|Home | About | Journals | Submit | Contact Us | Français|
Neuroendocrine (NE) phenotype, seen in >30% of prostate adenocarcinomas (PCa), and NE prostate tumors are implicated in aggressive prostate cancer. Formation of NE prostate tumors in the TRAMP mouse model was suppressed in mice lacking the ubiquitin ligase Siah2, which regulates HIF-1α availability. Cooperation between HIF-1α and FoxA2, a transcription factor expressed in NE tissue, promotes recruitment of p300 to transactivate select HIF-regulated genes, Hes6, Sox9 and Jmjd1a. These HIF-regulated genes are highly expressed in metastatic PCa and required for hypoxia-mediated NE phenotype, metastasis in PCa and the formation of NE tumors. Tissue-specific expression of FoxA2 combined with Siah2-dependent HIF-1α availability enables a transcriptional program required for NE prostate tumor development and NE phenotype in PCa.
Prostate adenocarcinomas (PCa) with neuroendocrine (NE) phenotype and NE prostate tumors are associated with poor prognosis and androgen independence. Here we demonstrate that formation of NE tumors and metastasis of PCa require the ubiquitin ligase Siah2. Through its role in the control of HIF-1α availability, Siah2 enables cooperation between HIF-1α and the NE-specific transcription factor FoxA2. Genes induced by HIF-1α/FoxA2 cooperation are expressed in NE lesions and in metastatic human PCa and required for formation of the NE phenotype, for metastasis of human PCa, and for the development of NE tumors. Tissue-specific cooperation between transcription factors that promote NE phenotype and prostate tumor development offers a paradigm for the development, progression and potential targeting of aggressive prostate tumors.
Prostate cancer is the second leading cause of cancer deaths among men in Western nations. Among the metastatic forms of prostate adenocarcinoma (PCa) are those that express neuroendocrine (NE) markers, often referred to as neuroendocrine differentiation (NED) or NE phenotype. NED is seen in over 30% of PCa and is associated with poor prognosis and androgen independence (Cindolo et al., 2007). A small percentage (0.5–2%) of human prostate tumors develops as highly aggressive NE tumors, which have a 35% survival rate in 2 years (Cindolo et al., 2007; Sella et al., 2000). Factors implicated in NED of LNCaP prostate cancer cells in vitro include IL-6 treatment (Deeble et al., 2001), androgen removal (Yuan et al., 2006) and ionizing radiation (Deng et al., 2008). In transgenic animals, T antigen expression or inactivation of p53 and Rb has been associated with prostate NE tumors (Huss et al., 2007; Zhou et al., 2006).
FoxA2, a member of the FoxA subfamily of forkhead box transcription factor, is expressed in mouse prostate NE carcinomas (Chiaverotti et al., 2008; Mirosevich et al., 2006) and NE foci of human PCa (Mirosevich et al., 2006). HIF-1α, the master regulator of the hypoxia response, is also expressed in NE tumors (Monsef et al., 2007). HIF-1α is regulated under normoxia by the E3 ligase pVHL and is also regulated under mild hypoxia (2–5% O2) by the ubiquitin ligase Siah2. Siah2 controls prolyl hydroxylase 1/3 (PHD) stability (Nakayama et al., 2004), thereby affecting PHD availability to modify HIF-1α, which is essential for HIF-1α’s association with and ubiquitination by pVHL (Ivan et al. 2001). Given that PHD function as cellular oxygen sensors (Aragones et al., 2009; Nakayama et al., 2009), Siah2 is expected to play a central role in controling hypoxia and related biological outcomes, including tumorigenesis and metastasis (Nakayama et al., 2009). Indeed, inhibition of Siah2 activity blocks formation of tumors (Ahmed et al., 2008; Moller et al., 2009; Qi et al., 2008; Schmidt et al., 2007). Further, Siah2’s contribution to melanoma metastasis is HIF-dependent (Qi et al., 2008).
Once stabilized, HIF-1α translocates to the nucleus and dimerizes with HIF-1β, the heterodimers then bind to hypoxia responsive elements (HREs) to regulate transcription of hypoxia-responsive genes (Semenza, 2003). Several transcription factors cooperate with HIF to regulate its transcriptional activity, HIF activity is enhanced by β-catenin (Kaidi et al., 2007) and repressed by FOXO3a (Emerling et al., 2008). HIF can also modulate activity of other transcriptional regulators: HIF-1α potentiates Notch signaling (Gustafsson et al., 2005) and represses c-Myc activity (Gordan et al., 2007).
Given the role of Siah2 in regulation of HIF-1α, we set to determine its role in prostate cancer development and metastasis.
