Androgen-independent DU145 and PC3 prostate carcinoma cell lines (ATCC, Rockville, MD) were cultured in RPMI containing 10% heat-inactivated FCS. DU145 subclones resistant to 5 μM 4HPR (R5) were established from cultures progressively exposed to increasing concentrations of 4HPR starting from 1 μM and cloning by limiting dilution following published procedures 17 . 4HPR (kindly provided by Dr. James A. Crowell, Division of Cancer Prevention, NCI, Bethesda and Dr. Gregg Bullard, McKessonBio, Rockville, MD) was dissolved in ethanol at a stock concentration of 10 mM and stored in aliquots at -20°. Wortmannin, LY294002, N-Acetyl-L-cysteine (NAC) and diphenyleneiodionium chloride (DPI) were from Sigma (Milano, Italy), IGF-1 and bone morphogenetic protein-2 (BMP-2) were from R&D (Minneapolis, MN, USA). VEGF protein released into the media by PC cells was measured using a commercial human ELISA kit (Biosource, Invitrogen, Carlsbad, CA, USA).

In vitro cell proliferation was performed on cells plated in 96-well plates at 3,000 cells/well dilution and grown in complete medium or treated as described. The medium was changed every two days. At different time points, the number of viable cells was evaluated by the crystal violet assay.

Chemotaxis and chemoinvasion assays were carried out in Boyden chambers as previously described 18 . The cells (12 × 10⁴/chamber) were extensively washed with PBS, resuspended in serum-free media (SFM) and placed in the upper compartment with or without selected molecules. In parallel experiments, trypan blue exclusion under all the conditions tested showed no altered cell proliferation or viability compared with controls during the five hours of chemotaxis test. The two compartments of the Boyden chamber were separated by 8 μm pore-size polycarbonate filters coated with 5 μg/50 μl/filter of collagen type IV (diluted in H₂O, 0.1% CH₃COOH) for the chemotaxis assay, or with Matrigel (60 μg/filter), a reconstituted basement membrane, for the invasion assay. Serum free 24 h-conditioned medium from human fibroblasts (FB-CM) was used as a chemoattractant in the lower chamber. After 5 hours of incubation at 37° in 5% CO₂, the filters were recovered, fixed in ETOH and stained by Toluidine blue after removal of the non-migrating cells on the upper surface. The migrated cells were quantified counting five to ten fields for each filter under a microscope. Graphical results are shown as percent inhibition as compared to a 100% untreated control.

Proteins were obtained from PC cells after four hours culture in the absence or presence of the indicated molecules as described in the text and in figure legends, or after 4 days in the presence of BMP-2 (50-100 ng/ml). To perform AKT activation, thirty min before the end of the incubation, the cells were stimulated with IGF-I (100 ng/ml). The cells were then lysed in RIPA buffer containing protease inhibitors. Protein concentration was determined with the DC Protein Assay kit (Bio-Rad). Equal amounts of samples were resolved by SDS-PAGE, transferred to nitrocellulose and probed at 4°C overnight with the following anti-human antibodies (Cell Signaling Technology, Beverly, MA): rabbit polyclonal anti-phospho-FAK (Tyr576/577), phospho-AKT-1 (Ser473), phospho-GSK3β (Ser9), β-catenin, phospho-β-catenin (Ser552), cyclin D1, survivin and E-cadherin. After washing, the blots were incubated for 1 h at room temperature with horseradish peroxidase-conjugated secondary antibodies (GE-Healthcare, Milano, Italy) and specific complexes were revealed by enhanced chemiluminescence (ECL, GE-Healthcare). An anti-GAPDH antibody conjugated to horseradish peroxidase (Novus Biologicals, Littleton, CO) or a mouse monoclonal anti-β-tubulin antibody (Sigma, Milano, Italy) were utilized as loading controls for all samples.

pCMV6-Myr.Akt (a constitutively active Akt) and control vector were kindly provided by Dr. Alex Toker of the Beth Israel Deaconess Medical Center of Boston. Transient transfections were performed with lipofectamine 2000 (Invitrogen). β-catenin was silenced using a lentiviral vector expressing short hairpin RNA (MISSION shRNA clones, Sigma) following manufacturer's instructions. This resulted in the generation of stable DU145 and PC3 cells showing decreased β-catenin expression. Control vectors contained scrambled sequences.

Data are expressed as means ± SD. The statistical significance between two data sets was determined by two-tailed unpaired Student's t test using the PRISM GraphPad software.

We first assessed the action of 4HPR on prostate cancer cell proliferation by treating DU145 and PC3 cells with a range of concentrations. All the concentrations tested inhibited cell growth, with statistically significant differences only after 96 h exposure (Fig 1A, B left panels).

