Application of the GRN-NFκB-CSF1 loop as a target for the preparation of products that inhibit prostate cancer lineage plasticity

Abstract

The present invention provides the use of the GRN-NFκB-CSF1 loop as a target in the preparation of products that inhibit the plasticity of the prostate cancer lineage, and belongs to the technical field of prostate cancer prevention and treatment. The present invention is verified by in vitro cell co-culture, in vivo targeted intervention, organoid drug resistance test and spatiotemporal information analysis in the tumor microenvironment. The results show that monocytes / macrophages promote the transition from adenocarcinoma to the VIM lineage with mesenchymal and stem cell-like characteristics through the paracrine mechanism of GRN / NF-κB signaling. Conversely, the VIM lineage promotes the high expression of GRN in monocytes / macrophages by secreting CSF1, forming a positive feedback cell communication loop. In summary, the research of the present invention reveals that the paracrine feedback loop of the GRN-NFκB-CSF1 axis can be used as a therapeutic target, and targeting this loop is expected to reverse the resistance to androgen receptor signaling targeted therapy.

Description

Technical Field

[0001] The present invention belongs to the technical field of prostate cancer prevention and treatment, and specifically relates to the application of the GRN-NFκB-CSF1 loop as a target in the preparation of products that inhibit prostate cancer lineage plasticity. Background Art

[0002] Prostate cancer is a common malignancy in men, characterized by its clinical heterogeneity. In patients with locally advanced or metastatic disease, the aggressive nature of the disease often leads to rapid resistance to commonly used androgen deprivation therapy (ADT), resulting in the emergence of castration-resistant prostate adenocarcinoma (CRPC-Adeno), characterized by high expression of androgen receptor (AR) signaling in the luminal lineage. AR promotes the reestablishment of transcriptional programs or the activation of alternative transcription factors, potentially contributing to ADT resistance. New drugs such as enzalutamide and abiraterone aim to prolong cancer survival by inhibiting AR signaling, but they may induce a variety of cancer phenotypes independent of AR signaling, including small cell carcinomas with neuroendocrine (NE) features or double-negative tumors. Lineage plasticity has emerged as a novel mechanism that promotes multilineage differentiation and acquisition of drug resistance, independent of AR activity. It is widely believed that lineage plasticity contributes to the stem-like epithelial-mesenchymal transition (EMT)-like transition of luminal adenocarcinoma cells, allowing these cells in this multilineage state to redifferentiate into various lineages, such as the NE lineage. Despite the known genetic and epigenomic alterations that may occur, drug targets for prostate cancer lineage plasticity remain elusive.

[0003] In contrast, increasing evidence suggests that the primary source of non-genetic regulation stems from the tumor microenvironment (TME). Non-genetic regulation can alter characteristics such as metastasis and drug resistance by determining the state of cancer cells. For example, SPP1, secreted by cancer-associated fibroblasts, promotes castration resistance in prostate cancer, while KIT ligands, secreted by stromal cells, drive lineage plasticity in prostate cancer. The TME is rich in a variety of inflammatory factors that can synergize with oncogenic mutations to promote tumor cell proliferation and malignancy. Macrophages, as the primary source of numerous inflammatory cytokines in prostate inflammation, are associated with poorly differentiated and plastic cells. For example, the human macrophage-like cell line THP-1 has been shown to promote prostate hyperplasia, while interleukin-6 (IL-6) secreted by macrophages supports the development of neuroendocrine prostate cancer (NEPC). However, the role of inflammatory factors released by macrophages in maintaining stem cell-like and multi-lineage states requires further investigation. Summary of the Invention

[0004] In view of this, the object of the present invention is to provide the use of the GRN-NFκB-CSF1 loop as a target in the preparation of products for inhibiting prostate cancer lineage plasticity.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0006] The present invention provides an application of a GRN-NFκB-CSF1 loop as a target in the preparation of a product for inhibiting prostate cancer lineage plasticity.

[0007] Currently, most of the reported drivers associated with prostate cancer lineage plasticity are not suitable as targets for drug intervention. The present invention proposes for the first time that granulin (GRN, Grn, granulin) is a ligand that drives lineage transition and AR-targeted therapy resistance. The prior art does not report whether GRNs secreted by macrophages have a role in promoting lineage plasticity. The present invention reveals for the first time that GRNs regulate NF-κB signaling activity in tumor cells, and that NF-κB signaling activation plays a role in transitioning to stem cell-like and EMT states, but does not directly redifferentiate into NE-like lineages. In addition, CSF1 expression is upregulated in specific malignant lineages. CSF1 is essential for the polarization of M2 macrophages and the production of GRNs, promoting the production of GRNs in a positive feedback manner. Blockade of the CSF-1 / CSF-1R axis in TRAMP mice verifies the key role of the integrity of the positive feedback loop in the lineage transition process. The present invention clarifies for the first time the key role of the GRN-NF-κB signaling-CSF1 loop in mediating prostate cancer lineage transition, providing a new therapeutic target for overcoming AR-targeted therapy resistance.

[0008] The present invention also provides the use of the GRN-NFκB-CSF1 loop as a target in the preparation of a product for treating prostate cancer.

[0009] In the present invention, the prostate cancer is preferably androgen receptor targeted therapy-resistant prostate cancer.

[0010] This study employed single-cell multi-omics analysis to reveal dynamic changes in immune cell infiltration, transcriptional regulation, and intercellular communication in prostate cancer. Results demonstrated that GRN-positive macrophages promote the transition of adenocarcinomas to a multilineage state with mesenchymal and stem-like features by activating intratumoral NF-κB signaling. Multilineage clones induce high GRN expression in macrophages through CSF1 secretion, forming a positive feedback loop for cellular communication. In vitro experiments demonstrated the biological function of GRN in mediating epithelial-mesenchymal transition, and organoid drug resistance studies demonstrated that GRN contributes to resistance to the androgen receptor-targeted drug enzalutamide. In the TRAMP mouse model, pharmacological blockade of the CSF-1 / CSF-1R axis reduced GRN expression in macrophages and inhibited the formation of multilineage subclones in prostate malignant cells. Multiple immunofluorescence staining of tumor samples from the TRAMP mouse model revealed that the VIM lineage is spatially closely associated with macrophages. These findings were further validated at the single-cell protein level using mass cytometry (CyTOF). In addition, there are three tumor-infiltrating subpopulations associated with disease recurrence, including DCN+ endothelial cells, CCL7+ fibroblasts, and IFIT1+ neutrophils. These research results of the present invention provide a possible therapeutic target for solving the resistance to AR-targeted therapy caused by lineage remodeling. In the present invention, the VIM lineage refers to a cell lineage with high expression of VIM, a cancer cell type derived from cancer epithelial cells and showing mesenchymal characteristics.

[0011] The present invention also provides the use of a GRN-NFκB-CSF1 loop inhibitor in the preparation of a product for inhibiting prostate cancer lineage plasticity.

[0012] The present invention also provides the use of a GRN-NFκB-CSF1 loop inhibitor in the preparation of a product for treating prostate cancer.

[0013] In the present invention, the prostate cancer is preferably androgen receptor targeted therapy-resistant prostate cancer.

[0014] In the present invention, the GRN-NFκB-CSF1 loop inhibitor preferably includes a GRN inhibitor or a CSF-1 / CSF-1R axis inhibitor. In the present invention, the CSF-1 / CSF-1R axis inhibitor preferably includes BLZ945; the GRN inhibitor preferably includes shRNA, and the nucleotide sequence of the shRNA is preferably as shown in SEQ ID NO. 1 or SEQ ID NO. 2.

[0015] The present invention also provides a shRNA that inhibits prostate cancer lineage plasticity. The nucleotide sequence of the shRNA is preferably as shown in SEQ ID NO.1 or SEQ ID NO.2.

[0016] In the present invention, the product preferably comprises a medicine.

[0017] Beneficial effects of the present invention:

[0018] The present invention explores the complexity of TME through single-cell multi-omics analysis and reveals the molecular mechanism behind the plasticity of prostate cancer lineage. The present invention is verified by in vitro cell co-culture, in vivo targeted intervention, organoid drug resistance test and spatiotemporal information analysis within TME. The results show that monocytes / macrophages promote the transition from adenocarcinoma to the VIM lineage with mesenchymal and stem cell-like characteristics through the paracrine mechanism of granulin (GRN) / NF-κB signaling. Conversely, the VIM lineage promotes the high expression of GRN in monocytes / macrophages by secreting CSF1, forming a positive feedback cell communication loop. In summary, the present invention reveals that the paracrine feedback loop of the GRN-NFκB-CSF1 axis can be used as a therapeutic target, and targeting this loop is expected to reverse the resistance to AR signal targeted therapy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Figure 2 shows the results of single-cell multi-omics analysis of 29 mouse samples, where A is a schematic diagram of the design of the present invention; B is a UMAP visualization diagram showing cell populations colored by annotated major cell types; C is a pie chart showing the cell distribution colored by annotated major cell types assigned by sample type (top) and the cell fractions (y-axis) of samples at different time points colored by annotated major cell types (bottom); D is the cell fractions (y-axis) of samples at different time points colored by annotated tumor microenvironment cell types; E and F are violin plots showing the pseudo-time of five malignant lineages (E) and seven time points (F); G is the cell fractions (y-axis) of samples at different time points colored by the five annotated malignant lineages.

