Designed compounds for recognition of 10 base Pairs of DNA with two at binding sites

Article abstract of DOI:10.1021/ja211628j

Short AT base pair sequences that are separated by a small number of GCs are common in eukaryotic parasite genomes. Cell-permeable compounds that bind effectively and selectively to such sequences present an attractive therapeutic approach. Compounds with

Full text of DOI:10.1021/ja211628j

Article
pubs.acs.org/JACS
Designed Compounds for Recognition of 10 Base Pairs of DNA with
Two AT Binding Sites
Yang Liu,† Yun Chai,†,§ Arvind Kumar,† Richard R. Tidwell,‡ David W. Boykin,† and W. David Wilson*,†
†Department of Chemistry, Georgia State University, Atlanta, Georgia 30302-4098, United States
‡Department of Pathology, University of North Carolina, Chapel Hill, North Carolina 27599, United States
S
* Supporting Information
ABSTRACT: Short AT base pair sequences that are separated by
a small number of GCs are common in eukaryotic parasite
genomes. Cell-permeable compounds that bind effectively and
selectively to such sequences present an attractive therapeutic
approach. Compounds with linked, one or two amidine-
benzimidazole-phenyl (ABP) motifs were designed, synthesized,
and evaluated for binding to adjacent AT sites by biosensor-surface
plasmon resonance (SPR). A surprising feature of the linked ABP
motifs is that a set of six similar compounds has three different
minor groove binding modes with the target sequences. Compounds with one ABP bind independently to two separated AT
sites. Unexpectedly, compounds with two ABP motifs can bind strongly either as monomers or as cooperative dimers to the full
site. The results are supported by mass spectrometry and circular dichroism, and models to explain the different binding modes
are presented.
If we focus on genes and their control sequences while
recognizing that not all genes are expressed in a specific cell,
that number is reduced. Although the design concepts are
general and could be extended to target DNA in a variety of
cells, we have specifically focused on the mitochondrial
kinetoplast genome of trypanosomes as an optimum, initial
test system.11,21−23 The kinetoplast consists of thousands of
interlocked AT-rich DNA minicircles that code for guide RNAs
that control mRNA editing.1,24 The mRNAs are coded by
maxicircle DNAs that are also part of the kinetoplast matrix.
The possibility of targeting the kinetoplast as a unique DNA
target structure is recognized,11,21−23 but new drugs to
selectively target the unusual structure have not yet reached
the final approval stage.25 For the small mitochondrial genomes
of these parasitic microorganisms, a 10 base pair target site can
give the desired selectivity.20 Trypanosomes cause sleeping
sickness (Trypanosoma brucei) in Africa25 and Chagas disease
(Trypanosoma cruzi) from the southern United States to
Argentina.26 The organisms infect and kill many thousands of
people each year and current drugs to treat the associated
diseases have serious limitations from toxicity to resist-
ance.11,25,27
INTRODUCTION
■
The concept of specific control of cellular gene expression by
designed compounds is a key goal of chemical biology and
offers many advantages in the development of new types of
drugs as well as agents for biotechnology. The control could be
affected, for example, by directly targeting genes with designed
oligomers, such as PNA,1,2 or with a variety of small
molecules.3−5 Sequences for most of the 20−25 thousand
human genes are now known and all of them are potentially
susceptible to external control.6 The sequences of the genomes
of a large number of disease causing microorganisms have also
been determined and offer selective strategies to control or
destroy the organisms.7,8 Designed compounds that have
biological activity against either duplex,4,5,9−12 triplex and
triplet-repeat DNA,13,14 or quadruplex DNA structures15−19
have also been identified and are under development.
The design of optimum compounds to target DNA for a
therapeutic outcome must obey several rules. The compounds
obviously must be cell permeable and this limits the group
types that can be combined on a core compound template. The
compounds must be large enough to give sufficient affinity and
selectivity to generate activity while maintaining low toxicity.
The cost of synthesis can also be a factor, especially for
compounds that are designed for use against diseases such as
malaria and sleeping sickness that affect many of the world’s
most financially depressed regions.
Both the kinetoplast of trypanosomes and the nuclear
genome of malaria are quite AT rich with significant regions of
AT base pairs separated by a small number of GC base
pairs.9,11,21,22,24 To simplify and reduce the costs of synthesis,
symmetric compounds have been prepared (Figure 1). With
appropriate linking groups, the symmetry of such compounds
With these constraints as limiting features, we have focused
in this work on the design of relatively simple compounds that
can selectively target a 10 base pair sequence in the minor
groove of DNA. To uniquely bind to a site in the entire human
genome, about 15−16 base pairs must be in the binding site.20
Received: December 13, 2011
Published: February 27, 2012
© 2012 American Chemical Society
5290
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
Figure 1. Structure of the compounds and the DNA sequences used in this study. For SPR experiments, 5′-biotin labeled DNA sequences are used.
Information. The purity of all compounds was verified by NMR and
elemental analysis.
makes them ideal for targeting two AT sites separated by GC
base pairs. The amidine-benzimidazole-phenyl (ABP) motif is
effective for targeting AT sites in DNA and we have
incorporated that motif in our design strategy. In an initial
set of test compounds, the symmetric compounds RT546,
DB2114, DB2115, and DB2119 were designed with two ABP
motifs, to specifically target two AT sequences with a flexible
linker to connect them across one or more GC base pairs
(Figure 1). Compounds DB184 and RT533, with single ABP
motifs, are included as controls. The concept of connecting
recognition units to recognize two or more sites in proteins or
DNAs has been used in other systems.28 In DNA, combilexins
connect an intercalator and minor groove binding agent;29−31
bis- or polyintercalators can intercalate and bind in both the
major and minor grooves;32−35 minor groove alkylating groups
can be placed on two linker groove recognizing units to alkylate
at two sites;12,36 and linked polyamide minor groove agents can
also bind at adjacent sites.37
The key questions that we wish to address with these
compounds are (i) what level of binding affinity can be
obtained with compounds of this type and DNA sequences
containing AT base pair sites separated by one or more GC
base pairs which are common in the parasitic organisms (Figure
1); (ii) do the compounds fold back through the flexible linkers
to bind to single AT sites or can a single compound bind to
both of the AT sites as desired; and (iii) for compounds that
can bind simultaneously to both sites, what correlation is there
between linker length, rigidity and the number of GC pairs
between the AT sites? These questions can be answered with
the designed set of compounds in Figure 1. The results
presented here are quite surprising and show that even this
small set of compounds has unique interactions with DNA
including three completely different binding modes, even with
the relatively simple set of DNA sequences in Figure 1.
