ERG/AKR1C3/AR Constitutes a Feed-Forward Loop for AR Signaling in Prostate Cancer Cells Katelyn Powell1, Louie Semaan1, M

Published OnlineFirst March 9, 2015; DOI: 10.1158/1078-0432.CCR-14-2352

Biology of Human Tumors Clinical Cancer Research ERG/AKR1C3/AR Constitutes a Feed-Forward Loop for AR Signaling in Prostate Cancer Cells Katelyn Powell1, Louie Semaan1, M. Katie Conley-LaComb1, Irfan Asangani2, Yi-Mi Wu2, Kevin B. Ginsburg1, Julia Williams3, Jeremy A. Squire4, Krishna R. Maddipati5, Michael L. Cher1,5,6, and Sreenivasa R. Chinni1,5,6

Abstract

Purpose: Intratumoral androgen synthesis in prostate cancer Results: We found that ERG regulated the expression of the ABE contributes to the development of castration-resistant prostate AKR1C3 in prostate cancer cells via direct binding to the AKR1C3 cancer (CRPC). Several enzymes responsible for androgen bio- gene. Knockdown of ERG resulted in reduced AKR1C3 expression, synthesis have been shown to be overexpressed in CRPC, thus which caused a reduction in both DHT synthesis and PSA expres- contributing to CRPC in a castrated environment. The TMPRSS2– sion in VCaP prostate cancer cells treated with 5a-androstane- ERG transcription factor has been shown to be present in primary dione (5a-Adione), a DHT precursor metabolite. Immunohisto- prostate cancer tumors as well as CRPC tumors. We hypothesize chemical staining revealed that ERG was coexpressed with that TMPRSS2–ERG fusions regulate androgen biosynthetic AKR1C3 in prostate cancer tissue samples. enzyme (ABE) gene expression and the production of androgens, Conclusions: These data suggest that AKR1C3 catalyzes the which contributes to the development of CRPC. biochemical reduction of 5a-Adione to DHT in prostate cancer Experimental design: We used a panel of assays, including cells, and that ERG regulates this step through upregulation of lentivirus transduction, gene expression, chromatin immunopre- AKR1C3 expression. Elucidation of ERG regulation of ABEs in CRPC cipitation and sequencing, liquid chromatography-mass spectro- may help to stratify TMPRSS2–ERG fusion-positive prostate cancer metric quantitation, immunocytochemistry, immunohistochem- patients in the clinic for anti–androgen receptor–driven therapies; istry, and bioinformatics analysis of gene microarray databases, to and AKR1C3 may serve as a valuable therapeutic target in the determine ERG regulation of androgen synthesis. treatment of CRPC. Clin Cancer Res; 21(11); 2569–79. 2015 AACR.

Introduction ERG has been shown to have several biologic functions that facilitate its oncogenic properties in prostate cancer. It has been The TMPRSS2–ERG fusion gene has been shown to be present shown to regulate cellular migration and invasion, epithelial-to- in approximately 40% of prostate cancer tumors, including pri- mesenchymal transition, dedifferentiation, and AR signaling (16– mary and advanced prostate cancer (1–6). The TMPRSS2–ERG 19); however, the biologic functions of ERG contributing to the fusion gene results in androgen receptor (AR)–induced overex- development of castration-resistant prostate cancer (CRPC) have pression of the ERG transcription factor (7). The presence of not been fully elucidated. Interestingly, ERG fusion-positive pro- TMPRSS2–ERG in prostate cancer has been shown to be associ- state cancer patients responded better to abiraterone treatment, a ated with poor clinical prognosis (2, 8–10). ERG fusion-positive CYP17A1 enzyme inhibitor used to treat CRPC, compared with prostate cancer patients had significantly lower disease-free sur- ERG fusion-negative patients (5). These data are consistent with vival in watchful waiting cohorts (11, 12), and significantly greater ERG fusion-positive patients having a higher PSA recurrence, and risk of biochemical PSA recurrence compared with fusion-nega- suggests that ERG may regulate the synthesis of intratumoral tive prostate cancer patients (13–15). androgens, and AR activation in CRPC. Upregulation of ABEs and intratumoral androgen synthesis have been shown to facilitate the development of CRPC (20–22). 1Department of Urology, Wayne State University School of Medicine, This can be seen with the ability of abiraterone treatment to reduce Detroit, Michigan. 2Department of Pathology, University of Michigan, serum and intratumoral dihydrotestosterone (DHT) and testos- 3 Ann Arbor, Michigan. Department of Pathology and Molecular Med- terone levels in CRPC patients (23–25). One ABE that has recently icine, Queen's University, Kingston, Ontario, Canada. 4Department of Pathology and Forensic Medicine, University of Sao Paulo at Ribeirao~ become of particular interest is AKR1C3, because it has been Preto, Brazil. 5Department of Pathology, Wayne State University shown to be highly upregulated in CRPC tumors (26). AKR1C3 6 School of Medicine, Detroit, Michigan. Department of Oncology, enzyme functions downstream of CYP17A1 enzyme in the DHT Wayne State University School of Medicine, Detroit, Michigan. synthesis pathway (22), and it may play a pivotal role in DHT Note: Supplementary data for this article are available at Clinical Cancer synthesis in CRPC. Intratumoral DHT synthesis bypasses the Research Online (http://clincancerres.aacrjournals.org/). synthesis of testosterone, and DHT is synthesized directly from Corresponding Author: Sreenivasa R. Chinni, Wayne State University School of androstanedione (5a-Adione) or androstanediol (22, 27, 28). fi Medicine, 540 East Can eld Avenue, Detroit, MI 48201. Phone: 313-577-1833; Inhibitors specific for AKR1C3 are currently being developed (29– Fax: 313-577-0057; E-mail: [email protected] 31), and may prove to be beneficial for treatment of CRPC. doi: 10.1158/1078-0432.CCR-14-2352 This study is focused on TMPRSS2–ERG transcription factor 2015 American Association for Cancer Research. upregulation of AKR1C3 expression and DHT synthesis in CRPC.

www.aacrjournals.org 2569

Powell et al.

lentivector was used for ERG overexpression (Open BioSys- Translational Relevance tems).TruncatedERGandfull-lengthERGwereclonedintothe Intracrine androgen synthesis is a viable target for clinical pLOC lentivector. Cloning primers are listed in Supplementary management of patients undergoing traditional androgen Table S1. Lentiviral particles were generated in HEK293T cells deprivation therapies. Despite newer therapies such as abir- using a Trans-Lentiviral Packaging Kit (Thermo Fisher Scientif- aterone acetate, a drug that targets androgen biosynthetic ic). VCaP cells were infected with pGIPZ lentiviral particles and enzyme (ABE) Cyp17, the disease eventually relapses (abir- selected with (0.3 mg/mL) puromycin (Research Products Inter- aterone resistance). Elucidating molecular mechanisms con- national Corporation). LNCaP and BPH-1 cells were infected tributing to the expression of biosynthetic enzymes in andro- with pLOC lentiviral particles and selected with (2–4 mg/mL) gen production helps to understand the biologic mechanisms blasticidin (Research Products International Corporation). A contributing to the development of castration-resistant pros- pooled lentiviral cell population was used for all lentiviral tate cancer (CRPC). ERG factor regulates key enzymes that experiments. Transient transfections were performed using participate in DHT production in prostate cancer cells. ERG Lipofectamine2000 Reagent (Invitrogen) in OptiMEM media speciﬁcally binds to AKR1C3 gene, regulates its expression, (Invitrogen). promotes DHT production via 5a-androstanedione and results in androgen receptor (AR) activation. Further, both Western blotting and antibodies ERG factor and AKR1C3 are coexpressed in human prostate Western blotting was performed using SDS-PAGE with gel TMPRSS2– tumors and predict lower survival in patients. As transfer to a nitrocellulose membrane. Membranes were blocked ERG fusions are prevalent in prostate cancer, including CRPC, with 5% non-fat milk and probed with primary antibody in 5% and our data present a link between ERG expression, androgen milk. Membranes were probed with horseradish peroxidase– biosynthesis, and AR activation, this study suggests that linked secondary antibody in 5% milk. Protein bands were – TMPRSS ERG fusion-positive patients may have favorable detected using enhanced chemiluminescence substrate and auto- responses to anti-AR therapies. radiography ﬁlm. Primary and secondary antibodies are listed in Supplementary Table S2. Densitometry was performed using ImageJ software (NIH).

