Upregulation of c-mesenchymal epithelial transition expression and RAS mutations are associated with late lung metastasis and poor prognosis in colorectal carcinoma.

The present study aimed to investigate whether c-mesenchymal epithelial transition factor (C-MET) overexpression combined with RAS (including KRAS, NRAS and HRAS) or BRAF mutations were associated with late distant metastases and the prognosis of patients with colorectal cancer (CRC). A total of 374 patients with stage III CRC were classified into 4 groups based on RAS/BRAF and C-MET status for comprehensive analysis. Mutations in RAS/BRAF were determined using Sanger sequencing and C-MET expression was examined using immunohistochemistry. The associations between RAS/BRAF mutations in combination with C-MET overexpression and clinicopathological variables including survival were evaluated. In addition, their predictive value for late distant metastases were statistically analyzed via logistic regression and receiver operating characteristic analysis. Among 374 patients, mutations in KRAS, NRAS, HRAS, BRAF and C-MET overexpression were observed in 43.9, 2.4, 0.3, 5.9 and 71.9% of cases, respectively. Considering RAS/BRAF mutations and C-MET overexpression, vascular invasion (P=0.001), high carcino-embryonic antigen level (P=0.031) and late distant metastases (P<0.001) were more likely to occur in patients of group 4. Furthermore, survival analyses revealed RAS/BRAF mutations may have a more powerful impact on survival than C-MET overexpression, although they were both predictive factors for adverse prognosis. Further logistic regression suggested that RAS/BRAF mutations and C-MET overexpression may predict late distant metastases. In conclusion, RAS/BRAF mutations and C-MET overexpression may serve as predictive indicators for metastatic behavior and poor prognosis of CRC.

Introduction

Colorectal cancer (CRC) is the third most commonly diagnosed malignancy and the fourth most frequent cause of cancer-associated mortality worldwide (1). It has recently been indicated that late distant metastases are common in CRC, particularly liver and lung metastases, which accounted for ~40% of all advanced patients (2). Although notable advances have been made in comprehensive therapy, the prognosis of metastatic CRC remains unfavorable (3). As the understanding of molecular mechanisms underlying tumorigenesis and progression of CRC develops, targeted therapy has already become a popular alternative to other, currently used treatments, representing a significant landmark in devising individualized treatment regimens.

It is known that epidermal growth factor receptor (EGFR) is an important molecular target in metastatic CRC (mCRC) (4). Furthermore, the success of cetuximab or panitumumab, agents that target EGFR, created a new milestone in precision medicine for mCRC (5). However, mutations of RAS genes (including KRAS, NRAS and HRAS) or BRAF may induce constitutive activation of downstream signaling pathways, independent of EGFR inhibition, which is associated with tumor proliferation and diffusion. Recent data (4) has demonstrated that KRAS exons 2, 3 and 4; NRAS exons 2 and 3; HRAS exon 2; and BRAF exon 15 occurs in ~50% of CRC patients, and exhibits facilitated neoplastic transformation in vitro of colorectal cells as well as resistance to anti-EGFR therapy (6). Therefore, screening of gene mutation profiling is important for appropriate therapeutic options and regular surveillance. Notably, the predictive and prognostic significance of RAS/BRAF mutations in CRC remains controversial. A recent retrospective study (7) indicated that distant metastasis was more likely to occur in patients with KRAS or BRAF mutation. In addition, Morris et al (8), previously demonstrated a trend toward lung metastasis and low survival for RAS/BRAF-mutant CRC. Conversely, certain studies have not demonstrated that mutations in RAS/BRAF were independent prognostic factors for CRC (9,10). Therefore, the association of RAS/BRAF status with late distant metastases and prognosis of CRC requires further investigation.

The c-mesenchymal epithelial transition factor (C-MET), a tyrosine kinase receptor for hepatocyte growth factor, is associated with diverse biological functions ranging from embryogenesis to wound healing (11). However, aberrant C-MET expression is closely correlated with tumor progression and metastasis via regulating cell proliferation, scattering and apoptosis (12). It is well known that C-MET gene is upregulated in a variety of human malignancies, including CRC (11). Recently, Lorenzon et al (13), reported that in KRAS wild-type patients with CRC, high C-MET expression appeared as a negative predictor for disease-specific survival and may interfere with anti-EGFR strategies, although the patient cohort analyzed in the research was small.

Currently, use of a combination of biomarkers as a better predictor of metastasis and prognosis in patients with CRC has attracted more attention due to the potential of identifying distinct tumor subtypes bearing different prognoses. However, the clinicopathological relevance of RAS/BRAF mutations combined with high C-MET expression in CRC is yet to be fully elucidated. The majority of studies focused on western populations (8,11–13) and, with few deriving data from Chinese patients (10). To improve the current knowledge, the present study comprehensively characterized RAS/BRAF mutations and C-MET overexpression in stage III CRC, alone and in combination, to provide an insight into the association between gene abnormalities and patient survival in Chinese populations.

