Structure-activity relationships in a series of bisquaternary bisphthalimidine derivatives modulating the muscarinic M2-receptor allosterically

Article abstract of DOI:10.1021/jm991136e

Hexane-bisammonium-type compounds containing lateral phthalimide moieties are well-established ligands of the common allosteric binding site of muscarinic M₂ receptors. Previous structure-activity relationships (SAR) revealed two positively charged centers and two lateral phthalimide moieties in a defined arrangement to be essential of a high allosteric potency. The purpose of this study was to replace one carbonyl group of the phthalimides with hydrogens, hydroxy, alkoxy, phenyl, benzyl, and benzylidene groups in order to check the influence of these substituents on the allosteric activity in antagonist-linked receptors. The analysis of the quantitative SAR indicated that a high allosteric potency is related to a certain amount of rigidity as well as polarizibility and the ability to form hydrophobic interactions.

Full text of DOI:10.1021/jm991136e

J . Med. Chem. 2000, 43, 2155-2164
2155
Str u ctu r e-Activity Rela tion sh ip s in a Ser ies of Bisqu a ter n a r y Bisp h th a lim id in e
Der iva tives Mod u la tin g th e Mu sca r in ic M2-Recep tor Alloster ica lly
Hector M. Botero Cid, Christian Tra¨nkle,† Knut Baumann, Rainer Pick, Elisabeth Mies-Klomfass,†
Evi Kostenis,† Klaus Mohr,† and Ulrike Holzgrabe*
Pharmaceutical Chemistry, Institute of Pharmacy, University of Wu¨rzburg, Am Hubland, 97074 Wu¨rzburg, FRG, and
Pharmacology and Toxicology, Institute of Pharmacy, University of Bonn, An der Immenburg 4, 53121 Bonn, FRG
Received December 2, 1999
Hexane-bisammonium-type compounds containing lateral phthalimide moieties are well-
established ligands of the common allosteric binding site of muscarinic M2 receptors. Previous
structure-activity relationships (SAR) revealed two positively charged centers and two lateral
phthalimide moieties in a defined arrangement to be essential of a high allosteric potency.
The purpose of this study was to replace one carbonyl group of the phthalimides with hydrogens,
hydroxy, alkoxy, phenyl, benzyl, and benzylidene groups in order to check the influence of
these substituents on the allosteric activity in antagonist-linked receptors. The analysis of the
quantitative SAR indicated that a high allosteric potency is related to a certain amount of
rigidity as well as polarizibility and the ability to form hydrophobic interactions.
In tr od u ction
compound W84 were replaced with corresponding satu-
rated skeletons.12 This structural change resulted in a
decrease of allosteric potency. Reduction in size de-
creases the potency by a factor of 400. Thus, we decided
to keep the main part of the aromatic phthalimide ring
but to replace, on one hand, one carbonyl functional
group with two hydrogens as well as hydroxy and alkoxy
groups and with benzyl and phenyl substituents. On the
other hand, different benzylidene isomers (E and Z) of
the phthalimide skeleton were synthesized in order to
investigate the possible stereoselectivity of the allosteric
modulation. Finally, an analysis of quantitative struc-
ture-activity relationships (QSAR) was performed in
order to explain the potency of the newly synthesized
compound and to further characterize the allosteric
binding site.
Several bisquaternary compounds belonging to dif-
ferent pharmacological groups are known to retard the
dissociation of antagonists, such as N-methylscopola-
mine (NMS), from the muscarinic acetylcholine recep-
tor.1,2 This effect, being indicative of an allosteric
modulation of the antagonist receptor complex, is caused
by the binding of an allosteric modulator in a receptor
whose orthosteric site is occupied by the antagonist.3,4
The comparison of the allosteric modulators alcuro-
nium,5 tubocurarine,6 gallamine,7 and W848 (a bisph-
thalimide compound derived from hexamethonium),
utilizing the Kohonen neural network, revealed the
dependence of the allosteric potency on the molecular
electrostatic potential (MEP) as well as steric param-
eters.9 To elucidate the binding mode a 3D quantitative
structure-activity analysis was subsequently performed
in an extended series of structurally diverse compounds
of different allosteric potency.10 Using the features of a
molecular shape analysis, alcuronium was taken as a
template molecule for comparison of W84-derivatives
and several bispyridinium-derived compounds (i.e.
WDUO, DUO-O, IWDUO). The pharmacophoric ele-
ments turned out to be two positively charged nitrogens
and two aromatic areas arranged in a sandwich-like
geometry. The biological activity within this extended
series was found to be related to the nonoverlap volume
and the MEP, described as integrated potential field
differences. These findings supported the hypothesis
that the allosteric binding site seems to have space for
voluminous substituents especially on the lateral het-
erocycle moieties of the bisquaternary molecules.11 The
purpose of this study is to check the influence of the
size, lipophilicity, and electronic properties of substit-
uents in position 3 of the phthalimide. In a preceding
step the aromatic lateral heterocycles of the lead
Ch em istr y
Syn th esis. The synthesis of the phthalimidine de-
rivative 2 started off from the lead compound W8413
(Scheme 1). Reduction using Zn powder in glacial acetic
acid and 10% propionic acid gave the methylene com-
pound 2 in a rather good yield. The synthesis of the
alkoxyphthalimidines started off from N-bromopropyl-
phthalimide. Selective reduction of one of the two
carbonyl functions was achieved by means of sodium
borohydride in methanol. The hydroxy group of the
obtained 3 was replaced with the appropriate alkoxy
groups by a nucleophilic substitution using the corre-
sponding alcohols.14 Then the exhaustive alkylation of
N1,N1,N,6N6-tetramethyl-1,6-hexanediamine with two
molecules of 4b-d gave the bisquaternary compounds
5b-d by refluxing for 5-24 h in a polar solvent. The
corresponding hydroxy compound 5a was achieved by
heating 2 mol of 3 with 1 mol of the bisamine in acetone.
The hydroxy groups in 5a were replaced with isoprop-
ylmercapto groups by refluxing in propanethiole under
acidic conditions to give 5e.
* Corresponding author. Tel: 0931/8885460. Fax: 0931/8885494.
E-mail: holzgrab@pharmazie.uni-wuerzburg.de.
† University of Bonn.
The introduction of a phenyl or methyl group into
position 3 of the phthalimide skeleton (Scheme 2) was
10.1021/jm991136e CCC: $19.00 © 2000 American Chemical Society
Published on Web 05/16/2000

2156 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11
Sch em e 1. Synthesis Pathway of the Compounds 5
Botero Cid et al.
Sch em e 2. Synthesis Pathway of the Compound 9
achieved by partially utilizing the synthesis pathway
described by Gabriel and Naumann.15 o-Benzoylbenzoic
acid was converted to phenylphthalazone 6 using hy-
drazine sulfate in sodium hydroxide solution. Subse-
quent reduction with Zn powder in hydrochloride acid
gave the 3-phenylphthalimidine 7, which was N-alky-
lated with 1,3-dibromopropane in the presence of so-
dium hydride in tetrahydrofurane to achieve 8. It was
impossible to obtain the corresponding methyl deriva-
tive because only a mixture of many products could be
isolated from the reaction solution. Two molecules of 8
were connected with N1,N1,N,6N6-tetramethyl-1,6-hex-
anediamine by refluxing in acetonitrile to give 9.
