A Genome-Edited ERα-HiBiT Fusion Reporter Cell Line for the Identification of ERα Modulators Via High-Throughput Screening and CETSA

Abstract

The estrogen receptor α (ERα) is a target of intense pharmacological intervention and toxicological biomonitoring. Current methods to directly quantify cellular levels of ERα involve antibody-based assays, which are labor-intensive and of limited throughput. In this study, we generated a post-translational reporter cell line, referred to as MCF7-ERα-HiBiT, by fusing a small pro-luminescent nanoluciferase (NLuc) tag (HiBiT) to the C-terminus of endogenous ERα in MCF7 cells. The tag allows the luminescent detection and quantification of endogenous ERα protein by addition of the complementary NLuc enzyme fragment. This MCF7-ERα-HiBiT cell line was optimized for quantitative high-throughput screening (qHTS) to identify compounds that reduce ERα levels. In addition, the same cell line was optimized for a qHTS cellular thermal shift assay to identify compounds that bind and thermally stabilize ERα. Here, we interrogated the MCF7-ERα-HiBiT assay against the NCATS Pharmacological Collection (NPC) of 2,678 approved drugs and identified compounds that potently reduce and thermally stabilize ERα. Our novel post-translational reporter cell line provides a unique opportunity for profiling large pharmacological and toxicological compound libraries for their effect on ERα levels as well as for assessing direct compound binding to the receptor, thus facilitating mechanistic studies by which compounds exert their biological effects on ERα.

Introduction

Estrogen receptors (ERs) belong to the class I nuclear receptor superfamily and function as ligand-inducible transcription factors. There are two ER isoforms, ERα and ERβ, which differ in expression pattern, ligand specificity, and have both common as well as distinct physiological functions.^1,2 ERα is activated by steroid hormones, in particular 17β-estradiol (E2), which results in either direct binding of ERα to estrogen response elements (ERE) in the promoter of estrogen-responsive genes or protein–protein interactions with other transcriptional modulators to regulate the expression of estrogen-responsive genes.^1,2 In turn, ligand binding to the ERα significantly modulates the receptor's stability via the ubiquitin-proteasome system.^3–5

ERα is the predominant isoform in the breast and is key to normal mammary gland development as well as to the initiation and progression of breast cancer among other diseases.⁶ ERα is expressed in 50%–80% of breast tumors, and if present is the primary therapeutic target for drug intervention with anti-estrogenic therapies designed to antagonize ERα function and signaling.⁷ Selective ER modulators (SERMs), such as Tamoxifen and Raloxifene, are competitive inhibitors of estrogen-receptor binding; however, many ERα-positive breast cancers are refractory to these agents by either intrinsic or acquired resistance. In addition, SERMs also function as agonists depending on the target tissue.⁸ On the other hand, selective ER down-regulators (SERDs), such as Fulvestrant, are pure antagonists with no known agonistic activity and function by inducing proteasome-mediated degradation of ERα.³ However, Fulvestrant and similar compounds have low oral bioavailability and limited clinical efficacy.⁹ Hence, the development of new SERDs with improved properties is an active field of research in cancer drug discovery.^10–12

In addition to the pharmacological implications described earlier, assessing the effect of compounds on ERα levels has important toxicological implications since disruption of ERα signaling by so-called “endocrine disruptors” has been shown to have profound effects on human health.^13,14 Thousands of compounds such as environmental chemicals are regularly tested by the United States Environmental Protection Agency (U.S. EPA) for their potential to be endocrine disruptors.^15–17

High-throughput cellular assays to assess ERα signaling typically involve transactivation assays, estrogen-responsive reporter cell lines in which a reporter gene is driven by an ERE-containing promoter.¹⁸ These assays identify compounds that directly and indirectly modulate ER activity. However, current methods to directly quantify cellular levels of the receptor involve antibody-based assays that are labor-intensive and of limited throughput.^10,19 Here, we used CRISPR/Cas9 genome-engineering to fuse a small reporter peptide to the C-terminus of endogenous ERα in MCF7 cells. The reporter consists of a nanoluciferase (NLuc) pro-luminescent tag of 11 amino acids (HiBiT), which allows for the detection and quantification of endogenous ERα protein by addition of the complementary Large NLuc (LgBiT) enzyme fragment.²⁰ This novel post-translational reporter cell line was used to optimize a high-throughput screening assay to identify compounds that reduce ERα levels. We interrogated the assay against the NCATS Pharmacological Collection (NPC)^21,22 of 2,678 approved drugs and identified 73 compounds that potently reduce ERα. In addition, we used this cell line to develop a high-throughput cellular thermal shift assay (CETSA) to identify compounds that bind and thermally stabilize ERα. The cell line described here provides a powerful tool for profiling large compound libraries and for detailed studies of mechanisms by which compounds exert their biological effects on ERα signaling and function.

