Discovery of Selective Estrogen Receptor Covalent Antagonists for the Treatment of ERα^WT and ERα^MUT Breast Cancer

Breast cancer is the second leading cause of cancer mortality among women worldwide, with more than 1 million new cases diagnosed each year and approximately 400,000 deaths annually (1). Nearly 70% of breast cancers express estrogen receptor alpha (ERα; encoded by ESR1), a key hormone-regulated transcription factor for normal and malignant breast cells. Nonclinical and clinical–epidemiologic studies highlight an important oncogenic role for ERα in the genesis and progression of breast cancer (2). Several ER-directed therapies have been developed to antagonize oncogenic ERα function including selective ER modulators (SERM; tamoxifen), selective ER downregulators (SERD; fulvestrant), selective nonsteroidal aromatase inhibitors (NSAI; anastrozole and letrozole), and steroidal aromatase inhibitors (exemestane; ref. 3), for use in women with locally advanced, recurrent, or metastatic cancer. Although these therapies have demonstrated antitumor efficacy in the clinic, innate and acquired resistance remains a major challenge. Several mechanisms of resistance to endocrine therapies have been identified, including “cross-talk” between ERα and other kinases, particularly HER2 (4); dysregulation of apoptosis- and cell-cycle–related regulators (5); aberrant expression of ERα coactivators/corepressors (6); and, most recently, recurrent mutations in ERα (ERα^MUT; refs. 7–9). The hotspot mutations in ERα, which are enriched in nearly 30% of endocrine therapy–resistant metastases, induce ligand-independent activation of the ERα pathway (7–12), confer partial resistance to existing classes of endocrine therapies, and are associated with more aggressive disease biology with shorter overall survival in patients relative to wild-type ESR1 (13). Because current endocrine therapies are only partially effective in the ERα mutant setting, and a significant proportion of endocrine therapy–resistant metastases continue to remain dependent on ERα signaling for growth/survival, there remains a critical need to develop the next generation of ERα antagonists that can overcome aberrant activities of both wild-type ERα (ERα^WT) and ERα^MUT.

Herein, we report the discovery of a covalent class of ERα antagonist referred to as selective estrogen receptor covalent antagonist (SERCA) that inactivates both ERα^WT and ERα^MUT by targeting a unique cysteine residue (C530) that is not conserved among the other steroid hormone receptors. H3B-5942 enforces a unique antagonist conformation in ERα and demonstrates increased ERα antagonism compared with standard-of-care (SoC) therapies in preclinical models. We further show that the activity of H3B-5942 can be improved in combination with CDK4/6 or mTOR inhibitors, suggesting a potential combination strategy for SERCAs in the clinical setting.

It has previously been shown that hotspot ERα mutations identified in endocrine refractory metastatic breast cancer promote constitutive activity in ERα (7–12). Consistent with previous reports, we also noted estradiol (E2)-independent ERα activity conferred by the hotspot mutations Y537S/C/N and D538G (Supplementary Fig. S1A–S1B). Among the hotspot mutations, the Y537S mutant demonstrated the greatest E2-independent reporter activity (Supplementary Fig. S1A–S1B). To further assess the functional role of ERα mutations in promoting resistance to currently marketed endocrine therapies, the ERα^WT-positive MCF7 breast cancer line was engineered to ectopically express ERα hotspot variants (Fig. 1A). Consistent with earlier data (Supplementary Fig. S1A–S1B), ectopic expression of ERα^Y537S or ERα^D538G constitutively activated ERα function as determined by an increased E2-independent expression of ERα target genes GREB1 and TFF1 (Supplementary Fig. S1C). In regular culture medium (10% FBS), treatment with the ERα antagonists raloxifene or fulvestrant led to a reduction in ERα pathway activity in control cells (vector and ERα^WT), whereas higher pathway activity persisted in ERα^Y537S- and ERα^D538G-expressing cells (Fig. 1B; Supplementary Fig. S1D). Importantly, enhanced pathway activity in mutant ERα–expressing lines was associated with partial resistance to the antiproliferative effects of 4-hydroxytamoxifen (4-OHT) and fulvestrant (Fig. 1C), confirming the functional role of ERα mutations as drivers of partial resistance to SoC endocrine therapies. Of note was the observation that the most constitutively activating mutation, Y537S, was also associated with the greatest resistance phenotype following 4-OHT treatment (Fig. 1C). In summary, ERα mutations constitutively activate ERα function and promote partial resistance to antiestrogen therapies.

Mechanistically, the Y537 and D538 mutations are located in the AF-2 helix of ERα and introduce a stabilizing interaction to shift the dynamic equilibrium toward the agonist conformation even in the absence of ligand (14, 15). We reasoned that for a compound to be more active in the mutant setting, it would necessarily have to overcome the stabilizing effects of the mutation and shift the equilibrium toward the destabilized antagonist conformation. In principle, manipulating this dynamic equilibrium could be achieved through enhancing the compound potency against the receptor, through either covalent or noncovalent interactions, or a combination of both, in a primary or allosteric binding site. By targeting a nonconserved cysteine at the 530 position at the end of helix H11 that presents in the ligand-binding pocket of ERα when the receptor is in the antagonist conformation (Fig. 1D), our drug-discovery efforts ultimately led to the identification of H3B-5942 (Fig. 1E).

H3B-5942 was confirmed to engage covalently with ERα by intact mass spectrometry with >95% covalent modification observed for ERα^WT and ERα^Y537S ligand-binding domains (LBD) after overnight incubation at 4°C (Fig. 1F). To confirm C530 as the site of modification, intact mass spectrometry was performed with ERα^WT and ERα^Y537S LBDs harboring a cysteine to serine substitution at position 530 (ERα^C530S and ERα^{C530S, Y537S}, respectively). As expected, H3B-5942 was found to specifically engage with C530, as no covalent adduct was detected for ERα harboring the C530S mutation and the abundance of unmodified proteins was equal to that observed for DMSO controls (Fig. 1F).

To further confirm covalent engagement at C530 and to better understand the structural impact of H3B-5942 binding on ERα conformation, we determined the cocrystal structure of ERα^Y537S bound to H3B-5942 at 1.89 Å resolution (PDB ID 6CHW; Supplementary Table S1; Fig. 1G). The continuous electron density between C530 and the Michael acceptor confirmed covalent engagement. Moreover, the crystal structure shows the receptor adopts the desired antagonist conformation in the mutant setting. The indazole core is anchored at the bottom of the pocket through an H bond to E353, and the flexible linker extends upward from the core to C530. An internal H bond within the linker side chain helps orient the Michael acceptor with respect to the cysteine. Overall, these data confirm H3B-5942 as an ERα-targeting small molecule that specifically engages the C530 residue covalently to enforce the antagonist conformation.

To profile the impact of covalent engagement on cellular ERα protein, we first performed western blot analysis for ERα in the ERα^WT-expressing MCF7 parental line and the ERα^Y537S/WT-expressing ST941 patient-derived xenograft (PDX) cell line (PDX-ERα^Y537S/WT; Supplementary Fig. S2), following 24-hour treatment with fulvestrant, 4-OHT, or H3B-5942. In contrast to fulvestrant, which degraded ERα, H3B-5942 and 4-OHT treatments showed no alteration (MCF7 parental cell line) to slight elevation (ST941 cell line) in ERα protein levels, suggesting that the latter two compounds may maintain the pool of ERα in the antagonist conformation (Fig. 2A).

