Introduction

Use of Standardized Official Symbols: We use HUGO (Human Genome Organisation)-approved official symbols for genes and gene products, including BRAF; EGFR; KRAS; PIK3CA; all of which are described at http://www.genenames.org.

Colorectal cancer represents a heterogeneous group of diseases, and its molecular classification is increasingly important. Colorectal cancers can be classified using mutations in oncogenes such as KRAS, BRAF and PIK3CA[1]. In addition, microsatellite instability (MSI) and epigenomic instability, such as the CpG island methylator phenotype (CIMP) and LINE-1 hypomethylation, have been associated with the oncogene mutations and clinical outcomes[1–4].

Approximately 30-40% of colorectal cancers harbor KRAS mutations, typically in codon 12 or 13[5–9]. Features of colorectal cancers with KRAS codon 12 and 13 mutations include associations with cecal location[5, 8], low-level CIMP (CIMP-low)[10–14], and PIK3CA mutation[15–18]. KRAS codon 12 and 13 mutations are widely accepted as a predictive biomarker of lack of response to anti-EGFR therapy in metastatic colorectal cancer[19–23], though a few exploratory studies suggest that codon 13 mutants may benefit from EGFR-targeted therapy[24, 25].

KRAS codons 61 and 146 are additional hotspots for mutation in colorectal cancer, and data from a small number of studies suggest that KRAS mutation at these sites may predict resistance to anti-EGFR therapy[26–28]. Recently, Douillard et al., utilizing existing clinical trial data, reported that KRAS mutations in codons 61, 146, and 117, and mutations in NRAS, might identify patients with metastatic colorectal cancer who fail to derive benefit from panitumumab plus FOLFOX4[29]. Despite growing clinical relevance, the clinicopathological and molecular features of colorectal cancers with KRAS codon 61 or 146 mutation remain largely unknown. It is of interest to examine the characteristics of colorectal cancers with KRAS mutations in codons 61 and 146, compared to those in codons 12 and 13, and KRAS-wild-type cases. In the near future, routine clinical testing of these additional KRAS codons may be warranted.

We therefore investigated the clinicopathological, molecular, and prognostic characteristics of tumors harboring KRAS codon 61 and 146 mutations, utilizing a molecular pathological epidemiology[30, 31] database of 1267 colorectal cancers from two U.S. nationwide prospective cohort studies. We also performed a comprehensive review on KRAS codon 61 and 146 mutations in colorectal cancer, and our curated literature data can be readily useful for public databases such as the COSMIC (Catalogue of Somatic Mutations in Cancer) database.

Results

KRAS codon 12, 13, 61 and 146 mutations, in relation to clinicopathological and molecular features

We detected KRAS mutations in 505 (40%) cases in 1267 colorectal cancers (Table 1). Codon 12 mutations were present in 344 cases (27%), codon 13 mutations in 115 cases (9.1%), codon 61 mutations in 19 cases (1.5%), and codon 146 mutations in 40 cases (3.2%). There were 493 cases with KRAS mutations identified in only one of codons 12, 13, 61 and 146, and 12 cases with KRAS mutations identified in two or more of the four codons (Table 1).

Table 1 Frequencies of KRAS mutations in 1267 colorectal cancer cases
Full size table

The baseline characteristics of study subjects are summarized in Table 2, according to tumor KRAS mutation status. Compared to KRAS-wild-type tumors, overall KRAS-mutated cancers were less likely to exhibit poor differentiation (5.8%, P < 0.0001), MSI-high (6.2%, P < 0.0001), and BRAF mutation (1.4%, P < 0.0001), and more likely to demonstrate cecal location (24%, P < 0.0001), CIMP-low (49%, P < 0.0001), and PIK3CA mutation (24%, P < 0.0001). Of note, these trends were generally evident across case groups with specific mutated codons (Table 2). KRAS mutation status was not significantly associated with sex, age, body mass index (BMI), year of diagnosis, family history of colorectal cancer, disease stage, peritumoral lymphocytic reaction, or tumor LINE-1 methylation level. There was no significant difference in any of the features between the cases with KRAS mutations identified in only one codon (N = 493) and those with KRAS mutations identified in two or more codons (N = 12), though statistical power was limited, given only 12 cases with KRAS mutations identified in multiple codons (Additional file1: Table S1).

