Characterization of HERG potassium channel inhibition using CoMSiA 3D QSAR and homology modeling approaches

Article abstract of DOI:10.1016/S0960-894X(03)00196-3

A data set consisting of twenty-two sertindole analogues and ten structurally diverse inhibitors, spanning a wide range in potency, was analyzed using CoMSiA. A homology model of HERG was constructed from the crystal structure of the open MthK potassium channel. A complementary relationship between our CoMSiA and homology models is apparent when the long inhibitor axis is oriented parallel to the longitudinal axis of the pore, with the tail region pointed toward the selectivity filter. The key elements of the pharmacophore, the CoMSiA and the homology model are: (1) The hydrophobic feature optimally consists of an aromatic group that is capable of engaging in π-stacking with a Phe656 side chain. Optionally, a second aromatic or hydrophobic group present in some inhibitors may contact an additional Phe656 side chain. (2) The basic nitrogen appears to undergo a π-cation interaction with Tyr652. (3) The pore diameter (12 A+), and depth of the selectivity loop relative to the intracellular opening, act as constraints on the conformation-dependent inhibitor dimensions.

Full text of DOI:10.1016/S0960-894X(03)00196-3

Bioorganic & Medicinal Chemistry Letters 13 (2003) 1829–1835
Characterization of HERG Potassium Channel Inhibition Using
CoMSiA 3D QSAR andHomology Modeling Approaches
Robert A. Pearlstein,a Roy J. Vaz,a,* Jiesheng Kang,a Xiao-Liang Chen,a
Maria Preobrazhenskaya,b Andrey E. Shchekotikhin,b Alexander M. Korolev,b
Ludmila N. Lysenkova,b Olga V. Miroshnikova,b James Hendrixa and David Rampea
aAventis Pharmaceuticals, 1041 Route 202/206 N, Bridgewater, NJ 08876, USA
bRussian Academy of Medical Sciences, Gause Institute of New Antibiotics, Bolshaya Pirogovskaya 11, Moscow 119867, Russia
Received 15 May 2002; accepted 31 October 2002
Abstract—A data set consisting of twenty-two sertindole analogues and ten structurally diverse inhibitors, spanning a wide range in
potency, was analyzed using CoMSiA. A homology model of HERG was constructed from the crystal structure of the open MthK
potassium channel. A complementary relationship between our CoMSiA and homology models is apparent when the long inhibitor
axis is oriented parallel to the longitudinal axis of the pore, with the tail region pointed toward the selectivity filter. The key ele-
ments of the pharmacophore, the CoMSiA and the homology model are: (1) The hydrophobic feature optimally consists of an
aromatic group that is capable of engaging in p-stacking with a Phe656 side chain. Optionally, a second aromatic or hydrophobic
group present in some inhibitors may contact an additional Phe656 side chain. (2) The basic nitrogen appears to undergo a p-cation
interaction with Tyr652. (3) The pore diameter (12 A+), and depth of the selectivity loop relative to the intracellular opening, act as
constraints on the conformation-dependent inhibitor dimensions.
# 2003 Elsevier Science Ltd. All rights reserved.
The HERG potassium channel (human ether-a-go-go-
related gene) is expressed in the human heart. The
channel is a key effector of cardiac repolarization and
contributes to the QT interval measured by the electro-
cardiogram. Inhibition of HERG can lead to a pro-
longation of the QT interval, widely considered a
critical risk factor for torsades de pointes arrhythmia.
