ORIGINAL ARTICLE

Multiplex Diagnosis of Oncogenic Fusion and MET Exon Skipping by Molecular Counting Using Formalin-Fixed Paraffin Embedded Lung Adenocarcinoma Tissues

Kuniko Sunami, MD,a,g Koh Furuta, MD, PhD,b Koji Tsuta, MD, PhD,c Shinji Sasada, MD, PhD,d Takehiro Izumo, MD, PhD,d Takashi Nakaoku, MD,a Yoko Shimada, MFSc,a Motonobu Saito, MD, PhD,a Hiroshi Nokihara, MD, PhD,e Shun-ichi Watanabe, MD, PhD,f Yuichiro Ohe, MD, PhD,e,g Takashi Kohno, PhDa,* aDivision of Genome Biology, National Cancer Center Research Institute, Tokyo, Japan bDivision of Clinical Laboratory, National Cancer Center Hospital, Tokyo, Japan cDivision of Pathology, National Cancer Center Research Institute, Tokyo, Japan dDepartment of Endoscopy, Respiratory Endoscopy Division, National Cancer Center Research Institute, Tokyo, Japan eDepartment of Thoracic Oncology, National Cancer Center Research Institute, Tokyo, Japan fDepartment of Thoracic Surgery, National Cancer Center Hospital, Tokyo, Japan gCourse of Advanced Clinical Research of Cancer, Juntendo University Graduate School of Medicine, Tokyo, Japan

Received 31 July 2015; revised 23 September 2015; accepted 13 October 2015

ABSTRACT detected oncogenic fusions in bronchial lavage fluid and transbronchial biopsy samples. Introduction: Fusions of the anaplastic lymphoma receptor kinase (ALK), ret proto-oncogene (RET), ROS Conclusions: The MC assay allows multiplex detection of proto-oncogene 1, receptor gene (ROS1), B- oncogenic fusion and exon-skipped transcripts in tumor Raf proto-oncogene, serine/threonine kinase gene (BRAF), samples, including in formalin-fixed paraffin-embedded and 1 gene (NRG1) and intronic MMNG HOS samples obtained in the clinic. Transforming gene (MET) mutations are druggable onco- 2015 International Association for the Study of Lung gene alterations in lung adenocarcinoma that cause Cancer. Published by Elsevier Inc. All rights reserved. expression of aberrant transcripts. Because these aberrant transcripts are both infrequent (incidence <5%) and mutually exclusive, multiplex assays are required to detect Keywords: Personalized medicine; Molecular counting; them in tumor samples. Oncogene fusion; MET exon skipping; Genetic diagnosis Methods: Aberrant transcripts of the six aforementioned oncogenes (36 transcripts in total) were examined in a Introduction molecular counting (MC) assay, which counts RNA mole- Analyses of lung adenocarcinoma (LADC) genomes cules by simultaneous hybridization of several probes. fi Forty-one samples of surgically resected lung adenocarci- have identi ed several oncogene alterations that cause noma tissue found to express one of these aberrant onco- expression of aberrant transcripts in tumor cells. A genic transcripts upon whole transcriptome sequencing representative example is the fusion of oncogenes to (test cohort: n ¼ 22) or reverse transcription polymerase chain reaction (validation cohort: n ¼ 19) analyses were subjected to MC, after which biopsies were performed on *Corresponding author. tumor tissue samples. Disclosure: The authors declare no conflicts of interest. Results: Address for correspondence: Takashi Kohno, PhD, Division of Genome Threshold values for the diagnosis of each of the Biology, National Cancer Center Research Institute, 1-1 Tsukiji 5- 36 transcripts were determined in frozen and formalin- chome, Chuo-ku, Tokyo 104-0045, Japan. E-mail: [email protected] fixed paraffin-embedded samples from the test cohort. On ª 2015 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. the basis of these threshold values, the MC assay diagnosed ISSN: 1556-0864 expression of oncogenic transcripts in the validation cohort http://dx.doi.org/10.1016/j.jtho.2015.10.005 samples with 100% accuracy. The assay also accurately

