Molecular spectrum of KRAS, NRAS, BRAF and PIK3CA mutations in Chinese colorectal cancer patients: analysis of 1,110 cases

Mutations in genes such as KRAS, NRAS, BRAF and PIK3CA have become an important part of colorectal carcinoma evaluation. The aim of this study was to screen for mutations in these genes in Chinese patients with colorectal cancer (CRC) and to explore their correlations with certain clinicopathological parameters. We tested mutations in the KRAS (exons 2, 3 and 4), NRAS (exons 2, 3 and 4), PIK3CA (exon 20) and BRAF (exon 15) genes using reverse transcriptase-polymerase chain reaction (RT-PCR) and Sanger sequencing in a large cohort of 1,110 Chinese CRC patients who underwent surgical resection at one of three major teaching hospitals located in different regions of China. The prevalence rates of KRAS, NRAS, BRAF and PIK3CA mutations were 45.4%, 3.9%, 3.1% and 3.5%, respectively. Mutant KRAS was associated with the mucinous subtype and greater differentiation, while mutant BRAF was associated with right-sided tumors and poorer differentiation. Our results revealed differences in the genetic profiles of KRAS, NRAS, PIK3CA and BRAF at mutation hotspots between Chinese CRC patients and those of Western countries, while some of these gene features were shared among patients from other Asian countries.

median age of the patient cohort was 62.1 years, ranging from 18 to 96 years. Regarding the histological subtypes, 92.3% (1024/1,110) of tumors were tubular adenocarcinomas, and 7.7% (86/1,110) were mucinous adenocarcinomas. The tumors were graded according to the WHO criteria (WHO Classification of Tumours of the Digestive System, Fourth Edition) as follows: 211 (19.0%) were well differentiated, 816 (73.5%) were moderately differentiated and 83 (7.5%) were poorly differentiated. The locations of the primary tumors included the rectum, the left side of the colon (including sigmoid colon and splenic flexure) and the right side of the colon (cecum, hepatic flexure, transverse colon). In total, 314 (28.3%) samples were located in the left colon, 249 (22.4%) were located in the right colon, and 547 (49.3%) were located in the rectum.

Consistency among ARMS-PCR, Sanger sequencing and NGS results. ARMS-PCR was successful
in all 1,110 cases, while Sanger sequencing failed in 13 cases. Detailed nucleotide changes detected by Sanger sequencing are listed in Supplementary Table 3. Inconsistencies between ARMS-PCR and Sanger sequencing occurred in 4.9% (54/1,110) of cases. In 20.4% (11/54) of the inconsistent cases, two mutations were detected by the ARMS-PCR assay, but only one of them was detected by Sanger sequencing. In the remaining 79.6% (43/54) of cases, the ARMS-PCR assay detected one mutation, while Sanger sequencing detected no mutations. Further validation with NGS sided with ARMS-PCR in 63.0% (34/54) of the inconsistent cases. A total of 9.3% (5/54) of the cases showed two mutations that were detected by ARMS-PCR, while only one of the mutations was detected by NGS. A total of 16.7% (9/54) of the cases showed one mutation that was detected by ARMS-PCR but were deemed mutation-negative by NGS. A total of 11.1% (6/54) of the cases failed to yield sufficient DNA for NGS analysis.
The presence of PIK3CA mutations was noted in 39 cases (3.5%, 39/1,110). Twenty-six tumors had both KRAS and PIK3CA mutations. PIK3CA mutations were present in 26 of 504 (5.2%) patients with KRAS mutations, compared with only 13 of 606 (2.1%) patients with wild-type KRAS (Supplementary Table 4). This finding suggests that PIK3CA and KRAS gene mutations represent partially overlapping subgroups in colon cancer.
BRAF exon 15 mutations were detected in 34 of the 1,110 CRC patients (3.1%). The mutual exclusivity of KRAS (exons 2, 3 and 4), NRAS (exons 2, 3 and 4) and BRAF mutations was confirmed, given that none of the patients with a KRAS or NRAS mutation harbored a simultaneous BRAF mutation. However, three tumors showed concurrent mutations in BRAF and PIK3CA.

