Clinical Characterization of Targetable Mutations (BRAF V600E and KRAS G12C) in Advanced Colorectal Cancer—A Nation-Wide Study

BRAF V600E and KRAS mutations that occur in colorectal cancer (CRC) define a subpopulation of patients with an inferior prognosis. Recently, the first BRAF V600E-targeting therapy has been approved and novel agents targeting KRAS G12C are being evaluated in CRC. A better understanding of the clinical characteristics of the populations defined by those mutations is needed. We created a retrospective database that collects clinical characteristics of patients with metastatic CRC evaluated for RAS and BRAF mutations in a single laboratory. A total of 7604 patients tested between October 2017 and December 2019 were included in the analysis. The prevalence of BRAF V600E was 6.77%. Female sex, primary in the right colon, high-grade, mucinous, signet cell, partially neuroendocrine histology, perineural and vascular invasion, and surgical tissue sample were factors associated with increased mutation rates. The prevalence of KRAS G12C was 3.11%. Cancer of primary origin in the left colon and in samples from brain metastases were associated with increased mutation rates. The high prevalence of the BRAF V600E mutation in cancers with a neuroendocrine component identifies a potential candidate population for BRAF inhibition. The association of KRAS G12C with the left part of the intestine and brain metastases of CRC are new findings and require further investigation.


Introduction
Metastatic colorectal cancer (CRC) is one of the leading causes of cancer-related mortality [1,2]. The mitogen-activated protein kinase (MAPK) cascade is the main pathway that transduces signalling from growth factor receptors. It is central in the regulation of proliferation, differentiation, and viability in healthy and malignant epithelial cells. Alterations increasing MAPK activity play a crucial role in cancer pathogenesis [3]. In addition to sustaining proliferative signalling, they are also closely related to the formation of metastases and drug resistance. Pathological MAPK activity is most commonly caused by activating mutations in genes in the RAS (Rat Sarcoma) family. Other pathomechanisms include epigenetic activation of growth factor receptors, activating mutations of genes for receptor tyrosine kinases, or activating mutations in signalling kinases downstream of kinase inhibitors (vemurafenib, dabrafenib) has been proven to be ineffective in metastatic colorectal cancer (mCRC) [11]. It has been shown to trigger the feedback loop that facilitates alternative MAPK activation mechanisms, particularly through increased EGFR and MEK activity [11]. Simultaneous inhibition of this feedback loop through inhibition of EGFR or MEK has been shown to suppress the feedback, thus restoring control over BRAF V600E-activated MAPK [12]. The efficacy of such a strategy was confirmed in clinical trials, most recently in a phase III BEACON study that demonstrated the feasibility of a combined BRAF blockade in a randomized setting. The study also demonstrated that triple inhibition (anti-EGFR + anti-BRAF V600E + anti-MEK with cetuximab, encorafenib, binimetinib) appears to be not more effective than double inhibition (anti-EGFR + anti-BRAF V600E with cetuximab, encorafenib) [13]. As a result, the cetuximab-encorafenib combination is now a standard of care in patients with pretreated BRAF V600E-mutated mCRC. Trials evaluating the first-line efficacy, alternative, and derivative combinations are also under way. Noticeably, another double inhibition strategy (anti-BRAF V600E + anti-MEK) was subjected to an early clinical evaluation [14]. Agents with anti-BRAF activity that have been investigated in this setting include Vemurafenib (Genentech/Roche), Dabrafenib (GlaxoSmithKline), Encorafenib (Array BioPharma/Pierre Fabre), Regorafenib (Bayer), Sorafenib (Bayer), Lifirafenib (BeiGene), Agerafenib (Roche), Belvarafenib (Hanmi Pharmaceutical), Ro 5126766 (Chugai Pharmaceutical), and LY3009120 (Eli Lilly).
