Gating mechanism of the human α1β GlyR by glycine

Summary

Glycine receptors (GlyRs) are members of the Cys-loop receptors that constitute a major portion of mammalian neurotransmitter receptors. Recent resolution of heteromeric GlyR structures in multiple functional states raised fundamental questions regarding the gating mechanism of GlyR, and generally the Cys-loop family receptors. Here, we characterized in detail equilibrium properties as well as the transition kinetics between functional states. We show that while all allosteric sites bind cooperatively to glycine, occupation of 2 sites at the α-α interfaces is sufficient for activation and necessary for high efficacy gating. Differential glycine concentration dependence of desensitization rate, extent, and its recovery suggests separate but concerted roles of ligand-binding and ionophore reorganization. Based on these observations and available structural information, we developed a quantitative gating model that accurately predicts both equilibrium and kinetical properties throughout the glycine gating cycle. This model likely applies generally to the Cys-loop receptors and informs on pharmaceutical endeavors.

Graphical Abstract

eTOC Blurb

Liu et al. characterized the steady-state and kinetic properties of human heteromeric GlyR throughout the gating cycle. Through strategically generated mutations, basic gating properties were identified, which led to a mathematical gating model. This quantitative model hints on the fundamental principles underlying heteromeric Cys-loop receptor family function.

Introduction

Cys-loop receptors are pentameric ligand-gated ion channels that constitute a major portion of ionotropic neurotransmitter receptors in the mammalian nervous systems, including the cation-selective nicotinic acetylcholine receptors (nAChR) and type 3 serotonin receptors (5-HT3), as well as the anion-selective glycine receptors (GlyR) and GABA(A) receptors^1,2. GlyRs mediate both locomotive and sensory signals in the spinal cord, whose disfunction causes hyperekplexia (startle syndrome)^3,4 and relates to chronic inflammatory pain^5,6, making them potential drug targets^7–10. Autism spectrum disorders are also genetically associated with defects of GlyRs, implicating their indispensable function in the brain^11,12.

GlyRs share a general gating scheme with other Cys-loop receptors with unclear underlying mechanism^13,14. Upon binding with glycine, GlyRs quickly transition from non-conductive resting state to conductive open state. The open state then turns into non-conductive desensitized state, and only recovers by the removal of glycine. In early developmental stages GlyRs are homomeric containing 5 α subunits, while in adult animals they are heteromeric with an invariant 4 α:1 β stoichiometry ^15–18. Recent heteromeric GlyR structures contained both fully and partially occupied glycine binding sites, all of which share an identical pseudo-symmetric extracellular glycine-binding domain (ECD) conformation, and with no clear correlation with the open or desensitized states^16–18. This raises the question of how many glycine bindings are required, and whether different subunit types contribute differentially to GlyR activation, like other Cys-loop receptors^19–24. In addition, how the same glycine bound ECD conformation results in different functional states (open and desensitized) is puzzling.

To address the above questions, we systematically characterized the kinetical and steady-state properties of the wild-type human α1β GlyR and its mutants throughout the glycine gating cycle. Based on the invariant 4:1 α:β stoichiometry, we generated strategic mutations and show that 2 allosteric sites at the α-α interfaces are necessary and sufficient for GlyR activation, while sites at the β subunits contribute through binding cooperativity. We also show simple mass action-based behavior for open and close, but a glycine concentration-independent desensitization rate, suggesting that the open to desensitization transition is purely within the transmembrane pore-forming domain (TM). The more extensive desensitization and slower recovery from higher glycine concentrations can be attributed to slower glycine binding/dissociation kinetics in the activated ECD conformation. These observations were consolidated into a mathematical model that encompasses all GlyR gating parameters and closely predicts channel behavior. Together, this work provides a quantitative description that addresses fundamental questions regarding GlyR gating mechanism and reconciles the seemingly discrepancies in structural and functional studies.

Results

2 α1-α1 allosteric sites are sufficient for activation and necessary for high efficacy gating

Recent heteromeric glycine receptor structures revealed that all the 5 ECD allosteric sites in each pentameric channel are capable of glycine binding^16–18 (Figure 1A, B). Although less than full occupancy seemed sufficient for activation¹⁸, the exact number was unknown. In addition, whether all three types of binding sites (Figure 1B), α(+)-α(−), β(+)-α(−) and α(+)-β(−), share identical contribution to channel activation is unclear. Since heteromeric GlyRs are found to have an invariable stoichiometry of 4α:1β^16–18, strategic allosteric site mutations in α and/or β subunits, as well as in their concatemers allowed for the evaluation of function of individual sites. For this, we generated multiple α1-β and α1-β-α1 concatemeric subunits with different linkers and identified the ones that retained the wild-type glycine activation profile (Supplementary Figure 1C and D) for further mutagenesis that abolish glycine binding in specific allosteric sites^24–27 (Figure 1B). Mutation in allosteric sites do not alter the invariant 4:1 α:β stoichiometry (Supplementary Figure 1F), which is likely determined by the unique glycosylation on the β subunit^16,27.

