Targeting the minor groove of DNA: Crystal structures of two complexes between furan derivatives of berenil and the DNA dodecamer d(CGCGAATTCGCG)2

Article abstract of DOI:10.1021/jm9604484

Crystal structures are reported of complexes of two novel furan derivatives of berenil with alkyl benzamidine groups bound to the DNA sequence d(CGCGAATTCGCG)₂. They have both been determined to 2.2 ? resolution and refined to R factors of 16.9

Full text of DOI:10.1021/jm9604484

4554
J . Med. Chem. 1996, 39, 4554-4562
Ta r getin g th e Min or Gr oove of DNA: Cr ysta l Str u ctu r es of Tw o Com p lexes
betw een F u r a n Der iva tives of Ber en il a n d th e DNA Dod eca m er
d (CGCGAATTCGCG)2
J ohn O. Trent,† George R. Clark,‡ Arvind Kumar,§ W. David Wilson,§ David W. Boykin,§ J ames Edwin Hall,|
Richard R. Tidwell,| Byron L. Blagburn,∇ and Stephen Neidle*,†
The CRC Biomolecular Structure Unit, The Institute of Cancer Research, Sutton, Surrey SM2 5NG, U.K.,
Chemistry Department, University of Auckland, Auckland, New Zealand, Department of Chemistry, Georgia State University,
Atlanta, Georgia 30303, Department of Pathology, School of Medicine, The University of North Carolina at Chapel Hill,
Chapel Hill, North Carolina 27599, and Department of Pathobiology, College of Veterinary Medicine, Auburn University,
Auburn, Alabama 36849
Received J une 21, 1996X
Crystal structures are reported of complexes of two novel furan derivatives of berenil with
alkyl benzamidine groups bound to the DNA sequence d(CGCGAATTCGCG)2. They have both
been determined to 2.2 Å resolution and refined to R factors of 16.9% and 18.6%. In both
structures the alkyl substituents, cyclopropyl and isopropyl, are found to be orientated away
from the floor of the minor groove. The drugs are located in the minor groove by two strong
amidinium hydrogen bonds, to the O2 of the thymines situated at the 5′ and 3′ ends of the
AT-rich region. The isopropyl-substituted derivative has a tight hydrogen-bonded water
network in the minor groove at one amidine site, which alters the orientation of the isopropyl
substituent. This compound has superior DNA-binding properties and activity against
Pneumocystis carinii and Cryptosporidium parvum infections in vivo compared to the
cyclopropyl derivative, which in turn is superior to the parent furan compound. We suggest
that the nature and extent of the interactions of these compounds in the DNA minor groove
play an important role in these activities, possibly in conjunction with a DNA-binding protein.
The overall effect of these alkyl benzamidine substitutions is to increase the binding of the
drugs to the minor groove.
In tr od u ction
currently used in the clinic, has toxic side effects, which
has prompted a search for related compounds with less
toxicity and greater efficacy. A correlation between
anti-PCP activity and DNA binding has been estab-
lished in several studies,10 which may be due to selective
interference with the normal processing of drug-DNA
complexes by the pathogen’s DNA topoisomerases.11
X-ray crystallographic studies on oligonucleotides com-
plexed to pentamidine and several analogues12 have
established that the drugs are minor groove binders,
confirming DNA footprinting and other biophysical
studies.13
There is now considerable evidence that the super-
family of drugs exemplified by netropsin, distamycin,
berenil, pentamidine, and Hoechst 33258 exerts their
biological effects by means of noncovalent binding to the
AT-rich regions in the minor groove of DNA.1-3 Inter-
ference with transcription control has been demon-
strated in several instances,4 especially with the TATA-
box binding proteins at the initiation of transcription.5
A number of these drugs have activity against parasitic
or fungal diseases; in these cases, inhibition of the
functioning of parasitic or fungal DNA topoisomerases
(especially topoisomerase I) may be the basis for their
selectivity.6 Antitumor effects have been found for
Hoechst 33258 itself and a number of alkylating ana-
logues,7 as well as for the alkylating distamycin ana-
logue tallimustine,8 where inhibition of basal transcrip-
tion has been proposed as the basis for its action. Other
distamycin alkylating analogues have been shown to
display increased cytotoxicity with enhanced sequence
selectivity.9
We have recently reported that several dicationic
diarylfuran compounds, which have a strong resem-
blance to pentamidine and berenil (2), have significant
activity against PCP in animal models.14 The parent
compound in this series, furamidine (3), which is
significantly more active and less toxic than pentami-
dine in the rat PCP model, also binds more strongly to
duplex DNA. We have shown15 by a combined X-ray
crystallographic, molecular modeling, and biophysical
study that this enhanced DNA affinity is in large part
due to additional nonbonded interactions between the
furan group and the walls of the narrow AT-rich minor
groove.
The bis(phenylamidinium) compound pentamidine (1)
has established activity against the Pneumocystis carinii
pathogen (PCP) which affects a high proportion of AIDS-
infected patients and is a major cause of mortality in
these individuals. This drug, which is one of the two
This study reports structural and molecular modeling
studies on two members of a further series of N-
alkylamidine furans which have the alkyl substituents
cyclopropyl (compound 4) and isopropyl (compound 5)
directly attached to the amidinium groups (Figure 1).
We examine here the effects of these substituents on
DNA binding and biological activity.
* Corresponding author. Tel: (44) 181 643 8901, ext. 4251. Fax:
(44) 181 643 1675. E-mail: steve@iris5.icr.ac.uk.
† The Institute of Cancer Research.
‡ University of Auckland.
§ Georgia State University.
| The University of North Carolina at Chapel Hill.
∇ Auburn University.
X Abstract published in Advance ACS Abstracts, October 15, 1996.
S0022-2623(96)00448-7 CCC: $12.00 © 1996 American Chemical Society

DNA Complexes of Two Bis-benzamidines
J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23 4555
treatment of C. parvum infections, we evaluated 3-5
in the neonatal mouse model. The data in Table 1
demonstrate that the three furans significantly reduce
the oocyst counts in this screen. Compound 5 is of
comparable activity to pentamidine, and 3 and 4 are
more active in this model for C. parvum infections.
Additional studies in other C. parvum models are in
progress. Collectively these results, and earlier ones,14,19
suggest that a strategy based upon design of molecules
which selectively recognize the DNA minor groove is a
promising approach for development of effective anti-
OI agents.
