Main

Colorectal cancer is the third most common cancer in the world,1 with nearly 1.4 million new cases diagnosed each year. An estimated, 700 000 people die from colorectal cancer worldwide each year. The 5-year overall survival rate is ~59%.2 About 45% of colorectal cancers are associated with activating mutations of the KRAS gene. Mutations are located mostly in codons 12 (~30%) and 13 (~8%) of exon 2.3, 4 These somatic mutations result in a constitutive activation of the epidermal growth factor receptor (EGFR) pathway and, therefore, confer resistance to anti-EGFR therapy. Previous studies have shown that KRAS mutations in exon 2 are associated with no clinical benefit from anti-EGFR monoclonal antibodies in metastatic colorectal cancers.5, 6 KRAS mutation has been also associated with worse disease-free survival in adjuvant setting.7, 8 Recently, other mutations in KRAS (exons 3 and 4) and NRAS genes (exons 2, 3, and 4) have been reported to be associated with resistance to anti-EGFR monoclonal antibodies too.9 Anti-EGFR monoclonal antibodies, panitumumab and cetuximab, were consequently restricted to patients with metastatic colorectal cancers harboring a complete wild-type RAS genotype (KRAS and NRAS exons 2, 3, and 4 testing).9, 10, 11, 12

Approximately 10% of colorectal cancer are BRAF-mutated.13 BRAF mutation in colorectal cancer is associated with poor prognosis, especially in metastatic colorectal cancer.14 In addition, mutations in RAS and BRAF genes are mutually exclusive.15, 16, 17, 18, 19 Colorectal cancers with deficient mismatch repair system account for ~15% of all colorectal cancers. Deficient mismatch repair system is due to germline mutation in a mismatch repair gene (Lynch syndrome) or more commonly due to epigenetic inactivation of the MLH1 gene (sporadic cases). As compared with proficient mismatch repair colorectal cancers, deficient mismatch repair colorectal cancers are associated with good prognosis in non-metastatic setting20 and high disease control with drugs that target the immune checkpoints in metastatic setting.21

Recent studies have shown that KRAS exon 2 and BRAF-mutated colorectal cancers have distinct clinical and pathological characteristics. KRAS exon 2-mutated colorectal cancers occur more frequently in older patients, with a predominance of male gender, and are frequently located in the proximal colon compared with KRAS wild-type colorectal cancers.22, 23 High frequency of KRAS-mutation has been reported especially in the cecum.22, 24 Nevertheless, limited number of patients is a major weakness of those studies, leading to inconsistent conclusions. For instance, some studies have mentioned an association between KRAS mutations and mucinous differentiation, whereas others have not.22, 23 Moreover, correlation between KRAS mutation subtypes and clinicopathological characteristics was never clearly appraised. Gonsalves et al. showed that KRAS codon 13 mutations (p.Gly13Asp) are associated with deficient mismatch repair status and poor histologic grade.15 BRAF mutations are significantly associated with advanced age, female gender, poor histologic grade, mucinous differentiation, proximal colon tumor site, and deficient mismatch repair status.15, 25

Up until now, few studies have evaluated the possible correlation between clinicopathological features and complete RAS status (exons 2, 3, and 4 of KRAS and NRAS). A small series of 264 metastatic colorectal cancers in Japanese patients reported some association between KRAS exon 2 mutations and others RAS mutations with the rectal tumor site.25 The aim of this study was to identify new distinct subsets of colorectal cancers based on clinicopathological features and complete RAS and BRAF genotype of a large cohort of 1735 colorectal cancer patients.

Materials and methods

Population

All patients harboring colorectal cancer with complete RAS testing, performed at the Molecular Cancer Genetics Platform of Poitiers (French National Cancer Institute (INCa)) from January 2011 to April 2015, were included in the study (Figure 1). Since 2006, INCa has been supporting a national network of 28 hospital molecular genetics platforms throughout France, offering to patients an access to all essential molecular genetics testing for cancers. Overall, 2813 consecutive colorectal cancer cases with a molecular testing were reviewed to implement the cohort. Tumors with multiple testing were included only once (n=36). Tumors with incomplete RAS/BRAF status (n=834) were excluded. Indeed, before December 2013 there was no complete RAS testing (only KRAS exon 2 mutations were analyzed). It became systematic for every colorectal cancer after December 2013. Other rare histological types than adenocarcinoma were also excluded (n=6). Tumors with both RAS and BRAF mutations, or with two different RAS mutations, were also excluded (n=7). Finally, 1735 tumors with complete RAS and BRAF testing were included in the study. Results of microsatellite instability analysis were also recorded (n=700).

