Defective Mismatch Repair Status as a Prognostic Biomarker of Disease-Free Survival in Stage III Colon Cancer Patients Treated with Adjuvant FOLFOX Chemotherapy

This article is featured in Highlights of This Issue, p. 7205

Colorectal cancer (CRC) is the second cause of cancer death in Western countries. Worldwide, approximately 1.2 million new cases of CRC are diagnosed every year (1). The main prognostic factor for survival and relapse after resection of nonmetastatic CRC is the pathologic tumor stage. Adjuvant chemotherapy based on 5-fluorouracil (5-FU) was shown in the 1990s to reduce the risk of recurrence and to prolong survival in patients with stage III colon cancer (2). More recently, adding oxaliplatin to 5-FU–based adjuvant therapy was shown to improve disease-free survival (DFS) and overall survival (OS) in this setting (3, 4).

Although the pathologic tumor stage remains the key determinant of CRC prognosis and treatment, there is considerable stage-independent variability in clinical outcome. Thus, new prognostic and predictive biomarkers are needed to guide the use (and choice) of adjuvant treatment. Although most cases of CRC develop via a chromosomal instability pathway, some 15% of cases are characterized by microsatellite instability (MSI), a molecular marker of defective DNA mismatch repair (MMR; ref. 5). Tumor MMR status is usually determined either by polymerase chain reaction (PCR)-based assay targeting specific microsatellites markers, or by immunohistochemical (IHC) analysis of the protein products of genes involved in DNA MMR, such as MLH1, MSH2, and MSH6 (6). Tumors with defective MMR (dMMR) status have distinct clinical and pathologic features, which include proximal colon predominance, frequent poor differentiation, and mucinous histology, and increased numbers of tumor-infiltrating lymphocytes (7, 8).

A systematic review has shown that patients with dMMR tumors have a more favorable stage-adjusted prognosis than patients with proficient MMR (pMMR) tumors (9). In addition, a growing body of evidence suggests that 5-FU–based adjuvant chemotherapy is ineffective in patients with dMMR tumors (10). In contrast, other studies have suggested that patients with dMMR tumors derive a similar or even a greater benefit from 5-FU–based adjuvant treatment than patients with pMMR tumors (10). These conflicting results were based on studies in which patients were not randomly assigned to 5-FU–based treatment versus watchful waiting after resection, thus raising the possibility of a selection bias. Finally, a recently reported independent dataset from randomized trials of adjuvant chemotherapy suggests that 5-FU–based adjuvant therapy does not improve survival in patients with dMMR tumors (11).

The relevance of all these studies is debatable, however, as the fluoropyrimidin-oxaliplatin combination is currently the standard adjuvant chemotherapy for patients with stage III colon cancer (3, 4). The prognostic impact of MMR status in patients with stage III colon cancer treated with adjuvant 5-FU–oxaliplatin combination chemotherapy has rarely been explored. The aim of the present study was to assess the prognostic significance of MMR status in a large series of patients receiving adjuvant FOLFOX chemotherapy after resection of stage III colon cancer.

This retrospective multicenter study included all consecutive patients with histologically confirmed stage III colon cancer and available tumor specimens who received curative surgical resection followed within 8 weeks by adjuvant FOLFOX chemotherapy, from June 2003 to December 2007; the study population included 92 patients previously studied by Zaanan and colleagues. (12) and 44 patients studied by Des Guetz and colleagues (13). The patients were treated in 4 university hospitals (Ambroise Paré: APA, Avicenne: AVI, Georges Pompidou: GP and Saint-Antoine: STA), 2 private hospitals (polyclinique Bordeaux: BOR and Institut Mutualiste Montsouris: IMM), one general hospital (Montfermeil: MTF), and 2 cancer centers (Institut Curie: CUR and Institut Gustave Roussy: IGR). Seventeen patients enrolled in the MOSAIC trial and treated in these centers with FOLFOX between October 1998 and January 2001 were also analyzed. Exclusion criteria were age less than 18 years, rectal cancer, and abdominopelvic radiotherapy. The study was approved by the Pitié Salpétrière Hospital ethics committee.

