PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of adtMary Ann Liebert, Inc.Mary Ann Liebert, Inc.JournalsSearchAlerts
Assay and Drug Development Technologies
 
Assay Drug Dev Technol. 2016 May 1; 14(4): 226–239.
PMCID: PMC4876501

Development and Implementation of a High-Throughput High-Content Screening Assay to Identify Inhibitors of Androgen Receptor Nuclear Localization in Castration-Resistant Prostate Cancer Cells

Abstract

Patients with castration-resistant prostate cancer (CRPC) can be treated with abiraterone, a potent inhibitor of androgen synthesis, or enzalutamide, a second-generation androgen receptor (AR) antagonist, both targeting AR signaling. However, most patients relapse after several months of therapy and a majority of patients with relapsed CRPC tumors express the AR target gene prostate-specific antigen (PSA), suggesting that AR signaling is reactivated and can be targeted again to inhibit the relapsed tumors. Novel small molecules capable of inhibiting AR function may lead to urgently needed therapies for patients resistant to abiraterone, enzalutamide, and/or other previously approved antiandrogen therapies. Here, we describe a high-throughput high-content screening (HCS) campaign to identify small-molecule inhibitors of AR nuclear localization in the C4-2 CRPC cell line stably transfected with GFP-AR-GFP (2GFP-AR). The implementation of this HCS assay to screen a National Institutes of Health library of 219,055 compounds led to the discovery of 3 small molecules capable of inhibiting AR nuclear localization and function in C4-2 cells, demonstrating the feasibility of using this cell-based phenotypic assay to identify small molecules targeting the subcellular localization of AR. Furthermore, the three hit compounds provide opportunities to develop novel AR drugs with potential for therapeutic intervention in CRPC patients who have relapsed after treatment with antiandrogens, such as abiraterone and/or enzalutamide.

Introduction

Castration-resistant prostate cancer (CRPC) is currently incurable, making prostate cancer the second most common cause of cancer death among men in the United States in 2012 with >28,000 deaths and >241,000 new cases diagnosed.1 Multiple studies have shown that the androgen receptor (AR) is activated in prostate cancer through several mechanisms, including AR overexpression, mutation, hypersensitization, and/or intratumoral androgen synthesis in patients relapsed after androgen deprivation therapy.2–8 Overexpression and knockdown studies have demonstrated that AR is a key molecular determinant and a validated therapeutic target for CRPC.9,10

The importance of AR as a target in the majority of CRPC patients is emphasized by the mechanisms of the two drugs most recently approved by the federal drug administration for the treatment of CRPC, abiraterone, a potent inhibitor of testosterone synthesis,11 and MDV3100 (Enzalutamide®), a novel AR antagonist.12,13 However, prostate cancers develop resistance to therapies, including the most recent second-generation antiandrogens.11,14–16 Also, some AR-positive prostate cancer cell models, such as 22Rv1, are insensitive to abiraterone and/or MDV3100.17–19 Therefore, there is a need for the development of more effective inhibitors of AR function to treat CRPC patients who have developed resistance to antiandrogens, including abiraterone and MDV3100.

As a member of the steroid receptor superfamily, AR is a ligand-dependent transcription factor that controls the expression of androgen-responsive genes.20 Intracellular trafficking is an important mechanism in the regulation of many transcription factors, including AR. To transactivate its target genes, AR must translocate from the cytoplasm into the nucleus, and retention of AR in the cytoplasm is one mechanism to prevent its transactivation activity. Thus, a key regulatory step in the action of AR is its nuclear translocation. AR contains one nuclear localization signal (NL1) within the DNA-binding domain and hinge region, one ligand-induced nuclear localization signal (NL2) within the ligand-binding domain (LBD), and a nuclear export signal in the ligand-free LBD.21–24 In addition, the N-terminal domain of AR contains amino acid sequences that can modulate subcellular localization.25,26 In androgen-sensitive cells, AR is localized to the cytoplasm in the absence of ligand.27 On exposure to androgens, AR translocates to the nucleus where it binds to specific androgen response element DNA sequences to transactivate target genes. However, in CRPC cells, AR remains in the nucleus even in the absence of androgens and transactivates androgen-responsive genes, leading to uncontrolled growth of prostate tumors.6,28 Therefore, approaches that can reduce the level of nuclear AR may provide an effective therapy against CRPC.

To date, no high-throughput screens to identify small molecules capable of specifically and effectively reducing the nuclear localization of AR in CRPC cells have been published. In this study, we report the development and implementation of the first high-throughput high-content screening (HCS) assay to identify small molecules capable of reducing AR nuclear localization in CRPC cells.

Materials and Methods

Reagents and Plasmid

Dimethyl sulfoxide (DMSO), 17-allylamino geldanamycin (17-AAG), formaldehyde and Lipofectamine™ were purchased from Sigma-Aldrich, St. Louis, MO. Hoechst 33342 was obtained from Invitrogen (Carlsbad, CA), phosphate-buffered saline (PBS) and RPMI-1640 medium from Corning Cellgro, fetal bovine serum (FBS) from Atlanta Biologicals (Flowery Branch, GA), l-glutamine from Gibco/Life Technology, and G418 from Gemini Bio-Products. The GFP-AR-GFP (2GFP-AR) expression vector was generated by adding another green fluorescent protein (GFP) cDNA at the C-terminus of the AR coding sequence of the GFP-AR expression vector, which is based on the expression vector pEGFP-C1 (Clontech).24 The 2GFP-AR expression vector was verified by DNA sequencing.

