A selective class of inhibitors for the CLC-Ka chloride ion channel

Article abstract of DOI:10.1073/pnas.1720584115

CLC proteins are a ubiquitously expressed family of chloride-selective ion channels and transporters. A dearth of pharmacological tools for modulating CLC gating and ion conduction limits investigations aimed at understanding CLC structure/function and ph

Full text of DOI:10.1073/pnas.1720584115

A selective class of inhibitors for the CLC-Ka chloride
ion channel
Anna K. Kostera,b,1, Chase A. P. Woodb,1, Rhiannon Thomas-Trana, Tanmay S. Chavanb, Jonas Almqvistc,
Kee-Hyun Choid, J. Du Boisa,2, and Merritt Madukeb,2
aDepartment of Chemistry, Stanford University, Stanford, CA 94305; bDepartment of Molecular and Cellular Physiology, Stanford University School of
Medicine, Stanford, CA 94305; cDepartment of Cell and Molecular Biology, Uppsala University, 751 24 Uppsala, Sweden; and dMaterials and Life Science
Research Division, Korea Institute of Science and Technology, 02792 Seoul, Republic of Korea
Edited by Richard W. Aldrich, The University of Texas at Austin, Austin, TX, and approved March 21, 2018 (received for review December 7, 2017)
CLC proteins are a ubiquitously expressed family of chloride-
selective ion channels and transporters. A dearth of pharmacolog-
ical tools for modulating CLC gating and ion conduction limits
investigations aimed at understanding CLC structure/function and
physiology. Herein, we describe the design, synthesis, and evalu-
ation of a collection of N-arylated benzimidazole derivatives (BIMs),
one of which (BIM1) shows unparalleled (>20-fold) selectivity for CLC-
Ka over CLC-Kb, the two most closely related human CLC homologs.
Computational docking to a CLC-Ka homology model has identified a
BIM1 binding site on the extracellular face of the protein near the
chloride permeation pathway in a region previously identified as a
binding site for other less selective inhibitors. Results from site-
directed mutagenesis experiments are consistent with predictions of
this docking model. The residue at position 68 is 1 of only ∼20 extra-
cellular residues that differ between CLC-Ka and CLC-Kb. Mutation of
this residue in CLC-Ka and CLC-Kb (N68D and D68N, respectively)
reverses the preference of BIM1 for CLC-Ka over CLC-Kb, thus show-
ing the critical role of residue 68 in establishing BIM1 selectivity. Mo-
lecular docking studies together with results from structure–activity
relationship studies with 19 BIM derivatives give insight into the in-
creased selectivity of BIM1 compared with other inhibitors and iden-
tify strategies for further developing this class of compounds.
without intervention, permanent neurological damage can occur (20,
21). Selectivity between small molecule inhibitors of CLC-Ka and
CLC-Kb is critical for any clinical application, as these two channels
are also expressed in the inner ear, where they play critical but re-
dundant roles: disruption of both homologs results in deafness, while
disruption of either individually has no effect on hearing (22).
The development of selective small molecules that specifically
target either CLC-Ka or CLC-Kb poses a formidable challenge in
ligand design due to the >90% sequence identity between the two
homologs. Prior reports of CLC inhibitors largely describe low-
affinity (midmicromolar to millimolar IC50) compounds that lack
selectivity. The most potent CLC inhibitor known is a peptide
toxin, GaTx2, which selectively inhibits CLC-2 with an IC50 of
∼20 pM. However, despite its potency, GaTx2 inhibition saturates
at ∼50% (23), thus limiting its use as a pharmacological probe.
Accordingly, we have been motivated to identify small molecule
inhibitors of specific CLCs. Two classes of small molecule inhibi-
tors have been reported to display moderate potency and selectivity
for CLC-Ka over CLC-Kb. The first, a collection of stilbene
disulfonates, is a promiscuous class of anion transport inhibitors
(24–26). One of these, 4,4′-diisothiocyano-2,2′-stilbenedisulfonic
acid (DIDS), was used as a chemical tool to corroborate the
dimeric architecture of the Torpedo CLC-0 channel (27). Sub-
sequently, our laboratories showed that DIDS hydrolyzes in
buffer solution to form polythiourea oligomers, which are more
chloride channel molecular probes electrophysiology
|
|
LC proteins are a class of highly selective chloride (Cl−) ion
channels and transporters that are essential for proper
Significance
C
physiological function, including electrical signaling in muscles
and neurons and regulation of ion homeostasis (1–4). CLCs
possess a unique homodimeric, double-barreled structure and
operate through distinct gating (opening/closing) mechanisms
(5–10), which differ markedly from those of sodium, potassium,
and calcium channels. While researchers in the field have made
important progress toward understanding general CLC structural
dynamics and physiology, achieving a precise mechanistic un-
derstanding of these processes has remained a challenge (11–13).
Small molecules that bind and modulate specific CLC homologs
could serve as valuable tools for untangling the complex gating
behaviors and diverse physiological functions within the CLC
family. Small molecule modulators of CLCs also hold promise as
therapeutic leads (14, 15). For certain kidney-related disorders, two
CLC homologs, CLC-Ka and CLC-Kb, are particularly compelling
targets due to their important roles in maintaining electrolyte
balance. In humans, CLC-Kb loss-of-function mutations result in
type III Bartter’s syndrome, a salt-wasting disorder characterized
by low blood pressure (16). This phenotype suggests CLC-Kb as
a potential target for antihypertensive therapeutics. Similarly,
mouse KO models show that CLC-Ka Cl− transport is necessary
for concentration of urine (17, 18) and thus suggest that inhibitors
could be used to treat hyponatremia (19), a condition in which
blood concentrations of sodium are low. Hyponatremia develops
when the kidneys fail to excrete water efficiently and can occur in
patients with heart failure, cirrhosis, renal failure, or other condi-
tions (19, 20). Treatment of hyponatremia can be challenging, and
Chloride ion channels and transporters (CLCs) are critical to car-
diovascular, neurological, and musculoskeletal function. Small
molecules capable of selectively inhibiting CLCs would serve as
valuable tools for investigating CLC function and would have
potential applications for treating CLC-related disorders. The lack
of such agents has impeded efforts to study this family of pro-
teins. This work introduces a class of inhibitors with unprece-
dented selectivity for a single CLC homolog, CLC-Ka. Insights
gained through experiments to validate a predicted ligand bind-
ing site and to evaluate structure–activity relationships ratio-
nalize inhibitor potency and CLC-Ka selectivity. Our findings
provide tools for studies of CLC-Ka function and will assist
subsequent efforts to advance specific molecular probes for
different CLC homologs.
Author contributions: A.K.K., C.A.P.W., J.A., K.-H.C., J.D.B., and M.M. designed research;
A.K.K., C.A.P.W., R.T.-T., T.S.C., J.A., and K.-H.C. performed research; A.K.K., C.A.P.W.,
R.T.-T., T.S.C., J.A., K.-H.C., J.D.B., and M.M. analyzed data; and A.K.K., J.D.B., and M.M.
wrote the paper with contributions from all authors.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Published under the PNAS license.
1A.K.K. and C.A.P.W. contributed equally to this work.
2To whom correspondence may be addressed. Email: jdubois@stanford.edu or maduke@
stanford.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1720584115/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1720584115
PNAS Latest Articles
|
1 of 10

NH2
NH
NO2
NO2
NH
potent antagonists of CLC-0, CLC-Ka, and CLC-ec1 than DIDS
itself (28). A DIDS pentamer (Fig. 1A) is the most potent and
selective nonpeptidic CLC inhibitor known, with an IC50 against
CLC-Ka of 0.5 μM and ∼100-fold increased IC50 against CLC-
Kb (28). However, structure–activity relationship (SAR) studies
and further development of DIDS-based inhibitors are severely
hampered due to the low water solubility of these oligomeric
molecules as well as the synthetic challenges of systematically
modifying the DIDS scaffold. A second class of CLC inhibitors
is based on an aryl-substituted benzofuran carboxylic acid (29,
30). The most notable of these compounds, MT-189 and SRA-
36 (Fig. 1B), are ∼10-fold less potent than the DIDS pentamer
but more amenable to synthetic manipulation. While SRA-
36 inhibits both CLC-Ka and CLC-Kb with equal potency, MT-
189 has a nearly threefold greater apparent affinity for CLC-Ka
over CLC-Kb (IC50 = 7 vs. 20 μM) (29).
