NGS-based liquid biopsy profiling identifies mechanisms of resistance to ALK inhibitors: a step toward personalized NSCLC treatment Estela Sanchez-Herrero1,2, Roberto Serna-Blasco1, Vadym Ivanchuk1, Rosario Garcıa-Campelo3, Manuel Do  mine Go  mez4, Jose M. Sa nchez5, Bartomeu Massutı6, Noemi Reguart7, Carlos Camps 8,9,10,11 1 , Sandra Sanz-Moreno , Silvia Calabuig-Farin ~ as8,9,12, Eloısa Jantus-Lewintre8,9,13, 14 Magdalena Arnal , Dietmar Ferna ndez-Orth 15 , Virginia Calvo16, Vıctor Gonza lez-Rumayor2, 1,16 1,16 Mariano Provencio and Atocha Romero 1 Liquid Biopsy Laboratory, Biomedical Sciences Research Institute Puerta de Hierro-Majadahonda, Spain 2 Atrys Health, Barcelona, Spain 3 Medical Oncology Department, Complejo Hospitalario Universitario A Corun ~a, Spain 4 Medical Oncology Department, Hospital Universitario Fundacion Jimenez Dıaz, Oncohealth Institute, Universidad Auto noma de Madrid, Spain 5 Medical Oncology Department, Hospital La Princesa, Madrid, Spain 6 Medical Oncology Department, Hospital Universitario de Alicante, ISABIAL, Alicante, Spain 7 Medical Oncology Department, Hospital Clinic of Barcelona, Spain 8 Molecular Oncology Laboratory, Fundacion Hospital General Universitario de Valencia, Spain 9 CIBERONC, Valencia, Spain 10 Department of Medical Oncology, Hospital General Universitario de Valencia, Spain ncia, Spain 11 Department of Medicine, Universitat de Vale ncia, Spain 12 Department of Pathology, Universitat de Vale ncia, Spain 13 Department of Biotechnology, Universitat de Vale 14 MARGenomics, IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain 15 European Genome-phenome Archive, Centre for Genomic Regulation (CRG), Barcelona, Spain 16 Medical Oncology Department, Hospital Universitario Puerta de Hierro-Majadahonda, Spain Keywords Despite impressive and durable responses, nonsmall cell lung cancer ALK-TKI; EML4-ALK; liquid biopsy; NGS; (NSCLC) patients treated with anaplastic lymphoma kinase (ALK) inhibi- NSCLC tors (ALK-Is) ultimately progress due to development of resistance. Here, we have evaluated the clinical utility of circulating tumor DNA (ctDNA) Correspondence A. Romero, Medical Oncology Department, profiling by next-generation sequencing (NGS) upon disease progression. Hospital Puerta de Hierro, Calle Joaquın We collected 26 plasma and two cerebrospinal fluid samples from 24 Rodrigo, 1, 28222 Majadahonda, Madrid, advanced ALK-positive NSCLC patients at disease progression to an Spain ALK-I. These samples were analyzed by NGS and digital PCR. A tool to Tel: +341917769 retrieve variants at the ALK locus was developed (VALK tool). We identi- E-mail: atocha10@hotmail.com fied at least one resistance mutation in the ALK locus in ten (38.5%) plasma samples; the G1269A and G1202R mutations were the most preva- (Received 13 January 2021, revised 21 May 2021, accepted 28 May 2021, available lent among patients progressing to first- and second-generation ALK-Is, online 18 June 2021) respectively. Overall, 61 somatic mutations were detected in 14 genes: TP53, ALK, PIK3CA, SMAD4, MAP2K1 (MEK1), FGFR2, FGFR3, doi:10.1002/1878-0261.13033 Abbreviations AF, allele frequency; ALK, anaplastic lymphoma kinase; ALK-Is, Anaplastic lymphoma kinase inhibitors; cfDNA, cell-free DNA; CNS, central nervous system; CNVs, copy-number variations; CSF, cerebrospinal fluid; CTC, circulating tumor cell; dPCR, digital PCR; ECOG, cooperative oncology group; LOD, limit of detection; MAF, mutant allele frequency; MNP, multiple-nucleotide polymorphism; NGS, next-generation sequencing; NPA, negative percentage agreement; NSCLC, nonsmall cell lung cancer; ORA, overall rates of agreement; OS, overall survival; PFS, progression-free survival; PPA, positive percentage agreement; SNVs, single nucleotide variants; UMIs, unique molecular identifiers; wt, wild-type. Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2363 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. NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. BRAF, EGFR, IDH2, MYC, MET, CCND3, and CCND1. Specifically, a deletion in exon 19 in EGFR, a non-V600 BRAF mutation (G466V), and the F129L mutation in MAP2K1 were identified in four patients who showed no objective survival benefit from ALK-Is. Potential ALK-I-resistance mutations were also found in PIK3CA and IDH2. Finally, a c-MYC gain, along with a loss of CCND1 and FGFR3, was detected in a patient progress- ing on a first-line treatment with crizotinib. We conclude that NGS analysis of liquid biopsies upon disease progression identified different putative ALK-I-resistance mutations in most cases and could be a valuable approach for therapy decision making. than genotyping solid biopsies. Next-generation 1. Introduction sequencing (NGS) enables the interrogation of a large Anaplastic lymphoma kinase (ALK) inhibitors (ALK- number of mutations and can be used with liquid Is) have dramatically improved outcomes of nonsmall biopsies [15]. cell lung cancer (NSCLC) patients whose tumors har- In this study, we have analyzed 26 plasma and two bor an ALK translocation [1,2]. A broad therapeutic cerebrospinal fluid (CSF) samples, from 24 patients, arsenal is currently available to treat ALK-positive collected upon disease progression while being treated NSCLC tumors, and sequential treatment with differ- with an ALK-I, in order to determine the clinical util- ent ALK-Is is the best therapeutic option for NSCLC ity of liquid biopsies for ALK-Is sequencing. In addi- patients with an ALK translocation [3,4]. However, it tion, we provide a pipeline specifically designed to remains unclear how ALK-Is should be sequenced. It detect somatic mutations in the ALK domain. Finally, has been proposed that treatment sequencing can be we evaluate the clinical significance of somatic alter- established according to clinical characteristics of the ations in genes other than ALK. patients or toxicity profile. In this way, second-genera- tion ALK-Is have shown impressive central nervous 2. Methods system (CNS) efficacy in ALK-positive NSCLC patients [5,6]. On the other hand, crizotinib is associ- ated with adverse events dominated by gastrointestinal 2.1. Study population and visual effects, increased transaminases and edema, whereas the most common adverse events of alectinib Between June 2015 and July 2019, 24 stage IV, ALK- are anemia, myalgia, and increased blood bilirubin [7]. positive NSCLC patients progressing on an ALK-I The tumor molecular profile can also determine the (the ALK cohort), were prospectively recruited from treatment response. In this way, among patients trea- six hospitals across Spain. The study protocol was ted with the ALK-I lorlatinib, the ALK fusion variant approved by the Hospital Puerta de Hierro Ethics 3 was associated with significantly longer progression- Committee (internal code 79-18) and was conducted in free survival (PFS) than variant 1 [8]. Finally, several accordance with the precepts of the Code of Ethics of resistance mutations have been identified upon pro- The World Medical Association (Declaration of Hel- gression to an ALK-I [9,10]. While some of them con- sinki). All patients provided their appropriate written fer resistance to specific ALK-Is, others do not [11–13]. informed consent to participate in the study prior to Paradoxically, therapy is seldom decided based on the