The Journal of Neurophysiology Vol. 78 No. 3 September 1997, pp. 1293-1304
Copyright ©1997 by the American Physiological Society

Are Some Minis Multiquantal?

Matthew Frerking, Salvador Borges, and Martin Wilson

Section of Neurobiology, Physiology and Behavior, Division of Biological Sciences, University of California, Davis, California 95616

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Frerking, Matthew, Salvador Borges, and Martin Wilson. Are some minis multiquantal? J. Neurophysiol. 78: 1293-1304, 1997. The amplitude distribution of miniature postsynaptic currents (minis) in many central neurons has a large variance and positive skew, but the sources of this variance and skew are unresolved. Recently it has been proposed that spontaneous Ca²⁺ influx into a presynaptic bouton with multiple release sites could cause spontaneous multiquantal minis by synchronizing release at all sites in the bouton, accounting for both the large variance and skew of the mini distribution. We tested this hypothesis by evoking minis with internally perfused, buffered Ca²⁺ and the secretagogue -latrotoxin, both in the absence of external Ca²⁺. With these manipulations, the synchronized release model predicts that the mini distribution should collapse to a Gaussian distribution with a reduced coefficient of variation. Contrary to this expectation, we find that mini amplitude distributions under these conditions retain a large variance and positive skew and are indistinguishable from amplitude distributions of depolarization-evoked minis, strongly suggesting that minis are uniquantal.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

It is widely accepted that transmitter release at synapses occurs in multimolecular, quantal packets. The main experimental basis for this hypothesis is that at the neuromuscular junction the amplitudes of postsynaptic responses to presynaptic stimulation occur in integer multiples of the amplitude distribution of spontaneously occurring "miniature" postsynaptic currents, or minis (del Castillo and Katz 1954). At many central synapses, however, this simple and compelling observation has not been reproduced. At those central synapses where the uniquantal distribution has been estimated from quantal analysis, quantal peaks are associated with very little variance (reviewed in Frerking and Wilson 1996a), implying that quanta are uniform in size. In contrast to this finding is the observation that minis are far from uniform in amplitude, so much so that one would expect quantal peaks to be unresolvable (reviewed in Stevens 1993). This expectation is reinforced by evidence suggesting that even single boutons produce highly variable minis (Bekkers et al. 1990; Liu and Tsien 1995; Vogt et al. 1995).

One possible explanation for this discrepancy is that the mini distribution is not a strictly uniquantal distribution. As recently proposed, the results from quantal analysis could be reconciled with the variance of the mini distribution if minis were not single quantal responses but rather resulted from the synchronous, probabilistic release of multiple quanta from neighboring, but separate, release sites (Edwards 1995a,b; Korn et al. 1994). According to the simplest rendering of this idea, the quanta composing minis and evoked responses ought to be identical and, as measured in evoked response amplitude histograms, almost invariant. Minis would show much greater variability than individual quanta, because the probabilistic release from multiple sites would generate minis that represented responses to a variable number of quanta. This scheme allows for the possibility that the distribution of mini amplitudes should have not only a high variance, but also a positive skew, as is actually the case for many kinds of central neurons (reviewed in Stevens 1993). It also provides an explanation for the observation that, in some preparations, the amplitude distribution of minis sometimes shows multiple peaks (Bekkers et al. 1990; Edwards et al. 1990; Frerking et al. 1995; Korn et al. 1993; Lewis and Faber 1996; Ropert et al. 1990; Silver et al. 1992; Ulrich and Lüscher 1993), although it is not clear whether these peaks are due to genuine multimodality or sampling error (Bekkers and Stevens 1994; Frerking et al. 1995).

How might the spontaneous, probabilistic, synchronous release of quanta be coordinated at multiple sites? The only endogenous signal known to cause synaptic transmitter release is Ca²⁺, and it is argued that such a synchronization could be generated by spontaneous and transient influxes of Ca²⁺ through single Ca²⁺ channels. Because presynaptic Ca²⁺ buffering in central neurons may be weak (Borst et al. 1995), these spontaneous Ca²⁺ transients could cause nearly simultaneous increases in Ca²⁺ at multiple neighboring release sites. A further measure of credibility for this model comes from the observation that single synaptic boutons sometimes contain more than one release site (Korn et al. 1994; Sorra and Harris 1993).

