Leverage Experimental and in Silico 3D Protein Structures to Accelerate SBDD Projects
This blog post will show you how to exploit 3decision structure registration capabilities to easily add protein structures to the database and quickly gain structural insights to enhance your structure-based drug discovery projects.
Introduction
Retinoid X receptors (RXRs) are ligand-activated nuclear receptors. They play a crucial role in gene expression regulation and have emerged as potential drug targets for the treatment of neurodegenerative and inflammatory diseases.
RXR agonists are typically fatty acids that mimic the native ligand of RXRs (9-cis retinoic acid). However, traditional RXR agonists generally exhibit poor safety profiles, pharmacokinetics, and physicochemical properties, requiring improvement to expand their clinical application. Currently, only one synthetic retinoid, bexarotene (compound 1), has been approved for clinical use, and due to its adverse effects, it is only used as a second-line treatment for a few types of cancers. Recent studies have reported a new series of RXR compounds with superior physiochemical properties that were developed using a structure-based approach.¹‚²
In this article, we describe some steps of these studies and reproduce them using Discngine’s software 3decision®, highlighting the key role of structural information in driving the development of novel compounds – in this specific use case - the next generation of RXR targeting compounds.
Structural Analysis of Ligand Binding
Scheme 1. Chemical structures of bexarotene (1) and compound 2 and reported literature values1 of respective EC50 (for RXRα), aqueous solubility, and cytotoxicity.
Compound 2 was selected as the lead for developing the new series of RXR agonists.¹ This ligand is a fatty acid mimetic and a known agonist of other nuclear receptors. The pyrimidine scaffold of 2 imparts favorable physiochemical properties, proving superior to the FDA-approved molecule bexarotene in terms of solubility, lipophilicity, and cytotoxicity.
Docking of 2 in the ligand binding site of RXRα was carried out to assess whether 2 could be accommodated in the target binding site. The scientists conducted their docking studies exploiting the publicly available structural data of RXRα in complex with bexarotene (PDB ID: 4k6i).
In the next sections, we will show how to use the capabilities of 3decision® to compare the docking pose with the experimental structures and gain the structural insights that scientists exploited to optimize the ligand. The docking was performed with the software suite MOE.
If you want to learn how to use 3decision to select the best structure for docking, read our Bonus section at the end of this article.
Docking Pose Deposition in 3decision
The docking pose of 2 in RXRα can be registered in 3decision, allowing easy comparison with other public or proprietary 3D structures in the database and making it available to the whole drug discovery team.
One key strength of the 3decision deposition is that it allows the registration of both in silico predicted and experimental structures. Both types of structures undergo the same processing upon registration in the database (more details on processing below), allowing standardized formatting of all structural data and centralization of relevant knowledge to exploit for advanced structural analysis and ideation.
From the 3decision interface, you can deposit a new structure with a few clicks. The registration process is intuitive and guided, allowing non-expert users to register and visually explore 3D structures quickly.
Video 1. 3decision Structure Registration process. See our How To video.
The method used for producing the 3D structure needs to be specified upon registration. The chosen method does not affect how the structure is processed, and it can be exploited for conducting and refining searches in the 3decision database. In silico structures are labeled as "Model" to distinguish them from experimental structures.
Figure 1. 3decision Structure Registration page, “Edit Metadata, General” section. The method used to produce the 3D protein structure in 3decision can be set as X-ray diffraction, Solution NMR, Electron Microscopy, or Model.
You can also add relationships between the newly deposited structure and structures already in the database. For docking poses, this helps to keep track of the related structures and allows quick comparison between them. In this case, for instance, we set a relationship between the docking pose and the structure of RXRα with bexarotene (PDB ID: 4k6i) that was used for the docking.
Figure 2. 3decision Structure Registration page, “Edit Metadata, Related Structures” section. Relationships between the new structure and other structures already included in the 3decision database can be set from this section. The related structure can be either “Parent” or “Child” of the new structure, and the available “Relation type” options are: Bioassembly, Refined, Derived, or Prepared for Docking. A “Relation description” field is available to provide a supplementary description of the relationship.
-
Protein Chain Sequence Mapping: The sequence is tightly mapped back to the UniProt sequence to allow display to sequence annotations in 3D. -
Pocket Detection: The protein surface is analyzed, and both ligand-defined and putative binding sites are predicted, characterized, and stores in the database (the fpocket algorith is used for this processing). -
Ligand-Protein interaction indexing: The protein-ligand interactions are detected, characterized, and stores in the database (a proprietary algorithm is used for this analysis).
