Subunit stoichiometry and arrangement in a heteromeric glutamate-gated chloride channel

Significance Cys-loop receptors (CLRs) are transmembrane ion channels activated by neurotransmitters to mediate chemoelectric excitation or inhibition throughout the nervous system. Hence, CLRs play a key role in our day-to-day life, from coordination of motions to cognition. Impairment of CLRs’ activity leads to various pathophysiological conditions. The CLR studied here is a glutamate-gated chloride-selective receptor (GluClR). GluClRs are unique to invertebrates, yet they are pharmacologically important because they serve as targets for ivermectin, an anthelmintic drug used to treat humans suffering from filarial diseases. This study provides better understanding of the subunit arrangement and stoichiometry of Glu-binding sites in GluClRs. The invertebrate glutamate-gated chloride-selective receptors (GluClRs) are ion channels serving as targets for ivermectin (IVM), a broad-spectrum anthelmintic drug used to treat human parasitic diseases like river blindness and lymphatic filariasis. The native GluClR is a heteropentamer consisting of α and β subunit types, with yet unknown subunit stoichiometry and arrangement. Based on the recent crystal structure of a homomeric GluClαR, we introduced mutations at the intersubunit interfaces where Glu (the neurotransmitter) binds. By electrophysiological characterization of these mutants, we found heteromeric assemblies with two equivalent Glu-binding sites at β/α intersubunit interfaces, where the GluClβ and GluClα subunits, respectively, contribute the “principal” and “complementary” components of the putative Glu-binding pockets. We identified a mutation in the IVM-binding site (far away from the Glu-binding sites), which significantly increased the sensitivity of the heteromeric mutant receptor to both Glu and IVM, and improved the receptor subunits’ cooperativity. We further characterized this heteromeric GluClR mutant as a receptor having a third Glu-binding site at an α/α intersubunit interface. Altogether, our data unveil heteromeric GluClR assemblies having three α and two β subunits arranged in a counterclockwise β-α-β-α-α fashion, as viewed from the extracellular side, with either two or three Glu-binding site interfaces.

The invertebrate glutamate-gated chloride-selective receptors (GluClRs) are ion channels serving as targets for ivermectin (IVM), a broad-spectrum anthelmintic drug used to treat human parasitic diseases like river blindness and lymphatic filariasis. The native GluClR is a heteropentamer consisting of α and β subunit types, with yet unknown subunit stoichiometry and arrangement. Based on the recent crystal structure of a homomeric GluClαR, we introduced mutations at the intersubunit interfaces where Glu (the neurotransmitter) binds. By electrophysiological characterization of these mutants, we found heteromeric assemblies with two equivalent Glu-binding sites at β/α intersubunit interfaces, where the GluClβ and GluClα subunits, respectively, contribute the "principal" and "complementary" components of the putative Glu-binding pockets. We identified a mutation in the IVM-binding site (far away from the Glu-binding sites), which significantly increased the sensitivity of the heteromeric mutant receptor to both Glu and IVM, and improved the receptor subunits' cooperativity. We further characterized this heteromeric GluClR mutant as a receptor having a third Glu-binding site at an α/α intersubunit interface. Altogether, our data unveil heteromeric GluClR assemblies having three α and two β subunits arranged in a counterclockwise β-α-β-α-α fashion, as viewed from the extracellular side, with either two or three Glu-binding site interfaces. allostery | Cys-loop receptors | ion channels | ivermectin | neurotransmitters G lutamate-gated chloride-selective receptors (GluClRs) are pentameric glutamate-gated chloride channels unique to invertebrates. They belong to the Cys-loop receptor superfamily of transmembrane oligomers that open an intrinsic cationic or anionic channel pore upon binding of neurotransmitters, such as ACh, serotonin, GABA, Gly, histamine, or Glu (1)(2)(3)(4)(5)(6)(7)(8)(9). GluClRs are specific targets for ivermectin (IVM), a broad-spectrum anthelmintic drug used to treat filarial diseases like onchocerciasis (river blindness) and elephantiasis (lymphatic filariasis) that afflict hundreds of millions of people worldwide (10,11). IVM is also broadly used in cattle, swine, and pets to kill gastrointestinal roundworms, lungworms, grubs, sucking lice, and mange mites (12). The high efficiency of IVM stems from its capacity to act as an agonist that keeps the receptor's ion channel continuously open (13)(14)(15)(16)(17)(18). Because the GluClR is chloride-selective, IVM causes sustained hyperpolarization across postsynaptic membranes in the parasitic nematodes. This long-term hyperpolarization leads to suppression of excitation in motor neurons and inhibition of locomotion (19); inhibition of the pharyngeal muscle activity, which interrupts with feeding behavior (20); and interruption of secretion processes that are crucial for evading the host immune system (21).
