A multi-parameter imaging assay identifies different stages of ligand-induced androgen receptor activation


Correspondence to: Martin E. van Royen, Department of Pathology, Josephine Nefkens Institute, Erasmus University Medical Centre, PO-Box 2040, 3000 CA Rotterdam, The Netherlands. E-mail: m.vanroyen@erasmusmc.nl


Androgens exert their key function in development and maintenance of the male phenotype via the androgen receptor (AR). Ligand-activated ARs also play a role in prostate cancer. Despite initial success of treatment by testosterone depletion or blocking of androgen binding to the AR using antiandrogens, eventually all tumors escape to a therapy resistant stage. Development of novel therapies by other antagonistic ligands or compounds that target events subsequent to ligand binding is very important. Here, we validate a fluorescence resonance energy transfer (FRET) based imaging assay for ligand-induced AR activity, based on the conformational change in the AR caused by interaction between the FQNLF motif in the N-terminal domain and the cofactor binding groove in the ligand-binding domain (N/C-interaction). We test the assay using known agonistic and antagonistic ligands on wild type AR and specific AR mutants. Our data show a strong correlation between the ligand-induced AR N/C-interaction and transcriptional activity in wild type AR, but also in AR mutants with broadened ligand responsiveness. Moreover, we explore additional readouts of this assay that contribute to the understanding of the working mechanism of the ligands. Together, we present a sensitive assay that can be used to quantitatively assess the activity of agonistic and antagonistic AR ligands. © 2013 International Society for Advancement of Cytometry


The androgen receptor (AR) is a ligand-dependent transcription factor and member of the steroid receptor (SR) subfamily of nuclear receptors (NRs). Like all SRs, the AR is composed of a central DNA-binding domain, an N-terminal transactivating domain (NTD), and a C-terminal ligand-binding domain (LBD; 1). In the absence of ligand, AR is mainly cytoplasmic, but translocates rapidly to the nucleus after ligand binding [2]. In the nucleus, ARs interact in a highly dynamic manner with promoters and enhancers of target genes, and recruit transcriptional coregulators to regulate gene expression [3-7]. Ligand binding of SRs induces repositioning of helix 12 in the SR-LBD resulting in the formation of a hydrophobic pocket on its surface. Many cofactors, like the p160 family, can bind to this groove via LxxLL-like motifs [8]. The AR-LBD preferentially with FxxLF motifs, which enables an extra level of AR regulation via a ligand-induced interaction with the FQNLF motif in the AR-NTD (AR N/C-interaction; 9–12). It has been suggested that the AR N/C-interaction plays a role in the stability of ligand binding and also protects the cofactor groove for untimely and unfavorable protein–protein interactions [9, 13, 14].

Activated ARs regulate genes involved in development and maintenance of the male phenotype [15]. AR is also a key factor in prostate cancer (reviewed in 16–19). Therapy of metastasized prostate cancers aims at testosterone depletion by chemical or surgical castration or at blocking of androgen binding to the AR using antiandrogens. Despite initial success of these treatments, eventually all tumors escape to a therapy resistant stage [19]. Importantly, endocrine therapy resistant (hormone refractory) prostate cancer remains dependent on a functionally active AR. Therapy resistance can be caused by several mechanisms, including AR mutations resulting in inappropriate activation by antiandrogens and low-affinity agonists, ligand independent activation of the AR and aberrant expression or properties of cofactors (reviewed in 17,18,20–22). Recently, it has been shown that in prostate cancer the expression of enzymes needed for dihydroxytestosterone (DHT) synthesis is increased [23].

Most AR mutations found in prostate cancer so far are localized in the AR-LBD, where the threonine at position 877 is a mutational hot-spot mostly mutated to alanine (AR T877A; 24). The T877A mutation decreases AR ligand specificity and allows other steroids like estrogens, progestogens, and adrenal androgens, but also antiandrogens like OH-flutamide to act as agonists [25-27]. Similarly, in patients treated with antiandrogen bicalutamide a tryptophan to cysteine substitution (W741C) can be found [28, 29], which leads to AR activation by bicalutamide [25, 30, 31]. A third prostate cancer AR mutation affecting its ligand specificity, a leucine–histidine substitution at position 701 (L701H), was found in two independent patients and the MDA PCa 2A prostate cancer cell line [32-34]. The L701H mutation makes the AR highly sensitive to the glucocorticoids, cortisol, and cortisone [35-37]. In the MDA PCa prostate cancer cell line, the L701H mutation was accompanied with the T877A LNCaP mutation, combining the ligand-binding characteristics of the two single mutants [34, 35, 37, 38]. Research on new steroid-based therapies and therapies based on AR inhibitors that target AR activity in another way is, therefore, of great relevance.

