Author: Casey C. Heirman, Kyle J. Lafata, Xiang Li, Breylon Riley, Jack B Stevens, Tammara Watts 👨🔬
Affiliation: Duke University, Department of Radiation Oncology, Duke University 🌍
Purpose: To leverage quantitative fluorescence imaging and spatial transcriptomics for characterizing the spatial and molecular heterogeneity of the tumor microenvironment (TME) in HPV+ head and neck squamous cell carcinoma (HNSCC).
Methods: We investigated compartment-specific mechanisms underlying immune modulation and therapeutic resistance by integrating deep learning-based fluorescence imaging and spatial transcriptomics to analyze immune-tumor interactions in HPV+ oropharyngeal cancer. Graph-based approaches were applied to quantify the spatial organization of CD45+ immune cells and PanCK+ tumor cells from high-resolution histological images, with nuclei centroids as nodes and spatial proximities as edges. Topological features, such as centrality and edge density, were extracted to characterize immune cell distributions across TME compartments (stroma, immune, and tumor). To further investigate immune-tumor interactions, we selected genes relevant to immune activity and HNSCC using publicly available data from ProteinAtlas and Ensembl's BioMart, then analyzed their compartment-specific expression patterns to explore immune dynamics and heterogeneity.
Results: In the stroma, ADAM8 gene expression exhibited a mean count of 12.0 (median = 6.5, IQR = 14.75), with 16.7% of patients showing expression levels above the threshold, suggesting stromal-driven immune activity in a subset of TMEs. In the immune compartment, CCL20 expression displayed a mean count of 23.1 (median = 14.0, IQR = 26.0), with 3.7% of patients exhibiting high expression levels, indicative of localized chemokine gradients mediating immune recruitment. In the tumor compartment, CXCL1 expression demonstrated significant heterogeneity, with a mean count of 55.4 (median = 24.5, SD = 91.24), and 10.7% of patients showing elevated levels associated with immune evasion and pro-tumor inflammation.
Conclusion: Compartment-specific analysis revealed distinct immune modulation mechanisms. By combining graph-based analysis of immune cell topology with spatially resolved gene expression, we provide a high-resolution map of immune-tumor interactions, offering insights into actionable therapeutic targets and personalized treatment strategies.