KOtBu
Fe/FeCl3
Cl
F
Cl
Cl
HOAc
DMF, 0 °C
Cl
Cl
NH2
EtOH, 70 °C
Cl
NH2
NH
N
N
N
N
CS2, KOH
30%
H O
2 2
SH
Cl
K
SO3
Cl
Cl
Cl
Cl
1 M KOH
MeOH
EtOH, H2O
70 °C
Cl
Scheme 1. Synthesis of BIMs.
BIM1 Exhibits Enhanced Homolog Selectivity. CLC-K channels were
expressed in Xenopus oocytes, and two-electrode voltage clamp
(TEVC) recording was used to measure currents before and
after perfusion of inhibitor solutions. At 100 μM, BIM1 is an
effective inhibitor of CLC-Ka but shows markedly reduced ac-
tivity toward CLC-Kb (Fig. 2 A and B and Table S1). The IC50
Results and Discussion
Inhibitor Design and Synthesis. Our design of N-arylated benz-
imidazole derivatives (BIMs) follows from previous reports by
Liantonio et al. (29, 30) that describe the inhibitory effects of
benzofuran carboxylate compounds against CLC-Ka and CLC-
Kb (29, 30). In these studies, the most potent CLC-K biaryl in-
hibitors were found to be conformationally restricted molecules
that favor a twisted, noncoplanar structure of the two aromatic
rings (Fig. 1B) (29). We hypothesized that an N-arylated benz-
imidazole would possess an analogous molecular topology. This
substituted heterocycle could be easily assembled from com-
mercially available building blocks (Scheme 1), facilitating access
to a collection of structurally disparate derivatives. Replacing the
carboxylate group of MT-189 with a sulfonate anion was also
considered an important design element to increase selectivity
for CLC-Ka over CLC-Kb, a decision based on our earlier
finding that the sulfonated CLC inhibitor DIDS exhibits greater
homolog selectivity than MT-189 (28). The resulting heterocyclic
structure, BIM1, is shown in Fig. 1C.
for BIM1 against CLC-Ka, 8.5
0.4 μM, is similar to that
reported for MT-189 (7.0 1.0 μM) (29). In contrast, the po-
tency of BIM1 against CLC-Kb is significantly diminished
[IC50 = 200 20 μM for BIM1 (Fig. 2C) vs. 20 2 μM for MT-
189 (29)]. The identification of BIM1, which displays >20-fold
selectivity for CLC-Ka over CLC-Kb, is notable given that these
two channels are ∼90% identical in amino acid sequence. The
next closest homologs within the CLC family, CLC-1 and CLC-2
(40–45% identical to the CLC-K channels), are insensitive to
BIM1 at 100 μM (Fig. 3). These data strongly suggest that BIM1
will selectively inhibit CLC-Ka over other ion channels and
transporters found in the plasma membrane, a notable advance
given the promiscuity of many Cl− channel inhibitors (31).
Computational Modeling to Predict the BIM Binding Site. To gain in-
sight into the location of the BIM1 binding site, we generated a
homology model of human CLC-Ka based on the crystal structures
of the eukaryotic CLC transporter Cyanidioschyzon merolae
(cm)CLC [Protein Data Bank (PDB) ID code 3org] (32) and the
water-soluble domain of human CLC-Ka [PDB ID code 2pfi (33)].
Computational docking of BIM1 to the extracellular surface of our
CLC-Ka homology model identified a binding site near residue 68
(Fig. 4), a site known to affect channel sensitivity to MT-189 (29, 34)
as well as a variety of other known CLC-Ka inhibitors (3-phenyl-p-
chlorophenoxy-propionic acid, DIDS, and flufenamic acid deriva-
tives) (34–36). In most CLC homologs, including CLC-Kb, this
residue is an aspartate. CLC-Ka is unique among the human homo-
logs in that it contains a neutral asparagine at this position, 1 of only
∼20 residues on the extracellular face of CLC-Ka that differ from
CLC-Kb. A second ligand–channel interaction identified by our model
is an electrostatic interaction of the BIM1 sulfonate group with K165,
a residue located in the extracellular vestibule of the channel (Fig. 4).
SO3
A
H
N
H
N
H
N
H
N
=
S
S
H2N
SO3
S
N
H
N
H
N
H
N
H
S
NH2
DIDS pentamer
O
O
N
N
B
C
CO2
CO2
Cl
SO3
Cl
Cl
Cl
MT-189
SRA-36
BIM1
Testing Predictions of the Computational Docking. In our CLC model,
the proximity of N68 to the sulfonate group of BIM1 (Fig. 4)
predicts that introduction of an acidic residue at this position will
weaken the CLC-Ka–BIM1 interaction. CLC-Ka N68D was
expressed in Xenopus oocytes, and the sensitivity of the mutant
channel to BIM1 was evaluated. Consistent with our model, the
N68D mutation decreased sensitivity to BIM1 from an IC50 of
8.5 0.4 to 114 14 μM (Fig. 5 and Table S2). This loss in po-
tency parallels that observed for MT-189 against this same mutant
85°
C10
C8
85°
C9
C8
C7
C7
C7
C9
N2
C8
C10
C9
MT-189
BIM1
Fig. 1. Design of the BIM scaffold using known CLC-K inhibitors as in-
spiration. (A) The DIDS pentamer inhibits CLC-Ka with an IC50 of 0.5 μM and
exhibits ∼100-fold reduced potency toward CLC-Kb. (B) The benzofuran car-
boxylates, which are synthetically more tractable than the DIDS oligomers,
inhibit CLC-Ka and CLC-Kb with apparent affinities in the low micromolar
range. MT-189 shows approximately threefold selectivity for CLC-Ka over CLC-
Kb. Lower illustrates the noncoplanar conformation of MT-189, which is pre-
dicted to be an essential structural feature for inhibition (29). (C) The design of
BIM1, a benzimidazole sulfonate, is inspired by the benzofuran lead com-
pounds shown in B and by the sulfonated DIDS inhibitors (A). Lower shows a
hypothesized noncoplanar conformation for BIM1.
(IC50 of 7.0 1.0 vs. 54
8 μM) (29). As another test of the
model, the complementary mutation, D68N, was introduced into
CLC-Kb. This mutation increased sensitivity to BIM1 from an
IC50 of 200 20 to 55 36 μM (Fig. 5 and Table S2). Thus, the
preference of BIM1 for CLC-Ka over CLC-Kb is eliminated with
this single-point mutation. This experiment shows that the amino
acid at position 68 is critical for establishing BIM1 selectivity
2 of 10
|
www.pnas.org/cgi/doi/10.1073/pnas.1720584115
Koster et al.

between CLC-Ka and CLC-Kb and is consistent with a predicted
direct interaction between BIM1 and this residue.
A second BIM1–protein interaction apparent in our CLC-Ka
homology model—that between the sulfonate group of the in-
hibitor and K165—was more challenging to validate through
mutagenesis experiments. It has been shown that mutation of
K165 to a neutral residue (A, C, H, or Q) prevents functional
expression (37). Thus, as an alternative approach to examining the
sulfonate–K165 interaction, we first varied the extracellular pH of
our recording solution to alter the protonation state of the lysine
residue. Changes in the BIM1 IC50 as a function of pH would be
expected if an attractive electrostatic interaction between
K165 and the anionic BIM1 sulfonate group is present. Consistent
with this prediction, raising the pH from 7.6 to 9.5 decreases in-
hibition of CLC-Ka by BIM1 (from 75 3 to 24 4% at 22 μM
BIM1) (Fig. 6 A and C). To test that the reduction in BIM1 po-
tency is due to the neutralization of K165 (as opposed to another
titratable residue on CLC-Ka), we examined the effect of an Arg
mutation at K165. According to previous work, K165R is the only
mutation tolerated at this position (37). At pH 9.5, the protonated
form of Arg (K165R) should be present to a greater extent than
the protonated form of Lys (K165) due to the inherent difference
in pKa between these two residues. The pKa values for Lys and
Arg in alanine pentapeptides are 10.4 and 12.3, respectively (38).
