Intrinsic Disorder in Biomolecular Systems : A Modelling Challenge

Abstract

For a long time, it has been thought that a specific and well-defined three-dimensional (3D) structure was a prerequisite for a protein to function. However, more recently, the existence of several proteins that lack a 3D structure and that are still functional has been well documented. Such proteins are called intrinsically disordered proteins. The disorder in them can affect the full or part of the protein and their plasticity allows them to interact with numerous biomolecules by adapting the shape of the unfolded part to the different partners. In this way, they can participate in diverse cellular processes such as regulation, maintenance of the shape of the cell (cytoskeleton), protein degradation and DNA condensation. This disorder can be native, but can also be induced by an external stimulus, like the interaction with other biomolecules, environmental changes and light. Structural characterization of such intrinsically disordered systems at atomic scale is very difficult, but crucial for a better understanding of their functional mechanism and also for the possible design of new drugs. However, the lack of structure, the lack of methods to inspect their states, and the lack of tools to analyze those, remains a considerable challenge. Though NMR is used in structural biology generally as a tool for structure determination, it can also help as a spectroscopic tool to define the foldedness and disorder of biomolecules and to characterize their dynamics. In this thesis, I show that NMR can be combined to computational modelling to characterize those intrinsically disordered systems. In such dynamic systems the NMR parameters are measured as time and ensemble averages, not unlike the situation in normal structure determination of proteins with internal mobility. Computational methods to deal with ensemble averaging to treat internal or side chain mobility have been developed for native proteins and small peptides previously. Nevertheless, for the highly flexible elements as present in intrinsically disordered proteins, these computational approaches are inadequate. The interpretation of the averaged data of grossly different geometries is more delicate and the large ensemble introduces a new modelling challenge. In the first part of my thesis, I introduce new computational methods that can better describe the partially unfolded state of proteins from NMR data than was possible till now. In the second part of this thesis, I focus on glycans, which represent another example of highly flexible systems, and their interaction with proteins. Along this thesis, I highlight the inadequacy of conformational energy, which is commonly used in molecular modelling to define the most likely states of biological systems, when applying this to highly flexible systems and when experimental information is lacking. For this case, it is suggested to disregard potential energy in the analysis, and to base the selection of the most representative ensemble from a large collection of structures on the direct comparison versus experimental data or by adopting a statistical approach. Both methods might contribute to overcome the challenging concept of “free energy” that is frequently met in molecular modelling.