Combined spatial Keratin expression profiles at the invasive front represent a prognostic classifier for head and neck cancer

Keratin proteins are intermediate filaments that provide mechanical support and guide a plethora of important functions in epithelial cells. Although Keratins can be effectively used to characterize and differentiate cell lineages in many epithelia, it is still unclear if combinations of Keratin types and their spatial expression patterns can aid prognostication of head and neck squamous cell carcinoma (HNSCC). Here, we have assessed a panel of Keratin markers in four HPV negative HNSCC patient-derived organoid (PDO) models and their corresponding primary patient tissue to define spatial expression profiles of basal and supra-basal Keratins during invasion in Collagen-I matrices. Based on these profiles we have devised a scoring classification consisting of the following profiles: (i) no expression (Negative), (ii) uniform expression (Homogenous), (iii) heterogeneous expression (Mosaic), (iv) expression in the tumor core (Core), and (v) expression at the invasive front (Edge). Subsequently, this 5-tier system was validated in a well-documented tissue cohort of 157 HNSCC patients. We observe that patients with a Keratin 13 (K13) Edge profile have a significantly better prognosis than patients with K13 Core expression or K13 Negative tumors. Interestingly, we also find that K14 mosaicism strongly correlates with a favorable prognosis. We show that the absence of K13 from the tumor edge, in combination with K14 homogenous expression is a strong biomarker for poor prognosis in HNSCC. In short, our work indicates that defining the spatial expression patterns of basal and supra-basal markers in invasive HNSCC can benefit patient prognostication.