Figure 1.

Glycine response of α1β GlyR with mutations in specific allosteric sites. 2 allosteric sites at the α-α subunit interface are sufficient and necessary for activation. (A) Side view and (B) top-down view of the α1β GlyR atomic models in complex with glycine. Glycine (yellow) binding at 3 types of pockets at the (+) and (−) sides of the α (sky blue/green) and β subunits (salmon) are shown with key amino acid residues. (C) Glycine dose-response of wild-type (α1: β) and indicated mutants (α1: α1-β-α1 FYF denotes α1: α1F207A-βY231A-α1F207A and α1: α1-β-α1 RFYF denotes α1: α1R65DF207A-βY231A-α1F207A. n = 8 - 15, mean ± S.E.M.). (D) Representative whole-cell recordings for derivation of dose-response curves shown in C. (E) Glycine dose-response of mutants containing less than 2 intact α1-α1sites (n = 7 - 10, mean ± S.E.M.). α1: α1-β-α1 FYF (hot pink dashed line) is shown for comparison. (F) Representative whole-cell recordings for E. Responses are normalized to max response at 2 mM glycine. The number and types of intact sites, together with EC₅₀S are shown in the table below. (G) Maximum current (mean ± S.E.M.) statistics of constructs in panel C and panel E. See also Figure S1, Figure S4, Figure S5, Table 1 and Table S1.

Key interactions between glycine and GlyR are conserved at α-β and α-α interfaces, constituting hydrogen bonding (αT204/ βT288, αR86/βR65, αS152/βS129, βY231) and cation-pi interaction (αF207, αF159/ βF182). Mutagenesis of these key amino-acid residues have been shown to dramatically increased EC50 (> 100 folds)^25,28,29. Combinations of (+) and (−) side mutations in the allosteric glycine binding pockets (Figure 1B) at the α1, β subunits, as well as α1-β and α1-β-α1 concatemers enumerated a panel of α1β GlyRs containing 1^~5 intact glycine binding sites (Figure 1C, Table 1 and Supplementary Figure 5A, B, C). In one mutant, α1:α1-β-α1 FYF(FYF denotes triple mutations α1F207A-βY231A+α1F207A) GlyR, 2 intact α1-α1 sites were sufficient for glycine activation, showing similar maximum current as the wild-type (4.5 ± 0.7 nA mutant v.s. 7.0 ± 0.9 nA wild-type, Figure 1C, D,G, Table 1 and Supplementary Figure 5G) and slightly increased EC₅₀ from ^~100 μM to ^~240 μM. Removing one more intact site (leaving only 1) as in α1:α1-β-α1 RFYF(α1R65D+F207A-βY231A-αF207A) rendered the channel inactive, while adding more sites (3 and 4 intact sites total) slightly decreased EC₅₀. This data demonstrates that a minimum of 2 intact sites are sufficient for heteromeric GlyR activation.

Table 1.

Hill fit parameters of glycine dose-response of wild-type and mutant GlyRs in figure1. See also Figure 1.

Constrcuts	EC50 (μM)	nh (Hill coefficent)	n (Cell)	Imax (nA)
α1:β	95 ± 5	2.5 ± 0.4	12	7.0 ± 0.9
α1:βY231A	138 ± 8	3.1 ± 0.5	19	5.0 ± 0.7
α1:βR86TY231A	206 ± 15	2.1 ± 0.3	14	7.1 ± 0.5
α1:α1-β-α1 FYF	267 ± 26	2.2 ± 0.4	8	4.5 ± 0.7
α1:α1-β-α1 RFYF	>1000	1.6 ± 2.0	8	0.8 ± 0.3
α1F207A:β	>1000	1.4 ± 1.5	10	0.8 ± 0.2
α1F207A:α1-β	>1000	1.0 ± 0.6	8	0.2 ± 0.1
α1F207A:α1-β-α1	>1000	1.8 ± 3.3	7	0.5 ± 0.1

Values are represented as mean ± S.E.M.

EC₅₀: medium effective concentration of glycine.

I_max: maximum glycine-induced current amplitude.

We further tested whether allosteric sites at α1-α1, α1-β or β-α1 subunit interfaces have the same contribution to activation (Figure 1E, F and Table 1). As expected, only 1 intact site at the β-α1 interface was insufficient for activation. Moreover, unlike at α1-α1, 2 intact sites at α1-β and β-α1 interfaces did not activate GlyR. Even the addition of one more α1-α1 site (3 in total) did not result in active GlyR. These recordings suggest that 2 α1-α1 allosteric sites are sufficient for activation and necessary for high efficacy gating, while α1-β or β-α1 sites are neither sufficient nor necessary, but only decrease EC₅₀. These findings are consistent with available structural data where activated states (open/desensitized) are always found with no less than 2 glycine bound in allosteric sites at the α-α interfaces.

