Quantitative structure-activity relationships on 5-substituted terbenzimidazoles as topoisomerase I poisons and antitumor agents

Article abstract of DOI:10.1016/S0968-0896(97)10021-9

Several 5-substituted terbenzimidazoles were synthesized and evaluated as mammalian topoisomerase I poisons and for cytotoxicity against a human lymphoblastoma cell line, RPMI-8402. No correlation was observed between topoisomerase I poisoning activity and the Hansch π value or the σ(meta) and σ(para) values associated with each substituent. These data suggest that electronic effects and relative lipophilicity of substituents at the 5-position of these terbenzimidazoles do not have a significant effect upon intrinsic topoisomerase I poisoning activity. There was, however, a good correlation between the relative π values for the various subtituents evaluated and cytotoxic activity. Experimentally determined log P values did not correlate well with either cytotoxicity or π values. Capacity factors (log k') as determined by high pressure liquid chromatography did correlate well with the π values of varied substituents and cytotoxicity. These data indicate that the relative lipophilic activity of substituents at the 5-position of these terbenzimidazoles can strongly influence relative cytotoxic activity.

Full text of DOI:10.1016/S0968-0896(97)10021-9

BIOORGANIC &
MEDICINAL
CHEMISTRY
Bioorganic & Medicinal Chemistry 6 (1998) 163±172
Quantitative structure±activity relationships on 5-substituted
terbenzimidazoles as topoisomerase I poisons and antitumor
agents
Jung Sun Kima, Qun Suna, Chiang Yub, Angela Liub, Leroy F. Liub,
Edmond J. LaVoiea*
aDepartment of Pharmaceutical Chemistry, Rutgers, The State University of New Jersey, Piscataway, NJ 08855, U.S.A.
bDepartment of Pharmacology, The University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School,
Piscataway, NJ 08855, U.S.A.
Received 25 July 1997; accepted 4 September 1997
Abstract
Several 5-substituted terbenzimidazoles were synthesized and evaluated as mammalian topoisomerase I poisons and for
cytotoxicity against a human lymphoblastoma cell line, RPMI-8402. No correlation was observed between topoiso-
merase I poisoning activity and the Hansch ꢀ value or the ꢁmeta and ꢁpara values associated with each substituent. These
data suggest that electronic e€ects and relative lipophilicity of substituents at the 5-position of these terbenzimidazoles
do not have a signi®cant e€ect upon intrinsic topoisomerase I poisoning activity. There was, however, a good corre-
lation between the relative ꢀ values for the various subtituents evaluated and cytotoxic activity. Experimentally deter-
mined log P values did not correlate well with either cytotoxicity or ꢀ values. Capacity factors (log k0) as determined by
high pressure liquid chromatography did correlate well with the ꢀ values of varied substituents and cytotoxicity. These
data indicate that the relative lipophilic activity of substituents at the 5-position of these terbenzimidazoles can strongly
in¯uence relative cytotoxic activity. # 1998 Elsevier Science Ltd. All rights reserved.
Keywords: Antitumor, QSAR, structure±activity, topoisomerase, terbenzimidazoles.
1. Introduction
lymphoblast cell line, RPMI 8402. In contrast, the pre-
sence of a 5-phenyl or 5-pyridyl substituent on these
terbenzimidazoles, as in 3 and 4, resulted in analogs that
retained activity as topoisomerase I poisons and pos-
sessed signi®cant cytotoxicity against several tumor cell
lines [12]. The basis for the lack of cytotoxicity observed
for 1 and 2 has been attributed to poor penetration into
cells [12,14].
Previous studies have established that varied sub-
stituents could be accommodated at the 5-position of
these terbenzimidazoles with retention of activity as
topoisomerase I poisons [12,14]. These data prompted
the present quantitative structure-activity relationship
(QSAR) study. Pharmacological data in earlier
studies did suggest that lipophilicity could in¯uence
the cytotoxic activity of terbenzimidazoles. Several
Topoisomerases are nuclear enzymes that regulate
topological and conformational changes in DNA cri-
tical to cellular processes such as replication and tran-
scription [1±3]. Mammalian topoisomerase poisons have
been recognized as e€ective cancer chemotherapeutics
[4±7]. Several bibenzimidazoles and terbenzimidazoles
have recently been identi®ed as topoisomerase I poisons
[8±14]. Despite exhibiting potent activity as topoisome-
rase poisons, speci®c terbenzimidazoles, such as com-
pounds 1 and 2 (Fig. 1), did not exhibit signi®cant
cytotoxicity as re¯ected in their e€ect upon the human
*Corresponding author.
0968-0896/98/$19.00 # 1998 Elsevier Science Ltd. All rights reserved
PII: S0968-0896(97)10021-9

164
J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
5-substituted terbenzimidazoles with varying Hansch ꢀ
values [15] were selected for synthesis to evaluate the
in¯uence of lipophilicity in enhancing the cytotoxic
potential of 5-substituted terbenzimidazoles. It has been
hypothesized that electronic e€ects could alter the abil-
ity of these terbenzimidazoles to stabilize the ternary
enzyme-DNA complex. To investigate whether speci®c
electronic parameters could modify biological activity,
5-substituted terbenzimidazoles with varying electronic
properties were, therefore, selected for synthesis and
pharmacological evaluation. The present study exam-
ines for several 5-substituted terbenzimidazoles the
relationship between these physicochemical parameters
and topoisomerase I poisoning activity as well as rela-
tive cytotoxicity in human lymphobast cells.
as well as ꢀ and ꢁpara veri®ed the divergence of these
parameters [16].
Methods for the preparation of 5, 6, and 7 have been
previously described [12,14]. Scheme 1 outlines the gen-
eral procedure used for the preparation of compounds
8, 9, 10, 11, 12, and 13. Treatment of 5-methoxy-
terbenzmidazole (12) with BBr3 in CH2Cl2 provided 5-
hydroxyterbenzimidazole (14). Catalytic hydrogenation
of 13 using 10% Pd/C provided 5-aminoterbenzimida-
zole (15).
