Identification of a prognostic biomarker predicting biochemical recurrence and construction of a novel nomogram for prostate cancer

Abstract

Background: Biochemical recurrence (BCR) is common in prostate cancer (PCa), but its prediction is based predominantly on clinicopathological characteristics with low accuracy. We intend to identify a potential prognostic biomarker related to the BCR and construct a nomogram for improving the risk stratification of PCa patients.

Methods: The transcriptome and clinical data of PCa patients were obtained from TCGA and GEO databases. Differential expression analysis and weighted gene co-expression network analysis (WGCNA) were used to screen out differentially expressed genes (DEGs) related to the BCR of PCa. Cox regression analysis was further applied to screen out DEGs related to BCR-free survival (BFS). Time-dependent receiver operating curve (ROC) analysis and Kaplan-Meier (K-M) survival analysis were conducted to assess the prognostic value. Then, a prognostic nomogram was established and evaluated. The clinicopathological correlation analysis, GSEA analysis, and immune analysis were used to explore the biological and clinical significance of the biomarker. Finally, the qRT-PCR, western blotting, and immunohistochemistry (IHC) were conducted to validate the expression of the biomarker.

Results: BIRC5 was identified to be the potential prognostic biomarker. The clinical correlation analysis and K-M survival analysis found that the BIRC5 mRNA expression was positively associated with disease progression and negatively associated with the BFS rate. Time-dependent ROC curves verified its accurate prediction performance. The GSEA and immune analysis suggested that the BIRC5 was related to immunity. A nomogram with an accurate prediction for BFS of PCa patients was constructed. qRT-PCR, western blotting, and IHC results validated the expression level of BIRC5 in PCa cells and tissues.

Conclusion: Our study identified BIRC5 as a potential prognostic biomarker related to BCR of PCa and constructed an efficacy nomogram for predicting BFS to assist clinical decision-making.

Keywords: biochemical recurrence; biomarker; nomogram; prognosis; prostate cancer.

Figure 1

The flowchart of our study.

Figure 2

Differentially expressed genes (DEGs) between prostate cancer (PCa) and normal prostate tissue. The volcano plots for DEGs in the TCGA-PRAD dataset (A) and the GSE46602 dataset (B). The significant up-regulated genes are indicated by red dots; The significant down-regulated genes are indicated by green dots; and the nonsignificant genes are indicated by blue dots.

Figure 3

Identification of co-expressed modules related to biochemical recurrence (BCR). A cluster tree of PCa samples with a color band displaying clinicopathologic values beneath the tree in the TCGA-PRAD dataset (A) and GSE46602 dataset (D). The cluster dendrogram indicated the different gene modules in the TCGA-PRAD dataset (B) and GSE46602 dataset (E). The correlation coefficients between the modules and clinical features in the TCGA-PRAD dataset (C) and GSE46602 dataset (F).

Figure 4

Differentially expressed genes (DEGs) correlated with biochemical recurrence (BCR). (A, B) Gene correlation scatter plots of the red module in the TCGA-PRAD dataset (A) and the turquoise module of the GSE46602 dataset (B). Module membership is represented by the X-axis, and the significance of the gene is represented by the Y-axis. (C) Venn diagrams of DEGs and co-expressed genes of TCGA-PRAD dataset and GSE46602 dataset.

Figure 5

Identification of the final key gene. (A, B) The univariate Cox regression analysis in the TCGA-PRAD dataset (A) and the GSE46602 dataset (B). (C) Venn diagram of prognostic-related genes from the GSE46602 dataset and the TCGA-PRAD dataset.

Figure 6

Prognostic evaluation of BIRC5 expression in various datasets. The K-M survival analysis indicated that patients with low expression of BIRC5 had a better BFS than those with high expression of BIRC5 in TCGA-PRAD dataset (A), GSE46602 dataset (B) and GSE70770 dataset (C). Time-dependent ROC curves demonstrated that the mRNA expression of BIRC5 acted as a powerful predictor of BFS for prostate cancer (PCa) patients in the TCGA-PRAD dataset (D), GSE46602 dataset (E) and GSE70770 dataset (F).

Figure 7

The correlation analysis between the mRNA expression of BIRC5 and different clinical characteristics. (A) Clinical T-stage. (B) Gleason score. (C) pathological T-stage. (D) pathological N-stage. (E) biochemical recurrence (BCR) status. (F) Age. (G) PSA.

Figure 8

Gene set enrichment analysis (GSEA) and immune analysis. (A, B) GSEA pathway enrichment analysis of BIRC5 in prostate cancer (PCa) patients. (C) Immune cell infiltration in high- and low-expression of BIRC5 groups. (D) Correlation analysis between the mRNA expression of BIRC5 and immune cells infiltrated. (E) The correlation analysis between the BIRC5 expression and immune checkpoints. (F) The differential analysis of Tumor Immune Dysfunction and Exclusion (TIDE) scores between the low- and high-BIRC5 expression of PCa. (^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001).

Figure 9

Establishment of a nomogram to predict biochemical recurrence-free survival (BFS) in PCa patients. (A) The univariate Cox regression analysis of clinicopathological characteristics and the mRNA expression of BIRC5. (B, C) The LASSO regression analysis identified the final prognostic factors. (D) A prognostic nomogram predicting 1-, 3-, and 5-years BFS of PCa patients. cT (clinical T-stage):1-T1, 2-T2, 3-T3, 4-T4. (E) The K-M survival analysis showed that patients with high risk had significantly shorter BFS than those with low risk. (F) The time-dependent ROC curves demonstrated the accurate prediction performance of the nomogram. (G–I) The calibration plots for 1-, 3-, and 5-year BFS of PCa patients.

Figure 10

Validation of the mRNA and protein expression of BIRC5 in prostate cancer (PCa). (A) The mRNA expression of BIRC5 in PCa tissues was significantly higher than in normal tissues in the UALCAN database. (B) The mRNA expression of BIRC5 in PCa cells (DU145 and PC3) was significantly increased than in the prostate normal epithelium cells (RWPE1). (C) The protein expression of BIRC5 in PCa cells (DU145 and PC3) was significantly lower than that in RWPE1 cells. (D, E) The two Immunohistochemistry (IHC) images from the same patient. The normal prostate tissue (400X, bar = 20 um) (D), the PCa tissue (400X, bar =20 um) (E). (F) IHC analysis demonstrated that the BIRC5 staining in the normal prostate tissues was stronger than the paired PCa tissues. (G) The PCa tissue with a Gleason score 6 (400X, bar = 20 um). (H) The PCa tissue with Gleason score 9 (400X, bar = 20 um). (I) The staining of BIRC5 in PCa tissues with a Gleason score ≤7 was stronger than PCa tissues with a Gleason score >7 (^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001).