Novel Binding of HuR and Poly(C)-binding Protein to a Conserved UC-rich Motif within the 3 (cid:1) -Untranslated Region of the Androgen Receptor Messenger RNA*

The androgen receptor (AR) mediates androgen action and plays a central role in the proliferation of specific cancer cells. We demonstrated recently that AR mRNA stability is a major determinant of AR gene expression in prostate and breast cancer cells and that androgens differentially regulate AR mRNA decay dependent on cell type (Yeap, B. B., Kreuger, R. G., Leedman, P. J. (1999) Endocrinology 140, 3282–3291). Here, we have identified a highly conserved UC-rich region in the 3-untranslated region of AR mRNA that contains a 5 (cid:1) -C(U) n C motif and a 3 (cid:1) -CCCUCCC poly(C)-binding protein motif. In transfection studies with LNCaP human prostate cancer cells,

The androgen receptor (AR) mediates androgen action and plays a central role in the proliferation of specific cancer cells. We demonstrated recently that AR mRNA stability is a major determinant of AR gene expression in prostate and breast cancer cells and that androgens differentially regulate AR mRNA decay dependent on cell type (Yeap, B. B., Kreuger, R. G., Leedman, P. J. (1999) Endocrinology 140, 3282-3291). Here, we have identified a highly conserved UC-rich region in the 3-untranslated region of AR mRNA that contains a 5-C(U) n C motif and a 3-CCCUCCC poly(C)-binding protein motif. In transfection studies with LNCaP human prostate cancer cells, the AR UC-rich region reduced expression of a luciferase reporter gene. The AR UC-rich region was a target for cytoplasmic and nuclear RNAbinding proteins from human prostate and breast cancer cells as well as human testicular and breast cancer tissue. One of these proteins is HuR, a ubiquitously expressed member of the Elav/Hu family of RNA-binding proteins involved in the stabilization of several mRNAs. Poly(C)binding protein-1 and -2 (CP1 and CP2), previously implicated in the control of mRNA turnover and translation, also bound avidly to the UC-rich region. Mutational analysis of the UC-rich region identified specific binding motifs for both HuR and the CPs. HuR and CP1 bound simultaneously to the UC-rich RNA and in a cooperative manner. Immunoprecipitation studies confirmed that each of these proteins associated with AR mRNA in prostate cancer cells. In summary, we have identified and characterized a novel complex of AR mRNA-binding proteins that target the highly conserved UC-rich region. The binding of HuR, CP1, and CP2 to AR mRNA suggests a role for each of these proteins in the post-transcriptional regulation of AR expression in cancer cells.
The interaction between androgens such as dihydrotestosterone (DHT) 1 and the androgen receptor (AR) plays a crucial role in the evolution of human prostate cancer, and modulation of this interaction represents a major therapeutic strategy (1,2). The biological action of androgen is mediated via binding to the AR, a member of the nuclear hormone receptor family of transcription factors that regulate expression of androgen-responsive genes (3). Androgens promote the growth of prostate cancer cells, and androgen deprivation results in regression of most prostate cancers, although this effect is eventually abrogated by the emergence of androgen-insensitive disease. In contrast, AR expression correlates with better clinical outcome in human breast cancer (4,5), and androgens have been shown to inhibit the growth of breast cancer cells expressing the AR (6,7). To better understand the mechanisms underlying the expression of the AR gene, we recently utilized LNCaP prostate cancer cells and MDA453 breast cancer cells to determine the relative importance of transcriptional and post-transcriptional mechanisms in androgen-regulated AR gene expression (8). In LNCaP cells, androgens down-regulate transcription of the AR gene, but stabilize AR mRNA. In MDA453 cells, androgens produce rapid destabilization of AR mRNA in the absence of transcriptional change. These data suggest an important role for post-transcriptional events (in particular, AR mRNA turnover) in the regulation of AR gene expression in prostate and breast cancer cells. We therefore decided to investigate the mechanisms regulating AR mRNA stability. mRNA turnover is a central mechanism in the control of gene expression (9), where interactions between specific sequences within mRNA (cis-acting elements) and cellular RNA-binding proteins (trans-acting factors) regulate ribonuclease action and subsequent mRNA decay rates. Several cis-acting elements have been described, and the best characterized is the adenylate/uridylate (AU)-rich element (ARE). The ARE most often resides in the 3Ј-untranslated region (3Ј-UTR) and contains AUUUA pentamers and/or UUAUUUA(U/A)(U/A) nonamers (10 -12). AREs have been implicated in regulating the stability of labile cytokine and growth factor mRNAs via interactions with RNA-binding proteins such as the 44-kDa AU-binding factor (13,14) and the 36-kDa HuR protein (15,16). HuR is a ubiquitously expressed member of the Elav/Hu family of RNAbinding proteins, whose other members include Hel-N1, HuC, and HuD (17). HuR has been shown to stabilize other mRNAs, including vascular endothelial growth factor (18) and p21 WAF1 (19). In addition, HuR functions as a nuclear/cytoplasmic transport protein for mRNAs (17, 20 -22). Other non-ARE cis-elements have been described, including the U-rich polypyrimidine tract with a C(U) n C motif (23). For example, a C(U) n C motif is present within a region of c-myc mRNA known to bind HuD, another member of the Elav/Hu family (24,25). In addition, a C-rich cis-element with a consensus CCUCC or CCCUCCC sequence has been identified in a variety of mRNAs, including ␣-globin and erythropoietin (26 -28). This C-rich element is the target for several poly(C)-binding proteins (CPs; ϳ43 kDa) (27)(28)(29). The CPs have been implicated in both regulation of mRNA stability and translation initiation (27)(28)(29)(30)(31).
