Targeting the Unique Methylation Pattern of Androgen Receptor (AR) Promoter in Prostate Stem/Progenitor Cells with 5-Aza-2′-deoxycytidine (5-AZA) Leads to Suppressed Prostate Tumorigenesis*

Background: There is a specific silencing of AR gene expression in prostate stem cells. Results: The induction of AR gene expression inhibited the self-renewal of the prostate stem cells. Conclusion: The induction of AR expression suppressed PCa stem cell-mediated tumorigenicity. Significance: This is the first attempt to suppress prostate cancer growth via epigenetic modification of AR genes in PCa stem cells. Androgen receptor (AR) expression surveys found that normal prostate/prostate cancer (PCa) stem/progenitor cells, but not embryonic or mesenchymal stem cells, expressed little AR with high methylation in the AR promoter. Mechanism dissection revealed that the differential methylation pattern in the AR promoter could be due to differential expression of methyltransferases and binding of methylation binding protein to the AR promoter region. The low expression of AR in normal prostate/PCa stem/progenitor cells was reversed after adding 5-aza-2′-deoxycytidine, a demethylating agent, which could then lead to decreased stemness and drive cells into a more differentiated status, suggesting that the methylation in the AR promoter of prostate stem/progenitor cells is critical not only in maintaining the stemness but also critical in protection of cells from differentiation. Furthermore, induced AR expression, via alteration of its methylation pattern, led to suppression of the self-renewal/proliferation of prostate stem/progenitor cells and PCa tumorigenesis in both in vitro assays and in vivo orthotopic xenografted mouse studies. Taken together, these data prove the unique methylation pattern of AR promoter in normal prostate/PCa stem/progenitor cells and the influence of AR on their renewal/proliferation and differentiation. Targeting PCa stem/progenitor cells with alteration of methylated AR promoter status might provide a new potential therapeutic approach to battle PCa because the PCa stem/progenitor cells have high tumorigenicity.

Changes in patterns of gene expression, except for alterations in DNA sequences, occur through epigenetic changes, including DNA methylation, histone modifications, and alterations in chromatin structure, as well as by microRNA-mediated mechanisms (1).
Methylation/demethylation of genes is one of the powerful epigenetic modifications that regulates gene transcription (2,3). Methylation/demethylation profiles of many genes have been linked with normal development and differentiation (4 -6) as well as with cancer initiation and progression. Systemic methylation changes in certain tumor-related genes were also suggested to be diagnostic markers for tumor development or prognosis (7,8). Moreover, the self-renewal of normal hematopoietic (9,10), neural (11), and embryonic stem cells, as well as cancer stem cells (12), is also suggested to be controlled by methylation of specific genes.
The AR 4 is expressed in most stem cells (13)(14)(15)(16)(17)(18). The AR in prostate stem/progenitor cells plays key roles in prostate development and differentiation processes (19,20) as well as in prostate cancer (PCa) initiation and progression (21,22). AR has CpG islands in the promoter and exon 1, and differential methylation patterns have been linked to various PCa cell lines with differential AR expression (23).
Early reports suggested that PCa stem/progenitor cells might have the potency to promote prostate tumorigenesis (24 -26). The linkage of AR methylation status in PCa stem/progenitor cells to PCa therapy, however, remains unclear. We report here that the epigenetic control might govern AR expression differentially in three types of stem/progenitor cells (embryonic, mesenchymal, and prostatic). We revealed potential mechanisms by which differential methylation patterns in AR promoters might contribute to the prostate tumorigenesis, suggesting that a new therapeutic approach via targeting AR methylation status in PCa stem/progenitor cells to better battle PCa might be generated.

MATERIALS AND METHODS
Cell Lines and Treatments-The mouse embryonic stem cell line, ESD3, was established from 129/Sv blastocysts and purchased from the American Type Culture Collection (ATCC) (Manassas, VA). The mesenchymal stem cell line, D1, was also purchased from ATCC. The mPrE cell line, a normal mouse progenitor cell line, was a kind gift from Dr. Ming Jiang (Vanderbilt University Medical Center, Nashville, TN). The immortalized prostate cancer stem cell (PCSC) line was purchased from Celprogen (San Pedro, CA). All cell lines were maintained in appropriate culture conditions. The LNCaP stem/progenitor cells were isolated by a magnetic sorting method using an antibody of the human stem cell marker, CD133, and confirmed using immunofluorescence (IF) staining. When necessary, cells were treated with 5-aza-2Ј-deoxycytidine (5-AZA) (Invivogen), ASC-J9 (5 M), or DHT (0, 1, and 10 nM) for the indicated time.
