Estrogen Receptor β2 Induces Hypoxia Signature of Gene Expression by Stabilizing HIF-1α in Prostate Cancer

The estrogen receptor (ER) β variant ERβ2 is expressed in aggressive castration-resistant prostate cancer and has been shown to correlate with decreased overall survival. Genome-wide expression analysis after ERβ2 expression in prostate cancer cells revealed that hypoxia was an overrepresented theme. Here we show that ERβ2 interacts with and stabilizes HIF-1α protein in normoxia, thereby inducing a hypoxic gene expression signature. HIF-1α is known to stimulate metastasis by increasing expression of Twist1 and increasing vascularization by directly activating VEGF expression. We found that ERβ2 interacts with HIF-1α and piggybacks to the HIF-1α response element present on the proximal Twist1 and VEGF promoters. These findings suggest that at least part of the oncogenic effects of ERβ2 is mediated by HIF-1α and that targeting of this ERβ2 – HIF-1α interaction may be a strategy to treat prostate cancer.


Introduction
Prostate cancer is a slowly progressing disease, initially treatable with androgen-deprivation therapy (ADT) [1], but usually recurring in a more aggressive form that is androgen independent [2,3]. Most aggressive prostate cancers express high levels of androgen receptor (AR) and, in addition, utilize a variety of mechanisms to activate AR in the absence of its ligand. For instance, the cancer can acquire the ability to synthesize AR ligands, phosphorylate AR or, through alternative splicing, create a constitutively active AR [4]. In contrast, expression of the main isoform of estrogen receptor β (ERβ/ESR2), ERβ1, is reduced during prostate cancer progression [5][6][7][8]. ERβ1 has been shown to down-regulate the expression of AR, so upon [15]. For construction of PC3 cells expressing biotin ligase, PC3 cells were transfected with pBirA plasmid expressing Escherichia coli biotin holoenzyme synthetase (BirA) from the actin promoter and selected with G418 at 500 μg/ml. After two weeks clones were isolated and assayed for biotinylation activity.

Western blotting
Twenty micrograms of protein were loaded on an SDS-PAGE 10% Bis-Tris gel with Tris running buffer and transferred to a nitrocellulose membrane after electrophoretic separation. Membranes were blocked with 5% non-fat powdered milk in 0.1% TBST buffer and probed with anti-ERβ2 (produced in the lab), Twist1 (sc-15393), and HIF-1α (sc-10790) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Primary antibodies were used at 1:200-1000 dilutions, and secondary antibody was used at 1:10,000.

In vitro translation and bacterial expression
HIF-1α and ERβ2 were translated using the TNT Quick Coupled Transcription/Translation Systems (Promega Madison, WI). Briefly, 0.2 μg of HIF-1α or ERβ2 expression vectors (T7 promoter) were added to an aliquot of the TNT Quick master mix and incubated in a volume of 50 μl for 60 minutes at 30°C. The in vitro synthesis of protein was verified by SDS-PAGE. For bacterial expression BL21-DE3 bacterial cells were used to express His-LBD-ERα, His-LBD-ERβ1, His-LBD-ERβ2 and His-LBD-ERβ2ΔCX expression plasmids.

In-vitro pull downs
2μl in vitro translated HIF-1α protein was incubated with 25 μl of bacterial lysate containing His-ERβ2 (LBD) overnight, and complexes were adsorbed onto nickel agarose beads for 2 h. Beads were washed three times with ice-cold NETN (20 mM Tris (pH = 8.0), 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40). Complexes were boiled with 20 μl of 2× sample buffer, subjected to SDS-PAGE and transferred to nitrocellulose membrane for Western analyses. The blot was probed with anti-HIF-1α antibody (Santa Cruz). Primary antibodies were at 1:1000 dilution, and secondary antibody at 1:10,000.

