Molecular Determinants of Ivermectin Sensitivity at the Glycine Receptor Chloride Channel*

Background: The ivermectin-binding site on the glutamate-gated chloride channel was recently resolved by crystallography. Results: Ivermectin binds in a similar orientation to the structurally related glycine receptor, although two H-bonds apparent in the crystal structure proved unimportant for binding to glycine receptors. Conclusion: Ivermectin-binding mechanisms vary among Cys-loop receptors. Significance: Understanding ivermectin-binding mechanisms may help in designing new drugs. Ivermectin is an anthelmintic drug that works by activating glutamate-gated chloride channel receptors (GluClRs) in nematode parasites. GluClRs belong to the Cys-loop receptor family that also includes glycine receptor (GlyR) chloride channels. GluClRs and A288G mutant GlyRs are both activated by low nanomolar ivermectin concentrations. The crystal structure of the Caenorhabditis elegans α GluClR complexed with ivermectin has recently been published. Here, we probed ivermectin sensitivity determinants on the α1 GlyR using site-directed mutagenesis and electrophysiology. Based on a mutagenesis screen of transmembrane residues, we identified Ala288 and Pro230 as crucial sensitivity determinants. A comparison of the actions of selamectin and ivermectin suggested the benzofuran C05-OH was required for high efficacy. When taken together with docking simulations, these results supported a GlyR ivermectin binding orientation similar to that seen in the GluClR crystal structure. However, whereas the crystal structure shows that ivermectin interacts with the α GluClR via H-bonds with Leu218, Ser260, and Thr285 (α GluClR numbering), our data indicate that H-bonds with residues homologous to Ser260 and Thr285 are not important for high ivermectin sensitivity or direct agonist efficacy in A288G α1 GlyRs or three other GluClRs. Our data also suggest that van der Waals interactions between the ivermectin disaccharide and GlyR M2–M3 loop residues are unimportant for high ivermectin sensitivity. Thus, although our results corroborate the ivermectin binding orientation as revealed by the crystal structure, they demonstrate that some of the binding interactions revealed by this structure do not pertain to other highly ivermectin-sensitive Cys-loop receptors.

Ivermectin is a semi-synthetic anthelmintic drug used widely in human medicine and veterinary practice (1). The biological target for ivermectin and related macrocyclic lactones is a glutamate-gated chloride channel receptor (GluClR) 5 that is expressed in the neurons and muscle cells of nematodes and some arthropods but is absent in vertebrates (2). Ivermectin irreversibly activates these GluClRs at low nanomolar concentrations, thereby inhibiting neuronal activity and muscle contractility and thus inducing death by flaccid paralysis (3). Unfortunately, ivermectin resistance is emerging as a serious problem in nematodes and arthropods (4 -6). In several instances, resistance has been shown to be caused by mutations that reduce the GluClR ivermectin sensitivity (7)(8)(9). Insight into the binding mechanisms of ivermectin at the GluClR may contribute to the understanding of ivermectin resistance mechanisms and to the development of a much needed new generation of anthelmintic drugs. A 3.3-Å crystal structure of the Caenorhabditis elegans ␣ GluClR with ivermectin bound has recently been published (10), revealing ivermectin's molecular interactions at atomic resolution.
GluClRs belong to the Cys-loop receptor superfamily that also includes the excitatory nicotinic acetylcholine (nAChR) and 5-hydroxytryptamine type 3 receptors, the inhibitory ␥-aminobutyric acid type A receptor (GABA A R), and the inhibitory GlyR. Cys-loop receptors are formed by five homologous subunits that each consist of an N-terminal ligand-binding domain (LBD) and a bundle of four transmembrane helices (M1-M4) that constitute the transmembrane domain (TMD). M2 helices contributed from each subunit line the central ion channel pore. Neurotransmitter ligand-binding sites lie at the interface of LBDs of adjacent subunits.
Ivermectin also interacts with many vertebrate Cys-loop receptors but usually only at high (micromolar) concentrations. For example, GABA A Rs and GlyRs are directly activated by ivermectin at 1-2 M (11,12), and acetylcholine-induced cur-rents at ␣7 nAChRs are potentiated by a pre-application of 30 M ivermectin (13,14). Insight into the binding mechanisms of ivermectin at human Cys-loop receptors may contribute to the characterization of novel therapeutic pharmacophores. For this reason, we sought to identify the molecular basis of ivermectin binding to the ␣1 GlyR.
The ␣ GluClR-ivermectin crystal structure (10) was published after the experiments described in Figs. 1-7A were completed. With prior knowledge of the crystal structure, our experimental design would have been different. However, because all of our data remain relevant, we describe our original experiments in the context of our original experimental design. Following this, we generate a structural model of the ␣1 GlyR ivermectin-binding site on the basis of our data, compare it with the crystal structure binding site, and then experimentally verify whether it can account for ivermectin binding to the ␣1 GlyR.

EXPERIMENTAL PROCEDURES
Molecular Biology-The human ␣1 GlyR subunit and the Hemonchus contortus GluClR ␣3〉 subunit cDNAs were subcloned into the pCIS and pcDNA3.1 plasmid vectors, respectively. Site-directed mutagenesis was performed using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA), and the successful incorporation of mutations was confirmed by DNA sequencing.
