Purpose: To determine the prevalence and prognostic value of mismatch repair (MMR) status and its relation to BRAF mutation (BRAFMT) status in metastatic colorectal cancer (mCRC).
Experimental Design: A pooled analysis of four phase III studies in first-line treatment of mCRC (CAIRO, CAIRO2, COIN, and FOCUS) was performed. Primary outcome parameter was the hazard ratio (HR) for median progression-free survival (PFS) and overall survival (OS) in relation to MMR and BRAF. For the pooled analysis, Cox regression analysis was performed on individual patient data.
Results: The primary tumors of 3,063 patients were analyzed, of which 153 (5.0%) exhibited deficient MMR (dMMR) and 250 (8.2%) a BRAFMT. BRAFMT was observed in 53 (34.6%) of patients with dMMR tumors compared with 197 (6.8%) of patients with proficient MMR (pMMR) tumors (P < 0.001). In the pooled dataset, median PFS and OS were significantly worse for patients with dMMR compared with pMMR tumors [HR, 1.33; 95% confidence interval (CI), 1.12–1.57 and HR, 1.35; 95% CI, 1.13–1.61, respectively), and for patients with BRAFMT compared with BRAF wild-type (BRAFWT) tumors (HR, 1.34; 95% CI, 1.17–1.54 and HR, 1.91; 95% CI, 1.66–2.19, respectively). PFS and OS were significantly decreased for patients with BRAFMT within the group of patients with pMMR, but not for BRAF status within dMMR, or MMR status within BRAFWT or BRAFMT.
Conclusions: Prevalence of dMMR and BRAFMT in patients with mCRC is low and both biomarkers confer an inferior prognosis. Our data suggest that the poor prognosis of dMMR is driven by the BRAFMT status. Clin Cancer Res; 20(20); 5322–30. ©2014 AACR.
This is the first pooled analysis on individual patient data to assess the role of the mismatch repair (MMR) status in relation to the BRAF mutation (BRAFMT) status in respect to prevalence and outcome in patients with metastatic colorectal cancer (mCRC). These patients participated in four large randomized prospective phase III studies, namely the CAIRO, CAIRO2, COIN, and FOCUS studies. We show that the prevalence of deficient MMR (dMMR) and BRAFMT is low in patients with mCRC. Both biomarkers confer an inferior prognosis. We observed a higher incidence of BRAFMT in dMMR tumors than reported for patients with early-stage dMMR colorectal cancer, and our data suggest that the poor prognosis of dMMR is driven by BRAFMT status.
Colorectal cancer is a heterogeneous disease arising through different pathways (1, 2). Three molecular pathways are well known to be involved in the multistep process of colorectal carcinogenesis, including the chromosomal instability (CIN) pathway, the mutator pathway [microsatellite instability (MSI)], and the epigenetic instability pathway or CpG island methylator phenotype (CIMP), the latter of which has substantial overlap with the other two.
MSI is the result of a deficient DNA mismatch repair (dMMR) system. A germline mutation in one of the MMR genes, most often MLH1 or MSH2, is the cause of dMMR in patients with Lynch syndrome, which comprises 0.8% to 5% of all colorectal cancers (3). dMMR is also observed in 10% to 20% of patients with sporadic colorectal cancer, of which the majority of dMMR tumors are due to inactivation of MLH1 (∼95%), caused by hypermethylation of the gene promoter, with MSH2 and MSH6 accounting for a smaller percentage (3–5). These dMMR tumors have distinct features, such as origin in the proximal colon, prominent lymphocytic infiltrate, poorly differentiated morphology, mucinous or signet ring differentiation (6), and association with a favorable prognosis in early-stage colorectal cancer (7). In metastatic colorectal cancer (mCRC), the prevalence of dMMR is low (3.5%; refs. 8, 9). This supports the hypothesis that dMMR tumors have a reduced metastatic potential (10, 11). Because of its lower frequency, the prognostic role of dMMR in mCRC has not been properly evaluated thus far.
