Oncogene (2008) 27, 4865–4876 & 2008 Macmillan Publishers Limited All rights reserved 0950-9232/08 $30.00 www.nature.com/onc ORIGINAL ARTICLE Identification of ETS-like 4 as a novel androgen target in prostate cancer cells

H Makkonen1,TJa¨ a¨ skela¨ inen1, T Pitka¨ nen-Arsiola1,4, M Rytinki1, KK Waltering2,MMa¨ tto¨ 3, T Visakorpi2 and JJ Palvimo1

1Institute of Biomedicine/Medical Biochemistry, University of Kuopio, Kuopio, Finland; 2Institute of Medical Technology, University of Tampere, and Tampere University Hospital, Tampere, Finland and 3Institute of Clinical Medicine/Clinical Microbiology, University of Kuopio, Kuopio, Finland

Transcriptional control byandrogens via androgen Introduction receptor (AR) is stronglyinvolved in prostate cancer development, but the critical target have remained Androgens have a key role in male sexual development elusive. We have characterized E twenty-six-like tran- and act via (AR) that belongs to the scription factor 4 (ELK4) (also known as serum response superfamily. Androgens have a crucial factor accessoryprotein 1) as a novel AR target in human role in development of prostate cancer, which is the prostate cancer cells. In-silico screening identified three most widespread neoplasm in men in Western countries. putative AR response elements (AREs) within À10 kb Androgen-dependency was first recognized by Huggins from the transcription start site of ELK4. Both ARE1 at and Hodges (1941), who showed that the removal À167/À153 and ARE2 at À481/À467 bound AR in vitro of androgens leads to regression of prostate cancer and mediated androgen induction as isolated elements in (Culig et al., 2002; Heinlein and Chang, 2004). There- transcription assays in non-prostate cells. However, fore androgen ablation therapy is used to delay cancer merelythe ARE2 that cooperates with a proximal development, but the therapy fails eventually and cancer forkhead box A1-binding site was critical for the AR- turns into a hormone-refractory disease (Stewart et al., dependent activation of ELK4 in prostate cancer 2005). The mechanism how prostate cancer transforms cells. Preferential loading of holo-AR onto the ARE2 and from an androgen-dependent to an androgen-indepen- concomitant recruitment of RNA polymerase II onto the dent stage is incompletely understood (Feldman and ELK4 promoter was confirmed in prostate cancer cells by Feldman, 2001). chromatin immunoprecipitation. Database searches indi- Forced overexpression of AR can transform andro- cated that the expression of ELK4 is markedlyincreased gen-dependent prostate cancer cells to androgen- in prostate cancers relative to normal prostates. More- independent ones (Chen et al., 2004). In concordance over, prostate cancer tissue immunostainings showed that to these experimental findings, common amplification nuclear ELK4 levels are significantlyincreased in andro- and overexpression of AR is associated with hormone- gen-refractoryprostate cancers compared to untreated refractory clinical prostate cancer specimens (Visakorpi tumours. Reduction of the amount of ELK4 in LNCaP et al., 1995; Linja et al., 2001). Also overexpression of cells byRNAi retarded cell growth. In conclusion, ELK4 AR co-activators has been implicated in progression of is a direct AR target in prostate cancer cells. Androgens prostate cancer grade, stage and decreased disease-free maythus contribute to the growth of prostate cancer via survival, but the results are inconsistent (Gregory et al., influencing ELK4 levels. 2001; Culig et al., 2004; Linja et al., 2004). Mutations Oncogene (2008) 27, 4865–4876; doi:10.1038/onc.2008.125; and polymorphisms in the AR may also play a role published online 12 May 2008 in the development of prostate cancer (Han et al., 2005). For example, mutations in the ligand-binding domain of Keywords: androgen receptor; ETS transcription factor; the AR have been found in prostate cancers treated with prostate cancer; FoxA1 antiandrogens, such as flutamide (Taplin et al., 1995). Only a small number of androgen-responsive genes have been well characterized in prostate cancer cells. The best studied gene is prostate-specific antigen (PSA) that is a member of kallikrein (KLK) gene family Correspondence: Professor JJ Palvimo, Institute of Biomedicine/ (Riegman et al., 1991). The level of PSA in Medical Biochemistry, University of Kuopio, PO Box 1627, Kuopio plasma is important for prostate cancer diagnostics and FI-70211, Finland. monitoring prostate cancer state (Maroni and Craw- E-mail: Jorma.Palvimo@uku.fi ford, 2006). PSA (KLK3) contains two AR response 4Current address: Department of Biosciences/Applied Biotechnology, University of Kuopio, PO Box 1627, FI-70211 Kuopio, Finland. elements (AREs) in its proximal promoter (PSAprom) Received 12 March 2007; revised 13 March 2008; accepted 25 March and a more distal ARE unit/enhancer, with the latter 2008; published online 12 May 2008 AR-binding unit being more important for the androgen 4866 Oncogene

8 ELK1 8 ELK3 8 ELK4a 8 PSA LNCaP LNCaP LNCaP LNCaP 7 7 7 7 6 6 6 6 5 5 5 5 4 4 4 4 3 3 3

foldchange 3 foldchange foldchange foldchange 2 2 2 2 1 1 1 1

0 0 0 0 of Regulation 0 1.5 36 12 18 0 1.5 36 12 18 0 1.5 36 12 18 0 1.5 36 12 18 time (h) time (h) time (h) time (h)

5 5 7 7 ELK4a ELK4a EtOH LNCaP EtOH LNCaP ELK4 ELK4b ELK4b R1881 R1881 Makkonen H 6 6 4 4 R1881+CHX R1881+Actinomycin D androgens by LNCaP PC-3 5 5

3 3 4 4 al et

2 2 3 3 fold change fold change fold change fold change 2 2 1 1 1 1

0 0 0 0 0123 0123 PSA ELK4a PSA ELK4a time (h) time (h)