Antibodies targeting EGFR (cetuximab and panitumumab) have transformed systemic treatment for advanced CRC and are now an integral part of first-line regimens. As a result, molecular testing for KRAS, NRAS and BRAF mutations has become an obligatory part of the standard diagnostic routine. As MAPK pathway deregulation in CRC becomes better understood, novel, effective therapies targeting this signalling cascade, in unique, molecularly-defined mCRC patient subpopulations, are emerging. The aim of this study is to define the clinical characteristics of mCRC patients harbouring readily targetable mutations (KRAS G12C or BRAF V600E) and to screen for potential clinical factors predictive of the occurrence of the mutations.

Population Characteristics
A total of 7798 cases tested between October 2017 and December 2019 were identified. After discarding 89 duplicates, 102 nondiagnostic results and 3 non-colorectal cancers, 7604 were included in the analysis ( Figure 1). Men made up 58.8%. A total of 75.5% were over 60 and 4.91% were over 80 years old. The distal primary location was more prevalent, with the rectum being the most common one. The majority of the analysed samples originated in the primary tumour (87.9%). Of the 685 samples not originating from primaries, the most common sample source sites were the liver, peritoneal implants, and local recurrence. Surgical samples comprised the majority of the cases (73.49%), with endoscopic biopsy being the second most common sampling method (22.91%). The BRAF V600E mutation occurred in 515 cases (6.77%), KRAS G12C in 237 (3.12%), and any other KRAS in 3591 (47.23%). The detailed characteristics of the general population and the subpopulations are outlined in Table 1 and Table S1.

BRAF V600E Mutation
The prevalence of BRAF V600E was 6.77%. The characteristics of mutated cases compared to those without mutations are outlined in Table 2. BRAF V600E-mutated cancers exhibited higher T and N scores, as well as higher rates of vascular and perineural invasion compared to non-mutated ones.
The mutation was present more often in women (odds ratio (OR) 2.009, 95% confidence interval (CI) 1.677-2.408, p < 0.001). A strong yet nonsignificant trend towards greater mutation risk in older age was observed ( Figure 2). The mutation was more likely to be found in high-grade cancers (OR 4.061, 95% CI 3.218-5.125, p < 0.001). It was more likely to occur in cancers with mucinous histology (OR 3.430, 95% CI 1.415-8.315, p = 0.006) or partially mucinous histology (OR 2.718, 95% CI 2.121-3.482, p < 0.001) or with the presence of signet cells (OR 1.544, 95% CI 1.104-2.160, p = 0.011). Mixed adeno-neuroendocrine cancers (MANEC) were more likely to carry the mutation (OR 9.211, 95% CI 3.071-27.626, p < 0.001). Proximal tumours, occurring in the caecum, ascending colon, hepatic flexure, or transverse colon, were more likely to carry the mutation as compared to distal ones occurring in the splenic flexure, descending colon, sigmoid colon, rectosigmoid junction, or rectum (OR 8.356, 95% CI 6.732-10.372, p < 0.001). Rectal/rectosigmoid tumours were less likely to carry the mutation as compared to the abdominal (proximal and distal combined) part of the colon (OR 0.231, 95% CI 0.179-0.299, p < 0.001), although the difference between rectal/rectosigmoid and distal abdominal tumours (occurring in splenic flexure or descending and sigmoid colon) was not significant (OR 1.32, 95% CI 0.928-1.876, p = 0.122). There was no significant difference in the mutation rate between primary, locally recurrent, and metastatic lesions, although a trend towards a higher rate was observed in primaries. The mutation was less likely to occur in samples from biopsies as compared with surgical ones. All samples secured during colonoscopy were 30.9% less likely to produce a positive result as compared to all samples secured during surgeries (OR 0.691, 95% CI 0.532-0.897, p = 0.006). All samples secured during needle biopsies were 58.7% less likely to produce a positive result as compared to all samples secured during surgeries (OR 0.691, 95% CI 0.532-0.897, p = 0.023). The same was true when restricted to metastatic lesions only. A biopsy was 57.1% less likely to produce a BRAF V600E-mutated sample than a surgery (OR 0.429, 95% CI 0.202-0.911, p = 0.028). Variability in the rate of BRAF V600E mutation was also observed according to geographical regions. Only the difference between the regions with the highest sample numbers (Lodzkie < Slaskie) was statistically significant (OR 0.669, 95% CI 0.481-0.930, p = 0.017). (Tables 3 and S2 (Table S2).

