J. Chem. Inf. Model.

2009, 49, 1925–1935 1925

Docking Studies on DNA-Ligand Interactions: Building and Application of a Protocol

To Identify the Binding Mode

Clarisse G. Ricci and Paulo A. Netz*

Instituto de Quı́mica, Universidade Federal do Rio Grande do Sul, av. Bento Gonçalves 9500,
91501-970, Porto Alegre, Brazil

Received April 30, 2009

Despite DNA being an important target for several drugs, most of the docking programs are validated only
for proteins and their ligands. In this paper, we used AutoDock 4.0 to perform self-dockings and cross
dockings between two DNA ligands (a minor groove binder and an intercalator) and four distinct receptors:
1) crystallographic DNA without intercalation gap; 2) crystallographic DNA with intercalation gap; 3)
canonical B-DNA; and 4) modified B-DNA with intercalation gap. Besides being efficient in self-dockings,
AutoDock is capable of correctly identifying two of the main DNA binding modes with the condition that
the target possesses an artificial intercalation gap. Therefore, we suggest a default protocol to identify DNA
binding modes which uses a modified canonical DNA (with gap) as receptor. This protocol was applied to
dock two different Tröger bases to DNA and the predicted binding modes agree with those suggested, yet
not established, by experimental data. We also applied the protocol to dock aflatoxin B1 exo-8,9-epoxide,
and the results are in complete agreement with experimental data from the literature. We propose that this
approach can be used to investigate other ligands whose binding mode to DNA remains unknown, yielding
a suitable starting point for further theoretical studies such as molecular dynamics simulations.

INTRODUCTION Certainly, this had shed new light on the potential of

automated docking programs for virtual screening of DNA
As the number of biological structures in data banks
binding agents. However, although self-dockings (i.e., using
rapidly increases, molecular docking is becoming an impor-
the original crystallographic target) are considered useful as
tant approach to evaluate or even to elucidate the interaction
a first indication of docking accuracy, they have proved to
between potential ligands and their macromolecular targets.1,2
provide little information about accuracy in real drug
It has been shown that several docking methods described
discovery.2 Indeed, the employment of docking techniques
so far can correctly reproduce the binding modes of cocrys-
to elucidate unknown DNA binding mechanisms - without
tallized ligands to their protein targets (self-dockings), but
any conclusive previous experimental data - remains a
none of them can be considered a universally applicable
challenge.
method.2
The known fact that DNA is not rigid but rather a very
While there is a large number of studies reporting
flexible molecule that can assume several structurally distinct
protein-ligand docking, much less research has been re-
isoforms16-18 combined with the fact that most of current
ported on docking of ligands to nucleic acids3-13 despite
automated docking programs do not take into account target
DNA being an important molecular target for a wide number
flexibility leads to a so far insufficiently explored issue: when
of antibiotics and antitumor drugs.14 Disappointingly, most
it comes to the docking of ligands whose binding mode to
of the scoring functions have been parametrized exclusively
DNA remains unknown, which oligomer conformation
with protein-ligands sets, and the programs have been
should be used as target? It is likely that several structural
validated only for proteins and their ligands.2,15 As well-
features of the chosen oligomer conformation will strongly
known, nucleic acids differ from proteins due to unique
impact docking performance, especially when ligand and
structural features such as high density charge and helix chiral
target interact through an induced-fit mechanism.
geometry. Also, nucleic acids do not present a single and
well-defined binding site (as occur with most of the proteins) In general, interaction of small molecules with DNA
and offer more solvent exposed binding pockets.2 As a occurs in two distinct ways: intercalation between base pairs
consequence, this leads to the question of whether docking or groove recognition.19 Although the major groove offers
programs validated for proteins can also produce reasonable more H-bonding donor and acceptor sites, it has been
results in ligand-DNA docking. This issue has been recently suggested that it provides a much larger and shallow binding
approached by Holt et al.,13 who has shown that AutoDock pocket for small molecule binding than the minor groove.
and Surflex can accurately reproduce the crystal structure Indeed, most of the small ligands bind to the minor groove,
of several ligands (minor groove binders and intercalators) while the major groove is the preferential binding site for
bound to DNA, within a resolution of approximately 2 Å. proteins and peptides.19,20
Among the main DNA binding modes, intercalation is the
* Corrresponding author fax: (++55)(51)33087304; e-mail: netz@ most common way through which small and rigid aromatic
iq.ufrgs.br. molecules recognize DNA. Most of the classical and simple
10.1021/ci9001537 CCC: $40.75  2009 American Chemical Society
Published on Web 08/06/2009
1926 J. Chem. Inf. Model., Vol. 49, No. 8, 2009 RICCI AND NETZ