Next, we evaluated the influence of SERCA binding on the conformational change of ERα in cells. It is well accepted that conformational changes of ERα in response to ligand binding can have a significant impact on its biological function through influencing sites that are ultimately accessible for coregulator binding (16, 17). We applied a mammalian 2-hybrid–based reporter assay with a panel of nuclear receptor coactivators and conformationally sensitive peptides to evaluate their ability to interact with H3B-5942-bound ERα (refs. 18–20; Fig. 2B; Supplementary Fig. S3; Supplementary Table S2). Compared with wild-type ERα, endocrine-resistant ERα^Y537S and ERα^D538G demonstrated a remarkably higher interaction with LXXLL-containing coactivators/peptides in the absence of E2, suggesting that the constitutive activity of mutant ERα is likely mediated by ligand-independent recruitment of coactivators (Supplementary Fig. S3). All of the ERα antagonists tested at 10 μmol/L, including H3B-5942, substantially reduced binding of these LXXLL-containing peptides to both ERα^WT and ERα^MUT (Supplementary Fig. S3). A noteworthy observation was that while SERMs and SERDs induced interactions between ERα^WT and the conformational peptides specific for the corresponding antagonists (17), the H3B-5942–ERα complex did not recruit any of the tested peptides, indicating that H3B-5942 induces a distinct conformation in ERα (Fig. 2B). Similarly, H3B-5942–bound ERα^Y537S and ERα^D538G also failed to interact with conformationally sensitive peptides that are specifically recruited by the SERM- or SERD-bound mutant ERα (Fig. 2B), further confirming a potentially unique conformational alteration induced by H3B-5942 that would represent a novel antagonistic mechanism for both wild-type and mutant ERα. Because it is well appreciated that the conformation of ERα is an essential determinant of its biological activity, these results imply that H3B-5942 may impart some divergent responses in ERα relative to SoC and other agents in the SERD/SERM class.

To interrogate the potential transcriptional impact of H3B-5942, we investigated its influence on genome-wide DNA-binding modes of ERα in MCF7 cells using chromatin immunoprecipitation followed by high-throughput DNA sequencing (ChIP-seq). This analysis revealed that the peak overlap of global ERα binding sites was highly consistent across different compound treatments, including E2, 4-OHT, fulvestrant, and H3B-5942 (Supplementary Fig. S4A). The read distribution correlation analysis indicated that the ERα-binding profile of H3B-5942 treatment was more related to that of 4-OHT treatment (Supplementary Fig. S4B). Moreover, motif enrichment analysis revealed that the classic estrogen response element (ERE) AGGTCA-N(3)-TGACCT was the top binding consensus for H3B-5942 (Supplementary Fig. S4C). Together, these data suggest that H3B-5942 induces a similar global DNA-binding pattern as other ERα ligands.

Next, we explored the gene expression profiles of H3B-5942 relative to SoCs under E2-containing (antagonist mode) and E2-depleted (agonist mode) conditions. RNA-sequencing (RNA-seq) analysis of parental MCF7 breast cancer cells cultured in media supplemented with FBS and treated with various compounds for 6 days revealed a large subset of genes commonly suppressed by all compounds, enriched for pathways involved in cell-cycle and estrogen signaling (Supplementary Fig. S5; Supplementary Tables S3 and S4). Interestingly, whereas fulvestrant regulated the expression of a number of unique genes not affected by other treatments, H3B-5942 treatment induced gene expression changes more similar to 4-OHT under these conditions (Supplementary Fig. S5A and S5B). These data indicate that H3B-5942, similar to the SERDs/SERMs profiled, can potently suppress ERα-dependent transcription in breast cancer cells.

As a SERM, 4-OHT can stimulate ERα signaling under certain contexts, such as in the absence of estradiol, in the normal uterus and the Ishikawa endometrial carcinoma line (21). To assess for potential SERM activity of H3B-5942 in the uterine tissue, H3B-5942 was dosed once daily for 3 days at 10, 30, and 100 mg/kg in postnatal day (PND) 19 Sprague–Dawley rat pups. In agreement with the global transcriptional studies above, H3B-5942 was comparable with SERM tamoxifen as both compounds similarly increased the absolute (data not shown) and relative [to body weight (BW)] uterine weights, and enhanced endometrial epithelium thickness (Supplementary Fig. S6).

Having demonstrated SERM activity of H3B-5942 in the normal uterus, we next aimed to investigate H3B-5942 activity in the endometrial carcinoma setting. As expected, 4-OHT was able to induce ERα activity in Ishikawa carcinoma cells, noted by the significant increase in PGR expression and enhanced proliferation (Fig. 2C–E). Consistent with its role as a pure antagonist, treatment with fulvestrant significantly suppressed PGR expression at all concentrations. Interestingly, although fulvestrant did not induce proliferation at low concentrations, it did have a stimulatory effect at 300 nmol/L, perhaps due to induction of nongenomic pathways (Fig. 2C–E). In contrast to 4-OHT, H3B-5942 did not robustly affect PGR expression or induce proliferation of Ishikawa cells at any concentration tested (Fig. 2C–E). In summary, these data suggest that although H3B-5942 shares features of SERMs, unique context-dependent functions are also observed, possibly driven by the unique ERα conformation induced upon H3B-5942 binding.

We next aimed to define the role covalent engagement plays in influencing ERα antagonist activity. To enable these studies, we synthesized the saturated analogue of H3B-5942, denoted here as H3B-9224, which lacks the electrophilic Michael acceptor (Fig. 3A). Comparison of the cocrystal structure of ERα^Y537S-H3B-9224 (PDB ID 6CHZ; Supplementary Table S1) with H3B-5942 shows similar binding modes for the pharmacophore and similar conformations for Helix-12. However, the linker conformations appear to diverge, showing a higher degree of flexibility in that part of H3B-9224. In contrast to H3B-5942, a potential interaction between the secondary amine group of H3B-9224 and ERα-D351 was observed, similar to the conventional SERMs such as 4-OHT (PDB ID 3ERT) and raloxifene (PDB ID 1ERR; Fig. 3A).

We next questioned whether the different linker conformations for H3B-5942 and H3B-9224 would affect residence time on ERα and, eventually, downstream antagonist activity. Given that we previously showed highly selective engagement of ERα C530 by H3B-5942 (Fig. 1F), ERα proteins containing the C381S and C417S double mutation were used for further kinetic characterization as they showed greater stability, allowing detailed in vitro characterization (22, 23). The K_d values of estradiol for ERα^WT and ERα^Y537S were determined to be 0.144 and 0.030 nmol/L, respectively, suggesting a 5-fold higher affinity of ERα^Y537S for estradiol (Supplementary Fig. S7A). These values are in agreement with a recent study (24) but are significantly lower than values published in previous studies (14, 25), likely due to differences in experimental design. Based on the precise determination of the affinity for estradiol, we determined that the K_i value for noncovalent H3B-9224 was ∼2-fold higher than 4-OHT for ERα^WT and ERα^Y537S (Supplementary Fig. S7B).

To estimate the residence time of various ERα-binding compounds, jump dilution experiments were performed using initially saturating concentrations of binders. An excess of ³H-estradiol was then used to titrate the number of ERα sites that were no longer bound to the initial binder after dilution. This experiment demonstrated that H3B-9224 and 4-OHT were competed off ERα^WT and ERα^Y537S similarly, whereas H3B-5942 was essentially irreversibly bound due to the covalent nature of its interaction (Fig. 3B). Thus, the longer residence time of H3B-5942 could conceivably play a role in its antagonistic activity toward wild-type and mutant ERα.

