ORIGINAL ARTICLE Assessment of a New ROS1 Immunohistochemistry Clone (SP384) for the Identification of ROS1 Rearrangements in Patients with Non–Small Cell Lung Carcinoma: the ROSING Study Esther Conde, MD, PhD,a Susana Hernandez, PhD,b Rebeca Martinez,b Barbara Angulo, PhD,a Javier De Castro, MD, PhD,c Ana Collazo-Lorduy, MD, PhD,b Beatriz Jimenez, MD,b Alfonso Muriel, PhD,d Jose Luis Mate, MD,e Teresa Moran, MD,f Ignacio Aranda, MD, PhD,g Bartomeu Massuti, MD,g Federico Rojo, MD, PhD,h Manuel Domine, MD, PhD,i Irene Sansano, MD,j Felip Garcia, MD,k Enriqueta Felip, MD, PhD,j Nuria Mancheño, MD,l Oscar Juan, MD, PhD,l Julian Sanz, MD, PhD,m Jose Luis Gonzalez-Larriba, MD, PhD,m Lidia Atienza-Cuevas, MD,n Esperanza Arriola-Arellano, MD,n Ihab Abdulkader, MD,o Jorge Garcia-Gonzalez, MD,o Carmen Camacho, MD,p Delvys Rodriguez-Abreu, MD,p Cristina Teixido, PhD,q Noemi Reguart, MD, PhD,q Ana Gonzalez-Piñeiro, MD,r Martin Lazaro-Quintela, MD,r Maria Dolores Lozano, MD, PhD,s Alfonso Gurpide, MD,s Javier Gomez-Roman, MD, PhD,t Marta Lopez-Brea, MD,t Lara Pijuan, MD, PhD,u Marta Salido, PhD,u Edurne Arriola, MD, PhD,u Amparo Company, MD,v Amelia Insa, MD,v Isabel Esteban-Rodriguez, MD, PhD,c Monica Saiz, MD,w Eider Azkona, MD,w Ramiro Alvarez, MD,x Angel Artal, MD, PhD,x Maria Luz Plaza, MD, PhD,y David Aguiar, MD,y Ana Belen Enguita, MD,z Amparo Benito, MD,aa Luis Paz-Ares, MD, PhD,bb Pilar Garrido, MD, PhD,cc Fernando Lopez-Rios, MD, PhDa,* and Takeda. Dr. Gonzalez-Piñeiro has received honoraria from Pfizer, *Corresponding author. Roche, and AstraZeneca. Dr. Lazaro-Quintela has received honoraria Disclosure: Grupo HM Hospitales has received honoraria from Roche, from Roche, Pfizer, Eli Lilly, Merck Sharp and Dohme, and Bristol- Pfizer, Thermo Fisher Scientific, Bristol-Myers Squibb, and AbbVie. Myers Squibb. Dr. Lopez-Brea has received travel expenses from Dr. Conde has received honoraria from Pfizer and Roche and travel Roche and Bristol-Myers Squibb. Dr. Arriola has received honoraria expenses from Roche, Merck Sharp and Dohme, and Pfizer. Dr. from Bristol-Myers Squibb, Roche, Merck Sharp and Dohme, Pfizer, Hernandez has received honoraria from Roche and Bristol-Myers Eli Lilly, AstraZeneca, and Boehringer Ingelheim; research grants Squibb and travel expenses from Thermo Fisher Scientific, Pfizer, from Roche, Pfizer, and Bristol-Myers Squibb; and travel expenses and Roche. Dr. Angulo has received travel expenses from Thermo from Bristol-Myers Squibb, Roche, Merck Sharp and Dohme, and Eli Fisher Scientific. Dr. Massuti has received honoraria from Lilly. Dr. Esteban-Rodriguez has received honoraria from Roche, Boehringer Ingelheim, Roche, Bristol-Myers Squibb, Merck Sharp and AstraZeneca, and Merck Sharp and Dohme and travel expenses from Dohme, AstraZeneca, Amgen, Pfizer, Merck Serono, and Janssen and Merck Sharp and Dohme. Dr. Paz-Ares has received honoraria from travel expenses from Roche, Merck Sharp and Dohme, AstraZeneca, Roche, Novartis, Eli Lilly, Boerhinger Ingelheim, AstraZeneca, and Boehringer Ingelheim. Dr. Rojo has received honoraria from Bristol-Myers Squibb, Pfizer, Merck Sharp and Dohme, Clovis Pfizer, Novartis, AstraZeneca, Merck Sharp and Dohme, Bristol- Oncology, Merck Serono, Amgen, Celgene, PharmaMar, and Sanofi. Myers Squibb, Merck, Genomic Health, Guardant Health, AbbVie, Dr. Garrido has received honoraria from Roche, Merck Sharp and and Roche. Dr. Sansano has received honoraria from Pfizer, Roche, Dohme, Bristol-Myers Squibb, Boerhinger Ingelheim, Pfizer, AbbVie, Merck Sharp and Dohme, AbbVie, Takeda, and AstraZeneca and Guardant Health, Novartis, Eli Lilly, AstraZeneca, Janssen, Sysmex, travel expenses from Pfizer, Roche, and AstraZeneca. Dr. Felip has Blueprint Medicines, Takeda, and Rovi and research grants from received honoraria from Roche, AbbVie, AstraZeneca, Bergenbio, Guardant Health and Sysmex. Dr. Lopez-Rios has received research Blueprint Medicines, Boehringer Ingelheim, Bristol-Myers Squibb, funding from Bristol-Myers Squibb, Pfizer, Roche, AbbVie, and Celgene, Eli Lilly, Guardant Health, Janssen, Medscape, Merck Thermo Fisher Scientific and travel expenses and honoraria from Serono, Merck Sharp and Dohme, Novartis, Pfizer, Prime Oncology, AbbVie, Bayer, Roche, AstraZeneca, Bristol-Myers Squibb, Merck Samsung, Takeda, and Touchtime. Dr. Juan has received honoraria Sharp and Dohme, Pfizer, and Thermo Fisher Scientific. The from Boehringer Ingelheim, Bristol-Myers Squibb, Merck Sharp and remaining authors declare no conflict of interest. Dohme, Roche/Genentech, AstraZeneca, and AbbVie. Dr. Garcia- Gonzalez has received honoraria from AstraZeneca, Pierre-Fabré, Address for correspondence: Fernando Lopez-Rios, MD, PhD, Bristol-Myers Squibb, Boehringer Ingelheim, Eli Lilly, Merck Sharp Pathology-Laboratorio de Dianas Terapeuticas, HM Sanchinarro Uni- and Dohme, Roche, and Ipsen and travel expenses from Bristol- versity Hospital, Calle de Oña, 10, 28050 Madrid, Spain. E-mail: Myers Squibb, Merck Sharp and Dohme, and Roche. Dr. Rodriguez- flopezrios@hmhospitales.com Abreu has received grants from Bristol-Myers Squibb and honoraria ª 2019 International Association for the Study of Lung Cancer. from Bristol-Myers Squibb, Merck Sharp and Dohme, Roche/ Published by Elsevier Inc. This is an open access article under the Genentech, AstraZeneca, Boehringer Ingelheim, Eli Lilly, and CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc- Novartis. Dr. Teixido has received honoraria from Roche, Takeda, nd/4.0/). Pfizer, and Bristol-Myers Squibb and research grants from Novartis. Dr. Reguart has received honoraria from Roche, Merck Sharp and ISSN: 1556-0864 Dohme, Bristol-Myers Squibb, Boerhinger Ingelheim, Pfizer, https://doi.org/10.1016/j.jtho.2019.07.005 Guardant Health, AbbVie, Ipsen, Eli Lilly, AstraZeneca, Novartis, Journal of Thoracic Oncology Vol. 14 No. 12: 2120-2132 December 2019 ROS1 Immunohistochemistry with Clone SP384 2121 a HM Sanchinarro University Hospital-CIBERONC, Madrid, Spain b HM Sanchinarro University Hospital, Madrid, Spain c La Paz University Hospital, Madrid, Spain d Ramon y Cajal University Hospital, IRYCIS and CIBERESP, Madrid, Spain e Germans Trias i Pujol University Hospital, Badalona, Spain f Catalan Institute of Oncology-Germans Trias i Pujol University Hospital, Universitat Autònoma Barcelona (UAB), Badalona- Applied Research Group of Oncology (B-ARGO), Badalona, Spain g General University Hospital-ISABIAL, Alicante, Spain h Institute of Health Research-Jimenez Diaz Foundation-CIBERONC, Madrid, Spain i Institute of Health Research-Jimenez Diaz Foundation, Madrid, Spain j Vall d’Hebron University Hospital, Barcelona, Spain k Quironsalud Hospital, Barcelona, Spain l La Fe University Hospital, Valencia, Spain m Clinico San Carlos University Hospital, Madrid, Spain n Puerta del Mar University Hospital, Cadiz, Spain o Clinico