Mapping a Molecular Link between Allosteric Inhibition and Activation of the Glycine Receptor

Abstract

Cys-loop ligand-gated ion channels mediate rapid neurotransmission throughout the central nervous system. They possess agonist recognition sites and allosteric sites where modulators regulate ion channel function. Using strychnine-sensitive glycine receptors, we identified a scaffold of hydrophobic residues enabling allosteric communication between glycine-agonist binding loops A and D, and the Zn²⁺ inhibition site. Mutating these hydrophobic residues disrupted Zn²⁺ inhibition, generating novel Zn²⁺ activated receptors and spontaneous channel activity. Homology modelling and electrophysiology revealed that these phenomena are caused by disruption to three residues on the ‘–’ loop face of the Zn²⁺ inhibition site, and to D84 and D86, on a neighbouring β3 strand, forming a Zn²⁺ activation site. We provide a new view for the activation of a Cys-loop receptor where, following agonist binding, the hydrophobic core and interfacial loops reorganise in a concerted fashion to induce downstream gating.

The Cys-loop ligand-gated ion channel superfamily includes nicotinic acetylcholine (nAChR), γ-aminobutyric acid type A (GABA_A), glycine (GlyR) and serotonin type 3 (5HT₃) receptors. Each subunit of these pentamers contains: a ligand-binding extracellular domain (ECD), formed by a sandwich of two β-sheets; a four α-helical membrane-spanning domain; and an intracellular region of unspecified quaternary structure ¹. The interior of the ECD is hydrophobic^2,3, and, as for most globular proteins, it is considered to be an entropic stabiliser of protein folding ⁴. Given the presumed stability of this hydrophobic core and its location between two sheets of rigid β-strands, it is usually regarded as a relatively inflexible structure. Thus, after agonist-induced activation, the core would move, if at all, as a rigid body ⁵. Accordingly, it would be the agonist-binding loops A-F, supported by the surrounding rigid β-strands, that would undergo a conformational change upon agonist binding to trigger rigid body movement and downstream channel opening^{2, 6, 7, 7–11}. An alternative view, based on structural and modelling data, suggests that substantial portions of the inner and outer β-sheets of the ECD shift their orientation relative to one another upon receptor activation^{1, 12}. Given the location of the hydrophobic core between the inner and outer β-sheets of each ECD it would then be expected that the core would reorganise, rather than move as a rigid body, to facilitate the reorientation of the β-sheets ^13–17. This movement may induce separation of important charge interactions along neighbouring receptor subunit interfaces, allowing the ECDs to twist and induce downstream channel opening^{8, 16, 18, 19}.

An ideal model system to investigate the role of the hydrophobic core in Cys-loop receptor activation is that involving Zn²⁺ inhibition of the GlyR. These receptors readily form homomers that are modulated by the physiological cation Zn²⁺ in a biphasic fashion. Zn²⁺ can be found in nanomolar concentrations in external medium and is also packaged into vesicles and released at synapses in sufficient amounts to endogenously modulate GlyRs, with low micromolar concentrations potentiating submaximal glycine responses, and higher concentrations causing inhibition ^20–22. Two Zn²⁺ binding sites have been identified; the potentiation site is contained solely on the outer β sheet of the ECD ²³, whereas the inhibition site spans neighbouring subunits on the inner β sheet of the ECDs ^24,25,26 (Fig. 1a-c). Inhibition by Zn²⁺ of GlyR function involves the stabilisation of charge interactions between neighbouring subunit ECD interfaces, thereby hindering their movement. This supports the notion that charge separation of neighbouring ECD interfaces is necessary for receptor activation, and that agonist binding must transduce a signal near to the Zn²⁺ inhibition site to evoke a conformational change in this area leading to receptor activation. As the hydrophobic core is located between the glycine binding site and the Zn²⁺ inhibition site, identifying the molecular requirements for Zn²⁺ inhibition will elucidate the roles of the hydrophobic core and the subunit ECD interface in receptor function.

Figure 1. Hydrophobic determinants of Zn²⁺ inhibition.

(a) Partial protein alignment of Cys–loop receptor ECDs showing: GlyR α1 hydrophobic residues that are examined (red); where hydrophobicity is retained in other Cys–loop receptors (red); conserved residues (boxed); and the GlyR Zn²⁺ inhibition binding site residues (underlined). Hydrophobic motifs are labelled and colour–coded in accordance with (b, c), the GlyR α1 ECD homology model based on TornAChRα₁ ¹. Clustered hydrophobic residues (orange) connect the Zn²⁺ inhibition site at the inner–subunit interface (black and light blue motifs – note, black motif is not a β–strand in the model but referred to as such in main text for clarity ²), to glycine binding loops A (green), D (dark blue) and E (mauve) of outer–subunit interface. Zn²⁺ inhibition site residues, His¹⁰⁷, His¹⁰⁹, Thr¹¹² and Thr¹³³, are in grey (nitrogen atoms are blue and oxygens are red). Dotted line depicts the interface with (+) and (–) subunit sides ²⁸. Glycine, displayed in CPK, spacefill, was fitted manually in accordance with ³⁵; Zn²⁺, fitted manually, in grey, spacefill. Exact side chain orientations presented here are speculative, based on best alignment and modelling estimations. Inset, blue box – pentamer plan view showing viewing angle (arrow) for main picture, and revealing the separation between the Zn²⁺ binding site (grey circle) and glycine binding site (mauve circle); enlarged plan view of interface shown in black box. (d, e) Zn²⁺ concentration response curves for modulation of EC₅₀ glycine currents on wild–type (WT; dashed line) and alanine–substituted receptors. α1^I120A did not traffic to cell surface so no recordings were made (Supplementary Figure 2). Curves fitted with Hill equation. All points are means ± s.e.m. (n = 3–6).

Here, we demonstrate that a cluster of residues forming a scaffold across the hydrophobic core are critical for Zn²⁺ inhibition and spontaneous opening of the human GlyR ion channel. Spontaneous opening was attributed to the apparent flexibility of a loop on the ‘–’ face of the Zn²⁺ inhibition site, which is exquisitely-sensitive to the molecular composition of the hydrophobic core. Disruption of this loop and the discovery of novel elements in a neighbouring β3 strand that are also important for receptor activation, demonstrate that charge redistribution at the ECD inner subunit interface is a key component of GlyR activation.

