Inactivation and inhibition of γ-aminobutyric acid aminotransferase by conformationally restricted vigabatrin analogues

Article abstract of DOI:10.1021/jm020134i

Four cyclohexene analogues of γ-aminobutyric acid (GABA) and β-alanine were designed as conformationally rigid analogues of the epilepsy and drug addiction drug vigabatrin and as potential mechanism-based inactivators of γ-aminobutyric acid aminotransfera

Full text of DOI:10.1021/jm020134i

J . Med. Chem. 2002, 45, 4531-4539
4531
In a ctiva tion a n d In h ibition of γ-Am in obu tyr ic Acid Am in otr a n sfer a se by
Con for m a tion a lly Restr icted Viga ba tr in An a logu es
†
Sun Choi and Richard B. Silverman*
Department of Chemistry, Department of Biochemistry, Molecular Biology, and Cell Biology, and the Drug Discovery Program,
Northwestern University, Evanston, Illinois 60208-3113
Received March 25, 2002
Four cyclohexene analogues of γ-aminobutyric acid (GABA) and â-alanine were designed as
conformationally rigid analogues of the epilepsy and drug addiction drug vigabatrin and as
potential mechanism-based inactivators of γ-aminobutyric acid aminotransferase (GABA-AT).
The corresponding cyclopentene analogues were previously reported to be inhibitors, but not
inactivators, of GABA-AT (Qiu, J .; Pingsterhaus, J .; Silverman, R. B. J . Med. Chem. 1999, 42,
4
1
725-4728). cis-3-Aminocyclohex-4-ene-1-carboxylic acid (3) and cis-2-aminocyclohex-3-ene-
-carboxylic acid (5) showed time- and concentration-dependent, irreversible inactivation of
GABA-AT. In both cases, the inactivations are protected by substrate, indicating that they are
active site-directed. trans-3-Aminocyclohex-4-ene-1-carboxylic acid (4) and trans-2-aminocy-
clohex-3-ene-1-carboxylic acid (6) are not inactivators but are competitive reversible inhibitors
of GABA-AT. Unlike the cyclopentene analogues, there appears to be sufficient ring flexibility
to allow inactivation to occur. The orientation of the carboxylic and amino groups of these
analogues is important for their binding to GABA-AT. Molecular modeling of GABA-AT with
3
-6 and molecular dynamics simulations with vigabatrin bound provide rationalizations for
the inhibitory properties of these compounds.
In tr od u ction
Presumably, the reason that two inactivation mecha-
nisms arise is because the orientations of the pyridoxal
ring system and the vinyl π-orbitals relative to the
electrons in the γ-C-H bond of γ-vinyl-GABA allow
γ-Aminobutyric acid (GABA) is an important inhibi-
tory neurotransmitter in the mammalian central ner-
1
vous system. The major pathway for its degradation is
delocalization of those electrons in either direction.13
A
via transamination with R-ketoglutarate catalyzed by
the pyridoxal 5′-phosphate (PLP)-dependent enzyme
GABA aminotransferase (GABA-AT, E. C. 2.6.1.19).
Inhibition of this enzyme results in an increase in
availability of GABA, which can have a beneficial effect
molecular model of vigabatrin bound to the PLP sup-
ported the possibility that the lysine residue that
removes the γ-proton (B in Scheme 1) could then become
2
+
the acid donor (B -H in Scheme 1) for either the
cofactor (pathway a) or the vinyl group (pathway b).14
3
on neurological disorders including epilepsy, Parkin-
4
4b,5
son’s disease, Huntington’s chorea,
and Alzheimer’s
6
disease. Recently, it has been found that an increase
7
in GABA also blocks the effects of drug addiction.
Selective inactivation of GABA-AT by 4-aminohex-5-
8
,9
enoic acid (1, Scheme 1; γ-vinyl-GABA or vigabatrin),
We wanted to design an inactivator that functioned
only through one inactivation mechanism and thought
that by constraining the vinyl π-orbitals in a conforma-
tionally rigid analogue of vigabatrin, such as 4-amino-
a mechanism-based inactivator10 of the enzyme, is
already being successfully applied in the treatment of
epilepsy and is in clinical trials for the treatment of
1
1
1
2
drug addiction. A detailed study of the mechanism of
inactivation of GABA-AT by vigabatrin revealed that it
acted by two principal inactivation pathways: a Michael
addition mechanism (pathway a), which accounts for
about 70% of the total inactivation, and an enamine
mechanism (pathway b, Scheme 1), which accounts for
about 30% of the inactivation. These two pathways
occur via deprotonation of the γ-carbon followed by
tautomerization through either the PLP ring (pathway
a, leading to a Michael addition) or the vinyl double
bond (pathway b, leading to an enamine formation).
2
-cyclopentene-1-carboxylic acid (2), one of the two
1
5
inactivation pathways may be excluded. Unfortu-
nately, 2 is so constrained that neither the cis nor the
trans analogue is an inactivator of GABA-AT, although
the cis isomer is an excellent substrate and the trans
isomer is a competitive inhibitor. A molecular model of
vigabatrin bound in GABA-AT indicated that prevention
of rotation of the vigabatrin vinyl group should prevent
the Michael addition pathway, so a slightly less
conformational constraint was needed. To provide more
flexibility for the double bond, the corresponding cyclo-
hexene vigabatrin analogues (3-6) were designed and
synthesized. Compound 3 was found to be a time-
dependent inactivator, which did indeed inactivate
GABA-AT by only one mechanism (analogous to path-
1
3
1
4
*
To whom correspondence should be addressed. Tel: (847)491-5653.
