Misregulation of Wnt signaling is common in human cancer. The development of small molecule inhibitors against the Wnt receptor, frizzled (FZD), may have potential in cancer therapy. During small molecule screens, we observed binding of carbamazepine to the cysteine-rich domain (CRD) of the Wnt receptor FZD8 using surface plasmon resonance (SPR). Cellular functional assays demonstrated that carbamazepine can suppress FZD8-mediated Wnt/β-catenin signaling. We determined the crystal structure of the complex at 1.7 Å resolution, which reveals that carbamazepine binds at a novel pocket on the FZD8 CRD. The unique residue Tyr52 discriminates FZD8 from the closely related FZD5 and other FZDs for carbamazepine binding. The first small molecule-bound FZD structure provides a basis for anti-FZD drug development. Furthermore, the observed carbamazepine-mediated Wnt signaling inhibition may help to explain the phenomenon of bone loss and increased adipogenesis in some patients during long-term carbamazepine treatment.
Misregulation of Wnt signaling is common in humancancer. The development of small molecule inhibitors against the Wnt receptor, frizzled (FZD), may have potential in cancer therapy. During small molecule screens, we observed binding of carbamazepine to the cysteine-rich domain (CRD) of the Wnt receptor FZD8 using surface plasmon resonance (SPR). Cellular functional assays demonstrated that carbamazepine can suppress FZD8-mediated Wnt/β-catenin signaling. We determined the crystal structure of the complex at 1.7 Å resolution, which reveals that carbamazepine binds at a novel pocket on the FZD8 CRD. The unique residue Tyr52 discriminates FZD8 from the closely related FZD5 and other FZDs for carbamazepine binding. The first small molecule-bound FZD structure provides a basis for anti-FZD drug development. Furthermore, the observed carbamazepine-mediated Wnt signaling inhibition may help to explain the phenomenon of bone loss and increased adipogenesis in some patients during long-term carbamazepine treatment.
Ligands belonging to
the Wnt family of secreted lipoproteins play
central roles in tissue morphogenesis and homeostasis through binding
to members of the frizzled (FZD) family of cell surface receptors.[1] Overexpression of FZD proteins has been observed
in cancers,[2,3] and FZD8 has been proposed as a therapeutic
target in humanlung cancer[4] and renal
cell carcinoma.[5] The anti-FZD antibodies
vantictumab (OMP-18R5), IgG-2919, and IgG-2921 have been taken into
phase 1 (ClinicalTrials.gov: NCT01345201) or preclinical trials for
cancers.[6] These antibodies usually target
multiple FZDs due to the high level of sequence conservation between
FZD proteins. For example, vantictumab interacts with FZD1, 2, 5,
7, and 8.[7] However, structural information
on antibody–FZD complexes can guide engineering to improve
specificity and avoid unwanted toxicity.[8] The development of small molecules against specific FZD receptors
has proved challenging. FZD proteins comprise an extracellular Wnt-binding
cysteine-rich domain (CRD), a seven helix transmembrane domain (TMD),
and a cytoplasmic tail. The small molecule FZM1 has been identified
as an FZD4 misfolding chaperone, which is likely to bind to the intracellular
loops of FZD4,[9] and its derivatives can
act as allosteric agonists of noncanonical Wnt signaling.[10] The recent FZD4TMD structure suggests that
FZDs are not amenable to targeting with traditional G protein-coupled
receptor (GPCR) small molecule ligands.[11] Although computer modeling can potentially be used for anti-FZD
drug development,[12,13] experimentally determined high-resolution
crystal structures of a target protein in complex with small molecule
inhibitors remain the gold standard for structure-guided drug design.
To date, no structural information about small molecules binding to
FZD receptors has been published. We have used surface plasmon resonance
(SPR) to screen a set of small molecules and report here the high-resolution
crystal structure of the extracellular FZD8cysteine-rich domain (FZD8CRD) bound with carbamazepine.
Results
Screening of
Small Molecules for Binding to FZD8CRD
Wnt ligands
bind to the FZD receptor CRD to initiate Wnt
signaling. Small molecule antagonists that bind to the FZD CRD could
therefore have therapeutic potential in cancers with upregulated Wnt
signaling.[1] We used SPR to screen for small
molecules that bind to FZD8CRD. Small molecules from the
Maybridge fragments library (58 compounds) and our own laboratory
collection (44 compounds) were screened. All of the compounds screened
are listed in Supporting Information, Table S1. Initial screening identified three hits showing SPR responses (Figure ): carbamazepine,
kahweol, and quinacrine (number 79, 87, and 99 in Supporting Information, Table S1). Further analysis of these hits with
other irrelevant protein controls eliminated quinacrine as a nonspecific
binder. Kahweol and carbamazepine were then carried forward to co-crystallization
with FZD8CRD. While kahweol did not yield a complex structure,
the structure of carbamazepine in complex with FZD8CRD was
determined (Figure ). In the absence of further evidence to support kahweol binding,
we focused all further work on carbamazepine.