We have evidence, suggesting that AKR1C3 functions as a key ABE RT-PCR and RT-qPCR responsible for DHT synthesis via biochemical reduction of 5a- RNA was extracted from VCaP and LNCaP cells using the TRizol Adione into DHT (bypassing testosterone). Our data indicate that (Invitrogen) method. For RT-PCR experiments, cDNA was ampli- 5a-Adione is the primary DHT precursor metabolite, and that 5a- fied using an Eppendorf Mastercycler Realplex2 qPCR machine. Adione can induce high levels of AR activation in CRPC cells. PCR products were loaded onto a 0.8% agarose gel containing Elucidation of TMPRSS2–ERG regulation of AKR1C3 may help to ethidium bromide. For RT-qPCR experiments, cDNA was ampli- stratify ERG fusion-positive prostate cancer patients toward anti– fied with SYBR green from the SensiFAST SYBR Kit obtained from AR-targeted therapies. In addition, our data support the notion Bioline (BIO-98002). Gene expression was quantified with the that AKR1C3 is an attractive therapeutic target for CRPC previously described qPCR method (33). Fold changes in message treatment. levels were quantified using the comparative Ct method (34). PCR primers are listed in Supplementary Table S2. Materials and Methods Cell lines and treatments Chromatin immunoprecipitation PCR Prostate cancer cell lines VCaP, LNCaP, and HEK293T cells were Chromatin immunoprecipitation (ChIP)-PCR was carried obtained from the ATCC. BPH-1 cells were obtained from Dr. out using the ChIP-IT Express kit obtained from Active Motif Shijie Sheng (Wayne State University). VCaP cells were main- as previously described with following modification (17). tained in ATCC DMEM media with 10% FBS (Thermo Fisher Briefly, five million VCaP cells were fixed with formaldehyde Scientific), and 1% penicillin–streptomycin (PS; Gibco). LNCaP and chromatin was extracted according to the manufacturer's and BPH-1 cells were maintained in Gibco RPMI media (Gibco) instructions. For immunoprecipication, 15 to 25 mgofchro- with 10% FBS and 1% PS. HEK293T cells were maintained in matin was incubated with 3 to 5 mgofanti-ERGprimary Gibco DMEM media with 10% FBS, 1% PS, and 1% sodium antibody ERG1/2/3 (H-95)X or normal IgG antibody (Santa pyruvate. All cell lines were tested for Mycoplasma contamination Cruz Biotechnology) and protein G agarose beads. Immuno- before use with the Venor-GeM Mycoplasma Detection Kit from precipitated chromatin was reverse cross-linked with proteinase Sigma Biochemicals. HPLC/MS-MS, immunocytochemistry, and K. PCR was performed using 750 to 1,000 ng DNA template per PSA activation experiments were performed using Gibco DMEM sample and ran at 30 to 40 cycles using an Eppendorf phenol red-free media supplemented with 1% charcoal stripped Mastercycler PCR machine. ChIP-PCR primers are listed in serum (CSS) and 1% PS. Supplementary Table S1. PCR products were analyzed onto a 0.8% agarose gel containing ethidium bromide. ChIP-seq Lentiviral plasmids and transduction methodology and data analysis were performed as previously Lentiviral plasmid selection and virus production were con- described (35). ducted as previously described with minor modifications (32). Briefly, the pGIPZ lentivector was used for shRNA ERG knock- Mass spectrometry (HPLC/MS-MS) down (50-TGGACAGACTTCCAAGATG-30) and shRNA AKR1C3 Androgen metabolites obtained from Steraloids and Sigma- knockdown (50-ACACCAGTGTGTAAAGCTA-30), and the pLOC Aldrich (Supplementary Table S3) were used for MS standards