Once the structural information is provided, the registration can be launched. In just a few seconds, the structure will be registered in the 3decision database, and you can immediately start working on it!
Docking Pose Analysis
When you open the newly registered structure in the 3decision workspace, you can immediately visualize it in the 3D Viewer. You can also quickly load the related experimental structure used to prepare the docking pose. Related structures and all other associated data - such as structure parameters, protein chain information, custom annotations, and attached files - can be found in the Information Browser. The experimental structure can be overlayed on the docking pose of 2 in one click, showing a similar binding mode of compound 2 and the co-crystallized bexarotene to RXRα (Figure 3).
Video 2. Import of a related structure in the 3decision Workspace and superposition.
3decision calculates the protein-ligand interactions upon structure registration and automatically displays them when opening the structures in the 3D Viewer to facilitate ligand binding analysis. Both bexarotene and compound 2 form polar contacts between their carboxylic acid group and the Arg316 and Ala327 residues on the protein (Figure 3A). Examining the binding pocket structure and properties, we can see that the chlorine substituent on 2 protrudes into a lipophilic sub-pocket, and the phenyl ring is placed in a highly lipophilic region (Image 3B).
Figure 3. Binding site comparison between docking pose of compound 2 in RXRα binding site (in orange) and X-ray crystal structure of bexarotene in complex with RXRα (PDB ID: 4K6I, in blue). A) Protein-ligand interaction patterns for compound 2 (in orange) and bexarotene (in blue). The grey arrows indicate the salt bridges between the carboxylic acid moieties, the N backbone of Ala327 and the side chain of Arg316. The protein-ligand interactions are calculated by the 3decision® software upon structure registration. B) Representation of the binding site surface, color-coded by hydrophobicity index: from red (hydrophilic) to green (hydrophobic). Surface calculated with the 3decision® software. Hydrophobic sub-pockets targeted for ligand optimization are indicated by the grey arrows. The phenyl ring and chlorine atom on compound 2 are indicated in orange.
In the original study, these structural considerations drove the optimization of compound 2. With the aim of targeting the identified lipophilic sub-pockets around the ligand, the chloride substituent on the 4-position of the pyrimidine core was replaced by a trifluoromethyl group, and the dimethylaniline was modified by closing a ring over the two methyl groups yielding a 4-aminoindane derivative. These modifications produced larger hydrophobic substituents that better fitted the binding site. Finally, replacing the thioether moiety in 2 (a potential labile and toxic group) with a longer ether linker yielded compound 3, which exhibited improved potency and physicochemical properties (Scheme 2).
Scheme 2. Chemical structures of compound 2 and compound 3 with reported literature values1 of EC50 (for RXRα), aqueous solubility, and cytotoxicity. Compound 3 was produced from compound 2 by structure-driven optimization. The modified substituents of compound 2 are indicated in the dashed lines.
Ligand Docking Pose Validation
In a second study from the same group,² the X-ray crystal structure of RXRα with the optimized compound 3 was obtained, elucidating the binding mode of this series of compounds.
This experimental structure is publicly available (PDB ID: 8pp0), so it is included in the 3decision database and can be easily added to the same workspace of the docking pose to continue the ligand binding comparison.
Video 3. Add a structure available in the 3decision database to the Workspace.
The experimental structure confirms the binding mode of the predicted docking pose (Figure 4). The salt bridge between the carboxylic acid of 3 and Arg316 and Ala327 is confirmed, and the molecule is oriented in the predicted pose. The trifluoromethyl group protrudes further into the small hydrophobic cavity, and the indane group is accommodated in the pocket occupied by the phenyl on 2. Overall, there is an excellent shape complementarity of 3 with the ligand binding site, and the interaction network between the pyrimidine portion of the molecule seems highly optimized.
Figure 4. Ligand binding model comparison between docking pose of compound 2 in RXRα binding site (in orange) and X-ray crystal structure of compound 3 in complex with RXRα (PDB ID: 8PP0, in white). Protein-ligand interactions are calculated by the 3decision® software upon structure registration. The grey arrows indicate the salt bridges between the carboxylic acid moieties, the N backbone of Ala327, and the side chain of Arg316. The binding site surface, color-coded by hydrophobicity index, is calculated with the 3decision® software. The phenyl ring and chlorine atom on compound 2 are indicated in orange, and the indane and trifluoromethyl moieties on compound 3 are indicated in white.
