AZD9496: An Oral Estrogen Receptor Inhibitor That Blocks the Growth of ER-Positive and ESR1-Mutant Breast Tumors in Preclinical Models

With over 70% of breast cancers expressing the estrogen receptor alpha (ERα), treatment with antihormonal therapies that directly antagonize ER function (e.g., tamoxifen) or therapies that block the production of the ligand, estrogen (e.g., aromatase inhibitors) are key to management of the disease both in adjuvant and metastatic settings (1, 2). Despite initial efficacy seen with endocrine therapies, the development of acquired resistance ultimately limits the use of these agents although the majority of tumors continue to require ERα for growth. Through the use of preclinical models and cell lines, changes in growth factor receptors have been implicated in driving resistance in cells, for example, receptor tyrosine kinases (HER2), EGFR, and their downstream signaling pathways, Ras/Raf/MAPK and PI3K/protein kinase B (AKT) signaling (3–5). In some instances, signaling from different growth factor receptor kinases leads to phosphorylation of various proteins in the ER pathway, including ER itself, leading to ligand-independent activation of the ER pathway (6).

In the metastatic setting, 500 mg fulvestrant, given as 2 × 250 mg intramuscular injections, has been shown to offer additional benefit over the 250 mg dose with improved median overall survival (7, 8). Recent analysis of presurgical data from a number of trials showed a dose-dependent effect on key biomarkers such as ER, progesterone receptor (PR), and Ki67 labeling index (9). Although fulvestrant is clearly effective in this later treatment setting, there are still perceived limitations in its overall clinical benefit due to very low bioavailability, the time taken to achieve plasma steady-state of 3–6 months and ultimately the level of ERα reduction seen in clinical samples (10, 11). It is widely believed that a potent, orally bioavailable SERD that could achieve higher steady-state free drug levels more rapidly would have the potential for increased receptor knockdown and lead to quicker clinical responses, reducing the possibility of early relapses (12). Moreover, the recent discovery of ESR1 mutations in metastatic breast cancer patients that had received prior endocrine treatments, where the mutated receptor is active in the absence of estrogen, highlights an important potential resistance mechanism. Preclinical in vitro and in vivo studies have shown a reduction in signaling from ESR1 mutants by tamoxifen and fulvestrant, but at higher doses than required to inhibit the wild-type receptor, which implies that more potent SERDs may be required to treat these patients or to use in the adjuvant setting to limit the emergence of mutations (13–15).

Achieving oral bioavailability has remained a key challenge in the design of ER downregulators and until now no oral estrogen receptor downregulators, other than GDC-0810 and RAD1901, have progressed into clinical trials, despite encouraging reports from others of preclinical activity in mouse xenograft models (16–21). Here, we describe a chemically novel, nonsteroidal ERα antagonist and downregulator, AZD9496, that can be administered orally and links increased tumor growth inhibition to enhanced biomarker modulation compared with fulvestrant in a preclinical in vivo model of breast cancer. Moreover, we show that AZD9496 is a potent inhibitor of ESR1-mutant receptors and can drive tumor growth inhibition in a patient-derived ESR1 mutant in vivo model (PDX). Hence, AZD9496 is anticipated to yield improved bioavailability and clinical benefit through enhanced ER pathway modulation.

All AZ cell lines were tested and authenticated by short-tandem repeat (STR) analysis. MCF-7 was obtained from DSMZ and last STR tested in February 2013. MDA-MB-361 (STR tested in October 2011), MDA-MB-134 (STR tested in October 2010), T47D (STR tested in June 2012), CAMA-1 (STR tested in March 2012), HCC1428 (STR tested in December 2011), HCC70 (STR tested in October 2013), LnCAP (STR tested in August 2011), MDA-MB-468 (STR tested in March 2014), BT474 (STR tested in July 2013), PC3 (STR tested in October 2012), HCT116 (STR tested in August 2011) were all obtained from the ATCC. HCC1428-LTED cell line (STR tested in November 2012) was kindly provided by Carlos Arteaga (Vanderbilt University, Nashville, TN; ref. 22). Cells were cultured in RPMI1640 media (Corning) supplemented with 10% heat-inactivated FCS or charcoal/dextran–treated CCS to deplete hormone and 2 mmol/L l-glutamine (Corning). Stable MCF-7 cell lines expressing FLAG-tagged ESR1-mutant proteins in pTRIPZ lentiviral vectors (GE Healthcare) under the control of the doxycycline-inducible Tet promoter were generated by lentiviral infection using standard techniques.