Figure 2.
Odds ratios for BRAF V600E mutation incidence according to selected factors: (a) According to patient's age; (b) According to tumour histological type; (c) According to tumour location in the large intestine; (d) According to procedure resulting in the assessed sample.

KRAS G12C Mutation
The prevalence of KRAS G12C in the overall population was 3.12% (6.60% of all KRAS mutations). The pathomorphological characteristics of the cases with the KRAS G12C mutation compared to the rest of the cohort and other KRAS-mutated cases are outlined in Table 4.

KRAS G12C Mutation
The prevalence of KRAS G12C in the overall population was 3.12% (6.60% of all KRAS mutations). The pathomorphological characteristics of the cases with the KRAS G12C mutation compared to the rest of the cohort and other KRAS-mutated cases are outlined in Table 4.  (Table S3)).
When comparing KRAS G12C mutant cases with the rest of the cohort, proximal tumours were less likely to carry the KRAS G12C mutation compared to distal tumours (OR 0.652, 95% CI 0.451-0.944, p = 0.024). Rectal/rectosigmoid tumours were more likely to carry the mutation as compared to those of the abdominal part of the colon (OR 1.470, 95% CI 1.122-1.925, p = 0.005), although the difference between the rectal/rectosigmoid and intraabdominal distal colon was borderline insignificant (OR 0.729, 95% CI 0.521-1.021, p = 0.066). CNS metastases were 21 times more likely to carry the G12C mutation than the rest of the cohort (OR 20.98, 95% CI 3.48-126.47, p = 0.001) (Tables 5 and S3, Figure 4).  (Table S4)).
When comparing KRAS G12C-mutant cases with the rest of the KRAS-mutant cases, proximal tumours were less likely to carry the KRAS G12C mutation compared to distal tumours (OR 0.618, 95% CI 0.425-0.898, p = 0.012), whereas rectal/rectosigmoid tumours were more likely to carry the mutation as compared to the abdominal part of the colon (OR 1.425, 95% CI 1.083-1.874, p = 0.011). CNS metastases were 15.5 times more likely to carry the G12C mutation than the rest of the cohort (OR 15.529, 95% CI 2.173-110.997, p = 0.006). Higher rates of G12C mutation were observed in samples taken from needle biopsies but not endoscopic biopsies (OR 2.05, 95% CI 1.096-3.833, p = 0.025). This finding was not repeated when restricting the comparison to only metastatic cases and is not present when comparing G12C cases with the general population. (Tables 6 and S4).
Right colon (proximal) included caecum, ascending colon, hepatic flexure, and transverse colon. Left colon (distal) included splenic flexure and descending and sigmoid colon. Rectum included rectosigmoid tumours. MANEC-mixed adeno-neuroendocrine cancer. p-for quantitative variables, Mann-Whitney test; for qualitative variables, chi-squared test or exact Fisher test. Expanded version available in the Supplementary Data (Table S4). When comparing KRAS G12C-mutant cases with the rest of the KRAS-mutant cases, proximal tumours were less likely to carry the KRAS G12C mutation compared to distal tumours (OR 0.618, 95% CI 0.425-0.898, p = 0.012), whereas rectal/rectosigmoid tumours were more likely to carry the mutation as compared to the abdominal part of the colon (OR 1.425, 95% CI 1.083-1.874, p = 0.011). CNS metastases were 15.5 times more likely to carry the G12C mutation than the rest of the cohort (OR 15.529, 95% CI 2.173-110.997, p = 0.006). Higher rates of G12C mutation were observed in samples taken from needle biopsies but not endoscopic biopsies (OR 2.05, 95% CI 1.096-3.833, p = 0.025). This finding was not repeated when restricting the comparison to only metastatic cases and is not present when comparing G12C cases with the general population. (Tables 6 and S4).

