Quantal currents at single-site central synapses (original) (raw)

J Physiol. 2000 Jul 1; 526(Pt 1): 3–11.

A Marty

*Arbeitsgruppe Zelluläre Neurobiologie, Max-Planck-Institut für biophysikalische Chemie, Göttingen, Germany

Department of Physiology, University College London, London, UK

*Arbeitsgruppe Zelluläre Neurobiologie, Max-Planck-Institut für biophysikalische Chemie, Göttingen, Germany

Corresponding author C. Auger: Department of Physiology, University College London, Gower Street, London WC1 E6BT, UK. Email: ku.ca.lcu@aecbgcu

Received 2000 Jan 17; Accepted 2000 Apr 28.

Abstract

The mode of operation of synaptic transmission has been primarily worked out at the vertebrate neuromuscular junction, thus providing a framework for the interpretation of studies at central synapses. However, differences have been found between the two systems, and a coherent model is still lacking for central synapses. Research in this area revolves around several questions. (1) Is the variability of quantal amplitudes determined pre- or postsynaptically? (2) What is the occupancy of postsynaptic receptors following the release of a synaptic vesicle? And (3) does multivesicular release occur at single release sites following one presynaptic action potential? To answer these questions, it is essential to investigate synaptic processes at the level of single release sites. This is technically difficult because of the complex morphology and small dimensions of central synapses. Nevertheless significant advances have been made in the past few years.

Miniature synaptic currents originating at a single release site

The issue

The amplitude distribution of miniature synaptic currents at the neuromuscular junction is described by a single Gaussian curve. However, in CNS neurones it is typically highly skewed, with most events squeezing into the first amplitude bins, while a significant proportion exhibit amplitudes that are 10 or 20 times higher than the mean. The exact origin of this amplitude scatter is still unclear (reviewed in Bekkers, 1994). One basic and controversial issue is whether the large variation observed comes from an intrinsic variability of quantal sizes at each release site), or whether it primarily reflects the participation of many release sites with diverse characteristics.

To address this problem, it is necessary to record synaptic currents coming from a single release site. If the variability of synaptic currents coming from a single release site is much smaller than that of the population of currents coming from all the release sites, its origin must lie in differences between release sites. A variable quantal amplitude between release sites could account for such observations. Alternatively, differential distortion of synaptic currents, due to different electrical distances between the postsynaptic sites and the recording electrode, is also expected to increase the spread of the synaptic currents’ amplitudes. If on the other hand the variability of currents coming from all the release sites is similar to that of the currents coming from a single release site, then it must reflect variations occurring at individual release sites.

Various approaches that have been developed recently to restrict measurements of miniature synaptic currents to single release sites are reviewed here.

Local perfusion

Miniature currents coming from a single release site were studied initially with local applications of hyperosmotic or K+-rich (hyperkalaemic) solutions in hippocampal cultures (Bekkers et al. 1990; Tang et al. 1994; Liu & Tsien, 1995) and slices (Bekkers et al. 1990). The preparation is bathed in a solution which does not sustain transmitter release (typically containing a very low Ca2+ concentration) and therefore keeps spontaneous synaptic activity at its lowest. Hyperosmotic solutions are well known to increase the frequency of miniature synaptic currents, although their mode of action is not understood. Hyperkalaemic solutions on the other hand depolarise synaptic terminals and evoke a Ca2+-dependent increase in the frequency of miniature synaptic currents. Local applications of hyperosmotic or hyperkalaemic solutions can therefore be used to trigger release from a single or a few release sites. The difficulty of the method is to ascertain that a single release site is stimulated. The localisation of the stimulated release sites may be studied by labelling presynaptic proteins immunocytochemically (Bekkers et al. 1990) or by using the fluorescent indicator FM1-43 (Liu & Tsien, 1995). However, neither of these two staining methods works satisfactorily in slices, so that most studies have been limited to neuronal cultures. In two cases (Bekkers et al. 1990; Liu & Tsien, 1995), the amplitude distribution of the miniature currents observed with this method was similar to the amplitude distribution of the miniatures coming from all the release sites. The coefficient of variation (CV =s.d./mean) was high for both local and population distributions, in the order of 40-50 %. The large variance of the amplitude distribution was therefore concluded to arise from variations in the neurotransmitter content of synaptic vesicles.

