Interleukin-19 as a novel target for predicting survival in glioblastoma patients using a radiogenomics approach

Immunotherapy has shown clinical benefit in several malignancies; however, its efficacy in glioma remains limited, partly due to the substantial heterogeneity of immunosuppressive mechanisms within the tumor microenvironment (TME). Tumor-infiltrating immune cells secrete diverse immunoregulatory cytokines that collectively shape tumor progression and influence clinical outcomes. Nevertheless, the integrated prognostic impact of these multiple cytokines in glioma has not been systematically characterized. Interleukin-19 (IL-19) has recently emerged as an immunosuppressive cytokine that attenuates T-cell activation and is associated with unfavorable prognosis in glioblastoma (GBM), highlighting its potential as a therapeutic target. In this study, we aim to delineate cytokine-expression patterns linked to patient survival and to develop a machine-learning–based radiogenomic model capable of distinguishing survival outcomes in patients with GBM according to IL19 expression levels within the TME. RNA-seq dataset from the TCGA glioma cohort (WHO grade II–IV) were analyzed using consensus clustering to define cytokine-associated immune subgroups. Immune cell enrichment was evaluated by GSEA using curated Panglao-derived marker sets. For radiogenomic modeling, histologically confirmed GBM (WHO grade IV) cases with available MRI data from TCIA and transcriptomic data (RNA-seq and microarray) from the TCGA cohort were included. Radiomic features were extracted from contrast-enhanced T1-weighted (T1 + C) and apparent diffusion coefficient (ADC) images. Bayesian logistic regression models with backward feature selection were constructed to predict IL19 expression and validated in independent GBM microarray and RNA-seq cohorts. Cytokine-based clustering identified four molecular subgroups with significantly different survival outcomes. Subgroups with the poorest prognosis showed marked enrichment of M2-like macrophage signatures and exhibited the highest IL19 expression levels. Consistently, IL19 was strongly associated with adverse prognosis and showed significantly increased expression in grade IV gliomas across both TCGA and CGGA datasets. Multiple radiomic features correlated with IL19 expression, predominantly within first-order intensity classes. Radiogenomic models achieved AUCs of 0.84–0.85 and accuracies of 0.77–0.79 for IL19 prediction, and independent validation showed that model-predicted IL19–high groups consistently exhibited significantly worse survival. IL-19 is a prognostic cytokine linked to immunosuppressive macrophage activity and glioma aggressiveness. Multiparametric MRI–based radiomic immunophenotyping enables the stratification of GBM patient survival according to intratumoral IL19 expression. This radioimmunogenomic framework provides a non-invasive approach for characterizing radiomic features associated with IL19 expression level and may serve as an exploratory platform for investigating immune heterogeneity in GBM.