Kinetic and mutational analysis of Zn²⁺ modulation of recombinant human inhibitory glycine receptors

Abstract

1. The effects of Zn²⁺ on glycine receptor (GlyR) currents were analysed in Xenopus oocytes and human embryonic kidney cells expressing homomeric human wild-type and mutant α1 subunit GlyRs.

2. Low concentrations (10 μm) of extracellular Zn²⁺ converted the partial agonist taurine into a high-efficacy agonist. Concentration-response analysis showed that the EC₅₀ for taurine decreased whereas the Hill coefficient increased under these conditions. In contrast, 50–500 μm Zn²⁺ showed an increased EC₅₀ value and reduced maximal inducible taurine currents. The potency of competitive antagonists was not affected in the presence of Zn²⁺.

3. Single-channel recording from outside-out patches revealed different kinetics of glycine- and taurine-gated currents. With both agonists, Zn²⁺ altered the open probability of the α1 GlyR without changing its unitary conductance. Low Zn²⁺ concentrations (5 μm) increased both the opening frequency and mean burst duration, whereas higher Zn²⁺ concentrations (> 50 μm) reduced GlyR open probability mainly by decreasing the open frequency and the relative contribution of the longest burst of the single-channel events.

4. Site-directed mutagenesis of the GlyR α1 subunit identified aspartate 80 and threonine 112 as important determinants of Zn²⁺ potentiation and inhibition, respectively, without affecting potentiation by ethanol.

5. Our data support the view that Zn²⁺ modulates different steps of the receptor binding and gating cycle via specific allosteric high- and low-affinity binding sites in the extracellular N-terminal region of the GlyR α1 subunit.

Zn²⁺ is a naturally occurring metal ion that, in the brain, is highly abundant in small synaptic vesicles (Perez-Clausell & Danscher, 1985). Upon stimulation, Zn²⁺ is released into the synaptic cleft (Assaf & Chung, 1984) to reach estimated concentrations of several micromolar (Frederickson, 1989). Central effects of Zn²⁺ deficiency are often manifested in behavioural disorders, in particular lethargy and depression, indicating a crucial role of Zn²⁺ in neuronal signal processing (Cuajungco & Lees, 1997).

Many lines of evidence indicate that Zn²⁺ acts as a neuromodulator at postsynaptic receptors. These include the N-methyl-D-aspartate (NMDA), amino-3-hydroxy-5-methylisoxazol-propionic acid (AMPA) and kainate subtypes of excitatory glutamate receptors, ATP-gated ion channels and inhibitory ionotropic GABA receptors (Smart et al. 1994). Analysis of the blocking effects of Zn²⁺ at the NMDA receptor have revealed both voltage-dependent and -independent mechanisms, which result from channel block and decreases in the frequency of channel opening, respectively (see Smart et al. 1994). The Zn²⁺ sensitivity of GABA_A receptors has been shown to depend on their subunit composition (see Smart et al. 1994). Histidine residues important for the inhibition of GABA_A and GABA_C receptors by Zn²⁺ have been identified on the α6 (Fisher & MacDonald, 1998), β1 (Wooltorton et al. 1997; Horenstein & Akabas, 1998) and ρ1 (Wang et al. 1995) subunits.

In previous reports, we and others have shown that Zn²⁺ also affects the function of the strychnine-sensitive glycine receptor (GlyR), another inhibitory ligand-gated ion channel (Bloomenthal et al. 1994; Laube et al. 1995a; Lynch et al. 1998). The GlyR is abundantly expressed in spinal cord and brainstem and mediates postsynaptic inhibition in many sensory and motor systems (Betz, 1992). Upon extracellular application, Zn²⁺ has a biphasic modulatory effect on glycine-gated chloride currents in cultivated embryonic spinal neurons and at recombinant GlyRs (Bloomenthal et al. 1994; Laube et al. 1995a). Low micromolar concentrations of Zn²⁺ potentiate glycine-gated chloride currents, whereas higher Zn²⁺ concentrations block glycine-induced chloride flux. The analysis of chimeric GlyR subunits indicates that the positive and negative effects of Zn²⁺ are exerted via different domains of the GlyR α1 subunit (Laube et al. 1995a). Although the glutamatergic mossy fibre axons in the hippocampus represent the most thoroughly studied Zn²⁺-containing neuronal system (Frederickson, 1989), there is histological evidence that Zn²⁺-accumulating neurons may also be localized in the dorsal spinal cord (Danscher, 1982), i.e. the region of highest GlyR expression in the central nervous system (Zarbin et al. 1981).

Since Zn²⁺ affects GlyR currents by changing the EC₅₀ of glycine (Laube et al. 1995a), we wished to investigate the underlying mechanisms. Here, we show that Zn²⁺ has no effect on the single-channel conductance of the α1 GlyR but significantly alters its open probability. Potentiation results from increases in both open frequency and burst duration of glycine-gated channels. In contrast, Zn²⁺ inhibition appears to be mainly due to a reduced channel opening frequency. Notably, Zn²⁺ also converts the partial agonist taurine into a high-efficacy activating ligand. Using site-directed mutagenesis, we identified two amino acid residues in the N-terminal extracellular domain of the GlyR α1 subunit which are crucial for the modulatory effects of Zn²⁺. Our data can be summarized in a schematic model of GlyR activation, where Zn²⁺ acts at different steps of the receptor binding and gating cycle.

METHODS

cRNA synthesis, mutagenesis and oocyte expression

Linearized plasmid DNAs were used for in vitro synthesis of cRNA (Laube & Betz, 1998) using the mCAP mRNA capping kit (Stratagene, La Jolla, CA, USA). Mutants were constructed using oligonucleotide-directed mutagenesis (QuikChange site-directed mutagenesis Kit, Stratagene) as described (Laube et al. 1997). About 50 nl of each cRNA (50–200 ng ml⁻¹) was injected into Xenopus laevis oocytes as detailed previously (Laube & Betz, 1998). For the isolation of stage V and VI oocytes, ovarian lobes were surgically removed from adult female frogs anaesthetized by immersion in 1% urethane (Sigma) by making a 1–2 cm incision in the abdomen. Following the ovariectomy, frogs were terminally anaesthetized and killed by decapitation. All protocols were approved by the local animal care and use committee (II25.3-19c20/15; RP Darmstadt, Germany). The ovarian lobes were then mechanically separated and incubated with collagenase Type IIA (Sigma) for 1–3 h to enzymatically dissociate the oocytes. Before injection, the layer of follicle cells surrounding the isolated oocytes was removed using forceps, and oocytes were maintained in Barth's medium (composition (mM): 88 NaCl, 1 KCl, 0.4 CaCl₂, 0.3 Ca(NO₃)₂, 8.2 MgSO₄, 2.4 NaHCO₃, 10 Tris, pH 7.2). Voltage clamp recording from injected oocytes was performed at a holding potential of −70 mV (Laube et al. 1997) in frog Ringer solution containing (mM): 115 NaCl, 1 KCl, 1.8 CaCl₂ and 10 Hepes (pH 7.2). Effects of Zn²⁺ on glycine-induced currents were analysed after superfusing the oocytes with Zn²⁺ for 5 s prior to and during agonist application. Dose-response curves of agonist-induced peak currents (I) were normalized to the maximal current value (I_max) obtained and fitted with the sigmoidal Hill equation I/I_max=cⁿ/(cⁿ+ EC₅₀ⁿ) using a Gauss-Marquardt iteration, where I/I_max represents normalized current, c glycine concentration, EC₅₀ the glycine concentration resulting in a half-maximal response, and n the Hill coefficient. Dose-response curves of Zn²⁺-modulated currents were determined in the presence of glycine concentrations eliciting a response corresponding to 25 % of the maximal inducible current (EC₂₅ value) and normalized to the current obtained in the absence of Zn²⁺. Data were fitted to the following equation:

(1)

using a Gauss-Marquardt iteration, where I/I_max represents normalized current, m₁ the maximal potentiation, c the Zn²⁺ concentration, EC₅₀ and IC₅₀ the 50% effective Zn²⁺ concentration for potentiation and inhibition, respectively, and n₁ and n₂ the corresponding Hill coefficients.

