Transcriptome analysis reveals a long non-coding RNA signature to improve biochemical recurrence prediction in prostate cancer

Abstract

Despite highly successful treatments for localized prostate cancer (PCa), prognostic biomarkers are needed to improve patient management and prognosis. Accumulating evidence suggests that long noncoding RNAs (lncRNAs) are key regulators with biological and clinical significance. By transcriptome analysis, we identified a set of consistently dysregulated lncRNAs in PCa across different datasets and revealed an eight-lncRNA signature that significantly associated with the biochemical recurrence (BCR)-free survival. Based on the signature, patients could be classified into high- and low-risk groups with significantly different survival (HR = 2.19; 95% CI = 1.67-2.88; P < 0.0001). Validations in the validation cohorts and another independent cohort confirmed its prognostic value for recurrence prediction. Multivariable analysis showed that the signature was independent of common clinicopathological features and stratified analysis further revealed its role in elevating risk stratification of current prognostic models. Additionally, the eight-lncRNA signature was able to improve on the CAPRA-S score for the prediction of BCR as well as to reflect the metastatic potential of PCa. Functional characterization suggested that these lncRNAs which showed PCa-specific expression patterns may involve in critical processes in tumorigenesis. Overall, our results demonstrated potential application of lncRNAs as novel independent biomarkers. The eight-lncRNA signature may have clinical potential for facilitating further stratification of more aggressive patients who would benefit from adjuvant therapy.

Keywords: BCR-free survival; lncRNA signature; prognosis; prostate cancer; transcriptome.

Figure 1. Kaplan–Meier estimates of the BCR-free survival of PCa patients using the eight-lncRNA signature

(A) Training cohort (n = 184). (B) Validation cohort (n = 123). (C) Entire TCGA cohort (n = 307). (D) Independent cohort GSE21034 (n = 140). The differences between the two curves were determined by the two-side log-rank test.

Figure 2. LncRNA risk score analysis of the entire TCGA cohort

(A) The distribution of the eight-lncRNA risk score. (B) Patients’ BCR status and time. (C) Heatmap of the eight lncRNA expression profiles. The dotted line represents the median lncRNA risk score cutoff dividing patient into low-risk and high-risk groups.

Figure 3. Kaplan–Meier estimates of the BCR-free survival of the entire TCGA cohort stratified by Gleason score and positive lymph node

(A) Patients with Gleason score ≤ 7. (B) Patients with Gleason score ≥ 8. (C) Patients with positive LNs = 0. (D) Patients with positive LNs > 0.

Figure 4. Comparison of sensitivity and specificity for survival prediction by the eight-lncRNA signature, Gleason score and positive lymph node

(A) ROC curves of the eight-lncRNA signature, Gleason score and the combination of the two factors. (B) ROC curves of the eight-lncRNA signature, positive LN and their combination. P values showed the AUC of Gleason score versus the AUC of the combination of signature and Gleason score and the AUC of positive LN versus the AUC of the combination of signature and positive LN.

Figure 5. The eight-lncRNA signature significantly improved the prediction value of the CAPRA-S score

The prediction capability of the eight-lncRNA risk score, CAPRA-S score and combination of the two factors is evaluated by the ROC curve in GSE21034 cohort. P value showed the AUC of CAPRA-S score versus the AUC of the combination of signature and CAPRA-S score.

Figure 6. Functional characterization of the lncRNAs in the signature

(A) Gene set enrichment analysis (GSEA) delineates biological processes correlated with the lncRNA signature. Nodes represent enriched GO terms and an edge represents existing genes shared between connecting GO terms. Nodes are grouped and annotated by their similarity according to related gene sets. (B) Box plot comparing the expression of the eight lncRNAs across 12 cancer types. PCAT7, SLC12A9-AS1, RGMB-AS1, PCAT1, AC025265.1 and LINC00593 were highly expressed in prostate cancer across twelve cancers. BLCA, Bladder Urothelial Carcinoma; BRCA, Breast Invasive Carcinoma; HNSC, Head and Neck Squamous Cell Carcinoma; KIRC, Kidney Renal Clear Cell Carcinoma; KIRP, Kidney Renal Papillary Cell Carcinoma; LGG, Brain Lower Grade Glioma; LIHC, Liver Hepatocellular Carcinoma; LUAD, Lung Adenocarcinoma; LUSC, Lung Squamous Cell Carcinoma; PRAD, Prostate Adenocarcinoma; STAD, Stomach Adenocarcinoma; THCA, Thyroid Carcinoma.