Multicellular immune ecotypes within solid tumors predict real-world therapeutic benefits with immune checkpoint inhibitors

作者：Xuefeng Wang, Tingyi Li, Islam Eljilany, Alex C. Soupir, Michael D. Radmacher, Phaedra Agius, Vineeth Sukrithan, Aakrosh Ratan, Martin D. McCarter, Bodour Salhia, Howard Colman, Abdul Rafeh Naqash, Igor Puzanov, Susanne M. Arnold, Julian A. Marin‐Acevedo, Shridar Ganesan, Michelle L. Churchman, Patrick Hwu, Paulo C. Rodrı́guez, William S. Dalton, George J. Weiner, Ahmad A. Tarhini · 发表于：Nature Communications · 年份：2025 · DOI：10.1038/s41467-025-65016-3 · 被引用次数：6 · 研究领域：Cancer Immunotherapy and Biomarkers、Ferroptosis and cancer prognosis、Melanoma and MAPK Pathways

Immune checkpoint inhibitors (ICIs) have transformed cancer treatment, yet predicting patient response remains a major challenge. Carcinoma ecotypes, which capture the cancer-immune interactions, show promise as prognostic biomarkers but remain untested in real-world settings. We compile and analyze the ORIEN Avatar ICI cohort of 1610 patients with matched gene expression data from a broader dataset of 14,997 individuals. Using EcoTyper-based immunophenotyping, we define ecotypes and assess their prognostic value across cancers, with a focused analysis in melanoma. Distinct cell states and ecotypes are consistently associated with survival outcomes across cancer types. We further develop a melanoma-specific ICI predictive model and validate it using data from the phase III ECOG-ACRIN E1609 trial as well as in external harmonized melanoma datasets. Together, these findings establish an ecotype-based framework and provide real-world evidence for their translational utility as clinically actionable biomarkers with prognostic and predictive value to guide ICI therapy. Cellular state cooccurrence signatures, such as carcinoma ecotypes may serve as potential biomarkers of response to cancer immunotherapy, however, their clinical utility remains unexplored. Here, the authors analyse large real world immunotherapy cohorts and gene expression data and develop a predictive model for response.