Materials and Methods

Cell culture

MCF7 cells were obtained from ATCC (HTB-22). Parental and ER-HiBiT reporter MCF7 cells were cultured in DMEM (4.5 g/L glucose, GlutaMAX™-I; Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Hyclone, GE), 100 U/mL penicillin, and 100 μg/mL streptomycin (Thermo Fisher Scientific) at 37°C in a humidified incubator maintained at 5% CO₂. Three days before experiments, media were removed and replaced with culture media containing 10% charcoal-stripped fetal bovine serum (Hyclone, GE). Cells were routinely tested for mycoplasma by using a MycoAlert detection kit (Lonza).

Generation of ERα-HiBiT reporter cell line

To allow RNP complex formation, 2.5 μL of 61 μM Alt-R S.p. Cas9 Nuclease V3 (Integrated DNA Technologies, IDT) and 1.88 μL 100 μM ESR1 gRNA 5′- mG*mC*mC* rArCrG rGrUrC rUrGrA rGrArG rCrUrC rCrCrG rUrUrU rUrArG rArGrC rUrArG rArArA rUrArG rCrArA rGrUrU rArArA rArUrA rArGrG rCrUrA rGrUrC rCrGrU rUrArU rCrArA rCrUrU rGrArA rArArA rGrUrG rGrCrA rCrCrG rArGrU rCrGrG rUrGrC mU*mU*mU* rU -3′ (IDT) were transferred to an Eppendorf tube containing 5.63 μL of sterile RNase-free TE Buffer (10 mM Tris-HCL, 0.1 mM EDTA), mixed lightly, and incubated for 20 min at room temperature (RT). Subsequently, 2 μL of 100 μM ESR1-HiBiT HDR DNA (5′-TCGCATTCCTTGCAAAAGTATTACATCACGGGGGAG GCAGAGGGTTTCCCTGCCACGGTCGTGAGCGGCTGGCGGCTGTTCAAGAAGATTAGCTGAGAGCTCCCTGGCTCCCACACGGTTCAGATAATCCCTGCTGCATTTTACCCTCATCATGCA-3′ IDT) was transferred to the tube, followed by the addition of 100 μL of a 10 × 10⁶ cells/mL solution of MCF7 cells in SE Cell Line Nucleofector Solution (Lonza). Nucleofection controls include positive control GFP [1 μL GFP vector diluted in 11 μL IDT Duplex Buffer (IDT)] and negative control (2.5 μL of 61 μM Alt-R S.p. Cas9 Nuclease V3 diluted in 9.5 uL TE Buffer). Experimental and control solutions were mixed, transferred to Nucleocuvette Vessels (Lonza), and transfected by using the EN-130 system protocol of the 4D-Nucleofector Core Unit (Lonza). After transfection, cells were incubated for 5 min at RT, and they were subsequently seeded in a 6-well dish with 200 μL pre-warmed culture media and incubated for 48 h at 37°C. Limiting dilution was performed to isolate monoclonal populations. Briefly, cells were harvested, diluted to a density of 0.5 cells/100 μL, plated into Corning CellBind Clear Bottom 96-well Microplates containing 100 μL of pre-warmed culture media, and incubated for 120 h at 37°C. Wells containing cells were trypsinized and resuspended in 100 μL pre-warmed culture media. To identify clones containing HiBiT-tagged cells, 50 μL of cell suspension was aliquoted into white Corning 96-well plates and luminescence was quantified by the addition of 50 μL/well of Nano-Glo HiBiT Lytic Detection Reagent (Promega) as per the manufacturer's instructions by using a ViewLux high-throughput CCD imager (PerkinElmer) equipped with clear filters. The remaining 50 μL of cell suspension was reseeded for expansion. Genomic DNA of positive clones as well as parental MCF7 cells was isolated by using DNeasy Blood & Tissue kit (Quiagen) as per the manufacturer's instructions. The genomic DNA was used to PCR amplify a fragment encompassing the insertion site. The amplicon was obtained by using a forward primer complementary to the HiBiT tag (HiBiT-Fw: 5′- CGGCTGTTCAAGAAGATTAGC-3′) and a reverse primer complementary to the ESR1 gene downstream of the insertion site (ESR1-Rv: 5′- CGGGAATCCTCACGCTTAGT-3′). The amplicon was gel purified by using the NucleoSpin Gel & PCR Clean-up kit (Takara) as per the manufacturer's instructions and sequenced with the HiBiT-Fw primer. The MCF7-ERα-HiBiT clone #2 cell line generated is also identified as NCATS-CL6791.