Materials and methods

Patients and follow-up

The observational model was developed in 374 stage III CRC samples (204 males and 170 females; age range, 23–92 years old) and corresponding non-cancerous tissues from patients who had undergone surgical resection at the department of gastrointestinal surgery of Guangdong General Hospital (Guangzhou, China) between January 2010 and October 2015. The inclusion criteria were as follows: All patients had to have undergone complete lesion removal, without having received any prior anticancer therapy. Patients were also required to have normal renal and hepatic function test results. Patients were excluded from the present study if they exhibited inflammatory bowel disease. All patients were classified into 4 groups: Group 1, RAS/BRAF-wild without C-MET overexpression; group 2, RAS/BRAF-wild with C-MET overexpression; group 3, RAS/BRAF-mutant without C-MET overexpression; and group 4, RAS/BRAF-mutant with C-MET overexpression. Genetic testing was performed as a part of integrated care and information on clinicopathological data were obtained from medical archives. Tumor grading was based on the American Joint Committee on Cancer TNM classification and pathological classification was in line with the World Health Organization criteria (14,15). Overall survival (OS) or disease-free survival (DFS) was calculated from the surgery of the primary CRC until death/censoring or local recurrence/late distant metastasis/censoring, respectively. Late distant metastasis was defined as metastasis that occurred during follow-up. Of the 374 participants, 272 (72.7%) received 5-fluorouracil (5-FU)-based postoperative adjuvant chemotherapy. An outpatient follow-up was conducted every 3 months in accordance with Response Evaluation Criteria in Solid Tumors 1.1 (16) during the initial 2 years following clinical treatments and subsequently every 6 months, until the end of a 3 year follow-up or mortality. Written, informed consent was obtained from all individual participants and the protocol was approved by the Ethics Committee of Guangdong General Hospital.

Tissue sampling and mutation assessment

Comprehensive genomic profiling was analyzed in 374 resected CRC tissue samples, which were fixed with 10% formalin overnight at room temperature and embedded in paraffin wax. Tissues were then sliced longitudinally to a thickness of 4 µm. Genomic DNA was isolated from each FFPE specimen using a QIAamp DNA FFPE Tissue Kit 56404 (Qiagen GmbH, Hilden, Germany) according to the manufacturer's protocol. In addition, cancer cell-rich regions were identified prior to sample DNA isolation via application of hematoxylin and eosin (HE) staining to ascertain that all cases exhibited enrichment of ≥70% malignant cells. HE staining was performed according to manufacturers' instructions. Following washing with xylene and dehydration with ethanol, the sections were rehydrated in distilled water and then stained with the alum haematoxylin (Shanghai XIBAO Biology Co., Ltd., Shanghai, China) for 13 min at room temperature. After rinsing under running tap water, slides were differentiated with 0.3% acid alcohol for 5 min and washed in running tap water for 10 sec. Next, the tissue sections were stained with eosin (Shanghai XIBAO Biology Co., Ltd.) for 1 min at room temperature, dehydrated and mounted in crystal mount. Staining was analyzed by two independent observers under an optical microscope (magnification, ×400; CX31; Olympus Corporation, Tokyo, Japan). Ultimately, extracted DNA concentration was determined using an ND-1000 spectrophotometer (NanoDrop; Thermo Fisher Scientific, Inc., Wilmington, DE, USA).

Each tumor specimen was examined for KRAS exon 2, 3 and 4; NRAS exon 2 and 3; HRAS exon 2; and BRAF exon 15 (codon 600). AmpliSeq Designer v.1.2.6 software (Thermo Fisher Scientific, Inc., Waltham, MA, USA) was used to design primer pairs for PCR amplification of each gene region of interest (17). DNA was amplified using GoTaq Hot Start Polymerase (Promega Corporation, Madison, WI, USA) and 0.2 µM each primer on the GeneAmp PCR System 9700 (Applied Biosystems; Thermo Fisher Scientific, Inc.). Cycling conditions were as previously described (18). Amplicons were finally Sanger sequenced bidirectionally on an ABI 3730XL genetic analyzer (Invitrogen; Thermo Fisher Scientific, Inc.). Primers and procedures were the same as previously reported (19).

Immunohistochemical (IHC) analysis of C-MET protein expression

Immunohistochemistry was performed as described previously (11). Briefly, slides were dewaxed, rehydrated and antigens were retrieved with EDTA (pH 8) by microwave heating at 95°C. Following the inhibition of endogenous peroxidase activity and blocking non-specific antibody binding, sections were incubated with lyophilized primary antibody against C-MET (1:100; EP1454Y; BD Biosciences, Franklin Lakes, NJ) overnight at 4°C. Following a 30-min incubation at room temperature with secondary antibodies (cat. no. sc-3699; 1:200; Santa Cruz Biotechnology, Inc., Dallas, TX, USA), immunoreaction was visualized using the streptavidin-biotin peroxidase complex method. Subsequently, slides were examined under an optical microscope (magnification, ×400, CX31; Olympus Corporation). C-MET staining was assessed according to Hercep Test guidelines (20) as follows: 0, no membrane staining or membrane staining in <10% of tumor cells; 1+, faint membrane staining; 2+, moderate and smooth membrane staining; 3+, strong and granular membrane staining in ≥10% of tumor cells. C-MET overexpression was defined as IHC 2+/3+. The results were judged by two independent pathologists.