Cell Culture and Stable Transfection

C4-2 cells were purchased from UroCor (Oklahoma City, OK).29 Cells were maintained in the RPMI-1640 medium supplied with 10% FBS and 1% l-glutamine at 37°C with 5% CO2. C4-2 cells were transfected with the 2GFP-AR expression vector using Lipofectamine according to the manufacturer's protocol (Invitrogen). The transfected cells were cultured in the presence of 800–1,000 μg/mL G418, individual C4-2 colonies expressing 2GFP-AR were selected, and the subcellular localization of 2GFP-AR was determined by fluorescence microscopy using a Nikon TE 2000U inverted microscope. The N3 C4-2-2GFP-AR clone was used for high-throughput screening.

Compound Library

A library of 219,055 compounds provided by the National Institutes of Health (NIH) was formatted at 10 mM concentration in DMSO and arrayed into 384-well microtiter master plates. The storage and use of the library in high-throughput screening were carried out as previously described.30

Image Acquisition and Analysis

The images of 2GFP-AR in C4-2 cells cultured in 384-well plates were captured using the ArrayScan VTI (AS-VTI) platform (Thermo Fisher Scientific, Waltham, MA) as described previously.31,32 Images of Hoechst-stained nuclei (Ch1) and 2GFP-AR (Ch2) in fixed cells were sequentially acquired on the AS-VTI using a 10X 0.3NA objective, and the 2 channel molecular translocation (MT) image analysis algorithm was used to quantify the expression of the 2GFP-AR biosensor in the digital images of the C4-2-2GFP-AR cells. Hoechst 33342 was used to stain the nuclei of the C4-2-2GFP-AR cells, and the image analysis segmentation used the intensity over background, width and area parameters of this fluorescent signal in channel 1 (Ch1) to define a nuclear mask for each cell as described previously.31 The nuclear mask was contracted from the edge of the identified nucleus to reduce cytoplasmic contamination within the nuclear area, and the reduced mask was used to segment and quantify the amount of target channel, 2GFP-AR (channel 2, Ch2) fluorescence within the nucleus. The nuclear mask was also expanded to cover as much of the cytoplasmic region as possible within the cell boundary. Removal of the original nuclear region from this expanded mask creates a ring mask that covers the cytoplasmic region. The outputs of the MT image analysis algorithm are quantitative data such as the total and average fluorescent intensities of the Hoechst-stained objects (Ch1), the selected object or cell count (SCC) from Ch1, the total and average fluorescent intensities of the 2GFP-AR (Ch2) signals in the nucleus (circ) or cytoplasm (ring) regions as an overall average value, or on an individual cell basis. To quantify the changes in subcellular distribution between the nucleus and cytoplasm of 2GFP-AR, we used the MT image analysis algorithm, which calculates a mean average intensity difference by subtracting the average 2GFP-AR intensity in the ring (cytoplasm) region from the average 2GFP-AR intensity in the circ (nuclear) region of Ch2; mean circ–ring average intensity difference in channel 2 (MCRAID-Ch2). To quantify and compare the expression levels and subcellular localization of 2GFP-AR in C4-2-2GFP-AR cells in the presence or absence of small molecules, we analyzed images with the MT algorithm.

Automated 2GFP-AR Subcellular Localization HCS Assay Protocol

The 2GFP-AR translocation HCS assay protocol was very similar to the protocol developed for dexamethasone-induced glucocorticoid receptor nuclear translocation HCS assay (Table 1).31 N3 C4-2-2GFP-AR cells were seeded in 384-well plates at density of 3,000 cells/well in complete media and cultured overnight. The cells were then treated overnight with 0.2% DMSO vehicle, 0.2% DMSO in the presence of 20 μM of the compounds in the library, or 0.2% DMSO plus 3.0 μM 17-AAG. The treated cells were subsequently fixed using 3.7% formaldehyde containing 2 μg/mL Hoechst 33342 in PBS at room temperature. Images of Hoechst-stained nuclei (Ch1) and 2GFP-AR (Ch2) expression in fixed cells were sequentially acquired on the AS-VTI using a 10 × 0.3NA objective and were analyzed using the MT image analysis algorithm. Data processing, visualization, statistical analysis, and curve fitting were conducted as described previously.31 Data processing for the 2GFP-AR nuclear localization screen was performed using ActivityBase™ (IDBS, Guildford, Alameda, CA) and CytoMiner. Processed data and HCS multiparameter features were visualized using Spotfire™ DecisionSite™ (Somerville, MA) software. An ActivityBase primary HTS template was created that automatically calculated % inhibition together with plate control signal-to-background ratios (S:B) and Z’-factor coefficients.33 For the 2GFP-AR nuclear localization screen, we utilized the MRAID-Ch2 values of the 0.2% DMSO maximum plate control wells (n = 32) and the MCRAID-Ch2 values of the 3.0 μM 17-AAG minimum plate control wells (n = 32) to normalize the MCRAID-Ch2 values of the compound data and to represent 100% and 0% activation or translocation of 2GFP-AR to the nucleus, respectively. We also constructed an ActivityBase concentration–response template to calculate percent inhibition together with plate control S:B ratios and Z’-factor coefficients for quality control purposes.31,33,34 For the cytotoxic and fluorescent outlier analysis, we used the plate-based statistical scoring method z-score that is estimated from the 320 compound wells (no plate control wells) on an assay plate.34 It is defined as z-score = (Xiequation eq1)/Sx. Where Xi is the raw measurement on the ith compound, equation eq2 and Sx are the mean and standard deviation of all the sample measurements on a plate. A deviation of −3 below the sample average on the plate is frequently utilized as a statistical threshold or cutoff for active compounds.34

Table 1.
Androgen Receptor Nuclear Localization HCS Assay Protocol

Western Blot Analysis

C4-2 cells were lysed in a modified radioimmune precipitation assay buffer containing 10 mM Tris-Cl (pH 8.0), 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl, and 1% protease inhibitor cocktail (Sigma-Aldrich).

