SOX9 Is Expressed in Normal Prostate Basal Cells and Regulates Androgen Receptor Expression in Prostate Cancer Cells

The human prostate is composed of prostatic glands with well-defined basal and luminal epithelial cell layers, which are embedded in a fibromuscular stroma. Cells within the luminal epithelium have a very low rate of proliferation and express high levels of androgen receptor (AR). These cells are differentiated with a primary function of seminal fluid synthesis, a process that is strongly androgen dependent, and undergo atrophy or apoptosis in response to androgen withdrawal. In contrast, basal cells have a higher rate of proliferation, express low or undetectable levels of AR, and are not androgen dependent (1–3). The precise functions of basal cells are not clear, but accumulating evidence indicates that they play critical roles in prostate organogenesis and homeostasis. Basal cells form a continuous single-cell layer between the fibromuscular stroma and the luminal epithelium, thus providing support and a barrier for the luminal cells. The basal cell layer is also believed to contain prostatic stem cells, which can differentiate into basal, luminal, and neuroendocrine cells (4–10). Knockout of p63, a p53 family protein expressed selectively in basal cells, disrupts formation of a basal cell layer and development of prostate glands (11, 12). Moreover, p63-expressing basal cells are capable of repopulating the luminal cell layer (13). The basal cell layer becomes discontinuous in prostatic intraepithelial neoplasia, which is believed to be a precancerous lesion. The complete loss of the basal cell layer is a defining feature of prostate cancer. Prostate cancer cells resemble luminal epithelial cells in their high-level expression of AR and androgen dependence, but they also express many protein markers characteristic of basal cells and clearly resemble basal cells with respect to increased proliferative capacity.

One group of genes that may regulate aspects of prostate development and function are members of the SOX [Sry-related high-mobility group (HMG) box] family. SOX proteins are a large family of transcription factors that share a homologous HMG DNA-binding domain and are key regulators of many developmental and tissue-specific processes (14). The HMG DNA-binding domain binds to DNA in the minor groove, and the architectural HMG proteins (HMG-1 and HMG-2) have been found to enhance steroid hormone receptor binding to DNA by locally altering DNA conformation and through direct protein-protein interactions (15–17). We have similarly identified direct interactions between AR and the sequence-specific HMG proteins SRY and T-cell factor 4 (18, 19). SOX9 in the developing gonad plays a critical role in male sex determination by stimulating expression of anti-Mullerian hormone (AMH, also known as Mullerian-inhibiting substance), a transforming growth factor β–like hormone that causes regression of the female Mullerian ducts (20). SOX9 interacts with steroidogenic factor 1 on the Amh promoter to directly stimulate AMH expression. Moreover, loss of SOX9-mediated AMH production contributes to XY sex reversal, whereas increased SOX9 expression in XX male (Odsex) mice or transgenic mice is sufficient to cause a female to male reversal even in the absence of Sry (21). Conditional SOX9 knockout in the developing gonad shows that SOX9, expressed by Sertoli cells, is also essential for Sertoli cell differentiation and seminiferous tubule formation (22, 23).

SOX9 is expressed in multiple other tissues during embryogenesis, including cartilage, neural crest, notochord, kidney, pancreas, and endocardial cushions of the heart. Heterozygous SOX9 mutations are the cause of the human disease campomelic dysplasia, a form of dwarfism characterized by extreme cartilage and bone malformation, which is frequently associated with XY sex reversal and other anomalies (24–27). Homozygous knockout of SOX9 in mice results in embryonic lethality, whereas SOX9 heterozygous knockouts exhibit the same skeletal anomalies as campomelic dysplasia patients (28, 29). The tissue-specific inactivation of SOX9 in limbs results in the absence of cartilage and bone formation, whereas overexpression of SOX9 induces chondrogenic cell differentiation, revealing that SOX9 is required for cartilage development (28–30). Consistent with these defects, major SOX9 target genes in cartilage cells include type II collagen (Col2a1), type XI collagen (Col11a2), and aggrecan, which are all important components of cartilage (29, 31).

SOX9 knockout in neural stem cells results in defects in specification of oligodendrocytes and astrocytes, indicating that the switch from neurogenesis to gliogenesis fails to take place (32–34). In the intestine, SOX9 is expressed in the crypt, which is populated by progenitor/stem cells. Moreover, SOX9 expression in intestinal crypt cells is regulated by the Wnt/β-catenin signaling pathway, which is required to maintain the stem cell compartment (35, 36). SOX9 is also found in the outer root sheath compartment of hair follicles, where its expression is regulated by sonic hedgehog signaling (37). SOX9 conditional knockout results in hair loss and absence of the stem cell compartment, further supporting a critical role for SOX9 in maintaining stem cells.