Binding, ER agonism, antagonism, downregulation, and cell proliferation assays were carried out as described previously (23). BIAcore affinity measurements and immunoblotting from compound-treated cells are described in Supplementary Methods.

Protein expression, purification, and crystallization of the ERα ligand-binding domain was carried out as described previously (24). X-ray diffraction data were collected at the ESRF on beamline ID23-1 on an ADSC detector. Data were processed using XDS (25) as implemented within EDNA (26) and scaled and merged using SCALA (27). The structure of ERα in complex with AZD9496 was solved by molecular replacement using AmoRE and an internal ERα structure as the search model (28). Quality checks on the protein structures were carried out using the validation tools in Coot (29). The final structure has been deposited in the Protein Databank with the ID code given in Supplementary Table S1.

Stable isotope labeling of amino acids (SILAC) assays were carried out as described previously (30) and detailed in Supplementary Methods. Compound-treated samples were analyzed by mass spectrometry using relative peptide quantification by selected reaction monitoring (SRM). Degradation half-life was measured using the one-phase exponential decay equation in GraphPad PRISM (Y = Span^.e^−K.X + Plateau), where X is time and Y is response, which starts out as Span + Plateau and decreases to Plateau with a rate constant K.

All studies involving animals in the United Kingdom were conducted in accordance with UK Home Office legislation, the Animal Scientific Procedures Act 1986, and AstraZeneca Global Bioethics policy or with US federal, state, and institutional guidelines in an Association for Assessment and Accreditation of Laboratory Animal Care–accredited facility. The rat uterine weight assay for the measurement of estrogenic activity was performed as described previously (31). Details of individual models, protein analysis of tumor fragments, and pharmacokinetic analysis of plasma samples are included in Supplementary Methods.

AZD9496 ((E)-3-(3,5-difluoro-4-((1R,3R)-2-(2-fluoro-2-methylpropyl)-3-methyl-2,3,4,9-tetrahydro-1H-pyrido[3,4-b]indol-1-yl)phenyl)acrylic acid; Fig. 1A) was identified through structure-based design and iterative medicinal chemistry structure–activity relationship (SAR) studies from a novel ER-binding motif (32). A directed ER-binding screen was established to identify novel binding motifs that could then be optimized for both cellular phenotype and potency while maintaining good pharmacokinetic properties (23). Unlike fulvestrant, and other well-known historical modulators of the ER receptor, including hydroxytamoxifen (Fig. 1B), AZD9496 lacks a phenolic moiety in which the phenol mimics the A-ring phenol group of the endogenous ligand estradiol but gains its potency through a different series of novel interactions with the ER protein (Fig. 1C). The indole NH picks up an interaction by forming a hydrogen bond to the carbonyl of Leu-346, with additional lipophilic interactions achieved by both the chiral methyl substituent, adjacent to the ring nitrogen, and the N-substituted isopropyl fluoro side-chain occupying a “lipophilic hole” in the ER protein. The acrylic acid side chain picks up an unusual acid–acid colocalization with Asp-351 in the helix-12 region of the ER that has been proposed to be crucial for achieving a downregulator–antagonist profile (33). AZD9496 showed high oral bioavailability across three species (F% 63, 91, and 74, rat, mouse, and dog, respectively) with generally low volume and clearance across species, albeit a higher clearance in mouse. The percent free levels in human plasma of 0.15% were 5-fold higher than those measured for fulvestrant.