Interestingly, the human AR mRNA does not contain an ARE (32). However, it does have a UC-rich region in the proximal 3Ј-UTR that shares significant homology with the U-and C-rich sequences described above and that could act as a ciselement and interact with RNA-binding proteins. Here, we show that this highly conserved UC-rich cis-element binds several cytoplasmic and nuclear proteins from LNCaP and MDA453 cells as well as from human testicular and breast cancer extracts. One of these trans-acting factors is HuR. Two other proteins interacting with this region are poly(C)-binding protein-1 and -2 (CP1 and CP2). These are the first described AR mRNA-binding proteins. Furthermore, these proteins bind to the UC-rich region simultaneously, and each is closely associated with AR mRNA in LNCaP prostate cancer cells. Thus, the UC-rich region of AR mRNA is a target for at least three RNA-binding proteins that may have significant impact on the mRNA stability and/or translation of AR mRNA.
Cell Culture and Tissue Specimens-LNCaP (American Type Culture Collection CRL-1740), DU145 (HTB-81), and MDA453 (HTB-131) cells were grown in phenol red-free RPMI 1640 medium supplemented with 10% fetal calf serum (dextran-charcoal stripped for experiments), 50 units/ml penicillin G, and 50 g/ml streptomycin (Invitrogen). DHT (Sigma) was used at a final concentration of 10 nM in cell culture media. Human normal testicular and breast cancer tissue was obtained from fresh surgical specimens after the required pathological examination was performed. Tissues were stored at Ϫ80°C until used for preparation of cytoplasmic extracts.
Transient Transfection and Luciferase Assay-LNCaP cells were grown to 70% confluence prior to transfection with 2 g of either pRSV-Luc or pRSV-LucϩUC using FuGENE 6 transfection reagent (Roche Molecular Biochemicals GmbH, Mannheim, Germany). Cells were cotransfected simultaneously with 0.5 g of pRSV-␤-gal as a transfection control. Luciferase assays were performed in a Berthold AutoLumat LB953 luminometer. ␤-Galactosidase activity was assayed as previously described (37).
Preparation of Cytoplasmic Extracts-Cytoplasmic extracts from LN-CaP, DU145, and MDA453 cells were prepared as previously described (38,39) and stored in aliquots at Ϫ80°C. Human testicular and breast cancer tissue was homogenized in buffer containing 0.32 M sucrose, 3 mM MgCl 2 , 40 mM KCl, 10 mM HEPES (pH 7.5), 1 mM DTT, 2 g/ml aprotinin, 0.5 mM PMSF, and 10 g/ml leupeptin using a sterile Dounce homogenizer. The homogenate was centrifuged at 4°C for 10 min at 1000 ϫ g, and the resultant supernatant was centrifuged again at 4°C for 1 h at 130,000 ϫ g, followed by storage of the final supernatant in aliquots at Ϫ80°C. Protein concentrations were estimated by the Bradford method (Bio-Rad).
Preparation of Nuclear Extracts-Nuclear extracts were prepared using a modified method of Ausubel et al. (40). Briefly, LNCaP cells were washed, suspended in cold phosphate-buffered saline, and centrifuged at 4°C for 5 min at 1800 ϫ g. The pellet was resuspended twice in hypotonic buffer (10 mM HEPES (pH 7.5), 1.5 mM MgCl 2 , 10 mM KCl, 0.2 mM PMSF, and 0.5 mM DTT) and kept on ice for 10 min. The cell suspension was homogenized with 10 strokes of a Dounce homogenizer and centrifuged at 4°C for 15 min at 1500 ϫ g. The resultant pellet of nuclei was resuspended directly in buffer containing 20 mM HEPES (pH 7.5), 25% glycerol, 1.5 mM MgCl 2 , 0.5 M KCl, 0.2 mM EDTA, 0.2 mM PMSF, and 0.2 mM DTT and kept on ice for 30 min with regular mixing. An equal volume of buffer containing 20 mM HEPES (pH 7.5), 25% glycerol, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.2 mM PMSF, and 0.2 mM DTT was added, followed by centrifugation at 4°C for 30 min at 14,000 ϫ g. The protein concentration of the supernatant was determined as described above prior to storage in aliquots at Ϫ80°C.
Recombinant HuR domains I and IϩII were prepared similarly to the GST-HuR fusion protein described above. Each construct was cloned into pGEX-4T1 and expressed in E. coli BL21-Codon Plus (Stratagene). After induction with 0.5 mM isopropyl-␤-D-thiogalactopyranoside to A 600 ϭ 0.8, cells were resuspended in 50 mM Tris (pH 7.8), 200 mM NaCl, 1 mM EDTA, and 2 mM PMSF. Cells containing HuR domain III were resuspended in phosphate-buffered saline (pH 7.3), 1 mM EDTA, 0.5% Triton X-100, and 2 mM PMSF. Cells were lysed with a French press; the lysate was centrifuged at 15,000 ϫ g for 30 min; and GST fusion proteins were purified using glutathione-agarose beads with eight washes in the resuspension buffer without PMSF and EDTA. HuR fragments were cleaved from GST by the addition of 2.5 units/ml thrombin. HuR domain I was further purified by size-exclusion chromatography. The protein solution was dialyzed against 1000 times the volume of 50 mM Tris (pH 7.8) and applied to a Bio-Rad Bio-Prep SE 1000/17 column equilibrated in the same buffer. HuR domain IϩII was further purified by size-exclusion and cation-exchange chromatography. The protein solution was dialyzed against 1000 times the volume of 50 mM MES (pH 6.2) and applied to the Bio-Rad Bio-Prep SE 1000/17 column equilibrated in the same buffer. The HuR domain IϩII preparation was then eluted from a Bio-Rad UNO S6 cation-exchange column with a linear gradient at 40% 50 mM MES (pH 6.2) and 1 M KCl. CP1 was prepared as previously described by Kiledjian et al. (41).