enrollment. Briefly, patients who were 18+ years of age tumor molecular profile upon disease progression, and and with a pathologically confirmed diagnosis of stage ALK-Is are usually prescribed empirically. Conversely, IV NSCLC with an EML4-ALK translocation were biomarker testing after treatment failure is routinely eligible for inclusion. The identification of EML4-ALK performed in EGFR-positive NSCLC patients, as rec- rearrangement was carried out by different methodolo- ommended by clinical guidelines [14]. gies, according to each participant center (Table S1). Repeat tumor biopsy upon disease progression is Specifically the ALK testing was performed by not always feasible. Nevertheless, there is considerable immunohistochemistry (IHC) using the Ventana ALK evidence showing that genotyping cell-free DNA (D5F3) CDx assay on a Ventana BenchMark XT (cfDNA) is a valid and significantly faster approach automated slide-processing system (Ventana Medical 2364 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. S anchez-Herrero et al. NGS-based liquid biopsy profiling in ALK+ NSCLC Systems, Tucson, AZ, USA) or the mAb ALK (5A4) calling, annotation, and filtering were performed on (NovocastraTM; Leica Biosystems, Newcastle Upon the Ion Reporter (v5.10, Thermo Fisher) platform Tyne, UK); Fluorescence In Situ Hybridization (FISH) using the Oncomine TagSeq Pan-Cancer Liquid Biopsy using Vysis LSI ALK Dual Color Break Apart FISH workflow (v2.1, Thermo Fisher). All candidate muta- probe kit (Vysis, Downers Grove, IL, USA) or ALK tions were manually reviewed using the Integrative (2P23) Break Apart FISH probe kit (cYTOtEST, Genomics Viewer v.2.3.40, (Broad Institute, Cam- Rockville, MD, USA). The nCounter analysis system bridge, MA, USA). The clinical significance of somatic (NanoString Technologies, Seattle, WA, USA) was variants was determined according to the Standards also used to detect EML4-ALK translocation. All and Guidelines for the Interpretation and Reporting of plasma samples were collected upon disease progres- Sequence Variants in Cancer [16]. sion, which was assessed according to RECIST criteria The mutations identified by NGS were confirmed by v.1.1. In total, 26 plasma and two CSF specimens were digital PCR (dPCR) using a QuantStudio 3D dPCR collected and analyzed. System (Applied Biosystems, South San Francisco, CA, USA). In accordance with the manufacturer’s specifications, dPCR reactions were performed in an 2.2. Laboratory procedures 18-lL volume comprising 9 lL of 20X QuantStudio Peripheral whole-blood samples were collected in a 10- 3D Master Mix, 0.45 lL of 409 commercially avail- mL Streck cfDNA BCT (Streck, Omaha, NE, USA) able predesigned or custom TaqMan assays and tube for cfDNA. CSF samples were collected in a 10- 8.55 lL of cfDNA (minimum amount 2 ng). Subse- mL sterile tube with no additives or anticoagulants. quently, 14.5 lL of the PCR reaction was loaded onto CSF samples were taken from patients progressing at a QuantStudio 3D dPCR 20K chip using QuantStu- the brain level. Samples were centrifuged at room tem- dioTM 3D dPCR Chip Loader. Each dPCR run perature in two consecutive centrifugations of 1500 g included a negative control DNA, as a wild-type (wt) for 10 min and 5000 g for 20 min in order to separate control, a blank (with no cfDNA) and a positive con- plasma or CSF from the cellular fraction. cfDNA was trol. PCR reactions were performed in a thermal cycler isolated using QIAamp Circulating Nucleic Acid Kit (Applied Biosystems) at 96 °C for 10 min, then 40 (QIAgen, Valencia, CA, USA) according to the manu- cycles at 56 °C for 2 min and 98 °C for 30 s, and a facturer’s instructions. Libraries were prepared from at final elongation step at 72 °C for 10 min. Finally, sam- least 15ng of input cfDNA using the OncomineTM Pan- ples were maintained at 22 °C for at least 30 min. Cancer Cell-Free Assay kit (Thermo Fisher, Palo Alto, Chips fluoresce was read twice two independent CA, USA) according to the manufacturer’s instruc- QuantStudioTM 3D dPCR Instruments. Results were tions. According to manufacturer, this amplicon-based visualized and analyzed using QuantStudio 3D Anal- targeted sequencing assay allows the detection of mul- ysis SuiteTM Cloud Software (Thermo Fisher). The tiple variants in ctDNA isolated from liquid biopsy automatic call assignments for each data cluster were samples with a limit of detection (LOD) down to manually adjusted when needed. The MAF was calcu- 0.1% mutant allele frequency (MAF). Specifically, the lated as the ratio of mutant DNA molecules to the panel interrogates variants in 52 genes (Table S2). sum of mutant and wt DNA molecules. AMPureXP magnetic beads (Beckman Coulter, Inc., The LOD and limit of quantitation (LOQ) of the Brea, CA, USA) were used to purify all libraries. Sub- dPCR assays for ALK variants were evaluated for four sequently, the individual libraries were quantified using custom TaqMan assays by mixing DNA from two the Ion Library TaqMan Quantitation Kit (Thermo different plasmids an one patient sample harboring the Fisher, Palo Alto, CA, USA) in a StepOnePlusTM specific ALK mutations in a background of wt DNA qPCR machine (Thermo Fisher) and adjusted to a (from healthy donors) at different mutant allele con- final concentration of 50 pM. Eight samples were centrations (i.e., 1%, 0.5%, 0.1%, 0.05%). LOD and pooled. Templating and Ion 550TM Chip loading were LOQ were estimated for an input of cfDNA of 7 ng. carried out with an Ion ChefTM System (Thermo Specifically, LOD and LOQ for ALK G1202R assay Fisher). Finally, an Ion GeneStudioTM S5 Sequencer were estimated using a patient sample. The mutation (Thermo Fisher) was used to sequence loaded Ion was previously confirmed by NGS. LOD and LOQ for 550TM chips. TORRENT SUITE Software (v5.12) was used ALK S1206Y were estimated using ALK_1 plasmid. to analyze the raw sequencing data. The COVERAGE Finally, LOD and LOQ for ALK L1196M and ANALYSIS (v. 5.12.0.0) plugin was used for sequencing G1269A assays were estimated using ALK_3 plasmid. coverage analysis (Thermo Fisher). Raw reads were Plasmids were designed by Classic GeneArt Gene Syn- aligned to the human reference genome hg19. Variant thesis Portal (Thermo Fisher). Schematic plasmid Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2365 NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. maps are available upon request. LOD and LOQ were of PCR product was subjected to Sanger sequencing calculated based on the standard deviation of the using PCR primers and the BigDyeTM Terminator v1.1 response and the slope according to ICH Q2 (R1) Cycle Sequencing Kit (Applied Biosystems). The ther- guidelines (Validation of analytical procedures: text mal program included an initial step of 94 °C for and methodology). The standard deviation of the 10 min, follow by 25 cycles of denaturation at 94 °C response was calculated based on standard error of the for 10 s, annealing at 50 °C for 10 s and extension at y-intercept. Results of the detection sensitivity of Taq- 60 °C for 2 min. The PCR product was purified with Man assays are displayed in Data S1. Overall, mutant the protocol of the BigDye XTerminatorTM Purifica- allele frequencies correlated with the expected mutant tion Kit (Applied Biosystems) and sequenced with the allele frequencies in all cases (Pearson’s