Estrogen receptor coregulator binding modulators (ERXs) effectively target estrogen receptor positive human breast cancers

The majority of human breast cancer is estrogen receptor alpha (ER) positive. While anti-estrogens/aromatase inhibitors are initially effective, resistance to these drugs commonly develops. Therapy-resistant tumors often retain ER signaling, via interaction with critical oncogenic coregulator proteins. To address these mechanisms of resistance, we have developed a novel ER coregulator binding modulator, ERX-11. ERX-11 interacts directly with ER and blocks the interaction between a subset of coregulators with both native and mutant forms of ER. ERX-11 effectively blocks ER-mediated oncogenic signaling and has potent anti-proliferative activity against therapy-sensitive and therapy-resistant human breast cancer cells. ERX-11 is orally bioavailable, with no overt signs of toxicity and potent activity in both murine xenograft and patient-derived breast tumor explant models. This first-in-class agent, with its novel mechanism of action of disrupting critical protein-protein interactions, overcomes the limitations of current therapies and may be clinically translatable for patients with therapy-sensitive and therapy-resistant breast cancers. DOI: http://dx.doi.org/10.7554/eLife.26857.001


Introduction
Endocrine therapies for estrogen receptor alpha (ER)-positive breast cancer involve modulation of ER signaling using either antiestrogens (AE) or aromatase inhibitors (AI). However, most patients develop resistance to these drugs, and disease progression is common, with progression to metastases (Musgrove and Sutherland, 2009;Ma et al., 2015). ER signaling is complex and involves coregulators (McDonnell and Norris, 2002;O'Malley and Kumar, 2009). Therapy-resistant tumors often retain ER-expression and ER-signaling. While multiple mechanisms maintain ER signaling in therapy-resistant tumors, ER signaling is mediated by the interactions between activated ER and critical coregulator proteins (Dasgupta and O'Malley, 2014;Lonard et al., 2007).
Thus, there is a strategic need for drugs that disrupt interactions between ER and critical coregulators to block ER signaling. In this study, we have synthesized a series of small organic molecules to emulate the nuclear receptor (NR) box motif, important for ER coregulator interactions. We have identified a small molecule named as ER coregulator binding modulator-11 (ERX-11), with potent anti-proliferative activity against ER-driven breast tumors. ERX-11 interacts with ER and blocks the interaction between ER and coregulators. In ER-expressing breast cancer cells, ERX-11 blocks the proliferation and induces apoptosis. ERX-11 has no activity against ER-negative breast cancer cells. eLife digest Around 70% of breast cancers in women need one or both of the female hormones (estrogen and progesterone) to grow. To treat these 'hormone-dependent' cancers, patients receive drugs that either block the production of estrogen or directly target a receptor protein that senses estrogen in the cancer cells. Unfortunately, many breast cancers develop resistance to these drugs. This resistance is often caused by genetic mutations that alter the estrogen receptor; for example, the receptor may develop the ability to interact with other proteins in the cell known as coregulators to promote tumor growth.
Developing new drugs that prevent estrogen receptors from interacting with coregulators may provide more options for treating hormone-dependent breast cancers. Here, Raj et al. developed a new small molecule named ERX-11 that is able to inhibit the growth of human breast cancer cells that are sensitive to existing drugs as well as cells that have become drug-resistant.
For the experiments, hormone-dependent breast cancer cells from humans were transplanted into mice. This procedure usually causes the mice to develop tumors, but giving the mice ERX-11 by mouth stopped estrogen receptors from interacting with coregulators and blocked the growth of tumors. Furthermore, ERX-11 does not appear to have any toxic effects on the mice, indicating that it may also be safe for humans.
The findings of Raj et al. suggest that ERX-11 is a promising new drug candidate for treating some breast cancers. The next steps are to examine the effects of ERX-11 on mice and other animals in more detail before deciding whether this molecule is suitable for clinical trials. In the longer term, molecules similar to ERX-11 could also be developed into drugs to treat other types of cancer that are also caused by abnormal interactions of coregulator proteins.  Table 1). Pathways analysis revealed that ERX-11binding proteins were involved in the activation of transcriptional regulation (Figure 2-figure supplement 1). Importantly, ERX-11 pulldown included a number of proteins other than ER, including a weak affinity for the progesterone receptor (PGR) and several ER-associated proteins ( Table 1). Immunoprecipitation analyses in MCF-7 cells validated the strong affinity of ERX-11 for ER, and weak affinity for the PR-A isoform but not GR, AR or PR-B isoforms ( Figure 2C).
The interaction between ER and ERX-11 within the cells was partially disrupted by high doses of tamoxifen ( Figure 2D). Further, in the tamoxifen-resistant cell line, MCF-7-TamR, even high doses of tamoxifen could not disrupt the interaction between ERX-11 and ER (Figure 1-figure supplement 3E). The differences between these results and the in vitro results may be attributed to the context in which ER is presented within the cell.
Using GST-fused ER domain constructs, we validated that ERX-11 interact with the GST-AF2 domain of ER but not with the GST-AF1 or GST-DNA-binding domain of ER ( Figure 2E). Further, ER-AF2 interaction with ERX-11 was disrupted by tamoxifen but not ICI ( Figure 2F). These data clearly establish the interaction between ER and ERX-11 through the AF-2 domain.

