Impact of circulating tumor DNA mutant allele fraction on prognosis in RAS-mutant metastatic colorectal cancer Elena Elez1,2, Chiara Chianese3 , Enrique Sanz-Garcıa1,2, Erica Martinelli4, Alba Noguerido1,2, Francesco Mattia Mancuso3, Ginevra Caratu  3, Judit Matito3, Julieta Grasselli1,5, Claudia Cardone4, Riziero Esposito Abate , Giulia Martini , Cristina Santos5, Teresa Macarulla1,2, Guillem Argile 6 1,2 s1,2, 1,2 1,2 1,2,5 7 6 Jaume Capdevila , Ariadna Garcia , Nuria Mulet , Evaristo Maiello , Nicola Normanno , Frederick Jones8, Josep Tabernero1,2, Fortunato Ciardello4, Ramon Salazar5 and Ana Vivancos3 1 Department of Medical Oncology, Vall d’Hebron Institute of Oncology, Barcelona, Spain 2 Department of Medical Oncology, Vall d’ Hebron University Hospital, Universitat Autonoma de Barcelona, Spain 3 Cancer Genomics Group, Vall d’Hebron Institute of Oncology, Barcelona, Spain 4 Medical Oncology, Department of Clinical and Experimental Medicine ‘F. Magrassi’, Universit a della Campania ‘L. Vanvitelli’, Napoli, Italy 5 Department of Medical Oncology, Catalan Institute of Oncology, Universitat de Barcelona, L´Hospitalet, Spain 6 Cell Biology and Biotherapy Unit, Istituto Nazionale Tumori ‘Fondazione Giovanni Pascale’ IRCCS, Napoli, Italy 7 Medical Oncology, IRCCS Casa Sollievo della Sofferenza, San Giovanni Rotondo (Foggia), Italy 8 Sysmex Inostics, Mundelein, IL, USA Keywords Despite major advances in the treatment of metastatic colorectal cancer circulating tumor DNA; MAF; metastatic (mCRC), the survival rate remains very poor. This study aims at exploring colorectal cancer; prognostic biomarker; the prognostic value of RAS-mutant allele fraction (MAF) in plasma in RAS analysis mCRC. Forty-seven plasma samples from 37 RAS-mutated patients with Correspondence nonresectable metastases were tested for RAS in circulating tumor DNA A. Vivancos, Cancer Genomics Group, Vall using BEAMing before first- and/or second-line treatment. RAS MAF was d’Hebron Institute of Oncology (VHIO), correlated with several clinical parameters (number of metastatic sites, hep- Cellex Building, C/ Natzaret 115-117, atic volume, carcinoembryonic antigen, CA19-9 levels, primary site location, Barcelona, 08035, Spain and treatment line) and clinical outcome [progression-free survival (PFS) and Tel: +34 93 2543450 Ext. 8647 overall survival (OS)]. An independent cohort of 32 patients from the E-mail: avivancos@vhio.net CAPRI-GOIM trial was assessed for clinical outcome based on plasma base- Chiara Chianese and Enrique Sanz-Garcıa line MAF. RAS MAF analysis at baseline revealed a significant correlation contributed equally. with longer OS [Hazard ratios (HR) = 3.514; P = 0.00066]. Patients with Ramon Salazar and Ana Vivancos are co- lower MAF also showed a tendency to longer PFS, although not statistically senior authorship. significant. Multivariate analysis showed RAS MAFs as an independent prognostic factor in both OS (HR = 2.73; P = 0.006) and first-line PFS (Received 21 October 2018, revised 11 May (HR = 3.74; P = 0.049). Tumor response to treatment in patients with higher 2019, accepted 15 July 2019, available MAF was progression disease (P = 0.007). Patients with low MAFs at base- online 31 July 2019) line in the CAPRI-GOIM group also showed better OS [HR = 3.84; 95% doi:10.1002/1878-0261.12547 confidence intervals (CI) 1.5–9.6; P = 0.004] and better PFS (HR = 2.5; 95% CI: 1.07–5.62; P = 0.033). This minimally invasive test may help in adding an independent factor to better estimate outcomes before initiating treatment. Further prospective studies using MAF as a stratification factor could fur- ther validate its utility in clinical practice. Abbreviations CEA, carcinoembryonic antigen; cfDNA, circulating free DNA; ctDNA, circulating tumor DNA; HR, hazard ratio; MAF, mutant allele fraction; mCRC, metastatic colorectal cancer; OS, overall survival; PFS, progression-free survival. Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1827 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. RAS MAFs in metastatic colorectal cancer E. Elez et al. MAF on OS and progression-free survival (PFS) in 1. Introduction these disease settings. We also included an indepen- The past two decades have witnessed significant pro- dent cohort from the CAPRI-GOIM trial that was gress in the treatment of metastatic colorectal cancer assessed for clinical outcome based on plasma base- (mCRC) partly due to a better selection of therapy line MAF (Ciardiello et al., 2014; Normanno et al., based on the tumor RAS mutation status. Nonethe- 2017). less, the 5-year survival rate in mCRC patients remains poor (Siegel et al., 2017). The large variability in sur- 2. Materials and methods vival shows that current routine prognostic evaluation of mCRC is insufficient and needs to be improved, for both resectable and nonresectable metastases. The 2.1. Study design development of reliable prognostic biomarkers is an increasingly pertinent tool in this setting. This multicentric study included both retrospective In mCRC, the detection of circulating tumor DNA and prospective patients: Retrospective patients were (ctDNA) is an emerging alternative to detect muta- recruited from two Spanish hospitals (Vall d’Hebron tions, thus avoiding biopsies from primary or meta- University Hospital and Catalan Institute of Oncology, static sites. We and others previously reported ~ 90% Duran I Reynals); prospective patients were recruited concordance of RAS-mutant status in paired plasma from the Vall d’Hebron University Hospital only. and tissue samples, as well as its predictive value in Additionally, an independent cohort of first-line plasma for anti-EGFR therapy response (Bettegowda patients derived from the CAPRI-GOIM trial (regis- et al., 2014; Grasselli et al., 2017; Siravegna et al., tration number: 2009-014041-81) were also included. 2015). Mutant allele fractions (MAFs) are a measure The study was approved by the ethics committee of all of the percentage of mutant alleles within the totality hospitals, and all patients signed written informed con- of alleles in any given sample. MAF estimations of dri- sent. This study was conducted in accordance with the ver genes have shown important clinical implications Declaration of Helsinki and Good Clinical Practice in various settings. In a retrospective analysis of the guidelines. CRYSTAL trial (Van Cutsem et al., 2011), mCRC patients with tumor RAS MAFs between 0.1% and 2.2. Patient characteristics < 5% were more likely to benefit from the addition of cetuximab to FOLFIRI. Likewise, resistance to anti- Patients from the TTD ULTRA clinical trial EGFR therapies in mCRC with KRAS MAFs < 1% (NCT01704703) were included. Of 110 mCRC plasma (Azuara et al., 2016; Laurent-Puig et al., 2015) and samples screened, 62 (56%) were identified as RAS- longer benefit with tyrosine kinase inhibitor therapy mutated by BEAMing in plasma. To obtain a homoge- were associated with higher MAFs in EGFR-mutated neous study population, we excluded patients with lung cancer patients (Ono et al., 2014; Zhou et al., liver-limited resected metastases, leaving 41 plasma 2011). samples (37%) from 37 RAS-mutated patients with The potential prognostic value of plasma MAFs in nonresectable metastases for analysis (Fig. S1); of mCRC has not been well established yet. Interestingly, them, 29 samples were prior to first-line therapy and we and others have observed that RAS MAF showed 12 prior to second-line treatment (Table S1). Baseline a trend to lower overall survival (OS) when plasma characteristics, number and location of metastasis, and levels were above a cutoff of 10% and 1%, respec- number and description of previous lines of therapy tively, although the population was heterogeneous in are summarized in Table 1. terms of treatment and time of analysis (El Messaoudi The CAPRI-GOIM trial, a nonprofit academic, et al., 2016; Siravegna et al., 2017; Vidal et al., 2017). open-label, multicenter study, enrolled 340 mCRC Of note, plasma was obtained at different disease patients, KRAS exon-2 wild-type, according to local stages and timing on treatment. pathology assessment, treated with FOLFOX plus To better define the predictive nature of RAS cetuximab vs FOLFOX at progression to first-line MAF levels, we performed a study in a homogeneous FOLFIRI plus cetuximab (Eudract number: 2009- group of patients with plasma samples collected sys- 014041-81) (Normanno et al., 2017). Of these, 33 tematically prior to the first or second treatment line, patients were found mutated according to their plasma to correlate RAS-mutant MAFs with clinical parame- sample and thus used in this study as an independent ters and to determine the impact of RAS-mutant validation set (Table S1) (Normanno et al., 2017). One 1828 Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Elez et al. RAS MAFs in metastatic colorectal cancer Table 1. Patient characteristics. 