Statistical analysis

Data analysis was performed using SPSS version 19.0 (SPSS, Inc., Chicago, IL, USA). Pearson's Chi-square (χ2) test was used to compare the correlation between RAS/BRAF mutations and clinicopathological variables. Kruskal-Wallis test or Mann Whitney U test were performed to compare treatment response. Survival curves of OS and DFS were plotted via Kaplan-Meier analysis with significance assessed using log-rank test. Univariate and multivariate proportional Cox models were performed to assess independent prognostic factors. Logistic regression using a backward stepwise method and receiver operating characteristic (ROC) analysis were performed to evaluate synchronous liver metastasis of patients with CRC. P<0.05 was considered to indicate a statistically significant difference.

Results

Frequencies of gene mutations and C-MET status in stage III CRC patients

Mutations in KRAS, NRAS and HRAS were observed in 43.9% (164/374), 2.4% (9/374) and 0.3% (1/374) of patients, respectively. In addition, as another vital component of the EGFR pathway, BRAF mutations were observed in 5.9% (22/374) cases. Mapping correlations between molecular biomarkers demonstrated that 4 patients carried concurrent KRAS and NRAS mutations (combinations were p.G12D/p.G12D, p.G12D/p.A18T and p.A146T/p.Q61L), and in another 4 patients, KRAS and BRAF mutations (combinations were all p.G12D/p.V600E) were concomitantly observed. However, no co-mutations of NRAS with BRAF were observed in the present study. Notably, the most prevalent mutation occurred in exon 2 (codons 12 and 13) of KRAS (38.0%, 142/374). The detailed distribution of KRAS and NRAS mutation subtypes is presented in Fig. 1A and B.

Figure 1.

Mutation subtypes frequency distribution of (A) KRAS and (B) NRAS.

In addition, the status of C-MET protein in all stage III CRC biopsies were investigated via IHC assay (Fig. 2). It was observed that 269 (71.9%) cases exhibited C-MET overexpression (Fig. 2B-D). In paired non-tumorous specimens, C-MET staining was either absent or present in the membrane of only a few cells (Fig. 2A).

Figure 2.

Analysis of C-MET expression by immunohistochemistry in colorectal carcinomas. C-MET expression was localized in the membrane and its expression was observed predominantly in cancer cells. (A) Negative C-MET staining in a cancerous tissue sample (magnification, ×100). (B) Positive C-MET staining in tumor cells (upper), with negative or weak staining in adjacent epithelial cells (lower) (magnification, ×100). (C) Strong C-MET staining in tumor nests (magnification, ×100). (D) Positive membrane staining, as observed in the majority of tumor cells (magnification, ×200). C-MET, c-mesenchymal epithelial transition factor.

Associations between RAS/BRAF mutations and C-MET overexpression with clinicopathological features

The present study evaluated the correlations of RAS/BRAF and C-MET status, alone or in combination, with the clinicopathological characteristics in patients with stage III CRC. Briefly, KRAS mutations were significantly correlated with vascular invasion (P<0.001) and late distant metastasis, particularly lung metastases (P=0.001). NRAS mutations were more likely to exhibit low COX-2 expression (P=0.001). Furthermore, BRAF exhibited a higher mutation rate in female patients than males (P<0.001) and right colon than other tumor locations (P=0.002; Table I). The present study demonstrated that, compared with low C-MET expression, C-MET overexpression was more likely to occur in cases with late nodal stage (P=0.019), vascular invasion (P=0.023) and late distant metastases, particularly lung and liver metastases (P<0.001; Table II). Considering both RAS/BRAF mutations and C-MET status, there were significant differences in the clinicopathological features distribution among different groups. For patients in group 4, vascular invasion (P=0.001), high carcino-embryonic antigen level (P=0.031) and late distant metastases (P<0.001) were observed at significantly higher levels than in the other groups (Table III).

Table I.

Correlation between mutation profile and clinicopathological features in 374 patients with stage III colorectal cancer.

