Comparison of ROS1-rearrangement detection methods in a cohort of surgically resected non-small cell lung carcinomas

Background: Patients with non-small cell lung cancer (NSCLC) harboring a ROS proto-oncogene 1 (ROS1)-rearrangement respond to treatment with ROS1 inhibitors. To distinguish these rare cases, screening with immunohistochemistry (IHC) for ROS1 protein expression has been suggested. However, the reliability of such an assay and the comparability of the antibody clones has been debated. Therefore we evaluated the diagnostic performance of current detection strategies for ROS1-rearrangement in two NSCLC-patient cohorts. Methods: Resected tissue samples, retrospectively collected from consecutive NSCLC-patients surgically treated at Uppsala University Hospital were incorporated into tissue microarrays [all n=676, adenocarcinomas (AC) n=401, squamous cell carcinomas (SCC) n=213, other NSCLC n=62]. ROS1-rearrangements were detected using fluorescence in situ hybridization (FISH) (Abbott Molecular; ZytoVision). In parallel, ROS1 protein expression was detected using IHC with three antibody clones (D4D6, SP384, EPMGHR2) and accuracy, sensitivity, and specificity were determined. Gene expression microarray data (Affymetrix) and RNA-sequencing data were available for a subset of patients. NanoString analyses were performed for samples with positive or ambiguous results (n=21). Results: Using FISH, 2/630 (0.3% all NSCLC; 0.5% non-squamous NSCLC) cases were positive for ROS1 fusion. Additionally, nine cases demonstrated ambiguous FISH results. Using IHC, ROS1 protein expression was detected in 24/665 (3.6% all NSCLC; 5.1% non-squamous NSCLC) cases with clone D4D6, in 18/639 (2.8% all NSCLC; 3.9% non-squamous NSCLC) cases with clone SP384, and in 1/593 (0.2% all NSCLC; 0.3% non-squamous NSCLC) case with clone EPMGHR2. Elevated RNA-levels were seen in 19/369 (5.1%) cases (Affymetrix and RNA-sequencing combined). The overlap of positive results between the assays was poor. Only one of the FISH-positive cases was positive with all antibodies and demonstrated high RNA-expression. This rearrangement was confirmed in the NanoString-assay and also in the RNA-sequencing data. Other cases with high protein/RNA-expression or ambiguous FISH were negative in the NanoString-assay. Conclusions: The occurrence of ROS1 fusions is low in our cohorts. The IHC assays detected the fusions, but the accuracy varied depending on the clone. The presumably false-positive and uncertain FISH results questions this method for detection of ROS1-rearrangements. Thus, when IHC is used for screening, transcript-based assays are preferable for validation in clinical diagnostics