In this paper, we test this "synchronized release" model in cultured retinal amacrine cells, a preparation in which neurons may be well voltage clamped and in which we have already shown that mini amplitudes are highly variable (Frerking et al. 1995). Because it is a transient, local increase in Ca²⁺ concentration that is postulated to synchronize release, we would expect that coordinated release from multiple sites would not be possible if release were triggered by internally perfused, buffered Ca²⁺ in the absence of external Ca²⁺ and that minis elicited by this method should become smaller, less variable, and distributed normally if the synchronized release model is correct. The same changes should also be observed if minis are elicited by Ca²⁺-independent means. We therefore examined minis elicited by internal Ca²⁺ perfusion and the Ca²⁺-independent secretagogue latrotoxin. Contrary to the predictions of the synchronized release model, we find that the mini distributions elicited under these conditions retain a significant skew, a large variance, and a nonnormal shape; moreover, they are not resolvably different from minis elicited by depolarization in the presence of external Ca²⁺.

	METHODS

Abstract Introduction Methods Results Discussion References

Cell culture and electrophysiology

Dissociated retinas of embryonic day 8 chick embryos were plated on culture dishes at low density as described previously (Gleason et al. 1993, 1994) and used after 7-10 days in culture. Amacrine cells were identified as described previously (Gleason et al. 1993, 1994), and cells with long processes (>60 µm) were not used to avoid problems with both space clamp and equilibration of internal perfusate. Unwanted currents were blocked in all solutions with sodium and potassium channel blockers. The solutions used for depolarization-evoked release contained: internally64.5 mM CsCl, 75.5 mM cesium methanesulfonate, 5.0 mM tetraethylammonium chloride (TEACl), 1.0 mM CaCl₂, 2.0 mM MgCl₂, 11.0 mM ethylene glycol-bis(-aminoethyl ether)-N,N,N,N-tetraacetic acid (EGTA), and 10.0 mM N-2-hydroxyethylpiperazine-N-2-ethanesulfonic acid (HEPES); externally50 mM N-methyl-glucamine chloride, 72.2 mM sodium methanesulfonate, 20.0 mM TEACl, 3.0 mM CaCl₂, 0.41 mM MgCl₂, 5.6 mM glucose, 3.0 mM HEPES, and 300 nM tetrodotoxin. All solutions had a pH of 7.4 and osmolarity of 270-290 mosM. Zero-Ca²⁺-containing external solutions used for testing the Ca²⁺ dependence of depolarization-evoked minis and currents in response to bath-applied -aminobutyric acid (GABA) were the same, except that 3.0 mM MgCl₂ was substituted for 3.0 mM CaCl₂. Zero-Ca²⁺-containing external solutions used during internal perfusion and latrotoxin application had the same substitution as well as 0.5 mM bis-(o-aminophenoxy)-N,N,N,N-tetraacetic acid (BAPTA) or EGTA to remove any residual Ca²⁺. Additionally, 61 mM N-methyl-glucamine was substituted for 61 mM sodium to prevent a reduction in internal free Ca²⁺ by the Na⁺/Ca²⁺ exchanger (Gleason et al. 1994). The same internal solution used to generate depolarization-evoked minis was used during bath applications of latrotoxin. All buffered Ca²⁺-containing solutions used for internal perfusion experiments contained (in mM) 135.0 CsCl, 5.0 TEACl, 10.0 HEPES, and either 5.0 Cs₄BAPTA (1 and 50 nM free Ca²⁺-containing solutions) or 5.0 K₄Br₂BAPTA (200 nM and 2 µM free Ca²⁺-containing solutions). The respective amounts of CaCl₂ and MgCl₂ required to achieve the desired free Ca²⁺ and Mg²⁺ concentrations were then calculated with the use of MaxChelator 6.0 (Bers et al. 1994); for the solutions used, these were as follows (in mM): 0.038 CaCl₂, 1.99 MgCl₂ (1 nM free Ca²⁺); 1.37 CaCl₂, 2.35 MgCl₂ (50 nM free Ca²⁺); 0.54 CaCl₂, 2.09 MgCl₂ (200 nM free Ca²⁺); and 2.74 CaCl₂, 2.04 MgCl₂ (2 µM free Ca²⁺). The buffers used were chosen for their fast binding of Ca²⁺ (Tsien 1980).

