For years, interactions between double stranded (duplex) DNA were presumed to be independent of the DNA structure and base pair sequence because the nucleotides are buried inside the double helix and shielded by the highly charged sugar-phosphate backbone. In discussion of such interactions, duplex DNA was explicitly or implicitly modeled as a uniformly charged cylinder. However, this concept was based on intuitive perception rather than experiments or rigorous theory. In reality, the experimental evidence, e.g., transformation of duplex DNA from a non-ideal helix with 10.5 base pairs/turn in solution into a nearly ideal helix with 10.0 bp/turn in aggregates, suggested that this concept may be wrong. Starting from the classical paper of Rhodes and Klug published in 1980, it became clear that interactions between duplex DNA not only depend on but also affect the double helix structure.? ? To account for possible effects of the structure of the sugar phosphate backbone on DNA-DNA interactions, over the last decade we have been developing a theory of electrostatic interactions between macromolecules with helical patterns of surface charges. Even the simplest models, which did not account for dynamic variations in the structure, e.g., due to the thermal motion, already suggested possible explanations for many observations. Such observations included the torsional deformation of the double helix upon aggregation mentioned above, counterion-specificity of DNA condensation, multiple liquid crystalline phases in DNA aggregates, and measured intermolecular forces. We, therefore, continued development of this theory and its applications to various phenomena. ? ? Most importantly, this theory predicted that the dependence of the backbone structure on the nucleotide sequence might be sufficiently strong to affect DNA-DNA interactions. In the last year, we confirmed this prediction by statistical analysis and comparison of known structures of DNA oligonucleotides in crystals (determined by x-ray diffraction) and in solution (determined by NMR). We evaluated the helical coherence length -- a cumulative parameter quantifying sequence-dependent deviations from the ideal double helix geometry. We found, e.g., that the solution structure of synthetic oligomers is characterized by 10-20 nm coherence length, which is similar to 15 nm coherence length of natural, salmon-sperm DNA extracted from fiber x-ray diffraction patterns. Packing of oligomers in crystals dramatically alters their helical coherence. The coherence length increases to 80-120 nm, consistent with its theoretically predicted role in interactions between DNA at close separations.? ? The effects of the sequence on interactions between duplex DNAs, e.g., the predicted direct recognition of sequence homology between 100 base pair (bp) or longer sequences, may obviously have significant biological implications. In particular, 100-300 bp sequence homology recognition is essential for avoiding recombination mistakes that lead to cancer, genetic disorders, etc. Experiments recently reported in the literature suggested that local, transient pairing of homologous sequences in intact DNAs may precede double strand breaks, further recognition by protein-covered single strands, and strand crossover. Direct interactions between duplex DNAs in nucleosome-free regions were proposed to be involved, but the mechanism and possibility of sequence homology recognition in such interactions remained unknown.? ? During the last year, we continued experiments probing sequence-dependent DNA-DNA interactions in vitro. We imaged different mixtures of two fluorescently tagged, double helical DNA molecules with identical nucleotide composition and length, but different sequences. In electrolytic solution at minor osmotic stress these DNAs formed discrete liquid-crystalline aggregates (spherulites). We observed spontaneous segregation of the two kinds of 294-bp-long DNA within each spherulite, revealing nucleotide sequence recognition between double helices separated by water in the absence of proteins. The segregation was observed for several different binary mixtures, suggesting that it was not specific to a particular pair. Consistent with our theoretical predictions, the segregation was much weaker for 150-bp-long DNAs. While these experiments unequivocally demonstrate the possibility of sequence homology recognition without unzipping the double helix, much work remains to be done to test whether the mechanism of this recognition is indeed as predicted by the theory and whether such recognition plays any role in DNA pairing in vivo. Further experiments are currently in progress.
