Cell culture

The identities of all cell lines were authenticated by DNA fingerprinting (Identifier Kit; Applied Biosystems, Grand Island, NY, USA) within 6 months of use. The BCK4 line is an ER+/AR + breast cancer line derived recently from a pleural effusion and has a nearly normal karyotype [25]. For the BCK4 cell line, the patient sample was acquired under a University of Colorado Institutional Review Board-approved tissue-acquisition protocol and patient-informed consent was obtained to acquire blood and tissue for research purposes. All other cell lines were obtained from the American Type Culture Collection; Manassas, Virginia, USA. BCK4 and MCF7 cells were grown in minimum essential media, 5% fetal bovine serum, non-essential amino acids, insulin and penicillin/streptomycin, and ZR75 cells in the same media with the addition of HEPES and l-glutamine. MCF7 cells express a wild-type AR, albeit with a shortened CAG repeat [26] that is often indicative of a more active receptor [27]. T47D cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum, l-glutamine penicillin/streptomycin. LNCaP cells were grown in RPMI, 5% fetal bovine serum and penicillin/streptomycin. All cells were grown in a 37°C incubator with 5% carbon dioxide. MDA-MB-453 and MDA-kb2 (a derivative of MDA-MB-453 stably expressing the AR-dependent MMTV-luciferase reporter gene construct; American Type Culture Collection) were cultured in Leibovitz’s L-15 media (Invitrogen, Carlsbad, CA, USA) containing 10% fetal bovine serum (Invitrogen) and penicillin/streptomycin. MCF7-TGL cells were generated by stable infection with retroviral SFG-NES-TGL vector, encoding a triple fusion of thymidine kinase, green fluorescent protein and luciferase and sorted for green fluorescent protein.

Tumor studies

MCF7 experiments with enzalutamide delivered in rodent chow were performed at the University of Colorado Anschutz Medical Campus and approved by the University of Colorado Institutional Animal Care and Use Committee (IACUC protocol 83611(03)1E). The MDA-MB-453 xenograft experiment in which enzalutamide was delivered by oral gavage was performed by AntiCancer Inc., San Diego, CA, USA and was approved by the Institutional Animal Care and Use Committee of AntiCancer Inc. All animal experiments were conducted in accordance with the National Institutes of Health Guidelines of Care and Use of Laboratory Animals.

For MCF7 xenograft experiments, 10⁶ MCF7-TGL cells that stably express a triple fusion of thymidine kinase, green fluorescent protein and luciferase (SFG-NES-TGL retroviral vector) for in vivo imaging purposes were mixed with Matrigel (BD Biosciences, Franklin Lakes, New Jersey, USA) and injected into the fourth inguinal mammary fat pad of female, ovariectomized athymic nu/nu or nonobese diabetic (NOD)/SCID mice (Taconic, Germantown, NY USA). At time of tumor injection, E2 pellets (60-day release, 1.5 mg/pellet; Innovative Research of America, Sarasota, Florida USA) or the nonaromatizable androgen 5-alpha-dihydrotestosterone (DHT) (8 mg/pellet, packed and sealed in silastic tubing) were implanted subcutaneously at the back of the neck. Tumor burden was assessed using an in vivo imaging system or caliper measurements (tumor volume was calculated as: length × width × depth/2). Once tumors were established, mice were matched into groups based on the total tumor burden as measured by in vivo imaging system or caliper. Groups receiving tamoxifen had a 90-day release, 5 mg/pellet (Innovative Research of America) implanted subcutaneously. Mice were administered enzalutamide in their chow (approximately a 50 mg/kg daily dose) or by oral gavage (10 or 25 mg/kg/day). Enzalutamide was mixed with ground mouse chow (catalog number AIN-76; Research Diets Inc., New Brunswick, NJ, USA) at 0.43 mg/g chow. The feed was irradiated and stored at 4°C before use. Mice in the control group received the same ground mouse chow but without enzalutamide. All mice were given free access to enzalutamide formulated chow or control chow during the entire study period and at an average of 3.5 g/day food intake. Feed was changed in the animal cages twice a week. Water and feed were prepared ad libitum. Two hours prior to sacrifice, mice were injected intraperitoneally with 50 mg/kg bromodeoxyuridine (BrdU; Sigma-Aldrich, St. Louis, MO, USA). Mice were euthanized by carbon dioxide asphyxiation followed by cervical dislocation, and the blood, tumors, colon, uteri and mammary glands were harvested.

