Effectors of epidermal growth factor receptor pathway: the genetic profiling ofKRAS, BRAF, PIK3CA, NRAS mutations in colorectal cancer characteristics and personalized medicine.

Mutations in KRAS oncogene are recognized biomarkers that predict lack of response to anti- epidermal growth factor receptor (EGFR) antibody therapies. However, some patients with KRAS wild-type tumors still do not respond, so other downstream mutations in BRAF, PIK3CA and NRAS should be investigated. Herein we used direct sequencing to analyze mutation status for 676 patients in KRAS (codons 12, 13 and 61), BRAF (exon 11 and exon 15), PIK3CA (exon 9 and exon 20) and NRAS (codons12, 13 and 61). Clinicopathological characteristics associations were analyzed together with overall survival (OS) of metastatic colorectal cancer patients (mCRC). We found 35.9% (242/674) tumors harbored a KRAS mutation, 6.96% (47/675) harbored a BRAF mutation, 9.9% (62/625) harbored a PIK3CA mutation and 4.19% (26/621) harbored a NRAS mutation. KRAS mutation coexisted with BRAF, PIK3CA and NRAS mutation, PIK3CA exon9 mutation appeared more frequently in KRAS mutant tumors (P = 0.027) while NRAS mutation almost existed in KRAS wild-types (P<0.001). Female patients and older group harbored a higher KRAS mutation (P = 0.018 and P = 0.031, respectively); BRAF (V600E) mutation showed a higher frequency in colon cancer and poor differentiation tumors (P = 0.020 and P = 0.030, respectively); proximal tumors appeared a higher PIK3CA mutation (P<0.001) and distant metastatic tumors shared a higher NRAS mutation (P = 0.010). However, in this study no significant result was found between OS and gene mutation in mCRC group. To our knowledge, the first large-scale retrospective study on comprehensive genetic profile which associated with anti-EGFR MoAbs treatment selection in East Asian CRC population, appeared a specific genotype distribution picture, and the results provided a better understanding between clinicopathological characteristics and gene mutations in CRC patients.

Introduction

Colorectal cancer (CRC) still causes majority of mortality in the world [1]. In mCRC tumors, exceedingly poor prognosis was observed. Fortunately, the rapid development in biological agents appears a promising future in treatment. Cetuximab or panitumumab, the monoclonal antibody (MoAb) targeted on epidermal growth factor receptor (EGFR), has been implemented in clinical practice, and emerged as an effective single agent or chemotherapy adjuvant approach for mCRC treatment [2]. These MoAbs blocks the downstream intracellular signaling of EGFR, which includes two main signaling pathways, RAS-RAF-MAPK axis, which mainly involved in cell proliferation, and the phosphatidylinositol 3-kinase (PI3K)-PTEN-AKT, key mediators of survival, and motility-invasion [3].

Although previous clinical trials have indicated that patients who carry KRAS mutations in codons 12 and 13 are non-responsive to the EGFR-targeted therapy [4], [5], [6], [7], and the wild-type status seems a response condition, some wild-type patients still fail to respond to anti-EGFR monoclonal antibody therapy [8], and the mechanism remains unclear. It is possible that mutations in the downstream effectors of the EGFR signaling pathway, such as BRAF, PIK3CA and NRAS, may induce a negative effect on the response in anti-EGFR targeted treatment [9], [10], [11].

To date, genetic profiling of individual tumors affect the selection of therapy and treatment response have been proven in clinical practice, however, major data about the frequency of oncogenes mutations were presented in Western populations and few data are available for the Chinese. Since gene mutation status alters with ethnic differences [12], we design this study to investigate the ethnicity-specific role of mutations in development and progression of CRC. KRAS, BRAF, PIK3CA and NRAS mutations in primary tumors from Chinese CRC patients were detected and their potential correlations with clinicopathological factors were analyzed. Furthermore, we collected the survival data of mCRC subgroup patients, in order to obtain an appropriate insight between gene mutation and survival status. We intended that these data could benefit the design of future clinical trials and individualized therapy in CRC patients.

