CpG Island Methylator Phenotype in Colorectal Cancers: Comparison of the New and Classic CpG Island Methylator Phenotype Marker Panels

Promoter CpG island hypermethylation is an important epigenetic change associated with gene silencing and is now recognized as an alternative mechanism to mutations for the inactivation of tumor suppressor genes or tumor-related genes in cancer cells.1 CpG island methylator phenotype (CIMP) is a subset of colorectal carcinoma (CRC) characterized by concordant hypermethylation of multiple promoter CpG island loci2 and peculiar clinicopathologic findings, such as close association with microsatellite instability (MSI), proximal tumor location, female preponderance, high frequency of mutation of the protooncogene BRAF, and low frequency of mutation of the proto- oncogene KRAS.2-9 Furthermore, CIMP-positive (CIMP+) CRCs have been reported to be associated with poor clinical outcomes.10-12

A new CIMP marker panel comprising the genes CACNA1G, IGF2, NEUROG1, RUNX3, and SOCS1 was recently introduced by the Laird laboratory7 as superior to the original CIMP marker panel, comprising CDKN2A, MINT1, MINT2, MINT31, and MLH1,5,13,14 in that the new CIMP panel easily detected a heavily methylated subset of CRCs that encompasses almost all BRAF mutants and sporadic MSI-high CRCs. However, prognostic power was not compared between the 2 CIMP panels. Moreover, both CIMP panels were tested on the platform of real-time- based, quantitative, methylation-specific polymerase chain reaction (PCR; MethyLight technology) rather than with traditional methylation-specific PCR (MSP). Methylated alleles of most CpG island loci are heterogeneous with respect to the methylation status of individual CpG sites,15-17 and thus the MethyLight assay detects a subset of DNA alleles that are heavily methylated in all CpG sites targeted by both primers and probe.18,19 The MethyLight assay cannot, however, detect alleles that are not fully methylated in the targeted CpG sites but are sufficiently methylated to inactivate gene expression.20 Traditional MSP detects both fully methylated alleles and substantially but incompletely methylated alleles. Thus, it remains to be determined whether this new CIMP panel outperforms the original panel using traditional MSP.

In the present study, we used traditional MSP to determine the CIMP status of 130 sporadic CRCs based on both the new and classic CIMP panels and explored the differences in clinicopathologic and molecular features between CIMP+ CRCs defined by each marker panel. Moreover, both CIMP panels were compared for their prognostic power.

MATERIALS AND METHODS

Tissue Samples

Our study was approved by the Seoul National UniversityHospital's Institutional Review Board. Tumor tissues were obtained from 130 patients with sporadic CRC who had undergone curative surgery at Seoul National University Hospital, Seoul, Korea, between 2001 and 2002. Formalin-fixed, paraffin-embedded tissues were used for DNA extraction. Through light microscopic examination, we marked tumor areas where tumor cells occupied 50% or more of all cells and represented the main histology and differentiation of the tumor. Ten serial 10-µm-thick histologic slides of methanol-fixed tumor and normal tissue blocks were used for manual microdissection. Dissected tissue samples were subjected to tissue lysis using proteinase K-containing lysis buffer. Mutation analysis of KRAS codons 12 and 13 and BRAF codon 600 and MSI analysis were previously performed and determined for these 130 samples.21 Clinicopathologic information, including age, sex, histologic type, tumor location, stage, and overall survival, was obtained from all patients. Tumor staging was based on the TNM staging system of the American Joint Committee on Cancer.22

