Abstract The microenvironment surrounding brain cancer cells, and their position relative to anatomical features such as the vasculature, is a critical contributor to tumor malignancy. Immunohistochemical staining is widely used to examine such multicellular interactions, and approaches like cyclic immunohistochemistry (cycIHC), multiplexed immunofluorescence, or mass-based imaging can include up to 100 channels of interest. However, large (often > 50 GB) datasets from cycIHC or comparable approaches currently require an expert data scientist to concatenate open-source tools for each step of image pre-processing, registration, and segmentation, or the use of expensive proprietary software. We generated an open-source, unified, and user-friendly workflow for processing and analyzing cycIHC data - Cyclic Analysis of Single-Cell Subsets and Tissue Territories (CASSATT). CASSATT registers scanned slide images across all rounds of staining, segments individual nuclei, and scores marker expression on each detected cell. Beyond straightforward data analysis outputs such as dimensionality reduction, clustering, cell population identification and quantification, CASSATT explores the spatial relationships between cell populations via a statistical analysis module that can be applied to other high-dimensional imaging data types. In brief, it helps users identify populations of cells that interact - or do not interact - at frequencies that are greater than those occurring by chance. It also identifies specific ‘neighborhoods’ based on the cell types that surround each cell in the sample. The presence and location of these neighborhoods can be compared across slides or within distinct regions within a tissue. Within a newly generated cycIHC dataset consisting of 6 glioblastoma tissue sections processed through 8 cycles of chromogenic IHC staining, CASSATT successfully identified populations detected in parallel flow cytometry analyses and revealed specific associations between tumor and immune populations. Additional successful benchmarking was completed on a published tissue microarray dataset consisting of 107 cores processed through 18 cycles of staining and imaging.
Read full abstract