Figure 1
SPR sensorgrams from
FZD8CRD-positive response compounds.
The biotinylated FZD8CRD protein (1000 response units)
was immobilized on a streptavidin (SA) sensor chip. Analyte compounds
carbamazepine (A), kahweol (B), and quinacrine (C) are in two concentrations
(50 and 5 μM).
Figure 2
Overall structure of
apo and carbamazepine-bound FZD8CRD and structural comparison.
(A) Four molecules of the apo structure
of FZD8CRD in the asymmetric unit (ASU) (PDB code 6TFM). (B) Two molecules
of FZD8CRD in complex with carbamazepine in the ASU (PDB
code 6TFB).
(C) Cartoon representation of FZD8CRD, rainbow-colored
from N-(blue) to C-(red) terminus. Residues originating from the Rhinovirus
3C cleavage site are colored in gray. In both our apo and carbamazepine-bound
crystal structures, these additional residues contribute to an antiparallel
β-strand (β2), which stabilizes β1. The bound carbamazepine
is shown as gray sticks (PDB code 6TFB). (D) Close-up view of loop_6 from the
two aligned carbamazepine-bound FZD8CRD molecules (gray
and brown) superimposed on two representative apo FZD8CRD molecules from the ASU (green and cyan, PDB code 6TFM). (E) Alignment
of two FZD8CRD copies from previously published complex
structures with Wnt8 (blue, PDB code 4F0A) and Wnt3 (red, PDB code 6AHY). The Wnt index
fingers are shown as Cα traces.
SPR sensorgrams from
FZD8CRD-positive response compounds.
The biotinylated FZD8CRD protein (1000 response units)
was immobilized on a streptavidin (SA) sensor chip. Analyte compounds
carbamazepine (A), kahweol (B), and quinacrine (C) are in two concentrations
(50 and 5 μM).Overall structure of
apo and carbamazepine-bound FZD8CRD and structural comparison.
(A) Four molecules of the apo structure
of FZD8CRD in the asymmetric unit (ASU) (PDB code 6TFM). (B) Two molecules
of FZD8CRD in complex with carbamazepine in the ASU (PDB
code 6TFB).
(C) Cartoon representation of FZD8CRD, rainbow-colored
from N-(blue) to C-(red) terminus. Residues originating from the Rhinovirus
3C cleavage site are colored in gray. In both our apo and carbamazepine-bound
crystal structures, these additional residues contribute to an antiparallel
β-strand (β2), which stabilizes β1. The bound carbamazepine
is shown as gray sticks (PDB code 6TFB). (D) Close-up view of loop_6 from the
two aligned carbamazepine-bound FZD8CRD molecules (gray
and brown) superimposed on two representative apo FZD8CRD molecules from the ASU (green and cyan, PDB code 6TFM). (E) Alignment
of two FZD8CRD copies from previously published complex
structures with Wnt8 (blue, PDB code 4F0A) and Wnt3 (red, PDB code 6AHY). The Wnt index
fingers are shown as Cα traces.
Structure of FZD8CRD and Its Complex with Carbamazepine
The amino acid sequence of FZD8CRD is fully conserved
between human and mouse. Therefore, although we used a mouse cDNA
sequence, the resulting structure is identical to humanFZD8CRD. Thus, we refer to it simply as FZD8CRD hereafter. The
apo FZD8CRD structure was determined at 2.3 Å resolution
in space group P21 with four molecules
in the asymmetric unit (ASU, Figure A). The carbamazepine complex structure (1.7 Å
resolution) crystallized in the same condition (see the Experimental Section) with similar lattice contacts to the
apo structure, but with higher crystallographic symmetry resulting
in space group P212121, with two
molecules in each ASU (Figure B).The overall structure of FZD8CRD, either
apo or in complex with carbamazepine, is almost identical to those
of previously reported apo or Wnt-bound FZD8CRD[14−16] except at the CRD C-terminus, which forms a β-hairpin with
the remaining residues of a Rhinovirus 3C protease cleavage site (used
to remove purification tags; Figure C and the Experimental Section). In both our structures, the ASUs contain dimers resulting from
two-fold noncrystallographic symmetry (Figure A,B). This dimeric arrangement, as well as
the other lattice packing interactions, differs from the distinctive
dimer, mediated by unsaturated fatty acyl–FZD CRD binding,
that has been observed for a number of FZD CRD structures, including
FZD8.[15,17−19]In the apo structure,
the four molecules of FZD8CRD in
the ASU fall into two conformations (“a” and “b”, Figure D) regarding loop_6
(residues R137 to L147). However, both copies of the carbamazepine-bound
FZD8CRD are in conformation “b” (Figure D). This suggests
that carbamazepine may stabilize conformation “b” upon
binding. The XenopusWnt8-[15] or humanWnt3-[16] bound FZD8CRD loop_6 corresponds
most closely to conformation “a” (Figure E). This suggests that Wnt binding may prefer
the loop_6 conformation “a”.