Importantly, the naturally occurring GluClR robustly responds to both Glu and IVM independently; therefore, it is considered to consist of both GluClα and GluClβ subunit types (13)(14)(15)(16)(17)(18). However, little is known about the stoichiometry and molecular arrangement of the subunits in heteromeric GluClRs. Furthermore, the aforementioned crystallographic observations (23) are consistent with earlier studies showing that Glu elicits current responses in homomeric GluClαRs only when applied after activation by IVM (14), which gives rise to the following question: Could an α/α intersubunit interface be formed in a heteromeric assembly, bind Glu, and functionally participate in the activation process even without IVM preassociation? To resolve this question, we clarified here the stoichiometry and positions of the α and β subunits in GluClα/βR heteromeric assemblies that carry mutations in both the putative Glu-and IVM-binding pockets.

Results
Can the Coupling Loops of the GluClα Subunit Mediate Channel Opening upon Glu Binding? Based on the capability of the WT homomeric GluClα (GluClαWT) receptor to respond to Glu only following exposure to IVM, it was suggested that IVM binding induces a conformational change that enables coupling of Glu binding at α/α intersubunit interfaces to the opening of the ion-channel gate (14,23). To explore this suggestion further, we used a strategy of microchimerism that is based on previous studies showing that in various Cys-loop receptors, the β1β2, Cys, and β8β9 loops of the LigBD interact with the M2-M3 loop of the pore domain to couple neurotransmitter binding to channel gating (23,(35)(36)(37)(38)(39)(40)(41)(42)(43)(44) (e.g., Fig. 1 A and C). These four loops are termed the coupling loops. Fig. 1D shows schemes of the WT GluClα and GluClβ subunits, as well as three microchimeric GluClβ subunits where we replaced the Cys loop, β8β9 loop, or both loops with the homologous loops of the GluClα subunit. These microchimeric subunits are termed GluClβ α-CysL , GluClβ α-β8β9L , and GluClβ α-Loops , respectively. Note that the C. elegans GluClα and GluClβ subunits share an identical β1β2 loop sequence, whereas their M2-M3 loop sequence is almost identical (Fig. S1).
CHO cells transfected with the GluClαWT subunit alone showed very weak responses to 10 mM Glu (135 ± 27 pA in eight cells; mean ± SEM), but responded well to 500 nM IVM ( Fig.  S2B; 14 cells). This observation is in line with the findings of Frazier et al. (45), who reported that HEK cells expressing GluClα homomers are responsive to IVM but not to Glu. CHO cells transfected with the GluClβWT subunit alone showed very weak, rare responses to 10 mM Glu (less than 230 pA in eight cells; Fig. S2A), in line with results obtained in HEK cells (45). No responses to 500 nM IVM in CHO cells transfected with the GluClβWT subunit alone were observed (10 cells), in agreement with the same observations in HEK cells (18,22). In contrast to these differential responses, cells cotransfected with both GluClαWT and GluClβWT subunits displayed robust responses to 1.5 mM Glu (EC 50 concentration) and 500 nM IVM (Fig.  S2C). We therefore deduce that robust responses to Glu and IVM (independently) in a cell cotransfected with the GluClαWT and microchimeric GluClβ subunits (Fig. S2 D-F) reflect the function of heteromeric GluClα/βR complexes. This deduction also applies for the site-specific mutants discussed further below. Fig. 2 shows representative current traces elicited by increasing Glu concentrations ( Fig. 2A) and the corresponding dose-response curves (Fig. 2B) for the heteromeric WT and microchimeric GluClRs. The Glu-EC 50 values specified in Table  S1 indicate that the apparent affinities of the GluClαWT/β microchimeric receptors for Glu were very close to the apparent affinity of the GluClαWT/βWT receptor. The Hill coefficients of all four receptors (Table S1) were >1, indicating their activation with positive cooperativity. Note that the Glu-EC 50 and the Hill coefficient determined here for the GluClαWT/βWT receptor (Table S1) are very close to those values determined in Xenopus oocytes [Glu-EC 50 = 1.36 ± 0.05 mM and Hill coefficient (n H ) = 1.7 ± 0.1] (13). Remarkably, Glu readily activates the GluClαWT/ β α-Loops receptor, all of whose LigBD's coupling loops are of the GluClα subunit ( Fig. 2 and Table S1). We thus conclude that the β1β2, Cys, and β8β9 loops of the GluClα subunit are inherently capable of coupling Glu binding to pore gating, with no need for IVM prebinding. Glu and IVM are shown as spacefilling models with carbon, oxygen, and nitrogen atoms colored in yellow, red, and blue, respectively. They are bound at the α(+)/α(−) intersubunit interface far away from each other: Glu in the extracellular LigBD and IVM in the upper part of the pore-domain periphery, between M1 (of the light gray subunit) and M3 (of the dark gray subunit). Note that in Cys-loop receptors, the principal and complementary faces of a neurotransmitter-binding pocket are formed by the (+) and (−) sides of two adjacent subunits, respectively. (B) Top view of the GluClα cryst pentamer showing five identical subunits, which are colored differently to highlight the intersubunit interfaces located between the (+) and (−) sides. (C) Space-filling