Cr ystallogr aph ic An alysis. DNA an d Dr u g Str u c-
t u r es. The crystal structures of the complexes of
compounds 4 and 5 with the d(CGCGAATTCGCG)2
duplex (Figure 2) are both isomorphous with other
drug-dodecanucleotide crystal structures, including
those with compounds 1-3. The DNA is in the right-
handed B form. Both 4 and 5 fit well into the minor
grooves at the center of the AT-rich regions in each
duplex. In each structure, in principle, there are two
possible orientations for the alkyl-substituted amidine
groups: one with the alkyl group located at the floor of
the groove and the other with it directed to the mouth
of the groove. It is notable that despite the alkyl
substitutions, both molecules fit deeply into the minor
groove, with the amidine alkyl groups oriented away
from the floor of the minor groove, as shown by their
unambiguous fit to the electron density in both cases
(Figure 3). The fit of the molecules to the density shows
that the diphenylfuran moieties are highly coplanar,
appreciably more so than they are in the dodecanucle-
otide complex of molecule 3. This is shown by the
average furan-phenyl dihedral angles (O-CA-C1-C6
and O-CA′-C1′-C6′), being 1°, 1°, and 10°, for 4, 5,
and 3, respectively. As has been observed previously
with compounds containing the same core aryl rings,
for example, the d(CGCGAATTCGCG)2:3 complex, both
4 and 5 form strong hydrogen bonds from N1 to O2 of
T20 and from N1′ to O2 of T8. We would expect the
overall effect of 4 and 5, being more planar than 3, to
be a narrowing of the minor groove in the center of
the AT region of their complexes, compared with the
corresponding complex with 3. The alkyl groups should
not widen the groove significantly since they could
form favorable hydrophobic interactions with the walls
of the groove. There are two possible orientations that
the alkyl groups can possess with respect to the minor
groove axis direction, one parallel and one perpendicu-
lar. The minor groove can accommodate alkyl groups
of this size (and larger ones) if they are orientated
parallel to the groove, having minimal unfavorable
interactions with the wall of the groove or with the alkyl
groups pointing out perpendicular to the groove. In the
case of the complex with 5, only the perpendicular
orientation is observed for the cyclopropyl group since
it is more rigid and spherical in cross section and cannot
have as narrow a profile as the isopropyl group of 4.
However, both orientations are observed for the isopro-
pyl group in the complex with 4, as seen in Figure 1. It
is apparent that a water hydration network is respon-
sible for the perpendicular orientation, in this case due
to hydrogen bonding between the water molecules and
amidinium groups (see below).
F igu r e 1. Structures of 1-5 with the numbering scheme of
4.
Resu lts a n d Discu ssion
Biologica l Resu lts. Table 1 contains the results
from a study of the interaction of 3-5 with DNA and
from evaluation of these compounds in animal models
for P. carinii and Cryptosporidium parvum infections.
The binding to poly(dA‚dT) was studied, and their DNA
complexes melt at >90 °C. In order to rank the relative
binding affinities for interactions with DNA, the duplex
oligomer d(GCGCAATTGCGC)2 (A2T2) was employed.
The melting temperature range for this oligomer allows
measurement of ∆Tm values for these compounds, even
though they exhibit strong binding affinities. The
increases in melting temperature on complex formation
with the dicationic molecules are related to their binding
affinities with nucleic acids.16 It is notable that intro-
duction of an alkyl group on the amidine nitrogen
results in an increase in the ∆Tm values; compare the
results for 3 with that for 4 and 5. We have previously
suggested that increases in ∆Tm values are due to
involvement of increased van der Waals interactions.17
The in vivo activity for the furan dications against P.
carinii roughly parallels the DNA affinity of these
compounds. The furans 3-5 are significantly more
active and less toxic than pentamidine in the rat model.
These results suggest that other alkylfurans should be
investigated; such studies are underway. Blagburn
and co-workers18 noted that analogues of pentamidine
showed activity in a neonatal mouse model for C.
parvum which employs an oral dosage regimen. In view
of the fact that no effective drug is available for

4556 J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23
Trent et al.
Ta ble 1. In Vivo Activity of Dicationic Diarylfurans against P. carinii and P. parvum
P. carinii
C. parvum
∆Tm
dosageb
(µmol/kg/day)
cyst/g of lungd
(% of control)
dosagee
(µmol/kg/day)
oocystsf
(% of control)
a
compd
saline
d(GCGCAATTGCGC)2
toxicityc
0
2
0
0
1
100.0
1.47
0.71
0.24
0.02
pentamidine
4.8
11.7
14.4
12.4
22.1
13.3
10.8
10.9
62
20.8
15
3
3
4
5
37.9
36.9
72.2
17
a
b
Increase in thermal melting of the oligomer d(GCGCAATTGCGC)2, see ref 4. Intravenous dosage. c See ref 5 for detailed explanation
of toxicity scale; generally, the larger the value the more severe the toxicity, and values greater than 2 indicate death of some animals.
d
Counted cysts in lung tissue; using a blinded protocol reported as percentage of saline-treated controls, see ref 5. e Oral dosage; no
adverse reactions resulting from treatment with the test compounds were observed. f Counted oocysts obtained from alimentary tract
tissue; using a blinded protocol reported as percentage of controls, see ref 3.
F igu r e 2. Structures of the complexes of (left) 4 and (right) 5 with d(CGCGAATTCGCG)2. Water molecules have been removed
for clarity.
Although the three core aryl rings in both complexes
are nearly planar, the amidinium groups themselves are
twisted significantly out from the aryl ring planarity.
This is to be expected and is consistent with molecular
mechanics calculations showing that the optimal in
vacuo deviation between the phenyl and alkyl-substi-
tuted amidine groups is ca. 42° but close to 0° for
unsubstituted amidines (a search of the small-molecule
Cambridge Crystallographic Database for compounds
with similar functionality confirms these orientations).