200 175 150 time (h) 125

0312 ELK4a 100 55 kDa ELK4 75

40 kDa relative 50

55 kDa mRNA expression -TUB 25 40 kDa 0

PC-3 LAPC4 DU145

LNCaP(AR+)LNCaP(AR-) Regulation of ELK4 by androgens H Makkonen et al 4867 response (Cleutjens et al., 1996, 1997; Schuur et al., ELK3 in LNCaP cells by employing quantitative RNA 1996). analysis with PSA as a reference gene. As shown in Recently, about half of the prostate cancers were Figures 1c and d, PSA mRNA and also ELK4 mRNA shown to contain genetic rearrangements leading to accumulated in LNCaP cells in response to treatment fusion of transmembrane protease, serine 2 (TMPRSS2) with synthetic androgen R1881. First signs of mRNA gene and either ERG, ETV1 or ETV4 gene that induction were seen within 3 h after the addition of E twenty-six (ETS) transcription factors. The most R1881, and the levels of the mRNAs continued to common form of these fusions is TMPRSS2:ERG. ERG increase for at least next 15 h (PSA B7-fold (Po0.001) is a putative oncogene whose expression is low in non- and ELK4 B5-fold (Po0.001) increase after 18 h). In malignant prostate, whereas TMPRSS2 is highly contrast to ELK4 mRNA, neither ELK1 nor ELK3 expressed in prostate. The expression of TMPRSS2 is mRNA was stimulated by R1881 (Figures 1a and b). If androgen-regulated, and therefore, the TMPRSS2:ERG anything, there was a modest decrease in their mRNA rearrangement leads to androgen-dependent and abun- levels after androgen treatment. ELK4 mRNA was dant expression of the fusion transcript in prostate similarly (B4-fold, Po0.05) induced by R1881 in VCaP cancer cells. These genetic alterations have given first prostate cancer cells. Two ELK4 mRNA splice variants hints at the strong androgen dependence of prostate a and b have been reported (Dalton and Treisman, cancer development (Tomlins et al., 2006). 1992). Both of the ELK4 isoforms were induced by In this study, we have characterized androgen- androgen in LNCaP cells with the accumulation of dependent regulation of a novel AR target gene, ETS- isoform a being somewhat more pronounced than that like transcription factor 4 (ELK4), that belongs to the of isoform b (Figure 1e). As expected, androgen did not ternary complex factor (TCF) subfamily of ETS domain induce ELK4 in AR-deficient PC-3 prostate cancer cells transcription factors (Dalton and Treisman, 1992; Shaw (Figure 1f). Moreover, there was a positive correlation and Saxton, 2003; Buchwalter et al., 2004). The latter between the expression level of ELK4 and the presence factors were first described in the context of c-fos gene of functional AR in several other prostate cancer cell regulation (Shaulian and Karin, 2002; Buchwalter et al., lines (Figure 1j). 2004). Interestingly, analysis of data- To test whether new protein synthesis is required bases indicates that the expression of ELK4 is increased for the hormonal induction of ELK4, LNCaP cells in prostate cancer. Moreover, we found that hormone- were treated with cycloheximide (CHX), a protein refractory prostate cancers contain significantly more synthesis inhibitor, during the induction with R1881. ELK4 in their nuclei compared to androgen-sensitive Interestingly, CHX attenuated the androgen induction prostate cancer specimens. We suggest that AR can of ELK4 mRNA to about half, but it also decreased the influence prostate growth also via ELK4-regulated induction of PSA mRNA, a classic direct androgen genes. target gene, by B40% (Figure 1g). Therefore, even though we cannot rule out the possibility that a labile androgen-induced protein is involved in the regulation of ELK4, we suggest that a major part of Results the androgen regulation occurs through a direct action of AR on the gene promoter. Moreover, since addition Androgens induce ELK4 in prostate cancer cells of actinomycin D, a transcription inhibitor, abolished As ETS domain transcription factor ELK4 has been the androgen induction of ELK4 mRNA in LNCaP recently listed among genes overexpressed in prostate cells (Figure 1h), androgens act at the level of cancer (Edwards et al., 2005) and as perturbed regula- transcription rather than influencing the stability of tion of at least two other ETS factor genes EGR and ELK4 mRNA. Immunoblotting of LNCaP cell extracts ETV1 has been implicated in a significant number of with anti-ELK4 antibody showed that the amount prostate cancer samples (Tomlins et al., 2005, 2007), we of immunoreactive ELK4 is clearly increased after 12- chose to analyse potential androgen regulation of ELK4 h induction with R1881, which is in agreement with the and the other closely related TCF genes ELK1 and mRNA data (Figure 1i).

Figure 1 Androgen-induced accumulation of E twenty-six (ETS)-like transcription factor 4 (ELK4) mRNA and protein in LNCaP cells. LNCaP human prostate carcinoma cells or PC-3 cells were treated with 10 nM R1881 for the indicated times. Quantitative RT–PCR was used to measure the mRNA levels of (a) ELK1,(b) ELK3,(c) ELK4 and (d) prostate-specific antigen (PSA) in LNCaP cells. ELK4 a and b mRNA isoforms were compared in (e) LNCaP cells and (f) PC-3 cells. (g) LNCaP cells were treated 18 h with 10 nM R1881 in the absence and presence of cycloheximide ( þ CHX, 10 mg/ml). (h) LNCaP cells were treated 18 h with 10 nM R1881 in the absence and presence of actinomycin D (1 mg/ml). Quantitative RT–PCR was used to measure the mRNA levels of ELK4 and PSA. Total RNA levels between samples were normalized using mRNA levels of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). ÀðDDCtÞ Fold changes were calculated using the formula 2 , where DDCt is DCt(R1881)ÀDCt(EtOH), DCt is Ct(gene X)–Ct(GAPDH) and Ct is the cycle at which the threshold is crossed. Columns represent the mean of three experiments and bars indicate standard deviations. (i) Immunoblot analysis of ELK4 in LNCaP cells at indicated times after addition of R1881. The lower panel shows a-tubulin levels in the same samples. Representative gels of three experiments are shown. The protein lanes originate from the same experiment and analysis, but they have been regrouped as indicted by dividing lanes. (j) Quantitative RT–PCR analyses to measure the relative mRNA levels of ELK4 in indicated androgen receptor (AR)-positive (LNCaP(AR þ ) and LAPC4) and AR-negative (LNCaP(ARÀ), DU154 and PC-3) prostate cancer cell lines were carried out as described (Linja et al., 2004).

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4868 Identification of putative androgen response elements in ARE3 ARE2 ARE1 -481 -467 the ELK4 promoter -6014AGGACAgaaCTTTCT-6000 GATACTgcaTGTTCC -167AGTACAttaTGCTCT-153 Next, we were interested in whether ELK4 is directly regulated by AR and revealing the regulatory elements ELK4 mediating the regulation. We performed in-silico analy- -7000 -6000 -5000 -4000 -3000 -2000 -1000 TSS sis up to À10 kb from the transcription start site of -6101ELK4dis-5867 -507ELK4-ARE2-330 -212ELK4-ARE1-21 ELK4 using CONSITE program (http://mordor.cgb.ki. se/cgi-bin/CONSITE/consite) to identify potential -507ELK4prom-21 AREs in the ELK4 promoter. The analysis suggested three putative AREs (ARE1–3) two of which (ARE1 C3(1) ELK4 and ARE2) are located in the PSAprom region and ARE ARE1 ARE2 ARE3 the third (ARE3) in the distal promoter (Figure 2a). To test capability of the putative AREs to bind AR, we performed electrophoretic mobility shift assays (EMSAs) with AR produced by ectopic expression in -+++ + - +++ + - ++++++ + - + COS-1 cells. The intronic ARE of the prostatic C3(1) AR gene that binds efficiently AR in vitro was used as a R1881 -- ++ + - -++ +--++-++ +- + positive control (Kallio et al., 1994). EMSAs showed non-spec ab -- -+ + - --+ +---+--+ +- + that AR does not interact with ARE3, but the receptor -- -- + - --- +------+ +- - is capable of binding to both ARE1 and ARE2, albeit anti-AR less efficiently than to the positive control ARE Figure 2 Binding of androgen receptor (AR) to putative E twenty- (Figure 2b). Quantification of the R1881-bound AR– six (ETS)-like transcription factor 4 (ELK4) AR response elements (AREs) in vitro.(a) The sequences of putative AREs and their DNA complexes (in the absence of antibodies) by schematic localizations. Bold regions depict the promoter regions phosphoimager analysis indicated that the intensity of amplified in chromatin immunoprecipitation (ChIP) assays. AR–ARE1 complexes was about one-fifth (19±1%) (b) Electrophoretic mobility shift assays were performed in and that of AR–ARE2 complexes about one-tenth the presence of AR-containing COS-1 cell extracts (AR þ )orextracts ± from empty vector-transfected cells (ARÀ), and 32P-labelled ds- (9 2%) of the intensity of AR–C3(1)ARE complexes. AREs. Protein–DNA complexes were resolved on non-denaturing To confirm that the observed DNA–protein complexes 4% polyacrylamide gels. Anti-AR antibody was used to confirm were caused by AR, not due to other AR-induced the specificity of AR–DNA complexes, and normal rabbit in COS-1 cell lysates, we used anti-AR immunoglobulin G (IgG) (non-spec ab) was used as non-specific antibody in EMSA reactions. As shown in Figure 2b, control. A representative gel of three experiments is shown. ARE1-, ARE2- and C3(1)ARE–protein complexes were clearly upshifted in the presence of the anti-AR efficiently than that containing ARE1 (Figure 3e). These antibody, but not with a non-specific antibody, whereas ChIP data verify that holo-AR is loaded onto the ELK4 ARE3 again showed no evidence for interaction promoter under bona fide chromatin conditions. with AR.

