The CpG island methylator phenotype (CIMP) is a distinct phenotype in colorectal cancer, associated with specific clinical, pathologic, and molecular features. However, most of the studies stratified methylation according to two subgroups (CIMP-High versus No-CIMP/CIMP-Low). In our study, we defined three different subgroups of methylation (No-CIMP, CIMP-Low, and CIMP-High) and evaluated the prognostic significance of methylation status on a population-based series of sporadic colon cancers. A total of 582 colon adenocarcinomas were evaluated using methylation-specific PCR for 5 markers (hMLH1, P16, MINT1, MINT2, and MINT31). No-CIMP status was defined as no methylated locus, CIMP-Low status as one to three methylated loci, and CIMP-High status as four or five methylated loci. Clinicopathologic and molecular characteristics were correlated to the methylation status. Crude and relative survival was compared according to methylation status. In the microsatellite-stable (MSS) group, CIMP-High was significantly associated with proximal location (P = 0.011) and BRAF mutation (P < 0.001). KRAS mutations were more associated with CIMP-High and CIMP-Low status (P = 0.008). A shorter 5-year survival was observed in MSS cancer patients with CIMP-Low or CIMP-High status. These results remained significant in multivariate analysis adjusted for age, stage, and BRAF and KRAS mutational status [CIMP-Low: hazard ratio (HR), 1.85; 95% confidence interval (95% CI), 1.37–2.51; CIMP-High, HR, 2.90; 95% CI, 1.53–5.49 compared with No-CIMP]. Within the high-level microsatellite instability group, no difference in survival was observed between the different CIMP groups. Our results show the interest of defining three subgroups of patients according to their methylation status (No-CIMP/CIMP-Low/CIMP-High). Methylation is an independent prognostic factor in MSS colon cancer. [Cancer Res 2008;68(20):8541–6]
- colonic neoplasm
Methylation of the CpG Island termed CpG Island methylator phenotype (CIMP), which leads to loss of tumor suppressor gene expression by transcriptional inactivation, has been described in colorectal cancers ( 1). CIMP tumors have distinct clinical, pathologic, and molecular features such as associations with female sex, old age, proximal location, mucinous and poor differentiation, BRAF mutation, and low rates of P53 mutations ( 2, 3). All these features are significantly associated with high-level microsatellite instability (MSI-H), suggesting a link between CIMP-High and microsatellite instability. In contrast, tumors with no methylation (designated as “No-CIMP”) are frequently associated with chromosomal imbalance. The absence of a consensus panel defining the CIMP groups and the use of many different technical approaches make the comparison of the literature data difficult. Methylation status has been reported to have bad prognostic significance in different studies ( 4– 7). In this study, we examined a population-based series of sporadic colon cancers in an attempt to determine their profiles in terms of CIMP status, and clinico pathologic, molecular characteristics, and prognosis.
Patients and Methods
Population. Tissue samples from patients living in the Côte-d'Or area resected for a colon adenocarcnioma between January 1998 and December 2002 were collected and stored in a frozen tissue bank. They were provided by the three pathology laboratories covering the area: Dijon University Hospital, Comprehensive Cancer Centre, and private Centre. Five fragments, considered surgical waste in accordance with French ethical laws (L.1211-3 to L.1211-9), were obtained from each tumor and from the adjacent normal mucosa and were immediately frozen in liquid nitrogen. Cases with rectal cancers (located within 15 cm from the anal verge) or with known familial adenomatous polyposis or hereditary nonpolyposis colorectal cancer syndrome, ulcerative colitis, or Crohn's disease were not eligible. A total of 582 cases (333 men and 249 women) were included in the analysis. Data were merged with that of the cancer Registry, which routinely collects data on patients' characteristics, cancer site, stage at diagnosis, treatment, recurrence, and survival. Tumors occurring between the caecum and the splenic flexure were defined as proximal (n = 248). Other sites were classified as distal (n = 334). Cancer extension at the time of diagnosis was classified according to the 5th Unio Internationale Contra Cancrum (UICC) version of the tumor-node-metastasis classification. Among included cases, 51 were classified as stage I, 252 as stage II, 154 as stage III, and 124 as stage IV.
