PTEN Loss Promotes Intratumoral Androgen Synthesis and Tumor Microenvironment Remodeling via Aberrant Activation of RUNX2 in Castration-Resistant Prostate Cancer

Most cases of advanced prostate cancer are initially responsive to androgen deprivation therapy (ADT); however, the majority relapses and eventually evolves into castration-resistant prostate cancer (CRPC). Accumulating evidence suggests that IAS plays a pivotal role in CRPC development and progression (1, 2). These findings have led to the development of second-generation anti-AR therapeutics, including the cytochrome P450 17A1 (CYP17A1) inhibitor abiraterone that inhibits intratumoral androgen synthesis (IAS) in prostate tumors (3). Despite the increased benefit of abiraterone in overall survival in clinic, most patients develop resistance after initial responses via unclear mechanisms (4).

Up to 70% of patients with advanced prostate cancer exhibit loss of PTEN function or activation of the PI3K/AKT pathway (5). Aberrant activation of AKT frequently associates with prostate cancer progression, poor prognosis, and high risk of recurrence (6). PTEN loss also associates with worse recurrence-free survival in patients treated with radical prostatectomy (7) or abiraterone acetate (ABA; ref. 8). These findings imply that PTEN loss may link to IAS. However, the molecular mechanisms by which loss of PTEN promotes IAS remain unexplored.

Invasion into the surrounding stromal niche is a prerequisite for distant metastasis of cancerous cells. The process requires the detachment of cancer cells from the primary tumor and the recapitulation of TME. The stroma-epithelium crosstalk plays a critical role in creating a vicious cycle of positive feedback, thereby facilitating aggressive local cancer growth and metastatic spread to distant organs.

Runt-related transcription factor 2 (RUNX2) is implicated in human prostate cancer. It is frequently overexpressed in cancers that metastasize predominantly to bone (9, 10). This transcription factor has been shown to promote cancer cell metastasis by increasing expression of various types of matrix metalloproteinases (9–11), thereby causing collagen degradation. It has been shown previously that FOXO1 is a negative regulator of RUNX2 (12, 13). Loss of PTEN or growth factor stimulation results in FOXO1 inactivation, which in turn induces aberrant activation of RUNX2 and promiscuous expression of its target genes [e.g., OCN, osteopontin (OPN), VEGF, and interleukin 8 (IL8)] involved in cell migration and invasion (12). To date, however, the exact mechanisms underlying RUNX2-induced prostate cancer cell aggressiveness remain to be elucidated. In this study, we demonstrated that abnormal activation of RUNX2 caused by PTEN inactivation or AKT activation induces IAS and TME remodeling, thereby promoting prostate cancer tumorigenesis and progression.

PC-3 and LNCaP cell lines were purchased from the ATCC, and C4-2 cell line was purchased from UroCorporation. All cell lines were cultured in RPMI-1640 medium supplemented with 10% FBS (Thermo Fisher Scientific) or otherwise specified plus 100 μg/mL penicillin–streptomycin–glutamine (Thermo Fisher Scientific) at 37°C with 5% CO₂. All cell lines were tested and authenticated (karyotyping, RUNX2 expression and PTEN loss status) within 6 months of the submission of the manuscript. Mycoplasma contamination was regularly examined using Lookout Mycoplasma PCR Detection Kit (Sigma-Aldrich). Plasmocin (Thermo Fisher Scientific) was routinely added to the cell culture medium to prevent or eliminate mycoplasma contamination. Cell transfection was performed by electroporation using an Electro Square Porator ECM 830 (BTX) as described previously or lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer's instructions. Transfection efficiencies of approximately 80% to 90% were routinely achieved.