3. Results and discussion
The Hansch ꢀ and ꢁ values as well as the relative
topoisomerase I poison activity and cytotoxicity in the
RPMI 8402 cells of the terbenzimidazoles evaluated in
this study are listed in Table 1. No ꢁpara values were
found in the literature for the naphthyl substituents
associated with the isomeric terbenzimidazole deriva-
tives, 5 and 6. In the case of the 5-piperidinyl terbenzi-
midazole, 9, no ꢁmeta values were located in the
literature for the piperidinyl moiety. While experimental
ꢀ values for 5-(1-naphthyl) and 5-(2-naphthyl) sub-
stituents were unavailable, calculated ꢀ values could be
obtained for 5 and 6 using Rekker's method [17].
2. Chemistry
A group representing 14 5-substituted terbenzimida-
zoles was selected for synthesis and biological evalua-
tion on the basis of their varied electronic e€ects as
indicated by their di€ering ꢁ values (Table 1). The
compounds selected for incorporation in this study also
display a cross section of ꢀ values representing di€ering
degrees of lipophilicity and are listed in Table 1 in order
of highest to lowest ꢀ value. Both ꢁpara and ꢁmeta values
relative to the NH at position 1 were taken into con-
sideration because of the tautomeric nature of benzimi-
dazoles. As shown in Fig. 2, the functional group R
can be viewed as para to the -NH of the primary benzi-
midazole ring in one tautomeric form while meta to the
-NH in the other. It is important in the selection of
compounds for synthesis and biological evaluation that
the ꢀ and ꢁ values not be highly correlated with each
other. Two-dimensional Craig plots of ꢀ and ꢁmeta
Topoisomerase I poisoning activity is assessed in a
subcellular assay that re¯ects the ability of the drug to
stabilize the ternary cleavable complex. The assay uses
isolated enzyme and plasmid DNA (see Experimental).
There is approximately a 15 min time period wherein the
added drug is permitted to interact with a mixture of
DNA and enzyme prior to the initiation of gel electro-
phoresis to measure the extent of DNA fragmentation.
In this measurement of the intrinsic potency as a topoi-
somerase I poison, factors such as drug stability and
potential to cross cellular membranes are usually
inconsequential. Steric factors, electronic e€ects, and the
presence of localized area of hydrophilicity or lipophili-
city would be anticipated as being among the principal
physicochemical parameters that could in¯uence topoi-
somerase I poisoning activity. With the exception of 5,
6, and 9 for which either ꢁpara or ꢁmeta values were not
available, a multiple regression analysis of the remaining
eleven compounds in Table 1 was performed to deter-
mine if there was a correlation between topoisomerase I
poisoning and either ꢀ or ꢁ values. As seen in the matrix
provided in Chart 1, topoisomerase I activity did not
correlate well with ꢀ or ꢁ values. The highest correlation
was 0.37 for ꢁp. The multiple regression model as shown
in Eq. (1) had a low correlation coecient and an F
value that is signi®cantly lower than the tabulated value
representing a model of signi®cance p>0.05. These
results indicate that there is no de®nitive relationship
between these physiochemical parameters and topoiso-
merase I activity.
X
N
N
N
N
H
H
N
N
H
1, X = H
3, X =
4, X =
2, X = CN
N
Fig. 1.

J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
165
Log 1=topo I ˆ 0:124 ‡ 0:098ꢀ ‡ 2:05ꢁp 0:912ꢁm
n ˆ 11; S:E: ˆ 0:573; F ˆ 0:61;
The relationship between physicochemical parameters
and cytotoxicity, expressed as IC50, for these eleven
compounds was also examined. Cytotoxicity, as shown
in the matrix listed in Chart 2, did not correlate well
with ꢁ values. In contrast, however, ꢀ values did corre-
late better with the cytotoxicity with a correlation coef-
®cient of 0.70 (Chart 2). The correlation slightly
improved by addition of terms based on ꢁm and ꢁp to
the ꢀ terms bringing the r ˆ 0:76. However, the F ratio
Table F ꢀ3; 7; 0:05† ˆ 4:35
r ˆ 0:46; r2 ꢀVariance in Log 1=topo I
explained by regression† ˆ 21%
ꢀ1†
Table 1
Physicochemical parameters and biological results associated with 5-substituted terbenzimidazolesa
R
N
N
N
N
H
H
N
N
H
Code
R
ꢀ
ꢁp
ꢁm
Topob
Cytotoxicityc
5
1-Naphthyl
2-Naphthyl
Phenyl
Propyl
Br
3.19d
3.19d
1.96
1.45
0.86
0.85
0.71
0.14
0.00
0.02
0.28
0.57
0.67
1.23
NAe
NAe
0.01
0.17
0.23
0.57
0.23
0.06
0.00
0.27
0.78
0.66
0.37
0.66
0.06
0.06
0.06
0.06
0.39
NAe
0.37
0.34
0.00
0.12
0.71
0.56
0.12
0.16
10
10
1
1.26
0.16
6
3
0.19
7
0.5
1
15.31
1.63
0.63
8
9
Piperidine
Cl
0.5
1
10
11
2
1.30
1.70
F
0.05
1
H
5.00
0.79
12
13
1
OCH3
NO2
CN
0.05
0.5
0.1
0.5
1
113.9
133.4
150.3
61.64
14
15
OH
NH2
aHansch ꢀ and ꢁ values were used except where otherwise noted [15].
bDetails provided in the Experimental section. Values provided re¯ect relative e€ective concentrations needed to produce a similar
degree of DNA fragmentation.
cValues represent the IC50 values (ꢂM).
dCalculated value using Rekker's method [17].
eNA indicates that values were not available in the literature.
N
N
N H
N H
H
N
N
R
R
N
N
N
N
N
H
H
N
H
Fig. 2. Tautomeric forms of 5-substituted terbenzimidazoles.