We employed the TRAMP mouse model, in which prostate-specific expression of SV40 T antigen results in prostate tumors that metastasize to lymph nodes, lung and liver (Gingrich et al., 1996), to assess the possible role of Siah in tumor growth and metastasis. Analysis of 8-month-old TRAMP mice with different Siah2 background (TRAMP/Siah2−/−, TRAMP/Siah2+/−, and TRAMP/Siah2+/+) revealed that the majority developed prostate masses (Figure 1A). Most primary masses were composed of benign proliferations of stroma and epithelium with atypical epithelial hyperplasia (AH) in TRAMP/Siah2−/− mice, compared with a preponderance of NE carcinoma in the TRAMP/Siah2+/− and TRAMP/Siah2+/+ mice (Figure 1B, 1C). Although TRAMP AH have been referred to as adenocarcinoma in some literature (e.g., Gingrich et al., 1996), we refer them as TRAMP AH in lieu of clear discrimination, detailed in Chiaverotti et al., (2008). NE carcinomas were identified by morphology (Figure 1B) and expression of the wellestablished NE markers synaptophysin, neuron-specific enolase (NSE) and FoxA2 (Figure 1D), consistent with the notion that the TRAMP tumors are primarily of NE origin (Chiaverotti et al., 2008; Mirosevich et al., 2006). In clear contrast, FoxA2, synaptophysin and NSE were undetectable in normal prostate glands (Figure 1D, arrows) or in AH (data not shown).
Since Siah1 also contributes to the regulation of PHD and consequently HIF availability (Nakayama et al., 2004; Qi et al., 2008), we also evaluated Siah1a function in prostate tumor formation. Siah2 and Siah1a doubly homozygous mutant mice are non-viable (Frew et al., 2003). We therefore established a TRAMP/Siah1a+/−Siah2−/− mouse line. TRAMP/Siah1a+/−Siah2−/− mice showed a prostate tumor incidence similar to that seen in TRAMP/Siah2−/− mice (Figure 1A) but lacked NE tumors (Figure 1C).
In the TRAMP mice, the early NE tumor lesions develop in the ventral prostate after 3 months and are recognized as foci that express NE markers. To examine the effect of Siah2 on the development of early stage NE tumors, we analyzed 5-month-old mice. NE foci were identified in 3/6 control mice but in none of the 10 TRAMP/Siah2−/− mice (Figure 1E), the difference is statisticaly significant. These data reveal that in the TRAMP model Siah is required for development of NE carcinomas.
AH predominantly develops in the dorsal lateral lobe of prostate and is identifiable in 1-month-old TRAMP mice (Chiaverotti et al., 2008). To evaluate whether Siah also involved in the progression of AH, we analyzed the dorsal prostate lobes from 1- and 3- month-old mice and found that lack of Siah2 delayed the progression from a normal prostate gland to the early and medium stages AH in 1-month-old mice and into the late stage AH in the 3-month-old mice (Figure S1A, S1B).
Consistent with Siah2 regulates HIF-1α availability, HIF-1α level is reduced in the very few NE carcinomas observed (Figure 1D) and in AH (Figure S1C) from TRAMP/Siah2−/− mice compared with TRAMP/Siah2+/− mice. HIF-2α staining was also reduced in AH from TRAMP/Siah2−/− mice (Figure S1C), whereas HIF-2α was undetectable in NE carcinoma from any genotype (data not shown). These findings are also consistent with the observation that HIF-1α but not HIF-2α is co-expressed with NE markers in prostate cancers (Monsef et al., 2007).
Potential changes in proliferation, apoptosis and angiogenesis were evaluated using PCNA, TUNEL/active caspase-3 and CD31, respectively. NE tumors, but not AH, from TRAMP/Siah2−/−mice showed reduced cell proliferation and increased apoptosis compared with those seen in TRAMP/Siah2+/− mice (Figure 1C, 1D, Table S1). Vascular density was similar in the two strains in both the NE tumor (Figure 1F, Table S1) and AH (Figure 1C, Table S1). Hence, loss of Siah and consequent downregulation of HIF levels appear to specifically govern proliferation and cell survival of NE tumors.
To directly assess a possible role for Siah2 and HIF-1α in tumorigenesis of TRAMP cells, we analyzed TRAMP-C cells derived from TRAMP tumors (Foster et al., 1997). These cells likely represent NE tumor-derived cells as they express multiple NE specific transcription factors (Figure 4A, and data not shown). The PHYL peptide binds to Siah’s substrate recognition site (House et al., 2003) and attenuates Siah2’s effect on PHD1/3, thus reduces HIF-1α levels under hypoxia (Moller et al., 2009; Qi et al., 2008). Expression of the PHYL peptide in TRAMP-C cells effectively abolished their ability to form tumors (Figure 1H, 1I), consistent with the finding that NE tumors do not form in TRAMP/Siah1a+/−Siah2−/− mice (Figure 1C). Forced HIF-1α expression in TRAMP-C cells expressing PHYL peptide by transfection (Figure 1G) partially recovered their ability to form tumors (Figure 1H, 1I). These data support the role of Siah2, in part through its regulation of HIF-1α levels, in formation of NE prostate tumors.