Analysis of CpG Island Methylation Status

The classic CIMP marker panel, consisting of hMLH1, p16, MINT1, MINT2, and MINT31, and the novel CIMP-marker panel of CACNA1G, IGF2, NEUROG1, RUNX3, and SOCS1 were analyzed using MSP.2,5,7,13,14 For MSP assays, DNA was subjected to bisulfite modification, as described.23 Primer sequences and PCR conditions of all 10 loci for both methylated and unmethylated forms are shown in Table 1. To amplify the bisulfite-modified promoter sequence of p16 and hMLH1, we used a PCR mixture containing 1× PCR buffer (10mM Tris, pH 8.3; 50mM KCl; and 1.5mM MgCl2), deoxynucleotide triphosphates (each 0.2mM), primers (10 pmol each), and bisulfite-modified DNA (30-50 ng) in a final volume of 25 µL. For the amplifications of MINT1, MINT2, MINT31, CACNA1G, IGF2, NEUROG1, RUNX3, and SOCS1 we used a PCR mixture containing 1× PCR buffer (16.6mM (NH4)^sub 2^SO^sub 4^; 67mM Tris, pH 8.8; 6.7mM MgCl2; and 10mM -mercaptoethanol), deoxynucleotide triphosphate (each 1mM), primers (10 pmol each), and bisulfite-modified DNA (30- 50 ng) in a final volume of 25 µL. Reactions were hot started at 98°C for 5 minutes before addition of 0.75 units of Taq polymerase (Intron Company, Seoul, Korea). Amplifications were carried out in a thermal cycler (Perkin-Elmer, Foster City, Calif) for 33 cycles (40 seconds at 95°C, 50 seconds at variable temperatures according to primer, and 50 seconds at 72°C) and were given a final 10- minute extension. Polymerase chain reaction products were electrophoresed in 2.5% agarose gels and visualized under ultraviolet illumination after ethidium bromide staining. Samples showing signals more intense than that of the size marker (3.5 ng) were scored as methylated, and samples showing signal weaker than that of the size marker were repeated twice, and those giving consistently positive PCR results were scored as methylated. For each MSP reaction, we used placental DNA treated with M.SssI methyltransferase (New England Biolabs Inc, Beverly, Mass) or distilled water (without template DNA) as positive and negative controls, respectively.

Sequencing Analysis

The PCR products were purified using a JETSORB gel extraction kit (Genomed, Bad Oeynhausen, Germany) and cloned into the pCR2.1-TOPO vector (Invitrogen, Carlsbad, Calif). Plasmid DNA was extracted from individual clones by an alkaline lysis plasmid minipreparation. The inserted PCR fragments of 4 individual clones from each sample were sequenced with both M13 reverse and M13 (-20) forward primers using an ABI Prism 377 DNA Sequencer (Perkin-Elmer, Norwalk, Conn).

Statistical Analysis

All statistical analyses were done using SPSS software versions 12.0 (SPSS Inc, Chicago, Ill). Pearson -2 test or Fisher exact test was used to test associations between clinicopathologic variables and genetic or epigenetic alterations (CIMP, MSI, mutation of KRAS or BRAF). To compare the means of numeric variables, the Student t test was used. Survival curves were constructed using the Kaplan-Meier method, and differences between Kaplan-Meier survival curves were tested using the log-rank test. Survival was measured from the date of resection of CRC to the date of death or the last clinical review before August 29, 2007. The average follow-up time (from surgery to death or last follow-up) was 48.2 months (range, 3-74 months). P values & lt; .05 were accepted as being statistically significant.

RESULTS

We assessed 130 sporadic CRCs for the methylation status of 5 new CIMP markers and 5 classic CIMP markers using MSP (Figure 1), and Figure 2 presents representative examples of the MSP results. To validate the adequacy of MSP primers of new CIMP markers, we performed bisulfite genomic sequencing of the representative MSP samples for each marker. All of the sequenced MSP products showed extensive methylation of CpG sites between the MSP primer sequences as well as those CpG sites within the primer sequences. The vast majority of CpG sites of all tested genes exhibited methylation at a frequency greater than or equal to 75%.

Frequency of Methylation at Markers Using Classic CIMP or New CIMP Panels

We determined the methylation frequencies of each marker (Table 2) and the distribution of the numbers of methylated markers (Figure 3) in the new and classic CIMP panels. We observed methylation at 1 or more of 5 markers in classic and new CIMP panels in 51.5% and 60.8% of CRCs, respectively. The distribution of tumors by the numbers of methylated markers showed similar patterns in both CIMP panels and was not of bimodal distribution (Figure 3).

Clinicopathologic Characteristics of Classic and Novel CIMPs in CRCs

Classic CIMP was usually defined as positive when CRCs showed methylation at 2 or more of 5 classic CIMP markers,5,13,14 whereas new CIMP was defined as positive when CRCs exhibited methylation at 3 or more of 5 novel CIMP markers. However, in the present study, both CIMP panels showed similar distribution patterns of CRCs based on the numbers of methylated markers. Thus, to compare clinicopathologic or molecular features of classic CIMP+ CRCs with those of new CIMP+ CRCs, we set 2 separate cutoff values (2 of 5 and 3 of 5) in the determination of classic and new CIMP status. With a cutoff value of 2, classic CIMP+ tumors and new CIMP+ tumors comprised 23.1% (39/130) and 23.1% (39/130), respectively, whereas with a cutoff value of 3, classic CIMP+ tumors and new CIMP+ tumors were 18.5% (24/130) and 16.9% (22/130), respectively (Table 3).

