Disruption of an intersubunit electrostatic bond is a critical step in glycine receptor activation

Abstract

Proper regulation of neurotransmission requires that ligand-activated ion channels remain closed until agonist binds. How channels then open remains poorly understood. Glycine receptor (GlyR) gating is initiated by agonist binding at interfaces between adjacent subunits in the extracellular domain. Aspartate-97, located at the α1 GlyR interface, is a conserved residue in the cys-loop receptor superfamily. The mutation of D97 to arginine (D97R) causes spontaneous channel opening, with open and closed dwell times similar to those of maximally activated WT GlyR. Using a model of the N-terminal domain of the α1 GlyR, we hypothesized that an arginine-119 residue was forming intersubunit electrostatic bonds with D97. The D97R/R119E charge reversal restored this interaction, stabilizing channels in their closed states. Cysteine substitution shows that this link occurs between adjacent subunits. This intersubunit electrostatic interaction among GlyR subunits thus contributes to the stabilization of the closed channel state, and its disruption represents a critical step in GlyR activation.

Glycine receptors (GlyR) are pentameric anion-conducting members of the cys-loop receptor superfamily, with their subunits arranged around a central ion pore. Each subunit consists of a large extracellular N-terminal ligand binding domain and four transmembrane segments (TM1–TM4); TM2 of each subunit lines the pore (1). When glycine binds to initiate channel opening it interacts with specific amino acids located at intersubunit interfaces, and spontaneous openings do not occur in the absence of neurotransmitter (2). Six loops of amino acids located on adjacent subunits constitute the known glycine binding site. On the plus (+) side of the interface on one subunit are loops A–C, whereas loops D–F are located on the minus (−) side of an adjacent subunit (3). In the related nicotinic acetylcholine receptor (nAChR), signal transduction after agonist binding is described as a “Brownian conformational wave” that travels down the interface between subunits (4). Auerbach and colleagues used ϕ analysis to show that the binding pocket region of the N-terminal domain is the first to move after ligand binds. Loops 2 and 7 (the cys-loop) of the N-terminal domain interact with the extracellular end of TM1 and the TM2–3 linker region to transmit binding signals to the channel gate (5–7). In the α1 GlyR subunit, D148 in loop 7 forms an electrostatic bridge with K276 in the TM2–3 linker (8).

Our previous work demonstrated that mutation of D97 in loop A results in spontaneous channel opening (9). This D97 residue is conserved within the cys-loop receptor superfamily (Fig. 1), suggesting its critical role in channel function. In this article we report on an electrostatic interaction between specific charged amino acid residues at the interfaces of adjacent subunits that contribute to the stabilization of the closed channel state in the absence of neurotransmitter. Disruption of these electrostatic bonds represents a critical step in GlyR activation.

Fig. 1.

Partial amino acid sequence alignment of selected members of the nAChR receptor subunit superfamily. The human GlyR α1 subunit sequence from residues 89–123 was compared with the same regions of selected human GABA_A, nACh, and 5-HT3 receptor subunits, as well as the Lymnaea stagnalis acetylcholine binding protein. Residues equivalent to GlyR α1 W94, P96, and D97 are invariably conserved in all members of the nAChR receptor superfamily. The amino acid residues equivalent to GlyR α1 D97, K116, and R119 are highlighted in bold.

Results

Charge Swap Mutations Illustrate Interactions Between D97 and K116/R119.

