CD74/SLC34A2-ROS1 fusion variants involving the transmembrane region predict poor response to crizotinib in non-small cell lung cancer independent of TP53 mutations

Variable partners and breakpoints have been reported in patients with ROS1-rearranged NSCLC. Here, we investigated the association of fusion partners and breakpoints with crizotinib efficacy in NSCLCs with common ROS1 fusions.DNA and RNA next-generation sequencing (NGS) and immunohistochemistry were performed to characterize ROS1 fusions.Using DNA NGS, we identified ROS1 fusions in 210 cases, comprising 171 common (CD74/EZR/TPM3/SDC4/SLC34A2-ROS1) and 39 uncommon (variants identified in <5%) ROS1 fusion cases. DNA NGS detected variable ROS1 genomic breakpoints in common ROS1 fusions, whereas RNA NGS found ROS1 breakpoints mainly occurring in exons 32, 34 and 35, resulting in long (exon 32) and short (exon 34 or 35) ROS1 fusions. ROS1 immunohistochemistry revealed that membranous and cytoplasmic staining was predominant in long ROS1 fusions, whereas cytoplasmic staining was predominant in short ROS1 fusions (p = 0.006). For patients who received first-line crizotinib, median progression-free survival (mPFS) was lower in patients with long ROS1 fusions than those with short ROS1 fusions (8.0 versus 24.0 mo, p = 0.006). Moreover, mPFS for patients with and without TP53 mutations was 8.0 and 19.0 months, respectively (p = 0.159); mPFS for patients with and without BIM deletion polymorphism was 5.0 and 22.0 months, respectively (p = 0.003). When analyzing together with fusion partners, patients with long CD74/SLC34A2-ROS1 fusions were found to have shorter PFS than those with other ROS1, regardless of the presence or absence of TP53 mutations (p < 0.001 and p = 0.002, respectively).Long CD74/SLC34A2-ROS1 fusions, which retain transmembrane regions in ROS1 and fusion partners, are associated with poor response to crizotinib independent of TP53 mutations.