Taxane-Induced Blockade to Nuclear Accumulation of the Androgen Receptor Predicts Clinical Responses in Metastatic Prostate Cancer

Cell culture and antibodies

The prostate cancer cell lines LNCaP, PC3 and PC3/AR were derived and maintained as previously described [21–23]. The PC3/mCherry-tubulin cell line was generated in our laboratory by stably transfecting PC3 prostate cancer cells with pcDNA3:mCherry-tubulin plasmid which was a generous gift from the laboratory of Dr. Tsien (UCSD, CA) and selecting transfected cells with 200 μg/ml neomycin (G418). The following antibodies were used for immunofluorescence staining and western blotting: rabbit polyclonal anti-AR (N-20) and goat polyclonal anti-PSA (C-19) from Santa Cruz Biotechnology, Inc (Santa Cruz, CA), mouse monoclonal anti-prostate-specific membrane antigen (PSMA) J591 [24, 25], rat monoclonal anti α-tubulin was from Novus Biologicals (Littleton, CO), mouse monoclonal anti-c-myc was from Oncogene Research Products (Darmstadt, Germany), mouse monoclonal antibody against dynein intermediate chain (IC74) was from Covance (Emeryville, CA). All the Alexa conjugated secondary antibodies were from Molecular probes (Eugene, OR). NLP-005 Methyltrienolone (R1881) was purchased from Perkin Elmer (Boston, MA). Paclitaxel and docetaxel were from Sigma.

Determination of secreted PSA levels

LNCaP cells were cultured on 6 well plates and treated overnight with either docetaxel at the indicated concentrations or the di-hydroxytestosterone (DHT) analog, R1881, at 10 nM for 1 hr or the combination of R1881 treatment following incubation with docetaxel. Supernatant were diluted 1:10 with Abbott Laboratories Free and Total PSA Specimen Diluent. Secreted PSA level measurements were made on Abbott Diagnostics IMx MEIA system (Abbott Park, IL) as previously described [18].

Measurement of PSA-luciferase activity

LNCaP cells were transiently co-transfected with a dual-luciferase reporter assay system in which luciferase is under the control of a promoter containing androgen-responsive elements (ARE-Luciferase) from the prostate-specific antigen (PSA) gene (Promega Corp., Madison, WI) [26] and pRL-TK-luc, Renilla luciferase reporter construct (kindly provided by P. Vertino, Emory University, Atlanta, GA), upon reaching 60% confluency on 6 well plates. Thirty hours post-transfection cells were incubated overnight with either DMSO (vehicle control) or taxanes (paclitaxel or docetaxel) at the indicated concentrations followed by 1 hr treatment with R1881 at either 1nM or 10nM concentration. Cells were harvested and cell lysates were prepared for luciferase assays. Each transfection experiment was performed in triplicate. Results represent an average of at least three independent biological repeats with data presented as relative PSA luciferase activity normalized to Renilla luciferase values.

Establishment of 1A9 cancer cell lines overexpressing AR

The parental ovarian cancer cells 1A9 and their derived beta-tubulin mutant, paclitaxel-insensitive clone PTX10 [27] were transfected with a pFLAG-hAR plasmid using lipofectamine (Invitrogen) following the manufacturer’s instructions. Cells were selected using G418 (300 ug/ml) and AR-expressing clones (as verified by Western Blot analysis) were named 1A9/AR and PTX10/AR cells, respectively. To evaluate AR trafficking to the nucleus, 1A9/AR and PTX10/AR cells were plated on Cell-tak-coated coverslips in RPMI 1640 containing 10% FCS and switched to medium containing 10% charcoal stripped serum (CS) for 72 hours. Following treatments without (control) or with 1) DHT (100 nM) for 2 hours; or 2) PTX (100 nM) for 2 hrs, followed by DHT (100 nM) for 2 hours, cells were fixed with PHEMO buffer [16] and immunostained using antibodies against AR (PG21, Millipore, 1:200) and alpha tubulin (1:1000) followed by Alexa 647 (1:1000) and Alexa 568 (1:500) secondary antibodies and DAPI staining.