Cell transfection and electrophysiological recording

Human embryonic kidney (HEK) 293 cells (ATCC CRL 1573) were transfected (Laube & Betz, 1998) with the human α1 subunit cDNA as well as mutants thereof. All cDNAs were inserted into the mammalian expression vectors pCIS2 (Gormann et al. 1990) or pRc/CMV (Invitrogen, San Diego, CA, USA). After 2 days, the transfected HEK 293 cells were viewed with an inverted microscope (Zeiss, Deissenhofen, Germany) and continuously perfused (0.5 ml min⁻¹) at room temperature (21–25°C) with Ringer solution containing (mM): 137 NaCl, 5.4 KCl, 1.8 CaCl₂, 1 MgCl₂ and 5 Hepes (pH 7.4). Patch pipettes contained (mM): 120 CsCl, 20 tetraethylammonium chloride, 1 CaCl₂, 2 MgCl₂, 11 EGTA and 10 Hepes (pH 7.2). Membrane currents were recorded from whole cells and cell-free membrane patches (Hamill et al. 1981) using an EPC-9 amplifier (HEKA Elektronik, Lambrecht, Germany). Whole-cell currents were sampled at 20 Hz and stored on disk. Single-channel recordings were stored on digital audiotape (DTR 1204; Biologic, Claix, France). The membrane potential was clamped at −70 mV in all experiments. Glycine was applied with the bath solution using a microcapillary application system (DAD-12; Adams and List, Westbury, NY, USA). All drugs were purchased from Sigma (Deissenhofen, Germany).

Single-channel analysis

Single-channel data were digitized at 10 kHz and analysed after low-pass filtering at 2.3 kHz (-3 dB) as described in Langosch et al. (1994). The shortest detectable single-channel event was 70 μs. Channel openings were identified using the 50% threshold-crossing criterion, and half-amplitude threshold analysis was done with a semiautomatic procedure using a program for ‘threshold analysis of continuous’ (TAC) single-channel records of the M2-LAB software package (Instrutech Corporation, Elmont, NY, USA) which forms part of the EPC9-program of the HEKA software (Lambrecht, Germany). Only the most frequent conductance state (main state) was analysed. Amplitude histograms were constructed from digitized traces and fitted with Gaussian curves using a non-linear least-squares routine (Langosch et al. 1994). The Gaussian fit provides an estimate of mean current amplitude.

The total open probability (NP_o; the product of number of channels and open probability times) was determined using the TAC program. Open probabilities were calculated as the sum of each dwell time multiplied by the level number and divided by the total recording time. The number of channels in the patch was estimated as being greater than or equal to the maximal number of simultaneous openings observed during recording. Values of NP_o were computed from 2 min recordings.

Distributions of the open period and closed time durations were displayed using the square root of the number of events per bin and fitted with a mixture of exponential densities by the method of maximum likelihood to define the respective time constants τ_i (ms) and areas a_i (%) (Colquhoun & Sigworth, 1995). The overall mean open time was calculated by the sum of the product of individual dwell times multiplied by the level numbers, divided by the number of opening transitions. Simultaneous channel openings were discarded from the analysis. To obtain estimates of channel burst duration, closed time distributions were created for each record using the TAC program. Histograms were fitted with three to four well-separated areas and time constants using the method of maximum likelihood. Visual inspection of the single-channel events revealed that only the shortest component of the closed time distribution (τ₁) seemed to be closures within bursts; thus only the first closed time population was considered within bursts and was therefore assumed to represent closures within a burst. In contrast, only bursts characterized by the longest time constant seemed to correspond to groups of successive openings, whereas the remaining burst populations always appeared to be isolated openings. Using the method of equal percentage of misclassified bursts (Colquhoun & Sigworth, 1995), values of 0.8-3 ms were used as a critical gap length (t_crit) to define bursts, such that successive openings that were separated by closures < t_crit were classified as closures within a single burst, whereas openings separated by closures >t_crit were assigned to discrete bursts. The constructed burst durations were used to define the total open time per burst as described for open times. Burst durations were binned logarithmically (bin width 1/10 of a log unit) and fitted with multiple exponential functions (Sigworth & Sine, 1987). From the ratio of the longest burst time constant to the shortest closed time constants, the number of closures per burst was calculated.

For evaluating the distinct microscopic effects of Zn²⁺, a model for receptor gating according to del Castillo & Katz (1957) was used, in which three agonist molecules must bind to the receptor for channel opening (see Fig. 5; see also Twyman & Macdonald, 1991 and Lewis et al. 1998). The rate constants k_{plus; 1} and k_-1 govern agonist binding and unbinding, whereas β and α control the actual opening and closing steps. The rate constants β, α and k_-1 for each of the steps in the activation process were estimated from analysing transitions within bursts (Colquhoun & Hawkes, 1981) using the maximum likelihood method:

(2)

(3)

(4)

where n_g is the mean number of shut periods per burst, t_g the length of gaps and t_b the length of bursts.

Figure 5. Model of the mechanisms and sites of Zn²⁺ action at the α1 subunit GlyR.

A, reaction scheme for GlyR gating indicating the steps at which Zn²⁺ affects glycine currents. In this scheme, three agonist binding sites assumed to display identical binding affinities (K; binding) have to be occupied to achieve a liganded state from which channel gating (E; gating) can occur with high probability (see Discussion). Accordingly, (1) potentiating Zn²⁺ concentrations increase agonist occupation by mainly decreasing the unbinding constant k_-1; (2) inhibitory concentrations of Zn²⁺ drastically decrease gating efficacy, thus generating mixed surmountable/non-competitive Zn²⁺ antagonism. A denotes the agonist, R the unoccupied receptor, A_nR the n-liganded closed channel states, and A₃R* the liganded open-channel state of the GlyR. The rate constants k_{+ 1} and k_-1 govern agonist binding and unbinding, whereas β and α control the actual opening and closing steps. B, schematic model of an α1 subunit GlyR illustrating the putative sites of Zn²⁺ action delineated above. A high-affinity Zn²⁺ site including the aspartate 80 residue (D80) is proposed to mediate the potentiating effect of Zn²⁺ that results from an increased receptor occupancy (1, illustrated in the left subunit). The inhibitory low-affinity site for Zn²⁺ is suggested to include threonine 112 (T112), i.e. a residue crucial for Zn²⁺ inhibition (illustrated in the right subunit). This site is thought to have a direct effect on the channel gating process (2, arrow). + and – indicate positive and negative modulatory effects of Zn²⁺ on GlyR activation. Gly, glycine. For clarity, only two α1 subunits are shown in detail, although the α1 GlyR is composed of five subunits (Kuhse et al. 1993).

Results represent means ±s.d. unless indicated otherwise. The significance of the data was evaluated using Student's paired t test and considered to be statistically significant at P < 0.05.

RESULTS

To gain insight into the mechanisms of GlyR modulation by Zn²⁺, we generated homo-oligomeric GlyRs by heterologous expression of human wild-type and mutant α1 subunits and analysed their agonist responses and single-channel properties in the presence of different Zn²⁺ concentrations.

Zn²⁺ converts taurine from a partial into a full GlyR agonist

Both potentiation and inhibition of GlyR currents by Zn²⁺ are fully reversible and also seen upon application of other glycinergic agonists, e.g. β-alanine and taurine (Laube et al. 1995a). Taurine is a partial agonist of the GlyR, which produces maximal current responses that are significantly lower than those obtained with glycine (Laube et al. 1995b). Here, superfusion of α1 subunit-expressing oocytes with taurine (10 mM) evoked maximal membrane currents (I_max) with an amplitude of 42 ± 13% of that obtained at saturating glycine concentrations; half-maximal activation was seen at a concentration (EC₅₀) of 0.92 ± 0.15 mM with a corresponding Hill coefficient of 1.66 ± 0.2. Addition of Zn²⁺ at concentrations up to 10 μm (Fig. 1a) produced a significant decrease in the EC₅₀ of taurine (0.42 ± 0.1 mM; P < 0.05), whereas a concentration of 100 μm Zn²⁺ (EC₅₀ for taurine: 2.7 ± 0.2 mM; P < 0.05; Fig. 1a) resulted in the opposite effect. Both Hill slopes were altered significantly (P < 0.05) with values of 2.2 and 1.2, respectively, compared with 1.66 in the absence of Zn²⁺, suggesting that Zn²⁺ might affect the cooperativity between the GlyR agonist binding sites (Laube et al. 1995a). These results are consistent with previously described changes in glycine responses (Bloomenthal et al. 1994; Laube et al. 1995a).

Figure 1. Effect of different concentrations of Zn²⁺ on agonist and antagonist responses of the recombinant α1 subunit GlyR.