2570 Clin Cancer Res; 21(11) June 1, 2015 Clinical Cancer Research

ERG Regulation of AKR1C3 Androgen Production and AR Activity

and in vitro steroid treatments. VCaP-shScrambled and shERG Gene expression omnibus database cells were plated at a cell density of 500,000 cells per well in a 6- Gene expression omnibus (GEO) publicly available micro- well plate. Cells were serum starved for 24 hours using 1% CSS array data were used to analyze correlation between ERG phenol red-free DMEM media. Cells were treated with andro- expression and AKR1C3 expression. The 25 metastatic prostate gen metabolites or 0.1% ethanol (vehicle control) for 24 hours cancer expression profiles were extracted from database in 1% CSS phenol red-free media. Cells were treated with (100 GDS2545 using ERG accession ID 914_g_at and AKR1C3 acces- nmol/L) androgen metabolite or a concentration gradient of sion ID 37399_at, and analyzed using "GEO2R." The metastatic (10 nmol/L; 50 nmol/L), or (100 nmol/L) androgen metabo- samples were obtained from warm autopsy program. The in vitro lite. To serve as an internal control, 1 ng of DHEA-D2 and cell line expression profiles were extracted from databases DHT-D3 was added to the media directly before cell lysate GSE22010, GSE39353, and GSE39354 using ERG accession preparation. Cells and media were collected and probe soni- ID's 8070297, 211626_x_at, and 211626_x_at respectively, and cated on ice. Lipids were purified using a Sep-pak cartridge AKR1C3 accession ID's 7925929, 209160_at, and 209160_at obtained from Waters Corporation (WAT023590). Lipid sam- respectively. ples were eluted with acetonitrile into HPLC/MS-MS vials obtained from MicroSOLV Technology (9512S-0CV vials, Oncomine database 9502S-10C-B tops). A cell count control plate was plated and The Oncomine database was used to extract survival data for cell number was determined on the day of sample collection. 363 prostate tumor specimens from Setlur and colleagues "pros- HPLC/MS-MS experiments and data analysis were performed at tate cancer" (2008; ref. 36). Survival data were given as dead or the Wayne State University Lipidomics Core Facility. alive at 1 year, 3 years, 5 years, and endpoint. For survival data past 5 years, the last time to follow-up was used as the endpoint. Immunocytochemistry Prostate tumor expression profiles were extracted from the GEO VCaP cells were plated on cover slips at a cell density of 400,000 database GSE8402 using AKR1C3 accession ID DAP4_3222. cells per cover slip. Cells were serum starved for 24 hours using 1% AKR1C3 expression was determined as high or low by taking the CSS phenol red-free DMEM media. Cells were treated with (100 upper and lower quartile expression value of 363 tumor speci- a a nmol/L) 5 -Androstanedione (5 -Adione) or 0.1% ethanol for mens. This gave a sample size of 91 high tumors and 90 low 24 hours in 1% CSS phenol red-free DMEM media. Cell coverslips tumors, with their respective available survival data taken from fi were washed with 1X PBS, xed with 4% formaldehyde, permea- the Oncomine database. For ERG analysis the presence or absence bilized with 0.5% Triton X-100, blocked with 2% horse serum, – of the TMPRSS2 ERG fusion gene was determined in the Setlur and incubated overnight at 4 C with either (i) AR antibody and colleagues study by FISH or PCR analysis, and these data were (rabbit) or (ii) AR antibody (mouse), 1:1000 in PBS (Santa Cruz made available in the Oncomine database as well as the GEO Biotechnology). Cover slips were washed with PBS and incubated database. Survival data and TMPRSS2–ERG fusion status data at room temperature in a dark box for 1 hour with either (i) Texas were available for analysis from 354 of the prostate tumor red antibody (rabbit; Vector Laboratories) or (ii) Alexa Fluor 633 specimens. antibody (mouse), 1:200 in PBS (Molecular Probes). Cover slips were mounted onto slides using Vectashield mounting medium Statistical analysis with DAPI (Vector Laboratories; H-1200). Slides were subjected to Statistical analyses were performed using GraphPad Prism confocal imaging at the Wayne State University School of Med- Version 5.0. icine Microscopy, Imaging, and Cytometry Resources Core Facility using the Ziess LSM 780 confocal microscope at a magnification of 40. Results TMPRSS2-ERG regulates ABE expression in prostate cancer Immunohistochemistry cells Human prostate cancer tissues from patients undergoing rad- To determine ERG regulation of ABEs in prostate cancer, we ical prostectomy were used for immunohistochemical analysis. performed shRNA knockdown of ERG in TMPRSS2-ERG fusion- Formalin-fixed, paraffin-embedded serial tissue sections from positive VCaP cells using lentivirus transduction. Lentiviral human prostate tumor tissue specimens and LuCaP xenograft shRNA knockdown of ERG resulted in decreased ERG expression tumors were deparaffinized with xylene and rehydrated in graded at the mRNA and protein levels compared with VCaP- EtOH. Endogenous peroxidase activity was blocked by incubating shScrambled control (Fig. 1A). Using VCaP-shScrambled and in 3% H2O2 for 20 minutes. Antigen retrieval was performed with shERG cells, we performed a screening of ABE gene expression. Antigen Retrieval Citra Plus Solution (BioGenex) in a steamer. Among the tested genes, expression of AKR1C3, HSD17B4, and Slides were then blocked with the Blocking Serum from ABC HSD17B6washighlyreducedinVCaPshERGcells,andexpres- Vectastain Kit (Vector Laboratories). Slides were incubated at 4C sion of HSD3B2 and HSD17B10 was also reduced to some overnight in a humidified chamber with antibodies directed extent (Fig. 1B). Using quantitative PCR, we showed that against ERG (1:100; Epitomics) or AKR1C3 (1:5,000; Sigma- AKR1C3, HSD17B4, and HSD17B6 expression was significantly Aldrich). After washing, sections were incubated with the ABC reduced in shERG knockdown cells(Fig.1C).ERGknockdown Vectastain Kit, according to the manufacturer's protocol, followed in VCaP cells resulted in reduced protein expression of AKR1C3 by incubation with 3,3-diaminobenzidine tetrahydrochloride (Fig. 1D) and HSD17B4 enzymes (Supplementary Fig. S1A), (Vector Laboratories). Nuclei were counterstained with Mayer's whereas HSD17B6 protein expression was undetectable (data hematoxylin (Sigma-Aldrich). Sections were then dehydrated not shown). with graded EtOH, washed with xylene, and mounted with Next, we overexpressed the truncated form of ERG that is most Permount (Sigma-Aldrich). commonly found in prostate cancer patients, TMPRSS2 exon 1

www.aacrjournals.org Clin Cancer Res; 21(11) June 1, 2015 2571

Powell et al.

A B VCaP scrambled VCaP shScr VCaP shERG VCaP shScr VCaP shERG VCaP shRNA-ERG 25 30 35 25 30 35 20 25 30 20 25 30

MOI 1 MOI 2 ERG AKR1C3 1.2 30 35 40 30 35 40 30 35 40 30 35 40 1.0 HSD17B6 0.8 GAPDH 0.6 30 35 40 30 35 40 20 25 30 20 25 30 0.4 HSD3B1 HSD17B4 0.2 30 35 40 30 35 40 30 35 40 30 35 40 0.0 mRNA fold change mRNA fold ERG ERG ERG ERG HSD3B2 AKR1C2 30 35 40 30 35 40 25 30 35 25 30 35 VCaP HSD17B1 AKR1C1 MOI 1 MOI 2 30 35 40 30 35 40 30 35 40 30 35 40 shScr shERG shScr shERG HSD17B2 HSD17B10 110.22 0.31 30 35 40 30 35 40 30 35 40 30 35 40 HSD17B3 CYP17A1 ERG 30 35 40 30 35 40 25 30 35 25 30 35 RDH5 SRD5A1 GAPDH 30 35 40 30 35 40 SRD5A2

C DE VCaP scrambled VCaP VCaP shRNA-ERG 1.2 VCaP shScr shERG LNCaP BPH-1 1.15 1 0.14 Ctrl ERG40 Ctrl ERG40 1.0 0.8 ERG ERG

0.6 0.94 1 0.39 115.17 5.08 0.4 AKR1C3 AKR1C3 0.2 mRNA fold change mRNA fold 0.0 GAPDH GAPDH

ERG ERG AKR1C3AKR1C3 HSD17B6HSD17B6HSD17B4HSD17B4

Figure 1. TMPRSS2–ERG regulates ABE expression in prostate cancer cells. A, RT-qPCR analysis of ERG in VCaP-shScrambled (shScr) and VCaP shERG lentiviral cells. Western blot analysis of ERG in VCaP shScr and shERG cells; MOI, multiplicity of infection for lentiviral transduction. B, RT-PCR analysis of ABE expression in VCaP shScr and shERG cells. PCR cycle numbers are indicated above each DNA gel. C, RT-qPCR analysis of HSD17B6, HSD17B4, and AKR1C3 in VCaP shScr and VCaP shERG cells. D, Western blot analysis of AKR1C3 in VCaP, VCaP shScr, and shERG cells. E, Western blot analysis of AKR1C3 in LNCaP ERG40 and BPH-1 ERG40 lentiviral cells. LNCaP Ctrl and BPH-1 Ctrl cells represent empty vector controls. For statistical analysis a two-tailed, paired, Student t test was performed (N ¼ 3); P < 0.01; P < 0.001. Numbers above Western blots represent fold changes determined by densitometry. Representative of three independent experiments.

