Allosteric binding site in a Cys-loop receptor ligand-binding domain unveiled in the crystal structure of ELIC in complex with chlorpromazine

Significance

Cys-loop receptors belong to a family of ion channels that are involved in fast synaptic transmission. Allosteric modulators of Cys-loop receptors hold therapeutic potential as they tweak receptor function while preserving the normal fluctuations in neurotransmitter signaling at the synapse. Here, we take advantage of a model Cys-loop receptor, the Erwinia ligand-gated ion channel (ELIC). We determined cocrystal structures of ELIC in complex with chlorpromazine (IC₅₀, ∼160 μM) and its brominated derivative bromopromazine, which unveil an allosteric binding site localized at the interface between the extracellular ligand-binding domain and the pore-forming transmembrane domain. Our results demonstrate that the different allosteric binding sites present in Cys-loop receptors form an almost continuous path stretching from top to bottom of the receptor.

Abstract

Pentameric ligand-gated ion channels or Cys-loop receptors are responsible for fast inhibitory or excitatory synaptic transmission. The antipsychotic compound chlorpromazine is a widely used tool to probe the ion channel pore of the nicotinic acetylcholine receptor, which is a prototypical Cys-loop receptor. In this study, we determine the molecular determinants of chlorpromazine binding in the Erwinia ligand-gated ion channel (ELIC). We report the X-ray crystal structures of ELIC in complex with chlorpromazine or its brominated derivative bromopromazine. Unexpectedly, we do not find a chlorpromazine molecule in the channel pore of ELIC, but behind the β8–β9 loop in the extracellular ligand-binding domain. The β8–β9 loop is localized downstream from the neurotransmitter binding site and plays an important role in coupling of ligand binding to channel opening. In combination with electrophysiological recordings from ELIC cysteine mutants and a thiol-reactive derivative of chlorpromazine, we demonstrate that chlorpromazine binding at the β8–β9 loop is responsible for receptor inhibition. We further use molecular-dynamics simulations to support the X-ray data and mutagenesis experiments. Together, these data unveil an allosteric binding site in the extracellular ligand-binding domain of ELIC. Our results extend on previous observations and further substantiate our understanding of a multisite model for allosteric modulation of Cys-loop receptors.

Chlorpromazine (CPZ) (Fig. 1), a phenothiazine-derived antipsychotic drug, was introduced in psychiatry in the early 1950s, revolutionizing the treatment of psychotic disorders (1, 2). The main mechanism of action of CPZ consists in the blockage of dopamine receptors (2–4), but the numerous side effects associated with this drug indicate that it interacts with other physiologically relevant targets. CPZ was indeed shown to interfere with several voltage- and ligand-gated channels: it inhibits neuronal voltage-gated K⁺ channels (5–7), BK_Ca channels (8), and the human α_1E subunit-mediated Ca²⁺ channels (9); CPZ was also shown to inhibit GABAergic currents (10, 11), specifically through GABA_A receptors (GABA_ARs) (12), and to inhibit serotonin type-3 receptors (5-HT₃Rs) (13, 14) and nicotinic acetylcholine receptors (nAChRs) (15, 16), members of the Cys-loop receptor family.

Fig. 1.

Structure and function of chlorpromazine (CPZ) and analogs. (A–C) Chemical structures of CPZ, bromopromazine (BrPZ), and methanethiosulfonate-promazine (MTS-PZ), respectively. (D) Electrophysiological recordings from Xenopus oocytes expressing ELIC. Channels were activated by the application of the agonist GABA at the EC₅₀ (20 mM). In the presence of 30 μM CPZ, this response was reduced. (E) Concentration–inhibition curve for CPZ on ELIC. Averaged data ± SEM are shown for three to nine different oocytes.

The Cys-loop receptor family is composed of membrane-spanning ligand-gated ion channels that are responsible for fast excitatory or inhibitory synaptic neurotransmission. They are composed of five identical or nonidentical subunits, each of them comprising an N-terminal extracellular domain, which contains the neurotransmitter binding site, four transmembrane helices, that when assembled allow ions to pass through the membrane, and an intracellular domain, responsible for channel conductance, receptor modulation, and trafficking (17, 18). Initial structural insight into the mechanism of Cys-loop receptor function derives from cryo-EM images of the Torpedo marmorata nAChR (19–22) as well X-ray crystal structures of the acetylcholine binding protein (AChBP) (23, 24). AChBPs are water-soluble homologs of the extracellular ligand-binding domain of the nAChR and lack the pore-forming transmembrane domain. To date, more than 100 cocrystal structures of AChBP in complex with different agonists, partial agonists, antagonists, and allosteric modulators have been determined, creating a wealth of information on the molecular determinants of ligand recognition in nAChRs (25). Subsequently, the identification of Cys-loop receptors in prokaryotes (26) allowed the first X-ray structure determination of integral Cys-loop receptors Erwinia ligand-gated ion channel (ELIC) (27) and Gloeobacter ligand-gated ion channel (GLIC) (28, 29), which likely represent a nonconducting and conducting conformation of the channel pore, respectively. Later on, X-ray crystal structures were determined for the first eukaryote Cys-loop receptors, including the Caenorhabditis elegans glutamate-gated chloride channel GluCl (30, 31), the human β3 GABA_AR (32), and the mouse 5-HT_3AR (33). More recently, the cryo-EM structure of the α1 GlyR was determined in closed, open, and desensitized conformations (34). Finally, the X-ray crystal structure of the α3 GlyR was determined in a strychnine-bound state (35).

In this study, we take advantage of the ELIC, a prokaryote homolog of vertebrate Cys-loop receptors, which is activated by primary amine molecules, including GABA (36, 37). The availability of several relatively high-resolution X-ray cocrystal structures of ELIC in complex with known ligands renders this channel a relevant model for the study of Cys-loop receptor modulation (36, 38–42).

