Prognostic Value of Methylator Phenotype in Stage III Colon Cancer Treated with Oxaliplatin-based Adjuvant Chemotherapy

Three major molecular phenotypes have been recognized in colon cancer: chromosomal instability (1), microsatellite instability owing to deficient mismatch repair system (dMMR; ref. 2), and methylator phenotype (3). This latter one is associated with hypermethylation of the CpG islands localized in the gene enhancer regions and is named “CIMP” for “CpG island methylator phenotype.” It represents about 10% to 20% of colon cancer (3–5). Clinical features more frequently associated with this phenotype are female gender, older age, and proximal location of the tumor; in addition, these tumors are more frequently BRAF-mutated, KRAS, and TP53 wild-type, and associated with a dMMR phenotype due to hypermethylation of the enhancer region of MLH1 gene.

Nonmetastatic colon cancer prognosis is primarily based on the TNM classification that guides adjuvant treatment indications (6). Several markers have been suggested to refine the prognostic classification of nonmetastatic colon cancer, such as dMMR, KRAS, and BRAF mutation status (2, 7). The prognostic impact of CIMP⁺ phenotype in nonmetastatic colon cancer is controversial with several studies showing a pejorative impact of this phenotype (8–13), whereas several others failed to demonstrate such impact (14–16). In addition, the efficacy of 5-fluorouracil (5FU) or oxaliplatin (FOLFOX)-based adjuvant chemotherapy regimen in this tumor phenotype is a matter of debate (8, 12, 14–20). To clarify the prognostic role of the CIMP⁺ phenotype, we characterized the tumor methylation status of patients included in the PETACC-8 randomized clinical trial, which compared FOLFOX4 with FOLFOX4 plus cetuximab regimen in patients with stage III colon cancer (21). We examined the relationship between CIMP⁺ phenotype and disease-free survival (DFS), overall survival (OS), and survival after recurrence (SAR).

PETACC-8 trial is a phase III randomized trial that included 2,559 patients with stage III colon cancer and compared FOLFOX4 (oxaliplatin, leucovorin, and 5-FU) regimen with FOLFOX4 + cetuximab in the adjuvant setting (21). Among these patients, 2,043 have signed an informed consent for translational research study, and tumor DNA extracted from formalin-fixed, paraffin-embedded sections after macrodissection was available for 1,907 of them in this study. Microsatellite instability, KRAS, NRAS, and BRAF status were reported previously (22).

To characterize CIMP status, the panel of five markers described by Weisenberger and colleagues was used: IGF2, CACNA1G, NEUROG1, RUNX3 and SOCS1 (5). CIMP⁺ phenotype was defined by methylation-positive status of at least 3 markers out of the 5 tested (5). In the literature, the definition of CIMP⁺ status is not standardized regarding the gene panel used and marker threshold (13). We decided to use this definition because of its better diagnostic performance compared with the classic panel (methylated in tumors (MINT)1, MINT2, MINT31, and MLH1; ref. 23) as demonstrated by Weisenberger and colleagues (5) and its frequent use in many recent studies (10, 13, 16, 18, 19, 24).

Bisulfite conversion was carried out with EZ-96 DNA Methylation-Lightning Kit (Zymo Research). The amount of DNA used was approximately 200 ng in an elution volume of 20 μL to obtain a final concentration of bisulfite-treated DNA of 10 ng/μL.

Primers used for the methylation-specific PCR (25) are given in Supplementary Table S1. Three PCRs were performed. First, we performed a multiplex PCR of the three following markers IGF2/CACNA1G/NEUROG1 using the corresponding 12 primers for the determination of the methylation status. A mix of 500 μL the 12 primers were done using 10 μL of each primer at 100 μmol/L, and 380 μL Tris–EDTA buffer. Briefly, the multiplex PCR was performed in 25 μL reaction volume using the Multiplex PCR Kit, (Qiagen Multiplex PCR Kit), with 2.5 μL of the primer mix, 12.5 μL of the multiplex PCR Master Mix, 2.5 μL of the Q Solution, 4 μL of the bisulfite-converted DNA, and 3.5 μL of RNAse-free water. PCR was performed as follows: 15 minutes at 95°C, followed by 40 cycles (30 seconds at 94°C, 90 seconds at 59°C, and 90 seconds at 72°C), and 10 minutes at 72°C. Second, MSP was performed in two separate reactions for the two remaining markers, owing to different annealing temperatures. The same protocol was used except for the primer mix, which contained 4 specific primers for the annealing temperature, which was 55°C for RUNX3, and 60°C for SOCS1.