Milliseconds to second open/close kinetics

Upon activation by glycine, heteromeric GlyRs transition from the closed resting state to the open conductive state(s), which in turn transforms to activated but non-conductive desensitized state(s)³⁰. The transition kinetics between these states directly influence total ion flux in transient physiological processes (such as action potentials) but have yet to be systematically characterized. We first determined the kinetics between the closed and open states (Figure 2).

Figure 2.

Open and close kinetic measurements of α1β GlyR at varied glycine concentrations. (A-D) Typical process of open, desensitization and close in response to the application and removal of (A) 40 μM, (B) 100 μM, (C) 200 μM and (D) 1000 μM glycine. (E-H) Open and close time courses (n = 5-13 cells) fitted with single-exponential with time constants (mean ± S.E.M.) shown for (E) 40 μM, (F) 100 μM, (G) 200 μM and (H) 1000μM glycine. (I) Fraction GlyR open calculated from open/close rates (blue, with propagated S.E.M.) approximates dose-response measurements (black). See also Figure S2 and Table 1.

The open/close kinetics of α1β GlyR roughly follows the laws of mass-action but deviates from simple second-order reaction. At lower glycine concentrations, the opening is slower with a larger time constant τ_open (^~0.6s at 40 μM glycine), which decreases approximately in proportion to increased glycine concentrations until reaching a minimum of 60 ms likely limited by instrument (Figure 2A–D). Although closing rate does not necessarily follow single-exponetial decay (please see Figure 5B and D), we fitted with single-exponential to show intuitively how it depends on glycine concentration. The close rate is fast with τ_close ^~ 0.3 s at 40 μM glycine, and showed a non-linear steep slowing down beyond 200 μM glycine to ^~1.1 s. The glycine concentration-dependent off rates deviate from simple second order reaction and is likely related to 2 glycine binding being sufficient and necessary for GlyR activation. Consistent with this, the fraction of open GlyR at thermal equilibrium as calculated from $\eta ={\tau }_{\text{close}}/\left({\tau }_{\text{close}}+{\tau }_{\text{open}}\right)$ very closely follows the dose-response curve obtained from glycine titration (Figure 2I). These measurements demonstrate that α1β GlyR quickly opens and closes following the binding and dissociation of glycine, which allows fast response to action potentials where glycine concentration is transiently elevated.

Desensitization rate is independent of glycine concentration

With sustained application of glycine, α1β GlyR quickly opens, followed by a decrease in conduction owing to transition into non-conductive desensitized state(s). We characterized the transition rate from open to desensitization under varying glycine concentrations (Figure 3A–D, Supplementary Figure 5F, See Supplementary Figure 2 for representative complete recordings). Clearly, desensitization followed a single exponential process with comparable time constants (^~ 5-8s) across all tested glycine concentrations spanning 40 μM to 10 mM (^~0.4 – 100 x EC₅₀). Such invariant kinetics clearly differs from open rate which scales with glycine concentration (Figure 2), suggesting that open to desensitized state transition is independent of the number bound glycine but an intrinsic property of activated α1β GlyR.

Figure 3.

Differential glycine-dependence of desensitization rate, extent, and its recovery. (A-D) GlyR α1β desensitization responses in the presence of glycine (A) 40 μM, (B) 100 μM, (C) 200 μM and (D) 1000 μM are shown. Typical time courses and best exponential fits are shown, with time constants τ listed (n = 5 – 13 cells, mean ± S.E.M.). (E, G, I) Representative recordings measuring the recovery after desensitization of GlyR α1β in the presence of (E) 40 μM, (G) 100 μM and (I) 1000 μM glycine. (F, H, J) Averaged desensitization-recovery curves from the sustained application of (F) 40 μM, (H) 100 μM and (J) 1000 μM glycine with exponential fits and time constants shown (n = 6-10 cells, mean ± S.E.M.). See also Figure S2, Figure S5 and Table 1.

Higher glycine concentration results in more desensitization and slower recovery

Although the rate of desensitization is unaffected, the extent of desensitization, as well as the recovery rates are directly determined by glycine concentrations. At 40 μM glycine, ^~66% of current remains after desensitization. This number decreases as glycine concentration increases, reaching ^~18% at 1 mM glycine and ^~3% at 10 mM glycine (Figure 3A–D and Supplementary Figure 2C). Coincidently, α1β GlyR recovers from desensitization faster when activated by lower glycine concentrations. The recovery time constant τ_rec at 40 μM glycine was ^~ 8 s but reached ^~ 57 s at 1 mM glycine (Figure 3E–J). This ^~ 7-fold increase in desensitization recovery time constant roughly accounts for the ^~ 4-fold increase in the extent of desensitization, as the open to desensitization rate stays constant.