CPZ, referred to as a noncompetitive antagonist of nAChRs (16), was also found to inhibit GABA-evoked responses in ELIC—although with low potency (IC₅₀, >100 μM)—suggesting a distinct pharmacology (43). Initial studies aimed at identifying the molecular determinants of CPZ binding in nAChRs showed that [³H]CPZ binds to the channel pore (44). Subsequently, [³H]CPZ became a widely used tool to probe the channel pore of nAChRs in closed, open, and desensitized states (16, 45–50). Initial photoaffinity labeling studies on the Torpedo nAChR in the desensitized state revealed that CPZ binds to a high-affinity binding site near the cytoplasmic end of the channel pore, comprising the 2′, 6′, and 9′ positions of the pore-lining M2-helix (46–49). Recently, Chiara et al. (50) extended on these observations and identified an additional binding site in the desensitized state for [³H]CPZ near the extracellular end of the channel pore, comprising the 16′, 17′, and 20′ positions of the M2 segment. In the closed state, [³H]CPZ labeling was observed at 5′, 6′, and 9′, with no labeling at 2′ (50). Additionally, a binding site for CPZ was identified in the intracellular domain, as photoaffinity labeling was also observed for residues αMet-386 and αSer-393, which are localized in the intracellular MA-helices (50). In contrast, in 5-HT₃Rs CPZ acts directly on the neurotransmitter binding site (13, 14) and competitively antagonizes the action of serotonin.

In this study, we set out to investigate the structural determinants of CPZ binding in ELIC. To facilitate structural studies, we used a brominated derivative of CPZ, termed bromopromazine (BrPZ) (Fig. 1B). Here, we report the X-ray crystal structures of ELIC in complex with CPZ or BrPZ at 3.7 Å resolution. We further characterize this interaction using two-electrode voltage-clamp (TEVC) recordings with a thiol-reactive methanethiosulfonate analog of CPZ (MTS-PZ) (Fig. 1C) on ELIC expressed in Xenopus oocytes and also perform molecular-dynamics simulations of the complex. Together, our results expand our current understanding of allosteric modulation in the family of pentameric ligand-gated ion channels.

Results

X-Ray Crystal Structures of ELIC in Complex with CPZ or BrPZ.

In agreement with previous observations (43), we determined that CPZ inhibits ELIC expressed in Xenopus oocytes with an IC₅₀ value of 158 ± 37 μM and a Hill coefficient of 1.6 ± 0.5 (n = 3–13; Fig. 1 D and E). To investigate the structural determinants of CPZ recognition in ELIC, we determined the X-ray cocrystal structures of ELIC in complex with CPZ or BrPZ (crystallographic statistics are reported in Table S1). We obtained diffraction data to a resolution of 3.7 Å and took advantage of the bromine atom in BrPZ to collect anomalous diffraction data, which allowed us to calculate a so-called anomalous difference density map and identify the density of the anomalously scattering electrons around the bromine atoms even at medium resolution. Both the fivefold averaged simple electron difference density maps (F_o-F_c) as well as the fivefold averaged anomalous difference density map allowed us to localize two distinct locations for the binding of CPZ or BrPZ (Fig. 2). Unexpectedly, we do not observe any electron density in the pore domain of the channel. Instead, we observe simple difference density in the extracellular ligand binding domain of ELIC at a site that is located near to the β8–β9 loop (Fig. 2 A–C). In eukaryote receptors, this loop together with the Cys-loop, the M2–M3 loop, and the pre-M1 region form the interface between the ligand binding domain and the pore domain of the channel. At each of the five sites in the pentamer, the simple difference density displays a curved shape (6σ) consistent with the curvature of the tricyclic 10H-phenothiazine ring in CPZ. The electron density for the dimethylpropylamino-moiety is visible only at lower σ levels, indicating it is more disordered in the crystal structure. The binding location of CPZ at this site is further substantiated by the presence of a strong anomalous peak (10σ) at each of the five sites of the pentamer in the ELIC+BrPZ cocrystal structure (Fig. 2 D–F). Importantly, the β8–β9 loop undergoes a conformational rearrangement to accommodate CPZ at this site, which will be discussed in further detail below. Additionally, in the ELIC+BrPZ cocrystal structure, we also observe anomalous difference density at a second location in the extracellular ligand binding domain, namely at the agonist binding site, which is located at the interface between each of two subunits (Fig. 2G). This site is lined by highly conserved aromatic residues localized on historically designated “loops,” termed loops A–B–C on the principal face of the binding site and loops D–E–F on the complementary face of the binding site. The highly conserved aromatic residues at this site form a so-called aromatic box, which creates an electronegative environment for agonist recognition. In ELIC, this site includes Y38 (loop D), F133 (loop B), and Y175 and F188 (both loop C). In the ELIC+BrPZ structure, the anomalous difference density at this site is slightly offset toward Y38 and F133, suggesting a possible location of the bromine atom in BrPZ. Importantly, the binding of BrPZ at this site is consistent with the earlier observation that CPZ binds directly at the neurotransmitter binding site in the related 5-HT₃ receptor (13, 14). Together, the ELIC X-ray crystal structures reveal that CPZ and BrPZ bind at two distinct sites in the extracellular domain, but not in the pore-forming transmembrane domain. The two binding sites are localized at functionally important domains, namely the β8–β9 loop, which contributes to coupling of ligand binding to channel opening, and the agonist binding site, which contains the structural determinants for agonist recognition. The lack of CPZ binding in the pore domain of ELIC can likely be explained by the presence of an unusual and bulky phenylalanine residue at the extracellular end (F16′) of the ELIC channel pore that restricts pore access of known pore blockers, including memantine (40) and likely also CPZ.

Table S1.

Crystallographic table

	ELIC+CPZ	ELIC+BrPZ (merged from multiple crystals)
Crystallographic statistics
Beamline	X06A (SLS)	PROXIMA-I (SOLEIL)
Wavelength, Å	0.9999	0.91983
Space group	P21	P21
a, b, c, Å	105.20, 266.65, 111.05	105.72, 267.91, 111.33
β, °	107.71	106.7
Resolution limits, Å	49.16–3.71 (3.91–3.71)	49.54–3.70 (3.86–3.70)
Rmerge, %	9.3 (49.3)	15.4 (211.9)
Rmeas, %	10.8 (57.3)	16.2 (222.3)
Rpim, %	5.5 (29.1)	3.5 (47.3)
<I/σ>	14.8 (3.3)	11.2 (1.4)
CC1/2, %	99.8 (88.1)	99.9 (64.5)
Multiplicity	3.9 (3.8)	21.4 (21.8)
Completeness, %	98.9 (96.7)	99.9 (100)
Total no. of reflections	236,959 (33,343)	1,351,729 (167,289)
No. unique reflections	60,919 (8,664)	63,054 (7,686)
Anomalous completeness		99.9 (100.0)
Anomalous multiplicity		10.7 (10.8)
Refinement and model statistics
No. of residues in ASU	3,070	3,070
No. of atoms in ASU	25,255	25,070
Rwork, %	22.00	20.68
Rfree, %	25.31	25.56
Rmsd bond distance, Å	0.005	0.004
Rmsd bond angle, °	0.933	0.924
Ramachandran analysis
Outliers, %	0.72	1.0
Favored, %	86.72	88.2
Molprobity score	2.29 (99th percentile)	2.25 (100th percentile)

Fig. 2.