Most compounds that prolong the QT interval do so via
a specific, high affinity blockade of HERG.1 Such che-
mically diverse drugs as terfenadine (Seldane1), cisa-
pride (Propulsid1), and astemizole (Hismanal1)2 have
been withdrawn from the market due to their HERG
blocking properties and attending QT interval pro-
longation. Thus, HERG inhibitory effects represent an
important safety consideration in drug discovery. The
prevalence of this problem reflects a significant prob-
ability of overlap between the HERG pharmacophore
and that for many primary target classes. Pharmacophore
and QSAR analysis studies have been hampered by the
lack of consistent, high quality data sets. Pharmacophore
models generated using Catalyst3 and CoMFA4 have
been reported in the recent literature. In Silico profiling
methods have also been proposed for generating biased
libraries enriched with non-HERG inhibitors.5
Sertindole, an antipsychotic drug exhibiting high affinity
HERG blockade,6 is a useful starting point for studying
HERG structure–activity relationships. The fluorophenyl–
piperidyl substituted indole scaffold of sertindole is con-
formationally constrained,7 making it an attractive tem-
plate for structural comparisons with other HERG
inhibitors. We probed the possible structural requirements
of HERG inhibition by synthesizing a series of sertindole
analogues in which structural features were systematically
omitted or changed. HERG IC50 values were deter-
mined for these analogues using patch clamp electro-
physiology. A comparative molecular similarity analysis
(CoMSiA)8 was performed for the sertindole series toge-
ther with a diverse set of known inhibitors (see Table 1).
Synthesis of the Sertindole Analogues
The synthesis schemes for the sertindole analogues listed
in Table 1 are reported in Schemes 1–3. Analogues 2, 3,
and 14 were prepared via the method of Perregaard et
*Corresponding author. E-mail: roy.vaz@aventis.com
0960-894X/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0960-894X(03)00196-3

Table 1. Structures and their hERG activities used in the analysis (*ref 15, **ref 16)
Compd
Structure
IC50 (nM)
3
Compd
Structure
IC50 (nM)
56
Compd
Structure
IC50 (nM)
36
Compd
Structure
IC50 (nM)
26,000
8
sertindole
terfenadine
16
1
2
3
4
5
6
7
88
10
fexofenadine
dofetilide
haloperidol
pimozide
cisapride
23,000
10
9
10
11
12
13
14
15
>10,000
6.2
17
18
19
20
21
22
23
1480
4550
7
32
460
1947
579
75,000
137
131
18
>10,000
23.5
15,700
2200
45
MK499
34*
143**
204
3500
verapamil
11
>10000

R. A. Pearlstein et al. / Bioorg. Med. Chem. Lett. 13 (2003) 1829–1835
1831
from these cells as described previously.12 All com-
pounds were added to the cells in ascending concentra-
tions, and IC50 values for HERG channel inhibition
were calculated.
HERG channel 3D-QSAR (CoMSiA) model
A CoMSiA model was constructed using IC50 values for
all but four of the 32 compounds shown in Table 1. Our
data set includes our sertindole series and several pub-
lished structurally diverse chemotypes, which together
span a wide range of HERG potency. The super-
imposed structures of representative compounds inclu-
ded in our CoMSiA model are shown in Figure 4. Four
compounds (1, 3, 7, and 21) representing the available
range in IC50 were withheld from the CoMSiA training
set as a means of validating the prediction accuracy of
our model.
Electrostatic calculations have shown that monovalent
cations are stabilized within the pore of the KcsA
potassium channel,13 and many potassium channels
including HERG are blocked by quaternary alkyl-
amines.14 Electrophysiological evidence suggests that
inhibitors bind within the intracellular lumen of the
potassium conduction pore,15 which is believed to con-
tain a large water-filled cavity. Although important dif-
ferences16 between HERG and other potassium
channels likely exist, it is reasonable to assume that a
preference for cationic species is shared among all
potassium channels. Therefore, our conformational
energy and CoMSiA calculations were performed using
the protonated form of the basic nitrogen, where pre-
sent. The active conformations of sertindole and MK-
499 were reconstructed based on those described pre-
viously.7,15 Conformational searching was performed
on all other inhibitors using the MMFF94 force field in
vacuum, as implemented in Sybyl 6.7.2 (available from
Tripos, Inc., St. Louis, MO). The candidate active con-
formations were selected from among those within
5 kcal of the lowest minimum energy conformation. The
final chosen conformation was that which gave the best
correlation in the CoMSiA model. This is particularly
evident in the case of terfenadine and its metabolite,
fexofenadine.