models of residues belonging to the coupling loops, which create an extensive bond network at the interface between the LigBD and the ion-channel pore domain. (D) Schemes of GluClR subunits used in this study. The M1-M4 transmembrane segments are numbered 1-4. Different colors reflect differences in amino acid sequences (Fig. S1). Contribution of the GluClα Subunit (−) Side to Glu Accommodation in Heteromeric GluClR Mutants. The aforementioned observations brought us to the recognition that a thorough study of how the GluClα subunit contributes to Glu binding in heteromeric GluClRs is necessary. Therefore, we first introduced mutations in the (−) side of the GluClα subunit based on the crystal structure of the homomeric GluClα cryst R (23) [the (−) and (+) subunit sides are defined in Fig. 1 A and B]. This structure indicates that the δ-guanidino groups of α(−)R98 and α(−)R117 are at an appropriate distance to form ion pairing with the αand γ-carboxyl groups of Glu, respectively (Fig. 3D). A mutation that eliminated the charge and drastically reduced the side-chain size of α(−)R117, but kept hydrophilicity at this position (i.e., R→S), did not provide a functional GluClαR117S/βWT receptor. We therefore replaced the two Args (one at a time) with a more conservative and bulkier hydrophilic residue, Asn or Gln, which can function as hydrogen bond donor (or acceptor) with no capability to form salt bridges. A mutant having an αR98N substitution (GluClαR98N/ βWT receptor) provided robust responses, but very high Glu concentrations were necessary to reach saturation [ Fig. 3A, traces (Right) and brownish dose-response curve (Left)]. Note that to dissolve Glu, it was titrated with equimolar concentrations of NaOH; therefore, we did not change the Nernst potential for Cl − during Glu application. However, the osmolarity and negative charge of the external solution drastically increased during the application of high Glu concentrations (for 0.6 s). Even so, we assume that these factors did not affect the responses, as discussed in SI Text, section 1, in conjunction with Fig. S3.
In the case of the GluClαR117N/βWT receptor, the current responses did not allow us to analyze the dose-response relation reliably because they were very low (∼300 pA at 1 M Glu) and did not reach saturation, unlike in the case of the GluClαR98N/ βWT receptor. In contrast, introducing Q at position α(−)117, which has a longer side chain than N, created a responsive GluClαR117Q/βWT receptor that enabled us to determine its Glu-EC 50 and Hill coefficient ( Fig. 3A and Table S1).
The crystal structure also indicates that α(−)S182 forms a hydrogen bond with the γ-carboxyl group of Glu (23) (Fig. 3D). Preventing this hydrogen bonding in the heteromeric GluClαS182A/ βWT receptor produced an effect similar to the effect observed with the α(−)R98N and α(−)R117Q substitutions ( Fig. 3A and Table  S1). The drastic effects exerted by mutations in the GluClα(−) side raised the question of whether mutations at the homologous positions in GluClβ would exert the same effects.
To examine this hypothesis, we mutated residues β(+)F122, β(+)T229, and β(+)Y232 that might contact Glu, as predicted by sequence alignments (17,23) and our 3D homology model (Fig.  3D). We then coexpressed the mutated β subunits (one at a time), together with the αWT subunit, and found that they shifted the dose-response curves rightward (Fig. 3C). Table S1 shows the extent of increase in the Glu-EC 50 values, with the most prominent shift in the GluClαWT/βT229N and GluClαWT/ βT229W receptors (by approximately eightfold compared with the GluClαWT/βWT receptor). We infer that the GluClβ(+) side in the heteromeric assemblies generated here contributes the principal Glu-binding components. Daeffler et al. (22) added to the homomeric GluClβT283S receptor a β(+)T229A mutation (no. 203 in ref. 22), which increased the Glu-EC 50 to a much larger extent than observed here for the heteromeric GluClαWT/βT229N or GluClαWT/βT229W receptor. This difference can be attributed to the nature of the replacing amino acids. In the current study, we did not wish to change the chemical properties of the amino acids too much. This approach was adopted because the GluClαWT/βWT receptor inherently displays low affinity for Glu, which would probably make a dramatic increase in Glu-EC 50 difficult to probe. Hence, we kept the capability of the replacing residues at position β(+)T229 to act as hydrogen bond donors (Asn, Trp) or a hydrogen bond acceptor (Asn). We expected that the greater size of the replacing residues would interfere with, but not abolish, Glu accommodation. This expectation emerged because position β(+)229 is located on loop C, which caps the putative Glu-binding pocket but, on the other hand, is considered to be flexible and mobile (46) (Fig. 3 D and E). As to the β(+)Y232S substitution, we probably eliminated a cationpi interaction that was recently suggested to be formed in a homomeric GluClβR, between the β(+)Y232 aromatic ring and the α-amino nitrogen of Glu (22). Still, one cannot exclude hydrogen bonding between the hydroxyl group of the Ser we introduced at this position [β(+)232] and the α-amino nitrogen of Glu, which could explain the moderate effect of the β(+)Y232S mutation.