Mutating aspartate-97 to arginine (D97R) in homomeric α1 GlyR produces receptors that exhibit increased background inward currents in the absence of glycine. This tonic current, not seen in WT α1 GlyR, is blocked by the competitive GlyR antagonist strychnine (Fig. 2A and Fig. S1). The mutation also increases the glycine EC₅₀ (Table 1), but this is likely a consequence of a decrease in the Hill slope in the D97R mutant, because the threshold glycine concentrations required to observe currents in WT and D97R receptors are quite similar. When this residue is mutated in the β2 subunit of α1β2 GABA_A receptors the GABA EC₅₀ also increases (10). Two other mutations, α1 K116D and α1 R119E, each resulted in receptors with decreased glycine sensitivities (Table 1) but did not exhibit spontaneously opening channels. The charge reversal mutant α1 D97R/K116D bearing both mutations on the same subunit showed a partial α1 D97R phenotype, with reduced holding currents and a smaller effect of strychnine, compared with α1 D97R (Fig. 2 B and D). In contrast, the α1 D97R/R119E charge reversal seemed to stabilize the receptor in the closed state, characterized by WT-like whole-cell holding currents and loss of the strychnine effect (Fig. 2 C and D). Despite the low sensitivity to glycine of the α1 R119E and α1 D97R/R119E mutants, the enhancing effects of a high concentration of glycine (500 mM) could still be antagonized by 1–10 mM strychnine (Fig. 2C). This high concentration of glycine acted specifically via the GlyR because concentrations as high as 1 M applied to uninjected oocytes did not elicit any currents. Consistent with the strychnine findings, the average holding currents when cells were clamped at −70 mV were much higher in oocytes expressing the D97R or D97R/K116D receptors, reflective of considerable spontaneous channel-opening activity (Fig. 2D).

Fig. 2.

Constitutive activity observed in some GlyRα1 mutants. Representative tracings show the responses to maximally effective glycine and strychnine concentrations for a series of α1 GlyR mutants. The lines and arrows over tracings refer to drug applications. Note the two different current scale bars for the glycine and strychnine applications in A–C. (A) Glycine applied for 45 s produced inward currents indicating channel opening in the D97R mutant. In contrast, strychnine applied for 1 min initially produced an outward current indicative of closure of spontaneously active channels, followed by an inward current once strychnine was removed. (B) The double mutant D97R/K116D also displays spontaneous channel-opening activity that can be antagonized by strychnine. (C) The D97R/R119E double mutant shows no evidence of strychnine antagonism of spontaneous activity in whole-cell recordings, although strychnine can block the effects of glycine on these receptors. (D) The holding current required to voltage clamp oocytes at −70 mV was measured and compared among mutants. WT α1 GlyR do not exhibit spontaneous channel opening and typically exhibit holding currents of approximately 100 nA. These low holding currents were also seen in the R119E, K116D, and D97R/R119E mutants. However, D97R and D97R/K116D mutants displayed significantly greater holding currents and differed from one another as measured using Student's t test [t(33) = 3.84, P < 0.001]. Values are reported as mean ± SEM of 8–24 oocytes.

Table 1.

Characterization of glycine concentration–response relationships of WT and mutated α1 GlyR expressed in Xenopus oocytes

α1 GlyR mutation	Response to glycine (EC50)	Hill coefficient	Possible electrostatic bonds
None (WT)	0.2 mM (n = 6)	1.4
D97R	1.9 mM (n = 3–8)	0.41
K116D	3.0 mM (n = 6)	1.01
R119E	87.9 mM (n = 6)	1.30
D97R/K116D	1.3 mM (n = 10)	0.65
D97R/R119E	390 mM (n = 6)	3.0

For each receptor, glycine concentration–response curves were generated, and the glycine EC₅₀s and Hill coefficients determined. The possible intersubunit electrostatic interactions among the amino acids studied (97, 116, and 119) are shown for each GlyR.

Single-Channel Recordings of D97R and D97R/R119E Mutants.

Outside-out patch-clamp recordings were made of homomeric α1 D97R GlyR and α1 D97R/R119E GlyR in the absence of exogenously applied glycine, as well as WT α1 GlyR exposed to 10 mM glycine (Fig. 3). The spontaneous activity observed in the D97R GlyR seemed quite similar to the effects of a saturating concentration of glycine on WT receptors. The D97R/R119E receptor behaved very differently, with many very brief opening events seen, as shown in the tracing on the left-hand side of Fig. 3C. The two mutant receptors displayed quite different open and closed dwell-time properties. D97R receptor spontaneous activity had a P_open of 0.91, very similar to that produced by glycine on WT GlyR (P_open = 0.9). In contrast, the D97R/R119E mutant had a much lower P_open of 0.14, which is likely an overestimate because multiple channels probably contributed to this measure. Open dwell-time histograms generated from the tracings were fit using two open-time components (τs) for the WT, α1 D97R GlyR, and α1 D97R/R119E GlyR (Fig. 3, Right; τ and amp values are provided as Table S1). Spontaneous openings were much longer in the D97R mutant, suggesting that reestablishing the electrostatic bond in the double D97R/R119E mutant acts to stabilize the closed state of the channel, albeit imperfectly. Supporting this argument, the histograms illustrating the fits of the closed dwell times for α1 D97R GlyR and α1 D97R/R119E GlyR, which were each fit using two closed-time components, indicate an increased prevalence of longer closed times in the double mutant, whereas very short closings predominated in the D97R mutant.

