Journal of

Structural

Journal of Structural Biology 148 (2004) 34–50

Biology

www.elsevier.com/locate/yjsbi

Structure of DNA sequence d-TGATCA by two-dimensionalnuclear magnetic resonance spectroscopy and restrained

molecular dynamicsq,qq

Ritu Barthwal,a,* Pamita Awasthi,a Monica,a Manpreet Kaur,a Uma Sharma,a

Nandana Srivastava,a S.K. Barthwal,b and Girjesh Govilc

a Department of Biotechnology, Indian Institute of Technology Roorkee, Roorkee 247 667, Indiab Department of Physics, Indian Institute of Technology Roorkee, Roorkee 247 667, India

c Chemical Physics Group, Tata Institute of Fundamental Research, Homi Bhabha Road, Navy Nagar, Colaba, Mumbai 400 005, India

Received 4 November 2003, and in revised form 24 May 2004

Available online 2 July 2004

Abstract

The 50 d-TpG 30 element is a part of DNA sequences involved in regulation of gene expression and is also a site for intercalation

of several anticancer drugs. Solution conformation of DNA duplex d-TGATCA containing this element has been investigated by

two-dimensional NMR spectroscopy. Using a total of 12 torsional angles and 121 distance constraints, structural refinement has

been carried out by restrained molecular dynamics (rMDs) in vacuum up to 100 ps. The structure is characterized by a large positive

roll at TpG/CpA base pair step and large negative propeller twist for AT and TA base pairs. The backbone torsional angle, c(O50–C50–C40–C30), of T1 residue adopts a trans-conformation which is corroborated by short intra nucleotide T1H6–T1H50 (3.7�A)

distance in nuclear overhauser effect spectroscopy (NOESY) spectra while the backbone torsional angle, b(P–O50–C50–C40), exists intrans as well as gauche state for T1 and C5 residues. There is evidence of significant flexibility of the sugar–phosphate backbone with

rapid inter-conversion between two different conformers at TpG/CpA base pair step. The base sequence dependent variations and

local structural heterogeneity have important implications in specific recognition of DNA by ligands.

� 2004 Elsevier Inc. All rights reserved.

Keywords: TpG/CpA basepair step; 2D-NMR of d-TGATCA; Restrained molecular dynamics; Sequence related conformation; DNA flexibility

Several natural antibiotics bind to DNA in diverse

ways but with a common ability to act as potent in-

hibitors of DNA transcription and replication. Many

synthetic compounds have been added to this list in

search for more potent drugs for use in chemotherapy.

Efforts have since been directed towards understanding

the molecular basis for DNA sequence specificity for

qThis work was supported by a research Grant No. 99/37/02/

BRNS-Cell from Board of Research in Nuclear Sciences (BRNS),

Department of Atomic Energy (DAE), and Government of India.qq Supplementary data associated with this article can be found in

the online version, at doi:10.1016/j.jsb.2004.05.005.* Corresponding author. Fax: +11-91-1332-273-560.

E-mail addresses: ritubfbs@iitr.ernet.in, ritubarthwal@yahoo.co.in

(R. Barthwal), .

1047-8477/$ - see front matter � 2004 Elsevier Inc. All rights reserved.

doi:10.1016/j.jsb.2004.05.005

binding by identifying the preferred binding sequences

of many key drugs with natural DNA (Arcamone, 1981;

Carter, 1982; Neidle et al., 1983; Searle, 1993). The foot

printing data have proved to be useful in identifying

short DNA sequences, which are the specific binding

sites of drugs (Chaires et al., 1990). The mechanism of

specific recognition involves reading of the local DNAtopography at these binding sites. Therefore the strength

and kinetics of specific binding of a ligand to a double

stranded DNA depends on the ligand matching these

and other contacts in a direct readout of the sequence

(Helene and Lancelot, 1982; Searle, 1993). However, the

sequence dependent conformation of DNA modifies the

three-dimensional pattern of contacts to be made as an

indirect readout of the sequence. Therefore the knowl-edge about sequence dependent variations in DNA

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 35

duplex geometry is essential to the understanding of thefunction of DNA in biological processes (Gronenborn

and Clore, 1985; Hagerman, 1990; Steitz, 1990; Trifo-

nov, 1985). Sequence specific variations arise from a

specific three-dimensional arrangement of hydrogen

bonding site in bases/base pairs due to geometric dis-

position of functional groups, DNA helical geometry,

local deformability or bending of DNA and effects of

major/minor groove width. Apart from direct hydrogen-bonding, van der Waal’s and hydrophobic interaction

appear to play a key role in stabilization of protein–

DNA and drug–DNA complex. The discrimination be-

tween G–C and A–T base pair on the basis of patterns

of potential hydrogen bond donors and acceptors in the

major groove is easily possible (Helene and Lancelot,

1982; Searle, 1993). Besides, the major groove also ex-

hibits non-polar patches in the C5 regions of pyrimi-dines with an obvious difference due to the hydrophobic

contact offered by thymine methyl group. In the minor

groove the difference is much lesser although guanine-

2-amino groups presents a way of discrimination between

G–C and A–T base pairs. Intrinsic major and minor

groove width may not be such an important factor in

DNA recognition as there have been evidences of in-

duced fit of bulky ligand in the minor groove so as toaccommodate the ligand in binding site of DNA (Searle,

1993). It is generally regarded that in the deoxyribose

sugar, the N-conformer (characterized by pseudorota-

tion PN ¼ 9�/18�, C30-endo) and S-conformer (C20-endogeometry) are rapidly inter-converting in solution. The

fraction of S-conformer in non-terminal pyrimidines

may be 75–90% while that in purines, may be higher,

about 90–100% (Searle, 1993). Since C30–C40 bond isshared between deoxyribose and sugar–phosphate bond,

the conformational change can be transmitted from the

base to the backbone through the base-sugar confor-

mational correlations. In addition, the conformational

flexibility of the double helix at specific sites may play an

important role in recognition of proteins/drugs/ligands,

perhaps by providing loci with a propensity for con-

formational change which is ultimately reflected in theconformation of DNA–ligand complex upon ligand

binding.

Sequence dependent structural variations have thus

been a subject of considerable interest in past (Celda

et al., 1989; Chary et al., 1987; Dornberger et al., 1998;

Nibedita et al., 1993; Radha et al., 1995; Ravi Kumar

-et al., 1985; Sarma et al., 1988; Schmitz et al., 1992a,b;

Tonelli et al., 1998; Ulyanov et al., 1992; Yanagi et al.,1991). Selection of DNA sequences with anomalously

low gel electrophoretic mobility revealed the importance

of dinucleotide base step TG/CA in DNA bending

(Beutal and Gold, 1992). A growing number of studies

have since suggested the existence of dynamically and

structurally unique TG/CA dimers (Cheung et al., 1984;

Donlan and Lu, 1992; McNamara et al., 1990; Timsit

et al., 1991) which may be involved in DNA bendingand flexibility. TG steps are believed to be crucial struc-

tural elements of various in vivo processes such as reg-

ulation of gene expression and are over-represented in

consensus protein binding sites (Trifonov and Brendal,

1986). The CAC/GTG element in specific binding region

of Cro with the OR3 operator evidently facilitates pro-

tein-induced DNA bending (Lyubchenko et al., 1993). It

has also been observed that two CAC/GTG elements atOL1 and OR2 sites improve interaction with both Cro

and C1 repressor proteins (Kabsch et al., 1982; Sarai

and Takeda, 1989). Possible binding mode of transcrip-

tional factors to regulatory DNA sequences are revealed

by the crystal structure of CAP–DNA complex exhib-

iting two 40o kinks between TG/CA base pairs (Schultz

et al., 1991; Steitz, 1990). The structural elements CATG

and TTTGCA are also a part of immunoglobulin pro-moter regions recognized by transcriptional factors

(Kemler et al., 1989) and octamer motif of immuno-

globulin genes (Weisz et al., 1994). The flexibility in base

triplets containing TG base step has also been attributed

to origin of genetic diseases (Bacolla et al., 1997; Cha-

stain and Sinden, 1998).

DNA bending has also been correlated to changes in

minor groove width (Chuprina et al., 1991) and otherhelicoidal parameters (Dornberger et al., 1998; Weisz

et al., 1994) and is believed to play an important role in

the process of recognition by proteins and minor groove

binding ligands, such as drugs. An understanding of the

structural heterogeneity of the DNA leads to the design

of new drugs targeted to interact directly with the DNA

sequence of interest. In this paper, we have carried out

structural analysis of d-TGATCA based on NMRspectroscopic techniques. This sequence contains 50d-TpG30 step in which intercalation of anticancer drugs—

adriamycin, daunomycin, and epiadriamycin takes

place, the evidence for which has been provided by foot

printing (Chaires et al., 1987, 1990), NMR (Barthwal

et al., 1994) and X-ray crystallography techniques

(d’Langlois Estaintot et al., 1992; Leonard et al., 1992,

1993; Nunn et al., 1991).

Materials and methods

The deoxyoligonucleotide sequence d-TGATCA was

purchased from DNA Chemical International USA and

used as such without any purification. The sample was

prepared by dissolving 5.5mg of DNA (�180 OD units)in 400 ll of D2O (deuterium oxide) containing 16.25mM

phosphate buffer (pH 7.0), 15mM NaCl and �0.1mM

ethylenediaminetetraacetic acid (EDTA). One microliter

of 0.1M sodium 2,2-dimethyl-2-silapentane-5-sulpho-

nate (DSS) was added as internal reference. The con-

centration of DNA duplex is approximately 3.68mM.