The key characteristics of a SERD is the ability to both antagonize and downregulate ER protein without inducing any partial agonist effects in breast cells. A series of in vitro cell assays were developed specifically to measure ERα downregulation, agonism, and antagonism through simultaneous measurement of ER and PR protein levels in cells after treatment with AZD9496 (23). PR expression is known to be regulated by ER transcription and was measured as a marker of ER activation in cells (34). The potency of AZD9496 approached that observed with fulvestrant in achieving pmol/L IC₅₀ in ERα binding, downregulation, and antagonism and significant cell growth inhibition in hormone-depleted growth conditions. Adding 250 nmol/L tamoxifen significantly lowered the IC₅₀ value, as expected, if AZD9496 and tamoxifen are competing for binding to the ligand-binding domain (LBD; Table 1). The potent downregulation of ERα was also verified by immunoblotting techniques (Supplementary Fig. S1).

To understand the binding kinetics of AZD9496 to human ERα LBD, binding studies were performed in the presence of increasing concentrations of AZD9496 and association (k_ass) and dissociation (k_diss) rate constants were calculated. AZD9496 was calculated as having an apparent dissociation rate constant (k_diss) for ERα LBD of 0.00043/sec, at 25°C equating to a half-life (t_1/2) of 27 ± 2 minutes and the affinity derived from the ratio of the rate constants gave a K_d of 0.33 nmol/L (Supplementary Table S2). This is in good agreement with the binding IC₅₀ of 0.82 nmol/L and correlates well with previously published data on estrogen and tamoxifen binding to ER LBD (35). AZD9496 showed pmol/L equipotent binding to both ERα and ERβ isoforms, as expected, given their high sequence homology and similarity in tertiary structure in their LBDs (36), and highly selective binding compared with progesterone (∼650-fold), glucocorticoid (∼11,223-fold), and androgen (∼36,375-fold) receptor LBDs (Supplementary Table S3). Selectivity for ER was also evident by measuring cell growth inhibition in a panel of ER⁺ and ER⁻ cell lines (Supplementary Table S4). No significant growth inhibition was seen in any ER⁻ cell lines (EC₅₀ > 10 μmol/L) in contrast to the low nanomolar activity seen in a range of ER⁺ cell lines.

To measure the rate at which AZD9496 downregulated ERα, SILAC experiments were performed in MCF-7 cells using mass spectroscopy analysis to detect ERα protein levels after treatment with compounds. Addition of AZD9496 or fulvestrant increased the rate of degradation of the isotope-labeled ERα compared with DMSO control. Levels of newly synthesized ERα peptide were also reduced following AZD9496 and fulvestrant treatment, presumably due to ongoing degradation of newly synthesized ERα protein by compound present, whereas ERα protein levels continued to increase over time in the presence of tamoxifen or DMSO (Fig. 2). The t_1/2 of ERα was decreased from 3 hours, in the presence of DMSO, to 0.75 hours with 100 nmol/L AZD9496 and 0.6 hours with 100 nmol/L fulvestrant (Supplementary Fig. S2). Downregulation of ER occurs through the 26S proteosomal pathway as no decrease in ERα protein level in MCF-7 cells was seen in the presence of the proteosome inhibitor MG132 (Supplementary Fig. S3). In addition, ERα downregulation was shown to be reversible in compound wash-out experiments, where ERα levels increased in a time-dependent manner back to basal levels over a 48-hour period following compound removal (Supplementary Fig. S4).