Since the crystal structures of these complexes are not
symmetric (the duplex plus drug molecule being the
asymmetric unit), the amidinium groups tend to have
varying conformations as shown by the N1-C7-C4-
C3 and N1′-C7′-C4′-C3′ dihedral angles: A2T2:4, 18°,
40°; A2T2:5, 26°, 30°; and A2T2:3, 20°, 24°. These
values are altered from those expected for the drug
alone due to the strong thymine-drug hydrogen bond-
ing overcoming low-energy barriers for dihedral angle
(O-CA-C1-C6 and O-CA′-C1′-C6′) rotations. In
each case the three aryl-linked rings are forced near to
coplanarity in order to maintain the optimum orienta-
tion and position of the amidine ends. Despite these
differing dihedral angles, the amidine nitrogen atoms
are in nearly the same position in all three complexes
(Figure 4), with about the same overall twist between
the two amidine ends of each molecule in the complexes.
The N1-C7-C7′-N1′ torsion angles are for 4, 51°; 5,
57°; and 3, 60°; the intramolecular drug N1-N1′
distances are for 4,12.5 Å; 5,12.5 Å; and 3,12.8 Å.
The locations of the drugs in the three furan com-
plexes are almost identical with respect to the A2T2
region of the minor groove. This is not surprising since
the two N1 and N1′ hydrogen bonds (to O2 of T8 and
T20) fix the location of the amidine groups. It is notable
that the minor groove can tolerate the binding of ligands
with extended regions of planar structure; the three aryl
rings of 4 and 5 have little specific interactions with
the groove other than van der Waals contacts.
DNA Str u ctu r e. The shape of the groove in the
structures of the A2T2 complexes of 4 and 5 appears to
have changed subtly from that in the complex with 3.
The duplexes in the first two complexes have a narrower
groove floor, while the mouth of the groove stays

DNA Complexes of Two Bis-benzamidines
J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23 4557
F igu r e 3. Fo - Fc electron density maps (drug atoms have been omitted from the calculation) contoured at the 1.0 s level showing
the “fit” of 4 and 5 to the density: (top left) side view of 4, (bottom left) top view of 4, (top right) side view of 5, and (bottom right)
top view of 5.
(Table 2). The flexibility of the base pair movement and
the rotation of the amidine group mean that these
nonbonding arrangements are variable, but there is a
preference for a close contact to the sugar ring oxygen
atoms C21 O4′ or C9 O4′.
The isopropyl and cyclopropyl groups have few close
contacts (Table 2) with the minor groove when in the
perpendicular orientation. The perpendicular oriented
isopropyl carbon atoms C9 and C10 of the complex
A2T2:4 generally weakly interact with the phosphate
backbone. The parallel oriented isopropyl group has
completely different interactions: the C9′ atom situated
in the minor groove interacts with the wall of the groove,
while the C10′ atom, situated in the mouth of the minor
groove, has no contacts under 4 Å. The cyclopropyl C9′
and C10′ atoms of the complex A2T2:5 have weak
interactions with the phosphate backbone that are
similar to the perpendicular oriented isopropyl group
in the A2T2:4 complex; however, the cyclopropyl C9 and
C10 atoms have only one interaction with the phosphate
backbone, despite being in the same orientation.
The two alkyl-substituted amidine structures are very
similar to each other, with a rms difference on super-
position of all the DNA atoms of just 0.39 Å (Figure 5).
The least variable region is in the AATT sequence where
the drugs 4 and 5 effectively hold it relatively rigid. A
comparison of helical parameters (calculated by the
program CURVES20) for the A2T2 complexes of 3-5
with the native dodecamer21 (Nucleic Acid Data Base22
code bdl001) shows that in terms of propeller twist the
two new structures with 4 and 5 are closer to the native
than to the A2T2:3 complex (Table 3). This is somewhat
surprising due to the structural similarity of 4 and 5 to
3, although it appears that the large twists between the
phenyl ring and the alkyl-substituted amidine groups
in 4 and 5 follow the isohelicity of the minor groove to
a greater extent and with less unfavorable interactions
than 3, thus not causing a significant deformation to
F igu r e 4. Superposition of bound drugs 3-5 showing phenyl
ring rotation and amidine rotation and position. Atoms used
in the superposition were C7, C4, C1, CA, CA′, C1′, C4′, and
C7′.
approximately at the same width. The effect of this can
be seen in the close contacts of the drugs and DNA
(Table 2). The close contacts (<3.5 Å) for 4 and 5 are
with the walls of the groove but with significantly fewer
interactions than in the A2T2:3 complex. The three aryl
rings of 3 have a noncoplanar arrangement which can
more effectively contact the groove walls.
The N1 and N1′ atoms of the drugs are involved in
several nonbonded interactions in addition to the hy-
drogen bonds to O2 of T8 and T20; these are each with
three other charge-complementary atoms: A6 O4′, A6
N3, C21 O4′; and A18 O4′, A18 N3, C9 O4′, respectively

4558 J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23
Ta ble 2. Drug-DNA Close Contacts (<3.5 Å) of 3-5 in
Trent et al.
a
Complexes with d(CGCGAATTCGCG)2
drug atom
contact base
contact atom
distance (Å)
A2T2:4
N1
N1
N1
N1
C5
C5
C6′
CA′
C5′
C5′
C5′
C5′
N1′
N1′
N1′
N1′
C9
A6
A6
O4′
N3
O2
O4′
O2
C2
O2
C4′
O4′
N3
C2
3.9
3.8
2.9b
3.3
2.9
3.3
3.3
3.4
2.9
3.2
3.4
3.1
3.8
3.4
3.2b
3.7
3.3
3.9
4.0
3.8
4.0
3.6
3.6
3.8
3.9
T20
C21
T20
A5
T19
T7
T19
A18
A18
T8
A18
A18
T8
C9
A6
T7
T7
C21
G22
A17
A17
A17
A18
O2
O4′
N3
O2
O4′
O3′
O2P
C5′
O3′
C5′
C4′
O3′
O4′
C5′
C9
C9
C10
C10
C9′
C9′
C9′
C9′
A2T2:5
N1
N1
N1
N1
C3
C5
CA
CA′
C6′
C5′
C5′
C5′
C5′
N1′
N1′
N1′
N1′
C9
A6
A6
O4′
N3
O2
O4′
C5′
O2
C5′
C5′
O2
O4′
N3
C2
3.8
3.3
3.0b
3.8
3.3
3.0
3.4
3.4
3.4
2.9
2.8
3.1
2.9
4.0
3.5
2.9b
3.0
3.6
4.0
3.7
3.6
3.7
T20
C21
C21
T20
T20
T8
T19
T19
A18
A18
T8
A18
A18
T8
C9
A6
F igu r e 5. Superposition of all DNA atoms of the complexes
A2T2:4 and A2T2:5. The total rms difference is 0.39 Å.
