Abstract B33: Discovering distinct immune environments defined by the states of single cells within tumor microenvironment

Tumor microenvironment (TME) plays an important role in determining the oncogenic behavior of tumors and their response to immune therapies. Profiling the cellular states of cells within a TME and further studying their interactions would shed light on complex mechanisms that define different TMEs and allow us to understand different immune evasion mechanisms. Currently, only a small fraction of head and neck tumors respond to contemporary immunotherapy, indicating that distinct immune environments (thereby distinct immune evasion mechanisms) exist among tumors. Thus, investigating different immune environments is of significant clinical importance. We developed a novel computational framework to study the cellular states of single cells and interactions among them. We have collected single-cell RNA sequence (scRNAseq) data of over 100,000 cells from 18 head and neck squamous carcinoma (HNSC) tumors, including tumor cells, stromal cells, and immune-related (CD45+) cells. Based on the assumption that the cellular state of a single cell can be represented by the activation status of biologic processes in the cell, which in turn can be represented by the expression status of gene modules regulated by the processes, we applied the nested hierarchical Dirichlet processes (NHDP) model to identify the gene modules potentially regulated by different biologic processes. Gene modules identified by NHDP model are shown to reflect activation status of biologic processes, such as activation or exhaustion of T cell, which enabled us to group cells according to cellular states. Discovery of gene modules from single cells enabled us to deconvolute bulk transcriptomic data from TCGA to infer the presence and abundance of cells with different cellular states. Based on such information, we discovered 4 subtypes of immune environments, and patients belonging to different subtypes exhibited significantly different survival outcomes. Furthermore, we applied NHDP to tumor and stromal cells and discovered gene modules expressed in nonimmune cells. The expression status of these gene modules is highly predictive of the expression status of the gene modules from immune cells in TCGA HNSC tumors. That is, this approach enables to study the interactions among different types of cells in a TME. Finally, based on express status of gene modules from tumor and stromal cells, we can accurately predict the immune subtype of a tumor. In summary, we developed a novel computational framework to investigate distinct immune environments, and such information can potentially be used to predict tumor response to contemporary immune therapies and develop novel immune therapy targets. Citation Format: Xueer Chen, Lujia Chen, Cornelius Kurten, Lazar Vujanovic, Aditi Kulkarni, Robert Ferris, Xinghua Lu. Discovering distinct immune environments defined by the states of single cells within tumor microenvironment [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 B33.