characterized (1214). Global estimates of hypermethylation and hypomethylation in the tumor in regard to the paired normal tissue have been used to investigate the possible association of DNA methylation profiles with genetic and clinicopathological parameters. Our results unveil different roles for hypermethylation and hypomethylation in colorectal cancer progression and clinical behavior.

RESULTS

Selection of samples and AIMS bands

A total of 93 carcinomas and 11 adenomas with their paired normal tissues produced reproducible and consistent AIMS fingerprints and were included in the analysis of DNA methylation indices. Five additional cases showed inconsistent results in one or more AIMS experiments and were not considered for analysis. Most failures could be attributed to genomic DNA degradation.

DNA methylation profiles were obtained in three AIMS experiments performed with different sets of primers. On the basis of band display consistency (see Materials and Methods), 208 sequences were tagged and considered for analysis (set A: 77 bands, set B: 62 bands and set C: 69 bands). An illustrative example is shown in Figure 1. Differential display of certain bands was observed among normal tissues, indicating the polymorphic nature of their representation. Eighty-four tagged bands (40%) were present in all normal tissues and therefore considered as non-polymorphic. Sixty-four bands were low polymorphic (31%) and 60 (29%) were high polymorphic (see Materials and Methods). Because it has been noted that DNA methylation in normal tissues may be related to aging (15), we investigated the possible association of apparent polymorphisms with age. Fifteen of the high polymorphic sequences exhibited age-related display (14 were lost and one gained associated to aging) and were excluded from the analysis. At the end, 193 tagged bands were scored for differential display between normal and tumor tissues, allowing the calculation of indices of hypermethylation (increase in the intensity of the band) and hypomethylation (decrease in the intensity of the band) in each tumor in relation to its paired normal tissue.

Tagged bands behavior

On average, a given tagged band was informative in 88±5 (range 72–93) normal–tumor pairs. All tagged bands showed differential DNA methylation in at least one tumor when compared with normal tissue and showed a wide distribution of the hypermethylation/hypomethylation ratios (Fig. 2A). Almost half of the tagged bands (46%) underwent more hypomethylations than hypermethylations (>2-fold), whereas only 16% of the tagged bands showed the opposite behavior. Hypermethylation and hypomethylation showed a negative correlation (r=−0.196 and P=0.006), indicating that most bands tended to be either hypermethylated or hypomethylated (Fig. 2A). Recurrent changes affecting >25% of the tumors appeared as hypomethylated in 13 bands and hypermethylated in four bands. Polymorphic and non-polymorphic sequences showed similar behavior in regard to the likelihood of change between normal and tumor, although the most recurrent hypermethylations and hypomethylations occurred in polymorphic and non-polymorphic bands, respectively (Supplementary Material, Figs S1 and S2). Isolation and sequence characterization have been performed in a subset of the tagged bands. These data are supplied in Supplementary Material (Appendix A).

DNA methylation indices in colorectal carcinomas

The DNA methylation indices were obtained from the analysis of 185±20 tagged bands per case (range 123–193). Hypomethylation was more prominent than hypermethylation (P<0.001) (Table 1), although a wide distribution of the hypermethylation/hypomethylation ratios was observed (Fig. 2B). The main associations between the indices of DNA hypomethylation and hypermethylation and different genetic and clinicopathological parameters are summarized in Table 1. No significant differences in the DNA methylation indices were observed regarding age (data not shown), tumor localization, p53 and K-ras mutations and MSI. However, a remarkable sexual dimorphism was observed regarding both the hypermethylation and the hypomethylation indices. The hypermethylation index was higher in tumors with distant invasion, although it did not reach statistical significance (P=0.062). Note that a strong correlation was found between the hypermethylation index and the number of hypermethylated specific gene promoters, suggesting that generalized hypermethylation equally affects sequences generated by AIMS and the promoter regions of selected genes with a putative role in carcinogenesis. Individual case results are available in Supplementary Material (Appendix B).

DNA methylation indices along the adenoma–carcinoma transition

Indices of DNA hypermethylation and hypomethylation (in reference to normal tissue) were obtained in 11 cases with paired adenoma and carcinoma. The number of tagged bands per case was 174±17 (range 124–180). Carcinomas showed enhanced indices of hypermethylation when compared with adenomas (P=0.030), whereas hypomethylation was similar in both adenomas and carcinomas (Fig. 3A). These results indicate that hypomethylation is an early event in tumor progression, whereas hypermethylation accumulates in more advanced stages. This trend is clearly illustrated in the representation of the hypermethylation/hypomethylation ratios of the adenoma–carcinoma pairs (Fig. 3B). In regard to the methylation status of the panel of six CpG islands specifically investigated (hMLH1, APC, p16, p14, MGMT and LKB1) and previously reported to be hypermethylated in cancer (16), a similar number of methylated genes was observed in carcinomas (1.5±0.8) and adenomas (1.3±0.8). The methylation profile was not always coincident in the adenoma–carcinoma pairs (data not shown), suggesting the basic role of stochastic components behind the occurrence and clonal expansion of these alterations.

