Analysis of Molecular Markers by Anatomic Tumor Site in Stage III Colon Carcinomas from Adjuvant Chemotherapy Trial NCCTG N0147 (Alliance)

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer worldwide and is a leading cause of cancer-related death (1). The majority of colon cancers, especially distal cancers, display chromosomal instability (CIN) in association with frequent loss of heterozygosity and aneuploidy (2). CIN tumors are enriched in KRAS mutations that occur as early events during tumorigenesis (3). Alternatively, a subset of CRCs have microsatellite instability (MSI) and hypermutation due to deficient DNA mismatch repair (dMMR) (4). These tumors most commonly develop dMMR due to epigenetic inactivation of the MLH1 MMR gene and the CpG island methylator phenotype (CIMP), along with frequent oncogenic BRAF (V600E) mutations (5, 6). Colon cancers with BRAF^V600E, including those with dMMR, are believed to arise via a serrated neoplasia pathway named for the pattern of mucosal crypts in precursor polyps, and have a strong predilection for the proximal colon (7). Mutations in BRAF^V600E are mutually exclusive with KRAS, and both lead to the activation of MEK/ERK signaling pathway (3, 8) that promotes tumor progression. While BRAF mutation is associated with adverse outcome in the metastatic CRC (9, 10), less is known about its association or that of KRAS with patient outcome in earlier stage disease.

The primary site of colon cancer is typically categorized as located proximal to the splenic flexure, at or distal to this flexure, or in the rectum based upon their embryological origin. Mutations in BRAF^V600E and dMMR are known to be enriched in cancers of the proximal colon (9, 10), although their subsite distribution and that of KRAS remain unknown. While the mechanisms underlying tumor site/subsite-related biological differences are poorly understood, distal colon cancers show more frequent chromosomal aberrations whereas proximal cancers have higher mutation rates even after excluding the subset of hypermutated dMMR tumors (11). There is a need to examine the subsite distribution of relevant molecular alterations and to examine their prognostic association by anatomic site to gain biologic insights and further guide prognostication. Mortality rates have been reported to be higher in proximal compared to distal colon cancers, including lymph node-positive, i.e., stage III, tumors (12–14). However, studies examining tumor site and prognosis generally lack data on key molecular markers that precludes their adjustment in multivariable models, and cohorts have not received standard adjuvant chemotherapy with the FOLFOX regimen.

In this report, we examined the frequency of molecular alterations by primary tumor subsite, and tested the hypothesis that the association of BRAF^V600E and KRAS mutations with clinical outcome depends on their anatomic site within the colon. We also determined whether tumor site-related differences in clinical outcome persist after adjustment for BRAF^V600E and KRAS mutations as well as by MMR status. Our study population consists of stage III colon cancers from an adjuvant chemotherapy trial that found no survival benefit for the addition of cetuximab to standard FOLFOX chemotherapy (15). In contrast to prior reports in the N0147 cohort that were limited to tumors lacking KRAS mutations based upon the primary endpoint analysis (15, 16), the current study includes patients with KRAS mutated tumors and reports mature overall survival (OS) data for the trial.

METHODS

STATISTICAL ANALYSIS

We examined the relationship between the anatomic location of the primary tumor and biomarkers with time-to-recurrence (TTR), disease-free survival (DFS), and OS. All biomarker assays were interpreted with investigators blinded to patient outcomes. Treatment arms were pooled for all biomarker analyses given the lack of statistically significant differences in patient outcome or interactions between any of the biomarkers and treatment. Analysis within the FOLFOX alone treatment arm was also performed based upon clinical relevance. The Cochran-Armitage test for trend was used to assess the relationship between tumor site and biomarker status. Kaplan-Meier methods were used to describe the distributions of outcome (20). Univariate Cox proportional hazard models (21) were used to explore the associations of patient/tumor characteristics and biomarkers with outcome. Multivariate models were adjusted for stratification factors (see Study Population), tumor site, age, sex, treatment, and status of MMR, BRAF, or KRAS. TTR and DFS were calculated as the time from randomization until disease recurrence, or were censored at the date of last disease assessment if no recurrence occurred. For DFS, patients having died without recurrence were considered an event using date of death. OS was calculated as the time from randomization until date of death; surviving patients were censored for OS at the date of last contact/follow-up. All patients were censored at 5 years (DFS, TTR) and 8 years post-randomization (OS). Patients were analyzed according to treatment arm assignment including those who withdrew or were deemed ineligible. Median follow-up for 2,394 (79%) surviving patients was 4.9 years (range 0.0–8). Two-sided p-values are reported; values <0.05 were considered statistically significant and were not adjusted for multiple comparisons. Analyses were performed using SAS version 9.3 (SAS Institute Inc, Cary NC) and R version 3.02 (22).

