Open Access

Jérémy Curuksu

• This PhD dissertation deals with the application of atomic-scale computer simulation (Molecular Dynamics) on different aspects of the conformational flexibility of Nucleic Acids (i.e. DNA, RNA). I have used diverse statistical and computational methods such as Umbrella Sampling and Replica Exchange to extract entropic properties, e.g. free energy, characterising (i) the bending of DNA on short length-scale, (ii) the folding of recently discovered RNA ribosomal motifs (the kink-turn motif) and (iii) backbone dihedral conformations accessible to damaged DNA. One achievement is the reproduction of the experimental curve for the probability of very high bend angles observed for short fragment of DNA which demonstrates a non linear (softer) bending flexibility of DNA. Results of my thesis predict that DNA kinks (local defects unstacking neighbour basepairs) occur in vivo and some of them induce a 90⁰-turn in the helix. They are associated with a systematic decrease of the local DNA stiffness constant (half an order of magnitude) which was quite unexpected. DNA bending up to 150⁰ on the 5 nm length scale requires on average 12 kcal/mol. It is slightly less expensive, ̃10 kcal/mol when a run of consecutive adenines is present (Atract sequences which occur frequently in every genome). The same kind of information is provided for a larger kink-turn motif in RNA (̃10 base pairs) for which an almost iso- energetic twist / bend coupled mode has been characterised. A methodological development of Hamiltonian replica exchange sampling techniques enables to characterize several competing DNA backbone conformations accessible to damaged DNA (an abasic site). The method can be used for any combination of backbone dihedral variables which constitute an example of DNA metastable states with very short relaxation time. More generally the PhD thesis presents the development of new methods to tackle the accurate sampling of particular nucleic acid helical propensities.