Diagnosis and genetic characterization of CNS metastases in lung cancer via shallow whole-genome sequencing of CSF cell-free DNA.

e14013 Background: The genetic characterization of CNS metastases in lung cancer is hindered by the limited sampling opportunities and genetic heterogeneity between intracranial and extracranial lesions. To address these challenges, this study uses cerebrospinal fluid (CSF) and blood samples, to detect and characterize CNS metastasis variants through shallow whole-genome sequencing (sWGS). Methods: From Dec. 2021 to Aug. 2023, 57 lung cancer patients were recruited, comprising 29 with leptomeningeal metastases (LMs) and 28 with only parenchymal brain metastases (PMs). A total of 66 blood and 112 CSF samples were collected for cfDNA sWGS to analyze chromosomal arm CNVs, focal amplifications (CN > 5) and SNV/InDels. Cytology was also conducted on 95 of these CSF samples. Variant differences between PMs and extracranial metastases (EMs) were analyzed using the MSK-MET database. Results: CSF cfDNA concentration was significantly lower than that in plasma (p < 0.0001), with 41.7% of samples unquantifiable, yet all samples successfully completed sWGS. The cfDNA tumor fraction in CSF was significantly higher than in paired plasma (p < 0.0001), and most arm CNVs and gene amplifications exclusively detected in CSF. Cytology, as the gold standard for LM diagnosis, had a sensitivity of 0.76. The sensitivity for LM diagnosis based on aneuploidy and genomic amplification fraction (GAF) were 0.51 and 0.80, respectively, without compromising specificity. Gains in 1p, 7p, 11q, 16p, and 17q, as well as losses in 9p and 9q, were significantly more frequent in PMs compared to EMs and further increased in the LMs (p < 0.05). Notably, losses in 1p, 4p, 5q, and 11p, significantly more frequent in PMs than in extracranial metastases, but showed a significant increase in gain frequency in LMs (p < 0.05). In the LM group, the frequency of EGFR L858R and amplification of nearly the entire 7p where the EGFR located, were significantly increased, whereas KRAS G12D frequency significantly decreased (p < 0.05). Conclusions: The characteristics of extremely low concentration but high tumor fraction of CSF cfDNA make it suitable for sWGS, enhancing LM diagnostic sensitivity without compromising specificity, and identifying molecular characteristics of CNS metastasis. These molecular features require further validation in larger prospective clinical cohorts, offering potential for personalized diagnosis and treatment of CNS metastases.