Synthesis of Radiolabeled RNA Transcripts-Linearized templates, ARWt, various mutant AR probes, and c-fos HuD were used in transcription reactions containing [ 32 P]UTP (3000 Ci/mmol; Amersham Biosciences) with T7 RNA polymerase (Invitrogen) as described (38,39) to produce transcripts with a specific activity of Ϸ1 ϫ 10 9 cpm/g of RNA.
Full-length transcripts were purified as described (39). Unlabeled transcripts were prepared as described above, but column-purified.
RNA Electrophoretic Mobility Shift Assay (REMSA)-REMSA was performed as described previously (35,38,39,42), and complexes were detected using a PhosphorImager and quantified using ImageQuant (Molecular Dynamics, Inc., Sunnyvale, CA). For competition assays, unlabeled RNA transcripts (75-450 ng) or RNA homopolymers (75-300 ng; poly(U), poly(C), poly(A), Amersham Biosciences) were added to the reaction mixture for 10 min at 22°C prior to incubation with labeled RNA probe. When REMSA was performed with GST fusion protein (ϳ100 -400 ng), the CEB buffer contained 4 mM DTT, MnCl 2 , ZnCl 2 , and 10 mM sodium orthovanadate. For antibody supershifts with cytoplasmic and nuclear extracts, 1 l of anti-HuR (ϳ7 mg/ml) or 0.5-1.0 l of anti-CP (ϳ2 mg/ml) antibody was added after the RNase T1 step. For antibody supershifts of labeled riboprobe⅐GST fusion protein complexes, the GST fusion proteins were incubated with the CEB buffer and 150 g of heparin for 10 min at 22°C prior to incubation with labeled riboprobe. The antibody was then added, and the reaction was incubated on ice for 1 h and analyzed by PAGE as described above.
UV Cross-linking and Immunoprecipitation Assay-The method for the UV cross-linking (UVXL) assay was as described (39,42). For UV cross-linking/immunoprecipitation (UVXL/IP) studies, 50 -100 g of cytoplasmic extract and 500,000 cpm of probe were used, and the reaction volume was increased to 300 l by addition of the CEB buffer after the RNase A incubation step. Two l of anti-HuR or 1.5 l of anti-CP1 or anti-CP2 antibody was added, and the mixture was rotated at 4°C for 1 h. Protein G-Sepharose (30 l; Sigma) was used to precipitate the anti-HuR antibody⅐RNA⅐protein complexes, and protein A-Sepharose (Amersham Biosciences) for the anti-CP antibody complexes. The RNA⅐protein complexes were resolved by 8.5% SDS-PAGE and visualized using a PhosphorImager.
Immunoprecipitation/Reverse Transcription-PCR Assay-LNCaP or MDA453 cells were grown to 70% confluence in RPMI 1640 medium supplemented with 10% fetal calf serum in 75-cm 2 flasks. The cells were harvested, washed twice with cold phosphate-buffered saline, and then lysed for 40 min on ice in 1.5 ml of CEB buffer containing standard protein inhibitors and 100 units of RNaseOUT (Invitrogen). The lysed cells were centrifuged at 12,000 ϫ g for 10 min at 4°C, and 0.75 ml of supernatant was added to 10 g of anti-HuR, anti-CP1, anti-CP2, or anti-p21 WAF1 antibody. A control tube containing only lysate was also prepared. After incubation at 4°C for 30 min, 5 g each of protein A and protein G beads were added to all tubes, which were gently mixed for a further 30 min at 4°C. The tubes were centrifuged at 2000 ϫ g for 2 min, and the supernatants were removed for RNA extraction. The precipitated beads were washed with cold CEB buffer (5 ϫ 1 ml) and then RNA-extracted with TRIzol reagent (Invitrogen). Reverse transcription with oligo(dT) primer was performed on the extracted RNA using standard procedures. PCR was performed for 35 cycles with denaturation at 95°C, annealing at 58°C, and extension at 72°C with primers 5Ј-ATGAACTTCGAATGAACTACATCAAGG-3Ј and 5Ј-GGAA-CATGTTCATGACAGACTGTAC-3Ј from exons 7 and 8 of the ARwt sequence. The PCR products were resolved on a 1% agarose gel.
Data Analysis-Transfection data are shown as means Ϯ S.E. Statistical analysis was performed using Student's t test with a p value of Ͻ0.05 regarded as significant.