ERX-11 blocks ER interactions with coregulators
Using an unbiased approach with IPMS, we showed that ERX-11 significantly disrupted the interactions of 91 nuclear ER-binding proteins with ER in MCF-7 cells (Figure 2-figure supplement 2A), including well-characterized ER coregulators, such as SRC1, SRC3, and PELP1. Global analyses revealed that these proteins may be involved in a number of critical cellular pathways including transcription, cell cycle and regulation of cell death ( Table 2). These findings were validated by IPMS studies in ZR-75 cells, which showed a significant overlap with MCF-7 cells in the coregulators disrupted by ERX-11 ( Figure 2-figure supplement 2B). Of the top 10 coregulators, whose interactions with ER were negatively influenced by ERX-11, five contained LXXLL motifs with serine at i-3/4 and i+7/8 flanking position of the LXXLL motifs Table 3. Interestingly in the MDA-MB-231 TNBC model cells, we found that biotinylated ERX-11 was able to stringently interact only with a small number of proteins (n = 8) (Figure 2-figure supplement 2C).
In MCF-7 cells, a majority of ER-binding proteins disrupted by ERX-11 were also blocked by tamoxifen (55/88 proteins or 62.5%) ( Figure 2G and Figure 2-figure supplement 3). Importantly, a significant number of ER-binding proteins were disrupted by ERX-11 but not tamoxifen (33/88 or 37.5%) ( Figure 2G and Co-immunoprecipitation studies validated that endogenous complexes containing ER and coregulator PELP1 do form in MCF-7 cells and that ERX-11 disrupts the formation of these complexes and ER using AutoDock Vina. Superimposition of the docked ERX-11 (green) on the crystal structure (PDB code 1L2I) of the LXXLL motif (orange) (D). Purified full-length ER was incubated with biotin-control or biotin-ERX-11 in the presence of E2. ERX-11 interaction with purified ER was analyzed using avidin bead pulldown and western blotting (E). Purified full-length ER was incubated with biotin-ERX-11 in the presence of E2 ± free ERX-11 (1 mM). ERX-11 ability to compete with the binding of biotin ERX-11 with ER was analyzed using avidin pulldown assay (F). Purified full-length ER was incubated with biotin-ERX-11 in the presence of E2 ±LXXLL peptides (1 mM) from various coregulators SRC1, SRC2, AIB1, and PELP1. LXXLL peptides ability to compete with the binding of biotin ERX-11 with ER was analyzed using avidin pulldown assay (G). Purified full-length ER was incubated with biotin-ERX-11 in the presence of E2 ± GDC0810, AZD-9496, ICI, and Tam (1 mM) and their ability to compete with the binding of biotin ERX-11 with ER was analyzed using avidin pulldown assay (H). DOI: 10.7554/eLife.26857.003 The following figure supplements are available for figure 1:   Figure 2. ERX-11 interacts with ER and blocks its interactome. Interaction with endogenous ER was evaluated in nuclear lysates prepared from ZR-75 cells stimulated with E2, incubated with biotin-control or biotin-ERX-11 and analyzed by avidin pull-down assay (A). Nuclear lysates from MCF-7 cells were incubated with biotinylated ERX-11 for 2 hr and then subject to a streptavidin column. The bound proteins were eluted and subjected to analyses by mass spectroscopy. The fold-enrichment in binding over basal is depicted for two independent replicates (x axis = replicate one and y Figure 2 continued on next page ( Figure 2H). Proximity ligation assays confirmed the disruption of the endogenous interaction between ER and several coregulators including PELP1, SRC1 and SRC3 ( Figure 2I, quantitation in Figure 2-figure supplement 4A). In contrast, ERX-11 has no effect on ER interaction with ARID1B ( Figure 2I, quantitation in Figure 2-figure supplement 4B). Using an in vivo structural complementation NanoBiT assay, we found that the direct interaction between ER and PELP1 was enhanced by E2 treatment and that ERX-11 significantly reduced the interaction between ER and PELP1 ( Figure 2J).
To confirm the specificity of ERX-11 for the ER AF-2 domain, we demonstrated using reporterbased assays, that ERX-11 failed to reduce the ERE-Luc reporter activity driven by a ERa-VP16 chimera that does not require AF-2 (Figure 2-figure supplement 4C). As expected, tamoxifen, did not affect the ERa-VP16 chimera-induced reporter activity, while ICI reduced the ERE-Luc reporter activity. Evaluation with an endometrial cancer cell line Ishikawa, which exhibits agonist activity via AF1, revealed that ERX-11 lacks AF1 agonist activity (Figure 2-figure supplement 4D). Collectively, these results confirm that the ERX-11 block signaling driven by functional AF2 domain but not by AF1 domain.
We then specifically evaluated whether interactions through the ER LXXLL motif was responsible for ERX-11 activity. Biotinylated ERX-11 was able to pull down both the wild-type ER and the L540Q point mutant ER (which retains E2 binding and does not interact with SRC1) and these interactions were not affected by tamoxifen ( Figure 2K). Using ER L540Q point mutant, we showed that the mutant ER still interacts with biotinylated ERX-11 ( Figure 2K). Interestingly, the ERX-11 binds strongly to ER~12 mutant (helix 12 deleted ER) and this binding is blocked efficiently by tamoxifen Figure 2 continued axis = replicate 2). Relative binding of ER and PGR are shown (B). MCF-7 nuclear lysates were incubated with biotinylated ERX-11 and then subject to a streptavidin column. The bound proteins were eluted and evaluated by western blotting compared to equivalent amount of input (C). ZR-75 cells were incubated with tamoxifen and its ability to interfere ERX-11 binding to ER was analyzed by immunoprecipitation followed by western blotting (D). To confirm ERX-11 binding to ER-ligand-binding domain (AF2), a GST pull-down assay was performed. Biotinylated ERX-11 interacted with the GST-AF2 domain of ER but not with the GST-AF1-or GST-DNA-binding domain of ER (E). Purified ER-AF2 domain was incubated with biotin-ERX-11 in the presence of E2 ±ICI or tamoxifen (1 mM). ICI and tamoxifen ability to compete with binding of biotin ERX-11 with ER-AF2 domain was analyzed using avidin pull-down and western blotting (F). Venn diagram shows the overlap between the ER-binding proteins immunoprecipitated from nuclear lysates from E2-stimulated MCF-7 cells following treatment with vehicle or ERX-11 or tamoxifen (G). Co-immunoprecipitation analyses show the effect of ERX-11 on the interaction of ER with coregulators PELP1, SRC1, TIF1a in MCF-7 cells (H). Proximity ligation assay validated the ability of ERX-11 and tamoxifen to disrupt the interactions between ER and coregulators such as PELP1, SRC1, SRC3/AIB1 and ARID1B in MCF-7 cells (I). NanoBiT assay: expression plasmids were created to express either ER or PELP1 in conjunction with the large bit or the small bit of the NanoBiT luciferase enzyme. If the proteins directly interact within the cell, the two parts of the NanoBiT luciferase enzyme come together and create a quantifiable luminescent signal. The effect of ERX-11 on the interaction between the two sets of ER and PELP1 constructs is shown (J). Validation of the binding of ERX-11 to the ER-AF2 domain was further explored using AF2-domain mutants of ER stably transfected in ER-negative MDA-MB-231 breast cancer cell lines. Biotinylated ERX-11 was then used to pull-down ER from these cell lines (K). Data shown are the means of ±SEM performed in triplicate wells. *p<0.05. DOI: 10.7554/eLife.26857.007 The following figure supplements are available for figure 2: ( Figure 2K). These data suggest that the presence of helix 12 may regulate the conformation of the binding pocket and account for differences in the binding of ERX-11 and tamoxifen to ER. Our data would suggest that removal of helix 12 enables access of ERX-11 to the same binding pocket as tamoxifen and may reflect the in vitro data, where tamoxifen and ERX-11 compete efficiently for ER binding. In contrast, neither GDC-0810 nor AZD-9496 were able to block ERX-11 binding to ER or its mutants, suggesting that their binding to ER occurs through distinct pockets (Figure 2-figure supplement 5A and B).
In competition assays, tamoxifen fails to dislodge ERX-11 from ER (Figure 2-figure supplement 5C, Figure 1-figure supplement 3E). Increasing concentrations of tamoxifen is only able to dislodge ERX-11 from ER~12 mutant at higher concentrations, suggesting that ERX-11 interaction with ER is through a second binding site (Figure 2-figure supplement 5D).
Further, using an ER restoration model, MTT cell viability assays revealed that introduction of either ER, ER~12 or ER-L540Q into MDA-MB-231 cells, restored ERX-11 growth inhibitory activity in Table 1. Top proteins pulled down by biotinylated ERX-11 in MCF-7 cells, as identified by IP-MS. The column marked E2 represents spectral counts for the protein bound to biotinylated control eluted from avidin column, under conditions of E2 stimulation. The column marked E2 +ERX-11 represents spectral counts for proteins bound to biotinylated ERX-11 eluted from an avidin column with E2 stimulation. The column marked E2 + ERX-11/E2 represents the ratio of binders.