To obtain the benzylidene and benzyl phthalimidine
derivatives 12 and 15, respectively, phthalic anhydride,
phenylacetic acid, and sodium acetate were heated to
250-300 °C to give the Z isomer of 3-benzalphthalide
10 (Scheme 3).16 The oxygen of the phthalide was
replaced with a nitrogen by refluxing with dimethyl-
aminopropylamine and subsequent dehydratization us-
ing formic acid. The so obtained 11 is a mixture of about
90% E and 10% Z isomer which was separated by means

SAR of Bisquaternary Bisphthalimidine Derivatives
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11 2157
Sch em e 3. Synthesis Pathway of the Compounds 12 and 15
of column chromatography. Each isomer was converted
to the corresponding bisquaternary benzylidene com-
pounds 12a ,b (E/E, Z/Z) with 1,6-dibromohexane. In
case of the E/Z isomer 12c, the E benzylidene phthal-
imidopropylamine was selectively alkylated with 1,6-
dibromohexane to give 13. At last, the Z-isomer is
connected to the opposite end of the hexamethonium
middle chain. For the benzyl derivative 15 the mixture
of isomers of 11 was hydrogenated with Pd/C/H2. Two
molecules of 14 were connected with 1,6-dibromohexane.
Ster eoch em istr y. Since the reduction of bromopro-
pylphthalimide created a center of chirality (f3; Scheme
1), it is expected that the last step of the synthesis
pathway will result in a mixture of two enantiomers and
a meso isomer (5a -e). We tried to find out the ratio of
the isomers by means of chiral column chromatography
and NMR spectroscopy (see Table 5, Experimental
Section).
formed. The defined melting points measured in each
case may further support this hypothesis. Because the
compounds 5 did not exhibit a higher potency than the
lead compound W84 (see below) and, due to N,O-acetal
function, the center of chirality is supposed to be
stereochemically unstable, a separation of the isomers
was not performed.
Accordingly, for the phenyl-substituted compound 9
only one set of signals could be detected in the NMR
spectra. Interestingly, the signals for the hydrogens next
to the imide nitrogens are very separated (3.32 versus
3.74 ppm; see Table 1). This is caused by the phenyl
ring in position 3, which is orientated to the bisquater-
nary middle chain, and in line with that, one of the
NPhth-methylene hydrogens is influenced by the ring
current effect resulting in a shielding of the hydrogen.
The same was found to be true for the benzyl derivative
15.
The 1H and 13C NMR of all chiral compounds does
neither in CDCl3 nor in DMSO-d6 show two sets of
signals corresponding to the racemate and the diaster-
eomeric meso-compound. Additionally, neither with
chiral shift reagents, such as Eu(tfc)3, nor with different
cyclodextrin derivatives could two or three sets of
signals be detected (although a complexation was ob-
served in either case). Column chromatography using
a chiral stationary phase, polymer phenylethylacryla-
mide, which has already been successfully applicated
in a similar case,11 shows only one broad peak. Similar
results were obtained with cyclodextrin-modified capil-
lary electrophoresis. Since there was no evidence of
diastereomers in either method, it may be assumed that
either the racemic mixture or the meso isomer was
The configuration of the isomers of 12 was determined
by means of NOE experiments. Saturation of the signal
of the methine proton in 12a resulted in an enhance-
ment of the NPhth-CH2CH2 signals and vice versa, which
indicates a spatial neighborhood of these groups, and
thus an E-configuration. Additionally, the signals of the
propylene chain, the N-methyl group, and partly the
hexane chain of the Z-isomer (12b) are upfield shifted
in comparison with the signals of the E-isomer (12a ).
The shielding is found to be most pronounced for the
NphthCH2CH2CH2N+- (∆δ ) 0.6-0,7 ppm). In turn, the
aromatic hydrogen next to the benzylidene group are
upfield shifted in case of the E-isomer (12a ) in compari-
son with the Z-isomer. These effects are caused by the
shielding ring current effect of the phenyl ring, which

2158 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11
Ta ble 1. Substituents and log(1/EC50) of the Compounds Studieda
Botero Cid et al.
compd
W84
2
5a
5b
5c
5d
5e
R1, R2
log(1/EC50) ( SEM
surf (Å2)
vol (Å3)
log P
MR (Å3)
pol (Å3)
dip (D)
ind
222
dO, -
H, H
OH, H
OCH3, H
5.85 ( 0.05
5.70 ( 0.18
5.22 ( 0.15
5.17 ( 0.35
5.64 ( 0.11
5.80 ( 0.47
5.90 ( 0.67
6.10 ( 0.04
7.15 ( 0.07
7.07 ( 0.05
7.18 ( 0.02
6.00 ( 0.04
274.15
276.71
271.40
310.47
340.02
354.39
376.16
313.26
342.99
341.34
342.17
353.88
489.12
490.96
506.08
554.78
611.20
639.61
663.70
677.27
713.02
709.35
711.19
739.08
1.46
1.55
1.45
1.72
2.07
2.48
2.92
3.40
2.96
2.96
2.96
3.65
43.51
43.39
44.42
49.17
53.91
58.33
64.51
67.67
73.81
73.81
73.81
72.42
13.38
12.80
13.28
14.78
16.07
17.32
19.18
20.60
23.97
24.58
24.28
21.87
2.878
3.729
3.167
3.840
3.932
3.675
4.174
3.779
3.194
3.040
3.117
3.746
1
0
0
0
0
0
0
0
1
1
1
0
13
9
9
9
9
OC2H5, H
O-i-C3H7, H
S-i-C3H7, H
C6H5, H
dCH-C6H5, -
dCH-C6H5, -
dCH-C6H5, -
CH2-C6H5, H
9
9
9
16
22
22
22
15
12a (E/E)
12b (Z/Z)
12c (Z/E)
15
a
Symbols: surf., surface; vol, volume; log P, ocatnol/water partition coefficient; MR, molar refractivity; pol, mean R-polarizability
(MOPAC); dip, dipole moment (MOPAC); ind, indicator variable for a double bond in position 3; 222, count of sp2-2-sp2 substructures
(cf. text).
of magnitude as the lead compound W84. Almost all
alkoxy compounds 5 are less active than W84. At a first
glance, within this series the potency seems to increase
with the increasing size of the substituents. However,
the phenyl- and benzyl-substituted compounds 9 and
15, respectively, exhibit the same potency as W84. But
even more interesting, the introduction of a double bond
in position 3, which makes the lateral moiety more rigid,
enhances the potency by more than 1 order of magni-
tude. To quantitatively explain the differences in the
allosteric potency of the compounds, studied herein, a
QSAR analysis was performed. Table 1 shows the
compounds under scrutiny, their corresponding poten-
cies, and molecular descriptors.
F igu r e 1. [3H]NMS dissociation from muscarinic M2 receptors
under control conditions and in the presence of the indicated
test compounds in Na,K,Pi buffer. Note that obidoxime at-
tenuates the allosteric action of 12a . Shown is a representative
set of experiments. Monoexponential curve fitting.
Since, in previous studies, the allosteric potency was
found to depend on the steric bulk and on the lipophi-
licity, parameters were considered which characterize
these properties, e.g. the STERIMOL parameters L, B1,
and B5 as well as MR (molar refractivity) and π.
