A new class of polyintercalating molecules

Article abstract of DOI:10.1021/ja9706108

We have synthesized a series of polyintercalating compounds, including the first known tetraintercalator, based on the 1,4,5,8-naphthalenetetracarboxylic diimide chromophore. The chromophores are attached in a head-to-tail arrangement by peptide linkers and are synthesized by standard solid phase peptide synthesis methods. We report evidence, based on UV-visible spectroscopy and viscometry, that the compounds are fully intercalated upon binding to double-stranded DNA. Using DNAse I footprinting experiments, the bisintercalator 2 was found to bind to DNA in a cooperative manner. The footprinting results as well as association and dissociation kinetics data reveal that the compounds exhibit a tremendous preference for GC over AT sequences. A mode of binding is proposed in which the compounds intercalate completely from the major groove, and not in a threading manner as may be suggested by their structures. A kinetic scheme is proposed that takes into account the observed cooperativity and fits the data for the dissociations of the polyintercalators from poly(dAdT), although a similar scheme could not adequately model their dissociations from poly(dGdC) or from calf thymus DNA.

Full text of DOI:10.1021/ja9706108

7202
J. Am. Chem. Soc. 1997, 119, 7202-7210
A New Class of Polyintercalating Molecules
R. Scott Lokey,† Yan Kwok,‡ Vladimir Guelev,† Christopher J. Pursell,§
Laurence H. Hurley,‡ and Brent L. Iverson*,†
Contribution from the Departments of Chemistry, Biochemistry, and Pharmacy,
The UniVersity of Texas at Austin, Austin, Texas 78712, and Department of Chemistry,
Trinity UniVersity, San Antonio, Texas 78212
ReceiVed February 25, 1997X
Abstract: We have synthesized a series of polyintercalating compounds, including the first known tetraintercalator,
based on the 1,4,5,8-naphthalenetetracarboxylic diimide chromophore. The chromophores are attached in a head-
to-tail arrangement by peptide linkers and are synthesized by standard solid phase peptide synthesis methods. We
report evidence, based on UV-visible spectroscopy and viscometry, that the compounds are fully intercalated upon
binding to double-stranded DNA. Using DNAse I footprinting experiments, the bisintercalator 2 was found to bind
to DNA in a cooperative manner. The footprinting results as well as association and dissociation kinetics data
reveal that the compounds exhibit a tremendous preference for GC over AT sequences. A mode of binding is
proposed in which the compounds intercalate completely from the major groove, and not in a threading manner as
may be suggested by their structures. A kinetic scheme is proposed that takes into account the observed cooperativity
and fits the data for the dissociations of the polyintercalators from poly(dAdT), although a similar scheme could not
adequately model their dissociations from poly(dGdC) or from calf thymus DNA.
Introduction
bisintercalation, however, have met with relatively few suc-
cesses.3,12,16-19 Of the known trisintercalators, most have
reported association constants within an order of magnitude of
their bisintercalating counterparts; one noteworthy exception is
a polyamine-linked triacridine synthesized by Laugaˆa and co-
workers20 that exhibited an unusually high binding affinity for
double-stranded DNA (Kapp ) 1014 M-1).
The propensity of a molecule to intercalate two or more
groups simultaneously appears to be dependent on both the
length and composition of the linking backbone.3,11,16 The
traditional polyintercalating design in which the aromatic units
are connected in parallel along a contiguous linker may restrict
polyintercalation due to unfavorable distortions in the backbone.
Takenaka et al.19 reported a different design in which three
acridine units were linked in series, like beads on a string. For
these trisintercalating molecules a “threading” type of binding
mode was proposed similar to the mono-21,22 and bisintercala-
tors23,24 previously known to extend into both grooves of DNA
simultaneously. We now report the solid phase synthesis and
DNA-binding properties of a new series of polyintercalators,
including what is to the best or our knowledge the first known
tetraintercalating molecule, in which the aromatic groups are
connected in series. These polyintercalators comprise a new
class of DNA-binding molecules in which properties such as
Numerous compounds, including some clinically used che-
motherapeutic agents, interact with DNA by intercalating one
or more aromatic groups between base pairs of the double helix.
Whereas initial studies focused on molecules with a single
intercalating moiety,1,2 the promise of improved antitumor
activity3-7 and/or sequence specificity8-10 has led researchers
to investigate compounds that contain more than one intercalat-
ing group.11-15 Efforts to extend the interaction beyond
* To whom correspondence should be addressed.
† Departments of Chemistry and Biochemistry, The University of Texas
at Austin.
‡ Department of Pharmacy, The University of Texas at Austin.
§ Trinity University.
X Abstract published in AdVance ACS Abstracts, July 1, 1997.
(1) Lerman, J. S. J. Mol. Biol. 1961, 3, 18-30.
(2) Wilson, W. D.; Jones, R. L. AdV. Pharm. Chemother. 1981, 18, 177-
222.
(3) Wakelin, L. P. G. Med. Res. ReV. 1986, 6, 275-340.
(4) Chaires, J. B.; Leng, F.; Przewloka, T.; Fokt, I.; Ling, Y. H.; Perez-
Soler, R.; Priebe, W. J. Med. Chem. 1997, 40, 261-266.
(5) Waring, M. J. In Molecular Aspects of Anticancer Drug-DNA
Interactions; Neidle, S., Waring, M. J., Eds.; CRC: Boca Raton, FL, 1993;
Vol 1, Chapter 7.
(6) Bosquet, P. F.; Bran˜a, M. F.; Conlon, D.; Fitzgerald, K. M.; Perron,
D.; Cocchiaro, C.; Miller, R.; Moran, M.; George, J.; Quian, X.-D.;
Keilhauer, G.; Romerdahl, C. A. Cancer Res. 1995, 55, 1176-1180.
(7) McRipley, R. J.; Burns-Horwitz, P. E.; Czerniak, P. M.; Diamond,
R. J.; Diamond, M. A.; Miller, J. L. D.; Page, R. J.; Dexter, D. L.; Chen,
S.-F.; Sun, J.-H.; Behrens, C. H.; Seitz, S. P.; Gross, J. L. Cancer Res.
1994, 54, 159-164.
(16) Gaugain, B.; Markovits, J.; Le Pecq, J.-B.; Roques, B. P. FEBS
Lett. 1984, 169, 123-126.
(17) Atwell, G. J.; Baguley, B. G.; Wilmanska, D.; Denny, W. A. J.
Med. Chem. 1986, 29, 69-74.
(8) Waring, M. J.; Fox, K. R. Molecular Aspects of Anticancer Drug
Action; Macmillan: London, 1983.
(18) Hansen, J. B.; Koch, T.; Buchardt, O.; Nielsen, P. E.; Norde´n, B.;
Wirth, M. J. Chem. Soc., Chem. Commun. 1984, 509-511.
(19) Takenaka, S.; Nishira, S.; Tahara, K.; Kondo, H.; Takagi, M.
Supramol. Chem. 1993, 2, 41-46.
(9) Delepiene, M.; Milhe, C.; Namane, A.; Dinh, T. H.; Roques, B. P.
Biopolymers 1991, 31, 331-353.
(10) Bailly, C.; Bran˜a, M.; Waring, M. J. Eur. J. Biochem. 1996, 240,
195-208.
(20) Laugaˆa, P.; Markovits, J.; Delbarre, A.; Le Pecq, J.-B.; Roques, B.
P. Biochemistry 1985, 24, 5567-5575.
(11) Denny, W. A.; Atwell, G. J.; Willmott, G. A.; Wakelin, L. P. G.
Biophys. Chem. 1985, 22, 17-26.
(21) Yen, S.; Gabbay, E. J.; Wilson, D. W. Biochemistry 1982, 21, 2070-
(12) Wirth, M.; Ole, B.; Torben, K.; Nielsen, P.; Norden, B. J. J. Am.
Chem. Soc. 1988, 110, 932-934.
2076.
(22) Williams, L. D.; Egli, M.; Gao, Q.; Bash, P.; van der Marel, G.;
van Boom, J. H.; Rich, A.; Frederick, C. A. Proc. Natl. Acad. Sci. U.S.A.
1990, 87, 2225-2229.