[0020] Figure 2Figure 2 shows the tumor heterogeneity results of malignant cells of five different cell lineages, where A is the unsupervised hierarchical clustering of the five malignant lineages; B is the survival probability of prostate cancer patients (from TCGA data) stratified by the expression level of the VIM lineage gene signature; C is the RNA velocity vector of the five malignant lineages on the t-distributed random neighbor embedding; D is a heat map showing the RNA velocity vectors based on the Hallmark Pathway gene set (top) and KEGG Pathway gene set (bottom) shows the two most enriched pathways in malignant lineages; E and F are violin plots showing AR, NEPC, EMT, stemness, and inflammatory signaling signatures in five malignant lineages (E) and three time points (F); G is pairwise correlation clustering of 35 gene expression programs (GEPs) in malignant cells, and the biological functions of the resulting nine consensus modules were predicted by GO analysis (top); H is the survival probability of prostate cancer patients (from TCGA data) stratified by the expression level of EMT module genes; I is a two-dimensional butterfly plot of the NE, synthesis, biosynthesis, cell cycle, and EMT modules in the five malignant lineages, with the visualized module scores as relative meta-module scores, and each quadrant corresponds to a module; the exact position of each cell reflects its relative signature score in all four modules; J is a heat map showing the Pearson correlation between the top differentially enriched pathways and the four modules (NE, synthesis, cell cycle, and EMT modules), asterisks indicate statistically significant comparisons (P value < 0.05), and the scale bar represents the Pearson correlation (r) (red = positive correlation, green = negative correlation).

[0021] Figure 3Figure 3 GRN+ macrophages promote prostate cancer lineage transition results, where A is a heat map showing the number of receptor-ligand interactions between tumor microenvironment cell types and the five malignant lineages; B is a UMAP visualization showing cell populations colored by annotated Mono / Macro / DC subpopulations; C is a volcano plot showing differentially expressed genes in Mono / Macro / DC between PRP_8-9weeks_Intact and WT; D is a volcano plot showing differentially expressed genes in Mono / Macro / DC between PRP_12-16weeks_Intact and PRP_8-9weeks_Intact; E is a heat map showing GRN gene expression scores of Mono / Macro / DC at different time points; F is a dot plot showing selected receptor-ligand pairs interacting with the five malignant lineages and Mono / Macro / DC subpopulations, and the dot size indicates the difference between the two groups by permutation test. The P value generated by the experiment is shown, and the color represents the average expression of each receptor-ligand pair; G is a dot plot showing the gene expression levels of receptor-ligand pairs in five malignant lineages and Mono / Macro / DC subsets; H is a UMAP visualization showing the GRN expression density in Mono / Macro / DC; I is a ring plot showing the proportion of Mono / Macro / DC subpopulations with GRN expression greater than 0; J is a violin plot showing the GRN expression levels of different Mono / Macro / DC subpopulations; K is a heat map showing the GRN gene expression scores of macrophage subsets in different sample types; L is a dot plot showing the inflammatory signal score and lineage transition signal score based on 39 clusters with a resolution of 2 in GEMMs; M is a two-dimensional plot showing the dynamic changes in CD44 and TNFRSF1A expression and the scores of EMT, stemness and TNFA-SIGNALING-VIA-NFKB characteristics.

[0022] Figure 4Figure 3 shows that GRN+ macrophage-derived GRN mediates the EMT program and drives AR-targeted therapy resistance, where A is the CCK8 assay to measure the cell viability of RM-1 cells after treatment with different doses of recombinant GRN; B is a schematic diagram of co-culture; C is quantitative RT-PCR to detect overexpression of the GRN gene in mouse macrophages RAW264.7, and the bar graph represents the fold change of mRNA levels relative to NC; D is immunoblotting to detect EMT proteins after co-culture of RM-1 cells with RAW264.7 overexpressing the Grn gene; E is quantitative RT-PCR to detect shRNA targeting of the Grn gene in mouse macrophages RAW264.7, and the bar graph represents the fold change of mRNA levels relative to NC; F is immunoblotting to detect EMT proteins after co-culture of RM-1 cells with RAW264.7 transfected with shRNA Western blot; G shows the morphological changes of RM-1 cells after co-culture with RAW264.7 cells overexpressing Grn; H shows the enriched pathways upregulated and downregulated in RM-1 cells after co-culture with RAW264.7 cells overexpressing GRN, sorted by P value; I shows representative images of human prostate cancer organoids with or without GRN after enzalutamide treatment, scale bar: 50 μm; J and K show the viability (n=9 wells, from 3 patients) and growth area (n=20 fields, from 3 patients) of human prostate cancer organoids with (K) or without (J) GRN after enzalutamide treatment, the viability was measured by CCK8 assay, and the growth area was measured by high-content assay; data are mean ± SD, **P < 0.01, ****P < 0.0001, using one-way analysis of variance.

[0023] Figure 5Figure 3. Results of the important TME components and lineage programs in CRPC patient biopsies, where A is a UMAP visualization showing all cells in CRPC patient biopsies, colored by the annotated major cell types; B is a UMAP visualization showing tumor cells in CRPC patient biopsies, colored by the annotated tumor cell types; C and D are UMAP visualizations showing GRN, TNFRSF1A, CD44 gene markers (C) and FOXA2, GHGB and POU3F2 (D); E is the survival probability of prostate cancer patients (from TCGA data) stratified by the expression level of the mixed lineage gene signature; F and G are heat maps showing the expression of the GSEA Hallmark Pathway gene set (F) and the KEGG The most enriched pathways in the malignant lineages of the Pathway gene set (G); H is a scatter plot showing the proportions of glandular cells (left), mixed lineage cells (middle) and NEPC (right) based on Monocle3 pseudo-time; I is a violin plot showing the scores of AR, NEPC, EMT, stemness and TNFA-SIGNALING-VIA-NFKB features in the three tumor lineages; J is a scatter plot showing tumor clusters (dots) (y-axis) with HALLMARK_TNFA_SIGNALING_VIA_NFKB and stemness (top) and EMT (bottom) features, with linear fits calculated for non-NEPC and NEPC respectively, and the corresponding Pearson correlations are indicated, and the clusters are annotated as non-NEPC (red) and NEPC (green).

[0024] Figure 6Figure 3. CSF1 protein derived from the VIM lineage promotes macrophage-induced lineage progression. A is a UMAP visualization showing mixed lineage cell populations from CRPC patient biopsies, colored by tumor cluster; B is the developmental trajectory of mixed lineages, colored by mixed lineage cluster (top) and pseudo-time score (bottom); C is a box plot showing the CytoTRACE score of mixed lineage clusters; D is a violin plot showing AR, NEPC, EMT, stemness and inflammatory signaling characteristics in mixed lineage clusters; E is GO enrichment analysis of tumor cluster-specific genes; F is a dot plot showing selected receptor-ligand pairs interacting with the VIM lineage and Mono / Macro / DC subsets in WT and GEMMs mice, the dot size indicates the P value generated by the permutation test, and the color indicates the average expression of each receptor-ligand pair; G is a violin plot showing the expression of luminal cells and five malignant lineages in GEMMs. Expression levels of Ccl3 (top) and Csf1 (bottom) in cells; H is a violin plot showing the expression levels of CCL3 (top) and CSF1 (bottom) in mixed lineage clusters from CRPC patient biopsies; I is a UMAP visualization showing the Csf1r expression density of Mono / Macro / DC in GEMMs biopsies; J is a ring plot showing the proportion of Mono / Macro / DC subpopulations with Csf1r expression greater than 0 in GEMMs biopsies; K is a violin plot showing the Csf1r expression levels of different Mono / Macro / DC subpopulations in GEMMs biopsies; L is a heat map showing the top ligands regulating Mono / Macro / DC in GEMMs biopsies inferred by NicheNet based on the VIM lineage; M is the enrichment of target genes expressed in Mono / Macro / DC based on predictions; Fisher test was used for statistical analysis.