Western blotting

Cells were collected after indicated treatment and lysed in RIPA buffer supplemented with 1X Halt protease inhibitor cocktail (Thermo Fisher Scientific). Lysates were incubated on ice for 30 min and clarified by centrifugation at 15,000 g for 15 min at 4°C. Samples were quantified by DC protein assay (Bio-Rad) as per the manufacturer's instructions. Samples were run on a 4%–12% Bis-Tris NuPAGE gel (Thermo Fisher Scientific) at 150 V for 50 min by using 1X MOPS buffer and transferred to an NC membrane by using an iBlot transfer system at 25 V for 7 min. For ERα and loading control GAPDH, membranes were blocked with a 5% milk solution in tris buffered saline-tween (TBS-T) (20 mM Tris HCl, pH 7.5; 150 mM NaCl; 0.1% Tween-20); primary antibodies (Anti-ERα antibody; Bio-Rad MCA1799T, 1:2,000) and Anti-GAPDH antibody (G8795; 1:10,000; Sigma Aldrich) were incubated overnight at 4°C in 5% milk solution. Membranes were washed with TBS-T (three washes of 5 min each), incubated with secondary antibody (Anti-mouse-HRP; Cell Signaling 7076; 1:10,000) in 5% milk solution, and incubated for 1 h at RT. Membranes were washed with TBS-T as described earlier, and chemiluminescent signal was detected by using SuperSignal West Dura solution (Thermo Fisher Scientific) and a Bio-Rad ChemiDoc Imager. For detection of NLuc, membranes were incubated in 10 mL Nano-Glo HiBiT Blotting System (Promega) for 30 min, as per the manufacturer's instructions, and chemiluminescent signal was detected by using a Bio-Rad ChemiDoc Imager.

HiBiT reporter assay

Protocol can be found in Supplementary Table S2. Briefly, MCF7-ERα-HiBiT cells were dispensed at a density of 3 × 10⁵ cells/mL (5 μL cell/well) into white solid-bottom TC-treated 1,536-well plates (Greiner, #789173) in assay medium by using a Multidrop Combi dispenser (Thermo Fisher Scientific). For the primary screen, compound and control solutions were added to the plates via pin tool transfer (Wako Automation) in the concentration range of 2.5 nM to 39.8 μM (7-point compound titration). For validation screens, compounds were added via acoustic transfer (Echo 550, Labcyte) in the concentration range of 0.8 nM to 49.7 μM (11-point compound titration) and tested in triplicate. Neutral control dimethyl sulfoxide (DMSO) and positive control Tamoxifen (49 μM final) were plated in column 1–4 in each plate. The plates were incubated for 24 h at 37°C in a humidified incubator maintained at 5% CO₂, media were subsequently removed, and 3 μL/well of NanoGlo HiBiT Lytic Detection Regent was added by using a Multidrop Combi dispenser. Luminescence was measured in a ViewLux plate reader as described earlier.

Cell viability assay

Protocol can be found in Supplementary Table S3. Briefly, MCF7-ERα-HiBiT cells were dispensed and treated as described earlier with two main differences: first, the positive control was Digitonin (99 μM final); second, viability was assessed by the addition of 3 μL of CellTiter-Glo (CTG, Promega) without media removal. Plates were read by using a ViewLux plate reader equipped with clear filters.

NLuc counterscreen

Protocol can be found in Supplementary Table S4. Briefly, 2 μL/well of media containing secreted NLuc²³ was dispensed into white solid-bottom 1,536-well plates (Greiner, #789173) by using a BioRaptr Workstation. Twenty-five nanoliters of compounds (final concentration range of 1 nM to 62 μM) and controls (neutral control DMSO or positive control Cilnidipine at a final concentration of 125 μM) were subsequently transferred via acoustic dispensing (Echo 550, Labcyte) and plates were incubated for 15 min at RT. Two microliters per well of NanoGlo HiBiT Lytic Detection Regent were dispensed into the plates and after a 5 min incubation at RT, luminescence intensity was measured by using a ViewLux High-throughput CCD imager.