Runx2 heterozygous (Runx2^+/−) mice were reported previously (14, 15). Pten conditional knockout (Pten^loxp/loxp or Pten^L/L) mice were generated originally in the laboratory of Dr. Hong Wu (University of California Los Angeles School of Medicine, Los Angeles, CA) and purchased from The Jackson Laboratory (Bar Harbor, Maine). PB-Cre transgenic mice were generated originally in the laboratory of Dr. Pradip Roy-Burman (Keck School of Medicine, University of Southern California, Los Angeles, CA) and acquired from the National Cancer Institute (NCI) Mouse Repository. Cohorts of Runx2^+/−;Pten^pc-/−, Runx2^+/− and Pten^pc-/−mice were generated from PB-Cre⁺;Runx2^+/−;Pten^L/L males and Pten^L/L females, which were obtained by cross breeding PB-Cre males with Runx2^+/− and Pten^L/L females. All procedures were approved by the Mayo Clinic Institutional Animal Care and Use Committee (IACUC) and conform to the legal mandates and federal guidelines for the care and maintenance of laboratory animals in accordance with recommendations as described (16, 17).

Genomic DNA from mouse tails was purified by phenol/chloroform extraction and used as templates for genotyping PCR reactions. Genotyping of wild-type and knockout allele of Runx2, conditional alleles of Pten, and Cre transgene was performed according to previously described PCR protocols (16, 17). Information for primers used is provided in Supplementary Table S1. Total RNA was isolated from cells or tissues with TRizol (Thermo Fisher Scientific) and cDNA was synthesized using the Super-Script kit from Thermo Fisher Scientific. Two-step real-time PCR was performed using the SYBR Green Mix (Bio-Rad) and an iCycler iQTM detection system (Bio-Rad) according to the manufacturers' instructions. Both forward and reverse primers were used at a final concentration of 200 nmol/L. Expression of Gapdh gene in each sample was used as an internal control.

Four-month-old Pten^pc-/− male mice were castrated for one month. Mice were randomly assigned to two groups (n = 5/group) and treated with intraperitoneal injections of vehicle (5% benzyl alcohol and 95% safflower oil) or ABA (0.5 mmol/kg/d in vehicle) daily for 21 days (18).

Western blotting was performed as described previously (16, 17). Antibodies used for Western blotting were diluted at 1:1,000 to 1:2,000.

Experiments were carried out with three or more replicates. Statistical analyses were performed using the Student t test unless otherwise stated. Values with P < 0.05 are considered statistically significant.

Addition methods are provided in Supplementary Information.

CYP11A1 and CYP17A1 are members of the cytochrome P450 enzyme family that are critical for steroid biosynthesis. Importantly, CYP17A1 is the target of the next-generation endocrine therapeutic agent ABA. Previous studies in mice showed that Ocn acts as a bone hormone that binds to G-protein coupled receptor family C group 6 member A (Gprc6a) in Leydig cells to stimulate testosterone biosynthesis in the testis by inducing expression of Cyp11a1 and Cyp17a1 (19). OCN is a well-established target gene transcriptionally regulated by RUNX2 (12, 20, 21). We and others previously showed that FOXO1 binds to and inhibits the transcriptional activities of RUNX2 in prostate cancer cells and osteoblasts, and this effect is abolished by activated AKT due to PTEN loss or insulin-like growth factor-1 (IGF-1)/insulin stimulation (12, 13). Osteomimictric properties, such as elevated RUNX2 expression and OCN secretion, have been well documented in PTEN-null prostate cancer cells (22). Furthermore, elevated expression of the GPRC6A gene is common in human prostate cancer cell lines (23) and strongly associated with prostate cancer risk (24). We therefore sought to determine whether expression of CYP11A1 and CYP17A1 genes is regulated by the AKT–RUNX2–OCN–GPRC6A signaling axis in PTEN-deficient prostate cancer cells.