However, using the same approach, Tang et al. (1994) observed high frequency bursts of mEPSCs in some experiments. Within a burst, isolated events had similar peak amplitudes. The amplitude distribution was Gaussian and the CV extremely low (6 % in one case). Amplitudes of overlapping events summed sublinearly, indicating that these currents arose from the same set of postsynaptic receptors, and therefore that they came from a single release site. The conclusion from this work was that the variability of miniature currents comes from differences between release sites. However, the results of this study may have been influenced by the use of cyclothiazide, which may have increased the occupancy of postsynaptic receptors and thus reduced the variability among EPSCs coming from single sites.

Loose patch clamp recording from single boutons

Due to the unavoidable limitations of the local perfusion approach, alternative methods to restrict recording to a single site are obviously desirable. An elegant solution to this problem was recently developed in hippocampal cultured neurones (Forti et al. 1997). Electron microscopy showed that in this preparation, synaptic boutons contain a single release site. Postsynaptic signals coming from a single release site were isolated by directly recording from synaptic boutons contained in the recording pipette (‘loose patch clamp’) in the presence of tetrodotoxin. The amplitude distribution of miniature currents recorded in these conditions was Gaussian. When the synaptic currents were recorded simultaneously in the soma in whole-cell configuration, the amplitude distribution was also Gaussian, even though the kinetics of the whole-cell currents appeared to be filtered. The variability of the single sites’ EPSCs was smaller than the variability of the general amplitude distribution of the miniature currents (respective mean CV values, 28 and 72 %; Fig. 1). Therefore, the main source of the population variability resided in differences among release sites in quantal sizes and/or electrical distances to the soma. This conclusion is in accord with one (Tang et al. 1994), but opposite to the other two studies (Bekkers et al. 1990; Liu & Tsien, 1995) using local perfusion in the same preparation (see above). One possibility to explain the discrepant results is that some of the synaptic parameters are affected by the culture conditions, giving rise to a variability not representative of the in vivo situation. Another possibility that cannot be excluded is that the local perfusion method can lead inadvertently to the stimulation of several release sites.

An external file that holds a picture, illustration, etc. Object name is tjp0526-0003-f1.jpg

Loose-patch recording of single quanta at individual hippocampal synapses

A, individual boutons are visualised along dendrites of cultured hippocampal neurones with the exocytosis-endocytosis marker FM1-43. A blunt pipette is placed on a well-isolated bouton for recording. B, during loose-patch recording, the pipette exclusively collects signals originating in the enclosed bouton. C, simultaneous local loose-patch recording (lp, upper trace) and somatic whole-cell recording (WC, lower trace). Miniature EPSCs originating in the enclosed bouton give rise to simultaneous, proportionally sized events in the lp and WC traces. D, amplitude histograms for simultaneously recorded signals from the lp and WC traces. Both histograms have Gaussian distributions and have similar CVs. Inset, cumulative histograms for the same data, showing in addition the population histogram recorded in the whole-cell trace. Note that the population histogram is much more broadly distributed than the histogram corresponding to the single bouton events. From Forti et al. 1997. Reprinted by permission from Nature (388,874-878) copyright 1997 Macmillan Magazines Ltd.

Dinapses

The variability of miniature IPSCs was studied in cultured amacrine cells (Frerking et al. 1995). Pairs of these cells form special types of synapses (dinapses), where the released neurotransmitter binds to receptors located on the presynaptic cell as well as to those belonging to a neighbouring neurone. Dinaptic currents were isolated from the synaptic background by their coincident detection in the two cells. The amplitude of the two simultaneous currents was correlated, suggesting that the source of the variability was presynaptic. Given the total number of release sites on those cells and the relative proportion of dinaptic currents, it was estimated that pairs of cultured amacrine cells possess at most two dinapses. Therefore variations of the number of postsynaptic receptors at dinapses could not generate the observed distribution and it was concluded that the skew of the amplitude distribution comes from variations in neurotransmitter content of the synaptic vesicles.