HEK-293 Cell Culture and Transfection-HEK-293 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) containing Serum Supreme (Lonza, Walkersville, MD) and penicillin/streptomycin (Sigma) and split onto glass coverslips in 35-mm culture dishes. The following day, cells were transiently transfected via a calcium phosphate method with the GlyR or GluClR cDNAs together with empty pEGFP plasmid vector (Clontech) as a fluorescent transfection marker. Typically, 250 ng of each plasmid was used to transfect each 3-cm dish. After 12-18 h in the transfection medium, cells were washed twice with calcium-free phosphate-buffered saline and returned to DMEM. Cells were used in experiments 24 -48 h later.
Electrophysiology and Data Analysis-An inverted fluorescence microscope was used to visualize cells for electrophysiological experiments. Cells expressing recombinant GluClRs or GlyRs were identified by their green fluorescence. Borosilicate glass capillary tubes (Vitrex, Modulohm, Denmark) and a horizontal pipette puller (P97, Sutter Instruments, Novato, CA) were used to pull patch clamp pipettes with tip resistances of 2-3 megohms when filled with pipette solution consisting of (in mM) 145 CsCl, 2 CaCl 2 , 2 MgCl 2 , 10 HEPES, and 10 EGTA, adjusted to pH 7.4 with 2 M NaOH. Drug solutions were prepared from the control solution consisting of (in mM) 140 NaCl, 5 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 HEPES, and 10 D-glucose, adjusted to pH 7.4 with 2 M NaOH. Cells were voltage-clamped at Ϫ40 mV in the whole-cell recording configuration, and membrane currents were recorded using an Axon Multiclamp 700B amplifier and pClamp 10 software (Molecular Devices, Sunnyvale, CA). Membrane currents were filtered at 500 Hz and digitized at 2 kHz. Stocks of glycine and L-glutamate, dissolved in water to 1 M and adjusted to pH 7.4 with NaOH, were maintained at 4°C. Ivermectin (Sigma) and selamectin (a gift from Pfizer Inc., Milwaukee, WI) were dissolved in dimethyl sulfoxide and stored as 10 mM stocks at Ϫ20°C. Thus, solutions containing 30 M ivermectin (the highest concentration routinely used) also contained 0.3% dimethyl sulfoxide. This concentration of dimethyl sulfoxide showed no effects on the membrane resistance of cells. Solutions were applied to cells via gravity-induced perfusion systems fabricated from polyethylene tubing. All experiments were performed at room temperature (19 -22°C). Agonist dose-response experiments were performed as described below. For each dose-response experiment, the halfmaximal agonist concentration (EC 50 ), Hill coefficient (n H ), and saturating current magnitude (I max ) values were determined by fitting individual dose-response relationships with the three-parameter Hill equation (SigmaPlot 11, Jandel Scientific, San Rafael, CA). Results are expressed as mean Ϯ S.E. from at least three experiments. Unpaired t tests (SigmaPlot 11, Jandel Scientific, San Rafael, CA) were used to compare these values, as described in the table legends.
GlyR Structural Modeling and Computational Docking-A homology model of the ␣1 GlyR pentamer was built based on a hybrid template, with the TMD and Cys-loop based on the bacterial ELIC channel structure (Protein Data Bank code 2VL0) (15) and the remainder of the LBD based on acetylcholinebinding protein (Protein Data Bank code 1I9B) (16), as described previously (17). Two distinctly different families of ivermectin conformers were predicted by MarvinSketch software (Chemaxon, Budapest, Hungary). AutoDock Vina (18) was used to explore feasible interactions of each of these two conformer families with WT and A288G mutant model GlyRs, within a 40 ϫ 40 ϫ 30-Å box surrounding the Ala 288 residue. Conformer 1 gave consistently stronger binding energies than conformer 2 in all docking experiments, so only the results for conformer 1 are presented here.
Following the publication of the ␣ GluClR-ivermectin crystal structure (10), in which the bound ivermectin was in the conformer 1 family, we built a new homology model of the ␣1 GlyR pentamer based on this structure (Protein Data Bank code 3RIF). Ivermectin in conformer 1 was then docked to this model, as described above. As a control, equivalent docking was carried out on 3RIF from which the bound ivermectin had been removed prior to docking.

Disruption of Ivermectin Efficacy via Mutations at the LBD-TMD
Interface-Because the conservative Y279F mutation in the ␣1 GlyR M2-M3 loop dramatically reduces ivermectin sensitivity (19), we investigated whether residues at the LBD-TMD interfacial region may contribute to ivermectin-binding or -gating mechanisms. We thus introduced cysteines one at a time for each residue from R271C (19Ј) to I283C. Each mutant GlyR was investigated by quantitating the steady-state current magnitude activated by 0.3, 3, and 30 M ivermectin as a fraction of the saturating glycine-activated current in the same cell (Fig. 1A). The wild type (WT) ␣1 GlyR was included as a positive control. Mean glycine EC 50 values and peak current magnitudes of all mutant receptors have previously been reported (20). As shown in Fig. 1A, most mutant GlyRs showed little or no activation at 0.3 M ivermectin but strong activation at 30 M ivermectin. Apart from Y279F, P275C was the only mutation that significantly reduced the magnitude of the current activated by 3 M ivermectin relative to the corresponding value observed at the WT GlyR (p Ͻ 0.05). The locations of Pro 275 and Tyr 279 in an ␣1 GlyR structural homology model are shown in Fig. 1B. We next quantitated the ivermectin EC 50 values for the P275C and Y279F mutant GlyRs. As ivermectin is a poorly reversible agonist and does not produce appreciable desensitization (12,21), dose responses were quantitated by applying progressively increasing ivermectin concentrations. Examples of the responses of WT and Y279F mutant GlyRs to saturating concentrations of glycine and progressively increasing concentrations of ivermectin are shown in Fig. 1C. Aver-aged ivermectin dose-response relationships for both receptors are shown in Fig. 1E with mean parameters of best fit summarized in Table 1. Both mutations dramatically reduced receptor sensitivity to both ivermectin and glycine (Tables 1 and 2). Unlike Y279F, the Y279C mutation had little effect on ivermectin sensitivity (Table 1) but produced a large rightward shift in the glycine EC 50 (19,20).