In the folded protein environment, these pKa values will shift (38);
however, the relative difference in pKa should persist. As such,
inhibition of K165R by BIM1 should be less sensitive to pH than
WT CLC-Ka. At pH 7.6, BIM1 inhibits the CLC-Ka K165R
channel with similar potency to the WT (75 3 vs. 72 5% in-
hibition at 22 μM BIM1) (Fig. 6). In contrast, at pH 9.5, K165R is
more sensitive to BIM1 than the WT channel is (inhibition of 54
7 vs. 26 3% at 22 μM BIM1) (Fig. 6). Together, these results
support the hypothesis that the positively charged K165 residue is
positioned in or near the inhibitor binding site.
Comparison with an Alternative Docking Model. In 2016, Liantonio
et al. (30) disclosed results from a computational study, in which MT-
189 and related inhibitors were docked to CLC-Ka. Our model of
CLC-Ka docked to BIM1 shares many similarities, including spe-
cific interactions with N68 and K165 (Fig. 4). However, the
model by Liantonio et al. also identifies ligand contacts with
I263 and H346. In our model, these loop residues are far re-
moved from the inhibitor binding site (Fig. 7A). The difference
between the two docking results likely arises from differences in
the structural templates used to generate the CLC-Ka homology
models, which vary significantly in their loop regions (discussion
is in SI Discussion of CLC-K Homology Models).
To distinguish between the two computational models, we
evaluated the contribution of H346 to BIM1 binding through
site-directed mutagenesis experiments. Application of 15 μM
BIM1 to H346A inhibited current by 65
3%, which is in-
distinguishable from the inhibition observed with WT CLC-Ka:
62 5% (Fig. 7 and Fig. S2). In addition, the H346A mutation
did not affect the sensitivity of channel binding to the benzofuran
derivative MT-189 (depicted in Fig. 1B). These results (Fig. 7
and Fig. S2) favor a model in which this residue is not involved in
either benzofuran or benzimidazole binding.
Comparison with Homology Models Based on Bovine CLC-K
Structures. After our initial mutagenesis studies to identify the
BIM binding site, two high-resolution structures of the bovine chloride
channel bCLC-K were disclosed (39) [The two structures, al-
though distinct, are highly similar to one another (details are in
Materials and Methods).] Because of the significantly greater
amino acid sequence similarity of bCLC-K with CLC-Ka and CLC-Kb
Fig. 2. BIM1 is selective for CLC-Ka over CLC-Kb. (A) Representative CLC-Ka
currents, recorded using TEVC, before application, after application, and after
washout of 100 μM BIM1 (Upper) or MT-189 (Lower). (B) Representative CLC-
Kb currents showing sensitivity to 100 μM BIM1 and MT-189 (as in A). (C)
Summary data for BIM1 inhibition at +60 mV. [Inhibition is not significantly
voltage-dependent (Fig. S1).] Each data point represents the mean SEM for
measurements on 3–15 oocytes at each concentration (Table S1). Oocytes were
obtained from multiple (>10) batches of oocytes injected and measured on
separate occasions. For CLC-Ka, the solid line is a fit of the data to the equa-
tion: I = (Imax × [BIM1]n)/(IC50n + [BIM1]n), where I is the percentage inhibition,
Imax is the maximal inhibition, IC50 is 8.5 0.4 μM, and n is the Hill coefficient
(0.99). For CLC-Kb, the solid line is a fit to the same equation but with Imax
and n fixed at 100 and 1.0, respectively, yielding a value of 200 20 μM for
the IC50 of BIM1 against CLC-Kb.
Koster et al.
PNAS Latest Articles
|
3 of 10

application of each compound at 100 μM (as in Fig. 2). Com-
pounds inhibiting >80% of the CLC-Ka current at 100 μM (Fig. 8
and Table S3) were also screened at 5 μM for a more accurate
comparison of potency (Fig. S4 and Table S4). BIM15 and
BIM16 were found to be slightly more potent than BIM1, inhib-
iting CLC-Ka by 54 and 72%, respectively, at 5 μM compared with
37% for BIM1; however, this was accompanied by a loss in se-
lectivity (discussed below).
Effects on inhibitor potency. To assess the influence of each BIM
substituent, the potency against CLC-Ka of every inhibitor was
compared with an equivalent compound bearing a hydrogen atom
at each of the positions R1–R6. To begin, R1 was substituted with a
carboxylate, a sulfonate, or a hydrogen atom. Consistent with an
electrostatic interaction in the inhibitor binding site (Fig. 4), a
negatively charged group at R1 is necessary for inhibitor efficacy
(comparing inhibition with 100 μM compound) (Fig. 8). Re-
−
−
placement of -SO3 or -CO2 with a hydrogen atom (BIM3 and
BIM6) renders these compounds inactive against CLC-Ka.
At position R2, the presence of either -Cl or -Br (BIM1 and
BIM11) moderately increases the percentage of inhibition rela-
tive to BIM10 (R2 = H) (Fig. 8 and Table S3). A similar effect is
noted for substitution at position R6 with -Cl or -Br in place of -H
(BIM1 and BIM7 vs. BIM8) (Fig. 8 and Table S3). If -Cl is in-
cluded at both R2 and R6 (BIM1), the effect on potency is ad-
ditive compared with monohalogenated compounds (BIM8 and
BIM10). The identity of the halogen atom at positions R2 and R6
is inconsequential, with both -Cl and -Br compounds showing the
same efficacy at 5 μM (Fig. S4 and Table S4). In comparison with
R2 and R6, substituting -Cl or -Br at R5 (BIM12 and BIM13) does
not improve BIM efficacy. These compounds are weakly potent
and have similar affinity to BIM8 (R6 = H) (Fig. 8 and Table S3).
To probe the steric environment of the binding pocket sur-
rounding the benzimidazole and N-phenyl rings, each of the
positions R3–R6 was substituted with a -Ph group. For com-
pounds substituted at position R5 (BIM 14), there is no effect on
potency relative to BIM1, while for position R6 (BIM4, BIM5),
there is a negative effect (Fig. 8 and Table S3). Substitution at R3
or R4 (BIM15 and BIM16) produces the opposite result to R6, an
enhancement in potency relative to BIM1 (Fig. 8 and Table S3).
By contrast, the introduction of a secondary amide group at an
equivalent position, R3 (BIM19), results in a reduction in po-
tency (Fig. 8). Collectively, these findings suggest that the N-aryl
group of the BIM ligands occupies a rather large void space in
the receptor site that can accommodate R3 and R4 substituents,
with hydrophobic phenyl groups favored over the more polar
amide substituent.
Fig. 3. Selectivity of BIM1 among mammalian CLC homologs. Representa-
tive currents from Xenopus oocytes expressing CLC-1 (A) or CLC-2 (B) shown
before and after application of 100 μM BIM1. (C) Summary inhibition data (
SEM) for 100 μM BIM1 on CLC-1 (n = 8), CLC-2 (n = 8), CLC-Ka (n = 9), and
CLC-Kb (n = 6). Inhibition is reported for data at +60 mV (CLC-Ka, CLC-Kb,
and CLC-1) or −120 mV (CLC-2). For CLC-1 and CLC-2, inhibition is not sig-
nificantly different from zero (P = 0.55 and P = 0.84, respectively).
