Categories
Prostate cancer

Stromal androgen signaling acts as tumor niches to drive prostatic basal epithelial progenitor-initiated oncogenesis

Loss of stromal AR in Gli1-lineage cells diminishes prostatic epithelial oncogenesis

Using in vivo tissue recombination assays, we first examined if stromal AR action in Gli1-lineage cells acts as a tumor niche to support prostate epithelial oncogenesis15. The urogenital sinus mesenchyme (UGM) of either ArL/Y:Gli1CreER/+ (ARKO) mice14, with tamoxifen (TM) induced Ar deletion in Gli1-expressing cells, or Gli1CreER/+ controls, and the urogenital sinus epithelium (UGE) of Ctnnb1L(ex3)/+:PBCre4 mice, a prostate cancer model, with stabilized β-catenin expressed in prostate epithelium were isolated at E16.5, and transplanted together under the renal capsule of NOD/SCID mice (Fig. 1a). Gross analyses after 12-week implantation showed noticeable impaired growth in grafts combined with AR-deficient UGM and Ctnnb1L(ex3)/+:PBCre4 UGE in comparison to counterparts with control UGM, which were transparent and significantly heavier (p = 0.012) and larger than the ARKO grafts (left panel, Fig. 1b, c, d vs 1h, i). Pathological lesions resembling typical mouse prostatic intraepithelial neoplasia 3–4 (mPIN3 to 4)16 appeared in the control UGM combined grafts (Fig. 1e). Immunohistochemical analyses (IHC) showed positive nuclear and cytoplasmic β-catenin in atypical cells within PIN lesions, implicating the oncogenic role of stabilized β-catenin in PIN formation (red arrows, Fig. 1f). Positive nuclear AR staining appeared in both atypical epithelial and stromal cells (pink and blue arrows, respectively, Fig. 1g). In contrast, grafts derived from ARKO UGM with Ctnnb1L(ex3)/+:PBCre4 UGE showed mild pathological lesions resembling prostatic hyperplasia and PIN1 (Fig. 1j). Accordingly, very few cells showed positive β-catenin nuclear staining (p < 0.001; red arrow, Fig. 1k; Supplementary Fig. 1a). Positive nuclear AR staining mainly appeared in epithelial cells but was lacking in stromal cells (pink arrows, Fig. 1l). Significantly fewer Ki67+ cells (p = 0.001) appeared in the above grafts than those derived from control UGM (right panel, Fig. 1b). In this study, we also performed tissue recombination assays using AR-deficient and control UGMs as described above with UGEs isolated from another prostate cancer mouse model, PtenL/L:PBCre4 mice17, and observed mild pathologic changes and impaired oncogenic growth in the samples recombined with AR-deficient UGMs (Supplementary Fig. 1b, c). Altogether, these data demonstrate a promotional role of stromal AR signaling in Gli1-expressing cells to support prostate epithelial oncogenesis and tumor initiation.

Fig. 1: Deletion of AR in Gli1-expressing cells in the UGM impairs UGE oncogenic xenograft growth.
figure 1

a Schematic of the experimental timeline for tissue recombination and kidney capsule transplantation. Urogenital sinus epithelium (UGE) from Ctnnb1L(ex3)/+:PBCre4 embryos and urogenital sinus mesenchyme (UGM) from Gli1CreER/+ or ArL/Y:Gli1CreER/+ embryos after activation of Gli1CreER/+ are recombined and grown under the renal capsule of host mice. DOC, day of conception; TM, tamoxifen; KCT, kidney capsule transplantation. b Weights of xenograft recombinants from the indicated UGE and UGM combinations (left panel). Quantification of the percentage of Ki67+ cells per total cells (right panel). Data are represented as mean ± SD of five biological replicates. Two-sided Student’s t test, *p < 0.05, **p < 0.01. Source data and the exact p-values are provided as a Source Data file. cl Representative gross (cd, hi), hematoxylin-eosin (H&E) (e, j), and immunohistochemistry (IHC) for β-catenin (β-cat) and androgen receptor (AR) (fg, kl) images of xenograft tissues from the indicated UGE and UGM combinations. Scale bars, 2 mm (cd, hi), 100 μm (e, j), and 20 μm (fg, kl). The nuclear expression of β-cat in the epithelium is indicated by red arrows. Pink and blue arrows indicate epithelial and stromal AR+ cells, respectively.

Deletion of stromal AR in Gli1-lineage cells impairs prostatic epithelial oncogenesis and tumor development

