Identification of indoles as potential endogenous ligands of ERRγ and their modulation on drug binding

Abstract

Estrogen-related receptor γ (ERRγ) is an orphan nuclear receptor in the ERR subfamily that plays a crucial role in regulating energy metabolism. To date, no endogenous ligand has been identified for ERRγ, posing a challenge for developing targeted therapeutics. Here, we identified that indole and skatole produced by the gut microbiota are potential endogenous ligands of ERRγ using biochemical, cellular, structural, and computational approaches. Indole and skatole increased ERRγ thermostability and directly bound to the ligand-binding domain (LBD) with a K_d of approximately 1–2 μM but had no significant effect or weak inhibitory activity on the transcriptional efficiency. However, RNA sequencing revealed that ERRγ could coregulate several lipid metabolism- and immune-related genes with indole, suggesting a role for ERRγ in the indole pathway. Interestingly, indole and skatole differentially attenuated the activities of ERRγ ligands: they both neutralized the agonistic activity of GSK4716, while indole reduced the antagonistic activity of 4-hydroxytamoxifen (4OHT) and GSK5182, and skatole affected the agonistic activity of endocrine disruptor bisphenol A (BPA). We further screened additional indole metabolites and analogs, resolved the complex structures of ERRγ-LBD with these compounds, and conducted molecular dynamics simulations to determine their binding site and elucidate their binding mechanisms. This study identified potential endogenous ligands of ERRγ, suggesting a novel link between the energy metabolism regulation and the indole pathway. Our findings highlight the need to consider endogenous ligands when designing and optimizing ERRγ-targeted drugs.

Introduction

Estrogen-related receptors (ERRs), a family composed of three nuclear receptors (ERRα, ERRβ, and ERRγ), are named for their high structural homology with estrogen receptor α and β (ERα and ERβ) [1]. Structural studies have revealed that ERRs adopt a transcriptionally active conformation in the absence of a ligand (apo-form), which enables their interaction with nuclear receptor coactivators such as PGC1α or corepressors such as receptor-interacting protein 140 (RIP140) [2, 3]. To date, no endogenous ligands of ERRs have been identified.

ERRγ is a key regulator of energy metabolism that influences fatty acid oxidation and glucose homeostasis, with significant implications for metabolic disorder-related health and disease [1]. In particular, ERRγ is critical for maintaining normal lipid levels and preventing fatty liver disease. Recent studies have shown that ERRγ modulates triglyceride secretion in the liver through phospholipase A2 G12B (G12B), a protein linked to the development of nonalcoholic fatty liver disease (NAFLD) [4, 5]. In addition, ERRγ is essential for cardiac function, bone development and homeostasis, age-related hearing loss (ARHL), and cancer biology, establishing its potential as a therapeutic target [6–8]. However, this receptor has a small ligand-binding pocket (LBP) that is less than 280 ų in volume [3], presenting a considerable challenge for identifying small molecules capable of modulating its activity.

Over 20 agonists and inverse agonists of ERRγ have been identified, but no clinical studies of their use have been reported. GSK4716 was the first selective agonist of ERRγ [9] that exhibits a binding affinity of 5 μM determined by isothermal titration calorimetry (ITC), and it increased the transcriptional activity by approximately 1.85-fold. However, GSK4716 has not shown sufficient selectivity for the ERR family [10]. Building on the efficacy of GSK4716, we developed and synthesized Compound 2e, which increased ERRγ transcriptional activity fivefold at 2.0 μM and promoted the browning of white adipose tissue, despite its lower binding affinity of 46 μM according to ITC determinations [11]. In addition, the environmental pollutant bisphenol A (BPA), an endocrine disruptor related to breast cancer, has been structurally identified as an agonist of ERRγ, demonstrating agonist activity of up to 125% with an EC₅₀ of 174 nM [12]. The antagonist 4-hydroxytamoxifen (4OHT), a known ERα inhibitor, was found to be a potent ERRγ antagonist with a high binding affinity (K_d of 35 nM) [13] but may cause potential side effects because of the involvement of ERRγ in multiple energy metabolic processes. GSK5182, designed on the basis of the 4OHT structure, specifically inhibits ERRγ transcriptional activity and suppresses liver cancer cell proliferation [14]. GSK5182 inhibits the binding of 4OHT to ERRγ with an IC₅₀ value of 79 nM [15].