RESULTS

DISCUSSION

We examined the association of mutations in BRAF^V600E and KRAS in stage III colon carcinomas with primary tumor subsite from patients treated in a large adjuvant chemotherapy trial. KRAS exon 2 mutations were more evenly distributed throughout the colon than were BRAF^V600E mutations. KRAS codon 12 mutations had the highest rates in tumors of the cecum and splenic flexure whereas the rate of codon 13 mutations was relatively similar across tumor subsites. KRAS mutations were detected in only 15% of dMMR tumors consistent with their more frequent association with pMMR and the CIN pathway (25). The distribution of BRAF^V600E mutations showed a marked increase with proximal migration and peak in frequency in tumors at the hepatic flexure. This distribution directly paralleled that of dMMR colon cancers that are known to have a strong predilection for the proximal colon (23, 26). Our findings in stage III cancers differ from those of Yamauchi et al (26) in all stages of CRC from a non clinical trial cohort where the frequencies of BRAF^V600E mutations, CIMP, and MSI-H increased linearly from the rectum to the ascending colon, although both studies found that KRAS mutations were most frequent in cecal cancers. Importantly, we found that the proximal colon predominance for BRAF^V600E mutations persisted when the cohort was restricted to pMMR cancers, consistent with the earlier development of BRAF^V600E-mutations that appear to develop in a site-specific manner during colonic tumorigenesis. BRAF mutant colon cancers are believed develop via the serrated neoplasia pathway from precursor lesions known as sessile serrated adenomas/polyps that are characterized by frequent BRAF mutations, CIMP-high, and proximal colon predominance (7). Analysis of stage II and III colon cancers from the PETACC-3 adjuvant study (11, 27) and the TCGA series (4) also found higher rates of BRAF^V600E and KRAS mutations in proximal vs distal colon cancers. While the mechanisms underlying the observed tumor site-related differences in the mutation rates of these oncogenes and dMMR remain unknown, they may reflect specific mechanisms of carcinogenesis and/or differences in embryologic origin with the proximal colon derived from the midgut and the distal colon from the hindgut. In addition, differences in colonic microbiota have been found in the proximal and distal regions of the colon that may influence molecular events and host immune responses during carcinogenesis in a site-specific manner (28, 29).

Colon cancers with BRAF^V600E mutations were significantly associated with higher rates of proximal site, age, female sex, high grade histology, dMMR status and N2 stage compared to mutant KRAS tumors. These clinicopathological features were similarly associated with dMMR tumors except for higher N stage. At 5 years of median follow-up, BRAF^V600E mutations were independently associated with adverse outcome, including OS, in the overall cohort (16) which is consistent with other studies (30–32). Studies also indicate that BRAF^V600E mutations are associated with worse patient survival after tumor recurrence (32–34); however, such an analysis in our cohort awaits more extended patient follow-up. In the overall cohort, we found that KRAS codon 12 mutations were significantly associated with worse outcome variables compared to tumors that were non-mutated for KRAS or BRAF. Mutations at codon 13 were also significantly associated with worse DFS as we reported previously (35), and a strong trend for worse TTR and OS was also found. KRAS codon 13 mutations were of borderline significance for poorer DFS in the PETACC-8 adjuvant study that had identical treatment arms to our trial, yet had fewer codon 13 mutated tumors (136 vs 231) (36). in a database from two nationwide prospective cohort studies, the lack of association of codon 13, in contrast to codon 12, mutations may also be due to sample size since only 108 codon 13-mutated colon cancers were analyzed (37). Patients with dMMR cancers showed a statistically significant delay in TTR and showed a trend for better DFS (p=0.10) and OS (p=0.18) rates compared to patients with pMMR tumors. These data are consistent with other studies where the association of dMMR or MSI-H status with favorable prognosis is attenuated in stage III compared to stage II colon cancer patients treated with 5-FU-based adjuvant chemotherapy (38). Among stage III disease, the attenuated effect of dMMR on survival may be related to our finding of a similar proportion of dMMR and pMMR N2 tumors where both N2 subgroups showed poorer outcome in the N0147 cohort (39).