intercalators such as ethidium bromide and proflavine do not

possess sequence selectivity since the binding of these ligands
to DNA depends basically on π-stacking and stabilizing
electrostatic interactions.20,21 It is also well-known that
intercalation imposes structural alterations to DNA in order
to open an intercalation gap between two consecutive base
pairs.19 Therefore, intercalative binding to DNA can be
considered as an induced-fit mechanism.
On the other hand, minor groove recognition by shape-
selective agents like distamycin and netropsin is more similar
to a lock-and-key mechanism since little or no apparent DNA
distortion is observed after binding.20 Due to their flexibility,
these naturally occurring polyamides assume a curved shape
that matches DNA topology20,22 and allows a snug fit of the
ligand in the minor groove. It is assumed that this interaction
depends on a combination of three main factors, in which
van der Waals contacts play a pivotal role assisted by Figure 1. Examples of DNA binding agents.
hydrogen bonding and electrostatic interactions.23,24 Also,
netropsin and distamycin have attracted considerable atten- to intercalate to DNA forming a subsequent covalent bond
tion due to their selective binding to AT-rich sequences in with a guanine in the major groove.32,33
minor groove, showing that it is possible to achieve If consistent with experimental data, we suggest that the
sequence-selective binding agents.22,23,25 Netropsin, for tested docking protocol can be used to investigate other
instance, is known to interact strongly with AT-rich se- ligands whose binding mode to DNA remains unknown,
quences, displaying large residence times.26 yielding a suitable starting point for further theoretical studies
Considering the preference that sequence-specific ligands as, for instance, molecular dynamics simulations in conjunc-
show for some sequences, it is important to observe that tion with free energy calculations.34
sequence selectivity can arise from i) direct readout of
H-bonding pattern, ii) indirect readout of sequence flexibility METHODOLOGY
(as occur with TATA-box Binding Protein27), or iii) a
combination of both (as occur with polyamides20). Therefore, Molecular docking experiments were performed with
docking methods that do not take into account DNA AutoDock 4.0, a software that uses an empirical scoring
flexibility are sensitive only to the first type of sequence function based on the free energy of binding.28,35 Among
selectivity, not to the second one. This is the main reason the stochastic search algorithms offered by the AutoDock
why we consider that, as a primary approach to docking suite, we chose the Lamarckian Genetic Algorithm (LGA)
studies of DNA-interacting agents, elucidating the binding which combines global search (Genetic Algorithm alone) to
mode may be more promising than investigating the existence local search (Solis and Wets algorithm36).
of preferential binding sites. Genetic algorithms are based on the evolutionary concept
in which the solution to an adaptative problem is spread
In this context, we decided to use the software AutoDock among a genetic pool. In molecular docking, the ‘solution’
4.028 to perform not only self-dockings but also cross corresponds to the best binding position for the ligand, and
dockings between two ligands (ellipticine and netropsin) and it is represented by a ‘chromosome’ file containing transla-
four structurally distinct oligomers (two canonical and two tion, orientation, and torsion ‘genes’. Basically, a genetic
crystallographic) in order to establish a docking protocol able algorithm creates a randomly placed population of individuals
to identify two of the main DNA binding modes: intercalation (ligands) and then applies cycles of genetic operators
and minor groove recognition. (mutation and crossover) giving rise to new generations until
We further applied this docking protocol to predict the a suitable solution is achieved.
binding modes of two Tröger bases: a symmetric base The ‘solutions’ are evaluated through their free energy of
derived from proflavine and an asymmetric base derived from binding. To achieve faster energy evaluation, AutoDock
proflavine and phenanthroline, each with two optical isomers. represents the macromolecule as a tridimensional grid, in
The data concerning these molecules in literature remain which each point stores precalculated affinity potentials for
inconclusive. For both molecules, the (-)-isomer has proved all atom types of the ligand. In this way, AutoDock allows
to preferentially bind B-DNA.29-31 However, although the flexibility to the ligand, whereas the macromolecule is kept
binding mode is not yet elucidated, spectroscopic and rigid and fixed during docking.
biochemical experiments suggest two different binding The Protein Data Bank37 was searched for ligand-DNA
modes for these Tröger bases: minor groove binding for the complexes, and two structures were selected: 1Z3F (a
symmetric base30 and a bimodal binding mode for the hexamer d(CGATCG)2 complexed with ellipticine) and
asymmetric Tröger base, with proflavine intercalated and 1DNE (a dodecamer d(CGCGATATCGCG)2 complexed
phenanthroline residing in minor groove.31 with netropsin). Ellipticine is a typical intercalating agent
Finally, we decided to test the docking protocol with a with antitumor activity,38 whereas netropsin is a naturally
molecule that binds covalently to DNA and compare the occurring antibiotic that interacts with DNA through minor
result with that from self-docking. In this way, we chose groove recognition.39,40 Ellipticine and netropsin structures
the aflatoxin B1 exo-8,9-epoxide, a potent carcinogen known are shown in Figure 1.
DOCKING STUDIES ON DNA-LIGAND INTERACTIONS J. Chem. Inf. Model., Vol. 49, No. 8, 2009 1927

Table 1. Summary of Tested Dockings

number of maximum of energy maximum of
docking runs evaluations generations
0 50 10,000,000 270,000
1 50 2,500,000 270,000
2 50 50,000,000 270,000
3 25 10,000,000 270,000
4 100 10,000,000 270,000
5 25 50,000,000 270,000
6 50 10,000,000 27,000
7 50 10,000,000 2,700
8 50 10,000,000 270
9 25 50,000,000 27,000

In a first stage, dockings with ellipticine and netropsin were

performed with the oligomers from crystallographic com- Figure 2. Results from dockings 1-9 (applied to netropsin). Above:
bars indicate the percentage of runs resulting in intercalation or
plexes 1Z3F and 1DNE, respectively (self-dockings). After other binding modes. Below: CPU time associated with each
the separation of the coordinates of ligands and DNA, polar protocol.
and aromatic hydrogens were added with the GROMACS
package41,42 using the GROMOS 53A6 force field,43 and
Gasteiger-Marsili charges44 were calculated with AutoDock
Tools (ADT).45
A grid box was created with 96 × 96 × 96 points and a
resolution of 0.375 Å, in order to include the entire DNA
fragment. After the grid box was centered in the macromol-
ecule, grid potential maps were calculated using module
AutoGrid 4.0.
Initially, each docking consisted of 50 independent runs,
with an initial population of 50 individuals, a maximum
number of 107 energy evaluations, and a maximum number
of 270,000 generations. This was considered docking zero.
Mutation and crossover were applied to the population at
rates of 0.02 and 0.80, respectively. For a local search, it
was set to the pseudo Solis and Wets algorithm, with a Figure 3. Results from dockings 1-9 (applied to the acridine
translational step size of 0.2 Å, an orientational step size of derivative). Above: bars indicate the percentage of runs resulting
5.0 degrees, and a torsional step size of 5.0 degrees. To the in minor groove or other binding modes. Below: CPU time
remaining parameters, default values were applied. Results associated with each protocol.
differing by less than 2.0 Å in root-mean-square deviation acridine derivative (PDB: 2GB948) whose structure is also
(rmsd) were grouped in the same cluster, which is represented shown in Figure 1. We chose this acridine derivative since
by the energetically most favorable conformation belonging it has more degrees of freedom than ellipticine and thus
to the cluster. Since the grid box enclosed not only the enhances the challenge of docking.
binding site but also the entire DNA fragment, we could not Basically, it was observed that the number of runs has no
use rmsd as an accuracy criterion. Instead, we opted for a clear effect upon the predictive skill of docking, although it
more subjective yet more representative criterion, which was increases the computational cost in a linear way (see dockings
to classify the resulting binding mode by visual inspection 3, 0, and 4 in Figures 2 and 3). In contrast, docking results
as intercalation, minor groove binding, or others (major are improved as the maximum number of energy evaluation
groove binding, interaction with phosphate groups, etc.). We increases. However, computational time increases strongly
also evaluated the pattern of hydrogen bonding for the best in this case (see dockings 1, 0, and 2 in Figures 2 and 3).
ranked netropsin complexes using the program VMD,46 with The maximum number of generations correlates weakly with
a donor-acceptor distance cutoff of 3.5 Å and an angle computational cost, but no prediction improvement was
cutoff of 40°. However, as hydrogen bonding also depends observed when changing from 27,000 to 270,000 generations
strongly on the flexibility of the target, we could not see the (compare dockings 8, 7, 6, and 0 in Figures 2 and 3).
characteristic bifurcated hydrogen bonds formed between Therefore, a set of standard docking parameters with the best
netropsin and two consecutive DNA bases.47 overall performance was established, consisting of 25 runs,
Next, to improve docking performance while reducing 50 × 106 energy evaluations, and 27,000 generations
computational cost, dockings were carried out with variations (docking 9). These parameters were kept fixed in further
in the number of runs, in the maximum number of energy dockings with netropsin and ellipticine.
evaluations, or in the maximum number of generations. The While most of the docking protocols shown in Figure 2
remaining parameters were kept as in docking zero. The total indicate that minor groove interaction is the most accessed
of tested dockings is summarized in Table 1, and the results binding mode for netropsin, it can also be observed that most
are shown in Figures 2 and 3. These dockings were carried of the docking results shown in Figure 3 point to an incorrect
out with netropsin and, instead of ellipticine, an intercalating binding mode for the acridine derivative. However, it is
1928 J. Chem. Inf. Model., Vol. 49, No. 8, 2009 RICCI AND NETZ