Profiles of DNA methylation in colorectal carcinomas

Next, we studied whether common patterns of DNA methylation could be used to classify tumors. Two-way hierarchical cluster analysis of AIMS data outlined two main groups of tumors and four main groups of tags (Fig. 4A). As expected, the two clusters of tumors were characterized by the prevalence of either hypermethylations (cluster 1) or hypomethylations (cluster 2) (Table 2). Thirty tagged bands exhibited differential behavior between the two groups of tumors: 12 tagged bands grouped as class 1 (Fig. 4A) showed hypermethylation in group 1 and hypomethylation or unchanged in group 2; 18 tagged bands grouped as class 2 showed hypomethylation in cluster 2 tumors and mostly unchanged in cluster 1. Bands with similar behavior in both clusters of tumors were classified as class 3 (hypermethylation) and 4 (hypomethylation). Interestingly, the two clusters of tumors displayed multiple distinctive molecular and phenotypic traits including tumor staging and specific spectra of mutations in the K-ras and p53 genes (Table 2). No statistically significant differences were observed for the rest of the parameters considered (data not shown).

DNA methylation indices as a prognostic factor

The indices of hypermethylation and hypomethylation were not indicators of prognosis in univariate and multivariate analyses. Cox regression analysis showed a trend for hypomethylation with poor prognosis and hypermethylation with good prognosis (data not shown), although the association was not significant. When the analysis was performed separately for the two clusters of tumors, a higher hypermethylation index was associated to good prognosis in tumors belonging to cluster 1, whereas in cluster 2, a higher hypomethylation index was associated to poor survival (Fig. 4B). Moreover, in cluster 1, the hypermethylation index (lower 50th percentile) was an independent predictor of survival (HR=3.2, CI 95% 1.2–8.3 and P=0.015) when compared with the Dukes' stage (C–D versus A–B, HR=2.5, CI 95% 0.9–6.8 and P=0.064).

DISCUSSION

Several conclusions may be drawn from the direct analysis of the raw data. Firstly, the DNA methylation pattern of the normal colonic mucosa is variable among individuals. The polymorphic nature of the band did not appear to affect its behavior in the tumor (hypomethylation or hypermethylation) and therefore we have made no distinction in the calculation of the indices, except for the age-related polymorphisms (8% of the screened CpG sites), which were excluded from further analysis. Secondly, the cancer-related changes tend to be unidirectional for each sequence: some bands preferentially show hypomethylations, whereas others display hypermethylations (Figs 2A and 4A). For most bands, a predominance of hypomethylation over hypermethylation was observed, which is in agreement with other studies (5,17). Note that many bands showed an erratic behavior, with hypermethylations in some tumors and hypomethylations in others. These results may be explained by the heterogeneous nature of the methylation in the sample due to allele-specific methylation (imprinting) or tissue heterogeneity (i.e. endothelial tissue may show different patterns than epithelial tissue). Therefore, the increase (or decrease) in the intensity of a band in the tumor when compared with the same preexisting band in the normal tissue sample should be interpreted as the homogenization of the methylated (or unmethylated) status rather than as a de novo methylation change.

DNA methylation indices in colorectal carcinomas

Our estimations of the overall abnormal DNA methylation in the tumor are a novel figure, and reveal the ubiquity of epigenetic alterations in colorectal neoplasia together with its wide range from tumor to tumor (7–44% of the tagged bands). The fraction of hypermethylated CpG sites in colorectal adenomas and carcinomas when compared with normal tissue (∼10% of the screened sequences) is in the same range as data reported by using CpG island array approaches in different types of human tumors (10,11). The hypermethylation index was associated with the number of methylated gene promoters analyzed by MSP (Table 1). Therefore, we can postulate that both the sequences represented in AIMS fingerprints and the genes selected because of their putative role in tumorigenesis display a common signature of the same process. In this sense, the AIMS approach has been recently used successfully to identify new hypermethylated genes in colon cancer (13). Regarding overall hypomethylation, most of the data in the literature have been generated by determination of the 5dC content in the genome. Note that the reported demethylation levels in colorectal neoplasia (4,1821) are in similar ranges as our estimations, suggesting that the AIMS fingerprints are also representative of the genomic global methylation status.

