Opposing Effects of Cyclooxygenase-2 (COX-2) on Estrogen Receptor β (ERβ) Response to 5α-Reductase Inhibition in Prostate Epithelial Cells*

Current pharmacotherapies for symptomatic benign prostatic hyperplasia (BPH), an androgen receptor-driven, inflammatory disorder affecting elderly men, include 5α-reductase (5AR) inhibitors (i.e. dutasteride and finasteride) to block the conversion of testosterone to the more potent androgen receptor ligand dihydrotestosterone. Because dihydrotestosterone is the precursor for estrogen receptor β (ERβ) ligands, 5AR inhibitors could potentially limit ERβ activation, which maintains prostate tissue homeostasis. We have uncovered signaling pathways in BPH-derived prostate epithelial cells (BPH-1) that are impacted by 5AR inhibition. The induction of apoptosis and repression of the cell adhesion protein E-cadherin by the 5AR inhibitor dutasteride requires both ERβ and TGFβ. Dutasteride also induces cyclooxygenase type 2 (COX-2), which functions in a negative feedback loop in TGFβ and ERβ signaling pathways as evidenced by the potentiation of apoptosis induced by dutasteride or finasteride upon pharmacological inhibition or shRNA-mediated ablation of COX-2. Concurrently, COX-2 positively impacts ERβ action through its effect on the expression of a number of steroidogenic enzymes in the ERβ ligand metabolic pathway. Therefore, effective combination pharmacotherapies, which have included non-steroidal anti-inflammatory drugs, must take into account biochemical pathways affected by 5AR inhibition and opposing effects of COX-2 on the tissue-protective action of ERβ.

Benign prostatic hyperplasia (BPH) 3 is a common disorder affecting elderly men. Although 50% of men over the age of 50 will develop histological BPH (1,2), less than half will go on to develop lower urinary tract symptoms (LUTS) (3,4) such as urinary frequency, urgency, and retention (5,6). Multiple factors including altered steroid hormone signaling and chronic inflammation contribute to the pathogenesis of BPH, ultimately leading to an increase in prostate size through tissue remodeling (4,(7)(8)(9).
The prostate epithelium is responsive to multiple sex steroids and expresses both androgen receptor (AR) and the ␤ isoform of estrogen receptor (ER␤) (9,10). The prostate, which is mainly an androgen-dependent organ, can convert testosterone to not only the potent AR ligand dihydrotestosterone (DHT) but also to ER␤ ligands, the most abundant of which is 5␣-androstane-3␤,17␤-diol (3␤-diol) (11,12). Although androgens regulate differentiation functions of the prostate and enhance proliferation of prostate cancer cells, these actions are opposed by estrogenic ligands, which limit proliferation and epithelial cell differentiation (12)(13)(14)(15).
The main pharmacotherapies for treating symptomatic BPH include ␣-adrenergic receptor blockers and 5␣-reductase (5AR) inhibitors, which limit conversion of testosterone to DHT (16,17). ␣-Adrenergic blockers alleviate LUTS, whereas 5AR inhibitors reduce prostate size. Several clinical trials (Medical Therapy of Prostatic Symptoms, Reduction by Dutasteride of Prostate Cancer Events, and Prostate Cancer Prevention Trial) have examined the use of single and combination therapies in prostate cancer patients with the concurrent examination of BPH progression. The Prostate Cancer Prevention Trial showed that finasteride, a 5AR type 2 inhibitor, decreased the number of BPH events over a 7-year period as compared with placebo (18). In the Reduction by Dutasteride of Prostate Cancer Events trial, dutasteride, a 5AR type 1 and 2 inhibitor, reduced the number of BPH events compared with placebo over a 2-year period (19). The Medical Therapy of Prostatic Symptoms trial examined the combination of ␣-adrenergic receptor blockers with 5AR inhibitors and the effectiveness of combination therapy over monotherapy (20). Treatment with either drug significantly reduced BPH symptoms compared with placebo, but the combination therapy performed signifi-cantly better in alleviation of LUTS compared with monotherapy. The SMART-1 study showed that ␣-adrenergic receptor blockers used initially could be discontinued after 6 months without any loss of improvement from the combination therapy (21). However, treatment-resistant disease is still a considerable problem, and many men still fail medical therapy and go on to require invasive surgical intervention for symptomatic relief.
Chronic inflammation has been implicated as a contributor to both BPH and prostate cancer (8,20). Inflammation in BPH is characterized by an increased production of proinflammatory cytokines and chemokines by prostate stromal cells in conjunction with infiltration by inflammatory immune cells (22)(23)(24). Additionally, oxidative stress mediated by nitric-oxide synthase (NOS) and cyclooxygenase (COX) is associated with BPH in patients exhibiting significant inflammation (25,26). Therefore, it has been hypothesized that combination therapy with non-steroidal anti-inflammatory drugs (NSAIDs) and 5AR inhibitors may improve LUTS. A small clinical study designed to investigate the short term effects of NSAIDs plus finasteride showed that patients treated with both drugs improved in both international prostate symptom score and urinary flow within the first 4 weeks; however, within 24 weeks, combination therapy showed no increased efficacy compared with monotherapy with finasteride (27).
