Gel retardation studies and other experiments indicate that DNA sequences containing the d(GA(4)T(4)C)(n) motif are curved, whereas those of identical composition but with a reverse sequence polarity, the d(GT(4)A(4)C)(n) motif, are straight. Hydroxyl radical cleavage experiments show that d(GA(4)T(4)C)(n) shows a unique signature, whereas d(GT(4)A(4)C)(n) behaves normally. To explain these results at a molecular level, molecular dynamics (MD) simulations were performed on the DNA duplexes d(G(5)-(GA(4)T(4)C)(2)-C-5) and d(G(5)-(GT(4)A(4)C)(2)-C-5) to 3.0 and 2.5 ns, respectively. The MD simulations are based on the Cornell force field implemented in the AMBER 4.1 modeling package and performed in a neutral solution of anionic DNA with K+, Cl- and Mg2+ at concentrations roughly comparable to a ligase buffer. Long range interactions were treated by the particle mesh Ewald method. Analysis of the results shows that the calculated dynamical structure of d(G(5)-(GA(4)T(4)C)(2)-C-5) exhibits strong gross curvature, consistent with the observed behavior. The most significant locus of curvature in the MD structure is found at the central C-15-G(16) step, with an average roll angle of 12.8 (+/-6.40)deg. The d(G(5)-(GT(4)A(4)C)(2)-C-5) MD structure exhibited significantly less gross curvature. Analysis of results indicates that the reduction in gross curvature in the d(G(5)-(GT(4)A(4)C)(2)-C-5) trajectory originates from the effect of the T-10-A(11) and T-20-A(21) average roll angles of 12.5(+/-5) deg. These three steps T-20-A(21), are half-helix turns away from one another, and their contributions to concerted bending cancel out. The A-tracts in the MD structure are essentially straight. The dynamical. structure of d(G(5)-(GA(4)T(4)C)(2)-C-5) exhibited minor groove deformation comprised of expansion at the 5' end of A-tracts and progressive narrowing towards the 3' end, consistent with and elaborating the interpretation of hydroxyl radical chemical probing results. (C) 1999 Academic Press.