For the MDA-453 tumor study, 6 × 10⁶ cells were injected into the fourth inguinal mammary fat pad of NOD-SCID-IL2Rgc^-/- female mice into which a DHT pellet (1.5 mg 60-day release; Innovative Research of America) was implanted subcutaneously. The tumor size was measured using calipers, and when tumors reached 100 mm³ the mice began receiving 10 mg/kg enzalutamide or vehicle by oral gavage. Once the tumors reached 400 mm³, another group was started on 25 mg/kg enzalutamide. At the end of the experiment, tumors were weighed and processed for embedding.

Neoadjuvant endocrine therapy study

The inclusion criteria and trial design are described elsewhere [8, 9]. Briefly, women with ER + breast cancers were enrolled in a randomized phase II clinical trial to receive exemestane alone (25 mg daily) or exemestane in combination with tamoxifen (20 mg daily) for 4 months prior to surgery. Women included in the trial were postmenopausal with newly diagnosed cancers of stage II/III, T2 to T3. Core needle biopsies were taken prior to treatment and tumor pieces from the final excision surgery were taken for analysis. The protocol was approved by the Colorado Multiple Institutional Review Board and informed consent was provided by all patients. The criteria for responders ranged from minor response to complete response, while nonresponders had stable or progressive disease.

Tamoxifen study

This study included 192 female patients diagnosed with ER + breast cancer at the Massachusetts General Hospital (Partners) between 1977 and 1993, who were offered tamoxifen treatment as part of their adjuvant therapy and were followed at the hospital through 1998. Patients were offered tamoxifen based on estrogen positivity (≥10 pmol/mg protein) determined using either a ligand binding assay or a radioactive enzyme-linked immunosorbent assay, the standard protocol in use during this time period. As part of the present study, archival formalin-fixed paraffin-embedded tumors collected under the Institutional Review Board protocol Molecular and Cellular Predictors of Breast Cancer were stained for AR and ER by immunohistochemistry. All slides were evaluated and the percentage and intensity of both AR and ER were recorded. Each slide was also scored using the Allred scoring method.

Contingency tables were used to study the associations between the AR/ER ratio and clinicopathologic variables. In this analysis, each clinicopathologic variable was divided into two or three categories (lymph node negative vs. lymph node positive; lymph node negative vs. one to three positive vs. four or more positive; patient age <50 years vs. ≥50 years; tumor size ≤2 cm vs. >2 cm; grade 1 vs. grade 2 vs. grade 3; PR negative vs. positive; ErbB2 ≤30% vs. >30%, MIB-1 < median vs. ≥median, mitoses/10 high-powered fields (mitotic index) < median vs. ≥median, epidermal growth factor receptor  < median vs. ≥median). Patients were followed from the date of diagnosis to the date of first failure (local recurrence or distant metastasis) as well as the date of death or last follow-up. Patients who died of causes other than breast cancer and patients who were lost to follow-up or whose last encounter was before the end of the study were censored at the date of death or last encounter for survival analyses. The AR:ER ratio was calculated using a manual receiver operator characteristic analysis where we investigated the ratio that produced the best difference between good and poor prognosis in relation to disease-free survival (DFS) to identify the cutoff point for this variable. The final AR:ER ratio cutoff point was determined to be 2.0. A Fisher’s exact test was used for all dichotomized variables and the chi-square test for all trichotomized variables to compare the AR:ER ratio with other predictive markers. Kaplan–Meier curves used the calculated AR:ER ratio. All statistics were calculated using SAS (version 9.3; SAS Institute, Cary, NC, USA). Significance was determined at P <0.05 and all tests were two-sided.

Immunohistochemistry

Slides were deparaffinized in a series of xylenes and ethanols, and antigens were heat retrieved in either 10 mM citrate buffer pH 6.0 (BrdU, Ki67) or 10 mM Tris/1 mM ethylenediamine tetraacetic acid buffer at pH 9.0 (AR, ER, caspase 3). Tissue for BrdU was incubated in 2 N HCl followed by 0.1 M sodium borate following antigen retrieval. Antibodies used were: AR clone 441 and ER clone 1D5 (Dakocytomation, Carpinteria, CA, USA), cleaved caspase 3 (Cell Signaling Technology, Danvers, MA, USA), Ki67 (sc-15402; Santa Cruz, Dallas, TX, USA) and BrdU (BD Biosciences, Franklin Lakes, NJ, USA). Envision horseradish peroxidase (Dakocytomation) was used for antibody detection.

