Different Binding Modes of Small and Large Binders of GAT1

Article abstract of DOI:10.1002/cmdc.201500534

Well-known inhibitors of the γ-aminobutyric acid (GABA) transporter GAT1 share a common scaffold of a small cyclic amino acid linked by an alkyl chain to a moiety with two aromatic rings. Tiagabine, the only FDA-approved GAT1 inhibitor, is a typical example. Some small amino acids such as (R)-nipecotic acid are medium-to-strong binders of GAT1, but similar compounds, such as proline, are very weak binders. When substituted with 4,4-diphenylbut-3-en-1-yl (DPB) or 4,4-bis(3-methylthiophen-2-yl)but-3-en-1-yl (BTB) groups, the resulting compounds have similar pK_i and pIC₅₀ values, even though the pure amino acids have very different values. To investigate if small amino acids and their substituted counterparts share a similar binding mode, we synthesized butyl-, DPB-, and BTB-substituted derivatives of small amino acids. Supported by the results of docking studies, we propose different binding modes not only for unsubstituted und substituted, but also for strong- and weak-binding amino acids. These data lead to the conclusion that following a fragment-based approach, not pure but N-butyl-substituted amino acids should be used as starting points, giving a better estimate of the activity when a BTB or DPB substituent is added.

Full text of DOI:10.1002/cmdc.201500534

DOI: 10.1002/cmdc.201500534
Full Papers
Different Binding Modes of Small and Large Binders of
GAT1
Thomas Wein,[a] Marilena Petrera,[b] Lars Allmendinger,[a] Georg Hçfner,[a] Jçrg Pabel,[a] and
Klaus T. Wanner*[a]
Well-known inhibitors of the g-aminobutyric acid (GABA) trans-
porter GAT1 share a common scaffold of a small cyclic amino
acid linked by an alkyl chain to a moiety with two aromatic
rings. Tiagabine, the only FDA-approved GAT1 inhibitor, is
a typical example. Some small amino acids such as (R)-nipecot-
ic acid are medium-to-strong binders of GAT1, but similar com-
pounds, such as proline, are very weak binders. When substi-
tuted with 4,4-diphenylbut-3-en-1-yl (DPB) or 4,4-bis(3-methyl-
thiophen-2-yl)but-3-en-1-yl (BTB) groups, the resulting com-
pounds have similar pKi and pIC50 values, even though the
pure amino acids have very different values. To investigate if
small amino acids and their substituted counterparts share
a similar binding mode, we synthesized butyl-, DPB-, and BTB-
substituted derivatives of small amino acids. Supported by the
results of docking studies, we propose different binding
modes not only for unsubstituted und substituted, but also for
strong- and weak-binding amino acids. These data lead to the
conclusion that following a fragment-based approach, not
pure but N-butyl-substituted amino acids should be used as
starting points, giving a better estimate of the activity when
a BTB or DPB substituent is added.
Introduction
g-Aminobutyric acid (GABA) is the most important inhibitory
neurotransmitter in the mammalian brain. It plays a major role
in the regulation of neuronal activity throughout the central
nervous system (CNS). The inhibitory action of GABA in the
synaptic cleft is terminated by the reuptake of GABA in neu-
rons and astroglial cells via GABA transport proteins. Four sub-
types of GABA transporters[1] (GATs) are currently known: only
GAT1 and GAT3 are present in the CNS in significant amounts,
whereas GAT2 and BGT1 are located primarily in the liver and
kidneys.[2] When cloned from mice these GABA transporters are
generally termed mGAT1, mGAT2, mGAT3, and mGAT4.[3]
Herein we use the nomenclature suggested by the Human
Genome Organization (HUGO) in which the corresponding
human GABA transport proteins are denoted as GAT1 (encod-
ing gene: SLC6A1), BGT1 (SLC6A12), GAT2 (SLC6A13), and GAT3
(SLC6A11), respectively.[4] The GATs are members of the SLC6
gene family of sodium- and chloride-dependent neurotransmit-
ter transporters, including transporters for glycine, dopamine,
norepinephrine, and serotonin,[5] which are of considerable
medical interest. These transporters ensure that the concentra-
tions of the neurotransmitters in the synapse are low, so that
post-synaptic receptors can detect neurotransmitter molecules
upon exocytotic release from pre-synaptic neurons. Inhibition
of neurotransmitter uptake from the synaptic cleft increases
the level of neurotransmitter in the synaptic cleft and enhan-
ces synaptic transmission. Hence, neurotransmitter transporters
provide important targets for therapeutic intervention. Inhibi-
tion of GABA uptake has been recognized as a therapeutic
strategy for the treatment of epileptic disorders,[6] for example.
Since the 1970s and supported in the 1990s by the first clon-
ing of human[7] and rat[8] brain GABA transporters, a huge me-
dicinal chemistry effort aimed at the development of GAT in-
hibitors has begun, resulting in tiagabine {(R)-1: (3R)-1-[4,4-
bis(3-methylthiophen-2-yl)but-3-en-1-yl]piperidine-3-carboxylic
acid}, as the only US Food and Drug Administration (FDA)-ap-
proved GAT inhibitor (selective for GAT1) for the therapy of ep-
ilepsy.[9] Several other GAT1 inhibitors such as NO711[10] (2),
SKF89976A[11] [(R)-3], and CI966[12] (4), sharing common scaf-
folds of a small cyclic amino acid linked by an alkyl or hetero-
alkyl chain to a lipophilic substructure with two aromatic rings
(Figure 1), have been identified without reaching approved
drug status. In addition, tiagabine [(R)-1] is being investigated
for other possible indications such as the treatment of general
anxiety[13] and depression.[14]
[a] Dr. T. Wein, Dr. L. Allmendinger, Dr. G. Hçfner, Dr. J. Pabel,
Prof. Dr. K. T. Wanner
As early as the 1970s (R,S)-nipecotic acid [(R,S)-5a] and guva-
cine (6a) were found to be inhibitors of GABA uptake in neu-
ronal and astroglial cell cultures,[6,15] where (R)-nipecotic acid
[(R)-5a] was known to be a more potent inhibitor than the
S isomer [(S)-5a] for GABA uptake in rat brain slices.[15c] The in-
hibitory effect on GABA transport of 2-[(3S)-pyrrolidin-3-yl]ace-
tic acid[16] [(S)-7a] (homo-b-proline) has also been known for
a long time. Other cyclic amino acids very similar to nipecotic
Department for Pharmacy—Center for Drug Research
Ludwig-Maximilians-Universität München
Butenandtstr. 7-13, 81377 Munich (Germany)
E-mail: klaus.wanner@cup.uni-muenchen.de
[b] Dr. M. Petrera
Galliera Hospital, Mura delle Cappuccine 14, 16128 Genova (Italy)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201500534.
ChemMedChem 2016, 11, 509 – 518
509
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 1. pKi and pIC50 values for GAT1 binding and inhibition and dock-
ing results.[a]
[30]
Entry
Compd
pKi[29]
pIC50
AutoDock4 pKi[b]
1
2
3
4
(R)-5a
(R)-5b
(R)-5c
(R)-5d
4.50Æ0.05
3.53Æ0.06
6.33Æ0.04
7.43Æ0.11
5.19Æ0.03
3.72Æ0.09
5.49Æ0.09
6.88Æ0.12
4.43
4.99
6.68
7.02
5
6
7
8
(S)-5a
(S)-5b
(S)-5c
(S)-5d
2.99Æ0.05
3.05Æ0.09
5.64Æ0.11
6.94Æ0.13
4.24Æ0.05
3.12Æ0.15
4.97Æ0.10
6.20Æ0.13
4.34
5.08
6.86
6.78
Figure 1. Structures of known GABA uptake inhibitors.
9
6a
6b
6c
6d
3.75Æ0.09
2.84Æ0.04
6.32Æ0.23
7.04Æ0.06
4.87Æ0.06
3.10Æ0.03
5.22Æ0.16
6.61Æ0.09
4.53
5.01
6.99
6.51
10
11
12
acid or guvacine, such as (R)- and (S)-proline [(R)-8a and (S)-
8a] or (R)- and (S)-2-(pyrrolidin-2-yl)acetic acid[17] [(R)-9a and
(S)-9a] (homoproline) show only weak or no inhibitory poten-
tial on GABA uptake. However, adding bulky lipophilic groups
to the nitrogen atom increases the inhibitory potential of
these small amino acids by several orders of magnitude,[16c,17,18]
nearly reaching the potency of tiagabine (Table 1, entry 33).
This is true for the uptake inhibition values (pIC50) as well as
for the binding activities (pKi) listed in Table 1.
