https://orcid.org/0000-0001-6689-6135
Figures & data
Figure 1. Weighted gene co-expression network analysis (WGCNA).
(A) This study conducted a systematic review of the literature and integrative analysis of 10 bulk-RNA sequencing or microarray studies and three single-cell RNA sequencing; (B) with weighted gene co-expression network analysis, seventeen modules were identified in 206 cases; (C) the expression of the above-clustered genes were highly correlated; (D) the cells from the prostate tissues were clustered into nine populations according to classical markers for indicated cell types; (E) the gene signature significance between significant modules and indicated prostate lesion were highly correlated; (F–H) the gene ontology biological process analyses of significant module genes showed that the chromatin remodeling and DNA damage repair pathways were exclusively overexpressed in the prostate cancer.
(A)The flowchart of the analysis process; (B) the cluster dendrogram of co-expression genes in different prostate tissues; (C) a heatmap visualizing the correlation of clustered genes; (D) module–trait relationships in different prostate diseases and normal controls. Each cell contains the corresponding correlation parameter and P-value. (E) Representative dot plot of gene signature significance between selected modules and indicated prostate lesion; (F–H) the G.O. biological process analyses of significant module gene in BPH (F), CaP (G), and CRPC (H). Norm, normal prostate; Adj, adjacent tissues; BPH, benign prostate hyperplasia; CaP, prostate cancer; CRPC, castration-resistant prostate cancer; G.O., gene ontology.
Figure 2. Analysis of copy number variations in benign prostate hyperplasia and prostate cancer.
(A) The canonical molecular markers for nine subpopulations from prostate tissues was presented; (B) UMAP embedding of jointly analyzed single-cell transcriptomes from normal and malignant prostate; (C) The single-cell RNA sequencing data was classified by disease state or cell types; (D) The comparison of CNV scores in luminal and basal cells between BPH and CaP showed that CNV was higher in the luminal cells. (E) The estimation of CNV by inferCNV in BPH and CaP was compared with normal epithelial of the prostate. (F) Recurrent CNV in the prostate cancer cohort was identified with GISTIC 2.0 (q-value < 0.2). (G) The top 5 recurrently altered regions are annotated, when the 10q23.31 was frequently deleted in prostate cancer, where PTEN was located.
(A) Distribution of canonical molecular markers for indicated subpopulations visualized in the dot plot; (B) Uniform Manifold Approximation and Projection (UMAP) embedding of jointly analyzed single-cell transcriptomes from normal prostate and prostate cancer tissues; (C) The UMAP embedding of jointly analyzed single-cell transcriptomes of epithelial cells grouped by cell type or disease states; (D) the boxplot indicating the comparison of CNVs scores in luminal cells and basal cells between BPH and CaP. (E) The estimation of copy number variants by inferCNV in BPH and CaP, regarding normal epithelial of the prostate. (F) Recurrent copy number aberrations in the prostate cancer cohort (TCGA- PRAD). Regions of recurrent copy number amplifications (violet) and deletions (green) in the targeted array were identified with GISTIC 2.0 (q-value < 0.2). (G) Dot plot visualizing the top 5 recurrently altered regions are annotated. UMAP, Uniform Manifold Approximation and Projection; BPH, benign prostate hyperplasia; CaP, prostate cancer.
Figure 3. The difference in transcription factors between BPH and prostate cancer.
(A) Heatmap of regulon activity was analyzed by SCENIC with default thresholds for binarization in BPH and CaP. The "regulon" refers to the regulatory network of transcription factors (TFs) and their target genes; (B) most TFs were generally active in both BPH and prostate cancer, except for the indicated exclusive TFs; (C,D) ridge plot visualizes the comparison of indicated regulon activity in basal cells and luminal cells among normal prostate, BPH tissues, and prostate cancer tissues. FOSL1 and MYC were active in BPH; (E) SPDEF was exclusively activated in luminal cells of prostate cancer; (F) XBP1 was active in both BPH and prostate cancer.
(A) Heatmap of regulon activity in BPH and CaP analyzed by SCENIC with default thresholds for binarization. The "regulon" refers to the regulatory network of T.F.s and their target genes; (B) UMAP projection of indicated regulon activity in basal cells and luminal cells from both BPH and prostate cancer samples; (C) ridge plot visualizes the comparison of indicated regulon activity in basal cells and luminal cells among normal prostate, BPH tissues, and prostate cancer tissues. BPH, benign prostate hyperplasia; CaP, prostate cancer; T.F.s, transcription factors.
Figure 4. The unique gene signatures and signaling pathways in BPH.
PPI, protein–protein interaction; DEGs, differential expression of genes; BPH, benign prostate hyperplasia; CaP, prostate cancer; CRPC, castration-resistant prostate cancer; G.O., gene ontology.
(A) The genes of differential expression of genes (DEGs) and significant modules from WGCNA were selected to find the overlap signatures; (B) five overexpressed genes (H2BC15, PSMA2, H4C3, LIMD2, ATP1B2) were found in the BPH DEGs and significant modules; (C) the five gene signatures were not significantly associated with the prognosis of prostate cancer; (D) the shared genes were found to be functionally active in molecular functions toward translation, post-translational modification, or metabolic pathways.
