Introduction

Over the last few decades, organisms have become increasingly exposed to environmental pollutants, particularly synthetic chemical substances known as endocrine disrupting chemicals (EDCs). Common EDCs include bisphenol A (BPA), polychlorinated biphenyls (PCBs), phthalates and polybrominated biphenyls (PBBs). These compounds are ubiquitous in the environment and can enter the human and animal bodies through ingestion, dermal absorption, and inhalation, posing risks to human health and animal welfare. In addition to their biological effects, EDCs can disrupt the ecological balance and environmental homeostasis1,2.

Among various EDCs, BPA has gained considerable global attention owing to its widespread industrial use and high production volume. BPA forms a key building block in the manufacturing of polycarbonate plastics and epoxy resins. BPA is commonly used in consumer products such as food packaging, water pipes, toys, and medical devices, leading to widespread human and environmental exposure3,4. In recent years, the production of BPA and its derivatives (analogs) has increased tremendously because of their widespread industrial applications. By 2022, the global demand for BPA reached approximately 10.6 million metric tons5. Given the substantial environmental impact of BPA, an estimated 100 tons is released annually worldwide6, and several structural analogs have been developed under the assumption that they are less harmful. However, many BPA analogs may pose similar risks to both human health and the environment7,8. Consumption of BPA via any route leads to health deterioration via endocrine disruption or deregulation9,10,11. Owing to its structural similarity to certain hormones such as estrogen, BPA can bind to nuclear receptors (NRs) similar to endogenous hormones. This interaction can disrupt the normal functioning of the endocrine system, leading to neurological, reproductive, immune and developmental disorders in both humans and livestock12,13,14.

NRs are a large family of ligand-regulated transcription factors that play essential roles in various physiological processes, including reproduction, inflammation, and metabolism. Based on their ligand binding characteristics, NRs are broadly classified as endocrine and orphan receptors. Endocrine receptors, such as estrogen receptors (ERs), typically exhibit high-affinity binding to endogenous ligands such as steroid hormones. In contrast, orphan NRs either lack identified endogenous ligands or are activated by low-affinity metabolites and xenobiotics. To date, only 48 NRs have been identified in the human genome. BPA and its structural analogs can interact with a range of NRs, functioning as either agonists or antagonists, modulating receptor-mediated gene expression and the associated biological functions7. Although the mode of bisphenol interaction varies depending on the receptor, some compounds act as agonists, while others act as antagonists. Bisphenols bind with high affinity to human constitutive androstane receptor (CAR) as reported by NR binding assays. Moreover, most BPA and its analogs have been reported to act as antagonists or inverse agonists of the CAR7,15.

Encoded by the NR1I3 gene, CAR is a ligand-activated nuclear receptor that is primarily expressed in the liver and small intestine. It plays a critical role in xenobiotic metabolism and detoxification by regulating the expression of genes involved in drug metabolism, such as cytochrome P450 enzymes, for example, CYP2B6 and CYP3A416,17. CAR dysregulation has been implicated in various pathological conditions, including hepatic steatosis, cholestasis, liver fibrosis, and altered drug responses18,19,20. CAR is a 358-amino acid protein organized into distinct functional domains that are characteristic of NRs. Its structure includes an N-terminal DNA-binding domain (DBD), a central ligand binding domain (LBD), and a ligand-dependent C-terminal activation function-2 (AF-2) domain. Previous reports on NRs have shown that ligand-induced conformational changes in the AF-2 (helix) domain are essential for the recruitment of co-regulators and the modulation of downstream transcriptional activity21.

Recent studies have investigated the binding modes and interaction mechanisms of BPA and its analogs with various NRs, including estrogen-related receptor gamma (ERRγ) and the androgen receptor, where BPA often acts as agonist22,23. In contrast, BPA and its analogs exhibit antagonistic or inverse agonistic activity toward CAR. Therefore, in this study, we examined the binding modes of bisphenol compounds with CAR in their inverse agonist-bound structural conformations using molecular docking and dynamic simulations (Fig. 1). We performed binding free energy calculations to understand the binding affinity and per-residue decomposition to identify hotspot residues contributing to the binding of bisphenols with CAR.

Fig. 1
figure 1

Schematic overview of the work flow employed in this study.