Materials and Methods

Patients

We investigated 676 consecutive patients who underwent surgery for colorectal cancer at the Cancer Institute/Hospital of the Chinese Academy of Medical Sciences (Beijing, China) between August 2010 and December 2011, all the patients were carried out primary resection in our hospital, and no patient had received chemotherapy before surgery. Each patient was contacted to provide available formalin-fixed, paraffin-embedded (FFPE) CRC tissues. Written informed consent was obtained from individual patients, and the experimental protocol was approved by the Institutional Review Board (IRB) in Cancer Institute/Hospital of Chinese Academy of Medical Sciences and Peking Union Medical College. CRC diagnosis was confirmed by hematoxylin and eosin (HE) staining and histological analysis. Overall survival was defined as the period from the start of diagnosed CRC until death from any cause or last follow-up. The patients’ demographic and clinicopathological data are presented in Table 1.

Table 1

Characteristics of 676 CRC patients and association of gene mutations with clinicopathological parameters.

Characteristics	Number (%)	KRAS		BRAF		PI3KCA		NRAS
		Mutations (%)	P	Mutations (%)	P	Mutations (%)	P	Mutations (%)	P
Sex
Male	407 (60.2)	131 (32.3)	0.02	28 (6.9)	0.93	34 (9.0)	0.35	16 (4.3)	0.93
Female	269 (39.8)	111 (41.3)		19 (7.1)		28 (11.3)		10 (4.0)
Age
≤60	342 (50.6)	109 (32.0)	0.03	23 (6.7)	0.81	25 (8.0)	0.11	15 (4.9)	0.38
>60	334 (49.4)	133 (39.9)		24 (7.2)		37 (11.9)		11 (3.5)
Mean	60±11
Range	23–86
Primary tumor site
Rectum	391 (57.8)	138 (35.4)	0.74	27 (6.9)	0.96	22 (6.1)	<0.001	19 (5.4)	0.09
Colon†	285 (42.2)	104 (36.6)		20 (7.0)		40 (15.1)		7 (2.6)
Tumor location*
Proximal	133 (19.7)	52 (39.1)	0.38	9 (6.8)	0.92	25 (19.8)	<0.001	4 (3.1)	0.51
Distal	542 (80.2)	189 (35.0)		38 (7.0)		37 (7.4)		22 (4.5)
Missing data	1 (0.1)
Tumor differentiation
Well	31 (4.6)	17 (54.8)	0.06	1 (3.2)	0.18	4 (14.3)	0.21	2 (6.9)	0.07
Moderate	556 (82.2)	197 (35.6)		36 (6.5)		46 (9.0)		17 (3.4)
Poor	87 (12.9)	27 (31.0)		10 (11.5)		12 (14.5)		7 (8.4)
Missing data	2 (0.3)
Tumor stage↑
I	92 (13.6)	33 (35.9)	0.56	5 (5.4)	0.61	8 (9.9)	0.29	4 (4.9)	0.03
II	238 (35.2)	86 (36.1)		17 (7.1)		20 (9.0)		8 (3.6)
III	288 (42.6)	107 (37.4)		19 (6.6)		25 (9.3)		7 (2.6)
IV	55 (8.1)	15 (27.3)		6 (10.9)		9 (18.0)		6 (12.2)
Missing data	3 (0.5)
Depth of invasion↑
T1	18 (2.7)	5 (27.8)	0.21	1 (5.6)	0.57	0 (0.0)	0.34	1 (6.3)	0.69
T2	105 (15.4)	37 (35.6)		4 (3.8)		9 (9.7)		4 (4.2)
T3	521 (77.1)	184 (35.4)		40 (7.7)		48 (9.9)		20 (4.2)
T4	30 (4.5)	16 (53.3)		2 (6.7)		5 (17.2)		0 (0.0)
Missing data	2 (0.3)
Lymph node↑
N0	342 (50.8)	123 (36.0)	0.88	23 (6.7)	0.48	30 (9.5)	0.75	12 (3.8)	0.89
N1	190 (28.0)	66 (34.9)		11 (5.8)		16 (9.4)		8 (4.7)
N2	142 (20.9)	53 (37.6)		13 (9.2)		16 (11.7)		5 (3.8)
Missing data	3 (0.3)
Distant metastasis
Yes	55 (8.1)	15 (27.3)	0.16	6 (10.9)	0.26	9 (18.0)	0.08	6 (12.2)	0.01
No	619 (91.6)	227 (36.8)		41 (6.6)		53 (9.2)		19 (3.3)
Missing data	2 (0.3)

Colon *: Colon is defined as right colon, transverse colon, left colon, sigmoid colon, rectosigmoid transition zone.

