A Novel Orally Active Inverse Agonist of Estrogen-related Receptor Gamma (ERRγ), DN200434, A Booster of NIS in Anaplastic Thyroid Cancer

Sodium iodide symporter (NIS) constitutes a transmembrane glycoprotein that induces iodide uptake in thyroid follicular cells (1, 2), consequent to increased thyroid-stimulating hormone stimulation. Ablative radioiodine therapy and nuclear medicine imaging of NIS are widely used in clinical settings for treatment and diagnosis of well differentiated thyroid cancers and their metastasis with minimal adverse effects. However, their practical application to anaplastic thyroid cancer (ATC), a poorly differentiated thyroid cancer exhibiting progressive dedifferentiation, cancer treatment resistance, and aggressive lung, bone, and regional lymph node metastasis, is hampered by insufficient NIS expression and function. Thus, restoring NIS expression and function in ATC and less differentiated thyroid cancers has been attempted through gene delivery or demethylating agent, nuclear receptor agonist, kinase inhibitor, or histone deacetylase inhibitor treatment to render cells responsive to radioiodine (2, 3), albeit without satisfactory therapeutic outcomes.

Estrogen-related receptors (ERRs) including ERRα, ERRβ, and ERRγ comprise constitutively active nuclear receptors belonging to the NR3B nuclear receptor superfamily. ERR isoforms are expressed in the brain, heart, pancreas, and liver. ERRγ, encoded by Esrrg, plays a key role in tissue metabolism. The peripheral circadian clock regulates ERRγ in muscle, white or brown adipose tissue, and liver (4); in turn, ERRγ enhances uncoupling protein 1 expression and fatty acid oxidation in brown adipose tissues (5). ERRγ is an essential factor for functional glucose-responsive β-cell maturation (6) but is also closely associated with cancer progression including breast and prostate cancers (7, 8). ERRγ is reported as a promising biomarker for evaluating breast cancer prognosis (7). Conversely, ERRγ leads to both androgen-sensitive and -insensitive prostate cancer cell suppression. Thus, ERRγ may be an attractive therapeutic target and biomarker for metabolic disorders and cancer.

We recently reported that the selective ERRγ inverse agonist GSK5182 enhances NIS-mediated radioiodine uptake in ATC cells with either KRAS or BRAF mutations, promoting enhanced radioiodine therapy responsiveness in vitro (9). Toward discovering novel ERRγ inverse agonists, we also reported several lead optimization studies including 4-hydroxytamoxifen analogue synthesis and biological evaluation (10, 11). This finding provided a rationale for exploring new ERRγ inverse agonists that effectively enhance NIS function in vivo and show potential for clinical translation to patients with ATC. Herein, we report the discovery of DN200434, a biocompatible, highly selective, and orally bioavailable ERRγ inverse agonist that acts as a NIS booster, as well as its applicability for ATC treatment in vitro and in vivo.

Metabolic stability assays were performed via incubation of human and selected animal liver microsomes (most often dog, rat, and mouse) at 37°C with a test compound at a final concentration of 1 μmol/L in the presence of 0.5 mg/mL microsomal protein and NADPH regeneration system in a total volume of 100 μL of 100 mmol/L phosphate buffer (pH 7.4). The incubation was started by the addition of the NADPH regeneration system and terminated with 40-μL ice-cold acetonitrile after 0 and 30 minutes. Precipitated proteins were removed by centrifugation at 10,000 × g for 5 minutes at 4°C. Aliquots of the supernatant were injected into a LC/MS-MS (Thermo Fisher Scientific) system. Incubations terminated prior to the addition of NADPH regeneration system (time point, 0 minute) were used as standards and defined as 100%. Percent of the parent compound remaining was calculated by comparing peak areas.

All incubations were performed in duplicate and the mean values were used for analysis. Phenacetin O-deethylase, tolbutamide 4-hydroxylase, S-mephenytoin 4-hydroxylase, dextromethorphan O-demethylase, and midazolam 1′-hydroxylase activity assays were determined as probe activities for CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A, respectively, using cocktail incubation and tandem mass spectrometry. Briefly, the incubation reaction was performed with 0.25 mg/mL human liver microsomes in a final incubation volume of 100 μL. The incubation medium contained 100 mmol/L phosphate buffer (pH 7.4) with probe substrates. The incubation mixture containing various inhibitors (10 μmol/L) was preincubated for 5 minutes, after which an NADPH regenerating system was added. After incubation at 37°C for 15 minutes, the reaction was stopped by placing the incubation tubes on ice and adding 40-μL ice-cold acetonitrile. The incubation mixtures were then centrifuged at 10,000 × g for 5 minutes at 4°C. Aliquots of the supernatant were injected into an LC/MS-MS system. CYP-mediated activity in the presence of inhibitors was expressed as percentages of the corresponding control values.

Parallel artificial membrane permeability assay was conducted with lipid trilayer polyvinylidene difluoride membranes according to the manufacturer's instructions (Gentest; Corning Inc.). Briefly, 300 μL of test compound diluted to 10 μmol/L in PBS (pH7.4) was added to the bottom wells of a 96-transwell plate while 200-μL PBS was added to the top wells. After incubation for 5 hours at 25°C, 20-μL aliquots from each well (bottom and top wells) were transferred into new tubes and mixed with 80-μL acetonitrile (Sigma-Aldrich) containing 4 μmol/L chlorpropamide (Sigma-Aldrich) as internal standard. Then, test compound concentrations were measured by LC/MS-MS (Thermo Fisher Scientific) and permeability rates were calculated using the equations reported previously.