KRAS, NRAS, PIK3CA and BRAF mutations and their correlations with clinicopathological findings.
A summary of the relationships among KRAS, NRAS, PIK3CA and BRAF mutations and various clinicopathological features is provided in Table 2. KRAS mutations were significantly more prevalent in tumors in the right side of the colon than in tumors in the left side or the rectum (53.4% vs. 35.0% vs. 47.7%, respectively; P < 0.001). Meanwhile, KRAS mutations were significantly associated with well and moderately differentiated tumors but less so with poorly differentiated tumors (48.3% vs. 46.1% vs. 31.3%, respectively; P = 0.023). KRAS mutations were also more prevalent in mucinous than tubular adenocarcinomas (61.6% vs. 44.0%, respectively; P = 0.002). However, no significant relationship was observed between KRAS mutations and sex (P = 0.288), age (P = 0.185), depth of invasion (P = 0.694), lymph node metastasis (P = 0.410), distant metastasis (P = 0.103) or TNM stage (P = 0.209).

Discussion
EGFR signaling plays a key role in the development and progression of CRC. In particular, this receptor triggers downstream signaling cascades, including the RAS-RAF-MAPK and PI3K-AKT pathways, to stimulate cell proliferation, differentiation, survival and invasion 15,16 . Recently, significant improvements have been made in the treatment of metastatic CRC, with the aid of monoclonal antibodies targeting EGFR, such as cetuximab and panitumumab 17,18 . However, these anti-EGFR therapies have been effective only in a subset of CRC patients 19 . KRAS mutations negatively predict the response to EGFR-targeted therapies in metastatic CRC patients 20,21 . However, only 40% to 60% of KRAS wild-type patients respond to treatment; thus, a considerable number of patients do not benefit from anti-EGFR therapy 18 . It is also important to identify other important molecular alterations that may impact anti-EGFR therapy. De Roock et al. studied the effect of other EGFR downstream mutations on the efficacy of cetuximab in the largest cohort of patients with chemotherapy-refractory metastatic CRC treated with cetuximab plus chemotherapy 12 . Their findings not only confirmed the negative effect of KRAS mutations on the outcome after cetuximab but also demonstrated that BRAF, NRAS, and PIK3CA exon 20 mutations were associated with a low response rate 12 . Gene mutations in the EGFR signaling pathway, such as mutations in KRAS, NRAS, BRAF and PIK3CA, have become an important part of CRC evaluation, and their alterations may determine the therapeutic response to anti-EGFR therapy 12,13 .
In the present study, we evaluated KRAS, NRAS, BRAF, and PIK3CA somatic mutation frequencies in a large cohort of 1,110 CRC patients from three major teaching hospitals located in different regions of China. The present study was a large retrospective, multicenter study that utilized three methods to analyze the mutational profile of Chinese CRC patients. Additionally, we assessed the correlations of these genetic mutations with clinicopathological features. The prevalence rates of KRAS, NRAS, BRAF, and PIK3CA somatic mutations were 45.4%, 3.9%, 3.5%, and 3.1%, respectively. In the remaining 44.1% of patients, no mutations were detected in any of the targets analyzed. These results suggest that approximately 10% of CRC patients harbor NRAS, BRAF, or PIK3CA mutations in a KRAS wild-type population.
In our study, KRAS mutations in codons 12, 13, 61, 117 and 146 were evaluated. The majority of the mutations occurred in codon 12 or 13 (87.9%). The KRAS G12D mutation was the most common, followed by the G12V, G13D, G12S, G12C, G12A, G12R, and G13C mutations. These data differ slightly from the studies of Western populations, suggesting that race might play a role in KRAS mutation patterns 39 .
Our findings suggest that tumors with KRAS mutations are significantly more associated with well-and moderately differentiated tumors compared to poorly differentiated tumors, which is consistent with a previous report 13 . In addition, KRAS mutations occurred more frequently in tumors on the right side of the colon compared to those on the left side and in the rectum. Interestingly, tumors with a KRAS mutation tended to occur more frequently in mucinous adenocarcinomas compared with the tubular subtype, which further supports previous findings. Other clinicopathological features, such as sex, age, depth of invasion, lymph node metastasis or TNM stage, did not exhibit any association with KRAS mutations, which is consistent with a recent report of Chinese patients 22 .
BRAF is a member of the RAF gene family and acts as a downstream effector of activated KRAS. The frequency of BRAF mutations varies widely from 1.1% to 25% worldwide [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38] . The V600E mutation frequency of 3.1% observed in this study is consistent with various Asian studies (1.1% to 5.8%) but is lower than several Western studies 33,34,36 . This difference in the mutation frequency may be attributable to different sample selections, ethnicities and geographical distributions. Notably, BRAF mutations were more common on the right side of the colon than on the left side of the colon and the rectum, which is similar to the findings of a previous study 36 . Furthermore, none of the BRAF-mutated samples harbored a concurrent KRAS mutation, indicating that these mutations are mutually exclusive, which is consistent with previous findings 29,40 (Supplementary Table 4). BRAF mutations were also more common in poorly differentiated tumors than well and moderately differentiated tumors, which supports previous findings 13 .