A, taurine dose-response curves in the presence of Zn²⁺. Oocytes expressing the α1 subunit were superfused with the indicated concentrations of taurine in the absence (•) and presence of 10 μm (▪) and 100 μm (♦) Zn²⁺. The respective EC₅₀ values, with Hill coefficients in parentheses, are 920 ± 150 μm (1.66 ± 0.2), 420 ± 100 μm (2.22 ± 0.3) and 2700 ± 200 μm (1.2 ± 0.2). B, maximal inducible currents of glycine (5 mM, •) and taurine (20 mM, ▪) in the presence of different Zn²⁺ concentrations. Data are plotted as a fraction of the maximal glycine current in the absence of Zn²⁺. Note biphasic dose dependence on Zn²⁺ for taurine currents compared with glycine-induced currents. C, strychnine antagonism of glycine currents in the absence (•) and presence of 1 μm (▪) and 100 μm (♦) Zn²⁺, determined at glycine concentrations corresponding to the EC₅₀ value of the respective Zn²⁺ concentration. The respective IC₅₀ values are 35 ± 2.5, 43 ± 5 and 39 ± 4 nM. Data are given as current relative to the glycine control response. Smooth curves represent least-squares fits to the Hill equation. Data are normalized to maximal currents obtained in the absence of antagonist and represent the means ±s.d. of at least three independent experiments.

A remarkable difference in the modulation of glycine- and taurine-gated currents became apparent, however, when comparing relative I_max values. In contrast to glycine, where Zn²⁺ up to 100 μm did not significantly alter maximal glycine-inducible whole-cell currents (Laube et al. 1995a; and Fig. 1B), the I_max value of taurine was increased in the presence of low concentrations of Zn²⁺ (Fig. 1B). At 1 μm and 10 μm Zn²⁺ maximal taurine responses were considerably higher than those in the absence of the metal ion (relative maximal amplitudes of 59 ± 9 and 81 ± 13%, respectively). Zn²⁺ antagonism was overcome by increasing agonist concentrations at Zn²⁺ concentrations less than 100 μm (Laube et al. 1995a; and Fig. 1a). However, for Zn²⁺ concentrations > 100 μm, inhibition was only partially relieved by high concentrations of both glycine and taurine (Fig. 1B). Again, Zn²⁺ showed a more dramatic effect on the I_max values elicited by taurine than on those obtained with glycine (Fig. 1B). In contrast, the potency of the competitive antagonist strychnine for inhibiting agonist currents at the corresponding EC₅₀ values of glycine was not affected at these Zn²⁺ concentrations (Fig. 1C).

Zn²⁺ binding to proteins often involves histidine residues (Vallee & Falschuk, 1993). This implies a strong pH dependence of Zn²⁺ binding. For GABA_A receptors, the inhibitory effect of Zn²⁺ has been shown to be sensitive to changes in extracellular pH (Smart & Constanti, 1982). To evaluate a putative interaction of Zn²⁺ with protonatable amino acid side chains, we determined the pH dependence of Zn²⁺ action on α1 subunit GlyRs. At a pH of 5.6, the current activated by 200 μm glycine was significantly reduced (> 20%) in comparison to neutral conditions (not shown). Acidification, by reducing the pH range from 7.2 to 5.6, shifted the glycine concentration-response curve to the right, thus increasing the EC₅₀ value for glycine without significantly altering the glycine I_max. Under the same conditions, the concentration-response profile of Zn²⁺ potentiation was not altered (not shown), whereas the inhibitory potency of Zn²⁺ was reduced at acidic pH (IC₅₀ 25 μm at pH 7.2, 283 μm at pH 5.6). This change might be attributed to the protonation of histidine residues, suggesting that histidine side chains contribute to GlyR inhibition but not potentiation by Zn²⁺. Alternatively, this pH effect could reflect partial complexation of Zn²⁺ by glycine (see below).

Zn²⁺ alters the probability of GlyR channel opening

To unravel the voltage-dependent and -independent components of Zn²⁺ modulation, we analysed glycine- and taurine-induced whole-cell currents and single-channel events in the absence and presence of different concentrations of Zn²⁺. Figure 2a and D show channels recorded from an outside-out patch excised from a HEK 293 cell transfected with the α1 subunit after application of 3 μm glycine and 30 μm taurine, respectively. No channels were detected in the patch before agonist application. Inspection of current traces as shown in Fig. 2a and D suggested that glycine and taurine induce a different gating behaviour of α1 subunit GlyRs. Whereas single-channel currents recorded in the presence of 3 μm glycine showed a burst-like appearance separated by longer intervals, single-channel records in the presence of 30 μm taurine revealed largely isolated opening events (compare Fig. 2a and D). The most prominent effect of Zn²⁺ on both glycine- and taurine-gated currents was an alteration in the probability of channel opening (Fig. 2B, C and E), with 5 μm Zn²⁺ causing an increase, and 50 μm Zn²⁺ a decrease, in the apparent open probability of GlyR channels. To substantiate this observation, we quantified the total open probability measured after exposure to various concentrations of Zn²⁺. A plot of the relative values of the total open probability vs. concentration obtained from the comparison of different cells (see also legend of Fig. 2) confirmed that Zn²⁺ indeed causes significant concentration-dependent changes in apparent open probability with both glycine and taurine (Fig. 2F; P < 0.05). Together, our data indicate that 5 μm Zn²⁺ increases the open probability of GlyR channels by about 2.5- to 5-fold, whereas 50 μm Zn²⁺ decreases it by about 70–90%.

Figure 2. Single-channel recordings of glycine and taurine currents in the absence and presence of Zn²⁺.

Outside-out patch recordings from a HEK 293 cell expressing the α1 subunit in the absence (A and D) and presence of 5 μm (B and E) and 50 μm (C) Zn²⁺ (for the respective gating characteristics see Fig. 3). The membrane potential was clamped at −70 mV. Traces represent single-channel inward currents induced by 3 μm glycine (A–C) and 30 μm taurine (D and E). F, normalized values of the total open probability for glycine (▪) and taurine (□) are plotted for the indicated Zn²⁺ concentrations. G, potentiation and inhibition of GlyR currents by Zn²⁺ are voltage independent. Current-voltage ramps of glycine-elicited currents recorded from a HEK 293 cell transfected with α1 cDNA before (a) and after application of 5 μm (b) or 50 μm (c) Zn²⁺. The cell was dialysed with standard intracellular solution resulting in Cl⁻ concentrations of 145 mM on both sides of the membrane. H-J, conductance states evoked by 3 μm glycine in the absence (H) and presence of 5 μm (I) or 50 μm (J) Zn²⁺ for the patch recorded from in A-C. In the absence of Zn²⁺, conductance states of 5.9 pA (77% of all events), 4.9 pA (9%), 3.6 pA (10%) and 1.5 pA (3%) were seen (H). In the presence of 5 μm (I) or 50 μm (J) Zn²⁺, the respective conductance states were 6.2 pA (83%), 4.7 pA (10%), and 3.2 pA (7%), or 5.8 pA (80%), 5.4 pA (13%) and 4.2 pA (7%), respectively.

To reveal possible channel-blocking effects of higher Zn²⁺ concentrations that could account for the non-competitive Zn²⁺ inhibition (see Fig. 1B), we analysed the current- voltage relationship of glycine-gated currents in the presence of 5, 50 (Fig. 2G) and 500 μm Zn²⁺ (not shown). The overall shape of the current-voltage relations of the α1 subunit GlyR was unaffected by the presence of Zn²⁺, irrespective of whether potentiating or inhibiting concentrations were used. This indicates a voltage-independent mechanism of Zn²⁺ inhibition and suggests that the effects of potentiating and inhibitory concentrations of Zn²⁺ on whole-cell glycine- and taurine-activated currents are a consequence of alterations in gating behaviour or ligand binding affinity rather than of changes in channel conductance.

Further analysis of the single-channel properties of the GlyR resulted in a more detailed picture of Zn²⁺ action. Amplitude histograms constructed from single-channel recordings obtained at different Zn²⁺ concentrations were fitted to the sum of multiple Gaussian functions to estimate channel amplitudes (Fig. 2H–J). At −70 mV and identical extra- and intracellular Cl⁻ concentrations, we observed glycine- and taurine-activated elementary inward currents with a predominant amplitude of about 6 pA, indicating that glycine and taurine activate similar single-channel conductances. Linear regression analysis of the slope of the main conductance amplitude for recordings obtained from control cells and from cells exposed to 5 and 50 μm Zn²⁺ revealed no apparent effect of Zn²⁺ on glycine-induced single-channel conductance. Averaged main-state conductances were about 85–90 pS (> 80%), with some smaller and less frequent conductances seen in both the absence and presence of Zn²⁺ (see for example Fig. 2H). For the patch shown in Fig. 2A–C, the relative frequency of each conductance state was not altered significantly. In the absence and presence of 5 and 50 μm Zn²⁺, similar conductance distributions were found (see legend of Fig. 2), corroborating that Zn²⁺ had no apparent effect on single-channel conductances and relative frequencies.

