Multidimensional, spatially resolved immunologic hallmarks of response to neoadjuvant immune checkpoint blockade (ICB) therapies.

Abstract

9562 Background: The groundbreaking neoadjuvant ICB clinical trials have established beyond doubt that ICB before surgery will become the new standard of care for metastatic melanoma. Importantly, the shift from radiographic to pathologic response scoring offers an unprecedented window of opportunity to interrogate a goldmine of ‘on-treatment’ biospecimens. To date, single-cell profiling and region-based spatial transcriptomics (ST) have highlighted the importance of tertiary lymphoid structures (TLS) and stem-like T-cells as positive predictors of ICB response. However, the molecular mechanisms by which tumor infiltrating lymphocytes communicate and organize multicellular immune hubs within the spatial context of the tumor ecosystem remains poorly understood. Methods: To identify robust biomarkers of response to neoadjuvant ICB, we assembled a cohort of 58 stage III melanoma patients treated with neoadjuvant ICB [24 ipilimumab-nivolumab (IPI-NIVO), 21 NIVO-relatlimab (RELA), 13 PD1 mono] and deployed transformative technologies and computational methods, including single-cell FFPE sequencing, multiplexed FISH single-cell ST (MERFISH 305 genes), digital pathology, 3D open-top light-sheet imaging and AI-based computational pathology, to decipher the complex neoadjuvant ICB tumor ecosystem in response to therapy. We also developed 2 critical computational tools, SCIRA (Spatial Cellular Interaction and Receptor Activation), to compute receptor-ligand (R-L) interactions in whole slide images, and GC-SCAN (graph-based spatial clustering against noise), a graph-based algorithm that quantifies locally clustered structures from spatial -omics data. Results: Our results showed that the quantity and size of hyper-expanded germinal center/TLS, with increased GC: non-GC B-cell ratio, plasma cells and spatially resolved stem-like T-cells are strongly associated with response. IPI-NIVO elicited significantly stronger GC proliferation compared to NIVO-RELA and PD1 monotherapy, suggesting anti-CTLA4 can robustly induce germinal center/TLS proliferation. SCIRA spatial R-L analyses revealed the critical chemokine R-L interactions that organize the GC- and T-cell zones in response to therapy. Lastly, 3D light-sheet imaging revealed remarkable morphologic heterogeneity in 3D, with interconnected GC-TLS networks that are indicative of long-range molecular gradients. Conclusions: Our investigations herein have provided a comprehensive characterization of the immune architectures, cellular communications and 3D large-scale morphologic organizations of the TME that drive response to neoadjuvant ICB therapy. We believe the results of this study will enable the development of robust predictive biomarkers to guide the design of next generation combination ICB therapies in the clinical trial setting for melanoma and other cancer types.