166
J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
O
C
N
N
X
NH2
NO2
X
NH2
NH2
H
N
+
N
H
H
8, X = Br
X
N
N
9, X = (CH2)5N
10, X = Cl
N
11, X = F
N
H
12, X = OCH3
13, X = NO2
H
N
N
H
Scheme 1.
was still too small to validate the relationshiop between
these parameters and cytotoxicity.
Log 1=IC50
ˆ
0:902 ‡ 1:078ꢀ 0:802ꢁm
n ˆ 10; S:E: ˆ 0:547; F ˆ 14:06;
Table F ꢀ2; 7; 0:01† ˆ 9:55
r ˆ 0:90; r2ꢀVariance in Log 1=IC50
explained by regression† ˆ 80%
Log 1=IC50
ˆ
1:504 ‡ 0:914ꢀ ‡ 2:265ꢁp 1:868ꢁm
ꢀ4†
ꢀ5†
n ˆ 11; S:E: ˆ 0:799; F ˆ 3:20;
Table F ꢀ3; 7; 0:05† ˆ 4:35
ꢀ2†
r ˆ 0:76; r2 ꢀVariance in Log 1=IC50
explained by regression† ˆ 58%
Log 1=IC50
ˆ
0:656 ‡ 1:016ꢀ 1:165ꢁp
n ˆ 10; S:E: ˆ 0:572; F ˆ 12:53;
Table F ꢀ2; 7; 0:01† ˆ 9:55
In an e€ort to improve the correlation between the
biological activities and the physicochemical para-
meters, an attempt was made to remove outliers. The 5-
propylterbenzimidazole analog, 7, which possesses
substantial streic ¯exibility, is an outlier. It is possible
that the ¯exible nature of the propyl group in terms of
conformation may represent a detriment to its pharma-
cological activity. Removal of 7, did improve the corre-
lation as seen in equations 3, 4, and 5 and in the
correlation matrix which showed the r ˆ 0:84 for ꢀ.
r ˆ 0:88; r2ꢀVariance in Log 1=IC50
explained by regression† ˆ 78%
Two variable multiple regression equations with either
the variables ꢀ and ꢁm or ꢀ and ꢁp gave better correla-
tions than the three variable equation with all three
variables ꢀ, ꢁm, and ꢁp. Equation (3) (which included ꢀ,
ꢁp, ꢁm) had a correlation coecient of 0.90 but F ratio
was signi®cant only at the 2.5% level. Equations (4) and
(5), on the other hand, had correlation coecients of
0.90 and 0.88, respectively, as well as F value that was
signi®cant at the 1% level.
Log 1=IC50
ˆ
0:960 ‡ 1:089ꢀ ‡ 0:267ꢁp 0:957ꢁm
n ˆ 10; S:E: ˆ 0:589; F ˆ 8:07;
ꢀ3†
Table F ꢀ3; 6; 0:025† ˆ 6:60; F ꢀ3; 6; 0:01† ˆ 9:78
r ˆ 0:90; r2 ꢀVariance in Log 1=IC50
explained by regression† ˆ 80%
However, results of the T test on the coecients of
the above equations showed ꢁm [Eq. (4)] and ꢁp [Eq. (5)]
Chart 1.
Chart 2.

J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
167
to have T values of 1.64 and 1.90, respectively (Table
T test ˆ 1:895, p ˆ 0:1). These T values were not sig-
ni®cant even at the 10% level. The coecient ꢀ, on the
other hand, had a T value of 4.817 [Eq. (4)] and 5.240
[Eq. (5)], which were signi®cant at the 1% level. Equa-
tion (6), a one variable simple linear equation, best
de®nes the relationship between these physicochemical
properites of these terbenzimidazoles and their cytotoxic
activity.
correlation coecient of 0.829 and F value of 22.06,
which was signi®cant at the 1% level and a T value of
4.697 for ꢀ, which was also signi®cant at the 1% level.
The addition of 5 did have a negative e€ect upon the
correlation observed in Eq. (6). In the instance where 5,
6, and 9 were taken into consideration, a correlation
coecient of 0.761 was obtained.
Electronic parameters (ꢁpara and ꢁmeta) did not con-
tribute to the cytotoxicity and to the topoisomerase I
activity of these compounds whereas lipophilicity (ꢀ)
did signi®cantly correlate with the cytotoxicity of
these 5-substituted terbenzimidazoles. Based on these
results, we decided to expand this study by further
investigating the e€ects of lipophilicity on the cyto-
toxicity of the 5-substituted terbenzimidazoles by
applying two experimentally determined lipophilic
parameters: the partition coecient, log Poct; and the
capacity factor, log k0.
Log 1=IC50
ˆ
0:949 ‡ 0:996ꢀ
n ˆ 10; S:E: ˆ 0:232; T ˆ 4:294;
Table T ꢀp ˆ 0:01† ˆ 3:335 F ˆ 18:44;
Table F ꢀ1; 8; 0:01† ˆ 11:26
r ˆ 0:84; r2 ꢀVariance in Log 1=IC50
explained by regression† ˆ 70%
ꢀ6†
These data indicate that the critical parameter asso-
ciated with the cytotoxicity is their ꢀ values. We inclu-
ded in these analyses, 5 and 6, using calculated ꢀ values
according to Rekker's method [17] and 9, employing the
Hansch ꢀ value associated with the 5-piperidinyl group.
The addition of 6 and 9 to this QSAR study did not
have a signi®cant e€ect upon correlation observed in
Eq. (6). Including these two compounds provided a
Table 2 summarizes the experimentally determined
log k0 and log P values of compounds in this study,
with the exception of the propyl analog, 7, together with
their reported or calculated ꢀ values. Similar experi-
mental conditions to those previously described for
determination of relative hydrophobicity constants by
reversed-phase HPLC were employed in this study
[18,19].