[0025] Figure 7Figure 3. Pharmacological blockade of CSF-1 inhibits the formation of the VIM lineage in the TRAMP mouse model. A is a schematic diagram of the experimental design of BLZ945 in 25-week-old TRAMP mice. After 7 days of treatment, prostate tumors were collected for single-cell RNA sequencing and multiplex immunohistochemistry. B is a UMAP visualization showing cell populations colored by the annotated major cell types of TRAMP mice. C is a heat map showing the similarity of cell clusters in GEMMs mice (x-axis) and TRAMP mice (x-axis). D is a top-ranked chart stratified by the expression level of the VIM lineage gene signature in TRAMP mice. Survival probability of adenocarcinoma patients (from TCGA data); E is a volcano plot showing differentially expressed genes between the VIM lineage and adenocarcinoma (Adeno); F is a gene set enrichment analysis (GSEA) (HallmarkPathway gene set) ranking genes by log2 fold change between the VIM lineage and adenocarcinoma (Adeno), NES, and normalized enrichment score; G is a violin plot showing EMT, stemness, inflammatory signaling signatures, and CytoTRACE scores in mixed lineage and adenocarcinoma clusters; H is a violin plot showing Csf in major cell types in TRAMP mice 1 expression level; I is a heat map showing the tissue preference of adenocarcinoma clusters and cell subpopulations of the VIM lineage in prostate tumor tissues of TRAMP mice treated with BLZ945 and control (vehicle); J is a violin plot showing the Csf1 expression level in the VIM lineage of TRAMP mice treated with BLZ945 and control; K and L are violin plots showing the expression levels of Csf1r (K) and Grn (L) in the main cell types of TRAMP mice; M is a UMAP visualization showing Mono / Macro / DC cell population; N is a heat map showing the tissue preference of adenocarcinoma clusters and cell subsets of the VIM lineage in prostate tumor tissues and blood samples of TRAMP mice treated with BLZ945 and controls; O is a violin plot showing the Grn expression level in prostate tumor tissues and blood samples of TRAMP mice treated with BLZ945 and controls; P is the RNA velocity vector of macrophages and monocytes in blood samples and prostate tumor tissues on t-distributed random neighbor embedding; Q is multiple immunofluorescence staining showing the co-localization of VIM+ cells and macrophages (F4 / 80+), the scale bar represents 10 μm.

[0026] Figure 8Figure 3 shows the mass spectrometry cytometry analysis results of phenotypic and compositional heterogeneity of CRPC patient biopsy cells, where A is a schematic diagram of the CyTOF experimental design; B is a UMAP projection showing cells from CRPC-Adeno (n=3) and CRPC-NEPC (n=4) patient biopsies, colored by FlowSOM clusters; C is a violin plot showing the signal intensity of Snail.1, ZEB2, Oct3-4, and vimentin in the ten major clusters; D is the cell type ratio (y-axis) of CRPC-Adeno and CRPC-NEPC patient biopsies, colored by the ten annotated cell types; E is a violin plot showing epithelial, mixed lineage, and mitochondrial lineages. Figure 3 Signal intensity of selected markers in GRNs and neuroendocrine cells; F is a violin plot showing the signal intensity of GRNs in the ten major clusters; G is a heat map showing the spatial distribution of the five clusters; H is a violin plot showing the AR and NEPC characteristics and marker genes in the five clusters; I is a heat map showing the spatial distribution of TAM gene markers (MSR1, CSF1R, CD163), GRNs, NEPCs, GRN-high Mono / Macro, stemness and TNFA-SIGNALING-VIA-NFKB characteristics; J is a proposed model system showing the coordinated multi-lineage state of granulin+ macrophages with mesenchymal-like and stem-like properties.

[0027] Figure 9 These are the results of cell marker genes, where A: bubble chart displays marker genes of different clusters, B: bubble chart displays marker genes of different cell types, and C: umap chart displays marker genes.

[0028] Figure 10 The infiltration patterns of the main cell types in prostate cancer mice in six independent prostate cancer cohorts, including A: correlation between cell abundance; B: correlation between Mono / Macro / DC marker gene scores and cell plasticity scores; C: the proportion of Mono / Macro / DC in different prostate tissue subtypes in GSE70768; D: the relationship between the abundance of immune cells and survival prognosis (overall survival) in different databases.

[0029] Figure 11 Annotation of myeloid cell subsets, where A: heat map shows the highly variable genes of myeloid subsets; B: UMAP shows myeloid subsets; C: UMAP shows the marker genes of myeloid subsets; D: bar chart shows the percentage changes of samples in different periods; E: bubble chart shows the expression of TNF family genes in different immune cell types; F: bubble chart shows the expression of TNF family genes in different cancer cell types.

[0030] Figure 12 The results of human prostate cancer cell clustering and annotation are shown, where A: human prostate cancer tissue clustering and annotated genes; B: human prostate cancer epithelial cell clustering and annotated genes.

[0031] Figure 13 The infiltration patterns of major cell types in human samples from six independent prostate cancer cohorts, including A: correlation between cell abundance; B: correlation between Mono / Macro / DC marker gene scores and cell plasticity scores; C: proportion of Mono / Macro / DC in different prostate tissue subtypes in GSE70768; D: relationship between immune cell abundance and survival prognosis (overall survival) in different databases.

[0032] Figure 14 The differentiation of human prostate cancer cells; A: monocle3 shows the differentiation trajectory of human prostate cancer cells, B: violin plot shows the psedotime of different prostate cancer cell subpopulations, C: core gene expression changes with psedotime, D: velocity shows the differentiation trajectory of different cancer cell types, E: umap shows the expression of different marker genes.

[0033] Figure 15 The results of the spatial transcriptome of prostate cancer, where A: markers of the spatial transcriptome, B: distribution of GRN in different spaces, C: RCTD deconvolution of idle rotation, D: CNV scores of different clusters, E: correlation between GRN and tumor-associated macrophage markers in the spatial transcriptome.

[0034] Figure 16 It is a marker gene for CyTOF analysis of human prostate cancer samples. DETAILED DESCRIPTION

[0035] The technical solutions provided by the present invention are described in detail below with reference to the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0036] In the following examples, unless otherwise specified, all methods are conventional.

[0037] Unless otherwise specified, the materials and reagents used in the following examples can be obtained from commercial sources.

[0038] Example 1

[0039] Test method:

[0040] (1) Collection and acquisition of mouse models and human prostate tumors

[0041] Prostate tumor samples from three patients with CRPC-Adeno and four patients with CRPC-NEPC were analyzed using mass cytometry (CyTOF). These patients underwent transurethral prostatectomy at Shanghai Tongji Hospital, and written informed consent was obtained. The clinical characteristics of the patients included in this method are shown in Table 1. TRAMP mice (male, 25 weeks, weighing 40 g) used for pharmacological blockade experiments were purchased from Jiangsu Huachuang Xinnuo Pharmaceutical Technology Co., Ltd.

[0042] Table 1 Clinical characteristics of included patients

[0043]

[0044] Single-cell RNA expression profiles of 29 mouse samples (9 wild-type [WT], 7 Pten- / -Rb- / - [PtR], and 13 Pten- / -Rb- / -Trp53- / - [PtRP]) and 12 human prostate cancer samples were obtained from Joseph M. Chan, GSE210358. Spatial transcriptome sequencing data were obtained from GSE230282. In addition, this method also utilized RNA-seq data and clinicopathological data from multiple sources, including the Cancer Genome Atlas (TCGA) PRAD cohort (483 prostate patients), GSE54460 (106 prostate patients), GSE116918 (248 prostate patients), GSE54460 (107 prostate patients), GSE70768 (13 CRPC tissues, 113 tumor tissues, 73 matched benign tissues), GSE70769 (93 prostate patients), GSE94767 (154 prostate cancer patients), and DKFZ-PRAD (268 prostate patients). These datasets were obtained from the TCGA database (https: / / portal.gdc.cancer.gov / ), cBioPortal (https: / / www.cbioportal.org / ), and GEO database (https: / / www.ncbi.nlm.nih.gov / gds / ).

[0045] (2) Single-cell RNA sequencing analysis of mouse models

[0046] First, 103,684 single cells from genetically engineered mouse models (GEMMs) were integrated. To ensure data quality, the Scrublet algorithm was used to identify and remove data that might be double cells. The integrated dataset was subjected to three quality filters: 1) the total UMI count per cell must be higher than 500; 2) the number of genes detected per cell must be higher than 250 and lower than 6,000; and 3) the percentage of mitochondrial genes must be lower than 20%. Based on these criteria, the 83,183 retained cells were analyzed using the R package Seurat v4.3.1. Principal component analysis was performed on the 1,500 highly variable genes identified, and a nearest neighbor graph was constructed based on the first 50 principal components (PCs) for clustering using the FindNeighbors function. Then, the 15 major cell types were annotated with marker genes using the FindClusters function. Marker genes used for annotation include: HOXB13, SBP, KRT4, FOXI1, SDC1, CALML3, and PATE4 for normal epithelial cell lineages (L1, L2, L3, Basal, Basal SV, and SV); KRT14, KRT5, EPCAM, TFF3, VIM, POU2F3, CHGA, and GFP for malignant cell lineages (Adeno, TFF3, POU2F3, VIM, and NEPC); CD79A, MS4A1, CD3E, and CD3D for lymphocytes (B and NK / T); COL1A1 and COL1A2 for mesenchymal cells (endothelial cells and fibroblasts); CD68 and CST3 for myeloid cells (Mono / Macro / DC); and CD14 and CSF3R for myeloid cells (PMNs). To further annotate cell subtypes, a second round of clustering was performed for each of the major cell types.