Functionality of the putative ELK4 AREs Loading of holo-AR to the ELK4 promoter in vivo Next, we assessed the individual importance of the To investigate protein occupancy of the chromatin ARE1 and ARE2 in living cells by performing reporter regions bearing the ELK4 promoter in vivo,we gene assays (RGAs) with ELK4 promoter fragments performed chromatin immunoprecipitation (ChIP) cloned in front of luciferase gene (LUC) and used 5.8-kb using antibodies specific for AR and RNA polymerase PSA enhancer/promoter-driven luciferase (PSA5.8- II (PolII) with primer pairs covering ELK4 proximal LUC) as a positive control. The constructs were (ELK4prom) and distal (ELK4dis) promoter transfected to LNCaP cells and the cells were grown (Figure 2a). LNCaP cells were treated with or without with or without androgen. ELK4 promoter fragment androgen for 120 min. After cross-linking and sonica- (À209/ þ 54) that contains only ARE1, yielded only a tion, chromatin samples were immunoprecipitated with minor (o2-fold) increase in luciferase activity in the specific antibodies or immunoglobulin G (IgG) control, presence of androgen, whereas ELK4 promoter frag- and recovered DNAs were used as templates in ment (À543/ þ 54) containing both ARE1 and ARE2 quantitative PCR. Promoter and enhancer regions of conferred significant B11-fold (Po0.001) androgen PSA served as positive controls (Figures 3a and b) induction in the LUC activity (Figure 4a). Surprisingly, (Kang et al., 2004). As shown in Figures 3c and d, a much longer ELK4 promoter fragment (À6086/ þ 54) loading of AR and recruitment of PolII onto the yielded practically no androgen induction, suggesting ELK4prom were clearly increased in response to that the upstream ELK4 region harbours elements androgen stimulation, whereas the occupancy of AR capable of silencing the promoter. To further test the or that of PolII on ELK4dis was negligible in importance of ARE1 and ARE2 in the androgen comparison to ELK4prom, also after the androgen induction, we mutated both AREs (ARE1m: AA- treatment. Higher-resolution ChIP analyses with shorter TAAAttaTGCTCT, ARE2m: GATACTgcaTTTTTC) chromatin fragments and additional primers showed individually or in combination in the context of the that the region bearing ARE2 recruited AR more À543/ þ 54 promoter fragment. In line with the above

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4869 80 0 min PSAenh 80 0 min PSAprom 80 0 min ELK4dis 70 120 min 70 120 min 70 120 min 60 60 60 50 50 50 40 40 40 % of input % of input 30 30 % of input 30 20 20 20 10 10 10 0 0 0 AR PolII IgG AR PolII IgG AR PolII IgG

80 0 min ELK4prom 12 0 min AR-ChIP 70 120 min 10 120 min 60 8 50 40 6

% of input 30 4 % of input 20 2 10 0 0 AR PolII IgG

-ARE2 -ARE1

ELK4 ELK4 Figure 3 Loading of androgen receptor (AR) and recruitment of PolII onto the E twenty-six (ETS)-like transcription factor 4 (ELK4) promoter in LNCaP cells. Cells were treated with 10 nM R1881 for the indicated times and chromatin immunoprecipitation assays were used to investigate the occupancy of AR and PolII on the ELK4 proximal (d) and distal (c) promoter regions (see Figure 2a). (e) Differentiation of AR binding between the ELK4 AR (ARE) 2 and the ELK4 ARE1 using higher-resolution chromatin immunoprecipitation (ChIP) assays with shorter chromatin fragments and primer pairs that are selective for two the AREs (Supplementary Table S1). ChIP samples were used as templates in quantitative PCR. Enhancer (a) and promoter (b) regions of prostate-specific antigen (PSA) were used as references. Results are shown as percentages of the input samples. The percentages ÀðDCtÞ were calculated using the formula 2 Â100, where DCt is Ct(ChIP-template)ÀCt(Input). Columns represent the mean±s.d. of three experiments.

RGA data, mutation of the ARE2 alone (or in induction in isolation, but they were both less active combination with the ARE1), but not that of the than the positive control ARE (Figure 4e; B6- and B5- ARE1, abolished the androgen induction of ELK4 fold induction, respectively, vs 28-fold induction). promoter in LNCaP cells (Figure 4a). Also in VCaP Mutation of ARE1 or that of ARE2 practically prostate cancer cells that contain endogenous AR, abolished the androgen induction. merely ARE2 was sufficient for conferring androgen induction (Figure 4b). Similar results were obtained in PC-3 cells co-transfected with an AR expression AR cooperates with FoxA1 in the regulation of ELK4 construct, albeit the androgen induction of the promoter promoter constructs was weaker in these cells than in the two AR- The above reporter gene data imply cell specificity in the positive prostate cancer cells (Figure 4c). Interestingly, function of the ARE2 and involvement of other the ELK4 promoter fragment À209/ þ 54 that showed promoter binding and cooperating transcription factors. practically no androgen induction in LNCaP cells was In-silico analyses predicted the presence of a forkhead induced by androgen in a fashion comparable to that of box A1 (FoxA1)-binding site at À449/À438 in the close the À543/ þ 54 construct in HeLa cervical carcinoma proximity of the ARE2. To test the importance of the cells co-transfected with AR expression construct putative FoxA1 site for the androgen induction of ELK4 (Figure 4d; B15- and B17-fold induction, respectively). promoter in prostate cancer cells, we changed the In HeLa cells, both ARE mutations clearly attenuated underlined Ts of the sequence À449ATATGTTT the androgen effect, albeit the effect of ARE2 mutation GCATÀ438 to the Gs in context of ELK4-543/ þ 54 was slightly more deleterious than that of ARE1 promoter fragment and compared the activity of mutation. To test the functionality of ARE1 and mutated fragment to that of the wild-type promoter ARE2 in isolation, we cloned individual AREs and fragment in conferring androgen induction. Interest- their mutated versions in front of a TATA box-driven ingly, the mutation resulted in B50% reduction in the LUC and co-transfected those with AR expression androgen induction of ELK4 promoter in LNCaP and vector to COS-1 cells. The C3(1)ARE–TATA construct VCaP cells, whereas, in non-prostate HeLa or COS-1 was used as a positive control. In agreement with the cells, it did not have any marked effect on the androgen EMSA data, the ELK4 ARE2 did not dramatically induction (Figure 5a). Notably, the mutation of the differ from the ELK4 ARE1 in conferring androgen FoxA1-binding site did not influence the activity of the

