Genomic determinants of response to pembrolizumab in head and neck squamous cell carcinoma (HNSCC).

Abstract

6009 Background: Somatic mutational load (ML) is associated with response to anti CTLA-4 and PD-1/-L1 immunotherapies in select tumors, due to formation of neoepitopes not subject to central immune tolerance. Neoepitopes specific to HPV, EBV virus infection are also present in some HNSCC. An IFNγ gene expression profile (GEP) characteristic of tumor inflammation is also related to response to anti PD-1/-L1 therapy. This study evaluated relationships between ML and clinical outcome and independent predictive values of ML and GEP in patients with HNSCC treated with pembrolizumab. Methods: Whole exome sequencing (WES) and GEP were assessed in FFPE tumor specimens of patients with HNSCC (KEYNOTE 012; subsets of B1 [PD-L1+, n = 34] and B2 [PD-L1+/‒, n = 73] cohorts). ML, neoantigen load (NL), HPV/EBV status and clonality were assessed by standard WES analytical methods. GEP score is a weighted sum of normalized expression values of 18 genes. Statistical testing of ML and response, and ML and GEP relationship by HPV/EBV status was prespecified. Results: There were 73 patients identified as HPV‒ and EBV‒ (n = 25 in B1; n = 48 in B2). In HPV‒ and EBV‒ patients in B1 and B2 cohorts, respectively, associations between ML and objective response (OR) (P = 0.029 and 0.055; AUROC 0.89 and 0.63), and GEP and OR (P = 0.064 and 0.01; AUROC 0.82 and 0.74) were statistically significant. In combined cohorts of HPV‒ and EBV‒ patients, ML and GEP were significantly associated with OR (P = 0.009 and 0.002; AUROC 0.70 and 0.76, respectively). ML and GEP were only weakly correlated (r = 0.173). In a joint model, ML was significantly associated with response (p = 0.020) after adjusting for GEP (also significant, p = 0.006). NL and clonality weighted ML were also significantly associated with response (P = 0.026 and 0.006, respectively). In HPV+ or EBV+subjects, OR association was not significant for ML, possibly due to a dominance of viral vs somatic neoepitopes; GEP was significant, likely due to tumor inflammation. Conclusions: ML and GEP are independently predictive of response to pembrolizumab in HPV‒/EBV‒ patients with HNSCC; GEP was predictive regardless of viral status. ML and GEP may have utility in characterizing responses to anti PD-1 therapies and novel cancer regimens in HNSCC. Clinical trial information: NCT01848834.