2.5. Mutation detection by BEAMing technology First-line Second-line CAPRI-GOIM in ctDNA (N = 29; (%) (N = 12; %) (N = 33; %) RAS status was determined in plasma using BEAMing Gender (Sysmex Corporation, Kobe, Japan). The commercially Male 19 (65) 7 (58) 16 (48) available, and previously validated (Grasselli et al., 2017), Female 20 (35) 5 (42) 17 (52) RAS-mutated CE-IVD BEAMing RAS plasma panel of mutations was KRAS 12 16 (55) 7 (58) 19 (58) evaluated (Table S2). Plasma was processed as previously KRAS 13 6 (21) 3 (25) 2 (6) described (Grasselli et al., 2017). Samples were considered KRAS (others) 3 (10) 1 (8) 7 (21) mutant according to a mutation rate threshold (0.02– NRAS 12 1 (4) 0 2 (6) 0.04%) based on the CE-IVD BEAMing RAS panel NRAS 13 0 0 0 assay, as per the manufacturer’s algorithm. NRAS (others) 3 (10) 1 (8) 3 (9) M1 metastatic sites 1 10 (35) 1 (8) 17 (52) 2.6. Statistics 2 16 (55) 7 (58) 14 (42) Statistical analyses were performed using R 3.4.1, R 3+ 3 (10) 4 (33) 2 (6) studio (v. 1.0.153, https://www.r-project.org/), and the Primary site CRAN R survival package. Data are summarized by fre- Right 9 (31) 5 (42) 7 (21) Left 12 (41) 0 16 (48) quency for categorical variables and by median and Rectum 8 (28) 7 (58) 10 (31) range for continuous variables. PFS was defined as the Treatment time from treatment start to disease progression or FOLFOX 26 (89) 3 (25) death. OS was defined as the time from mCRC diag- FOLFIRI 1 (4) 7 (58) nosis to death from any cause or the last follow-up Antiangiogenics 9 (31) 5 (42) visit. Response rate was assessed according to RECIST Othersa 3 (10) 2 (17) 1.1 (https://recist.eortc.org/). a First-line: 5-fluorouracil (5-FU) and capecitabine; Second-line: Mutant allele fractions were calculated as the num- irinotecan. ber of mutant beads divided by the total number of beads analyzed, and all samples were analyzed blinded patient was excluded for analysis due to lack of fol- to the study endpoints. Pearson’s correlation coeffi- low-up data. cients between MAF levels and selected clinical vari- ables were determined. Clinical variables analyzed 2.3. Sample collection included treatment line, primary site, number and loca- tion of metastatic sites, best response, carcinoembry- Blood samples (4 mL) were collected in CellSave onic antigen (CEA) and CA 19-9 levels, number of Preservative Tubes (Menarini-Silicon Biosystems, metastatic hepatic lesions, and hepatic lesion volume Bologna, Italy), and plasma was isolated within 48 h. (sum of the largest diameter of all hepatic lesions For nontrial patients, 10 mL of blood was collected in [maximum 10], according to RECIST v1.1). Significance EDTA tubes and plasma was isolated within 1 h. A was determined with nonparametric Kruskal–Wallis two-step centrifugation was performed with blood ini- tests; P-values < 0.05 were considered significant. tially centrifuged for 10 min at 1600 g at room tem- Hazard ratios (HR) and 95% confidence intervals perature. Supernatant was collected, avoiding the (CI) were calculated. Survival curves were estimated buffy coat, and then centrifuged again for 10 min at using the Kaplan–Meier method. Log-rank tests, room temperature at 3000 g to remove remaining cells. including univariate and multivariate Cox proportional Plasma supernatant was transferred into a 1.5-mL tube hazards models, were performed for key endpoints. and stored at 80 °C until use. 2.7. MAF cutoff 2.4. DNA purification The optimal MAF cutoff of 5.8% used in our cohort Circulating free DNA (cfDNA) was performed with was calculated based on our dataset using the R function the QIAamp Circulating Nucleic Acid Kit (QIAGEN, CUTP in the SURVMISC package (Contal and O’Quigley, Venlo, Netherlands) according to the manufacturer’s 1999; Mandrekar et al., 2003). This function determines instructions. DNA quality and concentration were the optimal cut point for a continuous variable in a measured with a NanoDrop 1000 Spectrophotometer coxph or survfit model under the null hypothesis that (Thermo Scientific, Waltham, MA, USA). the chosen cutoff does not predict survival. Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1829 RAS MAFs in metastatic colorectal cancer E. Elez et al. Using the same optimized cutoff of 5.8%, samples 3. Results with MAF < 5.8% showed significantly better OS (Fig. 2B). The difference in median OS between patients with RAS-mutant samples having MAF 3.1. Correlation of MAF with clinical parameters < 5.8% (N = 10) and those with MAF ≥ 5.8% was A total of 41 samples from 37 patients were ana- 26.7 months vs 11.4 months (HR: 3.5; 95% CI: 2.08– lyzed, 29 prior to first-line and 12 prior to second- 43.1; P = 0.0006). RAS MAF was still an independent line treatment. A wide range of plasma RAS MAFs variable for OS with a 1% cutoff, but at 10% cutoff, was seen for both the first- and second-line treatment HR was no longer significant (Fig. S3A–D). groups. Median MAF was 9.9% in the first-line In the second-line setting, the analyses show clearly group (range: 0.014–51.5%) and 1.8% (range: 0.03– that patients with MAF < 5.8% have both longer PFS 52.4%) in the second-line group (Fig. 1A), although and OS, with an HR of 6.6 and P = 0.00018 for both not statistically significant. MAF distribution did not variables, compared to those with MAF > 5.8% correlate with tumor right/left sidedness or the num- (Fig. S4A,B). RAS MAF remained an independent ber of metastatic sites (Fig. 1B,C) but, in contrast, variable with cutoffs of 1% or 10% (Fig. S4C,F). varied significantly according to the site of metastasis. Mutant allele fractions were significantly higher in MAF median values were calculated according to patients whose outcome was progression disease (PD), metastatic spread involving the liver, lung, lymph compared to those with partial response (PR; Fisher’s nodes, or peritoneum (most patients had more than test P = 0.002) or stable disease (SD; Fisher’s test one metastatic site). Median MAF was significantly P = 0.014; Fig. 2C). One-way ANOVA test draws lower in patients with metastases in the peritoneum identical conclusions (P = 0.007). compared to those with metastases in the liver Univariate analyses in the first-line cohort including (P = 0.0003), lung (P = 0.044), or lymph nodes different clinical factors such as tumor location, num- (P = 0.025; Fig. 1D), although this observation is lim- ber of metastatic sites, gender, and CEA levels showed ited by sample size. plasma MAF was the only statistically significant A multivariate Cox analysis showed that MAFs did prognostic factor for OS. Multivariate Cox analysis not significantly correlate with CA19-9 or CEA levels, considering the previous biomarkers showed that though a slight tendency to higher levels of CA19-9 plasma RAS MAF was the strongest prognostic factor was observed in samples with higher MAFs. CEA for both PFS (HR: 3.74; 95% CI 1.01–13.92; levels were overall higher than reference values; how- P = 0.049) and OS (HR: 2.73; 95% CI 2.35–182.53; ever, no association was observed with MAFs P = 0.006; Table 2). (Fig. S2, Table S3). Similarly, MAFs did not correlate Consistent with our results, in an independent with the number of metastatic hepatic lesions or hep- cohort at first-line treatment from the CAPRI-GOIM atic volume (Fig. S2). trial, samples with MAF < 5.8% showed significantly better OS (HR: 3.84; 95% CI 1.5–9.6; P = 0.004) and longer PFS (HR: 2.5; 95% CI: 1.07–5.62; P = 0.033; 3.2. Correlation of MAF with clinical outcomes Fig. 3). Similar results were obtained when using a Overall, higher MAF values correlated with shorter cutoff of 10% (Fig. S5). PFS (cor = 0.476; P = 0.009) and OS (cor = 0.506; P = 0.005; Fig. S3E). We decided to set a cutoff MAF 4. Discussion value that split patients with better vs worse prognosis in our cohort. This was done with the Cutpoint func- This is the first clinical study that aims at specifically tion (cutp) for a