[0047] (3) scRNA-seq analysis of human prostate cancer patients

[0048] Preprocessing of single-cell sequencing data from human prostate cancer samples was performed using the same methods used for single-cell RNA-seq analysis of GEMM samples. Five major cell types were annotated based on known marker genes: CLDN5, COL1A1, and VIM for mesenchymal cells; CD14, LYZ, and PTPRC for myeloid cells; CD2, CD3E, and CD3D for lymphocytes; AR, CDKN2A, EPCAM, and KLK3 for epithelial cells; and ALB and CRP for benign hepatocytes. Malignant cells were identified and further classified into three subtypes using the same methods as in the first round: AR and KLK3 for malignant epithelial cells; NOTCH1, KLF4, TGFB2, VIM, and SMAD2 for mixed-lineage cells; and ASCL1, CHGA, and SYP for NEPCs.

[0049] (4) scRNA-seq analysis of the TRAMP mouse model

[0050] The TRAMP mouse model develops poorly differentiated prostate cancer with features of nephropathy at 24 weeks of age. 25-week-old TRAMP mice were orally administered 200 mg / kg of the CSF-1R inhibitor BLZ945 daily for 7 days. A 20% captisol treatment was used as a control. Fresh prostate tissue was collected for single-cell transcriptome sequencing. Sequencing data from TRAMP mouse samples were pre-processed using the same method as single-cell RNA-seq analysis of GEMM samples. Ten major cell types were annotated based on known marker genes: Cd79a and Mzb1 for B cells, Cd3d, Cd3e, and Nkg7 for T / NK cells, Epcam for Adeno, Cd14 and Cd68 for monocytes / macrophages / dendritic cells (Mono / Macro / DC), S100a8 and S100a9 for PMNs, Gp9 and Pf4 for megakaryocytes, Top2a for circulating cells, Col1a1 and Col1a2 for fibroblasts, Pecam1 for endothelial cells, and Vim for VIM (exploitin-positive cells). Adeno and VIM were further re-clustered into 13 sub-cell types, and Mono / Macro / DC were further re-clustered into 7 sub-cell types, using the same methods as the first round. To assess the enrichment of cell types in each tissue, the observed number of cells in each cluster was compared with the expected number of cells.

[0051] (5) Pathway enrichment analysis

[0052] Three different approaches were used to assess enriched pathways: 1) Gene Ontology (GO) analysis: Using the “FindAllMarkers” function in the R package Seurat, as described above, genes with unique expression in relevant populations, clusters, or cell types were identified and subjected to GO enrichment analysis using gene sets from the Gene Ontology database (http: / / www.geneontology.org / ). This analysis was performed using the clusterProfiler package version 21. 2) Gene Set Enrichment Analysis (GSEA): Enrichment analysis of relevant populations, clusters, or cell types was performed using the Hallmark gene set (RRID: SCR_016863) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) gene set (https: / / www.genome.jp / kegg / ). Significantly enriched pathways with an FDR < 0.05 were identified using GSEA (http: / / software.broadinstitute.org / gsea / index.jsp). 3) Gene set signature scoring: For single-cell transcriptome data, multiple known signatures were acquired and cells were scored using AUCell. For spatial transcriptome data, GRN+mono / macro refers to cells whose GRN gene expression levels in GEMMs are higher than the average mono / macro level. These signature genes were obtained using COSG32. The gene signature score for each cell was determined by calculating the average expression value of all genes included in the signature.

[0053] (6) Survival analysis of new subclusters in bulk RNA-seq data

[0054] To analyze the survival outcomes of novel subclusters in bulk RNA-seq data, we performed the following steps: We selected specific signatures for different cell populations based on specific criteria, such as Fibroblast-c1-CCL7, Endothelial-c2-ACKR1, Endothelial-c3-DCN, PMN-c6-IFIT1, the VIM lineage in GEMMs, and mixed lineages in human prostate cancer patients. Selection criteria included a fold change greater than 2, an adjusted P value less than 0.01, and a percentage of cells expressing the signature greater than 0.1. A signature score for each cell population was determined using the mean expression value of the signature genes based on RNA sequencing data from the TCGA. Patients were then divided into two groups based on the median signature score. Kaplan-Meier survival curves were compared using a two-sided long-rank test.

[0055] (7) Cell infiltration analysis based on single-cell data

[0056] To estimate the infiltration of 10 major cell types defined in human CRPC patient samples from GEMM samples and the TCGA cohort (https: / / portal.gdc.cancer.gov / ), the DKFZ-PRAD cohort (https: / / www.cbioportal.org / ), and other public bulk RNA-seq and microarray datasets, a reference marker matrix of major cell types was generated using CIBERSORTx. Cell fractions were estimated using a permutation parameter setting of 500 replicates, while other parameters were kept at default values. Spearman correlation analysis was performed on the infiltration levels of cell types; correlation coefficients greater than 0.2 and false discovery rates (FDRs) less than 0.05 were considered statistically significant. To explore the prognostic value of cell population infiltration, the TCGA cohort (https: / / portal.gdc.cancer.gov / ) and the DKFZ-PRAD cohort (https: / / www.cbioportal.org / ) were divided into two groups based on the median infiltration level of each cell type. Kaplan-Meier survival curves were compared using a two-sided long-rank test.

[0057] (8) Evaluation of the association between NKFB signaling and lineage plasticity

[0058] For the single-cell transcriptome data of GEMMs, the inflammatory signal and lineage transition signature scores were calculated for each malignant cell, and then the cells were divided into 39 clusters at a resolution of 2 using the "FindClusters" function in Seurat. Spearman correlation analysis was performed to evaluate the relationship between the inflammatory signal and lineage transition signature scores of these clusters. For the single-cell transcriptome data of human CRPC patient samples, the TNFA_SIGNALING_VIA_NFKB signal, EMT, and stemness signature scores were calculated for each malignant cell, and then the cells were divided into 56 clusters at a resolution of 1.2 using the "FindClusters" function in Seurat. Spearman correlation analysis was performed to evaluate the relationship between the TNFA_SIGNALING_VIA_NFKB signal and EMT and stemness.

[0059] (9) Evaluation of the association between myeloid traits and lineage plasticity

[0060] Mono / Macro / DC signatures were selected from GEMMs, and myeloid markers were selected from human CRPC patients. Selection criteria included a fold change greater than 2, an adjusted P value less than 0.01, and a percentage of cells expressing greater than 0.1. The mean expression value of the signature genes was calculated based on RNAseq data from the TCGA to obtain a signature score. Spearman correlation analysis was then performed to assess the relationship between the Mono / Macro / DC or myeloid signature scores and the lineage plasticity signature scores.

[0061] (10) Development trajectory analysis

[0062] Monocle2 (v.2.12) was used to infer developmental trajectories. Differentially expressed genes (DEGs) were identified using the differential gene test function, and samples with a q value less than 1 × 10 -5 Cells were sorted based on pseudotime using the genes identified in the expression data. Dimensionality reduction was then performed using the DDRTree algorithm, which arranged cells along trajectories. Cell fates were mapped using Monocle3. The expression data were embedded using the Monocle function "reduce_dimension" with default parameters. Trajectory graphs were then inferred using the "learn_graph" function, with a minimum branch length of 15 and "close_loop" set to FALSE. Lineage trajectories were inferred using Slingshot software.

[0063] (11) cNMF analysis

[0064] It is well known that individual cells contain specific cellular functional programs. The cNMF algorithm was used to obtain consensus modules representing various cellular programs. To obtain the optimal solution, the cNMF algorithm was performed 100 times, with k values ​​ranging from 5 to 15 for each sample. The results showed that 9 gene expression programs (GEPs) were identified in PRP_16weeks_Intact, 4 in PR_24weeks_Intact, 3 in PR_30weeks_Intact, 4 in PR_47weeks_Intact, 3 in PRP_8weeks_Intact, 3 in PRP_9weeks_Intact, 4 in PRP_12weeks_Intact, 4 in PRP_12weeks_DHT, and 5 in PRP_12weeks_Intact. Then, each GEP gene set score was calculated for all malignant cells in each sample, and the GEP scores across samples were clustered. Consensus modules were identified using the "1-Pearson correlation coefficient" as the distance metric and "ward D2" linkage. To represent each GEP, the top 50 genes based on their unique expression were selected. GO analysis was performed on the top 50 genes of each GEP to further characterize these modules. Based on the RNA-seq data from TCGA, the average expression values ​​of the top 50 genes of the EMT program were divided into high-risk and low-risk groups. The Kaplan-Meier survival curves were compared using a two-sided long-rank test.

[0065] (12) Cell communication analysis

[0066] To investigate intercellular interactions between Mono / Macro / DC subsets and malignant cells in GEMM samples, cell-cell communication analysis was performed using the CellPhoneDB package. The significance of intercellular communication was determined based on the interactions and the normalized cell matrix obtained by scran normalization, with a significance threshold of P < 0.05. To determine the effects of potential ligands secreted by the VIM lineage on macrophages in GEMM samples, the regulatory network of the VIM lineage on macrophages was assessed using NicheNet. Macrophages from WT samples were considered as reference recipient cells.

