Advancing Education, Research, and Quality of Care for the Head and Neck oncology patient.
Background: Aberrant splicing events (ASEs) play a central role in oncogenesis of head and neck cancers. Understanding the neoantigens generated by these ASEs is important to inform targets for new immunotherapies. Using our previously published algorithm OutSplice, a novel tool to detect ASEs based on outlier expression of junctions using bulk RNA-seq data, we analyzed associations between the ASEs identified by OutSplice to tumor immune infiltration and prognosis in head and neck squamous cell carcinoma and five other cancer types.
Methods: RNA-seq data from The Cancer Genome Atlas (TCGA) for head and neck squamous cell carcinoma (HNSC), breast adenocarcinoma (BRCA), lung squamous cell carcinoma (LUSC), lung adenocarcinoma (LUAD), bladder carcinoma (BLCA), and colon adenocarcinoma (COAD) were analyzed by OutSplice to identify ASEs for each tumor. In each cancer, tumors were stratified into high/low tumor mutational burden (TMB) based on median mutation counts and into high/low splice burden based on the median number of ASEs. Tumor immune infiltration was quantified with Xcell, a cell enrichment analysis webtool; the degree immune infiltration by splice burden was compared using Student’s t-tests. HNSC-specific TCGA survival data was analyzed with log-rank tests and Cox proportional hazards regression models to determine associations between overall and progression-free survival and tumor splice and mutational burden. OutSplice performance was compared to Psichomics, another publicly available ASE quantification tool.
Results: OutSplice broadly identified significantly (p<0.05) lower mean immune cell enrichment values in tumors with high splice burden than low splice burden across all six cancers. Specifically, this pattern was observed in CD8 T-cells, and M1 and M2 macrophages in HNSC; CD8 and CD4 T-cells, M1 and M2 macrophages, and the overall immune and stromal scores in LUSC and BLCA; CD4 T-cells, M1 macrophages, and the overall stromal score in BRCA; and CD4 T-cells, M1 and M2 macrophages, and the overall immune and stromal scores in COAD and LUAD. In the HNSC survival analysis, higher splice burden had poorer overall survival (hazard ratio (HR)=1.47, p=0.01) (Figure 1). Stratifying this analysis by TMB, higher splice burden had significantly poorer overall survival in low TMB (HR=1.65, p=0.03) but not in high TMB (HR=1.36, p=0.14). Progression-free survival was not different by splice burden. In comparison, the analysis with Psichomics did not show any significant relationships between splice burden and immune infiltration or prognosis in the analyzed tumor types.
Conclusions: Across six cancer types, the analysis using our novel OutSplice algorithm uniquely identified lower immune infiltration in tumors with higher splice burden. To our knowledge, we are the first to identify this association between immune infiltration and splice burden in a pan-cancer analysis. The degree of immune infiltration is a potential explanation for differences in prognosis by splice burden; in HNSC specifically, our analysis with OutSplice revealed that high splice burden was associated with poorer survival, especially in tumors with low mutational burden. These data support the utility of OutSplice for defining clinically relevant splicing events that affect the tumor immune microenvironment and its potential to predict prognosis and immunotherapy response.