Figure 1
figure 1

Flow chart.

Full size image

Molecular Analyses

Tumoral DNA was extracted from paraffin-embedded tumor sections using KAPA Express Extract Kits (KAPA Biosystems, CliniSciences, Nanterre, France) and GeneAmp PCR System 9700 Thermal Cycler (Applied Biosystems, Waltham, MA, USA).

According to the manufacturer’s instructions, RAS and BRAF molecular testing was performed using polymerase chain reaction (PCR) amplification and pyrosequencing technology Q24 PyroMark system (Qiagen, Hilden, Germany) with CE-IVD-certified allele-specific (Qiagen) or homemade primers.26 Mutational status of KRAS (exon 2: codons 12 and 13, exon 3: codons 59 and 61, and exon 4: codons 117 and 146), NRAS (exon 2: codons 12 and 13, exon 3: codons 59 and 61, and exon 4: codons 117 and 146), and BRAF (exon 15: codon 600) was analyzed.

Mismatch repair system status was determined with Microsatellite Instability Analysis System (Promega, Charbonnières-les-Bains, France) and multiplex PCR fluorescence assay using a panel of five mononucleotide repeat markers (BAT-25, BAT-26, NR-21, NR-24, and MONO-27) known to be monomorphic in Caucasian population. Samples were analyzed by capillary electrophoresis and data were deciphered using fragment analysis software (GeneMapper Software, Applied Biosystems). Colorectal cancers with two or more unstable loci were classified microsatellite-unstable and, thus, with deficient mismatch repair system.

Pathological Features

The following pathological features were analyzed using a pathological report database: histologic subtype (classical adenocarcinoma, mucinous adenocarcinoma, signet-ring cell adenocarcinoma, medullary adenocarcinoma, adenosquamous carcinoma, and undifferentiated carcinoma), histologic grade (well, moderately, or poorly differentiated), tumor site (proximal colon, transverse colon, distal colon, and rectum), vascular invasion (lymphatic and/or venous invasion/embols), and perineural invasion. Primary tumors located in the cecum and ascending colon were defined as proximal tumors, whereas tumors located in the splenic flexure, descending colon, and sigmoid colon were defined as distal tumors.27 Mucinous colorectal adenocarcinoma was defined by at least 50% of the tumor volume composed of extracellular mucin.28, 29

Definition of RAS Status and Different Subgroups

RAS and BRAF mutations are considered to be mutually exclusive (at 99%).15, 16, 17, 18, 19 Consequently, tumors with a RAS mutation were considered BRAF wild type, if no BRAF testing was performed. In the same way, tumors with BRAF mutation were considered RAS wild type, if RAS testing was incomplete.

Then, different groups of colorectal cancers were set, whether they presented RAS mutation or BRAF mutation, or none (super wild-type colorectal cancers), and they were compared with each other. Subtypes of KRAS mutants (KRAS exon 2: KRAS codon 12 and codon 13, KRAS exon 3 and KRAS exon 4 mutants) were also compared with each other.

Statistical Analyses

Clinical, pathological, and molecular variables collected at baseline were described as means and s.d.’s for quantitative variables and percentages for qualitative variables. Associations between mutational status and patients or tumor characteristics were assessed using the χ2-test (or Fisher’s exact test if appropriate) for qualitative variables and using Student's t-test for continuous variables. For all analyses an adjustment for multiple testing has been performed using Bonferroni correction. Statistical analyses were performed using the Statview software (Statview for Windows, SAS Institut, version 5.0).

Results

Patient and Tumor Characteristics

RAS and BRAF genotype was available for 1735 patients. The median age was 70.5 years and 57.9% were of male gender (Table 1). Histologic subtypes were 88.7% of classical adenocarcinoma, 10.1% of mucinous adenocarcinoma, and 1.2% of other adenocarcinoma subtypes. The majority of colorectal cancers were moderately differentiated (55.2%). Tumor sites were proximal, transverse, distal colon, and rectum at 37.0%, 4.3%, 39.2%, and 19.5%, respectively.

Table 1 Association between RAS status and clinicopathological features
Full size table

There was 55.6% of KRAS mutation (n=965/1735), 2.1% of NRAS mutation (n=37/1735), 15.5% of BRAF mutation (n=269/1735), 26.7% of super wild-type colorectal cancers (n=464/1735), and 13.4% of deficient mismatch repair colorectal cancers (n=94/700).