Docking simulations model ERX-11 binding to ER
We have modeled ERX-11 interaction with ER using docking simulations of ERX-11 using known crystal structures ( Figure 2-figure supplement 6). In the agonist-bound conformation (3ERD.pdb), helix 12 is relocated and forms a hydrophobic cleft (i.e. AF2-binding site) that is surrounded by helices 3, 4, 5 and 12: ERX-11 can be modeled to make hydrophobic contact with the AF2 site with its two isobutyl side chain groups (Figure 2-figure supplement 6A(A)). In addition, the hydroxyl  Table 3. Selected pathways modulated by ERX-11 treatment. Differentially expressed genes were subjected to pathway analysis using IPA software and the selected top canonical pathways modulated by ERX-11 are shown. This data is related to Figure 3.    Figure 3. ERX-11 globally disrupts ER-mediated transcriptome. Total RNA was isolated from the ZR-75 cells that were treated with either vehicle or ERX-11 for 48 hr and subjected to RNA sequencing. The heat map of differentially expressed genes between vehicle and ERX-11 is shown (A). ZR-75 cells were treated with either vehicle or ERX-11 for 48 hr, and the selective genes representing each pathway were validated using RTqPCR (B). Gene set enrichment analysis (GSEA) testing correlation between ERX-11-regulated gene and both the tamoxifen-responsive (M3283) and estradiol- Figure 3 continued on next page group of ERX-11 may interact with a residue near AF2 domain. These data suggest that ERX-11 interacts with ER LBD differently than the agonist. In the antagonist-bound conformation (3ERT.pdb), 4-hydroxytamoxifen induces conformational change and makes helix 12 occupy the AF2-binding site, blocking both coactivator and corepressor binding: here, ERX-11 may interact with an alternate pocket formed by helices 5, 11 and 12, as shown in the Figure 2-figure supplement 6A(B). These data could explain why the interaction between ERX-11 and purified ER could be blocked by tamoxifen. However, in a cellular context, tamoxifen cannot block ERX-11 binding to ER suggesting that the secondary binding site of ERX-11 on ER may be stabilized by coregulators. Docking simulation of ERX-11 on human ERa with affinitytagged corepressor peptide (Figure 2-figure supplement 6A(C))(2JFA.pdb) or rat ERb crystal structure with ICI bound (Figure 2-figure supplement 6A(D))(1HJ1.pdb) showed that ERX-11 can still bind to the AF2 domain, in a similar manner as it does when the ligand is bound. These data further support our biochemical findings that ICI does not block ERX-11 interaction with ER.

Pathway p-Value Ratio Genes
Detailed evaluation showed that (1L2I.pdb) two isobutyl groups of ERX-11 may dock into the AF2 binding site (black dashed box) (Figure 2-figure supplement 6B). The hydroxyl group of ERX-11 may interact with Gln375 of the helix five through a hydrogen bond. The carboxamide group of ERX-11 was docked into a pocket formed with Gln542, Asp545 and Ala546 on the helix 12. Additional docking experiment on ER crystal structures without helix 12 (3ERT.pdb and 5ACC.pdb) showed that ERX-11 was found to bind to the tamoxifen-binding site in the ER~12 (

ERX-11 blocks ER-driven breast cancer signaling pathways
RNA-seq analyses revealed that ERX-11 altered the expression of 880 E2-regulated genes (p<0.01) in ZR-75 cells compared to vehicle control ( Figure 3A). Using stringent cutoffs (p<0.01 and RPKM FC > 1.5), ERX-11 down-regulated more genes (669) than upregulated (211) Figure 3H). Gene set enrichment analysis (GSEA) confirmed the correlation between ERX-11-regulated genes and both the tamoxifen-responsive and E2-responsive genes set ( Figure 3C and D). In addition, ERX-11-regulated genes correlated well with signatures of early and late response E2 targets as well as genes driven by consensus ER motifs ( Figure 3E and F).
Ingenuity pathway analysis (IPA) revealed that ERX-11 significantly down-regulated genes involved in ER-signaling, breast cancer, cell cycle, and MAPK signaling ( Table 3). ERX-11 upregulated gene set positively correlated with apoptotic genes, on GSEA analyses ( Figure 3G). Importantly, ERX-11 but not tamoxifen or ICI-treatment-induced apoptosis in ZR-75 ( Figure 3I  . GSEA analysis testing the correlation of ERX-11-regulated genes with signatures of early (M5906) and late (M5907) response estrogen targets as well as genes driven by consensus ER motifs (M17968 and M6101) (E, F). GSEA analysis of correlation of ERX-11 regulated gene set with an apoptotic gene set (M15912) (G). ERX-11 upregulated apoptotic genes were validated by RT-qPCR analyses (H). Data are represented as mean ±SEM. **p<0.01; ***p<0.001. Effect of indicated doses of Tam, ICI and ERX-11 on cell viability (CellTiter-Gloassay, Promega) and caspase3/7 activity (Caspase-Glo 3/7 Assay, Promega) in ZR-75cells (I). Data shown are the means of ±SEM performed in triplicate wells. ****p<0.0001. DOI: 10.7554/eLife.26857.018 The following figure supplement is available for figure 3: Apoptosis can be seen as early as 24 hr, however, the effect is more pronounced at 72 hr (Figure 3figure supplement 1E). These results suggest that ERX-11 both reduces the expression of genes involved in proliferation and enhances expression of genes that promote apoptosis.

ERX-11 inhibits ER-mediated transcription
Using ERE-Luc reporter based transcription assays, we found that ERX-11 significantly reduced the E2-induced ERE-Luc reporter gene activity in ZR-75 cells in a similar fashion as tamoxifen ( Figure 4A). In HEK-293T cells, expression of AIB1 and SRC1 enhanced ER-driven ERE-Luc reporter activity and was blocked by both tamoxifen and ERX-11 in both a ligand-dependent ( Figure 4B) and ligand-independent manner ( Figure 4C). Further, ZR-75 cells stably overexpressing either PELP1 or AIB-1 were responsive to ERX-11 ( Figure 4D). These results indicate that ERX-11 interferes with both ligand-dependent and ligand-independent transcriptional function of ER.
Using chromatin immunoprecipitation studies, we found that ERX-11 treatment significantly blocked ER recruitment to canonical ER target gene promoters following E2 treatment ( Figure 4E). In contrast, ERX-11 did not affect AR recruitment to AR target genes (Figure 4-figure supplement 1A). Since ERX-11 binds to ER, we hypothesized that the effect of ERX-11 on ER DNA binding may be mediated via disruption of ER dimerization. Using the NanoBiT assay, we demonstrated that ERX-11 efficiently blocks the dimerization of ER ( Figure 4F). In contrast, ERX-11 did not affect AR dimerization (Figure 4-figure supplement 1B).
Using E2 dendrimer conjugates (EDC), that are potent in activating ER non-genomic signaling but not ER genomic signaling , we showed that ERX-11 was unable to influence the EDC-mediated non-genomic activation of the Src, AKT and MAPK pathways ( However, similar to its inhibitory activity on ER transcription, after several days of ERX-11 treatment, decreased ER levels were detected ( Figure 4H). Accordingly, RTqPCR results showed that ERX-11 reduce the ER transcript levels under conditions of long-term treatment (7 days). These results reflect the inhibition of ER signaling indirectly affected autoregulation of ER transcript by E2-ER signaling (Figure 4-figure supplement 1E).