Microsatellite instability status and Mutation analyses. Microsatellite instability and mutational status for the KRAS gene and the BRAF gene was defined as previously described ( 8).
CIMP status. A minimum of 1 μg of DNA was treated as previously described ( 9) and was eluted in a final volume of 90 μL after purification using a Cleanmix DNA purification kit (Euromedex). To control bisulfite conversion of DNA, all treated samples were amplified with primers specific for COL2A in a region with no CG. Amplification of the five-gene panel described by Issa and colleagues ( 2) was performed once the bisulfite treatment (COL2A) was efficient. Methylation status was assessed for 5 loci [hMLH1 (D region; ref. 10), P16CDKN2A, MINT1, MINT2, and MINT31]. Methylation-specific PCRs (MSP) were performed separately for each gene with primers labeled with VIC or FAM (hMLH1: Fw-CGAATTAATAGGAAGAGCGGATAGCG, Rv-TCACCTCAATACCTCGTACTCACG; P16CDKN2A: Fw-TTATTAGAGGGTGGGGCGGATCGC, Rv-ACCCCGAACCGCGACCGTAA; MINT1: Fw-AATTTTTTTATATATATTTTCGAAGC, Rv-AAAAACCTCAACCCCGCG; MINT2: Fw-TTGTTAAAGTGTTGAGTTCGTC, Rv-AATAACGACGATTCCGTACG; MINT31: Fw-TGTTGGGGAAGTGTTTTTCGGC, Rv-CGAAAACGAAACGCCGCG). Three microliters of bisulfite treated DNA were amplified in a final volume of 15 μL, using 1.3 μL of PCR buffer, 0.5 μL of MgCl2 at 25 mmol/L, 0.6 μL of DNTP at 10 mmol/L, and 0.15 μL Hotstar Taq polymerase (Qiagen). The protocol was validated by amplification of DNA extracted from cell lines (SW48 for methylated control, V9P for unmethylated control). The reactions were performed according to the standard protocol with hybridization temperature set at 62°C for P16CDKN2A, MINT2 and MINT31, 64°C for hMLH1, and 52°C for MINT1. Each amplified plate bore a negative control. Amplified products were then pooled and diluted at 1/50. Amplicons were separated by capillary electrophoresis (ABI-3100). Two assays were performed for each locus, and a CpG-island was defined as methylated when both assays showed amplification. When one of two assays showed no amplification, a third assay was performed to conclude whether that locus was methylated or not.
Statistical analysis. To select the best cutoff for CIMP-High, we tested the homogeneity of groups with 1, 2, 3, 4, or 5 methylated loci, according to clinico pathologic and molecular features. For most of the features (Supplementary Table S1), a change of trend was observed between three and four methylated markers for age and location, and four and five methylated markers for other features. To refine our analyses, we performed them separately for microsatellite-stable (MSS) and MSI-H groups. In the MSS group (Supplementary Table S2), the changes of trend were rather observed between three and four methylated markers. In the MSI group, we could not evaluate the cutoff due to the small size of each subgroup. The cutoff was thus chosen between three and four methylated markers, and tumors were classified as CIMP-High when they showed four or more methylated loci, versus CIMP-Low with one to three methylated loci. No-CIMP tumors were characterized by lack of methylation.
Association between CIMP status and clinico pathologic and molecular features was analyzed using χ2 and Fisher exact test for heterogeneity. Crude survival was calculated using Kaplan-Meier method and estimates were compared using the Log-Rank test. Relative survival was defined as the ratio of the observed survival rates to the expected survival rates in a population with similar sex and age distribution derived from local mortality tables. Multivariate relative survival analyses were performed using a generalized linear model with a Poisson error structure based on collapsed data using exact survival times as proposed by Dickman and colleagues ( 11). The significance of the covariates was tested by the likelihood ratio test. Relative survival estimates were computed using STATA 10 software. Multivariate survival analyses in MSS group were adjusted for classic clinico pathologic variables (stage and age) and for molecular features that showed significant effect or borderline effect (P < 0.1) in univariate analyses (KRAS and BRAF mutational status) after merging stage I and stage II patients (only 1 patient belonged to CIMP-high and stage I subgroup).