Figure 3
Carbamazepine chemical
structure, electron density, and binding
pocket. (A) Chemical structure of carbamazepine with atom numbers
indicated. (B) Simulated annealing |Fo – Fc| omit electron density map
for carbamazepine contoured at 2.5σ from one molecule from the
ASU (the other has similar electron density and is not shown, PDB
code 6TFB).
(C) Surface representation of FZD8CRD with areas interacting
with carbamazepine (red), Wnt index finger (site 2, green), Wnt thumb
(site 1, cyan), and lipid palmitoleic acid (PAM, blue). (D) Electrostatic
properties of FZD8CRD. The protein surface is colored by
the electrostatic potential at ±5 kT/e (red, acidic; blue, basic;
carbamazepine shown as spheres: yellow, carbon; blue, nitrogen; and
red, oxygen).
Carbamazepine chemical
structure, electron density, and binding
pocket. (A) Chemical structure of carbamazepine with atom numbers
indicated. (B) Simulated annealing |Fo – Fc| omit electron density map
for carbamazepine contoured at 2.5σ from one molecule from the
ASU (the other has similar electron density and is not shown, PDB
code 6TFB).
(C) Surface representation of FZD8CRD with areas interacting
with carbamazepine (red), Wnt index finger (site 2, green), Wnt thumb
(site 1, cyan), and lipidpalmitoleic acid (PAM, blue). (D) Electrostatic
properties of FZD8CRD. The protein surface is colored by
the electrostatic potential at ±5 kT/e (red, acidic; blue, basic;
carbamazepine shown as spheres: yellow, carbon; blue, nitrogen; and
red, oxygen).
Carbamazepine Binds FZD8CRD at a Novel Pocket
Carbamazepine (5H-dibenzo[b,f]azepine-5-carboxamide),
sold under the trade names Tegretol, Equetro, Carbatrol, Epitol, and
Orteril, is a tricyclic compound (Figure A). The 1.7 Å resolution complex structure
unambiguously showed carbamazepine binding to the FZD8CRD. The simulated annealing omit electron density map showed clear
electron density for all atoms of carbamazepine (Figure B) and both molecules in the
ASU show similar quality density. FZD8CRD possesses two
well-documented Wnt binding sites: a hydrophobic groove that binds
the palmitoleic acid moiety (PAM) that is appended to the Wnt thumb
(site 1) and the Wnt index finger binding site (site 2).[15,16] The carbamazepine binding pocket is sandwiched by the two Wnt binding
sites (Figures C and 4A). This pocket is largely hydrophobic and neutral
in surface charge but is surrounded by positively and negatively charged
patches (Figure D).
Aside from Wnt, there are many FZD CRD binding proteins that have
been reported, but none of them bind at this pocket. The Wnt mimic
Norrin protein binds at site 2 on FZD4CRD (Figure A),[20,21] while all other reported binders bind at or near site 1. These include
the Clostridium difficile toxin B (TcdB[22]) that binds FZD2CRD (Figure B), the Wnt surrogate B12 module
that binds FZD8CRD[23] and the
peptide FZ7-21 that binds FZD7CRD[18] (Figure C), the
DARPin module DRPB that binds FZD8CRD,[24] and the antibody Fab F2.I that binds FZD5CRD[8] (Figure D). Carbamazepine does not overlap with any of the
reported FZD CRD ligands when the various structures are superimposed
(Figure ) and instead
binds to residues located between the two Wnt binding sites in a novel
binding pocket (Figures and 4).
Figure 4
Carbamazepine binding site (PDB code 6TFB) relative to other
FZD ligand binding
sites. (A) Wnt3 (PDB code 6AHY(15)) and Norrin (PDB code 5BQE(20)); (B) C. difficile toxin
B (TcdB, PDB code 6C0B(22)); (C) peptide FZ7-21(PDB code 5WBS(18)) and Wnt surrogate B12 module (PDB code 5UN5(23)); and (D) DARPin B (PDB code 6NDZ(24)) and the
Fab fragment of antibody F2.I (PDB code 6O39(8)). The carbamazepine-bound
FZD8CRD structure is shown as gray cartoon and spheres
as in Figure D. For
clarity, in all other ligand-bound structures, only our FZD8CRD–carbamazepine structure (gray ribbon and spheres) is shown.
All other FZD CRDs are hidden from view.