ERX-11 suppresses ER-driven breast tumor growth in vivo
Our prior studies indicated that our peptidomimetics are orally bioavailable (Ravindranathan et al., 2013). We detected no overt signs of toxicity after 14 days of treatment of C57BL/6 mice (n = 3) with 10, 50 or 100 mg/kg/day of ERX-11 via oral gavage. ERX-11 treatment neither caused weight loss nor have uterotrophic effects or any observable hematologic, liver and kidney abnormalities (data not shown). We designated our highest dose (100 mg/kg) as the maximum tolerated dose and used 10% of this dose (10 mg/kg/day) for testing as therapeutic dose, so that we would have at least a 10:1 therapeutic to toxicity ratio.
Established ZR-75 xenografts (n = 8 tumors/group) in the mammary fat pad of nude mice were randomized to feed via oral gavage 5 days/week with either 10 mg/kg ERX-11 or vehicle (30% Captisol). ERX-11 treatment resulted in significantly smaller tumors (63% reduction compared to control) ( Figure 5A). ERX-11-treated tumors exhibited less proliferation (Ki67 staining), and more apoptosis (TUNEL and caspase-3 staining) than controls ( Figure 5A,B). Further, ERX-11 treatment group had lower ER but similar PELP1 protein expression levels within the tumor, compared to control ( Figure 5B). The mice body weights in the control and ERX-11 treated groups were similar (Figure 5-figure supplement 1A). Mice treated with ERX-11 exhibited no uterotrophic effects, no changes in ovary, liver and kidney gross morphology on H and E staining, or acute phase injury to liver and kidney ( Figure 5-figure supplement 2A-D). These data indicate that ERX-11 is a potent inhibitor of the growth of ER-positive breast tumors in vivo with no overt signs of toxicity in mice.
To address the potential immunogenicity of ERX-11, D2A1 ER-positive breast cancer xenografts were established in a syngeneic BALB/c model system with an intact immune system. In D2A1 model cells, cellular gene int-5/aromatase in BALB/c mammary alveolar hyperplastic nodule (D2 HAN/D2 tumor cells) is activated as a result of mouse mammary tumor virus integration within the 3' untranslated region of the aromatase gene. Thus, these models also have ability to synthesize local estrogen via aromatase induction. Further, this model express ER, and represent a model of intra-tumoral R e la tiv e E R E -L u c a c tiv ity . Again, no overt signs of toxicity was noted in these mice; specifically, no enlargement of the spleen or evidence of immune complex deposition within the kidneys was detected (data not shown). These data further support the potential clinical translatability of ERX-11.
Importantly, ERX-11 had activity against ER-driven breast cancer cell lines that were either resistant to tamoxifen (MCF-7-TamR, Figure 6A, or MCF-7-HER2, Figure 6B) or to letrozole (MCF-7-LTLT, Figure 6C). In these cell lines, ERX-11 was still able to interact with the ER, both in the absence and presence of tamoxifen ( Figure 6D). In contrast to SERD (ICI), which had limited activity on the tamoxifen resistant cell lines, ERX-11 had potent activity ( Figure  We then evaluated the effect of ERX-11 against two prevalent ER mutants (MT-ESR1-Y537S, MT-ESR1-D538G) (Toy et al., 2013;Robinson et al., 2013;Jeselsohn et al., 2014;Merenbakh-Lamin et al., 2013). Using biotinylated ERX-11, we showed that ERX-11 interacts directly with ESR1-MT (ESR1-MT-D538G, ESR1-MT-Y537S), with high affinity comparable to the affinity to WT-ER ( Figure 6F). Using CRISPR/Cas9, we knocked down ER in ZR-75 cells and then stably transfected with WT-ESR1 or MT-ESR1 (Y537S, and D538G). While ESR1-MT expressing cells showed higher rates of proliferation than WT-ESR1-expressing cells ( Figure 6-figure supplement 1C), they were still inhibited by ERX-11 ( Figure 6G). Further the ability of these ESR1-MT to drive ligand-independent transcription from an ERE-Luc reporter was also efficiently blocked by ERX-11 ( Figure 6H). Further, these ESR1-MT expressing cells were resistant to tamoxifen, however, were sensitive to ERX-11-mediated growth inhibition ( Figure 6I). Further, oral ERX-11 administration had significant activity against the growth of ZR-75-ESR1MT-Y537S xenografts in vivo ( Figure 6J  . HEK-293T cells stably transfected with ERE-Luc vector were transiently transfected with control or coregulator-expressing vectors along with ER expression vector and after 24 hr treated with ERX-11. After 24 hr, the reporter gene activity was measured (C). Cell proliferation of ZR-75 model cells stably expressing PELP1 or AIB1 was measured using Cell Titer Glo assay (D, left panel). ZR-75 model cells stably expressing PELP1 or AIB1 were transfected with ERE-Luc reporter vector. After 48 hr, the cells were treated with ERX-11 and the reporter activity was measured 24 hr later (D right panel). The effects of ERX-11 on ER recruitment at ER target genes were examined using a ChIP assay in MCF-7 cells (E). The effect of ERX-11 on ER dimerization as evaluated by the NanoBiT luciferase assay is shown (F). ZR-75 cells were treated with E2 with or without ERX-11 (1 mM) for the indicated time, and the stability of ER was determined using western blotting. Quantitation of ER levels compared to control (E2-treated cells) was shown after normalizing to the levels of GAPDH (G). ZR-75 cells were treated with ERX-11 for 5 days, and the status of ER was determined using Western blotting (H). Data shown are the means of ±SEM performed in triplicate wells. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. DOI: 10.7554/eLife.26857.020 The following figure supplement is available for figure 4: ERX-11 has activity against primary patient derived breast tumor explants We recently developed an ex vivo culture model of primary breast and prostate tumors, which allows for the evaluation of drugs on breast tumors while maintaining their native tissue architecture Schiewer et al., 2012). In brief, surgically extirpated de-identified breast tissues are sliced into small pieces and grown ex vivo for a short term on a gelatin sponge in the absence or presence of ERX-11 ( Figure 7A). Incubation of ERX-11 with ER-positive breast tumor samples (patient characteristics detailed in Table 4) dramatically decreased their proliferation in 11/12 patients (Ki67 staining) compared to untreated controls ( Figure 7B). Further, ERX-11 treatment significantly reduced the ER staining ( Figure 7C), but not PELP1 staining (Figure 7-figure supplement 1) in 12/12 ER-positive tumors. Importantly, ERX-11 treatment had no effect on the proliferation on 6/6 triple negative breast cancer (TNBC) tumors ( Figure 7D). These results suggest that ERX-11 has the potential to selectively influence the growth of human breast tumors expressing ER.