Protein concentrations in cell lysates were measured using the bicinchoninic acid assay reagent. Lysates were subjected to SDS–PAGE, and Western blotting was carried out using antibodies against AR (N-20, sc-816) and prostate-specific antigen (PSA) (sc-7638) from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). To determine the effect of SID 3712502 on AR and PSA protein levels, we treated C4-2 cells with various concentrations of SID 3712502 for 48 h and then harvested for cell lysate preparation. An antibody against GAPDH (sc-25778; Santa Cruz Biotechnology, Inc.) or β-actin (sc-47778; Santa Cruz Biotechnology, Inc.) was used as a control to demonstrate equal protein loading.

Results

Generation and Characterization of 2GFP-AR Expressing Cell Lines in the C4-2 CRPC Cell Background

To establish a cell-based high-throughput screening assay to identify small molecules capable of reducing nuclear AR levels in prostate cancer cells, we transfected the C4-2 human prostate cancer cell line with the 2GFP-AR expression vector and isolated stably transfected clones. C4-2 was chosen because it is a well-characterized CRPC cell line derived from LNCaP cells.29 We transfected the 2GFP-AR expression vector into C4-2 cells and then used G418 selection to isolate stably transfected clones for subsequent studies. The 2GFP-AR expression vector was used because GFP-AR did not provide a green fluorescence signal strong enough for the HCS assay (data not shown). The N3 clone of C4-2-2GFP-AR cells was selected for the development of AR nuclear localization HCS assay. In addition to endogenous expression of AR, the stably transfected cell line exhibited a slower migrating larger molecular weight immunoreactive band on Western blots probed with the anti-AR antibody (Fig. 1A). Endogenous AR and 2GFP-AR were reported previously to predominantly localize to the nuclei of C4-2 cells cultured in an androgen-free medium.35 As expected, 2GFP-AR was predominantly localized to the nuclei of C4-2 cells cultured in androgen-free media, and exposure to DHT further enhanced that nuclear distribution (Fig. 1B). The androgen-independent nuclear localization of AR requires HSP90 function and can be effectively inhibited by the HSP90 inhibitor 17-AAG.35 Exposure to 17-AAG for 16 h appeared to cause some reduction in the protein levels of both endogenous AR and 2GFP-AR in the N3 clone of C4-2-2GFP-AR cells (Fig. 1A). Both the androgen-independent and, to a lesser extent, the DHT-induced, nuclear localization of 2GFP-AR in C4-2 cells were markedly reduced by exposure to 17-AAG (Fig. 1B, C). These observations indicate that the stability and subcellular localization of 2GFP-AR in the N3 clone of C4-2-2GFP-AR cells behave similarly to endogenous AR with respect to both androgen responsiveness and HSP90 inhibition by 17-AAG. These data further substantiated the selection of the N3 C4-2-2GFP-AR cell line for the development of the 2GFP-AR nuclear localization HCS assay and confirmed that 17-AAG was a suitable control compound that induced the desired cytoplasmic 2GFP-AR distribution phenotype in the C4-2 CRPC cell line.

Fig. 1.
Characterization of the N3 C4-2-2GFP-AR clone selected for assay development and screening. (A) Effects of 17-AAG on 2GFP-AR and endogenous AR protein levels in the stably transfected N3 clone of C4-2 cells. N3 C4-2-2GFP-AR cells were cultured in ligand-free ...

Development and Optimization of the 2GFP-AR Nuclear Localization HCS Assay

To evaluate whether the 2GFP-AR nuclear localization assay would be compatible with HCS, we acquired 10× images of Hoechst-stained nuclei (Ch1) and 2GFP-AR (Ch2) on the AS-VTI automated HCS platform and used the MT image analysis algorithm to quantify the 2GFP-AR subcellular distribution phenotype in N3 C4-2-2GFP-AR cells cultured in the presence or absence of 312 nM 17-AAG (Figs. 2 and 33). Consistent with the images presented in Figure 1 that were acquired on a Nikon TE 2000U inverted fluorescent microscope, 2GFP-AR was predominantly localized to the nuclei of N3 C4-2-2GFP-AR cells cultured in the androgen-free medium, and the colocalization of the Hoechst and 2GFP-AR signals was indicated by the light blue nuclear staining of the color composite images acquired on the AS-VTI (Fig. 2A). However, in N3 C4-2-2GFP-AR cells cultured overnight in the presence of 312 nM 17-AAG, the 2GFP-AR was predominantly localized to the cytoplasm of the cell as indicated by the dark blue nuclear staining of the color composite images (Fig. 2A). The MT image analysis algorithm was used to quantify the relative subcellular distribution of 2GFP-AR between the nucleus and cytoplasm regions of N3 C4-2-2GFP-AR cells (Fig. 2B). The MT algorithm generates a mask of the nucleus for each cell by segmenting the Ch1 Hoechst images into background and stained objects that were classified as nuclei because they exhibited average intensities above a preset threshold and also had suitable width and area measurements (Fig. 2B). The MT algorithm then uses the nuclear masks for each cell to segment the corresponding Ch2 2GFP-AR images into nuclear “circ” and cytoplasm “ring” regions for each cell (Fig. 2B).