Results

Hydrophobic residues are required for Zn²⁺ inhibition

A GlyR homology model was constructed to guide our site-directed mutagenesis studies into receptor activation. The GlyR protein sequence was first aligned with other Cys-loop receptors, revealing that the region encompassing the GlyR Zn²⁺ inhibition site is not conserved across any of the Cys-loop receptors, for which atomic resolution ECD templates are available, and furthermore, it contains a 2 – 3 amino acid insertion. Structural alignment of three ECD template structures (conotoxin-bound Aplysia californica acetylcholine binding protein, (Ctx-Ac-AChBP ²⁷); Torpedo (Tor) nAChR α₁ ¹; and mouse nAChR α1 subunit ¹⁴), revealed substantial structural variation at the two loops of the ‘+’ and ‘–’ face ²⁸ flanking the β5 strand (nomenclature of Brejc ²; Supplementary Fig. 1), which corresponds to the Zn²⁺ inhibition site. Using the variable loop regions as insertion points for the extra residues in our GlyR alignment (Fig. 1a) allowed us to generate a GlyR homology model (MODELLER-9.2 ²⁹; based on the TornAChR α₁ template) with His107 and His109 exposed at the subunit interface ²⁵ and Phe108 solvent accessible ³⁰, in accord with published data.

To probe the link between Zn²⁺ inhibition and GlyR activation, hydrophobic residues positioned between the Zn²⁺ inhibition site, defined by His107, His109, Thr112 and Thr133 ^{24, 26, 31}, and the three closest agonist binding loops (A, D and E; Fig. 1b,c), were substituted with alanine. Substituted GlyRs were expressed in human embryonic kidney (HEK) cells and their sensitivities to Zn²⁺ potentiation and inhibition assessed by whole-cell recording of glycine (EC₅₀) evoked responses in the presence of increasing concentrations of Zn²⁺. Substituting residues at the Zn²⁺ inhibition site ‘–‘ face (β5 F108A, I111A; β6’ I132A, L134A) and the loop A face (L98A, F99A, F100A), substantially reduced sensitivity to Zn²⁺ inhibition compared to wild-type, whilst Zn²⁺ potentiation remained the same (Fig. 1d,e). Substitutions in agonist binding loop D (V60A, I62A, L64A) also reduced Zn²⁺ inhibition with I62A causing ablation. This residue is orientated in the GlyR model towards other residues required for Zn²⁺ inhibition from loop A (Leu98, Phe99, Phe100) and those from β5 and β6’ adjoining the Zn²⁺ inhibition site (Phe108, Ile111, Ile132, Leu134). By contrast, substituting distally-located hydrophobic residues in agonist-binding loop E (Leu117 and Leu118; Fig. 1e and Table 1), and other hydrophobic residues located away from loops A/D and β5/6’ (Supplementary Fig. 3 and Supplementary Table 1), did not disrupt Zn²⁺ inhibition. All the substituted receptors with attenuated Zn²⁺ inhibition retained glycine EC₅₀s within 10-fold of the wild-type (Table 1) and comparable maximal glycine currents (I_max) and Hill slopes, with the exception of α1^F99A (I_max = 3.9 ± 0.4 nA; wild-type GlyR I_max = 6.8 ± 0.5 nA; P < 0.01).

Table 1.

Glycine activation, Zn²⁺ modulation and Zn²⁺ activation data from wild-type (WT) receptors and from GlyRs carrying alanine substitutions of hydrophobic residues at positions running from the Zn²⁺inhibition site to nearby agonist binding loops, expressed in HEK cells.

	Glycine				Zn2+
					Inhibition	Activation
	EC50 (μM)	nH	Imax (nA)	N	IC50 (μM)	EC50 (μM)	nH	Relative efficacy (% Gly Imax)	N
α1 WT	35 ± 5	2.7 ± 0.2	6.8 ± 0.5	6	15 ± 2	None	—	—	4
Agonist binding loop A
α1L98A	35 ± 6	1.9 ± 0.3	6.4 ± 0.5	6	> 1000	0.26 ± 0.08	1.1 ± 0.2	49 ± 5	4
α1F99A	250 ± 40	2.3 ± 0.3	3.9 ± 0.4	4	> 1000	0.13 ± 0.03	1.1 ± 0.1	86 ± 13	5
α1F100A	120 ± 20	1.8 ± 0.2	4.5 ± 0.7	5	> 1000	9.2 ± 0.6	0.5 ± 0.1	7 ± 4	4
Zn2+ binding site: β5 strand
α1F108A	15 ± 3	2.3 ± 0.1	4.5 ± 0.5	5	> 1000	2.3 ± 0.6	0.7 ± 0.1	25 ± 6	6
α1I111A	39 ± 4	1.3 ± 0.2	4.6 ± 0.8	4	> 1000	470 ± 130	0.7 ± 0.2	83 ± 3	3
Zn2+ binding site: β6’ strand
α1I132A	95 ± 15	2.7 ± 0.4	6.1 ± 0.7	6	> 1000	140 ± 30	1.2 ± 0.1	72 ± 5	5
α1L134A	58 ± 3	3.0 ± 0.4	6.0 ± 0.3	7	> 1000	0.06 ± 0.01	1.3 ± 0.1	38 ± 9	5
Agonist binding loop Δ
α1V60A	91 ± 15	3.5 ± 0.8	6.5 ± 1.3	4	10.1 ± 1.26	None	—	—	4
α1I62A	240 ± 7	2.5 ± 0.1	7.6 ± 0.7	3	> 1000	> 1000	—	3 ± 1	4
α1L64A	10 ± 3	2 ± 0.6	5.6 ± 0.8	4	70 ± 9	1.5 ± 1.3	1.1 ± 0.4	10 ± 3	3
Agonist binding loop E
α1L117A	4200 ± 100	1.67 ± 0.1	4.4 ± 0.7	8	13.3 ± 3.3	None	—	—	3
α1L118A	1060 ± 110	3.00 ± 0.2	6.3 ± 0.7	3	10.3 ± 0.36	None	—	—	3