		KRAS status			BRAF status			NRAS status
Clinicopathological features	Patients, n	Wild-type (n=210)	Mutation (n=164)	P-value	Wild-type (n=352)	Mutation (n=22)	P-value	Wild-type (n=365)	Mutation (n=9)	P-value	All wild-type (n=186)	Any mutation (n=188)	P-value
Sex				0.470			<0.001			0.951			0.076
Male	204	118 (57.8)	86 (42.2)		200 (98.0)	4 (2.0)		199 (97.5)	5 (2.5)		110 (53.9)	94 (46.1)
Female	170	92 (54.1)	78 (45.9)		152 (89.4)	18 (10.6)		166 (97.6)	4 (2.4)		76 (44.7)	94 (55.3)
Age, years				0.355			0.074			0.089			0.079
<65	188	110 (58.5)	78 (41.5)		181 (96.3)	7 (3.7)		186 (98.9)	2 (1.1)		102 (54.3)	86 (45.7)
≥65	186	100 (53.8)	86 (46.2)		171 (91.9)	15 (8.1)		179 (96.2)	7 (3.8)		84 (45.2)	102 (54.8)
Tumor location				0.360			0.002			0.622			0.300
Left colon	166	100 (60.2)	66 (39.8)		152 (91.6)	14 (8.4)		162 (97.6)	4 (2.4)		84 (50.6)	82 (49.4)
Right colon	46	24 (52.2)	22 (47.8)		40 (87.0)	6 (13.0)		44 (95.7)	2 (4.3)		18 (39.1)	28 (60.9)
Rectum	162	86 (53.1)	76 (46.9)		160 (98.8)	2 (1.2)		159 (98.1)	3 (1.9)		84 (51.9)	78 (48.1)
Differentiation				0.609			0.068			0.372			0.611
Well/Moderate	238	136 (57.1)	102 (42.9)		220 (92.4)	18 (7.6)		231 (97.1)	7 (2.9)		116 (48.7)	122 (51.3)
Poor	136	74 (54.4)	62 (45.6)		132 (97.1)	4 (2.9)		134 (98.5)	2 (1.5)		70 (51.5)	66 (48.5)
Depth of invasion				0.406			0.712			0.404			0.360
T1	2	0 (0.0)	2 (100.0)		2 (100.0)	0 (0.0)		2 (100.0)	0 (0.0)		0 (0.0)	2 (100.0)
T2	24	14 (58.3)	10 (41.7)		22 (91.7)	2 (8.3)		24 (100.0)	0 (0.0)		12 (50.0)	12 (50.0)
T3	284	158 (55.6)	126 (44.4)		266 (93.7)	18 (6.3)		275 (96.8)	9 (3.2)		138 (48.6)	146 (51.4)
T4	64	38 (59.4)	26 (40.6)		62 (96.9)	2 (3.1)		64 (100.0)	0 (0.0)		36 (56.3)	28 (43.8)
Nodal stage				0.299			0.941			0.489			0.126
N1	260	148(56.9)	112 (43.1)		244(93.8)	16 (6.2)		255 (98.1)	5 (1.9)		132 (50.8)	128 (49.2)
N2a	74	44 (59.5)	30 (40.5)		70 (94.6)	4 (5.4)		72 (97.3)	2 (2.7)		40 (54.1)	34 (45.9)
N2b	40	18 (45.0)	22 (55.0)		38 (95.0)	2 (5.0)		38 (95.0)	2 (5.0)		14 (35.0)	26 (65.0)
Vascular invasion				<0.001			0.222			0.160			0.001
No	308	186 (60.4)	122 (39.6)		292 (94.8)	16 (5.2)		299 (97.1)	9 (2.9)		166 (53.9)	142 (46.1)
Yes	66	24 (36.4)	42 (63.6)		60 (90.9)	6 (9.1)		66 (100.0)	0 (0.0)		20 (30.3)	46 (69.7)
Initial CEA, ng/ml				0.613			0.350			0.757			0.950
<20	100	54 (54.0)	46 (46.0)		96 (96.0)	4 (4.0)		98 (98.0)	2 (2.0)		50 (50.0)	50 (50.0)
≥20	274	156 (56.9)	118 (43.1)		256 (93.4)	18 (6.6)		267 (97.4)	7 (2.6)		136 (49.6)	138 (50.4)
Late distant metastases				0.001			0.628			0.574			0.001
No	46	36 (78.3)	10 (21.7)		44 (95.7)	2 (4.3)		46 (100.0)	0 (0.0)		34 (73.9)	12 (26.1)
Liver	126	76 (60.3)	50 (39.7)		116 (92.1)	10 (7.9)		121 (96.0)	5 (4.0)		61 (48.4)	65 (51.6)
Lung	68	29 (42.6)	39 (57.4)		63 (92.6)	5 (7.4)		67 (98.5)	1 (1.5)		24 (35.3)	44 (64.7)
Abdomen	72	42 (58.3)	30 (41.7)		69 (95.8)	3 (4.2)		70 (97.2)	2 (2.8)		41 (56.9)	31 (43.1)
Others	62	27 (43.5)	35 (56.5)		60 (96.8)	2 (3.2)		61 (98.4)	1 (1.6)		26 (41.9)	36 (58.1)
COX-2 expression				0.080			0.180			<0.001			0.126
Negative/Weak	32	24 (75.0)	8 (25.0)		31 (96.9)	1 (3.1)		28 (87.5)	4 (12.5)		21 (65.6)	11 (34.4)
Moderate	66	36 (54.5)	30 (45.5)		59 (89.4)	7 (10.6)		66 (100.0)	0 (0.0)		29 (43.9)	37 (56.1)
Strong	276	150 (54.3)	126 (45.7)		262 (94.9)	14 (5.1)		271 (98.2)	5 (1.8)		136 (49.3)	140 (50.7)
MSI				0.466			0.111			0.448			0.979
MSI-H	22	14 (63.6)	8 (36.4)		19 (86.4)	3 (13.6)		22 (100.0)	0 (0.0)		11 (50.0)	11 (50.0)
MSI-L/MSS	352	196 (55.7)	156 (44.3)		333 (94.6)	19 (5.4)		343 (97.4)	9 (2.6)		175 (49.7)	177 (50.3)

Data are presented as n (%), unless otherwise stated. COX-2, cyclooxygenase-2; CEA, carcinoembryonic antigen; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, stable MSI.

Table II.

Correlation between C-MET overexpression and clinicopathological features in 374 patients with stage III colorectal cancer.