Unfortunately, these parameters could not be found for
all substituents in the literature.18-20 To apply compa-
rable parameters, they were computed using the Hy-
perChem/ChemPlus package (see the methods). As a
model, corresponding N-methylphthalimide moieties
were built up using the model builder and the MM+
program. Next, the octanol/water partition coefficient
(log P), the steric parameters, volume and surface area,
as well as polarizability and refractivity, which is a
function of steric bulk and polarizability, were calcu-
lated. Moreover, an indicator variable which takes a
value of 1 if there is a double bond in position 3 and a
zero otherwise was added to the set of descriptors. This
indicator variable accounts for the altered rigidity of the
lateral moiety. After an initial screening of the descrip-
tors it turned out that polarizability and the indicator
variable were the most important descriptors. To im-
prove the QSAR models we recalculated the polariz-
ability of the substituents on a higher level of sophis-
tication using MOPAC21 and replaced HyperChem’s
estimate of the polarizability by MOPAC’s mean R-po-
larizability.22 The MOPAC calculation also yielded the
dipole moment of the substituents, which was also
included in the analysis. Furthermore, we added a
is orientated either to the hexamethonium chain (Z) or
to the phthalimidine ring (E). Similar results were
reported by Marsili et al.17 for corresponding benzylide-
nephthalimidines.
P h a r m a cology. The ability of the compounds to
retard the dissociation of the radioligand [3H]N-meth-
ylscopolamine ([3H]NMS) was measured in porcine
cardiac M2 receptors. All compounds reduced the ap-
parent rate constant of [3H]NMS dissociation k-1 con-
centration-dependently. The concentration of a test
compound inducing a reduction of k-1 to 50% of the
control value (EC50) served as a measure of the allosteric
effect (Table 1). At the EC50 concentration, occupancy
of the NMS-receptor complexes by the allosteric agent
can be considered as half-maximal. In this case, the
EC50 value is equivalent to the KD value of alloster
binding.4 To check the topology of binding, the most
potent compound (12a ) was tested to determine whether
the allosteric retardation of [3H]NMS dissociation was
sensitive toward the “allosteric antagonist” obidoxime.
The experimental design is illustrated in Figure 1.
Resu lts a n d Discu ssion
The allosteric potency of the phthalimidine deriva-
tives 2, 5a -e, 9, and 15 is found to be on the same order

SAR of Bisquaternary Bisphthalimidine Derivatives
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11 2159
Ta ble 2. Correlation Matrix of the Final Set of Descriptorsa
no.
variable
log(1/EC50
)
surf
vol
log P
MR
pol
dip
ind
1
2
3
4
5
6
7
8
surf
vol
log P
MR
pol
dip
ind
222
0.432
0.707
0.611
0.806
0.877
-0.478
0.787
0.939
0.837
0.747
0.759
0.691
0.396
0.009
0.264
0.952
0.982
0.946
0.084
0.240
0.668
0.938
0.877
0.172
0.113
0.596
0.987
-0.044
0.371
-0.179
0.498
0.859
-0.822
-0.591
0.778
0.800
a
The cluster of highly intercorrelated descriptors is set in bold letters.
Ta ble 3. Best One- and Two-Parameter QSAR Models, Ranked According to Q2 a
no.
x1
x2
sPress
Q2
s
R2
prand(Q2)
prand(R2)
â0
â1
â2
1
2
3
4
5
6
7
ind
222
ind
222
pol
pol
surf
MR
0.285
0.286
0.311
0.296
0.363
0.364
0.376
0.865
0.864
0.839
0.838
0.781
0.780
0.762
0.202
0.222
0.211
0.253
0.273
0.239
0.250
0.932
0.918
0.926
0.882
0.876
0.905
0.896
0.0002
0.0012
0.0002
0.0028
0.0032
0.0026
0.0042
0.0002
0.0002
0.0002
<0.0002
0.0002
0.0002
<0.0002
3.988
3.265
3.814
4.462
5.672
3.165
4.558
0.664
0.111
0.807
0.117
0.127
0.933
1.036
0.100
0.004
0.033
dip
vol
log P
-0.557
0.004
0.471
ind
ind
a
Symbols: xi, variables included in the model; the numbering scheme follows the one given in Table 2; sPRESS, cross-validated standard
deviation (the square root of PRESS (predicted error sum of squares) divided by the degrees of freedom); Q2, cross-validated coefficient
of determination; s, standard deviation (the square root of the residual sum of squares divided by the degrees of freedom); R2, coefficient
of determination; prand(Q2), probability that a scrambled response vector yields a higher Q2 than the original one; prand(R2), same as
p
rand(Q2) for R2; âi, regression coefficient for the ith variable; â0, intercept. All reported regression coefficients are significant (t-test, R )
0.1, two-sided).
particular SE-descriptor,23,24 referred to as sp2-2-sp2
(abbr, 222), expected to characterize the rigidity of the
subsituents even more thoroughly. The 222-descriptor
counts the number of unique substructural fragments
that start on a sp2-hybridized atom and end on a sp2-
hybridized atom separated by two bonds (e.g. H2Cd
CH-CHdCH2 f 222 ) 2). This descriptor can also be
viewed as a special atom pair.25 In this series of
molecules it characterizes, among other things, the
overall rigidity of the substituent, rather than just the
hybridization in position 3 like the indicator variable.
Moreover, it does not only characterize the number of
double bonds but also the arrangement of double bonds
in the molecule. A correlation matrix of the final set of
descriptors plus log(1/EC50) is shown in Table 2. The
correlation matrix deserves some comments: First, the
single variable correlation between 222 and log(1/EC50)
is very good. The corresponding regression model yields
a coefficient of determination (R2) of 0.882 and a cross-
validated coefficient of determination (Q2) of 0.838 and
is by far the best one parameter regression model that
can be built for the data set with this set of descriptors.
Second, there is a cluster of four descriptors, namely,
volume, log P, refractivity, and polarizability, that are
highly intercorrelated (r > 0.87). There was no sub-
stituent in this series that breaks this correlation. For
hydrophobic substituents it is known that there exists
a close interrelation between volume, surface, refractiv-
ity, and lipophilicity.26 As a consequence, we anticipate
that it will be difficult to distinguish the effect of one of
these descriptors from the others.
0.5) are shown. Additionally, the best one-parameter
regression model is also given. All regression models
given in Table 3 were computed with the entire set of
structures (n ) 12). The statistical parameters of the
equations are good, and particularly, the internal
predictive power (Q2, sPRESS) of the models is high (Q2
> 0.76). Nonetheless, it must be noted that all models
including the indicator variable predict (cross-valida-
tion) either compound W84 or compound 2 substantially
worse than the remaining compounds. Note also that
the high R2 and Q2 values indicate better models than
they actually are. R2 and Q2 are slightly inflated because
the data set is clustered into few highly potent and
many less potent compounds. Plots of the experimental
versus the predicted activities for the best two models
are included in the Supporting Information.
Variable selection may be prone to chance correlation.
Topliss and Edwards27 showed that screening a large
number of variables which in fact were not related to
the response (since they were random numbers) often
resulted in satisfactory regression equations. In this
study, the probability of chance correlation was reduced
by selecting variables that maximized Q2 rather than
selecting variables that maximized R2. Put another way,
the predictive power of the models was optimized and
not the fit. Nevertheless, to thoroughly assess the
probability that the reported models may have been
generated just by chance we also ran a permutation
test.27,28 The employed permutation test is based on the
repetitive randomization of the response vector. In each
cycle of the test the response vector is randomly rear-
ranged, the regression is carried out on the scrambled
data, and Q2 and R2 values are recorded for each cycle.
If the majority of Q2 and R2 values of the scrambled data
sets is much lower than the Q2 and R2 of the original
data set it can be concluded that the derived model is
relevant. The number of permutations for each test was
5000. The probability that a scrambled response vector
yields a larger Q2 or R2 value than the original vector
Since the object-to-variable ratio is rather low (12/7)
only a subset of the descriptors given above should be
used to derive a QSAR. Since all one-parameter regres-
sion models apart from the aforementioned one are
unsatisfactory with respect to Q2 (Q2 < 0.5), we com-
puted all two-parameter regression models. In Table 3
all two-parameter models with Q2 values larger than
0.7 and descriptor pairs correlated less than 0.5 (r <

2160 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11
Botero Cid et al.
is recorded in the variables prand(Q2) and prand(R2) of
Table 3. If a scrambled vector gave rise to a larger Q2
and R2 value, the correlation between the scrambled
vector and the original one was generally rather high.
