J. Mol. Biol. (1992) 227, 738-756

Crystal and Molecular Structure of the A-DNA Dodecamer d(CCGTACGTACGG) Choice of Fragment Helical Axis Craig A. Bingman1s2, Gerald Zon3 and M. Sundaralingam1~2t ‘Crystallography Laboratory, Department of Biochemistry University of Wisconsin, Madison, WI 53706, U.S.A. ‘Laboratory of Biological Macromolecular Structure Department of Chemistry and Biotechnology Center, Ohio State University Rightmire Hall, 1060 Carmack Road, Columbus, OH 43210-1002, U.S.A. ‘Applied

Biosystems, 850 Lincoln Center Drive Foster City, CA 94404, [J.S.A.

(Received 2 January

1992; accepted 26 May

1992)

The crystal structure of the dodecamer d(CCGTACGTACGG) has been determined at 2.5 A resolution. The crystals grow in the hexagonal space group P6,22, a = b = 46.2 A, c = 71.5 A with one strand as the asymmetric unit. Diffraction data were collected by the oscillation film method yielding 1664 unique reflections with an Rmergeof 904. The structure was solved by real-space rotationa translational searches with idealized helical models of A, R and Z-DNA. The best agreement was given by an A-DNA model with its dyad axis along bhe diagonal crystallographic dyad axis, with an R-factor 643 and correlation coefficient of 959 for data between 10 and 5 A. Iterative map fitting and restrained least-squares refinement and addition of 40 solvent molecules brought the R-factor t,o 615 and the correlation coefficient to 0.97 for all data between 8.0 and 2.5 A. The stereochemistry of the atomic model is good, with a root-mean-square deviation in bond distances of 0006 A. This is the first example of an A-DNA containing a full helical turn. The dodecamer displays a novel packing motif. In addition to the characteristic contacts between the terminal base-pairs and the minor grooves of symmetry-related molecules, there are also minor groove to minor groove interactions not previously observed. The packing leaves an approximately 25 A diameter solvent channel around the origin, along the c-axis. The presence of a prominent 3.4 A meridional reflection and other diffuse features in the diffraction pattern provided evidence for the presence of disordered B-DNA along the c-axis, which can be accommodated in these solvent channels. The molecular conformation of the dodecamer also displays novel features. The dyad-related halves of the molecule are bent at an angle of 20”. and the helical parameters are affected by this bend. Unlike the shorter A-DNA octamers, the dimensions of the major groove can be directly measured. Novel correlations between local helical parameters and global conformational features are presented. Most of the solvent molecules are associated with the major groove and the sugar-phosphate backbone. Keywords: A-DNA

dodecamer; alternating right-handed DNA; kinked DXA: major groove width; fragment helical axis

turn, we prepared the alternating Pur-Pur sequence d(GCGTACGTACGC) and the companion sequence d(CCGTACGTACGG) with a double Pyr at the 5’ start. The addition of a 5’.terminal purine to small oligomers strongly disfavors the formation of Z-DNA (Quadrifogilo et al., 1984; Jain & Sundaralingam, 1987). Since the dodecamer with the double pyrimidine start gave better crystals and intensities, its structure was first determined.

1. Introduction We showed in our earlier work that the alternating Pyr-Pur decamer d(CGTACGTACG) crystallized in the expected left-handed Z-DNA form (Brennan & Sundaralingam, 1985). In extending the sequence to a dodecamer to study a full-turn helical t Author to whom all correspondence should be addressed. 738 0022-2836/92/190738-19

$08.00/O

0 1992 Academic

Press Limited

Structure

of A-DNA

2. Methods (a) Synthesis, crystallization

and characterization

of crystals

The dodecamer d(CCGTACGTACGG) was synthesized by the phosphoramidite method both at Applied Biosystems and in our labortory, and was purified under denaturing conditions by reverse-phase chromatography with a Hamilton PRP-1 column (Zon & Thompson, 1986). Both products gave the same type of crystals by vapor diffusion against a wide range of 1-propanol concentrations, but only in the presence of cobaltic hexammine and spermine. It is noteworthy that cobaltic hexammine has been used to stabilize Z-DNA (for a review, see Jovin et al., 1983), and in the decamer d(CGTACGTACGT) cobaltic hexammine was found in the crystal (Brennan et al.. 1986). The dodecamer crystals were grown from a solution containing 3.5 mM-DNA duplex, 12 mM-cobaltic hexammine and 5 mM-spermine buffered with 10 mM-sodium cacodylate, pH 65. Preliminary precession photographs of the h0Z and hhl zones of the dodecamer d(CCGATACGTACGG) were taken using an Elliot GX 6 rotating anode in a cold room at 4°C. The unit cell constants obtained from the precession photographs were a = b = 462 A, c = 71.5 A (1 A = 91 nm), a = /l = 90, y = 120, with a space group of either P6,22 or P6,22. The crystals were stable at 4°C for at least 3 days in a nickel-filtered beam of a rotatinganode source operating at 40 kV, 40 mA. Oscillation and precession photographs showed, in addition to the sharp Bragg reflections, a strong diffuse 34 A meridional reflection along the c-axis and a cross-like pattern characterisbic of B-DNA fibers. However, the Bragg reflections were not especially intense around the diffuse meridional reflection. The cell constants, symmetry and diffraction pattern were nearly identical to those of the companion dodecamer d(GCGTACGTACGC) (S. Jain, unpublished results). (b) Data collection Attempts to collect the data on a CAD-4 diffractometer failed, both at room temperature and using a Nonius cryogenic cooling system to maintain the crystals at approximately 4°C. In both cases, rapid decay of the

Dodecamer diffraction pattern and obvious physical deterioration of the specimen occurred before data could be collected. It was necessary to collect the diffraction data in the cold room by rotation photography (Arndt & Wonacott, 1977). The crystals were sufficiently stable to collect the entire intensity set from 1 crystal. Each of the 8 film cassettes of the Enraf Nonius oscillation camera was packed with 3 films (Kodak DEF-5) and exposed over a 3.5” angular range at 4000 s/deg. with 8% overlap. A total of 17 packs were collected, over a range of 55”, well over the unique part of the diffraction pattern of 30”. The films were processed, digitized and scanned on an Optronics Film scanner using a 100 pm raster. Intensity data were extracted from the digitized images and the 3 films within the packs were scaled using the program DENZO (Otwinowski, et al., 1988). The data from different packs were merged together with the program SCALEPACK, written in house by S. T. Rao (unpublished results). Altogether, there were 4971 reflections processed with an Rmerge of 904, yielding 1664 independent reflections greater

than

2a.