Removing Zn²⁺ inhibition creates Zn²⁺-activated GlyRs

Although Zn²⁺ does not activate wild-type GABA_A or glycine receptors, it generated inward currents at GlyRs with impaired Zn²⁺ inhibition, i.e., those with alanine substitutions in the glycine binding loops A/D and Zn²⁺ inhibition binding strands β5/6’ (L98A, F99A, F100A, F108A, I111A, I132A, L134A, I62A and I64A; Fig. 2a). The Zn²⁺-activated currents reversed close to E_Cl (0.4 ± 1.4 mV, n = 5), and the current-voltage relationships were comparable to those for glycine-activated Cl^- currents at wild-type GlyRs (Fig. 2b). Zn²⁺ concentration response curves revealed that the potency and relative efficacy (maximal Zn²⁺ response as a percentage of maximal glycine response in the same cell), varied substantially between the substituted receptors (Fig. 2c and Table 1). α1^L134A exhibited the highest sensitivity to Zn²⁺ (EC₅₀ = 0.06 ± 0.01 μM, n = 5), whilst α1^F99A exhibited the highest relative efficacy (86 ± 13%, n = 5). By contrast, α1^I62A supported only 3 ± 1% maximal activation with 1 mM Zn²⁺ (Fig. 2c and Table 1). Entirely consistent with Zn²⁺ activating the substituted GlyRs, the anion-selective channel blocker, cyanotriphenylborate (CTB, 20 µM), abolished the Zn²⁺-activated currents (Fig. 2d). Furthermore, both strychnine, a selective GlyR competitive antagonist, and picrotoxin (PTX), a GABA_AR and GlyR allosteric blocker, also inhibited Zn²⁺ activation (Fig. 2e,f). This prompted the classification of these substituted GlyRs as Zn²⁺-activated GlyRs (ZAGs).

Figure 2. Direct Zn²⁺ activation of substituted GlyRs.

(a) Representative glycine currents (blue traces) in the absence of Zn²⁺ (EC₅₀ – 30 μM; I_max – 1000 μM) and Zn²⁺ currents (black traces) in absence of glycine (0.01 – 30 μM) for GlyR α1^F99A. (b) Zn²⁺ current–voltage (I–V) relationships (normalised to the current recorded at –60 mV and fitted by linear regression) for the six most efficacious ZAGs, where EC₅₀ Zn²⁺ responses were large enough to be recorded reliably, and for the wild–type (WT) receptor. (c) Zn²⁺ activation–response curves for alanine–substituted GlyRs; Zn²⁺ maxima are normalised to glycine maxima (10 mM) in same cell. (d) Representative α1^F108A Zn²⁺ I_max (100 μM) current co–applied with and without 20 μM CTB. Inhibition greater than 100% is due to additional background leak (discussed later in Fig. 6). (e, f) α1^L98A concentration inhibition curves for strychnine and picrotoxin, respectively on glycine or Zn²⁺ EC₅₀–activated currents (n = 3 –6). Wild–type receptor activation sensitivities to the antagonists are shown in orange.

Hydrophobicity and sensitivity to Zn²⁺ inhibition

To investigate the ZAGs further, we substituted loop A, Phe99 and Phe100, and β5 Phe108 (Fig. 3a) individually with either: tyrosine or tryptophan – both aromatic like Phe; or leucine or methionine – aliphatic. Although many substitutions reduced Zn²⁺ inhibition (Fig. 3b-d), only some generated ZAG behaviour (Fig. 3e-g). Of the three Phe residues, Phe100 is the most critical for maintaining wild-type sensitivity to Zn²⁺ inhibition; however, there was no correlation between the properties of the substituting residue and the disruption to Zn²⁺ inhibition (Fig. 3b-d and Supplementary Table 2), suggesting each position has unique chemical and physical requirements.

Figure 3. Zn²⁺ inhibition and Zn²⁺ activation in GlyRs with conservative substitutions.

(a) Homology model showing Phe99 and Phe100 from glycine binding loop A (green) and Phe108 from β5 strand of Zn²⁺ inhibition site (black). Dotted line: interface with (+) and (–) subunit sides ²⁸. Glycine (CPK, spacefill) was docked manually ³⁵; Zn²⁺ (representation only, grey spacefill). Inset – pentamer plan viewing angle (arrow), Zn²⁺ binding site (grey spot), glycine binding site (mauve spot). (b–d) Zn²⁺ concentration response curves for modulation of EC₅₀ glycine currents from GlyRs with conservative hydrophobic substitutions at positions Phe99 (b), Phe100 (c), and Phe108 (d). For comparison, wild–type (WT) inhibition profiles (orange) and alanine substituted receptor profiles (grey) are included. The same set of substituted GlyRs were assessed for maximal Zn²⁺ (1 mM)–activated currents (I_max), normalised to glycine I_max (10 mM) in the same cell: Phe99 (e), Phe100 (f) and Phe108 (g). Alanine substituted receptor Zn²⁺ I_max (black bars) is shown for comparison. Note: F99L, F100Y and F108Y had substantially attenuated Zn²⁺ inhibition (navy blue lines) but almost no Zn²⁺ activation (navy blue bars; < 2% Gly I_max). (n = 3 – 6)

Glycine EC₅₀s for α1 Phe100 and Phe108 substituted receptors remained within two-fold of wild-type with the exceptions of α1^F100Y and α1^F108W (increased 15- and 25-fold, respectively; P<0.05; Supplementary Table 2). For Phe99 substituted receptors, glycine EC₅₀s were significantly increased for α1^F99L, α1^F99W and α1^F99Y (3- to 30-fold; P<0.05), but surprisingly the α1^F99M receptor was 6-fold more sensitive to glycine (wild-type EC₅₀ = 35 ± 5 μM; α1^F99M EC₅₀ = 5.5 ± 0.5 μM, n = 4 – 6; P<0.05), suggesting an important role for this residue in determining glycine binding. Interestingly, the GlyR model positions Phe99 facing into the glycine binding site (Fig. 1b,c).

Zn²⁺ activation originates from a novel binding site

The switch from Zn²⁺ inhibition to activation in ZAGs could have arisen if the function of an existing modulatory Zn²⁺ binding site was altered enabling activation in response to Zn²⁺ binding. However, substitution of Zn²⁺ binding residues with non-Zn²⁺ coordinating alanines at either the Zn²⁺ inhibition or potentiation sites revealed that neither site was required for Zn²⁺ activation (Supplementary Fig. 4). However, given that Zn²⁺ inhibition was severely compromised in ZAGs, we reasoned that regions bordering the inhibition site may have become structurally perturbed, sufficient to form a new Zn²⁺ activation site.