Terminal deoxynucleotidyl transferase dUTP nick end-labeling (TUNEL) staining for apoptosis was performed using the ApopTag Plus Peroxidase In Situ Apoptosis Detection Kit (Millipore, Billerica, MA, USA), as per the manufacturer’s instructions. AR and ER staining was assessed by a pathologist (PJ or ADT) and the score is reported as intensity multiplied by percent positive cells, or in the case of the tamoxifen-treated cohort the Kaplan–Meier curve is based on percent positive cells, although results are similar and still significant when the intensity is multiplied by the percent positive cells. For ER, BrdU and TUNEL staining in xenograft studies, three separate 200× fields of each xenograft tumor were taken using an Olympus BX40 microscope (Center Valley, PA, USA) with a SPOT Insight Mosaic 4.2 camera and software (Diagnostic Instruments, Inc., Sterling Heights, MI, USA). A color threshold (RGB for positive staining nuclei, and HSB for total nuclei) was adjusted manually using ImageJ (National Institutes of Health, Bethesda, MD, USA) for each image, and particles created by the thresholds were analyzed for total area. The RGB area was divided by the HSB area and multiplied by 100 for each image. For analysis of the nuclear androgen receptor, cleaved caspase 3 and Ki67, slides were scanned at 20× on an Aperio Scan ScanScope XT, Leica Microsystems Inc. Buffalo Grove, IL United States. Mammary tumor tissue was traced separately for each tumor and necrotic areas of the tumor removed using a negative pen tool in Aperio’s Scanscope software. A nuclear algorithm was utilized to measure the percent positive cells for the Ki-67-stained and AR-stained slides and the data were exported. Cleaved caspase 3-stained slides were analyzed using a modified positive pixel count algorithm.

Immunoblotting

Whole cell protein extracts (50 μg) were denatured, separated on SDS-PAGE gels and transferred to polyvinylidene fluoride membranes. After blocking in 3% bovine serum albumin in Tris-buffered saline–Tween, membranes were probed overnight at 4°C. Primary antibodies utilized include: ERα (Ab-16, 1:400 dilution; Neomarkers, Fremont, CA USA), AR (PG-21, 1:400 dilution; Millipore (Bedford, Massachusetts USA) or EP6704, 1:10,000; Abcam (San Francisco, CA USA), glyceraldehyde 3-phosphate dehydrogenase (1:20,000 dilution; Sigma, St. Louis, MO USA), Topo 1 (C-21, 1:100 dilution; Santa Cruz) and alpha-tubulin (clone B-5-1-2, 1:30,000 dilution; Sigma). After incubation with appropriate secondary antibody, results were detected using Western Lightning Chemiluminescence Reagent Plus (Perkin Elmer, Waltham Massachusetts USA).

Cellular fractionation

For the MDA-kb2 cellular fractionation, cells were washed with ice-cold Dulbecco’s phosphate-buffered saline, pH 7.4, pelleted using centrifugation and resuspended in 2 volumes of ice-cold NSB (10 mM Tris · Cl, pH 7.4, 10 mM NaCl, 2 mM MgCl₂, 1× protease inhibitors). The volume was adjusted with ice-cold NSB to 15 times the initial volume and incubated for 30 minutes on ice. The cytoplasmic fraction was obtained by addition of NP-40 to a final concentration of 0.3%. Nuclei and cytoplasm were separated using a 0.4 mm clearance Dounce homogenizer. After centrifugation, the supernatant containing the cytoplasmic fraction was collected. The pellet containing the nuclear fraction was resuspended in a 250 mM sucrose solution containing 10 mM MgCl₂ and then 1 volume was added to 880 mM sucrose containing 5 mM MgCl₂ under the nuclear fraction. The nuclei were then purified by centrifugation through the sucrose cushion. For the MCF7s cells, cellular fractionation was performed using the NE-PER Nuclear and Cytoplasmic Extraction Kit, Pierce Biotechnology, Rockford, IL USA as per the manufacturer’s instructions.