The sample was lyophilized several times from D2O and

36 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

finally made up to 0.4ml with 99.95% D2O just prior torecording of NMR spectra. The NMR spectra were re-

corded at Bruker AM 500 spectrometer operating at

500MHz for proton at the National High Field NMR

facility located at Tata Institute of Fundamental Re-

search (TIFR), Mumbai. The one-dimensional nuclear

magnetic resonance (1D NMR) spectra were recorded at

different temperatures at an interval of 5K in the range

277–325K. All two-dimensional nuclear magnetic reso-nance (2D NMR) experiments, double quantum filtered

correlation spectroscopy (DQF COSY) and NOESY,

were carried out at 295K. The chemical shifts have been

expressed with respect to the internal standard DSS.

Typical parameter for one-dimensional NMR experi-

ments are: pulse width �10 ls, number of data points 8–

16K, spectral width 4000Hz, number of scans 64–128,

digital resolution 0.25–0.50Hz/point. The typical pa-rameters for 2D experiments are 1024–2048 data points

along the t2 dimension, 512 free induction decays in t1

dimension, pulse width �9.5–12 ls, spectral width

4000Hz, number of scans 64–128, digital resolution

2.3Hz/point and relaxation decay 1.0 s. 2D NOESY

experiments were carried out at variable mixing times,

sm, that is, 75, 150, 200, and 250ms.

Cross-peaks in the NOESY spectra were integratedand intensities at mixing times of 75, 150, and 200ms

were translated into inter-proton distances using C5H5–

C5H6 cross-peaks of cytosine as the reference (2.45�A).

A range of �0.2�A was provided to distance to account

for any errors in integration. Pseudoatom corrections

were used for methyl and other equivalent protons. The

NOEs were categorized as strong, medium, and weak

with corresponding distance ranges 1.8–2.4, 2.5–2.9, and3.0–4.0�A set for the respective intensities of cross-peaks.

Force constants of 10.0, 5.0, and 2.0 kcalmol�1 �A�2

were fixed for strong, medium, and weak peaks, re-

spectively. The coupling constants J (H10–H20) and

J (H10–H200) were transformed to torsional angles using

Karplus relationship. These were introduced as dihedral

restraints allowing a range of �10� on the calculated

values with force constant of 20.0 kcalmol�1 rad�2. Thetotal numbers of dihedral and distance restraints used in

the structure calculations using restrained molecular

dynamics were 12 and 121, respectively. An initial model

of double helical B-DNA structure was generated using

INSIGHT II version 97.0, (Molecular Simulations,

MSI, San Diego) using Silicon Graphics O2 workstation

R5000 and is referred to as B-DNA1. The force constant

for hydrogen bonds was fixed as 40 kcalmol�1 �A�2

throughout the simulations. The energy of the molecule

was minimized using 1000 steps each of Steepest Descent

and Conjugate Gradient Methods to remove any inter-

nal strain due to short contacts in starting structure

using CFF91 force fields (Maple et al., 1988, 1990) in

DISCOVER version 97.0 (MSI). Dielectric constant was

fixed as 1:0 � r (r¼ distance) for calculation of electro-

static interactions. Conformational search was per-formed by the following simulated annealing restrained

molecular dynamics protocol: The molecule was heated

to a temperature of 500K in steps of 100K. Molecular

Dynamics was carried out for 25 ps (1000 iterations

steps of 1 fs each) at 500K during which 25 structures

were saved at regular intervals. Each of them was then

slowly cooled to 300K in steps of 50K. The force con-

stants for NOEs for strong, medium, and weak peakswere held constant as 25, 15, and 10 kcalmol�1 �A�2,

respectively while that for experimental dihedral re-

straints was fixed at 50 kcalmol�1 rad�2. At each step of

cooling by 50K the molecule was equilibrated for 25 ps

(2500 iterations steps of 1 fs each). At the end of simu-

lated annealing, 1000 steps of Steepest Descent mini-

mized all the structures until a predefined convergence

limit of root mean square derivative of energy with re-spect to atomic coordinates of <0.01 kcalmol�1 �A�1 was

reached. The final structure thus obtained is referred to

as rMD-B1SA. The complete protocol of rMD studies

with simulated annealing at 500K was repeated

by taking another starting structure (referred to as

B-DNA2) built using AMBER 7.0 and is referred to as

rMD-B2SA. Since the results obtained by NMR indi-

cated existence of a structure within the B-family ofDNA structures, we have also carried out studies using

rather a mild rMD protocol with both the starting

structures B-DNA1 and B-DNA2. In this protocol

the temperature is maintained at 300K throughout the

process of restrained energy minimization and re-

strained molecular dynamics and the corresponding

resulting structures are referred to as rMD-B1 and

rMD-B2. Another rMD structure, obtained after 100 pssimulations with starting structure B-DNA1, is referred

to as rMD-B1-100. All molecular dynamics calculations

have been carried out in vacuum.

Results and discussion

Analysis of NMR spectra

The one-dimensional proton NMR spectra of

d-TGATCA in D2O were recorded at 500MHz at nine

different temperatures in the range 277–325K. Some of

the NMR spectra at 325, 305, and 295K are available as

supplementary material supplied. Most of the protons

shift downfield gradually with the temperature due to

duplex to single strand transition. The chemical shifts ofvarious protons at three different temperatures are given

in Table 1. The downfield shifts in T4CH3, C5H5, and

G2H8 protons on increasing temperature from 277 to

325K are 0.32, 0.29, and 0.17 ppm, respectively. The

changes are quite pronounced in some of the H10 pro-tons due to ring current shifts; the shift in C5H10,T1H10, and G2H10 protons with temperature is 0.33,

Table 1

Chemical shifts of nonexchangeable protons of d-TGATCA at three different temperatures: 295, 305, and 325K

H8/H6 H5/H2/CH3 H10 H20 H200 H30 H40 H50 H500

295K

T1 7.35 1.73 5.90 1.76 2.25 4.65 4.07 3.65 3.61

G2 8.02 — 5.57 2.79 2.84 5.06 4.35 4.07 3.98

A3 8.33 7.60 6.34 2.76 2.96 5.06 4.52 4.31 4.22

T4 7.26 1.49 6.04 2.02 2.39 4.86 4.20 4.19 4.18

C5 7.52 5.80 5.90 1.99 2.30 4.80 4.02 3.97 3.95

A6 8.28 7.77 6.34 2.76 2.53 4.74 4.23 4.10 4.06

305K

T1 7.35 1.73 5.95 1.76 2.23 4.62 4.03 3.64 —

G2 8.00 — 5.61 2.70 2.70 4.96 4.31 4.03 3.97

A3 8.31 7.79 6.35 2.70 2.93 5.04 4.47 4.27 —

T4 7.30 1.57 6.05 2.08 2.38 4.83 4.16 4.03 —

C5 7.51 5.85 5.99 2.08 2.28 4.83 3.97 3.97 —

A6 8.30 7.61 6.33 2.70 2.56 4.76 4.21 4.08 —

325K

T1 7.43 1.84 6.03 1.91 2.27 4.63 4.02 3.65 —

G2 8.15 — 5.77 2.59 2.59 4.91 4.27 4.02 3.98

A3 8.34 7.80 6.40 2.59 2.92 5.03 4.44 4.21 —

T4 7.39 1.73 6.12 2.19 2.40 4.82 4.16 4.02 —

C5 7.53 5.98 6.12 2.19 2.32 4.82 3.98 3.98 —

A6 8.30 7.62 6.32 2.54 2.59 4.72 4.16 4.07 —

Fig. 1. Expansions of selected portion of 500MHz NOESY spectra of

d-TGATCA showing specific through space correlations within base

H8/H6, H10, H5, H30, H40 with H30, H40, H20, H200, CH3 protons at

sm ¼ 250ms.

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 37

0.24, and 0.18 ppm, respectively. The mid point of

transition, Tm, is found to lie in the range 304–311K for

all protons so that a major fraction of d-TGATCA ex-

ists in double helical state at 295K. This compares well

with the Tm values obtained for other oligonucleotides

in literature (Hilbers and Patel, 1975; Patel, 1977). The

imino proton spectra in 90% water at 2–10 �C shows

(please see supplementary information) that the increasein line width with temperature for T1NH protons is

more than that for T4NH proton due to fraying at ends.

Sequence specific assignments of various protons

have been obtained using well established procedures

based on 2D COSY and NOESY spectra (Feigon, 1992;

Hare et al., 1983; Hosur et al., 1988; Van de Van and

Hilbers, 1988; Van Wijk et al., 1992). The spin system

within the individual nucleotide unit is identified fromthe DQF COSY spectrum (available as supplementary

information) following the intra-nucleotide J connec-

tivity pathways. The individual spin systems are corre-

lated with the nucleotide type (T, G, A or C) by use of

base to H20 and H200 connectivity in the NOESY spectra

(Fig. 1). Some of the expansions of NOESY spectra are

available as supplementary material.

Assignment of all cross-peaks in DQF COSY andNOESY spectra to specific protons indicates that only

one type of double helical structure exists in aqueous

solution. The two strands are equivalent as only one set

of signals for particular nucleotide unit are observed.

The fact that the right handed DNA strategies for res-

onance assignments have given an internally consistent

result indicates that the overall structure is that of a

right handed DNA. A detailed analysis of NMR spec-tra, however, has provided finer conformational details

with reference to the deoxyribose geometry, glycosidic

bond rotation, and base to base stacking of each residue

along the sequence of the molecule.

Sugar conformation

Deoxyribose pucker may be determined from the

knowledge of three bond proton–proton vicinalcoupling constants, 3J, or from the knowledge of intra-

38 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

sugar inter-proton distances. Due to limited digital res-olution (2.3Hz/point in present case), spectral overlap

and inherent line widths, there are several limitations.

We have used a combination of two methods and made

an attempt to get an estimate of sugar pucker without

doing rigorous calculations and spectral simulations.

