Stepwise androgen receptor dimerization

Regulation of gene expression is a dynamic process involving many tightly orchestrated consecutive steps. Androgen-regulated gene expression is mediated by the androgen receptor (AR). The AR is a ligand-activated transcription factor and a member of the steroid receptor (SR) subfamily of nuclear receptors (NRs). Like all SRs, the AR has a modular structure composed of an N-terminal domain (NTD), a conserved DNA-binding domain (DBD) and a C-terminal ligand-binding domain (LBD) (Brinkmann et al., 1989). Activated ARs regulate genes involved in the development and maintenance of the male phenotype. AR is also a key factor in prostate cancer. AR activity is not only regulated by ligand binding and DNA binding but also by intramolecular interactions between functional domains, by homodimerization and by interactions with cofactors. The best-characterized interactions between AR functional domains are the intra- and intermolecular NTD–LBD interaction (N/C interaction) that is mediated by the FQNLF motif in the NTD and the coactivator groove in the LBD, and the intermolecular DBD–DBD interaction mediated by the dimerization box (D-box) (Centenera et al., 2008). However, the spatiotemporal relationship of the different intra- and intermolecular AR domain interactions in androgen-regulated gene expression is currently unknown.

Using fluorescence resonance energy transfer (FRET) and combined fluorescence recovery after photobleaching (FRAP) and FRET analysis, initial studies on the spatiotemporal organization of AR protein–protein interactions have been performed (Schaufele et al., 2005; van Royen et al., 2007). FRET showed that in the cytoplasm the N/C interaction is in an intramolecular conformation initiated directly after ligand-binding and before translocation to the nucleus (Schaufele et al., 2005; van Royen et al., 2007). In the nucleus, the intramolecular N/C interaction is followed by an intermolecular N/C interaction (Schaufele et al., 2005). The N/C interaction preferentially occurs in mobile ARs and is lost when the AR is bound to DNA (van Royen et al., 2007). These observations indicate that the AR itself regulates the time and place of interactions with coregulators by preventing untimely protein interactions when the AR is mobile, and allowing coregulator binding when the AR is bound to DNA (Dubbink et al., 2004; He et al., 2001; van Royen et al., 2007).

The intramolecular and intermolecular N/C interactions are mediated by binding of the FxxLF peptide motif (FQNLF) in the AR NTD to the ligand-induced cofactor binding groove in AR LBD. The phenylalanine residues in the FxxLF motif are essential for strong N/C interaction and bind deep into the coactivator groove with van der Waals interactions, whereas the leucine residue in the peptide motif lies in a shallow ridge on the surface of the LBD, and the other two amino acid residues are exposed to the solvent (Dubbink et al., 2004; Hur et al., 2004; van de Wijngaart et al., 2006). In other SRs, N/C interactions are absent or weak. In a homodimer, ARs also interact through their D-boxes in the second zinc finger of the DBD. SR D-box interactions are sustained by a network of hydrogen bonds between individual amino acid residues in the D-box and by an extensive complementary surface. In the AR DBD, a serine residue at position 597 (S597), which is absent in other SRs, forms a hydrogen bond and Van der Waals contacts with its counterpart in the opposing D-box in an AR homodimer (Shaffer et al., 2004). An additional pair of symmetrical hydrogen bonds is formed between an alanine at 596 (A596) and a threonine at 602 (T602) in the opposing AR DBD and vice versa. These interactions result in a relatively strong AR D-box dimerization interface compared with those of other SRs (Shaffer et al., 2004). The importance of the D-box in AR function is highlighted by the large number of mutations in this domain found in androgen insensitivity syndrome (AIS) patients (Centenera et al., 2008) (http://androgendb.mcgill.ca/).

To study the molecular mechanisms underlying AR homodimerization and to investigate when and where domain interactions take place, we used confocal microscopy and quantitative microscopic techniques to examine cells expressing functional, single and double YFP- and CFP-tagged wild-type ARs and appropriate AR mutants. In addition, we investigated the role that these molecular mechanisms have in differential target gene expression.