In contrast to an AstraZeneca reference compound (AZ10557841), which has known ERα agonist activity in breast cells, AZD9496 was shown to cause no increase in PR protein levels but rather a slight decrease below the detected basal levels (Fig. 3A). Some endocrine therapies, such as tamoxifen, can act as partial ER agonists in certain tissues, for example, endometrium, due to conformational changes that take place at the ER and the resulting effect on cofactor recruitment and transcriptional gene activation from the tamoxifen-bound receptor (37). To test whether AZD9496 could act as a partial agonist in other tissues, agonist effects were measured in vivo in a previously validated immature female rat model designed to detect agonistic properties of test compounds by measuring increases in uterine weight (31). AZD9496, given once daily orally at 5 and 25 mg/kg produced statistically significant increases in uterine weight compared with the fulvestrant control (P < 0.001) but significantly lower than tamoxifen (P = 0.001; Fig. 3B). Histologic staining of uterine tissue samples also showed that the lengthening of the endometrial cells in the rat uteri appeared to have decreased compared with tamoxifen (Fig. 3C). In further studies, ERα protein levels were reduced with fulvestrant and AZD9496, but not tamoxifen, compared with relevant vehicle controls, whereas PR levels were significantly increased with tamoxifen alone (Fig. 3D).

The effect of chronic, oral dosing of AZD9496 was explored in MCF-7 human breast xenografts, as a representative ER⁺/PR⁺/HER2⁺ breast cancer model. Good bioavailability and high clearance gave a terminal t_1/2 of 5–6 hours after oral dosing in the mouse and resulted in significant dose-dependent tumor growth inhibition with 96% inhibition at 50 mg/kg and no toxicity or weight loss relative to the vehicle control group (Fig. 4A). To confirm that AZD9496 was targeting the ER pathway, PR protein levels were measured in tumor samples taken at the end of the study and a significant reduction in PR was seen, which correlated with tumor growth inhibition. A >90% reduction in PR was seen with both 10 and 50 mg/kg doses and a 75% decrease even with the 0.5 mg/kg dose, demonstrating that AZD9496 can clearly antagonize the ER pathway (Fig. 4B). Dosing 5 mg/kg of AZD9496, the minimal dose required to see significant tumor inhibition, gave greater tumor growth inhibition compared with 5 mg/mouse fulvestrant given 3 times weekly and tamoxifen given 10 mg/kg orally, daily (Fig. 4C). A series of pharmacodynamic in vivo studies were conducted to measure time taken to reach maximal inhibition of PR levels and time taken to recover back to basal levels. Three days of dosing with 5 mg/kg of AZD9496 gave 98% reduction of PR protein and continued to suppress protein levels at 48 hours (Fig. 4D), with full recovery by 72 hours (data not shown), which indicates a long pharmacologic half-life in vivo. Fulvestrant, given as 3 × 5 mg/mouse doses over one week, gave a 60% reduction in PR protein over a prolonged period with measured plasma levels approximately 8-fold higher than those achieved clinically, at steady state, with 500 mg fulvestrant (Fig. 4E). As estrogen itself is known to downregulate ER protein (38, 39), we were unable to detect further decreases in ERα protein compared with control animals with AZD9496 or fulvestrant in the MCF-7 model presumably due to the high circulating plasma levels of estrogen from implanted pellets at the time tumor samples were taken (Supplementary Fig. S5). A mouse-specific metabolite of AZD9496 was detected in circulating plasma at similar levels to AZD9496 and showed a similar pharmacokinetic profile. Testing this metabolite in the in vitro MCF-7 assays resulted in approximately 5-fold lower ERα antagonism activity and 7-fold lower ERα downregulation activity than AZD9496 (data not shown). Using a pharmacokinetic/pharmacodynamic model based on PR inhibition data, at the 5 mg/kg dose in vivo, which gives 98% inhibition of PR, the inhibitory activity that could be attributed to the parent compound alone was 85% when the activity of the metabolite was discounted.

