Identification of an immune-related long non-coding RNA signature and nomogram as prognostic target for muscle-invasive bladder cancer

Abstract

To identify an immune-related prognostic signature based on long non-coding RNAs (lncRNAs) and find immunotherapeutic targets for bladder urothelial carcinoma, we downloaded RNA-sequencing data from The Cancer Genome Atlas (TCGA) dataset. Functional enrichment analysis demonstrated bladder urothelial carcinoma was related to immune-related functions. We obtained 332 immune-related genes and 262 lncRNAs targeting immune-related genes. We constructed a signature based on eight lncRNAs in training cohort. Patients were classified as high-risk and low-risk according to signature risk score. High-risk patients had poor overall survival compared with low-risk patients (P < 0.001). Multivariate Cox regression suggested the signature was an independent prognostic indicator. The findings were further validated in testing, entire TCGA and external validation cohorts. Gene set enrichment analysis indicated significant enrichment of immune-related phenotype in high-risk group. Immunohistochemistry and online analyses validated the functions of 4 key immune-related genes (LIG1, TBX1, CTSG and CXCL12) in bladder urothelial carcinoma. Nomogram proved to be a good classifier for muscle-invasive bladder cancer through combining the signature. In conclusion, our immune-related prognostic signature and nomogram provided prognostic indicators and potential immunotherapeutic targets for muscle-invasive bladder cancer.

Keywords: immune-related; lncRNA signature; muscle-invasive bladder cancer; nomogram; prognostic model.

Figure 1

Workflow of this study. The study was carried out in TCGA (The Cancer Genome Atlas) BLCA (Bladder Urothelial Carcinoma) dataset. Immune-related genes were extracted from Molecular Signatures Database v4.0. LncRNAs targeting immune-related genes were identified according to Pearson correlation. DEGs (differentially expressed genes) were calculated between BCa (bladder urothelial carcinoma) samples and normal bladder samples in TCGA BLCA dataset. The training cohort was used to identify the lncRNAs targeting immune-related genes and establish a prognostic signature based on the prognostic lncRNAs. The prognosis analysis was validated in the testing cohort, entire TCGA BLCA cohort and Tianjin validation cohort, respectively. Nomogram was constructed by including the immune-related signature and other prognosis-related clinical features in training cohort. Immunohistochemistry from THPA (The Human Protein Atlas) and online analyses from GEPIA (Gene Expression Profiling Interactive Analysis) were used to validate four key immune-related genes (CTSG, CXCL12, LIG1 and TBX1). Functional enrichment analyses were utilized to explore immune-related functions.

Figure 2

Identification of differentially expressed genes (DEGs) and immune-related DEGs in TCGA (The Cancer Genome Atlas) BLCA (Bladder Urothelial Carcinoma) dataset. (A) Volcano plot of all DEGs; (B) Volcano plot of immune-related DEGs; (C) Heat map for immune-related DEGs.

Figure 3

GO (Gene Ontology) functional and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analyses of differentially expressed genes (DEGs) and immune-related DEGs. (A) Top 10 BP (biological process) terms of all DEGs; (B) Top 10 KEGG pathways of all DEGs; (C) Top 10 BP terms of immune-related DEGs; (D) Top 10 KEGG pathways of immune-related DEGs; (E) Venn diagram for overlapped BP terms; (F) Venn diagram for overlapped CC (cell component) terms; (G) Venn diagram for overlapped MF (molecular function) terms; (H) Venn diagram for overlapped MF (molecular function) KEGG pathways.

Figure 4

Evaluating the predictive power of the 8-lncRNA immune-related signature in the training cohort. (A–C) Distribution of risk score, survival status, and lncRNA expression of patients in the training cohort; (D) Kaplan-Meier survival curve of the high-risk and low-risk groups in the training cohort; (E) Time-dependent ROC (receiver operating characteristic) curves and AUC (area under curve) based on the training cohort for 5-year overall survival; (F) Forest plot for multivariate Cox regression analysis.

