Abstract:

Objective To establish a prognostic model based on differentially expressed genes （DEGs） and differentially expressed long non-coding RNAs（DELs） of the recurrent glioblastoma （GBM） patients using the Chinese Glioma Genome Atlas （CGGA） database， to discuss the relationships between the DEGs， DELs and clinicopathological features， and to verify the findings through experiments. Methods The transcriptome data and corresponding clinical data of GBM patients， including 203 patients with primary GBM （primary GBM group） and 103 patients with recurrent GBM （recurrent GBM group）， were downloaded from the CGGA database. A Perl script was used to merge the expression profiles across all samples into a matrix， which was annotated using the Ensembl database. R packages edgeR and DESeq2 were used to filter out low-expression transcripts and to extract and normalize the data. The expression differences in transcripts between primary GBM group and recurrent GBM group were compared， the fold change（FC） was calculated， and the DEGs and DELs were screened with the criteria of |log₁₀FC|>0.5 and P<0.05. A univariate Cox proportional hazards regression model was used to analyze the relationship between DEGs and DELs and the prognosis of GBM patients. The DEGs and DELs associated with prognosis were selected to construct a multivariate Cox proportional hazards regression model， and the risk scores were calculated for all patients. Based on the median risk score， the patients were divided into low-risk and high-risk groups. The Kaplan-Meier survival analysis and receiver operating characteristic （ROC） curve were used to evaluate the predictive efficacy of the risk score for overall survival （OS）in the patients， and the differences in risk scores in the patients with different clinicopathological features were analyzed and compared. Lasso regression analysis was used to evaluate the predictive efficacies of the prognosis-related DEGs and DELs for patient survival， and a nomogram was constructed. The tumor tissue samples from 6 primary GBM patients and 6 recurrent GBM patients were selected as primary group and recurrent group， respectively. Real-time fluorescence quantitative PCR （RT-qPCR） method and fluorescence in situ hybridization （FISH） method were used to detect the expression levels of RP11-283I3.2 and the numbers of positive cells in tumor tissues of the patients in two groups. Results A total of 151 DEGs and DELs were identified between patients in primary GBM groupand recurrent GBM group，among which 12 were downregulated and 139 were upregulated. Univariate Cox regression analysis identified that 4 DEGs （CHRNA1， HIST1H4I， SLC18A3， and TMEM61） and 8 DELs （AC002454.1， AC006262.6， AC123886.2，CDKN2B-AS1，FAM225B，HAR1A，LBX1-AS1，and RP11-283I3.2） were significantly associated with the prognosis of GBM patients.The Kaplan-Meier survival analysis results showed that the OS of GBM patients in high-risk group was significantly lower than that in low-risk group （P<0.001）. The time-dependent ROC curve results showed that the areas under the curve （AUCs） of the risk score for predicting 1-， 3-， and 5-year survival probability were 0.675， 0.750， and 0.772， respectively. The risk score of GBM patients with wild-type isocitrate dehydrogenase （IDH） was significantly higher than that of patients with mutant IDH （P<0.001）. The Lasso regression analysis results indicated that RP11-283I3.2 had the highest score among the prognosis-related DEGs and DELs. Based on the expression profile， the patients were divided into Cluster 1 group and Cluster 2 group. Compared with Cluster 1 group， the patients in Cluster 2 group had lower OS （P<0.01）. The RT-qPCR method and FISH method results showed that compared with primary group， the expression level of RP11-283I3.2 and the number of positive cells in tumor tissue of the patients in recurrent group were significantly increased （P<0.001）. Conclusion A total of 12 DEGs and DELs associated with the prognosis of GBM patients were identified， and a prognostic model with good predictive efficacy was constructed. The high expression of RP11-283I3.2 in tumor tissues in recurrent GBM patients suggests its potential as a molecular biomarker for GBM recurrence.

Key words: Glioblastoma, Recurrence, Differentially expressed genes, Long non-coding RNA, Bioinformatics, Survival analysis