Notably, in several Western population-based studies, mutant BRAF was either strongly associated with clinicopathological characteristics such as older age and female gender or at least showed a trend toward such associations, while this was not the case in the present study [41][42][43] . In our study, the age and gender distribution were identical in BRAF mutant and BRAF wild-type patients. This finding is interesting because in another large Chinese cohort study by Shen et al. such an association was neither established nor was a trend shown 13 . Thus, we conclude that such a discrepancy might not be the result of limited BRAF mutant cases in the sample, but rather a distinct difference between Western and Chinese populations. BRAF associations in western populations might be secondary to other primary unknown associations such as life style (diet, tobacco…) that are probably different between western and Chinese populations. However, a more in-depth investigation is needed to further elucidate the underlying mechanisms.
The PIK3CA gene encodes the p110 catalytic subunit of PI3K that regulates this signaling pathway 3 . Prognostic and treatment-predictive effects have not been firmly established. De Roock et al. reported that PIK3CA exon 20 mutations were significantly associated with a low response rate 12 . A recent meta-analysis also reported that PIK3CA exon 20 mutations were associated with reduced response rates and OS 44 . We detected PIK3CA exon 20  Table 3. Multivariate logistic regression analysis of the relationship between gene mutations and clinicopathological characteristics in CRC patient. OR: odds ratio; 95% CI: 95% confidence interval.
As one of the RAS family members, NRAS contains effector binding domains identical to KRAS. Thus, NRAS mutations in codons 12, 13, 61 and 146 yield effects similar to those observed as a result of KRAS activation 46 . Unlike KRAS mutations, which constitute a large percentage of the mutations in colorectal cancer, NRAS mutations are rare. To date, there are few data available concerning the prevalence of NRAS mutations. Irahara et al. reported a mutation incidence of 2.2% 47 , whereas a 4.19% mutation rate was reported in a Chinese study 13 , and a rate of 9.57% was reported in Greek and Romanian patients 48 . We detected NRAS mutations in 3.9% of the tumors in the present study. This discrepancy may be attributed to different diagnostic techniques, ethnicities, genetic factors, and geographical distributions. NRAS mutations can coexist with KRAS mutations, but no significant association between these mutations was noted in this study. Similarly, no significant associations between NRAS mutations and clinical characteristics were observed in our study (Supplementary Table 4).
Discrepancies in the prevalence of several mutations might raise concern. However, it is agreed upon that the development and progression of CRC is a multi-step process that involves the acquisition of numerous mutations. Thus, the geographical distribution of patients, which results in different physical, biochemical and environmental exposures to the colon, could have a profound impact on the somatic mutations of the tumor. The present study enrolled patients from 3 major teaching hospitals located in different regions (north, east and southeast) of China to optimize the geographical representativeness of CRC patients. Second, as the accumulation of mutations may be associated with more advanced stages of colorectal cancer, the proportions of advanced diseases in the sample might interfere with the prevalence of mutations. The present study enrolled consecutive, non-selected cases but was still biased by the fact that most of the patients had operable disease upon admission. In other words, most of the patients in the present study were absent for distant metastasis, thus representing the profile of patients with earlier disease. The above considerations might help to explain the lower mutation prevalence in the present study in comparison with other studies.
The present study suffered from the following limitations. First, as stated above, patient selection bias could not be completely eliminated despite our best efforts. Thus, our results should be interpreted with the understanding that our samples represented cases of earlier-stage disease in the Chinese population. Second, due to how recently our patients were diagnosed, follow-up information such as distant metastasis, recurrence, and therapeutic response was incomplete. Third, some hotspot mutations such as in exon 11 of the BRAF gene and in exon 9 of the PIK3CA gene were not screened due to the limitation of the ARMS-PCR kit. Additionally, genetic alterations on the mRNA and protein levels, as well as epigenetic deregulation of CRC, were not discussed. A study that enrolls more stage IV CRC patients at multiple institutions should be incorporated into the present study along with follow-up information to provide a clearer insight into targeted therapy for CRC patients in China. An in-depth investigation into other genetic deregulations in CRC may also aid in the identification of novel targets for personalized therapy in the future.

Conclusions
In conclusion, our results revealed differences in the genetic profiles of KRAS, NRAS, PIK3CA and BRAF at mutation hotspots between Chinese CRC patients and those from Western countries, while some of these gene features were shared among patients from other Asian countries.