(23) Veal, J. M.; Li, Y.; Zimmerman, S. C.; Lamberson, C. R.; Cory,
M.; Zon, G.; Wilson, W. D. Biochemistry 1990, 29, 10918-10927.
(24) Zimmerman, S. C.; Lamberson, C. R.; Cory, M.; Fairley, T. A. J.
Am. Chem. Soc. 1989, 111, 6805-6809.
(13) Atwell, G. J.; Leupin, W.; Twigden, S. J.; Denny, W. F. J. Am.
Chem. Soc. 1983, 105, 2913-2914.
(14) Hansen, J. B.; Koch, T.; Buchardt, O.; Nielsen, P. E.; Wirth, M.;
Norden, B. Biochemistry 1983, 22, 4878-4886.
(15) Denny, W. A. In Cancer Chemotherapeutic Agents; Foye, W. O.,
Ed.; American Chemical Society: Washington DC, 1995; pp 218-239.
S0002-7863(97)00610-0 CCC: $14.00 © 1997 American Chemical Society

A New Class of Polyintercalating Molecules
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7203
Figure 2. Polyintercalators 1 (n ) 0), 2 (n ) 1), 3 (n ) 2), and 4 (n
) 3), synthesized using standard solid phase peptide synthesis
techniques from monomers 5 and 6.
Table 1. Spectroscopic and Viscometric Data for Compounds 1-4
a
b
c
compd
ꢀF
ꢀB
ꢀDet
φd (deg)
me
1
2
3
4
20 000
27 400
39 200
51 300
10 400
20 100
27 700
39 500
23 400
44 000
74 600
96 000
12
24
37
48
0.41
0.84
1.4
1.9
5
a Extinction coefficients at 386 nm, determined in 10 mM TRIS
buffer with 1 mM EDTA and 50 mM NaCl. b Extinction coefficients
(386 nm) in same buffer containing large excess of CT DNA.
c Extinction coefficients (386 nm) in 2% SDS. d Unwinding angles.
e Helix extension parameters.
6
Figure 1. Synthesis of diimide amino acid precursor. Reagents: (a)
â-alanine benzyl ester p-toluene sulfonate salt, BOCN(CH2)2NH2, i-Pr2-
NEt; (b) H2, Pd/C; (c) TFA/CH2Cl2; (d) FMOC-Gly-OC6F5, HOBT,
2,6-lutidine; (e) pentafluorophenol, DCC; (f) diglycine (TEA salt), sulfo-
NHS, EDC; (g) pentafluorophenol, DCC.
Figure 3. Binding isotherm of 1 and calf thymus DNA, determined
spectrophotometrically using the change in absorbance of 1 at 386 nm
upon addition of DNA. The data were fit to the McGee-von Hippel
equation,28 where r ) the concentration of bound ligand divided by
the total concentration of DNA, and c ) the concentration of free ligand.
sequence specificity and cell permeability can be explored in a
general way using combinatorial methods and/or rational design
to change the sequence of the linker segments.
absorbance maxima at 382 and 384 nm, respectively. In the
absence of DNA, the molar extinction coefficients of 2-4 at
384 nm, divided by the number of chromophores in each
molecule, are significantly less than the extinction coefficient
of 1 at that wavelength (Table 1). In 2% SDS solution, however,
the absorbance spectra of 2-4 are simple multiples of the
spectrum of 1 (Table 1), both in terms of extinction coefficient
and lineshape.
All of the polyintercalators show, in accord with previous
studies using the same chromophore,21 significant hypo-
chromism upon binding to double-stranded DNA (Supporting
Information). This absorbance change in the ultraviolet spec-
trum upon intercalation was used to study the equilibrium and
kinetic aspects of the interactions of compounds 1-4 with DNA.
Binding Studies. The binding isotherm between 1 and CT
DNA was determined spectrophotometrically and fit according
to the McGee-von Hippel equation.28 A value of 1.4 × 104
M-1 was obtained for Keq, and the expected value29 of 2.0 base
pairs was obtained for n, the binding site size (Figure 3). Direct
equilibrium analyses could not be carried out with 2-4, because
free ligand could not be detected in a concentration regime low
enough to evaluate their association constants (>107 as estimated
from the DNAse I footprinting results, Vide infra).
Results
Synthesis. Our polyintercalating molecules are based on a
modification of the design we reported previously for the
construction of so-called “aedamers”, synthetic molecules
designed to fold into an abiotic secondary structure.25 The
design utilizes 1,4,5,8-naphthalenetetracarboxylic diimide, an
electron-deficient aromatic group that has been shown previously
to intercalate into double-stranded DNA.21 Molecular dynamics
simulations using a flexible peptide backbone on a bisinterca-
lated DNA template26 predicted that a four amino acid segment
between diimide moieties would have the optimum combination
of length and flexibility to span two base pairs, allowing
polyintercalation in accord with the general principle of nearest
neighbor exclusion. One lysine residue was placed on each
segment to provide electrostatic attraction to the DNA. The
amino acid building blocks 5 and 6 were synthesized (Figure
1), and compounds 1-4 were constructed using standard solid
phase peptide synthesis (SPPS) methodology27 (Figure 2). A
number of different studies were carried out to investigate the
DNA-binding behavior of the predicted polyintercalators 1-4.
Hypochromism. In aqueous buffer, the diimide chro-
mophores of compound 1, and compounds 2-4, give rise to
Unwinding Studies. The helix extension parameter, which
is a measure of the lengthening of linear fragments of double
stranded CT DNA, was determined for each compound visco-
(25) Lokey, R. S.; Iverson, B. L. Nature 1995, 375.
(26) Peek, M. E.; Lipscomb, L. A.; Bertrand, J. A.; Gao, Q.; Roques, B.
P.; Garbay-Jaureguiberry, C.; Williams, L. D. Biochemistry 1994, 33, 3794-
3800.
(27) Atherton, E.; Shepard, R. C. Solid Phase Peptide Synthesis; IRL:
Oxford, 1989.
(28) McGhee, J. D.; von Hippel, P. H. J. Mol. Biol. 1974, 86, 469-489.
(29) Crothers, D. M. Biopolymers 1968, 6, 575.
(30) Cohen, G.; Eisenberg, H. Biopolymers 1969, 8, 45-49.

7204 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Lokey et al.
Figure 4. Viscometric titrations of 1 (2); 2 ((); 3 (9); and 4 (b) with
sonicated calf thymus DNA, carried out in TE buffer (10 mM Tris, 1
mM EDTA) with 50 mM NaCl at 20 °C. Since the monomer was
only partially bound at the concentrations used in the viscosity
experiment, the [compound]/[DNA base pairs] values were corrected
based on the known association constant (Ka ) 1.0 × 105 M-1) obtained
from the spectrophotometric titration of 1 with calf thymus DNA.
Figure 5. Unwinding of closed circular supercoiled plasmid DNA by
compounds 1 (2); 2 ((); 3 (9); and 4 (b). For monomer 1, equivalence
point ratios at three different DNA concentrations were plotted vs the
concentration of DNA and the true equivalence point was extrapolated
according to Vinograd et al.31 Unwinding angles were calculated based
on an unwinding angle for ethidium bromide of 26°.
metrically.30 As shown in Figure 4, the slopes (m) increase in
increments according to the number of diimide groups per
molecule. The unwinding angles for compounds 1-4 were
determined using standard viscometric methods31 with closed
circular supercoiled plasmid (CCS) DNA (Figure 5). As with
the helix extension parameters, the unwinding angles for 2-4
are multiples of that found for 1.
Kinetics Studies. To investigate the dynamics of the
interactions of 1-4 with DNA, association and dissociation rates
were determined for CT DNA, poly(dAdT), and poly(dGdC)
(Tables 2 and 3, respectively). The association rates and
monomer dissociation rates required stopped-flow techniques,
while the slower off-rates for compounds 2-4 were measured
using a standard spectrophotometer. All dissociation rate
measurements required the presence of 2% SDS in the buffer
to sequester the dissociated intercalators. Control studies
revealed that compounds 1-4 did not reassociate with the
nucleic acids in 2% SDS, but at lower SDS concentrations,
reassociation was observed.