On an α1^L134A ZAG background (most Zn²⁺-sensitive ZAG), potential Zn²⁺ coordinating residues neighbouring the Zn²⁺ inhibition site were substituted with alanine and assessed for activation by 0.1 μM Zn²⁺ (EC₇₀ for α1^L134A; Fig. 4a,b). Of these substitutions, Asp84, Asp86 (strand β3) and Asp97 (loop A) virtually abolished Zn²⁺ activation from 38 ± 9% (α1^L134A) to 5 ± 1% (α1^{L134A, D84A}) and 2.4 ± 1.2% (α1^{L134A, D86A}), with no detectable activation for α1L^{L134A, D97A} (Fig. 4c). The Zn²⁺ EC₅₀s for α1^{L134A, D84A} and α1^{L134A, D86A} were increased 120- and 15-fold, respectively (Fig. 4c inset); whilst glycine EC₅₀s were only shifted 2-fold (Supplementary Table 3). Notably, using a wild-type receptor background, the substitutions D84A, D86A or D97A, caused only modest (< 2-fold) changes in GlyR sensitivity to Zn²⁺ inhibition and potentiation (Supplementary Fig. 5 and Supplementary Table 4).

Figure 4. Identifying potential residues for the Zn²⁺ activation binding site in ZAGs.

(a) Homology model showing potential Zn²⁺ coordinating residues at the subunit interface around the His107 /His109 /Thr112 /Thr133 Zn²⁺ inhibition site. Side chains in grey, nitrogens in blue, oxygens in red. Dotted line: interface with (+) and (–) sides ²⁸. Glycine (CPK, spacefill) was docked manually ³⁵; Zn²⁺ (representation only, grey, spacefill). Inset shows the viewing angle (arrow), with Zn²⁺ (grey spot) and glycine binding sites (mauve spot). (b) 0.1μM Zn²⁺–activation of alanine–substituted α1^L134A background GlyRs, expressed as a percentage Zn²⁺ I_max in the same cell. (c) Zn²⁺ activation concentration response curves for α1^{L134A, D84A}, α1^{L134A, D86A} and α1^{L134A, D97A} receptors normalised to the Gly I_max and also normalised to the Zn²⁺ I_max (inset). (d) Zn²⁺ activation response curves for the equivalent mutations on an α1^F99A background showing that although Asp84 and Asp86 are still required for Zn²⁺ activation, Asp97 is not. (n = 3–6).

According to the GlyR homology model (Fig. 4a), Asp84 and Asp86 are positioned approximately 17 Å from Asp97, which is too far for the three residues to coordinate a single Zn²⁺ ion ³². Furthermore, Asp97, which is conserved across the Cys-loop receptor family, probably supports loop B via the carboxyl side chain ³³, precluding its involvement in Zn²⁺ binding. To establish the importance of Asp84, Asp86 and Asp97 for Zn²⁺ activation, alanine substitutions were also made on an alternative ZAG background, α1^F99A (most efficacious ZAG). Substituting Asp84 or Asp86 again substantially reduced the sensitivity to Zn²⁺ activation, but substituting Asp97 was ineffective (Fig. 4d). Thus, the role of Asp97 in Zn²⁺ activation is more complex than just directly binding Zn²⁺.

Zn²⁺ activation site is not a potentiation site

As reagents and water are ubiquitously contaminated with glycine (~ 50 nM ³⁴), it is feasible that Asp84/Asp86 may actually form part of a (second) Zn²⁺ potentiation site, rather than an activation site. Occupancy of this site by Zn²⁺ would then enhance the receptor’s sensitivity to glycine allowing activation by very low contaminating glycine concentrations. To address this, an α1^L134A ZAG background was used with an extra mutation, E157A on glycine binding loop B³⁵, to produce a receptor with 50-fold reduced sensitivity to glycine. The threshold concentration for glycine was now >100 μM (Fig. 5a) and 2000-fold higher than the predicted level of glycine contamination. Nevertheless, the α1^L134,E157A ZAG showed only a modest 3-fold reduced sensitivity to Zn²⁺ activation (Fig. 5b) and retained comparable maximal responses to Zn²⁺ (Fig. 5c). Using an alternative F99A background, α1^F99A,E157A yielded identical Zn²⁺ sensitivity to α1^F99A, despite being insensitive up to 10 mM glycine (Fig. 5d,e). Thus, in the absence of glycine-mediated activation, Zn²⁺ activation is still apparent, suggesting it originates from a pure activation site, not an additional Zn²⁺ potentiation site.

Figure 5. Zn²⁺–activation of GlyRs lacking a high affinity glycine binding site.

Glycine (a) and Zn²⁺ (b) concentration response curves on an α1^L134A background with an extra mutation in the glycine binding site (E157A). (c) Maximum currents evoked by glycine (10 mM) and Zn²⁺ (1 mM) for α1^L134A and α1^{L134A, E157A}. Incorporating E157A on an α1^F99A background, ablated activation by up to 10 mM glycine, but induced only a modest, 3–fold decrease in sensitivity to Zn²⁺ activation (d) and no reduction in Zn²⁺ I_max (1 mM) (e) (n = 3–4).

ZAGs exhibit spontaneous channel activity

HEK cells expressing the most sensitive ZAGs, α1^L98A, α1^F99A, α1^F108A and α1^L134A, all exhibited sizable (0.5 – 3 nA) leak currents. A minor component was caused by Zn²⁺ contamination of the external solution (~200 nM ³⁶) activating the ZAGs, as this was reduced by the Zn²⁺ chelators, tricine (2.5 mM) or N,N,N’,N’-tetrakis-(2-pyridylmethyl)-ethylenediamine (TPEN; 100 μM; Fig. 6a, b). The remaining component depended on spontaneous GlyR channel activity since it was abolished by CTB (20 μM) to less than 50 pA standing current (considered full abolition of receptor-mediated leak). Strychnine also attenuated the leak, by 100 % for ZAG α1^F99A and by 90 ± 4% for α1^L134A. Strychnine was 10-fold less potent inhibiting the leak current compared to glycine-activated currents (Fig. 6c, d).

Figure 6. Spontaneous activation of alanine–substituted receptors.

