Comparative molecular field analysis of hydantoin binding to the neuronal voltage-dependent sodium channel

Article abstract of DOI:10.1021/jm980556l

Comparative molecular field analysis (CoMFA), a 3-D QSAR technique, is widely used to correlate biological activity with observed differences in steric and electrostatic fields. In this study, CoMFA was employed to generate a model, based upon 14 structur

Full text of DOI:10.1021/jm980556l

J . Med. Chem. 1999, 42, 1537-1545
1537
Com p a r a tive Molecu la r F ield An a lysis of Hyd a n toin Bin d in g to th e Neu r on a l
Volta ge-Dep en d en t Sod iu m Ch a n n el
Milton L. Brown,† Congxiang C. Zha,† Christopher C. Van Dyke,§ George B. Brown,‡ and Wayne J . Brouillette*,†
Departments of Chemistry and of Psychiatry and Behavioral Neurobiology, University of Alabama at Birmingham,
Birmingham, Alabama 35294, and Tripos, Inc., 1699 South Hanley Road, Suite 303, St. Louis, Missouri 63144
Received September 30, 1998
Comparative molecular field analysis (CoMFA), a 3-D QSAR technique, is widely used to
correlate biological activity with observed differences in steric and electrostatic fields. In this
study, CoMFA was employed to generate a model, based upon 14 structurally diverse
5-phenylhydantoin analogues, to delineate structural and electrostatic features important for
enhanced sodium channel binding. Correlation by partial least squares (PLS) analysis of in
vitro sodium channel binding activity (expressed as log IC50) and the CoMFA descriptor column
generated a final non-cross-validated model with R2 ) 0.988 for the training set. The final
CoMFA model explained the data better than a simpler correlation with log P (R2 ) 0.801) for
the same training set. The CoMFA steric and electrostatic maps described two general features
that result in enhanced binding to the sodium channel. These include a preferred 5-phenyl
ring orientation and a favorable steric effect resulting from the C5-alkyl chain. This model
was then utilized to accurately predict literature sodium channel activities for hydantoins 14-
20, which were not included in the training set. Finally the hydantoin CoMFA model was used
to design the structurally novel R-hydroxy-R-phenylamide 21. Synthesis and subsequent sodium
channel evaluation of compound 21 (predicted IC50 ) 9 µM, actual IC50 ) 9 µM), a good binder
to the sodium channel, established that the intact hydantion ring is not necessary for efficient
binding to this site. Thus R-hydroxy-R-phenylamides may represent a new class of ligands
that bind with increased potency to the sodium channel.
The anticonvulsant diphenylhydantoin (DPH), or
phenytoin, binds to the neuronal voltage-sensitive so-
dium channel (NVSC) at therapeutically relevant con-
centrations in a frequency- and voltage-dependent
manner.1 These observations are consistent with the
selective actions of DPH on hyperactive versus normal
neurons.
site. CoMFA samples the differences in steric and
electrostatic fields surrounding a set of ligands. Numer-
ous successes have been reported using this approach
for mapping receptor properties, including studies on
central nervous system (CNS) active compounds6-21 as
well as a NVSC study22 which evaluated the binding of
brevetoxins to neurotoxin site 5.
Our earlier studies2,3 on hydantoin analogues re-
vealed that, when log P is held constant, structural
variations lead to dramatic differences in sodium chan-
nel binding potency. This observation, when coupled
with the reported4 saturable effects of DPH on sodium
channels, suggests that DPH undergoes interaction with
a specific site on the NVSC.
Unfortunately, the exact location of this site and
structural requirements for optimum binding have not
yet been determined. Our previous observation of the
strong influence of log P, given an appropriate structural
type, on the in vitro binding of hydantoins to the NVSC
suggests that membrane entry may be required for
activity. This observation and the report that changes
in intracellular pH affect the activity of DPH on sodium
channels are consistent with an intracellular binding
site.5 Given the absence of high-resolution structural
data for the NVSC, in the present study we employed
comparative molecular field analysis (CoMFA) to help
define important 3-dimensional (3-D) properties associ-
ated with the optimum binding of hydantoins to this
In this study we utilized sodium channel binding data
for hydantoin analogues that we previously published,2
and we additionally synthesized and evaluated the
sodium channel binding activities for 5-alkyl-5-phenyl-
hydantoins 5-7 to extend the series to include larger
log P values. Sodium channel binding data for hydan-
toins 1-13 and DPH (Table 1) were used to evaluate
correlations with log P and derive a CoMFA model. The
CoMFA model derived from the training set in Table 3
was then used to predict the sodium channel binding
activities of hydantoin analogues 14-19 and carbam-
azepine (20) (Table 4) and to design a novel ligand,
hydroxyamide 21.
Meth od s
Biologica l Da ta . The structures and relative sodium
channel binding activities for compounds 1-13 and
DPH, which form the training set, are listed in Table
1. Table 4 describes the structures and sodium channel
binding activities of compounds 14-21, which form the
test set. In the sodium channel evaluation, the IC50,
which represents the micromolar concentration of com-
pound required to displace 50% of specifically bound
[3H]batrachotoxinin A 20-R-benzoate ([3H]BTX-B), was
determined in an in vitro assay using rat brain cerebral
* Address correspondence to this author.
† Department of Chemistry, University of Alabama at Birmingham.
‡ Department of Psychiatry and Behavioral Neurobiology, University
of Alabama at Birmingham.
§ Tripos, Inc.