For all of the association profiles, calculated fits represent
the minimum number of exponentials required to give purely
random residuals. Most of the dissociation profiles were also
adequately described by up to three exponentials (the kinetic
traces are available as Supporting Information). The dissociation
profiles of compounds 2-4 from CT DNA, however, were too
Figure 6. DNAse I footprinting analysis of 78-mer DNA with the
“+” (a) and “-” (b) strands labeled on its 5′ end, in the presence of
compounds 1-4. Lanes labeled AG and TC represent the purine- and
pyrimidine-specific sequencing reactions. Lane C0 contains DNA
without DNAse I. Lane C contains DNA with DNAse I but no
compound. For compound 1, lanes 1-5 contain 0.5, 5, 50, 100, and
500 µM compound, respectively. For compounds 2-4, lanes 1-5
contain 62.5, 125, 250, 500, and 1000 nM compound, respectively.
complex to be fit even to a triexponential decay process, and
reproducibly nonrandom residuals were obtained in each of these
cases. This complexity is not unexpected considering the
sequence heterogeneity of natural DNA and the large number
of potential binding sites. Compound 1, with only one diimide
(31) Reve´t, B. M. J.; Schmir, M.; Vinograd, J. Nature New Biol. 1971,
229, 10-13.

A New Class of Polyintercalating Molecules
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7205
Table 2. Association Rates
1a
DNA type kb (s-1) Ac (%) k (s-1) A (%) k (s-1) A (%) k (s-1) A (%)
unit, shows distinct kinetic behavior that is generally faster, both
in terms of on- and off-rates, and less complex than the
polyintercalators 2-4.
2
3
4
The kinetic data reveal that 1-4 exhibit a dramatic preference
for poly(dGdC) over poly(dAdT), reflected primarily in their
dissociation rates. While the dissociations of compounds 2-4
from poly(dAdT) are essentially complete within 200 s, the
longest (and most predominant) component of their dissociations
from poly(dGdC) has a half-life of at least 48 h. Interestingly,
compounds 2-4 have virtually identical dissociation rate
profiles, in terms of their time constants and relative amplitudes.
The slowest components for the dissociation of 2-4 from poly-
(dGdC) could not be determined with accuracy because of
baseline drift and some sample precipitation over the long period
required to observe complete dissociation of the complex.
Experiments were performed in which the compounds were
mixed with poly(dGdC) dissolved in 2% SDS, and in no case
could reassociation be observed, even over a period of several
days. This indicates that the dissociations from poly(dGdC)
go to completion and not to some partially bound equilibrium
state. To fit the dissociation curves for poly(dGdC), the
asymptote representing complete dissociation was set as a
constant according to the known extinction coefficients of the
compounds in 2% SDS.
DNAse I Footprinting. A synthetic oligonucleotide duplex
was used for the footprinting studies that contained three GC
sites of four, six, and eight base pairs in length, separated by
stretches of AT. The footprinting gels (Figure 6a,b) reveal the
predicted preferential binding at the GC tracts for all the
molecules including the monomer 1. For each of the poly-
intercalators 2-4, all three of the GC sites became occupied
simultaneously within a narrow range of polyintercalator
concentration. The dimer and tetramer show identical foot-
printing patterns; the only difference between corresponding
lanes is that a 2-fold higher concentration of 2 is required to
produce the same footprint made by 4. Although the footprint
of the trimer is similar to those of the dimer and tetramer, there
is an important difference: Whereas the dimer apparently binds
cooperatively to the GC sites (see discussion section below),
the footprint of the trimer appears to increase gradually in size
as a function of concentration. For example, the trimer displays
a significantly smaller footprint at 125 than at 250 nM, while
the dimer appears to be completely unbound at 125 nM and
completely bound to the GC sites at 250 nM.
poly(dAdT) 12.0
100 40.1
4.1
33 41.9
38 28.4
32
42
26
49
27
48
4.9
44
18
40
38
4.8
1.2
4.7
0.60
0.90 19
0.99
5.7
0.70
poly(dGdC) 29.2
8.3
22
78
6.6
1.3
37
25
0.28 38
0.083 22
0.094 24
CT DNA
38.5
9.7
42
43
15
3.9
1.2
0.25 43
36
31
9.6
2.1
23
41
36
8.4
1.6
0.28
29
38
33
0.78
0.28
1
cant upfield shift of the diimide hydrogens in the H NMR
spectra of 2-4 relative to 1, is indicative of intramolecular
association of the hydrophobic diimide groups in aqueous
solution. In 2% SDS solution, however, the absorbance spectra
of 2-4 are in fact simple multiples of the spectrum of 1 (Table
1), both in terms of extinction coefficient and lineshape.
Presumably, the SDS disrupts the hydrophobic interactions that
are responsible for the stacked conformations in aqueous buffer.
While in general it may not be appropriate to deduce the
degree of intercalation of a compound based on its absorbance
spectrum, trends in extinction coefficients (as well as other
physical properties, such as unwinding parameters) in a
homologous series of compounds contain information when
isolated measurements for any one compound would not. Such
is the case for the series 1-4 bound to DNA. For example, if
all of the chromophores in a homologous series such as 1-4
are intercalated, one would expect (in the absence of significant
perturbations between neighboring intercalated groups) the
absorption spectra of their DNA-bound states to follow the
simple equation
ꢀn(bound) ) nꢀ1(bound)
where ꢀ1(bound) is the extinction coefficient of the monomer bound
to DNA and ꢀn(bound) is the extinction coefficient of compound
n, with n chromophores, bound to DNA. This is the case for
compounds 1-4, as shown in Table 1. In this system, the
analysis is particularly sensitive to the existence of unbound
chromophores because the extinction coefficient of the unbound
monomer is over 2-fold higher than that of the bound monomer.
Any diimide groups remaining free in solution would be
expected to contribute a larger absorbance and thus increase
the value of ꢀn(bound) above that observed.
The helix extension parameters and unwinding angles of 2-4
are also simply multiples of the values measured for compound
1, suggesting that all of the molecules have each chromophore
intercalated. It is worth noting that the values for the compound
1, namely 0.41 and 12° for the helix extension parameter and
unwinding angle, respectively, are somewhat lower than what
has been observed for some intercalators.30 Nevertheless, our
values fall within the range observed for other intercalators that
contain the same diimide chromophore.21 Thus, it appears that
the peptide tether has little effect on the degree to which each
intercalating moiety lengthens double-stranded DNA. Taken
together, the above spectroscopic and unwinding data indicates
full intercalation of compounds 2-4. To the best of our
knowledge, this represents the first example, despite at least
one published attempt,12 of tetraintercalation by a single
molecule.
Chemical Footprinting. Chemical footprinting methods are
often employed to elucidate the groove and sequence preferences
of DNA-binding molecules. Using Fe-EDTA, which cleaves
DNA largely via the minor groove, and dimethyl sulfate (DMS),
which alkylates the N7 position of guanine and causes cleavage
via the major groove, footprinting experiments were carried out
with the same oligonucleotides used in the DNAse I studies.
Studies were performed under conditions in which 1-4 are
known to be bound, but in no case was protection from cleavage
by either chemical reagent observed. On the contrary, in the
DMS experiments enhancement of methylation-induced cleav-
age was seen at most of the bound guanosine positions (data
not shown), possibly indicative of distortion at the bound GC
sequences.
Discussion
Degree of Intercalation. In the absence of DNA, compounds
2-4 give rise to absorbance spectra lower in intensity and
different in shape from what would be expected by simply
multiplying the absorbance spectrum of 1 by the number of
chromophores present in each molecule. This, and the signifi-
Kinetics. In general, the similarity in the rates of association
of 2-4 with a given type of DNA (Table 2) can be related to
a rate-determining preassociation step that is independent of the
specific DNA-ligand interactions. For example, such similarity

7206 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Lokey et al.
Applying this reasoning to compounds 3 and 4 of our system
results in a structure, referred to here as model A, in which the
aromatic groups thread through the double helix such that their
linker segments reside in the major and minor grooves in an
alternating fashion (Figure 7a).
