Predictive Molecular Biomarkers for HPV-Associated Head and Neck Squamous Cell Carcinoma

Abstract

<h3>Purpose/Objective(s)</h3> We hypothesized that molecular characteristics of HPV-associated head and neck squamous cell carcinoma (HPV+ HNSCC) will predict tumor behavior and response to therapy. Using TCGA, we previously correlated genetic defects in NF-kB regulators with outcome. Here, we validate this prognostic biomarker and identify a gene expression classifier that correlates with outcome in HPV+ HNSCC. <h3>Materials/Methods</h3> Total (ribosomal-depletion) RNA and targeted DNA sequencing was performed on an institutional cohort of HPV+ OPSCC tumors. Defects in NF-kB regulator genes: <i>TRAF3, CYLD, TRAF2, MYD88, NF-kBIA, TNFAIP3, TRAF6, BIRC2, BIRC3, MAP3K14</i> were identified and correlated with patient outcome. Expression data from <i>TRAF3</i> and <i>CYLD</i> mutant tumors was used to identify an initial gene set of differentially expressed genes. This initial set was optimized using machine learning to classify tumors based on RNA expression to NF-kB active and inactive sub-groups. The final gene set, termed the NF-kB activity classifier (NAC), was then correlated with patient outcome. <h3>Results</h3> Genomic defects in regulators of NF-kB that correlated with outcome in HPV+ HNSCC were initially identified using TCGA data. Herein, we find this association to be again observed in an independent patient cohort. Specifically, patients whose tumors harbored defects in the above listed genes had significantly better overall (p=0.03) and progression-free survival (p=0.03) compared to patients whose tumors were wild-type for these genes. Because we noted differential expression of many genes in tumors with and without defects in TRAF3 and CYLD, we also created a gene expression classifier as a complimentary method to distinguish these subsets. Gene set enrichment analysis demonstrated that the optimized classifier gene set was markedly enriched in NF-kB target genes, thereby was named the NF-kB Activity Classifier (NAC). Using the NAC, two HPV+ HNSCC patient populations were identified using TCGA data. Patients whose tumors had high NF-kB activity had significantly better recurrence-free survival compared to patients whose tumors had low NF-kB activity. <h3>Conclusion</h3> To improve quality of life for HPV+ HNSCC survivors, therapeutic deintensification is a goal of physicians treating HPV+ HNSCC. A major barrier has been determining appropriate patients whose treatment can be de-escalated without sacrificing the high cure rate associated with standard high-dose concurrent chemoradiation. Currently, TNM classification and smoking history are the only criteria available without pathologic analysis of resected tumors and nodes or induction chemotherapy. Here, using an independent cohort, we validated a prognostic biomarker based on defects in NF-kB pathway genes and created a new RNA based prognostic classifier that also correlates with outcome.