Selective Androgen Receptor Modulator RAD140 Inhibits the Growth of Androgen/Estrogen Receptor–Positive Breast Cancer Models with a Distinct Mechanism of Action

This article is featured in Highlights of This Issue, p. 7435

Breast cancer is the second leading cause of cancer-related death in women, with an estimated 246,660 newly diagnosed cases and 40,450 deaths in the United States alone in 2016 (1). Breast cancer is a heterogeneous disease categorized into several histopathologic subtypes based on the status of estrogen receptor α (ER), progesterone receptor (PR), and HER2 receptor. Although ER-positive (ER⁺) breast cancers are now treated with standard-of-care agents targeting the ER axis, including tamoxifen, fulvestrant, and aromatase inhibitors (AI), and HER2-positive tumors are treated with HER2 inhibitors, such as trastuzumab, novel therapeutic approaches are still needed to address resistance emerging from these established regimens (2, 3). More recently, combined administration of ER-targeted therapies with the inhibitors of cyclin-dependent kinase (CDK) 4/6 (4) or mTOR (5) have yielded improved therapeutic efficacy in ER⁺ breast cancer, and these combination therapies now represent a new generation of standard of care for this indication.

Recent histopathologic studies revealed that the androgen receptor (AR) is the most commonly expressed hormone receptor in breast cancer, with 75% to 90% of ER⁺ and approximately 30% of ER⁻ breast cancers expressing AR (6, 7). Although AR is widely present in breast cancer tissue, accumulating evidence has demonstrated that the role of AR in these tumors is subtype dependent (8–11). AR antagonists including bicalutamide and enzalutamide have been shown to reduce the growth of AR-positive but ER⁻, including triple-negative, breast cancer in preclinical models (12–14) and in patients (11). In contrast, in ER⁺ breast cancers, AR has been considered antiproliferative and has been associated with a favorable prognosis. Clinically, until the 1970s, breast cancers were often treated with nonselective steroidal androgens including testosterone derivatives and danazol with response rates of 20% to 25% (9, 15–18). In line with these clinical experiences, preclinical studies have also shown treatment with classic androgens, including DHT, reduced the growth of AR and ER⁺ (AR/ER⁺) breast cancer cells in vitro (19–21) and in vivo (22). Classic androgen-based therapy for breast cancer declined due to its virilizing effects, potential risk of further aromatization to estrogens, and the emergence of ER-targeted agents including tamoxifen. However, it is worth noting that more recent clinical evidence showed androgen treatment also led to remission in patients who were progressing after ER-targeted therapy with objective response rate around 17% to 39% (23, 24). Fulvestrant at second line or later in large phase III trials elicited objective response rates of 2.1% to 11% (25–28).

The high prevalence of AR observed in ER⁺ breast cancers, along with the prior clinical efficacy demonstrated with classic androgens, provides a strong rationale for the development of a new generation of oral, selective AR agonists and to further exploit their therapeutic benefits in AR/ER⁺ breast cancer. Here, we describe the tissue-selective AR agonist activity of RAD140, an oral nonsteroidal selective AR modulator (SARM), and its distinct AR-mediated mechanism of action including the suppression of ESR1. RAD140 exhibited potent antitumor activity in the in vivo models of AR/ER⁺ breast cancer both as a single agent and in combination with the CDK4/6 inhibitor palbociclib. RAD140 represents a new generation of tissue-selective AR agonists and potentially a novel therapeutic option for AR/ER⁺ breast cancer.

The PolarScreen nuclear receptor competitor assays (Thermo Fisher Scientific) for AR, ER, PR, and glucocorticoid receptor (GR) were used to determine the binding affinity and specificity of RAD140. Briefly, RAD140 and fluorescence-labeled ligands (Fluormones) for AR, ER, PR, or GR, respectively, were added to each reaction well, and the plate was incubated for 4 hours at room temperature in the dark. Polarization values for each well were measured and plotted against the concentration of RAD140. As a reference, the binding affinity of the standard ligands DHT, 17β-estradiol, progesterone, and dexamethasone (for AR, ER, PR, and GR, respectively) was also measured. The relative binding affinity of RAD140 versus the 4 standard ligands is reflected by the ratios of IC₅₀ (standard ligand)/IC₅₀ (RAD140). A commercially available spectrum screen was performed for RAD140 (MDS Pharma Services). The binding of RAD140 (1 μmol/L) to a panel of cellular targets was assessed (Supplementary Table S1). Appreciable binding was defined as a larger than 50% inhibition of the reference radio-ligand.