The covalent adduct was obtained from the protein data bank
(PDB: 1MKL33), and the coordinates of the ligand were
separated from the coordinates of DNA of sequence
d(ACATCGATCT) · (AGATCGATGT). In order to perform
the docking using our generic protocol, a canonical oligomer
was generated using X3DNA, with a sequence identical to
the crystallographic oligomer and then modified to contain
an intercalation gap as previously described. The parameters
used in dockings of aflatoxin B1 exo-8,9-epoxide to DNA
were the same as used in the dockings of the Tröger bases.
Figure 4. Structure of the Tröger bases.

worthwhile to stress that qualitatiVely the cluster profiles RESULTS AND DISCUSSION
point to the correct binding mode, for in all docking protocols Among the 25 runs performed in each docking with
with the acridine derivative the best ranked conformations ellipticine and netropsin, the ten most favorable were
in terms of binding free energy were those resulting in analyzed (those presenting the most negative binding free
intercalative binding (not shown). energies). In Table 2 it is reported a summary of the results
In a second stage, the crystallographic receptors were for ellipticine or netropsin dockings with the four different
replaced by canonical B-DNA with similar sequences, DNA targets. A/E are dockings with crystallographic DNA
generated with X3DNA.49 Since structural changes in the original from the ellipticine-DNA complex (with intercalation
macromolecule are not allowed during docking, one canoni- gap); B/F are dockings with modified canonical B-DNA
cal oligomer was previously modified to include an “inter- containing an intercalation gap; C/G are dockings with
calation gap”. In this way, Swiss PDB Viewer50 was used crystallographic DNA from netropsin-DNA complex (with-
to pose an ellipticine molecule between two base pairs of out intercalation gap); and D/H are dockings with canonical
canonical DNA, parallel to the base rings. After that, the B-DNA (without intercalation gap). Thus, A, B, G, and H
complex was minimized by the steepest descent method, are direct dockings, while C, D, E, and F are cross-dockings.
using the GROMACS package with the GROMOS 53A6 Among direct dockings, only dockings A and G are self-
force field. The result was a modified B-DNA in which the dockings. Cluster profiles are shown in Figure 5, and the
base pairs flanking ellipticine were separated by 6.50 Å. best ranked conformations for each docking set are illustrated
Ellipticine was removed from the complex, and the modified in Figure 6.
B-DNA (see Figure 6B/F) was used as a target for docking Direct Dockings. As expected, the applied docking
with ellipticine. The sequences of the X3DNA generated protocol proved to be very efficient to predict binding modes
oligomers are d(CGCAATTGCG)2 (without gap) and d(CG- for ellipticine or netropsin in self-dockings (Table 2, dockings
GCATGCCG)2 (with gap, indicated in bold). A and G). All runs with ellipticine resulted in intercalation
We also decided to perform cross dockings: ellipticine was mode, with an average binding free energy of -8.71 kcal/
docked with crystallographic and canonical DNA fragments mol (Figure 5A), and all runs with netropsin resulted in minor
previously used as receptors for netropsin (without intercala- groove recognition, with an average binding free energy of
tion gap), whereas netropsin was docked with crystal- -9.13 kcal/mol and a very favorable best docked conforma-
lographic and modified canonical DNA fragments previously tion (-9.97 kcal/mol) (Figure 5G).
used as receptors for ellipticine (containing intercalation gap). However, when the original crystallographic targets are
In this way, we tested four different docking protocols, replaced by canonical DNA of similar sequence, distinct
each one consisting of a docking pair, i.e., the same target tendencies are observed for each kind of ligand (Table 2,
oligomer docked with ellipticine and netropsin. dockings B and H). Ellipticine docking with modified
We also applied one of these protocols to dock two Tröger B-DNA (containing an intercalation gap) shows that inter-
bases to DNA: a symmetric Tröger base derived from calation represents only 50% of the runs (Figure 5B),
proflavine and an asymmetric Tröger base derived from although it still corresponds to the preferential binding mode
proflavine and phenanthroline. Both structures are shown in in terms of binding free energy (-8.10 kcal/mol). In contrast,
Figure 4. Since these are chiral compounds, we constructed there is no change in the percentage of minor groove
both isomers (-)-(R,R) and (+)-(S,S) using GaussView.51 recognition when netropsin is docked to canonical B-DNA
The geometries were optimized with RHF/6-31G* using (Figure 5H) instead of crystallographic DNA (Figure 5G).
Gaussian.52 After that, polar and aromatic hydrogens were With respect to binding free energies, the cluster profile from
added with GROMACS using the GROMOS 53A6 force docking H proves to be quite similar to that of docking G,
field, Gasteiger-Marsili charges were calculated using ADT, with a very favorable average binding energy of -8.70 kcal/
and the grid was created as described above, including the mol.
entire DNA receptor. The same default docking parameters These results are in agreement with the characteristic
were used for the Tröger bases, except that we opted for mechanisms of each binding mode. Since ellipticine interacts
100 runs instead of 25. with DNA through an induced-fit mechanism that is not
Finally, in order to test if this docking protocol can also allowed to occur during docking, it is likely that ellipticine
be applicable to a molecule that binds covalently to DNA, docking to DNA will depend strongly on the target selected
we chose the carcinogen aflatoxin B1 exo-8,9-epoxide, an conformation. On the other hand, since netropsin is a very
intercalator which also forms a covalent bond with the N7 flexible ligand and its flexibility is taken into account during
position of guanine,33 probably through a SN2 mechanism.32,53 docking, netropsin can endure structural adaptations in order
DOCKING STUDIES ON DNA-LIGAND INTERACTIONS J. Chem. Inf. Model., Vol. 49, No. 8, 2009 1929

Figure 5. Cluster profiles from dockings A-H. Each cluster is represented by a bar in which the height corresponds to the number of
conformations in the cluster, and the color indicates the binding mode: minor groove in black, intercalation in gray, and other in hatched
pattern. The dashed line indicates the average binding free energy.