Tumor location *: Proximal tumor is defined as right colon and transverse colon; distal tumor is defined as left colon, sigmoid colon, rectosigmoid transition zone and rectum.

†: Seventh edition of the AJCC/UICC TNM staging systems.

DNA Extraction and Mutation Analysis

Before the extraction of genomic DNA, all CRC samples were identified by two pathologists in order to ensure the representative malignant cells exist in each sample, the tissue blocks were cut into 5 µm sections, then microdissection was performed to guarantee each tissue sample tested contained >80% cancer cells. DNA was extracted by the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions and stored at −80°C until use.

We detected the mutation hotspots in KRAS (codons 12 and 13), BRAF (exon15), PIK3CA (exon 9 and exon 20) and NRAS (codon 61), where the most mutations occur in these genes [13], [14], besides, rare types of mutants for KRAS (codon 61), BRAF (exon 11) and NRAS (codons 12 and 13) were also included. The program for the PCR amplification in KRAS, BRAF, NRAS and PIK3CA exon20 was as follows: 1 min of initial denaturation at 95°C, 35 cycles of amplification consisting of 30 s at 94°C, 40 s at 57°C, and 30 s at 72°C, with a final additional elongation at 72°C for 7 min. PIK3CA exon 9 amplification was carried out with a touchdown PCR program: 94°C (2 min); 3 cycles of 94°C (30 sec), 64°C (30 sec), 70°C (30 sec); 3 cycles of 94°C (30 sec), 61°C (30 sec), 70°C (30 sec); 3 cycles of 94°C (30 sec), 58°C (30 sec), 70°C (30 sec); 32 cycles of 94°C for (30 sec), 57°C (30 sec), 70°C (40 sec); 1 cycle of 70°C (5 min). When performing the PCR, a non-template control was included in each batch. After PCR reaction, the products were purified and subjected to direct sequencing (ABI 3500×L Genetic Analyzer; Applied Biosystems, Carlsbad, CA, USA).

Statistical Analysis

Statistical analysis was carried out by the SPSS 17.0 statistical software (SPSS, Inc., Chicago, IL, USA). The Chi-square (χ2) test was used to compare the proportion of gene mutations among groups with different clinicopathologic factors. Multiple logistic regression analysis was done to investigate the effects of covariates on gene mutations, using a backward stepwise (likelihood ratio) method with odds ratio (OR) calculated, and variables which showed statistically significant association with gene mutations were subjected to final regression analysis. Survival analysis was done with the Kaplan-Meier survival function with the method of log-rank test. The two-sided significance level was set at P<0.05.

Results

KRAS Mutation

KRAS mutation status could not be assigned to 2 of 676 (0.30%) samples, 35.9% (242/674) harbored a KRAS mutation, 25.7% (173/674) in codon12, 6.8% (46/674) in codon13, and 2.1% (14/672) in codon61. Moreover, one patient harbored a double KRAS mutation in both codon12 and 13 (GGT>GTT, GGC>AGC). The corresponding order for KRAS codon12 mutation frequency was G12D, G12V, G12A, G12C, G12S and G12R; in KRAS codon13, the most frequent mutation was G13D, followed by G13C and G13S. The major mutation subtype in codon61 was Q61H, and Q61L, Q61R were also found in this study (Figure 1). KRAS mutation appeared more frequently in female than male (41.3% vs 32.3%, P = 0.02), and patients older than 60 years also showed a higher rate of KRAS mutation (39.9% vs 32.0%, P = 0.03). We did not find other significant associations between KRAS mutation and patients’ clinicopathological characteristics (Table 1).