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4870 470x LNCaP 60 VCaP 360 27x 320 55 50 23x 18 45 16 14x 40 14 11x 35 5.0x 12 30 10 25 8 20 6 15 4 1.4x 10

relative luciferace activity 1.1x 0.9x 1.6x

relative luciferace activity 2.6x 2 0.6x 0.9x 5 0.7x 2.2x 4.1x 0 0 4 pGL3 PSA pGL3 PSA -209/+54-543/+54 ARE1mARE2m -209/+5-543/+54 ARE1mARE2m -6086/+54 ARE1+2m -6086/+54 ARE1+2m

PC-3 HeLa 60 41x 55 15x 50 50 10 45 9 2.1x 40 8 3.1x 35 7 30 6 25 5 15x 4 3.1x 20 17x 3 15 1.1x 1.3x 1.5x 10 2 1.1x 7.2x 6.2x relativeluciferaceactivity

relative luciferace activity 4.4x 1 5 0.3x 2.3x 0 0 4 4 pGL3 PSA pGL3 PSA -209/+54-543/+54 ARE1mARE2m -209/+54-543/+54 ARE1mARE2m -6086/+5 ARE1+2m -6086/+5 ARE1+2m

70 28x COS-1 50 30 10 9 8 6.3x 7 6 5 5.2x 4 3 2.6x 2 0.7x

relative luciferace activity 1 1.8x 1.5x 0

TATAC3(1)ARE1ARE2ARE3 ARE1mARE2m Figure 4 Transcriptional activity of the E twenty-six (ETS)-like transcription factor 4 (ELK4) promoter fragments as assessed by reporter gene assays. LNCaP (a), VCaP (b), PC-3 (c), HeLa (d) or COS-1 (e) cells were transfected with luciferase gene (LUC) constructs driven by indicated promoter regions of ELK4 or prostate-specific antigen (PSA). Co-transfection of pCMVb and b-galactosidase activity was used for normalization of transfection efficiency. For PC-3, HeLa, and COS-1 analyses, pSG5-hAR was co-transfected with the reporter constructs. Cells were treated with vehicle (ethanol) or 10 nM R1881 for 16 h before harvesting the cells for reporter analyses. Results are shown as relative LUC activity and fold inductions of androgen-treated samples in the relation to the activity of ethanol-treated samples are shown above the columns. Columns represent the mean±s.d. of three independent experiments.

ELK4-543/ þ 54 promoter in the absence of androgen. accumulation of ELK4 mRNA. Silencing of GATA2 ChIP assays with FoxA1-specific antibody indicated that has recently been implicated in the regulation of that in prostate cancer cells, the interaction of the some AR target genes was used as an additional control FoxA1 with the ELK4 promoter is enhanced in the (Wang et al., 2007). FoxA1 siRNA and GATA2 siRNA presence of androgen (Figure 5b). To confirm were similarly efficient in reducing their target protein the importance of FoxA1 in the regulation of ELK4, levels (Figure 5c). As shown in Figure 5d, the androgen we silenced the FoxA1 by RNAi in LNCaP cells and induction of ELK4 mRNA was severely compromised studied the effect of silencing on androgen-induced (Po0.001) in FoxA1-ablated cells in comparison to

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4871 RGA EtOH ChIP 100 4 FoxAmut 90 R1881 VCaP 80 ELK4prom 3 70 60

50 2 40 % of input

% of wt induction 30 1 20 10 0 0 FoxA1 IgG

VCaP HeLa LNCaP Cos-1

siRNA SCR FoxA1 EtOH ELK4a qRT-PCR 6 R1881 -+-+ R1881 LNCaP FoxA1 5

α-TUB 4

siRNA SCR GATA2 3 R1881 -+-+ fold change fold GATA2 2

α -TUB 1

0 siSCR siFoxA1 siGATA2 Figure 5 Role of forkhead box A1 (FoxA1) in the regulation of E twenty-six (ETS)-like transcription factor 4 (ELK4) by androgen receptor (AR). (a) LNCaP, VCaP, HeLa and COS-1 cells were transfected with luciferase gene (LUC) constructs driven by wild-type (wt) or FoxA1-binding site mutated (FoxAmut) ELK4-543/ þ 54 promoter. For HeLa and COS-1 cell, pSG5-hAR was co-transfected with the reporter constructs. Co-transfection of pCMVb and b-galactosidase activity was used for normalization of transfection efficiency. Cells were treated with vehicle (ethanol) or 10 nM R1881 for 16 h before harvesting for reporter analyses. Results are shown as percentage of androgen induction of wt ELK4-543/ þ 54 promoter. Columns represent the mean±s.d. of three independent experiments. (b) Binding of FoxA1 to the ELK4 promoter in VCaP cells. Cells were treated with 10 nM R1881 for 2 h and chromatin immunoprecipitation assays using anti-FoxA1 antibody were used to investigate the occupancy of FoxA1 on the ELK4 proximal promoter region. (c) Immunoblot analysis of FoxA1 siRNA-, GATA2 or negative control scrambled (SCR) siRNA-treated LNCaP cells grown in the presence ( þ ) and absence (À) of R1881. The lower panels show a-tubulin levels in the same samples. (d) Quantitative RT–PCR analysis of ELK4 mRNA levels in LNCaP cells transfected with FoxA1 siRNA, GATA2 siRNA, or control SCR siRNA for 72 h, and subsequently, treated with vehicle (ethanol) or 10 nM R1881 for 24 h. Columns represent the mean±s.d. of three experiments.

GATA2-silenced cells or scrambled (SCR) control androgen by B85% and in the presence of androgen by siRNA-treated cells. These results strongly suggest that B70% compared to control siRNA-treated levels, the cooperation between the ARE2 and the adjacent resulting in marked silencing effects on the ELK4 FoxA1 site is required for the androgen induction of the protein level (Figure 6a). Interestingly, the growth of ELK4 in prostate cancer cells. ELK4 siRNA-treated LNCaP cells was significantly retarded in the absence of androgen in comparison to ELK4 influences LNCaP cell growth the control siRNA-transfected cells (Po0.01 and Because of the potential involvement of ELK4 in cell Po0.05 for ELK4-1 and ELK4-2 siRNA, respectively, growth, we attenuated its expression by RNAi in at 72 h; Po0.001 for both ELK4 siRNAs at 96 h) LNCaP cells and monitored the effect of the silencing (Figure 6b). Addition of androgen stimulated the cell on cell proliferation. To that end, we used ELK4-specific growth, and also under these conditions, ELK4 siRNAs siRNAs that were designed and synthesized in a fashion showed a tendency to inhibit cell proliferation. These that minimizes off-target effects (Jackson et al., 2006). data indicate that the ELK4 can contribute to the The two most potent ELK4 siRNAs decreased the proliferation of LNCaP cells, especially under dimin- amount of ELK4 mRNA levels in the absence of ished levels of androgen.