Stoichiometry of the Glu-Binding Sites in a Heteromeric WT GluClα/βR.
The results presented in the previous sections suggest that a β(+)/α(−) intersubunit interface is involved in Glu accommodation; so, how many such functional interfaces exist per heteropentamer? The various single-site mutant receptors discussed so far share with the GluClαWT/βWT receptor Hill coefficients smaller than 2 but clearly larger than 1 (Table S1). This property suggests that there is more than one Glu-binding site per heteropentamer. To determine the number of functional sites and their microscopic equilibrium dissociation constants for Glu binding in the heteromeric GluClαWT/βWT receptor, we used an allosteric model based on the Monod-Wyman-Changeux (MWC) theory (47), as applied also by Karlin (48) to the nicotinic ACh receptor (nAChR) (reviewed in refs. 49 and 50). Because the GluClαWT/βWT receptor displays very slow and weak desensitization, we simplified the allosteric model by focusing on two major states as previously performed for weakly or nondesensitizing Cys-loop receptors such as: homomeric α7-nAChR mutants (51), homomeric α7-5HT 3A R chimeras (52), and heteromeric GABA receptors (53,54). If the GluClαWT/βWT receptor has two equivalent (identical) Glu-binding sites, then Scheme I describes its MWC allosteric activation mechanism as follows: where R and R* are resting (closed) and active (open) receptor conformational states, respectively; A is an agonist molecule (Glu) that can complex with the receptor; K d,R and K d,R* are the microscopic equilibrium dissociation constants for agonist binding to the closed and open receptor states, respectively; and L is the equilibrium constant of the two receptor states (closed and open) in the absence of ligands. L is calculated by R/R* based on quantitative determinations, as follows.
Unoccupied R* corresponds to spontaneously open channels. Spontaneous activity (I spont ) was measured as the fraction of the leak current that could be blocked by picrotoxin, an ion-channel pore blocker of GluClRs (55) (e.g., Fig. 3F, indicated by "a"; elaborated in SI Text, section 2). Unoccupied R is estimated based on the current elicited by saturating Glu concentrations [maximal current response (I max )]. That is, I max represents the activatable receptor population, which is at rest in the absence of Glu (Fig. 3F, indicated by "b"). However, I max might not represent all of the activatable channels because not all of them are necessarily open at saturating Glu concentrations. Therefore, we determined the maximum open probability (P o-max ) of the ion channel by single-channel recordings at a saturating Glu concentration (Fig. 3 G and H) and then calculated R by I max /P o-max . Thus, L = (I max /P o-max )·(1/I spont ). Experimental P o-max and L values of three receptors are specified in Table S2 (footnotes).
I spont and P o-max (0.64) were also used to normalize the doseresponse data points of the GluClαWT/βWT receptor to estimate its open probability (P open ) at varying Glu concentrations by [(I + I spont )/(I max + I spont )]·P o-max (Fig. 3I). Then, to assess the applicability of Scheme I to the WT receptor activation mode, a curve was fitted to the normalized data points using an MWC allosteric model with two equivalent Glu-binding sites (n = 2) and the experimental mean L value (85) (Fig. 3I, dashed black curve and Eq. 2). Table S2 provides the resulting K d values (in bold). At very high Glu concentrations, the theoretical maximum open probability P o-max * = 1/(1 + c n L), where c = K d,R* /K d,R (54). So, when n = 2, the theoretical P o-max * = 0.65 for the GluClαWT/βWT receptor, which closely predicts the experimental P o-max (0.64). In contrast, fitting curves using an MWC model with other n values (one or equivalent three, four, or five Glu-binding sites; Eq. 2) resulted in a theoretical P o-max * ≥ 0.68 (Table S2). Moreover, analysis of the second-order Akaike information criterion difference (ΔAICc) (56) (SI Materials and Methods) selected the allosteric model with n = 2 as the most suitable MWC model for curve fitting in the GluClαWT/βWT receptor case (Table S2). Hence, we infer that the GluClαWT/ βWT receptor has two functional equivalent Glu-binding sites. Taken together with the results shown in Fig. 3 A-C, we suggest that these two Glu-binding sites likely lie at two β(+)/α(−) intersubunit interfaces (Fig. 3J). Although one cannot absolutely exclude the possibility of a change in subunit stoichiometry due to mutations, we argue that such a change is unlikely to occur here (SI Text, section 3, Fig. S5, and Table S3).