13
14
15
16
(S)-7a
(S)-7b
(S)-7c
(S)-7d
4.25Æ0.02
2.86Æ0.03
6.61Æ0.13
6.98Æ0.09
5.57Æ0.03
3.21Æ0.03
5.99Æ0.11
6.69Æ0.01
4.65
4.83
5.88
6.62
17
18
19
20
(R)-8a
(R)-8b
(R)-8c
(R)-8d
0.27[c]
1.60Æ0.03
2.32Æ0.15
5.42Æ0.06
6.43Æ0.11
4.44
4.59
6.41
6.37
2.50Æ0.13
5.83Æ0.04
6.44Æ0.07
In 2005 the first X-ray structure of a bacterial leucine trans-
porter (LeuT) homologue to the SLC6 transporters was report-
ed,[19] giving insight into the three-dimensional (3D) structure
of this important class of transport proteins and serving as
a foundation for structure-based design through homology
modeling of the other SLC6 proteins. The X-ray structure of
LeuT revealed 12 transmembrane helices (TMs) of which the
protein core consists of segments TM1–TM5 and TM6–TM10,
which are related by a pseudo-twofold axis located in the
membrane plane. The substrate leucine could be clearly identi-
fied in an inner binding site called the S1 site almost in the
middle of the transmembrane protein, which includes two
bound sodium ions, of which one forms direct interactions
with the substrate. The S1 site is closed to the extracellular
side by a pair of aromatic residues—Tyr108 and Phe253—and
a salt bridge formed by Arg30 and Asp404. Another cavity is
located above these gating residues, called the S2 site, which
is currently the topic of debate as to whether it is an additional
high-affinity binding site for substrate molecules and whether
its occupation with an additional substrate is important for the
transport process.[20] Meanwhile, structures of LeuT with other
substrates and inhibitors have been reported in addition to an
open-to-out structure of LeuT.[21]
21
22
23
24
(S)-8a
(S)-8b
(S)-8c
(S)-8d
1.68Æ0.10
1.66Æ0.01
6.16Æ0.14
6.79Æ0.03
2.93Æ0.10
1.69Æ0.09
5.24Æ0.06
6.01Æ0.10
4.01
4.55
6.11
6.07
25
26
27
28
(R)-9a
(R)-9b
(R)-9c
(R)-9d
2.61Æ0.05
2.47Æ0.02
6.24Æ0.05
6.65Æ0.06
3.62Æ0.08
2.73Æ0.21
4.90Æ0.16
6.04Æ0.07
4.87
4.78
6.70
6.07
29
30
31
32
(S)-9a
(S)-9b
(S)-9c
(S)-9d
2.60Æ0.12
3.60Æ0.03
7.13Æ0.03
6.75Æ0.03
3.39Æ0.17
3.52Æ0.09
6.64Æ0.05
6.24Æ0.03
4.83
4.72
6.35
6.47
33
(R)-1[d]
7.43Æ0.11
6.88Æ0.12
7.02
[a] All biological testing results were processed and evaluated in tripli-
cate; Each pIC50 and pKi value is the mean ÆSEM of three independent
experiments. [b] AutoDock4 pKi was calculated from the AutoDock4 esti-
mated inhibition constant Ki, which is reported in the AutoDock4 output.
[c] Single experiment performed in triplicate; due to the high concentra-
tion required, the obtained value is of limited reliability. [d] Tiagabine.
Only two investigations have so far analyzed the binding of
small molecules in the S1 site of GAT1 using docking and sub-
sequent molecular dynamics (MD)[22] or flexible docking,[23] and
were able to predict the biological activities of a small set of
analyzed compounds. In both studies the nitrogen atom of ni-
pecotic acid was found to point to the intracellular gate of
GAT1. A modeling study using MD calculations to compare the
binding of nipecotic acid and tiagabine came to the same con-
clusion that nipecotic acid is bound to GAT1 in an orientation
by which the amino nitrogen atom faces mainly the intracellu-
lar side of the GAT1 protein.[24] In contrast, the N-substituted
nipecotic acid moiety in tiagabine is oriented with its nitrogen
atom facing the extracellular side, and the lipophilic group is
oriented into the vestibule (S2 site).
The data regarding binding affinities and inhibitory potential
toward GAT1 determined by uptake assays of pure amino
acids and those N-substituted with 4,4-diphenylbut-3-en-1-yl
(DPB) and 4,4-bis(3-methylthiophen-2-yl)but-3-en-1-yl (BTB)
groups clearly indicate that the gain in binding energy
brought by adding a large bulky lipophilic residue to small
amino acids does not show a linear correlation. In contrast, the
gain in inhibitory potential and binding affinity by adding the
same group, for example, BTB, to small amino acids raises the
measured values by one to nearly five log units. This large mis-
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
510
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
match is likely to indicate that some of the unsubstituted
amino acids, like (R)-nipecotic acid [(R)-5a] or guvacine (6a),
for which the gain in binding energy is relatively low, adopt
a favorable binding pose, yielding high binding affinities. Such
poses cannot be adopted if a bulky substituent is attached to
the nitrogen atom, an assumption that is underscored by the
few molecular modeling investigations available so far.
Results
Synthesis
To be able to compare the biological activity of the N-unsubsti-
tuted amino acids and their butyl derivatives with those of the
large inhibitors at GAT1 in uptake and binding assays, first the
respective compounds, i.e., butyl- [(R)-5b, (S)-5b, 6b, (S)-7b,
(S)-8b, (R)-8b, (S)-9b, (R)-9b], DPB- [6c and (S)-7c] and BTB-
substituted compounds [(S)-5d, (S)-7d and (R)-8d] had to be
synthesized. The synthesis of the new N-butyl derivatives was
accomplished by reductive amination, except in the case of
6b, for which a nucleophilic substitution strategy was applied,
as the double bond already present in the guvacine moiety
could suffer under the conditions of reductive amination. Thus,
(R)-5b, (S)-5b, (S)-8b, and (R)-8b were synthesized directly
from the respective commercially available amino acids
(Scheme 1) by treatment with butyraldehyde applying hydro-
To clarify the hypothesis that small and large binders of
GAT1 use partly different binding modes, we investigated a set
of small cyclic amino acids {(R)-nipecotic acid [(R)-5a], (S)-nipe-
cotic acid [(S)-5a], guvacine (6a), 2-[(3S)-pyrrolidin-3-yl]acetic
acid [(S)-7a], (R)- and (S)-proline [(R)-8a and (S)-8a], and (R)-
and (S)-2-(-pyrrolidin-2-yl)acetic acid [(R)-9a and (S)-9a]} and
their DPB-, BTB-, and butyl-substituted derivatives (Figure 2).
Figure 2. The selected amino acids and substituents.
The binding affinities of the pure amino acids range from pKi =
0.27 for (R)-8a up to pKi =4.50 for (R)-5a for GAT1 binding (~
4.2 log units; compare entry 17 with entry 1 in Table 1), where-
as the pKi values for the BTB-substituted compounds are all
within one log unit of pKi =6.44–7.43 (for (R)-8d and (R)-5d;
compare entry 20 with entry 4 in Table 1). Some preliminary
docking calculations indicated that amino acids substituted at
the nitrogen atom are able to occupy the main binding pocket
(S1) of GAT1, but that in the presence of a substituent of a cer-
tain size, such as an N-butyl residue, the amino group points
toward the extracellular side of GAT1, whereas the pure amino
acids investigated by Palló et al.[22] and Wein et al.,[23] according
to these authors, adopt orientations with the nitrogen atom
pointing to the intracellular side of the protein. Therefore, we
wanted to identify the binding and inhibition values of N-
butyl-substituted compounds in order to delineate possible
recommendations for the fragment-based drug design for new
GAT1 inhibitors.
Scheme 1. Synthesis of N-butyl amino acids via reductive amination. Re-
agents and conditions: a) H2 (5 bar), Pd/C, butyraldehyde, EtOH, RT.
gen (5 bar) in the presence of pallidum on charcoal. Pyrrolidin-
3-yl-acetic acid derived N-butylated compound (S)-7b, and pyr-
rolidin-2-yl-acetic acid derivatives (S)-9b and (R)-9b were also
obtained by reductive amination, albeit by starting from the
respective CBz-protected amino acids 10,[25] (S)-11, and (R)-11,
respectively, accomplishing deprotection and N-substitution
with the butyric moiety in “one pot” (Scheme 1).
The novel compounds bearing an olefinic moiety, either in
the amino acid (6b and 6c) or in the substituents to be at-
tached [(S)-5d·HCl, (S)-7c, and (S)-7d], were synthesized by
means of a conventional nucleophilic substitution reaction.
Thus, treatment of (S)-ethyl piperidine-3-carboxylate (12) or
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
511
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1,2,5,6-tetrahydropyridine-3-carboxylic acid methyl ester (14)
with the respective alkyl bromide [1-bromobutane (22b) or
4,4-diphenylbut-3-en-1-yl bromide (22c)] in the presence of
catalytic amounts of KI and K2CO3 as base gave the N-substi-
tuted amino acid methyl ester 13d (70%) and 15b (57%). The
following alkaline hydrolysis of these ester functions in 13d
and 15b, as well as in 15c[16c] provided amino acids 6b, 6c,
and (S)-5d·HCl (Scheme 2).