Table 2
Lipophilic parameters of 5-substituted terbenzimidazoles
N
N
H
N
R
N
N
H
N
H
Compound
R
ꢀa
Log k0b
Log Pc
5
1-Naphthyl
2-Naphthyl
3.19d
3.19d
1.96
0.86
0.85
0.71
0.14
0.00
0.02
0.28
0.57
0.67
1.23
0.509
0.448
0.173
0.011
0.088
0.014
0.259
0.316
0.253
0.159
0.262
0.573
0.546
1.76
1.61
1.43
2.31
0.89
1.45
3.05
2.42
1.73
1.21
2.28
1.17
1.46
6
3
Phenyl
Br
1-Piperidinyl
8
9
10
11
2
Cl
F
H
12
13
1
OCH3
NO2
CN
OH
NH2
14
15
aHansch ꢀ values obtained from literature [15].
bCapacity factors determined by HPLC.
cAverage of triplicates.
dCalculated value using Rekker's method [17].

168
J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
As shown in Fig. 3, there is a very good correlation
between the ꢀ values and the capacity factors deter-
mined by HPLC, giving the Eq. (7), where
index of these terbenzimidazoles. The major advantages
of this method are that it is not a€ected by self-aggre-
gation and that small di€erence in the purity of the
compounds do not radically alter the results.
Using log k0 as the lipophilicity parameter, the corre-
lation between log 1/IC50 and log k0 was examined.
Equation (8) was obtained where
Log k0 ˆ 0:2304 0:2328 with an r2 of 0:925 ꢀ7†
Log P values were determined experimentally under
standard experimental conditions [20]. The log P values
obtained did not correlate with either the ꢀ values or
their pharmacological activity. One explanation for this
may be that the activity of these compounds may be
related to the importance of a localized increase in lipo-
philicity at the 5-position of the molecule. Thus, the ꢀ
values may tend to correlate better than the overall
lipophilicity with cytotoxicity. In addition, there was a
great variation in the log P values ranging from 0.89 to
3.05 for these terbenzimidazoles which intuitively was
unexpected. In studies performed by Dr D. Pilch (per-
sonal communication) it has been observed that terben-
zimidazoles such as 3 can self-associate and form several
di€erent types of aggregates depending upon con-
centration. This physical property could signi®cantly
contribute to the unreliability of these experimentally
determined log P values. There are many factors that
could a€ect this physical phenomenon, including charge
and electronic properties of the 5-substituent. Alteration
of the pKa of the compound could increase or decrease
the association constant as well as alter the ¯uorescence
of these compounds.
Log 1=IC50 ˆ 2:500 log k0
0:358
ꢀ8†
which had a correlation of r2 ˆ 0:56 (Fig. 4). Removing
the 5-(1-naphthyl) analog, 5, resulted in Eq. (9) which
had an improved r2 ˆ 0:63.
Log 1=IC50 ˆ 2:971 log k0 0:239
ꢀ9†
This 1-naphthyl derivative, 5, like the propyl analog, 7,
appears to be an outlier. Its relative cytotoxicity is sig-
ni®cantly less than that of its 2-naphthyl isomer, 6. In
contrast to 6, there is a greater energy barrier associated
with the 5-(1-naphthyl) isomer adopting a conformation
wherein it resides within the same plane as the benzimi-
dazole ring to which it is attached. This is not surprising
in view of the potential of one of the peri-hydrogens on
the 1-naphthyl analog (H-8), to sterically interact with
H-4 or H-6 on the terbenzimidazole. This suggests that,
although lipophilicity of the 5-substituents can play a
major role in determining cytotoxicity of these 5-sub-
stituted terbenzimidazoles, other factors such as steric
e€ects may also in¯uence activity.
The lack of convenience and reliability of these log P
determinations prompted our investigation into utilizing
the reverse phase HPLC for obtaining the lipophilic
4. Experimental
Melting points were determined with a Thomas±
Hoover Unimelt capillary melting point apparatus.
Fig. 3. Plot of ꢀ versus log k0.
Fig. 4. Plot of log k0 versus log 1/IC50
.

J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
169
Column chromatography refers to ¯ash chromato-
graphy conducted on SiliTech 32±63 ꢂm, (ICN Biome-
dicals, Eschwegge, Ger.) using the solvent systems
indicated. Infrared spectral data (IR) were obtained on
dazo-50-yl)benzimidazole (119 mg, 0.45 mmol) were stir-
red in nitrobenzene (5 ml) at 145ꢀC overnight. The
cooled reaction mixture was then puri®ed directly by
column chromatography. Elution with 0±10 % metha-
nol/ethyl acetate provided 127 mg (66%) of a brownish-
white solid: mp>280ꢀC; IR (KBr) 3101, 1626, 1547,
1440, 1294, 1140; 1H NMR ꢃ 7.24 (dd, 1H, J ˆ 7:0, 1.5),
7.57 (d, 1H, J ˆ 9:0), 7.71±7.80 (m, 3H), 8.04±8.18 (m,
2H), 8.39 (s, 2H), 8.50 (s, 1H); 13C NMR (DMSO-d6+3
drops CF3COOH) ꢃ 114.1, 115.8, 116.2, 116.4, 117.0,
118.6, 123.5, 125.3, 126.2, 128.7, 128.9, 131.8, 132.0,
132.3, 133.1, 134.4, 138.3, 140.6, 151.1, 153.4; HRMS
(FAB) calcd for C21H14N6Br (MH+), 429.0463, found,
429.0454.