Figure 6. Best docked conformations for dockings A-H. Ellipticine is shown in pink and netropsin in orange. In DNA, CG-rich regions are
shown in gray, and AT-rich regions are shown in magenta.

to successfully fit a canonical minor groove, as proved by Cross Dockings. In cross dockings, the mechanistic
results from docking H. features observed for ellipticine and netropsin binding mode
1930 J. Chem. Inf. Model., Vol. 49, No. 8, 2009 RICCI AND NETZ

Table 2. Docking Results for Ellipticine (EL) or Netropsin (NE)c

docking ligand DNA Nminor (%) Nint (%) ∆Gbest ∆Gaverage
A EL a
crystallographic (with gap) 0 100 -8.710 (int) -8.71
B EL canonical (with gap) 50 50 -8.100 (int) -7.25
C EL crystallographic (without gap)b 100 0 -7.98 (minor) -7.98
D EL canonical (without gap) 100 0 -6.51 (minor) -6.51
E NE crystallographic (with gap)a 20 70 -7.62 (int) -6.74
F NE canonical (with gap) 90 0 -9.85 (minor) -7.17
G NE crystallographic (without gap)b 100 0 -9.97 (minor) -9.13
H NE canonical (without gap) 100 0 -9.47 (minor) -8.70
a
Original from the ellipticine-DNA complex. b Original from the netropsin-DNA complex. c Nminor, percentage of runs in minor groove
mode; Nint, percentage of runs in intercalating mode; ∆Gbest, most favorable binding free energy (kcal/mol); ∆Gaverage, average binding free
energy (kcal/mol).

Table 3. Average Values of Structural Double Helix Parametersd

become more evident. As expected, results from dockings
C and D (Table 2) clearly show that the applied docking DNA receptor twist (°) P-P (Å) rise (Å)
method cannot recognize intercalation binding mode when crystallographic (with gap)b 29.43 16.4 6.88a
DNA does not possess a proper intercalating gap. This canonical (with gap) 34.38 12.4 6.50a
limitation of all contemporary docking methods that do not crystallographic (without gap)c 36.22 11.1 3.42
canonical (without gap) 35.73 11.7 3.35
take into account receptor flexibility was already noted in
the literature.2,6 Indeed, when crystallographic (Figure 5C) a
Value for the base-pair step that contains the intercalation gap.
b
or canonical DNA (Figure 5D) without intercalation gap is Original from ellipticine-DNA complex. c original from netropsin-
DNA complex. d Twist, twist between two consecutive base pairs;
used as receptor, all ellipticine docking runs result in minor P-P, minor groove distance; rise, distance between two consecutive
groove recognition, although with binding free energies not base pairs. The analysis of these parameters was performed with
as favorable as the intercalating binding free energies from X3DNA.
docking A.
In contrast, netropsin cross dockings E and F show that
the minor groove can still be recognized as a possible binding Indeed, it is well-known that intercalation not only requires
mode for netropsin even when the DNA target presents an the opening of an intercalation gap but also partially unwinds
intercalation gap (Table 2, Figure 5E,F). However, docking the double helix.19,38 This occurs because base pairs are better
E points to an intercalative binding as the preferential binding overlapped in unwound DNA, resulting in increased π-stack-
mode for netropsin both quantitatively (70% of the runs) and ing interactions between base rings and aromatic rings of
qualitatively (best binding free energy). Even so, it is the intercalator.20 Consequently, reduced twist is probably
noteworthy that the binding free energies are far less one of the important features that favor intercalation in
favorable than those from netropsin dockings G and H and dockings A and E. (It is important to note that, although the
also that the free binding energy of the best docked classical score function of AutoDock does not explicitly take
conformation (-7.62 kcal/mol) is not as negative as those into account molecular polarizability, dockings A and B
of ellipticine intercalation in dockings A (-8.71 kcal/mol) suggest that π-stacking can be sufficiently well mimetized
and B (-8.10 kcal/mol). through van der Walls interactions).
In comparison to docking E, docking F presents a very On the other hand, considering that shape selective binders
different profile (Figure 5F) for it recognizes minor groove possess a natural curvature compatible to B-DNA, it is likely
recognition as the preferential and most accessed binding that netropsin affinity for DNA will be strongly affected by
mode for netropsin even in the presence of an intercalation a decrease in double helix twist. Moreover, unwinding of
gap. Of ten runs, only one did not result in minor groove DNA also affects minor groove width, as denoted by P-P
recognition, and none of the runs resulted in intercalation as distances in Table 3. For shape-selective polyamines, minor
a possible binding mode. The average binding free energy groove width is an important structural feature: a narrow
(-7.17 kcal/mol), although not comparable to those of minor groove allows snug fit of the ligand, consequently
docking G and H, was still more favorable than the average enhancing favorable van der Waals interactions as well as
free binding energy from docking E (-6.74 kcal/mol). Also, electrostatic attraction between the cationic groups of the
the minor groove binding from the best ranked run in docking ligand and the phosphate groups of DNA.20 Conversely, the
F presented a very favorable binding free energy of -9.85 very large minor groove of crystallographic DNA from
kcal/mol, comparable to the binding free energies from best ellipticine complex (16.4 Å) diminishes the contact surface
docked conformations in dockings G and H and also far more between netropsin and minor groove walls, decreasing the
negative than the binding free energy from the best inter- netropsin-DNA interactions. In other words, these structural
calated conformation from docking E (-7.62 kcal/mol). features - decreased twist associated with a very large minor
Conformational Analysis of Target Oligomers. In order groove - are probably the main reason behind the incorrect
to enlighten the docking results presented above, we used binding mode predicted in docking E, in which two of
X3DNA to evaluate the double helix parameters for the four netropsin pirrol rings are inserted in the intercalation gap
oligomers used as targets. The results are shown in Table 3. (see Figure 6).
As can be concluded by the values of twist, crystal- As also revealed by Table 3, the structural parameters for
lographic DNA from ellipticine complex is more unwound crystallographic and canonical oligomers without an inter-
(29.43°) when compared to canonical B-DNA (35.73°). calation gap are quite similar. As a result, it is not surprising
DOCKING STUDIES ON DNA-LIGAND INTERACTIONS J. Chem. Inf. Model., Vol. 49, No. 8, 2009 1931

