Background The pathway from discovery to clinical adoption of predictive spatial biomarkers (spatial signatures) for immunotherapy response requires a solution that bridges ultrahigh-plex discovery experiments with targeted high-throughput translational and clinical studies. A critical step toward ensuring the successful transition is the harmonizing of technologies for staining, imaging, and data analysis. The aim of this study is to demonstrate how the integration of spatial multiplexed imaging technologies from Akoya Biosciences® and associated data analysis methods provides an effective workflow for the spatial phenotyping of the tumor microenvironment (TME) across the discovery to clinical continuum.

Methods Here we profiled an array of human formalin-fixed, paraffin-embedded (FFPE) cancer tissues using ultrahigh-plex PhenoCode™ Discovery panels (PDP) comprising of cell lineage, structural, immune activation, and checkpoint markers on the PhenoCycler®-Fusion (PCF) spatial biology platform. This was followed by running PhenoCode™ Signature panels (PSP) targeted to key biomarkers of immune contexture, macrophage polarization, and T cell activation status identified from the discovery studies in high-throughput experiments using the PhenoImager® HT platform. Image analysis was performed on multiplexed images using deep learning algorithms developed in Visiopharm® to segment specific tissue regions of interest (ROI) and to perform accurate cell detection and classification of different cell phenotypes. Spatial interactions among the cell-types were also explored using spatial neighborhood analysis followed by identification of distinct spatial signatures.

Results We have phenotyped multiple cancer types and quantified immune cell distributions, and their spatial interactions. Our analyses of the ultrahigh-plex data revealed distinct spatial relationships within the different tumor types. Immune profiling coupled with complementary high-throughput signature panels revealed correlation to the ultrahigh-plex analysis, paving the way for a more targeted approach toward the identification of predictive spatial signatures for immunotherapy outcomes.

Conclusions The combination of ultrahigh-plex discovery panels, targeted high-throughput signature panels, and deep learning quantitative image analysis allows for deeper characterization of complex cellular interactions in the tumor microenvironment. These methodologies also shorten the developing and identifying spatial biomarker signatures with a predictive value. This study also further highlights the importance of Akoya’s robust end-to-end workflows to deliver optimal staining, imaging, and analysis to facilitate spatial phenotyping and interpretation across pathologically complex human tissue samples for the development of spatial signatures.