(a) Persistent leak currents for alanine substituted GlyRs (= % CTB blockable current (I_CTB)/(I_CTB + Gly I_Max)) in presence of individual and combined ion chelators. (b) Membrane currents for α1^F99A ZAG in the presence of either glycine (orange), CTB (black) or TPEN (blue), overlaid to demonstrate their relative contributions to the total current (c) Strychnine concentration curves for inhibiting leak currents for α1^F99A and α1^L134A ZAGs (c) and EC₅₀ glycine–activated current (d). Orange line indicates strychnine concentration inhibition curve for WT GlyRs, which are notably more sensitive as the agonist binding loops A and D are not perturbed. (e) Persistent leak current absent from receptors with alanine substitutions at Asp84 or Asp86 on α1^F99A and α1^L134A backgrounds. (f) Glycine sensitivity is modestly reduced in alanine–substituted Asp84 or Asp86 receptors, as compared to wild–type, α1^F99A and α1^L134A backgrounds. n = 3 –6.

Interestingly, Asp84 and Asp86, which are important for Zn²⁺ activation, were also required for spontaneous activation with α1^{F99A, D84A}, α1^{F99A, D86A}, α1^{L134A, D84A}, and α1^{L134A, D86A} failing to exhibit spontaneous channel activity (Fig. 6e). Furthermore, Asp84 and Asp86 also influenced glycine-induced receptor activation, as alanine substitutions induced a modest but consistent 2-fold reduction in glycine sensitivity of wild-type, α1^F99A and α1^L134A backgrounds (Fig. 6d and Supplementary Table 4). A double substituted receptor, α1^{D84A, D86A}, was non-functional.

To investigate whether spontaneous activity mimics agonist-induced activity, single channel currents were recorded in cell-attached mode (pipette potential +60 mV) for wild-type channels activated by glycine (20 μM; EC₃₀) and for spontaneously-gating α1^F99A ZAGs without glycine. TPEN (100µM) was present throughout to remove any activation by contaminating Zn²⁺. The single channel currents for each receptor population were comparable at 4 – 5 pA with estimated conductances of ~ 60 pS ³⁷ (Fig. 7a). The corresponding open time distributions were best fit by three Gaussian components with similar mean time constants and relative areas (P > 0.05; Fig. 7b and Table 2). The shut time distributions required five Gaussian components giving similar time constants for both receptors, with the exception of τ_C2 and τ_C3, which were 2-fold higher for α1^F99A receptors (Fig. 7c and Table 2). These changes will contribute to the lower open probability (P_O) for clusters of openings at α1^F99A (0.53 ± 0.07) compared to wild type (0.9 ± 0.03, n = 3) channels. With regard to the burst duration distributions, four Gaussian components were required with comparable time constants and relative areas, except for τ_B3, which was 2-fold longer for α1^F99A (Fig. 7d).

Figure 7. Spontaneous channel activation mimics agonist–induced activation.

(a) Single channel currents from cell–attached recordings of HEK cells expressing either WT α1 GlyRs or GlyR α1F99A, at a pipette potential +60 mV. Burst activity was recorded in the presence of glycine (EC₃₀) for WT and in its absence for spontaneously–activated α1^F99A. TPEN (100 μM) was present in the pipette solution to remove contaminating Zn²⁺. (b–d) Dwell time distributions for (b) open times, (c) shut times, and (d) burst durations.

Table 2.

Data from cell-attached single-channel recordings made from α1 wild-type (WT) or α1^F99A receptors expressed in HEK 293 cells. Includes: average durations of open, closed and burst time constants and areas of exponential components that fitted the distributions; single channel amplitudes; P_O for openings within bursts. n = 3. * denotes significant variations between the two receptor populations (P<0.05).

	α1 WT (+30 μM Gly)		α1F99A (no glycine)
Open times	τO (ms)	Area (%)	τO (ms)	Area (%)
1	0.28 ± 0.029	36 ± 11	0.4 ± 0.1	36.9 ± 9.3
2	1.5 ± 0.2	39 ± 3	1.7 ± 0.3	41.9 ± 4.0
3	8.9 ± 2.1	26 ± 10	5.2 ± 0.8	21.2 ± 7.2
Closed times	τC (ms)	Area (%)	τC (ms)	Area (%)
1	0.22 ± 0.03	44 ± 16	0.26 ± 0.026	37 ± 2
2	0.90 ± 0.24	35 ± 15	1.6 ± 0.17 *	30 ± 3
3	4.3 ± 1.6	8 ± 1	10.0 ± 2.9	18 ± 1 *
4	109 ± 6.2	10 ± 2	89 ± 39	12 ± 1
5	2200 ± 1000	4 ± 1	1200 ± 400	3 ± 3
Burst durations	τB (ms)	Area (%)	τB (ms)	Area (%)
1	0.3 ± 0.06	28 ± 7	0.3 ± 0.06	26 ± 2
2	1.4 ± 0.6	22 ± 4	1.4 ± 0.1	22 ± 8
3	6.9 ± 2.2	35 ± 5	17.6 ± 0.8 *	32 ± 8
4	63.4 ± 14.8	19 ± 7	77.1 ± 10.7	20 ± 1
Amplitude (pA)	4.5 ± 0.7		4.1 ± 0.6
PO	0.9 ± 0.03		0.53 ± 0.07

The ‘–’ face affects Zn²⁺ activation and spontaneous activity

Conceivably, Zn²⁺ activation and spontaneous channel activity may arise if the substitutions of hydrophobic residues exert a common conformational effect on a region that undergoes critical movement during channel gating. As the Zn²⁺ inhibition site is perturbed in ZAGs, it is the ideal region to examine for conformational flexibility. The top ten GlyR homology models (lowest distant-dependent atomic statistical potential (DOPE)³⁸ score from 100 models run in MODELLER-9.2) based on three related template structures, conotoxin-bound Ac-AChBP, TornAChR α₁ and mnAChR α1 ^{1, 14, 27}, showed much greater structural variability at the ‘+’ and ‘–’ loop faces surrounding the β5 strand of the Zn²⁺ inhibition site, compared to other more rigid β-strands and across the structure as a whole (Fig. 8a, insets and Supplementary Table 5). By using DOPE loop modelling to optimise the structures of the GlyR ‘+’ and ‘–’ face loops, using TornAChR α₁ and mnAChR α1 as templates, the ten best conformations for the ‘+’ loops were all comparable (Fig. 8b), whereas for the ‘–’ face loops, variable conformations were equally favoured with residues exhibiting multiple orientations at this location (Fig. 8b and Supplementary Table 6).