that netropsin dockings to these oligomers show similar confirms the assumption that docking of intercalators to DNA
cluster profiles (dockings G and H). However, the average is very sensitive to structural features such as the helix twist
P-P distances in Table 3 must be considered carefully in or small changes in the rise of the intercalation gap.
order to avoid an oversimplification. It may lead to the Comparing Docking Protocols. In order to evaluate the
conclusion that the crystallographic minor groove (11.1 Å) ability of each docking protocol in identifying binding modes,
is only slightly narrower than the canonical minor groove it is interesting to analyze each docking pair, i.e., the same
(11.7). However, it is important to remark that minor groove receptor docked with two different ligands, as in Figure 6.
dimensions in crystallographic DNA are not homogeneous The crystallographic oligomer with intercalation gap used
as occurs in canonical DNA. As should be expected, the in dockings A/E clearly favors intercalative binding mode,
central region ATAT in crystallographic DNA presents a probably because of DNA strongly distorted structure (Figure
significantly narrow minor groove (10.8 Å) in comparison 6A/E). On the other hand, oligomers without intercalation
to the peripheral CG rich (∼13.2 Å) regions in the same gap used in dockings C/G and D/H cannot lead to intercala-
oligomer. This decreased accessibility of the central region tion binding mode, resulting in minor groove recognition as
is very probably the reason why the best docked conforma- the preferred binding mode for both ligands (Figure 6 (parts
tion in self-docking G shows netropsin docked slightly above C/G and D/H)). However, it is important to remember that
the ATAT sequence, in a region where the minor groove is cluster profile for netropsin in docking E shows binding free
not so narrow (see Figure 6). In other words, although a energies that are on average less favorable when compared
narrow minor groove is considered to enhance ligand-DNA to minor groove binding free energies found in dockings F,
interaction in the formed complex, it may not represent the G, and H or to the intercalative binding free energies of a
most favorable intermediate geometry for the approximation real intercalator found in docking A. Analogous, cluster
and fit of the ligand. profile for ellipticine in docking D shows binding free
On the other hand, probably because canonical DNA from energies that are on average less favorable when compared
docking H presents uniform helix geometry with a not so to intercalating binding free energies found in dockings A
narrow minor groove, docking can correctly place netropsin and B or to minor groove binding free energies of a real
in the central AT-rich region (see Figure 6), which is known minor groove ligand found in docking H. Therefore, the
to be the preferential sequence for binding also because of binding free energy profiles from Figure 5 indirectly suggest
the pattern of hydrogen bonding. In summary, it seems that that the predicted binding modes for netropsin and ellipticine
in such rigid target dockings, the importance of hydrogen in dockings E and D are not the preferential binding modes
bonding is limited by the lack of target flexibility and plays for these molecules but artifacts arising from docking
only a secondary role in binding site prediction, which turns methodological limitations.
out to be mainly determined by helix geometry. Finally, the modified canonical B-DNA with intercalation
Moreover, this bold minor groove snuggling in the ATAT gap used in dockings B/F lead, for each ligand, to the correct
central region from crystallographic DNA is probably the binding mode as the energetically most favorable (Figure 6
reason behind the relatively favorable binding free energies (parts A/E and B/F)) and thus proved to be a suitable target
in ellipticine docking C, since ellipticine is placed exactly oligomer in DNA-ligand docking.
in the region where P-P distances reaches the minimum Predicting Tröger Bases Binding Mode. We decided to
value of 10.8 Å (see Figure 6). When ellipticine is docked use the modified canonical B-DNA (with gap) as a receptor
to canonical DNA (dockings B and D), the binding free for docking with the Tröger bases. Therefore, four different
energies from the minor groove recognition are far less dockings were performed: 1) (-)-(R,R) symmetric Tröger
favorable than those from intercalation, clearly indicating base; 2) (+)-(S,S) symmetric Tröger base; 3) (-)-(R,R)
that the former is not the preferential binding mode for asymmetric Tröger base; 4) (+)-(S,S) asymmetric Tröger
ellipticine but an artifact from docking flexibility limitations. base.
Regarding to the modified B-DNA, Table 3 shows that it Among the 100 runs performed in each docking, the 25
presents an almost canonical value of twist (34.38°), and most favorable in terms of binding free energy were
P-P distances are only slightly larger (12.4 Å) than canonical analyzed. Cluster profiles are shown in Figure 7, and the
minor groove distances (11.7 Å). Therefore, modified B- best ranked conformations for each docking are illustrated
DNA is more similar to canonical B-DNA than to the in Figure 8.
crystallographic DNA from ellipticine complex, indicating For both symmetric and asymmetric Tröger bases, cluster
that the artificial mechanism applied to open an intercalation profiles show that binding free energies are more negative
gap in canonical B-DNA implies structural adaptations that for the levorotatory form than for the dextrorotatory one,
are far more subtle than the changes caused by the real indicating an enantiospecific binding of the (-) isomer to
interaction with an intercalator. Hence, it is quite understand- B-DNA. This is in agreement with thermal denaturation
able that netropsin docks to modified B-DNA in a similar studies29,30 and can be explained by the intrinsic geometry
way as it docks to canonical and crystallographic DNA of the compounds; the (-) isomer presents a right-handed
without intercalation gap, as revealed by cluster profiles from helix shape which is similar to B-DNA helices, whereas the
dockings F, G, and H. (+) isomer presents a left-handed helix shape opposite to
Finally, as can be noted by comparing the values of rise B-DNA helices.
for target oligomers, the artificially created gap (6.50 Å) is Indeed, shape complementarity is the reason why docking
not as large as the crystallographic original gap (6.88 Å), of the (-) symmetric Tröger base resulted in binding of the
caused by the real presence of an intercalator. This com- two proflavine moieties to the minor groove with a very
parison contributes to explain why the results for ellipticine favorable binding free energy (-9.90 kcal/mol), while
in docking B were not as satisfactory as in docking A and dockings with the (+) symmetric Tröger base resulted only
1932 J. Chem. Inf. Model., Vol. 49, No. 8, 2009 RICCI AND NETZ

Figure 7. Cluster profiles from docking of Tröger bases to modified B-DNA. Each cluster is represented by a bar in which the height
corresponds to the number of conformations in the cluster, and the color indicates the binding mode: minor groove in black, intercalation
in gray, and other in hatched pattern bimodal binding mode (intercalation/minor groove recognition).

Figure 8. Best docked conformations for dockings of Tröger bases to modified B-DNA. Purple, (-)-(R,R) symmetric Tröger base; magenta,
(+)-(S,S) symmetric Tröger base; dark blue, (-)-(R,R) asymmetric Tröger base; cyan, (+)-(S,S) asymmetric Tröger base.
in intercalation of one proflavine moiety while the other sequence selectivity, suggesting a different binding mode for
proflavine is projected out from the major groove (see top the (+) enantiomer.30 This assumption is also corroborated
views in Figure 8). by our results which indicate that the (+) Tröger bases
Although docking with the (-) symmetric Tröger base interact with DNA via intercalative binding (a binding mode
also resulted in intercalation (-9.67 kcal/mol, 7 runs), minor which is usually associated with lack of sequence selectivity).
groove binding seems to be the preferred binding mode Concerning the (-) asymmetric Tröger base, all docking
qualitatively (-9.90 kcal/mol) and quantitatively (18 runs). runs pointed to a bimodal binding mode, with a large
This is in agreement with electric linear dichroism studies negative binding free energy (-9.84 kcal/mol), in which the
and with the lack of DNA unwinding activity reported by phenanthroline moiety is intercalated whereas the proflavine
Bailly et al.,30 who suggest that the (-) symmetric Tröger moiety is fitted in the minor groove (Figure 8). Again, this
base interacts with DNA through minor groove recognition. binding mode is only possible because ligand chirality is
Moreover, DNase footprinting studies with the symmetric similar to chirality of the B-DNA double helix. On the other
Tröger base proved that only the (-) enantiomer presents hand, dockings of the left-handed (+) asymmetric Tröger
DOCKING STUDIES ON DNA-LIGAND INTERACTIONS J. Chem. Inf. Model., Vol. 49, No. 8, 2009 1933