The 3Ј-UTR of AR mRNA Contains a Conserved UC-rich
Region That Is a cis-Acting Element-Based on our recent demonstration that AR mRNA stability plays an important role in the regulation of AR gene expression in at least two cancer cell types (prostate and breast) (8), we undertook a detailed analysis of the AR mRNA to identify any regulatory cis-elements. Examination of the AR mRNA sequence (32) revealed there were no AREs present. However, it does contain a UCrich region in the proximal portion of the 3Ј-UTR that is similar to other U-rich cis-elements. Interestingly, the location of the 51-nucleotide sequence is between 153 and 156 nucleotides from the stop codon in the human, mouse, and rat species (Fig.  1A). At the 5Ј-end of this region is a CUGGG motif conserved in all species. In the human AR mRNA, this 5-nucleotide sequence is followed immediately by a C(U) 9 C sequence, part of a 45-nucleotide UC-rich element comprising 29 uridine and 16 cytidine residues. This sequence is identical to that found in LNCaP cells (GenBank TM /EBI accession number M34233). The C(U) n C motif closely resembles the polypyrimidine tracts of Adenovirus 10 pre-mRNA and phosphatidylinositol 3,4,5trisphosphate RNA bound by the polypyrimidine tract-binding protein (23) (Fig. 1A). Additionally, Hel-N1 (human elav-like neuronal protein-1), a member of the Elav/Hu family of RNAbinding proteins, has been shown to interact with comparable U-rich regions that need not contain classical AUUUA or UUAUUUA(U/A)(U/A) sequences (43). Similarly, a C(U) n C stretch is present within a region of c-myc mRNA known to bind HuD, another member of the Elav/Hu family that is restricted to the central nervous system (24,25) (Fig. 1A). Interestingly, the 3Ј-end of the human AR mRNA UC-rich sequence contains a CCCUCCC motif that is conserved across species (Fig. 1A). The sequence CCUCC is a recognized human consensus RNA-binding site for the CPs (44). Based on these sequence homologies, we reasoned that the AR UC-rich region may function as a cis-element in the regulation of AR mRNA turnover and be a target for the ubiquitous HuR protein and the CPs.
To evaluate whether the AR UC-rich region may play a role in AR mRNA turnover, we generated constructs for use in a transfection assay utilizing a luciferase reporter in LNCaP cells. Initially, we examined the change in basal luciferase activity induced by the UC-rich region ( Fig. 2A). The presence of the AR UC-rich sequence significantly reduced reporter activity by 30% ( Fig. 2A). Although there was a trend toward increased reporter activity after the cells were treated with DHT, this was not significant. These data support the notion that the AR UC-rich region functions as a cis-acting element contributing to the regulation of AR mRNA turnover in LNCaP prostate cancer cells.
The AR mRNA UC-rich Element Is a Target for Cytoplasmic RNA-binding Proteins from a Variety of Tissues-As the AR UC-rich element was capable of regulating luciferase reporter activity, we investigated whether it is a target for RNA-binding proteins using REMSA. Two distinct RNA⅐protein complexes that bound to the 32 P-labeled wild-type AR UC-rich ( 32 P-ARWt) transcript were identified from LNCaP cells (Fig. 2B, lane 2). Competition studies with unlabeled RNA confirmed specificity for the AR UC-rich probe (Fig. 2B, lanes 3-8). Comparable RNA⅐protein complexes were found in cytoplasmic extracts from ARϩve MDA453 human breast cancer cells (Fig. 2B, lane 9), ARϪve DU145 human prostate cancer cells (lane 10), human testicular tissue (lane 11), and human breast cancer tissue (lane 12). Thus, the proteins binding the AR mRNA UC-rich region are widely distributed in normal human testicular tissue and human prostate tumor cells that are both ARϩve and ARϪve as well as in ARϩve breast cancer cells and breast cancer tissue. The intensity of the RNA⅐protein complexes was not modulated by DHT in LNCaP or MDA453 cells (data not shown).
To identify possible candidates for the AR mRNA-binding proteins, we next attempted to compete each AR RNA⅐protein complex with excess RNA homopolymer. Addition of poly(U) abrogated the slower migrating RNA⅐protein complex (Fig. 2C, lanes [2][3][4][5]. This effect was apparent at the lowest concentration of poly(U) RNA of 7.5 ng/l (Fig. 2C, lane 2). Interestingly, unlabeled poly(C) abolished the faster migrating RNA⅐protein complex (Fig. 2C, lanes 7-10), but neither complex was altered by addition of unlabeled poly(A) even at the highest concentration of 30 ng/l (lanes [11][12][13][14]. Taken together, these data suggested that each AR RNA⅐protein complex comprised separate distinct RNA-binding proteins. The slower migrating complex (RPC1) was likely to contain proteins binding specifically to U-rich sequences, whereas the faster migrating complex (RPC2) was likely to contain one or more CPs.
Multiple RNA-binding Proteins Interact with the AR mRNA UC-rich Sequence-UVXL analysis was performed to further define the proteins binding the UC-rich region. Multiple distinct RNA-binding proteins from LNCaP cytoplasmic extract with molecular masses of Ϸ110, 87, 60, 47, 43, and 36 kDa bound the 32 P-ARWt transcript (Fig. 2D, lane 2). Specificity of binding of each protein was determined as for described for REMSA (Fig. 2D, lanes 3-6). When excess unlabeled pARWt was added, there was a marked decrease in binding of all complexes (Fig. 2D, lane 3). Addition of excess poly(U) abolished the Ϸ110-, 87-, 60-, 47-, and 36-kDa complexes (Fig. 2D,  lane 4). Excess poly(A) had no effect on the binding pattern (Fig. 2D, lane 5). However, addition of poly(C) abrogated binding of the 43-kDa complex without affecting the 36-kDa com-plex or the higher molecular mass complexes (Fig. 2D, lane 6). Further resolution of the 40 -50-kDa RNA⅐protein complexes showed that unlabeled pARWt and poly(C) reduced specifically the 43-kDa complex (Fig. 2E, lanes 3 and 6). Taken together, these data suggest that the Ϸ110-, 87-, 60-, and 36-kDa complexes are likely to represent RNA-binding proteins that target AU-or U-rich mRNA motifs, whereas the 43-kDa complex may contain a protein(s) that has high affinity for poly(C) mRNA motifs.