Scheme 1. Compounds 1, 3–8, 10, and 13.
Scheme 2. Compounds 9 and 16–23.
Two minimum energy conformations are predicted for
terfenadine. These are a linear conformation denoted as
conf-1 and an intra-molecular H-bond stabilized con-
formation denoted as conf-2. Conf-2 is the lowest
energy conformation, and the only one observed in the
crystal structure of the isolated compound (data on
file at Aventis Pharmaceuticals, Inc.). Three minimum
energy conformations are predicted for fexofenadine
(Fig. 1). Two are similar to conf-1 (yellow) and conf-2
(magenta) predicted for terfenadine. The third
conformation denoted as conf-3 (cyan) consists of
the folded form, which is stabilized by a ‘salt bridge’
interaction between its carboxylate oxygen and pro-
tonated piperdyl nitrogen. This is the lowest
energy conformation for fexofenadine, and is observed
together with other conformations in the crystal struc-
ture of the isolated compound (data on file at Aventis
Scheme 3. Compound 12.
al.9 Compound 11 was prepared via the method of Carr
et al.10 Compound 15 was prepared via the method of
Andersen et al.11
HERG channel electrophysiology
Chinese hamster ovary cells stably expressing the
HERG cardiac potassium channel were used for these
experiments. Standard whole-cell patch clamp electro-
physiology was used to record HERG channel currents

1832
R. A. Pearlstein et al. / Bioorg. Med. Chem. Lett. 13 (2003) 1829–1835
used as landmarks for superimposing the compounds
onto sertindole using least-squares fitting. All com-
pounds in our data set contain at least two of these
landmarks. In cases where more than one superposition
mode is possible, all were tried and the possibility with
the higher q2 was retained in the model. Our CoMSiA
results are based on steric and electrostatic fields.
H-bond and hydrophobic CoMSiA fields, with or with-
out steric and electrostatic fields, did not result in
increased statistical significance of the model. This is
consistent with the non-orthogonality between the
hydrophobic/H-bond fields and the steric/electrostatic
fields.
The CoMSiA models are based on 470 grid points with
a 2 A separation and a minimum sigma of 2.0 kcal. Plots
of fitted versus actual pIC50 (-logIC50) are provided in
Figure 2. The statistics of the model are shown in Figure
3. The four compounds omitted from the training set
are shown as red points. The predicted values of these
compounds fall well within the standard errors of the
fitted model.
Figure 1. Conformations of fexofenadine tested in the CoMSiA
models (see text).
Pharmaceuticals, Inc.). This conformation is not acces-
sible to terfenadine, which lacks the carboxyl group.
The intra-pore environment is likely to influence inhi-
bitor conformational stability. However, there is no
obvious choice of conformational energy calculation
conditions (e.g., solvent versus vacuum, dielectric con-
stant) that best approximate this environment. Water
and ions present in the pore would likely destabilize the
intra-molecular hydrogen bond in conf-2. The carbox-
ylate group in conf-1 of fexofenadine likely requires an
inter-molecular counter-ion partner, which is unlikely to
exist based on the predicted hydrophobic side chain
composition of the pore lining (see the next section).
Environmental considerations are thus suggestive of
conf-1 and conf-3 as the active conformations of terfe-
nadine and fexofenadine, respectively. CoMSiA models
were constructed by incorporating different permuta-
tions of conf-1, conf-2, and conf-3 for each analogue.
The statistics for the models are listed in Table 2. The
model with the largest q2 (see the bold entry in Table 2)
is obtained for conf-1 and conf-3 of terfenadine and
fexofenadine, respectively. The non-folded fexofenadine
conformations (conf-1 or conf-2) result in models with
much lower statistical validity. Our CoMSiA results
provide indirect evidence that the intra-molecular salt
bridge-stabilized conformation (conf-3) persists under
biological conditions, and that the large difference in
HERG channel activity observed between the two ana-
logues can be largely explained in terms of conforma-
tional differences.
