Genomic alterations of cerebrospinal fluid cell-free DNA in leptomeningeal metastases of gastric cancer

Leptomeningeal metastases (LM) were rare in gastric cancer (GC), and GC patients with LM (GCLM) generally suffer from poor prognosis. Nevertheless, the clinical utility of cerebrospinal fluid (CSF) circulating tumor DNA (ctDNA) was underinvestigated in GCLM.

CSF ctDNA could more sensitively detect molecular markers and metastasis-related mechanisms than tumor tissues in GCLM patients, and our study sheds light on utilizing CSF ctDNA in prognostic estimation and clinical assessment in GCLM.

We retrospectively studied 15 GCLM patients, and all patients had paired primary tumor tissue samples and post-LM CSF samples while 5 patients also had post-LM plasma samples. All samples were analyzed using next-generation sequencing (NGS), and the molecular and clinical features were correlated with clinical outcomes.

CSF had higher mutation allele frequency (P = 0.015), more somatic mutations (P = 0.032), and more copy-number variations (P < 0.001) than tumor or plasma samples. Multiple genetic alterations and aberrant signal pathways were enriched in post-LM CSF, including CCNE1 amplification and cell cycle-related genes, and CCNE1 amplification was significantly associated with patients' overall survival (P = 0.0062). More potential LM progression-related markers were detected in CSF samples than in tumor samples, including PREX2 mutation (P = 0.014), IGF1R mutation (P = 0.034), AR mutation (P = 0.038), SMARCB1 deletion (P < 0.001), SMAD4 deletion (P = 0.0034), and TGF-beta pathway aberration (P = 0.0038). Additionally, improvement in intracranial pressure (P < 0.001), improvement in CSF cytology (P = 0.0038), and relatively low levels of CSF ctDNA (P = 0.0098) were significantly associated with better PFS. Lastly, we reported a GCLM case whose CSF ctDNA dynamic changes were well correlated with his clinical assessment. Conclusions: CSF ctDNA could more sensitively detect molecular markers and metastasis-related mechanisms than tumor tissues in GCLM patients, and our study sheds light on utilizing CSF ctDNA in prognostic estimation and clinical assessment in GCLM.