Comparison of subgroups of super wild-type, KRAS, NRAS, and BRAF mutants showed significant association with several features, namely age, gender, histologic subtype, histologic grade, tumor site, and microsatellite instability (all P<0.0001; Table 2). To highlight differences between mutational status, a 2 by 2 comparison was then conducted (mutated versus wild type).

Table 2 Association between KRAS, NRAS, BRAF, super wild-type status, and clinicopathological features
Full size table

Association Between RAS/Frequent KRAS Mutations and Clinicopathological Features

Among the 1735 colorectal cancers, 733 were RAS wild-type (42.2%) and 1002 were RAS mutants (57.8%). There was no age difference among patients according to RAS status, 70.4 years for RAS-mutated colorectal cancers, and 70.6 years for RAS wild-type colorectal cancers (Table 1). Compared with RAS wild-type colorectal cancers, RAS-mutated colo-rectal cancers were statistically associated with male gender (60.7% versus 54.0%; P=0.006), classical adenocarcinoma subtype (90.6% versus 86.1%; P=0.005), well/moderately differentiated tumors (93.4% versus 83.5%; P<0.0001), and microsatellite-stable phenotype (95.3% versus 75.8%; P<0.0001).

Secondly, we looked at the association between RAS mutation subtypes and clinicopathological features. Among RAS-mutated colorectal cancers, 965 were KRAS mutants (exons 2, 3, or 4; 96.3%) and 37 were NRAS mutants (exons 2, 3, or 4; 3.7%). Like RAS-mutated colorectal cancers, well/moderately differentiated tumors (P<0.0001) and microsatellite-stable phenotype (P<0.0001) were found to be associated with KRAS-mutated colorectal cancers as compared with KRAS wild-type colorectal cancers (Table 2). Moreover, results tended to show a possible association with male gender (P=0.016) and classical adenocarcinoma subtype (P=0.014). Among the colorectal cancer cohort, there were 909 KRAS exon 2 mutations (52.4%) distributed between 738 KRAS codon 12 mutants (42.5%) and 171 KRAS codon 13 mutants (9.9%; Table 3). KRAS exon 2 and KRAS codon 12 mutants, in comparison with RAS wild-type colorectal cancers, were also significantly associated with the same clinicopathological features as RAS-mutated colorectal cancers (classical adenocarcinoma subtype, well/moderately differentiated tumors and microsatellite stable phenotype; data not shown). Tumors with KRAS codon 13 mutants, in comparison with RAS wild-type colorectal cancers, were more frequently classical adenocarcinoma (95.3% versus 86.1%; P=0.004), with microsatellite-stable phenotype (90.9% versus 75.8%; P<0.0001), with perineural invasion (43.8% versus 29.9%; P=0.018) and located in the proximal colon (50.3% versus 37.4%; P=0.011).

Table 3 Association between KRAS-mutant status and clinicopathological features
Full size table

KRAS codon 13-mutated colorectal cancers, compared with KRAS codon 12-mutated colorectal cancers, were more frequently poorly differentiated (11.5% versus 5.5%; P=0.005) and were located in the proximal colon (50.3% versus 33.8%; P=0.002; Table 3).

Association Between Rare KRAS/NRAS Mutations and Clinicopathological Features

Rare KRAS mutations were either KRAS exon 3 mutants (n=23, 1.3%) or KRAS exon 4 mutants (n=33, 1.9%; Table 3). Compared with RAS wild-type colorectal cancers, results tended to show that rare KRAS-mutated colorectal cancers (n=56) were associated with rectal site (31.2% versus 16.9%; P=0.043) and microsatellite-stable phenotype (93.5% versus 75.8%; P=0.024).

When comparing the different KRAS mutation groups to each other, regarding distribution of histologic subtypes and tumor site, we found significant discrepancies (Table 3). KRAS exon 3-mutated colorectal cancers were more frequently associated with mucinous/rare histological subtypes (17.3% versus 9.5%; P=0.002) and tended to be associated with rectal tumor site (33.3% versus 21.1%; P=0.009). KRAS exon 4-mutated colorectal cancers were associated with mucinous/rare histological subtypes (21.2% versus 9.5%; P=0.002). NRAS mutation was present in 37 cases (2.1%). Clinicopathological features in NRAS-mutated colorectal cancers were not significantly different compared with NRAS wild-type colorectal cancers (Table 2). When comparing NRAS-mutated and KRAS-mutated groups, no significant disparity was observed (data not shown).