The ZR-75-1, HCC1428, and T47D breast cancer cells and LNCaP prostate cancer cells were purchased from ATCC and maintained in medium and condition as recommended by the vendor. All cell lines involved in the study were tested negative for mycoplasma contamination (MycoAlert Detection Kit) and subjected to annual authentication using STR profiling (Thermo Fisher Scientific). All the cells were under passage 15 at the time of experiments. HCC1428 long-term estrogen-depleted (LTED) cells were established by culturing these cells in medium with 10% charcoal dextran-stripped serum (CSS) for 16 weeks, and the expressions of AR, ER, and PR were confirmed by Western blotting (Supplementary Fig. S1A). For proliferation assay, cells were seeded in medium with 10% CSS at 30,000 cells per well in 24-well plates and subjected to treatment with RAD140, DHT, or DMSO for 14 days with treatments being renewed every 3 days. At the end of the treatment period, trypsinized cells were stained with Trypan blue before being subjected to live cell counting in a Nexcelom Cellometer (Nexcelom Bioscience). The PC3-AR cell line, described previously (29), was a generous gift from Dr. Steve Balk of Beth Israel Deaconess Medical Center (Boston, MA) and Dr. Changemeng Cai of University of Massachusetts (Boston, MA). For ZR-75-1 cell-based assays, cells were seeded and incubated in medium with 5% CSS for 48 hours. Cells were then treated with 17β-estradiol at 1 nmol/L final concentration or vehicle (ethanol) before being incubated with DMSO, RAD140 (Radius Health, Inc., as described previously; ref. 30), or DHT (Sigma Aldrich) in the presence or absence of enzalutamide or ARN-509 (Selleckchem) for 24 hours. The experiments were performed in duplicate. The nuclear and cytoplasmic fractions of the cells were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Fisher Scientific) following the manufacturer's instructions.

AR reporter gene assay was carried out using a Cignal Androgen Receptor Reporter (luc) Kit (Qiagen). ZR-75-1 breast cancer cells, LNCaP, and PC3-AR prostate cancer cells were transfected with a tandem androgen-responsive element (ARE)-driven firefly luciferase construct and a Renilla luciferase construct in RMPI1640 media containing 5% CSS. Forty-eight hours later, transfected cells were treated with DMSO, RAD140, or DHT. The luciferase activity was determined using the Dual-Luciferase Reporter Assay (Promega) as a representation of AR transcription activity and normalized to the activity of Renilla luciferase. Each set was carried out in triplicate. The values were presented as average fold changes ± SD over the vehicle control.

All study protocols were reviewed by Institutional Animal Care and Use Committees and Radius Health, Inc., and conducted in compliance with U.S. and European regulations of protection of laboratory animals. The HBCx-22, HBCx-21, and HBCx-3 breast cancer patient-derived xenograft (PDX) models were established and evaluated at XenTech by implanting tumor fragments subcutaneously in the flank female athymic nude mice (Foxn1nu, Envigo RMS). The ST897 breast cancer PDX model was established in female athymic nude mice (Foxn1nu, Charles River Laboratories) and studies conducted at South Texas Accelerated Research Therapeutics. These models were characterized as ER and PR positive and negative for HER2 overexpression based on IHC and later confirmed using Western blot analysis and qPCR (Supplementary Table S2; Supplementary Fig. S1). The efficacy study in T-47D breast cancer cell line–derived xenograft model, established in NOD/Shi-scid/IL-2Rγ^null mice, was conducted at WuXi AppTec. These xenograft models were supplemented with exogenous estradiol in drinking water (HBCx-3, HBCx-21, HBCx-22, and ST897) or subcutaneous pellet (0.18 mg/90-day release, Innovative Research America) implanted 3 days prior to tumor cell inoculation (T-47D). The PDX-bearing mice were randomized once their tumor volumes reached 60 to 200 mm³, as described previously (31). The cell line–derived xenografts with volumes between 125 and 250 mm³ were enrolled in treatment groups. RAD140, DHT, or fulvestrant were administered as single agent or in combination with palbociclib. RAD140, palbociclib, and the vehicle (0.5% carboxymethyl cellulose) were given orally. Fulvestrant (clinical formulation, AstraZeneca) was given as subcutaneous injection. DHT was administered as subcutaneously implanted pellets (12.5 mg/60-day release, Innovative Research America) from the day of randomization to the end of treatment (14). Tumor volume and body weight were measured twice a week. Tumor growth inhibition (TGI %) was calculated on the basis of the change of mean tumor volume from treatment day 1 to the last day of assessment and presented as a percentage of that of the vehicle control group [100 × (1- ΔTV_treatment/ΔTV_vehicle)], as an indication of antitumor efficacy. Statistical analyses were performed for the changes in individual tumor volume (ΔTV, from treatment day 1 to the last day of assessment) between the groups compared.