		C-MET overexpression
Clinicopathological features	Patients, n	No (n=105)	Yes (n=269)	P-value
Gender				0.690
Male	204	59 (28.9)	145 (71.1)
Female	170	46 (27.1)	124 (72.9)
Age, years				0.610
<65	188	55 (29.3)	133 (70.7)
≥65	186	50 (26.9)	136 (73.1)
Tumor location				0.699
Left colon	166	50 (30.1)	116 (69.9)
Right colon	46	13 (28.3)	33 (71.7)
Rectum	162	42 (25.9)	120 (74.1)
Differentiation				0.103
Well/Moderate	238	60 (25.2)	178 (74.8)
Poor	136	45 (33.1)	91 (66.9)
Depth of invasion				0.251
T1	2	0 (0.0)	2 (100.0)
T2	24	6 (25.0)	18 (75.0)
T3	284	75 (26.4)	209 (73.6)
T4	64	24 (37.5)	40 (62.5)
Nodal stage				0.019
N1	260	84 (32.3)	176 (67.7)
N2a	74	15 (20.3)	59 (79.7)
N2b	40	6 (15.0)	34 (85.0)
Vascular invasion				0.023
No	308	94 (30.5)	214 (69.5)
Yes	66	11 (16.7)	55 (83.3)
Initial CEA, ng/ml				0.072
<20	100	35 (35.0)	65 (65.0)
≥20	274	70 (25.5)	204 (74.5)
Late distant metastases				<0.001
No	46	23 (50.0)	23 (50.0)
Liver	126	27 (21.4)	99 (78.6)
Lung	68	7 (10.3)	61 (89.7)
Abdomen	72	26 (36.1)	46 (63.9)
Others	62	22 (35.5)	40 (64.5)
COX-2 expression				0.490
Negative/Weak	32	10 (31.2)	22 (68.8)
Moderate	66	22 (33.3)	44 (66.7)
Strong	276	73 (26.4)	203 (73.6)
MSI				0.167
MSI-H	22	9 (40.9)	13 (59.1)
MSI-L/MSS	352	96 (27.3)	256 (72.7)

Data are presented as n (%), unless otherwise stated. C-MET, c-mesenchymal epithelial transition factor; COX-2, cyclooxygenase-2; CEA, carcinoembryonic antigen; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, stable MSI.

Table III.

Association of combinational status of RAS/BRAF genes and C-MET protein with clinicopathological features.

Clinicopathological features	Patients, n	Group 1 (n=62)	Group 2 (n=124)	Group 3 (n=43)	Group 4 (n=145)	P-value
Gender						0.053
Male	204	32 (51.6)	77 (62.1)	27 (62.8)	68 (46.9)
Female	170	30 (48.4)	47 (37.9)	16 (37.2)	77 (53.1)
Age, years						0.068
<65	188	39 (62.9)	63 (50.8)	16 (32.6)	70 (50.3)
≥65	186	23 (37.1)	61 (49.2)	27 (67.4)	75 (49.7)
Tumor location						0.190
Right colon	46	8 (12.9)	10 (8.1)	5 (11.6)	23 (15.9)
Left colon/Rectum	328	54 (87.1)	114 (91.9)	38 (88.4)	112 (84.1)
Differentiation						0.293
Well/Moderate	238	33 (53.2)	83 (66.9)	27 (62.8)	95 (65.5)
Poor	136	29 (46.8)	41 (33.1)	16 (37.2)	50 (34.5)
Depth of invasion						0.310
T1+T2	26	2 (3.2)	12 (9.7)	4 (9.3)	8 (5.5)
T3+T4	348	60 (96.8)	112 (90.3)	39 (90.7)	137 (94.5)
Nodal stage						0.054
N1	260	50 (80.6)	82 (66.1)	34 (79.1)	94 (64.8)
N2	114	12 (19.4)	42 (33.9)	9 (20.9)	51 (35.2)
Vascular invasion						0.001
No	308	57 (91.9)	109 (87.9)	37 (86.0)	105 (72.4)
Yes	66	5 (8.1)	15 (12.1)	6 (14.0)	40 (27.6)
Initial CEA (ng/ml)						0.031
<20	100	16 (25.8)	34 (27.4)	19 (44.2)	31 (21.4)
≥20	274	46 (74.2)	90 (72.6)	24 (55.8)	114 (78.6)
Late distant metastases						<0.001
No	46	18 (29.0)	16 (12.9)	5 (11.6)	7 (4.8)
Yes	328	44 (71.0)	108 (87.1)	38 (88.4)	138 (95.2)
COX-2 expression						0.657
Negative/Weak	32	4 (6.5)	12 (9.7)	2 (4.7)	14 (9.7)
Moderate/Strong	342	58 (93.5)	112 (90.3)	41 (95.3)	131 (90.3)
MSI						0.523
MSI-H	22	5 (8.1)	5 (4.0)	4 (9.3)	8 (5.5)
MSI-L/MSS	352	57 (91.9)	119 (96.0)	39 (90.7)	137 (94.5)

Group 1, RAS/BRAF-wild without C-MET overexpression; group 2, RAS/BRAF-wild with C-MET overexpression; group 3, RAS/BRAF-mutant without C-MET overexpression; and group 4, RAS/BRAF-mutant with C-MET overexpression. Data are presented as n (%), unless otherwise stated. C-MET, c-mesenchymal epithelial transition factor; COX-2, cyclooxygenase-2; CEA, carcinoembryonic antigen; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, stable MSI.