An alternative possibility for the bound structure, referred to
here as model B, is one in which the diimide groups intercalate
into double-stranded DNA “edge-on”, while the linker segments
reside entirely within the same DNA groove (Figure 7b). The
helical twist of the DNA duplex could allow for considerable
intercalation of the diimides within the major groove, while still
providing enough space for the linker chains to extend from
each end of the diimide. Although the chemical footprinting
results were inconclusive regarding the groove preference of
compounds 2-4, we consider similar binding in the smaller
minor groove unlikely on steric grounds. Moreover, a recent
report by Waring et al.10 on LU799553, a naphthalimide
bisintercalator, indicates that the naphthalimide moiety in this
molecule probably intercalates via the major groove of DNA.
Some hybrid of modes A and B is also possible with 3 and 4.
In the absence of definitive structural data from either NMR
spectroscopy or X-ray crystallography (these methods have
yielded little information to date), we have relied on more
indirect routes to discern the binding modes of compounds 1-4.
As described below, based primarily on kinetic evidence, we
conclude that of the two model B is more consistent with the
available data.
All components of the stopped-flow association kinetics of
1-4 with poly(dGdC) are several times slower than those for
poly(dAdT), implying that transient breathing of the double helix
(an event that is about 25 times more likely for an AT than a
GC base pair)35,36 has an influence on the rates of association.
This would tend to favor model A; indeed, Fox and Waring
observe a similar trend in the association of the threading
intercalator nogalamycin,34 certain components of which they
ascribe to a rate-determining DNA breathing step. If, however,
model A is correct, then one would not expect such similar
kinetic behavior among all of the polyintercalating species 2-4.
Relative to the bisintercalator 2, the tris- and tetraintercalators
would require extra, presumably rate-limiting, steps to thread
into, and out of, the double helix. On the other hand, if DNA
binding proceeded according to model B, the accessibility of
the major groove would be similar to all intercalating units in
the ligand and could result in the observed similar rate profiles
for the polyintercalators.
Figure 7. Schematic representation of the binding of tetraintercalator
4 to double stranded DNA by A weaving into and out of the double
helix such that the linkers reside in both grooves or by B intercalation
solely from one side of the double helix.
in on-rates was observed among a series of mono-, bis-, and
trisacridines with poly(dAdT) and was attributed to rate-limiting
desolvation of both the ligand and the DNA to form a
preassociation complex.20,33 Since polyintercalators 2-4 exhibit
a significant degree of self-stacking in aqueous solution, it is
also possible that a slow unstacking step, likely intramolecular
but possibly having some intermolecular components at higher
concentrations, is required before these molecules are able to
intercalate. The monomer, which is not significantly aggregated
under the conditions of the association experiments (Vide supra),
displays significantly faster association rates than 2-4 for the
different DNAs, consistent with the above idea.
Cooperativity. A synthetic oligonucleotide duplex was used
for the footprinting studies that contained three GC sites of four,
six, and eight basepairs in length, separated by stretches of AT.
The footprinting gels (Figure 6a,b) reveal the predicted pref-
erential binding at the GC tracts for all the molecules including
the monomer 1. For each of the polyintercalators 2-4, all three
of the GC sites became occupied simultaneously within a very
narrow range of polyintercalator concentration. The dimer and
tetramer show identical footprinting patterns; the only difference
between corresponding lanes is that a 2-fold higher concentration
of 2 is required to produce the same footprint made by 4.
Importantly, at no concentration is any footprint smaller than
nine basepairs observed with 2 (or 4). That is, the cooperative
binding of two molecules of 2 simultaneously is more favorable
than binding of a single molecule of 2, thereby explaining the
nine basepair footprint eVen at the four basepair GC site.
Cooperative binding behavior has recently been reported for
the bisintercalating drug echinomycin.32
Binding Modes. The DNA-bound state of a tetraintercalating
polydiimide such as 4 can be represented by at least two distinct
schematic models (Figure 7). Wilson et al. observed the
intercalation of several compounds containing the 1,4,5,8-
naphthalenetetracarboxylic diimide moiety with bulky groups
such as adamantyl attached at both of the nitrogen atoms.21
These authors proposed that in order to intercalate, at least one
of the bulky groups must pass through the double helix, invoking
a dynamic model of DNA that allows for transient DNA
breathing, even at temperatures well below its melting temper-
ature. This type of “threading” intercalation has been attributed
to other DNA-binding compounds, namely the drug nogalamy-
cin,22,34 and Takenaka et al.19 reported a trisintercalating
molecule that was proposed to bind in a threading manner.
Specifically, according to model A, one would expect 3 and
4 to display significantly slower off-rates compared with 2. A
salient feature of the dissociation data is that, within a given
type of DNA, compounds 2-4 display remarkably similar
kinetic behavior. This includes all of the key parameters of
the fits, including the number of components observed, the
relative extent of each component, and the magnitudes of the
derived constants. No reasonable mechanistic model that
involves “unthreading” of either of the longer intercalators 3
or 4 could be produced that fits the dissociation data. Of course,
we cannot rigorously rule out model A. For example, it is
possible, albeit unlikely, that the “threading” steps do not
produce spectroscopic changes that are detectable by the
absorbance techniques used in these kinetic studies. Assuming
no such ambiguities exist, we propose a mechanism below for
(34) Fox, K. R.; Waring, M. J. Biochim. Biophys. Acta 1984, 802, 162-
168.
(32) Bailly, C.; Hamy, F.; Waring, M. J. Biochemistry 1996, 35, 1150-
1161.
(33) Capelle, N.; Barbet, J.; Dessen, P.; Blanquet, S.; Roques, B.; Le
Pecq, J.-B. Biochemistry 1979, 17, 3354-3366.
(35) Teitelbaum, H.; Englander, S. W. J. Mol. Biol. 1975, 92, 93-111.
(36) Teitelbaum, H.; Englander, S. W. J. Mol. Biol. 1975, 92, 55-78.
(37) Kaiser, E.; Colescott, R. L.; Bossinger, C. D.; Cook, P. I. Anal.
Biochem. 1970, 2324, 595.

A New Class of Polyintercalating Molecules
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7207
the dissociations of 2-4 from poly(dAdT) that is consistent with
all of the available data if the compounds are bound according
to model B.
Mechanism of Dissociation. We investigated whether a
mechanistic scheme within model B could be generated that
predicts the similarity in dissociation profiles between the
different compounds for a given type of DNA. We began with
the assumptions that (a) the elementary steps in a sequential
mechanism should be similar among all three of the polyinter-
calators 2-4, (b) the N-to C-terminal asymmetry of the
compounds could be ignored, and that (c) in the presence of
SDS, which solvates the unbound chromophores, bimolecular
reassociation rates are negligible. These assumptions allowed
us to prune the many branching steps that are theoretically
possible for a multiintercalating species down to a manageable
number. We then evaluated potential mechanisms by comparing
the resulting theoretical dissociation curves with the actual data.
Specifically, we constructed theoretical absorbance-vs-time
curves assuming a constant increase in absorbance per chro-
mophore expelled from the DNA. A rate equation for the
unbound species was generated for a given mechanism and
substituted into the absorbance expression (Supporting Informa-
tion). The resulting differential equation was numerically solved
for a given set of rate constants and plotted against the
experimental data. The rate constants were manually varied,
and the resulting plots were used to obtain an interpolative fit.
While this approach may seem rather primitive, the complexity
of the resulting analytical solutions prevented the use of a
mathematically more rigorous least-squares fit. Plotting the
numerical solutions for each set of differential equations, on
the other hand, allowed us to evaluate rapidly the viability of a
particular mechanism directly using the elementary step rate
constants with no restrictions on the complexity of the resulting
exponential function.
Figure 8. Simplified mechanistic scheme for the dissociation of
intercalators 2-4 in which the dimer 2 binds cooperatively in pairs.
All three compounds dissociate from DNA in a similar stepwise pattern,
in which the off-rates depend on the occupancy of neighboring
intercalated sites.
“flexible” rate constants whose values could be varied signifi-
cantly (within an order of magnitude) without a large impact
on the fit. These numbers show k3 to be the rate limiting step,
and attempts to reverse the relative size of the constants failed.