Cellular thermal shift assay

CETSA was performed as described.²⁴ Protocol tables can be found in Supplementary Tables S5 and S6. Briefly, MCF7-ERα-HiBiT cells were harvested and resuspended in CETSA buffer (Dulbecco's phosphate-buffered saline supplemented with 1 g/L glucose and 1X protease inhibitors). For 1,536-well format, cells were dispensed at a density of 5 × 10⁵ cells/mL (5 μL cell/well) into white solid-bottom TC-treated 1,536-well plates (Aurora, cyclic olefin polymer, cat no. EWB041000A) by using a Multidrop Combi dispenser. Compound and control solutions were added via acoustic transfer (Echo 550, Labcyte) in the concentration range of 0.7 nM to 39 μM (11-point compound titration) and tested in triplicate. Neutral control DMSO and positive control E2 (39 μM final) were plated in column 1–4 in each plate. The plates were incubated for 1 h at 37°C and subsequently heated at 45°C by using a custom heating block. Plates were cooled at RT, and 3 μL/well of NanoGlo HiBiT Lytic Detection Regent was added by using a Multidrop Combi dispenser. After a 10 min incubation, luminescence was measured in a ViewLux plate reader as described earlier. For the 384-well format, the assay was performed as described earlier with the following modifications: 15 μL cells/well were dispensed into white 384-well PCR plates (Roche). Plates were sealed, heated for 3.5 min, and cooled to 25°C by using an AB qPCR machine (Roche) with ramp speed of 1.5°C/s for heating phase and max ramp rate for the cooling phase. For luminescence detection, 10 μL of NanoGlo HiBiT Lytic Detection Regent was added per well.

Data analysis for quantitative high-throughput screening

Quantitative high-throughput screening (qHTS) data were analyzed by using software developed internally at NCATS. Data from each assay were normalized plate-wise to corresponding intra-plate controls, as previously described.²⁵ Briefly, for loss of signal assays (NLuc, CTG), raw plate reads for each titration point were first normalized relative to the positive control compound (−100%) and DMSO-only wells (0%) as follows: % Activity = [(V_CMP − V_DMSO)/(V_DMSO − V_POS)] × 100, where V_CMP denotes the compound well values, V_POS denotes the median value of the positive control wells, and V_DMSO denotes the median values of the DMSO-only wells. For gain of signal assays (CETSA), raw data were normalized to positive control (100%) and DMSO-only wells (0%) as follows: % Activity = [(V_CMP − VD_MSO)/(V_POS − V_DMSO)] × 100. The same controls were also used for the calculation of the Z′ factor index for each assay following the formula Z′ = 1 − (3 × SD_positive + 3 × SD_neutral)/|Mean_positive − Mean_neutral|, where SD is the standard deviation. Percent activity was derived by using in-house software (http://tripod.nih.gov/curvefit/). Concentration-response curves (CRCs) were classified as previously described ²⁶ and fitted by using the GraphPad Prism^® software (GraphPad, San Diego, CA). The derived compound response parameters (IC₅₀, Hill fit parameters, measures of efficacy, etc.) were stored in a central Oracle Database. Compounds exhibiting high-quality CRCs (class −1 and −2), IC₅₀ <10 μM, and efficacy <−50% in the NLuc readout and inactive (CRC class 4 or positive class 1 and 2) or active (CRC class −1 and −2) with IC_{50 (CTG/NLuc)} >50-fold or Δefficacy _(CTG-NLuc) >50% in the CTG readout were considered active. Visual inspection of curves was also performed to ensure that compounds meet the cutoffs.

Clustering

Structure similarity clustering of compounds was performed by using the TIBCO Spotfire software with the Lead Discovery plug-in, which performs hierarchical clustering of compounds by using the Tanimoto similarity index over generated compound fingerprints. A Tanimoto similarity of 0.7 was used as a pruning parameter.

Target enrichment analysis

Target enrichment analysis was done as previously described.²⁷ Briefly, we utilized multiple statistical comparisons of the occurrences of a compound's primary target for two sets of compounds (active compounds vs. all compounds tested), calculating adjusted probabilities by using Benjamini-Hochberg correction. Only targets with an adjusted p-value >0.05 in the first set (active compounds) were considered as statistically significantly different from the second set (all compounds tested).

Compounds

The NPC Library contains 2,678 approved chemical entities, as previously described.^21,22

Results and Discussion

Design, characterization, and miniaturization of an ERα post-translational reporter assay

