Gene expression signature of Gleason score is associated with prostate cancer outcomes in a radical prostatectomy cohort

Abstract

Prostate cancer (PCa) is a leading cause of cancer-related mortality worldwide. Gleason score (GS) is one of the best predictors of PCa aggressiveness, but additional tumor biomarkers may improve its prognostic accuracy. We developed a gene expression signature of GS to enhance the prediction of PCa outcomes. Elastic net was used to construct a gene expression signature by contrasting GS 8-10 vs. ≤6 tumors in The Cancer Genome Atlas (TCGA) dataset. The constructed signature was then evaluated for its ability to predict recurrence and metastatic-lethal (ML) progression in a Fred Hutchinson (FH) patient cohort (N=408; NRecurrence=109; NMLprogression=27). The expression signature included transcripts representing 49 genes. In the FH cohort, a 25% increase in the signature was associated with a hazard ratio (HR) of 1.51 (P=2.7×10-5) for recurrence. The signature's area under the curve (AUC) for predicting recurrence and ML progression was 0.68 and 0.76, respectively. Compared to a model with age at diagnosis, pathological stage and GS, the gene expression signature improved the AUC for recurrence (3%) and ML progression (6%). Higher levels of the signature were associated with increased expression of genes in cell cycle-related pathways and decreased expression of genes in androgen response, estrogen response, oxidative phosphorylation, and apoptosis. This gene expression signature based on GS may improve the prediction of recurrence as well as ML progression in PCa patients after radical prostatectomy.

Keywords: Gleason score; gene expression; metastasis; prostate cancer; recurrence.

Figure 1. Gene expression signature of Gleason score in The Cancer Genome Atlas (TCGA) dataset and the Fred Hutchinson (FH) cohort

(a) Box plots of the gene expression signature for patients in different Gleason score categories in TCGA. (b) Box plots of the gene expression signature for patients in different Gleason score categories in the FH cohort.

Figure 2. Gene expression signature of Gleason score and prostate cancer recurrence in the Fred Hutchinson cohort

(a) Kaplan-Meier curves of recurrence-free survival by quartiles (Q1-4) of the gene expression signature. The vertical dashed line shows the recurrence-free survival rate at 10 years after diagnosis. Shown in parentheses are the number of events/number of patients. (b) Area under the curve for prediction of recurrence for the gene expression signature alone (black) and for clinicopathological factors (age at diagnosis, Gleason score and pathological stage) with (red) and without (blue) the gene expression signature. (c-d) Same analyses as in Figure 2a and b for patients with Gleason score 7 tumors.

Figure 3. Area under the curve for prediction of metastatic-lethal prostate cancer for the gene expression signature alone (black) and for clinicopathological factors (age at diagnosis, Gleason score and pathological stage) with (red) and without (blue) the gene expression signature