HuR Binds to the UC-rich Region of AR mRNA-Given the close similarity between the UC-rich region and the reported RNA target sequences for Hel-N1 (43) and HuD (24, 25), we FIG. 1. A, sequence alignment of UC-rich tracts within the 3Ј-UTRs of the AR and several other mRNAs. The conserved CUGGG and CCCUCCC motifs in AR mRNAs from human, mouse, and rat sequences and in human tyrosine hydroxylase mRNA are boxed. The C(U) n C (n ϭ 9) motif in human and mouse AR mRNAs is shaded, as is the same motif in polypyrimidine tract-binding protein-bound sequences (n ϭ 8) and in the c-myc mRNA sequence (n ϭ 10) binding HuD (underlined). Consensus CCUCC-type sequences are underlined in human ␣-globin and erythropoietin mRNAs. Note that in human ␣-globin mRNA, the second underlined sequence reveals multiple, overlapping motifs. The consensus U(U/C)CCCU sequence is underlined in human and mouse AR mRNAs and rat and human tyrosine hydroxylase mRNAs. The two CCCUCUU motifs in the rabbit 15-lipoxygenase mRNA are shown. GenBank TM /EBI accession numbers are indicated in parentheses. Bothwell 1991 refers to Ref. 23. B, schematic detailing of the AR mRNA UC-rich sequence and mutant plasmid clones. The major U-and C-rich regions within the AR mRNA are highlighted in the diagram, and all of the regions that were targeted for mutation are underlined in the wild-type sequence to the right. The general location of each mutation (Mut-1-7) is indicated in the diagram with an "X," and the corresponding mutated sequence is shown underlined and in boldface to the right. The sequence numbers refer to the human AR nucleotide sequence (GenBank TM /EBI accession number M20132). Ad10, Adenovirus 10; PIP3, phosphatidylinositol 3,4,5-trisphosphate. examined whether the 36-kDa RNA⅐protein complex contains HuR, a ubiquitous member of the Elav/Hu family of RNAbinding proteins (34). A monoclonal antibody against HuR supershifted part of the slower migrating RNA⅐protein complex (RPC1) in REMSA with 32 P-labeled pARWt and LNCaP extract (Fig. 3A, lane 3). No supershift was observed upon addition of the control anti-p21 WAF1 antibody (Fig. 3A, lane 4). A supershift was also apparent with anti-HuR antibody and the 32 Plabeled c-fos HuD probe (Fig. 3A, lane 7), which contains a high affinity binding site for HuR (34). To further investigate this interaction, we generated recombinant HuR. GST-HuR fusion protein bound avidly to 32 P-labeled pARWt in REMSA (Fig. 3B,  lane 2). The GST-HuR⅐pARWt RNA complex was also supershifted with anti-HuR antibody (Fig. 3B, lane 3). A similar supershift was observed using anti-HuR antibody, the 32 Plabeled c-fos HuD probe, and GST-HuR (Fig. 3B, lane 7). To validate these results with another approach, we performed a UVXL/IP assay. Anti-HuR antibody specifically immunoprecipitated a complex at 36 kDa from LNCaP cells with 32 Plabeled pARWt (Fig. 3C, lane 3), confirming direct binding of HuR to the UC-rich region of AR mRNA and the identity of the 36-kDa band upon UVXL. As a control, the UVXL/IP experiment was performed with anti-EGFR antibody, which did not precipitate any complexes (Fig. 3C, lane 4).
Poly(C)-binding Proteins (CP1 and CP2) Bind to the UC-rich Region of AR mRNA-Several observations suggested that CP1 and/or CP2 might be targeting the AR UC-rich region. First, analysis of the UC-rich region revealed a conserved CCCUCCC sequence identified as a component of the CP-binding motif in erythropoietin mRNA (see Fig. 1A) (28). In addition, the consensus CP-binding motif in tyrosine hydroxylase mRNA, UUC-CCU (45), is also present contiguous to this conserved region in the human and further 5Ј in the mouse AR mRNA (Fig. 1A). Second, poly(C) homopolymer competition abolished the faster migrating RNA⅐protein complex in REMSA. Third, poly(C) competition in UVXL assay specifically reduced binding of the major 43-kDa RNA-binding protein.
Next, we used anti-CP1 and anti-CP2 antibodies in REMSA and UVXL/IP experiments and generated CP1 and CP2 fusion proteins. Anti-CP1 and anti-CP2 antibodies each produced a prominent supershift and significantly reduced binding of the faster migrating RNA⅐protein complex (RPC2) (Fig. 4A, lanes 3  and 4). Addition of increasing amounts of anti-CP1 and anti-CP2 antibodies together completely abolished RPC2 (Fig. 4A,  lanes 5 and 6). This confirmed binding of both CP1 and CP2 to the UC-rich probe and verified the identity of the majority of proteins within the RPC2 complex. No supershift was observed with anti-p21 WAF1 or anti-EGFR antibody (Fig. 4A, lanes 7  and 8).