To generate a post-translational reporter assay to monitor ERα protein levels, we fused an NLuc pro-luminescent HiBiT tag to the ERα locus of MCF7 cells, a breast cancer cell line shown to express ERα²⁸ (Fig. 1A). Specifically, we utilized CRISPR/Cas9 to insert the HiBiT tag before the stop codon located in exon 10 of the ESR1 gene (Fig. 1B). Endogenous ERα levels were then detected and quantified by addition of the complementary Large NLuc (LgBiT) enzyme fragment. By limited dilution, we identified two clonal isolates displaying NLuc activity (Fig. 1C). Targeted amplicon sequencing confirmed the insertion in the ESR1 locus (Supplementary Fig. S1). Western blotting of lysates estimates the expression of an HiBiT-tagged protein close to 60 kDa, which corresponds to the molecular weight of the protein identified by the anti-ERα antibody (Fig. 1D). The NLuc signal also detected smaller bands in lysates from both isolates (Supplementary Fig. S2A). To investigate whether these smaller proteins are the result of off-target integration events or are derived from HiBiT-tagged ERα (such as cleaved or degraded protein), we treated isolate #2 with the SERD Fulvestrant (Supplementary Fig. S2B). Fulvestrant decreased all bands in the NLuc blot, indicating that the smaller bands are derived from HiBiT-tagged ERα. Due to higher NLuc signal, we moved forward with clone #2 (referred to as MCF7-ERα-HiBiT). As evidenced by Western blotting, the treatment of MCF7-ERα-HiBiT cells with different concentrations of Fulvestrant led to reduced levels of ERα-HiBiT fusion protein (Fig. 1E). Fulvestrant had a similar effect on untagged ERα protein in parental MCF7 cells (Fig. 1F). This suggests that the HiBiT tag does not interfere with compound-receptor binding.

Fig. 1.

Generation and characterization of an ERα-HiBiT fusion reporter assay in MCF7 breast cancer cells. (A) Schematic representation of the MCF7-ERα-HiBiT assay. Endogenous ERα-HiBiT levels after compound treatment are detected and quantified by the addition of lysis buffer containing the complementary Large NLuc (LgBiT) enzyme fragment and furimazine substrate. (B) Diagram indicates the precise CRISPR/Cas9-mediated insertion of the HiBiT tag at the C-terminus of the ERα-encoding gene (ESR1). The tag was inserted before the stop codon on exon 10 to generate an in-frame ERα-HiBiT fusion protein. Limited dilution was subsequently performed to isolate clonal populations of NLuc-positive cells, which are then expanded for experimentation. (C) NLuc activity (top—image; bottom—quantification) of two MCF7-ERα-HiBiT isolates; n = 3. Wt MCF7 cells are included for comparison. (D) Detection of HiBiT-tagged proteins (NLuc) by Western blot in lysates from wt MCF7 and MCF7-ERα-HiBiT isolates. Anti-ERα antibody was used to identify ERα, and GAPDH was used as a loading control. (E) Pharmacologic characterization of MCF7-ERα-HiBiT assay after 24 h Fulvestrant treatment (1:4 dilution, concentration range 3 μM to 2.8 pM). The ERα-HiBiT levels were detected by Western blotting using NLuc and Anti-ERα antibody. GAPDH was used as a loading control. (F) Pharmacologic characterization of MCF7 wt cells assay after 24 h Fulvestrant treatment (1:3 dilution, concentration range 1.1 μM to 0.5 nM). The ERα was detected by Western blotting using Anti-ERα antibody. GAPDH was used as a loading control. ERα, estrogen receptor α; NLuc, nanoluciferase; wt, wild-type. Color images are available online.

Then, we miniaturized the MCF7-ERα-HiBiT assay to enable high-throughput screening of large compound libraries. Titration-based qHTS enables the identification and classification of compound activity based on potency and efficacy parameters extracted from the resultant CRCs. Hence, qHTS is preferable over screening approaches that test compounds at a single concentration.²⁵ The MCF7-ERα-HiBiT assay was scaled to a volume of 5 μL to effectively use qHTS in 1,536-well format.²⁹ Since the inhibition of ERα signaling leads to a reduction in cell viability,⁷ reduction in NLuc levels can be confounded by either on- or off-target compound-mediated cytotoxicity. To identify compounds that reduce ERα, it is critical to find a treatment window in which the NLuc signal is reduced but cell viability remains unaffected. To this end, we measured the effect of compound treatment on ERα levels and cell viability in parallel by quantifying NLuc activity and ATP levels (proxy for cell number), respectively. The treatment of MCF7-ERα-HiBiT cells with Fulvestrant for 24 h reduced NLuc activity without significant effects on cell viability. However, longer treatment times of 48 and 72 h reduced both NLuc levels and cell viability (Fig. 2A). Based on these findings, we implemented compound treatments of 24 h. The MCF7-ERα-HiBiT assay was further optimized for cell number and addition of LgBiT-containing reagent (i.e., either directly to the media or after media removal; Fig. 2B). We chose a cell number of 1,500 cells/well and removal of culture media before the addition of LgBiT-containing reagent, which not only improves signal intensity but also reduces costs by minimizing the volume of reagent needed per plate.

Fig. 2.