Fax: (847)491-7713. E-mail: Agman@chem.northwestern.edu.
†
Current address: Tripos, Inc. 1699 South Hanley Road, St. Louis,
MO 63144.
1
0.1021/jm020134i CCC: $22.00 © 2002 American Chemical Society
Published on Web 08/28/2002

4
532 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20
Choi and Silverman
Sch em e 1
Sch em e 2
Sch em e 3
way b in Scheme 1).14 Here, we report the inactivation
of GABA-AT by 5, the reversible inhibition by 4 and 6,
and rationalize their activity by molecular modeling.
were separated by repeated silica gel column chroma-
tography (Scheme 3). The alkyl carbamate and alkyl
ester groups in each isomer were deprotected simulta-
neously with iodotrimethylsilane, and 5 and 6 were
purified by ion exchange chromatography. The structure
of 5 was confirmed by X-ray crystallography (Figure 1).
Ch em istr y
2
0
The preparation of (d,l)-trans-3-azidocyclohex-4-ene-
1
-carboxylic acid (8) was carried out by SN2 azide
1
6
substitution of 7-oxabicyclo[3,2,1]-oct-2-en-6-one (7),
which was synthesized from cyclohex-3-ene-1-carboxylic
1
7
Resu lts a n d Discu ssion
Incubation of GABA-AT with 314 and 5 led to time-
and concentration-dependent inactivation of the en-
zyme. A Kitz and Wilson replot of the data gave the
kinetic constants shown in Table 1. After they were gel-
filtered, no return of enzyme activity was observed,
supporting the irreversible nature of the inactivation.
In the presence of GABA, the rate of inactivation
acid by the procedure of Marshall et al. (Scheme 2).
Subsequent reduction of the allyl azide with PPh3 in
1
8
tetrahydrofuran (THF) with 1 equiv of water gave
2
1
(d,l)-4.
A mixture of ethyl cis- and trans-6-carbomethoxy-2-
cyclohexen-1-yl carbamates (9 and 10, in a 4:1 ratio) was
prepared from the Diels-Alder reaction of ethyl trans-
1
9
1
,3-butadiene-1-carbamate and methyl acrylate, which

γ-Aminobutyric Acid Aminotransferase
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20 4533
F igu r e 1. ORTEP drawing of 5.
Ta ble 1. Kinetic Constants for the Cyclohexene Analoguesa
kinact
KI
(mM)
kinact/KI
Ki
(µM)
(min-1)
(min mM-1)
-1
compd
3a
4
5
0.01
0.03
0.24
2.3
13.0
0.85
0.0042
0.0024
0.28
100
125
6
vigabatrin
a
Choi, S.; Storici, P.; Schirmer, T.; Silverman, R. B. J . Am.
Chem. Soc. 2002, 124, 1620-1624.
diminished, indicating that the reaction occurs at the
active site of the enzyme. The trans isomers, 4 and 6,
however, showed no time-dependent inhibition of the
enzyme, but they were potent competitive reversible
inhibitors (Table 1).
Inactivation by 314 and 5 is quite remarkable, given
F igu r e 2. (A) Molecular modeling of vigabatrin bound to show
that Lys-329 is not in the correct orientation for Michael
addition. Only the γ-H and Schiff base N-H are explicitly
displayed; the rest of the hydrogens are omitted for clarity.
that the corresponding cyclopentene analogue 2 does not
inactivate the enzyme.15 By analogy to the two mech-
anisms of inactivation of GABA-AT by vigabatrin
(B) Molecular model to show that Michael addition requires
(
Scheme 1),13 5 could react by one or both of these
the vinyl group to rotate toward Lys-329.
possible mechanisms (Scheme 4). Expansion of the
cyclopentene ring by one methylene provides sufficient
conformational flexibility to allow for tautomerization
and rearrangement of the enamine (pathway b) but
presumably not for nucleophilic attack (pathway a).
Recently, we found that 3 inactivates GABA-AT exclu-
sively by the enamine mechanism.14 This was rational-
ized based on molecular modeling results, using coor-
dinates from the X-ray crystal structure of pig liver
GABA-AT,22 which show that the lysine residue that
binds to the PLP cofactor, known to be the residue
labeled by vigabatrin,23 is not aligned with the vinyl
group in its lowest energy orientation (Figure 2A);
rotation of the vinyl group is imperative for nucleophilic
attack by that lysine residue (Figure 2B). Consequently,
both tautomerization and inactivation pathways could
occur.
chain nitrogen of Lys-329 and the terminal carbon of
the vinyl group were examined from the recorded
trajectories of the simulation. After it reached equilib-
rium, the distances were mostly in the range of 3.2-
4.1 Å (Figure 3B), indicating that the rotatable vinyl
group would be close enough to the nucleophilic amino
group in the flexible side chain of Lys-329 to allow for
Michael addition (Figure 2). In the case of 3, the
conformational rigidity of the cyclohexene ring prohibits
the double bond from rotating sufficiently to allow
nucleophilic addition (data not shown), which results
in only pathway b as the inactivation pathway.14
Compound 5, however, can bind in two important
conformations, which could allow for tautomerization
by both pathways and multiple inactivation products.