X-ray crystal structures of ELIC in complex with chlorpromazine (CPZ) and bromopromazine (BrPZ). Side view (A) and top view (B) of ELIC in complex with CPZ in blue ribbon representation. The green mesh represents fivefold averaged F_o-F_c difference electron density contoured at a level of 6σ. The Inset (C) shows a detailed view of the β8–β9 loop binding site and its location relative to the Cys-loop, the M2–M3 loop, and the pre-M1 region. CPZ is shown in stick representation. Yellow is carbon, blue is nitrogen, green is chlorine, and orange is sulfur. Side view (D) and top view (E) of ELIC in complex with BrPZ in blue ribbon representation. The red mesh represents fivefold averaged anomalous difference electron density contoured at a level of 10σ. The Insets show a detailed view of the β8–β9 loop binding site (F) and the agonist binding site (G).

Conformational Change of the β8–β9 Loop.

The CPZ-bound ELIC structure superimposes well with the apo form of ELIC [Protein Data Bank (PDB) ID code 2VL0] with a rmsd of 0.9 Å for 2,954 out of 3,070 aligned residues. This suggests that the conformational state of ELIC remains unaltered after CPZ binding and corresponds to a closed nonconductive conformation of the receptor (27). However, detailed inspection of the simple electron density map (2F_o-F_c) reveals structural differences in the β8–β9 loop of the CPZ-bound structure, which was manually rebuilt and refined (Fig. 3A). A detailed view of a monomeric subunit superimposed for apo ELIC (yellow) and ELIC+CPZ (blue) is shown in Fig. 3B. As stated above, the overall structure of ELIC+CPZ is nearly identical to apo ELIC, except for a change in the β8–β9 loop. Most of the conformational change can be observed in the descending part of the β8–β9 loop, including residues Y148 to E155. Residues involved in forming the interface with the M2–M3 loop of the neighboring subunit as well as the ascending part of the β8–β9 loop remain unaltered. Detailed analysis of the interactions between CPZ and residues of the β8–β9 pocket reveals a wide range of mostly hydrophobic interactions (Fig. 3C). These include I20, N21, and I23 on the β1-strand, F126 on the β7-strand, and V147, T149, E150, E155, D158, W160, and I162 on the β8–β9 loop. Weak hydrogen bonds are formed between the dimethylamino-moiety of CPZ and the side-chain oxygen atoms of D158 (indicated with dashed lines in Fig. 3C) and between the 10H-phenothiazine nitrogen and the main-chain carbonyl atom of I23.

Fig. 3.

Conformational change of the β8–β9 loop in ELIC. (A) Stereo representation of the β8–β9 loop in ELIC. The ELIC backbone is shown as green ribbon. Chlorpromazine (CPZ) is shown in yellow sticks. The blue mesh is simple electron density (2F_o-F_c) contoured at a level of 1.4σ. (B) Superposition of a single monomer of apo ELIC in yellow (PDB ID code 2VL0) and ELIC in complex with CPZ in cartoon representation. The Inset shows a detailed view of the β8–β9 loop. CPZ is shown in sphere representation. White is carbon, blue is nitrogen, green is chlorine, and orange is sulfur. (C) Detailed view of amino acids involved in ligand interactions with CPZ. Dashed lines indicate hydrogen bonds. CPZ is shown in stick representation.

Cysteine-Scanning Mutagenesis of the β8–β9 Loop Binding Site.

To explore the contribution of individual amino acids in the β8–β9 loop binding site to molecular recognition of CPZ, we individually mutated each residue involved in the CPZ interaction to a cysteine residue in the background of a Cys-less ELIC variant, which is functionally identical to wild-type ELIC (Table 1). To determine the effect of CPZ binding at this specific site, and not elsewhere in the protein, we used a thiol-reactive analog of CPZ termed MTS-PZ.

Table 1.

Summary of the functional characterization of WT and mutant ELIC: GABA EC₅₀ and n_H and I_max ± SEM

ELIC construct	EC50, mM	nH	Imax, μA	n
Wild-type ELIC	21 ± 1.0	2.1 ± 0.20	21 ± 2.4	4
Cys-less ELIC	23 ± 3.2	2.3 ± 0.70	16 ± 1.9	4
I20C	62 ± 8.0*	2.0 ± 0.40	3.8 ± 0.70	5
N21C	30 ± 5.3	2.1 ± 1.0	1.3 ± 0.10	3
I23C	24 ± 3.6	2.3 ± 0.80	4.9 ± 2.90	2–3
F126C	14 ± 0.5*	2.2 ± 0.10	6.9 ± 3.60	3–5
V147C	41 ± 12.2	1.7 ± 0.90	21 ± 5.8	3
T149C	16 ± 1.1*	2.4 ± 0.30	31 ± 3.1	3
E150C	20 ± 7.8	2.0 ± 1.30	6.2 ± 1.40	3
E155C	20 ± 2.7	2.7 ± 0.70	11 ± 1.9	3–5
D158C	8.5 ± 2.00*	3.0 ± 2.30	6.6 ± 1.30	2–4
W160C	3.5 ± 0.30*	3.1 ± 0.80	0.20 ± 0.07	3–4
I162C	22 ± 5.5	2.3 ± 1.10	6.6 ± 3.20	3

^*P < 0.05, significantly different from Cys-less ELIC, Student’s t test.

First, we investigated the effect of the cysteine mutation alone on the function of ELIC by expressing each mutant in Xenopus oocytes and investigating the response to the agonist GABA using TEVC. All mutants were functional and responded to application of GABA. For each mutant, we determined a GABA concentration–activation curve and calculated EC₅₀ values (Table 1). We observe that EC₅₀ values varied from a threefold decrease in mutant I20C, to a fivefold increase in mutant W160C. EC₅₀ values were statistically compared between Cys-less ELIC and all of the Cys-mutants (Table 1): for mutants I20C, F126C, T149C, D158C, and W160C, EC₅₀ values were significantly different from Cys-less ELIC; the Hill coefficients of the entire set of mutants were not significantly different from Cys-less ELIC. These results point toward a functional role of β8–β9 loop residues to channel gating, which is consistent with the β8–β9 loop’s established contribution to channel gating. In addition, we observe that all mutants express at levels that are comparable to wild-type ELIC except for W160C, which expresses severalfold lower, suggesting that this mutation critically affects protein folding and/or trafficking.

