Induction of Prostatic Intraepithelial Neoplasia and Modulation of Androgen Receptor by ETS Variant 1/ETS-Related Protein 81

Analysis of tumor samples

Laser capture microdissection was performed on 122 and macrodissection on 40 fresh-frozen tissues. ETV1 expression was determined by microarray analyses (14). Formalin-fixed, paraffin-embedded tumor samples from an independent study were analyzed by quantitative reverse transcription-PCR (RT-PCR; ref. 15). Informed consent was obtained from all patients, and these studies were approved by the Mayo Clinic Institutional Review Board.

Transgenic mice

Coding sequence for full-length human ETV1 or its amino acids 132 to 477 (ΔETV1), which were preceded by a Kozak sequence and followed by an SV40 polyadenylation site and intron, was cloned downstream of the composite ARR₂PB promoter (16). These gene cassettes were injected into one-cell fertilized FVB mice embryos, which were then reimplanted into pseudopregnant females. Resulting pups that tested positive for transgene insertion by PCR were crossed with C57BL/6 mice. Prostates from transgenic animals were dissected, paraffinized, cut, and stained with H&E (17). These experiments were approved by the Mayo Clinic Institutional Animal Care and Use Committee.

Luciferase assays

LNCaP cells were seeded at 40% confluency in six-well plates and then transfected by the calcium phosphate coprecipitation method (18). One microgram of prostate-specific antigen (PSA)-luc that encompasses 5.8 kb of PSA upstream sequences or 1.5 μg of PSA-enhancer-luc reporter plasmid, in which the indicated PSA enhancer sequences were cloned into the tk80 luciferase construct, were used for transfection. As indicated, the empty vector pEV3S or ETV1 expression vectors were cotransfected, and pBluescript KS⁺ was used as a carrier to bring the total DNA amount up to 5 μg. The precipitate was washed away with PBS after 12 h. Cells were then incubated in medium containing 10% charcoal/dextran-treated fetal bovine serum (FBS) with or without mibolerone for 36 h, and luciferase activities were determined as described (19).

Similarly, RK13 cells were seeded at 25% confluency in 12-well plates and transfected with 0.5 μg of E74₃-tk80-luc (18) or the AR binding site–driven luciferase reporter ARE₄-luc containing four copies of the AR binding site AGCACTTGCTGTTCT (20). Where indicated, 100 ng of ETV1 expression plasmid or empty vector pEV3S and 500 ng (in case of E74₃-tk80-luc) or 20 ng (in case of ARE₄-luc) of pSG-AR expression vector or empty vector pSG6 were cotransfected.

Knockdown experiments

To down-regulate human ETV1, respective short hairpin RNAs (shRNA) were cloned into pSIREN-RetroQ. Within 2 d, LNCaP cells were thrice infected with retrovirus derived from these constructs and then incubated for 47 h with or without 0.5 nmol/L mibolerone in medium containing 10% charcoal/dextran-treated FBS (21). Protein and RNA analyses were then performed as described (22). The following sequences were targeted by shRNA: #1, GTGCCTGTACAATGT-CAGT; #4, TGTCAGTGCCTATGATCAG; and #5, TTCGATGGAGACAT-CAAAC.

Chromatin immunoprecipitation assays

These assays were performed as described (23). The following primers were used to amplify a 367-bp PSA enhancer fragment (5′-GGGGTTTGTGCCACTGGTGAG-3′ and 5′-GGGAGGCAATTCTCCATGGTT-3′), a 261-bp PSA promoter fragment (5′-TCTAGTTTCTGGTCTCAGAG-3′ and 5′-TTGCTGTTCTGCAATTACTAG-3′), and a 268-bp PSA 3′-untranslated region (3′-UTR) fragment (5′-TACTGGC-CATGCCTGGAGAC-3′ and 5′-TGGCTCACAGCCTTCTCTAG-3′). The PCR program used was as follows: 98°C for 2 min; 6 cycles of 98°C for 30 s, 64°C for 30 s (−1°C per cycle), 72°C for 25 s, followed by 32 cycles of 98°C for 30 s, 58°C for 30 s, 72°C for 25 s (+1 s per cycle); and finally by 72°C for 4 min.

