A Self-Consistent Formulation for Analysis and Generation of Non-Uniform DNA Structures

Abstract

A fully self-consistent formulation is described here for the analysis and generation of base-pairs in non-uniform DNA structures, in terms of various local parameters. It is shown that the internal “wedge parameters” are mathematically related to the parameters describing the base-pair orientation with respect to an external helix axis. Hence any one set of three translation and three rotation parameters are necessary and sufficient to completely describe the relative orientation of the base-pairs comprising a step (or doublet). A general procedure is outlined for obtaining an average or global helix axis from the local helix axes for each step. A graphical representation of the local helix axes in the form of a polar plot is also shown and its application for estimating the curvature of oligonucleotide structures is illustrated, with examples of both A and B type structures.     

Base sequence and helix structure variation in B and A DNA

The observed propeller twist in base-pairs of crystalline double-helical DNA oligomers improves the stacking overlap along each individual helix strand. But, as proposed by Calladine, it also leads to clash or steric hindrance between purines at adjacent base-pairs on opposite strands of the helix. This clash can be relieved by: (1) decreasing the local helix twist angle between base-pairs; (2) opening up the roll angle between base-pairs on the side on which the clash occurs; (3) separating purines by sliding base-pairs along their long axes so that the purines are partially pulled out of the stack (leading to equal but opposite alterations in main-chain torsion angle delta at the two ends of the base-pair); and (4) flattening the propeller twist of the offending base-pairs. Simple sum functions, sigma 1 through sigma 4, are defined, by which the expected local variation in helix twist, base roll angle, torsion angle delta and propeller twist may be calculated from base sequence. All four functions are quite successful in predicting the behavior of B DNA. Only the helix twist and base roll functions are applicable to A DNA, and the helix twist function begins to fail for an A helical RNA/DNA hybrid. Within these limits, the sequence-derived sum functions match the observed helix parameter variation quite closely, with correlation coefficients greater than 0.900 in nearly all cases. Implications of this sequence-derived helix parameter variation for repressor-operator interactions are considered.     

Crystallographic study of one turn of G/C-rich B-DNA

The DNA decamer d(CCAGGCCTGG) has been studied by X-ray crystallography. At a nominal resolution of 1.6 A, the structure was refined to R = 16.9% using stereochemical restraints. The oligodeoxyribonucleotide forms a straight B-DNA double helix with crystallographic dyad symmetry and ten base-pairs per turn. In the crystal lattice, DNA fragments stack end-to-end along the c-axis to form continuous double helices. The overall helical structure and, notably, the groove dimensions of the decamer are more similar to standard, fiber diffraction-determined B-DNA than A-tract DNA. A unique stacking geometry is observed at the CA/TG base-pair step, where an increased rotation about the helix axis and a sliding motion of the base-pairs along their long axes leads to a superposition of the base rings with neighboring carbonyl and amino functions. Three-center (bifurcated) hydrogen bonds are possible at the CC/GG base-pair steps of the decamer. In their common sequence elements, d(CCAGGCCTGG) and the related G.A mismatch decamer d(CCAAGATTGG) show very similar three-dimensional structures, except that d(CCAGGCCTGG) appears to have a less regularly hydrated minor groove. The paucity of minor groove hydration in the center of the decamer may be a general feature of G/C-rich DNA and explain its relative instability in the B-form of DNA.     

Inosine.adenine base pairs in a B-DNA duplex.

The structure of the synthetic deoxydodecamer d(C-G-C-I-A-A-T-T-A-G-C-G) has been determined by single crystal X-ray diffraction techniques at 2.5A resolution. The refinement converged with a crystallographic residual, R = 0.19 and the location of 64 solvent molecules. The sequence crystallises as a B-DNA helix with 10 Watson-Crick base-pairs (4 A.T. and 6 G.C) and 2 inosine.adenine (I.A) pairs. The present work shows that in the purine.purine base-pairs the adenine adopts syn orientation with respect to the furanose moiety while the inosine is in the trans (anti) orientation. Two hydrogen bonds link the I.A. base-pair, one between N-1(I) and N-7(A), the other between O-6(I) and N-6(A). This bulky purine.purine base-pair is incorporated in the double helix at two positions with little distortion of either local or global conformation. The pairing observed in this study is presented as a model for I.A base-pairs in RNA codon-anticodon interactions and may help explain the thermodynamic stability of inosine containing base-pairs. Conformational parameters and base stacking interactions are presented and where appropriate compared with those of the native compound, d(C-G-C-G-A-A-T-T-C-G-C-G) and with other studies of oligonucleotides containing purine.purine base-pairs.     