Discussion
Our study used one of the largest single-laboratory-generated databases comprising cases of advanced colorectal Caucasian patients worldwide. In the recent meta-analysis of 67 studies (32,407 patients) evaluating the clinical and pathological significance of the BRAF V600E mutation in colorectal cancer, the largest study collected less than 2000 cases. Our analysis comprises 7604 patients' records, and this population would represent approximately 23% of all summarized records in the aforementioned meta-analysis. Moreover, there is not a single large-scale study published to date evaluating the clinicopathological features associated with KRAS G12C mutation, which is expected to shortly become the first clinically targeted KRAS mutation in CRC.
The prevalence of BRAF V600E mutations (below 7% (6.77%)) is lower than expected, based on previously published data. The observed mutation rate varies considerably when compared across other studied cohorts. The aforementioned meta-analysis (n = 32,407) reported a mean rate of 11.34% (3.14-23.14%) [15]. In contrast, previous, much smaller studies conducted on the Polish population showed values similar to the one reported in our analysis. A study on non-metastatic CRC patients found the mutation in 2/163 (1.22%) cases [16]. Two Polish studies on advanced CRC reported the incidence of BRAF V600E mutations in 24/500 (4.80%) and 7/102 (6.86%), respectively [17,18]. A recently published large study on the Russian population reported the incidence at 6.7% [19]. The observed variability may be due to a number of reasons, described below. However, it must be underscored that our analysis was unique in its homogeneity, in that all of the tested patients were Caucasian.
Most studies found that the BRAF V600E mutation was more likely to occur in women and in proximal tumours, which is also a finding in the current study. An association of mutation risk with age has also been described, although it did not reach the significance threshold in this study [20]. In some studies, a higher clinical stage at the time of diagnosis was associated with an increased risk of mutation [15,21,22]. The distribution of these and other risk factors among different cohorts may explain the variability.
Another reason for the lower-than-expected BRAF V600E mutation rates may be survivor bias. The V600E mutation is associated with an aggressive phenotype, adverse metastasizing patterns, and rapid progression. Therefore, it has been shown to negatively impact OS in patients with radically treated and metastatic colorectal cancer. Interestingly, the same studies did not show a significant impact on disease-free survival (DFS) or progression-free survival (PFS) [23][24][25][26][27]. Therefore, it can be concluded that it is the fast rate of disease progression that drives the decrease in OS, whereas the effectiveness of treatment (surgical or systemic as measured in recurrence rates and PFS) does not appear to be affected by the mutation. This indirectly hints that the mutation rate in the radically treated population should remain unchanged in the subpopulation of patients who ultimately recur. In this cohort, patients were only tested for the BRAF mutation when presenting with metastatic disease. Effectively, only patients who maintained a good general condition for long enough to be diagnosed, tested, and treated were included in the analysis. Therefore, subpopulations with adverse prognostic factors for metastatic disease (including BRAF V600E mut, older age, higher tumour burden, poor performance status) are likely to be underrepresented. This selection practice resulted directly from the national reimbursement policy. The authors are aware that most clinical practice guidelines for CRC treatment recommend testing at the time of first diagnosis.
In this study, the samples secured during surgery contained significantly higher numbers of BRAF V600E mutations than the samples secured by endoscopy. A similar relationship was not observed for the KRAS G12C mutation. Therefore, in the authors' opinion, it is unlikely that the BRAF V600E mutation rate was affected by sample processing or the methodology for detecting mutations. The difference in the BRAF V600E rate between surgical and endoscopic samples may represent the differences in the possibilities of collecting material during colonoscopy. It is not always possible to reach the right side of the colon with the endoscope; therefore, the material from this location is less often collected during endoscopic examination. The BRAF-mutated tumours are associated with a right-sided location in the colon and such samples are less likely to be secured by endoscopy.