α-Latrotoxin

The black widow spider toxin, α-latrotoxin, has long been known to trigger vesicular release of neurotransmitter. The release is continuous and Ca2+ independent, or it occurs in bursts in a Ca2+-dependent manner. Both types of release may coexist in the same preparation. When applying the toxin at extremely low concentrations to cerebellar slices, high frequency bursts of mIPSCs were observed in stellate and basket cells, occurring randomly on a low background activity (Auger & Marty, 1997). As in the local perfusion study by Tang et al. (1994), peak amplitudes were fairly constant within a given burst. Overlapping events summed to less than twice the mean of individual events, i.e. sublinearly, and their amplitude distribution was well described by a single Gaussian curve. These observations indicated that events in a burst came from a single release site.

The CV of burst distributions was much lower than that of the whole population of miniature IPSCs recorded in these cells (respective mean values, 13 and 80 %). Moreover, the average amplitude varied from burst to burst over a factor of 10-fold. This indicates that most of the variability of miniature amplitudes in whole-cell recordings comes from differences between release sites.

It appears from this study that α-latrotoxin can be used to generate single-site miniature events at some central synapses. The toxin suddenly activates a burst of release at one site, increasing the release rate by orders of magnitude during a period of about 1 min. Separation among sites is simply provided by the stochastic nature of toxin action and at low toxin concentration, the probability of recording from more than one release site in the same time window is very low.

Summary

The above results clearly show that in some preparations the amplitude distribution of miniature currents at single release sites is described well by single Gaussian functions of relatively low CV. Thus differences among the mean amplitude at different synapses generates most of the scatter observed in whole-cell recordings. This conclusion is in line with a growing body of evidence indicating that the number of postsynaptic receptors varies greatly from one site to the next (Nusser et al. 1997; Schikorski & Stevens, 1997; reviewed in Walmsley et al. 1998; Nusser, 1999). However, additional work is clearly needed to see how general these conclusions are.

Synaptic currents evoked at single release sites by action potentials

Several recent studies have focused on evoked synaptic currents coming from a single release site, notably at the excitatory synapse between pyramidal cells and inhibitory interneurones of the CA3 region of the hippocampus (Gulyás et al. 1993; Arancio et al. 1994), at the excitatory synapse between mossy fibres and granule cells in the cerebellum (Silver et al. 1996) and at the inhibitory synapse between interneurones of the molecular layer of the cerebellum (Auger et al. 1998). In one study, the involvement of a single release site was demonstrated unambiguously using electron microscopy and serial reconstruction (Gulyás et al. 1993). In the other studies less direct evidence was obtained. The release probability was altered by changing the extracellular Ca2+ and Mg2+ concentrations and evidence for a single release site was based on the following argument. If multiple sites release neurotransmitter synchronously, decreasing the release probability will decrease the average number of sites releasing and shift the distribution towards small amplitudes. If, on the other hand, a single release site is involved, the number of release failures will increase, and the average amplitude including failures will therefore decrease, but the average amplitude of successful responses will be fundamentally unaffected (Arancio et al. 1994; Silver et al. 1996; Auger et al. 1998). Based on this criterium, single-site synapses were identified in all three preparations tested. In each case, the amplitude distribution of evoked currents coming from a single release site was Gaussian. Mean CV values were around 30 % in hippocampal synapses (Gulyás et al. 1993; Arancio et al. 1994), 23 % in cerebellar granule cells (Silver et al. 1996) and 19 % in cerebellar interneurones (Auger et al. 1998).

Another set of results concerns hippocampal CA3-CA1 synapses, which mostly involve single-site contacts (Shepherd & Harris, 1998). In this preparation some conflicting results were obtained. ‘Minimal stimulation’ protocols yielded either skewed (e.g. Raastad et al. 1992) or Gaussian (e.g. Bolshakov & Siegelbaum, 1995; Dobrunz & Stevens, 1997) distributions. The discrepancy may be due to the difficulty in avoiding contamination of the recordings from multisite stimulation, as discussed in Raastad (1995). Since amplitude distributions appear to be Gaussian when using paired recordings (Bolshakov & Siegelbaum, 1995), the single-site distribution is most probably Gaussian in this preparation as well.

Overall, these results indicate that the distribution of miniature and evoked currents at single sites are similar, and that both are well described by a single Gaussian distribution. This seems to hold true for both excitatory and inhibitory synapses.

What is the degree of occupancy of postsynaptic receptors after neurotransmitter release?