YFP and CFP fluorescent proteins were fused to the N- and C-terminus of a single AR (YFP–AR–CFP) or to separate ARs (YFP–AR and AR–CFP) to study, by FRET and FRAP, AR domain interactions and mobility in living human hepatoma Hep3B cells (Fig. 1A). Western blot analysis showed that all tagged ARs were of the expected size (Fig. 1B). All tagged ARs were able to induce expression of a luciferase (Luc) reporter gene driven by an androgen-regulated promoter, (ARE)2-TATA Luc (supplementary material Fig. S1) (van Royen et al., 2007). In the absence of hormone, double-tagged YFP–AR–CFP and single-tagged YFP–AR and AR–CFP were mainly located in the cytoplasm (Fig. 1C–E respectively, left panels). Upon hormone addition, single- and double-tagged ARs rapidly translocated to the nucleus (supplementary material Fig. S2). In the nucleus the ARs were distributed in a typical punctate pattern (Fig. 1C–E respectively, right panels). This punctate distribution pattern correlates with a transient immobilization of the AR and partially overlaps with sites of active transcription (Farla et al., 2005; van Royen et al., 2007).

We then analyzed cells expressing either double-tagged or single-tagged ARs by acceptor bleaching FRET (abFRET; Fig. 1F). In abFRET, the relative increase of the donor emission after acceptor photobleaching is a measure of the interaction between the tagged molecules under surveillance (Bastiaens and Jovin, 1996; Bastiaens et al., 1996; Karpova et al., 2003; Kenworthy, 2001; van Royen et al., 2009a). In the presence of the synthetic androgen, R1881, cells expressing double-tagged AR and cells expressing a combination of single-tagged YFP–AR and AR–CFP showed abFRET. FRET was not detected in the nucleus or cytoplasm in the absence of hormone (supplementary material Fig. S4). AbFRET was strongly reduced in double-tagged N/C-interaction-deficient mutant ARs, in which the N-terminal FQNLF motif was mutated to AQNAA (AR F23,27A/L26A), and completely lost in single-tagged N/C-interaction-deficient AR mutants (Fig. 1F). These results show that using abFRET it is possible to quantitatively study inter- and intramolecular N/C interactions in double-tagged ARs and intermolecular N/C interaction specifically between single-tagged ARs.

Confocal time-lapse microscopy showed that FRET, as measured by the YFP/CFP ratio in double-tagged AR (YFP–AR–CFP)-expressing cells increased rapidly after ligand addition, prior to translocation to the nucleus (supplementary material Fig. S2A) (Schaufele et al., 2005). By contrast, in cells expressing two different single-tagged ARs (YFP–AR and AR–CFP) the YFP/CFP ratio only increased following translocation of the tagged ARs to the nucleus (supplementary material Fig. S2B). These observations indicate that intramolecular N/C interactions were initiated rapidly after hormone binding, followed by nuclear translocation and initiation of intermolecular N/C interactions (see also Schaufele et al., 2005).

To determine the ratio of intermolecular and intramolecular interactions of nuclear ARs we co-transfected a YFP–AR–CFP expression vector with increasing amounts of a vector expressing untagged AR (Fig. 2A). In this setting, untagged AR competes with YFP–AR–CFP for intermolecular N/C interactions, because it will hetero-dimerize, resulting in a reduction of FRET. By contrast, the untagged AR does not compete for intramolecular N/C interactions, and therefore FRET of the intramolecular N/C interaction will not be reduced. Indeed, with increasing expression of untagged ARs the abFRET of YFP–AR–CFP decreased proportionally (Fig. 2B, red squares). Comparing the decrease of the experimental FRET data with theoretical FRET efficiencies, based on complete intramolecular or complete intermolecular N/C interactions (Fig. 2B, grey lines), showed that in steady state AR N/C interactions were mostly intermolecular. The curve of the experimental data dropped more than halfway to the theoretical curve of sole intermolecular FRET. The remaining clearly measurable AR fraction, estimated to be 20–40%, showed intramolecular N/C interactions.