Figure 1

Frequency of the various KRAS and BRAFmutations.

Panel A: KRAS mutations (codons12 & 13: n = 674; codon61: n = 672). Panel B: BRAF mutations (exon11: n = 676; exon15: n = 675). The data are presented as percentages (number of total samples).

BRAF Mutation

The status of BRAF mutation was detected in 99.8% (675/676) samples, 6.96% (47/675) harbored a BRAF mutation, 4.4% (30/675) in exon15 and 2.5% (17/676) in exon11. The V600E mutation in exon15 was the most frequent subtype (1.8%,12/675), and followed by V600M mutation and other types. In exon11, the mutations distributed widely, R461K and G465E were relatively more common in these mutations (Figure 1). BRAF and KRAS mutations were not mutually exclusive, 4.55% (11/242) of KRAS mutant tumors harbored a BRAF mutation (of which 7/242[2.89%] exon15 and 4/242[1.66%] exon 11 mutations). However, BRAF (V600E) only existed in KRAS wild types (0.0% vs 2.78% [12/431], P = 0.005). In this group, V600E mutation showed a strong association with primary tumor site, tumor in colon appeared more frequently to harbor a V600E mutation (9/285[3.2%] in colon vs 3/390[0.8%] in rectum; P = 0.020), besides, with the tumor differentiation getting worse, a higher V600E mutation rate emerged (5/87[5.7%] in poor differentiation vs 7/555[1.3%] in moderate differentiation; P = 0.030). No other significant association was found between BRAF mutation and patients’ characteristics (Table 1).

PIK3CA Mutation

PIK3CA mutation status could not be assigned to 7.54% (51/676) samples, 9.9% (62/625) harbored a PIK3CA mutation, 7.0% (45/643) in exon9 and 2.67% (17/636) in exon20. The E545K in exon9 appeared more frequently than any other mutation subtype, followed by E542K, E545G, Q546E and others. By contrast, nearly all mutations in exon20 was H1047R, only one sample was H1047Y mutation, the spectrum of these mutations was showed in Figure 2. We also detected one sample harbored a double mutation in exon9 (L540V and Q546E), besides, this sample also had a KRAS mutation (G13D). There was a strong significant association between PIK3CA exon9 and KRAS mutations (23/230[10.0%] in KRAS mutant vs 22/412[5.3%] in KRAS wild types, P = 0.027), whereas this association was not found in PIK3CA exon20 with KRAS mutation (P = 0.673). BRAF and PIK3CA mutations were not mutually exclusive, 10% (4/40) of BRAF mutation coexists with PIK3CA mutation (of which 3/40[7.5%] exon 9 and 1/40[2.5%] exon20). For the clinicopathological characteristics analysis, patients with tumor located in rectum had a significantly lower PIK3CA mutation rate than other sites in colon and rectosigmoid transition zone (6.1% vs 15.1%, P<0.001) and proximal tumors appeared a higher PIK3CA mutation rate (19.8% vs 7.4%, P<0.001). No other significant association was found in this analysis (Table 1).

Figure 2

Frequency of the various PIK3CA and NRAS mutations.

Panel A: PIK3CA mutations (exon9: n = 643; exon20: n = 636). Panel B: NRAS mutations (codons12 & 13: n = 630; codon61: n = 643). The data are presented as percentages (number of total samples).

