1283 In vivo phenotyping of the tumor immune microenvironment to predict melanoma outcomes

Tumor outcomes (progression, response to immunotherapies) are influenced by tumor-intrinsic and the tumor-immune microenvironment (TiME) factors. To overcome the existing limitations of phenotyping tumors using single-time, single-siteex vivo biopsies in predicting tumor behavior, we propose an integrated dynamic in vivo approach for phenotyping TiME by combining inflammation, vasculature/ angiogenesis and tumor-intrinsic features through dynamic high-resolution reflectance confocal microscopy (RCM) in patients. Melanoma patients (n=35) were imaged to characterize density and spatial distribution of tumor, inflammation, vasculature/angiogenesis and leukocyte trafficking. Unsupervised clustering (hierarchical clustering analysis, HCA and principal component analysis, PCA) was performed on manually evaluated TiME features to derive phenotypes, and correlated with immune cells (T-cells, B-cells) and tertiary lymphoid structures (TLS), and response to topical toll-like receptor agonist imiquimod. We demonstrate three main phenotypes, InflamHIGHVascHIGH, InflamHIGHVascLOW and InflamLOWVascHIGH in HCA and PCA (top contributors to PC1: inflammation while to PC2: vasculature). Highest T-cell and B-cell infiltration (p<0.01), and TLS presence (ns) seen in the InflamHIGH phenotype. Overall, lower lichenoid inflammation and number of vessels were found to be hallmarks of non-responders. Our preliminary results support presence of unique TiME phenotypes in melanoma that correlate with underlying immune states, and response to topical TLRA immunotherapy.