Structure of the B-DNA decamer C-C-A-A-C-G-T-T-G-G and comparison with isomorphous decamers C-C-A-A-G-A-T-T-G-G and C-C-A-G-G-C-C-T-G-G

The crystal structure of the DNA decamer C-C-A-A-C-G-T-T-G-G has been solved to a resolution of 1.4 A, and is compared with the 1.3 A structure of C-C-A-A-G-A-T-T-G-G and the 1.6 A structure of C-C-A-G-G-C-C-T-G-G. All three decamers crystallize isomorphously in space group C2 with five base-pairs per asymmetric unit, and with decamer double helices stacked atop one another along the c axis in a manner that closely approximates a continuous B helix. This efficient stacking probably accounts for the high resolution of the crystal data. Comparison of the three decamers reveals the following. (1) Minor groove width is more variable than heretofore realized. Regions of A.T base-pairs tend to be narrower than average, although two successive A.T base-pairs alone may not be sufficient to produce narrowing. The minor groove is wider in regions where BII phosphate conformations are opposed diagonally across the groove. (2) Narrow regions of minor groove exhibit a zig-zag spine of hydration, as was first seen in C-G-C-G-A-A-T-T-C-G-C-G, whereas wide regions show two ribbons of water molecules down the walls, connecting base edge N or O with sugar O-4' atoms. Regions of intermediate groove width may accommodate neither pattern of hydration well, and may exhibit a less regular pattern of hydration. (3) Base-pair stacking is virtually identical at equivalent positions in the three decamers. The unconnected step from the top of one decamer helix to the bottom of the next helix is a normal helix step in all respects, except for the absence of connecting phosphate groups. (4) BII phosphate conformation require the unstacking of the two bases linked by the phosphate, but do not necessarily follow as an inevitable consequence of unstacking. They have an influence on minor groove width as noted in point (1) above. (5) Sugar ring pseudorotation P and main-chain torsion angle delta show an excellent correlation as given by the equation: delta = 40 degrees cos (P + 144 degrees) + 120 degrees. Although centered around C-2'-endo, the conformations in these B-DNA helices are distributed broadly from C-3'-exo to O-4'-endo, unlike the tighter clustering around C-3'-endo observed in A-DNA oligomer structures.     

Analysis of local helix geometry in three B-DNA decamers and eight dodecamers

Local variations in B-DNA helix structure are compared among three decamers and eight dodecamers, which contain examples of all ten base-pair step types. All pairwise combinations of helix parameters are compared by linear regression analysis, in a search for internal relationships as well as correlations with base sequence. The primary conclusions are: (1) Three-center hydrogen bonds between base-pairs occur frequently in the major groove at C-C, C-A, A-A and A-C steps, but are less convincing at C-C and C-T steps in the minor groove. The requirements for large base-pair propeller are (1) that the base-pair should be A.T rather than G.C, and (2) that it be involved in a major groove three-center hydrogen bond with the following base-pair. Either condition alone is insufficient. Hence, a large propeller is expected at the leading base-pair of A-A and A-C steps, but not at A-T, T-A, C-A or C-C steps. (2) A systematic and quantitative linkage exists between helix variables twist, rise, cup and roll, of such strength that the rise between base-pairs can hardly be described as an independent variable at all. Two typical patterns of behavior are observed at steps from one base-pair to the next: high twist profile (HTP), characterized by high twist, low rise, positive cup and negative roll, and low twist profile (LTP), marked by low twist, high rise; negative cup and positive roll. Examples of HTP are steps G-C, G-A and Y-C-A-R, where Y is pyrimidine and R is purine. Examples of LTP steps are C-G, G-G, A-G and C-A steps other than Y-C-A-R. (3) The minor groove is especially narrow across the two base-pairs of the following steps: A-T, T-A, A-A and G-A. (4) In general, base step geometry cannot be correlated solely with the bases that define the step in question; the two flanking steps also must be taken into account. Hence, local helix structure must be studied in the context, not of two base-pairs: A-B, but of four: x-A-B-y. Calladine's rules, although too simple in detail, were correct in defining the length of sequence over which a given perturbation is expressed. Whereas ten different two-base steps are possible, allowing for the identity of complementary sequences, there are 136 different four-base steps. Only 33 of these 136 four-base steps are represented in the decamer and dodecamer structures solved to date, and hence it is premature to try to set up detailed structural algorithms. (5) The sugar-phosphate backbone chains of B-DNA place strong limits on sequence-induced structural variation, damping down most variables within four or five base-pairs, and preventing purine-purine anti-anti mismatches from causing bulges in the double helix. Hence, although short-range sequence-induced deformations (or deformability) are observed, long-range deformations propagated down the helix are not to be expected.     