Journal of Thoracic Oncology Vol. 11 No. 2: 203-212 204 Sunami et al Journal of Thoracic Oncology Vol. 11 No. 2 other partner , which causes constitutive activation multiplex assays allowing simultaneous identification of of protein kinases encoded by these oncogenes.1,2 These several alterations are required. Furthermore, in contrast alterations are a mutually exclusive driver mutation in to hot spot mutations such as EGFR, Kirsten rat sarcoma lung carcinogenesis and are a target for therapies based viral gene homolog gene (KRAS), and BRAF (which acti- on protein tyrosine kinase inhibitors (TKIs). A well- vate oncogenes in tumor cell), the presence of highly known example is the oncogenic epidermal growth diverse genomic fusion points located within intron factor receptor gene (EGFR) mutation, which is a target sequences hampers the detection of gene fusions by for therapy using EGFR-TKIs. Indeed, LADCs harboring multiplex genome polymerase chain reaction (PCR)- the anaplastic lymphoma gene based assays using next-generation sequencers.20 In (ALK) oncogene fusion (in 3%–5% of all cases of LADC) addition, identifying the intronic mutations responsible respond very well to ALK inhibitors such as ,3 for MET exon skipping using genomic DNA is difficult ,4 and .5 Ret proto-oncogene (RET) and because of their highly diverse locations and the occur- ROS proto-oncogene 1, receptor tyrosine kinase gene rence of passenger mutations.21 Multiplex reverse tran- (ROS1) oncogene fusions are also rare (each occurring in scription (RT)-PCR–based assays using tumor RNA 1%–3% of cases of LADC6–10), but they are a promising are often used to detect aberrant transcripts caused by target for therapy with TKIs. Several clinical trials gene fusions and exon skipping because there are focusing on RET and ROS1 fusion–positive LADC have few aberrant transcript patterns, despite the diversity been undertaken, and the results of these studies have with respect to the location of the responsible mutations. recently been reported.11,12 In addition, there is a report However, the poor quality of the RNA obtained from of single patient with RET fusion–positive LADC who FFPE tumor tissues, particularly from archived tissues, responded to RET TKI.13 These reports indicate the makes it difficult to develop reliable RT-PCR–based therapeutic effects of RET and ROS1 TKIs. Recently, we multiplex assays. and others identified B-Raf proto-oncogene, serine/ Molecular counting (MC) using the nCounter system threonine kinase gene (BRAF) and gene (NanoString Technologies, Seattle, WA) is a method of (NRG1) fusions as another therapeutic target in quantifying RNA that is not based on RT and PCR; this approximately 10% of patients with invasive mucinous method directly counts RNA molecules by means of LADC.14,15 Recent studies have also identified another simultaneous hybridization of multiple probes. Indeed, type of oncogene alteration that generates aberrant previous studies have demonstrated the feasibility of oncogenic transcripts in approximately 3% of LADCs: this method for quantifying RNA expression and skipping of the coding exon 14 in MMNG HOS Trans- detecting gene fusions.22,23 Here, we used this method to forming gene (MET) oncogene transcripts on account of detect 36 oncogenic transcripts, 35 of which were aberrant splicing, which is associated mainly with derived from five oncogenes (ALK, RET, ROS1, NRG1, and intronic mutations.16,17 The skipped transcript produces BRAF) fused to partner genes and the aforementioned a constitutively active MET protein lacking the E3 MET exon–skipped transcript above. To examine the ubiquitin protein ligase (CBL) binding domain that feasibility of this method, RNA samples obtained from negatively regulates MET kinase. Indeed, in recent re- 62 LADC tissue samples from 41 patients (39 FFPE and ports, a MET TKI showed antitumor activity in patients 23 snap-frozen tissue samples in total; Fig. 1B) were with LADCs expressing exon 14–skipped MET tran- subjected to MC. These samples were selected from scripts,18,19 thus suggesting that this alteration is also a 608 consecutive patients with LADC patients because promising therapeutic target. whole RNA sequencing or RT-PCR analyses of snap- Precision medicine for patients with LADC on the frozen tissues revealed that they either did or did not basis of the TKIs described earlier requires the diagnosis express aberrant oncogenic transcripts. of multiplex oncogene aberrations in tumor cells. Because the tumors of most patients subjected to TKI therapy are inoperable, diagnosis must be performed by using small Methods amounts of fixed formalin-fixed and paraffin-embedded Subjects (FFPE) biopsy samples or archived surgical specimens Consecutive patients with LADC who underwent for histopathological examination. In fact, techniques surgical resection between 1997 and 2008 at the Na- such as fluorescence in situ hybridization, which detects tional Cancer Center Hospital in Tokyo (n ¼ 608) were gross rearrangements, are often used to screened for EGFR, KRAS, BRAF, and HER2 hot spot diagnose gene fusions in such clinical specimens. How- mutations by the high-resolution melting method using ever, the recent discovery of a number of uncommon but genomic DNA obtained from snap-frozen tumor tis- druggable oncogene fusions in LADC, along with the sues.24 RT-PCR of RNA obtained from snap-frozen tumor finding that they are mutually exclusive, means that tissues as described previously was used to examine the February 2016 Multiplex Diagnosis of Oncogenic Fusions 205

Figure 1. (A) Selection of study subjects. From 608 consecutive patients who underwent surgical resection of lung adeno- carcinoma, we selected 41 who harbored oncogenic gene fusions or exon skipping. Patients were classified into two cohorts. Cases in which RNA isolated from snap-frozen tumor samples had been analyzed by whole transcriptome RNA sequencing were included in the test cohort (n ¼ 22), whereas those in which RNA isolated from snap-frozen tumors samples had been analyzed by reverse transcription polymerase chain reaction were included in the validation cohort (n ¼ 19). (B) Sample characteristics and molecular counting assay results. RNA was extracted from fresh frozen and/or formalin-fixed and paraffin- embedded tissues. Status was determined by counting the signals generated by fusion probes (F) or those generated by fusion and imbalance probes (FI). (Black box) Diagnosis based on counts generated by the fusion probe only. (Gray box) Diagnosis based on counts generated by the FI probes. (C) Strategy for diagnosing gene fusions and MET exon skipping. (Left) Twenty-six cases of gene fusion and one case of MET exon 14 skipping were diagnosed on the basis of counts generated by the fusion probes alone. (Right) Nine cases of gene fusion were diagnosed on the basis of counts generated by both FI probes. following fusions: echinoderm microtubule associated transcripts. The results were confirmed by RT-PCR as protein like 4 gene (EML4)-ALK and kinesin family described previously.21 member 5B (KIF5B)-ALK; KIF5B-RET and coiled-coil Sixty-four cases of LADC with ALK, RET, ROS1, NRG1, domain containing 6 gene (CCDC6)-RET; and CD74, ma- and BRAF gene fusions or MET exon skipping (Fig. 1A) jor histocompatibility complex, class II invariant chain were identified. After 23 cases been excluded because gene (CD74)-ROS1, ezrin gene (EZR)-ROS1, and solute of insufficient sample, the remaining 41 were divided carrier family 34 (type II sodium/phosphate cotrans- into two cohorts and subjected to MC analysis. The porter), member 2 gene (SLC34A2)-ROS1.8,10,25 Histo- test cohort included 22 cases in which whole RNA logical diagnosis was based on the latest edition of the sequencing with next-generation sequencers had been World Health Organization classification of lung tu- performed on RNA obtained from snap-frozen tumor mors.26 RNA sequencing was performed on the 160 tissues. From these cases, 13 frozen tissue RNA samples cases that yielded sufficient RNA to determine whether and 21 FFPE tissue RNA samples were available for they expressed oncogene fusions and exon-skipped MET study (see Fig. 1B). The validation cohort comprised 206 Sunami et al Journal of Thoracic Oncology Vol. 11 No. 2

19 cases in which RT-PCR analysis had been performed of the transcripts. Eight sets of probes were designed to to determine the presence of ALK, RET, ROS1, NRG1, hybridize with internal control genes: glyceraldehyde- and BRAF fusions and MET exon skipping. From these 3-phosphate dehydrogenase gene (GAPDH), omithine cases, 10 frozen tissue RNA samples and 18 FFPE tissue decarboxylase antizymel gene (OAZ1), polymerase RNA samples were available for study (see Fig. 1B). In (RNA) II (DNA directed) polypeptide A, 220lDa addition, two FFPE tumor tissues that had been obtained gene (POLR2A), and glucuronidase, beta gene (GUSB). by transbronchial biopsy and diagnosed as positive The nCounter assay was performed according to for the RET and ROS1 fusions, respectively, and 18 the manufacturer’s instructions. Briefly, 150 ng and bronchial lavage samples (including two harboring 400 ng of RNA from fresh frozen and FFPE tissues, the EML4-ALK fusion) that had been obtained by