Constant Degradation of the Androgen Receptor by MDM2 Conserves Prostate Cancer Stem Cell Integrity

While the androgen receptor (AR) is a master regulator of prostate development and disease, it is commonly believed that normal prostate stem cells and prostate cancer stem cells (CSC) express no/low AR and that their growth is androgen-independent (1). Our previous study demonstrated that AR⁻ pluripotent CSCs isolated from patient biopsies differentiated into prostatic glandular structures containing all three epithelial cell types including AR⁺ luminal secretory cells as well as AR⁻ basal and neuroendocrine cells when engrafted with embryonic mesenchyme under the renal capsule (2). Similarly, other studies report that AR⁻ normal prostate stem-like cells differentiate into three prostatic epithelial cell lineages (3). This pattern of AR⁻ expression mimics that observed in human and rodent prostate development where AR⁻ epithelial anlagen grow into the urogenital mesenchyme until AR protein is induced by, as yet unknown mechanisms, to initiate glandular lumen formation, epithelial cell specification, and androgen-mediated secretory activity (4, 5). Qin and colleagues determined that prostate specific antigen (PSA)^−/lo cells, a subpopulation isolated from LNCaP and LAPC9 prostate cancer cell lines, were AR^−/lo and exhibited stem-like properties including self-renewal and the ability to regenerate PSA⁺ cells (6). PSA^−/lo LAPC9 cells developed into therapy-resistant tumors (6).

Together, these observations imply that an AR⁻ phenotype is essential for maintaining CSC and normal prostate stem cell homeostasis and for promoting castration-resistant prostate cancer (CRPC). The mechanisms underlying this AR⁻ phenotype are unknown. Our previous study showed that biopsy-derived prostate cancer CSCs, referred to as HPET (human prostate epithelial cells expressing hTERT), expressed AR mRNA but not AR protein, suggesting that expression was regulated at the posttranscriptional level (2). The Qin and colleagues' study reported that in PSA^−/lo cells, both AR mRNA and protein were downregulated, implying that AR expression was regulated at the transcriptional level (6). Whether inhibition of AR protein expression occurs at the transcriptional and/or posttranscriptional level remains to be established.

Evidence from non-CSC, AR⁺ prostate cancer cell lines suggests that the ubiquitin–proteasome system (UPS) modulates the steady state of AR expression. For example in AR⁺ LNCaP, CWR-R1, and CWR22Rv1 cells, the E3 ligase MDM2 (mouse double minute 2 homolog) transiently modulates AR stability during transcriptional activity (7). Other E3 ligases, including NEDD4 (neural precursor cell expressed developmentally downregulated protein 4; ref. 8), CHIP (C-terminus of Hsp70-interacting protein; ref. 9), and SKP2 (S-Phase Kinase-Associated Protein 2; ref. 10) also regulate AR protein degradation in LNCaP, C4-2B, and CWR22Rv1 cells. Whether AR⁻ CSCs use these same ligases to block total AR expression remains to be explored.

The recent discovery of AR splice variants (AR-Vs) provides insight into the mechanisms promoting emergence of CRPC. These naturally occurring AR-Vs are identified in clinical prostate cancer biopsy specimens and non-CSC prostate cancer cell lines (reviewed in refs. 11, 12). AR-Vs contain the N-terminal and DNA-binding domains; however, they lack a ligand-binding domain, resulting in constitutive activation. Several reports demonstrate that AR-Vs are highly expressed in CRPC, metastasis, and prostate cancer cell lines not requiring androgens for cell growth. The most commonly expressed variant AR-V7 (a.k.a. AR3) is associated with the development of CRPC and drug resistance. Furthermore, AR-V7 promotes epithelial–mesenchymal transition (EMT) and induces expression of signature stem cell genes, including NANOG in LNCaP cells and LIN28B in DU-145 cells (13). Interestingly, MDM2 induces AR-V7 ubiquitination and protein degradation (14). Whether CSCs express AR-Vs remains to be investigated.

