High BRAF Mutation Frequency and Marked Survival Differences in Subgroups According to KRAS/BRAF Mutation Status and Tumor Tissue Availability in a Prospective Population-Based Metastatic Colorectal Cancer Cohort

RAS and BRAF mutations impact treatment and prognosis of metastatic colorectal cancer patients (mCRC), but the knowledge is based on trial patients usually not representative for the general cancer population. Patient characteristics, treatment and efficacy according to KRAS, BRAF and MSI status were analyzed in a prospectively collected unselected population-based cohort of 798 non-resectable mCRC patients. The cohort contained many patients with poor performance status (39% PS 2-4) and elderly (37% age>75), groups usually not included in clinical trials. Patients without available tissue micro array (TMA) (42%) had worse prognostic factors and inferior survival (all patients; 7m vs 11m, chemotherapy-treated;12m vs 17m). The 92 patients (21%) with BRAF mutation had a poor prognosis regardless of microsatellite instability, but receipt of 1-2nd chemotherapy was similar to wildtype BRAF patients. Median survival in this cohort varied from 1 month in BRAF mutated patients not given chemotherapy to 26 months in wildtype KRAS/BRAF patients <75 years in good PS. TMA availability, BRAF mutation and KRAS mutation were all independent prognostic factors for survival. The observed 21% BRAF mutation incidence is higher than the previously and repeatedly reported incidence of 5-12% in mCRC. Screening for BRAF mutations before selection of treatment is relevant for many patients, especially outside clinical trials. A BRAF mutation only partly explained the very poor prognosis of many mCRC patients. Survival in unselected metastatic colorectal cancer patients is extremely variable and subgroups have an extremely short survival compared to trial patients. Patients without available TMA had worse prognostic factors and shorter survival, which questions the total generalizability of present TMA studies and implies that we lack information on the biologically worst mCRC cases. Lack of available tissue is an important underexposed issue which introduces sample bias, and this should be recognized more clearly when conclusions are made from translational mCRC studies.


Introduction
In metastatic colorectal cancer (mCRC) trials, median survival now regularly exceeds 20 months and approaches 30 months [1][2][3][4]. Molecular and gene analyses on tumor tissue give prognostic and predictive information. Screening for mutations in RAS and BRAF oncogenes is used in clinical practice and impacts treatment selection. Mutations in the KRAS codon 12 of exon 2 are strongly predictive for lack of benefit from EGFR-inhibition [1][2][3]. Rare RAS mutations also indicate no benefit from EGFR-inhibition [4]. Mutations in the BRAF oncogene have been found in 5-12% of mCRC patients (S1 Table). The single substitution missense mutation V600E accounts for more than 95% of BRAF mutations. The BRAF mutation (mutBRAF) indicates a poor prognosis with median survival of 10 months in trial patients [2,3,[5][6][7][8][9][10]. The negative prognostic impact of mutBRAF after primary tumor resection in colorectal cancer seems to be restricted to patients with microsatellite stable tumors [11][12][13].
The molecular studies behind these results are mainly based on tumor specimens from patients included into trials. However, trial patients are selected and usually not representative for the general cancer population. In recent cancer registry data, median survival for mCRC patients was only 10 months, although it had increased compared to earlier years [14][15][16][17]. Patients included into trials are from sub-populations with better prognostic factors as they are younger, have better performance status (PS), less co-morbidity and better baseline prognostic factors [18]. Furthermore, population-based studies have shown that approximately one third of mCRC patients do not receive palliative chemotherapy at all [14,15,18,19]. Populationbased data on incidence, prognostic and predictive effects of BRAF and KRAS mutations in mCRC patients are therefore needed.
In the present study we analyzed the prognostic and predictive value of BRAF/KRAS mutations and tumor tissue availability in patients with non-resectable mCRC included in a prospective population-based study [18]. This cohort represents an unselected population of all non-resectable mCRC patients in a defined area, including patient groups usually not present in clinical trials and where molecular and clinical data are very scarce; patients with poor PS, elderly patients and patients not given chemotherapy. The aim of the study was: i: To describe the true incidence of BRAF mutations in unselected mCRC patients, ii: To investigate if BRAF mutations alone could explain the very poor prognosis for mCRC patients with poor PS, iii: To characterize patients without available tumor tissue for molecular analyses, to decide generalizability of prior TMA studies, and iv: To study heterogeneity of survival for unselected mCRC patients in comparison to selected trial patients.