The negative steric contour shown in red in Figure 4
corresponds to a (S.D.) times (coeff.) (referred to sub-
sequently as coefficients) of À0.005. This contour level
encompasses the region of fexofenadine that differ-
entiates it from terfenadine. Substituents on the hydro-
phobic groups corresponding to the indole phenyl and
N-phenyl rings of sertindole are also covered by this
contour. The addition of steric bulk to these regions is
predicted to reduce HERG activity. The +0.005 steric
contour shown in green in Figure 4 encompasses the
entire central inhibitor region and that corresponding to
the N-phenyl ring of sertindole. The hydrophobic fea-
ture corresponding to the N-phenyl ring of sertindole is
present in a subset of inhibitors, providing an opportu-
nity for variance in this region to contribute to the
correlation. On the other hand, co-variance between the
fields and activities cannot be observed in the region
corresponding to the indole phenyl ring of sertindole
The centroids of the indole-phenyl and N-phenyl rings,
nitrogen atoms, and N-protons (where existing) were
Table 2. Values of q2 from various CoMSiA trials, using combina-
tions of conformations of terfenadine and fexofenadine. The combi-
nation with the best q2 was used in the final model
Conf of terf
Conf of fexo
q2
SDEP
Components
conf-1
conf-1
conf-1
conf-2
conf-2
conf-2
conf-1
conf-2
conf-3
conf-1
conf-2
conf-3
0.424
0.514
0.571
0.398
0.313
0.481
1.034
0.949
0.892
1.08
1.085
1.003
3
3
3
4
1
3
Figure 2. Fitted (blue) and predicted (red) pIC50 values for the studied
compounds. The predicted compounds (withheld during CoMSiA
model construction) fall well within the fitted set of compounds.

R. A. Pearlstein et al. / Bioorg. Med. Chem. Lett. 13 (2003) 1829–1835
1833
The structure–activity relationships of our sertindole
series suggest that the HERG binding site is highly tol-
erant to large changes in the parent compound. Sig-
nificant potency is retained in analogues lacking the
ethyl cyclic urea tail group (compound 14) or containing
a partial piperdyl ring (compound 15). Complete loss of
activity occurs for only three highly truncated analogues
(compounds 9, 12, and 23).
Structure-basedinterpretation of the CoMSiA model
Figure 3. Statistical parameters q2, r2, standard error of prediction
(SEP) and Standard Error of Estimate (SEE) from the PLS model.
There is ample evidence in the literature that HERG
blockade by small molecules is caused by direct occu-
pation of the pore.16À18 A crystal structure of the open
bacterial MthK potassium channel was recently pub-
lished.6 We wanted to explore whether this structure
could be used to interpret and refine our CoMSiA-based
pharmacophore model. This work was intended to
address the following key issues: (1) orientation of the
pharmacophore in the binding site; (2) protein partners
associated with the proposed pharmacophore features
and the nature of the interactions; (3) reported sensitiv-
ity differences of mutated forms of HERG to terfena-
dine, cisapride, and MK-499.
A homology model of the tetrameric pore region of
hERG (inner and pore helices and selectivity loop) was
constructed from the MthK template and the reported
HERG-MthK sequence alignment17 using Modeler
(available from Accelrys, San Diego, CA). The side
chains of the MthK pore domain were not resolved in
the crystal structure, and were therefore modeled ab
initio. Complete 4-fold symmetry in the resulting
homology model was not obtained due to a slight
asymmetry in one of the monomers of the MthK struc-
ture. This was particularly apparent for the side chain
conformations of Phe656 and Tyr652. The conforma-
tions of these side chains were therefore adjusted in our
model by copying all corresponding side chain torsion
angles from the opposite monomer. Due to the low
resolution of the MthK structure and the uncertain
structural relationship between the MthK and HERG
channels, our homology model should be considered
highly qualitative.