To evaluate intrinsic stromal AR signaling in prostatic epithelial oncogenesis, we developed Hi-Myc:ArL/Y:Gli1CreER/+ mice, in which the human c-Myc transgene (hMycTg) expression controlled by the modified probasin promoter in prostate epithelium18 co-occurs with stromal Ar deletion regulated by Gli1-driven CreER in prostate stroma14 (Fig. 2a). A significant impairment of mPIN and adenocarcinoma development was observed in Hi-Myc:ArL/Y:Gli1CreER/+ mice in comparison with Hi-Myc:Gli1CreER/+ controls when they both received TM at postnatal day 14, P14 (Fig. 2b). Grossly, prostates of 2- and 6-month-old Hi-Myc:ArL/Y:Gli1CreER/+ mice were significantly smaller and weighed less (p = 3.97 × 10−8 and p = 0.003, respectively; Fig. 2c) than those of age-matched Hi-Myc:Gli1CreER/+ littermates (Figs. 2d and 2e vs 2f and 2g). Pathological analyses showed mPIN3 and 4 as well as prostatic adenocarcinoma lesions in 2- and 6-month-old Hi-Myc:Gli1CreER/+ mice, respectively, similar to the lesions reported in original Hi-Myc mice18 (Fig. 2d, e, h). In contrast, only prostatic hyperplasia and PIN1 or 2 lesions revealed in prostate tissues of age-matched Hi-Myc:ArL/Y:Gli1CreER/+ counterparts (Fig. 2f–h). IHC analyses showed specific nuclear Myc staining in atypical and tumor cells in prostate tissues of both 2- and 6-month-old HiMyc mice (red arrows, Fig. 2d, e), demonstrating Myc-induced oncogenesis. Positive AR staining appeared in both epithelial and stromal cells in the above samples (pink and blue arrows, Fig. 2d, e). In contrast, in samples of age-matched Hi-Myc:ArL/Y:Gli1CreER/+ mice, few atypical cells showed positive Myc, and AR staining appeared in epithelial cells but not in stromal cells (p < 0.01; pink arrows, Fig. 2f, g; Supplementary Fig. 1d). Additionally, less Ki67+ cells appeared in Hi-Myc:ArL/Y:Gli1CreER/+ mice than in Hi-Myc:Gli1CreER/+ counterparts (p < 0.01; Fig. 2d, e vs f, g; Supplementary Fig. 1e). In this study, we also examined stromal AR action in Gli1-lineage cells to promote PIN and tumor progression by injecting TM to 2-month-old Hi-Myc:ArL/Y:Gli1CreER/+ and Hi-Myc:Gli1CreER/+ mice and analyzed them at 6-months of age (Supplementary Fig. 1f). Mild pathological changes reflecting impaired PIN and prostate tumor development revealed in Hi-Myc:ArL/Y:Gli1CreER/+ mice in comparison with the age-matched Hi-Myc:Gli1CreER/+ counterparts (Supplementary Fig. 1g–j). Taken together, these results demonstrate a promotional role of stromal AR signaling in Gli1-lineage cells to support prostatic epithelial oncogenesis and tumor development, which also provides experimental evidence to use these relevant models for in-depth molecular analyses.

Fig. 2: Deletion of AR in Gli1-expressing cells impairs prostate epithelial oncogenic transformation and growth in vivo.
figure 2

a Schematic for generating Hi-Myc:Gli1CreER/+ and Hi-Myc:ArL/Y:Gli1CreER/+ mice. b Schematic of experimental timeline for tamoxifen (TM) injection and analysis. c Ratio of prostate weight (PW) versus whole body weight (BW) as percentages of 2-month-old mice (left panel; n = 5 biological replicates) and 6-month-old mice (right panel; n = 8 biological replicates). Data are represented as mean ± SD of indicated biological replicates. Two-sided Student’s t test, **p < 0.01. Source data and the exact p values are provided as a Source Data file. dg Representative images of gross urogenital tissues, H&E, and IHC staining for Myc, AR, and Ki67 in adjacent prostate tissue sections of Hi-Myc:Gli1CreER/+ and Hi-Myc:ArL/Y:Gli1CreER/+ mice. Red arrows indicate nuclear Myc+ cells. Epithelial and stromal AR+ cells are indicated by pink and blue arrows, respectively. h Table summarizing the pathological abnormalities in the prostates of 2- and 6-month-old Hi-Myc:Gli1CreER/+ and Hi-Myc:ArL/Y:Gli1CreER/+ mice after activation of Gli1CreER/+ at 2 weeks of age. PIN prostatic intraepithelial neoplasia.

Deletion of AR in stromal Gli1-lineage cells impairs prostatic basal epithelial cell-mediated oncogenesis

