Androgen receptor function is modulated by the tissue-specific AR45 variant

Abstract

A naturally occurring variant of the human androgen receptor (AR) named AR45 has been identified. It lacks the entire region encoded by exon 1 of the AR gene and is composed of the AR DNA-binding domain, hinge region and ligand-binding domain, preceded by a novel seven amino-acid long N-terminal extension. A survey of human tissues revealed that AR45 was expressed mainly in heart and skeletal muscle. In cotransfection experiments, AR45 inhibited AR function, an effect necessitating intact DNA- and ligand-binding properties. Overexpression of AR45 reduced the proliferation rate of the androgen-dependent LNCaP cells, in line with the repressive effects of AR45 on AR growth-promoting function. AR45 interacted with the AR N-terminal domain in two-hybrid assays, suggesting that AR inhibition was due to the formation of AR–AR45 heterodimers. Under conditions where the transcriptional coactivator TIF2 or the oncogene β-catenin were overexpressed, AR45 stimulated androgen-dependent promoters in presence of dihydrotestosterone. AR45 activity was triggered additionally by the adrenal androgen androstenedione in presence of exogenous TIF2. Altogether, the data suggest an important role of AR45 in modulating AR function and add a novel level of complexity to the mode of action of androgens.

Abbreviations

AR

androgen receptor

DAPI

4,6-diamidino-2-phenylindole

DBD

DNA-binding domain

DHT

dihydrotestosterone

ECL

electrochemiluminescence

ER

estrogen receptor

FBS

fetal bovine serum

GFP

green fluorescent protein

LBD

ligand-binding domain

MEM

minimal essential medium

NTD

N-terminal domain

PSA

prostate-specific antigen

The surprising finding that the human genome only codes for 20 000–25 000 genes suggests that several processes contribute to additional complexity levels. One such mechanism is the generation of multiple RNA forms from a single gene through use of different promoters or alternative splicing [1–3]. The corresponding protein variants may exert very different and sometimes opposite biological functions [1–3]. Abnormal splicing is responsible for about 50% of genetic disorders [4], including some forms of Parkinson's Disease, of cystic fibrosis, of polycystic kidney disease and progeria [5–8]. Changes in splicing patterns have also been linked to cancer [9]; for example a role of the insulin receptor, cyclin D1 and Mdm2 isotypes in tumour progression has been documented [10–12]. Consequently, therapeutic modalities aiming at correcting abnormal splicing events have already been envisaged [13].

Steroid receptors belong to a unique superfamily of ligand-activated transcription factors that have very pleiotropic effects [14]. They are organized in a modular fashion and act by binding to DNA response elements found in the regulatory regions of their target genes [14]. Due to the involvement of these receptors in several major diseases, extensive research has been carried out to identify agonistic and antagonistic ligands with beneficial effects. An unresolved issue concerns the tissue- and cell-specific activities observed with selective receptor modulators [15–17]. Different panels of cofactors may account for some of these effects [18]. Another possibility arises from the existence of variant forms of steroid receptors. The estrogen receptor (ER) probably belongs to the most diversified family. Two isoforms, ERα and ERβ, exist, and have different tissue distribution [15]. In addition, many ER splice variants have been described, mainly in tumours [19]. The biological function of ER variants has only been studied in a few cases. Examples include ERα46, an ERα form lacking the region encoded by the first exon, which modulates the activity of ERα in MCF7 cells [20], and ERβcx, a C-terminally truncated form containing 26 specific amino acids, which forms heterodimers with ERα to inhibit its function [21]. The progesterone receptor exists as two isoforms, PRA and PRB, differing by the length of the N-terminal end and originating from translation reinitiation at an internal methionine codon. Studies with transgenic mice show that the ratio of both forms is essential for proper development of the mammary gland [22]. Analysis of human endometrial tumours indicates that loss of the B-form is linked to poor prognosis [23]. Concerning the glucocorticoid receptor, a splice variant named GRβ that lacks the region encoded by the most 3′ exon and is unable to bind to glucocorticoids has been described. Overexpression of GRβ leads to glucocorticoid resistance in a number of pathological conditions [24]. In the case of the mineralocorticoid receptor a variant lacking the hinge and ligand-binding regions but able to increase the activity of the full-length receptor has been described [25].