Western Blot Analysis-Harvested cells were washed with PBS and lysed in radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.4, 1% Nonidet P-40, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, 1 mM Na 3 VO 4 , 1 mM NaF, 1 mM okadaic acid, and 1 mg/ml aprotinin, leupeptin, and pepstatin). Individual samples (40 g of protein) were separated on 8 -10% SDSpolyacrylamide gel and transferred to PVDF membranes (Millipore, Billerica, MA). Membranes were blocked in a PBS-Tween 20 solution with 5% fat-free milk for 1 h at room temperature, and then the membranes were incubated with appropriate dilutions of specific primary antibodies overnight at 4°C. After washing, the blots were incubated with HRP-conjugated anti-rabbit or antimouse IgG for 1 h. The blots were developed in ECL mixture (Vector Laboratories, Burlingame, CA) and visualized by Imager.
DNA Extraction and Modification by Sodium Bisulfite-Genomic DNAs were isolated from D1, ESD3, mPrE, and PCSC cells using the DNeasy tissue kit (Qiagen, Valencia, CA). Bisulfite modification of DNA was performed using the EpiTect bisulfite kit (Qiagen, Valencia, CA) following the manufactur-er's directions. With this treatment, all unmethylated cytosine residues are converted to thymine, but those that are already methylated (5-methylcytosine) are resistant to this treatment and remain as cytosine residues.
Bisulfite-specific PCR (BSP) and Methylation-specific PCR (MSP)-Bisulfite-treated DNAs were amplified by a nested PCR protocol using the primers described in the supplemental  15 cycles, and 72°C for 180 s. BSP products were purified using the Qiaquick TM gel extraction kit (Qiagen, Valencia, CA) and 1:1000 diluted samples were used as MSP template. After the reaction, PCR products were analyzed by gel electrophoresis. In these reactions, DNAs of DU145 and LNCaP cells were used as positive (methylated) and negative (unmethylated) controls, respectively; as blank controls of MSP reactions, water and unmodified DNAs were used as templates.
Subcloning and Bisulfite Sequencing-PCR products were purified from the gels after electrophoresis using the Qiaquick TM gel extraction kit (Qiagen, Valencia, CA), ligated into pGEM-T easy vector (Promega, Madison, WI), and then introduced into JM109 high efficiency competent cells (Promega, Madison, WI). Transformed cells were then plated on LB agar containing 100 g/ml ampicillin (Invitrogen) and incubated overnight at 37°C. 5-10 individual colonies were selected, and each was inoculated into 3 ml of LB broth containing 100 g/ml ampicillin (Invitrogen) and grown overnight at 37°C. The insert containing plasmid DNA was extracted from the cells using the Eppendorf FastPlasmid Mini-prep kit (Eppendorf, Westbury, NY). Each DNA sample was sequenced using the automated DNA Sequencer using the vector's forward primers (T7). DNA sequencing reactions were performed using the DNA dRhodamine Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Foster City, CA) and ABI3730xl sequencer (Applied Biosystems) according to the manufacturer's instructions.
