An amino acid substitution in the pore region of a glutamate-gated chloride channel enables the coupling of ligand binding to channel gating.

Many of the subunits of ligand-gated ion channels respond poorly, if at all, when expressed as homomeric channels in Xenopus oocytes. This lack of a ligand response has been thought to result from poor surface expression, poor assembly, or lack of an agonist binding domain. The Caenorhabditis elegans glutamate-gated chloride channel subunit GluClβ responds to glutamate as a homomeric channel while the GluClα subunit is insensitive. A chimera between GluClα and GluClβ was used to suggest that major determinants for glutamate binding are present on the GluClα N terminus. Amino acid substitutions in the presumed pore of GluClα conferred direct glutamate gating indicating that GluClα is deficient in coupling of ligand binding to channel gating. Heteromeric channels of GluClα+β may differ from the prototypic muscle nicotinic acetylcholine receptor in that they have the potential to bind ligand to all of the subunits forming the channel.

Many of the subunits of ligand-gated ion channels respond poorly, if at all, when expressed as homomeric channels in Xenopus oocytes. This lack of a ligand response has been thought to result from poor surface expression, poor assembly, or lack of an agonist binding domain. The Caenorhabditis elegans glutamate-gated chloride channel subunit GluCl␤ responds to glutamate as a homomeric channel while the GluCl␣ subunit is insensitive. A chimera between GluCl␣ and GluCl␤ was used to suggest that major determinants for glutamate binding are present on the GluCl␣ N terminus. Amino acid substitutions in the presumed pore of GluCl␣ conferred direct glutamate gating indicating that GluCl␣ is deficient in coupling of ligand binding to channel gating. Heteromeric channels of GluCl␣؉␤ may differ from the prototypic muscle nicotinic acetylcholine receptor in that they have the potential to bind ligand to all of the subunits forming the channel.
The prototype of a ligand-gated ion channels is the muscle nicotinic acetylcholine receptor where the major determinants of ligand binding are located on the ␣ subunits (2)(3)(4). Binding of two agonist molecules leads to channel gating. The number of agonist molecules necessary to gate ligand-gated chloride channels, as well as the subunits which bind ligand remain matters of controversy (5,6). GABA A receptors have an unknown stoichiometry and there is evidence for ligand binding determinants on ␣, ␤, ␥, and perhaps ␦ subunits. Glycine receptors are thought to have a stoichiometry of 3␣2␤, with the ␣ subunits carrying the ligand binding domains (5). However, very small responses to glycine in Xenopus oocytes injected with the glycine ␤ subunit suggest that a glycine binding site exists on the ␤ subunit (8).
When co-expressed in oocytes, GluCl␣ and GluCl␤ form heteromeric channels gated by glutamate and the widely used antiparasitic agent ivermectin (1). Ivermectin is a widely used broad spectrum anthelmintic/insecticidal agent used to control parasitic nematodes in man and animals (9). Ivermectin potentiates the glutamate response of heteromeric GluCl␣ϩ␤ channels indicating that ivermectin acts as an allosteric modulator of the channel (1, 10). Homomeric GluCl␤ channels are directly gated with glutamate demonstrating that GluCl␤ contains all the determinants for ligand binding and coupling to channel gating (1). Homomeric GluCl␣ channels are insensitive to glutamate, but are directly activated by ivermectin (1). We demonstrate that GluCl␣ is a ligand binding subunit deficient in coupling to channel gating. Heteromeric channels of GluCl␣ϩ␤ have the potential to bind ligand to all of the subunits forming the channel.

EXPERIMENTAL PROCEDURES
Electrophysiological Recordings-Xenopus oocytes were prepared, injected (in vitro RNA, 0.1-10 ng), and membrane currents recorded as described previously (10, 11). Recordings were made in standard frog saline consisting of (in mM): NaCl (115), KCl (2), MgCl 2 (1), CaCl 2 (1.8), HEPES (10) adjusted to pH 7.5 with NaOH. Agarose-cushion electrodes were made as described (12). The water soluble derivative 22,23-dihydroavermectin B1 a 4Љ-O-phosphate (IVMPO 4 ) was used for these studies (11). Concentration response curves were fit to a modified Michaelis-Menten equation: where I max is the maximal response, [D] is the drug concentration, EC 50 is the drug concentration for half-maximal effect, and h is the Hill coefficient. All data in text and figures are means Ϯ S.E. Current voltage (I/V) relationships were determined using 1-s voltage ramps from Ϫ100 to 60 mV in the presence and absence of IVMPO 4 . Drug-sensitive currents were generated by subtracting drug-free from drug-containing data. Ion substitutions were accomplished by replacing NaCl in standard frog saline with KCl, choline-Cl, or sodium isethionate as indicated. The bath was connected to the ground through a 3M KCl-agar bridge for experiments in which external chloride was replaced. To limit contributions from calcium-activated chloride currents to the I/V relationships, CaCl 2 was removed from the external solutions, and oocytes were preincubated containing 10 M BAPTA-AM (13).