Indole and its derivatives are small metabolites produced from the metabolism of tryptophan (Trp) by the gut microbiota and regulate multiple biological functions, including enhancing intestinal barrier function, modulating the immune system, and exerting antioxidant and anti-inflammatory effects [16]. Indoles are absorbed through the intestinal epithelium and are rapidly distributed in high concentrations to various organs. In a mouse oral administration assay, indole was detected at 414 nmol/g in the liver 2 h after ingestion [17]. Upon transport to the liver, indoles have been reported to influence liver metabolism and immune responses [18]. Indoles act through binding to several receptors. Indole has been reported as an agonist of the aryl hydrocarbon receptor (AhR) with a binding affinity of 0.87 μM [19] and is a very weak agonist of the pregnane X receptor (PXR) [20]. Upon binding to AhR, indoles modulate immune regulation and intestinal homeostasis, which is known as indole-AhR pathway [21]. Indoles can activate PXR to induce anti-inflammatory responses and enhance the intestinal barrier integrity. For example, indole-3-propionic acid acts as a PXR ligand, downregulating enterocyte TNF-α and upregulating junctional protein-coding mRNAs to improve intestinal permeability [22].

Through fragment screening, we found that indole and skatole directly bind to ERRγ and characterized their binding properties and cellular activity. We also applied RNA sequencing (RNA-seq) to detect the indole-induced gene transcription changes related to ERRγ. Given that endogenous metabolites can influence drug binding and serve as a baseline for optimizing drug selectivity and safety, we further examined the effects of indole binding on ERRγ and ERRγ-targeted drugs. Additionally, we investigated a range of indole derivatives and analogs, employing structural studies and molecular dynamics (MD) simulations to elucidate the molecular mechanisms underlying their impacts on ERRγ activity. Our findings not only reveal potential endogenous ligands of ERRγ but also have major implications for the development of novel therapeutic strategies targeting this receptor.

Materials and methods

Cloning, protein expression, and purification

The gene encoding the human ERRγ229-458 fragment (ERRγ-LBD, UniProt: P62508) was subcloned into the pET-15b vector, which carries an N-terminal hexahistidine (6xHis) tag sequence and a thrombin cleavage site. The protein was expressed and purified as previously described [11]. In brief description, this construct was transformed into BL21 cells grown in LB medium. Protein expression was induced by the addition of 0.3 mM isopropyl β-D-thiogalactopyranoside at 16 °C for 20 h. The target protein was purified by nickel-nitrilotriacetic acid (Ni-NTA) affinity chromatography, anion-exchange chromatography (Source Q, GE Healthcare), and gel filtration (Superdex^TM 75 Increase 10/300 GL column, GE Healthcare), which was equilibrated with a buffer containing 20 mM Tris–HCl, 150 mM NaCl and 0.5 mM TCEP at pH 7.5.

Thermal shift assay (TSA)

All reactions were set up in a final volume of 10 μL in a 96-well plate, containing 5× SYPRO Orange (Sigma, Germany), and incubated on ice for 30 min. 5000× SYPRO Orange (Sigma, Germany) was diluted using the gel filtration buffer (20 mM Tris–HCl, 100 mM NaCl, pH 7.5). Each reaction included 10 μM protein, 400 μM compounds. The samples were heated from 20 °C to 79.5 °C at a thermal ramping rate of 0.5 °C/s, and fluorescence signals were measured using a CFX96 Connect real-time PCR system (Bio-Rad). Data are presented as the mean ± SD of 3 independent experiments performed in triplicates.

All compounds, indole (J&K Scientific, China), skatole (Shanghai Yuanye, China), indole-3-acetic acid (IAA, Shanghai Yuanye, China), indole-3-carboxylic acid (ICA, Shanghai Yuanye, China), oxindole (J&K Scientific, China), 5-nitroindole (5NI, Bide Pharmatech Ltd., China), 6-nitroindole (6NI, Bide Pharmatech Ltd., China), 3-Br-6-nitroindole (Jiangsu Aikon, China), 3-methyl-6-nitroindole (J&K Scientific, China), 4028691 (691, ChemBridge, USA), 4034496 (496, ChemBridge, USA), GSK4716 (Sigma, Germany), 4OHT (Sigma, Germany), BPA (Bide Pharmatech Ltd., China) and GSK5182 (TargetMol, USA, T8370) were dissolved in DMSO to 20 mM.