Several variant forms of the androgen receptor (AR) have been found. A short AR-A form arising, as for PRA from internal translational reinitiation, has been described [26]. Recent data, however, question its existence in human tissues [27]. A few cases of exon skipping leading to androgen insensitivity have been reported [28]. Finally, polymorphisms affecting the length of glutamine, proline and glycine repeats exist [28]. They may lead to severe pathologies, as observed in Kennedy's disease, a progressive motor neuron degeneration caused by an extended polyglutamine repeat in the N-terminal domain of the AR.

Here we describe the identification and characterization of AR45, a human AR variant composed of a unique N-terminal extension linked to the DNA-binding domain (DBD), hinge region and ligand-binding domain (LBD) of the AR. The restricted expression pattern, the inhibitory effect on AR function and the cofactor-dependent activity suggest an important biological role of AR45 in modulating androgen action.

Results

Identification of AR45, a novel variant of the human AR

5′ Rapid amplification of cDNA ends (RACE) was performed on human placental RNA using a reverse primer directed against the hinge domain of the AR and a forward primer recognizing the common 5′ end of the reverse-transcribed RNA. This allowed the amplification of a 500 bp-long DNA fragment, much shorter than the expected AR product. DNA sequencing revealed that this fragment coded for the AR DBD and hinge region preceded by a novel, short N-terminal extension, rather than the 538 amino acid-long N-terminal moiety encoded by exon 1 of the AR gene (Fig. 1). A primer specific for the extreme 5′ region was then used together with a primer recognizing the 3′ end of the AR coding part to amplify the complete coding region of this novel AR form. The deduced amino acid sequence (Fig. 1A) revealed that it contained an intact DBD, hinge region and LBD. However, it lacked the entire region encoded by exon 1 of the AR which was replaced by a short, unique seven amino acid-long N-terminal extension. The deduced molecular mass was about 45 kDa, hence the name AR45.

Figure 1

Sequence of AR45 and position of exon 1B in the AR gene. (A) Deduced amino acid sequence of AR45. The unique N-terminal extension is in bold. The exon limits are indicated with converging arrowheads and the domain boundaries with colons. The residues mutated for functional studies are underlined. (B) DNA and deduced amino acid sequences of exon 1B. The open reading frame is shown in bold capital letters. The flanking intron regions are in italics. (C) Localization of exon 1B in the AR gene. Exon numbers are shown above the representation, intron sizes are given below in kb. These sequence data have been submitted to the DDBJ/EMBL/GenBank databases under accession number AX453758.

Homology search revealed that the AR45-specific coding region was located on chromosome Xq11, between exons 1 and 2 of the AR gene. The complete DNA sequence of this novel exon, which we named 1B, together with flanking intron regions is depicted in Fig. 1B. Exon 1B localized ≈ 22.1 kb downstream of AR exon 1 (Fig. 1C) and was not predicted by annotation programs of the human genome. Its extremities nevertheless conformed to the splicing rules.

AR45 is mainly expressed in heart

RT-PCR was carried out on a panel of human tissues to determine the expression pattern of AR45 mRNA. A primer corresponding to the AR45-specific upstream region was used together with a primer recognizing the AR common part. The strongest signal was observed in heart, followed by skeletal muscle, uterus, prostate, breast and lung. Weaker signals were seen in other tissues, including testis (Fig. 2A). In comparison, very low levels of the AR transcript were detected in heart, as compared to liver or testis (Fig. 2B). The transcript levels of the ribosomal S9 protein were determined as a control (Fig. 2C). Initial studies with an antibody directed against the AR LBD indicated that a band of about 45 kDa was present in human heart extracts (not shown). However, in the absence of an antibody recognizing the specific N-terminal extension of AR45, it cannot be excluded that this band corresponds to a degradation product of the AR.

Figure 2

Analysis of tissue distribution of AR45 by RT-PCR analysis of human RNA. First-strand cDNA from the indicated organs was used as template for PCR amplification. Transcript levels of (A) AR45, (B) AR and (C) ribosomal protein S9 (S9).

AR45 is bound by androgen and localizes to the cell nucleus

The hormone-binding properties of AR45 were assessed by competitive binding experiments in presence of ³H-labeled R1881 and increasing amounts of cold R1881 (Fig. 3A) CV-1 cells were transfected with a pSG5-based expression vector for AR45 or AR. The results showed specific R1881 binding to AR45 with a calculated IC₅₀ of 22 nm. This was very close to the IC₅₀ found for AR, which was 16 nm. When correcting for the amount of AR45 or AR protein expressed in the cells, as determined by Western blot analysis and quantification of electrochemiluminescence (ECL) signals, comparable levels of total [³H]R1881 bound to both forms were found (not shown). The R243H mutant form of AR45 was used as negative control. It corresponds to AR R774H, a mutant unable to bind to androgens [29,30]. As expected, no specific binding was measured (not shown).