Prostate cells with CSC-like properties have typically been isolated as side populations (∼0.1%–0.3% of prostate cancer cells) from established cell lines, for example, PC-3 (15), DU-145 (16), and LNCaP (17) cells, and from biopsy tissues (18). Here we use two CSC-like cell models to investigate the AR⁻ signature of prostate CSCs. Both HPET (2) and HuSLC (human stem like cells; previously termed HPE for human prostate epithelial; ref. 19) cell lines were isolated from high-grade Gleason 9 biopsy tissues from two unrelated individuals. Of note is that the HuSLC line arose spontaneously. Both cell lines express AR mRNA but not AR protein, and exhibit stem-like properties including pluripotency in vivo, differentiating into the three prostatic epithelial cell lineages, that is, luminal secretory AR⁺ cells, basal cells, and neuroendocrine cells (2, 19). Using these CSC-like cell models, we report that MDM2 is critical for conserving an AR⁻ signature and promoting stemness, while loss of MDM2 induces full-length AR expression, terminal CSC differentiation into AR⁺ luminal epithelial cells, and cell death.

Prostate cancer cells LNCaP, PC3, 22RV1, and VCaP were purchased from ATCC. Proteasome inhibitors MG132 (catalog no.: 3175-v), MG115 (catalog no.: 3170-V), and epoxomicin (catalog no.: 4381-v) were purchased from the Peptide Institute (Osaka, Japan). Matrigel (BD Biosciences; catalog no.: 10828028) was batch-tested by the Pluripotent Stem Cell Facility at Cincinnati Children's Hospital Medical Center (Cincinnati, OH). The human full-length wild-type (wt) AR expression vector pSVAR_o was a gift from Dr. Shutsung Liao, The Ben May Department for Cancer Research, University of Chicago, Chicago, IL. Additional materials and reagents are indicated below.

The parental HPET cell line was established by transducing primary human prostate epithelial cells (passage 2) cultured from deidentified human prostate cancer surgical waste material (Gleason 9, undifferentiated prostate cancer,) using pLenti-particles expressing the hTERT-EGFP gene (2). The HPET cell lines and their prostate epithelial and stem cell characteristics were authenticated both in vitro and in vivo as described in detail by Gu and colleagues (2). Vials containing HPET passage 73 were thawed and passaged approximately five times to complete this current study. A second cell line, termed HuSLC (Human Stem Cell-Like Cell) line arose spontaneously from human prostate epithelial (HPE) cells cultured from deidentified biopsy tissue (Gleason 9, undifferentiated prostate cancer) from an unrelated male donor. Their HPE-like and prostate cancer–generating characteristics were authenticated both in vitro and in vivo in detail in Williams and colleagues (Fig. 6; ref. 19). They were initially named “HPE,” but subsequently renamed “HuSLC” because of their stem-like characteristics. Vials containing HuSLC passage 32 and 34 were thawed and passaged approximately 5–6 times to complete this study. Cell lines were recently tested (June 2018) and found to be negative for Mycoplasma (Lonza MycoAlert Mycoplasma Detection Kit, catalog no.: LT07-218; Lonza Mycoalert Mycoplasma Assay Control Set, catalog no.: LT07-518).

Both the HPET and HuSLCs cell lines were cultured in under embryonic stem (ES) cell conditions using defined ESC medium, DMEM-F12 (Thermo Fisher Scientific, catalog no.: 11320033) supplemented with KnockOut Serum Replacement (Thermo Fisher Scientific, catalog no.: 10828028) and 4 ng/mL recombinant bFGF (ProSpec, catalog no.: CYT-218) on Matrigel-coated plates. Prostate cancer cells LNCaP, PC3, 22RV1, and VCaP were purchased from ATCC and cultured as recommended by ATCC. Cells from the company were expanded through approximately 4–5 passages and frozen down as stock vials. Stock vials were thawed and cells were passaged approximately 2–4 times to complete this study.

HPET or HuSLC cells were trypsinized using Trypsin-EDTA Solution (Thermo Fisher Scientific, catalog no.: 25200056), centrifuged for 5 minutes at 300 × g and the cell pellet resuspended in defined ESC medium (described above). Cells were seeded at 2,500 cells/mL in 6-well ultra-low attachment plates (Thermo Fisher Scientific, Corning Costar 3471, catalog no.: 07-200-601) and cultured for 10 days. Wells were photographed using phase contrast and total number of spheres/well were counted using NIH ImageJ software.