PCR amplification products were diluted (1:150 for the triplex PCR reaction, 1:50 for RUNX3 PCR reaction, and 1:200 for SOCS1 PCR reaction). A total of 1.5 μL of this diluted solution was mixed with 18.5 μL of formamide (HIDI Formamide, Applied Biosystems) and 0.1 μL of a size marker [Genescan 400 HD (ROX) size standard, Applied Biosystems). Capillary electrophoresis was carried out on a 3730xl DNA Analyzer (Thermo Fisher Scientific). Electrophoresis data were interpreted using the software GeneMapper (Thermo Fisher Scientific). For each gene, the peak height ratio of the methylated/unmethylated amplicon was measured. If the ratio was greater than 0.1 or if the peak height of methylated amplicon was greater than 500, the marker was considered methylated.

Because of the definition of CIMP status (methylation of at least three markers of the five tested), for each sample, the analysis was initially carried out on the triplex IGF2/CACNA1G/NEUROG1. If the three markers or none of these markers were methylated, the status was, respectively, CIMP⁺ and CIMP−; therefore, the analysis was stopped. If one or two markers were considered methylated, the analysis of SOCS1 and RUNX3 was carried out to further characterize CIMP status.

Categorical variables were compared using the χ² test. Continuous variables were compared using t test. OS was defined as the time from randomization to death. DFS was defined as the interval from the randomization to locoregional or metastatic recurrence, the appearance of a secondary colon or rectal cancer, or death, whichever occurred first. SAR was defined as the interval from the first locoregional or metastatic recurrence or the appearance of a secondary colon or rectal cancer to death. OS, DFS, and SAR curves were estimated with Kaplan–Meier method and were compared by a two-sided log-rank test.

A Cox proportional hazard model was achieved by adjusting for known prognostic factors: age, gender, WHO performance status (PS), bowel obstruction and/or perforation, tumor stage pT and pN, histopathology grade, MMR status, BRAF and RAS mutation status, and also treatment arm. Analyses were carried out according to a two-sided statistical significance level of 5%.

Among the 2,043 that signed the informed consent for translational research in PETACC8 trial, CIMP status was tested on 1,907 samples. In 40 cases, CIMP status could not be determined due to poor DNA quality and CIMP status determination could be achieved in 1,867 samples (97.9%) and was positive in 275 cases (14.7%; see Supplementary Fig. S1). The CIMP-analyzed group was well matched to those for whom we cannot obtain the CIMP status owing to the absence of tumor block or failure of CIMP determination (full details in Supplementary Table S2).

Patients in CIMP⁺ group were slightly older (60.8 years vs. 58.9 years; P = 0.002) and more frequently women than patients in CIMP⁻ group [48.7% vs. 42.1%, respectively; OR, 1.3; 95% confidence interval (CI), 1–1.7; P = 0.04]. CIMP⁺ tumors were more frequently located in the proximal part of the colon than CIMP⁻ tumors (75.3% vs. 32.4%, respectively; OR, 6.35; 95% CI, 4.7–8.7; P < 0.001), more frequently poorly differentiated (grade 3–4; 36.3% vs. 15.9%, respectively; OR, 3; 95% CI, 2.2–4.0; P < 0.001), with a greater proportion of pT3-4 stage (94.9% vs. 89.6%, respectively; OR, 2.2; 95% CI, 1.2–4.1; P = 0.005), and pN2 stage (46.2% vs. 36.1%, respectively; OR, 1.51; 95% CI, 1.2–2.0; P = 0.001). On a molecular level, CIMP⁺ tumors were more frequently dMMR than CIMP⁻ tumors (34.2% vs. 5.7%, respectively; OR, 8.6; 95% CI, 6.0–12.3; P < 0.001), BRAF mutated (50.4% vs. 4.3%, respectively; OR, 22.5; 95% CI,15.8–32.3; P < 0.001), and less frequently RAS mutated (36.3% vs. 50.5%, respectively; OR, 0.55; 95% CI, 0.4–0.7; P < 0.001). There was no significant difference for the remaining tested clinical variables (Table 1). Because these clinical variables and these genetic alterations are associated, we observed that dMMR- and BRAF-mutated right-sided tumors were 100 times more frequent in the CIMP⁺ tumor group than in the CIMP⁻ tumor group (20.6% vs. 0.2%, respectively) and that pMMR and BRAF-mutated right-sided tumors were 10 times more frequent in the CIMP⁺ tumor group as compared with CIMP⁻ tumor group (20.2% vs. 2.7%, respectively). Finally, pMMR and BRAF wild-type left-sided tumors were 6 times more frequent in CIMP⁻ tumor group than in CIMP⁺ tumor group (60% vs. 10.1%, respectively; see Supplementary Fig. S2).

