Androgenic Induction of Prostate-specific Antigen Gene Is Repressed by Protein-Protein Interaction between the Androgen Receptor and AP-1/c-Jun in the Human Prostate Cancer Cell Line LNCaP*

In exploring the possible mechanisms of androgen independence of prostate-specific antigen (PSA) gene expression, we investigated the effect of elevating AP-1 by both 12-O-tetradecanoylphorbol 13-acetate (TPA) treatment and transfection of the c-Jun expression vector in LNCaP cells. Transcription of PSA is initiated when ligand-activated androgen receptor (AR) binds to a region in the PSA promoter that contains an androgen-responsive element (ARE). It was found that TPA inhibited androgen-induced PSA gene expression by a mechanism that did not alter nuclear levels of AR protein. Overexpression of AP-1 (junand fos proteins) also inhibited androgen-induced PSA promoter activity. These observations were apparently related to the disruption of AR·ARE complexes as demonstrated by the results of electrophoretic mobility shift assays. Specifically, c-Jun inhibited the formation of AR·ARE complexes and conversely that AR-glutathioneS-transferase proteins inhibited the formation of c-Jun·TPA-responsive element (TRE) complexes. Consistent with the inhibitory effect of both proteins, anti-c-Jun antibody blocked the inhibition of AR·ARE complex formation by c-Jun. A similar, but less marked, effect was obtained when anti-AR antibody was used to prevent AR inhibition of c-Jun·TRE complex formation. These findings together with results obtained from co-immunoprecipitation experiments strongly suggest that mutual repression of DNA binding activity is due to direct interaction between the two proteins and that the degree of repression may be determined by the ratio of AR to c-Jun. The mechanism of repression studied in mutant analysis experiments yielded evidence of an interaction between the DNA- and ligand-binding domains of AR and the leucine zipper region of c-Jun. Thus, the AR is similar to other nuclear receptors in its ability to interact with AP-1. This association provides a link between AP-1 and AR signal transduction pathways and may play a role in the regulation of the androgen-responsive PSA gene.

Prostate-specific antigen (PSA) 1 belongs to the family of kallikrien-like serine proteases (for a review, see Ref. 1). In males, expression of PSA occurs exclusively in the prostate with serum levels of PSA glycoprotein being an important marker in the diagnosis and progression of prostate cancer (2,3). PSA is an androgen-induced gene that contains androgen response elements (AREs) to which the androgen receptor (AR) binds (4,5).
The AR belongs to the superfamily of nuclear receptors that mediate the responses of lipophilic ligands, including steroids, retinoids, vitamin D 3 , and thyroid hormones (6). These receptors contain a highly conserved DNA-binding domain comprised of zinc finger-like motifs responsible for sequence-specific DNA binding, as well as protein-protein interactions. There is evidence to suggest that the DNA-binding domain of a nuclear receptor may interact with the leucine zipper of AP-1 to result in mutual transrepression (7)(8)(9). However, this interaction between nuclear receptors and AP-1 may be cell-specific, gene-specific, and may involve various mechanisms, including protein-protein interaction and/or adjacent or overlapping binding sites (10).
AP-1 is a transcriptional factor whose components are nuclear proteins encoded by c-fos and c-jun proto-oncogenes induced by 12-O-tetradecanoylphorbol 13-acetate (TPA). AP-1 has been implicated in cell growth, differentiation, and development with its activity modulated by growth factors, cytokines, oncogenes, and tumor promoters activating protein kinase C (PKC) (10). AP-1 induces transcriptional activation through interaction with the TPA-responsive element (TRE or AP-1 DNA-binding site) (11,12). TREs are recognized by Jun homodimers and Jun/Fos heterodimers that are formed through the leucine zipper domain of both proteins (11). The basic region adjacent to the leucine zipper on Jun and Fos proteins mediates AP-1 DNA binding activity (13)(14)(15).
Since the zinc finger motifs of the AR share a high degree of homology with the same regions of other nuclear receptors and in the face of mounting evidence of AP-1 interaction with such motifs, the question arose whether AR function could also be affected by interaction with AP-1. Accordingly, in this study we examined the direct interaction of AP-1 with AR protein and the effects of elevated AP-1 on androgen-stimulated PSA gene expression.