Historically, most screening approaches for activating SR ligands or SR inhibitory compounds use target promoter driven reporter genes as assays. Although these assays do quantitatively assess SR transcriptional activity, these measurements do not provide mechanistical details on which of the consecutive steps, from ligand binding to transcriptional activity, is inhibited. More recently, a number of fluorescent indicators have been designed for ligand-mediated responses of SRs. Some of these quantify ligand-induced activity by nuclear translocation or nuclear mobility of fluorescently labeled receptors or receptor chimeras [39-42]. Others make use of ligand-induced conformational changes in the SR-LBDs [43]. An elegant system directly uses the ligand-induced repositioning of helix 12 in the estrogen receptor (ER) LBD to sense ER–ligand interactions [44]. A whole set of indicators has been designed to detect ligands for the ER, glucocorticoid receptor, progesterone receptor, and AR, but also the orphan receptor peroxisome proliferator-activated receptor γ [45-49]. These indicators make use of the conformational change in the LBD that leads to cofactor recruitment to the cofactor groove in the SR-LBD. The conformational change of YFP- and cyan fluorescent protein (CFP)-, double tagged SR-LBD–cofactor peptide chimeras, due to peptide interaction with the cofactor groove in the ligand-activated LBD, was detected by fluorescence resonance energy transfer (FRET).

A similar principle was used on full length ER and AR, tagged with YFP and CFP at both termini [14, 50-53]. In the AR, the conformational change is a direct result of the ligand-induced AR N/C-interaction [14, 52, 53]. Applying FRET on these double tagged ARs showed that the intramolecular AR N/C-interaction was initiated rapidly after ligand binding, followed by an intermolecular N/C-interaction [52]. The FRET approach on YFP- and CFP- double-tagged ARs, was previously used in a high throughput cellular conformation-based screening setting. This assay was used in a screen of FDA-approved drugs and natural products and identified compounds with previously unidentified antiandrogen activity [54-56]. Comparing FRET results from a compound screening of double-tagged ARs in two different cell lines with a traditional transcription reporter assay indicated that the FRET assay is less sensitive to nonspecific cellular variation than a typical promoter–reporter assay. Therefore, the authors suggest that this FRET-based assay is expected to give a higher specificity of the AR function, less noise and similar sensitivity in compound detection in larger screens [54].

We generated a similar FRET assay based on the AR N/C-interaction, extended the use of tagged AR, and identified different readouts of this microscope-based assay, enabling the extraction of additional mechanistical details of AR activity inhibition [14]. This assay provides information on both molecular structure and localization and identifies the activation of subsequent steps of AR function and could well be used in a screening setup for potential clinically relevant AR inhibitors. With this assay, we examined and validated the activation and inhibitory effects of the clinically used antiandrogens; OH-flutamide and bicalutamide on wild type AR and prostate cancer mutants T877A and W741C, respectively. In addition, we examined the response of different mutations at position 701 of the AR on naturally occurring nonandrogenic hormones like glucocorticoids.



The cDNA construct encoding for double tagged AR (pYFP-AR-CFP) was generated as described [14, 57]. In all AR fusion constructs, AR was separated from the fluorescent tag by a flexible (GlyAla)6 spacer further referred to as a single dash [5]. The F23,27A/L26A and F23,27L mutations were introduced by QuikChange mutagenesis (Stratagene) in YFP-AR-CFP. The FQNLF deletion mutant was generated by creating appropriate restriction sites by introducing silent mutations in the N-terminus and replacing a 53 bp fragment by a linker coding for this fragment with deleted FQNLF. The LBD mutation T877A was introduced by QuikChange mutagenesis. To generate the other LBD mutants, W741C, L701H, L701H/T877A, and L701M, the AR-LBDs of pYFP-AR-CFP or pYFP-AR(T877A)-CFP in the case of the double mutant were replaced by a pGFP-AR or pSVAR0 fragment containing the mutant LBD [25, 58]. All new constructs were verified by sequencing.

The (ARE)2-TATA Luc reporter was a gift from G. Jenster (Josephine Nefkens Institute, Rotterdam, The Netherlands).

Cell Culture, Transfection, and Luciferase Assay

Two days before analysis, Hep3B cells, lacking endogenous SR expression, were grown on glass coverslips in six-well plates in α-MEM (Cambrex) supplemented with 5% fetal bovine serum (FBS; HyClone), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. At least 4 h before transfection, the medium was substituted by medium containing 5% dextran charcoal stripped FBS. Transfections were performed with 1 μg/well AR or CFP-YFP expression constructs or 0.5 μg/well empty YFP or CFP expression vector in FuGENE6 (Roche) transfection medium. Four hours after transfection, medium was replaced by medium with 5% dextran charcoal stripped FBS with or without hormone.

For AR transactivation experiments, cells were cultured in 24-well plates on α-MEM supplemented with 5% dextran charcoal stripped FBS in the presence or absence of hormone and transfected using 50 ng AR expression construct and 100 ng luciferase reporter construct. Twenty-four hours after transfection, cells were lysed and luciferase activity was measured in a luminometer (GloMax Microplate luminometer; Promega). Hep3B stably expressing wild type YFP-AR-CFP was generated as previously described [14].

Western Blot Analysis

The size of the expressed AR fusion proteins is checked by Western blot analysis. Hep3B cells were cultured and transfected in six-well plates. Twenty-four hours after transfection, 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, 10mM DTT, and 0.001% Bromophenol blue). After boiling for 5 min, a 5-μL sample was separated on a 10% SDS-polyacrylamide gel and blotted to Immobilon-P Transfer Membrane (Millipore). Blots were incubated with anti-AR (1:2,000; mouse monoclonal F34.4.1) and subsequently incubated with HRP-conjugated goat antimouse antibody (DakoCytomation). Proteins were visualized using Super Signal West Pico Luminol solution (Pierce Chemical Co.), followed by exposure to X-ray film.