(c) Structure solution The structure was solved by molecular replacement using a real-space, brute-force structure-factor calculation program written in house by S. T. Rao & S. Jain (unpublished results), which had been successfully employed earlier in the solution of the tetragonal form of d(GTGTACAC) (Jain et al., 1989). The dodecamer duplexes were constrained to lie on 1 of the 2 classes of crystallographic 2-fold axes in the space groups P6,22 or P6,22 on the basis of the volume occupied by the asymmetric unit (Table 1). The most tightly packed oligonucleotide crystal was d(ATAT) (Viswamitra et al., 1978) with a volume of 986 A3 per base-pair. For the dodecamer structure a model having a duplex as the asymmetric unit would have a volume of 918A3 per base-pair. Such a low volume per base-pair in the crystal would indicate that the dodecamer was packed more tightly than any previously described oligonucleotide crystals, even Z-DKA crystals. If only a single strand of the dodecamer duplex is in the asymmetric unit, then the volume occupied would be 1836 A3 per base-pair, although this value is considerably

Table 1 Lattice volume per base-pair

Type Alternating duplext A-DNA A-DNA A-DNA A-DNA A-DNA A-DNA A-DNA/RNA B-DNA B-DNA Z-DNA Z-DNA

Sequence pATAT ‘CCGG GGCCGGCC GGBrUAB’UACC GTGTACAC ACCGGCCGGT CCGTACGTACGG r(GCG)d(TATACGC) CGCGAATTCGCG CCAAGATTGG CGCGCG CGCG

1, Viswamithra et al. (1978). 2, Dickerson & Drew (1981). 3, Wang et al. (1982a). 4, Shakked et al. (1981). 5, Jain & Sundaralingam (1989). 6, Frederick et al. (1989). 7 Two AT dinucleotide helical segments with alternating

in selected oligonucleotide Space group

p2, P4,2,2 P4,2,2

P’5, P6,22 P6,22 P6,22

p212121 p21212, c2 P2,2,2, c222,

crystals

Volume/base-pair (A7 986 1409 1391 1528 1561 1733 1836 1299 1253 1278 1048 1232

7, C. A. Binzman (unpublished 8, Wang et 2. (1982bf 9, Drew et al. (1982). 10, Prive et al. (1987). 11, Wang et al., 1979). 12, Drew et al. (1980). C-%endo and C-%endo sugars.

Reference

2 3 4 5 6 7 8 9 10 11 12 results)

C. A. Bingman et al.

740

(a 1

Figure 1. Sim-weighted 3F0-2Fc electron density maps of the region of the primary lattice contact in d(CCGTACGTACGG), before (a) and after (b) refinement, computed with data between 10 and 2.5 A. This was the region in which close intermolecular contacts were noted in the initial packing analysis of the struct,ure solution. A skeletal representation of the starting model is shown by broken lines. the refined model in heavy continuous lines, and the symmetry-related molecule in light continuous lines. The density for the terminal base-pairs 1s clearly shifted toward the refined structure in the first map, even though this region was included in the phasing model (broken lines). In the final map, after refinement (b), the terminal base-pair has been omitted from the phasing model. and positive density returns for this base-pair. higher than the values shown in Table 3 (below) it appears consistent with the fact that the diffraction pattern from these crystals extended only to 2.5 A, indicating a rather loose packing. Two-dimensional searches were carried out on each of the 2 distinct types of 2-fold axis with A, B and Z-DNA models constructed with idealized co-ordinates of the appropriate sequence (Arnott & Hukins, 1972, 1973; Arnott et al., 1980; Wang et al., 1979). The search was performed in both the enantiomeric space groups P6,22 or P6,22 with a 5” and 1 A grid. using only the data between 10 and 5 A. The highest correlation coefficient was found with an A-DNA model in P6,22 with the helical axis inclined 35” with respect to the a. b plane on t,he class of 2-fold axes passing through the short diagonal of the unit cell, at c = l/12. The correlation coefficient of this model was 0.59, with an R-factor of 0.431. A fine grid search of 2.5” and 0.25 A was used to position the model accurately. Inspection of this model by graphics showed that there were a few forbidden contacts involving the t,erminal base-pairs of one molecule and the sugar-phosphate backbone of a symmetry-related molecule. It is noteworthy that all A-DNA crystals described to date have precisely this type of contact, between the terminal base-pair and the narrow groove sugar-phosphate backbone without

being unduly short. This solution was further verified by calculation of a Sim-weighted 3F,-2F, map. The contact region of this map is shown in Fig. 1. The short contacts between symmetry-related molecules appeared to be relieved if the end base-pairs are moved 1.5 a away from the initial position. as suggested by the electron densit) map. Such clear. interpretable electron density maps containing information not only about the accurately modeled parts of the molecule, but also about the more poorly modeled parts are the most important characteristic of a correct molecular replacement, model. The projection of the molecular packing on the ab plane is consistent with the Harker section at 11% = l/S. which essentially is a 2-dimensional projection of the electron density down the c axis. This provides additional confirmation of the correctness of the molecular replacement solution. (d)

Structure

rejinement

The dodecamer model was rebuilt int,o the 3Fo-%Fc map using the program FRODO. The correlation coefficient of the model increased to 075 after the fitting. The rebuilt model was then refined using the restrained leastsquares refinement program NUCLSQ (Hendrickson & Konnert, 1981; Westof et al., 1985). Initially. a constant

Structure of A-DNA

Dodecamer

741

Table 2 Restrained parameters for d(CCGTACGTACGG) Restraint

NO.

r.m.s. dev. (A)?

Sigma

No. of dev. > 2u

Distances(A) 456 672 88 524 252 72

0406

O-025

0

0016 0.027 0038 0.027 0057

O-050 0050

0 0

O-075 0040

0 0

0100

0

32 35

0.206

0090 0090

2 18

456 672 88 524

1.750 2650 1.790 2.670

3mO 4500 3900 4500

0 0 0 0

(A) parameter (A’)

243 243

@020 0210

O-200

0

5ooo

0

(A) parameter (A*)

40 40

0.066 0.670

@200

0

5000

0

Sugar, base bonds Sugar, base angles

Phosphate bonds Phosphate angles, hydrogen bonds Planarity (A) Chiral centers (A3) Non-bonded contacts (A)

Single torsion Multiple torsion

0089

Thermal parameter (A’)

Sugar, base bonds Sugar, base angles Phosphate bonds Phosphate angles, hydrogen bonds

Non-crystallographic

DZJA Position Thermal Solvent Position Thermal

symmetry

t r.m.s. root-mean-square

thermal parameter of 30 A* was used, and the data were gradually extended from 10 to 5 A in shells of approximately 001 sin #/,i until positional convergence was achieved before the next shell of data was added. After 78 cycles, all the data in the 10 to 25 A resolution range were included, giving an R-factor of 0.30. At this point, bhe model was refitted to “omit” electron density maps. Typically 2 residues were omitted from the phasing model and both 3F,-2F, and F,-F, maps were computed. Some solvent molecules were apparent in these maps. Next, the solvents were included if their peak centres were 2.0 to 3.5 A from polar DNA atom and the peak heights were at least 20. At this point, both positional and thermal parameters for DNA and solvent atoms were refined. After the final cycles of map fitting, including solvent atoms the R-factor dropped to @15 with 40 water molecules per strand. No ions could be unequivocally identified. The final refinement statistics for the structure of the final are given in Table 2. The stereochemistry model is quite good, as exemplified by the deviation in sugar and base bond lengths of 0906 A from the library values and 1” in bond angles.

3. Results and Discussion (a) Molecular packing and co-existence of B-DNA A view of the contents of the unit cell is shown in Figure 2. The DNA duplexes crowd around the 31 axis and leave wide continuous solvent channels centered at the 6, axis. Each molecule of the dodecamer contacts a total of six symmetry-related molecules. There are two interactions between the terminal base-pairs of the reference molecule at the center and the minor grooves of the other molecules, two interactions of the minor groove of the reference molecule with the terminal base-pairs of neighboring molecules, and two glancing minor groove-

minor groove interactions. The first kind of interaction involves the stacking of the terminal basepairs against the minor groove of a symmetryrelated molecule (Fig. 3(a)). This type of interaction is characteristic of crystalline A -DNA oligomers and has been found in every known single crystal structure of A-DNA. The nearly flat surface of the minor groove allows tight interactions with the surface of the nearly planar terminal base-pairs of symmetryrelated molecules. Within the dodecamer crystals, the interaction of the terminal base-pairs with the minor groove of a gl-related duplex stabilizes the crystals along the a, b plane (Fig. 3(b)). Besides the above contacts, the dodecamer crystals also display a second type of packing interaction that has not been previously observed. Molecules related along the c direction around t)he 3, axes contact each other with a novel symmetrical “glancing” interaction involving the minor grooves of two molecules (Fig. 3(c)). Overall, the apolar portion of the minor groove surface is largely buried by close interactions with symmetry-related molecules. It has been proposed that such close contacts stabilize not only the crystal lattice but also the A conformation of the DNA. Structure analysis using the Bragg reflections revealed that the ordered crystalline lattice was made up of A-DNA dodecamers. The 34 A diffuse meridional reflections and other diffuse reflections in the diffraction photographs indicated the presence of B-DNA, coexisting with the A-DNA. The packing of the A-DNA molecules leaves a central cylindrical cavity of 25 A diameter at the origin. These large voids could easily accommodate the B-DNA with the helical axis lying along the crystal c-axis. The absence of features above la in an

742

C.