fused to ERG exon 4 (T1-E4; ref. 37). T1-E4 is also the truncated ERG regulates AKR1C3 expression via direct DNA binding to form of ERG expressed in VCaP cells (Supplementary Fig. S1B). the AKR1C3 gene We termed this truncated form "ERG40" because it has a deletion To investigate whether ERG regulation of AKR1C3 expression of the first 39 amino acids from the N-terminus (38). Lentivirus in prostate cancer occurs through direct DNA binding, we per- was used to overexpress the truncated ERG40 in LNCaP prostate formed ChIP-PCR in VCaP cells. Primers were used to access ERG cancer cells, and in BPH-1 cells (benign prostatic hyperplasia). binding to the AKR1C3 gene region, spanning from 2kb Overexpression of ERG40 resulted in a significant increase in upstream of the transcription start site (TSS) to þ2 kb downstream AKR1C3 mRNA compared with empty vector control (Supple- of the TSS. Several ERG consensus–binding sequence sites (50- mentary Fig. S1C). Overexpression of ERG40 caused an increase GGAA/T-30) were identified in the 2 kb 50untranslated region in AKR1C3 protein expression in both LNCaP cells and BPH-1 (UTR) and in the first intron of the AKR1C3 gene (Fig. 2A). Four cells (Fig. 1E). Interestingly, similar results were seen in LNCaP primer sets were designed to span the 2 kb promoter region (P1- cells overexpressing full-length (wild-type) ERG. LNCaP ERGwt P4) and the intron 1 region (I1–I4; Fig. 2A). ERG was shown to cells showed an increase in AKR1C3 mRNA compared with empty bind directly to the intron 1 region of the AKR1C3 gene, at a site vector control (Supplementary Fig. S1D). These data demonstrate located þ154 to þ670 downstream of the TSS (Fig. 2B). that ERG positively regulates the expression of DHT synthesis The known ERG target gene Plasminogen Activator Urokinase enzyme, AKR1C3, in prostate cancer cells. (PLAU; ref. 39) was also enriched in anti-ERG antibody

2572 Clin Cancer Res; 21(11) June 1, 2015 Clinical Cancer Research

ERG Regulation of AKR1C3 Androgen Production and AR Activity

A AKR1C3 -2 kb 5' UTR Intron 1

Promoter +252 +897 +1018 +1887 -858 +1939 -1860 -1642 -970 -570 -133 +178 +561 +740 +956 +1304 +1624 5' Exon 1 Exon 2 3'

P1 P2 P3 P4 l1 l2 l3 l4

Input ERG IgG NTC B Input ERG IgG NTC Input ERG IgG NTC C PLAU (+) Control P1 l1 AKR1C3 P2 l2 l1

P3 AKR1C3 (-) Control l3 -2.5 Kb

AKR1C3 P4 l4 - -6 Kb ( ) Control

D Scale 50 kb hg19

Chomosome 10: 5,070,000 5,090,000 5,110,000 5,130,000 5,150,000

AKR1C3

Figure 2. ERG binds directly to the AKR1C3 gene. A, diagram illustrating the AKR1C3 gene located on chromosome 10. Black tick marks indicate putative ERG-binding sites (50-GGAA/T-30) located in the region of 2Kbtoþ2 Kb relative to the TSS. B, ChIP-PCR analysis of putative ERG-binding sites in the AKR1C3 gene. ChIP was performed in VCaP cells using anti-ERG antibody; IgG antibody was used as a control for nonspecific binding. NTC is non-template control. C, ChIP-PCR analysis of a known ERG-binding site in PLAU promoter region using VCaP cells. Negative control primers were used in VCaP cells to demonstrate specificity of the ChIP-PCR technique. D, ChIP-Seq analysis performed in duplicate using anti-ERG antibody in VCaP cells. immunoprecipitates, suggesting the validity of the assay. To Androstanedione is a precursor metabolite for DHT in determine the specificity of ERG's interaction with the AKR1C3 TMPRSS2–ERG fusion-positive cells gene, we designed negative control primers that span regions of AKR1C3 participates in testosterone and DHT synthesis in the the AKR1C3 gene that do not contain putative ERG consensus– androgen biosynthetic pathway (Fig. 3A). DHT can be produced binding sites. As expected, there was no ERG binding to these directly from testosterone and indirectly through androstenedi- regions of the AKR1C3 gene (Fig. 2C). one without involving testosterone (testosterone bypass path- ChIP-Seq was performed in VCaP cells with two independent way). Using HPLC/MS-MS analysis, we investigated the produc- experiments using anti-ERG antibody. ERG-binding enrich- tion of DHT from its immediate precursors' as shown in Fig. 3A. ment peaks were observed at three different sites along the Previous studies support the notion that androstendione to 5a- AKR1C3 gene. Two of the binding peaks were located far Adione is the preferred pathway in prostate cancer cells for the upstream of the TSS in the 50-UTR of AKR1C3; one at approx- production of DHT (27). VCaP-shScrambled cells were treated imately 50 kb, and another at 75 kb, relative to the TSS. The with (10 nmol/L; 50 nmol/L), and (100 nmol/L) of 5a-Adione, third peak was shown to be located near the TSS and intron 1 androsterone, and androstanediol, and cells and media were region (Fig. 2D), consistent with the ChIP-PCR data. These data collected for MS analysis. A significant dose-dependent increase demonstrate that ERG directly binds to the AKR1C3 gene and in DHT synthesis was observed following treatment with 5a- regulates AKR1C3 gene expression in fusion-positive prostate Adione (Fig. 3B). These data show that in VCaP cells DHT is cancer cells. produced from the 5a-Adione pathway.

www.aacrjournals.org Clin Cancer Res; 21(11) June 1, 2015 2573

Powell et al.

CYP17A1 A Cholesterol

DHEA

Androstenediol

Androstenedione SRD5A1,2

Testosterone Figure 3. DHT Androstanedione is a precursor metabolite for DHT in TMPRSS2–ERG

SRD5A1,2 fusion-positive cells. A, diagram illustrating the DHT biosynthesis Androstanediol pathway and the enzymes that Androstanedione (5α-Adione) synthesize DHT. B, HPLC/MS-MS analysis in VCaP shSrc cells; DHT levels were quantiﬁed as pmol per one million Androsterone cells. C, HPLC/MS-MS analysis 16 in VCaP shScr and shERG cells; DHT BC16 10 nmol/L VCaP scrambled levels were quantiﬁed as pmol per one 14 14 VCaP shRNA-ERG million cells. Vehicle controls represent 50 nmol/L 0.1% EtOH-treated samples. For 12 12 100 nmol/L statistical analysis a two-tailed, 10 10 unpaired, Student t test was performed 8 8 (N ¼ 3); P < 0.05; P < 0.01; P < 0.001. DHT pmol 6 DHT pmol 6 4 4 2 2 0 0

Vehicle Vehicle 5a-Adione Androsterone Androsterone Androstanediol Androstenediol Androstanedione Androstenedione