Electrophoretic mobility shift assays

Reactions were set up in 10 μL of 20 mmol/L HEPES (pH 7.4), 25 mmol/L NaCl, 2 mmol/L DTT, 0.01% Tween 20, 12% glycerol, 0.1 μg/μL bovine serum albumin, and 0.05 μg/μL poly(deoxyinosinic-deoxycytidylic acid)·(deoxyinosinic-deoxycytidylic acid). Approximately 0.25 ng of ³²P-labeled oligonucleotides, bacterially expressed and purified ETV1 encompassing amino acids 249 to 477, and 0.5 μL of anti-ETV1 (C-20; Santa Cruz Biotechnology) antibody were used (24).

The following oligonucleotide pairs were hybridized to generate double-stranded PSA enhancer oligonucleotides and labeled with [³²P]dATP and Klenow DNA polymerase (25): ETS1, 5′-CAGTCTAGGTGGATGCTGTGCA-CACGGG-3′ and 5′-AACCCCGTGTGCACAGCATCCACCTAG-3′; ETS2/3, 5′-CTGAGATTAGGAATCCTCAATCTT-3′ and 5′-ATAAGATTGAGGATTCC-TAATCTC-3′; ETS4, 5′-GAAGATATTATCTTCATGATCTTGGATTGAAAAC-3′ and 5′-CTGTTTTCAATCCAAGATCATGAAGATAATATCTTC-3′; ETS5, 5′-CTACTCTGGAGGAACATATTGTATCGATTGTCCTTG-3′ and 5′-GTCAAG-GACAATCGATACAATATGTTCCTCCAGAGT-3′; ETS6, 5′-GACATTATCTT-TATTATCTAGGACAGTAAGCAAGCCTGGATCTGAG-3′ and 5′-CTCTCAGATCCAGGCTTGCTTACTGTCCTAGATAATAAAGATAATGTC-3′; ETS7, 5′-ATCATCTTGCAAGGATGCCTGCTTT-3′ and 5′-GTAAAGCAGG-CATCCTTGCAAGATG-3′; ETS8/9, 5′-CTTTACAAACATCCTTGAAACAA-CAATCCAGAA-3′ and 5′-TTCTGGATTGTTGTTTCAAGGATGTTTGTAA-3′; and ETS10, 5′-GTTCAGCCAGAGCCTTCCACCCTTGT-3′ and 5′-ACAAGGGTGGAAGGCTCTGGCTGA-3′.

Coimmunoprecipitations

Human embryonic kidney 293T cells grown on poly-L-lysine–coated dishes were transiently transfected by the calcium phosphate coprecipitation method (26). Cells were grown with or without 1 nmol/L mibolerone and coimmunoprecipitations were performed (22). Immunoprecipitated proteins were revealed as described (27).

Glutathione S-transferase pull-down assays

Glutathione S-transferase (GST) fusion proteins were produced in bacteria and affinity purified on glutathione agarose (28). Baculovirus encoding for AR endowed at the NH₂ terminus with both a Flag-tag and 6His-tag was used to infect Sf9 insect cells, and AR was subsequently affinity purified on Ni²⁺-NTA agarose. Binding of AR to GST fusion proteins was assessed as described (28).

ETV1 marks aggressive prostate cancer and induces PIN

Overexpression of ETV1 was reported in 6% to 16% of prostate tumors (4). To study its correlation with disease stage, we determined ETV1 expression in normal prostates, prostatic intraepithelial neoplasia (PIN), and carcinomas. Remarkably, ETV1 was found overexpressed only in Gleason pattern 4 and 5 tumors with a frequency of 3 of 46 (6.5%) and 2 of 22 (9.1%), respectively (Fig. 1A), suggesting that ETV1 expression is correlated with high-grade prostate cancer.

In addition, we studied ETV1 expression in a cohort of patients with high risk of systemic cancer progression (Gleason score 7–10) that underwent radical retropubic prostatectomy. In this cohort, slight ETV1 overexpression was observed in 3 of 77 (3.9%) of the cases remaining cancer-free for at least 8 years after prostatectomy, whereas 10 of 80 (14.7%) of the relapsing patients showed, in part dramatic, ETV1 overexpression (Fig. 1B, left). Although not quite statistically significant (P = 0.0797, Fisher’s exact test), this further supports the notion that ETV1 overexpression is indicative of aggressive prostate cancer. In contrast, an equal percentage of cancer-free (34 of 81, 42.0%) and relapsing patients (36 of 84, 42.9%) showed ERG overexpression (Fig. 1B, right).