The degree of occupancy of postsynaptic receptors following the release of the vesicular content is also an area of deep controversy (Frerking & Wilson, 1996). Unlike an active zone at the neuromuscular junction, each release site of CNS synapses is associated with a well-defined postsynaptic density containing a small number of postsynaptic receptors. Depending on the occupancy of the receptors following the release of a single vesicle of neurotransmitter, the synapse could operate either in a functional binary mode (failure vs. exocytosis of at least one vesicle) or if receptor occupancy is low, in a graded mode where the postsynaptic response grows accordingly to the number of quanta of transmitter released.

The competitive antagonist method

The issue of postsynaptic receptor occupancy was first addressed in two studies at excitatory synapses between hippocampal neurones in culture. Clements et al. (1992) introduced the use of a low affinity competitive antagonist to estimate the neurotransmitter concentration in the synaptic cleft. The displacement of the antagonist by the released neurotransmitter was measured and modelled with a kinetic scheme using estimated _K_on and _K_off values for the antagonist. It was estimated from those simulations that the transmitter concentration reaches about 1.1 mM (3 mM in a new analysis of the same data: Clements, 1996). Given the estimated time course of the glutamate transient, this concentration should be sufficient to saturate NMDA receptors. On the other hand, the AMPA/KA receptors which display a lower affinity for glutamate should reach about 60 % occupancy. Two other modelling studies are consistent with this estimate for AMPA/KA receptors (Holmes, 1995; Wahl et al. 1996), but one of them suggests that NMDA receptors may not be saturated if uptake is taken into consideration (Holmes, 1995).

Tong & Jahr (1994) also used an approach based on competition for receptor binding to study variations of the neurotransmitter concentration with release probability. The degree of inhibition by a low affinity competitive antagonist (in this case L-APV) depends on the neurotransmitter concentration competing for binding to the receptors. The inhibition by L-APV was shown to decrease with increasing release probability (e.g. with a higher extracellular Ca2+ concentration), indicating a build-up of the neurotransmitter concentration in the synaptic cleft. It was concluded that several vesicles can be released at a single release site. Moreover, the relative proportions of NMDA and AMPA/KA currents were maintained under high release probability conditions. Since the affinity of the two kinds of receptors are different it was concluded that both were saturated in control conditions. This study shows that the synaptic glutamate concentration increases with the release probability. However, it does not permit a distinction between multivesicular release or increased spillover from neighbouring synapses (Rusakov et al. 1999). Furthermore the analysis assumed similar properties for synaptic NMDA channels and for extrasynaptic channels studied in outside-out patches, whereas a subsequent study in hippocampal autaptic cultures indicates the contrary (Rosenmund et al. 1995). Finally, a recent study based on single-site optical recording in hippocampal slices led to the opposite conclusion, namely that NMDA receptors are far from saturated (Mainen et al. 1999).

The benzodiazepine method

At inhibitory synapses benzodiazepine-like molecules such as zolpidem have been used to estimate the degree of occupancy of postsynaptic receptors. These drugs increase the affinity of GABAA receptors for GABA. If postsynaptic receptors are already saturated by the neurotransmitter released in the synaptic cleft, the amplitude of the synaptic currents will not be modified by benzodiazepines. On the other hand, if the postsynaptic receptors are not saturated, the increased affinity will result in an increased peak amplitude of the currents.

It has been shown in several preparations (including hippocampal granule cells: De Koninck & Mody, 1994; CA3 pyramidal cells: Poncer et al. 1996) that benzodiazepines do not increase the peak amplitude of mIPSCs. Benzodiazepines do slow down the decay of synaptic currents, indicating that GABA molecules stay bound to the receptor for a longer period. In these preparations, it was therefore concluded that postsynaptic GABA receptors are saturated by the neurotransmitter released at the synapse. In other preparations, however (including cultured amacrine cells: Frerking et al. 1995; cerebellar stellate cells: Nusser et al. 1997; cultured cerebellar granule cells: Mellor & Randall, 1997; and cortical slices: Perrais & Ropert, 1999), an increase in the peak amplitude of mIPSCs was observed, indicating an incomplete occupancy of the postsynaptic receptors.