Confocal Imaging, YFP/CFP Ratio Imaging, and Acceptor Photobleaching FRET (abFRET) Analysis

Live-cell and immunofluorescence imaging was performed using a confocal laser-scanning microscope (LSM510; Carl Zeiss MicroImaging) equipped with a Plan-Neofluar 40×/1.3 NA oil objective (Carl Zeiss MicroImaging) at a pixel size of 100 nm. An argon laser was used for excitation of CFP and YFP at 458 and 514 nm, respectively. In all quantitative live cell imaging experiments, cells within a similar range of physiologically relevant expression levels of tagged ARs were selected for analysis [14, 57]. Image processing and analysis were performed with the Zeiss AIM software and Microsoft Excel.

N/C-interactions in YFP-AR-FP were assessed using YFP/CFP ratio imaging and abFRET (57 and references therein). In YFP/CFP ratio imaging, cells expressing YFP/CFP double-tagged AR (YFP-AR-CFP) were imaged with an interval of 30 s using a 458 nm excitation at low laser power to avoid bleaching at monitoring laser power. 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 is initiated after adding R1881 and detected by an YFP/CFP ratio change [(IYFPIbg) / (ICFPIbg), a numeric nonlinearly proportional to FRET efficiency], a direct consequence of the induction of energy transfer (FRET) between YFP and CFP at both AR termini. After background subtraction, FRET is simply calculated as: The relative nuclear intensity is determined simultaneously using the YFP emission and is calculated as Inucleus/ (Inucleus + Icytoplasm).

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 nuclear region of approximately 100 μm2 25 times at 514 nm at high laser power, covering the largest part of the nucleus. After photobleaching, a second YFP and CFP image pair was collected. Apparent FRET efficiency was estimated (corrected for incomplete YFP bleaching) using the equation:

display math

where CFPbefore and YFPbefore are the mean prebleach fluorescence intensities of CFP and YFP, respectively, in the area to be bleached (after background subtraction), and CFPafter and YFPafter are the mean postbleach fluorescence intensities of CFP and YFP, respectively, in the bleached area. To correct for bleaching of CFP due to monitoring, we express FRET efficiency relative to positive and negative controls (CFP-YFP fusion (abFRETCFP–YFP fusion) and free CFP and YFP (abFRET0) respectively):

display math

The formulae determining the apparent FRET efficiency, expressed as relative value to a positive and negative control as we used here, accounts for variation in CFP levels and bleaching, incomplete YFP bleaching and interexperimental variation due to varying laser intensities and other hardware related variations. Therefore, these formulae provide a confident quantitative and relative measure for the FRET efficiency.


Acceptor Bleaching FRET on Functionally Active YFP-AR-CFP Represents AR N/C-interaction in Living Cells

The molecular basis of the AR N/C-interaction is the interaction of the 23FQNLF27 motif in the AR-NTD with the cofactor groove in the AR-LBD. To be able to utilize the AR N/C-interaction in a FRET-based assay for ligand-induced AR activation, we tagged wild type and mutant ARs with the fluorescent protein YFP and CFP and transiently expressed them in Hep3B cells lacking endogenous SRs (Fig. 1A). Previously, we and others showed that similar double tagged ARs are applicable in studies on AR function [14, 52]. To determine whether FRET observed in double tagged wt-AR (Fig. 1A) was caused by N/C-interaction (Fig. 1B), we mutated the FQNLF motif to LQNLL and AQNAA in YFP-AR-CFP (YFP-AR(F23,27L)-CFP and YFP-AR(F23,27A/L26A)-CFP, respectively. In addition, we deleted the FQNLF motif (YFP-AR(ΔFQNLF)-CFP). Western blot analysis showed that all AR fusion proteins have the expected size (Fig. 1C). Mutating the FQNLF motif in the AR-NTD does not change the speckled distribution of the AR (Fig. 1D) indicating that AR FQNLF mutants are still able to stably bind DNA [5]. We applied abFRET to these cells where the readout was based on confocal images taken before and after photobleaching the acceptor (YFP) in which the subsequential increase of the donor (CFP) is measured [14, 59-62]. Weakening the N/C-interaction in YFP-AR-CFP by mutating the FQNLF motif indeed proportionally lowered the FRET signal (Fig. 1E). This decrease in apparent FRET efficiency is not due to variation in expression level as seen in Figure 1C, because cells were selected for analysis within a similar intensity range. Furthermore, in similar proportions the AR transcriptional activity on a transiently transfected luciferase reporter gene driven by an androgen-regulated promoter ((ARE)2TATA-luciferase) drops (Fig. 1F). These data indicate that the FRET measured on YFP-AR-CFP represents the AR N/C-interaction and that loss of N/C-interaction by mutating the FQNLF motif results in the loss of AR transcriptional activity.

Figure 1.