Figure 4 shows both conformers from two different
perspectives; conformer A does not allow Michael ad-
dition after tautomerization, but B does allow Michael
addition. The observation of pyridoxamine 5′-phosphate
(PMP) after denaturation of the inactivated enzyme
signals the Michael addition mechanism (see Scheme
4, pathway a), and the observation of modified cofactor
with a TR in the range of 34-35 min in our high-
performance liquid chromatography (HPLC) system
To gain further evidence to support the requirement
for vinyl group rotation, a molecular dynamics (MD)
simulation of the vigabatrin-enzyme complex was
carried out by raising the temperature to 500 K over a
period of 5000 fs to excite all of the bonds in the protein
randomly, followed by cooling to 300 K over a period of
5
000 fs to allow the bonds in the complex to reach
equilibrium (Figure 3A). The distances between the side

4
534 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20
Choi and Silverman
Sch em e 4
denotes the enamine mechanism (Scheme 4, pathway
the same electrostatic interactions of their carboxylate
b). In fact, inactivation of GABA-AT reconstituted with
group with the active site Arg-192 for binding, as shown
3
22,24
[
H]PLP by 5 followed by HPLC separation of the
by modeling of vigabatrin or GABA.
Our docking
coenzyme-derived products gave tritiated PLP, PMP,
and multiple products in the region where modified
coenzyme adducts occur (TR ) 35 min) (Figure 5),
studies also confirmed this. Another interesting result
observed from the docking of these regioisomers is their
internal hydrogen-bonding interactions. The S isomer
of inactivator 3 exhibits the normal internal hydrogen-
bonding pattern (i.e., between the aldimine N-H and
the hydroxyl O of PLP) (Figure 6A). However, the
internal hydrogen-bonding partner of the N-H in the
aldimine of inactivator 5 with the same stereochemistry
is switched to the carboxylate (Figure 6B). This H-bond
interchange appears to optimize the conformation of the
aldimine of 5 to maintain the binding interaction
between the carboxylate and the Arg-192 and at the
same time to orient the γ-H properly for reactivity with
Lys-329.
1
4
whereas 3 gave only a single peak (at TR ) 34 min).
It is not clear if the multiple products at the longer
retention times contain decomposition products of the
enamine pathway adduct or possibly carboxylated and
decarboxylated forms. The proposed modified cofactor
(11, Scheme 4) was synthesized by aldol condensation
of PLP with cyclohexanone and was shown to have the
same HPLC retention time and mass as the expected
product obtained by enzyme inactivation.
The inability of the trans isomers to inactivate GABA-
AT appears to be related to their binding orientations
at the active site. A homology modeling-deduced active
On the other hand, the R enantiomer of the trans
inhibitor 4 and the S enantiomer of 6 (which has the
same stereochemistry as the R enantiomer of 4) should
be the actual competitive inhibition species to accom-
modate their binding at the active site with the same
electrostatic interactions of their carboxylate group with
Arg-192 (Figure 7). When this salt bridge forms, how-
ever, the proton that must be abstracted in order for
inactivation to occur (the equivalent to the γ-proton in
GABA) is oriented away from Lys-329, preventing its
removal. Therefore, it competitively binds, but catalysis
cannot be initiated, so inactivation does not occur.
For comparison, the cis-cyclopentene vigabatrin ana-
logue (1R,4S)-(+)-2, which showed good substrate activ-
2
4
site structure of GABA-AT was previously reported,
which was used to explain the substrate specificity and
stereospecificity of the enzyme, as well as the unique
reactivity of the S isomer as compared to the R isomer
of vigabatrin as a mechanism-based inactivator. Only
the reactive S isomer of vigabatrin has the scissile
γ-C-H bond perpendicular to the coenzyme ring plane
and the γ-proton readily accessible to Lys-329, the
proposed general base of the reaction. The inactive R
isomer has the vinyl group directed toward Lys-329 and
the γ-C-H bond toward Arg-445, making the γ-proton
inaccessible to Lys-329.24 This same phenomenon was
demonstrated in modeling studies with the X-ray crystal
2
2
15
structure of pig liver GABA-AT, whose active site
residues, which are within a distance of 10 Å from the
cofactor, have identical counterparts with human brain
GABA-AT. It is probable that the active species in
racemic cis inactivator 3 has the S configuration as in
the case of vigabatrin, but because of the movement of
the carboxylate group in 5 to the carbon adjacent to the
amino group, the same stereochemistry in 5 is called
the R configuration to accommodate the R/S nomencla-
ture rules. These compounds probably are involved in
ity (actually better than GABA), was docked into the
enzyme (Figure 8A). It also binds well at the active site,
directed by the salt bridge interaction between the
carboxylate and the Arg-192. After it binds, its γ-H is
positioned properly for abstraction by the active site
base, Lys-329 (at a distance of 2.7 Å). Furthermore, this
proton is oriented perpendicular to the plane of the
pyridine ring (with a torsional angle CdN-C-H of
about 82°), making the γ-H-C bond easily broken
according to the Dunathan Postulate, i.e., the orienta-

γ-Aminobutyric Acid Aminotransferase
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20 4535
F igu r e 4. Molecular model of two low-energy conformations
of 5.
while maintaining the same electrostatic interaction of
the carboxylate with the active site Arg-192 for binding.
In summary, these results demonstrate that the
conformationally restricted, cyclohexene vigabatrin ana-
logues 3 and 5 irreversibly inhibit GABA-AT, and
inactivation occurs at the active site. Although 3 pro-
hibits the Michael addition pathway, 5 allows both the
Michael addition and the enamine pathways to operate.