Carbamazepine binding site (PDB code 6TFB) relative to other
FZD ligand binding
sites. (A) Wnt3 (PDB code 6AHY(15)) and Norrin (PDB code 5BQE(20)); (B) C. difficile toxin
B (TcdB, PDB code 6C0B(22)); (C) peptide FZ7-21(PDB code 5WBS(18)) and Wnt surrogate B12 module (PDB code 5UN5(23)); and (D) DARPin B (PDB code 6NDZ(24)) and the
Fab fragment of antibody F2.I (PDB code 6O39(8)). The carbamazepine-bound
FZD8CRD structure is shown as gray cartoon and spheres
as in Figure D. For
clarity, in all other ligand-bound structures, only our FZD8CRD–carbamazepine structure (gray ribbon and spheres) is shown.
All other FZD CRDs are hidden from view.The carbamazepine binding pocket comprises residues from helix
α3 (S90, M91), loop_2 (Y52, Q56), loop_4 (P94, P103, L104, P105,
P106), and loop_6 (R137). The residues L104 and R137 use their main
chain atoms to interact with carbamazepine, while the other interacting
residues use their side chains (Figure A). A distinctive feature of this pocket is a cluster
of four hydrophobic prolines. The interacting residues from the two
copies of FZD8CRD in the crystallographic ASU (“a”
and “b”) show similar interactions with carbamazepine,
except for the residue Q56. Q56 of molecule “a” interacts
with a carbamazepinenitrogen (N17) atom and azepine ring (C7, Figure B), while in the
other copy, “b”, Q56 interacts with the carbamazepineazepine and benzyl rings (Figure C). The difference in the Q56 side-chain conformation
may be due to crystal packing, as this area of molecule “a”
contacts a neighbor packing molecule. When two representative copies
of apo structures are aligned with carbamazepine-bound structures,
the Q56 side chain from the apo structure would sterically hinder
carbamazepine binding (Figure D). This suggests that Q56 may undergo a conformational change
upon carbamazepine binding. The carbamazepine–protein interactions
are mainly hydrophobic, especially those involving P105 and P103,
which form extensive hydrophobic interactions with the azepine ring
and two benzyl rings of carbamazepine (Figure E). Y52 forms a hydrogen bond through a water
molecule to the nitrogen (N17) of the carboxamide head of carbamazepine
(Figure A). There
are three additional ordered water molecules within the pocket that
are conserved between the two molecules in the ASU. The positions
of water molecules could guide carbamazepine modification to develop
more potent FZD8 inhibitors.
Figure 5
Interactions between carbamazepine and FZD8CRD. (A)
Details of the carbamazepine binding site (gray cartoon, from molecule
“a” of the ASU, PDB code 6TFB). The carbamazepine interacting amino
acid side chains are shown as gray sticks. Water molecules within
the pocket are shown as magenta balls, carbamazepine as salmon sticks,
and hydrogen bonds as yellow dashed lines. (B–D) Conformations
of Q56 from two carbamazepine binding molecules of the ASU, “a”
(gray, B) and “b” (brown, C), and superimposed with
two representative molecules from apo structures, (green and cyan;
D). Interactions are defined as distances between protein and carbamazepine
of less than 3.9 Å. The hydrophobic interactions are shown as
green dashed lines. (E) LIGPLOT[25] of FZD8CRD–carbamazepine interactions. The green lines indicate
hydrophobic interactions. Carbamazepine carbon atoms are shown as
black, nitrogen as blue, and oxygen as red spheres. Only molecule
“a” from the ASU is shown. (F) Sequence alignment of
human FZDs and sFRPs, with carbamazepine interacting FZD8 residues
and all matching residues highlighted. The residue numbers indicated
are for mouse FZD8.
Interactions between carbamazepine and FZD8CRD. (A)
Details of the carbamazepine binding site (gray cartoon, from molecule
“a” of the ASU, PDB code 6TFB). The carbamazepine interacting amino
acid side chains are shown as gray sticks. Water molecules within
the pocket are shown as magenta balls, carbamazepine as salmon sticks,
and hydrogen bonds as yellow dashed lines. (B–D) Conformations
of Q56 from two carbamazepine binding molecules of the ASU, “a”
(gray, B) and “b” (brown, C), and superimposed with
two representative molecules from apo structures, (green and cyan;
D). Interactions are defined as distances between protein and carbamazepine
of less than 3.9 Å. The hydrophobic interactions are shown as
green dashed lines. (E) LIGPLOT[25] of FZD8CRD–carbamazepine interactions. The green lines indicate
hydrophobic interactions. Carbamazepinecarbon atoms are shown as
black, nitrogen as blue, and oxygen as red spheres. Only molecule
“a” from the ASU is shown. (F) Sequence alignment of
human FZDs and sFRPs, with carbamazepine interacting FZD8 residues
and all matching residues highlighted. The residue numbers indicated
are for mouseFZD8.
