Multischaalmodellering van DNA: van basenparen tot chromosomen

Abstract

This thesis is devoted to the mechanical and topological properties of DNA, the repository of our genetic information. These properties are fundamental to abroad range of biological processes, essentially setting the ground rules for the organization of DNA and its interactions with the cellular machinery. We study physical properties of DNA and emergent phenomena spanning six orders of magnitude in length scale—from individual base pairs to entire chromosomes. Given this breadth, we develop effective coarse-grained descriptions tailored to each scale, retaining only the degrees of freedom essential to the phenomenon of interest, applying a synergistic approach that combines computational modeling and statistical mechanics throughout. At the shortest length scales (∼10 base pairs), we examine the mechanical properties of DNA, focusing in particular on the emergence of beyond-nearest-neighbor couplings between base pairs. Using a momentum-space formalism applied to all-atom molecular dynamics simulations, we characterize these off-site interactions and demonstrate that they are not unique to DNA but arise also in double-stranded RNA. A simple mechanical ladder model shows how these non-local couplings emerge naturally from the coarse-graining of atomic degrees of freedom into angular variables, suggesting they are a universal feature of double-stranded polymers. Momentum-space stiffnesses are further shown to be a powerful methodological tool for validating coarse-grained models. As an example, we compare two coarse-grained models of DNA—cgDNA and cgDNA+—against all-atom reference data. This analysis reveals that cgDNA+ accurately captures both short- and long-range mechanical properties of DNA, while the cgDNA model fails at large length scales. An important implication of beyond-nearest-neighbor couplings is that they enable local mechanical perturbations—such as those induced by protein binding—to propagate along the DNA to distant sites. We study this mechanism in the context of DNA-mediated allostery, in which two proteins binding at distant sites along the DNA interact indirectly through the mechanical properties of the intervening DNA linker. We develop a statistical mechanical model in which protein binding is treated as a local perturbation that either deforms or rigidifies the DNA. The model predicts three distinct allosteric mechanisms—two of enthalpic origin and one entropic—each with a characteristic dependence on the length of the intervening linker. Comparison with experimental measurements of DNA allostery reveals distinctive signatures that allow one to identify which of the proposed mechanisms best accounts for the data. At the kilobase-pair scale, we study the intricate interplay between DNA mechanics and topology within the context of DNA supercoiling. Exposing DNA to excessive torsional stress drives a buckling transition in which the molecule forms intertwined looped structures called plectonemes. We introduce a novel statistical mechanical framework—the multiplectoneme model—that extends the classical two-phase model by explicitly accounting for the configurational entropy associated with the spatial distribution and length partitioning of multiple coexisting plectonemic domains, balanced against a nucleation free energy penalty. With only two free parameters, the model accurately reproduces Monte Carlo simulation data of a coarse-grained DNA model for plectoneme counting statistics, torque responses, and length distributions, and shows good quantitative agreement with optical and magnetic tweezer experiments across a range of stretching forces. An overestimation of plectoneme number relative to fluorescence microscopy data is attributed to the limited spatial resolution of fluorescence-based detection rather than a deficiency of the model. At the megabase-pair scale, we infer the spatial organization of interphase chromosomes by analyzing open-source data from multiplexed fluorescence in situ hybridization (mFISH) experiments. Adopting a polymer physics perspective, our analysis is centered around pairwise distance distributions between genomic loci. We find these to be well-described by a two-population model, which we term the α- and β-phases. Through a scaling analysis of the characteristic radii of both phases, we find that they represent different folding modes of chromatin: the α-phase corresponds to a compact conformation consistent with a crumpled globule, while the β-phase reflects a spatially extended, geometrically confined conformation potentially reflective of a looped domain. Crucially, both phases are found to coexist within individual cells, suggesting a microphase seperation scenario. To test this hypothesis, we developed a statistical model of a microphase-separated polymer—the heterogeneous fractional Brownian motion (hfBm) model—which combines two fractional Brownian motions reflecting the distinct geometries of each chromatin phase. Despite its simplicity, the model quantitatively reproduces pairwise distance distributions, power-law scaling relations, and higher-order angular correlations.