Section 23.3.4.2. A-tract bending

It has long been known that introduction of short A-tracts into general-sequence B-DNA in phase with the natural 10–10.5 base-pair repeat produced overall curvature that could be detected via eletrophoretic gel retardation, ring-cyclization kinetics and other physical measurements in solution (Marini et al., 1982; Wu & Crothers, 1984; Koo et al., 1986; Crothers & Drak, 1992). However, the microscopic source of the observed macroscopic curvature remained unclear. Solution measurements alone cannot discriminate between three alternative curvature models: (1) local bending within the A-tracts themselves; (2) bending at junctions between A-tract B-DNA and general-sequence B-DNA; or (3) inherently straight and unbent A-tracts, with curvature resulting from removal of the normal writhe expected in general-sequence B-DNA (Koo et al., 1990; Crothers et al., 1990). The three curvature models are compared schematically in Fig. 10 of reference B77.

X-ray crystallographic results for DNA oligomers come down unequivocally in favour of model (3) above. Short A-tracts of four to six base pairs are straight and unbent in C-G-C-G--C-G-C-G (B1–B6), C-G-C--G-C-G (B20), C-G-C--G-C-G (B31), C-G-C--G-C-G (B17, B52), C-G-C-G--G-C (B64) and C-A-A-A-G -G (B105) (A-tracts are double-underlined). It has been claimed (Sprous et al., 1995) and disputed (Dickerson et al., 1994, 1996) that the observed straightness of crystalline A-tracts was only an artifact of crystal packing, or of the high levels of methyl-2,4-pentanediol (MPD) used in the crystallization. This concern now is put to rest by the observation that B-DNA packed against a protein molecule in its biological working environment behaves exactly the same as B-DNA packed against other DNA molecules in the crystal, as borne out by the roll/slide/twist studies of El Hassan & Calladine (1997) for DNA and of Dickerson (1998a,b,c) and Dickerson & Chiu (1997) for protein–DNA complexes. Added support has come from recent molecular-dynamics simulations by Beveridge and co-workers (Sprous et al., 1999), who have demonstrated that the duplex of sequence GGGGGGAAAATTTTAAAATTTTCCCCCC is severely curved because of a roll kink at the double-underlined central CG step, whereas the duplex GGGGGGTTTAAATTTAAACCCCCC is much less curved because the roll kink at CG is counterbalanced by roll kinks in the opposite direction at the two flanking TA steps. In both cases, A-tracts are straight and completely unbent. (Note that both roll kinks can involve compression of the major groove, as expected, because the kink sites are a half turn of helix apart.)

This similarity of behaviour of DNA in crystals and in protein–DNA complexes should come as no surprise, since the local molecular environments – close intermolecular contacts, partial dehydration, low water activity, low local dielectric constant, high ionic strength, presence of divalent cations – are similar in these two cases and quite different from that of free DNA in dilute aqueous solution. Far from being unwanted `crystal deformations', the local changes in structure resulting from intermolecular contacts in DNA crystals provide positive information about sequence-dependent deformability that is relevant to the protein recognition process. With regard specifically to A-tract behaviour, Occam's Razor would argue in favour of model (3) above for the behaviour of A-tracts in solution. The situation in dilute aqueous solution becomes of secondary importance if what is wanted is an understanding of A-tract B-DNA behaviour in protein–DNA complexes. Here, the answer is unambiguous: A-tracts in their biological setting are inherently rigid structural elements, chosen by natural selection when bending should be avoided.