When detected in the advanced stage, prostate cancer patients are treated with hormone therapies that reduce systemic androgen levels 123. This action palliates the symptoms of metastases, induces regression of metastatic lesions and slows prostate tumor growth 4. Over time, however, the cancer can recur in a castration resistant form (CRPC) that continues to grow despite the ability of hormone therapy to maintain systemic androgens at castrate levels and deaths from prostate cancer are inevitably associated with complications from this form of disease 5. Progression of prostate cancer to CRPC appears to involve a reactivation of androgen signaling in the cancer cells 678 and a variety of mechanisms may account for residual androgen signaling in a low androgen environment. These include expression of variant forms of androgen receptor (AR) that are transcriptionally active without ligand 910, acquisition of an ability to endogenously synthesize androgens by the tumor cells themselves 1112 or activation of aberrant AR transcriptional activity through cross-talk with alternate signaling pathways 613. While all of these mechanisms are of interest from a scientific viewpoint, the ones that are readily targetable by drugs are the most clinically imperative as they offer an opportunity to test novel therapies to treat a disease that will kill almost 28,000 men in the United States this year. Recent reports that Abiraterone, an inhibitor of androgen biosynthesis, has clinical effects against castration resistant prostate cancer, reflects a potential treatment advance that might target tumor cell androgen biosynthesis 14. Here we describe findings that suggest that inhibitors of the Hedgehog/Gli signaling pathway, currently in clinical testing for a variety of cancers, might also have a role for the treatment of castration resistant prostate cancer due to an ability to suppress reactivated androgen signaling in tumor cells.

Hedgehog (Hh) is best known for its role in tissue patterning and morphogenesis during embryonic development 15161718. In the developmental situation, Hh is a ligand-driven process in which a ligand (referred to as a Hedgehog) engages the Patched 1 (Ptch) receptor on the cell surface and this relieves repression of Smoothened (Smo), a member of the extended G protein coupled receptor family 18. Smo, when activated, then acts downstream to alter the processing and intracellular localization of Gli transcription factors and to increase Gli-mediated transcriptional activity. The plant-derived alkaloid, cyclopamine, is a prototype for a drug that antagonizes the Hh signaling process 19. Cyclopamine antagonizes Smo activation and this action explains the teratogenic effects of this drug when it is ingested during pregnancy 2021.

Aside from its role in development, Hh signaling also supports stem cells in adult tissues 222324. However, chronically hyperactive Hh/Gli signaling in adult tissues can be oncogenic, especially for the skin or brain 2526. Basal cell carcinoma of the skin and medulloblastoma are models for human Hh-mediated oncogenesis 27. The aberrant Hh activity in these tumors can result from a loss of the Ptch gene or its function 2829, mutations in Smo 30 or SuFu 31 that activate endogenous Hh signaling or cryptic overexpression of Gli proteins in tumor cells. For prostate cancer, the question as to whether Hh/Gli signaling plays any role is controversial. Although cyclopamine treatment or Gli knockdown suppresses the in vitro growth of prostate cancer cell lines or xenograft tumor growth in mice 323334, the commonly used prostate cancer cell lines show little, if any, evidence for active canonical Hh signaling activity when they are grown in standard culture conditions 3536. For the androgen-growth dependent LNCaP prostate cancer cells and its variants, C4-2 and C4-2B, however, the situation was found to be changed by chronic exposure of these cells to androgen depleted medium. Androgen deprivation highly upregulated the expression and secretion of Hh ligands and increased endogenous expression of Hh/Gli target genes in these cells 37. The clinical relevance of this observation is supported by the observation that Hh ligand production was found to be increased in prostate tumors by neoadjuvant hormone treatment 38. Since cyclopamine suppresses the expression of Hh target genes in androgen-deprived LNCaP cells (37), this also suggests that active Hh/Gli signaling activity is awakened by growth under androgen deprived conditions. Others have observed that the high basal expression of Hh/Gli target genes in androgen independent (AI) variants of LNCaP was reduced by cyclopamine 39 and, collectively, the outcomes of these studies imply that Hh signaling in LNCaP cells is restricted to the androgen deprived or AI state. The question remains as to whether active Hh signaling has any biological consequences for the androgen deprived or AI prostate cancer cell. Here we show that, by manipulating the activity of canonical Hh signaling in androgen deprived or AI prostate cancer cells, we also affected the expression of androgen regulated genes and the ability of these cells to grow in the absence of androgen. Our results indicate that Hh/Gli signaling activity supports androgen signaling and AI growth in prostate cancer under low/no androgen conditions. Furthermore, we report that Gli2 protein can bind to AR and this interaction might define the point of cross-talk between the two signaling pathways.