Patients were planned to receive 12 cycles of adjuvant FOLFOX chemotherapy within a 6-month period. The patients received a 2-hour infusion of 85 mg of oxaliplatin per square meter on day 1, in addition to the standard LV5FU2 regimen (FOLFOX4) or the simplified LV5FU2 regimen (modified FOLFOX6; refs. 3, 14). After surgery, tumor recurrence was detected by physical examination, serum carcinoembryonic antigen assay, and abdominal and thoracic imaging every 3 to 6 months for 3 years, every 6 months for the following 2 years, then annually. The duration of follow-up was defined as the time between surgery and disease recurrence, death, or last hospital contact (scheduled follow-up or telephone contact). The cutoff date for this analysis was December 2010.

Tumor MMR status was determined by MSI testing or immunohistochemistry. Defective MMR status was defined as the presence of either high-level tumor DNA MSI (MSI-H) or as the loss of tumor MLH1, MSH2, or MSH6 protein expression. Proficient MMR status was defined by tumor DNA microsatellite stable (MSS) status, low-level MSI (MSI-L) status, or normal tumor MLH1, MSH2, and MSH6 protein expression.

The paraffin tissue blocks were stored at room temperature. Blocks of formalin-fixed, paraffin-embedded adenocarcinoma tissue comprising an area of normal colonic mucosa adjacent to the tumor were selected in each case. The grade of differentiation was determined according to the World Health Organization criteria. Sections of 4 μm from the paraffin-embedded tissue samples were cut onto silane-treated Super Frost slides (CML) and left to dry at 37°C overnight. The slides were deparaffinized in xylene and rehydrated in pure ethanol. Endogenous peroxidase was blocked using 3% hydrogen peroxide in methanol for 30 minutes. Before immunostaining, antigen retrieval was done by immersing sections in citrate buffer (pH 6.0). Sections were then incubated for 15 minutes at room temperature with antibodies to MLH1 (dilution 1/70, clone G168-728, Pharmingen), MSH2 (dilution 1/100, clone FE11, Calbiochem, Oncogene Research Products) and MSH6 (dilution 1/100, clone 44, Becton Dickinson). The Bond polymer Refine Detection Kit (Leica) was used as the detection system. Nuclear staining was interpreted for each antibody. Protein expression was considered negative when there was a complete absence of nuclear staining of neoplastic cells (15). The staining of adjacent lymphocytes, normal epithelial cells, endothelial cells, or fibroblasts was used as a positive control. IHC assays were read by pathologists blinded to the clinical characteristics of the patients; discordant cases were reviewed by a supplementary pathologist to reach a consensus.

Fractions of blocks of formalin-fixed, paraffin-embedded tissue were selected by pathologists and microdissected. DNA was extracted using the QIAamp DNA Minikit (Qiagen) following the manufacturer's recommendations. Five quasimonomorphic mononucleotide markers (NR21, NR24, NR27, BAT 25, and BAT 26) were studied as previously described (16, 17). Each sense primer was end labeled with one of the fluorescent markers: FAM, HEX, or NED. The five loci were amplified by pentaplex PCR. The PCR conditions consisted of an initial 15-minute denaturation step at 95°C, followed by 35 cycles at 95°C for 30 seconds, 55°C for 30 seconds and 72°C for 30 seconds, with a final extension at 72°C for 10 minutes. Fluorescent products were separated on an ABI Prism3100-Avant Genetic Analyzer (Applied Biosystems), and their length was determined with Genotyper software (Applied Biosystems). Specimens with a minimum of 3 unstable markers were scored as highly unstable (MSI-H), whereas specimens with less than 3 unstable markers were scored as stable (MSS; ref. 16, 17).