The corresponding trans isomers 4 and 6 are competi-
tive reversible inhibitors of the enzyme because, once
bound, the proton that needs to be removed for the
reaction is not oriented toward the active site base (Lys-
F igu r e 3. (A) Total energy of the vigabatrin-enzyme complex
during MD simulation by raising the temperature to 500 K
over a period of 5000 fs and cooling to 300 K over a period of
5
000 fs. (B) Snapshot distances between the side chain
nitrogen of Lys-329 and the terminal carbon of the vigabatrin
vinyl group, measured at every 100 fs from the MD simulation.
tion of the bond to be cleaved should be perpendicular
to the cofactor-substrate π-system.25 This γ-C-H bond
breakage, however, does not lead to mechanism-based
inactivation of the enzyme because there is neither a
nucleophile available for Michael addition nor a proton
donor for the enamine mechanism. Instead, it is turned
over as a substrate, presumably via the mechanism
shown in Scheme 5. A comparison of a molecular model
of 2 with GABA at the active site of the enzyme explains
the stereospecificity of 2 (Figure 8B), namely, the same
as that for GABA (the pro-S proton on C4 of GABA).
Docking results show that the pro-S proton is properly
oriented for abstraction by Lys-329 (at a distance of 2.2
Å) with a torsional angle of CdN-C-H of about 73°,
3
29). Therefore, the binding orientations of the carbox-
ylic and amino groups are important to the type of
inhibition that occurs, which suggests future structural
modifications.
Exp er im en ta l Section
Gen er a l Meth od s. H and 13C NMR spectra were recorded
on Varian Gemini 300 MHz and Inova 500 MHz NMR
spectrometers. Chemical shifts are reported as δ values in
1
parts per million (ppm) downfield from Me
4
Si (δ 0.0) as the
internal standard in CDCl . For samples run in D O, the HOD
3
2
resonance was arbitrarily set at 4.80 ppm. X-ray crystal

4
536 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20
Choi and Silverman
3
F igu r e 5. HPLC profile of 5 after inactivation of [ H]PLP-GABA-AT. The retention time of authentic PMP and PLP matched
with the first and second peaks, respectively.
F igu r e 6. Molecular models comparing hydrogen-bonding
interactions of 3 and 5.
structure analysis (Bruker SMART-1000 CCD) and mass
spectral analyses (Micromass Quattro II electrospray triple
quadrupole) were performed by the Analytical Services Labo-
ratory in the Department of Chemistry, Northwestern Uni-
versity. Electron impact high-resolution mass spectra (HRMS)
F igu r e 7. Molecular models of 4 and 6.
were obtained with a VG70-250SE spectrometer. Combustion
analyses were performed by Oneida Research Laboratories,
NY. Thin-layer chromatography (TLC) was run with silica gel

γ-Aminobutyric Acid Aminotransferase
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20 4537
F igu r e 8. Comparison of molecular models of 2 and GABA.
Sch em e 5
6
0 F254 precoated glass plates (EM Separations Technology)
dC(5)H). 13C NMR (D
49.9 (C3), 125.8 (C4), 138.9 (C5), 187.8 (CO
2
O): δ 32.3 (C6), 33.8 (C2), 41.9 (C1),
). HRMS (EI) (m/
, 140.0712; found, 140.0718.
: C, 59.56; H, 7.85; N, 9.92. Found:
using standard visualization techniques. Flash column chro-
matography was carried out with silica gel 60 (230-400 mesh)
from Merck. Cation exchange chromatography was performed
on Dowex 50 resin (BioRad AG50W-X8, 100-200 mesh, biotech
grade). Unless otherwise noted, all reactions were carried out
under a nitrogen atmosphere. An Orion research model 701
pH meter with a general combination electrode was used for
pH measurements. Enzyme assays were recorded on a Perkin-
Elmer Lambda 10 UV/vis spectrophotometer. Radioactivity
was measured by liquid scintillation counting using a Packard
Tri-Carb 2100TR counter and Packard Ultima Gold scintilla-
tion cocktail. HPLC analyses were performed using a Beckman
System Gold system with a 125 solvent delivery module and
a 166 UV detector.
Rea gen ts. All chemicals used in the synthetic procedures
were purchased from Aldrich or Fischer Scientific and were
used as received unless otherwise noted. R-Ketoglutarate,
â-mercaptoethanol, NADP , Sephadex G50, buffer salts, and
other reagents used in the enzymological studies were pur-
chased from Sigma.
2
+
z): [M - H] calcd for C
7
H10NO
2
Anal. calcd for C
7
H
11NO
2
C, 59.26; H, 7.97; N, 9.77.
Syn th esis of cis- a n d tr a n s-2-Am in ocycloh ex-3-en e-1-
ca r boxylic Acid s (5 a n d 6). A mixture of ethyl cis- and trans-
6-carbomethoxy-2-cyclohexen-1-yl carbamates (9 and 10, in a
4:1 ratio) was prepared from the Diels-Alder reaction of ethyl
trans-1,3-butadiene-1-carbamate and methyl acrylate and
separated by repeated silica gel column chromatography (4:1
hexanes-ethyl acetate; using Silica gel 60 particle size 0.015-
0.040 mm from EM Science). To a 2 M solution of the cis
carbamate 9 (1.125 g, 4.95 mmol) in chloroform, sealed in a
reaction vessel under nitrogen, was added 2.5 equiv of iodo-
trimethylsilane (1.80 mL, 12.4 mmol) via dry syringe. The
reaction mixture was then heated at about 55 °C for 12 h.
