Characterization of cytokinome landscape for clinical responses in human cancers.

Dysfunctional intratumoral immune reactions are shaped by complex networks of cytokines (including chemokines), and how the cytokinome landscape coordinates with tumors has not been systematically investigated. Using high-dimensional datasets of cancer specimens, we explored the transcript abundance, biomarker potential, and prognostic impact of local cytokines across 19 tumor types. We found that most cytokines are highly locally dysregulated (p = 0.024), revealing spatiotemporal pattern of local cytokines in the development of cancers. In addition, we noted the significant downregulation of CCL14 and CXCL12 in 9 and 10 cancer types, respectively, implying their crucial roles in tumor pathogenesis. We also found that cytokines showed significantly higher specificity properties compared to other protein-coding genes (PCGs) in primary tumor specimens (p << 0.001), indicating that tissue context remains an issue when considering cancer cytokinomes. Finally, we linked concentrations of local cytokines to patient survival. Our results thus provide a panoramic view of pan-cancer cytokinomes, which highlights tumor type specificity of cancer-related cytokines and their impacts on disease prognosis.