Figure 4. Superposition of the studied compounds showing the steric
coefficients (coefficient values=Æ0.005) from the CoMSiA. The con-
tours of coefficients corresponding to the red region suggest desirable
substitution sites aimed at decreasing HERG affinity.
due to the absence of inhibitors lacking this feature in
the data set. On the basis of the absence of such inhibi-
tors, it is reasonable to conclude that this hydrophobic
feature is required for activity and is implicitly present
in the CoMSiA model. The 0.02 electrostatic coefficient
contour level shown in orange in Figure 5 overlaps with
the basic nitrogen feature. A positive correlation
between HERG activity and ‘increased’ positive charge
in this region is predicted. The other positive coefficient
contour is due to the acid group on compounds 4 and 5,
which when neutralized by esterification in compounds
7 and 8 respectively, results in increased inhibitory
activity. The electrostatic coefficients contribute 76% to
the CoMSiA model.
The homology model suggests that the lumen of the
pore consists of two layers of aromatic side chains con-
tributed by Phe656 and Tyr652. We qualitatively
docked the set of aligned inhibitor structures into the
homology model. The best correspondences between
pharmacophore and pore features were achieved using
the following criteria: (1) orientation of the inhibitor
long axis parallel to the pore axis, such that the hydro-
phobic feature(s) are located at the intracellular mouth,
and the tail region points toward the selectivity loop.
This results in spatial registration between the hydro-
phobic and basic nitrogen features proposed from our
CoMSiA model (6–8 A separation) with Phe656 and
Tyr652 side chains, respectively. (2) Overall shape com-
plementarity between pore and inhibitor surfaces based
on graphical visualization using Molcad (available from
Tripos). The inhibitor dimensions are qualitatively con-
sistent with the pore diameter (12 A+) and depth of the
Figure 5. Superposition of the studied compounds showing the elec-
trostatic contours of coefficients (coefficient values=Æ0.02) from the
CoMSiA. Modifications aimed at increasing the negative charge in the
orange regions and the positive charge in the cyan region are predicted
to decrease HERG affinity.

1834
R. A. Pearlstein et al. / Bioorg. Med. Chem. Lett. 13 (2003) 1829–1835
selectivity loop relative to the intracellular opening
(spatial constraints).
The proposed binding mode suggests an extended
interpretation of our CoMSiA model: (1) the hydro-
phobic features optimally consist of an aromatic group
that is capable of engaging in p-stacking with a Phe656
side chain. Optionally, a second aromatic or hydro-
phobic group present in some inhibitors may contact a
different Phe656 side chain. (2) The basic nitrogen is
proposed to undergo a p-cation interaction19 with
Tyr652. The variation in basicity of this feature across
the studied inhibitors is suggestive of a p-cation versus
salt bridge interaction, which would preferentially
involve a tertiary nitrogen. Though not present in our
data set, inhibitors that contain a hydrophobic or aro-
matic group in this position could theoretically engage
Tyr652 in hydrophobic or p-stacking interactions in lieu
of a p-cation interaction. (3) The tail region occupies the
pore between Tyr652 and the selectivity filter in our
docking model. This region is structurally and con-
formationally diverse, and extends maximally to the
selectivity loop. The tail region may provide additional
interaction sites with the pore lining that are not
apparent from our general CoMSiA model. As such,
this region may enhance the overall occlusion of the
pore. Structure–activity relationships about the tail
region are difficult to assess due to its greater diversity
and flexibility compared to the other regions. Our ser-
tindole analogues were largely aimed at exploring the
structure–activity relationships about this region. Sur-
prisingly, complete removal of the piperdyl N-sub-
stituent results in significant retention of activity (3–
200 nM IC50). The importance of the tail region to
HERG binding affinity should be considered on a case-
by-case basis. We believe that this region is best
addressed via ‘local’ pharmacophore models derived for
individual series of interest incorporating variation in
the tail region.