The second domain enabling an intermolecular interaction between two ARs is the dimerization box (D-box) in the second zinc-finger of the AR DBD (Fig. 3A). Three residues in the D-box interact with their counterpart in the corresponding AR DBD in an AR dimer (A596 with T602, S597 with S597 and T602 with A596; Fig. 3B). We introduced a combination of two D-box mutations that affect AR activity in individuals with partial androgen insensitivity syndrome (AIS), A596T and S597T, in single- and double-tagged ARs to study the role of the D-box in AR dimerization and in N/C interaction. Western blot analysis showed that the tagged mutant proteins were of the expected sizes (Fig. 3C–E). Mutating both residues (S597T and A596T) in YFP and CFP single-tagged ARs, abolishing all three hydrogen bonds between the D-boxes, resulted in a complete loss of FRET efficiency, indicating absence of N/C interaction (Fig. 3F). The complete absence of N/C interaction in the double D-box mutant strongly suggested that the AR DBD–DBD interaction drives the intermolecular N/C interaction. In other words, although the N/C interaction requires binding of the FQNLF motif in the AR NTD to the coactivator groove in the LBD, the D-box interaction is essential for intermolecular N/C interaction.

Substitution of the same D-box amino acid residues in YFP–AR–CFP hardly affected the FRET efficiency (Fig. 3G). As shown above, in wild-type YFP–AR–CFP most N/C interactions are intermolecular (Fig. 2). Together, our data indicate that an expected drop in FRET in double-tagged D-box mutants, because of the loss of intermolecular N/C interactions, was compensated by intramolecular N/C interactions. In conclusion, the intramolecular N/C interaction is independent of D-box dimerization, and more importantly, D-box dimerization is an essential step in intermolecular N/C interaction, possibly because of a conformational change in the AR induced by DBD–DBD interaction.

To study the role of DNA binding in AR dimerization we introduced a mutation in the α-helix in the first zinc-finger of the DBD, which binds the major groove of the androgen response element (ARE) half-sites. The arginine residue at position 585 (R585) within this helix and directly flanking the defined P-box makes base-specific van der Waals contacts with the thymine residue in a consensus ARE (Shaffer et al., 2004). Western blot analysis of double- and single-tagged ARs in which the arginine was replaced by either a lysine (R585K) as is found in complete AIS (Sultan et al., 1993) or by alanine (R585A; which has a more subtle effect because it probably retains the tertiary structure of the DBD) showed that the expressed YFP–AR–CFP, YFP–AR and AR–CFP mutants were of the expected size (Fig. 4A–C, respectively). High resolution imaging of Hep3B cells expressing the AR mutants showed a homogeneous nuclear distribution unlike the speckled pattern found for wild-type AR (Fig. 4D) and very similar to a previously published DNA-binding-deficient AR mutant (A573D) (Farla et al., 2004). Moreover, FRAP analysis (supplementary material Fig. S5) of both AR mutants (R585K and R585A) showed a recovery of intensity that was very similar to that of AR A573D (Fig. 4E,F). Previously it was shown that AR A573D lacked the transient immobilization that is caused by binding to chromatin as shown by wild-type AR that binds to DNA (Farla et al., 2004; Farla et al., 2005). As expected from DNA-binding-deficient AR mutants, the AR mutants (R585K and R585A) were unable to induce luciferase expression from the transiently transfected reporter (ARE)2TATA–Luc (supplementary material Fig. S3) (Sultan et al., 1993).

AbFRET analysis of single YFP- and CFP-tagged DNA-binding-deficient AR mutants (R585K and R585A) showed that loss of DNA binding did not abolish the intermolecular N/C interactions although the FRET value for the mutants was somewhat lower than that of wild-type AR (Fig. 4G). Together with our findings shown in Fig. 3 and described above, this observation indicates that the majority of D-box interactions occur prior to DNA binding of AR homodimers. AbFRET analysis of double-tagged ARs showed that the DNA-binding-deficient AR mutants were not diminished in their total intra- and intermolecular N/C interactions (Fig. 4H). Therefore, to summarize, transcriptionally inactive, DNA-binding-deficient AR mutants, mutated in an amino acid residue directly involved in AR–DNA contact, are able to show both intra- and more importantly intermolecular N/C interaction, the latter involving also the D-box interaction.

