Recognition of nine base pair sequences in the minor groove of DNA at subpicomolar concentrations by a novel microgonotropen

Article abstract of DOI:10.1016/S0968-0896(01)00272-3

The dsDNA interactions of the novel microgonotropen L1 have been characterized via spectrofluorometric titrations and thermal melting studies. A microgonotropen consists of a DNA minor groove binding moiety attached to a basic side chain capable of reachi

Full text of DOI:10.1016/S0968-0896(01)00272-3

Bioorganic & Medicinal Chemistry 10 (2002) 241–252
Recognition of Nine Base Pair Sequences in the Minor Groove of
DNA at Subpicomolar Concentrations by a Novel
Microgonotropen
Alexander L. Satz and Thomas C. Bruice*
Department of Chemistry and Biochemistry, University of California, Santa Barbara, CA 93106, USA
Received 13 July 2001; accepted 13 July 2001
Abstract—The dsDNA interactions of the novel microgonotropen L1 have been characterized via spectrofluorometric titrations
and thermal melting studies. A microgonotropen consists of a DNA minor groove binding moiety attached to a basic side chain
capable of reaching out of the minor groove and grasping the acidic DNA phosphodiester backbone. L1 was synthesized employing
solid-phase chemistry. L1 is shown to distinguish nine base pair A/T rich binding sites from sites possessing fewer than nine con-
tiguous A/T base pairs. Further, L1 binds its preferred dsDNA sequences at subpicomolar concentrations. The equilibrium con-
stant for complexation (K1) of a nine base pair A/T rich dsDNA binding site by L1 is roughly 1013
M
ꢀ1. Single base pair A/T ! G/
C substitutions within the nine base pair A/T rich binding site of L1 decreases the equilibrium constant for DNA binding by 1–2
orders of magnitude. The three proplyamine side chains of L1 enhance the agents free energy of binding by more than 5 kcal.
Molecular modeling suggests that L1 adopts a ‘spiral-like’ conformation which fits almost a full turn of the DNA helix. # 2001
Elsevier Science Ltd. All rights reserved.
Introduction
Our interest in the control of gene expression via inhi-
bition of transcription factor (TF) binding has also lead
Studies indicate that minor groove binding agents may
influence the regulation of gene expression by inhibiting
the binding of regulatory proteins to their double-
stranded DNA (dsDNA) binding sites.1,2 Interest in
controlling the expression of specific genes has spurred
efforts toward the development of agents with greater
sequence selectivity. It is expected that those agents
capable of recognizing longer dsDNA sequences will
exhibit the greatest specificity, however those few agents
targeted to longer dsDNA sequences have generally
lacked specificity.3ꢀ12 We recently reported a novel
minor groove binding agent, L3, capable of distin-
guishing nine bp A/T rich binding sites from sites
possessing fewer than nine contiguous A/T base pairs
(Chart 1).13 In contrast, the ‘classical’ minor groove
binding agents Hoechst 33258 (L9) and distamycin (L7)
cannot distinguish between these A/T rich binding
sites.1,14 L3 is proposed to adopt a ‘spiral’ shape which
allows it to conform to almost a full turn of the dsDNA
helix.13
to the development of microgonotropens.15,16 MGT-6a
(L6) is an example of a first generation MGT which was
based upon a tripyrrole polyamide minor groove bind-
ing moiety. To the central pyrrole unit of these MGTs is
attached a basic polyamine chain capable of associating
with the acidic phosphodiester backbone of dsDNA.
The tren polyamine chain of L6 is known to reach into
the major groove of dsDNA and grasp the DNA back-
bone.17 The polyamine tails of the MGTs endow them
with a superior ability to inhibit the binding of tran-
scription factors (TFs) to their dsDNA binding sites in
cell free assays.2,18,19 Second generation MGTs, or
fluorescent MGTs, were based upon a bisbenzimidazole
minor groove binding moiety which emits a significant
fluorescence signal upon complexation of A/T rich
dsDNA.20,21 Thus, interesting aspects of ligand–dsDNA
interactions can be readily investigated employing fluo-
rescence spectroscopy. Again, the polyamine chains of
these agents endow them with a superior ability to
inhibit TF binding in cell free assays.22
We sought to design a DNA binding agent which would
specifically recognize longer DNA sequences while also
possessing the superior TF inhibitory activity of the
*Corresponding author. Tel.: +1-805-893-2044; fax: +1-805-893-
2229; e-mail: tcbruice@bioorganic.ucsb.edu
0968-0896/02/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S0968-0896(01)00272-3

242
A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
We report the binding characteristics of L1, and its tri-
pyrrole polyamide precursor L4. Also, we have investi-
gated L2, a novel analogue of L1 possessing one less
pyrrole unit and its dipyrrole polyamide precursor L5.
We have investigated these compounds via spectro-
fluorometric titrations and thermal melting studies
employing seventeen different oligomeric duplexes. The
oligomeric duplexes 1–4 were employed to determine
the effect of binding site size and sequence on equili-
brium constants for dsDNA complexation by ligands
(Chart 2). The duplexes 5 and 6 were employed to
investigate the sequence selectivities of L4 and L6. The
oligomeric duplexes 7–17 were employed to investigate
the effect of single and double base substitutions of the
type A/T ! G/C within the nine bp A/T rich binding
site of 7 on binding by L1 (Table 5). 7 contains the
TATA box, which in eukaryotes consists of the con-
sensus sequence 50-TATAAAA-30, and is recognized by
the TBP (TATA binding protein).25
Materials and Methods
Materials
Purified DNA oligomers were purchased from the Bio-
molecular Resource Center, University of California at
San Francisco. L8, L9, and 0.05 wt% 3-(trimethylsilyl)-
propionic-2,2,3,3-d4 acid, sodium salt in D2O were pur-
chased from Aldrich Chemical Company and used
without further purification. L7 was purchased from
Sigma. Solvents and most reagents including triisopro-
pylsilane (TIS), dimethylformamide (DMF), dicyclo-
Chart 1.
MGTs. Thus, we replaced the pyrrole N-methyl sub-
stituents of L3 with propylamine chains to yield a novel
microgonotropen (L1). White et al. has recently shown
this novel MGT to be a more potent inhibitor of TF
complex formation (in cell free assays) than L6, L7, or
L8. L1 was also shown to inhibit gene expression in
whole cells, making it the first MGT to possess whole
cell activity.23 The first and second generation MGTs
show little activity in whole cells.24 Thus, it appears that
L1 possesses the ability to traverse the cell membrane
and localize in nuclear DNA.
hexylcarbodiimide
(DCC),
diisopropylethylamine
(DIPEA), acetonitrile (ACN), dimethyl sulfoxide
(DMSO), and hydroxybenzotriazole (HOBt) were pur-
chased from Aldrich Chemical Company. Rink amide
MBHA resin and benzotriazole-1-yl-oxy-tris-pyrroli-
dino-phosphonium hexafluorophosphate (PyBOP) were
bought from Novabiochem. The synthesis of com-
pounds L3,13 18,18 and 2220 have been previously pub-
lished.
DNA binding experiments
All DNA binding experiments employed 10mM potas-
sium phosphate pH 7.0buffer treated with Chelex and
filtered (0.45 mm). Solutions of differing ionic strengths
(m) were prepared by the addition of NaCl to buffer
solutions prior to treatment with Chelex and filtering.
Oligomeric duplexes were annealed and their molar
extinction coefficients determined as previously dis-
cussed.13 Solutions of known ligand concentrations
were prepared via NMR peak integration where the
ligand samples used contained a known quantity of the
internal standard 3-(trimethylsilyl)propionic-2,2,3,3-d4
acid, sodium salt.13,20 Thermal melting curves were
acquired on a Cary 100 Bio UV–vis spectrophotometer
equipped with a temperature programmable cell block.
Data points were taken every 1 ꢁC with a temperature
ramp of 0.5 ꢁC minꢀ1. Thermal melting temperatures
(tm) were determined as described by Marky and
Breslauer.26 ꢀHꢁ values for duplex formation were
Chart 2.