Cells were plated at 5 × 10³ cells/well using 24-well plates and either transfected with control plasmid or pSVAR_o to induce exogenous AR protein expression or treated with MG132 to induce endogenous AR, and treated with/without 10⁻⁸ mol/L DHT with/without 10⁻⁵ mol/L OHF or vehicle control (95% ethanol) as indicated in each assay. Cell numbers were determined using the Trypan Blue Viability assay (Thermo Scientific, HyClone Trypan Blue Solution, 0.4%, catalog no. SV3008401).

Cells were harvested using RIPA buffer (Invitrogen Inc., catalog no.: R0278) with 1% protease inhibitor cocktail (Thermo Fisher Scientific, cOmplete Protease Inhibitor Cocktail tablets/Roche, catalog no.: NC0939492) and phosphatase inhibitor cocktail (Thermo Fisher Scientific, EMD Millipore Calbiochem Phosphatase Inhibitor Cocktail Set I, catalog no.: 53-913-110VL). Lysates were centrifuged at 4°C for 10 minutes at 14,000 × g; 50 μg protein from each supernatant was subjected to 10% SDS-PAGE, transferred to polyvinylidene difluoride membrane and blocked and probed with primary antibody overnight at 4°C. Peroxidase-conjugated secondary antibody was added at a 1:4,500 dilution and blots were developed using the Enhanced Chemiluminescence (ECL) kit (Pierce/ThermoFisher Scientific, catalog no.: 32132). Antibodies used in this study are listed in Supplementary Table S1.

Cells were lysed in 300 μL of cold lysis buffer [1% Triton X-100, 150 mmol/L NaCl, 10 mmol/L NaHPO₄ (pH 7.2), 5 mmol/L NaF, 2 mmol/L EDTA, 1× HALT protease inhibitor cocktail (Thermo Fisher Scientific, catalog no.: 78440)] at 4°C in a cold room. Lysates were cleared by centrifugation and immunoprecipitation was performed by incubating 500 μg total protein with 10 μg rabbit anti-AR antibody or 5 μg mouse anti-hemagglutinin (HA) antibody (Santa Cruz Biotechnology, catalog no. sc-816 and catalog no. sc-805, respectively) overnight at 4°C. ImmunoPure Immobilized Protein G beads (Pierce Biotechnology, catalog no.: 44667) were used to pull down protein complexes. The immunoprecipitates were washed in 1× PBS, resuspended in 1× Laemmli buffer, and subjected to Western blot analysis.

Cells were transfected with the ARR₂PB-luc reporter (20) and Renilla luciferase vectors (Promega, catalog no.: E2231) and either treated with increasing concentrations of MG132 (to induce endogenous AR) or cotransfected with increasing concentrations of pSVAR_o (to induce exogenous AR). Twenty-four hours after AR induction, cells were treated with vehicle control (ethanol) or DHT (10⁻⁸ mol/L) with/without OHF (10⁻⁵ mol/L) for 24 hours, as described previously (21). Cells were lysed and luciferase activity was determined using the Promega Dual-Luciferase Reporter Assay System kit and protocol (Promega, catalog no.: E1910) according to the manufacturer's protocol. Cell lysate protein concentrations were determined using the Protein BCA Assay kit (Pierce/ThermoFisher Scientific, catalog no.: 23225).

Total RNA was extracted from HPET and HuSLCs using TRIzol reagent (Invitrogen/Thermo Fisher Scientific, catalog no.: 15596018) following the manufacturer's protocol. Total RNA concentrations (260/280 nm) were determined using the NanoDrop system (NanoDrop Technologies Inc.). RNAs were treated with DNase I (Invitrogen Inc., catalog no.: AM2222) to remove any traces of DNA contamination and cDNAs were synthesized from 1 μg of RNA per sample using the Fermentas Revertaid Kit (Fermentas/Thermo Fisher Scientific, catalog no.: K1621), according to the manufacturers' protocols.

