PATH-11. Detection of genetic and epigenetic alterations in Liquid Biopsies from pediatric brain tumor patients

Abstract

BACKGROUND: Liquid biopsies (LBs) hold great promise as a non-invasive method for tumor detection and disease monitoring. Their application in pediatric neuro-oncology patients has been challenging due to the blood-brain barrier limiting the amount of circulating tumor DNA (ctDNA) and low volumes of cerebrospinal fluid (CSF) and serum samples. Here, we assessed the feasibility of LBs in a pediatric brain tumor cohort (n=12 patients) by adapting pre-analytical protocols to low cell-free DNA (cfDNA) input and bioinformatic pipelines for genome-wide analyses. METHODS: Genomic DNA isolated from tumor tissues, including medulloblastoma (n=6), ependymoma (n=4), low-grade glioma (n=1), and choroid plexus papilloma (n=1), and cfDNA isolated from matched CSF and serum samples were subjected to low-coverage whole genome sequencing (lcWGS) and the EPIC DNA methylation array. For methylation analyses, samples were prepared using both traditional bisulfite- and recently developed enzymatic-based methylation conversion. Molecular tumor (sub)type predictions were performed using the Heidelberg Brain Tumor Classifier. RESULTS: lcWGS revealed tumor-specific copy number variations (CNVs) in both CSF and serum samples. The detection rate of tumor-specific CNVs (median 92% vs. 23%) and the ctDNA fraction (median 31% vs. 1.5%) were higher in CSF vs. serum samples. Bisulfite-converted as well as enzymatically converted tumor DNA and cfDNA samples could be readily analyzed using the Heidelberg Classifier. In comparison to bisulfite-based conversion, enzymatic-based conversion improved the quality of methylation analyses for minimal input cfDNA samples. CONCLUSIONS: Our results highlight the feasibility of liquid biopsies from both CSF and serum in brain tumor patients. In contrast to CNV-based tumor detection from serum samples, ctDNA analysis from CSF allowed a more comprehensive genetic and epigenetic profiling which is expected to assist in less-invasive tumor classification, identification of drug targets, or the investigation of tumor evolution. Altogether, this study provides the rationale for further implementation of LBs in pediatric neuro-oncology.