In SPR experiments, 5′-biotin labeled hairpin DNA oligomers were
used (DNA sequences shown in Figure 1B). In mass spectrometry
(MS), circular dichroism (CD), and Tm experiments, the hairpin DNA
oligomers used were AATT [5′-GGAATTCGCTCTCGAATTCC-3′],
AATTGAATT [5′-GGAATTGAATTCGCTCTCGAATT-
CAATTCC-3′] and AATTGCAATT [5′-GGAATTG-
CAATTCGCTCTCGAATTGCAATTCC-3′] with the hairpin loop
sequences underlined. All DNA oligomers were obtained from
Integrated DNA Technologies, Inc. (IDT, Coraville, IA) with
reverse-phased HPLC purification and mass spectrometry character-
ization. The CCL10 buffers used in CD and Tm experiments contained
0.01 M cacodylic acid, 0.001 M EDTA, and 0.1 M NaCl, pH 6.25. The
SPR experiments were performed in filtered, degassed CCL10 buffer
with 5 × 10−3% (v/v) Surfactant P20 and 0.1 M ammonium acetate,
pH 7.0, was used in MS experiments.
Biosensor-Surface Plasmon Resonance (SPR) Studies. SPR
measurements were performed with a four-channel Biacore T100 or
T200 optical biosensor system (Biacore, GE Healthcare, Inc.). 5′-
Biotin labeled DNA samples [AATT, AATTGAATT, AATTG-
CAATT, AATTGCGAATT, AATTGCGCAATT, AATTGCGC-
GAATT, AATTGCGCGCAATT hairpins] were immobilized onto
streptavidin-coated sensor chips (Biacore SA) as previously
described.40,41 Three flow cells were used to immobilize the DNA
oligomer samples, while a fourth cell was left blank as a control. SPR
binding analysis was performed with multiple injections of different
compound concentrations over the immobilized DNA surface at a flow
rate of 25 μL/min and 25 °C. For steady-state analysis, the number of
binding sites and the equilibrium constant were obtained from fitting
plots of RU versus Cfree. Binding results from the SPR experiments
were fit with one- (K2 = 0) or two-site interaction models:
2
2
r = (K C
1 free
+ 2K K C
1 2 free
)/(1 + K C
1 free
+ K K C )
1 2 free
(1)
where r represents the moles of bound compound per mole of DNA
hairpin duplex, K1 and K2 are macroscopic binding constants, and Cfree
is the free compound concentration in equilibrium with the
complex.41,42 Kinetics fits were used when no steady state was
reached as previously described.41,43
Electrospray Mass Spectrometry (ESI−MS) Studies. ESI−MS
experiments were performed on a Q-TOF micro (Waters Micromass,
Manchester, U.K.) with its standard ESI source and general published
procedures for DNA.44,45 The electrospray source was operated in the
MATERIALS AND METHODS
Materials. Syntheses of DB2114, DB2115, and DB2119 were done
much the same as with published compounds DB184, RT546, and
RT533,38,39 and the detailed syntheses are described in the Supporting
■
5291
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
Figure 2. SPR binding affinity: (A) SPR sensorgrams for the interaction of selected compounds with AATT (upper panel) and AATTGCAATT
(lower panel) DNA sequences. (B) Comparison of the SPR binding affinity for AATT (blue circles) and AATTGCAATT (red squares) DNA
sequences with different compounds. RU values from the steady−state region of SPR sensorgrams are converted to r (r = RU/RUmax) and are
plotted against the unbound compound concentration (flow solution) for DB184, RT546, RT533, and DB2119. The lines are the best fit values to a
single site or two-site interaction models and K values are in Table 1.
negative ion mode with a needle voltage of −2.2 kV. The cone voltage
was set to −30 V, and the RF lens1 voltage to −60 V. The hexapole
collision voltage of 3 V and the ion energy of 1.5 V were used for full
scan MS. Source block and desolvation temperatures were set to 70
and 100 °C, respectively. Spectra were acquired from 200 to 3200 m/z
and only a portion of the mass range was shown for clarity. DNA
oligomers were dialyzed in 0.1 M ammonium acetate, pH 7.0, using
Spectra/Por 7 dialysis membranes (molecular weight cut off: 1000 Da;
Spectrum Laboratories, Inc., CA) to remove the nonvolatile ions.
Experiments were generally conducted at a concentration of 5 × 10−6
M for hairpin DNA and the ratios of compound to DNA (1:1, 2:1, and
3:1) were obtained by adding compound to DNA solution.
CD Titration Studies. CD spectra were recorded using a Jasco J-
810 instrument with a 1 cm cell and a scan speed of 50 nm/min with a
response time of 1 s. The spectra from 500 to 220 nm were averaged
over five scans. A buffer baseline scan was collected in the same
cuvette and subtracted from the average scan for each sample. The
titration experiments were performed at 25 °C. The desired ratios of
compound to DNA were obtained by adding compound to the cell
containing a constant amount of DNA, and in all cases, the absorbance
was kept below 1.0. Data processing and plotting were performed with
Kaleidagraph software.
