A novel chemokine-based signature for prediction of prognosis and therapeutic response in glioma

Gliomas are the primary malignant brain tumor and characterized as the striking cellular heterogeneity and intricate tumor microenvironment (TME), where chemokines regulate immune cell trafficking by shaping local networks. This study aimed to construct a chemokine-based gene signature to evaluate the prognosis and therapeutic response in glioma.

In all, our results highlighted a potential 4-chemokine signature for predicting prognosis in glioma and reflected the intricate immune landscape in glioma. It also threw light on integrating tailored risk stratification with precision therapy for glioblastoma.

In this study, 1024 patients (699 from TCGA and 325 from CGGA database) with clinicopathological information and mRNA sequencing data were enrolled. A chemokine gene signature was constructed by combining LASSO and SVM-RFE algorithm. GO, KEGG, and GSVA analyses were performed for function annotations of the chemokine signature. Candidate mRNAs were subsequently verified through qRT-PCR in an independent cohort including 28 glioma samples. Then, through immunohistochemical staining (IHC), we detected the expression of immunosuppressive markers and explore the role of this gene signature in immunotherapy for glioma. Lastly, the Genomics of Drug Sensitivity in Cancer (GDSC) were leveraged to predict the potential drug related to the gene signature in glioma.

A constructed chemokine gene signature was significantly associated with poorer survival, especially in glioblastoma, IDH wildtype. It also played an independent prognostic factor in both datasets. Moreover, biological function annotations of the predictive signature indicated the gene signature was positively associated with immune-relevant pathways, and the immunosuppressive protein expressions (PD-L1, IBA1, TMEM119, CD68, CSF1R, and TGFB1) were enriched in the high-risk group. In an immunotherapy of glioblastoma cohort, we confirmed the chemokine signature showed a good predictor for patients' response. Lastly, we predicted twelve potential agents for glioma patients with higher riskscore.