Sulfide Analogues of Flupirtine and Retigabine with Nanomolar KV7.2/KV7.3 Channel Opening Activity

Article abstract of DOI:10.1002/cmdc.201900112

The potassium channel openers flupirtine and retigabine have proven to be valuable analgesics or antiepileptics. Their recent withdrawal due to occasional hepatotoxicity and tissue discoloration, respectively, leaves a therapeutic niche unfilled. Metabolic oxidation of both drugs gives rise to the formation of electrophilic quinones. These elusive, highly reactive metabolites may induce liver injury in the case of flupirtine and blue tissue discoloration after prolonged intake of retigabine. We examined which structural features can be altered to avoid the detrimental oxidation of the aromatic ring and shift oxidation toward the formation of more benign metabolites. Structure–activity relationship studies were performed to evaluate the K_V7.2/3 channel opening activity of 45 derivatives. Sulfide analogues were identified that are devoid of the risk of quinone formation, but possess potent K_V7.2/3 opening activity. For example, flupirtine analogue 3-(3,5-difluorophenyl)-N-(6-(isobutylthio)-2-(pyrrolidin-1-yl)pyridin-3-yl)propanamide (48) has 100-fold enhanced activity (EC₅₀=1.4 nm), a vastly improved toxicity/activity ratio, and the same efficacy as retigabine in vitro.

Full text of DOI:10.1002/cmdc.201900112

Accepted Article
Title: Sulfide Analogues of Flupirtine and Retigabine with Nanomolar
KV7.2/KV7.3 Channel Opening Activity
Authors: Andreas Link, Christian Bock, Abdrrahman S. Surur, Kristin
Beirow, Markus K. Kindermann, Lukas Schulig, Anja Bodtke,
and Patrick J. Bednarski
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemMedChem 10.1002/cmdc.201900112
Link to VoR: http://dx.doi.org/10.1002/cmdc.201900112
A Journal of
www.chemmedchem.org

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
Sulfide Analogues of Flupirtine and Retigabine with Nanomolar
KV7.2/KV7.3 Channel Opening Activity
Christian Bock,[a] Abdrrahman S. Surur,[a] Kristin Beirow,[a] Markus K. Kindermann,[a] Lukas Schulig,[a]
Anja Bodtke,[a] Patrick J. Bednarski[a] and Andreas Link*[a]
[a]
C. Bock, A. S. Surur, K. Beirow, Dr. M. K. Kindermann, L. Schulig, Dr. A. Bodtke, Prof. Dr. P. J. Bednarski, Prof. Dr. A. Link
Institute of Pharmacy
University of Greifswald
Friedrich-Ludwig-Jahn-Str. 17, 17489 Greifswald, Germany
Fax: (+49) (0)3834 4204895
E-mail: link@uni-greifswald.de
Supporting information for this article is given via a link at the end of the document.
Dedicated to Prof. Dr. Hans-Hartwig Otto on the occasion of his 80th birthday.
Abstract: The potassium channel openers flupirtine and retigabine
have proven to be valuable analgesics or antiepileptics. Their recent
withdrawal due to occasional hepatotoxicity and tissue discoloration,
respectively, leaves a therapeutic niche unfilled. Metabolic oxidation
of both drugs gives rise to formation of electrophilic quinones. These
elusive, highly reactive metabolites may induce liver injury in the case
of flupirtine and blue tissue discoloration after prolonged intake of
retigabine. We examined which structural features can be altered to
avoid the detrimental oxidation of the aromatic ring and shift oxidation
towards the formation of more benign metabolites. SAR studies were
performed to evaluate the KV7.2/3 opening activity of 45 derivatives.
Sulfide analogs were identified that are devoid of the risk of quinone
formation but possess potent KV7.2/3 opening activity. For example,
KV7 channels are slow activating and non-inactivating, especially
repetitive neuronal firing is intercepted, thus making them
validated drug targets for the control of pain and epilepsy.[5]
The five KV7.1-KV7.5 proteins form homo- and heterotetrameric
channels and are expressed throughout the human body with
definite distribution patterns for each protein. The predominant
form in the CNS are the KV7.2 and KV7.3 proteins, thus opening
of the heterotetramer KV7.2/3 is believed to be the main mode of
action of 1 for suppressing pain and convulsions. While the
analgesic properties of 1 have been shown in various studies in
humans,[6] the potential of 1 for the treatment of epilepsy has only
recently been considered based on results from rodent models
and indirectly by the approval and use of the identically
substituted triaminobenzene, retigabine (2), as an antiepileptic
drug.[7]
flupirtine
analog
3-(3,5-difluorophenyl)-N-(6-(isobutylthio)-2-
100-fold
(pyrrolidin-1-yl)pyridin-3-yl)propanamide (48) has
a
enhanced activity (EC50 = 1.4 nM), vastly improved toxicity/activity
ratio and the same efficacy as retigabine in vitro.
The mechanism underlying DILI after prolonged flupirtine use is
not entirely understood, but evidence indicates that the drug might
be metabolized to reactive intermediates (3 and its ortho-
azaquinone diimine isomer), that react with endogenous
nucleophiles.[8] Specifically, the detection of cysteine conjugates
in the urine of healthy subjects after oral administration of 1
strongly indicates the formation of reactive intermediates of
flupirtine.[9] Presumably, upon oxidation of 1, electrophile 3 is
formed which reacts with bionucleophiles, as could be shown in
vitro by oxidation of 1 with peroxidases and conjugation with
glutathione.[10] Surprisingly, 2 has not been associated with
hepatotoxicity but nevertheless was recently removed from the
European market for causing blue discolorations of various
tissues after extended intake.[11] A recent publication found
derivatives of 2 in the choroid of rats by imaging mass
spectrometry after treatment with the anticonvulsant.[12] The
authors suspect that 2 is also converted to quinone diimines (4),
which then dimerize to hydrophenazine structures and become
oxidized to phenazinium ions (e.g. 5), causing the blue
pigmentation (see Figure 1).[13] In contrast to 1, this pathway
seems to be dependent on melanin, as 5 could not be detected in
albino rats and thus is dominant in pigmented tissues as the skin
and eyes. The side effects of 2 are less severe than 1 and mostly
reversible.[14]
Introduction
In Germany, 62% of the 672 million prescribed daily defined
doses (DDD) for treating pain consisted of opioids while the other
34% received nonsteroidal anti-inflammatory drugs (NSAIDs).[1]
Patients that tolerate neither NSAIDs because of gastrointestinal
bleeding nor opioids because of adverse effects such as addiction,
obstipation or respiratory depression, were often treated with
flupirtine (1) (5.3 million DDD in 2017). Due to the distinctive mode
of action, the triaminopyridine 1 filled a therapeutic niche in these
patients for more than 30 years. However, in 2018 flupirtine was
withdrawn in the European Union because of the insufficient
control of drug induced liver injury (DILI) in rare but fatal cases.[2]
The analgesic effect of 1 is thought to be achieved by two
mechanisms. On the one hand, 1 is increasing the affinity of γ-
aminobutyric acid for GABAA receptors, especially in pain
mediating neurons.[3] On the other hand, flupirtine (1) is increasing
the opening probability of voltage-dependent potassium channels
of the KV7 family.[4] The activation of KV7 channels leads to a
potassium efflux, thus hyperpolarizing the membrane potential of
neurons, leading to reduced forwarding of action potentials. As
Aiming to develop safer alternatives for the treatment of pain and
epilepsy, we have attempted to separate activity from toxicity by
1
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
Figure 1. Structures of the approved drugs flupirtine (1) and retigabine (2) and their proposed toxification pathways via the quinone diimines 3 and 4.
NO2
NO2
Cl
altering the oxidative behavior of 1 and 2 through replacement of
the secondary amino group for a sulfide bridge. While the steric
properties are arguably retained, the oxidation of the aromatic
core should be avoided and instead the sulfur atom oxidized to
non-electrophilic and potentially less toxic metabolites. To gain
further insights into the SAR of the sulfide analogues, we
synthesized a limited library of compounds bearing a variety of
functional groups while keeping the methylene-sulfur-aromatic
core motif. After structural confirmation, we tested the compounds
for their ability to augment potassium currents in KV7.2/3 channels
and their toxicity in TAMH and HEP-G2 cells. Furthermore, we
characterized the compounds by cyclic voltammetry because a
correlation between oxidizability and activity was suspected
earlier.[15] To rationalize the results obtained from the biological
evaluation, we also applied homology modeling and molecular
docking to identify potential interactions within the binding pocket.
