Galbanic acid decreases androgen receptor abundance and signaling and induces G₁ arrest in prostate cancer cells

Abstract

Androgen receptor (AR) signaling is crucial for the genesis and progression of prostate cancer (PCa). We compared the growth responses of AR(+) LNCaP and LNCaP C4-2 vs. AR(−) DU145 and PC-3 PCa cell lines to galbanic acid (GBA) isolated from the resin of medicinal herb Ferula assafoetida and assessed their connection to AR signaling and cell cycle regulatory pathways. Our results showed that GBA preferentially suppressed AR(+) PCa cell growth than AR(−) PCa cells. GBA induced a caspase-mediated apoptosis that was attenuated by a general caspase inhibitor. Subapoptotic GBA down-regulated AR protein in LNCaP cells primarily through promoting its proteasomal degradation, and inhibited AR-dependent transcription without affecting AR nuclear translocation. Whereas docking simulations predicted binding of GBA to the AR ligand binding domain with similarities and differences with the AR antagonist drug bicalutamide, LNCaP cell culture assays did not detect agonist activity of GBA. GBA and bicalutamide exerted greater than additive inhibitory effect on cell growth when used together. Subapoptotic GBA induced G₁ arrest associated with an inhibition of cyclin/CDK4/6 pathway, especially cyclin D₁ without the causal involvement of CDK inhibitory proteins P21^Cip1 and P27^Kip1. In summary, the novelty of GBA as an anti-AR compound resides in the distinction between GBA and bicalutamide with respect to AR protein turnover and a lack of agonist effect. Our observations of anti-AR and cell cycle arrest actions plus the anti-angiogenesis effect reported elsewhere suggest GBA as a multi-targeting drug candidate for the prevention and therapy of PCa.

INTRODUCTION

Prostate cancer (PCa) is the second leading cause of cancer death in American men and is responsible for an estimated 30,000 deaths per year¹. Strong evidence suggests that androgen receptor (AR) plays a crucial role in the genesis and progression of PCa^2–4. Androgen deprivation therapies result in a remission of the primary cancerous lesions, but they will invariably relapse and progress to the castration-resistant prostate cancer (CRPCa), which is rather insensitive to current chemotherapeutic drugs⁵. The majority of the recurring PCa cells expresses AR and still depends on AR signaling for proliferation and survival in presence of castrate level of androgen. Laboratory studies demonstrate that increased AR is necessary and sufficient to promote the progression of PCa from androgen-dependent stage to CRPCa^{6, 7}. Recent clinical trials of Abiraterone Acetate (a CYP17 blocker to inhibit the synthesis of androgenic steroids) and MDV-3100 (an AR antagonist to inhibit ligand-binding activation of AR) have shown that they can induce a dramatic decline of serum level of prostate specific antigen (PSA), an AR transcriptional target, and/or radiologically detectable regression of tumors in patients who failed currently available therapeutic agents including docetaxel^{8, 9}. Overall, these suggest that targeting AR signaling remains an important and feasible approach for the therapy and prevention of PCa.

Ferula assafoetida is a perennial herb widely distributed throughout the Mediterranean area and Central Asia. Its resin has been used in traditional herbal medicine as antiseptic, antifungal, antibiotic, antioxidant, anti-carcinogenic, anti-inflammatory, anti-thrombotic, anti-hepatotoxic or laxative agents in Asian countries for thousands of years, although the active chemicals and their molecular targets are not well defined^10–12. Galbanic acid (GBA, also known as asacoumarin B, Fig. 1A), isolated from this herbal source, has antibiotic, anti--thrombotic and hepatoprotective properties^13–15. A parallel study¹⁶ carried out by our collaborative team suggests that GBA has strong anti-angiogenic activities, and daily administration of GBA by intraperitoneal (ip) injection with as little as 1 mg/kg body weight can inhibit the growth of Lewis lung carcinoma (LLC) allograft in syngenic mice. In addition, a previous study showed a good tolerance (50 mg/kg) of GBA in animals¹⁵. These observations suggest bioavailability of GBA and/or its metabolites to exert the biological activities in vivo.

Fig. 1.

A, Chemical structure of galbanic acid (GBA). B, GBA-induced growth inhibition (crystal violet staining of cellular protein) of prostate cancer cells. Each point represents mean ± sem, n=3. C, GBA-induced suppression of colony formation of LNCaP cells. Colonies were stained by crystal violet and counted (graph). D, Detection of apoptotic histone-associated DNA nucleosomal fragmentation by Death ELISA assay (left graph) and cleaved-PARP (cPARP) and cleaved-caspase-3 (cCaspase-3) by immunoblot (right image) in GBA treated LNCaP cells. Each column represents mean ± sem, n=3. E, Attenuation of GBA-induced apoptotic histone-associated DNA nucleosomal fragmentation by a general caspase inhibitor (zVADfmk). Each column represents mean ± sem, n=3. Statistical significance from control, *P<0.05, **P<0.01, ***P<0.001

