Gene Expression Analysis Reveals Prognostic Biomarkers of the Tyrosine Metabolism Reprogramming Pathway for Prostate Cancer

Abstract

Background: Tyrosine metabolism pathway-related genes were related to prostate cancer progression, which may be used as potential prognostic markers.

Aims: To dissect the dysregulation of tyrosine metabolism in prostate cancer and build a prognostic signature based on tyrosine metabolism-related genes for prostate cancer. Materials and Method. Cross-platform gene expression data of prostate cancer cohorts were collected from both The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). Based on the expression of tyrosine metabolism-related enzymes (TMREs), an unsupervised consensus clustering method was used to classify prostate cancer patients into different molecular subtypes. We employed the least absolute shrinkage and selection operator (LASSO) Cox regression analysis to evaluate prognostic characteristics based on TMREs to obtain a prognostic effect. The nomogram model was established and used to synthesize molecular subtypes, prognostic characteristics, and clinicopathological features. Kaplan-Meier plots and logrank analysis were used to clarify survival differences between subtypes.

Results: Based on the hierarchical clustering method and the expression profiles of TMREs, prostate cancer samples were assigned into two subgroups (S1, subgroup 1; S2, subgroup 2), and the Kaplan-Meier plot and logrank analysis showed distinct survival outcomes between S1 and S2 subgroups. We further established a four-gene-based prognostic signature, and both in-group testing dataset and out-group testing dataset indicated the robustness of this model. By combining the four gene-based signatures and clinicopathological features, the nomogram model achieved better survival outcomes than any single classifier. Interestingly, we found that immune-related pathways were significantly concentrated on S1-upregulated genes, and the abundance of memory B cells, CD4+ resting memory T cells, M0 macrophages, resting dendritic cells, and resting mast cells were significantly different between S1 and S2 subgroups.

Conclusions: Our results indicate the prognostic value of genes related to tyrosine metabolism in prostate cancer and provide inspiration for treatment and prevention strategies.

Figure 1

Expression of tyrosine metabolism pathway-related genes in prostate cancer. (a) Heatmap showed the expression landscape of tyrosine metabolism-related genes between prostate cancer and normal prostatic tissues; (b) heatmap showed the expression landscape of tyrosine metabolism-related genes between metastatic prostate cancer and primary prostate cancer tissues; (c) Pearson analyses showed the correlation between genes encompassed in tyrosine metabolism pathways; (d) the regulation network and pathway enrichment of tyrosine metabolism pathway-related genes using GeneMANIA.

Figure 2

Mutational landscape of genes in the tyrosine metabolism pathway in prostate cancer.

Figure 3

Establishment of the molecular subgroup of prostate cancer. (a) The unsupervised hierarchical clustering method established the molecular subgroup of prostate cancer based on the expression of 19 tyrosine metabolism-related enzymes; (b) principal component analysis (PCA) showed the clustering of two molecular subgroups; (c) Kaplan–Meier plot and logrank analysis showed the survival difference between S1 and S2 subgroups; (d) univariate Cox regression analysis revealed the prognostic role of tyrosine metabolism-related genes in prostate cancer; (e) the differentially expressed genes between S2 and S1 subgroups; (f) GSEA revealed the difference between S2 and S1 subgroups at the pathway level.

Figure 4

The frequently mutated genes between S1 and S2 subgroups. (a) Frequently mutated genes in the S1 group of TCGA prostate cancer. (b) Frequently mutated genes in the S2 group of TCGA prostate cancer. (c) The forest plot shows the differentially mutated genes between S1 and S2. (d) Mutations of SPOP in S1 (up) and S2 (down) groups, respectively.

Figure 5

Establishment of the signature based on tyrosine metabolism-related genes. (a) Partial likelihood deviance of different numbers of variables; (b) the LASSO coefficient profiles of the selected four features; (c) the heatmap plot showed the distribution of four selected genes between different clinicopathological features; (d) the Kaplan–Meier plot and log-rank analysis showed the survival outcome difference between low- and high-risk subgroups in the TCGA training cohort; (e) the ROC curve showed the predictive value of the four gene-based signatures in the training cohort; (f) the Kaplan-Meier plot and logrank analysis showed the survival outcome difference between low- and high-risk subgroups in the TCGA validation cohort; (g) the ROC curve showed the predictive value of the four gene-based signatures in the validation cohort.

Figure 6

External validation of the established four tyrosine metabolism-related genes in MSKCC (a) and GSE116918 (b) cohorts.

Figure 7

Nomogram model establishment. (a) The establishment of a nomogram model by integrating prostate cancer patients' age, pathological grade, T, N, M stages, and a risk score of four-gene models. An unsupervised hierarchical clustering method established the molecular subgroup; (b–d) the calibration curve for 1-year, 3-year, and 5-year PFS from the prognostic nomogram model.

Figure 8

A boxplot presented compositional differences of 22 immunocytes between the low-risk group (S1) and the high-risk group (S2).

Figure 9

The correlation between the risk score derived from the nomogram model and ratios of infiltrated immune cells.