The benzodiazepine method is convenient and valuable, but it rests on assumptions on the mechanisms of action of benzodiazepines that may hold true at some synapses and not at others, given the variety of GABAergic receptors in the CNS (see discussion in Perrais & Ropert, 1999). Also, unless a time-consuming calibration using somatic receptors is used, it is not quantitative (Perrais & Ropert, 1999). Perrais & Ropert furthermore observed that, whereas zolpidem increases the amplitude of the response to subsaturating GABA concentrations at room temperature, it fails to do so at physiological temperature. These results suggest that the benzodiazepine method may give erroneous conclusions if applied at physiological temperatures, as in some of the earlier studies, concluding that receptors are saturated (e.g. De Koninck & Mody, 1994; however, Poncer et al. 1996, as well as Mellor & Randall, 1997, describe results at room temperature).

The CV method

The coefficient of variation of amplitude distributions is also indicative of the neurotransmitter concentration reached in the synaptic cleft. If a large fraction of the postsynaptic receptors is occupied, variations of the neurotransmitter concentration cannot give rise to major variations in amplitude of the synaptic currents, and the stochastic properties of the postsynaptic receptors is the main source of variance. Assuming independent postsynaptic channels with common stochastic properties, the fluctuations from response to response has a CV given by:

equation image

where _P_o is the peak opening probability, and _N_o is the number of channels opened at the peak of the synaptic current. (_P_o is the product of occupancy with the probability that bound channels are open. _N_o is the product of _P_o with the total number of channels present at the site.) At central synapses _N_o is in the range 10-100. For small and medium (≤ 0.5) values of _P_o, application of the above equation predicts CV values of between 7 and 32 %. CV values below 7 % indicate _P_o values above 0.5, and occupancy values even higher. Only one study at single release sites has reported a CV value in this range, sufficiently small to be attributed to the noise of fully occupied channels (Tang et al. 1994; results obtained in the presence of cyclothiazide). In the other studies where Gaussian distributions were reported, there is a clear additional variance with a CV in the range 15-30 %, indicating incomplete occupancy of postsynaptic receptors. This applies both to excitatory (Gulyás et al. 1993; Arancio et al. 1994; Silver et al. 1996; Forti et al. 1997) and to inhibitory synapses (Auger & Marty, 1997). More specifically, Silver et al. (1996) obtained an upper estimate of 0.6 for the peak channel occupancy in cerebellar granule cells, based on a comparison between mean and maximum EPSC values.

The conclusions of the recent studies at single sites will be difficult to reconcile with the results of quantal analysis performed over the last two decades at multisite synapses. In these studies, amplitude distributions were fitted with multiple, regularly spaced Gaussian curves. The single-site variance estimated from these fits was generally small compared with the noise of the recordings, requiring CV values at individual sites substantially lower than 10 % and implying that the postsynaptic receptors are saturated or close to saturation (Jack et al. 1981; Edwards et al. 1990; Schneggenburger & Konnerth, 1992; Kraszewski & Grantyn, 1992; Jonas et al. 1993). An intriguing recent study carried out at excitatory synapses onto cerebellar granule cells indicates that some of the discrepancy can be attributed to differences in development (Wall & Usowicz, 1998). In this preparation, the distribution of quantal sizes is broad for immature synapses (Silver et al. 1996), but it is narrow in the adult (Wall & Usowicz, 1998).

The occlusion method

Examination of synaptic currents at single-site synapses allows another measurement of synaptic current occupancy. The method was first used by Tang et al. (1994) in hippocampal cultures. Its principle is illustrated in Fig. 2_A_. Suppose that two release events occur with an interval δ_t._ The first release results in a synaptic current having an amplitude A_1, and in the occupancy of a fraction ω of the postsynaptic receptors. ω relaxes back to 0 as the neurotransmitter dissociates from the receptor so that at time δ_t its value is ω (δ_t_), lower than ωo, the peak occupancy. The second release event gives rise to a current increment, A′_2, which is the product of A_1 with the fraction of receptors that are available at time δ_t. This fraction is 1 –ω(δ_t), so that:

equation image

An external file that holds a picture, illustration, etc. Object name is tjp0526-0003-f2.jpg

Measuring receptor occupancy from analysis of closely interspaced miniature events