GBA belongs to sesquiterpene coumarin-containing compounds, some of which (such as the pyranocoumarin decursin) have been demonstrated to exert anti-androgen and anti-neoplastic activities by our collaborative team^{17, 18}. Therefore, we hypothesize that GBA may have an inhibitory activity against PCa cell proliferation and survival through targeting AR signaling and/or other molecular pathways. We report here for the first time, to our knowledge, that GBA down-regulates AR protein abundance and signaling in LNCaP cells through mechanisms distinct from the clinically used AR antagonist drug bicalutamide (Bic) (Casodex) in spite of some predicted similarities on their binding to the AR ligand binding domain (LBD). We also show that GBA induces G₁ arrest in association with an inhibition of cyclin/CDK4/6/RB/E2F pathway.

MATERIALS AND METHODS

Compounds and reagents

Preparation of GBA (Fig. 1A) and chemical characterization have been reported elsewhere^{16, 19}. GBA (purity >99%) stock solution was prepared in DMSO and stored at −20°C. The stable androgen analog mibolerone was a gift kindly provided by Dr. Charles Young (Mayo Clinic, Rochester, MN) and dissolved in DMSO, and stored at −20°C. Decursin was prepared as previously described¹⁸. zVADfmk was purchased from R&D system (Minneapolis, MN). MG-132 and cycloheximide were obtained from Sigma-Aldrich (St Louis, MO). Bicalutamide was purchased from LKT Labs (St Paul, MN). All the commercial compounds were stored and prepared according to the vendor manuals except when indicated otherwise.

Cell culture

The human LNCaP, 22Rv1, DU145 and PC-3 cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA) and were maintained and treated under the ATCC recommended culture conditions except when indicated otherwise. The LNCaP C4-2 (an androgen-independent variant of LNCaP) cells were a generous gift from Dr. Donald Tindall (Mayo Clinic, Rochester, MN) and were cultured in the same condition as that of LNCaP cells. All cells were maintained under the standard 37°C and 5% CO2 humidified environment.

Cell Growth assay

Crystal violet staining of cellular proteins was used to evaluate the overall growth inhibitory efficacy of GBA on PCa cells as previously described²⁰. The PCa cells were treated with GBA and vehicle control (DMSO) for 72 h before analyses.

Colony formation assay

LNCaP cells were seeded onto 6-well plates (500 cells per well) and exposed to indicated concentrations of GBA in complete culture medium. The GBA treatment was refreshed twice a week. After 8–14 days of treatment, the cells remaining attached were fixed and stained with 0.02% crystal violet to visualize colonies for counting.

Cell cycle distribution and BrdU incorporation

PCa cells were exposed to indicated concentrations of GBA for 12, 24 or 48 h, and then BrdU was added to the culture medium (1µM final concentration). After 30 min of incubation with BrdU, the cells were collected for cell cycle distribution and BrdU incorporation assays as previously described²¹.

Immuno-blot

The whole cell lysate or subcellular fractions were prepared for immunoblot (IB) analyses as previously described^{18, 22}. Anti-PSA was obtained from DAKO (Glostrup, Denmark). Anti-AR, anti-ERα, anti-ERβ, anti-α-tubulin, anti-cyclin E₁, anti-CDK2, anti-CDK4, anti-CDK6, anti-P21^Cip1, anti-P27^Kip1 and anti-E2F1 were obtained from Santa Cruz Biotechnology Inc (Santa Cruz, CA). Anti-PARP, anti-cleaved PARP, anti-caspase-3, anti-cyclin D₁, anti-cyclin A, anti-phospho-RB (ser807/811), anti-phospho-RB (ser795), anti-P53, anti-VEGFR2, anti-AKT and anti-eNOS were obtained from Cell Signaling Technology (Beverly, MA). The band intensities were quantified with Quantity One software (Bio-Rad Laboratories, Hercules, CA).

Death ELISA for apoptotic DNA nucleosomal fragmentation

After exposure to GBA and a general caspase inhibitor (zVADfmk) for 48 h, the cells were collected for quantifying apoptotic histone-associated oligonucleosomes as previously described²⁰.

ELISA assays for secreted PSA

After exposure to GBA for 24 or 48 h, the secreted PSA in culture medium was quantified using a MAGIWEL PSA ELISA system (United Biotech, Mountain View, CA) as previously described²³.

Real-time PCR

Two (2) µg total RNA from each sample was reverse transcribed by using SuperScript™ II RT Kit (Invitrogen, Carlsbad, CA). Real-time PCR was performed by using FastStart Universal SYBR Green Master (Rox) system (Roche Diagnostics GmbH, Mannheim, Germany), AR primers: F 5 ’-CAGGAGGAAGGAGAGGCTTC-3’, R 5’-AGCAAGGCTGCAAAGGAGTC-3’. PSA primers: F 5’-CCCACT GCATCAGGAACAAA-3’, R 5’-GAG CGG GTG TGG GAA GCT-3’.