Fig. 3.

Single-channel recordings of D97R and D97R/R119E α1 GlyR mutants. Representative outside-out patch tracings made from (A) WT α1 GlyR exposed to 10 mM glycine, as well as spontaneous activity recorded from (B) D97R α1 GlyR and (C) D97R/R119E α1 GlyR in the absence of exogenously applied glycine. Each of the top lines represents 7 s of recording, and regions with the horizontal lines above them in each trace are expanded below. Downward deflections signify channel opening. The D97R/R119E trace seems to show multiple channels in the patch with very brief spontaneous openings. Open- and closed-time histograms generated from measurements of the WT GlyR in the presence of 10 mM glycine seemed quite similar to those describing the spontaneous activity of the D97R α1 GlyR. Both exhibited longer open times as well as shorter closed times than the D97R/R119E double mutant. Activity of each GlyR was adequately fit using two open-time or closed-time components (τs). Tau and likelihood numbers are provided in Table S1. The two thin lines in each histogram describe the individual components, whereas the thicker line is an overall fit.

Effects of Reduction and Cross-Linking on Cysteine Mutants.

To test our hypothesis that residues D97 and R119 are physically interacting we constructed α1 D97C, R119C, and D97C/R119C mutants to test for possible cysteine cross-linking. After first establishing stable maximally effective (1 M) glycine responses, the reducing agent DTT, the cysteine-bridging agent HgCl₂, or the oxidizing agent iodine were tested for their effects. Applications of 1 M glycine were 3–5 s long, during which a peak response was observed, followed by 12–15-min washout periods. Oocytes were then perfused with 10 mM DTT for 2 min, and after a 12-min washout 1M glycine was reapplied twice. Oocytes then received either a 2-min application of HgCl₂ (10 μM) or a 1-min application of iodine (0.5 mM), which favors disulfide bond formation. Control experiments followed the same experimental procedure but with the application of buffer (MBS) instead of DTT, HgCl₂, or iodine. The WT, D97C, and R119C single mutants showed no significant effects of either DTT or HgCl₂ (Fig. 4 A–C). In contrast, application of DTT to the α1 D97C/R119C double mutant significantly enhanced responses to glycine applied afterward, whereas HgCl₂ decreased the magnitude of glycine effects (Fig. 4 D and E). Reapplication of DTT after HgCl₂ resulted in the same high glycine-mediated currents that were observed after the initial DTT application.

Fig. 4.

Reduction and cross-linking experiments support an interaction between α1 D97C and R119C. (A) There was no effect of the reducing agent DTT on WT α1 GlyR (n = 3). Glycine was applied three times before 10 mM DTT was applied for 2 min, allowed to wash out for 10 min, and then glycine applied three more times. In addition, there were no significant effects of either DTT or HgCl₂ (10 μM for 2 min) on currents elicited by maximally effective glycine concentrations on either (B) D97C (n = 3–7) or (C) R119C (n = 3–11) single mutants. D97C GlyR exhibited average holding currents of 488 ± 64 nA, compared with 93 ± 18 for R119C GlyR and 63 ± 12 for the double mutant D97C/R119C. (D) In contrast, the sample tracing of the double mutant D97C/R119C α1 GlyR shows a significant increase in glycine current after exposure to DTT and a decrease in the glycinergic current after application of HgCl₂. (E) Bar graph illustrating the effects of DTT and HgCl₂ on D97C/R119C α1 GlyR responses. Data are reported as the mean ± SEM of 5–12 oocytes. (F) The oxidizing agent iodine can also decrease the effects of a maximally effective glycine concentration. Data are reported as mean ± SEM of six oocytes.