Having confirmed the irreversible binding of H3B-5942, we next examined the functional relevance of covalent engagement on ERα conformation, protein abundance, and transcriptional activity. Interestingly, H3B-9224 binding to ERα also induced a novel conformation relative to known ERα–ligand complexes using conformationally sensitive peptides (Supplementary Table S2; Supplementary Fig. S8A). However, it is unclear whether different conformational states in ERα may be induced between H3B-5942 and H3B-9224, because conformation-specific probes are currently not available for either of these ERα–ligand complexes. Importantly, although H3B-9224 did not influence overall ERα abundance in cells (Supplementary Fig. S8B), weaker cellular activity was noted (Fig. 3C and D). Gene expression analysis of two independent ERα target genes, GREB1 and TFF1, consistently showed reduced potency of H3B-9224 relative to H3B-5942 across both ERα^WT and ERα^MUT lines (Fig. 3C). Consistent with the gene expression analysis, H3B-9224 also showed weaker antiproliferative activity across ERα^WT and ERα^MUT lines, confirming the biological importance of covalent engagement for ERα antagonism (Fig. 3D). These data suggest that the increase in potency observed for H3B-5942 in the cellular setting is driven, at least in part, by covalent binding.

To further confirm the functional importance of covalency, we performed proliferation assays in MCF7 lines engineered to express ERα^WT or ERα^Y537S with C530S point mutations (Fig. 3D; Supplementary Fig. S8C). The C530S point mutation in ERα^WT (ERα^C530S) did not hinder basal or estradiol-mediated ERα activity (Supplementary Fig. S8D). Proliferation assays demonstrated that the C530S mutation in ERα conferred reduced activity to H3B-5942 but not H3B-9224 (Fig. 3D; Supplementary Fig. S9), reinforcing the observation that H3B-5942 activity is critically dependent on covalent engagement with C530 for ERα antagonism in breast cancer lines, a phenotype that may be potentially driven by a combination of unique conformation (Fig. 2B) and enhanced residence time on ERα (Fig. 3B).

Because we previously confirmed that H3B-5942 demonstrates weaker agonism relative to 4-OHT in Ishikawa endometrial carcinoma cells (Fig. 2C–E), we next aimed to profile ERα agonism induced by H3B-9224 relative to H3B-5942. H3B-9224 demonstrated modestly higher agonistic activity relative to H3B-5942 as assessed by expression analysis of a subset of E2-regulated genes, including PGR (Supplementary Fig. S10A and S10B). Collectively, these data imply that covalent engagement with C530 may enhance ERα antagonism while concomitantly tuning down agonist activity in carcinoma cells.

Having demonstrated a significant potency gain for H3B-5942 through covalent engagement, we next aimed to assess the antagonist potential of H3B-5942 relative to SoC and experimental agents in breast cancer models. H3B-5942 showed a high apparent affinity for both ERα^WT and ERβ^WT comparable with SoCs and GDC-0810 (Supplementary Fig. S11A). To study the impact of high-affinity ERα binding on coregulator recruitment, we extended our previous analysis (Supplementary Fig. S3) to assess coregulator interactions globally using two independent methods (Supplementary Fig. S11B and S11C). FRET-based analysis revealed (i) an increase in CoA peptide recruitment to ERα^WT following E2 treatment relative to the apo form, (ii) higher basal recruitment of CoAs to ERα^Y537S relative to ERα^WT in the absence of ligand supporting the constitutive activity observed for ERα^MUT, and (iii) that H3B-5942 broadly suppresses CoA recruitment with little impact on CoR interaction (Supplementary Fig. S11B). As further validation of the FRET-based coregulator interaction profiling, we screened WT ER coregulator interactions using the Microarray Assay for Real-time Coregulator–Nuclear receptor Interaction (26). These data support the TR-FRET analysis (Supplementary Fig. S11C), confirming that H3B-5942, in these settings, primarily prevents CoA recruitment with minimal influence on CoR interactions.

To quantitatively determine the impact of various treatments on CoA recruitment, an enzyme-fragment complementation-based cellular assay was performed to measure the interaction of ERα with the CoA PGC1α. Although all treatments significantly suppressed recruitment of PGC1α, H3B-5942 was superior to both 4-OHT and GDC-0810 (Supplementary Fig. S11A). This potent inhibition in the recruitment of PGC1α was selective to ERα^WT, as recruitment to other steroid hormone receptors was not significantly affected by H3B-5942 (Supplementary Fig. S11A).

As recruitment of CoAs to ERα is a prerequisite for ERα activity, we hypothesized that the potent inhibition in CoA recruitment by H3B-5942 would lead to enhanced suppression in cellular ERα signaling, comparable with or superior to existing classes of ERα antagonists. Indeed, H3B-5942 showed a significant dose-dependent suppression of the ERα target gene GREB1 in MCF7-ERα^WT, various MCF7-ERα^MUT lines, and the PDX-ERα^Y537S/WT line, which was superior to that observed for 4-OHT and GDC-0810 (Fig. 4A). Consistent with the transcriptional profile, H3B-5942 also exhibited a dose-dependent reduction in proliferation superior to 4-OHT and GDC-0810 (Fig. 4B).

We next further characterized H3B-5942 in the endocrine-therapy–resistant setting in MCF7 cells resistant to long-term estrogen deprivation (LTED), which either expressed ERα^WT (MCF7-LTED-ERα^WT) or spontaneously gained the ERα^Y537C mutation (MCF7-LTED-ERα^Y537C; ref. 27). In the absence of E2, H3B-5942 showed no significant impact on ER-mediated transcription in the MCF7-Parental (endocrine therapy–sensitive) and MCF7-LTED-ERα^Y537C lines but did result in a 1.5-fold (P = 0.03) increase in the MCF7-LTED-ERα^WT line. In contrast, 4-OHT showed no significant impact on ER-mediated transcription in all cell lines tested, whereas fulvestrant caused a 30% reduction in both MCF7-Parental and MCF7-LTED-ERα^WT lines (Supplementary Fig. S12). In the presence of E2, H3B-5942 showed a significant dose-dependent decrease in ER-mediated transactivation in all cell lines tested, comparable with fulvestrant and superior to 4-OHT (Supplementary Fig. S12). Consistent with this reduction in ER activity, H3B-5942 caused a concentration-dependent decrease in proliferation in all cell lines tested with GI₅₀ values of 0.5, 2, and 30 nmol/L in the MCF7-Parental, MCF7-LTED-ERα^WT, and MCF7-LTED-ERα^Y537C lines, respectively (Supplementary Fig. S12). Collectively, the CoA recruitment, ERα target gene expression, and proliferation data confirm in vitro activity of H3B-5942 across a range of models of endocrine-therapy resistance.

Based on the significant antiproliferative activity of H3B-5942 in breast cancer cell lines in vitro, we next aimed to assess its antitumor activity across various cell line–derived and PDX tumor models harboring ERα^WT or ERα^MUT. H3B-5942 dosed once (q.d.×1) orally at 30 to 300 mg/kg showed a dose-proportional increase in plasma and tumor exposure and a concomitant dose-proportional decrease in expression of the ERα target genes PGR and NPY1R in the ERα^Y537S/WT ST941 tumor model (Supplementary Fig. S13). Single or repeat dosing of H3B-5942 at 200 mg/kg suppressed a large panel of direct ERα target genes (ref. 28; Supplementary Fig. S14), with q.d.×1 dosing maintaining target gene suppression for up to 72 hours after dose, and q.d.×3 (3 daily doses) dosing demonstrating greatest suppression in PGR and NPY1R (Supplementary Fig. S14).

We first evaluated the efficacy of H3B-5942 administered orally in the MCF7 xenograft model in athymic female nude mice. Once-daily oral administration of 1, 3, 10, or 30 mg/kg H3B-5942 resulted in dose-dependent tumor growth inhibition [tumor growth inhibition (TGI) on day 17 of 19%, 41%, 68%, and 83%, respectively; Supplementary Fig. S15]. Whereas the TGI achieved by subcutaneous (s.c.) tamoxifen treatment at 1 mg/mouse every other day was comparable with that achieved with H3B-5942 (TGI of 80%), once-weekly (q.w.) s.c. fulvestrant treatment at 5 mg/mouse showed reduced TGI (TGI of 47%). In addition to the ERα^WT MCF7 model, H3B-5942 demonstrated significant efficacy across a panel of ERα^WT PDX models, inducing stasis or regression (Supplementary Fig. S16). As expected, no significant TGI was observed in ERα-negative PDX models (Supplementary Fig. S16).