C4-2B is an AR-positive PTEN-null CRPC cell line that expresses high levels of RUNX2 (25, 26). We treated C4-2B cells with the PI3K-mTOR dual inhibitor NVP-BEZ235. As expected, NVP-BEZ235 treatment inhibited AKT phosphorylation and importantly, it also decreased CYP11A1 and CYP17A1 mRNA expression (Fig. 1A). Similar results were obtained by treating C4-2B cells with the AKT inhibitor MK2206 (Fig. 1B). Next, we examined whether knockdown of RUNX2 decreases expression of these genes. As shown in Fig. 1C, endogenous RUNX2 was effectively knocked down by a pool of two independent shRNAs in C4-2B cells as demonstrated by Western blot analysis. Depletion of RUNX2 not only largely diminished mRNA expression of CYP11A1 and CYP17A1 genes, but also blocked NVP-BEZ235-mediated inhibition of these two genes (Fig. 1C; Supplementary Fig. S1A). In agreement with the finding that FOXO1 acts as a negative upstream regulator of RUNX2 in prostate cancer cells (12), expression of a constitutively active form of FOXO1 (FOXO1-A3), in which three AKT phosphorylation sites are mutated to three alanine residues, significantly decreased CYP11A1 and CYP17A1 mRNA expression in C4-2B cells, but had no overt effect on expression of RUNX2 mRNA and protein (Fig. 1D; Supplementary Fig. S1B).

The level of testosterone or the ratio of testosterone to estradiol is a crucial factor influencing prostate cancer development during aging of men (27). The finding that mice lacking Ocn and Gprc6a show a significant change in the testosterone/estradiol ratio supports the notion that GPRC6A and OCN are two prostate cancer relevant genes (19, 28). OCN is readily expressed in C4-2B cells (29). We therefore examined the effect of OCN knockdown on expression of CYP11A1 and CYP17A1 mRNA in this cell line. Endogenous OCN was effectively knocked down by a pool of two independent shRNAs in C4-2B cells and this effect was correlated with decreased expression of both CYP11A1 and CYP17A1 genes (Fig. 1E). In contrast, overexpression of OCN prevented RUNX2 knockdown-induced downregulation of these two genes (Supplementary Fig. S1C and S1D).

Previous studies show that GPRC6A is highly expressed in various prostate cancer cell lines (23). While the endogenous GPRC6A protein was only partially (approximately 50%) depleted by infection of C4-2B cells with a pool of two independent shRNA, GPRC6A knockdown substantially decreased CYP11A1 and CYP17A1 mRNA expression (Fig. 1F). Finally, it has been shown that stimulation of mouse Leydig cells with Ocn results in activation of the transcriptional factor Creb and knockout of Creb in Leydig cells decreases expression of testosterone biosynthesis genes (19). We found that ectopic expression of CREB increased expression of CYP11A1 and CYP17A1 in C4-2B cells (Fig. 1G). Chromatin immunoprecipitation coupled quantitative polymerase chain reaction (ChIP-qPCR) assays revealed the binding of CREB protein in the promoter of CYP11A1 and CYP17A1 genes in C4-2B cells (Fig. 1H). Importantly, CREB occupancy at both promoters was largely diminished by RUNX2 knockdown (Fig. 1H), although there is no Runx binding site detected in this region. LNCaP (AR-positive) and PC-3 (AR-negative) are other two PTEN-negative prostate cancer cell lines which also express high level of RUNX2, OCN and GPRC6A (23, 25, 30). Similar results were obtained in LNCaP and PC-3 cells (Supplementary Figs. S2 and S3). Together, these data indicate that activation of the AKT–RUNX2–OCN–GPRC6A–CREB signaling pathway induces expression of steroidogenesis genes CYP11A1 and CYP17A1 in human PTEN-null prostate cancer cell lines.