Full size image

Materials and methods

Bisphenol A and its analogs extraction

First, we reviewed the literature and considered bisphenol A and its analogs (22 compounds), as described in a recent study22,23. The two dimensional (2D) and three-dimensional (3D) chemical structures of compounds were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/). These chemical structures were imported into the Maestro interface and prepared using the LigPrep tool [LigPrep Schrödinger, LLC, New York, NY, 2017-1]. During this step, 32 possible conformations were generated for each bisphenol. The tautomeric and ionization states of all bisphenol analogs were standardized at physiological pH (pH 7.0 ± 2.0) using the Epik module, and the structures were subsequently energy-minimized with the OPLS-2005 force field before docking.

Inverse agonist bound conformation of CAR-LBD

The crystal structures of the human CAR-LBD are limited to only two structures available in the Protein Data Bank (PDB ID: 1XV9 and 1XVP), and the structures are in an agonist bound conformation. Bisphenol A and its analogs show antagonistic activity against the CAR15. Hence, the mouse CAR-LBD structure (PDB ID: 1XNX), which exists in an inverse agonist bound conformation complexed with androstenol, was used to model human CAR-LBD. The sequence of the CAR-LBD (110-351aa) is obtained from UniProt database (Q14994). The sequence identity between human and mouse CAR-LBDs was 70.66%. A 3D structural model was generated using the SWISS-MODEL web server (https://swissmodel.expasy.org/) along with androstenol. The CAR-LBD with the androstenol complex was energy-minimized using a chimera tool to remove the steric clashes and further removed the androstenol from the CAR-LBD complex for subsequent molecular docking.

Molecular Docking of bisphenols with CAR-LBD

To perform docking simulations, the protein (CAR-LBD structure) was preprocessed using the Protein Preparation Wizard in the Schrödinger suite [Protein Preparation Wizard, Schrödinger, LLC, New York, NY, 2017-1]. The prepared CAR-LBD and bisphenol analog datasets (22 ligands) were imported into the GLIDE module of the Schrödinger suite [GLIDE, Schrödinger, LLC, New York, NY, 2017-1] for molecular docking studies. Extra Precision (GLIDE-XP) docking was used to predict the binding orientation and affinity of the ligands for the protein. This sophisticated method identified potential compounds among bisphenol A and its analogs based on favorable docking scores and predicted molecular interaction profiles. In this study, the binding site residues were defined based on the androstenol (native) ligand present in the CAR-LBD structure. The native ligand was included as a control molecule for comparative analysis. This comparison validated the docking protocol by benchmarking the interaction patterns and binding affinities of the bisphenol dataset against those of a known control molecule.

MM-GBSA binding free energy (BFE) calculation

MM-GBSA calculations, a post-docking binding affinity protocol, were used to evaluate the binding affinities between bisphenols and the CAR-LBD protein in the docked complexes. For this purpose, the prime module within the Schrödinger suite was used by incorporating the VSGB (variable-solvent generalized Born) solvation model and the OPLS3 force field to ensure a reliable representation of molecular interactions. All the complexes were ranked based on their binding affinity scores to identify the most favorable interactions. The BFE (ΔGbind) for the complexes was calculated using the following equation:

$$\Delta {\text{G}}_{{{\text{bind}}}} = \Delta {\text{E}}_{{{\text{MM}}}} \, + \Delta {\text{G}}_{{{\text{solv}}}} \, + \Delta {\text{G}}_{{{\text{SA}}}}$$

