Correlation of crystallographically determined and computationally predicted hydrogen-bonded pairing configurations of nucleic acid bases.

Crystals of pairs of H-bonded nucleic acid bases are generally grown from nonaqueous solutions. We have been able to predict the H-bonded configuration of most of the base pairs in such crystals by using an empirical-potential function we recently developed for calculating the energetics of such interactions in chloroform solution. The following configurations were computationally predicted to predominate and are those observed in crystal structures: the Watson-Crick G.C configuration instead of two competing configurations; the Hoogsteen-type configurations for A.T, A.U, and A.br5U instead of Watson-Crick-type configurations; the Watson-Crick-type configurations for 2-aminopurine.br5U instead of the purine N3-type configuration; the Watson-Crick-type configurations for 8-bromo-2,6-diaminopurine.T instead of the Hoogsteen or purine-N3-type configurations; the syn-anti configuration for br8A.br8I instead of the anti-anti configuration; the Watson-Crick-type configurations for br8A.br5U instead of the Hoogsteen-type configurations; and the Hoogsteen-type configurations for me8A.T instead of the Watson-Crick configurations. In addition, the H-bonded base triplet br5U.2,6-diaminopurine.br5U was calculated to have Hoogsteen and Watson-Crick-type configurations but not the purine N3-type configuration. Apparently, lattice forces and chance nucleation of a minor base pairing configuration are not significant when the stability difference between the preferred and alternative configurations exceeds a relatively small value. In one case, in order to correctly predict the base pairing configuration in the crystal, it was necessary to include a contribution due to a C--H...O bond, suggesting that this type of H bond can make a significant contribution to base pair stability.