We next investigated possible ivermectin sensitivity determinants in the LBD regions (i.e. loop 2, conserved Cys-loop pre-M1 domain) that lie proximal to Pro 275 and Tyr 279 in our model ␣1 GlyR structure. In loop 2, we investigated A52C and E53C not only because of their proximity to Tyr 279 but also because of the known role of A52 as an alcohol sensitivity determinant (22,23). We also investigated D141G, L142C, and   Table 1. The L142C and Y222C mutations dramatically reduced receptor sensitivity to both ivermectin and glycine (Tables 1 and 2 and Fig. 1E). Given this nonselective action on glycine-and ivermectin-mediated currents, and the locations of these mutations in known agonist transduction pathways (24), we hypothesized that these mutations disrupted the ivermectin gating efficacy. To test this, we quantitated the potency with which ivermectin potentiated glycine-activated currents. This was performed by alternating low (EC 5 -EC 10 ) glycine applications with 5-s applications of increasing ivermectin concentrations (Fig. 1D). Averaged ivermectin potentiating dose-response relationships for the WT, L142C, Y222C, P275C, and Y279F GlyRs reveal that the potentiating effects of ivermectin occurred at lower concentrations at these mutant GlyRs than at the WT GlyR ( Fig. 1F and Table 3). Together, these results indicate that the L142C, Y222C, P275C, and Y279F substitutions disrupt ivermectin efficacy while having little effect on its affinity.
We then tested the effects of these mutations on the highly ivermectin-sensitive A288G mutant GlyR (21). As summarized in Table 4, the A288G mutation adds to the deleterious effect of these mutations on ivermectin efficacy.
Identification of TMD Residues That Influence Ivermectin Sensitivity-We have already shown that the A288F mutation eliminates ivermectin agonist sensitivity (25). However, it has yet to be investigated whether this mutation also affects the sensitivity with which ivermectin potentiates glycine currents. Using a protocol as described in Fig. 1D, we observed significant potentiation of glycine-gated currents only at ivermectin concentrations Ͼ10 M (Fig. 2, A and B). This is consistent with A288F disrupting an ivermectin-binding site.
As discussed below, our GlyR model places Ala 288 near the extracellular end of M3, facing across the subunit interface toward M1 of the adjacent subunit (26 -29). From our ELICbased model, we identified 10 residues as having side chains close to or directed toward Ala 288 as follows: six in the M1 domain (Ile 225 , Gln 226 , Ile 229 , Pro 230 , Leu 233 , and Ile 234 ) and three M2 side chains that face away from the channel pore (Val 260 , Thr 264 , and Gln 266 which correspond to the 8Ј, 12Ј, and 14Ј, positions, respectively) and an M3 residue, Leu 291 , located one helical turn below Ala 288 . We individually mutated each of these residues to tryptophan, on the grounds that the bulky tryptophan side chain may occlude access to a putative ivermectin-binding site in the vicinity of Ala 288 . Tryptophan substitution has been employed at GABA A Rs to elucidate transmembrane domain binding sites for alcohols and anesthetics Inhibition, see text Inhibition, see text a p Ͻ 0.05 by unpaired t test relative to WT GlyR values b p Ͻ 0.01. (30,31), at nAChRs to define transmembrane helical structure and dynamics (32,33) and at P 2 X receptors to investigate movements induced by ivermectin (34). Glycine dose-response relationships were quantitated for all these mutant GlyRs to establish whether the substitutions were well tolerated by the receptor. As cells transfected with the I234W mutant GlyR exhibited no response to glycine or ivermectin, this mutant was not investigated further. Averaged dose responses for M1 domain mutant GlyRs are shown in Fig. 3A (left panel) with parameters of best fit summarized in Table 2. Similarly, averaged dose responses for GlyRs incorporating mutations in the M2 or M3 domains are shown in Fig. 3B (left panel) with mean parameters of best fit summarized in Table 2. Mean glycine EC 50 values of all mutant GlyRs were within an order of magnitude of the WT GlyR value, although significant reductions in I max were observed at both the V260W and T264W GlyRs ( Table 2). Mean ivermectin dose-response relationships for these mutants are plotted in Fig. 3, A and B (right panels), with averaged parameters of best fit summarized in Table 1. Sample ivermectin dose-response relationships for the I229W, P230W, and L233W mutant GlyRs are shown in Fig. 3C. Most mutants did not differ significantly from WT in their ivermectin sensitivity. However, the P230W GlyR exhibited a significantly increased ivermectin EC 50 value and a significantly reduced I max value relative to the WT GlyR value (Table 1). Also, at the Q226W mutant GlyR, maximal ivermectin-activated currents were significantly larger than the WT GlyR value, although ivermectin sensitivity was significantly decreased (Table 1). These results implicate Gln 226 and Pro 230 as ivermectin sensitivity determinants. The L233W, L291W, and T264W mutant GlyRs were not activated by ivermectin at concentrations of up to 100 M ( Fig. 3C and Table 1).