A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
243
determined by van’t Hoff plots (ꢀHꢁ values are
provided in Tables 1 and 2). Gibbs free energies for
ligand binding (ꢀGꢁ) and corresponding apparent
equilibrium constants (K1) for DNA binding were
calculated from tm0 and tm values employing eq (1).27 K1
values are provided in Tables 3 and 5.
the cases of L8 and L9 whose emissions were monitored
at 450nm. The generation of isothermal binding curves
and their fitting to determine K1 values has been
previously discussed.13 In short, a nanomolar con-
centration of dsDNA is titrated with a relatively
Table 2. Melting temperatures (ꢀtm, ꢁC) for L1 Complexes of 7–7a
ꢀ
ꢁ
1
1
1
ꢀlna
ꢀG
Oligomeric duplexesb
ꢀꢀH(kcal) t0m(C) ꢀtm(ꢁC)
ꢀ
¼
þ
ð1Þ
Rn t0m tm
ꢀH ꢀHRtm
5’-GCGGTATAAAATTCGACG-3’ (7)
5’-GCGGCATAAAATTCGACG-3’ (8)
5’-GCGGTGTAAAATTCGACG-3’ (9)
5’-GCGGTACAAAATTCGACG-3’ (10)
5’-GCGGTATGAAATTCGACG-3’ (11)
5’-GCGGTATAGAATTCGACG-3’ (12)
5’-GCGGTATAAGATTCGACG-3’ (13)
5’-GCGGTATAAAGTTCGACG-3’ (14)
5’-GCGGTATAAAACTCGACG-3’ (15)
5’-GCGGTATAAAATCCGACG-3’ (16)
5’-GCGGTATAGGAATTCGCG-3’ (17)
138
159
141
131
135
138
132
143
149
151
128
56
61
59
59
57
55
57
57
56
57
56
21
16
14
13
16
15
15
14
17
17
13
t0m and tm are the melting points of the duplex without
and with the presence of ligand, respectively, and are
provided in Tables 1 and 2. a is the concentration of
ligand free in solution when the temperature of the
solution is equal to the tm of the complex (a was esti-
mated to be half the total ligand concentration). Calcu-
lated ꢀGꢁ and K1 values are at the complex’s tm and at
ionic strengths provided in the text, and are otherwise
at standard state conditions. Although K1 values in
Tables 3 and 4 are not extrapolated to a common tem-
perature, they are still very helpful when trying to
interpret the observed tm values.
aThermal melting curves acquired in 10mM potassium phosphate pH
7.0buffer, 150mM NaCl ( m=0.17). t0m values are melting tempera-
tures for 0.15 mM oligomeric duplex in the absence of ligand. ꢀtm are
differences in melting temperatures for oligomeric duplexes in the
absence and presence of two eq ligand. Standard deviation for ꢀtm
values are roughly Æ 1 ꢁC. ꢀHꢁ values for oligomeric duplexes have
standard deviations of Æ 10% and were determined by van’t Hoff
plots.
Fluorescence spectra were obtained on a Perkin-Elmer
LS50B fluorimeter equipped with a constant tempera-
ture water bath set at 26 ꢁC. Solutions were excited at
345 nm. Emissions were monitored at 475 nm except in
bThe ligand binding site is underlined. Substitution sites are in bold
type and larger font.
Table 3. Apparent equilibrium constants (K1 Â 10ꢀ9, Mꢀ1) for com-
plexation of oligomeric duplexes 1–6. K1 values were determined via
melting curves of ligand–dsDNA complexesa
Table 1. Melting temperatures (ꢀtm, ꢁC) for ligand complexes of 1–
6a
Ligand
None
m
1c
2
3
4
5
6
Ligand
m
1b
2
3
4
5
6
tm0
0.032
44
36
50
45
39 53 58
ꢂ65,000c
ꢂ7600c
1430
120
17
5.7
780
3.7
ꢀꢀHꢁ (kcal)
7373
56
100
7057 96 93
57
111
L1
0.032 1300
tm0
0.17
0.17
0.032
0.17
1.6
5.9
0.19
ꢀꢀHꢁ (kcal)
102
L2
L3
L4
L5
L6
L7
L8
L1
L2
L3
L4
L5
L6
L7
L8
0.032 ꢂ40b ꢂ51b
0.17
0.03
14
22
9
9
0
22
10
9
2
22
7
ꢂ33b
ꢂ38b
23
0.032 0.080
0.17 0.0033
0.032
0.17
0.032 0.080
5.9
4.1
0.29
1.1
1.8
3.3
0.14
0.17
0.032
0.17
0.032
0.17
0.032
0.17
0.032
0.17
0.032
0.17
0.032
0.17
0.29
24
16
36
15
0.048 Dtm=1d
2.2 0.092
0.17 0.0083
22
9
21 16
6
9
1
8
0
0
0.032
0.17
0.032 0.080
5.9
160
0.60
0.40
0.059
1.1
3.0
0.067
0.079
0.050 0.075 0.020 Dtm=0d
0.0092
0.17
0.032
0.17
0.021
0.11
0.021
31
11
13
6
16
5
21
6
23 15
0.18
0.041
0.043 0.074 Dtm=0d
0.037
9
4
10
4
8
2
11
9
4
7
aApparent K1 values are derived from melting curves of ligand–
dsDNA complexes and were calculated employing eq (1) as discussed
in the text. Ligand/dsDNA melting curves acquired in 10mM potas-
sium phosphate pH 7.0buffer with 10mM NaCl ( m=0.032) or 150
mM NaCl (m=0.17) are listed in bold and regular type, respectively.
tm values for oligomeric duplexes and their ligand complexes are pro-
vided in the supporting information. Standard deviations for K1 values
are less than Æ 60%.
12
5
aThermal melting curves acquired in 10mM potassium phosphate pH
7.0buffer, 10( m=0.032) or 150 mM NaCl (m=0.17). t0m values are
melting temperatures for 0.3 mM oligomeric duplex in the absence of
ligand. ꢀtm are differences in melting temperatures for oligomeric
duplexes in the absence and presence of two equiv ligand. Standard
deviation for ꢀtm values are roughly Æ 1 ꢁC. ꢀHꢁ values for oligo-
meric duplexes have standard deviations of Æ 10% and were deter-
mined by van’t Hoff plots.
bSome melting curves went off scale such that their ꢀtm values are
approximate (see Figure 1 for an example).
cDNA sequences are provided in Chart 2 of the manuscript and their
potential A/T rich binding sites are underlined.
bDNA sequences are provided in Chart 2 and their potential A/T rich
binding sites are underlined.
cSome melting curves went off scale such that their K1 values are
approximate (see the L1/2 complex melting curve in Figure 1).
dThere was little difference in tm values between the oligomeric duplex
with and without the presence of ligand (ꢀtm). Thus, K1 values could
not be calculated for these complexes.