Tumour histology is another factor that was associated with the appearance of the BRAF mutation. This study confirmed the well-documented link between the risk of the BRAF mutation and mucinous histology, high tumour grade, and vascular and perineural invasion. The presence of signet cells or partial neuroendocrine components also increased the odds of the mutation. The last finding, although previously described [28], suggests a potential therapeutic target in this rare and clinically difficult cancer subtype. There are already case studies describing responses to dual inhibition of BRAF and MEK in this population [29,30]. Recently, a dabrafenib-trametinib combination has gained tissueagnostic approval in the United States [31], which now allows any malignancy with the mutation to be treated with this regimen. Tumour sidedness has also been shown to affect the rate of BRAF mutation in this study, with proximal (right-sided) tumours manifesting higher rates. Notably, rectal/rectosigmoid tumours were associated with significantly lower mutation rates than tumours located in the abdominal part of the colon. Because the occurrence of the mutation did not differ significantly between the rectal/rectosigmoid and distal abdominal colon (only a moderate trend was present), this difference is likely a manifestation of well-documented biological differences between the proximal and distal parts of the large intestine. A study on a larger dataset may resolve this doubt. This study also confirms the link between the risk of BRAF mutation and female sex, mucinous histology, high tumour grade, vascular and perineural invasion, and high T and N scores, which is also well documented in the literature [32].
In this study, some geographical variability is evidenced in the rates of BRAF mutations. Only the difference between theŚląskie and Łódzkie regions is significant, but notably, those were the ones with the highest number of samples. Nonsignificant trends for other regions are also present. In the authors' opinion, this finding should be further evaluated in a larger study. This may provide insight into hereditary and environmental factors that influence the rates of BRAF mutations, but it may also be a manifestation of practice patterns that influenced patient selection for this cohort.
The frequency of KRAS G12C mutations in this study was 3.12%, which is similar to recently published datasets [33][34][35]. Surprisingly, the KRAS G12C mutation in the Polish population occurred more frequently in distal tumours, contrary to a known predisposition of RAS mutations for the proximal part of the colon [5][6][7][8] and data on KRAS G12C-mutated tumours in the Scandinavian population [35]. Additionally, samples from rectal/rectosigmoid cancers were more likely to carry the mutation compared to samples from the intraabdominal part of the intestine. These findings were true both in comparison to the rest of the cohort and in comparison to the rest of the KRAS-mutated cases. Given that there was only a borderline nonsignificant trend towards a higher mutation rate in the rectal/rectosigmoid tumours as compared to the left colon tumours, this should be interpreted with caution, as it may only be a manifestation of the aforementioned predisposition to the distal parts of the intestine. This predisposition for left-sidedness seems to be a new characteristic of the G12C mutation and should be confirmed in other datasets. The authors identified four recent publications that included sidedness in KRAS G12C vs. KRAS non-G12C population characteristics. In two of them, there was a trend towards a higher percentage of left-sided primaries in the KRAS G12C cohort compared to the cohort with the remaining RAS mutations [35,36]. In the remaining two, a higher percentage of right-sided primaries was reported in KRAS G12C-mutated cancers [33,37]. The aforementioned trends were not statistically significant.
Of the five samples originating from brain metastases in this study, two are KRAS G12C-mutated. Although the numbers are small, the rates of G12C occurrence in brain metastases differ significantly, both compared to the rest of the cohort and to the rest of the KRAS-mutated cases. Notably, another two of the brain metastasis samples harboured KRAS mutations (G12D and G13D). Evidence that KRAS mutations in general may play a role in the pathogenesis of CRC brain metastases has previously been described [38]. The finding in this study suggests that there may be a link between brain metastases and the G12C mutation, specifically. Recent studies involving brain metastases from lung cancer have reported similar findings [39,40]. If confirmed in further studies, the overrepresentation of the G12C mutation in brain metastases may be of clinical significance, in light of recent preliminary data on sotorasib being active in brain metastases [41].
A higher prevalence of KRAS G12C mutations in samples taken from needle biopsies is a puzzling finding. It only presented itself in the comparison between G12C and other KRAS mutations. A similar finding was not found for endoscopic biopsies and when the analysis was restricted to metastatic samples only. This may be a manifestation of selection bias, but the exact mechanisms are unclear and require further investigation.