Formalin-fixed paraffin-embedded sections of xenograft tumors underwent antigen retrieval and were then blocked using 5% goat serum and avidin blocking solution. IHC was performed in a Leica Bond III auto-stainer (Leica Biosystems) according to the manufacturer's instructions. The primary rabbit mAbs against ER (SP1) and PR (SP42) were obtained from Cell Marque. Rabbit mAb against Ki67 (D2H10) and phospho-retinoblastoma protein (phospho-Rb) were from Cell Signaling Technology. Mouse antibody against AR (441) was obtained from Dako. The AR IHC staining protocol included the use of ImmunoDNA Background Blocker (Bio SB). Sections shown in each figure were stained and photographed under identical conditions. Staining and scoring of the samples were performed by a pathologist blinded to the treatment groups.

Tumor lysates from the in vivo studies were prepared using Cell Lysis Buffer with protease and phosphatase inhibitors (Cell Signaling Technology) in a FastPrep Sample Preparation System (MP Biomedicals). Cell lysates were prepared using Cell Lysis Buffer as described above. Total protein was quantitated and subjected to Western blot analysis using antibodies against AR (PG-21, EMD Millipore; Clone 441, Thermo Fisher Scientific), ERα, PR, GAPDH, β-tubulin, and HDAC2 (Cell Signaling Technology).

Total RNA from xenograft samples and cells was extracted using the RNeasy Mini Kit (Qiagen), with a DNase incubation step included to ensure complete removal of genomic DNA. The expression of genes was evaluated using qPCR and RNA sequencing (RNA-seq). For qPCR, TaqMan primer and probe sets for AR, KLK2, FKBP5, ZBTB16, ESR1, PGR, TFF1, GREB1, BLM, FANCI, LMNB1, MCM2, MCM4, MCM7, 18S, and GAPDH were purchased from Thermo Fisher Scientific. The PCR reaction was set up with TaqMan Fast Virus 1-Step Master Mix in a Quant Studio 6 Flex qPCR system (Thermo Fisher Scientific). Samples from a minimum of 3 tumors from each treatment group or those from a minimum of 2 replicated cell-based assays were included for the analysis. The expression of genes of interest were normalized to that of GAPDH or 18S and presented as the mean fold change compared with the vehicle control group. For RNA-seq analysis, total RNA samples were converted into cDNA libraries using the TrueSeq Stranded mRNA Sample Preparation Kit (Illumina). Libraries are quantified, normalized, and pooled before being subjected to HiSeq 2 × 50 bp paired end sequencing on an Illumina sequencing platform. De-multiplexed FASTQ files were aligned to HG19 human genome using the STAR aligner version 2.4 and genetic features were quantified using RSEM version 1.2.14. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE104177. The AR and ER target genes with significant changes in expression (FDR < 0.05) were analyzed and log₂ fold changes presented in heatmap. Global pathway analysis was performed for genes with significant changes in expression (FDR < 0.001) using the functional annotation tools (Gene Ontology) available on the National Health Institute DAVID Bioinformatics Resources platform (https://david.ncifcrf.gov).

The binding affinity of RAD140 was assessed using fluorescence polarization-based competitive binding assays for AR, ER, PR, and GR and compared against the standard ligands for each nuclear receptor. As positive controls, assays evaluating the binding of the standard ligands, including DHT, 17β-estradiol (E2), progesterone, and dexamethasone to AR, ER, PR, and GR, respectively, were performed. In the competitive binding assay against the AR-Fluoromone, RAD140 exhibited 1.6-fold lower AR affinity than DHT (Fig. 1A). This suggests binding affinity of RAD140 to AR is slightly lower but comparable with that of natural androgen. RAD140 exhibited 33-fold lower affinity to PR compared with progesterone. No binding of RAD140 to ER or GR was detected. In addition, a cellular target spectrum screen was performed to determine the binding selectivity of RAD140 against a broad panel of 165 molecular targets that included many pharmacologically relevant molecules including ion channels, G-protein–coupled receptors, enzymes, and nuclear hormone receptors (Supplementary Table S1). No appreciable interaction of RAD140 with any targets screened, other than AR and PR, was detected. These results indicate RAD140 binds to AR with high affinity and specificity.

SARMs are selective AR ligands that have been shown to act as agonists or antagonists in a tissue-context–dependent manner (32). We previously demonstrated that RAD140 exhibits differential activity in prostate versus in the muscle (30). To further assess the activity of RAD140 on AR transcription in tissue context, an AR reporter assay was performed in ZR-75-1, a human breast cancer cell line expressing endogenous AR and ER (21), and the AR-positive prostate cancer cell line LNCaP (29). Dose responses by RAD140 were compared with the known AR agonist DHT. Treatment of steroid-depleted ZR-75-1 cells with 1, 10, and 100 nmol/L RAD140 induced AR transcription activity by 1.9- to 2.9-fold, comparable with the induction seen with DHT at the same concentrations (Fig. 1B, left). In contrast, in the low-passage, androgen-sensitive LNCaP cells treated with RAD140 at 0.1 to 10 nmol/L, no induction of AR activity was observed while DHT at 1 or 10 nmol/L led to more than a 10-fold induction of AR transcription activity (Fig. 1B, right). To rule out the possibility that the attenuated AR activity of RAD140 is due to the expression of a mutant AR T877A in LNCaP cells, we also performed AR reporter assay in PC3-AR cells, which express ectopic wild-type AR (29). Similar to the finding in LNCaP cells, RAD140 exhibited much attenuated activity on ectopic wild-type AR in prostate cancer cells compared with DHT (Supplementary Fig. S2). These results indicated that RAD140 is a tissue-specific AR agonist, with selective activity in breast cancer cells but not in prostate cancer cells.

