Prospective detection of mutations in cerebrospinal fluid, pleural effusion, and ascites of advanced cancer patients to guide treatment decisions Sergio Villatoro1 , Clara Mayo-de-las-Casas1, Nu  ria Jordana-Ariza1, Santiago Viteri-Ramırez2, Mo nica Garzo  n-Iban ~ ez , Irene Moya-Horno , Beatriz Garcıa-Pela 1 3 ez1, Marıa Gonza lez-Cao2, Umberto Malapelle , Ariadna Balada-Bel , Alejandro Martınez-Bueno , Raquel Campos1, 4 1 2 Noemı Reguart5, Margarita Majem6, Remei Blanco7, Ana Blasco8, Marıa J. Catala n1, Xavier Gonza lez , Giancarlo Troncone , Niki Karachaliou , Rafael Rosell 3 4 1,9 1,2,10 and Miguel A. Molina-Vila1 1 Laboratory of Oncology, Pangaea Oncology, Quiro n Dexeus University Hospital, Barcelona, Spain n Dexeus University Hospital, Barcelona, Spain 2 Dr Rosell Oncology Institute, Quiro s, Spain 3 Dr Rosell Oncology Institute, General Hospital of Catalonia, Sant Cugat del Valle 4 Department of Public Health, University of Naples Federico II, Italy 5 Department of Medical Oncology, Hospital Clınic, Barcelona, Spain 6 Medical Oncology Service, Hospital de Sant Pau, Barcelona, Spain 7 Medical Oncology Service, Hospital de Terrassa, Spain 8 Medical Oncology Department, Hospital General Universitario de Valencia, Spain 9 Dr Rosell Oncology Institute, University Hospital Sagrat Cor, Barcelona, Spain 10 Cancer Biology and Precision Medicine Program, Catalan Institute of Oncology, Germans Trias i Pujol Health Sciences Institute and Hospital, Badalona, Spain Keywords Many advanced cases of cancer show central nervous system, pleural, or ascites; cell-free DNA; cerebrospinal fluid; peritoneal involvement. In this study, we prospectively analyzed if cere- liquid biopsy; pleural effusion; solid tumors; brospinal fluid (CSF), pleural effusion (PE), and/or ascites (ASC) can be somatic mutations used to detect driver mutations and guide treatment decisions. We collected Correspondence 42 CSF, PE, and ASC samples from advanced non-small-cell lung cancer M. A. Molina-Vila, Laboratory of Oncology, and melanoma patients. Cell-free DNA (cfDNA) was purified and driver Pangaea Oncology, External Consultation mutations analyzed and quantified by PNA-Q-PCR or next-generation Building, Room 1.7, Quiro n Dexeus sequencing. All 42 fluid samples were evaluable; clinically relevant muta- University Hospital, C/ Sabino Arana 5-19, tions were detected in 41 (97.6%). Twenty-three fluids had paired blood 08028 Barcelona, Spain samples, 22 were mutation positive in fluid but only 14 in blood, and the Fax: +34 935460137 abundance of the mutant alleles was significantly higher in fluids. Of the 34 Tel: +34 93 546 01 40 E-mail: mamolina@panoncology.com fluids obtained at progression to different therapies, EGFR resistance muta- tions were detected in nine and ALK acquired mutations in two. The Rafael Rosell and Miguel A. Molina-Vila results of testing of CSF, PE, and ASC were used to guide treatment deci- shares last authorship sions, such as initiation of osimertinib treatment or selection of specific ALK tyrosine–kinase inhibitors. In conclusion, fluids close to metastatic (Received 14 May 2019, revised 6 sites are superior to blood for the detection of relevant mutations and can September 2019, accepted 13 September offer valuable clinical information, particularly in patients progressing to 2019, available online 11 October 2019) targeted therapies. doi:10.1002/1878-0261.12574 Abbreviations cfDNA, cell-free DNA; CNS, central nervous system; CSF, cerebrospinal fluid; NGS, next-generation sequencing; NSCLC, non-small-cell lung cancer; PE, pleural effusion; TKI, tyrosine–kinase inhibitor. Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2633 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. Detection of mutations in fluids of cancer patients S. Villatoro et al. 1. Introduction whether cfDNA purified from CSF, PE, and ASC can be reliably used to determine mutations and other With the advent of precision medicine, mutation test- alterations in cancer patients (Table S1) and have con- ing in tumor biopsies from advanced patients has cluded that these fluids are suitable for molecular pro- become a routine practice in several malignancies, filing (Li et al., 2018; Xu et al., 2018). However, these including non-small-cell lung cancer (NSCLC), meta- studies have generally been performed in a retrospec- static melanoma, or colorectal cancer. In addition, tive manner using fluid samples with paired biopsies genetic analysis of the so-called ‘liquid biopsies’, par- and, to the best of our knowledge, a recently published ticularly blood samples, is quickly gaining acceptance case report is the only example of ‘nonblood’ liquid in the clinical setting (Bedin et al., 2017; de Wit et al., biopsies actually being used to orient clinical decisions 2019 Goldman et al., 2018; Mayo-de-Las-Casas et al., (Melms et al., 2018). Therefore, the performance and 2018; Mayo-de-Las-Casas et al., 2017; Rolfo et al., utility in the clinical setting of prospective molecular 2018). Although liquid biopsies are unlikely to replace profiling using cfDNA from CSF, PE, and ASC ‘real’ biopsies, they provide valuable information that remain to be established. cannot be obtained in any other way. In patients In this article, we describe the results of the imple- where biopsies are impossible or do not yield enough mentation in the routine practice of mutation analysis tissue, biological fluids are the only samples available in CSF, PE, and ASC of advanced cancer patients for genetic testing. Moreover, in patients with multiple with central nervous system (CNS), pleural, and peri- metastases, it is not feasible to obtain biopsies of every toneal involvement, respectively. A total of 42 ‘non- site but genetic materials coming from all of them can blood’ liquid biopsies, 34 of them at progression to be found in blood. Finally, in contrast to ‘real biop- therapies, were systematically collected in several insti- sies’, liquid biopsies can be periodically obtained with tutions; cfDNA was purified