Western blotting and immunoprecipitation

Control untreated and treated cells were lysed in TNES buffer containing 50 mM Tris (pH 7.5) 100 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, and a 1X protease inhibitor mixture (Roche Applied Science). For the immunoprecipitation experiments, 0.5 mg of soluble cell extract was immunoprecipitated with either a rat α-tubulin or a mouse antibody directed against dynein intermediate chain (IC74) and their respective IgG controls, using protein-G plus agarose (Calbiochem, Darmstadt, Germany) as recommended by the manufacturer. Immunoprecipitated proteins and 50 μg of total cell extracts were resolved by 10% SDS-PAGE and immunoblotted for the indicated proteins. Immunoblots were analyzed with the Odyssey infrared imaging system (LI-COR, Lincoln, NE).

Dynamitin overexpression

For the dynamitin overexpression experiments, LNCaP and PC3-AR cells were plated on 12mm glass coverslips (Electron Microscopy Sciences, Hatfield, PA) and transiently transfected with c-myc-tagged pCMVH50m plasmid containing dynamitin, a kind gift from R. Vallee (Columbia University, New York, NY) [28]. All transfections were performed using FuGENE 6 transfection reagent (Roche Applied Science, Indiananpolis, IN) according to the manufacturer’s recommendations. C-myc-dynamitin transfected cells were processed for immunofluorescence labeling using the following primary antibodies: anti-AR rabbit polyclonal and an anti-c-Myc mouse monoclonal. The secondary antibodies used were Alexa 488-conjugated goat anti-rabbit immunoglobulin G (IgG; 1:500) and Alexa 568-conjugated goat anti-mouse IgG (1:500). Cells were then analyzed by confocal microscopy.

Microtubule co-sedimentation assay

LNCaP cells were lysed in low salt buffer (LSB) (20 mM Tris, 1 mM MgCl₂, 2 mM EGTA, 0.5% NP-40 and 1X cocktail protease inhibitors). 20 μl of MAP-rich bovine brain tubulin (5 mg/ml) (Cytoskeleton, Denver, CO) was supplemented with 1 mM GTP and 2.5 μl cushion buffer consisting of PEM buffer (80 mM PIPES, 1 mM MgCl₂, 1 mM EGTA ) in 50% glycerol and allow to polymerized at 35°C for 20 min. Microtubules were diluted in 200μl warm PEM plus 20 μM paclitaxel (PTX) and kept at room temperature. Alternatively, MAP-rich tubulin was kept on ice and supplemented with PEM buffer plus 20 μM colchicine to prevent microtubule polymerization. 20 μl of either microtubules or tubulin dimers were incubated with 50 μg of LNCaP total cell extracts for 30 min at 35°C or 4°C respectively. Samples were loaded into 100 μl cushion buffer and microtubules were pelleted by centrifugation at 100,000g at room temperature in an airfuge (Beckman Coulter, Inc., Fullerton, CA). The warm supernatant (WS) or cold supernatant (CS) were separated and the warm (WP) or cold pellet (CP) were resuspended in an equal volume of PEM buffer. 35 μl from each fraction were loaded in a 10% SDS-PAGE and western blotting was performed with antibodies against AR (1:200 dilution) and α-tubulin (clone YL1/2, 1:5000 dilution).