endo- respectively, were used in the assay to ensure that the bronchial ultrasound-guided transbronchial needle counts for an internal control gene (GPADH)were aspiration were also tested by MC assay. This study higher than 5000 per sample. RNA was hybridized was approved by the institutional review board of the with the probe sets for 16 hours at 67C. Samples were National Cancer Center, and all patients gave written processed using an automated nCounter Sample Prep informed consent. Station (NanoString Technologies, Inc.). Cartridges In snap-frozen tissue samples, the percentage of tu- containing immobilized and aligned reporter com- mor cells was determined to range from 10% to 90% plexes were subsequently imaged on an nCounter (mean 22%) by using the Global Parameter Hidden Digital Analyzer (NanoString Technologies, Inc.) that Markov Model method27 on the basis of somatic muta- had been set at a data resolution of 555 fields of view. tion and allelic imbalance data as described previously.25 Reporter counts were collected and normalized using The results were consistent with those for the FFPE nSolver analysis software version 2.0 (NanoString tissue samples estimated by hematoxylin and eosin Technologies, Inc.). staining. Thus, the percentage of tumor cells in the samples in the present study was determined to be higher than 10%. Determination of Gene Fusion and Exon Skipping Data were normalized in two steps as described pre- 22 RNA Preparation viously. First, six positive internal controls were used to remove potential systemic differences between individual RNA was extracted from FFPE tissue sections (five hybridization experiments. The sum of the signal intensity sections, each 5 mmthick)usingtheRNeasyFFPE (S ) from the six positive control probes was calculated (QIAGEN, Hilden, Germany) or from lavage fluid i for sample i and then scaled using the normalization using the RNeasy Mini kit (QIAGEN). Total RNA was factor, S /S . Second, the scaled signal intensity of extracted from grossly dissected, snap-frozen tissue mean i sample i was further normalized using four housekeeping samples using TRIzol (Invitrogen, Carlsbad, CA) ac- genes to remove any effects that might be attributable to cording to the manufacturer’s instructions. RNA was differences in the amount of input RNA. For example, if quantified using the Nanodrop 8000 (Thermo Scien- H represents the genes in sample i, the second normali- tific, Wilmington, DE). i zation factor would be defined as Hmean/Hi. Previous studies have suggested that aberrations in MC Analysis the six oncogenes examined herein are mutually exclu- The nCounter system uses a set of probes to detect sive.28,29 This suggestion was supported by the whole transcripts:asingle50 reporter probe (approximately transcriptome sequencing results for the test cohort 50-mer) linked to a fluorescence bar code tag (6-mer) samples examined; therefore, samples expressing one and a biotinylated 30 capture probe (approximately particular aberrant oncogene transcript were used as 50-mer). Seventy sets of probes (Supplementary negative controls for the other aberrant oncogene tran- Table 1) were designed to simultaneously detect 36 scripts. The threshold values used to determine the aberrant transcripts (i.e., 35 fusion transcripts for presence or absence of a particular aberrant transcript five oncogenes and one exon 14-skipped MET tran- were defined as the mean plus five standard deviations script) (Table 1). The probes were synthesized by (SDs) above the value for tumor samples harboring ab- NanoString Technologies (Seattle, WA). Of the 70 sets, errations in other transcripts. On the basis of these 38 were designed to hybridize with chimeric portions threshold values, the rate of false-positive detection was of the fusion transcripts, whereas the other 24 sets predicted to be less than 0.0002%. Threshold values were designed to hybridize with wild-type ALK, RET, were set for both fresh frozen and FFPE tissues. To and NRG1 transcripts to examine gene fusion–induced determine imbalances in the expression of the 50 and 30 imbalances in the expression of the 50 and 30 regions regions of gene transcripts, the 30 overexpression ratio February 2016 Multiplex Diagnosis of Oncogenic Fusions 207

