Abstract 547: Novel algorithms for spatial modeling of cellular interactions in the tumor microenvironment

Abstract

Abstract Background: Understanding the spatial interactions between tumor and immune cells in the heterogeneous tumor microenvironment is pivotal to improve clinical outcomes of immunotherapy. Various multiplexed imaging platforms have been recently developed to visualize the different immune cell subtypes with various distribution patterns that can impact the antitumor immunity. However, the lack of tools that allow for the spatial pattern modeling remains a major barrier for systematic analysis of cellular interactions in the TME. Methods: We used multiplexed immunohistochemistry (mIHC) to understand the cellular interplays in metastatic melanoma TME using FFPE section. This platform allows in situ quantitative single cell analysis with high-level of multiplexing while preserving the tumor heterogeneity. Single cell spatial data (position and location) in the context of TME can be obtained during mIHC, enabling further spatial modeling to characterize the cellular interactions between tumor immune cells. Novel spatial analysis algorithms are developed taking advantages of these spatial data. Results: Using R programming language, we developed two novel analytical approaches to reveal the spatial distributions of various immune cell subtypes relative to tumor cells. The cell neighborhood analysis algorithm uses the “applynbd()” function to traverse through every cellular point on a two-dimensional plane and creates neighborhoods of 24um in diameter centered by a tumor cell. This algorithm captures the counts of a specific type of immune cells that are in direct contact with a tumor cell, allowing the visualization and statistical analysis of cellular attraction/repulsion patterns. The cell aggregate analysis algorithm uses a sliding-window based approach to capture cell aggregates with certain densities and the areas within the TME covered by specific cell aggregates. Combined with the Ripley K-function, the cell aggregate algorithm captures the aggregation vs. segregation pattern between different cell types, enabling the clustering and distance analysis of immune cells relative to tumor cells. Our algorithms demonstrate that tumor cells either attract CD3+ tumor-infiltrating T cells or favor the aggregation of CD20+ B cell in peri-tumor areas depends on the levels of tumor cell human leukocyte antigen-1 (HLA-1) expression. Conclusion: Our novel spatial analysis algorithms enable the modeling of different interaction patterns between tumor and immune cells (direct contact vs. aggregation) at both single cell and TME level. It demonstrates that tumor cells with heterogeneous properties can impact the immune cells’ distributions in the TME with various biological outcomes. It also establishes tools that are necessary for systematic analysis of the TME, allowing the elucidation of the “homogeneous patterns” within the heterogeneous TME. Citation Format: Yiyi Yan, Alberto Santamaria-Pang, Michael Gerdes, Fiona Ginty, Anup Sood, Svetomir N. Markovic, Alexey Leontovich. Novel algorithms for spatial modeling of cellular interactions in the tumor microenvironment [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 547. doi:10.1158/1538-7445.AM2017-547