We developed Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ mice, further referred to as HiMyc and HiMyc-ARKO mice later, respectively (Supplementary Fig. 2a, b), which showed similar pathological lesions as observed in Hi-Myc:ArL/Y:Gli1CreER/+ and Hi-Myc:Gli1CreER/+ counterparts but express CreER activated membrane green fluorescent protein (mGFP) in Gli1-lineage cells19. Given HGPIN lesions and prostatic adenocarcinoma lesions developed in 2- and 6-month-old Hi-Myc:Gli1CreER/+ mice, we performed single-cell RNA sequencing (scRNA-seq) using prostate tissues of 3-month-old HiMyc and HiMyc-ARKO mice to assess the molecular mechanisms for stromal AR signaling in early prostate epithelial tumorigenesis. After the data processing (Supplementary Fig. 2c–j; also see “Methods”), both HiMyc and HiMyc-ARKO samples were visualized individually using the uniform manifold approximation and projection (UMAP) and then merged and clustered for analysis of molecular changes induced by stromal AR deletion Gli1-lineage cells20 (Fig. 3a). A total of 8 cell subsets were identified based on their transcriptomic profiles21,22,23,24 (Fig. 3b; Supplementary Fig. 2k–l) and aligned in separated UMAP plots of HiMyc and HiMyc-ARKO samples (Fig. 3c), demonstrating their comparable cellular properties. UMAP expression plots showed that hMycTg expression was comparable in luminal epithelial cell sets of both samples but reduced in basal epithelial cell sets in ARKO samples in comparison with HiMyc samples (pink and green arrows, Fig. 3c). Gli1-CreER activated mGFP expression was observed in fibroblasts and smooth muscle cell subsets of both samples, further confirming the stromal cell properties of Gli1-lineage cells (blue arrows, Fig. 3c). Reduction of Ar expression appeared in stromal cell subsets of HiMyc-ARKO mice (grey arrows, Fig. 3c). To gain higher resolution of hMycTg expression induced cellular and molecular changes, we separated epithelial cells from other non-epithelial cells, and re-clustered them. Ten epithelial cell clusters were identified, including 3 basal, 6 luminal, and 1 other epithelial cluster, OE (Fig. 3d; Supplementary Fig. 3a, b). The expression of hMycTg appeared comparable in luminal epithelial cells with high Krt8 expression in both HiMyc and HiMyc-ARKO samples, but reduced in Krt14+ basal epithelial cells in HiMyc-ARKO samples (arrows, Fig. 3e). Specifically, the percentage of hMycTg+ basal cells was significantly higher in HiMyc than in HiMyc-ARKO mice (p < 0.01; Fig. 3f). IHC analyses showed Myc+ cells were clearly overlaid with CK8+ cells in PIN lesions on adjacent prostate tissues sections of both HiMyc and HiMyc-ARKO mice (top panel, Fig. 3g). However, noticeable overlays between Myc+ and CK14+ cells only appeared in PIN lesions of HiMyc mice (red arrows, Fig. 3g). In contrast, CK14+ cells were mainly localized in basal cell layers in HiMyc-ARKO samples. Co-immunofluorescent analyses (Co-IF) further identified a significant increase of CK14+ Myc+ atypical cells in PIN lesions of HiMyc samples (p < 0.01; yellow arrows, Fig. 3h; Supplementary Fig. 3c). Myc+ luminal cells overlaying with AR and CK8 staining also appeared in both HiMyc and HiMyc-ARKO samples (Fig. 3h). Using triple-IF approaches, we further identified CK8+ CK14+ and CK8+ CK14+ Myc+ cells in HiMyc samples, indicating the intermediate cell properties of some CK14+ cells in the samples (white arrows, Supplementary Fig. 3d). Prostatic basal epithelial cells are directly adjacent to stromal cells and possess progenitor properties23,25. Specifically, Myc+ basal epithelial cells have been shown to function as prostate tumor-initiating cells in Hi-Myc mice26. Therefore, our data of identifying reduced atypical Myc+ basal cells in PIN tissues of HiMyc-ARKO mice implicate an underlying mechanism by which stromal AR deletion in Gli1-lineage cells impairs prostatic oncogenesis and tumor development.

Fig. 3: Loss of AR in Gli1-lineage stromal cells impacts the basal epithelial compartment at a single cell resolution.
figure 3

a Uniform Manifold Approximation and Projection (UMAP) visualization of 10,810 individual cells from Hi-Myc:R26mTmG/+:Gli1CreER/+ prostates (top left, gray), 6849 individual cells from Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ prostates (top right, dark blue), and integrated cells (bottom). b UMAP representation showing cell-type cluster identities based on gene expression profiles. The dotted line delineates the prostatic epithelial cells. c UMAP plots separated by indicated genotype from b (Far left) and gene expression levels in Hi-Myc:R26mTmG/+:Gli1CreER/+ (top) and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ (bottom). Color intensity indicates the scaled expression level in each cell. Pink and green arrows indicate luminal and basal epithelial cell clusters, respectively, and blue and grey arrows indicate stromal fibroblast and smooth muscle cell clusters, respectively. d UMAP visualization of epithelial cells sub-clustered, re-clustered, and labeled by cell type. BE, basal epithelial cells; LE, luminal epithelial cells; OE, other epithelial cells. e Gene expression levels of markers for human Myc transgene (hMycTg), LE (Krt8), and BE (Krt14). f Graph showing the percentage of hMycTg+ cells per total basal or luminal cells in each genotype. Source data are provided as a Source Data file. g Representative images of H&E and IHC staining for Myc, cytokeratin 8 (CK8), and cytokeratin 14 (CK14) in adjacent prostate tissue sections of 3-month-old Hi-Myc:R26mTmG/+:Gli1CreER/+ (top) and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ (bottom) mice. Scale bars, 20 μm. Red arrows indicate CK14+ cells infiltrating towards the lumen. h Colocalization of Myc with either AR, CK8, or CK14 from indicated prostate tissues. Yellow arrows indicate Myc+CK14+ cells. Scale bars, 20 μm.