Effects on inhibitor selectivity. To evaluate the effect of each sub-
stituent on inhibitor selectivity, we compared the ratio of CLC-
Kb/CLC-Ka percent inhibition at 100 μM for all 19 BIMs (Fig. 8
−
and Fig. S5). Substitution at R1 with -CO2 (BIM2) instead of
(84% sequence identity) compared with the cmCLC transporter on
which our original model was based (32% identity), we developed
updated CLC-Ka homology models derived from the bCLC-K struc-
tures. Consistent with our original model, docking of BIM1 to either of
these constructs favors an extracellular binding pocket composed of
residues N68 and K165, with H346 and I263 distant from the in-
hibitor binding site. Although both original and updated CLC-Ka
homology models locate N68 and K165 in the binding pocket, the
collection and orientation of amino acids within this region vary
significantly between models. A summary of these findings is pre-
sented in Fig. S3.
Fig. 4. BIM1 docking to a CLC-Ka homology model. The CLC-Ka homology
model viewed from the side. Subunits of the homodimer are colored purple and
cyan, and BIM1 is docked to each subunit near the extracellular side of the Cl−
permeation pathway. Right shows a close-up stereoview of the BIM binding site.
Residues predicted to interact with BIM1 and tested in mutagenesis experiments
(N68 and K165) are shown in stick representation. This initial model was con-
structed using cmCLC (PDB ID code 3org) as a template.
SAR Studies. To evaluate the effects of molecular structure on
inhibitor potency and selectivity, we synthesized a panel of 19 BIM
inhibitors substituted at positions R1–R6 (Fig. 8). Using TEVC,
CLC-Ka and CLC-Kb currents were measured before and after
4 of 10
|
www.pnas.org/cgi/doi/10.1073/pnas.1720584115
Koster et al.

Fig. 5. Testing the docking model: effect of residue 68 mutations. Represen-
tative currents for CLC-Ka and CLC-Kb N68/D68 mutants and the respective re-
sponse to BIM1. The summary graph shows the mean for measurements on two
to four oocytes at each concentration. Error bars show either the range of the
data points (for n = 2) or the SEM (for n = 3–4) (Table S2). Oocytes were from two
(CLC-Kb D68N) or three (CLC-Ka N68D) batches injected and measured on sep-
arate occasions. For comparison, results for WT channels are reproduced from
Fig. 2. Dashed curves are fit as described in Fig. 2, with IC50 values of 114 14 μM
(CLC-Ka N68D) and 60 5 μM (CLC-Kb D68N).
-SO3− results in a compound with similar selectivity for CLC-Ka
over CLC-Kb compared with BIM1. Replacing either R2 or R6
-Cl in BIM1 with a sterically larger -Br group (BIM7, BIM11)
has little to no effect on selectivity. Similarly, an isomer of
BIM1 in which R5 = Cl and R6 = H (BIM12) shows little change
in homolog selectivity (albeit the potency of this molecule is
reduced from BIM1). Strikingly, Ka/Kb selectivity is completely
eroded with the -Br analog of BIM12 (BIM13).
Selective inhibition of CLC-Ka is abolished when a -Ph group is
introduced in place of -Cl at R5 or R6 (BIM4, BIM5, BIM14). A
similar outcome has been noted with benzofuran inhibitors: the
R6-phenyl–substituted RT-93 has an IC50 of 7.0 0.9 μM against
Fig. 6. Testing the docking model: titrating the charge at residue 165. Repre-
sentative CLC-Ka WT (A) or K165R (B) currents at pH 7.6 or 9.5, showing re-
sponses to 22 μM BIM. (C) Summary graph shows the mean SEM for the WT,
pH 7.6 (71.8 4.6, n = 4); K165R, pH 7.6 (74.8 2.9, n = 5); the WT, pH 9.5 (26.3
3.4, n = 5); and K165R, pH 9.5 (54.2 7.9, n = 3). Oocytes were from two batches
(each injected with WT and K165R mRNA and each tested at pH 7.6 and pH 9.5).
P values were calculated using unpaired t tests.
Koster et al.
PNAS Latest Articles
|
5 of 10

structure of the heterocycle (benzimidazole vs. benzofuran).
Previous investigations with MT-189 analogs concluded that re-
placement of the benzofuran heterocycles with either indole or
benzothiophene does not alter potency toward CLC-Ka (29) and
that a sulfonate group in place of carboxylate anion reduces
potency (30). These studies did not query the effect of such
structural changes on homolog selectivity.
Based on our studies with DIDS derivatives, we had hypoth-
esized that the sulfonate group in BIM1 would confer increased
selectivity for CLC-Ka compared with the analogous carboxylate
derivative. To test this hypothesis, we synthesized a set of four
compounds and evaluated the potency of these ligands against
CLC-Ka and CLC-Kb (Fig. 9). The results clearly show that the
enhanced selectivity of BIM1 is dominated by the structure of
the heterocycle and not the sulfonate group. Both benzofuran
derivatives, MT-189 and AK2-168, exhibit reduced selectivity for
CLC-Ka over CLC-Kb compared with analogous benzimid-
azoles, BIM1 and BIM2. The choice of sulfonate or carboxylate
in both the benzofuran and benzimidazole series has little in-
fluence on potency and selectivity.
A Rationale for BIM1 Selectivity. What accounts for the enhanced
selectivity afforded by the benzimidazole heterocycle in BIM1? We
attempted to address this question by comparing the docking re-
sults for BIM1 and MT-189. Due to the 10-fold difference in
BIM1 and MT-189 potency toward CLC-Kb [IC50 = 200 20 μM
for BIM1 vs. 20 2 μM for MT-189 (29)], we first examined our
CLC-Kb models, looking for specific amino acid side-chain inter-
actions with the heterocyclic scaffolds that could explain the re-
duced inhibitor potency toward this homolog. (In this analysis, we
focused on the bCLC-K–derived homology models.) Analysis of
the top nine docking poses for each model suggests that both
BIM1 and MT-189 can adopt a variety of orientations in the
binding pocket and does not reveal a preference for the benzofuran
heterocycle over the benzimidazole heterocycle. Both BIM1 and
MT-189 dock to CLC-K models with a biaryl dihedral angle of
∼90° (Fig. S6). These observations are consistent with our initial
hypothesis that the BIM scaffold preferentially adopts a non-
coplanar conformation (Fig. 1C) but do not provide insight into the
selectivity difference between BIM1 and MT-189, as none of the
predicted ligand–protein interactions seem manifestly responsible
for selectivity. Estimating from the measured IC50 values, the dif-
ference in binding energy that accounts for the selectivity of BIM1
for CLC-Ka over CLC-Kb is likely <2 kcal/mol. By the same logic,
the energetic preference of MT-189 for CLC-Kb compared with
BIM1 is even smaller. Given the resolution limitations of homology
modeling, it is, therefore, not surprising that in silico modeling on
this system is insufficient to elucidate the elements responsible for
the enhanced selectivity of BIM1.
While our models do not reveal any specific protein–heterocy-
clic interactions to account for BIM1 selectivity, general differ-
ences are evident in the electrostatic surface potentials of the
ligand binding pockets in CLC-Ka and CLC-Kb. Models based on
either of the two bCLC-K structures indicate that the inhibitor
binding site in CLC-Kb is qualitatively more hydrophobic than
that of CLC-Ka (Fig. 10). To examine whether there is a link
between inhibitor hydrophobicity and CLC-Ka/CLC-Kb selectiv-
ity, we calculated logP (cLogP) values for the four analogous
benzimidazole and benzofuran derivatives (Fig. 9) using three
different cLogP calculators. To isolate the effect of the hetero-
cycle on overall hydrophobicity, we compared compounds with
equivalent anionic groups: BIM1 with AK2-168 and BIM2 with
MT-189. Across all calculation methods, the trend that emerges is
that benzofuran cLogP values are generally higher than those of the
corresponding benzimidazoles (Table 1). This trend is consistent
with the experimental LogP values reported for the unfunctional-
ized benzimidazole and benzofuran heterocycles (1.32 for
benzimidazole and 2.67 for benzofuran) (Table 1) (40), which
indicate that benzofuran is inherently more hydrophobic than
benzimidazole. Addition of nonpolar substituents, such as -Br
and -Ph, often negates any gains in selectivity achieved by the
Fig. 7. Testing the prediction of an alternative homology model. (A) In our CLC-
Ka homology model (Fig. 4), residues I263 and H346 on linkers I–J and K–L are far
removed from the inhibitor binding pocket. This is in contrast to an alternative
model (30), which places these residues in positions that interact with the bound
inhibitor. (B) Summary data showing that mutation of H346 has no effect on
inhibition by either BIM1 or MT-189. Error bars show the mean
SEM for
measurements on three to five oocytes. Oocytes were from two batches (with
one batch used in the experiments comparing BIM1 inhibition of WT and H346A
channels and the second batch used in the experiments with MT-189). P values
(0.53 for BIM1 and 0.66 for MT-189) were calculated using unpaired t tests.