The sexual dimorphism in the DNA methylation profile has been noted previously in the analysis of specific CpG islands, with higher frequencies of hypermethylation at specific loci in women (22,23). Conversely, we also report that hypomethylation is higher in men (Table 1). We have no clues on the causes and possible implications of the gender-specific differences, but they deserve further investigation.

As a whole, the indices of hypermethylation and hypomethylation displayed a continuous distribution and did not appear to differentiate subgroups of tumors with distinctive features. The CpG island methylator phenotype described by Toyota et al. (24) was not foreseen from our data. We also did not find the reported correlations between hypermethylation and right localization, MSI and K-ras mutations (22,24,25). The significance of this classification should be revised under the light of recent studies (26,27), whose conclusions are supported by our results.

DNA methylation changes along tumor progression

The comparison of the methylation profiles between adenomas and carcinomas has confirmed that hypomethylation is an early event in colorectal cancer (4,18,21,28), whereas cumulated hypermethylation is more prominent in carcinomas. The trend to accumulate methylations is also exhibited in carcinoma progression, where advanced Dukes' stages showed higher levels than early stages (Table 1). These trends are not clearly reflected in the methylation of specific CpG islands, although in most cases the number of hypermethylated CpG islands tends to be higher in carcinomas than in adenomas (our data) (21,24) and in high grade tumors (22). In any case, the hypermethylation profile of concurrent adenomas and carcinomas is fluctuating and probably reflects the concomitant implication of multiple factors including heterogeneity and the diversity of markers analyzed.

Because we are simultaneously detecting two opposite phenomena (hypermethylation and hypomethylation) in the same experiment, an apparent inverse correlation is obtained between the two indices of hypomethylation and hypermethylation; this result may indicate the alternative prevalence of one of the two converse processes in a group of tumors (the tails of the distribution). Nevertheless, the presence of both types of changes in all the tumors and the continuity of the hypermethylation/hypomethylation ratio distribution (Fig. 2B) suggests that gains and losses of DNA methylation are governed by different mechanisms and selective pressures and are therefore independent. Other studies have also noted the independent contribution of both processes to tumorigenesis in colorectal (21) and Wilms tumors (29).

Profiles of DNA methylation associate with molecular and clinicopathological features

We have defined groups of carcinomas by hierarchical clustering of the AIMS profiles. Tree-type classification of the tumors delineated two main branches that came up from the differential methylation of 30 tags. Tumors in each group showed, in addition to predictable differences in the hypermethylation/hypomethylation ratio, striking differences in the mutation profile of the K-ras and p53 genes. The hypermethylated tumors showed almost exclusively transition mutations (G>A) at codons 12 and 13 of the K-ras gene (Table 2). The association of MGMT promoter methylation with G>A transitions in the K-ras gene has been previously reported (3032). In our series, the MGMT gene was also more frequently methylated in cluster 1 (39%) than in cluster 2 (22%), but differences did not reach statistical significance (P=0.099). Hence, our data expand the association between the type of mutation in the K-ras gene and the methylation of MGMT promoter to a specific DNA methylation signature characterized by extensive hypermethylations. Preceding findings also sustain a relationship between the DNA methylation profile and the spectrum of mutations in the p53 gene. Transitions at non-CpG sites (more frequent in cluster 1) have also been linked to MGMT promoter hypermethylation (31,33,34). In addition, the prevalence of transitions at CpG sites in hypomethylating tumors suggests that endogenous mechanisms (called as defective repair of spontaneously deamynated 5mC) are likely to play a major role in p53 mutagenesis in this group of tumors (3538). Although our results are insufficient to sustain a categorization of tumors based on their DNA methylation profiles, the distinctive molecular associations here reported strongly support the involvement of hypomethylation and hypermethylation in alternative processes related to malignant transformation.

This analysis also provides a list of candidate targets showing generalized hypermethylations (Fig. 4, bands of class 3) or hypomethylations (bands of class 4), together with targets selectively affected in one or another cluster of tumors (classes 1 and 2). Ongoing studies in our laboratory are focused on the characterization and functional analysis of different recurrent alterations with a potential specific involvement in colorectal carcinogenesis.

DNA methylation and survival

To our knowledge, this is the first study investigating the prognostic value of global hypomethylation and hypermethylation in colorectal cancer. Although none of the indices of differential DNA methylation appears to have prognostic utility in itself, the alternative association of the hypermethylation and hypomethylation indices with patient outcome in each one of the clusters is consistent with the independent roles of hypermethylation and hypomethylation in tumor initiation and progression (39). The attainment of high levels of hypomethylation already in premalignant lesions together with its maintenance throughout tumor progression (18,21) suggests that this factor may play a key role in conferring the malignant potential since early stages.