and the subsequent highly limited risk of complications through the proce- molecular testing used to guide treatment decisions. dure. This repeated testing can be used to monitor the Our work demonstrates that prospective genetic analy- course of the disease and facilitate early evaluation of sis of cfDNA in CSF, PE, and ASC offers valuable response, detection of relapse, and characterization of clinical information and that these fluids are superior resistance mechanisms to targeted therapies (Tsui to blood for mutation testing in patients with certain et al., 2018; Ulz et al., 2017; Yi et al., 2017). This last metastatic sites. aspect is particularly relevant, since drugs active against some of the resistance mechanisms are approved or in late-stage clinical trials. Examples 2. Materials and methods include osimertinib, which targets the p.T790M muta- tion in EGFR emerging in NSCLC patients treated with EGFR tyrosine kinase inhibitors (TKIs); or sec- 2.1. Patients ond- and third-generation ALK TKIs, such as briga- Forty-two fluids collected during a period of 2 years in tinib or lorlatinib, that show efficacy in patients eight hospitals from 23 advanced NSCLC and two harboring resistance mutations in ALK (Zhang et al., metastatic melanoma patients were prospectively tested 2016). for genetic alterations in a central laboratory located Circulating free DNA (cfDNA) purified from serum in the Quir on Dexeus Hospital (Table 1). The study or plasma is the most widely used type of liquid was carried out in accordance with the principles of biopsy. Since it contains a fraction of tumor DNA the Declaration of Helsinki under an approved proto- (circulating tumor DNA) mixed with DNA from nor- col of the institutional review board of Quir on Hospi- mal cells, it is employed to identify mutations, amplifi- tals. Written informed consent was obtained from all cations, gene fusions, and other somatic alterations patients and documented and samples were deidenti- that modify the DNA sequence (Gonzalez-Cao et al., fied for patient confidentiality. 2015; Mayo-de-Las-Casas et al., 2017; Szallasi, 2017). However, the concept of ‘liquid biopsy’ is not restricted to blood-derived samples and includes other 2.2. Cell lines body fluids such as cerebrospinal fluid (CSF), pleural NCI-H1975, NCI-H23, DLD1, HT29, and NCI-H460 effusion (PE), and ascites (ASC). In all these cases, human tumor cell lines were purchased from the cfDNA can be isolated from the supernatants obtained American Type Culture Collection (Rockville, MD, after centrifugation of the fluid and submitted to USA) and grown under standard conditions; PC9 was genetic analyses. A number of studies have evaluated obtained from Hofmann-La Hofmann Roche Ltd with 2634 Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. S. Villatoro et al. Detection of mutations in fluids of cancer patients Table 1. Characteristics of the patients included in the study. list of mutations), we used a quantitative real-time All patients PCR (Taqman) assay in the presence of a PNA Characteristics N = 25 clamp (Eurogentec, Seraing, Belgium) designed to inhi- bit the amplification of the wild-type alleles (Gonzalez- Gender Cao et al., 2018; Karachaliou et al., 2015; Mayo-de- Male 8 Las-Casas et al., 2017). The assay also allows estima- Female 17 Primary tumor tion of the absolute and relative abundances of mutant Advanced NSCLC 23 alleles in positive samples. Briefly, amplification is per- Advanced melanoma 2 formed in a final volume of 12.5 µL, using 3 µL Type of mutation (~ 4.5 ng) for exon 21 analysis or 1 µL (~ 1.5 ng) for EGFR 19 exon 19 and p.T790M analysis of cfDNA; 6.25 µL of KRAS 1 Genotyping Master Mix (Applied Biosystems, Foster BRAF 3 City, CA, USA); 0.96 pmol of each primer; 1.2 pmol ALK 2 Patients with samples of probes and 6.25 pmol (for exon 21 and p. T790M) Only at presentation 1 or 62.4 pmol (for exon 19) of PNA. Samples are sub- Only at progression 19 mitted to 50 cycles of 15 s at 92 °C and 1.5 min at More than one sample 5 60 °C, in a QuantStudioTM 6 real-time PCR System Presentation and follow-up 1 (Applied Biosystems/Thermo Fisher Scientific, Wal- Follow-up and progression 4 tham, MA, USA). The sequence of the primers, probes, and PNAs used has been described elsewhere (Gonzalez-Cao et al., 2015; Karachaliou et al., 2015). the authorization of M. Ono. Cell lines were validated Analyses were carried out in duplicates using two sam- by genotyping them for EGFR, KRAS, PIK3CA, ples of purified cfDNA, when possible. In addition, all BRAF, and TP53 and used as positive or negative con- samples were assayed in the absence of PNA to con- trols (Table S3). DNA from the cell lines was purified firm the presence of cfDNA. Genomic DNAs from cell using the DNA Easy extraction kit (Qiagen, Hilden, lines at 5 ngµL 1 were used as positive and negative Germany), according to the manufacturer’s instruc- controls (Table S2). Extraction and nontemplate con- tions, and used as a control material. trols were added in each run. A sample was considered positive if the same mutant allele amplified in the two duplicates in the presence of PNA. If amplification 2.3. cfDNA isolation from fluid and plasma was only detected in one duplicate, samples were rean- samples alyzed and considered positive if again at least one of Fluid samples (3–500 mL) were collected in sterile con- the duplicates was positive for the same mutated allele. tainers and plasma samples (10 mL) in Vacutainer tubes (BD, Plymouth, UK). After a first centrifugation 2.5. NGS for mutation testing step (509 g, 10 min), the supernatants were transferred to a new tube and immediately submitted to a second In order to detect the ALK resistance mutations and centrifugation followed by cfDNA purification. The the G719X in EGFR, not covered by our Taqman sediments of the first and second centrifugation, if pre- assay (Table S3), we used next-generation sequencing sent, were extended on slides, stained by hematoxylin (NGS) with the GeneReader Platform (QIAGEN). and