Introduction of E-Score: A gene expression score to characterize immunogenic versus oncogenic HPV expression in HNSCC.

Abstract

6057 Background: Human Papillomavirus (HPV) integration in the human genome occurs in a majority of HPV(+) head and neck squamous cell carcinomas (HNSCC), with direct and indirect effects on host gene expression and tumorigenesis. Although previous studies showed that HPV integration in HNSCC is overrepresented in genes often somatically mutated in head and neck, lung, and urogenital cancers, reports are mixed for whether patients with HPV integration have worse clinical outcomes. These disparate results may be due to differences in how integration is defined. Integration can often lead to partial loss of the viral genome (e.g. HPV genes E1, E2, or E5) with retention of the oncogenes E6 and E7, which led to a low E2/E7 expression ratio being used as a biomarker for integration. Conversely, others have used the direct detection of DNA or RNA HPV-human breakpoints to identify tumors harboring integrated HPV. HPV integration and the retention or deletion of HPV genes can dramatically change the expression of HPV genes, and we hypothesize it also affects the host immune response. Hypothesis: Loss of non-oncogenic HPV genes relative to E6 and E7 levels reduces immunogenicity, leading to weakened immune cell infiltration and worse clinical outcomes. Methods: Bulk RNA sequencing data from HPV(+) HNSCC tumors (n = 136) was used to characterize the differences in retention or loss of HPV genes and calculate the E-Score as the sum of the logCPM of HPV genes E1, E2, E4, and E5 relative to E7. Cell type deconvolution was performed using ComBat to correct for batch effects, and CIBERSORTx to quantify cell type proportions within the tumor samples. Kaplan-Meier estimates of recurrence and overall survival (OS) were calculated, and Cox proportional hazards tests with covariates were used to test for association with recurrence and OS. To test the correlation between immune cell type inferred proportions and E-Score, T-Tests were performed between samples with high and low E-scores. Results: Low E-Score was significantly associated with recurrence (p = 0.012) but not with OS (p = 0.39) in the multivariate model accounting for age, tumor stage (AJCC8), smoking status, drinking status, BMI, marital status, and molecular tumor subtype (defined as IMU or KRT). For integration positive samples (n = 66), higher E-Score, suggesting lower rates of HPV gene loss, was correlated with higher percentages of dendritic cells (p = 0.044), macrophages (p = 0.021), and CD8+ T Cells (p = 0.024), but not CD4+ T cells (p = 0.14). This significance did not hold when looking at integration negative samples. Conclusions: These results suggest that HPV+ HNSCC patients with a loss of E1, E2, E4 and E5 tend to have less immune cell infiltration and are more likely to have recurrence. We conclude that E-Score is a potential prognostic biomarker of recurrence in HPV+ HNSCC and potential predictive biomarker for immunotherapy.