Figure 3

AR45 binds to androgen and localizes to the cell nucleus. (A) CV-1 cells were seeded in six-well plates and transfected with 6 µg of expression plasmids for AR (•) or AR45 (). For the binding test, the indicated concentrations of unlabeled R1881 were mixed with the tracer. The radioactivity of whole cell extracts was determined. The results are a representative of two separate experiments. (B) Subcellular localization of AR45. Left panel: PC-3 cells were seeded in 24-well plates and transfected with 2.5 µg of an AR45–GFP expression construct and treated with R1881. Fluorescence microscopy was used to visualize the chimeric AR45 protein. Right panel: nuclear staining with DAPI.

The subcellular localization of AR45 was determined by transfecting an expression vector coding for AR45 with an N-terminally fused green fluorescent protein (GFP) moiety into PC-3 cells. Following R1881 treatment, we observed an exclusively nuclear localization of AR45 (Fig. 3B, left panel). 4,6-Diamidino-2-phenylindole (DAPI) staining was performed to visualize the cell nuclei (Fig. 3B, right panel). Transfection of a control GFP plasmid gave fluorescent signals in both the cytoplasm and nucleus (not shown).

AR45 inhibits AR transcriptional activity

Initial cell-based transactivation studies with different cell lines and with constructs containing various androgen-responsive promoters showed that AR45 did not stimulate reporter gene activity in the presence or absence of ligand (not shown). We therefore sought to determine whether AR45 might be an inhibitor of AR function. Transient transfection studies were performed in CV-1 cells using AR45- and AR-expressing plasmids, in the presence of the MMTV-Luc reporter plasmid (Fig. 4A). Using increasing amounts of AR45 plasmid for transfection, a concentration-dependent inhibition of androgen-stimulated AR activity was observed. To find out whether androgen- and DNA-binding were important for this effect, we devised several mutant forms of AR45. AR45 R243H corresponds to AR R774H, a mutant form not bound by androgens [29,30]. Cotransfection experiments revealed that the AR45 R243H mutant did not inhibit AR function. AR45 C31G and ΔR84 correspond to AR C562G and ΔR615, two mutants that no longer bind to DNA [29,30]. These AR45 mutants also did not repress AR activity in transactivation assays. Suprisingly, a stimulatory effect was elicited by all three AR45 mutants at the highest plasmid concentration used. To show that repression was not limited to a single cell line, we performed similar cotransfection experiments in the prostate cancer cell line PC-3 and obtained comparable results (not shown).

Figure 4

AR45 inhibits AR function. (A) CV-1 cells were transfected with a reporter vector harbouring the MMTV promoter (40 ng), an expression vector for AR (10 ng) and with the indicated amounts (ng) of expression plasmids for AR45 or a mutated form. Treatment was with 1 nm R1881 (grey bars) or with vehicle (white bars). The results are a representative of three separate experiments and the bars are the mean ± SEM of sextuplicate values. (B) PC-3 cells were transfected with a reporter vector harbouring the PSA promoter (40 ng), an expression vector for AR (10 ng) and with the indicated amounts (ng) of expression plasmids for AR45 or a mutated form. Treatment was with 1 nm R1881 (grey bars) or with vehicle (white bars). The results are a representative of three separate experiments and the bars are the mean ± SEM of sextuplicate values. Position of the mutations was indicated relative to the AR45 amino sequence. For clarity, the corresponding location in the AR was also given in parentheses.

The results were further confirmed with the prostate-specific antigen (PSA) promoter (Fig. 4B). Here we also observed an inhibitory effect of AR45 on AR activity following overexpression in PC-3 cells. As before, this was not seen with the mutant AR45 C31G, which does not bind to DNA.

The results show that AR45 acts as an inhibitor of AR function. This may result from competition of AR45 homodimers or of AR45–AR heterodimers with AR homodimers for binding to DNA response elements.