All three mutants were also tested for possible disulfide bond formation by iodine (Fig. 4F). In both the D97C and R119C mutants, the application of 0.5 mM iodine for 1 min had no effects on glycinergic currents, suggesting that there were no available free cysteine residues close enough to form disulfide bonds. However, the double mutant D97C/R119C was able to form a disulfide bond when in an oxidizing environment because iodine (Fig. 4F) markedly decreased the effects of glycine in a DTT-reversible manner. For this disulfide bond formation to occur, the sulfhydryl groups of cysteine residues can be up to 15 Å (11) apart; once the bound has formed, the distance between α-carbons in the two cysteines is 5.6 Å (12).

Intersubunit Interactions Stabilize the Closed State of the Receptor.

We next determined whether the interaction between D97 and R119 occurs between adjacent subunits. To do this we mixed cDNAs containing two different single mutants, each expressing either the α1 D97C or R119C mutation. The D97C and R119C cDNAs were mixed in a ratio of 1:50, allowing us to predict the percentages of D97C and R119C subunits in receptor populations expressed by oocytes, assuming that the two mutants express and assemble with the same efficiencies. The equation P = 100 × p_D97Cⁿ × p_R119C^{(5 − n)} × 5!/[n!(5 − n)!] describes the percentage of receptors (13) bearing n D97C subunits, where n = 1–5. p_D97C and p_R119C represent the fractions of D97C and R119C cDNAs injected into oocytes; those numbers were 1/51 and 50/51, respectively. Thus the likelihoods of obtaining receptors bearing zero, one, or two D97C subunits in any individual receptor were 90.6%, 9.06%, and 0.34%, respectively. The homomeric α1 R119C single mutant is markedly insensitive to glycine, such that concentrations of glycine below 1 mM elicit no currents. In contrast, coexpression of the D97C and R119C mutants resulted in receptors sensitive to glycine concentrations as low as 10 μM (Fig. 5 A and C). Replacement of C119 with the WT R119 in even one of five subunits in a receptor is sufficient to restore sensitivity to low concentrations of glycine. Coexpression of α1 WT + R119C cDNAs at a 1:50 ratio produced the same results (Fig. 5C). This suggested that the responses obtained using low concentrations of glycine in the α1 D97C+R119C heteromeric GlyR were produced by receptors that bore one copy of the D97C mutation (Fig. 5B). It possesses a single WT D97-R119 intersubunit electrostatic bond and a single C97-C119 covalent bond, as well as three D97-C119 interfaces. Glycine (30 μM) was applied to D97C+R119C receptors to ascertain the effects of HgCl₂ and DTT on cross-linking between subunits (Fig. 5A). Initially these receptors displayed low μA currents in response to 30 μM glycine, which decreased significantly after HgCl₂ exposure. This intersubunit cross-linking was reversed by DTT (Fig. 5 A and D).

Fig. 5.

Disulfide bond formation between adjacent subunits. (A) A concentration of 30 μM glycine does not activate homomeric R119C α1 GlyR. However, in the D97C+R119C (1:50) receptors, in which the two single mutants D97C and R119C were coinjected in a cDNA ratio of 1:50, 30 μM glycine elicited currents. These currents are produced by receptors bearing four R119C subunits and a single D97C subunit as shown in B. A single bridging Hg atom introduced between adjacent D97C and R119C residues is thus sufficient to decrease glycinergic currents, in a manner that can be reversed by reduction using DTT. (C) Glycine concentration response curves were constructed for homomeric D97C and R119C GlyR (hollow symbols) as well as WT+R119C (1+50) and D97C+R119C (1+50) heteromeric receptors (solid symbols), in which the R119C subunit cDNA was coinjected with either the WT or D97C cDNAs but at 50 times the concentration. The two heteromeric GlyR glycine concentration–response curves thus depict the responses of homomeric R119C receptors, which are very insensitive to glycine, as well as a minority of heteromeric WT+R119C or D97C+R119C GlyR that exhibit greater sensitivity to glycine. The 10,000-fold glycine concentration range at which glycine effects are observed is evidence of this receptor heterogeneity. (D) Summary of data comparing the effects of HgCl₂ and DTT on homomeric receptors bearing both the D97C and R119C mutations on each subunit (D97C/R119C) as well as those bearing the two mutations on different subunits, at D97C+R119C ratios of 1:50. Data are shown as mean ± SEM obtained from three oocytes.