At the L233W and L291W mutant GlyRs, we noticed that responses to saturating glycine were reduced in magnitude or even absent when glycine was applied after ivermectin (e.g. Fig.  3C). At both mutant GlyRs, a 5-s application of 10 M ivermectin caused a significant (p Ͻ 0.05) decrease in the time taken for glycine-activated current to decay to half-maximum amplitude (Fig. 4, A and B). In addition, subsequent glycine-activated currents were dramatically decreased in magnitude (Fig. 4A). This effect of ivermectin proved to be irreversible. Although single applications of 1 or 3 M ivermectin had little effect on peak glycine-activated current magnitude, repeated applications of 1 M ivermectin produced a slowly developing but consistent increase in the desensitization rate (data not shown). Unfortunately, the slow onset and irreversibility of this effect made it difficult to quantitate its ivermectin sensitivity. To determine whether Leu 233 and Leu 291 were specifically required for ivermectin to exert an agonist effect, we investigated the effects of ivermectin on the highly nonconservative L233Q and L291Q mutant GlyRs. As summarized in Table 1, ivermectin activated both receptors with EC 50 and I max values that were not significantly different from WT values. Thus, the desensitization-enhancing effect of ivermectin appears to require bulky tryptophan side chains at either of the two sites. In the model presented below, these two residues lie opposite each other across the intersubunit interface.
The T264W (12Ј) mutant GlyR responded in an unusual manner to both glycine and ivermectin. Glycine activated a transient inward current that was followed by a distinct upward deflection that persisted for the duration of the glycine application (Fig. 4C). Consistent with a previous study that observed a similar phenomenon in mutant homomeric 1 GABA A Rs (35), we speculate that prolonged glycine application inhibited a leak current through these receptors. Ivermectin produced an irreversible dose-dependent inhibition of this leak current (Fig.  4C), with a mean IC 50 of 2.5 Ϯ 1.4 M and an n H value of 1.3 Ϯ 0.2 (both n ϭ 4). This IC 50 value, which is not significantly different from the WT value, suggests that ivermectin sensitivity is not dramatically affected by the mutation.
We next tested whether the M1 domain tryptophan substitutions impaired ivermectin sensitivity in the A288G mutant  GlyR. Whereas the P230W mutation produced an ivermectin EC 50 value 7-fold higher than the WT value, the P230W/ A288G value (5.6 Ϯ 0.3 M) was 175-fold greater than the value of 0.032 Ϯ 0.008 M at the A288G mutant GlyR. Similarly, mutations of residues surrounding Pro 230 on the same face of M1, I225W/A288G, Q226W/A288G, and I229W/A288G mutant GlyRs, exhibited EC 50 values 5-8-fold higher than A288G mutant GlyR alone, whereas the single mutants did not differ from WT values (Table 4). Thus, Pro 230 emerges as a crucial ivermectin sensitivity determinant at the A288G mutant GlyR. As discussed below, the above results are consistent with ivermectin binding at the M3-M1 intersubunit interface. As several other residues in this region have previously been implicated as binding sites for alcohols, anesthetics, and neurosteroids in GABA A Rs and GlyRs, we tested whether these resi-dues may also include ivermectin-binding sites. As Ser 267 (at the 15Ј position of the M2 domain) in the ␣1 GlyR has been implicated as an alcohol-binding site (36,37), we investigated the ivermectin sensitivity of the S267I mutant GlyR. As summarized in Table 1, this mutant GlyR exhibited an identical ivermectin sensitivity to the WT GlyR, although it did show a tendency toward higher I max and n H values. The potency with which ivermectin potentiated EC 10 glycine currents at the S267I mutant GlyR was also similar to WT (Table 3). Thus, Ser 267 does not contribute to an ivermectin site. In the GABA A R, an ␣1 subunit M1 domain 12Ј Thr residue (corresponding to Ile 234 in the GlyR) has been proposed to contribute to an intersubunit neurosteroid site that is eliminated via the Thr-to-Ile mutation and only slightly altered by the Thr-to-Ser mutation (38). As the ␣1 GlyR contains an endogenous Ile at this position, we investigated the effect of the reverse I234S  Table 1, this mutation had no effect on ivermectin sensitivity. A GABA A R etomidate-binding determinant at an ␣-subunit Met residue corresponding to Leu 233 in the ␣1 GlyR (27) can also be eliminated as a potential ivermectin sensitivity determinant on the grounds that the L233Q mutation had no effect on ivermectin potency (Table 1). Finally, a GABA A R neurosteroid-binding site at a ␤2-subunit Tyr residue corresponding to Trp 286 at the GlyR was shown to be eliminated by a Tyr-to-Phe substitution (38). However, the ␣1 GlyR W286F mutation produced only a moderate (2-fold) increase in the ivermectin EC 50 value (Table 1). We also investigated the less conservative W286A mutation, but it did not express. Thus, we conclude that none of these residues are likely to contribute to an ivermectin-binding site.