AR45 inhibits proliferation of LNCaP cells

We next determined the biological effects of AR inhibition by AR45 in LNCaP cells. First, the endogenous expression levels of AR45 mRNA were assessed in LNCaP cells by RT-PCR, using similar conditions as above. Specific AR45 transcripts were detected in LNCaP cells grown in the presence or absence of R1881 (Fig. 5A). When analyzing LNCaP nuclear extracts with an antibody directed against the AR LBD, several protein bands migrating ahead of full-length AR were seen, including one of a size compatible with that of AR45 (Fig. 5B, lane 4). Indeed, a protein band of similar size was also seen in PC-3 cells transfected with an AR45-expressing plasmid (lane 2), but not with an AR-expressing plasmid (lane 3). Also, AR45 produced by an in vitro translation procedure migrated at the same level (lane 1). Even though the identity of the 45 kDa protein detected in LNCaP cells was not demonstrated due to the unavailability of a specific antibody, the results showed that AR45 expression was at best low.

Figure 5

AR45 inhibits LNCaP cell proliferation. (A) Total RNA was purified from LNCaP cells grown in the presence (lane 1) or absence (lane 2) of 0.1 nm R1881 and reverse-transcribed. Primers specific for AR45 or AR were used for PCR amplification. The levels of actin were determined as control. (B) Nuclear extracts were prepared from LNCaP cells grown in presence of 1 nm R1881 and analyzed by Western blot using an antibody directed against the LBD (lane 4). Nuclear extracts from PC-3 cells transfected with an expression vector for AR45 or for AR were analysed in parallel (lanes 2 and 3). In vitro translated AR45 was loaded on lane 1. Molecular mass markers are indicated in kDa. (C) LNCaP cells grown in presence of the indicated R1881 concentrations were seeded in 10-cm Petri dishes and transfected with 20 µg of an AR45-expressing vector. The number of viable cells was measured after 3 days by quantifying the ATP levels (grey bars). In the control experiment empty pSG5 plasmid was used (white bars). The results are a representative of two separate experiments and the bars are the mean ± SEM of quadruplicate values.

Proliferation tests were carried out with LNCaP cells grown in the presence of 0.1 or 1 nm R1881. Following transfection of an AR45-expressing plasmid, a significant reduction in the number of viable cells was measured after three days at both hormone concentrations, indicating that inhibition of cell proliferation had taken place (Fig. 5C).

AR45 interacts with the N-terminal region of the AR

Having determined that AR45 inhibited AR activity, we sought to find out whether this was due to a direct interaction by performing a mammalian two-hybrid assay. Expression vectors coding for full-length AR45 and for a fusion protein between various domains of the AR N-terminal domain (NTD) and the activation domain of NF-κB were cotransfected into CV-1 cells. Using an MMTV reporter construct we observed a strong, androgen-dependent interaction between AR45 and the AR NTD (Fig. 6A). No interaction was observed between AR45 and the DBD, hinge region or LBD of the AR (not shown). Additional experiments were performed with a reporter construct harbouring the androgen-dependent Pem promoter [31]. The Pem gene codes for a homeobox protein of the paired-like family involved in male reproductive functions [32] and its promoter has recenly been shown to contain highly selective androgen-responsive elements [31,33]. By performing similar two-hybrid assays, we also observed an androgen–dependent interaction between AR45 and the AR NTD (Fig. 6B) but not with other regions of the AR (not shown).

Figure 6

AR45 interacts with the AR NTD. CV-1 cells were seeded in 96-well plates and transfected with expression vectors for AR45 and for a fusion between the AR NTD and the activation domain of NF-κB (40 ng each). Treatment was with 1 nm R1881. (A) Reporter plasmid (80 ng) containing the MMTV promoter was cotransfected. (B) Reporter plasmid containing the Pem promoter (80 ng) was cotransfected. The results are a representative of three separate experiments and the bars are the mean ± SEM of sextuplicate values.

These results indicate that AR45 binds to androgen-responsive promoters and directly interacts with the AR NTD.