Next, we investigated the functional effect of MTS-PZ binding at each of the individual Cys-mutants. To accomplish this, we used a protocol in which ELIC displayed stable channel activation following two consecutive applications of GABA at the EC₅₀. Next, we perfused the oocyte with 200 µM MTS-PZ for 2 min and washed out unreacted MTS-PZ during a 30-s washout. The change of channel activation after MTS-PZ modification was then measured with a third application of GABA at the EC₅₀. As expected, MTS-PZ did not affect the amplitude of the GABA response in Cys-less ELIC (Fig. 4). We observed that the GABA response was significantly reduced in four mutants, namely, E150C (34.0% ± 10.0, n = 3), D158C (52.0% ± 6.9, n = 3), W160C (55.0% ± 6.2, n = 3), and I162 (41.0% ± 13.0, n = 3). This is consistent with the ligand-binding pose in the ELIC cocrystal structure, which puts the thiol-reactive moiety of MTS-PZ (equivalent to the dimethylamino-moiety in CPZ) in close proximity to these residues. For the seven other mutants, MTS-PZ application did not affect the GABA response. This indicates that either the mutant did not react with MTS-PZ due to the increased distance between the MTS moiety and the sulfhydryl side chain, or that the mutant reacted with MTS-PZ but did not functionally affect the channel. In conclusion, we demonstrate that covalent modification of residues E150C, D158C, W160C, and I162 in Cys-less ELIC with MTS-PZ results in functional inhibition of the GABA-evoked channel response. This result suggests that CPZ binding at the β8–β9 loop binding site is involved in negative allosteric modulation of ELIC.

Fig. 4.

Cysteine-scanning mutagenesis of the β8–β9 loop in ELIC. (A) Electrophysiological recordings of Cys-less ELIC in response to repetitive pulses of GABA at the EC₅₀ (=20 mM) and application of 200 µM of a thiol-reactive CPZ derivative termed MTS-PZ. (B) Example traces of a Cys mutant, E155C, showing no effect of MTS-PZ. (C) Example traces of a Cys mutant, D158C, showing an inhibitory effect of MTS-PZ. (D) Summary of MTS-PZ–mediated channel inhibition on the different Cys mutants. Data represent the mean ± SEM of three to five experiments. *P < 0.05, significantly different from Cys-less ELIC, Student’s t test; **P < 0.01, significantly different from Cys-less ELIC, Student’s t test.

Binding Stability of CPZ.

To characterize the stability of CPZ binding and its effect on surrounding loops, two molecular-dynamics simulations were performed with and without CPZ bound to the structure (labels CPZ and apo, respectively).

Except for one subunit, the binding of CPZ was stable and the molecule remained within the allosteric binding site as measured by the distance from F126 and W160 to the C11 atom of CPZ (Fig. 5 A and B). The cavity volume, on the other hand, increased by 250 ų in CPZ-bound simulations compared with the crystal structure (893 ų) (Fig. 5C), whereas the volume of the same cavity in apo simulations decreased by 50 ų. The increase in cavity volume did not disrupt the weak hydrogen bond interactions of CPZ with the surrounding loops. The CPZ molecules spent on average ∼12% of the simulation time in contact with the β8–β9 loop (residues 148–162) and the β1-strand (residues 22–23) and ∼2% with the Cys-loop (residues 113, 126) (Fig. 5D). The predicted hydrogen bond interactions of CPZ with the β8–β9 loop include two residues, T149 and E155, that were not labeled by MTS-PZ when mutated to cysteine. Although these residues are in proximity of the bound CPZ, the conformations sampled by the side chains were not as favorable for interactions as other residues, that is, E150 and D158 (Fig. 5 E and F). Together, the molecular-dynamics simulations support a ligand-induced conformational change of the β8–β9 loop forming an allosteric binding site for CPZ. The reactivity of the specific side chains identified in the cysteine-scanning mutagenesis experiments is consistent with the rotamers sampled by these residues in simulations.

Fig. 5.

Binding stability of CPZ in simulations. (A and B) Binding of CPZ measured by the distance from F126 and W160 to the C11 atom of CPZ (solid and dashed lines, respectively) per subunit. (C) Allosteric cavity volume in presence and absence of CPZ; the crystal structure value is indicated by the dashed line. (D) Average hydrogen bond interactions of CPZ. (E) Probability distribution of side-chain angles relative to CPZ; measured from Cβ–Cα–NC2 atom positions. (F) MTS-PZ–mediated inhibition (purple, no significant effect; green, inhibition). Outliers (T149, E155) are marked with black arrows.

Because the molecular-dynamic simulations indicate an increase in the allosteric CPZ-binding pocket, which is larger than necessary to adapt CPZ, additional electrophysiological experiments were conducted to exclude the possibility of nonspecific CPZ-binding. IC₅₀ values were determined for 12 additional phenothiazine analogs, and some of these compounds indeed exhibited significantly different IC₅₀ values, indicating that the binding of phenothiazine analogs occurs in a structure-dependent manner (Table S2). Additionally, we observed that the presence of a piperazine group at position R1 gave rise to more potent inhibitors, most likely due to an increase of the interacting interface with the expanded CPZ-binding pocket.

Table S2.

Pharmacological properties of phenothiazine analogs on ELIC

N.S., not soluble after dilution in water of a 10-mM stock solution in DMSO.

Discussion

In the present work, we identify an allosteric binding site in the extracellular domain of the Erwinia pentameric ligand-gated ion channel ELIC using X-ray crystallography. In combination with cysteine-scanning mutagenesis and electrophysiological recordings of ELIC expressed in Xenopus oocytes, we demonstrate that the identified β8–β9 loop site is involved in negative allosteric modulation of ELIC. These results are further supported with molecular-dynamics simulations, which confirm our observations in the crystal structure and the mutagenesis experiments. These results extend on previous observations of allosteric binding sites in different Cys-loop receptors and substantiate our understanding of a multisite model of allosteric modulation in this family of ion channels.