Whole cell recording of GABA_Aergic autaptic minis was performed as described in Frerking et al. (1995). In internal perfusion experiments, the holding potential of the cell was generally kept at 70 to 60 mV, well below the voltage at which depolarization-evoked minis could be seen in the presence of external Ca²⁺. Currents from the patch-clamp amplifier (Axopatch 1A, Axon Instruments) were analog filtered (4-pole Bessel filter) at a corner frequency of 1 or 2 kHz and sampled at a rate of 2 kHz during recording to disk (Axotape, Axon Instruments). Voltage offsets generated by junction potentials were not corrected.

-Latrotoxin (1 nM; Calbiochem, La Jolla, CA) and ionomycin (5 µM; Calbiochem, La Jolla, CA) were added by bath application. Because these compounds are reported to have irreversible effects, new culture dishes were used after each application of these agents.

Data analysis

Minis were detected and analyzed as described in detail previously (Frerking et al. 1995). The threshold for mini detection was determined by the noise level, and in no case were data from a cell used if the resolution of minis appeared to be noise limited. Cells with highly atypical properties [for example, cells with a high frequency of minis at 70 mV or minis that persisted in the absence of external Ca²⁺ even in the absence of perfused, high internal Ca²⁺ concentrations, or cells with abnormally large (>1 nS) mean mini amplitude] were also excluded from this analysis. Depolarization-evoked mini frequencies were measured by counting the average number of minis present during repeated iterations of a 1.3-s voltage step from a holding potential of 70 mV. Latrotoxin- and perfusion-evoked mini frequencies were measured by breaking the entire data record into 5- or 10-s blocks and counting the number of minis in each block. Cells were included in our estimates of average maximum frequency even if they showed no minis in response to internal perfusion of Ca²⁺. In cells with very low internal Ca²⁺ concentrations (1 nM), where most cells showed no minis in response to perfusion, it was independently verified that these cells were capable of releasing quanta in response to depolarization in the presence of external Ca²⁺.

Amplitude distributions were analyzed as described previously (Frerking et al. 1995), with the use of the Kolmogorov-Smirnov test with a test statistic of  = 0.05. Amplitude distributions were constructed only from recording periods where the mini frequency was very roughly 1 Hz. Because two minis could be resolved as separate if they occurred >2 ms apart, this restriction allows us to calculate that <1% of the observed minis represent the chance occurrence of multiple minis unresolvably close together in time. This low frequency of random occurrences is critical for our analysis, and so, when we analyzed cells in which latrotoxin application or Ca²⁺ perfusion resulted initially in a mini frequency much higher than 1 Hz, minis were not analyzed until the frequency had run down to approximately this level (see Fig. 5 for example). Unless otherwise stated, all data are presented as means ± SD. Amplitude distributions were only measured during periods when the amplitude distribution was stationary and there was not an obvious slowing of the minis due to resealing of the patch, which occasionally occurred at high internal Ca²⁺ concentrations (2 µM). A minimum of 60 events, and more typically >100 events, were collected in all amplitude distributions used to compare mini amplitude distribution shapes between cells. The Mann-Whitney rank-sum test was used to compare the means of two distributions.

View larger version (27K): [in this window] [in a new window]   FIG. 5. Latrotoxin induces minis in the absence of external Ca2+. A: depolarization-evoked minis were recorded in the presence of extracellular Ca2+ in a representative cell held at 45 mV. As expected, when Ca2+ was replaced with Mg2+, these depolarization-evoked minis were blocked. However, after roughly 90 s of bath-applied -latrotoxin (1 nM) in the absence of external Ca2+, a rapid increase in frequency of minis was observed. In this cell, minis appear to decrease in frequency when latrotoxin is removed, but frequency rundown when observed in other cells is independent of whether latrotoxin was present or not, and could not be reversed even when Ca2+ was replaced in bath. For these reasons, we believe that this frequency rundown represents a latrotoxin-induced depletion of available quanta, as described elsewhere (Hurlbut et al. 1990). B: as in Fig. 3B, CVs (left) and standardized skews (right) are shown for 3 groups: cells with latrotoxin-evoked minis (open bars; n = 7), multiple iterations of 145 random samples from a Gaussian distribution (black bars; n = 20), and cells with depolarization-evoked minis (gray bars; n = 25). Results are presented as means ± SE. C: pooled standardized mini amplitude distribution from latrotoxin-evoked mini distributions of 7 cells (n = 1,070). Pooled standardized distribution of depolarization-evoked minis from Frerking et al. (1995) is also shown (n = 3,364); as with perfusion-evoked minis, the 2 distributions are not significantly different. In addition, there is no significant difference between pooled standardized amplitude distributions of perfusion-evoked minis and latrotoxin-evoked minis (not shown). LTx, latrotoxin.