To examine the action of Zn²⁺ on glycine-gated channel openings, we analysed our channel records in greater detail (Fig. 3; Table 1). For the patch shown in Fig. 2A–C, glycine alone activated channels with a mean open time of 2.01 ms; coapplication of 5 μm Zn²⁺ caused an increase, and of 50 μm Zn²⁺ a decrease, of this value to 2.28 and 1.51 ms, respectively. When comparing the results of three different patches, the mean open times determined in the presence of 5 μm (3.44 ± 1.41 ms) and 50 μm Zn²⁺ (1.25 ± 0.7 ms) were significantly different from each other (P < 0.05), but were not statistically different from those obtained in the absence of Zn²⁺ (2.16 ± 1.0 ms) (Fig. 3J). The distribution of closed times as binned in 0.2 ms intervals in the presence of 3 μm glycine could be fitted by four exponentials using the closed times listed in the insets to Fig. 3A, D and G (see for comparison Twyman & Macdonald, 1991; Takahashi et al. 1992; Lewis et al. 1998). The fastest component (τ₁, time constant of about 0.3-0.4 ms) always comprised the largest fraction of the total number of closures and presumably reflects the average duration of closings within bursts (Fig. 3A; Table 1). This component varied little between different patches and in the absence or presence of Zn²⁺, in contrast to the three slower components that showed more variability (Fig. 3; Table 1). Since the probability of opening increased in the presence of 5 μm Zn²⁺ (Fig. 2F), we compared the relative contribution of the closed time periods in the absence and presence of different Zn²⁺ concentrations. Zn²⁺ (5 μm) decreased the longest closed time constant (τ₄) and its relative area presumably as a consequence of an increased number of openings (Fig. 3D; Table 1). Thus, the increased open probability in the presence of Zn²⁺ might be attributed to the decrease in the longest component of the closed time constant. In contrast, higher concentrations of Zn²⁺ (50 μm) increased the longest closed time constant (τ₄) and the relative area (a₄) as a consequence of a reduced number of openings (Fig. 3G; Table 1). We also compiled open time histograms to examine whether the changes in open probabilities might be due to a shift in the open state duration of the channel. Open-time histograms were constructed from data obtained before and after exposure of the patches to 5 and 50 μm Zn²⁺, respectively (Fig. 3B, E and H). The mean time constants (τ₁, τ₂ and τ₃) for each component obtained by exponential fitting procedure were not significantly different in the absence and presence of Zn²⁺ (see Table 1). A comparison of the relative areas of each time constant in the histogram, which represent a measure of the relative frequency of openings contributed by each channel component, showed no major changes in respective relative areas in the presence of Zn²⁺ (Table 1).

Figure 3. Gating characteristics of the α1 subunit GlyR in the absence and presence of Zn²⁺.

A-I, histograms of closed time (A, D and G), open time (B, E and H), and burst duration (C, F and I) distributions, constructed from a total of 2377 (A and B) and 930 (C) glycine-induced openings in the absence (mean open time 2.01 ms, open probability 0.046), 6177 (D and E) and 2321 (F) openings in the presence of 5 μm Zn²⁺ (mean open time 2.28 ms, open probability 0.11), 488 (G and H) and 266 (I) openings in the presence of 50 μm Zn²⁺ (mean open time 1.51 ms, open probability 0.0067), respectively. Abscissae are binned logarithmically (bin width 1/10 of a log unit), and ordinates are scaled to display N, the square root of the number of events per bin. All three distributions were fitted with multiple exponentials (continuous curves). The estimated parameters (time constants τ (ms) and relative areas (%)) are shown in each panel. J and K, kinetic properties of glycine-evoked single-channel currents in the absence and presence of Zn²⁺. Absolute mean open and burst time durations of α1 homo-oligomeric GlyRs (J). Relative weights of the three different components of the burst distribution calculated for GlyRs in the absence and presence of different concentrations of Zn²⁺ (K). Note that τ₃ is affected by low, and τ₂ and τ₃ by high concentrations of the metal ion.

Table 1.

Kinetic analysis of glycine- and taurine-gated α1 GlyR currents in the absence and presence of Zn²⁺

	Glycine-gated
	α1			α1D80G		Taurine-gated α1
	Control	+5 μm Zn2+	+50 μm Zn2+	Control	+5 μm Zn2+	Control	+5 μm Zn2+
Closed times
τ1 (ms)	0.30 ± 0.09	0.34 ± 0.07	0.37 ± 0.1	0.37 ± 0.05	0.4 ± 0.1	0.65 ± 0.47	0.55 ± 0.25
(%)	61 ± 9	72 ± 10	52 ± 7	54 ± 14	72 ± 10	29 ± 4	47 ± 5
τ2 (ms)	2.4 ± 1.3	3.2 ± 1.3	2.7 ± 1.9	1.9 ± 0.9	2.5 ± 1.4	120 ± 105	65 ± 54
(%)	17 ± 11	10 ± 3	9 ± 2	23 ± 5	19 ± 3	62 ± 5	30 ± 7
τ3 (ms)	21 ± 15	31 ± 15	30 ± 25	20 ± 8.8	23 ± 22	800 ± 657	400 ± 287
(%)	3 ± 2	4 ± 5	2 ± 2	9 ± 2	6 ± 5	9 ± 7	23 ± 4
τ4 (ms)	250 ± 265	150 ± 154	600 ± 451	70 ± 82	90 ± 85	n.d.	n.d.
(%)	19 ± 7	14 ± 6	37 ± 11	14 ± 6	3 ± 3	n.d.	n.d.
Open times
τ1 (ms)	0.66 ± 0.14	0.60 ± 0.2	0.68 ± 0.3	0.45 ± 0.29	0.51 ± 0.09	0.70 ± 0.34	0.62 ± 0.3
(%)	68 ± 7	60 ± 11	83 ± 14	74 ± 8	82 ± 6	92 ± 7	41 ± 11
τ2 (ms)	3.6 ± 1.8	5.0 ± 0.9	2.9 ± 2.8	1.1 ± 0.5	2.4 ± 0.8	3.0 ± 2.9	3.2 ± 3.1
(%)	29 ± 8	36 ± 9	15 ± 6	23 ± 5	15 ± 4	8 ± 7	59 ± 11
τ3 (ms)	22 ± 17	32 ± 29	15 ± 19	8.8 ± 1.7	11 ± 2.2	n.d.	n.d.
(%)	3 ± 4	4 ± 5	2 ± 4	3 ± 3	3 ± 2	n.d.	n.d.
Burst duration
τ1 (ms)	0.46 ± 0.14	0.45 ± 0.25	0.41 ± 0.3	0.39 ± 0.19	0.40 ± 0.1	0.56 ± 0.24	0.54 ± 0.3
(%)	64 ± 19	46 ± 10	62 ± 9	54 ± 10	52 ± 4	82 ± 8	57 ± 13
τ2 (ms)	3.7 ± 1.8	5.7 ± 0.9	3.3 ± 2.8	4.5 ± 1.2	4.2 ± 1.4	4.2 ± 3.6	10.4 ± 9.8
(%)	24 ± 8	35 ± 2	35 ± 13	33 ± 9	32 ± 5	18 ± 8	43 ± 13
τ3 (ms)	27 ± 17	87 ± 29	40 ± 19	20 ± 18	23 ± 26	n.d.	84*
(%)	12 ± 8	19 ± 6	3 ± 5	13 ± 3	16 ± 3	n.d.	2

Time constants τ (ms) and their relative areas (%) obtained from fits of closed time, open time and burst length distributions at a glycine concentration of 3 μm for the α1 wild-type and α1^D80G mutant are given. The taurine concentration used was 30 μm. Numbers represent means ± s.d. of at least three determinations. n.d., not determined.

^*This time constant was only seen in one patch out of three.