10.1021/jm980556l CCC: $18.00 © 1999 American Chemical Society
Published on Web 04/08/1999

1538 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9
Ta ble 1. Data from the PLS Cross-Validated Analysis
Brown et al.
preliminary model, -log IC50
final model, -log IC50
torsion angle (deg)
compd
R1
R2
R3
n
N1,C5,C6,C7
obsd
pred
res
obsd
pred
res
1
2
3
4
5
6
7
8a
n-propyl
n-butyl
n-butyl
n-pentyl
n-hexyl
n-heptyl
n-nonyl
methyl
methyl
H
H
H
H
H
H
H
ethyl
ethyl
H
H
5
5
5
5
5
-2.21
-2.01
-2.45
-1.59
-1.11
-0.70
-0.70
-2.86
-2.86
-3.32
-3.32
-2.93
-2.93
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-2.40
-1.60
-2.00
-1.91
-1.90
-1.72
-1.29
-1.00
-0.63
-2.91
-3.80
-2.88
-2.71
-2.93
-2.75
-2.47
-2.57
-2.70
-2.57
-2.69
-2.81
-2.24
-2.37
-2.25
-2.38
-2.35
-2.43
-2.45
-2.10
-0.21
-0.10
-0.55
0.13
0.18
0.30
-0.07
0.05
0.20
-0.44
-0.61
0.00
-0.18
0.07
0.17
0.30
0.17
0.29
-2.21
-2.01
-2.45
-1.59
-1.11
-0.70
-0.70
-2.86
-2.03
-2.03
-2.01
-1.67
-1.12
-0.91
-0.72
-2.74
-0.18
0.02
-0.44
-0.12
0.01
0.21
0.02
-0.12
methyl
H
H
H
H
H
H
5
5
-45
-57
-36
-76
-43
-65
-17
-27
-44
-75
-93
-63
-10
-18
-25
-63
-73
-142
-179
8b
9a
9b
1
1
2
2
3
3
3
3
3
3
4
4
4
4
4
-3.32
-2.93
-2.40
-2.54
-2.58
-2.58
-0.68
0.19
10a
10b
11a
11b
11c
11d
11e
11f
12a
12b
12c
12d
12e
13a
13b
DPH
0.18
0.41
-0.16
-0.03
-0.15
0.02
-0.05
0.03
-2.40
-2.28
-0.12
-2.40
-1.60
-2.53
-2.20
0.13
0.06
0.05
0.50
phenyl
H
H
Ta ble 2. Cross-Validated PLS Analysisa
ring, for conformational searching. Previously reported2
low-energy conformers of 1, 2, 4, and 8-16 were used.
To determine the low-energy conformations for 3 and
5-7 about the C5-C6 bond, we utilized GRIDSEARCH
to rotate N1, C5, C6, and C7 over 360° in 1° increments
as previously reported.2 A similar approach was used
for 21. For carbamezapine (20)33 the conformation
reported for the X-ray crystal structure was utilized.
The results for conformational searches on the rota-
tionally restricted 5-phenyl analogues 9-13 using the
GRIDSEARCH routine were previously described.2 In
a similar fashion, the low-energy conformations of 17
were obtained by incorporating a cyclohexyl side chain
in a chair conformation (obtained from the SYBYL
fragment library). The atomic charges for all analogues
were calculated using AM1 (MOPAC).
Molecu la r Align m en t. All of the compounds in the
training set (Table 3) have identical hydantoin ring
atoms, and these were used as the basis for an align-
ment rule (atoms N1, C2, N3, C4, and C5 were fit onto
each other). Similarly, for 14-19 in the test set, we
aligned the hydantoin ring atoms N1, C2, N3, C4, and
C5 and the phenyl ring atom C6. The n-alkyl groups at
C5 for analogues 1-7 in the training set and for 14-
16, 20, and 21 in the test set approximated a fully
extended conformation following energy minimization,
which was arbitrarily selected for this study. For
carbamezapine (20)33 the imide carbonyl oxygen and,
for one aromatic ring, the two aromatic carbons that
are â to the ring N were fit onto the amide carbonyl
oxygen and the two aromatic carbons attached to C6 of
preliminary modelb
final modelc
components
s
R2
s
R2
1
2
3
4
5
0.561
0.406
0.342
0.314
0.293
0.301
0.648
0.762
0.808
0.840
0.531
0.457
0.404
0.404
0.430
0.573
0.736
0.821
0.844
0.847
a
s, standard error for the estimate of log IC50; R2, correlation
coefficient or press value ) 1.0 - {∑(Y predicted - Y actual)2
÷
(Y actual - Y mean)2 }. The preliminary model included single
conformations of 1-7 and multiple conformations of 8-13. c The
final model included single conformations of 1-7 and a single best
fit (smallest residual as determined from the preliminary model)
conformer of 8-13. Optimum number of components is 3 with R2
) 0.821.
b
cortex synaptoneurosomes. To calculate the IC50 value,
the percent inhibition at 5-7 concentrations which
spanned the IC50 value was determined in triplicate.
The IC50 value was obtained by a Probit analysis of the
concentration versus percent binding curve.
Con for m a tion a l An a lysis. All compounds were
energy-minimized with the Tripos force field,23 without
solvent, using default bond distances and angles and
neglecting electrostatics. The details were previously
described.2 While compounds were synthesized and
evaluated as racemic mixtures, for internal consistency
the geometries were modeled using only the R-config-
uration at C5. Since 5-phenyl ring orientation for
hydantoin analogues has previously been suggested to
be important for hydantoin binding to the NVSC,24 we
designated the C5-C6 bond, which rotates the phenyl

CoMFA of Hydantoin Binding
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9 1539
Ta ble 3. Sodium Channel Binding Activities for the Training Set Used in the Final Non-Cross-Validated CoMFA Modela
Na+ channel
inhibition
IC50 (µM)
-log IC50
predi
compd
R1
R2
R3
n
log P
obsd
res
1
n-propyl
H
H
H
1.96h
162
[136-193]e
103
[85-124]
285
[232-350]
39
[32-47]
13
[9-17]
5
[4-6]
5
[4-6]
720
2112
[2021-2207]
851
[761-953]
251
[225-280]
251
-2.21b
-2.16
-0.05
2
3
4
5
6
7
n-butyl
n-butyl
n-pentyl
n-hexyl
n-heptyl
n-nonyl
methyl
H
2.46h
3.01h
2.96h
3.40
-2.01b
-2.45b
-1.59b
-1.11d
-0.70d
-0.70c
-2.09
-2.39
-1.61
-1.19
-0.72
-0.52
0.08
0.06
H
methyl
H
H
H
H
H
H
0.02
H
0.08
H
3.96
0.02
H
4.96
-0.18
8
9
ethyl
2.02j
0.96j
-2.86c
-3.32k
-2.93
-3.32
0.07
0.00
1
2
3
4
10
11
12
1.46j
1.96j
2.46j
-2.93k
-2.4k
-2.4k
-2.82
-2.36
-2.36
-0.11
-0.04
-0.04
[228-277]
250
13
DPH
1.52j
2.46g
-2.4c
-1.6
-2.42
-1.61
0.02
0.01
phenyl
H
H
40f
a
b
Number of components ) 3, s ) 0.099, R2 ) 0.988. Reported IC50 value ( 1 standard deviation in ref 3. c Reported IC50 value in ref
24. The sodium channel value was determined in this study. e Numbers in brackets represent (1 standard deviation. f Data taken from
ref 1. Log P data taken from ref 30. Calculated using parent structure log P value in ref 3. Generated from the CoMFA non-cross-
validated run (see Methods). Calculated using parent structure log P value in ref 2. Reported IC50 value in ref 2.