A, principle of the measurement. Bursts of mIPSCs are elicited by a liminal application of α-LTX. Events that occur during the decay of a preceding mIPSC do not sum linearly with it; rather, they appear to have lower amplitudes than isolated events. For short values of the interval δ_t_, the foot-to-peak amplitude _A′_2 of the second event is smaller than that of the first event, while the baseline-to-peak amplitudes (_A_2 and _A_1) are similar. B, plot of A_2 as a function of δ_t, showing no significant increase at short time intervals. C, a plot of A′_2 as a function of δ_t can be approached with the sum of two exponentials. The slow exponential has a small amplitude and reflects cumulated desensitization of receptors during high frequency periods; the fast exponential component has a larger amplitude and reflects competition of successive release events for a common pool of postsynaptic receptors. The peak occupancy of postsynaptic receptors, ωo, can be calculated from the extrapolated amplitudes for each component, A and A′, by using the equation in the figure. Reproduced with permission from Auger & Marty (1997), copyright Cell Press.

Therefore an extrapolation to the origin of the plot of the ratio A′_2/A_1 as a function of δ_t_ gives the value of ωo, the peak occupancy.

Applying this method to consecutive mIPSCs during high frequency bursts triggered by α-latrotoxin yields the value ωo= 0.76 (Fig. 2_C_; Auger & Marty, 1997).

The occlusion method has obvious advantages compared with the less direct methods described above. It has some limitations though. First, it requires that nearly all analysed currents do arise from a single release site, and it is highly sensitive to contamination from foreign sites. Such contamination may arise if data are obtained with the α-latrotoxin method in a cell that has a substantial background rate of mini discharge. A second limitation is that significant receptor desensitisation occurring on the time scale of the synaptic currents will complicate the analysis.

Summary

An array of methods has been developed to measure receptor occupancy, but there is no clear picture emerging yet. Probably the results will vary according to receptor subtypes and synapse geometry. Available results at single-site excitatory or inhibitory synapses do not indicate full saturation; at inhibitory synapses, the receptor occupancy appears nevertheless to be substantial. It should, however, be stressed that many of these results were obtained at room temperature, and that the situation may be substantially different at physiological temperature.

One site-one vesicle hypothesis

The issue

It was shown in the central nervous system that the number of release sites equals N, one of the parameters of the binomial law used in quantal analysis to describe the amplitude distribution of evoked synaptic currents (see for review Redman, 1990). This identity means that the maximal number of quantal peaks in amplitude distributions is inferior or equal to the number of release sites. In the extreme case where the connection involves a single release site, a single peak is observed (as in Gulyás et al. 1993).

These observations together with electron microscopy structural data have been interpreted by Triller & Korn (1982) as indicating that only one vesicle can be released per release site after the arrival of an action potential (one site- one vesicle hypothesis). Nevertheless, as stressed by the same authors, near-saturation of postsynaptic receptors by the neurotransmitter content of a single vesicle could provide an alternative to the single site-single vesicle hypothesis (Korn et al. 1982). If postsynaptic receptors are close to saturation, multiple vesicular release would result in a small signal increment over monovesicular release as only few additional receptors can be recruited. Single release sites would therefore generate single peak amplitude histograms in spite of multivesicular release. If on the other hand postsynaptic receptor occupancy is significant but far from saturation, multivesicular release will result in multiple peaks at decreasing intervals as the number of free postsynaptic receptors decreases after each vesicular release. In practice, occupancy values of 0.5 or higher will generate amplitude histograms that are virtually indistinguishable from single peak distributions.

Multivesicular release in cerebellar single site recordings

Even if full saturation of postsynaptic receptors occurs, multivesicular release can in principle be demonstrated at single-site synapses. Vesicular release is a time-distributed process: synaptic current onsets are spread over a period of about 1 ms at room temperature (Katz & Miledi, 1965; Isaacson & Walmsley, 1995). Two successive exocytotic events should therefore result in a succession of a large and a small synaptic current, very much as in the case of two closely separated miniature events (Fig. 2). The early part of evoked IPSCs was examined at single-site synapses in cerebellar interneurones and smaller secondary currents were observed (Fig. 3; Auger et al. 1998). The postsynaptic receptors occupancy ωo was calculated using the same approach as for mIPSCs (Fig. 2). ωo for evoked IPSCs was similar to that previously inferred from the analysis of mIPSCs. Thus multivesicular release occurs at this particular synapse.