When analysing the single-channel events recorded in the presence of 3 μm glycine from the patch shown in Fig. 2A–C, bursts were fitted best when assuming three different components with time constants of 0.6 (τ₁), 4.6 (τ₂) and 38 (τ₃) ms. The time constants τ₁ and τ₂ did not change in the presence of 5 μm Zn²⁺. However, in all patches analysed a significant increase in the relative contribution of the longest time constant τ₃ was observed (Fig. 3K; Table 1), resulting in a significant prolongation in the mean burst time (16.5 ± 4.4 vs. 4.45 ± 2.01 ms; Fig. 3J). We therefore propose that the increase in mean burst duration may be largely attributed to both an increase in the relative contribution of the longest component (τ₃) and a corresponding decrease in τ₁ and τ₂ of the short burst components (Fig. 3K). In contrast, 50 μm Zn²⁺ produced a significant reduction of the relative area of τ₃ and of the relative current carried by the longest bursts (Fig. 3K; Table 1). Thus, higher concentrations of Zn²⁺ decreased the relative contribution of the longest burst duration, causing a decrease in mean burst duration (2.65 ± 1.3 ms; Fig. 3J). In conclusion, in addition to enhancing channel opening frequency, 5 μm Zn²⁺ increased the relative contribution of the longest burst population (Table 1; Fig. 3C, F and K), resulting in a prolonged mean burst time (Fig. 3J). In contrast, 50 μm Zn²⁺ altered the kinetics of glycine-gated channels mainly by changing the average mean burst time and the fraction of current carried by the longest burst (Table 1; Fig. 3C, I, J and K).

We also analysed the closed and open times of single-channel currents activated by 30 μm taurine in the absence and presence of 5 μm Zn²⁺ (Table 1). The distribution of closed times in the presence of 30 μm taurine was best fitted by three exponentials with closed times of 0.65 ± 0.47 ms (29%), 120 ± 105 ms (62%) and 800 ± 657 ms (9%) (Table 1). In contrast to glycine-induced currents, where closed time distributions were best fitted by the sum of four exponentials, the most rapid component (time constant of about 0.6 ms) did not correspond to the highest number of closures (< 30%); this presumably reflects the low abundance of closings within bursts. Application of 5 μm Zn²⁺ caused a 2.8-fold increase in mean open time, which is more pronounced than results obtained with glycine (see Fig. 3J). In the presence of 5 μm Zn²⁺, observed closed times were 0.55 ± 0.25 ms (47%), 65 ± 54 ms (30%) and 400 ± 287 ms (23%). Open time histograms were again constructed from data obtained before and after exposure of the patches to 5 μm Zn²⁺. Open times were fitted best when assuming two different components with time constants of 0.7 ± 0.34 ms (τ₁; 92%) and 3 ± 2.9 ms (τ₂; 8%) in the absence, and 0.62 ± 0.3 ms (41%) and 3.2 ± 3.1 ms (59%) in the presence, of Zn²⁺ (Table 1). The contribution of the mean time constant (τ₂) of the long openings showed a significant increase in the presence of Zn²⁺ (see Table 1). Using a critical gap duration of ∼3 ms, average burst durations of 0.56 ± 0.24 ms (82%) and 4.2 ± 3.6 ms (18%) were obtained. Interestingly, open time and burst time distributions were very similar, reflecting again the low abundance of closings within bursts. Bursts recorded in the presence of 5 μm Zn²⁺ had average durations of 0.54 ± 0.3 ms (57%) and 10.4 ± 9.8 ms (43%) (Table 1). Taking into account the burst durations (Table 1), the relative contributions carried by the longest bursts were 62 and 93% in the absence and presence of 5 μm Zn²⁺, respectively. Thus, Zn²⁺ modulation of taurine-gated currents yielded results similar to those seen with glycine (see Fig. 3K). For both agonists, a change in the frequency of channel opening, as well in mean burst duration, was found in the presence of 5 μm Zn²⁺. However, a significant prolongation of the mean open time and the longest time constant of burst duration was only seen for taurine responses (Table 1).

In conclusion, this analysis confirms the previously observed macroscopic effects (Laube et al. 1995a), as low concentrations of Zn²⁺ (5 μm) enhanced the open probability of the GlyR channel by a factor of 2–5, whereas higher concentrations (50 μm) decreased the former without changes in amplitude or relative frequency of single-channel conductances. Our data indicate that low concentrations of Zn²⁺ increase the probability of the channel being in the open state by increasing burst frequency and duration in the presence of glycine or taurine. Higher Zn²⁺ concentrations inversely decrease channel open probability by decreasing burst frequency and mean burst duration.

Mutations in the extracellular N-terminal region of the α1 subunit affect Zn²⁺ potentiation and inhibition differentially

To reveal possible correlations between agonist binding and Zn²⁺ modulation, we analysed the effects of Zn²⁺ on α1 subunit mutants known to display altered agonist responses. Substitution of arginine 271 by lysine (R271K) causes a ∼10-fold increase in the EC₅₀ for glycine (Langosch et al. 1994), whereas the F159Y mutant shows a 4-fold enhanced efficacy of glycine activation (Schmieden et al. 1993). Although the R271K mutant exhibits a drastically reduced affinity for glycine, currents evoked with concentrations of glycine corresponding to the EC₂₅ value were similarly modulated by Zn²⁺ when compared with that seen with the wild-type α1 subunit (Fig. 4a). As the R271K substitution has been shown to convert the partial agonist taurine into a fully competitive antagonist of glycine receptor activation (Laube et al. 1995b; Rajendra et al. 1995), we also analysed taurine inhibition in the presence of Zn²⁺. At different concentrations of Zn²⁺, the potency of taurine to block glycine-induced currents was not altered (data not shown). This indicates that Zn²⁺ mainly affects the agonistic properties of taurine whereas its antagonistic potency is not changed. Analysis of the high affinity mutant α1^F159Y in the presence of the corresponding EC₂₅ for glycine showed a similar biphasic modulation by Zn²⁺ as seen with the wild-type α1 subunit (Fig. 4a). Although very different EC₂₅ concentrations of glycine were required to activate the wild-type and mutant receptors, similar concentrations of Zn²⁺ potentiated and antagonized the agonist response. These data imply that the inhibitory effects of Zn²⁺ cannot be attributed to a chelation of the agonists by Zn²⁺, as both the low and high affinity mutants showed rather similar Zn²⁺ modulation profiles.

Figure 4. Analysis of α1 subunit mutations.

A, Zn²⁺ dose-response curves obtained from oocytes injected with either the wild-type α1 cRNA (dashed line) or the α1^R271K (•) and the α1^F159Y (▪) mutant cRNAs in the presence of glycine concentrations corresponding to the respective EC₂₅ values. B, Zn²⁺ dose-response curves obtained from oocytes injected with either the wild-type α1 cRNA (dashed line) or the α1^D80G (▴) and the α1^T112A (♦) mutant cRNAs in the presence of glycine concentrations corresponding to the respective EC₂₅ values. C, ethanol (EtOH) modulation of oocytes injected with either the wild-type α1 (▪) or the α1^D80G (□) cRNA in the absence and presence of 5 and 50 μm Zn²⁺ (n = 5). D, outside-out patch recording from a HEK 293 cell expressing the α1^D80G mutant protein in the absence and presence of 5 μm Zn²⁺. E-J, gating properties of α1^D80G GlyRs in the absence and presence of 5 μm Zn²⁺. Histograms of closed time (E and H), open time (F and I) and burst duration (G and J) were constructed from a total number of channel openings of 7887 (E and F) and 2768 (G) in the absence (mean open time 2.57 ms, open probability 0.15) and 7893 (H and I) and 1988 (J) in the presence of 5 μm Zn²⁺ (mean open time 1.88 ms, open probability 0.13); the glycine concentration used was 3 μm. Abscissae are binned logarithmically (bin width 1/10 of a log unit), and ordinates are scaled to display N, the square root of the number of events per bin. All three distributions were fitted with multiple exponentials (continuous curves). The estimated parameters (time constants τ (ms) and relative areas (%)) are shown for each panel.