Figure 6. Orthogonal views of terfenadine (yellow), cisapride (green),
and MK-499 (magenta) docked into the homology model of HERG.
MK-499 is predicted to approach closer to Val625 (in spacefill) which
showed decreased affinity for Val625Ala. Our model suggests a possi-
ble interaction with Val625 that may be unique among the three
compounds.
variable set of pharmacophore feature types having a
relatively constant, well-defined spatial relationship. For
example, the basicity of the nitrogen can vary within the
context of a p–cation interaction. It is conceivable that
an aromatic or hydrophobic group at this position
could also interact with Tyr652 via p–p or hydrophobic
mechanisms. Compounds 14 and 15 of our sertindole
series, which exhibit nanomolar potency in the absence
of the ethyl cyclic urea substituent, suggest that the tail
region can be less important than other features of the
pharmacophore.
A ‘drain plug’ picture of HERG channel inhibition
emerges from our modeling studies. The drain plug
shape is complementary to the cylindrical architecture
of the pore, with key interaction sites on the drain plug
‘handle’ [aromatic feature(s)] and along the length of the
plug (basic nitrogen and polar group). The depth of
penetration can vary, depending on the length of the tail
region, maximally extending to the intracellular end of
the selectivity filter. The scope of our model is surprising
in light of the complexity of the HERG channel. Possi-
bilities for additional binding modes not covered by our
model are entirely plausible given the complexity of the
protein, its state-dependent conformational changes,
and its interactions with the membrane and regulatory
proteins. Furthermore, such factors as membrane per-
meation required for accessing the intra-cellular lumen of
the channel and possible shifts in local pKa due to changes
in membrane potential further complicate the problem of
predicting HERG channel inhibitory activity. Never-
theless, it is hoped that our model can be used as a road-
map for explaining HERG channel liabilities in early lead
candidates as well as for designing out properties that
promote HERG channel binding in such compounds.
Phe656 and Tyr652 were shown via site-directed muta-
genesis to promote sensitivity of HERG to inhibition by
MK-499, terfenadine, and cisaparide.15,18 However,
only MK-499 blockade demonstrated sensitivity to
mutation of Val625. The alignment and conformations
predicted from our CoMSiA model suggest that the
methylsulfonamide group of MK-499 lies in a region
that is not occupied by terfenadine and cisapride. Fur-
thermore, the methylsulfonamide group is spatially
close to Val625 in the docked model (Fig. 6). The
Val625Ala mutation may eliminate a possible van der
Waals interaction with MK-499 that does not occur for
terfenadine or cisapride. Recent evidence suggests that
altering the relative positions of Phe656 and Tyr652 via
mutagenesis results in decreased sensitivity to cisa-
pride.18 This further supports our proposed CoMSiA-
homology model relationships.
References andNotes
HERG contains a promiscuous binding site for drug-
like molecules. This can be explained partly by the large
volume of the binding site and partly by the nature of
the pharmacophore. Our results suggest that diverse
inhibitor chemotypes can be described by a somewhat
1. Brown, A. M.; Rampe, D. Pharmaceutical News 2000, 7,
15.
2. (a) Roy, M.-L.; Dumaine, R.; Brown, A. M. Circulation
1996, 94, 817. (b) Rampe, D.; Roy, M.-L.; Dennis, A.; Brown,
A. M. FEBS Lett. 1997, 417, 28. (c) Zhou, Z.; Vorperian,

R. A. Pearlstein et al. / Bioorg. Med. Chem. Lett. 13 (2003) 1829–1835
1835
V. R.; Gong, Q.; Zhang, S.; January, C. Y. J. Cardiovasc.
Electrophysiol. 1999, 10, 836.