Thanks to the experimentally validated structural insights, the indane moiety of compound 3 was further optimized, guided by the structural consideration that a larger, more rigid, hydrophobic group could be accommodated in the pocket. The structure-guided optimization yielded the final compound 4, which carried a tetrahydroquinoline ring instead of indane and showed high potency, markedly superior solubility, and low toxicity.
Scheme 3. Chemical structure of compound 4 and reported literature values2 of EC50 (for RXRα), aqueous solubility, and cytotoxicity.
Conclusions
The described structure-guided optimization of RXR agonists led to the development of a new set of highly potent and soluble compounds that hold the premise for building the next-generation RXR-targeting drugs.¹‚² Structural information from both in silico and experimentally produced 3D protein structures were employed, demonstrating how their synergetic use could increase the speed of structure-based ligand optimization and, in general, of the drug discovery and development process.
This study showcases how structure-based drug design can effectively support the production of improved therapeutics and highlights some key aspects where a software solution can be implemented, providing structural insights and analytical tools.
3decision is a prime example, offering seamless integration of experimental and computational models through a single, unified platform, maximizing the use of structural data. Its unique structure registration system standardizes and centralizes 3D protein structures, making it easy to compare structures from multiple sources. In addition, 3decision offers advanced visualization and analytics capabilities that facilitate efficient examination of ligand binding, streamlining structure-based drug discovery workflows for faster and more efficient discovery of drugs within the pharmaceutical industry.
Figure 5. The centralization of 3D protein structures from different sources (e.g., in silico models and experimental structures) in a single repository like 3decision creates a structural knowledge database to exploit for structure-based drug discovery efforts. This maximizes the use of structural data, aiding ideation and lead optimization, ultimately enhancing the discovery of new drugs.
References
1. Pollinger, J.; Gellrich, L.; Schierle, S.; Kilu, W.; Schmidt, J.; Kalinowsky, L.; Ohrndorf, J.; Kaiser, A.; Heering, J.; Proschak, E.; Merk, D. Tuning Nuclear Receptor Selectivity of Wy14,643 towards Selective Retinoid X Receptor Modulation. J. Med. Chem. 2019, 62, 4, 2112–2126.
https://pubs.acs.org/doi/10.1021/acs.jmedchem.8b01848
2. Lewandowski, M.; Carmina, M.; Knümann, L.; Sai, M.; Willems, S.; Kasch, T.; Pollinger, J.; Knapp, S.; Marschner, J. A.; Chaikuad, A.; Merk, D. Structure-Guided Design of a Highly Potent Partial RXR Agonist with Superior Physicochemical Properties. J. Med. Chem. 2024, 67, 3, 2152–2164.
Bonus Section
Structure selection for docking
To select the best structure for docking studies in 3decision, you can exploit the multi-query advanced search capabilities. The known 3D protein structures of RXRα with bexarotene can be easily retrieved by combining a query for all the RXRα structures with the search of all the structures in complex with bexarotene (using the integrated sketcher to draw the ligand 2D chemical structure). This search will run through the entire 3decision database, searching for structures from the public and proprietary domains, as well as experimental and in silico models.
Figure 6. 3decision Advanced Search page. On the left, you can search by structure property or chemical structure. On the right, you have the launched queries. The Uniprot Code search for “RXRA_HUMAN retrieved 115 hits, and the Chemistry Search for the exact structure of bexarotene got 2 hits. The two queries combined by the logical operator “AND” (in the orange button) provided a total of 2 structures in the database matching both, as indicated in the blue button “Display 2 Results” on the bottom.
The combined queries return two hits. The 3decision results navigation page provides a concise overview of the structures, helping you to choose the most appropriate structure for docking. The PDB structure 4k6i was selected since it is a high-resolution structure of RXRα (NCOA2 is an activator) in complex with bexarotene, while the other one is a heterodimer of RXRα with PPARγ. The PDB file of the selected structure can be easily exported and readily used for docking.
Figure 7. 3decision Results Navigation page. For each structure a general overview is provided, with information such as: resolution, experimental methos, chains composing the protein, and ligand(s) 2D representation. Easy export of the structure(s) of interest is available.