Patient cohort
Prospective registration of patients with non-resectable, histologically confirmed mCRC referred to the oncology units of three university hospitals in Scandinavia was performed between October 2003 and August 2006. These hospitals received all oncology department referrals of their administrative area. Uppsala University Hospital in Sweden served a region of 280 000 inhabitants, Odense University Hospital in Denmark 475 000 inhabitants and Haukeland University Hospital in Norway 450 000 inhabitants. Clinical characteristics, blood tests, treatments, response to treatment, progression-free survival and overall survival were recorded. The cohort was originally set up with the intent to study trial inclusion. The cohort was later expanded to also include mCRC patients not seen at the oncology departments using regional cancer registries checking CRC deaths during October 2003-December 2008. Patients were included if initial mCRC diagnosis was between October 2003-August 2006. The present cohort therefore represents an unselected population of all non-resectable mCRC patients in a defined area. Baseline variables were from the patients' first consultation at the oncology department or at the diagnosing health care unit if not referred. Combination chemotherapy was irinotecan or oxaliplatin with a fluoropyrimidine. Written informed consent was given by all participants seen at the clinics. Patient information was anonymized and de-identified prior to analysis. The study was approved by the ethical committees in Norway (Regional Committee for medical and health research ethics-REC West), Sweden (Regional Ethical Committee Uppsala) and Denmark (The Regional Scientific Ethical Committees for Southern Denmark).

Tissue retrieval and tissue microarray generation
Paraffin embedded tissue blocks of the primary tumor or metastatic lesion were retrieved and corresponding hematoxylin-eosin stained glass slides were examined. Representative parts of the tumor were selected for tissue microarray (TMA) generation and DNA extraction. Tissue cores of 1 mm in diameter were extracted from the original block and transferred to a recipient paraffin block (for TMA construction) or to a clean vial for DNA extraction, after comparison with the marked HE-stained slide. On the glass slide, areas with viable tumor with as little as possible other cells admixed were marked. Care was taken to insure the correct alignment of the glass slide with the paraffin block [20]. TMAs were generated using a TMArrayer and the Beecher Instruments Manual Tissue Arrayer MTA-1.

Mutation and MSI analysis
DNA was extracted from the tissue cores using Recoverall Total Nucleic Acid Isolation (Ambion, Austin, Texas, USA). KRAS and BRAF pyrosequencing mutational analysis was performed according to the manufacturer's protocol for the PyroMark Q24 (QIAGEN GmbH, Hilden, Germany) and the use of PCR primers previously described for KRAS codon 12/13 and BRAF codon 600 [21]. Ten ng genomic DNA was used for each PCR reaction. The pyrosequencing analysis was performed as described [22]. Sequencing primer for KRAS codon 12/13 was 5´-AACTTGTGGTAGTTGGAGCT-3´and for BRAF codon 600 5´-TGATTTTGGTCTA GCTACA-3´. MSI status was determined using MSI Analysis System, version 1.

Statistics
Group comparisons were performed using the exact chi-square tests for dichotomous or nominal variables, the exact Mann-Whitney tests for ordinal variables, and Student's t-tests for continuous variables. Binary and ordinal logistic regression was used for dichotomous and ordinal outcome variables, respectively. Results are reported as odds ratios (OR) and 95% confidence intervals (CI). Overall survival time was the interval from the date of non-resectable metastatic disease to the date of death or censored if the patient was alive on February 4th 2014, using the Kaplan-Meier method with log-rank tests for univariate and Cox regression for multivariate analyses. For the multivariate survival analyses, backward stepwise selection of covariates excluding the most non-significant variable at each step according to the likelihood ratio test was used. Variables included in the initial multivariate survival analyses but not shown to be independent factors were: age, gender, smoking habits, study inclusion and location of metastases. Results are reported as hazard ratios (HR) and 95% confidence interval (CI). All analyses were performed with the statistical program SPSS v22. All statistical tests were two tailed, using the 5% significance level.