Next, we evaluated the antitumor activity of H3B-5942 administered orally q.d. in the ERα^Y537S/WT ST941 model in athymic female nude mice. Oral q.d. administration of 3, 10, 30, 100, and 200 mg/kg H3B-5942 resulted in dose-dependent inhibition of tumor growth (TGI on day 35 of 3%, 35%, 63%, 76%, and 81%, respectively; Fig. 4C; Supplementary Fig. S17A), and dose-dependent suppression in the ERα target genes PGR (PR) and SGK3, and the proliferation marker Ki67 in endpoint tumors (Fig. 4D; Supplementary Fig. S17B–S17C). Although tamoxifen and fulvestrant also inhibited tumor growth (TGI of 70% and 30%, respectively), H3B-5942 demonstrated superior activity relative to fulvestrant (Fig. 4C; Supplementary Fig. S17A). Consistent with the significant antitumor activity in the ST941 model, H3B-5942 also demonstrated superior activity over SoCs in an independent ERα^Y537S/− endocrine therapy–resistant WHIM20 model (Supplementary Fig. S18). Collectively, in vivo profiling confirmed H3B-5942 has significant antitumor activity in both the ERα^WT and ERα^MUT setting at well-tolerated doses (Supplementary Fig. S19).

Although H3B-5942 monotherapy showed potent inhibitory activity in both ERα^WT and ERα^MUT backgrounds, we next aimed to identify cellular pathways that, upon coinhibition, might further enhance SERCA potency. To this end, we performed an unbiased in vitro combination screen in the PDX-ERα^Y537S/WT breast cancer cell line using a fixed concentration of H3B-5942 at 1 μmol/L and multiple doses of a panel of reference compounds, including marketed drugs, compounds currently in clinical trials, and tool compounds. The panel consisted of 1,356 small-molecule inhibitors with 1,198 unique structures (11.6% sample redundancy) and 334 unique protein targets. After 6 days of exposure to compounds, cell viability/proliferation was determined, and the delta-AUC was calculated for each combination. Data are presented as a waterfall plot (Fig. 5A) showing the combination results as the fold change in standard deviations from the median for the middle 1,000 results. The FDA-approved CDK4/6 inhibitors palbociclib (n = 4), ribociclib (n = 2), and abemaciclib (n = 2) as well as several mTOR inhibitors clustered as highly synergistic combinations (Fig. 5A, inset and table; Supplementary Figs. S20–S21). To confirm the high-throughput combination screening results, traditional matrix-style combination tests were performed. As expected, large areas of synergy were observed with clear effects above 10 to 25 nmol/L for H3B-5942 and 25 pmol/L and above for the CDK4/6 inhibitors across multiple cell lines bearing ERα^WT or clinically frequent ERα mutations (Fig. 5B; Supplementary Fig. S20B). The enhanced antiproliferative potency observed in vitro was supported by greater suppression of proliferation markers MKI67 and CDC25A in the combination setting relative to either monotherapy alone (Supplementary Fig. S22).

To confirm the in vitro combination activity with CDK4/6 inhibition, we administered H3B-5942 alone or in combination with palbociclib and monitored efficacy in the MCF7 and ST941 tumor models. Consistent with the in vitro analysis, greater efficacy was achieved in the combination setting in both the ERα^WT and ERα^MUT models (Fig. 5C and D; Supplementary Fig. S23). In aggregate, these data suggest that although H3B-5942 as a monotherapy is very potent, combination therapy with agents that target CDK4/6 or potentially mTOR can further improve potency.

Endocrine therapy is a standard treatment option for patients with ER-positive breast cancer. Currently, the majority of patients with breast cancer with localized disease will experience long-term disease-free survival. Unfortunately, the clinical effectiveness can be limited because of high rates of intrinsic and acquired drug resistance during treatment. Those patients who present with or develop endocrine refractory metastatic diseases have a 5-year survival of less than 25% and are currently incurable (29). Although it appears that breast cancer develops resistance to endocrine therapies through multiple molecular processes, reactivation of ERα remains a prominent mechanism. The recent discoveries of hotspot ERα mutations in metastatic breast cancer further strengthen the key role of ERα in drug resistance and disease development. The recurrent mutations, mainly occurring on tyrosine-537 and aspartic acid-538 located in the LBD, are enriched in nearly 30% of endocrine therapy–resistant metastases and are associated with more aggressive disease biology with shorter overall survival relative to the wild-type ERα (13). Furthermore, ERα mutations induce ligand-independent activation of the ERα pathway and confer partial resistance to existing classes of endocrine therapies (7–13, 30). Therefore, there is a critical need to develop the next generation of ERα antagonists that can overcome aberrant activities of both ERα^WT and ERα^MUT.

The existing classes of ER antagonists, SERMs, and SERDs exploit distinct mechanisms to promote their antitumor activities. SERMs execute selective inhibition or stimulation of estrogen-like actions in a tissue-dependent manner through differential recruitment of coregulators to ERα, which in turn regulates ERα-dependent transcription. In contrast, SERDs trigger selective ER degradation upon binding to the LBD. Hence, SERDs are usually considered “pure” ER antagonists showing distinct tissue-selective activities compared with SERMs. Given the clinical success of fulvestrant, there have been significant efforts geared toward the development of orally bioavailable SERDs or SERD/SERM hybrids. Several oral SERDs have recently been tested in clinical trials for locally advanced or metastatic ER-positive breast cancer, including a nonsteroidal combined SERM and SERD, brilanestrant (GDC-0810, discontinued), elacestrant (RAD1901), and AZD9496. Although some are showing encouraging clinical data, none of them were originally developed to target both ERα^WT and ERα^MUT. Thus, identification of novel classes of ER antagonists that effectively target ERα^MUT is of great interest for patients with metastatic breast cancer with unmet medical need.

Previously, in order to study the ERα structure and map the LBD, affinity labeling ligands (estrogenic ketononestrol aziridine and antiestrogenic tamoxifen aziridine) that covalently engaged with ERα were developed (31). Interestingly, although both preferentially engaged with C530, these ligands were capable of labeling an alternate residue in the absence of C530 (32). Here, we positioned an internal electrophile so that covalent modification occurs only at C530 located in the LBD of ERα.

Similar to SERMs and SERDs, upon binding to ERα, SERCA H3B-5942 triggers global DNA binding of ERα to ERE-containing promoter and enhancer regions and induces a transcriptionally repressive conformation of ERα by evicting coactivators. H3B-5942 also shares some pharmacologic features with SERMs, e.g., the uterotrophic activity observed in immature rats. Importantly, H3B-5942 demonstrates a distinct mechanism of action (MoA) from other ERα antagonists. H3B-5942 induces a unique conformational change of ERα that is distinct from SERDs and SERMs as revealed by a peptide mapping reporter system. Subsequently, H3B-5942-bound ERα differentially regulates a subset of target genes in endometrial carcinoma cells, suggesting it may possess less ERα agonistic activity compared with SERMs in certain cellular contexts. The covalency of H3B-5942 appears to contribute at least some of these unique profiles because the chemical analogue of H3B-5942 lacking the Michael acceptor (H3B-9224) or genetic disruption of the covalency (C530S mutation in ERα) largely abolished the gain in anti-ERα potency by the unique MoA. However, it is currently unclear whether enhanced residence time on ER and/or potentially different conformational states that may exist following H3B-5942/-9224 binding are contributing to the differential functional profiles.

