Modeling of biomembranes: from computational toxicology to simulations of neurodegenerative diseases

Abstract: It was known from the middle of the last century that a cell-membrane is a lipid bilayer. Since that time a large number of experimental studies has been done in order to see how a certain molecule can penetrate through a membrane. Due to the complexity of laboratory experiments computational chemistry became a convenient tool for investigations involving this process. In a real life a compound has to pass through several membranes of different chemical composition before reaching the actual target. Such a diversity in constitution gives a various selectivity to cell-membranes: some molecules will penetrate through them and others will not. That is why the development and a choice of suitable models for lipid bilayers are important steps in such a research. In this thesis new all-atomistic models for polyunsaturated phospholipids in cis conformations have been derived and added to the SLipids force field. After a successful force field validation, the new lipid models were used in molecular dynamics and well-tempered metadynamics simulations of several problems, such as toxicity of hydroxylated polybrominated diphenyl ethers (OH-PBDE), behavior of cholesterol in various membranes, an aggregation of amyloid-β (Aβ) peptides. The significance of the presence of lipid unsaturation has been demonstrated by all computations. 2’-OH-BDE68 (ortho) showed the affinity to saturated lipid bilayer, but had more conformational variations in the center of the unsaturated membrane. Cholesterol did not exhibit the preference to polynsaturated lipid bilayers from free energy calculations, but the diversity in orientations of this molecule, depending on its locations was observed. The behavior of Aβ peptides was dependent on membrane saturation as well. The insertion of Aβ peptides was detected in lipid bilayers containing higher amounts of polyunsaturated phospholipids, while in systems with more saturated membranes amyloids aggregated on membrane surfaces. Moreover, a comparison of simulations for quadro- and mono-component lipid bilayers showed that the membrane built of 18:0-22:6 PC can serve as a good model for the ’healthy’ tissue of a human brain. Also the lipid bilayer built of 14:0-14:0 PC exhibited similar features as the quadro-lipid membrane representing the brain tissue affected by Alzheimer’s disease. Good agreement of some computational results with available experimental findings demonstrated the applicability of computer simulations to real life problems.