All analyses were carried out with a bilateral alpha type 1 error of 5%. Data were described as frequencies (percentages) or means and medians (range). The Fisher exact test was used to compare distributions of qualitative and ordinal variables. The primary endpoint was DFS, defined as the time between the date of surgery and the first event (local or distant disease recurrence or death from any cause, whichever occurred first). Patients who were alive and relapse free at the last contact were censored at the last follow-up date. OS was defined as the time elapsed from the date of surgery until death (all causes). Surviving patients were censored on the last follow-up date. Median follow-up and the 95% confidence interval (CI) were calculated using the reverse Kaplan–Meier method. Survival curve was estimated with the Kaplan–Meier method and compared using the log-rank test. The DFS rate at 3 years was reported according to MMR status with its 95% CI. Univariate and multivariate Cox proportional hazard regression models were used to estimate the HR and 95% CI. A multivariate Cox model was constructed according to the “one variable for 10 events” rule. Multivariate Cox analysis included all relevant clinical variables, whatever their univariate Cox P value, namely age, sex, tumor location, differentiation grade, bowel perforation or obstruction, tumor stage, median lymph node ratio (LNR), MMR status, and treatment center origin. The LNR was defined as the ratio of metastatic lymph nodes to all examined lymph nodes. The Harrell C-statistic was computed for discrimination (a Harrell C index of 0.5 indicates no predictive discrimination and a Harrell C index of 1.0 indicates perfect separation of patients). On the basis of a bilateral alpha type one error of 5%, an expected HR of 2.0 (tumor status: pMMR vs. dMMR), and an anticipated unbalanced distribution of pMMR (85%), it would be necessary to include 302 patients to achieve a 90% statistical power. Two-sided P values of less than 0.05 were considered statistically significant.

A total of 520 patients with stage III colon cancer were screened in the 9 centers. We excluded patients who died within 30 days after surgery (n = 6), patients treated by surgery alone without adjuvant chemotherapy (n = 75), patients treated with chemotherapy other than FOLFOX (n = 86), patients whose chemotherapy and/or surveillance after surgery were done in another center (n = 31), or when the delay between surgery and start of chemotherapy was more than 8 weeks (n = 3), paraffin tissue block was unavailable (n = 1), immunohistochemical test was uninterpretable (n = 2) or when clinical data were missing (n = 13; Supplementary Fig. S1). Therefore, the study population consisted in 303 patients with stage III colon cancer treated by FOLFOX. MMR status was done by immunohistochemistry in 210 patients, by MSI testing in 46 patients, and by both techniques in 47 patients. IHC analysis was conducted in APA center for all patients treated in APA and GP centers (n = 69); in IGR for all patients treated in IGR center (n = 13); and in STA for all patients treated in IMM and STA centers (n = 177). MSI analysis was conducted in AVI center for all patients treated in AVI, BOR, CUR, and MTF centers (n = 44), whereas immunohistochemistry was not done for these patients. MSI analysis was further conducted in GP for 8 patients treated in GP; and in STA for 39 patients treated in STA hospital. MSI testing and immunohistochemistry gave identical results in the 47 patients tested with both techniques. Thirty-four patients (11.2%) had dMMR tumors, as determined by immunohistochemistry (n = 23), MSI testing (n = 8), or both (n = 3). Among the 26 patients with dMMR tumors diagnosed by immunohistochemistry, 17 tumors exhibited a loss of MLH1 protein expression, and 9 a loss of both MSH2 and MSH6 protein expression. Among the 17 patients with a loss of MLH1 protein expression, 4 had a confirmed Lynch syndrome or positive family history according to revised Bethesda Guidelines (18), 10 had no Lynch syndrome or suggestive family history, and we have no information about family history or the existence of a genetic workup for the remaining 3 patients. Among the 9 patients with a loss of MSH2 and MSH6 protein expression, 6 had a confirmed Lynch syndrome or positive family history, and we have no information about family history or the existence of a genetic workup for the remaining 3 patients. Among the 8 patients with dMMR tumors tested exclusively by MSI analysis, one had a positive family history, 5 had no Lynch syndrome or suggestive family history, and no information about family history or the existence of a genetic workup was available for the remaining 2 patients.