Ta ble 4. Predicted and Actual Sodium Channel Binding Activities for Diverse Analogues Forming the Test Set
d
g
h
i
j
k
Na+ channel
-log IC50 (µM)
obsd
predd
-2.35 -2.27 -0.08
energy
(kcal/mol)
torsion angle (deg)
N1,C5,C6,C7
inhibitionc
compd
R1
R2
R3
R4
log P
IC50 (µM)
res
14(a )
n-butyl
methyl
H
H
H
H
H
10.8
12.4
8.2
-37
159
4
2.96e
225
[218-233]
225
[218-233]
58c
[57-60]
95
14(b)
15
n-butyl
methyl
2.96e
2.96e
2.96e
2.96
-2.35 -2.37
-1.76 -2.03
-1.98 -2.03
-1.76 -1.89
0.02
0.26
0.05
0.13
n-butyl
H
H
H
methyl
16
n-butyl
H
H
methyl
H
[86-106]
58
17
cyclohexyl
[41-82]
18(a )
18(b)
19(a )
40.4
39.1
27.0
86
-1
-1
2.01b
2.01b
2.51b
700a
-2.85 -3.11
-2.85 -2.92
-2.63 -2.72
0.26
0.07
0.09
700a
427a
[376-383]
19(b)
26.3
2.51b
427a
[376-383]
131f
20
21
2.13g
3.70
-2.12 -2.10 -0.02
-0.95 -0.95 0.00
9
[7-11]
a
b
Reported in ref 3. Calculated using parent structure log P value in ref 24 and π values in ref 32. c Numbers in brackets represent
(1 standard deviation. Data generated from the CoMFA non-cross-validated run (see QSAR Methods). e Calculated using parent structure
d
log P value in ref 2. f Reported in ref 1. Calculated using Clog P.
g
DPH (Figure 4C). The hydroxyamide 21 was aligned
such that the OH group was superimposed with N1 and
the carboxamide was aligned with C4 and N3.
CoMF A Ca lcu la tion s. CoMFA, using default pa-
rameters except where noted, was calculated in the
QSAR option of SYBYL 6.4 on a Silicon Graphics Octane
computer. The CoMFA grid spacing was 2.0 Å in the x,
y, and z directions, and the grid region was automati-
cally generated by the CoMFA routine to encompass all
molecules with an extension of 4.0 Å in each direction.

1540 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9
Ta ble 5. Selected Data for New Compoundsa
Brown et al.
isolated
mp (°C)
1H NMR
13C NMR
IR (cm-1
)
(CdO), KBr
compd yield (%) (recryst solvent)
(DMSO-d6)b
(DMSO-d6)c
5
6
7
59
21
95
138-140 [lit.
140-142]d
(ethanol)
124-126
(ethanol)
10.91-10.65 (s, 1H, NH), 8.79-8.56 (s, 1H, NH),
7.65-7.15 (m, 5H, Ph), 2.15-1.68 (s, 2H, CH2),
1.45-0.98 (m, 10H, CH2), 0.94-0.63 (m, 3H, CH3)
10.98-10.53 (s, 1H, NH), 8.70-8.44 (s, 1H, NH),
8.08-7.87 (m, 1H, Ph), 7.73-7.16 (m, 4H, Ph),
3.09-2.85 (s, 1H, CH), 2.11-1.68 (s, 2H, CH2),
1.68-1.43 (m, 12H, CH2), 1.40-1.04 (m, 12H, CH2),
0.99-0.61 (m, 3H, CH3)
7.74-7.43 (m, 2H, Ph), 7.43-7.10 (m, 3H, Ph),
6.62-6.27 (s, 1H, NH), 5.90-5.60 (s, 1H, NH),
3.54-3.17 (s, 1H, OH), 2.30-2.12 (m, 1H, CH),
2.08-1.89 (m, 1H, CH), 1.56-1.08 (m, 10H, CH2),
0.94-0.59 (m, 3H, CH3)
176.3, 156.5, 139.3, 128.4,
127.7, 125.3, 67.5, 31.1,
28.7, 28.5, 23.3, 22.2, 13.9
176.9, 156.1, 139.8, 129.0,
128.2, 125.9, 67.1, 36.3,
28.8, 28.4, 28.4, 28.3,
1755, 1710
1750, 1700
115-116
(ethanol)
28.2, 22.8, 21.7, 13.5
21
100
91-92
(toluene)
177.0, 142.4, 128.5, 127.8,
125.4, 78.8, 39.2, 31.8,
29.7, 29.2, 23.4, 22.6, 14.1
1660
a
All compounds were also analyzed by (1) GC/MS (EI, 70 eV), giving one peak in the GC and the expected molecular ion peak, and (2)
elemental analysis for C, H, N (within (0.4% of the calculated value). Compound 20 1H NMR taken in CDCl3. c Compound 20 13C NMR
taken in CDCl3. Compound reported in ref 28.
b
d
An sp3 carbon and a charge of +1.0 were used as probes
to generate the interaction energies at each lattice point.