Metastatic lesions in liver, lung and lymph nodes of TRAMP mice were identified as NE carcinomas, based on morphology (Figure 2A) and FoxA2/synaptophysin expression (Figure 2B). However, both the frequency and size of metastatic lesions were significantly reduced (6 fold) in the lung and were not found in liver and lymph nodes of TRAMP/Siah2−/− mice, compared with TRAMP/Siah2+/− or TRAMP/Siah2+/+ animals (Figure 2A, 2C). Furthermore, we found no metastases in TRAMP/Siah1a+/−Siah2−/− mice (Figure 2C). The very few lung metastases observed in TRAMP/Siah2−/− mice were smaller, showed reduced cell proliferation and enhanced cell death (Figure 2D, 2E). These findings point to the role of Siah2 in TRAMP tumor metastasis.
Since the NE-specific transcription factor FoxA2 is co-expressed with HIF-1α protein in nuclei of NE carcinoma cells (Figure 1D), we tested the possibility that these transcription factors cooperate with each other. While expression of exogenous FoxA2 in TRAMP-C cells did not alter the expression of an HRE-linked luciferase construct (HRE-Luc, data not shown), expression of exogenous HIF-1α alone elicited a modest increase in HRE-Luc activity as expected (Figure 3A). Significantly, co-expression of HIF-1α and FoxA2 led to a 6-fold increase in luciferase activity over expression of HIF-1α alone (Figure 3A). Co-expression of IPAS, a spliced form of HIF-3α with potent dominant-negative activity towards all HIFs (Makino et al., 2002) or inhibiting HIF-1α expression by expressing S2RM (a dominant-negative form of Siah2) or PHYL abolished FoxA2-potentiated HRE-Luc activity (Figure 3A). Importantly, knockdown of FoxA2 caused an approximately 40% reduction in HRE-Luc activity under hypoxia, although the degree of inhibition was lower than that seen using HIF-1α or HIF-1β siRNA (Figure 3B, Figure S2A). These data establish a role for FoxA2 in enhancing HIF-mediated transcriptional activity. In contrast, HIF-1α did not stimulate FoxA2 transcription activity (data not shown), indicating that HIF/FoxA2 transcriptional synergy may be restricted to the context of an HRE. Notably, FoxA1 did not increase HRE-Luc activity in the presence of HIF-1α (data not shown).
We then mapped the FoxA2 domains required for cooperation with HIF-1α. FoxA2 fragments consisting of the N-terminal transactivation domain (N-TAD), the central forkhead domain, or the C-terminal transactivation domain (C-TAD) were generated and evaluated for their effect on FOXA- and HIF-1α-dependent transcription activity (Figure S2B). FoxA2 mutants lacking either the N-TAD or C-TAD exhibited FOXA-dependent transcriptional activity similar to that of the wild-type FoxA2 (Figure 3C), indicating that one transactivation domain is sufficient for FoxA2 transcriptional activity. FoxA2 mutants lacking the C-TAD promoted HIF-dependent transcriptional activity to a level similar to wild-type FoxA2, whereas the FoxA2 mutant lacking the N-TAD was much less effective (Figure 3D). The FoxA2 C-terminus, which contains intrinsic chromatin remodeling activity (Cirillo et al., 2002), was dispensable for HRE activation (construct m-2 in Figure 3D, Figure S2B) thereby excluding the role of chromatin remodeling activity in FoxA2 cooperation with HIF. These data suggest that the N-TAD and forkhead domains of FoxA2 are required to stimulate HIF-mediated transcriptional activity.
Co-expression of HA-FoxA2 or HA-FoxA1 with Flag-HIF-1α in 293T cells followed by immunoprecipitation of Flag-HIF-1α identified HA-FoxA2,but not HA-FoxA1, as an HIFassociated protein (Figure 3E), consistent with the effect of FoxA2 but not FoxA1 on HIF-dependent transcriptional activity (data not shown). Importantly, endogenous FoxA2 was co-precipitated with endogenous HIF-1α (Figure 3F) or HIF-1β (Figure 3G) in TRAMP-C cells maintained under hypoxia. The association of HIF-1β with FoxA2 was abolished following HIF-1α knockdown (Figure 3G), indicating that HIF-1α recruits FoxA2 to the HIF complex under hypoxia. Furthermore, in vitro binding between purified His-FoxA2 and in vitro translated Flag-HIF-1α confirmed their direct interaction (Figure 3H).
In vitro binding using truncation mutants of HIF-1α (Figure S2C) and FoxA2 (Figure S2B) identified the bHLH-PAS domain of HIF-1α (Figure 3H) and the N-TAD of FoxA2 (Figure 3I) as the minimal regions mediating the HIF-1α-FoxA2 interaction. Although the bHLH-PAS domain of HIF-1α was found to interact with FoxA2 and the PAS domain of HIF-1α is known to interact with HIF-1β (Erbel et al., 2003), alternation of the FoxA2 level in vitro or in TRAMP-C cells, had no apparent effect on the interaction of HIF-1β with HIF-1α (Figure 3J and 3K).