High-resolution confocal images of Hep3B cells expressing wild-type and mutant ARs show a typical speckled pattern for the wild-type AR but a more homogeneous pattern is found of the AR D-box double mutant (Fig. 5A). As we previously showed using FRAP, this speckled pattern is always accompanied by a reduced mobility of the AR due to transient immobilization (Farla et al., 2005). Here we applied the strip FRAP procedure to study the mobility of the D-box mutants (Houtsmuller, 2005; van Royen et al., 2009b). There was a rapid redistribution of the D-box mutant (A596T/S597T; Fig. 5B, blue curve), similar to that of the DNA-binding-deficient mutant AR A573D (Fig. 5B, black curve), indicating that the D-box mutant is much more mobile than wild-type AR and that it lacks the relatively long transient immobilization of wild-type AR (Fig. 5B, grey curve). This finding correlates with less efficient target gene expression, but, as we have shown previously, such an association is not absolute (see Farla et al., 2004).

As a first step in exploring the role of the D-box dimerization in individual AR-regulated gene expression, we studied, using 1 and 100 nM R1881, the activity of AR mutants on transiently transfected luciferase reporter genes driven by minimal promoters containing different types of ARE. The D-box amino acid substitutions differentially affected AR transactivation capacity on different promoters in this assay (Fig. 5C,D). Both wild-type AR and the D-box mutant were able to activate a minimal promoter composed of two high-affinity AREs. The mutant AR showed a somewhat lower relative activity on a single ARE promoter (Fig. 5C, top panel for wild type and bottom panel for D-box mutant). Although the double-mutant AR (A596T/S597T) showed activity similarly to wild-type AR on a minimal promoter composed of two high-affinity consensus AREs [(ARE)2TATA-Luc; ARE sequence: 5′-TGTACAnnnTGTTCT-3′; Fig. 5C)], it was much less active on a minimal promoter driven by two probasin AREIIs (sequence: 5′-AGTACTnnnAGAACC-3′), or other weak AREs: SARG (specifically androgen regulated gene) AREs (sequence: 5′-TGTGCTnnnTGTTCT-3′) and TMPRSS2 (transmembrane protease, serine 2) AREs (sequence: 5′-AGGACAnnnCACTCT-3′; Fig. 5D, bottom panel) (Denayer et al., 2010). By contrast, wild-type AR was similarly active on all promoters and even somewhat more active on the promoter driven by two TMPRSS2 AREs (Fig. 5D, top panel). Also, complete lack of D-box interaction did not substantially affect AR activity on the MMTV LTR promoter that is composed of multiple AREs with variable AR affinity (Fig. 5E). In summary, complete loss of D-box interactions leads to lower AR activity or has no effect on activity. Importantly, the differential effects are ARE dependent.

Transcription activity of SRs is regulated not only by ligand binding and DNA binding but also by multiple protein–protein interactions including homodimerization and interactions with transcriptional coregulators (reviewed by Rosenfeld et al., 2006). A subgroup of SR coregulators interacts with a hydrophobic cleft in the LBD through LxxLL-like motifs (Dubbink et al., 2004; Hur et al., 2004). Unlike other SRs, the deep cofactor groove of AR preferentially binds bulky FxxLF motifs, enabling interactions with cofactors containing FxxLF-like motifs. The D-box in the second zinc finger of the DBD is a well-characterized dimerization interface of DNA-bound SRs (Dahlman-Wright et al., 1991). The AR contains the unique FQNLF sequence in the NTD as a second homodimerization motif that can bind to the coactivator groove in the LBD leading to AR N/C interaction. However, AR N/C interaction not only occurs intermolecularly, but also intramolecularly, so either between an FQNLF motif and a cofactor groove in the same AR molecule or between two ARs (Schaufele et al., 2005). In the present study we investigated the spatiotemporal association between D-box–D-box interaction and the N/C interaction in AR dimerization, by using quantitative live-cell imaging of cells expressing YFP and CFP single- and double-tagged wild-type and mutant ARs (Fig. 1A). Moreover, we investigated the role of AR monomers and AR dimerization in DNA binding. On the basis of our findings we propose a model for the dynamics of AR protein–protein interactions (Fig. 6). In the model, AR D-box interaction is an essential step between intra- and intermolecular AR N/C interaction.