The default value of 30 kcal/mol was used as the
maximum electrostatic and steric energy cutoff.
For the preliminary and final cross-validated CoMFA
models (Table 2), the lowest energy conformer for the
analogues 1-7 was used, which in all cases contained
a phenyl ring orientation similar to that found in the
X-ray crystal structure25 of DPH (see Table 1). For
compounds 8-13, as a first approach in the preliminary
CoMFA model, all low-energy conformations within 5.0
kcal/mol of the global minimum were included, and the
single conformer with the smallest cross-validated
residual value was selected (Table 1). These single
conformers of 8-13 were used in both the final cross-
validated CoMFA model (Table 2) and the non-cross-
validated model (Table 3).
logues 14-20, from the test set (Table 4), were pre-
dicted. The quality of these predictions prompted us to
design a structurally unique compound, R-hydroxy-R-
phenylamide 21, which met requirements of the model
but did not contain a hydantoin ring. As discussed
below, predictions regarding the relatively potent activ-
ity for this compound were correct.
Ch em istr y
The syntheses of analogues 1-4,3 8,2 11,29 12,29 13,2
14-16,2 17,15 and 18-193 were previously described.
Compounds 528 (59% yield), 927 (31% yield), and 1728
(15% yield) were prepared from commercially available
hexanophenone, 1-indanone, and cyclohexyl phenyl
ketone, respectively, using the Bucherer-Berg ap-
proach.28 Hydantoins 6 and 7 each were prepared in an
isolated yield of 95% from commercially available oc-
tanophenone and decanophenone, respectively, accord-
ing to literature methods26,28 (see Scheme 1).
The synthesis of the 2-hydroxy-2-phenylnonanamide
(21) was accomplished using a previously reported
procedure26 for the synthesis of 2-hydroxy-2-phenylpro-
panamide (Scheme 1). Briefly, octanophenone (22) was
treated under anhydrous conditions with cyanotrimeth-
ylsilane to give the cyanotrimethylsilyl ether 23, and
cleavage of the trimethylsilyl group in 23 with 5% HCl
gave cyanohydrin 24 in 100% yield. The hydrolysis of
24 with cold concentrated HCl and HCl gas resulted in
the precipitation of 2-hydroxy-2-phenylnonanamide (21)
in 100% yield.
We also used the final non-cross-validated CoMFA
model to predict the sodium channel binding activities
for all low-energy conformations of the test set com-
pounds 14-21 (Table 4).
P a r tia l Lea st Squ a r es (P LS) Regr ession An a ly-
sis. Cross-validated (preliminary and final models) and
non-cross-validated PLS analyses (final model) were
performed within the SYBYL/QSAR routine. Cross-
validation of the dependent column (log IC50) and the
CoMFA column was performed with 2.0 kcal/mol column
filtering. Scaled by the CoMFA standard deviation, the
final cross-validated analysis generated an optimum
number of components equal to 3 and R2 ) 0.821 (Table
4). PLS analysis with non-cross-validation, performed
with 3 components, gave a standard error of estimate
of 0.099, a probability (R2 ) 0) equal to 0.000, an F value
(n1 ) 4, n2 ) 9) of 286, and a final R2 ) 0.988. The
relative steric (0.748) and electrostatic (0.252) contribu-
tions to the final model were contoured as the standard
deviation multiplied by the coefficient at 80% for favored
steric (contoured in green) and favored positive electro-
static (contoured in blue) effects and at 20% for disfa-
vored steric (contoured in yellow) and favored negative
electrostatic (contoured in red) effects, as shown in
Figure 4.
Resu lts a n d Discu ssion
Our previous sodium channel SAR studies3 suggested
that log P was one of the parameters important for
enhanced binding of hydantoins to the NVSC. We thus
estimated the log P values for analogues 5-7 and 17-
20 (Tables 1 and 2) by the approach we employed
earlier, using reported experimentally measured log P
values for the parent structures, 5-phenylhydantoin (log
P ) 0.46)30 and 2-hydroxy-2-phenylpropanamide (log P
) 1.2),31 to which were added the appropriate π values
for the substituents at C5 and N1 to provide the final
log P.32 The π values used were 0.50 for each CH2 or
On the basis of this analysis, the sodium channel
binding activities for low-energy conformers of ana-

CoMFA of Hydantoin Binding
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9 1541
F igu r e 1. Plot of log P versus observed log IC50 for the
F igu r e 2. Plot of observed log IC50 versus predicted log IC50
training set.
values for the non-cross-validated CoMFA model in Table 3.
Sch em e 1
F igu r e 3. Plot of observed log IC50 versus predicted log IC50
for the test set in Table 4.
binding affinities to the NVSC (e.g., 3 vs 4). Thus a
number of points deviate from the regression line. Note
that, since the most effective inhibitors begin to ap-
proach unfavorable log P values, compounds in the
lower right quadrant which deviate to the left of the
correlation line are most likely to have clinical relevance
(high potency and moderate log P). We thus were
encouraged to use CoMFA to investigate the differences
in structural and electrostatic features of these ana-
logues in an effort to formulate a better model.
The sodium channel binding affinities and lipophi-
licities for the CoMFA training set are presented in
Table 3. PLS correlation of in vitro sodium channel
binding, expressed as -log IC50, and CoMFA descriptors
generated a CoMFA model with R2 ) 0.988 for the
training set (see Table 3). Furthermore, the CoMFA
CH3 in the alkyl chain at C5 and 0.56 for an N-CH3
substitute. For example, the log P for hydantoin 8 was
calculated as (log P5-phenylhydantoin + π5-methyl + πN1-methylene
+ πmethyl) ) (0.46 + 0.50 + 0.56 + 0.50) ) 2.02. Since
the ring closure of 8 provides the bicyclic hydantoin 13,
the log P of 13 was calculated as (log P8 + πring closure
)
) 2.02 + (-0.50) ) 1.52. The log P values for previously
reported compounds2 were similarly calculated.