NRAS Mutation

We detected NRAS mutation in 92.0% (621/676) samples, and 4.19% (26/621) harbored a NRAS mutation. Although NRAS is closely to KRAS which also included in Ras gene [13], unlike KRAS, most NRAS mutation occurred in codon61 (2.02%, 13/643), rather than in codon12 or 13 (1.75%, 11/630). The most frequently mutation subtype in codon61 was Q61R, and G12D in codon12/13 (Figure 2). Besides, one G15W and one G60E mutation were also detected in these samples. Moreover, we still found that NRAS mutation appeared a strong significant association with KRAS wild types (1/227[0.44%] in KRAS mutant vs 25/394[6.3%] in KRAS wild types, P<0.001). Interestingly, NRAS codon61 mutation only harbored in KRAS wild types (0.0% vs 3.2% [13/410], P = 0.006), whereas NRAS codon12 and 13 did not share this association (P = 0.063). Only one sample harbored a BRAF mutation (V600E) with a NRAS mutation (G15W), and 6.78% (4/59) PIK3CA mutation harbored a NRAS mutation (of which 2/59[3.39%] in codons 12 and 13, 2/61[3.28%] in codon61). Furthermore, NRAS mutation occurred more frequently in distant metastasis tumors (12.2% vs 3.3%, P = 0.010), and different tumor stage showed a different NRAS mutation rate (P = 0.030). (Table 1).

In the multivariate logistic regression analysis, we selected sex, age, primary tumor site, tumor differentiation, tumor stage and distant metastasis as covariates, and KRAS mutants appeared more frequently in patients older than 60 (P = 0.023), as well in female patients (P = 0.016). BRAF mutations did not show any significant association with characteristics (data not shown), while BRAF (V600E) mutants shared significant association with tumor differentiation (P = 0.016). As for PIK3CA mutations, colon cancer appeared a higher mutation rate than rectum cancer (P<0.001), however, NRAS mutations showed more frequently in rectum cancer (P = 0.031), although no significant association was found in univariate analysis (P = 0.09). Moreover, a strong significant association still existed between NRAS mutants and distant metastasis in the multivariate analysis (Table 2).

Table 2

Multivariate logistic regression in CRC patients between gene mutations and clinicopathological characteristics.

Characteristics	KRAS			BRAF (V600E)
	Adjusted odds ratio (95% CI)	LRT p value		Adjusted odds ratio (95% CI)	LRT p value
Sex	0.671(0.485–0.928)		0.016	0.627(0.197–1.994)	0.429
Age	1.450(1.052–1.999)		0.023	1.241(0.381–4.040)	0.720
Primary tumor site	1.066(0.768–1.480)		0.702	3.587(0.947–13.588)	0.060
Tumor differentiation	0.675(0.453–1.006)		0.053	4.101(1.298–12.957)	0.016
Tumor stage	1.098(0.864–1.396)		0.443	1.310(0.601–2.855)	0.496
Distant metastasis	0.681(0.362–1.280)		0.232	0.637(0.069–5.891)	0.691
Characteristics	PIK3CA			NRAS
	Adjusted odds ratio (95% CI)	LRT p value		Adjusted odds ratio (95% CI)	LRT p value
Sex	0.766(0.447–1.315)	0.334		1.394(0.587–3.311)	0.452
Age	1.491(0.869–2.559)	0.147		0.700(0.305–1.609)	0.401
Primary tumor site	2.773(1.604–4.792)	<0.001		0.348(0.134–0.907)	0.031
Tumor differentiation	1.025(0.534–1.970)	0.940		1.587(0.609–4.138)	0.345
Tumor stage	0.939(0.616–1.432)	0.771		0.708(0.384–1.305)	0.268
Distant metastasis	1.802(0.814–3.989)	0.146		4.930(1.817–13.375)	0.002

LRT: likelihood ratio test; 95% CI: 95% confidence interval.

Analysis of Gene Mutation in mCRC Patients

Fifty-five of 676 patients were confirmed as mCRC, and all these 55 samples were collected before chemotherapy. We further investigated the mutants distribution and clinicopathological characteristics association in this group. 27.3% (15/55) harbored a KRAS mutation, of which 93.3% (14/15) in codon12 and 6.7% (1/15) in codon13, respectively. The BRAF mutation rate was 10.9% (6/55), 66.7% (4/6) in exon15 and 33.3% (2/6) in exon11. PIK3CA mutation was detected in 18.0% (9/50) tumors, 66.7% (6/9) was detected in exon9 and 33.3% (3/9) in exon20. 12.24% (6/49) tumors were detected as NRAS mutants, of which 50% (3/6) in codons12 and 13, 33.3% (2/6) in codon61, besides, one sample harbored a G15W mutation. Statistical analysis indicated that KRAS mutation was significantly higher in the deeper invasion stage (5/7[71.4%] in T4 vs 10/48[20.8%] in T3; OR 9.500, 95% CI 1.599–56.426; P = 0.013), and tumor with poor differentiation showed a higher NRAS mutation rate than moderate differentiation (5/19[26.3%] vs 1/30[3.3%]; OR 10.357, 95% CI 1.103–97.266; P = 0.027). We did not find any other significant association between gene mutation (included subgroup analysis) and clinicopathological characteristics (data not shown).