Primers used in this study are listed in Supplementary Tables S2 and S3. One (1) μg of synthesized cDNA was added to 1 μmol/L random-specific primers (synthesized by IDT Inc.), and 12.5 μL of 2× Power SYBR Green PCR Master Mix (Applied Biosystems/Thermo Fisher Scientific, catalog no.: 4309155) to a final volume of 25 μL. PCR amplification was performed using an Applied Biosystems 7300 Real-Time PCR System (one cycle at 50°C for 2 minutes, one cycle of 95°C for 10 minutes, followed by 40 cycles of 15 seconds at 95°C and 1 minutes at 60°C). The dissociation curve was completed with one cycle of 1 minute at 95°C, 30 seconds of 55°C, and 30 seconds of 95°C. Non-reverse transcription control and no template control were included in the PCR program for quality control. RNA expression for the genes of interest were normalized to expression of the GAPDH gene and analyzed using the ΔΔC_t method (22).

GraphPad Prism v4.0 was used for all statistical analyses. Statistical parameters, including the types of tests, number of samples (n), descriptive statistics, and P values are reported in the figure legends.

Previously, we reported that HPET cells recapitulated the AR⁻ phenotype reported for CSCs (Fig. 1A; ref. 2) and similarly, the HuSLC line also expressed AR mRNA but not AR protein (Fig. 1B). To determine whether AR protein was constitutively being degraded, HPET and HuSLCs were transfected with increasing concentrations of pSVAR_o, which expresses human full-length/wt AR at low concentrations (≤4 μg; ref. 23). Unexpectedly, 30 μg pSVAR_o was required for detectable AR protein expression in HPETs (Fig. 1A; P < 0.0001). Moreover, addition of dihydrotestosterone (DHT, 10⁻⁸ mol/L) appeared to stabilize AR protein levels. Similarly, 40 μg pSVAR_o was required to induce detectable AR protein in HuSLCs (Fig. 1B; P < 0.0001), suggesting that at lower pSVAR_o concentrations, AR protein was actively being degraded.

To determine whether AR protein levels were downregulated by proteasomal degradation, HPET cells (Fig. 1C) and HuSLCs (Fig. 1D) were treated with increasing concentrations of the proteasome inhibitor MG132. In HPET cells, endogenous full-length AR protein was already detected at the lowest concentration of MG132 (1 μmol/L) tested. Similarly, AR protein was induced in HuSLCs following MG132 treatment (20 μmol/L). Furthermore, treatment with other proteasome inhibitors, MG115 and Epoxomicine, confirmed these observations (Supplementary Fig. S1). In transfection assays using the androgen-regulated probasin promoter linked to the luciferase reporter gene (ARR₂PB-luc; ref. 24), AR-mediated transcription was only induced when exogenous (Fig. 1A and B) or endogenous (Fig. 1C and D) AR protein was expressed.

The E3 ligase MDM2 is reported to regulate AR levels in prostate cancer cells (7). Therefore, MDM2 expression was analyzed in HuSLC and HPET cells and compared with two standard prostate cancer cell lines, LNCaP (where MDM2 modulates AR levels in a temporal manner to regulate AR-mediated transcription) and DU-145 (which do not express AR). Both HuSLC and HPET cells expressed higher levels of MDM2 (2.16- and 1.74-fold respectively) as compared with LNCaP and DU-145 cells (Supplementary Fig. S2). Furthermore, MDM2 levels were highest in HuSLCs, implying that AR degradation in HuSLCs was greater than in HPET cells. This is supported by the findings that more AR plasmid and higher amounts of proteasome inhibitor were required to induce HuSLC AR expression. Taken together, these observations support an active role for the proteasome in conserving the AR⁻ CSC signature.

Published primer sets spanning the genomic region from which AR-Vs are transcribed were used to determine whether HPET cells and HuSLCs expressed AR-V transcripts (25, 26). A universal forward primer, P1/P2/P3 (F), located in AR exon 2 was paired with one of three reverse primers (P1R, P2R, and P3R) located within in each variant exon (as outlined in Supplementary Table S2). This approach provided coverage for the known AR variants. 22Rv1 and VCaP cell lines served as positive controls.