Chromatin Immunoprecipitation (ChIP) Assay-The ChIP assay was performed using the Methyl-CpG Binding Domain Protein2 (MBD2) ChIP kit (EpiQuik TM , Biokits, Brooklyn, NY). DNA cross-linking was performed by adding 1% formaldehyde into the cell cultures at room temperature for 10 min, and glycine was then added (0.125 M final concentration) for 5 min to stop the cross-linking reaction. Cells were lysed with a lysis buffer with protease inhibitors and sonicated to shear genomic DNA to lengths between 200 and 1000 bp. One-tenth of the cell lysate was used for input control, and the rest was used for immunoprecipitation using MBD2 antibody. After collecting the immunoprecipitates using protein G-agarose columns, protein-DNA complexes were eluted and heated at 65°C to reverse the cross-linking. After digestion with proteinase K, DNA fragments were purified using spin columns and analyzed using PCR for 35 cycles in a sequence of 94°C for 30 s, 58°C for 30 s, and 72°C for 1 min. Specific primer sets were designed to amplify a target sequence within the mouse and human AR promoter as follows: mAR-F, 5Ј-TTAGGGCTGGGAAGGGT-CTAC-3Ј; mAR-R, 5Ј-GTCTCCTGCCTCTGCTGTAAAC-3Ј; hAR-F, 5Ј-CGACAGCCAACGCCTCTTG-3Ј; hAR-R, 5Ј-CCTTGCTTCCTCCGAGTCTTTAG-3Ј. PCR products were electrophoresed in a 1% agarose gel with ethidium bromide and visualized under ultraviolet light.
MTT Assay-Cells were plated onto 24-well plates. At various time points indicated, MTT solution (Promega, Madison, WI) was added onto cells for 2 h, and then media were removed, DMSO was used to dissolve the MTT salt, and OD values were measured at 570 nm.
IF Staining of Cells-Cells were seeded on the 4-well chamber slides and fixed with iced methanol. After fixation, cells were washed with PBS three times for 5 min each, and then cells were blocked with 5% BSA for 1 h. Cells were washed with PBS three times and then incubated with primary antibodies in 5% BSA in PBS overnight at 4°C. Antibodies used were as follows: anti-AR (N20, 1:250; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), anti-CK5 (1:250; Covance, Princeton, NJ), anti-CK8 (1:250; Abcam, Cambridge, MA), and anti-sca-1 (1:250; eBioscience, San Diego, CA). Cells were then incubated with 1:200 diluted biotinylated secondary antibody (Vector Laboratories, Burlingame, CA) and with fluorescent secondary antibodies for IF (either Alexa 594-or Alexa 488-tagged). IF intensity was quantified by ImageJ software.
Sphere Formation Assay-Sphere formation assays were performed as described earlier (27). Briefly, single cell suspensions (1 ϫ 10 3 , in 60 l of medium) were mixed with 60 l of cold Matrigel, and the mixture was placed along the rim of the 24-well plates with a minimum of triplicate experiments. The culture plates were placed in a 37°C incubator for 10 min to let the mixture solidify, and 500 l of medium was then added into the well. Sphere numbers were counted after 7-14 days under an Olympus light microscope, and size differences were also examined.
Colony Formation Assay-LNCaP stem/progenitor cells were resuspended in 0.4% Bactoagar (BD Biosciences), layered on top of 1 ml of 0.8% agarose in 6-well culture plates, and incubated with 1 ml of RPMI complete medium. After 3 weeks, the colonies were visualized by staining with 0.5% crystal violet. The experiment was analyzed in triplicate, and colonies larger than 100 mm in diameter were counted.
Lentivirus Infection-For the infection of GFP-containing lentivirus carrying either vector or AR, 293T cells were transfected with a mixture of DNAs (lentiviral vector pWPI) containing AR siRNA (scramble vectors were used as control), pAX2 packaging plasmid, and pMD2G envelope plasmid (at a 4:3:2 ratio) using Lipofectamine (Invitrogen). After infection, media containing the virus were replaced with normal culture media. Fluorescence microscopy was used to monitor the infection efficiency via checking the green fluorescence signal.
Orthotopic Implantation-PCSCs that had been treated with 5-AZA (5 M) for 6 days were directly injected into mice. Briefly, after anesthesia, the abdomens of 8-week-old male athymic nude mice were surgically opened in sterile environments. PCSCs (5 ϫ 10 5 /site) in 20 l of medium mixed with Matrigel 1:1 (v:v) were injected into lobes of anterior prostate by a 25-gauge needle, and the abdomens were closed by silk sutures. Five mice were used per group. Because DNA hypomethylation is time-dependent and requires 2-3 cell cycles to be effective and 5-AZA is known to have a short half-life in serum, mice received either control (DMSO) or 5-AZA continuously twice a week after the injection (intraperitoneal injections, 0.05 mg/kg 5-AZA and DMSO as vehicle control). Tumors were harvested at 3 weeks, and tumor sizes were compared. The tumor tissues were stained using AR antibodies of AR, Oct4, and CK8. All mice experiments were under a protocol approved by the Institutional Animal Care and Use Committee of the University of Rochester Medical Center.