The presence of a BRAF mutation (BRAFMT) in a dMMR tumor indicates a sporadic origin, and essentially excludes a diagnosis of Lynch syndrome (12, 13). In colorectal cancer, the overall prevalence of BRAFMT is approximately 10% (14). BRAFMT has a negative prognostic impact, although this may be restricted to patients with proficient MMR (pMMR) tumors (15, 16). Data on the role of BRAF in relation to MMR status in mCRC are scarce and are derived from small subsets of selected patients.
The current study was initiated to assess the role of MMR in relation to the BRAFMT status in respect to prevalence and outcome in patients with mCRC who participated in four large prospective phase III studies: CAIRO (17), CAIRO2 (18), COIN (19, 20), and FOCUS (21).
Materials and Methods
Patients and treatment
Data were derived from patients with mCRC included in four large phase III studies in first-line treatment: CAIRO (ClinicalStudys.gov; NCT00312000), CAIRO2 (ClinicalStudys.gov; NCT00208546), COIN (ISRCTN; 27286448), and FOCUS (ISRCTN; 79877428), of which the results have been published previously (17–21). Collection of formalin-fixed paraffin-embedded material (FFPE) of the primary tumor was part of the initial protocol in all four studies.
For samples of both CAIRO studies, immunohistochemistry (IHC) was performed on FFPE tissue with antibodies against MMR proteins hMLH1, hMSH2, hMSH6, and hPMS2. In addition, MSI analysis was performed where there was an absence of MMR protein expression or equivocal IHC results. dMMR status was determined using two microsatellite markers (BAT 25 and BAT 26). If only one of these markers showed instability, the analysis was extended with four additional markers (BAT 40, D2S123, D5S346, and D17S250). A tumor was defined as dMMR if at least two of the six markers showed instability or pMMR if none of the markers showed instability. Tumors with only one of the markers showing instability were defined as dMMR-low and included in the pMMR category. For samples from the COIN study, dMMR status was assessed using two microsatellite markers (BAT25 and BAT26). If only one of these markers showed instability, the tumor was defined as dMMR, and as pMMR if no instability was observed. For samples from the FOCUS study, dMMR status was based on loss of MLH1 and MSH2 protein expression, assessed by IHC. If either protein showed loss of expression, the tumor was defined as dMMR, and pMMR if no loss of expression was observed.
Hypermethylation status of the MLH1 gene promoter
Hypermethylation of the MLH1 gene promoter in patients with a dMMR tumor was analyzed in samples from the CAIRO and CAIRO2 studies only and therefore not included in the pooled analysis. The DNA methylation status of the MLH1 promoter region was determined after bisulfite treatment of the DNA using the EZ DNA Methylation Kit (ZYMO Research), as described previously (8).
The BRAF V600E mutation status was assessed in duplicate by high-resolution melting (HRM) sequencing analysis for tumor material in the CAIRO study (22) and by direct sequencing analysis in the CAIRO2 study (23). For samples of the COIN and FOCUS studies, the BRAF V600E mutation status was determined by Pyrosequencing (and Sequenom in COIN), and verified by Sanger sequencing as described previously (19, 24). Non-V600E BRAFMT detected by these assays (n = 19) were not included in the current analyses on outcome.
Individual patient data were included in the pooled analysis. Progression-free survival (PFS) was defined as the time from the date of randomization to first progression or death, whichever came first. Overall survival (OS) was defined as the time from randomization to the date of death. The primary outcome measure was the hazard ratio (HR) for PFS and OS in relation to MMR and BRAFMT status. For PFS and OS, all studies were included in a Cox regression model (proportional hazard model) by using the study as a factor in the model. In this way, dependence of the hazard on study could be modeled. The HR was corrected for study effect. Survival curves were plotted and log-rank tests were performed to compare survival for the different groups defined. A statistical interaction analysis for survival data of dMMR and BRAF status was performed. All analyses were conducted using the SAS system version 9.2; P < 0.05 was considered statistically significant.
Study population and MMR/BRAFMT status
Tumor and normal samples from 3,063 out of 6,155 randomized mCRC patients were available and suitable for analysis of both MMR and BRAFMT status. Of these 3,063 patients, 322 patients participated in the CAIRO study, 516 patients in the CAIRO2 study, 1,461 patients in the COIN study, and 764 patients in the FOCUS study.