We tested the hypothesis that the association of KRAS and BRAF^V600E mutations with clinical outcome depends on the anatomic site of the primary tumor. In the overall cohort, we found a statistically significant interaction between mutations in KRAS, but not BRAF^V600E, and primary tumor site for OS. Specifically, KRAS mutations were associated with adverse outcome in distal vs proximal cancers that reached statistical significance for OS. Non-mutated BRAF/KRAS tumors were enriched in distal tumors, and we previously reported that this patient subgroup had a favorable clinical outcome (35). We also observed a statistically significant interaction between MMR status and tumor site for TTR, DFS and OS. Specifically, proximal but not distal dMMR cancers were associated with significantly better outcomes. In a prior study, we also reported an interaction between MMR status and tumor site for survival in patients treated with 5-FU/leucovorin alone or combined with irinotecan (CALGB 89803) (16).

Since cetuximab has no role in the adjuvant therapy of colon cancer (15, 24), only data from the standard FOLFOX treatment arm are directly relevant to clinical practice. Accordingly, we performed a biomarker analysis restricted to the FOLFOX arm and found results that were highly consistent with those in the overall cohort. Specifically, the association of KRAS mutations with poorer outcomes remained more evident in distal vs proximal cancers. While the association of BRAF^V600E mutations with poorer OS was again limited to proximal tumors, the interaction test was not significant. Lastly, the better prognosis of dMMR vs pMMR tumors was limited to tumors of the proximal colon in the FOLFOX alone arm. Data from the PETACC-3 adjuvant trial that evaluated 5-FU/LV + irinotecan also found significantly better OS of patients with MSI-H vs MSS/MSI-L cancers of the proximal vs distal colon (38). The association of dMMR vs pMMR with significantly poorer outcome in distal tumors in the overall cohort was less evident in the FOLFOX alone arm potentially due to the small sample size.

An unresolved issue is whether primary tumor site-related differences in outcome persist after adjustment for MMR, KRAS and BRAF status. We found that tumor site was independently associated with prognosis after adjustment for BRAF^V600E and KRAS mutations as well as for MMR status in the overall cohort. Further analysis by tumor subsite revealed that anatomical migration of the primary tumor from the proximal to distal colon was significantly associated with a stepwise improvement in overall survival. Patients with cecal cancers had the worst outcome whereas those located in the sigmoid colon had the best survival rates. While the specific explanation for this finding is unknown, analysis of colon cancers from the PETACC-3 (11) and TCGA (4) studies found that proximal cancers have higher mutation rates even after excluding dMMR tumors, and potentially more deleterious mutations.

Data on molecular alterations by tumor subsite can improve our understanding of mechanisms of colon carcinogenesis with implications for clinical, epidemiological and pathological research. Our finding that primary tumor site remained prognostic after adjustment for molecular markers, suggesting its use as stratification factor for patients receiving adjuvant FOLFOX therapy. Strengths of our study include the large size of the clinical trial cohort, prospective analysis of biomarkers using uniform assay methodology in a single CLIA certified laboratory, treatment with the standard adjuvant FOLFOX regimen, and long term follow-up data. Only two other studies have examined molecular marker data from colon cancer patients treated with the adjuvant FOLFOX regimen (33, 40). Study limitations include that KRAS mutation analysis was limited to exon 2 and the inability to examine the predictive utility of the molecular markers given that all patients received adjuvant chemotherapy. We acknowledge potential sources of bias inherent in clinical trial cohorts including referral bias, the presence of strict eligibility criteria, and the requirement for tumor tissue in our study that may limit the generalizability of results.

In conclusion, mutations in BRAF or KRAS codon 12 were significantly more frequent in proximal cancers and associated with adverse outcome, whereas non-mutated BRAF/KRAS tumors were significantly enriched in distal tumors and associated with favorable outcome. The association of KRAS mutations with OS was dependent upon primary tumor site, with significantly poorer OS seen for patients with distal vs proximal cancers. The association of dMMR with significantly better outcome variables was limited to proximal tumors. These findings were maintained in the FOLFOX alone treatment arm which underscores their relevance to clinical practice. Taken together, our data demonstrate that use of KRAS and MMR status for prognostication in stage III colon cancer patients requires consideration of the location of the primary cancer within the colon.

Supplementary Material