Statistical Analysis-Values were expressed as mean Ϯ S.D. Student's t test and analysis of variance were used to calculate p values. p values were two-sided and were considered statistically significant when they were Ͻ0.05.

RESULTS
Differential AR Expression in Different Types of Stem/Progenitor Cells-We investigated AR expression in three different types of stem/progenitor cells (embryonic, mesenchymal, and prostatic) using the cell lines ESD3 (established embryonic) (28), D1 cells (mesenchymal) (29,30), mPrE cells (mouse normal prostatic), PCSCs (human PCa), and the human PCa LNCaP stem/progenitor cells (CD133 ϩ population). The mPrE FIGURE 2. Distinct methylation status of the AR promoter-associated CpG island regions in three types of stem cells. A, schematic map of the AR promoter-associated CpG island region. The locations of PCR primers used for BSP and MSP analyses are indicated. The CpG dinucleotide residues with respect to the transcriptional start site (marked as ϩ1) are shown with numbers (mouse AR from Ϫ22 to ϩ231, human AR from ϩ24 to ϩ291), and the CpG sites are shown by lollipop shapes and red font. The mouse AR promoter region contains 16 CpG loci, and human AR contains 20 loci. B, MSP analysis. The bisulfitemodified DNAs isolated from different types of stem cells, D1, ESD3, mPrE, and PCSCs, were amplified with AR primers specific for methylated (top) and unmethylated DNA (bottom). The lanes detecting PCR products containing methylated allele (M) in DNA and unmethylated allele (U) in DNA, respectively, are indicated. DNAs of DU145 (methylated) and LNCaP (non-methylated) cells were used as positive and negative controls, and non-bisulfitetreated PCSC DNAs (PCSC*) and H 2 O were used as blank control in performing MSP. C, direct DNA sequencing of AR in different types of stem cells. The unmethylated cytosines (C) were deaminated and converted to thymines (T), but the methylated 5-methylcytosines remained unaltered upon reaction. CpG sites are marked with stars. D, summarized bisulfite sequencing of AR promoter regions in various types of stem cells. At least five clones from each cell line were analyzed to obtain the percentage of methylation of the CpG dinucleotides. O, individual CpG sites in the CpG island. S/P, stem/progenitor. cell line was originally derived from the mouse prostate epithelial tissues and characterized as epithelial basal cells and 98% positive for stem/progenitor cell markers (31). The PCSCs were originally obtained from a human PCa patient and immortalized by Celprogen (San Pedro, CA) (26,32,33). We also analyzed these cells and found that they are homogenous and positive for stem cell markers, such as Oct4 and Nanog (data not shown). The LNCaP stem/progenitor cells were isolated by magnetic sorting method using an antibody of the human stem cell marker, CD133 (34), and confirmed by IF staining and selfrenewal test on Matrigel (data not shown).
Interestingly, we found little AR expression in the mPrE and PCSCs at both mRNA and protein levels as compared with the ESD3 and D1 stem cells (Fig. 1, A and B). Little AR expression was also detected in LNCaP stem/progenitor cells as compared with LNCaP non-stem/progenitor cells (Fig. 1, C and D). These results are consistent with the recent reports showing little AR expression in the PCa cell lines CD44 ϩ LAPC4 and CD133 ϩ / CD44 ϩ WPE1-NB26 (35)(36)(37) and the primary cells isolated from human PCa specimens (13,38,39).
To test whether the low expression of AR in normal prostate/ PCa stem/progenitor cells is responsive to androgen treatment, luciferase activity was determined using an MMTV-luciferase assay, and results showed that LNCaP stem/progenitor cells had a low response to DHT (1 and 10 nM, the human prostate DHT concentration before and after castration, respectively) (40) compared with parental LNCaP cells, but the mPrE and PCSCs cells failed to respond to DHT treatment (Fig. 1E). tion, and tested whether AR expression profiles in the three types of stem/progenitor cell are changed but found no significant differences (data not shown). We next investigated whether the AR expression difference is due to the methylation difference in the AR promoter-associated CpG island region because numerous studies demonstrated that the methylation of these CpG islands interferes with transcription, resulting in an absence of mRNA and protein expression (41,42).