Molecular dynamics simulations

To assess the dynamic behavior and stability of the bisphenol-bound CAR-LBD, molecular dynamics (MD) simulations were performed using GROMACS 2022.224,25. Docked complexes were selected based on their docking scores, binding energies, and hydrogen bonds. Nine complexes, including androstenol, were subjected to MD simulations using the CHARMM27 force field26. Ligand topologies compatible with the CHARMM27 force field were generated using the SwissParam web server (https://old.swissparam.ch/)27. Each complex was placed in a dodecahedral box with a minimum distance of 1.2 nm from the box wall. TIP3P water molecules and appropriate counter ions were added to neutralize the system. To mimic physiological conditions, additional buffer (NaCl) of 0.15 molar concentration were included. The LINCS algorithm28 was used to constrain the hydrogen bonds. The temperature (300 K) and pressure (1 atm) couplings were maintained using the V-rescale29 and Parrinello–Rahman algorithms30, respectively. Particle mesh ewald (PME) summation was applied for long-range electrostatics, and a cut off of 1.2 nm was used for short-range and van der Waals (VdW) interactions31. Energy minimization was performed using the steepest descent algorithm for a maximum of 50,000 steps with a maximum tolerance of 1000 kJ/mol. Following minimization, the systems were subjected to two phases of equilibration: NVT for 500ps and an NPT ensemble for 1ns, respectively. Final production simulations with a 2 fs time step integrated by a leapfrog were performed for 300ns, and 2ps coordinate data were saved for trajectory analysis21,32. Trajectory analyses of root mean square deviation (RMSD), root mean square fluctuations (RMSF), radius of gyration (Rg), and hydrogen bonds were performed using the built-in module of GROMACS. Principal component analysis (PCA) was performed for each system using the last 250 ns trajectory with 100ps coordinates. The gmx covar was used to construct and parameterize the covariance matrix, and the gmx aneig was used to analyze eigenvalues and eigenvectors, as reported in previous studies21. Free energy landscapes were determined using the gmx sham module. All the plots were generated using Excel and Mathematica. Molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) calculations were performed for the stable MD trajectory of each system with 100ps coordinates of 50ns (230–280ns) using the g_mmpbsa tool33 to calculate the BFE between the components (protein and ligand). The energy decomposition of each residue in CAR-LBD was also performed.

Results and discussion

Comparison of CAR-LBD agonist and inverse agonist bound conformation

Similar to other NR LBD structures, CAR-LBD contains over all 12 helices and one beta sheet including AF-2 helix at the C-terminal (Fig. 2). Because only the agonist-bound human CAR-LBD structure are available in the PDB (PDB IDs: 1XV9 and 1XVP)34, we modeled the 3D structure of the inverse agonist-bound conformation of human CAR-LBD from the mouse structure (PDB ID: 1XNX), along with the androstenol ligand35. Upon comparison between both the conformations (agonist and inverse agonist) of CAR-LBD, the rmsd of 0.7 Å for 186 Cα atoms suggests structurally both the conformations are similar, except, the kink region at the H10/11 and AF-2 helix mobility in the inverse agonist bound CAR-LBD (Fig. 2). The kinked region is unlikely to occur in other nuclear receptors, such as ERR gamma21. The AF-2 helix at the C-terminus is largely flexible in CAR-LBD compared to other NRs. This suggests that the inverse agonist-bound conformation may not be similar in all NRs, indicating strong functional specificity among NRs35. Interaction analysis suggested that the contacts between H4-H12 and HX-H12 for the agonist and H10-H4 and H11-HX contacts for inverse agonist-bound human CAR-LBD might be crucial (Fig. 2). For NR activation, the orientation of the H12 (AF-2) helix is crucial for coactivator binding. Therefore, the contacts H4-H12 and HX-H12 contacts may be responsible for maintaining the orientation of H12 upon agonist binding. However, the H12 (AF-2) helix for the inverse agonist bound conformation showed high flexibility and lost its interactions, unlike the agonist bound form, moving away from the core structure. These conformational differences at the AF-2 helix upon agonist/inverse agonist binding may lead to recruit either coactivator or corepressor respectively for further CAR activity. Although there are significant similarities between murine and human CAR, ortholog specificity is observed. A previous study showed that helix7 is crucial for strong agonist binding in human CAR compared to murine CAR, in which the H11–H12 and H4–H11 contacts are crucial for agonist and inverse agonist-induced activation, respectively36,37.

Fig. 2
figure 2

Structure of CAR-LBD. The structure of CAR-LBD consists of 12 helices and one beta sheet (left). The superimposition of agonist (pale cyan) and inverse agonist (wheat) bound conformations (right). The kink region (black circle), AF-2 helix (green color), and the ligands (sticks) are shown. The interaction between helices were shown in square boxes for both agonist and inverse agonist conformations.

