Application of a novel multiplex imaging-based immunotherapy panel and AI-powered analysis solution for predictive spatial biomarker identification on immunotherapy-treated melanoma patients.

Abstract

9524 Background: There is an urgent need for more robust methods to differentiate immunotherapy responders from non-responders. In this study, we present a novel multiplex imaging (MI)-based immunotherapy panel and a comprehensive analysis pipeline to characterize the spatial distribution and function of immune cells and its application for spatial biomarker detection in a cohort of immunotherapy-treated melanoma patients. Methods: We designed a 28-plex panel to perform sequential immunofluorescence (seqIF) on the COMET platform to target key biomarkers associated with tumor microenvironment (TME), immune cell infiltration, and immune checkpoint pathways. Pre-treatment biopsies were obtained from 12 patients with known long-term response or rapid progression to immunotherapy combination treatment from the SECOMBIT Trial (NCT02631447) and profiled utilizing Nucleai’s deep-learning-based MI analysis pipeline, aiming to identify spatial biomarkers that can differentiate between long-term responders and non-responders. We identified 15 cell types, including 10 immune cell populations, in addition to 10 cell state markers. Cells were assigned to the tumor area or TME, and spatial features were calculated based on cell type, marker positivity, and gross area assignment. Results: Our novel MI panel and analysis pipeline demonstrated highly balanced accuracy (> 0.8) and F1 scores (> 0.8) in cell typing and protein quantification for most cell types and markers. This analysis pipeline enabled the quantification of known biomarkers such as T cell activation states, T cell infiltration patterns, and tertiary-lymphoid structure maturation. A comparison of calculated spatial features between long-term responders and rapid progressors revealed distinct immune cell interactions and differences in activation status across the tumor areas associated with response. Within the tumor area, the reciprocal interactions of tumor cells, cytotoxic CD8 T-cells and antigen-presenting cells (APC) were associated with a better outcome. In contrast, a high percentage of proliferating regulatory T cells within the tumor invasive margin was associated with a worse outcome. In the adjacent TME, endothelial cell interactions with T-cells and macrophage proliferation were associated with immunotherapy resistance. In contrast, the interaction between HLA-DR-expressing macrophages and APC cells was associated with an improved clinical outcome. Conclusions: Integrating MI with AI analysis has the potential to enhance our understanding of treatment efficacy and resistance mechanisms. Our preliminary data demonstrate that area-specific immune niches contribute to the success or failure of immunotherapy response and highlight the importance of spatial biology in predicting immunotherapy outcomes. Clinical trial information: NCT02631447 .