examined under a microscope. Purification of Purified DNA (16.5 µL, ~ 40 ng) was used as a tem- cfDNA from supernatants (1.2 mL) was performed plate to generate libraries for sequencing with the Gen- with the QIAsymphony DSP Virus/Pathogen Midi eReadTM QIAact Lung DNA UMI Panel, according Kit using a QIAsymphony robot (Qiagen), following to manufacturer’s instructions. The panel is designed the manufacturer’s instructions. Extraction was per- to enrich specific target regions containing 550 variant formed in duplicates; cfDNA was purified from two positions in 19 selected genes frequently altered in lung aliquots of fluid. The final elution volume was 30 µL cancer tumors (AKT1, ALK, BRAF, DDR2, EGFR, per aliquot. ERBB2/HER2, ESR1, KIT, KRAS, MAP2K1, MET, NRAS, NTRK1, PDGFRA, PIK3CA, PTEN, ROS1, FGFR1, and RICTOR), including MET exon 14 skip- 2.4. PNA-Q-PCR assay for mutation testing ping mutations. Libraries were quantified using a In order to detect EGFR (exons 19, 20, 21), KRAS QIAxcel Advanced System, diluted to 100 pgµL 1, (exon 2), and BRAF (exon 15) mutations (Table S1, and pooled in batches of 6 (liquid biopsies) or 12 Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2635 Detection of mutations in fluids of cancer patients S. Villatoro et al. (tissues). Clonal amplification was performed on were used for comparisons and P-values of < 0.05 625 pg of pooled libraries by the GeneRead Clonal were considered to be statistically significant. Amp Q Kit using the GeneRead QIAcube and an automated protocol. Following bead enrichment, 3. Results pooled libraries were sequenced using the GeneRead UMI Advanced Sequencing Q kit in a GeneReader instrument. QIAGEN Clinical Insight Analyze soft- 3.1. Patients and samples ware was employed to perform the secondary analysis of FASTQ reads, align the read data to the hg19 refer- A total of 42 fluid samples were collected from the 25 ence genome sequence, call sequence variants, and gen- patients included in the study (Table 1, Fig. 1); eight erate a report for visualization of the sequencing ASC, 12 PE, and 22 CSF (Table 2). The type of fluid results. Variants were imported into the QIAGEN collected was dependent on the metastatic sites within Clinical Insight Interpret web interface for data inter- the patient, cytologies were positive only in 10 cases. pretation and generation of final custom report. Of the 42 fluids included in the study, 23 had paired blood samples (four PE, 12 CSF, and seven ASC). Importantly, tissue biopsies at the moment of the fluid 2.6. Quantification of the wild-type and mutant sample collection were not available in any case. alleles Regarding the time of collection, two fluid samples For absolute quantification, serially diluted genomic were obtained baseline, six during to follow-up, and DNAs from the appropriate positive or negative cell 34 at progression to different therapies. One baseline lines were analyzed by PNA-Q-PCR and the corre- sample corresponded to the CSF of a NSCLC patient sponding plot of log (genome equivalents per mL) vs. and was positive for the p.L858R in EGFR. The Ct was represented. The resulting linear equation was remaining baseline and four follow-up samples were used to interpolate the absolute concentration of CSFs from a NSCLC patient, all of them showed the mutant or wild-type genomes in the samples based on p.G12V mutation in KRAS mutation but significant the Ct experimentally obtained (Fig. S1A). To deter- changes in the allelic fraction were observed (see mine the relative amount of mutations (allelic frac- below). Finally, the two additional follow-up fluids tion), serial dilutions of positive in negative genomic were BRAF p.V600E-positive ASC from a melanoma DNA, corresponding to allelic fractions 1 : 20 to patient (Table 2). 1 : 20 000, were periodically run. In the PNA-Q-PCR Of the 34 fluids at progression included in the study, assay, the difference between the Ct of the mutated 14 were EGFR-mut CSFs or ASC collected from allele and the Ct (ΔCt) corresponds to a lower EGFR NSCLC patients relapsing to first or second-line TKIs. mutated allelic fraction, ΔCt = a (b 9 log ML), The p.T790M resistance was detected in 50% of them. where ML stands for the relative abundance of the Seven EGFR-mut fluids corresponded to patients in mutant allele, or mutation load (Gonzalez-Cao et al., progression to 3rd-generation EGFR TKIs, the 2015; Mayo-de-Las-Casas et al., 2017). In conse- p.T790M had disappeared in four, while two were pos- quence, we plotted the ΔCt vs. the logarithm of the itive for the p.T790M + p.C797S. Two additional allelic fraction in the serial dilutions and used it to cal- CSFs were collected from EML4-ALK patients in culate the parameters a and b in the linear equa- relapse to TKIs, both showed ALK resistance muta- tion ΔCt = a (b 9 log ML). Finally, the allelic tions. Six ASC and two CSFs corresponded to mela- fraction in the cfDNA samples was interpolated from noma patients progressing to immunotherapy and this linear equation (Fig. S1B). BRAF + MEK inhibitors, respectively, and they were In the case of the NGS samples, the allelic fraction all positive for the p.V600E mutation in BRAF. was directly estimated from the absolute amount of Finally, 3 fluids corresponded to NSCLC patients pro- reads of the wt and mutant alleles. gressing to chemotherapy, KRAS mutations were detected in two cases (Table 2). 