BRAF mutations are associated with distinctive clinical, pathological and molecular features of colorectal cancer independently of microsatellite instability status

BRAF is a member of the RAF family of kinases that acts upstream of the MEK1/2 kinases in response to RAS signals. Activating mutations in BRAF have been reported in 5–15% of colorectal carcinomas (CRC), with by far the most common mutation being a 1796T to A transversion leading to a V600E substitution [1‐3]. The BRAF V600E hotspot mutation is strongly associated with the microsatellite instability (MSI+) phenotype but is mutually exclusive with KRAS mutations [4‐7]. Interestingly, BRAF mutations are found only in MSI+ sporadic tumors that result from aberrant MLH1 promoter methylation and do not occur in MSI+ tumors from hereditary non-polyposis colorectal cancer (HNPCC) patients [5, 8‐10], thus providing a convenient discriminator between sporadic and familial cases. The majority of MSI+ sporadic tumors belong to a larger CRC group referred to as the CpG island methylator phenotype (CIMP+) that is characterised by widespread hypermethylation of CpG islands located with gene promoter regions [11]. Both MSI+ and CIMP+ tumors are thought to arise from large hyperplastic polyps and serrated adenomas [12, 13] and recent work has demonstrated a high frequency of BRAF mutations in these lesions [7, 14, 15].

Although the positive association with MSI+ and inverse association with KRAS mutation have been well documented, little is known about the other properties of tumors with BRAF mutation. In the present study we analysed for BRAF V600E mutations in a consecutive series of 275 CRCs that were well characterised for the major pathological and molecular features of this disease. Our results demonstrate that oncogenic BRAF mutation occurs preferentially within a subgroup of CRCs that have distinctive features. It could therefore be used as a convenient marker for the further characterisation of these tumors, particularly in relation to their prognosis and response to adjuvant chemotherapy.

Figure 1A shows representative Fluorescent-SSCP results for the screening of BRAF mutations in this CRC series, while Figure 1B shows DNA sequencing confirmation of the 1799T to A transversion resulting in the V600E mutation. The overall frequency of BRAF mutation was 8.4% (23/275), comparing favourably with frequencies of 9–11% reported for other large studies of this tumor type [6, 16, 17]. The mean age of patients with and without BRAF mutation was identical (Table 1). Strong associations were observed between BRAF mutation and tumor origin in the proximal side of the large bowel, poor histological grade, mucinous appearance and the presence of infiltrating lymphocytes. Higher frequencies of BRAF mutation were also observed in females and in node negative tumors but these did not reach significance.

BRAF mutations showed no association with TP53 mutations and were mutually exclusive with the presence of KRAS mutations (Table 2). In contrast, BRAF mutations were approximately 10-fold more frequent in MSI+ and CIMP+ tumors compared to tumors without these phenotypes. A strong association was also seen with methylation of the MLH1 gene promoter and in particular with methylation of its proximal region. We have previously examined the methylation status of 7 different CpG islands in this CRC series [18]. The mean number of these methylated sites was 3-fold higher in tumors with BRAF mutation compared to those without (2.6 ± 1.7 vs 0.8 ± 1.0; P < 0.001). Multivariate analysis revealed that MSI+ was the only significant independent predictor of BRAF mutation (RR = 6.3, 95%CI [1.2–32.3]; P = 0.028) in a model that included CIMP+, tumor site, histological grade, presence of infiltrating lymphocytes and mucinous appearance.

We next examined whether the characteristic features of tumors with BRAF mutation were still apparent following stratification into MSI and CIMP phenotypes. Although the statistical power of this subgroup analysis was limited, the morphological features of infiltrating lymphocytes, poor histological grade and mucinous appearance were clearly associated with BRAF mutation regardless of tumor MSI status (Table 3). Similarly, these features were each more common in tumors with BRAF mutation in both the CIMP- and CIMP+ subgroups (Table 4). Similar to previous observations in a separate CRC cohort [20], the frequency of KRAS mutation was lower in MSI+ compared to MSI- tumors (P = 0.034; Table 3), while the frequency of TP53 mutation was also considerably lower in MSI+ tumors with wildtype BRAF than in MSI- tumors with wildtype BRAF (P = 0.014).

