Skip to main content
PMC home page
The Journal of Physiology logoLink to The Journal of Physiology
. 2003 Sep 18;553(Pt 3):819–832. doi: 10.1113/jphysiol.2003.049577

Heterogeneity of postsynaptic receptor occupancy fluctuations among glycinergic inhibitory synapses in the zebrafish hindbrain

Jean-Michel Rigo *, Carmen Ionela Badiu *, Pascal Legendre *
PMCID: PMC2343629  PMID: 14500774

Abstract

The amplitude of glycinergic miniature inhibitory postsynaptic currents (mIPSCs) varies considerably in neurons recorded in the isolated hindbrain of 50-h-old zebrafish larvae. At this age, glycinergic synapses are functionally mature. In order to measure the occupancy level of postsynaptic glycine receptors (GlyRs) and to determine the pre- and/or postsynaptic origin of its variability, we analysed mIPSCs within bursts evoked by α-latrotoxin (0.1–1 nm). Two types of burst were observed according to their mIPSC frequencies: ‘slow’ bursts with clearly spaced mIPSCs and ‘fast’ bursts characterised by superimposed events. Non-stationary noise analysis of mIPSCs in some ‘slow’ bursts recorded in the presence or in the absence of Ca2+ denoted that mIPSC amplitude variance did not depend on the quantity of neurotransmitters released (presynaptic origin), but rather on intrinsic stochastic behaviour of the same group of GlyRs (postsynaptic origin). In these bursts, the open probability measured at the peak of the mIPSCs was close to 0.5 while the maximum open probability is close to 0.9 for the synaptic isoform of GlyRs (heteromeric α1/β GlyRs). In ‘fast’ bursts with superimposed events, a correlation was found between the amplitude of mIPSCs and the basal current level measured at their onset, which could suggest that the same group of GlyRs is activated during such bursts. Altogether, our results indicate that glycine synapses can display different release modes in the presence of α-latrotoxin. They also indicate that, in our model, postsynaptic GlyRs cannot be saturated by the release of a single vesicle.

In recent years, considerable effort has been made to analyse synaptic transmission in the central nervous system. From the many works aimed at understanding the mechanisms underlying the activity of single central synapses, it appears that these are functionally heterogeneous. This heterogeneity is exemplified by the broad and skewed distribution of amplitudes of miniature synaptic events observed at central excitatory or inhibitory synapses after the blockade of presynaptic action potentials (Bekkers et al. 1990; Manabe et al. 1992; Silver et al. 1992; Legendre & Korn, 1994; Auger & Marty, 1997). Accordingly, several hypotheses have been proposed to explain such a variability, including: a multi-vesicular release, fluctuations in the quantity of neurotransmitter released per vesicle, a difference in the number of receptors and/orreceptor subtypes among synapses, the clearance process of the neurotransmitter from the synaptic cleft or the stochastic properties of receptor channel gating (reviewed by Frerking & Wilson, 1996). The relative implication of these factors in controlling receptor occupancy at one synapse is, however, still poorly understood. Glycinergic synapses also display such a strong variability in the amplitudes of miniature inhibitory postsynaptic currents (mIPSCs), which has been observed in the mammalian brain stem (Singer & Berger, 1999) and spinal cord (Oleskevich et al. 1999) and in the zebrafish hindbrain (Ali et al. 2000).

A broad amplitude distribution can result from several factors that cannot be determined by the analysis of randomly occurring miniature postsynaptic events, as they originate from an unknown number of synapses. To address this issue, we used an approach based on the activation of vesicular release by α-latrotoxin (α-LTX) on inhibitory synapses located on reticular neurons of the hindbrain of the 50-h-old zebrafish larva. At this age, inhibitory glycinergic synapses are mature (Triller et al. 1997; Ali et al. 2000) and most of them possess one release site (Triller et al. 1997). α-LTX has a complex mode of action on vesicular release, but appears to stochastically activate single release sites, which resulted in the occurrence of isolated bursts of mIPSCs at central GABAergic synapses (Auger & Marty, 1997). A bursting behaviour of synaptic events evoked by brown widow spider venom was first reported at the frog neuromuscular junction (Pumplin & del Castillo, 1975) where each burst was suggested to arise from the activation of a single active zone. This was confirmed for black widow spider venom and α-LTX by combining intracellular and focal extracellular recordings in the same preparation (Fesce et al. 1986). α-LTX was purified from the black widow spider (Latrodectus mactans tredecimguttatus) venom and its targets seem localised at the presynaptic terminal (Auger & Marty, 1997). To be active, α-LTX needs to bind to two unrelated specific receptors: neurexin and latrophilin (reviewed by Sudhof, 2001). α-LTX was proposed to act either by a direct interaction with the release machinery (Ushkaryov, 2002) and/orby forming a cation-selective channel permeable to Ca2+ (Van Renterghem et al. 2000).

The goal of our study was to explore the origin of the amplitude fluctuations of the glycinergic mIPSCs in the zebrafish hindbrain. We first determined whether α-LTX can evoke some bursts of mIPSCs arising from a single release site at glycinergic synapses as it does at GABAergic synapses (Auger & Marty, 1997). We show that α-LTX can evoke bursts of glycinergic mIPSCs that are heterogeneous in terms of both mIPSC amplitude distribution and frequency. In bursts with mIPSCs of small amplitudes occurring with a relatively low frequency (‘slow’ bursts), we were able to determine the mean occupancy level of the postsynaptic receptors, their open probability and their number. These bursts are likely to reflect the activity of a single release site. When vesicular release occurred at a higher frequency (‘fast’ bursts), mIPSCs were superimposed and their amplitude variance had a complex origin.

METHODS

The experiments conform to the European Community guiding principles on the care and use of animals (86/609/CEE, CE Official Journal No. L358, 18 December 1986), to the French decree No. 97/748 of 19 October 1987 (Journal Officiel de la République Française, October 20 1987, pp. 12245–12248) and to the recommendations of the CNRS and of the University Paris VI.

Isolated brain preparation

The isolated intact zebrafish brain was prepared as previously described (Legendre & Korn, 1994). Briefly, 50-h-old zebrafish (Danio rerio) larvae were first anaesthetised by bath application of 0.1 mg ml−1 aminobenzoic acid ethyl ester (Sigma). The brains were then quickly dissected out and glued onto a coverslip using a plasma-thrombin embedding procedure. Before starting the experiments, brain preparations were stored for 15 min in an oxygenated (100 % O2) bathing solution containing (mm): NaCl 145; KCl 1.5; CaCl2 2; MgCl2 1; Hepes 10; glucose 10; pH 7.3 (330 mosmol l−1).

Whole-cell patch-clamp recordings

The isolated brains were placed in a recording chamber (0.5 ml) that was continuously perfused at room temperature (20–24 °C) with the oxygenated bathing solution (2 ml min−1). Standard whole-cell recordings (Hamill et al. 1981) were obtained from small reticular neurons located in the hindbrain under direct visualisation (Nikon Optiphot microscope). These cells differ from other neurons by their shape as well as by their location in the median part of the rhombomeres (Metcalfe et al. 1986). Patch-clamp electrodes were pulled from thick-wall borosilicate glass (GC150F-10: CEI, Harvard Apparatus Ltd, UK). They were then fire-polished and filled with (mm): CsCl 135; MgCl2 2; Na3ATP 4; EGTA 10; Hepes 10; pH 7.2 (290 mosmol l−1). Electrodes had resistances of 2–4 MΩ. Currents were recorded using an Axopatch 1D amplifier (Axon Instruments), filtered at 5 kHz, and stored using a digital tape recorder (DAT DTR 1201, Sony). Synaptic currents were recorded at a holding potential of −50 mV. Series resistances (6–20 MΩ) were 60–80 % compensated. Recordings with a series resistance > 20 MΩ were rejected.

Chemicals

mIPSC recordings were performed in the presence of tetrodotoxin (0.5-1 μm; Sigma), the glutamate receptor antagonists d-aminophosphonovalerate (APV; 2 μm; Tocris) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 20 μm; Tocris) and the GABAA receptor antagonist GABAzine (2 μm; RBI). α-Latrotoxin (α-LTX) was purchased from Latoxan (Rosan, France) and from Alomone Laboratories (Israel). These two batches of toxins had similar efficacies and α-LTX was used at a concentration of 0.1-1 nm. The toxin was dissolved in the recording solution in the presence of 0.1 % bovine serum albumin (Sigma).

