BRAF V600E and SRC mutations as molecular markers for predicting prognosis and conversion surgery in Stage IV colorectal cancer

Comprehensive genomic sequencing (CGS) enables us to detect numerous genetic alterations in a single assay. We aimed to identify molecular markers for predicting prognosis and conversion surgery in Stage IV colorectal cancer (CRC) using CGS. One-hundred eleven patients with Stage IV CRC who underwent primary tumor resection were analyzed. We retrospectively investigated genetic alterations using CGS of a 415-gene panel. Clinicopathological variables and genetic alterations were analyzed to identify independent prognostic factors of overall survival (OS). Forty-five of 111 patients had R0 resection; of these, 11 patients underwent conversion surgery. Univariate and multivariate analyses identified histopathological grade 3, R0 resection, BRAF V600E mutation, and SRC mutation as independent prognostic factors for OS (P = 0.041, P = 0.013, P = 0.005, and P = 0.023, respectively). BRAF V600E and SRC mutations were mutually exclusive, and SRC mutation was significantly associated with left-sided tumor and liver metastasis compared to BRAF V600E mutation (P = 0.016 and P = 0.025, respectively). Eleven of the 74 initially unresectable patients underwent conversion surgery for R0 resection, yet none harbored BRAF V600E or SRC mutations. BRAF V600E and SRC mutations are important molecular markers which can predict prognosis and conversion surgery in Stage IV CRC.

Residual tumor status and conversion surgery. Patients were classified according to residual tumor status. Patients who achieved R0 resection of both the primary lesion and distant metastasis were classified as "R0", while patients for whom R0 resection could not be achieved were classified as "R2". Conversion surgery was defined as R0 resection after systemic therapy for initially unresectable distant metastasis 3,11,12 . CGS analysis of primary tumors. As we previously reported, CGS of primary tumor was performed as follows [19][20][21] . Tumor content was evaluated by an independent pathologist for each sample using hematoxylin and eosin-stained slides to ensure >50% tumor content. Where applicable, unstained slides were macro-dissected to enrich for tumor content, and DNA was extracted using a BioStic FFPE Tissue DNA Isolation Kit (Mo Bio Laboratories, Inc., Carlsbad, CA). All sample preparations, CGS, and analytics were performed in a CLIA/ CAP-accredited laboratory (KEW Inc., Cambridge, MA). DNA fragment (50-150 ng) libraries were prepared and enriched for the CancerPlex 415-gene panel (KEW Inc.) [19][20][21] . Sequencing was performed on Illumina MiSeq and NextSeq platforms with an average 500X sequencing depth. Genomic data were then processed through a proprietary bioinformatics platform and knowledgebase to identify multiple classes of genomic abnormalities, including single nucleotide substitutions, small insertions/deletions, copy number variations, and translocations. www.nature.com/scientificreports www.nature.com/scientificreports/ Single nucleotide variant (SNV) and insertion or deletion (indel) calling were only performed in genomic regions intended to be captured by the assay (region of interest). We set a standard threshold of 10% allelic fraction for calling SNVs and indels to focus on primary truncal driver mutations and avoid subclonal events. Copy number variants were called for exons as well as globally. We segmented regions using a Fused-Lasso method and exported results to a VCF file. The threshold for gain was >2.5 fold, and for loss was <0.5 fold. Variants were filtered or flagged according to technical quality (e.g. coverage, allelic fraction, number of supporting reads), presence in previously characterized normal samples, or presence/absence in the following databases: dbSNP, ExAC, COSMIC, ClinVar, and KEW. SNVs and indels in VCF format were annotated using SnpEff and the output was adapted according to HGVS recommendations. Tumors were tested for the presence of microsatellite instability (MSI) based on an extended loci panel. In addition to the Bethesda panel, a collection of 950 regions consisting of tandem repeats of one, two, or three nucleotides of minimum length of 10 bases were examined. The number of indels within the regions of interest was calculated, and tumors were classified as MSI-high (MSI-H) or microsatellite stable (MSS).