de Santiago University Hospital, Santiago De Compostela, Spain p Insular Materno-Infantil University Hospital Complex, Las Palmas de Gran Canaria, Spain q Clinic Hospital, Barcelona, Spain r Alvaro Cunqueiro Hospital, Vigo, Spain s University of Navarra Clinic, Pamplona, Spain t Marques de Valdecilla University Hospital, Santander, Spain u Hospital del Mar, Barcelona, Spain v Clinico University Hospital, Valencia, Spain w Cruces University Hospital, Baracaldo, Spain x Miguel Servet University Hospital, Zaragoza, Spain y University Hospital of Gran Canaria Doctor Negrin, Las Palmas de Gran Canaria, Spain z 12 de Octubre University Hospital, Madrid, Spain aa Ramon y Cajal University Hospital, Madrid, Spain bb 12 de Octubre University Hospital-CIBERONC, Madrid, Spain cc Ramon y Cajal University Hospital-CIBERONC, Madrid, Spain Received 10 May 2019; revised 15 July 2019; accepted 16 July 2019 Available online - 23 July 2019 ABSTRACT (91% sensitivity and 100% specificity). ROS1 expression in normal lung was more frequent with use of the SP384 clone Introduction: The ROS1 gene rearrangement has become (p < 0.0001). The ezrin gene (EZR)-ROS1 variant was an important biomarker in NSCLC. The College of American associated with membranous staining and an isolated green Pathologists/International Association for the Study of Lung signal FISH pattern (p ¼ 0.001 and p ¼ 0.017, respectively). Cancer/Association for Molecular Pathology testing guide- lines support the use of ROS1 immunohistochemistry (IHC) Conclusions: The new SP384 ROS1 IHC clone showed as a screening test, followed by confirmation with fluores- excellent sensitivity without compromising specificity, so it cence in situ hybridization (FISH) or a molecular test in all is another excellent analytical option for the proposed positive results. We have evaluated a novel anti-ROS1 IHC testing algorithm. antibody (SP384) in a large multicenter series to obtain  2019 International Association for the Study of Lung real-world data. Cancer. Published by Elsevier Inc. This is an open access Methods: A total of 43 ROS1 FISH–positive and 193 ROS1 article under the CC BY-NC-ND license (http:// FISH–negative NSCLC samples were studied. All specimens creativecommons.org/licenses/by-nc-nd/4.0/). were screened by using two antibodies (clone D4D6 from Cell Signaling Technology and clone SP384 from Ventana Keywords: ROS1; Immunohistochemistry; FISH; Next- Medical Systems), and the different interpretation criteria generation sequencing; Lung carcinoma were compared with break-apart FISH (Vysis). FISH- positive samples were also analyzed with next-generation sequencing (Oncomine Dx Target Test Panel, Thermo Introduction Fisher Scientific). The ROS1 gene rearrangement has now become an Results: An H-score of 150 or higher or the presence of at important predictive biomarker for targeted tyrosine least 70% of tumor cells with an intensity of staining of 2þ kinase inhibitors in NSCLC. In March 2016, crizotinib or higher by the SP384 clone was the optimal cutoff value was approved by the U.S. Food and Drug Administration (both with 93% sensitivity and 100% specificity). The D4D6 (FDA) for the treatment of patients with advanced ROS1- clone showed similar results, with an H-score of at least 100 rearranged NSCLCs without the requirement of use of an 2122 Conde et al Journal of Thoracic Oncology Vol. 14 No. 12 FDA-approved companion diagnostic test.1 Soon after- This situation prompted us to evaluate a novel anti- ward, the drug was approved by the European Medicines ROS1 IHC antibody in a large multicenter series to Agency, with the statement that “an accurate and vali- obtain real-world data for the proposed ROS1 testing dated ROS1 assay is necessary for the selection of pa- algorithm. tients.”2 On the basis of the excellent results of the crizotinib clinical trials and the development of other ROS1 inhibitors with consistent efficacy results in Material and Methods this patient population, the importance of accurately Study Design and Tumor Samples identifying ROS1-positive lung cancer has never been The flow diagram is depicted in Figure 1. A total of 55 greater.3-8 ROS1-positive samples from patients with NSCLCs that Regarding the detection of ROS1 rearrangements, had been initially tested as part of routine clinical care in the recently updated College of American Patholo- 23 different institutions were used for this study (also gists/International Association for the Study of Lung known as ROSING [ROS IHC and NGS]). To confirm the Cancer/Association for Molecular Pathology (CAP/ ROS1-positive status, FISH analysis (the criterion stan- IASLC/AMP) molecular testing guidelines for the se- dard method) was performed at the referral institution lection of patients with lung cancer support the use (i.e., University Hospital HM Sanchinarro). Only cases of ROS1 immunohistochemistry (IHC) as a screening with enough tissue available (i.e., 50 tumor cells, as per test, followed by fluorescence in situ hybridization the FISH test requirements) and ROS1 FISH–confirmed (FISH) (traditionally considered to be the criterion positivity were included. In addition, 193 consecutive standard method)9 or a molecular test (i.e., reverse- ROS1 FISH–negative samples from NSCLCs tested at 14 transcriptase polymerase chain reaction or next- of the participating institutions as part of routine clinical generation sequencing [NGS]) in all cases with posi- care were included as negative controls. The material tive IHC results.10 To date, only one anti-ROS1 IHC available for all tumors was formalin-fixed and paraffin- clone has been commercially available, and there is embedded (FFPE). The specifics of formalin fixation no universally accepted criterion for the interpreta- were unknown. All cases were reviewed by two pa- tion of ROS1 IHC results.10,11 thologists (E. C. and F. L-R.). In addition to being ROS1-posive tumor samples ROS1-negave tumor samples n = 55 n = 193 Excluded (n = 12) - Less than 50 tumor cells (n = 4) - FISH not evaluable (n = 8) ROS1-posive samples by FISH* ROS1-negave samples by FISH* n = 43 n = 193 Excluded (n = 2) - Inaccesible medical records ROS1 FISH-posive samples with clinical data n = 41 ROS1 FISH-posive samples with ROS1 FISH-negave samples with Excluded (n = 16) - Early-stage tumors (n = 13) ROS1 IHC data ROS1 IHC data - Died before treatment (n = 2) n = 43 n = 193 - Chemotherapy alone (n = 1) Excluded (n = 9) ROS1 FISH-posive paents - Insufficient sequencing coverage treated with crizonib n = 25 ROS1 FISH-posive samples with Excluded (n = 4) NGS data - Died before evaluaon (n = 3) - Neoadyuvant crizonib in stage IIIB n = 34 (n = 1) ROS1 FISH-posive paents treated with clinical follow-up n = 21 Figure 1. Flowchart of patients in the ROSING study. *ROS1 fluorescence in situ hybridization (FISH)–positive samples were defined as those with