Deletion of AR in stromal Gli1-lineage cells impedes IGF1-induced Wnt/β-catenin activation in prostatic basal epithelia

To assess stromal AR-induced molecular changes in prostatic basal epithelium, we performed gene set enrichment analysis (GSEA) using the differentially expressed genes (DEGs) identified in hMycTg+ basal cells between HiMyc and HiMyc-ARKO samples (Supplementary Data 1). Significantly enriched IGF1 receptor (IGF1R), Wnt, and other signaling pathways directly related to prostate tumorigenesis were identified in hMycTg+ basal epithelial cells of HiMyc samples (Fig. 4a; Supplementary Fig. 3e). Emerging evidence has shown a prominent role of IGF1/IGF1R signaling activation in inducing prostate cancer development and progression in both mouse models and clinical studies27. IGF1R activation can induce Wnt/β-catenin signaling to promote prostate tumorigenesis28,29,30. Importantly, IGF1 and Wnt/β-catenin signaling activation directly contributes to androgen-induced prostate oncogenesis and tumor development31. Given these lines of relevant scientific evidence, we examined transcriptomic changes and identified increased expression of Igf1r and its downstream genes, Hras, Irs2, and Akt1 (p < 0.01; top panel, Fig. 4b), as well as Wnt signaling effectors and targets, Ctnnb1, Ccnd1, Cd44, and Tcf7l2 in hMycTg+ basal epithelial cells of HiMyc samples in comparison to HiMyc-ARKO samples (p < 0.01; bottom panel, Fig. 4b; Supplementary Fig. 3f, g). A significant positive correlation between the expression of Igf1r and its downstream target genes, Hras, Irs2, and Akt1 as well as Wnt signaling effectors, Ctnnb1, and targets, Ccnd1, Cd44, and Tcf7l2 (Spearman r = 0.3, 0.2, 0.3, 0.4, 0.2, 0.3, and 0.3, respectively; p = 1.44 × 10−29, 6.27 × 10−18, 7.15 × 10−35, 1.42 × 10−63, 2.93 × 10−5, 3.12 × 10−23, and 5.14 × 10−42, respectively) was identified in hMycTg+ basal epithelial cells of HiMyc mice but not in the counterparts of HiMyc-ARKO mice (Fig. 4c), demonstrating the specific activation of IGF1R and Wnt signaling pathways induced by stromal AR in prostatic basal cells. Real-time quantitative reverse transcription PCR (qRT-PCR) analyses further identified upregulated expression of hMycTg, Igf1r, and its downstream genes, Hras, Grb2, Irs2, and Akt1, as well as Ctnnb1 and its downstream targets, Cd44, Tcf7l2, and Ccnd1 in sorted prostatic basal epithelial cells of HiMyc mice versus those of HiMyc-ARKO mice (p < 0.05; Fig. 4d). It has been shown that aberrant activation of IGF1 signaling can either directly or via activating AKT and GSK3β axes stabilize cellular β-catenin to augment its activity for prostate tumor growth28,29,32. We, therefore, assessed the effect of IGF1 signaling in activating Wnt/β-catenin axes in prostatic atypical basal cells. IHC analyses further detected positive staining for IGF1R, phosphorylated IGF1R (pIGF1R), and CyclinD1, a downstream target of β-catenin, in atypical cells of PIN lesions in HiMyc mouse tissues (Supplementary Fig. 3h1–4 vs 3i1–4), suggesting activated IGF1R and Wnt signaling pathways. Using triple-IF approaches, we further identified co-expression of Myc and CK14, as well as pIGF1R, phosphorylated AKT, phosphorylated GSK3β, cytoplasmic and nuclear β-catenin, CyclinD1, or TCF7L2 in atypical basal cells in PIN lesions of HiMyc mice but not those of HiMyc-ARKO mice (p < 0.01; yellow arrows, top panel, Fig. 4e; Supplementary Fig. 3j). The above regulatory loop was further validated with co-IF analyses in HGPIN tissues of HiMyc mice (Supplementary Fig. 3k). These data demonstrate the impairment of IGF1R and canonical Wnt signaling pathways in prostatic basal cells of HiMyc-ARKO mice, providing mechanistic insight into stromal AR in Gli1-lineage cells to regulate prostatic epithelial tumorigenesis.

Fig. 4: Loss of AR in Gli1-lineage cells reduces IGF and Wnt signaling in basal epithelial cells.
figure 4

a Gene Set Enrichment Analysis (GSEA) enrichment plots depicting insulin-like growth factor 1 receptor (IGF1R) and Wnt signaling pathways significantly up-regulated in hMycTg+ basal epithelial cells in Hi-Myc:R26mTmG/+:Gli1CreER/+ versus Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+. See also “Methods” section. NES, normalized enrichment score; FDR, false discovery rate. b Box plots representing scaled expression data for IGF1R (top) or Wnt (bottom) downstream genes in hMycTg+ basal cells from indicated genotypes (HiMyc, n = 1363; HiMyc-ARKO, n = 36). Pink lines mark the median; top and bottom lines of boxes indicate the boundaries of the first and third quartiles, respectively; the top and bottom whiskers show the maximum and minimum values, respectively, excluding outliers. c Heatmap of pairwise Spearman correlation between the indicated gene expression in hMycTg+ basal cells from indicated genotypes. Colors reflect the level of the correlation coefficient. Numbers show the correlation coefficient. d qPCR analysis of gene expression as indicated in the figure shown as fold change in sort-purified basal cells from indicated genotypes. Data are represented as mean ± SD of three biological replicates. Two-sided Student’s t test, *p < 0.05, **p < 0.01. The exact p values and related source data are provided as a Source Data file. e Representative images of triple-immunofluorescent (IF) staining for indicated antibodies on prostate tissue sections from Hi-Myc:R26mTmG/+:Gli1CreER/+ (top) and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ (bottom). Yellow arrows indicate triple-positive stained cells. Scale bars, 25 μm.