Nuclear translocation assay

MDA-kb2 cells were seeded at 2 × 10³ cells/cm² in optical microplates in Leibovitz’s L-15 medium supplemented with 5% charcoal-stripped serum. After 3 days of cultivation the cells were pretreated with enzalutamide (1 or 10 μM) for 2 hours and then co-treated with 1 nM DHT for 1 hour in the presence of enzalutamide (total 3 hours of treatment with enzalutamide). The cells were washed with phosphate-buffered saline, fixed with 4% formaldehyde for 30 minutes at room temperature and permeabilized with 0.2% triton X-100. Samples were then blocked with 5% bovine serum albumin for 1 hour and incubated with an antibody against AR (N20, sc-815 1:100; Santa Cruz) in phosphate-buffered saline 0.1% triton overnight. Incubation with the secondary antibody anti-rabbit Alexa Fluor 488 (1:1,000) was performed in 2.5% bovine serum albumin for 2 hours at ambient temperature. The nuclei were stained with 4',6-diamidino-2-phenylindole (1 μg/ml) for 30 minutes. Cells were visualized with a 60× objective and a Qimaging digital camera coupled to an Olympus X71 fluorescence microscope using a yellow fluorescent protein filter (Chroma U-N31040; Center Valley, PA, USA). The nuclear distribution of AR (ratio of nuclear AR signal/total AR signal) was quantified in a minimum of 48 cells using ImageJ software (National Institutes of Health).

Real-time quantitative polymerase chain reaction

cDNA was synthesized from 1 μg total RNA, using M-Mulv reverse transcriptase enzyme (Promega, Fitchburg, WI, USA). For FASN, PRLR and GCDFP-15, SYBR green quantitative gene expression analysis was performed using the following primers: FASN forward, 5′-AAGGACCTGTCTGGATTTGATGC-3′ and FASN reverse, 5′-TGGCTTCATAGGTGACTTCCA-3′; PRLR forward, 5′-TATTCACTGACTTACCACAGGGA-3′ and PRLR reverse, 5′-CCCATCTGGTTAGTGGCATTGA-3′; GCDFP-15 forward, 5′-TCCCAAGTCAGTACGTCCAAA-3′ and GCDFP-15 reverse, 5′-CTGTTGGTGTAAAAGTCCCAG-3′; and 18S forward, 5′-TTGACGGAAGGGCACCACCAG-3′ and 18S reverse, 5′-GCACCACCACCCACGGAATCG-3′. For PR and stromal cell-derived factor 1 (SDF-1, also known as CXCL12), Taqman real-time polymerase chain reaction was performed using validated primer/probe sets from Applied Biosystems (assay ID: PR Hs01556702_m1, SDF-1 Hs00171022_m1, 18S Hs99999901_s1). Relative gene expression calculated using the comparative cycle threshold method and values were normalized to 18S.

Luciferase assays

MDA-kb2 cells were plated at 5 × 10³ cells/well in 96-well luminescence plates and incubated overnight. Cells were treated with 10-fold serial dilutions of enzalutamide (10, 1, 0.1 μM) and DHT (10, 1, 0.1, 0.01, 0.001 nM) that were prepared in dimethylsulfoxide. Following 24 hours of incubation, the luminescence levels were determined with the luciferase assay system (Promega). Three independent experiments were performed and the luminescence values were determined as relative units and normalized to vehicle. Values are expressed as the mean fold induction ± standard error.

A new method to examine AR relative to ER

To test the significance of AR and ER expression in breast cancer, we examined primary tumors from a group of tamoxifen-treated patients with clinical outcome data. This study included a cohort of 192 female patients diagnosed with breast cancer at the Massachusetts General Hospital (Partners) between 1977 and 1993, treated with adjuvant tamoxifen and followed at the hospital through 1998 under Institutional Review Board approval. The women ranged in age from 20 to 91 years at the time of cancer diagnosis with a median age of 68 years. Forty-eight (25.0%) of the women failed tamoxifen therapy. Women who relapsed while on tamoxifen were generally younger (median 64 years vs. 70 years for nonfailures, P = 0.007), had larger tumors (median 2.6 vs. 1.9 cm³; P = 0.003), had a higher proportion of grade 3 tumors (45.8% vs. 29.4%; P = 0.034), had more positive lymph nodes (median 2 vs. 1; P = 0.006), had a higher mitotic index (median 5 vs. 4; P = 0.007), and had lower levels of PR staining (median 5% vs. 45%, P = 0.048). There were no differences in MIB-1, HER2, or epidermal growth factor receptor staining percentages between the two groups. Women who failed had a median ER percent cells positive of 62.5%. This was significantly lower than the 92.5% percent cells positive in tumors that did not fail (P = 0.001). Although the AR percent cells positive was higher in tumors of women who failed (70% vs. 57.5% for nonfailures), the difference in AR staining percentage did not reach statistical significance.