In summary, the interplay of DNA hypermethylation and hypomethylation demonstrates two independent roles with significant associations with molecular and clinicopathological parameters in colorectal cancer. Characterization of DNA methylation signatures in tumors should be instrumental not only in the identification of markers with a potential applicability in diagnosis and prognosis, but also in the definition of the pathways of progression. Further studies should clarify the implications of genetic and epigenetic profiles in tumor management and treatment.

MATERIALS AND METHODS

Samples

A total of 98 colorectal carcinomas, with their paired non-adjacent areas of normal colonic mucosa, and 11 colorectal adenomas, synchronous with carcinomas included in the former series, were collected as fresh-frozen tissues within 2 h of removal and then stored at −80°C. All samples were obtained from the Hospital de la Santa Creu i Sant Pau (Barcelona, Spain). The study protocol was approved by the Ethics Committee. Colorectal carcinomas used in this study were characterized previously for microsatellite instability (MSI) (40), mutations in the p53 and K-ras genes (41), and the methylation status of the promoter region of the genes hMLH1, APC, p16, p14, MGMT and LKB1 by methylation-specific PCR (MSP) (19). Detailed individual data are available in Supplementary Material (Appendix B).

Analysis of differential DNA methylation by AIMS

AIMS method is based on the differential cleavage of isoschizomers with distinct methylation sensitivity and the selective amplification of the sequences flanked by two methylated SmaI (CCCGGG) sites that are ligated to an adaptor and a specific 3–4 nucleotide sequence (adjacent to the SmaI site) arbitrarily chosen by the investigator. Lack of methylation at either site prevents amplification of the band. A detailed description of the method may be found in Frigola et al. (12). All samples were analyzed in three independent AIMS experiments with different primer sets (sets A, B and C) using conditions previously described (12). In a preceding setting, four normal–tumor pairs were analyzed in quadruplicate to assess reproducibility of all three AIMS experiments. Only sharp bands that were reproducible and clearly distinguishable from the background were tagged and assessed for differential methylation in all the samples. Faint bands with inconsistent display due to small variations in gel electrophoresis quality were not included in the analysis. Double bands consisting of two strands of the same sequence where considered as a single tagged band. Because of the polymorphic display of some bands in the normal mucosa, we have defined three different types of bands according to their nature: non-polymorphic (bands present in all normal tissues), low polymorphic (present in >90% of the samples) and finally, high polymorphic (present in <90% of the samples). Referring to the last type, we studied a putative correlation with age. Tagged bands showing an age-related display in the normal tissue were not used for the calculation of indices of hypermethylation and hypomethylation in the tumor when compared with the normal tissue.

Tagged bands behavior

Differences in the intensity of tagged bands between the tumor and its paired normal tissue were ascertained by direct visual inspection of the film. Densitometric analysis was performed in a limited subset of experiments to validate band intensity differences (data not shown), but was considered unnecessary in normal–tumor comparisons as only clear changes were taken into account and considered as qualitative events. Therefore, a ‘behavior’ could be scored for all tagged bands in every comparison between the normal and the tumor. Three different behaviors were defined: ‘hypermethylation’ (increased intensity in the tumor), ‘hypomethylation’ (decreased intensity in the tumor) and ‘no change’ (no substantial differences of intensity). Selected DNA methylation changes detected by AIMS have been confirmed by bisulfite sequencing (12,13) (unpublished data).

DNA methylation indices

Indices of hypomethylation and hypermethylation in the tumor were calculated as the number of hypomethylated and hypermethylated tagged bands in the tumor (when compared to the normal pair) divided by the total number of tagged bands considered. A third index reflecting the sum of both indices (differential methylation index) was also calculated.

Statistical analysis

All results are expressed as mean ± SD. Statistical differences between variables were analyzed with unpaired/paired t-tests or analysis of variance (ANOVA), as appropriate. Contingency tables were analyzed by the Chi-square or Fisher's exact test. Survival curves were traced according to the Kaplan–Meier method using the 50th percentile as cutoff for groups of low and high indices of abnormal DNA methylation. The statistical significance between curves was tested using the log-rank test. Univariate and multivariate analyses were performed using Cox's proportional hazards model. Hazard ratios and their 95% confidence interval (CI) were derived from Cox's proportional models. All P-values are calculated from two-sided statistical tests. Hierarchical clustering using Euclidean distance between observations and complete linkage as agglomeration method was performed to analyze the presence of subsets of tumors with comparable patterns of methylation changes. Only tags with differential methylation in >25% of the tumors were included in cluster analysis.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online.

ACKNOWLEDGEMENTS

We thank Gemma Aiza and Mar Muñoz for excellent technical help. This work was supported by a grant from the Spanish Ministry of Education and Science (SAF2003/5821). J.F. was recipient of an FPU fellowship from the Ministry of Education and Science at the Universitat Autònoma de Barcelona.