Previously, we and others found that the intramolecular N/C interaction, but not the intermolecular N/C interaction is initiated rapidly after hormone binding before the AR translocates to the nucleus (Schaufele et al., 2005; van Royen et al., 2007). These findings are summarized in part I of the proposed model (Fig. 6). Only after nuclear translocation is the intramolecular N/C interaction followed by an intermolecular N/C interaction (supplementary material Fig. S2; Fig. 6) (Schaufele et al., 2005). In an in vivo abFRET-based competition assay where we added increasing amounts of untagged AR to YFP–AR–CFP, we showed that in steady state the majority of FRET in YFP–AR–CFP is by this intermolecular N/C interaction. This suggests that AR homodimers are the preferred conformation of AR in the nucleus. However, we calculated that a substantial percentage (20–40%) of nuclear AR shows an intramolecular N/C interaction (Fig. 2). This observation can hardly be explained as a transient intermediate population prior to AR intermolecular N/C interaction. Our AR N/C and D-box mutant analyses suggest a dynamic equilibrium between intra- and intermolecular N/C interaction in mobile AR. This raises the question of whether ARs with intramolecular N/C interaction are homodimers that are stabilized by D-box interaction. However, because stable AR D-box interaction has never been observed in AR DBDs in solution (see above) we favor the alternative explanation that the intramolecular N/C interaction in nuclear ARs represents a stable monomer subpopulation (Fig. 6, II).

Based on crystal structures of DBDs complexed with DNA, DBD–DBD dimerization through the D-box has been established as an important protein–protein interaction interface of SRs (Luisi et al., 1991; Roemer et al., 2006; Schwabe et al., 1993a; Shaffer et al., 2004). The most prominent amino acid residues involved in the AR dimerization in this complex are A596, S597 and T602 (Fig. 3B) (Shaffer et al., 2004). We showed here that mutation of two of these three amino acid residues completely abolished intermolecular N/C interaction, most probably because of complete absence of the D-box interaction. The mutations had no effect on intramolecular N/C interaction. In fact, in the absence of intermolecular N/C interaction in D-box mutants, an increased intramolecular N/C interaction was observed (Fig. 3F,G). These findings strongly suggest that D-box to D-box interactions drive the transition from intramolecular AR N/C interaction to intermolecular N/C interaction in nuclear AR (Fig. 6, III).

It is not known whether peptide motif interactions other than D-box interactions and N/C interactions can play a prominent role in AR dimerization. For AR the evidence for LBD–LBD interactions, as documented for other SRs, is limited. Although amino acid residues involved in glucocorticoid receptor (GR) LBD–LBD interactions are conserved in AR (Centenera et al., 2008), in crystallographic studies, the isolated AR LBD is present as a monomer in solution, in contrast to GR, progesterone receptor (PR) and estrogen receptor (ER) LBDs (Bledsoe et al., 2002; Matias et al., 2000; Sack et al., 2001; Tanenbaum et al., 1998; Williams and Sigler, 1998). However, a (weak) dimerization function in the hinge region as suggested for GR, or in the C-terminal extension of the AR DBD cannot be completely excluded (Centenera et al., 2008; Haelens et al., 2003; Savory et al., 2001).

It has long been disputed whether AR dimerization occurs before or after DNA binding (Centenera et al., 2008). We previously showed that the N/C interaction occurs predominantly when the ARs are mobile and is lost when the ARs are bound to chromatin (van Royen et al., 2007). Combined with the present observation that D-box interaction drives the intermolecular N/C interaction (Fig. 3) this indicates that the D-box interaction occurs before DNA binding (Fig. 6). These findings are in contrast to theories based on crystallographic studies which suggest that separate SR DBDs are monomeric in solution and show cooperative dimerization when bound to DNA (Freedman et al., 1988; Härd et al., 1990a; Härd et al., 1990b; Luisi et al., 1991; Schwabe et al., 1993a; Schwabe et al., 1993b; Shaffer et al., 2004). However, AR dimerization before DNA binding was confirmed by experiments carried out with the DNA-binding-deficient mutants (Fig. 4). Possibly, the stronger D-box to D-box interaction in AR, compared with other SRs, combined with the intramolecular N/C interaction, are of crucial importance in this regard.