A. Ringman

et al.

Figure 2. Contents of a single unit cell as viewed down the c-axis. The A-DNA dodecamers are clustered around the 3,-screw axes, and leave large voids along the 6,.axes. A molecule of B-DNA has been placed along one of the c-axes to illustrate that there is sufficient space for such disordered molecules at this position. Their presence within t,he rrystal has been inferred from the successful refinement of the A -DNA starting model: and the B-DNA fiber-like scatt,ering pattern from these crystals.

F, - F, map of this region indicates that the B-DNA molecules are highly disordered. This is not surprising, due to the intrinsic rotational disorder caused by placing an object with approximate lo,22 symmetry on a 6, axis. The volume occupied by a base-pair of the A-DNA lattice is 1836 A3. The value would be reduced to a limiting value of 1500 A3, if the solvent channels were completely filled by B-DNA. We have no reason to assign a specific occupancy to the B-DNA. The co-existence of disordered B-DNA within a lattice of A-DNA has been previously observed in the octamer d(GGB’UAB’UACC) (Doucet et al., 1989). (b) Overall molecular conformation and hydration The overall conformation of the dodecamer is presented in Figure 4. The major groove of the molecule is narrow compared to the expansive and shallow minor groove, as is typical of A-DNA. The backbone torsion angles of the dodecamer are given in Table 3. With the exception of the gamma torsion angle at the 5’ end, which is not constrained by additional 5’-helical nucleotides, all the torsion angles fall within the domains expected for A-DNA. There are no instances of the all-trans backbone geometry noted in other A-DNA structures (Wang et al., 1982a), in particular in the octamer d(GTACGTAC) Takusagawa, 1990), which has a sequence identical to the inner eight base-pairs of the dodecamer. All the deoxyribose sugars fall within the C-3’-endo domain. Although the backbone torsion angles are within the broad range associated with

the A-form, there are a large number of subtle departures in this structure from an ideal il-Dh’A helix. The overall pattern of the ordered watet molecules surrounding the dodecamer is shown in Figure 5. The deep major groove surface is covered with essentially a single layer of water molecules. For each strand there are 13 water molecules within 3.5 A of base polar atoms on the major groove side. Although there are many water-water hydrogen bonds linking the water molecules bound in the major groove, no regular motif such as the fused pentagonal network in d(GGB’UAB’UACC) (Kennard et al., 1986) was observed. Recently a 1.4 A resolution refinement of the hexagonal form of d(GTGTACAC) also showed no evidence of such a pentagonal network of water molecules major groove, so they are not an obligatory

in the feature

of the major groove hydration in A-DNA (Thota et aE.. 1992). There are only three water contacts per strand on the minor groove side of the bases. Large stretches of the aliphatic portion of the sugar-phosphate backbone are totally devoid of hydration, due in large part to interactions with symmetry-related molecules in the crystal lattice. There are 37 distances less than 3.5 A between the water

molecules

and polar

DNA

backbone

atoms

per strand. Of these, four involve O-4’ deoxyribose atoms, with distances in the range of 3.2 to 3.4 .A. longer

than

ideal

hydrogen-bonding

distIances.

There are five contacts with O-3’ atoms. four contacts with O-5’ atoms. seven contacts with O-II’ atoms, and 17 contacts with O-2P at,oms. Thus, as

Structure C$A-DNA

(a)

743

Dodecamer

(b)

Figure 3. Details of the packing motif. Each duplex is in contact with 6 neighboring duplexes within the crystal lattice. (a) Two symmetry-related molecules (light bonds) have their terminal base-pairs in contact with the minor groove of the reference duplex (heavy bonds). This type of interaction decreases the accessible surface area of the reference molecule by 215 AZ. (b) Two more molecules interact with the terminal base-pairs, which decreases the accessible surface area of the reference molecule by 223 A’. The types of packing interactions depicted in (a) and (b) are found in all A-DNA crystal structures described to date. (c) Additionally, 2 symmetry related duplexes participate in novel “glancing” interactions, which bring the minor groove into van der Waals’ contact with the minor grooves of 2 symmetry-related molecules. This type of interaction have not been previously described in an A-DNA crystal structure, but plays an important role in desolvating the minor groove surface of the dodecamer, decreasing the accessible surface area of the molecule by 300 A*. (d) All 3 types of interactions are displayed in stereo. The 2 lightly rendered molecules interact minor groove to minor groove with each other, and by their terminal base-pairs with the dark molecule.

expected, the phosphate anionic oxygens are more hydrated than other nucleic acid atoms (Kennard et al., 1986; Saenger et al., 1986). At 2.5 a resolution, it is difficult to discern the hydration shell surrounding the dodecamer with certainty. We therefore were cautious not to add too many water moIecules.

(c) Base stacking The base-stacking pattern of the dodecamer is important in determining its conformation. It is well known that for a 5’ purine-3’ pyrimidine step most of the stacking interactions occur between

bases in pyrimidine-3’ interactions

the

same strand,

whereas

in

a 5’

purine step most of the stacking involve the cross-strand purines. In

A-DNA, these stacking patterns are accentuated compared to B-DNA, due to the inclination of the base-pairs to the helical axis. The resultant basestacking interactions are shown in stereo pairs in Figure 6 and schematically in Figure 7. The schematic figure reveals that in the purely alternating

pyrimidine-purine

ten base-pair

core of the

dodecamer, this stacking pattern leads to three vertical “domains”. These vertical stacks of four bases are separated from neighboring stacks and contain two bases from each strand. This striking

(a)

(b)

Figure 4. Overall conformation of the dodecamer. (a) The helical axis is vertical and the dyad relating the 2 strands is in the plane of the page. (b) View down the dyad into the major groove. (c) View down the helical axis, showing the distinctive base-stacking pattern characteristic of A-DNA with an alternating purine-pyrimidine sequence.

