Geometries and energies of Watson-Crick base pairs in oligonucleotide crystal structures

We analyzed geometries and energies of 469 GC and 224 AT pairs occurring in 11 A-, 54 B- and 12 Z-DNA crystal structures. The most frequent hydrogen bond length conforms to the canonical value but distributions of the hydrogen bond lengths are unexpectedly wide. The average GC pair is rather compressed and opened into the double helix minor groove while an average AT pair has canonical dimensions. The extreme base pair geometrical parameters are the following: buckle -26 degrees and 39 degrees, propeller -33 degrees and 26 degrees, opening -23 degrees and 19 degrees, shear -1.4 A and 2.1 A, stretch -0.7 A and 0.4 A, and stagger -2.3 A and 1.8 A. The analyzed set contains complementary bases apparently bound by unusual bifurcated hydrogen bonds. The unusual base pairing geometries are of two kinds. In the first case, an extreme value of one parameter is compensated for by other geometrical parameters so that the energy of the resulting geometry is acceptable. However, there are also examples when the compensatory effect is missing and then the base pairing instability is dramatic. The most extensive base pair deformations occur in dodecamers, the d(GGATGGGAG) nonamer, the r(UUAUAUAUAUAUAA) 14-mer, and the d(ICCGG) tetramer whose common feature is a low structure resolution. However, very unstable base pairs are also present in the decamer d(CCAACGTTGG) and hexamer d(CGTACG) whose structures were solved at a relatively high resolution. As the deformations probably originate from crystal packing forces and/or data and refinement errors, we recommend to omit these structures from studies of DNA sequence-structure relationships. The remaining hexamers, octamers and decamers whose list and their Watson-Crick base pair characteristics are given in the article are suitable for this purpose because they exhibit no prohibitive energy deviations from the canonical hydrogen bonding properties.