Compounds (S)-7c and (S)-7d were synthesized starting
from Boc-protected amino acid 16. At first 16 was transformed
into the respective ester 17 by treatment with methyl iodide.
Removal of the Boc group of 17 with TFA gave amino acid
methyl ester hydrotrifluoromethanesulfonate 18, which was
subjected to nucleophilic substitution with either 22c (4,4-di-
phenylbut-3-en-1-yl bromide) or 22d [(4,4-bis(3-methylthio-
phen-2-yl)but-3-en-1-]yl bromide, to afford N-substituted
amino acid esters 19c and 19d, respectively. Subsequent hy-
drolysis with 12m NaOH yielded the desired amino acids (S)-7c
and (S)-7d. Finally, (R)-8d was synthesized similarly by starting
from (R)-proline methyl ester hydrochloride (20), introducing
the [(4,4-bis(3-methylthiophen-2-yl)but-3-en-1-]yl moiety by
treating 20 with bromide 22d to give 21d which, after hydrol-
ysis, yielded the desired proline derivative (R)-8d (85%,
Scheme 2).
Biological data
In the following, the binding affinities (pKi values) are discussed
in detail, but the conclusions drawn are equally well supported
by the observed pIC50 values from the uptake assays, which
are listed in Table 1 as well, for the sake of completeness. The
N-unsubstituted amino acids (R)- and (S)-5a, 6a, (S)-7a, (R)-
and (S)-8a, and (R)- and (S)-9a show pKi values between 0.27
and 4.50 (Table 1, entries 1, 5, 9, 13, 17, 21, 25, 29). When
linked with a DPB residue, the resulting compounds (R)-5c, (S)-
5c, 6c, (S)-7c, (R)-8c, (S)-8c, (R)-9c, and (S)-9c show remarka-
bly high binding affinities to GAT1, with pKi values in the range
of 5.64–7.13 (Table 1, entries 3, 7, 11, 15, 19, 23, 27, 31), no
matter whether the corresponding unsubstituted amino acids
possess low or high binding affinities. Even more remarkable
are the BTB-substituted compounds (R)-5d, (S)-5d, 6d, (S)-7d,
(R)-8d, (S)-8d, (R)-9d, and (S)-9d, which exhibit binding affini-
ties with pKi values between 6.44 and 7.43 (Table 1, entries 4,
8, 12, 16, 20, 24, 28, 32). The only exception where the DPB de-
rivative shows a slightly higher pKi than the BTB substituted
compound (S)-9b is (S)-9c (Table 1, compare entry 31 and
entry 32). These experimental results indicate that the pure
amino acids with low binding affinities (pKi) and low inhibition
values (pIC50) gain much more binding potential from the DPB
or BTB substituents than amino acids with already high bind-
ing affinities or inhibition values.
All other compounds required for this study that carry either
a 4,4-diphenylbut-3-en-1-yl [(S)-5c, (R)-5c,[16c] (S)-8c, (R)-8c (S)-
9c, (R)-9c],[17] or [(4,4-bis(3-methylthiophen-2-yl)but-3-en-1-]yl
substituent [(R)-5d,[16c] 6d,[26] (S)-8d, (S)-9d, (R)-9d[17] were re-
ported previously, and have been synthesized following pub-
lished procedures.
The GAT1 binding values of the N-butyl derivatives [(R)- and
(S)-5b, 6b, (S)-7b, (R)- and (S)-8b, (R)-9b, and (S)-9b] are all in
a range between 1.66 and 3.60 (Table 1, entries 2, 6, 10, 14, 18,
22, 26, 30). The binding constants of N-unsubstituted amino
Scheme 2. Synthesis of N-substituted amino acids via nucleophilic substitution. Reagents and conditions: a) 1-bromobutane (22b), 4,4-diphenylbut-3-en-1-yl
bromide (22c) or [(4,4-bis(3-methylthiophen-2-yl)but-3-en-1-]yl bromide (22d), KI, K2CO3, acetone, RT; b) aq. NaOH/EtOH; c) HCl (2m); d) Ba(OH)2·8H2O, EtOH/
H2O 1:1 (v/v); e) CH3I, K2CO3, DMF, RT; f) TFA, CH2Cl2, RT. The synthesis of compounds (S)-5d·HCl[26] and its free base[27] as well as 6c[28] were reported previous-
ly, but no spectroscopic data were made available.
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
512
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
acids, which are already good binders like (R)-nipecotic acid
[(R)-5a, pKi =4.50, entry 1], guvacine (6a, pKi =3.75, entry 9),
and 2-[(3S)-pyrrolidin-3-yl]acetic acid [(S)-7a, pKi =4.25,
entry 13], are decreased by 0.97, 0.91, and 1.59 log units, re-
spectively, indicating that these amino acids might be forced
to adopt a new binding mode by the butyl group. On the
other hand, the low pKi value (0.27) of (R)-proline (R)-8a
(Table 1, entry 17) is dramatically increased by adding the butyl
group. The butyl-substituted proline (R)-8b exhibits a pKi value
of 2.50 (Table 1, entry 18), which is now in the same range as
the pKi values for the other butyl-substituted compounds.
Notably, the small good binders (R)-5a, 6a, and (S)-7a, (en-
tries 1, 9, 13) lose binding affinity when N-substituted, whereas
the binding values of the small moderate and weak binders
(S)-5a, (S)-8a, (R)-9a, and (S)-9a, (entries 5, 21, 25, 29) do not
change very much. These findings seem to indicate that (R)-ni-
pecotic acid [(R)-5a], guvacine (6a), and (S)-pyrrolidin-3-yl-
acetic acid [(S)-7a] bind GAT1 in a very favorable binding pose
which cannot be adopted if they are N-substituted with
a butyl or bulkier substituent such as DPB or BTB. In contrast,
the binding affinities of the small moderate or weak binders
are not significantly affected by N-substitution with the butyl
substituent except for (R)-8a, for which this results in a gain
and not a loss of binding affinity. When N-substituted with
a BTB or DPB residue, all of the parent amino acids show
a gain in binding affinity, the increase of which is, however,
often strikingly different.
GAT1 without significant overlap with protein atoms and large
energy penalties, we investigated the binding while keeping
the gate-keeping side chains of the residues Phe293, Tyr140,
Arg69, and Asp451 flexible in order to obtain meaningful re-
sults. The binding energies calculated with AutoDock4 were
converted into pKi values and are listed in Table 1.
For all unsubstituted amino acids [(R)-5a, (S)-5a, 6a, (S)-7a,
(S)-8a, (R)-8a, (S)-9a, and (R)-9a,] that fit into the S1 site, the
best-scoring poses represent an orientation in which the nitro-
gen atom in the ring is directed toward the intracellular side of
the protein, which is in accordance with previous results of
Palló et al.,[22] Wein et al.,[23] and Skovstrup et al.[24] The carbox-
ylic acid moiety of all parent amino acids interacts with the
sodium atom Na1. The nitrogen atom of (R)- and (S)-nipecotic
acid [(R)-5a and (S)-5a], guvacine (6a) and 2-[(3S)-pyrrolidin-3-
yl]acetic acid [(S)-7a] is part of two hydrogen bonds to Tyr60-
C=O and Tyr60-OH oxygen. (R)- and (S)-proline [(R)-8a and (S)-
8a], (2S)-pyrrolidin-2-yl-acetic acid and 2-(2R)-pyrrolidin-2-yl-
acetic acid [(S)-9a and (R)-9a] are unable to adopt a conforma-
tion in which their nitrogen atom is part of two hydrogen
bonds to Tyr60. In their case, their amino group can only inter-
act with the backbone oxygen atom of Phe293-C=O, a fact
that explains the lower pIC50 and pKi values observed in the
biological tests (Figure 3).