MATERIALS AND METHODS
Cell Culture and Northern Blot Analysis-All chemicals were purchased from Sigma, unless stated otherwise. PC3 and CV-1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 5% FBS (Life Technologies, Inc., Burlington, Ontario, Canada). LNCaP cells were maintained in RPMI 1640 supplemented with 5% FBS. Cells between the 37th and 49th generation were used in the these experiments. For Northern blot analyses, LNCaP cells were down-shifted to RPMI 1640 containing 2% FBS plus 2% TCM™ (serum replacement obtained from Celox Corp., Hopkins, MN), for 10 -14 days and 5 ϫ 10 5 or 3 ϫ 10 5 cells were initially plated on 6-cm dishes or 6-well plates, respectively. Culture medium was changed to RPMI 1640 with 2% TCM™ (i.e. serum-free), with or without 10 nM R1881, when wells were 60 -70% confluent with the cells. Total RNA was extracted with acid guanidinium thiocyanate/phenol/chloroform (16) and fractionated by electrophoresis prior to blotting onto Hybond-N ϩ filters (Amersham, Oakville, Ontario, Canada). The 1.4-kb EcoRI fragments of the PSA cDNA (17), 0.7-kb EcoRI/HindIII fragments of human AR cDNA (18) and 1.9-kb PstI fragments of chicken ␤-actin (19) were labeled with [␣-32 P]dCTP by Random Primers DNA labeling kit (Life Technologies, Inc.). The 40-base oligonucleotides for either c-Jun or c-Fos (Ciderlane, Toronto, Ontario, Canada) were end-labeled by T4 polynucleotide kinase with [␥-32 P]ATP. Hybridization was performed as reported previously (20). Filters with PSA and ␤-actin were washed in 0.1 ϫ SSC, 0.1% SDS for 30 min at 65°C, while filters with AR, c-Jun and c-Fos were in 0.5 ϫ SSC, 0.2% SDS for 30 min at 55°C. Densitometric analyses of mRNA bands were performed using NIH image (National Institutes of Health) from scanned x-ray films.
Immunoblots-LNCaP cells were incubated in RPMI 1640 containing 2% TCM™ for 24 h prior to the addition of vehicle (0.1% ethanol), 10 nM R1881, or 1 nM TPA. After incubation with compounds, cytosolic and nuclear extracts were prepared as described by Antras et al. (21). Western blots were performed with approximately 40 g of protein in each lane. Immunoblots were blocked overnight in 5% dry milk (w/v) in 20 mM Tris-HCl, pH 7.4, containing 500 nM NaCl (TBS). Blots were incubated for 4 h with antibodies to AR (PG-21) (22), c-Jun (sc-045, Santa Cruz Biotechnology, Inc, Santa Cruz, CA), or c-Fos (sc-052, Santa Cruz Biotechnology, Inc.). The blots were then washed and incubated for 1 h with the secondary antibody (1:10,000). Antibodies were diluted in 5% milk/TBS solution. AR, c-Jun, and c-Fos proteins were detected using the ECL luminescence kit (Amersham Corp.). Densitometric analyses of protein bands were performed using NIH image from scanned x-ray films.
Co-immunoprecipitation-LNCaP and PC3 cells were incubated in RPMI 1640 and Dulbecco's modified Eagle's medium, respectively, containing 5% dextran-coated stripped serum, 10 nM R1881 for 24 h, and 10 nM TPA for 6 h. The cells were harvested and nuclear extracts were prepared as described by Antras et al. (21). The nuclei were solubilized in mild Nonidet P-40 buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.1% Nonidet P-40, 5 g/ml leupeptin 5 g/ml aprotinin, 5 g/ml trypsin inhibitor, 5 g/ml bacitracin, and 1 mM phenylmethylsulfonyl fluoride) for 15 min at 4°C. The nuclear extracts were precleared with protein A-Sepharose for 30 min and incubated with anti-AR antibody PG-21 (22) or 15071A (Pharmingen, San Diego, CA) for 1 h at 4°C. The antigen antibody complexes were collected by the addition of protein A-Sepharose. Immune complexes were washed once with mild Nonidet P-40 buffer prior to separating on a 10% SDS-PAGE. Western blots analyses were carried out with anti c-Fos antibody and anti c-Jun antibody as described above.
PSA Promoter Plasmid Constructs and Luciferase Assay-PSA 5Јflanking DNA was obtained by PCR-mediated amplification of human genomic DNA using oligonucleotide primers corresponding to the PSA gene. The sequences for primers were 5Ј-CATTGTTTGCTGCACGTTG-GAT-3Ј and 5Ј-TCCGGGTGCAGGTGGTAAGCTTGG-3Ј. The PCR fragments were purified by gel electrophoresis, blunt-ended, and ligated with EcoRV-digested pBluescript (pBS) SK(Ϫ) (Stratagene, La Jolla, CA). pBS containing the PSA 5Ј-flanking DNA (designated as pBS-PSA-630) was amplified and purified from transformed Escherichia coli DH-5␣ and sequenced by dideoxynucleotide chain termination method using double-stranded DNA cycle sequencing system kit (Life Technologies, Inc.). DNA fragments corresponding to Ϫ630/ϩ12 of the PSA 5Ј-flanking region were excised from pBS-PSA-630 with HindIII and inserted into the HindIII site of promoterless plasmid, pGL-2 basic (Promega, Madison, WI), which contains firefly luciferase as a reporter gene. This pGL-2 basic containing the Ϫ630/ϩ12 fragment of the PSA 5Ј-flanking DNA was referred to as pPSA-630.