Ta ble 3. Propeller Twists (Calculated using CURVES) for
Complexes of 3-5 with d(CGCGAATTCGCG)2 and for the
Native Dodecamer21
base pair
A2T2:4
A2T2:5
A2T2:3
native
C1-G24
G2-C23
C3-G22
G4-C21
A5-T20
A6-T19
T7-A18
T8-A17
C9-G16
G10-C15
C11-G14
G12-C13
-8
-8
-13
-12
-12
-16
-21
-24
-23
-14
-11
-8
3
-8
-17
-13
-4
O2
-12
-15
-17
-28
-18
-18
-15
-8
0
O4′
N3
O2
O4′
O3′
O3′
O5′
O3′
C4′
-10
-10
-14
-11
-16
-4
5
-18
23
-17
-27
-27
-24
-28
-25
-9
C9′
C9′
C10′
C10′
C9
G10
A18
A18
-16
2
-25
8
-27
-5
A2T2:3
average
-13
-14
-5
-18
N1
N1
N1
N1
C6
C6
C6
C5
C5
CA′
C1′
C6′
C6′
C5′
C5′
C4′
C7′
N1′
N1′
N1′
N1′
A6
A6
T20
C21
T20
T7
A5
T7
T20
T8
T8
O4′
N3
O2
O4′
O2
O4′
N3
O4′
O4′
C5′
O4′
O4′
O4′
O4′
O2
O4′
O4′
O4′
N3
3.5
3.4
3.2b
3.5
3.2
3.3
3.3
3.4
3.3
3.4
3.3
3.0
3.3
3.0
3.0
3.2
3.2
3.9
3.8
3.1b
3.0
the DNA. It is difficult to comparatively analyze groove
widths for all three as width is dependent on the
defining atom chosen. In spite of this complication,
there are several clear trends: (i) all three structures
have a wider groove width in the AT-rich region than
the native dodecamer (Figure 6), (ii) the isopropyl
complex produces the greatest amount of groove widen-
ing, and (iii) the cyclopropyl complex produces the least.
These effects are most apparant towards the 3′ end of
the AT region. It is possible to ascribe the differences
produced by the drugs to their dihedral twists and the
size of the substituent. The differences in groove
widening produced by the isopropyl and cyclopropyl
groups parallel their differences in steric volume, of 303
and 282 Å3, respectively.
T8
T19
T19
T8
T19
C9
A18
A18
T8
O2
O4′
The average thermal parameters for the complexes
of 3-5 with A2T2 (Table 4) show that the DNA follows
the expected trend of phosphates having more motion
than sugars, which in turn are more mobile than bases.
The thermal motions of the 4 and 5 molecules in their
C9
a
The longer contacts of interest (e4.0 Å) have been included
for the internal amidine nitrogen atoms N1 and N1′ and the alkyl
substituent carbon atoms C9, C10, C9′, and C10′. Indicates the
b
thymine hydrogen bonds at the 5′ and 3′ ends.

DNA Complexes of Two Bis-benzamidines
J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23 4559
F igu r e 7. Schematic representation of the A2T2:4 complex
showing the drug hydrogen bonds and the associated minor
groove water structure. The numbers represent the oxygen
atoms of the water molecules. Distances are in angstroms.
hence reduced overall binding. Ratios of the average
temperature factors for the drug to DNA for 3-5
complexes are 1.4, 1.0, and 1.1, respectively.
Hyd r a tion of A2T2:4 a n d A2T2:5. There is exten-
sive hydration of the DNA backbones and drug amidine
groups in both structures. The amidine nitrogen atoms
N2 and N2′ along the outer spine of the drug are
hydrogen-bonded to water molecules as are the amidine
N1 and N1′ atoms in the minor groove in the A2T2
complex with 4. There is a tight hydrogen-bonded
three-water network at the G4 end (N1 and N2 of 4) in
the minor groove that prevents the isopropyl group from
being parallel to the groove. These water molecules
interact with the A5 and G22 sugar O4′ and the G4,
A5, C21, and G22 bases (Figure 7). At the C9 end (N1′
and N2′ of 4) one water molecule interacts with the C9
and A17 bases with no further water network being
observed in the electron density maps.
The same water network is not observed in the
A2T2:5 complex, with only the amidine nitrogen atoms
N1, N2, and N1′ being hydrated by single waters, which
are not involved in further hydrogen-bonded networks.
Com p a r ison w it h St r u ct u r es fr om Molecu la r
Mod elin g. A molecular modeling study was performed
prior to the crystallographic analysis to investigate the
effect of alkyl substitution on the binding and to identify
possible orientations for the drug in the minor groove.
F igu r e 6. Minor groove widths (calculated using CURVES)
for the 3-5 complexes with (CGCGAATTCGCG)2 and the
native dodecamer: (top) using P atoms to measure width and
(bottom) using C1′ atoms to measure width; (0) A2T2:4, (])
A2T2:5, (O) A2T2:3, and (4) native A2T2.
Ta ble 4. Average Temperature Factors for A2T2:4, A2T2:5,
and A2T2:3 (in Å2)
group
A2T2:4
A2T2:5
A2T2:3
phosphates
sugars
bases
drug
water
37.3
30.1
21.6
28.7
41.5
49.2
39.5
29.7
43.8
55.9
33.0
27.6
21.9
37.6
39.9
complexes are comparable to the average motion of the
DNA (calculated by averaging the phosphate, sugar, and
base temperature factors), while 3 has a higher motion,
indicating a less complementary fit to the DNA and

4560 J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23
Trent et al.
(University of Edinburgh) and annealed before use. The drugs
4 and 5 were prepared as their hydrochloric salts.