Because H3B-5942 shows a clear dependence on covalent engagement, it is conceivable that resistant mechanisms may occur by mutation of C530 in ERα. Therefore, development of next-generation SERCAs with alternate interactions between the core of the molecule and the ligand-binding domain may influence binding affinity, residence time, and/or conformational state of the protein, all of which may potentially help to overcome C530-related resistance and maintain downstream potency associated with covalent engagement.

In summary, we identified a class of SERCAs that demonstrate preclinical evidence of covalent binding to ERα, potent inhibition of ERα-dependent transcription, and antitumor activity in both ERα^WT and ERα^MUT breast tumor models.

MCF7 BUS cells (33) were maintained in DMEM supplemented with 10% FBS, 4 mmol/L l-glutamine, and 1× nonessential amino acids. MCF7 lines engineered to overexpress ERα^WT, ERα^Y537S, ERα^Y537C, ERα^Y537N, ERα^E380Q, ERα^WT/C530S, and ERα^Y537S/C530S were derived from the MCF7 BUS cells and similarly maintained in culture. MCF7(P)-AP2, a serially in vivo passaged cell line derived from the MCF7-ATCC tumors, was cultured in Eagle Minimum Essential Medium supplemented with 10% FBS and used for xenograft studies. MCF7-Parental cell lines derived from ATCC were cultured in phenol red-free RPMI supplemented with 10% FBS and exogenous estradiol (1 nmol/L). The respective LTED derivatives, MCF7-LTED-ERα^WT and MCF7-LTED-ERα^Y537C, were cultured as previously described (34) in phenol red-free RPMI supplemented with 10% dextran charcoal-stripped FBS (DCC medium). MDA-MB-231 cells were maintained in Leibovitz's L-15 media supplemented with 10% FBS. The PDX-ERα^Y537S/WT cell line was derived from a patient-derived xenograft tumor model and was routinely cultured in DMEM supplemented with 20% FBS. Lenti-X 293T cells (Clontech, cat. #632180) were routinely cultured in DMEM supplemented with 10% FBS and 4 mmol/L l-glutamine. Ishikawa cells were maintained in Minimum Essential Medium supplemented with 2 mmol/L l-glutamine, 1% nonessential amino acids, and 5% FBS. All cells were maintained prior to and during experiments 37°C, 5% CO₂, and 95% relative humidity. Cells were passaged 2 to 3 times per week, and passage number was limited to between 6 and 20. For in vitro experiments, cells were seeded at appropriate densities to provide logarithmic growth during, and at least 24 hours beyond, the experiment duration. All cell lines were verified to be free of Mycobacterium contamination, and their identity was confirmed by short tandem repeat analysis of 9 markers.

Escherichia coli codon-optimized genes encoding the LBDs of the receptors His-TEV-ERα-WT (297-554), His-TEV-ERα-Y537S (297-554), His-TEV-ERα-C530S (297-554), His-TEV-ERα-C530S-Y537S (297-554), His-TEV-ERα-C381S-C417S (307-554), and His-TEV-ERα-C381S-C417S-Y537S (307-554) were synthesized by Genewiz and cloned into pET28a (EMD Millipore). Proteins were expressed in E. coli overnight at 12°C to 16°C after induction with 0.5 mmol/L IPTG at an OD600 of ∼0.8. Soluble protein was purified by Ni-NTA chromatography followed by size exclusion chromatography on a 26/60 Superdex S-200 column equilibrated in 50 mmol/L Tris–HCl, pH 8.0, 150 mmol/L NaCl, 5–10% glycerol, and 1 mmol/L TCEP. For mass spectrometry and crystallography applications, the His-tag was removed after the Ni-NTA step with overnight TEV protease incubation. The cleaved proteins were then injected on the gel filtration column as described above. Peak fractions were pooled and flash-frozen in liquid nitrogen.

His-TEV-ERα^Y537S (307-554) was cloned into pET-28a (EMD Millipore) and expressed in E coli. In this construct, two surface-exposed cysteine mutations, C381S and C417S, were introduced to improve protein behavior and yields. Soluble protein was obtained by overnight induction at 20° C using 0.1 mmol/L IPTG (35). Cells were harvested and protein was purified using Ni-NTA chromatography followed by overnight TEV cleavage of the His-tag and a polishing subtractive Ni-NTA step to remove the TEV. The flow-through was concentrated and injected on a sephacryl S-300 column equilibrated in 50 mmol/L Tris pH 8, 150 mmol/L NaCl, 1 mmol/L TCEP, and 10% glycerol. Peak fractions were pooled and concentrated to ∼12.4 mg/mL and flash-frozen in liquid nitrogen. Cocrystals were obtained by mixing compound and protein at a 2:1 molar ratio at room temperature for 1 hour to allow time for covalent bond to form, followed by filtration to remove aggregates. Sitting drops were set up using 0.5 μL protein + 0.5 μL reservoir and equilibrated over a reservoir containing 4% to 12% PEG 3350, 50 to 200 mmol/L MgCl₂, and 100 mmol/L imidazole pH 7.1. Crystals grew to full size in 1 to 3 weeks and were flash-frozen using reservoir solution supplemented with 20% ethylene glycol. Data were collected by Shamrock Structures LLC at the Advanced Photon Source, LS-CAT 21-ID-G. The structure was solved by molecular replacement using MOLREP (36) and refined using Refmac (37) with ligand coordinates generated using JLigand. The PDB identification code for H3B-5942 is 6CHW and for H3B-9224 is 6CHZ.

ERα^WT (297–554) and mutant (297–554) proteins were incubated in 50 mmol/L Tris pH 8.0, 150 mmol/L NaCl, 5% glycerol, and 1 mmol/L TCEP with a 2-fold excess of compound (2 μmol/L H3B-5942:1 μmol/L ERα protein solution) at 4°C overnight. Mass analyses were carried out on a Thermo Scientific Q-Exactive HRM (ESI source, 4.0 kV ionization voltage, 250°C capillary temperature, 10 arb sheath gas, S-lens RF level 65) coupled with an Accela Open AS 1250. Samples (10 μL) were desalted on a C4 column (Thermo Scientific Accucore 2.1 × 150 mm, 2.6 μm) with a gradient from 5% to 95% B over 10 minutes. Eluent A consisted of 0.1% formic acid in water, and eluent B consisted of 0.1% formic acid in acetonitrile. The flow was set to 400 nL/minute. All solvents were LC/MS grade (Thermo Scientific). The mass spectrometer was run in positive mode collecting full scan at R = 70,000 from m/z 500 to m/z 2,000. Data were collected with the Xcalibur 3.1 software.

Xcalibur raw files were processed using BioPharma Finder 2.0 (Thermo Scientific) and the ReSpect deconvolution algorithm. Peak was averaged over selected retention time to generate source spectra from Total Ion Chromatogram trace, and the chromatogram parameters set to m/z 700 to 2,000. Outputs for deconvolution algorithm include model mass range from 10,000 to 160,000, mass tolerance 20 ppm, charge state range from 10 to 100; target mass is the estimated mass of protein or protein + compounds, with noise rejection of 95% confidence.

