In western industrialized countries, PC presents a serious health problem and is the second leading cause of cancer deaths in elderly men. Although PC is very heterogeneous in its etiology, androgen signalling is a key element in the development and progression of PC. Initially, PC cells are largely dependent on androgens for growth and survival. As a consequence, endocrine therapy involving androgen depletion by surgical or medical castration, as well as the blockade of the androgen receptor with anti-androgens, has become a standard treatment for advanced or metastatic disease. However, the benefit from endocrine therapies of advanced PC is only transitory. Within a period of around 2 years, nearly all prostate cancers progress to a castration-resistant state of the disease where they do no longer respond to standard endocrine therapies. In the past, it has been hypothesized that the state of CRPC was due to a clonal selection of AR-negative cells. This assumption was mainly based on the Dunning rat tumor model where the development of an androgen-insensitive state is linked to the loss of the AR in tumor cells during androgen withdrawal. However, clinical studies showed that the AR is rarely lost but is often increased in CRPC tumor specimens and their metastases 
. The fact that in the absence of androgenic stimuli many of the same genes that are increased by androgens in androgen-dependent PC become elevated in CRPC supports the notion of constitutively active AR proteins in CRPC cells 
. This hypothesis is also supported by the observation that a disruption of AR function inhibits proliferation of CRPC cells grown in the absence of androgens 
In this report we used the AR-positive LNCaP sublines C4-2 and LNCaP-SSR which are both known to grow and survive under androgen-deprived conditions as an in vitro
tumor model for CRPC 
. We were able to show that, in contrast to the androgen-dependent parental LNCaP cells, castration-resistant LNCaP sublines C4-2 and LNCaR-SSR exhibited high levels of AR and PSA when grown in the absence of androgens. Most interestingly, the increase in intracellular AR protein levels was paralleled by an increase in GSK-3β, a ubiquitous serine threonine kinase shown to be an important modulator of AR stability in vitro 
. Although at the moment we are unable to say that the dysregulation of GSK-3β levels is responsible for the tendency of PC cells to increase intracellular AR-levels while becoming castration resistant, our in vitro
observations are in line with a recent clinical study showing an accumulation of GSK-3β in cells of castration-resistant tumors 
. The detection of high intracellular PSA levels in C4-2 and LNCaP-SSR devoid of androgenic stimuli suggests that the AR is functionally active in these cells even under androgen- deprived conditions.
In order to generate genomic signals, the AR has to be transported into the nucleus. In this study, the AR was found to be predominantly nuclear in LNCaP-SSR cells and C4-2 cells in the absence of androgens. The reason for the androgen-independent nuclear accumulation of full-length AR in CRPC cells remains largely unknown. Under normal conditions the nuclear translocation of the AR is dramatically increased upon hormone binding as demonstrated for LNCaP. Recent findings suggest that, in order to enter the nucleus, the AR has to undergo an intra-molecular conformation change that brings the N- and C-termini of the molecule into close proximity. This intra-molecular conformational change allowing activation and nuclear translocation of an AR monomer, prior to AR dimerization, is usually induced by ligand binding and only occurs in living cells, not in cell lysates 
, which suggests that this intra-molecular reorganization of the AR is not protein-autonomous but depends on cellular factors. To test this hypothesis, we subsequently transfected C4-2 and LNCaP cells with an expression construct coding for a wild type AR fused to a green Eos fluorescent protein (EosFP) 
. In the presence of DHT AR-EosFP was detectable in the nuclei of both LNCaP and C4-2. When grown in the absence of DHT, AR-EosFP was only detectable in the nuclei of castration-resistant C4-2 cells but not in the nuclei of androgen-dependent LNCaP cells. This observation is consistent with a similar experiment demonstrating a robust nuclear localization of a GFP-tagged AR transfected into C4-2 cells 
. Taken together, these findings suggest that the nuclear localization of full-length AR in CRPC does not necessarily require hormone binding and may instead be regulated by other factors.
As recently shown by our group, short term inhibition of the serine/threonine kinase GSK-3β by small molecule inhibitors induced a rapid, CRM1-dependent nucleocytoplasmic shuttling of the AR in androgen-treated cells 
. As GSK-3β is overexpressed in CRPC cells in vivo and in vitro
, we hypothesized that the enzyme might also be part of a putative protein complex involved in the nuclear accumulation of the AR. This hypothesis is supported by a previous finding showing that GSK-3β is highly phosphorylated at tyrosine 216 (Y216), the activation function of the enzyme, in C4-2 cells 
. The increase of GSK-3βY216
phosphorylation in LNCaP cells following DHT treatment () furthermore suggests an important role for GSK-3β in AR signalling under normal conditions. In an attempt to disrupt the predominant nuclear localization of the AR in CRPC cells, we treated castration-resistant C4-2 and LNCaP-SSR as well as the parental LNCaP with the maleimide SB216763. This highly potent compound is known to inhibit intramolecular tyrosine kinase activity of the GSK-3β molecule, the latter being responsible for its autophosphorylation at Y216 
. As seen in , the GSK-3β inhibitor SB216763 is able to decrease Y216-phosphorylation of GSK-3β in LNCaP as well as in C4-2 cells. Following treatment with SB216763, the nuclear accumulation of the AR in the castration- resistant LNCaP sublines C4-2 and LNCaP-SSR was dramatically reduced in the presence, and most importantly in the absence, of DHT, being most pronounced in C4-2 cells (). Nucleocytoplasmic shuttling of the AR after GSK-3β-inhibition in C4-2 cells grown in the absence of DHT could be reversed by leptomycin B (LMB), suggesting a CRM1-dependent export mechanism (). In a previous study, we identified a CRM1 binding site in the C-terminus of the AR 
. Due to its vicinity to the ligand binding domain, we hypothesized that hormone binding could regulate the accessibility of the CRM1 binding site, thereby modulating the nuclear export of the AR. The observation that the AR can be exported from the nuclei of CRPC cells growing in the absence of androgens is an important novel finding, clearly demonstrating that the nuclear export of the AR following GSK-3β inhibition is not due to a modulation of hormone binding properties of the receptor.