Table 1. Aberrant Transcripts for the Six Oncogenes Examined by Molecular Counting Target Fusion or Aberrant Detectable Cases in Test Cases in Validation Total Cases, Oncogene Transcripts Variants Cohort, n Cohort, n N ALK EML4-ALK E13:A20 2 3 5 E20:A20 1 1 2 E6:A20 1 0 1 E2:A20 0 0 0 E18:A20 0 1 1 KIF5B-ALK K15:A20 0 0 0 K17:A20 0 0 0 K24:A20 0 0 0 TFG-ALK T6:A20 0 0 0 RET KIF5B-RET K15:R11 2 2 4 K15:R12 0 0 0 K16:R12 1 0 1 K22:R12 1 0 1 K23:R12 1 0 1 K24:R8 1 0 1 K24:R11 0 0 0 CCDC6-RET C1:R12 0 1 1 ROS1 CD74-ROS1 C6:R32 0 0 0 C6:R34 6 2 8 EZR-ROS1 E10:R34 1 2 3 SDC4-ROS1 S2:R32 0 0 0 S2:R34 0 0 0 S4:R32 0 0 0 S4:R34 0 0 0 SLC34A2-ROS1 S4:R32 0 0 0 S4:R34 0 0 0 S13del:R32 0 1 1 GOPC-ROS1 G8:R32 0 0 0 G8:R34 0 0 0 TPM3-ROS1 T8:R35 0 0 0 LRIG3-ROS1 L16:R35 0 0 0 BRAF TRIM24-BRAF T5:B8 1 0 1 T9:B9 0 0 0 NRG1 CD74-NRG1 C8:N6 0 2 2 C6:N6 1 0 1 MET Exon 14 skip M13:M15 3 4 7 Total 22 19 41 EML4, echinoderm microtubule associated protein like 4 gene; ALK, anaplastic lymphoma receptor tyrosine kinase gene; KIF5B, kinesin family member 5B gene; TFG, TKK-fused gene; RET, Ret proto-oncogene; CCDC6, coiled-coil domain containing 6 gene; CD74, CD74, major histocompatibility complex, class II invariant chain gene; ROS1, ROS proto-oncogene 1, receptor tyrosine kinase gene; EZR, ezrin gene; SDC4, syndecan 4 gene; SLC34A2, solute carrier family 34 (type II sodium/phosphate cotransporter), member 2 gene; GOPC, golgi- associated PDZ and coiled-coil motif containing gene; TPM3, tropomyosin 3 gene; LRIG3, leucine-rich repeats and immunoglobulin-like domains 3 gene; BRAF, B-Raf proto-oncogene, serine/threonine kinase gene; TRIM24, tripartite motif containing 24 gene; NRG1, neu- regulin 1 gene; MET, MMNG HOS Transforming gene. was calculated as follows: 30-imbalanced probe counts/ deparaffinized using EZ Prep (Ventana), heat-induced 50-imbalanced probe counts. Threshold values for 30 epitope retrieval with CC1 (Ventana) was performed overexpression for a particular gene were defined as the and the slides were incubated with primary antibodies mean value plus five SDs above the value for tumor against c-MET (CONFIRM Anti-Total c-MET, clone SP44; samples with aberrations in other genes. Ventana). Immunoreactivity was detected by using the UltraView DAB Universal Detection Kit followed by Immunostaining for the MET Protein counter-staining with Hematoxylin II (Ventana) and The Bench-Mark XT automated slide processing sys- Bluing Reagent (Ventana). Immunopositive cases were tem (Ventana, Tucson, AZ) was used to stain for c-MET, as defined as those exhibiting cytoplasmic and/or mem- previously described.30 In brief, after tissue sections were branous staining in more than 10% of cells. 208 Sunami et al Journal of Thoracic Oncology Vol. 11 No. 2

Figure 2. Results for the validation cohort. The results for fresh frozen samples (A) and formalin-fixed and paraffin- embedded samples (B) are shown. (Upper) The reporter counts for the fusion probes are expressed as log2 (fusion probe counts/threshold counts). Values are shown when the counting results were greater than the threshold. (Lower) The results 0 0 0 0 0 of 3 overexpression are expressed as log2 ratios (3 /5 probe count ratio divided by the 3 /5 probe count threshold).

Results However, the fusion probes were not able to discrimi- 0 Test and Validation Cohorts nate nine fusion transcripts that shared 5 -partner gene exons. In these cases, oncogene fusion was determined A total of 41 cases of LADC, with one aberrant from the results of fusion probe assays in conjunction oncogenic transcript expressed in each, were selected with the results from the 30/50 imbalance assays from 608 consecutive surgical cases and divided into (Fig. 1C). two cohorts (22 in the test cohort and 19 in the vali- The 22 test cohort cases expressed 13 aberrant dation cohort). Samples were then examined in an MC transcripts: 12 fusion transcripts and 1 exon 14-skipped assay (see Fig. 1A). RNA from snap-frozen tissues from MET transcript (see Table 1). For 13 of the 22 cases, RNA the 22 test cohort samples had already been subjected samples from snap-frozen tissues were available and to whole transcriptome sequencing; therefore, each subjected to MC. The background counts were calculated RNA sample expressed 1 of the 36 aberrant transcripts from data obtained for negative samples, and the (see Table 1). RNA from snap-frozen tissues obtained threshold count for each