In this study, we show that SOX9 in normal adult prostate is expressed in basal epithelial cells, with no detectable expression in luminal epithelium. SOX9 was also expressed in a subset of primary prostate cancer in vivo, at increased frequency in recurrent prostate cancer, and in prostate cancer cell lines, and its expression in prostate cancer cells was regulated by the Wnt/β-catenin pathway. Similarly to other HMG proteins, SOX9 interacted with the AR and this interaction was dependent on the C-terminal extension (CTE) in the AR DNA-binding domain. Transient transfections and doxycycline-inducible expression of SOX9 in LNCaP prostate cancer cells showed that SOX9 expressed at very high nonphysiologic levels could suppress AR protein expression, but SOX9 expressed at lower levels enhanced AR protein expression. Moreover, small interfering RNA (siRNA)–mediated down-regulation of SOX9 in prostate cancer cells caused a decline in AR protein levels and suppressed cell growth. Taken together, these observations indicate that SOX9 expressed in normal prostate basal cells may play roles in maintaining a committed stem cell/progenitor cell compartment, or in regulating the expression of factors that support the luminal epithelium. Moreover, the expression of SOX9 protein in prostate cancer cells may be critical to maintain proliferative potential and growth independently of basal cells.

Plasmids, reagents, and cell lines. The pcDNA3-SOX9 plasmid, which expresses a Flag-tagged SOX9, was a gift from Dr. P. Berta (Human Molecular Genetics Group, Institut de Genetique Humane, Montpellier, France; ref. 20). The SOX9-regulated luciferase reporter (4x48-p89-luciferase) was provided by Dr. B. de Crombrugghe (Department of Molecular Genetics, University of Texas MD Anderson Cancer Center, Houston, TX; ref. 38). AR expression vectors (pSVARo), ARE₄-luciferase reporter, and Renilla luciferase control vector (pRL-CMV, Promega, Madison, WI) have been described previously (18). pcDNA6/TR and pcDNA4/TO/myc-His were purchased from Invitrogen (Carlsbad, CA). pcDNA4/TO/SOX9 was generated by inserting a BamHI-BamHI fragment from pcDNA3-SOX9 into the BamHI site of pcDNA4/TO/myc-His.

SB415286, a glycogen synthase kinase 3β (GSK3β) inhibitor, was purchased from Tocris Cookson (Avonmouth, Bristol, United Kingdom). Antibodies were from the following sources: anti-SOX9 (H90, C20 from Santa Cruz Biotechnology, Santa Cruz, CA, and 09-1, a kind gift from Dr. M. Wegner, Institut fur Biochemie, Emil-Fischer-Zentrum, Universitat Erlangen, Erlangen, Germany; ref. 32), anti–prostate-specific antigen (PSA; Biodesign, Saco, ME), anti–β-tubulin (Chemicon, Temecula, CA), anti–β-catenin (Sigma, St. Louis, MO), anti-AR (Upstate Biotechnology, Lake Placid, NY), anti-p27 (Santa Cruz Biotechnology), and anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Abcam, Cambridge, MA). Secondary anti-mouse and anti-rabbit antibodies were from Promega.

CV-1, LNCaP, PC3, and C3H10T cells were obtained from American Type Culture Collection (Manassas, VA). LNCaP cells were maintained in RPMI with 10% fetal bovine serum (FBS). PC3 and C3H10T cells were maintained in DMEM with 10% FBS. CWR22R3 cells were established by our laboratory from a CWR22 xenograft that relapsed after castration and bicalutamide treatment, and have been cultured long term (>2 years) in DMEM supplemented with 10% charcoal-dextran–treated FBS (39). CWR22Rv1 cells were from Dr. R.M. Sramkoski (Cancer Research Center, Case Western Reserve University, Cleveland, OH) and were maintained in DMEM with 10% FBS (40). RCS cells, a rat chondrosarcoma cell line, were a gift from Dr. de Crombrugghe and were maintained in DMEM with 10% FBS. C4-2 was a gift from Dr. L. Chung (Department of Urology, Emory, University School of Medicine, Atlanta, GA) and were maintained in T medium (Life Technologies, Inc., Gaithersburg, MD) supplemented with 5% FBS. Primary epithelial cells were purchased from Clonetics Corporation (San Diego, CA) and were maintained using the suggested prostate epithelial cell growth medium.