Binding Specificity and
Affinity of Carbamazepine
In
addition to the 10 FZDs, the human genome also encodes 5 secreted
FZD-related proteins (sFRPs), important Wnt regulatory proteins,[26] and all FZDs and sFRPs possess a conserved FZD
CRD. Sequence alignment of all FZD and sFRP CRDs shows that only one
proline residue (P94) is fully conserved among the residues interacting
with carbamazepine (Figure F). Another proline, P106, is conserved in all FZDs/sFRPs,
except for FZD9/10. The residue S90 is highly conserved but absent
in FZD3/6 and sFRP3/4. Other carbamazepine interacting residues are
only partially conserved among FZDs and sFRPs. FZD8 and FZD5 are closely
related (>80% sequence identity within the CRD), sharing most of
the
carbamazepine interacting residues, including Q56, which is unique
to these FZDs. However, the residue Y52 is found only in FZD8 (Figure F). The sequence
alignment suggests that of all of the other FZDs/sFRPs, FZD5 is the
most likely candidate to bind to carbamazepine, followed by FZD1,
2, and 7.We then used biophysical methods to measure the affinity
of the interaction between carbamazepine and FZD8CRD, as
well as the cross-reactivity of carbamazepine with the CRDs of FZD5
and 7. We have previously used a thermal shift assay (also known as
differential scanning fluorimetry) to investigate small molecule–protein
interactions.[27,28] However, we found that FZD CRDs
are highly thermostable (remaining folded at 95 °C), which precluded
the measurement of melting curves. We therefore turned to SPR as an
alternative method for the detection of small molecule–protein
interactions.[29,30] All three (FZD8, 5, and 7) CRD
constructs yielded correctly folded protein samples as evidenced by
their gel filtration profiles (Supporting Information, Figure S1). Biotinylated samples of each CRD
were immobilized on a streptavidin (SA) chip in separate flow chambers
with carbamazepine as an analyte. FZD8CRD has a clear concentration-dependent
response to carbamazepine (Figure ), with a calculated Kd of 17 μM. Surprisingly, carbamazepine does not interact with
the closely related FZD5CRD. This is despite FZD5CRD differing from FZD8CRD at only one carbamazepine interacting
residue, Y52, which corresponds to FZD5 H50. In fact, Y52 is unique
among all FZDs and sFRPs (Figure F), suggesting that this residue is a key determinant
of the specificity of the interaction between carbamazepine and FZD8.
As FZD7CRD shares only half of the interacting residues,
it is less surprising that it also does not bind detectably to carbamazepine
(Figure ).
Figure 6
SPR analysis
of carbamazepine interaction with FZD5, 7, 8 CRD.
Biotinylated mouse FZD5, human FZD7, and mouse FZD8 CRD were immobilized
on individual flow cells of a SA chip, respectively. A carbamazepine
concentration series was used as an analyte. The SPR sensorgrams are
shown in the right panels. ND, not detectable; RU, resonance units.
SPR analysis
of carbamazepine interaction with FZD5, 7, 8 CRD.
Biotinylated mouseFZD5, humanFZD7, and mouseFZD8 CRD were immobilized
on individual flow cells of a SA chip, respectively. A carbamazepine
concentration series was used as an analyte. The SPR sensorgrams are
shown in the right panels. ND, not detectable; RU, resonance units.
Carbamazepine Inhibits Wnt Signaling in a
Cellular Reporter
Gene Assay
We used cellular luciferase reporter assays to
assess the effects of carbamazepine on Wnt signaling. Carbamazepine
only binds to FZD8; however, the commonly used cell line (HEK293T)
expresses FZD8 at a very low level, and the HEK293T TOPFlash response
to canonical Wnt ligands is dependent mainly on FZD1, 2, and 7.[31] To observe the specific response from FZD8,
we therefore used an FZD1, 2 and 7 knockout HEK293T cell line[31] and introduced the full-length mouseFZD8 expression
cassette by lentiviral transduction.[32] The
cell line was further stably transformed with a T-cell factor/lymphoid
enhancer-binding factor 1 luciferase (TCF/LEF, TOPFlash) plasmid to
minimize reporter plasmid transfection variations. The cells were
then stimulated with a conditioned medium from mouse L-cells expressing
Wnt3a, in the presence of a carbamazepine concentration series (Figure ). Carbamazepine
starts to inhibit Wnt3a-induced TOPFlash luciferase activity at a
concentration of 8 μM (unpaired t-test, P <
0.0001), with greater inhibition seen at higher concentrations. However,
we found that carbamazepine can only partially suppress Wnt3a-induced
luciferase activity and even at the high concentration of 64 μM,
luciferase activity remained around 60% (Figure ).
Figure 7
Carbamazepine inhibits FZD8-mediated TOPFLASH
luciferase activity.
