An Immune-Based Genomic Classifier (GC) Prognosticates for Survival in Endemic Nasopharyngeal Carcinoma

Abstract

Nasopharyngeal carcinoma (NPC) is a distinct head and neck cancer that is associated with Epstein-Barr virus (EBV). Its histomorphology is characterized by epithelial carcinoma that is admixed with immune cells. This cellular pattern concurs with whole exome sequencing (WES) profiles of NPC that reveal several recurrent mutations involving NFKß and MHC-class genes, which are linked to poor outcomes. There is little information on the transcriptomic landscape of NPC. Here, we performed whole transcriptome sequencing (RNAseq) of a prospectively recruited cohort of NPC patients who were treated at a single institution.Patients with biopsy-proven NPC, staged using the UICC/AJCC 8th ed, were prospectively recruited under an institutional approved protocol. All formalin-fixed paraffin-embedded specimens were centrally reviewed. Specimens with tumor cellularity of > 70% were macrodissected and Ranse (Illumina Trusses, CA) was performed. All patients underwent radiotherapy (RT); concurrent (CRT) and/or induction (IC) or adjuvant chemotherapy (AC) was performed for patients with advanced NPC. Primary clinical endpoint was disease-free survival (DFS).Of 1,040 patients recruited, 145 patients had specimens that were successfully profiled by Ranse. Median age of this sub-cohort was 54.2 (IQR: 45.2-61.1) y; median follow-up was 4.8 (IQR: 3.3-8.1) y. Stage distribution was as such: 10 (6.9%) for Stage 1, 36 (24.8%) for Stage 2, 47 (32.4%) for Stage 3 and 41 (28.3%) for Stage 4A; 7 (4.8%) for Stage 4B; 4 (2.8%) were unclassified. 77 (53.1%) and 55 (37.9%) had EBV DNA titers of < 4,000 copies/ml and ≥4,000 copies/ml, respectively (13 [9.0%] missing). In terms of treatment received, 42 (29.0%) underwent RT alone; 50 (34.5%) and 41 (28.3%) received CRT and IC/AC+CRT; 12 (8.3%) had other treatment combinations. Unsupervised hierarchical clustering using expression of the top 1,000 most variable genes across the cohort revealed three distinct clusters that were characterized by differential expression of immune-related genes; incidentally, the cluster of patients with low immune score were enriched for recurrences, while the intermediate/high immune score groups had comparable prognoses. Subsequently, we performed gene-selection based on the low and intermediate/high immune score clusters and identified 511 genes that were significantly differentially expressed between them. We further clustered these genes to obtain a representative 5-gene GC that was prognostic for DFS (HR 3.28 [95% CI: 1.36-7.9]) on multivariable analyses, adjusted for T- and N-status and EBV DNA. Our GC was also prognostic for DFS in a validation dataset (N = 88, Zhang et al., 2017); HR 2.1 (95% CI: 0.73-6).Herein, we developed and validated an immune-based prognostic GC for NPC. Our findings underscore the importance of the immune landscape in EBV-positive NPC.