Live cell recording of AR nuclear translocation

For the analysis of GFP-AR trafficking in PC3-mCherry-tubulin cells, PC3-mcherry-tubulin cells were plated on 35mm MatTek dishes and microinjected with full-length GFP-AR. The cells were grown in complete RPMI media and switched to phenol-red free media supplemented with 5% charcoal-stripped medium for up to 3 hours before imaging. Cells were transferred to a live-cell chamber and maintained at 37°C and 5% CO2 using the Tokai Hit Temperature control unit INUG2-ZILCS. The AR ligand R1881 was added to the cells using a Tokait Hit perfusion pipe/tube set with Luer lock syringe. Live cell imaging was performed by acquiring 0.5 μm Z-sections through the entire cell depth at each time point using a Zeiss confocal microscope fitted with a Spinning Disk unit manufactured by Yokogawa Electric Corporation and run with the Axiovision software. Images were acquired using the 488 laser to monitor GFP-AR subcellular localization. Simultaneous imaging of both the GFP-AR and mCherry-tubulin fluorophores was not possible due to the low photostability of the mCherry fluorophore. Therefore, mCherry-tubulin was imaged once at the end of each recording to reveal the pattern of the microtubule cytoskeleton. Representative fields were selected and images (z-stacks) were acquired at 10 min intervals for a total of 2 hr. Maximum intensity projections were made with the Axiovision software, AxioVision files were saved as 16-bit images and exported to MetaMorph image analysis software. Background subtraction was performed using the mean pixel value of a noncell region on each image with the statistical correction function in MetaMorph and a sum projection was assembled from the images recorded at 10 min intervals. High-intensity pixels in noncellular regions or regions not of interest were also scaled out. Nuclear versus cytoplasmic AR values were obtained using integrated pixel intensity values from the sum projection and the percentage of nuclear AR was calculated using the following formula: % Nuclear AR = 100*Nuclear AR/Total AR.

Venous blood collection and isolation of circulating tumor cells (CTCs)

Following approval of the Weill Cornell Institutional Review Board (IRB), 10 mls of peripheral blood was collected, in sodium citrate tubes (BD, Franklin Lakes, NJ), from CRPC patients at various timepoints during taxane chemotherapy. CTCs were then isolated using the epithelial cell adhesion molecule (EpCAM)-based immunomagnetic capture CellSearch system as per the manufacturer’s instructions [29]. This enrichment step greatly facilitated our analysis as CTCs are very rare (1 in 10⁶ PBMCs) and require specialized technology for their detection. CellSearch has been the first and only FDA-approved CTC detection technology utilized for clinical prognosis. Following the isolation/enrichment step with the CellTracks AutoPrep System^®, captured CTCs were cytospun onto Cell-Tak-(BD Biosciences, San Jose, CA) -coated coverslips and processed for immunofluorescence staining. For our analysis, we performed only the first step of enrichment and omitted the second step of automated staining with cytokeratins, DAPI and CD45 which is used for CTC enumeration. In addition, to ensure analysis of PC CTCs among the leucocytes present in the EpCAM-enriched cell population, we stained for PSMA (prostate specific marker) and CD45 (leucocyte specific marker) and analyzed the PSMA+/CD45- cells for AR subcellular localization using multiplex confocal microscopy. Using this methodology we monitored longitudinally CRPC patients receiving taxane-chemotherapy and investigated the association between AR subcellular localization in CTCs and patients’ clinical response to therapy.

In parallel, CTCs were also identified following Ficoll separation of blood collected in Vacutainer Cell Preparation Tube with Sodium Heparin (BD, Franklin Lakes, NJ). The PBMC layer containing CTCs was collected and washed in 1X PBS. Hemolysis was performed using ammonium chloride buffer (150 mM NH₄Cl, 10 mM NaHCO₃, 1 mM Na₂EDTA, pH adjusted to 7.3). Cells were counted by Bright-Line hemacytometer (Hausser Scientific, Horshan, PA) and approximately 1 × 10⁶ cells were plated on cell-tak coated 8-mm glass coverslips (Electron Microscopy Sciences, Hatfield, PA) in 48 well plates, cytospun and allowed to attach. Subsequent to attachment, cells were fixed and processed for immunofluorescence and confocal microscopy as described below.