Transcript levels of the HIF targets VEGFA and Glut-1 were not altered following FoxA2 expression or co-expression of HIF-1α and FoxA2 in TRAMP-C, PC3, and HeLa cells (data not shown) suggesting that FoxA2/HIF-1α cooperation selectively affects HIF-regulated genes. We thus compared gene expression profiles of TRAMP-C cells transfected with pcDNA control vector, or expression vectors encoding FoxA2, HIF-1α, or HIF-1α plus FoxA2. Approximately 140 genes in TRAMP-C cells were upregulated by hypoxia (Table S2). Comparison of genes expressed in each condition identified 47 genes upregulated in the HIF-1α + FoxA2 group under hypoxia, in comparison with the other three groups (Table S3). Of these and other hypoxia-induced genes, we further assessed 30 genes for FoxA2-dependent expression, selected based on prostate- and NE tumor-specific expression. To confirm FoxA2-dependent transcription of this gene set we monitored changes in their expression in TRAMP-C cells expressing FoxA2 siRNA. qRT-PCR analysis of selected genes identified Hes6, Sox9, Jmjd1a and Plod2 among those that displayed FoxA2-dependent transcription under hypoxia (Figure 4A). Sox9, Jmjd1a and Plod2 are known HIF target genes (Amarilio et al., 2007; Beyer et al., 2008; Hofbauer et al., 2003). In contrast, transcription of Glut-1 and VEGFA was not altered following FoxA2 knockdown (Figure 4A). Overall, these results suggest that in TRAMP-C cells FoxA2 cooperates with HIF-1α to regulate a subset of HIF targets under hypoxia.
The transcription factor Hes6 is reportedly highly upregulated in human metastatic prostate cancers with a NE phenotype (Vias et al., 2008). Upregulation of Hes6 transcripts under hypoxia suggests that Hes6 is a HIF target gene. Indeed, knockdown of HIF-1α or HIF-1β reduced hypoxia-induced Hes6 transcription (Figure 4B). We identified 3 potential HREs and cloned the corresponding 1.25 kb of the mouse Hes6 promoter region upstream of a luciferase reporter. This construct was activated 2.5 fold by the hypoxia mimic DMOG (Figure S3A) and 1.8 fold by hypoxia (data not shown). Deletion analysis of the Hes6 promoter revealed that the HRE at −66bp was required for hypoxia-induced reporter activity (Figure S3A). Mutation of this HRE in the full-length 1.25 kb fragment resulted in loss of the response to DMOG (Figure 4C). FoxA2 knockdown repressed Hes6 transcription (Figure 4A, 4B), while FoxA2 overexpression increased Hes6 promoter activity following addition of DMOG, an effect not seen using the HRE-mutant promoter construct (Figure 4D). These observations strongly suggest that FoxA2-induced Hes6 promoter activity requires an HRE and thus HIF activity. Chromatin immunoprecipitation (ChIP) confirmed that HIF-1α and HIF-1β bound the −66 bp HRE but not the other two putative HREs in the HES6 promoter under hypoxia (Figure S3B). ChIP analysis confirmed the binding of FoxA2 to the same −66 bp HRE, which was impaired following HIF-1α knockdown (Figure S3C), suggesting that FoxA2 is recruited to the promoter through HIF-1α. Similarly, FoxA2 bound to HRE-containing promoter regions of Sox9 (Figure S3C) and Jmdj1a (data not shown) in a HIF-1α-dependent manner. These results indicate that FoxA2 regulates a subset of HIF target genes possibly through direct binding to HIF.
To analyze mechanisms underlying selectivity of FoxA2 cooperation with HIF-1α, we investigated whether FoxA2 expression changed HIF-1α levels, its asparagine hydroxylation, or its binding to the Hes6 promoter but found none (data not shown). To directly assess the contribution of FoxA binding sites to FoxA2/HIF-1α cooperation, we determined possible changes in FoxA2/HIF-1α-mediated transactivation using Hes6 promoter mutants (Figure S3D). While a single FoxA site was sufficient to retain full responsiveness to FoxA2/HIF-1α cooperation, a mutation within this element attenuated such cooperation (Figure S3D). Notably, analysis of the effect of FoxA2 on recruitment of p300, a co-activator of HIF transcriptional activity (Arany et al., 1996), revealed that binding of p300 to the Hes6 promoter was attenuated following FoxA2 knockdown (Figure 4E). FoxA2 knockdown also reduced the binding of p300 to HRE-containing promoters of Jmjd1a and VEGFA in both TRAMP-C cells and Rv1 cells (Figure 4F). However, p300 knockdown only reduced the transcript levels of Hes6 and Jmjd1a but not VEGFA (Figure 4G), which resembled changes seen upon knockdown of FoxA2 (Figure 4A) on the hypoxia-induced transcription of these genes. These results suggest that p300 serves a distinct FoxA2-dependent HIF transcriptional program. Consistent with this possibility, HIFinduced promoter activity of Hes6 was more sensitive to increased p300 levels than that of VEGFA (Figure 4H). Notably, p300, but not p300ΔCH1 (a p300 mutant that cannot interact with HIF-1α), could further increase the degree of HIF/FoxA2 effect on Hes6 promoter activity (Figure 4I). In contrast, FoxA2 and p300 showed no apparent effect on HIF-dependent activation of the VEGFA promoter (Figure 4J). These results suggest that recruitment of p300 is, in part, responsible for the degree and the selectivity of the transcriptional program elicited by HIF/FoxA2 cooperation.