Figure 1. AIMS analysis of two normal (N)–tumor (T) pairs. (A) A general image of the fingerprints generated with each of the three primer sets (SA, SB and SC). Boxed areas in (A) are shown enlarged in (B). (C) Band intensity changes in the tumor in regard to the paired normal tissue. Hypermethylations (gain of intensity) and hypomethylations are indicated by up and down arrowheads, respectively. No changes are represented by rectangles. Polymorphic bands are shown as empty geometric figures.

Figure 1. AIMS analysis of two normal (N)–tumor (T) pairs. (A) A general image of the fingerprints generated with each of the three primer sets (SA, SB and SC). Boxed areas in (A) are shown enlarged in (B). (C) Band intensity changes in the tumor in regard to the paired normal tissue. Hypermethylations (gain of intensity) and hypomethylations are indicated by up and down arrowheads, respectively. No changes are represented by rectangles. Polymorphic bands are shown as empty geometric figures.

Figure 2. (A) Distribution of the hypomethylation and hypermethylation rates in the 193 AIMS bands considered. (B) Distribution of the hypomethylation and hypermethylation indices in the 93 colorectal carcinomas analyzed.

Figure 2. (A) Distribution of the hypomethylation and hypermethylation rates in the 193 AIMS bands considered. (B) Distribution of the hypomethylation and hypermethylation indices in the 93 colorectal carcinomas analyzed.

Figure 3. (A) Box plot of hypermethylation and hypomethylation indices in paired adenomas and carcinomas. (B) Shift in the hypermethylation/hypomethylation ratio between synchronous adenomas (arrow origin) and carcinomas (arrow head). The log 2 transformation of the ratio is represented in the x-axis. Samples have been sorted (y-axis) according to the hypermethylation index of the adenoma (at the top, the lowest hypermethylation index).

Figure 3. (A) Box plot of hypermethylation and hypomethylation indices in paired adenomas and carcinomas. (B) Shift in the hypermethylation/hypomethylation ratio between synchronous adenomas (arrow origin) and carcinomas (arrow head). The log 2 transformation of the ratio is represented in the x-axis. Samples have been sorted (y-axis) according to the hypermethylation index of the adenoma (at the top, the lowest hypermethylation index).

Figure 4. (A) Tree-type classification and heat map representation of the tumors (columns) and bands (rows) in two-way hierarchical clustering. Hypomethylation is green-coded, hypermethylation is red-coded and no change is shown black. The two bars at right represent the average status of the four classes of tags in the two clusters of tumors following the same color-code. (B) Kaplan–Meier overall survival curves in the two clusters of tumors classified by their DNA methylation signatures. A high hypermethylation index (50th percentile) was an indicator of better outcome in cluster 1 tumors (upper panel). A high hypomethylation index (50th percentile) was an indicator of poor outcome in cluster 2 tumors (lower panel). Survival rates have been corrected by the number of survivors after each time-point. Curves have been traced to the last recorded event.

Figure 4. (A) Tree-type classification and heat map representation of the tumors (columns) and bands (rows) in two-way hierarchical clustering. Hypomethylation is green-coded, hypermethylation is red-coded and no change is shown black. The two bars at right represent the average status of the four classes of tags in the two clusters of tumors following the same color-code. (B) Kaplan–Meier overall survival curves in the two clusters of tumors classified by their DNA methylation signatures. A high hypermethylation index (50th percentile) was an indicator of better outcome in cluster 1 tumors (upper panel). A high hypomethylation index (50th percentile) was an indicator of poor outcome in cluster 2 tumors (lower panel). Survival rates have been corrected by the number of survivors after each time-point. Curves have been traced to the last recorded event.

Table 1.