With at least 3 markers methylated, new CIMP+ tumors were associated with proximal location (P=.02, Fisher exact test), poor differentiation (P=.05, Pearson -2 test), frequent MSI (P=.09, Fisher exact test), low frequency of KRAS mutation (P=.045, Fisher exact test), and high frequency of BRAF mutation (P< .001, Fisher exact test). The new CIMP panel outperformed the classic CIMP panel in detecting CRCs with poor differentiation and CRCs without KRAS mutation, whereas the classic CIMP detected CRCs with MSI more efficiently than the novel CIMP panel. With at least 2 markers methylated, both classic and novel CIMP+ CRCs were significantly associated with proximal tumor location, MSI, and BRAF mutation; however, the novel CIMP panel outperformed the classic CIMP panel in detecting these features.

Classic CIMP Status Is Significantly Implicated in Patient Survival

To compare the prognostication power of both marker panels, CRCs were divided into 4 molecular subtypes based on the status of CIMP and MSI: CIMP+/MSI+, CIMP-/MSI+, CIMP+/MSI-, and CIMP-/MSI-. Figure 4 displays the survival curves of these 4 groups obtained by the Kaplan-Meier method. Regardless of the cutoff value or type of CIMP panel, CIMP+/MSI- CRCs had the worst clinical outcome, followed by CIMP-/MSI-, CIMP-/MSI+, and CIMP+/MSI+, in order of increasing cumulative survival. The classic CIMP panel outperformed the novel CIMP panel in the prediction of clinical outcome, regardless of the cutoff value (2 or 3). The cutoff value of 2 better separated the survival curve of classic CIMP+/MSI- from that of classic CIMP-/MSI- compared with the value of 3. However, the cutoff value of 3 better separated the survival curve of CIMP+/MSI+ from CIMP-/MSI+, regardless of type of CIMP panel. It should be noted that CRCs classified as new CIMP+/MSI+ or classic CIMP+/MSI+ by the cutoff value of 3 had the best clinical outcome; up to the end of the follow-up, all of the patients belonging to this group were alive.

When the survival curve was analyzed for MSI- CRCs according to the number of methylated markers, the classic CIMP panel exhibited 3 subgroups by the survival curves (Figure 5): (1) CRCs with 1 or no marker methylated, (2) CRCs with 2 to 4 markers methylated, and (3) CRCs with all of the markers methylated. In contrast to the general tendency that CIMP+/MSI- CRCs had the worst clinical outcomes, MSI- CRCs with methylation of all 5 markers had the best clinical outcome. The survival analysis of the new CIMP panel did not show distinct grouping, as seen in the case of the classic CIMP panel analysis.

Survival Analysis by Methylation Status of Individual Markers and Multivariate Analysis 

When we evaluated the survival outcome of CRC patients by methylation status of individual methylation markers, MLH1 and SOCS1 exhibited significant associations between gene hypermethylation and patients' better prognosis (P=.03 and .03, respectively, Kaplan-Meier logrank method; Figure 6). Clinicopathologic features (age, sex, differentiation, and American Joint Committee on Cancer stage), genetic alterations (MSI and KRAS and BRAF mutations), and epigenetic alterations (10 individual methylation markers) were entered into a multivariate analysis to identify independent predictors of overall survival. Multivariate analysis using Cox proportional-hazards model revealed that MINT31 methylation (P=.001), SOCS1 methylation (P=.01), American Joint Committee on Cancer stage, and BRAF mutation were significant factors affecting overall survival of CRC patients (Table 4).

COMMENT

The concept of CIMP was first introduced by Issa's group2 to describe a subset of highly methylated CRCs and was based on 2 features of concordant hypermethylation of multiple CpG island loci and bimodal distribution of CRCs by the number of methylated CpG island loci. Initially, CIMP+ CRCs were defined by the positivity for hypermethylation of 3 or more of 7 MINT loci using the combined bisulfite restriction analysis assay.1 Such defined CIMP+ CRCs showed characteristic clinicopathologic or molecular features, including proximal location and MSI.2 Subsequently, a series of studies using various methods and reference CpG island loci have yielded results supporting the hypothesis that CIMP+ CRCs have characteristic clinicopathologic features. Notably, Ward et al10 reported an adverse prognostic effect of CIMP positivity in CRCs, adding an important clinical feature to CIMP+ CRCs. Recently, the Laird group7 and the Ogino group8 introduced new CIMP marker panels that were analyzed by real-time PCR-based MSP (the MethyLight technology). With these new marker panels analyzed by quantitative methodology, they observed a bimodal distribution of tumors based on the number of methylated markers that was not usually found in the traditional MSP analysis. The new CIMP marker panel introduced by the Laird group7 has demonstrated bimodal distribution of CRCs and performance superior to that of the classic CIMP panel in detecting the characteristic clinical or molecular features of CIMP+ CRCs. However, this comparison was based on real-time PCR-based MSP and has never been tested with traditional MSP analysis.