Br
N
11
Cl
Cl
N
6
a1-4
b
NO2
R1
NO2
R1
NO2
R1
c
b
HS
N
1 = NH
R2
N
17 35
N
1 = NH
S
-
R
7
8
9 R1
12
13
14 R1
R
2
2
R
1 = NHCH
= 1-pyrrolidinyl
R1 = NHCH
= 1-pyrrolidinyl
3
3
R
1 = 4-morpholinyl
R
1 = 4-morpholinyl
10
15
16
R
1 = CH
3
Scheme 1. Synthesis of sulfide analogues of flupirtine. a) 1. NH3, ethanol, 3 h
0 °C, 16 h rt; 2. morpholine, ACN, 30 min 0 °C, 3 h rt, 3. pyrrolidine, DCM, 1 h
0 °C, 16 h rt; 4. methylamine, Et3N, ACN, 10 min 0 °C, 30 min rt; yield 16–95%;
b) Na2S∙9H2O, S8, NaOH, ethanol, 3–5 h reflux, 32–75%; c) alkyl bromide, KOH,
DMF, 1–5 h rt; yield 13–97%, for R2 in 17-35 see table 1.
Results
Better results were obtained when the reduction was conducted
with tin-(II)-chloride dihydrate in ethanol. The observed reduction
was quantitative, yet the reaction times were quite long (up to
48 h) and not reproducible. The use of iron powder and
ammonium chloride shortened the reaction time to three hours,
but also produced some undefined by-products. The cleanest
conversion was achieved with Raney nickel under hydrogen
atmosphere, showing quantitative yields within a couple of hours.
Without any workup the formation of different amide and
carbamate derivatives was executed by dropwise addition of
triethylamine and acylating agents. While the formation of amides
was carried out at 0 °C to achieve a better selectivity towards N3-
acylation, the carbamoylation usually took place at 40 °C as the
reactivity of the ethyl chloroformate was significantly lower. The
compounds thus obtained are summarized in table 2.
Chemistry
The synthesis of the sulfide analogues 36-60 of flupirtine began
with 2,6-dichloro-3-nitropyridine (6) as we described in recent
publications for a limited set of such sulfides (see schemes 1 and
2).[16] After replacement of the 2-chloro substituent by different
nitrogen species, formation of the thiol in position six was
conducted. While thiourea as a sulfur source only yielded a small
amount of thiol (24% yield),[17] elemental sulfur and sodium sulfide
resulted in a nearly quantitative conversion, i.e. up to 93% of 12
could be obtained.[18] Under basic conditions 12 was
subsequently alkylated by a variety of different alkyl bromides in
a quantitative manner.[19] A summary of the intermediates is given
in table 1.
The last step included the reduction of the nitro group as well as
the selective carbamoylation of the resulting primary amino group
in position 3 (see scheme 2). We did not isolate the intermediate
diamino-pyridine as those compounds tend to form ortho-quinone
diimines and therefore show an insufficient stability. Our initial
attempt to reduce the nitro group with a palladium catalyst under
hydrogen atmosphere, which is convenient for nitrogen
analogues of flupirtine,[15] failed to achieve complete conversion.
R3
NH
R1
O
NO2
R1
a
1-3, b
R2
N
17 35
R2
N
S
S
-
-
36 60
Scheme 2. Reduction and acylation. a) 1. SnCl2∙2H2O, ethanol, 16–48 h 70 °C;
2. Fe, NH4Cl, H2O/2-propanol, 3–5 h reflux; 3. H2, Raney nickel, THF, 1–5 h rt;
b) acylating reagent, Et3N or EtN(iPr)2, 0–40 °C 1–24 h, yield 9–74%. For R1-3
in 36-60 see table 2.
2
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
Table 1. Overview of the synthesized precursors 17-35
We also wanted to investigate whether electron-withdrawing
substituents in position five of the pyridine ring would influence
the biological activity or the toxicity in comparison to 64. Therefore
a bromination was performed with N-bromosuccinimide and
ammonium chloride as described by Csuk et al..[20] But instead of
using the nitro precursor 17 we tried to directly brominate
compound 36 (scheme 4). The carbamate moiety proved to be
stable under the reaction conditions and 65 could be isolated in
29% yield.
Entry
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
R1
R2
NH2
4-fluorobenzyl
NH2
benzyl
NH2
3,4-difluorobenzyl
4-(trifluoromethyl)benzyl
4-(pentafluoro-λ6-sulfanyl)benzyl
cyclohexylmethyl
cyclohexylmethyl
isobutyl
NH2
NH2
O
O
O
O
NH2
NH
NH2
Br
NH
NH2
a
1-pyrrolidinyl
1-pyrrolidinyl
NH2
S
N
S
N
36
65
F
F
Scheme 4. Bromination of 36. a) NBS, NH4Cl, THF/ACN, 2 h 0 °C, yield 29%.
cyclopropylmethyl
cyclobutylmethyl
(pyridin-2-yl)methyl
(piperidin-1-yl)ethyl
2,4-dimethoxybenzyl
1,1′-biphenyl
NH2
Another possibility of reducing the reactivity of the aromatic
moiety to prevent the formation of dimers or conjugates with
endogenous nucleophiles is to exchange the pyridine core for
another heterocycle. We chose to use a pyrimidine motif, as the
ring position that equals position five of the pyridine is occupied
by nitrogen; as nitrogen is tetravalent when positively charged,
only, the pyrimidine ring is less likely to form covalent bonds with
other molecules. For the synthesis of the analogue 69 we started
with 5-nitrouracil (66, see scheme 5). The pyrimidine base was
reacted to form the corresponding dichloro derivative by refluxing
in phosphoryl chloride with dimethyl aniline as described by
Whittaker and Jones.[21] For the following amination, the
compound was dissolved in ethanol and treated with aqueous
ammonia at 0 °C to obtain 68 in 85% yield. The intermediates 67
and 68 resemble the structure of cyanuric chloride, thus it is not
surprising that the compounds are quite labile to hydrolysis and
require a rapid work-up and dry storage. The procedures for
sulfide formation and carbamoylation were similar to the ones of
the pyridine analogues described above.
NH2
NH2
NH2
NH2
NHCH3
4-morpholinyl
NH2
benzyl
4-fluorobenzyl
(tetrahydropyran-2-yl)methyl
(1,3-dioxolan-2-yl)methyl
4-fluorobenzyl
4-morpholinyl
CH3
Since position five of the pyridine ring is susceptible to substitution
reactions with endogenous proteins or peptides,[9,10] we
introduced a methyl group at this position to prevent such toxic
reactions. The synthesis of the corresponding derivate started
with 2-chloro-5-methyl-3-nitropyridine (61), which was converted
to the corresponding N-oxide using hydrogen peroxide - urea and
trifluoroacetic acid anhydride (see scheme 3). Next, 62 was
reacted with phosphoryl chloride to obtain 63. The following
amination, sulfide formation and ultimately the reduction and
carbamoylation were conducted in analogy to the previously
described pyridine derivatives.
NO2
O
NO2
Cl
NO2
NH2
HN
N
N
a
b
N
H
66
Cl
N
67
Cl
N
68
O
O
O
NH
N
S
N
NH2
F
69
NO2
NO2
NO2
a
b
Scheme 5. Synthesis of 69. a) POCl3, DMA, 2 h reflux, yield 53%; b) 25%
aqueous NH3, ethanol, 3 h 0 °C, yield 85%. Total yield of the four-step-synthesis
to obtain 69 from 68: 11%.
N
Cl
N
Cl
N
63
Cl
Cl
61
O
62
O
O
The phenyl ring core of 2 would appear to be a superior choice for
KV7 channel openers than a pyridine ring because 2 does not
show the same hepatotoxic potential as 1. Therefore, we also
synthesized the thio-analogue of 2 bearing a sulfur bridge. We
started from the commercially available 2-fluoro-4-bromo-aniline
(70, see scheme 6). In the first step, 70 was oxidized to the
corresponding nitro derivative (71) by using hydrogen peroxide in
trifluoroacetic acid.[22]
NH
NH2
N
S
64
F
Scheme 3. Synthesis of 64. a) urea-H2O2, TFAA, DCM, 30 min 0 °C, 36 h rt,
yield: 82%; b) POCl3, 5 h reflux, yield 74%. Total yield of the five-step-synthesis
to obtain 64 from 63: 7%.