When the reaction was completed (as determined by TLC), 4
equiv of HCl-saturated methanol was added (to quench the
reaction so that the amine was protonated as formed). After
volatile components were removed under reduced pressure, the
cis amino acid 5 was purified by a Dowex 50W-X8 cation
exchange column (by washing the column with water and then
eluting with 1 M pyridine) and then recrystallized from
ethanol/water (2:1), yielding 433 mg (62%). In the same
manner, the trans carbamate 10 (280 mg, 1.23 mmol) was
allowed to react with iodotrimethylsilane (440 µL, 3.08 mmol)
(55 °C for 12 h) and purified to afford the trans amino acid 6
1
9
+
2
0
Syn t h esis of tr a n s-3-Am in ocycloh ex-4-en e-1-ca r box-
ylic a cid (4). 7-Oxabicyclo[3,2,1]-oct-2-en-6-one (7) was syn-
thesized from cyclohex-3-ene-1-carboxylic acid by the procedure
1
7
3
of Marshall et al. and was allowed to react with NaN to
prepare (d,l)-trans-3-azidocyclohex-4-ene-1-carboxylic acid (8).16
A mixture of the resulting azide 8 (447 mg, 2.67 mmol) and
triphenylphosphine (770 mg, 2.94 mmol) in THF (15 mL) with
1
8
(99 mg, 57%). Compound 5: 1H NMR (D
m, CH and CH ), 2.60 (1H, m, CHCO ), 3.86 (1H, m, CHN),
5.60 (1H, m, J 10.2 Hz, dC(3)H), 6.0 (1H, m, J 10.2 Hz, d
C(4)H). 13C NMR (D
O): δ 21.8 (C5), 24.1 (C6), 42.2 (C1), 46.8
(C2), 121.8 (C3), 135.0 (C4), 181.0 (CO ). HRMS (EI) (m/z): [M
, 142.0868; found, 142.0869. Anal.
O: C, 58.08; H, 7.94; N, 9.68.
1
equiv of water (48 µL) was stirred at room temperature
2
O): δ 1.70-2.08 (4H,
for 97 h and then concentrated in vacuo. The residue was
partitioned between methylene chloride (15 mL) and 0.5 N HCl
2
2
2
(15 mL) in an ice bath. The aqueous phase was further washed
2
with methylene chloride (15 mL × 3). After the water was
2
+
removed, the resulting residue was chromatographed on a
+ H] calcd for C
calcd for C
7 2
H12NO
+
Dowex 50W (H ) cation exchange column by washing the
column with water and then eluting with 1 M pyridine. The
desired trans isomer 4 was recrystallized from ethanol/water
7
H
11NO
2
‚0.2H
2
Found: C, 58.04; H, 8.01; N, 9.65. X-ray crystal structure
analysis confirmed the structure, as shown in Figure 1: a
1
(2:1) to give 4 (200 mg, 53%). H NMR (D
2
O): δ 1.90-2.20
2
colorless square plate crystal of C
7
H
11NO
2
‚C
2 5
H OH having
(
4H, m, CH
2
and CH
2
), 2.50 (1H, m, CHCO
), 3.79 (1H, m,
dimensions of 0.22 mm × 0.22 mm × 0.09 mm with Z ) 4 and
3
1
CHN), 5.58 (1H, br d, J 9.9 Hz, dC(4)H), 6.05 (1H, m, 9.9 Hz,
the calculated density ) 1.24 g/cm . Compound 6: H NMR

4
538 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20
Choi and Silverman
(
D
2
O): δ 1.55-2.05 (4H, m, CH
CHCO ), 3.92 (1H, m, J 2.20 Hz, CHN), 5.50 (1H, dd, J 2.20,
0.3 Hz, dC(3)H), 5.97 (1H, m, J 10.3 Hz, dC(4)H). HRMS
2
and CH
2
), 2,31 (1H, m,
intervals over 60 min, 10 µL aliquots were removed and
2
assayed for enzyme activity.
1
(
1
Ir r ever sib le In a ct iva t ion of GABA-AT. GABA-AT (15
µL, 2.12 mg/mL) was incubated at 25 °C in 150 µL total volume
of a solution containing 5 mM 5, 5 mM R-ketoglutarate, 5 mM
â-mercaptoethanol, and 0.1 M potassium pyrophosphate at pH
8.5. Aliquots (10 µL) were removed at 0.5 and 17.5 h and
assayed for activity. A portion of the incubation mixture (100
µL) was applied to a Penefsky spin column30 prepared with
Sephadex G50 in a Bio-Spin disposable chromatography
column (BioRad) according to the published procedure. Small
molecules were removed by spinning for 2 min on an IEC
Clinical centrifuge. A 10 µL aliquot was removed and assayed
for enzyme activity. An identical sample, which contained no
inactivator, served as the control. Experiments were run in
duplicate.
+
EI) (m/z): [M + H] calcd for C
42.0862. Anal. calcd for C H11NO : C, 59.56; H, 7.85; N, 9.92.
7
H
12NO
2
, 142.0868; found,
7 2
Found: C, 59.17; H, 7.70; N, 9.87.
Syn th esis of th e Exp ected P r od u ct of En a m in e Rea c-
tion of 5 (11). PLP (50 mg, 0.2 mmol) and cyclohexanone (23
µL, 0.22 mmol) were dissolved in 8 mL of 0.5 M KOH, and the
solution was stirred for 24 h in the dark at room temperature.