Figure 8. Zn²⁺ and spontaneous channel activation originates via the Zn²⁺ inhibition site ‘–’ loop face.

(a) CE structural alignments ⁵⁵ of GlyR α1 homology models based on three predicted ‘closed’ conformation templates. Note, structural variability at the Zn²⁺ inhibition site (+) and (–) faces between the conotoxin–bound Ac–AChBP ²⁷ (green), mnAChR α1 ¹⁴ (red) and TornAChR α₁ ¹ (blue) templates. (a – insets) Overlays of 10 lowest energy models for each template reveal particular uncertainty over the fitting of the (–) face but less uncertainty for the (+) face. (b) 10 lowest energy loop conformations resulting from DOPE loop refinement ³⁸ of the (+) and (–) Zn²⁺ inhibition site loops starting from the best GlyR α1 model based on two different templates. This revealed one favourable conformation for the (+) face loop for each template–derived model, with only minor deviations in the peptide backbone (red) and side chain positions between models. In contrast, multiple peptide backbone conformations were equally preferred for the (–) face loop (pink or yellow for mnAChR α1; aquamarine or purple for TornAChR α₁) with clear divergence in side chain positions. Original model backbone conformations before loop refinement are shown in translucent red (mnAChR α1) and translucent blue (TornAChR α₁). (c) Zn²⁺ activation–response curves from GlyRs with potential Zn²⁺ coordinating residues on the (+) and (–) inhibition site faces substituted with alanines; Zn²⁺ maxima normalised to glycine maximal current (10 mM) in same cell. Only α1^T112A exhibited highly–sensitive and efficacious ZAG activity. ZAG α1^F99A (orange line; originally from Fig. 2c) included for comparison. Insets: DOPE loop refinement of GlyR α1 models from mnAChR α1 and TornAChR α₁ templates show how alternate favourable conformations on the (–) face loop could have a substantive impact on the organisation of Thr112 in the Zn²⁺ binding site. Zn²⁺ (fitted manually, dark grey, spacefill). (d) Significant spontaneous activity, measured as persistent leak currents (I_CTB/(I_CTB + Gly I_Max)) in the presence of 100 μM TPEN, was observed for GlyR α1^T113G and α1^D114G where ‘–’ loop flexibility was increased, but not for GlyR α1^T113P and α1^D114P where flexibility was reduced. (e) Glycine concentration response curves for receptors with glycine or proline substitutions at positions 112, 113 or 114 in the ‘–’ loop. Glycine increased agonist sensitivity compared to proline substitution at the equivalent position. Glycine curve for WT receptors included for comparison (orange line; n = 3–7).

To corroborate the modelling data, polar residues Thr112, Thr113 and Asp114 on the apex of the ‘–’ face, the point of greatest variability between DOPE loop-fitted structures, were individually substituted with alanine and examined for Zn²⁺ activation and spontaneous activity. Although T112A yielded a highly-sensitive and efficacious ZAG (Fig. 8c), no single alanine substitution generated a spontaneously-active receptor (data not shown). We altered the apex flexibility of the ‘–’ face by individually substituting Thr112, Thr113 and Asp114 with either glycine to increase, or proline to reduce, backbone flexibility ^{39, 40}. Whilst the proline-substituted receptors lacked spontaneous activity, two glycine-substituted receptors, α1^T113G and α1^D114G, exhibited 15 ± 4% and 43 ± 10% (n = 4 – 6) spontaneous activity, respectively (Fig. 8d). Furthermore, there was a 3–fold increase in glycine sensitivity for α1^T113G (EC₅₀ = 9 ± 2 μM) and α1^D114G (EC₅₀ = 10 ± 3 μM) compared to WT (EC₅₀ = 35 ± 5 μM; n = 4–6); whereas the proline substituted receptors all exhibited reduced sensitivities to glycine (Fig. 8e). Thus, increasing the flexibility of the ‘–’ face around Thr112–Asp114 increased the propensity of GlyR to open spontaneously, and in response to agonist binding, while decreasing flexibility by proline substitution had the opposite effect.

Discussion

This study identifies a scaffold of hydrophobic residues in the GlyR that functionally link glycine binding loops A and D with the Zn²⁺ binding β5 and β6’ strands of the Zn²⁺ inhibition site. Exchanging the hydrophobic residues, but not others outside the scaffold, initiated spontaneous channel opening, severely attenuated Zn²⁺ inhibition, and enabled Zn²⁺ to act as a novel activator of GlyRs. This suggests the hydrophobic scaffold is pivotally involved in receptor activation by stabilising one or more closed GlyR conformations. This is achieved by regulating the ‘–’ loop face of the Zn²⁺ inhibition site, as specific substitutions of polar residues in the ‘–’ face produced receptors with the same properties to those generated by alanine substitutions in the hydrophobic scaffold.

Structurally linking two discrete ligand binding sites

The current view of Cys–loop receptor activation is that agonist binding at the interface between subunits induces a rearrangement of interacting residues allowing the ECDs to twist relative to one another. The newly–orientated loops at the bases of the ECDs then promote rearrangement of opposing transmembrane domains to open the channel ^{6, 16, 18, 41–44}. By binding to its interfacial inhibitory site on GlyRs, Zn²⁺ stabilises subunit interfaces, preventing the ECDs from twisting and initiating activation. It is therefore plausible that by distorting the Zn²⁺ binding ‘–’ loop interface we will not only ablate Zn²⁺ inhibition, but also enable spontaneous channel activity, particularly if the distortion mimics the conformation that occurs in the activated receptor state. Thus, the attenuation of Zn²⁺ inhibition and appearance of spontaneous channel activation are intrinsically linked. The extent to which both properties are seen in mutated receptors, will depend on the degree by which each substitution perturbs the ‘–’ loop away from a closed towards an activated conformation.

Notably, the molecular pathway by which Zn²⁺ causes inhibition is entirely different to that for Zn²⁺ potentiation at GlyRs. The potentiation site resides very close to the Cys–loop where it may interact directly with Thr151 to facilitate channel gating ²³. This negates the need for any interaction with the hydrophobic scaffold identified here, explaining why Zn²⁺ potentiation was unaffected in this study.