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4872 siRNA SCR ELK4-1 ELK4-2 PSA value in untreated tumours, immunostainings of R1881 - + - + - + the TMAs showed that in androgen-refractory prostate cancers, nuclear ELK4 levels are significantly ELK4 (Po0.0001) increased in comparison to untreated α-TUB tumours (Table 1).

1.0 siSCR siELK4-1 siELK4-2 Discussion 0.8 ELK4 (SRF accessory factor-1, SAP-1) belongs to the 0.6 TCF subfamily of the ETS domain transcription factors that are classic targets for mitogen-activated protein A492 0.4 kinases (MAPKs). Other members in this subfamily are ELK1 and ELK3 (NET, SAP-2 or ERP). The TCFs bind and activate serum response elements together with 0.2 (SRF). The TCF–SRF complexes regulate many immediate early genes, such as c-fos, 0.0 encoding a subunit of AP-1 (Shaulian and Karin, 2002; R1881 --- + + + --- +++ Shaw and Saxton, 2003; Buchwalter et al., 2004). TCFs, 72h 96h especially ELK4 and ELK1, appear to have at least partly overlapping biological roles, which may explain Figure 6 Silencing of E twenty-six (ETS)-like transcription factor 4 (ELK4) retards the growth of LNCaP cells. (a) Immunoblot the relatively subtle defects of Elk1(À/À) and Elk4(À/À) analysis of ELK4 protein levels in LNCaP cells treated with 10 nM mice (Ayadi et al., 2001; Zheng et al., 2003; Cesari et al., R1881 or vehicle (ethanol) and transfected with two different 2004; Costello et al., 2004). ELK1 and other ETS ELK4 siRNAs (ELK4-1 or ELK4-2) or control scrambled (SCR) domain transcription factors ERG and PDEF have siRNA (Dharmacon ON-TARGETplus siRNAs) for 96 h. The recently been implicated in the growth of prostate cancer lower panel shows a-tubulin levels in the same samples. (b) Cell proliferation was measured using CellTiter96 AQueous cell cells (Oettgen et al., 2000; Xiao et al., 2002; Tomlins proliferation assay reagent (Promega) at indicated time points as et al., 2005). Interestingly, recent molecular concept absorbance at 492 nm. The experiment was repeated three times modelling of prostate cancer progression showed strong with comparable results. enrichment of ELK1-binding sites in the promoters of genes overexpressed in prostatic intraepithelial neopla- sia vs benign prostate (Tomlins et al., 2007). Activation Increased expression of ELK4 in prostate cancer of TCFs in response to MAPK signalling is well To reveal potential differences in the expression of characterized, but very little is known about their ELK4 between normal and tumour prostate, we regulation by hormones. Our results showing retrieved gene expression data from the Oncomine that ELK4, but not ELK1 or ELK3, is upregulated by website (www.oncomine.org). Statistical analysis of the androgens indicate that the TCFs are differentially data from Yu et al. (2004) revealed that ELK4 regulated, even though they are likely to have over- expression is significantly increased in prostate cancer lapping roles in transcription regulation. In the absence in comparison to non-malignant prostate, with meta- of active MAPK signalling, ELK3/NET has been shown static cancer showing higher expression than locally to act through its NET inhibitory domain (NID) and C- occurring cancer (Figure 7a). In addition, three other terminal-binding protein inhibition domain as a potent microarray studies display increased ELK4 levels in repressor of transcription (Buchwalter et al., 2004). prostate cancer compared to normal prostate tissue Since also ELK4 harbours an NID, it may similarly (Luo et al., 2001; Magee et al., 2001; Welsh et al., 2001). repress transcription in the absence of sustained growth We next wanted to analyse the cellular distribution of factor stimulation. ELK4 protein in prostate samples using immunohisto- In addition to direct genomic effects, androgens have chemical analysis. As shown in Figure 7b, ELK4 been reported to regulate cell physiology indirectly via localizes mainly to the epithelial cells in non-malignant influencing protein kinase pathways and/or other prostate, with the nuclear compartment showing more transcription factors (Heinlein and Chang, 2002). We intense immunoreactivity than the cytoplasm. We also therefore tested whether the androgen induction of examined whether the clinicopathological parameters ELK4 was conferred by direct binding of AR to the and the ELK4 levels correlate in untreated and hormone ELK4 promoter that was predicted to harbour three refractory cancers. To that end, tissue microarrays potential AREs. Two of the in silico-predicted AREs, corresponding to untreated prostate cancers and locally ARE1 (at À167/À153) and ARE2 (at À481/À467), recurrent hormone-refractory prostate cancers were bound AR in vitro and both of them were functional in analysed with anti-ELK4 antibody, and nuclear and transcription assays in HeLa and COS-1 cells. However, cytoplasmic stainings were scored separately. Although in AR-positive prostate cancer cells, merely the ARE2 these analyses did not reveal an apparent correlation was capable of conferring the androgen induction of the between ELK4 intensity and Gleason score, pT stage or ELK4 promoter, indicating cell type-specific differences

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4873 in the function of ELK4 AREs. These differences are DNA constructs due to cooperation between the ARE2 and its proximal Promoter regions of ELK4 were PCR-amplified from human FoxA1-binding site in prostate cancer cells. In addition genomic DNA using Phusion DNA polymerase (Finnzymes, to the binding of FoxA1, ChIP analyses verified the androgen-induced loading of AR and the recruitment of P<0.001 RNA polymerase II onto the ELK4 promoter under P<0.05 bona fide chromatin environment. Our results thus 0.3 confirm that the ELK4 is directly regulated by the AR, P<0.01 but similarly to some other AR target genes in prostate 0.2 cancer cells (Gao et al., 2003; Wang et al., 2007), the receptor has to cooperate with another DNA-binding factor, FoxA1 to induce the ELK4 gene. 0.1 The higher abundance of ELK4 in prostate as compared to most other human tissues (expressed 0.0 sequence tag information at http://genome-www5.stan- ford.edu) suggests that it has an interesting role in the -0.1 regulation of prostate physiology. Microarray data ELK4 mRNA expression indicate that ELK4 expression is further increased in -0.2 prostate cancers compared to non-malignant prostates BP PC MCP (Oncomine database) (Edwards et al., 2005). Our finding that the growth of LNCaP prostate cancer cells can be BPH retarded by reducing the cellular ELK4 suggests that the protein promotes proliferation of prostate cancer cells. Recently, other ETS transcription factors, especially ERG, have been shown to be overexpressed in prostate cancers due to genetic rearrangements leading to androgen regulation of the ERG (Tomlins et al., 2005). Since ELK4 itself is androgen-regulated, its enhanced expression in prostate cancer can take place without genetic alterations affecting ELK4. Our data showing that nuclear levels of ELK4 protein are significantly elevated in hormone-refractory tumours compared to untreated cancers further suggest that ELK4 could be an important contributor to the progression of prostate cancer. untreated PC

Materials and methods

Cell culture LNCaP, VCaP, HeLa, PC-3, COS-1, DU145 cells were obtained from American Type Culture Collection and main- tained as described (Linja et al., 2004; Karvonen et al., 2006). LAPC4 cells were kindly provided by Dr Charles Sawyers (UCLA).