The reaction mixture was adjusted to pH 7.0 with 70%
perchloric acid, and the solid KClO
4
was removed after
centrifugation. The remaining salt was washed twice with cold
water. The combined supernatants were concentrated and
applied to a Dowex 50W-X8 cation exchange column. Aliquots
of each fraction were spotted on silica gel thin-layer plates,
developed in 1-butanol-acetic acid-water (6:2:2 by volume), and
characterized with Gibbs reagent. Further purification was
carried out with a C18 reverse-phase HPLC column (with the
2
gradient system of 0.1% trifluoroacetic acid (TFA) in H O and
MeOH), and fractions containing the product were collected,
concentrated, and lyophilized to give the product 11 (26 mg,
3
retention time (T
the incubation mixture of 5 with GABA-AT occur. LC/MS
analysis using electrospray ionization showed the protonated
3
In a ctiva tion of [ H]P LP -Recon stitu ted GABA-AT by
2
6
5
a n d HP LC An a lysis. GABA-AT, which had been reconsti-
3
31
tuted with [ H]PLP as previously described, was incubated
at 25 °C and protected from light in 100 mM potassium
phosphate containing 112 mM 5, 5 mM R-ketoglutarate, and
5
mM â-mercaptoethanol at pH 7.4. A control was run with
6%). This PLP adduct from the aldol condensation showed a
the same concentrations of each reagent, excluding inactivator,
and a second control was run with 40 mM GABA containing
no inactivator or R-ketoglutarate. The first control should
release the cofactor as PLP, while the second should release
it as PMP. When the enzyme in the inactivator solution was
less than 7% active, excess small molecules were removed by
running the solutions over Sephadex G-50 using the Penefsky
R
) 35 min) where multiple products from
+
molecular ion peak of the product 11: [M + H] calcd for
C H18NO P, 327.27; found, 328.0.
14 6
Molecu la r Mod elin g. All procedures were performed using
SYBYL molecular modeling software version 6.5 or 6.6 (Tripos,
Inc., Saint Louis, MO) operating under IRIX 6.5. The coordi-
nates from the X-ray crystal structure of GABA-AT were used
for modeling studies. The docking experiment was performed
30
spin method. The pH of each solution was adjusted to 11-
1
2 using 1 M KOH. These were incubated at room temperature
2
2
for 1 h and added to enough TFA to quench the base and make
a 10% v/v TFA solution. After they stood at room temperature
for 10 min, the denatured enzyme solutions were placed into
prerinsed Centricon 10 microconcentrators and were centri-
fuged for 20 min at 5000 rpm in a DuPont Sorvall RC5B Plus
Centrifuge, using an SA 600 rotor to achieve complete separa-
tion of the protein and the effluents. Each Centricon was rinsed
with 1 mL of the buffer (containing 100 mM potassium
phosphate and 5 mM â-mercaptoethanol at pH 7.4), vortexed,
and then contrifuged for an additional 20 min. This was
repeated three times, and the rinses were added to the
supernatants and then freeze-dried. Analysis of the incubation
mixtures was carried out by dissolving the resulting solid in
2
7
using the FlexX program interfaced with SYBYL. Water
molecules were removed as they were found not to be located
within the active site region, and hydrogen atoms on the
backbone and side chains were added based on the standard
average bond angles and lengths. Dictionary charges from the
AMBER force field were added. The active site for docking was
defined as all amino acids within 6.5 Å proximity of the
cofactor PLP. The investigated ligands were created as exter-
nal aldimines starting from geometrically optimized frag-
ments. The potential energies were minimized to obtain a
comparable starting point for each compound. The Powell
conjugate gradient minimizer within the MAXIMIN procedure
was used with the parameters of the MMFF94 force field.
For the MD simulation, the docked complex of GABA-AT
and S-vinyl-GABA was used as the starting structure, and the
MMFF94 force field was used. The MD simulation was
performed by raising the temperature to 500 K over a period
of 5000 fs and cooling to 300 K over a period of 5000 fs with a
step size of 1 fs. Trajectories were recorded every 100 fs of
simulation and were analyzed.
1
00 µL of water, and then injecting the samples onto an Alltech
Econosil C18 reverse-phase HPLC column (4.6 mm × 250 mm,
0 µm). HPLC analysis was done using a mobile phase of H
1
2
O
containing 0.1% TFA flowing at 0.5 mL/min for 15 min. The
flow rate was increased to 1.0 mL/min over the next 5 min,
and then, a solvent gradient to 100% methanol with 0.1% TFA
was run over the next 20 min. The flow rate was further
increased to 2.0 mL/min over 5 min and maintained for 15
min, and then, a 10 min solvent gradient to 100% water with
0.1% TFA was run, followed by changing the flow rate to 0.5
mL/min over 5 min. Under these conditions, PLP elutes at 17
min and PMP at 8 min. Fractions were collected every minute,
and the elution of radioactivity was followed by liquid scintil-
lation counting. The incubation mixture with fractions, which
En zym es a n d Assa ys. GABA-AT was isolated from pig
2
8
brains as described previously. The purified enzyme was
approximately 95% homogeneous by sodium dodecyl sulfate
polyacrylamide gel electrophoresis (SDS-PAGE) (Coomassie
blue). Succinic semialdehyde dehydrogenase (SSDH) was
isolated from GABAse (Sigma), and GABA-AT activity was
monitored spectrophotometrically via a coupled assay as
3
during previous experiments with [ H]PLP-reconstituted GABA-
AT with 3 or 5 contained substantial amounts of radioactivity,
2
9
+
previously described.
was analyzed by LC/MS using electrospray (ES ) ionization,
Tim e-Dep en d en t In h ibition of GABA-AT. GABA-AT (15
µL, 2.12 mg/mL) was incubated at 25 °C in 150 µL total volume
of a solution containing various concentrations of 5, 1 mM
R-ketoglutarate, 5 mM â-mercaptoethanol, and 50 mM potas-
sium pyrophosphate at pH 8.5. An identical sample, which
contained no inhibitor, served as the control. At time intervals
over 60 min, 10 µL aliquots were assayed for enzyme activity.
Su bstr a te P r otection fr om In a ctiva tion of GABA-AT.