The prevalence of MMR status and BRAFMT status and their correlation are presented in Tables 1 and 2, respectively. dMMR was found in tumors of 153 (5.0%) patients and 250 (8.2%) patients had a BRAFMT (Table 1). There was no evidence of heterogeneity for the prevalence of dMMR and BRAFMT in the four studies; P = 0.614 and P = 0.943, respectively (Table 1). A BRAFMT was observed in 53 (34.6%) of patients with dMMR tumors compared with 197 (6.8%) of patients with pMMR tumors (P < 0.001; Table 2). There was heterogeneity for the prevalence of combined MMR and BRAFMT status between the four studies. In the CAIRO study, there were significantly more patients with a combined dMMR and BRAFMT (dMMR/BRAFMT) tumor compared with the other three studies (P = 0.002; Table 2).
Patient and tumor characteristics (sex, age, location of the primary tumor, performance status, and number of metastatic sites involved) for the different subgroups defined by the combined MMR and BRAFMT status are summarized in Supplementary Table S1. Hypermethylation of MLH1 was the main cause of dMMR in both CAIRO and CAIRO2 studies (30 out of 45 patients), this was associated with a high frequency of BRAFMT (73%) compared with tumors without MLH1 hypermethylation (7%).
The survival data of the individual studies, the pooled dataset, and the pooled analysis for patients with dMMR, pMMR, BRAFMT, and BRAF wild-type (BRAFWT) tumors are presented in Table 3. The median PFS and OS were significantly worse for patients with dMMR compared with pMMR tumors [PFS: 6.2 vs. 7.6 months, respectively; HR, 1.33; 95% confidence interval (CI) 1.12–1.57; P = 0.001; OS: 13.6 vs. 16.8 months, respectively; HR, 1.35; 95% CI, 1.13–1.61; P = 0.001). Median PFS and OS were also significantly worse for patients with BRAFMT compared with BRAFWT tumors (PFS: 6.2 vs. 7.7 months, respectively; HR, 1.34; 95% CI, 1.17–1.54; P < 0.001; OS: 11.4 vs. 17.2 months, respectively; HR, 1.91; 95% CI, 1.66–2.19; P < 0.001).
To determine a possible interaction between MMR and BRAF status, with respect to the survival, a Cox regression was performed by using the study as a factor in the model. For PFS and OS, all studies were included in a Cox regression model (proportional hazard model) by using the study as a factor in the model. Results are presented for MMR status in a BRAFMT and BRAFWT background, and vice versa for BRAF status in a dMMR and pMMR background in Table 4. Survival curves, as estimated by the Cox regression, are presented in Fig. 1. In BRAFMT tumors stratified by MMR status, there was no significant survival difference for patients with dMMR compared with pMMR tumors (PFS: 6.1 vs. 6.2 months, respectively; HR, 0.95; 95% CI, 0.62–1.46; P = 1.000; OS: 11.7 vs. 11.3 months, respectively; HR, 1.05; 95% CI, 0.68–1.63; P = 1.000). Also in BRAFWT tumors stratified by MMR status, there was no significant survival difference for patients with dMMR compared with pMMR tumors (PFS: 6.3 vs. 7.8 months, respectively; HR, 1.32; 95% CI, 1.00–1.75; P = 0.051; OS: 15.0 vs. 17.3 months, respectively; HR, 1.22; 95% CI, 0.91–1.65; P = 0.463). In dMMR tumors stratified by BRAF status, there was no significant survival difference for patients with BRAFMT compared with BRAFWT tumors (PFS: 6.1 vs. 6.3 months, respectively; HR, 1.07; 95% CI, 0.67–1.70; P = 1.000; OS: 11.7 vs. 15.0 months, respectively; HR, 1.51; 95% CI, 0.93–2.46; P = 0.155). In pMMR tumors stratified by BRAF status, there was a significantly decreased median PFS and OS for patients with BRAFMT compared with BRAFWT tumors (PFS: 6.2 vs. 7.8 months, respectively; HR, 1.34; 95% CI, 1.10–1.64; P < 0.001; OS: 11.3 vs. 17.3 months, respectively; HR, 1.94; 95% CI, 1.57–2.40; P < 0.001) The test for interaction between dMMR and BRAFMT was statistically not significant (PFS: HR, 0.79; 95% CI, 0.54–1.16; P = 0.234; OS: HR, 0.78; 95% CI, 0.52–1.15; P = 0.211).