Structure of A-DNA

Dodecamer

745

Table 3 Backbone torsion angles and sugar parameters St?pXlCt? Cl C2 G3 T4 A5 C6 GT T8 A9 Cl0 Gll G12

Alpha

Beta

Gamma

Delta

Epsilon

Zeta

Chi

P

288 301 282 275 330 264 316 284 289 304 285

161 162 185 181 168 174 168 164 176 176 191

139 77 77 61 79 30 82 35 73 6.5 47 63

84 74 84 78 79 84 80 78 78 82 81 79

222 221 212 195 217 204 212 206 218 209 191

278 272 291 283 273 286 280 282 282 281 292

190 187 193 205 201 212 194 205 196 193 197 205

16 28 26 24 14 13 27 18 17 10 23 11

stacking pattern results from the alternating base sequence in the center, the relative sizes of the two types of bases and the inclination of the base-pairs relative to the helix axis. The implications of the base-stacking pattern on the helical conformation of the dodecamer will be described in detail below. (d) Helical axis perturbation A superposition of the refined model on the starting model (Fig. 8) gave a root-mean-square deviation of @9 A, with the ends deviating more than 0.9 A and deflected towards the major groove side. This indicated that the dodecamer duplex was considerably “bent”. When local helical parameters (EMBO Workshop, 1989) of the dodecamer were calculated using NEWHEL91 (R. E. Dickerson, personal communication) they showed a pattern of pathological variation that seemed to be more dependent on location in the molecule, rather than base sequence. Most notable were roll angle, inclination, and the related base-pair buckle and dinucleotide cup angles. The dinucleotide role angles peaked at the center. Since a positive roll angle reflects an opening of adjacent base-pairs on the minor groove side, a large roll angle at the center would deflect the ends of the molecule in the observed direction. The inclination angles showed a sinusoidal variation with position. Many A-DNA structures exhibit high base-pair buckle angles at the ends of the double helix, where the base-pairing geometry is distorted by interactions with symmetry-related molecules. In the dodecamer, base-pair buckle and cup peak at the center of the double helix, and point to a possible unusual conformation near the center. It is interesting to note that the central base-pairs of the dodecamer are at the junction between the three distinct vertical base-stacking domains of the dodecamer. The two cross-strand purines G7 and G19 belong to the central stack, whereas their hydrogen-bonded partners, C6 and Cl8 belong to the flanking stacks. The angle between the base planes of C6 and Cl8 is 45”. In an attempt to model the bend in the dodecamer, the angles between the upper and lower halves of bhe molecule were computed with NEWHEL91

by using C-l’ atoms alternately from the upper and lower halves of the duplex in the helical axis calculations. The division of the molecule into two parts is quite reasonable in this case, since the 12-mer has a dyad. The angle between the crystallographic helical axes of the upper and lower halves is 20”, and the shortest distance between the half-molecule and overall axis is 911 A (Fig. 9). The bend can be described as a rotation of the upper and lower hexamers relative to each other, very nearly in the common plane of the molecular dyad and the overall helical axis. When the calculation is performed for the fiber model of A-DNA, the rotational and translational disrepancies of the fragment axis from the global helix axis was O*OO” and 000 A, indicating that the helical axis differences seen in the refined molecule reflect authentic features of the atomic model, and not limitations of the numerical algorithm used to compute the axes. (e) Helical parameters relative to overall and half-molecule axes The aberrations in the helical parameters of the dodecamer caused by using the global helical axis made their interpretation difficult. The parameters were substantially easier to understand when they were calculated with respect to the hexamer axes, and it was seen that in several cases, the most prominent features in the plots of the parameters relative to the global axis were artifacts created by the poor match of this axis to the local geometry of the molecule. The derived helical parameters for both cases are given in Table 4, and are compared in Figure 10. Perhaps the most fundamental parameters characterizing double-stranded DNA molecules are the helical rise and rotation per residue. With respect to the overall helical axis, the mean helical rise per residue of the dodecamer is 2.55 A, with a standard deviation of 0.43 A. The mean value is very close to the value for fiber diffraction A-DNA, 2.56 A. However, when the helical rise is computed with respect to the hexamer axis, the rise per residue increases to 2.97 A, with the standard deviation decreasing to 0.20 A, indicating that the hexamer

(b)

Figure 5. Solvation of the dodecamer duplex. All 80 water molecules are shown with a van der Waals’ radius of I4 ‘8 for the water oxygen atoms. (a) View into the major groove showing a monolayer of water molecules in the major groove. A region of the minor groove backbone (arrow) is devoid of hydration. This is caused by the close contact of molecules in the crystal related by the 3,-axes. (b) View is rotated 90” from top view. The same region devoid of hydration is seen. (c) View is into the minor groove, rotated 180” from the top view. Notice that the minor groove is less hydrated than the major groove of the molecule, due in part to intermolecular interactions. The region of the minor groove around the dyad axis is affected by 6,-related molecules.

P24

P23

P22

(d) A21 T4 =i@oM

P20

(e)

T20

A5

@rmca

P19

(f) Gf9

mta

Figure 6. Base-pair stacking diagrams of individual dinucleotide normal to the plane of the top base in strand 1. The top base-pair open bonds. (a) The stacking patt’ern for t#he Pur-Pur and Pyr-Pyr (c) and (e) The stacking for the Pur-Pyr steps.

C6

627

seps in the dodecamer d(CCGTACGTACGG) viewed is shown as filled bonds and the bottom base-pair as step. (b), (d) and (f) The stacking for Pyr-Pur steps.

748

G.

A. Bingman

Figure 7. Schematic representation of the base-stacking pattern in the dodecamer. Attention is drawn to the prominent vertical stacks of bases by enclosing them in a box. These stacking domains are clearly laterally separated. See also Fig. 4(c) for a view of the dodecamer down the global helical axis. The 3 vertical “domains” of stacked bases are clearly visible.

axis is a better choice. Figure 10 shows the helical rise for both the overall and hexamer axis. With respect to the overall helical axes, the maximum in this plot occurs at the central step of the molecule. A simple mechanical analogy seems to explain t,his observation. In A-DNA, the bases are displaced from the helix axis. If the two ends of the dodecamer were moved toward the major groove side of the molecule, the base-pairs at’ the location of such a hinge motion would be imagined to pull apart slightly. Since the bases are displaced away from the helix axis, this would cause a local increase in helical pitch. For the hexamer case, the helical rise is largest at the fourth step, the central step and the end step. The increase at the end of the molecule may be related to interaction with symmetryrelated molecules. The other two local maxima occur at 5’ R-Y steps in the molecule, and may be sequence related. To determine the helical rise at the central step, it was necessarv t,o consider the seventh base-pair, which is outside the range of residues used in the helical axis calculation. The variation in the local helical twist is diminished for the hexamer axis but is not entirely abolished. The t,wist is lowest at the ends of the molecule, reaches a maximum at the third step, and a local minimum in the center of the molecule. The decrease at the ends may be simply a conformational adjustment in the duplex to provide optimal contacts with the sugar-phosphate backbone of symmetry-related molecules. Tt should be noted that at the central step the parameters describing relationships between pairs of base-pairs, such as helical twist, roll and tilt angles, should be treated with caution, since they are derived from the relationship between base-pair 6, which was included in the helical axis calculation, and basepair 7, which was not. The inclination of the base-pairs with respect to the global helix axis shows two local maxima at the fourth and ninth residues. The variation is gradual, without apparent sequence dependence, and was extremely puzzling when initial attempts were made to rationalize the structure of the dodecamer.

et al.