HEK293T cells with FZD1, 2, and 7 knocked out and with mouse FZD8
and TOPFLASH expression cassettes introduced were transfected with
tk-Renilla luciferase and treated with carbamazepine at the indicated
concentrations and Wnt3a for luciferase induction. Each data point
represents the Renilla luciferase-normalized firefly luciferase fold
changes in quadruplicate. Error bars represent the standard deviation
of the mean and the Student’s t-test was used to calculate
statistical significance. Gray columns indicate no statistical difference
and black columns indicate a significant difference.
Carbamazepine inhibits FZD8-mediated TOPFLASH
luciferase activity.
HEK293T cells with FZD1, 2, and 7 knocked out and with mouseFZD8
and TOPFLASH expression cassettes introduced were transfected with
tk-Renilla luciferase and treated with carbamazepine at the indicated
concentrations and Wnt3a for luciferase induction. Each data point
represents the Renilla luciferase-normalized firefly luciferase fold
changes in quadruplicate. Error bars represent the standard deviation
of the mean and the Student’s t-test was used to calculate
statistical significance. Gray columns indicate no statistical difference
and black columns indicate a significant difference.
Discussion and Conclusions
FZD proteins, as essential
Wnt receptors, are a central point for
Wnt signaling intervention in diseases such as cancer. While macromolecules
like antibodies, peptides, the FZD5/8-binding B12 protein, and DARPin
molecules targeting FZD CRDs have been described,[8,18,23,24] small molecules
may offer advantages such as being easier to manufacture, more stable,
less expensive, and having the potential to be administered orally.
Among the methods of screening small molecules for binding to target
proteins, SPR provides a measure of direct molecular interactions,
allowing an effective triage for candidates with the highest potential
for generating complex crystal structures. Structural information
forms the basis for rational drug design. We have obtained the first
FZD-small molecule structure and discovered that carbamazepine specifically
binds to the FZD8CRD and not to the closely related FZD5.
To date, no antibodies or synthetic FZD binders have been reported
that can distinguish between FZD8 and FZD5. Conceivably, carbamazepine
may offer this potential.We noticed that the carbamazepine
can only partially suppress Wnt3a-induced
luciferase activity, even at a high concentration of 64 μM (Figure ), which is in agreement
with the recently reported weak Wnt inhibitory effects in mouse adipose
cells.[33] The binding pocket that we have
identified in FZD8CRD does not overlap with the known Wnt
binding sites. Thus, the mechanism of action is allosteric. However,
the carbamazepine binding pocket may still be important for Wnt signaling.
It was previously demonstrated that the disruption of the carbamazepine
binding region (insertion of a tripeptide, GSG, before the carbamazepine
interacting residues Y52 or R137) abolished Wnt8 binding to FZD8.[14] The carbamazepine binding affinity Kd value of 17 μM and starting inhibitory concentration
of 8 μM are comparable to the reported Wnt/β-catenin inhibitory
concentration of 5–10 μM in adipose cells[33] and colon cancer cells.[34] Carbamazepine is the primary drug used to treat epilepsy. Typical
doses are between 400 and 1000 mg per day and the carbamazepine plasma
concentration in patients during treatment reaches 20–40 μM,[35] which is higher than the Kd value and inhibitory concentration we observed.Epilepsy
is a common serious neurological disorder, affecting 1–2%
of the population (over 50 million people) worldwide.[36] The etiology of epilepsy is multifactorial and the involvement
of Wnt signaling misregulation has recently received attention.[37] Wnt/β-catenin signaling may play a role
in the development of temporal lobe epilepsy.[38] Conditional knockout of the major Wnt/β-catenin negative regulator
adenomatous polyposis coli (APC) gene in mice, which results in elevated
Wnt/β-catenin signaling, causes chronic seizures[39] with similar features to those seen in humans
with infantile spasms, a common syndrome in childhood epilepsy.[40] In a mouse model, both deletion and overexpression
of β-catenin have a significant impact on seizure susceptibility.[41] The kainic acid-induced epilepsy animal model
shows upregulated Wnt/β-catenin signaling.[42] Increased Wnt/β-catenin signaling is associated with
the increased number of neuronal stimuli,[43] while decreased Wnt/β-catenin signaling by the Wnt antagonist
Dickkopf-related protein 1 (DKK1) has been shown to be able to protect
against the development of hippocampal sclerosis, which is a hallmark
of temporal lobe epilepsy.[44] The antiepileptic
drug carbamazepine (an inhibitor of voltage-gated sodium channels[45]) has been shown to decrease Wnt/β-catenin
signaling in the humancolon adenocarcinomaSW480 cell line[34] and mouse adipocyte 3T3-L1 cells.[33] The results presented here contribute further