correlation forward and reverse PCR primers on ABI PRISM 310 coefficient, 0.9999, 0.9991, 0.9994, and 0.9994 for Genetic Analyzer (Applied Biosystems). Sequence data G1202R, S1206Y, L1196M, and G1269A, respec- were analyzed on SEQUENCING ANALYSIS Software 7 v7.0 tively). LOD were 0.05%, 0.23%, 0.19%, and 0.18% (Applied Biosystems). for G1202R, S1206Y, L1196M, and G1269A, respec- To increase the detection rate for variants at the tively. Additionally, 10 wt cfDNA from healthy ALK locus, we have developed a bioinformatic pipe- donors were used to evaluate the false-positive signals. line, called the VALK tool (available at GitHub: ALK mutations were not detected in any of the wt https://github.com/AtochaHUPH/VALK-tool-), which samples. is capable of fully automating the filtering generating Samples were considered to be positive when the a.csv file containing the output variants and a list of MAF was greater than or equal to 0.1% and when their properties. Specific conditions for single nucleo- there were at least 300 copiesmL 1 of wt DNA. A tide variants (SNVs), indels, multiple-nucleotide poly- negative control DNA (wt) and a blank sample (con- morphisms (MNP), fusions and copy-number variation taining no DNA) were included in every run. (CNV) calls were defined. The nonfiltered-on- In order to discard that mutations detected by NGS comine.tsv file with variants in ‘Variant Call Format’ were not clonal hematopoiesis-derived mutations was obtained for the 28 samples. All variants that have (specifically those in TP53), DNA from peripheral passed the Oncomine Variants (v.5.12, Thermo Fisher) blood mononuclear cells (PBMCs) was analyzed in 10 filter were included in the final analysis. In addition, samples by dPCR or Sanger sequencing (Table S3). To we performed a second filtering process in order to res- this aim, DNA was isolated from PBMCs using Max- cue other somatic variants ruled out by the Oncomine well RSC Whole Blood DNA Kit (Promega Corpo- Variants (v.5.12) filter. Specifically, parameters such as ration, Madison, WI, USA) according to the the overall error of the NGS assay, the LOD, the cov- manufacturer’s instructions. Custom primer sequences erage depth, the percentage of targeted bases to target specific TP53 regions (exon 2, exon 4, exon sequenced at that coverage depth, the total number of 7, and exon 8 splice site) were designed using the target reads covering a variant region, the number of Primer3plus software. PCR reactions were performed reads supporting a specific variant, and the clinical sig- in a 15-lL volume comprising 7.5 lL of 20X QuantS- nificance, among others, were taken into account for tudio 3D Master Mix, 5.5 lL of H2O, 0.5 lL of for- the selection of the different thresholds. Figure S1 ward primer, 0.5 lL of reverse primer, and 1 lL of shows the selection criteria based on certain variables isolated DNA. PCR amplifications were carry out in as presented in the nonfiltered-oncomine.tsv file. the VeritiPro Thermal Cycler (Applied Biosystems). All computations were performed in R v.3.6.3 (R The thermal program included an initial step of 94 °C Foundation for Statistical Computing, Vienna, Aus- for 10 min, follow by 40 cycles of denaturation at tria) using additional packages. Specifically, the Tcl/Tk 94 °C for 30 s, annealing at 60 °C for 1 min and package that was used to provide the end-user an intu- extension at 72 °C for 30 s, and finally, at 72 °C for itive graphical interface to carry out the entire filtering 7 min. Amplified DNA samples were subjected to elec- process. In addition, the Scales (v1.1.1; https://scales.r- trophoresis with 2% agarose gel, stained with GelRed lib.org/) package, which contains functions that con- TM Nucleic Acid Gel Stain, 10 000X in Water (Biotium, vert data values to perceptual properties, was used to Hayward, CA, USA), and examined under an UV transform the raw data obtained from the nonfiltered- transilluminator. Products revealing clear PCR bands oncomine.tsv files into interpretable values for the end- were purified by adding 2 µL of ExoSAP-IT reagent user. Positive and negative percentage agreement (PPA (GE Healthcare, Madison, WI, USA) for each 5 µL and NPA) and overall rates of agreement (ORA) of of preamplification volume and incubated for 15 min the VALK tool for detecting the ALK mutations speci- at 37 °C followed by 15 min at 80 °C. One microliter fied in Table S4 were calculated considering the 2366 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. S anchez-Herrero et al. NGS-based liquid biopsy profiling in ALK+ NSCLC imperfect reference standard the dPCR result and Table 1. Baseline characteristics of the study cohort. using the two independent data sets, the ALK cohort Feature N % and the Valencia cohort, which consists of 54 cfDNA samples from NSCLC patients. Age of diagnosis Median (range) 53 (36– (years) 72) Sex Male 10 41.7 2.3. Statistical analysis Female 14 58.3 Smoking status Current smoker 3 12.5 Median follow-up was estimated by the reverse Ex-smoker 6 25 Kaplan–Meier method. Overall survival (OS) was Never-smoker 15 62.5 defined as the time from the start of treatment with an ECOG performance 0 12 50 ALK inhibitor to death or last follow-up. PFS was status 1 11 45.8 2 1 4.2 defined as the time between the start of an ALK inhi- Histology Adenocarcinoma 23 95.8 bitor and disease progression (as ascertained by Neuroendocrine 1 4.2 RECIST criteria), death, or the censored date of the carcinoma last assessment, whichever occurred first. The log-rank Clinical stage at III 6 25 test was used to assess statistical differences between diagnosis IV 18 75 Kaplan–Meier survival curves. Hazard ratios were esti- mated from the Cox model using a multivariable approach adjusted for sex, cooperative oncology group (ECOG) performance status, and lines of ALK-TKI. 11.6 months (95% CI: 6.5–20.9 months) and Association between clinicopathological variables and 24.6 months (95% CI: 11.8–NR months), respectively. genomic features was assessed by Mann–Whitney and In addition, 12 samples corresponded to patients who Fisher’s exact test as needed. Values of P < 0.05 were had received previously crizotinib and were treated considered statistically significant. Statistical analyses with a second-generation ALK-I. Finally, the cohort were performed using Stata 15.1 (Stata Corporation, included two samples from patients progressing on lor- College Station, TX, USA). latinib after failure of a prior second-generation ALK- I and one patient progressing on alectinib who had previously received crizotinib plus two second-genera- 3. Results tion ALK-I. Detailed information about treatment lines is presented in Table S5. The median PFS and OS for patients progressing on a second or subsequent 3.1. Study cohort line with an ALK-I were 5.4 months (95% CI: 2–9.1) We collected and analyzed 26 plasma and two CSF and 11.2 months (95% CI: 3–NR) months, respec- specimens from 24 metastatic patients diagnosed with tively. an ALK-positive NSCLC who were progressing on an ALK-I. Baseline clinicopathological characteristics of 3.2. Next-generation sequencing analysis upon the study population (N = 24) are presented in disease progression Table 1. The median age at diagnosis was 53 years (range, 36–72 years) and 58.3% were females. The Overall plasma samples yielded higher cfDNA