However, the plot with respect to the hexamer axis shows that it is essentially flat, indicating that the local helical axis is a better choice. It is interesting to consider what would happen if the dodecamer structure were extended far beyond its present, limits by adding additional t)urns of unbent A-DNA to its termini. There would then be a sinusoidal variation in inclination, with rnaxima at points where the inclination angle of the base-pair was added t,o the offset in the local and global helical axis, the minima where t,he inclination angle was in opposition to this angle. This argument, seems to agree well with values of inclination angles calculated for the dodecamer. The mean inclination for the first six residues is about 12”. and each half of t,he dodecamer seems to be t,ilted with respect t,o the average helical axis by approximately 10”. From this argument. a maximum inclination of 22” and a minimum inclination of 2” are expect,ed. The actual maximum of 23” occurs at T4. Assuming 11 basrpairs per turn. t,he minimum value would be expect,ed one half turn away or ?5.5 residues before T4. The plot, of inclination is monotonicall\; decreasing towards the ends, and drops to a, value of 8” at, the first, residue, I.5 residues before t.hc expected minimum. In Figure 11 the x and :y components of the unit vect.ors normal to the least-squares best plane of each base-pair arc plotted from the origin (Otwinowski ef rrl.. 1988). The plot for fiber diffraction A -DNA shows t,hat unit vect’ors precess around a circle of radius 0.34, equal to t,he sine of the inclination angle. A similar plot for t)he base-pair normals of the dodecamer is more irregular. At t.hr ends of the molecule. the base-pair normals do not traverse a.s large a semicircular arc as the base--pairs at the center of the molecule. This is reflected in t.he small twist angles at the terminul steps. The progression of the base-pair normals seems to tab somewhat compressed along the .I’ axis of the plot. with a large vertical step at, the central base-pair. When these normals are plotted with respect to the local helical axis over residues 1 to 6. the arc described by the first six base-pair normals seems substantially more circular, although the arc’ dist)ance between the first t’wo base-pairs still is Iesh than would be rxpert,ed. The radius of the a.rc of the projection of t,hese normals is significant~ly IPSS than t,hat of fiber difIract,ion A -DNA. consistent with the smaller inclination angles measured for the first six base-pairs with the hexamer axis. Since the dinucleotide tilt angle represents a change in inclination, it is also expected to be perturbed by the choice of helical axis. For the overall axis, there is variation over a range of 14 ‘. with a monotonic decrease over residues 1 to 5. from 7 to -2, and an abrupt’ jump to zero at, the cent,ral residue. When t,he hexamer molecule axes are used. the variation in tilt angle is grea,tly reduced. to 2” over residues 1 to 5. There seems t)o be some residual sequence effect. on tip angles, with local maxima at R-Y steps. Since the tip and inclination represent orthogonal

Structure of A-DNA

Dodecamer

749

Figure 8. Superposition of the starting and the refined model. The refined model is shown in heavy bonds. the starting model in light bonds. (a) View is into the major groove along the dyad. Pjote the large changes in the position and orientation of the terminal base-pairs. The large relative rotation of the cross-strand pyrimidines at the central step is also apparent, while the cross-strand purines show little relative rotation. (b) The view is normal to the dyad and the helix axis. Notice that the ends of the final model are rotated toward the major groove.

rotations, both are affected There are large variations in but not entirely abolished, axis. The tip angles change switching from a global to with the above argument

by choice of helix axis. tip, which are damped, by using the hexamer by at most 10” when a local axis, consistent concerning inclination.

However, a substantial variation in tip remains. For the first half of the molecule, the outermost and innermost base-pairs are tipped the most. The conformation of the outermost base-pairs most probably is influenced by the stacking of these base-pairs against the sugar-phosphate backbone of the minor

750

C. A. Biqman

et, al.

Figure 9. The global helical axis (broken lines) for the dodecamer is shown with the hexamer axes (continuous lines) superposed on a skeletal model of the dodecamer. The molecular dyad is in the plane of the page. The angle between the 2 hexamer axes is 20”. and the distance between the global and hexamer axes is @l 1 A. Ordinarily, a helix axis deflect,ion 20” in this direction would lead to severe collisions of the phosphates at the center of the molecule but because of several conformational rearrangements involving the base-pair twist and helical rise, these collisions have been avoided (see the text).

groove of symmetry-related molecules. The large tip angles in the center of the molecule may be related to the bend at the center. The roll angle of the base-pair steps in the dodecamer is largest at the central step of the molecule, and this is still valid for the calculations performed with hexamer axes. The roll angles for the hexamer axis are somewhat smoothed, but there is still some systematic increase in this parameter from the ends of the duplex to the center. Furthermore, there seems to be an unusual pattern in the increase of this parameter. In pyrimidine-purine steps, the roll angle appears to rise, while in the purine-pyrimidine steps, the roll angle seems to remain constant. It has been previously observed by Calladine (1982) that propeller twisting causes a steric clash in the minor groove side of pyrimidine-purine steps, which may be alleviated by increasing the roll angle at this step. The values of propeller twist are not influenced by a poor match between the local and global helical axis over the range of conditions encountered in this molecule. The global shifts in propeller twist that are seen might be caused by changes in the relative position of the backbone of the two strands. Such effects might be most directly reflected in changes in the roll angle and the width of the minor groove of the dodecamer. This possibility will be considered when the groove geometry is discussed. Buckle and cup have been discussed in section (d), above.

(f) Major

groove dimensions

Early examples of crystalline il - DNA were found in studies on DNA octamers and tetramers. Tt was quickly noted that the apparent major groove dimensions of these oligomers deviated from models derived from fiber diffraction studies. It was dificult to measure precisely the major groove width of A-DNA duplexes comprising less than a full helical turn, since the closest approach of cross-strand phosphate groups occurs between t’he phosphate groups that are nine residues apart in the double helix. Consequently, the major groove dimensions could only be estimated by extending the atomic model using an overall helical axis, or could be compared with the next shortest approach in fiber diffraction models. There seem to be at least two distinct populations of A-DNA octamers, those crystallizing in the tetragonal space group P4,2,2, which have the widest major groove, estimated from phosphate groups seven residues apart of 8 to 10 A (Wang et al., 1982a; Conner et al., 1982, 1984: McCall et al., 1986; Heinemann et al., 1987; Jain ef al., 1989; Takusagawa, 1990) and the hexagonal A-DNA octamers, which have major grooves about 6 A across, again based on what is expected to the next to shortest phosphate-phosphate separation (Shakked et aE., 1983; McCall et al., 1985; Lauble et al., 1988; Jain et al., 1989; Jain & Sundaralingam, 1991). Therefore, it is interesting to have informa-

2.73

2.39

2.26

2.57

256

3.03

255

2.57

2.26

2.39

2.73

3.05

299

3.15

2-88

2.91

2.%

Half

281

33.2

348

342

326

30.7

326

34.3

347

332

28.1

All

30.3

31.7

32.9

33.3

32.6

2@7

Half

Twist (“)

&4

153

208

22.7

21.6

17.1

17.1

21.6

227

2@8

13.4

13.5

12.7

124

11.4

9.3

84

153

Half

(“)

All

Inclination

-7.0

-5.2

-1.7

1.0

45

0.0

- 4.5

- 1.0

1.7

52

7.0

All

5.0

-0.1

0.9

0.2

1.0

2.1

-l&O

-133

-7-4

1.2

3.8

4.4

-44

-38

-1.2

7.4

5.4

2.4

-0-7

1.9

4.0

7.8

18.0 133

Half

All

Tip 0

-4.4

-59

-9.2

-2.7

-0.5

9.1

-0.5

-28

-9.1

-59

-44

over-all

IO.1

3.1

3.2

-26

-2.2

- 3.9

Half

Roll (“) All

Table 4 d(CCGTACGTACGG)

Half

Tilt (“)

parameters,

- 3.6

-4.1

-4-1

-3.6

-3.8

0.0

-0.1

O-4

0.3

0.4

0.1

-@l

-0.4

-0.3

-0.4

0.1

0.1

0.1

@O

-0.4

0.1

0.3

-1.6

-1.8

-1.1

-1.4

-1.2

-1.4

-1.2

- 1.4

- 1.1

-1.8

-1.6

Slide (4

4.2

50

6.7

10.4

134

13.1

13.1

13.5

10.4

67

50

4.2

-Prop twist (“)

- 6.2

-0.9

1.1

1.4

@8

6.1

-6.1

-0.8

-1.3

-1.1

0.9

6.2

Buckle (7

axis, outside the range over which this axis was calculated.