aberrant transcript was set as from the 19 validation cohort samples had been sub- the mean plus five SDs (as described in the Methods jected to RT-PCR; therefore, each RNA sample expressed section). Indeed, with these criteria, all 13 samples 1 of the 36 aberrant transcripts. were verified as log2 (fusion probe counts/threshold counts) > 0, and the results of the imbalance assays for MC Assay to Determine the Expression of the RET and NRG1 genes enabled us to judge KIF5B-RET_ Aberrant Oncogene Transcripts (K15:R12) and CD74-ROS1 (C6:R34), respectively The expression of 26 fusion transcripts and one (Supplementary Fig. 1A). The threshold values for the 39 skipped MET transcript was ascertained by counting FFPE samples were recalculated in the same way; the the signals generated by the fusion probes alone. results clearly distinguished between cases that were February 2016 Multiplex Diagnosis of Oncogenic Fusions 209

Figure 2. (continued). positive or negative for gene fusion or MET skipping cases were accurately judged as expressing aberrant (Supplementary Fig. 1B). The expression values deter- transcripts in both snap-frozen and FFPE samples mined by MC showed a good correlation with those (Figs. 2A and 2B). In one NRG1 fusion–positive case (NR- determined by RNA sequencing, thus indicating the 002), CD74_ex6 and CD74-NRG1 (C8:N6) were the major quantitative nature of the MC assay (Supplementary transcripts detected. These results are consistent with Fig. 2). those obtained by RT-PCR, which showed that CD74- NRG1 (C6:N6) was the minor transcript coexpressed with CD74-NRG1 (C8:N6) as the major transcript in this MC Assay of the Validation Cohort Samples case. Next, the 19 cases in the validation cohort were tested by MC assay using the same threshold values and the criteria set from the test cohort. These cases MC Assay of Nonsurgical Samples expressed 11 aberrant transcripts: 10 fusion transcripts The MC assay was then used to detect aberrant and one exon 14-skipped MET transcript (see Table 1). transcripts in nonsurgical specimens that had been ob- The 19 cases yielded 10 RNA samples from snap-frozen tained from the daily lung cancer clinic and used for tissues and 18 from FFPE tissues. As expected, all 19 pathological diagnosis. Two FFPE transbronchial biopsy 210 Sunami et al Journal of Thoracic Oncology Vol. 11 No. 2

Figure 3. Results for the nonsurgical samples. The results from the biopsy samples (A) and lavage samples (B) are shown. The reporter counts of the fusion probes are expressed as log2 ratios (fusion probe counts/threshold counts). Values are shown when the counting results were greater than the threshold. specimens, which were diagnosed as CCDC6-RET and Discussion CD74-ROS1 fusion by RT-PCR and fluorescence in situ Here, we describe an MC assay that allows multiplex hybridization, respectively, were subjected to MC. The diagnosis of druggable aberrations for six oncogenes— results clearly showed that these FFPE tissues were ALK, RET, ROS1, BRAF, and NRG1 fusions and MET exon positive for both of these fusions (see Fig. 3A). In addi- 14 skipping—that are not detectable by NGS-based hot tion, 18 bronchial lavage samples were also subjected to spot sequencing of genomic DNA on account of the MC analysis. The results showed that two EML4-ALK highly diverse locations of the fusion and mutation fusion–positive samples had been diagnosed correctly points. Samples that had already been identified as (see Fig. 3B). either positive or negative were used to set the threshold February 2016 Multiplex Diagnosis of Oncogenic Fusions 211 values for the MC assay. The assay accurately diagnosed simultaneously detect aberrant transcripts and measure aberrant expression of transcripts for the six druggable the expression levels of other druggable genes are worth oncogenes in both frozen and FFPE samples. These re- developing. sults demonstrate the utility of MC assay for detecting 22,23 gene fusions by increasing the number of oncogenes Acknowledgments examined to date. In addition, we showed that the MC This work was supported in part by grants-in-aid from assay can be used to analyze nonsurgical specimens (e.g., the Japan Agency for Medical Research and Development samples obtained by transbronchial biopsy and bron- (AMED) for the Health, and Labor Sciences Research choalveolar lavage