Substituting Inosine for Guanosine in DNA: Structural and Dynamic Consequences

Synthesis and Purification of DNA Substrates

Standard and altered DNA phosphoramidites were purchased from Glen Research. The sequence mimics the U5 terminus of the HIV genome and the hairpins are named HIV-G and HIV-I hairpin. DNA substrates were synthesized using a 391 DNA synthesizer by Applied Biosystems. Standard synthesis protocols were used. ⁷ Deprotected oligonucleotides were purified by ion exchange using a Pharmacia 15Q PE column and then desalted using a GE HiTrap desalting column. Oligonucleotide concentrations were determined under denaturing conditions using extinction coefficients calculated from the sum of their mononucleotides. ⁵

T_m Thermal Melting Data

Ultraviolet melting curves were recorded at 260 nm in 1 cm cells on a Cary 100 Spectrophotometer using a temperature ramp of 0.5°C/min and analyzed as described previously. ⁸

NMR Spectroscopy

All NMR experiments were performed using a Bruker Avance 600 spectrometer with IDTG triple resonance and QXI probe heads. Samples in D₂O were in 50 mM NaCl, 10 mM sodium phosphate, 0.1 mM ethylenediamine tetraacetic acid (EDTA), at a pH* of 6.86 (meter reading). NOESY data were collected at 298 K using mixing times of 75, 150, and 250 milliseconds. Lowflip angle COSY, DQF COSY, TOCSY, and ³¹P-¹H HPCOR and constant-time NOESY (CT-NOESY) were recorded as described previously. ^9,10 For samples in water (90% H₂O:10% D₂O), imino proton spectra were obtained from samples in 50 mM NaCl, 10 mM sodium phosphate, and 0.1 mM EDTA at a pH of 6.6. Assignment and integration of two-dimensional spectra were done using SPARKY. ^{11 1}H spectra were referenced to internal DSS and ³¹P spectra were referenced to 85% H₃PO₄ in a capillary and recorded at 298K.

Basepair Lifetimes

The base pair lifetime for the HIV-I and HIV-G hairpins was determined from the T ₁ relaxation time of the imino protons in the presence of an ammonia catalyst. ^12-14 The base pair exchange rate, τ _ex, can be calculated from the following:

1/T ₁ = 1/τ _ex + 1/T ₁ ^°

The T ₁ experiments were performed at 278 and 298 K. T ₁ ^° is the relaxation time in the absence of any added catalyst. ¹³ The DNA samples of ~1 mM were prepared in a buffer of 50 mM NaCl, 10 mM NaP, 0.1 mM EDTA at a pH of 7.8. The pH was measured each time after catalyst was added. The NH₄OH had a 6 M concentration and the concentration of the ammonia catalyst was calculated from the following:

[NH₃] = [NH₄OH] × 10^(-pKa) / (10^(−pH) + 10^(−pKa))

The HIV-G-hairpin was titrated with the catalyst using catalyst concentrations of 2, 2.5, 3.5, 5, 10, 20, 40, and 100 mM. The HIV-I hairpin was titrated with the catalyst at catalyst concentrations of 5, 10, 15, 30, and 60 mM. The base pair lifetime at an infinite catalyst concentration can be estimated by the plot of 1/[cat] vs τ _ex:

τ _ex = τ _open +1/(K _d α[B]k _i)

where α is a correction factor for the accessibility difference of imino protons between a mononucleotide and the open state of an oligonucleotide (α ≈ 1). k _i is the imino proton transfer rate with a value of ~2×10⁸ s^-1 M^-1. K _d is the equilibrium constant of the open and closed base pair and [B] is the concentration of the catalyst. The y-intercept τ _open is the inverse of the opening rate and represents the base pair lifetime at an infinite concentration of catalyst. ¹³ The comparison of the K _d values for the I-C and the G-C pairs assumes similar values for the I and G residue.

Model Generation

The initial models for the hairpin structures were generated in the NAB module in AMBER as detailed previously, except that PARMBSC0 was used. ⁹ Briefly, LEaP was employed to generate topology and starting structure files. The DNA hairpins were then solvated in an octahedral box (10 Å from the solute) with 3453 TIP3P water molecules. Seventeen Na⁺ ions were added for electroneutrality and the system was then minimized with the DNA held rigid, force constant 500, prior to minimization of the entire system.

Structure Calculations

NMR assignments, restraint generation, restraint implementation, and structure calculation followed established protocols. ^9,15 Briefly, ¹H resonances were assigned via 2D ¹H NOESY pathways with the assistance of TOCSY spectra. ³¹P resonances were assigned based on HPCOR experiments. NOESY cross peak volumes were integrated in SPARKY using a Gaussian or Sum Over Box method. ¹⁶ A percentage error was manually assigned based on visual inspection of a projected cross slice overlaid with the integral trace. For unresolved peaks, a Sum Over Box integration method was used and a higher percentage error was assigned. Quantitative distance restraints were derived using an iterative RANDMARDI procedure using CORMA, MARDIGRAS, and AMBER cycles as described previously. ^9,15,17 R ^x values were calculated in CORMA using a correlation time (τ _C) of 3.5 nanoseconds for base and sugar protons. DNA sugar pucker and pseudorotation angles were assessed using the Altona graphical method. ^18,19 3JH1′- H2′1, 3JH1′-H2′2, 3JH1′-H3′, ∑ H1′, and ∑ JH3′ were measured from ³¹P decoupled low-flip angle COSY and DQF COSY experiments. Pseudorotation angles were derived for the dominant form of each deoxyribose and converted to torsion angle NMR restraints using the PUCKER script from AMBER. Backbone ε torsion restraints were calculated based on the ratio of peak heights from CT NOESY experiments as described in Tjandra and Bax. ¹⁰ Broad backbone torsion angle restraints for α, β, γ, and ζ were generated for nucleotides with typical Btype helix ³¹P chemical shifts using standard values for a B-type DNA helix. ²⁰ Fully restrained solvated molecular dynamic simulations were run for 10 nanoseconds using a parmbsc0 force field at 300 K. Final ensembles were generated by sampling 10 structures at the end of the simulation followed by individual fully restrained minimization. The final structure was selected based on having the lowest overall CORMA R ^x score while maintaining a low level of restraint violations.

Structure Analysis

Helical analysis was performed on each final structure using Curves + in order to obtain intra-basepair and backbone parameters. ²¹ The output file will not include base opening parameters if the hairpins are run as a single strand of DNA so the hairpin loop was first removed from each structure file and the residue names were made contiguous so that Curves + would read the nucleotides in the correct order. RMSF calculations were done for the final trajectories (10 nanoseconds) using VMD. ¹¹ The trajectory tool was used for the RMSF calculations for residues 4 and 15. For residue 4 the H2 on the inosine and the N2 and its hydrogens on the guanosine were purposefully excluded from the atom selection. Final structures were aligned according to the atom selection in each case prior to an all atom RMSD calculations.

Free MD Simulation

Each structure also underwent free MD (no restraints) at 300 K. ⁹ The final NMR derived hairpin structures were loaded into LEaP and solvated as already described under Model Generation but using PARMBSC1. System equilibration and fMD were carried out in the following order: initial minimization, then 200 picosecond isochoric isothermal MD to bring the temperature to 300 K, minimization, then four 125 picosecond isobaric isothermal MD (isothermal–isobaric) runs, with force constants of 5, 3, 2, and 1, each followed by minimization and finally a production MD (no constraints) for 1 microsecond with a 2 femtosecond time step. The resulting trajectories were loaded into VMD and RMSD calculations were performed as described in the Structure Analysis section.