Immunofluorescence and confocal microscopy

Following treatment and attachment, the cells, including PBMCs were fixed and processed for immunofluorescence labeling [16]. The following antibodies were used: rabbit polyclonal antiAR, mouse monoclonal anti-PSMA (J591) and rat monoclonal anti α-tubulin. The secondary antibodies used were Alexa 488-conjugated goat anti-rabbit immunoglobulin G (IgG; 1:500), Alexa 568-conjugated goat anti-mouse IgG (1:500) and Alexa 647-conjugated goat anti-rat IgG (1:500). DNA was stained with 13g ml⁻¹ 4, 6-diamidino-2-phenylindole, DAPI (Sigma, St. Louis, MO). The cells were mounted using Mowiol (Calbiochem, Darmstadt, Germany) and imaged using a Zeiss LSM 5 LIVE (Thornwood, NY) confocal microscope using 63x/1.4 Plan APOCHROMAT and 100x/1.4 Plan APOCHROMAT objectives. All images were acquired and analyzed using Zeiss LSM 5 LIVE software.

Quantitation of AR subcellular localization in CTCs

CTCs were obtained by CellSearch enrichment or Ficoll separation for each patient at each time point. CTCs were imaged with a Zeiss 5 LIVE confocal microscope using a 20X objective. A single plane image of each cell was taken and analyzed using the MetaMorph imaging software (Molecular Devices). The nuclear region of the CTC was identified by DAPI staining. Total cellular AR fluorescence intensity was measured and the amount of AR fluorescence inside the nuclear region was determined, allowing us to calculate the percentage of AR in the cytoplasm and the nucleus of each CTC. The total number of CTCs analyzed varied among patients, ranging from 3 to 27 CTCs per patient per time point. When possible we analyzed at least 10 CTCs per patient per time point.

Statistical Analysis

For CTCs isolated from the fourteen patients, we determined both the AR subcellular localization and total AR fluorescence intensity at each patient visit. AR localization was defined as follows: if greater than 50% of CTCs in a given sample have a meaningful percentage of AR in cytoplasm (>40%), we considered AR localization as cytoplasmic, otherwise, we considered AR localization as nuclear. Patients’ data with clear PSA outcomes (progressing or response/stable disease) was included in the analysis. Fisher’s exact test was used to correlate AR localization in the nucleus with the patients’ clinical response to treatment assessed using PSA working group 2 criteria. Considering the longitudinal nature of the data, the association between the AR measurements and the PSA outcome was more rigorously assessed using the generalized linear mixed effects models [30–32]. Effects with p-values <0.05 was considered statistically significant.

Taxanes inhibit AR nuclear accumulation and signaling

Since microtubules are involved in transcription factor trafficking, we investigated the effects of taxane treatment on AR nuclear translocation in LNCaP PC cells. Cells were treated with paclitaxel (PTX) followed by the addition of the di-hydrotestosterone (DHT) analog R1881 (Fig. 1A–C) and analyzed for evidence of PTX-induced microtubule bundling and R1881-induced AR nuclear accumulation by confocal microscopy. Interestingly, PTX induced a significant decrease in AR nuclear accumulation, at both baseline and following R1881 treatment (Fig. 1A–C). Quantitation of the extent of PTX-induced AR nuclear exclusion revealed a significant decrease in the percentage of cells with nuclear AR, following R1881 treatment, from 70% to <30% (Fig. 1B; p<0.001) and a concomitant 50% decrease in the total fluorescence intensity of nuclear AR staining (Fig. 1C; p<0.001). The profound cytoplasmic sequestration of AR following paclitaxel treatment implicated microtubules in the shuttling of the receptor from the cytoplasm to the nucleus. To explore the dynamics of ligand induced AR nuclear translocation, we introduced full-length GFP-AR into PC3 cells stably expressing mCherry tubulin (PC3:mCherry-tub) (Fig. 1D and time lapse confocal microscopy recordings available as movies S1 and S2). Paclitaxel treatment significantly reduced the extent and the rate of AR nuclear accumulation in cells with stabilized microtubules as early as 30 min post R1881 addition (p<0.01-0.001). Consistent with drug-induced AR nuclear exclusion, we also observed dose-dependent inhibition of AR transcriptional activity using a luciferase reporter assay in LNCaP cells expressing endogenous AR (Fig. 2A). Similarly, docetaxel (DTX) treatment inhibited ligand-induced protein expression (Fig. 2B) and secretion of prostate specific antigen (PSA; Fig. 2C), an AR target gene used clinically to monitor biochemical disease progression in men with PC [18, 33]. Together, these results suggested that taxanes inhibit AR signaling by preventing AR nuclear accumulation.