[0067] (13) RNA velocity analysis

[0068] Using the Python package velocyto, we calculated RNA velocity values ​​for each gene in each cell, mapped RNA velocity vectors into a low-dimensional space, and visualized these vectors on UMAP projections.

[0069] (14) Copy number variation (CNV) analysis

[0070] The initial CNV values of each Adeno and VIM single cell were estimated using infercnv, with randomly selected endothelial cells and fibroblasts as reference cells. The initial CNV values were transformed according to the following criteria, and then the average CNV value of each chromosome was calculated: for values of 0.85 < obs < 0.9, set to 2; for values of 0.9 ≤ obs < 0.95, set to 1; for values of 0.95 ≤ obs < 1.05, set to 0; for values of 1.05 ≤ obs < 1.1, set to 1; for values of 1.1 ≤ obs < 1.15, set to 2.

[0071] (15) Mass cytometry

[0072] Tumor samples were collected from 7 CRPC patients (3 CRPC - Adeno, 4 CRPC - NEPC) for mass cytometry analysis, and the clinical information is shown in Table 1. Live / dead cell staining was performed using 2 μM cisplatin (Fluidigm), and then quenched with CSB (Fluidigm). Subsequently, the cells were stained and fixed with Fix - I buffer (Fluidigm), and then secondary staining was performed using the Cell-ID TM 20 - Plex Pd barcode kit (Fluidigm) to minimize sample cross - reactivity. According to the instructions, the antibody was combined with x8 polymer kit (Fluidigm). For surface protein staining, cell counting should be performed and diluted to 1×10 6 cells / ml and cultured with the antibody mixture. Subsequently, the cells were washed, permeabilized with 80% methanol, and stained with the intracellular antibody mixture. After three washes with CSB, 0.125 μM iridium intercalator was added to the Fix and perm buffer (Fluidigm) containing the cells and incubated overnight at 4°C. Before data acquisition, the samples were washed and resuspended in deionized water containing 10% EQ 4 - element beads (Fluidigm), and the cell concentration was adjusted to 1×10 6 cells / ml. Data acquisition was performed using a Helios mass cytometer (Fluidigm). The raw FCS data was normalized, and the.fcs file of each sample was collected. Subsequently, all.fcs files were uploaded to Cytobank for data cleaning, and the single live cell population was exported as an.fcs file for further analysis.

[0073] For CyTOF data processing and analysis, manual gating was performed to identify live, intact, single cells, followed by Boolean gating and debarcoding using Cytobank (www.cytobank.org). Cells meeting these criteria were downloaded as FCS files and compensated using an arcsinh transformation (scaling factor 5) for subsequent analysis. Major clusters were generated using FlowSOM / Consensus ClusterPlus in CATALYST and visualized using the UMAPs and ComplexHeatmap packages.

[0074] (16) Spatial transcriptome data processing

[0075] Spatial transcriptome data were diagnosed for tissues with coexisting NEPC and HSPC25 and were normalized and adjusted using Sctransform according to the tutorial manual provided by Seurat (https: / / satijalab.org / seurat / articles / spatial_vignette.html). GRN+Mono / Macro cells were defined in GEMM as macrophages expressing GRN genes at levels higher than the average expression value. Using COSG2, the top 20 signature genes of GRN+Mono / Macro were defined as the GRN+Mono / Macro signature. The signature score was calculated based on the average expression of all signature genes. The Spearman correlation between GRN and TAM signatures (CD163, MSR1, CSF1R) was calculated using the function “cor()” in R.

[0076] (17) Cell and organoid culture

[0077] Mouse prostate cancer RM-1 cells were obtained from Procell Life Science & Technology Co., Ltd (Wuhan, China) and cultured in RPMI 1640 medium containing 10% fetal bovine serum (FBS). RAW264.7 and HEK293T cells were obtained from the National Cell Resource Infrastructure (Beijing, China) and cultured in Dulbecco's Modified Eagle Medium (DMEM) containing 10% FBS in an incubator at 37°C, 5% CO2, and 100% humidity. Prostate tumor samples for organoid culture were obtained from three patients who underwent radical prostatectomy, and written informed consent was obtained from each patient. Human prostate cancer organoids were produced and cultured according to the manufacturer's instructions and cultured in human prostate cancer organoid medium (Cat#PRS-OCR-100, Cat#PRS-ODR-100, Cat#PRS-TDD-2-100, Cat#PRS-TCR-1-100, Cat#PRS-TDE-2-100, Cat#PRS-PRCM-3D-100, Precedo.Matrigel Cat#356231, BD).

[0078] (18) Cell and organoid viability

[0079] Cell viability was determined using CCK8. Cells were seeded in 96-well plates and treated with ligand recombinant proteins (recombinant mouse TNFSF12, Cat#50174-M15H, Sino Biological; recombinant mouse IL-1β protein, Cat#50101-MNAE, Sino Biological; secreted recombinant mouse GRN protein, Cat#50396-M08H, Sino Biological) at specific concentrations for a certain period of time. Organoids were seeded in 24-well plates and treated with enzalutamide (Cat#T6002, TargetMol) with or without recombinant human GRN protein (Cat#10826-H08H, Sino Biological). The viability of cells and organoids was determined using a CCK-8 kit (Cat#B34302, Selleck) according to the instructions.

[0080] (19) High-content imaging and analysis

[0081] Detection data and images were acquired using an Operetta CLS high-content imaging and analysis system (PerkinElmer, Waltham, MA) and a 5x air objective. Six fields of view were acquired per well. All images were analyzed using Harmony 5.2 software (PerkinElmer) to calculate growth area.

[0082] (20) shRNA knockdown and ligand gene overexpression

[0083] Hairpin shRNAs targeting the coding sequence (CDS) of mouse Tnfsf12, Il-1b, and Grn transcripts were inserted into the pLL3.7 vector (plasmid Cat#11795, Addgene). The specific targeting sequences used for each hairpin shRNA are shown in Table 2, and the overexpression sequences of the ligands are shown in Table 3. To package lentiviral particles, 2 μg of VSVG and 5 μg of PAX2 were added to 10 μg of pLL3.7 shRNA construct or overexpression construct and transfected into HEK293T cells. Transfection was performed using Lipofectamine 3000 (Cat#L3000015, Thermo Fisher) according to the manufacturer's instructions. 72 hours after transfection, the supernatant containing the virus was collected and filtered through a PVDF membrane filter unit with a pore size of 0.45 μm (Millipore). The virus was then concentrated by centrifugation at 18,000 rpm and 4°C for 2 hours. The resulting pellet was resuspended in DMEM medium.

[0084] Table 2 Specific targeting sequences of hairpin shRNA

[0085] shRNA Targeting sequence shGRN-1 CCTAGAATAACGAGCCATCAT(SEQ ID NO.1) shGRN-2 ACTCATCCTGAGTCACCCTAT(SEQ ID NO.2)

[0086] Table 3 Overexpression sequences of ligands

[0087]

[0088]

[0089] (21) RNA extraction and real-time PCR analysis

[0090] Total RNA was isolated using an RNA extraction kit (Cat# DP451, TIANGEN) according to the instructions, and the Reverse transcription was performed using All-in-One First-Strand cDNA Synthesis SuperMix (Cat# AT341-01, TransGen). RT-qPCR was performed in triplicate using SYBR Green PCR Master Mix (Cat# A57156, Thermo Fisher Scientific). The primer sequences used are shown in Table 4.

[0091] Table 4 Primer sequences for RT-qPCR

[0092]

[0093] (22) Co-culture of macrophages and RM-1 cells

[0094] Treated mouse macrophages, RAW264.7, were seeded into 0.4 μm pore Transwell inserts (Cat#3412, Corning) and cocultured with RM-1 cells cultured in 6-well plates at a 1:1 ratio. After 72 hours of coculture, the Transwell inserts containing RAW264.7 cells were removed, and the RM-1 cells were harvested for subsequent analysis.

[0095] (23) Western blotting

[0096] The cells were lysed on ice for 20 minutes in stringent RIPA lysis buffer containing a protease inhibitor cocktail (Cat#B14001, Selleck), and then centrifuged at 12,000 rpm for 10 minutes at 4°C. The collected cell supernatant was suspended in SDS loading buffer and incubated at 100°C for 10 minutes. The primary antibodies used were rabbit anti-TNFRSF1A (Cat#21574-1-AP, Proteintech, 1:1000), rabbit anti-N-cadherin (Cat#13116, Cell Signaling Technology, 1:1000), rabbit anti-vimentin (Cat#5741, Cell Signaling Technology, 1:1000), rabbit anti-E-cadherin (Cat#3195, Cell Signaling Technology, 1:1000), rabbit anti-ZO-1 (Cat#8193, Cell Signaling Technology, 1:1000), and rabbit anti-GAPDH (Cat#5174, Cell Signaling Technology, 1:5000). Secondary antibodies conjugated to HRP were used at a concentration of 1:5000.

