Clinical, Pathological and Prognostic Features of Rare BRAF Mutations in Metastatic Colorectal Cancer (mCRC): A Bi-Institutional Retrospective Analysis (REBUS Study)

Simple Summary Somatic BRAF mutations occur in approximately 10% of metastatic colorectal cancers (mCRCs) and, according to the involved codon, are classified as V600E and in non-V600, accounting for 80% and 20%, respectively. Being the most frequent mutation, the BRAF V600E mutation has been extensively investigated and up to now its clinical, pathological and molecular phenotype and its prognostic impact have been clearly described. On the contrary, evidence concerning BRAF non-V600 is weaker. We retrospectively evaluated 537 mCRC patients treated at two Italian Institutions. This study corroborates and strengthens available evidence concerning phenotype and prognostic performance of BRAF non-V600 compared to BRAF V600E and BRAF wild-type mCRCs. This deeper insight on rare BRAF non-V600 mutated mCRC is a primary issue in the precision oncology era, since the wider application of NGS is expected to increase the identification of those aberrations. Abstract Recently, retrospective analysis began to shed light on metastatic colorectal cancers (mCRCs) harboring rare BRAF non-V600 mutations, documenting a distinct phenotype and a favorable prognosis. This study aimed to confirm features and prognosis of rare BRAF non-V600 mCRCs compared to BRAF V600E and BRAF wild-type mCRCs treated at two Italian Institutions. Overall, 537 cases were retrospectively evaluated: 221 RAS/BRAF wild-type, 261 RAS mutated, 46 BRAF V600E and 9 BRAF non-V600. Compared to BRAF V600E mCRC, BRAF non-V600 mCRC were more frequently left-sided, had a lower tumor burden and displayed a lower grade and an MMR proficient/MSS status. In addition, non-V600 mCRC patients underwent more frequently to resection of metastases with radical intent. Median overall survival (mOS) was significantly longer in the non-V600 compared to the V600E group. At multivariate analysis, only age < 65 years and ECOG PS 0 were identified as independent predictors of better OS. BRAF V600E mCRCs showed a statistically significant worse mOS when compared to BRAF wild-type mCRCs, whereas no significant difference was observed between BRAF non-V600 and BRAF wild-type mCRCs. Our study corroborates available evidence concerning incidence, clinicopathologic characteristics and prognosis of BRAF-mutated mCRCs.

Besides the more frequent BRAF V600E, mutations occurring outside of codon 600 (BRAF non-V600) account for around 20% [3] and have been less analyzed. Recent retrospective analysis shed light on mCRCs harboring BRAF non-V600, documenting a distinct phenotype (arising more frequently in younger and male patients and being mainly leftsided, low-grade, non-mucinous, with no peritoneal spread and microsatellite stable) and a favorable prognosis (with a significantly longer survival compared with BRAF V600Emutated mCRCs) [3,23].
Lately, BRAF mutants have been classified into 3 functional classes based on the outcome of mutation on kinase activity: class 1 (BRAF V600) and class 2 are constitutively activated and RAS-independent, whereas class 3 mutants have impaired kinase activity or are kinase-dead, being therefore sensitive to inhibition of activated RAS [24]. A recent analyses of BRAF-mutated mCRCs based on functional classes showed that class 3 only displays a divergent phenotype and a better prognosis compared to class 1, while class 2 behaves similarly to class 1 both in terms of features and prognosis [25].
The main limit of evidence concerning BRAF non-V600 mutations depends on the rareness of such aberrations. Thus, with the purpose to strengthen such evidence, we conducted a retrospective analysis aiming to confirm the features and prognostic performance of rare BRAF non-V600-mutated mCRCs compared to BRAF V600E-mutated mCRCs treated at two Italian Institutions. In addition, prognostic performances of BRAF-mutated mCRCs (either V600E or non-V600) were compared to BRAF wild-type datasets.