2.7. Statistical analyses 3.2. Comparison of fluid and plasma samples The results were statistically analyzed using R software (R Core Team, 2012). Shapiro–Wilk normality tests The total concentration of cfDNA in the fluid (n = 42) revealed that the distributions of absolute and relative and blood (n = 23) samples included in the study was quantities of mutations in the fluid samples analyzed estimated based on the Ct of the wt alleles (see Materi- were not normal. In consequence, Kruskal–Wallis tests als and Methods). The concentrations of cfDNA in 2636 Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. S. Villatoro et al. Detection of mutations in fluids of cancer patients Fig. 1. Flowchart of the patient population included in the study. None of the 25 patients had biopsies available at the time of fluid sample extraction. blood and CSF were found to be indistinguishable, 3.3. Analysis of paired samples with medians of 9.3 pgµL 1 of plasma and 12.6 pgµL 1 of CSF, respectively. In contrast, ASC Next, we specifically compared the results obtained in and PE showed higher cfDNA concentrations, with the 23 paired fluid and plasma samples. Of those, 14 medians of 82.1 and 250 pgµL 1 of fluid, respectively. were positive in both liquids and eight positive only in These differences were statistically significant in a fluid. The remaining case was negative in the two liq- Kruskal–Wallis test (Fig. 2A). uids, while we did not find any case of positivity exclu- Absolute concentrations of the mutations were cal- sively in plasma (Table S4). Interestingly, the eight culated from the Ct of the mutant alleles and expressed cases negative in blood and positive in fluid corre- as pg of mutant genomes per µL of fluid. We observed sponded to CSF samples extracted from patients with that ASC had the highest concentration of mutant brain metastases. genomes, with a median of 56.5 pgµL 1; followed by The concentrations of cfDNA were usually higher in PE, CSFs, and blood samples, with medians of 12.0, fluids vs. paired plasma samples, with medians of 37.2 0.7 and 0.002 pgµL 1, respectively. The differences and 9.5 pgµL 1, respectively (Fig. 2E). Regarding the between plasma and the rest of fluids were significant absolute and relative concentrations of mutant alleles, in all cases (Fig. 2B). they were almost invariably superior in fluids vs. The relative concentrations of mutations were deter- paired plasma; with medians escalating to 3.0 pgµL 1 mined based on the difference of Ct between the wt and 2.8% in fluids, compared to 0.002 pgµL 1 and and the mutant alleles. In contrast to absolute concen- 0.008% in blood (Fig. 2F). These differences were sig- trations, we did not find statistically significant differ- nificant in a paired Wilcoxon test. ences in the allelic fractions of the mutations in ASC, PE, and CSF, which showed medians of 15.2%, 3.2%, 3.4. Mutation analysis in fluids to guide and 3.2%, respectively. However, the mutation loads treatment decisions in the three types of fluid were significantly higher than in plasma, where the median allelic fraction only The results of the mutation analysis in fluids were used reached 0.008% (Fig. 2C). When classifying samples to guide clinical decisions (Table 3). In two EGFR-mut according to the type of mutation, we did not observe patients progressing to first- or second-generation significant differences in the allelic fractions of KRAS, EGFR TKIs with brain and pleural involvement, the EGFR, and BRAF mutations in the fluid samples p.T790M could be detected in CSF and PE, respec- (Fig. 2D). tively. Both patients received osimertinib deriving a Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2637 Detection of mutations in fluids of cancer patients S. Villatoro et al. Table 2. Characteristics of the fluid samples included in the study. olmutinib was discontinued and replaced by All chemotherapy or best supportive care. In the case of samples chemotherapy, patients subsequently experienced stabi- Characteristics N = 42 lization of the disease for 5–9 months. Two CSFs from patients in progression to brigatinib Type of fluid and lorlatinib were analyzed by NGS, showing de novo Pleural effusion 12 Cerebrospinal fluid 22 p.F1174C and p.L1196M mutations in ALK, respec- Ascites 8 tively. New ALK TKIs were then selected based on Cytology (presence of tumor cells) their inhibitory profile against ALK resistance muta- Positive 10 tions. The patients have subsequently derived a signifi- Negative 28 cant clinical benefit, in both cases ongoing. Not analyzed 4 Remarkably, a paired blood sample was negative in Paired blood samples the case of the p.F1174C patient, who has been in No 19 Yes 23 response to lorlatinib for 22 months. Pleural effusion 4 Cerebrospinal fluid 12 3.5. Mutation analysis of serial fluid samples Ascites 7 Collection time and type of mutation Two clinical cases where serial samples were obtained Presentation 2 will described in further detail. The first case is a 63- p.L858R 1 year-old man, heavy smoker, diagnosed in February p.G12V 1 Follow-up 6 2015 with a lung adenocarcinoma with leptomeningeal p.G12V 4 and medullary metastases and carrying a p.G12V Melanoma, p.V600E 2 mutation in KRAS. CSF samples were systematically Progression to 1st- and 2nd-generation EGFR TKIs 14 collected and analyzed. The patient was referred to EGFR-sensitizing mutation 7 our hospital in November 2015; CSF was positive for EGFR-sensitizing mutation + p.T790M 7 the p.G12V mutation with a 2.8% allelic fraction while Progression to 3rd-generation EGFR TKIs 7 plasma was negative. Nab-paclitaxel plus carboplatin EGFR-sensitizing mutation 4 EGFR-sensitizing mutation + p.T790M 1 and bevacizumab was administered with stabilization EGFR-sensitizing 2 of the disease, which was accompanied by a reduction mutation + p.T790M + p.C797S in the p.G12V mutation in CSF, which dropped to Progression to ALKi 2 0.5% in January 2016. However, a new CSF analysis p.F1174C 1 on March 2016 showed a sharp increase to 7.6%. p.L1196M 1 Imaging techniques subsequently revealed a radiologi- Progression to immunotherapy, p.V600E 6 cal progression in brain. Stereotactic radiotherapy was Progression to BRAF + MEKi, melanoma, p.V600E 2 Progression to chemotherapy 3 then applied with a partial response and a new decrease of the mutation load to 0.8%. A second increase in the allelic fraction to 87.5% in July was accompanied with a rapid progression with lep- clinical benefit. Importantly, blood was negative for tomeningeal carcinomatosis and cauda equina syn- the resistance mutation in the patient with p.T790M- drome. The patient was then administered positive