Immunohistochemistry. Five-micrometer sections from paraffin-embedded tissue blocks were deparaffinized, rehydrated, and underwent antigen retrieval by microwaving at high setting for 30 min in 10 mmol/L citrate buffer (pH 6.2). After cooling to room temperature, the tissue sections were blocked using 5% goat serum and avidin blocking solution (Vector, Burlingame, CA). Primary antibodies were then added and incubated overnight at 4°C. The anti-AR or anti-SOX9 antibodies were used at 1:50. After four washes in PBST (PBS with 0.05% Tween 20), the antibodies were detected using biotinylated goat anti-rabbit antibody at 1:400 followed by streptavidin-horseradish peroxidase (HRP) at 1:400 (Vector). After an additional four washes with PBST, slides were developed with 3,3′-diaminobenzidine and counterstained with hematoxylin.

Immunoblotting. Prostatectomy samples were excised and minced in PBS into 1-mm³ pieces, homogenized with a glass Dounce homogenizer, and sonicated in radioimmunoprecipitation assay (RIPA) lysis buffer [50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L sodium chloride, 1.0% NP40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mmol/L EDTA, and 1 mmol/L EGTA] containing protease and phosphatase inhibitors. Cultured cells were directly lysed with RIPA buffer containing protease and phosphatase inhibitors (antipain hydrochloride, 100 μmol/L; aprotinin, 0.2 μg/mL; AEBSF, 1 mmol/L; E-64, 10 μmol/L; leupeptin hemisulfate, 100 μmol/L; pepstatin, 1 μmol/L; glycerol phosphate, 1 mmol/L; NaPPO₄, 1 mmol/L; and sodium vanadate, 1 mmol/L). Protein quantity was determined by Bradford assay (Bio-Rad Laboratories, Inc., Hercules, CA). Proteins were separated by SDS-PAGE under reducing conditions and then transferred to 0.45-μm nitrocellulose membranes by electroblotting. The membranes were blocked with 5% nonfat powdered milk in PBS and then probed with primary antibodies at a 1:1,000 dilution in TBS containing 0.2% Tween 20 (TBST) with 5% milk. The membranes were then washed extensively with TBST and probed with HRP-conjugated secondary antibodies at 1:2,000 dilutions in TBST with 5% milk. After further washing in TBST, the membranes were developed with enhanced chemiluminescence Western blotting detection system (Pierce Biotech, Rockford, IL). The Image J program (Wayne Rasband, NIH, Bethesda, MD) was used to quantify protein band densities on immunoblots according to author's instructions, which were normalized to β-tubulin or GAPDH and expressed as relative density.

Dephosphorylation with calf intestinal phosphatase. CWR22Rv1 cells were plated in a six-well plate the day before collection. At the time of collection, the cells were washed once with cold PBS and scraped with a rubber policeman into passive lysis buffer (Promega) supplemented with protease inhibitors (1 μg/mL aprotinin, 1 μg/mL pepstatin, 0.5 μg/mL leupeptin, 0.2 mg/mL AEBSF, and 1.6 mg/mL iodoacetamide). Lysates were incubated with rotation for 20 min at room temperature and cleared by centrifugation for 15 min in a microcentrifuge at 4°C. Forty microliters of cleared cell lysates with or without added calf intestinal phosphatase (40 units, New England Biolabs) were incubated at 37°C for 5 h. The treated lysates were subsequently immunoblotted for SOX9 protein.

Transfection. One day before transfection, cells were plated into 24- or 48-well plates at a density of 70% to 80%. The cells were transfected with mixtures of DNA and LipofectAMINE 2000 (Invitrogen) for 24 h, according to the manufacturer's recommendations. Cells were then switched to fresh medium containing various treatment reagents for another 24 h and then lysed with passive lysis buffer and analyzed for luciferase activity using the Dual-luciferase measurement system (Promega). The siRNAs were similarly transfected at 40 pmol/mL using LipofectAMINE 2000, and the cells were studied at 48 to 72 h posttransfection. The following siRNAs were purchased from Dharmacon (Lafayette, CO). β-Catenin siRNAs were a mixture of four different siRNA duplexes (M-003482-00). SOX9 siRNAs were as follows (positive strands): A, 5′-GCAGCGACGUCAUCUCCAAUU-3′; B, 5′-CAACGAGUUUGACCAGUACUU-3′; and control siRNA (siCONTROL Non-Targeting siRNA 1, D-001210-01).

