Abstract Despite advances in our understanding of tumor biology as well as radiation, chemotherapy, and surgical treatments, head and neck squamous cell carcinoma (HNSCC) remains the sixth leading cause of cancer-related mortality and a major cause of morbidity, with over 550,000 new cases per year. Although HNSCC mortality has slightly declined with improved tolerance of therapy, the majority of deaths in HNSCC are related to advanced disease with locoregional lymph node (LN) or distant metastasis. Indeed, the presence of LN metastases reduces survival by approximately 50% in all HNSCC patients. Unfortunately, a firm biologic understanding of pathways that underlie metastasis in HNSCC (and many other solid tumors) remains elusive. Tumor heterogeneity in HNSCC is a major challenge in clinical management and can manifest as variability among tumors (intertumoral heterogeneity) or among cells harboring distinct mutations or expressing unique transcriptional programs from the same tumor (intratumoral heterogeneity). Prior studies have suggested that intratumoral heterogeneity is present in HNSCC based on comparative genomic hybridization assays and DNA ploidy analyses, with recent studies validating intratumoral heterogeneity using whole-genome sequencing. In addition, there is growing interest in subpopulations of HNSCC cells, which may have unique properties such as the ability to self-renew or invade, contributing to recurrence or metastasis, respectively. The clinical significance of intratumoral heterogeneity has recently been demonstrated with a quantitative measure of genetic heterogeneity based on whole-exome sequencing (WES). This measure, termed mutant-allele tumor heterogeneity (MATH), suggests that intratumoral heterogeneity is associated with tumor progression and poor responses to treatment, supporting the idea that greater genetic variability predicts a worse clinical outcome in HNSCC. However, whether intratumoral transcriptional heterogeneity also contributes to clinical outcomes remained an open question. We recently utilized single cell RNA-sequencing (scRNA-seq), a powerful new approach to characterize expression heterogeneity present within oral cavity HNSCC (Puram et al., Cell 2017). The strength of these analyses is the ability to capture the transcriptional profile of each individual cell without contamination from other cells that may be present within the sample. As a result, single-cell sequencing offers 1) more detailed, nuanced expression data than analyses of tumor fragments and 2) captures the transcriptional profile of every cell, even rare populations, that might be overlooked in pooled RNA. We sequenced ~6000 single cells from 18 samples, including five matched primary/lymph node pairs. Our analyses reveal that stromal and immune populations have well-defined and consistent expression programs across tumors. In contrast, malignant cells are highly heterogeneous with variable gene signatures related to cell cycle, stress, hypoxia, epithelial structures, and partial epithelial-to-mesenchymal transition (p-EMT). Cells expressing the p-EMT program spatially localized to the leading edge of primary tumors and in further in vitro analyses, we found that p-EMT cells were more invasive. Integration of single-cell transcriptomes with bulk expression profiles for >500 tumors from The Cancer Genome Atlas using a deconvolution approach identified a malignant cell-specific signature for each of these samples. Remarkably, this analysis identified the p-EMT program as the primary source of variability in HNSCC, while enabling us to redefine HNSCC subtypes by their malignant and stromal composition. Further integration with clinical annotations uncovered a significant association between the p-EMT program and nodal metastasis, grade, and adverse pathologic features including lymphovascular invasion and extracapsular extension. Consistent with these findings, the p-EMT score was a significant predictor of occult nodal metastases in clinically N0 patients, with predictive power equivalent to tumor thickness/size. Our studies provide a detailed map of expression heterogeneity in the HNSCC ecosystem and define an in vivo p-EMT program associated with metastasis and disease progression. Citation Format: Sidharth V. Puram. Understanding programs of invasion and metastasis in head and neck cancer [abstract]. In: Proceedings of the AACR-AHNS Head and Neck Cancer Conference: Optimizing Survival and Quality of Life through Basic, Clinical, and Translational Research; 2019 Apr 29-30; Austin, TX. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(12_Suppl_2):Abstract nr IA08.