To investigate whether ERG activation of AKR1C3 gene therefore, we investigated whether this had any functional con- expression drives androgen synthesis in fusion-positive pros- sequence on AR activation. AR activation can be determined by tate cancer cells, we used HPLC/MS-MS analysis and performed AR translocation into the nucleus, and by AR target gene expres- a substrate feeding experiment comparing VCaP-shScrambled sion. Immunocytochemistry in VCaP cells demonstrated that and shERG cells. ERG knockdown did not significantly change (100 nmol/L) 5a-Adione treatment induced a marked increase the production of testosterone from androstenedione (Supple- in AR translocation into the nucleus compared with vehicle mentary Fig. S2C). However, in ERG knockdown cells DHT controls (Fig. 4A and Fig. 4B). AR colocalization with the nucleus production from 5a-Adione and androsterone was significantly was quantified using Velocity software (PerkinElmer) analysis reduced compared with shScrambled controls (Fig. 3C). Inter- and represented as a global Pearson's correlation coefficient. In estingly, testosterone precursors androstenedione and andros- vehicle-treated cells AR had a weak to moderate correlation with tenediol were converted to DHT at very low levels in VCaP- the nucleus (r ¼ 0.38); however, in 5a-Adione–treated cells AR shScrambled or shERG cells (Fig. 3C), suggesting that the 5a- had a strong correlation with the nucleus (r ¼ 0.72; Supple- Adione pathway has a more profound effect on DHT synthesis. mentary Fig. S3A). RT-qPCR analysis showed a significant AKR1C3 and HSD17B3 are the enzymes responsible for cata- increase in PSA expression in VCaP-shScrambled cells following lyzing the biochemical reduction of 5a-Adione to DHT. How- treatment with (100 nmol/L) 5a-Adione compared with VCaP ever, in VCaP cells PCR data showed that HSD17B3 was shERG cells (Fig. 4C). In addition, RT-PCR data showed that expressed at very low levels, and AKR1C3 was shown to be activation of PSA expression was abrogated in VCaP shERG cells highly expressed. This suggests that AKR1C3 may be the treated with 5a-Adione (Fig. 4D); this further supports the enzyme responsible for the conversion of 5a-Adione to DHT notion that AKR1C3 is the enzyme responsible for the conver- in VCaP cells. Taken together, these data suggest that AKR1C3 sion of 5a-Adione to DHT. To study the direct role of 5a-Adione enzyme catalyzes the biochemical reduction of 5a-Adione to synthesis to DHT in AR activation, we performed shRNA knock- DHT, and that ERG regulates this step through upregulation of down of AKR1C3 in VCaP cells using shAKR1C3 lentivirus AKR1C3 expression in fusion-positive prostate cancer cells. (Supplementary Fig. S3C). RT-qPCR analysis showed a significant increase in PSA expression in VCaP-shScrambled cells Androstanedione induces AR activation through ERG regulated following treatment with (100 nmol/L) 5a-Adione compared AKR1C3 expression with VCaP shAKR1C3 cells (Fig. 4E). Taken together, these data The data above show that androstanedione (5a-Adione) show that ERG enhances 5a-Adione–induced AR activation treatment of VCaP cells caused an increase in DHT synthesis; through direct upregulation of AKR1C3 expression.

2574 Clin Cancer Res; 21(11) June 1, 2015 Clinical Cancer Research

ERG Regulation of AKR1C3 Androgen Production and AR Activity

ABAR:Texas Red Nucleus:DAPI AR/DAPI merge AR/DAPI merge

Vehicle 5a-Adione

Vehicle

5a- Adione

CD E VCaP scrambled VCaP scrambled VCaP shRNA-ERG VCaP shRNA-AKR1C3 VCaP VCaP Scrambled shERG 22 a a 18 20 5 - 5 - Vehicle Adione Vehicle Adione NTC 16 18 14 16 12 14 PSA 12 10 10 8 8 6

PSA mRNA 6 PSA mRNA 4 fold change fold change fold 4 GAPDH 2 2 0 0

Vehicle Vehicle Vehicle Vehicle 5a-Adione 5a-Adione 5a-Adione 5a-Adione

Figure 4. Androstanedione induces AR activation through ERG-regulated AKR1C3 expression. Immunocytochemistry in VCaP cells treated with vehicle or (100 nmol/L) 5a-Adione for 24 hours. Confocal images were shown in A and B; cells were stained with anti-AR, Texas red and DAPI as a green pseudo color. PCR ampliﬁcation of PSA expression is shown in C and D. C, RT-qPCR of PSA expression in VCaP shScr and VCaP shERG lentivirus–infected cells, D, RT-PCR analysis of PSA expression in VCaP shScr and shERG lentivirus–infected cells. NTC is non-template control. E, RT-qPCR analysis of PSA expression in VCaP shScr and shAKR1C3 lentivirus–infected cells. For statistical analysis a two-tailed, unpaired, Student t testwasperformedinC(N ¼ 5), and E (N ¼ 2); P < 0.05; P < 0.001.

ERG and AKR1C3 are coexpressed in human prostate tumor fusion-positive tumors. Analysis of 90 patients by low or high tissue specimens, and predict lower survival probability expression of AKR1C3 and ERG in prostate tumor tissues, showed To determine the expression of ERG and AKR1C3 in human that there was a significant decrease (P ¼ 0.022 for AKR1C3 and tumor tissues, immunohistochemical (IHC) staining of ERG and P < 0.0001 for ERG) in survival for AKR1C3 high-expression AKR1C3 in 64 human prostate tumor tissues was performed. IHC tumors, and for TMPRSS2–ERG fusion-positive tumors staining revealed that ERG and AKR1C3 were coexpressed in (Fig. 5E). The median survival for AKR1C3 low tumors or human prostate tumor tissue (Fig. 5A). The distribution of 64 TMPRSS2–ERG fusion-negative tumors was 12 to 13 years, where- patient tumor tissue samples for the presence or absence of ERG as the median survival for AKR1C3 high tumors or TMPRSS2– and AKR1C3 showed that there was significant coexpression of the ERG fusion-positive tumors was 7 to10 years. Analysis of these two genes (Fig. 5B). ERG and AKR1C3 were also coexpressed in patients by AKR1C3 low and ERG negative versus AKR1C3 high TMPRSS2–ERG fusion-positive LuCaP xenograft tissue specimens and ERG-positive showed that there was a significant decrease (Fig. 5C). Analysis of 25 metastatic prostate tumor specimens (P ¼ 0.017) in survival for AKR1C3 high and ERG-positive extracted from GEO publicly available microarray data revealed a tumors. The median survival for AKR1C3 low and ERG-negative significant correlation (, P ¼ 0.0001; r ¼ 0.69) between ERG tumors was 13 years, whereas the median survival for AKR1C3 and AKR1C3 expression (Fig. 5D). ERG and AKR1C3 expression high and ERG-positive tumors was 8.5 years. also correlated in GEO microarray data from in vitro prostate cell line experiments (Supplementary Fig. S4B–S4D). Data extracted from the Oncomine database containing 363 prostate tumor Discussion tissue specimens were used to perform Kaplan–Meier survival The TMPRSS2–ERG fusions persist in CRPC (low androgen plots of AKR1C3 low- versus high-expression tumors (Supple- state) and fuel tumor growth through oncogenic ERG mechan- mentary Fig. S4A), and TMPRSS2–ERG fusion-negative versus isms. This implies that in order for ERG to be expressed in tumors,

www.aacrjournals.org Clin Cancer Res; 21(11) June 1, 2015 2575

Powell et al.