a
Perkin-Elmer 1600 Fourier transform spectro-
photometer and are reported in cm 1. Proton (1H
NMR) and carbon (13C NMR) nuclear magnetic reso-
nance were recorded on a Varian Gemini-200 Fourier
Transform spectrometer. NMR spectra (200 MHz 1H
and 50 MHz 13C) were recorded in the deuterated sol-
vent indicated with chemical shifts reported in ꢃ units
down®eld from tetramethylsilane (TMS). Coupling
constants are reported in hertz (Hz). A few drops of
CF3COOH improved 13C NMR spectra by allowing for
increased solubility and formation of the protonated
form of the terbenzimidazoles, thereby eliminating tau-
tomeric di€erences among carbon atoms. Mass spectra
were obtained from Washington University Resource
for Biomedical and Bioorganic Mass Spectrometry
within the Department of Chemistry at Washington
University, St Louis, MO. The purity of all compounds
for which HRMS data are provided was determined by
analytical reverse-phase HPLC. Compounds were ana-
lyzed using both of the following conditions (method A)
a Vydac C-18 column (The Separations Group) using
methanol:H2O) (87:13) with a ¯ow rate of 1 ml/min;
(method B) a Microsorb C-8 column (Rainin Instru-
ment Co., Inc.) using methanol:0.1 M potassium phos-
phate bu€er (pH 7.0) (95:05) with a ¯ow rate of 1 ml/
min. HPLC analyses were performed with a Hewlett-
Packard 1090 liquid chromatograph equipped with a
diode array UV detection monitoring at 335 nm. The %
purities of these compounds were calculated from the
peak area assuming that the extinction coecient of the
compound of interest and the impurity are the same. On
the basis of these analyses, all the compounds were
found to be 98.0±99% pure in these systems. Combus-
tion analyses were performed by Atlantic Microlabs,
Inc., Norcross, GA, and were within ‹0.4% of the
theoretical value. Methods for the preparation of 1,12
2,12 3,12 4,12 5,14 6,14 and 712 have been previously
described.
4.1.1 5-(1-Piperidinyl)-2[20-(benzimidazol-500-yl)
benzimidazol-50-yl]benzimidazole (9)
2-Nitro-5-(1-piperidinyl)aniline [21] (400 mg, 1.8mmol)
was reduced by hydrogenation with Pd/C (80 mg) in
ethyl acetate (50 ml) overnight. After passing through
a bed of celite and concentrating in vacuo, 4-(1-piperi-
dinyl)-1,2-phenylenediamine was obtained. Without
further puri®cation, the crude diamine (341 mg,
1.8 mmol) and 5-formyl-2-(benzimidazo-50-yl)benz-
imidazole (200 mg, 0.76 mmol) in nitrobenzene (10 ml)
provided 104 mg (32%) of a brownish white solid:
mp >280ꢀC; IR (KBr) 3118, 2923, 2821, 1626,
1436, 1287, 1123; 1H NMR (DMSO-d6+3 drops
CF3COOH) ꢃ 1.70 (m, 2H), 1.89 (m, 4H), 3.58 (m, 4H),
7.67 (d, 1H, J ˆ 9:0), 7.84±7.823 (m, 5H), 8.46 (d, 1H,
J ˆ 9:0), 8.60 (s, 1H), 8.73 (s, 1H), 9.66 (s, 1H);
13C NMR (DMSO-d6+3 drops CF3COOH) ꢃ 21.5,
24.1, 54.9, 107.5, 111.8, 114.1, 115.3, 115.4,
115.9,116.1, 117.9, 120.4, 123.4, 125.2, 132.0, 133.4,
135.4, 139.5, 140.1, 148.1, 151.8, 153.1; HRMS
(FAB) calcd for C26H24N7 (MH+) 434.2093, found
434.2076.
4.1.2 5-Chloro-2[20-(benzimidazol-500-yl)benzimidazol-
50-yl]benzimidazole (10)
2-Nitro-5-chloroaniline (152 mg, 0.88 mmol) was
reduced by re¯uxing in absolute ethanol (20 ml) with
SnCl2 (1.68 g, 8.86 mmol) overnight under N2 to obtain
4-chloro-1,2-phenylenediamine. Without further pur-
i®cation, the crude diamine (125 mg, 0.88 mmol) and 5-
4.1 General procedures for the preparation of 5-
substituted-2[20-(benzimidazol-500-yl)benzimidazol-50-
yl]benzimidazole: 5-Bromo-2[20-(benzimidazol-500-
yl)benzimidazol-50-yl]benzimidazole (8)
formyl-2-(benzimidazo-50-yl)benzimidazole
(160 mg,
0.61 mmol) in nitrobenzene (5 ml) provided 167 (71%)
of a brownish-white solid: mp > 280ꢀC; IR (KBr) 3103,
2826, 1427, 1293; 1H NMR ꢃ 7.24 (dd, 1H, J ˆ 8:25,
2.0), 7.60±7.87 (m, 4H), 8.07±8.17 (m, 2H), 8.40 (s, 2H),
8.49 (s, 1H); 13C NMR (DMSO-d6+3 drops
CF3COOH) ꢃ 114.3, 114.4, 115.3, 115.5, 115.6, 116.1,
118.5, 123.0, 125.4, 125.5, 125.6, 129.4, 132.3, 132.9,
133.0, 135.2, 138.9, 140.8, 151.8, 153.5; HRMS (FAB)
calcd for C21H14N6Cl (MH+) 385.0968, found
385.0954.
2-Nitro-4-bromoaniline (170 mg, 0.78 mmol) was
reduced by re¯uxing in absolute ethanol (20 ml) with
SnCl2 (1.50 g, 7.91 mmol) overnight under N2. When the
reaction was complete, it was basi®ed to pH 11 with 2 N
NaOH, and then extracted with ether. After removing
the solvent in vacuo, 4-bromo-1,2-phenylenediamine
was obtained. Without further puri®cation, the crude
diamine (138 mg, 0.73 mmol) and 5-formyl-2-(benzimi-

170
J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
4.1.3 5-Fluoro-2[20-(benzimidazol-500-yl)benzimidazol-
50-yl]benzimidazole (11)
115.9, 118.7, 123.4, 124.3, 124.6, 124.8, 132.3, 133.4,
134.0, 137.2, 138.7, 142.7, 143.1, 152.3, 155.9; HRMS
(FAB) calcd for C21H14N7O2 (MH+) 396.1209, found
396.1209.
