Abstract 1467: Multiplexed immunofluorescence quantitation and validation of multiple immune cell types in colon cancer epithelium and stroma

Immune response to tumor growth and development is highly complex and varies with stromal and tumor cell composition and secreted factors. Different immune cell types can have opposing functions (tumor suppressive or immune suppressive) and their density, location and activation state can have profound effects on tumor outcome. Routine immune-profiling methods based on sample homogenization or limited in situ detection capabilities may be insufficient to measure this diversity. To fully decipher tumor immune response and contribution of different cell types, there is urgent need to develop new imaging tools for in situ immune cell-typing and quantification. MultiOmyx® technology allows repeated staining, imaging and cell-level analysis of multiple biomarkers (at least 60) on the same FFPE tissue section. Together with epithelial and stromal cell analysis, the relative location and quantity of immune cells can be established, enabling regional assessment of immune infiltration and activation status. Using this workflow, 747 stage I-III colorectal cancer patient samples in TMA format were multiplexed for a total of 61 proteins, including CD3 (all T-cells), CD8 (cytotoxic T-cells), CD20 (B-cells), CD68 (macrophages) and pan-cytokeratin (epithelial tumor cells). All images were registered, auto-fluorescence subtracted and illumination corrected. Individual cells were segmented using automated image analysis workflow, consisting of nuclei segmentation, epithelium segmentation and stroma-epithelial cell nuclei classification. We applied a probabilistic multi-class, multi-label classification algorithms based on multi-parametric models to build statistical models of CD protein expression and classify immune cells. Using selected images, expert annotations of the following immune cells were made [CD20+ (n = 4282 (cells)), CD3+ (n = 5600), CD8+ (n = 5346), and CD68+ (n = 4261), and defects (n = 1739)]. Support Vector Machines (SVM) were used to derive a statistical model for cell classification. To objectively evaluate the cell classification accuracy, we performed 10-fold cross validation. The methods described performed well in classifying cell-types, yielding ≥97% accuracy, relative to expert user annotations, for all immune cell types. Classification accuracy was slightly higher for lymphocytes vs. macrophages, likely due to more diffuse localization of CD68. In line with previous reports, higher T-cell infiltration was significantly correlated with recurrence-free survival in the entire cohort. In summary, we have developed a highly accurate quantitation method for in situ analysis of immune cells in tumor and stroma. Using the same workflow, additional cell lineage and functional markers can be added for deeper phenotyping and identification of innate and adaptive immune cell lineages, generating new insights into role of immune response in tumor progression. Citation Format: Christopher Sevinsky, Alberto Santamaria-Pang, Anup Sood, Yunxia Sui, Qing Li, Nicole LaPlante, Raghav Padmanabhan, Fiona Ginty. Multiplexed immunofluorescence quantitation and validation of multiple immune cell types in colon cancer epithelium and stroma. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1467.