Discussion
The majority of breast cancer is ER positive. Therapeutic agents that suppress oncogenic ER activation by depletion of hormone-driven growth signaling or by blocking synthesis of hormone have become the mainstay of systemic treatment for breast cancer. However, both de novo and acquired therapy resistance are major clinical challenges. Importantly, ER signaling is intact in these therapyresistant tumors. ER interaction with critical coregulator proteins appears to mediate ER signaling in these therapy-resistant and ER-positive metastatic tumors. While AIs and AEs may disrupt some of these ER-coregulator interactions, their ability to target these interactions is limited in therapy-resistant cells. In this study, we described the development of a novel inhibitor that targets ER interactions with coregulators. Using in vitro and in vivo assays, we demonstrated that (Musgrove and Sutherland, 2009) ERX-11 blocks both ligand-dependent and ligand-independent ER signaling, (Ma et al., 2015) ERX-11 effectively blocks ER signaling in both therapy-sensitive and therapy-resistant cells and (McDonnell and Norris, 2002) disruption of ER signaling by ERX-11 has biological activity against both therapy-sensitive and therapy-resistant cells, with minimal overt signs of toxicity in vitro and in vivo. Until recently, protein-protein interactions have been viewed as undruggable. However, the recent advent of a transformative class of compounds (peptidomimetics) allowed us to rationally design and synthesize small organic molecules that structurally emulate target protein sequences in defined conformations. We have developed a novel oligo-benzamide scaffold for mimicking helical protein segments (Ravindranathan et al., 2013;Ahn and Han, 2007). The rigid framework of the oligo-benzamide scaffold can present functional groups mimicking amino acid residues in a helical conformation (e.g. ones in i, i-3/4, and i + 7 positions). Furthermore, we have established efficient synthetic routes to make bis-and tris-benzamides as alpha-helix mimetics (Lee and Ahn, 2011; Figure 5 continued and quantitated. Apoptosis was measured using Caspase3 activation and by using TUNEL assay, and the number of TUNEL-positive and cleaved caspase 3 cells were counted in five different fields and plotted as histogram. DAPI was used to visualize the nuclei (A, B). Representative IHC analysis of ER and PELP1 performed on xenograft tumors that were treated with or without ERX-11 (B). Effect of ERX-11 on the growth of ER-positive D2A1 syngeneic tumors. Small pieces of D2A1 syngeneic tumors were implanted subcutaneously into the BALB/c mice. After 1 week, mice (n = 8) were treated with vehicle or ERX-11 (20 mg/kg/day). Tumor growth was measured at indicated time points. The body weights and extirpated tumor weights are shown. Ki-67 expression was analyzed by IHC and quantitated (C). The effect of ERX-11 on the coregulator-driven cell survival was measured by MTT assay using ZR-75 cells stably expressing SRC3/AIB1 or PELP1 (D). MCF-7-PELP1 cells were injected into the mammary fat pad of nude mice (n = 5) implanted with an estrogen pellet. After 3 weeks, mice were treated with vehicle or ERX-11 (10 mg/kg/day). Tumor volume, status of Ki-67 and apoptosis was shown (E). Data shown are the means of ±SEM. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. DOI: 10.7554/eLife.26857.022 The following figure supplements are available for figure 5: mM tamoxifen or 1 mM letrozole, respectively, and nuclear lysates were subject to biotin-ERX-11 pull down followed by western blotting with ER Figure 6 continued on next page Marimganti et al., 2009). Based on the helix mimicking tris-benzamide scaffold, we designed ERX-11 to block the interactions between ER and a subset of its coregulator proteins containing the NR box in a helical structure. The unbiased IPMS analyses suggest that ERX-11 interacts with ER and blocks the interactions of ER with multiple coregulators. While the broad effect of ERX-11 protein-protein interactions raises potential concerns about off-target activities, they are largely mitigated by the lack of overt signs of toxicity in cell culture models and multiple animal models tested to date. In addition, the multiplicity of targeted protein-protein interactions makes the development of resistance to the ERX-11 less likely. Thus, ERX-11 represents an exciting new mechanism to attenuate ER oncogenic functions.
Like tamoxifen, ERX-11 potently blocks the proliferation of therapy-sensitive cells. Unlike tamoxifen, ERX-11 has activity in multiple therapy-resistant models, including those driven by ER ligandbinding domain mutants. Unlike a classic SERD, ERX-11 does not cause immediate ER degradation, but appears to affect ESR1 levels over several days, by blocking its transcription.
Our data clearly indicate that ERX-11 binds to AF2 domain of ER. However, the exact interface between ERX-11 and ER has not been established. The ability of tamoxifen to compete efficiently for ERX-11 binding to purified ER in vitro and to ER within the cell suggests a significant overlap between the tamoxifen and ER-binding site on ER. In addition, the inability of the SERDs and ICI to disrupt the interaction between ER and ERX-11 suggests that ERX-11 may also interact with a secondary tamoxifen-binding site within ER. As shown for ER beta, tamoxifen has two distinct binding sites-one in the consensus ligand-binding pocket, and another in the hydrophobic groove of the coactivator recognition surface . Published studies using LXXLL peptide probes showed ER undergo distinct conformational changes as a result of binding with different ligands and such changes expose distinct surfaces on ER facilitating interaction with various coregulators (Paige et al., 1999). These studies also reported that tamoxifen binding can create unique surface on ER that facilitate binding of unique LXXLL-binding peptides (Tamrazi et al., 2002). Further investigations of the ERX-11-ER interface with co-crystallization studies are ongoing and are likely to clarify the precise nature of the interaction.
Our studies also suggest that ERX-11 can interact with ER, even in therapy-resistant cells. While therapy resistance can be attributed to multiple mechanisms, structural changes in ER via post-translational modifications or point mutations may create new binding surfaces on ER for coregulator interactions and potentially for ERX-11 interactions. The interaction between ERX-11 and ER-MTs in therapy-resistant cells supports a model for ERX-11 interaction with ER at an alternate secondary site distinct from the site of tamoxifen binding. These findings may explain the differences between ERX-11 and tamoxifen in their activity in both therapy-sensitive and therapy-resistant breast cancer cells.
The ability of ERX-11 to block ER dimerization may be responsible for its disruption of ER DNA binding and these findings are supported by published reports that LXXLL peptides may affect ER dimerization (Tamrazi et al., 2002). We have noted that ERX-11 has a weak affinity for the A-isoform of PGR but not for the B-isoform. We have also found that ERX-11 does not interact with other Figure 6 continued antibody (D). Following implantation and growth of ER-positive letrozole-resistant xenografts in nude mice (n = 8), mice were treated with control or ERX-11 (20 mg/kg/day). Tumor volume, tumor weight and Ki-67 status of control and treated tumors was shown (E). Nuclear extracts prepared from HEK-293T cells transiently transfected with WT-or MT-ER expression plasmids and analyzed for interaction between WT-and MT-ER to the biotin-ERX-11 using avidin pulldown followed by western blot analysis (F). Effect of ERX-11 on the cell viability of ZR-75 ESR1-KO cells stably expressing ESR1-WT or ESR1-Y537S mutant or ESR1-D538G mutant (G) was measured using MTT assay. ER-negative MDA-MB-231 cells were co-transfected with ERE reporter along with WT-ESR1 and MT-ESR1 plasmids. After 48 hr, the cells were treated with ERX-11 (500 nM) and the reporter activity was measured 24 hr later (H). Effect of ERX-11 and tamoxifen on the cell viability of ZR-75 cells stably expressing ER-Y537S mutant was measured using MTT assays (I). ZR-75 cells stably expressing ER-Y537S mutant were injected into the mammary fat pads of nude mice implanted subcutaneously with an estrogen pellet. After 2 weeks, mice with xenografts were treated with vehicle or ERX-11 (20 mg/kg/day, n = 6). Tumor growth was measured at indicated time points (J). Ki-67 expression was analyzed by IHC and quantitated (K). Data shown are the means of ± SEM. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. DOI: 10.7554/eLife.26857.025 The following figure supplement is available for figure 6:    Figure 7 continued on next page steroid receptors like GR or AR. In addition, ERX-11 failed to show activity on AR-expressing prostate cancer cells. Disruption of multiple protein-protein interactions by ERXs may result in significant toxicity. To evaluate this, we have initially tested toxicity using a dose of 100 mg/kg in immune competent C57BL/6 mice for 14 days. We did not observe any uterotropic activity or immune effects. Further, no overt signs of toxicity was found in various E2-responsive organs including the liver, lung, heart, and kidney. Similarly, in five separate studies using a dose of 10 mg/kg in tumor-bearing mice, we did not find any overt signs of toxicity; however, ERX-11 treatment significantly limited ER expression in the tumors and reduced tumor growth. To address potential issues with immunogenicity of these tris-benzamides, we also evaluated ERX-11 treatment in syngeneic models and found no overt signs of toxicity.
While the ER coregulator protein levels are tightly regulated under normal conditions, many coregulators are over-expressed in breast cancer (Lonard et al., 2007), substantially contribute to ERsignaling, drive disease progression (Singh and Kumar, 2005) and correlate with a poor prognosis (List et al., 2001;Azorsa et al., 2001;Tamrazi et al., 2002). The differential coregulator milieu within breast tumors and the dependence of breast tumors on ER and coregulator-driven signaling may explain why ERX-11 has potent antiproliferative activity within the tumor but does not have any overt signs of toxicity. Since ERX-11 blocked some but not all ER coregulator interactions, ERX-11 may function as a polypharmacology agent, and its activity may depend on the concentration and repertoire of coregulators present in a tumor cell.
Importantly, using ex vivo culture of patient-derived tumor tissues, we demonstrated that ERX-11 is effective in limiting proliferation of ER-positive but not ER-negative tumors. We also discovered that ERX-11 treatment of primary tumors within their native microenvironment also reduced ER levels within the tumor. None of the primary ER-negative tumors responded to ERX-11 therapy. These explant studies represent the first evaluation of a drug effect using tissues from primary breast cancer patients and are likely to show biologically relevant outcomes.  Table 4. Clinicopathologic characteristics of the 12 patients, whose ER+, PR + status in breast tumors were analyzed by Ki67 and ER staining. This data is related to Figure 7B,C. In response to hormone binding, ER interacts with multiprotein complexes containing coregulators and transcriptional regulators to activate transcription (McKenna et al., 1999;Collingwood et al., 1999;Tsai and O'Malley, 1994;Torchia et al., 1998). Even though coregulators modulate ER functions, each coregulator protein appears to play an important but not overlapping function in vivo (Han et al., 2006;Xu et al., 1998). Accordingly, the ERX-mediated blockage of coregulator interactions with ER resulted in both inactivation and activation of unique sets of genes and pathways modulated by ER oncogenic signaling, leading to tumor suppression. The pathways/ genes modulated by ERX-11 can be used to correlate with the outcome of its therapy, and they may serve as biomarkers that prognosticate response to these agents.
The biology of E2-ER signaling is complex and context dependent. Elegant studies have shown that ER, depending on the ligand and presence of unique set of coregulators can promote apoptosis (Jordan, 2015). In that scenario, antiestrogen and SERM are shown to inhibit apoptosis, hence, many of these drugs exhibit cytostatic response. Estrogen-induced apoptosis is shown to cause an increase in Fas receptor associated with the extrinsic pathway of apoptosis (Lewis-Wambi and Jordan, 2009). Similarly, a recent study found that inhibition of SRC family coregulators using small molecule inhibitor promote significant apoptosis (Song et al., 2016). We consistently observed activation of apoptosis by ERX-11 but failed to observe any apoptosis by tamoxifen or ICI in our assays. We believe that activation of apoptosis is not due to elimination of ER rather due to unique mechanism of action of ERX-11. Specifically, we predict that changes in the ER signaling is due to alterations in coregulator binding to ER. Accordingly, our RNAseq data showed that alterations in genes that contribute activation of apoptosis. However, further studies are needed to clearly identify the mechanisms by which ERX-11 promotes apoptosis.
Currently used drugs (AEs and AIs) are associated with an initial period of clinical response; however, most patients develop resistance with cancer progression. Recent studies suggested that selective estrogen receptor downregulators (SERDs), molecules that eliminate ER expression, may have utility for treating breast cancers that have progressed on AE and/or AIs (McDonnell et al., 2015). Several orally available SERDs (GDC-0810, AZD9496) were recently developed and shown to have utility in treating resistant tumors using preclinical models (Lai et al., 2015;Weir et al., 2016). However, it is important to note that each of the SERDs are only in phase I clinical trials (clinical trials. gov) and none of them are either FDA approved or have proven efficacy in patients. The only FDAapproved agent is Fulvestrant (ICI), which has known pharmacologic limitations as evidenced by its dosing regimen of intramuscular injections every 14 days. Recent studies also showed ESR1 mutations lead to constitutive activity and reduced sensitivity to ER antagonists, and mutations such as Y537S contribute to fulvestrant resistance in vivo (Toy et al., 2017). Since ERXs are small, stable, orally bioavailable small molecular inhibitors, their use as an alternative therapeutic approach may decrease therapy resistance and reduce side effects, which are the current limitations of AEs or AIs.
In summary, we have developed and tested the utility of ERX-11 as a novel therapeutic agent for ER-positive, therapy-sensitive and therapy-resistant breast cancers. Since ERX-11 is orally available and is well tolerated with fewer side effects, ERX-11 can be readily extended to clinical use as a therapeutic agent, and may enhance the survival of advanced breast cancer patients. . All of these cells were passaged in the user's laboratory for fewer than 6 months after receipt or resuscitation. Validation experiments were performed using a number of luminal, basal and TNBC cells with extensive prior available molecular profiles were a kind gift of Dr. John Minna. All the model cells utilized are free of mycoplasma contamination. Additionally, STR DNA profiling of the cells was used to confirm the identity using UTHSA and UT Southwestern core facilities. MCF-7-PELP1 cells (Vallabhaneni et al., 2011), MCF-7-HER2 cells (Nabha et al., 2005), MCF-7-TamR cells (Nabha et al., 2005), MCF-7-LTLTca cells (Macedo et al., 2008) and D2A1 cells (Tekmal and Durgam, 1997) were described earlier. MCF-7-LTLTca and MCF-7-TamR cells were cultured in Phenol red-free RPMI medium containing 10% dextran charcoal-treated serum supplemented with either 1 mmol/L of letrozole or 1 mmol/L of tamoxifen, respectively.