244
A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
Table 4. Equilibrium constants (K1 Â 10ꢀ9
M
ꢀ1, 26 ꢁC) for com-
plexation of oligomeric duplexes 1–7. Values were determined via iso-
curve may be generated by plotting fluorescence signal
versus the concentration of ligand free in solution ([L]f)
(Fig. 3). The [L]f term is calculated employing eq (2)
where n is the stoichiometry of the ligand/dsDNA com-
plex. For 1:1 and 2:1 stoichiometries, eqs (3) and (4) are
employed to fit the isothermal binding curve, respec-
thermal binding curvesa
Ligand mb EtOH
(%)
1c
2
3
4
7
L1
L3
L4e
0.17
0.77
0.17
0.77
0.17
0.77
0.17
0.17
0.17
0
25
ꢂ 5000d
ꢂ10,000d
5.4
tively. [L]Bound is the concentration of ligand bound to
Èf is the total fluorescence signal of the
P
0.028
2.6
dsDNA.
1.9g
0.024g
solution upon saturation of ligand binding sites. For
2:1 complexes, the equilibrium constants are given as
the product of K1K2 if separation of the individual
binding constants is not possible (due to similarity in
magnitude).
0
25
0
0
0
65
360
1.6
0.032
0.18
0.15
0.25
L6e
L7e
L8
K1=0.9f
K2=0.004f
6.0 0.045
0.77
0.17
25
0
0.0063
K1=0.050f
K2=0.008f
X
L9
½LꢁBound
F ¼
F ¼
Èf
Èf
ð2Þ
ð3Þ
n½DN AꢁT
aK1 values determined by nonlinear least squares fitting of isothermal
binding curves at 26 ꢁC employing eqs (3), (4), or (5). The stated
equilibrium constants have a standard deviation of Æ 50%.
bK1 values acquired in 10mM potassium phosphate pH 7.0buffer with
150mM NaCl ( m=0.17) are listed in bold type. K1 values acquired in a
solution of 75% 10mM potassium phosphate pH 7.0buffer with 1 M
NaCl and 25% ethanol (m=0.17) are listed in regular type.
cDNA sequences provided in Chart 2.
ꢀ
ꢁ
X
X
K1½Lꢁf
1 þ K1½Lꢁf
!
2
0:5K1½Lꢁf þK1K2½Lꢁf
1 þ K1½Lꢁf þK1K2½Lꢁf
F ¼
Èf
ð4Þ
2
dK1values estimated as discussed in text.
eK1 values determined via spectrofluorometric competition assays with
L8.
Competition assays were employed in an effort to
determine K1 values for L4, L6, L7, and in some cases
L1. 5 nM dsDNA, in the presence of a large excess of L8
(100–2000 nM) was titrated with ligand and decreases in
the solutions fluorescence signal measured as L8-
dsDNA complexes were replaced with ligand–dsDNA
complexes possessing lower quantum yields of fluores-
cence. Plots were made of the negative change in fluo-
fComplex stoichiometries are 2:1.
gThe L3–7 complex has a stoichiometry of 2:1.13 The K1 and K2 terms
for the L3–7 complex cannot be separated due to their similarity in
magnitude. Approximate K1 values in units Mꢀ1 are provided by tak-
ing the square root of the multiplied through K1K2 Mꢀ2 value.
Table 5. Apparent equilibrium constants (K1 Â 10ꢀ9
M
ꢀ1) for com-
plexation of oligomeric duplexes 7–17 by L1. K1 values were deter-
mined via melting curves of L1–dsDNA complexesa
rescence signal (ꢀꢀF) at 450nm versus [L] , where [L]f
f
was calculated according to eq (2) assuming n=1 (see
Fig. 4). The isothermal binding curves were fit employ-
ing eq (5).
Oligomeric duplexesb
K1
5-GCGGTATAAAATTCGACG-3 (7)
5-GCGGCATAAAATTCGACG-3 (8)
5-GCGGTGTAAAATTCGACG-3 (9)
5-GCGGTACAAAATTCGACG-3 (10)
5-GCGGTATGAAATTCGACG-3 (11)
5-GCGGTATAGAATTCGACG-3 (12)
5-GCGGTATAAGATTCGACG-3 (13)
5-GCGGTATAAAGTTCGACG-3 (14)
5-GCGGTATAAAACTCGACG-3 (15)
5-GCGGTATAAAATCCGACG-3 (16)
5-GCGGTATAGGAATTCGCG-3 (17)
970
180
18
ꢀ
ꢁ
X
K1½Lꢁf
ꢀꢀF ¼
Ff
ð5Þ
1 þ K1½Lꢁf þKHt½Htꢁf
5.6
43
The addition of L6 causes a slight enhancement in the
fluorescence emission of L8 free in solution. Thus, the
‘background’ signal acquired from titration of L8
(without the presence of dsDNA) with L6 was sub-
tracted from the agent’s spectrofluorometric titration
assays.
33
20
23
230
250
5.4
aApparent K1 values are derived from melting curves of ligand–
dsDNA complexes which were acquired in 10mM potassium phos-
phate pH 7.0buffer with 150mM NaCl ( m=0.17). K1 values were
calculated from melting curves employing Eq. (1) as discussed in the
text. tm values for oligomeric duplexes and their ligand complexes are
provided in the supporting information. Standard deviations for K1
values are less than Æ60%.
Model building
The molecular modeling program SYBYL was used to
construct plausible ligand–dsDNA complexes. For the
L1–dsDNA complex the agent was docked into the A/T
rich minor groove of a model B-DNA helix. The
dsDNA complexes of L4 and L6 were derived from X-
ray crystal structure coordinates.28
bA/T rich ligand binding sites are underlined. Substitution sites are in
bold type and larger font.
concentrated solution of ligand dissolved in dimethyl
sulfoxide. Formation of the ligand/dsDNA complex is
monitored via the solutions fluorescence emission. If the
total concentration of dsDNA employed ([DNA]T) is
sufficiently small compared to the ligand’s K1 value,
[DNA]T=1/K1, then a meaningful isothermal binding
Organic synthesis: general
1H and 13C NMR spectra were obtained on a Varian
Unity Inova 400 spectrometer at 400 and 100 MHz,
respectively. TLC was carried out on silica gel (Kie-
selger 60F254) glass backed commercial plates and