2-Nitro-4-¯uoroaniline (99 mg, 0.63 mmol) was
reduced by hydrogenation with Pd/C (20 mg) in ethyl
acetate (50 ml) for 3 h. After passing through a bed of
celite and concentrating in vacuo, 4-¯uoro-1,2-phenyle-
nediamine was obtained. Without further puri®cation,
the crude diamine (80 mg, 0.63 mmol) and 5-formyl-2-
(benzimidazo-50-yl)benzimidazole (84 mg, 0.32 mmol) in
nitrobenzene (5 ml) provided 85 mg (72%) of a brown-
ish-white solid: mp >280ꢀC; IR (KBr) 3050, 2817, 1625,
1553, 1429, 1131; 1H NMR (DMSO-d6+3 drops
CF3COOH) ꢃ 7.41±7.52 (m, 1H), 7.76 (dd, 1H, J ˆ 8:5,
2.0), 7.88±7.95 (m, 1H), 8.03±8.20 (m, 3H), 8.48 (dd, 1H,
4.1.6 5-Hydroxy-2[20-(benzimidazol-500-yl)
benzimidazol-50-yl]benzimidazole (14)
Compound 12 (88 mg, 0.29 mmol) was dissolved in
freshly distilled CH2Cl2 (50 ml) and cooled to 78ꢀC.
BBr3 (1.0 M CH2Cl2 solution, 0.58 ml) was slowly
added. Temperature was brought to room temperature
overnight under N2. The reaction was quenched with
H2O and CH2Cl2 removed in vacuo. After extracting
the reaction mixture with ethyl acetate, the H2O layer
was neutralized to pH 7.0 and extracted with ethyl
acetate. The ethyl acetate fractions were put together
and concentrated in vacuo after drying in Na2SO4.
Separation on a silica-gel column eluting with 10±20%
methanol/ethyl acetate gave 68 mg (64%) of brownish
solid: mp >280ꢀC; IR (KBr) 3409, 3104, 2973, 2813,
J ˆ 9:0, 1.5), 8.61 (s, 1H), 8.73 (s, 1H), 9.71 (s, 1H); 13
C
NMR (DMSO-d6+3 drops CF3COOH) ꢃ 100.7, 101.2,
112.4, 113.9, 114.1, 114.7, 115.7, 115.9, 116.2, 118.2,
120.3, 123.5, 125.4, 129.2, 131.7, 132.9, 133.2, 140.4,
151.3, 153.5, 162.8; HRMS (FAB) calcd for C21H14N6F
(MH+) 369.1264, found 369.1262.
1
1702, 1622, 1419, 1244, 1149; H NMR ꢃ 6.69 (dd, 1H,
4.1.4 5-Methoxy-2[20-(benzimidazol-500-yl)
benzimidazol-50-yl]benzimidazole (12)
J ˆ 8:5, 2.0), 6.94 (s, 1H), 7.40 (d, 1H, J ˆ 8:0), 7.69±
7.82 (m, 2H), 8.01 (s, 1H), 8.10 (d, 1H, J ˆ 8:0), 8.19±
8.50 (m, 3H), 9.18 (brs, 1H); 13C NMR (DMSO-d6+3
drops CF3COOH) ꢃ 98.3, 114.2, 114.8, 115.3, 115.8,
116.2, 116.3, 123.3, 125.3, 125.4, 126.4, 131.8, 133.2,
139.9, 140.3, 145.7, 148.4, 150.4, 153.3, 156.7, 161.6;
HRMS (FAB) calcd for C21H15N6O (MH+) 367.1310,
found 367.1309.
2-Nitro-4-methoxyaniline (168 mg, 1.0 mmol) was
reduced by hydrogenation with Pd/C (30 mg) in ethyl
acetate (50 ml) for 3 h. After passing through a bed of
celite and concentrating in vacuo, 4-methoxy-1,2-phe-
nylenediamine was obtained. Without further puri®-
cation, the crude diamine (136 mg, 0.99 mmol) and
5-formyl-2-(benzimidazo-50-yl)benzimidazole (131 mg,
0.5 mmol) in nitrobenzene (5 ml) provided 140 mg (74%)
of a brownish-white solid: mp >280ꢀC; IR (KBr) 3405,
4.1.7 5-Amino-2[20-(benzimidazol-500-yl)benzimidazol-
50-yl]benzimidazole (15)
1
3093, 2940, 1625, 1429, 1269, 1138; H NMR ꢃ 3.83 (s,
Compound 13 (35 mg, 0.09 mmol) was reduced by
hydrogenation with 10% Pd/C (7 mg) in ethyl acet-
ate:methanol (1:1, 50 ml) at 45 psi for 24 h. After passing
through a bed of celite, and concentrating in vacuo, the
crude mixture was separated on a silica-gel column
eluting with 10±20% methanol/ethyl acetate. 10 mg of
brownish solid (31%) was obtained: mp >280ꢀC; IR
(KBr) 3421, 3223, 2961, 2918, 1624, 1421, 1373; 1H
NMR (DMSO-d6+3 drops CF3COOH) ꢃ 7.46 (d, 1H,
J ˆ 9:0), 7.84 (s, 1H), 7.93 (d, 1H, J ˆ 8:5), 8.06±8.18
(m, 2H), 8.30 (dd, 1H, J ˆ 9:0, 2.0) 8.54 (dd, 1H,
3H), 6.85 (dd, 1H, J ˆ 8:5, 2.0), 7.12 (s, 1H), 7.51 (d,
1H, J ˆ 8:5), 7.73 (s, 1H), 7.79 (d, 1H, J ˆ 8:5), 8.05 (d,
1H, J ˆ 8:5), 8.15 (d, 1H, J ˆ 7:5), 8.38 (d, 1H, J ˆ 7:5),
8.39 (s, 1H), 8.48 (s, 1H): 13C NMR (DMSO-d6+3
drops CF3COOH) ꢃ 56.2, 96.5, 113.8, 115.0, 115.7,
115.8, 115.9, 116.4, 117.8, 122.9, 125.2, 125.2, 126.1,
126.3, 131.8, 133.0, 133.1, 139.0, 141.2, 149.4, 153.7,
158.4; HRMS (FAB) calcd for C22H17N6O (MH+)
381.1464, found 381.1451.