Limits of statistical resolution

Our general approach has been to compare the properties of mini amplitude distributions obtained under different experimental conditions (see below) with the properties of the uniquantal distribution expected if the synchronized release model is correct. We therefore summarize here the quantitative limits of this comparison.

First, we consider the degree by which we expect our experimental manipulations to reduce synchronized release. According to the synchronized release model, depolarization-evoked minis occur with a frequency limited by the opening of single Ca²⁺ channels localized to the synaptic boutons. During the brief period of Ca²⁺ entry, the release probability at all sites within a bouton would be very high, leading to the likelihood that >1 quantum would be released more or less simultaneously. In practice, for us to detect this as simultaneous release rather than a cluster of closely spaced events, all release must occur within a 1- to 2-ms window, implying that the instantaneous release rate from a single site during a Ca²⁺ transient is somewhere between 100 and 1,000 Hz. In the experimental condition where we elicit release with buffered internal Ca²⁺ or latrotoxin, release rates from all sites would be constant but orders of magnitude lower than this range, and measurable as the frequency of minis. Because under these conditions we observe minis at rates of ~1 Hz per cell and we estimate ~10 sites per cell (Borges et al. 1995), we estimate a release rate of ~0.1 Hz per site. At this rate the probability of two sites, by chance, releasing quanta within 2 ms of each other is negligible, implying that the observed mini amplitude distribution should correspond to the uniquantal distribution.

If the synchronized release model is correct, we would expect the properties of the uniquantal distribution to be given by the numerous quantal analyses that have been performed, the general conclusion of which is that the distribution is a Gaussian distribution (skew = 0) with a coefficient of variation (CV) of <30% (reviewed in Edwards 1995a,b; Stevens 1993). Alternatively, if the synchronized release model is not correct, we would expect the CV, skew, and shape of the mini amplitude distributions to be independent of the method by which minis are evoked, and identical to CV, skew, and shape of the depolarization-evoked mini amplitude distributions measured in Frerking et al. (1995).

To test the synchronized release model, then, we must have sufficient resolving power to distinguish between the properties of a Gaussian distribution and the distribution described in Frerking et al. (1995). The CV and skew for each cell was averaged across multiple cells to estimate the variability in these parameters and the averages were compared with average values for skew and CV either from cells with depolarization-evoked minis or from 20 iterations of random sampling from a Gaussian distribution, with each iteration composed of 145 samples. The expectation of the synchronized release model is that the average CV and average skew of the mini amplitude distributions evoked by buffered internal Ca²⁺ perfusion or latrotoxin should not be resolvably different from the average CV and average skew obtained by repeated sampling from a Gaussian distribution, but should be different from the average CV and average skew of mini amplitude distributions evoked by depolarization.

These simple expectations of the synchronized release model are not supported by our data, but for completeness we have also examined the shapes of the mini amplitude distributions from each cell, with the expectation that this distribution should be significantly different from the depolarization-evoked mini distribution from Frerking et al. (1995) if the synchronized release model is correct, and different from a Gaussian distribution if it is not. When using this test, it is imperative to know what sample sizes are required to distinguish between the Gaussian uniquantal distribution inferred from quantal analysis and the mini amplitude distribution generated by depolarization. The maximum difference between the cumulative frequency distributions of these two curves is ~15%; using the Kolmogorov-Smirnov test statistic, we calculate that a minimum of ~80 minis is required to distinguish between these two distributions with 95% confidence. As will be described below, all mini distributions with a sample size larger than this value are significantly different from a Gaussian, allowing us to reject the synchronized release model as an explanation for the large skew and variance of the mini amplitude distribution.