Human ether-à-go-go–related gene (hERG) channel–binding assays were performed using the predictor hERG fluorescence polarization assay (catalog no. PV5365; Invitrogen) according to the manufacturer's instructions. For measuring IC₅₀, compounds were serially diluted (16 points, threefold), followed by reactions for 4 hours at 25°C in a reaction mixture containing hERG membrane, fluorescence tracer red dye, and fluorescence polarization buffer. Fluorescence intensity (excitation at 530 nm, emission at 590 nm) was measured using the multi-mode microplate reader Synergy Neo (BioTek). E-4031 was used as the reference positive standard (IC₅₀ = 10–90 nmol/L).

The Ames MPF 98/100 Mutagenesis Assay Kit (Xenometrix) was used to test mutagenic activity and contains the following components, 2-nitrofluorene, 4-nitroquinolone-N-oxide, and 2-aminoanthracene as positive control, aroclor 1254–induced lyophilized rat liver S9 fraction as the exogenous metabolic activation, ampicillin (50 mg/mL), Salmonella growth medium, exposure medium, and indicator medium. The TA98/TA100 Salmonella strain was used to inoculate the bacterial culture medium and cultured overnight. The Salmonella test strain was then exposed to six concentrations of the test compound (including positive and negative controls). Compounds were prepared in 96-well plates in sterile DMSO (Sigma-Aldrich). Cultured bacteria, test compound, and S9 mixture are incubated for 90 minutes at 37°C in a shaking incubator. Next, the mixture solution was added with indicator medium and dispensed into 384-well plates. The plates were then incubated at 37°C for 48 hours, after which the number of revertant wells in each section was counted.

Sprague–Dawley rats were purchased from Korean Animal Technology (Koatech) and maintained in a specific pathogen-free facility (Laboratory Animal Center, Daegu-Gyeongbuk Medical Innovation Foundation, Daegu, South Korea) with unlimited access to water and food. Rats weighing 250–300 g and 7 weeks old were fasted for 16 hours and subsequently used for experiments. Before compound administration, blood was collected from the jugular vein and used as blank control. For oral administration, four rats received the compound suspended in 10% DMSO, 15% water, and 75% PEG400 at a dose of 10 mg/kg via oral gavage. For intravenous administration, the compound was injected at a dose of 1 mg/kg via the caudal vein. Dosing volume of the compound was 600 μL for oral and 200 μL for intravenous administration. Blood from the jugular vein was collected into heparinized tubes at 0.08, 0.25, 0.5, 1, 2, 4, 6, and 8 hours after compound administration. Plasma was prepared from blood samples by centrifugation at 12,000 rpm for 15 minutes, after which 20-μL plasma was mixed with 80-μL acetonitrile (Sigma-Aldrich) containing internal standard and centrifuged at 14,000 rpm for 5 minutes. Collected plasma supernatants were loaded into triple quadrupole LC/MS-MS (Triple Quad 5500; Applied Biosystems) to measure the compound concentrations. The standard curve range was 5–1,000 ng/mL and the lower limit of measurement quantification was 5 ng/mL. Pharmacokinetic parameters were analyzed with noncompartmental analysis using Phoenix WinNonlin ver. 6.4 (Pharsight).

The gene fragment of human ERRγ ligand binding domain (LBD; 222–458) was designed and codon optimized, after which it was ordered from Integrated DNA Technologies. The DNA fragment was cloned into PGEX-6P1 vector as a GST-tagged protein and expressed in Escherichia coli BL21(DE3) cells. The overnight culture from a single-plate colony was used to inoculate LB media containing 100 μg/mL ampicillin, after which the cells were grown for 1–2 hours to reach an A₆₀₀ = 0.7 at 37°C. The temperature was lowered to 25°C and 1 mmol/L isopropyl-1-thio-β-D-galactopyranoside was added to start protein expression. Cells were harvested after 4 hours and stored at −80°C. Cell pellets were lysed by resuspension and sonication in PBS buffer containing 1 tablet per 50-mL buffer of protease inhibitor cocktail (Roche) and 1 mmol/L PMSF (Sigma-Aldrich). The lysate was clarified by centrifugation and the protein solution was filtered and loaded onto a 4-mL Glutathione Sepharose 4B Column (GE Healthcare) with gravity flow. Bound protein was washed with PBS buffer and eluted with 50 mmol/L Tris (pH 8.0) and 10 mmol/L reduced glutathione. The ERRγ-containing fractions were pooled according to SDS-PAGE analysis. Pooled protein solutions were dialyzed against 50 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, and 1 mmol/L DTT at 4°C overnight. PreScission Protease (GE Healthcare) was added to the protein solution for cleavage of the GST tag with rocking incubation at 4 °C for 4 hours, after which the protein solution was loaded onto the 4-mL Glutathione Sepharose 4B column to separate tag-free ERRγ from GST-tagged ERRγ. Flow through containing tag-free ERRγ was pooled and then diluted fivefold with 50 mmol/L Tris (pH 8.0) and 1 mmol/L DTT. Diluted protein was loaded onto a 5-mL HiTrap Q HP (GE Healthcare) and eluted with a 75–600 mmol/L NaCl gradient over 10 column volumes. The ERRγ containing fractions were pooled according to SDS-PAGE analysis. The pooled material was further purified using a Superdex 200 with an elution buffer containing 25 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, and 1 mmol/L DTT. Peak fractions were collected by SDS-PAGE analysis and concentrated to 10 mg/mL. The final material was aliquoted, flash frozen in liquid N₂, and stored at −80°C until further use.

Purified ERRγ in Superdex 200 buffer was complexed with a threefold molar excess of the compound DN200434 for 1 hour on ice, followed by incubation at room temperature for 30 minutes. Any precipitated material was removed by centrifugation prior to initial screening trials. Initial crystallization trials using commercial screens for protein complex were carried out using mosquito liquid handler (TTP LabTech) followed by additive screening. Optimal crystals were grown in a buffer containing 0.2 mol/L NH₄F (pH 6.2), 20% polyethylene glycol 3350, and 0.01 mol/L ethylenediaminetetraacetic acid disodium salt dihydrate at 18°C. For drop size up, the hanging drop vapor diffusion technique was performed, in which drops containing 1-μL protein and 1-μL mother liquor were equilibrated above a 500-μL mother liquor reservoir.

