Conformational Properties of Specific Biomolecular Systems Investigated Using Molecular Dynamics Simulations

Molecular dynamics (MD) simulations play nowadays an increasingly important role in many areas of biology, chemistry and physics. Growing computational power and ongoing advances in methodology and model accuracy have enabled MD to become a valuable complement to experimental methods and, in some cases, even a substitute. Chapter 1 gives a brief introduction into MD, specifies the methods used in the present thesis, and provides key concepts concerning the two types of biomolecular systems considered, namely carbohydrates and lipids. Chapter 2 investigates the conformational equilibrium of the six-membered ring, which is a common structural feature in carbohydrates. The conformational equilibrium is studied in terms of the influence of non-polar ring substitution (one or two substituents of various bulkiness in different positions and orientations). There, the MD simulations with the local elevation umbrella sampling (LEUS) method are used in order to characterize all relevant conformers of the ring, and to calculate the free energy differences between them. The results are compared to those of quantum-mechanical calculations and to available experimental data. It is shown that for most isomers, the chair conformation with substituents oriented equatorially is the preferred one, and the preference is stronger for bulkier substituents. The exceptions are trans-1,2 isomers with bulky substituents, where the axial orientation is preferred in order to avoid vicinal-gauche repulsions between the substituents in equatorial orientation. Two compounds trans-1,3-di-tert-butylcyclohexane and cis-1,4-di-tert-butylcyclohexane are also found to adopt a non-negligible amount of non-chair conformations. The compound trans-bis-cyclohexylmethyl-cyclohexane, which is structurally similar to a branched trisaccharide, slightly prefers 4C1 over 1C4. This suggests that the destabilization of the 4C1 conformation of the central residue of LewisX, which is investigated in Chapter 3, is not due to a steric repulsion. Chapter 3 presents new GROMOS force-field parameters for 12 functionalized hexopyranoses including N-acetyl-2-deoxy-D-glucopyranose (GlcNAc), as well for the branched trisaccharide LewisX. The MD simulations with the LEUS method are used to investigate the conformational behavior of these compounds in terms of ring conformation and, in the case of GlcNAc and LewisX, 3JH;H-values and NOE-derived proton-proton distances, respectively. Although the functionalized hexopyranoses present the expected ring conformation along with 3JH;H-values in agreement with experimental data, inverted ring conformations of the GlcNAc residue in LewisX cause disagreements with experimental NOE-derived proton-proton distances. The reason of the discrepancy is the variation of the atomic partial charges upon functionalization, which dramatically affects the ringconformational equilibrium via third-neighbor electrostatic interactions around the ring. Two force-field modifications EXCL and TORS are proposed in order to reduce the dependence of the ring conformation upon the functionalization. The EXCL modification excludes the third-neighbor electrostatic interactions around the ring. The TORS modification implements the same exclusions as EXCL, but introduces in addition new torsional terms which substitute to the electrostatic interaction of the excluded pairs. The EXCL modification does not describe well the ring properties of unfunctionalized hexopyranoses. On the other hand, the TORS modification leads to similar conformational behavior as the GROMOS force field 56A6CARBO, but does not improve the conformational behavior of LewisX. Although the TORS approach could be a promising way to solve the ring conformational properties upon functionalization, further parameterization of the new torsional terms will be necessary. Chapter 4 investigates the glycosidic-linkage conformational preferences of 31 aldohexopyranose disaccharides using two types of theoretical approaches. The first approach, conformational scanning (CS), relies on performing systematic variations of the glycosidic and exocyclic dihedral angles, all other covalent parameters taking fixed values. The results are used to construct maps in the glycosidic space (CS-maps) characterizing separately : (i) the potential for steric clashes; (ii) the influence of short-range (including stereolectronic) effects local to the linkage; (iii) the potential for inter-residue hydrogenbonds. The second approach, the MD simulations with LEUS, characterize the systems in terms of free-energy maps in the gycosidic space (MD-maps). It is shown that the computationally inexpensive CS-maps reproduce well the numbers, positions and relative free energies of the main minima on the MD-maps. Steric clashes and short-range effects are recognized as the leading conformational forces, with virtually no influence of inter-residue hydrogen-bonding in water. All maps except for (1->6)-linkage present a single significantly populated basin, generally broad and possibly accompanied by one or two metastable states with high relative free energies. These observations suggest that the flexibility of oligo- and polysaccharide chains is due primarily to thermal fluctuations around a main glycosidic preference, with a secondary role for the occasional occurrence of alternative glycosidic states. Chapter 5 focuses on a different class of biological molecules, lipids. More specifically, it is concerned with bilayers of lipid glycerol-1-monopalmitate (GMP). The lipid bilayer can be in phases depending on the composition and thermodynamic conditions. Two phase transitions are investigated, involving the liquid crystal phase (stable at high temperature) and the gel phase (stable at low temperature in an aqueous environment) or the interdigitated phase (stable at low temperature in a water-methanol mixture). Three methods are tested in their ability to facilitate the determination of the main transition temperature between these phases, namely MD simulation initiated from a homogeneous phase, MD simulation initiated from a system consisting of two phases in contact (mixed phases) and temperature replica exchange MD (T-REMD). The MD with mixed phases is expected to decrease the computational cost by eliminating the time needed to form a nucleation or a disruption unit for the formation of the ordered or disordered phase, respectively. The T-REMD method should decrease the computational cost by allowing replicas to cross barriers at either high or low temperatures. It is observed that the T-REMD method does not lead to any significant enhancement of the convergence rate associated with the transition temperature determination in comparison to MD initiated from a single-phase, whereas MD with mixed phases shows a faster convergence and narrower prediction intervals. The reason for the unsuccessful enhancement by the T-REMD method is that the phase transition, as a first-order transition, has a discontinuity in the molar enthalpy as a function of the temperature. This causes a gap between the potential energy distributions of consecutive replicas which are in different phases. This results in scarce exchanges between these two replicas, and prevents all replicas to reach sufficiently low or high temperature at which the transition would occur within the simulation time