The clinical and pathologic characteristics of the patients are summarized in Table 1. As expected, compared with pMMR tumors, dMMR tumors were significantly associated with a proximal location and a poor differentiation grade (7, 19). Patients with dMMR tumors tended to be younger than patients with pMMR tumors, but the difference did not reach statistical significance. All other characteristics were well balanced between the MMR tumors groups (Table 1). Moreover, there was no significant difference in the distribution of clinical and pathologic characteristics according to the participating centers, except for sex (P = 0.03) and bowel perforation/obstruction (P = 0.02; Supplementary Table S1). The rate of dMMR tumors varied between 4.7 % and 21.2 % but the difference was not significant between the participating centers (P = 0.20; Supplementary Table S1). Patients included from the MOSAIC trial were exclusively treated in IMM and STA hospitals. The characteristics of these patients are summarized in the Supplementary Table S2. LNR was preferred to N stage, based on publications showing that LNR is the most significant prognostic factor for both OS and DFS (20–22). In addition, the median LNR has been identified as an adequate breakpoint for predicting DFS in nonmetastatic CRC patients (21). We stratified the population around the median LNR of 0.113 (<0.113 = low; ≥0.113 = high).

The mean number of FOLFOX cycles received was respectively 9.7 and 10.5 in the pMMR and dMMR groups. Median follow-up was respectively 48.2 (95% CI, 47.2–51.5) months and 41.7 (95% CI, 37.1–53.4) months.

At the end of follow-up, 26 patients (9.7%) with pMMR tumors and one patient (2.9%) with dMMR tumor had died. Early estimation of the 5-year OS rate was 96.8% in patients with dMMR tumors and 86.9% in patients with pMMR tumors (HR = 2.11; 95% CI, 0.63–7.05; P = 0.23). For DFS analysis, 77 patients (28.6%) with pMMR tumors and 3 patients (8.8%) with dMMR tumors had relapsed or died. The 3-year DFS rate was 90.5% (95% CI, 73.2%–96.9%) in patients with dMMR tumors and 73.8% (95% CI, 67.9%–78.8%) in patients with pMMR tumors (log-rank test, P = 0.027; Fig. 1). On the basis of a bilateral alpha type one error of 5%, an observed HR of 2.16, and an unbalanced distribution of dMMR (34 of 303 = 0.11), with 303 patients and 80 events, calculated post hoc power is about 73%.

Among the variables analyzed in the univariate Cox model (age, sex, tumor location, differentiation grade, bowel perforation, or obstruction, tumor stage, LNR, MMR status, and treatment center origin), only tumor stage (stage IIIA/B vs. IIIC: HR = 2.01; 95% CI, 1.30–3.12; P = 0.002), LNR (<0.113 vs. ≥0.113: HR = 2.62; 95% CI, 1.61–4.24; P = 0.001) and MMR status (dMMR vs. pMMR: HR = 3.41; 95% CI, 1.08–10.80; P = 0.037) were significantly associated with improved DFS (Table 2). A trend toward better DFS was observed among patients with no bowel perforation or obstruction (Table 2).

Six patients were excluded from the multivariate model because they had at least one missing pathologic data among the variables analyzed. Multivariate analysis of the remaining 297 patients showed that only MMR status retained significant prognostic value for DFS (Table 3). DFS was significantly higher in patients with dMMR tumor status than in patients with pMMR tumor status (HR = 4.48; 95% CI, 1.34–14.99; P = 0.015; Table 3).

This large multicenter study is the first to show a significant prognostic impact of MMR status on DFS in patients with stage III colon cancer treated with adjuvant FOLFOX chemotherapy. DFS was analyzed after a median follow-up of 3 years, as this is the standard endpoint used in most adjuvant trials involving colon cancer patients and correlates strongly with long-term OS (23, 24). Median follow-up in our study was too short to effectively compare OS in the dMMR and pMMR groups.