Figure 7 Expression of E twenty-six (ETS)-like transcription hormone-refractory PC factor 4 (ELK4) in prostate tumours. (a) Gene expression data of ELK4 were retrieved from the Oncomine website (www.oncomi- ne.org), and the data from Yu et al. (2004) was used for statistical calculations. The data were re-analysed to show expression levels of ELK4 in benign prostate (BP), prostate carcinoma (PC), and metastatic prostate cancer (MPC). Horizontal lines represent medians. Difference of expression compared to median of BP group is shown in the y axis. P-values were calculated using unpaired one-way analysis of variance (ANOVA) and Bonferroni post hoc test. (b) Representative examples of benign prostate hyperplasia (BPH), untreated and hormone-refractory prostate cancer samples stained with anti-ELK4 antibody (see Supplemen- tary Figure). In non-malignant prostate, moderate anti-ELK4 immunoreactivity can be detected both in cytoplasmic and nuclear compartments.

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4874 Table 1 Expression of ELK4 in prostate cancer according to immunohistochemistry Nuclearstaining a Cytoplasmic staininga No. (%) No. (%)

0–1 2–3 P-value 0–1 2–3 P-value

Untreated tumours 106 (53) 93 (47) 111 (56) 88 (44) Hormone-refractory tumours 14 (16) 76 (84) o0.0001b 52 (58) 38 (42) 0.7984b

Untreated tumours Gleason score o7 40 (38) 30 (33) 36 (33) 34 (40) 7 56 (53) 44 (49) 57 (52) 43 (50) >7 9 (9) 16 (18) 0.1575c 16 (15) 9 (10) 0.5267c

pT-stage 2 74 (70) 60 (65) 75 (69) 59 (67) 3 32 (30) 32 (35) 0.5435b 35 (31) 29 (33) 0.8796b

PSA (mean±s.d.) 17.4 (25.1) 15.7 (13.8) 0.8327d 19.0 (26.0) 13.6 (9.8) 0.1766d

For construction of tissue microarray (TMA) formalin-fixed paraffin-embedded tumour blocks of 199 untreated prostate cancers (prostatectomy specimens) and 90 locally recurrent hormone-refractory (transurethral resection specimens) prostate cancers were obtained from the Tampere University Hospital. The hormone-refractory tumours were from patients who had been treated with following modalities: 41 orchiectomy, 31 LHRH (luteinizing hormone-releasing hormone) analog, 3 estrogen, 1 bicalutamide, 2 orchiectomy+estrogen, 10 combined androgen blockade (CAB), 1 CAB+estramustine and 1 unknown The median time between diagnosis and progression was 34 months (range 4–160 months). The immunostaining was performed using anti-ELK4 antibody. Nuclear and cytoplasmic stainings were scored separately from 0 to 3. Staining without primary antibody was used as a negative control. aStaining 0–1, no or weak staining; 2–3, moderate to strong staining. bFisher’s exact test. cw2-Test. dMann–Whitney U-test.

Espoo, Finland) and cloned into pGL3-basic (Promega, (Fermentas, Vilnius, Lithuania) following manufacturer’s Madison, WI, USA). ARE–TATA constructs were produced instructions. cDNA was used as a template in quantitative by ligation of ds-oligomers into the TATA box-containing real-time PCR, which was carried out using M Â 3000P pGL3-basic. All sequences were verified by DNA sequencing (Stratagene), ABsolute QPCR SYBR Green Mix (Abgene, using the ALFexpress system. pSG5-hAR and pPSA5.8-LUC Epson, UK) and specific primers (Supplementary Table S1). have been described (Karvonen et al., 2006). pCMV encoding b-galactosidase was from Clontech (Mountain View, CA, Immunoblotting USA). Mutations were constructed with QuikChange II LNCaP cells were seeded, grown and treated as prior to Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, quantitative real-time RT–PCR. SDS PAGE and immuno- USA) according to manufacturer’s instructions. blotting were carried out as described (Karvonen et al., 2006). Antigen-antibody complexes were detected using ECL Plus Antibodies western Blotting Detection System (GE Healthcare) according Primary antibodies anti-AR (sc-816), anti-ELK4 (anti-Sap1a) manufacturer’s instructions. (sc-13030), anti-PolII (sc-899), anti-FoxA1 (sc-6553), anti- GATA2 (sc-9008), anti-a-tubulin (sc-5286) and normal rabbit Electrophoretic mobility shift assay IgG (sc-2027) were from Santa Cruz Biotechnology (Santa For production of AR for EMSAs, COS-1 cells were Cruz, CA, USA). Secondary antibodies were purchased from transfected either with empty pSG5 or pSG5-hAR and cell Zymed (South San Francisco, CA, USA): Zymax goat anti- extracts were prepared as described (Thompson et al., 2001). rabbit IgG (H þ L) horseradish peroxidase (HRP) conjugate Cell extract (12 mg) was used for 20-ml EMSA reaction and (81-6120) and Zymax goat anti-mouse IgG (H þ L) HRP incubated with or without R1881 (10 nM) and specific antibody conjugate (81-6520). or normal rabbit IgG (200 ng) for 10 min at 22 1C before addition of 32P-labelled ds-oligomer probe. Protein–DNA Quantitative real-time RT–PCR complexes were allowed to form for 1 h (Palvimo et al., LNCaP or PC-3 cells were seeded into six-well plates (330 000 1993). Probes contained the 29-bp native sequences cells per well) and grown 48 h in transfection medium (LNCaP: (Supplementary Table S1). The complexes were separated on RPMI 1640, 5% charcoal-stripped fetal bovine serum (FBS) 4% non-denaturing PAGE. The gels were dried and detected (CCS-FBS), 2 mML-glutamine, 1 mM sodium pyruvate, 10 mM using phosphoimager (FLA3000, Fuji). 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 25 mM glucose; PC-3: F-12 (Ham), 5% CCS-FBS). Cells were treated Chromatin immunoprecipitation with or without 10 nM R1881 (Perkin Elmer Inc., Waltham, The experiments were performed essentially as previously MA, USA). Total RNA was extracted using TRIZOL reagent described (Kang et al., 2004). LNCaP cells were seeded (Invitrogen Life Technologies, Carlsbad, CA, USA) and resulting at B50% confluence and allowed to grow in converted to cDNA using M-MuLV Reverse Transcriptase transfection medium for 48 h before treating with 10 nM