In vitro protein binding showed that FoxA2 enhanced HIF-1α/p300 interaction but had no effect on p300 binding to the HIF-1α 531–822 mutant that cannot associate with FoxA2 (Figures S2C, ,3H,3H, ,4K).4K). p300 lacking the CH1 domain neither interacted with the HIF/FoxA2 complex (Figure 4L) nor enhanced Hes6 transcription by HIF/FoxA2 (Figure 4I). These results suggest that FoxA2 promotes HIF and p300 interaction unlikely require HIF-1α N-TAD and p300 CH3 domain (Ruas et al., 2010).
To determine whether genes regulated by HIF/FoxA2 cooperation are important for tumorigenesis we employed retroviral vectors to re-express Hes6, Sox9 and Jmjd1a individually or in combination in TRAMP-C cells stably expressing PHYL or FoxA2 shRNA. While the endogenous expression of these genes was attenuated upon expression of PHYL peptide or FoxA2 shRNA, their transcript levels were restored following ectopic expression (Figure S4A, S4B). TRAMP-C cells expressing control (pBabe vector) or NxN, a mutant PHYL that cannot interact with Siah, but not those expressing PHYL or shFoxA2 could form colonies on soft agar (Figure 5A, 5B). Significantly, co-expression of Hes6, Sox9 and Jmjd1a effectively rescued the ability of TRAMP-C cells expressing either PHYL or shFoxA2 to form colonies in soft agar but re-expression of each individually was insufficient (Figure 5A, 5B). Consistent with the effect of PHYL being Siah-dependent, reduction of Siah1a and Siah2 (Figure S4C) reduced levels of HIF-1α protein (Figure S4D) and Hes6, Sox9 and Jmjd1a transcripts (Figure S4E), and attenuated the ability of these cells to form colonies in soft agar under hypoxia (Figure S4F). The reduced ability to form colonies in soft agar could be partially rescued upon re-expression of Hes6, Sox9 and Jmjd1a (Figure S5F).
Similar to the findings in culture, control but not the PHYL-expressing TRAMP-C cells were able to form tumors upon injection into the prostate of mice (Figure 5C, 5D, 5E). Whereas re-expression of Hes6, Sox9 or Jmjd1a individually in PHYL-expressing TRAMP-C cells failed to rescue tumorigenesis, expression of all three was able to restore (4/5 mice) tumorigenicity and partially rescue tumor growth (30% of tumor size) (Figure 5C, 5D, 5E). Orthotopic injection of shFoxA2-expressing TRAMP-C cells into the prostate also showed significant reduction in tumor formation, which was almost fully rescued by co-expression of Hes6, Sox9 and Jmjd1a (Figure 5F, 5G). These results further establish the importance of HIF/FoxA2 target genes, Hes6, Sox9 and Jmjd1a, in the development of NE prostate tumor.
We next assessed the importance of the pathway discovered using the TRAMP model in human PCa. To this end we selected CWR22Rv1 cells (Rv1), which were derived from a human prostate adenocarcinoma xenograft displaying an NE phenotype (Huss et al., 2004; Sramkoski et al., 1999). Rv1 cells grown under hypoxia showed upregulation of NE markers NSE and chromogranin B (ChgB) (Figure 6A, 6B) and demonstrated protrusion of neurite-like structures from cells located at the periphery of colonies (Figure 6C). Concomitant with the induction of NE phenotype was the upregulation of Hes6, Sox9 and Jmjd1a transcripts (Figure 6D). Significantly, inhibition of Siah or knockdown of FoxA2 attenuated hypoxia-induced upregulation of Hes6, Sox9 and Jmjd1a (Figure S5A, S5B). To determine whether Hes6, Sox9 and Jmjd1a are important for hypoxia-induced NE phenotype, we re-expressed these genes individually or in combination in PHYL- or shFoxA-expressing Rv1 cells (Figure S5A, S5B) and found only co-expression of all three could partially restore hypoxia-induced NSE upregulation and formation of neurite-like structures (Figure 6E, 6F, 6G). These results point to a requirement for Siah-HIF/FoxA2 regulated genes in the hypoxia-induced NE phenotype of human prostate adenocarcinoma cells.