Results
The study finally included 798 patients after exclusion of eleven patients due to re-evaluation or autopsy confirming other cancer than CRC (6 patients) or non-metastatic disease (5 patients). Forty-nine patients (6%) were never seen at the oncology departments. At last follow-up on February 2014, 24 patients (3%) were alive. First-line chemotherapy was given to 456 patients (57%), 77% of these received combination chemotherapy. Bevacizumab was given to 50 patients (11%) and EGFR-inhibitor to 111 patients (24%). Median survival (95%CI) for all 798 patients was 9 months (7.8-10.2), 15 months (13.4-16.6) for chemotherapy-treated patients and 2 months (1.5-2.5) for patients given best supportive care only. Poor PS or high age were the main reasons for not receiving chemotherapy. Tumor blocks containing tumor tissue were collected in 701 (88%) of 798 cases (Fig 1). For the remaining 97 cases, the tumor tissue blocks did either not contain cancer tissue (typically severe dysplasia in a colon/rectum tumor biopsy with metastatic disease and elevated CEA) or the blocks were displaced in the archive. TMA was possible to produce from 462 of 701 collected cases (58% of initial 798 cases), whereas the remaining 239 patients had biopsies being too small or containing necrotic tissue. TMA was almost exclusively made from the primary tumor; 5 were made from metastatic lesions. Gene analyses for BRAF failed in 16 patients and eventually 446 patients had reliable BRAF analyses and 442 patients had both BRAF and KRAS analyses. Among patients with evaluable BRAF status, 277 patients (62%) received chemotherapy.

TMA
Patient characteristics and treatment were significantly different in patients with available TMA, compared to patients where TMA was not possible to produce (Table 1). Patients without available TMA had worse prognostic factors; higher age, primary tumor less resected, more often synchronous metastatic disease, worse PS, more metastatic sites, more often weight loss and anorexia and more often abnormal baseline prognostic blood tests with elevated leucocytes, platelets and alkaline phosphatase [23]. Independent baseline characteristics by logistic regression for patients with available TMA vs. non-available TMA were metastatic sites >1: (OR 0.5, 95% CI 0.4-0.7, p<0.001), synchronous disease (OR 0.5, 95%CI 0.3-0.8, p<0.001), colonic primary (OR 3.6, 95%CI 2.2-6.0, p<0.001) and distant metastatic lymph nodes (OR 2.4, 95%CI 1.3-4.2, p = 0.002). Patients without available TMA less often received chemotherapy (50% vs 63%, p< 0.001) and when given, combination chemotherapy was less used (39% vs 48%, p = 0.0014). Furthermore, fewer were included into clinical trials and fewer received secondary surgery (Table 1). Response rate to 1 st -line chemotherapy was similar, whereas progression-free survival (PFS) was borderline significantly shorter (6.3 vs 8.0 months, p = 0.06). Median survival was inferior in patients without available TMA compared to patients with TMA in all patients (7 vs 11 months, log-rank p<0.001) and in chemotherapy-treated patients (12 vs 17 months, log-rank p<0.001) (Fig 2). Availability of tissue for TMA was an independent prognostic factor for survival in the multivariate analysis (p = 0.01, Table 2).