Design and Construction of Molecular Clones-Homologous positions between ligand-gated ion channel subunits were determined using the pretty plot alignment from the GCG package, version 7 (Genetic Computer Group, Madison, WI). The EcoRI restriction sites were introduced by methods described in Ref. 14. The chimeric receptor was constructed as a SalI-EcoRI fragment of GluCl␣ and a EcoRI-NotI fragment of GluCl␤. Base pair substitutions in the codon for amino acid 308 in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The Extracellular Domain of GluCl␣ may Contain Determinants of Glutamate Binding-The major determinants of the ligand binding site in ligand-gated ion channels are thought to reside on the large extracellular N-terminal (3)(4)(5)(6). Chimeric channels between the 5HT1c and ␣7 nicotinic acetylcholine receptor further confirmed this localization (18). To determine if the extracellular domain of GluCl␣ contains major components of a glutamate binding site we constructed a chimera between GluCl␣ and GluCl␤ (GluCl␣N-␤C) ( Fig. 2A). A glutamate-sensitive current with an EC 50 of 530 Ϯ 90 nM was observed when oocytes were injected with RNA from GluCl␣N-␤C (Fig. 2B). Ibotenic acid (500 nM) activated 82 Ϯ 3% (n ϭ 3) of the current activated with 500 nM glutamate. D-Glutamate, GABA, and glycine (1 M) were inactive. This pharmacology is consistent with the agonist profile of glutamate-gated chloride channels and the IVMPO 4 -dependent glutamate current of GluCl␣. Oocytes injected with the chimera also had a large holding current and a resting potential close to the Nernst potential for chloride (Ϫ31 Ϯ 2 mV, n ϭ 10). The holding current was inhibited to the level of a water-injected oocyte with 50 M picrotoxin (n ϭ 7, Fig. 2C), a non-competitive inhibitor of ligand-gated anion channels (10, 17,19,20). The current/voltage relationship of the leak current was strongly outwardly rectifying, similar to the rectification of the GluCl␤ pore (1). The results with GluCl␣N-␤C suggest that the N-terminal extracellular domain of GluCl␣ contains major determinants of glutamate binding. They also suggest that the C-terminal part of GluCl␣ does not couple ligand binding to channel gating.
Point Mutations in M2 Creates a Glutamate-sensitive Channel-Amino acid substitutions in the C-terminal part of the GluCl␣ channel showed that it was possible to obtain direct glutamate gating in the absence of an IVMPO 4 -induced conformational change. The amino acid threonine in position 308 was replaced with amino acids known to occur in the homologous position of ligand-gated ion channels (Ala, Gly, Ser, Cys, Pro, Val, and Leu) (Fig. 3A). The mutant channels were assayed for expression by using 5 M IVMPO 4 . All mutant subunits except T308L expressed robust (Ͼ1 A) IVMPO 4 -sensitive currents. The EC 50 for IVMPO 4 was not dramatically altered in the mutants as judged by the threshold concentration of IVMPO 4 needed to activate a current (50 -100 nM, data not shown). No glutamate-induced currents (10 and 50 mM) were observed for T308S, T308C, T308V, T308L. However, the mutants T308A, T308G, and T308P all responded to glutamate (Fig. 3B). The highest sensitivity to glutamate was observed in the mutant T308P with an EC 50 of 1.4 mM followed by T308A and T308G (Fig. 3B). Gating efficiency was estimated by normalizing the current activated with 50 mM glutamate to the current activated with 5 M IVMPO 4 (Fig. 3C). Gating efficiency was highest for T308P (90 Ϯ 9%) followed by T308G (37 Ϯ 5%) and T308A (14 Ϯ 2%). The three mutants that responded to glutamate were also activated with 1 mM ibotenate, while GABA, glycine, and D-glutamate (1-10 mM) were inactive. These results demonstrate that homomeric GluCl␣ channels contain, in the absence of IVMPO 4 , all the determinants for a glutamate binding site.
The amino acid substitutions at position 308 of GluCl␣ also altered the rate of desensitization of the glutamate-and IVMPO 4 -sensitive currents (Fig. 3, B and C). This result is consistent with the alterations in desensitization properties associated with this position in a Drosophila GABA receptor (21). The mutations T308G, T308A, T308C, and T308V all showed an increased rate of desensitization in the presence of IVMPO 4 when compared to GluCl␣ or T308P (Fig. 3C and data  not shown). The rate of desensitization of the IVMPO 4 response could not predict the occurrence of an IVMPO 4 -independent glutamate current. Furthermore, the desensitization rate of the glutamate current did not reflect the efficiency of channel gating (Fig. 3, B and C). These results suggest that the glutamate current observed in the mutants T308G, T308A, and T308P was independent of the alterations in desensitization.