We also performed REMSA supershifts with LNCaP cell nuclear extract (Fig. 4B). A similar profile of complexes was observed, with preserved relative intensities of RPC1 versus RPC2 (Fig. 4B, lane 2). A supershift with anti-HuR antibody was not observed (Fig. 4B, lane 3). Although addition of anti-CP1 and anti-CP2 antibodies abolished RPC2 (Fig. 4B, lanes  4 -6), a prominent supershift was observed only with anti-CP2 antibody (lane 5). To further investigate differences between LNCaP cytoplasmic and nuclear extract RNA-protein binding, we directly compared them in UVXL assays. Several complexes migrated similarly in both cytoplasmic and nuclear extracts (Fig. 4C, lanes 2 and 3, respectively). In particular, complexes at 36 and 43 kDa were observed in nuclear extract (Fig. 4C,  lane 3), consistent with a proportion of HuR, CP1, and CP2 being located in the nucleus of LNCaP cells. These data suggest that each of these proteins may have a role in binding AR mRNA in both the cytoplasm and nucleus. The reasons for the lack of supershift using anti-HuR and anti-CP1 antibodies in nuclear extract may relate to a lower relative concentration of each protein in the nucleus, despite their detection with the UVXL assay.
To further examine the contribution of CP1 and CP2 to AR mRNA⅐protein complexes, UVXL/IP was performed. Using the 32 P-labeled pARWt and LNCaP cytoplasmic extracts, we immunoprecipitated a single complex at 43 kDa with anti-CP1 antibody (Fig. 4D, lane 2). With anti-CP2 antibody, we identified a major complex at 43 kDa and a less intense band at ϳ39 kDa (Fig. 4D, lane 3). In repeat experiments, the 43-kDa complex from UVXL/IP migrated precisely with the 43-kDa complex upon UVXL (data not shown), confirming that this complex contains both CP1 and CP2. Taken together, these data indicate that CP1 and CP2 are the major components of the 43-kDa complex.

REMSA. As shown in
Binding Affinity of CP1 for AR mRNA-The studies above indicate that the CPs comprise a major component of the group of proteins that target the pARWt sequence in LNCaP cells. To further investigate the interaction between CP1 and AR mRNA, REMSA was performed using cleaved and purified CP1 and the pARWt probe. At a low concentration of CP1 (ϳ20 nM), binding to the probe was evident (Fig. 5A, lanes 5 and 6). At a higher concentration of CP1 (ϳ300 nM), the vast majority of the probe (ϳ78%) was incorporated into bound complex (Fig. 5, A,  arrow a; and B). Based on the relationship between bound and free fractions (Fig. 5, B and C), we estimate the K d for binding of purified CP1 to the pARWt probe to be ϳ28 nM. This compares favorably with interactions of the CPs with other mRNAs, including ␣-globin (27).
Mutational Analysis of the AR mRNA UC-rich Region-We generated several mutant probes designed to disrupt U-and C-rich sequences to determine the essential binding motifs for each of the proteins (see Fig. 1B). Using the Mut-1, Mut-6, and Mut-7 probes, we explored the importance of the U-rich sequences for HuR binding. Interestingly, HuR appears to bind to U-rich stretches that span the entire UC-rich sequence (Fig.  5D, lanes 2 and 3). The binding motif for HuR is a composite of the downstream U-rich sequences (Mut-6) (Fig. 5D, lane 2), with a more significant contribution from the upstream U-rich sequence (Mut-7) (lane 3). Mutation of the upstream U-rich sequence (Mut-1) alone had little impact on binding (data not shown). Thus, only when all of the U-rich sequences were abolished (Mut-7) was binding completely abrogated (Fig. 5D,  lane 3), suggesting that it is the combination of U-rich sequence motifs that is critical for optimal binding by HuR.
The CCCUCCC sequence contains a consensus motif for CP1 and CP2 binding (27)(28)(29). Mutations that abolished the CCC motif on either side of the central U (Mut-2 and Mut-3) (Fig.  1B) greatly reduced binding by purified CP1 (complex a, Mut-2, and Mut-3) (Fig. 5E, lanes 2 and 3). A similar result was obtained using GST-CP2 with Mut-2 and Mut-3, in which the major complex was almost obliterated (complex b) (Fig. 5E,  lanes 5 and 6, respectively). When CP1 and CP2 were added together to the probe, we found that a new complex was formed (Fig. 5E, lane 7, arrow c), with the complex for each of the individual proteins being significantly reduced (complexes b and d). Taken together, these data indicate that the CPs target the CCCUCCC motif in the UC-rich region and that loss of either the upstream or downstream CCC triplet abrogates binding. Furthermore, CP1 and CP2 may bind simultaneously to the C-rich motif sequence, as shown by the appearance of a new complex when the two proteins were added together in REMSA with the pARWt probe.
HuR and CP1 Bind to the AR UC-rich Sequence in a Cooperative Manner-HuR has three well characterized RNA-binding domains (I, II, and III). HuR contains an RNP/RNA-binding sequence in each of domains I and II, and several previous studies have verified the importance of these domains for binding AU-and U-rich sequences (24,46). However, the role for domain III is less clear. It may play a role in binding to the poly(A) tail (25). To investigate which domain(s) is responsible for binding to the U-rich sequences of AR mRNA, we generated three cleaved and purified GST-HuR proteins (HuR domain I, containing RNP-binding domain I; HuR domain IϩII, containing RNP-binding domains I and II; and HuR domain III, containing only domain III).