The BRAF V600E mutation has already been proposed as a convenient marker to discriminate between MSI+ tumors that are sporadic or HNPCC in origin [5, 8‐10]. This is a very important issue for population-based screening programs that aim to identify CRC associated with the HNPCC syndrome. Compared to the analysis of MLH1 promoter methylation, mutation at the BRAF V600E hotspot is relatively simple to detect using DNA sequencing, RFLP or the SSCP method used in the present work (Figure 1).

Similar to other studies [4, 5, 10, 16, 17] we observed BRAF mutation frequencies of 4% in MSI- tumors and 39% in MSI+ tumors (Table 1). The highest frequencies were seen in tumors showing methylation of the MLH1 promoter proximal region (46%) and in tumors with infiltrating lymphocytes (48%). BRAF mutation frequencies of up to 70–80% have been reported in sporadic MSI+, CIMP+ and MLH1-methylated CRC and polyps [7, 8, 15, 16]. For reasons that are still unclear, BRAF mutations are approximately 5–10-fold more frequent in tumors that have characteristic features of sporadic MSI+ (ie. MLH1 methylated) and CIMP+ phenotypes. These include proximal colon location, poor differentiation, mucinous histology and infiltrating lymphocytes [13, 19, 20]. Interestingly however, in the present study BRAF mutations never occurred in association with KRAS mutation, were present in only 3% of CIMP- tumors and showed no association with TP53 mutation (Table 2). The observation that BRAF mutations occur only very rarely in HNPCC-related MSI+ CRC demonstrates that defective DNA mismatch repair is not involved in causing this genetic alteration.

In order to determine whether the characteristic clinicopathological features of tumors with BRAF mutation were due to their close association with MSI+ and CIMP+, we stratified tumours according to these phenotypes. Despite having only 9 MSI-/BRAF mutant and 5 CIMP-/BRAF mutant tumors, the results showed that associations between BRAF mutation and the morphological properties of tumor-infiltrating infiltrating lymphocytes, poor histological grade and mucinous phenotype were retained (Tables 3 and 4).

The frequencies of BRAF mutation observed in MSI- (4%) and MSI+ (39%) tumors in the present study compare favourably (5% and 52%, respectively) to those reported recently in another large, population-based study [17]. Although BRAF mutations are much more frequent in MSI+ tumors, the comparative rarity of this phenotype means that a considerable proportion occur in MSI- tumors. In the present study, 43% of all BRAF mutations occurred in MSI- tumors compared to 48% in the study by Samowitz et al [17]. BRAF mutations were reported to show prognostic significance in MSI- but not in MSI+ CRC [17]. The lack of follow-up information on CRC patients in the current study and the small number of BRAF mutations (n = 21) meant that we were unable to evaluate the prognostic significance of BRAF mutation according to MSI status.

Findings from the present study and from previous work indicate that BRAF mutation is likely to be a convenient marker for the identification of a subset of CRCs with distinctive clinical, pathological and molecular features and which may originate in hyperplastic polyps and serrated adenomas [7, 14, 15]. In view of the strong associations between BRAF mutation and specific pathological (site, grade, mucinous, infiltrating lymphocytes) and molecular (methylated MSI+, CIMP+, wildtype KRAS) features, it will be interesting in future studies to determine the predictive significance of this marker for response to adjuvant therapies in CRC.

The 275 colorectal tumors investigated in this study were obtained from the Colorectal Unit of the Royal Adelaide Hospital. These were snap frozen in liquid nitrogen within 20–40 min after resection and stored at -70°C prior to extraction of DNA. Clinical data available for this series included patient age, sex and family history of CRC. Only one case was confirmed as HNPCC-related. Pathological data included nodal involvement, tumor site, histological grade, mucinous appearance and the presence of infiltrating lymphocytes. Evaluation of MSI+ [21], CIMP+ [18], KRAS mutation [22] and TP53 mutation [23] were performed as described previously by our group. Mutations in exon 15 of BRAF including the V600E hotspot were detected using the PCR primer sequences reported earlier [1], the F-SSCP method [22, 23] and confirmed by direct sequencing.

Statistical analyses were performed using SPSS Version 12.0 (Chicago, Illinois, USA). Associations between BRAF mutation and clinical, pathological or molecular features were evaluated using Fisher's exact or Pearson's chi-squared tests as appropriate. Multivariate analysis was performed using binary logistic regression with BRAF mutation as the dependent variable.