Both HPET cells and HuSLCs expressed full-length AR transcript (AR-fl; Fig. 2A); however, they did not express any of the other recognized AR-V transcripts (Fig. 2; Supplementary Fig. S3). We then determined whether androgen-mediated activity was required for induction of AR-V expression. AR-V1 and AR-V7 transcripts were not detected under any conditions tested (Fig. 2B and C). In contrast, full-length AR mRNA was expressed at similar levels regardless of treatment with vehicle or dihydrotestosterone (DHT) with/without hydroxyflutamide (OHF; Fig. 2A). Western blot analysis using an AR antibody toward the N-terminal (which recognizes AR-fl and AR-Vs) determined that HPET cells and HuSLCs expressed full-length 110 kDa endogenous AR following MG132 treatment; however an 80-kDa band corresponding to AR-Vs (27) was not detected in either cell line under any condition tested (Supplementary Fig. S4). Several bands at lower molecular weights (<68 kDa) were observed in HPET cells; however, they did not correspond to any known AR-Vs. Because these bands were already present prior to induction of AR expression, they may represent degraded/nonfunctional AR or nonspecific antibody interactions. In summary, CSC-like HPET cells and HuSLCs did not express AR-Vs, but instead, conserved expression of full-length AR.

A standard sphere formation assay was performed to determine whether AR signaling inhibited CSC stemness. HPET cells (Fig. 3A) and HuSLCs (Fig. 3B) were treated with MG132 with/without DHT to induce AR protein expression and activate AR-mediated signaling. Sphere formation was absent upon AR expression, and remained unaltered by DHT treatment. In contrast, sphere formation was rescued by addition of OHF. Similarly, exogenous expression of pSVAR_o inhibited sphere formation; and this could be rescued by OHF treatment (Fig. 3C and D). Furthermore, induction of AR dramatically decreased cell proliferation within 48 hours, and addition of DHT decreased cell proliferation even further down to baseline levels. In contrast, OHF-mediated inhibition of AR activity restored cell proliferation (Fig. 3C and D).

Collectively, these observations infer that an AR⁻ phenotype is essential for prostate CSC self-renewal and proliferation. Moreover, treatment with OHF alone is sufficient to significantly promote sphere formation. Therefore, absence of AR protein and the potential of antiandrogens to exert as yet undiscovered AR-independent effects on stimulating prostate CSC self-renewal and proliferation could facilitate the emergence of therapeutic resistance.

HPET and HuSLCs express numerous stem/progenitor cell markers, including the transcription factors OCT4, NANOG, and SOX2, which regulate pluripotency and self-renewal in human and mouse embryonic stem cells (2). They also express the progenitor cell markers Nestin (NES) and CD44 (2). To determine whether AR restricted their expression, cells were treated with MG132. As seen in Fig. 4A, induction of endogenous AR decreased OCT4, NANOG, SOX2, NES, and CD44 expression following DHT treatment (P < 0.05). In contrast, inhibition of AR signaling by OHF restored their expression. Transfection of HPET and HuSLCs with pSVAR_o confirmed these observations (Fig. 5A). Moreover, in both MG132-treated and pSVAR_o-transfected cells, OHF alone could upregulate these factors with OCT4, NANOG, and NES increasing >2-fold (P < 0.05) as compared with the vehicle control group, suggesting that antiandrogens might exert, as yet, unknown AR-independent actives that support CSC expansion.

These observations provide evidence that CSC-like cells lose stemness characteristics upon expression of AR. Because prostatic glandular epithelium consists of luminal secretory, basal, and neuroendocrine epithelial cells (2), we questioned whether loss of these factors would initiate epithelial cell specification. Indeed, androgen-regulated genes associated with luminal epithelial cell fate, including PSA and PAP (P < 0.001) as well as NKX3.1, TMPRSS2, and FGF5 (P < 0.05) were induced upon activation of endogenous AR (Fig. 4B) or pSVAR_o (Fig. 5B) following treatment with DHT. Furthermore, expression of these luminal cell markers was inhibited by treatment with OHF. Similarly, the AR target genes HES1 and HEY1 were also induced following induction of AR expression (Supplementary Fig. S5). In contrast, neither endogenous AR (Fig. 4C) nor pSVAR_o (Fig. 5C) induced expression of TP63, a common basal epithelial cell marker, or Chromogranin A (CHGA), typically expressed by prostate neuroendocrine cells, under any condition tested. Thus, AR appears to selectively promote luminal secretory cell fate.