Real-time reverse transcription-PCR. Total RNA was isolated from cultured cells using the RNeasy protect mini kit (Qiagen, Valencia, CA). The amount of total RNA was determined by spectrophotometer, and 100 ng of total RNA from each sample were used to determine the specific RNA level by TaqMan real-time reverse transcription-PCR (RT-PCR) using an ABI Prism 7000 (Applied Biosystems, Foster City, CA). The primer sequences for human SOX9 were as follows: 5′-TATGACTGGACCCTGGTG-3′ (forward); 5′-TGTGGCTTGTTCTTGCTGG-3′ (reverse) and 5′-FAM-TGCCGGTGCGCGTCAACG-3′ (probe). The primers for human AR were as follows: 5′-GGAATTCCTGTGCATGAAA-3′ (forward); 5′-CGAAGTTCATCAAAGAATT-3′ (reverse); and 5′-FAM-CTTCAGCATTATTCCAGTG-3′ (probe). The validated 18S human rRNA assay primer/probe set was purchased from Applied Biosystems. Relative quantitation with the comparative threshold cycle (C_t) method was done as recommended by ABI. The C_t is the fractional cycle number at which the amplified target reaches a fixed threshold. The amount of target gene normalized to an endogenous reference gene (18S rRNA) is given by 2^−ΔCt, where ΔC_t is C (target gene) − C_t (reference gene).

Establishment of LNCaP cell lines expressing inducible SOX9. The T-REx system (Invitrogen) was used to generate LNCaP cells that can be induced to express human SOX9 or SOX9-specific short hairpin RNA (shRNA). Two LNCaP clones that constitutively express high levels of tetracycline repressor were first established. Each cell line showed similar responsiveness as their parental cells to dihydrotestosterone (DHT), as indicated by the PSA level changes after androgen withdrawal or stimulation. SOX9-inducible lines were subsequently established by selecting clones that carry the pcDNA4/TO/SOX9. SOX9 shRNA–inducible lines were established by transduction with SOX9 shRNA in the tetracycline-regulated pSuperior vector (pSuperior-SOX9i, generated by cloning the above SOX9 siRNA A sequence into pSuperior), followed by puromycin selection.

Glutathione S-transferase pull down. DNA sequences encoding various AR protein fragments were generated by PCR and cloned into the pGEX-2T vectors. Glutathione S-transferase (GST)–AR fusion were purified on glutathione-agarose beads and used to pull down ³⁵S-labeled, in vitro transcribed and translated SOX9 or SRY proteins, as previously described (18).

Cell proliferation and cell cycle analysis. Cells were rinsed in PBS, trypsinized, washed thrice with PBS, and fixed with 95% ethanol at 4°C for 30 min. The cells were then treated with propidium iodide and RNase A at 37°C for 30 min and subsequently analyzed by flow cytometry. The results were analyzed using CellQuest-Pro software. Cell proliferation was studied with the cell proliferation ELISA, bromodeoxyuridine (BrdU) (colorimetric) kit from Roche Applied Science, according the manufacturer's recommendations. Briefly, CWR22Rv1 or C4-2 cells were plated in 96-well plates at 5 × 10³/well and transfected with 5 pmol of SOX9 siRNAs (A or B) or control siRNA. At the 48 h posttransfection, cells were labeled with 100 μmol/L BrdUrd for 24 h. After labeling, the cells were fixed with FixDenat solution at room temperature for 30 min and then incubated with anti–BrdU-POD antibody for 90 min. The cells were subsequently washed thrice with washing solution and incubated with substrate solution at room temperature for 5 to 10 min before the plate was read in an ELISA reader at 490 nm.

SOX9 protein is expressed by basal cells in normal prostate. Transcripts encoding a number of SOX proteins, such SRY, SOX7, and SOX9, have been reported in prostate, with further data indicating that SRY loss may contribute to prostate cancer (41–46). However, our immunohistochemical studies have not identified cells that express detectable levels of SRY protein in adult human prostate, despite the use of several anti-SRY antibodies and positive staining of SRY in human testis (data not shown). In contrast, immunohistochemistry with two independent anti-SOX9 antibodies showed strong nuclear and weaker cytoplasmic expression of SOX9 in the basal cell layer of normal prostate glands, with no staining in luminal epithelial cells (Fig. 1A). This SOX9 expression pattern is opposite that of AR, which is highly expressed in the luminal epithelium and is weakly expressed or absent in basal epithelial cells (Fig. 1A). SOX9 expression in testis was predominantly in the Sertoli cells, consistent with its previously reported pattern and supporting the specificity of the SOX9 immunohistochemistry (Fig. 1A).