Synaptic current analysis

Data were digitised off line using pCLAMP6 software at a sampling frequency of 40 kHz and were transferred onto a G4 Macintosh computer. Analysis was performed using Axograph 4.6 software (Axon Instruments) after digital filtering of the traces at 1–4 kHz. In ‘slow’ bursts (see Results) mIPSCs were detected using a form recognition procedure in which a typical mIPSC chosen in the current trace was used as a template. Events were only acquired when their amplitude exceeded three times the s.d. value of the basal noise. Amplitudes, 20–80 % rise times and half-widths were automatically measured on the detected events.

The coefficient of variation (CV) of the distribution of mIPSC amplitudes was calculated after subtracting the variance noise (varn) from the variance of the mIPSC amplitude distribution (vari) so that:

graphic file with name tjp0553-0819-m1.jpg (1)

where Imean is the mean of the mIPSC amplitude distribution.

A coefficient of skewness (Cs) of the mIPSC amplitude distribution was calculated for n ungrouped data as follows:

graphic file with name tjp0553-0819-m2.jpg (2)

In control conditions, the frequency of background events was randomly distributed. A burst of mIPSCs was defined as an isolated group of mIPSCs that occurred with an intraburst frequency exceeding that of the randomly occurring mIPSCs by at least fivefold.

Non-stationary noise analysis

To perform non-stationary noise analysis on synaptic events (Sigworth, 1980), mIPSCs must be fully isolated from each other to avoid time course or amplitude distortions. Glycinergic mIPSC decay phase in the zebrafish hindbrain can be described by a double exponential function with decay time constants of about 4 ms and 30 ms (Legendre, 1998). Therefore mIPSCs were selected by rejecting events occurring 50 ms before and after the detected mIPSC. Events with obvious artefacts were manually discarded. To minimise artefactual variance fluctuations due to contamination from the background mIPSC activity, events within a burst were further selected by using amplitude criteria as illustrated in Fig. 3B. These selected events were aligned on their rise time with respect to the time to 20 % of their maximum amplitude. Since the 20–80 % rise time of the events (≈ 0.2 ms) was 10–20 times faster that their half-width, non-stationary noise analysis was preferentially performed on the deactivation phase of the responses to prevent artefactual variance fluctuation due to small variations of the onset alignment. The averaged current and variance over time were computerised using Axograph software. The average current over time was calculated as:

Figure 3. Determination of the maximum open probability of postsynaptic glycine receptors (GlyRs) during ‘slow’ bursts.

Figure 3

Open in a new tab

Aa, example of mIPSCs occurring with a low frequency within a burst. Ab, plot of the baseline-to-peak mIPSC amplitudes as a function of the recording time. The dashed line represents the baseline current. B, plot of the instantaneous frequency of mIPSCs versus time within the burst shown in A. The mIPSC occurrence frequency periodically increased and occasionally reached 20–25 Hz in this example. C, histogram of the inter-event time intervals for the burst shown in A. The distribution of the inter-event time intervals was fitted by a single exponential curve with a time constant of 338 ms. D, the amplitude distribution of the detected events was fitted by a single Gaussian curve. Events attributed to background synaptic activity stand out of the Gaussian distribution. Non-stationary fluctuation analysis was performed on events standing within the Gaussian distribution (between arrows). The insert shows the distribution of the basal current noise giving a s.d. noise of 2.9 pA. E, plot of the mean variance versus the mean amplitude after subtracting the basal variance (8.4 pA2). Insert, example of 15 consecutive events selected for noise analysis (filter cut-off frequency, 2 kHz; VH, −50 mV). The parabolic fit gives an estimation of the number of the available postsynaptic GlyRs (N) and the elementary current generated by the activation of one receptor (i). In this case N = 46 and i = 2.36 pA. The elementary conductance of the GlyR channel (γ) was calculated for a Cl− equilibrium potential of 0 and a VH of −50 mV. The open probability of GlyR at the peak of the response (Po) was obtained by dividing the mean peak current by the estimated maximum mean amplitude when the variance reaches 0. In this case γ = 47.2 pS and Po = 0.56.

graphic file with name tjp0553-0819-m3.jpg (3)

where Imean(t) is the averaged current at time t, n is the number of sweeps and Ix(t) is the current at time t for the trace number x. The variance over time (vart) was calculated as follows:

graphic file with name tjp0553-0819-m4.jpg (4)

where vart is the variance at time t. Variance of the basal noise was subsequently subtracted. The point-to-point relationship between the mean variance at time t (vart) and the mean current at time t (Imean(t)) of the decay phase was fitted using a least square minimisation method with a parabola of the form:

graphic file with name tjp0553-0819-m5.jpg (5)

where i is the elementary current of the receptor channel and N is the total number of available GlyRs at the synapse. The single channel conductance

graphic file with name tjp0553-0819-m6.jpg (6)

was calculated with a Nernst Cl− equilibrium potential (ECl) of 0 mV and a holding potential (VH) of −50 mV.

Since only one conductance state has been described for the postsynaptic zebrafish GlyR and since GlyR desensitisation kinetics are too slow to influence the amplitude of mIPSCs and their time course (Legendre, 1998):

graphic file with name tjp0553-0819-m7.jpg (7)

where i is the elementary current of the GlyR channel, N is the number of available postsynaptic GlyRs and Po,max is the maximum open probability when GlyRs are saturated, i.e. the maximum fraction of receptor channels that are open. When the receptors are not saturated, the current Imean represents a fraction of Imax and:

graphic file with name tjp0553-0819-m8.jpg (8)

where Po represents the fraction of channels that are open at the peak of the response for a given concentration of agonist (A). The Po value depends on both Po,max and on the dissociation constant (KA) of the H binding sites (assuming all binding sites being equivalent). Accordingly:

graphic file with name tjp0553-0819-m9.jpg (9)

Since i and N can be considered as constant at a given synapse, assuming a single fully liganded open state, the relationship between Po and the agonist concentration can be described by a single binding isotherm function and Po will equal Po,max when receptors are saturated. Accordingly:

graphic file with name tjp0553-0819-m10.jpg (10)

When the point-to-point relationship between vart and Imean(t) of the decay phase deviated from the parabolic function predicted by the binomial law (see Fig. 4F), it was not possible to calculate N, i or Po. However, an estimation of the elementary current i of the receptor channels underlying mIPSCs could still be obtained by scaling the individual mIPSC amplitude to the averaged amplitude (Traynelis et al. 1993; Perrais & Ropert, 1999) before applying eqn (5) (see Fig. 4G). In this case, the point-to-point relationship between vart and Imean(t) was described by a parabolic function where vart tends to 0 when Imean(t) = 0 and when Imean(t) value approaches the mean amplitude value of mIPSCs. The i value was obtained by fitting data points by the eqn (5)

Figure 4. Non-stationary noise analysis of ‘slow’ bursts with large mIPSC amplitude variance recorded in the absence of external Ca2+.

Figure 4

Open in a new tab

A, example of a ‘slow’ burst recorded in the absence of external Ca2+. B, 25 consecutive mIPSCs occurring within the burst shown in A (filter cut-off frequency, 1 kHz; VH, −50 mV). C, the amplitude distribution of the detected events had a CV of 0.34 and a mean of 103.6 pA. The coefficient of skewness (Cs) was 0.16 (bin width, 10 pA; n = 67). The insert shows the distribution of the basal current noise giving a s.d. noise of 1 pA. D, correlation plot between 20–80 % rise times and mIPSC amplitudes. No significant correlation was found (Spearman rank order test P > 0.1). E, correlation plot between mIPSC half-widths and amplitudes. No significant correlation was found between these parameters (Spearman rank order test P > 0.1). Non-stationary analysis was performed on all events. F, plot of the mean variance versus the mean amplitude. Note that this plot deviates from the predicted parabolic function as shown in Fig. 3, indicating that the mIPSC amplitude variance cannot be explained only by intrinsic channel activity fluctuation. G, plot of the mean variance versus the mean amplitude of responses normalised with respect to their mean amplitude. The data points were fitted by a parabola allowing to extract the elementary current generated by a single GlyR channel giving an elementary conductance (γ) of 48.6 pS.