Several studies report that the UPS modulates transcription factor levels to regulate stem cell and CSC maintenance and differentiation (28–30). To determine whether poly-ubiquitination regulates AR protein levels, HPET cells were transfected with 30 μg pSVAR_o and increasing concentrations of the wild-type ubiquitin expression vector pRK5-HA-Ubiquitin-WT (HA-UbWt; Fig. 6A,i). Exogenous AR was degraded in a dose-dependent manner with complete degradation occurring at 50 μg HA-UbWt. In a similar manner, 20 μg HA-UbWt was capable of degrading endogenous AR protein induced with MG132 treatment (Fig. 6A, ii). Cells were then transfected with a mutant ubiquitin plasmid pRK5-HA-Ubiquitin-KO (HA-UbKO) that is incapable of adding ubiquitin molecules onto its target protein to determine whether inhibition of AR poly-ubiquitination prevented AR degradation. Endogenous AR protein was strongly expressed after transfection with HA-UbKO (Fig. 6A, iii). Collectively, these observations imply that the dynamic turnover of AR protein in prostate CSC-like cells is regulated by poly-ubiquitination.

The final step in the ubiquitination cascade is carried out by E3 ligases (31). Several E3 ligases, including MDM2 and NEDD4, are reported to regulate AR and/or AR-V protein levels in non-CSCs, AR⁺ prostate cancer cells, which comprise the bulk of prostate tumor cells and in AR⁺ prostate cancer cell lines derived from metastatic lesions. Whether MDM2 and/or NEDD4 degrade AR in prostate CSCs is unknown. Therefore, HPET cells were transfected with pSVAR_o and increasing concentrations of shMDM2 plasmid to determine the level of MDM2 knock-down and whether AR protein would be induced. As shown in Fig. 6B, MDM2 levels decreased in parallel with increasing concentrations of shMDM2; and AR expression was greatest when MDM2 protein was barely detectable. AR-mediated transcription was confirmed using the ARR₂PB-luc assay (Fig. 6C). In contrast, NEDD4 knockdown to undetectable levels did not induce AR protein, implying that it did not play a role in modulating AR levels in CSC-like cells.

Other E3 ligases not currently known to regulate AR protein include MARCHVII, associated with adult stem cells (32), and WWP2, reported to regulate embryonic stem cell factors, for example, OCT4 (33) and SOX2 (34). Transfecting HPET cells with shMARCHVII and shWWP2 in a dose-dependent manner determined that loss of MARCHVII and WWP2 expression did not induce AR protein expression under any concentrations of shRNA tested (Fig. 6B). Thus, MDM2 appears to selectively degrade AR in CSCs.

Immunoprecipitation (IP) analysis was performed to determine whether MDM2 directly binds AR (Fig. 6D). When AR⁻ HPET cells were transfected with HA-UbWt alone (lane 4), AR was not observed in the IP:AR fraction and MDM2 was absent in the IP:HA fraction, suggesting that AR expression was required for the formation of an AR/HA-UbWt/MDM2 binding complex. Once endogenous AR was expressed following MG132 treatment, both AR and MDM2 were observed following IP (lane 5), indicating that AR was necessary for AR/HA-UbWt/MDM2 complex formation. Similarly, an AR/HA-UbWt/MDM2 complex was observed in cells where expression of mutant HA-UbKO inhibited the degradation of endogenous AR (lane 6), confirming that AR could form a complex with ubiquitin and MDM2.

To determine the effects of MDM2 knockdown on CSC stemness, HPET cells were transfected with shMDM2 to prevent AR degradation. Knockdown of MDM2 alone was sufficient to abolish sphere formation and treatment with DHT did not alter these effects (Fig. 7A and B). Again, sphere formation was rescued by addition of OHF. In addition, MDM2 knockdown alone decreased cell proliferation (P < 0.05), addition of DHT inhibited cell proliferation even further, and proliferation was rescued with OHF treatment (Fig. 7C). Furthermore, MDM2 knockdown decreased OCT4 and potentially NANOG expression, and treatment with OHF alone increased their expression significantly (Fig. 7D). In parallel, expression of luminal epithelial cell–specific genes, PSA, PAP, NKX3.1, and TMPRSS2, increased with DHT treatment and decreased to basal levels with addition of OHF, while FGF5 expression was regulated in a similar manner, it did not reach statistical significance (Fig. 7E).

