Prognostic impact of mutation profiling in patients with stage II and III colon cancer

Development of colorectal cancer (CRC) associates with accumulation of genetic mutations include the epidermal growth factor receptor (EGFR) signaling pathway. However, whether mutations in KRAS together with downstream factors BRAF, PIK3CA and NRAS impact prognosis is still unclear for stage II-III colon cancer. In the present study a total of 228 stage II-III colon cancer samples were retrospectively collected, KRAS (codons 12, 13 and 61), BRAF (exon 11 and exon 15), PIK3CA (exon 9 and exon 20) and NRAS (codons 12, 13 and 61) status was detected by Sanger sequencing, 37.89% (86/227) tumors harbored a KRAS mutation, 7.02% (16/228) harbored a BRAF mutation, 13.18% (29/220) harbored a PIK3CA mutation and 0.89% (2/224) harbored a NRAS mutation. NRAS mutations existed only in stage II colon cancer. Older groups harbored a higher KRAS and BRAF mutation (P < 0.05), PIK3CA (exon9) mutations appeared more common in worse differentiation tumors (P = 0.032). Moreover, PIK3CA (E545K) mutation was significantly associated with tumor recurrence (P = 0.031) and acted independently prognostic for poor OS (P = 0.044), while only in stage III colon cancer. KRAS, BRAF and NRAS mutations do not have major prognostic value in stage II and III colon cancer, subtypes of gene mutations should be further investigated for a better understanding in CRC.

with prognostic value for stage II or III cancers. NRAS is closely to KRAS, which also a member of RAS gene 23 , although NRAS gene mutation is rare in CRC, approximately 2. 2-4.19% 16,22,24 , it appears a valuable prognostic biomarker in anti-EGFR MoAbs therapy for mCRC patients 5,25 , then NRAS status for indicating prognosis should be further confirmed in early stage cancers.
The tumorigenesis and development of CRC is a multistep process with different genetic mutation accumulation 26 , driver mutations included somatic changes in KRAS, BRAF, PIK3CA and NRAS represent the main aspect in genetic alternations for CRC 27 . To date, major of the presented data were derived from western populations and few data was available for the Chinese. Herein we design this study to investigate the mutation spectrum in stage II/III colon cancer and the potential correlations with clinicopathological characteristics, furthermore, survival data of patients treated with 5-fluorouracil (5-FU)/oxaliplatin-based chemotherapy was collected, which may provide an appropriate insight between gene mutation and survival status in Chinese populations.
NRAS (codons 12, 13 and 61) mutations existed only in stage II colon cancer, 0.89% (2/224) tumors harbored a NRAS mutation, of which were G12D in codon 12 and Q61R in codon 61. The rare mutants in NRAS did not yield any significant association with other mutants. Besides, we also found that NRAS and other gene mutations were mutually exclusive in this cohort.

Clinicopathological characteristics of KRAS/BRAF/PIK3CA/NRAS gene mutations.
In the univariate analysis, both KRAS and BRAF mutations were more frequent in older patients (P < 0.05), and female patients shared higher KRAS mutations (codon 12 and codon 61) frequency, meanwhile patients with tumor family history showed more KRAS (G12D) mutations (30.4% vs. 11.8%, P = 0.023) and stage II tumors harbored a higher KRAS codon 13 mutations (10.5% vs. 1.9%, P = 0.013). Moreover, with N stage accelerated, KRAS codon 13 mutations declined (P < 0.05). We found BRAF mutants only existed in non-smokers (P = 0.046) and no other significant differences were discovered when compared to wild type. As for PIK3CA mutations, no significant association appeared between gene mutants and the characteristics, while subgroup mutants such asPIK3CA (exon9) mutations appeared more common in worse differentiation tumors (P = 0.032), and exon 20 mutations did not share this (P = 0.593). Besides, exon 9 (E545K) mutation was significantly associated with tumor recurrence (10.7% vs. 0.0%, P = 0.031) in stage III patients, but not for stage II or combined together. NRAS mutations were too rare to attain significant associations with any clinicopathological characteristics in this study (Table 1).