For docking runs with butyl-substituted compounds [(R)-5b,
(S)-5b, 6b, (S)-7b, (R)-8b, (S)-8b, (R)-9b, and (S)-9b] the two
gate-keeping residues Tyr140 and Phe294 were kept flexible
during docking. All of these derivatives adopt orientations with
the butyl group, and hence the nitrogen atom of the amino
acids, pointing toward the extracellular side, which is exempla-
rily shown in Figure 4 for N-butyl derivative (R)-5b in compari-
Docking
To get a more detailed view of potential binding modes that
explain the measured pKi values, we conducted a series of
docking calculations using homology models of GAT1 as the
target. The first available 3D protein structure of a homologous
transporter protein was the leucine transporter from Aquifex
aeolicus (LeuTAa, PDB ID: 2A65).[19] The structure includes a sub-
strate leucine molecule trapped in an active site almost in the
middle of the protein, called the S1 site. The protein conforma-
tion in 2A65 is called “closed” because the active site S1 is not
accessible from the extracellular or intracellular sides. Some
years after the aforementioned structure had been reported,
a LeuT structure with a so-called “open-to-out” conformation
was published (PDB ID: 3F3A)[21c] in which the S1 site was
open and accessible from the extracellular side. This open S1
site, the S2 site, and the channel to the extracellular water was
occupied with two tryptophan molecules, one detergent mole-
cule, and three water molecules. We calculated homology
models of GAT1 based on either the “closed” conformation
(PDB ID: 2A65) or the “open-to-out” conformation of LeuT
(PDB ID: 3F3A). Subsequent docking calculations revealed that
the binding energies from the docking calculations obtained
using the “open-to-out” GAT1 model did not correspond as
well with the experimental pKi values as the docking binding
energies using the “closed” GAT1 model. However, as all BTB-
or DPB-substituted amino acids [(R)-5c, (S)-5c, 6c, (S)-7c, (S)-
8c, (R)-8c, (S)-9c, (R)-9c, (R)-5d, (S)-5d, 6d, (S)-7d, (S)-8d, (R)-
8d, (S)-9d, and (R)-9d] did not fit into the “closed” S1 site of
Figure 3. Best docking poses in the “closed” S1 site: (R)-nipecotic acid (R)-5a
in green, (S)-proline (S)-8a in magenta, 2-[(2S)-pyrrolidin-2-yl]acetic acid (S)-
9a in blue.
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
513
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
and binders of GAT1. Their N-substituted derivatives with 4,4-
diphenylbut-3-en-1-yl (DPB) [(R)-5c, 6c, (S)-7c] or with 4,4-
bis(3-methylthiophen-2-yl)but-3-en-1-yl (BTB) [(R)-5d, 6d, (S)-
7d] residues show some of the highest inhibition of GAT1
known so far. However, the gain in affinity due to the addition
of the N-substituents, the DPB or the BTB residue, is only in
the range of 1.84–2.56 (for DPB) and 2.73–3.29 (for BTB) log
units. In stark contrast, the very poor inhibitors of GAT1, such
as (R)- and (S)-proline [(R)-8a and (S)-8a] or (R)- and (S)-2-(-pyr-
rolidin-2-yl)acetic acid[17] [(R)-9a and (S)-9a] gain between 3.64
to 5.55 (DPB) and 4.05 to 6.17 (BTB) log units if N-substituted
with DPB [(R)-8c and (S)-8c, (R)-9c and (S)-9c] or BTB [(R)-8d
and (S)-8d, (R)-9d and (S)-9d] groups. This marked difference
in gain in binding affinity, a low increase for the more potent
and a high increase for the less potent N-unsubstituted amino
acids, suggest that either the binding mode of the N-unsubsti-
tuted good binders or even of both the good and weak bind-
ers, do not reflect the orientation these amino acids adopt
when substituted with DPB and BTB residues. In other words,
at least the N-unsubstituted amino acids that exhibit high
binding affinities display a binding pose that provides a high
binding affinity, but is not related to the orientation of this
unit in the DPB- and BTB-substituted derivatives. Consequently,
the high binding affinities that these N-unsubstituted amino
acids possess cannot be transmitted into the DPB- and BTB-
substituted derivatives.
Figure 4. Results from docking into the “closed” GAT1 conformation with
the side chains of Tyr140 and Phe293 flexible (shown as stick models): (R)-ni-
pecotic acid (R)-5a in green and (R)-N-butylnipecotic acid (R)-5b in cyan.
The important hydrogen bonds between nipecotic acid and Tyr60-C=O and
Tyr60-OH are depicted with dotted black lines. The hydrogen bond between
N-butylnipecotic acid and Phe293-C=O is shown with a dotted blue line.
The high inhibitory power of the known small GAT1 inhibi-
tors, i.e., of (R)-nipecotic acid [(R)-5a], guvacine (6a), and 2-
[(3S)-pyrrolidin-3-yl] acetic acid [(S)-7a] seems to be based on
the fact that these compounds might be able to accommodate
a perfect fit into the closed site S1 in GAT1 where the nitrogen
atom is directed toward the intracellular side of the channel
forming hydrogen bonds with the carbonyl and side chain
oxygen atoms of Tyr60 (Figure 4). The synthesis of the N-butyl-
substituted derivatives of the investigated amino acids was
performed, as it was expected that as soon as a substituent
like an N-butyl group is present, the nitrogen atom of the
amino acids must be oriented toward the extracellular side of
the protein. This way all small amino acid inhibitors of mGAT1
should be forced into a binding pose with an orientation simi-
lar to that of the amino acid and linker part of the large amino
acid GAT1 inhibitors, the latter of which differ largely from the
former only by the additional presence of the aromatic resi-
dues. As a consequence thereof, the binding affinities shown
by the N-butyl-substituted amino acids should more reliably
reflect their potential to serve as starting points for the devel-
opment of large mGAT1 inhibitors with aromatic residues.
Indeed, this appears to be the case. The variance of the gain in
binding energy for the transition from the N-butyl- to N-diaryl-
butenyl-substituted amino acids acting as GAT1 binders is far
smaller, ranging from 2.59 to 4.50 log units for the DPB and
from 3.15 to 5.13 log units for the BTB residue than before for
the N-unsubstituted versus the DPB- (1.84–5.55 log units) and
BTB-substituted compounds (2.73–6.17 log units). This is also
plausible taking the results from the docking experiments into
account. The strong small amino acid binders (R)-nipecotic
acid [(R)-5a], guvacine (6a), and 2-[(3S)-pyrrolidin-3-yl]acetic
son with the N-unsubstituted compound (R)-5a. In the case of
(R)-5b, only one hydrogen bond between the NH hydrogen
and Phe294-C=O is possible in this orientation, whereas for (R)-
5a it had been two (to Tyr60).
The compounds with BTB or DPB substituents still do not fit
into this model or only with a high energy penalty, rendering
the results unreasonable. For the BTB- and DPB-substituted
compounds it is necessary to open the channel even more by
making Tyr140 and Phe293 and the salt bridge Arg69 and
Asp451 flexible during docking in order to get negative scor-
ing energies. With this setup, docking experiments are able to
differentiate between the weak binders like the pure amino
acids [(R)-5a, (S)-5a, 6a, (S)-7a, (R)-8a, (S)-8a, (R)-9a, and (S)-
9a] and N-butyl-substituted compounds [(R)-5b, (S)-5b, 6b,
(S)-7b, (R)-8b, (S)-8b, (R)-9b, and (S)-9b] on one side, and the
good binders like the DPB- [(R)-5c, (S)-5c, 6c, (S)-7c, (R)-8c,
(S)-8c, (R)-9c, (S)-9c] and BTB-substituted compounds [(R)-5d,
(S)-5d, 6d, (S)-7d, (R)-8d, (S)-8d, (R)-9d, (S)-9d] on the other.
In addition, it nicely reproduces the rank order of binding po-
tency for the BTB-substituted compounds [(R)-5d, (S)-5d, 6d,
(S)-7d, (R)-8d, (S)-8d, (R)-9d, (S)-9d]. Again, as with the N-
butyl-substituted compounds, for the DPB- and BTB-substitut-
ed compounds the only hydrogen bond for the NH hydrogen
that exists is toward Phe293-C=O.
Discussion
(R)-Nipecotic acid [(R)-5a], guvacine (6a), and 2-[(3S)-pyrroli-
din-3-yl]acetic acid [(S)-7a] are known to be good inhibitors
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
514
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
two sodium atoms (Na1 and Na2) located close to the S1 binding
site were copied into the hGAT1 structures. A chloride ion was
placed into the putative chloride binding site proposed by Zomot
et al.[34] and Forrest et al.[35] However, the chloride ion has no direct
contact to the active site, and its presence has no impact on the
docking calculations. The models with the lowest Modeller objec-
tive function were checked with PROCHECK[36] for structural consis-
tency and further used for docking calculations.
acid [(S)-7a] profit from favorable interactions with Tyr60 com-
prising two hydrogen bonds. Upon substitution of the amino
nitrogen with alkyl residues such as an N-butyl moiety or
larger groups and the resulting reorientation of the system,
this favorable interaction is lost, and only one hydrogen bond
can be formed. Although the docking model used is able to
predict almost the order of potency of the BTB-substituted
compounds and can distinguish the small versus large binders,
the subtle differences between the pure amino acids and the
N-butyl compounds is not reproduced. However, the above-
mentioned effect is clearly observed in the pKi values of the N-
butylated amino acids (R)-5b, 6b, and (S)-7b (Table 1, entries 2,
10, 14) which decrease between one and more than two
orders of magnitude. In contrast, unsubstituted amino acids
that have no or only low inhibitory potency toward GAT1 such
as (R)-5a, (S)-5a, (S)-8a, (R)-9a, and (S)-9a (Table 1, entries 1, 5,
17, 21, 25, 29), the inhibition (pKi) values are not changed very
much upon N-butyl substitution [(S)-5b, (S)-8b, (R)-9b; en-
tries 6, 18, 26] or are even increased [(R)-5b, (S)-9b], as in their
case the orientation of the amino group toward the intracellu-
lar side associated with the interaction with Tyr60 is not
strongly favored or is not favored at all.