Intracellular M3-M4 loops do not affect α1β GlyR gating by glycine

The large intracellular M3-M4 loops have been proposed to modulate Cys-loop receptor function in multiple cases, especially in occasions where post-translational modification on these loops clearly affect GlyR function likely under disease situations^5,8,31. However, some functional effects are conflicting in specific cases^32–35. We set out to characterize the functional effects of α1 and β M3-M4 loops without any intentional post-translational modifications through truncation (α1_s) and insertion of an irrelevant protein (GFP for β_sg, Figure 4A).

Figure 4.

Gating property comparisons between the wild-type and M3-M3 large intracellular domain deleted α1β GlyR. (A) Domain organizations of wildtype (α1wt and βwt) and truncated (α1s and βsg) subunits. SP, signal peptide; ECD, extracellular domain; TM, transmembrane; ICD, intracellular domain; GFP, green fluorescent protein. (B) Glycine response curves of α1wtβwt and α1sβsg (n = 12–17cells, mean ± S.E.M.), normalized to 1 mM glycine in each cell. (C-E) Desensitization rate measurements of truncated α1sβsg GlyR at (C) 40 μM, (D) 100 μM and (E) 1000 μM glycine, with typical exponential fits and time constants shown (n = 6 cells, mean ± S.E.M.). (F) Representative desensitization recovery recording of GlyR α1sβsg from sustained application of 1000 μM glycine. (G) Recovery rate of α1sβsg GlyR fitted to exponential with time constant shown (orange, n = 6-10, mean ± S.E.M.). Wild-type recovery is shown as dashed line for comparison. See also Figure S3, Figure S5 and Table 1.

Modifications of α1 and β M3-M4 loops did not show appreciable functional effects in the steady-state or kinetics of α1β GlyR activation by glycine. The apparent glycine affinities were indistinguishable with EC₅₀ ^~ 100 μM (Figure 4B). In addition, a similar invariant rate but glycine concentration-dependent extent during desensitization was observed (Figure 4C and Supplementary Figure 5H). Both wild-type and mutants had similar slow recovery rate with ^~ 50–60 s time constants when activated with 1 mM glycine (Figure 4D and Supplementary Figure 5G). Clearly, the modifications we made did not have functional effects on α1β GlyR in glycine activation. These findings suggest that the intracellular loops likely do not intrinsically modulate heteromeric GlyR function. Instead, post-translational modifications on these loops are more relevant.

A mechanistic model for α1β GlyR gating

The peculiar gating properties of α1β GlyR point to complicated and confounding gating mechanisms. For instance, how does β subunit contribute to gating as its allosteric sites are neither necessary nor essential for activation? How does the same glycine-activated ECD conformation result in distinct open and desensitized states? How does glycine concentration modulate the extent and recovery rate of desensitization without changing its forward rate? To address these questions, we propose a gating model that consolidates all the above gating properties with minimal parameterization.

The proposed model is shown in Figure 5A, with specific states in Figure 5B. The 3 α-α and 2 α-β / β-α sites allow 0-3 glycine binding on α-α sites and 0-2 glycine binding on α-β / β-α sites, making the total possible binding states 4 * 3 = 12. 6 out of these 12 states have less than 2 occupied α-α sites and should be in the closed state (Figure 1). The other 6 should have activated ECDs either in the open, or the desensitized state. This leads to a total of 6 closed states, 6 open states and 6 desensitized states. Following are the rationales to parameterize this model seeking to conform to current knowledge with minimal variables.

Figure 5.

A gating model of the α1β GlyR. (A) Schematic of the gating cycle. (B) Reaction schematic of the gating model, with rate constants shown. Glycine equilibrates quickly at allosteric sites at both α-α (association, dissociation rate constants: k_aα, k_dα) and α-β (k_aβ, k_dβ) interfaces with cooperativity index n_α and nβ and dominates close-open transitions. ECD activates when 2 or more α-α sites are occupied. Open TM spontaneously relaxes to desensitized TM with intrinsic constant rates ($(k_{\text{des}}=1/{\tau }_{\text{des}},k_{{\text{des}}_{-}\text{r}}=1/{\tau }_{\text{des}\_\text{r}})$), which scales (slows) kinetics by n_des fold. Please see main text and methods for determination of these parameters. (α)x(β)γ represents x glycine bound at α/α interfaces and y glycine bound at α/β or β/α interface. (C) Overlay of GlyR activation time-course predicted by model with experimental recordings at indicated glycine concentrations. Black and blue arrows denote the point of glycine application and washout, respectively. Green stars show max activity during glycine application. Glycine-dependent desensitization recovery at (D) 40 μM and (E) 1000 μM glycine concentrations (Black: experimental recordings, red: predicted by model). (F) Experimental (black, mean ± S.E.M.) and predicted glycine dose-response (red), with Hill fit to the predicted activity. Fitted parameters are shown. See also Figure S2.