We here discuss our results in the context of previously determined Cys-loop receptor crystal structures in which allosteric binding sites were revealed (Fig. 6 A–D). To enhance clarity in this figure several structures were grouped and classified as “closed,” “open,” and “desensitized,” although we emphasize that subtle and important differences exists in, for example, the GLIC locally closed state (51) and the GluCl apo state (31), which we both classified as closed structures, but most likely represent “intermediate” conformational states (31, 51). For a detailed discussion of the known conformational differences in currently available Cys-loop receptor structures, we refer to recent reviews (52, 53). Notably, several additional sites have been identified using other methods, such as photoaffinity labeling and mutagenesis, but due to space limitation these results extend beyond the scope of the current discussion. Finally, it should be noted that the agonist binding site loop F, which precedes the β8–β9 loop site, adopts significantly different conformations in different Cys-loop receptors, including ELIC and different forms of GLIC.

Fig. 6.

Overview of allosteric binding sites in different conformational states of Cys-loop receptors. Overview of allosteric binding sites in the closed, open, and desensitized states of Cys-loop receptors. (A) Green ribbon representation of α7-AChBP structure as an example of “ligand binding domain-only” structures (54). (B) Green ribbon presentation of the ELIC ion channel as a representative example of a closed state (27). (C) Green ribbon presentation of the GluCl+ivermectin ion channel structure as a representative example of an open state (30). (D) Green ribbon presentation of the β3 GABA_A receptor structure as a representative example of a desensitized state (32). Allosteric modulators identified in the different conformational states are shown in sphere representation. Identical color codes have been used for overlapping sites in the different states, for example, orthosteric site in yellow, vestibule site in firebrick, etc. Detailed explanation of PDB ID codes, allosteric modulator color codes, and references for all structures used in this figure are given in Table S3.

Table S3.

Overview of allosteric binding sites in Cys-loop receptors

Protein + ligand	PDB ID code	Color ligand	Ref.
A, α7-AChBP
α7-AChBP + lobeline	5AFH	Yellow	Ref. 54
α7-AChBP (+ lobeline) + fragment 1	5AFJ	White and light pink	Ref. 54
α7-AChBP (+ lobeline) + fragment 4	5AFM	Firebrick	Ref. 54
α7-AChBP + epibatidine	3SQ6	Not shown	Ref. 55
α7-AChBP + α-bungarotoxin	4HQP	Not shown	Ref. 56
B (“closed state”)
ELIC + bromo-flurazepam	4A98	Yellow	Ref. 36
ELIC (+ GABA) + flurazepam	2YOE	Firebrick	Ref. 36
α3 GlyR + strychnine	5CFB	Not shown	Ref. 35
α1 GlyR + strychnine	3JAD	Not shown	Ref. 34
ELIC + Ba2+	2YN6	Red	Ref. 39
ELIC + bromoform	3ZKR	Magenta	Ref. 41
GLIC locally closed + xenon	4ZZB	White	Ref. 61
ELIC + CPZ	5LG3	Cyan	This paper
ELIC F16′S + memantine	4TWD	Orange	Ref. 40
ELIC + isoflurane	4Z90	Marine	Ref. 42
GLIC “closed” neutral pH	4NPQ	Not shown	Ref. 60
GluCl apo	4TNV	Not shown	Ref. 31
C (“open state”)
GluCl (+ ivermectin) + glutamate	3RIF	Yellow	Ref. 30
GLIC “open” acidic pH	3EHZ, 3EAM	Not shown	Refs. 28 and 29
α1 GlyR + glycine	3JAE	Not shown	Ref. 34
GLIC + (bromo)acetate	4QH1	Not shown	Ref. 58
GLIC + ketamine	4F8H	Light pink	Ref. 57
GLIC A13′F + Cs+	4ILA	Firebrick	Ref. 59
GLIC + Ni2+	4NPP	Not shown	Ref. 60
GluCl + ivermectin (+ glutamate)	3RIF	Violet	Ref. 30
GLIC F14′A + ethanol	4HFE	Light orange	Ref. 63
GLIC + propofol	3P50	Blue	Ref. 62
GLIC + desflurane	3P4W	Not shown	Ref. 62
GLIC F14′A + bromoform	4HFD	Not shown	Ref. 63
GLIC + bromo-lidocaine	2XQ3	Dark red	Ref. 63
GLIC + tetraethylarsonium	2XQ5	Light blue	Ref. 63
GLIC + Cs+	2XQ6	Red	Ref. 63
GLIC + Zn2+	2XQ8	Wheat	Ref. 63
GLIC + Cd2+	2XQ7	Not shown	Ref. 63
GluCl (+ ivermectin) + picrotoxinin	3RI5	Not shown	Ref. 30
D (“desensitized state”)
β3 GABAAR + benzamidine	4COF	Yellow	Ref. 32
GLIC (+ bromopropylamine) + isoflurane	4Z91	Marine	Ref. 42
GluCl + POPC	4TNW	Not shown	Ref. 31
α1 GlyR + ivermectin	3JAF	Violet	Ref. 34

First, we discuss allosteric binding sites unveiled in a chimera of the Lymnaea AChBP and the ligand binding domain of the α7 nAChR, α7-AChBP (green ribbon, Fig. 6A). Using a fragment-based screening approach, Spurny et al. (54) discovered three allosteric binding sites in α7-AChBP, which are remote from the orthosteric binding site occupied by the agonist lobeline in these structures (yellow spheres, Fig. 6A), and also the agonist epibatidine (55) or the competitive antagonist α-bungarotoxin in other α7-AChBP cocrystal structures (56). One fragment molecule, fragment 1, was identified that binds at the interface between the N-terminal α-helix and a loop that corresponds to the main immunogenic region (MIR) in the α1 muscle nAChR (white spheres, Fig. 6A). This site was termed the “top site” and is involved in negative modulation of the α7 nAChR (54). The same fragment molecule also occupies an allosteric binding site that is located just below the orthosteric binding site and that was termed the “agonist subsite” (pink spheres, Fig. 6A) (54). This site corresponds to the ketamine binding site reported in the GLIC (57), where ketamine also binds just below the orthosteric agonist binding site and is involved in inhibition of GLIC (pink spheres, Fig. 6C). Another fragment molecule, fragment 4, was identified that occupies an allosteric binding site accessible from the vestibule of the receptor and was termed the “vestibule site” (firebrick spheres, Fig. 6A) (54). This fragment is involved in negative modulation of the α7 nAChR (54). The same site was unveiled in the crystal structure of the ELIC in complex with flurazepam (firebrick spheres, closed state, Fig. 6B), which is involved in positive modulation of ELIC (36). The importance of the same site was also confirmed in the crystal structure of GLIC in complex with acetate (firebrick spheres, open state, Fig. 6C) (58).

