Abstract 156: Tumor microenvironment profiling of multiplexed images in head and neck cancer reveals spatial features associated with immunotherapy response

Abstract Introduction: Current objective response rates of recurrent or metastatic head and neck squamous cell carcinoma (RM HNSCC) to immune checkpoint inhibition (ICI) remain less than 20%, highlighting the innate variability of patient tumors that preclude their immunogenic clearance. Such outcomes in HNSCC have driven the exploration of the immune and tumor centric properties that drive such refractory responses to ICIs. This study expands on this knowledge base by leveraging highly multiplexed imaging combined with a deep learning cell type classification pipeline to comprehensively evaluate the spatial properties of tumor and immune cells associated with ICI response. Methods: 65 RM HNSCC pre-ICI biopsies on tissue microarrays were imaged using highly multiplexed immunofluorescence performed on the Akoya PhenoCycler platform. Cell type classification was performed using a deep learning platform developed by Nucleai. Marker positivity was used to further assign cell type functional states. Supervised spatial clustering and tumor cell density contours were used to define tissue regions in TMA cores. Cell frequencies, tumor-immune cell ratios and cell-cell interactions were computed within these regions as input for tests against clinical correlates. Results: 95, 000 tumor cells and 76, 000 immune cells were phenotyped for both functional and metabolic states. Macrophages and CD4 T cells were the most abundant immune cell types in the cohort, with both CD4 and CD8 T cells demonstrating highly OXPHOS+ metabolically active states. Univariate analysis for progression free survival indicated CD8 T cells in tumor regions were associated with better PFS, while CD4 Tregs expressing ICOS in stromal regions correlated with poorer PFS. Interestingly, both B cells and Granulocytes at the tumor-stroma interface correlated with poorer PFS. Furthermore, the interaction between tumor cells and highly metabolically active fibroblasts within tumor regions was also associated with poorer PFS. Conclusions and future directions: Our study interrogated spatial compartments in HNSCC tissues for cellular features that may predict response to ICI therapy. Such approaches are limited by the provision of sufficient cells to derive meaningful biology. Future aims for this study include deep learning methods to contrast tissue subgraphs to identify regions that differ significantly between patients, and thereby provide novel insights into the relevant spatial communities that drive cancer immunotherapy responses in patient biopsies. Citation Format: Ettai Markovits, James Monkman, Aaron Kilgallon, Thazin Nwe Aung, David Rimm, Arutha Kulasinghe. Tumor microenvironment profiling of multiplexed images in head and neck cancer reveals spatial features associated with immunotherapy response [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 156.