Modeling.

Our initial use of the homology model was to predict residues in the vicinity of D97 that could form a salt bridge with it and that would be suitable for charge-reversal mutations. A second use, as described in Methods, was to refine the alignment of the GlyR sequence with that of the template, GLIC (Fig. 6A). The ambiguity in alignment is that GlyR has four more residues in the region of interest than does GLIC. Although many other regions in the total alignment have important conserved residues or prominent secondary structural features (14), the region of interest here does not. Even secondary structure prediction for GlyR in this region does not help because it is an interrupted β strand, designated β5–β5′ [throughout, we use the nomenclature suggested for the nicotinic acetylcholine binding protein by Brejc et al. (15)]. We built models based on three alignments; the four gaps spaced along the β5–β5′ strand of GLIC, the four gaps clustered after β5′ (increased the size of loop 6), and the four gaps clustered before β5 (increased the size of loop 5). Only the last alignment produced a good model because loop 5 could expand without bad overlaps with other residues (Fig. 6B). In contrast, the former two models resulted in conflicts with other regions; in particular the expanded loop 6 intercalated with loop 3. We then used the homology model to observe possible interactions of D97 with K116 and R119 on the opposite subunit interface. Fig. 6B shows that both K116 and R119 are in proximity to D97 and that fluctuations of torsion angles within each side chain produce a variety of possible electrostatic interactions.

Fig. 6.

Homology model of the GlyR intersubunit interface. (A) Partial amino acid sequence alignment of the GlyR α1 (residues 68 through 131) and GLIC sequences. Arrows highlight D97, K116, and R119 in the GlyR α1 sequence. (B) A homology model of GlyR was built as described in Methods. Here two of the five subunits of the homopentamer are shown (side view) with a ribbon structure to highlight the intersubunit interface where D97 is located in proximity to K116 and R119. The (+) and (−) interfaces of adjacent subunits are labeled, as are amino acids D97, K116, and R119. The backbone of the D97 residue is red, whereas the backbones of the K116 and R119 residues are blue. Nitrogen atoms are depicted in light blue, oxygen in red, carbon in gray, and hydrogen in white.

Discussion

It is critical for the proper regulation of neurotransmission that ligand-activated ion channels remain closed until a binding signal is received by the receptor. The receptor complex must exist in conformations that will keep the channel from opening in the absence of agonist while still allowing for rapid conformational changes to occur within microseconds after neurotransmitter binding. The opening of the GlyR pore is initiated by the binding of the glycine molecule at interfaces between loops A, B, and C, on the “+” side of one subunit and β sheet segments D, E, and F on the “−” side of an adjacent subunit (3). Glycine binding is stabilized by its interactions with the side chains of several amino acids in the “+” and “−” sides of adjacent subunits, including R119 (3, 16–18). This residue is also implicated in GABA binding (19). The 2BG9 model of the nAChR (5) and the prokaryotic acetylcholine receptor homolog GLIC (6) provide a structural framework to aid in explaining how ligand binding signals could be transmitted to the pore (5). Electrostatic interactions are thought to occur between residues in extracellular loops 2, 7 (the conserved cys-loop), and 9, with amino acids in the pre-TM1 region, the TM2–3 extracellular loop, and post-TM4 residues (20) linking ligand binding to channel opening. In the GABA_A receptor specific electrostatic interactions between D57 and D149 residues in loops 2 and 7 with K276 in the TM2–3 linker region affect gating (21); later work also implicated a residue in the pre-TM1 region (22). However, in the homomeric α1 GlyR, direct electrostatic interactions between D53 or E57 of loop 2, or D148 of loop 7, with K276 in the TM2–3 linker were not observed (23). Xiu et al. (20) concluded that interactions between extracellular domain amino acids with those in pre-TM1, TM2–3, and post-TM4 do not generally seem to involve specific amino acids, but rather, overall clusters of positively and negatively charged residues mediate interactions between these domains. In a series of publications Auerbach and colleagues studied the relative timing of movements of domains of the nAChR initiated by ligand binding that result in the transitioning of channels from closed to open states. Upon acetylcholine binding the transmitter binding region involving loops A, B, and C moves first (24), loops 2 and 7 then move (25, 26), and this is followed by the almost simultaneous movements of the TM2–3 linker region (26) and TM2 (4, 27). It is believed that a “Brownian conformational wave” involving sequential movements of portions of receptor subunits ultimately links ligand binding to channel opening. Although comparable ϕ analyses have not been conducted in the GlyR, one may hypothesize that a similar process links ligand binding to channel gating. The D97 residue is part of loop A in the α1 subunit and thus, by analogy with the nAChR findings, would be expected to move soon after ligand binding.