FRET on YFP-AR-CFP indicates AR N/C-interaction in living cells. (A) Schematic representation of the YFP-, CFP- double tagged AR (YFP-AR-CFP), and the positioning of the FQNLF mutations. (B) Scheme of the conformational change based on the N/C-interaction. Agonist binding by the AR induces the formation of a binding site in the LBD enabling the N-terminal FQNLF motif to interact with the LBD (N/C-interaction). (C) Western analysis of YFP-AR-CFP (mutants). (D) High resolution confocal images of cells expressing YFP-AR-CFP mutants show that loss of N/C-interaction does not interfere with the speckled nuclear AR distribution. Bars represent 5 μm. Brightness and contrast of the images were enhanced for presentation only. (E) AbFRET on YFP-AR-CFP. Mutating the N-terminal FQNLF to LQNLL or AQNAA, or deletion of the FQNLF motif results in the loss of FRET efficiency indicating that the FRET represents the AR N/C-interaction. Means + 2*SEM of minimally 50 cells measured in two independent experiments are shown. (F) Weakening or loss of AR N/C-interaction results in a less transcriptionally active AR on transiently transfected (ARE)2 TATA Luc reporter gene. Means + 2*SEM of 2 independent experiments are shown.

AR N/C-interaction-based FRET is Hormone Dose Dependent

The dependency of the N/C-interaction on hormone concentration was compared with the efficiency of AR nuclear translocation and AR transcriptional activity. In absence of hormone, the AR was mainly localized in the cytoplasm and translocated to the nucleus only after incubation with a sufficient dose of R1881 (2; Fig. 2A). The Hep3B cells stably expressing YFP-AR-CFP were treated for minimally 12 h to allow AR distribution to reach steady state. Without hormone, the AR shows a 30% nuclear fraction (Fig. 2B). The minimal R1881 concentration necessary to induce the nuclear translocation was 0.1 nM (∼60% nuclear fraction). A plateau in relative nuclear fraction (∼80% nuclear fraction) was reached by culturing the cells in medium containing at least 1 nM R1881 (Fig. 2B). Acceptor bleaching FRET showed a very similar pattern where no N/C-interaction was detected in the absence of hormone, 0.1 nM R1881 induced a substantial FRET signal, but already 0.01 nM R1881 induces some N/C-interaction (Fig. 2C). Also, for initiating transcriptional activity on a transiently transfected luciferase reporter gene driven by a minimal promoter [(ARE)2TATA-luciferase], minimal 0.1 nM R1881 is necessary to induce transcriptional activity of YFP- and CFP- double tagged AR (Fig. 2D).

Figure 2.

The AR N/C-interaction-based FRET efficiency is hormone dose dependent. (A) Overview images of cells stably expressing YFP-AR-CFP in the absence and presence of increasing quantities of R1881 (1 pM–1 μM, 12-h-treatment). With minimal 0.1 nM R1881, the AR efficiently translocates to the nucleus. Bars represent 50 μm. Brightness and contrast of the images were enhanced for presentation only. (B) Quantitative analysis of YFP-AR-CFP nuclear translocation in presence of increasing amounts of R1881. Without R1881, about 30% of the ARs is nuclear. With a high dose of R1881, the nuclear fraction increases to ∼ 80%. Means + 2*SEM of 10 cells measured are shown. (C) AbFRET on YFP-AR-CFP indicates a dose dependency of the AR N/C-interaction. The minimal dose necessary to induce the AR N/C-interaction is 0.01 nM R1881 and reaches a plateau with 1 nM R1881. Means + 2*SEM of 20 cells measured in minimally two independent experiments are shown. (D) Normalized transcription activity of YFP-AR-CFP with increasing doses of R1881. The transcriptional activity of YFP-AR-CFP is about 30% of the untagged variant (Supporting Information Fig. 1). The minimal dose of R1881 to induce transcription is the same for untagged and double tagged ARs (0.1 nM) indicating that the dose dependency of AR activity is not changed by the two tags on YFP-AR-CFP (Supporting Information Fig. 1). Means + 2*SEM of two independent experiments are shown.

Comparing the hormone dose dependency of YFP-AR-CFP with untagged ARs indicate that both were activated at the same minimal concentration of R1881 (0.1 nM), but that YFP-AR-CFP had a lower maximal activity, indicating that the tags on YFP-AR-CFP do not interfere with hormone-binding efficiency but rather affect the maximal activity of the AR (Supporting Information Fig. 1). Furthermore, the absence of N/C-interaction and transcriptional activity but also diminished nuclear translocation in the presence of low concentrations of R1881 suggests an insufficient ligand-binding efficiency by the AR-LBD at these concentrations. We can conclude from these data that the FRET-detected N/C-interaction is a bonafide measure for ligand-induced activity of wild type AR.

Agonist and Antagonist Action Assessed by FRET

Next to AR agonists and AR antagonists, a third class of AR ligands is that of the partial antagonists. These ligands are able to activate the AR at high concentration whereas at lower concentrations these ligands function as antagonists. To investigate whether partial antagonists are able to induce the AR N/C-interaction in living cells, we applied abFRET on Hep3B cells stably expressing YFP-AR-CFP in the presence of 1 μM CPA or RU486. Similar to R1881, DHT-bound AR is distributed in the typical nuclear speckled pattern, not seen for antagonist (bicalutamide and OH-flutamide) bound AR. A very similar speckled distribution pattern is found for partial antagonist RU486, but CPA showed a weaker, less prominent, speckled pattern (Fig. 3A). Confirming previous data, three classes of ligands, agonists, partial antagonists, and full antagonists, indeed showed three levels of transcription activation of AR on a transiently transfected ARE driven luciferase reporter gene. Where both agonists (R1881 and DHT) were able to activate AR and both full antagonists (bicalutamide and OH-flutamide) were not, both partial antagonists (CPA and RU486) showed a minimal AR activation on this reporter gene (Fig. 3B; 63).