LNCaP cells (2-2.5 ϫ 10 5 ) were plated on 6-well plates and incubated in RPMI 1640 with 5% FBS for 3 days, resulting in 50 -60% confluence. Plasmid DNA was mixed with 5 l of Lipofectin agent (Life Technologies, Inc.) and incubated for 15-20 min at room temperature. The total amount of plasmid DNA used was normalized to 3 g/well by the addition of empty plasmid. Medium was replaced after 24 h by RPMI 1640 with 2% TCM™ with or without R1881 (i.e. serum-free media). Cells were collected after 48-h incubation using cell lysis buffer (100 mM potassium phosphate (pH 7.8), 0.2% Triton X-100, and 1 mM dithiothreitol). Luciferase activities were measured using a commercial kit from Promega according to the manufacturer's protocol, and activities were normalized by either protein concentration determined by the method of Bradford (23) or ␤-galactosidase activities measured with Galacto-Light (Tropix Inc.). Luciferase activities are expressed as relative luminescent units/mg of protein/min. All transfection experiments were carried out in triplicate wells and repeated two to eight times using at least two sets of plasmids prepared separately.
Other Plasmids-Full-length rat AR cDNA (amino acid 1-902) was cloned into pRc-CMV whose transcription is driven by cytomegalovirus promoter and this plasmid was referred to as pAR6. In our previous report, transcriptional efficacy of pAR6 was determined by expression of mRNA and binding assay in AR-negative PC3 cells when stably transfected with pAR6 (24). Several mutant rat AR expression plasmids were constructed in our laboratory. Regions of amino acids encoded by these mutant AR expression plasmids are as follows: pAR4 (232-649), pAR5 (390 -649), and pAR7 (232-902). pRSV-c-Jun and pRSV-c-Fos are wild type c-Jun and c-Fos expression plasmids, respectively. pRSV-c-Jun-⌬1 is lacking a small portion of the N-terminal activator domain, -c-Jun-⌬3 lacks the N-terminal domain, and -c-Jun-⌬LZ lacks the leucine zipper. pRSV-Jun B, -Jun D, -Fra 1, and -Fra 2 are Jun B, Jun D, Fra 1, and Fra 2 expression plasmids, respectively. pRSV-0 is the empty plasmid for pRSV series of plasmids. ARR 3 -tk-luciferase reporter construct consists of three congruent rat probasin AREs (Ϫ244 to Ϫ96) ligated in tandem into the HindIII site of the pT81 luciferase vector (ATCC, Rockville, MD) as described by us previously (25). pCH110 (Amersham), a ␤-galactosidase expression plasmid, was co-transfected as an internal marker for normalizing efficacy of transient transfection.
Preparation of Recombinant c-Jun Protein and AR-GST Fusion Proteins-The prokaryotic expression vector pET-8c/c-Jun coding for the full-length c-Jun was transformed into E. coli BL21(DE3)pLysS. The recombinant protein was induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside and extracted from inclusion bodies according to the protocol of Lin and Cheng (26) with slight modifications. Briefly, the inclusion bodies were isolated and subjected to several rounds of sonication in the appropriate buffers. The insoluble pellet was solubilized in 5 M guanidinium HCl, followed by stepwise dialysis against buffers containing 2 M, 1 M, and 0.5 M guanidinium HCl. Yields were typically 14 mg of c-Jun protein per 200 ml of bacterial culture with 90% purity. AR1 and AR2 were expressed in E. coli as isopropyl-␤-D-thiogalactopyranoside-induced fusion protein with glutathione S-transferase (GST), purified through glutathione affinity chromatography, and calculated to have greater than 90% purity when assayed by Coomassie Blue staining of polyacrylamide gels (25). AR1 encodes amino acids 524 -902 (rat AR) encompassing the DNA-binding domain, hinge region, and ligand-binding domain. AR2 encodes amino acids 524 -649 (rat AR) encompassing the DNA-binding domain and hinge region.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts from LNCaP cells or purified proteins were used for EMSA studies. Nuclear extracts were prepared from cells treated with vehicle (0.1% ethanol), 10 nM R1881, or TPA (1 and 10 nM) for 6 h before harvesting. DNA binding reactions were carried out in a total volume of 25 l, containing DNA binding buffer (20 mM HEPES, pH 7.9, 20% (v/v) glycerol, 100 mM KCl, 0.2 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 500 ng of poly(dI-dC) (Pharmacia Biotech Inc.)). Approximately 1.5 fmol of double-stranded 32 P-labeled TRE oligonucleotide (5Ј-CGCTTGATGAGTCAGCCGGAA-3Ј), PSA ARE oligonucleotide (5Ј-TTGCAGAACAGCAAGTGCTAGCTC-3Ј), PSA mutant ARE (5Ј-TTG-CAAAAAAGCAAGTGCTAGCTC-3Ј) were used. SP-1 oligonucleotide (5Ј-ATTCGATCGGGGCGGGGCGAG-3Ј) and the TRE oligonucleotide were obtained from Promega. All lanes were normalized for additions of purified proteins by correction for the amount of buffer and total protein. For competition experiments 100-fold excess unlabeled oligonucleotide was used. DNA-protein complexes were separated under nondenaturing conditions in a 8% polyacrylamide gel (29:1) containing 2.5% glycerol and run in 0.5 ϫ TBE (1 ϫ ϭ 89 mM Tris borate, 89 mM boric acid, and 2 mM EDTA, pH 8.3) at 200 V. Bands from dried EMSA gels were quantified by the STORM 860 PhosphorImager (Molecular Dynamics). Experiments employing antibodies were first preincubated for 30 min at room temperature with the antigen and antibody before the probe was added.