CSF, who was in almost complete response to immunotherapy in combination with radiotherapy, but osimertinib for 18 months. In contrast, three EGFR- was exitus in September (Fig. 3A). mut patients progressing to first- or second-generation The second case is a 62-year-old woman, never smo- EGFR TKIs only showed the sensitizing mutation in ker, diagnosed in 2013 with stage IV NSCLC with fluid samples. In consequence, chemotherapy or radio- bone and asymptomatic brain metastases at presenta- therapy was selected, with partial responses or stabi- tion, which harbored an EGFR deletion in exon 19 lization of the disease for 5–6 months. Paired blood and received erlotinib. On June 2014, a single site of samples were negative in most cases. progression in lung was surgically resected; genetic Five of the CSF samples included in our study cor- analysis revealed the presence of the p.T790M. The responded to patients progressing to 3rd-generation patient remained with erlotinib until April 2015, when TKIs, the sensitizing mutation was present but the a significant impairment in performance status was p.T790M was either absent or present in cis configura- accompanied by leptomeningeal carcinomatosis. tion with the p.C797S. In consequence, osimertinib or Osimertinib was then administered and the patient 2638 Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. S. Villatoro et al. Detection of mutations in fluids of cancer patients Fig. 2. (A–C) Box plots representing (A) the dynamic ranges of total cfDNA, (B) the concentrations, and (C) the allelic fractions of mutant alleles in PE, CSF, ASC, and plasma samples. (D) Dynamic ranges of allelic fractions of EGFR, BRAF, and KRAS mutations in fluid samples. Solid lines indicate medians and numbers level of significance (P) in a Kruskal–Wallis test. (E, F) Box plot representing the dynamic ranges of (E) total cfDNA and (F) allelic fractions in fluid vs. plasma paired samples. Solid lines indicate medians and the numbers level of significance (P) in a paired Wilcoxon test. Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2639 2640 Table 3. Clinical decisions based on results of a mutation analysis in CSF, PE, or ascites. LF, loss to follow-up; NA, paired blood not available for analysis; NEG, negative; NR, not recorded; VAF, variant allelic fraction; WBRT, Whole brain radiation therapy. Type VAF in Patient Date of of Mutation Previous lines Type of code analysis fluid Gen Mutation status in fluid VAF in fluid status in blood blood of treatment Clinical decision response PFS 6462 02/02/2017 CSF EGFR L858R+ 5.6% NEG 0% 1-Erlotinib Change to PR 18 months osimertinib T790M+ 2.3% 0% 3970 23/11/2016 PE EGFR Del 24 pb+ 0.2% NA – 1-Erlotinib Change to PR 9 months osimertinib T790M+ 0.2% 4427 20/11/2017 CSF EGFR L858R+ 46.4% NEG 0% 1-Gefitinib Change to SD 5 months Detection of mutations in fluids of cancer patients chemotherapy T790M- 0% 0% 8415 27/05/2018 CSF EGFR E709K/G719S+T790M- 3.5%0% NA – 1-Afatinib Change to LF LF chemotherapy 4827 22/09/2017 PE EGFR Del 15 bp+T790M- 2.7%0% Del 0.1%0% 1-Erlotinib Lung radiotherapy PR 6 months 15 bp+T790M- and chemotherapy 6606 26/10/2018 CSF EGFR Del 15 pb+T790M- 83.5%0% NEG 0%0%0% 1-Gefitinib2- Change to LF LF C797S- 0% Osimertinib chemotherapy 4249 03/05/2018 CSF EGFR Del 15 pb+T790M- 46.5%0% NEG 0%0%0% 1-Afatinib2- Change to SD 9 months C797S- 0% Osimertinib chemotherapy 5057 20/06/2017 CSF EGFR Del 9 pb+T790M-C797S- 3.7%0%0% Del 9 pb+T790M- 0.6%0% 1-Erlotinib2- Stop olmutinibBest SD 6 months C797S- 0% Olmutinib supportive care 6737* 10/03/2017 CSF EGFR Del 15 pb+T790M+ 16.7% Del 15 pb+ 0.5%0.3% 1-Erlotinib2- Change to SD 5 months 16.5% T790M+ 0.2% Osimertinib chemotherapy + C797S+ (cis) 11.2% C797S+ (cis) avastin 0965 25/09/2015 CSF EGFR L858R+T790M-C797S- 2.2%0%0% NEG 0%0%0% 1-Four lines of Stop WBRTChange SD 3 months EGFR TKIs2- to best supportive WBRT care 0145 17/04/2017 CSF ALK F1174C 4.0% NEG 0% 1-Seven lines of Change to lorlatinib PR 22 months treatment2- (ongoing) Brigatinib 8828 08/01/2019 CSF ALK L1196M 51.8% NA – 1-Lorlatinib Change to brigatinib PR 4 months (ongoing) *Patient presented in Fig. 3C. S. Villatoro et al. Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. S. Villatoro et al. Detection of mutations in fluids of cancer patients remained in response for 2 years, when she presented 4. Discussion a CNS progression. A first analysis of CSF, in April 2017, revealed the presence of the deletion in exon 19 Genetic testing in cfDNA purified from plasma of and the p.T790M at allelic fractions of 7% and 5%, advanced cancer patients is quickly becoming a stan- together with the p.C797S in cis with the p.T790M, at dard practice in many hospitals (Janku and Kurzrock, a 0.7% allelic fraction. This result indicated the emer- 2016; Janku et al., 2017). However, PE, CSF, or ASC gence of a p.T790M + p.C797S tumor cell population can also be employed for mutation testing. PE and resistant to osimertinib, which could not be rechal- ASC are usually available in large quantities, by mini- lenged with other EGFR TKIs. Carboplatin, taxol, mally invasive procedures, in patients with pleural or and avastin were then added to the treatment. A CSF peritoneal involvement. In contrast, CSF is more diffi- sample collected 5 months later, in September 2017, cult to extract, and the volumes obtained are fre- presented the sensitizing mutation, the p.T790M and quently limited. However, CSF is in direct contact the p.C797S at significantly higher allelic fractions with the brain and constitutes a valuable material for 12% and 16%, respectively. The patient underwent a genetic testing in patients with SNC metastases or rapid progression and was exitus a month later brain tumors (De Mattos-Arruda et al., 2015; Mou- (Fig. 3B). liere et al., 2018; Nevel et al., 2018; Panditharatna Fig. 3. Mutation analysis in serial fluid samples from selected patients correlates with the course of the disease. (A) Evolution of the allelic fractions of KRAS p.G12V mutation in CSF of a 63-year-old man, heavy smoker, with a lung adenocarcinoma with leptomeningeal and medullary metastases. (B) Clinical course of a 62-year-old woman, never smoker, diagnosed in 2013 with stage IV NSCLC with bone and asymptomatic brain metastases at presentation. Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2641 Detection of mutations in fluids of cancer patients S. Villatoro et al. et al., 2018; Pentsova et al., 2016; Seoane et al., 2019; Although some studies suggest that the spatial hetero- Zhao et al., 2016). Despite the potential of CSF, PE, geneity in the distribution of the p.T790M can lead to or ASC as liquid biopsies, attention at the clinical level positivity in the blood and negativity in the CSF of has been focused on blood and the performance and patients relapsing to EGFR TKIs (Hata et al., 2014), clinical utility of prospective molecular profiling in in our cohort all cases p.T790M negative in CSF were those fluids has not been adequately investigated. also negative in blood. The fact that the majority of In this study, we systematically collected and the patients with CSF analyzed were progressing exclu- prospectively analyzed 22 CSF, 12 PE, and eight ASC sively in SNC might explain this finding. In addition samples obtained from advanced NSCLC and mela- to fluid samples at relapse to first- and second-genera- noma patients with pleural, peritoneal, and brain tion EGFR TKIs, CSF from seven patients with brain involvement. A majority of those ‘nonblood’ fluids progression after treatment with third-generation TKIs were obtained at progression to different therapies, was prospectively tested. In two cases, only the initial particularly TKIs. Importantly, patients could not be sensitizing mutation was detected at allelic fractions as biopsied at the moment of sample collection, and fluids high as 45%, indicating disappearance of the p.T790M were the only material available. More than 50% of as a mechanism of resistance. In two other samples, the CSF, PE, and ASC samples analyzed had negative the p.C797S was found in cis with the p.T790M, cytologies and, in those where tumor cells were present, excluding rechallenge with EGFR TKIs (Chic et al., they were often insufficient for genetic analyses. 2017; Niederst et al., 2015). Although some studies have reported more than 50% The results obtained in our paired samples confirm, invalid samples when analyzing CSF from lung cancer in a prospective manner, that CSF, PE, and ASC are patients (Kawahara et al., 2019), the 42 fluid samples superior to blood for mutation testing in patients with included in our study yielded valid results after cen- SNC, pleural, and peritoneal involvement. The median trifugation, cfDNA isolation, and mutation testing. allelic fractions of the mutations were 350 times higher Clinically relevant mutations could be detected in 41/42 in fluids than in paired blood samples and, in eight cases (98%), a percentage significantly superior to the patients with SNC metastases, plasma samples tested 70–80% sensitivity usually reported for mutation test- negative but CSF was clearly positive for clinically rel- ing in cfDNA isolated from blood (Mayo-de-Las-Casas evant mutations. In addition, allelic fractions were et al., 2017). In addition, mutation burden in fluid sam- below 0.01% in a significant number of positive paired ples could be easily calculated, allowing for monitoriza- blood samples, and mutations could only be detected tion of the course of the disease in those patients where because of the highly sensitive Q-RT-PCR used in our serial sampling was feasible. Our results are concordant study. Commercially available kits and NGS panels with several retrospective studies, generally including would have probably missed some of them, due to low number of samples, reporting good sensitivity (64– insufficient sensitivity (Malapelle et al., 2017; Mayo- 100%) for EGFR KRAS or BRAF mutation detection de-Las-Casas et al., 2017). In contrast, the allelic frac- in the CSF (Ballester et al., 2018; Sasaki et al., 2016; tions in nonblood fluids, with a median value of 2.8%, Yang et al., 2014; Zhao et al., 2016) or PE (Kang are adequate to be detected by these platforms. Our et al., 2015; Liu et al., 2014; Liu et al., 2018; Shin study also indicates that the results of mutation testing et al., 2017) of NSCLC and melanoma patients with in fluids, not in blood, should be preferentially consid- brain or pleural involvement, either baseline or after ered for the selection of antitumor treatments in can- progression to targeted therapies. cer patients with particular metastatic sites. An EGF- Of the 42 ‘nonblood’ liquid biopsies included in our mut patient with brain metastasis negative for the study, 34 where obtained at relapse to antitumor ther- p.T790M in blood but positive in CSF derived an 18- apies, constituting the largest reported cohort of fluid month clinical benefit from osimertinib, while the samples at progression. Clinically relevant mutations detection of a p.F1174C mutation in an EML4-ALK were detected in all the 34 samples, and the results of patient relapsing to brigatinib allowed for the selection the testing were used to guide treatment decisions. A of lorlatinib as the next-line treatment (Gainor et al., significant number of cases were CSFs or PEs from 2016) with a 22-month response, still ongoing. Again, EGFR-mut patients progressing to first- and second- blood was negative for ALK mutations. generation EGFR TKIs, where the presence of the p.T790M was used to select for osimertinib treatment. 