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4875 R1881 for 2 h. For higher-resolution ChIP analyses, prolonged LNCaP cells were transfected with siRNAs (40 nM final sonication of chromatin yielded fragments of an average concentration) using Lipofectamine 2000 according to manu- length of 200 bp instead of 600 bp. ChIP samples were used as facturer’s protocol (Invitrogen). For silencing of ELK4, templates in quantitative real-time PCR. ChIP template (5 ml) Dharmacon’s (Lafayette, CO, USA) chemically modified was used per PCR. Specific primers for different promoter ON-TARGETplus siRNAs 50-CGACACAGACATTGATTC regions are listed in Supplementary Table S1. PCRs were ATT-30 (ELK4-1) and 50-GAGAATGGAGGGAAAGATAT carried out with ABsolute QPCR SYBR Green Mix and with T-30 (ELK4-2) were used (Jackson et al., 2006), with ON- M Â 3000P. TARGETplus SCR siRNA as a control. For proliferation assays, LNCaP cells were seeded onto 96-well plates (5000 cells Reporter gene assays per well) in transfection medium. After 24 h, LNCaP cells were LNCaP cells were seeded onto 12-well plates (140 000 cells per transfected with siRNAs (40 nM final concentration) using well) and grown overnight in transfection medium. Cells were Lipofectamine 2000 according to manufacturer’s protocol transfected with reporter gene constructs (1.8 mg pLUC and (Invitrogen) and the cells were supplemented with 10 nM 0.2 mg pCMVb per well) using jetPEI transfection reagent R1881 or vehicle (ethanol). After 72 and 96 h, samples were (Polyplus-transfection). PC-3 (140 000 cells per well), HeLa analysed using CellTiter96 AQueous cell proliferation assay (60 000 cells per well), and COS-1 (140 000 cells per well) cells reagent (Promega) according to manufacturer’s instructions. were seeded onto 12-well plates and grown overnight in culture The quantity of formazan reaction product as measured by the medium. The medium was replaced 4 h before transfection absorbance at 492 nm is directly proportional to the number of with transfection medium (F-12 (Ham) containing 5% CCS- living cells in culture. In parallel, portions of the cells were FBS for PC-3 cells, DMEM containing 2.5% CCS-FBS and collected for qRT–PCR and immunoblotting. non-essential amino acids for HeLa cells, DMEM containing 2.5% CCS-FBS for COS-1 cells), and the cells were co- Prostate cancer tissue stainings transfected with LUC construct (0.46 mg per well), pSG5-hAR Tissue microarrays were prepared as described (Kononen (0.02 mg per well) and pCMVb (0.02 mg per well) using et al., 1998) and analysed using anti-ELK4 antibody. Trans IT-LT1 transfection reagent (Mirus Bio Corporation, Specificity of ELK4 staining was confirmed by pre-adsorbtion Madison, WI, USA). One day after transfection, cells were with bacterially produced and purified GST fusion of ELK4 grown for 16 h with either vehicle (ethanol) or R1881 and lysed amino acids 154–320 corresponding to the epitope region of in Reporter Lysis Buffer (Promega) and LUC activity was the antibody (Supplementary Figure S1). measured with Luciferase Assay System (Promega) using Luminoskan Ascent (Thermo Electron, Helsinki, Finland) luminometer and b-galactosidase activity as described (Palvi- Acknowledgements mo et al., 1993). This work was supported by grants from the Academy of RNAi and cell proliferation assays Finland, Association for International Cancer Research, For silencing of FoxA1 and GATA2, Qiagen custom siRNAs Finnish Cancer Foundation, Sigrid Juse´ lius Foundation, and control SCR were used. Target sequences for the FoxA1, Kuopio Naturalists’ Society, Helena Vuorenmies Foundation, GATA2 and SCR were 50-GAGAGAAAAAATCAACAGC- Karjalan Sivistysseura and Finnish Union of Experts in 30,50-ACCCTTAGCAGCCCAGCAT-30 and 50-AATTCTCC Science. We thank Merja Ra¨ sa¨ nen for skilful technical GAACGTGTCACGT-30, respectively (Wang et al., 2007). assistance.

References

Ayadi A, Zheng H, Sobieszczuk P, Buchwalter G, Moerman P, Alitalo Culig Z, Comuzzi B, Steiner H, Bartsch G, Hobisch A. (2004). K et al. (2001). Net-targeted mutant mice develop a vascular Expression and function of androgen receptor coactivators in phenotype and up-regulate egr-1. EMBO J 20: 5139–5152. prostate cancer. J Steroid Biochem Mol Biol 92: 265–271. Buchwalter G, Gross C, Wasylyk B. (2004). Ets ternary complex Culig Z, Klocker H, Bartsch G, Hobisch A. (2002). Androgen transcription factors. Gene 324: 1–14. receptors in prostate cancer. Endocr Relat Cancer 9: 155–170. Cesari F, Brecht S, Vintersten K, Vuong LG, Hofmann M, Klingel K Dalton S, Treisman R. (1992). Characterization of SAP-1, a protein et al. (2004). Mice deficient for the ets transcription factor elk-1 recruited by serum response factor to the c-fos serum response show normal immune responses and mildly impaired neuronal gene element. Cell 68: 597–612. activation. Mol Cell Biol 24: 294–305. Edwards S, Campbell C, Flohr P, Shipley J, Giddings I, Te-Poele R Chen CD, Welsbie DS, Tran C, Baek SH, Chen R, Vessella R et al. et al. (2005). Expression analysis onto microarrays of randomly (2004). Molecular determinants of resistance to antiandrogen selected cDNA clones highlights HOXB13 as a marker of human therapy. Nat Med 10: 33–39. prostate cancer. Br J Cancer 92: 376–381. Cleutjens KB, van der Korput HA, van Eekelen CC, van Rooij HC, Feldman BJ, Feldman D. (2001). The development of androgen- Faber PW, Trapman J. (1997). An androgen response element in a independent prostate cancer. Nat Rev Cancer 1: 34–45. far upstream enhancer region is essential for high, androgen- Gao N, Zhang J, Rao MA, Case TC, Mirosevich J, Wang Y et al. regulated activity of the prostate-specific antigen promoter. Mol (2003). The role of hepatocyte nuclear factor-3 alpha (Forkhead Box Endocrinol 11: 148–161. A1) and androgen receptor in transcriptional regulation of prostatic Cleutjens KB, van Eekelen CC, van der Korput HA, Brinkmann AO, genes. Mol Endocrinol 17: 1484–1507. Trapman J. (1996). Two androgen response regions cooperate in Gregory CW, He B, Johnson RT, Ford OH, Mohler JL, French FS steroid hormone regulated activity of the prostate-specific antigen et al. (2001). A mechanism for androgen receptor-mediated prostate promoter. J Biol Chem 271: 6379–6388. cancer recurrence after androgen deprivation therapy. Cancer Res Costello PS, Nicolas RH, Watanabe Y, Rosewell I, Treisman R. 61: 4315–4319. (2004). Ternary complex factor SAP-1 is required for Erk-mediated Han G, Buchanan G, Ittmann M, Harris JM, Yu X, Demayo FJ et al. thymocyte positive selection. Nat Immunol 5: 289–298. (2005). Mutation of the androgen receptor causes oncogenic