Isothermal titration calorimetry (ITC)

Isothermal titration calorimetry experiments were performed using a MicroCal ITC200 (GE Healthcare) at 20 °C. The protein eluted from gel filtration was concentrated using a 10 kDa centrifugal filter (Millipore Amicon, Ireland) to 8 mg· mL⁻¹. The protein and the compound were diluted to 300 μM and 20 μM, respectively, using the gel filtration buffer. All solutions were degassed before titrations were performed.

The sample cell was filled with 300 μL of compound, while the syringe was filled with 60 μL of the protein. The titration consisted of an initial injection of 0.4 μL followed by 13 injections of 3.0 μL, with a spacing of 150 s between injections. Experiments were repeated three times. Data were analyzed using Origin 7.0 software (Origin Lab) with the one-site model. Thermodynamic parameters were calculated using the equation ΔG = ΔH − TΔS = −RTlnK, where ΔG, ΔH, and ΔS represent the changes in free energy, enthalpy, and entropy of binding, respectively.

Luciferase reporter assay

HEK293T cells were cultured in high-glucose DMEM supplemented with 10% fetal bovine serum at 37 °C in 5% CO₂. Full-length human ERRγ was inserted into the modified-pCAG vector encoding a His-tag and a strep-tag in the N-terminus of the proteins (pFLERRg). The 3×ERRE sequence (5’- GCTCAAGGTCACG-3’) was inserted into the pGL3-promoter vector between KpnI and XhoI as an ERRE-driven luciferase reporter. Additionally, the 3×G12B sequence (5’-GTTTCACCTTTGTCCTC-3’) was inserted into the pGL3-promoter vector, using the same restriction enzymes.

HEK293T cells were seeded onto 96-well plates at a density of 10,000 cells per well. At 70% confluence, cells were then transfected with 25 ng protein plasmid or pCAG vector (empty vector, (EV)), 25 ng pGL3 reporter plasmid, and 5 ng Renilla plasmid as an internal control using Lipofectamine 2000 (Thermo Fisher Scientific, USA) and Opti-MEM. After 24 h of transfection, cells were assayed for dual luciferase activities using a Dual-Luciferase Assay System, with results normalized to Renilla luciferase activity in the same well. Dual-Glo luciferase assay kit (Promega, USA) was used to detect the luciferase signal according to the manufacturer’s instructions. For the drug tests, after cells were transfected for 6 h, the test compounds were then added and the cells were cultured for additional 20 h. The final concentration of DMSO of all wells was maintained at 1‰. Data are presented as the mean ± SD of experiments performed at least in triplicates. The experiments were performed at least three times independently. Statistical analyses were performed with GraphPad Prism.

RNA-sequencing (RNA-seq) analysis

HEK293T cells were seeded onto 10-cm culture dishes at a density of 1,500,000 cells. Cells were then transfected with 30 μg pCAG vector as control or 30 μg protein plasmid pFLERRg. 6 h later, 100 μM indole or DMSO was added and the cells were cultured for additional 20 h. Total RNA was isolated manually from cultured cells using TRIzol reagent (Invitrogen). RNA sequencing and data analysis were performed by GENEWIZ (China).

Crystallization and structure determination

For crystallization, ERRγ-LBD in the gel filtration buffer (25 mM Tris–HCl, 150 mM NaCl, 5 mM DTT, 5% 1, 2-Propanediol, pH 8.0) was incubated with RIP140 peptide (NNSLLLHLLKSQTIP) and ligands in a molar ratio of 1:1:1.2. The mixtures were incubated at RT for 1 h and concentrated to 15 mg· mL^-1 before crystallization screening. High-quality diffraction crystals were obtained by optimizing the pH, additives, and concentration of precipitant in the reservoir. Prior to data collection, crystals were soaked in mother liquor with the addition of ligands to a final concentration of 1 mM. Complete diffraction data sets for the crystals were collected at the Shanghai Synchrotron Radiation Facility (SSRF) at 100 K.

Diffraction data obtained from autoPROC were scaled using Aimless from the CCP4 program suite [23–25]. Five percent of the data were randomly selected for the R-free calculation. The initial structure solution was obtained by molecular replacement using PDB 2GP0 as template by MolRep [26]. Then the model was refined by REFMAC5 and Coot [27, 28]. Structure figures were prepared using the program PyMOL [29].