AR45 stimulates androgen-dependent promoters in presence of cofactors and adrenal androgen

The absence of stimulatory activity of AR45 might be linked to poor coactivator recruitment, due to the absence of the NTD. This might be overcome by the overexpression of cofactors. To test this hypothesis, we analyzed the effects of two AR cofactors known to interact with the AR LBD. TIF2 and AR45 were overexpressed in CV-1 cells (Fig. 7A). Subsequent dihydrotestosterone (DHT) treatment led to a sevenfold increase of MMTV-driven reporter activity. When coexpressing the oncoprotein, β-catenin, a fourfold stimulation was seen (Fig. 7A). A similar induction was also seen when using the S33F β-catenin mutant, a form with enhanced stability previously identified in several tumours, including prostate carcinoma (Fig. 7A). There was no significant effect of AR45 and cofactor overexpression on reporter gene activity in the absence of DHT (not shown). To rule out that the observed AR45 activity was cell- or promoter-specific, we coexpressed AR45 and TIF2 in PC-3 cells, and used the PSA promoter as well as the MMTV promoter in the reporter plasmids. In these cells, TIF2 overexpression enhanced DHT-dependent AR45 activity sixfold on the MMTV promoter and twofold on the PSA promoter (Fig. 7B,C). In the absence of DHT, no significant difference was noted for MMTV or PSA promoter activity following AR45 and TIF2 overexpression (not shown). Finally we tested the effects of androstenedione on AR45 function. This hormone is synthesized primarily by the adrenal glands and is not suppressed by chemical castration used to treat prostate cancer patients, which led to speculations about a role in refractory tumours [27]. TIF2 and AR45 were overexpressed in PC-3 cells. Following androstenedione treatment, we measured a twofold stimulation of the MMTV promoter (Fig. 7D).

Figure 7

Activation of AR45 by cofactors and androstenedione. Transfections were carried out in 96-well plates in presence of 30 ng of AR45 plasmid, 120 ng of cofactor plasmid and 40 ng of reporter plasmid. (A) CV-1 cells were transfected with expression vectors for AR45 and for the indicated cofactors. A reporter construct containing the MMTV promoter was cotransfected. Treatment was with 1 nm DHT (grey bars) or with vehicle (white bar). (B) PC-3 cells were transfected with expression vectors for AR45 and TIF2. A reporter construct containing the MMTV promoter was cotransfected. Treatment was with 1 nm DHT (grey bars) or with vehicle (white bar). (C) PC-3 cells were transfected with expression vectors for AR45 and TIF2. A reporter construct containing the PSA promoter was cotransfected. Treatment was with 1 nm DHT (grey bars) or with vehicle (white bar). (D) PC-3 cells were transfected with expression vectors for AR45 and TIF2. A reporter construct containing the MMTV promoter was cotransfected. Treatment was with 1 nm androstenedione (grey bars) or with vehicle (white bar). The results are a representative of two separate experiments and the bars are the mean ± SEM of sextuplicate values.

Altogether these data document that under conditions where coactivators are overexpressed, a stimulation of AR45 by androgens such as DHT or androstenedione may be observed.

Discussion

Here we report the cloning and characterization of human AR45 cDNA, which codes for a variant form of the AR. AR45 lacks the entire region encoded by exon 1 of the AR gene and possesses instead a unique seven amino acid-long stretch. This AR45-specific region is entirely encoded by a hitherto undescribed exon 1B lying between the first and the second exon of the AR gene. AR45 may therefore originate from transcription controlled by a novel promoter region lying upstream of exon 1B. Interestingly, the consensus binding sequence CAAGTG for the heart-specific transcription factor Nkx2.5 [34] and motifs with 90% identity to the binding site GGGRNNYCCC for p65, a subunit of the NF-κB transcription factor that regulates gene expression in heart [35], are found in the region immediately upstream of exon 1B (not shown). Detailed studies are now needed to determine whether this region directs heart-selective expression of AR45. An alternative splicing event is less likely, because no sequence corresponding to an additional, more upstream exon was found in our 5′ RACE-PCR experiments. Nonetheless this cannot entirely be ruled out and tissue-specific splicing mechanisms have already been described for other steroid receptors [36,37].

The role of androgens in the heart has been documented by many studies. Hypertension and myocardial ischemia are associated with elevated androgen levels [38]. Direct modulatory effects of androgens and estrogens on the left-ventricular mass have been postulated [39]. Due to the comparatively low levels of AR in heart, AR45 may play an important regulatory role, as an inhibitor of AR function or possibly also as an activator in its own right, depending on the promoter context and availability of cofactors. AR45 may act as a dominant-negative inhibitor of AR function. As intact ligand- and DNA-binding regions are mandatory for this, the formation of AR45 homodimers and of AR45–AR heterodimers on DNA response elements may both account for the effect. The formation of AR45–AR heterodimers is likely as previous experiments with a truncated rat AR form lacking most of the NTD have been shown to interact with full-length AR in electrophoretic mobility shift assays [40,41]. If such heterodimers cannot recruit the full coactivator set needed for activity, dominant-negative effects may be observed. The important role of the NTD, which is absent in AR45, has been documented by several studies [40–42]. The hormone-dependent interaction between this region and the LBD of the AR is essential for activity. Its disruption by introducing appropriate mutations in the NTD or by overexpressing an N-terminal AR peptide leads to the impairment of AR function [43,44]. The amino- to carboxy-terminus interaction probably forms a platform for recruitment of several important cofactors [45–47]. Chaperones (e.g. Bag1L), cofactors (e.g. ARA160 and ART27) and signaling effectors (e.g. Akt) have been reported to bind to the AR NTD [48]. It is therefore likely that AR45 functions mainly as a repressor of AR function.