Note that for the trimer and tetramer the proposed order of
dissociation from the intercalated sites would require the
existence of intermediates in which dissociated chromophores
are tethered on both sides by intercalated groups (Figure 8).
However, given the small difference between k2 and k3, it is
possible that any “middle” bound residue in the trimer and
tetramer would be held tightly in position by the linkers on both
sides, giving rise to an alternative mode of dissociation (Figure
9, pathway b).
Our mathematical model does not explicitly differentiate
between intermediates such as a and b in the dissociation
reaction of compound 3, as the absorbance curve is simulated
by simply counting the number of unbound chromophores in
each species. For the dimer 2, where the two corresponding
intermediates (Figure 10) theoretically give rise to kinetically
distinct pathways, we observed an equally good fit for both
mechanisms (via intermediates c or d, corresponding to a and
b for the trimer, respectively), with the same set of rate constants.
This is not surprising considering the negligible reverse rate
constants (k-2a ≈ k-3b ≈ 0.0005 s-1) in both instances. In light
of the above, we propose that in reality the dissociation could
proceed through a combination of these two pathways. Since
each elementary step is virtually irreversible in the poly(dAdT)
fits, the constants obtained are possibly linear combinations of
the elementary step rate constants.
Initially we attempted to reproduce the dissociation profiles
of all three compounds with a simple mechanistic scheme in
which the intercalating groups dissociate in no particular order
and at approximately the same rate. This model was unable to
produce satisfactory fits for any set of experimental data and
was quickly abandoned.
In an alternative approach, we introduced as a constraint the
fact that the dimer 2 binds cooperatively in units of two (as
observed in the DNAse I footprinting experiments.) In such a
scenario, the rates of dissociation of the individual intercalating
groups are largely dependent upon the occupancy of neighboring
sites. To make the math tractable, we postulated a stepwise
nonbranching mechanism (Figure 8), based on the following
assumptions: First, since cooperativity was observed with the
dimer, we inferred that binding to the third, but not necessarily
to the fourth, occupied site is cooperative for each of the
compounds. Second, the fact that the Keq for the dimer is less
than the square of the Keq of the monomer implied that the
binding of adjacent groups (separated by two base pairs) is non-
or perhaps even anticooperative. In terms of off-rates these data
suggested the order kthird < kfirst ≈ ksecond ≈ kfourth for the
dissociation from the respective intercalated sites.
The best fits (available as Supporting Information) obtained
for the dissociation of compounds 2-4 from poly(dAdT)
according to our model (Figure 8) yielded the rate constants
listed in Table 4 (note the similarity in the rate constant sets
for the three compounds). However, the postulated mechanistic
scheme failed to produce satisfactory fits for the poly(dGdC)
data.
The existence of such alternate pathways may account for
the more complex dissociation curves observed for poly(dGdC).
The slower experimental off-rates measured for all three
The results shown in Table 4 for the poly(dAdT) data support
our initial assumptions. Numbers in parentheses indicate

7208 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Lokey et al.
Table 3. Dissociation Rates
1
2
3
4
k (s-1
)
A (%)
k (s-1
0.12
0.021
)
A (%)
k (s-1
0.097
0.021
)
A (%)
k (s-1
0.11
0.019
)
A (%)
poly(dAdT)
poly(dGdC)
15.4
6.6
0.98
27
73
23
77
8
92
8
15
77
34
35
31
12
88
6
15
79
37
35
28
7
93
5
15
80
39
46
15
5.9 × 10-4
4.6 × 10-5
3.8 × 10-6
5.5 × 10-3
5.0 × 10-4
1.0 × 10-4
6.9 × 10-4
8.2 × 10-5
4.5 × 10-6
6.0 × 10-3
5.5 × 10-4
1.1 × 10-4
4.6 × 10-4
3.7 × 10-5
2.3 × 10-6
4.3 × 10-3
5.0 × 10-4
8.1 × 10-5
0.047
CT DNA
5.1
1.0
0.23
24
36
40
a Numbers refer to compounds 1-4. b Association and dissociation rate constants determined by fitting the spectrophotometrically-determined
decay traces to up to three exponentials (see Experimental Section). c Relative amplitudes from curve fits as percentage of total.
Table 4. Rate Constants Obtained from Fitting Dissociation
Curves of Compounds 2-4 with poly(dAdT) According to
Mechanism in Figure 8
N- to C-terminal asymmetry of the polyintercalators, a factor
that may be important on the much slower time scale of the
poly(dGdC) dissociations. In order to take this branching into
account, many more kinetic constants would have to be
introduced, which seriously limits the credibility of any deduced
model. Finally, it may be that 2-4 adopt a somewhat different
mode of binding to poly(dGdC) compared with poly(dAdT).
compd
2
3
4
a
k1
0.12
0.11
k-1
k2
k-2
k3
k-3
k4
7 × 10-4
5 × 10-4
7.0 × 10-2
0.12
7.0 × 10-2
(5 × 10-4
)
(5 × 10-4
)
(5 × 10-4
)
b
Conclusion
2.1 × 10-2
2.1 × 10-2
2.1 × 10-2
(5 × 10-4
)
(5 × 10-4
)
We have described a new class of polyintercalating com-
pounds and the first known tetraintercalator. Using DNAse I
footprinting, we found that cooperativity is an important feature
in the binding of our polyintercalators to double-stranded DNA.
Using kinetic data obtained from stopped-flow and manual
mixing techniques, we were able to propose a mode of binding
for compounds 2-4 and a mechanism for their dissociations
from poly(dAdT). The solid phase synthesis of these com-
pounds, in which the intercalating groups are attached to one
another by peptide linkers, suggests the possibility of developing
novel DNA-binding compounds using the rapidly advancing
techniques of combinatorial and solid phase chemistry. With
this design we may also tap into the burgeoning field of non-
peptide solid-phase synthesis to generate polyintercalators
separated by rigid, highly functionalized linkers in which
properties such as sequence specificity or enzyme inhibition (i.e.,
topoisomerase inhibition) could be selected from a library or
rationally designed.
(0.12)
(0.10)
(0.12)
a Rate constants (s-1). b Parentheses denote numbers that can be
varied by at least an order of magnitude without greatly affecting the
closeness of the fit.
Figure 9. Proposed branching pathways for the dissociation of trimer
3, according to the calculated rate constants (Table 4). A theoretical
model based on hypochromism does not differentiate between pathways
a and b.
Experimental Section
General Considerations. Routine NMR spectra were recorded on
a Bruker 250 MHz. Spectra of compounds 2-4 were recorded on a
Varian 500 MHz instrument. Solvents and starting materials were used
without further purification unless otherwise stated. Organic chemicals
were obtained from Aldrich and nucleic acids were obtained from
Sigma. Dissociation profiles were simulated in Mathematica 2.2
(Wolfram Research) on a UNIX Sun SPARC 2000 station.
N-(2-tert-Butoxycarbonylaminoethyl)-N′-(2-carboxyethyl)-1,4,5,8-
naphthalenetetracarboxylic Diimide. 1,4,5,8-Naphthalenetetracar-
boxylic dianhydride (8.96 g, 33.4 mmol) was suspended in 450 mL of
i-PrOH. Mono-tert-butoxycarbonylaminoethylamine (5.34 g, 33.4
mmol) and â-alanine benzyl ester p-toluenesulfonate salt (11.72 g, 33.4
mmol) were added to the mixture, followed by N,N-diisopropylethyl-
amine (6.4 mL, 37 mmol). The mixture was heated to reflux for 24 h,
after which it was concentrated in Vacuo. The resulting solid was
partitioned between CH2Cl2 (450 mL) and 0.1 M sodium citrate buffer
(pH 4.5, 500 mL). The organic layer was washed with citrate buffer
(2 × 100 mL), saturated NaHCO3 (3 × 100 mL), and then brine (100
mL) and dried over anhydrous NaSO4. MeOH (100 mL) was added
to the mixture, and it was filtered through a short pad of silica gel.
The filter was rinsed with 10% MeOH/CH2Cl2 (100 mL), and the
supernatant concentrated to yield the expected statistical mixture of
products (18.7 g, 91%). This mixture was carried on to the next step
without further purification.