Previously we reported evidence for conditional activation of canonical Hh signaling in androgen sensitive human prostate cancer cells by culture in an androgen depleted conditions 37. Here, we used androgen sensitive parental LNCaP cells, other derivatives of LNCaP that are less dependent on androgens for growth (C4-2, LN3, LNCaP-AI) or androgen responsive VCaP cells that are unrelated to LNCaP, to study the effects of Hh signaling manipulation on the expression of androgen regulated genes in these cells. The LNCaP-AI variant cells that we used were independently isolated in our lab following long-term (> 1 year) culture of parental LNCaP cells in androgen depleted medium. These cells downregulate basal expression of Ptch1 when treated with cyclopamine (Additional file 1, Figure S1) so they appear to have basal-active Hh signaling activity similar to other AI derivatives of LNCaP that were previously described (39). Initially, we tested the effects of the classic Hh inhibitor drug, cyclopamine on androgen regulated gene expression. All experiments were done using a medium that was depleted for androgens (phenol red-free RPMI with charcoal-stripped FBS) that could be re-supplemented with androgen (R1881) to mimic androgen-stimulated conditions. For parental LNCaP cells grown in androgen supplemented medium (+R1881), the presence of cyclopamine had no significant effects on the expression of four model androgen-regulated genes; KLK2, KLK3 [PSA] and PGC (androgen-inducible), or SHH that is repressed by androgen (Figure 1A). However, when these cells were switched to androgen depleted medium (-R1881) for 3 days, cyclopamine treatment had a distinct dose-dependent effect that further suppressed expression of KLK2, KLK3 and PGC and further increased expression of SHH (Figure 1A). Likewise, cyclopamine significantly downregulated expression of KLK2, KLK3 and PGC in the LNCaP-AI cells that are normally propagated in androgen-free medium, and it upregulated the expression of SHH in these cells (Figure 1A). Cyclopamine also suppressed the expression of KLK2 and KLK3 in VCaP, LN3 or C4-2B cells grown in androgen depleted medium for 3 days (Additional file 1, Figure S2), so the effects of cyclopamine on androgen regulated genes were not limited to LNCaP or its derivatives. We also tested whether a more water-soluble cyclopamine derivative, KAAD-cyclopamine, had a similar effect and found that this drug (at 0.5 or 1 μM) was as effective in reducing KLK2/3 and PGC expression in androgen-deprived LNCaP or LNCaP-AI cells as the 5 or 10 μM dose of cyclopamine (Additional file 1, Figure S3). Finally, we found that cyclopamine also significantly diminished the expression of a reporter gene (luciferase) from either of two androgen dependent promoters (Probasin [PRB] or PGC) in LNCaP or LNCaP AI cells in androgen depleted medium (Figure 1B) in a dose dependent manner. As for endogenous androgen-regulated genes, cyclopamine did not affect the expression of the reporter when cells were cultured in medium supplemented with 10 pM R1881 (data not shown).

Cyclopamine represses Hh signaling through its ability to antagonize Smo activation so we also tested whether Smo expression knockdown (using siRNA) could mimic the effects of cyclopamine with regards to suppression of androgen-inducible gene expression. LNCaP cells were transfected either with control (non-targeting) siRNA or with siRNA targeting AR or Smo and were thereafter maintained in androgen-depleted medium. AR siRNA selectively reduced expression of AR mRNA and protein (Figures 2A, C) but did not reduce the expression of Smo. Likewise, Smo siRNA reduced Smo mRNA levels but did not affect expression of AR mRNA or protein (Figure 2C). However, both AR and Smo siRNAs similarly reduced expression of KLK2 and KLK3 (Figure 2A). Further assessment of the effects of AR or Smo siRNA on expression of a luciferase reporter from either a Gli- or androgen-responsive promoter showed that AR knockdown selectively reduced expression of the androgen reporter but did not affect expression of the Gli reporter (Figure 2B). In contrast, Smo knockdown significantly reduced expression of both the Gli and androgen reporters (Figure 2B) in androgen deprived LNCaP cells. In summary, the above data shows that suppression of Hh signaling with a Smo antagonist, cyclopamine, or by reduction of Smo expression itself, suppresses expression of androgen inducible genes and induces expression of androgen repressed genes, but only when these human prostate cancer cells were cultured in a medium lacking androgen. The fact that Smo knockdown reduced expression of androgen regulated genes but did not affect expression of AR mRNA or protein suggests that some aspect of Hh signaling regulates the activity of the AR rather than its expression.