Luciferase reporter assays

LNCaP cells stably expressing a luciferase reporter (PSA-luc), named as LNCaP-PSA-luc, were exposed to GBA to determine the effect of GBA on AR-dependent transcription as previously described^{23, 24}. PSA-luc is a luciferase reporter driven by a 6.0 kb (−5824/+12) PSA genomic promoter sequence.

The transient transfection assays of AR-luc and E2F-luc activities were carried out to determine the effect of GBA on transcriptional activity of AR promoter and E2F-dependent transcription, respectively. AR-luc is a luciferase reporter driven by an 1.8 kb (−1380/+577) human AR genomic promoter (kindly provided by Dr. Charles Young, Mayo Clinic, Rochester, MN). E2F-luc is a luciferase reporter driven by a composite promoter containing six E2F binding sites and kindly provided by Dr. Kristian Helin (European Institute of Oncology, Milan, Italy)²⁵. Transfection was carried out as previously described²⁶.

Measurement of AR nuclear translocation

The relative AR abundance in the nuclear and cytosolic fractions of LNCaP cells was evaluated by immunoblot to determine the nuclear translocation of AR as previously described¹⁸.

Measurement of AR degradation

LNCaP cells were treated with cycloheximide to block new protein synthesis in the absence or presence of GBA for 3, 6, 9 and 12 h. The AR abundance in cellular lysate was measured by immunoblot. MG-132 (an inhibitor of 26S proteasomal-dependent degradation) was used to block the 26S proteasomal degradation pathway as previously described¹⁷. Decursin, which has been demonstrated to promote AR protein degradation¹⁷, was used as a positive control.

Molecular Modeling

In silico docking was done using the Schrödinger Suite 2009 (Schrödinger, LLC)²⁷. The induced fit docking (IFD)²⁸protocol, which takes into consideration the ligand-induced receptor conformational change, was used for all docking studies. Residues within 5 Å from the ligand were allowed to be flexible. The docking results were scored using the Extra-Precision mode of Glide version 5.0 (Schrödinger, LLC)²⁹. The AR protein structure was obtained from the protein databank (PDB ID: 3B5R). The induced fit docking protocol and parameters were first validated by separate docking of dihydrotestosterone (DHT) and bicalutamide (Bic) to AR. The docking results of both molecules excellently reproduced the protein-ligand binding in their corresponding complex crystal structures (PDB ID: 3L3X, 1Z95, respectively). The same protocol and parameters were then used to study the docking of GBA to AR.

Comparison of GBA with bicalutamide on AR signaling and cell growth

To test for AR agonist activity of GBA, LNCaP cells (1×10⁵ per well) were seeded into 6-well plates in phenol red-free medium supplemented with 5% char-coal stripped serum (CSS) as well as Bic or GBA in increasing concentrations. The DHT analog mibolerone (Mib) was added to additional wells to establish concentration-response patterns for PSA readout and cell growth. After 48 h exposure, 100 µL medium was collected for detection of secreted PSA as a read-out for AR signaling. The cells were maintained for another 6 days, then stained with crystal violet to evaluate the overall growth inhibitory efficacy as previously described²⁰.

To compare the effect of combination of GBA with Bic, LNCaP cells (1×10⁵ per well) were seeded onto 6-well plates in complete growth medium and treated with either agent alone or both combined at equal concentration. After 24h exposure, 100 µL medium was collected for detection of secreted PSA. The cells were maintained for another 7 days, then stained with crystal violet of cellular proteins to evaluate the growth inhibitory efficacy as previously described²⁰.

Overexpression of cyclin D₁ and knock-down of P21^Cip1 and P27^Kip1

Stable overexpression of cyclin D₁ in LNCaP cells was carried out as previously described for DU145 cells³⁰. Knockdown of P21^Cip1 and P27^Kip1 by small interference RNA (siRNA) were carried out as previously described³¹. All siRNA were obtained from Santa Cruz Biotechnology Inc (Santa Cruz, CA). After 24 h of transfection, the cells were fed fresh complete growth medium supplemented with GBA for another 24 h, and then the cells were collected for IB, cell cycle distribution and BrdU incorporation analysis.

Statistical analyses

Student’s two-tailed t-test was used to determine the statistical significance between the specific groups.

RESULTS

1. GBA inhibited PCa cell growth with a preferential potency against AR(+) PCa cells

To investigate the direct effect of GBA on PCa cells, human LNCaP, LNCaP C4-2, DU145 and PC-3 cells were treated with different concentrations of GBA for 72 h. The number of cells remaining attached was measured by crystal violet staining of cellular proteins. As shown in Fig. 1B, the number of AR(+) LNCaP and LNCaP C4-2 cells decreased in response to exposure of increasing GBA in a linear manner without an obvious threshold, with an IC₅₀ of ~ 80 µM. The number of AR(−) DU145 and PC-3 cells was minimally or not affected at 80 µM GBA. The IC₅₀ in these AR(−) PCa cell lines was likely > 120 µM each. The greater sensitivity of AR(+) cells was not related to the wild type P53 status in these cells as GBA treatment did not change their total P53 (supplement Fig. S2B) or phosphorylation status of Ser15 (not shown).