Figure 5

Evaluating the predictive power of the 8-lncRNA immune-related signature in the testing cohort. (A–C) Distribution of risk score, survival status, and lncRNA expression of patients in the testing cohort; (D) Kaplan-Meier survival curve of the high-risk and low-risk groups in the testing cohort; (E) Time-dependent ROC (receiver operating characteristic) curves and AUC (area under curve) based on the testing cohort for 5-year overall survival; (F) Forest plot for multivariate Cox regression analysis.

Figure 6

Evaluating the predictive power of the 8-lncRNA immune-related signature in the entire TCGA (The Cancer Genome Atlas) BLCA (Bladder Urothelial Carcinoma) cohort. (A–C) Distribution of risk score, survival status, and lncRNA expression of patients in the entire TCGA BLCA cohort; (D) Kaplan-Meier survival curve of the high-risk and low-risk groups in the entire TCGA BLCA cohort; (E) Time-dependent ROC (receiver operating characteristic) curves and AUC (area under curve) based on the entire TCGA BLCA cohort for 5-year overall survival; (F) Forest plot for multivariate Cox regression analysis.

Figure 7

Stratified survival analyses and Clinical characteristics with 8-lncRNA prognostic signature in the entire TCGA (The Cancer Genome Atlas) BLCA (Bladder Urothelial Carcinoma) cohort. (A–G) Kaplan-Meier survival curves in subgroups stratified by different clinical characteristics; (H) Distribution of clinicopathologic features, and lncRNA expression in low-risk and high-risk groups; (I) Risk score comparison between alive and dead patients; (J) Risk score comparison between different tumor stages. *** P-value < 0.005.

Figure 8

PCA (Principal components analysis) and GSEA (Gene set enrichment analysis). PCA based on the eight lncRNAs indicated low-risk and high-risk groups were generally distributed in two different directions in (A) the training cohort and (B) the testing cohort, respectively; (C) PCA based on the whole gene set indicated these two groups did not show significant distinctions; (D, E) GSEA indicated significant enrichment of immune-related phenotype in the high-risk group patients. FDR false discovery rate; NES normalized enrichment score.

Figure 9

Immunohistochemistry from THPA (The Human Protein Atlas) and online analyses from GEPIA (Gene Expression Profiling Interactive Analysis) were used to explore four key immune-related genes (LIG1, TBX1, CTSG and CXCL12). (A) Immunohistochemistry between normal bladder tissues and BCa (bladder urothelial carcinoma) tissues; (B) Gene expression level between normal bladder tissues and BCa tissues for LIG1, TBX1, CTSG and CXCL12, respectively; (C) Kaplan-Meier survival curve of LIG1, TBX1, CTSG and CXCL12, respectively. * P-value < 0.05.

Figure 10

Building and validating the nomogram to predict prognosis in TCGA (The Cancer Genome Atlas) BLCA (Bladder Urothelial Carcinoma) dataset. (A) The nomogram was constructed based on age, UICC (Union for International Cancer Control) stage, histological grade and the immune-related signature in the training cohort; (B) The calibration plot for internal validation of the nomogram; (C–E) Kaplan-Meier survival curves between high-nomogram-score and low-nomogram-score groups in the training cohort, testing cohort and entire TCGA BLCA cohort, respectively; (F–H) Time-dependent ROC (receiver operating characteristic) curves and AUC (area under curve) for 1-year, 3-year, and 5-year overall survival based on the training cohort, testing cohort and entire TCGA BLCA cohort, respectively.

Figure 11

External validation of 8-lncRNA immune-related signature and nomogram in Tianjin cohort. (A) Kaplan-Meier survival curve of 8-lncRNA immune-related signature in Tianjin validation cohort; (B) Time-dependent ROC (receiver operating characteristic) curves and AUC (area under curve) of 8-lncRNA immune-related signature based on Tianjin validation cohort for 3-year overall survival; (C) Forest plot for multivariate Cox regression analysis; (D) Kaplan-Meier survival curve of nomogram in Tianjin validation cohort.