Statistics-The Student's t test was used for statistical analysis. The significance levels were set at: ***, p Ͻ 0.001; **, p Ͻ 0.01; and *, p Ͻ 0.05, unless stated otherwise in the figure legend.

TPA Decreases Levels of PSA mRNA and Increases c-Jun and
c-Fos mRNA-TPA alters the transcriptional activity of numerous nuclear receptors. Therefore, to examine the biological effects of TPA upon androgen regulation of PSA gene expression, we exposed LNCaP cells to TPA and performed Northern blot analyses. Cells exposed to the synthetic androgen, R1881 (10 nM), showed a 5-fold increase in accumulated PSA mRNA as compared with control levels (Fig. 1A). Androgen-stimulated cells exposed to TPA for 9 h showed a dose-dependent decrease in PSA mRNA levels. However, TPA concentrations of 10 and 100 nM induced "apoptosis-like" cell death of LNCaP cells consistent with the observations of Day et al. (27) and Young et al. (28). TPA concentrations of 1 nM and less did not alter morphology nor reduce cell viability (data not shown). Therefore, all subsequent experiments employed a TPA concentration of 1 nM. ␤-Actin mRNA levels did not appear to be altered in cells undergoing apoptosis, as compared with untreated cell levels, after 9 h of treatment with 10 and 100 nM TPA, which is consistent to observations with HeLa cells undergoing programmed cell death (29).
TPA is thought to increase AP-1 transactivation through the PKC signal transduction pathway (30). Thus, inhibition of PKC activity by staurosporine, a PKC inhibitor, should block TPA induction of AP-1. In Fig. 1B, a 50 nM concentration of staurosporine completely blocked the TPA-associated decrease of androgen-induced PSA mRNA levels. Staurosporine (50 nM) alone had no effect on androgen-induced PSA mRNA levels.
A time-dependent decrease of androgen-induced levels of PSA mRNA was observed when cells were exposed to 1 nM TPA (Fig. 1C). These decreases were apparent after 6 h of TPA treatment and levels continued to decline for the duration of the experiment at 24 h. ␤-Actin mRNA levels remained consistent, and androgen-induced levels of PSA mRNA remained elevated in the absence of TPA for the duration of the study.
TPA has been shown to induce c-jun and c-fos mRNA levels in LNCaP cells (27). In agreement, TPA (1 nM) caused a transient increase in both c-jun and c-fos mRNA levels ( Fig. 1D), regardless of the presence of androgen (R1881). This suggests that androgen stimulation and the subsequent activation of AR does not interfere with the signal transduction pathway leading from PKC to c-jun and c-fos induction in LNCaP cells.
A time course study of TPA induction of c-Jun and c-Fos proteins in LNCaP cells showed that maximum levels were achieved after 4.5 h of exposure ( Fig. 2). At this time point, c-Jun levels were 6.6-fold higher and c-Fos levels 49-fold higher than levels in untreated cells. c-Jun levels remained 3-4-fold higher in TPA-treated cells, as compared with levels in untreated cells, for the duration of the experiment (32 h).
TPA Effects on Nuclear and Cytosolic Levels of Androgen Receptor-Androgen induction of PSA mRNA has been shown to be mediated by the AR which binds to AREs on the PSA promoter (4,5). To determine whether the TPA-associated decreases in androgen-induced PSA mRNA demonstrated in Fig.  1A were due to reduced expression of AR, we examined levels of AR protein. AR protein was detected as a band at approximately 110 kDa in both nuclear and cytosolic extracts (Fig. 3). Nuclear extracts prepared from cells exposed to R1881 had increased levels of nuclear AR, as compared with cells not exposed to androgen. Cytosolic and nuclear levels of AR were not altered by treatment with TPA (1 nM) for 24 h. Therefore, TPA does not appear to decrease PSA gene expression by a mechanism that involves decreasing the nuclear levels of AR protein.
Inhibition of Androgen-induced PSA Promoter Activity by TPA and Overexpression of c-Jun and c-Fos-PSA promoter activity was examined by transient transfection of LNCaP cells with a PSA promoter-luciferase reporter plasmid (pPSA-630). LNCaP cells express endogenous AR, and the addition of 10 nM R1881 to cells resulted in a 5-fold increase of PSA promoter activity (data not shown). Co-transfection of cells with the rat wild type AR (pAR6) expression plasmid (0.5 g/well) resulted in a 34-fold increase in androgen-induced PSA promoter activity in the presence of 10 nM R1881 (Fig. 4A). Therefore, all subsequent studies measuring androgen-induced PSA promoter activities were performed with cells transiently transfected with pAR6.