Classic AR agonists are known to induce AR conformation change and subsequent localization to the nuclei, while pure AR antagonist, such as enzalutamide and ARN-509 act partially via blocking AR nuclear localization (33, 34). We then examined whether RAD140 induces the nuclear expression of AR in breast cancer cells. Treatment of ZR-75-1 cells with RAD140 led to increased AR expression in the nuclear fraction while the estrogen (E2) showed no apparent effect (Fig. 1C). The positive control, DHT, substantially increased the nuclear expression of AR. ARN-509 substantially reduced the nuclear expression induced by DHT and RAD140, suggesting a direct antagonism against these two AR agonists. This provided further evidence that RAD140 is an AR agonist in breast cancer cells.

It has been previously reported that DHT inhibits the proliferation of ER⁺ breast cancer cells driven by supplemental estradiol (21). We evaluated the activity of RAD140 and DHT in an estrogen-independent AR/ER⁺ breast cancer line HCC1428 LTED, in which ER signaling appears to be active as judged by PR expression (Supplementary Fig. S1). RAD140 exhibited inhibitory effect on the proliferation of these endocrine-resistant breast cancer cells at concentrations as low as 10 nmol/L, with an apparent maximal effect seen at 100 nmol/L (Fig. 1D). This was comparable with the inhibitory effect seen with DHT in the 0.1 to 10 nmol/L range. To further assess how the AR agonist activity of RAD140 translates into antitumor activity in vivo, RAD140 was evaluated as a single agent in AR/ER⁺ breast cancer PDX models. Oral administration of RAD140 induced significant TGI of 76%, 59%, and 58% when compared with the vehicle control groups in HBCx-22, HBCx-3, and HBCx-21, respectively (Fig. 2A, Supplementary Fig. S3A and S3B). In the HBCx-22 model, fulvestrant, a selective ER downregulator (SERD) and standard of care for ER⁺ breast cancer dosed at 1 mg weekly, exhibited a statistically significant antitumor activity, as judged by a TGI of 59%, comparable with a previous report of this model and indicative of its dependency on ER (31). It has been previously determined that the 1 mg weekly administration of fulvestrant achieves exposure equivalent to the 500 mg monthly dose in human (35). The changes in tumor volume of the RAD140- or fulvestrant-treated HBCx-22 xenografts appeared to be lower compared with those treated with vehicle (Fig. 2B). To further verify the concept of inhibiting the growth of AR/ER⁺ breast cancer tumors with AR agonists, T-47D cell line–derived xenografts were treated with DHT, RAD140, or fulvestrant (Supplementary Fig. S3C). Treatment with DHT or RAD140 inhibited the growth of these T-47D tumors to a comparable extent. Fulvestrant also elicited potent growth inhibition. In addition, the activity of the combination of fulvestrant with RAD140 was evaluated in HBCx-22 model, but no appreciable improvement in tumor growth inhibition was observed, suggesting a potential overlap in the pathway inhibition induced by these two agents in this model (Supplementary Fig. S3D). This observation is consistent with previous clinical experience with anastrozole–fulvestrant combination and preclinical data of fulvestrant–RAD1901 combination, both of which showed such intense ER blockade failed to yield additional benefit (35, 36). Together, these results indicated that RAD140 as a single agent is effective in inhibiting the growth of AR/ER⁺ breast cancer xenografts.

We then examined the changes in common breast cancer biomarkers in the HBCx-22 xenografts treated with RAD140 or fulvestrant using IHC. As shown in Fig. 2C, HBCx-22 tumors in the control group were positive for AR, ER, and PR, consistent with prior characterization of this model (Supplementary Fig. S1). An apparently increased AR expression, predominantly localized in the nuclei, was seen in RAD140-treated tumors. This demonstrated that RAD140 possesses AR agonist properties, such as stabilizing AR and promoting its nuclear localization, which are commonly seen with classic androgens (37, 38). Interestingly, RAD140 treatment led to substantially decreased expression of ER and its downstream target PR. A profound decrease in Ki67 expression was also seen in these RAD140-treated tumors, indicating a suppressive effect on the proliferation of these tumor cells. Treatment with fulvestrant also led to substantial decreases in ER, PR, and Ki67 expression, to levels seemingly comparable with those seen with RAD140. Furthermore, the epithelial nature of the terminal tumor mass was confirmed using IHC assays for cytokeratin (CK) 18 and vimentin (Supplementary Fig. S4). This suggests SARM and SERD both inhibit the ER pathway and breast cancer cell proliferation.