[0097] (24) Multiple immunofluorescence staining

[0098] Multiplex immunofluorescence staining of prostate cancer tissues from TRAMP mice for Epcam, CSF1, F4 / 80, GRN, and vimentin was performed using the Opal 7-Color Automation IHC Kit (Cat#NEL821001KT; Akoya Biosciences). Briefly, 5 μm formalin-fixed, paraffin-embedded (FFPE) sections were stained sequentially according to the reference protocol. All antibodies and their respective fluorochromes were as follows: Epcam (1:500, Dye Opal 480, Cat#ab213500, Abcam), F4 / 80 (1:200, Dye Opal 570, Cat#70076, Cell Signaling Technology), and vimentin (1:200, Dye Opal 690, Cat#5741, Cell Signaling Technology). Multispectral images of stained sections were scanned at high magnification (40×, 0.25 μm / pixel) using the PhenoImager HT system from Akoya Biosciences.

[0099] Test results:

[0100] (1) Dynamic tumor microenvironment (TME) composition promotes cellular heterogeneity in prostate cancer (the experimental method involved is (1) in the experimental method)

[0101] To investigate the role of TME composition in lineage plasticity, we reanalyzed scRNA-seq data from 29 mice, including 9 wild-type (WT), 7 Pten- / -Rb- / - (PtR), and 13 Pten- / -Rb- / -Trp53- / - (PtRP), which recapitulate the transition from prostate cancer to neuroendocrine prostate cancer (NEPC) (see Figure 1 A). We integrated 29 scRNA-seq datasets and manually annotated these single cells into 15 different cell lineages (see Figure 1 B in Figure 9 The cellular composition at different time points has been investigated, particularly during the transition from adenocarcinoma to NEPC (see Figure 1 C in ). It was found that Mono / Macro / DC increased significantly in the early stage (PRP_8-9weeks_Intact), but decreased in the later stage (PRP_12-16weeks_Intact) (see Figure 1D in Figure ). We further investigated the differentiation trajectory of cancer cells and identified Adeno as an early lineage stage, TFF3, VIM, and POU2F3 as intermediate lineage stages, and NEPC cell types as late lineage stages by pseudotime analysis (see Figure ). Figure 1 E in ). The trend of lineage transitions is consistent with the timing of cancer evolution (see Figure 1 F in ), which is also consistent with the changes in cell composition during tumorigenesis (see Figure 1 G). These results indicate that the composition of malignant and non-malignant cells exhibits significant dynamic changes during lineage transition.

[0102] (2) Tumor heterogeneity across the prostate cancer spectrum (involving experimental methods (2), (11), and (13) in the experimental methods section)

[0103] We then investigated the tumor heterogeneity of malignant cells representing five different cell lineages. Unsupervised hierarchical clustering analysis revealed that among the five malignant lineages, the VIM lineage exhibited a unique transcriptome pattern ( Figure 2 Meanwhile, based on TCGA RNA-seq data, VIM lineage signatures were associated with poor patient survival ( Figure 2 To identify the molecular pathways activated in the VIM lineage, GSEA analysis was performed, which revealed significant enrichment in inflammatory response, cytokine-receptor interaction, and EMT signaling in the VIM lineage ( Figure 2 D in Figure 1). EMT features have been shown to possess stem cell-like properties, allowing for redifferentiation into new lineages. Using EMT, stemness, and NF-κB signaling signatures to score cancer cells, it was observed that the VIM lineage exhibited mesenchymal and stem cell-like properties, accompanied by overactivation of NF-κB signaling ( Figure 2 E), which is consistent with the stage of cancer progression ( Figure 2 F), similar to previously reported drug-persistent mesenchymal stem-like prostate cancer cells (MSPC). The dynamic transdifferentiation model proposes that adenocarcinoma cells undergo a partial EMT and transition to a stem-like state before redifferentiating into new lineages, such as the NE lineage. Similarly, RNA velocity analysis consistently revealed a trajectory linking the VIM lineage to the Adeno and NEPC lineages ( Figure 2 C in ). Using the cNMF algorithm, malignant cells were further decomposed into nine consensus modules, each of which was annotated with specific biological functions ( Figure 2G in Figure 5). Importantly, EMT, biosynthesis, cell cycle, and NE2 correspond to cancer lineage states and are projected onto two-dimensional branches. Interestingly, the Adeno and TFF3 lineages are primarily enriched in the biosynthesis and cell cycle modules, while the VIM lineage is enriched in the EMT module. In contrast, POU2F3 and NEPC cell types are enriched in the NE2 module, which exhibits lineage-specific biological functions in distinct subpopulations ( Figure 2 Based on TCGA RNA-seq data, the characteristics of the EMT module are associated with poor prognosis ( Figure 2 Consistent with the enrichment of inflammatory signaling pathways in the VIM lineage, a significant positive correlation was observed between the EMT module and NF-κB signaling ( Figure 2 These observations characterize a unique VIM lineage that exhibits a mixture of mesenchymal and stem cell-like properties accompanied by activated NF-κB signaling.

[0104] (3) Infiltration of GRN+ macrophages is associated with lineage transition (the experimental methods involved are (5), (6), (7), (8), (9), and (10) in the experimental methods section)

[0105] During lineage transition, the composition of Mono / Macro / DC changes significantly (see Figure 1 D in Figure 3). To explore the interplay between TME composition and malignant cells, we used CellPhoneDB to infer receptor-ligand (RL) interactions. The analysis revealed that Mono / Macro / DCs are the cell types that primarily secrete ligands targeting malignant cells, particularly those of the VIM lineage ( Figure 3 A in Figure ), suggesting that Mono / Macro / DC may play a key role in shaping the VIM lineage. To determine the relationship between cell populations in prostate cancer, CIBERSORTx was used to predict the abundance of cell subtypes in the TCGA cohort, the DKFZ-PRAD cohort, and other public bulk RNA-seq and microarray datasets. Spearman correlation analysis was used to examine the infiltration patterns of 10 major cell types in six independent prostate cancer cohorts. A significant positive correlation was observed between Mono / Macro / DC and the VIM lineage in most cohorts ( Figure 10 In addition, a significant positive correlation was found between Mono / Macro / DC characteristics and plasticity traits, especially VIM lineage characteristics ( Figure 10 Notably, the abundance of Mono / Macro / DC cells was significantly increased in CRPC patients compared with adenocarcinoma patients and controls ( Figure 10To assess the clinical relevance of cell type infiltration in the TME, we analyzed the association between Mono / Macro / DC infiltration and disease-free survival (DFS) in the TCGA and Dkfz cohorts. In these patient cohorts, Mono / Macro / DC infiltration was significantly associated with poor DFS ( Figure 10 D). These results indicate that Mono / Macro / DC play a key role in regulating the malignant cell state and generating the VIM lineage.

[0106] Re-clustering and classification of Mono / Macro / DC into 13 subgroups ( Figure 3 B in Figure 11 AC in PtRP mice). Compared with WT mice, the composition of Mono / Macro / DC in PtRP mice was positively correlated with the early progression stage and expanded in PRP_8weeks_Intact ( Figure 11 D in Figure 3). We then performed differentially expressed gene (DEG) analysis of Mono / Macro / DC paracrine inflammatory genes between PRP_8-9weeks_Intact and WT, and between PRP_12-16weeks_Intact and PRP_8-9weeks_Intact. We found that some paracrine genes, such as Spp1 and Grn, were uniquely expressed by Mono / Macro / DC during the lineage transition stage and decreased at the end of reprogramming ( Figure 3 C and D in Figure ). High expression of SPP1 is associated with vascular-related genes and has been shown to be significantly associated with poor survival prognosis in multiple types of tumors. GRN, encoded by the Grn gene, is secreted by inflammatory cells in wound response and is involved in tumorigenesis and tumor progression. Nielsen and Quaranta et al. demonstrated that GRN promotes the activation of mesenchymal programs, leading to immune checkpoint resistance and metastasis in pancreatic cancer. Consistent with the results of DEG analysis, GRN expression was upregulated in Mono / Macro / DC during lineage progression and downregulated in the terminal stage ( Figure 3 E), suggesting that GRN may play a role in driving the EMT program and the development of the VIM lineage. Furthermore, using CellPhoneDB to assess the ligand-receptor interaction between the Mono / Macro / DC subset and the five malignant lineages, we found that this interaction is mediated by the GRN-TNFRSF1A interaction axis ( Figure 3 F in Figure ). Consistent with the CellPhoneDB results, Mono / Macro / DC expressed the highest level of GRN in the TME composition ( Figure 3 The G in Figure 11 E in ), and the Tnfrsf1a receptor is mainly expressed in the VIM lineage ( Figure 3 F in Figure 11 In addition, to identify myeloid subsets that highly express Grn, high expression of Grn was observed in macrophages ( Figure 3 H in ), macrophages constitute the major myeloid cell subset expressing GRN ( Figure 3 It is noteworthy that Grn is highly expressed mainly in FCNA+ macrophages ( Figure 3 These macrophages exhibit MDSC-like characteristics and play a tumor-promoting role in colon cancer, liver cancer, and lung cancer. In the macrophages of the PRP group, Grn was mainly expressed at high levels ( Figure 3 K in this group), the tissues in this group showed higher malignancy compared with the PR and WT groups, indicating that macrophages secrete GRNs to promote cancer progression in a highly malignant microenvironment, consistent with previous findings in the VIM lineage. Next, the trajectories of CRPC-Adeno and NEPC were analyzed to obtain dynamic signaling scores of transition states along the lineage. Plasticity characteristics such as stemness and EMT scores and TNFRSF1A expression were highest in the intermediate stage ( Figure 3 M in the middle stage), especially TNFA_SIGNALING_VIA_NFKB signaling showed the highest expression score in the middle stage ( Figure 3 In addition, inflammatory and lineage transition signaling scores were calculated for each malignant cell and grouped into 39 clusters. The TNFA_SIGNALING_VIA_NFKB signaling was significantly positively correlated with the plasticity signature score and CD44 expression score, but significantly negatively correlated with the AR signature score. NF-κB signaling has been reported to activate genes that control cancer stemness and motility and is associated with poor survival in prostate cancer patients. These data highlight the important role of macrophage-derived GRNs in driving malignant lineage states.