An external file that holds a picture, illustration, etc. Object name is tjp0526-0003-f3.jpg

Multivesicular release at a single site synapse

A, recording of evoked IPSCs in a cerebellar interneurone following electrical stimulations in the molecular layer. Responses to 50 consecutive stimulations. Note the high proportion of failures (66 %), and the total lack of interference of the results with spontaneous synaptic currents. B, amplitude histograms for the responses in A. This histogram can be fitted with a single Gaussian distribution (CV = 20 %). The histogram of the recording noise is also shown. C, when displayed on a fast time scale, about 5 % of the successful responses in A can be seen to include two events in close succession. D, occurrence histograms of event pairs (as illustrated in C) as a function of interevent interval. The frequency of pairs falls off abruptly over a period of 3 ms. The dashed line illustrates the frequency of pairs expected on the basis of random summation of evoked events with the background activity due to unstimulated synapses. E, A_2/A_1 ratio as a function of interevent interval, from the same data. Reproduced with permission from Auger et al. (1998).

Since the release probability is a function of time, early release events are more likely to be followed by a second release event than late ones. Thus multivesicular IPSCs will tend to have early latencies. Their rise times should be slowed down because of temporal summation of the rising phases of the underlying individual components. They will also have long durations, because larger and longer agonist pulses tend to prolong the decay time of GABAergic currents. Finally, they will have large amplitudes because of the partial summation of successive events (due to the fact that ωo is less than 1). Conversely, monovesicular events will tend to be associated with late first latencies, fast rise times, short decay times and small amplitudes. A combination of substantial receptor occupancy (ωo≡ 0.7) and multivesicular release is therefore expected to generate complex relations between IPSC amplitudes and shape at single-site synapses. Correlations between those synaptic parameters were indeed observed (Kondo & Marty, 1998_b_; Auger et al. 1998).

Due to limited resolution, all multivesicular release events were not detected, so that the method of Fig. 3 gives a severe underestimate of the rate of occurrence of double release events. Better estimates can be derived from analyses of the amplitude distribution or of the correlation between latencies and mean amplitudes at single-site synapses. It was concluded from this analysis that more than 30 % of synaptic events are multiple (Auger et al. 1998).

Conclusion

The concerted efforts deployed in the past years to study single release sites have been already quite successful. One major conclusion emerging from these studies is that the distribution of quantal sizes at single sites is compact and Gaussian and thus differs strongly from the dispersed and skewed distribution found at the whole-cell level. This finding allows the gap to narrow significantly between the classical tenets of quantal analysis, as originally described at the neuromuscular junction, and apparently discrepant experimental results in the CNS. However, unlike active zones at the neuromuscular junction, single-site contacts at CNS synapses have their own individuality, which is determined by the number of postsynaptic receptors (Nusser, 1999; Walmsley et al. 1998) and by their specific biophysical properties (Auger & Marty, 1997). It is the summation of site-specific signals, combined with dendritic filtering, that gives rise to the variable amplitudes and kinetics observed in whole-cell recordings.

The occupancy of the postsynaptic receptors and the mean number of vesicles released per action potential are other parameters which shape synaptic transmission at central synapses. High occupancy of postsynaptic receptors at a single site results in sublinear summation of synaptic currents. Moreover, multivesicular release was shown to occur at some single release sites, contrary to the widely held idea that one site can only release one vesicle. Multivesicular release and receptor occupancy interact in a subtle way to shape the distributions of latencies, amplitudes, rise- and decay times of synaptic currents. Although these effects can be relatively mild, multivesicular release will also increase intersynaptic diffusion of neurotransmitter and crosstalk between synapses.

Once a solid and coherent description of the response to single action potentials is available for single site-synapses, it will be important to see how this response is modulated by applications of neuromodulators or during expression of synaptic plasticity. A first attempt to address this issue was an investigation by Kondo & Marty (1998_a_) of the modulation of cerebellar single-site synapses by noradrenaline. The next challenge will be to utilize our changing views on the elementary processes occurring at single-sites synapses to understand better the properties of multisite synaptic connections.

Acknowledgments

We thank Drs L. Forti and D. Ogden for constructive criticisms of the manuscript.

References

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