In Fig. 1, A and C, it was demonstrated that TPA decreased androgen-induced PSA mRNA. In agreement with these data, androgen-induced PSA promoter activity was also inhibited by 39% in LNCaP cells exposed to 1 nM TPA for 24 h (data not shown). Numerous nuclear receptors have been shown to interact with TPA-inducible proteins, c-Jun and/or c-Fos (10). Therefore, to determine whether c-Jun and/or c-Fos are involved in TPA repression of androgen-induced PSA promoter activity, pPSA-630, pAR6, and increasing amounts of c-Jun and/or c-Fos expression plasmids were transfected into LNCaP cells. Androgen-induced PSA promoter activity was inhibited in a dose-dependent manner with transfection of increasing amounts of c-Jun and c-Fos, c-Jun, or c-Fos expression plasmids (in order of descending potency) (Fig. 4B). Overexpression of c-Jun and c-Fos had negligible effects on the empty plasmid GL-2 basic (data not shown) that originates from pUC18 plasmid reported to contain a TRE (31). In addition, overexpression of c-Jun and c-Fos also had no effect on the pGL2-Control vector (Promega), which contains the SV40 promoter inserted into the same Hin-dIII site as the PSA promoter (data not shown).
Inhibition of Androgen-induced PSA Promoter Activity Is Conserved among the jun Family-To investigate the specificity of c-Jun and c-Fos inhibition, we examined the effects of other members of the jun and fos families on androgen-induced PSA promoter activity by transfection of the respective expression vectors into LNCaP cells. All members of the jun family examined significantly attenuated androgen-induced PSA promoter activity relative to RSV-0 values (Fig. 5). However, for members of the fos family investigated, only c-Fos significantly inhibited androgen-induced PSA promoter activity in the absence of jun members. Maximum inhibition of androgen-induced PSA promoter activity was generally observed in cells co-transfected with expression plasmids from both jun and fos families (lanes 8 -16).
Mutual Inhibition of DNA Binding Activity by Co-incubation of AR and c-Jun-EMSA with radiolabeled synthetic oligonucleotide containing a TRE consensus site and nuclear extracts prepared from LNCaP cells exposed to TPA for 6 h showed comparable DNA binding activities in vehicle-treated (0.1% ethanol) and 10 nM R1881-treated cells (Fig. 6A). Thus, nuclear extracts from TPA-treated LNCaP cells are characterized by enhanced AP-1 DNA-binding activity which is reduced by R1881.
The work of Kallio et al. (32) indicated that AR inhibits c-Jun DNA binding activity, while c-Jun does not affect AR DNA binding to the C3 ARE. Similarly, the c-Jun DNA binding activity seen in our experiments (Fig. 6B, lane 3) was also inhibited by peptide fragments of the AR. AR1-GST, containing both the ligand-and DNA-binding domains, was more potent in the inhibition of c-Jun DNA binding activity (lanes 4 -8) than the AR2-GST which contains only the DNA-binding domain (lanes 10 -13). A 5-fold increase in molar ratio of AR1-GST to c-Jun resulted in a 100% decrease in c-Jun DNA binding (lane 6); by comparison, at the same ratio AR2-GST seemed to be less effective, inhibiting c-Jun binding by 94% (lane 11). At equimolar concentrations of AR1 and AR2 to c-Jun, there was an 82 and 75% inhibition of c-Jun DNA binding activity, respectively, which is comparable with that reported by Kallio et al. (32). AR inhibition of c-Jun DNA binding activity could be partially blocked by preincubation of the AR with an antibody to the AR DNA-binding domain before incubation with c-Jun (compare Fig. 6B, lanes 11 and 12, and inset, lanes 6 and 7). Specificity of c-Jun binding to the TRE oligonucleotide was shown by a supershift of the c-Jun⅐c-Jun⅐TRE complex with an antibody to c-Jun (lane 2). AR-GST proteins did not bind to the TRE oligonucleotide (lanes 8 and 13).
To examine whether the presence of c-Jun could alter AR DNA binding activity to the PSA ARE, we investigated several different molar ratios of c-Jun to AR. c-Jun at a molar concentration 10-fold higher than AR1-GST protein levels, caused a 44% decrease in AR1-GST binding to the PSA ARE (Fig. 6C,  compare lanes 2 and 7). c-Jun inhibition of AR DNA binding activity could be completely blocked by preincubation of c-Jun with an antibody to c-Jun prior to incubation with the AR (compare lanes 7 and 9). Specificity of AR-GST protein binding to the PSA ARE was confirmed using three different approaches. These included supershift of the AR⅐ARE complex with an antibody to the AR DNA binding domain (lane 11), lack of AR binding to the ARE oligonucleotide containing a mutated AR half-site (lane 12), and by competition experiments using 100-fold excess unlabeled SP-1 oligonucleotide (nonspecific competitor) (lane 13) and ARE oligonucleotide (specific competitor) (lane 14).