Since we had previously observed that AR mRNA and protein decrease with treatment in tumors responsive to neoadjuvant endocrine therapy, but did not decrease in nonresponsive tumors [8, 9] (Figure S1, left in Additional file 1), we decided to examine nuclear AR as compared with ER. The median AR:ER ratio in pretreatment biopsies of responsive tumors (Figure S1A in Additional file 1) in the neoadjuvant study was 1.00, with a statistically significant positive correlation between AR and ER expression (P = 0.006) (Figure S1A in Additional file 1). However, in nonresponsive tumors (Figure S1B in Additional file 1), the median AR:ER ratio was 3.79 with no significant correlation between AR and ER. Interestingly, in adjacent uninvolved epithelium (Figure S1C in Additional file 1), the median ratio of AR to ER expression was 0.94, again with a significant positive correlation between the two receptors (P = 0.0003).

Based on these intriguing results in the small neoadjuvant study, we decided to examine the amount of AR relative to ER in the larger cohort of 192 female patients diagnosed with ER + breast cancer that received adjuvant tamoxifen therapy. To identify the best cutoff point for separating patients into good and poor survival, a manual receiver operator characteristic analysis based on time to first failure (disease-free interval, DFS) was performed for the AR:ER ratio – and the optimal cutoff point of 2.0 was determined. In addition, since the AR:ER ratio was not in a log-linear relationship with the hazard function, it was necessary to use the dichotomized variable in the Cox proportional hazard models. Both AR percent cell staining and ER percent cell staining contribute to the AR:ER ratio. AR showed strong positive correlation (r = 0.86, P < 0.0001) with the ratio, while ER showed moderate negative correlation (r = -0.36, P < 0.0001). The AR:ER ratio with a cutoff value of 2.0 was significantly different between the two groups (failed tamoxifen versus nonfailed), with 27.1% of women who failed having an AR:ER ratio >2.0 compared with only 6.3% of nonfailures (P < 0.0001).

High AR:ER ratio indicates poor response to traditional endocrine therapy and overall survival

We then compared study variables with tamoxifen failure by 5 years, and overall DFS and overall disease-specific survival (DSS). By univariate analyses, the tumor size, ER percent staining and AR:ER ratio were significantly associated with all survival outcomes (Table 2), while nodal positivity was significant only for tamoxifen failure and DFS. Notably, the AR:ER ratio was the most significant marker of poor survival (hazard ratio (HR) = 4.43 for tamoxifen failure, P < 0.0001; HR = 4.40 for DFS, P < 0.0001; and HR = 3.66 for DSS, P < 0.0001). In contrast, the ER percent cell staining was associated with reduced risk (HR = 0.98 for tamoxifen failure P < 0.0002; HR = 0.99 for DFS, P < 0.0004; and HR = 0.99 for DSS, P < 0.0001) (see Table 2 for 95% confidence intervals and Figure 1A,B for Kaplan–Meier curves). A number of factors were independently predictive of survival in a Cox proportional hazards model. For tamoxifen failure these variables include tumor size (HR = 1.92 for tumors >2 cm, P = 0.03), lymph node positivity (HR = 3.41, P = 0.01), and ER percent staining (HR = 0.98, P = 0.0002) (Table 2).

	Tamoxifen failure 5 years			DFS overall		DSS overall
Variable	n	HR (95% CI)	Pvalue	HR (95% CI)	P value	HR (95% CI)	P value
Age <50	16	1.00		1.00		1.00
Age ≥50	175	0.49 (0.22, 1.08)	0.08	0.69 (0.33, 1.46)	0.33	0.79 (0.33, 1.87)	0.58
Tumor size ≤2 cm	105	1.00		1.00		1.00
Tumor size >2 cm	86	1.92 (1.08, 3.42)	0.03	1.95 (1.18, 3.24)	0.01	2.39 (1.32, 4.31)	0.004
Tumor grade 1	15	1.00		1.00		1.00
Tumor grade 2	112	1.78 (0.42, 7.54)	0.43	1.33 (0.47, 3.74)	0.59	1.05 (0.37, 3.02)	0.92
Tumor grade 3	63	3.33 (0.78, 14.2)	0.10	2.05 (0.71, 5.90)	0.18	1.52 (0.51, 4.51)	0.45
LN-negative	63	1.00		1.00		1.00
LN-positive	83	3.41 (1.40, 8.31)	0.01	2.42 (1.19, 4.94)	0.02	2.12 (0.95, 4.75)	0.07
%ER, continuous	192	0.98 (0.98, 0.99)	0.0002	0.99 (0.98, 0.99)	0.0004	0.99 (0.98, 0.99)	0.001
AR = 0%	20	1.00		1.00		1.00
AR >0%	171	0.61 (0.27, 1.36)	0.23	0.91 (0.42, 2.01)	0.82	1.20 (0.43, 3.33)	0.73
AR/ER < 2	169	1.00		1.00		1.00
AR/ER ≥ 2	22	4.43 (2.33, 8.42)	<0.0001	4.40 (2.47, 7.83)	<0.0001	3.66 (1.94, 6.93)	<0.0001
PR-negative	22	1.00		1.00		1.00
PR-positive	108	0.43 (0.18, 1.04)	0.06	0.62 (0.27, 1.43)	0.26	0.69 (0.26, 1.84)	0.46
MIB-1 <21.3	93	1.00		1.00		1.00
MIB-1 ≥21.3	95	1.17 (0.66, 2.07)	0.59	0.98 (0.59, 1.61)	0.93	0.98 (0.55, 1.73)	0.94
Mitotic index number ≤4	94	1.00		1.00		1.00
Mitotic index number >4	92	1.64 (0.92, 2.94)	0.10	1.54 (0.93, 2.55)	0.10	1.34 (0.77, 2.37)	0.30
erbB2 ≤30%	152	1.00		1.00		1.00
erbB2 >30%	16	1.02 (0.36, 2.84)	0.98	0.71 (0.26, 1.96)	0.51	0.69 (0.22, 2.24)	0.54
EGFR = 0%	139	1.00		1.00		1.00
EGFR >0%	26	1.31 (0.60, 2.83)	0.50	1.01 (0.50, 2.07)	0.97	1.17 (0.54, 2.51)	0.70