Representative data traces are shown in Fig. S2.
CLC-Ka and an IC50 of 6.0 0.9 μM against CLC-Kb (29). In
addition, the recently reported inhibitor SRA-36 (Fig. 1B),
substituted with a benzyl group at R6, is equipotent against CLC-
Ka and CLC-Kb (30). Loss of Ka/Kb selectivity is also noted for
R3 and R4 -Ph– and -Br–substituted BIMs (BIM15 to BIM18).
Because BIM15 and BIM16 are more potent than BIM1, a
comparison of these compounds at a nonsaturating concentra-
tion is important to confirm that they are less selective than BIM1.
Accordingly, we evaluated inhibition of CLC-Kb by these BIMs at
5 μM (Table S4) and compared selectivity with BIM1 (Fig. S5).
These results confirm that BIM15 and BIM16 are substantially less
selective between CLC-Ka and CLC-Kb than BIM1.
Our collective findings indicate that the addition of large hydro-
phobic substituents at positions R3–R6 on BIM decreases inhibitor
selectivity between the two CLC homologs. The incorporation of -Cl
groups at R2 and R6, however, is needed to ensure high affinity and
selectivity of BIM1 toward CLC-Ka. Other substituents at these two
sites that are similar in size and/or polarity to -Cl may further en-
hance potency and selectivity in next generation inhibitor designs.
Identifying the Heterocycle as Critical for CLC Homolog Selectivity.
The structure of BIM1 differs from MT-189 in both the nature of
the anionic substituent at R1 (sulfonate vs. carboxylate) and the
6 of 10
|
www.pnas.org/cgi/doi/10.1073/pnas.1720584115
Koster et al.

N
N
N
N
CLC-Ka
SO3
Cl
SO3
N
N
Cl
SO3
Cl
Cl
Cl
Cl
BIM1
H
BIM1
H
N
N
SO3
Cl
BIM1
Cl
N
N
CO2
Cl
N
N
Cl
SO3
Cl
Cl
Br
BIM17
N
N
BIM2
SO3
Cl
Br
Cl
BIM18
R5
R6
N
N
N
N
H
R1
R2
Cl
Cl
N
N
CLC-Kb
BIM15
SO3
Cl
Cl
N
N
SO3
Cl
BIM3
Cl
R4
N
R3
SO3
Cl
NH
O
N
BIM19
BIM16
BIM4
N
N
N
SO3
SO3
Br
N
N
N
Cl
Cl
Cl
CO2
Cl
Benzimidazole (BIM)
BIM1
BIM11
H
Cl
N
N
N
N
N
N
N
N
N
N
N
BIM5
SO3
Cl
SO3
Cl
SO3
H
SO3
SO3
Cl
SO3
Cl
N
Cl
Br
H
Cl
H
N
N
H
Cl
BIM8
BIM12
BIM1
BIM7
BIM9
BIM10
Br
N
N
N
N
N
N
N
SO3
Cl
SO3
Cl
SO3
Cl
SO3
Cl
N
BIM6
H
BIM13
BIM14
BIM4
BIM8
Fig. 8. SARs of BIM inhibitors; 19 BIM variants were synthesized and evaluated. Inhibitor structures are shown, such that each boxed group contains var-
iations at one of the ring positions: R1 (blue), R2 (red), R3 (pink), R4 (orange), R5 (magenta), or R6 (green). The ring positions are correspondingly labeled and
colored on the diagram of the BIM scaffold shown in Middle Center. Bar graphs show summary inhibition data (mean
SEM for measurements on 3–
17 oocytes) (Table S3).
BIM heterocycle (Fig. 8 and Fig. S5). These observations support
our hypothesis that the more polar BIM scaffold (BIM1) disfa-
vors the more hydrophobic CLC-Kb binding site compared with the
analogous benzofuran derivative (MT-189) and inhibitors bearing
large hydrophobic substituents (BIM4 and BIM5 as well as BIM14
to BIM16). Such insights will aid efforts to further develop CLC-Ka
inhibitors with increased potency and homolog selectivity.
the inhibitor binding site. We also provided evidence that the
polarity of the heterocyclic scaffold is driving inhibitor se-
lectivity for CLC-Ka over CLC-Kb. Next generation CLC-Ka
inhibitors will explore additional variations of the BIM core,
maintaining the optimum twisted biaryl geometry of BIM1
while minimizing the addition of hydrophobic substituents.
Inhibitor design will be guided by our CLC-K models based on
high-resolution CLC-K structures in combination with protein
mutagenesis experiments. This work serves as a foundation for
developing and optimizing potent small molecule modulators
of specific CLC homologs.
Conclusions
The CLC family represents an important class of membrane
proteins for which our understanding of physiological function is
incomplete and the therapeutic potential is unexplored. We have
developed a class of benzimidazole inhibitors that shows un-
precedented selectivity toward a single CLC homolog, CLC-Ka,
which shares >90% sequence identity with its nearest relative.
The modularity of our ligand design and the short synthetic
route that we developed provide ready access to a range of
inhibitor structures. Through computational modeling and
experimental mutagenesis studies, we identified and validated
Materials and Methods
Chemical Synthesis. All reagents were obtained commercially unless other-
wise noted. Organic solutions were concentrated under reduced pressure
(∼20 torr) by rotary evaporation. Air- and moisture-sensitive liquids and solu-
tions were transferred via syringe or stainless steel cannula. Chromatographic
purification of sulfonate products was accomplished using HPLC on a
C18 column (Alltima C18, 10 μM, 22 × 250 mm or SiliaChrom AQ C18, 5 μM,
Koster et al.
PNAS Latest Articles
|
7 of 10

CLC Expression in Xenopus Oocytes. Human CLC-Ka, CLC-Kb, CLC-1, and Barttin
constructs subcloned into the psGEM vector (41) were linearized with NheI
(New England Biolabs) and cleaned using the DNA Clean & Concentrator-5
Kit (Zymo Research) or the Monarch PCR and DNA Reaction Cleanup Kit
(New England Biolabs). The linearized plasmids were then transcribed in
vitro using the mMESSAGE mMACHINE T7 Kit (Ambion; Life Technologies)
and cleaned using the RNeasy MinElute Cleanup Kit (Qiagen) or a classical
chloroform:phenol extraction (42). The Y98A mutation was introduced into
the Barttin construct, because this mutation increases surface expression of
CLC-K channels (22). All mutations were generated using the QuikChange
Lightning Site-Directed Mutagenesis Kit (Agilent Technologies). The entire
gene was sequenced to confirm occurrence of a desired mutation as well as
the absence of additional mutations.
Benzofuran
Benzimidazole
N
O
CO2
CO2
N
Cl
Cl
Cl
Cl
MT-189
BIM2
N
O
Rat CLC-2 subcloned into the pTLN plasmid vector (43) was linearized with
MluI (New England Biolabs). The linearized plasmid was then transcribed in
vitro using the mMESSAGE mMACHINE SP6 Kit (Ambion; Life Technologies)
SO3
Cl
SO3
N
Cl
Cl
Cl
AK2-168
BIM1
Fig. 9. Investigating the role of the heterocycle and anion substituent on CLC-
Ka/CLC-Kb selectivity. Inhibitors synthesized to systematically vary the hetero-
cycle (benzofuran vs. benzimidazole) and the anionic substituent (carboxylate vs.
sulfonate) are shown in Top. Data for inhibition of CLC-Ka and CLC-Kb by 5 μM
of each compound are summarized below (mean SEM for measurements on
4–16 oocytes) (Table S4). Each set of experiments was performed on oocytes
from two or more separate batches. Bottom (purple bars) shows selectivity as
reflected by the ratio of percent inhibition of CLC-Kb and CLC-Ka by 5 μM BIM
compound. Error bars show 90% CIs computed by the method of Fieller (55)
using GraphPad software. The selectivity ratio is shown as CLC-Kb/CLC-Ka, be-
cause the inverse ratio would not allow accurate estimation of CIs (56).