Our previous analysis of α1/β chimeric GlyR subunits has shown that the positive and negative effects of Zn²⁺ are exerted via different regions of the extracellular N-terminal domain of the α1 subunit. This led to the proposal that two different Zn²⁺ binding sites exist, and that residues which are not conserved between the α1 and β subunit, are involved in Zn²⁺ potentiation (Laube et al. 1995a). Here, we employed site-directed mutagenesis to positions of the N-terminal extracellular region of the recombinant human α1 subunit, which differ between the α1 and β sequence, in order to identify residues responsible for the modulatory effects of Zn²⁺. The non-conserved residues aspartate 80 (D80) and tyrosine 75 (Y75) were considered as candidate side chains because these amino acids interact with Zn²⁺ in other proteins and lie in a segment previously shown to be required for Zn²⁺ potentiation (Vallee & Falschuk, 1993; Laube et al. 1995a). Replacement of aspartate 80 by glycine generated glycine-gated receptor channels which showed no apparent differences in the EC₅₀ value of glycine (0.2 ± 0.04 mM) or the IC₅₀ value of strychnine (not shown) but markedly differed in Zn²⁺ potentiation (Fig. 4B, Table 2). The glycine response of the D80G mutant was inhibited at higher concentrations of Zn²⁺, whereas Zn²⁺ potentiation was completely abolished. In contrast, the Y75F mutant required higher concentrations of glycine for efficient gating (EC₅₀ 1.2 vs. 0.26 mM), but showed no impairment in the ability of Zn²⁺ to either potentiate or inhibit currents activated by EC₂₅ concentrations of glycine (Table 2).

Table 2.

EC₅₀ values of glycine and Zn²⁺ effects at wild-type and mutant α1 subunit GlyRs

		Zinc
cRNA injected	Glycine EC50 (mm)	EC50 (μm)	max. pot. (%)	IC50 (μm)	n
Wild-type	0.26 ± 0.02	0.89 ± 0.3	260	25 ± 4	10
Y75F	1.1 ± 0.2	0.96 ± 0.28	245	43 ± 21	3
Y78F	0.36 ± 0.1	2.8 ± 1.7	179	13 ± 9	5
P79R	0.31 ± 0.08	0.95 ± 0.4	90	21 ± 12	3
D80G	0.2 ± 0.04	no pot.	0	14 ± 3	6
D81A	0.29 ± 0.08	7.2 ± 3.8	195	78 ± 24	4
D84A	0.13 ± 0.03	0.56 ± 0.32	50	350 ± 170	4
D91A	0.25 ± 0.09	5.1 ± 3.0	148	38 ± 16	3
N102D	0.9 ± 0.2	0.61 ± 0.2	160	90 ± 31	3
F108A	0.07 ± 0.03	0.35 ± 0.14	400	20 ± 5	4
E110A	0.34 ± 0.06	2.92 ± 0.95	180	395 ± 110	4
T112A	0.12 ± 0.05	0.42 ± 0.06	260	no inhib.	5
E157A	n.f.	—	—	—	10
F159Y	0.1 ± 0.04	0.86 ± 0.17	280	14 ± 7	3
E165A	n.f.	—	—	—	10
E169A	0.27 ± 0.03	3.4 ± 2.1	130	45 ± 23	3
F207Y	0.55 ± 0.08	0.49 ± 0.12	280	87 ± 34	3
R271K	2.3 ± 0.5	1.2 ± 0.6	200	20 ± 9	4

EC₅₀ values for glycine were determined at a holding potential of −70 mV. Zn²⁺ modulation was investigated using a glycine concentration corresponding to the respective EC₂₅ value. From the resulting effect-response curves we operationally estimated apparent EC₅₀ and IC₅₀ values for Zn²⁺ potentiation and inhibition, respectively. Values are means ± s.d. of n determinations. n.f., nonfunctional; max. pot., maximal potentiation; no pot., no potentiation; no inhib., no inhibition.

In an attempt to delineate additional residues involved in Zn²⁺ modulation, we introduced several substitutions in the vicinity of D80. In most cases (Y78F, P79R and D81A), these substitutions did not drastically alter Zn²⁺ modulation or glycine affinity (Table 2). One mutant (D84A) showed a decreased Zn²⁺ potentiation of glycine currents, although the EC₅₀ values for glycine and Zn²⁺ were only modestly changed (Table 2). Substitution of threonine 112 by alanine (T112A), which results in an marginal increase in glycine affinity (Maksay et al. 1999), was insensitive to Zn²⁺ inhibition even at concentrations as high as 1 mM (Fig. 2B). In addition, inhibition by Ni²⁺, which inhibits α1 GlyR currents in a fashion resembling that of Zn²⁺ (Laube et al. 1995a), was abolished even when tested at concentrations as high as 1 mM (data not shown). Substitution of several side chains in the vicinity of residues involved in agonist binding (Vandenberg et al. 1992; Schmieden et al. 1993), i.e. the regions around positions 160 and 200 of the α1 subunit (mutants F159Y, E169A, F207Y) failed to alter Zn²⁺ modulation, although substitution of two glutamate residues at positions 157 and 165 resulted in a loss of the glycine response (Table 2). Substitution of aspartate 84 (D84A), which also affects potentiation, and of glutamate 110 (E110A) produced glycine currents indistinguishable from wild-type; however, Zn²⁺ inhibition was about 15-fold reduced. Collectively these data indicate that negatively charged residues in the extracellular N-terminal region, i.e. aspartates 80 and 84, are essential for high-affinity Zn²⁺ potentiation, whereas inhibition by Zn²⁺ requires threonine 112 and, to some extent, aspartate 84 and glutamate 110.

As the D80G substitution produced a complete loss of Zn²⁺ activation, we examined single-channel events in the absence and presence of Zn²⁺ after expression of the α1^D80G cDNA in HEK 293 cells (Fig. 4D and E). For the patch shown in Fig. 4, the distribution of closed times in the presence of 3 μm glycine was fitted by four exponentials. It can be seen that the mutation had no major effect on the distribution of open, burst and closed times (Table 1 and Fig. 4E–J). In addition, the burst durations and the exponential densities of the time constants in the absence and presence of Zn²⁺ for the D80G substitution were rather similar to the values for α1 wild-type receptors, confirming a selective loss of Zn²⁺ potentiation. This result supports the idea that the potentiating effect of Zn²⁺ is in part mediated via charge neutralization of the γ-carboxylate of aspartate 80.

To unravel whether the loss of Zn²⁺ potentiation is a specific consequence of the D80G substitution, we tested the effect of another positive modulator of the GlyR, i.e. ethanol. Ethanol potentiates glycine-induced currents at a concentration of 100 mM; this has been attributed to residues located in the transmembrane regions (Mihic et al. 1997; Ye et al. 1998). Both in the presence of potentiating (5 μm) and inhibitory (50 μm) Zn²⁺ concentrations, 100 mM ethanol produced potentiation of glycine-activated currents in both the α1 wild-type GlyR and the α1^D80G mutant (Fig. 4C). This indicates that the effects of Zn²⁺ and ethanol are additive, and that mutation of D80 selectively alters Zn²⁺ potentiation. Thus, our data suggest that ethanol and Zn²⁺ act at different binding sites.

DISCUSSION

This study describes an electrophysiological approach to elucidate the mechanism and molecular determinants of Zn²⁺ modulation of the recombinant α1 GlyR. By analysing the response properties of the principal agonist glycine and the partial agonist taurine at the whole-cell and single-channel levels, we found that potentiation of glycine currents by Zn²⁺ is caused primarily by an increase in channel open probability and burst duration, whereas Zn²⁺ inhibition is due to a reduction of these parameters. Our mutational analysis indicates that distinct residues in the N-terminal extracellular domain of the GlyR α1 subunit are crucial for the positive and negative effects of Zn²⁺. This corroborates the previously predicted existence of separate high- and low-affinity Zn²⁺ binding sites and establishes Zn²⁺ as a dual allosteric modulator of the GlyR.

Mechanism of reversible Zn²⁺ modulation of glycine currents

The present findings and those from our previous work (Laube et al. 1995a) are consistent with the idea that Zn²⁺ modulation of the GlyR is mediated by sites which affect neurotransmitter binding and/or channel gating. Dose- response analysis revealed a strong dependence of the EC₅₀ values of both glycine and taurine on Zn²⁺ concentration (Laube et al. 1995a; Lynch et al. 1998). In contrast, inhibition by the competitive antagonist strychnine was not altered in the presence of Zn²⁺. In additon, Zn²⁺ does not appear to affect ion permeation through the channel because there was no change in the unitary current amplitude and the relative abundance of subconductance states. Thus, Zn²⁺ must affect agonist binding/unbinding and/or channel opening/closing reactions. Here, we used different approaches to discriminate between these alternatives at both potentiating and inhibitory Zn²⁺ concentrations. First, we compared the effects of Zn²⁺ under whole-cell and single-channel recording conditions, and second, we used a partial agonist in an attempt to further separate effects on agonist binding from effects on channel gating.