A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
245
visualized by UV light. Fast atom bombardment mass
spectra, HRMS and LRMS, were obtained on a VG
analytical, VG-70E double focusing mass spectrometer,
with an Ion Tech Xenon Gun FAB source, and an
OPUS/SIOS data interface and acquisition system.
High-pressure liquid chromatography was accom-
plished using a Hewlett-Packard Series 1050 HPLC
equipped with a diode array detector. For preparative
separations an Alltech Macrosphere 300A, C8, silica, 7
mm, 250 Â 10mm reverse-phase column was used. For
analytical separations an Alltech Macrosphere 300A,
C18, silica, 7 mm, 250 Â 4.6 mm reverse-phase column
was used.
purified by flash chromatography (silica, DCM) to give
1
20 (767 mg, 92%, white powder). H NMR (CDCl3) d
1.44 (s, 9H, tBoc), 2.02 (m, 2H, Ar–C–CH2-C–), 3.19
(m, 2H, –CH2–N–tBoc), 4.42 (t, J=7.1 Hz, Ar–CH2–),
4.75 (bm, 2H, –NH–tBoc), 7.77 (s, 1H, ArH), 7.94 (s,
1H, ArH); 13C NMR (CDCl3) d 28.44 (-tBoc), 31.83
(Ar–C–C–C–), 37.49 (Ar–C–C–C–), 48.33 (Ar–C–C–C–),
following 4 signals were detected in the aromatic region
116.31+118.95+129.43+136.19, 155.79 (–N–C(=0)–),
156.48 (–Ar–C(C=0)–). HRMS (ESI) 502.1017, M+
Na+ (502.1014 calcd for C19H18F5N3O+6 Na).
Pentafluorophenyl 1-[3-[N-(tert-butyloxycarbonyl)ami-
no]propyl]-4-[(9-fluorenylmethoxycarbonyl)amino]-2-pyr-
rolecarboxylate (21). Compound 20 (200 mg, 0.41
mmol) was dissolved in 30mL 1:1 ethyl acetate/ethanol,
5 mL dioxane, and 2 mL saturated NaHCO3. To this
colloidal solution was added 25 mg of 10% palladium
on carbon catalyst. The solution was stirred under an
atmosphere of hydrogen gas (ꢂ1 atm) for 24 h. Reac-
tion progress was monitored by TLC. The solution was
filtered through Celite and the filtrate reduced to a
volume of ꢂ 5–10mL. To the reduced filtrate was then
added 5 mL dioxane, 2 mL saturated NaHCO3, and
FMOC-Cl (127 mg, 0.49 mmol). The solution was stir-
red for 24 h. Product was extracted into diethyl ether,
washed with water, and purified by flash chromato-
graphy (silica, 20:1 DCM/ethyl acetate) to give 21 (100
mg, 36%, white foam). 1H NMR (CDCl3) d 1.43 (s, 9H,
tBoc), 1.94 (m, 2H, Ar–C–CH2-C–), 3.1 (m, 2H,–CH2–
N–tBoc), 4.2–4.3 (m, 3H, Ar–CH2– and Ph2–CHR–),
4.51 (d, 2H,–CH2OC(C¼O)N–Py), 4.69 (bm, 2H,–NH–
tBoc), 6.66 (s, 1H, ArH), 7.03 (s, 1H, ArH) 7.31 (m, 2H,
FMOC ArH), 7.40(t, 2H, FMOC ArH), 7.61 (d, 2H,
FMOC ArH), 7.78 (d, 2H, FMOC ArH); 13C NMR
(CDCl3) d 28.56 (-tBoc), 32.02 (Ar–C–C–C–), 37.72
Solid-phase synthesis (SPS): general
SPS synthesis was accomplished using MBHA resin and
standard manual solid-phase FMOC techniques.29
Coupling reactions for 21 were accomplished using 1.6–
2 equiv of 21, 2 equiv HOBt, and 4 equiv of DIPEA in
anhyd DMF and were run for 24 h. High coupling
yields (70–100%) were measured by absorption at 290
nm of deprotected FMOC after resin was treated with a
20% piperidine/DMF solution. After each coupling,
unreacted terminal amines were capped with a DMF
solution of acetic anhydride and triethylamine. Coupling
reactions for 22 (Scheme 1) were accomplished employ-
ing 2.6–3 equiv of 22, 5.2–6 equiv PyBOP, 2 equiv
HOBt, and 6 equiv DIPEA and were run for 24 h. Resin
cleavage was accomplished in 2–4 h using a 95% TFA,
2.5% water, and 2.5% TIS solution. All final products
synthesized via SPS were purified by HPLC chroma-
tography (silica, reverse phase) with an increasing
gradient of acetonitrile in 0.1% aq TFA solution.
Product purity was checked by analytical HPLC analysis.
Product was lyopholized and then reconstituted in a
minimal amount of methanol. Product precipitated out
of solution by the addition of diethyl ether followed by
bubbling of the colloidal solution with HCl (g). Product
was then collected via centrifugation as the HCl salt,
reconstituted in H2O, and lyopholized to dryness.
(Ar–C–C–C–),
46.96+47.33
(Ar–C–C–C–+–
COC(=O)N-Py), 67.28 (–C–COC(=O)N–Py), the fol-
lowing 6 signals were detected in the aromatic region
120.27+125.13+127.35+128.03+141.56+143.87, two
signals belonging to carbonyl carbons were detected, a
weak signal at 153.86 and a more intense signal at
156.25 ppm; HRMS (FAB) 671.203 (671.205 calcd for
C34H30F5N3O6).
Pentafluorophenyl 1-[3-[N-(tert-butyloxycarbonyl)ami-
no]propyl]-4-nitro-2-pyrrolecarboxylate (20). To 6 mL
solution (3:2:1, 1 M NaOH(aq)/dioxane/etoh) was added
18 (580mg, 1.7 mmol). The solution was heated to
60 ꢁC and stirred for 30min. The disappearance of
starting material 18 was monitored by TLC (silica,
DCM). The solution was acidified with dilute HCl until
a white precipitate appeared (pH ꢂ 2). Compound 19
was extracted out of the aq solution with ethyl acetate.
The combined fractions of organic solvent were washed
with water, dried over Na2SO4, and evaporated off to
give a thick bright yellow paste/solution. The triethyla-
mine salt of 19 has been previously characterized.18
Without further purification, 19 was dissolved in 10mL
DMF to which was added DCC (421 mg, 2.0mmol) and
pentafluorophenol (626 mg, 3.4 mmol). The solution
was stirred overnight at room temperature. Solvent was
removed under vacuum and the remaining paste re-dis-
solved in diethyl ether after which reacted DCC (dicy-
clohexylurea) was removed by filtration. The diethyl
ether was evaporated and the resulting crude product
L5. Synthesis employed MBHA rink amide resin (18
mg, 0.009 mmol loading sites). L4 was collected as its
1
HCl salt (0.0036 mmol, 40%, white powder). H NMR
(D2O) d 2.10–2.14 (two overlapping signals, 7 H, –
C(C¼O)–CH3)+Npy–C–CH2–C–), 2.92 (t, J=7.02 Hz,
4 H, Npy–C–C–CH2–), 4.37 (m, 4H, Npy–CH2–C–C–),
6.79 (m, 1H, ArH), 6.86 (m, 1H, ArH), 7.22 (m, 1H,
ArH), 7.27 (m, 1H, ArH). UV spectrum: lmax=305,
l
min=262 nm; LRMS (ESI) 391 (M+H)+.
L2. Synthesis employed MBHA rink amide resin (18
mg, 0.009 mmol loading sites). Product was collected as
1
its HCl salt (0.0028 mmol, 31%, yellow powder). H
NMR (D2O+DMSO–d6) 1.80+1.92 (two sets of mul-
tiplets, 2H/multiplet, Npy–C–CH2–C–N), 2.13 (m, 2H,
Ph-O–C–CH2–C–), 2.39 (bt, 2H, Ph–O–C–C–CH2–),
2.67+2.80(two sets of triplets, signal at 2.67 obscured
by solvent, signal at 2.80integrates for 2H, N py–C–C–