This study presents the largest dataset on the role of tumor MMR status and BRAFMT status in respect to prevalence and outcome in a population of patients (n = 3,063) with mCRC who participated in four prospective phase III studies. We found that dMMR and BRAFMT in mCRC each have a low prevalence (5% and 8.2%, respectively), and that both biomarkers indicate a poor prognosis. Given the absence of a statistically significant interaction between BRAFMT and dMMR, our data suggest that the poor prognostic value of dMMR is driven by the BRAFMT status.
Several aspects of our study warrant further discussion. In this pooled analysis, different methods for detecting dMMR were applied, which, however, have all been validated for the detection of dMMR in colorectal cancer. In both CAIRO studies, an approach based on test methods described in the Bethesda criteria, used for standard clinical practice for patients suspected for Lynch syndrome, has been applied (25). The COIN study analyzed the BAT25 and BAT26 mononucleotide markers, which have a high sensitivity (94%) and specificity (98%), and the use of these two markers alone identifies 97% of MSI tumors (26). The FOCUS study evaluated MLH1 and MSH2 protein expression by IHC, which is a sensitive (92.3%) and specific (100%) method for screening for dMMR (27).
We acknowledge that the difference in MMR detection methods represents a weakness of our study; however, the comparable prevalence of the dMMR status among the four studies in this pooled analysis, ranging from 4.4% to 5.6%, argues against this. The results from the individual studies show that the patient population with dMMR tumors is heterogeneous. The observed difference in the prevalence of a BRAFMT in dMMR tumors suggests a possible difference in the origin of dMMR, sporadic versus hereditary. Unfortunately, data on the hypermethylation status of the MLH1 gene promoter, which could differentiate between these two groups, are not available of all four studies.
Furthermore, different methods for detecting the BRAF V600E mutation were applied. HRM sequencing, Sanger sequencing, and Pyrosequencing have all shown to be reliable methods (22, 28). Data from systematic studies to assess the test accuracy or reproducibility of the different techniques used for BRAFMT testing are not available.
Another issue is the difference in availability of tumor samples among the trials. This is partly caused by nonavailability of an extra paraffin-embedded block for DNA analysis, and partly due to nonresected primary tumors in patients with synchronous disease. In these patients, often only a diagnostic biopsy was performed, which does not provide sufficient material for further molecular analysis for research purposes. This is an important, underexposed issue that may introduce a sample/case bias not only in our analysis, but in other translational studies in mCRC as well.
The low prevalence of dMMR in mCRC can be explained by the reduced potential of stage I–III dMMR tumors to metastasize (10, 11). However, the underlying mechanisms of this low metastatic potential are yet to be elucidated. It has been suggested that a greater immunoreactivity of dMMR tumors (29, 30) or decreased tumor cell viability due to excessive DNA damage (31) may play a role. In mCRC, data about the prevalence of BRAFMT in dMMR tumors are scarce, but in line with our results (32, 33). The strong inter-relationship between BRAFMT and dMMR is well established in early-stage colorectal cancer (14, 34); however, the etiology of both alterations still needs to be elucidated.
We observed a higher prevalence of BRAFMT in mCRC dMMR tumors (34.6%) than reported for early-stage dMMR colorectal cancer tumors (24%; 16). Patients with early-stage dMMR in general have a better prognosis compared with patient with early-stage pMMR; however, within the group of dMMR, patients with BRAFMT tumors have a worse prognosis (35). Subsequently, this may lead to a shift in the dMMR/BRAFMT ratio in patients with mCRC. There is increasing evidence identifying BRAFMT as a significant poor prognostic factor in early stage and mCRC (18, 36–38). BRAF is an oncogene and it is known that the mutations constitutively activate the MAPK pathway for cell growth, in the absence of extracellular stimuli. However, by itself BRAF is not sufficient for cancer and must cooperate with other processes to induce the fully cancerous state (39). Another explanation for the inferior prognosis of BRAFMT tumors might be their distinct pattern of metastatic spread. Previous studies have demonstrated a significantly increased rate of peritoneal and distant lymph node metastases and a decreased rate of lung metastases compared with BRAFWT tumors (9, 40).
