BIOM-77. MUTATION PROFILE CONCORDANCE BETWEEN TARGETED NEXT-GENERATION SEQUENCING OF PEDIATRIC CENTRAL NERVOUS SYSTEM TUMORS AND LONGITUDINAL PLASMA AND CEREBROSPINAL CIRCULATING TUMOR DNA SAMPLES

Abstract Accurate detection and quantification of circulating tumor DNA (ctDNA) in the plasma and cerebrospinal fluid (CSF) of children with central nervous system (CNS) tumors provides critical clinical information. Utilization of targeted next-generation sequencing to identify mutational profiles across tumor, plasma, and CSF samples increases confidence in successful tumor-derived ctDNA detection. A total of 223 plasma and 47 CSF longitudinal samples were collected from 49 pediatric CNS tumor patients. Cell-free DNA (cfDNA) extraction was performed on samples from a cohort of 16 patients followed by targeted next-generation sequencing. Tumor types include diffuse midline glioma (n=6, 37%), high-grade glioma (n=5, 31%), medulloblastoma (n=4, 25%), and ependymoma (n=1, 6%). CSF collection occurred by lumbar puncture (n=7, 47%), during surgical resection (n=4, 27%), or via ommaya catheter (n=2, 13%) or external ventricular drain (n=2, 13%). Ten (66%) CSF samples and 51 (76%) plasma samples were collected from patients with radiographic evidence of disease. Cell-free DNA (cfDNA) was successfully extracted in 67/67 (100%) plasma samples (average yield 11.7 ng) and 10/15 (66%) CSF samples (average yield 62.9ng). Targeted sequencing identified at least one non-intronic variant in thirteen (81%) of the tumors with 31 distinct mutations across 21 genes. The most frequently mutated genes were H3-3A (n=6), TP53 (n=6), ATRX (n=4), PIK3CA (n=3), BRAF (n=2) and NOTCH1 (n=2). An average of 2.9 mutations were detected per tumor with a mean allele frequency of 48% (SD 26%). Mutations of varying allele frequencies were found in four (27%) CSF and 19 (28%) plasma longitudinal samples, some of which correlated clinically to disease status. We conclude that cfDNA can be reliably extracted from plasma and CSF samples. Targeted sequencing can identify tumor-specific mutations in these longitudinal samples and serve as a surrogate biomarker to monitor the patient’s clinical course. Further refinement of this methodology may expand clinical applications.