Since there was no detectable cell death in PCa cells exposed to 80 µM of GBA (Fig. 1D), we hypothesized that anti-proliferation, but not pro-apoptosis, should be primarily responsible for the growth inhibition induced by lower concentrations of GBA. A colony formation assay was carried out to assess the anti-proliferative activity of GBA over several cell cycles (Fig. 1C). After 14 days of exposure, GBA dramatically decreased the colony number of LNCaP cells with an IC₅₀ < 10 µM. In these cells, androgen signaling is crucial for their proliferation to form colonies, as demonstrated by the charcoal stripped serum (CSS) control in which depletion of steroids almost completely prohibited colony formation. GBA at 20 µM afforded a comparable inhibitory efficacy as androgen deprivation. Overall, these observations suggested that GBA could inhibit PCa cell growth with a preferential potency against AR(+) PCa cells.

2. GBA induced caspase-mediated apoptosis in LNCaP cells

Induction of apoptosis is an important mechanism for anti-cancer agents to rid of cancer cells. To investigate the pro-apoptotic activity of GBA, we exposed LNCaP cells to increasing levels of GBA for 48 h, and then analyzed the treated cells for apoptosis by either ELISA of histone-associated nucleosomal fragments (Fig. 1D, graph) or immunoblot detection of cleaved caspase-3 and cleaved PARP (Fig. 1D, image). GBA at 80 µM or lower did not induce detectable apoptotic cell death; but GBA between 100 and 120 µM dramatically increased apoptotic parameters. A general caspase inhibitor (zVADfmk) almost completely attenuated the GBA-induced apoptosis (Fig. 1E). These observations suggested that GBA could induce caspase-mediated apoptosis in LNCaP cells at > 80 µM. Henceforth, subapoptotic GBA will refer to ≤ 80 µM.

3. Subapoptotic GBA down-regulated AR and inhibited AR-dependent transactivation in PCa cells

Since GBA exerted a preferential inhibitory activity against the growth of AR(+) PCa cells (Fig. 1B), we hypothesized that the AR pathway might be one molecular target of GBA. Indeed, subapoptotic concentrations of GBA decreased AR abundance detectable by immunoblot in concentration- and time-dependent manners in AR(+) PCa cells including LNCaP, LNCaP C4-2 and 22Rv1 cells (Fig. 2A). PSA, one of the AR transcriptional targets, was decreased by GBA treatment at the cellular protein level (Fig. 2A), its secretion into the medium (Fig. 2B) and at the steady state mRNA level (Fig. 2C). The reduction of secreted PSA could not be completely explained by the cell number reduction. For example after 48 h treatment with 20 and 40 µM GBA, the reduction of secreted PSA was 65% and 90%, respectively (Fig. 2B), whereas after 72 h treatment the cell number reduction for these respective levels of GBA was 10% and 30% (Fig. 1B). In addition, GBA inhibited the PSA promoter-luciferase reporter activity (PSA-luc, which is driven by a 6.0 kb human PSA genomic promoter containing AR binding response elements) (Fig. 2D), providing further evidence that GBA could inhibit AR-mediated transactivation.

Fig. 2.

A, Immunoblot detection of AR and cellular PSA levels in GBA-treated AR(+) LNCaP, LNCaP C4-2 and 22Rv1 cells. AR and PSA ratios: the band intensities were normalized to those of beta-actin. B, ELISA detection of secreted PSA in conditioned medium of GBA-treated LNCaP cells. C. Real-time RT-PCR detection of PSA mRNA in GBA-treated LNCaP cells. D, Effect of GBA on the 6.0 kb PSA promoter-driven luciferase reporter (PSA-luc) activity in stable transfected LNCaP cells. The reporter activities were normalized to protein concentrations, and then converted into the relative folds to the mibolerone (Mib) stimulated control. Real-time PCR, PSA ELISA and luciferase reporter assays were done in triplicate. Statistical significance from control, *P<0.05, **P<0.01, ***P<0.001. E, Lack of effect of GBA on AR nuclear and cytosolic distribution in Mib stimulated LNCaP cells. Decursin (DEC) was used as a positive control. PARP and tubulin were used as the markers of the nuclear and cytosolic fractions, respectively. C: cytosolic fraction, N: nuclear fraction. F, Immunoblot detection of AR protein in GBA-treated LNCaP cells after new protein synthesis was blocked by cycloheximide (CHX). G, Effect of 26S proteasomal inhibitor (MG-132) on GBA-induced AR protein degradation in LNCaP cells. Decursin (DEC) was used as a positive control. AR ratio: the band intensity of AR was normalized to that of beta-actin.