Microtubule stabilization is a prerequisite for taxane-induced inhibition of AR signaling

To determine whether the taxane-induced cytoplasmic sequestration of AR required prior microtubule stabilization, we utilized an isogenic model of paclitaxel- sensitive and -resistant human ovarian cancer cells, 1A9 and 1A9/PTX10, respectively, stably transfected with wild-type GFP-AR. Drug resistance in 1A9/PTX10 cells is conferred by an acquired tubulin mutation (Fβ270V) at paclitaxel’s binding site [27]. Paclitaxel treatment of taxane-resistant 1A9/PTX10 cells had no effect on GFP-AR subcellular localization (Fig. 3, right panel) (consistent with the drug’s inability to stabilize microtubules), whereas DHT-induced GFP-AR nuclear accumulation was robustly inhibited by paclitaxel in parental drug-sensitive 1A9 cells (Fig. 3, left panel). These results demonstrated that the microtubule stabilizing activity of taxanes is required to inhibit AR translocation and downstream signaling; further implying that in cases where the drug-target interaction is impaired AR signaling will remain unaffected. To further characterize the interaction of AR with microtubules, we performed high resolution multiplex confocal microscopy in PC3 cells stably expressing AR, and observed a clear co-localization of cytoplasmic AR with intact microtubules. Following taxane treatment AR remained associated with bundled microtubules in the perinuclear region (Fig. 4A). The tubulin-AR association was confirmed by coimmunoprecipitation of tubulin and AR in LNCaP cells (Fig. 4B) as well as by a microtubule cosedimentation assay [14] in which AR was found preferentially associated with the microtubule polymer fraction (WP) and not the fraction containing soluble tubulin dimers (WS), lending further support to the role of microtubules as tracks for AR transport towards the nucleus (Fig. 4C).

The microtubule-associated motor protein, dynein, mediates AR trafficking

To test this hypothesis, we investigated the involvement of the minus-end-directed microtubule-motor protein dynein, which transports cargo from the cytoplasm towards the nucleus, in AR trafficking [34]. Coimmunoprecipitation experiments revealed that AR associated with dynein, and that this interaction was increased upon ligand (R1881) induced AR nuclear translocation (Fig. 5A). In addition, dynein co-fractionated with AR and microtubules in the co-sedimentation assay (Fig. S1A), suggesting that both dynein and AR associate preferentially with microtubule polymers. Cytoplasmic dynein, in order to drive subcellular motile functions, works in concert with several accessory proteins including dynactin, an adapter that mediates the binding of dynein to cargo structures enhancing the motor’s processivity. Overexpression of the dynactin associated protein dynamitin, which disrupts the dynein-cargo interaction [28], markedly reduced AR nuclear accumulation following R1881 induction (Fig. 5B and Fig. S1B), suggesting that following ligand binding, AR is shuttled from the cytoplasm to the nucleus via microtubules and its associated motor complexes.