Figure 3.

Agonist and antagonist action on wild type AR. (A) High resolution confocal images of Hep3B cells stably expressing YFP-AR-CFP show that agonist (R1881 and DHT) bound ARs are distributed in a typical speckled pattern, which is not present in antagonist (bicalutamide and OH-flutamide) bound ARs. Partial antagonist RU486 is also able to induce the AR speckled pattern, whereas partial antagonist CPA-bound AR only shows a weak speckled pattern. Bars represent 5 μm. Brightness and contrast of the images were enhanced for presentation only. (B) Normalized transactivation activity on a transiently transfected (ARE)2TATA-luciferase reporter gene. Identical to results of the FRET assay, only pure agonists (black bars) are able to induce a strong AR transcription activity. Partial antagonists (grey bars) only minimally activate the AR, whereas pure antagonists (bicalutamide and OH-flutamide; white bars) do not induce AR activity. (C) AbFRET on YFP-AR-CFP. Both pure agonists (black bars) are able to induce the AR N/C-interaction, whereas both partial antagonists (grey bars) only show a weak FRET signal. Means ± 2*SEM of two independent experiments are shown.

Both agonists (DHT and R1881), but not the full antagonists (OH-flutamide and bicalutamide), were able to induce the AR N/C-interaction. In contrast to in vitro pull-down and mammalian two-hybrid assays, both partial antagonists (CPA and RU486) showed a detectable, although moderate, FRET efficiency, indicating that these partial antagonists did induce the AR N/C-interaction (Fig. 3C; 64–66). This apparent discrepancy suggests a higher sensitivity of the FRET assay compared to the previously used assays, enabling the detection of short and possibly weak interactions.

In conclusion, the FRET-based AR ligand assay identified three classes of AR ligands on the basis of the induction of the N/C-interaction. Furthermore, the FRET data on the N/C-interaction status of ARs bound with partial and full antagonists and agonists correlates very well with the AR transcriptional activity.

FRET Reveals Agonist and Antagonist Activity and Ligand Competition on Wild Type and Mutant AR

To further explore the FRET-based assay, two AR-LBD mutations, W741C [28] and T877A [24], found in prostate cancer from patients treated with bicalutamide and OH-flutamide, respectively, were introduced in YFP-AR-CFP (Fig. 4A). Western blot analysis showed that all fusion proteins were of expected size (Fig. 4B). Previously, it has been show that these mutations enable antagonists to act as agonists by restoring the coactivator groove in antagonist-bound AR [30, 31]. Corroborating previous data, both AR antagonists do not induce a speckled AR distribution in Hep3B cells transiently expressing wild type YFP-AR-CFP, as is found for R1881, but rather show a homogeneous distribution like is previously shown in DNA-binding deficient mutants, for example, A573D (Fig. 4C; 5). In contrast, antiandrogen (bicalutamide and OH-flutamide) bound double tagged AR W741C and AR T877A, respectively, do show the typical speckled pattern (Fig. 4C). As shown before, the nuclear distribution of wild type AR and these AR mutants bound with R1881 and these antiandrogens is correlated with nuclear mobility and transcriptional activity [25]. Indeed, transcriptional activity shows a very similar effect of R1881 activation of wild type AR and both mutants, and activation by bicalutamide of only AR W741C and by OH-flutamide of AR T877A (Fig. 4D). The same effect is found in the N/C-interaction status of these ARs (Fig. 4E). Again, R1881 induces the N/C-interaction in wild type AR and both W741C and T877A mutants, as shown by abFRET. Antiandrogens bicalutamide and OH-flutamide induce the N/C-interaction only in mutants W741C and T877A, respectively. The decrease in apparent FRET efficiency is not due to variation in expression level as seen in Figure 4B, because cells were selected within a similar range of AR expression levels. These data show that also for antiandrogens and ARs with mutations in the ligand-binding pocket, the degree to which N/C-interaction occurs is strongly correlated to the transcriptional activity of the AR mutants and their ligands.

Figure 4.