concen- majority of the patients were never smokers (62.5%) trations than CSF. Despite the small sample size of and the most frequent histology was adenocarcinoma the study cohort, the median concentration of cfDNA (95.8%). ECOG Performance Status at study entry isolated from CSF was significantly lower varied from 0 to 2. As shown in Fig. 1, two samples (0.20 nglL 1) than median concentration of cfDNA from two patients were collected upon disease progres- obtained from plasma samples (1.94 nglL 1; sion while on two consecutive lines of treatment with P = 0.039; calculated by Mann–Whitney test). an ALK-I; three samples were obtained from one The average mapped reads per sample was patient upon failure to three consecutive lines of ALK- 9 775 623, resulting in a median overall sequencing I; all 21 other members of the cohort each provided a depth of 25 322. The median read coverage per sample single sample. was 21 261 and the median molecular coverage per As presented in Fig. 1, 13 samples corresponded to sample was 1670.8. Regarding the two CSF samples, ALK-I-na€ıve patients who progressed on a first-line taken from patients with CNS disease, the average crizotinib (N = 11) or alectinib (N = 2) treatment. For mapped reads per sample was 8 268 194, resulting in a these patients, the median PFS and OS were median overall sequencing depth of 9318. Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2367 NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. Fig. 1. Swimmer chart showing the individual treatment responses of the study cohort. Blue, red, purple, green, and orange bars correspond to response duration of patients who were treated with crizotinib, alectinib, ceritinib, brigatinib, and lorlatinib, respectively, and from whom a blood sample was taken at the time of disease progression (syringe icon). Gray bars denote treatments responses to therapy for which samples at disease progression could not be analyzed (out of study). Tumor progression is denoted by triangle, and patient’s death is denoted by a squared. Mutations in BRAF, EGFR, and MAP2K1 are indicated by a lightning. In total, 61 somatic variants in ctDNA from 24 further tested in PBMCs in order to discard a clonal samples were detected. In three patients, no mutations hematopoietic origin (Table S3). were found. One of the patients with undetectable plasma ctDNA had progressed exclusively at the brain 3.3. Identification of acquired resistance level. The average number of mutations per patient mutations in the ALK locus upon disease was 2.18 and the median MAF was 0.39%, with a progression minimum MAF of 0.02% and a maximum of 5.09%. As expected, SNPs were the most frequent mutation To increase the sensitivity for detecting somatic muta- type (N = 48). In addition, we identified 10 indels and tions in the ALK locus, we developed an algorithm three CNVs (Fig. 2). Specifically, a c-MYC gain in named VALK tool. The tool has been specifically conjunction with a CCND1 and an FGFR3 loss was designed for the analysis of the NGS data obtained detected in a patient progressing on a first line with from liquid biopsies. Among other parameters, the crizotinib. This patient also harbored a mutation in algorithm takes into account the molecular depth and TP53 (Fig. 2). molecular counts as well as specific regions that are As illustrated in Fig. 2, somatic mutations were more likely for false-positive calls. In order to test the detected in 14 genes: TP53, ALK, PIK3CA, SMAD4, analytical performance of the tool, all SNPs in the MAP2K1 (MEK) FGFR2, FGFR3, BRAF, EGFR, ALK locus that were present in the nonfiltered-on- IDH2, MYC, MET, CCND3, and CCND1. All the comine.tsv file were analyzed by dPCR. In total, 19 variants detected are listed in Table S6. Thirteen vari- ALK variants from 22 samples were evaluated ants (12 of which were in the ALK locus and one was (Table S4). Considering the nonreference standard the in the EGFR gene) were categorized as being of strong dPCR result PPA, NPA, and overall percent agree- clinical significance. Mutations in TP53 gene were ment of ALK mutation detection for the VALK tool 2368 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. S anchez-Herrero et al. NGS-based liquid biopsy profiling in ALK+ NSCLC Fig. 2. Co-mutation plot according to ctDNA profiling by NGS. Each column represents data from a single patient. Each row represent data from a specific gene. Indels are represented in blue whereas missense mutations are represented in green. CNVs are colored in orange. Heat map at the bottom illustrates type of treatment received. were 67% (95% CI: 35–90%), 93% (95% CI: 75– case of the patient progression on ceritinib (patient 5), 99%), and 85% (95% CI: 69–94%), respectively this mutation was not detected at treatment initiation (Table S7). To validate the final algorithm (locked confirming its role as a resistance mutation prior to analyses), a second independent batch of 54 (Table S11). In addition, the S1206Y mutation was samples provided by a different laboratory (Valencia detected along with the G1202R mutation in one of cohort) was used. In this case, the PPA, NPA, and the aforementioned alectinib-progressing patients. This ORA were 100% (95% CI: 29–100%), 98% (95% CI: patient had been treated with crizotinib before initiat- 90–100%), and 98% (95% CI: 90–100%), respectively. ing alectinib treatment. The low MAF of the S1206Y Cross-tables describing PPA, NPA, and ORA for the mutation suggests that it could be responsible for the VALK tool as well as for the Oncomine variants v5.10 previous crizotinib failure and it could have been using ALK and Valencia cohorts are presented in decreased since alectinib therapy. In addition, the Table S7–S10. G1269A mutation was detected upon crizotinib failure Overall, in the ALK cohort we detected at least one in two cases and the L1196M mutation was identified ALK mutation in 10 (38.5%) plasma samples collected after progression to crizotinib and lorlatinib. The latter upon disease progression (Table 2, and Fig. S2). Nota- was detected together with the R1275Q mutation in a bly, the Oncomine variants v5.10 filter only detected patient diagnosed with an ALK-positive neuroen- ALK mutations in three patients (Table 2). docrine carcinoma (patient 19). Of note, neither the The G1202R mutation was identified in four L1196M mutation nor the R1275Q mutation was patients who had progressed on alectinib (N = 3) and detected in the pre-treatment plasma sample from ceritinib (N = 1; patient number 5). Specifically, in patient 19 (Table S11). Finally, the A1200V mutation Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2369 NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. Table 2. Somatic mutations detected at the ALK locus upon treatment failure. MAF No. of Nucleotide Amino acid Progression Line of MAF NGS dPCR Total dPCR patient change change to treatment Filter (%) (%) input (ng) Patient 2 c.3599C>T p.A1200V Crizotinib First line nonfiltered- 0.02 0.04 42.07 oncomine.tsv Patient 3 c.3806G>C p.G1269A Crizotinib First line Oncomine 5.10/ 3.36 2.82 4.64 VALK tool Patient 5 c.3806G>C p.G1269A Crizotinib First line Oncomine 5.10/ 0.88 0.42 7.20 VALK tool Patient 5 c.3604G>A p.G1202R Ceritinib Second line Oncomine 5.10/ 1.28 2.12 54.04 VALK tool Patient 6 c.3586C>A p.L1196M Crizotinib First line nonfiltered- 