To explore mechanistic effects on ER pathway regulation further, a gene expression analysis study was performed using tumor samples dosed at 10 mg/kg for 3 days with AZD9496 or tamoxifen and fulvestrant given as a single dose 5 mg dose subcutaneously. Tumors were taken 24 hours after the last dose of AZD9496 and tamoxifen and mRNA levels measured using a Human Transcriptome Array (HTA 2.0). A number of known ER-regulated genes were examined to see whether AZD9496 could antagonize these genes in a similar manner to fulvestrant and tamoxifen (40–42). The heatmap shows that AZD9496 could indeed downregulate the mRNA levels of estrogen-responsive genes (Supplementary Fig. S6A). As differences in the levels of mRNA downregulation in vivo could be due to different pharmacokinetic profiles of the molecules, the effects of compounds on the mRNA levels of a subset of estrogen-regulated genes were also measured in MCF-7 cells that had been treated with or without estrogen (Supplementary Fig. S6B–S6E). For two of these genes, TFF1 and AREG, tamoxifen was shown to increase mRNA levels in the absence of estrogen but not with either fulvestrant or AZD9496 (Supplementary Fig. S6D and S6E). Where transcript levels were increased to varying levels in the presence of estrogen, fulvestrant and AZD9496 inhibited transcript levels in a dose-dependent fashion compared with tamoxifen. No significant effects were seen in ESR1 mRNA levels in vitro, indicating that the protein downregulation described previously is due to downregulation of the protein and not a decrease in transcript levels per se (Supplementary Fig. S6F).

Acquired resistance to aromatase inhibitors and antiestrogens is driven through a variety of mechanisms, which include effects on the ER pathway itself, i.e., downregulation of ER expression and altered expression of ER coregulators, as well as activation of alternative prosurvival or proliferative pathways and cross-talk between growth factor receptors and ER pathways. This has led to a number of clinical trials with PIK3, mTOR, AKT inhibitors, as well as CDK4/6 inhibitors in combination with aromatase inhibitors or fulvestrant and early results have shown increased progression-free survival (PFS) in many cases (43). Combinations of AZD9496 with AZD2014 (dual mTORC1/2), AZD8835 (PIK3Cα/δ) and palbociclib (CDK4/6) inhibitors were tested in the MCF-7 in vivo model and in all cases tumor regressions seen with the combination arms alone compared with tumor stasis with monotherapy (Fig. 5A–C). AZD9496 was also tested in a long-term estrogen-deprived model (LTED), using the HCC-1428 LTED cell line that was previously adapted to grow in the absence of estrogen and as such represents a model of aromatase inhibitor resistance (22). Tumor regressions were seen with both AZD9496 and fulvestrant (Fig. 5D) and both compounds completed ablated ER protein levels in the end-of-study tumors (Supplementary Fig. S7).

The recent identification of ESR1 mutations in patients that had received one or more endocrine therapy treatments has led to the discovery that these mutations can drive cell proliferation in the absence of estrogen and may constitute a resistance mechanism to some endocrine agents. In vitro binding studies were performed using wt, D538G, and Y537S LBDs and both AZD9496 and fulvestrant were able to bind to mutant LBD with nmol/L potency although 2- to 3-fold reduced compared with wt (Table 2). The different IC₅₀ values obtained for binding of AZD9496 and fulvestrant to wt ERα compared with the value in Table 1 is likely due to different sources of protein used in the Lanthascreen kit assays versus the laboratory-derived wt and ESR1-mutant proteins. Downregulation of ER-mutant proteins was measured using MCF-7 doxycycline-inducible stable cell lines in vitro. ER protein levels increased on induction with 1 μg/mL doxycycline and led to increased levels of PR in the absence of estrogen. AZD9496 and fulvestrant were able to downregulate all mutant ER proteins and decreased PR levels, whereas tamoxifen appeared to stabilize ER protein levels and reduced PR levels to a lesser extent (Fig. 6A). To explore activity against an ESR1 mutant in vivo, tumor growth inhibition studies were performed in a PDX model, CTC-174, which harbors a D538G mutation in the LBD and constituted 31% of the transcripts variants identified by RNA-seq analysis. Tumors were grown from implanted tumor fragments either with or without estrogen and in both cases viable tumors grew. Given published data that higher doses of SERM/SERDs may be needed to inhibit ESR1 mutants (14, 15), 25 mg/kg of AZD9496 was used alongside 5 mg/mouse of fulvestrant, which was known to achieve higher pharmacokinetic levels in mouse than the current clinical dose. AZD9496 and fulvestrant inhibited tumor growth by 66% and 59% respectively, compared with tamoxifen, which gave 28% inhibition. Efficacy correlated with antagonism of the pathway with AZD9496 giving a 94% decrease in PR levels compared with 63% with fulvestrant (Fig. 6B and C). In the absence of estrogen, and given the significant PR changes seen with 25 mg/kg, the dose of AZD9496 was lowered but a 5 mg/kg dose still achieved growth inhibition of approximately 70% and led to a significant decrease of 73% in ERα protein levels at the end of the study (Fig. 6D and E). Given the similarities in growth inhibition either with or without estrogen, residual tumor growth is likely to be due to either additional mutations (e.g., PIKCA N345K) and/or growth factor signaling pathways involved in driving tumor growth in addition to the ER pathway.