To our knowledge, only 3 retrospective studies have evaluated the impact of MMR status on survival, each including about 100 patients treated with FOLFOX (12, 13, 25). In these studies, among patients with stage II or III CRC treated with adjuvant FOLFOX, DFS was similar or slightly longer in those with dMMR tumors compared with those patients with pMMR tumors, without reaching statistical significance (8). These results could be explained on the one hand by the low statistical power of these studies, and on the other hand by the inclusion, in 2 of the 3 studies, of stage II CRC patients, in whom adjuvant chemotherapy is less beneficial than in stage III patients (especially among patients with dMMR tumors; ref. 3, 11, 26). Moreover, because only 10% to 15% of patients with CRC have dMMR tumors, large studies are needed to obtain a representative population of dMMR patients. For these reasons, we chose to study only patients with stage III colon cancer treated with standard 5-FU–oxaliplatin combination chemotherapy. We also included sufficient patients to show a significant difference in DFS, in view of our initial statistical hypothesis.

MMR status as a predictive or prognostic biomarker according to the mechanism of MMR deficiency (MLH1 promoter silencing by hypermethylation vs. germline MMR gene mutation) has recently been evaluated in 5-FU adjuvant treatment but not in the FOLFOX setting yet. Indeed, the predictive and prognostic impact of presumed germline versus sporadic origin of dMMR tumors have recently been published on the basis of a large database set up from randomized trials of 5-FU–based adjuvant chemotherapy (27). After adjustment for patient age, prognosis was similar for suspected sporadic and germline dMMR tumors (27). In colon cancer stage III patients with suspected germline tumors, a greater DFS benefit was observed for 5-FU–based treatment compared with those receiving observation or no 5-FU, whereas no treatment benefit was observed in patients with sporadic tumors (27). These data suggest that differences in 5-FU response may be indirectly related to the mechanism of MMR deficiency. In our study, we observed that among 34 patients with dMMR tumors, 11 (32.4%) had a confirmed Lynch syndrome or positive family history, 15 (44.1%) had no Lynch syndrome or suggestive family history (corresponding to sporadic cases), and we had no information for eight patients (23.5%) about their family history or the existence of a genetic workup. Among the 3 dMMR patients with disease recurrence, one had a confirmed Lynch syndrome with MSH2 germline mutation, and the other two patients were probably sporadic cases without family history. The low proportion of patients with dMMR tumors and the low number of events did not allow us to conclude on the prognostic impact according to the mechanism of MMR deficiency. Furthermore, while the majority of dMMR cases arise on a defect in MLH1 deficiency due to silencing of its promoter by hypermethylation, which is the typical pattern of sporadic MSI cases, a significant proportion of patients displaying MSH2/MSH6 deficiency were observed in our series. Such an underrepresentation of sporadic cases in patients with dMMR tumors likely results from the fact that several dMMR sporadic cases were excluded as they were not given FOLFOX adjuvant chemotherapy because they were aged, frail, or with comorbidity. In addition, several participating centers have a cancer genetics department which could contribute to a recruitment bias in favor of hereditary cases. Consequently, in the group of patients with dMMR tumors, the beneficial effect of FOLFOX treatment may be contributed by the relative enrichment of presumed germline cases, which have recently been shown to be more sensitive to 5-FU chemotherapy than sporadic cases (27). Thus, adjunction of oxaliplatin to 5-FU could be amplifying the effect of adjuvant chemotherapy with 5-FU alone for hereditary MSI cancer, as established in the overall population stage III colon cancer (3, 4). Regarding the sporadic MSI cases who do not benefit from 5-FU adjuvant chemotherapy (27), the low recurrence rate (13.3 %) observed among presumably sporadic MSI cases treated with FOLFOX in our study (2 events among 15 patients) may suggest that oxaliplatin could overcome their tumor resistance to 5-FU. Nevertheless, these hypotheses could not be adequately assessed because this study was not randomized between FOLFOX and 5-FU, and included a too low number of patients with dMMR tumors.

Because IHC analysis with PMS2 antibodies was not conducted routinely in all participating centers, we did not include this marker in our protocol. The absence of immunohistochemistry with PMS2 staining may preclude detecting some cases with Lynch syndrome. However, MMR system alterations in Lynch syndrome are mainly due to constitutional mutations in MLH1 or MSH2 genes, more rarely to mutations in MSH6 gene, and very rarely to mutations in PMS2 gene (28). In keeping, no isolated MSH6 defect was detected in our series. In addition, no isolated PMS2 defect was detected in the three centers (APA, GP, and IGR) where PMS2 staining (dilution 1 of 100, clone A16-4, Pharmingen) is routinely assessed by immunohistochemistry (corresponding to 249 patients tested among whom 82 were included in our study). Thus, the risk of missing cases of Lynch syndrome due to PMS2 mutations can be considered as reasonably low in this work.