more than 25 (50%) break-apart (BA) signals or isolated green signals (IGS) in tumor cells. ROS1 FISH– negative samples were defined as those with fewer than five (10%) BA signals or IGS in tumor cells. ROS1 FISH–positive samples were defined as those with more than 25 (50%) break-apart (BA) signals or isolated green signals (IGS) in tumor cells. IHC, immunohistochemistry; NGS, next-generation sequencing. December 2019 ROS1 Immunohistochemistry with Clone SP384 2123 investigated by FISH, all specimens (negative and posi- ROS1-positive external controls were included in all tive) were independently screened for ROS1 expression the slides. by using two IHC antibodies. ROS1 FISH–positive cases The slides were reviewed by two pathologists (E. C. were also tested by NGS. Clinical data from patients with and F. L-R.) who were blinded to the FISH results. When ROS1 FISH–positive tumors were collected. The institu- a discrepancy was observed, the final result was ob- tional ethics committee at Grupo HM Hospitals reviewed tained by consensus. Staining intensity was defined as and approved this study. Each referring institution follows: strong cytoplasmic staining (3þ), which was regulated the need for additional specific consent, as clearly visible with use of a 2 or 4 objective; mod- ROS1 testing is part of routine clinical care. Clinical data erate staining (2þ), which required use of a 10 or 20 were retrieved from the patient clinical records. objective; weak staining (1þ), which involved use of a 40 objective; and negative staining (0), which was defined as absence of expression.12 The percentages of FISH for ROS1 Rearrangements tumor cells with each staining intensity were also eval- FISH was repeated centrally on unstained 4-mm-thick uated. Membrane staining was recorded when observed. FFPE tumor tissue sections from all positive and nega- ROS1 IHC staining results with both clones were finally tive cases. The Vysis 6q22 ROS1 Break Apart FISH Probe interpreted by using four previously described criteria: Kit (Abbott Molecular, Des Plaines, IL) was used (1) an H-score with a threshold for ROS1 positivity following the manufacturer’s instructions as previously defined as at least 100,11,13 (2) an H-score cutoff of at described.12 The ROS1 FISH assay was independently least 150,11,14 (3) an intensity criterion with the cutoff of captured and scored with the automated BioView Duet positivity defined as 2þ or higher in any tumor scanning system (BioView, Rehovot, Israel) by an expe- cells,11,15,16 and (4) a positive status based on an in- rienced lung pathologist (E. C.) and molecular biologist tensity of 2þ or higher in at least 30% of the total tumor (S. H.). The interpretation criteria strictly followed highly cells.17 Intratumoral staining heterogeneity was also recommended criteria.11 A minimum of 50 tumor nuclei evaluated. It was defined as the presence of areas of were counted. ROS1 FISH–positive cases were defined as staining with an intensity of 0 or 1þ in positive cases.16 those with more than 25 break-apart (BA) signals (50%) The positivity of normal lung tissue was recorded when (separated by at least one signal diameter) or an isolated it was present on the sections. green signal (IGS) in tumor cells. ROS1 FISH–negative samples were defined as those with fewer than five (10%) BA signals or IGS in tumor cells. ROS1 FISH cases NGS for ROS1 Rearrangements were considered borderline if five to 25 (10%–50%) For each FFPE tumor sample, freshly cut 5-mm-thick cells were positive. In the case of borderline results, a sections were collected for nucleic acid extraction: five second reader evaluated the slide, added cell count sections for surgical specimens and 12 sections for small readings from the already automatically captured im- biopsy specimens. The first and last sections were ages, and calculated the percentage out of 100 cells. If stained with hematoxylin and eosin and reviewed by two the percentage of positive cells was lower than 15%, the pathologists (E. C. and F. L.-R.) to assess the percentage sample was considered negative. If the percentage of of tumor cells. RNA extraction was performed with the positive cells was higher than or equal to 15%, the RecoverAll Total Nucleic Acid Isolation Kit (Thermo sample was considered positive.11 Fisher Scientific, Vilnius, Lithuania) following the man- ufacturer’s instructions. The RNA was then purified and concentrated by using the GeneJET RNA cleanup and IHC for ROS1 Expression concentration micro kit (Thermo Fisher Scientific). Automated IHC for ROS1 expression was performed The Oncomine Dx Target Test panel (Thermo Fisher for all cases on a BenchMark ULTRA staining instrument Scientific) was the selected approach because it requires (Ventana Medical Systems, Tucson, AZ). FFPE tumor very little input RNA and it was the first FDA-approved tissues were sectioned at a thickness of 4 mm and NGS test. The protocol for the NGS analysis followed stained with two different anti-ROS1 clones: SP384 the manufacturer0 s instructions, and a minimum of 5000 (Ventana Medical Systems) and D4D6 (Cell Signaling mapped fusion panel reads was required for ROS1 fusion Technology, Danvers, MA). Briefly, the Ventana ROS1 analysis. Consent was granted only for the RNA part of (SP384) ready-to-use rabbit monoclonal primary anti- the procedure. body was applied with the OptiView DAB IHC Detection Kit and OptiView Amplification Kit, following the man- ufacturer’s instructions. The D4D6 clone was used at a Statistical Analysis 1:50 dilution. Detection was performed with the same On the basis of all the valid data obtained, we per- OptiView detection-amplification kit. FISH-validated formed a descriptive analysis of all the variables of 2124 Conde et al Table 1. Performance of ROS1 IHC Using the Previously Published Criteria to Predict ROS1 Rearrangements by FISH ROS1 FISH Sensitivity, % Specificity, % LR-Positive LR-Negative ROS1 IHC IHC Result FISH-Positive FISH-Negative Total, % (95% CI) (95% CI) (95% CI) (95% CI) Clone SP384 Criterion 1: H-score 100 IHC-positive 40 1 41 (17.4) 93 (81–98) 99 (97–100) 180 (25.4–1270) 0.1 (0–0.2) IHC-negative 3 192 195 (82.6) Criterion 2: H-score 150 IHC-positive 40 0 40 (16.9) 93 (81–98) 100 (98–100) 0.1 (0–0.2) IHC-negative 3 193 196 (83.1) Criterion 3: 2þ staining IHC-positive 40 31 71 (30.1) 93 (81–98) 84 (78–89) 5.8 (4.1–8) 0.1 (0–0.2) IHC-negative 3 162 165 (69.9) Criterion 4: 2þ staining IHC-positive 40 1 41 (17.4) 93 (81–98) 99 (97–100) 180 (25.4–1270) 0.1 (0–0.2) in 30% of tumor cells IHC-negative 3 192 195 (82.6) Clone D4D6 Criterion 1: H-score 100 IHC-positive 39 0 39 (16.5) 91 (78–97) 100 (98–100) 0.1 (0–0.2) IHC-negative 4 193 197 (83.5) Journal of Thoracic Oncology Criterion 2: H-score 150 IHC-positive 37 0 37 (15.7) 86 (72–95) 100 (98–100) 0.1 (0.1–0.3) IHC-negative 6 193 199 (84.3) Criterion 3: 2þ staining IHC-positive 39 14 53 (22.5) 91 (78–97) 93 (88–96) 12.5 (7.5–20.9) 0.1 (0–0.2) IHC-negative 4 179 183 (77.5) Criterion 4: 2þ staining in 30% of IHC-positive 37 0 37 (15.7) 86 (72–95) 100 (98–100) 0.1 (0.1–0.3) tumor cells IHC-negative 6 193 199 (84.3) CI, confidence interval; FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; LR, likelihood ratio. Vol. 14 No. 12 December 2019 ROS1 Immunohistochemistry with Clone SP384 2125 A 1. 