Binding stoichiometry and structure of RecA-DNA complexes studied by flow linear dichroism and fluorescence spectroscopy. Evidence for multiple heterogeneous DNA co-ordination

The interaction between RecA and DNA (in the form of unmodified single-stranded DNA, fluorescent single-stranded DNA and double-stranded DNA) is studied with linear dichroism and fluorescence spectroscopy. RecA is found to form a complex with single-stranded DNA with a binding stoichiometry of about four nucleotides per RecA monomer, in which the DNA bases appear to have a random orientation. Addition of ATP gamma S (a non-hydrolyzable analog of ATP) reduces the stoichiometry to about three nucleotides per RecA and causes the DNA bases to adopt an orientation preferentially perpendicular to the fiber axis. This complex can incorporate an additional strand of single-stranded DNA or double-stranded DNA, yielding a total stoichiometry of six nucleotides or three nucleotides and three base-pairs, respectively, per RecA. RecA, in the presence of ATP gamma S, is also found to interact with double-stranded DNA, with a stoichiometry of about three base-pairs per RecA. In all studied complexes, the tryptophan residues in the RecA protein are oriented with their planes preferentially parallel to the fiber axis, whereas in complexes involving ATP gamma S the planes of the DNA bases are oriented preferentially perpendicular to the fiber. This virtually excludes the possibility that the tryptophan residues are intercalated in the DNA helix. On the basis of these results, a model for the research of homology in the RecA-mediated, strand-exchange reaction in the genetic recombination process is proposed.     

Orientation relaxation of DNA restriction fragments and the internal mobility of the double helix

The orientation relaxation of 15 DNA restriction fragments (43-4361 base-pairs) is characterized by measurements of linear dichroism using high electric field pulses. The off-field relaxation of fragments of 84 base-pairs or less can be described by single exponentials, which are related to the transverse rotational diffusion of the helix. Fragments of 95 base-pairs or greater exhibit an additional fast component with time constants around 100 ns for fragments of approx. 100 base-pairs, increasing with chain length to about 700 ns for a fragment with 258 base-pairs. The amplitude of this process increases from virtually zero at low fields (approximately equal to 10 kV) to a substantial limit contribution at high fields. According to these results, we suggest that electric fields induce stretching of the DNA fragments from a weakly bent to a more straight form and that the fast component reflects the internal mobility of the DNA chain. The slow off-field components of the orientation are discussed in terms of different models. The data up to helix lengths of about 400 base-pairs can be described by the 'weakly bending rod' model from Hearst using 3.4 A rise per base-pair and 13 A axial radius of the helix. Both the weakly bending rod according to Hearst and the 'wormlike chain' according to Hagerman and Zimm provide a persistence length of 500 A. The on-field relaxation is slower than the corresponding off-field process at low field strengths, but the on-field process is accelerated substantially at high electric fields. These observations are compared with model calculations of Schwarz.     

Refinement of the solution structure of the DNA hexamer 5'd(GCATGC)2: combined use of nuclear magnetic resonance and restrained molecular dynamics

The solution structure of the self-complementary DNA hexamer 5'd(GCATGC)2 comprising the specific target site for the restriction endonuclease Sph 1 is investigated by using nuclear magnetic resonance spectroscopy and restrained molecular dynamics. All the nonexchangeable proton resonances are assigned sequentially, and from time-dependent nuclear Overhauser enhancement measurements a set of 158 approximate interproton distances are determined. These distances are used as the basis of a structure refinement using restrained molecular dynamics in which the interproton distances are incorporated into the total energy function of the system in the form of an effective potential term. Two restrained molecular dynamics simulations are carried out, starting from classical B- and A-DNA [atomic root mean square (rms) difference 3.3 A]. In both cases convergence is achieved to essentially identical structures satisfying the experimental restraints and having a root mean square difference of only 0.3 A between them, which is within the rms fluctuations of the atoms about their average positions. These results suggest that the restrained molecular dynamics structures represent reasonable approximations of the solution structure. The converged structures are of the B type and exhibit clear sequence-dependent variations of helical parameters, some of which follow Calladine's rules and can be attributed to the relief of interstrand purine-purine clash at adjacent base pairs. In addition, the converged restrained dynamics structures appear bent with a radius of curvature of approximately 20 A. This bending appears to be due almost entirely to the large positive base roll angles, particularly at the Pyr-Pur steps. Further, the global and local helix axes are not coincident, and the global helix axis represents a superhelical axis which the bent DNA, when extended into an "infinite" helix by repeated translation and rotation, wraps around.     