Survival analysis

By May 1, 2017, the end of follow-up period, 68.4% (256/374) of patients had succumbed. The median follow-up duration was 32.0 months (range, 0.6–76.3 months) and 19 (5.1%) patients were lost to follow-up. The potential influence of RAS/BRAF mutations and C-MET status on survival was analyzed. In the entire study cohort, it was concluded that OS and DFS for RAS/BRAF mutant patients, particularly those exhibiting BRAF mutation, were significantly reduced compared with those of cases with all wild-type. The any-other-KRAS/NRAS-mutated group exhibited longer median OS and DFS (27.2 and 21.4 months, respectively) than the other two mutational groups (Fig. 3A and B). As compared with C-MET low expression cancers (median OS and DFS, 38.7 and 32.3 months, respectively), C-MET overexpression cases (median OS and DFS, 26.4 and 21.2 months, respectively) were correlated with worse OS (P=0.004) and DFS (P=0.036; Fig. 3C and D). Notably, patients in Group 2 exhibited a more favorable survival than those in Group 3, indicating that tumors which harbor single RAS/BRAF mutations demonstrate higher malignant potential in comparison with cases carrying a single C-MET overexpression. Therefore RAS/BRAF mutations may have a more powerful impact on OS and DFS than elevated C-MET (Fig. 4A and B).

Figure 3.

Kaplan-Meier survival curves of patients with stage III colorectal carcinoma. (A) OS and (B) DFS in all wild-type vs. RAS/BRAF mutations. (C) OS and (D) DFS in low C-MET expression vs. C-MET overexpression of entire study population. OS, overall survival; DFS, disease-free survival; C-MET, c-mesenchymal epithelial transition factor; CI, confidence interval.

Figure 4.

Kaplan-Meier survival curves of patients with colorectal carcinoma classified according to RAS/BRAF mutations and C-MET status. (A) OS and (B) DFS based on the combinational status of RAS/BRAF and C-MET. OS, overall survival; DFS, disease-free survival; C-MET, c-mesenchymal epithelial transition factor.

Furthermore, the Cox proportional hazards model was applied to estimate prognostic factors. As confirmed by multivariate analyses, RAS/BRAF mutations emerged as independent risk factors for OS [hazard ratio (HR), 2.045; 95% confidence interval (CI), 1.276–3.279; P=0.003)] and DFS (HR, 1.976; 95% CI, 1.230–3.175; P=0.005), whereas C-MET overexpression only exerted a significant prognostic effect on OS (HR, 2.837; 95% CI, 1.103–6.053; P=0.031; Table IV).

Table IV.

Univariate and multivariate analyses of OS and DFS for 374 patients.

		OS univariate analysis		OS multivariate analysis		DFS univariate analysis		DFS multivariate analysis
Parameter	Variables	HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value
Gender	Male vs. female	1.041 (0.701–1.545)	0.843			1.061 (0.714–1.576)	0.771
Age, years	<65 vs. ≥65	1.258 (0.845–1.874)	0.258			1.048 (0.706–1.554)	0.817
Tumor location	Left/right colon vs. rectum	0.911 (0.623–1.377)	0.658			1.076 (0.871–1.330)	0.496
Differentiation	Well/moderate vs. poor	1.062 (0.702–1.605)	0.776			1.061 (0.085–1.000)	0.771
Depth of invasion	T1+T2 vs. T3+T4	1.011 (0.818–1.250)	0.916			1.140 (0.765–1.700)	0.520
Nodal stage	N0+N1 vs. N2a+N2b	1.042 (0.806–1.347)	0.752			1.123 (0.868–1.453)	0.377
Vascular invasion	No vs. yes	0.982 (0.782–1.234)	0.879			0.968 (0.772–1.214)	0.779
Initial CEA, ng/ml	<20 vs. ≥20	1.154 (0.890–1.497)	0.281			1.186 (0.916–1.536)	0.195
Late distant metastases	No vs. yes	3.334 (2.139–5.197)	<0.001	2.678 (1.655–4.334)	<0.001	3.291 (2.092–5.178)	<0.001	2.782 (1.678–4.435)	<0.001
COX-2 expression	Negative/weak vs. moderate/strong	0.991 (0.758–1.294)	0.946			0.991 (0.759–1.293)	0.946
MSI	MSI-H vs. MSI-L/MSS	0.713 (0.345–1.471)	0.360			0.619 (0.300–1.277)	0.194
C-MET overexpression	No vs. yes	3.032 (1.323–6.948)	0.009	2.837 (1.103–6.053)	0.031	2.642 (1.154–6.045)	0.021	2.382 (0.892–4.753)	0.083
RAS/BRAF mutations	No vs. yes	2.459 (1.617–3.739)	<0.001	2.045 (1.276–3.279)	0.003	2.222 (1.460–3.382)	<0.001	1.976 (1.230–3.175)	0.005
Anti-EGFR therapy	No vs. yes	0.497 (0.229–1.080)	0.077			0.396 (0.182–0.864)	0.020	1.055 (0.411–2.710)	0.911

OS, overall survival; DFS, disease-free survival; HR, hazard ratio; CI, confidence interval; COX-2, cyclooxygenase-2; CEA, carcinoembryonic antigen; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, microsatellite stability; C-MET, c-mesenchymal epithelial transition factor; EGFR, epidermal growth factor receptor.