3. Ekins, S. L.; Crumb, W. J.; Sarazan, R. D.; Wikel, J. H.;
Wrighton, S. A. J. Pharmacol. Exp. Ther. 2002, 301, 427.
4. Cavalli, A.; Poluzzi, E.; De Ponti, F.; Racanatini, M. J.
Med. Chem. 2002, 45, 3844.
5. Roche, O.; Trube, G.; Zuegge, J.; Pflimlin, P.; Alanine, A.;
Schneider, G. ChemBioChem. 2002, 3, 455.
6. Rampe, D.; Murawsky, M. K.; Grau, J.; Lewis, E. W. J.
Pharmacol. Exp. Ther. 1998, 286, 788.
7. (a) Andersen, K.; Liljefors, T.; Hyttel, J.; Perregaard, J. J.
Med. Chem. 1996, 39, 3723. (b) Anderson, K.; Liljefors, T.;
Gundertofte, K.; Perregaard, J.; Bogeso, K. P. J. Med. Chem.
1994, 37, 950.
8. (a) Klebe, G.; Abraham, U.; Mietzner, T. J. Med. Chem.
1994, 37, 4130. (b) Cramer, R. D., III; Patterson, D. E.;
Bunce, J. D. J. Am. Chem. Soc. 1988, 110, 5959. (c) Cramer,
R. D. III; DePriest, S. A.; Patterson, D. E.; Hecht, P. In 3-D
QSAR in Drug Design; Kubinyi, H., Ed.; ESCOM, Leiden,
1993; pp 443. (d) Klebe, G. In 3-D QSAR in Drug Design;
Kubinyi, H.; Folkers, G.; Martin, Y. C. Eds.; Kluwer Aca-
demic: Great Britain, 1998; Vol. 3, p 87.
10. Compound 11 was prepared via the method of Carr et al.
Krauss, R. C.; Strom, R. M.; Scortichini, C. L.; Kruper, W. J.;
Wolf, R. A.; Carr, A. A.; Rudisill, D. E.; Panzone, G.; Hay,
D. A.; Wu, W. W. WO 9500480, 1995.
11. Compound 15 was prepared via the method of Andersen
et al.Andersen, K.; Perregaard, J.; Liljefors, T.; Hyttel, J. J.
Med. Chem. 1996, 39, 3723.
12. Kang, J.; Wang, L.; Cai, F.; Rampe, D. Eur. J. Pharma-
col. 2000, 392, 137.
13. Roux, B.; MacKinnon, R. Science 1999, 285, 100.
14. (a) Suessbrich, H.; Schonherr, R.; Heinemann, S. H.;
Lang, F.; Busch, A. FEBS Lett. 1997, 414, 435. (b) Tamargo,
J. Jpn. J. Pharmacol. 2000, 83, 1.
15. Mitcheson, J. S.; Chen, J.; Lin, M.; Culberson, C.; San-
guinetti, M. C. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 12329.
16. Keating, M. T. K.; Sanguinetti, M. C. Cell 2001, 104, 569.
17. (a) Jiang, Y.; Lee, A.; Chen, J.; Cadene, M.; Chait, B. T.;
MacKinnon, R. Nature 2002, 417, 515. (b) Jiang, Y.; Lee, A.;
Chen, J.; Cadene, M.; Chait, B. T.; MacKinnon, R. Nature
2002, 417, 523.
18. Chen, J.; Seebohm, G.; Sanguinetti, M. Proc. Natl. Acad.
Sci. U.S.A. 2002, 99, 12461.
9. Sertindole analogues 2, 3, and 14 were prepared via the
method of Perregaard et al. Perregaard, J.; Arnt, J.; Boegesoe,
K. P.; Hyttel, J.; Sanchez, C. J. Med. Chem. 1992, 35, 1092.
19. Zhong, W.; Gallivan, J. P.; Zhang, Y.; Li, L.; Lester, H.;
Dougherty, D. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95,
12088.