continuous variable in a Coxph or assessing the prognostic potential of measuring RAS Survfit model. MAFs in cfDNA in a homogenous group of mCRC In the first-line setting, using an optimized MAF patients. Previously, we and others observed that cutoff of 5.8%, PFS was not significantly better in patients with lower OS tend to have a plasma RAS samples with MAF < 5.8% (Fig. 2A); however, a MAF above 10% (El Messaoudi et al., 2016; Gras- trend toward lower PFS was observed in samples with selli et al., 2017; Vidal et al., 2017). The present a higher MAF. The difference in median PFS between study aimed to accurately define the impact of RAS patients with RAS-mutant samples with MAF < 5.8% MAF in a homogenous cohort of mutated mCRC (N = 10) and those with MAF ≥ 5.8% was patients, thereby excluding potential confounding fac- 10.7 months vs 7.0 months with an HR of 2.2 (95% tors, in the context of specific clinical parameters and CI: 0.94–7.20; P = 0.06). survival outcomes. We correlated RAS MAF with 1830 Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Elez et al. RAS MAFs in metastatic colorectal cancer A 60 P = 0.374 B P = 0.329 60 50 50 40 40 MAF (%) MAF (%) 30 30 20 20 10 10 0 0 1st Line 2nd Line right left/rectum Treatment line Primary site C D P = 0.323 P = 0.092 P = 0.0003 (***) 60 60 P = 0.044 (*) 50 50 40 40 P = 0.025 (*) MAF (%) MAF (%) 30 30 20 20 10 10 0 0 1 2 3 Liver Lung Lymph nodes Peritoneum Number of metastatic site distributions Type of metastatis Fig. 1. MAF distribution. Representation of MAF (%) distributions according to: (A) the two lines of treatment; (B) tumor laterality; (C) number of metastatic lesions; and (D) metastatic site. Box plots show the interquartile range (IQR) with median, 25th and 75th percentile, outliers, and P-values. Continued lines (in graph D) indicate the comparison between two variables. Statistically significant P-values are marked with a (*). Samples are represented by light blue dots. several clinical characteristics including previously multivariate analysis revealed that tumor sidedness proposed prognostic biomarkers, such as laterality, was not a prognostic factor in our cohort. This appar- CEA, CA19.9, hepatic tumor volume, number of ent discordance with previous publications might be metastatic sites, and previous lines of therapy, to gain accounted for by a bias concerning the study popula- a better understanding of the biological basis of tions; our cohort is relatively small and included only plasma MAFs in mCRC patients. However, no linear RAS-mutated mCRC samples, whereas other studies correlations were found with any of these parameters were based on patients with RAS wild-type mCRC —an outcome which warrants further research with (Arnold et al., 2017) or did not take RAS mutational expanded cohorts. status into account (Petrelli et al., 2017). To date, there is plausible evidence that the primary While MAF distribution was independent of hepatic tumor side might have prognostic value in mCRC tumor volume and the number of metastatic hepatic (Arnold et al., 2017; Petrelli et al., 2017). Our lesions, the presence of metastases in the liver, lung, or Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1831 RAS MAFs in metastatic colorectal cancer E. Elez et al. 1.0 1.0 A B MAF < 5.8% (n = 10) MAF < 5.8% (n = 10) MAF > 5.8% (n = 19) MAF > 5.8% (n = 19) 0.4 0.6 0.8 0.4 0.6 0.8 Probability Probability 0.2 0.2 HR = 2.19 HR = 3.514 P = 0.057 P = 0.00066 0.0 0.0 0 5 10 15 20 0 10 20 30 PFS (months) - 1st Line OS (months) - 1st Line C P = 0.007 (**) P = 0.014 (*) 60 P = 0.002 (**) 50 40 MAF (%) 30 20 10 0 PR SD PD Best response Fig. 2. PFS and OS analyses in first-line treatment. Survival curves are shown for samples with MAF < 5.8% (black line) and MAF> 5.8% (red line) in terms of PFS (A) and OS (B) in the 1st line. HR and P-values are shown. (C) MAF distribution according to best response to treatment. Table 2. Multivariate analysis for first-line PFS and OS. PFS OS Risk factor HR 95% CI P-value HR 95% CI P-value Gender 2.37 0.77–7.38 0.135 1.20 0.33–4.38 0.778 Laterality 0.54 0.17–1.74 0.303 0.32 0.08–1.19 0.088 CEA 0.99 0.99–1.00 0.656 0.99 0.99–1.00 0.711 No. of hepatic lesions 1.04 0.32–3.37 