Development of models to describe the dynamics and interaction with water molecules in protein-ligand binding

Wassermoleküle formen Netzwerke, die einen erheblichen Einfluss auf die Protein-Ligand Bingung nehmen. Sowohl strukturelle als auch thermodynamische Daten werden miteinander korreliert. Durch Molekulardynamische Simulationen werden Netzwerke erfolgreich reproduziert und vorhergesagt.

Saved in:
Bibliographic Details
Main Author: Betz, Michael
Contributors: Klebe, Gerhard (Prof. Dr.) (Thesis advisor)
Format: Doctoral Thesis
Published: Philipps-Universität Marburg 2015
Online Access:PDF Full Text
Tags: Add Tag
No Tags, Be the first to tag this record!

In modern drug design, lead optimization is usually performed in a step- wise fashion by adding and replacing functional groups at a given lead scaffold. Promising lead scaffolds with systematically varied substituents are synthesized and the structural variations are correlated with observed trends in binding affinity. Assigning contributions to particular functional groups, to estimate by how much an added group will enhance the binding affinity, is on first sight a very pragmatic but often misleading approach and does not follow simple additivity rules. The resulting structure–activity relationships frequently do not reflect the expected gradual changes in a congeneric ligand series. Often enough, the thermodynamic binding profiles of such ligands are modulated or can switch unpredictably and unexpected candidates surprisingly turn out to be the highest affinity binders of a series. The binding affinity to be optimized is a Gibbs free energy, which is the sum of an enthalpic and an entropic contribution, which usually cannot be factorized into simple additive functional group contributions. Regarding the complexity of the protein-ligand complex formation, one must consider that changes in ∆G, ∆H and T∆S reflect how the system is modified as a whole. Often such effects on binding affinity are not obvious, as they are frequently masked by a mutual compensation of binding enthalpy and entropy resulting in only a minor change in the Gibbs free energy. Nevertheless, a thoroughly combined analysis of both, thermodynamic terms along with structural information, as presented in many studies described in this thesis, provides profound insights into the underlying structure–activity relationships. In the present thesis two aspects have been studied that are recognized as increasingly import for rational computer-aided drug design. The first aspect relates to the influence of water on the non-additivity of functional group contributions to the binding affinity and it is important to elucidate the driving forces behind the so-called hydrophobic effect. Using high resolution crystal structures allowed us to explain and correlate observed changes in the thermodynamic binding profiles with changes in the local solvation structure on the molecular level. As a model system to investigate the influence of water on thermodynamic binding profiles we used the zinc metalloprotease thermolysin (TLN). Because the TLN has itself a rather rigid geometry and because the studied congeneric ligands differ only slightly, many important aspects of the binding process, e.g. ligand desolvation and differences in ligand conforma- tional properties can be assumed as virtually identical in the comparison of the protein-ligand binding event. The most important observation is the fact that water molecules do not interact with the solutes as separate and independent particles, but rather as “supramolecular assemblies” reflecting repeatedly occurring geometric arrangements. These local assemblies are parts of a large hydrogen-bonded network. Undoubtedly, the completeness and quality of the formed local network influences binding entropy and enthalpy in a systematic manner. This suggests that, rational approaches will have to incorporate concepts how to regard and optimize solvent exposure of molecular surfaces. For this purpose, we developed a molecular dynamics (MD) simulation protocol, with which we successfully predicted the detailed solvation structure around different TLN-inhibitor complexes. In reality, however, most proteins are more flexible than TLN, for which the factorization of the thermodynamic profile correlates remarkably well with high resolution crystal structures. Thus, in the second part of this thesis, it was investigated whether such correlations can also be observed for more flexible proteins. Therefore, the properties of several lin-benzopurine derivatives binding to the tRNA-modifying enzyme tRNA-guanine transg- lycosylase (TGT) were investigated. In a straight-forward computer-aided drug design procedure, e.g. docking, the protein part is regarded as virtually rigid while the ligands are flexible. More sophisticated docking methods take protein side chain flexibility into account, to reflect, e.g. an induced-fitting upon ligand binding. But such methods are not capable to describe the highly dynamic changes as observed for TGT. To inhibit TGT, which is only active as a homodimer, we followed different concepts by either blocking the active site (Chapter 7), but also by destabilizing the protein-protein interface of the homodimer. We demonstrated that ligands differing merely by an exocyclic amino group trigger large-scale rearrangements in the protein structure. Therefore, changes in the thermodynamic profile, which relate to the entire protein-ligand binding process and the involved protein flexibility, prevent that individual contributions can be assigned to single interactions or individual binding steps. Despite of the available high-resolution crystal structure of TGT and its complexes, some uncertainties in the models remain. For example, active-site ligands that insert a needle-like rigid substructure into the protein-protein interface, to perturb certain hot spot regions in order to destabilize the homodimer, are not fully visible in the crystal structures due to an ill-defined electron density. By use of different modeling techniques such as MD simula- tions we provide a model in agreement with experimental data. The simulations reveal how a loop region in TGT, despite its high flexibility, prevents solvent access of the residues forming the protein-protein interface. The research in the present thesis about the two aspects — the influence of water and protein flexibility — allow for a better understanding of funda- mental aspects of protein-ligand binding. Taking such aspects into account will make computational simulations more complex, but hopefully better in predicting ligand-binding affinity, as well as the molecular recognition process.