Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Jan 20;8(3):578-592.
doi: 10.1039/c6md00680a. eCollection 2017 Mar 1.

The Molecular Basis of the Interactions Between Synthetic Retinoic Acid Analogues and the Retinoic Acid Receptors

Affiliations
Free PMC article

The Molecular Basis of the Interactions Between Synthetic Retinoic Acid Analogues and the Retinoic Acid Receptors

Hesham Haffez et al. Medchemcomm. .
Free PMC article

Abstract

All-trans-retinoic acid (ATRA) and its synthetic analogues EC23 and EC19 direct cellular differentiation by interacting as ligands for the retinoic acid receptor (RARα, β and γ) family of nuclear receptor proteins. To date, a number of crystal structures of natural and synthetic ligands complexed to their target proteins have been solved, providing molecular level snap-shots of ligand binding. However, a deeper understanding of receptor and ligand flexibility and conformational freedom is required to develop stable and effective ATRA analogues for clinical use. Therefore, we have used molecular modelling techniques to define RAR interactions with ATRA and two synthetic analogues, EC19 and EC23, and compared their predicted biochemical activities to experimental measurements of relative ligand affinity and recruitment of coactivator proteins. A comprehensive molecular docking approach that explored the conformational space of the ligands indicated that ATRA is able to bind the three RAR proteins in a number of conformations with one extended structure being favoured. In contrast the biologically-distinct isomer, 9-cis-retinoic acid (; 9CRA), showed significantly less conformational flexibility in the RAR binding pockets. These findings were used to inform docking studies of the synthetic retinoids EC23 and EC19, and their respective methyl esters. EC23 was found to be an excellent mimic for ATRA, and occupied similar binding modes to ATRA in all three target RAR proteins. In comparison, EC19 exhibited an alternative binding mode which reduces the strength of key polar interactions in RARα/γ but is well-suited to the larger RARβ binding pocket. In contrast, docking of the corresponding esters revealed the loss of key polar interactions which may explain the much reduced biological activity. Our computational results were complemented using an in vitro binding assay based on FRET measurements, which showed that EC23 was a strongly binding, pan-agonist of the RARs, while EC19 exhibited specificity for RARβ, as predicted by the docking studies. These findings can account for the distinct behaviour of EC23 and EC19 in cellular differentiation assays, and additionally, the methods described herein can be further applied to the understanding of the molecular basis for the selectivity of different retinoids to RARα, β and γ.