Table 2 gives the presence/absence of cross-peaks pat-

terns in DQF COSY spectra at 295K in deoxyribose of

each residue. The DQF COSY cross-peaks pattern forthe rigid conformations have been simulated in litera-

ture using SPHINX-LINSHA (Celda et al., 1989; Sch-

mitz et al., 1990; Widmer and Wuthrich, 1987) and

SICOSS (Majumdar and Hosur, 1992) programs. A vi-

sual comparison of the simulated cross-peak patterns for

various spin systems with the experimental off diagonals

cross-peaks has been made. Absence of H200–H30 cross-peak in T1, G2, T4, and C5 (that is all expect A3 and A6residues), indicates that active coupling is <2.3Hz and

therefore the pseudorotation phase angle lies in the

range 90–225� (Celda et al., 1989; Majumdar and Hosur,

1992; Schmitz et al., 1992; Widmer and Wuthrich, 1987).

Presence of H30–H40 cross-peak pattern for all residues

except G2 residue indicates that P lies in range 90–108�.The P value is thus restricted to the range of 90–108�,i.e., close to O10 endo sugar pucker for most of the

Table 2

Presence (+) and absence ()) of intra-residue cross-peak patterns involving d

295K

H10–H20 H10–H200 H20–

T1 + + +

G2 + + +

A3 + + +

T4 + + +

C5 + + +

A6 + + +

Table 3

Observed values of spin–spin coupling constants, J (Hz), sums of H10, H20,(degrees), and mole fraction of major S-conformer, vS, of deoxyribose proto

J (H10–H20)a J (H10–H200)aP

H10P

H20

325K

T1 8.2 6.0 14.2 —

G2 7.4 7.1 14.5 —

A3 6.7 6.7 13.4 —

T4 8.2 5.7 13.9 29.2

C5 8.2 5.5 13.7 29.2

A6 7.5 6.8 14.3 —

305K

T1 8.6 6.0 14.6 —

G2 9.2 5.7 14.9 —

A3 6.9 6.7 13.6 26

T4 7.8 6.7 14.5 29

C5 7.5 6.6 14.1 29

A6 8.7 5.9 14.6 —

a J values are measured to an accuracy �0.5Hz.bPS and vS values are an average estimate only.

residues. Since the observed H10–H20, H10–H200, andH20–H30 cross-peak patterns do not qualitatively agree

with the corresponding simulated patterns for P �90–

108�, we looked into possibility of existence of rapid

inter-conversion on the NMR timescale between two

sugar conformations observed in several studies on de-

oxyoligonucleotides (Majumdar and Hosur, 1992; Sch-

mitz et al., 1990; Schmitz et al., 1992a,b; Tonelli et al.,

1998; Widmer and Wuthrich, 1987; Wuthrich, 1986)instead of a rigid structure of P angle restricted to a

small domain. The resulting coupling constant, Jobs, can

be described as a weighted average of the two states:

Jobs ¼ vSJS þ vNJN; ð1Þ

where vS and vN are the fractional population of S and

N conformers (vS þ vN ¼ 1), respectively and JS and JNare the proton–proton vicinal coupling constants for S

and N conformers, respectively. Some of the observed

values of J and their sums measured directly from the

1D NMR spectra using J (H20–H200)¼ 14.0Hz andJ (H30–31P)¼ 5.8Hz are given in Table 3. The values of Jand summation of coupling constants at 295K were

found to be close to that obtained at 305K. The ob-

served values ofP

H10 lie in the range 13.4–15.0Hz in-

dicating occurrence of dominant S-type conformer with

eoxyribose sugar protons in d-TGATCA from DQF COSY spectra at

H30 H200–H30 H30–H40

) +

) )+ +

) +

) +

+ +

H200, and H30 coupling constants (Hz), estimate of pseudorotation, PS

ns at 325 and 305K in d-TGATCAP

H200P

H30 vSb PS

b

22.8 17.9 0.77 144–153

— — 0.81 189–198

— 16.3 0.64 180–189

22.3 18.2 0.72 144–162

22.6 18.2 0.68 153–162

— 21.2 0.78 117–126

19.0 — 0.83 126–153

— — 0.87 153–180

— — 0.67 180–189

22.8 — 0.81 126–153

22.3 — 0.75 135–144

— — 0.83 153–180

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 39

vS P 0:63. The value of vS is estimated using the relation(Van Wijk et al., 1992)

vS ¼X

H 0�

� 9:4�=ð15:7� 9:4Þ; ð2Þ

while other couplings have been used to get an estimate

of pseudo rotation phase angle of S-conformer (Van

Wijk et al., 1992). Some of the results are given in Table

3. Since J=P

J are measured to an accuracy of �0.5Hz

(and are not obtained by simulations of spectra), noattempt is made to estimate the values of PN, /N, and /S

and the average value of 18�, 36�, and 36�, respectivelyhave been used. The visual inspection of H10 peaks at

305K shows clearly that G2 is closest to B-DNA

structure and that difference between J (H10–H20) and

J (H10–H200) decreases in the order of G2>A6>T1>T4>C5>A3 which is in agreement with the measure-

ment ofP

H10 and hence calculated values of vS whichare 0.87, 0.83, 0.83, 0.81, 0.75, and 0.67 for six residues

G2, A6, T1, T4, C5, and A3, respectively. But the DQF

COSY at 295K shows that B-DNA character decreases

in the order G2>T1, T4, C5>A6, A3 since in the A3

and A6 residues, both the H200–H30 and H30–H40 cross

peaks are observed. This is due to the fact that J (H200–

H30) and J (H30–H40) values are sensitive to variation in

PS and vS both. Therefore A6 residue has PS value,which is different from that in T1 residue while vS for

both these residues are nearly same. In fact for A3 and

A6 residues, two sets of PS, that is, 99–108�, 180–189�,and 153–180�, 117�, respectively fit into the experimental

results. The value indicated in Table 3, that is 180–189�and 153–180� for A3 and A6 residues, respectively were

selected out of the two sets on the basis of the intensity

of H10–H40 cross-peak in NOESY spectra, discussedlater. We have used the values of PS and vS obtained in

Table 3 to carry out back calculations of all the coupling

constants. It is found that whenever H200–H30 and/or

H30–H40 cross peak pattern was not observed in DQF

COSY spectra at 295K, the calculated value of J was in

the range 1.5–2.2Hz which isbelow the experimental

resolution of 2.3Hz/point. This shows that the estimates

of PS and vS obtained on the basis of 1D NMR spectraare consistent with the DQF COSY spectra.

It is generally believed that the non-terminal pyrimi-

dines in the B-DNA conformation are characterized by

the presence of major S-conformer with vS ¼ 0:75–0.90while the mole fraction of S-conformer for purines tends

to be higher, being �0.90 to 1.00 (Searle, 1993). The

results obtained are contrary to this belief. We observed

that A3 residue has vS ¼ 0:67, a value much lower thanexpected (0.90–1.0). Similar result was also obtained in

our studies on DNA hexamer d-CGATCG (Barthwal

et al., 2003) in which A3 residue had vS lower than 0.9–

1.0; the value of vS was 0.83. Further the present results

show that vS in d-TGATCA (at 305K) decreases in

the order G2>A6, T1>T4>C5>A3 while that in

d-CGATCG is found to decrease in the order G2>A3,T4>G6, C5, C1. The B-DNA character, as judged on

the basis of presence/absence of H200–H30 and H30–H40

DQF COSY peak patterns in d-TGATCA decreases in

the order G2>T1, T4, C5>A3, and A6 while that in d-

CGATCG (Barthwal et al., 2003) decreases in the order

G2>A3, T4>G6, C1, and C5. This clearly shows se-

quence dependent variations in conformation of oligo-

nucleotides. The G2 residue in both the DNA sequencesd-CGATCG and d-TGATCA, happens to possess the

maximum B-DNA character. It is an interesting obser-

vation in view of the fact that the anticancer drugs

adriamycin, daunomycin and epiadriamycin, bind

maximally to d-C/TpG step with direct hydrogen bonds

with G2 residue (Chaires et al., 1987; d’Langlois

Estaintot et al., 1992; Leonard et al., 1992, 1993; Nunn

et al., 1991). We obtained vS fromP

H10 values of allresidues as a function of temperature in the range 295–

325K (part of data are shown in Table 3). It was found

that percentage of S-conformer decreased with temper-

ature in all the residues in d-TGATCA. The decrease

was most significant for T4 residue (DvS � 0:09) and was

minimum for A3 residue (DvS � 0:04). Thus the equi-

librium N$S between the two inter converting con-

formers is affected by temperature and shifts towardsN-conformer at high temperatures. This question has

been raised by Van Wijk et al. (1992) but there are no

explicit reports on change in N$S equilibrium with

temperature in oligonucleotides in literature. NMR

spectroscopy methods are most suitable to address this

aspect. In d-CGATCG we observed that the terminal

G6 residue showed maximal change in vS (�0.09)

while pyrimidines were less affected by temperature.These results once again highlight the sequence de-

pendent subtle variations in conformation of oligonu-

cleotides.