Despite the enhanced clinical benefits of directly antagonizing and downregulating ER and thereby targeting any ligand-independent ER-driven pathways, the pharmacokinetic and IHC biomarker data for ER, PR, and Ki67 from clinical trials with fulvestrant would suggest that a plateau of effect has not been reached, and there is scope to further improve on the 50% reduction in ER levels seen after 6 months of treatment (8, 12). As one of the potential drawbacks associated with fulvestrant is the very low bioavailability, which is seen both in patients and in preclinical animal models, a potent, orally bioavailable SERD could have the potential to significantly antagonize ER activity and increase ERα receptor knockdown versus fulvestrant due to higher exposures. Our approach focused on identifying novel ER motifs with established drug-like properties that could then be optimized for cellular phenotype and potency and led to the identification of AZD9496, which showed picomolar potency in vitro in ER⁺ cell lines only. Previous structural analysis of a number of SERMs with ERα LBD identified key interactions with Asp 351, which was shown to be critical for the antagonist activity of antiestrogens such as raloxifene, lasofoxifene, and GW5638 through shielding or neutralization of the negative charge. Unlike tamoxifen, however, GW5638 repositioned residues helix-12 through its acrylate side chain, increasing the exposed hydrophobic surface of ERα, resulting in destabilization and ultimately 26S proteosomal degradation of the receptor. The acrylic acid moiety of AZD9496 is colocalized with the Asp 351 residue in the helix-12 region of ER in an unusual acid–acid interaction, and it is therefore proposed that AZD9496 also repositions residues associated with the helix-12 via specific contacts with the N-terminus region, resulting in a mixed antagonist and downregulator profile (33, 44). Evidence that AZD9496 was directly binding and downregulating the receptor at the same site as other antiestrogens was seen by the reduced potency observed on prior addition of tamoxifen to the downregulation assay (Table 1). Mechanistically, we have shown that AZD9496 has very similar effects on antagonism, ERα downregulation, and agonism to fulvestrant. Even where ERα was overexpressed in MCF-7 cells, unlike tamoxifen, AZD9496 was able to downregulate and antagonize the overexpressed receptor (Fig. 6A). Our transcriptional studies in both MCF-7 cells and the in vivo model also demonstrated antagonism of the transcriptional activity of the receptor as seen by decreased mRNA levels of ER-activated genes in a dose-dependent manner (Supplementary Fig. S6).

Unlike fulvestrant, AZD9496 did show some degree of agonism in endometrial tissue (Fig. 3). This effect was dose independent and would indicate that there are subtle differences in how AZD9496 interacts with ERα compared with tamoxifen and fulvestrant and therefore the extent to which tissue specific coactivators can interact with the receptor. Given the widespread use of tamoxifen in the adjuvant setting for 5–10 years, the low level of agonism seen with AZD9496 is not viewed as a deterrent for clinical development. A number of studies have looked at the incidence of endometrial cancer in patients taking tamoxifen over a number of years and despite the small increase in the incidence of endometrial cancer, the benefits of tamoxifen greatly outweigh any risks of developing uterine malignancy in postmenopausal patients (45, 46).