3
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
Table 2. Overview of the synthesized sulfide analogues 36-77 in comparison to 1 and 2.
a
b
e
En
try
R1
R2
R3
Epa
EC50
[µM]
Efficacyc
[%]
LD25d [µM]
LogD7.4
[mV]
TAMH
HEP-G2
74 ± 40
269 ± 166
> 40
1
251
205
464
530
530
430
487
455
460
459
452
0.918 ± 0.099
0.147 ± 0.012
1.279 ± 0.293
2.134 ± 0.457
4.544 ± 1.457
0.260 ± 0.082
0.066 ± 0.013
1.746 ± 0.347
0.771 ± 0.076
0.237 ± 0.022
0.240 ± 0.066
100
103 ± 47
> 400
> 40
2.96
2.57
3.54
3.43
3.49
3.90
4.00
3.45
3.66
3.91
4.10
2
134 ± 16
93 ± 13
140 ± 11
118 ± 28
100 ± 23
114 ± 16
110 ± 15
96 ± 9
36
37
38
39
40
41
42
43
44
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
4-fluorobenzyl
4-fluorobenzyl
4-fluorobenzyl
4-fluorobenzyl
4-fluorobenzyl
benzyl
ethoxy
butyl
19 ± 11
> 63
> 63
isobutyl
> 63
3,4-difluorophenyl
3,5-difluorobenzyl
ethoxy
8 ± 5
4 ± 1
> 5
> 7.5
91 ± 10
25 ± 3
13 ± 6
> 10
97 ± 72
> 30
3,4-difluorobenzyl
4-(trifluoromethyl)benzyl
ethoxy
ethoxy
115 ± 19
78 ± 10
> 20
4-(pentafluoro-λ6-
ethoxy
> 10
sulfanyl)benzyl
45
46
47
48
NH2
cyclohexylmethyl
cyclohexylmethyl
isobutyl
butyl
558
n.o.f
n.o.f
n.o.f
1.320 ± 0.440
0.021 ± 0.006
0.0245 ± 0.0069
0.0014 ± 0.0001
120 ± 19
162 ± 9
6 ± 2
102 ± 61
> 125
3.97
5.10
4.70
4.73
1-pyrrolidinyl
1-pyrrolidinyl
1-pyrrolidinyl
3,5-dimethoxybenzyl
3,5-dimethoxybenzyl
16 ± 5
> 125
35 ± 27
141 ± 11
136 ± 13
> 125
isobutyl
(3,5-
38 ± 18
difluorophenyl)ethyl
49
NH2
cyclopropylmethyl
(3,5-
477
0.515 ± 0.035
117 ± 22
155 ± 86
75 ± 25
3.86
difluorophenyl)ethyl
50
51
52
53
54
55
56
57
NH2
cyclobutylmethyl
(pyridin-2-yl)methyl
(piperidin-1-yl)ethyl
3,5-dimethoxybenzyl
1,1′-biphenyl
3,5-difluorobenzyl
ethoxy
514
497
605
449
447
542
651
598
0.021 ± 0.003
> 10
107 ± 7
44 ± 23
503 ± 270
> 500
41 ± 23
405 ± 157
284 ± 178
48 ± 37
132 ± 97
47 ± 36
> 20
4.03
2.62
2.04
3.61
4.34
4.60
4.73
3.68
NH2
-
NH2
ethoxy
> 10
-
NH2
ethoxy
> 10
-
75 ± 55
14 ± 4
> 125
NH2
ethoxy
0.253 ± 0.042
0.109 ± 0.078
0.0009 ± 0.0004
1.348 ± 0.185
69 ± 9
119 ± 10
127 ± 9
106 ± 17
4-morpholinyl
4-morpholinyl
NH2
4-fluorobenzyl
ethoxy
4-fluorobenzyl
3,5-difluorobenzyl
3,5-difluorobenzyl
> 5
(tetrahydropyran-2-
yl)methyl
> 125
108 ± 61
58
59
60
64
65
69
77
4-morpholinyl
NHCH3
(1,3-dioxolan-2-yl)methyl
benzyl
3,5-difluorobenzyl
3,5-difluorobenzyl
ethoxy
596
422
n.o.f
406
487
636
406
0.122 ± 0.031
0.015 ± 0.002
0.269 ± 0.031
0.447 ± 0.293
> 10
112 ± 11
147 ± 9
129 ± 3
72 ± 4
-
141 ± 99
> 7.5
> 10
119 ± 75
> 10
3.85
4.23
3.85
3.94
4.17
2.71
3.24
CH3
4-fluorbenzyl
> 30
> 15
> 20
> 7.5
> 80
> 7.5
> 80
1.024 ± 0.298
0.166 ± 0.035
122 ± 4
128 ± 2
> 50
> 50
a Measured as 10 mmol/L solution in TRIS-buffer (100 mmol/L, pH 7.4). b EC50 values were obtained in HEK293 cells overexpressing KV7.2/3 channels, shown are
means and standard deviations (n≥3). c Calculated relative to the maximum flupirtine-induced fluorescence signal. d LD50 values were determined by MTT assay
f
after a 24 h exposure, shown are means and standard deviations (n≥3). e LogD7.4 values were determined by a HPLC method (n≥2). No oxidation took place
between −0.5 V and 1.0 V.
4
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
NH2
a
NO2
F
NO2
NH2
isomers were formed in equal amounts, thus a workup of the 1:1
mixture did not seem promising.[25] Correspondingly, we Boc-
protected the primary amine of 75 to obtain 76. Hydrogenation
using Raney nickel as a catalyst and following carbamoylation
with ethyl chloroformate and N,N-diisopropylethylamine produced
the correct carbamate. After chromatographic workup, the Boc-
groups were removed using trifluoroacetic acid in
dichloromethane, resulting in the formation of 77.
b
Br
F
Br
Br
70
71
72
c
NO2
NO2
NH2
N
N
H2
O2
S
By replacing the secondary amine of 1 and 2 by sulfur, it is very
likely that the metabolic pathways of the compounds are different.
Organically bound sulfur is easy oxidizable and tends to form
sulfoxides and sulfones catalyzed by flavin monoxygenases
(FMO) or cytochrome P450 (CYP) enzymes as observed in
albendazole.[26] Correspondingly, our sulfide analogues are likely
to undergo facile sulfur oxidation in vivo. To investigate the activity
and the toxicity of those putative metabolites, the corresponding
sulfoxides and sulfones were also prepared. A variety of
sulfoxides and sulfones were synthesized (see table 3) by
oxidation of the sulfides using meta-chloroperoxybenzoic acid (m-
CPBA). The addition of one equivalent of oxidizing agent resulted
in almost exclusive formation of racemic sulfoxides, as proven by
SFC-MS separation by the same method as reported by
Hofstetter et al..[27] A larger excess (>2 equivalents) lead to the
formation of the optically inactive sulfones (see scheme 7). The
clean conversion of the sulfides under oxidative conditions also
underlines our hypothesis that the introduction of a sulfur bridge
as replacement for an amine shifts oxidation away from the
formation of quinone diimines.
NH2
HS
S
74
73
d
e
NO2
NH2
S
75
F
f
NO2
Boc
S
N
76
Boc
F
g, h, i
O
O
R2
O
NH
NH
a
b
NH2
S
77
R1
N
36 60
NH2
S
F
-
Scheme 6. Synthesis of 77. a) H2O2 (30%), TFA, 30 min rt, 1 h reflux, yield
71%; b) NH3, ethanol, 3 h 0 °C, 16 h rt, yield 86%; c) 1,2-ethanedithiol,
CuSO4•5H2O, KOH, DMSO/H2O, 23 h 60 °C, yield of 73: 38%; d) NaBH4,
K2CO3, 4-fluorbenzylbromide, 90 min rt, yield 74%; e) 4-fluorbenzyl bromide,
NaOH, DMF, 45 min rt, yield 91%; f) Boc2O, THF, 25 h rt, yield 89%; g) H2,
Raney-Nickel, 1 h rt, 4 h 40 °C; h) ethyl chloroformate, EtN(iPr)2, 40 h 40 °C; i)
TFA, DCM, 30 min rt. Combined yields of g), h) and i): 9%.
R2
R2
NH
NH2
O
O
NH
O
S
R1
R1
N
78
NH2
N
S
-
-
86
89 92
O
O
Subsequently, the fluoro substituent was replaced selectively via
nucleophilic substitution with gaseous ammonia, yielding 72.
While the formation of the thiol went smoothly with the pyridine
derivatives (vide supra), the introduction of a sulfur atom into 72
was more troublesome. At first, again, elemental sulfur as well as
sodium sulfide were used for the reaction yielding only 26% of 73.
The majority of the product was converted into the disulfide,
yielding 43% of 74 after workup by flash chromatography. In a
second attempt, 1,2-ethanedithol was used as a sulfur source and
copper-(II)-sulfate as a catalyst for thiol formation as reported by
Liu et al..[23] Again, the majority of the product formed consisted
of disulfides and only 38% of 73 could be isolated. The formation
of disulfides could possibly be prevented by the addition of
dithiothreitol, yet we did not try to improve the reaction any further
as also 74 could be reacted to the desired sulfide 75 in a one pot
reaction. By reduction of the S-S bond using sodium borohydride
and subsequent alkylation with 4-fluorobenzyl bromide, 75 could
be obtained in 74% yield.[24] While a patent describes the
synthesis of 2 by hydrogenation of the 2-nitro-aniline intermediate
and carbamoylation of the resulting ortho-diamine, our attempt of
reacting 75 to 77 failed, as only a mixture of 77 and its isomer
could be isolated. SFC-DAD investigations showed that both
O
O
O
O
NH
NH
S
N
S
N
NH2
O
87
O
88
F
F
Scheme 7. Synthesis of sulfoxides and sulfones. a) m-CPBA (1.1 eq), DCM,
0 °C 1–4 h, yield 16–89%; b) m-CPBA (>2 eq), DCM, 0 °C 2-6 h, yield 25–55%.
For R1-2 in 78-92 see table 3.