Migration of cancer cells is one of the key factors responsible for cancer metastasis. To metastasize, cancer cells must migrate from the original growth site, invade surrounding tissues and locate to other parts of the body through the blood or the lymphatic system. The effects of 4HPR on PC cell migration and invasion were then tested in a broad dose/response experiment. Untreated PC cells migrated (Fig. 1A, B middle; DU145 and PC3 respectively) and invaded through Matrigel (Fig. 1A, B right; DU145 and PC3 respectively) in response to fibroblast conditioned medium (FB-CM). A very short exposure (5 h) to micromolar concentrations of 4HPR significantly inhibited migration and invasion of DU145 (Fig. 1A) and PC3 (Fig. 1B) cells. To explore the possibility that ROS induction by 4HPR was a mechanism underlying these effects, PC3 cells were pretreated 1 h with the ROS scavengers NAC and DPI (at 10 mM and 1 μM respectively) and then subjected to chemotaxis in the presence of 4HPR at different concentrations. As shown in Fig. 1C, both compounds did not modify 4HPR effects. Additional Fig. 1 (Fig. 1S) shows that 30 min pretreatment with the ROS scavenger N-Acetyl Cysteine (NAC) at 10 mM significantly decreases ROS production induced by 4HPR treatment (1 h at 5 μM).

Focal adhesion kinase (FAK) is a non-receptor tyrosine kinase that plays an important role in signal transduction and is a key regulator of survival, proliferation, migration and invasion. Overexpression and/or increased activity of FAK are common in a wide variety of human cancers, implicating a role for FAK in carcinogenesis. DU145 and PC3 cells express high levels of activated FAK, which was rapidly (4 hours) downregulated by 4HPR (Fig. 2A, B). The survival and migratory signaling mediated by FAK operates via activation of the PI3K/AKT pathway 19 20 that in turn promotes prostate cancer cell migration and invasion 21 . Exposure to 4HPR rapidly decreased AKT phosphorylation in both cell lines (Fig. 2A, B). In agreement, prostate cancer cell migration towards FB-CM was significantly inhibited in the presence of the specific PI3K inhibitors Wortmannin (Wort, 200 nM) and LY294002 (LY, 10 μM). Co-exposure of the inhibitors with 4HPR produced more pronounced effects being the combination Wotmannin/4HPR the most effective. (Fig. 2C, DU145 cells; PC3 cells gave similar results). These data suggest that these pathways are partially independent of each other. As AKT overexpression correlates with increased VEGF levels and prostate tumor angiogenesis 22 , we determined VEGF release in 4HPR-treated cells by ELISA. We noted that 4HPR-induced AKT downregulation is associated with reduced VEGF secretion (Fig. 2D).

The AKT activator IGF-1 is a potent mitogenic and motogenic factor and has a prominent role in protection against apoptosis and cell survival. IGF-1 has been implicated in the initiation and progression of several different cancers including prostate cancer 23 . Chemotaxis assays showed that IGF-1 stimulates androgen-independent DU145 prostate cancer cell migration and co-exposure to 4HPR completely abrogated the effect (Fig. 3A). Accordingly, activation of the AKT signaling pathway by a short exposure to IGF-I (30 min at 100 ng/ml), as detected by western blot analysis, was lowered by 4HPR pretreatment (Fig. 3B).

Next, we determined the effect of overexpression of myristoylated Akt (Myr.Akt), which is anchored to the plasma membrane and has a constitutively active kinase activity. Cells transfected with the empty vector were used as controls. Western blot analysis of extracts from DU145 cells transiently transfected with constitutively active Akt showed high levels of total Akt and phosphorylated (Ser 473)-Akt as compared with the empty vector-transfected control cells (Fig. 3C). Ectopic expression of constitutively active Akt abrogated 4HPR-mediated inhibition of DU145 cell migration (Fig. 3C).

β-catenin is a multifunctional protein not only involved structurally in the adherens junction complex, but also acting as a signaling molecule promoting cancer cell proliferation, survival and migration. β-catenin signaling is aberrantly activated in greater than 70% of colorectal cancers and has been shown to play a causative role in prostate cancer 24 . As β-catenin turnover is triggered by GSK-3β, a well known target of AKT, we made β-catenin signaling a focal point of our investigation. We measured the soluble levels of β-catenin by western blot analysis in 4HPR-treated DU145 cells (Fig. 4A, left) and evaluated cyclin D1 and survivin as typical transcriptional targets of the β-catenin/TCF/LEF complex 25 26 (Fig. 4A, middle). As shown in Figure 4A, DU145 cells exhibited high levels of β-catenin, cyclin D1 and survivin expression. A short exposure to 4HPR (4 hours) resulted in a significant reduction in the levels of these proteins. Pre-incubation of cells with the ROS scavenger NAC (1 h at 10 mM) did not compromise the ability of 4HPR to modulate β-catenin thus suggesting a redox-independent signaling (Fig. 4A right). Similar results were obtained with PC3 cells (data not shown). These data suggest that the highly activated β-catenin signaling was suppressed by short-term 4HPR treatment in prostate tumor cells.