Deletion of AR expression induces robust IGFBP3 expression in stromal Gli1-lineage cells

To gain in-depth insight into prostatic Gli1-lineage cells as stromal niches, we separated stromal cells from other types of cells and re-clustered them into five fibroblast (FB) and smooth muscle cell clusters (Fig. 5a, b; Supplementary Fig. 4a–b). Reduced Ar expression was detected in stromal cell clusters of HiMyc-ARKO mice while comparable Gli1-CreER driven mGFP expression appeared in stromal cells of both samples, demonstrating selective deletion of Ar expression in Gli1-lineage stromal cells (Fig. 5c–d). We then analyzed the DEGs of mGFP+ FB between HiMyc and HiMyc-ARKO samples to explore their role as cancer-associated fibroblasts (CAF)33 (Supplementary Data 2). Intriguingly, we identified the Igfbp3, insulin-like growth factor-binding protein 3, as one of the most highly expressed genes in FB of HiMyc-ARKO (the bottom panel, Fig. 5d, e). IGFBP3 is a key regulator of IGF1 pathways and can bind IGF1/2 to block their access to their receptors for activation34. Increased Igfbp3 expression was further identified in mGFP+ FB of HiMyc-ARKO samples (Fig. 5f). Intriguingly, an inverse correlation between Ar and Igfbp3 expression was identified in mGFP+ FB (middle and bottom panels, Fig. 5d; Supplementary Fig. 4c, d). A significant increase in Igfbp3 alongside reduced Ar expression was further confirmed in sorted mGFP+ cells from HiMyc-ARKO prostate tissues in comparison to those from HiMyc counterparts using qRT-PCR analyses (p = 1.09 × 10−6 and p = 3.36 × 10−6, respectively; Fig. 5g). In alignment with the above results, co-IF analyses showed the significant co-occurrence of reduced AR and increased IGFBP3 expression in mGFP+ and VIM+ fibroblasts of HiMyc-ARKO prostate tissues (p = 3.35 × 10−6 and p = 1.09 × 10−6, respectively; left vs. right panel, Fig. 5h; Supplementary Fig. 4e). These data demonstrate an inverse relationship between AR and IGFBP3 expression in prostatic stromal Gli1-lineage cells, implicating a regulatory mechanism by which stromal AR regulates IGF1 signaling to facilitate prostatic epithelial oncogenesis and tumor development.

Fig. 5: Identifying AR loss-induced molecular changes in Gli1-lineage stromal cells.
figure 5

a–b UMAP plots of stromal fibroblast (FB) and smooth muscle (SM) cell clusters from Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+. c Gene expression levels of mGFP and Ar in UMAP plots separated by indicated genotypes. Color intensity indicates the scaled expression level in each cell. Blue arrow indicates Ar-expressing cells from FB1 in Hi-Myc:R26mTmG/+:Gli1CreER/+. d Violin plots visualizing the expression levels of mGFP, Ar, and insulin-like growth factor binding protein 3 (Igfbp3) in stromal clusters of each genotype. e Gene expression levels of Igfbp3 in UMAP plots of stromal FB and SM cells. Color intensity indicates the scaled expression level in each cell. f Violin plots of mGFP+ FB cells. g qPCR analysis of Ar and Igfbp3 expression shown as fold change in FACs-sorted mGFP+ cells from prostate tissues of indicated genotypes. h Co-localization of IGFBP3 with either AR, mGFP, or Vimentin (Vim) from indicated prostate tissues. Yellow and pink arrows indicate AR+ epithelial and stromal cell, respectively. Blue arrows indicate IGFBP3+ stromal cells co-expressing with either mGFP or Vim. Scale bars, 20 μm. i Sp1 Chromatin immunoprecipitation qPCR analysis of the Igfbp3 promoter regions, and negative control (Untr4) shown as percent input. j qPCR analysis of cancer-associated fibroblast (CAF) markers shown as fold change in sort-purified mGFP+ cells. k Representative images of triple-IF staining for indicated antibodies on prostate tissue sections from indicated genotypes. Blue arrows indicate cells showing double- or triple-positive with the indicated proteins. Scale bars, 20 μm. l Heatmap displaying gene expression of Ar, Vim, and CAF markers across FB clusters of Hi-Myc:R26mTmG/+:Gli1CreER/+. m Heatmap depicting expression of genes associated with indicated pathways. n GSEA showing the enrichment of indicated gene signatures in mGFP+Ar+ FB1 cells from Hi-Myc:R26mTmG/+:Gli1CreER/+ versus mGFP+Ar FB1 cells from Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+. See also “Methods” section. NES, normalized enrichment score; FDR, false discovery rate. In g, i, and j, data are represented as mean ± SD of three biological replicates. Two-sided Student’s t test, **p < 0.01. The exact p values and related source data are provided in the “Source Data file”.