For display purposes, variance versus mean amplitude plots were binned by calculating the mean variance ±s.d.versus the mean amplitude interval ±s.d. for a given time interval. Amplitude increments between 2 and 4 pA were used, depending on the amplitude of the responses.

Simulation of synaptic currents

The simulations of bursts of mIPSCs were performed using Axograph 4.8 (Axon Instruments) on a G4 Macintosh computer. A previously developed kinetic model of glycine receptors expressed by M-cells was used (Legendre, 1998) to generate mIPSCs. The number of available GlyRs was adjusted to allow comparisons with experimental data.

The kinetic model had two equivalent agonist binding steps and the doubly liganded closed state provides access to a reluctant doubly liganded closed state. These two doubly liganded closed states can provide access to two independent open states (Legendre, 1998). The rate constants were adjusted to obtain theoretical responses with a time course similar to the experimental data. The association rate constant and the dissociation rate constant were 10 μm−1 s−1 and 2000 s−1, respectively. The opening rate constant and the closing rate constant linking the doubly liganded close state and one open state were 9000 s−1 and 850 s−1, respectively. The on rate constant and the off rate constant between the doubly liganded close state and the reluctant doubly liganded closed state were adjusted to 500 s−1 and 100 s−1, respectively. The opening rate constant and the closing rate constant linking the reluctant doubly liganded close state and the other open state were 3180 s−1 and 2000 s−1, respectively.

To generate simulated mIPSCs, the duration of the glycine concentration pulse was set to 0.5 ms with exponential ascending and descending time courses (0.1 ms time constant). The number of available channels was adjusted depending on the simulation performed. For each trial, GlyR channels were allowed to behave stochastically.

To simulate ‘fast’ bursts with many superimposed events, concentration pulses were generated randomly. To allow many simulated events to be superimposed, 200 concentration pulses were generated per 2.5 s run. The number of available GlyRs was determined depending on the type of simulation performed. Glycine concentration was adjusted according to experimental data to obtain GlyR occupation level for each trial of 40, 50 or 80 % depending on the type of simulation we wanted to perform. To simulate the activity of two independent release sites, two runs were combined. To determine the underestimation level of the mean Po measurement when two independent synaptic boutons were synchronously active, three types of simulation were performed. The first simulation run supposed that the two independent release sites activate different receptor clusters having the same number of GlyR but with identical Po. The second simulation run supposed that they are different receptor clusters with different numbers of GlyR but with identical Po. The third simulation run supposed that both Po and the number of postsynaptic GlyRs per cluster varied.

Baseline current and mIPSC amplitude measurements on bursts with many superimposed events

Event amplitude measurements were performed manually on experimental data, after a digital filtering at 2 kHz, using Axograph 4.8 software and they were also performed on simulated events. Baseline currents were measured at the onset of the events and mIPSC amplitudes were obtained by measuring the onset-to-peak amplitudes (see Fig. 6). The time a which the mIPSC amplitude was measured after the onset ranged from 0.6 to 0.8 ms. Decay time constants of mIPSCs were calculated on the few events that could be isolated within a burst. In all bursts analysed, isolated mIPSCs were characterised by a double exponential decay. The fast and the slow decay time constants of the mIPSCs varied among bursts and ranged from 4.2 to 7 ms and from 30 to 60 ms, respectively. The relative ratios of the two decay components also varied and ranged, for the fast decay component, from 90 to 40 % among bursts. Between two events, the ratio of the current decrease of the first event at the onset of the next event (A) was calculated as follows:

Figure 6. Correlation between mIPSC amplitude and current baseline within ‘fast’ bursts.

Figure 6

Open in a new tab

A, ‘fast’ burst recorded in the presence of 1.3 mm external Ca2+. Ba, portion of the ‘fast’ burst shown in A with many superimposed events. Bb, enlarged portion of this ‘fast’ burst (box in Ba) where the parameters measured for the occupancy analysis are shown: a1 is the amplitude of the baseline measured from the zero baseline current determined before and after the burst, and a2 is the peak amplitude of the mIPSCs measured from the baseline current at the onset of the response. C, mIPSC (upper graph) and baseline current (lower graph) amplitude fluctuations with recording time. Note that mIPSC amplitude has a tendency to decrease when the baseline current increases. D, plot of the peak response amplitude (a2) versus the baseline current amplitude (a1). These values were then corrected to account for the decay occurring in the first mIPSC while the second rose to peak (see Methods). The data points in grey are individual measurements showing the fluctuation of mIPSC amplitudes for a given baseline value, while points in black are the mean amplitudes of mIPSCs obtained for a 5 pA change in the baseline amplitude. The amplitudes of mIPSCs and their fluctuation decreased when the baseline increased, indicating that superimposed mIPSCs are likely to result from the activation of a given group of postsynaptic glycine receptors. Single linear regression of the plot yielded a significant correlation between the mean peak amplitudes of mIPSCs and the baseline current amplitudes (Spearman rank order test P < 0.001). The slope of the fit was −0.445. E, no significant correlation was found between the mean percentage of receptor occupancy (slope in D) and the mean amplitude of mIPSCs for baseline current = 0 pA (Spearman rank order test P > 0.1).

graphic file with name tjp0553-0819-m11.jpg (11)

where a1 and a2 are the relative ratios of the two exponential components, τ1 and τ2 are the corresponding time constants and T is the time interval between the peak of the first event and the onset of the second event.

The ratio of the current decrease of the first event at the peak of the second event (A’) was then calculated as:

graphic file with name tjp0553-0819-m12.jpg (12)

where ttp is the time-to-peak (0.6-0.8 ms) from the onset at which the amplitude of the second mIPSC was measured.

The corrected baseline current amplitude (Ib’) at time T+ ttp was then obtained as follows:

graphic file with name tjp0553-0819-m13.jpg (13)

where Ib is the baseline current measured at the onset of the event.

mIPSC amplitudes (Ievent) were then corrected (Ievent’) as follows:

graphic file with name tjp0553-0819-m14.jpg (14)

Results are expressed as mean values ± standard deviation (s.d.) throughout, except where otherwise stated.

RESULTS

Miniature inhibitory synaptic events in small reticular neurons

The analysis of bursts of mIPSCs induced by α-LTX was performed on small reticular neurons located in the median part of the zebrafish hindbrain. These cells had an input resistance ranging from 400 to 900 MΩ. They were electrically compact with a limited dendritic tree and had a small input capacitance (≈ 10 pF), allowing a good recording resolution. These neurons were preferred to large reticular neurons as their glycinergic mIPSCs occurred with a lower frequency (0.3-2 Hz) compared with those in large neurons (≈ 10 Hz; Ali et al. 2000). Cells with a mIPSC frequency < 2 Hz were used for the analysis of the bursts of mIPSCs induced by the application of α-LTX.

In the presence of TTX, GABAzine (a GABAA receptor antagonist), APV (an NMDA receptor antagonist) and CNQX (an AMPA/kainatereceptor antagonist), a glycinergic mIPSC activity was recorded that is characterised by isolated events of highly variable amplitudes (Fig. 1Aa). In the example shown in Fig. 1Ab, the mIPSC amplitudes were broadly distributed, ranging from 10 to 500 pA with a peak near 20 pA. The mean amplitude value of mIPSCs ranged from 44 to 178 pA (106.1 ± 50.2 pA; n = 17 cells) with a CV of 0.926 ± 0.245 (n = 17 cells). These values are very similar to those described in the Mauthner cell (mean amplitude of 140 pA with a CV of 0.9; Ali et al. 2000). In all recorded cells, amplitude distributions were positively skewed (Fig. 1Ab), as also observed in large reticular neurons. Skewness of the fit was estimated by calculating a coefficient of skewness (Cs; see Methods), which yielded a mean value of 1.8 ± 0.72 (n = 17 cells), as expected for a positively skewed distribution. The 20–80 % rise times of the mIPSCs had a mean of 0.25 ± 0.04 ms (n = 17 cells) and were not correlated to the amplitudes (R = 0.172± 0.068; n = 17 cells), indicating that cable attenuation could not account for the mIPSC amplitude variability.