Median follow-up durations were 6.1 years (range 0–10.1) and 6.6 years (range 0–10.3) for patients with CIMP⁺ tumors and CIMP⁻ tumors, respectively.

In the overall population, DFS was not significantly different between the two groups of patients with CIMP⁺ tumors and CIMP⁻ tumors (HR, 1.18; 95% CI, 0.94–1.47; P = 0.15) with 3-year DFS at 70.4% and 75.4%, respectively. OS was significantly shorter in patients with CIMP⁺ tumors than CIMP⁻ tumors (HR, 1.46; 95% CI, 1.15–1.87; P = 0.002) with 5-year OS at 72.6% and 83.2%, respectively. SAR was significantly shorter in patients with CIMP⁺ tumors as compared with patients with CIMP⁻ tumors (HR, 2.14; 95% CI, 1.63–2.80; P < 0.0001) with 1-year SAR at 54.7% and 82.2%, respectively (Fig. 1).

Because CIMP⁺ tumors are strongly associated with dMMR phenotype and BRAF mutations, we analyzed the survival data separately in patients with pMMR and dMMR tumors and in patients with BRAF wild-type and mutated tumors. In patients with pMMR tumor: DFS was significantly shorter in patients with CIMP⁺ tumors than CIMP⁻ tumors (HR, 1.44; 95% CI, 1.12–1.87; P = 0.005) with 3-year DFS at 63.7% and 75.3%, respectively. OS was significantly shorter in patients with CIMP⁺ tumors than CIMP⁻ tumors (HR, 1.92; 95% CI, 1.46–2.51; P < 0.0001) with 5-year OS at 64.5% and 83.4%, respectively. SAR was significantly shorter in patients with CIMP⁺ tumors than CIMP⁻ tumors (HR, 2.80; 95% CI, 2.06–3.80; P < 0.0001) with 1-year SAR at 55.4% and 82.3%, respectively (Fig. 2). Similar results were observed when the analysis was performed independently for proximal and distal tumors (Table 2). In patients with dMMR tumors, no significant difference was observed between patients with CIMP⁺ and CIMP⁻ tumors for DFS, OS, and SAR (respectively, P = 0.75; P = 0.85; and P = 0.53; see Supplementary Fig. S3).

In patients with BRAF wild-type tumors, DFS, OS, and SAR were shorter in patients with CIMP⁺ than in CIMP⁻ tumors (HR, 1.26; 95% CI, 0.93–1.70; P = 0.13; HR, 1.57; 95% CI, 1.4–2.18; P < 0.007; and HR, 1.90; 95% CI, 1.30–2.78; P < 0.001, respectively). No significant difference was observed in the BRAF-mutated tumors.

Cox proportional hazard model was performed including the following variables: age, gender, WHO PS, bowel obstruction and/or perforation, treatment arm, tumor stage pT and pN, grade, MMR status, and BRAF and RAS mutation status. In the overall population, OS and SAR were significantly shorter in patients with a CIMP⁺ tumor than in patients with a CIMP⁻ tumor (respectively, HR, 1.46; 95% CI, 1.02–1.94; P < 0.04 and HR, 1.76; 95% CI, 1.20–2.56; P < 0.004; see Table 3), in contrast to DFS, which was not significantly different between the two groups (HR, 1.15; 95% CI, 0.86–1.54; P = 0.34).

In multivariate analysis, by adjusting on the same prognostic factors apart from MMR status, in the pMMR population, OS and SAR remained shorter in patients with CIMP⁺ tumor as compared with patients with CIMP⁻ tumor (HR, 1.40; 95% CI, 1.00–1.96; P = 0.05 and HR, 1.98; 95% CI, 1.33–2.95; P < 0.001, respectively).