00 B 1.00 SP384 ROS1 IHC SP384 ROS1 IHC 0.75 0.75 H-score ≥ 150 ≥ 2+ staining in ≥ 70% of total tumor cells Sensitivity Sensitivity 0 .50 0.50 0.25 0.25 Area under ROC curve = 0.9738 0.00 Area under ROC curve = 0.9595 0.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 1 - Specificity 1 - Specificity C D 1. 00 1.00 D4D6 ROS1 IHC 0.75 D4D6 ROS1 IHC 0.75 H-score ≥ 100 ≥ 2+ staining in ≥ 30% of total tumor cells Sensitivity Sensitivity 0.50 0.50 0.25 0.25 0.00 0.00 Area under ROC curve = 0.9438 Area under ROC curve = 0.9496 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 1 - Specificity 1 - Specificity Figure 2. Receiver operating characteristic (ROC) curve analyses identified an H-score of 150 or higher (A) or the presence of at least 70% of cells with a staining intensity of 2þ or higher (B) with use of the SP384 clone as the optimal cutoff value for identifying ROS1 translocations by fluorescence in situ hybridization (both with 93% sensitivity and 100% specificity). Regarding the D4D6 clone, the optimal cutoff value was an H-score of at least 100 (C) (with 91% sensitivity and 100% specificity), followed by the presence of at least 30% of cells with a staining intensity of 2þ or higher (D) (with 86% sensitivity and 100% specificity). IHC, immunohistochemistry. interest. The test used for comparison of categorical Results variables was the Pearson chi-square test (frequency <5 ROS1 Rearrangements Assessed by FISH [Fisher exact test]). For comparison of means we used the Of the 55 ROS1-positive lung carcinoma specimens, Mann-Whitney test. The sensitivity and specificity of both four were excluded for lack of sufficient tumor tissue ROS1 IHC clones versus those of FISH were obtained. and eight were excluded because the FISH results were Receiver operating characteristic (ROC) curves were used not evaluable (i.e., no or weak hybridization signals). Of to determine the optimal cutoff value that discriminates the 193 ROS1-negative NSCLC specimens, all were between patients with ROS1-rearranged and ROS1-non- included in the study (see Fig. 1). Among the 43 ROS1 rearranged tumors. We also analyzed the correlation be- FISH–positive cases analyzed, 27 tumors (62.8%) had a tween the different ROS1 fusion variants and BA pattern and 16 (37.2%) showed an IGS pattern. The clinicopathologic features. Survival analysis was per- total number of tumor cells analyzed was 50 in all cases formed by using the Kaplan-Meier method with the log- (97.7%) except in one specimen (2.3%) (a case with rank test and Cox regression. All analyses were done initial borderline results in which 100 nuclei had to be with Stata 15.1 software (StataCorp, College Station, TX) scored). In ROS1 FISH–negative cases, the mean per- and were two sided; p values less than 0.05 indicated centage of positive tumor cells was 0.4% (median 0%, statistical significance. range 0%–10%). In the ROS1 FISH–positive tumors, the 2126 Conde et al Journal of Thoracic Oncology Vol. 14 No. 12 Figure 3. Most of the ROS1-positive tumors showed homogeneous staining with the SP384 clone (A [top inset]), whereas intratumoral heterogeneity was more frequently observed with the D4D6 antibody (B [upper inset]). Moreover, as shown in the lower insets, ROS1 expression was more frequent in nonneoplastic type II pneumocytes with use of the SP384 clone (A) than with use of the D4D6 antibody (B). mean percentage of positive cells was 82.3% (median ROS1 Immunoreactivity by IHC 86%, range 49%–98%). There were no significant dif- The IHC results using the previously published ferences in the percentages of positive cells between the criteria are summarized in Table 1. In addition, the ROC two patterns of positivity. analyses showed that an H-score of 150 or higher Figure 4. Representative images of the different topographic immunohistochemistry patterns. A tumor with a linear membranous accentuation staining with the SP384 (A) and D4D6 (B) clones, respectively. Another case with a diffuse and granular cytoplasmic staining using the SP384 clone (C) and the D4D6 antibody (D). December 2019 ROS1 Immunohistochemistry with Clone SP384 2127 (criterion 2) or the presence of at least 70% of cells with Table 2. Clinicopathologic Features of Patients with ROS1 at least 2þ staining by the SP384 clone was the optimal Rearrangements cutoff value for identifying ROS1 translocations by FISH Patients, n (%)a (both with 93% sensitivity and 100% specificity). Feature (n ¼ 43) Regarding the D4D6 clone, the optimal cutoff value was Tumor histologic type criterion 1 (with 91% sensitivity and 100% specificity), AC 39 (90.7) followed by criterion 4 (Fig. 2). The IHC concordance SCC 1 (2.3) between observers was almost perfect (data not shown). NSCLC-NOS 3 (7) In accordance with the defined optimal criteria, when Specimen type the SP384 clone was used, 40 cases (16.9%) were pos- Surgical 28 (65.1) itive and 196 cases (83.1%) were negative. The mean Small biopsy 11 (25.6) Cell block 4 (9.3) H-score of the SP384 ROS1–positive cases was 291 Age at diagnosis, y a (median 300, range 160–300), and the mean of cells with Mean 59 staining scored as 2þ or higher was 98.9% (median 100, Median 60 range 70–100). Interestingly, in 37 out of 40 SP384 Range 32–83 ROS1–positive cases (92.5%), immunoreactivity was Sexa evident as diffuse staining with an intensity of 2þ or Male 24 (58.5) Female 17 (41.5) higher. Heterogeneity was present in 7.5% of cases Smoking statusa (Fig. 3A). With the D4D6 clone, we observed 39 positive Nonsmoker 26 (63.4) cases (16.5%) and 197 negative tumors (83.5%). The Smoker 15 (36.6) mean H-score of the D4D6 ROS1–positive cases was 243 Stage at initial diagnosisa (median 260, range 100–300), and the mean percentage I 8 (19.5) of tumor cells with staining graded 2þ or higher was II 5 (12.2) 82.3% (median 90, range 10–100). Of the 39 D4D6 III 10 (24.4) IV 18 (43.9) ROS1–positive cases, 22 (56.4%) showed intratumoral Metastasis sites for stage IV diseasea 26 heterogeneity (Fig. 3B). Interestingly, in the positive Lung 3 (11.5) cases the difference in intratumoral heterogeneity be- Brain 1 (3.8) tween the two clones was statistically significant (p < Bone 3 (11.5) 0.0001). Lymph nodes 1 (3.8) Regarding the SP384 ROS1–negative tumors, Pleural 3 (11.5) Multiple organs 12 (46.2) immunoreactivity ranged from absent (133 of 196 Other or unknown 3 (11.5) [67.9%]) to focal and weak (1þ) or moderate (2þ) Crizotinib treatment