0.947 1.05 0.30–3.70 0.935 Plasma MAF 3.74 1.01–13.92 0.049 2.73 2.35–182.53 0.006 lymph nodes was significantly associated with higher prognostic value in mCRC (Riihim€ aki et al., 2016; MAFs. Although there is currently no clear explana- Vidal et al., 2017; Yaeger et al., 2015). tion for this phenomenon, our results are in line with Evaluation of best response to treatment showed previous studies reporting that the site of metastatic that patients with higher MAFs had PD or SD rather spread rather than the number of lesions has than PR. Our study also indicates that patients with 1832 Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Elez et al. RAS MAFs in metastatic colorectal cancer 1.0 1.0 A B MAF < 5.8% (n = 23) MAF < 5.8% (n = 23) MAF > 5.8% (n = 9) MAF > 5.8% (n = 9) 0.4 0.6 0.8 0.4 0.6 0.8 Probability Probability 0.2 0.2 HR = 2.46 HR = 3.84 P = 0.033 P = 0.004 0.0 0.0 0 5 10 15 20 0 10 20 30 40 50 PFS (months) OS (months) Fig. 3. PFS and OS analyses in the validation cohort (CAPRI-GOIM trial). Survival curves are shown for samples with MAF < 5.8% (black line) and MAF > 5.8% (red line) in terms of PFS (A) and OS (B). HR and P-values are shown. higher RAS MAFs do present with more resistant along with tumor and patient characteristics, could tumors to conventional therapies. New or experimental provide a useful noninvasive decision-making tool in approaches should be considered for them. the first-line setting. After demonstrating the feasibility In our multivariate statistical model, RAS MAF for implementing liquid biopsies in routine care (Gras- did not correlate with either CEA or CA19-9, unlike selli et al., 2017), we propose RAS MAFs as a novel recent observation that elevated CA19-9 levels repre- independent prognostic biomarker for mCRC. sented a strong prognostic marker (Rahbari et al., 2017). However, we did observe that patients with Acknowledgements higher MAFs also tended to have higher levels of both CA19-9 and CEA, though not reaching statisti- The authors acknowledge the medical editing assis- cal significance. tance of Sarah MacKenzie (PhD). FC acknowledges The most striking finding of our analysis is that the AIRC (Associazione Italiana per la Ricerca sul Can- estimation of RAS MAF in liquid biopsies correlates cro) for funding: IG-2016-ID:14800. NN acknowledges with predicting life expectancy in this mCRC popula- AIRC (Associazione Italiana per la Ricerca sul can- tion. Our data provide evidence that baseline patients cro). This research project was supported by ESMO with higher RAS MAFs in cfDNA tend to progress [Translational Research Fellowship] with the aid of a after a shorter time and have significantly shorter OS. grant from Amgen. Any views, opinions, findings, con- An independent cohort from the CAPRI-GOIM trial clusions, or recommendations expressed in this mate- was analyzed, and plasma MAF at baseline resulted rial are those solely of the author(s) and do not statistically significant for prognosis in both OS and necessarily reflect those of ESMO or Amgen. This PFS. An improved prognostic value in PFS was work was supported partially by the Instituto de Salud observed in this first-line setting. A MAF cutoff was Carlos III (Ministerio de Economıa y Competitividad) selected, using the cutp algorithm, based on the ability and ‘Fondo Europeo de Desarrollo Regional to better segregate outcomes in terms of PFS and OS (FEDER), una manera de hacer Europa’ grants [FIS (5.8% MAF). Additional cutoffs of 1% and 10% PI12-01589 to RS] and RETICC Cancer. The funders MAF used in previous publications (El Messaoudi had no role in study design, data collection and et al., 2016; Siravegna et al., 2017; Vidal et al., 2017) analysis, decision to publish, or preparation of the were also evaluated, being 5.8% the one that overall manuscript. provided better prognostic value in our patient cohorts. The sample size is indeed relatively small and Conflict of interest larger prospective studies to confirm our analyses and JT has served in a consulting or advisory role for further evaluate clinical parameters will be valuable. Amgen, Boehringer Ingelheim, Celgene, Chugai, ImClone, Lilly, Merck, S.L., Madrid, Merck Serono, 5. Conclusion