TL;DR
This paper tackles the challenge of predicting drug side effects by exploring how ligands can bind to multiple sites across different pentameric ligand-gated ion channels (pLGICs). Using a multimodal computational approach, the researchers reveal that many drugs can interact with homologous sites in different channels or multiple distinct sites within the same channel, highlighting a significant potential for polypharmacology and informing future drug development.
What Problem Does This Paper Address?
Developing new drugs is a complex process, and a major hurdle is predicting and mitigating adverse drug effects. Pentameric ligand-gated ion channels (pLGICs) are a large family of crucial targets for many therapeutic areas, including anesthetics, anxiolytics, and smoking cessation aids. However, these channels are also implicated in a wide range of neuropsychiatric adverse events, from seizures to mood changes. The challenge lies in the fact that pLGICs have numerous binding sites, and many drugs are not highly selective, leading to unintended interactions (polypharmacology). This promiscuity can be due to ligands binding to similar sites across different pLGIC subtypes (homologous sites) or to multiple distinct sites within the same receptor (non-homologous sites). Current experimental screening methods are often limited, especially for allosteric binding sites, making it difficult to fully understand and predict these off-target effects. This paper aims to address this gap by developing and applying computational tools to systematically map and predict ligand interactions across the pLGIC superfamily, particularly focusing on sites where experimental screening data is scarce.
Methods
The researchers employed a comprehensive, multimodal computational approach to investigate ligand binding across pLGICs. Key techniques included:
- Structure-Based Pharmacophore Screening: Using existing 3D structures of pLGICs with bound ligands from the Protein Data Bank (PDB), they generated pharmacophore models representing the essential features of these binding sites. These models were then used to screen a large database of drugs to predict potential interactions.
- Descriptor Calculation and Analysis: They calculated physicochemical descriptors (z-Scales and MSWHIM) for amino acids within binding sites. Hierarchical clustering of these descriptors was used to assess the similarity between different binding pockets, both within and across pLGIC families.
- Conformation Analysis: Recognizing that binding sites can adopt different shapes, they analyzed the conformational flexibility of these pockets using Principal Component Analysis (PCA) to understand how ligand binding might be influenced by these dynamic states.
- Variant Data Mining: They analyzed genetic variant data (from GnomAD) near binding sites to assess the potential impact of human genetic diversity on drug interactions.
- Homology Modeling and Computational Docking: For specific sites of interest, particularly where experimental data was limited, they built homology models and performed computational docking to predict how ligands might fit into these pockets.
This combination of methods allowed them to explore binding site similarities, predict drug interactions, and investigate the influence of protein flexibility and genetic variation.
Datasets
- Protein Data Bank (PDB): A comprehensive collection of experimental 3D structures of pLGICs with bound ligands was used as the primary source for structural analysis and pharmacophore generation. The study integrated structures from 31 GABAARS, 41 GlyRs, 26 nAChRs, and 16 5-HT3Rs.
- DrugBank and PubChem: A database of 168 investigated drugs was compiled from these sources for pharmacophore screening.
- GnomAD (Genome Aggregation Database): Genetic variant data was used to assess the prevalence of missense variants in and around pLGIC binding sites.
Key Insights and Ideas
- Ligand Promiscuity is Widespread and Multifaceted: The study strongly suggests that ligands exhibit two primary forms of promiscuity: binding to homologous sites across different pLGIC families and binding to multiple distinct sites within the same receptor. This challenges the notion of highly selective drug action and highlights the complexity of predicting off-target effects.
- Transmembrane Domain (TMD) Sites are Highly Conserved and Promiscuous: The TMD, particularly the upper TMD interface, emerged as a region with broad ligand overlap across many pLGIC superfamily members. This is attributed to its structural conservation and the presence of sites that can accommodate diverse chemical scaffolds.
- Extracellular Domain (ECD) Sites Show More Variability: While ECD sites are crucial for orthosteric binding, they also contain structurally variable regions, including INDELs, making them more challenging for broad computational predictions but potentially offering opportunities for subtype-specific targeting.
- Conformational Flexibility Matters: The study emphasizes that static structural snapshots are insufficient. Ligand binding can induce conformational changes, and considering these dynamic states is crucial for accurate in silico drug screening.
- Phytocannabinoids as a Case Study: The investigation into Δ⁹-THC revealed its potential to interact with multiple binding sites within pLGICs, including TMD sites and a lipid-associated site, underscoring the need for multimodal approaches to understand complex drug interactions.
Key Results
- The computational analysis identified known and novel binding sites within pLGICs, with many in the TMD being broadly conserved and thus likely to interact with a wide range of ligands.
- Pharmacophore screening predicted that the phytocannabinoid Δ⁹-THC could bind to multiple pockets across the pLGIC superfamily, including sites in the upper and lower TMD and a lipid-associated site.
- Descriptor analysis of binding sites generally recapitulated known pharmacology, differentiating between orthosteric and allosteric sites, and providing insights into similarities across receptor families.
- Conformational analysis revealed that while some sites are relatively rigid, others, particularly in the TMD, exhibit flexibility that can influence ligand binding.
Limitations and Future Work
- Limited Experimental Data for Some Sites: The authors acknowledge that in vitro assays suitable for high-throughput screening are critically lacking for many allosteric binding sites, particularly in the TMD. This limits the ability to experimentally validate all computational predictions.
- INDELs in ECD Sites: The presence of variable regions and INDELs in ECD sites makes them challenging targets for current computational methods, requiring more advanced approaches for accurate prediction.
- Drug Library Completeness: The drug library used for screening, while representative, is not exhaustive, meaning some potential interactions might have been missed.
- Future Work: The authors suggest that their findings provide a basis for improved structure-guided in silico drug development and derisking. Future work could involve developing more sophisticated computational methods to handle INDELs and exploring the role of endogenous ligands in interacting with multiple pLGIC sites. Experimental validation of the predicted binding modes for compounds like Δ⁹-THC is also a crucial next step.
Summary updated