The sulfoximine moiety is an underrepresented sulfur functional
group in medicinal chemistry.[28] It is comparable to sulfones, yet
incorporating
a slightly basic nitrogen atom and showing
increased water solubility. Thus, we aimed to prepare the
sulfoximine analogue of 36, keeping close to the structure of lead
1. The synthesis of sulfoximines usually begins with the
corresponding sulfoxides, which are reacted with sodium azide or
other nitrogen sources.[29] Our attempt to react 78 to give the
sulfoximine
95
using
ammonium
carbamate
and
diacetoxyiodobenzene failed as the carbamate moiety is not
stable under these conditions. Therefore, we went a step back
5
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
NO2
NH2
and tried to directly convert 17 to the sulfoximine using a method
reported by Xie and coworkers.[30]After chromatographic workup
51% of the desired product 93 could be obtained. The downside
of this reaction cascade is the need for carbamate formation at
the end of the reaction sequence. As already mentioned, the
sulfoximine nitrogen is basic and can react with electrophiles as
for example ethyl chloroformate. Nevertheless, we waived the use
of protecting groups and tried to reduce 93 and form the ethyl
carbamate. After chromatographic workup 95, its isomer 96 and
the biscarbamoylated product 94 were isolated (see scheme 8).
In a preceding publication, Lemmerhirt et al. investigated a
possible correlation between toxicity and oxidizability of 1 and a
series of close analogues, in analogy to the mechanism of toxicity
of acetaminophen and its metabolite NAPQI.[15] While the initial
hypothesis could not be proven, a relation between the anodic
peak potential (Epa) obtained by cyclic voltammetry (CV) and the
activity of their compounds was observed; substances that were
oxidized at voltages below 400 mV were able to open KV7
channels, while compounds that were oxidized at higher voltages
failed to show activity in the cell-based assay. To gain further
insight into this hypothesis, the oxidizability of the sulfur
containing analogues was measured again by CV in the same
manner as reported. In general, the substitution of the secondary
amine by sulfur lead to an increase in oxidation potentials; i.e.
S
N
N
17
F
F
a
NO2
NH2
NH
S
O
93
b
O
O
O
O
NH
NH2
NH2
N
S
O
N
S
O
O
O
N
N
NH2
96
94
F
F
O
O
NH
NH2
NH
S
N
O
F
95
comparing
1
(Epa = 251 mV) to its sulfur congener 36
Scheme 8. Sulfoximine formation. a) PhI(OAc)2, H2NCO2NH4, methanol, 1 h rt,
yield 51%; b) SnCl2∙2H2O, 2 h 70 °C, ethyl chloroformate, Et3N, 2 h 40 °C, yield
of 95: 1%.
(Epa = 464 mV), the oxidizability changed by almost 200 mV (see
figure 2). The same applies to 2 and 77, where also a difference
of 200 mV was observed.
Table 3. Overview of sulfoxides 78-88, sulfones 89-92, and sulfoximine 95.
a
b
e
Entry
R1
R2
Epa
EC50
[µM]
Efficacyc
[%]
LD25d [µM]
TAMH HEP-G2
LogD7,4
[mV]
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
95
4-fluorobenzyl
benzyl
ethoxy
653
645
639
626
709
442
654
628
707
n.o.f
463
717
709
719
711
n.d.
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
> 125
> 125
878 ± 414
> 250
> 100
> 500
244
2.64
2.50
2.79
3.20
2.49
2.81
3.72
3.24
3.24
3.02
3.10
2.91
2.75
3.07
3.42
n.d.
ethoxy
713 ± 294
> 250
3,4-difluorobenzyl
4-(trifluoromethyl)benzyl
4-fluorobenzyl
3,5-dimethoxybenzyl
1,1′-biphenyl
ethoxy
ethoxy
> 100
butyl
> 500
ethoxy
78 ± 70
41 ± 20
77 ± 37
135 ± 23
> 500
ethoxy
> 63
4-fluorobenzyl
4-fluorobenzyl
3,4-difluorphenyl
3,5-difluorbenzyl
> 125
73 ± 45
> 500
> 500
62 ± 20
> 500
> 125
> 50
> 500
4-fluorobenzyl
ethoxy
ethoxy
ethoxy
ethoxy
146 ± 127
118 ± 8
> 125
benzyl
3,4-difluorobenzyl
4-(trifluoromethyl)benzyl
> 30
> 125
> 125
a Measured as 10 mmol/L solution in TRIS-buffer (100 mmol/L, pH 7.4). b EC50 values were obtained in HEK293 cells overexpressing KV7.2/3 channels, shown are
means and standard deviations (n≥3). c Calculated relative to the maximum flupirtine-induced fluorescence signal. d LD50 values were determined by MTT assay,
shown are means and standard deviations (n≥3). e LogD7.4 values were determined by a HPLC method (n≥2). f No oxidation took place between −0.5 V and 1.0 V.
6
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
In addition, a large difference between the different sulfur
oxidation states was also observed, as seen in figure 3. When
oxidizing 36 to 78 an increase of again almost 200 mV was
detected (Epa = 464 mV vs. 653 mV). The further oxidation to the
corresponding sulfone 89 increased the oxidation potential by
additional 60 mV (Epa = 717 mV). The exchange of the alkyl
residues of the sulfides had barely any influence on the Epa,
whereas replacing the carbamate moiety by amides markedly
reduced the oxidizability, see the comparison of 36 with 37
(Epa = 464 mV vs. 530 mV, respectively). The aromatic core also
had an influence on the oxidative behavior. While the phenyl and
pyridyl cycles were oxidized at 406 and 464 mV, respectively, the
pyrimidine ring was oxidized at 636 mV. We also obtained
compounds that were resistant to oxidation at all applied voltages
up to 1 V. All of them (46, 47, 48 and 60) are lacking the primary
amino group and instead incorporate a pyrrolidinyl or methyl
residue.
concentration (see figure 4).[31] The efficacy of the compounds
was calculated relative to the maximal flupirtine (1) induced
fluorescence signal. Representative concentration-activity curves
are shown in Figure 4. For a summary of the activity data see
tables 2 and 3. Retigabine (2) shows ca. 6 times greater potency
and 34% greater efficacy than the flupirtine (1). Interestingly, the
sulfide analogon 36 is only slightly less potent than 1 (EC50
=
1.28 µM vs. 0.92 µM) with similar efficacy. This result goes to
show that the substitution of the secondary amine by a sulfide is
generally well tolerated. Thus we modified the substituents of the
pyridine core and the core itself to obtain more active compounds,
derive structure-activity relationships of this class of compounds
and investigate their hepatotoxic potential. First, the 4-
fluorobenzylic moiety was modified to overcome the minimal loss
of potency. The removal of the 4-fluoro atom lowered potency
even further (41, EC50 = 1.75 µM), while the introduction of an
additional fluorine atom lead to a modest increase (42, EC50
=
0.78 µM). Comparable to the data obtained by Kumar et al., a
trifluoromethyl as well as a pentafluorosulfanyl group improved
potency around 6-fold (43 and 44).[32] The drawback of 44 is its
lower efficacy (78%). The introduction of a second phenyl moiety
improves potency in a similar manner, while also showing lower
efficacy (54, EC50 = 0.25 µM, 69% efficacy). An attempt to
introduce more hydrophilic groups compared to the 4-
fluorobenzylic motif, i.e., pyridinomethyl (51) and piperidinoethyl
(52), yielded inactive compounds.
2.5
1.5
0.5
-0.5
-1.5
-0.5
0
0.5
1
U [V]
Figure 2. Cyclovoltammograms of flupirtine (1, grey spheres) and sulfide
analogue 36 (black triangles).
9
7
5
3
1
Figure 4. Concentration-response curves of flupirtine (1, black squares),
retigabine (2, dark grey triangles) and compound 48 (light grey spheres).
-1
-3
As especially lipophilic compounds seemed to have favourable
activities, we decided to determine LogD7.4 values of our
compounds via an HPLC method. A correlation could be found in
this set of compounds as potencies increased with increasing
lipophilicity. Compounds with a LogD7.4 > 3.5 were active, with
only one exception: the quite lipophilic dimethoxy derivative 53
(LogD7.4 = 3.61) turned out to be inactive. As a rule, the
compounds with LogD7.4 < 3.5 showed no channel opening
properties.
Next we aimed to increase potency by exchanging the carbamate
moiety for more stable amide structures. Although bulky aliphatic
amides have been reported to be highly active,[33] our sulfide
derivatives with butyric acid (37 and 45) as well as pivalic acid
amides (38) only yielded potencies in the micromolar range
(EC50 = 1.32–4.54 µM). However, in case of 37 significantly better
-0.5
0
0.5
1
U [V]
Figure 3. Cyclovoltammograms of 36 (black triangles), 78 (grey spheres) and
89 (light grey squares).