Our cellular assays show, however, that in an environment where a cofactor such as TIF2 or an oncoprotein such as β-catenin is overexpressed, AR45 may stimulate the expression of androgen-dependent promoters following DHT and also adrenal androgen binding. This might have implications in the progression of prostate carcinoma, a disease in which androgens and the AR play the main role [49–51]. Enhanced TIF2 protein levels have been found in the majority of recurrent prostate cancers [52]. Elevated transcript levels of β-catenin, including mutant forms coding for proteins with enhanced stability, have also been found in prostate tumours [53]. In addition, a nuclear colocalization with the AR and a strictly ligand-dependent interaction have been reported for β-catenin [54]. Enhanced TIF2 or β-catenin expression may therefore lead to aberrant AR45 activity and the stimulation of prostate tumour cell proliferation. Our additional finding that androstenedione, a low-affinity adrenal androgen, may stimulate AR45 when cofactors are overexpressed, suggests a mechanism by which treatment-refractory prostate tumours may bypass testosterone deprivation. In the light of a recent study demonstrating that the AR is a major player in both early and late androgen-independent stages of prostate cancer [50], a survey of AR45 levels in tumour resections (surgical removal of tissue) should help clarify the role of this variant form in disease progression.

In conclusion, we have identified AR45, a naturally occurring variant of the AR. It may either repress or stimulate AR activity, depending on the respective levels of each protein and of cofactors such as TIF2 and β-catenin. This study raises the possibility of a role of AR45 in modulating androgen effects in heart, where this form is most abundantly expressed, or in prostate tumours, where aberrant expression of cofactors may modify its activity.

Experimental procedures

Chemicals and plasmids

Methyltrienolone (R1881), dihydrotestosterone (DHT) and androstenedione (4-androstene-3,17-dione) were from Dupont NEN (Boston, MA, USA). RPMI 1640, minimal essential medium (MEM), OPTI-MEM, streptomycin, penicillin, geneticin and l-glutamine were obtained from Gibco BRL Life Technologies (Eggenstein, Germany). Fetal bovine serum (FBS) was from PAA (Pasching, Austria). FuGene 6 was from Roche Molecular Biochemicals (Mannheim, Germany). The oligonucleotides were purchased from MWG Biotech (Ebersberg, Germany) or Roth (Karlsruhe, Germany). Plasmids were from Stratagene (pSG5, pCMV-BD; Amsterdam, the Netherlands), Invitrogen (pCR-TOPO II; Karlsruhe, Germany) or Promega (pGL3-Basic, pGL3-Promoter; Mannheim, Germany).

Cloning procedures and site-directed mutagenesis

Human placental RNA (1 µg; Stratagene) was reverse-transcribed using the 5′-SMART RACE kit (BD Bioscience Clontech; Heidelberg, Germany) and used as the template for PCR amplification. This was performed with the Advantage-2 PCR kit (BD Bioscience Clontech) using a reverse primer directed against the AR hinge region (5′-CAGATTACCAAGCTTCAGCTTCCG-3′) and the forward 5′-Smart II primer that binds to the common end of the SMART cDNA population. Reaction conditions were: five cycles of 5 s at 94 °C, 3 min at 72 °C; five cycles of 5 s at 94 °C, 10 s at 70 °C, 3 min at 72 °C; 27 cycles of 5 s at 94 °C, 10 s at 68 °C, 3 min at 72 °C. Following agarose gel separation, a 500 bp-long DNA fragment was generated, as visualized after gel electrophoresis. It was cloned into the pCR-TOPO II plasmid and sequenced. The corresponding full-length cDNA was amplified from placental DNA using a forward primer derived from the specific 5′ extremity and a reverse primer recognizing the 3′ end of the AR. This allowed the amplification of 1200 bp-long fragment which was purified on a gel, cloned and sequenced.