246
A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
CH2–N), 3.00 (s, 3H, CH3–NR2), 3.1–3.3 (m, obscured
by solvent, Ph–N(C–CH2–)2), 3.70+3.85 (two sets of
broad signals, 6H, Ph–N(CH2–)2+Ph–O–CH2–),
3.98+4.18 (two sets of broad signals, 4H, Npy–CH2–),
signals detected between 6.5 and 8 ppm are due to ArH
protons, 6.53 (d)+6.77 (s)+ 6.98 (m)+7.04 (s)+7.10
(s)+7.26 (m)+7.35–7.42 (m)+7.56 (d)+7.63 (s)+7.75
by reaction of the free amine with FMOC-Cl gave the
activated monomer 21. Coupling yields between pyrrole
units of 21 were observed to be between 70 and 100%.
Pyrrole monomer coupling reactions were monitored by
UV absorbance of the deprotected FMOC. The tradi-
tional Kaiser test is not compatible with the aromatic
amine of the pyrrole ring.30 Synthesis of L2 and L1
proceeded via coupling of 22 to the terminal amines of
di- and tripyrrole polyamides, respectively. Capping of
unreacted amines with acetic anhydride after each cou-
pling reaction greatly simplified the purification of final
products, particularly for L1. Coupling yields for the
addition of 22 were estimated to be in the range of 30–
40% by HPLC chromatography of resin cleaved pro-
duct.
(s); UV spectrum:
shoulder at 350nm,
(M+H)+.
l
l
max=308 with
a distinctive
min=280nm; LRMS (ESI) 841
L4. Synthesis employed MBHA rink amide resin (21
mg, 0.010 mmol). Product was collected was as its HCl
salt (0.0017 mmol, 17%, white product). L5, a bypro-
duct of this reaction, was also purified and collected
1
(0.0015 mmol, 15%, white powder). H NMR (D2O) d
2.0(two overlapping signals,
9
H,–C(C ¼O)–
Apparent equilibrium constants for complexation of 1–6
were determined via melting curves of ligand–dsDNA
complexes (Table 3)
CH3)+Npy–C–CH2–C–), 2.78 (m, 6 H, Npy–C–C–CH2–
), 4.37 (bm, 6H, Npy–CH2–C–C–), 6.92 (m, 2H, ArH),
7.04 (m, 1H, ArH), 7.30 (m, 1H, ArH), 7.38 (m, 1H,
ArH), 7.41 (m, 1H, ArH). UV spectrum: lmax=305,
Melting curves were acquired in 10mM potassium
phosphate pH 7.0buffer at 10( m=0.032) or 150
(m=0.17) mM NaCl concentrations. Representative
melting curves for ligand–dsDNA complexes acquired
at m=0.17 are presented in Figure 1. Observed tm values
were converted to apparent equilibrium constants (K1,
l
min=262 nm; LRMS (ESI) 556 (M+H)+.
L1. Synthesis employed MBHA rink amide resin (46
mg, 0.023 mmol). Product was collected as its HCl salt
(0.0036 mmol, 16%, yellow powder). 1H NMR
(D2O+DMSO-d6) 1.95+2.08 (two sets of overlapping
multiplets, 6H, Npy–C–CH2–C–N+Ph–O–C–CH2–C–),
2.46 (obscured by solvent, Ph–O–C–C–CH2–), 2.72 (m,
6H, Npy–C–C–CH2–N), 2.88 (s, 3H, CH3–NR2),
3.05+3.20 (two sets of multiplets, 2H/multiplet, Ph–
N(C–CH2–)2), 3.55+3.89 (two sets of multiplets, 2H/
multiplet, Ph–N(CH2–)2), 4.15 (obscured by HOD, Ar–
O–CH2–), 4.32 (broad multiplet, 6H, Npy–CH2–), sig-
nals detected between 6.8 and 9 ppm are due to ArH
protons, 6.86 (s)+6.95 (s)+ 7.05 (s)+7.15 (bd)+7.24
M
ꢀ1) for DNA binding employing eq (1).27
L1 is shown to complex a nine bp A/T rich binding site
at subpicomolar concentrations. The K1 value for
complexation of 2 by L1 is 8 Â 1012 Mꢀ1 at
m=0.17 (the oligomeric duplex 2 has the binding
site -AAAAAAAAA-). Additionally, L1 is shown to
distinguish between the nine bp binding site of 2 and the
shorter five bp A/T rich site of 1 (the oligomeric duplex
1 has the binding site -AAATT-). For L1, K1 for com-
plexation of 2 is 5000-fold greater than for 1. In com-
(s)+7.28–7.38
(bm)+7.39
(s)+7.53
(t)+7.71
(bd)+7.80–7.98 (bm)+8.07 (bd)+8.49 (s); UV spec-
trum: lmax=320with a slight shoulder at 355 nm,
l
min=278 nm; LRMS (ESI) 1006 (M+H)+.
Results
Synthesis
Synthesis of final products was accomplished in a step-
wise manner from MBHA rink amide resin by employ-
ing FMOC chemistry and standard manual solid-phase
synthetic techniques (Scheme 1).29 It was previously
found that N-methyl pyrrole carboxylic acid derivatives,
analogous to 19, do not reduce cleanly via catalytic
hydrogenation.13 However, N-methyl pyrrole penta-
fluorophenyl esters, analogous to 20, do reduce cleanly
without polymerization during catalytic hydrogenation
or during carbamoylation with FMOC-Cl. Still, in the
presence of HOBt catalyst these pyrrole penta-
fluorophenyl esters undergo coupling reactions via solid
phase synthesis to form di- and tripyrrole polyamides.
Thus, the activated monomer 21 was synthesized start-
ing with the functionalized pyrrole 18. The pyrrole 18
was hydrolyzed to give 19 and esterified to give 20.
Reduction of the nitro functional group of 20 followed
Scheme 1. (i) NaOH, 60 ꢁC, H2O; (ii) pentafluorophenol, DCC; (iii) a.
H2(g) 1 atm, 10% Pd/C, b. FMOC-Cl, NaHCO3; (iv) deprotection,
20% piperizine in DMF; (v) 21, HOBt, DIPEA; (vi) a. capping with
acetic anhydride, triethylamine, b. resin cleavage in 95% TFA; (vii) a.
22, PyBOP, HOBt, DIPEA, b. resin cleavage.

A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
247
same change in ionic strength increases K1 values for
complexation of 2 by L7 and L8 to a lesser extent, 7-
and 30-fold, respectively. Notably, at m=0.032 com-
plexation of 2 by L1 occurs at femtomolar concentra-
tions (K1 is 6 Â 1013
M
ꢀ1). However, at this low ionic
strength the K1 for complexation of 2 by L1 is only 50-
fold greater than for complexation of 1, compared to a
5000-fold difference when the ionic strength is m=0.17.
Thus, the sequence specificity of L1 is sensitive to
changes in ionic strength.
Apparent equilibrium constants for complexation of 1–4
were determined via isothermal binding curves. Determi-
nation of K1 values for L1 was complicated by its very
large magnitude. Even at nanomolar dsDNA con-
centrations, the lower limit for spectrofluorometric
titrations, the titration of oligomeric duplexes [particu-
larly 2 and 7 in 10mM potassium phosphate buffer pH
7.0with 150mM NaCl ( m=0.17) and 26 ꢁC] with L1
gave plots unusable for determination of K1 (see Chart 2
for DNA sequences). This is because the plots showed a
linear relationship between fluorescence intensity and
ligand concentration up to the point of binding site
saturation. However, the titrations did show that L1
formed 1:1 complexes with both 2 and 7.
Figure 1. Thermal melting curves for oligomeric duplex 2 (0.3 mM)
and its ligand complexes (0.6 mM ligand) (10mM potassium phos-
phate pH 7.0buffer, 150mM NaCl ( m=0.17)). Plots show change in
absorption units versus ꢁC. 1 is dsDNA with no ligand, 2 is dsDNA
with L8, 3-L7, 4-L6, 5-L4, 6-L3, 7-L2, and 8-L1.
parison to classical minor groove binders, K1 for com-
plexation of 2 by L1 is 5 orders of magnitude greater
than for complexation by L7 or L8. Also, as judged by
K1 values, neither L7 or L8 can distinguish between the
‘small’ five bp binding site of 1 and the ‘large’ nine bp
site of 2.
The propylamine chains of L1 are shown to greatly
enhance its K1 values for DNA binding while not
adversely affecting the sequence specificity of the agent.
At m=0.17, K1 for complexation of 2 by L3 is roughly 3
orders of magnitude less than for complexation by L1
(L3 lacks the propylamine chains of L1). Also for L3,
K1 for complexation of 2 is 1700-fold greater than for 1.
Thus L1 and L3 possess the same magnitude of specifi-
city for the longer A/T rich binding site of 2 relative to
the 5 bp binding site of 1.
A competition assay was conducted between L8 and L1
in an attempt to acquire a usable isothermal binding
curve. We chose to employ L8 in competition assays
because it possesses a less intense fluorescence signal
when free in solution than its more widely employed
analogue L9.31 Although both L1 and L8 fluorescence
in the 400–500 nm range when bound to dsDNA, the
quantum yield of fluorescence for L8 is greater than that
of L1. Thus, titration of dsDNA in the presence of a
large excess of L8 with L1 leads to a decrease in emis-
sion signal. However, even in the presence of 400 equiv
of L8, where the solution contained 5 nM 2, 2000 nM
L8, and m=0.17, the titration showed a linear relation-
ship between fluorescence signal and L1 concentration
(Fig. 2). This linear relationship indicates that K1 for com-
plexation of 2 by L1 is much greater than 400 times the K1
for complexation of 2 by L8. Thus, K1 for complexation
of 2 must be much greater than 4 Â 1011 Mꢀ1. Attempts
L2 possesses one less pyrrole subunit than L1. The K1
for complexation of 2 by L2 is ꢂ 60-fold less than for
complexation by L1. Also for L2, K1 for complexation
of 2 is 630-fold greater than for 1, indicating that L2
possesses less specificity than L1. The tripyrrole poly-
amide microgonotropen L4 lacks the bisbenzimidazole
moiety of L1. At m=0.17, K1 for complexation of 2 by
L4 is 2100-fold less than for complexation by L1 but
still roughly 70-fold greater than for complexation by
either of the classical minor groove binders L7 or L8. L4
shows only a slight preference for complexation of 2
relative to 1, indicating that unlike L1, L4 shows little
preference for ‘longer’ A/T rich binding sites. However,
L4 is still specific for A/T rich binding sites relative to
G/C rich dsDNA. For L4 at m=0.032, K1 for com-
plexation of 5 is 24-fold greater than for 6. The
oligomeric duplex 5 contains an A/T rich binding
site (-AATT-) whereas 6 does not contain an A/T rich
binding site. L5 possesses one less pyrrole subunit than
L4. At m=0.17, K1 for complexation of 1 by L5 is
equivalent to K1 values for DNA binding by the classi-
cal minor groove binders L7 and L8.
Figure 2. Titration of 5 nM 2 in the presence of 2000 nM L8 with L1
[10mM potassium phosphate pH 7.0buffer, 150mM NaCl ( m=0.17),
26 ꢁC]. Plot shows the negative change in the fluorescence signal at 450
nm versus the concentration of L1. The two straight lines intersect at
an L1–2 stoichiometry of 1:1.
Decreasing the ionic strength from m=0.17 to m=0.032
increased the K1 values for complexation of 2 by L4 and
L6 by 210- and 270-fold, respectively. In contrast, the