Impact of static and dynamic A-form heterogeneity on the determination of RNA global structural dynamics using NMR residual dipolar couplings

We examined how static and dynamic deviations from the idealized A-form helix propagate into errors in the principal order tensor parameters determined using residual dipolar couplings (rdcs). A 20-ns molecular dynamics (MD) simulation of the HIV-1 transactivation response element (TAR) RNA together with a survey of spin relaxation studies of RNA dynamics reveals that pico-to-nanosecond local motions in non-terminal Watson-Crick base-pairs will uniformly attenuate base and sugar one bond rdcs by approximately 7%. Gaussian distributions were generated for base and sugar torsion angles through statistical comparison of 40 RNA X-ray structures solved to <3.0 A resolution. For a typical number (>or=11) of one bond C-H base and sugar rdcs, these structural deviations together with rdc uncertainty (1.5 Hz) lead to average errors in the magnitude and orientation of the principal axis of order that are <9% and <4 degrees, respectively. The errors decrease to <5% and <4 degrees for >or=17 rdcs. A protocol that allows for estimation of error in A-form order tensors due to both angular deviations and rdc uncertainty (Aform-RDC) is validated using theoretical simulations and used to analyze rdcs measured previously in TAR in the free state and bound to four distinct ligands. Results confirm earlier findings that the two TAR helices undergo large changes in both their mean relative orientation and dynamics upon binding to different targets.     

Super Secondary Structure Consisting of a Polyproline II Helix and a β-Turn in Leucine Rich Repeats in Bacterial Type III Secretion System Effectors

Leucine rich repeats (LRRs) are present in over 100,000 proteins from viruses to eukaryotes. The LRRs are 20–30 residues long and occur in tandem. LRRs form parallel stacks of short β-strands and then assume a super helical arrangement called a solenoid structure. Individual LRRs are separated into highly conserved segment (HCS) with the consensus of LxxLxLxxNxL and variable segment (VS). Eight classes have been recognized. Bacterial LRRs are short and characterized by two prolines in the VS; the consensus is xxLPxLPxx with Nine residues (N-subtype) and xxLPxxLPxx with Ten residues (T-subtype). Bacterial LRRs are contained in type III secretion system effectors such as YopM, IpaH3/9.8, SspH1/2, and SlrP from bacteria. Some LRRs in decorin, fribromodulin, TLR8/9, and FLRT2/3 from vertebrate also contain the motifs. In order to understand structural features of bacterial LRRs, we performed both secondary structures assignments using four programs—DSSP-PPII, PROSS, SEGNO, and XTLSSTR—and HELFIT analyses (calculating helix axis, pitch, radius, residues per turn, and handedness), based on the atomic coordinates of their crystal structures. The N-subtype VS adopts a left handed polyproline II helix (PPII) with four, five or six residues and a type I β-turn at the C-terminal side. Thus, the N-subtype is characterized by a super secondary structure consisting of a PPII and a β-turn. In contrast, the T-subtype VS prefers two separate PPIIs with two or three and two residues. The HELFIT analysis indicates that the type I β-turn is a right handed helix. The HELFIT analysis determines three unit vectors of the helix axes of PPII (P), β-turn (B), and LRR domain (A). Three structural parameters using these three helix axes are suggested to characterize the super secondary structure and the LRR domain.     

DNA bending and its relation to nucleosome positioning

X-ray and solution studies have shown that the conformation of a DNA double helix depends strongly on its base sequence. Here we show that certain sequence-dependent modulations in structure appear to determine the rotational positioning of DNA about the nucleosome. Three different experiments are described. First, a piece of DNA of defined sequence (169 base-pairs long) is closed into a circle, and its structure examined by digestion with DNAase I: the helix adopts a highly preferred configuration, with short runs of (A, T) facing in and runs of (G, C) facing out. Secondly, the same sequence is reconstituted with a histone octamer: the angular orientation around the histone core remains conserved, apart from a small uniform increase in helix twist. Finally, it is shown that the average sequence content of DNA molecules isolated from chicken nucleosome cores is non-random, as in a reconstituted nucleosome: short runs of (A, T) are preferentially positioned with minor grooves facing in, while runs of (G, C) tend to have their minor grooves facing out. The periodicity of this modulation in sequence content (10.17 base-pairs) corresponds to the helix twist in a local frame of reference (a result that bears on the change in linking number upon nucleosome formation). The determinants of translational positioning have not been identified, but one possibility is that long runs of homopolymer (dA) X (dT) or (dG) X (dC) will be excluded from the central region of the supercoil on account of their resistance to curvature.     