Indices of DNA hypermethylation and hypomethylation in colorectal carcinomas

 Categoriesa Hypermethylation index P-valueb Hypomethylation index P-valueb 
All carcinomas  0.096±0.060  0.146±0.081  
Sex Female (40) 0.113±0.072 0.021 0.123±0.082 0.015 
 Male (53) 0.084±0.047  0.164±0.077  
Localization Left (66) 0.101±0.064 0.258 0.147±0.089 0.813 
 Right (27) 0.082±0.050  0.143±0.059  
Duke stage A–B (49) 0.085±0.053 0.062 0.157±0.079 0.192 
 C–D (44) 0.109±0.066  0.135±0.083  
p53 gene Wild-type (58) 0.096±0.060 0.916 0.140±0.077 0.210 
 Mutated (35) 0.095±0.062  0.162±0.089  
K-ras gene Wild-type (55) 0.097±0.054 0.728 0.154±0.081 0.231 
 Mutated (37) 0.093±0.070  0.133±0.081  
MSI status Stable (83) 0.096±0.062 0.464 0.149±0.082 0.660 
 Instable (5) 0.075±0.044  0.165±0.083  
Methylated genesc <3 (65) 0.084±0.056 <0.001 0.156±0.081 0.091 
 ≥3 (10) 0.159±0.059  0.106±0.091  
 Categoriesa Hypermethylation index P-valueb Hypomethylation index P-valueb 
All carcinomas  0.096±0.060  0.146±0.081  
Sex Female (40) 0.113±0.072 0.021 0.123±0.082 0.015 
 Male (53) 0.084±0.047  0.164±0.077  
Localization Left (66) 0.101±0.064 0.258 0.147±0.089 0.813 
 Right (27) 0.082±0.050  0.143±0.059  
Duke stage A–B (49) 0.085±0.053 0.062 0.157±0.079 0.192 
 C–D (44) 0.109±0.066  0.135±0.083  
p53 gene Wild-type (58) 0.096±0.060 0.916 0.140±0.077 0.210 
 Mutated (35) 0.095±0.062  0.162±0.089  
K-ras gene Wild-type (55) 0.097±0.054 0.728 0.154±0.081 0.231 
 Mutated (37) 0.093±0.070  0.133±0.081  
MSI status Stable (83) 0.096±0.062 0.464 0.149±0.082 0.660 
 Instable (5) 0.075±0.044  0.165±0.083  
Methylated genesc <3 (65) 0.084±0.056 <0.001 0.156±0.081 0.091 
 ≥3 (10) 0.159±0.059  0.106±0.091  

aNumbers in parentheses indicate number of cases in each group.

bTwo tailed t-test.

cSix genes were included in the analysis: hMLH1, APC, p16, p14, MGMT and LKB1.

Table 2.

Molecular and phenotypic characteristics of tumors classified by the DNA methylation profile

Parameter Categories Cluster 1 (n=47) Cluster 2 (n=46) P-valuea 
Hypermethylation index — 0.116±0.066 0.076±0.047 0.001 
Hypomethylation index — 0.098±0.057 0.196±0.071 <0.001 
DMI — 0.214±0.070 0.272±0.081 <0.001 
Dukes' stage A–B 19 30 0.022 
 C–D 28 16  
K-ras point mutationb Transition 17 11 0.032 
 Transversion  
 Negative 29 26  
p53 point mutationc Transition at CpG 12 0.034 
 Other point mutation  
 Negative 27 27  
Methylated genesd <3 26 34 0.076 
 ≥3  
Parameter Categories Cluster 1 (n=47) Cluster 2 (n=46) P-valuea 
Hypermethylation index — 0.116±0.066 0.076±0.047 0.001 
Hypomethylation index — 0.098±0.057 0.196±0.071 <0.001 
DMI — 0.214±0.070 0.272±0.081 <0.001 
Dukes' stage A–B 19 30 0.022 
 C–D 28 16  
K-ras point mutationb Transition 17 11 0.032 
 Transversion  
 Negative 29 26  
p53 point mutationc Transition at CpG 12 0.034 
 Other point mutation  
 Negative 27 27  
Methylated genesd <3 26 34 0.076 
 ≥3  

aTwo tailed t-test, Fisher's exact test or Chi-square as appropriate.

bMutations at codons 12 and 13.

cSeven tumors showing insertions or deletions in the coding sequence of the p53 gene were not included.

dSix genes were included in the analysis: hMLH1, APC, p16, p14, MGMT and LKB1.