Overall Survival Analysis in mCRC Patients

Overall survival of patients in this subgroup was analyzed with the Kaplan-Meier method, in this relative small subgroup (n = 55), survival information was collected successfully in only 45 patients, of whom 37 had received chemotherapy after surgery, either with infusional fluorouracil, leucovorin and irinotecan (FOLFIRI) or infusional fluorouracil, leucovorin, and oxaliplatin (FOLFOX4). However, the relative small sample size did not present any significant result between gene mutation and OS, including gene subsets analysis (data not shown).

Discussion

During the past decades, large amounts of research data emerged from molecular basis [15], drug investigation and usage [3], genetic profiling effects [16] studies have led to the thriving research on identification of multiple molecular subsets and targeted therapy in colorectal cancer. Following the discovery that mutant KRAS tumors were resistant to anti-EGFR antibodies, patients with metastatic colorectal cancer are now recommended to detect the KRAS codons12 and 13 mutation status before MoAbs therapy [8], [17], [18]. However, even in KRAS wild-type tumors, up to 65% patients were still resistant to anti-EGFR monoclonal antibodies [8]. Besides, although the detection of KRAS mutation status before MoAbs therapy is widely accepted, there is little agreement on its predicted and prognostic role, for published studies provided different results in the relationship between KRAS mutation and clinical outcomes in CRC, and the main effectors in downstream signaling pathway of KRAS, such as BRAF, PIK3CA and NRAS were already studied in many clinical trials, which showed the capability to present as potential predictive or prognostic biomarkers [14], [19].

To our knowledge, this study investigated the first time gene mutation type distribution in Chinese CRC population, and involved not only KRAS, but also BRAF, PIK3CA, NRAS together for comprehensive analysis between gene mutation and clinicopathological characteristics, in addition, the overall survival of metastatic colorectal cancer. Previous studies usually focused on KRAS, BRAF, PIK3CA [20], [21], [22], but not include NRAS mutation, or study sample size was too small to draw confirmed conclusions [20]. Many studies could not collect enough appropriate samples to describe a relative complete outline for Chinese CRC patients in genetic profile, and our investigation aimed to present the key mediated gene mutation of CRC, to some extent, representing the East Asian population.

KRAS gene encodes a small G protein which acts as a key transducer in EGFR pathway, mutations in KRAS gene lead to constitutive signaling through the EGFR pathway and active downstream MAPK and PIK3CA dependent pathways [18], [23]. Previous studies have analyzed KRAS mutation distribution from western population, which indicated that G12D was the most frequent mutation subtype in codon12, followed by G12V, G12C, G12S, G12A and G12R [24]. However, in present study, the corresponding order for KRAS codon12 mutation frequency was G12D, G12V, G12A, G12C, G12S and G12R. As for codon13, the difference remained in subtype distribution (G13/D/C/R in western population vs G13/D/C/S in this study). In addition to gaining more information and expanding the recognition of the KRAS mutation, the sample size of our series allowed us to investigated the rare codon61 mutation, since mutant tumors with KRAS codon61 led to significantly lower response rate than wild types (0.0% vs 35.7%, P = 0.0055) [14], while the mutation incidence (2.1%) was even higher than some codon12 and 13 mutations, we then suggested that codon61 detection should be taken into consideration during clinical practice. This study showed a 35.9% KRAS mutation rate, which was similar to previous studies [4], [9], [14], [22], and patients older than 60 appeared more frequently to harbor a KRAS mutation. Meanwhile, in mCRC patients, KRAS mutation was significantly higher in the deeper invasion stage (OR 9.500, 95% CI 1.599–56.426; P = 0.013), which was consistent with Li HT and colleagues [25] reported that KRAS mutation had a strong association with Dukes’ staging, with the highest mutation rate in Dukes’ D staging tumors. Our molecular data provided an evaluation of possibility for disease progression. Then the patients with rapid distant metastasis seemed more likely to be initial resistant to anti-EGFR MoAbs, because KRAS mutation maintained throughout the CRC development, progression and metastasis, with a high (95%) concordance presenting at the primary and related metastatic sites [26]–[27].