(A) The shared genes among signatures generated from DEGs and WGCNA between the BPH and CaP; (B) a heatmap of the gene expression of five gene signatures from the overlap of DEGs in BPH; (C) the association of the above five gene signatures with the overall survival and disease-free survival in prostate cancer from TCGA-PRAD dataset; (D) the interaction network of G.O. terms generated by the Cytoscape plug-in ClueGO. BPH, benign prostate hyperplasia; CaP, prostate cancer; G.O., gene ontology; DEGs, differential expression of genes.
Figure 5. The role of mutual gene signatures between BPH and CRPC in prognosis and anti-PD1 response.
Forty-two genes were shared between the BPH and CRPC based on differential gene analysis and WGCNA; (B) The shared genes were mainly focused on MAPK phosphorylation and MYC signaling; (C) The protein-protein interaction network and clusters analysis of the shared gene signatures were conducted; (D) The cross-validation error rates were good. (E) Lasso coefficient profiles of the shared gene signatures associated with the overall survival of prostate cancer (F) Boxplots indicating the expression of DDA1 varied in various prostatic lesions; (G) The association with the overall survival and disease-free survival in prostate cancer was not significantly different; (H) The expression of DDA1/C14orf1 were significantly and negatively correlated with the survival probability of cancer patients receiving anti-PD1 treatment.
(A) The shared genes among signatures generated from DEGs and WGCNA between the BPH and CRPC; (B) the interaction network of G.O. terms generated by the Cytoscape plug-in ClueGO; (C) the PPI network and clusters analysis of the shared gene signatures between the BPH and CRPC; (D) plots of the cross-validation error rates. (E) Lasso coefficient profiles of the shared gene signatures associated with the overall survival of prostate cancer. (F) Boxplots indicating the expression of DDA1 in different prostatic lesions; (G) the association of the above four gene signatures with the overall survival and disease-free survival in prostate cancer from TCGA-PRAD dataset; (H) the correlation between the expression of DDA1/C14orf1 and the survival probability of patients with renal cancer receiving anti-PD1 treatment.
Figure 6. The comparison of immune microenvironment between BPH and CRPC.
(A) To further explore the role of immunotherapy, the immune infiltration was statistically compared among prostate lesions; (B) correlation of the immune cells in prostate tissues was high in macrophage and T cells; (C–F) the violin plot showed the difference in the Immune Score (C), ESTIMATE Score (D), Stromal Score (E), and Tumor Purity (F) of every case was presented and colored by the type of prostatic lesions; (G,H) the box plot showed the difference in expression of CD274 (G) and TIGIT (H) among different prostate lesions were increased in prostate cancer, but not BPH and CRPC; (I) compared with CaP, the infiltration of CD8 + T cells, M1 macrophage, and NK cells were significantly increased in BPH and CRPC.
(A) A heatmap of enrichment levels of 22 immune-related cells and types among the normal prostate, adjacent tissues, BPH, CaP, and CRPC tissues; (B) correlation of the immune cells in prostate tissues. The size of the section area and the gradient of colors represented the correlation coefficient R; (C–F) the violin plot showed the difference in the Immune Score (C), ESTIMATE Score (D), Stromal Score (E), and Tumor Purity (F) of every case was presented and colored by groups; (G,H) the box plot showed the difference in expression of CD274 (G) and TIGIT (H) among different prostate lesions; (I) the differences in the infiltrating immune cells between five clusters. ssGSEA, single-sample Gene Set Enrichment Analysis; BPH, benign prostate hyperplasia; CaP, prostate cancer; CRPC, castration-resistant prostate cancer; *P < 0.05, **P < 0.01, ***P < 0.001, ns, none significance.
Figure 7. The correlation between selected gene signatures with the tumor immune microenvironment.
(A–D) The expression of OXAL1 (A), ERG28 (B), DDA1 (C), and OGFOD1 (D) were all negatively and significantly correlated with the infiltration of CD8+ T cells according to the TCGA dataset; (E) the mRNA expression of OXAL1, ERG28, DDA1, and OGFOD1 and CD8+ T cells were significantly correlated in prostate cancer tissues with adjacent normal tissues; (F) the copy number variations (CNV) of OXAL1, ERG28, and OGFOD1 were significantly correlated with CD8+ T cells in prostate cancer tissues; (G) the GSVA score of the four gene signatures was significantly correlated with infiltration of CD8 + T, DC, or NK cells in prostate cancer tissues.
(A–D) The correlation between the mRNA expression of OXAL1 (A), ERG28 (B), DDA1 (C), and OGFOD1 (D) and different infiltrating immune cells in prostate cancer tissues with adjacent normal tissues; (E) the correlation between the mRNA expression of OXAL1, ERG28, DDA1, and OGFOD1 and CD8 + T cells in prostate cancer tissues with adjacent normal tissues; (F) the correlation between the copy number variations (CNV) of OXAL1, ERG28, and OGFOD1 and CD8+ T cells in prostate cancer tissues; (G) the correlation between the GSVA score of the four gene signatures and infiltration of indicated immune cells in prostate cancer tissues. CNV, copy number variations.
Supplemental material
Supplemental Material
Download MS Word (3.8 MB)Supplemental Material
Download MS Excel (228.5 KB)Data availability statement
The original contributions presented in the study are included in the article/Supplementary Table S1, further inquiries can be directed to the corresponding author. All the scripts used in this study could be obtained by request to the corresponding author.