base resulted only in intercalation with binding free energy

(-9.35 kcal/mol) not as favorable as those from bimodal
binding mode of the (-) isomer.
It has already been suggested that the (-) asymmetric
Tröger base does not interact with DNA through minor
groove recognition since DNase footprinting assays showed
that it presents only a moderated level of sequence selectiv-
ity.31 This moderate sequence selectivity could be explained
by intercalation of one ring moiety, as was suggested by
docking results and also by topoisomerase I inhibition assays
reported by Baldeyrou et al.31 However, our results only
partially agree with experimental data since circular dichro-
ism and electric linear dichroism studies suggest a bimodal
binding mode in which proflavine is intercalated, while
phenanthroline resides in minor groove.31
It is possible that docking results pointed to an alternative
bimodal binding mode (with phenanthroline intercalated) due
to docking flexibility limitations already mentioned. As previ-
ously discussed, the minor groove of an artificial canonical DNA
containing an intercalation gap is not as large as the minor
groove of a crystallographic DNA complexed with an interca-
Figure 9. Cluster profiles from docking of aflatoxin B1 exo-8,9-
lating agent. In other words, the minor groove width of an epoxide with crystallographic DNA (A) and with canonical modified
artificially modified canonical DNA may be too narrow to bear DNA containing intercalation gap (B). Each cluster is represented
the phenanthroline moiety, which means that the target con- by a bar in which the height corresponds to the number of
formation may have guided the docking to this analogous conformations in the cluster, and the color indicates the binding
binding mode in which proflavine - instead of phenanthroline mode: minor groove in black, intercalation in gray, and other in
hatched pattern.
- binds the minor groove. In order to investigate this hypothesis,
we have also performed dockings between the (-) asymmetric
Tröger base and a crystallographic DNA from the ellipticine
complex (data not shown). Among the 25 most favorable runs,
all pointed to the bimodal binding mode suggested by the
literature (with proflavine intercalated and phenanthroline fitted
in minor groove), with a binding free energy of -10.35 kcal/
mol.
Testing Docking Protocol with Aflatoxin B1 exo-8,9-
Epoxide. We decided to apply the generic docking protocol
with a modified canonical B-DNA (with gap) to dock
aflatoxin B1 exo-8,9-epoxide and compare the result with
the result from the self-docking. Therefore, two different
dockings were performed: A) aflatoxin and crystallographic
DNA (1MKL) and B) aflatoxin and modified canonical DNA
(with gap).
Among the 100 runs performed in each docking, the 25
most favorable in terms of binding free energy were
analyzed. Cluster profiles are shown in Figure 9, and the
best ranked conformations for each docking are illustrated
in Figure 10.
The cluster profiles in Figure 9 show similar results for
dockings A and B. Very promising is the fact that both
dockings not only resulted in intercalation binding mode for
the aflatoxin B1 exo-8,9-epoxide but also placed the epoxide
in close proximity and in the proper orientation to the N7 Figure 10. Best docked conformations for dockings of aflatoxin
position of guanine (see Figure 10), thus consistent with the B1 exo-8,9-epoxide with crystallographic DNA (A) and with
modified canonical DNA with intercalation gap (B). CPK repre-
proposed SN2 mechanism of alkylation.32,54 In docking A, sentation in DNA indicates the guanine which forms the bond with
the epoxide carbon which is to be attacked is placed at 3.51 aflatoxin B1 exo-8,9-epoxide. In part (A), the covalent adduct from
Å from the N7 position with a free binding energy of -7.61 original complex is also shown in CPK. Aflatoxin B1 exo-8,9-
kcal/mol, and, in docking B, the same carbon is placed at epoxide is shown in red. Below: conformations are shown in detail,
3.09 Å from the N7 position with a binding free energy of with the carbon which is to be attacked by the N7 position from
guanine indicated by an arrow.
-7.47 kcal/mol.
These very similar results are also noteworthy if we crystallographic DNA from the covalent adduct presents a
consider the structural differences between the two oligomers significant bending toward the major groove (20° according
used as targets. As can be observed in Figure 10, the to Giri et al.33), which does not occur in the modified
1934 J. Chem. Inf. Model., Vol. 49, No. 8, 2009 RICCI AND NETZ