Glycine binding and dissociation rates follow laws of mass-action (Figure 2). Since glycine binding pockets are nearly identical for α-α or α-β / β-α sites, the same association (k_aα = k_aβ) and dissociation (k_dα = k_dβ) rate constants are assumed. Since glycine-induced ECD conformation change is also highly similar across α and β subunits, the same contributions to binding cooperativity is assumed, in the form of scalers (n_a = n_b) of dissociation rate constants. The open and closed states are constantly under fast exchange as single channel recordings show millisecond time scale kinetics between these states ^16,22,36,37. Binding/dissociation is the rate-limiting step in population shift upon change of glycine concentration (Figure 2).

Desensitization rate being independent of glycine concentration (Figure 3A–D and Supplementary Figure 5F) suggests that the transition from open to desensitized state mainly involves structural rearrangement in the transmembrane but not the glycine-binding ECD. This is consistent with current structural data showing that the open and desensitized states share an identical ECD conformation ^{16,17,27,38–43} but differ in TMs. According to this, we assumed a single exponential decay from open to desensitized TM at constant rate $k_{\mathit{\text{des}}}=1/{\tau }_{\mathit{\text{des}}}$. For the same reason, the reverse (desensitized to open) transition should also be glycine-independent, and was modeled with constant rate $k_{\mathit{\text{des}}\_r}=1/{\tau }_{\mathit{\text{des}}\_r}$.

Without change in the desensitization rates, how higher glycine concentrations lead to more desensitization (Figure 3A–D) becomes puzzling. Available structures are incompatible with the notion that higher glycine occupancies favor desensitized state over the open state because: 1. the open and desensitized states share the same ECD (glycine-bound) conformation; 2. activated ECD stays in the same conformation regardless of how many glycine sites are occupied ^{16,17,27,38–43}. For this, we propose that the recovery of the desensitized state relies on the dissociation of glycine from the activated ECD, analogous to the recovery of inactivated Kv channels that require closing through repolarization^44–46. Since glycine is much more engulfed in the activated ECD conformation ^{16,17,27,38,43}, glycine binding/dissociate kinetics should be slower. We use a rate scaler, n_des to describe this.

Under the above simplified assumptions, this model contains 6 unique parameters in addition to glycine concentration as the independent variable, including association rate constants k_aα = k_aβ, dissociation rate constants k_dα = k_dβ, binding cooperativity scalers n_a = n_b, desensitization forward and reverse rate constants k_des and k_{des_r}, and rate scaler after desensitization n_des. These parameters are interdependent in this network and non-trivial to determine. Luckily, matching against full traces of recording that encompass open, desensitization, close, and desensitization recovery yielded the following value selections: k_aα = k_aβ = 0.04 ± 0.01, k_dα = k_dβ = 10 ± 2, n_a = n_b = 0.6 ± 0.1, k_des = 0.18 ± 0.05, k_{des_r} = 0.007 ± 0.002 and n_des = 0.006 ± 0.007.

This model closely follows α1β GlyR open, desensitization and close kinetics (Figure 5C). Lower concentration of glycine (Figure 5C left, ^~0.4 EC₅₀) leads to a slower open rate, less desensitization extent, and a fast close rate. Intermediate glycine (Figure 5C middle, ^~1 EC₅₀) increases open rate and the extend of desensitization, while decreasing close rate and not changing desensitization rate. Very high glycine (Figure 5C right, ^~ 100 EC₅₀) induces rapid open, slower close, and almost complete desensitization, also without changing desensitization rate (Supplementary Fig. 2C). The residual ^~ 3% activity arises from the very slow transition from desensitized to open state k_{des_r} The recovery from desensitization through glycine dissociation is faster after low glycine activation (Figure 5D), and dramatically slows down as glycine concentration increases (Figure 5E).

This model also recapitulates dose-response properties during glycine activation (Figure 5F). The max currents after application of glycine (green stars in Figure 5C) correlates very well with the fraction of open GlyR predicted by this model (Figure 5F red). A Hill fit to model-predicted activities yielded very similar EC₅₀ (99 ± 2 model v.s. 95 ± 5 μM experimental) and apparent Hill slope (2.0 ± 0.1 model v.s. 2.5 ± 0.4 experimental, see Table 1). The cooperativity between α-α and α-β / β-α allosteric sites plays an important role in achieve such similar dose response. n_a = n_b = 0.6 ≈ 1/1.7 implies each binding increases apparent affinity by roughly 1.7 folds. This is consistent with available structures information where all allosteric sites are always in the same conformation.