Mutation in the IVM-Binding Pocket Gives Rise to a Third Glu-Binding
Site. During our research, we identified a mutation in the putative IVM-binding site (αL279W; position α(−)L218 in GluClα cryst R) that decreased the Glu-EC 50 of the GluClαL279W/βWT receptor by ∼25-fold, compared with the GluClαWT/βWT receptor   Table S1). This mutation increased the Hill coefficient to 2.6, suggesting that the number of occupiable Glubinding sites in the receptor mutant is probably not less than three. Intrigued by this possibility, we initially examined an MWC allosteric model with either two or three equivalent Glu-binding sites. To this end, we determined the values of I spont , I max , P o-max , and L for the GluClαL279W/βWT receptor [ Fig. 4 C and E and Table S2 (footnotes)] and estimated its P open at varying Glu concentrations, all as described above for the GluClαWT/βWT receptor. Then, a curve was fitted to the normalized dose-response data points using an MWC allosteric model with n = 2 and the experimental mean L value (81) (Fig. 4G, salmon-colored curve and Eq. 2). The resulting K d values (Table S2, same line of "2, 0") were applied to calculate the theoretical P o-max * by 1/(1 + c 2 L) = 0.98, which turned out to be much higher than the experimental P o-max (0.86). Extrapolating the salmon-colored curve in Fig. 4G (model with n = 2) until the theoretical P o-max * is reached indicates a strong deviation of this curve from the Hill plot at high Glu concentrations. Alternatively, a curve was fitted to the normalized doseresponse data points using an MWC allosteric model with three equivalent Glu-binding sites (n = 3) and the same L value (81) (Fig. 4G, cyan-colored curve and Eq. 2). The resulting K d values ( Table S2, same line of "3, 0") were used to calculate the theoretical P o-max * by 1/(1 + c 3 L) = 0.96, which is also much higher than the experimental P o-max (0.86). Extrapolation of the cyancolored curve in Fig. 4G (model with n = 3) until the theoretical P o-max * is reached indicates a strong deviation of this curve from the Hill plot at high Glu concentrations. Curve fitting with other values for n (one, or equivalent four or five Glu-binding sites) resulted in a theoretical P o-max * ≥ 0.95 (Table S2). We therefore applied an MWC allosteric model with two equivalent and a third distinct Glu-binding sites (n = 2, m = 1), using the same L value (81) (Fig. 4G,  where c = K d,R* /K d,R and c′ = K′ d,R* /K′ d,R. In this case, the theoretical P o-max * = 1/(1 + c n c′ m L) = 1/(1 + c 2 c′ 1 L) = 0.89, which is much closer to the experimental P o-max (0.86) than in cases of curve fitting with other numbers of Glu-binding sites ( Table S2). Analysis of the ΔAICc selected the allosteric model with two equivalent and a third distinct Glu-binding sites (n = 2, m = 1) as the most appropriate MWC model for curve fitting in the GluClαL279W/βWT receptor case (Table S2).
The allosteric mechanism suggested above does not provide details regarding the subunit types that form the third Glubinding site interface in the GluClαL279W/βWT receptor, however. If the fifth subunit is GluClβ, then it will give rise to α(+)/β(−) and β(+)/β(−) intersubunit interfaces (envisioned in Fig. 3J); however, based on the aforementioned results, the GluClβ(−) side is less likely to contribute to Glu binding. If the fifth subunit is GluClα, then it will give rise to α(+)/α(−) and α(+)/β(−) intersubunit interfaces (envisioned in Fig. 4G, Right); so, the α(+)/α(−) intersubunit interface remains a reasonable candidate to form the third Glu-binding site. However, this working hypothesis required further experimental investigation. Because the GluClα(−) side was inferred to line the two Glubinding pockets (Fig. 3 and main text), we introduced an α(+)T258N mutation (in loop C), in addition to the αL279W mutation. The homologous mutation [β(+)T229N] in the GluClαWT/βT229N receptor was shown to increase the Glu-EC 50 by approximately eightfold, compared with the GluClαWT/βWT receptor (Fig. 3C and Table S1; presented again in Fig. 4A in gray for convenience). Hence, an α(+)T258N mutation was anticipated to affect a potential α(+)/α(−) intersubunit Glu-binding site, without directly interfering with Glu binding at the two β(+)/α(−) sites. Fig. 4A shows that the dose-response curve of the GluClα (L279W,T258N)/βWT receptor is significantly shifted to the right relative to the curve of the GluClαL279W/βWT receptor, with an ∼57-fold increase in the Glu-EC 50 and a decrease of the Hill coefficient to n H = 1.6 (Table S1). These macroscopic properties resemble the properties displayed by the GluClαWT/ βWT receptor, which has two equivalent Glu-binding sites.