The deoxyribose sugar conformation can also be de-

termined from intra residue inter-proton distances in-

volving sugar protons (Tables 4 and 5). Except for

distances H10–H40 and H200–H40, all other distances donot show significant variation with PS and vS (Wuthrich,1986). The H10–H40 distance is about the same (3.0�A)

for P ¼ 18� and 162� and least (2.0�A) for P ¼ 99–108� sothat it can be used as a marker for PS and not vS. Itincreases gradually from 3.0 to 3.8�A as P is varied from

162� to 225�. Therefore PS for T1, T4, and C5 residues

having a distance of 2.7�A is estimated as �144� while

for G2, A6, and A3 residues PS ¼ 162�, 171�, and 180–

189�, respectively (distance H10–H40 increases in theorder G2>A6>A3). This excludes the possibility of

PS ¼ 99–108� and PS ¼ 117� for A3 and A6 residues,

respectively. The H200–H40 distance decreases from 3.9�Afor PS ¼ 198�/189� to 3.8�A for PS ¼ 162�, and further to

2.3�A for PS ¼ 18�/9� (Wuthrich, 1986). Therefore both

decrease of PS as well as vS will influence the intensity of

the cross-peak in the same direction, that is, to reduce

Table 5

Some of the inter-proton distances (�A) from inter-residue sequential

connectivities observed in NOESY spectra of d-TGATCA at 295K

T1H10–G2H500 3.1 A3H8–T4CH3 2.9

T1H10–G2H8 3.4 T4H10–C5H500 3.4

T1H20–G2H8 >4.2 T4H10–C5H6 3.5

T1H200 –G2H8 3.6 T4H20–C5H6 3.4

G2H10–A3H500 3.6 T4H30–C5H6 4.0

G2H10–A3H8 3.1 T4H20 –C5H6 2.4

G2H20–A3H8 3.4 T4H200 –C5H5 3.1

G2H200 –A3H8 2.3 T4H200 –C5H5 4.0

A3H10–T4H500 3.1 T4H6–C5H5 3.7

A3H10–T4H6 3.3 C5H10–A6H500 3.1a

A3H20–T4H6 3.2 C5H10–A6H8 3.5

A3H200 –T4H6 2.9 C5H20–A6H8 4.2

A3H20–T4CH3 2.9 C5H200–A6H8 3.7

aOverlap of resonance.

Table 4

Some of the inter-proton distances (�A) from intra-residue connectivities observed in NOESY spectra of d-TGATCA at 295K

T1 G2 A3 T4 C5 A6

H10–H20 2.9 — 2.8 2.7 3.1 2.8

H10–H200 2.2 2.2 2.4 2.2 2.2 2.4

H10–H30 — 3.4 3.4 — — 3.8

H10–H40 2.7 2.9 3.1 2.7 2.7 3.1

H200–H30 2.7 2.6 2.5 2.8 2.9 2.5

H200–H40 3.2 — — 3.2 3.2 3.7

H8/H6–H10 3.2 3.3 3.2 3.3 3.4 3.1

H8/H6–H20 2.4 2.2 2.2 2.1 2.2 2.4

H8/H6–H200 3.2 3.4a 3.2 2.8 3.4 3.3

H8/H6–H30 3.6 3.3 3.2 3.7 3.5 3.2

H8/H6–H40 — — 4.0 — —

H8/H6–H50 3.7 — — — — —

H8/H6–H500 — — 3.3 — — 3.2

aOverlap of resonance.

40 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

the H200–H40 distance. The observed distance decreases

in the order G2>A3>A6>T1, T4, and C5. This cor-

roborates the results shown in Table 3. Since G2 residue

has vS ¼ 0:87 (maximum value) and PS ¼�162�, it is

expected to have maximum distance. A3 has larger

distance due to higher P value, being 180–189�. The

other residues are showing an effect due to combination

of both PS and vS. The difference in distance of T1 andA6 residue having same vS ¼ 0:83 is due to the lower

average P value in T1 (144�) as compared to corre-

sponding value for A6 (171�) residue. The distances

H200–H30 increases gradually from 2.5 to 2.9�A on either

decreasing vS or PS. Our experimental results show that

A3 and A6 residues have most intense H200–H30 cross-peak having a distance of 2.5�A which may be attributed

to a value of PS being 180–189�. The distance of 2.6�Afor G2 residue is reasonable if vS ¼ 0:87 and PS ¼ 162�.The distance of 2.7, 2.8, and 2.9�A for T1, T4, and C5

residues, all having PS �144� is directly related to ob-

served decrease in vS to values of 0.83, 0.81, and 0.75 for

T1, T4, and C5, respectively (Table 3). Thus it was

found that by and large the distance analysis from

NOEs gives results, which are consistent with that ob-

tained from spin–spin coupling data.

Glycosidic bond rotation

The glycosidic bond rotation, v, may be estimated

from the distance of base H6/H8 proton to different

sugar protons (Wuthrich, 1986). Among purines and

pyrimidines, the A6 and T1 residues are found to have

lowest distance H8/H6–H10 and may adopt high anti

conformation. The base–H20 distance does not vary

appreciably with v. The base–H200 distance could not be

used to ascertain v as it also varies with vS. The very factthat we see base–H30 cross-peaks in NOESY spectra,

indicates that mixed N- and S-conformers are present in

deoxyribose sugar since these distances for B-DNA are

>4.2�A (the resolution in our experiment).

Base to base stacking

The observed chemical shifts and relative intensities

of some of the cross-peaks in NOESY spectra showsome typical trends which are characteristic of good

base to base stacking. The H10 protons of G2 residue are

shifted upfield to a greater extent as compared to that in

A3 and A6 residues. This is consistent with the general

observation that purines produce large ring current

shifts as compared to pyrimidines. For the same reason

T1 and C5 residues are more upfield shifted (both ap-

pearing at 5.90 ppm) as compared to T4 residue (ap-pearing at 6.04 ppm). The cross-peaks, A3H8–T4CH3

and T4H6–C5H5 cross-peaks are clearly seen in NO-

ESY spectra. These observations are indicative of good

base to base stacking at these base pair steps. The py-

rimidine H20 protons are shifted more upfield than the

corresponding purine H20 protons. The H20 protons ofpyrimidines resonate in the range 1.76–2.02 ppm while

that of purines resonate in the range 2.76–2.79 ppm.Also the difference in chemical shifts of H20 and H200

protons of pyrimidines is much larger (0.1–0.51 ppm)

than that for purines (0.05–0.23) ppm. These general

guidelines observed in literature (Chary et al., 1987;

Feigon, 1992; Hilbers and Patel, 1975; Patel, 1977;

Van de Van and Hilbers, 1988) are helpful in spectral

assignments.

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 41

Restrained molecular dynamics studies

Restrained molecular dynamics permits the system

to undergo conformational and momentum changes so

that different parts of the phase space accessible to the

molecule can be explored and stable conformations are

identified by energy minimization. Simulations were

carried out by starting with two different structures: B-

DNA built using INSIGHT II and other built usingAMBER 7.0 software package. Simulated annealing

was carried out up to a temperature of 500K but it

was found that convergence to a final structure could

be readily achieved even at 300K. The stereo view of

various rMD structures is shown in Fig. 2A. The rMD-

B1SA structure superimposed over the standard B-

DNA structure is shown in Fig. 2B indicating the

difference between the two structures. Several struc-tures obtained after equilibration at different time in-

tervals during the dynamics run at 500 and 300K were

examined. It was observed that they differ only mar-

ginally from each other in their overall feature irre-

spective of the rMD protocol followed and starting

structure. Further no significant drift in either potential

energy or restraint deviations was observed during the

final equilibration. It can therefore be concluded thatthe system reached a minimum energy conformation. A

structure obtained without introducing NMR con-

straints and by following same steps of rMD protocols

was found to be quite different. This confirms that the

structure is indeed defined by experimental restraints

and not the refinement procedure or variables used.

The root mean square deviation in any of the rMD

Fig. 2. Sequential stereo view of the: (A) various rMD structures and (B) rMD

structure.

structures and either of the starting structures is quitelarge but among various final structures was very low.

This is generally acknowledged as an indication that

convergence has been achieved. Table 6 indicates an

assessment of refined structures after equilibration (at

the end of 25/100 ps) in terms of energetics including

restraint violations energies and root mean square de-

rivative of energy with respect to atomic coordinates.

The total potential energy of the rMD-B1, rMD-B1SA,rMD-B2, rMD-B2SA, and rMD-B1-100 structures lie

in the range of 404–428 kcalmol�1 which is signifi-

cantly lower than the corresponding energies of initial

B-DNA structures (695–709 kcalmol�1). The forcing

potential, which indicates contribution to potential

energy due to violations of both experimental distances

and torsional angle data, exhibits a decrease from 934–

2858 to 198–200 kcalmol�1 after restrained energyminimization and restrained molecular dynamics. The

energy gradient with respect to atomic coordinates lies

in the range 0.07–0.17 kcalmol�1 �A�1 in various struc-

tures. A summary of experimental restraints and sta-

tistical analysis of family of structures generated by

restrained molecular dynamics (rMD) is shown in

Table 7.

Conformational features of rMD structures

All helical parameters, backbone torsional angles,

and sugar conformations of the resulting rMD struc-

tures were thoroughly analyzed with the program

CURVES, version 5.1 (Lavery and Sklenar, 1996). The

structural characteristics of various rMD structures are

discussed in the following.

-B1SA structure of d-TGATCA superimposed with standard B-DNA

Table 6

Energy terms (kcalmol�1) and root mean square gradient of energy with respect to atomic coordinates (kcalmol�1 �A�1) for starting models and rMD

structures of d-TGATCA

Structure Total Bond Angle Dihedral Vdw Repul-

sion

Disper-

sive

Electro-

static

Re-

straintaGradi-

entbDdavc

B-DNA1 695 74 466 105 343 1607 )1264 )295 934 25.16 0.55

rMD-B1 422 75 475 133 14 957 )942 )280 200 0.117 0.24

rMD-B1SA 424 75 476 132 15 953 )938 )278 199 0.133 0.27

rMD-B1-100 428 75 484 131 11 954 )942 )276 195 0.070 0.25

B-DNA2 709 98 451 106 343 1606 )1264 )289 2858 37.74 0.58

rMD-B2 414 83 473 132 17 945 )928 )294 198 0.077 0.24

rMD-B2SA 404 81 472 128 14 954 )940 )293 199 0.171 0.26

a This contribution has been subtracted from the total energy.bRoot mean square gradient of energy with respect to atomic coordinates (kcalmol�1 �A�1).c Average distance deviation.

Table 7

Summary of experimental restraints and statistical analysis of family of structures generated by restrained molecular dynamics (rMD)

Parameter Residue No. of restraints

Distance restraints

Intra-residue T1 17

G2 13

A3 19

T4 15

C5 15

A6 16

Inter-residue T1–G2 4

G2–A3 4

A3–T4 6

T4–C5 8

C5–A6 4

Dihedral restraints 12

Average pairwise RMSD B-DNA1¼ 0, rMD-B1¼ 0.87, rMD-B1SA¼ 0.97, rMD-B1–100¼ 1.01, B-DNA2¼ 0, rMD-

B2¼ 0.85, rMD-B2SA¼ 0.87.