Docking
The molecules for docking were exported from our in-house In-
stant JChem 5.4 database[37] as 2D MDL .mol files. Protonation at
pH 7.4 and the 2D-to-3D conversion was done using ChemAxon
Marvin 5.4.0.1[38] “cxcalc” and “molconvert”. The command line tool
“prepare_dpf42.py” was used for conversion into AutoDock4
.pdbqt input files. Appropriate protein input files were prepared
with AutoDock tools.[39] Docking grids were calculated with “auto-
grid4”, and docking was performed with AutoDock4 version
3
4.2.3.[40] The binding region was defined by a 18”18”18 Š box
centered between the two gate-keeping residues Tyr140 and
Phe294. Ten poses for each molecule were generated and scored
with the AutoDock4 scoring function.[40b]
Conclusions
Synthesis
Docking studies have underlined the hypothesis that the nitro-
gen atom of N-unsubstituted nipecotic acid [(R)-5a and (S)-5a]
is oriented toward the intracellular side of the binding pocket,
and, in contrast, the substituted nitrogen atom of N-butylnipe-
cotic acid or tiagabine must be oriented toward the extracellu-
lar side. In consequence, a fragment-based approach for the
construction of potent GAT1 inhibitors must take possible dif-
ferent binding modes of the amino acid fragments into consid-
eration, depending on the substitution of the amino nitrogen.
Small N-unsubstituted amino acids with already high binding
potential may not gain as much binding potential as those
with very low values, whereas small amino acids with low or
very low binding activity can be transformed into very good
binders. Considering the fact that N-butyl-substituted amino
acids, although only displaying a small N-substituent, always
adopt a binding pose with the amino nitrogen pointing
toward the extracellular side, which is thus analogous to that
of the large binders, this investigation suggests that fragments
already carrying an N-butyl substituent represent a much more
valid basis for the design of new potent inhibitors.
Anhydrous reactions were carried out in vacuum-dried glassware
under a nitrogen or argon atmosphere. CH3CN was freshly distilled
from CaH2 and EtOH from sodium under nitrogen with addition of
diethyl phthalate. Other common solvents were p.a. quality or
freshly distilled before use. The phosphate buffer (pH 7) was pre-
pared by dissolving NaH2PO4·H2O (77.32 g) and Na2HPO4·2H2O
(58.36 g) in H2O (1 L).
Purchased reagents were used without further purification. TLC
was carried out using plates purchased from Merck (silica gel
60 F254 on aluminum sheet). Column chromatography (CC) was car-
ried out using Merck silica gel 60 (mesh 0.040–0.063 mm) as sta-
tionary
phase.
Compounds
were
stained
with
5%
(NH4)6Mo7O24·4H2O, 0.2% Ce(SO4)2·4H2O, and 5% concd H2SO4. Op-
tical rotation: Polarimeter 241 MC, Polarimeter ADP440+ at l=
589 cmÀ1. Melting points: mp (uncorrected) were determined with
a Büchi 510 Melting Point apparatus. IR spectroscopy: FT-IR Spec-
trometer 1600 and Paragon 1000 (PerkinElmer), oils were measured
as film, solid samples as KBr pellets for measurements. Mass spec-
trometry: Mass spectrometer 5989 A with 59980 B particle beam
LC–MS interface (Hewlett Packard). High-resolution (HR) mass spec-
trometry was performed on either a Thermo Finnigan LTQ FT (ESI)
1
or a JMS GCmate II JEOL (EI). H and 13C NMR spectra were record-
ed with a JNMR-GX (JEOL, 400 or 500 MHz) using TMS or dioxane
as internal standard and integrated with the MestReNova (Version
5.2.5-4119, Mestrelab Research S.L., 2008).
Experimental Section
Homology modeling
Synthesis and experimental data of compounds (S)-5d·HCl, 6c, (S)-
7c, (S)-7d, (R)-8d, 13d, 19c, 19d, and 21d are provided in the
Supporting Information.
For homology modeling of a “closed” GAT1 conformation, the 3D
structure of PDB ID: 2A65[19] was taken from the RCSB Protein Data
Bank.[31] The only available 3D structure of an “open-to-out” confor-
mation (PDB ID: 3F3A[21c]) was chosen as a template for an “open”
GAT1 conformation. The human GAT1 sequence (SwissProt[32] ac-
cession number P30531) was aligned to the LeuT sequence using
the alignment from Skovstrup et al.[24] Modeller software version
9v8[33] was used to generate 30 structures of each conformer. The
General Procedure 1 (GP1): A mixture of the respective amino
acid (1 equiv) and butyraldehyde (1.2 mmol) in absolute EtOH
(10 mL) under argon atmosphere was hydrogenated at 5 bar in the
presence of 10% Pd/C (50 mg) for the time given. The suspension
was filtered, washed with EtOH, and the filtrate was evaporated
under reduced pressure.
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
515
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
General Procedure
2
(GP2): The respective ester derivative
1.73–1.79 (m, 2H, CH3CH2CH2), 2.47–2.53 (m, 2H, CHCH2), 2.93–2.97
(m, 2H, CH3(CH2)2CH2), 3.07–3.13 (m, 2H, CHCH2CH2), 3.72–3.77 (m,
2H, NCH2C), 6.84–6.86 ppm (m, 1H, CHC); 13C NMR (125 MHz,
CDCl3): d=13.7 (CH3), 20.3 (CH3CH2), 22.5 (CHCH2), 26.1
(CH3CH2CH2), 47.3 (CHCH2CH2), 49.8 (NCH2CC), 55.2 (CH3(CH2)2CH2),
129.6 (CCH), 131.0 (CCH), 170.0 ppm (C=O); IR (KBr): n˜ =3440,
2961, 2935, 2875, 1671, 1583, 1364, 734 cmÀ1; MS (CI, CH5+) m/z
(%): 184 (100) [M+H]+, 166 (7), 140 (9), 86 (2); HRMS-EI+ m/z [M]+
calcd for C10H17N1O2: 183.1259, found: 183.1221.
(1 equiv) in EtOH (0.5m) was cooled to 08C, and aqueous NaOH
(12m, 2 equiv) was added dropwise. The ice bath was removed,
and the resulting mixture was stirred at room temperature for the
time given. The reaction mixture was again cooled to 08C and
acidified (pHꢀ6) using HCl (0.25m). H2O was then added, followed
by extraction with CH2Cl2. The combined organic layers were dried
over MgSO4 and concentrated in vacuo.
(R)-1-Butylpiperidine-3-carboxylic acid [(R)-5b]: Prepared accord-
ing to GP1 starting from (R)-(À)-nipecotic acid (129 mg, 1.0 mmol)
(S)-2-(1-Butylpyrrolidin-3-yl)acetic acid [(S)-7b]: Prepared accord-
ing to GP1 starting from (S)-2-{1-[(benzyloxy)carbonyl]pyrrolidin-3-
yl}acetic acid (10, 144 mg, 0.55 mmol) and butyraldehyde (59.8 mg,
and butyraldehyde (86.5 mg, 1.2 mmol); reaction time: 4 h. Yield:
20
D
184 mg (99%) white solid; mp: 84–868C; ½aŠ +10.58 (c=0.21 in
1
MeOH); H NMR (500 MHz, NaOD in D2O): d=0.89 (t, J=7.4 Hz, 3H,
0.83 mmol) in absolute EtOH (4.3 mL); reaction time: 24 h. Yield:
20
D
CH3), 1.23–1.32 (m, 1H, NCH2CHCH2), 1.23–1.32 (m, 2H, CH3CH2),
1.43–1.54 (m, 2H, CH3CH2CH2), 1.43–1.54 (m, 1H, NCH2CHCH2CH2),
1.68–1.73 (m, 1H, NCH2CHCH2CH2), 1.87–1.96 (m, 1H, NCH2CHCH2),
1.87–1.96 (m, 1H, NCH2CH), 1.87–1.96 (m, 1H, CH(CH2)2CH2), 2.30–
2.36 (m, 1H, NCH2CH), 2.30–2.36 (m, 2H, CH3(CH2)2CH2), 2.89 (dbr,
J=11.9 Hz, 1H, CH(CH2)2CH2), 3.02–3.05 ppm (m, 1H, NCH2CH);
13C NMR (125 MHz, NaOD in D2O): d=13.9 (CH3), 21.0 (CH3CH2),
24.7 (CH(CH2)2CH2), 28.3 (CH3CH2CH2), 28.5 (NCH2CHCH2), 45.5
(NCH2CH), 53.5 (CH(CH2)2CH2), 56.5 (NCH2CH), 58.7 (CH3(CH2)2CH2),
184.1 ppm (C=O); IR (film): n˜ =3396, 2957, 2874, 1698, 1588, 1453,
1410, 1359, 1120 cmÀ1; MS (CI, CH5+) m/z (%): 186 (100) [M+H]+,
168 (6), 142 (8); HRMS-EI+ m/z [M]+ calcd for C10H19N1O2:
185.1416, found: 185.1418.