4.1.5 5-Nitro-2[20-(benzimidazol-500-yl)benzimidazol-50-
yl]benzimidazole (13)
4-Nitrophenylenediamine (87 mg, 0.57 mmol) and 5-
J ˆ 9:0, 1.5), 8.75 (s, 1H), 8.84 (s, 1H), 9.72 (s, 1H); 13
C
NMR (DMSO-d6+3 drops CF3COOH) ꢃ 110.7, 114.4,
115.4, 115.5, 115.6, 115.9, 116.2, 116.3, 118.9, 120.4,
124.9, 125.5, 131.6, 133.3, 133.4, 134.1, 137.3, 138.7,
140.0, 150.7, 153.1; HRMS (FAB) calcd for C21H16N7
(MH+) 366.1471, found 366.1484.
formyl-2-(benzimidazo-50-yl)benzimidazole
(100 mg,
0.38 mmol) in nitrobenzene (5 ml) provided 70 mg (47%)
of a brownish-white solid: mp >280ꢀC; IR (KBr) 3118,
1
1626, 1549, 1436, 1330, 1287, 1133; H NMR (DMSO-
d6+3 drops CF3COOH) ꢃ 7.89 (d, 1H, J ˆ 9:0), 8.02 (d,
1H, J ˆ 8:5), 8.15 (d, 1H, J ˆ 9:0), 8.23 (dd, 1H,
J ˆ 9:0, 2.0), 8.35 (dd, 1H, J ˆ 8:5, 1.5), 8.46 (dd, 1H,
J ˆ 9:0, 1.5), 8.57 (d, 1H, J ˆ 2:0), 8.66 (s, 1H), 8.78 (s,
1H), 9.65 (s, 1H); 13C NMR (DMSO-d6+3 drops
CF3COOH) ꢃ 112.0, 114.1, 114.3, 114.7, 114.9, 115.6,
4.2 Determination of capacity factor, k0
The capacity factor, k0, was experimentally-deter-
mined under similar conditions to those previously
described [18,19]. A Hewlett-Packard 1090 liquid chro-
matograph equipped with diode array UV detection was

J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
171
used for obtaining the capacity factors, k0, for com-
pounds used in the QSAR study and for determining
compound homogeneity of compounds for which
HRMS data are provided. The samples used in this
study were prepared by dissolving about 1 mg of the
compound in 100 ml of methanol (HPLC grade). The
Pipes bu€er was prepared by dissolving 1.52 g of PIPES
(purchased from Aldrich) in 1.0 l of 18 mꢀ water and
adjusting the pH to 7.4 with 1 N KOH using a pH
meter. Each sample in methanol was injected (50 ꢂl) on
a Vydac C-18 column at a ¯ow rate of 1.0 ml/min
employing a mobile phase of 30% Pipes bu€er (5 mM,
pH 7.4) in methanol. The detector wavelength was
330 nm. The mobile phase hold-up time, t0 was deter-
mined by injecting the mobile phase. Each sample was
injected three times and the average of the three reten-
tion times (tR) were used in calculating the capacity
factor, k0, from the equation
bu€er, it was important that separate calibration curves
were obtained. The initial studies were not reproducible
and this was attributed to the compounds sticking onto
the surface of the reactivials. After the glassware used in
the study was siliconized with Aquasil1 (purchased
from Pierce), the results were reproducible. Based on the
solubility studies of the two compounds, stock solutions
of each of the compounds used in the QSAR study were
prepared by accurately weighing about 1 mg of the
compounds and dissolving in 100 ml of HPLC grade
methanol in a volumetric ¯ask. 300 ꢂl of each stock
solution was evaporated in reactivials, and 1.5 ml of
Pipes bu€er (5 mM, pH 7.4), and 0.5 ml of octanol
saturated with the bu€er were added to the vials and
stirred for 24 h at 37ꢀC. The phases were separated and
200 ꢂl of the octanol layer was diluted to 50 ml with
methanol while 0.5 ml of bu€er layer was diluted to 2 ml
with Pipes bu€er. Additional dilutions were made as
required. Concentrations of both phases were deter-
mined by measuring the ¯uorescence and using the
standard curves obtained for each compound in both
phases. Standard curves for each compound for both
the octanol layer and the bu€er layer were obtained by
preparing separate standard solutions from the stock
solutions. The standard solutions for the octanol layer
were prepared by taking 200 ꢂl each of the stock solu-
tion, evaporating o€ the methanol with nitrogen gas,
and diluting with 0.2% octanol in methanol to the
desired concentrations. The standard solutions for the
bu€er layer were prepared similarly except that bu€er
was used for the dilutions. The partitioning study was
done in triplicates for all the compounds and log P was
determined from the average concentrations according
to the equation
k0 ˆ ꢀtR t0†=t0
4.3 Determination of partition coecient, log P
Partition coecients were performed using similar
methods to those previously detailed [20]. Pipes bu€er
was prepared by dissolving 1.52 g in 1.0 l of 18 mꢀ water
and adjusting the pH to 7.4 as described previously. The
octanol used in this study was presaturated with the
Pipes bu€er by stirring overnight at 37ꢀC and then
separating the layers in a separatory funnel. Both the
saturated octanol and bu€er were kept in a 37ꢀC water
bath throughout the study. Analysis was done on a
Shimadzu RF 5000U spectro¯uorophotometer using
excitation wavelength of 340 nm and emission wave-
length of 380 nm. The bandwidth was adjusted from
5 nm to 10 nm depending on the analysis. Before per-
forming the partition coecient study, the solubility of
the compounds in octanol and water had to be deter-
mined so that partitioning is not a€ected by saturation
of the compounds in either layers. However, because
sucient amount of compounds were not available,
only compounds 1 and 8 were used in determining the
solubility. Compounds 1 and 8 were each saturated in
2 ml of octanol and 2 ml of Pipes bu€er separately in
di€erent reactivials by continuously adding the com-
pounds until there was excess of solid that did not dis-
solve. The excess solid was ®ltered o€ with a glass ®ber
and the octanol or bu€er solutions were analyzed by
¯uorescence. The calibration curves for both the octanol
and bu€er solutions were obtained with known con-
centrations of the compounds. Several trial and errors
were repeated to obtain the calibration curve for the
range that ®ts the solubility of the compounds. Because
the ¯uorescence behavior of these terbenzimidazoles
were very di€erent in octanol from that in methanol or
log P ˆ log ‰cŠo=‰cŠw
where ‰cŠ ˆ concentration of compounds analyzed,
o ˆ octanol phase, w ˆ aqueous phase.