Association Between BRAF Status and Clinicopathological Features

Patients with BRAF-mutated colorectal cancers were significantly older than patients with BRAF wild-type colorectal cancers (respectively, 74.7 and 68.2 years) and were associated with female gender (respectively, 59.5% and 38.9%; all P<0.0001; Table 2). Moreover, BRAF-mutated colorectal cancers were statistically associated with proximal tumor site (68.2% versus 31.0%), mucinous differentiation (24.5% versus 7.4%), poorly differentiated tumors (31.6% versus 7.0%), and microsatellite instability (57.0% versus 6.2%; all P<0.0001).

Clinicopathological Features of RAS and BRAF Wild-Type Colorectal Cancers

Patients with super wild-type colorectal cancers were significantly younger than patients with RAS or BRAF mutations (68.3 versus 71.3 years; P<0.0001). Moreover, super wild-type tumors were significantly associated with classical adenocarcinoma subtype (94.0% versus 86.8%; P<0.0001) and distal tumor site (54.5% versus 34.0%; P<0.0001) compared with RAS- or BRAF-mutated tumors (Table 2).

Discussion

To our knowledge, this retrospective study is one of the first to analyze associations between complete RAS and BRAF mutational status and clinicopathological features in a large cohort of colorectal cancers. As compared with RAS wild-type tumors, RAS mutants, KRAS mutants, KRAS exon 2 mutants, and KRAS codon 12 mutants tended to be associated with the same clinicopathological features, ie, male gender, classical adenocarcinoma subtype, well/moderately differentiated tumors, and microsatellite stable phenotype. KRAS codon 13 mutant colorectal cancers were more frequently classical adenocarcinoma subtype and were with microsatellite stable phenotype. For the first time, we highlighted some associations between rare RAS mutations (KRAS exons 3 and 4; NRAS mutations) and clinicopathological features. As compared with other KRAS mutations, KRAS exon 3-mutated colorectal cancers were more frequently associated with mucinous/rare histological subtypes and, most likely to the rectal tumor site, whereas KRAS exon 4-mutated colorectal cancers were associated only with mucinous/rare histological subtypes. In contrast, there was no significant association between NRAS mutation and any clinicopathological feature.

Mutation rates of RAS, KRAS, NRAS, and BRAF genes were 57.7%, 55.6%, 2.1%, and 15.5%, respectively. Among all of them, as expected, KRAS exon 2 (codons 12/13) mutation was the most frequent with 52.4%. These mutation rates were slightly above data in the literature. In recently published studies, around 50% of colorectal cancers presented a RAS mutation with an average of 45% of KRAS mutants and 40% of KRAS exon 2 mutants.3, 7, 8, 30 BRAF mutation rate in the literature is around 10%, which is slightly lower than the BRAF mutation rate found in our cohort of patients (15.5%).14, 31, 32 Sensitivity of the different molecular techniques used could explain the disparity. Since 2010, our genomic platform has been using pyrosequencing, a robust technique known to be more sensitive than Sanger sequencing used in some other platforms. In addition, our population may be enriched in RAS/BRAF mutants as colorectal cancers with incomplete RAS testing or colorectal cancers with RAS wild-type status but no BRAF testing or colorectal cancers with BRAF wild-type status but no RAS testing were automatically excluded (colorectal cancers analyzed between 2011 and 2013). Concerning rare RAS mutations (KRAS exons 3 and 4 and NRAS exons 2, 3, and 4), 5.4% colorectal cancers were concerned in our study. Rare RAS mutations were scattered between KRAS exon 3 mutants (1.3%), KRAS exon 4 mutants (1.9%), and NRAS mutants (2.1%). These rates are lower than the 10% rate previously reported by various studies.3, 9, 25, 33 Apart from the exclusion of colorectal cancers with incomplete RAS testing and from the use of different techniques with different specificity and sensitivity, no clear explanation can be drawn. The 13.4% of colorectal cancers with deficient mismatch repair system was in accordance with data found in the literature (≈12%).34, 35 In our study, patients and tumor characteristics were consistent with previously published data.30, 36, 37, 38, 39, 40