Figure 1. Glycinergic mIPSCs recorded in control conditions and after α-latrotoxin application.

Figure 1

Open in a new tab

Aa, example of mIPSC activity recorded in control conditions in the presence of 0.5 μmm TTX, 2 μmd-aminophosphonovalerate (APV) 20 μm 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and 2 μm GABAzine. Note the large variation in the mIPSC amplitudes. Ab, amplitude histogram of mIPSCs recorded for a period of 300 s The distribution is highly skewed with a coefficient of skewness (Cs) of 1.66 and a coefficient of variation (CV) of 0.94. The insert shows the basal noise amplitude distribution with a s.d. noise of 2 pA. Ba, burst of mIPSCs after the application of 1 nmα-latrotoxin (α-LTX). Bb, the amplitude distribution of mIPSCs within the burst Ba could be fitted by a Gaussian curve showing their homogeneity compared to control mIPSCs. The insert shows the basal noise amplitude distribution after α-LTX application. Records in Aa and Bb are from the same reticular neurons (holding potential (VH), −50 mV).

Such a large variation in glycinergic mIPSC amplitudes can theoretically depend on either pre- and/orpostsynaptic mechanisms. According to the latter hypothesis, it was previously proposed to result from differences in the number of postsynaptic receptors among synaptic boutons (Oleskevich et al. 1999; Suwa et al. 2001). However, postsynaptic GlyRs do not appear to be fully occupied (Suwa et al. 2001). Therefore, it is likely that such a large variation in mIPSC amplitude could result from complex interactions between pre- and postsynaptic factors. In order to gain new insights into the mechanism(s) underlying mIPSC amplitude variability, we decided to test the effects of α-LTX, an agent known to evoke isolated bursts of mIPSCs arising from a single release site (Auger & Marty, 1997).

Bursts of glycinergic mIPSCs evoked by α-latrotoxin

After a 5 min period of control recording of mIPSCs, as described in the previous paragraph, α-LTX was applied for 3–5 min at a concentration of 0.1-1 nm and was then washed out. As previously reported for these concentrations of α-LTX (Auger & Marty, 1997), changes in mIPSC activity induced by the toxin did not occur immediately, but rather 2–10 min after the end of the application. Two main effects of α-LTX were previously described on the GABA synapse of cerebellar interneurons (Auger & Marty, 1997). The first one, consisting of the occurrence of single channel openings and closures reflecting the insertion of α-LTX channels in the membrane of the postsynaptic cell, was rarely observed in our preparation. The second effect of α-LTX is its ability to evoke isolated bursts of mIPSCs (Figs 1Ba, 2A, 3Aa, 4A, 5A and 6A) resulting from its action on the release machinery of the presynaptic terminals (Ushkaryov, 2002). These bursts were usually infrequent, but in some cells their rate of occurrence increased with time to a point where bursts could not be clearly isolated. Only recordings where bursts were well isolated have been analysed. These bursts can be easily isolated since the frequency of mIPSCs within a burst was much higher than in control conditions. Within an α-LTX-evoked burst, the mIPSC amplitude distribution is homogeneous and, in some cases, it could be fitted by a Gaussian curve (Fig. 1Bb and Fig. 3D). Among bursts, however, the mean amplitude of mIPSCs can strongly vary. Figure 2A shows two examples of bursts recorded from the same cell in the presence of 1.3 mm external Ca2+. The mean amplitudes of mIPSCs within these bursts are ≈ 305 and ≈ 31 pA, respectively, for Fig. 2Aa and Ab. The overall distribution of the mean amplitudes of mIPSCs coming from 36 bursts recorded in the presence of 1.3 mm external CaCl2 is shown in Fig. 2B. This distribution is positively skewed (Cs = 1.93) and its mean current amplitude (100.7 pA) and CV (1.08) are similar to the values measured for mIPSC amplitude distributions recorded in control conditions (see Fig. 1). There are no significant differences between the averaged control mIPSC amplitude distribution and the distribution of the mean amplitudes of mIPSCs within α-LTX-evoked bursts (Kolmogorov Smirnov test, P > 0.1)

Figure 2. Mean amplitude of mIPSCs can strongly vary among α-LTX-induced bursts of mIPSCs.

Figure 2

Open in a new tab

Aa, example of a burst with large amplitude mIPSCs (VH, −50 mV). In this burst, mIPSC frequency decreased with time while at the beginning of the burst many synaptic events were superimposed. Ab, another burst with mIPSCs of much smaller amplitudes was recorded on the same cell as in Aa (VH, −50 mV). B, distribution of the mean amplitude values of mIPSCs obtained for 36 different bursts. This amplitude distribution is highly skewed (Cs, 1.93) with a CV of 1.08, as observed in control conditions (see Fig. 1Aa and Ab).

Figure 5. Frequency pattern of mIPSCs within α-LTX-induced ‘fast’ bursts.

Figure 5

Open in a new tab

A, example of a burst with many superimposed events (‘fast’ burst) recorded in the absence of external Ca2+ (VH, −50 mV). B, plot of the instantaneous frequency of mIPSCs versus time within the burst shown in A. The mIPSC occurrence frequency periodically increased and reached 920 Hz in this example. C, histogram of the inter-event time intervals for the burst shown in A. The distribution of the inter-event time intervals was fitted by the sum of two exponential curves with time constants of 14.3 ms (59.2 %) and 42.4 ms. D, amplitude histogram of the detected events. This distribution has a CV of 0.36 and a mean of 74.7 pA (between arrows).

The variation of α-LTX-evoked bursts not only concerns mIPSC amplitudes, but also the total duration of the bursts (ranging from 0.1 s to 1 min) and the mIPSC frequency within a burst. In terms of their mIPSC intraburst frequencies, three types of bursts were observed. The first type of burst (‘slow’ burst) was made of mIPSCs occurring at a low frequency that remained stable or that only slightly declined over time (Fig. 3A). Inter-event interval distribution of these bursts was well fitted by a single exponential curve (Fig. 3C) with time constants ranging from 0.12 to 0.6 s among bursts. Some rare bursts started with events occurring at high frequency that eventually reached a stable value in the latter part of the burst (Fig. 2A). The so-called ‘fast’ bursts were characterised by mIPSCs occurring with a high frequency (inter-event distribution time constant < 0.05 s; see section headed ‘Bursts of mIPSCs occurring at a high frequency’) and often superimposed that did not decline over time (see Fig. 5 and Fig. 6). All types of bursts, thus also including ‘fast’ bursts, were observed even in the absence of external Ca2+ (see section headed ‘Bursts of mIPSCs occurring at a high frequency’). Since neither qualitative nor quantitative differences were observed among the bursts recorded in the presence and in the absence of external Ca2+, theses two conditions were pooled for statistical analysis.

Open probability of postsynaptic GlyRs for mIPSCs occurring with low frequency within a burst

If mIPSCs within α-LTX-evoked bursts originate from a single release site facing a single set of receptors, the open probability of postsynaptic GlyRs should be able to be measured using a method known as the non-stationary fluctuation analysis developed by Sigworth (1980). Such an analysis is aimed at extracting information about the elementary conductance of the postsynaptic receptor, the number of postsynaptic receptors present at the synapse and the open probability of the receptor per vesicle released. It also helps to reveal whether the main source of mIPSC amplitude variations comes from single receptor channel kinetics fluctuations. If this is true, the plot of the current variance versus the current amplitude needs to be fitted by a parabola with an initial slope value compatible with known GlyR elementary current. As non-stationary noise analysis can only be carried out on events with no contamination due to mIPSC superimpositions, bursts with a low mIPSC frequency were selected for this analysis. The occurrence of such ‘slow’ bursts could not be controlled since, in the zebrafish hindbrain, decreasing the external Ca2+ concentration did not markedly reduce the mIPSC frequency within a burst evoked by α-LTX (see section headed ‘Bursts of mIPSCs occurring at a high frequency’ and Fig. 4).