Molecular dynamics simulations

The ERRγ-LBD structures solved by X-ray crystallography in the presence of ligands were used to obtain the starting conformations for the corresponding simulations. The initial structure for the unliganded form was obtained by removing the ligand from the indole-bound ERRγ-LBD structure. The three systems were solvated with ~ 23,000 TIP3P water molecules in a dodecahedron box of 725 nm³. Total charges were neutralized with Na⁺/Cl⁻ ions. The Amber ff14SB force field was used for describing the protein and GAFF for the ligands. All the simulations were performed with the Amber 20 software [30]. The time step was set to 2 fs and LINCS constraints were applied to all bonds. Long-range electrostatic interactions were treated with the particle Mesh Ewald method whereas the Van der Walls interactions were implemented with a cutoff of 10 Å. The velocity-rescaling thermostat and the Parrinello–Rahman barostat were used to maintain constant temperature and pressure at values of 300 K and 1 bar, respectively. The MD system was minimized for 2500 steps using steepest descent, followed by another 2500 steps of conjugate gradient minimization, with other atoms constrained by a harmonic force of 50 kcal· mol⁻¹ ·Å⁻². The harmonic force was then adjusted to 25, 5, and 0 kcal ·mol⁻¹ ·Å⁻² during subsequent minimization. With the protein restrained to its initial coordinates, the system was heated to 310 K over 500 ps in the NVT ensemble, using a 2 fs step size and an Andersen thermostat with a heat bath coupling time constant of 1 ps. After transitioning to the NPT ensemble, positional restraints on the protein were gradually removed over 6 ns, while Langevin dynamics with a collision frequency of 2.0 ps⁻¹ was employed for temperature regulation.

Each production MD was performed for 200 ns with a 2 fs step size, constant pressure periodic boundary conditions, isotropic pressure scaling, and Langevin dynamics. The production simulations were conducted in triplicates. During the simulations, the SHAKE method constrained covalent bonds involving hydrogen atoms, and the Particle Mesh Ewald method was used for long-range electrostatic interactions, with cutoff distances for long-range electrostatic and van der Waals interactions set to 8 Å. MD trajectory analysis with CPPTRAJ (version 18.0) [31] included calculations of Cα atom root-mean-square deviations to assess the stability of the ERRγ model.

Relative binding free energies (RBFE) calculations

The relative binding free energies (RBFE) of compounds (indole, 6-nitroindole, 5-nitroindole, skatole, 4034496, 4028691) were calculated using the RBFE module in XtalPi ID4 software. Initially, we prepared the protein and optimized the hydrogen bonding network. Subsequently, we prepared the ligands using the conformations from the crystal structures or docking results. Indole was used as the reference molecule for atom mapping and pair generation. The calculation method employed was topology_change2. Small molecules were parameterized using gaff-am1bcc. Finally, we submitted the RBFE task to calculate the ΔΔG values [32].

Limited proteolysis

Recombinant purified ERRγ-LBD (68 μg) was initially mixed with 2 μg of trypsin (Sigma, Germany) in gel-filtration buffer. Then the sample was sequentially diluted with a 9-fold volume of protein solution, and the mixtures were incubated at RT for 1 h. Aliquots of 20 μL were mixed with 5 μL of 5× loading buffer containing β-mercaptoethanol. The results were verified via SDS-PAGE gel.

Results

Identification of indoles as ligands that directly bind to ERRγ

In nuclear receptors, the LBP of ERRγ is relatively small. Fragment screening revealed that indole can bind to ERRγ-LBD. Thermal shift assay (TSA) was applied to measure the ERRγ-LBD stability by tracking unfolded protein fractions across temperature variations. The differential melting temperature (ΔTm) for indole was +4.0 °C, indicating ligand-induced stabilization. Using ITC, we confirmed that indole bound directly to the ERRγ-LBD, with a K_d of 2.3 µM (Fig. 1a, b).

Fig. 1. Functional characterization of compounds with indole rings.

a, b Chemical structures of the evaluated compounds and their affinity values, which were measured using isothermal titration calorimetry (ITC). Each ITC experiment was repeated three times. ND, not Detected. c Binding affinities of the compounds to ERRγ assessed using thermal shift assays. The data are shown as the mean ± SD. All the experiments were repeated at least three times.