Purified HuR domains I and IϩII bound the pARWt sequence avidly as a single large complex (Fig. 6, lanes 2 and 3,  arrow a), consistent with the known consensus RNP-binding motifs within these proteins. However, purified HuR domain III did not bind the probe at all (Fig. 6, lane 4). Uncleaved GST-HuR also bound, but less actively (Fig. 6, lane 5). A similar pattern of binding activity was found with cleaved and purified CP1 and GST-CP1 (Fig. 6, lanes 6 and 7, respectively). When we added HuR domain IϩII and CP1 to the same REMSA tube, we found that each of the individual complexes (a and b) shifted up to form a new slower migrating complex (Fig.  6, lane 8, arrow c). The intensity of this complex was greater than that of the complexes with HuR domain IϩII or CP1 alone, and no unbound probe remained, suggesting a cooperativity of binding in the formation of the complex. Data sugges-tive of a cooperative interaction were also observed with GST-HuR and CP1 (Fig. 6, lane 9) and with GST-HuR and GST-CP1 (lane 10). In particular, CP1 was clearly shifted to a new position in the presence of GST-HuR (Fig. 6, lane 9), and GST-HuR was shifted to a new position in the presence of GST-CP1, with less probe unbound (lane 10). Taken together, these data indicate that HuR and CP1 bind to the AR UC-rich sequence simultaneously and in a cooperative manner. In addition, they confirm that HuR domain I is sufficient for binding the AR mRNA.
Binding of HuR, CP1, and CP2 to AR mRNA from LNCaP Cells-To examine the interactions of HuR, CP1, and CP2 with the AR mRNA in whole cells, we utilized an immunoprecipitation/reverse transcription-PCR assay with primers designed to bind upstream of the AR UC-rich region (see "Experimental Procedures"). Using anti-HuR, anti-CP1, and anti-CP2 antibodies, we were able to specifically co-immunoprecipitate AR mRNA from LNCaP cells (Fig. 7, lanes 1-3, respectively). However, no AR-specific PCR product was identified using the anti-GST or anti-p21 WAF1 antibody (Fig. 7, lanes 4 and 5, respectively). Several controls were routinely included in each assay: positive (assay of immunoprecipitated supernatant (e.g. Fig. 7, lane 6) and plasmid AR cDNA (e.g. lane 9)) and negative (reactions that did not contain reverse transcriptase (lane 7) and total cellular cDNA (lane 8)). These data provide definitive information that HuR, CP1, and CP2 each interact closely with AR mRNA in LNCaP cells and support our supershift data and the in vitro fusion protein studies outlined above. DISCUSSION Although the AR plays a major role in androgen action, and mRNA stability is a major contributor to the regulation of AR gene expression in hormone-dependent cancer cells (8), little is known of the molecular mechanisms governing AR mRNA decay. Here, we have identified a group of proteins (36 -110 kDa) that interact with a highly conserved novel UC-rich target in the 3Ј-UTR of AR mRNA. We have definitively identified three of this group that, to our knowledge, are the first described AR mRNA-binding proteins. HuR binds to U-rich portions of the UC-rich region, whereas CP1 and CP2 bind to C-rich portions and constitute a major component of binding activity for the AR UC-rich region as a whole. Importantly, HuR and CP1 can bind the UC-rich region at the same time. Taken together, these data suggest a model in cancer cells in which the UC-rich region of AR mRNA is a target for simultaneous binding by multiple RNA-binding proteins, including HuR, CP1, and CP2.
The UC-rich motif in the proximal 3Ј-UTR of the AR mRNA is highly conserved between species in both richness of UC content and its position relative to the stop codon. The presence of a C(U) n C (n ϭ 9) motif within an exclusively UC-rich sequence is noteworthy, as is the presence of a wholly conserved distal CCCUCCC motif representing the consensus CP-binding site (Fig. 1A). There is no classical AU-rich cis-acting mRNA stability-modifying element in this region. Significantly, the UC-rich region produced a moderate reduction in the expression of the luciferase reporter gene, consistent with a potential role as a cis-acting element. Although we cannot exclude that a component of this reduction in reporter activity is due to translational change, the design of the construct and our previous experience with this vector suggest a close association between luciferase activity and luciferase mRNA levels (39). The presence of binding proteins in REMSA with LNCaP (ARϩve) and DU145 (ARϪve) human prostate cancer cells, MDA453 (ARϩve) human breast cancer cells, and normal human testicular and surgical breast cancer tissue reflects wide tissue expression of AR UC-rich region mRNA-binding proteins. Furthermore, it suggests that these proteins may play an important role in the regulation of AR mRNA stability and/or translation in prostate and breast cancer.
HuR is a member of the Elav/Hu family of RNA-binding proteins, each of which contains three RNA recognition motifs (46,47) and whose members include Hel-N1 (HuB), HuC, and HuD (17). The Hu protein is expressed as an autoantigen in small cell lung cancer, being associated with paraneoplastic syndromes of neurological dysfunction (48 -50). Although Hu proteins typically bind AREs, Hel-N1 also binds short RNA transcripts with C(U) [3][4][5] (A/C) motifs (43). HuD also binds to a 28-nucleotide sequence in the 3Ј-UTR of c-myc containing a CUUUUUUUUUUC motif (24,25). Our mutation data indicate that HuR binds to the U-rich sequences throughout the UCrich region. When the downstream Us were mutated (Mut-6), there was some reduction in binding (Fig. 5D); however, the majority of binding was abolished when both the downstream and upstream U-rich sequences were mutated (Mut-7). This suggests that optimal binding of HuR to the UC-rich region involves interactions with at least two of these U-rich sequences.
