Content uploaded by Alexandre M J J Bonvin
Author content
All content in this area was uploaded by Alexandre M J J Bonvin
Content may be subject to copyright.
Available via license: CC BY-NC 2.0
Content may be subject to copyright.
Published online 5 May 2009 Nucleic Acids Research, 2009, Vol. 37, Web Server issue W235–W239
doi:10.1093/nar/gkp287
3D-DART: a DNA structure modelling server
Marc van Dijk and Alexandre M. J. J. Bonvin*
Bijvoet Center for Biomolecular Research, Science Faculty, Utrecht University, The Netherlands
Received January 30, 2009; Revised and Accepted April 14, 2009
ABSTRACT
There is a growing interest in structural studies of
DNA by both experimental and computational
approaches. Often, 3D-structural models of DNA
are required, for instance, to serve as templates
for homology modeling, as starting structures
for macro-molecular docking or as scaffold for
NMR structure calculations. The conformational
adaptability of DNA when binding to a protein is
often an important factor and at the same time a
limitation in such studies. As a response to the
demand for 3D-structural models reflecting the
intrinsic plasticity of DNA we present the 3D-DART
server (3DNA-Driven DNA Analysis and Rebuilding
Tool). The server provides an easy interface to
a powerful collection of tools for the generation of
DNA-structural models in custom conformations.
The computational engine beyond the server
makes use of the 3DNA software suite together
with a collection of home-written python scripts.
The server is freely available at http://haddock.
chem.uu.nl/dna without any login requirement.
INTRODUCTION
DNA often changes its conformation as a result of inter-
actions with various ligands; especially binding to proteins
can result in large conformational changes such as helical
kinks (1) or local helical untwisting (2). These play an
important roll in providing complementarity to the pro-
tein binding surface and contributing to the interaction
specificity (3). In order to fully understand the nature of
the conformational changes taking place upon complex
formation, 3D, atomic-resolution structures are required.
Experimental methods such as X-ray crystallography and
nuclear magnetic resonance spectroscopy (NMR) but
also computational approaches such as macro-molecular
docking are important techniques for obtaining such 3D
structures or models.
Most techniques make use of 3D-structural models of
DNA at some point along the structure calculation pipe-
line. NMR for instance can benefit from the regularity in
the structure of double-stranded DNA by using a model
as starting point for structure calculations, thereby com-
pensating for the lack of long range structural informa-
tion. For macro-molecular docking, a starting model
is often required as experimental structures might not
be available. Often, starting from multiple models with
different conformations improves the results. Finally, as
last example, homology-modeling programs require a
template model as starting point for the homology build-
ing process.
The regularity in the structure of double-stranded
DNA makes it especially suitable for modeling. Various
software packages are available that convert a user speci-
fied base-pair sequence into a 3D structure using regular
nucleotide building blocks (4–8). However, most of these
software packages, some of which are available via web-
servers (4,7), are only able to generate models in ideal
canonical conformations (5,7) or in conformations mim-
icking that of a free unbound structure (4).
The structures of double-stranded DNA in complex
with various ligands often show considerable conforma-
tional changes compared to their unbound counterparts
(9–12). This plasticity originates at a ‘local’ level in the
orientation of one base relative to its Watson–Crick part-
ner and of two base pairs relative to one another. These
‘local’ changes accumulate and result in bending and
twisting of the structure at a ‘global’ level. Only a few
existing programs, such as NAMOT (8) and NAB (6),
offer options to introduce custom bends in the generated
DNA conformation and give control over all local param-
eters; they however require some expertise from the user
and are not available as web servers. Here we describe the
3D-DART web server (3DNA-Driven DNA Analysis and
Rebuilding Tool) which we developed to allow for the easy
generation of 3D-structural DNA models with a defined
conformation by providing control over both ‘global’ and
‘local’ conformational features.
The generation of models is accomplished by modifica-
tion of the well-established rotational and translational
parameters that describe the position of one base to its
Watson–Crick counterpart and of two successive base
pairs relative to one another (13). It has been demon-
strated in the past that rebuilding a double-stranded
DNA structure using these parameters results in a near
native structure (5). The only exceptions are local changes
in the sugar and phosphate backbone conformation.
*To whom correspondence may be addressed. Tel: +31 30 2533859; Fax: +31 30 2537623; Email: a.m.j.j.bonvin@uu.nl
ß2009 The Author(s)
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
The 3D-DART server uses the Roll, Tilt and Twist
parameters to introduce bends into the structure. The
other parameters can be used to ‘fine-tune’ the conforma-
tion of the structure. Note that 3D-DART does not pro-
vide custom control of the sugar-phosphate backbone
conformation (in contrast for example to NAMOT). The
3DNA software (5) is used to generate a 3D-structural
model from the modified parameters.