Figure 3A shows one of the 31 ‘slow’ bursts that were analysed. In the example of Fig. 3, the mIPSC amplitude distribution was fitted by a single Gaussian curve (Fig. 3D). Contamination arising from synaptic background activity was determined and, hence, discarded as classes of events being out of the range of the Gaussian amplitude distribution (delimited by the two arrows in Fig. 3D; in this example, five events were not considered as being part of the α-LTX-induced burst).

In 17 out of the 31 ‘slow’ bursts, the plot of the time-averaged values of mIPSC mean and variance amplitudes was fitted by a parabola (Fig. 3E). The mean amplitudes of the mIPSCs ranged from 11.7 to 88 pA. The initial slope of the fit was used to calculate the elementary channel current (i): i ranged from 2.05 to 2.9 pA (2.53 ± 0.28 pA; n = 17 cells), which gives, with a Cl− equilibrium potential at 0 mV and a holding potential of −50 mV, elementary GlyR conductances (γ) ranging from 41 to 58 pS (50.6 ± 5.6 pS; n = 17 cells). These measurements are in accordance with the conductance range already described for zebrafish α/β heteromeric GlyRs (≈ 45 pS; Legendre, 1997), which are known components of the inhibitory synapses in the zebrafish hindbrain (Ali et al. 2000). The parabola fit also helped to determine the number of available postsynaptic GlyRs (N), which ranged from 7 to 74 GlyRs (33 ± 20 GlyRs, n = 17 cells) together with the open probability (Po), which ranged from 0.3 to 0.82 (0.508 ± 0.151; n = 17 cells). Assuming a Po,max of 0.9 for heteromeric α/β GlyRs (Legendre, 1998; see also Methods), the mean value of Po (0.501) derived from non-stationary noise analysis suggests that all postsynaptic GlyRs were not fully activated at these glycinergic inhibitory synapses.

In the 14 other bursts analysed, which will be called irregular ‘slow’ bursts hereafter (Fig. 4), the amplitude distribution was usually complex (Fig. 4C). The amplitude distribution had low Cs values (ranging from −0.11 to 0.54; 0.26 ± 0.23; n = 14 cells) and no correlation was found between 20–80 % rise times and mIPSC amplitudes (Fig. 4D), The plot of the variance as a function of the mean current amplitude of isolated mIPSCs could not be fitted with a parabola indicating that N and/orPo varied from one mIPSC to another (Fig. 4F). Fitting these plots with eqn (5) gave an infinite value for N and an i value > 10 pA. This i value gave an unrealistic single channel conductance level for GlyRs ≥ 200 pS.

Since there were no correlation between the decay and the amplitude of the mIPSCs, which ensures that the expected waveform is common for all mIPSCs (Fig. 4E), peak-scaled non-stationary fluctuation analyses were performed to estimate unitary GlyR current (i) (see Methods; Fig. 4G). Values for i ranged from 2.21 to 3.1 pA (2.64 ± 0.3 pA; n = 14 cells), yielding unitary conductances (γ) ranging from 44.2 to 62 pS (52.7 ± 5.8 pS; n = 14 cells). We also observed that the mean amplitude of the mIPSCs within these bursts, which ranged from 32 to 442 pA (152.6 ± 138.4 pA; n = 14 cells) was significantly higher than the mean amplitude values of mIPSCs (39.1 ± 23.6 pA; n = 17 cells) within the 17 bursts for which Po could be successfully calculated (unpaired t test P < 0.01). Their respective mIPSC rising phases were not significantly different (unpaired t test P > 0.1): the 20–80 % rise time was 0.27 ± 0.07 ms (n = 14 cells; filter cut-off frequency = 2 kHz) for large amplitude bursts and 0.256 ± 0.06 ms (n = 17 cells) for lower amplitude bursts.

Irregular ‘slow’ bursts could therefore reflect a large variation in the number of neurotransmitter molecules released at a single release site or the synchronous activity of several release sites or even both.

Bursts of mIPSCs occurring at a high frequency

In ‘fast’ bursts, mIPSCs occurred at a high frequency and appeared superimposed as clusters of events of short time intervals separated by longer time intervals (Fig. 5 and Fig. 6). Similar to the ‘slow’ bursts, these ‘fast’ bursts were observed in the absence and in the presence of external Ca2+. Time intervals among mIPSCs were exponentially distributed suggesting that the events occurred randomly (Fig. 5C). In 16 ‘fast’ bursts, the best fit was given by a sum of two exponential curves (Fig. 5C) with time constants τ1 and τ2 of 16.5 ± 5.0 and 67.7 ± 38.0 ms (n = 16 cells), respectively. In four other bursts, the time interval distribution was best fitted with a single exponential curve with a time constant of 39.9 ± 24.0 ms (n = 4 cells).

As in ‘slow’ bursts, mIPSC amplitude variability within such ‘fast’ bursts is lower than the one observed for mIPSCs occurring randomly in the absence of α-LTX. In all ‘fast’ bursts, most of the events were superimposed, but peak amplitude of events arising from a baseline close to 0 pA (according to the basal noise fluctuation) could be measured, allowing us to analyse mIPSC amplitude fluctuations (Fig. 5D). Mean mIPSC amplitudes ranged from 21 to 419 pA (mean = 96.9 ± 87.9 pA; n = 20 cells). The individual mIPSC amplitude distribution showed a CV ranging from 0.31 to 0.72 (0.459 ± 0.11; n = 20 cells) with a coefficient of skewness (Cs) ranging from −0.38 to 0.84 (0.24 ± 0.39; n = 20 cells). The CV values of ‘fast’ bursts were significantly higher than those of ‘slow’ bursts for which non-stationary noise analysis were successfully performed (unpaired t test P < 0.01), which ranged from 0.12 to 0.37 (0.22 ± 0.08; n = 17 cells). These data suggest that mIPSC amplitude fluctuation within ‘fast’ bursts could have a more complex origin.

We then determined whether interactions between superimposed mIPSCs within ‘fast’ bursts could occur. As shown in Fig. 6Aa–Bb, several synaptic events need to superimpose before their amplitude apparently declines. This is also exemplified by the tendency of mIPSC amplitudes to decrease when the baseline current increased (Fig. 6C). This decrease is unlikely to be the consequence of GlyR desensitisation since the absolute mIPSC amplitude did not decline with time and since heteromeric α/β GlyRs have slow desensitisation kinetics (Legendre, 1998). For all bursts tested, there were no correlation between the ratio of the amplitude of two adjacent mIPSCs (amplitude of the second mIPSC divided by the amplitude of the first one) and their time interval. This rules out the occurrence of a fast desensitisation process for postsynaptic GlyRs. For example, such a calculation gave a correlation coefficient of 0.06 for the burst shown in Fig. 6A.

Hence, we measured the relationship between the amplitude of mIPSCs (a2 in Fig. 6Bb) and the baseline value at the onset of the response (a1 in Fig. 6Bb). These values were then corrected to account for the decay occurring in the first mIPSC while the second rose to peak (see Methods). For the 20 ‘fast’ bursts studied, this relationship could be fitted as a first approximation by a linear regression (Fig. 6D). The slope of the linear regression ranged among bursts from 0.146 to 0.611 (0.396 ± 0.145; n = 20 cells). As illustrated in Fig. 6E, there was no significant correlation between the slope of the linear regression and the mean amplitude of the mIPSCs for baseline current ≈ 0 pA (Spearman rank order test P > 0.1).

Fast bursts of mIPSC arising from a single release site were simulated with synaptic events having an amplitude representing 50 % of the maximum current (i.e. Imean/Imax, where Imean is the mean amplitude of events when the baseline equals zero and Imax is the maximum current when the receptors are saturated, i.e. when GlyR Po approaches Po,max; see Methods). The slope of the linear regression (0.497) was similar to the expected Imean/Imax value (data not shown).