References

1
Fearon, E.R. and Vogelstein, B. (
1990
) A genetic model for colorectal tumorigenesis.
Cell
 ,
61
,
759
–767.
2
Bird, A.P. (
1986
) CpG-rich islands and the function of DNA methylation.
Nature
 ,
321
,
209
–213.
3
Plass, C. (
2002
) Cancer epigenomics.
Hum. Mol. Genet.
 ,
11
,
2479
–2488.
4
Feinberg, A.P., Gehrke, C.W., Kuo, K.C. and Ehrlich, M. (
1988
) Reduced genomic 5-methylcytosine content in human colonic neoplasia.
Cancer Res.
 ,
48
,
1159
–1161.
5
Ehrlich, M. (
2002
) DNA methylation in cancer: too much, but also too little.
Oncogene
 ,
21
,
5400
–5413.
6
Eden, A., Gaudet, F., Waghmare, A. and Jaenisch, R. (
2003
) Chromosomal instability and tumors promoted by DNA hypomethylation.
Science
 ,
300
,
455
.
7
Jones, P.A. and Baylin, S.B. (
2002
) The fundamental role of epigenetic events in cancer.
Nat. Rev. Genet.
 ,
3
,
415
–428.
8
Esteller, M. (
2002
) CpG island hypermethylation and tumor suppressor genes: a booming present, a brighter future.
Oncogene
 ,
21
,
5427
–5440.
9
Feinberg, A., Cui, H. and Ohlsson, R. (
2002
) DNA methylation and genomic imprinting: insights from cancer into epigenetic mechanisms.
Semin. Cancer Biol.
 ,
12
,
389
.
10
Huang, T.H., Perry, M.R. and Laux, D.E. (
1999
) Methylation profiling of CpG islands in human breast cancer cells.
Hum. Mol. Genet.
 ,
8
,
459
–470.
11
Costello, J.F., Fruhwald, M.C., Smiraglia, D.J., Rush, L.J., Robertson, G.P., Gao, X., Wright, F.A., Feramisco, J.D., Peltomaki, P., Lang, J.C. et al. (
2000
) Aberrant CpG-island methylation has non-random and tumour-type-specific patterns.
Nat. Genet.
 ,
24
,
132
–138.
12
Frigola, J., Ribas, M., Risques, R.A. and Peinado, M.A. (
2002
) Methylome profiling of cancer cells by amplification of inter-methylated sites (AIMS).
Nucleic Acids Res.
 ,
30
,
e28
.
13
Paz, M.F., Wei, S., Cigudosa, J.C., Rodriguez-Perales, S., Peinado, M.A., Huang, T.H. and Esteller, M. (
2003
) Genetic unmasking of epigenetically silenced tumor suppressor genes in colon cancer cells deficient in DNA methyltransferases.
Hum. Mol. Genet.
 ,
12
,
2209
–2219.
14
Sadikovic, B., Haines, T.R., Butcher, D.T. and Rodenhiser, D.I. (
2004
) Chemically induced DNA hypomethylation in breast carcinoma cells detected by the amplification of intermethylated sites.
Breast Cancer Res.
 ,
6
,
30
.
15
Ahuja, N., Li, Q., Mohan, A.L., Baylin, S.B. and Issa, J.P. (
1998
) Aging and DNA methylation in colorectal mucosa and cancer.
Cancer Res.
 ,
58
,
5489
–5494.
16
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.
17
Issa, J.P. (
2002
) Epigenetic variation and human disease.
J. Nutr.
 ,
132
,
2388S
–2392S.
18
Goelz, S.E., Vogelstein, B., Hamilton, S.R. and Feinberg, A.P. (
1985
) Hypomethylation of DNA from benign and malignant human colon neoplasms.
Science
 ,
228
,
187
–190.
19
Esteller, M., Fraga, M.F., Guo, M., Garcia-Foncillas, J., Hedenfalk, I., Godwin, A.K., Trojan, J., Vaurs-Barriere, C., Bignon, Y.J., Ramus, S. et al. (
2001
) DNA methylation patterns in hereditary human cancers mimic sporadic tumorigenesis.
Hum. Mol. Genet.
 ,
10
,
3001
–3007.
20
Paz, M.F., Avila, S., Fraga, M.F., Pollan, M., Capella, G., Peinado, M.A., Sanchez-Cespedes, M., Herman, J.G. and Esteller, M. (
2002
) Germ-line variants in methyl-group metabolism genes and susceptibility to DNA methylation in normal tissues and human primary tumors.
Cancer Res.
 ,
62
,
4519
–4524.
21
Bariol, C., Suter, C., Cheong, K., Ku, S.L., Meagher, A., Hawkins, N. and Ward, R. (
2003
) The relationship between hypomethylation and CpG island methylation in colorectal neoplasia.
Am. J. Pathol.
 ,
162
,
1361
–1371.
22
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.
Gastroenterology
 ,
122
,
1376
–1387.
23
Laghi, L., Bianchi, P. and Malesci, A. (
2003
) Gender difference for promoter methylation pattern of hMLH1 and p16 in sporadic MSI colorectal cancer.
Gastroenterology
 ,
124
,
1165
–1166.
24
Toyota, M., Ahuja, N., Ohe-Toyota, M., Herman, J.G., Baylin, S.B. and Issa, J.P. (
1999
