What is the basis of sequence-directed curvature in DNAs containing A tracts?

A variety of solution and gel experiments show that DNAs containing tracts of 4-8 A's repeated in phase with the helix repeat are curved. Several independent analyses of these experiments argue that curvature resides in the A tracts themselves. In x-ray crystallographic studies of several DNAs containing A tracts, however, the A tracts are uncurved, leading to models in which curvature resides in the non-A tracts. This "curved general sequence model" has several problems, in our view. We review those, and we describe recent experiments that show that the dehydrating agents commonly used in x-ray crystallography markedly reduce curvature in gels and in solution, calling into question the ability of crystallography to determine the structural basis of DNA curvature. Finally, we discuss the critical role of hydration in curved DNAs and suggest new experiments that we hope could finally determine exactly which sequences are responsible for curvature.