Figure 10. Theoretically distinct pathways in the dissociation of dimer
2, corresponding to the branching proposed for compounds 3 (Figure
9) and 4 (not shown).
compounds suggest both smaller dissociation and larger reverse
(reassociation) constants, which may lead to more complex
kinetic behavior, than for poly(dAdT). Furthermore, one or
more of the assumptions that allowed a fit of the poly(dAdT)
data may not be valid in the case of the dissociations from poly-
(dGdC). For example, a branching of the mechanistic steps
would have to be introduced in order to take into account the

A New Class of Polyintercalating Molecules
J. Am. Chem. Soc., Vol. 119, No. 31, 1997 7209
The above tan solid was suspended with sonication in absolute EtOH
(250 mL) and purged with Ar. Pd/C (14 g, 10%) was added, followed
by 15 mL of 1,4-cyclohexadiene added dropwise. The mixture was
charged with H2 at atmospheric pressure and stirred for 48 h. After
concentrating the reaction to 50 mL (bath temperature 30 °C, 20 Torr),
250 mL of 15% TEA/CH2Cl2 was added. The resulting slurry was
stirred for 20 min and filtered through Celite. The filter was rinsed
with 10% TEA/CH2Cl2 until the rinsings were colorless, and the solution
was evaporated in Vacuo. The resulting solid was taken up into CH2-
Cl2 with just enough MeOH to allow for complete solvation, and the
product mixture was deposited onto silica by evaporation. Chromato-
graphic separation using a gradient of 0 f 10% MeOH in 10% TEA/
CH2Cl2 yielded the TEA salt of the BOC-protected amino acid as the
second, bright orange band. After concentration the solid was taken
up into 10% MeOH/CH2Cl2 and 20 mL of HOAc was added. Hexanes
(300 mL) were slowly added to precipitate the product. After cooling
for several hours, the mixture was filtered, and the precipitate was rinsed
with MeOH and dried in Vacuo to yield the carboxylic acid (5.10 g,
27%) as a fine white powder: 1H NMR (DMSO-d6) δ 8.58 (s, 4H),
6.90 (br t, 1H), 4.25 (t, J ) 7.5 Hz, 2H), 4.12 (br t, 2H), 3.27 (br q,
2H), 2.62 (t, J ) 7.4 Hz, 2H), 1.20 (s, 9H); 13C NMR (DMSO-d6) δ
172.4, 162.7, 162.4, 155.8, 130.4, 130.3, 126.4, 126.1, 77.5, 37.6, 36.1,
32.0, 28.1; HRMS (FAB) m/z 482.1547 (482.1563 calcd for C24H24N3O8).
N-2-(Nr-9-Fluorenylmethoxycarbonylglycyl)aminoethyl-N′-(2-
carboxyethyl)-1,4,5,8-naphthalenetetracarboxylic Diimide. The above
BOC-protected amino acid (2 g, 4.16 mmol) was suspended in 15 mL
of CH2Cl2, and 15 mL of TFA was added slowly. After standing for
10 min the solution was evaporated, and the residual TFA removed by
azeotropic evaporation (2×) from heptane. The resulting solid was
triturated with ether, filtered, and dried in Vacuo. The solid was
suspended in DMF (15 mL), and N-9-fluorenylmethoxycarbonylglycine
pentafluorophenyl ester (1.92 g, 4.16 mmol) was added, followed by
1-hydroxybenzotriazole (HOBT, 561 mg, 4.16 mmol) and 2,6-lutidine
(890 mg, 8.32 mmol). After stirring for 4 h the mixture was poured
slowly into rapidly stirred H2O (150 mL), and the resulting suspension
was allowed to stand for several hours. The mixture was filtered, and
the resulting yellow solid was rinsed with H2O and dried in the presence
of P2O5 in a vacuum desiccator overnight. The crude product was
triturated with Et2O several times (to remove the residual 1-hydroxy-
benztriazole and pentafluorophenol), filtered, and dried in Vacuo to yield
the fully protected glycine adduct (86%) as a yellow powder. 1H NMR
(300 MHz, DMF-d7) δ 8.66 (s, 4H), 8.12 (t, 1H, J ) 5.7 Hz), 7.88 (d,
2H, J ) 7.4 Hz), 7.68 (d, 2H, J ) 7.3 Hz), 7.4 (m, 1H), 7.42 (t, 2H,
J ) 7.4 Hz), 7.30 (t, 2H, J ) 7.3 Hz), 4.40 (t, 2H, J ) 7.3 Hz), 4.31
(t, 2H, J ) 5.6 Hz), 4.20 (d, 2H, J ) 6.9 Hz), 4.11 (t, 1H, J ) 6.5 Hz),
3.75 (d, 2H, J ) 5.8 Hz), 3.66 (br dt, 2H, J1 ) 5.5), 2.81 (t, 2H, J )
7.7 Hz); 13C NMR (300 MHz, DMF-d7) δ 173.0, 170.4, 163.6, 162.9,
157.3, 144.8, 141.6, 131.1, 131.0, 128.3, 127.7, 127.2, 127.0, 126.0,
120.6, 67.0, 47.5, 44.6, 40.6, 37.0, 36.2, 32.5; HRMS (FAB) 661.1934
(661.1934 calcd for C36H29N4O9, M+ + H).
Nr-9-Fluorenylmethoxycarbonyl-(Nε-tert-butoxycarbonyl)lysyl-
glycylglycine. N-9-Fluorenylmethoxycarbonyl-(Nꢀ-tert-butoxycarbo-
nyl)lysine (0.9 g, 1.92 mmol) was dissolved, with N-hydroxysulfo-
succinimide (458 mg, 2.11 mmol), in 5 mL of 10% H2O/DMF. 1-Ethyl-
3-[3-dimethylaminopropyl]carbodiimide hydrochloride (EDC) (405 mg,
2.11 mmol) was added, and the mixture was stirred for 4 h. The
triethylammonium salt of diglycine (492 mg, 2.11 mmol.) (It was
prepared as follows: Diglycine was dissolved in water, and excess
triethylamine was added. EtOH was added until the solution became
homogeneous, and the solution was evaporated to dryness in Vacuo.
The resulting solid was suspended in MeOH, filtered, rinsed with Et2O,
and dried in Vacuo.) was dissolved in 5 mL of H2O and added dropwise
to the reaction mixture. After stirring for 16 h the reaction mixture
was partitioned between EtOAc and Na citrate buffer (0.2 M, pH 4.5).
The aqueous phase was extracted with EtOAc (3 × 25 mL), and the
combined organic layers were dried over anhydrous Na2SO4, filtered,
and concentrated in Vacuo. The resulting white foam was purified by
chromatography on silica gel with 5% MeOH/CH2Cl2. Crystallization
by the slow addition of pentane to a solution of the product in 10%
EtOAc/CH2Cl2 (25 mL) yielded FMOCN-Lys(ꢀ-BOC)-Gly-Gly-OH as
a white powder (735 mg, 69%). 1H NMR (300 MHz, DMSO-d6) δ
8.17 (t, 1H, J ) 5.4 Hz), 8.09 (t, 1H, J ) 5.8 Hz) 7.87 (d, 2H, J ) 7.4
Hz), 7.72 (dd, 2H, J ) 7.2, 4.3 Hz), 7.53 (d, 1H, J ) 7.8 Hz), 7.41 (t,
2H, J ) 6.9 Hz), 7.32 (t, 2H, J ) 7.0 Hz), 6.76 (t, 1H, J ) 5.3 Hz),
4.32-4.19 (m, 3H), 3.96 (br q, 1H), 3.74 (t, 4H, J ) 6.1 Hz), 2.88 (q,
2H, J ) 4.5 Hz), 1.7-1.1 (c, 6H), 1.35 (s, 9H); 13C NMR (300 MHz,
DMSO-d6) δ 172.3, 171.1, 169.1, 156.1, 155.6, 143.9, 143.8, 140.7,
127.6, 127.1, 125.3, 120.1, 77.3, 65.6, 54.7, 46.7, 40.6, 39.7, 31.4, 29.2,
28.3, 22.8; HRMS (FAB) m/z 583.2793 (583.2768 calcd for C30H39N4O8,
M+ + H).