248
A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
were made to conduct competition assays employing
even larger excesses of L8 in an attempt to generate a
usable isothermal binding curve, however, L8 slowly
precipitates out of solution at greater concentrations.
High salt concentrations and the presence of ethanol are
both known to weaken DNA–L9 interactions.32 A ‘high
salt–25% ethanol’ solution where m=0.77 (75% 10 mM
potassium phosphate pH 7.0buffer, 1 M NaCl and 25%
ethanol) was found to lessen K1 values for DNA binding
by L1 to magnitudes measurable via spectro-
fluorometric titration (Fig. 3 and Table 4). K1 values at
26 ꢁC are provided in Table 2. The extent to which the
‘high salt–25% ethanol’ solution (m=0.77) decreases K1
values for L1 relative to ‘normal’ pH 7.0buffer where
m=0.17 was estimated by determining K1 values for L4
in both solutions (the tripyrrole moiety of L1). Upon
going from the ‘normal’ to the ‘high salt–25% ethanol’
solution, the K1 for complexation of 1 by L4 decreased
by 2030-fold. Applying this value to L1, the K1 for
complexation of 2 by L1 in ‘normal’ pH 7.0buffer
Figure 3. Titration of 2 nM 2 with L1 in a solution of 75% 10mM
potassium phosphate pH 7.0buffer, 1 M NaCl with 25% ethanol
(m=0.77), 26 ꢁC. Fluorescence signal at 475 nm versus concentration
of L1 free in solution as calculated by eq (2). The data have been fit by
a nonlinear least squares fitting routine employing eq (3) to determine
the equilibrium constant for complex formation (R2=0.991).
where m=0.17 is estimated to be ꢂ5 Â 1012 Mꢀ1
,
roughly 3 to 4 orders of magnitude greater than for
complexation by L8 and L9. Also, in ‘normal’ buffer the
K1 for complexation of 1 by L4 was found to be ꢂ260-
fold greater than for complexation by L6, L7, or L8 (see
Figure 4 for a representative competition assay).
The two minor groove binding moieties of L1 may
complex dsDNA with positive or negative cooperativity
with respect to one another. In an attempt to char-
acterize the cooperativity between the two moieties,
thermodynamic cycles such as shown in Figure 5 were
constructed. Thermodynamic cycles at 26 ꢁC were con-
structed for complexation of 2 in both ‘normal’ pH 7.0
buffer where m=0.17 (cycle presented in Fig. 5) and the
‘high salt–25% ethanol solution’ where m=0.77 (cycle
not shown). Free energies of dsDNA binding for the
tripyrrole moiety of L1 were estimated via K1 values for
complexation of 1 by L4 (Table 4). Similarly, free ener-
gies of binding for the bisbenzimidazole moiety were
estimated via K1 values for complexation of 1 by L8.
Further, the ability of two minor groove binding mole-
cules to simultaneously occupy the minor groove of 2
was investigated by the determination of K1 and K2
values for complexation of 2 by L8 and L9. For both
ligands, K1 was roughly two orders of magnitude
greater than K2.
Figure 4. Titration of 5 nM 4 in the presence of 100 nM L8 with L4 ([0
mM potassium phosphate pH 7.0buffer, 150mM NaCl ( m=0.17),
26 ꢁC]. Negative change in fluorescence signal at 450nm versus con-
centration of L4 free in solution as calculated by eq (2) with n=1. The
data have been fit by a non-linear least squares fitting routine
employing eq (5) to determine the equilibrium association constant for
complex formation (R2=0.993).
K1 values for complexation of 3 (-AAATT-) and 4
(-TATAA-) by L6 and L8 are also provided in Table 2.
The five bp A/T rich binding sites of 3 and 4 are repre-
sentative of either side of the nine bp A/T rich binding
site of 7 (for dsDNA sequences see Chart 2). L6 and L8
show the same type of selectivity, with the magnitude of
their K1 values for DNA binding in the order 3 > 1 >
4. Both had K1 values for complexation of 3 two orders
of magnitude greater than for complexation of 4. Inter-
estingly though, the K1 values derived at elevated tem-
peratures via thermal melting curves (Table 3) suggest
little or no selectivity by either ligand. It is reported that
the stacking of A-tracts is disrupted by temperature
(see the discussion section for a brief description of
Figure 5. Schematic thermodynamic cycle showing the formation of
the L1/2 complex. ꢀGꢁ values (kcal) are calculated from equilibrium
constants for DNA binding determined via spectrofluorometric titra-
tions provided in Table 2 [10mM potassium phosphate pH 7.0buffer,
150mM NaCl ( m=0.17), 26 ꢁC]. (A) unbound L1 (B) 2 bound by the
tripyrrole moiety of L1. ꢀGꢁ estimated from K1 for complexation of 1
by L4. (C) 2 bound by the bisbenzimidazole moiety of L1. ꢀG esti-
mated from K1 for complexation of 1 by L8. (D) experimentally
determined ꢀG value for complexation of 2 by L1. (E) Hypothetical
ꢀG value for noncooperative complexation of 2 by L1. The difference
in ꢀGꢁ between D and E is the magnitude of the ligand’s negative
cooperativity.