The relatively strong dimerization of the AR enables activation of promoters containing different types of ARE (Fig. 5) (reviewed by Centenera et al., 2008; Claessens et al., 2008; Denayer et al., 2010; Shaffer et al., 2004). We showed that ARs without appropriate D-box interaction cannot activate promoters driven by two probasin AREIIs, SARG AREs or TMPRSS2 AREs, although a promoter with high affinity AREs can be stimulated (Fig. 5C–E) (Denayer et al., 2010). On the basis of our findings it is tempting to speculate that promoters with high affinity AREs can be activated both by AR homodimers and by consecutive binding of AR monomers that subsequently dimerize on the DNA. By contrast, promoters with low-affinity AREs would preferentially be activated by AR homodimers (Fig. 6, IV). If this is the case then AR monomers in the nucleus are of functional importance.

Recently, genome-wide chromatin immunoprecipitation (ChIP) approaches indicated the presence of thousands or even tens of thousands of AR binding sites in the human genome (Jia et al., 2008; Takayama et al., 2011; Wang et al., 2007; Wang et al., 2009; Yu et al., 2010). Interestingly, the majority of the AR binding regions found in these studies, and AR binding sites identified by ChIP in promoter and enhancer regions of androgen-regulated genes, apparently contain ARE half-site motifs, low-affinity AREs or ARE half-sites with suboptimal spacing and not obvious high-affinity AR binding motifs (Massie et al., 2007; Wang et al., 2007). Loss of AR dimerization might directly result in less AR binding to these sites (Fig. 5B). Genome-wide ChIP-seq approaches combined with global gene expression profiling in cells that exclusively express AR monomers compared with cells that contains AR mainly in the homodimer conformation would provide more detailed information on the role of monomers in AR-regulated gene expression. The AR monomer to dimer ratio in a nucleus might be a mechanism of regulation of specificity in gene expression. One obvious parameter that affects the dimerization status of the activated nuclear AR population, is its concentration. This hypothesis can well be extended to a role for specific cofactors that differentially interact with AR monomers and dimers or cofactors that regulate AR dimerization (Bai et al., 2005).

In summary, previous data on the AR intra- and intermolecular N/C interaction lead to a model in which the intramolecular N/C interaction is initiated in the cytoplasm directly after hormone binding, followed by intermolecular N/C interaction in the nucleus (Schaufele et al., 2005). Using quantitative imaging techniques, we elucidated the essential role of D-box dimerization in the transition from intramolecular to intermolecular N/C interaction (Fig. 6). The D-box dimerization and the shift from intramolecular to intermolecular N/C interaction might occur as one event or two separate events, but both independent of DNA binding. Together with our observations showing that the AR N/C interaction is lost in DNA-bound AR enabling cofactor interactions (van Royen et al., 2007), data in the present study elucidated the spatiotemporal relationship of the consecutive AR intra- and intermolecular domain interactions in living cells (Fig. 6). Moreover, the model proposes a dynamic equilibrium of AR homodimers and monomers in the nucleus, which can be an important mechanism of AR-regulated gene expression.

In all constructs expressing AR fusion proteins the AR was separated from the fluorescent tag by a flexible (Gly-Ala)₆ spacer (Farla et al., 2004) indicated by a single dash. Constructs coding for wild-type and A573D variants of YFP–AR–CFP and AR–CFP were generated as previously described (van Royen et al., 2007). The construct expressing N-terminally YFP-tagged AR was generated by replacing EGFP in pGFP-AR (Farla et al., 2004) with EYFP-C1 (Clontech Laboratories, Inc., Mountain View, CA). The construct expressing untagged AR was obtained by inserting the AR cDNA from pAR0 (Brinkmann et al., 1989) into pEGFP-C1 from which EGFP was deleted. The F23,27A/L26A mutation of YFP–AR–CFP, AR–CFP and untagged AR was introduced using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA). In YFP–AR–CFP an LBD–CFP fragment was replaced with an AR–LBD fragment from YFP–AR to obtain YFP–AR (F23,27A/L26A). The DBD mutations R585K, R585A and A596T/S597T were introduced by QuikChange mutagenesis in pYFP-AR-CFP. The LBD mutation E897A in untagged AR was also generated with QuikChange mutagenesis. For mutagenesis primers see supplementary material Table S1. To generate the single-tagged DBD mutant ARs, the AR DBDs of pYFP-AR and pAR-CFP were replaced with a pYFP-AR-CFP fragment containing the mutant DBD.