Oncogene Regulation of ELK4 by androgens H Makkonen et al 4876 transformation of the prostate. Proc Natl Acad Sci USA 102: Riegman PH, Vlietstra RJ, van der Korput HA, Romijn JC, Trapman 1151–1156. J. (1991). Identification and androgen-regulated expression of two Heinlein CA, Chang C. (2002). The roles of androgen receptors and major human glandular kallikrein-1 (hGK-1) mRNA species. Mol androgen-binding proteins in nongenomic androgen actions. Mol Cell Endocrinol 76: 181–190. Endocrinol 16: 2181–2187. Schuur ER, Henderson GA, Kmetec LA, Miller JD, Lamparski HG, Heinlein CA, Chang C. (2004). Androgen receptor in prostate cancer. Henderson DR. (1996). Prostate-specific antigen expression is Endocr Rev 25: 276–308. regulated by an upstream enhancer. J Biol Chem 271: 7043–7051. Huggins C, Hodges GV. (1941). Studies on prostate cancer. I. The Shaulian E, Karin M. (2002). AP-1 as a regulator of cell life and death. effect of castration, estrogen and androgen injections on serum Nat Cell Biol 4: 131–136. phosphatases in metastatic carcinoma of the prostate. Cancer Res 1: Shaw PE, Saxton J. (2003). Ternary complex factors: prime nuclear 293–297. targets for mitogen-activated protein kinases. Int J Biochem Cell Biol Jackson AL, Burchard J, Leake D, Reynolds A, Schelter J, Guo J et al. 35: 1210–1226. (2006). Position-specific chemical modification of siRNAs reduces Stewart AB, Lwaleed BA, Douglas DA, Birch BR. (2005). Current ‘off-target’ transcript silencing. RNA 12: 1197–1205. drug therapy for prostate cancer: an overview. Curr Med Chem Kallio PJ, Palvimo JJ, Mehto M, Ja¨ nne OA. (1994). Analysis of Anticancer Agents 5: 603–612. androgen receptor-DNA interactions with receptor proteins pro- Taplin ME, Bubley GJ, Shuster TD, Frantz ME, Spooner AE, Ogata duced in insect cells. J Biol Chem 269: 11514–11522. GK et al. (1995). Mutation of the androgen-receptor gene in Kang Z, Ja¨ nne OA, Palvimo JJ. (2004). Coregulator recruitment and metastatic androgen-independent prostate cancer. N Engl J Med histone modifications in transcriptional regulation by the androgen 332: 1393–1398. receptor. Mol Endocrinol 18: 2633–2648. Thompson J, Saatcioglu F, Ja¨ nne OA, Palvimo JJ. (2001). Disrupted Karvonen U, Ja¨ nne OA, Palvimo JJ. (2006). Androgen receptor amino- and carboxyl-terminal interactions of the androgen receptor regulates nuclear trafficking and nuclear domain residency of are linked to androgen insensitivity. Mol Endocrinol 15: 923–935. corepressor HDAC7 in a ligand-dependent fashion. Exp Cell Res Tomlins SA, Mehra R, Rhodes DR, Cao X, Wang L, Dhanasekaran 312: 3165–3183. SM et al. (2007). Integrative molecular concept modeling of prostate Kononen J, Bubendorf L, Kallioniemi A, Ba¨ rlund M, Schraml P, cancer progression. Nat Genet 39: 41–51. Leighton S et al. (1998). Tissue microarrays for high- Tomlins SA, Mehra R, Rhodes DR, Smith LR, Roulston D, Helgeson throughput molecular profiling of tumor specimens. Nat Med 4: BE et al. (2006). TMPRSS2:ETV4 gene fusions define a third 844–847. molecular subtype of prostate cancer. Cancer Res 66: 3396–3400. Linja MJ, Porkka KP, Kang Z, Savinainen KJ, Ja¨ nne OA, Tammela Tomlins SA, Rhodes DR, Perner S, Dhanasekaran SM, Mehra R, Sun TL et al. (2004). Expression of androgen receptor coregulators in XW et al. (2005). Recurrent fusion of TMPRSS2 and ETS prostate cancer. Clin Cancer Res 10: 1032–1040. transcription factor genes in prostate cancer. Science 310: 644–648. Linja MJ, Savinainen KJ, Sarama¨ ki OR, Tammela TL, Vessella RL, Wang Q, Li W, Liu XS, Carroll JS, Ja¨ nne OA, Keeton EK et al. Visakorpi T. (2001). Amplification and overexpression of androgen (2007). A hierarchical network of transcription factors governs receptor gene in hormone-refractory prostate cancer. Cancer Res 61: androgen receptor-dependent prostate cancer growth. Mol Cell 27: 3550–3555. 380–392. Luo J, Duggan DJ, Chen Y, Sauvageot J, Ewing CM, Bittner ML et al. Welsh JB, Sapinoso LM, Su AI, Kern SG, Wang-Rodriguez J, (2001). Human prostate cancer and benign prostatic hyperplasia: Frierson HF et al. (2001). Analysis of gene expression identifies molecular dissection by gene expression profiling. Cancer Res 61: candidate markers and pharmacological targets in prostate cancer. 4683–4688. Cancer Res 61: 5974–5978. Magee JA, Araki T, Patil S, Ehrig T, True L, Humphrey PA et al. Visakorpi T, Hyytinen E, Koivisto P, Tanner M, Keina¨ nen R, (2001). Expression profiling reveals hepsin overexpression in Palmberg C et al. (1995). In vivo amplification of the androgen prostate cancer. Cancer Res 61: 5692–5696. receptor gene and progression of human prostate cancer. Nat Genet Maroni PD, Crawford ED. (2006). Screening for prostate cancer in 9: 401–406. 2006: PSA in the 21st century. NC Med J 67: 136–139. Xiao D, Qu X, Weber HC. (2002). GRP receptor-mediated immediate Oettgen P, Finger E, Sun Z, Akbarali Y, Thamrongsak U, Boltax J early gene expression and transcription factor Elk-1 activation in et al. (2000). PDEF, a novel prostate epithelium-specific ets prostate cancer cells. Regul Pept 109: 141–148. transcription factor, interacts with the androgen receptor and Yu YP, Landsittel D, Jing L, Nelson J, Ren B, Liu L et al. (2004). activates prostate-specific antigen gene expression. J Biol Chem Gene expression alterations in prostate cancer predicting tumour 275: 1216–1225. aggression and preceding development of malignancy. J Clin Oncol Palvimo JJ, Kallio PJ, Ikonen T, Mehto M, Ja¨ nne OA. (1993). 22: 2790–2799. Dominant negative regulation of trans-activation by the rat Zheng H, Wasylyk C, Ayadi A, Abecassis J, Schalken JA, Rogatsch H androgen receptor: roles of the N-terminal domain and heterodimer et al. (2003). The transcription factor Net regulates the angiogenic formation. Mol Endocrinol 7: 1399–1407. switch. Genes Dev 17: 2283–2297.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Oncogene