Differences in FGFR2 fusion detection between DNA and RNA NGS in a large Chinese lung cancer cohort

Background: Fibroblast growth factor receptor 2 (FGFR2) fusions are rare driver gene alterations in non-small cell lung cancer (NSCLC), with an overall incidence of approximately 0.02%. FGFR2 fusion breakpoints typically occur within exons 17-19, and numerous fusion partner genes have been identified. While next-generation sequencing (NGS) is the primary method for fusion detection, the differences in the detection spectrum between DNA- and RNA-based NGS for FGFR2 fusions remain inadequately characterized in large-scale cohorts. Methods: This retrospective study analyzed a large Chinese lung cancer cohort comprising 12,974 tumor samples. Among these, 3,578 samples were tested by DNA-NGS for fusion detection, and 9,396 samples were tested using RNA-NGS. Only FGFR2 fusions classified as pathogenic or likely pathogenic were included in the final analysis. Results: A total of 18 FGFR2 fusion events were identified, yielding an overall detection rate of 0.14%. The patient cohort was predominantly male (72.22%), with a median age of 65 years (range: 52-78). Histologically, most cases were diagnosed as lung adenocarcinoma (44.44%) and lung squamous cell carcinoma (38.89%), with 3 cases of undetermined pathology. All FGFR2 fusions were characterized by 5'-3' rearrangements. DNA-NGS identified 5 FGFR2 fusions (0.14%), all in lung adenocarcinoma, with heterogenous fusion partners including FGFR2-INPP5F, FGFR2-TACC2 (non-canonical breakpoints: exon 6), FGFR2-PCAT2, FGFR2-KCNH5, and one complex fusion involving FGFR2-NME9 and NME9-FGFR2. Meanwhile, RNA-NGS detected 13 FGFR2 fusions (0.14%), with a partner distribution distinct from DNA-NGS: 11 cases of FGFR2-ATE1 (84.62%) involving non-canonical breakpoints (exon2/5/6/9), as well as FGFR2-BICC1 and FGFR2-TACC2. All FGFR2-ATE1 fusion patients were male, and 7 of the 8 cases with specified pathology were lung squamous cell carcinomas, indicating a potential enrichment. Importantly, the FGFR2-ATE1 fusion lacked the FGFR2 kinase domain, and the vast majority (90.91%) had supporting reads <100, suggesting low expression levels. These features imply that FGFR2-ATE1 may exert oncogenic effects via kinase-independent mechanisms, and such fusions are unlikely to respond well to FGFR kinase inhibitors. The pathogenicity of atypical, low-expressed fusion transcripts requires cautious evaluation and reclassification. Conclusions: This study reveals distinct FGFR2 fusion detection patterns between DNA and RNA NGS. When interpreting complex or atypical fusions for targeted therapy guidance, integrating results from both DNA and RNA NGS is recommended to optimize patient stratification and treatment decisions.