We next carried out immunoblotting to confirm that SOX9 protein was expressed in adult prostate. As shown in Fig. 1B (left), extracts from three independent histologically normal frozen human radical prostatectomy blocks clearly contained a protein that migrated as a doublet with the same mobility as SOX9 from RCS cells (a rat chondrosarcoma cell line expressing high levels of SOX9). To further assess SOX9 expression in normal prostate basal cells, we examined primary epithelial cells cultured short term from human prostate. These cells are phenotypically basal cells that express a basal cell pattern of cytokeratins and are negative for AR and PSA proteins (47). Immunoblotting again showed a doublet that migrated identically to the RCS SOX9 protein (Fig. 1B , right). Significantly, three independent SOX9 antibodies against different regions of the protein recognized this protein, which supports the conclusion that it is SOX9.

SOX9 is expressed in prostate cancer in vivo and in prostate cancer cell lines. Although SOX9 expression was not detected by immunohistochemistry in normal prostate luminal epithelium, it could be clearly observed in prostate cancer samples (Fig. 1C). The staining in prostate cancer was predominantly nuclear and varied in intensity and distribution from patchy to diffusely positive. In a series of randomly collected primary prostate cancer samples from radical prostatectomies (n = 29), tumor cells with clear nuclear SOX9 expression were readily found in 17 cases. The variation in SOX9 staining did not seem to reflect differences in fixation, as the negative tumor samples still showed basal cell SOX9 staining in the adjacent normal glands (Fig. 1C,, bottom left) and strong nuclear AR staining in the tumor (Fig. 1C,, bottom right). Interestingly, SOX9 expression was further increased in a series transurethral resection of the prostate samples from patients with recurrent prostate cancer after androgen deprivation therapy (n = 37), with >90% of these tumors having SOX9-positive tumor cells (Fig. 1D). Moreover, a larger fraction of the tumor cells were SOX9 positive in these recurrent tumors versus the initial primary tumors, and this difference was statistically significant (Table 1). These results show that SOX9 is expressed by a large fraction of primary prostate cancer, and indicate that there is positive selection for increased SOX9 expression in prostate cancer that relapse subsequent to androgen deprivation therapy.

Based on these results, SOX9 expression in a series of prostate cancer cell lines was next assessed by immunoblotting. Significantly, each of the cell lines examined (CWR22R3, CWR22Rv1, C4-2, LNCaP, and PC3) expressed readily detectable levels of SOX9, which migrated identically to SOX9 in RCS cells and C3H10T cells (a mouse mesenchymal line reported to express SOX9; Fig. 2A). As an approach to further confirm SOX9 expression in the prostate cancer cells, we used siRNA to down-regulate SOX9. Expression of SOX9, as assessed by immunoblotting in CWR22Rv1 cells, was markedly reduced by both of the siRNAs against SOX9, but not by a control siRNA (Fig. 2A,, bottom left). Finally, the migration of SOX9 as a doublet on most gels suggested that it was phosphorylated, as has been previously reported (48). Consistent with this hypothesis, SOX9 ran as a faster migrating single band after treatment with calf intestinal phosphatase (Fig. 2A , bottom right). Taken together, these results show that tumor cells express SOX9 in a large fraction of prostate cancer. However, SOX9 expression is confined to basal cells in normal prostate.

SOX9 expression in prostate cancer cells is modulated by Wnt/β-catenin signaling. Recent studies indicate that SOX9 expression in intestinal crypts and hair follicles is required to maintain stem cell compartments, and that SOX9 expression in these sites is regulated by Wnt or Shh signaling, which are the important pathways for stem cell maintenance (35, 37). Although the identity of stem cells for prostate epithelium remains unclear, the prostate basal cell layer seems to contain precursor cells that can give rise to luminal epithelial cells. These observations suggested that SOX9 expression may be similarly regulated by the Wnt/β-catenin pathway and may contribute to maintaining proliferative potential in normal prostate and in prostate cancer. Therefore, we next examined whether the Wnt/β-catenin pathway regulates SOX9 expression in prostate cancer cells.

Wnt signaling inhibits GSK3β-mediated β-catenin phosphorylation, reducing β-catenin ubiquitylation and proteasome degradation and resulting in an accumulation of β-catenin in the nucleus and transactivation of its target genes (49). Akt also inhibits GSK3β, so that β-catenin is stabilized in PTEN-deficient prostate cancer cells (including PC3, LNCaP, and C4-2, which were derived from LNCaP) due to activation of the phosphatidylinositol 3-kinase/Akt pathway (50, 51). To stimulate the Wnt/β-catenin pathway in prostate cancer cells with intact PTEN, we treated the DU145 prostate cancer cell line with a direct GSK3β inhibitor, SB415286, which caused a marked and dose-dependent increase in β-catenin protein levels (Fig. 2B). Significantly, SB415286 also strongly increased the expression of SOX9 protein, consistent with Wnt/β-catenin regulation of SOX9 expression. Similarly, inhibition of GSK3β by SB415286 in CWR22Rv1 cells led to an increase in β-catenin and SOX9 protein levels (Fig. 2B).