FRET identifies agonist and antagonist action and ligand competition on AR (mutants). (A) Schematic representation of YFP-AR-CFP (mutants). (B) Western blot analysis of AR mutants W741C and T877A. (C) High resolution confocal images of YFP-AR-CFP or its prostate cancer variants AR W741C and AR T877A. AR mutants W741C and T877A treated with, respectively, bicalutamide or OH-flutamide, and not vice versa, show a typical speckled AR distribution. Bars represent 5 μm. Brightness and contrast of the images were enhanced for presentation only. (D) Normalized transactivation activity YFP-AR-CFP (mutants). AR agonist R1881 (0.1 μM) is able to activate wild type AR and AR mutants W741C and T877A, but not DNA-binding deficient AR mutant A573D. Prostate cancer AR mutations W741C and T877A enable AR activation by the antagonists, bicalutamide and OH-flutamide (1 μM), respectively. (E) AbFRET efficiency of YFP-AR-CFP in the presence of an agonist (0.1 μM R1881) or an antagonist (1 μM Bicalutamide or OH-flutamide). AR prostate cancer mutants W741C and T877A show an AR N/C-interaction in the presence of the antagonists bicalutamide or OH-flutamide, respectively and not vice versa (see also D). (F) AbFRET efficiency of YFP-AR-CFP treated with both R1881 and OH-flutamide. Simultaneous treatment of 0.1 nM R1881 with 0.01 μM to 0.01 mM OH-flutamide results in the inhibition of R1881-induced AR N/C-interaction. Means ± 2*SEM of two independent experiments are shown. (G and H) Time laps [YFP] / [CFP] ratio imaging of cells expressing wild type YFP-AR-CFP treated for minimal 12 h with either 0.1 μM R1881 (dark grey curve in G) 0.1 nM R1881 (H), or 1 μM OH-flutamide (black and light grey curves in G). At t = 0 min, a second hormone was added. The addition of neither a high dose of OH-flutamide (1 μM) to 0.1 μM R1881 nor a low dose of R1881 (0.1 nM) to 1 μM OH-flutamide results in a change of the YFP / CFP ratio. Only 0.1 μM R1881 was able to compete with 1 μM OH-flutamide and induce the AR N/C-interaction. In contrast, only a high dose (1 μM) of OH-flutamide was able to compete with a low dose (0.1 nM) of R1881 and diminish the AR N/C-interaction in a period of about 10 h (H).

The antagonistic effect of antiandrogens is only detectable in competition with an agonist. Cells expressing YFP-AR-CFP were simultaneously treated with a low concentration of the agonist R1881 (0.1 nM) and increasing concentrations of OH-flutamide. AbFRET analysis showed the proportional loss of the N/C-interaction with increasing amounts of antagonist OH-flutamide (Fig. 4F). The addition of 10 μM OH-flutamide together with 0.1 nM R1881 resulted in an identical FRET efficiency to cells treated with 10 μM OH-flutamide only (Fig. 4F).

The competition between agonists and antagonist was studied in more detail by time lapse YFP / CFP ratio imaging, monitoring the effect of antagonist or agonist addition to cells expressing YFP-AR-CFP, primarily treated with either an agonist or an antagonist, respectively. In this assay, instead of abFRET we used (increase of) YFP / CFP ratio as a measure for the AR N/C-interaction induction. A low concentration of R1881 (0.1 nM) was not sufficient to induce the N/C-interaction if cells were already treated with a high dose of OH-flutamide (1 μM; Fig. 4G, light grey curve), nor was a concentration of 1 μm OH-flutamide sufficient to compete with a relative high dose of R1881 (0.1 μM) and N/C-interaction, therefore, was not abolished (Fig. 4G, dark grey curve). In contrast, addition of this same dose of R1881 (0.1 μM) at t = 0 min, to YFP-AR-CFP treated for minimally 12 h with a high dose of OH-flutamide resulted in a rapid increase in YFP / CFP ratio from ∼1.3 to ∼1.8, indicating a rapid induction of the AR N/C-interaction (Fig. 4G, black curve). In the opposite set-up, initial treatment of cells expressing YFP-AR-CFP with 0.1 nM R1881 moderately induced the AR N/C-interaction (see also Fig. 2C). Adding 1 μM OH-flutamide (at t0) to these cells showed a decrease of YFP/CFP ratio, although it took about 10 h to reach the new steady state at the YFP / CFP ratio level similar to the initial level of 1 μM OH-flutamide treatment (Fig. 4H).

Together, these data show that this FRET-based assay on YFP-AR-CFP is applicable for the detection of agonistic and antagonistic activities of ligands. These types of analyses can provide relative binding efficiencies of different ligands in living cells.

FRET Detection of Broadened Ligand Responsiveness of AR L701 Mutants

A third type of AR mutant found in prostate cancer [32, 33] is a mutation at position 701 in the AR-LBD (L701H), which results in a broadened ligand responsiveness of the AR to corticosteroids [37]. This mutant and a variant (L701M) were transiently expressed in Hep3B cells and tested for their ability to induce the AR N/C-interaction (Fig. 5A). Western blot analysis showed that all fusion proteins were of the expected size (Fig. 5B). As described recently, both mutants (L701H and L701M) and also a double mutant (L701H/T877A; 34), found in the prostate cancer cell line MDA PCa, showed that agonist R1881 was able to activate all AR variants (Figs. 5C–5F, black bars). 11-Desoxycorticosterone was able to induce the N/C-interaction in all three AR mutants (Figs. 5D–5F, dark grey bars) but not wild type AR (Fig. 5C, red bar). Similarly, cortisol induced the N/C-interaction in two AR mutants (L701H and L701H/T877A; Figs. 5D and 5E, light grey bars) and corticosterone in the AR L701/T877A double mutant only (Fig. 5E, medium grey bar). In concordance, the AR mutants in Figure 4, AbFRET analysis of YFP-AR-CFP to analyze the induction of the N/C-interaction in wild type AR and the L701 mutants completely mirrors the ability to induce transcriptional activity (Figs. 5G–5J). Strikingly, the ability of most of these hormones to induce the N/C-interaction in the different ARs correlated with the subnuclear speckled distribution of these ARs in the presence of these ligands (Fig. 5K). The remaining combinations of ligand-bound ARs (11-desoxycorticosterone-bound wild type AR, corticosterone-bound wild type AR, AR L701H, and AR L701M) do not show the AR N/C-interaction and the typical speckled pattern but a more homogeneous distribution. Two of the four exceptions in this are 11-desoxycorticosterone-bound AR L701H, which shows a less pronounced speckled pattern and AR L701M, of which no speckled pattern can be detected. More importantly, two other exceptions, wild type AR and AR L701M in the presence of cortisol, show limited nuclear translocation (Fig. 5K). The decrease in apparent FRET efficiency in Figures 5G–5J is not due to variation in expression level as seen in Figure 5B, because cells within a small range of AR expression were selected for analysis. In conclusion, most data obtained with this FRET assay correlates well with the AR subnuclear distribution and with transcriptional activity. Moreover, the FRET data highly correlates with previously described AR activity [58].