To evaluate the biological significance of NE phenotype for human prostate cancer in vivo, we injected the above Rv1 transfectants into the prostate of nude mice. Surprisingly, unlike TRAMP-C cells, Rv1 cells expressing PHYL only showed about 30% reduction in the tumor size which could not be rescued by co-expression of Hes6, Sox9 and Jmjd1a (Figure 6H). Similarly, Rv1 cells expressing shFoxA2 did not exhibit any notable decrease in tumor formation (Figure 6H). These results indicate that Hes6, Sox9 and Jmjd1a are not involved in tumorigenesis of Rv1 cells. The 30% reduction of tumor size by PHYL was correlated with a 35% reduction in the tumor vessel density (Figure S5C, Table S4) and 40% reduction of VEGFA transcript in PHYL-expressing Rv1 cells in vitro (Figure S5D). Similarly, colony formation in the soft agar assay was comparable for PHYL-expressing Rv1 cells and control cells (Figure S5E). Significantly, the circulating PHYL- or shFoxA2-expressing Rv1 cells in the blood were reduced, indicating impaired intravasation of these cells, which was partially rescued by co-expression of Hes6, Sox9 and Jmjd1a (Figure 6I). Although Rv1 cells exhibited limited ability to form liver metastasis (2 out of 10 mice injected with Rv1 cells), they were very efficient in periaortic lymph node metastases. Expression of PHYL or shFoxA2 abolished lymph node metastases of Rv1 cells, which could be largely restored by co-expression of Hes6, Sox9 and Jmjd1a (Figure 6J). These findings suggest that genes co-regulated by HIF and FoxA2 play a key role in metastasis of prostate adenocarcinoma cells.
IHC analyses of Rv1 orthotopic tumors revealed that HIF-1α and NSE staining were concentrated around the necrotic regions, which are known to be highly hypoxic, whereas FoxA2 staining was evenly distributed (Figure 6K). As expected, expression of PHYL resulted in reduced HIF-1α staining and loss of NSE staining in the hypoxic regions (Figure 6K). FoxA2 shRNA reduced the NSE staining in the hypoxic regions without affecting HIF-1α levels (Figure 6K), consistent with our in vitro data (Figure 6E, 6F, 6G). Importantly, consistent with our in vitro results, co-expression of Hes6, Sox9 and Jmjd1a in PHYL- or shFoxA2-expressing Rv1 cells restored the NSE staining in the hypoxic regions (Figure 6A to 6G). Strong NSE staining was primarily seen within the more hypoxic regions, proximal to the necrosis, of the primary tumor (Figure 6K) and metastatic lesions in liver and lymph nodes (Figure 6L), implying that the NE-differentiated Rv1 cells may be responsible for the metastasis. These results establish that genes co-regulated by HIF and FoxA2 play a key role in hypoxia-induced NE phenotype of PCa in vitro and in vivo, and that NE phenotype is tightly associated with PCa metastasis.
We next asked whether FoxA2/HIF-1α transcriptional targets were expressed in NE tumors. The very few NE tumors identified in a TRAMP/Siah2−/− mice had lower transcript levels of Hes6, Sox9, Jmjd1a and Plod2 compared with TRAMP/Siah2+/−-derived NE tumors (Figure 7A), consistent with HIF-1α-dependent expression.
The NE phenotype seen in human PCa can be classified to three types based on IHC staining of NE markers (Cindolo et al., 2007; Hirano et al., 2005; Shimizu et al., 2007). Focal -where NE markers distinguish clusters of cells - found in low-grade and moderately differentiated PCa; general staining -where larger tumor areas are positive for NE markers - found in high-grade and poorly differentiated PCa; and single cells that are stained positively for NE markers (Hirano et al., 2005). We examined 15 cases of human PCa and found 10 to exhibit NE phenotype. Two of the 10 samples displayed focal staining of NE marker NSE, where FoxA2, Hes6 and Sox9 were concentrated in the NE foci (Figure 7B, Figure S6A). Eight of the 10 specimens showed a general staining of NSE, with co-staining of FoxA2, Hes6 and Sox9 in 5/8 cases. Co-expression of FoxA2, Hes6 and Sox9 found in 70% of PCa specimens with NE phenotype was statistically significant (p<0.05; Figure 7C).
We next performed IHC staining of a human prostate TMA consisting of 79 cases representing prostatic intraepithelial neoplasia (PIN) and different Gleason stages of PCa. A higher NSE staining was found in high-grade PCa (G4 and G5), which also showed increased staining of Siah2, FoxA2, Hes6 and Sox9, compared with low-grade tumors (PIN and G3) (Figure 7D). The difference in the expression of NSE, Siah2, FoxA2, Hes6 and Sox9 between high-grade and low-grade tumors is statistically significant and correlated with pathoclinical staging. In addition, IHC staining of 3 human prostate NE tumors revealed co-expression of synaptophysin, FoxA2, HIF-1α, Hes6 and Sox9 in all cases (Figure S6B). These findings suggest that NE-positive tumors are associated with the more malignant human prostate cancers, in which the HIF/FoxA2 target genes are expected to play a role in the development of the NE phenotype. In agreement, gene expression data from human prostate adenocarcinoma identified a marked increase in Hes6, Plod2, Jmjd1a and FoxA2 expression in metastatic prostate tumors, compared with primary prostate tumors and normal prostate tissues (Figure 7E). The expression of these genes correlates with the expression of NE markers, further illustrating the link between their expression, NED and metastatic prostate tumors (Figure 7E).