Varying amounts of ³H-estradiol (Perkin Elmer, cat. #NET517250UC) were diluted into 1,480 μL of binding buffer (25 mmol/L Tris pH 8.0, 10% glycerol, 1 mmol/L TCEP, 0.3 mg/mL ovalbumin) in a 2 mL microfuge tube (VWR, cat. #10025-738). His-TEV-ERα-C381S-C417S (307-554) or His-TEV-ERα-C381S-C417S-Y537S (307-554) was then added to the ³H-estradiol solution at a final concentration of 25 pmol/L (WT) or 5 pmol/L (Y537S). The final volume of the mixture was 1,500 μL. Binding was allowed to equilibrate at 4°C for 18 hours. Following the 18-hour incubation, 300 μL of hydroxyapatite (HAP) resin (Bio-Rad, cat. #1300150) was added to the binding mixture, and the 2-mL tube was rotated at room temperature for 1 hour. The HAP resin with ³H-estradiol–ERα complex bound was then washed in the 2-mL microfuge tube 3 times with 900 μL of 25 mmol/L Tris pH 7.2. For the ERα^WT–estradiol binding measurements, the HAP resin was resuspended in 200 μL of dH₂O and transferred to a scintillation vial containing 3 mL of MicroScint-PS scintillation fluid (Perkin Elmer, cat. #6013631) prior to reading with a MicroBeta2 scintillation counter. Due to the very low amounts of ERα-Y537S used in each 2-mL tube, the HAP resin from two 2-mL binding reactions was combined into one scintillation vial prior to reading so as to achieve a suitable signal-to-noise ratio.

K_d values were determined in Prism using the following equation:

A total of 100 pmol/L His-TEV-ERα-C381S-C417S (307-554) or His-TEV-ERα-C381S-C417S-Y537S (307-554), 1 nmol/L ³H-estradiol and varying concentrations of competitor ligand were mixed in binding buffer (25 mmol/L Tris pH 8.0, 10% glycerol, 1 mmol/L TCEP, 0.3 mg/mL ovalbumin) at a final volume of 1,200 μL. ERα–ligand binding was allowed to equilibrate for 18 hours at 4°C. HAP (300 μL) was then added to the binding mixture and the 2-mL tube was rotated at room temperature for 1 hour. The HAP resin with the ³H-estradiol–ERα complex bound was then washed in the 2-mL microfuge tube 3 times with 900 μL of 25 mmol/L Tris pH 7.2. The HAP resin was then resuspended in 200 μL of dH₂O and transferred into a scintillation vial containing 3 mL of scintillation fluid and read on a scintillation counter as described above. K_i values were calculated using the Munson and Rodbard equation (38).

His-TEV-ERα-C381S-C417S (307-554) or His-TEV-ERα-C381S-C417S-Y537S (307-554) and ligand were diluted into 75 μL of binding buffer (25 mmol/L Tris pH 8.0, 10% glycerol, 1 mmol/L TCEP, 0.3 mg/mL ovalbumin) at a concentration of 4 nmol/L ERα protein and 10 nmol/L ligand, in a 2-mL microfuge tube. The ERα–ligand mixture was allowed to sit at room temperature for 5 hours to achieve binding equilibrium.

After 5 hours, the binding mixture was diluted to a final volume of 1,500 μL with binding buffer in the presence of 7.5 nmol/L ³H-estradiol. The final concentrations of ERα and ligand were 200 pmol/L and 500 pmol/L, respectively. Under these final diluted conditions, the concentration of ³H-estradiol is 15-fold greater than competitor ligand. The diluted binding mixture was placed at 4°C for 22 hours, during which time the free ³H-estradiol replaces the prebound ligand after it dissociates from ERα. After the 22-hour incubation at 4°C, the ERα–³H-estradiol complex was bound to HAP resin, washed, and read on a scintillation counter as described above.

Both enzyme-fragment complementation (EFC)–based assays were performed at DiscoverX using components of the HitHunter Estrogen Assay Kit (cat. #90-0019). Purified recombinant full-length human ERα and ERβ proteins were obtained from Life Technologies (ERα: cat. #A15674; ERβ: cat. #A15664). During assay optimization, the concentration of ERα or ERβ protein was titrated into the assay based on the molecular weight of each protein (ERα: 53.4 kDa and ERβ: 66.4 kDa) to produce a comparable EC₅₀ with the reference ligand 17β-estradiol. All test compounds were serially diluted (1:3) in DMSO, prepared at 46× the final assay concentration. In brief, 2 μL of compound titration (10 μmol/L starting dose; 11-point dose response, plus DMSO-only control; 4 replicates/dose), 20 μL ED reagent, and 30 μL of a 2.5 μg/mL stock of ERα or 1.6 μg/mL of ERβ protein were added to each well of a 96-well assay plate. The assay components were incubated for 90 minutes at room temperature. Following incubation, HitHunter Detection reagent was prepared as recommended by the manufacturer (DiscoverX) and then added to each plate, followed by a 1-hour incubation at room temperature in the dark. The assay plate was read on an Envision plate reader (Perkin Elmer), and data were analyzed using the GraphPad Prism software. Data from each sample (average of quadruplicate wells) were normalized to the max value for the reference sample, 17β-estradiol, to calculate percentage activity. The EC₅₀ was calculated using the equation: log(agonist) versus response − variable slope (4 parameters). Three independent assays were conducted on separate days.

Assay buffer was composed of 100 mmol/L potassium phosphate, pH 7.4, 5 mmol/L DTT, 0.1 mg/mL bovine gamma-globulin, and 0.001% pluronic F-127. A 2× working stock of His-ERα-LBD was prepared by diluting the protein to 4 nmol/L in assay buffer. Solutions of 5× anti-6xHis-Terbium antibody (CisBio, 61HISTLA) and 3.33× fluorescently labeled peptide (Thermo Fisher) were prepared separately such that their final concentrations were 3 and 125 nmol/L. Fluorescently labeled biotin was also included to control for nonspecific binding.

An acoustic dispenser delivered 2 nL compound or DMSO for a final concentration of 2 μmol/L in 384-well assay plates (Corning, 3820). Then, 5 μL of 2× protein working stock or controls were added. Plates were centrifuged, and incubated for 1 hour at room temperature, followed by the addition of 3 μL of fluorescently labeled peptide and 2 μL of 5× antibody. Plates were covered, centrifuged, and incubated for an additional hour. The positive control contained His-ERα-LBD protein, PGC1α peptide, and 50 nmol/L (final) estradiol or DMSO for the wild-type or mutant protein, respectively. The negative controls were the same but lacked ERα.

The TR-FRET data were recorded with an Envision plate reader (Perkin Elmer), using the settings recommended by Thermo Fisher. Normalization of TR-FRET signal to percent activation and Z′ values were determined from the positive and negative control's recruitment of FITC-PGC1α. Experiments were performed in triplicate, and data were analyzed in GraphPad Prism 7.

ERα from the lysate of serum-starved MCF7 was functionally analyzed by MARCoNI as described previously (39). Each array was incubated with ERα containing lysate in the presence of compound or solvent (DMSO) only. The negative control (DMSO) and positive control (17β-estradiol) were analyzed using 4 technical replicates (arrays) each. 4-OHT and H3B-5942 test compounds were analyzed using 3 technical replicates. Binding was detected with fluorescently labeled ERα antibody. From each array, a series (increasing exposure time) of tiff images were obtained by the CCD camera in the PamStation (incubator). The signal intensity (arbitrary units fluorescence) of each interaction/array/exposure time/array was quantified using dedicated BioNavigator software (PamGene). A circle was placed around the spot boundaries, and the signal within (foreground) and just outside (background) was determined. For each interaction, the exposure time series signal minus background values were used for a linear fit, from which the signal at 100 ms was extracted. Next, we applied a cutoff at the bottom of the signal range to eliminate noise; all signals <50 are set to 50.

