Development and validation of a combined metabolic and immune prognostic classifier for head and neck cancer

Background: Genomic characterisation of head and neck cancer (HNC) has identified 3-5 subgroups with distinct biological properties, including metabolic profile and immune status. Both facets could impact on response to standard and novel targeted therapies for HNC, but are not currently considered for treatment selection due to lack of validated biomarkers. Methods: A 54-gene metabolic-immune signature (MIGS) was constructed. Gene expression was analysed in silico using the TCGA HNC dataset (whole transcriptome RNA-Seq, n = 275) and validated using two independent cohorts (Chicago microarray [Agilent], n = 130; Birmingham targeted RNA-Seq [Illumina] on FFPE tissue, n = 123). We then evaluated MIGS in a cohort of anti-PD-1 treated R/M HNC patients. Immunohistochemistry (IHC) was used to investigate the utility of a surrogate protein signature. Spatial distribution of metabolic and immune markers was examined using Opal/Vectra multiplex immunofluorescent staining. Results: Analysis of TCGA dataset using unsupervised hierarchical clustering identified three patient subgroups with distinct metabolic-immune phenotypes and survival profiles: (1) immunehigh/metaboliclow, (2) metabolichigh/immunelow and (3) intermediate, with 5-yr overall survival (OS) rates of 71%, 51% and 49% respectively (p = 0.0015). The prognostic nature of MIGS was replicated in both validation cohorts (Table). Protein IHC signature was not prognostic. Metabolic and immune markers showed inverse expression patterns on multiplex staining. Preliminarily, presence of metabolichigh/immunelow signature was associated with ~ 20% worse OS at 1yr in PD-1 treated R/M HNC patients. Conclusions: We developed and validated a prognostic molecular classifier based on metabolic profile and immune status. This classifier may have clinical application to guide use of metabolic modification and targeted immunotherapies for HNC treatment.