GCT-02. Classification of intracranial germ cell tumor using genomic and epigenomic analysis of cerebrospinal fluid

Background Currently, there is no clinical method for molecular classification of intracranial germ cell tumors (IGCT), where proper diagnosis is essential to differentiate the prognostically-distinct germinoma and non-germinoma (NGGCT) subtypes. DNA methylation profiling is the gold standard for molecular classification of other brain tumors, and prior studies of IGCT tissue have found distinct differences in global methylation between germinoma and NGGCTs. However, current array-based methods are reliant on large amounts of input DNA, necessitating extraction from tissue and limiting their utility for IGCT classification, where access to the tumor tissue is limited. Cerebrospinal fluid (CSF), containing cell-free tumor DNA (cfDNA), presents an alternative source of DNA for methylation-based IGCT diagnosis and disease monitoring. To overcome the limitation of low input DNA, novel enzymatic DNA methylation sequencing (EM-seq) methods have been developed to accurately quantify methylation from 1ng of DNA. Methods Using the NEBNext EM-seq kit on cfDNA (1-10ng) from archival CSF samples, we performed methylation sequencing on 14 IGCT cases. Data was processed using the methylseq pipeline for quantification of site-specific methylation. For tumor classification, we utilized MNP-Flex, a platform-­agnostic version of the Heidelberg methylation classifier. Results We successfully generated genome-wide methylation profiles using the EM-seq method. Using the MNP-Flex classification method, originally trained using array data from tissue, we were able to successfully classify germinoma samples with high confidence (score > 0.9). NGGCT subtypes largely were not classifiable, likely due to differences between tumor DNA and the cfDNA found in the CSF, as well as missing NGGCT subtypes in the training data. Global methylation differences in cfDNA alone could be used to easily separate germinoma from NGGCT. Conclusions Despite the limitations of the MNP-Flex classifier for use with cfDNA, we classified germinoma with high confidence in this pilot study. Additional work is being performed to further separate NGGCT subtypes.