Materials and Methods
This is an observational, retrospective study conducted at two Italian Institutions: Fondazione Policlinico Universitario "Agostino Gemelli" IRCCS, Roma, Italy, and Azienda Ospedaliera Universitaria Integrata, Verona, Italy. The study was approved by the Ethics Committee of Fondazione Policlinico Universitario "Agostino Gemelli" IRCCS, (Roma, Italy).
The primary objective of the study was to assess and compare demographic, clinical, pathological and molecular features of mCRCs harboring rare BRAF non-V600 mutations to those of BRAF V600E-mutated mCRCs. Secondary objectives were to estimate the prognostic performance of BRAF non-V600 compared to BRAF V600E in mCRCs, and to compare the prognostic performance of BRAF V600E-and BRAF non-V600-mutated mCRCs to that of two additional datasets of BRAF wild-type mCRCs (respectively RASmutated and RAS wild-type) treated at Fondazione Policlinico Universitario "Agostino Gemelli" IRCCS, Roma, Italy, during the same time period.
Clinical charts of patients affected by mCRCs treated between January 2013 and December 2018 were reviewed. Patients with tumors harboring BRAF missense mutations, treated with at least one line of chemotherapy and evaluable for survival, were enrolled. Baseline demographic and clinical characteristics, as well as information concerning treatments and survival, were collected from clinical charts. Pathological and molecular features were retrieved from histological reports. The following data were collected: gender, age and ECOG PS at diagnosis of metastatic disease, onset of metastatic disease, primary tumor location, sites of metastases, mucinous histology, grade of differentiation, BRAF missense mutation, RAS mutational status and MSI/MMR status. Moreover, information concerning received treatments and date of death were retrieved. RAS and BRAF gene mutational status was assessed by means of NGS or pyrosequencing on formalin-fixed, paraffin-embedded (FFPE) archival tumor tissue samples from primary tumor or metastases. Expression of MMR proteins was performed by immunohistochemistry: expression of MLH1, MSH2, MSH6 and PMS2 was tested.

DNA Extraction and Assessment of RAS and BRAF Mutational Status
DNA was extracted from three 10 µm-slides from FFPE tumor tissue samples using the QIAamp DNA FFPE Tissue Kit (QIAGEN, Milan, Italy), following the manufacturer's protocol. In order to minimize contamination by normal cells, the tumor areas dissected for DNA extraction contained at least 70% of tumor cells. KRAS, NRAS and BRAF mutational analyses were carried out by pyrosequencing on chemo-naïve primary or metastatic samples, as previously described [26]. RAS and BRAF mutational analysis were performed using the therascreen KRAS Pyro Kit, therascreen RAS Extension Pyro Kit and therascreen BRAF Pyro Kit (Qiagen, Milan, Italy). Mutation status was determined by pyrosequencing on the Qiagen PyroMark Q24.

Assessment of MMR/MSI Status
MMR status was assessed on FFPE tumor tissue samples using antibodies against MLH1, MSH2, MSH6 and PMS2, as previously described [27]. Briefly, FFPE sections (4 µm thick) were mounted on positive charged glass slides. For antigen retrieval, deparaffinized and rehydrated sections were treated with citric acid buffer (pH 6.0), two cycles of 3 min each at 500 W, followed by inhibition of endogenous peroxidase with 3% H 2 O 2 for 5 min. Then, sections were incubated for 1 h at room temperature with mouse monoclonal anti-MLH-1, anti-MSH2 and anti-MSH6 (clone M1; clone G219-1129; clone 44; Roche, Monza, Italy) and mouse monoclonal anti-human PMS2 (clone 2G5; Novus Biologicals, Milan, Italy). The primary antibodies were visualized using the avidin-biotin-peroxidase complex. Samples were stained more than once, and results were highly reproducible. Staining patterns of MMR proteins were evaluated using normal epithelial, stromal or inflammatory cells, or centers of lymphoid follicles as internal controls. Deficiency of any products of these four MMR proteins was stated as MMR-deficient, while proficient MMR was determined if all MMR proteins were expressed.
MSI analysis was also carried out on all samples. DNA from tumor tissues was analyzed with the Titano MSI kit (Diatech Pharmacogenetics, Jesi, Italy) following the manufacturer's protocol. MSS was defined if no instability at any of the loci was detected, MSI-low (MSI-L) was defined if instability at a single locus was detected and MSI-high (MSI-H) was defined if two or more loci demonstrated instability.

Statistical Analysis
Endpoint for prognostic assessment was OS. OS was defined as the period from the date of diagnosis of metastatic disease to the date of death due to any cause or was censored at the date of last follow-up for alive patients. Differences in baseline characteristics were compared with the use of the chi-square test. Survivals were estimated according to the Kaplan-Meier method and survival curves were compared using the log-rank test. The correlation of baseline characteristics with OS was assessed in univariate analyses: gender (male vs. female), age and ECOG PS at diagnosis of metastatic disease (<65 vs. >65 years; 0 vs. 1-2), metastases onset (metachronous vs. synchronous), number of metastatic sites (1 vs. >1), primary tumor location (left vs. right), presence of liver, lymph node, lung and peritoneal metastases (no vs. yes), MSI/MMR status (MSI-H/MMR-deficient vs. MSS/MMR-proficient), BRAF mutational status (BRAF non-V600 vs. BRAF V600E), mucinous histology (yes vs. no) and tumor grading (2 vs. 3). Variables with a p-value of 0.05 or lower at the univariate analysis were included as covariates at multivariate Cox regression analysis, to examine the effect of baseline prognostic characteristics on OS. Statistical significance was set at p = 0.05. All analyses were performed using MedCalc software version 9.2, MedCalc Software Ltd, Ostend, Belgium.