The fact that short term inhibition of GSK-3β (max. 4 hours) leads to a rapid export of the AR prompted us to analyze the effects of a long term inhibition of the enzyme in AR-positive CRPC-cells. Silencing of GSK-3β using specific shRNA led to a depletion of nuclear AR in C4-2 cells. Nuclear depletion of AR protein was less pronounced in C4-2 cells grown in the presence of the AR-stabilizing hormone DHT. Although these findings support the assumption that inhibition of GSK-3β triggers the nucleocytoplasmic shuttling of the AR, the data derived from the long term shRNA-experiments must be interpreted carefully. Short term inhibition of GSK-3β activity by small molecules like SB216763 was repeatedly shown to induce a rapid, CRM1-dependent nuclear export of the AR in PC cells 
. In contrast to these findings, the long-term inhibition of GSK-3β caused a proteasomal degradation of the AR protein 
. Because GSK-3β triggers AR localization as well as AR stability, we hypothesize that the dramatic depletion of nuclear AR in C4-2 cells following GSK-3β silencing is due to a combination of nuclear export and proteasomal degradation of the receptor molecule.
The fact that inhibition of GSK-3β affects AR function as well as AR stability prompted us to analyze the effects of a GSK-3β inhibition on the proliferation of PC cells in vitro. Response to treatment of PC cells with SB216763 was most pronounced in AR-positive CRPC cells (inhibition of cell growth: C4-2>LNCaP-SSR>LNCaP>PC3). The ability of SB216763 to inhibit cellular proliferation was paralleled by its ability to induce a CRM1-dependent export of the AR in PC cells.
In order to test the efficacy of GSK-3β inhibitors as potential therapeutic agents, we extended our studies to a CAM-xenograft model. To our surprise the AR was predominantly nuclear in LNCaP xenografts as well as in C4-2 xenografts. A possible explanation for this could be the production of androgens by the host organism. Moreover, the AR of LNCaP and its derivatives is carrying a point mutation (T877A) leading to a promiscuous AR that can, at least in part, be activated by various non-androgenic steroids 
. Nevertheless, the AR of C4-2 as compared to the AR of the parental LNCaP is more active in the CAM-Xenograft model, as documented by intracellular PSA levels (cells positive for PSA: LNCaP 3,1% and C4-2 16.8%). Following a 48-hour treatment with SB216763, nuclear AR levels of C4-2 cells were significantly reduced, which was paralleled by a considerable reduction in intracellular PSA levels. In contrast to results in C4-2 cells the effects on AR and PSA in LNCaP cells remained insignificant.
Taken together our study demonstrates that (1) overexpression of the AR in CPRC cells is paralleled by an increase in intracellular GSK-3β, (2) the effects of GSK-3β on AR signalling are not due to a modulation of DHT binding to the AR, (3) nuclear accumulation of the AR in CRPC cells is not an autonomous function of a mutated AR receptor but is mediated by deregulated cellular factors, such as GSK-3β, (4) GSK-3β and CRM1 are part of a putative complex controlling the nuclear localization of the AR in CRPC cells, and (5) inhibition of GSK-3β activity by small molecule inhibitors induces a rapid nuclear export of the AR in CRPC cells in vitro and in vivo thereby modulating AR signalling.
The dependence of CRPC on transcriptionally active AR has resurged the interest in developing inhibitors targeting the AR or androgen axis. Next generation hormone therapies recognize the fact that in CRPC the AR can be activated by intrinsic production of tissue androgens as well as by peptide growth factors. Among hormonal agents currently being tested are CYP17 inhibitors like abiraterone, TAK-700 and TOK-001 or the second generation anti-androgen MDV-3100 
. Small molecule inhibitors like EPI-001 and sintokamide targeting the transactivation domain at the N-terminus of the AR have shown encouraging results in vitro as well as in experimental animal models 
. Our results suggest that simultaneous targeting of AR nuclear export as well as AR-stability by GSK-3β inhibitors is a further valuable strategy to diminish AR signalling in CRPC. Moreover, the combination of GSK-3β inhibitors with novel AR inhibitors might be a useful approach for the treatment of advanced CRPC.