Table 2 Clinicopathological, and molecular characteristics according to KRAS mutation status in 1267 colorectal cancer cases
Full size table

KRAS mutation status and patient survival in BRAF-wild-type cases

To examine the prognostic role of KRAS mutation independent of BRAF mutation, within 1067 BRAF-wild-type cases (excluding BRAF mutants), we compared KRAS-mutated cancers to cases with wild-type KRAS in all four codons 12, 13, 61 and 146 (Additional file2: Table S2). We evaluated clinicopathological, molecular and survival data of 51 cases with KRAS codon 61 and 146 mutations (Additional file3: Table S3). There were 514 deaths, including 307 colorectal cancer-specific deaths, during a median follow-up of 11.7 years (interquartile range, 8.3-16.1 years) for censored cases.

The 5-year colorectal cancer-specific survival probabilities were 80.6% for cases with KRAS-wild-type/BRAF-wild-type tumors, 67.9% for cases with codon 12 mutations, 75.8% for cases with codon 13 mutations, 79.4% for cases with codon 61 mutations, and 76.7% for cases with codon 146 mutations. Specific KRAS mutations were significantly associated with patient survival in Kaplan-Meier analysis (log-rank P = 0.0014, Figure 1). In multivariate analysis, compared to KRAS-wild-type/BRAF-wild-type tumors, we observed a significant prognostic association for KRAS codon 12 mutation [multivariate hazard ratio (HR) = 1.45; 95% confidence interval (CI), 1.12-1.87; P = 0.0048; Table 3). However, neither mutation of KRAS codon 61 nor codon 146 was associated with patient outcome (Table 3). For cases with the 10 most common KRAS mutations across all four codons examined, those with the c.34G>T (p.G12C) mutation, and those with the c.35G>T (p.G12V) mutation experienced significantly higher colorectal cancer-specific mortality in Cox regression analysis [multivariate HR = 2.33; 95% CI, 1.36-3.99; P = 0.0021 for c.34G>T (p.G12C); multivariate HR = 2.13; 95% CI, 1.47-3.09; P < 0.0001 for c.35G>T (p.G12V); Table 3], even after adjusting a statistical significance level for multiple testing (P < 0.005). None of the three most common KRAS mutations in codons 61 and 146 [c.183A>C (p.Q61H), c.436G>A (p.A146T) and c.437C>T (p.A146V)] was associated with patient prognosis (Table 3), although statistical power was limited. Subgroup analyses of stage I-II cases (N = 544, Additional file4: Table S4), and stage III-IV cases (N = 414, Additional file5: Table S5) yielded similar results, although statistical power was limited.

Figure 1
figure 1

Kaplan-Meier curves for colorectal cancer patients with BRAF- wild-type tumors, according to tumor KRAS mutation status. (A) Colorectal cancer-specific survival. (B) Overall survival. Table indicates the number of patients who were alive and at risk of death at each time point after diagnosis of colorectal cancer.

Full size image
Table 3 Colorectal cancer patient mortality according to KRAS mutation status in 1067 BRAF -wild-type cases
Full size table

Discussion

Although a number of studies have examined codon 61 or 146 hotspot mutations in colorectal cancer (Additional file6: Table S6)[26–29, 32–74], clinicopathological, molecular, and prognostic characteristics of those mutations have not been well investigated. Our data, from 1267 tumors, suggest that approximately 5% of all colorectal cancers harbor KRAS mutations in codon 61 or 146, and those colorectal cancers generally show similar characteristics to tumors with KRAS mutations in codon 12 or 13 (including associations with cecal location, CIMP-low and PIK3CA mutations).