An ELISA-based assay (Cayman Chemical) is commonly used for testosterone measurement in Leydig cells, serum and recurrent prostate cancer tissues (1, 19, 31–33). We used this well-established approach to determine whether the PTEN–AKT pathway regulates testosterone level in prostate cancer cells. We cultured PC-3 cells in charcoal-stripped (androgen-depleted) medium to minimize the interference of exogenous testosterone from supplemented FBS. We found that treatment of PC-3 with the PI3K–mTOR dual inhibitor NVP-BEZ235 significantly decreased testosterone level (Fig. 2A and B). Similar results were obtained when PC-3 cells were treated with the AKT inhibitor MK2206 (Fig. 2A and C). Consistent with a previous report (2), a very low level of testosterone was detected in prostate tissues in castrated wildtype mice (Fig. 2D). However, the testosterone level was significantly elevated in Pten-knockout prostate tumors in castrated mice compared with that in normal prostate tissues from castrated wildtype littermates (Fig. 2D; Supplementary Fig. S4A and S4B). In agreement with this observation, Pten deletion increased expressions of Cyp11a1 and Cyp17a1 mRNA in prostate tumors in castrated mice (Fig. 2E). We further showed that the testosterone concentration in serum was comparable between castrated Pten knockout mice and castrated wildtype littermates, but was much lower compared with that in serum of uncastrated wildtype mice (Fig. 2F). These data suggest that the source of the increased testosterone levels observed within Pten-null tumors was most likely originated from the tumor itself. Together, our data demonstrate that testosterone is synthesized in PTEN-deficient prostate cancer cells in culture and in mice and this event is regulated by the PTEN–AKT pathway.

Given that the RUNX2–OCN–GPRC6A axis regulates expression of steroidogenesis genes CYP11A1 and CYP17A1 in RUNX2-expressing cell lines (Fig. 1; Supplementary Figs. S2 and S3), we sought to employ a loss-of-function approach to determine whether this pathway affects testosterone levels in PTEN-deficient prostate cancer cells. We found that knockdown of endogenous RUNX2 decreased testosterone levels in PC-3 cells (Fig. 3A). Similar results were obtained in OCN- or GPRC6A-knockdown PC-3 cells (Fig. 3B and C). Importantly, in agreement with the regulation of CYP11A1 and CYP17A1 gene expression by the RUNX2–OCN pathway, OCN overexpression prevented RUNX2 knockdown-induced decrease in testosterone levels in C4-2B cells (Supplementary Fig. S1C–S1E). Next, we examined the effect of Runx2 knockout on testosterone levels in Pten-knockout tumors in castrated mice. Given that Runx2 homozygous deletion mice die right after birth (14), we generated Runx2 heterozygous mice in the background of prostate-specific Pten homozygous deletion. Although Runx2 was only heterozygously deleted, surprisingly, Runx2 mRNA and protein were largely depleted in the prostates (Supplementary Fig. S4A, S4C, and S4D), and this result is consistent with the previous finding that Runx2 is a autoregulatory gene (34). In agreement with the findings in cultured human prostate cancer cells, Runx2 heterozygous deletion substantially decreased testosterone levels in Pten-knockout prostate tumors in castrated mice (Fig. 3D). Accordingly, deletion of Runx2 decreased expression of Cyp11a1 and Cyp17a1 genes in prostate tumors in castrated Pten-knockout mice (Fig. 3E). It is worth noting that the expression level of Cyp11a1 was much higher than that of Cyp17a1 in Pten-deficient prostate tumors in castrated mice (Figs. 2E and 3E), but the exact underlying mechanism warrants further investigation. In agreement with these observations, heterozygous deletion of Runx2 also decreased expression of Ocn, a upstream regulator of these two genes in Pten-null prostate tumors in mice (Fig. 3F and G). While Runx2 heterozygous deletion had no overt effect on expression of Creb, it largely decreased the level of phosphorylated (active) Creb in Pten-null prostate tumors (Fig. 3H–K). Together, these data suggest that expression of RUNX2 is required for IAS in both human and mouse PTEN-null prostate cancer cells.