Univariate analysis of outcomes predictors.
The median follow-up for outcomes analysis in this cohort was 46.0 months, 55.9% (66/118) stage II and 84.9% (79/93) stage III patients had received 5-FU/ oxaliplatin-based standard chemotherapy after surgery, consequent treatment of the other 17 patients were unclear. Only patients who received adjuvant chemotherapy after surgery were recruited in the final analysis and none of them had accepted radiotherapy or anti-EGFR/VEGF MoAbs targeted therapy. No significant difference in DFS was observed between patients with KRAS wild type and mutants (P = 0.727), furthermore subgroups such as codon 12/13 and mutation types (G12D/G12V/G12C/G12S/G13D) did not appear as prognostic factors for DFS, stratified by tumor stage or analyzed together, we also failed to discover significant associations between BRAF/PIK3CA/NRAS mutations and DFS (Fig. 1). However, KRAS codon 61 and PIK3CA (E545K) mutations were prognostic for DFS in stage III alone, but not in stage II or the whole population. KRAS mutation status was not prognostic for OS, no matter analyzed together or separately. Other gene mutations also did not obtain significance with OS ( Fig. 2), while similarly KRAS codon 61 and PIK3CA (E545K) mutants showed prognostic for OS (Table 2). We also discovered that tumor differentiation had significant relationship with DFS and OS, in patients for stage III and combined together (P < 0.05).

Multivariate analysis of outcome predictors.
In the multivariate analysis, we selected sex, age, tumor differentiation, nodal stage, KRAS, BRAF and PIK3CA gene mutations in relation to DFS and OS by stage and in both stages combined together. The KRAS codon 61 mutation still achieved significance after adjustment, which acted as an independent prognostic factor for DFS and OS in stage III alone (P = 0.026 and 0.018 respectively).    Table 3).

Discussion
In this retrospective study, we evaluated common hot spot gene mutations located downstream of EGFR signaling pathway, the potential prognostic value for KRAS, BRAF, PIK3CA and NRAS was assessed in 228 stage II-III colon cancer. Since anti-EGFR MoAbs improved survival in mCRC patients based on specific RAS gene mutations 5,28 , biomarkers such as gene mutations or amplifications has sparked thriving researches for outcomes prognosis. As for stage II-III colon cancer, KRAS and BRAF gene mutations were evaluated frequently in different studies, however, no consistence was observed [10][11][12]29 . Gene mutation profiling alters for different ethnics was confirmed in previous studies 22,30,31 , then potential prognostic value of gene mutations should be reconsidered in Chinese populations. Our data presented here demonstrated that PIK3CA (E545K) mutation was significantly associated with tumor recurrence in stage III, besides, E545K also played as an independent prognostic gene biomarker for adverse outcomes in stage III colon cancer alone. Although KRAS codon 61 appeared significant association with DFS and OS for stage III cancer, while only one patient died with mutant KRAS codon 61 tumor, considering the rare death event in this group, the magnitude of KRAS codon 61 status for prognosis was confused, which should be investigated further in the future. We did not discover any other gene mutant had impact for DFS or OS on stage II-III colon cancer patients treated with 5-FU/oxaliplatin-based standard chemotherapy, KRAS tumor mutation status has no major prognostic value for survival was consistent with a previous analysis in a large cohort, though BRAF mutations achieved significance for OS in their study 11 . Associations between KRAS or BRAF mutation and worse outcomes were reported in North Central Cancer Treatment Group N0147 trial, in which prognostic impact of KRAS and BRAF mutants were involved with stage III colon cancer patients received FOLFOX with or without cetuximab 32 . The adverse effects of KRAS or BRAF mutation in stage II/III colon cancer was not observed in this study, which may be due to the small sample size and consequent lower mutant tumors. However, considering the different treatment strategy in these studies (i.e. FU/FA alone or combined with irinotecan 11 ), comparison of the prognostic value of these biomarkers should be more careful. Previous studies which involved Eastern Asia populations showed KRAS mutation was associated with worse prognosis in stage III or high-risk stage II colon cancer patients 10 , while lack of significant association between BRAF mutation and DFS was in accordance with our results and other clinical trial researches based on western populations 11,12,33 . Different populations and treatment regime may be responsible for the final outcomes, we should also notice that diverse therapies were performed after tumor relapse, the over-all survival of patients would be influenced   We found PIK3CA (E545K) was prognostic for poor OS independently in stage III colon cancers, the adverse effect of PIK3CA in early stage colon cancers were previously reported 34,35 , although other studies presented no associations between PIK3CA mutation and survival 36 . Reasons for the differences may include different target populations, covariates studied in the Cox model, diverse drugs usage and the methodologies for mutations detection. PIK3CA mutation (exon9) significantly associated with KRAS mutant tumors, which was consistent with reported results 16,22,37 , the exon9 mutations were highly dependent on RAS-GTP binding, especially in E542K and E545K 38 , which may suggest that PIK3CA mutations occur after KRAS mutations, in a later step during the tumorigenesis of CRC 39 . E545K also related to the tumor recurrence in stage III cancer patients, all the three E545K mutations appeared only in relapse tumors (10.7% vs. 0.0%, P = 0.031), while E545K was rare in this cohort because of the small sample size, and which consisted a relative small proportion (23.1%, 3/13) in PIK3CA exon9 mutants, then precise conclusions on the prognostic value of E545K should be investigated more in the future. NRAS mutations were only 0.89% (2/224) in this cohort, which was obviously lower than reported results (2.9-4.0%) 13,36 , the different detection method and study population may explain the distinct NRAS mutant frequency. We did not find significant associations between NRAS mutation and patients' survival, which was consistent with previous studies 12,13,36 . In multivariate analysis, tumor differentiation appeared independently prognostic value in colon cancer patients for stage III and combined together, poor differentiation indicated more malignant tumors, and the results were in accordance with a recent study 40 , though the tumor differentiation acted as prognostic for survival only in univariate analysis in their study.
The frequencies of mutations in KRAS, BRAF, PIK3CA and NRAS were compared with previous researches, the KRAS and BRAF mutation frequencies were close to early studies 11,12,18,41 , however, data based on another population presented lower KRAS and BRAFmutants 10 , more sigmoid and rectum tumors involved in their cohort may contribute to the lower mutation frequency. Older patients harbored a higher gene mutation showed aberration accumulation with aging 12 . The PIK3CA mutation (13.18%) was lower than some previous data, which ranged from 12.2 to 20.18% 12,35,36,[42][43][44] , ethnic population, various detection methods and whether rectum cancer involved may explain the difference, the PIK3CA mutation did not significantly associated with clinicopathological characteristics, which was consistent with reported results 35,43,44 . However, subtype of PIK3CA mutation may impact the tumor recurrence and survival, because few E545K mutants in our cohort, the relationship involved in this analysis may be not robust, then further studies or pool analysis of reported researches may focus on this. NRAS mutations were only 0.89% in present study, while another study based on high sensitive detection method showed a 2.9% mutation frequency 12 , and both studies did not discover any correlation between NRAS mutation and clinicopathological characteristics or patients' survival. Considering rare data provided for NRAS mutations in stage II/III colon cancer, large population cohort investigation for this gene mutation detection should be performed, as NRAS closely related to KRAS and NRAS mutation was prognostic for EGFR MoAbs therapy inmCRC 28 . Several limitations were included in present study such as relatively small sample size and limited information of rare gene mutants to support confirmed conclusions. Though KRAS codon 61 appeared as prognostic biomarker in stage III cancer, the rare event may indicate indeterminacy at present, considering the low codon 61 mutation frequency in CRC 22,27 , confirmation of the prognostic value required further investigation with a larger sample size. Additionally, other potentially biomarkers for prognosis were not evaluated, which contained DNA mismatch repair (MMR) and microsatellite instability (MSI) status, or the CpG island methylator phenotype (CIMP). Reported studies provided the relationship between survival and these biomarkers, though the conclusions were still controversial 10,12,18,32,35 . We should strive to obtain more precise and simpler genome  Table 3. Multivariate analysis of outcome predictors.
aberrations for prognosis, even in cancers marked as a complicated multistep disease, more investigations should be implemented to bring us a step closer to personalized medicine in the future.
In conclusion, we conducted this study to describe the gene mutation profiling in stage II/III colon cancer in Eastern Asia populations, PIK3CA (E545K) mutation was independently prognostic for OS in stage III and different subtypes of gene mutants should be further investigated in colon cancers, additional analysis with these molecular prognostic factors may provide a better insight and help us understand the mutation profiling in the evolution of this disease.

