Objective： To explore intercellular communication-related signaling molecules in tumor microenvironment (TME) of castration-resistant prostate cancer (CRPC) and analyze their relationship with tumor occurrence and development. Methods： The main components of the TME in CRPC was analyzed using singlecell RNA sequencing (scRNA-seq) data (GSE137829) from CRPC patients found in the Gene Expression Omnibus (GEO) database. CellPhoneDB was used to examine interactions between different cells within the microenvironment, with a focus on revealing the connections between cancer-associated fibroblasts (CAFs) and malignant epithelial cells. CellChat was applied to analyze signaling molecules involved in communication between CAFs and malignant epithelial cells. Data from The Cancer Genome Atlas Prostate Adenocarcinoma (TCGA-PRAD) were integrated to identify signaling molecules closely related to the prognosis of CRPC patients. siRNA was used to knock down the expression of the signaling molecule Delta-like canonical Notch ligand 3 (DLL3) in CRPC cells, and the knockdown efficiency was validated by real-time quantitative PCR and Western blotting. The effects of DLL3 on the biological behaviors of CRPC cells, including proliferation, migration, and invasion, were investigated using CCK8 assays, EdU assays, and Transwell assays with Matrigel, respectively. Singlegene differential analysis, overrepresentation analysis (ORA), and gene set enrichment analysis (GSEA) were also performed to identify pathways enriched with DLL3-related differential expression genes. Results： The results of single-cell transcriptomic sequencing indicated that the main cellular components of the TME in CRPC include malignant epithelial cells, CAFs and immune cells. Close communication between CAFs and malignant epithelial cells was mediated by interaction-related signaling molecules, with 63 signaling molecules identified as closely related to the communication between CAFs and malignant epithelial cells. Univariate Cox regression analysis identified key signaling molecules, including tumor necrosis factorrelated ligand superfamily member 12 (TNFSF12) and DLL3, whose transcriptional expression were significantly associated with the prognosis of CRPC patients. Notably, high expression of DLL3 was significantly associated with poor prognosis in CRPC patients. In vitro experimental results demonstrated that knockdown of DLL3 significantly inhibited the growth, proliferation, migration and invasion capabilities of CRPC cells. KEGG and GSEA enrichment analyses indicated that differentially expressed genes related to DLL3 were significantly enriched in metabolic pathways. Conclusion： In the TME, CAFs and malignant epithelial cells communicate closely through intercellular signaling molecules. Among these signaling molecules, the expression of DLL3 is negatively correlated with cancer prognosis and affects the biological behavior of CRPC cells, suggesting that DLL3 may be a key signaling molecule in the malignant progression of human CRPC.
［Key words］castration-resistant prostate cancer; scRNA-seq; tumor microenvironment; cell communication; DLL3

Objective: To construct a prognostic risk model for predicting bone metastasis in prostate cancer (PCa) based on bioinformatics approach, and to investigate the effect of risk score on prognosis and immune infiltration. Methods: Clinical information and microarray expression data of PCa patients were downloaded from the Gene Expression Omnibus (GEO), and genes related to the bone morphogenetic protein (BMP) pathway were downloaded from the Molecular Signature Database (MSigDB). The differential expression genes (DEGs) of PCa bone metastasis samples and primary PCa samples were screened by differential analysis, and the intersection with BMP pathway-related genes was taken to obtain BMP-related DEGs, and the prognostic risk model was constructed by screening risk genes through LASSO regression analysis, and independent prognostic factors of PCa were screened by multifactorial regression analysis. The risk scores of patients were calculated according to the risk model and subsequently divided into two groups of high and low risk scores by corresponding score median values, and DEGs were identified for functional enrichment analysis to compare the differences in immune infiltration between high and low risk groups. Results: After the intersection of DEGs and BMP-related gene sets in 3 055 differentially expressed genes of PCa bone metastasis samples, 13 BMP-related DEGs were obtained. Four gene were screened by LASSO regression to construct a prognostic risk model, in which secreted frizzled-related proteins 2 (SFRP2), endoglin (ENG), and follistatin-like protein 1 (FSTL1) were risk genes and chordin like 1 (CHRDL1) was a protective gene. Kaplan-Meier analysis showed that metastasis-free survival and biochemical recurrence-free survival of patients in the high-risk group were significantly lower than that of patients in the low-risk group. Multifactorial regression analysis identified risk score as an independent prognostic factor for PCa bone metastasis, and patients in the high-risk group were more likely to develop bone metastasis and biochemical recurrence, and the risk score was positively correlated with the PSA value, the Gleason score, and the T stage. The DEGs of high-risk group were mainly involved in the activation of the WNT/BMP signaling pathway, and were significantly enriched in the proliferation of epithelial cells and extracellular matrix bioprocesses. GSEA analysis showed that, the differentially expressed genes were upregulated in the gene sets of mesenchymal transition, inflammatory response, angiogenesis, and multiple tumor metastasis gene sets. Immune infiltration analysis showed that PCa bone metastasis samples and patients in the high-risk group possessed a higher degree of immune cell infiltration, cancer associated fibroblasts (CAFs), macrophages, Estimate scores, stromal scores, and immune scores than the lowrisk group, and lower tumor purity than the low-risk group. Conclusion: The prognostic risk model of BMP-related DEGs constructed based on the LASSO regression analysis can effectively predict the occurrence of PCa bone metastasis, and the high-risk score is closely related to the prognosis and immune infiltration of PCa.
［Key words］prostate cancer; bone metastasis; BMP signaling pathway; prognostic risk model; immune invasion