Patients' Characteristics
Patients' characteristics according to BRAF mutational status are shown in Table 2.

BRAF Prognostic Performance
At a median FU of 13 months, 25 death events were reported in the V600E cohort. Median FU for the non-V600 cohort was 37 months and 4 death events were reported.

BRAF Prognostic Performance
At a median FU of 13 months, 25 death events were reported in the V600E cohort. Median FU for the non-V600 cohort was 37 months and 4 death events were reported.
BRAF V600E mCRCs showed a statistically significant worse mOS when compared to BRAF wild-type mCRCs, for both RAS wild-type (20.  Figure S4).

Discussion
Our study corroborates available evidence concerning incidence and clinical, pathological and molecular phenotypes of BRAF-mutated mCRCs (both V600E and non-V600).
In our study population, non-V600 mutations accounted for 20% of BRAF mutations. Considering the whole cohort (including both BRAF-mutated and BRAF wild-type patients), BRAF non-V600 mutations were reported in 2% of cases. These results are in line with previous reports by Jones et al. [3].
Besides incidence, the clinical-pathological phenotype of both BRAF V600E-and BRAF non-V600-mutated mCRCs reported in our analysis was consistent with available evidence. Notably, BRAF V600E mCRCs occurred more frequently in female and elderly patients, arising mainly from the right colon and being associated with high tumor burden (involving multiple rather than single metastatic sites) and with high differentiation grading. On the other hand, BRAF non-V600 mCRCs developed mainly in male and younger patients, arising more frequently from the left colon and being associated to a low tumor burden (with a single metastatic site, involving mainly the liver) and a low tumor grading. Comparing patients' and tumors' features of V600E and non-V600 cohorts, differences in demographic features (gender and age) did not achieve statistical significance, while left primary tumor location, lower tumor burden and lower tumor grading resulted significantly more frequent in the non-V600 cohort (p = 0.017, 0.026 and 0.045, respectively). Moreover, non-V600-mutated mCRCs displayed more frequently an MSS/MMR-proficient status (p = 0.00016). In 2015, Cremolini et al. [23] showed that BRAF 594/596-mutated mCRCs arose mainly from the rectum, did not develop peritoneal spread and displayed more frequently a non-mucinous histology and a MSS status. In a large, multicenter, retrospective cohort study published by Jones et al. in 2017 [3], BRAF non-V600-mutated mCRCs, as compared to BRAF V600E-mutated, arose more frequently in younger and male patients, displayed mainly a low grade of differentiation and primary tumor location less frequently involved the right side. More recently, Schirripa et al. [25] analyzed clinical, pathological and molecular phenotypes of BRAF-mutated mCRC, according to functional classes based on kinase activity [24]. The analysis reported distinguished features for mCRCs belonging to class 3 and class 2. Indeed, class 2 mutated mCRCs did not diverge from class 1, whereas class 3 mCRCs were more frequently left-sided, with rare lymph node and peritoneal involvement. Moreover, in the analysis by Schirripa et al. [25], class 3 mutated mCRCs did not display any concomitant RAS mutation. The model of class 3 BRAF mutants by Yao et al. [24] postulates that signaling dysregulation by these mutants relies on concomitant mechanisms to support RAS activation: coexisting RAS mutations or NF1 deletions or mutations in melanomas, and RAS activation by aberrant expression or activity of receptor tyrosine kinase in epithelial tumors (lung or CRC). In our analysis, no coexisting RAS mutation was described for class 3 mutants. Therefore, our results are consistent with data by Yao et al. [24] and Schirripa et al. [25] regarding the possible role of receptor tyrosine kinase aberrant expression or activity in supporting RAS activation in class 3 BRAF mutants. Wider genomic profiling would allow deeper insights to address this issue.
Looking at the treatments administered in our patients' population, patients belonging to the non-V600 cohort underwent more frequently to surgical resection of metastases compared to those belonging to the V600E cohort (66.6% vs. 17.4%; p = 0.001912). Interestingly, this data is consistent with the previous retrospective analysis by Cremolini et al., which documented a higher rate of radical resections in patients with BRAF 594/596-mutated mCRCs compared with BRAF V600E mCRCs (60% vs. 12%; p = 0.001) [23]. Anti-EGFRbased treatments were administered in 5 (10.5%) V600E-mutated mCRCs and in 2 (22.2%) non-V600-mutated mCRCs. Best response was disease progression in 80% of the V600E cohort and partial response in 100% of the non-V600 cohort. No difference in activity and efficacy of anti-EGFR agents was observed between class 2 and 3. However, given the small sample size, clear conclusions cannot be drawn.
