TIBS 23 – SEPTEMBER 1998 341 Copyright © 1998, Elsevier Science Ltd. All rights reserved. 0968 – 0004/98/$19.00 PII: S0968-0004(98)01265-1 32 Hiyama, H., Iavarone, A. and Reeves, S. A. (1998) Oncogene 16, 1513–1523 33 Chinery, R. et al. (1997) Nat. Med. 3, 1233–1241 34 Somasundaram, K. et al. (1997) Nature 389, 187–190 35 Zeng, Y-X., Somasundaram, K. and El-Deiry, W. S. (1997) Nat. Genet. 15, 78–82 36 Mal, A. et al. (1996) Nature 380, 262–265 37 Zerfass-Thome, K. et al. (1996) Oncogene 13, 2323–2330 38 Funk, J. O. et al. (1997) Genes Dev. 11, 2090–2100 39 Jones, D. L., Alani, R. M. and Münger, K. (1997) Genes Dev. 11, 2101–2111 40 Li, X-S. et al. (1996) Cancer Res. 56, 5055–5062 41 Timchenko, N. A. et al. (1997) Mol. Cell. Biol. 17, 7353–7361 42 Maki, C. G. and Howley, P. M. (1997) Mol. Cell. Biol. 17, 355–363 43 Chen, I-T., Smith, M. L., O’Connor, P. M. and Fornace, A. J., Jr (1995) Oncogene 11, 1931–1937 44 Warbrick, E., Lane, D. P., Glover, D. M. and Cox, L. S. (1997) Oncogene 14, 2313–2321 45 Chuang, L. S-H. et al. (1997) Science 277, 1996–2000 46 Chen, J., Chen, S., Saha, P. and Dutta, A. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 11597–11602 47 Levin, D. S. et al. (1997) Proc. Natl. Acad. Sci. U. S. A. 94, 12863–12868 48 Watanabe, H. et al. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 1392–1392 49 Alevizopoulos, K., Vlach, J., Hennecke, S. and Amati, B. (1997) EMBO J. 16, 5322–5333 50 Hofmann, F. and Livingston, D. M. (1996) Genes Dev. 10, 851–861 REVIEWS LOCAL BENDING OF DNA can con- tribute extensively to the specificity of biological events such as gene regu- lation and packaging 1 . In contrast to tra- ditional structural polymorphism (e.g. the B-, A- and Z-DNA structures), bend- ing is a localized micropolymorphism in which the original B-DNA structure is distorted but is not modified extensively. Broadly speaking, the DNA segments that are involved in the protein-induced and/or inherent DNA bending that occurs in many promoters, enhancers and si- lencers are about 10–50 bp in length. DNA molecules in this size range are dif- ficult to model because they are longer than those that can be described easily by atomic-resolution molecular model- ling or quantum-mechanical approaches. Equally, they are shorter than those that can be meaningfully handled by tradi- tional elastic models, which successfully describe macroscopic behaviour (such as supercoiling) in longer DNA seg- ments 2,3 . Also, local DNA conformations and recognition by DNA-binding pro- teins are clearly sequence dependent, so conventional elastic-rod models of DNA, which do not explicitly represent the dependence of the elasticity on the base sequence, cannot tell us much about these conformations. Here, we review briefly the advantages and limitations of rod models of DNA, particularly with regard to elastic modelling of local bending phenomena. Static-geometry models Rod models are the simplest form of DNA models and represent DNA as a cylindrical rod of constant diameter. The shape, in this case, is the path or trajec- tory of the longitudinal z-axis, which can be either straight or curved (Fig. 1a). The common philosophy of rod models is to divide the rod into short cylindrical segments (e.g. the size of a base pair) and then to compute a given rod param- eter on the basis of segment parameters that have to be known a priori. Dinu- cleotide models define the base-pair-size unit as two adjacent base pairs. There are therefore 16 possible units, or 10 if we allow strand symmetry (e.g. AA = TT). Trinucleotide models define the unit around the central base pair of a given trinucleotide. This yields 64 or 32 differ- ent units, again depending on whether or not strand symmetry is allowed. Static models are rigid rod models that only consider the static geometry of a segment. Curvature in B-DNA was originally believed to be a consequence of A n (n = 4–6) tracts that were repeated in phase with the helical repeat in DNA. Two static models were proposed ini- tially, to explain the phenomenon. In the nearest-neighbour model, the axial de- flections of successive AA/TT dinu- cleotides sum to produce a curve 4 (Fig. 1b). In the so-called junction model, cur- vature is produced at the junction be- tween the modified B-DNA structure (consisting of A n tracts) and adjacent unmodified B-DNA structure 5 . More re- cently, it became clear that DNA curva- ture also involves other sequence el- ements 6 , and more sophisticated models, which included different geometries for all 16 dinucleotides, were proposed 7,8 . Current static models consider dinu- cleotide geometries that are derived from direct measurements such as X-ray crystallography 9–12 or NMR 13,14 , from Rod models of DNA: sequence-dependent anisotropic elastic modelling of local bending phenomena Mircea G. Munteanu, Kristian Vlahovicek, Subbiah Parthasarathy, István Simon and Sándor Pongor Local bending phenomena can be predicted by elastic models that incor- porate sequence-dependent anisotropic-bendability (SDAB). SDAB models consider DNA to be an initially straight, segmented, elastic rod, in which the flexibility of each segment is greater towards the major groove than it is in other directions. While local bending can be predicted by static- geometry models as well, SDAB models, in addition, qualitatively explain such phenomena as the affinity of protein binding and kinking. A set of prediction tools is available at http://www.icgeb.trieste.it/dna M. G. Munteanu, K. Vlahovicek, S. Parthasarathy and S. Pongor are at the International Centre for Genetic Engineering and Biotechnology, Padriciano 99, 34012 Trieste, Italy; and I. Simon is at the Institute of Enzymology, BRC, Hungarian Academy of Sciences, H-1518 Budapest, PO Box 7, Hungary. Email: pongor@icgeb.trieste.it v v