We next used quantitative real-time RT-PCR to determine whether SOX9 message levels were increased in response to GSK3β inhibition by SB415286. As shown in Fig. 2C, SB415286 caused an increase in SOX9 message levels at 6 h, with a further marked increase at 24 h, which indicates that Wnt signaling increases SOX9 transcription. Finally, as GSK3β inhibition can modulate many proteins in addition to β-catenin, we used β-catenin siRNA to determine whether SOX9 expression was regulated by β-catenin. Transfection with β-catenin siRNA caused a marked reduction of β-catenin protein levels in CWR22Rv1 cells compared with control siRNA (Fig. 2D). Significantly, there was a corresponding decrease in SOX9 protein, which indicates that SOX9 was positively regulated by β-catenin. Taken together, these results show that SOX9 is a Wnt/β-catenin pathway–regulated gene in prostate cancer cells.

SOX9 directly interacts with AR. We previously reported that SRY and AR could interact through their respective DNA-binding domains, and that SRY could suppress AR transcriptional activity (18). Given the homology between the HMG DNA-binding domains of SRY and SOX9, we tested whether there was also an interaction between SOX9 and AR by using a series of GST-AR DNA-binding domain fusion proteins. As shown in Fig. 3A, in vitro transcribed/translated SOX9 could bind specifically to the GST-AR DNA-binding domain fusion proteins. Significantly, SOX9 binding to GST-AR DNA-binding domain was markedly reduced when eight amino acids immediately after the second zinc finger in AR DNA-binding domain were absent (removing amino acids 629–636). This eight-amino-acid sequence was previously identified as a part of the CTE that is critical for steroid hormone receptors, such as AR and progesterone receptor, to interact with the HMG-1 and HMG-2 proteins (Fig. 3C; refs. 15–17, 52). The observation that SOX9 binding was abrogated by deletion of this CTE from the AR DNA-binding domain fusion protein indicates that SOX9 interacts with AR by a mechanism similar to the HMG proteins. In contrast, SRY binding to GST-AR DNA-binding domain was more efficient and was not affected by removing the eight amino acids of CTE, suggesting that additional sites on SRY might be involved in AR binding (Fig. 3B).

SOX9 interaction with AR was next assessed functionally in transient transfections experiments. SOX9 is a transcription factor with a COOH-terminal transactivation domain, and SOX9 transfection stimulated expression of a reporter gene regulated by a multimerized SOX9 binding site (Fig. 4A). In contrast, SOX9 transfection inhibited DHT-stimulated AR transactivation of an androgen-responsive element–regulated reporter gene (Fig. 4B). This inhibition was not a nonspecific effect, as DHT-independent activity or activity of an internal control cytomegalovirus-regulated Renilla luciferase reporter were not repressed (Fig. 4B and data not shown). Significantly, immunoblotting showed that AR protein expression was decreased at the high levels of SOX9 that repressed AR transcriptional activity (>0.8 ng), whereas lower amounts of SOX9 seemed to moderately enhance AR protein expression (0.16–0.8 ng; Fig. 4C). In contrast, SOX9 had no effect on a control transfected protein [green fluorescent protein (GFP)]. Additional transfection experiments confirmed this biphasic response, with lower levels of transfected SOX9 causing an increase in AR protein expression (Fig. 4D). These results support the hypothesis that AR and SOX9 can interact, although the down-regulation of AR activity at very high levels of SOX9 (due to decreased AR protein or possibly squelching) is probably not a physiologic function.

SOX9 positively regulates endogenous AR expression in prostate cancer cells. Based on the above transfection results, we next examined the effect of SOX9 on endogenous AR protein in prostate cancer cells. For these experiments, we first established LNCaP cell lines expressing tetracycline-inducible SOX9. The induction of SOX9 expression was very rapid, with substantial amounts of SOX9 detected by 2 to 3 h after doxycycline treatment (Fig. 5A). Only the much lower level of endogenous SOX9 was detected when cells were not treated with doxycycline, or in the control cell line that was similarly established with a control vector. Consistent with the transient transfection results, AR protein was modestly reduced after 3 to 6 h when the cells expressed high SOX9 levels, but then recovered after 18 h despite the continued high levels of SOX9 (Fig. 5A). We next used real-time RT-PCR to determine whether SOX9 was regulating AR message levels. As shown in Fig. 5A (bottom), AR message levels did not decline in response to SOX9 induction, indicating that the decrease in AR protein was due to decreased translation or increased degradation. Interestingly, AR message levels increased after ∼18 h, and this corresponded to the recovery in AR protein levels.