Molecular Docking. Molecular docking and visualization experi-
ments were performed with the SYBYL-X1.2 software package on a
Windows 8 processor Workstation.46 The initial DNA duplexes 5′-
d(GGAATTGAATTCG)-3′ and 5′-d(GGAATTGCAATTCG)-3′
were constructed in the Biopolymer-Build DNA Double Helix module
employing regular B-DNA parameters. To accommodate two ligands
binding in the minor groove of the duplex 5′-d-
(GGAATTGCAATTCG)-3′, the define distance-dependent con-
straints option was used to widen the minor groove. Two different
models were constructed. The middle six base pairs were widened for
the overlapped model (see Results for model descriptions), while only
four base pairs were widened for the offset model. The groove was
then allowed to decrease to the original width. These three DNA
sequences were next energy minimized for a maximum of 100
iterations using the conjugate gradient algorithm and Tripos force
field, with a termination gradient of 0.1 kcal/(mol Å). RT546 was
assigned Gasteiger-Huckel charges and minimized using the Tripos
̈
force field until a terminating conjugate gradient of 0.01 kcal/(mol Å)
or the maximum 1000 iterations was reached.47
To help position the bound compound in the AATTGAATT site,
the crystal structure, 3GJH, a 6-amidine-2-(4-amidino-phenyl)indole
in complex with 5′-d(CGCGAATTCGCG)-3′ from the protein data
bank, was selected. The DNA duplex containing the AATTGAATT
site was aligned to the crystallized sequence such that AATT sites were
overlaid and the DNA sequence from 3GJH was removed. This
process was then repeated at the second AATT. A protomol was
generated using a ligand-based approach with the Surflex-Dock
module. During the protomol generation, the two parameters
“Threshold” and “Bloat” were set to 0.5 and 2, respectively.
“Threshold” determines how much the protomol extends into the
minor groove, while “Bloat” impacts how far the protomol expands in
three-dimensional space. All other parameters were left at the default
values. After generation of the protomol, the GeomX module was used
to dock RT546 (Figure 1).48,49
The two AATTGCAATT duplexes were used with the RT546
dimer and the compound was manually docked to the widened minor
groove. The complex was minimized with DNA fixed until a derivative
of 0.01 kcal/(mol Å) was reached. Before each docking, the RT546
dimer was moved to a separate molecular area, so that the ligands can
be moved independently of the DNA.47 The genetic algorithm of the
Flexidock module was employed implementing five different random
numbers and 678 000 generations.47,50 Both the ligands and the bound
DNA were permitted torsional flexibility in the docking process.
Atomic charges were computed using the Kollman all-atom for DNA
and Gasteiger-Huckel for the ligands. All the possible hydrogen-bond
̈
sites were selected for the DNA−compound complex. From each
docking, the 20 lowest-energy structures were selected.
RESULTS
■
Biosensor-Surface Plasmon Resonance: Determining
Binding Affinity, Stoichiometry, and Cooperativity. SPR
experiments were used to quantitatively compare binding
5292
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
Table 1
a
b
K = (KlK2)0.5. CF (Cooperativity Factor) = (K2/K1) × 4.
affinity, stoichiometry, and cooperativity of compound
interactions with immobilized AATT and AATTGCAATT
sequences (Figure 2).40,41,43 Essentially, the same moles of the
DNA oligomers were immobilized on the surface of each sensor
chip so that the sensorgram saturation levels can be compared
directly for stoichiometry. The differences in interaction
strength and binding stoichiometry for four example com-
pounds are visualized in a plot of r (moles of compound
bound/mol of hairpin DNA) versus Cfree, the free compound
concentration (Figure 2B). The plots were best fit with a one-
site binding model for AATT and a two-site binding model for
AATTGCAATT (K values are summarized in Table 1).
The relative values of the macroscopic equilibrium constants,
K1 and K2, also reflect the cooperativity of the interaction. A
cooperativity factor to assess the degree of cooperativity is
defined as CF = (K2/K1) × 4. CF is equal to 1 for interactions
with no cooperativity, greater than 1 for positive cooperativity,
and less than 1 for negative cooperativity. As can been seen in
Figure 2 and Table 1, DB184 binds AATT as a monomer and
AATTGCAATT at two sites without cooperativity, as expected.
In a similar manner, RT533 binds AATT very strongly as a
monomer, while it binds AATTGCAATT as a 2:1 complex
with no significant cooperativity. Surprisingly, the linked
compound, RT546, binds in a completely different manner,
as a weak monomer to AATT, but as a strong dimer with
positive cooperativity to AATTGCAATT (CF > 3000).
DB2119, with a more rigid linker, does not bind to AATT
while it also binds to AATTGCAATT as a dimer with positive
cooperativity, as with RT546. These SPR results are supported
by thermal melting studies. DB184 has a very small increase in
Tm in agreement with its weak binding by SPR. Both RT546
and DB 2119 have biphasic curves in agreement with their
positive binding cooperativity. RT533 has a monophonic curve
in agreement with its lack of binding cooperativity (Supporting
Information, Figure S1). Tm curves for all of the two-site
compounds do not have an observable high temperature
baseline and the Tm values cannot be determined accurately. It
is clear, however, that the Tm increases scale in the same general
order as binding constants.
SPR sensorgrams were also obtained for two additional
DNAs; a sequence of similar length without a GC linker, A5T5,
and a sequence with one AATT binding site mutated to AGTC,
AATTGCAGTC (Figure S2, Supporting Information). RT546
and DB2119 bind strongly as monomers to A5T5 (Figure S2,
Supporting Information). RT533 binds A5T5 as a 2:1 complex
with negative cooperativity. With AATTGCAGTC, RT546 and
5293
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
RT533 bind as monomers with binding constants very close to
the values for a single site AATT (Table 1 and Figure 2). As
with AATT, no binding is observed for DB2119 with
AATTGCAGTC (Figure S2, Supporting Information).
In additional studies with DNA sequences that have an
extended GC linker length, it was observed that the linker
length has no significant effect on binding of the control
compounds, DB184 and RT533 (Table 1). As the GC length
increases, RT546 still forms a 2:1 complex but with decreasing
cooperativity. Figure S3 (Supporting Information) shows that
both the association and dissociation rates increase for RT546
binding with 0 to 3 GC. DB2119, however, loses significant
binding affinity and cooperativity with increasing GC length
and has no detectable binding for the 5 and 6 GC base pair
spacer (Table 1).