Mutations at position 308 had little effect on the pore properties of homomeric channels (Fig. 4). The IVMPO 4 -sensitive current of GluCl␣ (1) and the T308P mutant had linear I/V curves, while the I/V curve of T308A was slightly outwardly rectifying. All were anion selective since replacement of extracellular sodium with potassium or choline had no effect on E rev , or the shape of the current/voltage relationship (data not shown). The reversal potentials after replacement of extracellular NaCl with sodium isethionate were 2 Ϯ 1 (GluCl␣), 3 Ϯ 4 (T308P), and 1 Ϯ 1 mV (T308A), indicating a permeability ratio of isethionate to chloride of approximately 0.2 (1,22,23). DISCUSSION We present three independent lines of evidence that homomeric GluCl␣ channels contain a glutamate binding site with pharmacology common to glutamate-gated chloride channels. First, we observe an IVMPO 4 -dependent glutamate current indicating that IVMPO 4 induced a conformational change that allowed gating of the channel with glutamate (Fig. 1). Second, we show that the large extracellular loop of GluCl␣ is sufficient to cause the gating of a GluCl␤ pore (Fig. 2). This result suggests that major determinants of glutamate binding are present on the GluCl␣ N-terminal. Most importantly, we were able to restore the complex mechanism of ligand binding to channel gating in a homomeric GluCl␣ channel by replacing one amino acid in the pore region (Fig. 3). Changing the naturally occurring threonine in M2 to a proline, glycine, or alanine enabled glutamate gating of GluCl␣ homomeric channels. Gating was independent of the side chain volume or the chemical properties of the amino acid. This result suggests that a subtle conformational change, like the introduction of a bend by proline or flexibility by glycine, was sufficient to reduce the energy needed for channel gating. A reduction in the free energy necessary to couple ligand binding to channel gating has also been observed for point mutations in the second membrane spanning domain of nicotinic acetylcholine receptors (24,25). GluCl␣ is a clear example of a ligand binding subunit which fails to couple ligand binding to channel gating as a homomeric channel. Other subunits of the ligand-gated ion channel superfamily that express poorly with low apparent affinity as homomeric channels may simply have poor coupling of ligand binding to channel gating.
Our results indicate that two separate subunits in heteromeric GluCl␣ϩ␤ channels carry independently all the determinants for ligand binding. It is tempting to speculate that coassembly with GluCl␤ provokes a subtle conformational change which leads to the coupling of the GluCl␣ binding site to channel gating. The interactions of these binding sites in a hetero- meric channel remains to be determined. Perhaps, binding to one type of ligand binding site determines channel opening while binding to another type influences desensitization. Alternatively, subconductance states commonly observed in ligandgated chloride channels (26,27) may be governed by number and type of ligand binding sites occupied.
The scheme of binding sites on more than one type of subunit probably carries over to other members of the ligand-gated chloride channel family. For example, there is evidence for homomeric channel formation of glycine ␤ subunits (8). Homomeric ␣ and ␤ subunits of the GABA A receptor have been expressed, however, with poor efficiency (28,29). Furthermore, ligand binding determinants for GABA binding may be present on the ␣, ␤, ␥, and ␦ subunits (6). Our results suggest that heteromeric GluCl␣ϩ␤ channels differ from prototypical muscle nicotinic acetylcholine receptor in that more than one type of subunit carries major determinants for ligand binding.  (30), and human glycine receptor ␣1 subunit (27). The amino acids Glu, Ala, Ser, Cys, Pro, Val, and Leu have been introduced in the position T308. B, current traces at the approximate EC 50 concentration of glutamate are shown. Dose-response relationships were obtained using 5-6 concentrations of glutamate from 0.1 to 50 mM. For the T308A and T308P mutants the current elicited with 50 mM glutamate was less than I max . We estimated I max by fitting individual dose-response curves to Equation 1 with I max as a free parameter. The estimated I max was then used to normalize data for each experiment. The smooth curves represent fits of the normalized data to Equation 1 for T308P (q, EC 50 ϭ 1.4 mM, h ϭ 1.3), T308A (f, EC 50 ϭ 8.7, h ϭ 1.1), and T308G (å, EC 50 ϭ 9.7, h ϭ 1.4). At least 4 dose-response curves were performed for each construct. C, glutamate-and IVMPO 4 -sensitive currents for GluCl␣, T308A, T308G, and T308P. Currents were recorded in glutamate and 1 min later in IVMPO 4 . The small response to 50 mM glutamate in oocytes injected with GluCl␣ was also observed in water-injected oocytes.