The correlation of log P versus sodium channel
binding (log IC50) for 1-13 and DPH, shown in Figure
1, resulted in a better quality model than a previously
reported2 log P model, which used fewer compounds and
spanned a smaller log P range. However, comparisons
within the present model of hydantoins with the same
log P and different structures often show very different

1542 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9
Brown et al.
model provided a better correlation than that for the
model (observed IC50 ) 58 µM). Thus even cyclic groups
in the 5-alkyl region favor sodium channel binding.
Additionally, the potential utility of this CoMFA
model for designing new, potent sodium channel binders
was evaluated. Using the CoMFA model, the R-hydroxy-
R-phenylamide 21 was designed as a new structural
class that met all major requirements of the model yet
did not contain the hydantoin ring. Compound 21
contains, like 1-7, a relatively unrestricted phenyl ring
and a seven-carbon 5-alkyl side chain. The success of
this compound as a potent sodium channel binder (IC50
) 9 µM) and the accuracy of prediction by the CoMFA
model validate the alignment rule and reveal that the
intact hydantion ring is not required for good sodium
channel binding activity.
We also used this CoMFA model to predict the activity
of carbamazepine (20), one of the most common anti-
convulsants known to act on the sodium channel, to
further verify the value of the model. The alignment
used for 20 was previously described and is illustrated
in Figure 4C. Despite significant structural differences
between carbamazepine and the hydantoins, the model
accurately predicts the sodium channel binding activity
(actual log IC50 ) 2.12, predicted log IC50 ) 2.10) for
20.
log P model (R2 ) 0.801) for the same data set.
Investigations of the CoMFA steric map revealed that
n-alkyl substitution at the C5 position appears to be
required for tight binding to the hydantoin site. The
contour map in Figure 4 reveals a favorable steric
volume near the 5-alkyl side chain with an optimum
length of 6-7 carbons (further increases gave no en-
hancement in binding). For hydantoins 8-12, which
contain conformationally restricted phenyl rings, the
CoMFA model supports the conclusion that 5-phenyl
ring orientation is important for binding.2 Compounds
9 and 10 have 5-phenyl rings restrained to spatial
arrangements that occupy unfavorable steric regions of
the model and are therefore less active sodium channel
binders. Analogues 11 and 12 have greater phenyl ring
flexibility and can adopt low-energy conformations that
allow for occupation of more favorable steric regions,
consistent with significantly increased binding potency
as compared to hydantions 9 and 10. Figure 4A, which
contains highly active analogue 7 (orange) and the
weakly active hydantoin 9 (green) within the electro-
static and steric CoMFA fields, respectively, illustrates
the effects of proper 5-phenyl ring conformation on
binding.
In conclusion, the CoMFA model in this study was
predictive of sodium channel binding affinities for a
number of structurally diverse hydantoins, as well as
carbamazepine, that were not included in the training
set. Furthermore, the CoMFA model was shown to have
potential utility for designing new agents in that a
structurally novel R-hydroxy-R-phenylamide was also
accurately predicted. Thus this CoMFA model, based
only on electrostatic and steric effects, may prove useful
for designing new and more potent sodium channel-
directed agents.
An important property of a QSAR model is the ability
to accurately predict the relative binding potencies of
new analogues. The predictive ability of the CoMFA
model described here was first evaluated by calculating
activities of previously published monocyclic hydantoins
14-16,2 hydantoin 17, and bicyclic hydantoins 18 and
19.3 As shown in Table 4, calculations agreed remark-
ably well with the experimental results. Hydantoins
14-16 (Table 4), which have essentially identical log P
values, were designed to confirm the importance of
5-phenyl ring orientation independent of changes in log
P.2 Compound 14 contains high-energy phenyl ring
rotamers that are significantly less populated as com-
pared to the less restricted analogues 15 and 16. As
shown in Figure 4B, the phenyl ring of the lowest energy
conformer of 14(a) does not occupy a sterically favored
region of space. This is consistent with a predicted and
observed binding affinity for 14(a) that is 4 times less
potent than that of 15 (for both a and b conformations
of the methyl group, Table 4) and twice as inactive as
16.
Exp er im en ta l Section
Melting points were recorded on an Electrothermal melting
apparatus and are uncorrected. IR spectra were recorded on
Beckman Acculab 6 and Nicolet IR/42 spectrometers, and
elemental analyses were performed by Atlantic Microlabs of
1
Norcross, GA. H and 13C NMR spectra were recorded on GE
300 MHz FT and Bruker 300 FT MHz NMR spectrometers in
DMSO-d6 (for hydantoins) and CDCl3 (for all other compounds)
at ambient temperature and referenced internally to tetra-
methylsilane (TMS). The GC/MS analyses were performed on
a Hewlett-Packard 5885 GC/MS. Statistics and graphs were
generated with SigmaStat (version 1.02a) and Sigma Plot
(version 1.0) for Windows from J andel, Inc. [3H]Batrachotoxi-
nin A 20-R-benzoate ([3H]BTX-B) with a specific activity of 30
Ci/mol was obtained from New England Nuclear (Boston, MA).