85 mg (84%) light-yellow solid; mp: 51–538C; ½aŠ À7.38 (c=0.45
in MeOH); 1H NMR (500 MHz, CD3OD): d=0.99 (t, J=7.4 Hz, 3H,
CH3CH2), 1.39–1.47 (m, 2H, CH3CH2), 1.66–1.72 (m, 2H, CH3CH2CH2),
1.77 (dq, J=13.3, 8.7 Hz, 1H, NCH2CHCH2), 2.25–2.31 (m, 1H,
NCH2CHCH2), 2.33 (dd, J=15.7, 7.1 Hz, 1H, CHCH2CO), 2.42 (dd, J=
15.7, 5.7 Hz, 1H, CHCH2CO), 2.66–2.75 (m, 1H, NCH2CH), 3.09–3.18
(m, 2H, N(CH2)2CH2), 3.09–3.18 (m, 1H, NCH2CH), 3.21–3.27 (m, 1H,
CHCH2CH2), 3.40–3.48 (m, 1H, CHCH2CH2), 3.40–3.48 ppm (m, 1H,
NCH2CH); 13C NMR (125 MHz, CD3OD): d=14.0 (CH3CH2), 21.0
(CH3CH2), 29.2 (CH3CH2CH2), 30.2 (NCH2CHCH2), 35.3 (NCH2CH), 42.3
(CHCH2CO), 54.9 (CHCH2CH2), 55.7 (N(CH2)2CH2), 60.1 (NCH2CH),
179.5 ppm (CO); IR (film): n˜ =3428, 2964, 2935, 2875, 1745, 1702,
1627, 1418, 1386, 1288, 1245 cmÀ1; MS (CI, CH5+) m/z (%): 186 (46)
[M+H]+, 129 (66), 111 (100); HRMS-ESI+ m/z [M+H]+ calcd for
C10H20N1O2: 186.1494, found: 186.1489.
(S)-1-Butylpiperidine-3-carboxylic acid [(S)-5b]: Prepared accord-
ing to GP1 starting from (S)-(+)-nipecotic acid (129 mg, 1.0 mmol)
and butyraldehyde (86.5 mg, 1.2 mmol); reaction time 4 h. Yield:
180 mg (97%) white solid; mp: 85–878C. Analytical data (1H NMR,
(S)-1-Butylpyrrolidine-2-carboxylic acid [(S)-8b]: Prepared accord-
ing to GP1 starting from l-proline [(S)-8a, 115 mg, 1.0 mmol] and
13C NMR, IR, MS) are in accordance with those of the R enantiomer
butyraldehyde (86.5 mg, 1.2 mmol); reaction time: 5 h. Yield:
20
D
20
D
(R)-5b; ½aŠ À11.48 (c=0.21 in MeOH).
168 mg (98%) light-yellow solid; mp: 71–738C; ½aŠ À56.68 (c=
0.47 in MeOH); 1H NMR (400 MHz, CD3OD): d=0.98 (t, J=7.4 Hz,
3H, CH3), 1.42 (sextet, J=7.3 Hz, 2H, CH3CH2), 1.62–1.77 (m, 2H,
CH3CH2CH2), 1.88–2.00 (m, 1H, CHCH2CH2), 2.04–2.16 (m, 1H,
CHCH2CH2), 2.04–2.16 (m, 1H, CHCH2), 2.37–2.47 (m, 1H, CHCH2),
3.05–3.13 (m, 1H, CH(CH2)2CH2), 3.05–3.13 (m, 1H, CH3(CH2)2CH2),
3.20–3.27 (m, 1H, CH3(CH2)2CH2), 3.73 (ddd, J=11.2, 7.3, 3.7 Hz, 1H,
CH(CH2)2CH2), 3.84 ppm (dd, J=9.3, 6.0 Hz, 1H, NCH); 13C NMR
(100 MHz, CD3OD): d=14.0 (CH3), 21.0 (CH3CH2), 24.5 (CHCH2CH2),
29.0 (CH3CH2CH2), 30.4 (CHCH2), 56.1 (CH(CH2)2CH2), 56.6
(CH3(CH2)2CH2), 70.8 (NCH), 173.6 ppm (C=O); IR (film): n˜ =3406,
2962, 2937, 2875, 1624, 1459, 1389 cmÀ1; MS (CI, CH5+) m/z (%):
172 (100) [M+H]+, 126 (15); HRMS-EI+ m/z [M]+ calcd for
C9H17N1O2: 171.1259, found: 171.1256.
Methyl 1-butyl-1,2,5,6-tetrahydropyridine-3-carboxylate (15b):
1,2,5,6-Tetrahydropyridine-3-carboxylic acid methyl ester (14)[41]
(212 mg, 1.5 mmol) was stirred with K2CO3 (829 mg, 6.0 mmol), KI
(50 mg, 0.3 mmol) and 1-bromobutane (22b, 411 mg, 3.0 mmol) in
dry acetone (6 mL) at room temperature for 24 h. The solvent was
evaporated, then water was added, and the solution was extracted
with CH2Cl2 (3”6 mL). The organic layers were collected, dried
over Na2SO4 and concentrated. The residue was purified by flash
chromatography over silica gel (n-heptane/Et2O 7:3, 1% N-ethyldi-
methylamine, Rf =0.2). Yield: 169 mg (57%) yellow oil: 1H NMR
(400 MHz, CDCl3): d=0.93 (t, J=7.3 Hz, 3H, CH3), 1.30–1.39 (m, 2H,
CH3CH2), 1.51–1.58 (m, 2H, CH3CH2CH2), 2.33–2.38 (m, 2H, CHCH2),
2.45–2.49 (m, 2H, CH3(CH2)2CH2), 2.53 (t, J=5.7 Hz, 2H, CHCH2CH2),
3.18–3.19 (m, 2H, NCH2C), 6.99–7.02 ppm (m, 1H, CHC); 13C NMR
(100 MHz, CDCl3): d=14.1 (CH3), 20.8 (CH3CH2), 26.6 (CHCH2), 29.2
(CH3CH2CH2), 49.1 (NCH2CC), 51.5 (CHCH2CH2), 51.6 (OCH3), 58.2
(CH3(CH2)2CH2), 129.0 (CCH), 138.0 (CCH), 166.4 ppm (C=O); IR
(film): n˜ =2954, 2932, 2872, 2807, 2765, 1717, 1657, 1437, 1262,
1194, 1141, 1089, 1048, 719 cmÀ1; MS (CI, CH5+) m/z (%): 198 (100)
[M+H]+, 168 (2), 154 (10), 141 (2); HRMS-EI+ m/z [M]+ calcd for
C11H19N1O2: 197.1416, found: 197.1407.
(R)-1-Butylpyrrolidine-2-carboxylic acid [(R)-8b]: Prepared accord-
ing to GP1 starting from d-proline [(R)-8a, 115 mg, 1.0 mmol] and
butyraldehyde (86.5 mg, 1.2 mmol); reaction time: 5 h. Yield:
168 mg (98%) light-yellow solid; mp: 71–738C; Analytical data
(1H NMR, 13C NMR, IR, MS) are in accordance with those of the S en-
20
D
antiomer (S)-8b; ½aŠ +58.18 (c=0.47 in MeOH).
(S)-2-(1-Butylpyrrolidin-2-yl)acetic acid [(S)-9b]: Prepared accord-
ing to GP1 starting from (S)-2-{1-[(benzyloxy)carbonyl]pyrrolidin-2-
yl}acetic acid[42] [(S)-11, 105 mg, 0.4 mmol] and butyraldehyde
1-Butyl-1,2,5,6-tetrahydropyridine-3-carboxylic acid (6b): Methyl
1-butyl-1,2,5,6-tetrahydropyridine-3-carboxylate
(15b,
70 mg,
(43.3 mg, 0.6 mmol) in absolute EtOH (3 mL); reaction time: 24 h.
20
0.35 mmol) was stirred with Ba(OH)2·8H2O (221 mg, 0.70 mmol) in
EtOH/H2O (1:1; 2.8 mL) at room temperature overnight. Dry ice was
then added until precipitation of barium was complete, and the re-
sulting milky suspension was filtered and concentrated. Yield:
44 mg (69%) white solid; mp: 105–1088C; 1H NMR (500 MHz,
CDCl3): d=0.94 (t, J=7.4 Hz, 3H, CH3), 1.33–1.40 (m, 2H, CH3CH2),
Yield: 66 mg (90%) light-yellow solid; mp: 76–778C; ½aŠ À101.38;
D
(c=0.15 in MeOH); 1H NMR (400 MHz, CD3OD): d=1.00 (t, J=
7.4 Hz, 3H, CH3), 1.35–1.52 (m, 2H, CH3CH2), 1.65–1.76 (m, 2H,
CH3CH2CH2), 1.78–1.87 (m, 1H, CHCH2CH2), 1.98–2.14 (m, 2H,
CHCH2CH2), 2.25–2.34 (m, 1H, CHCH2CH2), 2.50 (dd, J=16.6, 3.8 Hz,
1H, CH2CO), 2.70 (dd, J=16.6, 5.6 Hz, 1H, CH2CO), 2.93–2.99 (m,
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
516
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[1] M. Palacín, R. EstØvez, J. Bertran, A. Zorzano, Physiol. Rev. 1998, 78, 969–
1054.