It has been proposed that androgens suppress ER signaling in AR/ER⁺ breast cancer cells, which may contribute to the antiproliferative activity (19, 21). To better understand the mechanism of action of the SARM RAD140 in AR/ER⁺ breast cancer models, we examined the pharmacodynamic changes with a focus on the AR and ER pathways. In the HBCx-22 model, a profound induction of AR target genes, including FKBP5, KLK2, and ZBTB16, was seen in RAD140-treated tumors (Fig. 3A, top) while the expression of the AR gene was not affected. Fulvestrant, as expected, did not induce the expression of these AR target genes, although a slight increase in AR message was seen. The ER target genes PGR, TFF1, and GREB1 were found to be substantially suppressed in RAD140-treated tumors to levels comparable or even lower than that seen with fulvestrant (Fig. 3A, bottom). Interestingly, in these RAD140-treated tumors, the mRNA expression of ESR1, the coding gene of ER, was profoundly suppressed. In contrast, the fulvestrant treatment did not lead to appreciable change in ESR1 gene expression. At protein level, both RAD140 and fulvestrant exhibited a profound effect in decreasing the expression of ER and PR in HBCx-22 xenografts (Fig. 3B, left), consistent with the IHC findings described above. Similar decrease in ER and PR expression was seen in HBCx-3 AR/ER⁺ xenografts treated with RAD140 (Fig. 3B, middle). In the T-47D xenografts, treatment with RAD140, DHT, or fulvestrant also led to decreased ER and PR expression (Fig. 3B, right).

Next, the modulation of AR and ER target genes by RAD140 was further examined in vitro. As determined in ARE-luc reporter assay and proliferation assay (Fig. 1B and D), RAD140 at 100 nmol/L exhibited maximal effects, comparable with DHT at 10 nmol/L; therefore, 100 nmol/L was selected as the RAD140 concentration for further in vitro assays. In ZR-75-1 cells treated with RAD140, substantial induction of FKBP5 and ZBTB16 messages was seen (Fig. 3C), which was comparable with that seen with DHT. The induction of these AR targets by RAD140 or DHT was completely blocked by the antagonist enzalutamide. The expression of the AR gene did not seem to be affected by RAD140 or DHT treatment. The ER target genes PGR and TFF1 were induced by 1 nmol/L of E2. Treatment with either RAD140 or DHT was found to repress the E2-induced expression of these genes. Similar to the effects seen in xenograft models, treatment with RAD140 or DHT decreased ESR1 mRNA. The effects of RAD140 and DHT on AR pathway and ER pathway genes appear to be comparable, suggesting an AR-specific effect. Furthermore, competitive blockade of AR activation by enzalutamide effectively reversed the AR and ER pathway modulation by RAD140 or DHT. The incomplete reversal of ESR1 suppression by enzalutamide may be due to the relatively lower affinity of this compound compared with AR agonists RAD140 and DHT. These in vivo and in vitro observations further demonstrated the AR agonist activity of RAD140 in AR/ER⁺ breast cancer models. More importantly, RAD140 and DHT both inhibited ER downstream targets, as well as the ESR1 gene, suggesting a novel mechanism of action of AR agonists in inhibiting the ER pathway in breast cancer cells.

To further understand the global effect of RAD140 on signaling cascades in AR/ER⁺ breast cancer cells, RNA-seq analysis of HBCx-22 xenografts was performed. We examined a broader range of AR and ER target genes with significantly altered expression in RAD140-treated tumors (Fig. 4A). The results indicated that in addition to the targets examined by qPCR, RAD140-treated tumors had higher levels of AR-activated genes in addition to those shown in qPCR. A subset of the ER target genes was found to be suppressed in RAD140-treated tumors, consistent with the qPCR findings. Using a stringent filtering criteria (FDR < 0.001), we found that globally 86 genes were significantly upregulated and 77 were downregulated by at least 2-fold in RAD140-treated HBCx-22 xenografts (Fig. 4B; Supplementary Tables S3 and S4). In the Gene Ontology analysis, these upregulated genes were found to be most significantly enriched in the metabolic process. This was consistent with the previous report describing the role of AR as a master regulator of central metabolism and biosynthesis in prostate cancer cells (39). The downregulated genes in RAD140-treated tumors were found to be enriched in categories associated with cell division, DNA replication, and cell-cycle progression. This negative regulation of gene transcription was in line with previous reports on the transcription suppressor role of AR (40, 41). These data further confirmed the regulation of AR and ER pathways by RAD140 and suggested a unique mechanism of action of RAD140 via the AR-mediated transcription repression. These data offer evidence for further uncovering the underlying mechanism of RAD140 as a single agent in AR/ER⁺ breast cancer models.