Biology
To determine the KV7.2/3 channel opening properties of our
synthesized compounds, we used a cellular assay with HEK293
cells overexpressing these channels. Compound potency was
determined by measuring the intensity of the fluorescence signal
after thallium entered the cells through KV7 channels and
complexed with a specific dye, as a function of compound
7
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
efficacy (140%) was noted. To further increase the potency of the
sulfides, we incorporated the template published by Dalby-Brown
et al. bearing a 3,5-difluorophenylacetic acid amide and a
morpholine moiety.[34] The corresponding sulfide analogue 56
showed a high potency in opening KV7.2/3 channels with an EC50
of 0.0009 µM. Keeping the amide structure constant, the rest of
the molecule was altered, again showing a correlation between
activity and lipophilicity. The least hydrophobic analogue 57
(LogD7.4 = 3.68) is less active than 1 (EC50 = 1.35 µM). With
increasing lipophilicity (57<58<40<50<59<56) again the channel
opening activity increased, peaking in 56 (LogD7.4 = 4.73) being
about 1,000 times more active than 1 and showing a higher
efficacy (147%). The carboxylic amide structure was also
shortened by removing the methylene bridge yielding 39, which is
three times less active than 40. When prolonging the amide chain
by a methylene group, mixed effects were observed. While the
simultaneous substitution of the 4-fluorbenzyl residue by a
cyclopropylmethyl group (49) lowered the potency eightfold,
introduction of an isobutyl moiety and replacement of the NH2-
group by pyrrolidine yielded one of the most potent compounds of
the entire series (48, EC50 = 0.0014 µM). To investigate the
influence of electron-donating substituents in the amide chain,
fluorine was replace by methoxy groups; the resulting derivatives,
bearing a pyrrolidine motif and isobutyl (46, EC50 = 0.021 µM) or
cyclohexylmethyl residues (47, EC50 = 0.025 µM), were highly
active and more efficacious than 1.
observed up to the maximal soluble concentration of 40 µM (see
figure 5).
Figure 5. Concentration-viability curves of 1 (grey squares) and 36 (black
triangles) in HEP-G2 cells.
Oxidizing the sulfides to sulfoxides or sulfones led to lower toxicity
in the hepatic cell lines, e.g. compare 41 with its respective
sulfoxide (79) and sulfone (90). While the sulfide 41 was as toxic
as 1, the sulfoxide 79 was ten times less toxic. The corresponding
sulfone 90 was less toxic in TAMH cells compared to 41 and none
toxic in the HEP-G2 line up to the maximum soluble concentration
of 500 µM. Surprisingly big differences between both cell lines
could be observed for a couple of compounds. While 45 was less
toxic than 1 in HEP-G2 cells (LD25 = 102 µM) it exceeded the
toxicity of flupirtine 16-fold in TAMH cells (LD25 = 6 µM). Moreover,
the reversed case was observed as 49 was half as toxic in TAMH
cells as in HEP-G2 cells (LD25 = 155 vs. 75 µM). In general, the
TAMH cell line was more susceptible to toxicity than the HEP-G2
line.
To complete the SAR, we also modified the pyridine core by either
adding substituents or replacing it with other rings. When adding
a
methyl group in position
5
(64) potency doubled
(EC50 = 0.447 µM), while efficacy (72%) decreased. Even
stronger effects were observed when the methyl group was
shifted into position 2, replacing the primary amino group. The
resulting sulfide 60 was more potent and efficacious than 64. On
the other hand, introduction of the sterically more demanding,
electron-withdrawing bromine atom in position 5 (65) abolished
the KV7 opening properties completely (EC50 >10 µM). Similar to
1 and 2, the replacement of the pyridine for a phenyl ring (77)
increased the potency about eight fold, while the introduction of a
second nitrogen into the ring resulted in a moderately active
pyrimidine analogue (69).
Finally, we determined the KV7 channel opening properties of the
sulfoxides and sulfones to obtain information about the biological
activity of the putative metabolites of the sulfide derivatives. None
of the compounds showed channel opening activity at
concentrations up to 10 µM (table 3).
Computations
Until now there are no reports of X-ray crystal structures of the
KV7 family of ion channels. Nevertheless, we were interested in
visualizing the binding site to help rationalize probable
interactions of our compounds with binding site amino acids.
Towards this goal, homology modeling with the structure of the
crystallized KV1.2/2.1 channel and the sequences of KV7.2 and
KV7.3 was used as described in earlier publications to obtain a
heterotetramer (see figure 6).[36]
In addition to improved activity, it was important to also determine
the toxicity of the compounds because it was due to toxicity that
the two drugs 1 and 2 were removed from the market. For this
purpose, in vitro cell-based assays (MTT assay) with two cell lines
of hepatic origin, the TAMH or HEP-G2 cells, were used. Both cell
lines have been used in the past as in vitro tools for the prediction
of liver toxicity.[35] The poor water solubility of some of the
compounds turned out to be a problem, thus the rather high
concentrations needed for determination of less pronounced toxic
effects could often not be achieved. We therefore decided to
express the toxicity as LD25 values (concentration which reduced
cell viability to 75%) instead of the more commonly used LD50.
The toxicity data is summarized in tables 2 and 3. A general trend
on the impact of replacing the secondary amine by sulfur could
not be found. While 1 showed a LD25 of about 100 µM in both cell
lines, little decrease in cell viability by the sulfur analogue 36 was
In accordance with previously published modelled binding sites,
our model is of a rather hydrophobic character with leucine and
phenylalanine being the predominant species. Especially the 4-
fluorobenzyl moiety and the carbamate residue are pointing
towards extended hydrophobic pockets. On the other hand, the
space around the sulfide is quite limited, as is the cavity where
the pyridine core is located. Our modeling approach revealed a
hydrogen bond between the primary amino group of 36 and L299
of KV7.2. Furthermore, hydrogen bonds between the hydroxyl
group of S303 and the primary amine as well as the carbonyl
moiety of 36 are predicted.
In order to gain a better understanding as to where in the molecule
oxidation takes place, we decided to calculate the highest
occupied molecular orbitals (HOMO) of 1, 36 and 47. The
electrons of the HOMO are the energetically least favourable ones
8
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
Figure 6. Predicted orientations of 36 (cyan) and 48 (brown) in the proposed binding site. For a better comparison the ligand interaction diagrams of both structures
are given in the right illustration.
of a molecule and thus are more likely to be lost during oxidation.
As shown in figure 7, the HOMO of 1 is located at the pyridine ring,
which underlines the hypothesis that azaquinone diimines are
formed upon oxidation of 1. These electronic features change
significantly through the exchange of the secondary amine, as
shown by the calculated HOMOs of 36. The lone pairs of the
sulfide are estimated to form the HOMO, whereas the π-electrons
of the pyridine ring are calculated to constitute the fourth highest
occupied molecular orbital (HOMO−3). Thus, it would be
predicted that 36 undergoes sulfur oxidation to form sulfoxides
and sulfones rather than ring oxidation, which is well in line with
the products we obtained by oxidation using m-CPBA. Similar
results were found when the HOMOs of the structurally more
divergent compound 47 were calculated. The electrons of the
sulfide lone pairs are forming the HOMO, while the HOMO-1 is
located at the carbonyl oxygen. These calculations predict that
the pyridine moiety should be less susceptible to oxidation, as the
corresponding π-electrons are forming the HOMO-2, which is
energetically much more favourable (−6.88 eV vs. −6.21 eV).
Discussion
Our results show that the secondary amino group of flupirtine (1)
could be replaced by a sulfide bridge without losing the KV7
Figure 7. Left: Comparison of the five highest occupied molecular orbitals of 1 and two of its sulfide analogues (36 and 47). Right: The calculated lowest unoccupied
molecular orbital (LUMO) and HOMOs of 47.
9
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
channel opening properties. While the structure of the sulfide
residue had some impact on the potency of the compounds,
altering the carbamate moiety was found to have a greater
influence on activity. The replacement of the NH by a sulfur atom
also strongly modified the oxidative behavior, making the
compounds more stable, thus higher voltages were needed for
oxidation in voltammetry experiments. The loss of oxidizability
made us expect a loss of activity as well, as a correlation between
oxidation potential and KV7.2/3 opening activity was published
earlier.[15] However, this was not the case. While the inactivity of
the sulfoxides and sulfones was well in line with the hypothesis,
as they were difficult to oxidize, all of the sulfides showed anodic
peak potentials above 400 mV but still the majority of them were
active in opening the KV7.2/3 channel. The most potent
compounds were not even oxidizable, showing half maximal
current amplification at low nanomolar concentrations. On the
other hand, structures with low Epa (e.g. 51, 53 and 65) failed to
modulate the potassium channel. Therefore, a simple relationship
between oxidizability and activity could not be established. In
search for other criteria for activity, we decided to determine the
lipophilicity of the compounds at pH 7.4 (equals the pH of the
activity determination), since extended hydrophobic amide chains
resulted in high potencies. When comparing the LogD7.4 values
obtained by HPLC with the activity data, we again found no simple
correlation. While four of the most potent compounds (46, 47, 48
and 56) also were the ones with the highest LogD7.4, another
compound, 64, which is also quite hydrophobic, did not open
KV7.2/3 channels. On the other hand, the least lipophilic
compounds 51 and 53 were inactive, as could have been
expected. Nevertheless, compound 77 had the third lowest
LogD7.4 among the sulfides and still showed potent channel
modulation.