A variety of methods have been used for KRAS codon 61 and 146 analyses (Additional file6: Table S6)[26–29, 32–74], which might have contributed to a wide variation in the prevalence of those mutations. Generally, nonsequencing methods make it cumbersome to confirm multiple independent mutations, and make it difficult to detect multiple variations at one allele without employing an expanded panel of probes or primers. Of the sequencing-based methodologies, pyrosequencing has been shown to be more sensitive than Sanger sequencing in paraffin-embedded archival tissue, with the capacity to reliably detect mutant alleles at low abundance (5-10% mutant), which is common in solid tumors[75].

The association between cecal cancers and KRAS mutations is intriguing. Emerging data suggest that gut luminal contents and microbiota, which change along bowel subsites, play important roles in colorectal carcinogenesis[8, 76]. Our recent study on colorectal cancers in detailed subsites (from cecum to rectum) has shown that tumor molecular features (including BRAF mutation, MSI and CIMP-high) change along the bowel subsites, and that cecal cancers are associated with KRAS codon 12 and 13 mutations[5, 8]. In our current study, cecal cancers appeared to be significantly associated with overall KRAS mutation status, and this trend was evident across all four mutated codons. Further studies are needed to elucidate why KRAS mutations, irrespective of mutated codon, are particularly common in cecal cancers.

Examining associations of tumor molecular features can provide insights into carcinogenesis processes, and is important in cancer research[77–83]. Previous studies have demonstrated that KRAS codon 12 and 13 mutations are associated with aberrant DNA methylation patterns, namely CIMP-low[10, 11]. Our current study suggests that KRAS mutation, irrespective of mutated codon (statistical power was limited for codon 61 mutants), is associated with CIMP-low. It remains to be investigated why KRAS mutations are associated with CIMP-low in colorectal cancer. KRAS have been positively associated with PIK3CA mutations in colorectal cancer[15–18]. Our data suggest that KRAS mutations, irrespective of mutated codon, are associated with PIK3CA mutations. It has been reported that activated RAS signaling potentiates PI3K (phosphatidylinositol-4,5-bisphosphonate 3-kinase)/AKT signaling, which is augmented by the presence of PIK3CA mutations[84]. Considering a possible role for PIK3CA mutation as a predictive biomarker of response to adjuvant aspirin therapy in colorectal cancer[16], our finding may be of interest. KRAS codon 12 and 13 mutations have been inversely associated with BRAF mutation in colorectal cancer[17, 26, 33, 41]. Our current data suggest that KRAS mutations, irrespective of mutated codon, are inversely associated with MSI-high and BRAF mutations in colorectal cancer. LINE-1 methylation level is a surrogate marker for global DNA methylation, and has been reported to be associated with MSI-high and CIMP-high in colorectal cancer[85]. This study showed that LINE-1 methylation level in average did not significantly differ according to KRAS mutation status.

Experimental studies are consistent with our observations that both KRAS codon 61 and 146 mutations can contribute to carcinogenesis in a similar manner to oncogenic mutations in codons 12 and 13. As KRAS codon 12 and 13 mutations, codon 61 mutation results in oncogenic RAS with impaired GTPase activity, resulting in constitutive activation[86, 87]. KRAS codon 146 mutation-transfected HEK-293FT cells showed a larger amount of RAS-GTP compared to KRAS-wild-type-transfected cells[28]. These experimental data provide an insights into plausible functional roles of codon 61 and 146 mutations in carcinogenesis.

In our current survival analysis, there was no significant association between KRAS codon 61 and 146 mutations, and patient outcome. The prognostic value of KRAS mutation in colorectal cancer remains controversial[7, 88–92]. Of note, in our current study, when we separately examined specific KRAS mutations, codon 12 mutations [especially c.34G>T (p.G12C) and c.35G>T (p.G12V)] were significantly associated with inferior survival, which is consistent with the ‘RASCAL II’ meta-analysis[88]. Accordingly, the prognostic associations of KRAS mutations in colorectal cancer may vary by specific mutation. Considered in conjunction with evidence that KRAS codon 61 and 146 mutations possess weaker transforming potential than codon 12 mutations[40], it may be the case that KRAS codon 61 or 146 mutation is not associated with patient prognosis. However, considering the limited case and event numbers for KRAS codon 61 and 146 mutations, our survival analyses should be considered exploratory. Additional larger studies, perhaps necessitating pooling of data, are required to definitively assess the prognostic roles codon 61 and 146 mutations in colorectal cancer.