10 × 250 mm). TLC was performed on EM Science silica gel 60 F254 plates
(250 mm). Visualization of the developed chromatogram was accom-
plished by fluorescence quenching and by staining with aqueous potas-
sium permanganate or aqueous ceric ammonium molybdate solution.
NMR spectra were acquired on a Varian Inova spectrometer operating at
400, 500, or 600 MHz for 1H and are referenced internally according to re-
sidual solvent signals. Data for 1H NMR are recorded as follows: chemical
shift (δ; parts per million), multiplicity (singlet, doublet, triplet, quartet, quintet,
multiplet, broad), coupling constant (hertz), and integration. Compound con-
centrations were determined by quantitative NMR in DMSO-d6 using N,N-
dimethylformamide as an internal standard. IR spectra were recorded as thin
films using NaCl plates on a Thermo-Nicolet 300 FT-IR spectrometer and are
reported in frequency of absorption. Low-resolution and high-resolution mass
spectra were obtained from the Vincent Coates Foundation Mass Spectrometry
Laboratory at Stanford University. Detailed synthesis protocols and characterization
data for all inhibitors are provided in Dataset S1.
Fig. 10. A rationale for BIM selectivity—evaluating hydrophobicity of in-
hibitor binding sites. Electrostatic surface maps of the CLC-Ka and CLC-Kb
homology models viewed from the extracellular side; scale is from −10 kT/e
(red; negative) to +10 kT/e (blue; positive). These models, based on the two
cryo-EM structures of bCLC-K (c1 and c2), reveal the CLC-Kb inhibitor binding
pocket to be more hydrophobic than that of CLC-Ka. Molecules shown in the
binding sites are a collection of representative docking poses for BIM1, BIM2,
MT-189, and AK2-168 within each of four models.
8 of 10
|
www.pnas.org/cgi/doi/10.1073/pnas.1720584115
Koster et al.

out”), and after switching to 100 mM NaI (“NaI”). To obtain background-
corrected currents, the NaI currents (or currents from uninjected oocytes as
described above) were subtracted from each of the initial, inhibited, and
washout currents. Percentage inhibition was determined using these
background-corrected currents. Data reported in the figures reflect in-
hibition determined by dividing the background-corrected inhibited current
by the initial current (and multiplying by 100). Experiments in which there
was poor reversibility of inhibition were discarded; poor reversibility was de-
fined as cases in which reversal current differed by >20% from initial current
and/or inhibition calculated using washout current differed by >10% from
inhibition calculated using initial current. Percentage inhibition was plotted as
a function of inhibitor concentration and fitted to a sigmoid function using
Sigmaplot 13.0 (Systat Software) to obtain IC50 values. Summary data report
inhibition at +60 mV unless otherwise noted (Fig. 3 and Fig. S1). To cal-
culate percentage inhibition of CLC-1 or CLC-2 current, an average of 5 ms
(CLC-1) or 17 ms (CLC-2) of the current trace within the final portion of each
test pulse was determined.
Table 1. Evaluating inhibitor hydrophobicity
Experimental
LogP
cLogP
cLogP
cLogP
Compound
QikProp MolInspiration ChemDraw
BIM1
AK2-168
BIM2
MT-189
Benzimidazole
Benzofuran
—
—
2.89
3.27
3.56
4.07
1.32
2.61
1.74
2.55
3.94
5.08
1.43
2.26
2.34
2.98
5.42
4.79
1.57
2.70
—
—
1.32
2.67
Hydrophobicity of benzimidazoles vs. benzofurans. Three different com-
putational algorithms were used to determine the cLogP values [for the set
of inhibitors examined in Fig. 9 and for unsubstituted benzimidazole and
benzofuran, for which experimental values are available (Table S5)].
and purified as described for the other constructs. For some CLC-2 experi-
ments, an alternate construct was used in an attempt to boost expression
levels. Rat CLC-2 was codon optimized for Xenopus and subcloned into the
pUNIV plasmid with the addition of a 3′ β-globin UTR at the end of the gene
(44). The pUNIV construct was linearized with NotI (New England Biolabs) and
transcribed in vitro using the mMESSAGE mMACHINE T7 Kit (Ambion; Life
Technologies). Expression levels were found to be similar for both constructs.
Homology Model Generation. First generation human CLC-Ka and CLC-Kb
models were created from the structural PDB templates of the eukaryotic
CLC transporter cmCLC (PDB ID code 3org) and the water-soluble cytoplasmic
domain of human CLC-Ka (PDB ID code 2pfi). Target template amino acid
sequence alignments were created using the FFAS03 protocol for profile–
profile alignments (46) and used for the initial modeling procedure. The
template sequences were ∼30% identical (40% similar) to the target CLC-Ka
and CLC-Kb sequences. C-α atoms in the two subunits were symmetrically
constrained during model generation given the C2 symmetry of dimeric CLC
protein structures. One Ca2+ ion was added to cytoplasmic loop residues (47)
E259, E261, D278, and E281 in each subunit using spatial restraints.
One thousand preliminary models were built using MODELER 9.12 (48).
The five models with the lowest MODELER objective function were kept for
further evaluation by the homology model-scoring program ProQ (49) and
the crystallographic model assessment tools Procheck (50) and What-If (51).
The overall best-scoring model was selected as the final model for
docking simulations.
Electrophysiology. Defolliculated Xenopus oocytes were injected with 27.6–
50.6 nL of cRNA at 0.2–1.5 μg/μL: 5–40 ng for CLC-K:Y98A-Barttin (2:1 mixture),
4–8 ng for CLC-1, and 19–37 ng for CLC-2. Oocyte storage and recording solu-
tions are described in SI Materials and Methods. All data were collected using
an Oocyte Clamp OC-725C amplifier (Warner Instruments) interfaced with
Digidata 1322A and pClamp 9.2 software (Axon Instruments) at room tem-
perature (∼20 °C). The data were sampled at 10 kHz and low pass-filtered at
1 kHz. Currents were measured using voltage protocols as indicated in the
figures. For CLC-Kb, the test pulse duration (500 ms) was longer than that used
for CLC-Ka (200 ms) because of the slower kinetics of CLC-Kb and hence, the
longer time required to reach steady state at negative voltages. The membrane
potential was held at the resting potential (usually −30 mV) during the inter-
sweep interval between each voltage step, with a total time of 1–2 s per sweep.
Solution changes were done by perfusing the chamber with 4–5 mL of each
solution (∼15 times the volume of the chamber) at ∼1 mL/min using Baxter
Control-A-Flo regulators or manual pipetting. The fraction of current
remaining after steady-state inhibition (percentage inhibition) was calculated
using the currents measured in the presence and absence of inhibitors. For
most experiments, background current from endogenous Xenopus anion
currents was subtracted using an iodide block protocol that is based on the
observation that iodide inhibits CLC-K currents but not endogenous Xenopus
anion currents (35). At the end of each experiment on CLC-K channels, oocytes
were exchanged into a solution containing 100 mM NaI (with 5 mM MgSO4,
5 mM Na-Hepes, pH 7.6, adjusted with NaOH). The contribution of endoge-
nous currents was evaluated based on current remaining in the presence
of 100 mM iodide at +60 mV. Endogenous currents at lower voltages (down
to −140 mV) were estimated by linear extrapolation of the endogenous cur-
rents at +60 mV (35). Endogenous currents were subtracted from recorded
currents with and without inhibitor present. This background correction was
performed for all CLC-K experiments with two exceptions. (i) For the
19-compound screen at 100 μM (Fig. 8), the absence of background correction in
this set of experiments results in a slight underestimation of percentage in-
hibition but still allows for across the board comparison of relative efficacy of
each BIM derivative examined. (ii) For experiments using 300 μM BIM1 (Figs. 2
and 5), the NaI subtraction protocol became inaccurate, since BIM1 at this high
concentration was found to inhibit endogenous currents. To correct for back-
ground currents in experiments with 300 μM BIM1, currents from three to five
uninjected oocytes from the same batch were averaged and used for subtraction
of background current and subsequent calculation of percentage inhibition as
described below. For CLC-1 and CLC-2 experiments, the NaI subtraction protocol
cannot be performed, since these homologs are not completely blocked by io-
dide (43, 45). Given that BIM1 does not significantly inhibit either CLC-1 or CLC-2
(Fig. 3), subtraction of background currents was not necessary.