Nr-9-Fluorenylmethoxycarbonyl-(Nε-tert-butoxycarbonyl)lysyl-
glycylglycine Pentafluorophenyl Ester (6). The above carboxylic acid
(735 mg, 1.26 mmol) was dissolved in dry dioxane (6 mL) and
pentafluorophenol (349 mg, 1.89 mmol) followed by dicyclohexylcar-
bodiimide (DCC, 285 mg, 1.39 mmol) was added. After stirring for 5
h, the reaction was filtered, the precipitate was rinsed with dry dioxane
(5 mL), and the solution was evaporated in Vacuo. The resulting solid
was triturated with hexanes, filtered, and dried in Vacuo to yield the
pentafluorophenyl ester (803 mg, 1.07 mmol, 85%) as a white
crystalline solid. 1H NMR (300 MHz, CD2Cl2) δ 7.75 (d, 2H, J ) 7.4
Hz), 7.58 (d, 2H, J ) 7.4 Hz) 7.44 (br m, 1H), 7.39 (t, 2H, J ) 7.2
Hz), 7.28 (t, 2H, J ) 7.4 Hz), 7.06 (br m, 1H), 5.96 (br m, 1H), 4.81
(br m, 1H), 4.38 (d, 2H, J ) 7.1 Hz), 4.29 (d, 2H, J ) 5.9 Hz), 4.19
(t, 1H, J ) 6.7 Hz), 4.05 (br t, 1H), 3.97 (d, 2H, J ) 5.6 Hz), 3.06 (br
dt, 2H), 1.95-1.3 (c, 6H) 1.30 (s, 9H); 13C NMR (300 MHz, DMSO-
d6) δ 172.4, 169.8, 166.6, 156.1, 155.6, 143.9, 143.8, 140.7, 127.6,
127.0, 125.3, 120.0, 77.3, 65.7, 54.7, 46.7, 41.7, 32.9, 31.4, 29.2, 28.8,
22.8; HRMS (FAB) m/z 749.2579 (749.2610 calcd for C36H38N4O8F5,
M+ + H).
General Procedure for the Solid Phase Synthesis of Compounds
1-4. Lysine-functionalized 2-chlorotrityl polystyrene resin (Advanced
Chemtech, 335 mg, ∼0.31 mmol/g loading) was added to a 10 mL
fritted-glass filter flask equipped with a screw cap and Teflon stopcock.
The resin was shaken for a few minutes with 50% Et2N/DMF (5 mL)
and then rinsed with DMF (3 × 5 mL), i-PrOH (3 × 5 mL), and then
DMF (3 × 5 mL). The resin was suspended in 2 mL of DMF, and 5
(212 mg, 0.26 mmol) was added. HOBT (35 mg, 0.26 mmol) was
added, and the mixture was shaken for 3 h using a simple oscillating
shaker. (Completion of the coupling could not be monitored using
the Kaiser test37 because compounds containing the 1,4,5,8-tetracar-
boxylic diimide moiety consistently gave a false positive reaction. The
use of a 3 h coupling time and 2.5 equiv of the pentafluorophenyl esters,
however, generally gave high coupling yields based on analysis of the
final products.) The resin was rinsed (DMF/i-PrOH/DMF) and treated
with 50% Et2N/DMF for 30 min. The resin was rinsed again (DMF/
i-PrOH/DMF), and then FMOC-Lys(ꢀ-BOC)-Gly-Gly-OPfp 6 (193 mg,
0.26 mmol) was added along with HOBT (35 mg, 0.26 mmol). After
shaking for 3 h the resin was rinsed, deprotected, and rinsed again as
described above. The process was repeated until the desired number
of units had been coupled. After the final deprotection step the resin
was treated with 1:1:3 HOAc/2,2,2-trichloroethanol/CH2Cl2 (7 mL) for
1 h. The resin was filtered using
N-2-(Nr-9-Fluorenylmethoxycarbonylglycyl)aminoethyl-N′-(2-
pentafluorophenoxycarbonylethyl)-1,4,5,8-naphthalenetetracarbox-
ylic Diimide (5). The above carboxylic acid (1.11 g, 1.68 mmol) was
dissolved in dry DMF (5 mL) and pentafluorophenol (335 mg, 1.85
mmol), followed by dicyclohexylcarbodiimide (381 mg, 1.85 mmol),
were added. After stirring for 16 h the reaction was evaporated in
Vacuo at room temperature to a viscous sludge, to which was added
dry dioxane (10 mL). The resulting suspension was stirred for 5 h
and then filtered. The bright yellow filtrate was evaporated in Vacuo,
and the resulting solid was triturated with pentane. The mixture was
filtered, and the product was rinsed with pentane and dried in Vacuo to
yield the pentafluorophenyl ester (1.21 g, 1.47 mmol, 87%) as a yellow
powder. 1H NMR (300 MHz, DMF-d7) δ 8.67 (s, 4H), 8.13 (t, 1H, J
) 5.6 Hz), 7.86 (d, 2H, J ) 7.4 Hz), 7.65 (d, 2H, J ) 7.3 Hz), 7.47
(t, 1H, J ) 6.0 Hz), 7.39 (t, 2H, J ) 7.3 Hz), 7.27 (t, 2H, J ) 7.4 Hz),
4.55 (t, 2H, J ) 7.1 Hz), 4.30 (t, 2H, J ) 5.6 Hz), 4.16 (d, 2H, J ) 6.8
Hz), 4.09 (d, 1H, J ) 6.4 Hz), 3.72 (d, 2H, J ) 5.9 Hz), 3.63 (dt, J )
6.1 Hz), 3.35 (t, 2H, J ) 7.2 Hz); 13C NMR (300 MHz, DMF-d7) δ
170.4, 168.4, 163.6, 162.9, 157.3, 144.8, 141.6, 131.1, 128.3, 127.7,
127.4, 127.1, 127.0, 126.9, 126.0, 67.0, 47.5, 44.6, 40.7, 37.6, 36.3,
34.3; HRMS (FAB) 827.1775 (827.1776 calcd for C42H28N4O9F5, M+
+ H).

7210 J. Am. Chem. Soc., Vol. 119, No. 31, 1997
Lokey et al.
a cintered glass funnel, and the supernatant was evaporated to yield an
olive green solid. To ensure that all of the compound had been cleaved,
the resin was treated with 5% TFA/CH2Cl2 for 10 min and filtered.
The filtrate was evaporated in Vacuo to yield an orange oil and
combined with the original material that had been cleaved from the
resin. To this mixture was added 1:1 TFA/CH2Cl2 (10 mL), and, after
10 min, the solution was evaporated in Vacuo. The resulting solid was
evaporated from heptane (2 × 10 mL), taken up in H2O (20 mL), and
filtered through a 0.45 m nylon syringe filter. The solution was loaded
onto a reversed phase C18 column (PepRPC 15 m, Pharmacia) and eluted
with a gradient of 10 mM TFA/H2O f 0.8% TFA/CH3CN over 180
min with a flow rate of 0.75 mL/min. The main fraction was collected
and determined to be homogeneous by HPLC (C18, same conditions as
above). The product fractions were pooled and lyophilized to yield
21, 22, 23, or 24 as a light green solid, with yields in the range of
25-35% based on the initial loading of the resin.
via syringe in increments to a solution of 1 (7.2 µM in the same buffer).
The decrease in the monomer’s absorption at 386 nm due to intercala-
tion was measured using a Hewlett-Packard 8452A diode array
spectrophotometer. The data were converted into Scatchard form using
10 375 M-1 cm-1 as the molar extinction coefficient for bound 1,
derived from a Beer’s Law titration of 1 into a 1.5 mM solution of CT
DNA. To derive values for Ka and n, the data were fit to the McGee-
von Hippel28 equation using the KaleidaGraph curve fitting routine.