It has been speculated that the worse prognostic value of dMMR tumors in mCRC may be related to a difference in metastatic spread. Earlier studies showed a reduced rate of liver metastases for dMMR tumors in mCRC (40), and a higher incidence of peritoneal metastases; these factors are known to be related to prognosis (41, 42). This was confirmed by a previous analysis of the COIN study (9), but these data are not available from the other studies of our analysis.
Finally, due to the different treatment regimens among the four studies of this pooled analysis, the predictive role of dMMR and BRAFMT in mCRC could not be addressed.
In conclusion, dMMR and BRAFMT each have a low prevalence in mCRC, and both biomarkers confer a poor prognosis. Our data suggest that the poor prognosis of dMMR is driven by the BRAFMT status. However, we caution against a firm conclusion on this issue because our study was not sufficiently powered to test this interaction.
Disclosure of Potential Conflicts of Interest
J.P. Cheadle reports receiving a commercial research grant from Merck. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S. Venderbosch, T.S. Maughan, C.J.A. Punt, M. Koopman
Development of methodology: S. Venderbosch, I.D. Nagtegaal, C.J.A. Punt, M. Koopman
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I.D. Nagtegaal, T.S. Maughan, C.G. Smith, J.P. Cheadle, R. Kaplan, P. Quirke, M.T. Seymour, S.D. Richman, G.A. Meijer, B. Ylstra, D.A.M. Heideman, M. Koopman
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Venderbosch, I.D. Nagtegaal, C.G. Smith, J.P. Cheadle, D. Fisher, P. Quirke, M.T. Seymour, A.F.J. de Haan, C.J.A. Punt, M. Koopman
Writing, review, and/or revision of the manuscript: S. Venderbosch, I.D. Nagtegaal, T.S. Maughan, C.G. Smith, J.P. Cheadle, D. Fisher, R. Kaplan, P. Quirke, M.T. Seymour, S.D. Richman, G.A. Meijer, B. Ylstra, D.A.M. Heideman, A.F.J. de Haan, C.J.A. Punt, M. Koopman
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Kaplan, D.A.M. Heideman, M. Koopman
Study supervision: P. Quirke, C.J.A. Punt
Other (laboratory analyses and interpretation of FOCUS data): P. Quirke
CAIRO: This study was supported by the Dutch Colorectal Cancer Group (DCCG), a grant support from the Commissie Klinisch Toegepast Onderzoek (CKTO) of the Dutch Cancer Foundation (KWF), unrestricted scientific grants from Roche, Aventis, Sanofi, and Pfizer, the Cornelis Visser Foundation and the framework of CTMM, the Center for Translational Molecular Medicine, DeCoDe project (grant 03O-101).
CAIRO2: This study was supported by the DCCG, and grants for data management and analysis from the CKTO of the Dutch Cancer Foundation and unrestricted research grants from Roche, Merck Serono, Sanofi-Aventis. DxS.
COIN: This study was supported by the Bobby Moore Fund from Cancer Research UK, Cancer Research Wales, and the NISCHR Cancer Genetics Biomedical Research Unit. The COIN trial was funded by Cancer Research UK and the MRC. An unrestricted educational grant from Merck Serono provided additional support to this work and the trial.
FOCUS: This study was supported by Merck KGgA, Yorkshire Cancer Research, Leeds Experimental Cancer Medicine Center, and the Leeds CRUK Center.
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.
The authors thank the patients and their families who participated in all the studies and gave their consent for this research, and the investigators and pathologists who submitted samples for assessment. The CAIRO studies were conducted with the support of the network of the Dutch Colorectal Cancer Group (DCCG). The COIN and FOCUS studies were conducted with the support of Cancer Research UK, the UK Medical Research Council (MRC) and the National Cancer Research Networks.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
- Received February 8, 2014.
- Revision received June 28, 2014.
- Accepted August 6, 2014.
- ©2014 American Association for Cancer Research.