Several studies have examined the predictive value of KRAS mutation in codon 61 and/or 146 in metastatic colorectal cancer treated with anti-EGFR therapy (cetuximab or panitumumab)[26–28, 41, 43]. Pentheroudakis et al. did not observe any association between KRAS codon 61 or 146 mutation (N =11) and survival[41]. De Roock et al. showed that KRAS mutation in codon 61 (N =13), but not that in codon 146 (N =11), was significantly associated with lack of response to cetuximab[27]. Seymour et al. reported that KRAS codon 146 mutations (N =17) were not associated with overall or progression-free survival[43]. In contrast, Loupakis et al. reported that, among BRAF-wild-type cancers, KRAS codon 61 or 146 mutant cases (N = 8) experienced a significantly lower response rate and progression-free survival[26]. Indeed, a few experimental studies also reported that tumors harboring KRAS mutations in codons 61 and 146 were resistant to anti-EGFR therapy[28, 93]. In addition, a recent published study reported by Douillard et al., showed that RAS mutants (N = 108) with any mutation in KRAS codons 61, 117 and 146, or NRAS codons 12, 13, 61, 117 and 146, did not benefit from combined panitumumab plus FOLFOX4 chemotherapy[29]. In our dataset, due to scarcity of data on cancer treatment, we were unable to examine the important question of the predictive value of KRAS mutations in relation to anti-EGFR therapy. Further clinical studies in this area are clearly required.

The question arises as to whether it is worth investigating these relatively rare mutations in the clinical setting. Given that over 250,000 individuals each year die of colorectal cancer in Europe and the U.S., and most of these unfavorable outcomes are due to distant metastases, we estimate that every year approximately 10,000 cases have KRAS mutations in codon 61 or 146, and would be regarded as KRAS-wild-type through current KRAS codon 12 and 13 testing protocols. Considering that KRAS codon 61 and 146 mutations may also confer resistance to EGFR inhibitors[26–29, 93], patients who have metastatic colorectal cancer with KRAS mutation in codon 61 or 146 could receive more tailored management through clinical testing of these additional KRAS codons.

A limitation of this study is the absence of data on KRAS codon 117 mutation and NRAS mutations. As a result, we could not refine purer RAS-wild-type (both KRAS- and NRAS-wild-type in codons 12, 13, 61, 117 and 146), or examine clinicopathological, molecular and prognostic features of those whole RAS mutations in this study. Considering that RAS mutations in those codons have been reported to predict lack of response to anti-EGFR therapy in colorectal cancer[29], further studies are necessary to answer important questions about features across various RAS mutants. Nonetheless, KRAS codons 61 and 146 are the most frequent mutational hotspots after KRAS codons 12 and 13. In addition, our current analysis (N>1200) represents a large single study to date (Additional file6: Table S6)[26–29, 32–74], examining KRAS codon 61 and 146 mutations, in relation to other important molecular features in colorectal cancers, such as status of CIMP, MSI, BRAF and PIK3CA mutations. Sample size is a critical issue when assessing these relatively infrequent mutations. Indeed, smaller studies (N < 300, Additional file6: Table S6) demonstrate considerable variability in the frequencies and distribution of reported KRAS mutations, ranging from 0.4% to 9.3% for KRAS codon 61 mutations, and from 1.3% to 6.6% for KRAS codon 146 mutations (Additional file6: Table S6)[26–29, 32–74]. Given the relatively low frequencies of these mutations, a large sample size is a prerequisite for assessing the prevalence of these mutations and their associations with other tumor molecular characteristics.