We then screened additional indole metabolites using TSA (Fig. 1c). The ΔTm values for the selected derivatives were +2.4 °C (3-methylindole, (skatole)), 0 °C (indole-3-acetic acid, (IAA)), 0 °C (indole-3-carboxylic acid, (ICA)), and -0.5 °C (oxindole), indicating that skatole induced stabilization of ERRγ-LBD (Fig. 1c). ITC experiments confirmed the binding affinity of skatole, with a K_d of 2.2 µM for ERRγ-LBD. Most indole derivatives derived from Trp possess a substituent at the 3-position [33]. However, our primary experiments revealed that even small groups, such as formic acid or acetic acid, at the 3-position of the indole ring abolish binding to ERRγ, suggesting that indoles with larger substituents at 3-position are unlikely to interact with ERRγ. Furthermore, oxindole, with a hydroxyl group at the 2-position, failed to bind ERRγ as shown by TSA and ITC (Fig. 1c and S1). Other indole metabolites with larger substituents at the 3-position or those with transient properties were not tested.

On the basis of the above results, we tested additional indole analogs (Fig. 1a). The ΔTm values for these compounds ranged from +3 °C (6-nitroindole) to +0.2 °C (4034496) (Table S1). Notably, both 5-nitroindole (5NI) and 6-nitroindole (6NI) displayed relatively high binding affinities to ERRγ-LBD (Fig. 1a), with 6NI showing the strongest affinity (K_d 0.82 μM). However, when both the 3rd and 6th positions are substituted simultaneously, the thermal stability does not increase further (Fig. 1c).

These findings indicate that indole and skatole bind to ERRγ with high affinity. Given that high concentrations of indoles can be rapidly distributed to organs [17], they may serve as potential endogenous ligands of ERRγ in these organs, such as the liver.

Regulatory activity of indoles on ERRγ

Compounds with potent binding activity were further evaluated using the full-length ERRγ (pFLERRg) in a luciferase reporter assay to determine their regulatory effects on the classical ERR element (ERRE). To accurately measure the luciferase signal, all the compounds were tested at a concentration of 100 μM, which was determined in cell viability tests to not inhibit cell proliferation. As a control, 1 μM 4OHT was used as an antagonist (Fig. 2a, b). The results revealed constitutive ERRγ activity, evidenced by the increased basal luciferase activity compared with that in cells transfected with the empty control vector and the ERRE vector. Among the tested compounds, 6NI exhibited transcriptional inhibition of ERRγ on the ERRE comparable to that of 1 μM 4OHT, reducing ERRγ activity by approximately twofold, whereas skatole caused a modest reduction in activity of approximately 20% (Fig. 2a), and this inhibition was dose-dependent (Fig. 2b). Other compounds, including indole, 5NI, 4034496 (496), and 4028691 (691), did not significantly impact the transcriptional activity of ERRγ (Fig. 2a).

Fig. 2. ERRγ cellular activity with or without compound treatment evaluated via the luciferase reporter-based transcription assay.

a Luciferase activity was measured in HEK293T cells transfected with the ERRE-Luciferase reporter, pFLERRg construct or pCAG empty vector (EV) and then treated with drugs at 100 μM. 1 μM 4OHT was used as an antagonist control. b The dose-dependent effects of indole, skatole and 6NI evaluated in the same system. c The dose-dependent effects of indole and 6NI evaluated in the same system upon displacement of the luciferase reporter sequence from the ERRE to the DR1 element of G12B. Luciferase activity was normalized to Renilla luciferase activity and is expressed as a fold change compared with the control. All the data are expressed as the mean ± SD. Statistical analyses were performed with two-tailed Student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001, and ***P < 0.0001 compared with the DMSO group. All experiments were independently repeated at least three times.

ERRE is a half-site element. We recently reported the structure of the direct repeats one element (DR1) of mouse G12B (G12Bpro) in complex with ERRγ [34]. Given that the element type may influence signal transduction between the DNA-binding domain and LBD, we further examined the effects of indole, 6NI (the strongest inhibitor among our tested compounds), and 4OHT (control) on DR1. The results revealed a profile similar to that observed with ERRE (Fig. 2c), indicating that both indole and 6NI likely have the same effects on genes with promoters containing ERRE or DR1. These findings indicate that indole has a high binding affinity for ERRγ but low modulatory effect on ERRE or DR1, potentially due to its molecular size.