Fig. 2.
Image acquisition and analysis on the ArrayScan-VTI automated HCS platform. (A) Image Acquisition. Grayscale images of Hoechst-stained nuclei (Ch1) and 2GFP-AR expression (Ch2) in N3 C4-2-2GFP-AR cells that had been fixed in paraformaldehyde after being ...
Fig. 3.
Quantitative data extracted from the digital images of the Hoechst-stained nuclei and 2GFP-AR expression (Ch2) in N3 C4-2-2GFP-AR cells by the MT image analysis algorithm. 10× images of Hoechst-stained nuclei (Ch1) and 2GFP-AR expression (Ch2) ...

The MT image analysis algorithm extracts and outputs a large number of quantitative parameters such as the integrated and average fluorescent intensities of the Hoechst-stained objects (Ch1), the SCCs from Ch1, the total and average fluorescent intensities of the 2GFP-AR (Ch2) signals in the nucleus (circ) or cytoplasm (ring) regions as an overall well average value, or on an individual cell basis.31 The selected object counts from Ch1 correspond to the number of cells included in the analysis and provide a useful readout of compound-mediated cytotoxicity. For example, compared to low androgen medium control wells, there were ~30% fewer cells in the images of N3 C4-2-2GFP-AR cells that had been cultured overnight in the presence of 312 nM 17-AAG (Fig. 3A). In these same cell populations, cells that were exposed to 312 nM 17-AAG exhibited ~ a 75% reduction in the mean “circ” average intensity of 2GFP-AR in the nucleus compared to low androgen media controls (Fig. 3B), but only ~ a 10% reduction in the mean “ring” average intensity of 2GFP-AR in the cytoplasm (Fig. 3C). The mean circ–ring average intensity difference (MCRAID) (Fig 3D) and the mean circ:ring average intensity ratio (Fig. 3E) readouts both indicated that compared to low androgen medium controls, 17-AAG treatment significantly reduced the nuclear localization of 2GFP-AR. We selected the MCRAID readout as our primary indicator of 2GFP-AR nuclear localization.

To further optimize the 2GFP-AR nuclear localization assay, we evaluated several variables to determine if they would improve the performance of the 2GFP-AR nuclear localization assay (Fig. 4). Consistent with previous findings,35 17-AAG treatment of N3 C4-2-2GFP-AR cells significantly reduced the 2GFP-AR nuclear localization in cells cultured in medium prepared with both charcoal stripped FBS and normal FBS (Fig. 4A). Since the assay signal window between medium controls and 17-AAG-treated cells was slightly larger in the medium prepared with normal FBS, we elected to use normal FBS for all further assay development experiments. In an effort to reduce and/or control for cell loss in the automated HCS protocol, we tested the impact of different types of 384-well plates, including poly-D- and poly-L-lysine-coated plates from different suppliers (Fig. 4B). All of the manually and commercially coated poly-D- and poly-L-lysine plates worked well in the automated 2GFP-AR nuclear localization screening assay, with only the CellBIND plates, which do not have peptide-coated surfaces, exhibiting increased variability and therefore a lower assay signal window (Fig. 4B). We selected commercially available poly-L-lysine plates for all further assay development experiments. Next, we evaluated the effects of the 384-well N3 C4-2-2GFP-AR cell seeding density in the range from 1.25 × 103 to 10 × 103 cells per well on assay performance (Fig. 4C). Although all of the seeding densities produced very comparable results, the variability of the low androgen media controls appeared to be slightly higher with the 5 × 103 and 10 × 103 seeding densities (Fig. 4C). On the basis of these data and to reduce the cell culture burden, we selected a 384-well cell seeding density of 3 × 103 N3 C4-2-2GFP-AR cells per well for the remaining assay development experiments. DMSO has two major effects on cells that could significantly impact the 2GFP-AR nuclear localization screening assay: at DMSO concentrations >1% but <5%, the cell morphology may change from a well-attached spread phenotype to a more rounded smaller less well-attached phenotype; and at DMSO concentrations >5%, there may be significant cell loss due to cytotoxicity and/or loss of attachment.31 Cell rounding and cell shrinking would severely diminish the ability of the MT algorithm to accurately segregate the nuclear and cytoplasm regions of cells, and significant cell loss would reduce the statistical power of the measurements. The 2GFP-AR nuclear localization HCS assay proved to be exquisitely sensitive to DMSO (Fig. 4D) and we selected 0.2% DMSO as the maximum DMSO concentration for the primary screen. In our experience, most cell-based assays do not tolerate DMSO concentrations ≥1.0%. Nuclear translocation and localization HCS assays are perhaps even more susceptible to interference by DMSO because even at concentrations <1%, there may be cell morphology changes, including cell rounding and cell shrinkage. Although we conduct most of our HCS assays at DMSO concentrations ≤0.5%, C4-2 cells may be particularly susceptible to the effects of DMSO on cell rounding and cell shrinkage. We conducted 17-AAG concentration–response assays using the optimized conditions for the 2GFP-AR nuclear localization assay (Fig. 4E). Under these optimized assay conditions, the N3 C4-2-2GFP-AR cells appeared more resistant to 17-AAG with respect to cell loss and also required a concentration of 3 μM to effectively inhibit 2-GFP-AR nuclear localization (Fig. 4E). Therefore, we selected 3 μM 17-AAG for the minimum plate controls for the 2GFP-AR nuclear localization HCS. In a 3-day variability and Z-factor coefficient determination test, the Z-factor coefficients for the 2GFP-AR nuclear localization assay on days 1, 2, and 3 were 0.52, 0.49, and 0.34, respectively (Fig. 4F and Table 2). The signal to background ratios ranged between 4.8-fold and 7.6-fold across all of the 12 × 384-well plates (Table 2), and together these data indicated that the assay was sufficiently robust and reproducible for screening.