- 5.4

-45

-3.8

-42

-4.6

-49

-49

-46

- 4.2

-3.8

-4.5

0.0

-4.3

-5.4

Half

All

Half

All

X-disp.

Y-disp. (A)

axes

(il)

and half-molecule

Normal type: overall axis. Bold type: half-molecule axis over the valid range of the axis. Italic type: half-molecule for comparison of these parameters. See EMBO Workshop (1989) for definitions of helical parameters.

G12

Gil

Cl0

A9

T8

G7

C6

A5

T4

G3

c2

Cl

All

Rise (A)

Helical

See Fig. 10

-53

-1.9

-0.2

0.5

-52

122

-5.2

0.5

--o-2

-1.9

-53

cup (3

C. A. Bingman

752 3.25 ,

I

25 L

et al.

I

-10 CCGTACGTACGG

CCGTACGTACGG

Sequence

Sequence

sequence 35

27

-15 -

CCGTACGTACGG

CCGTACGTACGG

‘: $

CCGTACGTACGG

Sequence

Sequence

12.5

; 10.0 .t 7.5 & %a 5.0 : 7

2.5

-10 ’

0:o CCGTACGTACGG

J CCGTACGTACGG

-10 -

Sequence

Sequence

Figure 10. Comparison of descriptive helical parameters for the overall and local the overall helical axis are given by open circles while those for the local axis are axis, the parameters for the entire dodecamer are given, including the dyad-related the tilt, tip and buckle, which have a dyad normal to the paper through the center have the dyad in the plane of the paper. The helical axes were calculated using all axis and the first 6 base-pairs for the half-molecule.

tion on structures of longer A-DNA oligomers to see if this trend of wider than expected major grooves persists. The major and minor groove dimensions of the dodecamer are shown in Figure 12, superimposed on a cylindrical projection of the backbone atoms. At first glance, it seems that the major groove of the dodecamer is quite similar to that expected from fiber diffraction studies. There are two independent measurements of this distance in the dodecamer, of 2.7 w and 4.3 A, respectively. The corresponding distance in fiber model of A-DNA is 2.2 A (Chandrasekaran et al., 1989). In the dodecamer, the smallest separation is similar to fiber diffraction A-DNA, while only one base-pair above and below these shortest contacts, the separation increases by l-6 A. There are two other A-DNA structures that allow a direct measurement of this shortest P-P separa-

CCGTACGTACGG

hexamer helical axes. Parameters for given by filled circles. For the overall half. Note that with the exception of of the plot, the remaining parameters the C-l’ to C-l’ vectors for the overall

tion, that of the A-DSA decamer d(ACCGGCCGGT) (Frederick et al.. 1989) and the A-DNA dodecamer d(CCCCCGCGGGGG) (Verdaguer et al.. 1991). In the the shortest’ cross-strand P-P first structure, distance is shifted by one residue relative to fiber diffraction A-DNA. The shortest approach is 3.0 A. The shift in the register of cross-strand phosphates does not occur in our dodecamer. No information on the groove dimensions of the dodecamer d(CCCCCGCGGGGG) is available. It is also noteworthy that there is a large discrepancy between the observed width of the major groove of our dodecamer and the width of estimated the major groove in d(GTACGTAC), a tetragonal octamer with a sequence identical to the inner eight base-pairs of the dodecamer. It is difficult to say if conclusions drawn from the major groove dimensions of either the decamer d(ACCGGCCGGT) or the dodecamer d(CCGTACGTACGG) have direct implications to

Structure of A-DNA

,a;

t

.\

. ._-. . . + ..._. - .. i j

I, ! .-.. ..j ..- ..,..,._.i.. /:’ ‘?I....:c /I i

i :.i /

-0.6 ~‘,,~1,,,,1,~,‘I’~~,‘I’,,,I,‘,,I,,,,I,,’~’,,I,’,J -06

-0.5 04

-0c3 -0.2 -0.1 0.0

0.1

0.2 0.3

04

0.5

Cos (AX)

Figure 11. Plot of the 5 (AX) and y (AY) components of unit vectors normal to the base-pair planes of the dodecamer in 3 cases. The points for the starting fiber diffraction model are shown by squares and, as expected, they trace a circular trajectory. If the base-pair normals were parallel to the helical axis, this plot would only show a single point at the center of the Figure, as is the case for idealized B-DNA. The points corresponding to refined models of the dodecamer are indicated by triangles. The filled triangles are for the overall helical axis, while the open triangles are for the hexamer axis.

the major groove dimensions of some “standard” A-DNA polymer. However, together they currently form the lower limits in major groove dimensions observed from single crystal studies. (g) Minor groove dimensions and correlations The phosphate groups of residues four base-pairs apart in the double helix delimit the minor groove

Dodecamer

753

of A-DNA. Accordingly, minor groove width is primarily a result of Watson-Crick base-pair geometry. Therefore, the minor groove dimensions of A-DNA crystal structures generally agree much better with fiber diffraction models than the dimensions of the major groove. Most workers have concluded that the width of the minor groove in A-DNA oligomers is quite monotonous. However, we find some variation in the dimensions of the minor groove of A-DNA crystal structures. There

are correlations between the minor groove width and the average roll angle, the minor groove width and average propeller twist and between the average propeller twist and average roll angles. These relationships have not been discussed before. The minor groove width of the dodecamer increases from the ends of the molecule towards the center, where the roll angle also reaches its maximum value. The phosphate groups that delimit the width of the minor groove comprise three doublehelical steps. Figure 13(a) shows that there is a good correlation between averaged roll angle of the threebase-pair steps and the corresponding minor groove width. This correlation is expected, since a larger positive roll angle opens adjacent base-pairs on the minor groove side. There is also a relationship between the roll angle and the propeller twist angles of the two base-pairs defining that roll angle (Fig. 13(b)). This can be rationalized by Calladine’s (1982) analysis of the forces shaping the relative positions of adjacent base-pairs in the double helix. Another striking relationship is that between the width of the minor groove and the average propeller twist of the four intervening base-pairs as shown in Figure 13(c)). For this A-DNA dodecamer, the relationship between propeller twist and the width of the minor groove is the opposite direction to the general trend seen in the isomorphous family of B-DNA dodecamers, where high propeller twists occur in the A+T-rich core of the molecule, which has the narrowest minor groove. All three of these correlations exist even when parameters calculated

Figure 12. Cylindrical projection of the sugar-phosphate backbone atoms of the dodecamer, computed at r = 10 A> and covering an angular range of 540” (more than once around the molecule). This allows the major and minor grooves to be shown continuously in 1 plot. The major and minor grooves widths are shown and were obtained by subtracting 58 A from the P-P separation, to account for the van der Waals’ radii of the phosphate groups.