In this report, we examine the impact of combined 5AR and COX-2 inhibition on the biological functions of ER␤ in a human BPH-derived cell line (i.e. BPH-1). Our results show that COX-2 promotes ER␤ activity through the regulation of steroidogenic enzyme gene expression, ultimately leading to the enhanced production of ER␤ ligands. Our results suggest that the anti-inflammatory benefits of NSAIDs in prostatic disease may be counterbalanced by a reduction in tissue-protective effects of ER␤. Therefore, any combination pharmacotherapies that attempt to limit the inflammatory component of benign or malignant prostate disease may need to include compounds that maintain ER␤ signaling.
RNA Isolation and qPCR-RNA was isolated from cells using TRIzol (Life Technologies) and chloroform (Sigma-Aldrich) with procedures described previously (30). cDNA synthesis was performed using the iScript cDNA synthesis kit containing a mixture of oligo(dT) and random hexamer primers (Bio-Rad) as described by the supplier. Gene-specific primers were used to validate gene expression levels using the comparative C T method (Table 1).
Western Blotting-Cell-free protein lysates were obtained from cells using radioimmune precipitation assay buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, 0.5 mM EGTA, 140 mM NaCl, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS), and total protein levels were quantified using the Lowry protein assay (Bio-Rad). Equivalent amounts of total proteins were resolved by 10% SDS-PAGE (Bio-Rad), transferred to PVDF membranes (Millipore, Danvers, MA), and subjected to Western blotting analysis as described (31). Blots were blocked in 5% nonfat dried milk in PBS ϩ 0.1% Tween 20 (Sigma) and incubated with the appropriate primary and secondary antibodies. Antibodies against COX-2 were purchased from Cayman Chemicals, and antibodies against Smad3 and pSmad3 were purchased from Cell Signaling Technology (Danvers, MA). Antibodies against GAPDH were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).
Indirect Immunofluorescence-Cells were plated on glass coverslips, fixed in 4% paraformaldehyde, and then permeabilized with 0.1% Triton X-100. Cells were blocked with Super-Block Blocking Buffer (Pierce) and then incubated with E-cadherin (Cell Signaling Technology) antibody overnight. Cells were washed with PBS and incubated with Alexa Fluor 488 secondary antibody (ThermoFisher Scientific, Grand Island, ACGGTTTAGCTTGGGTGTCT GCAGCCCAAGGAAACAAAGT NY). For F-actin labeling, cells were incubated for 30 min with Alexa Fluor 488-conjugated phalloidin (ThermoFisher Scientific). Nuclei were stained with DAPI, and slides were mounted with Fluoromount-G (SouthernBiotech, Birmingham, AL). Images were captured using an Olympus Fluoview 1000 confocal microscope.
Lentivirus Infection-Lentivirus sets containing five unique shRNA sequences were purchased from the University of Pittsburgh Cancer Institute Lentiviral Core Facility (Pittsburgh, PA). Cells were plated in 6-well plates and allowed to grow until 50% confluence at which point culture medium was replaced with 1 ml of complete medium containing 16 g/ml Polybrene and 1 ml of lentivirus expressing shRNAs of interest (i.e. COX-2, ESR2, and SMAD4) or scrambled control containing a puromycin resistance gene marker. Following an overnight infection, virus-containing medium was replaced with complete medium containing 1 g/ml puromycin to select for resistant cells. Cells were maintained in selection medium until experiments.
Chromatin Immunoprecipitation and qPCR-BPH-1 cells were treated at 80% confluence with either control (EtOH) or dutasteride (0.1 M) for 24 h. The medium was replaced with fresh growth medium containing 1% formaldehyde for crosslinking DNA-protein complexes and incubated for 10 min at 37°C. The cross-linking reaction was halted with the addition of glycine to a final concentration of 125 mM, and the cells were incubated for 10 min at room temperature. Cells were sonicated at 3 ϫ 15 s at 30 A and then at 3 ϫ 15 s at 40 A on ice. Fragment size was verified by DNA gel electrophoresis. Chromatin was then incubated with antibodies against phospho-Smad3 (Cell Signaling Technology), a portion was collected as an input, and the remainder was then linked to Protein A-Sepharose beads (GE Healthcare). Beads were then washed, and cross-linking was reversed. DNA was isolated using phenolchloroform-isoamyl alcohol and resuspended in Tris-EDTA buffer containing RNase A. DNA was analyzed using RT-qPCR using primers listed in Table 2. RT-qPCR results were calculated as relative enrichment over input control.
Flow Cytometry-Apoptosis was quantified by flow cytometry using annexin V labeling to detect the phosphatidylserine found on the outer membrane of apoptotic cells. 2.5 ϫ 10 5 cells were plated in 6-well plates and allowed to grow overnight prior to being subjected to treatments. Cells were then harvested, washed with ice-cold PBS, and then resuspended in 100 l of binding buffer (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 ) containing 5 l of FITC-conjugated annexin V (BD Biosciences) and 5 l of 1ϫ propidium iodide (Sigma-Aldrich). Cells were gently mixed and incubated for 15 min in the dark, and cell volumes were brought up to 500 l with binding buffer prior to flow cytometry. Cells were counted using an LSRII flow cytometer (BD Biosciences), and data were analyzed with FlowJo (FlowJo, Ashland, OR). Only cells labeled singly with annexin V were considered apoptotic; cells labeled with propidium iodide were considered dead.