Mammalian 2 hybrid assay
ERβ2 was cloned into the Gal4DBD expressing plasmid PM2 in frame with the Gal4DBD domain using BamHI and HindIII restriction sites making PM2-ERβ2. The pVP16 plasmid (Clontech) was used to clone the N-terminal HIF-1α (amino acids 1-401), oxygen destabilization domain (amino acids 401-603) and the C-terminal domain (amino acids 603-826) in frame with the VP16 activation domain using BamHI and PstI restriction sites. For interaction assays in PC3 cells the Gal4 responsive reporter FR-LUC 300 ng, PM2-ERβ2 500 ng and VP16 fused HIF-1α domains 200 ng was transfected using PEI method of transfection as described by Longo et al. [29] Microarray and bioinformatic analyses The two-color comparative microarray analysis covers the fully known protein-coding transcriptome of 39,600 transcripts and variants by using Human OpArray arrays purchased from Microarrays Inc. (Huntsville, AL). The array was complemented with 133 oligonucleotides specifically synthesized for detailed and robust analysis of nuclear receptors, splice variants, and coregulators. The microarray analysis was performed essentially as in Richter et al. [30]. For each cDNA synthesis, 20 μg of total RNA were used, and each comparison was replicated and dye swapped. Bioinformatic analyses were performed using GenePix, R, and Pathway Studio software (Elsevier, Philadelphia, PA). Differentially expressed genes were defined using a pvalue cut-off of p = 0.005 in combination with an M-value (log2 of fold-change) cut-off of +/-0.3. Enrichment analysis of gene ontology (GO) of the biological processes was performed with

Chromatin Immunoprecipitation (ChIP)
Sub-confluent PC3 and 22Rv1 cells (90%) were grown in a 150 mm dish and washed twice with cold PBS and then fixed in PFA (1%) for 10 minutes. Cells were washed twice with cold PBS + protease inhibitor, scraped with 150 μl of resuspension buffer, and centrifuged (4000rpm, 10 minutes). Pellets were resuspended in 750 μl ChIP lysis buffer (HEPES 50 mM, NaCl 140 mM, triton 1%, protease inhibitor) and cells kept on ice for 30 minutes. Cell lysates were sonicated (60 pulses, 30 seconds each) with 30 seconds break at 4°C. Following sonication, 25 μl samples were collected and stored at -20°C (input). The rest of the 125 μl cell lysate fraction was used for the following steps. Samples were pre-absorbed in 20 μl of G protein coupled FLAG tagged magnetic beads at 4°C for 2 hours and supernatant collected. Samples were then divided into two fractions and antibodies against FLAG (2.5 μg) or IgGs (2.5 μg) were added and incubated overnight. The next day, the beads were washed three times with ChIP lysis buffer, once with ChIP wash high salt buffer (HEPES 50 mM, NaCl 500 mM, triton 1%), and finally once with ChIP wash buffer (Tris 10 mM, LiCl 250 mM, NP40 0.5%, EDTA 0.1 mM). Beads were resuspended in 40 μl elution buffer (Tris 50 mM, SDS 1%, EDTA 10 mM) and incubated at 65°C overnight with a small hole on the cap. The input collected previously was also placed at 65°C. The next day, samples were purified using PCR purification kit (Qiagen) and used for qPCR analysis using the following primers. pHRE-Twist1 promoter

Statistics
The values are expressed as the mean with 95% confidence intervals. An unpaired two-tailed ttest was used to compare the differences between two groups. The significance is presented as Ã p<0.05, ÃÃ p<0.005, and ÃÃÃ p<0.001, and non-significant differences are presented as NS.
To validate these data, we analyzed the expression of HIF-1α in PC3 and another prostate cancer cell line-22Rv1, which are both over-expressing ERβ2. We observed an increase in HIF-1α protein level in ERβ2 expressing versus control cells, but no changes in corresponding mRNA level (Fig 1A-1D). This agrees with our microarray results, which did not detect change in HIF-1α transcription levels, but indicated a change in HIF-1α functions (based on enriched pathways). Further, ERβ2-dependent increase in HIF-1α protein levels was confirmed to be accompanied by ERβ2-dependent increase in HIF-1α activity. Luciferase activity of a reporter driven by the HIF-1α-dependent hypoxia inducible gene 2 (HIG2/HILPDA) promoter (HIG2-A-luc) [26] was increased in the presence of transfected ERβ2 expression vector (Fig  2A-2C). In a previous study, we showed that ERβ2-expressing PC3 cells have increased level of Twist1 [15]. Here, we show that ERβ2 also increased activity of a luciferase reporter driven by the HIF-1α-dependent Twist1 promoter in LNCaP prostate cancer cells and that this effect was dependent upon the HIF-1α response element (Fig 3A-3C). This demonstrates that HIF-1α activity in general is enhanced in prostate cells expressing ERβ2.