C. A. Bingman et al.

754

with respect to the global helical axis are used. For the two correlations that are affected by the axial choice (Fig. 13, top and center), the correlations are more robust when the half-molecule axis parameters are used. (h) Conclusions This

work has provided the structure of an dodecamer. The duplex is slightly more than one full helical turn and has allowed direct measurement of the major groove width. A novel minor groove to minor groove packing interaction has been observed, in addition to the ubiquitous terminal base-pair to minor groove packing observed in other A-DNA structures. The duplex is bent by 20”. Without additional struct,ural information, it is impossible to conclude that this bend is or induced by crystal packing either intrinsic, forces. Irrespective of the origin of the bend. t)he dodecamer has also been a rich source of new correlations between global parameters and local helical parameters. Analysis of its structure has also illustrated the importance of an appropriate choice of helical axis. The dodecamer structure presented here suggest,s several possible criteria in assessing the utility of fragment versus global helical parameters. The most obvious possibility is to use t,he standard deviations of t,he rise per residue and helical twist. For thp dodecamer, the standard deviation in the rise per residue actually decreased by a factor of 2 when calculated for the hexamer axis, and that of thr helical rotat,ion decreased by a third. It is notcworthy t,hat simply dividing an oligonucleot’ide st,ruct’ure into smaller helical fragments does not, always guarantee that the st,andard deviations of the helical parameters will actually decrease. For example. in t)he tetragonal A -DNA octamer d(GTG CGCAC), the standard deviation of the helical rise actually increased by 11 (+, when t’he molecule was divided into two halves (C‘. Bingman. unpublished observations). Soumpasis it nl. (1991) ha,ve also suggested monitoring a similar parameter ((6)) in their helical axis algorithm. A second somewhat less objective criterion is t,he degree to which the derived helical parameters drift) when calculated with respect t,o the global axis, and the degree to which systematic drifts are corrt%cted by using local helical axes. Tn the case of’ our dodecamer, dividing t,he molecule into two dyatlrelated halves for calculating the helical parameters was clearly just,ified. Large aberrations in the parameters. of the order of 13” for inclination. 10” for tip, 5” for tilt and roll angles. and 04 A for rise per residue, are caused by a poor fit of an overall helical axis to the structure. Artifacts of this magnitude are

A-DNA

-7-6-5-4-3-2-l

0 Average

I

2

roll angle

3

4

5

6

(deg.)

(0)

45

6

7

0

-Average

9

IO

propeller

II twist

I2

13

14

(deg.)

(b)

6

7

8 -Average

9

IO propeller (c

II twst

I2

13

14

(deg.)

I

Figure 13. Correlations relating the width of the minor groove, base-pair propeller twist and roll angles. (a) Minor groove width as a function of included roll angles. The 3 roll angles separating the phosphate groups determined the width of the minor groove have been averaged. The data points for the global helix axis are indicated by open circles. and the least-squares regression line is broken. The correlation coefficient is 990. The points for the hexamer axes are indicated by filled circles, and the least-squares line is continuous. The correlation coefficient is 994. (b) Roll angle versus propeller twist. The propeller twists to the 2 base-pairs defining a roll angle have been averaged, and the global (correlation coefficient = 666) and hexamer axis (correlation coefficient = O-85) cases are

-_-indicated as above. (P) Propeller t,wist vwsus minor groove. The propeller twists of the whose phosphate groups determine the width groove have been averaged. The correlation 0.98.

width of the 4 base-pairs of the minor coefficient is

Structure of A-DNA bound to obscure or even totally conceal underlying sequence-structure correlations that are of interest. We know of no other case in the literature where such striking departures have been described for an oligonucleotide crystal structure. Accordingly, this structure may become a paradigmatic example of a bent DNA molecule, and of great use in developing and testing procedures for describing helical parameters in terms of continuously variable helical axis (e.g. see Lavery & Skelnar, 1988, 1989). A third criterion, which unfortunately can only be applied a posteriori, is the ability to direct new and stereochemically reasonable correlations between parameters when they are calculated for a fragment VeraUs global axes. For example, the correlation coefficient observed between the included roll angle and the minor groove width in this dodecamer increased from @90 to @94 when the roll angles were calculated with respect to hexamer axis rather than the global helical axis. Calculations based on an overall helical axis may have the advantage that the same set of assumptions can be used on all DNA oligomer structures. However, this work has shown that the standard deviations in the parameters can be reduced by choosing an obvious fragment axis rather than a global axis. Therefore, a fragment axis for a duplex should be explored, which represents the first logical step in the continuum between the fiber diffraction global helical axis and a continuously, smoothly changing helical axis through the molecule, such as in the case of circularized DNA, the DNA in the CAP co-crystal structure (Schultz et al., 1991) and DNA wrapped around a histone core particle (Richmond et al., 1984). We thank Dr David Jebaratnam for his assistance in oligonucleotide synthesis Dr S. T. Rao and Dr Sanjeev Jain for discussions and programs, Dr Paul Sigler and Dr Z. Otwinowski for the use of their Optronics film scanner and oscillation film processing program DErjZO, and the National Institutes of Health for research grant GM-17378.

References Arndt, U. W. & Wonacott, A. J. (1977). The Rotation Method in North-Holland, Crystallography, Amsterdam. Arnott, S. & Hukins, D. W. L. (1972). Optimized parameters for RNA double-helices. Biochem. Biophys. Res. Commun. 48, 1392-1398. Arnott, 8. & Hukins, D. W. L. (1973). Refinement of the structure of B-DNA and implications for the analysis of X-ray diffraction data from fibers of biopolymers. J. Mol. Biol. 81, 93-106. Arnott, S., Chandrasekaran, R., Birdsall, D. L., Leslie, A. G. W. & Ratliff, R. L. (1980). Left-handed DNA helices. Nature (London), 283, 743-745. R. G. & Sundaralingam, Brennan, M. (1985). Crystallization and preliminary crystallographic studies of the decadeoxyoligonucleotide (dCpGpTpApCpGpTpApCpG). J. Mol. Biol. 181, 561-563. Brennan, R. G., Westof, E. & Sundaralingam, M. (1986). Structure of a Z-DNA with two different backbone

Dodecamer

755

chain conformations. Stabilization of the decadeoxynucleotide d(CGGTACGTACG) by [CO(NH,),]~‘. J. Biornol. Struct. Dynam. 4, 649-665. Calladine, C. R. (1982). Mechanics of sequence-dependent stacking of bases in B-DNA. J. Mol. Biol. 161, 343352. Chandrasekaran, R., Wang, M., He, R.-G., Pugianer, L. C., Byler, M. A., Millane, R. P. & Arnott, S. (1989). A re-examination of the crystal structure of A-DNA using fiber diffraction data. J. Biomol. Struct. Dynam. 6, 1189-1202. Conner, B. N., Takano, T., Tanaka, S., Itakura, K. & Dickerson, R. E. (1982). The molecular structure of d(‘CpCpGpG), a fragment of right-handed soluble helical A-DNA. Nature (London), 295, 294-299. Conner, B. N., Yoon, C., Dickerson, J. L. & Dickerson, R. E. (1984). Helix geometry and hydration in A-DNA tetramer: *C-C-G-G. J. Mol. Biol. 174, 663695. Dickerson, R. E. & Drew, H. R. (1981). Structure of a B-DNA dodecamer. II. Influence of base sequence on helix structure. J. Mol. Biol. 149, 761-786. Drew, H. R., Takano, T., Takana, S., Itakura, K. & Dickerson, R. (1980). High-salt d(CpGpCpG), a lefthanded 2’ DNA double helix. Nature (London), 278, 755-758. Drew, H. R., Samson, S. & Dickerson, R. E. (1982). Structure of a B-DNA dodecamer at 16 K. Proc. Nat. Acad. Sci., U.S.A. 79, 4040-4044. Duocet, J., Benoit, J. P., Cruse, W. B., Prange, T. & Kennard, 0. (1989). Coexistence of A- and B-form DNA in a single crystal lattice. Nature (London), 337, 190-192. EMBO Workshop (1989). Definitions and nomenclature of nucleic acid structure parameters. EMBO J. 8, l-4. Frederick, C. A., Quigley, G. J., Teng, M.-K., Coll, M., van der Marel, G. A., van Boom, J. H., Rich. A. & Wang, A. H.-J. (1989). Molecular structure of an A-DNA decamer d(ACCGGCCGGT). Eur. J. Biochem. 181, 295-307. Heinemann, U., Lauble, H., Frank, R. t Blocker, H. (1987). Crystal structure analysis of an A-DNA fragment at 1.8 A resolution: d(GCCCGGGC). Nucl. Acids Res. 15, 9531-9550. Hendrickson, W. A. & Konnert, J. (1981). Stereochemically restrained crystallographic leaatsquares refinement of macromolecule structures. In Bkmwlecular Structure, Conformation, Function and Evolution (Srinivasan, R., ed.), pp. 43-57. Pergamon Press, Oxford. Jain, S. & Sundaralingam, M. (1987). The potentially Z-DNA-forming sequence d(GTGTACAC) crystallizes as A-DNA. J. Mol. Biol. 197, 141-145. Jain, S. & Sundaralingam, M. (1991). Hexagonal crystal structure of the A-DNA octamer d(GTGTACAC) and its comparison with the tetragonal st’ructure: correlated variations in helical parameters. Biochemistry, 30, 3567-3576. Jain, S., Zon, G. & Sundaralingam, M. (1989). Base only binding of spermine in the deep groove of the A-DNA octamer d(GTGTACAC). Biochemistry, 28, 236& 2364. Jovin, T. M., McIntosh, L. P., Arndt-Jovin, D. J., Zarling, D. A., Robert-Nicoud, M., van de Sande, J. H., Jorgensoon, K. F. t Eckstein, F. (1983). Left-handed DNA: from synthetic polymers to chromosomes. J. Biomol. Struct. Dynum. 1, 21-57. Kennard, O., Cruse, W. B. T., Nachman, J., Prange, T., Shakked, Z. & Rabinovich, D. (1986). Ordered water