To investigate how the linker length in RT546 affects the
DNA recognition pattern, two longer compounds with one
(DB2115) and two (DB2114) additional −CH2− groups were
synthesized. For the linker [-O-(CH2)n-O-], the binding affinity
and cooperativity for DNA sequences can vary by a surprisingly
large amount as n increases from 3 to 5 (Figure 3, Table S1,
highest binding affinity and cooperativity for a 2:1 complex
with AATTGCAATT.
Electrospray Ionization Mass Spectrometry (ESI−MS):
Characterizing the Binding Stoichiometry and Cooper-
ativity. ESI−MS experiments allow the resolution of complex
mixtures and determination of stoichiometries and any binding
cooperativity for complexes that are present simultaneously in
an injected sample.44,45 Figure S4A (Supporting Information)
shows ESI−MS spectra of AATT with the three linked
compounds RT546, RT533, and DB2119 at different
compound to DNA ratios. A high intensity 1:1 complex is
detected for RT533, only a low intensity 1:1 complex peak is
observed for RT546, and no peak is observed for DB2119.
Figure S4B (Supporting Information) shows ESI−MS spectra
of AATTGAATT with the same compounds. Peaks for both a
1:1 (labeled as blue) and a 2:1 complex (labeled as red) can be
observed for RT533, even at low compound to DNA ratios. As
the ratio is increased, the intensity of the 2:1 complex peak
increases, but no higher order complexes are observed. For
RT546 and DB2119, only a 1:1 complex peak is observed as the
ratio increases, indicating a monomer complex. ESI−MS results
for AATTGCAATT with the three compounds RT546, RT533,
and DB2119 are summarized in Figure S4C (Supporting
Information). For RT546, the peak for the 2:1 complex is much
larger than the 1:1 complex peak indicating strong cooperative
dimer formation. With RT533, the peak for the 1:1 complex is
much larger than the peak for the 2:1 complex at low
compound to DNA ratios. With increasing ratio, the peak for
the 2:1 complex increases, and the 1:1 complex peak decreases.
This clearly shows that RT533 binds to AATTGCAATT as a
2:1 complex with no cooperativity. For DB2119, only the peak
of the 2:1 complex is detected as the compound to DNA ratio
increases, indicating that DB2119 binds to AATTGCAATT as
a dimer with very strong positive cooperativity. To visualize the
different stoichiometries for AATTGCAATT and AATT
binding with the three linked compounds more easily, ESI−
MS spectra of three DNA sequences with RT546, RT533, and
DB2119 at a 2:1 ratio (compound to DNA) are compared in
Figure 4. All of these results are in excellent agreement with the
findings from SPR.
Circular Dichroism (CD): Probing the Binding Mode
and Binding Ratio. CD spectra monitor the asymmetric
environment of the compounds when bound to DNA and
therefore can be used to obtain information on the binding
mode.51 Free RT546 does not exhibit CD signals; however,
upon addition of the compound to AATTGCAATT,
substantial, positive induced CD signals arise in the compound
absorption region between 300 and 400 nm (Figure S5A,
Supporting Information). These positive induced CD signals
for the complexes are characteristic patterns for minor groove
binding in AT sequences.51 CD changes are significantly
smaller for AATT than for AATTGCAATT. CD spectra for the
complexes of RT533, RT546, and DB2119 with AATTG-
CAATT sequence are compared in Figure S5B (Supporting
Information) and display large positive induced CD spectral
changes, as expected, for minor groove complexes. For all three
compounds, the saturation maximum is 2 compounds/hairpin
for AATTGCAATT, and these values are consistent with the
results from SPR experiments. The induced CD signals for
DB2119 with a more rigid linker are larger than for the more
flexible compounds, suggesting a stronger dipole interaction
with the bases in this compound.51
Figure 3. The effect of compounds linker length changes on DNA
binding: (A) SPR sensorgrams for the interaction of DNA sequences
(AATT, AATTGAATT, and AATTGCAATT) binding with RT546
[-O-(CH2)3-O-], DB2115 [-O-(CH2)4-O-], and DB2114 [-O-(CH2)5-
O-]. (B) Comparison of the SPR binding affinity for DNA sequences
binding with compounds with different linker length. The lines are the
best fit values to a single site or two site interaction models.
Supporting Information). RT546, n = 3, binds AATT as a weak
monomer, for example, while no binding is observed for the
longer compounds, DB2115 and DB2114 (n = 4 or 5). With
AATTGAATT, all three compounds bind strongly as
monomers and with AATTGCAATT all three bind as
cooperative 2:1 dimers. As can be seen in Figure 3B and
Table S1, the three compounds slightly lose binding affinity for
AATTGCAATT when the linker length is increased and the
cooperativity for binding decreases significantly (about 100
times) as n increases from 3 to 5. Apparently, RT546 has an
optimum linker length in this general structure and has the
5294
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
Figure 4. Comparison of ESI−MS spectra of DNA sequences with RT546, RT533, and DB2119 at a 2:1 ratio (compound to DNA): (A−D) the
AATT DNA sequence; (E−H) the AATTGAATT DNA sequence; (I−L) the AATTGCAATT DNA sequence. More detailed MS spectra are in
Supporting Information.
Figure 5. Models for RT546 docked into the (A) AATTGAATT site, Model 2 in Figure 6; and (B and C) the AATTGCAATT sequence (see
Model 3 in Figure 6). The complex in (B) is highly overlapped while in (C) the complex is offset. For clarity, the terminal bases are not displayed in
the images, and the images are of only the lowest-energy conformation. The DNA backbone is shown by yellow ribbons and the GC base pairs are
displayed by ball and stick type in magenta, while the ligand is displayed as spacefill type in CPK colors. More detailed modeling Figures of the
models are in Supporting Information, Figure S6.