Low-energy conformations of the separate molecules,
obtained by a torsion angle (C5-C4-C7-N2, C4-C7-
N2-C8, C7-N2-C8-C9, C5′-C4′-C7′-N2′, C4′-C7′-
N2′-C8′, C7′-N2′-C8′-C9′) systematic search proce-
dure, were placed in an ideal B-DNA molecule (generated
by Macromodel 5.0) in a similar position to 3 in its A2T2
complex. A full continuum-solvated torsion search of
the same angles, with minimization of the duplex-drug
complex at each point, gave the global minima. This
was checked by taking the two global minima for
complexes of 4 and 5 and subjecting them to a standard
mechanics/dynamics/mechanics protocol which yielded
the drugs in the same orientation as 3. The two global
minima have the alkyl groups in the parallel position,
which correctly predicts the A2T2:5 complex structure
but only part of the A2T2:4 structure. The GB/SA
continuum model23 uses implicit waters, so it is not
surprising that it did not correctly identify the unusual
perpendicular orientation of the isopropyl group, which
is tightly held in position by the network of waters in
the minor groove.
Im p lica tion s for Liga n d Design . The three dia-
rylfuran compounds 3-5 show a graduation in DNA
binding abilities that parallel their in vivo activities
against both P. carinii and C. parvum (Table 1). The
isopropyl analogue 4 is consistently the superior com-
pound in the series providing credence to the hypothesis
that the minor groove drugs act by directly interfering
with DNA function and processing.4 The observation
from the three X-ray crystallographic structures is that
the principal structural differences between the DNA
complexes of 3-5 are the extent of minor groove
deformation and nonbonded drug-DNA contacts, with
compound 4 producing the largest structural changes
compared to the native DNA dodecanucleotide struc-
ture. This in turn suggests that it may primarily be
the involvement of the widened DNA minor groove in
ternary complexes with a regulatory or processing
protein and drug that ultimately results in the thera-
peutic responses detailed here. The superior DNA
binding of 4 is due to the larger size of the isopropyl
group and consequently its greater surface area and
larger number of nonbonded contacts with the minor
groove surface. The involvement of a network of tightly
bound minor groove water molecules in the binding of
one end of the drug may also play a role in enhancing
the binding.
The complexes were grown from hanging drops at 286 K as
colorless prismatic blocks. The crystal of the A2T2:4 complex
used for data collection was grown from a drop containing 1
µL of 50% (w/v) 2-methylpentane-2,4-diol, 2 µL of 5 mM 4, 3
µL of 15 mM MgCl2, 1 µL of 2.5 mM spermine, and 3 µL of 2
mM A2T2 equilibrated against a reservoir containing 1 mL of
55% (w/v) 2-methylpentane-2,4-diol. The crystal of the A2T2:5
complex used for data collection was grown from a drop
containing 3 µL of 30% (w/v) 2-methylpentane-2,4-diol, 3 µL
of 5 mM 5, 3 µL of 50 mM MgCl2, 3 µL of 2.5 mM spermine,
and 3 µL of 2 mM A2T2 equilibrated against a reservoir
containing 1 mL of 50% 2-methylpentane-2,4-diol. The A2T2
solution was prepared using 30 mM sodium cacodylate buffer
at pH 7.0. The crystals employed for the X-ray studies were
obtained after about 3 months of growth.
Cr ysta llogr a p h ic Da ta Collection . The colorless crystals
of A2T2 complexes of 4 and 5 used for data collection were of
approximate dimensions 0.18 × 0.18 × 0.12 and 0.60 × 0.28
× 0.26 mm, respectively, and were mounted in 0.5 mm
Lindemann glass capillaries with small amounts of mother
liquor. Intensity data were collected at 287 K using a Siemens-
Xentronics multiwire area detector with a rotating anode X-ray
generator and a graphite monochromator. A crystal-to-detec-
tor distance of 10 cm and a swing angle of 15° were used to
collect data to a maximum possible resolution of 2.2 Å. Data
were collected with ø set at 45°, while the crystal was rotated
through 100° in ω at φ values of 0° and 60°; 180 s frames were
recorded every 0.20° step. The crystals did not suffer any
observable decay during the data collection. Data processing
was carried out using the program package XENGEN, version
1.3. After merging, the data for the complex with 4 comprised
3146 of a possible 3592 unique reflections to 2.2 Å (87.6%) with
a merging R value of 7.3%, and the data for 5 comprised 3427
of a possible 3748 unique reflections to 2.2 Å (91.4%) with a
merging R value of 8.8%.
Str u ctu r e Refin em en t. The unit cell dimensions of the
A2T2 complexes of 4 and 5 were a ) 24.60 Å, b ) 40.07 Å, c
) 65.45 Å, and a ) 25.43 Å, b ) 40.66 Å, c ) 66.13 Å,
respectively, in the space group P212121. These cells are close
to that reported for the native d(CGCGAATTCGCG)2 dodecam-
er (Dickerson and Drew, 1981; a ) 24.87 Å, b ) 40.39 Å, c )
66.2 Å) and other groove-bound dodecamer-drug complexes,
suggesting that the present crystals are isomorphous with
them. The dodecamer coordinates used as starting models for
the structure refinements were those of the native d(CGC-
GAATTCGCG)2 structure.21
The crystallographic refinement was carried out using the
program X-PLOR,24 version 3.1. Rigid body refinement of the
DNA molecule as one constrained group was performed with
the resolution range of the data increased from 8.0-4.0 Å (522
and 566 reflections) to 8.0-3.5 Å (804 and 874 reflections) for
data for A2T2:4 and A2T2:5, respectively. The resolution
range was gradually increased from 8.0-3.5 to 8.0-2.2 Å (2591
and 2817 reflections), the R factors being 26.6% and 25.1%
for A2T2:4 and A2T2:5, respectively. Electron density maps
were calculated and displayed using the graphics package
TOM/FRODO,25 version 3.2.
The two structures presented in this paper reveal
three features significant for the future design of
ligands: (i) the minor groove can tolerate a planar
orientation for the linked aromatic moieties of a drug if
they are energetically favored, i.e., the drug molecule
itself may not be in a global minimum but the increase
in drug-DNA binding energy can compensate for this;
(ii) the orientation of the amidine alkyl substituents
places them in a specific location which may be used as
the starting point for the design of other substituents
that increase binding and extend base pair selectivity;
and (iii) the water network in the A2T2:4 complex may
be exploited to devise new ligands that employ similar
interactions to increase binding and base pair selectiv-
ity.