Hence, the probability that the reported models may
have been generated just by chance is low.
(A)
What do we learn from the QSAR models about the
structural requirements for allosteric potency? First of
all, it is worth noting that each model contains a
descriptor characterizing rigidity (Ind or 222). Even the
222-descriptor alone yields a good model. Thus, we
hypothesize that a certain amount of rigidity of the
lateral substituent is important for allosteric potency.
The fact that rigidity is important can most easily be
seen by comparing the potency of the benzylidene-
substituted compounds 12 to the benzyl-substituted
compound 15. Compound 15 was found to be 1 order of
magitude less potent than compounds 12. Since the
most striking difference between both compounds is the
higher rigidity of 12 created by the double bond, it is
likely that the higher affinity of 12 is caused by a
smaller loss of conformational entropy29 on binding to
the receptor molecule.
(B)
F igu r e 2. (a) Obidoxime-induced attenuation (open diamonds,
right ordinate) of the effect of 12a (0.1 µM, filled diamonds,
left ordinate) on [3H]NMS dissociation. Left ordinate: The
effect of 12a alone (filled diamonds) is expressed as percentage
of the maximum possible effect, i.e., k-1 ) 0. Right ordinate:
Effect of 0.1 µM 12a in the presence of increasing concentra-
tions of obidoxime; the respective k-1 in the presence of
obidoxime alone was set to k-1,normalized ) 1.0. Simultaneous
curve fitting according to ref 28. (B) Schild plot for 12a ; dose
ratios DR ) EC0.5,obidoxime/EC0.5,control and the line were derived
from Figure 4A. W84 data were taken from ref 3.
Second, in the case of the indicator variable, good
models are obtained if it is combined with polarizability,
molar refraction, volume, or log P. This was actually
within expectation, since these parameters are highly
intercorrelated. Thus, we expect that all four param-
eters in combination with the indicator variable encode
one underlying phenomenon. Whether, the set of pa-
rameters encodes the substituents’ ability to interact
with the target receptor via van der Waals forces
(polarizability, molar refraction) or via hydrophobic
interactions (log P) cannot be determined unequivocally
from these models. High polarizability is usually ad-
vantageous for induction forces (permanent dipole-
induced dipole) or dispersion forces (instantaneous
dipole-induced dipole), whereas subsituents with a high
log P value are likely to exhibit hydrophobic interac-
tions.29,30 Model 6, which uses polarizability and dipole
moment as explanatory variables can, however, shed
some light on the interaction forces. Since dipole mo-
ment enters the respective regression equation with a
negative sign, more polar molecules are less potent
according to this equation. Thus, it is more likely that
nonpolar interactions (dispersion forces) and hydropho-
bic interactions drive the binding process rather than
inductive interactions. The latter type of interaction
would require high polarizability and high dipole mo-
ment.
of the prototype allosteric agents using the common
allosteric site of M2, receptors was antagonized by
obidoxime, a similar pKB ) 4.61 ( 0.05, n ) 5, was
found. The equal sensitivity of 12a and W84 against
obidoxime suggests that 12a acts at the same site as
W84 does, i.e., the common allosteric site.
Since the geometric isomers of the benzylidene-
substituted compound 12, the EE-, the ZZ-, and the EZ-
isomer, exhibit the same allosteric potency, the interac-
tion of these groups with the receptor is not stereo-
selective. As can clearly be seen in Figure 3, in these
compounds the phenyl rings in position 3 are pointing
to different directions in space when adopting the
pharmacophoric conformation described in the Intro-
duction. On the other hand, it is conceivable that the
benzylidene system instead of the phthalimide occupies
the aromatic area of the pharmacophore. To check this
hypothesis molecular modeling studies were performed
using the already developed concept10 using alcuronium
as a template. Since all hexamethonium compounds
proved to be highly flexible,10 no systematic search for
a global minimum energy conformation was performed.
The matching procedure onto alcuronium revealed
rather good fits for all diastereomers of 12 using both
aromatic rings, either the phthalimidine or the ben-
zylidene (representatively shown for 12a in Figure 3).
The rms (root mean square) values calculated for six
atoms are less than 1 and the energy of the conforma-
tions is similar in either case. It is tempting to speculate
that both conformations of each isomer can be accom-
modated in the receptor protein, which is an indication
of a sort of multiple binding mode phenomenon:32
Compounds 12 appears to have a 20-fold higher
affinity for NMS-occupied receptors than the parent
compound W84. Previously, it was shown that structur-
ally closely related compounds may reveal a divergent
mode of allosteric action.3 In view of the large size of
the benzylidene moieties, we aimed to check whether
12a still utilizes the common allosteric site.31 Therefore,
it was tested whether the allosteric action of 12a is
sensitive toward the “allosteric antagonist” obidoxime
(Figure 2a,b).3 Obidoxime interfered with the action of
12a in a competitive fashion. The equilibrium dissocia-
tion constant of obidoxime binding derived from the
experiments amounted to pKB ) 4.34 ( 0.52, n ) 4
(mean ( SEM). When the action of W84, which is one

SAR of Bisquaternary Bisphthalimidine Derivatives
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11 2161
Ta ble 4. Reaction and Physical Data of Compounds 2, 5a -e,
9, 12a -c, and 15
reaction temp yield
mp
compd mmol
solvent
HOAc
time (h) (°C)
(%)
(°C)
2
0.5
4
reflux
45
246
propionic acid
acetone
ethanol
acetone
MeCN
i-propanethiol
MeCN
MeCN
5a
5b
5c
5d
5e
22
0.5
reflux
reflux
reflux
reflux
reflux
rta
35
22
28
53
60
40
57
64
52
54
162-65
83
60
3.5
2.0
3.2
1.4
1.1
2.0
3.2
24
18
5.0
2.5
72
130
148-54
212-15
234
223
230
9
12b
12a
12c
15
4-5
80
MeCN
4-5
80
0.17 MeCN
1.8 MeCN
72
rta
4-5
80
255
a
rt ) room temperature.
and hydrophobic interactions were deemed important
for ligand-receptor interaction. Thus, annellation of
further aryl rings to either the imide or the benzylidene
moiety should increase the potency. Corresponding
investigations are planed.
Since no stereoselectivity concerning the substituent
in position 3 could be detected, the allosteric site at the
receptor is supposed to be rather large and is likely to
be located at the entrance of the ligand binding pocket
of the muscarinic receptor protein.
Exp er im en ta l Section
Melting points were determined with a Dr. Tottoli point
apparatus (Bu¨chi) and were not corrected. 1H and 13C NMR
spectra were recorded on Varian XL 300 (1H NMR, 299.956
MHz; 13C NMR, 75 MHz) spectrometers. The centers of the
peaks of CDCl3 and DMSO-d6 were used as internal references.
Abbreviations for data quoted are d, doublet; t, triplet; q,
quartet; quin, quintet; and m, multiplet. Coupling constants
are given in hertz. IR spectra, recorded as KBr disks, were
obtained using a Perkin-Elmer 298 spectrometer. Dry solvents
were used throughout. Chemicals were of analytical grade and
purchased from Aldrich, Steinheim, or Merck, (Darmstadt,
FRG). Drugs: [3H]NMS was purchased from NEN-Dupont
(Bad Homburg, FRG). The other drugs, atropine sulfate and
(-)-scopolamine methylbromide, were obtained from Sigma
Chemicals (Mu¨nchen, FRG). The analytical results of the
compounds are within (0.4% of the theoretical values.