As a nuclear transcriptional factor, the activity of AR is closely correlated to its nuclear translocation from the cytoplasm where it binds to its androgen ligand and forms homodimer³². To examine whether GBA interferes with this translocation process, we stimulated the androgen-deprived LNCaP cells with a synthetic androgen (mibolerone) in the presence or absence of GBA, and then examined the AR protein level in the cytosolic and nuclear fractions. Decursin, which has been shown to retard AR nuclear translocation¹⁸, was used as a positive control. Unlike decursin, GBA did not change the ratio between the cytosolic AR and nuclear AR (Fig. 2E). Taken together, these data suggested that GBA down-regulated AR protein and inhibited AR-dependent transactivation without interfering with ligand-stimulated nuclear translocation in PCa cells.

4. Subapoptotic GBA promoted AR protein degradation

To delineate the potential mechanism(s) by which GBA down-regulated AR protein abundance, we first examined its effect on the AR mRNA level. GBA moderately decreased AR mRNA abundance in LNCaP and LNCaP C4-2 cells (Supplement Fig. S1A). However, it did not affect the human AR promoter-luciferase reporter activity (AR-luc), which was driven by a 1.8 kb human AR genomic promoter (Supplement Fig. S1B). Since the moderate decrease of AR mRNA levels (10~20% for 40 µM, and 30~40% for 80 µM, Supplement Fig. S1A) was not sufficient to account for the dramatic decrease of AR protein (Fig. 2A), we hypothesized that a post-translational degradation might also be involved in the GBA-induced down-regulation of AR. Indeed, GBA treatment shortened the half-life (t_1/2) of AR protein from > 12 h to 3 h < t_1/2 < 6 h after new protein synthesis was blocked by cycloheximide (CHX) in LNCaP cells (Fig. 2F).

The 26S proteasomal pathway has been known to be responsible for the degradation of cytosolic AR protein which is stabilized by binding to the chaperone protein heat shock protein-90³³. Inhibition of this proteasomal pathway by MG-132 attenuated the GBA-induced down-regulation of AR protein (Fig. 2G). In this aspect, GBA resembled the action of decursin, which has been shown by our group to promote AR degradation via proteasomal degradation¹⁷. Taken together, these data suggested that, in addition to the moderate post-transcriptional reduction of AR mRNA, the 26S proteosomal-dependent protein degradation was primarily responsible for the down-regulation of AR protein induced by GBA. It is noteworthy that the promoting activity on AR degradation of GBA was lacking from bicalutamide (Bic), which could stimulate AR nuclear translocation³⁴ and stablize AR¹⁸ in the LNCaP cells.

5. Docking of GBA to AR

To gain structural insight into how GBA targets AR function, we carried out docking studies of GBA against the wild type AR ligand binding domain (LBD) (Fig. 3A) and compared the results with the endogenous ligand DHT (Fig. 3C) and antagonist drug bicalutamide (Bic) (Fig. 3B). Induced fit docking protocol was used to mimic the conformational change of the AR in response to ligand/drug binding. Both DHT- and Bic-bound AR LBD complex crystal structures have been solved. Comparing the docked GBA-AR LBD structure with these two crystal structures (Fig 3) showed similar binding mode with several key residues shared by all three ligands. Except for their size differences, these three ligands essentially bind within the same pocket of the LBD (Fig. 3D). The binding of DHT is stabilized by forming hydrogen bonds with ARG752, GLN711 and THR877. The same hydrogen bonds were observed in the GBA docked results. On the other hand, the hydrogen bonds formed with HIS874 contributed to the binding of both GBA and Bic (Fig. 3, A–C). Interestingly, while the docking results suggest an overall similar AR binding mode of GBA as Bic, the multiple hydrogen bonds formed between AR residues (LEU704, ASN 705) and the middle part of Bic molecule were not observed in the docked GBA-AR results; instead, a strong hydrophobic group of GBA occupied the middle area of the binding pocket. The binding free energies between AR and these three ligands were further estimated based on their IFD results using the same wild type AR LBD structure (PDB ID: 3B5R). The calculated binding free energy for GBA, DHT and Bic were approximately −12.9, −12.0 and −12.1 kcal/mol, suggesting a comparable or even higher binding affinity of GBA than the other two known AR ligands.

Fig. 3.

Structural representation of ligand-AR LBD interactions (A–D). Ligands were represented as solid sticks, protein residues near ligands were shown in wires. A, GBA-AR interaction based on induce-fit-docking results; B&C, Bicalutamide (Bic)-AR and DHT-AR interactions from their complex crystal structures (PDB ID: 1Z95, 3L3X, respectively). D. Structural alignment of A–C (only protein residues from A were shown) with carbon atoms of GBA, Bic and DHT colored as green, grey and orange, respectively. Key residues for ligand binding were labeled. Hydrogen bonds were shown in dashed yellow lines. E. Increasing DHT analog mibolerone (Mib) relieved GBA-induced growth inhibition of LNCaP cells. LNCaP cells were cultured in RPMI-1640 medium supplemented with 5% charcoal stripped serum with the indicated concentrations of Mib and GBA for 9 days. DMSO was used as the vehicle control (Veh). F. Comparison of GBA with Bic on secreted PSA (first 48 h) and cell growth (8 days) under steroid-deprived condition. Mib dose-response was included to show disparity of AR signaling and growth stimulation responses. G. Combination effect of GBA and Bic on secreted PSA (first 24 h) and cell growth (8 days) in complete medium. mean ± sem, n=3.