To quantify the effect of the α(+)T258N mutation further, we determined the values of I spont , I max , P o-max , and L for the GluClα (L279W,T258N)/βWT receptor [ Fig. 4 D and F and Table S2 (footnotes)] and estimated its P open at varying Glu concentrations, all as described above for the GluClαWT/βWT receptor. Then, a curve was fitted to the normalized dose-response data points using an MWC allosteric model with two equivalent Glubinding sites (n = 2) and the experimental mean L value (203) (Fig. 4H, dashed black curve and Eq. 2). The resulting K d values are provided in Table S2 (in bold). The theoretical and experimental maximum open probabilities were found to be equal (0.60), whereas other values for n (one, or equivalent three, four, or five Glu-binding sites) resulted in higher theoretical P o-max * values (Table S2). In addition, the analysis of the ΔAICc selected the allosteric model with n = 2 as the most suitable MWC model for curve fitting in the GluClα(L279W,T258N)/βWT receptor case (Table S2). Hence, the results imply that this double-mutant receptor lost the third Glu-binding site, and its remaining two equivalent Glu-binding sites display slightly lower affinity for Glu than the GluClαWT/βWT receptor [Table S2 (in bold)]. Provided that the mutations have not changed the subunit stoichiometry (as argued in SI Text, section 3), the two Glu-binding sites of the GluClα(L279W,T258N)/βWT receptor likely lie at β(+)/α(−) intersubunit interfaces (Fig. 4H, Right). As discussed above, the GluClβ(−) side is less likely to contribute to Glu binding, and so is an α(+)/β(−) intersubunit interface. We therefore infer that the α(+)T258N mutation is likely located at an α(+)/α(−) intersubunit interface. Taken together, our results suggest that in the GluClαL279W/βWT receptor, an α(+)/α(−) intersubunit interface likely forms a third Glu-binding site (Fig. 4G, Right), whereas Glu binding to this interface is impaired by adding the α(+)T258N mutation (Fig. 4H, Right). β(+)T229N is the homologous mutation of α(+)T258N. Combining the αL279W mutation with the β(+)T229N mutation, to give a GluClαL279W/βT229N receptor, led to a fivefold rightward shift of the dose-response curve relative to the GluClαL279W/ βWT receptor (Fig. 4A and Table S1). This shift is much smaller than the 57-fold rightward shift observed in the GluClα(L279W, T258N)/βWT receptor relative to the GluClαL279W/βWT receptor ( Fig. 4A and Table S1). This difference is in line with the above conclusion that an α(+)/α(−) intersubunit interface forms the third Glu-binding site in the GluClαL279W/βWT receptor.
Effect of the αL279W Mutation on the Responsiveness of the Heteromeric GluClαL279W/βWT Receptor to IVM. The crystal structure of the homomeric GluClα cryst R indicates that the backbone carbonyl oxygen of αL279 (L218 in GluClα cryst R) forms a hydrogen bond with hydroxyl O13-H of IVM, whereas the αL279 side chain does not interact with IVM (23) (Fig. 4B, Left). Three-dimensional homology modeling predicts that a Trp side chain introduced at position α279 might form multiple contacts with IVM (Fig. 4B,  Right). If so, how might this mutation affect the responsiveness of the GluClαL279W/βWT receptor to IVM? To answer this question, we had to determine the IVM EC 50 for the WT and mutant receptors. However, unlike the fully reversible responses to Glu, after activation by IVM, the response could not be reproduced by reapplication of IVM even when the first IVM application was followed by a long-term wash (up to 30 min). Other groups also observed this phenomenon when the wash was applied for several minutes (13) or an hour (18). Hence, to quantify the effect of the mutation, we first used the methodology of Lester and coworkers (18) to determine the time constant of conductance development following IVM application. To this end, voltage ramps were carried out during the application of various IVM concentrations, with each application in a different cell. Fig. 5A shows an example of such an experiment. Superimposition of the output currents of the voltage ramps shows a sharp increase in slopes that reflects the robust IVM-induced conductance and a clear leftward shift (decrease) of the reversal potential that occurs mainly after the conductance reached its maximum (Fig. 5B). The shift of the reversal potential indicates a change in the Nernst potential for Cl − and in the electrochemical driving force acting on Cl − ions. The chloride conductance is defined by the slope of the current-voltage (I/V) relations extracted from the output currents of the voltage ramps, and could be determined at several membrane voltage spans. Fig. S6A shows the slope conductance determined between −75 mV and −65 mV, around the reversal potential, and between +10 mV and +20 mV, as a function of time. The rise time of the conductance increment was found to be similar for all of the three aforementioned voltage spans (Fig. S6A). Notably, during the applications of high IVM concentrations, the conductance rise was followed by a decrease in the conductance to a steady state in all voltage spans and in both the WT and indicated mutant receptors (Fig. S6A). Because the current decay under high IVM concentrations was faster at −65 mV than at +20 mV (Fig. S6B), and because the exponential fits of the conductance rise time were very similar at the different membrane voltage spans, we chose to analyze the conductance development further between +10 mV and +20 mV. Fig. 5 C and D shows the development of the conductance under the application of different IVM concentrations in different representative cells.