Millennium Pharmaceuticals, Inc., Novartis, Roche, Our data strongly support that RAS MAFs have inde- Sanofi, and Taiho. RS has served in a consulting or pendent prognostic value for CRC survival and that, advisory role for Amgen, Merck, S.L., Madrid, and Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1833 RAS MAFs in metastatic colorectal cancer E. Elez et al. Roche Dx and obtained research funding from Roche El Messaoudi S, Mouliere F, Du Manoir S, Bascoul- Dx. AV has served in a consulting or advisory role Mollevi C, Gillet B, Nouaille M, Fiess C, Crapez E, for Merck, S.L., Madrid, Merck Serono, and Sysmex. Bibeau F, Theillet C et al. (2016) Circulating DNA as All remaining authors have declared no conflicts of a strong multimarker prognostic tool for metastatic interest. colorectal cancer patient management care. Clin Cancer Res 22, 3067–3077. Grasselli J, Elez E, Caratu G, Matito J, Santos C, Author contributions Macarulla T, Vidal J, Garcia M, Vieitez JM, Paez D et al. (2017) Concordance of blood- and tumor-based AV and EE conceived of the presented idea. AV, detection of RAS mutations to guide anti-EGFR EE, TT, and FC developed the study design. ES, therapy in metastatic colorectal cancer. Ann Oncol 28, EM, AN, JG, GC, RE, GM, CS, TM, GA, JC, and 1294–1301. NM contributed to sample acquisition. GC and JM Laurent-Puig P, Pekin D, Normand C, Kotsopoulos SK, performed the experiments. FM, AG, and AV per- Nizard P, Perez-Toralla K, Rowell R, Olson J, formed the quality control of data and algorithms. Srinivasan P, Le Corre D et al. (2015) Clinical FM, EE, AV, and RS contributed to data analyses relevance of KRAS-mutated subclones detected with and the interpretation of the results. FM, EM, and picodroplet digital PCR in advanced colorectal cancer NN performed the statistical analysis. FM prepared treated with anti-EGFR therapy. Clin Cancer Res 21, the figures. CC wrote and edited the manuscript 1087–1097. with support from EE and AV. All authors provided Mandrekar J, Mandrekar S and Cha S (2003) Cutpoint critical feedback and helped shape the research, anal- Determination Methods in Survival Analysis using ysis, and manuscript. SAS Proc 28th SAS Users Gr Int. Conf (SUGI), 261– 228. Normanno N, Esposito Abate R, Lambiase M, Forgione References L, Cardone C, Iannaccone A, Sacco A, Rachiglio AM, Arnold D, Lueza B, Douillard JY, Peeters M, Lenz HJ, Martinelli E, Rizzi D et al. (2017) RAS testing of Venook A, Heinemann V, Van Cutsem E, Pignon liquid biopsy correlates with the outcome of metastatic JP, Tabernero J et al. (2017) Prognostic and colorectal cancer patients treated with first-line predictive value of primary tumour side in patients FOLFIRI plus cetuximab in the CAPRI-GOIM trial. with RAS wild-type metastatic colorectal cancer Ann Oncol 29, 112–118. treated with chemotherapy and EGFR directed Ono A, Kenmotsu H, Watanabe M, Serizawa M, Mori K, antibodies in six randomized trials. Ann Oncol 28, Imai H, Taira T, Naito T, Murakami H, Nakajima T 1713–1729. et al. (2014) Mutant allele frequency predicts the Azuara D, Santos C, Lopez-Doriga A, Grasselli J, Nadal efficacy of EGFR-TKIs in lung adenocarcinoma M, Sanjuan X, Marin F, Vidal J, Montal R, Moreno harboring the L858R mutation. Ann Oncol 25, 1948– V et al. (2016) Nanofluidic digital PCR and extended 1953. genotyping of RAS and BRAF for improved selection Petrelli F, Tomasello G, Borgonovo K, Ghidini M, of metastatic colorectal cancer patients for Anti-EGFR Turati L, Dallera P, Passalacqua R, Sgroi G and therapies. Mol Cancer Ther 15, 1106–1112. Barni S (2017) Prognostic survival associated with Bettegowda C, Sausen M, Leary RJ, Kinde I, Wang Y, left-sided vs right-sided colon cancer. JAMA Oncol 3, Agrawal N, Bartlett BR, Wang H, Luber B, Alani RM 211. et al. (2014) Detection of circulating tumor DNA in Rahbari NN, Sch€ olch S, Bork U, Kahlert C, Schneider M, early- and late-stage human malignancies. Sci Transl Rahbari M, B€ uchler MW, Weitz J and Reissfelder C Med 6, 224ra24. (2017) Prognostic value of circulating endothelial cells Ciardiello F, Normanno N, Maiello E, Martinelli E, in metastatic colorectal cancer. Oncotarget 8, 37491– Troiani T, Pisconti S, Giuliani F, Barone