GABA-AT (15 µL, 2.12 mg/mL) was incubated at 25 °C in 150
µL total volume of a solution containing 10 mM 5, 5 mM
R-ketoglutarate, 5 mM â-mercaptoethanol, 0-10 mM GABA,
and 0.1 M potassium pyrophosphate at pH 8.5. A sample in
which the inhibitor was omitted served as the control. At time
which confirmed the occurrence of the decarboxylated modified
+
cofactor: [M + H] calcd for C14
for inactivator 5.
6
H18NO P, 327.27; found, 328.1
Com p etitive In h ibition of GABA-AT. The activity of
GABA-AT upon introduction of varying concentrations (0, 50,
100, 250, 375, 500, 750, and 1000 µM) of 4 or 6 was determined
spectrophotometrically by measuring the formation of NADH
upon the conversion of SSDH at 25 °C and at several GABA
concentrations (4, 6, 10, 15, and 25 mM) in 100 mM potassium
pyrophosphate buffer at pH 8.5, containing 5 mM â-mercap-
3
2
toethanol. Kinetic parameters were derived from Dixon and
Cornish-Bowden33 plots with at least five different concentra-
tions of 4 or 6 for every concentration of GABA.

γ-Aminobutyric Acid Aminotransferase
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 20 4539
Ack n ow led gm en t. We thank the National Insti-
tutes of Health (Grant NS15703) for financial support
of this research.
(16) Murahashi, S.-I.; Taniguchi, Y.; Imada, Y.; Tanigawa, Y. Pal-
ladium(0)-Catalyzed Azidation of Allyl Esters. Selective Syn-
thesis of Allyl Azides, Primary Allylamines, and Related Com-
pounds. J . Org. Chem. 1989, 54, 3292-3303.
(
(
(
17) Marshall, J . A.; Xie, S. Synthesis of a C22-34 Subunit of the
Refer en ces
Immunosuppressant FK-506. J . Org. Chem. 1995, 60, 7230-
7
237.
(
1) (a) Krnjevic, K. Chemical nature of synaptic transmission in
vertebrates. Physiol. Rev. 1974, 54, 418-540. (b) Faingold, C.
L.; Gehlbach, G.; Caspary, D. M. On the role of GABA as an
inhibitory neurotransmitter in inferior colliculus neurons: ion-
tophoretic studies. Brain Res. 1989, 500, 302-312. (c) Bohme,
I.; Luddens, H. The inhibitory neural circuitry as target of
antiepileptic drugs. Curr. Med. Chem. 2001, 8, 1257-1274.
2) Cooper, A. J . L. Glutamate-γ-aminobutyrate transaminase.
Methods Enzymol. 1985, 113, 80-82.
18) Nagarajan, S.; Ganem, B. Chemistry of Naturally Occurring
Polyamines. 11. Unsaturated Spermidine and Spermine Deriva-
tives. J . Org. Chem. 1987, 52, 5044-5046.
19) Overman, L. E.; Taylor, G. F.; Houk, K. N.; Domelsmith, L. N.
Diels-Alder Reactions between trans-1-N-Acylamino-1,3-dienes
and Methyl Acrylate. A Correlation between Diene Photoelectron
Ionization Potentials and Reactivity, Stereoselectivity, and
Regioselectivity. J . Am. Chem. Soc. 1978, 100, 3182-3189.
20) (a) J ung, M. E.; Lyster, M. A. Conversion of Alkyl Carbamates
into Amines via Treatment with Trimethylsilyl Iodide. J . Chem.
Soc. Chem. Commun. 1978, 315-316. (b) J ung, M. E.; Lyster,
M. A. Quantitative Dealkylation of Alkyl Esters via Treatment
with Trimethylsilyl Iodide. A New Method for Ester Hydrolysis.
J . Am. Chem. Soc. 1977, 99, 968-969.
(
(
(
(
3) Gale, K. GABA in epilepsy: the pharmacologic basis. Epilepsia
1
989, 30 (Suppl. 3), S1-S11.
4) (a) Hornykiewicz, O.; Lloyd, K. G.; Davidson, L. The GABA
system, function of the basal ganglia and Parkinson’s disease.
In GABA in Nervous System Function; Roberts, E., Chase, T.
N., Tower, D. B., Eds.; Raven Press: New York, 1976; pp 479-
4
85. (b) Kleppner, S. R.; Tobin, A. J . GABA signaling: Thera-
(21) Kitz, R.; Wilson, I. B. Esters of Methanesulfonic Acid as
Irreversible Inhibitors of Acetylcholinesterase. J . Biol. Chem.
1962, 237, 3245-3249.
peutic targets for epilepsy, Parkinson’s disease and Huntington’s
disease. Emerging Ther. Targets 2001, 5, 219-239.
(
5) Perry, T. L.; Hansen, S.; Kloster, M. Huntington’s chorea:
deficiency of γ-aminobutyric acid in brain. N. Engl. J . Med. 1973,
(
22) Storici, P.; Capitani, G.; De Biase, D.; Moser, M.; J ohn, R. A.;
J ansonius, J . N.; Schirmer, T. Crystal Structure of GABA-
aminotransferase, a Target for Antiepileptic Drug Therapy.
Biochemistry 1999, 38, 8628-8634.
23) (a) De Biase, D.; Barra, D.; Bossa, F.; Pucci, P.; J ohn, R. A.
Chemistry of the Inactivation of 4-Aminobutyrate Aminotrans-
ferase by the Antiepileptic Drug Vigabatrin. J . Biol. Chem. 1991,
2
88, 337-342.