The exponential fits of the conductance rise time (e.g., Fig. 5 C and D, orange curves) provide the time constant of conductance development (τ), whose reciprocal (1/τ) increased linearly with the increase in IVM concentration (Fig. 5E). Because IVM does not readily dissociate from the receptor (13,18) and the number of possible intermediate IVM-bound closed states is not known, the simplest possible kinetic model that could describe the activation mechanism by IVM would be one in which the channel opens when IVM binds and closes after a very long time when IVM dissociates. Scheme III describes this kinetic model: where R is the unoccupied closed receptor, IVM·R* is the IVMbound open receptor, and 1/τ = k f [IVM] + k b . The slope of the curves in Fig. 5E corresponds to the IVM association rate constant (k forward, k f ). The IVM dissociation rate constant (k backward, k b ) is the extrapolated intercept of the linear curve with the y axis in Fig. 5E. The apparent K d would be k b /k f , giving 73 × 10 −9 M for IVM binding to the GluClαWT/βWT receptor (k b = 5.3 × 10 −2 s −1 and k f = 7.3 × 10 5 s −1 ·M −1 ). In contrast, the apparent K d for IVM binding to the GluClαL279W/βWT receptor was 9.7 × 10 −9 M (k b = 3.4 × 10 −2 s −1 and k f = 3.5 × 10 6 s −1 ·M −1 ), which indicates that the affinity of the mutant receptor for IVM is 7.5-fold higher than the affinity of the WT receptor for IVM. Note that because no experiments revealed that IVM could be washed out of the receptor (13,18), the k b values are expected to be on the order of <10 −4 s −1 . However, the values here were found to be on the order of 10 −2 s −1 , implying that IVM should be removable. We therefore cannot exclude the possibility that after opening of the GluClR ion channel by IVM, a subsequent conformational change leads to trapping of IVM between the transmembrane helices irreversibly.

Discussion
To determine unequivocally the subunit stoichiometry and arrangement in native GluClα/βRs, high-resolution X-ray crystallography of heteropentameric receptors purified from the organisms that naturally express them is necessary. To the best of our knowledge, such a determination is yet out of reach. Hence, an alternative methodology must be considered. In Cys-loop receptors, the neurotransmitter-binding pockets lie at the interface between adjacent subunits (1-9). One could therefore use site-specific mutagenesis and biophysical characterization of acti-vation mechanisms in recombinant receptors to find the types of subunits that line the agonist-binding pockets. By working with recombinant receptors, however, one cannot exclude the possibility that the ratio of subunit cDNA transfected, the type of the expressing cell, or a mutation might influence the receptor's subunit composition (e.g., 45,57,58). We nevertheless argue that the specific mutations we introduced are less likely to change the subunit stoichiometry of the recombinant receptors studied here (SI Text, section 3).
In various Cys-loop receptors, the β1β2, Cys, and β8β9 loops were shown to play a key role in transducing the agonist-binding energy into ion-channel gating force (35)(36)(37)(38)(39)(40)(41)(42)(43)(44). Here, we first demonstrated that although the homomeric GluClαR is not responsive to Glu, the β1β2, Cys, and β8β9 loops of the GluClα subunit are fully capable of coupling Glu binding to channel gating in a heteromeric GluClα/β microchimera that has the sequences of the α-subunit loops. Subsequently, we undertook to identify the intersubunit interfaces involved in Glu accommodation by heteromeric GluClα/βRs. Taking advantage of the crystal structure of a truncated homomeric GluClα cryst R as a template, we built a 3D homology model for the GluClβ subunit. Then, based on the two structures, we introduced single-site mutations in the (−) side of either the GluClα subunit or the GluClβ subunit at positions carrying residues that putatively interact with Glu. Characterization of the effects of these mutations on the receptor function allowed us to suggest that in the heteromeric GluClα/ βRs studied here, the (−) side of the α subunit, rather than the (−) side of the β subunit, contributes complementary components to Glu binding. Single-site mutations and functional analysis of heteromeric GluClα/βRs carrying mutations in the (+) side of the β subunit imply that this side contributes principal components to Glu binding.
When considering the GluClαWT/βWT receptor in terms of the MWC allosteric mechanism, we infer that a maximum of two equivalent binding sites can be occupied by Glu [Fig. 3I, Scheme I, and Table S2 (in bold)]. Provided that the aforementioned singlesite mutations introduced at the intersubunit interfaces have not changed the subunit stoichiometry (as argued in SI Text, section 3), Glu binding likely takes place at two β(+)/α(−) intersubunit interfaces. Hence, one can envision a subunit arrangement as illustrated in Fig. 3J for a recombinant GluClαWT/βWT receptor expressed in CHO cells, with no information regarding the type of the fifth subunit.
When considering the GluClαL279W/βWT receptor in terms of the MWC allosteric mechanism, we infer that Glu can occupy three sites ( Fig. 4G and Scheme II). These sites are (i) two equivalent Glu-binding sites that are likely located at β(+)/α(−) intersubunit interfaces and display considerably higher affinity for Glu than their homologous binding sites in the GluClαWT/βWT receptor and (ii) a third distinct site with slightly lower Glu-binding affinity, in both the resting (closed) and active (open) receptor states [Table S2 (in bold)]. We argue that the third Glu-binding site is formed between two adjacent α subunits; the arguments for that conclusion are as follows: i) In CHO cells, a WT GluClβ subunit does not assemble into a homopentamer capable of responding to Glu or IVM, which indicates that the β subunit has difficulties in creating Glu-binding β(+)/β(−) intersubunit interfaces ( Fig. S2A and Table S1). ii) In the heteromeric GluClα/βRs studied here, three singlesite mutations in the β(−) side did not lead to drastic effects on the receptor activation by Glu, unlike the case of the same mutations introduced at the homologous positions in the α(−) side. iii) The homomeric GluClαL279W receptor responds to very high Glu concentrations (Fig. S7), indicating the capability of an α(+)/α(−) intersubunit interface to accommodate Glu (with no need for IVM prebinding).