0.02 0.06 27.10 oncomine.tsv Patient 16 c.3617C>A p.S1206Y Alectinib Second line VALK tool 0.06 0.01 26.50 Patient 16 c.3604G>A p.G1202R Alectinib Second line VALK tool 0.04 0.04 26.50 Patient 19 c.3586C>A p.L1196M Lorlatinib Third line nonfiltered- 0.05 0.05 29.24 oncomine.tsv Patient 19 c.3824G>A p.R1275Q Lorlatinib Third line nonfiltered- 0.03 0.04 21.80 oncomine.tsv Patient 20 c.3604G>A p.G1202R Alectinib Second line VALK tool 0.05 0.32 14.11 Patient 22 c.3604G>A p.G1202R Alectinib Fourth line VALK tool 0.03 0.01 48.05 Patient 24 c.3538G>C p.V1180L Alectinib First line nonfiltered- 0.37 0.35 16.67 oncomine.tsv (N = 1) was detected in a sample collected after crizo- BRAF mutation was absent in the plasma sample col- tinib failure, and the V1180L mutation (N = 1) was lected before ceritinib initiation (Table S11). Likewise, detected in a patient progressing on a first-line treat- two patients harboring the F129L mutation in ment with alectinib (Table 2). MAP2K1 (MEK1) obtained little benefit from second- line ALK-I (Fig. 1, Table 3). This mutation was detected upon disease progression to alectinib (patient 3.4. Other molecular mechanisms underlying 10) and lorlatinib (patient 23). In case of patient 10, resistance to ALK-I we could confirm that this mutation was not present A deletion in exon 19 of the EGFR gene, a non-V600 in the pre-treatment plasma sample and tumor biopsy BRAF mutation, and the F129L mutation in MAP2K1 (Table S11). Noteworthy, the median PFS and OS for (MEK1) were identified in four patients who showed second-line treatment for these patients were less than no objective survival benefit from ALK-Is. None of one month (0.97) and 3 months, respectively, whereas these patients had a secondary mutation in ALK locus. median the PFS and OS for patients progressing on a Notably, the patient harboring the E746_A750del second or subsequent lines with an ALK-I but without mutation in the EGFR gene had a PFS time of mutations in MAP2K1 were 5.9 and 11.2 months (P 1.8 months under first-line crizotinib treatment. The log-rank < 0.05 in both cases; Fig. S3). patient was subsequently treated with alectinib but Potential ALK-I resistance mutations were also found tumor progression was assessed 3.1 months later in IDH2, PIK3CA, and MYC (Table 3). Specifically, the prompting a switch of treatment to lorlatinib, but that oncogenic mutations E545K and E545A in the PIK3CA also failed after 1.8 months, suggesting that the tumor gene were detected in the plasma sample of two patients had primary resistance to ALK-Is (Fig. 1, Table 3). progressing on ceritinib and brigatinib (Table 3). These Similarly, a non-V600 BRAF mutation, namely mutations were not detected in the pre-treatment sample G466V, was identified in the CFS collected upon dis- (Table S11). Likewise, the gain-of-function mutation in ease progression to ceritinib. While PFS with first-line IDH2, R140Q, was detected upon disease progression crizotinib was 21 months, disease progression was to first-line alectinib treatment. Finally, as previously assessed within 3 months of starting second-line ceri- mentioned, a c-MYC amplification was detected jointly tinib treatment (Fig. 1). Remarkably, the non-V600 with a loss of CCND1 and of FGFR3. 2370 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. S anchez-Herrero et al. Table 3. Resistance mutations detected in loci other than ALK upon tumor progression. Treatment HUGO Amino acid Variant class Patient Treatment line Sample symbol change Nucleotide change/CNV Type (Tier) Transcript rs COSM Patient Alectinib 1st Sample IDH2 p.R140Q c.419G>A SNV Potential NM_002168.3 – COSM41590 8 10 clinical significance Patient Ceritinib 2nd Sample PIK3CA p.E545K c.1633G>A SNV Potential NM_006218 rs104886003 COSM763 14 16 clinical significance Patient Brigatinib 2nd Sample PIK3CA p.E545A c.1634A>C SNV Potential NM_006218 rs121913274 COSM12458 18 20 clinical significance Patient Ceritinib 2nd Sample BRAF p.G466V c.1397G>T SNV Potential NM_004333.4 rs121913351 COSM451 3 4 clinical significance Patient Alectinib 2nd Sample MAP2K1 p.F129L c.385T>C SNV Potential NM_002755.3 rs1057519805 COSM1570285 10 12 clinical significance Patient Crizotinib 1st Sample MYC Gain (3.08) CNV Potential NM_005359.5 – 12 14 clinical significance Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. Patient Crizotinib 1st Sample EGFR p.E746_A750del c.2235_2249delGGAATTAAGAGAAGC InDel Strong NM_005228.4 – COSM6223 19 21 clinical significance Patient Lorlatinib 2nd Sample MAP2K1 p.F129L c.385T>C SNV Potential NM_002755.3 rs1057519805 COSM1570285 23 27 clinical significance NGS-based liquid biopsy profiling in ALK+ NSCLC 2371 NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. ALK-Is [11]. Recently, Noe et al. [20] reported a 53% 4. Discussion ALK mutation detection rate in samples obtained post- ALK-Is have dramatically improved outcomes in progression on alectinib in which G1202R was the most NSCLC patients [17]. However, despite the impressive frequent mutation. In our cohort, more than one muta- responses they elicit, patients invariably relapse due to tion in ALK locus was detected in two samples collected acquired resistance mutations. Solid biopsies remain during second- and third-generation ALK-I treatment. the gold standard for biomarker testing. However, Likewise, it has been described that ALK resistance logistics for obtaining repeat tumor biopsies are com- mutations become more frequent with each successive plicated and seldom feasible leading to an empirical generation of ALK-I as sequential treatment may pro- prescription of sequential ALK-Is. Nevertheless, blind- mote the appearance of resistance mutation at the ALK ing treatment sequential strategies might have a delete- locus [21]. rious effect on patient’s survival due to the A reduced number of studies analyzing samples col- incompletely overlapping ALK mutation coverage of lected upon progression to an ALK-I by NGS have so far different ALK-Is. In this exploratory analysis, we show been conducted [11,18]. To our knowledge, there are only that plasma NGS is feasible and we propose several two studies describing NGS analysis of EML4-ALK different mechanisms which may underlie resistance to NSCLC liquid biopsy samples using OncomineTM Pan- ALK Inhibitors (ALK-Is). Unfortunately, we did not Cancer Cell-Free Assay and both are case reports [22, have available data or plasma/tissue samples collected 23]. Overall all, our results are in agreement with other at baseline in 18 of the 24 patients included in the studies [11]. However, some genes such as IGF-1R and study, which may constitute an important limitation. SRC, which have been described to be relevant for ALK Importantly, we also provide an algorithm capable of resistance, were not included in the used panel. Genomic retrieving somatic mutations in the ALK locus that alterations responsible of treatment failure can be would otherwise be discarded by the commercial bioin- detected analyzing different components of bloodstream, formatic pipeline. As presented in Table 2, the com- most notably circulating tumor cells (CTCs) and cfDNA mercial pipeline only detected three out of 12 [19,24,25]. However, it has been proposed that cfDNA is mutations. MAFs of variants detected by the commer- the best strategy for sequencing analyses [26]. Nonethe- cial pipeline were 2.8%, 2.1%, and 0.4%. According less, assays for the detection of EML4-ALK fusion pro- to the manufacturer’s specifications, the LOD, in terms tein in CTCs have been developed [27]. Consistent with of MAF, for mutations is 0.1%. However, in our our results, mutations in TP53, FGFR2, PIK3CA, and hands, mutations with a MAF below 0.5% are seldom MET have been identified in the tumor biopsy of patients detected by the commercial pipeline. By using the progressing on ceritinib [11]. The E545K and E545A VALK pipeline, some mutations that would otherwise mutations in PIK3CA have not only been detected upon have been missed can be rescued. Yet, confirmation progression in advanced ALK-positive NSCLC patients, using an alternative technique such as dPCR would be but also in EGFR-positive NSCLC patients [28,29]. The required to rule out false-positive calls. IDH2 R140Q detected in our cohort is known to trans- Regarding acquired mutations in the ALK locus, our form cells in vitro and induces myeloid and lymphoid results are consistent with those of previous studies. neoplasms in mice [30,31]. This mutation is also frequent Specifically, secondary mutations were detected in the in angioimmunoblastic T-cell lymphoma [32]. In plasma samples of 4 of the 11 (36%) patients treated NSCLC, IDH1/2 mutations are rarely detected in pri- with first-line crizotinib, with the G1269A mutation mary tumors but it has been suggested that they could be being detected in two cases. In this regard, mutation branching drivers leading to subclonal evolution, based detection rate after crizotinib failure might vary from on the MAFs at which these mutations are detected [33]. 60% [18] to 24% [19], G1269A being the most prevalent In this way, Zhao et al. [34] described a case of an ALK- mutation. We also detected the L1196M and S1206Y positive tumor in which an IDH1 variant was detected mutations, which have been reported to occur in 7% upon disease progression. and 2% of cases, respectively, of ALK-positive NSCLC Also, we found the E746_A750del mutation in one patients treated with crizotinib [11]. Finally, we detected patient who did not benefit from treatment with ALK- the A1200V mutation after crizotinib failure in one Is. Unfortunately, we could not orthogonally validate patient. This mutation is also known to appear upon the presence of the ALK translocation nor the EGFR crizotinib progression [18]. In addition, we found that variant at baseline due to lack of tissue, which may the G1202R mutation was identified in three of the 10 constitute a limitation of the present study. However, patients (30%) progressing on alectinib. This mutation it has been previously reported that mutations in is known to arise after treatment with second-generation EGFR in some NSCLC tumors coexist alongside ALK 2372 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. S anchez-Herrero et al. NGS-based liquid biopsy profiling in ALK+ NSCLC rearrangements, although this is, at best, a rare event 5. Conclusions [35,36]. Concomitant ALK and EGFR alterations may lead to primary resistance to ALK-I [37]. Likewise, a In conclusion, our data show that molecular mecha- non-V600 BRAF mutation was detected after nisms underlying treatment failure seem to involve dif- 3 months of treatment with second-line ceritinib treat- ferent pathways. NGS analysis of liquid biopsies ment, suggesting that resistance of the tumor to the collected upon disease progression is feasible and a ALK-I could be due to the acquisition of the BRAF valuable approach toward personalized that will lead mutation. It has been reported that ceritinib enhances to better care for ALK-positive NSCLC patients. the efficacy of trametinib, a MEK inhibitor, in BRAF/ NRAS-wt melanoma cell lines [38], which makes it Acknowledgements plausible that ceritinib would not have any effect in BRAF-mutated cells. Finally, two patients in whose The authors wish to thank the donors, and the BIO- plasma sample the F129L-activating mutation in BANK HOSPITAL UNIVERSITARIO PUERTA DE MAP2K1 (MEK1) was detected, exhibited marked HIERRO MAJADAHONDA (HUPHM)/INSTI- resistance to second- and third-generation ALK-Is. Of TUTO DE INVESTIGACION  SANITARIA note, we did not detect this mutation at baseline in PUERTA DE HIERRO-SEGOVIA DE ARANA one of these patients, supporting its role as a resistance (IDIPHISA) (PT17/0015/0020 in the Spanish National mutation. In this way, this mutation has been identi- Biobanks Network) for the human specimens used in fied as the molecular mechanism underlying MEK/ this study. This study has been funded by Instituto de ERK pathway activation in resistant clones of human Salud Carlos III through the project ‘PI17/01977’ (Co- HT-29 colon cancer cells [39]. Moreover, the activation funded by European Regional Development Fund/ of this downstream pathway is critical to the survival European Social Fund ‘A way to make Europe’/ of ALK-positive NSCLC cells [40,41]. Indeed, the com- ‘Investing in your future’). The work presented in this bination of ALK and MEK inhibition was highly effec- paper also received funding from the European tive at suppressing tumor growth in a preclinical Union’s Horizon 2020 research and innovation pro- model of EML4-ALK NSCLC [42]. Taken together, it gram under grant agreement No 875160. ES was is plausible that the F129L-activating mutation in funded by the Consejerıa de Ciencia, Universidades e MAP2K1 is an acquired mutation that leads to tumor Innovaci on of the Comunidad de Madrid (Doctorados resistance to ALK-Is. Industriales of the Comunidad de Madrid IND2019/ Mutations in the FGFR2 and FGFR3 genes were BMD-17258). RS was funded by the Consejerıa de detected in two patients progressing on ALK-Is, sug- Educaci on, Juventud y Deporte of the Comunidad de gesting sensitivity to fibroblast growth factor receptor Madrid and by the Fondo Social Europeo (Programa inhibitors. It has been reported that alectinib, despite Operativo de Empleo Juvenil, and Iniciativa de being a potent ALK-I, has limited inhibitory activity Empleo Juvenil, PEJD-2018-PRE/BMD-8640). against other protein kinases such as FGFR2 [43]. In this way, clinical trials evaluating the efficacy of com- Conflict of interest binations of ALK-Is with FGFR inhibitors would be of particular interest. MP reports personal fees from Roche, BMS, MSD Pfi- Three CNVs in c-MYC, CCND1, and FGFR3 were zer, Lilly, Novartis, and Takeda grants and personal detected upon disease progression in one patient, who fees from AstraZeneca, and Boehringer during the was being treated with crizotinib. Remarkably, c-MYC conduct of the study. VC reports personal fees from amplification determines many oncogenic effects [44] Roche BMS, MSD, Pfizer, Lilly, AstraZeneca, Boeh- and it has been identified as a potential mechanism of ringer, Novartis, Takeda, during the conduct of the primary resistance to crizotinib in ALK-rearranged study. MD reports personal fees from Astra-Zeneca, NSCLC patients [45]. It has been previously suggested BMS, Boehringer Ingelheim, MSD, Pfizer and Roche. by Alidousty et al. that co-occurrence of early TP53 The rest of the authors have declared no conflict of mutations in ALK+ NSCLC can lead to chromosomal interest. instability. Specifically, authors reported that, in a sub- set of 53 ALK+ tumors, up to a quarter of TP53-mu- Data accessibility tated tumors showed amplifications of known cancer genes such as MYC or CCND1 [46]. Consistent with This paper is available as a preprint in Research this, we detected the P92A and V157F mutations in the Square server, https://doi.org/10.21203/rs.3.rs-86055/v1 TP53 gene in the same plasma sample of this patient. (https://www.researchsquare.com/article/rs-86055/v1). Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2373 NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. et al. (2017) Alectinib versus crizotinib in untreated Author contributions ALK-positive non–small-cell lung cancer. N Engl J Med AR and MP conceived and /or designed the work. ES 377, 829–838. and RS have carried out statistical analyses. VI has 7 Camidge DR, Dziadziuszko R, Peters S, Mok T, Noe J, developed VALK tool. All authors have made substan- Nowicka M, Gadgeel SM, Cheema P, Pavlakis N, de tial contributions to the acquisition and interpretation Marinis F et al. (2019) Updated efficacy and safety data of data. AR and ES have drafted and revised the and impact of the EML4-ALK fusion variant on the manuscript. All authors have approved the final ver- efficacy of alectinib in untreated ALK-positive advanced non-small cell lung cancer in the global phase III ALEX sion. Each author agreed to be accountable for all study. J Thorac Oncol 14, 1233–1243. aspects of the work in ensuring that questions related 8 Lin JJ, Zhu VW, Yoda S, Yeap BY, Schrock AB, to the accuracy or integrity of any part of the work Dagogo-Jack I, Jessop NA, Jiang GY, Le LP, Gowen are appropriately investigated and resolved. K et al. (2018) Impact of EML4-ALK variant on resistance mechanisms and clinical outcomes in ALK- Peer Review positive lung cancer. J Thorac Oncol 14, 1233–1243. 9 Doebele RC, Pilling AB, Aisner DL, Kutateladze TG, The peer review history for this article is available at Le AT, Weickhardt AJ, Kondo KL, Linderman DJ, https://publons.com/publon/10.1002/1878-0261.13033. Heasley LE, Franklin WA et al. (2012) Mechanisms of resistance to crizotinib in patients with ALK gene rearranged non-small cell lung cancer. Clin Cancer Res References 18, 1472–1482. 1 Shaw AT, Yeap BY, Solomon BJ, Riely GJ, Gainor J, 10 Katayama R, Shaw AT, Khan TM, Mino-Kenudson Engelman JA, Shapiro GI, Costa DB, Ou S-HI, M, Solomon BJ, Halmos B, Jessop NA, Wain JC, Yeo Butaney M et al. (2011) Effect of crizotinib on overall AT, Benes C et al. (2016) Cancer: mechanisms of survival in patients with advanced non-small-cell lung acquired crizotinib resistance in ALK-rearranged lung cancer harbouring ALK gene rearrangement: a cancers. Sci Transl Med 4, 120ra17. retrospective analysis. Lancet Oncol 12, 1004–1012. 11 Gainor JF, Dardaei L, Yoda S, Friboulet L, Leshchiner 2 Shaw AT, Kim DW, Nakagawa K, Seto T, Crin o L, I, Katayama R, Dagogo-Jack I, Gadgeel S, Schultz K, Ahn MJ, De Pas T, Besse B, Solomon BJ, Blackhall F Singh M et al. (2016) Molecular mechanisms of et al. (2013) Crizotinib versus chemotherapy in resistance to first- and second-generation ALK advanced ALK-positive lung cancer. N Engl J Med 368, inhibitors in ALK -rearranged lung cancer. Cancer 2385–2394. Discov 6, 1118–1133. 3 Gadgeel SM, Gandhi L, Riely GJ, Chiappori AA, West 12 Okada K, Araki M, Sakashita T, Ma B, Kanada R, HL, Azada MC, Morcos PN, Lee R-M, Garcia L, Yu Yanagitani N, Horiike A, Koike S, Oh-hara T, L et al. (2014) Safety and activity of alectinib against Watanabe K et al. (2019) Prediction of ALK mutations systemic disease and brain metastases in patients with mediating ALK-TKIs resistance and drug re-purposing crizotinib-resistant ALK-rearranged non-small-cell lung to overcome the resistance. EBioMedicine 41, 105–119. cancer (AF-002JG): results from the dose-finding 13 Li J, Huang Y, Wu M, Wu C, Li X & Bao J (2018) portion of a phase 1/2 study. Lancet Oncol 15, 1119– Structure and energy based quantitative missense 1128. variant effect analysis provides insights into drug 4 Duruisseaux M, Besse B, Cadranel J, Perol M, resistance mechanisms of anaplastic lymphoma kinase Mennecier B, Bigay-Game L, Descourt R, Dansin E, mutations. Sci Rep 8, 10664. Audigier-Valette C, Moreau L et al. (2017) Overall 14 Ettinger DS, Wood DE, Aisner DL, Akerley W, survival with crizotinib and next-generation ALK Bauman J, Chirieac LR, D’Amico TA, DeCamp MM, inhibitors in ALK-positive non-small-cell lung cancer Dilling TJ, Dobelbower M et al. (2017) Non-small cell (IFCT-1302 CLINALK): a French nationwide cohort lung cancer, version 5.2017: clinical practice guidelines retrospective study. Oncotarget 8, 21903–21917. in oncology. J Natl Compr Canc Netw 15, 504–535. 5 Huber RM, Hansen KH, Paz-Ares Rodrıguez L, West 15 Provencio M, Perez-Barrios C, Barquin M, Calvo V, HL, Reckamp KL, Leighl NB, Tiseo M, Smit EF, Kim Franco F, Sanchez E, Sanchez R, Marsden D, D-W, Gettinger SN et al. (2020) Brigatinib in crizotinib- Cristobal Sanchez J, Martin Acosta P et al. (2020) refractory ALK+ NSCLC: 2-year follow-up on systemic Next-generation sequencing for tumor mutation and intracranial outcomes in the phase 2 ALTA trial. J quantification using liquid biopsies. Clin Chem Lab Thorac Oncol 15, 404–415. Med 58, 306–313. 6 Peters S, Camidge DR, Shaw AT, Gadgeel S, Ahn JS, 16 Li MM, Datto M, Duncavage EJ, Kulkarni S, Kim DW, Ou SI, Perol M, Dziadziuszko R, Rosell R Lindeman NI, Roy S, Tsimberidou AM, Vnencak-Jones 2374 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. S anchez-Herrero et al. NGS-based liquid biopsy profiling in ALK+ NSCLC CL, Wolff DJ, Younes A et al. (2017) Standards and markers: an update to a clinical routine practice. Cells guidelines for the interpretation and reporting of 10, 168. sequence variants in cancer. J Mol Diagnostics 19, 4–23. 27 Rossi E, Aieta M, Tartarone A, Pezzuto A, Facchinetti 17 Gambacorti Passerini C, Farina F, Stasia A, Redaelli S, A, Santini D, Ulivi P, Ludovini V, Possidente L, Ceccon M, Mologni L, Messa C, Guerra L, Giudici G, Fiduccia P et al. (2021) A fully automated assay to Sala E et al. (2014) Crizotinib in advanced, detect the expression of pan-cytokeratins and of EML4- chemoresistant anaplastic lymphoma kinase-positive ALK fusion protein in circulating tumour cells (CTCs) lymphoma patients. J Natl Cancer Inst 106, djt378. predicts outcome of non-small cell lung cancer 18 Yanagitani N, Uchibori K, Koike S, Tsukahara M, (NSCLC) patients. Transl Lung Cancer Res 10, 80–92. Kitazono S, Yoshizawa T, Horiike A, Ohyanagi F, 28 Romero A, Serna-Blasco R, Alfaro C, Sanchez-Herrero Tambo Y, Nishikawa S et al. (2020) Drug resistance E, Barquın M, Turpin MC, Chico S, Sanz-Moreno S, mechanisms in Japanese anaplastic lymphoma kinase- Rodrigez-Festa A, Laza-Briviesca R et al. (2020) positive non–small cell lung cancer and the clinical ctDNA analysis reveals different molecular patterns responses based on the resistant mechanisms. Cancer upon disease progression in patients treated with Sci 111, 932–939. osimertinib. Transl Lung Cancer Res 9, 532–540. 19 Horn L, Whisenant JG, Wakelee H, Reckamp KL, 29 Chaft JE, Arcila ME, Paik PK, Lau C, Riely GJ, Qiao H, Leal TA, Du L, Hernandez J, Huang V, Catherine Pietanza M, Zakowski MF, Rusch V, Sima Blumenschein GR et al. (2019) Monitoring therapeutic CS, Ladanyi M et al. (2012) Coexistence of PIK3CA response and resistance: analysis of circulating tumor and other oncogene mutations in lung adenocarcinoma- DNA in patients with ALK+ lung cancer. J Thorac rationale for comprehensive mutation profiling. Mol Oncol 14, 1901–1911. Cancer Ther 11, 485–491. 