1H, CH3(CH2)2CH2), 3.11 (dt, J=11.4, 8.0 Hz, 1H, CH(CH)2CH2), 3.33–
3.39 (m, 1H, CH3(CH2)2CH2), 3.53–3.59 (m, 1H, NCH), 3.65–3.72 ppm
(m, 1H, CH(CH)2CH2); 13C NMR (125 MHz, CD3OD): d=14.0 (CH3),
21.0 (CH3CH2), 23.2 (CHCH2CH2), 29.2 (CH3CH2CH2), 30.5 (CHCH2CH2),
35.8 (CH2CO), 54.0 (CH(CH2)2CH2), 54.1 (CH3(CH2)2CH2), 66.7 (NCH),
177.7 ppm (C=O); IR (KBr): n˜ =3434, 2963, 2937, 2876, 1593, 1468,
1448, 1379, 1366, 1036, 704 cmÀ1; MS (CI, CH5+) m/z (%): 186 (100)
[M+H]+, 142 (4), 126 (63); HRMS-EI+ m/z [M]+ calcd for
C10H19N1O2: 185.1416, found: 185.1407.
[2] a) Y. Zhou, S. Holmseth, R. Hua, A. C. Lehre, A. M. Olofsson, I. Poblete-
Naredo, S. A. Kempson, N. C. Danbolt, Am. J. Physiol.: Renal. Physiol.
2012, 302, F316–F328; b) Y. Zhou, S. Holmseth, C. Guo, B. Hassel, G.
Hofner, H. S. Huitfeldt, K. T. Wanner, N. C. Danbolt, J. Biol. Chem. 2012,
287, 35733–35746.
[3] a) Q. R. Liu, B. López-Corcuera, S. Mandiyan, H. Nelson, N. Nelson, J. Biol.
Chem. 1993, 268, 2106–2112; b) Y. Zhou, N. C. Danbolt, Front. Endocri-
nol. 2013, 4, 165.
[4] K. K. Madsen, H. S. White, A. Schousboe, Pharmacol. Ther. 2010, 125,
394–401.
(R)-2-(1-Butylpyrrolidin-2-yl)acetic acid [(R)-9b]: Prepared accord-
ing to GP1 starting from (R)-2-{1-[(benzyloxy)carbonyl]pyrrolidin-2-
yl}acetic acid[1] ([(R)-11, 118 mg, 0.45 mmol) and butyraldehyde
(48.7 mg, 0.67 mmol) in absolute EtOH (3.5 mL); reaction time:
24 h. Yield: 73 mg (88%) light-yellow solid; mp: 76–778C; Analyti-
[5] a) J. Masson, C. SagnØ, M. Hamon, S. E. Mestikawy, Pharmacol. Rev.
1999, 51, 439–464; b) F. Chen, M. B. Larsen, H. A. Neubauer, C. Sµnchez,
P. Plenge, O. Wiborg, J. Neurochem. 2005, 92, 21–28; c) U. Gether, P. H.
Andersen, O. M. Larsson, A. Schousboe, Trends Pharmacol. Sci. 2006, 27,
375–383; d) A. S. Kristensen, J. Andersen, T. N. Jorgensen, L. Sorensen,
J. Eriksen, C. J. Loland, K. Stromgaard, U. Gether, Pharmacol. Rev. 2011,
63, 585–640; e) T. Kvist, B. Christiansen, A. A. Jensen, H. Bräuner-Os-
borne, Comb. Chem. High Throughput Screening 2009, 12, 241–249.
[6] P. Krogsgaard-Larsen, E. Falch, O. M. Larsson, A. Schousboe, Epilepsy Res.
1987, 1, 77–93.
[7] H. Nelson, S. Mandiyan, N. Nelson, FEBS Lett. 1990, 269, 181–184.
[8] J. Guastella, N. Nelson, H. Nelson, L. Czyzyk, S. Keynan, M. C. Miedel, N.
Davidson, H. A. Lester, B. I. Kanner, Science 1990, 249, 1303–1306.
[9] a) B. S. Meldrum, A. G. Chapman, Epilepsia 1999, 40, S2–6; b) N. O.
Dalby, Eur. J. Pharmacol. 2003, 479, 127–137.
[10] P. D. Suzdak, K. Frederiksen, K. E. Andersen, P. O. Sorensen, L. J. Knutsen,
E. B. Nielsen, Eur. J. Pharmacol. 1992, 224, 189–198.
[11] L. M. Yunger, P. J. Fowler, P. Zarevics, P. E. Setler, J. Pharmacol. Exp. Ther.
1984, 228, 109–115.
[12] M. R. Pavia, S. J. Lobbestael, D. Nugiel, D. R. Mayhugh, V. E. Gregor, C. P.
Taylor, R. D. Schwarz, L. Brahce, M. G. Vartanian, J. Med. Chem. 1992, 35,
4238–4248.
[13] C. K. Thoeringer, S. Ripke, P. G. Unschuld, S. Lucae, M. Ising, T. Bettecken,
M. Uhr, M. E. Keck, B. Mueller-Myhsok, F. Holsboer, E. B. Binder, A. Er-
hardt, J. Neural Transm. 2009, 116, 649–657.
cal data (1H NMR, 13C NMR, IR, MS) are in accordance with those of
20
D
the S enantiomer (S)-9b. ½aŠ +101.78 (c=0.27 in MeOH); HRMS-
EI+ m/z [M]+ calcd for C10H19N1O2: 185.1416, found: 185.1427.
(S)-tert-Butyl 3-(2-methoxy-2-oxoethyl)pyrrolidine-1-carboxylate
(17): (S)-2-(1-(tert-Butoxycarbonyl)pyrrolidin-3-yl)acetic acid (16,
230 mg, 1.00 mmol) dissolved in DMF (3.0 mL) was stirred with
K2CO3 (416 mg, 3.0 mmol) and CH3I (456 mg 3.2 mmol) at room
temperature for 66 h. The solvent volume was significantly de-
creased in vacuo, and H2O (10 mL) was added. The aqueous phase
was extracted with EtOAc (5”10 mL), and the combined organic
layers were washed once with H2O (10 mL), Na2S2O3 solution (10%,
10 mL) and brine (10 mL) then dried (Na2SO4), filtered and concen-
20
D
trated in vacuo. Yield: 219 mg (90%) orange oil: ½aŠ +25.38 (c=
0.7 in MeOH); 1H NMR (400 MHz, CDCl3, 508C): d=1.46 (s, 9H,
(CH3)3), 1.56 (m, 1H, NCH2CH2), 2.05 (m, 1H, NCH2CH2), 2.30–2.47
(m, 2H, NCH2CH, CH2COO), 2.56 (m, 1H, CH2COO), 2.95 (m, 1H,
NCH2CH) 3.30 (m, 1H, NCH2CH2), 3.44 (m, 1H, NCH2CH2), 3.59 (m,
1H, NCH2CH), 3.68 ppm (s, 3H, OCH3); 13C NMR (100 MHz, CDCl3,
508C): d=28.6 (CH3)3), 31.0 (NCH2CH2), 35.0 (CH2COO), 37.6
(NCH2CH), 45.2 (NCH2CH2), 51.2 (NCH2CH), 51.6 (OCH3), 79.2
(C(CH3)3), 154.6 (NC=O), 172.5 ppm (CO); IR (KBr): n˜ =3447, 2960,
2780, 2495, 1736, 1678, 1439, 1202 cmÀ1; MS (ESI+) m/z: 244 [M+
H]+; HRMS-ESI+ m/z [M+H]+ calcd for C12H22NO4: 244.1543,
found: 336.1953.
[14] G. X. Liu, G. Q. Cai, Y. Q. Cai, Z. J. Sheng, J. Jiang, Z. Mei, Z. G. Wang, L.
Guo, J. Fei, Neuropsychopharmacology 2007, 32, 1531–1539.
[15] a) P. Krogsgaard-Larsen, Mol. Cell. Biochem. 1980, 31, 105–121; b) P.
Krogsgaard-Larsen, G. A. R. Johnston, J. Neurochem. 1975, 25, 797–802;
c) G. A. Johnston, P. Krogsgaard-Larsen, A. L. Stephanson, B. Twitchin, J.
Neurochem. 1976, 26, 1029–1032.