These observations primarily recapitulated those observed in HPET and HuSLCs expressing AR following treatment with MG132 or transfection with pSVAR_o. However, one difference was that not all of the stem cell factors decreased in response to DHT treatment (Fig. 7D), suggesting that MDM2 may exert specificity in maintaining the steady-state expression of select stem/progenitor cell proteins. Whether MDM2-mediated degradation of AR involves p53 is unclear. HPET cells do not express p53 while HuSLCs express p53, suggesting that p53 activity is not essential for degrading AR (Supplementary Fig. S6).

Mechanisms that prevent AR expression in normal prostate stem cells and prostate CSCs remain largely unknown. Our study provides the first evidence that in stem-like AR⁻ CSCs isolated from prostate cancer biopsies, MDM2 promotes the constant degradation of AR protein, thereby maintaining prostate CSC pluripotency and inhibiting epithelial cell lineage specification (summarized in Fig. 7F). The AR⁻ signature also facilitates CSC proliferation and expansion, while induction of AR via MDM2 downregulation selectively induces a luminal epithelial cell phenotype and loss of cell growth.

Most studies on prostate CSCs and AR have been performed in side-fractions of CSC-like cells isolated from LNCaP, LNCaP derivative, and LAPC9 cell lines (35, 36). Both AR mRNA and AR protein are downregulated In LNCaP and LAPC9-derived PSA^−/lo cells (6). In the CAstration-Resistant NKX3.1-expressing cells (CARN) mouse model, CARNs expressed AR; however, genetic deletion of AR did not alter their luminal progenitor/stem cell properties (37). Only the rate of proliferation during prostate regeneration was reduced (37). Taken together, these observations imply that AR is not required for prostate CSC or normal prostate progenitor/stem cell function.

Our study supports the role of MDM2 in blocking AR protein expression and proposes that MDM2 exerts fundamentally different functions in AR⁻ prostate CSCs as compared with non-CSC, AR⁺ prostate cancer cells that comprise the bulk of prostate tumor cells (summarized in Supplementary Table S4). In AR⁻ prostate CSCs, MDM2 continuously degrades AR to maintain an AR⁻ phenotype, self-renewal, and proliferative potential. MDM2 is also reported to promote stemness properties in other tissue-derived stem cells, for example, in generating induced pluripotent stem cells from p53-deficient murine embryonic fibroblasts (MEF; ref. 38) and suppressing differentiation of human mesenchymal stem cells into osteoblasts; whereas, MDM2 knockdown increases osteoblast differentiation (38). In a similar manner, MDM2 knockdown in HPET and HuSLCs induced terminal differentiation to a luminal epithelial phenotype. However, in non-CSC AR⁺ prostate cancer cells, it is well-documented that MDM2 temporally modulates AR protein levels to attenuate AR-mediated transcription during normal cellular function and to regulate cell-cycle progression while retaining basal levels of AR expression (7). Inhibition of MDM2 expression in LNCaP and androgen-resistant LNCaP (LNCaP-Res) cell lines downregulates AR protein levels and decreases AR activity; however, total AR expression is not lost during this process (39). Thus, the mechanism regulating MDM2-mediated knockdown of HPET and HuSLC AR protein to undetectable levels remains to be elucidated.

Therapeutic resistance remains a persistent challenge in the treatment of prostate cancer. Consequently, targeting the AR and androgens is still central to the management of advanced prostate cancer (reviewed in ref. 40). The recent discovery that prostate cancer cells synthesize steroids de novo has resulted in considerable interest in drugs that inhibit androgen synthesis (41). Regrettably, because CSCs do not appear to require AR, none of the second-generation ADT and antagonist drugs, for example, abiraterone and enzalutamide, would theoretically eliminate CSCs from the prostate cancer cell pool. Moreover, our study suggests that antiandrogen treatment alone paradoxically increases CSC self-renewal and cell proliferation. If indeed, induction of AR expression causes CSC differentiation into AR⁺ luminal cells, then this could potentially resensitize CSCs to ADT and eliminate them along with the bulk of responsive AR⁺ prostate cancer cells. Induction of AR could also promote terminal differentiation and eliminate CSCs through this mechanism.