Co-immunoprecipitation of AP-1 with AR-The above results from transfection experiments and EMSA suggest that mutual interference between AR and AP-1 may be due to direct protein-protein interaction. To investigate this possibility, co-immunoprecipitation studies were performed using LNCaP and PC3 cells. The PC3 cell line is a poorly differentiated prostate cancer cell line that does not express PSA or AR (33) and was included as a control. In Fig. 7A (lane 5), c-Jun is immunoprecipitated with the AR from LNCaP cells when using an antibody to the first 21 N-terminal amino acids of the AR (PG-21). As expected, c-Jun could not be immunoprecipitated from PC3 cells when using the PG-21 antibody (lane 4), because this cell line lacks AR. Since the DNA-binding domains of other steroid receptors have been suggested to be involved in protein-protein interaction with c-Jun, we used another antibody that binds to the AR DNA binding domain (15071A). c-Jun could not be immunoprecipitated with the AR when using this antibody (PC3 cells, lane 2; LNCaP cells, lane 3). This suggests that the AR DNA-binding domain is involved in protein-protein interaction between the AR and c-Jun. Similar results were obtained with c-Fos (Fig. 7B).
c-Jun Mutant Analyses-To determine the region(s) of c-Jun that may be involved in protein-protein interaction with the AR, we transfected expression vectors for various c-Jun mutants into LNCaP cells and examined their effects upon androgen-induced PSA promoter activity. Androgen-induced PSApromoter activity observed in cells transfected with pAR6 was inhibited 60% by transfection with wild type c-Jun (Fig. 8).
Deletion of a small region of the N-terminal activator domain (c-Jun-⌬1) did not alter the ability of c-Jun to repress androgen-induced PSA promoter activity. c-Jun inhibition was impaired when the N terminus region of c-Jun was deleted (c-Jun-⌬3). The most pronounced effect was observed upon deletion of the leucine zipper (c-Jun-⌬LZ), which completely abolished the protein's ability to inhibit androgen-induced PSA promoter activity. Therefore the leucine zipper motif of c-Jun may interact with the AR.
AR Mutant Analyses-To define possible AR domains involved in AP-1 repression, the expression plasmid for wild type c-Jun was co-transfected with expression plasmids for several AR mutants. Consistent with Fig. 4, transfection of the wild type AR (pAR6) resulted in over 30-fold induction of PSApromoter activity by R1881 (Fig. 9, lane 2 7. Co-immunoprecipitation of c-Jun (A) and c-Fos (B) with the AR. LNCaP and PC3 cells were exposed to 10 nM R1881 for 24 h prior to the addition of 10 nM TPA for an additional 6 h. Nuclear extracts were precleared with protein A-Sepharose for 30  sulted in slight androgen induction of PSA promoter activity (lane 11) that was not affected by co-transfection with c-Jun (lane 12). Thus, the ligand-binding domain of the AR may be required for inhibition by c-Jun.
Promoter and Cell Specificity-A previous report examining nuclear receptors and AP-1 interaction showed different results depending on the promoter and cell line examined (34). Therefore, we investigated whether AP-1 could inhibit another androgen-induced promoter in LNCaP cells. These cells were transfected with the ARR 3 -tk-luciferase reporter construct, that contains three repeats of the rat probasin ARE1 and ARE2 region ligated in tandom in a luciferase reporter (25). ARR 3 -tkluciferase activity was induced 72-fold in cells treated for 48 h with 10 nM R1881, as compared with cells exposed to vehicle only. LNCaP cells transfected with 0.5 g of c-Jun expression plasmid showed 43% inhibition of androgen-induced ARR 3 promoter activity (Fig. 10A, lane 2). Cells transfected with 0.5 g of c-Fos expression plasmid had 71% inhibition of androgeninduced ARR 3 promoter activity (lane 3). Co-transfection of both c-Jun and c-Fos expression plasmids resulted in an 89% inhibition of androgen-induced ARR 3 promoter activity (lane 4). Hence, there was a difference between the PSA and ARR 3 promoter in AP-1 inhibition of androgen-induced activity in LNCaP cells. c-Jun was more potent than c-Fos in the inhibition of androgen-induced PSA promoter activity (Fig. 4B), while c-Fos was more potent than c-Jun in the inhibition of androgen-induced ARR 3 activity (Fig. 10A).
The cellular specificity of AP-1 inhibition of PSA promoter activity was examined using CV-1 cells (kidney, African green monkey) that are considered to be glucocorticoid receptor (GR)deficient. Co-transfection of pAR6 was required for androgen induction of PSA promoter activity in CV-1 cells. PSA promoter activity was induced 32-fold in cells transfected with 0.5 g pAR6 and exposed to 10 nM R1881 (data not shown). CV-1 cells co-transfected with 0.5 g of both pAR6 and c-Jun resulted in 38% inhibition of androgen-induced PSA promoter activity (Fig. 10B, lane 2). These levels were comparable with those in cells co-transfected with 0.5 g of both pAR6 and c-Fos (43% inhibition of androgen-induced PSA promoter activity) (lane 3). Co-transfection of both c-Jun and c-Fos resulted in a 68% inhibition of androgen-induced PSA promoter activity (lane 4) that was similar to the percentage inhibition seen in LNCaP under the same conditions (Fig. 4B). These results infer that AP-1 inhibition of androgen-induced PSA promoter activity is cell-specific.