4.4 QSAR calculations
QSAR-PC [22] was used for the carrying out regres-
sion analyses. Data ®les were constructed using
DATAWELL where ꢀ, ꢁp, and ꢁm were chosen as the
variables. The correlation matrices were obtained by
using ALLREGRESS and multiple linear regression
QSAR computations were preformed using REGRESS.
4.5 Chromatography
The retention times needed to calculate k0 were
determined on a Vydac C18 column at a ¯ow rate of
1.0 ml/min employing a mobile phase of 30% Pipes
bu€er (5 mM, pH 7.4) in MeOH. The void volume was
determined by injecting methanol.

172
J. S. Kim et al./Bioorg. Med. Chem. 6 (1998) 163±172
4.6 Topoisomerase I-mediated DNA cleavage assays
References
[1] Wang JC. Annu Rev Biochem 1985;54:665.
[2] Liu L. Annu Rev Biochem 1989;58:351.
[3] D'Arpa P, Liu L. Biochim Biophys Acta 1989;989:163.
[4] Bodley AL, Liu LF. Bio/technology 1988;6:1315.
[5] Schneider E, Hsiang Y-H, Liu LF. Adv in Pharmacol
1990;21:149.
[6] Chen AY, Liu LF. Annu Rev Pharmacol Toxicol
1994;34:191.
[7] Wang JC. Annu Rev of Biochem 1996;65:635.
[8] Chen AY, Yu C, Bodley AL, Peng LF, Liu LF. Cancer
Res 1993;53:1332.
[9] Chen A, Yu C, Gatto B, Liu LF. Proc Natl Acad Sci USA
1993;90:8131.
[10] Beerman TA, McHugh MM, Sigmund R, Lown JW,
Rao KE, Bathini Y. Biochim Biophys Acta
1992;1131:53.
Human topoisomerase I was expressed in E. coli and
puri®ed as previously described [23]. Plasmid YEpG was
also puri®ed by the alkali lysis method followed by
phenol deproteination and CsCl/ethidium isopycnic
centrifugation as described [24,25]. The end-labeling of
the plasmid was accomplished as previously described
[25]. The cleavage assays were performed as previously
reported [8]. The drug and DNA in the presence of
topoisomerase I was incubated for 30 min at 37ꢀC. After
development of the gels, 24 h exposure was typically
used to obtain autoradiographs outlining the extent of
DNA fragmentation.
[11] Sun Q, Gatto B, Yu C, Liu A, Liu LF, LaVoie EJ. Bioorg
Med Chem Lett 1994;4:2871.
4.7 Cytotoxicity assay
[12] Sun Q, Gatto B, Yu C, Liu A, Liu LF, LaVoie EJ. J Med
Chem 1995;38:3638.
[13] Kim JS, Gatto B, Yu C, Liu A, Liu LF, LaVoie EJ. J Med
Chem 1996;39:992.
[14] Kim JS, Yu C, Liu A, Liu LF, LaVoie EJ. J Med Chem
1997;40:2818.
[15] Hansch C, Leo A, Hoekman D. Exploring QSAR-hydro-
phobic, electronic, and steric constraints; American Che-
mical Society, Washington, DC, 1995.
The cytotoxicity as listed for several of the com-
pounds which were prepared as part of this study was
determined using Human lymphoblast RPMI 8402 cells
and the MTT-microtiter plate tetrazolium cytotoxicity
assay (MTA) [26±28]. The cytotoxicity assay was per-
formed using 2000 cells/well, in 200 ꢂl of growth med-
ium, which were grown in suspension at 37ꢀC in 5%
CO2 and maintained by regular passage in RPMI med-
ium supplemented with 10% heat inactivated fetal
bovine serum, L-glutamine (2 mM), penicillin (100 U/
ml), and streptomycin (0.1 mg/ml). The cells were
exposed continuously for 4 days to di€erent drug con-
centrations, and were assayed at the end of the fourth
day. Each assay was performed with a control which did
not contain any drug. All assays were performed at least
twice in six replica wells.
[16] Craig PN. J Med Chem 1971;14:680.
[17] Rekker RF, Mannhold R. Calculations of drug lipophili-
city; VCR: New York, 1992.
[18] Loewe H, Urbanietz J. Arzneim-Forsch 1974;24:1927.
[19] Nowotnik DP, Feld T, Nunn AD.
1993;630:105.
J Chromatogr
[20] Minick DJ, Sabatka JJ, Brent DA. J Liq Chromatogr
1987;10:2565.
[21] Dearden JC, Bresnen GM. Quant Struct±Act Relat
1988;7:133.
[22] Coburn RA. QSAR-PC:PAR; BIOSOFT, Cambridge,
1987.
[23] Gatto B, Sanders MM, Yu C, Wu H-Y, Makhey D,
LaVoie EJ, Liu LF. Cancer Res 1996;56:2759.
[24] Maniatis T, Fritisch EF, Sambrook J. Molecular Cloning,
a Laboratory Manual; Cold Spring Harbor Laboratory:
Cold Spring Harbor, 1982:149±185.
[25] Liu LF, Rowe TC, Yang L, Tewey KM, Chen GL. J Biol
Chem 1983;258:15365.
[26] Mosmann TJ. J Immunol Methods 1983;65:55.
[27] Carmichael J, DeGra€ WG, Gazdar AF, Minna JD,
Mitchell JB. Cancer Res 1987;47:936.
Acknowledgements
Mass spectrometry was provided by the Washington
University Mass Spectrometry Resource, an NIH
Research Resource (Grant No. P41RR0954). This study
was supported by Grant CA 39662 from the National
Cancer Institute (L. F. L.) and a fellowship grant from the
Johnson & Johnson Discovery Research Fund (E. J. L.).
[28] Denizot F, Lang R. J Immunol Methods 1986;89:271.