The emergence of therapeutic resistance is also attributed to production of AR-Vs (11, 12, 42). AR-Vs are expressed in clinical prostate cancer biopsy specimens and prostate cancer cell lines (13, 26) and their expression is upregulated in metastatic and treatment-resistant disease (11, 43). One of the most commonly expressed variants is AR-V7. It is considered a valid therapeutic target in the treatment of CRPC; however, the mechanisms by which AR-V7 drives CRPC progression remains to be elucidated (42). Our study observed that biopsy-derived CSCs only expressed AR-fl, but not AR-V, in response to blocking MDM2 activity. Further studies are required to determine a putative role of AR-Vs in prostate CSCs. In addition, HPET cells do not express p53 while HuSLCs express p53, yet AR is continuously degraded and self-renewal and proliferation are maintained in both CSC-like cell lines, suggesting that p53 activity is not essential for these processes. Thus, therapeutic approaches that target MDM2–p53 interactions would likely be ineffective in inhibiting prostate CSC growth. Other reported p53-independent MDM2 activities include the promotion of cancer progression through EMT in AR⁻ DU-145 cells that express a mutant/nonfunctional p53 (44), and clonogenic survival in MCF7, SJSA, and Panc1 cell lines (38). This is in contrast to AR⁺p53⁺ cells where targeting MDM2/p53 interactions inhibit tumor cell growth. The small-molecule inhibitor MI-219 selectively disrupts MDM2/p53 interactions, thereby activating p53 signaling and inducing apoptosis in LNCaP cells in vitro and inhibiting LNCaP xenograft growth in vivo (45). Similarly, the MDM2 inhibitor Nutlin-3 activates p53 and inhibits the growth of SJSA-1 osteosarcoma xenografts by 90% (46). In CRPC, lncRNA (AR-repressed long noncoding RNA) is upregulated and binds AR protein to impair AR–MDM2 interactions. Consequently, AR is not ubiquitinated and degraded, resulting in the upregulation of AR transcriptional activity and increased CRPC cell growth (47).

Collectively, these studies infer that selectively targeting MDM2 activity, and not MDM2–p53 or AR-p53 interactions, could potentially eliminate CSCs more effectively than second-generation ADT or antagonist drugs. Our study suggests that MDM2 conserves the AR⁻ CSC signature and that this may be a critical step toward stimulating CSC expansion during the emergence of therapeutic resistance. Furthermore, treatments that promote antiandrogenic activities may signal CSCs to initiate proliferation and expansion (48). Thus, selectively blocking MDM2 expression and/or MDM2-mediated activity in combination with AR/androgen-targeted treatments may offer a novel strategy for eliminating AR⁻ CSCs as well as the bulk of AR⁺ prostate cancer cells to decrease tumor burden and metastasis, and/or inhibit the emergence of therapeutic resistant disease.

No potential conflicts of interest were disclosed.

Conception and design: P. Vummidi Giridhar, S. Kasper

Development of methodology: P. Vummidi Giridhar, K. Williams, S. Kasper

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Vummidi Giridhar, K. Williams, A.P. VonHandorf, P.L. Deford, S. Kasper

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Vummidi Giridhar, K. Williams, A.P. VonHandorf, S. Kasper

Writing, review, and/or revision of the manuscript: P. Vummidi Giridhar, K. Williams, A.P. VonHandorf, P.L. Deford, S. Kasper

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Williams, A.P. VonHandorf, P.L. Deford, S. Kasper

Study supervision: S. Kasper

The authors thank Katherine A. Burns, PhD for her invaluable assistance in the preparation and proofreading of the manuscript. This work was funded by the National Institute of Diabetes & Digestive & Kidney Diseases (R01 DK60957 to S. Kasper) and the United States Department of Defense (W81XWH-08-1-0662 to S. Kasper).

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.