Given the observed effect of RAD140 on cell cycle and DNA replication–related genes, we hypothesized that combined administration of this SARM and a CDK4/6 inhibitor, which inhibits the phosphorylation and subsequent degradation of Rb, thus blocking cell-cycle entry (42–44), may produce improved antitumor activity in AR/ER⁺ breast cancer models. Palbociclib, a CDK4/6 inhibitor recently approved for breast cancer treatment in combination with ER-targeted agents in as early as first line of therapy (45–47), was selected for this study. Two AR/ER⁺ PDX models, HBCx-3 (Fig. 5A) and ST897 (Fig. 5B), were treated with RAD140, palbociclib, or a combination of RAD140 with palbociclib. Compared with the vehicle-treated group, RAD140 as monotherapy produced statistically significant inhibition of tumor growth in these models compared with vehicle control. Palbociclib as a single agent also inhibited the growth of these xenografts to a comparable level. Importantly, RAD140 and palbociclib when administered together elicited improved efficacy over palbociclib or RAD140 monotherapies. Similarly, RAD140 and palbociclib combination produced improved efficacy compared with palbociclib alone in an additional AR/ER⁺ PDX model, HBCx-22 (Supplementary Fig. S5A). An additional TGI analysis performed for HBCx-3 xenografts treated with RAD140 and palbociclib with lower starting volume (≤108 mm³) versus those with higher starting volume (>108 mm³) did not reveal appreciable difference between the two subgroups, suggesting the combination treatment had similar effect in tumors of different starting size. As the combination of RAD140 with palbociclib induced apparent regression in ST897 tumors, we examined the expression of apoptotic marker cleaved caspase-3 in ST897 and HBCx-3 tumors treated with RAD140, palbociclib, or combination but did not observe clear signs of apoptosis (Supplementary Fig. S5B), suggesting a cytostatic but not cytotoxic effect. Together, these results indicate that the combination of RAD140 with palbociclib led to improved antitumor activity compared with each of these agents given as single agent.

The expressions of AR, ER, PR, Ki67, and phospho-Rb in the HBCx-3 PDX were assessed using IHC (Supplementary Fig. S6). Similar to that seen in HBCx-22 (Fig. 2C), RAD140 treatment led to apparent enrichment of AR in the nuclei, suggesting AR activation. ER and PR expression was profoundly decreased in tumors treated with a RAD140-containing regimen. Palbociclib alone also appeared to decrease the expression of ER and PR, a finding similar to the decrease of ER in breast cancer cells treated with a pan-CDK inhibitor roscovitine (48). The phosphorylation of Rb was found to be substantially suppressed in tumors treated with palbociclib as a single agent and in combination with RAD140, while RAD140 alone did not seem to affect phospho-Rb level. Ki67 expression was suppressed in tumors treated with RAD140 or palbociclib, and this suppression appeared to be further enhanced in tumors that received both agents. These results further confirmed the inhibitory effect of RAD140 on ER signaling cascade. Gene expression analysis of the HBCx-3 xenografts treated with RAD140, palbociclib, or their combination showed that the AR target genes FKBP5 and ZBTB16 were robustly induced in RAD140-treated tumors, while no appreciable effect on the AR gene was seen (Fig. 5C, top). The ER target genes PGR and TFF1 were substantially suppressed in the tumors of the RAD140-treated groups (Fig. 5C, middle). Notably, a profound suppression of the ESR1 mRNA level was also observed in these RAD140-treated HBCx-3 xenografts. Palbociclib alone, in contrast, did not have an apparent effect on AR target genes but did seem to repress PGR and induce TFF1 and ESR1 to a relatively modest degree compared with the changes induced by RAD140 treatment.

We next explored the underlying mechanism for the improved efficacy and enhanced inhibition of cell proliferation seen with the RAD140–palbociclib combination. It has been recently reported that in castration-resistant prostate cancer cells, activated AR recruits hypophosphorylated Rb to the loci of genes implicated in DNA replication and suppress their transcription (41). We hypothesized that AR activated by RAD140 inhibits DNA replication–related genes in breast cancer cells, and this effect is enhanced by coadministration of palbociclib. Indeed, the HBCx-3 tumors treated with RAD140 alone had decreased expression of the DNA replication–related genes including Bloom syndrome recQ like helicase (BLM), Fanconi anemia complementation group gene (FANCI), laminin B1 (LMNB1), and minichromosome maintenance genes 2, 4, and 7 (MCM2, MCM4, and MCM7; Fig. 5C, bottom). These decreases were comparable with those seen with palbociclib alone. In tumors treated with both RAD140 and palbociclib, the expression of these genes appeared to be further suppressed. Together, the enhanced effect of RAD140–palbociclib combination on the expression of DNA replication–related genes may have contributed to the enhanced inhibition on tumor growth.