Ch em istr y. 2-{3-[1-(6-{1,1-Dim eth yl-1-[3-(1-oxo-2,3-d i-
h yd r o-1H -2-isoin d olyl)p r op yl]a m m on io}h exyl)-1,1-d i-
m eth yla m m on io]p r op yl}-1-isoin d olin on e Dibr om id e (2).
W84 (0.35 g, 0.5 mmol) and Zn powder (1.8 g, 27.6 mmol) were
suspended in 25 mL of glacial acid and 2.5 mL of propionic
acid. The mixture was refluxed for 4 h and hot and cold
filtered, and the filtrate was evaporated to dryness under
reduced pressure at 60 °C. The obtained oil was crystallized
and recrystallized from ethanol. For analytical and spectro-
scopic data, see Tables 4 and 5.
N-(3-Br om opr opyl)-3-h ydr oxy-1-isoin dolin on e (3). Solid
sodium borohydride (0.7 g, 18.5 mmol) was carefuly added to
a solution of N-(3-bromopropyl)phthalimide (10.0 g, 37.3 mmol)
in dried methanol under cooling. The solution was allowed
stand at room temperature for 1 h. Afterward water was added
and the resulting mixture was extracted with methylene
chloride. The organic layer was dried over Na2SO4 and
evaporated in vacuo. The oil was crystallized from ethanol/
benzine to give 33% 3, mp 94 °C.
Gen er a l P r oced u r e for th e Syn th esis of N-(3-Br o-
m op r op yl)-3-m et h oxy-1-isoin d olin on e (4b ), N-(3-Br o-
m op r op yl)-3-eth oxy-1-isoin d olin on e (4c), a n d N-(3-Br o-
m op r op yl)-3-isop r op oxy-1-isoin d olin on e (4d ). A mixture
of N-(3-bromopropyl)-3-hydroxy-1-isoindolinone (3.62 g, 13,4
mmol), p-toluenesulfonic acid (0.22 g, 1.3 mmol), and the
corresponding alcohol (26 mmol) was heated under reflux for
F igu r e 3. Alignment of 12(EE) onto alcuronium using the
positively charged nitrogens and the two aromatic carbon
atoms at alcuronium and the phthalimidine moiety (a) or the
benzylidene ring (b).
Herein, the ligand uses the same points of attachment
of the receptor protein, but different moieties of the
ligands are involved.
The fact that the geometric isomers of 12 exhibit the
same allosteric potency might be due to the above-
described flexibility of the hexamethonium middle
chain, which enables the molecule to drive the phar-
macophoric elements in the right position upon binding
to the receptor protein. By analogy, it might be con-
cluded that the optical isomers of 5, 9, and 15 would
also exhibit the same allosteric potency. However, the
potency is low and it was, thus, not worthwhile to
separate the isomers.
Con clu sion
Summing up, electrostatic, van der Waals, and hy-
drophobic interactions are crucial for receptor recogni-
tion. The two positive charges in the middle of the
molecules are likely to be responsible for approaching
the receptor because they form long-range ion-ion
interactions. Substituents of the lateral phthalimido
rings in position 3 sensitively influence the allosteric
potency, indicating an additional point of interaction.
This was shown with benzylidene-substituted com-
pounds. These compounds are much more potent than
previously known allosteric modulators. Rigidity, result-
ing in a smaller loss of conformational entropy upon
binding to the receptor protein, polarizability, offering
the possibility of dispersive van der Waals interactions,

2162 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11
Botero Cid et al.
8 h. After evaporation of the excess of alcohol in vacuo, water
was added to the residue. The obtained mixture was basified
with NaHCO3 to pH 8 and extracted three times with meth-
ylene chloride. The organic layer was washed with water, dried
over Na2SO4, and evaporated in vacuo. The obtained oil was
purified by the means of column chromatography (silica gel;
mobile phase, hexane/ethyl acetate/methanol ) 10:4:1). The
residue was recrystallized from methylene chloride/hexane )
1:1) to give 45-60% pure 4b-d , mp (4b) 98 °C, mp (4d ) 63
°C.
Gen er a l P r oced u r e for th e Syn th esis of 3-Hyd r oxy-2-
{3-[1-(6-{1,1-d im eth yl-1-[3-(3-h yd r oxy-1-oxo-2,3-d ih yd r o-
1H -2-isoin d olyl)p r op yl]a m m on io}h exyl)-1,1-d im et h yl-
a m m on io]p r op yl}-1-isoin d olin on e
Dibr om id e
(5a ),
3-Methoxy-2-{3-[1-(6-{1,1-dimethyl-1-[3-(3-methoxy-1-oxo-2,3-
d ih yd r o-1H-2-isoin d olyl)p r op yl]a m m on io}h exyl)-1,1-d i-
m eth ylam m on io]pr opyl}-1-isoin dolin on e Dibr om ide (5b),
3-E t h oxy-2-{3-[1-(6-{1,1-d im et h yl-1-[3-(3-et h oxy-1-oxo-
2,3-d ih yd r o-1H-2-isoin d olyl)p r op yl]a m m on io}h exyl)-1,1-
d im eth yla m m on io]p r op yl}-1-isoin d olin on e Dibr om id e
(5c), 3-Isop r op oxy-2-{3-[1-(6-{1,1-d im eth yl-1-[3-(3-isop r o-
poxy-1-oxo-2,3-dih ydr o-1H-2-isoin dolyl)pr opyl]am m on io}-
h exyl)-1,1-dim eth ylam m on io]pr opyl}-1-isoin dolin on e Di-
br om id e (5d ) a n d 3-P h en yl-2-{3-[1-(6-{1,1-d im eth yl-1-[3-
(3-phenyl-1-oxo-2,3-dihydro-1H-2-isoindolyl)propyl]ammonio}-
h exyl)-1,1-dim eth ylam m on io]pr opyl}-1-isoin dolin on e Di-
br om id e (9). Bis(dimethylamino)hexane was refluxed with
phthalimidine derivatives 3, 4b-d , and 8 in a polar solvent
(see Table 4). The solvent was removed in vacuo. The obtained
oil was dissolved in 2 mL of ethanol. After addition of diethyl
ether again an oil was obtained and the solvent removed. This
procedure was repeated until the phthalimidine derivatives
(3, 4b-d , 8) were completely removed from the residue. The
obtained solid residue was recrystallized from a mixture of
ethanol/methylene chloride/diethyl ether. For analytical and
spectroscopic data, see Tables 4 and 5.
Syn th esis of 3-Isop r op ylth io-2-{3-[1-(6-{1,1-d im eth yl-
1-[3-(3-isop r op ylth io-1-oxo-2,3-d ih yd r o-1H-2-isoin d olyl)-
p r op yl]a m m on io}h exyl)-1,1-d im eth yla m m on io]p r op yl}-
1-isoin d olin on e Dibr om id e (5e). 5a (1.0 g, 1.4 mmol) was
dissolved in 2-propanethiol (3.0 g, 39.5 mmol). To the solution
were added 5 drops of concentrated hydrochloridic acid.
Subsequently, the reaction mixture was refluxed for 3 h. The
obtained oil was separated from the solution and washed with
ethyl acetate. The oil was dissolved in ethanol. After addition
of diethyl ether again an oil was obtained, which was washed
with ethyl acetate and crystallized from ethanol/ethyl acetate.