5. Conclusions In contrast, when the EGFR-sensitizing mutation was detected in significant allelic fractions but the We have presented the results of the prospective muta- p.T790M was absent, osimertinib was excluded. tion testing of 42 CSF, PE, and ASC samples from 2642 Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. S. Villatoro et al. Detection of mutations in fluids of cancer patients advanced NSCLC or melanoma patients with CNS, Raventos C, Tang J et al. (2015) Cerebrospinal fluid- pleural, or peritoneal involvement. We have shown derived circulating tumour DNA better represents the that fluids closer to metastatic sites are superior to genomic alterations of brain tumours than plasma. Nat blood for the detection of clinically relevant mutations Commun 6, 8839. and that genetic testing of these ‘nonblood’ fluids can de Wit S, Rossi E, Weber S, Tamminga M, Manicone M, be implemented in the routine clinical practice, being Swennenhuis JF, Groothuis-Oudshoorn CGM, Vidotto particularly useful in patients progressing to targeted R, Facchinetti A, Zeune LL et al. (2019) Single tube therapies. liquid biopsy for advanced non-small cell lung cancer. Int J Cancer 144, 3127-3237. Gainor JF, Dardaei L, Yoda S, Friboulet L, Leshchiner I, Acknowledgements Katayama R, Dagogo-Jack I, Gadgeel S, Schultz K, The authors thank patients, clinical, and laboratory Singh M et al. (2016) Molecular mechanisms of staff. The study reported in this publication was sup- resistance to first- and second-generation ALK ported by the ‘Centro para el Desarrollo Tecnol ogico inhibitors in ALK-rearranged lung cancer. Cancer e Industrial’ of the Spanish Ministry of Science, Inno- Discov 6, 1118–1133. vation and Universities under award number IDI- Goldman JW, Noor ZS, Remon J, Besse B and Rosenfeld N (2018) Are liquid biopsies a surrogate for tissue 20170023. The funding source had no role in study EGFR testing? Ann Oncol 29, i38–i46. design; in the collection, analysis, and interpretation of Gonzalez-Cao M, de Las Mayo, Casas C, Jordana Ariza data; in the writing of the report; and in the decision N, Manzano JL, Molina-Vila MA, Soriano V, to submit the article for publication Puertolas T, Balada A, Soria A et al. (2018) Early evolution of BRAFV600 status in the blood of Conflict of interest melanoma patients correlates with clinical outcome and The authors declare no conflicts of interest. identifies patients refractory to therapy. Melanoma Res 28, 195–203. Gonzalez-Cao M, Mayo-de-Las-Casas C, Molina-Vila MA, Author contributions De Mattos-Arruda L, Munoz-Couselo E, Manzano JL, Cortes J, Berros JP, Drozdowskyj A, Sanmamed M SV, CMC, NJA, MGI, RR, and MAMV conceived et al. (2015) BRAF mutation analysis in circulating and designed the study. CMC, NJA, MGI, BGP, RC, free tumor DNA of melanoma patients treated with MJC, and MAMV conducted the experiments, BRAF inhibitors. Melanoma Res 25, 486–495. acquired, and analyzed data. SVR, IMH, MGC, UM, Hata A, Masago K, Katakami N, Imai Y and Yatabe Y AMB, NR, MM, RB, AB, XG, GT, NK, and RR (2014) Spatiotemporal T790M heterogeneity in a provided clinical samples and collected clinical infor- patient with EGFR-mutant non-small-cell lung cancer. mation. SV, CMC, and MAM wrote the manuscript. J Thorac Oncol 9, e64–65. All authors reviewed and/or revised the manuscript. Janku F and Kurzrock R (2016) Bringing blood-based molecular testing to the clinic. Clin Cancer Res 22, 5400–5402. References Janku F, Zhang S, Waters J, Liu L, Huang HJ, Subbiah V, Ballester LY, Glitza Oliva IC, Douse DY, Chen MM, Lan Hong DS, Karp DD, Fu S, Cai X et al. (2017) C, Haydu LE, Huse JT, Roy-Chowdhuri S, Luthra R, Development and validation of an ultradeep next- Wistuba II et al. (2018) Evaluating circulating Tumor generation sequencing assay for testing of plasma cell- DNA from the cerebrospinal fluid of patients with free DNA from patients with advanced cancer. Clin melanoma and leptomeningeal disease. J Neuropathol Cancer Res 23, 5648–5656. Exp Neurol 77, 628–635. Kang JY, Park CK, Yeo CD, Lee HY, Rhee CK, Kim SJ, Bedin C, Enzo MV, Del Bianco P, Pucciarelli S, Nitti D Kim SC, Kim YK, Park MS and Yim HW (2015) and Agostini M (2017) Diagnostic and prognostic role Comparison of PNA clamping and direct sequencing of cell-free DNA testing for colorectal cancer patients. for detecting KRAS mutations in matched tumour Int J Cancer 140, 1888–1898. tissue, cell block, pleural effusion and serum from Chic N, Mayo-de-Las-Casas C and Reguart N (2017) patients with malignant pleural effusion. Respirology Successful treatment with gefitinib in advanced non- 20, 138–146. small cell lung cancer after acquired resistance to Karachaliou N, Mayo-de las CasasC, Queralt C, de osimertinib. J Thorac Oncol 12, e78–e80. Aguirre I, Melloni B, Cardenal F, Garcia-Gomez R, De Mattos-Arruda L, Mayor R, Ng CK, Weigelt B, Massuti B, Sanchez JM et al. (2015) Association of Martinez-Ricarte F, Torrejon D, Oliveira M, Arias A, EGFR L858R Mutation in Circulating Free DNA Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2643 Detection of mutations in fluids of cancer patients S. Villatoro et al. With Survival in the EURTAC Trial. JAMA Oncol 1, Nevel KS, Wilcox JA, Robell LJ and Umemura Y (2018) 149–157. The utility of liquid biopsy in central nervous system Kawahara A, Abe H, Murata K, Ishii H, Azuma K, malignancies. Curr Oncol Rep 20, 60. Takase Y, Hattori S, Naito Y and Akiba J (2019). Niederst MJ, Hu H, Mulvey HE, Lockerman EL, Garcia Screening system for epidermal growth factor receptor AR, Piotrowska Z, Sequist LV and Engelman JA mutation detection in cytology cell-free DNA of (2015) The allelic context of the C797S mutation cerebrospinal fluid based on assured sample quality. acquired upon treatment with third-generation EGFR Cytopathology 30, 144-149. inhibitors impacts sensitivity to subsequent treatment Li YS, Jiang BY, Yang JJ, Zhang XC, Zhang Z, Ye JY, strategies. Clin Cancer Res 21, 3924–3933. Zhong WZ, Tu HY, Chen HJ, Wang Z et al. (2018) Panditharatna E, Kilburn LB, Aboian MS, Kambhampati Unique genetic profiles from cerebrospinal fluid cell- M, Gordish-Dressman H, Magge SN, Gupta N, free DNA in leptomeningeal metastases of EGFR- Myseros JS, Hwang EI, Kline C et al. (2018) Clinically mutant non-small-cell lung cancer: a new medium of relevant and minimally invasive tumor surveillance of liquid biopsy. Ann Oncol 29, 945–952. pediatric diffuse midline gliomas using patient-derived Liu D, Lu Y, Hu Z, Wu N, Nie X, Xia Y, Han Y, Li