The schematic map of the AR promoter-associated CpG island region is shown in Fig. 2A. A 253-bp fragment (from Ϫ22 to ϩ231) and a 267-bp fragment (from ϩ24 to ϩ291) located near the transcription start site of the mouse AR and human AR gene (43) were used for direct sequencing, which was correlated with the AR expression (NCBI accession number mouse AR NM-013476, human AR NM-000044). These sites are within the minimal region of the AR promoter necessary for AR expression.
The methylation profiles in three types of stem/progenitor cells were investigated using two methods, MSP (44) and the DNA sequencing method. In performing MSP, two MSP primer sets were designed to specifically amplify either methylated or unmethylated sulfite-modified sequence in the AR promoter-associated CpG islands (supplemental Table 1). Amplification of DNA with either the methylated or unmethylated set of primers would result in PCR products depending on the methylation status of the CpG dinucleotides. Fig. 2B shows the PCR products obtained with different types of stem/progenitor cells. In D1 and ESD3 cells, which express higher levels of AR, the PCR products were obtained only with the unmethylated primer set, indicating no methylation at AR CpG islands. In contrast, the PCR products with the methylated primer set were only obtained in mPrE and PCSCs, suggesting the presence of dense methylation at the AR CpG islands in these cells.
In the D1 and ESD3 cell lines, unmethylated cytosines in the AR promoter region were seen in 16 CpG loci (98.75 and 97.5%), but higher percentages of the methylated cytosines in the AR promoter region were detected in the mPrE (55%) and PCSC cells (93%) (Fig. 2, B-D).
Taken together, results from Fig. 2, B-D, showed that there are significant differences in methylation in CpG islands of AR promoter in the various types of stem/progenitor cells, which may explain the differential AR expression in the various types of stem/progenitor cells (41,42).

Different Expressions of DNA Methyltransferases (DNMTs) and MBD2 (Methyl-CpG-binding Domain Protein 2) Binding Are Correlated with AR Expression/Silencing in Three Types of
Stem/Progenitor Cells-To dissect the potential molecular mechanism(s) responsible for the difference in methylation in the CpG islands of AR promoter in various types of stem/progenitor cells, we analyzed those proteins that could influence the methylation process (45,46) as described in the legend to Fig. 3A. We first found mRNA expression levels of DNMT1 and DNMT3 to be much higher in mPrE cells as compared with those found in ESD 3 and D1 cells (Fig. 3B). We also found that the MBD2 level is significantly higher in mPrE cells compared with the other two types of stem cells. However, when we examined expressions of the MBD1, -3, and -4 mRNAs in three types of stem/progenitor cells (ESD3, D1, and mPrE), we did not found significant differences (supplemental Fig. 2). We then selected MBD2 and applied the ChIP test to determine if the MBD2 binding to AR promoter contributed to its differential methylation pattern in various types of stem/progenitor cells. As shown in the PCR results of the AR promoter (Fig. 3C), MBD2 binding to the AR promoter was shown to be higher in normal prostate/PCa stem/progenitor cells but not detectable in mesenchymal D1 cells or embryonic ESD3 cells.
Together, the results from Fig. 3, A-C, concluded that higher expression of DNMT1/3 and higher binding of MBD2 to the AR promoter might lead to higher methylation levels in the AR promoter in normal prostate and PCa stem/progenitor cells.
Targeting Methylated AR with a Demethylating Agent, 5-AZA, Led to Induced AR Expression in Prostate Stem/Progenitor Cells-To test whether the silenced AR expression could be reversed by demethylation in prostate stem/progenitor cells, a demethylating agent, 5-AZA, was introduced into the cells. AR expression was induced in normal prostate/PCa stem/progenitor cells upon 5-AZA treatment (Fig. 4, mRNA level (A) and protein level (B); mPrE (a), PCSCs (b), and LNCaP stem/progenitor (c)). In contrast, little induction was observed in LNCaP non-stem/progenitor cells (Fig. 4, A and B, d). The induction of AR expression was shown to be in a time-dependent manner (Fig. 4B), and the maximal induction was obtained by adding 5 M 5-AZA for 6 days of incubation.
