Mechanical properties of symmetric and asymmetric DNA A-tracts: implications for looping and nucleosome positioning

A-tracts are functionally important DNA sequences which induce helix bending and have peculiar structural properties. While A-tract structure has been qualitatively well characterized, their mechanical properties remain controversial. A-tracts appear structurally rigid and resist nucleosome formation, but seem flexible in DNA looping. In this work, we investigate mechanical properties of symmetric AnTn and asymmetric A2n tracts for n = 3, 4, 5 using two types of coarse-grained models. The first model represents DNA as an ensemble of interacting rigid bases with non-local quadratic deformation energy, the second one treats DNA as an anisotropically bendable and twistable elastic rod. Parameters for both models are inferred from microsecond long, atomic-resolution molecular dynamics simulations. We find that asymmetric A-tracts are more rigid than the control G/C-rich sequence in localized distortions relevant for nucleosome formation, but are more flexible in global bending and twisting relevant for looping. The symmetric tracts, in contrast, are more rigid than asymmetric tracts and the control, both locally and globally. Our results can reconcile the contradictory stiffness data on A-tracts and suggest symmetric A-tracts to be more efficient in nucleosome exclusion than the asymmetric ones. This would open a new possibility of gene expression manipulation using A-tracts.

between any two ions was chosen to be 3.5Å and the minimum distance between any ion and the DNA molecule was 5Å.
This procedure was applied to all the DNA oligomers studied. As a result, we obtained initial configurations of the simulated systems, each with a different DNA oligomer of 18 base pairs, immersed in explicit water and ions at physiological salt concentration. Each of the systems consisted of about 33,000 atoms in total. Prior to the production MD run, all the systems were equilibrated.

S1.2 Equilibration procedure
The equilibration procedure included several energy minimizations and short MD runs with positional restraints of decreasing strength imposed on the DNA. The initial minimization was followed by slow heating, each of the next minimizations was followed by a short equilibration MD run. All the steps were carried out using sander module of Amber 10.
The initial minimization consisted of 500 steps of steepest descent followed by 500 steps of conjugated gradients at constant volume periodic boundary conditions, with 25 kcal.mol −1Å−2 positional restraints on the DNA. After that, a 100 ps MD run was performed at constant volume with the same positional restraints. During the first 10 ps, the system was heated from 100 K to 300 K and then kept at 300 K. The Berendsen weak-coupling algorithm (coupling constant 0.2 ps) was used for temperature regulation, the nonbonded cutoff distance was set to 9Å. The bonds which contain hydrogen atoms were constrained using the SHAKE algorithm. The MD time step was 2 ps.
All the subsequent minimizations again comprised 500 steps of steepest descent and 500 steps of conjugated gradients, with decreasing values of positional restraints (5, 4, 3, 2, 1 and 0.5 kcal.mol −1Å−2 ). Each minimization was followed by a short MD, using the same positional restraints as in the preceding minimization. The MD runs were 50 ps each, carried out at 300 K and 1 atm, with pressure and temperature regulation using the Berendsen algorithm (coupling constants 0.2 ps). The last step of the equilibration phase was another 50 ps MD run at 300 K and 1 atm with no positional restraint and with the Berendsen coupling constants increased to 5.0 ps. The final density fluctuated closely around 1 g.cm −3 and the total volume around 334×10 3Å3 for all the studied systems.

S1.3 Production
The production phase was performed using the pmemd module of Amber 10. Periodic boundary conditions were used, the constant temperature 300 K and pressure 1 atm were maintained by the Berendsen algorithm with coupling constants 5.0 ps, time step was 2 ps. The nonbonded cutoff was set to 9Å and all the bonds containing hydrogen atoms were treated with the SHAKE algorithm. Translation of the center of mass was removed every 5000 time steps. DNA snapshots were extracted from the simulated trajectories using the ptraj module of Amber Tools 1.0. Figure S1: Diagonal stiffness constants related to propeller deformation for the full, non-local rigid base model (upper row) and the unidimensional stiffness constants defined by eq (9) in the main text (lower row). Errors are mean differences between values for the whole trajectory and for its halves. Figure S2: Unidimensional stiffness constants defined by eq (9) in the main text for roll, twist and slide.  Figure Figure S7: Energy profiles due to threading of isolated A 6 tract (red) and A 6 tract together with its flanking sequences (12 bp in total, see Table 1 of the main text -cyan). The energy is higher for the embedded tract, since the sequence is longer. However, the phasing of the periodic profile is very similar.   Table S2: Bending magnitude and direction of the A-tracts alone (without flanking sequences) and the corresponding parts of the control. The direction of 0 • indicates bending towards the major groove, 180 • means bending towards the minor groove. Errors are mean differences between values for the whole trajectory and for its halves (MD) or standard deviations over the NMR models (1fzx).