mutation. As shown in
Evidence for Ivermectin-binding Site Spanning Adjacent Subunits-To discriminate experimentally between intrasubunit and intersubunit locations of the ivermectin-binding site, we examined the ivermectin sensitivity of GlyRs formed from co-expression of P230W/A288G and A288F mutant subunits, which have low (5.6 M) or no ivermectin sensitivity, respectively. Our rationale was that if the ivermectin-binding site is formed within a single subunit, then the resulting mixed receptors should be ivermectin-insensitive (i.e. EC 50 Ͼ5 M) as each subunit is individually ivermectin-insensitive. Alternatively, if the resultant recombinant receptors are potently activated by ivermectin, then this site must be formed at those subunit interfaces that contain the two residues "permissive" for ivermectin sensitivity; (ϩ) M3 288-Gly (from P230W/A288G mutant) and (Ϫ) M1 230-Pro (from A288F mutant). Assuming subunits recombine randomly to produce receptors with all possible stoichiometries, the number of putative intersubunit ivermectin sites (i.e. interfaces containing Gly 288 on one face and Pro 230 on the other) will range from 0 to 2 per receptor (17). As shown in the example in Fig. 5A, ivermectin did indeed potently activate currents in receptors formed by co-expression of P230W/ A288G and A288F mutant subunits. These currents exhibited a mean ivermectin EC 50 value of 0.7 Ϯ 0.3 M, an n H of 1.5 Ϯ 0.1,  and an I max of 3.1 Ϯ 1.0 nA (all n ϭ 4). This EC 50 shows higher sensitivity than the sensitivity of each subunit when expressed alone (Fig. 5B), suggesting that ivermectin sites are located at subunit interfaces and that potent receptor activation can be achieved with two bound ivermectin molecules. Our model below places these two residues directly opposite each other at the opening to a cavity at the intersubunit interface.
Selamectin Exhibits Reduced Agonist Efficacy at ␣1 GlyR and ␣3B GluClR-In an attempt to define the molecular interaction between ivermectin and the GlyR, we sought to identify moieties crucial for its potency and efficacy. We previously showed that the direct agonist EC 50 and I max values for doramectin, emamectin, eprinomectin, and moxidectin did not differ significantly from those for ivermectin at either the WT or A288G GlyRs (21). All these compounds share a common structure at the benzofuran moiety but vary in structure at other groups. As selamectin differs in structure from these compounds at the benzofuran moiety, we compared the effects of selamectin and ivermectin at the ␣1 GlyR. Selamectin contains an NOH group at the C05 position, whereas the other derivatives contain an OH group (abbreviated hereafter as C05-NOH and C05-OH, respectively). Selamectin, at concentrations up to 30 M, activated no current at the WT GlyR. However, as shown in the example in Fig. 6A, it potently potentiated glycine currents. The averaged selamectin potentiating dose-response relationship, together with that for ivermectin reproduced from Fig. 1F, is shown in Fig. 6B. Selamectin potentiation exhibited a mean EC 50 of 3.6 Ϯ 0.9 M and an n H of 3.2 Ϯ 0.7 (both n ϭ 5). As both values did not differ significantly from those for ivermectin at the WT GlyR (Table 4), we tentatively conclude that selamectin binds with a similar potency to ivermectin but is unable to directly gate the receptor.
The A288G GlyR was directly activated by selamectin with a mean EC 50 value of 2.0 Ϯ 0.3 M, an n H value of 2.1 Ϯ 0.1, and an I max of 2.6 Ϯ 0.2 nA (all n ϭ 4). As this selamectin EC 50 value was significantly higher than that of ivermectin at this mutant GlyR (p Ͻ 0.001), we concluded that the ivermectin C05-OH group was essential for high efficacy at highly ivermectin-sensitive Cys-loop receptors as well.
Molecular Modeling of a Putative Ivermectin-binding Site-To model the ivermectin-binding site, we carried out computational docking of ivermectin to our ELIC-based (Protein Data Bank code 2VL0) ␣1 GlyR model within a large box surrounding Ala 288 . We found no significant differences in docking results in the WT relative to the A288G mutant GlyR (Fig. 7, A-C). As noted above, Ala 288 and Pro 230 (Fig. 7, shown in red and magenta, respectively) face each other across the subunit interface, either side of the opening to a cavity at the intersubunit interface. We hypothesized that the ivermectin-binding site may be within this cavity, but access of ivermectin to the cavity may require significant conformational change, not catered for in our model, even though we allowed flexible side chains for seven residues surrounding Ala 288 . Consistent with this hypothesis of a cavity site, Thr 264 (from (ϩ) side, shown in green) and Gln 266 (from (Ϫ) side, shown in blue) contribute to the lining of the cavity, and Trp substitutions of these residues caused an inhibitory effect of ivermectin and an increased ivermectin sensitivity, respectively. Other residues that, in the A288G background, showed reduced ivermectin sensitivity when substituted with Trp, Ile 225 , Gln 226 , and Ile 229 also surround the entrance to the cavity (shown in Fig. 7, A-C, light pink). The residues Leu 233 and Leu 291 (shown in yellow), that when substituted with Trp abolished activation by ivermectin but retained sensitivity to ivermectin in the form of increased desensitization, also face each other across the subunit interface one helical turn below Pro 230 and Ala 288 , respectively.