DISCUSSION
Cross-talk between signal transduction pathways of nuclear receptors and AP-1 have been reported for the GR (7,8,35,36), retinoic acid receptor (37,38), estrogen receptor (39,40), vitamin D 3 receptor (38,41), and thyroid hormone receptor (42,43). To date however, very little has been reported about interaction between the AR and AP-1 (32,34,44). Interaction between nuclear receptors and AP-1 appears to involve numerous mechanisms including: 1) overlap of DNA binding sites of the nuclear receptor with AP-1; 2) a composite DNA binding site to which both nuclear receptor and AP-1 bind; 3) mutual transcriptional inhibition via protein-protein interaction (10); and 4) possibly sequestering of a common co-activator such as the CREB-binding protein (45). However, interaction between nuclear receptors and AP-1 may be gene-specific, cell-specific, and depend on endogenous levels and/or ratios of Jun to Fos and/or the composition of the AP-1 dimers (9,34).
TPA is a tumor promoter that induces AP-1 transcription via the PKC signal transduction pathway (30). Indeed in our present experiments we show a TPA-related decrease in androgeninduced PSA gene expression (Fig. 1, A-C) with a preceding increase in AP-1 (Figs. 1D and 2). However, we did not observe reduced nuclear levels of androgen-stimulated AR protein (Fig.   3). Since the AR is a phosphoprotein (46,47) that is hyperphosphorylated upon androgen stimulation (48), TPA alteration of PKC signal transduction pathways may lead to decreased transcriptional activity of AR due to alterations in its phosphorylation state or other proteins involved in the AR signal transduction pathway leading to increased PSA gene expression. However, to date there are no reports describing decreased transcriptional activity of the AR due to alteration of its phosphorylation state. Furthermore, the data presented here support the hypothesis that TPA induction of AP-1 leads to inhibition of androgen-induced PSA gene expression.
Inhibition of androgen-induced PSA promoter activity by overexpression of c-Jun and c-Fos suggests that TPA-associated decreases in PSA gene expression may be the result of increased AP-1 and interaction between the AR and these proteins. The order of potency of inhibition from co-transfection experiments observed in LNCaP cells was: c-Jun and c-Fos Ͼ c-Jun alone Ͼ c-Fos alone (Fig. 4B). These results would be consistent with cooperativity between c-Jun and c-Fos. Alternatively, they are compatible with a requirement for AP-1 dimerization, since Jun homodimers are considerably less stable than the heterodimeric Jun/Fos complex (49,50). However, dimer formation would theoretically mask the leucine zipper of c-Jun, a domain that our results (Fig. 8) suggest may be involved in AR/AP-1 interaction. Another study examining AP-1 modulation of androgen induction found that an increase in the intracellular levels of c-Jun stimulated, while c-Fos inhibited the transcriptional activation induced by androgen in numerous cell lines (34). However, these responses appeared to be cell-specific, which we also see when comparing PSA promoter activity in LNCaP cells (Fig. 4) and CV-1 cells (Fig. 10B), and gene-specific as shown by differences in AP-1 inhibition of PSA promoter activity (Fig. 4) and ARR 3 activity (Fig. 10A) in LN-CaP cells.
GR studies show mutual inhibition of DNA binding as the result of protein-protein interaction between the GR and AP-1 and do not show GR binding to TREs or vice versa (7,8). Similarly our EMSA results with the AR and AP-1 do not show any evidence for c-Jun binding to the PSA ARE (Fig. 6C) nor was there evidence for AR binding to the TRE (Fig. 6B). In agreement with the lack of AP-1 binding to the PSA ARE, the sequence of this region bears no homology to the TRE consensus site. Despite the lack of DNA sequence homology between the consensus TRE and the PSA ARE, incubation of the AR with c-Jun leads to loss of mutual DNA binding activities. DNA binding activity could be partially to fully recovered by preincubation of either protein with specific antibodies. Collectively, these data strongly suggest that the mutual inhibition of DNA binding activities may result from the formation of abortive heterodimers or complexes between the two proteins. This hypothesis was supported by the demonstration that c-Jun and c-Fos are co-immunoprecipitated with the AR (Fig. 7).