Stromal AR in Gli1-lineage cells represses Sp1-mediated Igfbp3 expression

Regulation of Igfbp3 transcription is mediated mainly through Sp1 transcription factor35. Previous studies have demonstrated that the AR can interfere Sp1 binding to the target promoters and repress Sp1-mediated transcription36,37. To examine the regulatory role of AR on IGFBP3 expression, we performed chromatin immunoprecipitation (ChIP) assays to assess the direct involvement of AR on Sp1-mediated IGFBP3 transcription. We observed increased recruitment of Sp1 within the Igfbp3 promoter regions (p = 0.002) in sorted AR-deficient mGFP+ cells from HiMyc-ARKO mice in comparison to the samples from HiMyc mice, but no difference in the control Untr4 locus38 (Fig. 5i). Accordingly, less AR occupancy on the Igfbp3 promoter regions revealed in samples of HiMyc-ARKO mice compared to those of HiMyc and wild-type control mice (p = 0.011 and p = 0.003, respectively; Supplementary Fig. 4f). These data implicate that AR deletion attenuates the repression of Sp1 recruitment on the Igfbp3 promoter in Gli1-lineage FB, augmenting Igfbp3 transcription. Increased IGFBP3 expression was also identified in prostatic stromal cells of HiMyc-ARKO mice when the Ar was deleted at 2 months of age (Supplementary Fig. 4g1–i3), corresponding to the impaired PIN and prostate tumor development in these mice (Supplementary Fig. 1f–j). Taken together, these data uncover a regulatory mechanism by which AR deletion induces IGFBP3 expression in Gli1-lineage FB to block IGF1-signaling activation, further hindering prostatic basal epithelial cell-mediated oncogenesis and tumor development.

Stromal AR in Gli1-lineage FB converts the cellular properties of CAF

To assess the role of stromal AR in Gli1-lineage FBs, we analyzed DEGs between FBs of HiMyc and HiMyc-ARKO samples and identified expression of CAF cellular markers33, including Pdgfrβ, Twist1, Sox9, Foxf1, Il11, and Cxcl10 (Supplementary Fig. 4j). Using qRT-PCR analyses, we further demonstrated reduced expression of those CAF markers in sorted mGFP+ cells from HiMyc-ARKO samples in comparison to those of HiMyc counterparts (p < 0.001; Fig. 5j). Triple-IF analyses also showed the specific co-localization between PDGFRβ or Twist1 with AR and Gli1CreER activated mGFP expression in PIN tissues of HiMyc mice (blue arrows in the top panel, Fig. 5k; Supplementary Fig. 4k1–4), but not in ones of HiMyc-ARKO mice (p < 0.01; bottom panel, Fig. 5k; Supplementary Fig. 4l1–4m). Significantly up-regulated expression of CAF cellular markers, including Pdgfrβ (p = 9.13 × 10−7), Twist1 (p = 2.05 × 10−32), Sox9 (p = 2.05 × 10−32), Foxf1 (p = 8.40 × 10−37), Il11 (p = 3.24 × 10−36), and Cxcl10 (p = 3.38 × 10−50) was further identified in the FB1 cell cluster of HiMyc samples (Fig. 5l; Supplementary Fig. 4n, o). Analyses of transcriptomic changes in FB1 cluster between mGFP+ Ar+ FB of HiMyc and mGFP+ Ar– FB of HiMyc-ARKO showed up-regulation of Ar, AR downstream target genes, and Wnt ligands, and down-regulation of P53 and Sp1 downstream targets as well as IGFBP3 and 7 (Fig. 5m; Supplementary Data 3). GSEA based on the above DEGs identified the enrichment in upregulated pathways that promote prostate cancer (p = 0.006) and androgen signaling activation (p = 0.016), and downregulated pathways related to cell hypoxia (p = 0.019), negative responses to insulin and IGF1 (p = 0.015), and P53-mediated signaling (p = 0.025) in the mGFP+ Ar+ FB1 of HiMyc samples (Fig. 5n). These data provide in-depth mechanistic insight into the regulatory role of stromal AR in Gli1-lineage FB acting as CAF to support prostate epithelial tumor cell growth and progression.