Intrigued by the docking prediction, we tested whether increasing androgen level could relieve LNCaP cells from GBA-induced growth inhibition under steroid deprived condition. As shown in Fig 3E, the DHT analog mibolerone (Mib) stimulated LNCaP cell growth at both 10 and 50 pM. GBA (1–30 µM) concentration-dependently inhibited the LNCaP colony growth stimulated by 10 pM Mib. However, it required much higher GBA concentration (10–30 uM) to suppress colony growth induced by 50 pM Mib. The data were consistent with GBA competing for binding with androgen ligand to the AR LBD.

5. Comparison of GBA and bicalutamide effects on LNCaP AR signaling and cell growths

We compared the effects of GBA and Bic under steroid deprived condition on LNCaP cell colony growth and on PSA secretion as a read-out for AR signaling for possible agonist activity. As shown in Figure 3F, Mib caused a concentration dependent stimulation of PSA secretion (48 h), and stimulated cell colony growth more than 3 fold at 0.03 nM and lost potency as its concentration was increased to 0.13 nM. The results were in good agreement with apoptosis induction by supra-physiological level of DHT analog R1881 in LNCaP cells³⁵. Whereas Bic in the range of 5 to 20 µM did not decrease PSA secretion, GBA caused a concentration dependent lowering of secreted PSA. GBA decreased the cell number more effectively than did Bic, both in a concentration-dependent manner. Overall, the data showed that GBA lacked agonist activity.

Next we evaluated the growth inhibitory effect of GBA and Bic and their combinations in complete medium. As shown in Fig. 3G, Bic exerted a concentration-dependent suppression of cell number within the range of 10–40 µM, but the slope of the response curve was steeper for GBA in the same concentration range. Combining Bic with GBA at 5 µM each exerted greater growth suppression than each alone at 20 µM. Combining 10 µM each out-performed the growth suppression than each alone at 20 µM. Similarly, the combination of 5 µM each of Bic and GBA in the first 24 h of exposure suppressed secreted PSA equal to or more than each at 20 µM (Fig. 3G). These results therefore suggested mechanistic differences between GBA and Bic to allow greater than additive action when combined.

6. Subapoptotic GBA induced G₁ arrest in association with an inhibition of cyclin/CDK4/6 pathway in PCa cells

In terms of cell cycle effects of subapoptotic GBA, Fig. 4A showed that GBA increased G₀/G₁ phase cells accompanied by a reduction of S and G₂/M phase cells in a concentration- and time-dependent manner (i.e., not affected at 12 h [data not shown], induced G₁ arrested at 24 h (Fig. 4A) and 48 h [data not shown]) in LNCaP cells. The distribution of BrdU(+) cells followed the same pattern as S phase cells (data not shown), supporting a simple G₁ arrest in GBA-treated LNCaP cells. The minimal concentration of GBA to cause a statistically significant G₁ arrest was 20 µM in LNCaP cells, whereas, the less sensitive AR(−) DU145 cells underwent moderate G₁ arrest at 80 µM of GBA exposure at 24 h (Fig. 4B).

Fig. 4.

A and B, Cell cycle distribution of GBA-treated LNCaP and DU 145 cells. C and D, Immunoblot detection of effect of GBA on the G₁ phase cyclin(s) and CDK(s) levels in LNCaP vs. DU 145 cells. E, Immunoblot detection of E2F1 and cyclin A levels (images), and inhibition of E2F-luc reporter activity in GBA-treated LNCaP cells (graph). Cell cycle distribution and luciferase reporter assays were done in triplicate. Statistical significance from control, *P<0.05, **P<0.01, ***P<0.001.

The cyclin/CDK/RB/E₂F pathway constitutes the basic machinery controlling G₁-S transition and S phase entry³⁶. To delineate the mechanism(s) by which GBA caused G₁ arrest, we examined the key components of cyclin/CDK/RB/E₂F pathway in GBA-treated LNCaP and DU145 cells. GBA at 40 and 80 µM down-regulated CDK4 (a HSP90 client protein) abundance at 6 h and the levels of CDK6, CDK2, cyclin D₁, and cyclin E in LNCaP cells at 24 and 48 h (Fig. 4C). GBA de-phosphorylated CDK substrate protein RB at multiple sites with similar kinetics (Fig. 4C). Although GBA did not affect E₂F1 level (Fig. 4E), it inhibited the E₂F-dependent trans-activation in LNCaP cells, measured with a luciferase reporter activity driven by a constructed promoter containing six repeats of canonical E2F binding site (Fig. 4E). In addition, the absence of detectable cyclin A, an S phase cyclin which is a transcriptional target of E2F, in GBA-treated cells (Fig. 4E) also supported the inhibition of cyclin/CDK/RB/E2F pathway signaling activity by GBA to prevent S phase entry.