Whereas Siah2 staining was higher in high-grade PCa, HIF-1α staining was found in both low-grade and high-grade tumors (Figure 7D). Despite high level of HIF-1α expression in all grades of tumors, the level of Glut-1, a common readout for HIF transcriptional activity, was high only in high-grade PCa, pointing to a correlation between Siah2 expression and HIF-1α activity (Figure 7D). In agreement, gene expression analyses revealed an increased Siah2 transcript and enhanced HIF activity in metastatic PCa as reflected by increased transcript of HIF target genes such as CA9, VEGFA and Glut-1, compared with primary PCa (Figure 7E). These results substantiate the correlation between Siah2 expression and HIF activity in human PCa, consistent with the role of Siah in regulation of PHD3 and factor-inhibiting HIF-1 (FIH) stability, which control HIF-1α availability and activity (Nakayama et al., 2004; Fukuba et al., 2008). The cooperation between HIF and FoxA2 in determining NE phenotype can be attributed to a higher level of Siah2 which increases HIF stability and activity, and availability of FoxA2 in the high-grade PCa.
Our results provide insight into regulation and function of the FoxA2/HIF-1α complex in determining NE prostate tumor formation and NE phenotype, an important component of metastatic prostate adenocarcinomas. These results also point to a role for Siah2 in determining tumor differentiation. Siah2 loss has little effect on development and growth of the prostate luminal epithelium but decreases initiation of NE carcinomas and, consequently, the metastatic burden in the TRAMP model. We show that partial deletion of Siah1a on a Siah2-null background fully ablated NE tumor formation, suggesting that both Siah2 and Siah1 are required to enable the development of prostate NE tumors.
As HIF-1α is stabilized under hypoxia and FoxA2 is expressed in NE tissues, our findings suggest conditional and spatial cooperation between these two factors under specific tissue and oxygen requirements. Siah2-dependent regulation of HIF coupled with NE-specific expression of FoxA2 provides a framework for a specific tumor differentiation program associated with a highly metastatic phenotype. Among the four FoxA2/HIF targets identified in this study, Hes6 is reportedly highly upregulated in metastatic prostate cancers displaying NE markers (Vias et al., 2008) and Plod2 is overexpressed in metastatic prostate NE tumors (Shah et al., 2004). Sox9 expression is increased in relapsed androgen-refractory prostate cancers and is associated with enhanced growth, invasion and angiogenesis (Wang et al., 2008). It is noteworthy that Hes6, Sox9 and Jmjd1a have been shown to regulate differentiation of stem/progenitor cells (Eun et al., 2008; Loh et al., 2007; Nowak et al., 2008), raising the possibility that FoxA2/HIF cooperation initiates a transcriptional program that regulates neuroendocrine differentiation of normal and/or prostate cancer stem cells. In agreement, the NE phenotype of PCa is also associated with expression of the stem/progenitor markers, supporting the notion that the NE-like cells may harbor prostate cancer stem cells (Bonkhoff, 1998; Helpap et al., 1999; Sotomayor et al., 2008).
Several plausible mechanisms could underlie HIF-1α/FoxA2 transcriptional synergy, we ruled out that the intrinsic chromatin remodeling activity of FoxA2 is important (Cirillo et al., 2002) and that FoxA2 displaces HIF-1β. Our data implicate FoxA2 in enhanced recruitment of p300 for the selective activation of a subset of HIF-target genes. Consistent with our findings, MEFs from mice expressing a p300/CBP mutant that cannot interact with HIF exhibited attenuated HIF activity in luciferase reporter assays but they showed attenuated expression of only a small subset of HIF target genes, not including VEGFA and Glut-1 (Kasper et al., 2005).
The importance of Hes6, Sox9 and Jmjd1a for NE phenotype is demonstrated in human PCa cells that exhibit NE phenotype under hypoxia. Inhibition of these genes attenuates the NE phenotype and PCa metastasis, whereas their co-expression rescues the NE phenotype and metastasis even upon knockdown of FoxA2 or Siah2 - their upstream regulators. Notably, the requirement of hypoxia for NE phenotype was confirmed by IHC analyses of PCa in vivo where level of NSE coincided with that of HIF, FoxA2 and their regulated genes. Of equal significance, inhibition of mouse prostate tumor growth by attenuating Siah2 activity could be overcome upon co-expression of Hes6, Sox9 and Jmjd1a, further illustrating the importance of their cooperation for prostate tumor development.