Fig. 4.
Development and optimization of the 2GFP-AR nuclear localization HCS assay. (A) Charcoal-stripped versus complete medium. N3 C4-2-2GFP-AR cells were cultured in charcoal-stripped (CS) media and complete (CM) media and treated overnight in the presence ...
Table 2.
GFP-AR Nuclear Localization HCS Assay 3-day Variability and Z-Factor Coefficient Determination

High-Throughput Screening of an NIH Library of 219,055 Compounds

Having developed and optimized the 2GFP-AR nuclear localization HCS assay on the AS-VTI platform, we then proceeded to screen 219,055 small molecules of the NIH MLSCN library at 20 μM (Fig. 5 and Table 3). Figure 5A is a scatter plot of the 2GFP-AR MCRAID raw data of a single representative 384-well assay plate from the primary HCS campaign. The 2GFP-AR MCRAID values of 0.2% DMSO Max controls were ~4.5-fold higher and well separated from the 17-AAG controls, and all 320 compound-treated wells produced signals similar to the Max controls, indicating that they failed to alter or inhibit 2GFP-AR nuclear localization and would therefore be classified as inactive (Fig. 5A). Figure 5B is an overlay scatter plot of the normalized % inhibition data from a single screening operation run of 9,600 compounds that were arrayed and screened at 20 μM in 30 × 384-well assay plates. The normalized 2GFP-AR MCRAID responses of the Max and Min plate controls were reproducible and provided a robust assay signal window (Fig. 5B). The majority of the tested compounds failed to alter or inhibit 2GFP-AR nuclear localization, but 17 compounds achieved the activity criterion of ≥60% inhibition of 2GFP-AR MCRAID values (Fig. 5B). A cutoff of 60% inhibition was selected to ensure that the HCS active rate was ≤0.5% after cytotoxic or fluorescent outlier compounds had been excluded (Table 3). A frequency distribution plot of the normalized % inhibition data for the compound-treated wells for the 219,055 compound primary HCS campaign illustrates that 99.55% of the compounds exhibited 2GFP-AR MCRAID responses comparable to Max controls, and they were therefore classified as inactive (Fig. 5C and Table 3). The % inhibition data presented in Figure 5C were before the removal of any cytotoxic or fluorescent outlier compounds. To identify and eliminate cytotoxic and autofluorescent compounds that might be scored as false positives or could interfere with the 2GFP-AR nuclear localization assay format, we calculated z-scores for several of the parameter outputs by the MT image analysis algorithm (Table 3), as described previously.31 We found a number of compounds (828, 0.38%) that were either cytotoxic or reduced N3 C4-2-2GFP-AR cell adherence sufficiently at 20 μM that they exhibited average cell counts per image z-scores ≤−4 (Table 3). Compounds that significantly increased the mean average Hoechst-stained nuclear total and/or average intensity values in Ch1 and produced z-scores >4 (2,591, 1.8%) were also eliminated (Table 3), because these compounds fluoresce strongly in Ch1 and interfere with the ability of the MT image analysis algorithm to accurately create the nuclear mask that is used to create the nuclear and cytoplasm regions in Ch2 images. In general, compounds with the mean average Hoechst-stained nuclear total and/or average intensity z-scores <−4 were also flagged for cytotoxicity. A small number of compounds (150, 0.07%) produced z-scores >4 for the Ch2 2GFP-AR total and average “ring” (cytoplasm) or “circ” (nucleus) intensity values, respectively (Table 3). Compounds that fluoresce brightly within cells in Ch2 have the potential to obscure any compound effect on 2GFP-AR nuclear localization. A total of 980 (0.45%) actives remained after we had eliminated cytotoxic and autofluorescent compounds, which interfered with the 2GFP-AR nuclear localization HCS assay format (Table 3). Nearly two hundred (182, 18.6%) of the cherry-picked samples were confirmed active because they reproducibly inhibited 2GFP-AR nuclear localization by ≥50% at 20 μM (n = 3) (Table 3). All of the confirmed actives were then tested in a 10-point twofold dilution series of concentration-dependent inhibition of 2GFP-AR nuclear localization (IC50) assays starting at a maximum concentration of 40 μM. Of these compounds, 163 (89.6%) produced calculable IC50s < 40 μM for inhibition of 2GFP-AR nuclear localization (Table 2).