Materials and Methods
We created a dedicated retrospective database of consecutive patients with metastatic colorectal cancer from 107 Polish institutions whose tissue samples had been tested for KRAS, NRAS, and BRAF mutations in a single laboratory. The cohort consisted of samples that were tested as part of standard practice. Although there was no active selection strategy, patients were passively selected, as testing was driven by the restrictive inclusion criteria of a national reimbursement policy. Clinical data were extracted from available records, including basic demographics (age, sex, geographic region), TNM scores (8th edition, 2016), histology (NOS, mucinous, partially mucinous, signet cell, cribriform, comedolike, partially neuroendocrine, other), primary location (right or proximal colon including caecum, ascending colon, hepatic flexure and transverse colon; left or distal colon including splenic flexure, descending and sigmoid colon; rectum/rectosigmoid junction), sample origin (primary, local recurrence, metastatic, according to location of metastases), and sampling method (endoscopic biopsy, needle biopsy, surgery).
The suitability of the samples (amount of tissue, neoplastic cells content) was assessed by a pathologist. The appropriate tissue fragments were collected from microscopic slides by macrodissection. The minimum requirement for analysis was 20% of neoplastic cells, and in most cases (90% of samples) the percentage of neoplastic cells exceeded 50%. A validated Sanger sequencing technique with a detection limit of 10% of the mutant DNA (approx. 20% of cells) was used to test the mutations. The mutation analysis included selected gene fragments: KRAS exon 2 (including codons 10-22), KRAS exon 3 (including codons 58-64), KRAS exon 4 (including codons 117, 146), NRAS exon 2 (including codons 10-18), NRAS exon 3 (including codons 58-64), NRAS exon 4 (including codons 117, 146), BRAF exon 15 (including codons 594-601). The appropriate DNA fragments were amplified by the polymerase chain reaction and subsequently sequenced with the BigDye Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher Scientific) on a sequencer (3730 DNA Analyzer, Applied Biosystems). In parallel to the main Sanger sequencing technique, an independent screening method (SSCP or HRM) was used as an intra-laboratory control in each case. The mutational analysis was performed by a single laboratory that successfully participates in external quality assessment trials organized by ESP and GenQA.
Statistical analysis was performed in the R environment v 4.1.3 [42]. Qualitative variables were compared between groups using the chi-square test (with Yates correction for 2 × 2 tables). Fisher's exact test was used when low expected quantities occurred in the tables. Quantitative variables were compared between groups using the Mann-Whitney test. A p-value of 0.05 was considered statistically significant.

Conclusions
Our nationwide study clearly demonstrates that BRAF V600E and KRAS G12C colorectal tumours show significantly detrimental clinicopathological features. On the other hand, the same mutations are highly predictive of the clinical benefit of novel targeted therapies.
The characteristics of the population with BRAF V600E-mutated CRC were generally consistent with the literature, apart from the lower overall prevalence of the mutation, the anomalies in its geographical distribution, and differences between sample types. These require further investigation. The increased prevalence of the BRAF V600E mutation in partially neuroendocrine cancers may identify a new subpopulation eligible for BRAF inhibition. The characteristics of the KRAS G12C mutation revealed two potentially novel findings: the higher prevalence of the KRAS G12C mutation in distal (left-sided tumours) contrary to other KRAS mutations and the higher prevalence of the KRAS G12C mutation in brain metastases compared to other KRAS mutations. Those require confirmation in other studies but may become clinically relevant with the ongoing development of specific KRAS G12C inhibitors.
With the rise of therapies targeting signal transduction in pathways central to CRC pathogenesis, new subpopulations defined by the ever-growing number of molecular biomarkers emerge. The clinical characterization of these subpopulations is both a requirement for the introduction of new drugs into clinical practice and a source of hypotheses that may affect the direction of future research. The authors present the first results from a newly created database, which hopefully will be expanded with more cases, survival data, and biomarkers to fuel future research.  Informed Consent Statement: Patient consent was waived due to the retrospective nature of the study and local regulations.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author. The data are not publicly available due to ethical and legal concerns.