5-Hep tyl-5-p h en ylh yd a n toin (6). A mixture containing
octanophenone (22) (1.5 g, 7.5 mmol), cyanontrimethylsilane
(TMSCN) (0.70 g, 7.5 mmol), and ZnI2 (5-10 mg) was stirred
at room temperature under a N2 atmosphere. The reaction
progress was monitored by the disappearance of the CdO
stretching peak (1670 cm-1) in the IR spectrum of the reaction
mixture. After 2 h this peak was absent, indicating complete
conversion to give 2-(trimethylsilyloxy)-2-phenylnonanecar-
bonitrile (23). The TMS ether 23 was hydrolyzed to the
cyanohydrin 24 by adding ether (10 mL) and 15% HCl (10 mL),
and the resulting mixture was stirred vigorously for 1 h. The
acidic layer was washed with ether (3 × 20 mL). The extracts
were combined and evaporated to give cyanohydrin 24 (1.8 g,
100%): IR (neat) 3400 (OH), 2250 (CN) cm-1. The cyanohydrin
24 was converted to hydantoin by a previously reported
literature procedure for similar compounds.2 Using this pro-
The substitution of a p-methyl group in compound 16
does not appreciably alter phenyl ring orientation
relative to a nonsubstituted phenyl ring. Consistently,
compound 16 (IC50 ) 95 µM) exhibited essentially the
same binding potency as 2 (IC50 ) 103 µM), which has
the same size n-butyl side chain. Comparisons of
analogue 16 (same log P and phenyl ring torsion angle)
with hydantoin 4 suggest that, given the same log P and
phenyl torsion angle, longer alkyl side chains provide
better sodium channel binders. Further, m-methyl may
be favored over p-methyl substitutions, as suggested by
the moderately increased binding potency of 15 as
compared to 16 (same log P and phenyl ring torsion).
Compound 17 was designed to test the influence of a
more bulky, cyclic alkyl group in the 5-alkyl region.
Interestingly, the sodium channel binding potency of
compound 17 was correctly predicted by the CoMFA

CoMFA of Hydantoin Binding
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9 1543
F igu r e 4. Stereoviews (relaxed) of the electrostatic and steric CoMFA fields for selected compounds. For electrostatic contours increased binding results from placing more
positive (+) charge near blue and negative (-) charge near red. For steric contours increased binding results from placing more bulk near green and less bulk near yellow.
A. Active analogue 7 (orange) and inactive analogue 9 (green). B. Active analogue 15 (red) and less active analogue 14 (blue). C. Carbamazepine (20; red) and DPH (blue).

1544 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9
Brown et al.
cedure, cyanohydrin 24 (1.8 g, 9.0 mmol) and (NH4)2CO3 (3.6
g, 3.6 mmol) were dissolved in 50% ethanol (30 mL) while
stirring under a N2 atmosphere. The mixture was heated
slowly between 40 and 60 °C for 12 h. The basic mixture was
evaporated to one-half volume and cooled to room temperature.
The precipitate was filtered and recrystallized from hot ethanol
to give 6 (0.50 g, 21%): mp 124-126 °C; IR (KBr) 1710, 1755
[3H]BTX-B. Each data point used for generating the IC50 curve
was the mean of triplicate experiments, and each experiment
included a control tube containing 40 µM DPH. The IC50 values
were determined from a Probit analysis of the dose-response
curve and excluded doses producing less than 10% or greater
than 90% inhibition.
(CdO), 3200 (NH) cm-1
.
Ack n ow led gm en t. This work was completed in
partial fulfillment of the requirements for the Ph.D.
degree in Organic Chemistry by M.L.B. This work was
supported in part by financial support from the Patricia
Robert Harris Fellowship, the National Consortium for
Educational Access, the UAB Comprehensive Minority
Faculty Development Program, and the UAB Depart-
ment of Chemistry. We also thank Ms. Bereaval Webb,
an Alabama Alliance for Minority Participation Summer
Intern, for technical support.
5-Non yl-5-p h en ylh yd a n toin (7). To a stirring solution of
50% ethanol (50 mL) were added decanophenone (3.0 g, 13
mmol), KCN (1.7 g, 26 mmol), and (NH4)2CO3 (5.9 g, 52 mmol).
The solution was warmed to 50-65 °C for 12 h. The precipitate
was filtered, and the filtrate was acidified (pH 2) using
concentrated HCl. The resulting solid was filtered, and the
filtrate was made basic (pH 8) using 3% KOH. This was
concentrated to one-half volume and filtered again. The
combined solids were recrystallized from hot ethanol to give
pure 7 (3.7 g, 95% yield): mp 115-116 °C; IR (KBr) 1700, 1750
(CdO), 3200 (NH) cm-1
.
2-Hyd r oxy-2-p h en yln on a n a m id e (21). The cyanohydrin
24 was prepared as described in the synthesis of 6, except that
octanophenone (22) (4.1 g, 2.0 mmol), TMSCN (2.0 g, 2.0
mmol), ZnI2 (5-10 mg), ether (20 mL), and 5% HCl (10 mL)
were used. Cyanohydrin 24 was added to cold concentrated
HCl (10 mL) and saturated with HCl gas while stirring. After
4 h a precipitate began to form, and the mixture was then
allowed to stand overnight without stirring. The solid was
filtered, washed on the filter with cold water (3 × 50 mL), and
recrystallized from hot toluene to give pure 21 (5.3 g, 100%):
mp 91-92 °C.
Refer en ces
(1) Willow, M.; Catterall, W. A. Inhibition of Binding of [3H]-
Batrachotoxinin A 20-R-Benzoate to Sodium Channels by the
Anticonvulsant Drugs Diphenylhydantoin and Carbamazepine.
Mol. Pharmacol. 1982, 22, 627-635.
(2) Brown, M. L.; Brown, G. B.; Brouillette, W. J . Effects of Log P
and Phenyl Ring Conformation on the Binding of 5-Phenylhy-
dantoins to the Voltage-Dependent Sodium Channel. J . Med.
Chem. 1997, 40, 602-607.
(3) Brouillette, W. J .; J estkov, V. P.; Brown, M. L.; Aktar, M. S.;
Delorey, T. M.; Brown, G. B. Bicyclic Hydantoins with Bridge-
head Nitrogen. Comparison of Anticonvulsant Activities with
Binding to the Neuronal Voltage-Dependent Sodium Channel.
J . Med. Chem. 1994, 37, 3289-3293.