Cell viability assays
The effects of ERX analogues on cell viability were measured using the MTT Cell Viability Assay in 96-well plates. Breast cancer cells were seeded in 96-well plates (1 Â 10 3 cells/well) in phenol redfree RPMI medium containing 5% dextran-coated charcoal-treated fetal bovine serum (DCC-FBS) serum. After an overnight incubation, cells were treated with varying concentrations of the ERX analogues in the presence or absence of E2 (1 Â 10 À8 M) for 7 days. For some experiments viability was also measured using Cell Titer-Glo Luminescent Cell Viability Assay (Promega) in 96-well, flat, clearbottom, opaque-wall micro plates according to manufacturer's protocol. For some experiments, apoptosis was measured using Caspase-Glo 3/7 Assay (Promega) using manufacturer's protocol.

Animal studies
All animal experiments were performed after obtaining UTHSA IACUC approval and using methods in the approved protocol. For xenograft tumor assays, 2 Â 10 6 ZR-75, or ZR-75-ER MT-Y537S, or MCF-7-PELP1 cells were mixed with an equal volume of matrigel and implanted in the mammary fat pads of 6-week-old female athymic nude mice as described (Cortez et al., 2012). Based on our previous data as well as published findings, the number of mice needed were chosen to demonstrate differences in tumor incidence or treatment effect. Calculations are based on a model of unpaired data power = 0.8; p<0.05. Once tumors reached measurable size, mice were divided into control and treatment groups (n = 5-8 tumors per group). The control group received vehicle and the treatment groups received ERX-11 (10 mg/kg/day) in 30% Captisol orally. Dose were selected based on pilot MTD study of 10, 50 and 100 mg/kg of ERX-11 for 14 days using C57BL/6 mice. The mice were monitored daily for adverse toxic effects. For MCF-7-LTLT xenograft studies, MCF-7-LTLT model cells were first injected into the mammary glands of nude mice implanted with androstenedione pellets. When the tumor was established, it was dissected into small pieces and they were again implanted subcutaneously into nude mice implanted with androstenedione pellets. Fulvestrant treatment was used as a positive control and MCF-7-LTLT xenografts were treated with 200 mg/kg/ 2 days a week/sc. For syngeneic mice studies, D2A1 cells were first injected into the mammary glands of BALB/c mice. When the tumor was established, it was dissected into small pieces and they were again implanted subcutaneously into the BALB/c mice. After three days of tumor tissue implantation, mice were randomly selected to receive control (n = 7-8) and treatment (n = 7-8) with 20 mg/kg/ day of ERX-11 orally. Tumor growth was measured with a caliper at 3-4 day intervals. At the end of each experiment, the mice were euthanized, and the tumors were removed, weighed and processed for IHC staining.

Patient-derived explant (PDEx) studies
UTSW Patients provided written consent allowing the use of discarded surgical samples for research purposes according to an institutional board-approved protocol. De-identified patient tumors were obtained from the UTSW Tissue Repository after institutional review board approval (STU-032011-187). Excised tissue samples were processed and cultured ex vivo as previously described (Mohammed et al., 2015). Briefly, tissue samples were incubated on gelatin sponges for 48 hr in culture medium containing 10% FCS, followed by treatment with either vehicle or E2 (10 nM) in the absence or presence of 10 mM ERX-11 for 48 hr (see Table 4 for clinicopathologic characteristics of these tumors). Representative tissues were fixed in 10% formalin at 4˚C overnight and subsequently processed into paraffin blocks. Sections were stained with hematoxylin and eosin and examined to confirm and quantify the presence/proportion of tumor cells. Immunohistochemistry was then performed.