Complete RAS status with analyses of exons 2, 3, and 4 underlined some distinct clinicopathological and molecular characteristics. RAS-mutated colorectal cancers, and more precisely KRAS exon 2 and KRAS codon 12 mutants, were significantly associated with classical adenocarcinoma subtype, well/moderately differentiated tumors, and microsatellite stable phenotype. Most of our results were consistent with the literature concerning KRAS exon 2, notably the association with well/moderately histologic grade.25 Nevertheless, the data found in literature present some disparities, and some dissimilarity can be observed concerning KRAS mutations and tumor sites or histologic subtypes. Some studies have shown an association between KRAS mutation and either right colon8 or rectal tumor sites.25 Other studies precisely determined that cecal cancers exhibited significantly higher frequency of KRAS mutations than other tumor sites.22, 24 Rosty et al. has also observed gender-related distribution of KRAS-mutated carcinoma between different colonic segments.22 Association between KRAS mutation and mucinous differentiation was affirmed in some studies but denied in others.19, 20 In the same way, two studies observed that KRAS exon 2-mutated colorectal cancers seemed to occur more frequently in elderly patients, but other studies did not.22, 23 These inconsistent results could arise from a low patient number and lack of robustness of some studies. We analyzed a large cohort and found no association between KRAS exon 2 mutations with tumor site, neither with age nor with mucinous differentiation. No significant association between KRAS mutations and tumor site in our study could be explained by the fact that no data on cecum tumor site were available. Yamauchi et al. introduced a new concept of continuum colorectal cancer characterized by linear gradual changes of key tumor molecular frequencies from the rectum to the ascending colon.41 However, cecal cancers did not follow the same continuum trend and stand for a unique colorectal cancer subtype characterized by high frequency of KRAS mutation.24 Indeed, there are some differences among proximal colorectal cancers, which is a heterogeneous subgroup.

A recent publication showed that, when compared with other KRAS mutations, KRAS codon 13 mutation was associated with deficient mismatch repair phenotype and poor histologic grade.15 In our population, KRAS codon 13 mutation was correlated with poor tumor differentiation, but also with the proximal colon site. To our knowledge, there exist no published data concerning the association of rare RAS mutations (KRAS exons 3 and 4 and NRAS mutations) and clinicopathological features. We demonstrated that, in comparison with colorectal cancers with other KRAS mutations, KRAS exon 3-mutated colorectal cancers and KRAS exon 4-mutated colorectal cancers were both associated with mucinous/rare histological subtypes. Moreover, KRAS exon 3-mutated colorectal cancers tended to be associated with rectal tumor site. Finally, there was no significant association between NRAS mutations and clinical or pathological features. Therefore, correct and complete determination of the RAS genotype is required as clinicopathological features are associated with colorectal cancer prognosis value and varied according to RAS mutation.

According to previous studies, patients with BRAF-mutated colorectal cancers are aged, mostly women, with right colon tumor site, either poorly differentiated or mucinous tumors, and had tumor with deficient mismatch repair status.15, 25, 42, 43 Our results were in complete accordance with the literature.

Concerning super wild-type colorectal cancers, they were significantly associated with younger age, classical adenocarcinoma subtype, and distal colon tumor site. It is worth noting that no other study has ever evaluated the correlation between clinicopathological features and super wild-type colorectal cancers. As expected, the correlations found for clinicopathological features were opposite to those discovered for KRAS and BRAF mutated-colorectal cancers.

Part of our study limitations is because of missing clinical, histological, molecular data, or incomplete information on tumor site, explained by the fact that it is an observational retrospective study. Nevertheless, aside from vascular and perineural invasions, the major data were efficiently collected with less than 20% of missing information. Having no data on the exact proximal location site (ie, cecum or not) is one of the limitations of our study. Because of unique molecular status of cecal tumors, detailed site information in future molecular studies in colorectal cancers is necessary. Our study can be criticized on the choice made to consider RAS and BRAF mutations as mutually exclusive. However, only 1–2% colorectal cancers may carry both mutations, which is a too small percentage to change the statistical results obtained, considering the size of our study.15, 16, 17, 18, 19 On the other hand, the strength of our study was to collect and statistically correlate data from one of the largest cohorts of colorectal cancers (n=1735). To take into account multiple testing, we used Bonferroni correction but some P-value are above the alpha level and we cannot exclude true association. Above all, this is the first report correlating complete RAS and BRAF analysis with colorectal cancers’ clinicopathological features.

In conclusion, this study provides a novel broader approach of clinicopathological features according to mutational status in colorectal cancers. In particular, KRAS exon 2 (codon 13), exon 3, and 4 mutations have distinct clinical, pathological, and molecular characteristics, and must be carefully considered when assessing the prognosis value of RAS status and in clinical trials in colorectal cancers.