) CpG island methylator phenotype in colorectal cancer.
Proc. Natl Acad. Sci. USA
 ,
96
,
8681
–8686.
25
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.
26
Yamashita, K., Dai, T., Dai, Y., Yamamoto, F. and Perucho, M. (
2003
) Genetics supersedes epigenetics in colon cancer phenotype.
Cancer Cell
 ,
4
,
121
–131.
27
Sieber, O.M., Heinimann, K. and Tomlinson, I.P. (
2003
) Genomic instability—the engine of tumorigenesis?
Nat. Rev. Cancer
 ,
3
,
701
–708.
28
Wynter, C.V., Walsh, M.D., Higuchi, T., Leggett, B.A., Young, J. and Jass, J.R. (
2004
) Methylation patterns define two types of hyperplastic polyp associated with colorectal cancer.
Gut
 ,
53
,
573
–580.
29
Ehrlich, M., Jiang, G., Fiala, E., Dome, J.S., Yu, M.C., Long, T.I., Youn, B., Sohn, O.S., Widschwendter, M., Tomlinson, G.E. et al. (
2002
) Hypomethylation and hypermethylation of DNA in Wilms tumors.
Oncogene
 ,
21
,
6694
–6702.
30
Esteller, M., Toyota, M., Sanchez-Cespedes, M., Capella, G., Peinado, M.A., Watkins, D.N., Issa, J.P., Sidransky, D., Baylin, S.B. and Herman, J.G. (
2000
) Inactivation of the DNA repair gene O6-methylguanine-DNA methyltransferase by promoter hypermethylation is associated with G to A mutations in K-ras in colorectal tumorigenesis.
Cancer Res.
 ,
60
,
2368
–2371.
31
Kim, S.G., Chan, A.O., Wu, T.T., Issa, J.P., Hamilton, S.R. and Rashid, A. (
2003
) Epigenetic and genetic alterations in duodenal carcinomas are distinct from biliary and ampullary carcinomas.
Gastroenterology
 ,
124
,
1300
–1310.
32
Lees, N.P., Harrison, K.L., Hall, C.N., Margison, G.P. and Povey, A.C. (
2004
) Reduced MGMT activity in human colorectal adenomas is associated with K-ras GC>AT transition mutations in a population exposed to methylating agents.
Carcinogenesis
 ,
25
,
1243
–1247.
33
Esteller, M., Risques, R.A., Toyota, M., Capella, G., Moreno, V., Peinado, M.A., Baylin, S.B. and Herman, J.G. (
2001
) Promoter hypermethylation of the DNA repair gene O(6)-methylguanine-DNA methyltransferase is associated with the presence of G : C to A : T transition mutations in p53 in human colorectal tumorigenesis.
Cancer Res.
 ,
61
,
4689
–4692.
34
Yin, D., Xie, D., Hofmann, W.K., Zhang, W., Asotra, K., Wong, R., Black, K.L. and Koeffler, H.P. (
2003
) DNA repair gene O6-methylguanine-DNA methyltransferase: promoter hypermethylation associated with decreased expression and G : C to A : T mutations of p53 in brain tumors.
Mol. Carcinog.
 ,
36
,
23
–31.
35
Tornaletti, S. and Pfeifer, G.P. (
1995
) Complete and tissue-independent methylation of CpG sites in the p53 gene: implications for mutations in human cancers.
Oncogene
 ,
10
,
1493
–1499.
36
Schmutte, C., Yang, A.S., Nguyen, T.T., Beart, R.W. and Jones, P.A. (
1996
) Mechanisms for the involvement of DNA methylation in colon carcinogenesis.
Cancer Res.
 ,
56
,
2375
–2381.
37
Schmutte, C., Yang, A.S., Beart, R.W. and Jones, P.A. (
1995
) Base excision repair of U : G mismatches at a mutational hotspot in the p53 gene is more efficient than base excision repair of T : G mismatches in extracts of human colon tumors.
Cancer Res.
 ,
55
,
3742
–3746.
38
Laird, P.W. and Jaenisch, R. (
1994
) DNA methylation and cancer.
Hum. Mol. Genet.
 ,
3
,
1487
–1495.
39
Gaudet, F., Graeme, J.G., Eden, A., Jackson-Grusby, L., Dausman, J., Gray, J.W., Leonhardt, H. and Jaenisch, R. (
2003
) Induction of tumors in mice by genomic hypomethylation.
Science
 ,
300
,
489
–492.
40
Gonzalez-Garcia, I., Moreno, V., Navarro, M., Marti-Rague, J., Marcuello, E., Benasco, C., Campos, O., Capella, G. and Peinado, M.A. (
2000
) Standardized approach for microsatellite instability detection in colorectal carcinomas.
J. Natl Cancer Inst.
 ,
92
,
544
–549.
41
Tortola, S., Marcuello, E., Gonzalez, I., Reyes, G., Arribas, R., Aiza, G., Sancho, F.J., Peinado, M.A. and Capella, G. (
1999
) p53 and K-ras gene mutations correlate with tumor aggressiveness but are not of routine prognostic value in colorectal cancer.
J. Clin. Oncol.
 ,
17
,
1375
–1381.