GBM CODEX clustered single cell data
We imaged in total 58 glioblastoma explants using CODEX technology with a 55-marker panel. The imaging data was segmented into single cells based on best-focus nuclear staining using DAPI and DRAQ5 as reference. Spatial (X/Y/Z) coordinates can locate each cell within a specific slide. Cell type annotation was thereafter performed using both VORTEX clustering and manual supervision on the slides (refer to software package from the NOLAN lab, Stanford (https://github.com/nolanlab/CODEX). A previously established highly multiplexed tissue cytometry platform called CO-Detection by indEXing (CODEX) was re-engineered here here to create multidimensional imaging datasets of glioblastoma (GBM) bioreactor explants. In this procedure, DNA-barcoded antibodies bound to antigens present in the tissue were iteratively rendered visible by hybridizing complementary fluorescent DNA oligonucleotides. An algorithmic pipeline was used to process raw imaging data, segment and identify single cells and their localizations within tissues, and quantify their marker expression. Unsupervised clustering, followed by manual curation of clusters based on marker expression, morphology and tissue localization, was used to call out specific cell types. Expression of selected markers per cell were manually gated in CellEngine (https://cellengine.com). Cellular neighborhoods were algorithmically identified. We identified seven conserved, distinct tissue compartments (TCs) –a collection of components characteristic of the GBM iTME. This study provides a framework for interrogating how glioblastoma patients could potentially profit from local neoadjuvant immunotherapies. Clustering_results.csv table contains cell types annotation, expression profiles, coordinates of all segmented objected identified in the explants analyzed in this study. Further, it contains information on the biopsy location (c=center, p= periphery of the tumor), treatment info of the explants (control, anti-CD47, anti-PD1, combination), and the respective anonymized tumor internal tumor ID.
Steps to reproduce
Raw, unprocessed imaging data is available upon request from the authors. Expression of selected markers per cell were manually gated in CellEngine (https://cellengine.com). Segmentation, clustering and annotation was performed by the software packages provided by the Nolan lab: https://github.com/nolanlab/CODEX. The here provided, annotated, spatial data can be loaded into CytoMAP software (https://gitlab.com/gernerlab/cytomap/-/wikis/home) for further interrogation and analysis. The processed CytoMAP File is available upon request by the authors.