Usually, methylated DNA alleles are quite heterogeneous with respect to the methylation status of each CpG site within a given CpG island locus (eg, CDH1 CpG island locus).15 Alleles detected by MethyLight technology contain highly methylated CpG sites within a short segment of DNA (usually >9 CpG sites in DNA sequences of <100 bp) targeted by the MethyLight primers and probe; thus, the MethyLight assay detects near-perfectly methylated DNA alleles, which may not comprise a major portion of heterogeneously methylated DNA alleles (depending on the types of CpG island loci). However, partially methylated DNA alleles not detected by MethyLight technology may exert epigenetic effects to downregulate gene expression.15-17,20 The traditional MSP assay can detect both partially methylated alleles as well as the highly methylated alleles detected by MethyLight technology. Until now, the superiority that the Laird group's novel CIMP panel showed in detecting the characteristic features of CIMP+ CRCs compared with the classic CIMP panel has not been proven using traditional MSP.

In the present study, we designed MSP primers that targeted DNA sequences within 50 bp from the DNA segment targeted by the MethyLight primers of the new CIMP panel markers,7 and we used traditional MSP to compare the performance of the new CIMP panel with that of the classic CIMP panel in detecting clinicopathologic or molecular features of CIMP+ CRCs. In addition, we compared the prognostication power of both CIMP panels according to the grouping based on the status of CIMP and MSI. The results showed that the new CIMP panel performed better than the classic CIMP panel in detecting tumors with poor differentiation and tumors without KRAS mutation when tumors showing methylation at 3 or more of 5 markers were defined CIMP+. With a cutoff value of 2, the novel CIMP still outperformed the classic CIMP panel in detecting proximal tumor location, MSI, and BRAF mutation. However, survival analysis revealed the superiority of the classic CIMP panel over the novel CIMP panel: the survival rate of CIMP+/MSI- CRCs was lower than that of the other 3 CRC molecular subtypes, which was more evident when the CIMP was defined by the classic CIMP panel rather than the novel CIMP panel, regardless of the cutoff value to define CIMP status. Superiority of the classic CIMP panel over the novel CIMP panel in prognostication value was also observed in survival analysis that was performed in MSI- CRCs alone (data not shown).

Of the 4 molecular subtypes, CIMP+/MSI- CRCs had the worst clinical outcome, regardless of the type of CIMP panel or the cutoff value (2 or 3). The close association of CIMP with adverse prognostic effect has been reported by other researchers10-12 and was confirmed in the present study. However, in the survival analysis of the MSI- CRCs by the number of methylated markers, CIMP+ CRCs appeared to be inconsistent with respect to clinical outcome. Regardless of the type of CIMP panel, CRCs with 2 or 3 markers methylated showed worse clinical outcome, whereas CRCs with methylation of 4 or more markers had a better outcome. The survival curve of patients with CRCs carrying 4 methylated markers displayed an early rapid drop in survival with prolonged survival in the remainder, which raises the possibility of the presence of 2 heterogeneous groups within this subset of CRCs. Exploration of the molecular alterations in MSI- CRCs with methylation of 4 or more of the novel CIMP markers or of the classic CIMP markers revealed that worse outcome was associated with the presence of BRAF or KRAS mutations and that patients having methylated CRCs without KRAS/BRAF mutations were all alive up to the end of the follow-up. Although the present study is limited because of the small number of cases, the findings suggest that CIMP-high, MSI- CRCs may show quite different clinical outcomes depending on the presence of KRAS/BRAF mutations. A large-scale study is required to clarify this issue.

In conclusion, we compared the clinicopathologic and molecular differences between new and classic CIMP markers in 130 cases of CRC. A new CIMP marker panel showed good correlation with tumor location and mutation of KRAS or BRAF. The classic CIMP marker panel had prognostic value, especially when MSI status was known.

This study was supported by a grant from the Seoul National University Hospital Research Fund (06-2007-070-9), a grant from the National R&D Program for Cancer Control, Ministry of Health & Welfare, Republic of Korea (0720540), and by the second stage Brain Korea 21 project.

© 2008 College of American Pathologists Provided by ProQuest LLC. All Rights Reserved.