Cocrystals of ERRγ with DN200434 were directly mounted onto cryoloops and quickly frozen in liquid N₂. Glycerol concentration of the well buffer was slowly increased in a stepwise fashion for cryopreservation of the crystals from a concentration of 8%–32%, followed by mounting the crystals onto cryoloops, and quickly freezing in liquid N₂. All data were collected at the 5C beamline of the Pohang Accelerator Laboratory. Data processing and scaling were carried out using the HKL2000.

The ERRγ crystal structure in complex with DN200434 was solved by molecular replacement using CCP4 program suite. Model building and refinement were carried out using COOT⁷ and Refmac5⁸, respectively. Water molecules were added automatically with the ARP/wARP function in Refmac5, then manually examined for reasonable hydrogen bonding possibilities.

(E)-4-(5-hydroxy-1-(4-(4-isopropylpiperazin-1-yl)phenyl)-2-phenylpent-1-en-1-yl)phenol dihydrochloride salt ¹H NMR (CD₃OD, 400MHz) δ 7.18 (m, 5H), 7.03 (d, J = 8.5 Hz, 2H), 6.83–6.77 (m, 4H), 6.70 (d, J = 8.8 Hz, 2H), 3.78 (m, 2H), 3.58–3.53 (m, 2H), 3.43 (t, J = 6.8 Hz, 2H), 3.24 (t, J = 9.7 Hz, 2H), 2.97 (t, J = 11.8 Hz, 2H), 2.53 (m, 2H), 1.55 (m, 2H), 1.41 (d, J = 6.7 Hz, 6H). ¹³C NMR (CD₃OD, 100MHz) δ 156.07, 145.18, 142.60, 139.74, 138.73, 138.59, 134.39, 131.69, 130.26, 129.44, 127.61, 125.81, 116.27, 114.60, 61.62, 58.46, 32.10, 31.64, 15.64. HRMS (ESI⁺) m/z calcd for C₃₀H₃₇N₂O₂ 457.2850; found 457.2849.

For histologic analyses of human thyroid tissues, we constructed tissue microarrays (TMAs) consisting of 38 normal thyroid samples, 96 papillary thyroid carcinoma (PTC), and 26 poorly differentiated or anaplastic thyroid carcinoma (PD/ATC) samples. Representative areas of the cancer lesions were carefully selected on hematoxylin and eosin–stained sections and two tissue cores (2 mm in diameter) were obtained from each paraffin block. IHC staining of the TMAs was performed using anti-human ERR gamma (mouse mAb; PP-H6812-00; dilution 1:100; R&D Systems), NIS (goat polyclonal antibody; sc-48052; dilution 1:100; Santa Cruz Biotechnology), thyrotropin receptor (TSHR goat polyclonal antibody; sc-7816; dilution 1:100; Santa Cruz Biotechnology), thyroglobulin (rabbit polyclonal antibody; A0251; dilution 1:5,000; Dako), and thyroid peroxidase (TPO rabbit polyclonal antibody; NBP1-80670; dilution 1:200; Novus Biologicals). This study was approved by the Institutional Review Board of the Seoul National University Hospital (Seoul, South Korea) and was conducted in accordance with the Declaration of Helsinki (approved ID: H-1107-060-369).

Histologic data were obtained from formalin-fixed, paraffin-embedded tumor samples. IHC staining using a Benchmark XT Slide Stainer (Ventana Medical Systems, Inc.) was performed according to the manufacturer's instructions. Anti-hNIS (Thermo Fisher Scientific) was applied to whole-sectioned tumor slides. The stained sections were then reviewed without any knowledge of the experimental data.

Specific pathogen-free, 6-week-old, female BALB/c nude mice were obtained from SLC, Inc.. All animals were maintained and used in accordance with the Guidelines for the Care and Use of Laboratory Animals of the Institute of Laboratory Animal Center, Daegu-Gyeongbuk Medical Innovation Foundation. The animal studies were conducted after approval by the institutional reviewer board on the Ethics of Animal Experiments of the Daegu-Gyeongbuk Medical Innovation Foundation (approval number: DGMIF-17120802-00).

ATC cells, CAL62 and BHT101, were purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen). CAL62 and BHT101 cells were maintained in DMEM (high glucose) supplemented with 10% FBS (Gibco FBS; Thermo Fisher Scientific) and 1% antibiotic-antimycotic (HyClone; Thermo Fisher Scientific) at 37°C in a 5% CO₂ atmosphere. Mycoplasma testing was regularly performed every month using a BioMycoX Mycoplasma PCR Detection Kit (CellSafe)

pFR-luc (Stratagene) was used as a Gal4-driven luciferase reporter. Expression vectors for the wild-type ERRγ LBD and the mutants D273A, E275A, Y326A, and N346A were constructed in pCMX-Gal4DBD (DNA binding domain). ERRγ mutants were generated using the QuikChange Site-Directed Mutagenesis Kit (Stratagene).

HEK 293T cells were cultured in DMEM supplemented with 10% FBS. Cells were then transiently transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Luciferase activity was measured after treatment with vehicle, GSK5182 (1 μmol/L), and DN200434 for 18 hours and normalized to β-galactosidase activity.

The change in radioiodine avidity was determined as described previously (9).

Cells seeded in 24-well plates were treated with DN200434 for 24 hours at a concentration of 0, 6, and 12 μmol/L. After aspirating the drug-containing medium, cells were washed with 1-mL Hank's Balanced Salt Solution (HBSS) and incubated in 500-μL HBSS containing 0.5% BSA (bHBSS) and 74 kBq of ¹⁸F-FDG per milliliter for 30 minutes at 37°C. The cells were then washed twice with ice-cold bHBSS and lysed in 500-μL 2% SDS. Radioactivity was measured using a gamma counter (Packard Cobra II Gamma Counter; PerkinElmer). Cell radioactivity was normalized using total protein concentrations determined with a Pierce BCA Kit (Thermo Fisher Scientific).