[0107] (4) GRN+ macrophage-derived GRN mediates EMT and drives resistance to AR-targeted therapy (for experimental methods involved, see (17), (18), (19), (20), (21), (22), and (23) in the experimental methods section)

[0108] To verify whether GRN can enhance the survival of RM-1 cells, RM-1 cells were treated with recombinant mouse Grn protein (purchased, Cat#10826-H08H, Sino373 Biological) for 24 h, 48 h, and 72 h. CCK8 assay showed that the OD value of RM-1 cells treated with recombinant mouse Grn protein was significantly higher than that of control cells (P < 0.01, Figure 4Subsequently, RAW264.7 and RM-1 cells were co-cultured in a non-contact Transwell system, which creates an in vivo-like cell culture environment and allows cells to carry out metabolic activities in a more natural manner without cell contact ( Figure 4 In order to test whether Grn+ macrophages can promote the EMT program of RM-1 cells, a Grn overexpression plasmid was constructed and transfected into mouse macrophage RAW264.7 cells. RT-qPCR confirmed that Grn mRNA was significantly overexpressed in RAW264.7 cells ( Figure 4 C). RM-1 cells co-cultured with Grn+macrophages were significantly longer than those in the control group ( Figure 4 G in the figure). Western blot showed that Grn+ macrophages significantly inhibited the expression of epithelial markers E-cadherin and ZO-1, and upregulated the expression of mesenchymal marker Vimentin ( Figure 4 In contrast, lentivirus carrying Grn shRNA was transfected into RAW264.7 cells, and RT-qPCR confirmed that lentivirus-mediated Grn shRNA significantly downregulated Grn mRNA in RAW264.7 cells ( Figure 4 E), the results showed that Grn-macrophages increased the expression of epithelial proteins E-cadherin and ZO-1, but suppressed the expression of mesenchymal markers Vimentin and N-Cad ( Figure 4 F in Figure ). Taken together, these macrophage-derived GRNs promote the EMT program in prostate adenocarcinoma cells. To explore the signaling pathways activated in RM-1 cells when co-cultured with GRNs and macrophages, large-scale RNA sequencing was performed. This is consistent with previous findings in the VIM lineage ( Figure 2 D), GSEA analysis showed that inflammatory pathways were significantly activated in RM-1 cells co-cultured with GRN+macrophages, especially the NF-κB signaling pathway ( Figure 4 In conclusion, GRN+ macrophages induce EMT and morphological changes by activating the NF-κB signaling pathway in prostate adenocarcinoma cells.

[0109] To determine whether GRN can promote resistance to AR-targeted drugs in prostate cancer patients, three patient-derived prostate cancer organoids were established. Recombinant GRN and a clinical-grade AR signaling inhibitor (enzalutamide) were used to investigate the functional impact of GRN on the efficacy of AR-targeted drugs. The study found that enzalutamide significantly inhibited the growth and activity of organoids ( Figure 4 H, I, J, and K in Figure ). In contrast, GRN promoted resistance to enzalutamide and maintained organoid number and viability ( Figure 4H, I, J, and K in Figure 5). This reveals a potential role for GRN in enhancing resistance to AR-targeted drugs.

[0110] (5) Key TME components and lineage programs were validated in human CRPC (for experimental methods, see (3) in the Experimental Methods section)

[0111] To investigate the relationship between transcriptional changes and plasticity progression in human prostate cancer, we reanalyzed scRNA-seq data from three CRPC-Adeno and four CRPC-NEPC patients from a previous study. Single cells that passed all quality control filters were divided into twenty clusters and manually annotated into five known cell types ( Figure 12 A in Figure 5 As hypothesized, myeloid cells in human samples specifically expressed GRN ( Figure 5 C in ). Then, malignant cells were extracted and divided into 13 different clusters ( Figure 12 B), further annotated into three major cell types: malignant epithelial cells, mixed lineages, and NEPCs ( Figure 5 B). Mixed lineage cells specifically express the receptor TNFRSF1A ( Figure 5 C in the figure) and highly expressed multiple EMT, stemness, and NE genes, including VIM, TGFB2, KLF4, CD44, FOXA2, GHGB, and POU3F2 ( Figure 5 C and D in Figure 12 B), indicating that they are in a mixed-lineage state with multi-lineage potential. Consistent with the scRNA-seq findings in GEMMs, a significant positive correlation between myeloid cells and mixed lineages was observed in most patient cohorts ( Figure 13 Myeloid cells were also positively associated with plasticity traits, particularly VIM lineage traits ( Figure 13 In addition, the abundance of myeloid cells is significantly increased in CRPC patients ( Figure 13 C in ), which was associated with poor overall survival in patients in the TCGA and Dkfz cohorts ( Figure 5 The E in Figure 13 D in Figure 3). Inflammatory response and EMT markers were also found to be significantly enriched in the mixed lineages ( Figure 5 Using Monocle3, we inferred lineage states and confirmed that the hybrid lineages were intermediate lineages based on pseudo-time ordering ( Figure 14 The proportion of mixed-lineage cells initially increased but then decreased, while the proportion of NEPC cells gradually increased and the proportion of Adeno cells gradually decreased along the lineage progression ( Figure 5H in ). Meanwhile, we found that TNFRSF1A expression and AR, NEPC, EMT, stemness, and TNFA / NF-κB signature scores were consistent with those found in GEMMs and showed consistent trends with pseudo-time sorting ( Figure 14 C in the figure), and the NECP lineage was found to be derived from a mixed lineage in mouse tissues ( Figure 14 The highest levels of EMT, stemness, and TNFA / NF-κB program scores were observed in mixed lineages, whereas AR and NEPC signature scores in mixed lineages showed intermediate levels between malignant epithelial and NEPC ( Figure 5 In addition, the TNFA / NF-κB signaling, EMT, and stemness characteristics of malignant cells were calculated and classified into 56 clusters. The scatter plot of these clusters showed a significant positive correlation between TNFA / NF-κB signaling and EMT and stemness ( Figure 5 These findings are consistent with observations from scRNA-seq of GEMMs, indicating that the EMT program and NF-κB signaling play a key role in lineage switching in human CRPC.

[0112] (6) Granulin-NFκB-CSF1 axis promotes macrophage-induced lineage switching (for experimental methods, see (10) in the experimental methods section)

[0113] To investigate the origins of human admixed lineages, we further clustered the admixed lineages and identified seven distinct cell populations ( Figure 6 Using monocle2 and CytoTRACE, we confirmed that the C1 population exhibited the highest stem cell properties and represented a mixed lineage origin ( Figure 6 Scoring of human mixed lineages revealed that the C1 population exhibited mixed mesenchymal and stem cell-like features with highly activated NF-κB signaling, similar to the VIM lineage in GEMMs, suggesting that this cell state has the potential for multilineage redifferentiation into other lineages ( Figure 6 Interestingly, cluster C1 was enriched for cytokine production and chemotaxis in GO analysis ( Figure 6 E), suggesting that activated NF-κB signaling in this cell state may induce the production of cytokines and chemokines, attracting and taming inflammatory cells to maintain a pro-tumor TME.