AR cytoplasmic localization in CTCs correlates with clinical response to taxane chemotherapy

To assess whether taxanes inhibit AR nuclear translocation in vivo, we isolated CTCs from the blood of CRPC patients and examined PSMA (prostate specific membrane antigen) [24, 25, 35, 36] expressing CTCs by high resolution confocal microscopy. We first confirmed colocalization of AR with the microtubule cytoskeleton in patient samples (Fig. 6A). We next investigated whether perturbation of the microtubule-AR axis in CTCs identified in CRPC patients receiving taxane chemotherapy would correlate with clinical response. As shown in Figure 6B (top panel), CTCs from a patient who was refractory to paclitaxel chemotherapy exhibited an unperturbed microtubule network with AR present in both the nucleus and cytoplasm. In contrast, in a second patient who was responding to docetaxel chemotherapy, CTCs demonstrated bundled microtubules and AR was exclusively in the cytoplasm with intense AR perinuclear staining (Fig. 6B, lower panel). These results in 2 patients suggested that taxane-induced microtubule bundling and AR cytoplasmic localization correlated with clinical response to chemotherapy. We therefore expanded our analysis to additional patients and used CTCs enriched from the blood of CRPC patients using EpCAM-based immunomagnetic capture (CellSearch) [29]. To identify PC CTCs among the leucocytes present in the EpCAM-enriched cell population, we stained for PSMA and CD45 (leucocyte specific marker) and analyzed the PSMA+/CD45− cells for AR subcellular localization using multiplex confocal microscopy (Fig. 6C). Using this methodology, we monitored longitudinally CRPC patients receiving taxane-chemotherapy and investigated the association between AR’s subcellular localization in CTCs and patients’ clinical response to therapy. Figure 6D shows a representative example of CTCs isolated from a metastatic CRPC patient in which AR was localized in the nucleus at baseline, consistent with this patient’s rising serum PSA (Fig. 6E, open symbols), but was consistently shifted to the cytoplasm during PSA decline following chemotherapy (Figs. 6D, 6E). In another patient (Fig. 6E, filled symbols), AR was detected in the cytoplasm as early as 1 hour following paclitaxel administration which was followed by a significant decline in PSA. This result emphasizes fast and effective drug-target engagement in this patient’s sample.

Table 1 summarizes the results from our CTC analyses performed on sequential blood samples obtained from fourteen patients receiving taxane chemotherapy and followed longitudinally. PSA outcomes were assigned by clinicians independent of CTC enumeration using modified Prostate Cancer Working Group 2 criteria [33]. Responders were defined as having at least a 30% reduction in PSA [33, 37, 38]; progressors had >25% increase in PSA with a minimum absolute increase of 2 ng/mL and those meeting neither criteria were considered stable. Among 18 samples obtained during clinical progression, 13 (72%) demonstrated nuclear AR localization; while among 17 responding/stable samples, 12 (70.6%) showed cytoplasmic AR localization (p-value=0.02) (Table 1). We also examined the association between longitudinal quantitative AR measurements (either in terms of the percentage of AR in cytoplasm or the total AR intensity) and the PSA outcome using a generalized linear mixed-effects model that accounts for possible clustering of data among cells from the same sample and among samples from the same subject. This analysis revealed that the odds of responding for those with AR completely in cytoplasm was estimated to be increased by 30% compared to patients with no AR in the cytoplasm (OR=exp(0.26) =1.30, P<0.001). Similarly, using log transformed total AR intensity as the predictor, we found that increased total AR intensity is associated with decreased odds of responding. With each unit increase in the log AR intensity, the odds of responding is 24% lower (OR=exp(−0.28)=0.76, P<0.001). Interestingly, we observed AR nuclear localization in all four baseline patient samples indicative of active AR signaling. In contrast, AR cytoplasmic localization was detected as early as 1 hr following the first dose of paclitaxel administration (patient 4) showcasing rapid and effective perturbation of the microtubule-AR axis. Altogether, these results demonstrate that AR subcellular localization can be resolved in CTCs and can serve as a biomarker to monitor clinical response to chemotherapy.