For analytical and spectroscopic data, see Tables 4 and 5.
4-P h en yl-1-p h th a la zon e (6). A solution of hydrazine
sulfate (4.0 g, 30.4 mmol) and sodium hydroxide (2.4 g, 60.8
mmol) in 20 mL of water was heated on a steam bath for 20
min. The hot solution was added to a hot aqueous solution of
benzoylbenzoic acid (6.9 g, 30.4 mmol). The reaction mixture
was heated under reflux for 1 h. After cooling and addition of
hydrochloride acid, the residue obtained was filtered and
washed with water. The crystalline solid was recrystallized
from ethanol to give 5.1 g (76%) of 6: Mw ) 222.3, mp 212 °C;
1H NMR (60 MHz, DMSO-d6) δ 7.23-8.53 (m, 9H, arom), 16.43
ppm (s, 1H, -NH).
3-P h en yl-1-isoin d olin on e (7). To a mixture of phen-
ylphthalazone (4.0 g, 18 mmol), Zn powder (17.6 g, 270 mmol),
and 10 mL of water was added 40 mL of concentrated
hydrochloric acid carefully. The reaction mixture was heated
under reflux for 1 h. To the cooled solution was added water.
The mixture was extracted with methylene chloride. The
organic layer was dried over Na2SO4 and evaporated in vacuo.
The residue was purified by the means of column chromatog-
raphy (silica gel; mobile phase, benzine/ethyl acetate/methanol
) 9:4:1; Rf ) 0.4). The crystalline solid was recrystallized from
1
ethanol to give 1.7 g (49%) of 7: Mw ) 209.2, mp 218 °C; H
NMR (60 MHz, DMSO-d6) δ 5.73 (s, 1H, -NCH-), 7.18-7.90
(m, 9H, arom), 9.10 ppm (s, 1H, -NH).
N-(3-Br om op r op yl)-3-p h en yl-1-isoin d olin on e (8). To a
solution of 3-phenyl-1-isoindolinone (2.0 g, 9.6 mmol) in 50 mL

SAR of Bisquaternary Bisphthalimidine Derivatives
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11 2163
of dried THF was added an equimolar amount of sodium
hydride. The mixture was heated under reflux for 2 h. 1,3-
Dibromopropane (5.8 g, 28.8 mmol) was added and the solution
was heated under reflux for additional 5 h. After THF and
the excess 1,3-dibrompropane were removed under reduced
pressure, the residue was purified by means of column
chromatography (silica gel; mobile phase, benzine/ethanol/
methanol ) 9:4:1; Rf ) 0.51). The crystalline residue was
recrystallized from ethanol in 26% yield; mp 106 °C.
ylidene-2-{3-[1-(6-{1,1-dimethyl-1-[3-(3-(E)-benzylidene-1-oxo-2,3-
d ih yd r o-1H-2-isoin d olyl)p r op yl]a m m on io}h exyl)-1,1-d i-
m eth ylam m on io]pr opyl}-1-isoin dolin on e Dibr om ide (12a),
3-(Z)-Be n zylid e n e -2-{3-[1-(6-{1,1-d im e t h yl-1-[3-(3-(Z)-
ben zylid en e-1-oxo-2,3-d ih yd r o-1H-2-isoin d olyl)p r op yl]-
a m m on io}h exyl)-1,1-d im et h yla m m on io]p r op yl}-1-iso-
in d olin on e Dibr om id e (12b), a n d 3-Ben zyl-2-{3-[1-(6-{1,1-
dim eth yl-1-[3-(3-ben zyl-1-oxo-2,3-dih ydr o-1H-2-isoin dolyl)-
p r op yl]a m m on io}h exyl)-1,1-d im eth yla m m on io]p r op yl}-
1-isoin d olin on e Dibr om id e (15). According to the general
procedure for the syntheses of 5a -d , the reaction of 1,6-
dibromohexane with the phthalimide derivatives 14 and 11-
(E/Z), respectively, for 5-24 h in a polar solvent, gave 15 and
12a ,b. For analytical and spectroscopic data, see Tables 4 and
5.
Syn th esis of 3-[Z]-Ben zyliden e-2-{3-[1-(6-{1,1-dim eth yl-
1-[3-(3-[E]-ben zyliden e-1-oxo-2,3-dih ydr o-1H-2-isoin dolyl)-
p r op yl]a m m on io}h exyl)-1,1-d im eth yla m m on io]p r op yl}-
1-isoin d olin on e Dibr om id e (12c). Equimolar amounts (1.8
mmol) of 13 and 11b were refluxed in acetonitrile for 4-5 h.
The solvent was removed in vacuo. The obtained solid residue
was dissolved in water and extracted five times with diethyl
ether. The water was removed in vacuo and the remaining oil
dissolved in methylene chloride. 12c precipitates upon addition
of diethyl ether. The precipitate was recrystallized from a
mixture of methylene chloride/ethyl acetate. For analytical and
spectroscopic data, see Tables 4 and 5.
P h a r m a cology.A homogenate of porcine heart ventricular
myocardium was prepared as decribed previously in more
detail.33 The membranes were stored in 50 mM Tris-HCl pH
7.4 or in distilled water at -80 °C until use. [3H]NMS (0.15
nM) equilibrium binding was measured in 3 mM MgHPO4, 50
mM Tris, pH 7.3, 37 °C (“Mg,Tris,Cl,Pi” buffer) after 2 h.
Nonspecific binding was determined in the presence of 1 µM
atropine and did not exceed 4% of total binding. Membrane-
bound radioactivity was separated by rapid filtration and
washed with 2 × 5 mL of ice-cold incubation buffer. Radioac-
tivity on the filters (Schleicher & Schu¨ll, No. 6; Dassel, F.R.G)
was determined by liquid scintillation counting. [3H]NMS
3-Ben za lp h th a lid e (10). A mixture of phthalic anhydride
(10.0 g, 67.5 mmol), phenylacetic acid (11.0 g, 81 mmol), and
sodium acetate (2.76 g, 33.7 mmol, dry) was heated for 3-4 h
up to 250-300 °C. After cooling to room temperature, the solid
reaction mixture was crystallized from ethanol to give 10 g
(70%) of the Z isomer of 3-benzalphthalide 10: Mw ) 222.2,
1
mp 98-100 °C; H NMR (60 MHz, DMSO-d6) δ 6.40 (s, 1H),
7.16-8.06 ppm (m, 9H, arom).
3-(E/Z)-Ben zylid en -N-[3-(N,N-d im eth yla m in o)p r op yl]-
1-isoin d olin on e [11 (E) a n d 11 (Z)]. A solution of 3-benzal-
phthalide (6.0 g, 27.0 mmol) and N,N′-dimethylaminopropyl-
amine (2.8 g, 27.0 mmol) in 10 mL ethanol was refluxed for 4
h. After evaporation of ethanol in vacuo, 50 mL of water was
added to the residue. The obtained mixture was extracted with
methylene chloride. The organic layer was dried over Na2SO4
and evaporated in vacuo. The residue was crystallized from
diethyl ether to give N-[3-(N,N-dimethylamine)propyl]-3-hy-
droxy-3-benzyl-1-isoindolinone. The benzylidenephthalimidine
derivative 11 was obtained after dehydration with formic acid
(3 g). The mixture was heated at 130-140 °C for 30 min. After
cooling, water was added. The resulting mixture was basified
with NaHCO3 and extracted with methylene chloride. The
organic layer was washed with water, dried over Na2SO4, and
evaporated. The obtained product 11 is a mixture of about 90%
E and 10% Z isomers. The diastereomers were separated by
means of column chromatography on silica gel (mobile phase,
methylene chloride/methanol ) 5:1; Rf(Z) ) 0.57, Rf(E) ) 0.51).