[0114] CellphoneDB analysis was performed to determine which cytokines are released by the VIM lineage that educated macrophages in the TME. It was found that the VIM lineage interacted with macrophages more frequently in the tumor group compared with the WT group, among which CSF1_CSF1R and several chemokine receptor interactions, such as CCL3_CCR5 and CCL3_CCR1, were significantly enriched in the tumor group ( Figure 6 F). CCL3 has been reported to recruit macrophages, whereas CSF1 has been reported to promote M2 macrophage polarization. Consistent with this, VIM lineages showed high levels of CSF1 and CCL3 expression in malignant and adenoid cells of GEMMs ( Figure 6 G in ). Meanwhile, CSF1 and CCL3 are also highly expressed in the C1 group of the mixed CRPC lineage in humans ( Figure 6 Next, we investigated the expression of Csf1r receptor in Mono / Macro / DC and found that Csf1r was mainly expressed in macrophages ( Figure 6 I and J in Figure 3), especially in FCNA+ macrophages that highly express GRN ( Figure 6 K), indicating that FCNA+ macrophages with high expression of GRN are the main receptor cells that respond to the overexpression of the ligand CSF1 by malignant cells.

[0115] We further investigated the key regulatory molecules and downstream activation targets from the VIM lineage to macrophages. Using NicheNet, we found that the VIM lineage exhibited elevated CCL3 ligand activity, interacting with macrophage-encoded receptors Ccr1 and Ccr5. Interestingly, the VIM lineage also showed high Csf1 ligand activity, binding to the Csf1r receptor, leading to increased expression of the target gene Grn in macrophages ( Figure 6 In addition, the top 100 predicted target genes in macrophages were enriched in TNF signaling pathway, IL-17 signaling pathway and NF-κB signaling pathway ( Figure 6 Previous studies have shown that IL-17 and NF-κB signaling can induce an M2 phenotype. Together, these findings suggest a positive feedback loop between GRN+ FCNA+ macrophages and the CSF1+ VIM lineage, promoting malignant lineage switching and macrophage recruitment and polarization.

[0116] (7) Pharmacological blockade of the CSF-1 / CSF-1R axis can inhibit the formation of the VIM lineage (for experimental methods involved, see (14), (24) and (4) in the experimental methods section)

[0117] Based on previous findings, we wanted to investigate whether disrupting the positive feedback loop could inhibit prostate cancer lineage transition. TRAMP mice were treated with BLZ945, a small molecule inhibitor of CSF-1R, for 7 days. Then, prostate tumor tissues from placebo-treated and BLZ945-treated (tail vein injection) were collected for scRNA-seq ( Figure 7 Based on typical gene markers, the transcriptome was classified into 10 major cell types ( Figure 7 B in Figure 15 Using ClusterFoldSimilarity, we calculated the similarity scores of cell populations and found that the VIM lineages of GEMMs and TRAMP mice showed extremely high similarity ( Figure 7 C in Figure 3). To confirm whether the VIM lineage originated from malignant cells, we re-clustered the Adeno and VIM lineages and calculated large-scale chromosomal copy number variations (CNVs) using Infercnv. We found that the VIM lineage and all Adeno populations showed significantly higher CNV levels than random control cells ( Figure 15 D), indicating that the VIM lineage in these TRAMP mice may be converted from malignant cells. It is important to note that according to TCGA RNA-seq data, VIM lineage characteristics are significantly associated with poor clinical prognosis (P = 0.03, Figure 7 Compared to Adeno, the VIM lineage significantly upregulated genes related to myeloid growth factors (Csf1), chemokines (Ccl2), and mesenchymal markers (Vim, Col1a2, Col3a1) ( Figure 7 E in the figure). Biomarker genes of the VIM lineage are enriched in GO terms such as inflammatory response, angiogenesis, and interferon response ( Figure 7 Furthermore, the VIM lineage possesses mesenchymal and stem cell-like features and has activated NF-κB signaling ( Figure 7 In conclusion, the VIM lineage in TRAMP mice shares similar biological features with that in GEMMs.

[0118] As expected, Csf1 was expressed primarily in the VIM lineage in all cell populations ( Figure 7 H), BLZ945-treated TRAMP mice showed reduced enrichment of the VIM lineage and decreased Csf1 expression ( Figure 7 I and J in Figure 3), Mono / Macro / DC expressed the highest levels of Csf1r and Grn among all cell populations ( Figure 7 K and L in ). Mono / Macro / DC are re-clustered into 7 groups ( Figure 7 M in tumor tissues), BLZ945 reduced the enrichment of Grn+ macrophages in tumor tissues ( Figure 7N in ), and reduced the expression of Grn ( Figure 7 O in the middle). These results indicate that BLZ945 disrupts the positive regulatory loop, leading to reduced macrophage Grn expression and thus inhibiting VIM lineage formation. By integrating scRNA-seq data from blood and tumor tissues, ancestor-descendant relationships can be inferred. Using Velocyto, the trajectory map of Mono_Cd14+_Grn+ and Macro_Cd14+_Grn+ was revealed ( Figure 7 P in the tumor). Tumor-infiltrating Mono_Cd14+_Grn+ differentiates into Macro_Cd14+_Grn+ and upregulates the expression of the Grn gene ( Figure 15 Further multiple immunofluorescence staining of tumor samples from TRAMP mice confirmed that macrophages and VIM+ cells were spatially close in many sites ( Figure 7 Q in ).

[0119] (8) Mass spectrometry cytometric analysis of CRPC-adenocarcinoma and CRPC-NEPC patient tissues (for experimental methods involved, see (15) and (16) in the experimental methods section)

[0120] To verify the above findings at the single-cell protein level, CRPC patients (n=3) and NEPC patients (n=4) were recruited, and tumor tissues were collected for time-of-flight mass spectrometry cytometry analysis (CyTOF) ( Figure 8 Low-quality cells were excluded using Cytobank (www.cytobank.org), and the remaining cells were clustered using FlowSOM based on a 41-parameter panel. CyTOF analysis identified 10 major cell types, which were visualized using t-SNE projections ( Figure 8 B in Figure 16 A mixed lineage was discovered that highly expressed EMT and stem cell markers, including Snail.1, Vimentin, ZEB2, and OCT3-4 ( Figure 8 C), which is consistent with scRNA-seq ( Figure 3 ) and human CRPC ( Figure 5 ) were similar in VIM lineage and mixed lineage. In addition, the proportion of cells of Mono / Macro and mixed lineages was increased in NEPC patients ( Figure 8 D). The expression levels of epithelial (EpCAM) and NE (Synatophys) markers in the mixed lineage are between those of malignant epithelial and NEPC, but the expression levels of EMT markers (Snail.1, Vimentin, ZEB2) and stem cell markers (OCT3-4) are the highest ( Figure 8E in ). Notably, the NF-κB activating factor NFkB.p65 was significantly overexpressed in the mixed lineage ( Figure 8 E in Mono / Macro), while granulin is highly expressed ( Figure 8 In short, these observations are consistent with the results of scRNA-seq data analysis.

[0121] To further validate the above findings at the spatial level, we obtained spatial transcriptome data of tissues defined as NEPC coexisting with HSPC25. These cells were classified into five groups based on gene markers, AR signatures, and NEPC signatures ( Figure 8 G and H in Figure 3). Tumor-associated macrophage marker genes such as MSR1, CSF1R, and CD163 are highly expressed in the NEPC_TME region ( Figure 8 In addition, GRN expression, GRN+Mono / Macro, TNFA_SIGNALING_VIA_NFKB and stem cell characteristics were enriched in NEPC_TME, showing significant spatial colocalization ( Figure 8 I in Figure 15 E in ). Then, we integrated the scRNA-seq and spatial transcriptome data of GEMM using MIA, and the results showed that GRN_high Mono / Macro and VIM were enriched in NEPC_TME regions, while GRN-low Mono / Macro was enriched in non-NEPC_TME regions ( Figure 15 Together, these data suggest that GRN+ macrophages and the VIM lineage are enriched in the NEPC microenvironment and contribute to transformation-induced plasticity.

[0122] It is well known that the cell state in the tumor phenotype is affected by both cell-intrinsic factors (genetic and epigenetic changes) and cell-extrinsic factors (autocrine, paracrine). The present invention reveals significant changes in the cell composition in the TME and the transition to multiple malignant lineages during the progression of prostate cancer. A key GRN-positive macrophage population was discovered. These cells secrete GRN and transform tumor cells into a multi-lineage state by activating NF-κB signaling. Subsequently, the multi-lineage state induces macrophages to express high levels of GRN by releasing CSF1, establishing a positive feedback loop for intercellular communication ( Figure 8 J in ).

[0123] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. Use of a GRN-NFκB-CSF1 loop inhibitor in the preparation of a product for inhibiting prostate cancer lineage plasticity, characterized in that: The inhibitor is shRNA, and the nucleotide sequence of the shRNA is shown in SEQ ID NO. 1 or SEQ ID NO.

2.

2. Use of a GRN-NFκB-CSF1 loop inhibitor in the preparation of a product for treating prostate cancer, characterized in that: The inhibitor is shRNA, and the nucleotide sequence of the shRNA is shown in SEQ ID NO. 1 or SEQ ID NO.

2.

3. The use according to claim 2, characterized in that The prostate cancer is androgen receptor targeted therapy-resistant prostate cancer.

4. A shRNA that inhibits prostate cancer lineage plasticity, characterized in that The nucleotide sequence of the shRNA is shown in SEQ ID NO. 1.

Images