To further evaluate the impact of the indole-induced transcript changes related to ERRγ, we performed RNA-seq analysis on HEK293T cells transfected with the ERRγ/control vector and subsequently treated with or without indole. Among the 133 genes regulated by ERRγ, 11 were significantly changed by the addition of indole, including genes related to lipid metabolism and transfer, such as the upregulation of CYP1A1, PLIN4 and the downregulation of ABCA1 (Table S2). Notably, CYP1A1 is reported as a downstream gene of the indole-AhR pathway [35], suggesting that ERRγ coregulates this gene with AhR. Additionally, compared with cells either treated with indole or overexpressing ERRγ alone, ERRγ overexpression in the presence of indole significantly reduced the transcription of interferon-induced protein with tetratricopeptide repeats (IFIT) genes. This finding suggests that ERRγ and indole mutually suppress the regulation of IFIT genes associated with autoimmune disease [36].

Interferences of indole and skatole on the regulatory activity of drugs on ERRγ

As a gut microbiota metabolite that is present in high concentrations in the body and has a strong affinity for ERRγ, indole was hypothesized to interfere with the binding of ERRγ to drug molecules. To test this hypothesis, luciferase activity was measured in HEK293T cells transfected with the ERRE-luciferase reporter and pFLERRg construct, followed by treatment with drugs in the presence of indole or skatole. The ERRγ agonist and antagonist controls were 10 μM GSK4716 and 1 μM 4OHT, respectively. Both indole and skatole effectively neutralized the activation effects of GSK4716 (Fig. 3a). On the other hand, only indole altered the activity of 4OHT, which increased by approximately 40%. Additionally, the effects of indole and skatole on BPA and the antagonist GSK5182 were tested. Skatole, which has inhibitory activity, was observed to inhibit the BPA-activated transcriptional activity, while the effect of indole was not significant enough to be detected. This is likely because indole does not have appreciable agonistic or inhibitory activity on ERRγ on its own, and BPA is a weak agonist with limited agonistic activity (around 20%), rendering indole’s contribution insignificant in this context (Fig. 3b, c). However, an in vitro ITC assay demonstrated that the binding of BPA to ERRγ was significantly inhibited in the presence of indole (Fig. S2). Like 4OHT, only indole specifically opposed GSK5182 inhibition (Fig. 3b, c). Together, these findings confirm that indole and skatole occupy the ERRγ’s binding pocket, to interfere with the binding of other ligands and attenuate their regulatory effects on ERRγ transcriptional activity (Fig. 3d). Specifically, indole can mitigate antagonist activity better than skatole.

Fig. 3. The interferences of indole and skatole on drug binding as determined with luciferase assays with the ERRE.

a Luciferase activities of the co-treatment of GSK4716 or 4OHT with indole or skatole. b Luciferase activities of the co-treatment of BPA or GSK5182 with indole. c Luciferase activities of the co-treatment of BPA or GSK5182 with skatole. d Model depicting how indole and skatole affect drug activity. “ns” indicates not significant. All the data are expressed as mean ± SD. Statistical analyses were performed with two-tailed Student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001. All experiments were independently repeated at least three times.

Structural mechanism by which indoles recognize and bind to ERRγ

To identify the ligand binding sites of ERRγ and investigate the molecular mechanisms underlying the distinct physiological effects of these structurally similar small molecules, we successfully obtained the structures of ERRγ-LBD complexed with indole, 6NI, 5NI, 496, 691, and the short peptide RIP140. The resolution of these complex structures ranged from 1.5 Å to 1.9 Å (Fig. 4a–c and S3; Table S3). The overall protein conformation remained largely unchanged upon ligand binding compared with that of the unliganded receptor, notably retaining activation helix H12 in its transcriptionally active position (Fig. S4).

Fig. 4. Compounds bind to ERRγ-LBD but have a limited structural impact on its constitutively active conformation.

The interactions of indole (a) and 6NI (b) with LBP residues reveal their different binding modes on ERRγ. Oxygen, and nitrogen atoms are colored red and blue, respectively. Hydrogen bonds are indicated by yellow dashed lines. c 2Fo–Fc maps for indole and 6NI contoured at 1σ are depicted in blue. d The interaction between 6NI and ERRγ were evaluated using LigPlot [37]. e, f Structural superimposition of the LBP of ERRγ complexed with 6NI, ERRα (PDB ID: 3D24), and ERRβ (PDB ID: 6LN4). g Molecular dynamics simulations of compound binding to the ERRγ-LBD. The RMSD of the backbone atom positions were measured for the apo- (black), indole-bound (red) and 6NI-bound (green) ERRγ-LBD. h, i Superposition of multiple PDB files from molecular dynamics trajectories of ERRγ with indole and 6NI.