The cellular EFC-based CoA recruitment assay was performed at DiscoverX using conditions previously established. In brief, CHO-K1 cells expressing full-length human ERα fused to an enzyme donor and full-length PGC1α containing steroid receptor coactivator peptide domains fused to an enzyme acceptor were counted and resuspended in PathHunter Cell Plating Reagent 13 (cat. #93-0563R13A) containing charcoal–dextran stripped FBS and seeded into 384-well assay plates at 10,000 cells per well in 20 μL. Compounds were serially diluted (1:3) in PBS + 1% BSA in 12-point dose-response format, 4 replicates per concentration. The last well in the dilution series was DMSO only. Intermediate dilutions of ERα antagonists from the original stock were performed in DMSO, followed by a 1:20 dilution in PBS + 0.1% BSA to generate a 5× stock. Compound dilution (5 μL) was added to the appropriate wells and incubated for 1 hour at 37°C. Cells were then stimulated by exposure to 17β-estradiol (20 nmol/L final concentration; 5 μL of a 6× stock) for 6 hours at 37°C. PathHunter chemiluminescent detection reagents were prepared as recommended by the manufacturer; 15 μL of reagent was added per well, and the plates incubated at room temperature for 1 hour in the dark. Luminescence signal was captured for each well on an Envision plate reader (Perkin Elmer). Signals were normalized to basal activity of the 17β-estradiol agonist control curve on each plate. Three independent assays were conducted on separate days.

The selectivity assay principle is similar to that of the ERα CoA recruitment assay described above. However, the control agonists and antagonists varied for each nuclear hormone receptor tested. The control agonists used were 6α-fluorotestosterone for androgen receptor (AR), aldosterone for mineralocorticoid receptor (MR), dexamethasone for glucocorticoid receptor (GR), and norgestrel for progesterone receptor alpha/beta (PRα/β). The control antagonists used were mifepristone for GR and geldanamycin for AR, MR, and PRα/β. In brief, cells were seeded in a total volume of 20 μL into white 384-well microplates and incubated at 37°C prior to testing. Assay media contained charcoal-dextran filtered serum to reduce hormone levels. Intermediate dilution of test agent stocks was performed to generate 5× test agent in assay buffer. Five microliters of 5× compound was added to cells and incubated at 37°C for 1 hour. Vehicle concentration was 1%. Five microliters of 6× EC₈₀ agonist (EC₈₀ concentrations were 0.03 μmol/L 6α-fluorotestosterone, 0.005 μmol/L aldosterone, 0.12 μmol/L dexamethasone, 0.005 μmol/L norgestrel) in assay buffer was added to the cells and incubated at 37°C for 6 hours. Assay signal was generated through the addition of 15 μL (50% v/v) of PathHunter Detection reagent cocktail, followed by 1-hour incubation at room temperature. Microplates were read following signal generation with an Envision plate reader (Perkin Elmer) for chemiluminescent signal detection.

Estradiol (E2), 4-hydroxytamoxifen (4-OHT), tamoxifen, and fulvestrant were purchased from Sigma. For the estrogen receptor (ER) conformation profiling assay, a panel of validated cofactors was tested for their ability to interact with ER when bound by mechanistically distinct ER antagonists. In addition to the receptor interacting domains (ID), isolated NR boxes (e.g., ACTR NR boxes 1 and 2) from the aforementioned cofactors were tested in isolation from the IDs. Various recombinant LXXLL (coactivator–receptor binding motif) and corepressor nuclear receptor (CoRNR) box peptides were also tested in this manner. HepG2 cells were maintained in Basal Medium Eagle containing 8% FBS. For mammalian two hybrid–based ER cofactor assay, cells were seeded in 96-well plates and transfected with VP16-ER (900 ng), 5XGalLuc3 (900 ng), Gal4 interactor (900 ng), and Renilla-Luc (300 ng) using Lipofectin. Cells were then treated with saturating concentrations of ligand (10 μmol/L for ER antagonist) for 48 hours, at which point dual luciferase assays were performed. The data were standardized to avoid bias due to signal strength and clustered with the Ward hierarchical clustering method using JMP (SAS). The hierarchical cluster dendrogram was ordered by the first principal component. Data were clustered using the unweighted pair group method with arithmetic mean (UPGMA) method, and correlation distance was measured using Spotfire (spotfire.tibco.com).

Cells were lysed in loading buffer (Invitrogen, cat. #NP0007) containing protease inhibitor (Roche, cat. #05892791001) and DTT (Invitrogen, cat. #NP0009), sonicated, and subsequently boiled for 5 minutes. Approximately 30 μg of protein was loaded per lane and resolved by SDS polyacrylamide electrophoresis. Protein was transferred onto nitrocellulose membranes, blocked in 5% low-fat milk, and probed overnight with antibodies to ER SP1 (Spring Bioscience, cat. #M3012), Rb (4H1; Cell Signaling, cat. #9309), phospho-Rb (Ser780; D59B7; Cell Signaling, cat. #8180), cyclin D1 (EPR2241; Abcam, cat. #ab134175), α-Tubulin (Sigma, cat. #T6199), and GAPDH (Sigma, cat. #G9545). Membranes were incubated with horseradish peroxidase (HRP)–conjugated anti-rabbit secondary antibody (Cell Signaling, cat. #7074) or anti-mouse secondary antibody (Cell Signaling, cat. #7076) for 1 hour, and signal was developed using the enhanced chemiluminescence method (GE Healthcare).

When preparing the MCF7(P)-AP2 cells for in vivo studies, the cells were harvested and washed with PBS, incubated with 0.25% trypsin-EDTA, and suspended in a 1:1 mixture of Matrigel (Corning 354234) and Hank's Balanced Salt Solution at a final concentration of 5 × 10⁷ cells/mL. To generate xenografts, 0.2 mL of the inoculum was injected into the third mammary fat pad of 6- to 8-week-old female BALB/c nude mice (Balb/cOlaHsd-Foxn1^nu), giving a final concentration of 1 × 10⁷ cells/mouse. Three days prior to inoculation, each mouse was implanted with a 0.72 mg 90-day release estrogen pellet (Innovative Research of America). When the mean tumor volume (TV) reached approximately 150 to 200 mm³, animals were selected based on TV and randomized into treatment groups of 6 to 8 animals per group. Single-agent or combination treatments were started on day 0 and continued for the duration of the study. H3B-5942 was administered orally q.d., tamoxifen was given s.c. Q2D, fulvestrant was given s.c. q.w., and palbociclib was administered orally q.d. Each treatment was administered based on BW (10 mL/kg). H3B-5942 was formulated daily in 10% 2-Hydroxypropyl-β-CycloDextrin (HPβCD) in 5% dextrose, tamoxifen was formulated in 95% peanut oil/5% ethanol (EtOH), clinical-grade fulvestrant was administered, and palbociclib was formulated in 25 mmol/L sodium bicarbonate, 15 mmol/L lactic acid solution with 2% Cremophor EL. The BW measurements were performed daily, and tumor measurements were recorded twice a week.

The TV in mm³ was calculated according to the following formula:

length: largest diameter of tumor (mm)

width: diameter perpendicular to length (mm)

The TGI% was calculated according to the following formula:

where day X is any day of measurement.

All procedures relating to animal care, handling, and treatment were performed according to guidelines approved by the Institutional Animal Ethics Committee (IAEC) of Aurigene Discovery Technologies Ltd. All doses and regimens were well tolerated with no clinical signs observed in all studies presented.