We noted that the WxPD motif in which D refers to D97 in the α1 GlyR is invariantly conserved among all members of the cys-loop family, including the snail acetylcholine binding protein (Fig. 1). Mutation of this highly conserved D97 residue to arginine destabilizes the receptor closed state, resulting in channels that exhibit constitutive activity. We sought to determine whether the D97 residue was stabilizing the closed-channel state of the α1 GlyR by forming an electrostatic bond with a positively charged residue and identified R119 on an adjacent subunit as a plausible candidate, on the basis of molecular modeling. Reversing the charge at D97 to arginine results in tonic channel opening that can be markedly reduced by a second charge reversal mutation at R119. Interestingly, the single R119E mutation does not result in spontaneous channel opening, and this may be because the D97 residue can also interact with K116. We hypothesize that the electrostatic bond between D97 and R119 (or K116) breaks after glycine binds, allowing for the uninterrupted propagation of the conformational wave to occur, ultimately leading to the opening of the channel pore. The aspartate residue equivalent to D97 in the nAChR was hypothesized to stabilize loop B via hydrogen bonding (28, 29).

The interaction between D97 and R119 is also illustrated by cross-linking these two residues. Application of DTT to D97C/R119C α1 GlyR results in markedly enhanced glycine-mediated currents, consistent with a breakage of disulfide bonds (Figs. 4 and 5). Application of the oxidizing agent iodine or the cysteine-bridging agent HgCl₂ decreases the magnitudes of glycine-mediated currents, suggesting that linking of D97C and R119C constrains the channels from opening. Importantly, none of these effects are seen with either the D97C or R119C single mutants or in WT receptors. We next addressed the issue of whether the D97 and R119 interactions were occurring between or within α1 subunits. Our mixed cDNA data, involving the coexpression of two α1 GlyR mutants each bearing either a D97C or R119C mutation, show that reduction and cross-linking effects are conserved despite the fact that no single subunit contains two cysteine mutations. When D97C and R119C cDNAs are injected in a 1:50 ratio, we would expect approximately 9% of the resulting receptors to consist of a single α1 D97C subunit and four R119C subunits (Fig. 5B). These receptors would possess three distinct intersubunit interfaces: one C97–C119 interface capable of forming a disulfide bond, a single WT D97–R119 interface, and three D97–C119 interfaces that would not interact either covalently or electrostatically. The D97C+R119C receptors become dramatically more sensitive to glycine compared with R119C homomeric mutants; this suggests that the single WT-like D97:R119 interface is able to increase glycine sensitivity. After application of HgCl₂, a single covalent link formed between D97C and R119C on adjacent subunits is sufficient to limit the abilities of these subunits to move relative to one another, and this results in a significant decrease in receptor activation (Fig. 5D). The breakage of this disulfide bond by DTT can reverse this. The responses we observe after injecting the 1:50 D97C+R119C cDNA ratio are due to heteromeric receptors rather than some combination of homomeric D97C and homomeric R119C for the following reasons: (i) no response to 30 μM glycine is seen in R119C homomeric receptors, although D97C homomeric GlyR are still sensitive to this concentration of glycine; (ii) D97C homomeric receptors are tonically open, whereas the D97C+R119C receptors are not; and (iii) if the data we obtained in response to 30 μM glycine after injecting the D97C+R119C subunits was somehow due solely to D97C homomers, then we would have observed DTT and HgCl₂ responses identical to those shown in Fig. 4B. Instead, the responses of homomeric D97C (as well as R119C) and heteromeric D97C+R119C receptors to application of DTT and HgCl₂ differ markedly.