Q, liquid biopsy. Clin Cancer Res 24, 5850–5859. Zhu G and Bai C (2014) Malignant pleural effusion Pentsova EI, Shah RH, Tang J, Boire A, You D, Briggs S, supernatants are substitutes for metastatic pleural Omuro A, Lin X, Fleisher M, Grommes C et al. (2016) tumor tissues in EGFR mutation test in patients with Evaluating cancer of the central nervous system advanced lung adenocarcinoma. PLoS ONE 9, e89946. through next-generation sequencing of cerebrospinal Liu L, Shao D, Deng Q, Tang H, Wang J, Liu J, Guo F, fluid. J Clin Oncol 34, 2404–2415. Lin Y, Peng Z, Mao M et al. (2018) Next generation R Core Team (2012) R: A Language and Environment for sequencing-based molecular profiling of lung Statistical Computing. R Foundation for Statistical adenocarcinoma using pleural effusion specimens. J Computing, Vienna, Austria. Thorac Dis 10, 2631–2637. Rolfo C, Mack PC, Scagliotti GV, Baas P, Barlesi F, Malapelle U, Mayo de-Las-CasasC, Rocco D, Garzon M, Bivona TG, Herbst RS, Mok TS, Peled N, Pirker R Pisapia P, Jordana-Ariza N, Russo M, Sgariglia R, De et al. (2018) Liquid biopsy for advanced non-small cell Luca C et al. (2017) Development of a gene panel for lung cancer (NSCLC): a statement paper from the next-generation sequencing of clinically relevant IASLC. J Thorac Oncol 13, 1248–1268. mutations in cell-free DNA from cancer patients. Br J Sasaki S, Yoshioka Y, Ko R, Katsura Y, Namba Y, Cancer 116, 802–810. Shukuya T, Kido K, Iwakami S, Tominaga S and Mayo-de-Las-Casas C, Garzon Ibanez M, Jordana-Ariza Takahashi K (2016) Diagnostic significance of N, Garcia-Pelaez B, Balada-Bel A, Villatoro S, cerebrospinal fluid EGFR mutation analysis for Malapelle U, Karachaliou N, Troncone G, Rosell R leptomeningeal metastasis in non-small-cell lung cancer et al. (2018) An update on liquid biopsy analysis for patients harboring an active EGFR mutation following diagnostic and monitoring applications in non-small gefitinib therapy failure. Respir Investig 54, 14–19. cell lung cancer. Expert Rev Mol Diagn 18, 35–45. Seoane J, De Mattos-Arruda L, Le Rhun E, Bardelli A Mayo-de-Las-Casas C, Jordana-Ariza N, Garzon-Ibanez and Weller M. (2019). Cerebrospinal fluid cell-free M, Balada-Bel A, Bertran-Alamillo J, Viteri-Ramirez tumour DNA as a liquid biopsy for primary brain S, Reguart N, Munoz-Quintana MA, Lianes-Barragan tumours and central nervous system metastases. Ann P, Camps C et al. (2017) Large scale, prospective Oncol 30, 211-218. screening of EGFR mutations in the blood of Shin S, Kim J, Kim Y, Cho SM and Lee KA (2017) advanced NSCLC patients to guide treatment Assessment of real-time PCR method for detection of decisions. Ann Oncol 28, 2248–2255. EGFR mutation using both supernatant and cell pellet Melms JC, Ho KW, Thummalapalli R, Tyler J, Brinker TJ, of malignant pleural effusion samples from non-small- Singh V, Sengupta S, Mier J and Izar B (2018) cell lung cancer patients. Clin Chem Lab Med 55, Implementation of cell-free tumor DNA sequencing 1962–1969. from the cerebrospinal fluid to guide treatment in a Szallasi Z (2017) Detecting mutant KRAS in liquid patient with primary leptomeningeal melanoma: a case biopsies: a biomarker searching for a role. Ann Oncol report. Mol Clin Oncol 9, 58–61. 28, 677–678. Mouliere F, Mair R, Chandrananda D, Marass F, Smith Tsui DWY, Murtaza M, Wong ASC, Rueda OM, Smith CG, Su J, Morris J, Watts C, Brindle KM and CG, Chandrananda D, Soo RA, Lim HL, Goh BC, Rosenfeld N (2018) Detection of cell-free DNA Caldas C et al. (2018) Dynamics of multiple resistance fragmentation and copy number alterations in mechanisms in plasma DNA during EGFR-targeted cerebrospinal fluid from glioma patients. EMBO Mol therapies in non-small cell lung cancer. EMBO Mol Med 10, e9323. Med 10, e7945. 2644 Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. S. Villatoro et al. Detection of mutations in fluids of cancer patients Ulz P, Heitzer E, Geigl JB and Speicher MR (2017) Patient status of paired cerebrospinal fluid and plasma samples monitoring through liquid biopsies using circulating in EGFR mutant non-small cell lung cancer with tumor DNA. Int J Cancer 141, 887–896. leptomeningeal metastases. Cancer Chemother Xu Y, Hu M, Zhang M, Zhong W, Yin X, Sun Y, Chen Pharmacol 78, 1305–1310. M, Zhao J, Si X, Wang H et al. (2018) Prospective study revealed prognostic significance of responses in leptomeningeal metastasis and clinical value of Supporting information cerebrospinal fluid-based liquid biopsy. Lung Cancer Additional supporting information may be found 125, 142–149. online in the Supporting Information section at the end Yang H, Cai L, Zhang Y, Tan H, Deng Q, Zhao M and of the article. Xu X (2014) Sensitive detection of EGFR mutations in Fig. S1. Examples of linear plots used for absolute (A) cerebrospinal fluid from lung adenocarcinoma patients and relative (B) quantification of mutant and wt alleles with brain metastases. J Mol Diagn 16, 558–563. in liquid biopsy samples. Yi X, Ma J, Guan Y, Chen R, Yang L and Xia X (2017) Table S1. Overview of reports on detection of muta- The feasibility of using mutation detection in ctDNA to tions and other genetic alterations in cfDNA purified assess tumor dynamics. Int J Cancer 140, 2642–2647. from fluids of advanced cancer patients. Zhang S, Anjum R, Squillace R, Nadworny S, Zhou T, Table S2. Cell lines used as positive controls in the Keats J, Ning Y, Wardwell SD, Miller D, Song Y et al. (2016) The potent ALK inhibitor brigatinib PNA-Q-PCR assay. Negative controls were genomic (AP26113) overcomes mechanisms of resistance to first- DNAs from the BxPC-3 or the PC-3 cell lines. and second-generation ALK inhibitors in preclinical Table S3. Mutations detected by the PNA-Q-PCR models. Clin Cancer Res 22, 5527–5538. assay. Zhao J, Ye X, Xu Y, Chen M, Zhong W, Sun Y, Yang Z, Table S4. Comparison of the mutational status in flu- Zhu G, Gu Y and Wang M (2016) EGFR mutation ids and plasma in paired samples. Molecular Oncology 13 (2019) 2633–2645 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2645