The (ARE)2-TATA Luc reporter, containing two high-affinity AREs (underlined in the following sequence: 5′-CCGGGAGCTTGTACAGGATGTTCTGCATGCTCTAGATGTACAGGATGTTCTGGTA-3′) was a gift from G. Jenster (Rotterdam, Netherlands). The other reporters were generated by swapping the ARE fragment in (ARE)2-TATA Luc with a fragment containing a single high affinity ARE as present in the (ARE)2-TATA Luc reporter (5′-CCGGGAGCTTGTACAGGATGTTCTGCATGCTCTAGAGGTA-3′), two probasin AREIIs (5′-CCGGGAGCTAGTACTGGAAGAACCGCATGCTCTAGAAGTACTGGAAGAACCGGTA-3′), two SARG AREs (5′-CCGGGAGCTTGTGCTGGATGTTCTGCATGCTCTAGATGTGCTGGATGTTCTGGTA-3′), or two TMPRSS2 AREs (5′-CCGGGAGCTAGGACAGGACACTCTGCATGCTCTAGAAGGACAGGACACTCTGGTA-3′). The MMTV-Luc reporter construct was described previously (de Ruiter et al., 1995). All new constructs were verified by sequencing. Sizes of expressed ARs were verified by western blotting.

For 2 days before microscopic analyses, Hep3B cells, lacking endogenous AR expression, were grown on glass coverslips in six-well plates in α-MEM (Cambrex, East Rutherford, NJ) supplemented with 5% fetal bovine serum (FBS; HyClone), 2 mM L-glutamine, 100 IU/ml penicillin and 100 μg/ml streptomycin. At least 4 hours before transfection, the medium was replaced with medium containing FBS stripped with 5% dextran-coated charcoal (DCC-FBS). Transfections were performed with 1 μg/well AR expression constructs or 0.5 μg/well empty YFP or CFP expression vector in FuGENE6 (Roche Molecular Biochemicals, Indianapolis, IN) transfection medium. Four hours after transfection, the medium was replaced with 5% DCC-FBS with or without 100 nM R1881. In the abFRET competition experiments 1 μg YFP–AR–CFP was co-transfected with increasing amounts of untagged AR (ratio YFP–AR–CFP: AR 1:0, 4:1, 2:1, 1:1, 1:2 and 1:4). Different vector sizes were taken into account. The amounts of CMV promoters and total transfected DNA were corrected by co-transfecting pcDNA3 (CMV) and pTZ19 vectors.

For the AR transactivation experiments, Hep3B cells were cultured in 24-well plates in α-MEM supplemented with 5% DCC-FBS in the absence or presence of R1881 (1 or 100 nM) and transfected using 50 ng AR expression construct and 100 ng luciferase reporter construct. After 24 hours, cells were lysed and luciferase activity was measured in a luminometer (GloMax Microplate luminometer; Promega Corporation, Madison, WI).

Hep3B cells were cultured and transfected in 6-well plates. After 24 hours, cells were washed twice in ice-cold PBS and lysed in 200 μl Laemmli sample buffer (50 mM Tris-HCl, pH 6.8, 10% glycerol, 2% SDS, 10 mM DTT and 0.001% Bromophenol Blue). After boiling for 5 minutes, a 5-μl sample was separated on a 10% SDS-polyacrylamide gel and blotted to Immobilon-P transfer membrane (Millipore, Billerica, MA). Blots were incubated with anti-AR (1:2000; mouse monoclonal antibody F39.4.1) and subsequently incubated with HRP-conjugated goat anti-mouse antibody (DakoCytomation, Glostrup, Denmark). Protein bands were visualized using Super Signal West Pico Luminol solution (Pierce Chemical Co., Rockford, IL), followed by exposure to x-ray film.

