Computational identification of immune-related lncRNA signature for predicting the prognosis and immune landscape of human glioblastoma multiforme

Abstract

Emerging evidence shows immune-related long noncoding RNAs (ir-lncRNAs) perform critical roles in tumor progression and prognosis assessment. However, the identification of ir-lncRNAs and their clinical significance in human glioblastoma multiforme (GBM) remain largely unexplored. Here, a designed computational frame based on immune score was used to identify differentially expressed ir-lncRNAs (DEir-lncRNAs) from The Cancer Genome Atlas (TCGA) GBM program. The immune-related lncRNA signature (IRLncSig) composed of prognosis-related DEir-lncRNAs selected by Cox regression analysis and its clinical predictive values were verified, which was further validated by another dataset from the Gene Expression Omnibus database (GEO). Subsequently, the association between IRLncSig and immune cell infiltration, immune checkpoint inhibitor (ICI) biomarkers, O6-methylguanine-DNA methyltransferase (MGMT) gene expression, and biological function were also analyzed. After calculation, five prognosis-related ir-lncRNAs were included in the establishment of IRLncSig. The risk assessment based on IRLncSig indicated that the high-IRLncSig-score group was significantly associated with poor prognosis (p < 0.001), significant aggregation of macrophages (p < 0.05), higher ICI biomarker expression, and MGMT gene expression (p < 0.05). Signature-related lncRNAs may be involved in immune activities in the tumorigenesis and progression of GBM. In summary, the novel IRLncSig shows a promising clinical value in predicting the prognosis and immune landscape of GBM.

Keywords: effectiveness; glioblastoma multiforme; immune checkpoint inhibitors; immune-related long noncoding RNAs; prognosis.

Figure 1

Study flow chart.

Figure 2

Construction and evaluation of the prognostic model. (A) The prognostic analysis of patients in the low-IRLncSig-score and high-IRLncSig-score groups by the Kaplan–Meier test of TCGA patients. (B) The Wilcoxon test of IRLncSig scores between the long-term survival group (overall survival >36 months) and short-term survival group (overall survival ≤36 months) of GEO patients. (C) The univariate and multivariate Cox regression prognostic analysis. (D) Time-dependent ROC curve analysis of IRLncSig and the other three features at 1 year. (E) Time-dependent ROC curve analysis of IRLncSig and the other three features at 2 years. (F) Time-dependent ROC curve analysis of IRLncSig and the other three features at 3 years. IRLncSig, immune-related lncRNA signature; TCGA, The Cancer Genome Atlas; GEO, Gene Expression Omnibus database; time-dependent ROC, time-dependent receiver operating characteristic.

Figure 3

Estimation of the correlation between tumor-infiltrating cells and the risk assessment model. Wilcoxon signed-rank analysis to calculate the infiltration difference of immune cells between low- and high-IRLncSig-scores in different algorithms [(A) MCP-counter; (B) EPIC; (C) CIBERSORT-ABS; (D) xCELL; (E) TIMER; (F) quanTlseq]. IRLncSig, immune-related lncRNA signature.

Figure 4

The correlation between the ICI biomarkers, the MGMT gene, TMB, and the risk assessment model. The high-IRLncSig-score group had significantly higher expression of (A) CTLA4, (B) HAVCR2, and (C) PD-1 but not (D) LAG3. (E) PD-L1 expression was higher in the High-IRLncSig-score group than in the low-IRLncSig-score group. (F, G) TMB does not have significant differences between the two groups and is lower than 10/MB in most cases. (H) The low-IRLncSig-score group was associated with lower expression of the MGMT gene. ICI biomarkers, immune checkpoint inhibitor biomarkers; MGMT, O6-methylguanine-DNA methyltransferase; TMB, tumor mutation burden; IRLncSig, immune-related lncRNA signature; CTLA4, cytotoxic T-lymphocyte-associated antigen-4; HAVCR2, hepatitis A virus cellular receptor 2; PD-1, antiprogrammed cell death protein-1; LAG3, lymphocyte activation gene 3 protein; PD-L1, programmed death ligand-1.

Figure 5

Identification of modules associated with the IRLncSig score and play enrichment analysis of ir-lncRNAs enrolled in IRLncSig. (A) The dendrogram of all genes is clustered based on the WGCNA algorithm. The color shows different clustering modules. (B) Heatmap of the correlation between the module eigengenes and clinical and molecular traits of GBM. (C) Correlation analysis of module membership in the selected module and IRLncSig score. (D) KEGG enrichment analysis is based on the genes in the selected module. (E–G) GO enrichment analysis is based on the genes in the selected module (molecular function (MF, Figure 6E), biological process (BP, Figure 6F), and cellular component (CC, Figure 6G)). IRLncSig, immune-related lncRNA signature; ir-lncRNAs, immune-related long noncoding RNAs; GBM, glioblastoma multiforme; KEGG, Kyoto Encyclopedia of Genes and Genomes, GO, Gene Ontology.