Although these findings confirmed that very high levels of SOX9 could decrease AR, the level of SOX9 in these cells after induction is clearly nonphysiologic. Therefore, to determine whether SOX9 at physiologic levels is a regulator of AR expression in prostate cancer cells, we tested whether SOX9 down-regulation by shRNA had an effect on AR levels in LNCaP cells. For this experiment, we used LNCaP cells expressing the tetracycline repressor and stably transduced them with pSuperior-SOX9i, expressing doxycycline-inducible SOX9 shRNA. As shown in Fig. 5B, doxycycline induction resulted in a marked decrease (∼10-fold) in the level of SOX9 protein. Significantly, there was also a decrease in AR expression (∼40%) and a corresponding decrease in expression of the AR-regulated PSA protein (Fig. 5B).

We similarly examined the CWR22Rv1 prostate cancer cell line, which expresses AR and higher levels of SOX9 than LNCaP cells (although it does not express detectable levels of PSA protein). Transfections with two different SOX9 siRNAs (A and B), but not a control siRNA (C), caused a marked decline in SOX9 protein, with a corresponding decline of ∼40% to 80% in AR protein expression (Fig. 5C). Finally, real-time RT-PCR was used to determine whether AR message levels were decreased in response to the SOX9 knockdowns. Significantly, no decline in AR message was observed using either of the SOX9 siRNAs (Fig. 5D). Taken together, these results indicate that endogenous SOX9 enhances AR protein expression in prostate cancer cell lines through translational or posttranslational mechanisms.

SOX9 down-regulation inhibits prostate cancer cell proliferation. The persistent expression of SOX9 in prostate cancer cell lines indicated that it might play a role in supporting cell growth in vitro as well as in vivo. As an initial approach to assess the importance of SOX9, we determined whether transient SOX9 down-regulation by siRNA had effects on proliferation. Significantly, SOX9 siRNAs caused a decrease in proliferation in both C4-2 and CWR22Rv1 cells, as assessed by BrdU incorporation (Fig. 6A and B). Consistent with the decreased BrdU incorporation, cell cycle analysis in CWR22Rv1 cells showed an increase in the G₀-G₁ population at 48 and 72 h, and a corresponding decrease in the S-G₂-M fraction (Fig. 6C). Moreover, immunoblotting showed a substantial increase in expression of the p27 cyclin–dependent kinase inhibitor (Fig. 6D), consistent with a block at G₀-G₁ and further supporting the hypothesis that SOX9 enhances prostate cancer cell proliferation.

Previous studies have established critical roles for SOX9 in the development of multiple tissues and in maintaining the stem cell compartments in adult tissues. In this study, we found that SOX9 was primarily expressed in basal epithelial cells in adult prostate, with no detectable expression in luminal epithelium. In contrast, immunohistochemistry showed that SOX9 was expressed in a subset of primary prostate cancer and was expressed at higher frequency in recurrent prostate cancer. The expression of SOX9 in prostate cancer cells was stimulated by suppression of GSK3β and repressed by β-catenin silencing, indicating that the Wnt/β-catenin pathway regulates SOX9 in prostate cancer. Similar to other HMG proteins, SOX9 interacted with the AR DNA-binding domain. Although a high level of overexpression of SOX9 in transient transfections or in LNCaP prostate cancer cells could suppress AR protein expression, SOX9 expressed at lower physiologic levels enhanced AR protein expression. More significantly, SOX9 siRNA caused a decline in AR protein levels without decreasing AR message levels, indicating that SOX9 can regulate AR expression through a posttranscriptional mechanism. Finally, siRNA down-regulation of SOX9 caused a decrease in proliferation and an increase in p27 expression, indicating that SOX9 has a role in supporting prostate cancer cell growth. Taken together, these observations indicate that SOX9 in normal prostate may play roles in maintaining the proliferative potential of basal cells or in supporting the luminal epithelium, and that these functions may be important for prostate cancer.

Although SOX9 is critical for maintaining the stem cell compartment in other tissues and may function similarly in prostate epithelium, it is clearly not just a marker of prostate stem cells because stem cells are present at very low frequency in adult prostate, whereas SOX9 is expressed diffusely in the basal cell layer. Instead, the relatively uniform expression of SOX9 by most prostate basal cells indicates that it may be critical for broader functions and regulate one or more basal cell–specific proteins. These proteins might include growth factors, cell surface proteins, or extracellular matrix components that are critical for stem/progenitor cell maintenance, or for supporting the overlying luminal epithelium. In either case, we postulate that SOX9-regulated genes expressed by prostatic basal cells are critical to support the generation or survival of luminal epithelial cells and that SOX9 expression in prostate cancer allows the tumor cells to maintain their proliferative potential and survival in the absence of basal cell support.