756

C. A. Bingmun

structure in an A-DNA octamer at 1.7 A resolution. J. Biomol. Struct. Dynam. 3, 623-647. Lauble, H., Frank, R., Blocker, H. 6 Heinemann, U. (1988). Three-dimensional structure of d(GGGATCCC) in the crystalline state. Nucl. Acids Res. 16, 7799-7816. Lavery, R. & Skelnar, H. (1988). The definition of generalized helicoidal parameters and of axis curvature for irregular nucleic acids. J. BiomoE. Strict. Dynam. 6, 63-91. Lavery, R. & Skelnar, H. (1989). Defining the structure of irregular nucleic acids: conventions and principles. J. Biomol. Struct. Dynam. 6, 655-667. McCall, M. Brown, T. & Kennard, 0. (1985). The crystal structure of d(GGGGCCCC): a model for poly(dG) .poly(dC). J. Mol. Biol. 183, 385-396. McCall, M., Brown, T., Hunter, W. N. & Kennard, 0. (1986). The crystal structure of d(GGATGGGAG) forms an essential part of the binding site for transcription factor IDA. Nature (London), 322, 661664. Otwinowski, Z., Schevitz, R. W., Zhang, R. G., Lawson. C. L., Joachimiak, A., Marmorstein, R. Q., Luisi, B. F. & Sigler, P. B. (1988). Crystal structure of trp repressor/operator complex at atomic resolution. Nature (London), 335, 321-329. Prive, G. G., Heinemann, U., Chandrasegaran, S., Kan, L.-S., Kopka, M. L. & Dickerson, R. E. (1987). Helix geometry, hydration, and GA mismatch in a B-DNA decamer. Science, 238, 498-504. Quadrifoglio, F., Manazini, G. & Yathindra? N. (1984). Short oligodeoxynucleotides with d(G-C)n sequence do not assume left-handed conformation in high salt conditions. J. Mol. Biol. 175, 419-423. Richmond, T. J., Finch, J. T., Rushton, B., Rhodes, D. & Klug, A. (1984). Structure of the nucleosome core particle at 7 A resolution. Nature (London), 311, 532-537. Saenger, W.. Hunter, W. N. & Kennard, 0. (1986). DN’A conformation is determined by economics in the hydration of phosphate groups. Nature (London), 324, 385-388. Schultz, S. C., Shields, G. C. & Steitz, T. A. (1991). Crystal structure of a CAP-DNA complex: the DNA is bent by 90”. Science, 253, 1001-1007. Shakked, Z., Rabinovich, D., Cruse, W. B. T., Egert., E.. Kennard, O., Sala, G., Salisbury. S. A. & Viswamitra. M. A. (1981). Crystalline A-DNA: the X-ray analysis of the fragment d(G-G-TA-T-A-C-C). Proc. Roy. SOC. ser. B, 213, 479-487. Shakked, A., Gurstein-Guzikelvich, G., Eisenstein; MM.,

et al.

Frolow, F. CG.Rabinovich, D. (1989). The conformation of the DNA double helix is dependent on its environment. Nature (London), 342, 456-460. Shakked, Z., Rabinovich, D., Kennard, O., Cruse. W. B. T., Salisbury, S. A. & Viswamitra, M. A. (1983). Sequence dependent conformation of an A-DNA double helix: the crystal structure of the octamer d(GGTATACC). J. Mol. Biol. 166, 183-204. Soumpasis, D. M., Tung, C.-S. & Garcia. A. E. (1991). Rigorous description of DNA structures. II. On the computation of best axes, planes and helices from atomic coordinates. J. Biomol. Struck Dynam. 8, 867-888. Takusagawa, F. (1990). The crystal structure of d(GTACGTAC) at 2.25 A resolution: are the A-I)E’As always unwound approximately 10 degrees at the C-G steps? J. Biomol. Struct. Dynam. 7, 795-809. Thota, N., Li, X. H., Bingman, C. & Sundaralingam, S. (1992). High resolution refinement of the hexagonal A-DNA octamer d(GTGTACAC) at 1.4 A resolution. Acta Crystallogr. in the press. Verdaguer, N.. Aymami, J., Fernandez-Forner, I).. Fita. 1.. Coil, M., Huyuh-Dinh, T., Igolen, J. & Subirana. J. 9. (1991). Molecular structure of a complete turn of A-DNA. J. Mol. Biol. 221, 6233635. Viswamitra, M. A., Kennard. O., Jones, I’. G.! Sheldrick. G. M., Salisbury, S., Falvello, L. & Shakked. Z. (1978). DNA double helical fragment at atomic resolution. Nature (London), 273, 687-688. Wang, A. H.-J., Quigley, G. J., Kolpak, F. ,J.. Crawford. J. L.. van Boom, ,J. H., van der Marel, G. & Rich. A. (1979). Molecular structure of left-handed doublehelical DNA fragment at atomic resolution. d(CGC GCG). Nature (London), 282, 680-686. Wang, A. H.-,J., Fujii, S., van Boom. J. H. & Rich, A. (1982a). Molecular structure of the octamer d(GGC CGGCC): modified A-DNA. Proc. Mat. Acad. Sci.. I,‘.S.A. 79. 3968-3972. Wang. A. H.-J ,, Fujii, S.. van Boom, ,J. H., van der Marrl. G. A.. van Boeckel, S. A. A. & Rich, A. (1982b). Molecular structure of r(GCG)d(TATACG(!): a DNA-RNA hybrid helix joined to double-helical DNA. Nature (London), 299, 601-604. Westhof, E.. Dumas, P. & Moras. I). (1985). Crystallographic refinement of yeast aspartic acid transfer RNA. J. Mol. RioE. 184, 119-145. Zon, G. Rr Thompson, J. A. (1986). Review of highperformance liquid chromatography in nucleic acids research. Isolation, purification. and analysis of oligodeoxyribonucleotides. Riochromatography. 1. 22-3 1,

Edited by M. F. Moody