The DNA molecules fitted the density well, and a lobe of
density was clearly visible in the minor groove in each
structure. The lobes of density were in the positions predicted
for both 4 and 5 from molecular modeling using Macromodel,26
version 4.5. In the case of the structure of A2T2:4, there was
a second small lobe of density in the minor groove close to the
large lobe, and all possible orientations of the alkyl amidine
group were fitted to it. Electrostatic charges for 4 and 5
were calculated by MOPAC,27 version 6.0, using the electro-
static potential option. Force field parameters were interpo-
lated from previous molecular modeling studies in this labo-
ratory.
The drug molecules were included, and the refinement was
repeated stepwise from low- to high-resolution data. The R
factor at this point for the structure A2T2:5 was 22.7%. It
was possible to unambiguously determine the orientation of
the alkyl amidine group for structure A2T2:4 by the difference
Exp er im en ta l Section
Cr ysta llogr a p h y. The DNA dodecamer d(CGCGAAT-
TCGCG)2 (A2T2) was purchased from Oswell DNA Service

DNA Complexes of Two Bis-benzamidines
J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23 4561
in R factors, 24.2% for the alkyl group oriented deep in the
minor groove and 23.8% for the alkyl group in an opposing
orientation, at the mouth of the groove. The later was used
for further refinement.
Solvent positions were included toward the end of the
refinement and assigned as water molecules. The criteria for
acceptance of solvent were proximity to the DNA-drug
complex (within 4 Å), peak height > 3σ in difference maps,
potential hydrogen-bonded neighbors (2.2-3.4 Å), and sensible
thermal parameters.
Hz), 7.95 (d, 4H, J ) 8.54 Hz), 7.42 (s, 2H), 2.87 (brm,
2H), 1.09-0.85 (m, 8H); 13C NMR (DMSO-d6) δ 163.9,
152.3, 133.7, 129.1, 126.6, 123.5, 111.3, 24.7, 6.5. Anal. Calcd
(C24H24N4O‚2HCl): C,63.02; H, 5.73; N, 12.25. Found: C,
62.89; H, 5.95; N, 12.00.
Ack n ow led gm en t. This work was supported by the
Cancer Research Campaign Programme Grant SP1384
(to S.N.) and by NIH Grant AI-33363 (to D.W.B.).
At the end of the refinements a total of 81 and 51 water
molecules had been included and gave final R factors of 16.9%
and 18.6% for structures A2T2:2 and A2T2:3, respectively, for
all data in the range 8.0-2.2 Å. Coordinates and structure
factors have been submitted to the Nucleic Acid Data
Base (identification numbers GDL044 and GDL045).
Molecu la r Mod elin g. The conformation spaces of 4 and
5 were searched extensively by rotating torsion angles OA-
CA-C1-C6, C5-C4-C7-N1, and C7-N1-C8-C9 in 5°
increments through 360°. Macromodel, v4.5, running on an
SGI Challenge Workstation, using the modified AMBER* force
field with a distance-dependent dielectric constant of 4, gave
very similar global minima structures for the two drugs in the
minor groove. The AMBER* structures of 4 and 5 were placed
in a generated ideal B-form d(CGCGAATTCGCG)2 duplex
using the position suggested by the crystal structure of A2T2:
3. A torsion angle search and minimization using AMBER*,
similar to the previous study on the drugs alone but eliminat-
ing the bad-contact structures, was used to obtain the starting
structure for molecular dynamics calculations. A 40 ps mo-
lecular dynamics run at 300 K with a 1.5 fs time step was
sampled 100 times, and the average structure was further
minimized using a standard protocol.
Refer en ces
(1) Brown, D. G.; Sanderson, M. R.; Garman, E.; Neidle, S. Crystal
structure of a berenil-d(CGCGAAATTTCGCG)2 complex: An
example of drug-DNA recognition based on sequence-dependent
structural features. J . Mol. Biol. 1992, 226, 481-490.
(2) Nunn, C. M.; J enkins, T. C.; Neidle, S. Crystal structure of
d(CGCGAATTCGCG)2 complexed with propamidine, a short-
chain homologue of the drug pentamidine. Biochemistry 1993,
32, 13838-13843.
(3) Edwards, K. J .; J enkins, T. C.; Neidle, S. Crystal structure of a
pentamidine-oligonucleotide complex: Implications for DNA-
binding properties. Biochemistry 1992, 31, 7104-7109.
(4) (a) Welch, J . J .; Rauscher, F. J ., III; Beerman, T. A. Targeting
DNA-binding drugs to sequence-specific transciption factor DNA
complexes. J . Biol. Chem. 1994, 269, 31051-31058. (b) Wong,
S. S. C.; Sturm, R. A.; Michel, J .; Zhang, X.-M.; Danoy, P. A. C.;
Mcgregor, K.; J acobs, J . J .; Kaushal, A.; Dong, Y.; Dunn, I. S.;
Parsons, P. G. Transcriptional regulation of differentiation,
selective toxicity and ATGCAAAT binding of bisbenzimidazole
derivatives in human melanoma cells. Biochem. Pharmacol.
1994, 47 (5), 827-837. (c) Henderson, D.; Hurley, L. H. Molec-
ular struggle for transcription control. Nature Med. 1995, 1,
525-527. (d) Mote, J ., J r.; Ghanouni, P.; Reines, D. A minor
groove-binding ligand both potentiates and arrests transcription
by RNA polymerase II. J . Mol. Biol. 1994, 236, 725-737.
(5) Chiang, S.-Y.; Welch, J .; Rauscher, F. J ., III; Beerman, T. A.
Effects of minor groove binding drugs on the interaction of TATA
box binding protein and TFIIA with DNA. Biochemisty 1994,
33, 7033-7040.
(6) (a) Chen, A. Y.; Yu, C.; Gatto, B.; Liu, L. F. DNA minor groove-
binding ligands: A different class of mammalian DNA topoi-
somerase I inhibitors. Proc. Natl. Acad. Sci. U.S.A. 1993, 90,
8131-8135. (b) Beerman, T. A.; McHugh, M. M.; Sigmund, R.;
Lown, J . W.; Rao, K. E.; Bathini, Y. Effects of analogues of the
DNA minor groove binder Hoechst 33258 on topoisomerase II
and I mediated activities. Biochim. Biophys. Acta 1992, 1131,
53-61.
(7) (a) Sun, Q.; Gatto, B.; Yu, C.; Liu, A.; Liu, L. F.; LaVoie, E. J .