RNA Sequencing and Analysis-Libraries were prepared using the Illumina Tru-seq Stranded Total RNA kit according to the manufacturer's protocol (Illumina, San Diego, CA). Ribosomal RNA was depleted from total RNA using Ribo-Zero Gold, and the remaining RNA was fragmented for cDNA synthesis. Adapters were ligated to purified cDNA, and DNA fragments were enriched using AMPure XP Beads (Beckman Coulter, Brea, CA). Libraries were single end-sequenced using Illumina HiSeq 2000 (Tufts core facility, Medford, MA). All sequence mapping and filtering were performed in house. FASTQ files were aligned to the human genome (hg19) using TopHat (32). Cufflinks was then used to assemble RNA reads into transcripts and calculate fragments per kilobase of transcript per million mapped to determine differential gene expression (33).
Steroid Metabolism-BPH-1 cells stably expressing a scrambled shRNA (BPH-1 scr) or BPH-1 shCOX-2 cells were seeded in 12-well plates at 50% confluence 1 day before steroid metabolism analysis. On the day of analysis, the cells in each well were incubated at 37°C with 1 ml of medium containing 0.1 M testosterone or dihydrotestosterone with 1,000,000 cpm 3 H-labeled steroid (PerkinElmer Life Sciences). Aliquots (0.3 ml) of medium were removed at 2, 4, and 12 h, and steroids were extracted with 1 ml of dichloromethane and dried under a nitrogen stream. Steroids were analyzed using an Agilent 1260 Infinity high performance liquid chromatography (HPLC) system with UV detector and ␤-RAM4 in-line scintillation counter (LabLogic, Brandon, FL). Extracted steroid products were dissolved in 20 l of methanol, and 5-l injections were resolved with a 50 ϫ 2.1-mm, 2.6-m, C 8 Kinetex column (Phenomenex, Torrance, CA) at a flow rate of 0.4 ml/min and a methanol/ water linear gradient: 27% methanol from 0 to 0.5 min, 39% to 16 min, 44% to 20 min, 60% to 22 min, 71% to 30 min, 75% to 30.5 min, and 27% to 33 min. Products were identified by retention times of external standards chromatographed at the beginning and end of the experiments. The flow rate of the scintillation mixture (Bio-SafeII, Research Products International, Mount Prospect, IL) was 1.2 ml/min, and the data were processed with Laura4 software (LabLogic (34)).
Prostaglandin E 2 (PGE 2 ) Accumulation-Cells were plated in 10-cm plates at 7.5 ϫ 10 5 cells and incubated overnight. Cells were then subjected to appropriate treatments for 24 h in complete medium. Cells were scraped and stored as a lysate in 2 ml of PBS at Ϫ80°C for prostaglandin analysis. The internal standard, PGE 2 -d 9 (100 ng), was added to the cell lysate and allowed to equilibrate for 5 min before adding 5 ml of CHCl 3 /MeOH (2:1). Samples were then vortexed and set to shake on low for 10 min before centrifugation at 2800 ϫ g for 10 min. The organic phase was transferred to a clean vial and dried under a stream of nitrogen. Samples were reconstituted in 100 l of methanol for stable isotope dilution LC-MS analysis of PGE 2 . Samples (10 l) were injected into a Shimadzu HPLC (Columbia, MD) and separated on a Phenomenex C 18 (2) octadecyl-silica column (2.1 ϫ 150 mm, 5-m bead size, 100-Å pore size) before analysis on an AB Sciex (Framingham, MA) 5000 triple quadrupole mass spectrometer. The solvent system used aqueous 0.1% acetic acid (A) and 0.1% acetic acid in acetonitrile (B). The 25-min gradient operated at a 0.25 ml/min flow rate and started at 35% B, ramping up to 65% B at 20 min. This was followed by a wash using 100% B for 3 min before returning to starting conditions. MS analyses by electrospray ionization were run in negative mode with the collision gas set at 4 units, curtain gas at 40 units, ion source gases at 40 units, ion spray voltage at Ϫ4500 V, and temperature at 550°C. The declustering potential was set to Ϫ50, entrance potential was set to Ϫ5, collision energy was set to Ϫ17, and collision exit potential was set to Ϫ18.4. Single reaction monitoring was used for sample analysis and quantification. Method development using standards confirmed baseline separation of PGE 2 and PGD 2 for accurate quantification using a standard curve based on the following transitions: PGE 2 , 351.2 3 271.2 and PGE 2 -d 9 , 360.2 3 280.2. PGE 2 is reported as ng/10 6 cells.

Results
Dutasteride-mediated Apoptosis in BPH-1 Cells Requires ER␤-Within the normal prostate epithelium, testosterone can be metabolized to a potent AR ligand (DHT) and ER␤ (e.g. 3␤-diol) ligands through different steroid-converting enzymes ( Fig. 1) (35,36). Therefore, normal prostate size and differentiated function are maintained in the presence of testosterone through the opposing actions of AR and ER␤. Perturbations of this balance, either by up-regulation of AR-dependent pathways (i.e. proproliferative) or down-regulation of ER␤-dependent pathways (i.e. proapoptotic), may contribute to BPH progression.