To gain further insight into the binding of the compounds we
attempted to visualize the binding site via homology modeling. In
our model, the 4-fluorobenzyl moiety of 36 stretches into a
hydrophobic pocket composed of leucine and phenylalanine (see
figure 6). This is consistent with the experimental data that
confirmed that the exchange of the benzyl motif for the more
hydrophilic pyridyl of piperidyl scaffolds abolished KV7 channel
opening activity, as seen in the case of 51 and 52. On the other
hand, introduction of lipophilic substituents (see 43, 44 and 54)
increased the potency of the compounds, highlighting the
importance of hydrophobic interactions within the binding pocket.
When replacing the ethoxy group of the carbamate by different
benzyl moieties, resulting in amide derivatives, higher potencies
were observed as well (see 40 and 59). Also, the carbamate
extends into a hydrophobic cleft in our model. Replacement of the
primary amino group by bulky groups increased the potency of
the compounds dramatically as shows the comparison between
48 and 49 or 36 and 55. When using bulky amides and replacing
the benzylic sulfide moiety by small aliphatic residues, highly
potent compounds were still obtained (46, 47, 48 and 50). The
docking simulations of those compounds exhibited a different
binding mode as for the carbamate derivatives (e.g. 36 vs. 47, see
figure 6). The most stable pose was found to be flipped, with the
amide chain pointing towards L314 and the small sulfide chain
being positioned in the bigger cleft near F254. The compounds of
the present work were not only designed to achieve high
potencies, but also to abolish adverse drug reactions of 1 and 2.
It could be shown by calculations and by chemical oxidation, that
the formation of quinone diimines should be limited to a minimum
by replacing the secondary amine for a sulfide. Thus, it would be
expected that the hepatotoxicity and the formation of blue
discoloration should be reduced relative to flupirtine and
retigabine, while the pharmacological activity of the new
compounds should be retained or even increased. Future studies
will focus on the in vivo effects of the new lead compounds in
animal models.
Conclusions
The synthesis and characterization of novel sulfide analogues of
flupirtine (1) and retigabine (2) yielded highly active openers of
KV7.2/3 heterotetramers showing high efficacy and potency in
single digit nanomolar concentrations. The replacement of NH by
sulfur increased the resistance to oxidation, thus potentially
suppressing the in vivo toxification, especially since the putative
sulfoxide und sulfone metabolites were found to be less toxic than
the parent compounds. Since formation of quinone diimines is
highly unlikely with the new compounds, and the activity/toxicity
ratio could be improved considerably, sulfide derivatives of 1 and
2 might prove valuable templates for the design of safe
replacements for the withdrawn drugs.
Experimental Section
Synthesis: Detailed information for the synthetic procedures for
compound 7-10, 12-60, 62-65, 67-69, 71-93, 95, and 96 including melting
points, NMR and LC-MS data can be found in the Supporting Information.
All commercial reagents and solvents were used without further
purification. Reaction control via TLC was performed on silica gel 60 F254
aluminium plates from Merck by using UV-light (λ = 254 nm and 366 nm)
for visualization. NMR spectra were recorded with a Bruker Avance III
instrument at 400 MHz (1H-NMR) and 101 MHz (13C-NMR), respectively.
The samples were measured in DMSO-d6 solution (unless stated
otherwise) with tetramethylsilane as an internal standard at 25 °C. If in
doubt, two-dimensional NMR measurements (HSQC, HMBC) were
applied to assign signals to the corresponding atoms. HRMS were
obtained with an LC-IT-TOF (Shimadzu) by using electrospray ionization.
IR spectra were recorded on an Alpha FT-IR spectrometer from Bruker
with the help of ATR technology (diamond probe head). Melting points
were measured automatically with the Melting Point M-565 instrument
(Büchi). The compound purities were determined by HPLC with DAD
(applying the 100% method at 220 nm). Flash chromatography was
performed (unless stated otherwise) on a glass column using silica gel 60
from Carl Roth with a particle size of 20−45 µm. The flow rate was adjusted
to approximately 25 mL/min by applying pressure with compressed air.
Preparative HPLC was performed using a preparative chromatography
system from Shimadzu and a LiChrospher 100 RP-18 endcapped
(250×25 mm) column from Merck.
Selected synthetic procedures for ligands 36 and 48: Compound 16
(5 mmol, 1.4 g, see Supporting Information) was suspended in water
(30 mL) and 2-propanol (50 mL). Iron powder (3 eq, 15 mmol, 0.84 g) and
NH4Cl (4.5 eq, 22.5 mmol, 1.2 g) were added and the suspension refluxed
for 5 hours. The resulting dark brown mixture was filtered and evaporated
to dryness. The residue was diluted with water (150 mL) and extracted with
ethyl acetate (3×60 mL). The combined organic layers were washed with
brine (150 mL) and dried over Na2SO4. Et3N (1.4 eq, 7 mmol, 0.97 mL)
was added and the solution cooled to 0 °C.
A solution of ethyl
chloroformate (1.2 eq, 6 mmol, 0.57 mL) in ethyl acetate (5 mL) was
added dropwise. After 2 hours additional ethyl chloroformate (1.2 eq,
6 mmol, 0.57 mL) was added at rt, 5 hours later, ethyl chloroformate
(0.6 eq, 3 mmol, 0,29 mL) and Et3N (0.7 eq, 3.5 mmol, 0.49 mL) were
10
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
added. After 2 hours the reaction was completed and the solution washed
with water (200 mL) and brine (200 mL). After drying with Na2SO4 and
evaporation to dryness the resulting solid was recrystallized from
dichloromethane. The product 36 was dried at 40 °C under reduced
pressure. For the synthesis of 48, compound 23 (1 mmol, 0.28 g, see
Supporting Information) was suspended in THF (10 mL) and one spatula
of Raney nickel was added. The mixture was stirred under hydrogen
atmosphere for 1 hour at rt. Subsequently, triethylamine (1.4 eq, 1.4 mmol,
0.2 mL) as well as a solution of (3,5-difluorophenyl)propanoyl chloride
(1.0 eq, 1.0 mmol, 0.20 g) in dichloromethane (1.1 mL) were added
dropwise and the suspension stirred for 30 min at rt. The catalyst was
filtered off and the filtrate diluted with water (100 mL). The mixture was
extracted with diethyl ether (100 mL) and washed with water (100 mL), sat.
solution of NaHCO3 (100 mL) as well as brine (100 mL). The organic
phase was evaporated to dryness and the resulting solid recrystallized
from ethanol/water (9:1) to yield compound 48.
10 µg/mL gentamicin sulfate (PAN Biotech). Cell number of suspension
was evaluated using EVE™ Automated Cell Counter (NanoEnTek) and
20,000 cells/well/200 µL DMEM/F12 media were seeded into 96 well
plates (#83.3924.300, Sarstedt). Plates were incubated until cells became
confluent (usually 48 h after seeding). Medium was removed and replaced
by 200 µL/well DMEM/F12 media supplemented with 5% PANEXIN NTA
(PAN Biotech), 10 mmol/L nicotinamide (Sigma Aldrich) and 10 µg/mL
gentamicin sulfate (PAN Biotech) and containing a serial dilution series of
test compounds (5-9 concentrations) prepared in DMSO (BioScience-
grade) (Carl Roth). In control wells 1% DMSO was present. Wells without
cells were used to determine background signal and were treated as
control wells. Every respective condition was run in triplicate. After 24 h
exposure to test compounds, medium was aspirated off, 125 µL
DMEM/F12 media supplemented with 20% (V/V) of 2.5 mg/mL solution of
3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) (Alfa
Aesar) in phosphate-buffered saline (PBS) was added to wells and plates
were incubated 2.5-4 h in a humidified incubator at 37 °C. Again, medium
was removed from wells and generated formazan crystals were dissolved
in 50 µL DMSO (for synthesis) (Carl Roth). Absorbance was measured at
λ = 570 nm with SpectraMax 190 microplate reader (Molecular Devices,
San Jose, USA). Cell viability was obtained by dividing absorbance values
for each treatment condition through that measured for control wells
expressed in % (both corrected for background before). LD25 of a
compound represents its concentration where cell viability attains 25%.
This value was determined by plotting (log)concentration of test compound
versus related cell viability and performing linear regression analysis in
Microsoft Excel (2013). When even the highest tested concentration of a
compound could not reduce cell viability to/below 25%, LD25 was indicated
as greater than the highest tested concentration. Determined values are
the mean of at least three independent experiments ± standard deviation
(SD).
Cyclovoltammetric analyses: Oxidation potentials of the compounds
were obtained by using a 797 VA Computrace device from Metrohm.