There are advantages in utilizing the molecular pathological epidemiology[30, 31] database of the two U.S. nationwide prospective cohort studies to assess prevalence and associations of KRAS codon 61 and 146 mutations. Selection bias is an inevitable issue when analyzing cases identified from a few academic hospitals, since patients have selected hospitals based on referral, health insurance applicability, and/or their own preference. In contrast, a large population-based or multicenter study is desirable to decrease the degree of such selection bias. In this study, cohort participants who were diagnosed with colorectal cancer were treated at hospitals throughout the U.S., and thus constitute a more representative sample of colorectal cancers in the U.S. population than patients in a few academic hospitals.

Conclusions

Our data from over 1200 colorectal cancers demonstrate that KRAS codon 61 or 146 hotspot mutations are present in approximately up to 5% of colorectal cancers, and those cancers exhibit similar clinicopathological and molecular features to cancers with KRAS codon 12 or 13 mutation. Our current findings suggest that additional large-scale studies are warranted to assess clinical utility of KRAS codon 61 and 146 testing in colorectal cancer.

Materials and methods

Study population

We utilized two prospective cohort studies, the Nurses’ Health Study (N = 121,701 women followed since 1976) and the Health Professionals Follow-up Study (N = 51,529 men followed since 1986)[16]. Every two years, cohort participants have been sent follow-up questionnaires to identify newly diagnosed cancers in themselves and their first degree relatives. The National Death Index was used to ascertain deaths of participants as well as unreported lethal cancers. The cause of death was assigned by study physicians. Formalin-fixed paraffin-embedded tissue blocks were collected from hospitals where participants with colorectal cancer had undergone colorectal resection or diagnostic biopsy (for preoperatively-treated rectal cancers). We used 1267 colorectal cancer cases, diagnosed up to 2006, based on the availability of KRAS sequencing data. In order to examine the prognostic role of specific KRAS mutations, independent of BRAF mutation, BRAF-mutated cancers (N = 184), cases with missing BRAF mutation status (N = 5), and tumors with KRAS mutations identified in two or more of codons 12, 13, 61 and 146 (N = 11) were excluded. As a result, a final total of 1067 BRAF-wild-type cases were used for survival analyses (Figure 2, Additional file2: Table S2). Informed consent was obtained from all study subjects. This study was approved by the Human Subjects Committees at Harvard School of Public Health and Brigham and Women’s Hospital. All clinicopathological and molecular analyses were performed blinded to other data, including patient outcome.

Figure 2
figure 2

Flow chart of the current study. Cases with BRAF mutation (N = 184) and those without available BRAF mutation data (N = 5), were excluded from survival analyses. In addition, cases with KRAS mutations identified in two or more of codons 12, 13, 61 and 146 (N = 11) were excluded, in order to assess a prognostic effect of specific KRAS mutations individually.

Full size image

Histopathological evaluation

Hematoxylin and eosin-stained sections of all cases were examined by a pathologist (SO) unaware of other data. Tumor differentiation was categorized as well-moderate or poor (>50% vs. ≤50% gland formation). Peritumoral lymphocytic reaction was examined as previously described[94].

Sequencing of KRAS codons 61 and 146

DNA was extracted from paraffin embedded tissue as previously described,[16] and polymerase chain reaction (PCR) and pyrosequencing, targeted for KRAS codons 61 and 146, were performed. The PCR primers for amplifying KRAS codon 61 were, 5′-biotin-TGGAGAAACCTGTCTCTTGGATAT-3′ (for forward primer), and 5′-TACTGGTCCCTCATTGCACTGTA-3′ (for reverse primer), and those for KRAS codon 146 were 5′-ATGGAATTCCTTTTATTGAAACATC-3′ (for forward primer), and 5′-biotin-TTGCAGAAAACAGATCTGTATTTAT-3′(for reverse primer). The sequencing primers were 5′-TCATTGCACTGTACTCCTC-3′ (for codon 61), and 5′-AATTCCTTTTATTGAAACATCA-3′ (for codon 146). Dispensation orders were designed such that all possible mutations would be detected (Additional file7: Figure S1). All mutations were confirmed by replicate analysis.