Sod iu m Ch a n n el Bin d in g Assa y. The details for the
sodium channel binding assay were reported in a previous
paper.32 Synaptoneurosomes were prepared from rat cerebral
cortex as follows. The cerebral cortex (gray matter) was
obtained by removing the white matter and other subcortical
structures. An incubation buffer was prepared which contained
130 mM choline chloride, 50 mM HEPES, 5.5 mM glucose, 0.8
mM MgSO4, and 5.4 mM KCl, adjusted to pH 7.4 using Tris
base. The tissue (about 1 g) was then homogenized in 2 mL of
incubation buffer using 10 full strokes of a glass-glass
homogenizer. The tissue preparation was transferred to a
centrifuge tube, the homogenizer was rinsed with 3 mL of the
incubation buffer, and the rinse was combined with the
preparation. The preparation was then centrifuged at 1000g
for 15 min at 4 °C. The pellet was resuspended in a total
volume of 20 mL of the incubation buffer and transferred to a
50-mL homogenizer. Three full strokes were used to homog-
enize the tissue. The suspension was gently filtered through
three layers of 160-µm nylon mesh and then through a
Whatman #4 filter paper using house vacuum. The filtrate was
centrifuged at 1000g for 30 min at 4° C. The pellet was
resuspended with 5 mL of isotonic sucrose solution, which
contained 10 mM NaH2PO4 and 0.32 mM sucrose at pH 7.4
(Tris base). Additional isotonic solution was added as needed
to obtain an absorbance of 1 at 280 nm. The isotonic suspen-
sion was stored in a freezer for up to 3 months at -70 °C until
needed. Prior to the binding assay the tissue was thawed, the
suspension was centrifuged at 1000g for 20 min at 4 °C, and
the pellet was resuspended in the HEPES incubation buffer
(using the same volume as the isotonic buffer used to store
the tissue).
(4) Francis, J .; Burnham, W. M. [3H]Phenytoin Identifies a Novel
Anticonvulsant-Binding Domain on Voltage-Dependent Sodium
Channels. Mol. Pharmacol. 1992, 42, 1097-1103.
(5) Barber, M. J .; Starmer, C. F.; Grant, A. O. Blockade of Cardiac
Sodium Channels by Amitriptyline and Diphenylhydantoin.
Evidence for Two Use-dependent Binding Sites. Cir. Res. 1991,
69, 677-696.
(6) Greco, G.; Novellino, C. S.; Vittora, A. Comparative Molecular
Field Analysis on a Set of Muscurinic Agents. QSAR 1991, 10,
289-299.
(7) Kim, K. H.; Greco, G.; Novellino, E.; Silipa, C.; Vittoria, A. Use
of the Hydrogen Bond Potential Function in a Comparative
Molecular Field Analysis (CoMFA) on a Set of Benzodiazepines.
J . Comput.-Aided Mol. Des. 1993, 7, 263-280.
(8) Giovanni, G.; Ettore, N.; Fiorini, I.; Nacci, B.; Campiani, G.;
Ciani, S. M.; Garofalo, A.; Bernasconii, P.; Mennini, T. Compara-
tive Molecular Field Analysis Model for 6-Arylpyrrolo[2,1-d][1,5]-
Benzothiazepines Binding Selectively to the Mitochondrial
Benzodiazepine Receptor. J . Med. Chem. 1994, 37, 4100-4108.
(9) Wong, G.; Koehler, K. F.; Skolnick, P.; Gu, Z.-Q.; Ananthan, P.;
Scho¨nholzer, W.; Hyunkeler, W.; Zhang, W.; Cook, M. Synthetic
and Computer-Assisted Analysis of the Structural Requirements
for Selective, High-Affinity Ligand Binding to Diazepam-
Insensitive Benzodiazepine Receptors. J . Med. Chem. 1993, 36,
1820-1830.
(10) Andre, M. M.; Charifson, P. S.; Constance, E. O.; Kula, N. S.;
McPhail, A. T.; Baldessarini, R. J .; Booth, R. G.; Wyrick, S. D.
Conformational Analysis, Pharmacophore Identification, and
Comparative Molecular Field Analysis of Ligands for the Neu-
romodulatory δ3 Receptor. J . Med. Chem. 1994, 37, 4109-4117.
(11) Ablordeppy, S. Y.; El-Ashmaway, M. B.; Glennon, R. A. Analysis
of the Structure Activity Relationships of Sigma Ligands. Med.
Chem. Res. 1991, 1, 425-438.
For the sodium channel binding assay, synaptoneurosomes
(∼1 mg of protein) from rat cerebral cortex were incubated
for 40 min at 25 °C with the test compound (seven different
concentrations spanning the IC50) in a total volume of 320 µL
containing 10 nM [3H]BTX-B (specific activity of 30 Ci/mol)
and 50 µg/mL scorpion venom. Incubations were terminated
(12) Thomas, B. F.; Compton, B. R. M.; Semus, S. F. Modeling the
Cannabinoid Receptor: A Three-Dimensional Quantative Struc-
ture Activity Analysis. Mol. Pharmacol. 1991, 40, 656-661.
(13) Semus, S. F.
A Computer Graphic Investigation into the
Structural Requirements for Interaction with the Putative
Cannabinoid Receptor Site. Med. Chem. Res. 1991, 1, 454-460.
(14) Chen, J . M.; Sheldon, A. Structure-Function Correlations for
Calcium Binding and Calcium Channel Activities Based on
3-Dimensional Models of Human Annexins I, II, III, V, VII.
Biomol. Struct. Dynam. 1993, 10, 1067-1089.
(15) Rustici, M.; Bracci, L.; Lozzi, P.; Nari, P.; Santucci, A.; Sodani,
P.; Spreafico, A.; Niccolai, N. A Model of the Rabies Virus
Glycoprotein Active Site. Biopolymers 1993, 33, 961-969.
by dilution with ice-cold buffer and filtration through
a
Whatman GF/C filter paper, and the filters were washed four
times with ice-cold buffer. Filters were counted in a Beckmann
scintillation counter. Specific binding was determined by
subtracting the nonspecific binding, which was measured in
the presence of 300 µM veratridine, from the total binding of

CoMFA of Hydantoin Binding
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 9 1545
(16) Nordvall, G.; Hacksell, U. Binding Site Modeling of the Musca-
rinic M1 Receptor: A Combination of Homology-Based and
Indirect Approaches. J . Med. Chem. 1993, 36, 967-976.