Average residuewise RMSD T1¼ 2.51, G2¼ 1.28, A3¼ 1.04, T4¼ 1.06, C5¼ 0.92, and A6¼ 0.90.

42 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

Helicoidal parameters

Plot of various helix axis parameters (global, unless

specified otherwise) of d-TGATCA in different struc-

tures as a function of residue position in duplex is shown

in Fig. 3 along with that for two classical structures of

A-DNA and B-DNA. Base sequence dependent varia-

tions in helicoidal parameters are evident. The results on

rMD structures are closer to that of B-DNA as com-pared to that of A-DNA. The helicoidal parameters are

classified into three categories: global base pair-axis

parameters, intra-base pair or global base–base param-

eters and the inter-base pair or base pair-step parame-

ters Among the base pairs-axis parameters, the

x-displacement (dx) are �)1.0�A for all residues which is

close to a value of )0.7�A as seen in canonical B-DNA

structure. The y-axis displacement (dy) varies from +0.2to )0.2�A along the sequence. The base pairs are inclined

(g) at an angle of )5� to )7� with no significant variation

with base sequence. The tip angle (h) fluctuates from )5�to )7� at 50 end to �+7� at 30 end. Such variations have

been reported in literature in deoxyoligonucleotides

(Mujeeb et al., 1993; Weisz et al., 1994). Among the

intra-base parameters, the shear (Sx) and stretch (Sy)values do not vary very much from their ideal values.

The shear lies within � 0.2�A while the shear lies with in

the range 0 to )0.2�A. The stagger (Sz) values lies withinthe range 0.1–0.4�A but have large positive values for the

two middle base pairs. The buckle varies within �0.3�A.

The variations in the propeller twist are significantly

large. For the two base pairs in the middle, that is, A3T4and T4A3, large negative values in the range 23–28� areobserved. The base pairs at either end show a small

positive propeller twist. Large negative values of pro-

peller twist in AT and TA base pairs have been reported

in literature (Dornberger et al., 1998; Mujeeb et al.,

1993; Weisz et al., 1994) and may possibly occur to

avoid steric clashes between the CH3 group of thymine

and 50 neighboring sugar in AX/XT base pairs step(Hunter, 1993). The positive value of propeller twist for

terminal base pairs may be attributed to fraying effects

at both ends of oligonucleotides (Fig. SI 4, supplemen-

tary information). The base pair opening lies between

)8� and )17� for all base pairs and is larger in magni-

tude for non-terminal base pairs.

Fig. 3. Helical parameters for d-TGATCA calculated for canonical A-DNA (—), B–DNA (- - - -) and different structures obtained by restrained

molecular dynamics simulations: rMD-B1 (d), rMD-B1SA (s), rMD-B2 (j), rMD-B2SA (�), and rMD-B1-100 (n).

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 43

In regular A-DNA and B-DNA geometries, global

values of the inter-base pair parameters shift (Dx), slide

(Dy), roll (q), and tilt (s) are essentially zero. For dif-

ferent rMD structures, the shift does not show signifi-

cant variations with the base sequence. The slide is

minimum for central base pair step, A3pT4, and about0.2�A for all other base pair steps. The rise per residue

(Dz) is about �3.2�A for all base pair steps except

A3pT4, for which it is somewhat large, being �3.7–

3.8�A. The tilt angle fluctuates within �4� and does not

appear to be sequence dependent. The roll angle, on the

other hand, shows large positive value of �8–12� at ei-ther end of the helix and is )8� to )9� at the central

A3pT4 base pair step. A positive roll opens the angle

between base pairs towards the minor groove, as a result

a wider minor groove and bending towards major

groove causing a curvature in helix axis occurs. Thelarge positive roll at T1pG2 steps indicates reduced base

stacking and hence demonstrates a flexibility at pyrim-

idine 30–50 purine step (Dornberger et al., 1998) which is

discussed later in detail. But the negative value of roll

angle at the A3pT4 step of helix indicates widening of

Fig. 4. Backbone torsional angles calculated for canonical A-DNA (—),

B–DNA (- - - -) and different structures of d-TGATCA obtained by

restrained molecular dynamics simulations: rMD-B1 (d), rMD–B1SA

(s), rMD-B2 (j), rMD-B2SA (�), and rMD-B1-100 (n).

44 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

major groove and narrowing of minor groove at thesesteps which is compensated by decrease in propeller

twist so as to prevent destacking of bases (Dornberger et

al., 1998). The twist angle for all base pairs in different

structures lies in the range 34–36� except the central basepair in which it is �37�. The CURVES software reports

both the global and local base helical parameters, and a

correlation between them is expected (Lavery and

Sklenar, 1989). The local helical parameters show trendsin roll and twist angle similar to that observed in global

parameters. The global axis curvature obtained for

various rMD structures shows that helix axis bends by

an angle of 5–6� at each end, which is consistent with the

observation of positive roll at these steps. The overall

helix bend leading to curvature is �5�.

Groove width

The groove widths of the double helix are defined

using the coordinates of the phosphate atoms (Bhat-

tacharya and Bansal, 1992; Mujeeb et al., 1993). In the

hexanucleotide only one major and one minor groove

exists in the middle of the helix. The width and depth of

major groove are 12.2–12.3 and 3.8–4.2�A, respectively.

The corresponding values for minor groove are 6.0–6.4

and 4.5–4.7�A, respectively. Thus the major groove iswider by �1.2�A. This corroborates with the negative

value of roll angle at A3pT4 step leading to widening of

major groove which is compensated by a large decrease

in propeller twist, observed at this step. The width and

depth of minor groove, on the other hand increases to a

lesser extent.

Torsional angles

The variation of torsional angles with base sequence

(Fig. 4) show equivalence of strands of duplex. The

general features and trends with base sequences for all

rMD structures are strikingly identical and a difference

in absolute values of torsional angles with different

starting structures or different rMD protocols (with and

without simulated annealing) are slightly different for

end residues only. This further confirms that all rMDstructures are uniquely defined by the restraints ob-

tained by the NMR spectra. The backbone torsional

angles (except d) are not defined directly by NOE dis-

tance constraints. We have used a low force constant to

permit a smooth search, which is conformationally

compatible with other structural features. It is observed

that the angles a through f show variations with base

sequence but have values close to that of B-DNA formost residues rather than the corresponding A-DNA

structure. The angle a varies in a narrow range )60� to79� and is largest in magnitude for T4 base in all

structures. The angle b does not vary appreciably with

base sequence and adopts a trans-conformation. The

variation of c is in the range 50–70� except that at 50 endit is unusually large; it adopts trans-conformation, which

is discussed later. The angle d reflects the deoxyribose

puckering and adopts a conformation between that of

A-DNA and B-DNA. In most of the rMD structures

obtained, it is lowest for T1 residue (119–122�) and

highest for A6 residue (�130�) although the variation is

within a small range. The angle e is in trans-conforma-

tion for all residues while n is close to the corresponding

value in B-DNA in all rMD structures. The analysis ofspin–spin coupling constant and inter-proton distances

reveal that deoxyribose sugars are in dynamic equilib-

rium between two conformers and the dominant

S-conformer has pseudo rotation angle P in the range

126–180� (Table 3). The value of P angle in five struc-

tures resulting from our restrained rMD simulations is

essentially consistent with that and shows that it lies in

the range 138–178�. In most of the rMD structures T1residue had minimum P value (�138–140�) while A6

residue had maximum value (�178�). T1 invariably

adopted C10-exo conformation. Further in case of A3

residue, NMR results yield P ¼ 180–189� with vS ¼ 0:67(Table 3) while the rMD studies give corresponding

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 45

value of P as 164�, respectively. Thus, in case of residuewith high population of minor conformer, average tor-

sional angles fail to provide adequate structural de-

scription of sugar conformation. The average value of

pseudo rotation phase angle and hence the sugar puck-

ering differ for the purine and pyrimidine nucleotides;

the average value being 164� and 146�, respectively. Thepyrimidine nucleotides thus prefer smaller phase angle

as compared to purines (Weisz et al., 1994). Sugarpucker and glycosidic torsional angles are typically the

parameters best defined by NMR and define the geom-

etry within a nucleotide. It is observed that glycosidic

bond rotation in all structures lies in the range )94� to133�. At the 50 end of the helix there is tendency for v to

adopt high anti-conformation of �)94� in all structures,

which was also manifested by the lowest base H6–H10

distance observed in T1 residue in NOESY spectra.

Structural features of TpG/CpA base pair step

Structural flexibility of DNA at TpG/CpA base pair

step has been reported by NMR (Cheung et al., 1984;

Donlan and Lu, 1992; Dornberger et al., 1998; Weisz

et al., 1994), X-ray crystallographic (Gorin et al., 1995;

Grzeskowiak, 1996) and other techniques (Beutal and

Gold, 1992; Lyubchenko et al., 1993) in literature. Astriking feature of X-ray crystallographic structures is a

large twist angle of 50� observed at TpG/CpA step

(Gorin et al., 1995; Grzeskowiak, 1996). As a conse-

quence of high twist and slide values, the inter-proton

distance, T3H200–G4H8 and C7H200–A8H8 at these steps

in decanucleotide d-CATGGCCATG were found to be

5.2�A. Further, the distance between base protons H8/

H6 at TpG and CpA base steps is P6.2�A (Grzeskowiak,1996). Alternatively two subsets, CAþ and CA�, withmean twist value of 30.5� and 49�, respectively have beenshown to exist in the X-ray crystal structures of DNA

samples (Gorin et al., 1995). Our NMR results do not

give any evidence of the existence of these structural

features. We have examined time-dependent fluctuations

of several inter-proton sequential distances for the TpG/

Table 8

Comparison of some of the structural features of TpG and CpA base pair s

Present NMR results NMR

(Dornberger et al.)