The molecular pathway identified here is the first to be described in a Cys–loop ligand–gated ion channel that functionally connects two distinct binding sites, linking the agonist binding site to downstream activation. The importance of the hydrophobic scaffold in mediating GlyR activation is emphasised by the common kinetics of spontaneously–active α1^F99A and agonist–activated wild–type GlyRs. Specifically, for α1^F99A, it is the alanine substitution that artificially perturbs loop A to induce activation, whilst for wild–type GlyRs it is presumably agonist binding that similarly perturbs loop A to cause activation. Although a critical role for loop A in directing receptor activation is evident for GABA_ARs ⁴⁵ and nAChRs ⁴⁶, loop C, possibly via transmission of a conformational change along the outside of the ECDs (β7, 9 and 10 strands), has also been suggested to mediate activation upon agonist binding ^{8, 10}. Our data does not preclude this scenario, but advocates loop A as an important contributor to the conformational wave that precedes channel opening ⁴⁷.

At the glycine binding site, Phe99 appears ideally positioned to directly influence the receptor’s sensitivity to glycine, possibly via a cation–π interaction ⁴⁸. The action of Phe99 to induce GlyR activation in response to agonist binding may then be mediated via the hydrophobic scaffold and subsequent ‘–’ loop face of the Zn²⁺ inhibition site. Indeed, Phe99 probably does this via Leu98 and Phe100, which are predicted to face, opposite to Phe99, into the hydrophobic scaffold towards the residues supporting the ‘–’ face loop. Such an interaction with Phe99 would explain why Phe100 could also influence the receptor’s sensitivity to glycine (Supplementary Figure 6; Supplementary Table 2). The ability of Phe99 to influence important residues within the hydrophobic scaffold may explain why it produces the most efficacious ZAG and the most spontaneously–active receptors when substituted with alanine.

From the perspective of the polar residues at the Zn²⁺ site’s ‘–’ loop face, substituting Thr112 or Ile111 produced a receptor that was insensitive to Zn²⁺ inhibition (cf ^{25, 31}) and capable of Zn²⁺ activation. Isoleucine 111 faces into the core, in close proximity with the other residues comprising the hydrophobic scaffold. Thus Thr112, via Ile111, is ideally located to act as a relay following perturbation of the hydrophobic scaffold. Sequential substitution of Thr113 and Asp114 within the ‘–‘ loop by glycine, but not by alanine or proline, also yielded spontaneously–active receptors. These residues must also be ideally located to respond to perturbations of the hydrophobic scaffold, with increased loop flexibility enabling the receptor to shift to an activated state, while imposed rigidity (e.g., proline insertion or when Zn²⁺ binds to stabilise this region) hinders receptor activation.

β5 loop movement during GlyR activation

Although we propose that the ‘–’ face loop undergoes a conformational change to facilitate receptor activation, comparative structural evidence does not, so far, support this idea. Overlaying crystal structures of Aplysia californica AChBP bound to either α–conotoxin PnIA (‘inactive conformation’) or HEPES (‘active conformation’ ²⁷) does not reveal any variation around the corresponding ‘–’ face loop region in the GlyR model (Supplementary Figure 7). Furthermore, structurally aligning Torpedo nAChR α₁ and α₂ subunits (presumed closed conformation), compared to β, δ and γ subunits (presumed open) for the pentamer, reveals only a small degree of variability around the corresponding ‘–’ face loop region (Supplementary Figure 8). Of course, as static structures, it is possible that neither the HEPES–bound AChBP nor βδγ TorAChR subunits represent fully–activated receptors. Alternatively, they might undergo different conformational rearrangements after activation compared to GlyRs. Simulation studies on nAChRs also do not support movements in the ‘–’ face region ⁴⁹, although the nanosecond timescales for these studies are as yet too short to encompass all conformational rearrangements in pentameric Cys–loop receptors.

Despite the caveats, the ‘–’ loop face of the GlyR Zn²⁺ inhibition site was predicted to adopt multiple conformations and side chain orientations with the potential to influence receptor function. Moreover, previous functional studies support a role for the ‘–’ face loop in the GlyR activation process, notably: Thr112 is important in determining partial agonist efficacy ⁵⁰; it is accessible to Cys–scanning mutagenesis, resulting in dynamic disruption to agonist–evoked responses ³⁰; and Zn²⁺ binds between subunits at the ‘–’ loop face to stabilise the GlyR closed conformation, suggesting this interface is mobile during receptor activation ^{25, 31}.

Creating a Zn²⁺ activation site

The Zn²⁺ activation site was localised to Asp84/Asp86 on strand β3, directly above the ‘–’ face loop. Structural perturbation of the ‘–’ loop face may therefore have a knock–on effect on the β3 region, allowing Asp84/Asp86 to form a novel Zn²⁺ binding site that aids movement of the subunit interfaces, rather than hinders them, so inducing activation. This provides further evidence that charge dispersal at subunit interfaces plays an important role in regulating Cys–loop receptor excitability ^{16, 19, 46, 51}, and also indicates that a dynamic interaction occurs between the ‘–’ loop face and the β3 strand to facilitate activation of wild–type receptors. The variable potency and efficacy of Zn²⁺ at different ZAGs further indicates that hydrophobic residues within the scaffold differentially affect the ‘–’ loop face, and consequently the juxtaposed β3 strand, so determining the efficiency with which Asp84/Asp86 forms a new Zn²⁺ activation site.

The general activation mechanism presented here for the GlyR is in accord with the hydrophobic scaffold and ‘–’ face loop dynamically responding to agonist binding. This provides a new vista on Cys–loop receptor activation whereby, during activation, the reorganisation of the hydrophobic scaffold and ‘–’ face facilitate the re–alignment of the inner and outer β–sheets relative to one another ^{12, 13, 16, 18}. This then initiates movement of subunit interfaces, which is subsequently transmitted to the transmembrane domains for receptor activation ⁴².

Methods

cDNA constructs and mutagenesis

We used human (h) GlyR α1L cDNA constructs and the mutant cDNAs were prepared using the Stratagene Quikchange kit. Mutated cDNAs were sequenced using an ABI sequencer.