CIMP status was successfully assessed in all cases. Methylation frequencies in the 582 patients were 12.2% for hMLH1 (71 cases), 26.3% for P16 (153 cases), 21.7% for MINT1 (126 cases), 29.9% for MINT2 (174 cases), and 37.1% for MINT31 (216 cases). Among included cases, 286 were No-CIMP (49.1%), 199 were CIMP-Low (34.2%), and 97 were CIMP-High (16.7%).
Clinicopathologic and molecular features according to methylation status among the whole population. The characteristics of tumors according to their methylation status are summarized in Table 1 .
CIMP status was significantly associated (P < 0.001) with sex, location of the cancers, age, MSI status, BRAF mutation, and KRAS mutation.
CIMP-H patients were more often associated with female sex, proximal location, advanced age, MSI-H status, BRAF mutation, and KRAS wild-type.
No-CIMP and CIMP-Low patients were more frequently associated with male sex distal colon, MSS status, and BRAF wild-type. CIMP-Low patients were also more often mutated for the KRAS gene.
Clinicopathologic and molecular features according to methylation status among MSS group and MSI groups. In the MSI group, significant associations were seen between CIMP-High and proximal location (P = 0.015), BRAF mutation (P < 0.001), and KRAS wild-type (P = 0.001; Table 2 ).
In the MSS group, a significant association was observed between CIMP-High status, proximal location of the cancers (P = 0.011), and BRAF mutation (P < 0.001; Table 2). KRAS mutation was more common in methylated tumors (CIMP-Low and CIMP-High) than in No-CIMP tumors (P = 0.008).
CIMP status and patient survival. The prognostic value of methylation status is shown in Table 3 for the whole population stratified according to MSI.
Within the MSI-H cancers ( Table 3), no difference was observed between the different CIMP groups. Survival in the 3 MSI-H subgroups was quite similar to that in the MSS/No-CIMP group. Within the MSS cancer group, the survival curves showed that increasing levels of tumor methylation were significantly associated with decreased survival ( Fig. 1 ). This trend was most apparent with tumors showing methylation at more than four methylated loci (CIMP-High) but was also noted in tumors with one to three methylated sites (CIMP-Low; P < 0.001). Multivariate analysis in the MSS group was summarized in Table 4 . UICC stage, KRAS mutational status, and methylation status were significantly associated with decreased survival (P < 0.001; Table 4).
For loci taken individually (hMLH1 excluded), methylation was significantly associated with poor relative survival for each marker in MSS group (P < 0.001; Fig. 2 ). In multivariate analysis, the results remained significant for P16 [hazard ratio (HR), 1.80; 95% confidence interval (95% CI), 1.29–2.49; P < 0.001] and MINT31 (HR, 1.43; 95% CI, 1.06–1.94; P = 0.019). Relative survival was not estimated for hMLH1 locus considering the small number of patients in that subgroup.
The CIMP was originally described in 1999 by Toyota and colleagues ( 1) who defined a subgroup of cancers with a high rate of hypermethylation. From this moment, the majority of papers concerning DNA methylation dealt with the most suitable methodology to use (technique and markers). Although Methylight is fashionable ( 12), MSP is the most commonly used technique in the literature. It is effective and specific and does not require specific equipment ( 13), and this justifies our choice. In this study, we used the five markers (hMLH1, P16CDKN2A, MINT 1, MINT 2, and MINT 31) defined by Issa ( 2). In 2005, Weisenberger ( 14) recommended a new panel. Nevertheless, the evaluation of Ogino and colleagues ( 15) and Shen and colleagues ( 16) did not support the notion that this new panel outperformed the classic panel. It also emphasized that differences in primer design and PCR conditions may substantially change sensitivity and specificity of a particular marker. At present, there is no consensus definition for the panel, although the five genes described by Issa ( 2) provide a simplified and representative approach in defining CIMP. All these genes have been reported to correlate with CIMP and accurately define CIMP in colon cancer ( 3, 13) and in many other cancers as well ( 17– 20).