AKR1C3 ERG AB*** P < 0.0001 N = 64 AKR1C3+ AKR1C3–

ERG+ 24 9

ERG– 6 25

C AKR1C3 ERG D GEO: GDS2545 Metastatic prostate cancer 300 N = 25 P = 0.0001 250 *** 200 r = 0.69 150 ERG 100 50 0 10 100 1,000 10,000 AKR1C3

E AKR1C3 Low (N = 90) TMPRSS2-ERG (–) (N = 290) AKR1C3 Low and TMPRSS2-ERG (–) (N = 78) AKR1C3 High (N = 91) TMPRSS2-ERG (+) (N = 61) ARK1C3 High and TMPRSS2-ERG (+) (N =19) 110 110 110 100 100 90 100 90 90 80 P < 0.0001 P < 0.05 80 P < 0.05 80 70 70 70 60 60 60 50 50 50 40 40 40 30 30 30 20 20 20 Percent survival Percent 10 survival Percent 10 survival Percent 10 0 0 0 0 100 200 300 0 100 200 300 0 50 100 150 200 250 Month elapsed Months elapsed Months elapsed after diagnoses after diagnoses after diagnoses

Figure 5. ERG and AKR1C3 are coexpressed in human prostate tumor tissue specimens, and predict lower survival probability. A, immunohistochemical analysis of AKR1C3 and ERG in human prostate tumor tissue specimens. B, distribution of 64 patients by presence or absence of AKR1C3 and ERG in prostate tumor tissues, P < 0.0001. C, immunohistochemical analysis of AKR1C3 and ERG in LuCaP xenograft specimens. D, the correlation plot of AKR1C3 and ERG expression using microarray data from 25 metastatic prostate cancer samples. Data were extracted from GEO database. E, the Kaplan–Meier survival plot of (left) AKR1C3 low tumors versus AKR1C3 high tumors (middle), ERG fusion-negative tumors versus ERG fusion-positive tumors, and (right) AKR1C3 low and TMPRSS2–ERG– negative versus AKR1C3 high and TMPRSS2–ERG–positive tumors. Data were extracted from Oncomine database. For statistical analyses of Kaplan–Meier survival, a Mantel–Cox test and a Gehan-Breslow–Wilcoxon test were performed.

there should be at least some active androgen synthesis and AR It has been shown previously that androgens can regulate signaling. As advanced tumors often have low systemic androgen AKR1C3 expression. Speciﬁcally, high androgen levels suppress exposure, it is well accepted that tumors gain the capability of activation of AKR1C3 expression, whereas androgen-starved con- androgen synthesis for tumor growth, but the transcriptional ditions activate expression of AKR1C3 (22, 40). This provides mechanisms contributing to the androgen biosynthetic machin- negative and positive regulation where in the presence of high and ery are not known. Here, we describe a model in which AR low androgen, respectively, AKR1C3 expression is regulated as a activates expression of the TMPRSS2–ERG transcription factor, means to balance intracellular androgen levels. Androgen inhi- which then activates androgen synthesis via regulation of ABE, bition of AKR1C3 is thought to be regulated, in part, by AR- AKR1C3. Under this model, AKR1C3 can use 5a-Adione as a induced recruitment of lysine-speciﬁc demethylase 1 (LSD1) to substrate and synthesize intratumoral DHT to provide continual the AKR1C3 gene (40). Whether TMPRSS2–ERG plays a role in AR activation in a castrate environment. This can fuel AR signal- androgen-induced regulation of AKR1C3 remains unclear; ing, as well as provide a feed-forward activation loop although, microarray data from Supplementary Fig. S4B suggest of TMPRSS2–ERG and AKR1C3 expression in prostate tumors that ERG positively regulates AKR1C3 expression independent of (Fig. 6). androgen status.

2576 Clin Cancer Res; 21(11) June 1, 2015 Clinical Cancer Research

ERG Regulation of AKR1C3 Androgen Production and AR Activity

DHT that the ERG fusion-gene product regulates the key enzymes in DHT Castration-resistant DHT DHT prostate cancer the pathway. Consistent with this idea, AKR1C3 expression is DHT growth low or absent in fusion-negative cells and higher in fusion- positive cells (26). This gene is highly expressed in CRPC DHT AR AR patients, thus, fusion status may dictate expression of AKR1C3 DHT and DHT production. On the basis of our data, fusion-positive cells have higher DHT TMPRSS2: propensity to synthesize DHT from DHEA ! androstendione ERG ! 5a-Adione ! DHT. In this pathway 3b-hydroxysteroid DHT dehydratase (HSD3B) catalyzes the rate-limiting step in con- shRNA- version of adrenal steroids such as DHEA to androstenedione. ERG Tumors have been shown to express a gain-of-function muta- AKR1C3 tion in HSD3B1, which increases the stability of the protein and fl 5a-Adione shRNA- provides a higher ux of DHT production from adrenal steroid AKR1C3 precursors (42). Our studies also show that, in VCaP cells, ERG regulates HSD3B2 (Fig. 1B) gene expression; thus, not only Figure 6. enzyme stability, but overexpression of this enzyme may con- Schematic representation of proposed role of TMPRSS2–ERG fusions in tribute to DHT production. Taken together, our data demon- androgen biosynthesis and AR activation via feed-forward activation. strate that ERG participates in the expression of multiple ABEs in the DHT biosynthetic pathway. Our data support the hypothesis that, in TMPRSS2–ERG fusion-positive cells, DHT is produced from the 5a-Adione The ChIP and ChIP-seq data from our study (Fig. 2) show a pathway, and this drives AR activation. However, the finding novel transcription mechanism, whereby ERG transcription factor that AKR1C3 may catalyze 5a-Adione into DHT is a novel directly binds to the AKR1C3 gene locus and activates gene finding that lends further support for therapeutically targeting expression. We cannot rule out direct ERG factor regulation of AKR1C3 in prostate cancer patients. It may be beneficial in the AR activity, as Yu and colleagues (19) demonstrated that ERG future to screen prostate cancer patients for levels of 5a-Adione can occupy AR targets and regulate their expression. In such and AKR1C3 in their tumor biopsies. This study suggests cases, ERG can override AR repressive activity and directly that TMPRSS2–ERG fusion-positive prostate cancer patients regulate AKR1C3 expression. AKR1C3 genome analysis (Fig. may have favorable response rates to anti–AR-targeted thera- 2A and B) supports this idea as shown by direct binding of ERG pies, such as enzalutamide, abiraterone, or AKR1C3-specific to AKR1C3 intron and 50-UTR. Another implication of Yu and inhibitors. colleagues study is that ERG inhibits AR signaling involving prodifferentiation function and shifts cells to stem cell dedif- Disclosure of Potential Conflicts of Interest ferentiation and oncogenesis. As our data support the notion I. Asangani is a consultant/advisory board member for Oncofusion. M.L. that ERG activates androgen synthesis through AKR1C3 expres- Cher reports receiving speakers bureau honoraria from Astellas-Medivation sion, the tumor-expressed androgens can mediate growth func- and is a consultant/advisory board member for Bayer, Dendreon, and tion through AR activation. As ERG depends on AR for its own Janssen Biotech. No potential conflicts of interest were disclosed by the expression, ERG-induced androgens are a link for sustained other authors. expression of ERG in advanced tumors via a feed-forward loop (Fig. 6). Sustained expression of ERG, restored AR activity, and Authors' Contributions expression of ABEs exist together in CRPC (41); our study Conception and design: M.L. Cher, S.R. Chinni provides a mechanistic link among these entities, which in Development of methodology: M. Katie Conley-LaComb, Y.-M. Wu, J.A. concert promote tumor growth. Squire, K.R. Maddipati, S.R. Chinni Intracrine androgen synthesis has been charged with activa- Acquisition of data (provided animals, acquired and managed patients, tion of AR and tumor growth despite gonadal depletion of provided facilities, etc.): K. Powell, L. Semaan, M.K. Conley-LaComb, I. Asangani, K.R. Maddipati, S.R. Chinni androgens. Recent reports demonstrate that DHT production in Analysis and interpretation of data (e.g., statistical analysis, biostatistics, tumors sustains AR-mediated tumor growth, and strongly computational analysis): K. Powell, L. Semaan, M. Katie Conley-LaComb, implicate the involvement of ABEs in DHT production (27). I. Asangani, K.B. Ginsburg, K.R. Maddipati, S.R. Chinni Specifically, this study demonstrates that DHT synthesis Writing, review, and/or revision of the manuscript: K. Powell, M. Katie Conley- bypasses testosterone production and that 5a-Adione is readily LaComb, J. Williams, K.R. Maddipati, M.L. Cher, S.R. Chinni converted to DHT. Our data show that, among various DHT Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Semaan, I. Asangani, J. Williams, S.R. Chinni precursors, 5a-Adione is more potent in making DHT in ERG- Study supervision: M.L. Cher, S.R. Chinni positive cells (Fig. 3). In addition, AKR1C3 has been shown to be highly expressed in VCaP cells (41) and our data demonstrate that ERG regulates AKR1C3 expression in fusion-positive Acknowledgments cells. Even though it can participate in production of both The authors thank Dr. Arul Chinnaiyan and Dr. Xuhong Cao for sharing fi testosterone and DHT, our data suggest that ERG-knockdown ChiP-Seq data. The authors thank Dr. Daniel Bon l for providing LuCaP xenograft tumor slides for IHC analysis. The authors thank Dr. Michael cells have severe impairment in production of DHT whereas Tainsky and Dr. Raymond Mattingly of Wayne State University for critically testosterone production is maintained. This suggests that tumor reading the article. The authors thank lipidomics core facility for perform- cells have preference for DHT production over testosterone and ing MS analysis androgens and androgen precursors. Lipidomics core was