Concerning survival analysis, mOS was significantly longer in the non-V600 cohort compared to the V600E cohort (61.3 vs. 20.4 months, HR 0.41, 95% CI 0.18-0.93; p = 0.05). We must highlight that the mOS for the BRAF V600E cohort observed in our analysis reached 20.4 months, overcoming historical reports of an average of 12 months [9][10][11][12]. This might be due to the prevalence of cases (56.5%) belonging to the low-risk category according to the BRAF Be Cool score [11]. This class has been associated to a less grim prognosis, with mOS reaching up to 30 months. In spite of this, our analysis confirmed the expected negative prognostic impact of BRAF V600E mutation compared to BRAF wildtype datasets. Indeed, mOS of the BRAF V600E cohort was significantly shorter compared to BRAF wild-type mCRCs, both RAS wild-type and RAS-mutated (p = 0.002 and p = 0.023, respectively). Finally, we should highlight the representativity and the wide sample size of our BRAF wild-type datasets compared to those of previous analyses. Indeed, the paper by Jones et al. (which is the most comprehensive analysis published up to now) reported survival data only for 249 of the total of 9435 patients included in the analysis (2.6%) [3].
We can hypothesize that the prognostic value of BRAF non-V600 mutations could depend on the different distribution of clinical-pathological features between the two cohorts. Indeed, tumors belonging to the non-V600 cohort displayed a statistically significant prevalence of features which are usually associated with good prognosis in the mCRC setting. Primary location was more frequently left-sided (p = 0.017), tumors displayed mainly a low tumor grade (p = 0.045) and presented with a low tumor burden (involving a single metastatic site) (p = 0.026). As a result of limited metastatic disease involving mainly the liver, a high proportion (66.6%) of patients belonging to the non-V600 cohort underwent liver surgery. This unbalance in clinical-pathological features and surgery with radical intent might have impacted on survival. Univariate and multivariate analyses were carried out in order to address this issue. For univariate analyses, besides BRAF non-V600 mutations, other factors positively impacted on survival: age < 65 years, ECOG PS of 0, low tumor burden, absence of lung metastases and low tumor grading. Notably, in accordance with previous evidence, primary tumor location did not appear to have a prognostic value in BRAF-mutated mCRC [11]. For multivariate analyses, only ECOG PS of 0 and age < 65 years retained a positive prognostic impact. Interestingly, ECOG PS is the main prognostic factor included in the simplified BRAF Be Cool score [11].
Up to now, BRAF mutational status has been usually assessed by means of RT-PCR and pyrosequencing. These techniques enable an assessment limited to exons 15 (codons 599, 600 and 601) and 11 (codons 464-469). During the last years, the wide application of diagnostic next-generation sequencing (NGS) allowed a more extensive analysis [29], increasing the identification of BRAF non-V600 missense mutations which would have otherwise been underdiagnosed [3]. Within BRAF missense mutations, a role as a poor prognostic factor and target for molecular treatment [20] pertains only to V600E mutation. Thus, assessment of BRAF mutational status could involve exclusively codon 600 instead of requiring an extensive analysis. Nevertheless, it is important to know the clinicalpathological, prognostic and predictive features of rare non-V600 mutated mCRCs, whose identification is expected to progressively increase, due to the wider application of NGS.
The main limitation of our study relies on both the retrospective nature and on the small sample size, which might have caused selection biases. Thus, prospective confirmation is warranted. Nevertheless, our results are consistent with previous evidence and allow to corroborate and strengthen data available up to now [3,23,25]. Moreover, given the low incidence of BRAF mutations in the setting of mCRC (around 10%) and the even lower incidence of non-V600 mutations (around 2%), it would be difficult to design and carry out prospective analysis to investigate this specific subgroup. Indeed, as previously hypothesized [30], it can be postulated that in the precision oncology era, retrospective analysis might represent a reliable and effective tool to allow an assessment of the prognostic value of rare molecular signatures. In order to limit the biases connected with the retrospective nature and the small sample size, a collaboration within national and international consortia is deemed mandatory and correlations with preclinical data are recommended.

Conclusions
Our results provide a deeper insight on BRAF-mutated mCRC, enlarging evidence on clinical, pathological and prognostic features of this rare entity. Specifically, our results corroborate two hot topics: • BRAF V600E mutation is associated with poor prognosis in the setting of mCRC. • Prognosis of BRAF-mutated mCRC significantly varies depending on the specific missense mutation. Namely, rare non-V600 mutations identify a distinct subgroup of patients with favorable prognosis and peculiar phenotype.
These findings highlight the need for routine evaluation of BRAF mutational status and furthermore, for determining the specific missense mutation, in order to guide clinicians in decision making.