Since cyclopamine suppressed residual/reactivated androgen gene expression in androgen deprived and AI prostate cancer cells, we also sought evidence that this effect had biological consequences relevant to AI growth. First, we tested whether the presence of cyclopamine might prevent the development of AI cells from parental LNCaP cells chronically maintained in androgen depleted medium. LNCaP cells were seeded onto 10 plates at low density and then 5 plates each were switched to androgen depleted medium supplemented with vehicle (EtOH) or with 5 μM cyclopamine. The media were changed every 3 days. Within 2 months, cell numbers in the cyclopamine-treated cultures were significantly reduced compared to vehicle-treated cultures and most surviving cells in the cyclopamine-treated cultures were shrunken with optically dense nuclei that contrasted with the neuroendocrine cell-like appearance of cells in vehicle-treated cultures (Figure 3A). By the third month, cyclopamine-treated cultures had less than 1% of the cells of vehicle-treated cultures and all remaining cells showed the presence of the optically dense nuclei. No cells remained on cyclopamine-treated plates by 4 months of culture but the cells in the vehicle-treated cultures were increasing in numbers by this time and these cultures gave rise to growing lawns of cells by 6 months that typify AI growth. For overt LNCaP-AI cells, we found that treatment with 5 μM cyclopamine significantly inhibited their growth over a 10 day period (Figure 3B) but when cyclopamine treatment was accompanied by supplemental androgen (10 pM R1881), the growth rate of these cells was no different than vehicle treated cells. This indicates that the presence of androgen can overcome the growth-inhibiting effects of cyclopamine on overt AI cells.

Finally, we sought to test whether overexpression of Gli1 or Gli2, transcription factors that lie at the endpoint of the Hh signaling process, might act oppositely to Smo antagonism/inhibition to increase androgen signaling or AI growth when LNCaP cells were grown in androgen free medium. Parental LNCaP cells were transduced with a Gli1- or Gli2- (Gli2ΔN) expressing lentivirus and these cells were compared to control cells transduced with empty virus to determine the effects of Gli overexpression on androgen regulated gene expression and cell growth in androgen depleted medium. The Gli overexpressing variants of LNCaP were also found to express significantly higher levels of KLK2 or KLK3 when compared to control (vector transduced) cells in androgen depleted medium (Figure 4A). Gli1 or Gli2 overexpressing LNCaP cells also expressed significantly higher levels of luciferase reporter from both AR and Gli dependent promoters compared to control cells (Figure 4B). Despite higher basal expression of androgen regulated genes, the Gli transduced cells expressed AR protein at equivalent levels to the control cells (Figure 4C) so here again, these effects appear to be independent of changes in AR expression. The Gli transduced LNCaP cells also showed significant increased growth in androgen depleted medium compared to the control cells (Figure 4D), though Gli2 cells appeared to be more robust than Gli1 in this regard. Regardless of this differential hierarchy, this data shows that Gli function supports androgen regulated gene expression in a low androgen environment as well as AI growth.

The evidence that Gli1 or Gli2 overexpression upregulates androgen inducible gene expression and AI growth of androgen deprived LNCaP cells without affecting AR expression suggests that some function of the Gli proteins may support AR transcriptional activity in a low androgen environment. We tested for some potential direct interaction between these Gli and AR proteins in co-immunoprecipitation experiments. Human 293FT cells were transfected with an expression plasmid for full-length human AR, myc-tagged Gli2 or a combination of these plasmids. Forty-eight hrs later, extracts from the cells were immunoprecipitated with anti-AR or anti-myc antibody and the immunoprecipitates (IPs) were analyzed by Western blot for the presence of AR or myc-tagged Gli2. When the Western blot was probed with anti-AR, we found that AR co-immunoprecipitated with myc-tagged Gli2 only in extracts from cells co-transfected with both plasmids (Figure 5). Similarly, myc-tagged Gli2 was co-immunoprecipitated in the AR IPs from extracts of cells co-transfected with both plasmids (Figure 5). This apparent interaction between Gli2 and AR in the 293FT cells was not diminished by supplementation with 1 nM R1881.