www.aacrjournals.org Clin Cancer Res; 21(11) June 1, 2015 2577

Powell et al.

supported by National Center for Research Resources, NIH grant The costs of publication of this article were defrayed in part by the S10RR027926. payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate Grant Support this fact. The work was supported by the U.S. Department of Defense, W81XWH-09-1- 0250, and NIH-NCI grant CA151557, The Fund for Cancer Research NIH-NCI Cancer Center support grant CA-22453, and a pre-doctoral training grant Received September 9, 2014; revised January 22, 2015; accepted February 19, T32CA008531 (to K. Powell). 2015; published OnlineFirst March 9, 2015.

References 1. Mehra R, Tomlins SA, Shen R, Nadeem O, Wang L, Wei JT, et al. 18. Leshem O, Madar S, Kogan-Sakin I, Kamer I, Goldstein I, Brosh R, et al. Comprehensive assessment of TMPRSS2 and ETS family gene aberra- TMPRSS2/ERG promotes epithelial to mesenchymal transition through tions in clinically localized prostate cancer. Mod Pathol 2007;20: the ZEB1/ZEB2 axis in a prostate cancer model. PloS ONE 2011;6: 538–44. e21650. 2. Perner S, Demichelis F, Beroukhim R, Schmidt FH, Mosquera JM, Setlur S, 19. Yu J, Mani RS, Cao Q, Brenner CJ, Cao X, Wang X, et al. An integrated et al. TMPRSS2:ERG fusion-associated deletions provide insight into the network of androgen receptor, polycomb, and TMPRSS2-ERG gene fusions heterogeneity of prostate cancer. Cancer Res 2006;66:8337–41. in prostate cancer progression. Cancer Cell 2010;17:443–54. 3. Han B, Mehra R, Lonigro RJ, Wang L, Suleman K, Menon A, et al. Fluo- 20. Pfeiffer MJ, Smit FP, Sedelaar JP, Schalken JA. Steroidogenic enzymes and rescence in situ hybridization study shows association of PTEN deletion stem cell markers are upregulated during androgen deprivation in prostate with ERG rearrangement during prostate cancer progression. Mod Pathol cancer. Mol Med 2011;17:657–64. 2009;22:1083–93. 21. Rasiah KK, Gardiner-Garden M, Padilla EJ, Moller G, Kench JG, Alles 4. Rickman DS, Chen YB, Banerjee S, Pan Y, Yu J, Vuong T, et al. ERG MC, et al. HSD17B4 overexpression, an independent biomarker of poor cooperates with androgen receptor in regulating trefoil factor 3 in prostate patient outcome in prostate cancer. Mol Cell Endocrinol 2009;301: cancer disease progression. Neoplasia 2010;12:1031–40. 89–96. 5. Attard G, Swennenhuis JF, Olmos D, Reid AH, Vickers E, A'Hern R, et al. 22. Mitsiades N, Sung CC, Schultz N, Danila DC, He B, Eedunuri VK, et al. Characterization of ERG, AR and PTEN gene status in circulating tumor Distinct patterns of dysregulated expression of enzymes involved in cells from patients with castration-resistant prostate cancer. Cancer Res androgen synthesis and metabolism in metastatic prostate cancer tumors. 2009;69:2912–8. Cancer Res 2012;72:6142–52. 6. Baweja R, Calhoun S, Singareddy R. Sleep problems in children. Minerva 23. Efstathiou E, Titus M, Tsavachidou D, Tzelepi V, Wen S, Hoang A, et al. Pediatr 2013;65:457–72. Effects of abiraterone acetate on androgen signaling in castrate-resistant 7. Tomlins SA, Rhodes DR, Perner S, Dhanasekaran SM, Mehra R, Sun X-W, prostate cancer in bone. J Clin Oncol 2012;30:637–43. et al. Recurrent Fusion of TMPRSS2 and ETS Transcription Factor Genes in 24. Ryan CJ, Smith MR, de Bono JS, Molina A, Logothetis CJ, de Souza P, et al. Prostate Cancer. Science 2005;310:644–8. Abiraterone in metastatic prostate cancer without previous chemotherapy. 8. FitzGerald L, Agalliu I, Johnson K, Miller M, Kwon E, Hurtado-Coll A, et al. N Engl J Med 2013;368:138–48. Association of TMPRSS2-ERG gene fusion with clinical characteristics and 25. de Bono JS, Logothetis CJ, Molina A, Fizazi K, North S, Chu L, et al. outcomes: results from a population-based study of prostate cancer. BMC Abiraterone and increased survival in metastatic prostate cancer. N Engl Cancer 2008;8:230. J Med 2011;364:1995–2005. 9. Attard G, Clark J, Ambroisine L, Fisher G, Kovacs G, Flohr P, et al. 26. Hamid AR, Pfeiffer MJ, Verhaegh GW, Schaafsma E, Brandt A, Sweep FC, Duplication of the fusion of TMPRSS2 to ERG sequences identifies fatal et al. Aldo-keto reductase family 1 member C3 (AKR1C3) is a biomarker human prostate cancer. Oncogene 2007;27:253–63. and therapeutic target for castration-resistant prostate cancer. Mol Med 10. Yoshimoto M, Joshua AM, Cunha IW, Coudry RA, Fonseca FP, Lud- 2012;18:1449–55. kovski O, et al. Absence of TMPRSS2:ERG fusions and PTEN losses in 27. Chang KH, Li R, Papari-Zareei M, Watumull L, Zhao YD, Auchus RJ, prostate cancer is associated with a favorable outcome. Mod Pathol et al. Dihydrotestosterone synthesis bypasses testosterone to drive 2008;21:1451–60. castration-resistant prostate cancer. Proc Natl Acad Sci U S A 2011; 11. Hagglof C, Hammarsten P, Stromvall K, Egevad L, Josefsson A, Stattin P, 108:13728–33. et al. TMPRSS2-ERG expression predicts prostate cancer survival and 28. Mohler JL, Titus MA, Wilson EM. Potential prostate cancer drug target: associates with stromal biomarkers. PloS ONE 2014;9:e86824. bioactivation of androstanediol by conversion to dihydrotestosterone. 