A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
249
A-tracts), and that A-tracts undergo a pre-melting
transition around 30–37 ꢁC to a structure little different
from that of ‘normal’ B-DNA (see ref 33 and references
therein). Thus, it would appear that the effect of 5-TpA-
3 steps on minor groove binding cannot be detected via
analysis of thermal melting curves when the DNA
duplexes employed have t0m values much greater than
30 ꢁC. The ligand complexes of 3 and 4 melt at tem-
peratures between 40and 50 ꢁC (see Table 1 for tm
values).
The longer side chain of L6 is capable of reaching a
greater distance beyond the phosphodiester backbone
than the shorter propylamine side chains of L4. When
the propylamine chains of L4 are fully extended, the
N–P distances between the propylamine nitrogens and
the DNA phosphorous atoms are at most 6–7 A. Also,
the three propylamine chains of L4 are spaced at
intervals equivalent to the P–P distances of the DNA
backbone. Alternatively, N–P distances for L6 average
ꢂ11 A (for the three nitrogen atoms N1, N2, and N3 as
labeled in Chart 1). Also, relatively short distances
between these three nitrogen atoms prohibits them
from simultaneously associating with three different
phosphate groups.
Apparent equilibrium constants for complexation of 7–17
by L1 were determined via melting curves of L1–dsDNA
complexes (Table 5). The ability of L1 to distinguish
between a nine bp A/T rich binding site, as for 7, and
related binding sites containing a single A/T!G/C bp
substitution, as for oligomeric duplexes 8–16, was
investigated. Apparent K1 values are derived from L1–
dsDNA complex melting curves employing eq (1) (see
Table 5 for K1 values, tm values for L1–dsDNA com-
plexes are provided in the supporting information).
Apparent K1 values decrease substantially, 20- to 200-
fold, for single bp mismatch duplexes 9–14. Substitu-
tions at either end of the A/T rich binding site (8, 15,
and 16) have little effect on K1. The double bp substitu-
tion in 17 causes a 180-fold decrease in K1.
Discussion
Spectrofluorometric and thermal denaturation experi-
ments were employed to investigate stoichiometries and
equilibrium constants (K1) for formation of ligand–
dsDNA complexes. L1 is observed to form 1:1 com-
plexes with oligomeric duplexes 2 and 7 at subpicomolar
concentrations (see Chart 2 for DNA sequences). Both 2
and 7 contain binding sites of nine contiguous A/T base
pairs. Also, L1 possesses the ability to distinguish
between its preferred nine bp A/T rich binding sites and
shorter sites with fewer than nine contiguous A/T bases
pairs. For L1, the K1 for complexation of 2 is 5000-fold
greater than for complexation of 1 whereas the classical
minor groove binders L7 and L8 demonstrate no sig-
nificant preference for either duplex. The specificity of
L1 was further investigated employing the oligomeric
duplexes 8–16, each of which contains a single A/T !
G/C bp substitution within the -TATAAAATT- bind-
ing site of 7 (Table 5). Single bp substitutions at either
termini of the A/T rich binding site of 7 had little effect
on the agents K1 values. However, substitutions else-
where (for duplexes 9–14) led to significant 20- to 200-
fold drops in K1.
Equilibrium constants for complexation of 7 by L1
and L3 were determined via isothermal binding curves
(Fig. 3)
As mentioned previously, K1 values for L1–dsDNA
complexes where determined in a ‘high salt–25% etha-
nol’ solution where m=0.77 (75% 10 mM potassium
phosphate pH 7.0buffer, 1 M NaCl and 25% ethanol)
because K1 values in a ‘normal’ phosphate pH 7.0buffer
where m=0.17 were too large to measure via spectro-
fluorometric titrations. K1 values for complexation of 7
by L1 and L3 are provided in Table 4. K1 for com-
plexation of 7 by L1 in ‘normal’ pH 7.0buffer where
m=0.17 is estimated to be 1 Â 1013 Mꢀ1 in the same
manner as described earlier for estimation of K1 for
complexation of 2 by L1. Direct comparison of equili-
brium constants for complexation of 7 by L1 and its
analogue L3 (L3 lacks the propylamine chains of L1) is
slightly complicated due to differences in the binding
stoichiometries of the two agents. L1 and L3 form 1:1
and 2:1 complexes with 7, respectively.13 However, K1
values for complexation of 7 by L3 in units Mꢀ1 can be
approximated by taking the square root of the multi-
plied through K1K2 Mꢀ2 values. Thus in ‘normal’ pH
7.0buffer where m=0.17, K1, in units Mꢀ1, for com-
plexation of 7 by L3 is determined to be ꢂ5200-fold less
than for complexation of 7 by L1. Thus, the three
propylamine chains of L1 increase the agent’s free
energy of binding by approximately 5.1 kcal.
The propylamine chains of L1 increase its free energy of
binding (ꢀGꢁ) by about 5.1 kcal as calculated from K1
values for complexation of 7 by L1 and L3 (Table 4).
The overall effect of the propylamine chains is to
strengthen L1–dsDNA interactions without a loss of
sequence specificity. For example, K1 values for both
molecules are three orders of magnitude less for com-
plexation of the 5 bp A/T rich site of 1 than the 9 bp site
of 2 (see Table 3). Initially the development of minor
groove binders capable of recognizing longer dsDNA
binding sites was hindered because longer molecules
failed to conform to the curvature of the minor groove.8
This problem was at least partially overcome by con-
necting ‘rigid’ minor groove binding segments, such as
polypyrroles or bisbenzimidazole units with flexible lin-
kers such as for extended polyamides3,5 or the poly-
amide-bisbenzimidazole conjugate L3. In the case of L3,
model building demonstrated how the molecule could
adapt a ‘spiral-like’ conformation which matched that
of the DNA minor groove.13 Figure 6 shows a plausible
structure for an L1–dsDNA complex where it too is
shown to adapt to the curvature of the minor groove.
Construction of ligand–dsDNA models. A plausible
structure for a L1–dsDNA complex was constructed
and its binding site size measured to be 9 bp (Fig. 6).
Models for L4 and L6 dsDNA complexes were also
constructed (Fig. 7). In this case, the basic side chains of
the ligands are shown reaching out of the minor groove.

250
A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
ꢀGꢁ values, calculated from the K1 values presented in
Table 4, were employed to characterize the nature of
binding cooperativity between the tripyrrole and bis-
benzimidazole moieties of L1. As shown in Figure 6, the
two moieties of L1 are estimated to possess a negative
cooperativity of 9 kcal. It is assumed that the free
energy of binding for either moiety of L1 is equivalent
to that for complexation of 1 by either L4 or L8. Thus,
the ꢀGꢁ for complexation of 2 by the tripyrrole peptide
or bisbenzimidazole moieties of L1 are estimated to be
ꢀ14.9 and ꢀ11.6 kcal, respectively, at 26 ꢁC and an
ionic strength of m=0.17. Hypothetically, if no coop-
erativity existed between the two moieties of L1, an L1–
2 complex would possess a ꢀGꢁ of ꢀ26.5 kcal and a K1
of 1 Â 1019
M
ꢀ1 (represented by a dotted line in Fig. 5).
Instead, the complex is measured to have a ꢀGꢁ of 7.5
kcal and a K1 of 5 Â 1012 Mꢀ1
.
At first one might be quick to assume that the negative
cooperativity demonstrated between the two moieties of
L1 is due to a poorly designed linker which prevents the
agent’s moieties from correctly registering with the
DNA bases. After all, this is the problem most often
ascribed to longer or extended minor groove binders.8
However, that formation of the (L8)2–2 complex also
occurs with negative cooperativity suggests otherwise.
K1 for complexation of 2 by L8 is two orders of magni-
tude less than K2 for complexation of 2 by a second
molecule of L8 (Table 2). Similar results are observed
for L9. It seems that the nine bp binding site of 2 is
simply not large enough for two L8 molecules to com-
plex in a linear end-to-end manner. Judging only by
length, two molecules of L8 should be capable of fitting
within a nine bp site. Still, complexation of the first
molecule may cause conformational changes in the
duplex which adversely affects binding by the second.
Another possibility is that the molecule’s preferred
binding site exists in the middle of the nine bp binding
site of 2. Complexation by two molecules then forces
both to either end of the A/T rich sequence.
Our results suggest that an optimized molecule similar
to L1 could possess an equilibrium constant as large as
1019–1020 if the source of negative cooperativity between
the two moieties could be removed. The source of
negative cooperativity appears to be inherent in the
targeting of a nine base pair binding site — that is, the
two minor groove binding moieties need to be spaced
further apart and the molecule directed towards a
longer A/T rich binding site.
We were interested in determining the effect of 50-TpA-
30 dinucleotide steps on the minor groove binding moi-
eties of L1. Particularly, we wished to determine if the
binding orientation of L1 in its complex with 7 could be
predicted from the relative selectivities of its two minor
groove binding moieties — that is, does L1 exclusively
bind the -TATA- region of 7 with its tripyrrole or bis-
benzimidazole moiety. Towards this end, K1 values were
determined for complexation of 3 and 4 by L4 and L8.
Oligomeric duplexes 3 and 4 contain the A/T rich bind-
ing sites -TATAA- and -AAATT-, respectively, similar
to either half of the nine bp site of 7. It has been sug-
gested that 50-TpA-30 dinucleotide steps produce a DNA
structural alteration which discourages minor groove
binding.31,34,35 Dickerson and co-workers have reported
that A/T rich regions of dsDNA usually adopt
poly(dA)/poly(dT)-like structures (referred to as ‘A-
tracks’).33,36 A-tracks consist of successive adenine
bases which stack in a rigid and unbent column.
Emphasized is that the A-track structure can incorpo-
rate 50-ApT-30 steps, but is broken by 50-TpA-30 steps.
For both L4 and L6, K1 values for complexation of 3
were roughly 100-fold greater than for complexation of
4 (Table 2). The specificity of the bisbenzimidazole and
tripyrrole moieties of L1 for an A-track appears to be
Figure 6. Computer generated model of proposed L1–dsDNA complex.