Theoretical assumptions

In principle, several mechanisms could potentiate receptor function within the framework of a sequential agonist binding-channel activation scheme (del Castillo & Katz, 1957). In an attempt to characterize Zn²⁺ effects on the kinetics of α1 GlyRs, we fitted our experimental data to the gating model shown in Fig. 5A. This model was based on the following considerations. (i) Since the α1 GlyR is a homo-oligomeric receptor, we assumed that its different agonist binding sites are identical. (ii) In the absence of detailed kinetic studies on the mechanism of GlyR gating (see also Twyman & Macdonald, 1991), the high Hill coefficients reported by different investigators (n = 2.5-4.5; Schmieden et al. 1993; Bormann et al. 1993; Lewis et al. 1998) were taken as evidence for occupation of at least three agonist binding sites being required for efficient opening. This assumption is also consistent with the quarternary structure of the adult-type GlyR which is composed of three α1 and two β subunits (Kuhse et al. 1993). Accordingly, activation of the GlyR (R) can be formally described as a sequential binding of three agonist molecules (A_n; n = 3), followed by channel gating from the liganded closed (A₃R) to an open (A₃R*) state. However, we are aware that the model prediction of only one open state might be an oversimplification since three open time components are present (see Fig. 3). As the multiple agonist binding sites were assumed to have the same affinity (K) for agonist, unitary binding constants were used for analysing the mechanism of Zn²⁺ action. According to Colquhoun & Hawkes (1981), the equilibrium constant for channel opening (E) is defined by the quotient of the actual opening and closing rate constants (β/α). Continuous exposure to agonist might cause receptor desensitization. However, under our conditions this seemed not to be relevant, since desensitization of the GlyR in oocytes (i) is slow, (ii) was not altered in the presence of Zn²⁺ (B. Laube, unpublished observations), and (iii) requires higher concentrations of agonist than used for single-channel recording. These assumptions are consistent with our findings and the five-state model used by Chang & Weiss (1999) for GABA_C receptor activation which, like the α1 GlyR, is composed of five identical subunits and shows only little desensitisation. We therefore considered the simple receptor activation model shown in Fig. 5a appropriate for further analysis.

Mechanism of Zn²⁺ potentiation

When evaluating the effect of Zn²⁺ on single-channel events induced by low concentrations of glycine or taurine, Zn²⁺ was found to increase the open probability induced by either 3 μm glycine or 30 μm taurine. This is consistent with data from whole-cell experiments (Laube et al. 1995a) and binding studies (Lynch et al. 1998), where Zn²⁺ decreased the agonist binding affinity of the receptor. As changes in either agonist affinity or channel opening efficacy can alter the EC₅₀ values of agonists (see Colquhoun & Farrant, 1993), we calculated open time and closed time distributions to obtain estimates of rate constants by analysing the apparent single-channel kinetics of the recombinant GlyR (see Methods). Open time distributions of glycine-gated currents were well fitted by a triple-exponential function, whereas closed time distributions could be described by four exponential components. This is in good agreement with GlyR single-channel recordings obtained from different preparations (Takahashi et al. 1992; Bormann et al. 1993). Zn²⁺ did not appear to change the apparent intraburst kinetics because there was no significant change in the duration of gaps within bursts. Twyman & Macdonald (1991) have shown that the relative contributions of the burst duration constants of native GlyRs are dependent on the concentration of agonist, whereas time constants are not affected. We found remarkable changes of both parameters in the presence of Zn²⁺. The observed overall mean number of openings per burst was 11.2 ± 3.8 and 30 ± 7.6 in the absence and presence of 5 μm Zn²⁺, respectively. Estimates of the rate constant value k_-1 obtained from analysis of transitions within bursts in the absence and presence of 5 μm Zn²⁺ resulted in values which differ by a factor of 2–3 (about 200 and 70 s⁻¹, respectively). However, the apparent individual opening (β) and closing (α) rate constants seemed not to be affected in the presence of 5 μm Zn²⁺ (for β in the range 3000–4000 s⁻¹ and for α in the range 250–400 s⁻¹). This indicates that in the presence of potentiating Zn²⁺ concentrations gating is only marginally affected. Moreover, the calculated 3-fold decrease in the dissociation rate constant k_-1 of glycine in the presence of 5 μm Zn²⁺ is in agreement with the observed increase in the affinity for glycine. We therefore conclude that the potentiation of glycine currents observed in the presence of low concentrations of Zn²⁺ is primarily due to a slowing of agonist dissociation.

Mechanism of Zn²⁺ inhibition

Our analysis of the inhibitory effects of Zn²⁺ revealed two types of GlyR channel antagonism. At Zn²⁺ concentrations < 100 μm, inhibition was overcome by high agonist concentrations, and coapplication of Zn²⁺ caused only parallel shifts of the glycine dose-response curves to higher glycine concentrations (Laube et al. 1995a). In contrast, Zn²⁺ concentrations > 100 μm also reduced the maximal current response. Since glycine may chelate Zn²⁺, we calculated the free amounts of glycine and Zn²⁺ present in the solutions; complexation values were taken from Grenstine & Winitz (1961). Notably, at 3 μm glycine about 99% of the Zn²⁺ is present in free form. Accordingly, Zn²⁺ cannot merely act by chelating glycine. This is consistent with α1 subunit mutants displaying high or low affinity for glycine showing unaltered Zn²⁺ responses. Single-channel recording in the presence of different Zn²⁺ concentrations revealed no changes in main conductance state compared with values found in previous studies (Takahashi et al. 1992; Bormann et al. 1993). Also, Zn²⁺ did not change the linear dependence of glycine currents on membrane potential, providing an argument against a direct voltage-dependent block of the GlyR channel at higher metal ion concentrations. This suggests that the dual effect of Zn²⁺ inhibition may be mediated via the same or similar mechanisms.

To define the site of action of inhibitory concentrations of Zn²⁺, we measured single-channel events induced by 3 μm glycine in the absence and presence of 50 and 500 μm Zn²⁺. Consistent with the increase in the EC₅₀ for glycine seen in whole-cell experiments (Laube et al. 1995a) and binding studies (Lynch et al. 1998), the open probability of channel opening was decreased in the presence of high concentrations of Zn²⁺. Calculation of the dissociation rate constant for glycine indicates that higher concentrations of Zn²⁺ have no further effect on k_-1 (∼50 s⁻¹) in comparison to that seen in the presence of 5 μm Zn²⁺. Rather, high concentrations of Zn²⁺ (50 μm) impaired the efficacy of gating as indicated by a reduction in the mean channel open time that results from a 6-fold increased closing rate constant α (2500 s⁻¹). We therefore conclude that higher concentrations of Zn²⁺ primarily affect gating of the GlyR channel.

In conclusion, our estimates of the individual rate constants α and β, as well as of the resulting apparent equilibrium constant of channel opening (E), indicate that potentiating concentrations of Zn²⁺ have only a modest effect on channel gating but strongly enhance glycine binding by decreasing the agonist dissociation rate constant. In contrast, inhibitory concentrations of Zn²⁺ mainly affect channel gating by increasing α, thus producing a decrease in E.

Zn²⁺ modulation of partial agonist responses

Results similar to but more complex than those obtained with glycine were also found when analysing single-channel events induced by the partial agonist taurine. Although glycine- and taurine-activated single-channel currents had the same unitary conductances, their kinetic properties were significantly different. In the absence of Zn²⁺, the estimated E values were in the range 10–20 and 1–4 for glycine and taurine, respectively, indicating that taurine is only a low-efficacy agonist (see also Lewis et al. 1998). Notably, however, Zn²⁺ increased the maximal current response to taurine ∼2-fold; this is in contrast to our results obtained with glycine, where Zn²⁺ did not increase maximal currents. This can be explained by analysing our single-channel data according to the model shown in Fig. 5A. When estimating the rate constant k_-1 for taurine in the absence and presence of 5 μm Zn²⁺, values of ∼400 and ∼200 s⁻¹, respectively, were found. This reduction by a factor of 2 is similar to that observed with glycine and shows that taurine dissociation is also slowed by potentiating concentrations of Zn²⁺. In addition, and in contrast to the results with glycine, we found significant differences in the kinetics of taurine gating that reflect an increase in β (900 vs. 1800 s⁻¹). The observed changes in burst frequency and the substantial lengthening of the apparent burst duration in the presence of 5 μm Zn²⁺ (Table 1) indicate a taurine-specific alteration of gating by 5 μm Zn²⁺. Analysis of taurine-gated single channels in the presence of high concentrations of Zn²⁺ proved very difficult due to the low frequency of opening events. Consequently, only data obtained in the presence of potentiating Zn²⁺ concentrations could be evaluated. Collectively, our results are consistent with low concentrations of Zn²⁺ both slowing dissociation of and facilitating gating by the partial agonist taurine.