In the DU145 cells, WB analyses showed that GBA, at 80 µM, but not 40 µM, down-regulated the abundance of CDK4, CDK6, cyclin D₁, and cyclin E but de-phosphorylated RB at 48 h without affecting CDK2 (Fig. 4D). Taken together, these observations suggested that GBA induced G₁ arrest in association with an inhibition of cyclin/CDK activity, resulting in accumulation of hypo-phosphorylated growth inhibitory RB to suppress E2F pathway in PCa cells, with a preference for AR(+) PCa cells.

7. GBA-induced G₁ arrest was associated with down-regulation of cyclinD₁, but independent of P21^Cip1 and P27^Kip1

The G₁ phase CDKs (CDK4, CDK6 and CDK2) activities are positively regulated by the bound cyclins such as cyclin D and E, but negatively by the CDK inhibitory proteins (CKIs) such as P21^Cip1 and P27^{Kip1 36}. To investigate if these molecular events were causally related to the GBA-induced G₁ arrest, we overexpressed cyclin D₁ as stable clone (Fig. 5A) or knocked down P21^cip1 and P27^Kip1 (Fig. 6B) in LNCaP cells, and then examined the response of these cells to GBA. As shown in Fig. 5B and C, stable over-expression of cyclin D₁ attenuated GBA-induced G₁ arrest.

Fig. 5.

A, Immunoblot detection of expression of cyclin D₁ in LNCaP cells stable transfected with cyclin D₁ plasmid (LN-cyclinD₁). B, cell cycle distribution of GBA-treated LN-vector (left graph) vs. LN-cyclin D₁ cells (right graph). C, BrdU incorporation in GBA-treated LN-vector (left graph) vs. LN-cyclin D₁ cells (right graph). Assays were done in triplicate. Statistical significance from control, *P<0.05, ***P<0.001.

Fig. 6.

A, Immunoblot detection of P21^Cip1 and P27^Kip1 levels in GBA-treated LNCaP cells. B, Immunoblot confirmation of knock-down of P21^Cip1 and P27^Kip1 by siRNA in LNCaP cells (24 h after transfection). C, lack of attenuation effect of knock-down of P21^Cip1 and/or P27^Kip1 on GBA-induced G₁ arrest in LNCaP cells. D, a proposed scheme of GBA-induced AR-dependent and/or AR–independent molecular and cellular pathways in PCa cells. T: testosterone/DHT; RTKs: Receptor tyrosine kinases.

There was a moderate increase of P21^Cip1 and P27^Kip1 after exposure to 80 µM GBA, but not at 40 µM in the LNCaP cells (Fig. 6A). However, knocking down of P21^Cip1 and/or P27^Kip1 (Fig. 6B) did not modify GBA-induced cell cycle arrest (Fig. 6C) or BrdU incorporation (data not shown). Taken together, the data suggested that GBA induced G₁ arrest in association with, at least in part, the down-regulations of CDK4/6 and cyclin D₁, but independent of P21^Cip1 and P27^Kip1.

DISCUSSION

The AR signaling plays a paradoxical role in the prostate, being essential for normal epithelial functional differentiation and survival, and subsequently essential for driving the malignant behavior in PCa including the incurable CRPCa. Studies suggest that the AR-dependent regulatory mechanisms are subverted but not bypassed in CRPCa^{3, 4}. CRPCa compensate for imposed androgen deficiency in several ways: 1) up-regulated or mutant AR are hypersensitive to broadened ligands or bypass the need for ligand, 2) altered coregulators facilitates the transcriptional activity of AR, 3) protein kinases (e.g., Src) and growth factor pathways aberrantly activate AR, and 4) altered androgen synthesis and metabolism pathways increase intra-tumoral androgen levels. Furthermore, recent studies suggest that the AR can be activated by interleukin class of cytokines through other transcriptional factors such as STAT3 and a NF-kB isoform p52^{37, 38}. These observations support that AR still plays a crucial role for the progression of CRPCa. The current AR antagonist drugs including bicalutamide (Bic) succumb to resistant mechanisms outlined above. Therefore, there exists a need for alternative strategies to block AR signaling³⁹. Decursin^{17, 18} and MDV-3100⁹ are some examples of these newer compounds without agonist effects.

Our data suggest that GBA may be a potential AR inhibitor with distinct actions from Bic. Salient findings concerning GBA include subapoptotic GBA down-regulated AR protein through promoting proteosomal degradation (Fig. 2), which was lacking by Bic¹⁸, and preferentially inhibited proliferation of AR(+) PCa cells through an induction of G₁ arrest (Fig. 1 and 4). Whereas the predicted binding of Bic and GBA to AR LBD (Fig. 3) suggested some similarities as well as differences in their binding modes to AR LBD, the results on PSA readouts of GBA vs. Bic under androgen-deprived condition (Fig. 3F) and the greater than additive anti-proliferative action of combining Bic with GBA than each agent alone (Fig. 3G) argue for their mechanistic distinctions.