Fig. 5.
Primary HCS to identify inhibitors of 2GFP-AR nuclear localization. (A) Scatterplot of the mean circ–ring average intensity difference (MCRAID-Ch2) intensity data from a single representative HCS assay plate. The MCRAID-Ch2 data from a single ...
Table 3.
GFP-AR Nuclear Localization HCS Campaign Summary

Validation of Selected Hit Compounds

The candidate small molecules identified in the HCS campaign were evaluated for potential biological promiscuity, which could limit their potential for drug development. Small molecules with promiscuous biological activity identified in a cross target query of the PubChem database often interfere with the HTS assay format36 or with multiple signaling pathways and therefore were not considered further. Any compound with <95% purity and IC50 > 10 μM was also eliminated. To further prioritize the remaining hits, we classified and clustered their chemical structures, applied computational filters (PAINS/REOS) to identify and eliminate nuisance compounds and to predict their drug-like characteristics and potentially adverse ADME/Tox properties,37,38 and considered their chemical tractability. A total of 23 small-molecule candidates were selected for further analysis. After overnight exposure, some of the small molecules appeared to reduce the overall expression levels of 2GFP-AR, rather than enhancing the cytoplasmic localization of the 2GFP-AR in C4-2 cells, and a few of the compounds exhibited significant cytotoxicity (data not shown). Two structurally related compounds (SID 14730725 and SID 14742211) inhibited the nuclear localization of 2GFP-AR in C4-2 cells (Fig. 6A–D). The concentration responses for inhibition of 2GFP-AR nuclear localization by SID 14730725 and SID 14742211 are presented in Figure 6B. In the presence of ≥10 μM concentrations of these two compounds, the subcellular localization of 2GFP-AR was predominantly shifted into the cytoplasm region of C4-2 cells (Fig. 6C, D). SID 3712502 (Fig. 6A) exhibited a significant inhibition of nuclear 2GFP-AR levels in C4-2 cells without increasing the levels of cytoplasmic 2GFP-AR (Fig. 6E). This suggests that this compound produced a global downregulation of endogenous AR expression that might thereby suppress the activity of AR. To test this hypothesis, we treated C4-2 cells with SID 3712502 at the indicated concentrations for 48 h. The effect of SID 3712502 on AR and PSA protein levels was assessed by Western blotting analysis. AR protein levels were significantly reduced at the 10 and 50 μM concentrations of SID 3712502 (Fig. 6E). In contrast, the concentration-dependent effects of SID 3712502 on PSA levels were biphasic (Fig 6F). However, PSA protein levels were significantly decreased at 2, 10, and 50 μM concentrations of SID 3712502, suggesting that this compound can inhibit the AR activity at 2 μM and cause downregulation of AR protein expression at 10 and 50 μM concentrations. This result indicates that SID 3712502 is also capable of inhibiting AR signaling.

Fig. 6.
Prioritized Hits from the 2GFP-AR nuclear localization HCS campaign (A) selected hits. Chemical structures, PubChem substance identifiers (SID), PubChem compound identifiers (CID), IC50 values, and the results of a cross target query of the PubChem database ...

Discussion

New approaches to identify novel small-molecule inhibitors of AR function are critically important because of the need to inhibit CRPC cells resistant to first- and second-generation antiandrogens such as flutamide, nilutamide, bicalutamide, and the more recently approved enzalutamide. In this study, we describe a high-throughput HCS campaign to identify small molecules capable of inhibiting the level of nuclear AR using C4-2 CRPC cells stably transfected with 2GFP-AR. Using this AR nuclear localization HCS assay, we identified three novel small-molecule hits from a library of 219,055 compounds that were capable of reducing the level of nuclear AR in C4-2 cells. Two of the three small-molecule hits are structurally related and effectively reduced the nuclear localization of 2GFP-AR in a concentration-dependent manner. The third screening hit did not induce the cytoplasmic localization of 2GFP-AR but inhibited the expression levels of AR proteins in C4-2 cells. The high-throughput HCS assay based on C4-2-2GFP-AR subcellular localization performed well in screening and based on our experience, this HCS assay could be used to screen additional small-molecule libraries.

Using the MT image analysis algorithm of the ArrayScan VTI HCS platform, we were able to quantify the relative intensity of 2GFP-AR in the nuclear and cytoplasm regions of the C4-2 CRPC cell line and to calculate the extent of the altered subcellular distribution phenotype produced by exposure to the HSP90 inhibitor 17-AAG. An interesting observation from this quantitative analysis is the finding that 17-AAG inhibited the nuclear intensity of 2GFP-AR without apparently increasing the cytoplasmic intensity of 2GFP-AR (Fig. 3). Since the 17-AAG reduction of average nuclear intensity did not translate into an increase in cytoplasmic intensity of 2GFP-AR in C4-2 cells, the mechanism of 17-AAG inhibition might be primarily mediated through a degradation of nuclear AR rather than inducing the translocation of nuclear AR to the cytoplasm. Although 17-AAG affects diverse signaling pathways, it reproducibly produced a cytoplasmic subcellular distribution phenotype for 2GFP-AR localization in C4-2 cells. Thus, it provided a reliable positive control for the screening assay. However, 17-AAG can cause cell death since it inhibits multiple signaling pathways. It was important to determine an appropriate concentration and adequate length of exposure such that 17AAG reduced 2GFP-AR nuclear localization without inducing significant cell death. Additional studies will be required to clarify the mechanism by which 17-AAG affects the subcellular localization of AR in CRPC cells. It is possible that other small molecules capable of inhibiting androgen-independent nuclear localization of AR may also function through selective degradation of AR in CRPC cells.