lineb staining (63 of 196 [32.1%]), with a mean H-score of First 12 (48) 10.6 (median 0, range 0–130) and a mean percentage Second 8 (32) of cells with staining graded 2 or higher of 1.9% Third 5 (20) (median 0, range 0–40). With the D4D6 clone, 157 out Response rate of crizotinibc of 197 ROS1 IHC–negative cases (79.7%) showed PD 3 (14.3) Stable disease 1 (4.8) absence of immunoreactivity, whereas the remaining PR 16 (76.2) cases (40 of 197 [20.3%]) exhibited a focal staining CR 1 (4.8) pattern graded 1þ to 2þ. The mean H-score was 3.8 a Clinical information was available for 41 out of 43 patients. (median 0, range 0–75) and the mean percentage of b Patients with stage IV disease treated with crizotinib (n ¼ 25). c cells with staining graded 2þ or higher was 0.6% Patients who were treated with crizotinib and had clinical follow-up in- formation available (n ¼ 21). (median 0, range 0–20). AC, adenocarcinoma; CR, complete response; NOS, not otherwise specified; We observed the same topographic staining pattern PR, partial response; PD, progressive disease; SCC, squamous cell carcinoma. with both ROS1 IHC antibodies. A granular or diffuse cytoplasmic staining was present in all cases with immu- noreactivity (ROS1-positive and ROS1-negative cases), nodule or in a subpleural location) was statistically more whereas a linear membranous accentuation was observed frequent with use of the SP384 clone (104 of 107 only in ROS1-positive tumors (14 of 40 examined by SP384 [97.2%]) than with use of the D4D6 antibody (63 of 107 IHC [35%] and 14 of 39 examined by D4D6 IHC [35.9%]) [58.9%]) (p < 0.0001) (see Fig. 3). (Fig. 4). There was no significant association between the topographic IHC patterns and the FISH patterns. ROS1 Rearrangements Assessed by NGS Finally, ROS1 expression in nonneoplastic type II Analysis by NGS was successful in 34 tumors (79%). pneumocytes (especially in the periphery of the tumor Results could not be assessed in nine cases owing to 2128 Conde et al Journal of Thoracic Oncology Vol. 14 No. 12 insufficient sequencing coverage: four cases had a very with use of all the criteria. Conversely, immunoreac- limited tumor cell content (i.e., 5%–10%), and in five tivity was absent with use of the D4D6 ROS1 antibody. cases results could not be obtained because of RNA Clinically, the patient was a 67-year-old smoking male degradation (for example, one of the biopsy specimens with a stage IV lung AC that had been diagnosed in a cell was a decalcified bone sample). Fourteen cases (41.2%) block; he received crizotinib with a partial response. An had a CD74 molecule gene (CD74)-ROS1 fusion (11 cases NGS result was not available. corresponding to CD74 exon 6–ROS1 exon 34 fusion and In addition, if we consider criteria 2 and 4, two ROS1 three corresponding to CD74 exon 6–ROS1 exon 32 FISH–positive cases were clearly positive according to fusion), nine cases (26.5%) showed an ezrin gene (EZR) SP384 antibody staining (i.e., H-scores of 230 and 300 exon 10–ROS1 exon 34 fusion, six cases (17.6%) had a and with 95% and 100% of tumor cells with staining syndecan 4 gene (SDC4) exon 2–ROS1 exon 32 fusion, graded 2þ or higher, respectively), whereas they should four cases (11.8%) presented a solute carrier family 34 be considered negative according to staining with the member 2 gene (SLC34A2)-ROS1 fusion (three corre- D4D6 clone (i.e., H-scores of 105 and 100 with 20% and sponding to SLC34A2 exon 13–ROS1 exon 32 fusion and 10% of tumor cells with staining graded 2þ or higher, one corresponding to SLC34A2 exon 13–ROS1 exon 34 respectively). NGS confirmed the ROS1 fusions (EZR- fusion), and finally one sample (2.9%) contained a ROS1 and CD74-ROS1 variants, respectively). Clinically, tropomyosin 3 gene (TPM3) exon 7–ROS1 exon 35 both patients were nonsmoking males with ACs who fusion. Interestingly, among the nine EZR-ROS1–positive received crizotinib, resulting in objective responses. tumors, eight (88.9%) showed membranous accentua- All discordant cases were independently reviewed (F. tion staining with both ROS1 IHC antibodies and six L-R.), and the results were confirmed. Remarkably, the (66.7%) presented an IGS FISH pattern. Both associa- results for all ROS1 NGS–positive tumors were in tions were statistically significant (p ¼ 0.001 and p ¼ agreement with those obtained by FISH. 0.017, respectively). CD74-ROS1–positive tumors more frequently exhibited cytoplasmic staining with both Correlation between ROS1 Rearrangements and ROS1 IHC clones (12 versus two [p ¼ 0.009]) and a BA Clinicopathologic Data FISH pattern (10 versus four [p ¼ 0.495]). The results of The clinicopathologic features of the 43 ROS1 FISH– all three assays in the FISH-positive cases are detailed in positive tumors are detailed in Table 2. Briefly, 31 Supplementary Table 1. cases (72.1%) were diagnosed as primary lung cancer (cancer originating in the lung), whereas 12 (27.9%) Discordances between ROS1 Assays were metastases from different sites. Thirty-nine tumors Of the 43 ROS1 FISH–positive tumors, three showed (90.7%) were ACs, one (2.3%) was a squamous cell absent (0) or focal (1þ) cytoplasmic staining with both carcinoma, and the remaining three (7%) were NSCLCs antibodies and were therefore considered ROS1 IHC– not otherwise specified. Of the 39 ACs, 20 (51.3%) were negative with use of all the criteria. Unfortunately, NGS observed to have a predominant acinar pattern, 14 results were not available for these cases. Clinically, all (35.9%) presented a solid architecture, two (5.1%) had a three patients were males with a history of smoking. predominant lepidic pattern, one (2.6%) showed a Interestingly, one patient had metastatic poorly differ- papillary growth, and one (2.6%) had a predominant entiated squamous cell carcinoma diagnosed by a micropapillary pattern. Mucinous and/or signet ring bronchial biopsy (i.e., p40 positive by IHC), with a pre- cells were observed in six out of 39 ACs (15.4%). dominantly BA FISH pattern (78% of positive cells); the Interestingly, psammomatous calcifications and pleo- patient received crizotinib treatment but had progres- morphic features were frequently observed (in 18.6% sive disease. The remaining two patients had adenocar- and 30.2% of tumors, respectively). cinomas (ACs) that were diagnosed by surgical Clinical data were available for 41 patients (see Fig. 1 specimens (i.e., lung and bone resections) with an IGS and Table 2). Briefly, the overall response rate was 81% FISH pattern (90% and 52% of positive cells, respec- and the disease control rate was 85.7%. At the time of tively). Only one of these two patients received crizotinib report, the median progression-free survival (PFS) and and had progressive disease. overall survival times were 10.8 and 16.6 months, Moreover, one ROS1 FISH–positive case (i.e., 98% of respectively. There were no relevant associations