TpG CpA TpG/CpA

Twist (�) 34.6 35.3 34.9, 34.8

Slide, dy

(�A)

0.13 �0.13 0.21, 0.24

Roll (�) 8–12 9–10 10–15

B-200 (�A) 3.6 3.7 3.1, 3.4

B-10 (�A) 3.4 3.5 —

B-20 (�A) >4.2 4.2 4.6, 4.5

B-B (�A) 5.1a 5.3a —

B, base proton H8/H6.a From rMD-B1SA structure.

CpA base pair step during 100 ps rMD simulations withdifferent starting geometries. We found that the varia-

tion in T1H200–G2H8 and C5H200–A6H8 lies within

�0.3�A. The twist angle at TpG/CpA step in several

conformers saved at different time intervals in an rMD

simulation showed deviation within �4�. These are in

agreement with those obtained in literature by NMR

techniques (Dornberger et al., 1998; Weisz et al., 1994).

We have observed NOE cross-peaks for sequential NOEconnectivities H8–H10, H20, and H200 at CpA step and

H8–H10, H200 at TpG step even at low mixing times with

distances lying in the range 3.4–4.2�A (Table 8). The

twist, roll, and slide values in various rMD structures

are �35�, 8–12�, and )0.13�A, respectively. These results

compare well with those obtained by NMR techniques

(Dornberger et al., 1998; Weisz et al., 1994) and a twist

angle of �34� reported by electrophoretic techniques(Kabsch et al., 1982).

The NMR conformers have been characterized by a

large positive roll angle at TpG step resulting in opening

towards a major groove and a concomitant local ex-

pansion of minor groove (Weisz et al., 1994). This in-

creases the inter-strand distance between A3H2 and

C5H10 protons to a value >4.5�A compatible with ab-

sence of cross-peak in the 2D NOESY spectra. We didnot observe NOE cross-peak for any of the inter-strand

or intra-strand inter-proton distances A3H2–C5H10,A6H2–G2H10, and A3H2–T4H10 (Chuprina et al.,

1991), the corresponding distance is apparently >4.2�A,

the resolution in our NOESY spectra. We observed that

large positive roll of 9–12� at TpG step is accompanied

by a tilt of 3–4� in our structures, which is consistent with

the Tilt-Roll plot relationship at TpG step shown earlier(Dornberger et al., 1998). Large positive roll angle up to

�20� were observed in several NMR structures of DNA

duplexes at the TpG steps (Dornberger et al., 1998;

Mujeeb et al., 1993; Weisz et al., 1994) and were also

predicted by Monte Carlo Calculations (Zhurkin et al.,

1991) emphasizing the unusual conformational behavior

and flexibility of this pyrimidine 30–50 purine step. The

teps

NMR

(Weisz et al.)

X-ray

(Grzeskowiak et al.)

X-ray

(Gorin et al.)

TpG/CpA TpG/CpA TpG/CpA

36, 40, 37 50 30.5 CAþ

49.0 CA�

0.1–0.1 High —

20–5, 4 — —

— 5.2 —

— — —

— — —

— P 6.2 —

46 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

preference of positive roll angle in TpG/CpA base pairstep to avoid a cross-strand steric clash in minor groove

has also been shown (Hunter, 1993). The observed

stacking patterns between base pairs at five individual

base pair steps viewed along the global helix axis for

rMD-B1-100 structure are shown in Fig. 5. It is noted

that overlap of bases at TpG step is particularly low. The

high positive roll at TpG step indicates reduced base

stacking interactions which leads to a non-parallelism ofthe base normal vectors of residues T1 and G2. High

flexibility and low stacking interactions have been de-

duced earlier by gel circularization assays (Lyubchenko

et al., 1993), gel-electrophoresis (Beutal and Gold, 1992),

and NMR studies (Cheung et al., 1984; Donlan and Lu,

1992; Dornberger et al., 1998). These results suggest that

the high flexibility at TpG/CpA step in solution can be

explained in terms of large positive roll rather than twistbased low stacking interaction. The unusually large twist

angle (49–50�) observed in X-ray crystal structures

(Gorin et al., 1995; Grzeskowiak, 1996) may be attrib-

uted to the crystal packing forces which disturb the local

conformation at TpG step. This conformational behav-

ior at TpG/CpA step has been shown to cause an upfield

shift in CH10 resonance of TpG/CpA junction along with

a pronounced temperature dependence (Weisz et al.,1994). The C5H10 resonance in our experiments shows a

change in chemical shift of 0.33 ppm with temperature

which is considerably larger than that observed for H10

protons of other residues.

The backbone torsional angles do not vary signifi-

cantly with the residue and are largely constrained to

preferred conformations observed generally in B-DNA.

However it was observed that the c torsional angleabout C50–C40 bond for T1 residue was �160� for most

rMD structures and �90� for few structures (Fig. 4),

indicating a large deviation from +gauche conformation

generally found to exist in B-DNA. The corresponding

torsional angles b of T1 residue was found to be in trans

conformation. In order to examine this aspect, we

monitored the variations of all the torsional angles with

time in rMD run up to 100 ps. It was found that only

Fig. 5. Stereo view of five individual base pair steps viewed down the

global helix axis showing stacking between adjacent base pairs in

d-TGATCA.

three torsional angles, that is, a, b, and c, for T1 and C5residues showed a significant variation with time

(Fig. 6A). The angle c for T1 residue rapidly flipped

between two conformations having torsional angle of

90� and 160�. The corresponding angle b showed fluc-

tuations between two conformational states having

torsional angle of )180� and )85�. It is a concerted

change from b(t) c(gþ) to b(g�) c(t) conformational

state. On the other hand, the time-dependent fluctua-tions in C5 residues exhibited an altogether different

behavior. The angles a, b, and c were )80� (g�),)170�(t), and +60�(gþ), respectively for first 75 ps and

then flipped to )150�(t), )90�(g�), and +90�(gþ), re-

spectively and stabilized at these torsional angles for

next 25 ps run. The conformation of backbone thus

changed from a state normally found in B-DNA, that is,

a(g�) b(t) c(gþ) to another state characterized by a(t)b(g�) c(gþ) for C5 residue. Corresponding to these

variations, small shift �10� was observed in d and eangles after 75 ps. Thus there is a bimodal distribution

involving two distinct conformational states (named as 1

and 2 in Table 9). It may be noted that a somewhat

similar behavior has been reported for G residue in the

TpGpC element in a decanucleotide earlier (Weisz et al.,

1994). It was found that a and c undergo a stronglycoupled reversible motion during rMD simulations from

trans to more common a(g�) and c(gþ) conformation,

respectively while b remained unchanged in trans-state

for central G residue in TpGpC element. The G6 residue

thus shifted from a(t) b(t) c(t) to a(g�) b(t) c(gþ) con-formation. The effect of backbone perturbations ap-

parently extended to the 30 linked C residue in TpGpC

element, which exhibited single conformation but a largestandard deviation. Similar trans-conformation has been

observed in CpG step in a crystal structure of self

complementary A-DNA octamer and its 30 methylene

phosphate analogue (Olson and Lu, 2001). The crystal

structure of topoisomerase II poison 9-amino-[N-(2-

dimethylamino)ethyl]acridine-4-carboxamide bound to

d-CGATCG (Adams et al., 1999) also shows that c for

G2 residue (at the site of intercalation) exists in trans-conformational state.

The change in torsional angles c, indicating rotation

about C50–C40 bond, from +gauche to trans-state brings

H50 and H500 much closer to the base H8/H6 and H30

protons, respectively (Weisz et al., 1994). In the c(t)state, the substituents H50, H500, and O50 atoms at the

C50 atom position remain staggered and hence are in

energetically favorable conformation. The intra nucle-otide distances (d) for a(g�) b(t) c(gþ) conformational

state occurring normally in B-DNA are: d(H8/H6–H50)>4.2�A and d(H30–H500) �3.6�A. When c angle adopts

trans-conformation the distances are considerably re-

duced; being d(H8/H6–H50) �3.4�A and d(H30–H500)<3.0�A. Our experimental results show cross-peaks

corresponding to H30–H500 NOE connectivities for all

Fig. 6. Trajectory analysis of 100 ps restrained molecular dynamics simulations in rMD-B1-100 structure of d-TGATCA showing variations in: (A)

torsional angles a, b, c, d, �, and f in each nucleotide residue: T1, magenta; G2, green; A3, black; T4, blue; C5, red; A6, orange. (B) Intra nucleotide

inter-proton distances H8/H6–H10, H8/H6–H50, H10–H40, H200–H40, and H30–H500 as well as inter-nucleotide inter-proton distance H8/H6–H200 atdifferent base pair steps: T1pG2, green; G2pA3, blue; A3pT4, red; T4pC5, black; C5pA6, orange. (For interpretation of the references to color in this

figure legend, the reader is referred to the web version of this paper.)

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 47

residues. Some of the distances could not be ascertained

accurately due to overlap of peaks but they were foundto lie in the range 3.0–3.6�A. On the other hand, the

NOE cross-peak corresponding to base H8/H6–H50 wasobserved only for T1 residue, the distance being 3.7�A(Table 4). This is consistent with the results on rMD

simulations which show that c shifts from gþ to t state

for T1 residue only. For C5 residue it alters by �30� butremains close to g+ state. Unfortunately there are no

direct inter-proton distances in our NOESY spectra,

which could ascertain the change in conformation dueto change in torsional angle a and b. It may thus be

concluded that due to conformational flexibility at TpG/

CpA steps, the T1 and C5 residues change from b(t)c(gþ) to b(g�) c(t) and a(g�) b(t) c(gþ) to a(t) b(g�) c(gþ)states, respectively.