Prognostic factors.
To identify factors influencing overall survival (OS) after surgery, genetic alterations (identified using the 415 gene panel) and 15 clinicopathological variables (Table 1) were tested in all 111 patients. In this study, genetic alterations that occurred at a frequency of more than 5% were evaluated for their prognostic impact in univariate and multivariate analyses of 5-year OS. Regarding BRAF mutation, we separately evaluated BRAF V600E and non-V600E.
Genetic alterations in EGFR pathway. To investigate the association between efficacy of anti-EGFR therapy and genetic alterations in EGFR pathway, genetic alterations of TK receptors (ERBB2, MET, EGFR, FGFR1, and PDGFRA), MEK/ERK pathway (KRAS, NRAS, HRAS, BRAF, and MAPK2K1), and PI3K pathway (PTEN and PIK3CA) were analyzed using CGS of the 415-gene panel. We defined patients who had no alterations in any of the 12 genes as "all wild-type"; theoretically, these patients should respond to anti-EGFR therapy 23,24 . Statistical analysis. Statistical analyses were performed with IBM SPSS Statistics 22 (IBM Japan, Inc., Tokyo, Japan). Five-year OS rates were estimated using the Kaplan-Meier method. The log-rank test was used to assess for significant differences between subgroups by univariate analysis. To investigate independent prognostic factors for OS, factors with a P-value of less than 0.05 in univariate analyses were entered into a multivariate analysis. The Cox proportional hazards regression model was used to identify factors that were independently associated with OS after surgery. Pearson's chi-squared test or Fisher's exact test was used to evaluate the associations between clinicopathological characteristics and genetic alterations evaluated using CGS. P-values less than 0.05 were considered statistically significant.

Association between Metastatic sites and genetic alterations.
Of the 111 patients, metastasis to the liver, lung, peritoneum, and distant lymph nodes at initial assessment were identified in 88, 34, 22, and 21 patients, respectively (Supplementary Table 1). Metastasectomy was performed for 49 of 111 patients. Among the 49 patients, simultaneous metastasectomy was performed for 23 patients, and metachronous metastasectomy was performed for 26 patients. Typically, chemotherapy was administered according to the guidelines 3,7,25,26 . Median number of lines before resection of metastasis was 1 line (range: 1-4 lines). In 31 genes altered at a frequency of more than 5%, liver metastasis was significantly associated with BRAF wild-type (P = 0.026); lung metastasis was significantly associated with APC wild-type (P = 0.030); peritoneal metastasis was not significantly associated with any genetic alterations; distant lymph node metastasis was significantly associated with ATM mutant-type and TP53 mutant-type (P = 0.010 and P = 0.024, respectively).

Factors influencing OS after primary tumor resection.
Of 111 patients, 81 patients were died of cancer while 4 patients were died of other causes. The OS rate after primary tumor resection in 111 patients was 84.6% at one year, 40.8% at three years, and 11.6% at five years. Forty-five of the 111 patients had R0 resection at both primary and metastatic sites (Fig. 1A). Of these, 11 patients had undergone conversion surgery after systemic chemotherapy (Fig. 1A). Univariate analyses revealed that histopathological grade 3, N1, 2 category, BRAF V600E mutation ( Fig. 2A, Supplementary Fig. 1A), and SRC mutation (Fig. 2B, Supplementary Fig. 1B) were associated with worse OS (Table 1) and R0 resection was associated with better OS (Table 1). Regarding BRAF non-V600E mutations, OS rate of six patients with BRAF non-V600E mutations showed no significant difference compared to BRAF wild-type ( Fig. 2A). Multivariate analysis identified histopathological grade 3, R0 resection, BRAF V600E mutation, and SRC mutation as independent prognostic factors for OS (P = 0.041, P = 0.013, P = 0.005, and P = 0.023, respectively; Table 1).

Stratification by BRAF V600E and SRC mutation status in Stage IV CRC. BRAF mutations were
identified in 13 patients (seven with V600E and six with a non-V600E mutation), and an SRC mutation was identified in 14 patients (13 with amplification and one with deletion). BRAF V600E and SRC mutations were mutually exclusive. Stratification of the 111 patients according to BRAF V600E and SRC mutation status found patients wild-type for both having significantly better OS (P < 0.001; Fig. 3 www.nature.com/scientificreports www.nature.com/scientificreports/ terms of clinicopathological characteristics, SRC mutation was significantly associated with left-sided tumor and liver metastasis compared to BRAF V600E mutation (P = 0.016 and P = 0.025, respectively; Table 2).