The server accepts a nucleotide sequence, a base pair
(step) parameter file or a DNA-containing PDB coordi-
nate file as input. The server returns 3D-structural models
with the desired conformation as well as a collection of
analysis and intermediate files. Several additional and con-
venient functions are available to control the markup of
the resulting PDB coordinate files, for instance to prepare
them for use in the macro-molecular docking program
HADDOCK (14,15) also developed in our group. For
the same purpose the server can automatically generate
a DNA restraint file (16) as an additional feature. The
server is freely available at http://haddock.chem.uu.nl/
dna without any login requirement.
3D-DART MODELING PROCEDURE
Local bending is often at the origin of double-stranded
DNA distortions when in complex with various proteins
(9–11). This type of bending can be described in terms of
the vector between two successive base pairs (a base pair
step). The length of this vector (Figure 1, thick black
arrows) describes the distance between the two base
pairs in a base-pair step, also known as Rise. It usually
does not vary much. In unbent canonical DNA, these
vectors align with the Z-axis that represents the main heli-
cal path of the structure (Figure 1A). When the DNA is
bent, then the position of the vector relative to the
global reference frame describes the magnitude and orien-
tation of the bend angle. The magnitude of the bend
corresponds to the vector component projected on the
Y–Zplane and its orientation to the component projected
on the Y–Xplane (Figure 1B). The accumulation of
successive vectors then determines the overall bend in
the structure.
Such a bend vector can be decomposed into a Roll, Tilt
and Twist base pair step parameter contributions. The
3D-DART server uses these parameters to introduce
bends in the structure. The underlying algorithm is
based on the transformation of the global bend vector in
Euclidean space into Roll and Tilt values in the local base
pair step reference frame (Figure 1C). This transformation
is accomplished using the following steps:
(i) The definition of bend vectors requires the definition
of an origin in Euclidean space set to an arbitrary
base pair in the sequence (ri) such that the main
helical path of the structure is aligned with the
Z-axis. By default the central base pair is chosen
(Figure 1A, cyan).
(ii) Because the helical DNA structure is twisted the ori-
entation component of the bend vector (O
i
,08!
3608) at a given base-pair step ineeds to be
corrected for the local Twist value. In Figure 1A
Figure 1. One block per base-pair Calladin–Drew plot of DNA illus-
trating the relation between the local Twist (V), Roll (r) and Tilt (t)
values and the global bend angle for a given base-pair step. Vector
projections are normalized for illustrative purposes. (A) The correction
of the orientation component of the bend vector for the local Twist
value at base pair iand i+ 1 (blue circle parts). The red arrow indi-
cates the value of the orientation component (O
i
) before Twist correc-
tion and the blue arrow (corO
i
aligned with Y-axis) after correction. (B)
A bend in the structure as a result of a different bend angle vector
(thick black arrows) between every successive base-pair step. The blue
arrow illustrates the orientation component of the vector (Y–Xplane)
and the red arrow the magnitude (Y–Zplane). (C) provides a detailed
view of the local base-pair step reference frame between base iand
i+ 1. The global bend vector (thick black arrow) is decomposed into
a Tilt (red arrow, Y0–Z0plane) and Roll (blue arrow, X0–Z0plane)
contribution.
W236 Nucleic Acids Research, 2009, Vol. 37, Web Server issue
the red arrow illustrates the direction of the orien-
tation component before Twist (V) correction
and the blue one after correction. In Equation (1)
the orientation component O
i
is corrected for the
Twist value at base pair step iby subtracting
the accumulated Twist from ri to i(Figure 1A,
blue circle partitions). The accumulated Twist
value is positive above and negative below the refer-
ence base pair.
corOi¼OiþX
i
ri¼1
i1
(iii) The bend vector with the corrected orientation
component (corO
i
) at base pair step iis the result
of the Roll (r) and Tilt (t) in the local base pair step
reference frame of i(Figure 1C). The base pair tilt
caused by Roll (X0–Z0plane) is orthogonal to the
base pair tilt originating from the Tilt parameter
(Y0–Z0plane). Together they can span 3608.
Equation (2) defines the fractional contributions of
the Roll (frr
i
) and Tilt (frt
i
) to the vectors
orientation:
fri¼cosðcorOi=180Þ
fri¼sinðcorOi=180Þ2
The in Figure 1C blue arrow represents the tilt frac-
tion and the red arrow the roll fraction.
(iv) Finally the actual value of the Roll and Tilt param-
eters at base pair iare calculated (Equation 3)
to reflect both the magnitude and orientation com-
ponents of the vector representing global bend
angle A
i
.
¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A2
ijfrij
q
Si
¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A2
ijfrij
q
Si
3
where:
Si¼1iffrior fri>0;Si¼1iffrior fri<0;
Si¼0iffriand fri¼0
Using this algorithm the magnitude and orientation of a
bend angle can be calculated for every individual base
pair step in the sequence of a double-stranded DNA struc-
ture. The algorithm is based on strict geometrical
principles and the resulting structures are not energy mini-
mized. Therefore it is important to emphasize that the
conformation does not need to reflect an energetically
favorable state. Figure 2 illustrates the procedure with a
few examples. Introducing a 38bend with the same orien-
tation in a 20 base pair structure create a smooth bend of
608(Figure 2A). Restricting a 308bend to only the central
3 base pair steps in the same sequence creates a kink
(Figure 2B). Using different angle values for every base
pair step in the sequence gives precise control over the
DNA conformation (Figure 2C).
3D-DART WEB SERVER
Input
The generation of custom DNA 3D-structural models
is based on the manipulation of the 6 bp and 6-bp step
parameters describing the conformation of the structure.
The first step in using the server is the definition of a
source for the base-pair (step) parameter template file
that serves as a starting point for the modeling process.
For this three different options are supported:
The first option consists of inputting a nucleotide
sequence defined as the bases belonging to the 50–30
template strand. The server then uses the ‘fiber’
module of 3DNA (5) to generate a canonical A- or
B-DNA structure with the defined sequence and the
‘find_pair’ and ‘analyze’ module of the same software
to generate the base pair (step) parameter file.
The second input option consists of uploading an user-
defined PDB coordinate file. The DNA in this file is
analyzed and a base pair (step) parameter file repre-
senting its conformation is generated. This allows
for the introduction of custom changes in an already
existing structure.
Finally, as third input option, the server also accepts a
predefined base pair (step) parameter file in 3DNA
format.
Parameters
The base pair (step) parameter file that results from the
input data is subsequently used to start the modeling
phase. At this stage, the user can introduce bends into
the structure. There are two modes in which this can be
accomplished referred to as ‘Global’ and ‘Local’:
In the ‘Global’ mode the defined bend angle is evenly
distributed over all base pair steps in the user-defined
zone (Figure 2A and B). The ‘Global’ modeling mode
accepts ranges of parameters so that multiple models
within a given bend angle and/or bend angle orienta-
tion range can be generated. Six models from 108to
408with steps of 58for example.
In the ‘Local’ mode the bend angle and its orientation
in Euclidean space can be defined uniquely for every
base pair step in the user-defined zone (Figure 2C).
Next to the introduction of bends the user can define
custom values for the various base pair and base pair
step parameters. These values are subsequently used for
every base pair or base pair step in the sequence. If a bend
is introduced then of course the Roll and Tilt values will
be substituted by the ones needed to introduce the bend. If
location-specific values for the parameters need to be
introduced it is advised to start from a base pair step
parameter file containing these values or from a PDB
coordinate file reflecting these values.
The combination of the possible input sources together
with the precise modeling options makes the 3D-DART
server very versatile, allowing the generation of both
ideal DNA models as well as fully customized ones by
Nucleic Acids Research, 2009, Vol. 37, Web Server issue W237
introducing local conformational changes in already
existing structures.
Output
The collection of base pair (step) parameter files resulting
from the modeling phase are converted into a 3D structure
in PDB format by the ‘rebuild’ module of 3DNA. The 3D
structure is built based on the nucleotide coordinates for
the common bases as determined by Arnott et al. (17). The
generated 3D-structural models are returned in a zipped
archive containing in addition a collection of analysis
and other useful intermediate files. These include the pro-
vided input files, the base pair (step) parameter files from
which the models were generated, various bend and
3DNA analysis files and the 3D-DART log file.
In addition, the server provides a few convenience
functions to further customize the output. These include
options to change the PDB markup such as changing
chain ID and renumbering residues. The server also
offers the option to output the structures in a format
and notation consistent with the macro-molecular
docking program HADDOCK (14,15). In that case, an
additional restraint file is generated to maintain the
DNA conformation during the flexible refinement stage
of the docking. These functionalities are actually used
by the HADDOCK web server (http://haddock.chem.
uu.nl/haddock) to automatically process DNA/RNA
input structures.
ACKNOWLEDGEMENTS
We would like to thank the group of Janusz M. Bujnicki
(International Institute of Molecular and Cell Biology,
Warsaw, Poland) for providing helpful feedback while
testing the server.
FUNDING
European Community (FP6 STREP project ‘Extend
NMR’, contract no. LSHG-CT-2005-018988, FP6 I3
project ‘EU-NMR’, contract no. RII3-026145 and FP7
I3 project ‘eNMR’, contract no. 213010-e-NMR); VICI
grant from the Netherlands Organization for Scientific
Research (NWO) (to A.M.J.J.B.) (grant no. 700.96.442).
