ABSTRACT Head and neck squamous cell carcinoma (HNSCC) is the seventh most common cancer worldwide. In the oral cavity, most cases of squamous cell carcinoma begin as a precursor lesion classified by the World Health Organization (WHO) as an “oral potentially malignant disorder (OPMD).” Oral leukoplakia is the most common OPMD, with a global prevalence of 4.1% and a malignant transformation rate between 0.1%-34.0%, and the malignant transformation rate increases to 40% in oral dysplasia. These wide ranges of malignant transformation rates suggest an unmet need to develop prognostic biomarkers that can better differentiate benign from premalignant lesions and predict the risk of transformation of premalignant lesions to invasive cancer. During the development of cancer, immunoediting occurs within the tumor microenvironment. Immunoediting consists of three phases: elimination, equilibrium, and escape. We posit that there are defined immune signatures within a lesional microenvironment that correlate with the transformation of precancerous cells to invasive cancer based on the known phases of immunoediting. The current standard of care tools to establish benign from premalignant oral lesions include conventional hematoxylin/eosin (HE) and single staining immunohistochemistry (IHC), which limits the number of immune cells which can be evaluated at any one time. Thus, we developed an innovative spatial omics technology (SAFE) that facilitates the comprehensive and deep multiplexing of whole tissue sections and incorporates artificial intelligence and machine learning approaches to accelerate the analysis of ~45 molecular and immune signatures within oral lesions in a clinically appropriate timeframe. We propose to perform single-cell RNA sequencing to identify the unique cellular and immunological transcriptional programs that distinguish benign and premalignant oral lesions (N=60) and will assess relevant protein expression and spatial localization via SAFE (Aim 1). Subsequently, in Aim 2, we will evaluate the defining molecular and/or immunological signature(s) in a unique set of patients (N=55) with serial biopsies documenting transformation from premalignancy to cancer over 20 years against a separate cohort of benign and premalignant lesions (N=155). From our dataset, we will develop an oral cancer progression model that incorporates the host immune response for the first time to improve risk assessment for malignant progression to a degree superior to what is currently possible.

NARRATIVE This proposal aims to improve the early detection of head and neck cancers by applying an innovative, deep spatial omic technology and computational tools to quantitatively differentiate benign from premalignant lesions and identify prognostic immune-based signatures that assess the risk of transformation from premalignant to malignant disease. A better understanding of the molecular and immune characteristics of dysplastic lesions is needed for improving oral lesion classification relevant to predicting cancer progression, which can impact cancer prevention and treatment.