Other simulations helped us to examine the effect of variable Po (fraction of receptor channels that are open) and N on the slope of the linear regression. Results obtained by the simulation of two independent release sites will be detailed. Simulation with more than two independent release sites resulted in larger measurement distortions. When two independent release sites were simulated but with similar Po and N, the slope of the linear regression resulted in a 42 % underestimation of the true Imean/Imax. When Po values were similar for the two release sites, but N different (by a factor of 2), the slope of the linear regression resulted in a 37 % underestimation of the true GlyR Imean/Imax. Similar N and different Po values (by a factor of 2) resulted in a 33 % underestimation of the Imean/Imax. Finally when the two independent release sites were simulated with both different Po and N values (each by a factor of 2), which is the most likely physiological situation, the averaged Imean/Imax was underestimated by 55 %. Therefore, if two independent release sites were to be considered, Imean/Imax underestimation will range from 33 to 55 % depending on the N and/orPo variability among release sites. Hence, our calculated mean slope of 0.396 would mean a real Imean/Imax ranging from 0.59 to at least 0.88.

DISCUSSION

Compared to their highly skewed distribution in control conditions, mIPSC amplitudes are more homogeneously distributed within α-LTX-evoked bursts. However, such a distribution was not necessarily the hallmark of the activity of a single active zone. Two main types of bursts were distinguished on the basis of their inter-event time intervals. ‘Slow’ bursts were characterised by mIPSCs that can be individualised and ‘fast’ bursts by superimposed events. In a subclass of ‘slow’ bursts having low mIPSC amplitudes, we succeeded in demonstrating that mIPSCs reflect the activity of a single active zone and that postsynaptic receptors were not saturated, the mean occupancy level for GlyRs being close to 50 %.

Postsynaptic GlyR open probability in ‘slow’ bursts

For some bursts in which mIPSCs occurred with a slow frequency, a parabolic relationship between the variance and the mean amplitude of mIPSCs was obtained. Such bursts were recorded both in the presence and in the absence of external Ca2+ indicating that their occurrence did not depend on Ca2+ entry at the presynaptic terminal, as observed in ‘slow’α-LTX-induced bursts arising from GABAergic inhibitory synapse in the rat cerebellum (Auger & Marty, 1997). Two assumptions are implicit in this approach: (1) the same postsynaptic receptor matrix with the same number of available receptors must be considered during each mIPSC, and (2) the change of Po as a function of time and also the quantity of neurotransmitters released per trial (mIPSC) must be closely the same. Accordingly, when a parabolic relationship was obtained, this suggested that bursts of mIPSCs originate from the activity of a single active zone (Auger & Marty, 1997) and, in this case, the mIPSC amplitude variance has a strict postsynaptic origin, i.e. it results from the intrinsic stochastic properties of receptor channel gating (Sigworth, 1980). Additional sources of variance, due, for example, to large variations in the quantity of neurotransmitters released per trial, would give a non-parabolic relationship and an important overestimation of the elementary conductance value (γ) of the receptor channel, as was exemplified by the plot shown in Fig. 4F. However, a small fluctuation in the quantity of neurotransmitters released per trial can also give an additional source of variance that does not impair a parabolic plot. This small additional source of variance would also result in some overestimation of the elementary conductance value. Since the α/β heteromeric GlyR, which is the postsynaptic GlyR isoform at glycinergic synapses, has a single conductance state of γ≈ 45 pS (Legendre, 2001), a parabolic plot giving a larger γ value would suggest some additional source of variance. In our experiments, γ ranged from 41 to 58 pS among bursts. This could reflect an overestimation of γ ranging from 0 to 30 %, which suggests that although the intrinsic stochastic properties of receptor channel gating remains the major source of variance, a small fluctuation in the quantity of neurotransmitters released per mIPSC could account for the small additional variance.

In our preparation, in general, postsynaptic GlyRs did not appear to be saturated after the release of a vesicle. Indeed, the GlyR open probability ranged from 0.3 to 0.8 at the peak of mIPSCs whereas the Po,max of zebrafish heteromeric α/β GlyR is 0.9 (Legendre, 1998). Such a conclusion arises from the hypothesis that extra-synaptic GlyRs analysed using outside-out recordings (Legendre, 1998) showed behaviour similar to that of postsynaptic GlyRs. There are several pieces of evidence suggesting that these receptors are functionally closely identical: they are both heteromeric α/β GlyRs, they have the same main conductance level and the same opening time constant (Ali et al. 2000). Moreover, α/β heteromeric GlyRs recorded in outside-out patches had slow desensitisation kinetics (Legendre, 1998). Since mIPSCs or outside-out currents evoked by short concentration steps (1 ms) using a fast-flow application system had closely similar activation and deactivation time courses (Legendre, 1998), this suggests that postsynaptic receptors did not display fast desensitisation processes. The Po,max of zebrafish postsynaptic heteromeric α/β GlyRs has never been estimated directly. However, the highest Po value (0.8) obtained using non-stationary noise analysis on ‘slow’ bursts suggests that both types of receptors had closely similar Po,max.

This lack of GlyR saturation is consistent with our previous study in the zebrafish hindbrain that showed that mIPSC amplitudes increased when GlyR affinity for glycine was increased by zinc application (Suwa et al. 2001). The fraction of postsynaptic receptor that is activated at inhibitory synapses has been reported to be highly variable among preparations. For example, a high fraction of available postsynaptic receptors was activated for GABA synapses in hippocampal slices (De Koninck & Mody, 1994; Poncer et al. 1996) or in cerebellar slices (Auger & Marty, 1997). Conversely, a low fraction of available postsynaptic GABAA receptors was activated in amacrine cell cultures (Frerking et al. 1995), in the visual cortex (Perrais & Ropert, 1999) and in cultured striatal neurons (Rumpel & Behrends, 2000). The fraction of available receptors that are activated by a single vesicle release might actually dependent on several pre- and/orpostsynaptic factors (Kruk et al. 1997) including: the size of postsynaptic receptor clusters (Nusser et al. 1997; Auger & Marty, 1997), the concentration of the neurotransmitter released in the synaptic cleft (Frerking et al. 1995) and the clearance of the neurotransmitter from the synaptic cleft (Clements, 1996; Suwa et al. 2001).

Our observations also contrast with previously published data on GABAergic synapses showing that a parabolic relationship between mIPSC variance and mean amplitude was correlated with postsynaptic receptor saturation (Auger & Marty, 1997). Furthermore, the mIPSC amplitude distribution of the ‘slow’ bursts followed a Gaussian distribution with low CV (0.22). In glutamatergic synapses, non-saturation of postsynaptic receptors was, on the opposite, associated with a large variation in synaptic event amplitudes (Liu et al. 1999; Hanse & Gustafsson, 2001). Taken together, our results could indicate that there is little variation in the amount of neurotransmitters released at these glycinergic synapses. Two hypotheses could explain such a limited variation in neurotransmitter release: either there are little size variations in vesicle volume or the release of the vesicle content is incomplete as may occur in the ‘kiss and run’ release mechanism where rapid openings and closures of a transient fusion pore (Almers & Tse, 1990; Fesce et al. 1994; Burgoyne & Barclay, 2002) between synaptic vesicles and the plasma membrane prevent the full release of the vesicle contents (Machado et al. 2000; Graham et al. 2000).

Some ‘slow’ bursts had a large mIPSC amplitude variance, which precluded non-stationary noise analysis. The non-parabolic relationship between the variance and the mean amplitude of mIPSCs would suggest that these complex bursts could either reflect fluctuations in the amount of neurotransmitters released per vesicle (variable Po) and/orthe synchronous activity of several independent active zones and/orsynaptic boutons facing different numbers of available postsynaptic receptors (variable N).

Nevertheless, it also has to be mentioned that some of these bursts have very large mean mIPSC amplitudes (up to 500 pA), which could be the hallmark of a large number of available postsynaptic GlyRs. Such large GlyR clusters facing extended active zones are known to occur in the goldfish hindbrain (Triller et al. 1985, 1990) and have been suggested in the mammalian spinal cord (Alvarez et al. 1997) and in the zebrafish hindbrain (Suwa et al. 2001).