All of the small molecules, characterized by possessing an indole ring structure, occupied the same binding site of the LBP and adopted a similar orientation, with the five-membered heterocyclic ring directed toward a critical water molecule (W1) and stabilized by forming interactions with residues Arg316 and Glu275 (Fig. 4a, b and S3). The indole derivatives exhibited a binding mode in which two phenolic groups interacted with the hydrophobic residues lining the LBP. In each complex structure, Arg316 and Glu275 formed a hydrogen bond network with key water molecules, stabilizing the LBP of ERRγ. Analysis of these structures revealed that the compounds bind with minimal disruption to the active conformation of the ERRγ LBD. The increased Tm values observed from the TSAs reflect that ERRγ LBD is stabilized upon ligand binding but does not undergo a change in conformation.

Interestingly, both 6NI and 5NI possess large hydrophilic nitrate groups at their tail ends, which are trapped in a hydrophobic environment created by surrounding hydrophobic amino acids (Fig. 4d). The electron density maps indicate that this binding mode likely results in a more rigid conformation than that upon indole binding (Fig. 4c). Among the compounds, 6NI adopts the most favorable orientation, with a distance of 2.9 Å between its N-atom and W1, potentially explaining its highest binding affinity (Fig. 4d). In contrast, the electron density map of indole is slightly diffuse due to the lack of substituent groups to stabilize its conformation. This structure further reveals that in this state, both the water molecule W1 and Glu275 are shifting away from their positions in the 6NI-bound structure (Fig. S5). Based on the refinement results, we ultimately modeled indole in the conformation in which its N-atom faces Leu268 and forms a hydrogen bond with the main-chain oxygen, which provides a more favorable interaction. We also compared the structure of ERRγ-6NI with those of the apo-forms of ERRα-LBD and ERRβ-LBD. The substitution of Ala272 to Phe232 in the LBP of ERRα occupies the indole binding site (Fig. 4e), whereas ERRβ shares high structural identity with the LBP of ERRγ (Fig. 4f), suggesting that ERRβ may be a potential target of indole.

To gain additional insights into the ligand-induced stabilization of the ERRγ-LBD, we performed molecular dynamics (MD) simulations of the ERRγ-LBD with the PGC1α peptide in its unliganded form with indole, skatole, 5NI and 6NI. The root-mean-square deviations (RMSD) analysis shows that the binding of 6NI induces greater instability in the helix H12 of ERRγ (Fig. 4g). As a result, 6NI reduces the ability of ERRγ to associate with coactivators, effectively acting as an inhibitor. In contrast, indole has a minimal effect on the stability of H12 or the coactivator binding interface. Uniform extraction of multiple PDB files of molecular dynamics trajectories revealed that indole exhibits significant positional fluctuations within the binding pocket of ERRγ, as does skatole and 5NI (Fig. 4h and Fig. S6). In contrast, the position of 6NI remained relatively stable (Fig. 4i). This finding suggested that 6NI can consistently influence the conformation of H12, thereby exerting an inhibitory effect. We further calculated the binding energy difference between indole and 6NI, 5NI, 496, and 691 by free energy perturbation using the relative binding free energy [32], and these data also suggested that the 6-nitrate substitution is favorable for binding (Table S4).

Our structural and computational data collectively indicate that ligand binding induces global stabilization of the LBD rather than specifically affecting the dynamics or repositioning of the activation helix H12, as commonly observed in classical nuclear receptors.

Discussion

Most nuclear receptors (NRs), such as estrogen receptor and glucocorticoid receptor, are hormone-sensing transcription factors that translate dietary or endocrine signals into changes in gene expression [38]. Among the 48 human NRs, 24 have identified ligands, while the remaining 24 are classified as “orphans” or “adopted orphans”, indicating that a likely ligand has been proposed [39]. Members of the ERR family are well-known “orphan” NRs with a small LBP. Owing to the small LBP, developing ERR-targeted therapeutics is challenging. Exploring the potential for small endogenous ligands of ERR is therefore of great interest. In this study, we unexpectedly discovered that indole and skatole can directly bind to the ERRγ-LBD at 2 μM, stronger than agonists GSK4716 and Compound 2e. TSAs and limited proteolysis experiment (Fig. S7) demonstrated that indole could stabilize ERRγ. Additionally, we resolved the crystal structure of ERRγ-LBD in complex with indole, confirming its binding to the LBP. Thus, by occupying ERRγ-LBP, indoles may contribute to protein stabilization and increase protein turnover time. Based on our structural alignment analysis, ERRβ could be a potential target for indole compounds. This finding suggests that there may be a molecular basis for interaction between indoles and ERRβ. However, it is noted that ERRβ mRNA is predominantly expressed during embryonic development. Given this expression pattern, the functional impacts of indoles on ERRβ may be significantly different from those of ERRγ [40].