Structure activity of topoisomerase I poisons related to Hoechst
33342. Bioorg. Med. Chem. Lett. 1994, 4, 2871-2876. (b) Atwell,
G. J .; Yaghi, B. M.; Turner, P. R.; Boyd, M.; O’Conner, C. J .;
Ferguson, L. R.; Baguley, B. C.; Denny, W. A. Synthesis, DNA
interactions and biological activity of DNA minor groove targeted
polybenzamide-linked nitrogen mustards. Bioorg. Med. Chem.
1995, 3, 679-691. (c) Gravatt, G. L.; Baguley, B. C.; Wilson, W.
R.; Denny, W. A. DNA-directed alkylating agents. 6. Synthesis
and antitumor activity of DNA minor groove-targeted aniline
mustard analogues of Pibenzimol (Hoechst 33258). J . Med.
Chem. 1994, 37, 4338-4345.
(8) (a) D’Incalci, M. DNA-minor-groove alkylators, a new class of
anticancer agents. Ann. Oncol. 1994, 5, 877-878. (b) Bellorini,
M.; Moncollin, V.; D’Incalci, M.; Mongelli, N.; Mantovani, R.
Distamycin A and tallimustine inhibit TBP binding and basal
in vitro transcription. Nucleic Acids Res. 1995, 23, 1657-1663.
(9) Wyatt, M. D.; Lee, M.; Garbiras, B. J .; Souhami, R. L.; Hartley,
J . A. Sequence specificity of alkylation for a series of nitrogen
mustard-containing analogues of distamycin of increasing bind-
ing site size: evidence for increased cytotoxicity with enhanced
sequence specificity. Biochemistry 1995, 34, 13034-13041.
(10) (a) Tidwell, R. R.; J ones, S. K.; Geratz, J . D.; Ohemeng, K. A.;
Cory, M.; Hall, J . E. Analogues of 1,5-bis(4-amidinophenoxy)-
pentane (pentamidine) in the treatment of experimental Pneu-
mocystis carinii pneumonia. J . Med. Chem. 1990, 33, 1252-
1257. (b) Cory, M.; Tidwell, R. R.; Fairley, T. A. Structure and
DNA binding activity of analogues of 1,5-bis(4-amidinophenoxy)-
pentane (pentamidine). J . Med. Chem. 1992, 35, 431-438. (c)
Donker, I. O.; Tidwell, R. R.; J ones, S. K. Pentamidine congeners.
2. 2-Butene-bridged aromatic diamidines and diimidazolines as
potential anti- Pneumocystis carinii pneumonia agents. J . Med.
Chem. 1994, 37, 4554-4557.
Syn th esis. 2,5-Bis{[4-(N-isop r op yl)a m in d in o]p h en yl}-
fu r a n Dih yd r och lor id e (4). Dry isopropylamine (0.47 g,
0.008 mol) was added to a suspension of imidate ester (1.3 g,
0.003 mol), prepared as previously described20 from 2,5-bis(4-
cyanophenyl)furan,28 in 45 mL of absolute ethanol. Within 0.5
h the imidate ester dissolved. After ca. 3 h a white solid
precipitated; the slurry was stirred overnight at room tem-
perature. The solvent was removed under reduced pressure,
diluted with water, filtered, and washed with water. After
the solid was dried, it was recrystallized from an ethanol/ether
mixture to yield a white solid (0.9 g, 78%): mp 233-4 °C; IR-
(KBr) 3249, 3069, 2997, 1600, 1383, 1210 cm-1 lH NMR
;
(DMSO-d6/60 °C) δ 7.79 (brs, 8H), 7.11 (s, 2H), 6.25 (br, 4H),
3.81 (br, 2H), 1.14 (d, 6H, J ) 5.9 Hz); 13C NMR (DMSO-d6/60
°C) δ 159.6, 152.4, 136.6, 130.4, 126.8, 122.8, 108.7, 43.5, 22.8;
MS m/e 388(M+).
The free base (0.78 g, 0.002 mol) was dissolved in 10 mL of
absolute ethanol, treated with 10 mL of ethanol saturated with
hydrogen chloride, and warmed for 2 h. The mixture was
reduced in volume to 5 mL. Addition of 20 mL of dry ether
produced a bright yellow precipitate which was filtered,
washed with 3 × 5 mL of dry ether, and dried in vacuo at 65
°C for 2 h to yield 0.8 g (87%): mp 276-7 °C dec; IR (KBr)
3407, 3132, 3065, 1669, 1611, 1388, 1128 cm-1 lH NMR
;
(DMSO-d6) δ 9.72 (s, lH), 9.69 (s, lH), 9.57 (s, 2H), 9.24 (s,
2H), 8.06 (d, 4H, J ) 8.1 Hz), 7.86 (d, 4H, J ) 8.1 Hz), 7.47 (s,
2H), 4.14 (sep, 2H, J ) 6.6 Hz), 1.29 (d, 12H, J ) 6.6 Hz); 13
C
NMR (DMSO-d6) δ 161.1, 152.3, 133.6, 129.2, 127.7, 123.5,
111.3, 45.1, 2l.l. Anal. Calcd (C24H28N4O‚2HCl‚1.25H2O): C,
59.57; H, 6.76; N, 11.57. Found: C, 60.00; H, 6.80; N, 11.52.
2,5-Bis{[4-(N-cyclop r op yla m id in o)p h en yl]}fu r a n Di-
h yd r och lor id e (5). The free base was prepared from cyclo-
propylamine and the imidate ester as described for 4 in a yield
of 79%: mp 185-186 °C dec; IR (KBr) 3464, 3320, 3080, 1610,
1510, 1364, 1022, 848, 791 cm-1; lH NMR (CDCl3) δ 7.71 (brs,
8H), 6.78 (s, 2H), 5.3 (v br, 4H), 2.6 (brm, 2H), 0.87-0.81 (m,
4H), 0.67-0.62 (m, 4H); 13C NMR (CDCl3 + DMSO-d6) δ 159.6,
152.2, 134.8, 130.7, 126.4, 122.6, 107.7, 25.7, 6.04; MS m/e 384
(M+).
(11) Tidwell, R. R.; J ones, S. K.; Naiman, N. A.; Berger, L. C.; Brake,
W. B.; Dykstra, C. C.; Hall, J . E. Activity of cationically
substituted bis-benzimidazoles against experimental Pneumocys-
tis carinii pneumonia. Antimicrob. Agents Chemother. 1993, 37,
1713-1716.