3-(E)-Ben zylid en e-N-{3-[1-(6-{1-b r om oh exyl})-1,1-d i-
m eth yla m m on io]p r op yl}-1-isoin d olin on e Br om id e (13).
A mixture of N-[3-(N,N-dimethylamino)propyl]-3-(E)-benzylide-
nephthalimidine (0.54 g, 1.765 mmol) and 1,6-dibromohexane
(6.7 g, 27.6 mmol) in 15 mL of acetonitrile was stirred at room
temperature for 72 h. Acetonitrile was removed in vacuo. The
obtained oil was dissolved in 4 mL of methylene chloride. After
addition of diethyl ether again an oil was obtained and the
solvent removed. This procedure was repeated until the
bisammonium compound (12a ) was completely removed from
the residue (TLC control; eluent, methanol/concentrated NH3
) 4.5:1.5; Rf(13) ) 0.133, Rf(12a ) ) 0.00). The obtained solid
residue was recrystallized from methylene chloride/ethyl
binding under control conditions was characterized by KD
0.5 nM and Bmax ≈ 100 fmol/mg of protein, respectively.
≈
All test compounds were studied in Mg,Tris,Cl,Pi buffer for
their ability to retard the dissociation of [3H]NMS from cardiac
muscarinic M2-receptors. After incubating radioligand with the
membranes for 30 min, [3H]NMS dissociation was visualized
by addition of 1 µM atropine, alone or in the presence of the
test compounds. Dissociation proceeded monophasically under
control conditions (t1/2 ≈ 2 min) and in the presence of the test
compounds. The dissociation data were analyzed by nonlinear
regression analysis applying a monoexponential decay function
to determine the apparent dissociation rate constant k-1 of the
radioligand. All test compounds retarded [3H]NMS dissociation
concentration-dependently. A plot of the apparent rate con-
stant of dissociation k-1 versus the concentrations of the test
compounds provided concentration-effect curves for the retard-
ing action on the dissociation. Curves were fitted to a four-
parameter logistic function. To study the sensitivity of the
allosteric action of 12a for the antagonistic action of obidoxime,
[3H]NMS dissociation experiments were carried out in a buffer
consisting of 4 mM Na2HPO4, 1 mM KH2PO4, pH 7.4, 23 °C
(“Na,K,Pi” buffer; [3H]NMS binding characteristics, KD ≈ 0.1
nM, Bmax ≈ 100 fmol/mg protein, dissociation t1/2 control ∼ 2.5
min). The data were analyzed as described previously in
detail.34
1
acetate to give 0.5 g (52%) of 13: Mw ) 470, mp 150 °C; H
NMR (300 MHz, CDCl3, δ (ppm)) 7.74 (td, 7.5, 0.9, 1H-arom),
7.25-7.50 (m, 8H-Arom), 6.89 (s, -CdCH), 4.03 (t, 6.5, Od
CNCH2-), 3.69 (m, OdCNCH2CH2CH2N+), 3.54 (m, +NCH2-
), 3.34 (s, +N(CH3)2), 3.30 m (t, 6.7, -CH2Br), 2.29 (m,
OdCNCH2CH2-), 1.75 (m, -CH2CH2Br), 1.66 (m, +NCH2CH2-
), 1.36 (m, +NCH2CH2CH2CH2-); 13C NMR (CDCl3, δ (ppm))
166.67 (-CdO), 135.02 (C-Arom), 134.81 (C-Arom), 134.52
(C-Arom), 131.72 (CH-Arom), 129.54 (CH-Arom), 129.39
(C-Arom), 129.39 (CH-Arom), 129.25 (CH-Arom), 128.48
(CH-Arom), 127.84 (CH-Arom), 123.18 (CH-Arom), 122.88
(CH-Arom), 111.82 (-CdCH-), 64.19 (+NCH2-), 61.37
(OdCNCH2CH2CH2N+), 51.39 (+N(CH3)2), 36.46 (OdCNCH2-
), 33.59 (-CH2Br), 32.15 (-CH2CH2Br), 27.54 (+NCH2CH2CH2-
), 25.34 (+NCH2CH2CH2CH2-), 22.61 (OdCNCH2CH2-), 22.48
(+NCH2CH2-).
Molecu la r Mod ellin g. QSAR An a lysis. The lateral varied
N-methylimides were built up using the model builder in
HyperChem 5.1 (Hypercube Inc. Waterloo, ON, Canada, 1997).
Geometry optimization was achieved using the MM+ progam
implemented in HyperChem. By means of ChemPlus (Hyper-
cube Inc.) the following physicochemical properties were
calculated: the octanol/water partition coefficient (logP), the
steric parameters volume and surface area, as well as polar-
izability and refractivity, a chameleon parameter comprising
both steric bulk and polarizability. The MOPAC polarizabilities
3-Ben zyl-2-[3-(N,N-d im eth yla m in o)p r op yl]-1-isoin d oli-
n on e (14). A mixture of diastereomers of 11 (1.3 g, 4.2 mmol),
10% Pd/C (300 mg), 300 mL of ethanol, and hydrogen was
shaken under 4.4 bar pressure at room temperature for 16 h.
After removal of the catalyst by filtration, solvent was removed
in vacuo. The product was purified by means of column
chromatography (silica gel; mobile phase, methylene chloride/
methanol ) 5:1) to give 82% of 14, mp 147-150 °C.
Gen er a l P r oced u r e for t h e Syn t h esis of 3-(E)-Ben z-

2164 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 11
Botero Cid et al.
(9) Holzgrabe, U.; Wagener, M.; Gasteiger, J . Comparison of struc-
turally different allosteric modulators of muscarinic receptors
by self-organizing neural networks. J . Mol. Graph. 1996, 14,
185-193.
(10) Holzgrabe, U.; Hopfinger, A. J . Conformational analysis, mo-
lecular shape comparison, and pharmacophore identification of
different allosteric modulators of muscarinic receptors. J . Chem.
Inf. Comput. Sci. 1996, 36, 1018-1024.
(11) Bejeuhr, G.; Blaschke, G.; Holzgrabe, U.; Mohr, K.; Su¨rig, U.;
Terfloth, G. A stable and highly potent hexamethonium deriva-
tive which modulates muscarinic receptors allosterically in
guinea-pig hearts. J . Pharm. Pharmacol. 1994, 46, 108-112.
(12) Staudt, M.; Tra¨nkle, C.; Mohr, K.; Holzgrabe, U.; Contributions
of lateral substituents of heptanebisammonium derivatives to
the allosteric modulation of antagonist binding to muscarinic
M2-receptors. Life Sci. 1998, 62, 423-429.
(13) Wassermann, O. Habilitationsschrift, University of Kiel, 1970.
(14) Dunet, A.; Willemart, A. Produits de reduction du phthalimide;
Hydroxyphthalimidine et son. Bull. Soc. Chem. 1948, 189, 887-
889.
were calculated from the molecular mechanics optimized
geometries using the following command line: AM1 EF
GNORM)0.1 POLAR. The mean R-polarizabilities of MOPAC
version 6.12 were used. The allosteric potencies (EC50) were
transformed in the QSAR form log(1/EC50). The QSARs (vari-
able selection, regression, and permutation test) were calcu-
lated with in-house written Matlab scripts (Matlab 5.3, The
MathWorks Inc. 24 Prime Park Way, Natick, MA) which are
available on request.
P h a r m a cop h or e Sea r ch . The isomers of compound 12
were built up using model builder in HyperChem 5.1; the so
obtained elongated geometries were optimized by means of the
force field calculation MM+ program.