To examine the impact of 5AR inhibition on ER␤ action, we used human BPH-derived cell lines BPH-1 and BHPrE1 (28,29). BPH-1 and BHPrE1 cells express functional ER␤ as treatment with either a natural (3␤-diol) or synthetic (WAY20070) ER␤ agonist efficiently induced expression of E-cadherin, an established ER␤ target gene ( Fig. 2A). These data corroborate previous studies from other laboratories that revealed functional ER␤ in BPH-1 cells (37,38). Because ER␤ activation by a potent synthetic ligand enhances apoptosis in normal human prostate epithelial and stromal cells in mouse xenograft studies (38), we examined the contribution of ER␤ ligands to apoptosis in cultured BPH cell lines. As shown in Fig. 2B, treatment with either ER␤ ligand did not trigger a significant increase in apoptosis. To eliminate the possible contribution of endogenous ER␤ ligands to basal apoptosis, we treated BPH-1 and BHPrE1 cells with the 5AR type 1 and 2 inhibitor dutasteride, which would block the production of potent DHT-derived ER␤ ligands and potentially reduce basal apoptosis in these cells. Surprisingly, dutasteride treatment (0.5 M) on its own increased apoptosis (Fig. 2C). This induction of apoptosis is not observed upon treatment of BPH-1 cells with the 5AR type 2-specific inhibitor finasteride (Fig. 2D). The expression of both 5AR enzymes was detected in BPH-1 cells by quantitative RT-PCR (see Fig. 8E). 5AR inhibitors increase apoptosis in prostate cancer cells (39,40) through limiting the action of DHT on the prosurvival effects of AR. The limited expression of AR in BPH-1 cells makes it unlikely that dutasteride-induced apoptosis in these cells is mediated by reduced AR action (28). Furthermore, dutasteride-induced apoptosis in BPH-1 cells requires ER␤ as shown by the lack of this response upon stable shRNAmediated ablation of ER␤ (Fig. 3, A and B). The results of poly-(ADP-ribose) polymerase 1 cleavage assays (Fig. 3C) corroborate our conclusion that dutasteride induces apoptosis as revealed by annexin V staining. Furthermore, dutasteride does not exert a significant effect on the generation of the cleaved form of LC-3, arguing against an effect on autophagy (Fig. 3D). The contribution of ER␤ to dutasteride-induced apoptosis is further supported by enhanced apoptosis in BPH-1 cells treated with a suboptimal concentration of dutasteride and the potent synthetic ER␤ ligand WAY20070 (Fig. 3E), either of which alone does not promote apoptosis. Stable ablation of ER␤ eliminated the proapoptotic effect of combined suboptimal concentrations of dutasteride and WAY20070 (Fig. 3F). In summary, our results confirm previous results establishing the role for ER␤ in triggering apoptosis in BPH-derived cells and tissues (38) but uniquely reveal an ER␤-dependent effect on apoptosis in these cells in response to inhibition of 5AR types 1 and 2. The inability of finasteride alone to trigger apoptosis in BPH-1 cells suggests that dutasteride may be having an off-target effect to drive apoptosis, which prompted further investigation into the mechanism of dutasteride effects on ER␤-dependent apoptosis.
Dutasteride-mediated Repression of E-cadherin mRNA Expression in BPH-1 Cells Requires ER␤-To further examine the impact of dutasteride on ER␤ action, we measured the expression of the ER␤ target gene E-cadherin in response to dutasteride treatment. Estrogenic ligands such as 3␤-diol activate ER␤ in prostate epithelial cells and maintain epithelial cell differentiation in part by limiting epithelial-mesenchyme transition (EMT) (41,42) through the induction of cell adhesion proteins such as E-cadherin (43). As shown in Fig. 4A, dutasteride does not promote or enhance the induction of E-cadherin but rather represses its expression in BPH-1 cells. In the absence of ER␤, there is no significant change in E-cadherin expression in any treatment group assessed (Fig. 4B), indicating that dutasteride repression of E-cadherin mRNA expression requires ER␤. Because a combined treatment with testosterone does not alter dutasteride repression of E-cadherin mRNA expression (Fig. 4A), inhibition of 5ARs appears to be required for the effect of dutasteride on E-cadherin expression. This conclusion is further supported by the elimination of the repressive effect of dutasteride on E-cadherin expression by a combined treatment with DHT, the product of 5AR. Treatment of BPH-1 scr cells with dutasteride results in a decrease in total E-cadherin protein, an effect not observed in ER␤-ablated BPH-1 cells (Fig. 4C). However, indirect immunofluorescence analysis did not reveal any dramatic effects on plasma membrane localization of E-cadherin or organization of F-actin into stress fibers (Fig. 4D) following a 24-h treatment with dutasteride. Thus, dutasteride treatment alone is not sufficient to drive EMT in BPH-1 cells. Treatment with the natural (3␤-diol) or synthetic (WAY20070) ER␤ ligand significantly increases E-cadherin mRNA levels in BPH-1 cells and overcomes the repressive effect of dutasteride (Fig. 4E). The induction of E-cadherin mRNA expression by 3␤-diol and WAY20070 and repression by dutasteride are not observed in ER␤-ablated BPH-1 cells (Fig. 4F). Thus, dutasteride generates opposing effects on ER␤ responses in BPH-1 cells, acting to promote ER␤-dependent apoptosis ( Fig. 2) but limiting one component of the EMT-inhibitory action of the receptor as reflected in reduced E-cadherin expression.