Discussion

Through systematic electrophysiology measurements and mathematical modeling, we quantitatively characterized the gating mechanism of α1β GlyR and identified underlying fundamental principles. We show that glycine binds cooperatively to all allosteric sites and activates ECD when 2 or more sites at α-α interfaces are bound. Activated ECD drives the TM into a semi-stable open conformation, which collapses spontaneously into a more stable and kinetically slow desensitized state that is compatible with the same activated ECD conformation. Slow glycine kinetics in the activated ECD conformation leads to the much slower (compared to open-close transition), and glycine concentration-dependent, desensitization recovery rate.

The contribution of individual allosteric sites in GlyR activation have been unclear due to the lack of such information in the 5-fold symmetric homomeric GlyR structures^38–43,47, and further confounded by the dispute over the α:β stoichiometry^23,24,48–⁵². Recently published heteromeric GlyR structures showed an invariable 4:1 α:β stoichiometry, and a minimal of 2 occupied α-α sites in the open or desensitized states^16–18, consistent with our finding of 2 α-α sites being necessary and sufficient for activation. The necessity of α-α sites is likely related to the asymmetric opening caused by the widening of α-α, but not α-β subunit interfaces¹⁸. In addition, 2 α-α sites out of 5 being sufficient means multiple ligand binding states and routes lead to GlyR activation, consistent with functional observations^36,47. Coincidentally, structures of other receptors in the Cys-loop family also show a 2-site activation pattern. For instances, in α1β3²¹ and α1β3γ2⁵³ GABA(A) receptors, only 2 GABA are bound at the β-α interfaces, while in nicotinic acetylcholine receptors 2 nicotine binding at the α-β interfaces activates the channel¹⁹. 2-site activation lowers the apparent Hill slope and likely benefits receptor response to lower ligand concentrations.

Given their high structural similarities^16,27,28, it is interesting that α-α and α-β/β-α sites contribute differently to activation. Mutation of a single α-α or α-β site increased EC₅₀ a similar extent (from ^~100 μM to ^~140 μM, Supplementary Fig.S4), indicating similar glycine affinities. The asymmetric opening mechanism²⁷ likely underly the differential contribution of α-α and α-β sites in GlyR activation: Opening of heteromeric GlyR involves widening of α-α, but not α-β subunit interface. Binding of glycine at α-α interface may exert some local effects, while glycine at α-β interface does not. Clearly more structural and mechanistic investigations are required for a definitive conclusion.

How ligand binding to the ECD drives GlyRs to both open and desensitized states have been puzzling, given that the ECD conformation in both states are identical. More profound desensitization at higher glycine concentrations seemingly points to a mechanism where lower occupancy leads to open and higher occupancy causes desensitization. However, this is inconsistent with ECD conformation being the same for both open and desensitized states – how would the TM know which state to be in when the ECD, messenger of glycine binding, is the same? Glycine concentration-independent desensitization rate further suggest that desensitization is a process intrinsic to TM.

Our model shows the activation of ECD drives the TM into a meta-stable open state, which then relaxes to a more stable desensitized state (k_des = 0.18 ≈ 1 / 5.5 s, see Figure 3A–D). Since the open to desensitization transition mainly involves TM, the rate is naturally unaffected by glycine concentrations (Figure 5C, Figure 3A–D, Figure 4C–E decay of solid lines). The more stable desensitized TM conformation locks the ECD in the activated state, from which the dissociation of glycine is much slower (Figure 5B, rate constants scaled by 1/n_des = 1/0.006 ≈ 170 folds). This leads to the much slower recovery from desensitization through glycine dissociation. Higher glycine concentrations occupy more sites, which require longer to dissociate and cause slower recovery (Figure 3E–J, Figure 4F, G and Figure 5D, E). Slower recovery increases the extent of desensitization, since the forward desensitization rate is constant. Direct desensitization to open transition is very slow (k_{des_r} = 0.007), contributing to ^~ 3–5% of total residual current, which is only prominent at very high glycine concentrations (Supplementary Figure 2C and Figure 5C right). At this point, our model faithfully recapitulated all the peculiar desensitization properties of GlyR.

The quantitative model we proposed here closely follows experimental behavior of α1β GlyR through the full gating cycle. We believe this model reflects fundamental principles underlying GlyR gating, as well as Cys-loop receptors in general (with receptor-specific parameters). Of course, due to intrinsic limitations of mutagenesis experiments (for example, mutations usually lower affinities instead of ideally abolish completely, see Supplementary Figure 5E), small variations are possible. In addition, we did not systematically test whether specific positions of allosteric sites, or combinations of them, have differential effects in activation, which might be important considering the asymmetric structures and gating mechanism recently proposed^16,17,27.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Weiwei Wang (Weiwei.Wang@UTSouthwestern.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Experimental model and study participant details

Mach1 competent E.coli and HEK293T cells were used in this study.