Our single-channel recordings on the charge reversal mutant D97R/R119E illustrate that restoration of the electrostatic bond in this double mutant prevents the receptor from opening to the same extent as D97R in the absence of ligand. In fact the spontaneously opening D97R mutant seems to behave very much like a WT GlyR exposed to a saturating concentration of glycine. The D97R receptor has a P_open of 0.91 and, like the WT GlyR, openings are grouped into long clusters containing very brief closing events. In both, GlyR clusters seem to be terminated by entries into longer-lived desensitized states. Histograms generated of the open and closed dwell times are also very similar between spontaneously opening D97R and fully activated WT GlyR. Spontaneous channel opening was also reported by Miller et al. (30), who studied the nearby F99A mutant. This mutation produced spontaneously opening channels due to a movement in loop A, and we wonder whether this mutation may also produce a weakening of the electrostatic bond between D97 and R119.

By restoring the attractive force between these residues in the D97R/R119E GlyR we were able to increase the closed state stability of the double mutant (Fig. 3). One logical inference that might be made from our findings is that the brief intraburst closings seen in maximally activated WT α1 GlyR do not involve the temporary restorations of electrostatic bonds such as the one between the D97 and R119 residues. Very brief closings were also prevalent in the D97R mutant, presumably because of the inability of the channel to reestablish salt bridges that would stabilize longer-lived closed states. Longer-lived closed states seen in WT GlyR at lower glycine concentrations might, however, involve restoration of these electrostatic bonds. Interactions other than D97/R119 within this region do occur, such as the hydrophobic interactions noted by Miller et al. (30), and we also see this in our partial phenotype mutant D97R/K116D, where the tonic activity is reduced but not decreased to the same extent as in the D97R/R119E mutant. We hypothesize that this intersubunit region of charges may be implicated in agonist binding, where the interaction between charged residues on adjacent subunits regulates closed to open state kinetics after agonist binds within the binding pocket. The dramatic reduction of receptor sensitivity to glycine in all R119 mutants lends support to this idea. The competing electrostatic interactions shown in Fig. 6B may help explain why the D97R mutation was so deleterious, whereas the K116D and R119E mutations were better tolerated from the view of spontaneous opening. This compensating interaction is similar to what was observed between loops 2, 7, and the TM2–3 linker in the GABA_AR (21). In fact, Xiu et al. (20) suggested that such redundant electrostatic interactions represent a common motif for transduction of gating energy. In conclusion, our study demonstrates the importance of intersubunit electrostatic coupling for normal receptor activation, and that disruption of this electrostatic bond may represent a critical step in GlyR activation.

Methods

Xenopus leavis Oocyte Isolation, cDNA Injection, and Whole-Cell Electrophysiology.

The human α1 GlyR subunit cDNA used in these studies is described in Mihic et al. (31). Oocytes were surgically removed and injected with cDNAs (0.03–3 ng/30 nL) and characterized electrophysiologically as described previously (32). Additional details are provided in SI Methods.

Patch Clamp Electrophysiology, Acquisition, and Analysis.

Outside-out patch recordings were made according to standard methods (33) and are described by Welsh et al. (34). Single-channel data were acquired and analyzed as described previously (34) using the single-channel analysis programs in QuB (35, 36) (version 1.4.0.125) and are described in more detail in SI Methods.

Modeling.

A homology model of GlyR α1 was built by threading the GlyR primary sequence onto an x-ray crystal structure template using the Modeler module of Discovery Studio as previously described (37). The template we used was the prokaryotic ligand-gated ion channel homolog GLIC (PDB ID 3EAM) (6), in part because the x-ray structure of GLIC has higher resolution than the previous cryoelectron microscopy structure of torpedo nAChR (Protein Data Bank ID 2BG9). This is described in more detail in SI Methods.