We investigated the impact of treatment (FOLFOX4 vs. FOLFOX4 plus cetuximab) according to CIMP status of the patients’ tumors. In univariate analysis, nonsignificant trends in DFS, OS, and SAR differences in disfavor of the FOLFOX4 + cetuximab arm were observed in the group of patients with CIMP⁺ tumors (Fig. 3). These marginal differences are retained after adjustment. The respective adjusted HR for the experimental arm (FOLFOX4 + cetuximab) was 1.35 (95% CI, 0.87–2.10), P = 0.19; 1.49 (95% CI, 0.92–2.41), P = 0.10 and 1.23 (95% CI, 0.66–2.29), P = 0.52 for DFS, OS, and SAR. On the contrary, in patients with CIMP⁻ tumors, DFS and OS were very similar between the two treatment arms. The respective adjusted HR of DFS and OS were 0.97 (95% CI, 0.80–1.18), P = 0.80; 1.13 (95% CI, 0.90–1.42) P = 0.30. The HR for SAR was of the same magnitude than that observed in CIMP⁺ tumors: 1.30 (95% CI, 1.00–1.68), P = 0.05.

Finally, we tested an interaction between the CIMP status and the treatment arm in the group of patients with RAS and BRAF wild-type pMMR tumors (601 patients), to eliminate the possible interactions with BRAF and dMMR status. After adjustment on age, gender, PS, tumor stage (pT and pN), tumor grade, and tumor location, we observed a nonsignificant trend for interaction (P = 0.10) in disfavor of FOLFOX4 + cetuximab arm for patients with a CIMP⁺ tumors.

Using the most recent described panel of markers to determine the CIMP status, the prevalence of the CIMP⁺ phenotype in this large series of stage III colon cancer was 14.7% (95% CI, 13.1%–16.3%). This is in accordance with the prevalence observed in the literature, which ranges between 9.6% and 23.1% (9, 16, 18, 20, 25). In addition, we observed similar associations between CIMP status, clinical variables, and tumor characteristics that have been shown in other studies (3, 5, 26, 27). Indeed, the CIMP⁺ phenotype seems more frequently associated with a female gender, a right-sided, poorly differentiated, BRAF mutated, and dMMR tumor. These associations reflect the overlap between CIMP⁺ and dMMR phenotype due to the methylation of the enhancer region of MLH1 gene. More than 50% of the tumors with a dMMR phenotype are CIMP⁺ and in the CIMP⁺ tumors those that shared a proximal location, a BRAF mutation and a dMMR-positive status are 100 times more frequent than in CIMP⁻ tumors.

These associations could explain the discrepant result in the literature on the prognostic impact of the CIMP status. In our whole series, the CIMP⁺ status is significantly associated with a shorter OS and SAR after adjustment on the main confounding factors. It is important to note that among the studies limited to stage III colon cancer, this study is the only one that have adjusted the prognostic value of CIMP⁺ phenotype for all main clinical, pathologic but also extensive molecular prognostic markers’ characterization (9, 17). Subgroup analyses showed on the one hand that the prognostic impact of the CIMP status is limited to the group of tumors with a pMMR status in univariate and multivariate models. On the other hand, the prognostic impact of the CIMP status is observed both in right- and left-sided pMMR tumors. Moreover, this is the first study evaluating the prognostic value of CIMP⁺ status on a homogenous large population of patients with stage III colon cancer included in a phase III trial with oxaliplatin-based adjuvant chemotherapy. Cohen and colleagues (16) evaluated CIMP phenotype in a small phase III trial (n = 293) with oxaliplatin-based adjuvant chemotherapy including patients with stage II or III colon cancer. Likely due to a lack of power, this study failed to demonstrate a prognostic value of CIMP⁺ status (in univariate analysis, HR for OS was 1.27; 95% CI, 0.58–2.80; P = 0.55). Shiovitz and colleagues (17) evaluated the CIMP status on a larger trial (n = 615) that compared 5-FU and leucovorin associated or not with irinotecan in patients with stage III colon cancer, showing that patients with CIMP⁺ tumor had a shorter OS in a univariate analysis, but failed to show any significant difference in their multivariate model. However, a significant interaction with treatment arms was observed: patients with CIMP⁺ and pMMR tumors benefitted from the addition of irinotecan to 5-FU and leucovorin therapy.