It has been well established that AR plays a crucial and indispensable role in the G₁-S transition of AR(+) PCa cells⁴⁰. In addition, AR has been shown to be a licensing factor of DNA replication in AR-dependent PCa cells⁴¹. In the present study, the close parallel between the inhibition of AR signaling (Fig. 2) and the induction of G₁ arrest (Fig. 4) suggested that the inhibition of AR signaling might, at least in part, be responsible for the anti-proliferative activity of subapoptotic GBA (See signaling scheme in Fig. 6D). The cyclin/CDK/RB/E2F pathway is a master regulator of G₁-S transition and S phase entry during the cell cycle progression (see signaling scheme in Fig. 6D). E2F is released in late G₁ phase from RB when the latter is hyperphosphorylated by holoenzyme complexes made up of G₁ phase cyclins, such as cyclin D and cyclin E, their associated CDKs, such as CDK4, CDK6 or CDK2 and CKIs, such as P21^Cip1, P27^Kip1, etc. The released E2F induces multiple transcriptional targets including E2F itself, CDK2, cyclin E, cyclin A, thymidylate synthase, ribonucleotide reductase-2 and DNA polymerase-α, to propel G₁-S transition, S phase entry, DNA synthesis and carcinogenesis⁴². Previous studies suggest a cross-talk between the AR and cyclin/CDK/RB/E2F pathways during the cell cycle progression of AR(+) PCa cells. For example, AR signaling can up-regulate CDK4, CDK2 and cyclin D₁ to promote cell cycle progression^{43, 44}. Conversely, elevated cyclin D₁, cyclin E and CDK6 may deregulate AR activity^45–47. In the present study, we detected multiple changes in G₁ phase kinases (CDK4 and CDK6) and cyclins (cyclin D₁ and cyclin E), and the consequent dephosphorylation of RB and attenuation of E₂F-dependent transcription in GBA-treated LNCaP cells (Fig. 4C and E). Although these changes might be the consequences of the suppression of AR signaling by GBA, we could not exclude the direct impact of GBA on these molecules. Indeed, higher concentration and longer exposure of GBA induced moderate changes in CDK4, CDK6, cyclinD₁, cyclinE and cell cycle distribution in AR(−) DU145 cells (Fig. 4D), albeit the extent of which was less than that in AR (+) LNCaP cells (Fig. 4C). Further work is needed to delineate the relationships between the AR signaling axis and the cell cycle regulatory pathway in their responses to GBA.

Interestingly, several molecules down-regulated by GBA in PCa cells such as AR (Fig. 2) and CDK4 (Fig. 4) and estrogen receptor ERα (supplement Fig. S2B) are HSP90 client proteins⁴⁸. As a chaperone protein, HSP90 regulates more than 200 client proteins in various signal pathways. An important function of HSP90 is to stabilize cytosolic AR to evade proteosomal degradation. On the other hand, several other HSP90 client proteins such as P53, VEGFR2 or eNOS were not affected by GBA within a similar time frame (supplement Fig. S2B). Further studies are needed to test whether GBA might be a specific HSP90 modulator for some client proteins.

Aside from the G₁ arrest action (anti-proliferation) that may primarily contribute to the growth inhibition of PCa cells, GBA at higher levels induced caspase-mediated apoptosis. In addition, a parallel study¹⁶ has shown an in vivo efficacy against Lewis lung carcinoma in association with an anti-angiogenic activity of GBA, documenting in vivo bioavailability of GBA and/or its metabolites. It is noteworthy that a similar exposure range (10~40 µM) was required for the in vitro anti-angiogenic activities of GBA in cell culture models (HUVEC). Taken together, our observations suggest that GBA may be a PCa drug candidate through direct inhibition of AR abundance and signaling and anti-proliferative actions on the cancer cells (CDK4/6, cyclin D₁), and indirect actions on the tumor microenvironments (such as suppressing tumor angiogenesis). Further studies are warranted to validate the efficacy and molecular targets of GBA against PCa in vivo.

Supplementary Material

Supp Fig S1

Supplement Fig. S1. A, Real-time RT-PCR detection of AR mRNA in GBA-treated LNCaP and LNCaP C4-2 cells. B, lack of effect of GBA on the 1.8 kb AR promoter-driven luciferase reporter (AR-luc) activity in LNCaP cells. Real-time PCR and luciferase assays were done in triplicate. Statistical significance from control, *P<0.05, **P<0.01.

Supp Fig S2

Supplement Fig. S2. A. Immunoblot detection of effect of subapoptotic GBA on the steroid receptor ERβ in LNCaP cells. B, effect of subapoptotic GBA on several representative HSP90 client proteins in LNCaP, PC-3 and human umbilical vein endothelial cell (HUVEC) cells, respectively.