(17) Caroll, F. I.; Mascrella, S. W.; Kuzemko, M. A.; Goa, Y. G.;
Abraham, P.; Lewin, A. H.; Boja, J . W.; Kuhar, M. J . Synthesis,
Ligand-Binding and QSAR (CoMFA and Classical) Study of 3â-
(3′-Substituted Phenyl), 3â-(4′-Substituted Phenyl), 3â-(3′,4′-
Disubstituted Phenyl)tropane-2â-Carboxylic Acid Methyl-Esters.
J . Med. Chem. 1994, 37, 2865-2873.
(18) Agarwal, A.; Taylor, E. W. 3-D QSAR for Intrinsic Activity of
5-HT1A Receptor Ligands by the Method of Comparative Molec-
ular Field Analysis. J . Comput. Chem. 1993, 14, 237-245.
(19) Langer, T.; Wermuth, C. G. Inhibitors of Prolyl Endopeptidase:
Characterization of the Pharmacophoric Pattern Using Confor-
mational Analysis and 3D-QSAR. J . Comput.-Aided Mol. Des.
1993, 7, 253-262.
(20) Allen, M. S.; LaLoggia, A. J .; Dorn, M. J .; Constantino, G.;
Hagen, T. J .; Koehler, K. F.; Skolnick, P.; Cook, J . M. Predictive
Binding of 0-Carboline Inverse Agonist and Antagonists via the
CoMFA/GOLPE Approach. J . Med. Chem. 1992, 35, 4001-4010.
(21) Calder, J . A.; Wyatt, J . A.; Frenkel, D. A.; Casida, J . E. CoMFA
Validation of the Superposition of Six Classes of Compounds
Which Block GABA Receptors Non-Competitively. J . Comput.-
Aided Mol. Des. 1993, 7, 45-60.
(26) Grunewald, G. L.; Brouillette, W. J .; Finney, J . Synthesis of
R-Hydroxyamides via the Cyanosilylation of Aromatic Ketones.
Tetrahedron Lett. 1980, 21, 1219-1220.
(27) Sarges, R.; Schnur, R. C.; Belletire, J . L.; Peterson, M. J . Spiro
Hydantoin Aldose Reductase Inhibitors. J . Med. Chem. 1988,
31, 230-243.
(28) Novelli, A.; Lugones, Z. M.; Velasco, P. Hydantoins III. Chemical
Constitution and Hypnotic Action. Anales Asoc. Quim. Argentina
1942, 30, 225-231.
(29) Huisgen, R.; Ugi, I. Polycyclische Systeme mit Heteroatomen.
Liebigs Ann. Chem. 1957, 610, 57-66.
(30) Lien, E. Structure-Activity Correlations for Anticonvulsant
Drugs. J . Med. Chem. 1970, 13, 1189-1191.
(31) Hernandez-Gallegos, Z.; Lehmann, P. A. F. Partition Coefficients
of Three New Anticonvulsants. J . Pharm. Sci. 1990, 79, 1032-
1033.
(32) Brouillette, W. J .; Brown, G. B.; Delorey, T. M.; Shirali, S. S.;
Grunewald, G. L. Anticonvulsant Activities of Phenyl-Substi-
tuted Bicyclic 2,4-Oxazolidinediones and Monocyclic Models.
Comparison with Binding to the Neuronal Voltage-Dependent
Sodium Channel. J . Med. Chem. 1988, 31, 2218-2221.
(33) Lowes, M. M. J .; Caira, M. R.; Lo¨tter, A. P.; Van Der Watt, J . G.
Physicochemical Properties and X-ray Structural Studies of the
Trigonal Polymorph of Carbamazepine. J . Pharm. Sci. 1987, 76,
744-752.
(34) Camerman, A.; Camerman, N. Stereochemical Basis of Anti-
convulsant Drug Action II. Molecular Structure of Diazepam.
J . Am. Chem. Soc. 1972, 94, 268-272.
(35) Hanson, A. W.; Ro¨hrl, M. The Crystal Structure of Lidocaine
Hydrochloride Monohydrate. Acta Crystallogr. 1972, B28, 3567-
3571.
(22) Rein, K. S.; Baden, D. G.; Gawley, R. E. Conformational Analysis
of the Sodium Channel Modulator, Brevetoxin A. Comparison
with Brevetoxin B Conformations, and a Hypothesis About the
Common Pharmacophore of the “Site 5” Toxins. J . Org. Chem.
1994, 59, 2101-2106.
(23) Clark, M. D.; Cramer, R. D., III; Opdenbosch, N. V. Validation
of the General Purpose Tripos 5.2 Force Field. J . Comput. Chem.
1989, 10, 982-1012.
(36) Zima´nyi, I.; Weiss, S. R. B.; Lajtha, A.; Post, R. M.; Maarten, E.
A. R. Evidence for a Common Site of Action of Lidocaine and
Carbamazepine in Voltage-Dependent Sodium Channels. Eur.
J . Pharmacol. 1989, 167, 419-422.
(24) Brouillette, W. J .; Brown, G. B.; Delorey, T. M.; Liang, G. Sodium
Channel Binding and Anticonvulsant Activities of Hydantoins
Containing Conformationally Constrained 5-Phenyl Substitu-
ents. J . Pharm. Sci. 1990, 79, 871-874.
(25) Camerman, A.; Camerman, N. The Stereochemical Basis of
Anticonvulsant Drug Action. I. The Crystal and Molecular
Structure of Diphenylhydantoin, a Noncentrosymmetric Struc-
ture Solved by Centric Symbolic Addition. Acta Crystallogr.
1971, B27, 2205-2211.
(37) Postma, S. W.; Catterall, W. A. Inhibition of Binding of 3H-
Batrachotoxin A 20-0-Benzoate to Sodium Channels by Local
Anesthetics. Mol. Pharmacol. 1984, 25, 219-227.
J M980556L