A pharmacokinetic/pharmacodynamic efficacy model was established, integrating diverse endpoints such as drug exposure, PR modulation, and inhibition of tumor growth (data not shown). The model reflected relative contributions of AZD9496 and its active mouse metabolite to the overall biomarker modulation and efficacy. Given the half-life of the parent drug and metabolite are around 5 hours, to explain the prolonged pharmacodynamic effect observed with AZD9496 at 48 hours, an indirect response model for PR was developed, which estimated a degradation rate constant of around 23 hours for PR. This modulation in PR was linked directly to tumor growth inhibition by a proportional decrease of the tumor growth constant, and correlated well with the wide range of measured efficacies.

We consistently saw enhanced tumor growth inhibition of AZD9496 compared with fulvestrant in the MCF-7 in vivo model (Fig. 4). This model contains high circulating levels of plasma estradiol over the dosing period of the study as a result of the estrogen pellets used to drive growth of the MCF-7 cells as a xenograft. With plasma levels between 750 and 1,000 pg/mL over day 21–35, this model is more representative of the premenopausal setting where estrogen levels can vary between 20 and 400 pg/mL (47). As the receptor binding data does not suggest a significant difference in affinity for the receptor between AZD9496 and fulvestrant, increased efficacy is likely due to the higher free drug levels of AZD9496 versus fulvestrant in this model. It would therefore be interesting to test the activity of potent oral SERDs, such as AZD9496, in the premenopausal setting where there is no enhanced risk of endometrial cancer observed for tamoxifen.

We were unable to detect a reduction in ER protein levels in the MCF-7 in vivo model with AZD9496 at any dose tested (data not shown). Having shown in vitro that estradiol can efficiently downregulate ERα protein, we postulate that downregulation of ERα by estradiol is already taking place in vivo given the high levels of circulating estrogen and that any additional changes by addition of a SERD are difficult to detect using the current methods. This was supported by measuring ERα downregulation in in vivo models, such as CTC-174 and HCC1428LTED, not requiring estrogen supplements, and showing potent downregulation of ERα by AZD9496 (Fig. 6E, Supplementary Fig. S7). The level of ERα reduction seen in the HCC1428 model is greater than seen clinically, with pharmacokinetic levels 5-fold higher than steady-state plasma levels of 500 mg fulvestrant, and supports the concept that if increased plasma levels of an active SERD agent could be achieved clinically, greater ERα knockdown may be possible. Preclinically, this appears to translate to enhanced tumor growth inhibition.

Achieving increased serum levels of a biologically active SERD could be particularly beneficial in the recently identified ESR1-mutant patient setting. Preclinical in vivo studies using PDX models derived from ESR1-mutant patients grown in the absence of estrogen supplementation have shown that high doses of fulvestrant (>500 mg/kg) can lead to tumor regression and significant decreases in ER protein levels but at significantly higher doses than could be achieved clinically (48). Our preclinical data would indicate that both AZD9496 and fulvestrant can potently bind and downregulate D538G and Y537C/N/S ERα proteins in vitro and significant tumor growth inhibition rather than regression seen in vivo. This is in contrast to published work showing relatively little downregulation of Y537N mutant protein in vitro with increasing doses of fulvestrant (15) but could be attributed to different expression systems and cell lines used for the analysis. It is notable that the doxycycline-inducible mutant receptors do not appear to be constitutively downregulated in a similar manner to wt ER exposed to estradiol (Supplementary Fig. S1). This may be due to the level of overexpression of the receptor in each stable cell line and it will be interesting to compare these findings from engineered expression systems to those from genetically altered mutant cell lines where expression is under control of the native promoter. Although very high doses of both compounds were not tested in the PDX CTC-174 ESR1-mutant model, the lack of any significant dose response might indicate the need to inhibit other growth factor–activated pathways in addition to the ER pathway to achieve significant antitumor growth. As trials commence to explore treatment of ESR1-mutant patients with new oral SERDs, data will hopefully emerge that will indicate whether lack of response is due to inability to fully inhibit mutated ER receptors or whether, in these advanced metastatic patients, other mutations and pathways are or become dominant in driving tumor growth.