The spectrum of targets regulated by SOX9 in basal cells remains to be determined, but our data indicate that AR is one of the SOX9-regulated proteins in prostate cancer cells. Consistent with previous studies of other HMG proteins, we found that AR could interact directly with SOX9 and that the interaction was dependent on a short peptide at the carboxyl terminus of AR DNA-binding domain, termed the CTE. This CTE is present in other steroid hormone receptors, suggesting that SOX9 may similarly interact with these receptors (15–17, 52). The AR-SOX9 interaction may serve to stabilize binding to a subset of genes coregulated by AR and SOX9, but such genes have not yet been identified and we have not observed SOX9 enhancement of AR activity using standard AR reporter genes. High-level overexpression of transfected SOX9 could suppress AR protein levels, possibly reflecting degradation of AR-SOX9 complexes; however, this is not physiologic. More importantly, lower levels of exogenous SOX9 could enhance AR expression, and SOX9 siRNA caused a decrease in endogenous AR protein expression without decreasing AR message level, indicating that SOX9 at physiologic levels in prostate cancer cells enhances AR translation or stability.

Interestingly, AR message levels are generally increased in response to treatments that cause AR protein reduction (such as androgen withdrawal). Therefore, the failure of AR message levels to increase after SOX9 down-regulation and AR protein reduction indicates that SOX9 may, in fact, also positively regulate AR message levels. A previous study found that SOX9 expression in SV40 T antigen–transformed prostate epithelial cells (M12 cells) could be induced by insulin-like growth factor binding protein–related protein, which is expressed at increased levels in senescent human prostate epithelial cell cultures (46). Significantly, overexpression of transfected SOX9 in the M12 cells was found to induce expression of message for AR, PSA, and N-cadherin, although cellular proliferation in these cells was reduced by SOX9 overexpression. This latter inhibitory effect on proliferation may reflect differences in the cells or be due to higher expression of transfected SOX9.

Although SOX9 can interact with and regulate AR expression in prostate cancer cells, which express high levels of AR, the significance of SOX9-AR interaction in normal prostate remains to be determined. Indeed, luminal epithelial cells express high levels of AR and are SOX9 negative, clearly demonstrating that SOX9 is not required for AR expression. Conversely, basal epithelial cells express only low levels of AR message and protein, demonstrating that SOX9 alone is not sufficient for high-level AR expression. Our interpretation of these observations is that SOX9 may initiate low-level AR gene expression in cells that are precursors to luminal epithelial cells, and other transcription factors subsequently take over and stimulate high-level AR expression in fully differentiated luminal epithelial cells. SOX9 expression and its regulation of AR in prostate cancer cells would then be consistent with the hypothesis that SOX9-positive prostate cancer cells represent an intermediate developmental stage between basal and luminal epithelial cells.

In summary, this study shows that SOX9 protein is expressed in adult prostate basal epithelium and may play roles in maintaining the committed stem cell compartment, differentiation, and/or supporting the overlying luminal epithelium. Expression of SOX9 in prostate cancer cells indicates that SOX9-regulated genes may similarly play critical roles in supporting prostate cancer growth independently of basal cells. SOX9 expression in prostate cancer cells is Wnt/β-catenin regulated, and AR is one identified downstream target, although the precise mechanisms by which SOX9 regulates AR remain to be determined. The further identification of SOX9-regulated genes should provide new insights into mechanisms that regulate the development of normal prostate and prostate cancer, as well as provide new therapeutic targets.

Grant support: NIH grants R01CA65647 (S.P. Balk) and K01DK64739 (X. Yuan), Department of Defense grant DAMD17-03-0164 (X. Yuan), and the Hershey Family Prostate Cancer Research Fund.

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.

We thank Drs. P. Berta, B. de Crombrugghe, R.M. Sramkoski, and M. Wegner for providing reagents; Drs. R. Rittmaster and M. Gleave (Prostate Centre, Vancouver General Hospital, Vancouver, British Columbia, Canada); Drs. P.A. Abrahamsson, A. Bjartell, and N. Dizeyi (Department of Urology, Malmo University Hospital, Lund University, Malmo, Sweden) for providing paraffin sections of locally recurrent prostate cancers; M. Regan for statistical analyses; and Balk laboratory members for helpful discussions.