Funding for open access charge: VICI grant from the
Netherlands Organization for Scientific Research (NWO)
(to A.M.J.J.B.) (grant no. 700.96.442).
Conflict of interest statement. None declared.
REFERENCES
1. Flick,K.E., Jurica,M.S., Monnat,R.J. Jr and Stoddard,B.L. (1998)
DNA binding and cleavage by the nuclear intron-encoded homing
endonuclease I-PpoI. Nature,394, 96–101.
2. Patikoglou,G. and Burley,S.K. (1997) Eukaryotic transcription
factor-DNA complexes. Annu. Rev. Biophys. Biomol. Struct.,26,
289–325.
3. Allemann,R.K. and Egli,M. (1997) DNA recognition and bending.
Chem. Biol.,4, 643–650.
Figure 2. Two block per base Calladine–Drew plots of a 20-nt B-DNA structure. The black dotted line defines the main helical path. Two examples
of bending using the ‘Global’ mode of the server are shown: (A) A DNA conformation with a smooth bend of 608distributed evenly over the
entire structure and (B) a conformation with a 308kink in which the bend was restricted between base pairs 10 and 12. (C) A DNA conformation
with custom angles values for every base pair step generated using the ‘Local’ mode of the server. The asterisk indicates the reference base pair
used in the algorithm to generate the bend. The pictures were created using 3DNA (5) and PyMol (DeLano Scientific LLC, Palo Alto, CA, USA).
W238 Nucleic Acids Research, 2009, Vol. 37, Web Server issue
4. Dixit,S.B. and Beveridge,D.L. (2006) Structural bioinformatics of
DNA: a web-based tool for the analysis of molecular dynamics
results and structure prediction. Bioinformatics,22, 1007–1009.
5. Lu,X.J. and Olson,W.K. (2003) 3DNA: a software package for the
analysis, rebuilding and visualization of three-dimensional nucleic
acid structures. Nucleic Acids Res.,31, 5108–5121.
6. Macke,T. and Case,D.A. (1998) In Leontes,N.B. and
SantaLucia,J.J. (eds), Molecular Modeling of Nucleic Acids.,
American Chemcial Society, Washington DC, pp. 379–393.
7. Munteanu,M.G., Vlahovicek,K., Parthasaraty,S., Simon,I. and
Pongor,S. (1998) Rod models of DNA: sequence-dependent aniso-
tropic elastic modelling of local bending phenomena. Trends
Biochem. Sci.,23, 341–346.
8. Tung,C.S. and Carter,E.S. II (1994) Nucleic acid modeling tool
(NAMOT): an interactive graphic tool for modeling nucleic acid
structures. Comput. Appl. Biosci.,10, 427–433.
9. Dickerson,R.E. (1998) DNA bending: the prevalence of kinkiness
and the virtues of normality. Nucleic Acids Res.,26, 1906–1926.
10. Dixit,S.B. and Beveridge,D.L. (2005) Axis curvature and ligand
induced bending in the CAP-DNA oligomers. Biophys. J.,88,
L04–L06.
11. El Hassan,M.A. and Calladine,C.R. (1998) Two distinct modes of
protein-induced bending in DNA. J. Mol. Biol.,282, 331–343.
12. Jones,S., Heyningen,P., Berman,H. and Thornton,J.M. (1999)
Protein-DNA interactions: a structural analysis. J. Mol. Biol.,287,
877–896.
13. Dickerson,R.E. (1989) Definitions and nomenclature of nucleic acid
structure parameters. J. Biomol. Struct. Dyn.,6, 627–634.
14. Dominguez,C., Boelens,R. and Bonvin,A.M. (2003) HADDOCK: a
protein-protein docking approach based on biochemical or
biophysical information. J. Am. Chem. Soc.,125, 1731–1737.
15. de Vries,S.J., van Dijk,A.D., Krzeminski,M., van Dijk,M.,
Thureau,A., Hsu,V., Wassenaar,T. and Bonvin,A.M. (2007)
HADDOCK versus HADDOCK: new features and performance
of HADDOCK2.0 on the CAPRI targets. Proteins,69, 726–733.
16. van Dijk,M., van Dijk,A.D., Hsu,V., Boelens,R. and Bonvin,A.M.
(2006) Information-driven protein-DNA docking using
HADDOCK: it is a matter of flexibility. Nucleic Acids Res.,34,
3317–3325.
17. Chandrasekaran,R.A. and Arnott,S. (1989) In Saenger,W. (ed.),
Landolt-Bo
¨rnstein Numerical Data and functional Relationships in
Science and Technology. Vol. VII/1b, Springer-Verlag, pp. 31–170.
Nucleic Acids Research, 2009, Vol. 37, Web Server issue W239