BRAF mutation
The observed 21% mutBRAF incidence is clearly higher than the previously and repeatedly reported incidence of 5-12% in mCRC (S1 Table). Screening for BRAF mutations before selection of treatment strategy is therefore relevant for many patients, especially in patients cared for outside clinical trials. As our study reveals that patients without available TMA have a much worse prognosis and poor patient characteristics, one could speculate that many of these patients could harbor a BRAF mutation. The true incidence of mutBRAF in the general mCRC population might therefore be even higher than the observed 21%. New treatment options for patients with mutBRAF are evolving as use of FOLFOXIRI in first-line treatment seems to increase survival and combinations of EGFR-inhibitors and BRAF/MEK-inhibitors are promising [24,25]. If mutBRAF is more frequent than anticipated, upfront BRAF testing before 1 st -line palliative chemotherapy becomes increasingly relevant. Ethnic differences in mutBRAF incidences have been reported [26]. However, mutBRAF does not seem to have a higher incidence in Scandinavians as among the 56 Nordic 7 trial patients in this cohort [2], the incidence of mutBRAF was 12% as in other populations. Although the frequency of mut-BRAF in CRC in the Cancer Genome Atlas Network was only 3% [27], population-based studies on mainly stage II-III colorectal or colon cancer patients have found a 12-28% mutBRAF incidence [28][29][30][31]. MutBRAF is mainly found in colon tumors, and the relative distribution of colon vs rectal primary in mCRC studies could affect the reported incidence of mutBRAF. The incidence of mutBRAF was 27% in colonic primaries in our study, still higher than the 10-18% reported by other mCRC studies (S1 Table). The BRAF assay used in our study has a 2-5% limit of detection of mutated DNA and is more sensitive than Sanger sequencing. Many different assays are used among prior stage IV CRC studies, but still all report mutBRAF frequencies ranging from 5-12% (S1 Table). We thus believe the higher frequency observed has more to do with the unselected cohort than the method used. Clinical and molecular data based on trial patients alone will not include data on subgroups with poor performance status, elderly and non-treated patients. The present cohort included patients with poor performance status (39% PS 2-4) and elderly (37% age >75), groups underrepresented in clinical trials. Our results, which show a considerable discrepancy in the frequency of BRAF mutations among mCRC patients compared with results from prospective clinical trials, clearly underline the continuous need for observational cohort studies/studies based on registries. As in prior studies, patients with mutBRAF tumors had more often proximal colon tumors, higher age, female gender, poor PS, multiple metastatic sites, distant lymph node metastasis and synchronous disease but less often lung metastases [6,7,10,[32][33][34][35][36]. Two novel observations in our study are that patients with mutBRAF tumors more often had weight loss and anorexia, whereas baseline prognostic serum markers as elevated alkaline phosphatase were not associated with mutBRAF. Different underlying mechanisms may therefore lie behind the prognostic role of these markers. Although a higher incidence of mutBRAF has been reported among smokers and a causal relationship suggested [37,38], we did not observe this similar to a recent study [29]. In our study, patients with mutBRAF tumors received 1 st and 2 nd line therapy comparable to wtBRAF patients as prior observed [10]. However, they received significantly less 3 rd line chemotherapy and only one patient had secondary surgery. mutBRAF tumors are only present in 0-2% in mCRC lung or liver resections [33,[39][40][41]. There seems to be a progressive selection of patients with mutBRAF; 21% or more in the general population as reported here for the first time, 5-12% in trial patients and 0-2% in patients undergoing liver or lung resections. Response to chemotherapy in mutBRAF patients was similar and PFS on 1st-line chemotherapy was only borderline significantly worse. Overall survival was significantly shorter in mutBRAF patients, but varied from 1 month in patients receiving only best supportive care to 14 months in PS 0-1 patients younger than 75 years receiving chemotherapy. Most prior studies have shown a shorter PFS for mutBRAF patients [1,2,5,9,42,43], but some report similar PFS [3,21,44]. Trial patients with mutBRAF have repeatedly had a shorter median survival, usually around 10 months (S1 Table). Initially we hypothesized that mutBRAF could explain the very poor prognosis for many subgroups of mCRC, e.g. in young patients with poor PS and not given chemotherapy. However, as mutBRAF only tended to be more frequent among these patient groups, mutBRAF can only partly explain the very poor prognosis among mCRC patient subgroups. Other candidate genes and molecular factors must be sought. In our study we found that 33% of mutBRAF tumors were MSI-H, which is in accordance with the 29% observed by others [33]. An adverse prognosis for CRC (mainly stages I-III) has been reported with presence of MSS with mutBRAF, but not for MSI-H [11][12][13]. In our study MSI-status did not affect prognosis in mutBRAF stage IV CRC patients, in accordance with other studies [9,35].