Our binding site hypothesis independently corroborates the recently published ␣ GluClR crystal structure with ivermectin bound (10). We then employed the same procedure as above to dock ivermectin onto a model of the ␣1 GlyR based on the ␣ GluClR structure (Protein Data Bank code 3RIF). As shown in Fig. 7, D and E, this produced a binding orientation very similar to that seen in the crystal structure, suggesting that ivermectin binds in a similar pose to both receptors. This in turn suggests that the M1 and M3 domains in our original ELIC-based model were located too close to each other to allow ivermectin access to its binding site in the intersubunit cavity. Interestingly, we saw only slight differences in the orientation of ivermectin docked to WT and A288G GlyRs, colored gray and green, respectively, in Fig. 7, B and C, with only a slight shift to accommodate the extra bulk of the Ala 288 methyl group. Its predicted binding energies to the WT and A288G models were also similar.
Comparison of ␣1 GlyR with ␣ GluClR Ivermectin Binding Interactions- Fig. 8 shows a sequence alignment of M1, M2, and M3 domains of the ␣1 GlyR and the crystallized C. elegans ␣ GluClR. It also includes several other GluClRs to be considered below. Using ␣ GluClR numbering, Leu 218 , Ser 260 , and Thr 285 were seen to form crucial H-bonds with ivermectin in the crystal structure (10). These residues (plus their homologues in other receptors) are colored blue in the alignment of Fig. 8, and the residues seen to form van der Waals interactions with ivermectin are colored green. The residues identified in this study as crucial ivermectin determinants in the ␣1 GlyR (i.e. Pro 230 and Ala 288 ) are colored red in Fig. 8.
The H-bond with Leu 218 is formed with the backbone carbonyl so its role in ivermectin binding is not readily tested by mutagenesis. Nevertheless, as the availability of the Leu 218 carbonyl for H-bonding is due to helical disruption by the conserved M1 proline one helical turn lower, it is likely that this bond is conserved in other Cys-loop receptors.
As we had not probed the role of the M3 H-bonding residue (Thr 285 in ␣ GluClR and Leu 292 in ␣1 GlyR) above, we investigated the effect of the L292T (␣1 GlyR3 ␣ GluClR) mutation. If ivermectin binds identically to ␣1 GlyRs and ␣ GluClRs, this mutation should introduce an H-bond that increases ivermec-tin affinity. Conversely, even if no H-bond is formed, this mutation may enhance ivermectin affinity by increasing the space available for ivermectin to bind in the TMD interface crevice. The L292T mutant ␣1 GlyR was found to exhibit a mean ivermectin EC 50 of 0.29 Ϯ 0.01 M (n ϭ 4 cells), significantly lower than that of the WT GlyR (p Ͻ 0.05).
In the ␣ GluClR structure, Ser 260 (Ser 15 Ј) forms an H-bond with the ivermectin C05-OH that was proposed to be essential for direct agonist activation of Cys-loop receptors (10). Although our docking suggested a similar H-bond exists in the GlyR, we showed above that mutation of the equivalent residue S267I, which eliminates any H-bonding propensity, had no effect on ivermectin sensitivity (Table 1), indicating this bond is either not functionally important or is not present in the ␣1 GlyR. The membrane-exposed surface of the TMD is lighter in color. Residues Ala 288 and Pro 230 are shown in red and magenta, respectively. Residues surrounding Pro 230 , Ile 225 , Gln 226 , and Ile 229 are shown in light pink, and residues Leu 233 and Leu 291 from opposing subunits are shown in yellow. B, closer view of the region surrounding these two residues, with best docks of ivermectin at the WT and A288G mutant GlyRs shown in gray and green, respectively. C, ivermectin is docked as in B but the model is rotated 90°to be viewed from the extracellular space with the membrane plane parallel to the plane of the page and a slice taken through the model to reveal a cavity at the subunit interface with the entrance bordered by Ala 288 and Pro 230 and residues Thr 264 (ϩ), shown in green, and Gln 266 (Ϫ), shown in blue, contributing to the lining of the cavity. D, this view corresponds to B but is based on the ␣ GluClR model instead of ELIC. E, this view corresponds to C but is based on the ␣ GluClR model instead of ELIC.

DISCUSSION
Functional Evidence for ␣1 GlyR Ivermectin-binding Site at the TMD Subunit Interface-The A288F substitution abolished ␣1 GlyR sensitivity to both the direct agonist and glycine potentiating effects of ivermectin. The P230W substitution in M1 decreased sensitivity to the direct ivermectin agonist effect by 7-fold in the WT GlyR and by 100-fold in the A288G mutant GlyR. The P230W mutation also decreased the glycine potentiating effects of ivermectin (Table 3). Molecular modeling places Ala 288 and Pro 230 residues from the neighboring subunits into close proximity at the TMD intersubunit interface, at either side of the entrance to an interface cavity (Fig. 7). Our experiments employing heteromeric mutant subunits (Fig. 5) provide independent experimental support for an ivermectin site at this interface. However, computational docking using our ELIC-based GlyR model did not reveal a convincing binding site for ivermectin that would be significantly disrupted by either the A288F or P230W mutations.