Binding of HuR regulates the stability of multiple AREcontaining mRNAs (15)(16)(17)51). This includes vascular endothelial growth factor, in which HuR binds both UC-and AUrich sequences within a cis-element and stabilizes the mRNA (18). In colon cancer cells, HuR binds to a cis-element in the 3Ј-UTR and stabilizes p21 WAF1 mRNA (19). In addition, HuR possesses a shuttling domain and functions as a nuclear/cytoplasmic shuttling protein, enabling stabilization of mRNA within the cytoplasm following export from the nucleus (17, 20 -22). Thus, the interaction of HuR with the AR UC-rich cis-element may be important in the context of either AR mRNA stability or cellular transport, both of which may impact on AR protein expression and thus on prostate and breast cancer cell growth. Recent evidence suggests that DHT regulates the protein level and subcellular distribution of HuR in HepG2 cancer cells, kidney, and submaxillary salivary gland (52). In prostate cancer cells, DHT may exert similar effects on HuR, thereby providing an additional point of regulation for AR gene expression.
Several lines of evidence support the conclusion that CP1 and CP2 bind to the AR UC-rich region: (i) competition in REMSA by poly(C) homopolymers (28,53), (ii) abolition of the 43-kDa complex by poly(C) in UVXL assays (27), (iii) independent and additive effects of anti-CP1 and anti-CP2 antibodies in supershift assays, (iv) CP1⅐pARWt and CP2⅐pARWt complex migration at 43 kDa upon UVXL/IP, (v) binding of recombinant CP1 and CP2 fusion proteins to ARWt mRNA, and (vi) REMSA studies with mutations of the C-rich sequence. Mutation of the FIG. 7. Immunoprecipitation/reverse transcription-PCR assay using anti-HuR, anti-CP1, and anti-CP2 antibodies in LNCaP cells. LNCaP cells were immunoprecipitated with anti-HuR, anti-CP1, or anti-CP2 antibody, and the RNA was extracted and then subjected to reverse transcription-PCR using oligo(dT) and AR-specific primers as described under "Experimental Procedures." The PCR products from the beads and supernatant (S/N) were resolved on an agarose gel and photographed. Controls included no added reverse transcriptase (ϪRT; lane 7) and PCR of the pCMV-AR3.1 plasmid (AR plasmid; lane 9). AR mRNA denotes the PCR product of the expected 416 nucleotides. Primers denotes detection of excess PCR primers.
CCC triplet upstream or downstream of the middle U in the consensus sequence abrogated binding, suggesting that each is critical for maintaining binding site integrity. The CPs, which are members of the heterogeneous nuclear RNP K homology domain family of RNA-binding proteins (54), play an important role in the regulation of stability of a variety of genes. An ␣-globin RNA⅐CP complex (an ␣-complex) regulates turnover of the stable ␣-globin transcript, and disruption of CP binding by mutation of the CCUCC motif results in RNA destabilization (53). Binding of the CPs to a UCCCCU element within the UC-rich sequence of the tyrosine hydroxylase mRNA 3Ј-UTR modulates hypoxic stabilization of the transcript (Fig. 1A) (29,45). Furthermore, the proximal 3Ј-UTR that stabilizes erythropoietin mRNA (55) contains a conserved CCCUCC motif that binds CP1 and CP2 (28). Both CCCUCCC and U(U/C)CCCU motifs central for CP binding to erythropoietin and tyrosine hydroxylase transcripts, respectively, are present in the AR mRNA 3Ј-UTR UC-rich region, supporting a role for the CPs in regulating AR mRNA turnover.
The CPs also play an important role in regulating translation initiation for 15-lipoxygenase and human papilloma virus type 16 L2 mRNAs (30,31) by binding to UC-rich motifs (Fig.  1A). Recombinant CP1 represses translation of each of these mRNAs. Additionally, CP1 and CP2 bind stem-loop secondary structures formed by the 5Ј-UTR of poliovirus RNA to facilitate (rather than inhibit) translation and modulate RNA stability (56,57). Thus, the CPs have the capacity to bind different RNA sequences to repress or facilitate translation and induce divergent effects on gene expression. Our finding that the CPs interact with AR mRNA via the UC-rich region is of considerable interest, as it represents a potential alternative pathway of mRNA processing distinct from the Elav/Hu proteins.
The observation that HuR and CP1 bind AR mRNA simultaneously suggests that the UC-rich region is capable of docking with several RNA-binding proteins at once. To our knowledge, this is the first description of a short defined stretch of RNA that contains binding sites for both HuR and the CPs. We know little about the regulation of these RNA-protein interactions and the determinants of nuclear versus cytoplasmic binding for AR mRNA. Elucidating the kinetics of these interactions and the functional impact of these proteins individually and together on AR mRNA stability, translation, nuclear export, and cancer cell proliferation will be of great interest.
In summary, we have identified a highly conserved UC-rich cis-element in the 3Ј-UTR of AR mRNA that is the target for a group of RNA-binding proteins, including HuR, CP1, and CP2. These novel AR mRNA-binding proteins are widely distributed in multiple human cell lines and tissues. HuR and the CPs target different motifs within the UC-rich sequence and bind simultaneously to the region. These findings suggest a model in which binding by HuR and the CPs is cooperative in nature. Characterization of the individual and combined roles of HuR, CP1, and CP2 in the regulation of AR expression in prostate and breast cancer cells should provide valuable insight into the mechanisms underlying cell proliferation for each of these cancers.