evidence that Wnt signaling modulation may be involved in carbamazepine
treatment of epilepsy. This warrants further studies of the role of
Wnt-FZD8 signaling modulation in the therapeutic mechanism of carbamazepine
against epilepsy.It is noteworthy that long-term antiepileptic
drug (AED) treatment
can have side effects, including disorders of bone metabolism leading
to bone loss.[46−50] The exact mechanism of this pathological change is not completely
understood. It is commonly accepted that intact Wnt signaling is essential
for proper bone formation and remodeling.[51] The strength and integrity of the human skeleton depend on a delicate
equilibrium between bone resorption by osteoclasts and bone formation
by osteoblasts. Wnt/β-catenin signaling can inhibit osteoclastogenesis[52] and increase bone mass. On the other hand, inhibition
of Wnt signaling can lead to bone loss. Among the 10 FZD receptors,
FZD8 and FZD9 are particularly relevant to bone metabolism. While
FZD9 regulates osteoblast function through noncanonical Wnt signaling,[53,54] FZD8 mediates canonical Wnt/β-catenin signaling and inhibits
osteoclastogenesis.[52] The discovery that
carbamazepine can inhibit FZD8-mediated Wnt signaling may help to
explain the loss of bone density associated with long-term treatment
with carbamazepine. In addition, weight gain affecting patients treated
with carbamazepine has also been linked to Wnt signaling inhibition.[33] Carbamazepine use has also been shown to reduce
the risk of prostate cancer[55,56] and synergistically
inhibits breast cancer cell proliferation when combined with other
anticancer treatment,[57] although the exact
mechanisms remain to be investigated.In summary, we have identified
carbamazepine as a specific ligand
for the Wnt receptor FZD8 using an SPR screen of small molecules.
The high-resolution crystal structure of the complex reveals a novel
binding site in FZD8CRD that allows small molecule interactions
to discriminate between closely related FZDs. As well as potentially
explaining the loss of bone density observed in patients following
long-term treatment with carbamazepine, our carbamazepine–FZD8CRD structure also provides a new avenue to explore the design
and development of FZD specific inhibitors.
Experimental
Section
Protein Production and Crystallization
MouseFZD8 (UniProtKB
Q61091) residues A28-R153 (identical amino acid sequence to humanFZD8CRD) with glycosylation site mutations (N49E and N152E)
were PCR amplified from the template cDNA (Source Bioscience, Nottingham
U.K.; clone ID 40130820) and cloned into a stable cell line vector
pNeo-sec.[58] with a Rhinovirus 3C cleavage
site, a monoVenus fluorescent protein, and 6xHis tags. The resulting
3C protease treated protein contains two additional amino acids (GT)
from KpnI cloning site and six amino acids (LEVLFQ) from the 3C cleavage
site at the C-terminus. HEK293S GnTI– cells (ATCC CRL-3022)
were co-transfected with this plasmid and a PhiC31 integrase expression
vector (pCB92/pgk-φC31).[59] The polyclonal
population of cells following G418 (1 mg/mL) selection was cultured
in a CompacT automated cell culture system.[60] The secreted proteins were captured by Talon Co2+ affinity
resin (Takara Bio Europe SAS) after the conditional media were dialyzed
against PBS. The protein-bound Talon beads were washed with 20 mM
imidazole in PBS and eluted with 300 mM imidazole in PBS. The eluted
fusion protein was treated with Rhinovirus 3C protease. After reverse
tag capture with Talon beads (to remove the monoVenus and His tags),
the protein was further purified by size-exclusion chromatography
using a Superdex 200 16/60 column (GE Healthcare) and concentrated
to 10 mg·mL–1 in 10 mM HEPES, pH 7.4, 50 mM
NaCl buffer. Crystallization screening was carried out using the sitting-drop
vapor diffusion method in 96-well plates, and crystals from both FZD8CRD protein alone or mixed with carbamazepine (∼10 μg.ML–1) grew in a condition containing 0.1 M sodium acetate,
pH 5, and 1 M ammonium sulfate.
Data Collection and Structure
Determination
Crystals
were flash-frozen by immersion in a reservoir solution supplemented
with 25% v/v glycerol followed by transferring to liquid nitrogen.
Data sets were recorded from crystals at 100 K at the Diamond Light
source (Didcot U.K., beamline I04-1) and processed with Xia2.[61] Structures were determined by molecular replacement
using Molrep (CCP4) using the previously reported FZD8CRD structure (PDB code 1IJY(14)) as a search model. The
model was then manually built with COOT[62] and refined with Phenix. Data collection and refinement statistics
are shown in Table . We noticed that the Rfree for both of the structures
are relatively high, and we then used Zanuda[63] to check if the space group for each structure was misassigned;
refinement in lower symmetry space groups (P1 and P21 for the apo and complex structures, respectively)
did not result in significantly lower R factors and both structures
passed Zanuda tests. The PyMOL Molecular Graphics System (Schrödinger,
LLC.) was used for preparing figures.