IGFBP3 represses IGF1/IGF1R signaling-induced prostate epithelial cell growth

Identifying the regulation of stromal AR in IGFBP3 expression in Gli1-lineage FB is intriguing, suggesting an underlying mechanism for androgen-mediated IGF1 signaling activation in prostate oncogenesis. To directly assess the effect of IGF1 on prostatic basal epithelium, we developed organoid cultures using enriched basal epithelial cells through cell sorting from microscopically confirmed PIN tissues of HiMyc and HiMyc-ARKO mice (see the “Methods”39,40; Supplementary Fig. 5a–c). Organoids derived from HiMyc-ARKO mice showed retarded growth in comparison with those from HiMyc mice at earlier culture time points (p = 7.36 × 10−5, p = 0.004, and p = 0.25, respectively; Supplementary Fig. 5d). However, the growth difference gradually diminished at day 6, implicating the effects of microenvironment on their growth. To mimic the niche effects as observed in our in vivo models, we treated epithelial organoids with IGF1, IGF1 + IGFBP3, or vehicle. Measuring organoid forming efficiency and average sizes of individual organoids showed significant increases in IGF1-treated samples compared to IGF1 + IGFBP3-treated counterparts in both HiMyc and HiMyc-ARKO mice (p < 0.001), demonstrating the critical role of IGFBP3 in neutralizing IGF1 activity (Fig. 6a–c; Supplementary Fig. 5e1, 5f1). There is no significant difference between IGF1-treated epithelial organoids derived from HiMyc and HiMyc-ARKO samples (forming efficiency, p = 0.590 and organoid size, p = 0.658). Histologically, IGF1-treated groups in both genotypes showed severe pathological lesions featuring multilayer atypical cells growing into the lumen and as cribriform structures, whereas IGF1 + IGFBP3- or vehicle-treated groups displayed mild pathologic changes (bottom panel, Fig. 6a; Supplementary Fig. 5e2–3, 5f2–3). These results suggest epithelial organoids derived from atypical basal cells of HiMyc and HiMyc-ARKO samples, after growing in a new microenvironment, possessing comparable abilities in response to IGF1-induced cell growth. IHC analyses showed more Myc+ atypical cells in IGF1-treated organoids than in IGF1 + IGFBP3-treated or vehicle-treated counterparts in both genotype samples (p < 0.001; top panel, Fig. 6d; Supplementary Fig. 5g1, h1, i). In the IGF1-treated organoids, CK14+ cells appear to overlay with Myc+ cells and infiltrate in PIN lesion, in contrast to CK14+ cells only localized on the basal layers in the IGF1 + IGFBP3 treated organoids (middle panel, Fig. 6d). Both AR and CK8 positive cells were also observed in the above basal cell derived organoids (Fig. 6d), demonstrating the ability of CK14+ cells to differentiate into luminal cells. Positive staining for pIGF1R and cytoplasmic and nuclear staining of β-catenin revealed in atypical cells in the above IGF1-treated organoids from HiMyc and HiMyc-ARKO epithelia, but negative or reduced staining of those proteins appeared in IGF1 + IGFBP3- or vehicle-treated counterparts (bottom panels, Fig. 6d; Supplementary Fig. 5g3–4, 5h3–4). The above data of identifying increased cytoplasmic and nuclear β-catenin expression in IGF1-treated organoid samples further demonstrates the promotional role of Wnt/β-catenin activation in prostatic epithelial growth.

Fig. 6: Identifying the reciprocal effect of stromal AR action on prostatic epithelial organoid oncogenic transformation and growth.
figure 6

a Organoid culture of prostatic epithelial cells from 12-week-old Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ mice, following treatment with vehicle, IGF1 (100 ng/ml), or IGF1 (100 ng/ml) + IGFBP3 (1 μg/ml). Representative of brightfield images and HE staining showing organoid morphology and structure observed in indicated groups. Scale bars, 400 μm and 25 μm. b Quantification of individual organoid size. Organoids (n = 100) per group examined over three independent experiments. The center line represents the median value in each group. c Quantification of organoid forming efficiency showing the percentage of organoids above 50 μm diameter per total cells seeded at day 0 in a well. The center line represents the median value, the box borders represent the lower and upper quartiles (25% and 75% percentiles, respectively) and the ends of the bottom and top whiskers represent the minimum and maximum values, respectively, for eight independent samples over three biological replicates. d Representative images of IHC staining with the indicated antibodies. Scale bars, 25 μm. e–f Organoid culture of prostatic epithelial cells from 12-week-old Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ mice, following treatment with IGF1 (100 ng/ml), IGF1 (100 ng/ml) + ICG-001 (5 μM), or IGF1 (100 ng/ml) + iCRT3 (5 μM). Representative brightfield images and HE staining showing organoid morphology and structure observed in indicated groups. Scale bars, 400 μm and 25 μm. b, c Two-sided Student’s t-test for IGF1-treated groups versus IGF1 + IGFBP3-treated groups in HiMYC or HiMYC-ARKO, **p < 0.01. Source data and the exact p-values are provided in the “Source Data file”.

To directly assess the effect of Wnt/β-catenin in IGF1-induced prostatic basal epithelial growth, we treated organoids with Wnt inhibitors, ICG-00141 and iCRT342 in combination with IGF1 (Fig. 6e, f; Supplementary Fig. 6a–c). It has been shown that iCRT3 can disrupt the interaction between AR and β-catenin and inhibit AR-mediated transcription and cell growth in prostate cancer cells42. We observed more and larger organoids in IGF1-treated samples than those treated with IGF1 and Wnt inhibitors (Top panel, Fig. 6e, f). Measuring organoid forming efficiency and average size of individual organoids showed a significant decrease in samples treated with IGF1 and Wnt inhibitors, ICG-001 or iCRT3, in comparison to those treated with IGF1 only (p < 0.001; Supplementary Fig. 6a, b). Histological analyses also showed minor pathological changes in the former but typical HGPIN lesions in the latter (bottom panels, Fig. 6e, f). IHC further showed decreased expression of cytoplasmic and nuclear β-catenin, β-catenin downstream targets, Myc, Cyclin D1, and Ki67 in samples treated with Wnt inhibitors and IGF1 compared to those treated with IGF1 only (Supplementary Fig. 6c). Taken together, these data demonstrate the promotional role of Wnt/β-catenin activation in IGF1-induced prostate tumor cell growth.