Second, we discuss allosteric binding sites unveiled in the closed state of Cys-loop receptors, and we show the ELIC structure as a representative example because most of the available modulator cocrystal structures have been determined with this ion channel (green ribbon, Fig. 6B) (27). Similar to other Cys-loop receptors, the orthosteric binding site in ELIC is occupied by the partial agonist GABA (36) or the competitive antagonist bromo-flurazepam (yellow spheres, Fig. 6B) (36). The orthosteric site is also occupied by the competitive antagonist strychnine in the closed α3 GlyR (35) and closed α1 GlyR structures (34). The occupancy of the vestibule site by flurazepam in ELIC was already mentioned in the previous paragraph (firebrick spheres, Fig. 6B) (36). It was reported that ELIC is negatively modulated by divalent cations, including Ca²⁺ and Ba²⁺, and it was demonstrated that these cations bind at three distinct binding sites (39) (red spheres, Fig. 6B). The first Ba²⁺ site is located at the outer rim of the “vestibule pocket” where it is coordinated by two residues at the end of the β4-strand, namely S84 on the principal subunit and D86 on the complementary subunit (39) (red spheres, Fig. 6B). This site overlaps with a Cs⁺ binding site in the GLIC A13′F mutant (59) and a Ni²⁺ binding site in wild-type GLIC (60) (red spheres, open state, Fig. 6C). A second Ba²⁺ site is located at the subunit interface about 15 Å below the orthosteric agonist site (red spheres, Fig. 6B). This site is formed by residues at the end of the β6-strand on the principal subunit and the loop connecting the β8- and β9-strand on the complementary subunit (39). This site is just 8 Å distant from the CPZ binding site (β8–β9 loop) reported in this paper (cyan spheres, Fig. 6B). The third Ba²⁺ site is located at the extracellular entrance of the channel pore where it binds at the 20′ position of the M2-helix (39) (red sphere, Fig. 6D). Another allosteric binding site near to the CPZ binding site is occupied by either bromoform in ELIC (magenta sphere, Fig. 6B) (41) or xenon in locally closed GLIC (gray sphere, Fig. 6B) (61). Additional binding sites for xenon in locally closed GLIC are located at the intrasubunit general anesthetic binding site (blue sphere, Fig. 6B), the inner-interfacial sites (gray spheres, Fig. 6B), and the outer-interfacial sites (gray spheres, Fig. 6B) (61). Finally, xenon also occupies the 9′ pore site in locally closed GLIC (gray sphere, Fig. 6B) (61). Two additional binding sites for bromoform have been localized in the ELIC cocrystal structure, namely at an intersubunit transmembrane site (magenta spheres, Fig. 6B) and at the 13′ pore site (magenta sphere, Fig. 6B) (41). The 13′ pore site for bromoform overlaps with one of the pore sites (13′) of the anesthetic isoflurane site in ELIC (light blue spheres, Fig. 6B) (42). Isoflurane simultaneously occupies the 6′ pore site in ELIC. The pore blocker memantine has been identified at the 16′ pore position in ELIC F16′S (orange spheres, Fig. 6B) (40).

Third, we discuss allosteric binding sites identified in the open state of Cys-loop receptors, and we show the GluCl+ivermectin crystal structure as a representative example (green ribbon, Fig. 6C). In GluCl, the orthosteric binding site is occupied by glutamate (yellow spheres, Fig. 6C) (30), and in open GLIC, this site is occupied by acetate (58). Allosteric binding sites for xenon have been determined for open GLIC, which overlap with those described in locally closed GLIC (gray spheres, Fig. 6B), except for the pore site, and these were omitted in Fig. 6C for clarity. The allosteric binding sites in the extracellular domain for Cs⁺ (red spheres, Fig. 6C) (59), ketamine (pink spheres, Fig. 6C) (57), and acetate (firebrick spheres, Fig. 6C) (58) were already mentioned in the previous paragraphs. In the transmembrane domain of open GLIC, an intrasubunit binding site has been identified for general anesthetics, including propofol (blue spheres, Fig. 6C) (62), desflurane (62), and bromoform (63), and is localized at the upper half of the interface between the M1- and M3-transmembrane helix. In an engineered F14′A mutant of GLIC, the ethanol binding site was identified (light orange spheres, Fig. 6C) (63), which localizes at the upper half of the transmembrane domain at the interface between two neighboring M2-subunits (63). The ethanol binding site partially overlaps with the ivermectin binding site in open GluCl (violet spheres, Fig. 6C), but in the latter ivermectin wedges in between the M1-helix of one subunit and the M3-helix of a neighboring subunit (30). The frontal ivermectin molecule in Fig. 6C was omitted for clarity because it obscures view on the pore. The ivermectin binding site was also found in the structure of α1 GlyR in complex with ivermectin and glycine (34). Different pore blocker sites have been identified in open GLIC, including at the 13′ position for bromo-lidocaine (brown sphere, Fig. 6C) (64), at the 6′ position for tetraethylarsonium (light blue sphere, Fig. 6C) (64), at the 2′ position for Cs⁺ (red sphere, Fig. 6C) (59, 64), and at −2′ position for Zn²⁺ and Cd²⁺ (wheat sphere, Fig. 6C) (64). The −2′ pore site, which forms the ion selectivity filter, also overlaps with the picrotoxinin binding site in open GluCl (30).

Fourth, we discuss allosteric binding sites identified in the desensitized state of Cys-loop receptors and the β3 GABA_AR structure is shown as a representative example (green ribbon, Fig. 6D) (32). In this structure, the orthosteric binding site is occupied by the β3-agonist benzamidine (yellow spheres, Fig. 6D). This site is also occupied by the agonist 3-bromopropylamine (3-BrPPA) in the ELIC complex with 3-BrPPA and isoflurane (42). Isoflurane occupies the 13′ and 6′ pore sites (light blue spheres, Fig. 6D), which are identical to those for the closed state (Fig. 6B). Finally, ivermectin in desensitized α1 GlyR occupies the same site as in the open GluCl structure (violet spheres, Fig. 6D) (34). The ivermectin site overlaps with the binding site for the lipid 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) in an intermediate GluCl structure (31).