be- positive cells with an IGS FISH pattern) showed immu- tween ROS1 fusion variants and clinicopathologic char- noreactivity with use of the SP384 clone (with an H- acteristics, except for a nonsignificant trend toward score of 160 and staining in 70% of tumor cells graded better PFS in patients with the EZR-ROS1 variant 2þ or higher) and was considered ROS1 IHC–positive (p ¼ 0.199). December 2019 ROS1 Immunohistochemistry with Clone SP384 2129 Discussion cell block (the only true discordant positive sample be- This multicenter study provided real-world data on tween both clones). Accordingly, pathologists should try ROS1 rearrangements in patients with NSCLC. To the to select blocks for ROS1 IHC testing that contain normal best of our knowledge, this series represents one of the lung, and extreme caution must be taken afterward so as largest ROS1-positive lung cancer cohorts ever assem- not to overinterpret the immunoreactivity in such bled. Considering that patients with ROS1 rearrangement normal or hyperplastic pneumocytes.11,15 Along these represent only 1% to 2% of the overall population with lines, positivity with D4D6 has been described in ROS1- NSCLC, few reports contain more than 50 patients.18-23 nonrearranged tumors having lepidic patterns of growth Moreover, a careful review of published studies identi- or containing EGFR mutations (see the next para- fied only two larger series in which positive tumors had graph).14,30 This potentially confounding situation could been investigated with more than two methodolo- be used to our advantage when searching for external gies.19,22 One potential caveat of our work is that it is a positive controls. retrospective series and therefore conclusions regarding Although our findings in the ROS1-nonrearranged ROS1 inhibition are limited. To partially overcome this cohort should be interpreted with extreme caution to shortcoming, it is relevant to emphasize that all samples avoid sample size bias,31,32 we truly believe that the were initially tested with intention to treat, so our results might represent the clinical reality (i.e., these findings represent the clinical reality. In fact, the clinical were not referral cases, and we chose not to use tissue results are in complete agreement with those of other microarrays). The specificity of the two clones could series.4,24 Moreover, we used commercially available very well be 100% if very stringent interpretation tools, so our findings could be replicated elsewhere. criteria were used. The best option would be an H-score Although the recently updated CAP/IASLC/AMP mo- of at least 100 for D4D6, but the higher sensitivity of lecular testing guidelines allow the use of ROS1 IHC for SP384 comes at a cost and higher cutoffs are needed to screening purposes, there has been only one antibody avoid what could be considered an excessive number of available to date (D4D6).9-11 The sensitivity for this orthogonal tests (98% versus 100% specificity). How- clone was controversial, probably reflecting the different ever, a broadly held consensus on the interpretation interpretation criteria and the small numbers that were criteria required for a positive IHC score has yet to tested in most studies (reviewed in9,10,25-28). The recent emerge.10,11 There are several lines of evidence that are release of a new clone (SP384), with only one published worth considering when addressing this matter. Un- report available to date, provides an in vitro diagnostic questionable ROS1 IHC expression (i.e., even strong but alternative.23 focal) with D4D6 has been described in ROS1-non- Several conclusions can be drawn from our study. rearranged cases containing other druggable alterations SP384 is more sensitive than D4D6 when compared with (mainly EGFR mutations, but also KRAS mutations, BRAF FISH, regardless of the criterion used. There are two mutations, ALK receptor tyrosine kinase gene [ALK] fu- differential features of SP384 that can be extremely sions, and erb-b2 receptor tyrosine kinase 2 gene [HER2] useful to reduce the risk of a false-negative result. First, abnormalities), and we have had anecdotal analogous there is the extremely frequent homogeneous staining experience with SP384 (E. Conde, unpublished obser- (>92%) for ROS1. Considering the small size and limited vation, 2019).14-16,25,30,33,34 Therefore, it is not surpris- number of fragments of most lung biopsy specimens, the ing that the analytical comparison data of SP384 versus sensitivity of some predictive IHC tests on small biopsy FISH released by the manufacturer achieves the best specimens has been challenged owing to heterogeneous balance between negative and positive agreement at the expression.29 Therefore, it is tempting to speculate that a 50% cutoff, a result similar to that of our ROC curve less heterogeneous pattern of staining is an advantage in analyses.35 Nonetheless, the interreader precision of this setting. Second, there is the constant staining of SP384 has been reported as high even with use of a nonneoplastic type II pneumocytes (>95%), which can lower cutoff (30%), so higher cutoffs should not be an be used as an in situ performance control. External interpretation challenge in the real clinical world.17 positive controls should not be used to rule out a false- Accordingly, a recent study has also reported a high negative result caused by suboptimal preanalytical pa- interpathologist agreement when interpreting both rameters.12 No matter how much one monitors this clones.23 In the light of the aforementioned, extreme phase of the procedure, samples will occasionally fail. caution is advised in settings with a very high incidence Along these lines, all but one of the IHC false-negative of patients with EGFR mutation (or other druggable non- samples in our series were precisely specimens that ROS1 genomic drivers, for that matter) so as not to are usually more prone to preanalytical artifacts: two render useless the screening value of ROS1 IHC (see later surgical resections, a decalcified bone specimen, and a in this discussion). 