Immunofluorescence imaging of Hep3B cells expressing tagged ARs was performed using a confocal laser-scanning microscope (LSM510; Carl Zeiss MicroImaging, Inc., Göttingen, Germany) equipped with a Plan-Neofluar 40×/1.3 NA oil objective (Carl Zeiss MicroImaging, Inc.) at a lateral resolution of 100 nm. An argon laser was used for excitation of CFP and YFP at 458 and 514 nm, respectively. In all quantitative imaging experiments cells with a physiologically relevant expression level of tagged ARs were selected for analysis (van Royen et al., 2007; van Royen et al., 2009a).

N/C interactions of double-tagged YFP–AR–CFP, or co-transfected YFP–AR and AR–CFP were assessed using YFP/CFP ratio imaging and acceptor photobleaching FRET (abFRET) (van Royen et al., 2009a and references therein). In YFP/CFP ratio imaging cells expressing YFP and CFP double-tagged AR or a combination of YFP–AR and AR–CFP with initially similar signal ratios to YFP–AR–CFP were imaged with an interval of 30 seconds using a 458 nm excitation at low laser power to avoid monitor bleaching. YFP and CFP emissions were detected using a 560 nm longpass emission filter and a 470–500 nm bandpass emission filter, respectively. The AR N/C interaction was initiated by adding R1881 to the cell culture. After subtraction of background FRET was calculated as: I_YFP/I_CFP. The relative nuclear intensity was determined simultaneously using the YFP emission and was calculated as: I_nucleus/(I_nucleus+I_cytoplasm).

In abFRET, YFP and CFP images were collected sequentially before photobleaching of the acceptor. CFP was excited at 458 nm at moderate laser power, and emission was detected using a 470–500 nm bandpass emission filter. YFP was excited at 514 nm at moderate laser power, and emission was detected using a 560 nm longpass emission filter. After image collection, YFP in the nucleus was bleached by scanning a region of ~100 μm² 25 times at 514 nm at high laser power, covering almost the complete nucleus. After photobleaching, a second YFP and CFP image pair was collected. Apparent FRET efficiency was estimated (correcting for the amount of YFP bleached) using the equation abFRET=[(CFP_after−CFP_before)×YFP_before]×[(CFP_after×YFP_before)−(CFP_before × YFP_after)]⁻¹, where CFP_before and YFP_before are the mean prebleach fluorescence intensities of CFP and YFP, respectively, in the area to be bleached (after subtraction of background), and CFP_after and YFP_after are the mean postbleach fluorescence intensities of CFP and YFP, respectively, in the bleached area (Dinant et al., 2008). The apparent FRET efficiency was finally expressed relative to control measurements in cells expressing either free CFP and YFP (abFRET₀) or the CFP–YFP fusion protein (abFRET_{CFP–YFP fusion}): apparent FRET efficiency=(abFRET−abFRET₀)× (abFRET_{CFP-YFP fusion}−abFRET₀)⁻¹ . For statistical analysis, the abFRET data sets were compared using the one-tailed Student's t-test.

The mobility of interacting proteins was studied using FRAP (supplementary material Fig. S5) (van Royen et al., 2009b). A narrow strip spanning the nucleus was scanned at 458 nm excitation (because of simultaneous CFP recording in FRET FRAP) (van Royen et al., 2007) using short intervals (100 ms) at low laser power (YFP is sufficiently excited at this wavelength) (van Royen et al., 2007). Fluorescence intensity of YFP was recorded using a 560-nm longpass filter. After 40 scans, a high-intensity, 100-ms bleach pulse at 514 nm was applied to photobleach YFP inside the strip. Subsequently, scanning of the bleached strip was continued at 458 nm at low laser intensity. The curves were normalized using the equation I_norm=(I_raw−I₀)/(I_pre−I₀), where I_pre and I₀ are the fluorescent intensities before and immediately after the bleach, respectively.