The structures of all three small molecules are very different from androgens or the known AR antagonists. Therefore, the mechanisms of their action are likely to be very different from the existing AR antagonists, which target the LBD of AR. AR inhibition by the three small-molecule hits could be mediated through direct mechanisms targeting AR or by indirect mechanisms such as targeting an AR cofactor or the expression of an AR cofactor. A determination of the mechanism(s) of action of the small-molecule hits could provide new insights into how AR subcellular localization and/or turnover are regulated in CRPC cells. This should be an important future research direction.

In summary, we have developed a high-throughput HCS assay for small molecules capable of reducing the level of nuclear 2GFP-AR in C4-2 cells. We have also identified three small molecules from two structural clusters as inhibitors of 2GFP-AR nuclear localization and/or AR function in CRPC cells. Further characterization of these screening hits may lead to new AR antagonists with the potential to treat CRPC patients.

Author Contribution

All authors have contributed to the manuscript and approved the final version of the manuscript.

Abbreviations Used

17-AAG
17-allylaminogeldanamycin
2GFP-AR
GFP-AR-GFP
AR
androgen receptor
AS-VTI
ArrayScan VTI automated HCS platform
Ch1
channel 1
Ch2
channel 2
Circ
nucleus region
CRPC
castration-resistant prostate cancer
DMSO
dimethyl sulfoxide
FBS
fetal bovine serum
GFP
green fluorescent protein
HCS
high-content screening
LBD
ligand-binding domain
MCRAID-Ch2
mean circ–ring average intensity difference in channel 2
MLSCN
molecular library screening center network
MT
molecular translocation
NIH
National Institutes of Health
NL1
nuclear localization signal 1
NL2
nuclear localization signal 2
PBS
phosphate-buffered saline
PSA
prostate-specific antigen
Ring
cytoplasm region
SCC
selected object or cell count

Disclosure Statement

No competing financial interests exist.