The expression of β-catenin is rather low in non-cancer cells as GSK-3β causes its phosphorylation and consequent degradation by the ubiquitin-proteasome pathway. GSK-3β activity is suppressed when it is phosphorylated on serine 9 by AKT. In agreement with the decreased phosphorylation of AKT after 4HPR treatment (Fig. 2A, B, 3B), GSK-3β phosphorylation was decreased by 4HPR (Fig. 4A, left). As a confirmation, a short exposure (2 hrs) to the specific PI3K/AKT inhibitors Wortmannin and LY294002 reduced GSK3β phosphorylation,β-catenin and cyclin D1 levels (Fig. 4B). AKT was also shown to directly phosphorylate β-catenin at Ser552, independent of GSK3β. Phosphorylation at Ser552 induces β-catenin accumulation in the nucleus, increases its transcriptional activity and promotes cell invasion 27 28 . According to AKT inhibition by 4HPR exposure, β-catenin phosphorylated at Ser552, probably representing the residual nuclear pool escaping proteasomal degradation, was decreased (Fig. 4A, left). Activation of AKT by IGF-I exposure (Fig. 4C) suppressed GSK3β activity, stabilized β-catenin, increased Ser552 phosphorylation and cyclin D1 levels, yet pretreatment with 4HPR abolished the effects of IGF-I stimulation (Fig. 4C). Together, all these data suggest that 4HPR-induced pAKT down-regulation exerts a multi-level control on total and nuclear β-catenin, dependent and independent of GSK3β activity.

In order to understand whether the observed signaling was univocally related to sensitivity of cell lines to 4HPR, our analysis was extended to DU145/R5, a cell line resistant to 5 μM 4HPR generated by in vitro incubation of DU145 cells with increasing concentrations of 4HPR. Supplementary Fig. 2 (Fig. 2S) shows that DU145/R5 cells grow (96 h treatment, panel A) and migrate (5 h treatment, panel B) in the presence of 2.5 and 5 μM 4HPR. This 4HPR-resistant phenotype is associated to very high levels of phosphorylated AKT (panel C).

We first investigated whether specifically reducing the levels of β-catenin resulted in decreased proliferation, migration and invasiveness. For this analysis, RNA interference with shRNAs directed against β-catenin (sh-βcat) was used and comparisons were made with controls with scrambled sequences (sh-NT). As shown in Fig. 5, shRNA targeting of β-catenin (insets panel A) resulted in a 40% and 30% decrease in cell proliferation seen at 96 h (Fig. 5A, DU145 and PC3 cells respectively). β-catenin silencing significantly reduced also cell migration (Fig. 5B, DU145 and PC3 cells respectively) and invasiveness (Fig. 5C, DU145 and PC3 cells respectively).

Next, we determined whether constitutively active Akt (Myr.Akt) could abrogate the effects of β-catenin silencing on cell migration and invasion. β-catenin-silenced DU145 cells transiently expressing Myr.Akt showed an enhanced migratory and invasive phenotype as compared to control vector transfected cells (Fig. 5D). In these cells GSK3β phosphorylation and β-catenin levels remained unaltered (Fig. 5D, right panel), as previously reported 29 , suggesting that under these conditions activation of Akt alone is not able to restore β-catenin levels through increased expression or stabilization.

Metastasis and angiogenesis are strictly related processes. We reported that the TGF-β family member BMP-2 is a mediator of the antiangiogenic activity of 4HPR 14 , controlling tumor growth. Since the role of BMPs in the formation of prostate cancer metastases remains unknown and controversial, we tested whether BMP-2 could influence PC cell growth, migration and invasion. We previously found 14 that the exposure of endothelial HUVE cells to 5 μM 4HPR caused the release of BMP-2 in the culture medium (50-100 ng/ml concentrations). When DU145 and PC3 cells were exposed to the same BMP-2 concentrations in long-term experiments (4 days and in the absence of 4HPR), cell growth, migration and invasion were significantly decreased in both cell lines (Fig. 6A, B). BMP signaling downregulates the β-catenin pathway in cancer cells 30 31 . While negative regulation on the β-catenin pathway by BMP signaling has been recognized to have a role in intestinal tumorigenesis in mice and humans 30 31 information is lacking about the relationships between the two pathways in prostate tumors. Western blot analysis of nuclear extracts from PC3 cells exposed for 4 days to BMP-2 (50-100 ng/ml) showed a dose-dependent decrease of nuclear β-catenin (Fig. 6C). We then looked at the possible mechanisms controlling β-catenin accumulation. We found that BMP-2 modulates AKT phosphorylation (Fig. 6C), but not that of pGSK3β (data not shown), further confirming that in prostate cancer cells β-catenin nuclear signaling is mainly controlled by AKT activity. As β-catenin promotes transcription of the proliferation gene cyclin D1, we also noted that BMP-2-treated cells exhibited significantly lower levels of cyclin D1 (Fig. 6D).

E-cadherin enforces cell-cell contacts forming the adherens junctions and is anchored to actin filaments by β and α catenin. E-cadherin loss promotes metastasis by enabling the first step of the metastatic cascade: the disaggregation of cancer cells from each another. BMP-7, another member of the BMP family, has been reported to be a potent inhibitor of prostate cancer metastasis 32 also acting through E-cadherin induction. PC3 cells exposed to BMP-2 exhibited increased levels of E-cadherin (Fig. 6D) suggesting that BMP-7 and BMP-2 have similar mechanisms of action. We cannot exclude, however, that β-catenin downregulation controls E-cadherin expression as reported for other compounds 33 .