Characteristics of internal perfusion

In internal perfusion experiments, single cells were held in the whole cell configuration, and internal solutions with buffered internal Ca²⁺ were introduced into the cell via the patch pipette while the spontaneous occurrence of minis in the absence of external Ca²⁺ was observed. After patch rupture, there was a pause of 30-60 s before recording, during which time series resistance and capacitance corrections were applied. To check that the internal solution exchanged with the cytoplasm during this period, three tests were performed. First, the cells were patched in the absence of external Ca²⁺ at a holding voltage of 60 to 70 mV so that the minis seen after patch rupture could be unambiguously identified as due to the internal perfusate. It generally took ~10-30 s after patch rupture before an increase in mini frequency could be seen, and the mini frequency always reached its maximum <1 min after patch rupture (unpublished observation). Second, we generated a compartment model of diffusion into the cell with the use of measured geometries of the longest dendrites of cells from which we recorded, similar in design and implementation to the one described in Frerking et al. (1995). This model predicts that the exchange of buffer with cytoplasm occurs over several minutes, but that within 30 s the concentration of buffer was 1 mM everywhere within the cell. Additionally, this model predicts that the free Ca²⁺ concentration will equilibrate within the cell in ~20 s, because this is the amount of time necessary for the exogenous buffer to accumulate to such a degree that the ratio of bound to unbound buffer controls the free Ca²⁺ concentration (data not shown). Because these two forms of exogenous buffer diffuse at the same rate, their ratio is expected to be roughly constant. Our final test to see that internal perfusion occurs on a relevant time scale was to compare the predictions of the compartment model with a measured increase in fluorescence seen at the tip of a dendrite when Lucifer yellow CH (1 mg/ml) was included in the patch pipette. We found that the model could reproduce the increase in fluorescence well, and the increase in fluorescence generally took ~3-5 min to stabilize (n = 5, data not shown). These three points suggest that the Ca²⁺ concentration in the cytoplasm is buffered to near its calculated concentration by the start of our recording period, although we predict, on the basis of both the time course of the increase in fluorescence during Lucifer yellow perfusion and the results from the compartment model, that the buffering capacity will increase during the course of recording from 1 to 5 mM.

Modeling Ca²⁺ microdomains

The influx of Ca²⁺ into a synaptic bouton during spontaneous intracellular Ca²⁺ release was estimated by modeling a radially symmetrical bouton as 10 compartments, each representing a concentric shell with a thickness of 50 nm centered on a spherical first compartment with a radius of 50 nm. Each compartment contained preequilibrated concentrations of free Ca²⁺, free buffer, and bound buffer, with a total buffer concentration of 2 mM for Br₂BAPTA or 50 µM for simulated native buffer, and an initial free Ca²⁺ concentration of 200 nM. On a constant influx of Ca²⁺ into the central compartment of the bouton, the concentration of each species in each compartment was calculated with the use of a separate differential equation. The three differential equations in each compartment x were as shown in Eq. 1, A-C

(1A)

(1B)

(1C)

where Ca²⁺ denotes free Ca²⁺, B denotes free buffer, and BCa denotes Ca²⁺-bound buffer. D_y represents the diffusion coefficient of each species (D_Ca²⁺ = 600 µm²/s; D_B = 171 µm²/s) (Adler et al. 1991), k_on represents the association rate constant of Ca²⁺ and buffer (k_on = 608 µM¹s¹ for BAPTA compounds) (Tsien 1980), k_off represents the dissociation rate constant of bound buffer [k_off = 60 s¹ for BAPTA and 960 s¹ for Br₂BAPTA; calculated from dissociation constants (K_ds) given in Adler et al. 1991], L is the thickness of each compartment (L = 50 nm), A_x is the surface area separating compartment x from compartment x + 1, and V_x is the volume of compartment x. The three terms in each equation account for unbinding of Ca²⁺ from buffer, binding of Ca²⁺ to buffer, and diffusion of each species between compartments, respectively. The 30 simultaneous differential equations describing the bouton were solved with the use of a fourth-order Runge-Kutta routine in Mathcad 5.0+ (MathSoft, Cambridge, MA). The influx of Ca²⁺ into the cytosol during Ca²⁺-induced Ca²⁺ release (CICR) was modeled as the spontaneous influx of Ca²⁺ into the central compartment for 2 ms through a single CICR channel with a current 10 times larger than the 0.8 pA estimated for single voltage-gated Ca²⁺ channels in the plasma membrane (Roberts 1994). This large single-channel current was chosen because the average single-channel conductance of intracellular Ca²⁺ release channels is roughly 10 times that of voltage-gated Ca²⁺ channels (Bezprozvanny and Ehrlich 1994; Hess et al. 1986; Tinker and Williams 1992), and the duration of the influx was set at 2 ms because, assuming that the opening of a single channel causes rapid release of a quantum, the release of a second quantum elsewhere in the bouton must occur in <2 ms to be unresolvable as a second, independent release event.

