High-resolution EPR distance measurements on RNA and DNA with the non-covalent Ǵ spin label

Abstract Pulsed electron paramagnetic resonance (EPR) experiments, among them most prominently pulsed electron-electron double resonance experiments (PELDOR/DEER), resolve the conformational dynamics of nucleic acids with high resolution. The wide application of these powerful experiments is limited by the synthetic complexity of some of the best-performing spin labels. The recently developed \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\bf\acute{G}$\end{document} (G-spin) label, an isoindoline-nitroxide derivative of guanine, can be incorporated non-covalently into DNA and RNA duplexes via Watson-Crick base pairing in an abasic site. We used PELDOR and molecular dynamics (MD) simulations to characterize \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\bf\acute{G}$\end{document}, obtaining excellent agreement between experiments and time traces calculated from MD simulations of RNA and DNA double helices with explicitly modeled \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\bf\acute{G}$\end{document} bound in two abasic sites. The MD simulations reveal stable hydrogen bonds between the spin labels and the paired cytosines. The abasic sites do not significantly perturb the helical structure. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\bf\acute{G}$\end{document} remains rigidly bound to helical RNA and DNA. The distance distributions between the two bound \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\bf\acute{G}$\end{document} labels are not substantially broadened by spin-label motions in the abasic site and agree well between experiment and MD. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\bf\acute{G}$\end{document} and similar non-covalently attached spin labels promise high-quality distance and orientation information, also of complexes of nucleic acids and proteins.


MD simulations Minimization and equilibration
Energy minimizations were carried out in the sander program as implemented in Am-ber16. 1 Firstly, the solvent and ions were relaxed by 500 steps of steepest descent minimization and by 500 steps of conjugate gradient minimization. The nucleic acid atom positions and theǴ (G-spin) atom positions were restrained with a force constant of 500 kcal·mol −1 · Å −2 during this step. Non-bonded interactions were treated with a cut-off of 12 Å. Relaxations were performed with constant box volume. Secondly, the restraints were exclusively kept at the nucleic acid atoms, while relaxing the rest of the system. Thirdly, the whole system was free to relax with 1000 steps of steepest descent minimization and 1500 steps of conjugate gradient minimization.
The system was equilibrated in 50,000 steps with a time step of 2 fs (= ∧ 100 ps). Weak atom position restraints (10 kcal·mol −1 · Å −2 ) were applied at the atoms of the nucleic acid atoms. The solvent, ions, andǴ molecules were free to equilibrate. Random velocities were drawn from a Maxwell-Boltzmann distribution. Long-range electrostatic interactions were treated with the particle-mesh Ewald summation and a cut-off of 12 Å was applied for non-bonded real-space interactions. Covalent bonds to hydrogen atoms were constrained with the SHAKE algorithm. Langevin dynamics with a collision frequency of 1.0 ps −1 slowly heated the system up from 0 K to 300 K. In a second equilibration step, we switched from the sander to the pmemd.cuda engine in Amber16. The complete system was relaxed in 500 ps with a constant temperature of 300 K using Langevin dynamics (γ = 1.0 ps −1 ).
Isotropic position scaling with a relaxation time of 2 ps ensured an average pressure of 1 atm. A random seed was set at the restart. Non-bonded interactions were truncated after 12 Å and hydrogen atoms were constraint with SHAKE.

Atom types and partial charges of abasic sites andǴ spin label
Supplementary

Collective variable describingǴ conformations
The dihedral angle d1 describes the accessibleǴ conformations in a dsNA helix. It has been shown that a TEMPO spin label covalently attached at a cytosine base 2 could be characterized via rotations of two torsion angles Φ 1 and Φ 2 . Following this approach, the potential energy surface (PES) for theǴ was constructed, based on the torsion angles Φ 1 and Φ 2 (SI Figure 3). Six minima were found, where Φ 1 adopts 180 • (minima 1, 2, 5, and 6) or 0 • (minima 3, and 4). As an initial structure, d1 was constrained to 0 • . The structure was minimized in 500 steps with the steepest descent algorithm followed by 1500 steps with conjugate gradient algorithm. The non-bonded cut-off was set to 999 Å. The resulting structure was then equilibrated in 500 ps (∆t = 1 fs). The system was slowly heated up from 0 K to 300 K using Langevin dynamics with a collision frequency of 1.0 ps −1 . A production run was performed for 100 ns.
TheǴ structure with Φ 1 =d1=0 • was used to generate starting structures for d1 = 60 • and −60 • , which where the used as starting structures for d1 = 120 • and −120 • , and the structure with d1 = 120 • for a structure with d1 = 180 • . These structures were then taken as starting structures for the various windows, which were simulated as described for The initial structures ofǴ molecules inside dsDNA were obtained by using the struc- G was rotated to d1 = 120 • , afterwards to 60 • and finally to 0 • , following the described protocol as in dsDNA. With bothǴ molecules inside dsRNA having d1 =0 • , the previously described protocol for dsDNA was repeated for dsRNA.

Evaluation: Classical description of d1 rotation inǴ
We were interested if the force field employed for theǴ molecule is able to capture the main characteristics of the collective variable d1. The overall shape of the potential energy surface at a PBE0/N07D level of theory, describing a rotation around d1, was used as a reference (SI Figure 7). Distances were derived with DeerAnalysis. 4 Background color coding indicates the confidence, with decreasing confidence from green (<3 nm) to red (>5 nm, where distances cannot be resolved).

Average helix parameter
Helix and base pair parameters were calculated with the do_x3dna program package. The average parameter for twist, roll, slide, inclination, h-rise, and propeller were directly calculated from do_x3dna. The distributions of the values are peaked and the average values for χ, α, β, γ, δ, and ζ angles were calculated according to the mean of circular quantities for every frame (Eq. 1) whereθ is the average angle over n θ angles. The individual base pair averages were then averaged over the whole sequence (SI Table 4). The errors were calculated as the standard deviation of the base pair averages to the overall average, divided by the number of base pairs (σ/ √ N).