C, Cartenı 37501. G, Rachiglio AM et al. (2014) Clinical activity of Riihim€aki M, Hemminki A, Sundquist J and Hemminki K FOLFIRI plus cetuximab according to extended gene (2016) Patterns of metastasis in colon and rectal mutation status by next-generation sequencing: findings cancer. Sci Rep 6, 29765. from the CAPRI-GOIM trial. Ann Oncol 25, 1756– Siegel R, Miller KD, Fedewa SA, Ahnen DJ, Meester 1761. RGS, Barzi A and Jemal A (2017) Colorectal cancer Contal C and O’Quigley J (1999) An application of statistics, 2017. CA Cancer J Clin 67, 177–193. changepoint methods in studying the effect of age on Siravegna G, Marsoni S, Siena S and Bardelli A (2017) survival in breast cancer. Comput Stat Data Anal 30, Integrating liquid biopsies into the management of 253–270. cancer. Nat Rev Clin Oncol 14, 531–548. 1834 Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Elez et al. RAS MAFs in metastatic colorectal cancer Siravegna G, Mussolin B, Buscarino M, Corti G, Fig. S2. Correlation analysis. Dot plots depicting the Cassingena A, Crisafulli G, Ponzetti A, Cremolini C, correlation between MAF (%) and the following Amatu A, Lauricella C et al. (2015) Clonal evolution parameters: CA19.9, CEA, hepatic volume and num- and resistance to EGFR blockade in the blood of ber of metastatic sites. Pearson correlation coefficient colorectal cancer patients. Nat Med 21, 795–801. (cor) and P-values are reported. Van Cutsem E, K€ ohne CH, Lang I, Folprecht G, Nowacki Fig. S3. PFS and OS analyses at different MAF cut- MP, Cascinu S, Shchepotin I, Maurel J, Cunningham offs in first-line treatment. Survival curves are shown D, Tejpar S et al. (2011) Cetuximab plus irinotecan, for samples with MAF < 1% (black line) and MAF fluorouracil, and leucovorin as first-line treatment for > 1% (red line) in terms of PFS (A) and OS (B), as metastatic colorectal cancer: updated analysis of well as for samples with MAF < 10% (black line) and overall survival according to tumor KRAS and BRAF MAF > 10% (red line) in terms of PFS (C) and OS mutation status. J Clin Oncol 29, 2011–2019. (D). HR and P-values are shown. Correlation between Vidal J, Muinelo L, Dalmases A, Jones F, Edelstein D, PFS/OS and MAF is reported (E). Pearson correlation Iglesias M, Orrillo M, Abalo A, Rodrıguez C, Brozos E et al. (2017) Plasma ctDNA RAS mutation analysis coefficient (cor) and the P-value are reported. for the diagnosis and treatment monitoring of Fig. S4. PFS and OS analyses in second-line treatment. metastatic colorectal cancer patients. Ann Oncol 28, Survival curves are shown for samples with: (a) MAF 1325–1332. < 5.8% (black line) and MAF > 5.8% (red line) in Yaeger R, Cowell E, Chou JF, Gewirtz AN, Borsu L, terms of PFS (A) and OS (B); (b) MAF < 1% (black Vakiani E, Solit DB, Rosen N, Capanu M, Ladanyi M line) and MAF > 1% (red line) in terms of PFS (C) et al. (2015) RAS mutations affect pattern of and OS (D); (c) MAF < 10% (black line) and MAF metastatic spread and increase propensity for brain > 10% (red line) in terms of PFS (E) and OS (F). HR metastasis in colorectal cancer. Cancer 121, 1195–1203. and P-values are shown. Zhou Q, Zhang XC, Chen ZH, Yin XL, Yang JJ, Xu CR, Fig. S5. PFS and OS analyses at different MAF cut- Yan HH, Chen HJ, Su J, Zhong WZ et al. (2011) offs in the validation cohort (CAPRI-GOIM trial). Relative abundance of EGFR mutations predicts Survival curves are shown for samples with MAF benefit from gefitinib treatment for advanced non- < 1% (black line) and MAF > 1% (red line) in terms small-cell lung cancer. J Clin Oncol 29, 3316–3321. of PFS (A) and OS (B), as well as for samples with MAF < 10% (black line) and MAF > 10% (red line) Supporting information in terms of PFS (C) and OS (D). HR and P-values are shown. Additional supporting information may be found Table S1. Description of patients used in the study. online in the Supporting Information section at the end Table S2. RAS panel of mutations for BEAMing anal- of the article. ysis. Fig. S1. Sample selection. Flowchart of selection steps Table S3. MAF tendency according to CA 19-9 and for the analysis population. CEA levels. Molecular Oncology 13 (2019) 1827–1835 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1835