Full size image

Molecular Docking and molecular interaction profile of BPA and its analogs with CAR-LBD

In this study, molecular docking was used to predict the binding orientation and affinity of BPA and its analogs (n = 22) for human CAR-LBD. The 2D chemical structures of BPA and its analogs are shown in Fig. S1. The inverse agonist conformation of the CAR-LBD structure and bisphenol dataset were prepared using the Protein Preparation Wizard and LigPrep modules within the Schrödinger suite. To validate the docking protocol, the co-crystallized androstenol ligand was redocked into the CAR-LBD binding pocket, and the resulting pose were compared with initial model constructed from the crystal structure (PDB ID: 1XNX) (Fig. S2). The binding energy and molecular interaction profile were also predicted. The bisphenol analogs with favorable docking scores and binding energies, indicating stronger affinities, were prioritized. However, three compounds, BTUM, Bisphenol A bis (diphenyl phosphate), and Phenol, 4,4′- sulfonylbis[2-(2- propenyl), exhibited very weak binding or not fit to bind at the binding pocket of the CAR-LBD domain and were therefore excluded from the study. The binding mode of all the bisphenols were shown in Fig. S3. The obtained docking scores for the bisphenol dataset (n = 19) ranged from − 7.31 to − 10.54 kcal/mol, and the MM-GBSA binding energies ranged from − 41.54 to − 82.72 kcal/mol. The molecular interaction profiles of BPA and its analogs, including favorable docking scores, MM-GBSA binding energies, and key interactions, such as hydrogen bonds, hydrophobic contacts, and other non-covalent interactions, are provided in Table 1. The molecular interaction profiles of BPA and its analogs with CAR-LBD are shown in Figs. 3 and 4.

Table 1 Molecular interaction profiles of selected bisphenols with the CAR protein, including their pubchem IDs, and molecular weights, Docking scores, MMGBSA binding energies, and key hydrogen bond interactions.
Full size table
Fig. 3
figure 3

Molecular interactions between CAR-LBD and bisphenols. Two-dimensional (2D) molecular interaction diagrams generated using GLIDE, depicting key interactions between the CAR-LBD and nine bisphenols along with the androstenol. Various types of molecular interactions were represented using color-coded schemes for clarity.

Full size image
Fig. 4
figure 4

Molecular interactions between CAR-LBD and bisphenols. Two-dimensional (2D) molecular interaction diagrams generated using GLIDE, depicting the key interactions between the CAR-LBD and ten bisphenols. Various types of molecular interactions were represented using color-coded schemes for clarity.

Full size image

BPA and seven bisphenol analogs were selected for further analysis based on their docking scores, MM-GBSA binding energies, and potential molecular interactions within the binding site. Although bisphenol A exhibited a docking score of − 8.60 kcal/mol, it was selected for further analysis because all bisphenol A analogs were derived from this scaffold. The co-crystallized androstenol ligand, used as a control, exhibited a docking score of − 9.49 kcal/mol and an MM-GBSA binding energy of − 73.32 kcal/mol. A single hydrogen bond was observed between the hydroxyl group (− OH) of the androstenol ligand and the aromatic amino acid residue Phe161, with a bond distance of 2.12 Å. Among the selected bisphenol analogs, Bisphenol PH (BPPH) demonstrated the most favorable docking score of − 10.54 kcal/mol and a considerable MM-GBSA binding energy of − 70.72 kcal/mol, comparable to that of the control molecule. The BPPH-protein complex was stabilized by two hydrogen bonds, involving Asn165 (2.87 Å) and Thr225 (2.62 Å). The substantial number of hydrophobic contacts further increases the stability of the complex. The next top-ranked analog, Bisphenol P (BPP), yielded a docking score of − 9.83 kcal/mol and formed a single hydrogen bond with Leu243, with a contact distance of 2.96 Å. The MM-GBSA binding energy for this complex was − 69.58 kcal/mol, which is also consistent with the androstenol ligand binding profile. The Bisphenol AF (BPAF)-bound CAR protein complex attained a comparable docking score of − 9.81 kcal/mol and an MM-GBSA binding energy of − 49.48 kcal/mol. This represented the lowest BFE among the selected bisphenol analogs. Two hydrogen bonds were observed in the complex formed by the polar residues Thr225 and Gln333, with bond distances of 2.71 Å and 2.95 Å, respectively. Bisphenol AP (BPAP) docked with the binding pocket residues of the CAR protein, attained a docking score of − 9.77 kcal/mol, and MM-GBSA binding energy of − 55.9 kcal/mol. A single H-bond interaction was observed with Leu243, with a relatively long contact distance of 3.08 Å in the docked complex. Another bisphenol analog that formed strong and stable interactions with the CAR protein was Bis[2-(4-hydroxyphenylthio) ethoxy] methane, which achieved a docking score of − 9.70 kcal/mol and a high MM-GBSA binding energy of − 82.72 kcal/mol. In this complex, a single hydrogen bond was formed between the polar residue Gln333 and the hydroxyl group (–OH), with a bond distance of 2.92 Å. The three analogs ranked at the bottom are Bisphenol B (BPB), Bisphenol Z (BPZ), and Bisphenol A (BPA). BPB interacted with the CAR-LBD through two conventional hydrogen bonds with Asn165 and Thr225, achieving a docking score of − 9.39 kcal/mol and an MM-GBSA binding energy of − 53.35 kcal/mol. Two conventional hydrogen bonds were observed in this complex, formed by Asn165 and Thr225, with bond distances of 2.90 Å, and 2.65 Å, respectively. Meanwhile, BPZ exhibited a docking score of − 9.31 kcal/mol and MM-GBSA binding energy of − 57.63 kcal/mol. This complex was stabilized by a single hydrogen bond formed between the residue Leu243 and the hydroxyl group of BPZ, with a bond distance of 2.78 Å, indicating a moderately stable interaction. In the case of BPA, the lowest docking score of − 8.60 kcal/mol and MM-GBSA binding energy of − 47.51 kcal/mol were observed among the selected compounds. Two hydrogen bonds formed by the residues Asn165 (2.87 Å) and Thr225 (2.63 Å) were observed in the docked complex. Furthermore, the interaction profiles of the remaining bisphenol analogs are provided in Fig. 4, which depicts the molecular interactions formed by CAR-LBD binding pocket residues with bisphenol analogs. The molecular interaction profile suggests that BPPH may serve as a strong binder, displaying an interaction pattern comparable to that of the androstenol ligand and among the selected bisphenol analog-bound CAR complexes, thereby potentially stabilizing the inverse agonist conformation.