Point mutations selectively abolish potentiation or inhibition by Zn²⁺

Potentiation of α1 GlyRs by Zn²⁺ was voltage independent and not affected by decreasing the pH, suggesting that histidines are not involved in co-ordinating Zn²⁺ ions for potentiation. Inhibition by Zn²⁺ was also voltage independent but impaired by decreasing the pH. This suggests that residues in the channel forming M2 region are not involved in co-ordinating Zn²⁺ ions but histidines might play a role for the inhibitory effect of Zn²⁺. In addition, the site of Zn²⁺ action seems to be distinct from the agonist binding site because the amino acid substitutions F159Y and R271K, which drastically alter the EC₅₀ values of glycine, had little effect on Zn²⁺ modulation. Our mutagenesis experiments identified two residues (aspartate 80 and threonine 112) in the extracellular N-terminal region of the GlyR α1 subunit, whose substitution produced a selective loss of potentiation and inhibition, respectively. The structural modifications resulting from these substitutions seem to be discrete, since responses to glycine and channel properties were indistingishable from those of wild-type receptors. Residue D80 seems to be specifically implicated in Zn²⁺ potentiation as ethanol potentiation was not altered in the D80G mutant, indicating that the potentiating effects of Zn²⁺ and ethanol are mediated by different sites (or mechanisms). Notably, high Zn²⁺ concentrations caused neither shifts in dose- response curves nor a reduction of glycine-evoked currents in the T112A mutant. Thus, the complete loss of Zn²⁺ inhibition in T112A receptors across a wide range of Zn²⁺ concentrations is consistent with this residue being part of an inhibitory Zn²⁺ subsite. However, at present we cannot exclude the possibility that the D80G and T112A mutations affect Zn²⁺ potentiation and inhibition by conformational changes, as is probably the case with the D84A mutant, where both the potentiating and inhibitory effects of Zn²⁺ were altered. Indeed, the T112A substitution as well as mutations of neighbouring residues have recently been found to directly cause modest increases in agonist affinity (Schmieden et al. 1999).

During the preparation of this manuscript, Lynch et al. (1998) also reported a loss of Zn²⁺ potentiation of glycine-induced currents upon substituting residue D80 in the α1 subunit of the GlyR. In addition, these authors found an impairment of Zn²⁺ potentiation when introducing substitutions in the loop connecting transmembrane domains 1 and 2 of the α1 subunit. As the respective intracellular residues are unlikely to bind extracellularly applied Zn²⁺, it was concluded that Zn²⁺ potentiation of the GlyR is mediated via indirect allosteric mechanisms. However, our data indicate that potentiation by Zn²⁺ is likely to involve direct effects on agonist binding.

A schematic model of Zn²⁺ action

This study confirms the proposal that potentiation and inhibition of the GyR by Zn²⁺ are mediated by at least two distinct Zn²⁺ binding sites which differ in their concentration dependence and mechanisms of action (Laube et al. 1995a; Lynch et al. 1998). Accordingly, Zn²⁺ has multiple macroscopic effects on GlyR channels which can be formally explained by changes in some of the microscopic constants of a sequential agonist binding and channel gating scheme. The schematic model of Zn²⁺ action shown in Fig. 5 proposes that (1) potentiation of glycine currents results from an increase in agonist affinity that reflects a decrease in agonist dissociation rate(s) k_-1; and (2), at higher Zn²⁺ concentrations, a voltage-independent block is caused by a decrease in the efficacy of channel opening (Fig. 5a). At the molecular level, these effects of Zn²⁺ can be attributed to the high- and low-affinity Zn²⁺ binding sites defined by residues D80 and T112, respectively (Fig. 5B).

With the partial agonist taurine, potentiating concentrations of Zn²⁺ increased not only agonist affinity (1), but in addition the co-operativity of taurine gating, due to an altered Hill slope. We attribute this additional effect to an increase in β in the presence of Zn²⁺. Indeed, Hill coefficients of taurine activation of the α1 GlyR consistently were increased in the presence of the metal ion and were similar to those seen with the full agonist glycine (n values of 2.22 ± 0.3 compared with 1.66 ± 0.2 in the absence of Zn²⁺). Our data might also be explained by a sequential gating model adopted from tetrameric ion channels, i.e. cyclic nucleotide- and glutamate-gated ion channels (Ruiz & Karpen, 1997; Rosenmund et al. 1998). For the latter, differently liganded receptor populations have been shown to contribute fractionally to the total conductance. By analogy, the three burst time constants observed here upon glycine application might correspond to mono-, di- and tri-liganded states of the GlyR (see Fig. 5a) which open with different probabilities and time constants. Accordingly, full occupancy of the receptor should result in the longest open time. With elevated glycine concentrations, the fraction of multi-liganded receptors is predicted to rise, thereby producing a prolongation in mean burst duration. Such a model implies that the observed concentration dependence of burst duration constants (Twyman & Macdonald, 1991) would be the consequence of a change in the relative proportions of the multiple open channel states. Correspondingly, potentiating Zn²⁺ concentrations should stabilize the fully liganded receptor, thus resulting in the observed increase in mean burst time. In the case of the partial agonist taurine, only two burst time populations were detected in the absence of Zn²⁺, suggesting that taurine may only partially occupy the α1 GlyR. In the presence of Zn²⁺, the slowed dissociation of taurine might allow saturation of all agonist binding sites, which becomes apparent in a prolongation of the mean burst time, the occurrence of a third burst time constant (which was seen here only in one out of three patches; see Table 1) and an increase in co-operativity. Again, this conclusion is consistent with both agonist binding and, upon binding of partial agonists, channel opening being modulated via the same high-affinity site defined by aspartate 80 (Fig. 5B). We therefore suggest a common mechanism for Zn²⁺ potentiation of GlyR currents elicited by full and partial agonists, which may affect both agonist affinity and co-operativity via allosteric conformational transitions (Changeux & Edelstein, 1998). It should, however, be emphasized that our data cannot exclude the existence of an alternative transduction pathway for taurine, as proposed by Lynch et al. (1998). In addition, recent data suggesting that the partial agonism of taurine may be due to self-inhibition (Schmieden et al. 1999) also may be indicative of a more complex interaction of taurine and the D80 high-affinity Zn²⁺ binding site.

Implications for the physiological role of Zn²⁺ modulation

Zn²⁺ is one of the few potent allosteric modulators of the GlyR that increases its channel open probability in a manner consistent with an enhanced affinity of the receptor for glycine. An important question is whether the modulatory effects of Zn²⁺ are likely to occur in vivo. Assuming that the time course of glycine receptor-mediated synaptic currents is determined by the channel burst length, as found for other neurotransmitter receptors (Segal & Barker, 1984), evoked release of Zn²⁺ from its synaptic vesicle storage sites should substantially slow the decay of glycine receptor-mediated synaptic currents. In the presence of non-saturating concentrations of neurotransmitter, an increase in glycinergic synaptic currents could in addition occur. However, several lines of evidence indicate that upon presynaptic stimulation the peak concentration of neurotransmitter in the synaptic cleft is likely to saturate postsynaptic receptors. In this case, low concentrations of Zn²⁺ should not alter the maximal response to glycine. Due to transmitter overspill (Clements, 1996), however, submaximal glycine concentrations may be present at neighbouring synapses as well as at extrasynaptic GlyRs; there Zn²⁺ might potently reinforce glycinergic inhibition. This leads us to speculate that Zn²⁺ might have a neuromodulatory function in particular for tonic inhibitory control mediated by glycine or taurine, an endogenous inhibitory amino acid thought to be implicated in developmental and/or trophic functions (Flint et al. 1998). At higher Zn²⁺ concentrations generated upon repetitive release of synaptic Zn²⁺, even direct inhibition of glycinergic transmission may occur. These modulatory actions of Zn²⁺ may be important for spinal sensory information processing including nociception. Experiments targeting the Zn²⁺ binding sites identified in this paper in vivo might help to unravel the physiological consequences of such a dual regulation.