The zinc finger motifs of GR are thought to interact with the DNA binding and/or leucine zipper domains of c-Jun (7,8). The homology between the GR and AR protein is 77% in the zinc finger regions. It is thus not surprising that results from EMSA and mutant analyses with both AR (Fig. 9) and c-Jun (Fig. 8) imply that the same structural components of AR and c-Jun may interact. In addition, the results of co-immunoprecipitation studies indicate that c-Jun and c-Fos are co-precipitated with the AR by an antibody to the AR N terminus region (Fig.  7). In contrast, neither c-Jun nor c-Fos could be co-precipitated with AR by an antibody to the AR DNA-binding domain (Fig.  7). These observations draw attention to the importance of the AR DNA-binding domain for AR and AP-1 interaction. Furthermore, the AR ligand-binding domain also appears to be in- FIG. 10. Promoter and cell specificity. A, AP-1 inhibition of androgen-induced ARR 3 activity in LNCaP cells. These cells were transfected with ARR 3 -tk-luciferase (1 g/well), pAR6 (0.5 g/well) with or without c-Jun (0.5 g), c-Fos (0.5 g), or c-Jun plus c-Fos (0.5 g of each) expression plasmids. Subsequent to transfection, cells were incubated for 48 h in serum-free RPMI 1640 containing 2% TCM™ in the presence of 10 nM R1881. The percent inhibition of relative luciferase activities is expressed using controls (with RSV-0) set at 0% and represent the averages of three individual experiments. All lanes represent samples exposed to R1881. Lanes: 1 pAR6; 2, pAR6 plus c-Jun; 3, pAR6 plus c-Fos; 4, pAR6 plus c-Jun plus c-Fos. B, AP-1 inhibition of androgeninduced PSA promoter activity in CV-1 cells. These cells were transfected with pPSA-630 (1 g/well), pAR6 (0.5 g/well) with or without c-Jun (0.5 g), c-Fos (0.5 g), or c-Jun plus c-Fos (0.5 g of each) expression plasmids. Subsequent to transfection, cells were incubated for 48 h in serum-free RPMI 1640 containing 2% TCM™ in the presence of 10 nM R1881. The percent inhibition of relative luciferase activities is expressed using controls (with RSV-0) set at 0% and represents the averages of three individual experiments. All lanes represent samples exposed to R1881. Lanes: 1, pAR6; 2, pAR6 plus c-Jun; 3, pAR6 plus c-Fos; 4, pAR6 plus c-Jun and c-Fos. volved in the interaction between the AR and c-Jun (Figs. 6 and 9). EMSA data supporting the view that the AR DNA-and ligand-binding domains may interact with c-Jun is depicted in Fig. 6B. In these studies, two different AR-GST proteins were employed and the AR protein consisting of both the DNA-and ligand-binding domains (AR1) was more potent in repression of AP-1 DNA binding activity than the identical AR protein lacking the ligand-binding domain (AR2). However, the AR2 protein, which contains the DNA-binding domain and hinge region, was capable of inhibiting c-Jun DNA binding activity. Furthermore, the observation that preincubation of the AR with an antibody to its DNA-binding domain only partially recovered c-Jun DNA-binding activity, implies that another domain besides the DNA-binding domain may play a role in the interaction with AP-1. In agreement with our results, a recent study using a form of the yeast two-hybrid system demonstrated interaction between c-Jun and the DNA-binding domain and hinge region of the AR (44).
The leucine zipper structure of c-Jun mediates dimer formation, which in turn facilitates DNA binding by the basic region. Thus, deletion of this structure would render the protein biologically inactive complicating interpretation of transfection data. However, results from transfecting c-Jun deletion mutants have suggested interaction between the GR and leucine zipper structure of c-Jun (8). Similarly, we have found a loss of c-Jun inhibition of androgen-induced PSA promoter activity when cells are transfected with the expression vector for a mutant c-Jun protein lacking the leucine zipper region (Fig. 8). These data support the view that the leucine zipper is of key importance in AR and c-Jun interaction. Consistent with this hypothesis is the observation of conservation of inhibition of androgen-induced PSA promoter activity in the jun family. Our results indicate that c-Jun, Jun B, and Jun D were all similar in potency for repression of androgen-induced PSA promoter activity (Fig. 5), which implies that conserved Jun domains are involved. Homology is shared between Jun proteins in the C terminus region, including DNA-binding and leucine zipper domains (51).
While the jun family was similar in potency of repression of androgen-induced PSA promoter activity, it is curious that from the fos family only c-Fos significantly reduced androgeninduced PSA promoter activity (Fig. 5). The DNA-binding domains of c-Fos and c-Jun are the only regions of these proteins that are conserved (52). Similarly the DNA-binding domains of c-Fos, Fra 1, and Fra 2 are also conserved (53), but we did not detect repression of androgen-induced activity with either Fra 1 or Fra 2, thereby suggesting that fos family proteins may interact with the AR via other domains yet to be established.
In summary, our experiments have shown that the AR and c-Jun each can inhibit the ability of the other to bind to its respective DNA-binding site in vitro. We propose that this mutual inhibition may result from the formation of abortive heterodimers or complexes between the two proteins, similar to complex formation between the GR and AP-1 in some systems. If this mechanism is active in vivo, mutual transrepression of AR and AP-1 would constitute a mechanism of transcriptional regulation of genes, e.g. PSA and collagenase, containing either an ARE or TRE or both. It follows that such a mechanism might depend on the ratio of AR to c-Jun not excluding other AP-1 proteins or co-regulators. The possible importance of the ratio of AR to AP-1 protein is especially relevant in prostate cells in which the promoter region of the PSA gene may contain several TREs (4,54). In a situation where the ratio of c-Jun to AR is high, there would be less AR available for binding to any ARE to initiate transcription. In contrast, there would be excess c-Jun available for binding to PSA TREs, a condition that conceivably might result in androgen-independent stimulation of PSA gene expression. In this regard, it is of interest that the PC3 cell line, which has progressed to an advanced stage of androgen-independence, is characterized by a 7-fold greater intracellular concentration of c-Jun relative to that in the more differentiated LNCaP cell line. 2 The possible role of c-Jun and its related family of proteins in contributing to the androgenindependent regulation of the PSA gene appears worthy of further investigation.