In order to understand the flexibility of backbone in

d-TGATCA, we examined the inter-proton distances of

Table 9

Two conformational states of T1 and C5 residues in d-TGATCA observed in 100 ps restrained molecular dynamics simulations

Conformer H10–H40

(�A)

H200–H40

(�A)

H6–H50

(�A)

H30–H500

(�A)

a(�)c

b(�)c

c(�)c

d(�)

e(�)

v(�)

T1a 1 3.1 3.8 4.0 3.6 — )180 (t) 90 (gþ) 140 170 )89.72 2.9 3.7 3.9 3.2 — )85 (g�) 160 (t) 120 160 )86.1C5b 1 3.1 3.7 4.8 3.7 )80 (g�) )170 (t) 60 (gþ) 125 178 )105.32 2.9 3.6 5.2 3.8 )150 (t) )90 (g�) 90 (gþ) 115 170 )99.0aConformers 1 and 2 correspond to conformational states adopted during time intervals 0–40 ps and 41–72ps, respectively.b Conformers 1 and 2 correspond to conformational states adopted during time intervals 1–75 ps and 76–100 ps, respectively.c The conformational state occurring normally in B-DNA is a(g�), b(t), and c(gþ).

48 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

various intra-residue and inter residue base to sugar

protons in rMD simulations up to 100 ps. The base H8/

H6 to H10 distance within residue depends only upon

glycosidic bond rotation, v. This distance showed neg-

ligible variations with time (Fig. 6) and therefore v does

notchange appreciably. The base H8/H6 to H20, H200,and H30 distances within a residue are a function of

pseudorotation P and v. They are also found to be verystable with time. The distances H10–H20, H10–H200, H10–H30, H20–H30, and H20–H40 do not show any variation

with time for any of the residues. The distances H10–H40

and H200–H40 show noticeable change for only T1and C5

residue (Fig. 6). For T1 residue, distances H10–H40 andH200–H40 fluctuate within the range 3.1–2.9 and 3.8–

3.7�A, respectively. As discussed earlier, the distance

H10–H40 is a marker for PS value and does not vary withmole fraction of N- or S-conformer while distance H200–H40 decreases significantly with increase in mole fraction

of N-conformer. The observed distances, H10–H40 andH200–H40, for T1 and C5 residue in two conformational

states during rMD simulations (Table 9) correspond to a

change in P from 162�/180� to �153� (also reflected as

variation in d, Fig. 6) and a mixing of 5–10% of N-

conformer to the predominant S-conformation. Thus itmay be inferred that P and vS are flexible only to a

limited extent. The corresponding change in distance

H8/H6–H50 and H30–H500 during rMD simulations for

T1 residue show that torsional angle c does take up

trans-conformation as these distances are �3.9 and

3.2�A at time interval of 41–72 ps. The distance base H8/

H6–H50 for C5 residue however is large in both the

states during 1–75 and 76–100 ps (Table 9) and thereforethere is no evidence of existence of trans-conformer for

angle c. Further it is noteworthy that the base H8/H6–

H200 distance at various base pair steps monitored with

time shows corresponding fluctuations/flexibility only at

TpG and CpA steps (Fig. 6).

The trajectory analysis thus shows that the amplitude

and frequency of motions are different at various sites in

d-TGATCA. The orientation of base with respect tosugar and the sugar puckers themselves do not show

variation through the course of simulation. The rela-

tively large variation in sequential distance and back-

bone torsional angles a, b, and c at TpG and CpA steps

show that backbone is relatively more dynamic. This

cannot be attributed to the end effect (due to fraying at

either 50 end or 30 end of the helix) as in that case dis-

tances and torsional angles corresponding to A6 residue

would be expected to show variation with time. The

flexibility at TpG and CpA base pair step is corrobo-

rated by experimental results obtained by NMR, which

correspond to the conformational averaging over dif-

ferent states. The complexities and variability in themotion in DNA contribute to the sequence dependent

flexibility and deformability of DNA. The conclusions

drawn have important implications in protein–DNA

and drug–DNA interactions. Our findings are particu-

larly significant in view of the occurrence of TpG/CpA

element in several DNA sequences involved in regula-

tion of gene expression as well as binding to ligands/

drugs (Bacolla et al., 1997; Chastain and Sinden, 1998;d’Langlois Estaintot et al., 1992; Kabsch et al., 1982;

Kemler et al., 1989; Leonard et al., 1992, 1993; Ly-

ubchenko et al., 1993; Nunn et al., 1991; Sarai and

Takeda, 1989; Schultz et al., 1991; Trifonov and Bren-

dal, 1986; Weisz et al., 1994). The existence of torsional

angle c in trans-state (instead of conventional +gauche

state) in a specific drug–DNA complex (Adams et al.,

1999) at the ligand binding site only, further demon-strates that studies on sequence dependent variations are

directly relevant to recognition of binding sites at DNA.

Acknowledgments

The facilities provided by the National Facility for

High Field NMR located at Tata Institute of Funda-

mental Research (TIFR), Mumbai are gratefully ac-

knowledged. The authors are grateful to Dr. R.P. Ojha,

Physics Department, DDU Gorakhpur University for

help in running FELIX and DISCOVER software

program. Financial support from CSIR is gratefullyacknowledged.

References

Adams, A., Guss, J.M., Collyer, C.A., 1999. Crystal structure of the

Topoisomerase II poison 9-amino-[N-(2-dimethylamino)ethyl]

R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50 49

acridine-4-carboxamide bound to the DNA hexanucleotide

d(CGATCG)2. Biochemistry 38, 9221–9233.

Arcamone, F., 1981. Doxorubicin Anticancer Antibiotics in Medicinal

Chemistry. Academic Press, New York.

Bacolla, A., Gellibolian, R., Shimizu, M., Amirhaeri, S., Kang, S.,

Ohshima, K., Larson, J.E., Harvey, S.C., Stollar, B.D., Wells,

R.D., 1997. Flexible DNA: genetically unstable CTG.CAG and

CGG.CCG from human hereditary neuromuscular disease genes.

J. Biol. Chem. 272, 16783–16792.

Barthwal, R., Mujeeb, A., Govil, G., 1994. Interaction of daunomycin

with deoxydinucleotide d-CpG by two dimensional proton mag-

netic resonance techniques. Arch. Biochem. Biophys. 313, 189–205.

Barthwal, R., Monica, Awasthi, P., Srivastava, N., Sharma, U., Kaur,

M., Govil, G., 2003. Structure of DNA hexamer sequence d-

CGATCG by two-dimensional nuclear magnetic resonance spec-

troscopy and restrained molecular dynamics. J. Biomol. Struct.

Dyn. 21, 407–423.

Beutal, B.L., Gold, L., 1992. In vitro evolution of intrinsically bent

DNA. J. Mol. Biol. 228, 803–812.

Bhattacharya,D., Bansal,M., 1992. J. Biomol. Struct.Dyn. 10, 213–226.

Carter, S.K., 1982. Anthracycline Antibiotics in Cancer Therapy.

Nijhof, Boston.

Celda, B., Widmer, H., Leupin, W., Chazin, W.J., Denny, W.A.,

Wuthrich, K., 1989. Conformational studies of d-

(AAAAATTTTT)2 using constraints from nuclear overhauser

effects and from quantitative analysis of the cross peak fine

structures in two-dimensional 1H nuclear magnetic resonance

spectra. Biochemistry 28, 1462–1471.

Chaires, J.B., Fox, K.R., Herrara, J.E., Waring, M.J., 1987. Site and

sequence specificity of the daunomycin–DNA interaction. Bio-

chemistry 26, 8227–8236.

Chaires, J.B., Herrera, J.E., Waring, M.J., 1990. Preferential binding

of daunomycin to 50TACG and 50TAGC sequences revealed by

footprinting titration experiments. Biochemistry 29, 6145–6153.

Chary, K.V.R., Hosur, R.V., Govil, G., Zu-kun, T., Miles, H.T., 1987.

Systematic study of nuclear overhauser effect vis-�a-vis local helical

parameters, sugar puckers, and glycosidic torsions in B DNA:

insensitivity of NOE to local transitions in B DNA oligonucleo-

tides due to internal structural compensations. Biochemistry 26,

1315–1322.

Chastain, P.D., Sinden, R.R., 1998. CTG repeats associated with

human genetic disease are inherently flexible. J. Mol. Biol. 275,

405–411.

Cheung, S., Arndt, K., Lu, P., 1984. Correlation of Lac-operator

imino proton exchange kinetics with its function. Proc. Natl. Acad.

Sci. USA 81, 3665–3669.

Chuprina, V.P., Lipanov, A.A., Fedoroff, O.Y., Kim, S.G., Kintanar,

A., Reid, B.R., 1991. Sequence effects on local DNA topology.

Proc. Natl. Acad. Sci USA 88, 9087–9091.

d’Langlois Estaintot, B., Gallios, B., Brown, T., Hunter, W.N., 1992.

The molecular structure of a 40-epiadriamycin complex with

d(TGATCA) at 1.7�A resolution: comparison with the structure

of 40-epiadriamycin d(TGATCA) and d(CGATCG) complexes.

Nucleic Acids Res. 20, 3561–3567.

Donlan, M.E., Lu, P., 1992. Transcriptional enhancer related DNA

sequences: anomalous 1H NMR NOE crosspeaks. Nucleic Acids

Res. 20, 525–532.

Dornberger, U., Flemming, J., Fritzsche, H., 1998. Structure determi-

nation and analysis of helix parameters in the DNA decamer

d(CATGGCCATG)2 comparison of results from NMR and

crystallography. J. Mol. Biol. 284, 1453–1463.

Feigon, J., 1992. 1H NMR spectroscopy of DNA. Methods Enzymol.

211, 235–253.

Gorin, A.A., Zhurkin, V.B., Olson, W.K., 1995. B-DNA twisting

correlates with base-pair morphology. J. Mol. Biol. 247, 34–48.