We next examined the causal role of RUNX2 in PTEN loss-induced prostate cancer growth using Pten knockout mice as a model. Consistent with the inhibitory effect on IAS in castrated mice (Fig. 3D), Runx2 heterozygous deletion largely decreased the incidence of high-grade prostatic intraepithelial neoplasia (HGPIN) and cancerous lesions under the castration condition (Fig. 4A–C). We further demonstrated that deletion of Runx2 significantly decreased Ki-67 staining in Pten-deficient prostate tumors in castrated mice (Fig. 4D and E). Runx2 heterozygous deletion in Pten-null prostate tumors from castrated mice also markedly increased cell apoptosis as demonstrated in IHC staining of cleaved caspase 3 (Fig. 4F and G). Thus, Runx2 reduction inhibited tumorigenesis in the Pten-knockout prostate in castrated mice by decreasing IAS and cell proliferation and increasing apoptotic cell death.

It was shown previously that Pten deletion results in downregulation of Ar protein in the mouse prostate and AR target gene expression (5, 17, 35). Although its expression level is downregulated, Ar protein remains functional and is required for growth of Pten-deficient prostate cancer cells under androgen-depleted conditions (36). In agreement with these observations, we demonstrated that expression of AR target genes such as Probasin, Nkx3.1, and Tmprss2 was downregulated in Pten-knockout prostate tissues compared to wildtype counterparts (Fig. 4H). Importantly, expression of these genes was further reduced by codeletion of Runx2 (Fig. 4H). It has been well established that androgen depletion decreases AR protein level (37). Consistent with these findings, Runx2 heterozygous deletion decreased testosterone synthesis gene expression and intratumoral testosterone level in Pten-deficient prostate tumors in castrated mice (Fig. 3D and E) and also decreased Ar protein expression (Fig. 4I). These data demonstrate that Runx2 heterozygous deletion decreases Ar protein level and inhibits its transcriptional activity in Pten-deficient prostate tumors in castrated mice.

ABA impairs AR signaling by inhibiting adrenal and intratumoral androgen biosynthesis in vivo (38). Given that deletion of Runx2 decreased expression of Cyp17a1 mRNA and testosterone synthesis in Pten-knockout prostate cancer in castrated mice, we examined the effect of ABA-mediated inhibition of Cyp17a1 on testosterone synthesis in the same mouse model. As expected, treatment of mice with ABA significantly decreased testosterone levels in prostate cancer tissues in castrated Pten-knockout mice (Fig. 5A and B). Similar to the effect of Runx2 heterozygous deletion, ABA treatment also decreased tumor burden by substantially reducing the incidence of HGPIN and cancerous lesions in castrated Pten-knockout mice (Fig. 5C and D). ABA treatment also largely decreased Ki-67 staining and elevated cleaved caspase 3 expression (Fig. 5E–H). These data suggest that ABA treatment decreases prostate tumor burden in the Pten-knockout model by inhibiting cell proliferation and increasing cell death. In accordance with the results that ABA treatment reduced intratumoral testosterone level (Fig. 5B), ABA treatment also largely decreased expression of Ar protein and its target genes examined in Pten-deficient prostate tumors in castrated mice (Fig. 5I and J), indicating that AR activity is inhibited by ABA treatment in PTEN-null tumors in mice.

Remodeling and stiffening of the stromal microenvironment offer an appropriate niche for cancer cell invasion and progression. As a transcription factor involved in osteoblast differentiation (22), RUNX2 has been studied intensively for its role in tumorigenesis in various types of cancer (9–11). In vitro studies suggest that RUNX2 promotes the aggressiveness of CRPC by enhancing cell invasive capability toward collagen (39). However, the role of RUNX2 in regulating stromal microenvironment of prostate cancer in vivo is poorly undetermined.