Cloning of TIF2 and β-catenin coding sequences was carried out using primers flanking the coding region and human universal cDNA (QUICK-Clone II, BD Bioscience Clontech) as the template. PCR amplification was performed with the Herculase enhanced DNA polymerase (Stratagene). Following purification on an agarose gel, the amplified DNA was cloned into pCR-TOPO II (Invitrogen) and sequenced.

Site-directed mutagenesis of AR45 to generate the C31G, ΔR84 and R243H forms, and of β-catenin to generate the S33F form was performed with the QuickChange kit (Stratagene) using appropriate mutating oligonucleotides, following the manufacturer's instructions.

RT-PCR analysis of AR45 mRNA expression

AR45 RNA levels were determined by semiquantitative RT-PCR. Total RNA from different human organs was obtained from BD Bioscience Clontech. LNCaP total RNA was purified using the RNeasy kit (Qiagen, Hilden, Germany). Reverse-transcription was carried out using the First-Strand cDNA kit (Stratagene) and semiquantitative PCR analysis performed with the Advantage-2 kit (Stratagene). The following primers were used for AR45: 5′-ACAGGGAACCAGGGAAACGAATGCAGAGTGCTCCTGACATTGCCTGT-3′ and 5′-TCACTGGGTGTGGAAATAGATGGGCTTGA-3′. Reaction conditions were: five cycles of 5 s at 94 °C, 3 min at 72 °C; five cycles of 5 s at 94 °C, 10 s at 70 °C, 3 min at 72 °C; 23 cycles of 5 s at 94 °C, 10 s at 68 °C, 3 min at 72 °C. The following primers were used for AR, 5′-GGGTGAGGATGGTTCTCCCC-3′ and 5′-CTGGACTCAGATGCTCC-3′. The reaction conditions were as above, except that the annealing temperatures were 54 °C for five cycles, 52 °C for five cycles and 50 °C for 27 cycles. The amplification products were separated on a 1% (w/v) agarose gel and stained with 0.5 µg·mL⁻¹ ethidium bromide.

Western blot analysis

Following separation on precast 4–12% Bis/Tris gels (NuPAGE Novex, Invitrogen), the proteins were transferred onto poly(vinylidene difluoride) membranes (Roche Molecular Biochemicals). Incubation with the primary antibody AR C-19 (sc-815; Santa Cruz Biotechnology, Santa Cruz, CA, USA) was for 12 h at 4 °C and at a 1 : 2000 dilution. For the secondary, peroxidase-labeled anti-rabbit whole IgG (A-0545, Sigma-Aldrich, Munich, Germany), incubation was for 1 h at room temperature and at a 1 : 5000 dilution. Detection was performed using the electrochemiluminescence (ECL) kit and ECL hyperfilms (Amersham Pharmacia Biotech, Freiburg, Germany).

In vitro translation

AR45 cDNA cloned into the pCRII-TOPO plasmid (1 µg) was in vitro-translated using the TNT T7/SP6 Coupled Reticulocyte Lysate system and the SP6 RNA polymerase, following the manufacturer's instructions (Promega). The reaction products were stored at −70 °C in small aliquots.

Cell culture and transfections

Cell lines were obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany), except the PC-3/AR cells, a stable, AR-expressing cell line derived from PC-3 cells. CV-1 cells were grown at 37 °C in a 5% (v/v) CO₂ atmosphere in MEM, 10% (v/v) FBS, 100 U·mL⁻¹ penicillin, 100 µg·mL⁻¹ streptomycin, 4 mm l-glutamine. PC-3 cells were grown at 37 °C in 5% (v/v) CO₂ in F-12K medium, 5% (v/v) FBS, 100 U·mL⁻¹ penicillin, 100 µg·mL⁻¹ streptomycin, 2 mm l-glutamine. PC-3/AR cells were grown at 37 °C in 4.5% (v/v) CO₂ in RPMI 1640, 10% (v/v) FBS, 100 U·mL⁻¹ penicillin, 100 µg·mL⁻¹ streptomycin, 4 mm l-glutamine, 600 µg·mL⁻¹ geneticin. LNCaP cells were grown at 37 °C in 5% (v/v) CO₂ in RPMI 1640, 5% (v/v) FBS, 100 U·mL⁻¹ penicillin, 100 µg·mL⁻¹ streptomycin, 4 mm l-glutamine, 0.1 nm R1881.