Molecular Docking. To provide ideas for a better
understanding of the two site binding modes of symmetric
compounds, RT546 was docked into the two AATT sites with
G and GC spacing sequences (Figure 5) as described in the
Materials and Methods (additional views of the models are
shown in Figure S6). The monomer docking at AATTGAATT
is similar to classical minor groove binding compounds (Figure
5A). The groove is relatively narrow along the entire two site
length. The ABP modules fit nicely into the AATT sites with
amidine and benzimidazole H-bonds to the floor of the groove.
The trimethylene linker binds into the groove at the GC base
pair as expected. Although minor variations of this model are
possible, and certainly occur on a dynamic basis, any major
movement of RT546 in this site results in a significant loss of
binding interactions and is unlikely.
For the AATTGCAATT site, two different models with
favorable interactions for the RT546 2:1 complex are presented.
In Figure 5B, the stacked molecules are slightly offset such that
the amidine of the top molecule of the figure stacks over the
benzimidazole of the lower molecule. At the other end of the
dimer, the stacking is reversed. This allows amidine and
benzimidazole −NHs of the lower molecule to H-bond to AT
base pairs at the floor of the groove, but the imidazole −NH
and phenyl of the upper molecule are pulled into the GC region
of the sequence. The alkyl group of the upper molecule is at the
top GC base pair in the site. For the lower molecule, the
stacking is reversed. All four amidine groups in the stacked
dimer are able to H-bond to AT base edges at the floor of the
groove. The central six base pairs of the minor groove in the
AATTGCAATT sequence must be widened to accommodate
the stacked dimer. The groove then rapidly narrows to the
5295
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
Figure 6. Schematic representation of three different two-site binding modes for the compounds of Figure 1. Model 1: single-site complexes, two
molecules bind with “ideal” negative cooperativity. Model 2: two-site complexes, single molecules bind to the entire sequence. Model 3: two-site
complexes, two molecules bind with strong positive cooperativity to the entire sequence. The coloring scheme is green triangles for amidine groups,
light blue ovals for benzimidazole groups, red ovals for phenyl groups, and dark blue lines for the linker.
normal B-form width at the ends of the AT binding site and
flanking sequences.
two-site kinetoplast DNA binding sites, however, clearly reveal
that the different compounds have quite different structures in
complex with DNA.
In Figure 5C, the AATT regions of groove are narrow, as in
all known structures, while the groove widens at the GC site,
also as expected. To accommodate two stacked RT546
molecules into this asymmetric groove structure, the compound
stacking must be significantly offset. This moves the amidines
far apart and reduces electrostatic repulsion. It allows stacking
of two benzimidazole-phenyl units in the wide GC groove.
Both amidine and benzimidazole −NHs in the GC center point
toward the GN3 (amidine) or C keto (benzimidazole NH)
groups and are within H-bonding distance. The phenyl oxygens
of these stacked systems are rotated such that the methylene
and remaining ABP units are brought to the center of the
AATT sites. The amidine and benzimidazole −NHs that are
not in the GC sequence H-bond to the floor of the groove at
AN3 and T keto groups. The central four base pairs have a wide
groove to accommodate the stacked modules and the groove
then quickly narrows to contact the remaining reactions of the
compounds in the AT sequences of the offset dimer. The two
stacked dimer complexes at the AATTGCAATT sequence
were docked to provide the best interactions with the DNA
sequence. Movement of the compounds away from these
positions results in a significant loss of H-bonding and is
unlikely. To optimize H-bonding with base pair edges at the
floor of the groove, the compounds in the dimer must be
moved in specific jumps from interactions with one base pair to
the next. This restriction means that there are very few
optimized dimer positions in the two site sequences and the
two shown have the best energetics.
The simple control compound, DB184, binds relatively
weakly to all of the DNAs. It binds to one site with the AATT
DNA and to two sites with no cooperativity with the two site
DNAs by all methods, as expected (Figure 2). This compound,
thus, has essentially “ideal” independent, multisite binding
properties42 and is a useful control for comparison with the
more complex results for the other compounds (Figure 2).
With the linked compounds, a surprisingly large variation in
interactions with the model DNAs of Figure 1 is obtained. The
analysis with different DNAs makes it possible to dissect the
binding modes of the single ABP module control compound,
RT533. It binds to AATT as a monomer, about 100 times
stronger (K > 108 M−1) than DB184, suggesting that the
phenylamidine can fold back to stack on the benzimidazole-
phenyl module with favorable interactions in a single GAATTC
site. A schematic model to illustrate this binding mode is shown
in Figure 6 (Model 1). We suggest that the stacked part of the
complex will be closer to the wider GC minor groove and will
be quite dynamic. While the compound may be able to fit the
AATT site without stacking, this would require unfavorable
conformational changes in the linker, and given the strong
binding, such a complex seems unlikely. RT533 binds to all of
the two site sequences as a strong 2:1 complex indicating that
the compound can bind to individual AATT sites in the two
site sequences in much the same way as with the single AATT.
It has significant negative cooperativity in the interaction with
the closely spaced sites in AAAAATTTTT and AATTGAATT,
perhaps due to electrostatic repulsion and unfavorable steric
effects. The negative cooperativity decreases with the GC
spacer length increase in AATTGCAATT and no significant
cooperativity is observed for the DNAs with longer GC
sequences. On the whole, the compound binds similarly to
DB184 except that it has larger K values.
DISCUSSION
■
The six compounds in Figure 1 were designed to probe the
types of linked structures that can bind strongly to the minor
groove in adjacent AT sites of the kind found in the DNA of
parasitic microorganisms. The results provide important and
somewhat surprising information on how compound molecular
structure affects binding to complex, closely related DNA
sequences. The fundamental unit in the compound design is an
ABP, amidine-benzimidazole-phenyl, motif and all compounds
have at least one copy of the motif (Figure 1, Tables 1 and S1).