The ST941 PDX model representing an ERα^Y537S/WT mutated human ER⁺ breast cancer was propagated in mice. To generate patient-derived xenografts, solid-tumor tissues from the ERα^Y537S/WT-positive xenograft model (passage 6) were cut into 70-mg pieces, mixed with Matrigel (Corning, 354234), and subcutaneously implanted into the right flank of 12-week-old female athymic nude (Crl:NU(NCr)-Foxn1nu) mice supplied with drinking water containing estradiol (Sigma-Aldrich, cat. #E1024-25G). When the mean TV reached approximately 125 to 200 mm³, animals were selected based on TV and randomized into treatment groups of 6 to 8 animals per group. H3B-5942 was administered orally q.d. (same formulation as above), tamoxifen was given s.c. three times a week, fulvestrant was given s.c. q.w., and palbociclib was dosed orally q.d. Each treatment was administered based on BW (10 mL/kg), except tamoxifen and fulvestrant which were flat-dosed. Tamoxifen was formulated in 90% peanut oil/10% EtOH, and clinical-grade fulvestrant was administered without further dilution. Beginning 3 days prior to treatment and for the remainder of the study, exogenous estradiol was removed from the drinking water. BW measurements were performed daily, and TV measurements were recorded twice per week. All studies were performed under guidelines set forth by the South Texas Accelerated Research Therapeutics (START) Institutional Animal Care and Use Committee and defined in the START Animal Care and Use Program (Protocol 09-001). All doses and regimens were well tolerated.

End-of-study tumor samples were formalin-fixed and paraffin-embedded (FFPE) at START and subsequently shipped to Cancer Genetics Inc. for sectioning, staining, and RNA extraction. In brief, 5-μm sections of FFPE tissue xenograft samples were stained for ERα, progesterone receptor (PR), and Ki67 using the Ventana rabbit monoclonal primary antibodies ER (SP1), PR (1E2), and Ki67 (30-9). IHC assays were run on the Ventana BenchMark IHC/ISH automated slide staining instrument with and ultraView Universal DAB detection kit. ER, PR, and Ki67-stained slides were scored for % positive tumor cells by a board-certified MD–pathologist at Cancer Genetics Inc. Scanned images, pathologist scores, and interpretations were made available to H3 Biomedicine Inc. for review and data analysis.

For gene expression analysis, a custom NanoString nCounter code set was used with 300 ng of purified total RNA extracted from 60 μm of FFPE blocks using the Qiagen RNAeasy FFPE kit according to the manufacturer's instructions. The RNA was hybridized with the codeset and processed according to the NanoString's instructions. The nCounter Digital Analyzer counted and tabulated the signals of reporter probes, and the raw counts were normalized to housekeeper.

The WHIM20 PDX model representing an ERα^Y537S mutated human ER⁺ breast cancer was propagated in mice. For the current study, solid-tumor tissues were depleted of necrotic components, cut into fragments, mixed with Matrigel, and subcutaneously implanted into the right flank of 6- to 8-week-old female SCID-bg mice. The precise number of fragments and volume of Matrigel was determined on a case-by-case basis. When the average TV reached approximately 350 to 400 mm³, animals were selected based on TV and randomized into treatment groups of 7 to 9 animals per group. H3B-5942 was administered orally q.d., tamoxifen (formulated in 95% peanut oil/5% EtOH) was administered s.c. Q2D, and clinical-grade fulvestrant was administered s.c. q.w. Each treatment was administered based on BW (10 mL/kg). BW measurements were performed daily, and TV measurements were recorded twice per week. Mice with at least 20% BW loss compared with day 0 BW were euthanized to prevent any pain or suffering to the animal. All procedures relating to animal care, handling, and treatment were performed according to guidelines approved by the IAEC of Aurigene Discovery Technologies Ltd.

In vitro data are expressed as mean ± SEM or mean ± SD as indicated. Statistical significance was determined by two-sided t test analysis. Uterotrophic data are presented as mean ± SD and were analyzed using a two-sided t test. Efficacy data are expressed as mean ± SEM for TV. The differences in TV on the final day of TV measurements between the vehicle and treatment groups were analyzed by multiple unpaired t tests with significance determined using the Holm–Sidak method with alpha set to 0.05 and without assuming a consistent standard deviation. Statistical analyses were performed using GraphPad Prism version 5.04 (GraphPad Software).

Microarray, RNA-seq, and ChIP-seq data presented in this article have been deposited at the National Center for Biotechnology Information Gene Expression Omnibus with the accession number GSE115611.

Additional experimental procedures are listed in the Supplementary Methods.

G.Z. Zheng is a Senior Director at Stealth Biotherapeutics. L.-A. Martin reports receiving a commercial research grant from Radius Pharm. No potential conflicts of interest were disclosed by the other authors.

Conception and design: X. Puyang, C. Furman, G.Z. Zheng, P. Fekkes, M.-H. Hao, S. Irwin, C. Karr, N. Larsen, S. Prajapati, D.J. Reynolds, F.H. Vaillancourt, J. Wang, M. Warmuth, P.G. Smith, P. Zhu, M. Korpal

Development of methodology: X. Puyang, C. Furman, G.Z. Zheng, K. Aithal, S. Agoulnik, D.M. Bolduc, S. Das, M.-H. Hao, R. Houtman, S. Irwin, C. Karr, N. Kumar, P. Kumar, N. Larsen, T.-V. Nguyen, S. Prajapati, S. Sivakumar, V. Subramanian, F.H. Vaillancourt, S. Yao, P.G. Smith, P. Zhu, M. Korpal

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Puyang, C. Furman, G.Z. Zheng, D. Banka, K. Aithal, S. Agoulnik, D.M. Bolduc, B. Caleb, S. Eckley, M.-H. Hao, R. Houtman, A. Kim, G. Kuznetsov, W.G. Lai, N. Larsen, L.-A. Martin, D. Melchers, A. Moriarty, T.-V. Nguyen, J. Norris, S. Pancholi, N. Rioux, R. Ribas, A. Siu, S. Sivakumar, V. Subramanian, M. Thomas, S. Wardell, M.J. Wick, S. Yao, P. Zhu, M. Korpal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Puyang, C. Furman, G.Z. Zheng, Z.J. Wu, K. Aithal, D.M. Bolduc, S. Das, S. Eckley, P. Fekkes, M.-H. Hao, A. Hart, R. Houtman, S. Irwin, J.J. Joshi, C. Karr, A. Kim, W.G. Lai, N. Larsen, T.-V. Nguyen, J. Norris, M. O'Shea, S. Rajagopalan, D.J. Reynolds, V. Rimkunas, A. Siu, S. Sivakumar, V. Subramanian, F.H. Vaillancourt, S. Wardell, L. Yu, M. Warmuth, P.G. Smith, P. Zhu, M. Korpal

Writing, review, and/or revision of the manuscript: C. Furman, Z.J. Wu, D. Banka, K. Aithal, M.-H. Hao, R. Houtman, J.J. Joshi, C. Karr, A. Kim, N. Larsen, A. Moriarty, S. Prajapati, D.J. Reynolds, V. Rimkunas, N. Rioux, S. Sivakumar, F.H. Vaillancourt, L. Yu, P.G. Smith, P. Zhu, M. Korpal

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Puyang, C. Furman, S. Irwin, A. Kim, W.G. Lai, C. Mackenzie, S. Sivakumar

Study supervision: G.Z. Zheng, S. Buonamici, W.G. Lai, A. Moriarty, S. Rajagopalan, M.J. Wick, M. Warmuth, P.G. Smith, P. Zhu, M. Korpal

Other (performed and developed multiple biochemical assays including the peptide fingerprinting experiments shown in the supplementary material): S. Irwin

We would like to thank Raghuveer Ramachandra for assistance with the MCF7 tumor model development and subsequent efficacy studies. We would also like to thank Drs. Tarek Sahmoud and Rinath Jeselsohn for helpful discussions. This work was supported by H3 Biomedicine, Inc. Additional funding was generously provided by the Breast Cancer Now Toby Robins Research Centre, and NHS funding to The Royal Marsden Hospital's NIHR Biomedical Research Centre (L.-A. Martin, S. Prajapati, and R. Ribas).