canonical B-DNA. Since such helix features already proved base derived from proflavine and an asymmetric base derived
to be of major importance during target rigid docking, one from proflavine and phenanthroline, each with two optical
may have expected that bending would strongly affect isomers.
aflatoxin dockings and lead to different cluster profiles, which The choice of modified canonical DNA as a receptor for
was not the case. docking with Tröger bases gave rise to very promising
In this particular case, it seems that the helix bending does profiles. Besides resulting in very favorable binding free
not result from an induced fit promoted by intercalation per energies, these dockings were capable of reproducing the
se but arises as an effect from subsequent covalent bonding. stronger DNA binding affinity that the Tröger levorotatory
In other words, the bending does not interfere significantly isomers possess when compared to the dextrorotatory ones.
in aflatoxin dockings because it is not required for the first Moreover, binding modes suggested by the docking are in
noncovalent interaction. This is in agreement with Giri et agreement with the binding modes suggested by experimental
al., who argue that DNA helix bending arises from the change data found in the literature.
in N7 hybridization when the covalent adduct is formed.33 Finally, we tested the proposed docking protocol with
Finally, the fact that both dockings placed the epoxide in aflatoxin B1 exo-8,9-epoxide, an intercalative binding agent
an orientation so favorable for a subsequent alkylation in that also alkylates DNA in the major groove, and compared
the N7 position is in agreement with the hypothesis that the result with that from a self-docking. This comparison
aflatoxin B1 was optimized during evolution to enhance its clearly shows that the choice of modified B-DNA as target
reactivity in DNA environment. This was already supported not only resulted in a binding profile very similar to that
by experimental studies which showed the reaction of from the self-docking but also correctly placed the carbon
aflatoxin B1 exo-8,9-epoxide to be more than 2000 times which is to be attacked in close proximity of the N7 position
more efficient in DNA than in an aqueous solution with free of guanine, in the major groove.
2′-deoxyguanosine.54 Based also on thermodynamic analysis, Therefore, we propose that a default protocol using a
Brown et al. proposed that intercalative binding plays a major modified canonical DNA with an artificial intercalation gap
role guiding the formation of the later adduct between can be successfully applied to investigate other ligands whose
aflatoxin B1 and DNA,54 which is clearly in agreement with binding mode remains unknown, yielding a suitable starting
the noncovalent complexes from dockings A and B. point for further theoretical studies such as more refined
dockings or molecular dynamics simulations.
CONCLUSIONS
ACKNOWLEDGMENT
We have performed several docking studies using ligands
which interact with DNA as intercalators (ellipticine and We would like to thank Hermes L. N. de Amorim and
acridine derivative) or as a minor groove binder (netropsin). Melina Mottin for critical review of the manuscript. We also
acknowledge the financial support from Coordenação de
Four distinct DNA structures were used as targets: crystal-
Aperfeiçoamento de Pessoal de Nı́vel Superior (CAPES) and
lographic DNA from ellipticine complex, crystallographic
Conselho Nacional de Desenvolvimento Cientı́fico e Tec-
DNA from netropsin complex, constructed B-canonical nológico (CNPq).
DNA, and modified B-canonical DNA containing an inter-
calation gap. Direct and cross dockings were carried out with
an optimized set of parameters using AutoDock suite. REFERENCES AND NOTES
Docking results from validation stage are in agreement with (1) Kitchen, D. B.; Decornez, H.; Furr, J. R.; Bajorath, J. Docking and
each binding mechanism (induced-fit or lock-and-key) when scoring in virtual screening for drug discovery: Methods and applica-
tions. Nat. ReV. Drug DiscoVery 2004, 3, 935–949.
the system is subjected to the flexibility limitations of the (2) Moitessier, N.; Englebienne, P.; Lee, D.,; Lawandi, J.; Corbeil, C. R.
docking method: the docking with ellipticine was found to be Towards the development of universal, fast and highly accurate
more sensitive to the selected target conformation than the docking/scoring methods: a long way to go. Br. J. Pharmacol. 2008,
153, S7–S26.
docking with netropsin. In cross dockings, this observation is (3) Grootenhuis, P. D. J.; Kollman, P. A.; Seibel, G. L.; DesJarlais, R. L.;
more evident: netropsin can dock even to the minor groove of Kuntz, I. D. Computerized selection of potential DNA binding
a structurally distorted DNA, whereas ellipticine cannot inter- compounds. Anti-Cancer Drug Des. 1990, 5, 237–242.
(4) Grootenhuis, P. D. J.; Roe, D. C.; Kollman, P. A.; Kuntz, I. D. Finding
calate when the receptor does not possess a proper intercalation potential DNA-binding compounds by using molecular shape. J. Com-
gap. Also, our results make more evident the important role put.-Aided Mol. Des. 1994, 8, 731–750.
that DNA features such as twist, rise, or groove width play in (5) Chen, K.; Adelstein, S. J.; Kassis, A. I. Molecular modeling of the
interaction of the iodinated Hoechst analogs with DNA: implications
rigid dockings, reinforcing the necessity to treat target flexibility for new radio-pharmaceutical design. J. Mol. Struct.-THEOCHEM
in dockings concerning nucleic acids. 2004, 711, 49–56.
Nevertheless, docking results also proved that the current (6) Rohs, R.; Bloch, I.; Sklenar, H.; Shakked, Z. Molecular flexibility in
ab initio drug docking to DNA: binding-site and binding-mode
limitations of docking methods can be overcome by a proper transitions in all-atom Monte Carlo simulations. Nucleic Acids Res.
choice of the target conformation. We found that, provided 2005, 33, 7048–7057.
(7) Tuttle, T.; Kraka, E.; Cremer, D. J. Docking, triggering and biological
that target DNA presents an artificial intercalation gap, activity of dynemicin A in DNA: a computational study. J. Am. Chem.
dockings pointed to the correct binding mode as the Soc. 2005, 127, 9469–9484.
energetically most favorable result, also suggesting that the (8) Evans, D. A.; Neidle, S. Virtual screening of DNA minor groove
binders. J. Med. Chem. 2006, 39, 4232–4238.
AutoDock score function is efficient to evaluate ligand-DNA (9) Sobhani, A. M.; Amini, S. R.; Tyndall, J. D. A.; Azizi, E.; Daneshtalab,
interactions at least in a qualitative way. M.; Khalaj, A. A theory of mode of action of azolylalkylquinolines
We further applied this docking protocol using a modified as DNA binding agents using automated flexible ligand docking. J.
Mol. Graphics Modell. 2006, 25, 459–469.
canonical B-DNA with an intercalation gap as receptor to (10) Lauria, A.; Patella, C.; Ippolito, M.; Almerico, A. M. Docking and
predict the binding modes of two Tröger bases: a symmetric synthesis of pyrrolopyrimidodiazepinone derivatives (PPDs) and their
DOCKING STUDIES ON DNA-LIGAND INTERACTIONS J. Chem. Inf. Model., Vol. 49, No. 8, 2009 1935