Endogenous ligands do not always show strong activity and binding affinity, such as silent ligands that may function through competing with endogenous agonists or inverse agonists [41]. In our study, indole did not significantly affect ERRγ transcription activity in the luciferase assay, while the RNA-seq results revealed complicated regulatory modes. These findings suggest that ERRγ may be involved in the regulation of indole pathway by competing with other indole-targeting proteins (e.g., AhR) because of its high binding affinity for indole.

Recent research has shown that indole and its derivatives can be rapidly distributed to the liver, plasma, and both large and small intestines and can even cross the blood–brain barrier [17]. This distribution pattern suggests that indoles may interact with target proteins in these organs. In addition to the liver, since ERRγ is highly expressed in the nervous system and regulates the dopaminergic neuronal phenotype [42], the effects of indoles on the brain should be considered. Additionally, recent studies have speculated that the gut microbiota is associated with hearing loss, proposing the concept of a gut–inner ear axis [43, 44]. Given the connection between ERR and hearing loss, the role of indole-metabolites in the development of ERRγ-regulated ARHL deserves further investigation. Together, our findings reveal a potential link between ERRγ and indoles, highlighting a novel connection between dietary components, energy regulation, and related diseases in organs where ERRγ and indoles can coexist.

Endogenous metabolites regulate biological functions by interacting with protein targets, and the efficacy of drugs often hinges on their interactions with these metabolites. This highlights the critical importance of accounting for endogenous ligands during drug design and optimization, especially when their interference with drug binding can impact efficacy [45]. In our study, indoles disrupted the binding and activity of drugs and environmental endocrine disruptors both in vitro and in vivo. Most drugs have significantly higher binding affinity for their primary targets than the corresponding endogenous metabolites. Therefore, the effects of indoles on ERRγ drug binding may need to be considered during drug development. This differential affinity can serve as a baseline in drug design, guiding the optimization of selectivity and safety.

In terms of the design of indole-like drugs, we found that modification at the 3-position of the indole ring should be limited. For example, although skatole can effectively bind to ERRγ, larger substituents, such as those in ICA and IAA, significantly impair binding. However, the presence of hydrophilic groups at the 5- and 6-positions is acceptable. Furthermore, the inability of oxindole to bind ERRγ, together with the observation that Compound 496 (featuring an amino group at the 2-position) has a 20-fold lower binding affinity than indole does, suggests that the 2-position is highly critical for binding. Therefore, the five-membered heterocyclic ring is likely critical for making optimal interactions with LBP.

In conclusion, in our study, we identified indoles as potential endogenous ligands of ERRγ, with the ability to disrupt the binding of drugs or environmental contaminants related to cancer. These findings have significant implications for understanding the role of ERRγ in metabolic regulation and developing innovative therapeutic strategies targeting ERRγ. These results also underscore the need for further research into the roles of indoles on modulating ERRγ function and influencing drug responses.

Author contributions

TTX and NW designed the experiments. YYS, HYZ, RC, NW, PD, BLW, YD, and MZS performed the experiments. YYS, HYZ, TTX, JSL, XSW, YX, and YZ performed the analyses. JSL, TTX edited the manuscript. NW and TTX wrote the manuscript.

Funding

This work was supported by Guangdong Basic and Applied Basic Research Foundation (2023A1515030039, 2021A1515220013), Queensland-Chinese Academy of Sciences Collaborative Science Fund (188GJHZ2023069MI), Guangzhou Basic and Applied Basic Research Foundation (2025A04J5436, 202201010711), State Key Laboratory of Respiratory Disease (SKLRD-Z-202519), the Youth Innovation Promotion Association of CAS (2021357), and Guangdong Provincial Key Laboratory of Biocomputing Grant (2016B030301007).

Data availability

Coordinates and structure factors for ERRγ in complex with indole, 6NI, 5NI, 4028691, and 4034496 have been deposited in the RCSB protein data bank with ID 9KND, 9KNC, 9KNE, 9KNF, and 9KNG, respectively. Raw diffraction data are available on the Guangzhou Institutes of Biomedicine and Health Supercomputing Center server.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41401-025-01550-6.