Stopped-Flow Kinetics. Experiments were performed using a
Dionex D-130 stopped-flow apparatus with a 20 mm pathlength. The
monochromator was set at 386 nm, and the signal from the photomul-
tiplier was amplified with a digital oscilloscope. All experiments were
performed at 25 °C in 10 mM Na PIPES buffer at pH 7.0 with 50 mM
NaCl (PIPES 05E). For the association experiments, solutions of 1-4
at concentrations of 10, 13.3, 20, and 40 µM, respectively, were mixed
upon actuation with equal volumes of CT DNA, poly-dAdT, and poly-
dGdC, at concentrations of 0.2, 0.15, and 0.15 mM, respectively. For
the dissociation experiments, the solutions resulting from the association
experiments were collected and mixed upon actuation with equal
volumes of 4% SDS. For compounds 2-4, little or no dissociation
was observed from any of the DNAs over a 50 s time interval, indicating
that direct mixing could be used to observe all components of their
dissociation kinetics. All kinetic traces and residuals representing the
best up to three exponentials are available as Supporting Information.
Slow Dissociation. In a typical experiment, 0.5 mL of a 4% SDS
solution in PIPES 05E was mixed with 0.5 mL of a solution of DNA-
ligand complex in a quartz cuvette of 10 mm pathlength, taking care
to avoid trapping air bubbles along the walls of the cuvette. Initial
ligand concentrations were the same as those used in the stopped flow
experiments, and initial concentrations of DNA were at 0.2 mM for
CT DNA, poly(dAdT), and poly(dGdC). Within 20 s of mixing, the
absorbance at 386 nm was recorded as a function of time using a
Hewlett-Packard 8452A spectrophotometer.
Compound 1. 1H NMR (500 MHz, 10% D2O/H2O) δ 8.64 (d, 2H,
J ) 7.6 Hz), 8.59 (d, 2H, J ) 7.6 Hz), 8.33 (t, 1H, J ) 5.8 Hz), 8.27
(d, 1H, J ) 7.5 Hz), 4.41 (m, 2H), 4.32 (t, 2H, J ) 5.9 Hz), 4.23 (dt,
1H, J ) 5.2, 7.9 Hz), 3.71 (s, 2H), 3.64 (dt, 2H, J ) 5.9, 5.8 Hz), 2.92
(br m, 2H), 2.72 (t, 2H, J ) 6.8 Hz) 1.72 (m, 2H), 1.61 (m, 2H), 1.34
(p, 2H, J ) 7.6); HRMS (FAB) m/z 567.2222 (567.2203 calcd for
C27H31N6O8, M+).
Compound 2. 1H NMR (500 MHz, D2O) δ 8.41-8.32 (m, 8H),
4.48-4.29 (m, 5H), 4.25 (t, 2H, J ) 8.9 Hz), 4.20 (t, 2H, J ) 9.0 Hz),
4.11-3.83 (m, 6H), 3.74 (s, 2H), 3.59 (m, 4H), 3.03 (t, 2H, J ) 7.7
Hz), 2.93 (t, 2H, J ) 7.6 Hz), 2.86 (m, 2H), 2.71 (t, 2H, J ) 7.1 Hz),
1.94-1.3 (m, 12H); HRMS (FAB) m/z 1229.4638 (1229.4652 calcd
for C89H101N22O26, M+ - H).
Compound 3. 1H NMR (500 MHz, D2O) δ 8.43-8.22 (m, 12H),
4.32-4.11 (m, 15H), 3.99-3.82 (m, 12H), 3.77 (s, 2H), 3.68 (m, 6H),
3.01 (t, 4H, J ) 7.5 Hz), 2.94 (t, 2H, J ) 7.6 Hz), 2.77 (m, 4H), 2.66
(t, 2H, J ) 7.2 Hz), 1.90-1.29 (m, 18H); HRMS (FAB) m/z 1891.7175
(1891.7101 calcd for C89H101N22O26, M+ - 2H).
DNAse I Footprinting. T4 polynucleotide kinase, DNAse I, and
[γ-32P]ATP were purchased from Amersham. Electrophoretic reagents
were purchased from J. T. Baker, Inc. The 78-mer DNA was
synthesized on an Applied Biosystems 381A DNA synthesizer using
the phosphoramidite method. The oligonucleotides were cleaved from
the column material by concentrated ammonium hydroxide and
deprotected overnight at 55 °C. The single stranded DNA fragments
were purified by a 12% preparative denaturing polyacrylamide gel.
Purified oligonucleotides were 5′-labeled with [γ-32P]ATP and T4
polynucleotide kinase and annealed with their complementary strands
to make duplex DNA. The labeled duplex DNA was purified using
native 8% PAGE.
Compound 4. 1H NMR (500 MHz, D2O) δ 8.4-8.1 (m, 16H),
4.38-4.1 (m, 20H), 3.98-3.70 (m, 18H), 3.77 (s, 2H), 3.55 (m, 8H),
3.02 (t, 6H, J ) 7.5 Hz), 2.96 (t, 2H, J ) 7.5 Hz), 2.75 (m, 6H), 2.66
(t, 2H, J ) 7.3 Hz), 1.90-1.3 (m, 24H); MS (FAB) m/z 2554 (2554
calcd for C120H136N30O35, M+ - 4H).
Unwinding Angles. The unwinding angles for compounds 1-4
were determined with a Cannon-Ubblehode dilution viscometer (size
75) by the method of Reve´t et al.31 Flow times were measured in
triplicate with a hand-held digital stopwatch; the average deviation of
a set of measurements was 0.5-0.7 s or 0.5%. The 11 kilobase-pair
plasmid pMC1403, a donation of professor Ian Molineux at the
University of Texas, was amplified in E. coli DH5R cells and purified
by CsCl density gradient centrifugation in the presence of propidium
iodide. All unwinding experiments were performed in TE buffer with
50 mM NaCl at pH 7.5. The slope of the Vinograd plot for ethidium
bromide was considered to represent an unwinding angle of 26°. To
determine the unwinding angle of compound 1, the slope of its Vinograd
plot was compared with that of ethidium bromide. For compounds
2-4, titrations performed at multiple DNA concentrations, from 0.15
to 0.25 mM, gave identical unwinding equivalence points, confirming
that these compounds are completely bound at these concentrations.
DNA Lengthening. Titrations were performed with the viscometric
technique described above, and the data were treated by the method of
Cohen and Eisenberg.30 Calf thymus DNA (Sigma, Type XV) was
sonicated in a low power sonicator on ice for 1.5 h, filtered through a
0.45 micron nylon filter, and dialyzed against TE buffer with 50 mM
NaCl using 10 000 MWCO dialysis tubing. The DNA concentration
was 0.1 mM base pairs. As determined in the supercoiled DNA
unwinding experiments and in titrations with excess CT DNA,
compounds 2-4 were completely bound to the DNA at the concentra-
tions studied; therefore, lengthening data for these compounds was not
corrected for binding constant effects. For monomer 1, a binding
constant of 1.0 × 105 (Vide infra) was used to calculate the value r for
each titration.
The 5′-32P-labeled 78-mer DNA (500 nM base pairs) was incubated
with various concentrations of compounds 1-4 for 30 min at room
temperature in 20 µL of a solution containing 20 mM sodium phosphate
(pH 7.5) and 2 mM MgCl2. Then 0.1 U of DNAse I was added to the
reaction mixture and incubated for 1 min. The reaction was quenched
by addition of 180 µL solution containing 3 mM EDTA, 0.3 M sodium
acetate, and 5 µg tRNA. The cleaved DNA was ethanol precipitated
and loaded onto a 12% denaturing sequencing gel. The gels were
exposed on X-ray film and/or phosphorimager screens and analyzed
by ImageQuaNT 4.1 software from Molecular Dynamics.
Acknowledgment. This material is based in part upon work
supported by the Texas Advanced Research Program under
Grant No. ARP-147 and the Welch Foundation (Grant F-1188).
We would also like to thank Ms. Jessica Hernandez for synthetic
and analytical help.
Supporting Information Available: Stopped flow associa-
tion and dissociation kinetic traces of compounds 1-4, absor-
bance spectrum of tetramer 4; UV dissociation profiles of
compounds 2-4, and description of calculational methods (11
pages). See any current masthead page for ordering and Internet
access instructions.
Equilibrium Constant. A stock solution of sonicated calf thymus
DNA (1.5 mM base pairs in TE buffer with 50 mM NaCl) was added
JA9706108