Our study has two main limitations. First, owing to its retrospective nature, we cannot exclude a selection bias. However, the prognostic value of tumor MMR status remained significant even after adjusting for validated prognostic factors in multivariate model analysis. Second, the lack of a patient group treated with 5-FU alone meant we could not fully assess the predictive effect of MMR status in the response to oxaliplatin-based adjuvant chemotherapy in stage III colon cancer. As prospective trials comparing FOLFOX and 5-FU adjuvant treatment are unlikely to be conducted, specifically in dMMR patients, the only step towards can now be obtained by analyzing tissue samples from previously completed prospective randomized trials such as the MOSAIC (3) and NSABP-C07 (4) studies, a goal that is currently pursued.

Various preclinical data suggested that the effect of 5-FU chemotherapy was dependent on the MMR status (29, 30). Several studies have shown that 5-FU induces G₂ arrest less efficiently in MMR-deficient cells than in their MMR-proficient counterparts (31–33). More recently, it was shown that, in the absence of a functional MMR system, repair may only occur through “base excision repair” system, a process that is less affected by the dNTP disequilibrium induced by 5-FU, possibly explaining why dMMR cells are more resistant to 5-FU (34). In contrast to 5-FU, sensitivity to oxaliplatin does not seem to be influenced by MMR status. Although oxaliplatin and cisplatin form DNA adducts resulting in similar distortion of secondary DNA structure (35), oxaliplatin-induced adducts differ from those created by cisplatin and are poorly recognized by MMR complexes (36). Defective MMR cells are thus expected to be as sensitive as pMMR cells to oxaliplatin (36–38). These observations have been confirmed in vitro using various CRC cell lines (39) and in vivo using a nude-mouse xenograft tumor model with dMMR and pMMR embryonic stem cells (40). The limitations of these preclinical works are that study of the cytotoxicity of 5-FU and oxaliplatin has not been evaluated in combination treatment and/or according to the mechanism of MMR deficiency (MLH1 promoter silencing by hypermethylation vs. germline MMR gene mutation). Finally, a possible biological explanation for the efficacy of oxaliplatin on dMMR tumors is the induction of an antitumor immune response. Indeed, oxaliplatin has been shown to induce immunogenic death of colon cancer cells, and this effect seems to be important in its therapeutic efficacy in patients with colon cancer (41). It is noteworthy in this respect that tumor-infiltrating lymphocytes are particularly abundant in patients with dMMR tumors (7).

In conclusion, this is the first study to show that MMR status is an independent prognostic biomarker in patients with stage III colon cancer treated with adjuvant FOLFOX chemotherapy. The predictive effect of MMR status in the response to FOLFOX adjuvant chemotherapy could not be evaluated because there is no control arm of patients with stage III colon cancer treated by 5-FU alone in our study. The prognostic and predictive value of MMR status on survival of patients treated with oxaliplatin-based adjuvant chemotherapy should be further adequately assessed by the undergoing analyses of tissue samples from previously completed randomized trials comparing 5-FU plus oxaliplatin with 5-FU alone, such as the MOSAIC and NASBP-C07 studies. The underlying biological mechanisms of MMR effect on adjuvant chemotherapy remain to be identified.

No potential conflicts of interest were disclosed.

The authors thank Prof. Emmanuel Tiret, head of the Surgical Department of Saint-Antoine Hospital, Prof. Anne Berger, head of the Surgical Department of Georges Pompidou European Hospital, Prof. Bernard Nordlinger, head of the Surgical Department of Ambroise Paré Hospital, Prof. Brice Gayet, head of the Surgical Department of Mutualiste Montsouris Institute, Dr. Pascale Mariani, Surgical Department of Curie Institute, and Dr. Mostefa Bennamoun, Medical Oncology Department of Montfermeil Hospital, for their contributions.

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.