Here we provided evidence that aspects of the canonical Hh signaling pathway can play a role in supporting residual/reactivated androgen signaling in androgen deprived and AI prostate cancer cells and this finding has important implications with regards to both the mechanistic basis for AI growth in the castration resistant prostate cell and for treatment strategies for CRPC in patients. Smo inactivation by cyclopamine, a cyclopamine variant drug (KAAD-cyclopamine) or reduction in Smo expression by siRNA downregulated androgen inducible genes in androgen deprived and AI prostate cancer cells and these findings suggest that some action of Smo might be important for reactivation of androgen signaling under low androgen conditions. The effects of cyclopamine on androgen regulated genes was common to several types of human prostate cancer cell lines grown under androgen deprived conditions so the effect was not limited to LNCaP cells and derivatives. Cyclopamine also suppressed expression of reporter genes from two different androgen responsive promoters in LNCaP cells in androgen depleted medium and these findings support the idea that Smo activity supports AR-mediated transcriptional activity in the androgen deprived state. Finally, the modulatory effects of cyclopamine on AR regulated gene expression were consistent with the effect of this drug on AI growth. Chronic cyclopamine treatment prevented the development of androgen growth independent cells from parental androgen growth-dependent LNCaP cells and significantly inhibited the growth of an overt AI variant of LNCaP. The cyclopamine-mediated growth suppression was reversed by returning a low level of androgen to the cells, providing further evidence that effects of cyclopamine on development and growth of AI cells are based upon cyclopamines' actions on residual androgen signaling.

Smo action ultimately drives transcription by Gli family proteins so we also tested whether exogenous expression of active Gli had opposite effects of cyclopamine or Smo reduction. Here, our findings that Gli1 or Gli2 overexpression enhanced androgen regulated gene expression in androgen depleted medium and enabled AI growth for androgen growth-dependent cells strongly argues that the active Gli proteins resulting from Hh signaling play the most critical role in Hh-support of residual/reactivated androgen signaling regulation. Although the Gli2 overexpressing LNCaP cells exhibited more robust androgen independent growth than the Gli1 overexpressing cells, it is not possible to rank the effectiveness of the Gli proteins on growth control from this study since the cells may be expressing different amounts of transcriptionally active Gli protein. However, the recent report that Gli2 protein was abundantly expressed in tumor cells from patients with AI (CRPC) prostate cancer 40 does provide further support for the idea that Gli2 protein expression might have a specific role in AI cancer cell growth in CRPC patients and Gli2 may be the preferred target for CRPC treatments.

With regards to the potential mechanism(s) through which Hh/Gli cross-talks to the androgen signaling pathway, it does not appear to involve changes in the expression of AR mRNA or protein as this was not affected by cyclopamine, Smo knockdown or Gli overexpression. However, the finding that Gli2 and AR proteins co-immunoprecipitate when they were co-expressed in 293T cells does suggest that Gli2 might directly interact with AR to influence the expression of AR target genes in the same manner that other co-activator proteins support AR function 41. Previously Gli2 was shown capable of binding to CREB or to Zic family transcription factor proteins 4243 so this finding extends the potential repertoire of transcription factors capable of interacting with Gli2. It is of further interest that the interaction between AR and Gli2 proteins was not diminished by androgen supplementation. Therefore, the lack of effects of cyclopamine on androgen regulated gene expression in androgen supplemented LNCaP cells might be due to some additional role of other upstream elements of the Hh signaling pathway that are only manifest in androgen depleted cells. Additionally, we must consider the possibility that Hh/Gli signaling is involved in the endogenous production of androgen (intracrine androgen biosynthesis) that is reportedly associated with AI prostate cancer cells 11, especially since Hh signaling is required for steroidogensis in the testis and for androgen production by other types of cells 4445. This is an aspect that we will test for in future experiments.

Regardless of the mechanism(s) involved, the outcome of this research suggests that Hh/Gli inhibitors offer a specific means to target reactivated androgen signaling in CRPC and to test the idea that inhibition of anomalous androgen signaling in CRPC cells has therapeutic benefit for patients. Although cyclopamine is difficult to use as a therapeutic agent, several pharmaceutical companies are in the process of developing similar drugs that are easier to use in the clinical setting and some of these drugs are through Phase I testing 46. Therefore, translation of these experimental studies to patients should be able to proceed fairly rapidly. Alternatively, there are non- canonical signaling pathways that increase Gli activity in cancer cells 47 so a clinical focus on Smo antagonists may not be sufficient to deal with all forms of CRPC. Reports of small molecular inhibitors of Hh/Gli signaling that act independently of Smo antagonism 48, suggests that Hh/Gli signaling provides a rich array of targets for the development of more effective treatments for CRPC.

Modulation of Hh signaling in prostate cancer cells by reduction of Smo expression or activity or by overexpression of active Gli proteins affected androgen signaling and the expression of androgen regulated genes in these cells but only when they were cultured in a low androgen medium. The effects of Hh modulation on androgen regulated gene expression in prostate cancer cells were consistent with the coordinate effects on AI cancer cell development and growth in low androgen medium but these effects were reversed by the presence of androgens. Since we have found that Gli2 protein, at least, interacts with the AR protein, the mechanism through which Hh signaling affects AR-dependent gene expression and AI cell growth may involve a direct interaction of AR with Gli proteins.