iv) Adding a mutation in loop C (+ side) of the αL279W subunit gave rise to an α(L279W,T258N)/βWT receptor that lost the third Glu-binding site (Fig. 4H), whereas the remaining two equivalent Glu-binding sites display microscopic equilibrium dissociation constants slightly higher than the microscopic equilibrium dissociation constants of the GluClαWT/βWT receptor [ A third Glu-binding site located at an α(+)/α(−) intersubunit interface requires that the fifth subunit would be a GluClα subunit. We therefore suggest that the subunits of the recombinant heteromeric GluClα/βRs studied here assemble in an anticlockwise β-α-β-α-α fashion, as viewed from the extracellular side ( Fig.  4 G and H, Right). Notably, previous studies show that expressing the heteromeric α4β2 nAChR under conditions that favor an (α4β2) 2 α4 stoichiometry (three α4 and two β2 subunits) results in a receptor having two α4(+)/β2(−) interfaces with high agonist sensitivity and a third binding site at the α4(+)/α4(−) interface that displays low agonist sensitivity (59)(60)(61).
As discussed in Results, the function of heteromeric receptors containing the αL279W mutation, together with a Thr→Asn substitution in loop C of the α subunit, β subunit, or both subunits, suggests that the two intersubunit interface types, α(+)/α(−) and β(+)/α(−), likely affect each other allosterically. Possible structural reasons for this mutual allosteric influence are provided in SI Text, section 4. Interestingly, an allosteric relationship between different extracellular intersubunit interfaces was proposed for the heteromeric α1β2γ2 GABA A receptor (62). In the latter case, conformational movements induced by benzodiazepine binding at the α/γ extracellular interface were suggested to propagate across the α1 subunit to the β/α GABA-binding site interface (62).
In the GluClα cryst R, L218 (αL279 in the full-length subunit used here) is part of the IVM-binding pocket located between M1 and M3 of adjacent subunits (23) (Fig. 4B, Left). The clear increase in the affinity of the GluClαL279W/βWT receptor for IVM ( Fig. 5E; 7.5-fold) implies that the IVM-binding pockets of the heteromeric receptor are homologous to the IVM-binding pockets of the homomeric GluClα cryst R. The structural mechanism underlying the effect of the αL279W mutation in the IVMbinding site is not clear. However, the microscopic equilibrium dissociation constants for Glu binding determined here imply that the conformational change induced by this mutation in the IVM-binding pocket propagates to the Glu-binding pockets and affects their affinity for Glu. It is not known whether Glu and IVM induce the same conformational change in the coupling loops. In the heteromeric α1β2γ2 GABA A receptor, for example, it was demonstrated that positive benzodiazepine modulators induce movements in loop F (β8β9 loop) of the γ2 subunit near the transmembrane channel domain (63). Such movements were not triggered by the binding of GABA, the allosteric modulator pentobarbital, or the inverse agonist methyl-6,7-dimethoxy-4-ethylβ-carboline-3-carboxylate (63).
In conclusion, our study provides evidence that the C. elegans heteromeric GluClR contains three α subunits and two β subunits arranged in an anticlockwise β-α-β-α-α fashion, as viewed from the extracellular side, with two Glu-binding sites located at the β(+)/α(−) intersubunit interfaces. The α(+)/α(−) intersubunit interface creates a third "dormant" Glu-binding site that becomes functional upon a conformational change induced by a mutation in the IVM-binding pocket.

Materials and Methods
Additional experimental procedures and data analyses are described in SI Materials and Methods.
Data analysis and mathematical modeling were performed using the Clampfit 10 program implemented in pClamp 10, and GraphPad Prism software.
Dose-response curves were fitted to the data points by a nonlinear regression using the Hill equation (Eq. 1): where I is the current response, I max is the maximal current response, EC 50 is the agonist effective concentration that elicits 50% of the maximal current response, [Glu] is the concentration of Glu, and n H is the Hill coefficient. For the allosteric modeling, Eq. 2 was used: where P open is the open probability estimated at varying Glu concentrations (54) (main text).
[Glu] is the concentration of the agonist (Glu) for which there are n equivalent binding sites, each with a microscopic equilibrium dissociation constant of K d,R in the resting (closed) state and K d,R* in the active (open) state. L is the equilibrium constant of the two states in the absence of ligands. The L values were determined by functional experiments, as described in the main text. For a receptor phenotype that does not behave as a receptor having only n equivalent Glu-binding sites, Eq. 3 [cf. Karlin (48) where m is the number of sites that Glu binds with microscopic equilibrium dissociation constants, K′ d,R in the closed state and K′ d,R* in the open state.