Identifying adverse molecular features of HPV+ head and neck cancers using patient-derived models.

Abstract

6057 Background: Advancing therapy for human papilloma virus-related (HPV+) head and neck squamous cell carcinoma (HNSCC) is hindered by inadequate preclinical models and risk stratification tools. This study addresses both barriers by characterizing a panel of patient-derived xenografts (PDXs) that includes nine new models of HPV+ disease. Methods: Exome-sequenced genetic features of the PDXs were compared to their growth properties and the outcomes of their patients of origin. Genetic traits with potential prognostic utility were validated in multiple retrospective patient cohorts. Results: The HPV+ PDXs avoided known artifacts in HPV+ cell lines, including 3q amplifications and loss of PIK3CA mutants, while enriching for alterations in H3K4 methyltransferases and Notch pathway genes. A positive association emerged between PDX tumor mutational burden (TMB) and their growth efficiency both in vivo and as organoids. This finding led to identification in The Cancer Genome Atlas (TCGA) of an association between high TMB and worse survival of early-stage HPV-negative cancers but not HPV+ ones. Insight into aggressive HPV+ disease came from a PDX established from a patient before lethal relapse. The reduced levels of viral E7 and p16INK4A present in this model were also detected in early lethal HPV+ cases in TCGA as well as the recurrences in a second HPV+ HNSCC cohort (JHU). This observation suggested a diminished contribution of viral oncogenes to cell cycle dysregulation in aggressive cases. To evaluate this possibility, hierarchical clustering of both cohorts was performed based on expression of E2F target genes. This analysis discovered a distinct cell cycle-related transcriptional pattern in the clusters of cases containing the recurrences and early lethal events. Furthermore, a subset of these transcripts proved to be stage-independent predictors of survival for the HPV+ HNSCCs in both TCGA and JHU cohorts. Conclusions: Characterizing the most HPV+ patient-derived models described to date revealed novel prognostic utility for E2F target expression in HPV+ HNSCCs and TMB in HPV-negative HNSCCs. These features have potential for application to risk stratification, biomarker development, and trial design.