BRAF gene is a member of the RAF gene family, which encodes a serine-threonine protein kinase, acts as a downstream effector of activated KRAS. Previous studies reported that KRAS and BRAF mutations were mutually exclusive in mCRC, and BRAF mutation occurring in approximately 5%–10% tumors [14], [28], [29]. However, in this study, we found that KRAS and BRAF mutations were not mutually exclusive, 4.55% (11/242) of KRAS mutant tumors harbored a BRAF mutation, our data were supported by Mao C and colleagues [20], although they gained an extremely high BRAF mutation incidence (25.4%,15/59), for reported studies of BRAF mutations usually presented a higher mutation frequency in western population (8.5%–13.9%) [29]–[30] than the Chinese (1.1%–7.0%) [25], [31]. In our study, BRAF mutation (6.96%) was consistent with previous results, and the lower BRAF mutation frequency may attribute to the patients population studied. However, difference in mutation frequency also indicates that geographical and ethnical variations play a role in gene mutation distribution. In the reported studies, most of the data were collected from only BRAF V600E mutation type, other types such as V600M were not included [14], [29], while we also confirmed this point for BRAF (V600E) only existed in KRAS wild types (0.0% vs 2.78% [12/431], P = 0.005). However, BRAF mutation associated with poor clinical outcomes were proven in several studies [10], [14], [29], herein we reported a BRAF mutation rate for 4.4% (30/675) in exon15 and 2.5% (17/676) in exon11, which were both higher than KRAS codon61 (2.1%) mutation. As De Roock W and colleagues [14] recommended BRAF should be tested subsequently after KRAS, we supposed both BRAF exon15and 11 need to be taken into consideration, in order to select better suitable subgroup patients. In addition, V600E mutation was significantly higher in colon cancer than rectum cancer (OR 4.035, 95% CI 1.062–15.330; P = 0.041) and poor differentiation tumor harbored a higher V600E mutation (P = 0.030). These data indicated that colorectal cancer treatment should be regarded from a deeper extent, for colon and rectum cancer required different therapy in different stage.

We confirmed the association between KRAS and PIK3CA mutations in CRC, which was comparable with previous studies [14], [20], [25], [32], and only exon9 (not included exon20) shared a strong association with KRAS mutation [14]. This was consistent with the findings that the gain of function by exon9 mutations (the helical domain) was highly dependent on RAS-GTP binding, especially in E542K and E545K, while exon20 mutations (the kinase domain) active was likely in the absence of RAS-GTP binding [33]. The PIK3CA mutation frequency varies between 13.6%–18.0% in western population [34]–[35], while we reported a relative low mutation frequency (9.9%). Studied population may lead to this difference mainly, for other studies which based on Chinese population also showed lower mutation frequency (4.9%–8.2%) [20], [36]. In the logistic regression analysis, PIK3CA mutation appeared more frequently in colon cancer than rectum at the same time, which was supported by a recent study [37]. Previous studies indicated that PIK3CA mutation existence implied negative prognosis, either a shorter median progression-free survival (PFS) [38], or a shorter median OS [39]–[40]. However, since the PIK3CA mutation effect seemed to be considered together, the separate effect of each subtype appeared unclear, for several studies had showed that exon9 and exon20 mutation led to different results [14], [41]. The large European consortium study indicated than only exon20 mutation was associated with worse clinical outcome [14], and was supported by other research data [16], [42]. But because exon20 mutation was relatively low compared with exon9 (2.96% vs 9.96%) [14], and in our study (2.67% vs 7.0%). The reported data should be regarded as clinical related hypothesis and required confirmation, based on further genetic profiling and clinical trials investigation.