Protein interaction analyses
String analyses were performed for human PELP1 using the http://string-db.org website, with the evidence view at the highest stringency for no more five interactors.

Conformational analysis of ERX-11
A Monte Carlo conformational search was performed using the torsional sampling method (MCMM) implemented in MacroModel (version 9.0, Schrö dinger, New York, NY) with automatic setup options. The calculation was done with the maximum number of steps set to 5000 using 100 steps per rotatable bond and an energy cutoff of 21 kJ/mol above the global energy minimum. The searches were done using MM3 force field (chosen for its accuracy with organic molecules) combined with the GB/ SA water solvation model with standard settings and the following cut-offs: van der Waals, 8.0 Å ; electrostatic, 20.0 Å ; and hydrogen bond, 4.0 Å . The observed conformations were minimized by 500 iterations of Polak-Ribiere Conjugate Gradient (PRCG) algorithm (a conjugate gradient minimization scheme that uses the Polak-Ribiere first derivative method with restarts every 3N iterations) (0.05 kJ/mol).

Molecular docking of ERX-11 to ER
AutoDock 4.2 software package, as implemented through the graphical user interface called Auto-DockTools (ADT), was used to create input PDBQT files of a receptor and a ligand. The input file of ER was prepared using the published coordinates (PDB 1L2I). Water molecules were removed from the protein structure and hydrogen was added. All other atom values were generated automatically by ADT. The docking area was assigned visually around the peptide ligand. A grid box of 24 Å x 20 Å x 24 Å was calculated around the docking area using AutoGrid. The x,y,z coordinates of the center of the grid box were set to x = À9.0, y = 14.0 and z = 26.0, respectively. The input file of ERX-11 was created from its energy-minimized conformation using ADT. Docking calculations were performed with AutoDock Vina 1.1.2. A search exhaustiveness of 16 was used and all other parameters were left as default values.

Reporter gene assays
Briefly, cells were transiently co-transfected with 200 ng of ERE-Luc reporter with 100 ng of ER-WT, ER-MT, PELP1, SRC1, SRC2, SRC3 or control vectors using Turbofect transfection reagent (Thermo Scientific, Waltham, MA). After 24 hr, cells were treated with either vehicle or ERX-11 for an additional 24 hr. b-galactosidase reporter (50 ng) plasmid was co-transfected and used for data normalization. Cells were lysed in Passive Lysis Buffer, and luciferase activity was measured using the luciferase assay system (Promega, Madison, WI) in a luminometer.

RNA sequencing and RT-qPCR
RNA-seq was performed using the UTHSA core-established protocol. Briefly, ZR-75 cells were treated with either vehicle or ERX-11 for 48 hr, and total RNA was isolated using RNAesy mini kit (Qiagen) according to the manufacturer's instructions. Differential expression analysis was performed by DEseq and significant genes with at least 1.5-fold change with p<0.01 were chosen for analysis. The interpretation of biological pathways using RNA-seq data was performed with IPA software using all significant and differentially expressed genes. RNA-seq data have been deposited in the GEO database under accession number GSE75664. To validate the selected genes, reverse transcription (RT) reactions were performed by using SuperScript III First Strand kit (Invitrogen, Carlsbad), according to manufacturer's protocol. Real-time PCR was done using SybrGreen on an Illumina Real-Time PCR system, using primers listed in Table 5.

ChIP
Chromatin immunoprecipitation (ChIP) analysis was performed using antibodies specific for the ER (Santa Cruz). Briefly, MCF-7 (7 Â 10 6 ) or T-47D (2 Â 10 7 ) cells were plated in 150 mm dishes, starved in unsupplemented phenol red-free DMEM for 24 hr and then treated for 2 hr with either ethanol or E2 after prior incubation with either DMSO or ERX-11. Relative recruitment was determined by qPCR of purified ChIP and input DNA in triplicate. The results presented are representative of two independent experiments.

NanoBiT luciferase studies
The NanoBiT assay utilizes a structural complementation-based approach to monitor protein-protein interactions within living cells. Large BiT (LgBiT; 18 kDa) and Small BiT (SmBiT; 1 kDa) subunits of NanoLuc Luciferase were optimized for the analysis of protein interaction dynamics. When LgBiT and SmBiT subunits are separated, the Large BiT part loses the majority of luciferase activity. However, when the direct interaction between fusion proteins on LgBiT and SmBiT occurred, the interaction promotes structural complementation between LgBiT and SmBiT and results in full luciferase activity. Protein-protein interaction are then monitored in living cells following addition of the Nano-Glo Live Cell Reagent, a non-lytic detection reagent containing the cell-permeable furimazine substrate and observed luminescent signals.
To generate different NanoBiT fusion constructs, human ER and PELP1 coding sequences were amplified by PCR and separately subcloned into NB-MCS vectors (Promega). To test the protein- Table 6. Analyses of the amino acids at the flanking sequences of top ER binders whose interactions are blocked by ERX-11 in MCF-7 and ZR-75 cells, as determined by unbiased IP-MS. protein interaction between ER and PELP1 by using the NanoBiT assay, C-LgBit-ESR1 paired with C-SmBit-PELP1 or C-LgBit-PELP1 with C-SmBit-ESR1 constructs were transiently transfected into HEK-293T cells by using Fugene HD transfection reagent (Promega). To test the ER dimerization, C-SmBit-ESR1 were cotransfected with either N-LgBit-ESR1 or C-LgBit-ESR1 constructs. On the day after transfection, the medium for the HEK-293T cells was changed to phenol red-free DMEM containing 1% charcoal-stripped FBS. After a 24 hr incubation, the cells were treated with DMSO or ERX-11 (10 mM) for 2 hr and then treated cells with EtOH or E2 (10 nM) for 30 mins. After treatment, Nano-Glo live cell reagents were added into cells and luminscence was measured after 10 mins.
Proximity ligation assays MCF-7 cells were cultured on collagen-coated cover slips. After treated with vehicle or 10 mM ERX-11 for 30 min, cells are treated with vehicle or 10 nM E2 for 30 min. After the treatment, cells were washed with PBS and then fixed with 10% buffered formalin for 20 min and permeabilized with ice cold methanol at À20˚C for 5 min. The cells were then blocked with blocking solution (provided with Duolink In Situ PLA probe, Sigma) for 30 min at 37˚C followed by incubating with primary antibodies for ER from different species at 37˚C for 2 hr. After washing twice with Wash Buffer A (Duolink In Situ Reagents) at room temperature, cells were then incubated with the appropriate anti-species secondary antibodies to which oligonucleotides had been conjugated (anti-rabbit PLA probe PLUS and anti-mouse PLA probe MINUS) for 1 hr at 37˚C, followed by treatment with Duolink ligationligase solution for 30 min at 37˚C. Finally, the cells were incubated with the Duolink amplificationpolymerase solution for 60 min at 37˚C, followed by washing and mounting on slides with Duolink Mounting Medium with DAPI. Images were taken using a Nikon Fluorescence Microscope.

