Molecular origin of 31P-NMR chemical shifts of phosphate groups with bivalent counter ions

The electrostatic interactions of phosphate groups and counter ions critically affect the structure, function and reactivity of DNA or RNA. We present a joint experimental-theoretical investigation of dimethyl phosphate (DMP-) in aqueous solution, an established model system of the sugar-phosphate backbone. Utilizing 31P-NMR spectroscopy as probe of phosphate-ion association, variations of Mg2+ and Ca2+ content exhibit a systematic shielding of the 31P chemical shift (δiso(31P)) with moderate temperature dependence. Enhanced sampling molecular dynamics (MD) and ab initio (GIAO-DF-LMP2) level of theory are used to reveal the microscopic mechanism. Simulations are performed for a configurational ensemble of DMP-ion geometries and their first solvation shells, demonstrating (i) the spatial convergence of changes of the nuclear shielding constant σiso(31P), (ii) the intramolecular geometric origin of short-timescale σiso(31P) fluctuations and (iii) an average shift of σiso(31P) of about 3-5 ppm upon contact ion pair formation with Mg2+ or Ca2+ ions. A quantitative analysis of δiso(31P) for varying ion content and temperature allows us to extract the temperature-dependent fraction of the contact ion pair species, indicating that solvent separated or free ion pairs are the energetically preferred species. The results impose boundary conditions for improvements of phosphate ion force fields and establish the interactions underlying the changes of δiso(31P).

💡 Research Summary

This study combines 31P‑NMR spectroscopy with enhanced‑sampling molecular dynamics (MD) and high‑level ab initio calculations to elucidate how divalent cations (Mg2+ and Ca2+) influence the chemical shift of dimethyl phosphate (DMP–) in aqueous solution, a widely used model for the sugar‑phosphate backbone of nucleic acids.

Experimental measurements were performed on 0.2 M DMP– solutions containing a broad range of Mg2+ or Ca2+ concentrations (0.1–25 equivalents) at three temperatures (25 °C, 35 °C, 45 °C). The isotropic 31P chemical shift (δiso) displayed a systematic up‑field (shielding) change of roughly 3–5 ppm as the divalent ion concentration increased, while the temperature dependence was modest (≈0.02 ppm °C‑1). Relaxation times (T1, T2) and linewidth analyses suggested the presence of at least two distinct spectral components, hinting at different ion‑pairing environments.

To interpret these observations, the authors carried out classical MD simulations using the Amber18/24 suite with the parmbsc1 force field, TIP4P‑FB water, and the 12‑6‑4 Lennard‑Jones parameters for Mg2+ and Ca2+. Systems were built in truncated octahedral boxes, neutralized, and equilibrated before 10 ns of unbiased production runs. Umbrella sampling was then applied along the distance between the non‑bridging oxygen O1 of DMP– and the metal ion, covering 2.0–6.0 Å in 0.1 Å increments. The resulting potential of mean force (PMF) revealed a deep minimum near 2.1 Å, corresponding to a contact ion pair (CIP) where the metal directly coordinates a non‑bridging oxygen. Beyond ~3.4 Å the free energy rises, indicating that solvent‑separated ion pairs (SSIP) become the dominant species.

Quantum‑chemical shielding constants (σiso) were computed on the GIAO‑DF‑LMP2 level using Molpro 2024.1. Correlation‑consistent core‑valence basis sets up to quintuple‑ζ (cc‑pCV5Z) were tested for convergence. For a representative DMP–‑Mg2+ CIP geometry, σiso increased by 3–5 ppm relative to free DMP–, matching the experimental δiso shift. The calculations demonstrated that this shielding change originates from the direct metal‑oxygen interaction and the associated redistribution of electron density in the first solvation shell. Adding water molecules up to a 4.5 Å radius around the solute was sufficient for σiso convergence; more distant solvent contributed negligibly.

By fitting the experimental δiso data to a two‑component model, δiso = X·δCIP + (1 − X)·δSSIP, where X denotes the fraction of CIPs, the authors extracted temperature‑dependent CIP populations. Across the studied concentration range, X varied between 0.05 and 0.15, decreasing with increasing temperature, indicating that SSIPs (or free ions) are thermodynamically favored at higher temperatures. Mg2+ and Ca2+ displayed virtually identical X values, suggesting that the current force fields do not differentiate their binding propensities sufficiently.

The work highlights several key insights: (1) 31P‑NMR chemical shifts are highly sensitive to the formation of contact ion pairs; (2) the magnitude of the shielding change (≈3–5 ppm) can be quantitatively reproduced by GIAO‑DF‑LMP2 calculations that include the first solvation shell; (3) CIP formation is energetically favorable at short metal‑oxygen distances but is less populated than SSIP under the experimental conditions; (4) existing 12‑6‑4 ion parameters tend to over‑stabilize SSIPs and under‑represent CIPs, pointing to a need for re‑parameterization guided by the present experimental‑computational dataset.

Overall, the study provides a rigorous benchmark for the interplay between phosphate groups and divalent cations, establishes 31P‑NMR as a quantitative probe of ion‑pairing equilibria, and offers concrete guidance for improving force fields used in simulations of nucleic acids and related biomolecular systems.