ERβ2 directly interacts with HIF-1α causing stabilization
Since ERβ2 increases expression of HIF-1α protein but not mRNA, we speculated that HIF-1α stabilization occurs through protein-protein interaction between ERβ2 and HIF-1α. To test this hypothesis, we attached bacterially expressed LBDs of ERα, ERβ and ERβ2 to a solid support and determined interactions with in vitro translated HIF-1α. HIF-1α strongly bound to ERβ2, but not to equivalent amounts of wild type ERα or ERβ LBDs. This interaction was independent of ERβ2 specific sequences because truncation of amino acids encoded by the ERβ2-specific exon did not eliminate HIF-1α binding (ERβ2ΔCX) (Fig 4A). Pull-down of biotin-tagged ERs expressed in PC3 cells revealed specific interaction of transfected HIF-1α with ERβ2 and ERβ2ΔCX and only modest interaction with ERβ1 ( Fig 4B). Since ERβ5 has the same amino acid sequence and is truncated at the same amino acid as ERβ2, i.e. only the small ERβ2 Stabilizes HIF-1α in Prostate Cancer C-terminal peptides differs between the two variants we decided to test this variant and investigate its interaction with HIF-1α. As shown in Fig 4B, ERβ5 interacts strongly with HIF-1α.

The oxygen destabilizing domain (ODD) of HIF1α is not required for interaction with ERβ2 and ERβ5
To investigate, if the oxygen destabilizing domain of HIF-1α was required for interaction with ERβ2 we constructed an expression plasmid of HIF-1α with amino acids 401-603 deleted as described earlier [31]. We then co-immunoprecipitated this domain deleted HIF-1α with biotinylated ERβ2 and ERβ5. As can be seen in Fig 5A the HIF-1α with deleted ODD (Δ401-603) interacts as strongly as the wild-type indicating that the ODD domain is dispensable for the ERβ2 Stabilizes HIF-1α in Prostate Cancer interaction. To map the interaction further we used mammalian two-hybrid assay where the N-terminus, ODD domain and C-terminus of HIF-1α were fused to VP16 and transfected into PC3 cells together with Gal4DBD fused ERβ2 and a reporter with Gal4 binding sites in front of luciferase. As can be seen in Fig 5B the interaction occurs preferably with the N-terminus of HIF-1α. It is well known that prolines 405 and 564 of HIF-1α under normoxia are hydroxylated by proline hydroxylases (PHD's) and then recognized by the VHL factor and subsequently targeted for degradation [32]. We wanted to find out if ERβ2 and ERβ5 interaction with HIF-1α is dependent on hydroxylation of these prolines so that the interaction would even occur during normoxic conditions. In Fig 6, we showed that WT and P405/A-P564/A mutated HIF-1α interact strongly with ERβ2 irrespective of destruction of prolyl hydroxylation sites suggesting that the interaction occurs both during normoxia and hypoxia.
ERβ2 is recruited to HIF-1α response elements of the Twist1 and VEGF promoters Finally, we explored whether ERβ2 can bind to the HIF-1α response elements using ChIP-qPCR. We detected increased levels of ERβ2 recruited at the HIF-1α response elements on the HIF-1α dependent Twist1 and VEGF promoters in both PC3 and 22Rv1 cells. However, ERβ2 does not bind to the promoters of the classical estrogen response element in the pS2 gene ( Fig  7A and 7B).
In conclusion, our findings indicate that ERβ2 binds to and stabilizes HIF-1α and furthermore, is co-recruited to key HIF-1α elements in HIF-1α regulated genes thereby enhancing the activity of HIF-1α during normoxic conditions in prostate cancer cells.