As expected from their shape, chemical groups, and charge, CD
studies show that all of these compounds bind to the DNA
minor groove.51 The results of biosensor-SPR and mass
spectrometry experiments with oligomer DNA models for
The symmetric compounds have two ABP modules and they
exhibit very different binding modes with the different DNA
sequences. RT546, which has an O-(CH2)3-O linker, is the
simplest of these compounds and in contrast to RT533 binds to
the single AATT DNAs in a relatively weak 1:1 complex. The
restricted conformational space of this compound does not
allow it, or any other symmetric ABP derivative, to fold into a
favorable complex with single AATT sites. RT546, however,
binds to AATTGAATT as a strong 1:1 complex, again in
5296
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
contrast to RT533 which binds as a 2:1 complex. These results
again indicate that RT546 is too large to fold back and bind to a
single AATT site, as in Figure 6, Model 1. The results indicate,
however, that it can extend its linker through the minor groove
at a GC base pair to bind effectively as a monomer to both
AATT sites in the AATTGAATT sequence, as in Model 2 of
Figure 6. The compounds were designed to bind in this
manner, ABP units at AATT and the linker at the GC base pair,
and the observed complex is quite satisfying.
To provide a molecular model to help visualize the
interactions in the 1:1 complex, RT546 was docked into the
AATTGAATT sequence (Figure 5A). The docked model for
the interaction of RT546 with the single GC sequence is similar
to classical minor groove complexes at pure AT sites. The
single GC between the narrow AATT units does not
significantly widen the minor groove but forms a binding site
for the somewhat more bulky methylene groups. The lack of
widening of the minor groove by a single GC in A-tract
sequences may be a common feature of DNA complexes. A
similar lack of groove widening by a single GC was noted with
the Fis transcription factor−DNA complex while insertion of
three GC base pairs did widen the groove with substantial loss
of Fis binding affinity.52 There are limited favorable interactions
in this GC part of the complex in Figure 5A and current design
efforts are focused on possible GC recognition units to improve
the interaction. The ABP modules fit into the AATT sites with
van der Waals, H-bonding, and water displacement contribu-
tions to binding affinity. The modules do not, however,
completely fill the length of the AATT site and it is clear that
the affinity for the AATT sequences in AATTGAATT can be
also improved.
Analysis of the binding results with AATTGCAATT
provided an unexpected surprise. The expected result was
either a 1:1 complex whose strength depended on how the
linking chain could fit into the groove at GC, or no complex if
the linker did not fit. Instead, we observed a highly cooperative
2:1 complex. Complexes at the same ratio but with decreasing
positive cooperativity were observed for the other DNAs in
Table 1. A schematic model to explain these results is shown in
Figure 6 (Model 3). The stacked dimer in Model 3 is an
entirely new binding mode for dicationic minor groove
compounds that bind in A-tract type sequences. In the
molecular docking results in Figure 5, two models for the 2:1
complex of RT546 at AATTGCAATT are described. The
results show that the minor groove with two GC base pairs
between AATT widens such that a stacked complex of RT546
is more energetically favorable that the monomer binding with
a single GC. This widening also agrees with the Fis−DNA
results for more than one GC between AT sequences.52 The
two complexes for RT546 are both able to form a significant
number of favorable interactions with base edges at the floor of
the minor groove as well as cation-π interactions in the stacked
dimer. The offset model of Figure 5C appears more favorable
than the overlapped model in Figure 5B for two primary
reasons: (i) the charged amidines are close in Figure 5B and
this will give significant electrostatic repulsion; and (ii) to
accommodate the model in Figure 5B, the AATT minor groove
must widen to bind the stacked dimer units but this sequence
strongly resists dimer complexes in other systems. Wemmer
and co-workers, for example, observed monomer binding in
AATT and only found the distamycin dimer in longer
sequences of A-tract type structures.53 Monomer binding is
observed in all AATT minor groove complexes in the protein
database. The model in Figure 5C retains a narrow groove in
the AATT sites but has the expected wider GC groove. The
models again suggest that both GC and A-site recognition can
be significantly improved. The AT recognition part of the
complex does not completely fill the site and expansion of the
ABP unit can improve these interactions. For stacked dimer
recognition of the GC spacer, we must make nonsymmetrical
molecules, as with expanded RT533 analogues that have an H
bond acceptor for interaction with the G-NH2. The other
molecule of the dimer should have a donor for interaction with
the complementary CO group. Initial examples of such
molecules are in preparation.
Another important observation with RT546 is that as the GC
linker length increases, K1 for the cooperative complex
increases while K2 decreases. As a result, the overall binding
constant (K = (K1 × K2)1/2) does not significantly change but
the cooperativity factor (CF = K2/K1 × 4) decreases. This is
probably caused by decreased stacking of the ABP modules
with decreased van der Waals interactions (K2 decrease), but
favorable, decreased amidine electrostatic repulsion (K1
increase). Adjustments to optimize binding energetics are also
possible in the alkyl linking chain.
It is somewhat surprising that in the symmetric compounds,
changing the length of the linking chains generally causes a
fairly small difference in DNA affinity (Tables 1 and S1). With
the AATTGAATT sequence and the alkyl linked compounds,
for example, the five methylene compound, DB2114, has
slightly weaker binding than the four and three methylene
derivatives (Table S1). In contrast, the more rigid five carbon
linker in DB2119, with a central phenyl, causes a 10-fold
decrease in binding affinity. This finding suggests that the
length and/or flexibility of the phenyl linker are not optimum
for binding with a single GC base pair spacer. With the
AATTGCAATT sequence, DB2119 binds much more strongly
and clearly has a better length for this sequence. It binds
similarly to the three GC spacer DNA but the binding constant
decreases marked for the longer DNAs in contrast to the
flexible linker compounds. The phenyl liker has much larger
restrictions on its two site interactions, but with appropriate
optimization, this could provide a powerful ability to target
specific DNA sites, especially since this compound does not
significantly bind to single AATT sites. The flexible compound
RT546 has much more similar affinities for all of the DNA GC
spacer lengths and can thus adapt its cooperative dimer
formation to a range of sequences. Such compounds might be
best for targeting kinetoplast DNA where binding to multiple
two-site sequences is an advantage.