Total RNA was extracted using TRizol Reagent (Invitrogen) and then 2 μg of total RNA was reverse transcribed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). qRT-PCR was carried out with the SYBR Green PCR Master Mix (Applied Biosystems) using a ViiA 7 Real-Time PCR System (Applied Biosystems) with the following primer sets, NIS forward, 5′-CTG CCC CAG ACC AGT ACA TGC C-3′ and reverse, 5′-TGA CGG TGA AGG AGC CCT GAA G-3′; TSHR forward, 5′-ACC CTG ATG CCC TCA AAG AGC-3′ and reverse, 5′-GCT TCA GTG TCA AGG TTT CAT TGC-3′; TPO forward, 5′-CCT CTG CAA AGA TGT GAA CGA-3′ and reverse, 5′-TCC CGG AGT CTA CGC AGG TT-3′; TG forward, 5′-TCT AAC CGA TGC TCA CCT CTT CTG-3′ and reverse, 5′-AGA TGA TGG CAC CTC CTT GAA CC-3′; and acidic ribosomal protein 36B4 forward, 5′-CCA CGC TGC TGA ACA TGC T-3′ and reverse, 5′-TCG AAC ACC TGC TGG ATG AC-3′. Target genes were normalized to the endogenous reference gene 36B4 and relative mRNA expression levels were calculated using the test and control samples.

The change in membrane NIS and other protein levels was evaluated as described previously (9).

To evaluate the effects of the reporter gene on cell proliferation, cell proliferation assays were performed using the Cell Counting Kit (CCK)-8 (Dojindo Laboratories). CAL62/effluc cells were plated at 1 × 10⁴ cells per well in 96-well plates. After 24 hours, DN200434 at various concentrations was used to treat the cells. At 24 hours after treatment, d-luciferin solution was added to each well and then in vitro bioluminescence imaging (BLI) was conducted using IVIS Lumina III.

Washed and fixed cells were stained in the dark with 0.5-mL propidium iodide/RNase staining buffer for 15 minutes at room temperature. DNA content, cell-cycle profiles, and forward scatter were analyzed using a Becton Dickinson LSRFortessa (BD Diagnostics) with emission detection at 488 nm (excitation) and 575 nm (peak emission). Data were analyzed using FlowJo (BD Biosciences).

Cells were washed with PBS containing 1% horse serum and stained with the FITC-Annexin V Apoptosis Detection Kit I (BD Biosciences) at room temperature for 30 minutes in the dark. DNA content, cell-cycle profiles, and forward scatter profiles were determined using the Becton Dickinson LSRFortessa flow cytometer and analyzed with FlowJo software.

Cells were plated on 6-well plates and incubated for 48 hours. After treatment with 12 μmol/L DN200434 for 24 hours, the drug-containing medium was discarded, and cells were washed twice with PBS. The medium was then replaced with DMEM in the presence or absence of 50 μCi ¹³¹I (Korea Institute of Radiological and Medical Sciences, Seoul, Korea) for 6 hours. Cells were washed with cold bHBSS and incubated in complete culture medium for a time duration corresponding to six doublings. Finally, cells were fixed in 4% paraformaldehyde solution and stained with 0.05% crystal violet. Control and treated colonies with more than 50 cells were counted.

Mice were subcutaneously challenged with CAL62/effluc cells (5 × 10⁶ cells/mouse). When tumor formation was detected via inspection and palpation, tumor-bearing mice were orally administered vehicle (polyethylene glycol) or DN200434 (100 and 200 mg/kg) dissolved in solution (10% ethanol, 10% cremophor, and 80% PEG) once a day for 6 days. CAL62/effluc tumor–bearing mice received radioactive ¹²⁵I (10 μCi per mouse for the biodistribution experiment) or ¹²⁴I [50 μCi per mouse for positron emission tomography (PET) imaging] via the tail vein. After 2 hours of circulation time, biodistribution analysis or PET/CT imaging was performed.

For biodistribution analysis, organs including tumors, liver, lung, heart, kidney, intestine, and others were removed, weighed, and tested for radioactivity using a gamma counter. The results were expressed as the percentage of injected dose per gram of tissue (%ID/g).

For PET/CT imaging, 20-min scans were performed using the Triumph II PET-CT System (LabPET8; Gamma Medica-Ideas). All mice were anesthetized using 1%–2% isoflurane gas during imaging. CT scans were performed using an X-ray detector immediately following the acquisition of PET images. PET images were reconstructed by 3D-OSEM iterative image reconstruction and CT images were reconstructed using filtered back-projections. PET images were coregistered with anatomic CT images using 3D image visualization and analysis software VIVID (Gamma Medica-Ideas). To measure uptake for the volumes of interest (VOI), the VOIs from each image were manually segmented from coregistered CT images using VIVID.

Tumor-bearing mice were divided into four groups as follows, group 1: vehicle; group 2: 200 mg/kg DN200434 daily for 6 days; group 3: single dose of 1 mCi ¹³¹I; and group 4: 200 mg/kg DN200434 daily for 6 days followed by a single dose of 1 mCi ¹³¹I. After the final DN200434 treatment, tumor-bearing mice received radioactive ¹³¹I. In vivo BLI was performed regularly at the indicated time intervals to monitor therapy response. Mouse body weight was also monitored at the same time.

For the BHT101 xenograft model, in vivo therapy was conducted as described for the CAL62 xenograft model when the tumor was detectable via inspection and palpation. Tumor size was measured with a caliper at the indicated time points and tumor volume (mm³) was calculated using the following formula, tumor volume (mm³) = d² × D/2, where d and D are the shortest and longest diameter in mm, respectively.