Discussion
Response of normal tissues to hypoxia is crucial for proper vascularization and for subsequent blood supply to oxygenate and provide nutrients to the tissue. A growing tumor becomes hypoxic when reaching more that 1mm in size [33]. The immediate response to hypoxia is a decreased prolyl-hydroxylation of HIF-1α leading to increased stabilization of this protein. This is because VHL, which induces ubiquitinylation and subsequent degradation of HIF-1α under normoxic conditions, cannot bind to HIF-1α in the absence of prolyl-hydroxylation. Stabilization of HIF-1α allows regulation of genes involved in angiogenesis such as VEGF, which is secreted from the tumor, attracting endothelial cells by binding to their surface VEGF receptors thus allowing increased tumor vascularization and growth. HIF-1α also increases expression of genes involved in glycolysis, to adapt the tumor to survive under conditions of low oxygen [34], and genes involved in invasion and metastasis such as Twist1 [25]. expressing doxycycline-regulated ERβ2 expression, this ERβ isoform binds to HRE (HIF-1α response element in Twist1 promoter) and VEGF promoter (HIF-1α response element in VEGF promoter) but not to pS2 (ERE) promoter. ChIP-qPCR results are shown with a non-specific IgG and a specific anti-M2-FLAG antibody immunoprecipitation. ERβ2 binding is enriched only in HIF-1α response element containing Twist1 and VEGF promoters but not in the pS2 promoter lacking HIF-1α response element. The graph shows the data as an increase in binding of FLAG to the Twist1 (pHRE) and VEGF promoter (mean of three separate experiments (±s.e.m.) calculated using Student's t-test, *p 0.028, ##p 0.041 (PC3 cells) and **p 0.0049, ##p 0.0481 (22Rv1 cells). Expression of ERβ2 correlates with worse prognosis in prostate cancer [14] and ERα-negative breast cancer [35], but possible oncogenic actions of ERβ2 are not understood. We show here that there is a strong correlation between ERβ2 expression and HIF-1α protein level but not mRNA level in the prostate cancer cell lines PC3 and 22Rv1. ERβ2 enhances the activity of HIF-1α dependent promoters and is recruited to the Twist1 and VEGF promoters, likely by tethering to HIF-1α. Together, the findings outlined above suggest that HIF-1α mediates the oncogenic effect of ERβ2 through a non-classical signaling pathway in which ERβ2 binds to, and stabilizes HIF-1α under normoxic conditions. This is in line with other studies showing that HIF-1α interacting proteins stabilize HIF-1α by blocking its degradation [22][23][24].
Surprisingly, although we found that ERβ1 does not interact efficiently with HIF-1α, correlating with its inability to induce HIF-1α responsive genes, the unique last exon of ERβ2 is not required for HIF-1α binding. Apparently, truncation of the ERβ1 C-terminus exposes a new protein surface mediating interaction with HIF-1α.
It remains to be shown if ERβ2 expression correlates with levels of Twist 1 or other HIF-1α target genes in clinical samples. A recent study by Ragnum et al. [36] shows that androgen-deprivation therapy (ADT) causes a reduced expression of many genes linked to hypoxia. We propose that expression of ERβ2 or ERβ5 could counteract the effect of ADT on hypoxia-linked genes in a prostate tumor and thus cause castration resistance resulting in a poorer outcome. Indeed, our findings of non-hypoxic stabilization of HIF-1α suggest that it might be possible to develop a small molecule interfering with the ERβ2 and ERβ5-induced HIF-1α stabilization for use in therapy of certain aggressive forms of prostate cancer. An overview of the described findings is shown in Fig 8. In conclusion, expression of the ERβ variants ERβ2 and ERβ5 has previously been shown to correlate with aggressive prostate cancer. Here we show that ERβ2 and ERβ5 stabilize the HIF-1α protein and induce hypoxic gene expression under normoxic conditions, proposing a mechanism for the oncogenic effect of ERβ2 and ERβ5.