To assess whether overexpression of full-length ETV1 leads to neoplasia, we generated transgenic mice in which human ETV1 cDNA was expressed under the control of a composite rat probasin promoter (ARR₂PB) that is androgen inducible and prostate specific (16). We analyzed prostates at 12 months of age in four mice that were derived from four independent transgenic ETV1 lines. Based on nuclear atypia, architectural changes such as cribriforming, and focality of the lesions, PIN was identified in all four transgenic ETV1 mice analyzed (Fig. 1C; Supplementary Table S1); please note that age-matched wild-type mice do not display PIN or hyperplasia (Fig. 1D). Further, RT-PCR of whole prostates indicated that the level of transgene expression (human ETV1) was similar to the level of endogenous murine ETV1/ER81 expression (Supplementary Table S2). Because ETV1 expression is elevated from 2-fold to >10-fold in human prostate tumors (4, 29), the level of ETV1 transgene expression is not unreasonably high. In conclusion, these results indicate that overexpression of full-length ETV1 is causally involved in prostate carcinogenesis.

Overexpression of ETV1 and the ΔETV1 truncation, which lacks the first 131 amino acids of full-length ETV1, occurs with roughly equal frequency in prostate tumors (3–5). Therefore, we also generated transgenic mice expressing ΔETV1 and analyzed four mice derived from three independent ΔETV1 transgenic lines. We observed the development of PIN in one and hyperplasia in the other three ΔETV1 mice (Supplementary Table S1); moreover, ΔETV1 and ETV1 transgene expression were not drastically different as determined by RT-PCR (Supplementary Table S2). These data indicate that both truncated and full-length ETV1 are capable of initiating prostate cancer formation. While this study was in progress, another report confirmed that overexpression of ΔETV1 induces PIN (4).

The only other organ apart from prostate displaying activity, albeit at a lower level, of the ARR₂PB promoter driving expression of the ETV1 and ΔETV1 transgenes is the seminal vesicle (16). Notably, we observed macroscopic differences in the seminal vesicles in one of four ETV1 and two of three ΔETV1 transgenic lines (Supplementary Table S1; Supplementary Fig. S1), including different color, enlargement, or anatomically different location in the abdomen. Moreover, pathologic examination of these abnormal seminal vesicles revealed hyperplasia or neoplasia, suggesting that ETV1/ΔETV1 overexpression may cause neoplastic transformation in the seminal vesicles.

Stimulation of PSA expression by ETV1

AR and ETS binding sites reside frequently close together in the genome (30), suggesting that the ETS protein ETV1 may cooperate with AR in regulating gene transcription. We noted the presence of 10 putative ETS sites in the enhancer of the PSA, a serum marker for prostate cancer whose expression is regulated by multiple AR sites (31, 32). Thus, we wondered if overexpression of ETV1 might affect PSA transcription. We overexpressed ETV1 in AR-positive LNCaP prostate cancer cells and assessed the activity of a luciferase reporter construct driven by 5.8 kb of PSA upstream sequences, including its enhancer and proximal promoter. As shown in Fig. 2A, a synthetic AR agonist, mibolerone, induced PSA luciferase activity by 14.7-fold. Further, overexpression of ETV1 stimulated PSA luciferase activity by 2.3-fold in hormone-depleted medium. Importantly, ETV1 and hormone synergized to affect a 41.3-fold stimulation of luciferase activity. These data suggest that ETV1 cooperates with AR in the regulation of PSA transcription.

ETV1 is activated through phosphorylation by mitogen-activated protein (MAP) and MAPKAP kinases on six serine/threonine residues (8, 33, 34), and LNCaP cells express high levels of the MAP kinase stimulator HER2/Neu (35). Accordingly, mutation of all six ETV1 phosphorylation sites to alanine suppressed the ability of ETV1 to induce PSA luciferase activity on hormone treatment in LNCaP cells (6×A mutant in Fig. 2A). Western blotting revealed that this effect was not due to different protein levels of the 6×A mutant compared with wild-type ETV1 (Fig. 2B). Thus, activation of ETV1 by MAP kinases is required for maximal PSA transcription. Interestingly, in the absence of hormone treatment, the 6×A mutant was slightly more active than wild-type ETV1, suggesting that ETV1 phosphorylation may suppress basal PSA expression.

