Abstract 4727: Molecular heterogeneity in glioblastoma multiforme influences variant clonality and neoantigen prediction accuracy

Abstract

Abstract Gliomas originate from neuroglial progenitor cells in the brain or spinal cord and have an incidence of ~6 cases per 100,000 individuals each year in the United States. Approximately half of glioma cases present as an aggressive type of astrocytoma known as glioblastoma multiforme (GBM). Due to its location and diffuse morphology, GBM has only a 30% two-year survival rate and additional treatment strategies are needed. Clinical trials for GBM personalized neoantigen vaccines have recently reported neoantigen-specific T cell responses, but treatment efficacy remains low. We posit that this may be due to GBM's high molecular heterogeneity, which could complicate the identification of clonal neoantigens for vaccine design. To address this topic, we performed whole-exome sequencing (WES) and RNA-seq on 30 human brain tumors. This cohort contains 21 primary/recurrent GBMs and 19 brain metastases derived from breast cancer, non-small cell lung cancer, or melanoma. We integrated variant calls, RNA expression, and HLA typing data for input into pVACtools, a suite of tools for neoantigen prediction, to obtain candidate neoantigen sequences with IC50 < 500nM. Mutation and neoantigen burden varied widely between the two tumor types. The range of variant counts for GBM samples is 7-169 (median 81), while the brain metastases have a range of 158-1317 (median 274.5). In addition, the former's class I and II neoantigen counts range from 1-312 (median 14), while the latter's range from 3-1159 (median 23.5). The percentage of variants that created neoantigens per sample ranged from 0.0-54.0% (median 7.1%). These counts exclude a recurrent GBM with two distinct hypermutated subclones that contain missense mutations in MSH6, a canonical DNA damage repair gene. By comparing intratumoral heterogeneity levels, we found that GBM patients had a lower proportion of clonal variants and neoantigens than tumors that had metastasized to the brain. In addition, the proportion of variants and neoantigens missed by single sector sequencing was significantly higher in GBM compared to brain metastases, indicating substantial heterogeneity in this tumor type. Variation in immune cell infiltrate was assessed using transcriptome data, and while many tumors demonstrated cellular heterogeneity, there was no clear correlation between this metric and tumor type. We are performing TCR sequencing to further investigate the immune landscape heterogeneity by determining TCR clonality diversity per region. Overall, this analysis demonstrates that GBM contains cellular and molecular heterogeneity, which can complicate clonal neoantigen selection. Multi-sample sequencing as a standard for GBM genomic analysis could alleviate these inconsistencies to improve neoantigen vaccine design and increase treatment efficacy. Citation Format: Megan M. Richters, Maximilian Schaettler, Zach L. Skidmore, Cody A. Ramirez, Matthew C. Mosior, Obi L. Griffith, Malachi Griffith, Gavin P. Dunn. Molecular heterogeneity in glioblastoma multiforme influences variant clonality and neoantigen prediction accuracy [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4727.