Spatial proteomics enables identification of prognostic biomarkers in papillary renal cell carcinoma

Abstract Background Papillary renal cell carcinoma (PRCC) is the second most common adult kidney cancer histology, constituting 15-20% of cases. While some patients may have indolent PRCC tumors that grow slowly, other tumors rapidly metastasize. Some PRCC tumors respond to immune checkpoint inhibitors. Thus, understanding the immune tumor microenvironment and how it correlates with patient outcomes in this relatively rare disease is a critical need. Spatial interrogation of patient samples has the potential to offer novel insights into the tumor-immune axis and provide avenues for enhanced diagnosis and treatment. Addressing questions of spatial arrangement within tumors has remained a technical and biological challenge, but emerging spatial biology technologies provide molecular data at single cell resolution. We hypothesized that the spatial interaction of immune, stromal and tumor cells would be prognostic for PRCC patients. Methods A tissue microarray was assembled from an archive of ~100 patients presenting with PRCC. This dataset was assayed with PhenoCycler/CODEX (Akoya Biosciences) using a 31-antibody panel with immune- and cancer-related proteins. We have developed methodology and novel algorithms to perform signal normalization, cell segmentation, and cell typing. We computed neighborhoods for each of 2.5 million cells and performed network analysis to identify spatial clusters. This method allowed us to identify clusters consistently present across multiple TMA spots from the same patient and across multiple patients. Results Using spatial neighborhood analysis, we have identified diverse spatial clusters of potential clinical relevance, including five distinct M2 macrophage spatial clusters. We have described one of the clusters as being physically associated with helper T cells, which is visualized on a PRCC spot and shows co-occurence of M2-macrophages (CD163) and helper T cells (CD4). This pattern was replicated in additional TMA spots from the same patient. In contrast, other clusters of M2-macrophages (M2-M2 spatial cluster and M2-M1 macrophages) are visualized to show vastly different cellular neighborhoods. Clinical associations show that the patients with presence of the M2-T helper cluster have a poor cancer-associated survival (p=0.005). In comparison, the total proportion of M2 macrophages is not associated with survival (p=0.4), highlighting the importance of characterizing spatial interactions beyond cell type quantitation. Conclusions In summary, we highlight the utility of spatial biology to explain the heterogeneity in patients’ tumors and to uncover novel correlates of clinical phenotypes, establishing a platform for future discovery in this field and the identification of additional spatial correlates of patient outcome and clinical response. CDMRP DOD Funding: yes