Overall, our study offers a paradigm underlying formation of NE phenotype and NE prostate tumor development. This pathway requires the ubiquitin ligase Siah2, which determines the level and activity of HIF-1α, which then cooperates with the NE-specific transcription factor FoxA2. It is the conditional and spatial regulation of these factors that transactivates a subset of genes critical for NE phenotype and metastasis of PCa as well as for development of NE prostate tumors. Common to NE prostate tumors and PCa harboring the NE phenotype is their strong propensity to metastasize, and the poor outcome associated with these more aggressive forms of prostate cancer. Our findings unveil mechanisms underlying their development and progression, respectively, while identifying possible targets for therapy and markers for improved detection and monitoring of these tumors.
Prostate tumor samples representing NE tumors and/or prostate adenocarcinomas were obtained as part of approved clinical studies from the University of California Davis (IRB # 200312072), University of California Irvine (SPECS project, IRB # 20005-4806) and Northwestern University (SPORE tissue banking protocol, IRB # NCI01×2: STU00009126). In all cases informed consent was obtained from all subjects.
All animals were housed in the Sanford-Burnham Institute’s animal facility and the experiments with live animals were approved by our institute animal committee (IACUC # 04–135, 04–141, 07–132) and conducted following the Institute’s animal policy in accordance with NIH guidelines.
TRAMP-C cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 5% Nu-serum IV, 5% fetal bovine serum (FBS), 5 µg/ml insulin and antibiotics. CWR22 Rv1 cells were maintained in RPMI1640 medium with 5% FBS and antibiotics.
Siah2+/− mice (129 SVJ strain) were crossed with TRAMP transgenic mice (C57/Bl6 strain) to obtain Siah2 heterozygotes carrying the TRAMP transgene (C57/Bl6 and 129 SVJ mixed strain). Female Siah2+/−/TRAMP mice were crossed with male Siah2+/− mice to generate male TRAMP mice of three genotypes (Siah2 +/+, +/−, −/−), which were predominantly of the 129 strain. Female Siah2+/−/TRAMP mice were also crossed with male Siah1a+/−Siah2+/− mice to generate male TRAMP mice with a Siah1a+/−Siah2−/− genotype. Siah/TRAMP mice were analyzed at 8 months of age.
Antibodies to HIF-1α, HIF-2α and Sox9 (NOVUS), to Hes6, Jmjd1a, p300 and NSE (Abcam), to synaptophysin (BD Bioscience), to FoxA2, HIF-1β and CD31, Chromogranin B (Santa Cruz), to active caspase-3 (Chemicon), to PCNA (Cell Signaling), to FLAG, HA, α-tubulin and β-actin (Sigma) were used according to manufacturers’ recommendations. An ApopTag peroxidase in situ apoptosis kit was obtained from Chemicon
Student’s t-test or Fisher’s exact test was used for the statistical analyses.
We thank members of the Ronai Lab for helpful discussions. We thank Drs. Lorenz Poellinger, Hueng-Sik Choi, Wei Gu, Andreas Möller, Collin House, Robert Abraham, Gary Chiang, Norman Greenberg, James Jacobberger, Marja Nevalainen, for reagents, Jeremy Mathews for preparation of the prostate tumor TMA, Ling Wang for help with intraprostatic injection, Joan Massague for protocol of retroviral infections. Support by NCI grant CA111515 (to Z.R.), P50CA090386 (K.K.), and U01CA114810 (to D.M.) is gratefully acknowledged. J.Q. was supported by a CIHR fellowship.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Accession number. Microarray data was deposited in the GEO database (GSE18478).
Jianfei Qi, Signal Transduction Program, Sanford-Burnham Medical Research Institute, La Jolla, CA, 92037, USA.
Koh Nakayama, MTT program, Medical Research Institute, Tokyo Medical and Dental University, Tokyo 113-8510, Japan.
Robert D. Cardiff, Center for Comparative Medicine and Department of Pathology, School of Medicine, University of California, Davis, CA 95616.
Alexander D. Borowsky, Center for Comparative Medicine and Department of Pathology, School of Medicine, University of California, Davis, CA 95616.
Karen Kaul, NorthShore University Health System, Evanston Hospital, Evanston, IL 60201 and Pritzker School of Medicine, University of Chicago, Chicago, IL 60637.
Roy Williams, Signal Transduction Program, Sanford-Burnham Medical Research Institute, La Jolla, CA, 92037, USA.
Stan Krajewski, Signal Transduction Program, Sanford-Burnham Medical Research Institute, La Jolla, CA, 92037, USA.
Dan Mercola, Translational Cancer Biology, University of California, Irvine, CA 92697.
Philip M. Carpenter, Translational Cancer Biology, University of California, Irvine, CA 92697.
David Bowtell, Research Division, Peter McCallum Cancer Centre, Melbourne 8006, VIC, Australia.
Ze’ev A. Ronai, Signal Transduction Program, Sanford-Burnham Medical Research Institute, La Jolla, CA, 92037, USA.