IPMS
MCF-7 or ZR-75 cells were grown in RPMI-1640 medium supplemented with 10% FBS. After cells reach 90% confluence, the medium was changed to phenol red-free RPMI1640 containing 1.5% charcoal stripped serum for 48 hr to starve the cells. After starvation, the cells were treated with vehicle or 10 mM ERX-11 for 2 hr followed by a 2 hr treatment with vehicle or 10 nM E2. Then, the cells were washed using PBS and incubated with Pierce IP Lysis Buffer (Thermo Fisher Scientific Inc.) at 4˚C for 20 min. The cell lysates were centrifuged at 14,000 rpm at 4˚C for 10 min, and the supernatants were used for IPMS analysis. Dynabeads Protein G (Invitrogen) are pre-coupled with anti-ER antibody (Santa Cruz, sc-8002) and then incubated with 1.5 mg cell lysate over night at 4˚C. Then the Dynabeads were washed using PBS and PBST (0.01% Tween 20) and eluted in 30 mL NuPAGE LDS sample buffer at 95˚C for 15 min. The final eluents containing ER and associated proteins were separated by using SDS-PAGE. The obtained proteins were proteolytically digested and subjected to mass spectrometry (MS) analysis.
Gel band samples were digested overnight with trypsin (Promega) following reduction and alkylation with DTT and iodoacetamide (Sigma). The samples then underwent solid-phase extraction cleanup with Oasis HLB plates (Waters) and the resulting samples were analyzed by LC/MS/MS, using an Orbitrap Fusion Lumos (Thermo Electron) coupled to an Ultimate 3000 RSLC-Nano liquid chromatography systems (Dionex). Samples were injected onto a 75 mm i.d., 50 cm long Easy Spray column (Thermo) and eluted with a gradient from 0% to 28% buffer B over 60 min. Buffer A contained 2% (v/v) ACN and 0.1% formic acid in water, and buffer B contained 80% (v/v) ACN, 10% (v/v) trifluoroethanol, and 0.08% formic acid in water. The mass spectrometer operated in positive ion mode with a source voltage of 2.2 kV and capillary temperature of 275˚C. MS scans were acquired at 120,000 resolution and up to 10 MS/MS spectra were obtained in the ion trap for each full spectrum acquired using higher-energy collisional dissociation (HCD) for ions with charge 2-7. Dynamic exclusion was set for 25 s.
Raw MS data files were converted to a peak list format and analyzed using the central proteomics facilities pipeline (CPFP), version 2.0.3. Peptide identification was performed using the X!Tandem and open MS search algorithm (OMSSA) search engines against the human protein database from Uniprot, with common contaminants and reversed decoy sequences appended. Fragment and precursor tolerances of 20 ppm and 0.1 Da were specified, and three missed cleavages were allowed. Carbamidomethylation of Cys was set as a fixed modification and oxidation of Met was set as a variable modification. Label-free quantitation of proteins across samples was performed using SINQ normalized spectral index Software.

Immunohistochemistry
For immunohistochemical analysis sections were incubated overnight with the ER (1:50) or PELP1 (1:200) or Ki67 (1:100) antibody in conjunction with proper controls. The sections were then washed three times with 0.05% Tween in PBS for 10 min, incubated with secondary antibody for 1 hr, washed three times with 0.05% Tween in PBS for 10 min, visualized by DAB substrate and counterstained with hematoxylin QS (Vector Lab, Burlingame, CA). A proliferative index was calculated as the percentage of Ki-67-positive cells in five randomly selected microscopic fields at 40X per slide. TUNEL analysis was done using the in situ Cell Death Detection Kit (Roche, Indianapolis, IN) as per the manufacturer's protocol, and five randomly selected microscopic fields in each group were used to calculate the relative ratio of TUNEL-positive cells.
For the PDEx samples, 5-mm sections were de-waxed, rehydrated and endogenous peroxidases were blocked with hydrogen peroxide. Sections were then boiled in citrate and blocked in 5% serum for 1 hr. Primary antibodies were incubated overnight at 4˚C at 1:200 for Ki67 (Vector Laboratories Burlingame, CA) and at 1:400 for ER (Santa Cruz Biotechnology, CA). Biotinylated anti-rabbit secondary antibodies (DAKO Carpentaria, CA) were incubated for 60 min at room temperature after slides were washed for 1 hr in PBS. Slides were incubated in ABC-HRP complex (Vector Laboratories Burlingame, CA) for 30 min. Bound antibodies were then visualized by incubation with 3,3' diaminobenzidine tetrahydrochloride (liquid DAB, DAKO). Slides were then rinsed in tap water and counterstained with hematoxylin, and then cover slide were mounted. Tumor cells with nuclear staining were recorded as positive manually per tissue core, by a reviewer who was blinded to the clinical data. The numbers of Ki67-positive tumor cells were counted in three high-power fields (x40). The ER immunostaining was registered semi-quantitatively in two ways. Staining intensity (0, no staining; 1, weak staining; 2, moderate staining; and 3, intense staining) and the proportion of stained cells (0, no staining; 1, 1-25% staining; 2, 26-50%; 3, 51-75%; and 4, if more than 75% of the tumor cells were positive.

Statistics
GraphPad Prism 6 software (GraphPad Software, SanDiego, CA) was used to analyze all data. Data represented in the bar graphs is shown as mean ± SEM. t-test was performed for all pairwise comparisons. A value of p<0.05 was considered as statistically significant. The multiple groups' statistical data were analyzed with one-way ANOVA. RNA-seq data were analyzed using IPA software.

Ethics
Animal experimentation: All animal experiments were performed after obtaining UTHSCSA IACUC approval and using methods in the approved protocol (15039X).

Major datasets
The following dataset was generated: General procedure for the hydrolysis of esters A 500 mL round-bottomed flask was charged with methyl 3-alkoxy-4-nitrobenzoate 2 (31 mmol), MeOH (200 mL), THF (200 mL) and 10% NaOH (30 mL). Alternatively, 1N LiOH (60 mL) was used for the synthesis of 3 f. The reaction mixture was stirred at room temperature for 12 hr. The volatile was removed under reduced pressure, and the resulting residue was acidified with 1 N HCl (200 mL) and then extracted with EtOAc (300 mL). The organic layer was separated and washed with 1 N HCl (100 mL) and brine (100 mL). The organic layer was then dried over Na 2 SO 4 , filtered, and concentrated under reduced pressure to give the compound three as yellow solid.

General procedure for cleavage of tris-benzamide
Resin-bound tris-benzamide seven was washed with DMF (3 Â 2 min) and DCM (3 Â 2 min), and dried in vacuo. The dried resin was treated with a cleavage mixture of TFA/TIS/H 2 O (95:2.5:2.5, 8 mL) for 90 min. The TFA solution was then filtered, and the resin was washed with TFA (2 mL). The combined TFA solution was concentrated to a volume of approximately 0.5 mL with a gentle stream of nitrogen, and the tris-benzamide was precipitated with cold diethyl ether (10 mL). The resulting precipitate was collected by centrifugation and the ether solution was decanted. Washing with cold diethyl ether was repeated and the tris-benzamide was dried in vacuo.
To a solution of the carboxylic acid (300 mg, 0.35 mmol) and DIEA (0.24 mL, 1.4 mmol) in DMF (10 mL) was added HATU (179 mg, 0.46 mmol) and the mixture was stirred at room temperature for 1 hr. N-Boc-ethylenediamine (112 mg, 0.70 mmol) was added and the resulting mixture was stirred at room temperature for 24 hr. The solution was then poured into 5% citric acid (20 mL) and extracted with EtOAc (20 mL x 2). The combined organic extracts were washed with saturated NaHCO 3 (20 mL x 2), dried over Na 2 SO 4 , filtered, and concentrated under reduced pressure. Recrystallization from EtOAc/hexanes (1:1) gave compound 12 as a light yellow solid (152 mg, 44% A solution of compound 12 (50 mg, 0.050 mmol) in TFA/TIS/H 2 O (95:2.5:2.5, 5 mL) was stirred at room temperature for 30 min. The excess of TFA were removed with a gentle stream of nitrogen, and a white solid was precipitated with cold diethyl ether. The precipitate was filtered, washed with diethyl ether and dried in vacuo. The resulting yellow solid was used in the next reaction without further purification.