‘Fast’ bursts and mechanisms of action of α-LTX

The inter-event time interval can strongly vary among α-LTX-evoked bursts or even within a single burst. Such variations could depend on the multiple mechanisms of action of α-LTX on neurotransmitter release (Ashton et al. 2001). The first of these mechanisms is independent of the external Ca2+ concentration (Ashton et al. 2001) and has been proposed to reflect neurotransmitter efflux through α-LTX pores, as described for noradrenaline from synaptosomes (Ashton et al. 2001). It was, however, not observed for GABA (Storchak et al. 2002) and cannot explain Ca2+-independent fast synaptic responses, as previously observed in a variety of preparations (Longenecker et al. 1970; Rosenthal & Meldolesi, 1989; Capogna et al. 1996; Khvotchev & Sudhof, 2000), unless the pore is formed within the synaptic cleft. The two other mechanisms are Ca2+-dependent and one of them has been proposed to correspond to a Ca2+ entry through the α-LTX-forming membrane pore. This mechanism is unlikely to explain the occurrence of both ‘slow’ and ‘fast’ bursts in our preparation, since these α-LTX-induced bursts were also observed in the absence of external Ca2+.

The third mechanism of α-LTX action, which involves presynaptic intracellular calcium stores (Ashton et al. 2001) and has been proposed to evoke fast neurotransmitter release by acting on readily releasable vesicles (Ashton et al. 2001), is thus the only one that can be taken into account in our model. In such a case, the distinction between ‘slow’ and ‘fast’ bursts could stem from intrinsic differences in release probabilities or from different metabolic states of synaptic boutons. mIPSC amplitude fluctuations are wide both within and, on average, among ‘fast’ bursts. Again, although these bursts can be well isolated, their heterogeneity could have different origins.

Our simulations predict a 33–55 % underestimation of Imean/Imax if only two independent release sites were simultaneously active. Imean/Imax varied from 0.146 to 0.611 among bursts and we therefore cannot ascertain that the bursts with low Imean/Imax truly reflected the activity of a single active zone. In fact, two groups of ‘fast’ bursts could be identified: one with a low Imean/Imax (0.23) and a large CV (0.59) and another group with a higher Imean/Imax (0.485) and a significantly lower CV (0.39; unpaired t test P < 0.01). According to our simulations, the real mean Imean/Imax within ‘fast’ bursts with lower CV should range from 0.724 to more than 1 if they reflect the activity of two independent release sites. A Imean/Imax higher than 1 is unrealistic. The lower corrected value was calculated from simulations supposing that only Po varied among release sites while the number of available GlyRs remained constant. This is unlikely to be the case as GlyR cluster size varies considerably in the zebrafish brain stem (Suwa et al. 2001). Moreover, both N and Po varied among release sites, as shown for ‘slow’ bursts with low CV. This class of ‘fast’ bursts may therefore reflect the activity of a single active zone. If so, the relatively low mIPSCs amplitude fluctuation (CV = 0.39) could reflect variations in the number of neurotransmitter molecules released, as previously proposed for glutamatergic synapses when postsynaptic receptors are not saturated (Liu et al. 1999; Hanse & Gustafsson, 2001).

Physiological implications

The present findings give new insights into the heterogeneity of the functional properties of glycinergic synapses. First, they indicate that different vesicular release mechanisms could occur at inhibitory glycinergic synapses and that mIPSC amplitudes can strongly vary from one group of synapses to another. Second, the lack of GlyR saturation has a physiological significance in that it might allow the control of the efficacy of inhibitory synapses by allosteric modulators such as zinc (Suwa et al. 2001).

Acknowledgments

This work was supported by INSERM, CNRS (P. Legendre and C. I. Badiu) and INSERM-CFB agreement (P. Legendre and J. M. Rigo). C. I. Badiu was supported by a Ministère de la Recherche fellowship. J. M. Rigo was supported by a CNRS fellowship.

