Free-energy perturbation calculations of DNA destabilization by base substitutions:: The effect of neutral guanine thymine, adenine cytosine and adenine difluorotoluene mismatches

The ability to determine the stability of non-Watson-Crick base pairs in DNA by computer simulation is a necessary prerequisite for quantitative theoretical studies of DNA replication fidelity. Here we report calculations of the relative free energy of nucleotide mismatches in DNA. Our study evaluated the "solvation" free energy cost associated with thymine (T) --> difluorotoluene (F), T --> cytosine (C), and C --> T transformations in aqueous solution and in a DNA duplex. The free energy differences were evaluated by the free energy perturbation (FEP) method based on classical all-atom simulations in a spherical surface-constraint water model. The dependence of the calculated free energies on the DNA sequence, the simulation length, size of the water droplet, and phosphate charges was examined. The calculations were carried out for both the negatively charged DNA backbone neutralized by explicit Na+ counterions and the neutral backbone with no counterions. Although the latter model provided overall free energy differences that were less sensitive to the radius of the simulation sphere and to the total simulation time, the results obtained with short 150 ps simulations for the fully charged system containing counterions in 18 A water sphere showed the best overall agreement with the corresponding experimental results. The overall Watson-Crick geometry of an adenine thymine (A.T) base pair has not changed during its transformation into an A.F base pair, where F is a nonpolar isostere of thymine. The calculated change in the duplex binding free energy at 25 degreesC (Delta DeltaG degrees (bind)) for this "mutation" was 5.1 kcal/mol, compared to 3.6 +/- 1.7 kcal/nlol determined previously from the observed DNA melting thermodynamics. The transformation of C, forming a Watson-Crick pair with guanine (G), into T yielded a wobble G.T mismatch analogous to the one observed by X-ray crystallography. Delta DeltaG degrees (bind) obtained for this mismatch (relative to the Watson-Crick G.C base pair) by FEP calculations were in a 1.0-3.4 kcal/mol range depending on DNA sequence and simulation protocol used. This result is in reasonable agreement with the experimental estimate of about 3.6 kcal/mol. Starting from an A.T base pair, mutation of the thymine carbonyl group into the amino group led to a change of the overall geometry of the base pair from Watson-Crick to reverse wobble. The Delta DeltaG degrees (bind) Of the resulting neutral A.C mismatch (relative to the Watson-Crick A.T base pair) calculated using the neutral and ionic DNA model was 10 and 7 kcal/mol, respectively. Using the observed binding affinity and pK(a) constants of the N1-protonated A(+).C base pair, the corresponding experimental Delta DeltaG degrees (bind) Of the neutral A.C base pair was determined to be 7.3 +/- 1.5 kcal/mol. Our ability to reproduce reasonably stabilities of non-Watson-Crick base pairs by "first principle" calculations will be used in future calculations of DNA polymerase fidelity.