Stromal AR promotes prostatic basal epithelial oncogenesis through the IGFBP3-IGF1/IGF1R and Wnt/β-catenin regulatory loop

To directly examine the regulatory role of increased IGFBP3 on IGF1 in prostate stromal cells, we isolated mGFP+ stromal cells from both HiMyc and HiMyc-ARKO mice through cell sorting and measured the level of IGF1 in the stromal cell-conditioned medium. Higher levels of secreted IGF1 were detected in the stromal cell-conditioned medium prepared from HiMyc mice (p = 1.79 × 10−6; Fig. 7a). Prostate epithelial organoids treated with the stromal cell-conditioned medium prepared from HiMyc samples showed a significant increase in cell growth in comparison with those treated with the conditioned medium of HiMyc-ARKO stromal cells (Fig. 7b). Pre-incubation of IGFBP3 with the conditioned medium reduced its effects in inducing organoid cell growth when both organoid forming efficiency and average sizes of individual organoids were measured (p < 0.001; Supplementary Fig. 6d, e). IHC analyses showed positive staining for phosphorylated IGF1R, cytoplasmic and nuclear β-catenin, Myc, and Cyclin D1 in samples treated with the conditioned medium from HiMyc stromal cells but not in other samples (Supplementary Fig. 6f), directly demonstrating IGF1 activity in the above-conditioned medium. Specifically, positive staining for cytoplasmic and nuclear β-catenin and its downstream target, Cyclin D1 in the above samples further demonstrate the regulatory mechanisms for IGF1-induced Wnt/β-catenin activation in prostate epithelial cells. Altogether, these data provide additional evidence to demonstrate the role of stromal AR in Gli1-lineage cells, acting as a niche to support prostate epithelial cell growth through IGF1-induced Wnt signaling activation.

Fig. 7: Stromal AR in Gli1-lineage cells induces basal epithelial IGF and Wnt pathways to promote oncogenic growth through paracrine action.
figure 7

a IGF1 concentration in conditioned media (CM) from 12-week-old Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ GFP+ cells. Data are represented as mean ± SD of four independent samples over two biological replicates. Two-sided t test, **p < 0.01. Source data and the exact p values are provided as a Source Data file. b Organoid culture of prostatic epithelial cells from 12-week-old Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ mice, following treatment with Hi-Myc:R26mTmG/+:Gli1CreER/+ CM, Hi-Myc:R26mTmG/+:Gli1CreER/+ CM + IGFBP3 (1 μg/ml), or Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ CM. Representative brightfield images and HE staining showing organoid morphology and structure observed in indicated groups. Scale bars, 400 μm and 25 μm. c Representative images of triple-IF staining for indicated antibodies on prostate tissues from Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ mice. Yellow or white arrows indicate CK14+Myc+ or CK14+pIGF1R+ basal epithelial cells respectively. Blue arrows indicate IGFBP3+ mGFP+ stromal cells. Scale bars, 20 μm. d Schematic of hypothetic models by which stromal AR in Gli1-lineage cancer-associated fibroblast (CAF) cells regulate prostate tumorigenesis via IGF, Wnt, and other signaling pathways in prostatic basal epithelial cells. e GSEA results from pre-ranked differentially expressed genes list comparing basal cells versus luminal cells in human prostate cancer samples, Hi-Myc:R26mTmG/+:Gli1CreER/+ and Hi-Myc:R26mTmG/+:ArL/Y:Gli1CreER/+ mice. See also “Methods” section.

Using triple-IF analyses, we further assessed the reciprocal regulation between prostatic stromal IGFBP3 expression and epithelial IGF1R activation. Multilayered atypical Myc+CK14+ basal epithelial cells were observed adjacent to mGFP+ Gli1-lineage cells within PIN lesions of HiMyc samples in contrast to the single layer of Myc+CK14+ basal epithelial cells in HiMyc-ARKO samples (yellow arrows, Fig. 7c), in which IGFBP3+ mGFP+ Gli1-lineage cells were mainly observed (blue arrows, Fig. 7c). Accordingly, positive staining of pIGF1R only appeared in HiMyc but not in HiMyc-ARKO samples (white arrows, Fig. 7c). Increased Ki67+ staining also appeared in prostatic basal epithelium in HiMyc versus HiMyc-ARKO samples, indicating the promotional role of IGF1R activation in prostatic tumor epithelia (p = 5.66 × 10−5; Fig. 7c; Supplementary Fig. 6g). These similar results were also observed in prostate tissues of 6-month-old HiMyc and HiMyc-ARKO mice when they received TM at 2 months of age (Supplementary Fig. 6h), validating the regulatory role of IGFBP3 on IGF1R1 signaling. Taken together, these data in combination with other results elucidate that AR loss attenuates the repression on Sp1-mediated IGFBP3 transcription resulting in elevated IGFBP3 in Gli1-lineage FB, which reduces IGF1 binding to IGF1R on adjacent prostatic basal epithelial cells and diminishes IGF1R-mediated Wnt/β-catenin activation (Fig. 7d).

Emerging evidence has shown the prostatic stem/progenitor cell properties of human prostatic basal epithelial cells1. Human prostatic cancer cells with basal cell gene signatures are also linked to advanced, metastatic, and castration-resistance tumor phenotypes43. To explore the clinical relevance of stromal AR action as a tumor niche, we compared GSEA based on DEGs of basal versus luminal cell in human prostate tumors with those in our mouse models, and identified similar enriched signaling pathways between human samples and HiMyc mice. Specifically, both IGF1 and canonical Wnt signaling pathways were significantly enriched (Fig. 7e and Supplementary Data 4). These data provide additional and clinically relevant evidence demonstrating the critical role of reciprocal interactions between stromal AR signaling in Gli1-lineage cells and prostatic basal epithelial tumorigenesis.