Sequencing of KRAS codons 12 and 13, BRAF, and PIK3CA, and MSI analysis

We performed PCR and pyrosequencing targeted for KRAS (codons 12 and 13)[75], BRAF (codon 600) and PIK3CA (exons 9 and 20) as previously described[16]. MSI analysis was performed using 10 microsatellite markers (D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67 and D18S487)[8]. MSI-high was defined as instability in ≥30% of the markers. MSI-low (<30% unstable markers) tumors were grouped with microsatellite stable (MSS) tumors (no unstable markers) because we have previously demonstrated that these two groups show similar features[8].

Methylation analyses for CpG islands and LINE-1

Using validated bisulfite DNA treatment and real-time PCR (MethyLight), we quantified DNA methylation in eight CIMP-specific promoters [CACNA1G, CDKN2A (p16), CRABP1, IGF2, MLH1, NEUROG1, RUNX3 and SOCS1][8]. CIMP-high was defined as the presence of ≥6/8 methylated promoters, CIMP-low as 1-5/8 methylated promoters, and CIMP-negative as the absence of methylated promoters, according to established criteria[8]. In order to accurately quantify LINE-1 methylation levels, we used bisulfite pyrosequencing as previously described[8].

Statistical analysis

All statistical analyses were performed using SAS (Version 9.2, SAS Institute, Cary, NC). All P-values were two-sided. Univariate analyses were performed to investigate clinicopathological and molecular characteristics according to KRAS mutation status; a chi-square test or Fisher’s exact test was used for categorical data, while a Wilcoxon or Kruskal-Wallis test was applied to continuous data (age and LINE-1 methylation). To account for multiple hypothesis testing in associations between KRAS mutation and other 14 covariates, the P-value for significance was adjusted by Bonferroni correction to P = 0.0036 (=0.05/14).

The Kaplan-Meier method and log-rank test were used to estimate survival distribution according to KRAS mutation status. Cases were observed until death, or January 1st 2011, whichever came first. For analyses of colorectal cancer-specific mortality, deaths as a result of other causes were censored. Cox proportional hazards regression models were used to compute mortality HRs for specific KRAS mutations. A multivariate model initially included the following clinicopathological and molecular variables with less than 10% of patients showing missing information among those we have previously published; sex, age (continuous), BMI (<30 vs. ≥30 kg/m2), year of diagnosis (continuous), family history of colorectal cancer in any first-degree relative (present vs. absent), tumor location (cecum vs. ascending colon to sigmoid colon vs. rectum), tumor differentiation (well-moderate vs. poor), peritumoral lymphocytic reaction (absent-minimal vs. mild-marked), MSI (high vs. low/MSS), CIMP (high vs. low vs. negative), PIK3CA mutation (present vs. absent) and LINE-1 methylation (continuous), with stratification by disease stage (I, II, III, IV or unknown) was performed using the “strata” option in the SAS “proc phreg” command. A backward elimination was performed with a threshold of P = 0.20, to avoid overfitting. Cases with missing information for any of the categorical covariates [BMI (0.2%), tumor location (1.0%), tumor differentiation (0.7%), peritumoral lymphocytic reaction (4.6%), MSI (1.6%), CIMP (6.7%), and PIK3CA (7.6%)], were included in the majority category of the given covariate to avoid overfitting. We confirmed that excluding cases with missing information in any of the covariates did not substantially alter results (data not shown). To account for multiple hypothesis testing in associations between KRAS mutations and patient outcome, the P-value for significance was adjusted by Bonferroni correction to P = 0.025 [=0.05/2, for the two groups of codons (codons 12 and 13, and codons 61 and 146)], P = 0.013 (=0.05/4, for the four codons), or P = 0.005 (=0.05/10, for the 10 most common mutations). The proportionality of hazards assumption was satisfied by evaluating time-dependent variables, which were the cross products of the KRAS indicator variables and survival time (all P- values>0.07).

Literature search

A systematic literature search was performed in Pubmed, up to April 5, 2014, using combinations of the following search terms; KRAS, codon, (61 or 146), (colon, rectal or colorectal), and (cancer, carcinoma or adenocarcinoma). All eligible publications were retrieved, and their references were checked to identify further relevant studies. In addition, we contacted some corresponding authors to obtain detailed data.