Structure of a B-DNA dodecamer. II. Influence of base sequence on helix structure

Detailed examination of the structure of the B-DNA dodecamer C-G-C-G-A-A-T-T-C-G-C-G, obtained by single-crystal X-ray analysis (Drew et al., 1981), reveals that the local helix parameters, twist, tilt and roll, are much more strongly influenced by base sequence than by crystal packing or any other external forces. The central EcoRI restriction endonuclease recognition site, G-A-A-T-T-C, is a B helix with an average of 9.8 base-pairs per turn. It is flanked on either side by single-base-pair steps having aspects of an A-like helix character. The dodecamer structure suggests several general principles, whose validity must be tested by other B-DNA analyses. (1) When an external bending moment is applied to a B-DNA double helix, it bends smoothly, without kinks or breaks, and with relatively little effect on local helix parameters. (2) Purine-3′,5′-pyrimidine steps open their base planes towards the major groove, pyrimidine-purine steps open toward the minor groove, and homopolymer (Pur-Pur, Pyr-Pyr) steps resist rolling in either direction. This behavior is related to the preference of pyrimidines for more negative glycosyl torsion angles. (3) CpG steps have smaller helical twist angles than do GpC, as though in compensation for their smaller intrinsic base overlap. Data on A-T steps are insufficient for generalization. (4) G.C base-pairs have smaller propellor twist than A · T, and this arises mainly from interstrand base overlap rather than the presence of the third hydrogen bond. (5) DNAase I cuts preferentially at positions of high helical twist, perhaps because of increased exposure of the backbone to attack. The correlation of the digestion patterns in solution and helical twist in the crystal argues for the essential identity of the helix structure in the two environments. (6) In the two places where the sequence TpCpG occurs, the C slips from under T in order to stack more efficiently over G. At the paired bases of this CpG step, the G and C are tilted so the angle between base planes is splayed out to the outside of the helix. This TpC is the most favored cutting site for DNAase I by a factor of 4.5 (Lomonossoff et al., 1981). (7) The EcoRI restriction endonuclease and methylase both appear to prefer a cutting site of the type purine-purine-A-T-T-pyrimidine, involving two adjacent homopolymer triplets, and this may be a consequence of the relative stiffness of homopolymer base-stacking observed in the dodecamer.