Dynamics of the BPA and its analogs with CAR-LBD

To better understand the stability of the bisphenols at the CAR-LBD binding pocket and the conformational changes of the protein, we performed MD simulations for a 300ns time scale on the selected docked complexes of CAR-LBD-bisphenol analogs. The selection based on favorable docking scores, MM-GBSA binding energies and intermolecular contacts. Nine complexes (androstenol, BPPH, BPP, BPAF, BPAP, Bis[2-(4-hydroxyphenylthio) ethoxy] methane, BPB, BPZ, and BPA bound CAR-LBD) were considered for the MD simulations. To assess the conformational stability of all complexes during the simulations, the MD trajectories were analyzed for RMSD and radius of gyration (Rg) using backbone atoms. In addition, the RMSD of the ligand was assessed. The RMSD of the protein backbone atoms of all nine complexes suggested that equilibrium was reached at least after ~ 60ns. However, the BPB and BPAP-bound complex exhibited larger deviations after ~ 180ns and ~ 220ns respectively. The RMSD density distribution for all the complexes suggests a unimodal distribution and cluster around 0.228–0.432 nm (Fig. 5A), indicating less fluctuations during simulations. The androstenol complex was more stable, with less deviation after ~ 60ns. Meanwhile, all the docked complexes (BPPH, BPP, BPAF, BPAP, Bis[2-(4-hydroxyphenylthio) ethoxy] methane, BPB, BPZ, and BPA) showed larger fluctuations. The overall RMSD of the protein suggests that ligand binding might stabilize the protein, reducing the flexibility of the residues surrounded by the pocket. The ligand RMSD showed that androstenol was more stable in the binding pocket than bisphenol ligands. This may be attributed to the structural conformations of the ligands. In particular, BPAP, BPZ and BPA showed large fluctuations with high RMSD values, indicating conformational differences between these molecules and androstenol as seen in Fig. 5B. The average RMSD of the ligands ranges from 0.03 nm to 0.2 nm, suggesting conformational variations in the ligand molecules in the binding pocket of CAR-LBD.

Fig. 5
figure 5

Trajectory analysis. (A) RMSD of CAR backbone atoms. (B) RMSD of bisphenols. (C) Rg of CAR backbone atoms. The corresponding distribution were shown for RMSD and Rg. The average values and its standard deviations were also provided.

Full size image

To assess the compactness of the protein molecules during simulations, the radius of gyration (Rg) of the protein backbone atoms was measured, as shown in Fig. 5C. The Rg suggests the compactness of the protein molecule was stable with less difference in the fluctuations and clusters around 1.81 nm to 1.86 nm and directly correlated with the RMSD of the complexes. Furthermore, Rg analysis demonstrated that the androstenol bound complex showed higher structural compactness compared to the bisphenol analog-bound complexes.