In recent years, a number of fluorescent indicators have been developed to study ligand or compound-mediated responses of SRs. Most of these fluorescent indicators report on consecutive steps during SR transcription activation, including hormone-induced conformational changes and dimerization (FRET), nuclear translocation (time lapse imaging), cofactor interactions (FRET), and DNA binding (fluorescence recovery after photobleaching (FRAP) (reviewed in 50,57,67). For AR, two early events induced by ligand binding are the formation of a groove in the C-terminal LBD and subsequent interaction of the N-terminal FQNLF motif with the groove (N/C-interaction; 9,11,12,26,52). Here, we validated a FRET-based assay for ligand-induced AR activity that quantitatively detects the N/C-interaction using cells expressing YFP-AR-CFP.

The N/C-interaction in Wild Type and Mutant ARs Reflects Transcriptional Activity

Mutational analysis of the FQNLF motif showed that FRET efficiency of YFP-AR-CFP is a measure for the N/C-interaction (Figs. 1E and 1F). Hormone dose responsiveness of the FRET-based detection of the N/C-interaction was very similar compared to its transcriptional activity (Fig. 2). Furthermore, the N/C-interaction status in wild type AR and a set of specific cancer-derived AR mutants with a broadened ligand responsiveness (T877A, W741C, L701H/T877A, L701H, and a variant L701M) bound with a variety of hormones very well reflected the AR transcriptional activity (Figs. 3-5;[24-26, 30, 36, 37, 58]).

The correlation between transcriptional activity and the N/C-interaction status in antagonist-bound ARs can be explained by the formation of a functional coactivator binding groove in the LBD. Binding of antagonist but possibly also nonandrogenic ligands displaces helix 12 over the coactivator-binding groove, as is shown for other antagonist-bound NRs [68-71]. This displacement results in a nonfunctional coactivator groove and, as a consequence, lack of N/C-interaction and transcriptional activity due to loss of coactivator binding either directly to the groove or indirectly via the N/C-interaction to the NTD. AR mutations L701H, L701M, W741C, and T877A results in broadened ligand responsiveness by providing a different set of pocket-ligand interactions or by avoiding steric hindrance for ligand binding. The overall conformation of these AR-LBD mutants bound to their cognate antagonists and possibly also nonandrogenic ligands is very similar to that of agonist-bound wild type AR-LBD [31, 38, 72, 73]. These mutations restore the coactivator groove in antagonist or nonandrogenic ligand-bound AR enabling N/C-interaction and transcriptional activity (Figs. 4 and 5; 24,26,30,31).

Figure 5.

FRET detection of broadened ligand responsiveness of AR L701 prostate cancer mutants. (A) Schematic representation YFP-AR-CFP (mutants). (B) Western blot analysis of AR L701 mutants. (C- -F) Normalized transactivation activity of wild type and AR-LBD mutants. AR agonist R1881 (0.1 μM) is able to activate all wild type AR and AR L701 mutants. Mutations on position L701 (L701H, L701H/T877A and L701M) differentially enable AR activation by the corticosteroids; 11-desoxycorticosterone, corticosterone, and cortisol (1 μM), respectively. Means ± 2*SEM of two independent experiments are shown. (G- -J) In concurrence with the data on transcriptional activity, AbFRET shows differential ability of the corticosteroids; 11-desoxycorticosterone, corticosterone and cortisol, to induce FRET in cells expressing wild type YFP-AR-CFP or AR-LBD mutants (L701H, L701H/T877A, and L701M). Means ± 2*SEM of two independent experiments are shown. (K) High resolution confocal images of Hep3B cells expressing wild type YFP-AR-CFP of its AR L701H, L701H/T877A, and L701M variants in the presence of R1881 or the corticosteroids; 11-desoxycorticosterone, corticosterone, and cortisol. Most hormones translocate the AR to the nucleus with exception of wild type AR and, in a lesser degree, AR L701M in the presence of cortisol. Bars represent 5 μm. Brightness and contrast of the images was enhanced for presentation only.