20 Noe J, Lovejoy A, Ou SHI, Yaung SJ, Bordogna W, 30 Wang F, Travins J, DeLaBarre B, Penard-Lacronique Klass DM, Cummings CA & Shaw AT (2020) ALK V, Schalm S, Hansen E, Straley K, Kernytsky A, Liu mutation status before and after alectinib treatment in W, Gliser C et al. (2013) Targeted inhibition of mutant locally advanced or metastatic ALK-positive NSCLC: IDH2 in leukemia cells induces cellular differentiation. Pooled analysis of two prospective trials. J Thorac Science 340, 622–626. Oncol 15, 601–608. 31 Mylonas E, Janin M, Bawa O, Opolon P, David M, 21 Dagogo-Jack I, Rooney M, Lin JJ, Nagy RJ, Yeap BY, Quivoron C, Bernard OA, Ottolenghi C, DeBotton S & Hubbeling H, Chin E, Ackil J, Farago AF, Hata AN Penard-Lacronique V (2014) Isocitrate dehydrogenase et al. (2019) Treatment with next-generation ALK (IDH)2 R140Q mutation induces myeloid and lymphoid inhibitors fuels plasma ALK mutation diversity. Clin neoplasms in mice. Leukemia 28, 1343–1346. Cancer Res 25, 6662–6670. 32 Cairns RA, Iqbal J, Lemonnier F, Kucuk C, De Leval 22 Sanchez-Herrero E, Clemente MB, Calvo V, Provencio L, Jais JP, Parrens M, Martin A, Xerri L, Brousset P M & Romero A (2020) Next-generation sequencing to et al. (2012) IDH2 mutations are frequent in dynamically detect mechanisms of resistance to ALK angioimmunoblastic T-cell lymphoma. Blood 119, 1901– inhibitors in ALK-positive NSCLC patients: a case 1903. report. Transl Lung Cancer Res 9, 366–372. 33 Rodriguez EF, De Marchi F, Lokhandwala PM, 23 Berger LA, Janning M, Velthaus JL, Ben-Batalla I, Belchis D, Xian R, Gocke CD, Eshleman JR, Illei P & Schatz S, Falk M, Iglauer P, Simon R, Cao R, Forcato Li MT (2020) IDH1 and IDH2 mutations in lung C et al. (2018) Identification of a high-level MET adenocarcinomas: evidences of subclonal evolution. amplification in CTCs and cfTNA of an ALK-positive Cancer Med 9, 4386–4394. NSCLC patient developing evasive resistance to 34 Zhao P, Peng L, Wu W, Zheng Y, Jiang W, Zhang H, crizotinib. J Thorac Oncol 13, e243–e246. Tong Z, Liu L, Ma R, Wang L et al. (2019) Carcinoma 24 Pailler E, Faugeroux V, Oulhen M, Mezquita L, of unknown primary with EML4-ALK fusion response Laporte M, Honore A, Lecluse Y, Queffelec P, to ALK inhibitors. Oncologist 24, 449–454. NgoCamus M, Nicotra C et al. (2019) Acquired 35 Yang JJ, Zhang XC, Su J, Xu CR, Zhou Q, Tian HX, resistance mutations to ALK inhibitors identified by Xie Z, Chen HJ, Huang YS, Jiang BY et al. (2014) single circulating tumor cell sequencing in ALK- Lung cancers with concomitant egfr mutations and rearranged non–small-cell lung cancer. Clin Cancer Res ALK rearrangements: Diverse responses to EGFR-TKI 25, 6671–6682. and crizotinib in relation to diverse receptors 25 Alix-Panabieres C, Pantel K & Authors C (2021) phosphorylation. Clin Cancer Res 20, 1383–1392. Liquid biopsy: from discovery to clinical application. 36 Ulivi P, Chiadini E, Dazzi C, Dubini A, Costantini M, Cancer Discov 11, 858–873. Medri L, Puccetti M, Capelli L, Calistri D, Verlicchi A 26 Hofman P (2021) Detecting resistance to therapeutic et al. (2016) Nonsquamous, non-small-cell lung cancer ALK inhibitors in tumor tissue and liquid biopsy patients who carry a double mutation of EGFR, Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 2375 NGS-based liquid biopsy profiling in ALK+ NSCLC E. S anchez-Herrero et al. EML4-ALK or KRAS: frequency, clinical-pathological Ferrarini A, Buson G et al. (2019) MYC amplification characteristics, and response to therapy. Clin Lung as a potential mechanism of primary resistance to Cancer 17, 384–390. crizotinib in ALK-rearranged non-small cell lung 37 Tiseo M, Gelsomino F, Boggiani D, Bortesi B, cancer: a brief report. Transl Oncol 12, 116–121. Bartolotti M, Bozzetti C, Sammarelli G, Thai E & 46 Alidousty C, Baar T, Martelotto LG, Heydt C, Ardizzoni A (2011) EGFR and EML4-ALK gene Wagener S, Fassunke J, Duerbaum N, Scheel AH, mutations in NSCLC: a case report of erlotinib- Frank S, Holz B et al. (2018) Genetic instability and resistant patient with both concomitant mutations. recurrent MYC amplification in ALK-translocated Lung Cancer 71, 241–243. NSCLC: a central role of TP53 mutations. J Pathol 38 Verduzco D, Kuenzi BM, Kinose F, Sondak VK, 246, 67–76. Eroglu Z, Rix U & Smalley KSM (2018) Ceritinib enhances the efficacy of trametinib in BRAF/NRAS- wild-type melanoma cell lines. Mol Cancer Ther 17, 73– Supporting information 83. Additional supporting information may be found 39 Wang H, Daouti S, Li WH, Wen Y, Rizzo C, Higgins online in the Supporting Information section at the end B, Packman K, Rosen N, Boylan JF, Heimbrook D of the article. et al. (2011) Identification of the MEK1(F129L) Fig. S1. Flowchart of the bioinformatic pipeline opti- activating mutation as a potential mechanism of mized for the processing and assessment of variants at acquired resistance to MEK inhibition in human the ALK gene locus. cancers carrying the B-Raf V600E mutation. Cancer Fig. S2. Frequency of ALK missense mutations identi- Res 71, 5535–5545. fied in the study population. 40 Hrustanovic G & Bivona TG (2015) RAS-MAPK in ALK Fig. S3. PFS and OS curves according to F129L targeted therapy resistance. Cell Cycle 14, 3661–3662. (MAP2K1) mutation status. 41 Hrustanovic G, Olivas V, Pazarentzos E, Tulpule A, Asthana S, Blakely CM, Okimoto RA, Lin L, Neel DS, Table S1. Identification of EML4-ALK translocation. Sabnis A et al. (2015) RAS-MAPK dependence Table S2. Genes included in the NGS panel used. underlies a rational polytherapy strategy in EML4- Table S3. DNA genotyping of PBMCs. ALK-positive lung cancer. Nat Med 21, 1038–1047. Table S4. List of mutations in ALK locus tested by 42 Crystal AS, Shaw AT, Sequist LV, Friboulet L, dPCR. Niederst MJ, Lockerman EL, Frias RL, Gainor JF, Table S5. ALK-Is treatments of the study cohort. Amzallag A, Greninger P et al. (2014) Patient-derived Table S6. List of all somatic mutations detected in the models of acquired resistance can identify effective drug study cohort. combinations for cancer. Science 346, 1480–1486. Table S7. Cross-table describing PPA, NPA and ORA 43 Sakamoto H, Tsukaguchi T, Hiroshima S, Kodama T, for ALK cohort using VALK tool. Kobayashi T, Fukami TA, Oikawa N, Tsukuda T, Ishii Table S8. Cross-tables describing PPA, NPA and N & Aoki Y (2011) CH5424802, a selective ALK ORA for Valencia cohort using VALK tool. inhibitor capable of blocking the resistant gatekeeper Table S9. Cross-tables describing PPA, NPA and mutant. Cancer Cell 19, 679–690. ORA for ALK cohort using the Oncomine Filter. 44 Lin CY, Loven J, Rahl PB, Paranal RM, Burge CB, Table S10. Cross-tables describing PPA, NPA and Bradner JE, Lee TI & Young RA (2012) ORA for Valencia cohort using the Oncomine Filter. Transcriptional amplification in tumor cells with Table S11. List of the mutations detected at disease elevated c-Myc. Cell 151, 56–67. progression and status at baseline or in previous sam- 45 Rihawi K, Alfieri R, Fiorentino M, Fontana F, Capizzi ple. E, Cavazzoni A, Terracciano M, La Monica S, 2376 Molecular Oncology 15 (2021) 2363–2376 ª 2021 The Authors. Published by FEBS Press and John Wiley & Sons Ltd.