2130 Conde et al Journal of Thoracic Oncology Vol. 14 No. 12 Although break-apart FISH has traditionally been A review of published studies in the light of our the criterion standard test for detection of ROS1 rear- findings suggests that there are two scenarios that can rangements, the ROS1 FISH is especially difficult to have important clinical consequences when ROS1 IHC is interpret and may be prone to both false negatives and to be used as the primary screening method for ROS1 false positives.9,11,19,36-40 To increase the robustness of therapy: (1) a ROS1 FISH–false-positive result in a pa- the results, we decided to repeat all FISH tests in house tient with another druggable alteration that is causing and score them with an outstanding automated FISH the ROS1 IHC positivity (awareness of the FISH poten- scanning system using a high threshold for positivity. tial pitfalls is essential [i.e., percentage of positive The mean and median numbers of positive cells in pos- nuclei around the cutoff, 30 isolated pattern], and if the itive tumors were very high (>80%, which is well above result is inconsistent, it is sensible to use a third the threshold) and obviously contributed to the excellent methodology [i.e., NGS], which will potentially discover correlation with FISH, but it must be emphasized that the reason for the IHC positivity) and (2) a ROS1 NGS– some rare fusion partners ([GOPC], which is also known negative or failed report in a ROS1-rearranged sample as FIG, occurs in 3% of ROS1 patients and is not repre- that exhibited intense and homogeneous IHC stain- sented in the present study) are a well-known source of ing.38,44 The choice of RNA-based NGS can reduce the false-negative FISH results.14,39,41,42 Conversely, we and risk of false negatives, and using another sample or a others have reported that bona fide ROS1-nonrearranged third technology (i.e., FISH) when the initial NGS tumors can contain a number of positive nuclei (10%– approach fails is mandatory to confirm those positive 12%), which is close to the 15% cutoff used in many IHC results.39,47 studies.9,15,16 At least some published reports with a In conclusion, the new SP384 clone showed high high prevalence of concomitant oncogene mutations may sensitivity without compromising specificity, so it is reflect problems with the FISH interpretation.43,44 The another excellent analytical option for the proposed use of imaging systems and/or a higher threshold for CAP/IASLC/AMP molecular testing algorithm. Consider- positivity are strategies that should ensure ation of the clinical problem of NSCLC highlights the specificity.9,11,15,16 need to be aware of how the methods that we use In the last phase of the study, we performed an RNA- perform in the real-world setting.46 based NGS assay in FISH-positive cases to understand the molecular epidemiology of the different rearrangements and try to correlate them with the clinical and patho- Acknowledgments Funding for this study was provided by Instituto de logical features. It must be emphasized that this was not a formal comparison study between different methodolo- Salud Carlos III (ISCIII) (Fondos FEDER and Plan Estatal de IþDþI 2013-2016 [PI14-01176 and PI17-01001], gies. Overall, the variety and prevalence of ROS1 partners 2018-2021 [PI18/00382], and PT17/0015/0006]) and identified were similar to those described.24,37,45 The percentage of cases in which the suboptimal RNA quality the iLUNG Program (B2017/BMD-3884) from the Comunidad de Madrid. Ventana Medical Systems pro- and/or quantity resulted in low sequencing coverage vided the clone SP384 free of charge. Thermo Fisher highlights the need for an evidence-based algorithmic approach.39,46,47 The fusion partner can influence both Scientific provided the Oncomine Dx Target Test panel free of charge. Dr. F. Lopez-Rios thanks T. Crean for his the IHC staining and the FISH pattern, with the EZR constant support. variant usually being associated with a membranous accentuation and isolated 30 signals, respec- tively.13,14,25,45 This latter association could explain some Supplementary Data of the cases with false-negative FISH results that were Note: To access the supplementary material accompa- found to contain the EZR-ROS1 transcript, as this atypical nying this article, visit the online version of the Journal of pattern is in fact the most difficult to score because the Thoracic Oncology at www.jto.org and at https://doi. isolated 30 signals can sometimes be absent or barely org/10.1016/j.jtho.2019.07.005. visible.9,13,40 Finally, our nonsignificant trend of better PFS for patients with the EZR-ROS1 fusion might be in References alignment with the results of series in which almost every 1. U.S. Food and Drug Administration. FDA approves crizo- patient with an IGS achieved a complete response and tinib capsules. https://www.fda.gov/drugs/resources- information-approved-drugs/fda-approves-crizotinib- with the recently published differential efficacy of crizo- capsules. Accessed May 5, 2019. tinib in the non–CD74-ROS1 group.24,48 Unfortunately, 2. European Medicines Agency. Xalkori, INN-crizotinib. this is still a controversial topic that would need larger https://www.ema.europa.eu/en/documents/product- multicenter series with longer follow-up and standard- information/xalkori-epar-product-information_en.pdf. ized NGS to draw definitive conclusions.22,37 Accessed April 29, 2019. December 2019 ROS1 Immunohistochemistry with Clone SP384 2131 3. Shaw AT, Ou S-HI, Bang Y-J, et al. Crizotinib in ROS1- 19. Lin JJ, Ritterhouse LL, Ali SM, et al. ROS1 fusions rarely rearranged non–small-cell lung cancer. N Engl J Med. overlap with other oncogenic drivers in non-small cell 2014;371:1963–1971. lung cancer. J Thorac Oncol. 2017;12:872–877. 4. Mazières J, Zalcman G, Crinò L, et al. Crizotinib therapy 20. Wu Y-L, Yang JC-H, Kim D-W, et al. Phase II study of for advanced lung adenocarcinoma and a ROS1 rear- crizotinib in east asian patients with ROS1-positive rangement: results from the EUROS1 cohort. J Clin advanced non–small-cell lung cancer. J Clin Oncol. Oncol. 2015;33:992–999. 2018;36:1405–1411. 5. Shaw AT, Felip E, Bauer TM, et al. Lorlatinib in 21. Park S, Ahn B-C, Lim SW, et al. Characteristics and non-small-cell lung cancer with ALK or ROS1 rearrange- outcome of ROS1-positive non–small cell lung cancer ment: an international, multicentre, open-label, single- patients in routine clinical practice. J Thorac Oncol. arm first-in-man phase 1 trial. Lancet Oncol. 2018;13:1373–1382. 2017;18:1590–1599. 22. Shaw AT, Riely GJ, Bang Y-J, et al. Crizotinib in ROS1- 6. Drilon A, Siena S, Ou S-HI, et al. Safety and antitumor rearranged advanced non-small-cell lung cancer activity of the multitargeted pan-TRK, ROS1, and ALK (NSCLC): updated results, including overall survival, inhibitor entrectinib: combined results from two phase I from PROFILE 1001. Crizotinib in ROS1-rearranged trials (ALKA-372-001 and STARTRK-1). Cancer Discov. advanced non-small-cell lung cancer (NSCLC): updated 7. Lim SM, Kim HR, Lee J-S, et al. Open-label, multicenter, results, including overall survival, from PROFILE 1001 phase II study of ceritinib in patients with non-small-cell [e-pub ahead of print]. Ann Oncol. https://doi.org/10.1 lung cancer harboring ROS1 rearrangement. J Clin 093/annonc/mdz131/5448502. accessed May 6, 2019. Oncol. 2017;35:2613–2618. 23. Hofman V, Rouquette I, Long-Mira E, et al. Multicenter 8. Remon J, Ahn M-J, Girard N, et al. Advanced stage non- evaluation of a novel ROS1 immunohistochemistry assay small cell lung cancer: advances in thoracic oncology (SP384) for detection of ROS1 rearrangements in a large 2018. J Thorac Oncol. 