Tumor-bearing mice received d-luciferin via intraperitoneal injection and then BLI was performed 10 minutes after substrate injection using the IVIS Lumina III In Vivo Imaging System (PerkinElmer). All mice were anesthetized using 1%–2% isoflurane gas during imaging. Grayscale photographic images and bioluminescent color images were superimposed using LIVINGIMAGE (version 2.12; PerkinElmer) and IGOR Image Analysis FX Software (WaveMetrics). BLI signals were expressed in units of photons per cm² per second per steradian (p/cm²/s/sr).

All data are expressed as the mean ± SD of at least three representative experiments, and statistical significance was determined using an unpaired Student t test. P < 0.05 was considered statistically significant. To determine the significance of in vivo radioiodine avidity (PET/CT imaging experiment) between the vehicle and DN200434 groups, paired Student t test was adopted. To compare trends of ERRγ, NIS, TSHR, thyroglobulin, and TPO expression in normal thyroid samples with PTC and PD/ATC samples, the linear-by-linear association test was used.

We evaluated ERRγ and NIS protein expression in human ATC carcinomas and normal thyroid tissues via immunoblotting and found that only the former demonstrated strong ERRγ protein expression (Supplementary Fig. S5); conversely, only the normal thyroid tissues exhibited strong NIS expression. Moreover, Western blotting showed that ERRγ expression was detected in two of five normal thyroid tissues. To confirm our results, we attempted to examine as many tumor tissues as possible via immunoblotting; however, it is difficult to obtain a sufficient number of fresh frozen tissues and thus we performed IHC to confirm the differences in ERRγ protein expression levels according to pathologic type using TMA consisting of 38 normal, 96 PTC, and 26 PDC/ATC tissues. As shown in Fig. 1A and B, high ERRγ protein expression in poorly differentiated ATC and samples from patients with thyroid cancer was confirmed by TMA analysis with an ERRγ-specific antibody (Supplementary Table S1). On the other hand, ERRγ expression was downregulated in normal thyroid tissue and PTC. As expected, iodide-handling gene expression, including NIS, TPG, and TSHR, was lower in PD/ATC than in normal tissues or PTC carcinomas. These results indicate an ERRγ association with ATC pathophysiologic characteristics, supporting our hypothesis of ERRγ as a promising ATC therapeutic target in vivo.

We synthesized an approximately 300-compound focused library based on structural motifs of the hit compound GSK5182, an active metabolite of tamoxifen, an FDA-approved drug optimized for ER⁺ breast cancer. Lead optimization of this series toward DN200434 will be published separately. Basic structure–activity relationships were tracked using a combination of biochemical assays with time-resolved fluorescence resonance energy transfer (TR-FRET) for binding potency and subtype selectivity, and cell-based functional assays using cotransfection systems for cellular activity.

We first performed a quick screening based on ERRγ binding potency and then checked binding selectivity over related subtypes. For the selected compounds, inverse agonistic characteristics were validated in cell-based assays including functional IC₅₀. Between-assay values showed relatively good correlations (Fig. 2A). Briefly, analogues distal to the N,N-dimethylaminoethyloxy group were mainly examined, yielding significant binding selectivity and cellular potency, as well as in vitro/in vivo absorption, distribution, metabolism, excretion, and toxicity (ADMET) property improvements. After several iterative design cycles, DN200434 (Fig. 2B; Supplementary Figs. S1–S4) was identified as the best tool compound for subsequent biological evaluations including X-ray crystallography and in vivo therapy. DN200434 exhibited promising ERRγ binding affinity (IC₅₀ = 0.040 μmol/L) and comparable subtype selectivity with no prominent ERRα, ERRβ, or ERα affinity (IC₅₀ > 10, 1.330, and 1.240 μmol/L, respectively) in in vitro binding assays (Fig. 2C–F). DN200434 was the most potent ERRγ inverse agonist with a functional IC₅₀ of 0.006 μmol/L activity, 12-fold more than that of GSK5182 (Fig. 2G).

DN200434 showed comparable in vitro safety profiles with those of the hit compound, including representative 1A2, 2C9, 2C19, 2D6, and 3A4 in CYP subtype inhibition, hERG binding and current inhibition, and mini-Ames test (Supplementary Tables S2 and S3). Additional in vitro ADMET studies, such as parallel artificial membrane permeability assays (Supplementary Table S2), Caco-2 cell permeability, plasma protein binding, plasma stability, and P-glycoprotein inhibition, indicated DN200434 as the new lead compound for treating ERRγ-related disorders (data not shown). Although in vitro microsomal stability data were indicative of a metabolically soft nature (Supplementary Table S2), it showed appropriate in vivo pharmacokinetics profiles for subsequent in vivo efficacy validation, including an extended half-life (3.8 hours) as well as enhanced oral AUC (0.81 μmol/L/hour) and bioavailability (21.4%; Supplementary Table S4). As the oral bioavailability of GSK5182 was only 8.4%, we theorized that the improved bioavailability of DN200434 resulted from a small change in the C-ring attachment, producing one of the key differences in the drug-like properties of the new lead (Supplementary Table S4; ref. 11).