samples) that are collected in the Expenses for Commission (the Practical Research for daily lung cancer clinics and used for pathological diag- Innovative Cancer Control); a grant-in-aid from the Jap- nosis. Notably, the aberrations in the six oncogenes anese Society for the Promotion of Science and Scientific examined in the present study were mutually exclusive; Research (B: 26293200); and the National Cancer Center they were also mutually exclusive with hot spot– Research and Development Fund for NCC BioBank activating mutations in the EGFR, HER2, and BRAF on- (15cK0106012h0002). cogenes in our LADC cohort (Supplementary Fig. 3). The high accuracy and robustness of the MC assay, even when used with FFPE tumor tissues, indicate that it will Supplementary Data complement other diagnostic assays used to detect hot Note: To access the supplementary material accompa- spot EGFR, HER2, and BRAF oncogene mutations and nying this article, visit the online version of Journal of thereby improve the diagnosis of LADC. Other novel Thoracic Oncology at www.jto.org and at http://dx.doi. druggable (but infrequent) oncogene fusions have been org/10.1016/j.jtho.2015.10.005. identified in lung cancer.31–34 The fact that the probe set used for the MC assay can be easily tailored to the genes References being examined is a great advantage. On the other hand, 1. Shaw AT, Hsu PP, Awad MM, Engelman JA. Tyrosine kinase the MC assay has a limitation in that it required a rela- gene rearrangements in epithelial malignancies. Nat Rev tively high amount of RNA (150 ng for frozen tissues and Cancer. 2013;13:772–787. 400 ng for FFPE tissues), which means that it is not 2. Oxnard GR, Binder A, Janne PA. New targetable onco- – applicable for samples from which an insufficient genes in non small-cell lung cancer. J Clin Oncol. 2013;31:1097–1104. amount of RNA is obtained. 3. Kwak EL, Bang YJ, Camidge DR, et al. Anaplastic lym- The present study also focused on MET exon 14 phoma kinase inhibition in non–small-cell lung cancer. skipping, a novel druggable aberration that is being New Engl J Med. 2010;363:1693–1703. examined in clinical trials as a therapeutic target.18,19 4. Shaw AT, Kim DW, Mehra R, et al. Ceritinib in ALK- Among seven such cases, we found that two were nega- rearranged non–small-cell lung cancer. New Engl J Med. tive after immunostaining FFPE tumor tissues for the 2014;370:1189–1197. MET protein (Supplementary Table 2 and Supplementary 5. Seto T, Kiura K, Nishio M, et al. CH5424802 (RO5424802) for patients with ALK-rearranged advanced non– Fig. 4); therefore, immunostaining is not a suitable small-cell lung cancer (AF-001JP study): a single-arm, method for detecting MET activation caused by exon open-label, phase 1-2 study. Lancet Oncol. 2013;14: skipping. Aberrant RNA splicing associated with intronic 590–598. mutations may be a mechanism that drives carcinogen- 6. Takeuchi K, Soda M, Togashi Y, et al. RET, ROS1 and ALK esis by activating oncogenes and inactivating tumor fusions in lung cancer. Nat Med. 2012;18:378–381. suppressor genes. Thus, the MC assay will be suitable for 7. Janne PA, Meyerson M. ROS1 rearrangements in lung detecting aberrantly spliced transcripts as well as gene cancer: a new genomic subset of lung adenocarcinoma. J Clin Oncol. 2012;30:878–879. fusions. 8. Kohno T, Ichikawa H, Totoki Y, et al. KIF5B-RET fusions in In summary, we have developed an assay that allows lung adenocarcinoma. Nat Med. 2012;18:375–377. simultaneous diagnosis of multiple druggable oncogene 9. Lipson D, Capelletti M, Yelensky R, et al. Identification of aberrations in FFPE lung cancer tissues with a high de- new ALK and RET gene fusions from colorectal and lung gree of accuracy and robustness. The advantage of this cancer biopsies. Nat Med. 2012;18:382–384. assay is that probe sets can be changed easily to 10. Yoshida A, Kohno T, Tsuta K, et al. ROS1-rearranged lung accommodate changes in the genes being investigated. cancer: a clinicopathologic and molecular study of 15 – The MC assay can be used not only to detect aberrant surgical cases. Am J Surg Pathol. 2013;37:554 562. 11. Drilon AE, Sima CS, Somwar R, et al. Phase II study of transcripts but also to examine levels, for patients with advanced RET-rearranged fi as exempli ed previously by PAM50, a Food and Drug lung cancers [abstract]. Presented at: 51st Annual Administration–approved assay used to classify intrinsic Meeting of the American Society of Clinical Oncology. breast cancer subtypes.35 Thus, multiplexed assays that May 29–June 2, 2015; Chicago, IL. 212 Sunami et al Journal of Thoracic Oncology Vol. 11 No. 2