To show that endogenous ETV1 affects PSA transcription, we knocked down ETV1 with three different shRNAs in LNCaP cells. All of these shRNAs specifically and efficiently down-regulated ETV1 expression both in the presence or in the absence of mibolerone (Fig. 2C); please note that ETV1 expression is up-regulated by mibolerone because the entire ETV1 gene locus is translocated into an androgen-responsive region in LNCaP cells (4). Importantly, both PSA mRNA and protein levels were reduced on ETV1 down-regulation in the presence and absence of mibolerone (Fig. 2C). Further, down-regulation of ETV1 did not significantly affect AR levels (Supplementary Fig. S2). In conclusion, ETV1 enhances endogenous PSA expression.

To assess if the PSA gene is targeted by ETV1, we performed chromatin immunoprecipitation assays in LNCaP cells. As expected, AR binding to the PSA enhancer dramatically increased after 1 or 8 hours of hormone stimulation (Fig. 2D). Consistent with a previous report (36), AR binding to the proximal PSA promoter was negligible, and none was observed at the 3′-UTR of the PSA gene. Neither an irrelevant antibody (anti-Rcl) nor omission of antibody led to the precipitation of PSA enhancer, promoter, or 3′-UTR DNA. Like AR, some ETV1 interacted with the PSA enhancer before hormone stimulation (Fig. 2D). On hormone stimulation, strongly increased binding of ETV1 to the PSA enhancer was evident, whereas negligible or no binding of ETV1 to the PSA proximal promoter and 3′-UTR was observed. Collectively, these data indicate that PSA is a target gene of ETV1 and that ETV1 and AR coordinately bind to the PSA enhancer on androgen stimulation.

Binding of ETV1 to the PSA enhancer

To assess whether the PSA enhancer is regulated directly by ETV1, we cloned it in front of a thymidine kinase basal promoter and the luciferase gene and transfected this reporter construct with and without ETV1 into LNCaP cells. Indeed, ETV1 enhanced luciferase activity from 0.02 to 0.12 in the absence of hormone and from 21.7 to 43.14 on hormone stimulation (Fig. 3A, −4324/−3875 construct). We then divided the enhancer into two parts: one containing the pivotal AR binding site III (−4324/−4119 construct) and the other encompassing ETS sites 6 to 10 (−4127/−3875 construct). Both constructs displayed a dramatically decreased response to hormone stimulation. Importantly, the effect of ETV1 overexpression was curtailed (Fig. 3A). These data suggest that ETS sites 6 to 10 cooperate with the AR binding site III in regulating the PSA enhancer.

To elucidate which ETS sites in the PSA enhancer might be bound by ETV1, we performed electrophoretic mobility shift assays. Oligonucleotides spanning 1 or 2 of the 10 ETS sites within the PSA enhancer were generated and incubated with recombinant ETV1 in vitro. We observed weak binding of ETV1 to ETS sites 8/9 (Fig. 3B, white arrow) and the known ETV1 site within the matrix metalloproteinase-1 (MMP-1) promoter. However, inclusion of an anti-ETV1 antibody, which stabilizes the binding of ETV1 to selected ETS sites such as in the MMP-1 or Smad7 promoters (8, 37), resulted in much more pronounced DNA binding to both ETS sites 8/9 and the MMP-1 target site (Fig. 3B, black arrow). Moreover, DNA binding was now also found at other ETS sites in the following order of decreasing strength: ETS8/9 > ETS7 > ETS5 > ETS1,6 > ETS10. Furthermore, mutation of ETS site 8 but not site 9 abrogated binding of ETV1 to the ETS8/9 oligonucleotide (Supplementary Fig. S3A), indicating that ETS site 8 but not ETS site 9 is targeted by ETV1; this result is also consistent with the fact that ETS site 8 matches the ETV1 consensus DNA binding site much better than ETS site 9 (Supplementary Fig. S3B). In conclusion, ETV1 can bind to several sites within the PSA enhancer, with ETS site 8 being the strongest binder.

To determine the functional importance of ETS site 8, we mutated it in the PSA enhancer. Both in the presence and in the absence of hormone, this led to a reduction of enhancer activity (Fig. 3C). Whereas ETV1 was still able to enhance the activity of the mutated PSA enhancer in the absence of hormone, albeit the degree of activation was less than with the wild-type enhancer, ETV1 over-expression no longer enhanced PSA transcription in the presence of hormone (Fig. 3C). Thus, ETS site 8 seems to be especially crucial for the androgen-inducible expression of the PSA gene.