REFERENCES

  1. Ali DW, Drapeau P, Legendre P. Development of spontaneous glycinergic currents in the Mauthner neuron of the zebrafish embryo. J Neurophysiol. 2000;84:1726–1736. doi: 10.1152/jn.2000.84.4.1726. [DOI] [PubMed] [Google Scholar]
  2. Almers W, Tse FW. Transmitter release from synapses: does a preassembled fusion pore initiate exocytosis. Neuron. 1990;4:813–818. doi: 10.1016/0896-6273(90)90134-2. [DOI] [PubMed] [Google Scholar]
  3. Alvarez FJ, Dewey DE, Harrington DA, Fyffe RE. Cell-type specific organization of glycine receptor clusters in the mammalian spinal cord. J Comp Neurol. 1997;379:150–170. [PubMed] [Google Scholar]
  4. Ashton AC, Volynski KE, Lelianova VG, Orlova EV, Van Renterghem C, Canepari M, Seagar M, Ushkaryov YA. α-Latrotoxin, acting via two Ca2+-dependent pathways, triggers exocytosis of two pools of synaptic vesicles. J Biol Chem. 2001;276:44695–44703. doi: 10.1074/jbc.M108088200. [DOI] [PubMed] [Google Scholar]
  5. Auger C, Marty A. Heterogeneity of functional synaptic parameters among single release sites. Neuron. 1997;19:139–150. doi: 10.1016/s0896-6273(00)80354-2. [DOI] [PubMed] [Google Scholar]
  6. Bekkers JM, Richerson GB, Stevens CF. Origin of variability in quantal size in cultured hippocampal neurons and hippocampal slices. Proc Natl Acad Sci U S A. 1990;87:5359–5362. doi: 10.1073/pnas.87.14.5359. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Burgoyne RD, Barclay JW. Splitting the quantum: regulation of quantal release during vesicle fusion. Trends Neurosci. 2002;25:176–178. doi: 10.1016/s0166-2236(00)02126-3. [DOI] [PubMed] [Google Scholar]
  8. Capogna M, Gahwiler BH, Thompson SM. Calcium-independent actions of α-latrotoxin on spontaneous and evoked synaptic transmission in the hippocampus. J Neurophysiol. 1996;76:3149–3158. doi: 10.1152/jn.1996.76.5.3149. [DOI] [PubMed] [Google Scholar]
  9. Clements JD. Transmitter timecourse in the synaptic cleft: its role in central synaptic function. Trends Neurosci. 1996;19:163–171. doi: 10.1016/s0166-2236(96)10024-2. [DOI] [PubMed] [Google Scholar]
  10. De Koninck Y, Mody I. Noise analysis of miniature IPSCs in adult rat brain slices: properties and modulation of synaptic GABA(A) receptor channel. J Neurophysiol. 1994;71:1318–1335. doi: 10.1152/jn.1994.71.4.1318. [DOI] [PubMed] [Google Scholar]
  11. Fesce R, Grohovaz F, Valtorta F, Meldolesi J. Neurotransmitter release: fusion or ‘kiss and run’. Trends Cell Biol. 1994;4:1–4. doi: 10.1016/0962-8924(94)90025-6. [DOI] [PubMed] [Google Scholar]
  12. Fesce R, Segal JR, Ceccarelli B, Hurlbut WP. Effects of black widow spider venom and Ca2+ on quantal secretion at the frog neuromuscular junction. J Gen Physiol. 1986;88:59–81. doi: 10.1085/jgp.88.1.59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Frerking M, Borges S, Wilson M. Variation in GABA mini amplitude is the consequence of variation in transmitter concentration. Neuron. 1995;15:885–895. doi: 10.1016/0896-6273(95)90179-5. [DOI] [PubMed] [Google Scholar]
  14. Frerking M, Wilson M. Saturation of postsynaptic receptors at central synapses. Curr Opin Neurobiol. 1996;6:395–403. doi: 10.1016/s0959-4388(96)80125-5. [DOI] [PubMed] [Google Scholar]
  15. Hamill OP, Marty A, Neher E, Sakmann B, Sigworth FJ. Improved patch-clamp techniques for high-resolution current recording from cells and cell-free membrane patches. Pflugers Arch. 1981;391:85–100. doi: 10.1007/BF00656997. [DOI] [PubMed] [Google Scholar]
  16. Hanse E, Gustafsson B. Quantal variability at glutamatergic synapses in area CA1 of the rat neonatal hippocampus. J Physiol. 2001;531:467–480. doi: 10.1111/j.1469-7793.2001.0467i.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Khvotchev M, Sudhof TC. α-Latrotoxin triggers transmitter release via direct insertion into the presynaptic plasma membrane. EMBO J. 2000;19:3250–3262. doi: 10.1093/emboj/19.13.3250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kruk PJ, Korn H, Faber DS. The effects of geometrical parameters on synaptic transmission: a Monte Carlo simulation study. Biophys J. 1997;73:2874–2890. doi: 10.1016/S0006-3495(97)78316-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Legendre P. Pharmacological evidence for two types of postsynaptic glycinergic receptors on the Mauthner cell of 52-h-old zebrafish larvae. J Neurophysiol. 1997;77:2400–2415. doi: 10.1152/jn.1997.77.5.2400. [DOI] [PubMed] [Google Scholar]
  20. Legendre P. A reluctant gating mode of glycine receptor channels determines the time course of inhibitory miniature synaptic events in zebrafish hindbrain neurons. J Neurosci. 1998;18:2856–2870. doi: 10.1523/JNEUROSCI.18-08-02856.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Legendre P. The glycinergic inhibitory synapse. Cell Mol Life Sci. 2001;58:760–793. doi: 10.1007/PL00000899. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Legendre P, Korn H. Glycinergic inhibitory synaptic currents and related receptor channels in the zebrafish brain. Eur J Neurosci. 1994;6:1544–1557. doi: 10.1111/j.1460-9568.1994.tb00545.x. [DOI] [PubMed] [Google Scholar]
  23. Liu G, Choi S, Tsien RW. Variability of neurotransmitter concentration and nonsaturation of postsynaptic AMPA receptors at synapses in hippocampal cultures and slices. Neuron. 1999;22:395–409. doi: 10.1016/s0896-6273(00)81099-5. [DOI] [PubMed] [Google Scholar]
  24. Longenecker HE, Jr, Hurlbut WP, Mauro A, Clark AW. Effects of black widow spider venom on the frog neuromuscular junction. Effects on end-plate potential, miniature end-plate potential and nerve terminal spike. Nature. 1970;225:701–703. doi: 10.1038/225701a0. [DOI] [PubMed] [Google Scholar]
  25. Machado JD, Segura F, Brioso MA, Borges R. Nitric oxide modulates a late step of exocytosis. J Biol Chem. 2000;275:20274–20279. doi: 10.1074/jbc.M000930200. [DOI] [PubMed] [Google Scholar]
  26. Manabe T, Renner P, Nicoll RA. Postsynaptic contribution to long-term potentiation revealed by the analysis of miniature synaptic currents. Nature. 1992;355:50–55. doi: 10.1038/355050a0. [DOI] [PubMed] [Google Scholar]
  27. Metcalfe WK, Mendelson B, Kimmel CB. Segmental homologies among reticulospinal neurons in the hindbrain of the zebrafish larva. J Comp Neurol. 1986;251:147–159. doi: 10.1002/cne.902510202. [DOI] [PubMed] [Google Scholar]
  28. Nusser Z, Cull-Candy S, Farrant M. Differences in synaptic GABA(A) receptor number underlie variation in GABA mini-amplitude. Neuron. 1997;19:697–709. doi: 10.1016/s0896-6273(00)80382-7. [DOI] [PubMed] [Google Scholar]
  29. Oleskevich S, Alvarez FJ, Walmsley B. Glycinergic miniature synaptic currents and receptor cluster sizes differ between spinal cord interneurons. J Neurophysiol. 1999;82:312–319. doi: 10.1152/jn.1999.82.1.312. [DOI] [PubMed] [Google Scholar]
  30. Perrais D, Ropert N. Effect of zolpidem on miniature IPSCs and occupancy of postsynaptic GABA(A) receptors in central synapses. J Neurosci. 1999;19:578–588. doi: 10.1523/JNEUROSCI.19-02-00578.1999. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Poncer JC, Durr R, Gahwiler BH, Thompson SM. Modulation of synaptic GABA(A) receptor function by benzodiazepines in area CA3 of rat hippocampal slice cultures. Neuropharmacology. 1996;35:1169–1179. doi: 10.1016/s0028-3908(96)00055-x. [DOI] [PubMed] [Google Scholar]
  32. Pumplin DW, del Castillo J. Release of packets of acetylcholine and synaptic vesicle elicited by brown widow spider venom in frog motor nerve endings poisoned by botulinum toxin. Life Sci. 1975;17:137–141. doi: 10.1016/0024-3205(75)90249-0. [DOI] [PubMed] [Google Scholar]
  33. Rosenthal L, Meldolesi J. α-Latrotoxin and related toxins. Pharmacol Ther. 1989;42:115–134. doi: 10.1016/0163-7258(89)90024-7. [DOI] [PubMed] [Google Scholar]
  34. Rumpel E, Behrends JC. Postsynaptic receptor occupancy during evoked transmission at striatal GABAergic synapses in vitro. J Neurophysiol. 2000;84:771–779. doi: 10.1152/jn.2000.84.2.771. [DOI] [PubMed] [Google Scholar]
  35. Sigworth FJ. The variance of sodium current fluctuations at the node of Ranvier. J Physiol. 1980;307:97–129. doi: 10.1113/jphysiol.1980.sp013426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Silver RA, Traynelis SF, Cull-Candy SG. Rapid-time-course miniature and evoked excitatory currents at cerebellar synapses in situ. Nature. 1992;355:163–166. doi: 10.1038/355163a0. [DOI] [PubMed] [Google Scholar]
  37. Singer JH, Berger AJ. Contribution of single-channel properties to the time course and amplitude variance of quantal glycine currents recorded in rat motoneurons. J Neurophysiol. 1999;81:1608–1616. doi: 10.1152/jn.1999.81.4.1608. [DOI] [PubMed] [Google Scholar]
  38. Storchak LG, Linetska MV, Himmelreich NH. Does extracellular calcium determine what pool of GABA is the target for α-latrotoxin. Neurochem Int. 2002;40:387–395. doi: 10.1016/s0197-0186(01)00107-3. [DOI] [PubMed] [Google Scholar]
  39. Suwa H, Saint-amant L, Triller A, Drapeau P, Legendre P. High-affinity zinc potentiation of inhibitory postsynaptic glycinergic currents in the zebrafish hindbrain. J Neurophysiol. 2001;85:912–925. doi: 10.1152/jn.2001.85.2.912. [DOI] [PubMed] [Google Scholar]
  40. Traynelis SF, Silver RA, Cull-Candy SG. Estimated conductance of glutamate receptor channels activated during EPSCs at the cerebellar mossy fiber-granule cell synapse. Neuron. 1993;11:279–289. doi: 10.1016/0896-6273(93)90184-s. [DOI] [PubMed] [Google Scholar]
  41. Triller A, Cluzeaud F, Pfeiffer F, Betz H, Korn H. Distribution of glycine receptors at central synapses: an immunoelectron microscopy study. J Cell Biol. 1985;101:683–688. doi: 10.1083/jcb.101.2.683. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Triller A, Rostaing P, Korn H, Legendre P. Morphofunctional evidence for mature synaptic contacts on the Mauthner cell of 52-hour-old zebrafish larvae. Neuroscience. 1997;80:133–145. doi: 10.1016/s0306-4522(97)00092-4. [DOI] [PubMed] [Google Scholar]
  43. Triller A, Seitanidou T, Franksson O, Korn H. Size and shape of glycine receptor clusters in a central neuron exhibit a somato-dendritic gradient. New Biol. 1990;2:637–641. [PubMed] [Google Scholar]
  44. Ushkaryov Y. α-Latrotoxin: from structure to some functions. Toxicon. 2002;40:1–5. doi: 10.1016/s0041-0101(01)00204-5. [DOI] [PubMed] [Google Scholar]
  45. Van Renterghem C, Iborra C, Martin-Moutot N, Lelianova V, Ushkaryov Y, Seagar M. α-Latrotoxin forms calcium-permeable membrane pores via interactions with latrophilin or neurexin. Eur J Neurosci. 2000;12:3953–3962. doi: 10.1046/j.1460-9568.2000.00282.x. [DOI] [PubMed] [Google Scholar]

Articles from The Journal of Physiology are provided here courtesy of The Physiological Society

RESOURCES