Refined crystal structure of an octanucleotide duplex with G . T mismatched base-pairs

Single crystal X-ray diffraction techniques have been used to determine the structure of the DNA octamer d(G-G-G-G-C-T-C-C) at a resolution of 2.25 A. The asymmetric unit consists of two strands coiled about each other to produce an A-type DNA helix. The double helix contains six G . C Watson-Crick base-pairs and two G . T mismatched base-pairs. The mismatches adopt a "wobble" type structure in which both bases retain their major tautomer forms. The double helix is able to accommodate this G . T pairing with little distortion of the overall helical conformation. Crystals of this octamer melt at a substantially lower temperature than do those of a related octamer also containing two G . T base-pairs. We attribute this destabilization to disruption of the hydration network around the mismatch site combined with changes in intermolecular packing. Full details are given of conformational parameters, base stacking, intermolecular contacts and hydration involving 52 solvent molecules.     

Crystal structure analysis of an A-DNA fragment at 1.8 A resolution: d(GCCCGGGC).

Single crystals of the self-complementary octadeoxyribonucleotide d(GCCCGGGC) have been analysed by X-ray diffraction methods at a resolution of 1.8 A. The tetragonal unit cell of space group P4(3)2(1)2 has dimensions of a = 43.25 A and c = 24.61 A and contains eight strands of the oligonucleotide. The structure was refined by standard crystallographic techniques to an R factor of 17.1% using 1359 3 sigma structure factor observations. Two strands of the oligonucleotide are related by the crystallographic dyad axis to form a DNA helix in the A conformation. The d(GCCCGGGC) helix is characterized by a wide open major groove, a near perpendicular orientation of base pairs to the helix axis and an unusually small average helix twist angle of 31.3 degrees indicating a slightly underwound helix with 11.5 base pairs per turn. Extensive cross-strand stacking between guanine bases at the central cytosine-guanine step is made possible by a number of local conformational adjustments including a fully extended sugar-phosphate backbone of the central guanosine nucleotide.     

The energy of formation of internal loops in triple-helical collagen polypeptides

The sequence-dependent local destabilization in the interior of the collagen triple helix has been evaluated by means of conformational energy computations. Using a model poly(Gly-Pro-Pro) triple helix as the reference state, a method was developed for generating local loops, i.e., internal deformations, and analyzing their conformations. A seven-residue Gly-Pro-Pro-Gly-Pro-Pro-Gly fragment was replaced by the Gly-Pro-Ala-Gly-Ala-Ala-Gly sequence in one, two, or all three of the strands of the loop region. A set of loop conformations was generated in which the ends of the loop were initially fixed in the triple-helical structure. The potential energy of the entire deformed triple helix was then minimized, resulting in a variety of structures that contained deformed loops. The conformations of the triple helices at the two ends of the loops remained essentially unchanged in many of the low-energy conformations. In numerous high-energy conformations, however, the triple-helical segments were also partially or totally disrupted. The minimum-energy conformations of the whole structures were compared in terms of rms deviations of atomic coordinates with respect to the original triple helix, and of the shapes of the loops (using a distance function derived from differential geometry). Three new geometrical parameters—stretch S, kink K, and unwinding U—were defined to describe the changes in the overall orientation of the triple helices at the two ends of the loop. It is shown that, when the number of Pro residues in a short fragment is reduced, the triple helical structure can accomodate internal loops (i.e., distortions) within a 5 kcal/mol cutoff from the essentially unperturbed triple helical structure. For structures with a Gly-Pro-Ala-Gly-Ala-Ala-Gly sequence in all three strands, the probability of finding conformations with internal loops is small, i.e., 0.06. Internal loops affect the overall orientation of these structures, as measured by the helix-distortion parameters S, K, and U. © 1995 John Wiley & Sons, Inc.     