The PETACC-8 study showed no difference in OS and DFS between the group treated by FOLFOX4 and the group treated by FOLFOX4 associated with cetuximab in the adjuvant setting for stage III colon cancer, even when excluding KRAS-mutant patients (21). When selecting patients wild-type for KRAS, NRAS and BRAF, an HR of 0.76 was observed but still not significant (P = 0.11; ref. 28). Here, we observed a HR greater than 1 for DFS (HR, 1.35; 95% CI, 0.87–2.10) and OS (HR, 1.49; 95% CI, 0.92–2.41), suggesting a potential deleterious effect of cetuximab in CIMP⁺ patients, even if the difference remains not significant. This analysis was not restricted to the RAS and BRAF wild-type subgroup due to the small number of patients subsequent to the overlapping between BRAF-mutant and CIMP⁺ status. Furthermore, we showed a nonsignificant trend for interaction in the subgroup of RAS and BRAF wild-type and pMMR patients in disfavor of cetuximab suggesting a possible deleterious effect of the association of FOLFOX4 plus cetuximab in CIMP⁺ patients. These findings could be explained by the association between CIMP⁺ phenotype and the downregulation of two EGFR ligands gene expression, EREG and AREG (29, 30). As anti-EGFR mAbs are currently used in RAS wild-type patients in the metastatic setting, to assess the predictive value of CIMP status for cetuximab efficacy in the treatment of patients with metastatic colon cancer may be interesting.

We found significantly shorter SAR and OS in patients with CIMP+ tumors, unlike DFS, showing that CIMP+ status has a prognostic value only for patients with recurrence, which could be explained by the prognostic value of CIMP+ status in the metastatic setting.

Validation of our results in the NCCTG N0147 trial (31), comparing mFOLFOX6 with mFOLFOX6 plus cetuximab in stage III colon cancer, would be interesting.

In conclusion, in a large clinically and molecularly well-defined stage III colon cancer population treated with standard adjuvant therapy, the methylator phenotype accounts for about 15% of patients and is associated with shorter OS and SAR. However, no impact of CIMP status on DFS was observed. These results suggest that the CIMP⁺ phenotype confers a more aggressive phenotype at the metastatic stage of the disease or a resistance to some of the drugs used in this setting.

J. Tabernero is a consultant/advisory board member for Amgen, Boehringer Ingelheim, Celgene, Chugai, Imclone Systems, Lilly, Merck, Millenium, Novartis, Roche/Genentech, Sanofi, and Taiho. G. Folprecht reports remuneration from Merck Serono; Roche/Genentech; Bayer; Lilly; Servier; Mundipharma; Bristol, Myers, Squibb; MSD; and Shire. C. Lepage reports receiving speakers bureau honoraria from Amgen, Novartis, and Bayer, and is a consultant/advisory board member for Novartis and Halio-DX. P. Laurent-Puig is a consultant/advisory board member for Merck Serono; Amgen; Boerhinger Ingelheim; Biocartis; Roche; Bristol, Myers, Squibb; and MSD. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Gallois, J. Taieb, K. Le Malicot, J. Tabernero, J.F. Seitz, P. Laurent-Puig

Development of methodology: C. Gallois, J. Taieb, D. Le Corre, K. Le Malicot, J. Tabernero, J.F. Seitz, P. Laurent-Puig

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Gallois, J. Taieb, J. Tabernero, C. Mulot, J.F. Seitz, T. Aparicio, G. Folprecht, C. Lepage, E. Mini, J.-L. Van Laethem, J.F. Emile, P. Laurent-Puig

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Gallois, J. Taieb, K. Le Malicot, J.F. Emile, P. Laurent-Puig

Writing, review, and/or revision of the manuscript: C. Gallois, J. Taieb, D. Le Corre, K. Le Malicot, J. Tabernero, C. Mulot, J.F. Seitz, T. Aparicio, G. Folprecht, C. Lepage, E. Mini, J.-L. Van Laethem, J.F. Emile, P. Laurent-Puig

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Gallois, J. Taieb, K. Le Malicot, J. Tabernero, J.F. Seitz, T. Aparicio, G. Folprecht, C. Lepage, E. Mini, J.-L. Van Laethem, J.F. Emile, P. Laurent-Puig

Study supervision: C. Gallois, J. Taieb, K. Le Malicot, P. Laurent-Puig

The PETACC8 study was supported by the Fédération Francophone de Cancérologie Digestive (FFCD) that was responsible for the study management. Merck KGaA and Sanofi-Aventis supported the PETACC8 study: Merck KGaA provided the study cetuximab and financial support for study management; Sanofi-Aventis provided financial support for the provision of oxaliplatin to Belgian sites when necessary. We thank all participating patients and their families and the PETACC8 investigators from the participating countries (Supplementary Data).

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.