To address ligand binding to the ERRγ LBD (222–458), we solved the ERRγ crystal structure in complex with DN200434 at 2.5 Å resolution by molecular replacement using a search model (PDB ID: 1S9Q). The collected data were refined and converged to final R_work = 19.93% and R_free = 24.79% using Coot (12) and Phenix (13). All figures were generated by Chimera (version 1.11.2, build 41376; ref. 14). Structure analysis showed an ERRγ-DN200434 binding mode similar to that of ERRγ-GSK5182 (Fig. 2H; ref. 15). Compared with GSK5182 (key protein residue distance to ligand, TYR326: 2.56 Å, ASN346: 2.55 Å, and GLU275: 2.6 Å), the DN200434 binding affinity in the ERRγ ligand binding pocket was enhanced owing to closer key residue (Y326, N346, and E275) ligand proximity. A trimer state was observed in the solved structure labeled as A, B, and C chains. In A and B chains, DN200434 alcohol functionality interacts with TYR326 at 2.4 Å and ASN346 at 2.5 Å via hydrogen bonding. Phenol interacts with GLU275 at 2.4 Å and the basic chain interacts with ASP273. HIS434 and PHE435 movement and DN200434 OH tip tilting in a different orientation from that of GSK5182 occurs only for the C chain (Fig. 2I); this induces binding pocket movement near the ligand and tighter ligand packing into the binding pocket, respectively. In addition, PHE435 in ERRγ-DN200434 moves toward the ligand benzene ring at 3.95 Å, compared with that at 5.25 Å in ERRγ-GSK5182, and among the three chains, only one shows π-H bonding between the DN200434 benzene ring and the PHE435 hydrogen. Hence, π-H bonding likely enhances (33% probability) DN200434-binding affinity, yielding higher DN200434 potency than for GSK5182. Consequently, the four-way DN200434 interactions with key ERRγ-binding pocket residues via hydrogen, ion, and π-H bonding enhance DN200434 potency.

According to the crystal structure in complex with DN200434 and ERRγ, E275, Y326, and N346 were the key residues interacting with DN200434. To investigate the role of these residues, we substituted each amino acid with alanine and performed luciferase reporter assays with wild-type ERRγ and its mutants E275A, Y326A, N346A, and additional mutant D273A. As depicted in Fig. 2J, DN200434 exerted an inhibitory effect on E275A, Y326A, and N346A, whereas the inhibitory effect of GSK5182 was dependent on E275A and Y326A. Specifically, ERRγ D273A was not inhibited by GSK5182 and DN200434, indicating that the substitution of D273 with alanine considerably impeded the interaction with GSK5182 and DN200434 compared with that of other substitutions. This finding suggests that the ion interactions of D273 with the basic chain of both GSK5182 and DN200434 are stronger than the hydrogen bonding interactions of E275, Y326, and N346. These results also indicate that π-H bonding between the DN200434 benzene ring and PHE435 hydrogen may play a key role in the interaction, enhancing the inhibitory effect of DN200434.

We examined the effects of DN200434 on the proliferation of CAL62 cells expressing luciferase as a bioluminescent reporter gene. In vitro BLI showed DN200434 dose-dependent growth inhibition (Supplementary Fig. S6). With 11% and 29% relative growth inhibition at 24 hours after treatment with 6 and 12 μmol/L DN200434, respectively; these concentrations were selected for subsequent experiments. As DN200434 led to dose-dependent growth inhibition of CAL62 cells, we examined whether DN200434 can induce apoptosis. Western blot analysis showed that DN200434 increased both expression of cleaved PARP and cleaved caspase-3 after treatment with 6 and 12 μmol/L DN200434 (Supplementary Fig. S7A). We further employed flow cytometry to demonstrate whether DN200434 results in apoptosis. Flow cytometry analysis with Annexin V/PI staining showed that treatment with DN200434 resulted in an increase in the number of cells undergoing apoptosis in a dose-dependent manner (Supplementary Fig. S7B). Consistently, a dose-dependent increase in the sub-G₁ cell population was also uncovered for DN200434-treated CAL62 cells (Supplementary Fig. S7C).

Furthermore, DN200434 significantly increased CAL62 cell radioiodine uptake in a dose-dependent fashion, with a 1.5- and 2.1-fold relative increase at 6 and 12 μmol/L DN200434, respectively, compared with that of vehicle-treated cells (Fig. 3A). Conversely, 6 μmol/L GSK5182 did not effectively increase radioiodine uptake (Fig. 3B), indicating that DN200434 is more potent at conferring ATC cell radioiodine avidity. As dose-dependent analysis did not detect increased radioiodine uptake above 12 μmol/L DN200434, this concentration was used for assessing time-dependent radioiodine uptake. Significantly increased iodide uptake occurred at 3 hours (∼1.4-fold), reaching a peak at 24 hours (∼2.7-fold; Fig. 3C). Specific NIS inhibition by potassium perchlorate (KClO₄) completely inhibited DN200434-stimulated radioiodine uptake to basal levels (Fig. 3D), suggesting that increased radioiodine uptake is associated with DN200434-mediated NIS function modulation. However, vehicle and vehicle + KClO₄–treated cells exhibited similar radioiodine uptake, revealing the lack of NIS-mediated radioiodine uptake in CAL62 cells.

Similar to GSK5182, DN200434 also significantly decreased endogenous CAL62 cell ERRγ expression as evidenced by immunoblotting with an ERRγ-specific antibody (Fig. 3F; Supplementary Fig. S8A). In addition, as GSK5182-induced pERK1/2 activation is directly associated with enhanced NIS function, DN200434 also dose dependently increased pERK1/2 levels in CAL62 cells (Fig. 3F; Supplementary Fig. S8B). Furthermore, the selective MEK inhibitors PD98059 or U0126 inhibited DN200434-stimulated CAL62 cell radioiodine uptake to basal levels, whereas neither inhibitor alone promoted increased iodide uptake (Supplementary Fig. S9A). Accordingly, Western blot analysis also revealed that PD98059 or U0126 inhibited DN200434-upregulated p-ERK1/2 in CAL62 cells (Supplementary Fig. S9B and S9C). Collectively, these results indicate that enhanced NIS function through DN200434-mediated ERRγ protein downregulation as well as MAP kinase activation is directly involved with radioiodine avidity stimulation.