Second generation human CLC-Ka and CLC-Kb models were created from
structural templates of bCLC-K. The structure of this template protein, de-
termined using single-particle cryo-EM, revealed two different conformations
denoted c1 and c2 (PDB ID codes 5tqq and 5tr1, respectively) (39). The major
difference between the two conformations is an ∼6° rigid body rotation of the
transmembrane regions within each subunit that results in a change at the
subunit interface; the Cl− permeation pathways within each of the two con-
formations are similar to one another (39). Both the c1 and c2 conformations
were used to model CLC-Ka and CLC-Kb, rendering a total of four protein
models based on bCLC-K. Because the overall target template sequence
identity is ∼84%, these template structures likely represent a better structural
framework for modeling than PDB ID code 3org.
Electrostatic surface maps were generated using the Adaptive Poisson-
Boltzmann Solver plugin for PyMol and are depicted as surface potentials in Fig. 10.
Molecular Docking. Inhibitor molecules were drawn using MarvinSketch
14.10.13 and converted to .pdbqt format using AutoDock. Initial docking
trials of our first generation CLC-K models were performed using with
AutoDock 4.2 (52) using a grid spacing of 0.375 Å. A search box surrounding
the extracellular surface of the protein model was defined, in which
1,000 searches were carried out using the Lamarckian Genetic Algorithm. For
each search, 2.5 × 106 energy evaluations were calculated. After structural
rms clustering, lowest docked energy conformations of inhibitors were
found docked at the N68/D68-K165 site. Final docking trials were performed
using AutoDock Vina (53). Searches were centered at the N68/D68-K165 site
using a search box of 22 × 22 × 22 Å in x, y, and z dimensions. Two docking
protocols were carried out: one treating the model residues as rigid (“rigid
docking protocol”) and another where residues in the vicinity of the in-
hibitor binding pocket were allowed flexibility during the docking search
algorithm (“flexible docking protocol”). In the flexible docking protocol,
residues N/D68, K165, F250, L253, and K355 were allowed flexibility. Both
docking protocols yielded similar results.
Molecular docking to our second generation models based on bCLC-K was
carried out using Autodock Vina. Preliminary docking trials identified the
region near N68 and F250 in both the c1 and c2 models discussed above. Rigid
and flexible docking protocols were carried out using a search box of 28 ×
28 × 28 Å that encompasses this region in our protein models for CLC-Ka and
CLC-Kb. Two flexible docking protocols were performed. As a direct com-
parison with our original homology model, the same five residues were
allowed flexibility: N/D68, K165, F250, L253, and K355. In the second flexible
To calculate percentage inhibition of CLC-K current, currents measured in
response to the voltage steps (steps that follow the prepulse to +60 mV)
(protocols are depicted in Fig. 2) were analyzed. An average of 15 ms of
data, taken from within the last 25 ms of each of the current traces, was
determined for currents measured before addition of inhibitor (“initial”),
after addition of inhibitor (“inhibited”), after washout of inhibitor (“wash-
Koster et al.
PNAS Latest Articles
|
9 of 10

docking protocol, 13 (CLC-Ka) or 14 (CLC-Kb) residues were allowed flexi-
bility: N/D68, I/V71, E72 (CLC-Kb), V75, L155, L163, K165, S258, Q260, S353,
M354, K355, L358 , M427. These residues were chosen based on the results of
our rigid docking protocol to our bCLC-K models together with our experi-
mental site-directed mutagenesis data. The top-scoring BIM1 docking pose
for the c1 CLC-Ka model predicts BIM1 interactions with N68 and K165 as
validated through site-directed mutagenesis, and the remaining residues are
those within 8 Å of the bound molecule for the top-scoring pose. The second
flexible docking protocol was performed only on the c1 CLC-Ka and CLC-Kb
models. The top nine poses (with the lowest docking energies) from each set
of docking protocols were compared as described in the text and Fig. S6.
(Table S5) are the preferred values compiled by Hansch et al. (40) for the
neutral form of each molecule unless otherwise noted. For comparison and
validation of the cLogP calculation methods, values were calculated for mole-
cules with known experimental LogP values in the test set reported in Table S5.
ACKNOWLEDGMENTS. We thank Chien-Ling Lee for performing the experi-
ments to test sensitivity of CLC-1 and CLC-2 to BIM1. We also thank Dr. Thomas
Jentsch for providing CLC-Ka, CLC-Kb, Barttin, CLC-1, and CLC-2 DNA. This re-
search was supported by American Heart Association (AHA) Grants
10GRNT3890045 (to M.M.) and 1GRNT16940072 (to M.M.) and a Stanford Bio-
X Seed Grant (to M.M. and J.D.B.). A.K.K. was supported by the Stanford Center
for Molecular Analysis and Design (CMAD) and a Stanford Interdisciplinary Grad-
uate Fellowship (SIGF) through the Stanford Chemistry, Engineering, & Medicine
for Human Health (ChEM-H) Institute. C.A.P.W. was supported by an HHMI un-
dergraduate fellowship and a Stanford Graduate Fellowship in Science and En-
gineering. R.T.-T. is the recipient of a Graduate Research Fellowship from the
National Science Foundation. T.S.C. was supported by a Stanford School of Med-
icine Dean’s Postdoctoral Fellowship and AHA Grant 17POST33670553. J.A.
thanks Verket För Innovationssystem for financial support.
Partition Coefficient Analysis. Octanol/water partition coefficients (cLogP) were
calculated by three different methods for the protonated form of each mol-
ecule (Table 1): ChemDraw software suite (version 14.0), MolInspiration (www.
molinspiration.com; 2017), and Schrödinger QikProp (version 2017–3). For
Schrödinger QikProp, molecular structures were first optimized using the
OPLS3 force field (54). Experimental LogP values for related heterocycles
1. Jentsch TJ (2015) Discovery of CLC transport proteins: Cloning, structure, function and
pathophysiology. J Physiol 593:4091–4109.
2. Stölting G, Fischer M, Fahlke C (2014) CLC channel function and dysfunction in health
and disease. Front Physiol 5:378.
3. Stauber T, Weinert S, Jentsch TJ (2012) Cell biology and physiology of CLC chloride
channels and transporters. Compr Physiol 2:1701–1744.
29. Liantonio A, et al. (2008) Molecular switch for CLC-K Cl- channel block/activation:
Optimal pharmacophoric requirements towards high-affinity ligands. Proc Natl Acad
Sci USA 105:1369–1373.
30. Liantonio A, et al. (2016) Kidney CLC-K chloride channels inhibitors: Structure-based
studies and efficacy in hypertension and associated CLC-K polymorphisms. J Hypertens
34:981–992.
4. Jentsch TJ (2008) CLC chloride channels and transporters: From genes to protein
structure, pathology and physiology. Crit Rev Biochem Mol Biol 43:3–36.
5. Abeyrathne PD, Chami M, Stahlberg H (2016) Biochemical and biophysical approaches
to study the structure and function of the chloride channel (ClC) family of proteins.
Biochimie 128-129:154–162.
6. Miller C (2015) In the beginning: A personal reminiscence on the origin and legacy of
ClC-0, the ‘Torpedo Cl(-) channel’. J Physiol 593:4085–4090.
7. Accardi A (2015) Structure and gating of CLC channels and exchangers. J Physiol 593:
4129–4138.