Selectivity of Ivermectin Analogues-The most useful technique for functionally characterizing molecular interactions is mutant cycle analysis. Unfortunately, as commercially available ivermectin analogues invariably differ from one another by more than one molecular group, mutant cycle analysis is not currently feasible for probing ivermectin-binding interactions. We have previously shown that emamectin, eprinomectin, moxidectin, and doramectin, which all vary from ivermectin at both disaccharide and spiroketal groups, exhibit similar potencies to ivermectin at WT and Ala 288 -substituted GlyRs (21). Selamectin, however, demonstrated a greatly decreased potency at the ␣1 GlyR and a moderately reduced potency at the ␣3B GluClR. Selamectin contains a monosaccharide instead of the disaccharide of ivermectin, a phenyl ring on the spiroketal, and an oxime in place of the hydroxyl at C05. Its decreased potency suggests that the C05-OH is crucial for interactions with the TMD. This is reflected in the decreased anthelmintic potency of ivermectin analogues with C05 substituents (39,42). Consistent with this, our model predicted that the ivermectin C05-OH interacts with the TMD intersubunit site.
Comparison of Our Functional Data with ␣ GluClR Crystal Structure-Although the entry to the TMD intersubunit cavity in our ELIC-based model was not wide enough for docking ivermectin, the ␣ GluClR-ivermectin crystal structure (10) clearly shows ivermectin lodged within this cavity. This new structure provided a template for us to refine our modeling and docking of ivermectin to the GlyR, and indeed our docking orientation proved very similar to that seen in the crystal structure (Fig. 7). The ␣ GluClR-ivermectin structure shows a distance of 9.4 Å between M1 and M3 (Leu 218 -Gly 281 C␣s), whereas the corresponding distances in ELIC and GLIC are 7.4 and 6.4 Å, respectively. Note that ␣ GluClR and GLIC are in the putative open state, whereas ELIC is in the putative closed state. Thus cavity width does not appear to correlate with channel state, although "wedging" the cavity open with ivermectin (10) certainly seems to favor channel opening.
The GlyR A288G mutation dramatically lowered the ivermectin EC 50 , although the docking studies showed only slight differences in binding orientations and predicted binding ener-gies at WT and A288G GlyRs (Fig. 7). Thus, our model provides no clear explanation for the observed EC 50 differences. We postulate that rather than affecting binding per se, the methyl group of Ala 288 reduces access to the interface cavity and consequently reduces the on-rate and apparent affinity for ivermectin. The inhibition of access to the cavity would be greatly amplified in the A288F mutant, with the possible additional effect of direct steric inhibition of ivermectin binding.
Hibbs and Gouaux (10) identified an H-bond between Ser 260 (Ser 15 Ј) and ivermectin C05-OH in the ␣ GluClR that they proposed was crucial for both high ivermectin affinity and direct agonist efficacy. However, we showed this bond is not required for high affinity ivermectin activation of the ␣1 GlyR. Moreover, because the H. contortus ␣ GluClR, the C. elegans ␣3B GluClR, and the H. contortus ␣3B GluClR all contain endogenous Ala residues at the corresponding 15Ј position (Fig. 8) and are at least as ivermectin-sensitive as the ␣ GluClR (40,43), it is evident that this H-bond is not necessarily required for high ivermectin sensitivity or direct ivermectin activation in those receptors as well. If a slightly greater distance cutoff is allowed for defining H-bonds, then other potential H-bonds can be identified between ivermectin C05-OH and Gln 219 in M1 and Asn 264 in M2 of the ␣ GluClR structure and equivalent residues Gln 226 and Arg 271 in our 3RIF-based GlyR model. Perhaps these H-bonds render the H-bond with Ser 15 Ј functionally redundant.
Hibbs and Gouaux (10) also identified a crucial H-bond between the ivermectin spiroketal oxygen and an M3 Thr (Thr 285 using ␣ GluClR numbering). As the ␣1 GlyR contains a non-H-bonding leucine (Leu 292 ) at the corresponding position (Fig. 8), it is evident this H-bond is not required for high ivermectin sensitivity at the WT or A288G ␣1 GlyRs. Moreover, as several other GluClRs contain an endogenous Ala at this position (Fig. 8), it is evident this H-bond is not required for high ivermectin sensitivity at some GluClRs as well. Nevertheless, we found that the ␣1 GlyR L292T mutation did increase ivermectin sensitivity, possibly consistent with a role for this H-bond.
The third H-bond, between ivermectin and the Leu 218 backbone of the ␣ GluClR, occurs because of the break in the helix due to the highly conserved M1 Pro. As the ␣1 GlyR contains this Pro and a conserved residue (Ile 225 ) at the Leu position, we infer the same H-bond is likely to exist in the ␣1 GlyR. Although not directly testable by conventional mutagenesis, our data (Tables 1 and 3) and docking studies are consistent with this notion.