KRAS mutations, wildtype KRAS/BRAF
It has been recommended that mCRC cohorts should be divided into 3 groups: mutBRAF, mut (K)RAS and wtBRAF/(K)RAS [3,5,8,10]. In most studies KRAS mutations have not been prognostic. However, the comparisons have often been done with a mixed wtKRAS population including mutBRAF. In our study, survival was similar comparing all wtKRAS/BRAF patients to all mutKRAS patients, regardless of treatment. However, in patients aged<75 years with good PS, being the target group for intensive chemotherapy, survival was significantly longer for wtKRAS/BRAF patients (26m vs 18m). Other groups have also found an inferior survival for mutKRAS compared to wtKRAS/BRAF [5,8,10]. Maugham et al, found a median survival of 14.4 months in mutKRAS mCRC patients, compared to 20.1 months in wtKRAS/BRAF patients [5]. Our observed survival of 26 months in a fit mCRC population (similar to trial patients) is similar to a recent study, where median survival was 28 months in wtRAS/BRAF patients [3].

TMA
In the present study, we could produce TMA from 462 patients (58%) of initially 798 cases, and BRAF analyses for 446 patients (56%). Similar incident CRC case numbers and retrieval rates have been reported in molecular epidemiology studies embedded within other large cohorts. In two US studies on 901 and 1255 CRC cases, respectively, tissue specimens were retrieved from 58% and BRAF status available for 42% and 45% [30,31,38]. In a Swiss register on 1420 CRC patients, resection specimens were retrieved from 42% and BRAF analyses possible for 26% [28]. In these studies, no significant differences in characteristics were observed between patients whose tissue specimens could be retrieved or not, however only few prognostic factors were compared and a minority of the patients had stage IV disease. Among our mCRC patients, we found large differences as patients without available TMA had worse prognostic factors, received less treatment and had a shorter survival. Even for patients receiving chemotherapy, survival was shorter in patients without available TMA. In our study 91% of patients with a resection of the primary had available TMA compared to 46% of patients without a resection. In most instances, TMA cores can be sampled more easily utilizing surgical resection samples, whilst biopsy specimens, especially if small or necrotic, often do not yield enough tissue for processing. It is likely that surgery is less often performed in patients with higher tumor burden where PS is often affected. Our results question the generalizability of present TMA studies, and imply that we lack information on the biologically worst mCRC cases. In a recent study on 6155 randomized mCRC trial patients, tumor samples and BRAF analyses were available from 3063 patients (50%) [9]. The authors state that lack of available tissue is an important underexposed issue which may introduce sample bias in translational studies in mCRC. Our study clearly underlines this. To ensure the representativity of molecular studies more information is needed on patients without available TMA. Liquid biopsies with cell-free DNA analysis may in the future be a way to obtain molecular data from this large undescribed subgroup of mCRC patients.
In conclusion, BRAF mutations are more frequent in the general mCRC population than prior anticipated. Screening for BRAF mutations before selection of treatment strategy is therefore relevant for many patients, especially outside clinical trials. Our study shows a heterogenic median survival in the mCRC population and subgroups have an extremely short survival compared to trial patients. MutBRAF only partly explains the very poor prognosis of some mCRC patients and additional underlying molecular mechanisms may be involved. Patients without tumor tissue available for TMA and molecular analyses have more frequently poor prognostic characteristics, respond less to treatment and have a worse survival. Lack of available tissue is an important underexposed issue which introduces sample bias, and this must be recognized more clearly when conclusions are made from translational studies of mCRC trial patients.
Supporting Information S1