Comparison of genetic alterations between initially resectable and initially unresectable groups.
Thirty-seven and 74 patients were diagnosed as initially resectable and initially unresectable, respectively (Fig. 1A). No significant differences were observed between the two groups in genetic alterations altered at frequency of more than 10%.
Genetic alterations of initially resectable group. Of 37 patients who were diagnosed as initially resectable metastatic disease, three patients did not undergo R0 resection because of progressive disease after preoperative chemotherapy; ARID1A and ATM mutations were significantly associated with progressive disease (P = 0.026 and P = 0.042, respectively), however, no significant differences were observed in the other genetic alterations at frequency of more than 10%. Regarding MSI status, one of the three patients who showed progressive disease after preoperative chemotherapy had MSI-H tumor.

Comparison of genetic alterations between conversion and non-conversion groups in initially unresectable patients.
Of 74 patients who were diagnosed as initially unresectable metastatic disease, 11 patients underwent conversion surgery (Fig. 1A, Table 3) and had significantly better OS than R2 patients (P = 0.004; Fig. 4A). Regarding the association between metastatic sites and conversion surgery, 8 of 29 patients (28%) with liver-limited disease received conversion surgery, while only 3 of 45 patients (7%) with the other initially unresectable metastases received conversion surgery. BRAF or SRC mutations were completely absent in the 11 R0 patients that underwent conversion surgery compared with 20 of the 63 R2 patients harboring a mutation in one of these genes (P = 0.023; Fig. 4B), however, no significant differences were observed in the other genetic alterations at frequency of more than 10%.
Genetic alterations in EGFR pathway and response to anti-EGFR therapy. In this cohort, 8 of 74 patients with initially unresectable metastatic disease received anti-EGFR therapy with cytotoxic agents (Fig. 1A). CGS revealed that 5 patients were "all wild-type" in EGFR pathway and theoretically should respond  www.nature.com/scientificreports www.nature.com/scientificreports/ to anti-EGFR therapy. Three of the 5 patients (60%) showed significant response to anti-EGFR therapy: 1 patient showed complete response and alive with no evidence of disease, and 2 patients received conversion surgery after anti-EGFR therapy (Fig. 5A,B). Conversely, no patients with "mutant-type" showed complete response to anti-EGFR therapy or received conversion surgery.