There are two key points to consider in evaluation of the
interaction of these first-generation compounds with two
closely spaced AT sites in terms of designing the second
generation of linked modules for binding to two closely spaced
AT sites. First, longer flexible linkers (as in DB2114) have
weaker binding probably due to loss of conformations and
entropy on forming a complex. Such linkers should be avoided
when possible. Second, the current linkers in the ABP
compounds do not have favorable GC interactions and serve
primarily to let the two ABP modules fit into the AT sites.
Optimization of compounds with appropriate linker lengths
that have specific GC recognition motifs is, thus, a requirement
for second generation compounds. Synthesis of second
generation compounds, based on these design principles, is in
progress and should provide new agents with both increased
affinity and specificity.
5297
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
(10) Franks, A.; Tronrud, C.; Kiakos, K.; Kluza, J.; Munde, M.;
Brown, T.; Mackay, H.; Wilson, W. D.; Hochhauser, D.; Hartley, J. A.;
Lee, M. Bioorg. Med. Chem. 2010, 18, 5553.
(11) Wilson, W. D.; Tanious, F. A.; Mathis, A.; Tevis, D.; Hall, J. E.;
Boykin, D. W. Biochimie 2008, 90, 999.
(12) Brucoli, F.; Hawkins, R. M.; James, C. H.; Wells, G.; Jenkins, T.
C.; Ellis, T.; Hartley, J. A.; Howard, P. W.; Thurston, D. E. Bioorg.
Med. Chem. Lett. 2011, 21, 3780.
(13) Ryan, B.; Melander, C.; Puckett, J. W.; Son, L. S.; Wells, R. D.;
Dervan, P. B.; Gottesfeld, J.M.. Proc. Natl. Acad. Sci. U.S.A. 2006, 103,
11497.
(14) Hebert, M. D. Biochimie 2008, 90, 1131.
(15) Balasubramanian, S.; Hurley, L. H.; Neidle, S. Nat. Rev. Drug
Discovery 2011, 10, 261.
(16) Siddiqui-Jain, A.; Grand, C. L.; Bearss, D. J.; Hurley, L. H. Proc.
Natl. Acad. Sci. U.S.A. 2002, 99, 11593.
In summary, linked two-site binding compounds such as
those in Figure 1 can have three major, different types of
complexes with A-tract sequences in DNA. There are two
particularly unexpected features of the results. First is the switch
from monomer binding by the compounds with two ABP
motifs, when there is a single GC between AATT target sites, to
dimer binding with a two GC spacer in the DNA sequence.
Clearly, the widening of the groove with the two GCs allows
the compound to improve the total free energy of binding by
inserting two stacked molecules into the binding site. Second,
the AATT site has typically maintained a narrow, A-tract type
minor groove, and a cooperative minor groove dimer has not
previously been observed for this sequence. This suggests that
the wider central GC sequence may allow compound module
stacking while maintaining a relatively narrow minor groove
with only single compound modules inserted, as in Model 3 in
Figure 5C. The compounds in this research have provided
exciting new information on recognition of adjacent AT
binding sites as well as numerous suggestions for design of
second generation agents that have improved affinity and
selectivity.
(17) Collie, G. W.; Sparapani, S.; Parkinson, G. N.; Neidle, S. J. Am.
Chem. Soc. 2011, 133, 2721.
(18) Koirala, D.; Dhakal, Soma; Ashbridae, B.; Sannohe, Yuta;
Rodriguez, R.; Sugiyama, H.; Balasurbramanian, S. Nat. Chem. 2011, 3,
782.
(19) De Cian, A.; Lacroix, L.; Douarre, C.; Temime-Smaali, N.;
Trentesaux, C.; Riou, J. F.; Mergny, J. L. Biochimie 2008, 90, 131.
(20) Dervan, P. B. Science 1986, 232, 464.
(21) Roy Chowdhury, A.; Bakshi, R.; Wang, J.; Yildirir, G.; Liu, B. Y.;
Pappas-Brown, V.; Tolun, G.; Griffith, J. D.; Shapiro, T. A.; Jensen, R.
E.; Englund, P. T. PLoS Pathog. 2010, 6, e1001226.
(22) Motta, M. C. Curr. Pharm. Des. 2008, 14, 847.
ASSOCIATED CONTENT
■
S
* Supporting Information
Synthesis and characterization of symmetric compounds;
figures for comparison of thermal melting; SPR experiments
for A5T5 and AATTGCAGTC binding; the effect of GC length
change on RT546 binding; ESI−MS spectra; comparison of
CD spectra; effect of compound linker length on the DNA
recognition pattern. This material is available free of charge via
the Internet at http://pubs.acs.org.
(23) Gonzal
699.
(24) Lukes,
Englund, R. T. Eukaryotic Cell 2002, 1, 495−502.
́
ez, M.; Cerecetto, H. Expert Opin. Ther. Pat. 2011, 21,
̌
́
́
J.; Guilbride, D. L.; Votypka, J.; Zíkova, A.; Benne, R.;
(25) Paine, M. F.; Wang, M. Z.; Generaux, C. N.; Boykin, D. W.;
Wilson, W. D.; De Koning, H. P.; Olson, C. A.; Pohlig, G.; Burri, C.;
Brun, R.; Murilla, G. A.; Thuita, J. K.; Barrett, M. P.; Tidwell, R. R.
Curr. Opin. Invest. Drugs 2010, 11, 876.
(26) Le Loup, G.; Pialoux, G.; Lescure, F. X. Curr. Opin. Infect. Dis.
2011, 24, 428.
(27) Brun, R.; Don, R.; Jacobs, R. T.; Wang, M. Z.; Barrett, M. P.
Future Microbiol. 2011, 6, 677.
(28) Corson, T. W.; Aberle, N.; Crews, C. M. ACS Chem. Biol. 2008,
AUTHOR INFORMATION
■
Corresponding Author
wdw@gsu.edu
Notes
The authors declare no competing financial interest.