12. Demichelis F, Fall K, Perner S, Andren O, Schmidt F, Setlur SR, et al. Clin Cancer Res 2011;17:5844–9. TMPRSS2:ERG gene fusion associated with lethal prostate cancer in a 29. Liedtke AJ, Adeniji AO, Chen M, Byrns MC, Jin Y, Christianson DW, et al. watchful waiting cohort. Oncogene 2007;26:4596–9. Development of potent and selective indomethacin analogues for the 13. Barwick BG, Abramovitz M, Kodani M, Moreno CS, Nam R, Tang W, et al. inhibition of AKR1C3 (Type 5 17beta-hydroxysteroid dehydrogenase/ Prostate cancer genes associated with TMPRSS2-ERG gene fusion and prostaglandin F synthase) in castrate-resistant prostate cancer. J Med Chem prognostic of biochemical recurrence in multiple cohorts. Br J Cancer 2013;56:2429–46. 2010;102:570–6. 30. Jamieson SM, Brooke DG, Heinrich D, Atwell GJ, Silva S, Hamilton EJ, et al. 14. Nam RK, Sugar L, Yang W, Srivastava S, Klotz LH, Yang LY, et al. Expression 3-(3,4-Dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acids: highly potent of the TMPRSS2:ERG fusion gene predicts cancer recurrence after surgery and selective inhibitors of the type 5 17-beta-hydroxysteroid dehydroge- for localised prostate cancer. Br J Cancer 2007;97:1690–5. nase AKR1C3. J Med Chem 2012;55:7746–58. 15. Bonaccorsi L, Nesi G, Nuti F, Paglierani M, Krausz C, Masieri L, et al. 31. Adeniji AO, Chen M, Penning TM. AKR1C3 as a target in castrate resistant Persistence of expression of the TMPRSS2:ERG fusion gene after pre- prostate cancer. J Steroid Biochem Mol Biol 2013;137:136–49. surgery androgen ablation may be associated with early prostate specific 32. Singareddy R, Semaan L, Conley-Lacomb MK, StJohn J, Powell K, Iyer M, antigen relapse of prostate cancer: preliminary results. J Endocrinol Invest et al. Transcriptional regulation of CXCR4 in prostate cancer: significance of 2009;32:590–6. TMPRSS2-ERG fusions. Mol Cancer Res 2013;11:1349–61. 16. Carver BS, Tran J, Gopalan A, Chen Z, Shaikh S, Carracedo A, et al. Aberrant 33. Chinni SR, Sivalogan S, Dong Z, Filho JC, Deng X, Bonfil RD, et al. CXCL12/ ERG expression cooperates with loss of PTEN to promote cancer progres- CXCR4 signaling activates Akt-1 and MMP-9 expression in prostate cancer sion in the prostate. Nat Genet 2009;41:619–24. cells: the role of bone microenvironment-associated CXCL12. Prostate 17. Cai J, Kandagatla P, Singareddy R, Kropinski A, Sheng S, Cher ML, et al. 2006;66:32–48. Androgens induce functional CXCR4 via ERG factor expression in 34. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using TMPRSS2-ERG fusion positive prostate cancer cells. Transl Oncol 2010;3: real-time quantitative PCR and the 2(-Delta Delta C(T)) method. Methods 195–203. 2001;25:402–8.

2578 Clin Cancer Res; 21(11) June 1, 2015 Clinical Cancer Research

ERG Regulation of AKR1C3 Androgen Production and AR Activity

35. Asangani IA, Dommeti VL, Wang X, Malik R, Cieslik M, Yang R, et al. 39. Tomlins SA, Laxman B, Varambally S, Cao X, Yu J, Helgeson BE, et al. Role Therapeutic targeting of BET bromodomain proteins in castration-resistant of the TMPRSS2-ERG gene fusion in prostate cancer. Neoplasia 2008;10: prostate cancer. Nature 2014;510:278–82. 177–88. 36. Setlur SR, Mertz KD, Hoshida Y, Demichelis F, Lupien M, Perner S, 40. Cai C, He HH, Chen S, Coleman I, Wang H, Fang Z, et al. Androgen receptor et al. Estrogen-dependent signaling in a molecularly distinct gene expression in prostate cancer is directly suppressed by the androgen subclass of aggressive prostate cancer. J Natl Cancer Inst 2008;100: receptor through recruitment of lysine-speciﬁc demethylase 1. Cancer Cell 815–25. 2011;20:457–71. 37. Wang J, Cai Y, Ren C, Ittmann M. Expression of variant TMPRSS2/ERG 41. Cai C, Wang H, Xu Y, Chen S, Balk SP. Reactivation of androgen receptor- fusion messenger RNAs is associated with aggressive prostate cancer. regulated TMPRSS2:ERG gene expression in castration-resistant prostate Cancer Res 2006;66:8347–51. cancer. Cancer Res 2009;69:6027–32. 38. King JC, Xu J, Wongvipat J, Hieronymus H, Carver BS, Leung DH, et al. 42. Chang KH, Li R, Kuri B, Lotan Y, Roehrborn CG, Liu J, et al. A gain-of- Cooperativity of TMPRSS2-ERG with PI3-kinase pathway activation in function mutation in DHT synthesis in castration-resistant prostate cancer. prostate oncogenesis. Nat Genet 2009;41:524–6. Cell 2013;154:1074–84.

www.aacrjournals.org Clin Cancer Res; 21(11) June 1, 2015 2579

ERG/AKR1C3/AR Constitutes a Feed-Forward Loop for AR Signaling in Prostate Cancer Cells

Katelyn Powell, Louie Semaan, M. Katie Conley-LaComb, et al.

Clin Cancer Res 2015;21:2569-2579. Published OnlineFirst March 9, 2015.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-14-2352

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2015/03/18/1078-0432.CCR-14-2352.DC1

Cited articles This article cites 42 articles, 10 of which you can access for free at: http://clincancerres.aacrjournals.org/content/21/11/2569.full#ref-list-1

Citing articles This article has been cited by 6 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/21/11/2569.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/21/11/2569. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.