To evaluate whether the newly expressed AR proteins induced by the demethylation process are functional, transcriptional activity was conducted at various concentrations of DHT. The increased luciferase activity at different concentrations of DHT in three kinds of normal prostate/PCa stem/progenitor cells (Fig. 4C) indicated that the newly expressed AR proteins are functional. MSP and DNA sequencing analysis results confirmed the 5-AZA-induced demethylation of the CpG islands in the AR promoter (Fig. 4, D and E).

Induced AR Expression Led to Reduced Stem Cell Marker Expression but Increased Expression of Differentiation Markers-
After treating normal prostate/PCa stem/progenitor cells with 5-AZA for 10 days, we observed the morphologic changes under a microscope. As shown in Fig. 5A, some cells acquired the morphology of terminally differentiated luminal cells. We therefore questioned if the 5-AZA-mediated AR induction can trigger cell differentiation. We examined the expressions of the stem/progenitor and differentiation markers before FIGURE 5. Demethylation of AR affects stemness/differentiation of normal prostate/PCa stem/progenitor cells. A, morphology of normal prostate/PCa stem cells (before and after 5-AZA treatment). B, qPCR analysis. mPrE and LNCaP-stem/progenitor (S/P) cells were treated with 5-AZA (5 M, 4 -6 days), and expressions of stem (Sca-1 for mPrE; CD133 and integrin for LNCaP stem/progenitor) and differentiation markers (CK5, CK8, Nkx3.1, and PSA/PSP94) were analyzed. C, IF staining. mPrE and PCSC cells were treated with 5-AZA as in B, and marker expressions, before and after 5-AZA treatment, were examined by IF staining (the inset represents DAPI staining) (magnification, ϫ200). Quantitation was done by ImageJ software. The y axis label IOD denotes integrated optical density. D, qPCR analysis. PCSCs were infected with lentivirus carrying AR/vector, and expressions of the stem cell markers, Oct4, Nanog, and CXCR4, were analyzed. Error bars, S.D. *, p Ͻ 0.05, **, p Ͻ 0.01, ***, p Ͻ 0.001. NOVEMBER 16, 2012 • VOLUME 287 • NUMBER 47 JOURNAL OF BIOLOGICAL CHEMISTRY 39961 and after 5-AZA treatment. As shown with mRNA (Fig. 5B) and IF staining (Fig. 5C, quantitation shown at right) analyses, the expressions of the stem cell markers, sca-1 (mouse mPrE cells) and CD133 (human PCSCs), were decreased upon 5-AZA treatment.

Unique Methylation in Prostate Stem Cell AR
The normal prostate/PCa stem/progenitor cells are generally considered as originating from CK positive basal epithelial cells (47), although the presence of the luminal epithelial originated stem cells has also been reported (48). As expected, we found decreased expression of CK5 in the 5-AZA-treated PCSCs and LNCaP stem/progenitor cells but increased expression of differentiation markers (CK8 and Nkx3.1 in mPrE cells; CK8 and PSA in LNCaP stem/progenitor cells) (Fig. 5B).
5-AZA treatment might affect expressions of other proteins needed for stemness. Therefore, we tested whether AR depletion is essential in maintaining stemness in prostate stem cells. When we added AR into PCSCs by lentiviral infection, significant down-regulation of stem cells markers (Oct4, Nanog, and CXCR4) was observed (Fig. 5D), suggesting that the stemness changes upon 5-AZA treatment in PCa stem/progenitor cells may be mainly due to the induction of the AR expression. Together, the results from Fig. 5, A-D, suggested that the AR methylation status in the AR promoter in normal prostate/PCa stem/progenitor cells is critical in maintaining stemness versus protecting from differentiation.
Inhibition of Self-renewal/Growth of Prostate Stem/Progenitor Cells upon 5-AZA Treatment-To determine if increased AR expression could also influence cell self-renewal/growth in those three kinds of prostate stem/progenitor cells, we carried out MTT assays, and results showed that the 5-AZA treatment resulted in significant inhibition of mPrE and LNCaP stem/ progenitor cell growth (Fig. 6A, a and b) and sphere formation in PCSCs and LNCaP stem/progenitor cells (Fig. 6B). In contrast, we found little difference in the growth of the LNCaP non-stem/progenitor cells upon 5-AZA treatment (Fig. 6A, c), suggesting that the induction of AR expression by 5-AZA is critical in blocking self-renewal of prostate stem/progenitor cells but not the growth of non-stem/progenitor cells.