To examine the residue flexibility of the CAR-LBD, the RMSF of the backbone atoms was analyzed for the last 250ns of the trajectory from each complex (Fig. 6). The observed RMSF suggested that the overall residues of the CAR-LBD showed similar flexibility during the simulations. However, certain regions of the CAR-LBD show differences in flexibility among bound molecules. The major differences in flexibility among the complexes were approximately 132–152aa (Fig. 6), which corresponds to a loop connecting the helical segments (H2 and H3) of CAR-LBD. Some of the other regions, 222–225 aa, 234–237 aa, 303–310 aa, show minimal differences in flexibility, suggesting that these regions might be influenced by the ligand binding. In particular, 222–225 aa region are involved in the binding site of the protein. The AF-2 helix at the C-terminal region is largely flexible in all complexes because of the loss of its helical nature during the simulations. However, structural flexibility and conformation of the AF-2 helix are crucial for the antagonistic activity of NRs. The overall dynamic stability and reduced fluctuations of CAR-LBD in complex with potent bisphenols, such as BPPH and Bis[2-(4-hydroxyphenylthio)ethoxy]methane, indicate ligand-induced conformational restriction, which may hinder the binding of coactivators required for CAR activation.

Fig. 6
figure 6

Residue flexibility during simulations. The RMSF of CAR backbone atoms was analyzed for all bisphenol-bound complexes.

Full size image

Dominant motions and conformational changes of CAR-LBD complexes

To assess the differences in the global motions of the bisphenol-bound CAR-LBD, PCA was performed on the backbone atoms using the last 250 ns of the MD trajectory of each complex. From the PCA, the obtained eigenvalues suggested that the top three PCs largely contributed to the global motions, followed by a reduction in the amplitude and leading to local motions. The percentage of cumulative variance of the top 20 PCs was > 80%, (except for androstenol and BPPH) and among which the percentage of the first three PCs was > 50% (Fig. 7A). Because the top eigenvectors suggested the overall dynamics of the system, we generated porcupine plots for PC1 to observe the conformational changes in the complexes during the simulations. Conformational changes were observed in the AF-2 helical region, as shown in Fig. 7B. Some of the loop regions showed substantial variations, particularly the long loop between helices 2 and 3. However, the observed variations were ligand-dependent. Androstenol and BPPH-bound CAR-LBD showed fewer conformational changes at the AF-2 helix than the other bisphenols. This strongly suggested that each bisphenol has its own specific binding capacity, which induces conformational changes in the AF-2 helix, leading to antagonistic activity. To assess the metastable states of all complexes during the simulations, we performed free energy landscape (FEL) analysis based on top PCs (PC1 and PC2). This suggested that all the complexes were distributed with more metastable states (Fig. 7C). However, one or two large energy basins was observed for the BPP, BPAF, BPAP, Bis[2-(4-hydroxyphenylthio) ethoxy] methane, BPB, BPZ, and BPA-bound complexes indicating less conformational transitions of protein occurred during simulations compared to androstenol and BPPH-bound complexes which showed few small energy basins. These results indicate that bisphenol-induced conformational changes in CAR-LBD may vary depending on the type of bisphenol bound.

Fig. 7
figure 7

Principal components and free energy landscape (FEL) analyses. (A) The first 20 eigenvalues were plotted along with cumulative variance. (B) Conformational changes were visualized using porcupine plots. The large magnitude observed for the AF-2 helix and long loop between helix2 and 3. (C) FELs were plotted for PC1 and PC2.

Full size image

Hydrogen bond analysis

The hydrogen bonds between bisphenols and CAR-LBD during simulations were assessed with a distance ≤ 3.5 Å, and the angle between the donor and acceptor was ≤ 30°. On average, one to two hydrogen bonds were observed (Fig. S4). Further assessment of hydrogen bond occupancy showed that only a few residues in all complexes had more than 10% occupancy. This strongly indicated that hydrophobic residues in the binding pocket of CAR-LBD are largely responsible for the binding of bisphenols (Fig. 8). Phe161, Asn165, His203, Thr225, Ser233, Leu243, His250, Asn327 Gln333, Gln338 are the key residues of CAR-LBD forming hydrogen bonds with occupancy more than 10% during simulations, which is similar to agonist bound form38. The hydrogen bonds were stronger in the bisphenols than in androstenol bound complex, which showed only two hydrogen bonds (>10% occupancy) with Phe161 (occupancy 35%) and Gln338 (occupancy 13%). BPP, BPAP, and BPA exhibited four hydrogen bond interactions during the simulations and were common residues. BPPH, BPAF, and BPZ form three hydrogen bonds, whereas BPB and Bis[2-(4-hydroxyphenylthio) ethoxy] methane forms two hydrogen bonds. Superimposition of the initial and final low energy complexes suggests that bisphenols do not deviate significantly from the binding pocket. However, variations in the bisphenol conformations were observed, indicating that ligand flexibility is possible within the pocket (Fig. S5).