precursors: New scaffolds for DNA-interacting agents. J. Mol. Struct.- (34) Bren, U.; Václav, M.; Florián, J. Free energy simulations of uncatalyzed
THEOCHEM 2007, 819, 26–31. DNA replication fidelity: structure and stability of T · G and dTTP · G
(11) Navarrete, J. T.; Casado, J.; Ram, F. J. Electronic spectroscopy study terminal DNA mismatches flanked by a single dangling nucleotide.
and molecular docking simulation of the interaction of terthiophene J. Phys. Chem. B 2006, 110, 10557–10566.
with DNA. J. Mol. Struct. 2007, 834, 176–181. (35) Huey, R.; Morris, G. M.; Olson, A. J.; Goodsell, D. S. A semiempirical
(12) Yan, Z.; Sikri, S.; Beveridge, D. L.; Baranger, A. M. Identification of free energy force field with charge-based desolvation. J. Comput.
an aminoacridine derivative that binds to RNA tetraloops. J. Med. Chem. 2007, 28, 1145–1152.
Chem. 2007, 50, 4096–4104. (36) Solis, F. J.; Wets, J.-B. Minimization by random search techniques.
Math. Oper. Res. 1981, 6, 19–30.
(13) Holt, P. A.; Chaires, J. B.; Trent, J. O. Molecular docking of (37) http://www.rcsb.org/pdb/home/home.do (accessed July 13, 2009).
intercalators and groove-binders to nucleic acids using AutoDock and (38) Canals, A.; Purciolas, M.; Aymami, J.; Coll, M. The anticancer agent
Surflex. J. Chem. Inf. Model. 2008, 48, 1602–1615. ellipticine unwinds DNA by intercalative binding in an orientation
(14) Hurley, L. H. DNA and its associated processes as targets for cancer parallel to base pairs. Acta Crystallogr., Sect D: Biol. Crystallogr.
therapy. Nat. ReV. Cancer 2002, 2, 188–199. 2005, 61, 1009–1012.
(15) Sternberg, M. J. E.; Gabb, H. A.; Jackson, R. M. Predictive docking (39) Finlay, A. C.; Hochstein, F. A.; Sobin, B. A.; Murphy, F. X. Netropsin,
of protein-protein and protein-DNA complexes. Curr. Opin. Struct. a new antibiotic produced by Streptomyces. J. Am. Chem. Soc. 1951,
Biol. 1998, 8, 250–256. 73, 341–343.
(16) Dickerson, R. E.; Drew, H. R.; Conner, B. N.; Wing, R. M.; Fratini, (40) Coll, M.; Aymami, J.; van der Marel, G. A.; van Boom, J. H.; Rich,
A. V.; Kopka, M. L. The anatomy of A-, B-, and Z-DNA. Science A.; Wang, A. H. Molecular structure of the netropsin-d(CGC-
1982, 216, 475–485. GATATCGCG) complex: DNA conformation in an alternating seg-
(17) Dickerson, R. E. DNA structure from A to Z. Methods Enzymol. 1992, ment. Biochemistry 1989, 28, 310–320.
211, 67–111. (41) Lindahl, E.; Hess, B.; van der Spoel, D. GROMACS 3.0: a package
(18) Phan, A. T.; Kuryavyi, V.; Patel, D. J. DNA architecture from G to for molecular simulation and trajectory analysis. J. Mol. Model. 2001,
Z. Curr. Opin. Struct. Biol. 2006, 16, 288–298. 7, 306–317.
(19) Hannon, M. J. Supramolecular DNA recognition. Chem. Soc. ReV. (42) GROMACS, Version 3.3; Groningen, Netherlands, 2007. Available from
2007, 36, 280–295. http://www.gromacs.org/ (accessed July 13, 2009).
(20) Tse, W. C.; Boger, D. L. Sequence-Selective DNA Recognition: (43) Oostenbrink, C.; Soares, T. A.; van der Vegt, N. F. A.; van Gunsteren,
Natural products and Nature’s lessons. Chem. Biol. 2004, 11, 1607– W. F. Validation of the 53A6 GROMOS force field. Eur. Biophys. J.
1617. 2005, 34, 273–284.
(44) Gasteiger, J.; Marsili, M. Iterative partial equalization of orbital
(21) Waring, M. J.; Bailly, C. DNA recognition by intercalators and hybrid electronegativity - a rapid access to atomic charges. Tetrahedron 1980,
molecules. J. Mol. Recognit. 1994, 7, 109–122. 36, 3219–3228.
(22) Dervan, P. B. Molecular recognition of DNA by small molecules. (45) Sanner, M. F. Python: a programming language for software integration
Bioorg. Med. Chem. 2001, 9, 2215–2235. and development. J. Mol. Graphics Modell. 1999, 17, 57–61.
(23) Bailly, C.; Chaires, B. Sequence-specific DNA minor groove binders. (46) Humphrey, W.; Dalke, A.; Schulten, K. VMD - Visual Molecular
Design and synthesis of netropsin and distamycin analogues. Biocon- Dynamics. J. Mol. Graphics 1996, 14, 33–38.
jugate Chem. 1998, 9, 513–538. (47) Coll, M.; Frederick, C. A.; Wang, A.D.-J.; Rich, A. A bifurcated
(24) Dolenc, J.; Borstnik, U,; Hodoscek, M.; Koller, J.; Janezic, D. Ab hydrogen-bonded conformation in the d(A · T) base pairs of the DNA
initio QM/MM study of the conformational stability of the complexes dodecamer d(CGCAAATTTGCG) and its complex with distamycin.
formed by netropsin and DNA. The importance of van der Waals Proc. Natl. Acad. Sci. U.S.A. 1987, 84, 8385–8389.
interactions and hydrogen bonding. J. Mol. Struct.-THEOCHEM. 2005, (48) Hopcroft, N. H.; Brogden, A. L.; Searcey, M.; Cardin, C. J. X-ray
718, 77–85. crystallographic study of DNA duplex crosslinking: simultaneous
(25) Pindur, U.; Jansen, M.; Lemster, T. Advances in DNA-ligands with binding to two d(CGTACG)2 molecules by a bis(9-aminoacridine-4-
groove binding, intercalating and/or alkylating activity: Chemistry, carboxamide) derivative. Nucleic Acids Res. 2006, 34, 6663–6672.
DNA-binding and biology. Curr. Med. Chem. 2005, 12, 2805–2847. (49) Lu, X.-J.; Olson, W.-K. 3DNA: a software package for the analysis,
rebuilding and visualization of three-dimensional nucleic acid struc-
(26) Bren, U.; Hodoscek, M.; Koller, J. Development and validation of
tures. Nucleic Acids Res. 2003, 31, 5108–5121.
empirical force field parameters for netropsin. J. Chem. Inf. Model.
(50) Guex, N.; Peitsch, M. C. SWISS-MODEL and the Swiss-PdbViewer:
2005, 45, 1546–1552.
na environment for comparative protein modeling. Electrophoresis
(27) de Souza, O. N.; Ornstein, R. L. Inherent DNA curvature and flexibility 1997, 18, 2714–2723.
correlate with TATA Box functionality. Biopolymers 1998, 46, 403– (51) Dennington, R., II; Keith, T.; Millam, J. GaussView, Version 4.1;
415. Semichem, Inc.: Shawnee Mission, KS, 2007.
(28) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.; (52) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb,
Belew, R. K.; Olson, A. J. Automated docking using Lamarckian M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.;
Genetic Algorithm and an empirical binding free energy function. Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels,
J. Comput. Chem. 1998, 19, 1639–1662. A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone,
(29) Tatibouët, A.; Demeunynck, M.; Andraud, C.; Collet, A.; Lhomme, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.;
J. Synthesis and study of an acridine substituted Tröger’s Base: Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.;
preferential binding of the (-)-isomer to B-DNA. Chem. Commun. Morokuma, K.; Salvador, P.; Dannenberg, J. J.; Malick, D. K.; Rabuck,
1999, 2, 161–162. A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.;
(30) Bailly, C.; Laine, W.; Demeuninck, M.; Lhomme, J. Enantiospecifc Baboul, A. G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.;
recognition of DNA sequences by a proflavine Tröger base. Biochem. Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-
Biophys. Res. Commun. 2000, 273, 681–685. Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill,
(31) Baldeyrou, B.; Tardy, C.; Bailly, C.; Colson, P.; Houssier, C.; P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.;
Charmantray, F.; Demeuninck, M. Synthesis and DNA interaction of Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian
a mixed proflavine-phenanthroline Tröger base. Eur. J. Med. Chem. 98 (ReVision A.1x); Gaussian, Inc.: Pittsburgh, PA, 2001.
2002, 37, 315–322. (53) Bren, U.; Guengerich, F. P.; Mavri, J. Guanine alkylation by the potent
carcinogen aflatoxin B1: quantum chemical calculations. Chem. Res.
(32) Iyer, R. S.; Coles, B. F.; Raney, K. D.; Thier, R.; Guengerich, F. P.; Toxicol. 2007, 20, 1134–1140.
Harris, T. M. DNA adduct by the potent carcinogen aflatoxin B1: (54) Brown, K. L.; Bren, U.; Stone, M. P.; Guengerich, F. P. Inherent
mechanistic studies. J. Am. Chem. Soc. 1994, 116, 1603–1609. stereospecificity in the reaction of aflatoxin B1 8,9-epoxide with the
(33) Giri, I.; Jenkins, M. D.; Schnetz-Boutaud, N. C.; Stone, M. P. Structural deoxyguanosine and efficiency of DNA catalysis. Chem. Res. Toxicol.
refinement of the 8,9-dihydro-8-(N7-guanyl)-9-hydroxy-aflatoxin B1 2009, 22, 913–917.
adduct in a 5′-CpAFBG-3′ sequence. Chem. Res. Toxicol. 2002, 15,
638–647. CI9001537