Predictive value of RAS/BRAF mutations and C-MET overexpression to late metastasis in patients with CRC

As distant metastasis was significantly associated with malignant progression and poor survival in patients with CRC, the potential predictors for late metastasis were investigated using unconditional logistic regression and ROC curves. Items that were verified to be statistically significant were regarded as independent variables. It was observed that RAS/BRAF mutations [yes=1, no=0; odds ratio (OR), 2.544; P=0.002], C-MET overexpression (yes=1, no=0; OR, 3.408; P=0.003) and depth of invasion (T3+T4=1, T1+T2=0; OR, 3.363; P<0.001) were all significantly correlated with the occurrence of late distant metastases (Table V).

Table V.

Logistic regression analysis of factors associated with late distant metastases in patients with colorectal cancer.

Characteristics	OR	95% CI	P-value
Depth of invasion: T3+T4 vs. T1+T2	3.363	1.911–5.916	<0.001
RAS/BRAF mutations: Yes vs. no	2.544	1.402–4.613	0.002
C-MET overexpression: Yes vs. no	3.408	1.527–7.604	0.003
Constant	0.001

CI, confidence interval; OR, odds ratio; C-MET, c-mesenchymal epithelial transition factor.

The number of cases included the whole study population. With ROC curve analysis, the sensitivity and specificity of RAS/BRAF mutations alone, C-MET overexpression alone, depth of invasion alone, or their combination for predicting late distant metastasis among patients with CRC were evaluated. The predictive findings presented in Fig. 5, demonstrated that the combination of RAS/BRAF mutations, C-MET overexpression and depth of invasion [area under curve (AUC), 0.734; 95% CI, 0.672–0.797; P<0.001] exhibited a better predictive value compared with single RAS/BRAF mutations (AUC, 0.618; 95% CI, 0.545–0.691; P=0.003), C-MET overexpression (AUC, 0.600; 95% CI, 0.531–0.670; P=0.011) or depth of invasion (AUC, 0.628; 95% CI, 0.553–0.702; P=0.001).

Figure 5.

ROC curves for the predictive ability of RAS/BRAF mutations and C-MET overexpression to late distant metastasis. ROC, receiver operating characteristic curve; C-MET, c-mesenchymal epithelial transition factor.

Efficacy of anti-EGFR therapies

In the present study, 342 patients suffered from late distant metastasis and/or recurrence during the follow-up period, 46 of whom received cetuximab combined with first-line FOLFIRI (irinotecan/5-Fu/leucovorin) or FOLFOX6 (oxaliplatin/5-Fu/leucovorin) chemotherapy, including 1 patient in group 1, 41 in group 2 and 4 in group 4. No instances of patient complete response (CR) were observed; 1 case in group 1 and 7 cases in group 2 exhibited partial response (PR); 24 cases in group 2 exhibited stable disease (SD), whereas 4 cases in group 4 exhibited all progressive disease (PD) for the first response evaluation at 3 months. The disease control rate (including CR, PR and SD) was 69.6% (32/46). Therefore, the efficacy of anti-EGFR therapy in RAS/BRAF wild-type patients were better than that in mutant counterparts, although no statistical significance was observed. However, the influence of C-MET status on anti-EGFR therapies were not assessed due to the low number of suitable cases.

Discussion

CRC is a clinically and pathologically heterogeneous malignancy, presenting high incidence of metastasis and a consequent poor clinical outcome on account of its invasive nature (1). Despite the complexity of carcinogenesis, a number of molecular studies have been performed in search of more specific and feasible markers with predictive and prognostic significance. As a result, multiple genes, such as vascular endothelial growth factor, cyclooxygenase-2, PIK3CA, protein kinase B and ERBB2 (7,21), have been considered as biomarkers of the aggressiveness of CRC. In recent years, increasing attention has been given to extended RAS and C-MET status, whose abnormalities have been demonstrated to contribute to uncontrolled cell growth and malignant transformation in CRC (18,22). To the best of our knowledge, this is the first study where a combined analysis of RAS/BRAF mutations plus C-MET overexpression was performed, which clarified their clinical value in a large cohort of Chinese patients with stage III CRC.

According to the present data, mutations in KRAS, NRAS, HRAS, BRAF and C-MET overexpression were observed in 43.9% (164/374), 2.4% (9/374), 0.3% (1/374), 5.9% (22/374) and 71.9% (269/374) of cases, respectively. The prevalence of genetic abnormalities was in accordance with previous publications (7,23–26). Different from intra-tumoral heterogeneity of KRAS mutations and rare NRAS or HRAS mutations, BRAF aberrance exhibited relative intra-tumoral homogeneity. In addition, the present study also demonstrated that mutations in RAS/BRAF oncogenes were not mutually exclusive, although the findings conflicted with several reports from other populations (27–29). One likely explanation for this may be the disparity of sample sources (Chinese vs. European population). Notably, emerging studies (30,31) have observed a high concordance of RAS/BRAF mutations between primary CRCs and corresponding metastases, indicating that these genetic changes existed early in tumorigenesis, and maintained their status during development (21). However, the level of concordance for C-MET expression was controversial (22,32). Shoji et al (31), previously indicated that c-MET protein was more highly expressed in liver metastases than in paired primary tumors. In contrast, another study (33) revealed that C-MET expression in late metastases tended to be decreased, which supported the outcome of the present study. Therefore, more studies in ethnically-diverse populations are required.