ETV1 forms complexes with AR

The apparent in vivo corecruitment of ETV1 and AR to the PSA enhancer after hormone stimulation suggested that these factors may be part of one complex. To explore this possibility, we expressed Flag-tagged AR with or without Myc-tagged ETV1 and performed anti-Myc immunoprecipitations. Probing the precipitate with anti-Flag antibodies revealed that AR coprecipitated with ETV1 (Fig. 4A). Moreover, this interaction of ETV1 and AR was independent of the presence or absence of hormone. To confirm this interaction, we also performed coimmunoprecipitation experiments with endogenous AR and ETV1 in LNCaP cells. Anti-ETV1 antibodies, but not control antibodies, led to the coprecipitation of AR (Fig. 4B), indicating that endogenous AR and ETV1 form complexes in vivo.

To assess whether the interaction between ETV1 and AR is direct, we produced ETV1 as a GST fusion protein in bacteria, affinity purified it, and bound it to glutathione agarose beads. Incubation of these beads with affinity-purified, baculovirus-expressed AR revealed that AR bound to GST-ETV1, but not to GST or ETV1 amino acids 182 to 334, in a hormone-independent manner (Fig. 4C, left). Thus, ETV1 and AR are capable of binding directly to each other. Similarly, ERG and AR interacted with each other in vitro and in vivo (Fig. 4C (right) and D).

Next, we sought to determine the interaction domain(s) within AR. Thus, we expressed full-length, Flag-tagged AR or various truncations thereof and probed for coprecipitation of Myc-tagged ETV1 after anti-Flag immunoprecipitation. Only amino acids 650 to 919, which encompass the ligand binding domain of AR, coprecipitated ETV1 with a similar efficiency as full-length AR_1–919 (Fig. 5A). Conversely, we probed which domain(s) in ETV1 is responsible for the interaction with AR (Fig. 5B). The NH₂-terminal activation domain of ETV1 encompassing amino acids 1 to 182 did not bind to AR, but ETV1 lacking these amino acids (ETV1_182–477) did, indicating that the NH₂-terminal activation domain of ETV1 is not involved in the interaction with AR. Moreover, ETV1_249–477 was unable to bind to AR. Thus, amino acids 182 to 249 are crucial for the interaction with AR, yet not sufficient as ETV1_1–249 did not bind to AR. Further, ETV1_1–334 was unable to bind to AR (Fig. 5B), indicating that amino acids 334 to 477 are also required for the interaction with AR. However, these amino acids alone were not sufficient because ETV1_333–477 did not bind to AR. In conclusion, ETV1 amino acids in the center domain cooperate with the COOH-terminal 144 amino acids in AR binding.

Reciprocal coactivation by ETV1 and AR

We next assessed if ETV1 and AR may recruit each other as cofactors, meaning that only one of these transcription factors binds to DNA, whereas the other is recruited by protein-protein interactions. To this end, we used the E74 site that is derived from the Drosophila E74 gene and is tightly bound by ETV1 (18) but does not contain an AR binding site. Moreover, we used rabbit kidney RK13 cells that do not contain AR but have been often used to study ETV1/ER81 (8, 18, 34). AR expression or mibolerone stimulation, either alone or in combination, did not activate E74-dependent transcription in the absence of ectopic ETV1 (Fig. 6A). Similarly, neither AR expression nor mibolerone stimulation alone raised luciferase activities in the presence of ectopic ETV1. However, AR activated by mibolerone stimulated E74-dependent transcription by 2.1-fold in the presence of ectopic ETV1. This suggests that ETV1 recruits AR to the E74 site.

Conversely, we tested whether AR could use ETV1 as a coactivator. To this end, we used an AR binding site–dependent luciferase reporter that was stimulated 116-fold by AR and mibolerone in the absence of ectopic ETV1 (Fig. 6B). In the presence of ectopic ETV1, which did not affect AR protein levels (Fig. 6B, inset), AR and mibolerone stimulated luciferase activity by 330-fold, implicating that ETV1 is recruited by AR to the AR binding site.

Finally, we also tested whether ERG may cooperate with AR. Surprisingly, ERG expression led to a more than 2-fold decrease of the ability of AR to induce the AR binding site–dependent luciferase reporter in the presence of mibolerone (Fig. 6C). Likewise, ERG inhibited the induction of the PSA promoter by hormone in LNCaP cells (Supplementary Fig. S4). Together, these data suggest that ERG and ETV1 exert opposing effects on their common binding partner, AR.