qPCR analysis demonstrated that DN200434 dose dependently increased NIS, TSHR, TPO, and TG mRNA expression levels in CAL62 cells (Fig. 3E). TG, TPO, and TSHR protein expression similarly increased compared with that of vehicle-treated cells (Fig. 3F; Supplementary Fig. S10), although a slight decrease in TG levels occurred under 12 μmol/L DN200434. DN200434-treated CAL62 cells showed total and membranous NIS protein upregulation with fully (∼95 kDa) and partially (∼50 kDa) glycosylated forms (Fig. 3G; Supplementary Fig. S11), revealing a 1.9- and 2.4-fold relative increase in the fully glycosylated membrane NIS form at 6 and 12 μmol/L DN200434, respectively. Conversely, U0126 or PD98059 reversed the increase in membrane NIS protein expression to basal levels (Supplementary Fig. S12), suggesting that MAK kinase is important for enhancing DN200434-mediated NIS expression and functional activity in ATC cells. Thyroid-regulating gene expression kinetics at various time points after DN200434 treatment demonstrated a time-dependent increase in protein expression (Fig. 3H; Supplementary Fig. S13), although expression peaked at different time points.

Among key glucose metabolism proteins, DN200434-treated CAL62 cells showed a dose-dependent decrease in GLUT-1 and GLUT-4 (Fig. 3F; Supplementary Fig. S14). In vitro F-18-FDG uptake assays also showed a dose-dependent decrease in glucose uptake in DN200434-treated CAL62 cells with a 1.5- and 1.7-fold relative decrease at 6 and 12 μmol/L DN200434, respectively (Fig. 3I). Together, these results indicate that DN200434 exhibits a unique ability to redifferentiate ATC by restoring thyroid-regulating genes and downregulating glucose metabolism.

Biodistribution and ¹²⁴I-PET/CT imaging analyses were performed in an ATC tumor mouse model following oral DN200434 administration (Fig. 4A; Supplementary Fig. S15A). DN200434 dose dependently increased CAL62 tumor radioiodine incorporation (Supplementary Fig. S15B). No differences were observed in other organs between vehicle-treated and DN200434-treated groups (Supplementary Fig. S15C).

To noninvasively and quantitatively evaluate DN200434-mediated CAL62 tumor radioiodine avidity changes in vivo, we conducted ¹²⁴I-PET/CT imaging using 200 mg/kg DN200434 as the concentration yielding the highest radioiodine incorporation level from biodistribution studies (Fig. 4A). CAL62 tumors exhibited a distinct increase (relative 4.4-fold) in DN200434-mediated radioiodine incorporation (Fig. 4B–D), whereas CAL62 tumors of vehicle-treated mice showed no increase. Immunoblotting analysis with an NIS-specific antibody revealed higher endogenous NIS protein expression in tumors from DN200434-treated than vehicle-treated mice (Fig. 4E). Similarly, IHC examination with an anti-hNIS antibody demonstrated strong NIS membrane staining in DN200434-treated but not in vehicle-treated CAL62 tumors (Fig. 4F). These results indicate the feasibility of orally bioactive DN200434 as a potent drug for inducing an effective increase in ATC radioiodine avidity by enhancing NIS protein expression. Furthermore, in vitro clonogenic assays indicated that either DN200434 or ¹³¹I alone induced a minimal cytotoxic effect (Supplementary Fig. S16A and S16B). Conversely, marked cytotoxic effects were observed in the combination group.

We next performed in vivo therapy to determine whether DN200434-mediated radioiodine avidity provides enhanced ATC susceptibility to therapeutic radioiodine therapy as described in Fig. 4G. Marked tumor growth was observed in the vehicle-treated group. ¹³¹I-treated mice also exhibited noticeable tumor growth resulting from lack of radioiodine avidity for CAL62 tumors (Fig. 4H). In DN200434-treated mice, slight tumor growth inhibition was apparent. In vivo BLI demonstrated marked tumor growth inhibition in the combination group, unlike single therapy. Furthermore, we investigated the therapeutic outcomes of a combination of DN200434 and radioiodine therapy in another ATC cancer model, BHT101 harboring BRAF^V600E mutation (Supplementary Fig. S17A). Similar with CAL62 tumor–bearing mice, mice receiving a single treatment of either ¹³¹I or DN200434 exhibited fewer therapeutic effects (Supplementary Fig. S17B; gray and blue line). However, combination treatment led to marked tumor growth reduction in the BHT101 xenograft model up to 32 days posttreatment (Supplementary Fig. S17B; gray and red line). Abnormal behavior or body weight loss was not detected over the entire treatment course (Supplementary Fig. S18).

ERRγ regulation represents a promising therapeutic target in various metabolic/cardiac diseases and certain types of cancers (16–19). Recently, we found that ERRγ modulation by its inverse agonist GSK5182 enhances NIS function and radioiodine therapeutic effects in ATC cells in vitro (9). However, GSK5182 was not effective in an in vivo ATC model, possibly owing to poor in vivo pharmacokinetic profiles (e.g., low bioavailability) and biocompatibility. Thus, we searched for novel drugs to selectively and effectively modulate ERRγ, thereby inducing effective radioiodine avidity and improving radioiodine therapeutic outcomes in in vitro and in vivo ATC models.

Herein, we extended the clinical scope of ERRγ modulating activities by identifying a novel ligand, DN200434, that enhances NIS function and is potently effective in ATC radioiodine treatment in vivo. This newly developed ERRγ inverse agonist was highly selective over other subtypes in binding assays and exhibited the most potent functional activity in a cell-based assay system. DN200434 also showed improved in vitro/in vivo ADMET profiles in required standard discovery studies including CYPs and hERG inhibition, mini-Ames test, in vivo PK, and in vivo safety. Structural assessment confirmed specific DN200434 binding to ERRγ, based on four-way interactions of DN200434 with corresponding residues at precise positions within the binding pocket. For clinical application of DN200434 in ATC, identifying the relationship between target protein expression levels (ERRγ) and thyroid cancer staging is essential. Notably, we found a reverse relationship of ERRγ expression between patient-derived normal and ATC tumor tissues, strongly suggesting that ERRγ constitutes a novel biomarker and therapeutic target for thyroid cancer management.