12. Shaw AT, Ou SH, Bang YJ, et al. Crizotinib in ROS1- 24. Kinno T, Tsuta K, Shiraishi K, et al. Clinicopathological rearranged non–small-cell lung cancer. New Engl J Med. features of non–small cell lung carcinomas with BRAF 2015;372:683–684. mutations. Ann Oncol. 2014;25:138–142. 13. Falchook GS, Ordonez NG, Bastida CC, et al. Effect of 25. Saito M, Shimada Y, Shiraishi K, et al. Development of the RET inhibitor in a patient with RET lung adenocarcinomas with exclusive dependence on fusion-positive metastatic non–small-cell lung cancer oncogene fusions. Cancer Res. 2015;75:2264–2271. [e-pub ahead of print]. J Clin Oncol. http://dx.doi.org/ 26. Travis W, Brambilla E, Muller-Hermelink HK, Harris CC. 10.1200/JCO.2013.50.5016 accessed January 4, 2016. Pathology and Genetics: Tumours of the Lung, Pleura, 14. Nakaoku T, Tsuta K, Ichikawa H, et al. Druggable onco- Thymus and Heart. Lyon, France: IARC; 2004. gene fusions in invasive mucinous lung adenocarcinoma. 27. Li A, Liu Z, Lezon-Geyda K, et al. GPHMM: an integrated Clin Cancer Res. 2014;20:3087–3093. hidden Markov model for identification of copy number 15. Fernandez-Cuesta L, Plenker D, Osada H, et al. CD74- alteration and loss of heterozygosity in complex tumor NRG1 fusions in lung adenocarcinoma. Cancer Discov. samples using whole genome SNP arrays. Nucl Acids Res. 2014;4:415–422. 2011;39:4928–4941. 16. Kong-Beltran M, Seshagiri S, Zha J, et al. Somatic mu- 28. Kohno T, Nakaoku T, Tsuta K, et al. Beyond ALK-RET, tations lead to an oncogenic deletion of met in lung ROS1 and other oncogene fusions in lung cancer. Transl cancer. Cancer Res. 2006;66:283–289. Lung Cancer Res. 2015;4:156–164. 17. Onozato R, Kosaka T, Kuwano H, Sekido Y, Yatabe Y, 29. Seo JS, Ju YS, Lee WC, et al. The transcriptional land- Mitsudomi T. Activation of MET by gene amplification or scape and mutational profile of lung adenocarcinoma. by splice mutations deleting the juxtamembrane domain Genome Res. 2012;22:2109–2119. in primary resected lung cancers. J Thorac Oncol. 30. Tsuta K, Kozu Y, Mimae T, et al. c-MET/phospho-MET pro- 2009;4:5–11. tein expression and MET gene copy number in non–small 18. Paik PK, Drilon A, Yu H, et al. Response to MET inhibitors cell lung carcinomas. J Thorac Oncol. 2012;7:331–339. in patients with stage IV lung adenocarcinomas harboring 31. Vaishnavi A, Capelletti M, Le AT, et al. Oncogenic and MET mutations causing exon 14 skipping. Cancer Discov. drug-sensitive NTRK1 rearrangements in lung cancer. Nat 2015;5:842–849. Med. 2013;19:1469–1472. 19. Frampton GM, Ali SM, Rosenzweig M, et al. Activation of 32. Shan L, Jiang P, Xu F, et al. BIRC6-ALK, a novel fusion MET via diverse exon 14 splicing alterations occurs in gene in ALK break-apart FISH-negative lung adenocar- multiple tumor types and confers clinical sensitivity to cinoma, responds to crizotinib. J Thorac Oncol. MET inhibitors. Cancer Discov. 2015;5:850–859. 2015;10:e37–e39. 20. Mizukami T, Shiraishi K, Shimada Y, et al. Molecular 33. Jang JS, Lee A, Li J, et al. Common oncogene muta- mechanisms underlying oncogenic RET fusion in lung tions and novel SND1-BRAF transcript fusion in lung adenocarcinoma. J Thorac Oncol. 2014;9:622–630. adenocarcinoma from never smokers. Sci Rep. 2015;5: 21. Network CGAR. Comprehensive molecular profiling of 9755. lung adenocarcinoma. Nature. 2014;511:543–550. 34. Ou SH, Chalmers ZR, Azada MC, et al. Identification of a 22. Lira ME, Kim TM, Huang D, et al. Multiplexed gene novel TMEM106B-ROS1 fusion variant in lung adenocar- expression and fusion transcript analysis to detect cinoma by comprehensive genomic profiling. Lung Can- ALK fusions in lung cancer. JMolDiagn. 2013;15: cer. 2015;88:352–354. 51–61. 35. Dowsett M, Sestak I, Lopez-Knowles E, et al. Comparison 23. Lira ME, Choi YL, Lim SM, et al. A single-tube multi- of PAM50 risk of recurrence score with oncotype DX and plexed assay for detecting ALK, ROS1, and RET fusions in IHC4 for predicting risk of distant recurrence after lung cancer. J Mol Diagn. 2014;16:229–243. endocrine therapy. J Clin Oncol. 2013;31:2783–2790.