Canonical TGF␤ Signaling Is Required for Dutasteride Repression of E-cadherin mRNA Expression-EMT is a key component in the pathogenesis of BPH and occurs through a TGF␤-dependent mechanism (41,44). In the prostate, TGF␤1 regulates cell growth, EMT, apoptosis, and carcinogenesis (45,46). Tissue specimens from human BPH patients show increased pSmad3 staining in areas of the prostate undergoing the phenotypic changes associated with EMT (44). Treatment of BPH-1 cells with dutasteride increases the phosphorylation of Smad3 significantly within 30 min with no changes in total Smad3 protein (Fig. 5A). Pretreatment of BPH-1 cells with TGF␤-neutralizing antibody inhibited the dutasteride-mediated repression of E-cadherin (Fig. 5B). In the canonical TGF-␤ signaling pathway, Smad2 and Smad3 can either form homo-or heterodimers upon activation. However, phosphorylated Smad2/3 associates with Smad4, translocates to the nucleus, and then binds to Smad-binding elements of TGF␤-regulated genes. Given that the inhibition of E-cadherin induction by dutasteride is blocked by a TGF␤-neutralizing antibody, we chose to ablate Smad4 in BPH-1 cells to inhibit this pathway (Fig. 5C). As shown in Fig. 5D, the repression of E-cadherin mRNA expression by dutasteride is not observed in the Smad4ablated cells (Fig. 5D) with expression levels similar to that of cells pretreated with TGF␤-neutralizing antibody (Fig. 5B). Therefore, dutasteride treatment of BPH-1 cells triggers a rapid activation of the canonical TGF␤ signaling pathway, which then mediates the ER␤-dependent repression of E-cadherin mRNA expression. TGF␤-mediated Induction of COX-2 in Response to Dutasteride Limits ER␤ Activation-BPH is commonly characterized as a disease of chronic inflammation and is associated with overexpression of COX-2 in luminal epithelial cells (25). Consistent with the contribution of COX-2 to BPH pathology, we found that basal COX-2 protein expression is much higher in BPH-derived cell lines than in the normal prostate epithelial cell line RWPE-1 (Fig. 6A). As shown in Fig. 6B, dutasteride treatment of BPH-1 cells induces COX-2 expression. This induction is lost with the addition of DHT but not testosterone, supporting the dependence on 5AR inhibition for this response. Furthermore, dutasteride induction of COX-2 in BPH-1 cells requires TGF␤ as it is not observed in Smad4-ablated BPH-1 cells (Fig. 6C). Pretreatment with TGF␤-neutralizing antibody also inhibits the dutasteride-mediated induction of COX-2 (Fig. 6D). Several putative Smad-binding elements have been identified in the promoter region of COX-2 (47,48). Examination of three Smad-binding elements by chromatin immunoprecipitation (ChIP) qPCR revealed that dutasteride treatment increases the binding of pSmad3 complexes to the COX-2 promoter (Fig. 6E). In contrast to these direct effects of the canonical TGF␤ pathway on COX-2 gene induction by dutasteride, repression of E-cadherin gene expression by TGF␤ is likely to be indirect, involving Smad3 regulation of E-cadherin targeting microRNAs (49).
Previous studies from our laboratory revealed an inhibitory effect of COX-2-generated ROS on the transcriptional response of ER␤ in the DU145 prostate cancer cell line (50). To determine whether ER␤ action in BPH-1 cells is likewise affected by ROS, we treated cells with a physiologic concentration of H 2 O 2 (i.e. 50 M) along with the ER␤ ligand WAY20070. As shown in Fig. 7A, ER␤ induction of E-cadherin is inhibited by H 2 O 2 . ROS may further promote EMT in BPH as evidenced by the induction of EMT markers Snail and Slug in BPH-1 cells (Fig. 7, B and C). These EMT-promoting effects of ROS cannot be overcome by treatment with WAY20070.
