Loss of androgen receptor signaling in prostate cancer-associated fibroblasts (CAFs) promotes CCL2- and CXCL8-mediated cancer cell migration

Abstract

Fibroblasts are abundantly present in the prostate tumor microenvironment (TME), including cancer-associated fibroblasts (CAFs) which play a key role in cancer development. Androgen receptor (AR) signaling is the main driver of prostate cancer (PCa) progression, and stromal cells in the TME also express AR. High-grade tumor and poor clinical outcome are associated with low AR expression in the TME, which suggests a protective role of AR signaling in the stroma against PCa development. However, the mechanism of this relation is not clear. In this study, we isolated AR-expressing CAF-like cells. Testosterone (R1881) exposure did not affect CAF-like cell morphology, proliferation, or motility. PCa cell growth was not affected by culturing in medium from R1881-exposed CAF-like cells; however, migration of PCa cells was inhibited. AR chromatin immune precipitation sequencing (ChIP-seq) was performed and motif search suggested that AR in CAF-like cells bound the chromatin through AP-1-elements upon R1881 exposure, inducing enhancer-mediated AR chromatin interactions. The vast majority of chromatin binding sites in CAF-like cells were unique and not shared with AR sites observed in PCa cell lines or tumors. AR signaling in CAF-like cells decreased expression of multiple cytokines; most notably CCL2 and CXCL8 and both cytokines increased migration of PCa cells. These results suggest direct paracrine regulation of PCa cell migration by CAFs through AR signaling.

Keywords: CCL2; CXCL8; EMT; androgen receptor; cancer cell migration; cancer-associated fibroblasts; prostate cancer.

Figure 1

Stromal androgen receptor (AR) expression in PCas is associated with Gleason score and metastatic disease. (A) Immunohistochemistry staining for AR (nuclear; brown) in human PCa (left of the red boundary) and stroma (top). Double staining for AR (nuclear; purple) and the fibroblast marker PDGFRβ (cytosol; brown) (bottom). Insets show magnification of the stromal area. Arrows indicate PDGFβ‐positive fibroblasts with nuclear AR staining. (B) Percentage of AR‐positive cells in the tumor‐associated stroma and stroma in a healthy region of prostatectomies with tumors with a high (≥ 8) Gleason score, compared to tumors with an intermediate (7) Gleason score (top; n = 11; error bars represent SEM *P = 0.032). Percentage of AR‐positive cells in the tumor‐associated stroma and stroma in a healthy region of prostatectomies with tumors associated with metastases to locoregional lymph nodes compared to tumors without metastases (bottom; n = 19; error bars represent SEM; **P = 0.007, ***P = 0.004).

Figure 2

Patients and fibroblast characteristics. (A) Characteristics of the three patients of whom fibroblasts were cultured from biopsies of a cancer‐affected side of the prostates. Side selection for taking biopsies was based on the highest proportion of tumor‐containing diagnostic biopsies, multiparametric MRI images, and palpation of the tumor. (B) Representative phase‐contrast image of cells isolated from human PCa specimens shows fibroblast‐like morphology. (C) Copy number analysis in PCDF‐1 and PCDF‐2 cells (top) and two representative PCas (bottom).

Figure 3

PCa‐derived fibroblasts have CAF‐like features and express functional AR. (A) Immunohistochemical staining for the fibroblast marker PDGFRβ (top) and AR (bottom), in PC346C PCa cells (left), PCa‐derived PCDF‐1 fibroblasts (middle), and SW1573 lung cancer cells. (B) Western blot for PDGFRβ, SMA‐α, AR, Vimentin, PSA, and β‐actin expression in SW1573 lung cancer cells, U937 histiocytic lymphoma cells, hTERT‐BJ fibroblasts and PCDF 1, 2, 3 cells. (C) Chromatin fractionation of hormone‐deprived PC346C PCa cells and PCDF‐1 fibroblasts, treated for 4 h with R1881 or DMSO control, and AR is stained. Histone 3 is used as loading control. (D) Scratch assay in human PCa CWR‐R1 cells. Cells cultured in CM of CAF‐like cells stimulated with vehicle, R1881 alone, or R1881 in combination with RD162. (E) Quantification of the scratch assay. Error bars show standard deviation of three replicates. Percentage of repopulation of the scratch surface after 96 h of culturing. ** represents P < 0.01.

Figure 4

Androgen receptor occupies distinct chromatin sites in PCa‐derived fibroblasts. (A) Venn diagrams depicting overlap of AR (top) and H3K27Ac (bottom) binding sites in PCa‐derived fibroblasts, under vehicle and R1881 conditions. (B) Genome browser snapshot of AR and H3K27Ac ChIP‐seq in PCa‐derived fibroblasts. (C) Correlation heatmap of AR and H3K27Ac peaks, using supervised hierarchical clustering. (D) Heatmap depicting AR binding sites in PCa‐derived fibroblasts, LNCaP cells, and prostate tumors. All AR sites found in fibroblasts shown, grouped into ‘fibroblast‐unique’ and ‘shared’ sites, overlapping with LNCaP and PCa cells. Data are centered on the top of the AR peak within a 5‐kb window, where all data are vertically aligned. (E) Genomic distributions of AR binding sites relative to the most‐proximal gene, unique for fibroblasts (top), or shared between fibroblasts and PCa cells (bottom). (F) Motif analyses for the two separate AR peak subsets. Shared between fibroblasts and PCa cells (left) and unique for fibroblasts (right). (G) Scatter plot depicting enrichment scores for transcription factor overlap with ReMap analysis and scores from motif analysis.

Figure 5

Effect of AR actions on gene expression. (A) Gene ontology terms analysis for biological process of differentially expressed genes in CAF‐like cells upon testosterone stimulation. (B) Ingenuity pathway analyses of differentially expressed genes suggest decreased expression of cytokines critically involved in cancer cell functions, such as cell migration and chemotaxis. (C) mRNA downregulation of CCL2 and CXCL8 upon R1881 stimulation. (D) AR and H3K27ac binding sites in CCL2 and CXCL8 gene regions. AR shows specific bindings upon R1881 stimulation.

Figure 6

AR signaling in fibroblasts reduces PCa cell migration. (A) Decreased CCL2 and CXCL8 at the protein level upon R1881 stimulation for 24 h. Addition of RD162 restored the levels to unexposed cells, suggesting an AR signaling‐dependent regulation of cytokine expression. Average of three experiments. Error bars show standard deviation. (B) Scratch assay in CWR‐R1 cells. Addition of CCL2 or CXCL8 cytokines strongly increased cell migration at 1 pg·mL⁻¹ (left). Quantification of the scratch assay (right). Percentage of repopulation of the scratch surface after 96 h of culturing. Error bars show standard deviation of three replicates. (C) Transwell migration assay. The migration of CWR‐R1 cells induced by fibroblast‐CM was reduced when αCCL2‐ and/or αCXCL8‐neutralizing antibodies were added in the lower chamber of the transwell. Normal medium (NM) was used as control. A representative of two independent experiments is shown. Error bars show standard deviation, and *, **, ***, **** represents P value < 0.05, < 0.01, < 0.001, < 0.0001, respectively. (D) Transwell invasion assay. The invasion of CWR‐R1 cells induced by fibroblast‐CM was reduced when αCCL2‐and/or αCXCL8‐neutralizing antibodies were added in the lower chamber of the transwell. Normal medium (NM) was used as control. A representative of two independent experiments is shown. Error bars show standard deviation, and *, **, ***, **** represents P value < 0.05, < 0.01, < 0.001, < 0.0001, respectively.