Holliday Junction Thermodynamics and Structure: Coarse-Grained Simulations and Experiments

Abstract

Holliday junctions play a central role in genetic recombination, DNA repair and other cellular processes. We combine simulations and experiments to evaluate the ability of the 3SPN.2 model, a coarse-grained representation designed to mimic B-DNA, to predict the properties of DNA Holliday junctions. The model reproduces many experimentally determined aspects of junction structure and stability, including the temperature dependence of melting on salt concentration, the bias between open and stacked conformations, the relative populations of conformers at high salt concentration, and the inter-duplex angle (IDA) between arms. We also obtain a close correspondence between the junction structure evaluated by all-atom and coarse-grained simulations. We predict that, for salt concentrations at physiological and higher levels, the populations of the stacked conformers are independent of salt concentration, and directly observe proposed tetrahedral intermediate sub-states implicated in conformational transitions. Our findings demonstrate that the 3SPN.2 model captures junction properties that are inaccessible to all-atom studies, opening the possibility to simulate complex aspects of junction behavior.

Figure 1. Idealized schematic of the junction conformations.

The stacked (a) iso-I and (b) iso-II conformations predominate at modest and high salt concentrations. (c) The open conformation is mainly observed at low salt concentration, and also potentially acts as an intermediary of conformational changes between iso-I and II. The colors denote the strands of the J3 junction, labeled X, R, B, and H, following the convention of Lilley and co-workers. See methods for the complete sequence. In the lower panel, the arrows highlight the inter-base distances for non-complementary bases of distinct strands at the junction center. In iso-I, the TT and CC separations are small (red arrows), while the AG distances are large (blue arrows); the opposite occurs for iso-II. The distances are essentially identical in the open conformation. We take advantage of these differences to distinguish conformations.

Figure 2. Melting properties of the J34 junction.

Comparison of the single-strand fraction α from (a) simulations and (b) experiments. We fit the data using the van’t Hoff equation (equation 1), indicated by the solid lines. Following convention, we define T_M as the temperature at which α = 0.5.

Figure 3. Salt concentration dependence of the junction melting temperature T_M.

At all concentrations, T_M differs by <3% between 3SPN.2 simulations (green circles) and absorption experiments (red diamonds). The lines are only intended as a guide to the eye. The figure also shows the Debye screening length used by the 3SPN.2 model in the Debye-Hückel approximation to the electrostatic interactions. The plateau of the melting temperature coincides with strong screening of electrostatic interactions.

Figure 4. Thermodynamics of junction melting.

The entropy ΔS and enthalpy ΔH of melting obtained by fitting the experiments (red diamonds) and simulations (green circles) to the van’t Hoff equation (equation 1). The inset shows an approximate linear entropy-enthalpy compensation relation. Both the experimental and the simulated data show better agreement with the van’t Hoff equation evaluated at T_M (eq. 2, blue line).

Figure 5. Criterion for distinguishing junction conformations.

Distribution of inter-base separation at the middle of the junction for (a) the AG bases, where a small separation identifies the iso-II conformer, and (b) TT or CC pairs, where a small separation identifies the iso-I conformer. The longer distance peak at low salt concentration is due to open conformations; at higher salt, it arises primarily from the complementary stacked form. The vertical dotted line indicates the cutoff criterion we use to subsequently identify conformational states of individual configurations.

Figure 6. Molecule-to-molecule variations in conformational sampling and conformational transition probabilities.

(a) Example time series of junction conformations for five of the 100 ensemble members at [Na⁺] = 300 mM. The open isoform is short lived, and acts a transition state between iso-I and iso-II. (b) Matrix of the transition probabilities from a given starting state to final state at [Na⁺] = 300 mM. The transition probabilities to the same state (diagonal elements) are not shown, since the tendency to remain in the current state dominates the scale of other transition probabilities. Note that the transition probabilities for iso I→II (and vice-versa) are nearly zero. (c) Salt concentration dependence of the four key transition probabilities.

Figure 7. Junction conformations, abundance, and structure.

(a) Representative conformations observed in our simulations; only the DNA backbone is shown, for simplicity. (b) The relative abundance of the primary isoforms as a function of salt concentration. Isoform identities are defined by the base separations at the junction interior, as described in the text. (c) Inter-duplex angle (IDA) for each conformation. Note that the open planar conformation is only predominant at low salt. At higher salt, the small fraction of open conformations sampled adopt (on average) a tetrahedral conformation.

Figure 8. Junction structure comparison between all-atom and coarse-grained models.

The IDA for the all-atom AMBER model (black) and 3SPN.2 coarse-grained model (red) at T = 283 K and [Na⁺] = 200 mM. The main panel shows the distribution of sampled IDA values; solid lines are the calculated frequency, and dotted lines are a normal distribution with the same mean and standard deviation as the data. The inset shows the original time series for each model, from which the distributions are determined. Note that for the 3SPN.2 model, we have 2 μs of data, not all of which are shown in the inset.

Figure 9. Sequence of the junction J34.

Each arm is 17 bp long, a truncated version of the J3 junction.