For the transactivation assays, the CV-1 and PC-3 cells were seeded in 96-well plates at a concentration of 10 000–15 000 cells per 100 µL per well in medium supplemented as above except that 5% (v/v) charcoal-stripped FBS was used. Transfections were carried out 18–19 h later using FuGene 6 in OPTI-MEM and 40 ng of reporter plasmid based on pGL3-Basic. For LNCaP cells, seeding was in six-well plates (5 × 10⁵ cells·well⁻¹) for 22 h. Transfection was with X-treme GENE (Roche Molecular Biochemicals) using 10 µg of reporter plasmid. Expression plasmids for AR45, AR or cofactors were cotransfected as indicated. The amount of transfected plasmids was kept constant by adding the appropriate concentrations of pSG5 containing a neutral insert. Induction was performed 5 h later by adding 1 nm (CV-1 and PC-3 cells) or 2 nm R1881 (LNCaP cells). Alternatively 1 nm androstenedione was used. Measurement of luciferase activity was carried out after 18 h in a Lumicount luminometer, following the addition of 100 µL of LucLite Plus reagent (both Packard, Dreieich, Germany). The activity of a constitutively active luciferase vector was determined in parallel to assess transfection efficiency. For all points the average value of six wells treated in parallel was taken. The experiments were repeated independently at least three times.

For the androgen binding test, CV-1 cells were seeded in six-well plates (2 × 10⁵ cells·well⁻¹) and transfected 24 h later with pSG5-AR45 or pSG5-AR (6 µg·well⁻¹). A mutant corresponding to the nonandrogen-binding form R774H was used in the control experiment. The level of AR45 or AR was determined by Western blot analysis as above. Quantification of the ECL signal was performed in a Kodak Image Station (Rochester, NY, USA).

For determination of the nuclear localization, PC-3 cells (1.25 × 10⁵ cells in 0.5 mL) were seeded in 24-well plates on glass coverslips, transfected with 2.5 µg of a GFP-AR45 expression construct and treated 5 h later with 1 nm R1881. After 24 h, the cells were fixed with 3.7% (v/v) formaldehyde and examined by fluorescence microscopy at 507 nm. Alternatively, staining with DAPI (Calbiochem, Bad Soden, Germany) was performed to visualize the nuclei.

For the proliferation tests, 3 × 10⁵ LNCaP cells were seeded into 10-cm Petri dishes. Transfection was with 20 µg of pSG5-AR45 plasmid in Lipofectamine 2000 (Invitrogen). The number of viable cells was determined after 3 days in a Lumicount luminometer (Packard) using the Cell Titer-Glo assay (Promega).

The two-hybrid assay was performed in CV-1 cells in the 96-well format (10⁴ cells·well⁻¹). Transfection was for 4 h with FuGene reagent using 80 ng of reporter plasmid, 40 ng of pSG5-AR45 and 40 ng of a pCMV-BD-based plasmid containing different domains of the human AR. Treatment was with 1 nm R1881 and luciferase activity was determined after 23 h.

Androgen-binding assay

For the binding test, 200 µL of tracer (0.86 µCi of ³H-labeled R1881 per well, 83.5 Ci·mmol⁻¹; NEN) mixed with 200 µL of different concentrations of cold R1881 were added to transfected CV-1 cells for 2 h at 37 °C. The cells were washed, lysed with low-salt buffer [2% (w/v) SDS, 10% (v/v) glycerol, 10 mm Tris/HCl pH 8] for 5 min, transferred into scintillation tubes, supplemented with 3.5 mL of Atomlight scintillator liquid (PerkinElmer, Rodgau-Jügesheim, Germany) and incubated for 18 h in the dark. Scintillation counting was performed in a 1500 Liquid Scintillation Analyzer (Packard).

Acknowledgements

We are indebted to Prof G. Stock and Dr U.-F. Habenicht for continuous interest in this project. We thank Prof W.-D. Schleuning and Dr K. Bosslet for support, and Dr D. Zopf, Dr K. Barbulescu and Dr D. Mumberg for fruitful discussions. The expert technical assistance of F. Knoth, I. Schüttke and M. Wostrack was much appreciated.

The PC-3/AR cell line and the human AR cDNA were obtained from Prof A. Cato (FZ Karlsruhe, Germany), the PSA promoter construct from Prof J. Trapman (Erasmus Medical Center, Rotterdam, the Netherlands).