Definitions and analysis of DNA Holliday junction geometry

A number of single-crystal structures have now been solved of the four-stranded antiparallel stacked-X form of the Holliday junction. These structures demonstrate how base sequence, substituents, and drug and ion interactions affect the general conformation of this recombination intermediate. The geometry of junctions had previously been described in terms of a specific set of parameters that include: (i) the angle relating the ends of DNA duplexes arms of the junction (interduplex angle); (ii) the relative rotation of the duplexes about the helix axes of the stacked duplex arms (J_roll); and (iii) the translation of the duplexes along these helix axes (J_slide). Here, we present a consistent set of definitions and methods to accurately calculate each of these parameters based on the helical features of the stacked duplex arms in the single-crystal structures of the stacked-X junction, and demonstrate how each of these parameters contributes to an overall conformational feature of the structure. We show that the values for these parameters derived from global rather than local helical axes through the stacked bases of the duplex arms are the most representative of the stacked-X junction conformation. In addition, a very specific parameter (J_twist) is introduced which relates the relative orientation of the stacked duplex arms across the junction which, unlike the interduplex angle, is length independent. The results from this study provide a general means to relate the geometric features seen in the crystal structures to those determined in solution.     

Interaction of paired homologous series of diacridines and triacridines with deoxyribonucleic acid

Viscometric measurements using covalently closed circular DNA and sonicated rod-like DNA fragments were performed to investigate unwinding and extension of the DNA helix associated with binding of paired homologous series of diacridines and triacridines. The maximum interchromophore distance for members of the diacridine series spans from 15.1 to 27.5 A, permitting the largest of these ligands to cover up to 4 or 5 base-pairs, allowing for helical twist and local unwinding in a bisintercalated complex lacking severe bending or kinking of the DNA backbone. Helix unwinding angles and increments in DNA contour length are characteristic of bifunctional reaction for all the diacridines studied, the DNA lattice appearing to saturate with one ligand molecule bound per 4 base-pairs. The triacridines, whose maximum end-to-end interchromophore distances are the same as those of their paired diacridines, have maximum central-to-terminal interchromophore distances covering the range 7.5-13.8 A and thus have the potential to form trisintercalated complexes with one or two base-pairs sandwiched between each chromophore. However, helix extension and unwinding parameters for the triacridines are similar to those of the diacridines, and we find no evidence of a transition from bifunctional to trifunctional reaction as the homologous series is ascended. In general, the binding site size appears to be 5 base-pairs for the triacridines. The stereochemical requirements for trisintercalation of triacridines are discussed with reference to the present findings and to the work of others.     

Structure of the pyrimidine-rich internal loop in the poliovirus 3'-UTR: the importance of maintaining pseudo-2-fold symmetry in RNA helices containing two adjacent non-canonical base-pairs

Formation of non-canonical base-pairs in RNA often plays a very important functional role. In addition they frequently serve as factors in stabilizing the secondary structure elements that provide the frame of large compact RNA structures. Here we describe the structure of an internal loop containing a 5'CU3'/5'UU3' non-canonical tandem base-pair motif, which is conserved within the 3'-UTR of poliovirus-like enteroviruses. Structural details reveal striking regularities of the local helix geometry, resulting from alternating geometrical adjustments, which are important for understanding and predicting stabilities and configurations of tandem non-canonical base-pairs. The C-U and U-U base-pairs severely contract the minor groove of the sugar-phosphate backbone, which might be important for protein recognition or binding to other RNA elements.     

A refined prediction method for gel retardation of DNA oligonucleotides from dinucleotide step parameters: reconciliation of DNA bending models with crystal structure data

The development and assessment of a prediction method for gel retardation and sequence dependent curvature of DNA based on dinulcleotide step parameters are described. The method is formulated using the Babcock-Olson equations for base pair step geometry (1) and employs Monte Carlo simulated annealing for parameter optimization against experimental data. The refined base pair step parameters define a stuctural construct which, when the width of observed parameter distributions is taken into account, is consistent with the results of DNA oligonucleotide crystal structures. The predictive power of the method is demonstrated and tested via comparisons with DNA bending data on sets of sequences not included in the training set, including A-tracts with and without periodic helix phasing, phased A4T4 and T4A4 motifs, a sequence with a phased GGGCCC motif, some "unconventional" helix phasing sequences, and three short fragments of kinetoplast DNA from Crithidia fasiculata that exhibit significantly different behavior on non-denaturing polyacrylamide gels. The nature of the structural construct produced by the methodology is discussed with respect to static and dynamic models of structure and representations of bending and bendability. An independent theoretical account of sequence dependent chemical footprinting results is provided. Detailed analysis of sequences with A-tract induced axis bending forms the basis for a critical discussion of the applicability of wedge models,junction models and non A-tract, general sequence models for understanding the origin of DNA curvature at the molecular level.