Table 1
Data Collection
and Refinement Statisticsa
data collection
data set
FZDCRD
FZDCRD-carbamazepine
PDB code
6TFM
6TFB
wavelength (Å)
0.916
0.916
space group
P21
P21 21 21
unit cell dimensions
(Å)
a = 52.3, b = 66.0, c = 72.7; α = γ = 90°, β = 90.1°
a = 52.0, b = 68.1, c = 73.8; α = β = γ = 90°
resolution (Å)
72.7–2.34 (2.38–2.34)
52–1.68 (1.71–1.68)
unique reflections
20741 (1024)
25589 (1212)
Rmerge
0.19 (0.83)
0.13 (---)
< I >
/ < σI>
7.3 (2.3)
9.4 (1.0)
CC half
0.99 (0.89)
0.99 (0.84)
completeness (%)
98.9 (98.5)
99.2 (95.3)
redundancy
6.5 (5.9)
11.5 (8.5)
Numbers in parentheses
refer to
the highest resolution shell of data.
Numbers in parentheses
refer to
the highest resolution shell of data.
SPR Equilibrium Binding Studies
HumanFZD5 (UniProtKB
- Q13467) CRD (P31- Y150), humanFZD7 (UniProtKB - O75084) CRD (Q33-G173),
and mouseFZD8CRD (A28-R153) were cloned into a pHL-Avi3
vector.[64] To produce biotinylated proteins,
these plasmids were co-transfected with the pDisplay_BirA-ER plasmid[65] into HEK293T cells with the media supplemented
with 20 μM biotin. This procedure allows in vivo biotinylation
to occur.[65] The dialyzed conditioned media
were directly used for the immobilization of ligands. The affinity
was measured at 25 °C in 10 mM HEPES, pH 7.4, 150 mM NaCl, 0.005%
Tween20, and 2% DMSO using a Biacore S200 machine (GE Healthcare).
The biotinylated ligands (1000 RU each) were coupled to a SA sensor
chip (GE Healthcare), and the analyte carbamazepine was tested using
a two-fold serial dilution. The response was plotted against the analyte
concentration and fitted by nonlinear regression to a one-site saturation
binding model (Sigma Plot, Systat software, Inc.)
Cellular Wnt
Signaling TOPFlash Luciferase Assay
The
HEK293T cell line with FZD1/2/7 knocked out[31] was a kind gift from Professor Michel Boutrous (DFKZ, Germany).
The mouse full-length FZD8 was introduced into this cell line using
a lentiviral vector with an IRES-GFP sequence after the expression
cassette.[32] The resulting cell line was
further transformed using a Super TOPFlash firefly luciferase[66] gene expression cassette in a stable cell line
vector, followed by blasticidine selection. The polyclonal population
from the blasticidine selection (20 μg mL–1) was seeded in a 96-well plate (105 cells/well) and transfected
with a constitutive Renilla luciferase plasmid (pRL-tk; Promega) at
a concentration of 10 ng mL–1 with lipofectamine
2000 (Invitrogen). Twenty-four hours after transfection, the media
were replaced with carbamazepine dilution series diluted in Wnt3a
conditional media from the Wnt3a producing mouse L Wnt-3A cell line
(ATCC CRL-2647, ref (67)). The culture medium from L-cells served as the Wnt3a control medium.
The firefly and Renilla luciferase activities were measured 24 h later
using the Dual-Glo luciferase reporter assay system (Promega) with
an Ascent Lunimoskan luminometer (Labsystems). The firefly luciferase
activity was normalized with the constitutive Renilla luciferase activity.
Authors: Yuguang Zhao; William Mahy; Nicky J Willis; Hannah L Woodward; David Steadman; Elliott D Bayle; Benjamin N Atkinson; James Sipthorp; Luca Vecchia; Reinis R Ruza; Karl Harlos; Fiona Jeganathan; Stefan Constantinou; Artur Costa; Svend Kjær; Magda Bictash; Patricia C Salinas; Paul Whiting; Jean-Paul Vincent; Paul V Fish; E Yvonne Jones Journal: ACS Chem Neurosci Date: 2022-06-22 Impact factor: 5.780
Authors: Alexandra McMellen; Elizabeth R Woodruff; Bradley R Corr; Benjamin G Bitler; Marisa R Moroney Journal: Int J Mol Sci Date: 2020-06-16 Impact factor: 5.923
Authors: Elliott D Bayle; Fredrik Svensson; Benjamin N Atkinson; David Steadman; Nicky J Willis; Hannah L Woodward; Paul Whiting; Jean-Paul Vincent; Paul V Fish Journal: J Med Chem Date: 2021-03-30 Impact factor: 7.446