Spatial Single-Cell Transcriptomic Analysis in Breast Cancer Reveals Potential Biomarkers for PD-1 Blockade Therapy

Introduction: Spatially defined cellular interaction and crosstalk are eminently important in deciphering key molecular messages driving oncogenesis and disease progression. To date, methods enabling high-plex true single-cell resolution profiling under spatial settings are gradually becoming available and those majorly include the expansion of spatial transcriptomics (ST) being utilized. Results: Through in-depth spatial single-cell profiling on four breast cancer (BC) tissue samples bearing distinct biological characteristics, we evaluated the analytical performance benchmarked against conventional pathology and by selecting pre-defined region-of-interests (ROIs), we consolidated the technical robustness of this method in defining different molecular subtypes at the transcript level matching with canonical immunohistochemistry. Moreover, we demonstrated that high-dimensional ST data is capable of identifying a major cellular network inter-wired via macrophage and cytotoxic T cells interaction in tumor adjacent cellular neighborhood via PD-L1/CD80 and CD86/CTLA4 axis, a phenomenon reflecting an improved PD-1 mediated drug response observed clinically. By incorporating open-source computational methods (Tangram and SpaGE), we found compatible inference tools for in-situ expression imputation, an approach generalizable to enable deeper spatial profiling using Xenium in-situ or other parallel approaches. Discussion: Our spatial single-cell ST sets as a technical and analytical prototype for those using similar approaches for high-dimensional in-situ profiling work. Materials: We applied a newly developed spatial single-cell technology (Xenium in-situ) to interrogate the spatial single-cell architecture of the complex tumor microenvironment on a set of breast cancer patient tissues (luminal-type, HER2 2+/HR- and triple negative breast cancer, TNBC) and benchmarked against multiple clinicopathological features using bioinformatic tools.