Fig. 8
figure 8

Binding pocket residues of CAR-LBD. Residues located within 4Å of the ligands in the binding pocket of CAR-LBD were shown along with the hydrogen bond occupancy for all simulated complexes. Two bars of His203 (BPP) and Gln333 (BPAF) indicates two hydrogen bonds with ligand.

Full size image

Binding energy calculations and energy decomposition

To understand the binding energy of each bisphenol with CAR-LBD, we performed MM-PBSA calculations using 500 frames (100ps coordinate) of a 50 ns (230–280 ns) MD trajectory. Table 2 shows the contribution of each energy component and the total binding free energies of the bisphenol compounds. The van der Waals energy contributing to the binding energy was higher than the electrostatic energy, suggesting that the binding pocket was surrounded by many hydrophobic residues. This was directly correlated with the interaction pattern between bisphenol and CAR-LBD (Fig. 9). Compared to BPA (− 86.4 kJ/mol), the binding free energy of the BPA analogs was higher. BPPH (− 141.2 kJ/mol), BPP (− 136.8 kJ/mol) and Bis[2-(4-hydroxyphenylthio) ethoxy] methane (131.56 kJ/mol) showed higher binding energy compared to other bisphenol A compounds (BPAF, BPAP, BPB, and BPZ) which are around − 100 kJ/mol. Among the bisphenol A analogs, BPAF showed the lowest binding energy of − 94.13 kJ/mol. The binding energy of bisphenol compounds indicates that strong binding of bisphenols is an inverse agonist of CAR-LBD, leading to antagonistic activity7,15. The per-residue decomposition of CAR-LBD suggests that most residues greater than + 3 kJ/mol and less than − 3 kJ/mol are hydrophobic residues located in the vicinity of the binding pockets of H4, H10/H11, and HX. Bisphenol compounds are mainly stabilized by Phe161, Asn165, Leu206, Phe217, Tyr224, Thr225, L246, I334, and I337 and are common for at least three bisphenols (Fig. 9). Among these residues, Phe161, Asn 165 and Thr225 formed a hydrogen bond interaction with some of the bisphenols during the simulation with greater than 10% occupancy. The higher binding free energies of BPPH, BPP, and Bis[2-(4-hydroxyphenylthio)ethoxy]methane compared to BPA reflect their stronger tendency to stabilize the inactive form of CAR-LBD.

Table 2 MM/PBSA based total binding energy of the bisphenols bound CAR-LBD complexes.
Full size table
Fig. 9
figure 9

Per-residue energy decomposition. The plot shows the energy contribution of each residue. The identified residues with criteria of greater than + 3 kJ/mol and less than − 3 kJ/mol were displayed on the structure using stick models (blue for positive, red for negative residues).

Full size image

Conclusions

Endocrine disruptor chemicals such as bisphenol A and its analogs are ubiquitous in the environment because of their usage in commercial industries. They have immense potential to modulate the activity of NR, such as the CAR, via antagonization, posing health issues in humans and livestock. Our study describes the binding modes and interaction patterns between bisphenols and human CAR using several computational approaches. In this study, we discussed the residues involved in the binding of bisphenols and the conformational changes in CAR-LBD upon bisphenol binding. Although BPA and its analogs showed strong binding affinities, the binding affinity of BPA was lower than that of its analogs, suggesting that BPA analogs also harm organisms and the environment. Notably, BPPH, BPP, and Bis[2-(4-hydroxyphenylthio)ethoxy]methane demonstrated the strongest CAR-LBD inverse agonistic potential, as evidenced by their favorable docking scores, high binding free energies, and stable interaction profiles during MD simulations. We hope that this study will provide a better understanding of the structure–function relationship between bisphenols and CAR, which, in turn, will help to identify better therapeutics.