Gronenborn, A., Clore, G.M., 1985. Investigation of the solution

structure of short nucleic acid fragments by means of nuclear

overhauser enhancements measurements. Prog. NMR Spectrosc.

17, 1–32.

Grzeskowiak, K., 1996. Sequence dependent structural changes in B-

DNA. Chem. Biol. 3, 785–790.

Hagerman, J.P., 1990. Sequence-directed curvature of DNA. Annu.

Rev. Biochem. 59, 755–781.

Hare, D.H., Wemmer, D.E., Chou, S.H., Drobny, G., Reid, B.R.,

1983. Assignment of non-exchangeable proton resonances of

d(CGCGAATTCGCG)2 using two dimensional methods. J. Mol.

Biol. 171, 319–336.

Helene, C., Lancelot, G., 1982. Interactions between functional groups

in protein–nucleic acid associations. Prog. Biophys. Mol. Biol. 39,

1–68.

Hilbers, C.W., Patel, D.J., 1975. Proton nuclear magnetic resonance

investigations of the nucleation and propagation reactions associ-

ated with the helix-coil transition of d-ApTpGpCpApT in H2O

solution. Biopolymers 14, 2656–2660.

Hosur, R.V., Govil, G., Miles, H.T., 1988. Application of two-

dimensional NMR spectroscopy in the determination of solution

conformation of nucleic acids. Magn. Res. Chem. 26, 927–944.

Hunter, C.A., 1993. Sequence-dependent DNA structure: the role of

base stacking interactions. J. Mol. Biol. 230, 1025–1054.

Kabsch, W., Sander, S., Trifonov, E.N., 1982. The helical twist angle

of B DNA. Nucleic Acids Res. 10, 1097–1104.

Kemler, I., Schreiber, E., Muller, M.M., Matthias, P., Schaffner, W.,

1989. Octamer transcription factors bind to two different sequence

motifs of the immunoglobulin heavy chain promoter. EMBO J. 8,

2001–2008.

Lavery, R., Sklenar, H., 1996. CURVES 5.1. Helical Analysis of

Irregular Nucleic Acids, Laboratory of Theoretical Biology.

CNRS, Paris.

Lavery, R., Sklenar, J., 1989. The definitions of generalized helicoidal

parameters and of axis curvature for irregular nucleic acids. J.

Biomol. Struct. Dyn. 6, 655–667.

Leonard, G.A., Brown, T., Hunter, W.N., 1992. Anthracycline binding

to DNA: high resolution structure of d(TGTACA) complexed with

40-epiadriamycin. Eur. J. Biochem. 204, 69–74.

Leonard, G.A., Hambley, G.A., Mc Aulley Hecht, K., Brown, T.,

Hunter, W.N., 1993. Anthracycline–DNA interactions at unfavor-

able base-pairs triplet binding sites: structures of d(CGGCCG)/

daunomycin and d (TGGCCA)/adriamycin complexes. Acta Cryst.

B D49, 458–467.

Lyubchenko, Y.L., Shlyakhtenko, L., Appella, S.E., Harrington, R.E.,

1993. CA runs increase DNA flexibility in the complex of k cro

protein with the OR3 site. Biochemistry 32, 4121–4127.

Majumdar, A., Hosur, R.V., 1992. Simulation of 2D NMR spectra for

determination of solution conformations of nucleic acids. Prog.

NMR Spectrosc. 24, 109–158.

Maple, J.R., Dinur, U., Hagler, T., 1988. Derivation of force field for

molecular mechanics and dynamics from ab initio energy surfaces.

Proc. Natl. Acad. Sci. USA 85, 5350–5354.

Maple, J.R, Thacher, T.S., Dinur, U., Hagler, A.T., 1990. Biosym

forcefield research results in new techniques for the extraction of

inter and intra molecular forces. Chem. Design Automation News

5, 5–10.

McNamara, P.T., Bolshoy, A., Trifonov, E.N., Harrington, R.E.,

1990. Sequence-dependent kinks induced in curved DNA. J.

Biomol. Struct. Dyn. 8, 529–538.

Mujeeb, A., Kerwin, S.M., Kenyon, G.L., James, T.L., 1993. Solution

structure of conserved DNA sequence from the HIV-I genome:

restrained molecular dynamics simulation with distance and

torsion angle restraints derived from two-dimensional NMR

spectra. Biochemistry 32, 13419–13431.

Neidle, S., Sanderson, M.R., Waring, M.J., 1983. Molecular Aspects

of Anticancer Drug Action. MacMillan, London.

Nibedita, R., Kumar, R.A., Majumdar, A., Hosur, R.V., Govil, G.,

1993. Solution structure of GCCAAT recognition motif by 2D

50 R. Barthwal et al. / Journal of Structural Biology 148 (2004) 34–50

NMR, spectral simulation, molecular modeling, and distance

geometry calculations. Biochemistry 32, 9053–9064.

Nunn, C.M., Meervelt, L.V., Zhang, S., Moore, M.H., Kennard, O.,

1991. DNA–drug interactions. The crystal structures of d(TGT-

ACA) and d(TGATCA) complexed with daunomycin. J. Mol. Biol.

222, 167–177.

Olson, W.K., Lu, X.J., 2001. 3DNA Version 1.4.1, A 3-dimensional

Nucleic Acid Structure Analysis and Rebuilding Software Package.

Department of Chemistry, Rutgers University, NJ.

Patel, D.J., 1977. Mutagen–nucleic acid complexes at the poly-

nucleotide duplex level in solution: intercalation of proflavin into

poly(dA–dT) and the melting transition of the complex. Biopoly-

mers 16, 1635–1656.

Radha, P.K., Madan, A., Nibedita, R., Hosur, R.V., 1995. Solution

structure of the conserved segment of myb cognate DNA sequence

by 2D NMR, spectral simulation, restrained energy minimization,

and distance geometry calculations. Biochemistry 34, 5913–5922.

Ravi Kumar, M., Hosur, R.V., Roy, K.B., Miles, H.T., Govil, G.,

1985. Resonance assignment of the 500MHz proton NMR

spectrum of self complementary dodecanucleotide d-GGATCCG-

GATCC: altered conformation at BamHI cleavage sites. Biochem-

istry 34, 7703–7711.

Sarai, A., Takeda, Y., 1989. Lambda repressor recognizes the

approximately 2-fold symmetric half-operator sequences asymmet-

rically. Proc. Natl. Acad. Sci. USA 86, 6513–6517.

Sarma, M.H., Gupta, G., Sarma, R.H., 1988. Structure of a bent

DNA: two-dimensional NMR studies on d(GAAAATTTTC)2.

Biochemistry 27, 3423–3432.

Schmitz, U., Sethson, I., Egan, W.M., James, T.L., 1992a. Solution

structure of a DNA octamer containing the Pribnow box via

restrained molecular dynamics simulation with distance and

torsion angle constraints derived from two-dimensional nuclear

magnetic resonance spectral fitting. J. Mol. Biol. 227, 510–531.

Schmitz, U., Zon, G., James, T.L., 1992b. Deoxyribose conformation

in [d(GTATATAC)]2: evaluation of sugar pucker by simulation of

double quantum filtered COSY cross peaks. Biochemistry 29,

2357–2366.

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.

Searle, M.S., 1993. NMR studies of drug–DNA interactions. Prog.

NMR Spectrosc. 25, 403–480.

Steitz, T.A., 1990. Structural studies of protein–nucleic acid interac-

tion: the sources of sequence-specific binding. Q. Rev. Biophys. 23,

205–280.

Timsit, T., Vilbois, E., Moras, D., 1991. Nature 354, 167–170.

Tonelli, M., Ragg, E., Bianucci, A.M., Lesiak, K., James, T.L., 1998.

Nuclear magnetic resonance structure of d(GCATATGA-

TAG)�d(CTATCATATGC): a consensus sequence for promoters

recognized by rK RNA polymerase. Biochemistry 37, 11745–

11761.

Trifonov, E.N., Brendal, V., 1986. GNOMIC—A Dictionary of

Genetic Codes. vch Verlagsgesellschaft Weinheim, Germany.

Trifonov, E.N., 1985. Curved DNA. CRC Crit. Rev. Biochem. 19, 89–

106.

Ulyanov, N.B., Goria, A.A., Zhurkin, V.B., Chen, B., Sarma, M.H.,

Sarma, R.H., 1992. Novel solution conformation of DNA observed

in d(GAATTCGAATTC) by two-dimensional NMR spectroscopy.

Biochemistry 31, 3918–3930.

Van de Van, F.J.M., Hilbers, C.W., 1988. Nucleic acids and nuclear

magnetic resonance. Eur. J. Biochem. 178, 1–38.

Van Wijk, J., Huckriede, B.D., Ippel, J.H., Altona, C., 1992.

Spectroscopic methods for analysis of DNA: furanose sugar

conformation in DNA from NMR coupling constants. Methods

Enzymol. 211, 286–307.

Weisz, K., Shafer, R.H., William, E., James, T.L., 1994.

Solution structure of octamer motif in immunoglobin genes via

restained molecular dynamics calculations. Biochemistry 33, 354–

366.

Widmer, H., Wuthrich, K., 1987. Simulated two-dimensional NMR

cross peak fine structures for 1H spin systems in polypeptides and

polydeoxynucleotides. J. Magn. Reson. 74, 316–336.

Wuthrich, K., 1986. NMR of Proteins and Nucleic Acids. Wiley, New

York.

Yanagi, K., Prive, G.G., Dickerson, R.E., 1991. Analysis of local helix

geometry in three B DNA decamers and eight dodecamers. J. Mol.

Biol. 217, 201–214.

Zhurkin, V.B., Ulyanov, N.B., Gorin, A.A., Jernigan, R.L., 1991.

Static and statistical bending of DNA evaluated by Monte Carlo

simulations. Proc. Natl. Acad. Sci. USA 88, 7046–7050.