In the present study, the association between combinational status of RAS/BRAF plus C-MET and clinicopathological features were investigated. Briefly, it was indicated that KRAS mutations and C-MET overexpression, or their combination, may be important indicators to identify subsets of CRC with vascular invasion and late distant metastases. Particularly, 35% of patients in the present study developed liver metastases during their disease course and >50% of cases exhibited metastases in other sites, including lung metastases. Of the cases with liver metastases, 39.7% had KRAS mutations and 78.6% exhibited high C-MET expression. By contrast, genetic abnormalities were more closely associated with lung metastases. In addition, NRAS mutations were correlated with low COX-2 expression, suggesting the reduced aggression of tumors carrying NRAS mutations compared with those with other RAS/BRAF mutations. This is in accordance with previous studies (10,23). Recently, a retrospective study (34) reported that BRAF mutations were observed more frequently in right colon and female patients, which supported the conclusions of the present study. Numerous experimental model systems have confirmed RAS/BRAF mutations and upregulated C-MET collaboration, or their interactions, contributed to cell proliferation and the invasion-metastasis cascade, which may yield tumor aggressiveness and distant organ involvement (6,35). Furthermore, Bradley et al (22), recently illustrated that small interfering RNA-mediated knockdown of c-MET inhibited the migration and invasion potential of CRC cells, thereby suppressing tumor progression and metastasis in vivo. These outcomes indicated that genetic abnormalities are important in promoting CRC malignancy.

The initiation and development of CRC is a complex, multi-step process that is accompanied by the accumulation of diverse gene alterations (3,6). RAS/BRAF mutations are typically the most frequent driver mutations in CRC (36), C-MET overexpression is regarded as adjuvant pro-metastatic marker, both of which represent the principle aspect of somatic genetic changes (37,38). Another focus of the present study was further exploring the predictive value of RAS/BRAF mutations and C-MET status. In one prior study (39), KRAS exon 2-mutated CRC patients exhibited a marked propensity for lung metastases. Similar results have also been described by Morris et al (8), in which all RAS/BRAF mutant cases harbored the trend towards distant metastases. The present data highlighted that RAS/BRAF mutations combined with C-MET overexpression were significant predictors for higher risk of late distant metastasis, suggesting their importance in distinguishing CRCs with highly aggressive behavior from low metastatic lesions. The results also demonstrated that these mutations provide powerful insights into the complexity of tumor foci genotype and provide a rationale for the combination therapeutic strategies. Previous studies have proposed that the block of C-MET, the HDAC inhibitor and CDK1 inhibition may markedly attenuate CRC development (40–42).

Previously, KRAS mutation was regarded as an adverse prognostic indicator in 1990 (43). Only in the last several years has the prognostic value of extended RAS mutations in CRC received more attention. Conversely, high C-MET expression has been documented to be associated with lower survival in diverse human tumors (12,32). A previous study (31) has demonstrated that C-MET overexpression indicated a poor outcome in terms of the risk of recurrence and mortality in patients with mCRC following metastasectomy. Similarly, the present data also revealed that C-MET overexpression and RAS/BRAF mutations, particularly BRAF mutation, were significantly associated with shorter OS and DFS in the entire study population. Notably, compared with C-MET overexpression, RAS/BRAF mutations appeared to be more powerful prognostic markers of a short interval to low survival and late metastasis following surgery. Furthermore, as the National Comprehensive Cancer Network recommends patients with mCRC and RAS/BRAF wild-type for anti-EGFR treatment (44), the present results also illustrated wild-type cases may gain survival benefits from cetuximab. Regarding C-MET status, Inno et al (32) previously proposed that C-MET overexpression was significantly associated with a worse outcome and anti-EGFR resistance; whereas in the present study, too small sample size in low C-MET expression patients treated with cetuximab prevented the elucidation of potential therapeutic importance of C-MET. A focus on this issue is required in future studies.

In view of the retrospective nature of the current methodology, there has been an inevitable selection bias in the present outcomes. Firstly, certain participants and their medical record documentation may have been lost to follow-up, particularly for those who were not hospitalized following first-line chemotherapy. Secondly, the patients were heterogeneous and selected according to the availability of genetic detection, which limited data analyses. Therefore, further prospective studies are required to confirm the present conclusions.

In conclusion, the status of RAS/BRAF and C-MET may serve as significant predictors for metastatic behavior and refining prognosis in CRC. Accordingly, radiological diagnosis in combination with RAS/BRAF and C-MET detection may help in the prognostic evaluation for postoperative stage III CRC cases, as well as devised appropriate individualized medicine in the future.