To further prove that the 5-AZA induced inhibition of selfrenewal/growth of stem/progenitor cells was mainly through demethylation of AR promoter, we co-treated cells with two molecules that could target AR specifically, AR siRNA and an AR degradation enhancer, ASC-J9 (49,50) to determine if suppression of AR could block the 5-AZA effect on inhibiting cell self-renewal/growth. We observed that the inhibitory effect of 5-AZA on the self-renewal ability of prostate stem/progenitor cells was reversed upon co-administration of ASC-J9 (Fig.  6C) or AR siRNA (Fig. 6D), suggesting that the self-renewal inhibitory effect exerted by 5-AZA was mainly through the induction of AR expression.
Targeting Methylated AR with 5-AZA Leads to Reduced Prostate Tumorigenicity-Our recent published results (37) and others (51) suggested that the PCa stem/progenitor cells have higher tumorigenic ability than the non-stem/progenitor cells. It was also shown in the orthotopic xenograft mouse studies that the LNCaP stem/progenitor cell-derived tumors were dramatically reduced when AR was incorporated into the cells before the injection (37). We were, therefore, interested to see if targeting methylated AR in PCa stem/progenitor cells could reduce their tumorigenicity. As expected, we found that the 5-AZA treatment greatly reduced the colony forming ability of the PCa stem/progenitor cells (Fig. 6E), indicating that demethylation of AR promoter reduced tumorigenicity of PCa stem/progenitor cells.
We further extended this in vitro tumorigenesis assay into in vivo orthotopic xenograft mice studies using PCSCs. We used PCSCs because their endogenous AR expression level is lower than that of LNCaP stem/progenitor cells, so we expected to observe the AR effect more significantly in these cells. Mice were divided into two groups; one group of mice were injected with PCSCs that had been treated with 5-AZA (5 M) for 6 days, and the other group of mice were given cells treated with vehicle. After cell injections, mice of each group were continuously treated with 5-AZA or vehicle (intraperitoneally). After 3 weeks of treatments, the mice were sacrificed, and the tumor developments were compared. As shown in Fig. 6F, smaller tumors were observed in the 5-AZA-treated mice compared with the control group mice. When we performed histological examination of tumor tissues, high numbers of the AR positively stained cells were detected in the 5-AZA-treated mouse tissues, confirming the induction of AR expression by 5-AZA treatment (Fig. 6G). We also examined expressions of the stem cell marker (Oct4) and differentiation marker (CK8). The expression of Oct4 was shown to be significantly decreased, whereas the expression of CK8 was increased in tissues of the 5-AZAtreated mice compared with the control group mice tissues (Fig. 6G). PCSCs were infected with lentivirus carrying AR siRNA (siAR) or scrambled (sc) control plasmids and then treated with 5 M 5-AZA for 5 days. Self-renewal abilities of these two groups of cells were analyzed by a sphere formation assay in a similar manner. E, soft agar assay. LNCaP stem/progenitor cells were treated with 5-AZA (5-20 M), and colony forming ability was tested on soft agar. F, mouse tumor weight analysis. Mice were injected orthotopically with PCSCs (treated with 5-AZA or DMSO) into anterior prostate lobes of nude mice, and mice were divided into two groups. One group of mice received 5-AZA, and the other group of mice were injected with DMSO (control) continuously twice a week after the injection of cells (intraperitoneal injections, 0.05 mg/kg 5-AZA and DMSO as vehicle control). Tumors were harvested at 3 weeks, and tumor sizes were compared, quantitation at right. G, immunohistostaining results in tumor tissues. Tumor tissues were harvested, and AR, Oct4, and CK8 expression was investigated. Quantitation shows percentage of the positively stained cell numbers. Significance was defined as p Ͻ 0.001 (***), p Ͻ 0.01 (**), and p Ͻ 0.05 (*), unless marked in figure, with Student's t test, quantitation in lower panel. Error bars, S.D.