Finally, Hibbs and Gouaux (10) reported a network of van der Waals interactions between the ivermectin disaccharides and residues in the ␣ GluClR M2-M3 loop that they thought might be important for allosteric interactions with the LBD. We have previously reported that moxidectin, which lacks both sugars, is equipotent with ivermectin at the ␣1 GlyR (21), arguing against a crucial role for these interactions in the GlyR.
Ivermectin Gating Determinants at the LBD-TMD Interface-If ivermectin interacts with the TMD of the ␣1 GlyR, why is its ability to open the channel affected by mutations at the LBD-TMD interface? The L142C, Y222C, and Y279F mutations each caused a large rightward shift in the ivermectin activation EC 50 value, although ivermectin potentiation of glycine currents showed a similar sensitivity as seen in the WT GlyR. The same mutations also produced large rightward shifts in glycine activation EC 50 values (Table 2). Similar effects on ivermectin and glycine sensitivities were observed when the above three mutations were investigated on the background of the A288G mutation. Together, these results strongly suggest that the L142C, Y222C, and Y279F mutations do not disrupt an ivermectinbinding site but rather reduce ivermectin gating efficacy. This is consistent with the known roles of the conserved Cys-loop, the pre-M1 domain, and the M2-M3 domain in gating the GlyR (44) and suggests that ivermectin activation involves a global conformational change that propagates to the LBD rather than simply a localized TMD conformational change. We therefore speculate that other ivermectin sensitivity determinants that have previously been identified in the TMD and interfacial domains of GluClRs (7,9,40) also disrupt the gating rather than the binding mechanisms of ivermectin.
Allosteric Effects of TMD Mutations Near the Ivermectin Site-The P230W mutation decreased both glycine sensitivity and ivermectin sensitivity. It appears the dominant effect of this mutation is to rearrange the M1 helix so as to abolish the H-bond between ivermectin and the Ile 225 backbone. It is interesting that the Q226W, P230W, and I234W mutant GlyRs all showed decreased glycine sensitivities (with no function at all for I234W) compared with the I225W, I229W, and L233W mutant GlyRs, suggesting that the effects of P230W are not specific to ivermectin. However, the decrease in glycine sensitivity caused by the Trp substitution at Gln 226 , Pro 230 , and Ile 234 might simply reflect their apposition with M3 of the adjacent subunit, whereas the Trp substitutions at Ile 225 , Ile 229 , and Leu 233 are less disruptive because their side chains are directed toward the surrounding lipids (45). However, it is possible that Pro 230 mutations might also affect ivermectin gating efficacy for two main reasons. The first is that Pro 230 substitutions will remove a central kink in the M1 ␣-helix that will alter interactions between the helix and neighboring domains throughout its length. This could have wide ranging effects on allosteric interactions involving this domain. The second related reason is that the corresponding Pro-to-Ala mutation in the ␤2 subunit of the ␣1␤2␥2 GABA A R affects both GABA sensitivity and barbiturate potentiation (46), consistent with a disruption of the linkage between the GABA-binding site and the channel gate.
The threshold concentrations at which ivermectin begins to enhance desensitization at L233W and L291W mutant GlyRs are 1-10 M, not remarkably different from WT ivermectin EC 50 values. The glycine sensitivities of both mutant receptors are also significantly higher than WT GlyR values ( Table 2). The nonconservative L233Q and L291Q substitutions had no effect on the mode of action of ivermectin, implying that the large hydrophobic Trp side chain is required for this effect, potentially blocking the conformational change normally favored by ivermectin. We therefore conclude that ivermectin binds with normal affinity to the L233W and L291W mutant GlyRs but promotes a desensitized state, rather than an open state. From our results, it is unclear if this effect is due to an altered ivermectin binding conformation imposed by the Trp substitutions or a preference of these mutants for a glycinebound desensitized state.
Conclusion-Our docking simulations predict that ivermectin binds to the ␣1 GlyR in a similar pose as observed in the ␣ GluClR crystal structure. Ivermectin binding to the ␣ GluClR was shown to be mediated by H-bonds with Leu 218 , Ser 260 , and Thr 285 and a network of van der Waals interactions (10). We showed that H-bonds with residues equivalent to Ser 260 and Thr 285 are not required for high ivermectin sensitivity at either the ␣1 GlyR or three other GluClRs. As the H-bond with the Leu 218 or equivalent residues (e.g. GlyR Ile 225 ) is via the carbonyl backbone, its role cannot be readily tested and is likely to be conserved along with the conserved Pro that exposes this carbonyl. We also show that van der Waals interactions between the ivermectin sugars and ␣1 GlyR M2-M3 loop residues are not important for high ivermectin affinity or efficacy.
The ␣1 GlyR A288F mutation eliminated the direct activation and glycine-potentiating effects of ivermectin, presumably by blocking access to the cavity. The effect of this mutation contrasted with those of several mutations in the conserved Cys-loop, pre-M1 domain, and M2-M3 loop that disrupted only the direct activation but not the glycine-enhancing effects of ivermectin. We conclude these later residues disrupted ivermectin gating efficacy, consistent with the known role of this domain in mediating glycine agonist transduction.
In summary, our results indicate that the ivermectin-binding mechanisms as revealed by the ␣ GluClR structure cannot be generalized to other Cys-loop receptors with comparable ivermectin sensitivities. Understanding these mechanisms will be crucial for designing new drugs as anthelmintics and as therapies for a wide range of human neurological disorders.