This study demonstrates for the first time that RAD140, an orally available SARM, is an AR agonist in breast cancer cells and suppresses the growth and proliferation of multiple AR/ER⁺ breast cancer cell line and xenograft models. The AR pathway was found to be activated in RAD140-treated breast cancer cells and xenografts, while genes within the ER pathway, including ESR1, were suppressed. In addition, RAD140 treatment was found to decrease the expression of DNA replication–related genes in breast cancer cells, consistent with previous report that these genes were suppressed in androgen-treated prostate cancer cells. Combined administration of RAD140 with the CDK4/6 inhibitor palbociclib was more efficacious compared with either of the agents used alone. These findings suggest a distinct mechanism of action of RAD140, which includes the AR-mediated suppression of ESR1 in inhibiting AR/ER⁺ breast cancer growth.

Accumulating evidence in the recent years has further defined the role of AR in breast cancer and has led to renewed interests in evaluating AR-targeted agents for breast cancer. The high prevalence of AR in the predominant ER⁺ subtype of breast cancers (6), clinical benefit rates as high as 39% seen with androgen therapy in breast cancer patients progressing on ER-targeted treatments (23, 24), along with prior experience with steroidal androgens in breast cancer together lend support to the development of a new generation of oral, nonsteroidal AR agonists for the treatment of AR/ER⁺ breast cancer.

SARMs are tissue-selective AR agonists by design and may offer a novel approach to inhibit the growth of AR/ER⁺ breast cancers, with substantially attenuated side effects commonly seen with classic nontissue selective androgens (32, 49). The activity of RAD140 are AR specific, as evidenced by the similar effects observed with DHT and RAD140 on AR and its downstream targets, along with the reversal of these effects by AR antagonists (Figs. 1C and 3C). In addition, due to its nonsteroidal structure, the SARM RAD140 is not subject to further conversion to estrogens by CYP19 aromatase, or to DHT by 5α-reductase, thus reducing the potential risk of stimulating ER⁺ tumor growth or increasing virilization. As a proof-of-concept study, we report here that RAD140 treatment inhibited the growth of the breast cancer xenograft models supplemented by exogenous estrogen. Of note, RAD140 and its classic androgen comparator, DHT, also inhibited the proliferation of an endocrine-resistant breast cancer cell line model, HCC1428 LTED, suggesting AR agonists may inhibit the growth of ER⁺ breast cancer models with or without estrogen supplementation. Furthermore, in multiple PDX models, RAD140 inhibited tumor growth as a single agent, and this effect was enhanced by combining with palbociclib. PDX models have been demonstrated to closely recapitulate human tumors with regards to tumor heterogeneity and response to standard-of-care agents (31, 50). Indeed, heterogeneity was observed in the PDX models used in this study, as indicated by variable levels of AR, ER, and PR expression, suggesting a good representation of the patient tumors with a spectrum of expression levels of these nuclear receptors. Also, the growth pattern and response to RAD140 seen with HBCx-3 and HBCx-22 PDX models, each evaluated in two independent studies (Figs. 2A and 5A; Supplementary Figs. S3A and S5A), seemed to be consistent. This suggests good reproducibility of these studies. Therefore, the efficacy seen with RAD140 in these PDX models may be suggestive of the potential therapeutic benefit in breast cancer patient population. It is worth noting that although RAD140 and fulvestrant led to similar tumor growth inhibition in HBCx-22 tumors, further studies are needed to evaluate the efficacy of RAD140 relative to that of fulvestrant or tamoxifen in hormone-dependent models to predict its value in first-line setting. It is interesting to note that although HBCx-3 (Supplementary Fig. S6) and HBCx-22 (Fig. 2) xenografts contain comparable percentage of AR-positive cells, RAD140 induced higher degree of inhibition in growth and Ki-67 expression in HBCx-22 than in HBCx-3. This may be due to the presence of higher percentage of ER⁺ cells in HBCx-22 tumors than in HBCx-3 (Supplementary Table S2). As summarized previously, AR agonists exhibit inhibitory effect on AR/ER⁺ breast cells but not on AR⁺/ER⁻ cells (9, 51). It is conceivable that the varying degree of growth inhibition seen with SARM may be attributed to different ER positivity of these models.

The tissue selectivity of RAD140 is evidenced by potent androgen-like effects in bone and muscle, with much attenuated effects in other androgen-responsive tissues, such as prostate and seminal vesicles (30). Notably, when administered along with testosterone, RAD140 was found to partially antagonize the growth effect of testosterone in prostate and seminal vesicles, suggesting that RAD140 acts as a competitive AR ligand with attenuated activity. Here, we demonstrated for the first time that the SARM RAD140 is a potent AR agonist in breast cancer cells. It exhibited much attenuated activity on AR in prostate cancer cells, a finding consistent with the previously described attenuated effect on prostate growth (30). It has been proposed that the tissue-selective AR activity of SARMs is due to the differential allosteric activation of AR compared with classic androgens, which impacts the recruitment of coactivators needed for AR-mediated gene transcription (52, 53). An earlier study on the selective estrogen receptor modulator tamoxifen showed that its tissue selectivity may be attributed to the differential expression of ER coactivators such as SRC-1 in breast versus endometrial cancer cells (54). Similarly, it is conceivable that AR upon binding to RAD140 assumes a conformation that only allows the interaction with a subset of coregulators that are uniquely expressed in breast epithelial cells but not in the prostate.