Disruption of COX-2 Alters BPH-1 Response to Dutasteride-To more definitively examine whether COX-2 contributes to ER␤ action in BPH-1 cells, we used both pharmacological and molecular approaches to reduce its activity and expression. As shown in Fig. 8A, stable shRNA-mediated ablation of COX-2 expression (Fig. 8C) or specific inhibition of COX-2 activity by NS398 increased apoptosis in BPH-1 cells in response to dutasteride. Given the ER␤ dependence of dutasteride-induced apoptosis (Fig. 3D), these results suggest that COX-2 may indeed limit the proapoptotic action of ER␤ in prostate epithelial cells. Furthermore, the ER␤-dependent repression of E-cadherin by dutasteride is lost upon COX-2 ablation (Fig. 8D). The sensitization of BPH-1 cells to dutasteride-mediated apoptosis is also observed for finasteride (Fig. 8B), which is ineffective on its own  1 shESR2) with an shESR2-expressing lentivirus. A Student's t test was performed. *, p value Ͻ0.05; n ϭ 3. B, dutasteride triggers apoptosis in BPH-1 scr cells but not BPH-1 shESR2 as measured by flow cytometry. A two-way ANOVA followed by Bonferroni post hoc test was performed. *, p value Ͻ0.05; n ϭ 3. C, dutasteride triggers apoptosis in BPH-1 scr cells but not BPH-1 shESR2 as measured by poly(ADP-ribose) polymerase (PARP) cleavage. A one-way ANOVA followed by Dunnett's post hoc test was performed. *, p value Ͻ0.05; n ϭ 3. D, dutasteride does not induce autophagy in either BPH-1 scr or BPH-1 shESR2 as measured by Western blotting. A Student's t test was performed. E, apoptosis is not increased when BPH-1 cells are treated with 3␤-diol, WAY20070, or a suboptimal dose of dutasteride (1 M). However, a combination treatment with WAY20070 and dutasteride significantly increases apoptosis. A one-way ANOVA followed by Tukey's multiple comparison test was performed. *, p value Ͻ 0.05; n ϭ 3. F, treatment with 3␤-diol, WAY20070, and dutasteride either singly or in combination does not induce apoptosis in ER␤-ablated BPH-1 cells. n ϭ 3. Error bars represent S.E. ns, not significant. (Fig. 2B) to promote apoptosis. Examination of 5␣-reductase reductase type 1 (SRD5A1) and type 2 (SRD5A2) genes showed that they are both expressed in BPH-1 (Fig. 8E). However, expression of SRD5A1 is significantly increased upon COX-2 knockdown. This suggests that the increased efficacy of dutasteride and finasteride could be due in part to an increase in expression of the target enzyme. Additionally, the increase of COX-2 expression as observed in BPH tissue may negatively influence the ability of ER␤ to drive apoptosis and limit EMT in prostate epithelial cells.
COX-2 Regulates Steroid Hormone Metabolism-The impact of COX-2 on various ER␤ responses in dutasteride-treated BPH-1 cells could be due to multiple mechanisms of action. For example, although 5AR inhibition by dutasteride will limit the production of ER␤ ligands from DHT, other ligands could be generated via backdoor pathways utilizing alternative steroid precursors and steroidogenic enzymes (36). We therefore initially confirmed that BPH-1 cells contained the enzymatic machinery to generate DHT from testosterone and ER␤ ligands from DHT as initially described by Hayward et al. (28) (Fig. 9, A  and B). Surprisingly, although COX-2 ablation did not affect the conversion of testosterone to DHT in BPH-1 cells, the production of ER␤ ligands from DHT is severely compromised in COX-2-ablated cells (Fig. 9, C and D). Therefore, although COX-2 may limit ER␤ activity through the production of ROS, it is required for efficient production of endogenous ER␤ ligands in BPH-1 cells.
To identify potential genetic alterations caused by COX-2 ablation in BPH-1 cells, RNA-seq was performed on both BPH-1 scr and BPH-1 shCOX-2 cells. Sequencing was performed on three biological replicates of each cell line, and only genes that were greater than one fragment per kilobase of transcript per million mapped were considered to be sequenced above baseline. To be considered differentially expressed between the two conditions, expression differences had to be greater than 2-fold in at least two replicates with a p value Ͻ0.05. Of the 16,007 genes sequenced that were greater than one fragment per kilobase of transcript per million mapped, 6,952 genes were differentially expressed. Among these, expression of 10 steroidogenic enzyme genes was altered when COX-2 was ablated in BPH-1 cells (Table 3). Specifically, the expression of two enzymes responsible for the conversion of DHT to 3␤-diol and 3␣-diol was significantly reduced. Although both AKR1C1 and AKR1C2 can metabolize DHT to adiols, AKR1C1 is predominantly responsible for the metabolism of DHT to 3␤-diol, and AKR1C2 is responsible for the

. Dutasteride repression of E-cadherin mRNA is mediated by canonical TGF␤ signaling. A, dutasteride treatment of BPH-1 cells represses E-cadherin mRNA expression. Dutasteride-induced repression of E-cadherin mRNA expression is overcome by a co-treatment with DHT but not testosterone.