In the literature, the problem of the cutoff to stratify CIMP (CIMP-High versus CIMP-Low) has not been sufficiently discussed, leading to different thresholds in various papers. In this population-based study, the cutoff was assessed according to clinico pathologic and molecular features. CIMP-High was defined when four or five markers were methylated. This allowed us to identify a group with extensive methylation (CIMP-High), which not only included MSI-H tumors (61%) but also MSS tumors (39%). Besides, we also identified the No-CIMP group (tumors without methylation) and the CIMP-Low group with one to three methylated loci.
In the whole population, CIMP-High was clearly associated with clinico pathologic features of MSI-H tumors such as female sex, older age, proximal location, and BRAF mutation. In the MSS group, only associations with proximal location and BRAF mutation remained significant, suggesting that associations with sex and age were due to MSI-H status.
For all patients, tumors with no methylation or a low level of methylation (No-CIMP or CIMP-Low) were significantly associated with men, distal location, MSS status, and BRAF wild-type. CIMP-Low tumors also showed a higher rate of KRAS mutation. These results are in accordance with other reports using other technologies ( 21), suggesting that MSP is as efficient as Methylight to define CIMP subgroups provided a stringent cutoff is used. In the MSS group, associations remained significant except for men, although a trend was observed.
In this paper, three factors influence relative survival in MSS group, the UICC stage, the KRAS mutational status, and the methylation status, with a major effect of stage IV (HR, 14.33) and CIMP-High status (HR, 2.90).
To our knowledge, our study is the first to evaluate survival in three clearly defined subgroups of methylation (No-CIMP, CIMP-Low, and CIMP-High) in MSS and MSI-H tumors in a population-based study. A key finding of this study is that individuals with methylated MSS tumors have a significantly worse outcome than those with nonmethylated MSS tumors. The higher the level of methylation, the poorer the survival. This highlights the importance of identifying the CIMP-Low group, which was not done in the six previous papers dealing with the prognosis of methylated colon cancers ( 3– 7, 22). It is all the more important because the clinico pathologic features of No-CIMP and CIMP-Low are quite similar. This emphasizes the importance of integrating genetic and epigenetic analysis to resolve the molecular heterogeneity in cancers ( 16). Taken individually, the methylated loci P16 and MINT31 are significant predictors of a poor prognosis. The association between levels of DNA methylation and poor survival in the MSS group has not yet found a clear explanation. In the MSI group, no significant result on survival according to methylation status was observed. The issue is complicated by the association of two antagonist factors on the prognosis (methylation and MSI status). Meta-analyses and team collaborations are necessary to gather a larger cohort of MSI-H colorectal cancer tumors and to clarify the interactions between MSI status, methylation and prognosis.
In conclusion, our work on a population-based series clearly shows the need to distinguish between the 3 groups of CIMP. This paper shows that methylation is an independent predictor of a poor prognosis in MSS colon cancers. It highlights the importance of analyses including genetic and epigenetic markers to take into account the different molecular subgroups of cancers. Further studies are now required to understand the cause of aberrant methylation and the way that acts on carcinogenesis and disease outcome.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: “Ligue nationale de lutte contre le cancer (Comité de Côte d'Or)” (N° 2006-19), by the “Fondation de France” (N° 2007006503), and by a PHRC 2003 (epigenetic and genetic alterations in colorectal cancer).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank the Biological Resource Center of Dijon (Ferdinand Cabanne) and the participating pathologists of Comprehensive Cancer Center (Dr. L. Arnould and Dr. F. Collin), Dijon Pathological Private Center (Dr. F. Drouot, Dr. L. Dusserre, Dr. D. Michiels-Marzais, Dr. F. Morlevat, Dr. T. Petrella, and Dr. T. Ponnelle), L. Barbier, P. Bastable, M. Chevarin, E. Lanier, and S. Normand for their contribution to this work.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received April 2, 2008.
- Revision received July 10, 2008.
- Accepted July 16, 2008.
- ©2008 American Association for Cancer Research.