AR, androgen receptor; CI, confidence interval; DFS, disease-free survival; DSS, disease-specific survival; EGFR, epidermal growth factor receptor; ER, estrogen receptor; LN, lymph node; PR, progesterone receptor. Bold data are significant.

Figure 1 shows Kaplan–Meier curves with survival separated into two groups: AR:ER ratios <2.0 (blue squares) and those with ratios ≥2.0 (red circles). By the end of 10 years, the observed DFS was 10% for patients with a higher AR:ER ratio compared with approximately 70% for women with a lower ratio (Figure 1A; log-rank P < 0.0001). Overall, 27% (6/22) of women with high ratios remained disease free by the end of the study or at the time they were censored compared with 72% (47/169) of women with low ratios. The DSS by the end of the study was about 30% for women with higher AR:ER ratios compared with about 60% for those with lower ratios (Figure 1B; P < 0.0001). The majority of women with high ratios (59%; 13/22) died from their breast cancer during the study period; only 21% (36/167) of women with low ratios died (Figure 1B). As shown in Figure 1C, there is a significant difference in the time to recurrence, with patients having tumors with high AR:ER ratio failing approximately 11 months earlier than those with a low (<2) ratio. The significance does not hold up for DSS; however, patients with high AR:ER ratios died from their breast cancer on average 10 months earlier than patients with low ratios (Figure 1D). The number of patients at risk at each time point is reflective of the number of patients censored due to no further follow-up data at each time point (underneath Figure 1A,B,C,D). Representative AR/ER staining in the <2 or ≥2 categories is shown (Figure 1E).

To investigate whether the AR:ER ratio was merely a reflection of the level of ER positivity, we tested various cutoff points for ER% cell staining. Using 20% cell staining positive for the ER cutoff point, we determined that although those patients with little ER were of course more likely to have a high AR:ER ratio (10/15), there were 12/165 tumors with high ER levels that also had a high AR:ER ratio (>2.0). A high AR:ER ratio is therefore not merely a consequence of low ER. In the multivariate setting, while the dichotomization of ER at <20% versus ≥20% was significant alone, when the AR:ER ratio was added ER percent cell staining lost its significance.

Androgens are proliferative in ER+/AR + breast cancer lines and the AR signaling inhibitor enzalutamide inhibits androgen-mediated proliferation and tumor growth in vivo

Lysates from four luminal ER + breast cancer cell lines were probed for AR and ER (Figure 2A). The prostate cancer cell line LNCaP and the molecular apocrine breast cancer cell line MDA-MB-453, which express high levels of AR [20, 28, 29], were used as positive controls for AR expression. MCF7 cells and the newly derived BCK4 cell line express both AR and ER (Figure 2A) and the new androgen receptor signaling inhibitor enzalutamide prevents ligand-mediated stabilization of AR protein in MCF7 cells (Figure 2B). Both cell lines proliferate in response to DHT (Figure 2C,D). Unlike androstenedione and testosterone, DHT is not aromatizable to estrone or E2 [30–32]. DHT-stimulated proliferation was blocked by enzalutamide in both the MCF7 and BCK4 lines (Figure 2C,D). Enzalutamide inhibited DHT-mediated nuclear translocation of AR within 3 hours as determined by nuclear and cytosolic fractionation (Figure 2E).