Together, results from Fig. 6, E-G, demonstrated that targeting methylated AR promoter in PCa stem/progenitor cells with 5-AZA could lead to (a) suppression of prostate tumorigenesis and (b) promotion of stem cells into a higher differentiation stage.

DISCUSSION
This study revealed differential AR expression in three types of stem/progenitor cells, as suggested by the different methylation patterns of CpG islands on the AR promoter. We also found that the low expression of AR in normal prostate/PCa stem/progenitor cells is important in maintaining their stemness, protecting them from differentiation, and accelerating their self-renewal/proliferation.
A question arises as to why normal prostate/PCa stem/progenitor cells express little AR, whereas stem/progenitor cells from other tissues, including the earlier developmental stage of embryonic stem cells, all expressed higher levels of AR. It is possible that this unique epigenetic regulation of AR methylation in normal prostate/PCa stem/progenitor cells is essential in maintaining their stemness, and the induction of AR by demethylation may then drive stem/progenitor cells into the differentiation axis. In contrast, maybe molecules other than AR play key roles to maintain stemness or drive differentiation in other types of stem/progenitor cells. For example, Dlk1/Pref-1 was reported to be critical (52) in differentiation of the mesenchymal stem cells, and Dax1 and SF-1 are suggested to be critical in embryogenesis (53) and steroidogenesis (54), respectively, for the differentiation of embryonic stem cells. But why different key factors are required in different stem/progenitor cells to maintain their stemness versus driving into differentiation remains unclear. Nevertheless, the finding that the low expression of AR in normal prostate/PCa stem/progenitor cells indeed played a key role in driving stem/progenitor cells (CD133 ϩ , integrin ϩ , CK5 ϩ , and AR Ϫ/low ) into transit-amplifying cells/intermediate cells (CD133 Ϫ , integrin ϩ , CK5 ϩ , CK8 ϩ , and AR low ) and finally into well differentiated epithelial luminal cells (CD133 Ϫ , integrin Ϫ , CK5 Ϫ , CK8 ϩ , and AR high ) (47, 55, 56) (supplemental Fig. 1) may provide potential targets to alter the normal prostate/PCa differentiation axis.
The in vitro and in vivo experimental results demonstrating the decreased self-renewal and tumorigenicity of PCa stem/ progenitor cells upon induction of AR expression further indicate that this finding can be applied to develop a therapeutic approach to treat PCa. Indeed, targeting stem/progenitor cells has emerged as a novel potential approach to battle PCa (25,57,58). We believe that using 5-AZA to induce AR expression in PCa stem/progenitor cells as developed here has several advantages over other approaches based on several reasons. (a) The strategy of using 5-AZA to induce AR expression in PCa stem/ progenitor cells to suppress effectively their self-renewal/proliferation without affecting AR-mediated proliferation in the PCa non-stem/progenitor cells provides a stem/progenitor cell-specific therapy. (b) This approach yields few side effects compared to the toxicity raised via other approaches to target stem/progenitor cells (59,60). (c) Importantly, this approach can be used to combine the classic androgen deprivation therapy that leads to suppression of most of non-stem/progenitor cells yet promotes stem/progenitor cells (61), which would simultaneously suppress stem/progenitor and non-stem/progenitor cells to battle PCa at stages before and after castration resistance.
In summary, we demonstrated the unique methylation pattern of the AR promoter in PCa stem/progenitor cells and its influence on their renewal/proliferation and differentiation, and this finding could be applied to target PCa stem/progenitor cells by altering AR expression levels via demethylation. This strategy will be of significance in providing a new potential therapeutic approach to battle PCa in the future.
5-AZA has been used in preclinical and clinical trials with reduced hormonal response and modest effect in alleviation of symptoms observed (2,62). The 5-AZA effect was interpreted as a result of induction of tumor suppressor gene expression (62). It was also hypothesized that the AR expression may be down-regulated by this treatment; thereby, hormonal response was reduced (62). We emphasize here the use of 5-AZA to target PCa stem cells by inducing AR expression in PCa stem cells. This is an unrecognized concept for a therapeutic approach using 5-AZA to battle PCa.