2019;14:1134–1155. cohort of lung adenocarcinoma patients. J Thorac Oncol. 9. International Association for the Study of Lung Cancer. 2019;14:1204–1212. IASLC atlas of ALK and ROS1 testing in lung cancer. 24. Li Z, Shen L, Ding D, et al. Efficacy of crizotinib among https://www.iaslc.org/publications/iaslc-atlas-alk-and- different types of ROS1 fusion partners in patients with ros1-testing-lung-cancer. Accessed April 29, 2019. ROS1-rearranged non–small cell lung cancer. J Thorac 10. Lindeman NI, Cagle PT, Aisner DL, et al. Updated mo- Oncol. 2018;13:987–995. lecular testing guideline for the selection of lung cancer 25. Su Y, Goncalves T, Dias-Santagata D, Hoang MP. Immu- patients for treatment with targeted tyrosine kinase nohistochemical detection of ROS1 fusion. Am J Clin inhibitors. J Thorac Oncol. 2018;13:323–358. Pathol. 2016;147:77–82. 11. Bubendorf L, Büttner R, Al-Dayel F, et al. Testing for 26. Rossi G, Jocollé G, Conti A, et al. Detection of ROS1 ROS1 in non-small cell lung cancer: a review with rec- rearrangement in non-small cell lung cancer: current ommendations. Virchows Arch. 2016;469:489–503. and future perspectives. Lung Cancer (Auckl). 12. Conde E, Suárez-Gauthier A, Benito A, et al. Accurate 2017;8:45–55. identification of ALK positive lung carcinoma patients: 27. Viola P, Maurya M, Croud J, et al. A validation study for novel FDA-cleared automated fluorescence in situ hy- the use of ROS1 immunohistochemical staining in bridization scanning system and ultrasensitive immuno- screening for ROS1 translocations in lung cancer. histochemistry. PLoS One. 2014;9:e107200. J Thorac Oncol. 2016;11:1029–1039. 13. Boyle TA, Masago K, Ellison KE, Yatabe Y, Hirsch FR. ROS1 28. Rogers T-M, Russell PA, Wright G, et al. Comparison immunohistochemistry among major genotypes of non– of methods in the detection of ALK and ROS1 rear- small-cell lung cancer. Clin Lung Cancer. 2015;16:106– rangements in lung cancer. J Thorac Oncol. 2015;10:611– 111. 618. 14. Yoshida A, Tsuta K, Wakai S, et al. Immunohistochemical 29. Gniadek TJ, Li QK, Tully E, Chatterjee S, Nimmagadda S, detection of ROS1 is useful for identifying ROS1 rear- Gabrielson E. Heterogeneous expression of PD-L1 in rangements in lung cancers. Mod Pathol. 2014;27:711– pulmonary squamous cell carcinoma and adenocarci- 720. noma: implications for assessment by small biopsy. Mod 15. Sholl LM, Sun H, Butaney M, et al. ROS1 immunohisto- Pathol. 2017;30:530–538. chemistry for detection of ROS1-rearranged lung 30. Zhao J, Chen X, Zheng J, Kong M, Wang B, Ding W. adenocarcinomas. Am J Surg Pathol. 2013;37:1441– A genomic and clinicopathological study of non-small- 1449. cell lung cancers with discordant ROS1 gene status by 16. Mescam-Mancini L, Lantuéjoul S, Moro-Sibilot D, et al. fluorescence in-situ hybridisation and immunohisto- On the relevance of a testing algorithm for the detection chemical analysis. Histopathology. 2018;73:19–28. of ROS1-rearranged lung adenocarcinomas. Lung Cancer. 31. Sabour S. Reliability assurance of EML4-ALK rearrange- 2014;83:168–173. ment detection in non-small cell lung cancer: a meth- 17. Hanlon Newell A, Liu W, Bubendorf L, et al. MA26.07 odological and statistical issue. J Thorac Oncol. ROS1 (SP384) immunohistochemistry inter-reader preci- 2016;11:e92–e93. sion between 12 pathologists. J Thorac Oncol. 32. Mahe E. Comment on “Testing for ALK rearrangement in 2018;13:S452–S453. lung adenocarcinoma: a multicenter comparison of 18. Lin JJ, Shaw AT. Recent advances in targeting ROS1 in immunohistochemistry and fluorescent in situ hybridi- lung cancer. J Thorac Oncol. 2017;12:1611–1625. zation” Mod Pathol. 2014;27:1423–1424. 2132 Conde et al Journal of Thoracic Oncology Vol. 14 No. 12 33. Warth A, Muley T, Dienemann H, et al. ROS1 expression 41. Gainor JF, Shaw AT. Novel targets in non-small cell lung and translocations in non-small-cell lung cancer: clini- cancer: ROS1 and RET fusions. Oncologist. 2013;18: copathological analysis of 1478 cases. Histopathology. 865–875. 2014;65:187–194. 42. Suehara Y, Arcila M, Wang L, et al. Identification of 34. Selinger CI, Li BT, Pavlakis N, et al. Screening for ROS1 KIF5B-RET and GOPC-ROS1 fusions in lung adenocarci- gene rearrangements in non-small-cell lung cancers us- nomas through a comprehensive mRNA-based screen for ing immunohistochemistry with FISH confirmation is an tyrosine kinase fusions. Clin Cancer Res. 2012;18:6599– effective method to identify this rare target. Histopa- 6608. thology. 2017;70:402–411. 43. Wiesweg M, Eberhardt WEE, Reis H, et al. High 35. Huang R, Smith D, Richardson B, et al. P2.09-13 corre- prevalence of concomitant oncogene mutations in lation of ROS1 (SP384) immunohistochemistry with ROS1 prospectively identified patients with ROS1- rearrangement determined by fluorescence in situ hy- positive metastatic lung cancer. J Thorac Oncol. bridization. J Thorac Oncol. 2018;13:S766. 2017;12:54–64. 36. Shan L, Lian F, Guo L, et al. Detection of ROS1 gene 44. Savic S, Rothschild S, Bubendorf L. Lonely driver ROS1. rearrangement in lung adenocarcinoma: comparison of J Thorac Oncol. 2017;12:776–777. IHC, FISH and real-time RT-PCR. PLoS One. 2015;10: 45. Yoshida A, Kohno T, Tsuta K, et al. ROS1-rearranged e0120422. lung cancer: a clinicopathologic and molecular study 37. Michels S, Massutí B, Schildhaus H-U, et al. Safety and of 15 surgical cases. Am J Surg Pathol. 2013;37:554– efficacy of crizotinib in patients with advanced or met- 562. astatic ROS1-rearranged lung cancer (EUCROSS): a Eu- 46. Sholl LM. Recognizing the challenges of oncogene fusion ropean phase 2 clinical trial. J Thorac Oncol. 2019;14: detection: a critical step toward optimal selection of 1266–1276. lung cancer patients for targeted therapies. J Thorac 38. Clavé S, Rodon N, Pijuan L, et al. Next-generation Oncol. 2018;13:1433–1435. sequencing for ALK and ROS1 rearrangement detection 47. Benayed R, Offin M, Mullaney K, et al. High yield of in patients with non–small-cell lung cancer: implications RNA sequencing for targetable kinase fusions in lung of FISH-positive patterns. Clin Lung Cancer. 2019;20: adenocarcinomas with no driver alteration detected by e421–e429. DNA sequencing and low tumor mutation burden. Clin 39. Davies KD, Le AT, Sheren J, et al. Comparison of mo- Cancer Res. 2019;25:4712–4722. lecular testing modalities for detection of ROS1 rear- 48. Dugay F, Llamas-Gutierrez F, Gournay M, et al. Clinico- rangements in a cohort of positive patient samples. pathological characteristics of ROS1- and RET- J Thorac Oncol. 2018;13:1474–1482. rearranged NSCLC in Caucasian patients: data from a 40. Kerr KM, López-Ríos F. Precision medicine in NSCLC and cohort of 713 non-squamous NSCLC lacking KRAS/EGFR/ pathology: how does ALK fit in the pathway? Ann Oncol. HER2/BRAF/PIK3CA/ALK alterations. Oncotarget. 2016;27(suppl 3):iii16–iii24. 2017;8:53336–53351.