A previous report (40) shows that the periacinar fibroblast layers appear to be wider in PIN lesions as an early sign of reactive stroma, implying a local activation of periacinar fibroblasts in response to PIN development. We examined the impact of the PTEN–RUNX2 pathway on stromal microenvironment using castrated Pten/Runx2 double knockout mice as a working system. We first performed IHC for smooth muscle actin (SMA). In the prostate of wildtype mice (Fig. 6A1), we detected a highly condensed layer of SMA-positive stroma, which is regarded as a natural barrier against the invasion of tumor cells into the adjacent stromal tissue. In contrast, we observed much loosened layers of SMA-positive stroma surrounding malignant epithelial cells in the prostate of Pten-knockout mice (Fig. 6A2). Indeed, invasive adenocarcinoma often induces a desmoplastic response, in which collagenous material is diffusely distributed in the surrounding stromal tissue. We performed Masson Trichrome stain that marks keratin and muscle fibers red and collagen blue. We detected a relatively condensed layer of collagen surrounding the stroma in the prostate of wildtype mice (Fig. 6B1). On the contrary, we found the existence of multiple layers of collagen (blue) mingled with layers of small muscle fibers (pink) in the prostate of Pten-knockout mice (Fig. 6B2). Thus, loss of PTEN promotes the recruitment of connective tissues to the diseased areas, which may likely lead to an invasive phenotype of tumor by commutating with the adjacent acini via collagen. SMA IHC showed that Runx2 heterozygous deletion alone had no overt effect on composition of SMA-positive stromal tissues in the prostate of mice without Pten loss (Fig. 6A3) and this result is not surprising because Runx2 expression level is very low in the prostate of wildtype mice (26). Notably, Runx2 depletion almost completely reversed Pten knockout-induced formation of multiple layers of SMA-positive stroma surrounding malignant epithelial cells (Fig. 6A4). Similar results were obtained in Trichrome staining (Fig. 6B4). These findings in mice reveal an important role of RUNX2 in modulating PTEN loss-induced remodeling of prostate cancer stromal environment under castration conditions. Similar to the effect of Runx2 heterozygous deletion, ABA treatment also largely inhibited PTEN loss-induced acquisition of loosened layers of SMA-positive stroma surrounding malignant epithelial cells (Fig. 6C). We further showed that ABA administration blocked deposition of collagen in the stromal environment (Fig. 6D). These findings suggest that RUNX2 depletion or ABA treatment prohibits the local invasive potential of prostate cancer in the context of PTEN deletion by reducing extracellular matrix remodeling in the stromal compartment.

The intimate interactions among each cell type in the neoplastic niche, such as epithelium, macrophage, fibroblast, and other stromal cells, contribute significantly to tumorigenesis and cancer progression (41). Cross-talk between cytokine and chemokine pathways in the stroma can elicit tumorigenic responses. To investigate the molecular mechanisms by which ABA treatment induces TME remodeling, we examined the effect of ABA on expression of chemokine and cytokine genes in stroma of prostate tumor in the Pten-knockout mouse model. Using the method reported previously (42), we successfully isolated fibroblasts and macrophages from Pten-knockout prostate tumors in mice treated with or without ABA for a month (Supplementary Fig. S5A). The authentication of isolated fibroblasts and macrophages was confirmed by their expression of platelet derived growth factor receptor α (PDGFRα) and colony stimulating factor-1 receptor (CSF-1R), respectively (Supplementary Fig. S5B).