A one-way ANOVA followed by Tukey's multiple comparison test was performed. *, p value Ͻ 0.05; n ϭ 3. B, in the absence of ER␤, neither dutasteride treatment nor dutasteride and testosterone co-treatment represses E-cadherin mRNA expression. n ϭ 3. C, E-cadherin protein decreases upon dutasteride treatment only in the BPH-1 scr cell line. A representative Western blot is shown below. A Student's t test was performed. *, p value Ͻ 0.05; n ϭ 3. D, indirect immunofluorescence of E-cadherin in BPH-1 scr cells treated with dutasteride showed no significant change in cell surface localization. Dutasteride treatment also did not result in the reorganization of F-actin (phalloidin staining) into stress fibers in BPH-1 scr cells. E, dutasteride repression of E-cadherin mRNA expression is relieved upon co-treatment with an ER␤ ligand (3␤-diol or WAY20070). A one-way ANOVA followed by Tukey's multiple comparison test was performed. *, p value Ͻ0.05; **, p value Ͻ0.01; n ϭ 3. F, E-cadherin mRNA expression is not affected by dutasteride or ER␤ ligands upon ablation of ER␤ in BPH-1 cells. n ϭ 3. Error bars represent S.E. ns, not significant. metabolism of DHT to 3␣-diol (51). Additionally, the expression of the enzyme responsible for the metabolism of 3␤-diol to its inactive metabolite 5␣-androstanetriol, CYP7B1, was elevated in shCOX-2 cells (Fig. 9E). The alterations of these steroidogenic enzyme genes agree with the reduction of ER␤ ligands as determined by mass spectrometry (Fig. 9, C and D). Therefore, the absence of COX-2 limits ER␤ activation through a dual mechanism: reduction of the metabolism of DHT to adiols as well as increased metabolism (inactivation) of any 3␤-diol that is produced.
Inhibition of COX-2 Influences ER␤ Activity through Mediation of Prostaglandin Synthesis-In addition to the alteration in steroid metabolism enzymes, the RNA-seq data also revealed differential expression of two prostaglandin synthase genes, PTGES and PTGES2; ablation of COX-2 leads to decreased expression of these enzymes. Previous studies have shown that several steroid metabolism enzymes are dependent on PGE 2 activation of the PKA pathway (52,53). Examination of PGE 2 accumulation in control and COX-2-ablated BPH-1 cells (both genetically and pharmacologically) by mass spectrometry showed that, in the absence of COX-2, intracellular PGE 2 levels are reduced (Fig. 9F). Treatment with a COX-2-specific inhibitor in the COX-2-ablated cell line showed no additional decrease in PGE 2 accumulation, indicating that the genetic ablation and the pharmacological inhibition of COX-2 is at a maximal level. NS398 treatment of BPH-1 cells results in the same levels of PGE 2 accumulation as is seen in shCOX-2 cells. Interestingly, treatment with a COX-1 inhibitor or a nonspecific COX inhibitor, aspirin, showed an even greater decrease in PGE 2 accumulation, suggesting that there is some COX-1 compensation when COX-2 is ablated in BPH-1 cells.

Discussion
Current monotherapies for symptomatic BPH have limited efficacy in decreasing prostate size and reducing LUTS due to the complex etiology of the disease. BPH progression is linked to both chronic inflammation and a down-regulation of ER␤dependent pathways. In this report, we examined the influence of one inflammatory mediator, COX-2, on ER␤ function in BPH-derived epithelial cells. We initially focused on a current standard pharmacotherapy approach for BPH, which seeks to limit the action of AR through inhibition of the enzymes that convert testosterone to the more potent AR ligand DHT. Although both dutasteride (inhibitor of 5AR types 1 and 2) and finasteride (inhibitor of 5AR type 2 only) prevent the conversion of testosterone to DHT, they also decrease the production of ER␤ ligands generated from DHT and therefore could limit potential tissue-protective effects of this receptor manifested partly through an ER␤-dependent promotion of apoptosis (Fig.  10). However, we found that dutasteride, but not finasteride, induces apoptosis in BPH-derived cells in an ER␤-dependent manner. In nearly all experiments presented in this report, the response of BPH-1 cells to dutasteride was overcome by the addition of DHT but not testosterone, confirming that the inhibition of 5AR activity and production of DHT was responsible for the effects observed. Therefore, ER␤ may promote apopto-sis due to activation by ligands generated by backdoor pathways that do not require 5AR activity or through ligand-independent action that is enhanced in the absence of potent DHT-derived ER␤ ligands.
The fact that finasteride on its own does not trigger apoptosis in BPH-1 cells suggests either that inhibition of testosterone to DHT must be complete for ER␤ to promote apoptosis or that dutasteride is having off-target effects to promote ER␤-depen-

FIGURE 7. ROS blocks ER␤ induction of E-cadherin and induces the expression of EMT markers.
A, treatment of BPH-1 cells with hydrogen peroxide prevents the induction of E-cadherin mRNA by the ER␤ synthetic ligand WAY20070. A one-way ANOVA followed by Dunnett's post hoc test was performed. **, p value Ͻ0.01; n ϭ 3. B, treatment of BPH-1 cells with hydrogen peroxide induces the mRNA expression of Snail, a marker of EMT, which cannot be reversed by co-treatment with WAY20070. A one-way ANOVA followed by Dunnett's post hoc test was performed. ***, p value Ͻ0.001; n ϭ 3 C, treatment of BPH-1 cells with hydrogen peroxide induces the mRNA expression of the EMT marker Slug, which cannot be reversed by co-treatment with WAY20070. A one-way ANOVA followed by Dunnett's post hoc test was performed. **, p value Ͻ0.01; n ϭ 3. Error bars represent S.E. JULY 8, 2016 • VOLUME 291 • NUMBER 28 dent apoptosis and gene expression responses. Dutasteride treatment of BPH-1 cells leads to rapid activation of the canonical TGF␤ pathway, which Smad4 ablation experiments revealed contributes to some of the biological effects of dutasteride (e.g. repression of E-cadherin). Prostate samples collected from BPH patients that failed ␣-blocker and dutasteride combination therapy showed a decrease in E-cadherin expression compared with patients that failed ␣-blocker monotherapy along with an increase in pSmad3 (54). Therefore, dutasteride activation of the TGF␤ pathway, as observed in our studies with BPH-1 cells, can occur in patients subjected to 5AR inhibition, although it is unlikely on its own to drive EMT as supported by our in vitro findings. Nonetheless, these clinical results as well as data reported herein suggest that the clinical utility, or lack thereof, of 5AR inhibitors to treat symptomatic BPH is not only related to its reduction in AR signaling but perhaps influenced by the activation of TGF␤ within individual BPH patients.