An important question concerns the relevance of the ELIC+CPZ cocrystal structure reported in this paper in relation to other prokaryote and eukaryote Cys-loop receptors. First, the existence of the CPZ binding site near the β8–β9 loop was predicted in the crystal structure of GLIC at neutral pH, which represents a “closed/resting state” (60). It was recently confirmed as a possible xenon binding site in GLIC (60). Importantly, a highly conserved aromatic residue of the β8–β9 loop, namely, Trp in cationic receptors (W160 in ELIC) or Phe/Tyr in anionic receptors, forms part of the conserved GEW sequence motif, which was previously demonstrated to be implicated in the positive allosteric modulation of the neuronal α7 nAChR by regulatory Ca²⁺ ions (65). This β8–β9 loop site is distinct from the widely known pore blocker site of CPZ, which has been extensively studied in the Torpedo nAChR using electrophysiological, mutagenesis, and photoaffinity labeling studies (16, 45–50). In the present study, we could not observe CPZ binding in the ELIC pore, and this can be likely explained by the unusual and bulky Phe residue at the 16′ pore position in ELIC, which prevents pore access to noncompetitive pore blockers such as memantine (40) and probably also CPZ. Therefore, it is possible that the β8–β9 loop site identified in our study on ELIC corresponds to an “external” binding site for CPZ described in one of the pioneering studies on mouse C2 muscle-type nAChRs and that is distinct from the high-affinity “internal” pore blocker site (66).

Collectively, the results from these structural studies offer a landscape view of different allosteric binding sites in Cys-loop receptors with different sites localized at the extracellular ligand binding domain, the pore domain and the transmembrane domain. The different allosteric sites form an almost continuous path stretching from one extreme end at the top of the N-terminal α-helix to the bottom of the intracellular entrance of the channel pore. With the structure determination of ELIC in complex with CPZ, an important and missing gap is filled, namely, at a site that forms the interface between the ligand binding domain and the pore-forming transmembrane domain. The β8–β9 loop site is structurally and functionally important as it affects coupling between ligand binding and channel opening.

Methods

ELIC was expressed as a N-terminal fusion with maltose-binding protein (MBP) in C43 Escherichia coli cells. The fusion protein was purified on amylose resin (New England Biolabs), and ELIC was cleaved off with C3V protease. Concentrated protein (10 mg/mL) was supplemented with E. coli lipids and cocrystallized with 1–10 mM CPZ or BrPZ using the vapor diffusion crystallization technique. The X-ray cocrystal structures of ELIC were solved using molecular replacement. Cysteine-scanning mutagenesis and current recordings were carried out on ELIC mutants expressed in Xenopus oocytes using the TEVC technique. Details on protein purification, X-ray crystallography, electrophysiological recordings, and molecular-dynamics simulations are reported in SI Methods.

SI Methods

Chlorpromazine and Analogs.

Chlorpromazine hydrochloride was purchased from Sigma-Aldrich. Bromopromazine hydrochloride was custom synthesized and purchased from VillaPharma Research. MTS-promazine or N-(3-methanethiosulfonylpropyl)phenothiazine was custom synthesized and purchased from Toronto Research Chemicals. All other phenothiazine analogs were purchased from Santa Cruz Biotechnology or Sigma-Aldrich.

Protein Expression and Crystallization.

ELIC was expressed and purified as previously described (36), with minor modifications. ELIC was cloned into pET-11a (Novagen) and N-terminally fused to maltose-binding protein (MBP) with a 3CV protease cleavage site for removal of MBP. The MBP–ELIC fusion protein was expressed in the C43 Escherichia coli strain. Membranes were solubilized with 2% (wt/vol) anagrade n-undecyl-β-d-maltoside (Anatrace), and the soluble fraction was purified by affinity chromatography on amylose resin (New England Biolabs). Affinity-bound protein was cleaved by 3CV protease and further purified on a Superdex 200 10/300 GL (GE Healthcare) column equilibrated with buffer containing 10 mM Na-phosphate (pH 8.0), 150 mM NaCl, and 0.15% n-undecyl-β-d-maltoside. Concentrated protein (10 mg/mL) was supplemented with 0.5 mg/mL E. coli lipids (Avanti Polar Lipids). Crystallization of ELIC was carried out at 4 °C by vapor diffusion of sitting drops. The crystallization buffer was composed of 200 mM ammonium sulfate, 50 mM N-(2-acetamido)iminodiacetic acid (ADA) (pH 6.5), and 9–12% (vol/vol) PEG4000. Cocrystallization trials of ELIC in complex with chlorpromazine (CPZ) or bromopromazine (BrPZ) were set up at final ligand concentrations of 1–10 mM in the crystallization drop. Before crystal harvesting, 30% glycerol was added to the mother liquor as a cryoprotectant. Crystals were cryocooled in liquid nitrogen.

X-Ray Crystal Structure Determination.

ELIC in complex with CPZ or BrPZ was crystallized using the sitting-drop vapor diffusion method in the presence of 1–10 mM ligand. For ELIC in complex with CPZ, a diffraction dataset to a resolution of 3.7 Å was collected (crystallographic statistics are shown in Table S1). Data were indexed and integrated in XDS (67), and scaling was done in Scala in the CCP4 suite (68). The structure was determined using molecular replacement with Phaser and a previously published ELIC structure from which ligands were removed (PDB ID code 4A97). Model building and refinement were carried out by iterative cycles of manual rebuilding in Coot (69) and refinement in Phenix.refine (70), using one translation libration screw body per subunit and fivefold noncrystallographic symmetry (NCS) restraints. The simple difference density map revealed clear features of the CPZ molecules to allow building of ligand in the ELIC+CPZ structure. The data collection strategy for crystals of ELIC in complex with BrPZ was optimized to collect highly redundant X-ray diffraction data. The clearest anomalous peaks could be observed when multiple datasets from three different crystals were merged to calculate anomalous maps (see statistics for the merged dataset in Table S1). Merging of 10 complete datasets, resulting in a multiplicity of more than 20, was required to distinguish clear peaks in the anomalous difference map after fivefold NCS averaging. The anomalous difference density map shown in Fig. 2 was calculated in Phenix.refine using reflections between 25 and 4.5 Å, after fivefold NCS averaging in Coot. The peaks in the anomalous difference map were used to guide building of bromine atoms in the ELIC+BrPZ structure. The simple difference density was insufficiently clear to build the rest of the ligand. Model validation was done using Molprobity (71), and all figures were prepared using PyMOL.

