Improving the diagnosis of leptomeningeal metastases by molecular profiling of cell-free DNA from cerebrospinal fluid

Background: The limited sensitivity of cerebrospinal fluid (CSF) cytology leads to delayed or missed diagnoses of leptomeningeal metastases (LM). This study explores the diagnostic potential of next generation sequencing (NGS) in diagnosing LM by analyzing primary tumor mutations in CSF. Methods: CSF-cell-DNA, CSF-cell-free-DNA (cfDNA) and plasma-cfDNA from ten patients with cytological confirmed LM were used to examine the best DNA source for analysis. Next, CSF samples from 85 patients with a known primary tumor mutation and a suspicion on LM were examined to compare diagnostic sensitivities of cytology, circulating tumor cells (CTC) and NGS using CSF-cfDNA (cfDNA-NGS). Results: While mutation detection in CSF-cell-DNA depended on the tumor cell percentage (TCP) in CSF, mutation detection in CSF-cfDNA was independent of TCP. Variant allele frequencies (VAF) of the primary tumor mutations were significantly higher in CSF-cfDNA as compared to CSF-cell-DNA suggesting CSF-cfDNA was predominantly tumor-derived. Plasma-cfDNA did not contribute to the high tumor fraction in CSF-cfDNA. Analyzing CSF from 85 patients resulted in a sensitivity of 82 %, 93 % and 100 %, for cytology, CTC-analysis and cfDNA-NGS respectively. For one LM-negative patient (4 %), primary tumor mutations were detected in CSF-cfDNA, possibly due to an active brain metastasis (BM) bordering the CSF. However, six other LM-negative patients with active BM showed no mutations in CSF-cfDNA. Conclusions: Molecular profiling of CSF-cfDNA is a powerful tool to diagnose LM from tumors harboring a known driver mutation, exceeding the sensitivity of cytology and CTC-analysis. Caution is warranted regarding patients diagnosed with active BM bordering the CSF.