In metastatic breast cancer, the median survival time is still around two years as a result of acquired endocrine resistance, highlighting the need for additional therapies (43). This has led to a number of clinical trials using PI3K–Akt–mTOR pathway inhibitors in combination with AIs as well as cyclin-dependent kinase inhibitors such as palbociclib, which was recently approved for combination treatment with letrozole in the metastatic breast cancer. Results from the phase III BOLERO-2 trial also demonstrated the PFS benefits of a steroidal AI plus everolimus in patients that had progressed on a nonsteroidal AI (49). More recently, further trials have been initiated using fulvestrant as the combination endocrine partner, which will allow analysis and understanding of any additional benefit from combination with a SERD versus AIs in this setting. Our data clearly demonstrated enhanced combination effects on tumor growth of an oral SERD agent with either dual mTORC1/2, PIKCα/δ or CDK4/6 inhibitors and furthermore opens up the possibility of using a combined endocrine and targeted inhibitor approach with longer term dosing in the adjuvant setting, assuming emerging data from ongoing metastatic trials supports the testing of combinations in early breast disease. The identification of a potent, orally bioavailable compound such as AZD9496 is a step forward in the next generation of SERD agents to undergo clinical evaluation as a future monotherapy or combination partner of choice.

J.O. Curwen is an associate principal scientist at AstraZeneca. G. Davies has ownership interest (including patents) in AstraZeneca as a shareholder. S. Powell is a team leader at Astrazeneca. G. Richmond has ownership interest (including patents) in AstraZeneca. All authors are current or former AstraZeneca employees and shareholders.

Conception and design: H.M. Weir, R.H. Bradbury, M. Lawson, A.A. Rabow, J.O. Curwen, R.A. Norman, G.E. Walker, C. Sadler, G. Richmond, C. De Savi

Development of methodology: H.M. Weir, R.H. Bradbury, M. Lawson, C.S. Donald, L.J.L Feron, T. Moss, S.E. Pearson, G. Davies, G.E. Walker, S. Powell, C. Sadler, B. Ladd

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.M. Weir, M. Lawson, R.J. Callis, M. Hulse, C.S. Donald, P. MacFaul, T. Moss, R.A. Norman, S.E. Pearson, M. Tonge, G. Davies, G.E. Walker, Z. Wilson, R. Rowlinson, B. Ladd, E. Pazolli, A.M. Mazzola

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.M. Weir, M. Lawson, A.A. Rabow, D. Buttar, R.J. Callis, J.O. Curwen, C. de Almeida, P. Ballard, M. Hulse, P. MacFaul, R.A. Norman, M. Tonge, G. Davies, G.E. Walker, Z. Wilson, R. Rowlinson, S. Powell, G. Richmond, A.M. Mazzola, C.M. D'Cruz

Writing, review, and/or revision of the manuscript: H.M. Weir, M. Lawson, D. Buttar, P. Ballard, C.S. Donald, R.A. Norman, S.E. Pearson, M. Tonge, G. Davies, G.E. Walker, C. Sadler, G. Richmond, B. Ladd, A.M. Mazzola, C.M. D'Cruz, C. De Savi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.M. Weir, M. Lawson, P. Ballard, S.E. Pearson, S. Powell, A.M. Mazzola

Study supervision: H.M. Weir, J.O. Curwen, R.A. Norman, G.E. Walker, C.M. D'Cruz

Other (chemistry route development): G. Karoutchi

The authors thank Elaine Hurt, Haifeng Bao, Sanjoo Jalla for access to the CTC-174 in vivo model developed at MedImmune.The authors also thank Paul Hemsley and Emma Linney for providing ER-mutant binding assay data, Helen Gingell for supporting the crystallography work and Zara Ghazoui and Henry Brown for the transcript data and analysis at AstraZeneca. The authors also thank Carlos Arteaga for use of the HCC1428-LTED cell line.

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.