Identification and validation of a novel eight mutant-derived long non-coding RNAs signature as a prognostic biomarker for genome instability in low-grade glioma

Abstract

Long non-coding RNAs (lncRNAs) comprise an integral part of the eukaryotic transcriptome. Alongside proteins, lncRNAs modulate lncRNA-based gene signatures of unstable transcripts, play a crucial role as antisense lncRNAs to control intracellular homeostasis and are implicated in tumorigenesis. However, the role of genomic instability-associated lncRNAs in low-grade gliomas (LGG) has not been fully explored. In this study, lncRNAs expression and somatic mutation profiles in low-grade glioma genome were used to identify eight novel mutant-derived genomic instability-associated lncRNAs including H19, FLG-AS1, AC091932.1, AC064875.1, AL138767.3, AC010273.2, AC131097.4 and ISX-AS1. Patients from the LGG gene mutagenome atlas were grouped into training and validation sets to test the performance of the signature. The genomic instability-associated lncRNAs signature (GILncSig) was then validated using multiple external cohorts. A total of 59 novel genomic instability-associated lncRNAs in LGG were used for least absolute shrinkage and selection operator (Lasso), single and multifactor Cox regression analysis using the training set. Furthermore, the independent predictive role of risk features in the training and validation sets were evaluated through survival analysis, receiver operating feature analysis and construction of a nomogram. Patients with IDH1 mutation status were grouped into two different risk groups based on the GILncSig score. The low-risk group showed a relatively higher rate of IDH1 mutations compared with patients in the high-risk group. Furthermore, patients in the low-risk group had better prognosis compared with patients in the high-risk group. In summary, this study reports a reliable prognostic prediction signature and provides a basis for further investigation of the role of lncRNAs on genomic instability. In addition, lncRNAs in the signature can be used as new targets for treatment of LGG.

Keywords: genomic instability-associated lncRNAs signature (GILncSig); long non-coding RNA (lncRNA); low-grade glioma (LGG); prognosis; risk score.

Figure 1

Study flow chart of genomic instability-related lncRNAs construction.

Figure 2

Identified and functionally interpreted genomic instability-associated lncRNAs in patients with low-grade gliomas. (A) An unsupervised clustering among 529 patients with low-grade glioma was performed based on the expression patterns of 59 candidate genomic instability-associated lncRNAs. The GS group is shown in blue on the left, whereas the GU group is shown in red on the right. (B) Box plots for somatic mutations of GS and GU groups. Cumulative somatic mutations in the GU group were significantly higher compared with those in the GS group. (C) Box plots of the expression levels of UBQLN4 in the GU and GS groups. Expression levels of UBQLN4 were significantly lower in the GU group compared with the levels in the GS group. (D) Pearson correlation coefficient analysis based genomic instability-associated lncRNA and mRNA co-expression network. (E–H) GO and KEGG functional enrichment analysis of lncRNA co-expression mRNA through.

Figure 3

LncRNAs signature of the genomic instability used to predict outcomes in the training set. (A) Lasso Cox analysis identified 16 lncRNAs associated with genomic instability that were highly associated with prognosis. (B) Determination of the optimal value of penalty parameters through 1000 replicates of cross-validation. (C) Kaplan-Meier estimation of GILncSig-predicted overall survival of low- or high-risk patients in the training set. (D) Time-dependent ROC curves of GILncSig at 1, 2 and 3 years. (E) Distribution of cumulative somatic mutations and expression of UBQLN4 in high- and low-risk groups in the GILncSig model of low-grade glioma patients. (F) Box plot for distribution of cumulative somatic mutations in the low- and high-risk groups of LGG patients. (G) Box plots for UBQLN4 gene expression in low- and high-risk groups of LGG patients.

Figure 4

Validation of the lncRNA signature for genomic instability used to predict outcomes in the testing and TCGA set. (A) Validation of overall survival in low- or high-risk patients predicted by pooling GILncSig with Kaplan-Meier estimates. (B) Time-dependent ROC curves of GILncSig at 1, 2 and 3 years in the testing group. (C) Verification of LncRNA expression patterns, the profile of somatic mutations and UBQLN4 expression in patients in low- and high-risk groups. (D–E) Box plots for the distribution of somatic mutations and UBQLN4 expression in high- and low-risk groups of patients. (F–J) Verification of the above results using the TCGA set.

Figure 5

(A–B) Kaplan-Meier survival curves analysis for the eight genomic instability-associated lncRNAs using the training set (A) and validation set (B) of patients with low-grade glioma. (C–E) Nomograms for the eight genomic instability-associated lncRNAs for each factor in the training set, predictions of patient survival at 1, 3, and 5 years. Nomograms were evaluated using calibration curves and DCA curves. (F–H) Plot nomogram plots in the validation set and evaluation of nomograms using calibration curves and DCA curves.

Figure 6

Correlation analysis of the 8 genomic instability-associated lncRNAs with infiltration of each subtype of immune cells. (A) Correlation coefficients for M0 macrophages, M1 macrophages, memory resting CD4 T cells, and CD8 T cells for the training set were 0.23, 0.26 0.27, 0.22 (p < 0.001), whereas the correlation coefficient for activated mast cells was R = –0.21 (p < 0.05) and the correlation coefficient for monocytes was R = –0.29 (p < 0.001). (B) The correlation coefficients among the risk scores for M1 macrophages, memory resting CD4 T cells, activated mast cells and monocytes for the validation set were 0.31, 0.29, –0.27 and –0.2, respectively (p < 0.05).

Figure 7

The correlation between GILngSig signature and 64 microenvironment infiltrating immune cells using xcell platform (A) training set (B) validation set.

Figure 8

Evaluation of the performance of the GILncSig partial gene using two external independent CGGA mRNA-seq-693 and GSE16011 datasets. (A–B) Box plots for gene expression levels of AC064875.1 and H19 for patients at different ages (< = 41 and > 41 years), tumor grade, IDH1 mutation status, 1p19q chromosome union deletion status, gender and MGMT methylation status in patients from the CGGA mRNA-seq-693 set. (C) Box plots for expression of AC131097.4 for patients of different ages in the CGGA mRNA-seq-693 set. (D) Box plots for gene expression levels of FLG-AS1 for patients of different tumor grades and gender in the CGGA mRNA-seq-693 set. (E). Box plots expression level of lncRNA H19 in patients with different tumor grades and IDH1(R132) mutation status in the GSE16011 dataset.

Figure 9

(A–C) ROC analysis of overall survival at 1-, 2- and 3- years for GILncSig, LilncSig, LilncSig and QiulncSig.

Figure 10

Verification that the GILncSig is an independent prognostic factor. (A–C) The result of univariate Cox regression showed that the age, grade, diagnostic types of gliomas and GILncSig signature were significant prognostic factors for LGG patients. (B–D) while only the factor of age, diagnostic types and GILncSig signature were also associated with overall survival in the multivariate Cox regression model, indicating that the GILncSig signature was the independent prognostic biomarker for predicting the survival of LGG patients.

Figure 11

Stratified analysis by age and tumor grade. (A–B) Kaplan-Meier curve analysis of OS in the high- and low-risk groups for patients in the two age groups. (< = 41 and >41 years). (C–D) Kaplan-Meier curve analysis for OS in high- and low-risk groups for Grade II and Grade III groups.

Figure 12

Correlation between GILncSig and IDH1 somatic mutations. (A–C) Proportion of IDH1 mutations in the high- and low-risk groups using the training set, testing set and TCGA set. (D) Kaplan-Meier curve analysis of OS of patients with IDH1 mutant status and wild-type status for the combined GS and GU groups.