Working electrode: glassy carbon, reference electrode: Ag/AgCl/KCl 3 M,
auxiliary electrode: Pt. The compound (10 µmol) was suspended in 10 mL
of TRIS buffer (100 mmol/L, pH 7.4) and sonicated for 3 min. The mixture
was poured into the glass sample holder and degassed by nitrogen for 3
min at room temperature. Starting at –0.5 V the voltage was increased to
1.0 V with a speed of 0.1 V/s, while the current was recorded in 0.05 V
steps. Once 1.0 V was reached, the voltage was inverted until –0.5 V with
the same speed to obtain a full cycle. Five of those cycles were recorded
but only the first sweep was used to determine the anodic peak potential
(Epa) as it gave the strongest signal. The consecutive cycles showed lower
currents, most likely because of polymerization of the oxidized compounds
on the surface of the working electrode. Thus, the graphite surface had to
be polished after every measurement with an alumina oxide suspension in
water which was distributed on a patch of cloth. The Epa was determined
automatically using the software: 797 VA Computrace 1.3.1.
The HEP-G2 cells were detached from flask by Trypsin-EDTA (Sigma
Aldrich) when ~70% confluent and suspended in RPMI 1640. Cells were
counted with an EVE™ Automated Cell Counter (NanoEntek) and 15000
cells/well/100 μL RPMI 1640 media were seeded into 96 well plates
(Sarstedt). Cells were allowed to attach for 24 h in a humidified incubator
at 37 °C. Test compounds and controls were then added to the wells of
the plates as previously described for TAMH cells.
Lipophilicity determination: The lipophilicity of the compounds at pH 7.4
was determined by using an RP-HPLC method. First, a calibration curve
was established by plotting the log(retention factor) (logk′) of several
references against their literature logP. With the help of the resulting
logarithmic equation, the LogD7.4 values of our compounds could be
calculated after determination of their logk′. More details are given in the
Supporting Information.
HEK293 cells were detached from flask by TrypLE Express (Thermo
Fisher Scientific) when ~80% confluent and resuspended in MEM media.
After centrifugation the cell pellet was again resuspended in MEM media.
Cell number of suspension was counted with EVE™ Automated Cell
Counter and 60,000 cells/well/100 µL MEM media were seeded into black
walled 96 well plates with clear bottom (4titude). Cells were allowed to
attach for 24 h in a humidified incubator at 37 °C and formed a near
confluent monolayer. FLIPR® Potassium Assay Kit (Molecular Devices)
was used to determine KV7.2/3 channel opening activity of test compounds
according to manufacturer's instructions. Briefly, equal amount of Loading
Buffer containing 5 mmol/L probenecid (Santa Cruz Biotechnology) (pH
7.4) was added to wells and incubated for 1 h at room temperature in the
dark. Serial dilution series of test compounds (5-9 concentrations) were
prepared in DMSO (BioScience-grade) (Carl Roth) and incubated with
cells for another 30 min at room temperature in the dark. Control wells
contained 1% DMSO. Plates were placed in an Infinite® F200 Pro plate
reader (Tecan) and background fluorescence (baseline) for each well was
bottom read at excitation/emission wavelength: 485nm/535 nm for
approximately 20 s. Then, 50 µL Stimulus Buffer (c(K+) = 25 mmol/L, c(Tl+)
= 15 mmol/L) was added to each well and fluorescence intensity was
recorded immediately every 2.5-3 s continuously for further 2 min. The
recorded data were transferred to Microsoft Excel (2013). The ratio
readout of maximal fluorescence intensity change of a compound's
concentration at a specific time point was divided through the average of
fluorescence intensities of baseline (ΔF/F) and analogous determined
value for control was subtracted (corr. ΔF/F). At the chosen specific time
point the sum of ΔF/F of all concentrations belonging to a dilution series of
In vitro toxicity: TGF-α transgenic mouse hepatocytes (TAMH) were a
gift from Prof. Sidney D. Nelson (formerly Department of Medicinal
Chemistry, University of Washington, Seattle, USA). HEK293 cells
expressing Kv7.2/3 were purchased from SB Drug Discovery (Glasgow,
UK). TAMH cells were maintained in a T75 flask with Dulbecco's Modified
Eagle's Medium/Nutrient Mixture F-12 (DMEM/F12 1:1 mix) (PAN Biotech),
supplemented with 5% PANEXIN NTA (PAN Biotech), 10 mmol/L
nicotinamide (Sigma Aldrich) and 10 µg/mL gentamicin sulfate (PAN
Biotech). HEK293 cells were cultured in a T75 flask with minimum
essential medium with non-essential amino acids (MEM) (Thermo Fisher
Scientific), supplemented with 10% heat-inactivated fetal bovine serum
(FBS) (Sigma Aldrich), 2 mmol/L L-glutamine (Sigma Aldrich), 4 µg/mL-1
blasticidin S-HCl (Merck), 1% penicillin/streptomycin (Thermo Fisher
Scientific) and 0.78 mg/mL G418 sulfate (Carl Roth). Both cell lines were
incubated at 37 °C in a humidified incubator with 95% air/5% CO2. HEP-
G2 cell line was cultured in T75 flask (Sarstedt) with RPMI 1640 (PAN
Biotech), supplemented with 10% heat-inactivated fetal bovine serum
(FBS) (Sigma Aldrich) and 1% penicillin/streptomycin (PANBiotech) and
were incubated at 37 °C in a humidified incubator with 95% air/5% CO2.
TAMH cells were detached from flask when ~90% confluent with Trypsin-
EDTA (Sigma Aldrich), resuspended with 0.25 mg/mL soybean trypsin
inhibitor (Sigma Aldrich) in DMEM/F12. After centrifugation the cell pellet
was resuspended in DMEM/F12 media supplemented with5% PANEXIN
NTA (PAN Biotech), 10 mmol/L nicotinamide (Sigma Aldrich) and
a
compound became maximal. Corr. ΔF/F were plotted versus
log(concentration) of a compound, fitted to a four parameter logistic
11
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
equation and EC50 values were calculated thereof with GraphPad Prism 6
(La Jolla, California, USA). Indicated EC50 values are the mean of at least
three independent experiments ± standard deviation (SD).
ce_Risk_Assessment_Committee/WC500146103.pdf; b) European
Medicines Agency, “Withdrawal of pain medicine flupirtine endorsed”,
can
be
found
under
http://www.ema.europa.eu/docs/en_GB/document_library/Press_releas
e/2018/03/WC500246353.pdf.
Molecular modelling: Molecular modelling studies were performed with
the Molecular Operating Environment (MOE) software suite (version
2018.01).[37] The default parameters were used if not explicitly specified.
The sequences of the voltage-gated potassium channels KV7.2 and KV7.3
were taken from the UniProt entries O43526 and O43525. Both were
aligned to the KV1.2 template (PDB 2R9R) in an alternating manner. Only
the first 351 amino acids, containing the membrane-associated S1-S6
subunits, were taken into account. All four, pore-forming chains were
modelled at once to prevent unusual sidechain positions. A total of 10 main
chain models with 10 sidechain models each were calculated at a
temperature of 300 K. The final homology model was protonated with
Protonate3D and energy-optimized. The binding site was defined as all
residues within 4.5 Å of residue W265, which was previously described as
an important amino acid for the binding of retigabine. AMBER force field
parameters and AM1-BCC charges were assigned to all ligands prior to
molecular docking. The placement of 30 poses was performed with the
Triangle Matcher with London dG scoring and flexible ligands. All
generated binding poses were refined with an Induced Fit and MM/GBVI
scoring. A total of 10 final poses for each ligand was calculated, visually
inspected and cross-checked with additional 20 compounds from literature
and patents using the same procedure and the obtained retigabine binding
mode as template. The ligands were obtained as follows:
[3]
[4]
[5]
[6]
F. Klinger, M. Bajric, I. Salzer, M. M. Dorostkar, D. Khan, D. D. Pollak, H.
Kubista, S. Boehm, X. Koenig, Br. J. Pharmacol. 2015, 172, 4946–4958.
F. Klinger, P. Geier, M. M. Dorostkar, G. K. Chandaka, A. Yousuf, I.
Salzer, H. Kubista, S. Boehm, Br. J. Pharmacol. 2012, 166, 1631–1642.
W. Dalby-Brown, H. Hansen, M. Korsgaard, N. Mirza, S.-P. Olesen, Curr.
Top. Med. Chem. 2006, 6, 999–1023.
a) M. A. Überall, G. H.H. Mueller-Schwefe, B. Terhaag, Curr. Med. Res.
Opin. 2012, 28, 1617–1634; b) H. Lee, V. de Vito, M. Giorgi, H. Yun, Eur.
J. Pharmacol. 2015, 762, 350–356.
[7]
a) T. Zhang, M. S. Todorovic, J. Williamson, J. Kapur, Ann. Clin. Trans.
Neurol. 2017, 4, 888–896; b) D. Sampath, R. Valdez, A. M. White, Y. H.
Raol, Neuropharmacology 2017, 123, 126–135.
[8]
[9]
C. Bock, A. Link, Future Med. Chem. 2019, 11, 337–355.
E. Scheuch, K. Methling, P. J. Bednarski, S. Oswald, W. Siegmund, J.
Pharm. Biomed. Anal. 2015, 102, 377–385.