The RAS gene (KRAS, NRAS, HRAS) encodes a series of GTP/GDP related switches that convey extracellular signals, resulting in regulating growth and survival of cells [43]. As one of the RAS family, NRAS shared close relations with KRAS [13], while unlike KRAS mutation occupies such a large percentage in colorectal cancer, NRAS mutations were rare. Irahara N and colleagues [44] reported a 2.2% (5/225) mutation incidence, and 2.64% (17/644) mutation rate in another study [14], while we detected 4.19% (26/621) tumors harbored a NRAS mutation. The higher NRAS mutation incidence presented a specific characteristic for Chinese population. As rare data was reported in Chinese patients for NRAS mutation status, our study may provide some original contribution. However, future investigations are needed to draw a better picture in this area. NRAS mutations were not mutually exclusive with BRAF and PIK3CA mutation, although another study did not share this [44]. NRAS mutation coexisted with KRAS wild-type (P<0.001), of note, codon61 mutation only appeared in KRAS wild-type tumors (P = 0.006), and codon12 and 13 had a significantly higher mutation rate in distant metastatic tumors (P = 0.016). These data can partially help explain the anti-EGFR MoAbs resistance in KRAS wild-type patients, as NRAS mutations were significantly associated with lower disease control rate and response rate to MoAbs [14], [45], and we recommended NRAS mutation detection should be taken into consideration before MoAbs treatment, especially in KRAS wild-type tumors. However, considering the low mutation incidence, the magnitude of NRAS mutation effect was still confused, larger sample size or preselected patients investigation seemed essential in future design.

There were several limitations in this retrospective study, including the relatively small number (n = 45) of patients in the survival analysis, then the limited information could not support confirmed conclusions in present study. Additionally, other potentially negative factors such as loss of expression of phosphatase and tensin homologue (PTEN) should be involved, thus essential effects of these biomarkers in clinical practice stayed further validation. Moreover, as epigenetic status or microsatellite instability (MSI) plays a significant role in CRC tumors, these features should be involved into analysis. Besides, gene expression in the key effectors, different tumor locations may provide information for better understanding in CRC, either in carcinogenesis or tumor progression and these should be taken into consideration in future studies. As high throughput detecting method has been implemented in screening gene variants or sequencing, different types of gene alternations have been investigated comprehensively in colorectal cancer [46], these studies have provided potential genes which need further investigations.

In a recent randomised trial [47], patients were preselected for only KRAS codon12, 13 and 61 wild-type tumors, however, therapy with panitumumab to irinotecan did not improve the overall survival compared with irinotecan alone, then refinement of molecular selection was required considering patients’ welfare. Another multicenter randomised placebo-controlled trail tested a novel multikinase inhibitor (Regorafenib) [48], although the study obtained a significant result in prolonging median OS (6.4 vs 5.0 months, hazard ratio 0·77; 95% CI 0·64–0·94; one-sided p = 0·0052), in view of the small incremental survival benefit, potentially exposed to toxic effects and heavy economic burden, the new agent seemed not to be a cost-effective option, while selecting the subset of patients who would really benefit from Regorafenib based on the identification of biomarkers was a high priority. We have already known that genotype subgroup would lead to different clinical outcomes in mCRC MoAbs treatment [14], [49], [50], all the data indicated that more precise classification of genetic profile should be implemented to enhance the clinical targeted therapy, then our study here was in order to bring us a step closer to personalized medicine.

In conclusion, this study presented a clear genotype distribution picture scroll in East Asian CRC population, involving potential molecular predictors KRAS, BRAF, PIK3CA, NRAS, which showed a specific characteristic. However, prospective randomised trials are needed to provide proposals and validate conclusions. More comprehensive genomic analysis and molecular classification should be performed, to recognize the genetic profile better and to improve the clinical choice smarter.