The free base was converted to the salt in a yield of 80%
(yellow solid): mp >310 °C dec; IR (KBr) 3369, 3181, 3037,
1665, 1607, 1502, 1032, 782, 674 cm-1; lH NMR (DMSO-d6) δ
10.24 (s, 2H), 9.86 (s, 2H), 9.27 (s, 2H), 8.06 (d, 4H, J ) 7.94

4562 J ournal of Medicinal Chemistry, 1996, Vol. 39, No. 23
Trent et al.
(12) (a) Nunn, C. M.; J enkins, T. C.; Neidle, S. Crystal structure of
γ-oxapentamidine complexed with d(CGCGAATTCGCG)2. The
effects of drug structural change on DNA minor-groove recogni-
tion. Eur. J . Biochem. 1994, 226, 953-961. (b) Nunn, C. M.;
Neidle, S. Sequence-dependent drug binding to the minor groove
of DNA: crystal structure of the DNA dodecamer d(CG-
CAAATTTGCG)2 complexed with propamidine. J . Med. Chem.
1995, 38, 2317-2325.
(13) (a) Fox, K. R.; Sansom, C. E.; Stevens, M. F. G. Footprinting
studies on the sequence-selective binding of pentamidine to
DNA. FEBS Lett. 1990, 266, 150-154. (b) Greenidge, P. A.;
J enkins, T. C.; Neidle, S. DNA minor groove recognition proper-
ties of pentamidine and its analogues: A molecular modeling
study. Mol. Pharmacol. 1993, 43, 982-988. (c) J enkins, T. C.;
Lane, A. N.; Neidle, S.; Brown, D. G. NMR and molecular
modeling studies for the interaction of berenil and pentamidine
with d(CGCAAATTTGCG)2. Eur. J . Biochem. 1993, 213, 1175-
1184.
(19) (a) Lombardy, R. L.; Tanious, F.; Ramachandran, K.; Tidwell,
R. R.; Wilson, W. D. Synthesis and DNA interactions of benz-
imidazole dications which have activity against opportunistic
infections. J . Med. Chem. 1996, 39, 1452-1462. (b) Tidwell, R.
R.; J ones, S. K.; Naiman, N. A.; Berger, L. C.; Brake, W. B.;
Dykstra, C. C.; Hall, J . E. Activity of cationically substituted
bis-benzimidazoles against experimental Pneumocystis carinii
pneumonia. Antimicrob. Agents Chemother. 1993, 37, 1713-
1716.
(20) Lavery, R.; Sklenar, H. The definition of generalized helicoidal
parameters and of axis curvature for irregular nucleic acids. J .
Biomol. Struct. Dyn. 1988, 6, 63-91.
(21) Drew, H. R.; Samson, S.; Dickerson, R. E. Structure of a B-DNA
dodecamer at 16-K. Proc. Natl. Acad. Sci. U.S.A. 1982, 79, 4040-
4044.
(22) Berman, H. M.; Olson, W. K.; Beveridge, D. L.; Westbrook, J .;
Gelbin, A.; Demeny, T.; Hsieh, S.-H.; Srinivasan, A. R.; Schneider,
B. The nucleic acid database.
A comprehensive relational
(14) Boykin, D. W.; Kumar, A.; Spychala, J .; Zhou, M.; Lombardy,
R. J .; Wilson, W. D.; Dykstra, C. C.; J ones, S. K.; Hall, J . E.;
Tidwell, R. R.; Laughton, C.; Nunn, C. M.; Neidle, S. Dicationic
diarylfurans as Anti-Pneumocystis carinii Agents. J . Med. Chem.
1995, 36, 912-916.
(15) Laughton, C.; Tanious, F.; Nunn, C. M.; Boykin, D. W.; Wilson,
W. D.; Neidle, S. A crystallographic and spectroscopic study of
the complex between d(CGCGAATTCGCG)2 and 2,5-bis(4-gua-
nylphenyl)furan and an analogue of berenil. Structural origins
of enhanced DNA-binding affinity. Biochemistry 1996, 35, 5655-
5661.
(16) Crothers, D. M. Statistical thermodynamics of nucleic acid
melting transtions with coupled binding equilibria. Biopolymers
1971, 10, 2147-2160.
(17) Czarny, A.; Boykin, D. W.; Wood, A. A.; Nunn, C. M.; Neidle, S.;
Zhao, M.; Wilson, W. D. Analysis of van der Waals and
electrostactic contributions in the interactions of minor groove
binding benzimidazoles with DNA. J . Am. Chem. Soc. 1995, 117,
4716-4718.
database of three-dimensional structures of nucleic acids. Bio-
phys. J . 1992, 63, 751-759.
(23) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T.
Semianalytical treatment of solvation for molecular mechanics
and dynamics. J . Am. Chem. Soc. 1990, 112, 6127-6129.
(24) Brunger, A. T.; Kuriyan, J .; Karplus, M. Crystallographic
R-factor refinement by molecular dynamics. Science 1987, 235,
458-460.
(25) Cambillau, C.; Horjales, E. TOM: a FRODO subpackage for
protein-ligand fitting with interactive energy minimization. J .
Mol. Graph. 1987, 5, 174-177.
(26) MacroModel V4.5: Mohamadi, F.; Richards, N. G. J .; Guida, W.
C.; Liskamp, R.; Lipton, M.; Caufield, C.; Chang, G.; Hendrick-
son, T.; Still, W. C. Macromodel - an integrated software system
for modeling organic bioorganic molecules using molecular
mechanics. J . Comput. Chem. 1990, 11, 440-467.
(27) Stewart, J . P. P. Mopac 6.0 (QCPE), available from the Quantum
Chemistry Program Exchange, Indiana University, Bloomington,
IN 47405.
(18) Blagburn, B. L.; Sundermann, C. A.; Lindsay, D. S.; Hall, J . E.;
Tidwell, R. R. Inhibition of Cryptosporidium parvum in neonatal
Hsd:(lCR)BR swiss mice by polyether ionophores and aromatic
amidines. Antimicrob. Agents Chemother. 1991, 35, 1520-1523.
(28) Bajic, M.; Kumar, A.; Boykin, D. W. Synthesis of 2,5-Bis(4-
cyanophenyl)furan. Heterocycl. Commun., in press.
J M9604484