Figures

Fig. 1
Fig. 1. Left, chemical structures of endogenous retinoids ATRA and 9CRA, synthetic retinoids EC23 and EC19, and synthetic retinoid esters EC23Me and EC19Me., Right, proposed investigation into the interactions of natural and synthetic retinoids by combining in silico docking studies with an in vitro binding assay.
Fig. 2
Fig. 2. (Left) The conformations of 9CRA observed during docking into RARα, RARβ and RARγ, in order of their observation. (Right) Conformation clustering histograms summarising the distribution of the conformations of ; 9CRA that were observed in docking to RARα (blue), RARβ (red) and RARγ (green). 53 potential docking solutions were examined in RARα (11 discarded due to atypical positioning), 78 potential docking solutions were examined in RARβ (five discarded due to atypical positioning), and 73 potential docking solutions were examined in RARγ (none discarded).
Fig. 3
Fig. 3. (Left) The conformations of ATRA observed during docking into RARα, RARβ and RARγ with respect to the alkene stereochemistry, in order of their observation. (Right) Conformation clustering histograms summarising the distribution of the conformations of ATRA that were observed in docking to RARα (blue), RARβ (red) and RARγ (green). 222 potential docking solutions were examined in RARα, 223 potential docking solutions were examined in RARβ, and 185 potential docking solutions were examined in RARγ. None of these solutions were discarded due to unrealistic positioning. The conformation frequency distribution profiles differ significantly between each receptor (Fisher's exact test, P < 0.001).
Fig. 4
Fig. 4. Ligand interaction diagram of the most frequent encountered ATRA conformation, ATRA-conformation 1, in the RARγ (PDB: 2LBD) binding pocket (mesh), highlighting the predominantly hydrophobic nature of the binding of retinoids to RARα, -β and -γ. A cluster of polar residues (K236, R278 and S289 in the case of RARγ) at the base of the pocket anchor the retinoid via the carboxylate. Helix 12 (H12, in yellow) is positioned in the holo conformation, covering the opening to the pocket (hydrogen-atoms were omitted for clarity in this and all other figures).
Fig. 5
Fig. 5. Structural overlay of the binding poses of ATRA in the binding pocket of RARγ (mesh) calculated using X-ray diffraction (PDB: 2LBD) and docking (conformation 1 and conformation 7). The highest scoring examples of conformations 1 and 7, respectively, show the same position of the polyene chain to that in the crystal structure.
Fig. 6
Fig. 6. Highest scoring docking pose of 9CRA docked into the RARα binding pocket. The RARα-specific residue Ser232 is highlighted, demonstrating the close distance between the ring methyl and the serine side chain.
Fig. 7
Fig. 7. Highest scoring docking poses of 9CRA and ATRA overlaid with the binding pose of ; 9CRA in the RARγ binding pocket as determined by X-ray diffraction (PDB: ; 3LBD). The same ; 9CRA conformation (; 9CRA conformation 1) was selected by docking as in the crystal structure, with similar positioning.
Fig. 8
Fig. 8. Structural overlay of the highest scoring docking poses of the most frequent binding conformations of ATRA, and EC23 in the binding pocket of RARγ (PDB: 2LBD) (mesh) calculated by docking simulation. The cyclohexenyl rings of ATRA conformation 1 and ATRA conformation 7 are in positions that are both straddled by the tetrahydronaphthalene hydrophobic region of EC23.
Fig. 9
Fig. 9. Structural overlay of the highest scoring docking poses of EC23 and EC19 in the binding pocket of RARγ (PDB: 2LBD) (mesh) calculated by docking. EC19 exhibits an opposite position of the hydrophobic region due to the need to accommodate the meta-carboxylate.
Fig. 10
Fig. 10. Structural overlay of the highest scoring docking poses of EC23 and EC19 in the binding pocket of RARβ (PDB: 1XAP) (mesh) calculated by docking simulation. The EC19 hydrophobic region is clearly oriented towards the RARβ-specific cleft at the top of the pocket.
Fig. 11
Fig. 11. Comparison between the highest scoring docking poses of EC23 and EC23Me in RARγ (PDB: 2LBD) (mesh). Docking predicts the same positioning for the two ligands, but with EC23Me moved back slightly towards the entrance to the pocket in order to accommodate the methyl ester group.
Fig. 12
Fig. 12. Comparison of the highest scoring docking poses of EC19 and EC19Me in RARα (PBD: 3KMR) (mesh). The ester of EC19Me is forced into an alternative positioning, which sacrifices hydrogen bonds with the polar cluster side chains but significantly enhances a polar contact with the main chain amide. EC19Me is positioned closer to the hydrophobic roof of the pocket.
Fig. 13
Fig. 13. Hydrogen bonding characteristics of each retinoid in RARα (PDB: 3KMR) (mesh) according to the docking study. The majority of the binding pocket is represented in mesh form, the polar cluster residues are shown in a stick representation without unipolar hydrogen atoms, and the ligands are shown in a stick representation without hydrogen atoms. Both ATRA and EC23 are well placed for polar contacts via both carboxylate oxygens. Due to their positioning, only one oxygen of the carboxylates of EC19 and ; 9CRA are available for polar interactions. The esters EC23Me and EC19Me are positioned further from the polar cluster at the bottom pocket, and in the ester group of EC19Me is pushed towards the side of the pocket, enabling a short hydrogen bond interaction with the main chain Phe286-Ser287 amide group.
Fig. 14
Fig. 14. Three-parameter sigmoid curves fitted to binding-assay data for ATRA ([black circle]), 9-CRA (○), EC23 (△) and EC19 ([blacktriangledown]) with RARα (a; blue lines), -β (b; red lines) and -γ (c; green lines). Ordinate axes are FRET ratio values normalised to the lower asymptote for each curve. Binding-assay simulations are shown in (d). The binding assay results show that the EC50 of EC23 for RARα was lower than ATRA, 9-CRA and EC19. For RAR-β, the EC50 of EC23 was slightly lower than ATRA, 9-CRA and EC19. With RAR-γ, the EC50 of EC23 was similar to ATRA, but 9-CRA and EC19 had a lower binding affinity. EC50 values estimated from these binding curves are given in Table 2. In the binding assay simulations with COPASI (d), the red line represents ATRA as the basal condition, the blue line models an increased affinity of co-activator for the ligand–LBD complex, the green line models an increased affinity of ligand for LBD and also co-activator for ligand–LDB, the black line represents an increase in affinity of ligand for the LBD alone, and dashed lines represent higher ligand–LBD affinity with progressively lower affinities of co-activator for ligand–LBD (see ESI for details). In (d), the ordinate units are arbitrary.

Similar articles

See all similar articles

Cited by 5 articles

Feedback