(
6) (a) Aoyagi, T.; Wada, T.; Nagai, M.; Kojima, F.; Harada, S.;
Takeuchi, T.; Takahashi, H.; Hirokawa, K.; Tsumita, T. In-
creased γ-aminobutyrate aminotransferase activity in brain of
patients with Alzheimer’s disease. Chem. Pharm. Bull. 1990,
(
3
8, 1748-1749. (b) Weiner, M. F. Treatment of behavioral/
2
66, 20056-20061. (b) De Biase, D.; Bolton, J . B.; Barra, D.;
psychological symptoms in Alzheimer’s disease. Exp. Rev. Neu-
rother. 2001, 1, 70-80.
Bossa, F.; J ohn, R. A. Stoichiometry and stability of the adduct
formed between human 4-aminobutyrate aminotransferase and
(
7) (a) Kushner, S. A.; Dewey, S. L.; Kornetsky, C. The irreversible
γ-aminobutyric acid (GABA) transaminase inhibitor γ-vinyl-
GABA blocks cocaine self-administration in rats. J . Pharmacol.
Exp. Ther. 1999, 290, 797-802. (b) Dewey, S. L.; Morgan, A. E.;
Ashby, C. R., J r.; Horan, B.; Kushner, S. A.; Logan, J .; Volkow,
N. D.; Fowler, J . S.; Gardner, E. L.; Brodie, J . D. A novel strategy
for the treatment of cocaine addiction. Synapse 1998, 30, 119-
4
1
-aminohex-5-enoate: sequence of a labelled peptide. Biochimie
989, 71, 491-495.
(
24) Toney, M. D.; Pascarella, S.; De Biase, D. Active Site Model for
γ-Aminobutyrate Aminotransferase Explains Substrate Specific-
ity and Inhibitor Reactivities. Protein Sci. 1995, 4, 2366-2374.
(25) Dunathan, H. C. Conformation and Reaction Specificity in
Pyridoxal Phosphate Enzymes. Proc. Natl. Acad. Sci. U.S.A.
1966, 55, 712-716.
(26) Ross, J . H. 2,6-Dichloroquinone 4-Chloroimide as a Reagent for
Amines and Aromatic Hydrocarbons on Thin-layer Chromato-
grams. Anal. Chem. 1968, 40, 2138-2143.
(27) Rarey, M.; Kramer, B.; Lengauer, T.; Klebe, G. A Fast Flexible
Docking Method Using an Incremental Construction Algorithm.
J . Mol. Biol. 1996, 261, 470-489.
1
29.
(
8) Lippert, B.; Metcalf, B. W.; J ung, M. J .; Casara, P. 4-Amino-
hex-5-enoic Acid, a Selective Catalytic Inhibitor of 4-Aminobu-
tyric-Acid Aminotransferase in Mammalian Brain. Eur. J .
Biochem. 1977, 74, 441-445.
(
9) Nanavati, S. M.; Silverman, R. B. Design of Potential Anticon-
vulsant Agents: Mechanistic Classification of GABA Ami-
notransferase Inactivators. J . Med. Chem. 1989, 32, 2413-2421.
(
10) (a) Silverman, R. B. Mechanism-Based Enzyme Inactivation:
Chemistry and Enzymology; CRC Press: Boca Raton, 1988; Vols.
I and II. (b) Silverman, R. B. Mechanism-Based Enzyme Inac-
tivators. Methods Enzymol. 1995, 249, 240-283.
(
(
(
28) J ohnson, T. R.; Silverman, R. B. Syntheses of (Z)- and (E)-4-
Amino-2-(trifluoromethyl)-2-butenoic Acid and Their Inactiva-
tion of γ-Aminobutyric Acid Aminotransferase. Bioorg. Med.
Chem. 1999, 7, 1625-1636.
(
(
(
11) Davies, J . A. Mechanism of action of antiepileptic drugs. Seizure
29) Silverman, R. B.; Bichler, K. A.; Leon, A. J . Mechanisms of
inactivation of γ-aminobutyric acid aminotransferase by 4-amino-
1
995, 4, 267-271.
12) Brennan, M. Treating substance abuse. Chem. Eng. News 1999,
August 23, 8.
5
1
-fluoro-5-hexenoic acid. J . Am. Chem. Soc. 1996, 118, 1241-
252.
13) Nanavati, S. M.; Silverman, R. B. Mechanism of Inactivation of
γ-Aminobutyric Acid Aminotransferase by the Antiepilepsy Drug
γ-Vinyl GABA (Vigabatrin). J . Am. Chem. Soc. 1991, 113, 9341-
30) Penefsky, H. S. A centrifuged-column procedure for the mea-
surement of ligand binding by beef heart F1. Methods Enzymol.
1
979, 56, 527.
9
349.
(
14) Choi, S.; Storici, P.; Schirmer, T.; Silverman, R. B. Design of a
Conformationally Restricted Analogue of the Antiepilepsy Drug
Vigabatrin that Directs Its Mechanism of Inactivation of γ-Ami-
nobutyric Acid Aminotransferase. J . Am. Chem. Soc. 2002, 124,
(31) Fu, M.; Silverman, R. B. Isolation and characterization of the
product of inactivation of γ-aminobutyric acid aminotransferase
by gabaculine. Bioorg. Med. Chem. 1999, 7, 1581-1590.
(32) Dixon, M. The Determination of Enzyme Inhibitor Constants.
Biochem. J . 1953, 55, 170-171.
1
620-1624.
(
15) Qiu, J .; Pingsterhaus, J . M.; Silverman, R. B. Inhibition and
Substrate Activity of Conformationally Rigid Vigabatrin Ana-
logues with γ-Aminobutyric Acid Aminotransferase. J . Med.
Chem. 1999, 42, 4725-4728.
(33) Cornish-Bowden, A. Fundamentals of Enzyme Kinetics; Portland
Press: London, 1995; pp 102-108.
J M020134I