COX-2 Regulation of ER␤ in Benign Prostate
One of the TGF␤ responses to dutasteride in BPH-1 cells that may be of particular relevance to BPH, given the importance of inflammation in this disease, is the induction of COX-2 expression. Inflammation associated with BPH may trigger an increase in COX-2 expression. Because COX-2-derived ROS can limit ER␤ action in a prostate cancer cell line (50), we predicted that reducing COX-2 expression in BPH-1 cells would potentiate ER␤ responses in these cells. In fact, the ER␤-dependent induction of apoptosis by dutasteride in BPH-1 cells was promoted by either molecular or pharmacological inhibition of COX-2. Furthermore, finasteride, which is ineffective on its own to promote apoptosis in BPH-1 cells, enhanced ER␤-dependent apoptosis upon COX-2 ablation. Therefore, the inhibition of COX-2 may be required to optimize any pharmacotherapies that seek to limit BPH symptoms through reduced production of DHT or activation of ER␤. A recent clinical trial found that ER␤ agonist treatment alone was insufficient to reduce LUTS in BPH patients (55), which was unexpected given the presumed benefit of ER␤ derived from numerous preclinical studies in rodents (9). These disappointing clinical trial data highlight the need to understand more fully the various signaling pathways that impact ER␤ action in prostate cells to develop effective combination pharmacotherapies for individual BPH patients.  Although COX-2 plays an important role in BPH disease, the clinical benefits of NSAIDs are indeterminate. A small clinical trial showed that combination therapy with NSAIDs and a 5AR inhibitor alleviated LUTS more quickly than monotherapy (27). However, because this effect does not persist with long term treatment, ablation of COX-2 may generate potential compensatory mechanisms that alter treatment efficacy. We uncovered one outcome of COX-2 inhibition that would oppose its beneficial effects in prostate and thereby potentially limit its therapeutic benefit. Specifically, we found that ablation of COX-2 in BPH-1 cells leads to a dramatic alteration in endogenous ER␤ ligand metabolism. Specifically, although COX-2 does not impact 5AR activity (i.e. conversion of testosterone to DHT), it is required for the efficient generation of potent ER␤ ligands in BPH-1 cells.

COX-2 Regulation of ER␤ in Benign Prostate
term efficacy of COX-2 inhibitors in clinical management of BPH.
Other genes that are differentially regulated by COX-2 in BPH-1 cells include two prostaglandin synthase genes, PTGES and PTGES2. The expression of some steroid metabolism enzymes is PGE 2 -dependent through the activation of the PKA pathway (52,56,57). In the absence of COX-2, there was a significant decrease in PGE 2 accumulation. Therefore, reduced PGE 2 accumulation would limit activation of steroid metabolism enzymes that are PGE 2 -dependent, providing yet another mechanism through which chronic COX-2 inhibition serves to alter the intraprostatic hormonal milieu. In addition to selective COX-2 inhibitors, nonspecific COX inhibitors have been evaluated for their effectiveness in limiting BPH symptoms (58 -60). Treatment of BPH-1 cells with a COX-1 inhibitor and a nonspecific COX inhibitor showed an even greater decrease in PGE 2 accumulation than inhibition of COX-2 alone. Even though COX-2 is the inducible isoform and COX-1 is constitutively expressed, these data and our results in BPH-1 cells suggest that within the prostate epithelial cells COX-1 is participating in the production and accumulation of PGE 2 . Additionally, it has been shown that COX inhibition by NSAIDs can inhibit the AKR1C enzymes, further reducing the metabolism and accumulation of ER␤ ligands (51).
Combination therapy with NSAIDs and 5AR inhibitors was considered a promising treatment for men with symptomatic BPH, but their clinical effectiveness has been disappointing. Although 5AR inhibitors do increase apoptosis in BPH cells and could potentially decrease prostate size, these compounds also alter the steroid hormone milieu through decreased production and increased metabolism of ER␤ ligands that normally function to protect and maintain prostate size and differentiation. Additionally, in an effort to treat the inflammation inherent in BPH, treatment with NSAIDs also further decreases the protection provided by ER␤ activation through a dual mechanism: in the absence of COX-2, there is a significant decrease in steroid metabolism enzymes necessary for the production of ER␤ ligands as well as a decrease in PGE 2 accumulation, which serves to limit the activation of PGE 2 -dependent steroid-metabolizing enzymes. Future therapies designed to limit proliferation and chronic inflammation but maintain ER␤ activation may provide more rational treatment strategies and thus serve to provide better clinical outcomes for patients with symptomatic BPH.