To determine whether enzalutamide inhibits androgen-mediated growth in vivo, MCF7 cells constitutively expressing luciferase (MCF7-TGL) were injected into the mammary fat pad of ovariectomized mice implanted with DHT pellets and the tumor burden was measured using luminescent imaging and caliper measurements. Once tumors were established, mice were matched based upon tumor imaging (day -2) into two treatment groups, one receiving control chow and the other receiving chow containing 50 mg/kg enzalutamide on day 0. Tumors in the DHT-treated mice on control chow continued to grow, while mice receiving DHT plus enzalutamide showed regression of tumors by the in vivo imaging system (Figure 3A) and caliper measurement (data not shown). On the final day of imaging (day 19), tumors had regressed to near undetectable levels, with an 83.2% decrease in luminescence in mice receiving DHT plus enzalutamide as compared with the DHT group (Figure 3A,B). As determined by BrdU incorporation and immunostaining, proliferation in the enzalutamide-treated tumors was 31.3% lower than in tumors treated with DHT alone (Figure 3C). TUNEL staining indicated a 50% increase in apoptotic cells in enzalutamide-treated tumors (Figure 3D). A dramatic (92.5%) decrease in AR nuclear localization was observed in the tumors treated with enzalutamide (Figure 3E), consistent with the ability of enzalutamide to impair nuclear entry of AR in prostate cancer [33]. Similar results to the above with MCF7 xenografts were obtained in mice administered enzalutamide by oral gavage, where tumor burden decreased in a dose-dependent manner (data not shown).

Enzalutamide inhibits androgen-mediated growth in ER– breast cancer cells in vitro and in vivo

The MDA-MB-453 breast cancer cell line represents the ER– molecular apocrine or luminal androgen receptor subtype of breast cancer with high levels of AR [17, 18, 20, 34]. In this line, AR contains a point mutation (Q865H) reported to decrease sensitivity to DHT [35]. Nonetheless, these cells proliferate in response to androgens [28, 29] and we therefore sought to determine whether enzalutamide could block DHT-mediated effects on proliferation and gene expression. Indeed, enzalutamide completely abrogated proliferation induced by DHT (Figure S2A in Additional file 2) and expression of known AR-regulated genes [29], such as fatty acid synthase, gross cystic disease fluid protein (also called prolactin inducible protein) and prolactin receptor, was reduced by enzalutamide (Figure S2B in Additional file 2). Further, in MDA-MB-453 cells that stably express an androgen responsive luciferase reporter (MDA-kb2) [36], enzalutamide inhibited luciferase reporter activity in a dose-dependent manner (Figure S2C in Additional file 2). Enzalutamide impairs ligand-mediated nuclear import of AR in prostate cells [33], and in MDA-kb2 cells it reduced the ratio of nuclear to total AR (Figure S2D in Additional file 2). Immunoblotting for AR in nuclear and cytoplasmic lysates demonstrates that the same is true in wild-type MDA-MB-453 cells (Figure S3 in Additional file 3).