[16] a) L. Nielsen, L. Brehm, P. P. Y. Krogsgaard-Larsen, J. Med. Chem. 1990,
33, 71–77; b) O. M. Larsson, P. Thorbek, P. Krogsgaard-Larsen, A. Schous-
boe, J. Neurochem. 1981, 37, 1509–1516; c) F. E. Ali, W. E. Bondinell,
P. A. Dandridge, J. S. Frazee, E. Garvey, G. R. Girard, C. Kaiser, T. W. Ku,
J. J. Lafferty, G. I. Moonsammy, H.-J. Oh, J. A. Rush, P. E. Setler, O. D.
Stringer, J. W. Venslavsky, B. W. Volpe, L. M. Yunger, C. L. Zirkle, J. Med.
Chem. 1985, 28, 653–660.
[17] G. H. Fülep, C. E. Hoesl, G. Hçfner, K. T. Wanner, Eur. J. Med. Chem. 2006,
41, 809–824.
[18] a) M. Sindelar, T. A. Lutz, M. Petrera, K. T. Wanner, J. Med. Chem. 2013,
56, 1323–1340; b) G. Quandt, G. Hofner, K. T. Wanner, Bioorg. Med.
Chem. 2013, 21, 3363–3378.
(S)-3-(2-Methoxy-2-oxoethyl)pyrrolidin-1-ium
trifluoroacetate
(18): To (S)-tert-butyl 3-(2-methoxy-2-oxoethyl)pyrrolidine-1-carbox-
ylate (17, 204 mg, 0.838 mmol) dissolved in CH2Cl2 (4.8 mL) was
added TFA (744 mg, 6.53 mmol) at 08C. The ice bath was removed,
and the mixture was stirred for 17 h at room temperature. The sol-
vent and excess TFA were removed in vacuo. Yield: 216 mg (100%)
20
yellow oil: ½aŠ +11.88 (c=0.5 in MeOH); 1H NMR (400 MHz,
D
CD3OD, 198C): d=1.67 (dq, J=12.8, 9.0 Hz, 1H, NCH2CH2), 2.25
(dtd, J=12.8, 7.3, 4.2 Hz, 1H, NCH2CH2), 2.53 (dd, J=16.7, 8.0 Hz,
1H, CH2COO), 2.60 (dd, J=16.7, 6.4 Hz, 1H, CH2COO), 2.67 (m, 1H,
CH2CH), 2.91 (dd, J=11.8, 8.9 Hz, 1H, NCH2CH2), 3.23 (ddd, J=11.6,
9.2, 7.6 Hz, 1H, NCH2CH), 3.38 (ddd, J=11.6, 8.5, 4.2 Hz, 1H,
NCH2CH), 3.53 (dd, J=11.8, 7.6 Hz, 1H, NCH2CH2), 3.68 ppm (s, 3H,
OCH3); 13C NMR (100 MHz, CDCl3, 508C): d=31.0 (NCH2CH2), 35.5
(NCH2CH), 37.1 (CH2CO), 46.2 (NCH2CH2), 50.9 (NCH2CH), 52.3
(OCH3), 173.7 ppm (CO); IR (KBr): n˜ =2976, 2875, 1739, 1695, 1407,
1166 cmÀ1; MS (ESI+) m/z: 144 [M+H]+, 401 [2M+TFA+H]+;
HRMS-ESI+ m/z [M+H]+ calcd for C7H14NO2: 144.1019, found:
144.1019.
[19] A. Yamashita, S. K. Singh, T. Kawate, Y. Jin, E. Gouaux, Nature 2005, 437,
215–223.
[20] a) C. L. Piscitelli, H. Krishnamurthy, E. Gouaux, Nature 2010, 468, 1129–
1132; b) S. K. Singh, A. Yamashita, E. Gouaux, Nature 2007, 448, 952–
956; c) T. Beuming, J. Kniazeff, M. L. Bergmann, L. Shi, L. Gracia, K. Ranis-
zewska, A. H. Newman, J. A. Javitch, H. Weinstein, U. Gether, C. J.
Loland, Nat. Neurosci. 2008, 11, 780–789; d) M. Quick, L. Shi, B. Zehnp-
fennig, H. Weinstein, J. A. Javitch, Nat. Struct. Mol. Biol. 2012, 19, 207–
211; e) M. Quick, A.-M. L. Winther, L. Shi, P. Nissen, H. Weinstein, J. A. Jav-
itch, Proc. Natl. Acad. Sci. USA 2009, 106, 5563–5568; f) L. Shi, M. Quick,
Y. Zhao, H. Weinstein, J. A. Javitch, Mol. Cell 2008, 30, 667–677.
[21] a) E. Gouaux, Philos. Trans. R. Soc. London Ser. B 2009, 364, 149–154;
b) H. Krishnamurthy, E. Gouaux, Nature 2012, 481, 469–474; c) S. K.
Singh, C. L. Piscitelli, A. Yamashita, E. Gouaux, Science 2008, 322, 1655–
1661.
Keywords: docking · GABA transporter · GAT1 · homology
modeling · tiagabine
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
517
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[22] A. Palló, A. Bencsura, L. HØja, T. Beke, A. Perczel, J. Kardos, A. Simon, Bio-
chem. Biophys. Res. Commun. 2007, 364, 952–958.
[23] T. Wein, K. T. Wanner, J. Mol. Model. 2010, 16, 155–161.
[24] S. Skovstrup, O. Taboureau, H. Bräuner-Osborne, F. S. Jørgensen, Chem-
MedChem 2010, 5, 986–1000.
[33] M.-Y. Shen, A. Sali, Protein Sci. 2006, 15, 2507–2524.
[34] E. Zomot, A. Bendahan, M. Quick, Y. Zhao, J. A. Javitch, B. I. Kanner,
Nature 2007, 449, 726–730.
[35] L. R. Forrest, S. Tavoulari, Y.-W. Zhang, G. Rudnick, B. Honig, Proc. Natl.
Acad. Sci. USA 2007, 104, 12761–12766.
[25] J. Eustache, A. Grob, C. Lam, O. Sellier, G. Schulz, Bioorg. Med. Chem.
Lett. 1998, 8, 2961–2966.
[36] R. A. Laskowski, M. W. Macarthur, D. S. Moss, J. M. Thornton, J. Appl.
Crystallogr. 1993, 26, 283–291.
[26] K. E. Andersen, C. Braestrup, F. C. Gronwald, A. S. Jorgensen, E. B. Niel-
sen, U. Sonnewald, P. O. Sorensen, P. D. Suzdak, L. J. Knutsen, J. Med.
Chem. 1993, 36, 1716–1725.
[27] “N-Substituted Azaheterocyclic Carboxylic Acids and Their Esters”, W. E.
Bondinell, J. J. Lafferty, C. L. Zirkle (Smithkline Beckman Corp.), Eur. Pat.
No. EP 0066456 A1, 1982.
[28] “Preparation of Tiagabine, Its (S)-Isomer, and Racemate”, J. Zhang, C.
Jiang, R. Wen, G. Lin (Shanghai Institute of Organic Chemistry, Chinese
Academy of Sciences), Pat. No. CN1651426A, 2005.
[37] Instant JChem, version 5.4, ChemAxon, 2010: https://www.chemaxon.
com/products/instant-jchem-suite/instant-jchem/.
[38] Marvin, version 5.4.0.1, ChemAxon, 2010: https://www.chemaxon.com/
products/marvin/.
[39] M. F. Sanner, J. Mol. Graphics Modell. 1999, 17, 57–61.
[40] a) G. M. Morris, D. S. Goodsell, R. S. Halliday, R. Huey, W. E. Hart, R. K.
Belew, A. J. Olson, J. Comput. Chem. 1998, 19, 1639–1662; b) R. Huey,
G. M. Morris, A. J. Olson, D. S. Goodsell, J. Comput. Chem. 2007, 28,
1145–1152.
[29] a) C. Zepperitz, G. Hçfner, K. T. Wanner, ChemMedChem 2006, 1, 208–
217; b) G. Hçfner, K. T. Wanner, J. Chromatogr. B 2010, 878, 1356–1364.
[30] A. Kragler, G. Hçfner, K. T. Wanner, Eur. J. Med. Chem. 2008, 43, 2404–
2411.
[41] I. Coldham, K. M. Crapnell, J. C. Fernandez, J. D. Moseley, R. Rabot, J.
Org. Chem. 2002, 67, 6181–6187.
[42] J. M. Cassal, A. Furst, W. Meier, Helv. Chim. Acta 1976, 59, 1917–1924.
[31] H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig,
I. N. Shindyalov, P. E. Bourne, Nucleic Acids Res. 2000, 28, 235–242.
[32] B. Boeckmann, A. Bairoch, R. Apweiler, M.-C. Blatter, A. Estreicher, E. Gas-
teiger, M. J. Martin, K. Michoud, C. O’Donovan, I. Phan, S. Pilbout, M.
Schneider, Nucleic Acids Res. 2003, 31, 365–370.
Received: November 13, 2015
Published online on January 25, 2016
ChemMedChem 2016, 11, 509 – 518
www.chemmedchem.org
518
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim