Conformational characteristics of DNA: Empirical classifications and a hypothesis for the conformational behaviour of dinucleotide steps

This paper is concerned with an investigation of the geometry and structure of DNA as revealed by X-ray diffraction of single crystal oligomeric structures. A database of atomic coordinates of 60 naked (i.e. not bound to any protein or drug) DNA oligomers (25 dodecamers, 18 decamers, 16 octamers and 1 tetramer) is set up and carefully described. An extensive empirical study of the geometries of DNA dinucleotide steps in the database, involving only unmodified Watson-Crick base pairs (A-T and G-C), is reported, and a number of new correlations and classifications are described in detail. The main conclusions include the kinematic classification of dinucleotide steps into two main classes: rigid and loose (or flexible or bistable). 'Continuously flexible' steps are shown to exercise their flexibility along a well-defined single-degree-of-freedom pattern, with roll, slide and twist all correlated linearly. The rigid steps are AA/TT, AT and GA/TC, and the loose (bistable) steps are GG/CC, GC, CG while the loose (continuously flexible) steps are CA/TG and TA. AC/GT is the least clear of all steps and it is perhaps best described as neither a rigid nor a loose step but rather an 'intermediate' step. The base-pair parameters are also carefully examined and the resulting pivotal correlation between the average propeller and the flexibility of the step (equals the standard deviation of slide), that we have recently described elsewhere (El Hassan & Calladine 1996), is examined in some detail. A simple two-parameter scheme for the description of the conformation of the sugar phosphate backbone is given and used to classify the sugar phosphate backbones in all entries of our database into A-backbone and B-backbone conformations. The role of the backbone in determining the conformational preferences of the dinucleotide steps is examined by demonstrating that whereas the B-backbone conformation permits a fairly narrow channel in the roll/slide/twist conformational space, with all three parameters linearly correlated, the A-backbone allows only a small 'box' region near the high-roll, low-twist, low-slide end of the space. Finally, the empirically determined conformational characteristics of the various dinucleotide steps are accounted for in terms of (a) mechanical stacking effects associated with propeller-twisting of constituent base pairs (the propeller-flexibility correlation), (b) chemical stacking effects associated with the special electrostatic charge distributions and pi-pi effects in homogeneous G/C steps (Hunter 1993), and (c) backbone-dictated effects that govern in the absence of (a) and (b).