31. Hartzell HC (2010) Chloride channels: An historical perspective. Physiology and
Pathology of Chloride Transporters and Channels in the Nervous System: From
Molecules to Diseases, eds Alvarez-Leefmans F, Delpire E (Academic, San Diego), 1st
Ed, pp 3–15.
32. Feng L, Campbell EB, Hsiung Y, MacKinnon R (2010) Structure of a eukaryotic CLC
transporter defines an intermediate state in the transport cycle. Science 330:635–641.
33. Markovic S, Dutzler R (2007) The structure of the cytoplasmic domain of the chloride
channel ClC-Ka reveals a conserved interaction interface. Structure 15:715–725.
34. Gradogna A, Pusch M (2010) Molecular pharmacology of kidney and inner ear CLC-K
chloride channels. Front Pharmacol 1:130.
8. Zifarelli G, Pusch M (2007) CLC chloride channels and transporters: A biophysical and
physiological perspective. Reviews of Physiology, Biochemistry and Pharmacology, ed
Amara SG (Springer, Berlin), pp 23–76.
35. Picollo A, et al. (2004) Molecular determinants of differential pore blocking of kidney
CLC-K chloride channels. EMBO Rep 5:584–589.
9. Miller C (2006) ClC chloride channels viewed through a transporter lens. Nature 440:
484–489.
36. Liantonio A, et al. (2006) Activation and inhibition of kidney CLC-K chloride channels
by fenamates. Mol Pharmacol 69:165–173.
10. Maduke M, Miller C, Mindell JA (2000) A decade of CLC chloride channels: Structure,
mechanism, and many unsettled questions. Annu Rev Biophys Biomol Struct 29:411–438.
11. Poroca DR, Pelis RM, Chappe VM (2017) ClC channels and transporters: Structure,
physiological functions, and implications in human chloride channelopathies. Front
Pharmacol 8:151.
12. Zaika O, Tomilin V, Mamenko M, Bhalla V, Pochynyuk O (2016) New perspective of ClC-Kb/2
Cl- channel physiology in the distal renal tubule. Am J Physiol Renal Physiol 310:F923–F930.
13. Imbrici P, et al. (2015) ClC-1 chloride channels: State-of-the-art research and future
challenges. Front Cell Neurosci 9:156.
14. Verkman AS, Galietta LJV (2009) Chloride channels as drug targets. Nat Rev Drug
Discov 8:153–171.
15. Suh KS, Yuspa SH (2005) Intracellular chloride channels: Critical mediators of cell vi-
ability and potential targets for cancer therapy. Curr Pharm Des 11:2753–2764.
16. Simon DB, et al. (1997) Mutations in the chloride channel gene, CLCNKB, cause
Bartter’s syndrome type III. Nat Genet 17:171–178.
37. Gradogna A, Pusch M (2013) Alkaline pH block of CLC-K kidney chloride channels
mediated by a pore lysine residue. Biophys J 105:80–90.
38. Pace CN, Grimsley GR, Scholtz JM (2009) Protein ionizable groups: pK values and their
contribution to protein stability and solubility. J Biol Chem 284:13285–13289.
39. Park E, Campbell EB, MacKinnon R (2017) Structure of a CLC chloride ion channel by
cryo-electron microscopy. Nature 541:500–505.
40. Hansch C, Leo A, Hoekman D (1995) Exploring QSAR: Hydrophobic, Electronic, and
Steric Constants (Am Chem Soc, Washington, DC).
41. Martinez GQ, Maduke M (2008) A cytoplasmic domain mutation in ClC-Kb affects
long-distance communication across the membrane. PLoS One 3:e2746.
42. Sambrook J, Russell DW (2006) Purification of nucleic acids by extraction with phenol:
chloroform. CSH Protoc 2006:pdb.prot4455.
43. Lorenz C, Pusch M, Jentsch TJ (1996) Heteromultimeric CLC chloride channels with
novel properties. Proc Natl Acad Sci USA 93:13362–13366.
44. Venkatachalan SP, et al. (2007) Optimized expression vector for ion channel studies in
Xenopus oocytes and mammalian cells using alfalfa mosaic virus. Pflugers Arch 454:155–163.
45. Rychkov GY, Pusch M, Roberts ML, Jentsch TJ, Bretag AH (1998) Permeation and block of
the skeletal muscle chloride channel, ClC-1, by foreign anions. J Gen Physiol 111:653–665.
46. Jaroszewski L, Rychlewski L, Godzik A (2000) Improving the quality of twilight-zone
alignments. Protein Sci 9:1487–1496.
17. Uchida S (2000) Physiological role of CLC-K1 chloride channel in the kidney. Nephrol
Dial Transplant 15(Suppl 6):14–15.
18. Matsumura Y, et al. (1999) Overt nephrogenic diabetes insipidus in mice lacking the
CLC-K1 chloride channel. Nat Genet 21:95–98.
19. Denton JS, Pao AC, Maduke M (2013) Novel diuretic targets. Am J Physiol Renal
Physiol 305:F931–F942.
47. Gradogna A, Fenollar-Ferrer C, Forrest LR, Pusch M (2012) Dissecting a regulatory
20. Buffington MA, Abreo K (2016) Hyponatremia: A review. J Intensive Care Med 31:
223–236.
calcium-binding site of CLC-K kidney chloride channels. J Gen Physiol 140:681–696.
48. Sali A, Blundell TL (1993) Comparative protein modelling by satisfaction of spatial
ꢀ
21. McGreal K, Budhiraja P, Jain N, Yu ASL (2016) Current challenges in the evaluation
and management of hyponatremia. Kidney Dis (Basel) 2:56–63.
22. Estévez R, et al. (2001) Barttin is a Cl- channel β-subunit crucial for renal Cl- re-
absorption and inner ear K+ secretion. Nature 414:558–561.
23. Thompson CH, et al. (2009) Isolation and characterization of a high affinity peptide
inhibitor of ClC-2 chloride channels. J Biol Chem 284:26051–26062.
24. Salhany JM (1996) Allosteric effects in stilbenedisulfonate binding to band 3 protein
(AE1). Cell Mol Biol 42:1065–1096.
25. Romero MF, Chen A-P, Parker MD, Boron WF (2013) The SLC4 family of bicarbonate
(HCO3−) transporters. Mol Aspects Med 34:159–182.
26. Cabantchik ZI, Greger R (1992) Chemical probes for anion transporters of mammalian
cell membranes. Am J Physiol 262:C803–C827.
restraints. J Mol Biol 234:779–815.
49. Wallner B, Elofsson A (2003) Can correct protein models be identified? Protein Sci 12:
1073–1086.
50. Morris AL, MacArthur MW, Hutchinson EG, Thornton JM (1992) Stereochemical
quality of protein structure coordinates. Proteins 12:345–364.
51. Hooft RWW, Vriend G, Sander C, Abola EE (1996) Errors in protein structures. Nature
381:272.
52. Morris GM, et al. (2009) AutoDock4 and AutoDockTools4: Automated docking with
selective receptor flexibility. J Comput Chem 30:2785–2791.
53. Trott O, Olson AJ (2010) AutoDock Vina: Improving the speed and accuracy of
docking with a new scoring function, efficient optimization, and multithreading.
J Comput Chem 31:455–461.
27. Miller C, White MM (1984) Dimeric structure of single chloride channels from Torpedo
electroplax. Proc Natl Acad Sci USA 81:2772–2775.
28. Matulef K, et al. (2008) Discovery of potent CLC chloride channel inhibitors. ACS Chem
Biol 3:419–428.
54. Harder E, et al. (2016) OPLS3: A force field providing broad coverage of drug-like
small molecules and proteins. J Chem Theory Comput 12:281–296.
55. Fieller EC (1940) The biological standardization of insulin. Suppl J R Stat Soc 7:1–64.
56. Motulsky H (2007) Intuitive Biostatistics (Oxford Univ Press, New York).
10 of 10
|
www.pnas.org/cgi/doi/10.1073/pnas.1720584115
Koster et al.