In cancer-associated fibroblasts (CAF) in Pten-knockout prostate in castrated mice, ABA treatment induced significant downregulation of several cytokine and chemokine genes examined (Fig. 7A and B), including interferon γ-induced protein IP-10 (or termed mCXCL10), macrophage inflammatory protein-1a (MIP-1a or termed mCCL3), macrophage inflammatory protein-1b (MIP-1b or termed mCCL4), regulated upon activation, normal T-cell expressed and secreted (RANTES or termed mCCL5), interleukin-10 (mIL-10), transforming growth factor β1 (mTGFβ1). ABA treatment also significantly upregulated expression of several cytokine and chemokine genes in CAFs (Fig. 7C), including tissue inhibitor of metalloproteinase 1 (mTIMP1), keratinocyte-derived chemokine (KC or termed mCXCL1), chemokine (C-X-C motif) ligand-2 (mCXCL2), thymus and activation regulated chemokine (TARC or termed mCCL17), stromal cell derived factor-1 (mSDF1), and interleukin-1 β (IL-1β). We also detected similar expression changes of these cytokine and chemokine genes in tumor-associated macrophages, such as significant downregulation of several cytokines and chemokines in tumor-associated macrophages (Fig. 7D), including mCXCL10, mCCL3, mCCL4, mCCL5, mIL-10, mCXCL2, and IL-1β. We further showed that ABA induced significant upregulation of several cytokine and chemokine genes in tumor-associated macrophages (Fig. 7E).

The functional importance of the CAFs in response to neoplastic epithelium can be attributed, at least in part, to the dynamics of expression of cytokines and chemokines in stromal cells and the feedback impact on cancerous epithelium. Matrix metallopeptidase 9 (MMP9) is a collagenase involved in degradation of the extracellular matrix. We demonstrated that Mmp9 protein expression was decreased in ABA-treated Pten-null prostate tumors compared to mock treatment (Fig. 7F and G). Runx2 heterozygous deletion also reduced Mmp9 expression in Pten-knockout prostate tumors in castrated mice (Supplementary Fig. S5C). Thus, both deletion of Runx2 and ABA treatment decrease Mmp9 expression in Pten-null tumors in mice, providing a plausible explanation of the effect of ABA treatment or Runx2 heterozygous deletion on TME remodeling.

Increased synthesis of testosterone and further conversion to dihydrotestosterone (DHT) in prostate cancer cells, through either de novo route after castration or from weak androgens produced by the adrenal gland, is one of the major mechanisms that likely contribute to AR reactivation during prostate cancer progression. Findings from multi-institutional integrative clinical sequencing studies reveal that the PTEN gene is frequently lost in advanced prostate cancer (43, 44). Moreover, PTEN loss is associated with worse recurrence-free survival after radical prostatectomy (7) and androgen depletion therapy of patients (8). Thus, it can be postulated that PTEN loss in prostate cancer may contribute to IAS. Indeed, Pten inactivation in mice inevitably results in the emergence of castration-resistant, but AR-dependent prostate cancer, although the underlying mechanisms were largely unclear (36, 45). We demonstrated previously that PTEN loss-induced inactivation of FOXO1 promotes oncogenic activation of RUNX2, thereby driving promiscuous expression of its downstream target genes involved in cell migration and invasion and prostate cancer progression (12). OCN is a well-established RUNX2 target gene (12, 46). Intriguingly, previous studies in mice show that OCN functions as a bone hormone that binds to the G protein–coupled receptor GPRC6A in Leydig cells to promote testosterone biosynthesis in the testis by inducing expression of androgen biosynthesis enzyme genes such as CYP11A1 and CYP17A1 (19). Based upon our findings in vitro and in mice in this study, we envision a model (Supplementary Fig. S5D) wherein the activation of RUNX2 caused by PTEN loss promotes IAS in prostate cancer via activation of the OCN–GPRC6A–CREB–CYP17A1 axis. Importantly, we provide in vivo evidence that treatment of Pten-null tumors in castrated mice with the CYP17A1 inhibitor ABA not only decreased testosterone levels in tumor tissues, but also caused inhibition of tumor growth and progression by inducing apoptosis in prostate tumors. Thus, our findings identify a signaling cascade that could be a potential target for effective treatment of PTEN-null CRPC.