Despite the preclinical evidence and favorable clinical outcome seen with androgens in ER⁺ breast cancer (9, 10), improved understanding of the underlying mechanism is needed to further improve the design and development of new AR agonist-based therapy. It has been proposed that activated AR suppresses ER signaling by competing with ER for transcriptional coactivators or directly competing for binding sites and subsequently leads to inhibited cell proliferation (55, 56), suggesting a functional interference between these two nuclear receptors. Here, we show AR agonists, including RAD140 and DHT, activate AR target genes in breast cancer cells in vitro and in vivo, while suppressing a subset of ER target genes. Importantly, we found that treatment with these AR agonists also led to substantial suppression of ER at both mRNA and protein levels. This suggests that AR may negatively regulate ER signaling by two mechanisms: while AR may compete with ER for coactivators (functional interference), it may directly downregulate ER expression (direct suppression). The reduction of ER expression by AR agonists has not been extensively documented but can be seen in a study by Poulin and colleagues (57), in which DHT treatment in ZR-75-1 cells for 8 or 12 days led to decreases of ER message by 50% or 65%, respectively. Another study also showed reduced ER protein level in ZR-75-1 cells treated with DHT, along with decreased expression of classic ER targets PGR and TFF1 (56), although the mechanism for the decrease in ER expression had remained to be further elucidated. Our data show both RAD140 and DHT treatment led to decreased ESR1 expression at as early as 24 hours, which may present a novel mechanism by which RAD140 suppresses ER signaling and differentiates from traditional ER-targeted agents that act via inhibiting ER protein function or leading to its degradation. This may translate into unique therapeutic benefit of SARMs in overcoming resistance to ER-targeted therapy seen in tumors with reactivated ER signaling, either via posttranslational modification or enrichment of ESR1 alterations. The inhibitory effect of RAD140 on the proliferation of HCC1428 LTED cells shed light on its efficacy in such E2-independent, albeit ER-active, models but further studies are needed to fully evaluate the activity of SARMs in similar models in vivo. In addition to the suppression of ER signaling, our RNA-seq analysis suggests that RAD140 may inhibit tumor growth via other AR-mediated mechanisms. Of note, RAD140 treatment suppressed genes implicated in DNA replication and cell-cycle progression, which were further suppressed by concurrent CDK4/6 inhibition. Gao and colleagues (41) described the role of AR as a tumor suppressor in castration-resistant prostate cancer (CRPC) cells by inhibiting DNA replication–related genes via the recruitment of hypophosphorylated Rb. In addition, the clinical and preclinical efficacy seen with high-dose androgen therapy in CRPC was also attributed to androgen-induced stabilization of AR, which prevents the relicensing of DNA for replication and subsequent cell-cycle progression (58, 59). In the current study, the AR-mediated suppression of ESR1 and its downstream pathway, along with the suppression of DNA replication- and cell cycle–related genes seen with RAD140 treatment suggest a distinct mechanism of action of this SARM in AR/ER⁺ breast cancer cells. The potent antitumor activity and tissue-selective AR activity, along with overall tolerability in animal models and oral availability together lend support to further clinical investigation of RAD140 in AR/ER⁺ breast cancer patients.

C.P. Miller and G. Hattersley hold ownership interest (including patents) in Radius Health. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Z. Yu, S. He, J.L. Brown, G. Hattersley, J.C. Saeh

Development of methodology: Z. Yu, S. He, H.K. Patel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Yu, S. He, H.K. Patel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z. Yu, S. He, D. Wang, J.L. Brown, G. Hattersley, J.C. Saeh

Writing, review, and/or revision of the manuscript: Z. Yu, S. He, D. Wang, C.P. Miller, J.L. Brown, G. Hattersley, J.C. Saeh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Yu, J.L. Brown

Study supervision: Z. Yu, J.L. Brown, J.C. Saeh

The authors would like to acknowledge Dr. Steven P. Balk of Beth Israel Deaconess Medical Center and Dr. Changmeng Cai of University of Massachusetts Boston for their comments on the manuscript. The authors thank Drs. Fiona Garner and Teeru Bihani, and Margaret Ward, of Radius Health, Inc., for their comments and suggestions on the project and the manuscript. The authors thank Dr. Jim Humphreys and the histopathology team at Precision Pathology (San Antonio, TX) for the IHC work.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.