Mutagenesis and TEVC Recordings.

To further characterize the binding site of CPZ on ELIC, the substituted-cysteine accessibility method (72) was applied. Based on the X-ray crystal structure of ELIC in complex with CPZ, 11 amino acids were identified as putative binding sites for CPZ and individually mutated to cysteine: I20, N21, I23, F126, V147, T149, E150, E155, D158, W160, and I162. For expression in Xenopus laevis oocytes, ELIC cDNA was cloned into pGEM-HE. Before any residue substitution, all endogenous cysteines (positions 299 and 312 in TM4) were replaced by a serine residue (Cys-less ELIC, C299S+C312S). All further mutations were generated in the background of Cys-less ELIC through a QuikChange method (Stratagene) and verified by sequencing (LGC Genomics). cRNA was transcribed in vitro using the mMESSAGE mMACHINE T7 transcription kit (Ambion).

Stage V–VI Xenopus laevis oocytes were harvested and stored in ND-96 solution containing 2 mM KCl, 96 mM NaCl, 1 mM MgCl₂, 1.8 mM CaCl₂, 5 mM Hepes (pH 7.4), supplemented with 50 mg/L gentamicin sulfate. Oocytes were injected with 27 nL of cRNA at a concentration of ∼1 µg/µL using a microinjector (Drummond). Whole-cell currents were recorded 1–2 d after injection using the TEVC technique. All electrophysiological experiments were conducted at a constant temperature (18 °C) using an Axoclamp 900A amplifier (Molecular Devices) controlled by a pClamp 10.2 data acquisition system (Molecular Devices). Data were sampled at 100 Hz and low-pass filtered at 10 Hz using a four-pole Bessel filter. Voltage and current electrodes were backfilled with 3 M KCl, and their resistances were kept as low as possible (<1 MΩ). Oocytes were clamped at −60 mV throughout all recordings. The bath perfusion solution (ND-96) contained the following (in mM): 96 NaCl, 2 KCl, 1.8 CaCl₂, 1 MgCl₂, and 5 Hepes (pH 7.4).

Modification of Engineered Cys Residues in ELIC with MTS-PZ.

MTS-PZ was prepared as a 1 M stock solution in 100% DMSO and stored at −20 °C. Upon use, one aliquot of MTS-PZ was thawed and diluted into the working solution (ND-96), which was kept on ice. The 200 µM MTS-PZ was applied to Cys-less ELIC and to each of the 11 mutants, and its effect on the GABA-induced current at the half-maximal concentration (EC₅₀) was investigated. After a stable GABA response was observed for two or more consecutive pulses (≤10% difference in current amplitude), MTS-PZ was applied for 2 min, which was followed by a 30-s washout. GABA response was determined again and the effect of MTS-PZ was calculated as follows:

$$\%\hspace{0.17em}\text{effect}=\left(\frac{I_{\text{after}}}{I_{\text{initial}}}-1\right)\times 100.$$

The mean and SEM for a series of oocytes were plotted against agonist or antagonist concentration and iteratively fitted to the following equation:

$$I_{\text{A}}=I_{\text{min}}+\frac{I_{\text{max}}-I_{\text{min}}}{1+{10}^{n_{\text{H}}\hspace{0.17em}(\log {\text{A}}_{50}-\log \text{A})}},$$

where A is the ligand concentration; I_A is the current in the presence of ligand concentration A; I_min is the current when A = 0; I_max is the current when A = ∞; A₅₀ is the concentration of ligand that evokes a current equal to (I_max + I_min)/2; and n_H is the Hill coefficient.

Data analysis was performed using Clampfit 10.2 (Molecular Devices), Excel 2007 (Microsoft), Origin 7.0 (OriginLab), and Prism (GraphPad). Statistical differences were assessed using Student’s t test.

Molecular-Dynamics Simulations of ELIC.

The first biological assembly of the structure cocrystallized with CPZ was used to construct the simulations. One missing residue in N termini was fixed by using the chain with lowest rmsd as a template in Modeler 9.14 (73). CPZ bound and apo simulations were prepared and equilibrated independently.

The protein was embedded in a POPC bilayer containing 364 lipids by aligning the center of mass of the transmembrane domain to the center of the bilayer, using the Gromacs g_membed tool (74). After removal of overlapping lipids, both system contained 308 lipids. Systems were solvated with ∼40,000 TIP3P water molecules in a rectangular box 105 Å wide and 160 Å high. In total, 102 Na⁺ and 72 Cl⁻ ions were added to neutralize the system at a concentration of 100 mM.

Simulations were performed with Gromacs 5.0.6 (75), using the Amber99sb-ILDN force field (76) for the protein and Berger (77) for POPC lipids. Parameters for CPZ were generated with the STaGE automatic topology generator using the GAFF force field (78). All bonds were constrained with the LINCS algorithm. Systems were minimized for 10,000 steps, followed by relaxation with pressure coupling; three consecutive 15-ns simulations with 2.5-fs time step were performed under 1,000 kJ⋅mol⁻¹⋅nm⁻² position restraints force applied to heavy atoms, backbone atoms, and Cα atoms, respectively. The CPZ ligands were restrained during the relaxation. Pressure was adjusted using a semiisotropic Berendsen barostat to a pressure of 1 bar. Temperature was coupled separate for protein-and-ligands, lipids, and water-and-ions to 310 K with a Bussi velocity-rescaling thermostat (79). Particle mesh Ewald electrostatics was used with a 10-Å cutoff and 1.25-Å fast Fourier transform grid spacing. After relaxation, all restraints were removed, and production runs extended to 500 ns each. Pocket calculations were performed with mdpocket module of Fpocket (80). The cavity definition was extracted from CPZ-bound simulations as previously described (81), and the boundaries of the cavity were set to accommodate both CPZ binding and the cavity entrance (between loop F and β1). Although this definition might lead to larger volumes than CPZ, the relative changes in volume were considered rather than the absolute values. Images were rendered in VMD 1.9.2 (82) using Tachyon rendering (83). Plots were generated using GraphPad Prism 6.0h for Mac OSX (GraphPad Software; www.graphpad.com).