Methods
Patients. We totally recruited 124 stage II and 104 stage III colon cancer patients who received complete resection of tumor at the Cancer Hospital of the Chinese Academy of Medical Sciences (Beijing, China) between September 2010 and October 2011. None of them was treated with chemotherapy or radiotherapy previously and each of them was contacted to provide formalin-fixed, paraffin-embedded (FFPE) tumor tissues. Disease-free survival (DFS) was defined as time from operation to the first confirmed relapse, or alive without recurrence at last contact, and time to death or last follow-up from surgery was defined as overall survival (OS). The patients' demographic and clinicopathological data are presented in Table 1. Statistical analysis. Univariate relationships between gene mutation status and clinicopathological characteristics were performed by the Chi-square (χ 2) test. Multivariate relationships were evaluated by fitting logistic regression analysis, using a backward stepwise (likelihood ratio) method and variables which showed statistically significant association with gene mutations were subjected to final regression analysis. The clinical database was last updated in March 2015. DFS and OS were done with the Kaplan-Meier survival function with the method of log-rank test, the hazard ratios were calculated with Cox proportional hazard regression models. All statistical tests were two-sided and the significance level was set at P < 0.05. Data analysis was performed by the SPSS 17.0 statistical software (SPSS, Inc., Chicago, IL, USA).