Cell culture, transfection and electrophysiology

By using a calcium phosphate transfection method (3 GlyR α1:1 eGFP) we expressed GlyR in HEK cells (ATCC CRL1573) grown on poly–L–Lysine–coated coverslips at 10% confluence. Whole–cell membrane currents were recorded after 24 h at 20–22°C from single HEK cells held at –40 mV using the patch clamp technique (Axopatch 200B, Molecular Devices). For rapid drug applications (exchange rate ~50–100 ms), we used a Y–tube. Patch electrodes (4 – 5 MΩ) were filled with (mM): 140 KCl, 2 MgCl₂, 1 CaCl₂, 10 HEPES, 11 EGTA, and 2 ATP, pH 7.2 (≈ 300 mOsm). External solution contained (mM): 140 NaCl, 4.7 KCl, 1.2 MgCl₂, 2.5 CaCl₂, 10 HEPES, and 11 D–Glucose, pH 7.4 (≈ 300 mOsm). For GlyRs exhibiting nanomolar sensitivities to Zn²⁺ activation, the tricine (2.5 mM; Zn²⁺ complexation K_D = 10^–5 M, ref. ⁵²), was added to the saline to remove Zn²⁺ contamination ³⁶. For single channel analysis, thick–walled electrodes were used (10 – 20 MΩ) and filled with external saline solution containing 100 μM TPEN, and 10 mM TEA to block endogenous potassium channel activity. Single–channel recordings were made in cell–attached mode at +60 mV pipette potential.

Data acquisition and analysis

Membrane currents were filtered using a high–pass Bessel filter at 3 kHz (–36dB per octave) and series resistance compensation was routinely achieved up to 70%. Data were recorded in 20 s epochs directly to a Pentium IV, 3.5 GHz computer into Clampex 8.0 via a Digidata 1322A (Axon instruments) sampling at 200 μs intervals. Due to Zn²⁺ activation in many of the receptors, Zn²⁺ inhibition profiles were measured by prolonged (4 s) co–application of Zn²⁺ with glycine, with response measurements being taken at the 4 s time point (to allow Zn²⁺ inhibition sufficient time to reach equilibrium ²⁶). The digitised membrane current records were analysed offline using Axoscope 8.2. The concentration response relationships for glycine and Zn²⁺ were fitted with modified Hill equations as previously described ²⁶.

For the single channel data analyses, stored pre–filtered (2.7 kHz Bessel) single channel data were digitised at 33 kHz prior to analysis. A fixed time resolution based on the dead time of the system was set at 80 μs. The analysis of the single channel current amplitudes was performed by fitting Gaussian components to the amplitude distributions to determine the mean single channel current, standard deviation and the total area of the component using a non–linear least–squares fitting routine. Single–channel conductances were calculated from the mean unitary current and the difference between the patch potential and glycine response reversal potential. The patch potential was estimated in cell–attached recordings, by estimating the cell membrane potential.

All open and shut durations were measured with a 50% threshold cursor applied to the main single channel current amplitude (WinEDR v2.8.9). The duration of events that were included in the analysis was not less than 200 μs before fitting the dwell time histograms. Frequency distributions were constructed from the measured individual open and shut durations and analysed by fitting a mixture of exponentials, defined by:

$$y(t)={\displaystyle \sum _{i=1}^n(A_i/{\tau }_i)}\cdot \exp (-t/\tau )$$

where A_i represents the area of the i^th component to the distribution and τ_i represents the respective time constant. The areas, time constants and standard errors of the individual components of the distribution were determined. The burst duration analysis required the determination of a critical shut time (τ_crit) ⁵³ determined between the shut time constants, τ_C3 and τ_C4 by solving:

$$\exp (-t_{crit}/{\tau }_{C3})=1-\exp (-t_{crit}/{\tau }_{C4})$$

Channel open probability (P_O) was calculated as the percentage of time that the channel spent in the open state within a cluster. All statistical comparisons used an unpaired t test and P<0.05 was considered significant.

Homology and loop modelling

We used ClustalW ⁵⁴ to produce protein sequence alignments. Aplysia californica acetylcholine binding protein, Ac–AChBP (2br8 ²⁷; conotoxin–bound form), Torpedo (Tor)nAChR α₁βδα₂γ (2BG9) ¹, and mouse (m)nAChR α1 (2QC1 ¹⁴) were used for the Combinatorial Extension (CE) structural alignment method ⁵⁵, which helped identify divergent regions in the GlyR α1 model. The final alignment reflected both alignment strategies. The Torpedo nAChR α₁ subunit was selected as the final template structure to guide the homology modelling of the GlyR α subunit, as it has only two less residues around the β5 strand (the GlyR Zn²⁺ inhibition site), compared to three less residues for Ac–AChBP; and also, the structure of TornAChR α₁ was determined as part of a pentamer, whereas mnAChR α1 was crystallized as a non–physiological monomer with several artificial point mutations ¹⁴. The TornAChR α₁ pentamer was built by overlaying a second α₁ subunit over the α₂ subunit and then using Chimera ⁵⁶, to build a five–fold symmetric pentamer. Using MODELLER–9.2 ²⁹, 100 hGlyR α1 models were generated with Cys bridges added into the agonist binding loop C (C198–C209), for the principle TornAChR α₁ pentamer template, and also for the Ac–AChBP–conotoxin–bound pentamer and mnAChR α1 monomer. Side chain configurations were generated using SCWRL3 ⁵⁷. Fifty loops were generated using DOPE loop modelling in Modeller 9.2, for each loop before (‘+’) and after (‘–’) the β5 strand for each of the subunit templates (‘+’ loop residues ¹⁰²NEKGAH¹⁰⁷; ‘–’ loop residues ¹¹⁰EITTDN¹¹⁵). Models were evaluated using MolProbity ⁵⁸ and gave good general agreement with each other. Uncertainty regarding the short β5 strand, ascribed as a β–strand in Ac–AChBP (1UW6) and nAChR α1 (2QC1), but not in TornAChR α₁ (2BG9), was considered unimportant, as it had little effect on side chain positioning, and a PSIPRED ⁵⁹ secondary structure prediction of the GlyR sequence gave low confidence for the presence of a β–strand, suggesting neither one nor other template was more likely to be correct. All 3–D images were prepared and rendered using Chimera ⁵⁶.