Invasion of carcinoma cells into the surrounding stromal niche is a prerequisite for distant metastasis. A highly condensed layer of SMA-positive stroma, which is regarded as a natural barrier against the invasion of tumor cells into the adjacent stromal tissue, circuits around the acini in the normal prostate. We found much loosened layers of SMA-positive stroma surrounding malignant epithelial cells in the prostate of Pten-knockout mice. Runx2 heterozygous deletion almost completely reversed Pten knockout-induced formation of multiple layers of SMA-positive stroma outside malignant epithelial cells, suggesting a causal role of Runx2 in Pten loss-induced remodeling of TME. In agreement with the role of Runx2 in testosterone synthesis in Pten-knockout prostate tumors in mice, blockade of testosterone synthesis by ABA also abolishes Pten loss-induced remodeling of stromal niche in castrated mice. Thus, our findings demonstrated for the first time that RUNX2 plays an important role in modulation of TME and this effect is likely mediated through the impact on the stroma compartment imposed by RUNX2-mediated testosterone synthesis in tumor cells.

Our finding that Pten-knockout prostate cancer cells intratumorally produce significant amount of testosterone provides a molecular explanation for the castration-resistance phenotype of prostate tumors in Pten-knockout mice. The importance of IAS in castration-resistant growth of Pten-null tumors is further supported by the observation that ABA treatment substantially reduces tumor burden by decreasing intratumoral testosterone levels. The effectiveness of ABA treatment is also manifested by decreased expression of Ar protein in mouse prostate tumors, a phenomenon consistent with previous finding in cultured cells that androgens induce accumulation of AR protein (37, 47).

ADT regulates the state of active stroma and modulates expression of chemokines and cytokines by blocking AR signaling pathway in stromal cells in the context of PTEN haploinsufficiency (48). A number of chemokines and cytokines have been shown to be secreted by different stromal and epithelial cell types and play important roles in microenvironment-mediated promotion of tumorigenesis (49, 50). We uncovered that ABA treatment substantially decreased expression of the collagenase MMP9, which plays a major role in degradation of the extracellular matrix. Our data suggest that while ABA treatment abrogates the AR signaling, it also induces TME remodeling by affecting expression of extracellular matrix modulators such as MMP9.

In summary, we demonstrated that PTEN loss or activation of AKT promotes IAS and this process is mediated by hyperactivation of RUNX2 and its downstream OCN–GPRC6A–CREB signaling cascade and increased expression of steroid biosynthesis enzyme genes CYP11A1 and CYP17A1. We also showed that Pten deletion induces TME remodeling in castrated mice. However, the effects of PTEN loss on androgen biosynthesis and TME remodeling in prostate tumors are completely blocked by Runx2 heterozygous deletion or treatment of mice with ABA, a prodrug of the CYP17A1 inhibitor. Therefore, our findings not only uncover a potential mechanism of emergence of the intrinsic castration-resistant phenotype in PTEN-null prostate cancer cells, but also identify an important role of the interplay between epithelial and stromal cells in development of resistance to abiraterone therapy in PTEN-deficient CRPC.

R.J. Karnes reports receiving other commercial research support from GenomeDx. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.J. Westendorf, R. Zhu, H. Huang

Development of methodology: Y. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Yang, Y. Bai, Y. He, J. Chen, L. Ma, Y. Pan, J. Zhang, R.J. Karnes, M. Kohli

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Yang, Y. He, Y. Zhao, J. Zhang, M. Kohli

Writing, review, and/or revision of the manuscript: M. Hinten, J. Zhang, R.J. Karnes, J.J. Westendorf, B. Li, R. Zhu, H. Huang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Yang, Y. Zhao, W. Xu

Study supervision: J. Zhang, H. Huang, W. Xu

This work was supported in part by grants from NIH (CA134514, CA130908, and CA193239; to H. Huang), DOD (W81XWH-14-1-0486; to H. Huang), Natural Science Foundation of China (81270022, 81611130070, and 81771898; to W. Xu), and Natural Science Foundation of Heilongjiang Province of China (no. QC2014C111; to Y. Yang). ABA was kindly provided by the Janssen Research and Development, LLC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.