§On leave from Institute of Medicinal Biotechnology, Chinese
Academy of Medical Sciences and Peking Union Medical
College, Beijing 100050, China
3, 677.
(29) Bourdouxhe-Housiaux, C.; Colson, P.; Houssier, C.; Waring, M.
J.; Bailly, C. Biochemistry 1996, 35, 4251.
(30) David-Cordonnier, M. H.; Hildebrand, M. P.; Baldeyrou, B.;
Lansiaux, A.; Keuser, C.; Benzschawel, K.; Lemster, T.; Pindur, U. Eur.
J. Med. Chem. 2007, 42, 752.
(31) Zhang, R.; Wu, X.; Guziec, L. J.; Guziec, F. S.; Chee, G. L.;
Yalowich, J. C.; Hasinoff, B. B. Bioorg. Med. Chem. 2010, 18, 3974.
(32) Wakelin, L. P. G. Med. Res. Rev. 1986, 6, 275.
(33) Holman, G. G.; Zewail-Foote, M.; Smith, A. R.; Johnson, K. A.;
Iverson, B. L. Nat. Chem. 2011, 3, 875.
(34) Zolova, O. E.; Mady, A. S. A.; Garneau-Tsodikova, S.
Biopolymers 2010, 93, 777.
(35) Fox, K. R.; Gauvreau, D.; Goodwin, D. C.; Waring, M. J.
Biochem. J. 1980, 191, 729.
(36) Rahman, K. M.; James, C. H.; Thurston, D. E. Nucleic Acids Res.
ACKNOWLEDGMENTS
■
This research was supported by National Institutes of Health
grant AI064200. Biacore instruments were purchased with
partial support from the Georgia Research Alliance.
REFERENCES
■
(1) Mohammed, H. S.; Delos Santos, J. O.; Armitage, B. A. Artif.
DNA PNA XNA 2011, 2, 43.
(2) Hu, J.; Corey, D. R. Biochemistry 2007, 46, 7581.
(3) Højfeldt, J. W.; Van Dyke, A. R.; Mapp, A. K. Chem. Soc. Rev.
2011, 40, 4286.
2011, 39, 5800.
(37) Kers, I.; Dervan, P. B. Bioorg. Med. Chem. 2002, 10, 3339.
(38) Czarny, A.; Wilson, W. D.; Boykin, D. W. J. Heterocycl. Chem.
1996, 33, 1393.
(4) Tietjen, J. R.; Donato, L. J.; Bhimisaria, D.; Ansari, A. Z. Methods
Enzymol. 2011, 497, 3.
(5) Chenoweth, D. M.; Dervan, P. B. J. Am. Chem. Soc. 2010, 132,
14521.
(39) Tidwell, R. R.; Geratz, J. D.; Dann, O.; Volz, G.; Zeh, D.; Loewe,
H. J. Med. Chem. 1978, 21, 613−623.
(6) He, Y.; Hoskins, J. M.; McLeod, H. L. Trends Mol. Med. 2011, 17,
244.
(7) Kissinger, J. C. Trends Parasitol. 2006, 22, 240.
(8) Mu, J.; Seydel, K. B.; Bates, A.; Su, X. Z. Curr. Genomics 2010, 11,
279.
(40) Liu, Y.; Wilson, W. D. Methods Mol. Biol. 2010, 613, 1.
(41) Nguyen, B.; Tanious, F. A.; Wilson, W. D. Methods 2007, 42,
150.
(42) Van Holde, K. E.; Johnson, W. C.; Ho, P. S. In Principles of
Physical Biochemistry; Pearson Prentice Hall: Upper Saddle River, NJ,
2006.
(9) Purfield, A. E.; Tidwell, R. R.; Meshnick, S. R. Malar. J. 2009, 8,
104.
5298
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299

Journal of the American Chemical Society
Article
(43) Nanjunda, R.; Munde, M.; Liu, Y.; Wilson, W. D. In Methods For
Studying DNA/Drug Interactions; CRC Press-Taylor & Francis Group:
Boca Raton, FL, 2011.
(44) Rosu, F.; De Pauw, E.; Gabelica, V. Biochimie 2008, 90, 1074.
(45) Lombardo, C. M.; Martínez, I. S.; Haider, S.; Gabelica, S. V.;
Pauw, E. D.; Moses, J. E.; Neidle, S. Chem. Commun. 2010, 46, 9116.
(46) SYBYL Molecular Modeling Software, Version XI.2; Tripos, Inc.:
St. Louis, MO, 2010.
(47) Collar, C. J.; Lee, L.; Wilson, W. D. J. Chem. Inf. Model. 2010,
50, 1611.
(48) Holt, P. A.; Chaires, J. B.; Trent, J. O. J. Chem. Inf. Model. 2008,
48, 1602.
(49) Bates, P. J.; Laber, D. A.; Miller, D. M.; Thomas, S. D.; Trent, J.
O. Exp. Mol. Pathol. 2009, 86, 151.
(50) Liu, Y.; Collar, C. J.; Kumar, A.; Stephens, C. E.; Boykin, D. W.;
Wilson, W. D. J. Phys. Chem. B. 2008, 112, 11809.
(51) Ardhammar, M.; Norden, B.; Kurucsev, T. In DNA-Drug
interactions; Wiley: New York, 2000.
(52) Stella, S.; Cascio, D.; Johnson, R. C. Genes Dev. 2010, 24, 814.
(53) Wemmer, D. E.; Geierstanger, B. H.; Fagan, P. A.; Dwyer, T. J.;
Jacobsen, J. P.; Pelton, J. G.; Ball, G. E.; Leheny, A. R.; Chang, W. H.;
Bathini, Y.; Lown, J. W.; Rentzeperis, D.; Marky, L. A.; Singh, S.;
Kollman, P. In Structural Biology: The State of the Art; Adenine Press:
Albany, NY, 1993.
5299
dx.doi.org/10.1021/ja211628j | J. Am. Chem. Soc. 2012, 134, 5290−5299