Method details

Glycine receptor constructs

The human glycine receptor β1 (NCBI: NP_001139512.1) and β (NCBI: NP_000815.1) sequence were amplified from cDNA clones (McDermott Center, UT Southwestern Medical center). The human glycine receptor β1 was subcloned in the pEG-Bacmam vector and glycine receptor β was subcloned in the pLVX-IRES-ZsGreen1 vector (Clonetech). We generated human α18Qβ, α18QGGS3, α18QGGSGGS3, α18QGGSGGS3Q3, α118Aβ, α119Aaβ concatemer construct by connecting human glycine receptor β1 and β (without signal peptide) with QQQQQQQQ, QQQQQQQQGGS,QQQQQQQQGGSGGS,QQQQQQQQGGSGGSQQQ,LHPGSIGGSGGSGRA T and LHPGSIGGSGGSGGSGRAT peptide(figure supplement 1C). The peptide was made by PCR overlap extension, β1 and each α1-β concatemer were co-expressed and their glycine EC₅₀ was measured by electrophysiology respectively. Results showed that α 1:α18QGGSGGSβ had a similar glycine EC₅₀ to wild type GlyR, so α18QGGSGGSβ was referred to α1-β. Therefore, 8QGGSGGS was also used to link α1, β(no signal peptide) and α1(no signal peptide) subunits to form the concatemer α1-β-α1 (as shown in supplement figure 1C). All α1 and β mutants was generated using site-directed mutagenesis. The α1s sequence was derived by substitution of M3/M4 loop (residues R316-P381) by GSSG peptide. For the βsg construct, residues N3344-N377 (M3/M4 loop) was removed from β and GGSSAAA-mEGFP-SGSGSG was inserted. GlyR α1β structure from Liu et al., glycine bound open, PDB ID code 8DN5. The sequence of the reading frame in all constructs was confirmed by Sanger sequencing of the full open frame by Eurofins Genomics.

HEK293T cell culture and transfection

HEK 293T cells (Purchased from ATCC) were grown at 37 °C, 5% CO2 incubator in DMEM (Gibco) containing with 10% (v/v) heat-inactivated fetal bovine serum (Corning), 100 units/ml of penicillin G, 100 μg/ml of streptomycin sulfate (Invitrogen). Cells were passaged after reaching around 90% confluence. The cells were passaged every 2–3 d. For expression, cells were plated on 35mm tissue culture dishes (Scientific Laboratory Supplies) with 0.2M/ml containing 2 ml of DMEM supplemented with 10% (v/v) FBS, and then transfected using Lipofectamine 3000 reagent (Invitrogen) with total 0.8μg plasmids /plate at 1α1:3β ratios coding for the above mentioned GlyRs. GFP fluorescence from β constructs was used to identify the cells expressing the heteromeric α1β GlyRs. Current recordings were conducted after 17-24h transfected at room temperature.

Electrophysiology

Glycine-induced currents were recorded in the whole-cell patch clamp configuration. The bath solution contained (in mM): 10 HEPES pH 7.4, 10 KCl, 125 NaCl, 2 MgCl₂, 1 CaCl₂ and 10 glucoses. The pipette solution contained (in mM): 10 HEPES pH 7.4,150 KCl, 5 NaCl, 2 MgCl₂,1 CaCl₂ and 5 EGTA. Experiments with varying external glycine concentrations were conducted. The resistance of borosilicate glass pipettes between 2^~7 MΩ. For current recorded, voltage held at −70 mV and a Digidata 1550B digitizer (Molecular Devices) was connected to an Axopatch 200B amplifier (Molecular Devices). Analog signals were filtered at 1 kHz and subsequently sampled at 20 kHz and stored on a computer running pClamp 10.5 software. We used local perfusion system purchased from Automate Scientific for agonist application (https://www.autom8.com/perfusion-systems-overview/valvelink8-2-controller/). The response time is 10 ms per manufacturer specification. In experiments based on GlyR being impermeable for phosphates and sulphates ^55,56, we determined a higher limit of τ = 57 ± 3 ms for solution exchange. For glycine EC₅₀ values calculation, Hill1 equation was used to fit the dose-response data and derive the EC₅₀ (k) and Hill coefficient (n). For glycine activation, we used $I=I_0+(I_{max}-I_0)\frac{x_n}{k^n+x^n}$, where I is current, I₀ is the basal current (spontaneous opening current and leak, very close to 0), Imax is the maximum current and x is glycine concentration. All start point is fixed at 0 during fit. Measurements were from 7-14 cells, average and S.E.M. values were calculated for each data point. Glycine EC₅₀ data analysis was performed by OriginPro2018 software (Origin Lab). For recovery tau time after desensitization, ExpAssoc1 equation was used to fit the time of GlyR from desensitization state to apo state data and derive the tau time. Measurements were from 5-10 cells, average and S.E.M. values were calculated for each data point. Data analysis of tau time from desensitization to apo state was performed by Origin software (Origin Lab). For tau time calculation of GlyR at open, desensitization and close states, exponential(standard) function was used to fit the trace of open, desensitization and close and calculate tau time. Measurements were from 5-8 cells, average and S.E.M. values were calculated for each data point. Data analysis of tau time was performed by pClamp 10.5 software.