References

1. Siegel R., Naishadham D., Jemal A.: Cancer statistics, 2013. CA Cancer J Clin 2013;63:11–30 [PubMed]
2. Visakorpi T., Hyytinen E., Koivisto P, et al. : In vivo amplification of the androgen receptor gene and progression of human prostate cancer. Nat Genet 1995;9:401–406 [PubMed]
3. Ford OH., 3rd, Gregory CW., Kim D., Smitherman AB., Mohler JL.: Androgen receptor gene amplification and protein expression in recurrent prostate cancer. J Urol 2003;170:1817–1821 [PubMed]
4. Brown RS., Edwards J., Dogan A, et al. : Amplification of the androgen receptor gene in bone metastases from hormone-refractory prostate cancer. J Pathol 2002;198:237–244 [PubMed]
5. Veldscholte J., Berrevoets C., Ris-Stalpers C, et al. : The androgen receptor in LNCaP cells contains a mutation in the ligand binding domain which affects steroid binding characteristics and response to antiandrogens. [Review]. J Steroid Biochem Mol Biol 1992;41:665–669 [PubMed]
6. Gregory CW., Johnson RT., Jr, Mohler JL., French FS., Wilson EM.: Androgen receptor stabilization in recurrent prostate cancer is associated with hypersensitivity to low androgen. Cancer Res 2001;61:2892–2898 [PubMed]
7. Mohler JL.: Castration-recurrent prostate cancer is not androgen-independent. Adv Exp Med Biol 2008;617:223–234 [PubMed]
8. Titus MA., Schell MJ., Lih FB., Tomer KB., Mohler JL.: Testosterone and dihydrotestosterone tissue levels in recurrent prostate cancer. Clin Cancer Res 2005;11:4653–4657 [PubMed]
9. Chen CD., Welsbie DS., Tran C, et al. : Molecular determinants of resistance to antiandrogen therapy. Nat Med 2004;10:33–39 [PubMed]
10. Zegarra-Moro OL., Schmidt LJ., Huang H., Tindall DJ.: Disruption of androgen receptor function inhibits proliferation of androgen-refractory prostate cancer cells. Cancer Res 2002;62:1008–1013 [PubMed]
11. de Bono JS., Logothetis CJ., Molina A, et al. : Abiraterone and increased survival in metastatic prostate cancer. N Engl J Med 2011;364:1995–2005 [PMC free article] [PubMed]
12. Salem M., Garcia JA.: Abiraterone acetate, a novel adrenal inhibitor in metastatic castration-resistant prostate cancer. Curr Oncol Rep 2011;13:92–96 [PubMed]
13. Tran C., Ouk S., Clegg NJ, et al. : Development of a second-generation antiandrogen for treatment of advanced prostate cancer. Science 2009;324:787–790 [PMC free article] [PubMed]
14. Scher HI., Fizazi K., Saad F, et al. : Increased survival with enzalutamide in prostate cancer after chemotherapy. N Engl J Med 2012;367:1187–1197 [PubMed]
15. Beer TM., Armstrong AJ., Rathkopf DE, et al. : Enzalutamide in metastatic prostate cancer before chemotherapy. N Engl J Med 2014;371:424–433 [PMC free article] [PubMed]
16. Ryan CJ., Smith MR., de Bono JS, et al. : Abiraterone in metastatic prostate cancer without previous chemotherapy. N Engl J Med 2013;368:138–148 [PMC free article] [PubMed]
17. Dehm SM., Schmidt LJ., Heemers HV., Vessella RL., Tindall DJ.: Splicing of a novel androgen receptor exon generates a constitutively active androgen receptor that mediates prostate cancer therapy resistance. Cancer Res 2008;68:5469–5477 [PMC free article] [PubMed]
18. Martin SK., Banuelos CA., Sadar MD., Kyprianou N.: N-terminal targeting of androgen receptor variant enhances response of castration resistant prostate cancer to taxane chemotherapy. Mol Oncol 2015;9:628–639 [PMC free article] [PubMed]
19. Liu C., Lou W., Zhu Y, et al. : Intracrine androgens and AKR1C3 activation confer resistance to enzalutamide in prostate cancer. Cancer Res 2015;75:1413–1422 [PMC free article] [PubMed]
20. Zhou Z., Wong C., Sar M., Wilson E.: The androgen receptor: an overview. [Review]. Recent Prog Horm Res 1994;49:249–274 [PubMed]
21. Zhou Z., Sar M., Simental J., Lane M., Wilson E.: A ligand-dependent bipartite nuclear targeting signal in the human androgen receptor. Requirement for the DNA-binding domain and modulation by NH2-terminal and carboxyl-terminal sequences. J Biol Chem 1994;269:13115–13123 [PubMed]
22. Jenster G., Trapman J., Brinkmann AO.: Nuclear import of the human androgen receptor. Biochem J 1993;293:761–768 [PubMed]
23. Picard D., Yamamoto KR.: Two signals mediate hormone-dependent nuclear localization of the glucocorticoid receptor. EMBO J 1987;6:3333–3340 [PubMed]
24. Saporita AJ., Zhang Q., Navai N, et al. : Identification and characterization of a ligand-regulated nuclear export signal in androgen receptor. J Biol Chem 2003;278:41998–42005 [PubMed]
25. Dar JA., Eisermann K., Masoodi KZ, et al. : N-terminal domain of the androgen receptor contains a region that can promote cytoplasmic localization. J Steroid Biochem Mol Biol 2014;139:16–24 [PMC free article] [PubMed]
26. Dar JA., Masoodi KZ., Eisermann K, et al. : The N-terminal domain of the androgen receptor drives its nuclear localization in castration-resistant prostate cancer cells. J Steroid Biochem Mol Biol S2014;143:473–480 [PMC free article] [PubMed]
27. Georget V., Lobaccaro JM., Terouanne B., Mangeat P., Nicolas JC., Sultan C.: Trafficking of the androgen receptor in living cells with fused green fluorescent protein-androgen receptor. Mol Cell Endocrinol 1997;129:17–26 [PubMed]
28. Zhang L., Johnson M., Le KH, et al. : Interrogating androgen receptor function in recurrent prostate cancer. Cancer Res 2003;63:4552–4560 [PubMed]
29. Thalmann GN., Anezinis PE., Chang SM, et al. : Androgen-independent cancer progression and bone metastasis in the LNCaP model of human prostate cancer. Cancer Res 1994;54:2577–2581 [PubMed]
30. Dudgeon DD., Shinde S., Hua Y, et al. : Implementation of a 220,000-compound HCS campaign to identify disruptors of the interaction between p53 and hDM2 and characterization of the confirmed hits. J Biomol Screen 2010;15:766–782 [PubMed]
31. Johnston PA., Shinde SN., Hua Y., Shun TY., Lazo JS., Day BW.: Development and validation of a high-content screening assay to identify inhibitors of cytoplasmic dynein-mediated transport of glucocorticoid receptor to the nucleus. Assay Drug Dev Technol 2012;10:432–456 [PMC free article] [PubMed]
32. Hua Y., Shun TY., Strock CJ., Johnston PA.: High-content positional biosensor screening assay for compounds to prevent or disrupt androgen receptor and transcriptional intermediary factor 2 protein-protein interactions. Assay Drug Dev Technol 2014;12:395–418 [PMC free article] [PubMed]
33. Zhang JH., Chung TD., Oldenburg KR.: A simple statistical parameter for use in evaluation and validation of high throughput screening assays. J Biomol Screen 1999;4:67–73 [PubMed]
34. Shun TY., Lazo JS., Sharlow ER., Johnston PA.: Identifying actives from HTS data sets: practical approaches for the selection of an appropriate HTS data-processing method and quality control review. J Biomol Screen 2011;16:1–14 [PubMed]
35. Saporita AJ., Ai J., Wang Z.: The Hsp90 inhibitor, 17-AAG, prevents the ligand-independent nuclear localization of androgen receptor in refractory prostate cancer cells. Prostate 2007;67:509–520 [PMC free article] [PubMed]
36. Johnston P.: Redox cycling compounds generate H2O2 in HTS buffers containing strong reducing reagents—real hits or promiscuous artifacts? Curr Opin Chem Biol 2011;15:174–182 [PMC free article] [PubMed]
37. Baell JB., Holloway GA.: New substructure filters for removal of pan assay interference compounds (PAINS) from screening libraries and for their exclusion in bioassays. J Med Chem 2010;53:2719–2740 [PubMed]
38. Huggins DJ., Venkitaraman AR., Spring DR.: Rational methods for the selection of diverse screening compounds. ACS Chem Biol 2011;6:208–217 [PMC free article] [PubMed]

Articles from Assay and Drug Development Technologies are provided here courtesy of Mary Ann Liebert, Inc.