Th e F ittin g P r oced u r e. Using this starting geometry the
atoms to be overlay have been tethered to the cartesian
coordinates of the corresponding alcuronium atoms. After
geometry optimization, all applied restraints have been re-
moved and den MM+ calculations were repeated. The so-
obtained structures were fitted onto alcuronium using the RMS
Fit option of HyperChem 5.1. In alcuronium both positively
charged nitrogens and aromatic atoms in indole skeletons
adjacent to the pyrrolidine ring were used as fitting atoms. In
12, both positively charged nitrogens, and either in the
phthalimidine skeleton an ipso-atom and the corresponding
atom in para position or in the benzylidene moiety the ipso-
carbon atom and the atom in para position were chosen for
fitting and calculation of the rms value.
(15) Gabriel, S.; Naumann, A. U¨ ber Derivate des Phthalizins und
Isoindols Ber. Dt. Chem. Ges. 1893, 26, 705-713.
(16) Gabriel, S. Zur Kenntnis des Benzylidenphthalids. Ber. Dt.
Chem. Ges. 1885, 18, 1251-2433.
(17) Marsili, A.; Scartoni, V. The structures of 3-benzylidene-
phthalidesPrimary amine adducts and of o-phenylacetylbenzoic
acid. Gazz. Chim. Ital. 1972, 102, 507-516.
(18) Seydel, J . K.; Schaper, K.-J . Chemische Struktur und biologische
Aktivita¨t; Verlag Chemie: Weinheim, 1979.
(19) Kubinyi, H. QSAR: Hansch Analysis and Related Approaches;
Mannhold, R., Krogsgaard-Larsen, P., Timmerman, H., Eds.;
VCH: Weinheim, 1993.
(20) Hansch, C.; Leo, A. Exploring QSAR-Fundamentals and Ap-
plications in Chemistry and Biology; ACS Professional Reference
Book; ACS: Washington, DC 1995.
(21) Stewart, J . J . P. MOPAC: A semiempirical molecular orbital
program. J . Comput.-Aided Mol. Design, 1990, 4, 1-105.
(22) Karelson, M.; Lobanov, V. S.; Katritzky, A. R. Quantum-chemical
descriptors in QSAR/QSPR studies, Chem. Rev. 1996, 96, 1027-
1043.
(23) Clerc, J . T.; Terkovics, A. L. Versatile Topological Structure
Descriptor for Quantitative Structure/property Studies. Anal.
Chim. Acta 1990, 235, 93-102.
(24) Baumann, K. Uniform-length molecular descriptors for quantita-
tive structure-property (QSPR), quantitative structure-activity
(QSAR), classification studies and similarity searching. TrAC
1999, 18, 36-46.
(25) Carhart, R. E.; Smith, D. H.; Venkataraghavan, R. Atom pairs
as molecular features in structure-activity studies: Definitions
and applications. J . Chem. Inf. Comput. Sci. 1985, 25, 64-73.
(26) Kubinyi, H. The Quantitative Analysis of Structure-Activity
Relationships. In Burger’s Medicinal Chemistry and Drug
Discovery; Wolff, M. E., Ed.; Wiley: New York, 1995; Vol. 1, part
IV, 14, pp 497-572.
(27) Topliss, J . G.; Edwards, R. P. Chance Factors in Studies of
Quantitative Structure-Activity Relationships. J . Med. Chem.
1979, 22, 1238-1244.
(28) Wold, S.; Eriksson, L. Statistical Validation of QSAR Results,
In: Chemometric Methods in Molecular Design (van de Water-
beemd, H. Ed.), VCH Verlagsgesellschaft, Weinheim, Germany,
1995, Chapter 5, p 307-318.
(29) Andrews, P. R. Drug-Receptor Interactions In 3D QSAR in Drug
DesignsTheory Methods and Applications; Kubinyi, H., Ed.;
ESCOM: Leiden, 1993; pp 13-40.
(30) Testa, B.; Carrupt, P.-A.; Gaillard, P.; Tsai, R.-S. Intramolecular
Interactions Encoded in Lipophilicity: Their Nature and Sig-
nificance In Lipophilicity in Drug Action and Toxicology; Pliska,
V., Testa, B., van de Waterbeemd Eds.; VCH Verlagsgesell-
schaft: Weinheim, Germany, 1996; Chapter 4, pp 48-71.
(31) Ellis J .; Seidenberg, M. Two allosteric modulators interact at a
common site on cardiac muscarinic receptors. Mol. Pharmacol.
1992, 42, 638-641.
(32) Kubinyi, H. Lock and Key in the Real World: Concluding
Remarks. Pharm. Acta Helv. 1994, 69, 243-258.
(33) Tra¨nkle, C.; Kostenis, E.; Burgmer, U.; Mohr, K. Search for lead
structures to develop new allosteric modulators of muscarinic
receptors. J . Pharmacol. Exp. Ther. 1996, 279, 926-933.
Ack n ow led gm en t. Thanks are due to the Deutsche
Forschungsgemeinschaft, DFG, and to the Fonds der
Chemischen Industrie, FRG, for financial support, to the
KAAD for the grant given to M.H.B.C., to an anonymous
referee, who greatly improved the presentation of the
QSAR results, and to Modest von Korff for valuable
discussions, as well as to Ilona Knoblauch, Iris Witten,
and Frauke Mo¨rschel for skillful technical assistance.
1
Su p p or tin g In for m a tion Ava ila ble: Spectra (IR and H
and 13C NMR) of the intermediates and final products as well
as the conformational energy and rms values of the different
fit conformations of the isomers of 12 are available free of
charge via the Internet at http://pubs.acs.org.
Refer en ces
(1) Holzgrabe, U.; Mohr, K. Allosteric modulators of ligand binding
to muscarinic acetylcholine receptors. Drug Discovery Today
1998, 3, 214-222.
(2) Christopoulos, A.; Lanzafame, A.; Mitchelson, F. Allosteric
interactions at muscarinic cholinoceptors. Clin. Exp. Pharmacol.
Physiol. 1998, 25, 185-19.
(3) Tra¨nkle, C.; Mohr, K. Divergent modes of action among allosteric
modulators of muscarinic M2 receptors. Mol. Pharmacol. 1997,
51, 674-682.
(4) Tra¨nkle, C.; Mies-Klomfass, E.; Botero Cid, M. H.; Holzgrabe,
U.; Mohr, K. Identification of a [3H]ligand for the common
allosteric site of muscarinic acetylcholine M2 receptors. Mol.
Pharmacol. 1998, 54, 139-145.
(5) Proska, J .; Tucek, S. Mechanism of steric and cooperative actions
of alcuronium on cardiac muscarinic acetylcholine receptors. Mol.
Pharmacol. 1994, 45, 709-717.
(6) Waelbroeck, M.; Robberecht, P.; De-Neef, P.; Christophe, J .
Effects of d-tubocurarine on rat muscarinic receptors. J . Recept.
Res. 1988, 8, 787-808.
(7) Ellis, J .; Seidenberg, M.; Brann, M. R. Use of chimeric muscarinic
receptors to investigate epitopes involved in allosteric inter-
actions. Mol. Pharmacol. 1993, 44, 583-588.
(8) J epsen, K.; Lu¨llmann, H.; Mohr, K.; Pfeffer, J . Allosteric
stabilization of 3H-N-methylscopolamine binding in guinea-pig
myocardium by an antidote against organophosphorus intoxica-
tion. Pharmacol. Toxicol. 1988, 63, 163-168.
J M991136E