Optimization of the MCF7-ERα-HiBiT assay in 1,536-well format. (A) Cell viability (top) and NLuc activity (bottom) of MCF7-ERα-HiBiT cells treated with a Fulvestrant titration for 24, 48, or 72 h. The CRCs are normalized to DMSO (100%) and 99 μM Digitonin for viability or 49.7 μM Tamoxifen for NLuc (0%); data are represented as mean ± SD, n = 3. (B) NLuc signal intensity of MCF7-ERα-HiBiT cells seeded at the indicated density and assayed by adding the NanoGlo HiBiT-Lytic reagent straight to the well (no media removal) or after removing media from the well (media removal). Data are represented as mean ± SD, n = 32. (C) Pharmacological characterization of the MCF7-ERα-HiBiT assay against ligand (E2), DHP, and Tamoxifen. The CRCs are normalized as in (A); data are represented as mean ± SD n = 3. CRCs, concentration-response curves; DHP, 5α-dihydroprogesterone; DMSO, dimethyl sulfoxide; SD, standard deviation.

Utilizing the conditions described earlier, we characterized the pharmacological response to three SERDs (Fulvestrant, Brilanestrant, and AZD-9495), three SERMs (Tamoxifen, Lasofoxifene, and Raloxifene), and two control compounds, the ERα natural ligand E2 and the androgen receptor ligand 5α-dihydroprogesterone (DHP) (Fig. 2C and Supplementary Fig. S3). Although E2 potently reduced ERα at all concentrations tested (IC₅₀ <0.8 nM), DHP was far less potent (IC₅₀ = 1.7 μM). As expected, although all SERDs potently reduced NLuc activity, SERMs had no effect on ERα levels. All SERMs also showed cytotoxicity at high concentrations. These results indicate that the MCF7-ERα-HiBiT assay recapitulates pharmacological-induced ERα degradation.

qHTS strategy to identify compounds that reduce ERα levels

Next, we interrogated the MCF7-ERα-HiBiT assay against the NPC library of 2,678 approved agents formatted for qHTS (seven concentrations), for which assay performance statistics are found in Figure 3A. Applying the criteria cutoffs as detailed in Materials and Methods, we identified 124 active compounds that reduced NLuc without affecting cell viability. To eliminate potential NLuc artifacts, compounds that directly inhibit NLuc reporter activity, we performed a counterscreen by using an NLuc enzymatic assay. One NLuc artifact (Erythrosin B) was identified and removed from the active set (Supplementary Fig. S4). Of note, the anthracyclines Epirubicin, Idarubicin, Nemorubicin, and Daunorubicin displayed weak inhibitory activity (IC₅₀ >20 μM) in the NLuc enzymatic assay. However, they were still considered as hits since these compounds reduced ERα-HiBiT levels with IC₅₀ <1 μM. Follow-up confirmatory screens, in which fresh compound solutions were re-tested at a higher resolution range (11 concentrations), validated a total of 73 compounds (Supplementary Table S1).

Fig. 3.

qHTS identifies approved drugs that reduce ERα-HiBiT levels. (A) Assay performance statistics in primary (n = 15) and follow-up validation (n = 3) screens for both readouts (ERα-HiBiT via NLuc and cell viability by CTG). S/B: signal/background. (B) Target-based analysis of agents that reduce ERα-HiBiT levels. Y-axis represents the Mechanism of Action (MOA), and x-axis indicates the number of compounds in the MOA category. (C) Example of compounds that reduced ERα-HiBiT levels. The CRCs are normalized to DMSO (100%) and 99 μM Digitonin for viability or 49.7 μM Tamoxifen for NLuc (0%). Data are represented as mean ± SD, n = 3. (D) Structure similarity clustering based on the Tanimoto coefficient of 73 hits that reduced ERα-HiBiT levels in NLuc readout without decreasing cell viability (CTG readout). Compounds belonging to one of the five identified chemotypes are highlighted with a blue triangle (Tanimoto coefficient of 0.7 or greater). Structures of representative compounds belonging to each of the five chemotypes are shown: Rubitecan (1), Danourubicin (2), Estriol (3), Digoxin (4), Nandrolone phenpropionate (5), and Norethynodrel (6). Compound potency (IC₅₀) in the NLuc and CTG readout of the ERα-HiBiT assay is indicated as a blue gradient; inactive compounds are indicated in grey. CTG, CellTiter-Glo; qHTS, quantitative high-throughput screening. Color images are available online.