[10] K. Methling, P. Reszka, M. Lalk, O. Vrana, E. Scheuch, W. Siegmund, B.
Terhaag, P. J. Bednarski, Drug Metab. Dispos. 2009, 37, 479–493.
[11] S. Clark, A. Antell, K. Kaufman, Ther. Adv. Drug Saf. 2015, 6, 15–19.
[12] M. R. Groseclose, S. Castellino, Chem. Res. Toxicol. 2019, DOI
10.1021/acs.chemrestox.8b00313.
[13] A. S. Surur, C. Bock, K. Beirow, K. Wurm, L. Schulig, M. K. Kindermann,
A. Bodtke, W. Siegmund, P. J. Bednarski, A. Link, Org. Biomol. Chem.
2019, submitted.
Conformational Analysis: Gaussian ’09 program package, Revision
E.01 and Gaussian 16, Revision A.03 were used to perform conformational
analysis and molecular geometry computations.[38] The B3LYP hybrid
exchange-correlation function along with the Pople’s 6-311++G (d,p) or 6-
311++G (3df,2p) split-valence basis set was used. The calculations took
place at a high level with respect to the basis set of the ab initio calculation
and were accordingly time-consuming. The important molecular orbitals of
1, 36 and 47 were calculated from their corresponding most stable
conformers. The B3LYP hybrid exchange-correlation function along with
the Pople’s 6-311++G(d,p) was employed. For better clarity and
comparability, the electron ratios were calculated after assembly of natural
bonding orbital (NBO) model. In most cases, the standard Hartree-Fock
(HF) method resulted in electron density distributed over the entire
molecule, assignment to a specific region of a molecule was accordingly
difficult. The electron ratios were converted with the help of NBO method
into a vivid orbital, natural bonding orbital (NBO), which made it possible
to assign orbitals to individual structural elements.
[14] S. V. Mathias, B. W. Abou-Khalil, Epilepsy Behav. Case Rep. 2017, 7,
61–63.
[15] C. J. Lemmerhirt, M. Rombach, A. Bodtke, P. J. Bednarski, A. Link,
ChemMedChem 2015, 10, 368–379.
[16] a) C. Bock, K. Beirow, A. S. Surur, L. Schulig, A. Bodtke, P. J. Bednarski,
A. Link, Org. Biomol. Chem. 2018, 16, 8695–8699; b) A. S. Surur, K.
Beirow, C. Bock, L. Schulig, M. K. Kindermann, A. Bodtke, W. Siegmund,
P. J. Bednarski, A. Link, ChemistryOpen 2019, 8, 41–44.
[17] K. R. Romines, G. A. Freeman, L. T. Schaller, J. R. Cowan, S. S.
Gonzales, J. H. Tidwell, C. W. Andrews, D. K. Stammers, R. J. Hazen,
R. G. Ferris et al., J. Med. Chem. 2006, 49, 727–739.
[18] H. Guo, Q. Xu, O. Kwon, J. Am. Chem. Soc. 2009, 131, 6318–6319.
[19] V. A. Artemov, V. L. Ivanov, A. V. Koshkarov, A. M. Shestopalov, V. P.
Litvinov, Chem. Heterocycl. Compd. 1998, 34, 96–101.
[20] R. Csuk, S. Sommerwerk, J. Wiese, C. Wagner, B. Siewert, R. Kluge, D.
Ströhl, Z. Naturforsch. B 2012, 67, 1297–1304.
[21] N. Whittaker, T. S. G. Jones, J. Chem. Soc. 1951, 1565–1570.
[22] M. Schlosser, A. Ginanneschi, F. Leroux, Eur. J. Org. Chem. 2006, 2006,
2956–2969.
[23] Y. Liu, J. Kim, H. Seo, S. Park, J. Chae, Adv. Synth. Catal. 2015, 357,
2205–2212.
Acknowledgements
[24] S. V. Fedoseev, M. Y. Belikov, O. V. Ershov, V. A. Tafeenko, Russ. J.
Org. Chem. 2016, 52, 1784–1787.
AL and PJB are recipients of grants DFG LI 765/7-1 and DFG BE
1287/6-1 by the Deutsche Forschungsgemeinschaft. The supply
of computing resources by the HPC computational centre,
University of Rostock is gratefully acknowledged. We thank Ms.
Anne Schüttler, Ms. Maria Hühr and Mr. Michael Eccius for
excellent technical assistance.
[25] R. Hofstetter, M. Hasan, G. M. Fassauer, C. Bock, A. S. Surur, C. W.
Grathwol, S. Behnisch, F. Potlitz, T. Oergel, W. Siegmund et al., J.
Chrom. A 2019, submitted.
[26] a) H. C. Rawden, G. O. Kokwaro, S. A. Ward, G. Edwards, Br. J. Clin.
Pharmacol. 2001, 49, 313–322; b) A. S. Surur, L. Schulig, A. Link, Arch.
Pharm. Chem. Life Sci. 2018, 352: e1800248.
[27] R. Hofstetter, G. M. Fassauer, A. Link, J. Chrom. B 2018, 1076, 77–83.
[28] U. Lücking, Angew. Chem. Int. Ed. 2013, 52, 9399–9408; Angew. Chem.
2013, 125, 9570–9580.
Keywords: flupirtine • retigabine • KCNQ • KV7 • sulfide
References
[29] M. Zenzola, R. Doran, L. Degennaro, R. Luisi, J. A. Bull, Angew. Chem.
Int. Ed. 2016, 55, 7203–7207; Angew. Chem. 2016, 128, 7319-7323.
[30] Y. Xie, B. Zhou, S. Zhou, S. Zhou, W. Wei, J. Liu, Y. Zhan, D. Cheng, M.
Chen, Y. Li et al., ChemistrySelect 2017, 2, 1620–1624.
[31] C. D. Weaver, D. Harden, S. I. Dworetzky, B. Robertson, R. J. Knox, J.
Biomol. Screen. 2004, 9, 671–677.
[1]
[2]
U. Schwabe, D. Paffrath, W.-D. Ludwig, J. Klauber, Arzneiverordnungs-
Report 2018, Springer, Berlin, Heidelberg, 2018.
a) European Medicines Agency, “Assessment report for flupirtine
containing
medicinal
products”,
can
be
found
under
[32] M. Kumar, N. Reed, R. Liu, E. Aizenman, P. Wipf, T. Tzounopoulos, Mol.
Pharmacol. 2016, 89, 667–677.
http://www.ema.europa.eu/docs/en_GB/document_library/Referrals_do
cument/Flupirtine-
containing_medicines/Recommendation_provided_by_Pharmacovigilan
12
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
[33] J. E. Davoren, M. M. Claffey, S. L. Snow, M. R. Reese, G. Arora, C. R.
Butler, B. P. Boscoe, L. Chenard, S. L. DeNinno, S. E. Drozda et al.,
Bioorg. Med. Chem. Lett. 2015, 25, 4941–4944.
[34] W. Dalby-Brown, C. Jessen, C. Hougaard, M. L. Jensen, T. A. Jacobsen,
K. S. Nielsen, H. K. Erichsen, M. Grunnet, P. K. Ahring, P.
Christophersen et al., Eur. J. Pharmacol. 2013, 709, 52–63.
[35] a) M. Davis, B. D. Stamper, Biomed. Res. Int. 2016, 2016, 4780872; b)
A. van Summeren, J. Renes, F. G. Bouwman, J.-P. Noben, J. H. M. van
Delft, J. C. S. Kleinjans, E. C. M. Mariman, Toxicol. Sci. 2011, 120, 109–
122.
[36] a) R. Syeda, J. S. Santos, M. Montal, J. Biol. Chem. 2016, 291, 2931–
2937; b) V. Barrese, F. Miceli, M. V. Soldovieri, P. Ambrosino, F. A.
Iannotti, M. R. Cilio, M. Taglialatela Clin. Pharmacol. 2010, 2, 225–236.
[37] Molecular Operating Environment (MOE), 2018.01; Chemical Computing
Group ULC, 1010 Sherbrooke St. West, Suite #910, Montreal, QC,
Canada, H3A 2R7, 2018.
[38] Gaussian 09, Revision E.01, M. J. Frisch, G. W. Trucks, H. B. Schlegel,
G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone,
B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P.
Hratchian, A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M.
Hada, M. Ehara, et al., Gaussian, Inc., Wallingford CT, 2009.
13
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201900112
ChemMedChem
FULL PAPER
Entry for the Table of Contents
O
O
O
O
N
NH
NH2
N
N
NH2
S
N
O
F
F
Road to
Safe KV7.2/3
Openers
O
O
NH
X
N
NH2
F
The road to hell is paved with oxidation. Flupirtine may be innocuous by itself, but suffers from oxidative biotransformation to toxic
metabolites, leaving its validated target (KV7 channels) vastly under-exploited. We present a library of sulfur-analogues that change
the track from toxic quinone imine metabolism in favour of benign sulfoxide formation. The rational of this approach is supported by
structure-activity relationships detailed here.
14
This article is protected by copyright. All rights reserved.