Robust prediction of glioma prognosis by hypoxia-induced ferroptosis genes: VEGFA-XBP1 co-expression for salvage therapy

Abstract

Hypoxia as a hallmark of solid malignancies compromises therapeutic efficacy and prognosis. This study deciphers the functional role of hypoxia-induced ferroptosis in glioma prognosis. Hypoxia-related transcripts and ferroptosis markers were curated from public databases. ConsensusClusterPlus identify hypoxia-based molecular subtypes, while LASSO-penalized Cox regression integrated with limma-based differential expression analysis screened prognostic ferroptosis genes. Subsequent risk modeling was validated against clinical parameters and extended through nomogram construction. Protein-protein interaction networks centered on HIF-1αidentified high-confidence interactors, with parallel immune correlation analysis completing the systems-level investigation.Based on 27 hypoxia-associated genes, we stratified samples into three distinct hypoxic clusters. Differential analysis of 123 ferroptosis markers across clusters, combined with univariate Cox regression and LASSO regression, identified 23 hypoxia-induced ferroptosis genes for constructing a prognostic model. The model demonstrated robust predictive accuracy with AUC values of 0.80 (1-year), 0.86 (3-year), and 0.86 (5-year). GSEA revealed significant enrichment in ECM-receptor interactions, focal adhesion, JAK-STAT signaling, and p53 signaling pathways, suggesting their involvement in hypoxia-induced ferroptosis. Our risk model significantly outperformed conventional clinical parameters (pathology, grade, age, primary/recurrent status). Protein-protein interaction analysis incorporating HIF-1αand the 23-model genes identified XBP1 and VEGFA co-expression as significant positive prognostic factor. The immune infiltration analysis further indicated that M0 macrophages may participate in the regulation of the prognosis of VEGFA-XBP1.Hypoxia-induced ferroptosis modulation emerges as a prognostic factor in gliomas, with XBP1 and VEGFA representing druggable nodes for novel combination therapies.

Keywords: Hypoxia; ferroptosis; glioma; immunity; prognosis.

Figure 1.

(a) Clustering based on hypoxia genes; (b) Hypoxia gene expression in response to different clinical conditions; (c) Prognostic differences between clusters; (d) Differences in hypoxia gene expression between the three clusters. All genes exhibited three expression gradients (high→low). Expression patterns aligned as follows: EIF2B3, EIF2B4, EIF2B5, HRAs, PRKCB, EIF1: cluster 1→Cluster 3 →cluster 2; EIF1AX, EIF2B1, EIF2B2, EIF2S1, EIF2S2, EIF2S3, ELAVL1, FLT4, HIF-1α, KDR, NOS3, PIK3CA, PLCG1, PRKCA, SHC1, VHL, ARNT: cluster 3→Cluster 1→Cluster 2; FLT1, PIK3CG, VEGFA: cluster 3→Cluster 2→Cluster 1.

Figure 2.

(a-c) Univariate prognostic analysis and Lasso regression; (d-e) Association between risk score and prognosis; (f) Survival curve based on the median value of the risk score; (g) The ROC curve assessing the predictive efficacy of the risk model; (h) Principal component analysis(PCA) to distinguish high- and low-risk groups.

Figure 3.

(a-c) The risk model in different grades of glioma; (d) Expression of risk scores in different glioma pathological types; (e) ECM receptor interaction, FOCAL adhesion, JAK-STAT, and P53 signaling pathway enhancement; (f) Expression of 23 genes in response to different clinical conditions.

Figure 4.

(a-b) Clinical risk assessment based on univariate and multivariate Cox analysis; (c) Comparison of the risk model to other clinical risk factors; (d) The AUC of 1, 3, and 5-year survival in the nomogram; (e) Calibration chart of 1, 3, and 5 year survival; (f) Nomogram.

Figure 5.

(a) PPI based on 23 prognostic ferroptosis and HIF-1α; (b) Expression of VEGFA and XBP1 in each grade of glioma; (c-e) correlation between VEGFA, XBP1, and HIF-1α; (f-g) Impact of XBP1 and VEGFA co-expression on glioma prognosis in the training and validation groups.

Figure 6.

(a) Immune cell visualization of each sample; (b) The differential immune cell types in clusters 1 and 3; (d-g) Correlation between naïve B cells, memory B cells, monocytes, and M0 macrophages and glioma prognosis; (c) expression of four immune genes in different grades of glioma.

Figure 7.

Correlation between M0 macrophages and VEGFA and XBP1 gene expression.