Gating model and calculations

Differential equation system describing transition between states using rate constants shown in Figure 5B were resolved numerically using NDSolve and ParametricNDSolve in Mathematica (Wolfram Research). Parameters were determined by matching model prediction to experimental recordings upon application and removal of glycine, to recapitulation open, desensitization, close, and desensitization recovery at the same time. These processes at 40 μM (^~0.5 EC₅₀), 100 μM (^~1 EC₅₀) and 10 mM (^~100 EC₅₀) glycine concentrations were considered at the same time to encompass a wide range of ligand concentrations. Fitting was performed using FindFit in Mathematica, and errors were estimated from fitted parameters using different measurements to better reflect experimental errors. Plots were generated using Mathematica and OriginPro (OriginLab).

Quantification and statistical analysis

The results are presented as the mean ± S.E.M., where n represents cell number. Analytical statistics were performed using the SPSS (version 20) software package. Differences were considered statistically significant at *p < 0.05, **p < 0.01, ***p< 0.001.

Table 2.

Summary of channel kinetical and steady-state properties during glycine activation. See also Figure 2, Figure 3, Figure 4.

	Glycine(μM)	τopen(s)	τclose(s)	τdes(s)	τdes(s)	I des / Imax	n(cells)
	40	0.6 ± 0.3	0.3 ± 0.1	4.8 ± 0.9	7.7 ± 0.6	0.66 ± 0.04	5
	100	0.18 ± 0.03	0.3 ± 0.1	7.5 ± 1.2	15.7 ± 0.6	0.39 ± 0.04	5
α1wtβwt	200	0.07 ± 0.01	1.1 ± 0.1	7.8 ± 0.6		0.23 ± 0.03	8
	1000	0.06 ± 0.01	1.0 ± 0.1	6.5 ± 0.4	57 ± 8	0.18 ± 0.02	13
	40	0.9 ± 0.1	0.4 ± 0.1	5.4 ± 1.2		0.74 ± 0.05	5
α1sβsg	100	0.5 ± 0.1	0.5 ± 0.1	7.0 ± 1.3		0.38 ± 0.05	5
	1000	0.06 ± 0.01	0.9 ± 0.3	8.8 ± 1.8	49 ± 7	0.17 ± 0.03	5

Values are represented as mean ± S.E.M.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Mach1 Competent E.coil	ThermoFisher Scientific	Cat#C862003
Chemicals, peptides, and recombinant proteins
n-Dodecyl-β-D-Maltopyranoside	Anatrace	Cat# D310
Glycine	Fisher BioReagents	Cat# BP381
Lipofectamine 3000	Thermo Fisher	Cat# L3000008
DMEM, high glucose, GlutaMAX™ Supplement	GIBCO	Cat# 10566016
Fetal Bovine Serum	Corning	Cat# 11023001
Penicillin-streptomycin solution	HyClone	Cat# SV30010
0.25% Trypsin-EDTA(1X)	Corning	Cat# 18318005
Experimental models: Cell lines
HEK293T	ATCC	ATCC CRL-3216
Oligonucleotides
Custom DNA oligos, de-salted (Table S2)	This paper	N/A
Recombinant DNA
α1 construct and mutants	This paper	N/A
Βconstruct and mutants	This paper	N/A
α1s construct	This paper	N/A
βsg construct	This paper	N/A
α1-β concatemer and mutants	This paper	N/A
α1-β-α1 concatemer and mutants	This paper	N/A
Software and algorithms
OriginPro	OriginLab	https://www.originlab.com/
UCSF ChimeraX	Pettersen et al., 202154	https://www.rbvi.ucsf.edu/chimerax
pClamp10.5	Molecular Devices, LLC	https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite
Prism	GraphPad	https://www.graphpad.com/features

Highlights.

1. Glycine binding at 2 α1-α1 allosteric sites is sufficient and necessary for gating

2. Desensitization is dominated by intrinsic properties of the TM instead of ECD

3. Glycine dissociation underly steady-state desensitization extent and its recovery

4. Quantitative model recapitulating complete gating cycle