We analyzed whether the mechanism of action or target class associated to each compound is enriched in the set of hits compared with the entire library (Fig. 3B). This analysis indicated that the three ER antagonists in the library, Fulvestrant, Epitiostanol and its pro-drug Mepitiostane, were among the set of hits (Fig. 3C and Supplementary Fig. S5). The ER agonists (i.e., Estriol, Fig. 3C) were also enriched among the set of hits. In contrast, none of the nine SERMs present in the library reduced ERα-HiBiT levels. Altogether, these results agree with the pharmacological characterization of the assay shown in Figure 2C and Supplementary Figure S3. Agonists of the Progesterone and Androgen receptors, such as Norethinodrel and Nandrolone, respectively, and steroidal cardiac glycosides, such as Deslanoside, are also enriched in the set of hits (Fig. 3B, C and Supplementary Fig. S5). Given that these compounds are steroidal derivatives, it is likely that they directly bind the ERα and induce receptor downregulation. DNA topoisomerase inhibitors such as Irinotecan (Supplementary Fig. S5) are also enriched among hits. Although it is possible that these inhibitors function by disrupting the transcriptional activity of ERα, which, in turn, regulates its own expression,³⁰ DNA topoisomerase inhibitors have been previously found to reduce the expression of HiBiT-tagged MHC class I and are likely disrupting gene expression in a non-specific manner.³¹ We found that all four proteasome inhibitors in the library also reduced ERα levels (exemplified by Ixazomib in Supplementary Fig. S5). It has been previously shown that ERα protein levels are tightly controlled by the 26S proteasome and proteasome inhibitors, such as Bortezomib effectively reduces ERα levels.^32,33 All three HDAC inhibitors in the library also reduced ERα levels (Fig. 3B and exemplified by Vorinostat in Fig. 3C). In addition, the HSP90 inhibitor Retaspimycin also reduced ERα levels (Fig. 3B, C). Accordingly, mechanisms involving HDAC and HSP90 inhibition have been documented to reduce ERα levels.³⁴ It has been recently shown that SERMs, in addition to binding ERα, also bind and disrupt tubulin polymerization, suggesting that SERMs can modulate microtubule assembly and inhibit cell proliferation independently from estrogens.³⁵ Although not enriched, we found that compounds that disrupt tubulin polymerization, such as Eribulin mesylate, also reduce ERα levels (Supplementary Fig. S5), further strengthening a connection between microtubule assembly and ERα.

A hierarchical clustering of hits based on compound structure indicated that most hits contained a steroidal scaffold. A pruning cutoff of 0.7 Tanimoto revealed the presence of six major clusters composed of two or more compounds (Fig. 3D). Clusters 1 and 2 contain DNA Topoisomerase inhibitors, specifically camptothecin derivatives and anthracyclines, respectively. Clusters 3 to 6 contain steroidal derivatives. These clusters comprise ER agonists, cardiac glycosides, Androgen receptor agonists, and Progesterone receptor agonists, respectively. In addition, smaller subclasses of compounds containing terpenoids and nucleoside analogs were also found. The remaining compounds were not classified into common clusters.

Development of CETSA to monitor compound-ERα engagement

To investigate whether the hits described earlier are acting by directly engaging ERα, we established a CETSA. CETSA enables the detection of direct drug–target interactions by quantifying the changes in the thermal stability of proteins on ligand binding in cells.³⁶ Based on a high-throughput CETSA platform that utilizes the split NLuc approach,²⁴ we developed a protocol to quantify compound-bound thermally stable ERα-HiBiT (Fig. 4A). MCF7-ERα-HiBiT cells were dispensed into multiwell plates, treated with compounds for 1 h at 37°C, and subsequently heated to induce protein aggregation. Cells were then lysed, and Lg-BiT was added to quantify soluble compound-bound ERα-HiBiT. The thermal melt profile of DMSO-treated cells indicated an ERα-HiBiT melting temperature (Tm) of 43°C (Fig. 4B). ERα-HiBiT stabilization was observed with Fulvestrant treatment, which increased the Tm to 46.6°C. However, overall NLuc signal intensity in Fulvestrant-treated cells was lower compared with DMSO-treated cells, consistent with Fulvestrant inducing degradation of ERα. Based on the melting profile of ERα-HiBiT, 70%–80% of the protein is melted at 45°C so we chose this temperature to carry out high-throughput CETSA (isothermal format). The assay was miniaturized to both 384- and 1,536-well plates (Supplementary Fig. 6A). We first characterized the pharmacological response to three SERDs (Fulvestrant, Brilanestrant, and AZD-9495), three SERMs (Tamoxifen, Lasofoxifene, and Raloxifene), E2, and DHP, plated in 11-point titrations (Fig. 4C and Supplementary Fig. S6A, B). All SERMs and E2 stabilized ERα-HiBiT, whereas DHP had no effect, which is consistent with receptor-binding patterns. Although Fulvestrant showed ∼50% stabilization, the other two SERDs tested showed no apparent stabilization at the concentrations tested, which is consistent with SERDs inducing receptor degradation. We then tested the validated 73 hits that reduced ERα-HiBiT levels and found that 22 compounds were able to thermally stabilize ERα (S/B = 2.3, Z′ = 0.24; compound activity is shown in Supplementary Table S1). Not surprisingly, hits belong to the Estrogen, Androgen, and Progesterone receptor modulator categories. The most potent hits (AC₅₀ <0.5 μM) include ER agonists such as Estradiol and Estriol and the non-steroidal ERα binder Dienestrol. Fulvestrant was also one of the most potent hits in addition to the Progesterone receptor agonist Norethynodrel (Fig. 4D). The only two compounds not annotated as hormone-receptor binders that showed ERα stabilization, although weak, were the terpenoid Cantharidin and the tetracycline antibiotic Minocycline (Supplementary Fig. S6C). To our knowledge, these two compounds have not been shown to interact with ERα. Interestingly, none of the steroidal cardiac glycosides, despite being potent at reducing receptor levels, thermally stabilized the receptor. Overall, no potency correlation was observed between the compound's activity in NLuc versus CETSA (Supplementary Fig. S6D), which is indicative of the different assay readouts, that is, the modulation of ERα levels versus ERα engagement and thermal stabilization.