To evaluate the biological potential of DN200434, we determined its effects on radioiodine uptake in CAL62 cells carrying KRAS gene mutations, which have been utilized to examine the therapeutic efficacy of several agents, allowing cellular cytotoxicity induction and iodide-handling gene restoration (20–22). DN200434 induced a dose- and time-dependent increase in radioiodine incorporation in CAL62 cells, which was inhibited to basal levels by KClO₄ as an NIS-specific inhibitor (1, 23), suggesting that the increased radioiodine uptake was linked to DN200434-modulated NIS function. Moreover, DN200434 was more potent at inducing increased radioiodine avidity than did GSK5182. We therefore expected that enhanced radioiodine avidity may also be due to iodide-handling gene restoration, including NIS, TPO, TG, and TSHR. Accordingly, DN200434 drastically enhanced iodide-handling gene expression at both mRNA and protein levels.

Poorly differentiated (or dedifferentiated) thyroid cancers exhibit high F-18-FDG uptake with upregulated glucose transporter expression, whereas primary or metastatic differentiated thyroid cancers demonstrate low F-18-FDG uptake and high radioiodine avidity (24–26), indicating that glucose uptake is inversely proportional to thyroid cancer radioiodine avidity or differentiation level (27). Here, DN200434-treated cells exhibited downregulated Glut-1 and Glut-4 expression levels and decreased F-18-FDG uptake, suggesting that DN200434 exhibits a unique ability to reinduce differentiation of poorly or undifferentiated ATC cells.

For successful radioiodine therapy, effective β-ray–emitting radioiodine incorporation should occur in thyroid cancer. The in vitro finding of DN200434-mediated increase in radioiodine avidity prompted us to further evaluate its effects on in vivo radioiodine uptake restoration in ATC tumors. To confirm the ability of DN200434 to enhance radioiodine avidity in living mice, we used PET/CT imaging, which shows radioiodine incorporation levels in thyroid tumors in a noninvasive and quantitative manner. DN200434-treated tumors exhibited a striking conversion of lesions from negative to positive on ¹²⁴I-PET/CT following DN200434 administration and accompanied by increased NIS protein expression, as assessed by immunoblotting and IHC. Thus, DN200434-mediated restoration of radioiodine avidity may lead to potential therapeutic effects of radioiodine therapy in ATC tumor–bearing mice. Accordingly, the restoration of radioiodine incorporation in DN200434-treated ATC tumors significantly inhibited tumor growth. Thus, ERRγ-modulated iodide-handling gene recovery by DN200434 may afford a reasonable strategy for achieving radioiodine accumulation, allowing for acceptable radioiodine therapeutic outcomes.

According to TMA of 38 normal, 96 PTC, and 26 PDC/ATC tissues, ERRγ was also found to be expressed in normal thyroid tissues but only in up to approximately 10% of cells; such low expression levels may lead to a lack of responsiveness to ERRγ inverse agonists. On the other hand, PD/ATC exhibit high expression levels of ERRγ (up to 40%), which may lead to good responsiveness to ERRγ inverse agonists, thereby decreasing ERRγ functional activity. Although we cannot fully explain the mechanism by which ERRγ appears to drive the dedifferentiated phenotype of ATC, we presume that ERRγ inverse agonists may be effective in ATC partially due to these reasons.

In conclusion, we reported the discovery of DN200434, which to our knowledge is the most cell active and first orally bioavailable inverse agonist of ERRγ, and uncovered its unique cocrystal structure with ERRγ. Notably, we showed that in an ATC tumor model, DN200434 enhanced ATC tumor radioiodine avidity, as determined by ¹²⁴I PET/CT imaging, via iodide-handling gene upregulation. Furthermore, radioiodine avidity enhancement by DN200434 allowed successful radioiodine therapy of conventional radioiodine therapy–refractive ATC tumors. We are currently investigating whether DN200434 can also enhance radioiodine avidity, as well as radioiodine therapeutic effects in a papillary thyroid cancer model (harboring RET/PTC rearrangements and BRAF^V600E mutation), which accounts for more than 90% of all thyroid cancer cases.

No potential conflicts of interest were disclosed.

Conception and design: T.D. Singh, J. Song, J. Kim, J. Chin, H.-S. Choi, S.J. Cho, Y.H. Jeon

Development of methodology: T.D. Singh, J.H. Yu, G.S. Yoon, S.J. Cho, Y.H. Jeon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.D. Singh, J. Song, J. Kim, J. Chin, H.D. Ji, S.B. Lee, H. Yoon, S.K. Kim, H. Hwang, H.W. Lee, J.M. Oh, S.-Y. Na, W.-I. Choi, Y.J. Park, Y.S. Song, Y.A. Kim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.D. Singh, J. Song, H. Yoon, S.K. Kim, S.-W. Lee, H.-S. Choi, S.-Y. Na, W.-I. Choi, Y.J. Park, Y.S. Song, Y.A. Kim, Y.H. Jeon

Writing, review, and/or revision of the manuscript: T.D. Singh, J. Song, H.D. Ji, S.-W. Lee, J. Lee, Y.J. Park, Y.S. Song, S.J. Cho, Y.H. Jeon

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.D. Singh, H.D. Ji, J.-E. Lee, S.-W. Lee, J. Lee, I.-K. Lee, Y.H. Jeon

Study supervision: J. Song, S.-W. Lee, J. Lee, S.J. Cho, Y.H. Jeon

This study was supported by the following grants, HI16C1501, HT16C0001, HT16C0002, 2017R1D1A1B03028340, 20110018305 (National Creative Research Initiatives Grant), 2017R1C1B1005599, 2017M3A9G7073088, NRF-2017R1A2B3006406, and KDDF-201612-16.

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