In total, the strong correlation between the N/C-interaction and transcriptional activity qualifies the FRET assay as a bonafide ligand-induced AR activation assay.

Role of N/C-interaction in Transcriptional Activity

The strong correlation between the N/C-interaction and transcriptional activity contrasts the lack of a known direct functional role for the N/C-interaction in transcription activation. Moreover, it is generally accepted that, in contrast to other SRs, the transactivation function in the AR-LBD (AF-2) is poorly active and that the major transactivation function of the AR is localized in the AR-NTD (AF-1), raising questions on the role of the AR-LBD coactivator groove. Therefore, it is surprising that the N/C-interaction status reflects so well the transcriptional activity on a transiently transfected reporter driven by a minimal promoter. The most likely explanation is that the N/C-interaction has a function in facilitating cofactor binding to the AR-NTD, by exposing regions in the NTD. If so, mutations in the FQNLF motif disabling the N/C-interaction, possibly results in less efficient binding of coactivators to the NTD. Alternatively, the N/C-interaction may regulate cofactor binding to the coactivator groove in the AR-LBD by blocking the groove when interactions are not required [14]. Loss of N/C-interaction then disables regulation of coactivator binding which may result in unfavorable protein interactions and inefficient coactivator binding to the coactivator-binding groove. This is possibly reflected by the remaining transcriptional activity in N/C-interaction deficient mutants (Fig. 1).

Limitations of the FRET-based Assay for Ligand-induced Activity

It is important to note that the correlation between the AR N/C-interaction and transcriptional activity does not necessary point to a causal relationship, but rather that the N/C-interaction reflects an individual step leading to a functionally active AR. The N/C-interaction can, therefore, be used to identify initial activating or repressing activity of AR ligands, but additional functional steps, like DNA binding and cofactor recruitment could be disabled resulting in aberrant activity of the AR or total lack of AR activity, without any effect on the N/C-interaction. For example, we have shown that aberrant AR activity of, for example, AR DNA binding deficient or dimerization mutants is not necessarily reflected by the N/C-interactions status [14, 57]. As a result, it is conceivable that compounds that do not act like ligands, are able to efficiently inhibit the AR activity, but do not interfere with the N/C-interaction and, therefore, will not be detected in this assay.

Moreover, the causal relationship between the N/C-interaction and transcriptional activity is blurred, because the requirement for the N/C-interaction is suggested to be promoter specific, implicating that the choice of promoter could be important [74]. In addition, ChIP data indicated that N/C-interaction deficient AR mutants were able to bind to plasmid-based AREs but not chromatin, but this could not be confirmed by FRAP (data not shown; 75).

FRET-based Detection of Antagonist Activity

Antagonist activity of wild type YFP-AR-CFP was detected in a ligand competition setting. OH-flutamide was able to inhibit an R1881 induced N/C-interaction in a dose dependent manner (Fig. 4F). A more detailed analysis of ligand competition showed a rapid (<10 min) replacement of OH-flutamide by a sufficient dose of R1881, and very slow (>10 h), complete replacement of a low dose of R1881 by a high dose of OH-flutamide in reverse set-up (Figs. 4G and 4H). This type of analysis enables the detection of relative ligand-binding efficiencies for agonists versus antagonists.

A Quantitative Microscopy Approach of this Assay Provides Additional Mechanistical Data on Ligand Activity

The quantitative microscopic approach, as used here, not only enables single cell selection to allow analysis of cells with physiological relevant AR expression levels, but also provides additional relevant parameters together with the AR N/C-interaction, including nuclear translocation, and subnuclear distribution. These parameters provide extra information on the mechanism of inhibition of antiandrogens or other antagonistic compounds. The speckled distribution of DHT-, R1881-, and RU486-bound AR, the less pronounced speckled pattern of CPA-bound AR and the lack of speckles in OH-flutamide- and bicalutamide-bound AR, reflect DNA-binding characteristics of ARs bound with these ligands as previously found with FRAP (Fig. 3A; 25,76).

In general, N/C-interaction status in wild type AR and AR-LBD mutants correlated with their speckled nuclear distribution (Figs. 3-5). The only exceptions are the 11-desoxycorticosterone-bound AR L701H, which shows a less pronounced speckled pattern and AR L701M, of which no speckled pattern can be detected (Fig. 5K). Exactly these combinations show relatively lower FRET efficiency comparable to their transcriptional activity, suggesting that FRET-based detection of ligand-induced activation is more sensitive than speckle detection (Figs. 5D and 5F). Importantly, most used ligands induced nuclear translocation of ARs, and thus bind the AR, independent of their ability to induce N/C-interaction. In contrast, the strongly inhibited nuclear translocation in wild type AR and AR L701M in the presence of cortisol indicated that these ARs probably had a low affinity for cortisol (Figs. 3-5).

In conclusion, this FRET-based assay using double tagged ARs provides a sensitive application in screens for both agonistic and antagonistic AR ligands but also other compounds. Because the N/C-interaction initiation is an early transition following ligand binding, this assay could very well be used as surrogate assay for ligand-induced AR activation. Although it was not shown here, in larger screening programs, FRET-based techniques (e.g., YFP/CFP ratio imaging or sensitized emission) in general are especially applicable in high-throughput settings [54, 55].