A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
251
equivalent. Thus, neither moiety of L1 likely binds
either region of 7’s binding site, exclusively.
greater than for complexation by L6. In Figure 7, the
basic side chains of both ligands are shown fully exten-
ded and reaching out from the minor groove. Still, the
basic amines of L4 are in close proximity to DNA
phosphate groups on either side of the minor groove
(N–P distances are at most 6–7 A). Additionally, the
nitrogen atoms of L4 are spaced such that each propy-
lamine chain may simultaneously form a salt bridge
with a different phosphate group. In contrast, the poly-
amine side chain of L6 is much longer with N–P dis-
tances averaging ꢂ11 A (for N1, N2, and N3 as labeled
in Chart 1). Further, N1, N2, and N3 are not spaced at
distances equal to the DNA phosphate groups. Inter-
estingly though, L6 is still as potent an inhibitor of TF
binding as L4.23 This may be because the long poly-
amine chain of L6 wraps around the phosphodiester
backbone and reaches into the major groove17 where it
may sterically hinder TF binding. In contrast, the
shorter propylamine chains of L4 are not long enough
to reach that far.
Substitution of the tripyrrole moiety of L1 with a
dipyrrole yields L2 and a 30-fold decrease in K1 for
complexation of 2. The specificity possessed by L2 for
the nine bp A/T rich site of 2 relative to the five bp site
of 1 remains equivalent to that of L1 (Table 3). L4 is the
tripyrrole polyamide moiety of L1. The K1 for com-
plexation of 2 by L4 is 2100-fold less then for com-
plexation by L1. Also, L4 possess little specificity for
longer A/T rich binding sites. Comparison of K1 values
derived via spectrofluorometric titrations show L4 to
possess a K1 for complexation of 1 at least 260-fold
greater than for the classical minor groove binders L7
and L8 and 360-fold greater than for the fellow micro-
gonotropen L6 (Table 4).
Molecular models are employed in an effort to explain
why the K1 for complexation of 1 by L4 is 260-fold
Acknowledgements
This work was supported by a grant from the National
Institute of Health (5R37DK09171-36).
References and Notes
1. Reddy, B. S. P.; Sondhi, S. M.; Lown, J. W. Pharmacol.
Ther. 1999, 84, 1.
2. Chiang, S. Y.; Bruice, T. C.; Azizkhan, J. C.; Gawron, L.;
Beerman, T. A. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 2811.
3. Trauger, J. W.; Baird, E. E.; Dervan, P. B. J. Am. Chem.
Soc. 1998, 120, 3534.
4. Trauger, J. W.; Baird, E. E.; Mrksich, M.; Dervan, P. B. J.
Am. Chem. Soc. 1996, 118, 6160.
5. Kissinger, K. L.; Dabrowiak, J. C.; Lown, J. W. Chem.
Res. Tox. 1990, 3, 162.
6. Chen, Y. H.; Yang, Y. W.; Lown, J. W. J. Biomol. Struct.
Dyn. 1996, 14, 341.
7. Turner, J. M.; Baird, E. E.; Dervan, P. B. J. Am. Chem.
Soc. 1997, 119, 7636.
8. Kelly, J. J.; Baird, E. E.; Dervan, P. B. Proc. Natl. Acad.
Sci. U.S.A. 1996, 93, 6981.
9. Singh, M. P.; Plouvier, B.; Hill, G. C.; Gueck, J.; Pon,
R. T.; Lown, J. W. J. Am. Chem. Soc. 1994, 116, 7006.
10. Beerman, T. A.; Woynarowski, J. M.; Sigmund, R. D.;
Gawron, L. S.; Rao, K. E.; Lown, J. W. Biochim. Biophys.
Acta 1991, 1090, 52.
11. Filipowsky, M. E.; Kopka, M. L.; Brazilzison, M.; Lown,
J. W.; Dickerson, R. E. Biochemistry 1996, 35, 15397.
12. Lown, J. W.; Krowicki, K.; Balzarini, J.; Newman, R. A.;
Declercq, E. J. Med. Chem. 1989, 32, 2368.
13. Satz, A. L.; Bruice, T. C. J. Am. Chem. Soc. 2001, 123,
2469.
14. Zimmer, C.; Wahnert, U. Prog. Biophys. Mol. Biol. 1986,
47, 31.
¨
15. He, G. X.; Browne, K. A.; Groppe, J. C.; Blasko, A.; Mei,
H. Y.; Bruice, T. C. J. Am. Chem. Soc. 1993, 115, 7061.
16. He, G. X.; Browne, K. A.; Blasko, A.; Bruice, T. C. J.
Am. Chem. Soc. 1994, 116, 3716.
17. Blasko, A.; Browne, K. A.; Bruice, T. C. J. Am. Chem.
Soc. 1994, 116, 3726.
Figure 7. Computer generated models of L4 (above) and L6 (below)
1:1 complexes with dsDNA. L4 and L6 are shown bound within the
minor groove. The positively charged chains of both agents are shown
fully extended and reaching out of the minor groove. Only relevant
DNA bases are shown.

252
A. L. Satz, T. C. Bruice / Bioorg. Med. Chem. 10 (2002) 241–252
18. Xue, T. H.; Browne, K. A.; Bruice, T. C. Bioconjugate
Chem. 1995, 6, 82.
28. Coll, M.; Frederick, C. A.; Wang, A. H.; Rich, A. Proc.
Natl. Acad. Sci. U.S.A. 1987, 84, 8385.
19. Bruice, T. C.; Sengupta, D.; Blasko, A.; Chiang, S. Y.;
Beerman, T. A. Bioorg. Med. Chem. 1997, 5, 685.
20. Satz, A. L.; Bruice, T. C. Bioorg. Med. Chem. 2000, 8,
1871.
29. Fields, G. B.; Noble, R. L. Intl. J. Pep. Prot. Res. 1990,
35, 161.
30. Sarin, V. K.; Kent, S. B. H.; Tam, J. P.; Merrifield, R. B.
Anal. Biochem. 1981, 117, 147.
21. Satz, A. L.; Bruice, T. C. Bioorg. Med. Chem. Lett. 1999,
9, 3261.
31. Satz, A. L.; White, C. M.; Beerman, T. A.; Bruice, T. C.
Biochemistry 2001, 40, 6465.
22. White, C. M., Gawron, L., Satz, A. L., Bruice, T. C., Beer-
man, T. A. Biochim. Biophys. Acta submitted for publication.
23. White, C. M., Satz, A. L., Bruice, T. C., Beerman, T. A.
Proc. Natl. Acad. Sci. U.S.A., in press.
32. Loontiens, F. G.; Regenfuss, P.; Zechel, A.; Dumortier,
L.; Clegg, R. M. Biochemistry 1990, 29, 9029.
33. Dickerson, R. E.; Goodsell, D.; Kopka, M. L. J. Mol.
Biol. 1996, 256, 108.
24. White, C. M., Beerman, T. A. unpublished results.
25. Latchman, D. S. Gene Regulation: A Eukaryotic Perspec-
tive, 3rd ed.; Stanley Thornes: Cheltenham, 1998.
26. Marky, L. A.; Breslauer, K. J. Biopolymers 1987, 26, 1601.
27. Crothers, D. M. Biopolymers 1971, 10, 2147.
34. Ward, B.; Rehfuss, R.; Goodisman, J.; Dabrowiak, J. C.
Biochemistry 1988, 27, 1198.
35. Chen, F. M.; Sha, F. Biochemistry 1998, 37, 11143.
36. Goodsell, D. S.; Kaczorgrzeskowiak, M.; Dickerson, R. E.
J. Mol. Biol. 1994, 239, 79.