Fig. 4.

Optimization of a high-throughput CETSA to determine ERα-HiBiT target engagement. (A) Schematic overview of the high-throughput CETSA. (B) Thermal stability of ERα-HiBiT. MCF7-ERα-HiBiT cells were treated with either DMSO or Fulvestrant (10 μM) for 1 h and heated at the indicated temperatures. (C) Dose-dependent stabilization of ERα-HiBiT by and ligand (E2), Tamoxifen, and DHP. The CRCs are normalized to DMSO (0%) and 39 μM E2 (100%); data are represented as mean ± SD, n = 3. (D) CRCs of the most potent hits. Data are normalized as in (C). CETSA, cellular thermal shift assay. Color images are available online.

Conclusion

The assays presented here constitute an important addition to the toolbox of cellular assays that are amenable for high-throughput screening of ERα modulators, especially activities aimed at identifying ER degraders and endocrine disruptors. Traditional transactivation assays report on ERα activity rather than levels by means of ER-binding elements located in the promoter of reporter genes. Hence, these assays report on direct as well as indirect modulation of ERα levels and/or activity. In addition, these reporter genes are randomly inserted in the genome of choice and are potentially subject to further epigenetic interference. In contrast, by fusing a HiBiT tag to the endogenous ERα-coding locus, this strategy allows for direct quantification of true endogenous ERα levels rather than activity. Modulation of endogenous ERα levels can be achieved by direct binding to the receptor as well as by indirect mechanisms. To this end, the same assay serves also as a means to monitor compound-mediated thermal stabilization of the receptor by measuring ERα-HiBiT levels at temperatures in which the receptor is normally denatured. This CETSA constitutes, as of today, to the best of our knowledge, the only non antibody-based cellular assay available to quantify compound-receptor binding in high-throughput. We found that both assays are complementary. For example, Estrogen, Androgen, and Progesterone receptor modulators both reduced and thermally stabilized ERα, whereas other known ERα-binders, such as Tamoxifen, did not reduce receptor levels but induced thermal stabilization. Conversely, steroidal cardiac glycosides reduced ERα without inducing thermal stabilization. Altogether, these assays can be used to shed light on the mechanism by which compounds exert their biological effects on ERα signaling and function.

Abbreviations Used

CETSA

cellular thermal shift assay

CRC

concentration response curve

CTG

CellTiter-Glo

DHP

5α-dihydroprogesterone

DMSO

dimethyl sulfoxide

DPBS

Dulbecco's phosphate-buffered saline

ERα

estrogen receptor α

ERE

estrogen response elements

NLuc

nanoluciferase

NPC

NCATS Pharmacological Collection

qHTS

quantitative high-throughput screening

RT

room temperature

SD

standard deviation

SERD

selective estrogen receptor degrader

SERM

selective estrogen receptor modulator

TBS-T

tris buffered saline-tween

U.S. EPA

United States Environmental Protection Agency

wt

wild-type

Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported by the NIH Intramural Research Program of NCATS.

Supplementary Material

Supplementary Table S1

Supplementary Table S2

Supplementary Table S3

Supplementary Table S4

Supplementary Table S5

Supplementary Table S6

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5

Supplementary Figure S6