Discussion
The present study had three main findings. First, BRAF V600E and SRC mutations were identified as independent prognostic factors for OS. Second, BRAF V600E and SRC mutations were mutually exclusive, and patients with Stage IV CRC were stratified according to BRAF V600E and SRC mutation status. Third, only patients with wild-type for both BRAF V600E and SRC mutations underwent R0 conversion surgery. These results imply that BRAF V600E and SRC mutations are important molecular markers in Stage IV CRC.
CGS can detect numerous important genetic alterations in many solid cancers 8 . Genetic alterations in the MEK/ERK pathway, such as in KRAS, NRAS, and BRAF V600E, are benchmarks used to determine treatment strategies for patients with metastatic CRC. National Comprehensive Cancer Network (NCCN) guidelines state that all patients with metastatic CRC should have tumor tissue genotyped for KRAS, NRAS, and BRAF V600E mutations, and any patient with a known KRAS or NRAS mutation should not be treated with anti-EGFR therapy such as cetuximab and panitumumab 7 . AJCC 8 th edition states KRAS, NRAS, and BRAF V600E mutations are poor prognostic factors in CRC (level II evidence) 2 . In the present analysis, we identified BRAF V600E and SRC mutations as independent prognostic factors for OS, which may have the potential to stratify Stage IV CRC and predict conversion surgery after systemic therapy. We suggest CGS be utilized in clinical practice to facilitate precision medicine.
CGS allows the simultaneous detection of BRAF V600E mutations as well as BRAF non-V600E mutations in a single assay. BRAF V600E mutation occurs at approximately 5 to 10%, and is one of the important molecular subtypes of CRC 2,3,7 . On the other hand, BRAF non-V600E mutations occur at approximately 1 to 5%, and show  www.nature.com/scientificreports www.nature.com/scientificreports/ different clinicopathological characteristics, such as left-sided tumor and better prognosis, than BRAF V600E mutation [27][28][29][30] . Hence, we separately evaluated BRAF V600E and non-V600E mutations in this analysis.
To the best of our knowledge, this is the first report of clinical significance of SRC mutation in CRC using CGS. Previous reports demonstrated that activating mutation of SRC was found in 12% of CRC patients 31,32 , and indicated that SRC mutation is associated with distant metastasis and drug resistance in CRC [14][15][16][17][18] . In this analysis, we reported that 14 of 111 patients (13%) had SRC mutation, which was significantly associated with left-sided tumor and liver metastasis compared to BRAF V600E mutation. Multivariate analysis revealed that SRC mutations were independent prognostic factors on OS in Stage IV CRC. Moreover, BRAF V600E and SRC mutations were mutually exclusive. Taken together, we consider that SRC mutation might be an important molecular subtype in CRC.  www.nature.com/scientificreports www.nature.com/scientificreports/ Although this analysis found patients with BRAF V600E or SRC mutations had a poorer prognosis and seem poor candidates for conversion surgery, BRAF V600E and SRC mutations are regarded as druggable. In melanoma, the BRAF V600E mutation is the target of the BRAF-mutant inhibitor vemurafenib 33 . However, BRAF V600E mutant CRCs are not sensitive to BRAF-mutant inhibitors due to the feedback reaction of EGFR 34 . As such, clinical trials are planned for metastatic CRC using combinations of BRAF-mutant inhibitors, MEK and EGFR inhibition 35,36 . SRC activity is considered to play a key role in CRC development and metastasis, and there are several trials using SRC inhibitors such as dasatinib 37,38 . Thus, targeting BRAF V600E and SRC may represent a future treatment strategy for Stage IV CRC, which has the possibility to improve OS and rate of conversion surgery in these dismal molecular subtypes.
Emerging CGS technologies enable the detection of numerous genetic mutations in a single assay, and we can now analyze the clinical significance regarding not only "single mutation" such as BRAF V600E and SRC, but also various combinations of genetic alterations using CGS. Recently, clinical impact of "double mutation" has been reported in patients with CRC. Yamashita et al. reported that double mutation of APC and PIK3CA was associated with response to preoperative chemotherapy and poor survival for colorectal liver metastasis 39 . Chun et al. reported that double mutation of RAS and TP53 was associated with shorter OS after hepatectomy for colorectal liver metasstasis 40 . Chow et al. reported that double mutation of KRAS and TP53 was associated with lymph node metastasis in Stage II/III rectal cancer patients who received chemoradiotherapy followed by total mesorectal excision 41 . Although we did not find clinical significance of these double mutations (APC and PIK3CA, RAS and TP53, KRAS and TP53) in our cohort, we consider that, not only "single mutation", but also "double mutation" might be an important concept in the era of precision medicine.
To achieve conversion surgery for initially unresectable patients, a regimen leading to high response rates and/or a large tumor size reduction is recommended, and a cytotoxic doublet plus an anti-EGFR antibody is applied for patients with RAS wild-type disease 3 . Although RAS mutations are established biomarkers of efficacy to anti-EGFR therapy, anti-EGFR therapy is not effective for all patients with a RAS wild-type tumor. Genetic alterations in TK receptors, the RAS pathway (other than KRAS and NRAS mutations), and the PI3K pathway are other possible mechanisms of resistance to anti-EGFR therapy 25,26 . In the present analysis, we defined patients who had no alterations in any of the 12 genes as "all wild-type", and demonstrated that "all wild-type" in the EGFR  www.nature.com/scientificreports www.nature.com/scientificreports/ pathway might be a predictor for significant response to anti-EGFR therapy and subsequent conversion surgery. Thus, we consider that, in addition to BRAF and SRC mutations, genetic alterations in EGFR pathway are also important for precision medicine of Stage IV CRC. This analysis has several limitations. First, it was a retrospective analysis performed at two institutions and included a relatively small number of patients. Second, this analysis included patients who underwent various systemic therapies with or without targeted therapy. Third, this analysis did not include the patients who did not undergo primary tumor resection. Fourth, this analysis did not include patients whose treatment was tailored based upon their tumor's individual genetic alterations. As we are now in the era of precision medicine, future analyses need to assess the value of CGS in patients whose treatment has been tailored to their tumor's genetic alterations.
In conclusion, BRAF V600E and SRC mutations are important molecular markers which can predict prognosis and conversion surgery in Stage IV CRC.