Final Report Summary - DRUGPROFILBIND (Chemogenomic profiling of drug-protein binding by shape, enthalpy/entropy and interaction kinetics)
The ERC project DrugProfilBind was aimed at a comprehensive understanding of structural, thermodynamic and interaction-kinetic properties that drive protein-ligand binding. The required thermodynamic signature and binding-kinetic profile, a putative drug molecule has to meet to optimally fight a given disease, can be quite different and depend on the mode-of-action, spatial location and life span of the target protein. Already during lead optimization, purposeful selection of the best and most appropriate candidates, discovered in a screening campaign, desperately needs predictive biophysical parameters to decide which hits to take to the next level of development and which properties to improve to optimally address the target in terms of binding affinity and target residence time.
From a theoretical point of view, enthalpic optimization of ligand binding is a promising and advisable concept for lead selection and optimization. We therefore assembled a knowledge base of crystallographically determined protein-ligand complexes along with carefully collected isothermal titration calori¬metry (ITC) and binding kinetic SPR data, supplemented by MD simulations of congeneric ligand series. The experimental data unfortunately comprise all changes involving the entire protein-ligand binding event. This makes factorization and subsequent assignment of the profile changes to individual interactions or functional group contributions extremely difficult. Furthermore, enthalpy/entropy data can only be compared across narrow congeneric ligand series. Frequently attempted global comparisons across large data samples extracted from literature fail due to problems of scaling such heterogeneous data with respect to a common reference state. Surprisingly, changes in the local water structure from complex-to-complex have a major impact on the enthalpy/entropy signature, e.g. a single water molecule can shift the ΔH/TΔS profile easily by ± 7-8 kJ/mol, whereas the Gibbs free energy is much less affected. Accordingly, optimization strategies that only focus on affinity data can hardly regard information about the local influence of the water structure and design hypotheses must therefore remain limited. In consequence, any interpretation of enthalpy/entropy profiles or reliable predictions of binding poses require detailed information about the water inventory of the complexes being compared.
Nonetheless, pairwise comparisons of complexes relative to each other show that hydrogen bonding relates to an enthalpic signal, which increases with growing charge assistance. Hydrophobic binding (“hydrophobic effect”) strongly depends on the properties of the water molecules being replaced upon ligand binding and span the entire thermodynamic range from more entropy to more enthalpy-driven signature. Poorly or insufficiently solvated binding pockets can result in high-affinity binding of entirely hydrophobic ligand portions. Water-mediated protein-ligand contacts strongly modulate the enthalpy/entropy signature but much less affect the Gibbs energy of binding.
Every formed protein-ligand complex creates a new common surface, which becomes wetted by the surrounding solvent. The quality and perfection of the formed surface water network that wraps around the partly exposed substituents of the bound ligand strongly influences affinity, the thermodynamic signature and the binding kinetic properties of the complex. As the modulations of the surface water network result directly from the properties of the bound ligand and its partly solvent-exposed substituents, fine-tuning of thermodynamic and binding kinetic properties can be accomplished by well-established medicinal chemistry optimization steps. We developed a computer approach to allow simulation of these surface water networks, and thus to design complexes with the desired properties.
To characterize thermodynamic properties of fragment binding, an ITC displacement titration protocol has been established. Remarkably, the binding of fragments is in most instances determined by the local water structure and affinity or enthalpic ligand efficiency is not a reliable criterion to select the most promising fragment candidates for further optimization. Weaker and less enthalpic binding fragments can easily outperform more potent candidates during fragment-to-lead evolution.
From a theoretical point of view, enthalpic optimization of ligand binding is a promising and advisable concept for lead selection and optimization. We therefore assembled a knowledge base of crystallographically determined protein-ligand complexes along with carefully collected isothermal titration calori¬metry (ITC) and binding kinetic SPR data, supplemented by MD simulations of congeneric ligand series. The experimental data unfortunately comprise all changes involving the entire protein-ligand binding event. This makes factorization and subsequent assignment of the profile changes to individual interactions or functional group contributions extremely difficult. Furthermore, enthalpy/entropy data can only be compared across narrow congeneric ligand series. Frequently attempted global comparisons across large data samples extracted from literature fail due to problems of scaling such heterogeneous data with respect to a common reference state. Surprisingly, changes in the local water structure from complex-to-complex have a major impact on the enthalpy/entropy signature, e.g. a single water molecule can shift the ΔH/TΔS profile easily by ± 7-8 kJ/mol, whereas the Gibbs free energy is much less affected. Accordingly, optimization strategies that only focus on affinity data can hardly regard information about the local influence of the water structure and design hypotheses must therefore remain limited. In consequence, any interpretation of enthalpy/entropy profiles or reliable predictions of binding poses require detailed information about the water inventory of the complexes being compared.
Nonetheless, pairwise comparisons of complexes relative to each other show that hydrogen bonding relates to an enthalpic signal, which increases with growing charge assistance. Hydrophobic binding (“hydrophobic effect”) strongly depends on the properties of the water molecules being replaced upon ligand binding and span the entire thermodynamic range from more entropy to more enthalpy-driven signature. Poorly or insufficiently solvated binding pockets can result in high-affinity binding of entirely hydrophobic ligand portions. Water-mediated protein-ligand contacts strongly modulate the enthalpy/entropy signature but much less affect the Gibbs energy of binding.
Every formed protein-ligand complex creates a new common surface, which becomes wetted by the surrounding solvent. The quality and perfection of the formed surface water network that wraps around the partly exposed substituents of the bound ligand strongly influences affinity, the thermodynamic signature and the binding kinetic properties of the complex. As the modulations of the surface water network result directly from the properties of the bound ligand and its partly solvent-exposed substituents, fine-tuning of thermodynamic and binding kinetic properties can be accomplished by well-established medicinal chemistry optimization steps. We developed a computer approach to allow simulation of these surface water networks, and thus to design complexes with the desired properties.
To characterize thermodynamic properties of fragment binding, an ITC displacement titration protocol has been established. Remarkably, the binding of fragments is in most instances determined by the local water structure and affinity or enthalpic ligand efficiency is not a reliable criterion to select the most promising fragment candidates for further optimization. Weaker and less enthalpic binding fragments can easily outperform more potent candidates during fragment-to-lead evolution.