To determine whether enzalutamide inhibits androgen-induced tumor growth of ER– breast cancer cells, MDA-MB-453 xenografts were grown at the orthotopic site in mice implanted with a DHT pellet and the tumor size was measured by caliper. Once tumors reached 100 mm³, mice were treated with 10 mg/kg/day enzalutamide or vehicle by oral gavage (Figure 4A, green arrow). DHT stimulates tumor growth as previously reported [20], but in mice treated with DHT plus enzalutamide (10 mg/kg by oral gavage) tumors did not significantly differ from mice that received vehicle control (Figure 4A). Another group of mice received a higher dose of enzalutamide (25 mg/kg/day) starting when tumors reached an average of 400 mm³ (Figure 4A, blue arrow). At this higher dose, there was a trend towards decreased tumor size, although this did not reach statistical significance (Figure 4A). Tumor weights in either the low-dose or high-dose enzalutamide treatments were significantly lower than mice treated with DHT only, an 85.2% and 65.0% decrease respectively (Figure 4B), indicating that the caliper measurements for a high dose of enzalutamide underestimates the decreased tumor burden in this group. Interestingly, there was a statistically significant increase in apoptosis in both enzalutamide treatment groups versus DHT (60.0% and 54.3% increase in low-dose and high-dose groups respectively), as measured by cleaved caspase 3 (Figure 4C, quantification on left and representative images on right), but there was no difference in the proliferation rate of any of the groups, as measured by Ki67 staining (not shown). Thus, in MDA-MB-453 tumors, DHT protects cells against apoptosis and enzalutamide impairs this anti-apoptotic effect. Consistent with the in vitro data, enzalutamide decreased ligand-mediated nuclear entry of AR such that there is a significant decrease (50.0% in low dose and 44.3% in high dose) in the number of AR-positive nuclei in the enzalutamide-treated tumors (Figure 4D, quantification on left and representative images on right). Similarly, when an MDA-MB-453 xenograft study was performed with low-dose and high-dose enzalutamide treatments initiated when the tumors reached 100 mm³ (Figure S4A in Additional file 4), tumor growth was decreased in a dose-dependent manner (Figure S4B in Additional file 4) and was associated with significantly reduced nuclear AR staining in enzalutamide-treated tumors (Figure S4C in Additional file 4). Steady-state concentrations of enzalutamide, including the pharmacologically active metabolite N-desmethyl-MDV3100, in the MDA-MB-453 xenograft studies were only moderately lower than what has been reported in patients receiving 160 mg/day enzalutamide (Cmax values for enzalutamide and the pharmacologically active metabolite, N-desmethyl enzalutamide, were 16.6 μg/ml and 12.7 μg/ml, respectively).

Enzalutamide inhibits estrogen mediated growth in vitro and in vivo

While enzalutamide has high affinity binding for AR, it does not significantly bind to either ERα or ERβ as determined by ligand binding assays (Table S1 in Additional file 5). However, originally as a negative control in experiments where we were antagonizing DHT with enzalutamide, we combined enzalutamide with E2 in ER+/AR + breast cancer cells. Surprisingly, enzalutamide significantly inhibited E2-induced proliferation of both MCF7 and BCK4 cells in vitro (Figure 5A,B). Enzalutamide also inhibited E2-induced upregulation of PR and SDF-1, two estrogen-responsive genes (Figure 5C). In stark contrast, although bicalutamide effectively inhibited DHT-mediated proliferation in MCF7 cells (Figure 5D), it had the opposite effect on E2 signaling, as it significantly increased E2-mediated proliferation (Figure 5E) and increased the E2-mediated induction of PR and SDF-1 mRNA (Figure 5F).

To determine the effect of enzalutamide on E2-stimulated breast tumor growth in vivo, a xenograft study was performed injecting MCF7-TGL cells in ovariectomized mice implanted with an E2 pellet. Cells were injected orthotopically and once tumors were established (arrow, average size of 100 mm³), mice were matched into three groups: control chow; control chow and a tamoxifen pellet; and chow containing 50 mg/kg enzalutamide (Figure 6A). Enzalutamide significantly inhibited E2-mediated MCF7 tumor growth as effectively as tamoxifen, with a decrease in tumor luminescence of 59.9% for the tamoxifen group and 70.3% in the enzalutamide group at day 11. Day 11 was the final day of imaging for the E2-only group since these mice had to be euthanized due to tumor burden. Luminescence flux for individual animals (Figure 6B) and images of mice (Figure 6C) are shown for the day of matching (day -3) and the last imaging day when all mice were alive (day 11). Both drugs significantly decreased cell proliferation, with a 46.4% decrease in the E2 plus tamoxifen group and a 54.2% decrease in the E2 plus enzalutamide group compared with the E2 group, as measured by BrdU incorporation (Figure 6D). In contrast to what was observed in DHT-mediated tumor growth, enzalutamide did not increase apoptosis when opposing E2-stimulated growth (data not shown). Interestingly, ER protein levels in the MCF7 xenograft tumors were affected differently by tamoxifen versus enzalutamide (Figure S5 in Additional file 6). ER immunostaining was quantified with ImageJ and by pathologist (PJ) scoring in a blinded manner for percent cells positive for nuclear ER. By both methods, ER was extremely low in the E2-alone group, but significantly increased with the addition of tamoxifen. However, in the E2 plus enzalutamide group, ER levels are not significantly different from E2 alone, indicating that enzalutamide does not elicit upregulation of ER like tamoxifen (Figure S5 in Additional file 6) and suggests that enzalutamide affects ER by a different mechanism than the competitive antagonist tamoxifen. This intriguing finding will be the focus of a subsequent study.

Importantly, mean animal weights during and at the end of all in vivo studies showed no differences across treatment groups, indicating no adverse effects on the general health of the mice (Figure S6 in Additional file 7).