	RESULTS

Abstract Introduction Methods Results Discussion References

Our general approach to testing the synchronized release model is to compare the minis evoked by mechanisms not involving Ca²⁺ transients with both the Gaussian uniquantal distribution implied if the synchronized release model is correct and the depolarization-evoked mini amplitude distribution described in Frerking et al. (1995). First, we consider whether the properties of depolarization-evoked minis are consistent with the synchronized release model. Second, we compare depolarization-evoked minis with those evoked by internal perfusion of buffered Ca²⁺-containing solutions in the absence of external Ca²⁺. Finally, we compare depolarization-evoked minis with those evoked by the Ca²⁺-independent actions of -latrotoxin.

Minis evoked by depolarization require external Ca²⁺

Before we can test the synchronized release model, we must confirm that the minis evoked by depolarization depend on Ca²⁺ influx. Two experimental observations suggest that they do. The first observation is that the frequency of autaptic mini release is steeply dependent on the holding potential. The potential at which the frequency of minis detectably increased was in the range 55 to 45 mV in the six cells in which this effect was carefully quantified, as shown by a representative cell in Fig. 1A. This range corresponds roughly to the foot of the activation curve for voltage-gated Ca²⁺ channels in these cells (Gleason et al. 1994). The second observation suggesting that minis require Ca²⁺ influx is that autaptic minis evoked by depolarization were abolished by removal of external Ca²⁺ (n = 16; Fig. 1B). These results indicate that the minis we have previously described are Ca²⁺ influx dependent, indicating that the synchronized release model can be tested with the use of the amacrine cell system.

View larger version (15K): [in this window] [in a new window]   FIG. 1. Depolarization-evoked minis require external Ca2+. A: relationship between autaptic mini frequency and membrane voltage for a representative cell. Minis first start to occur at around 50 mV, and increase in frequency as membrane is depolarized. Frequency was measured as average number of minis observed in response to iterated depolarizations from a holding voltage of 70 mV for 1.3 s. B: depolarization-evoked minis are blocked by removal of extracellular Ca2+. In control (top) and wash (bottom) traces, frequent minis (inward currents) are observed. When extracellular solution was replaced with a solution containing no Ca2+, minis were abolished (middle trace). This effect is unlikely to be due to changes in postsynaptic responsiveness, because minis can be seen in the absence of Ca2+, on internal perfusion of Ca2+, and on bath application of latrotoxin (see below). Holding potential for this cell: 45 mV.

Internal perfusion of Ca²⁺ elicits minis

A total of 50 single amacrine cells was internally perfused with Ca²⁺-containing solutions. In a majority of these cells, within 10-30 s after patch rupture, a pronounced increase in the frequency of minis occurred in the absence of external Ca²⁺. Minis were qualitatively similar in both time course and mean conductance to those evoked by depolarization, with the exception that mini frequency was no longer sensitive to changes in holding voltage.

It seems likely that the minis observed were evoked by the increase in internal free Ca²⁺, but to be certain on this point we used several internal free Ca²⁺ concentrations with the expectation that higher free Ca²⁺ concentrations would elicit higher mini frequencies. Because of the short time window over which we recorded minis and the technical difficulties inherent in changing the solution inside the patch pipette, each cell was examined in the presence of only one internal free Ca²⁺ concentration, and data from different cells were averaged together to produce an estimate of the mini frequency at each free Ca²⁺ concentration. Because at high internal free Ca²⁺ concentrations rapid rundown of mini frequency was observed (data not shown), we characterized mini frequency by measuring the average maximum rate of release seen in each cell.

The summary of these experiments is shown in Fig. 2. At internal free Ca²⁺ concentrations of ~1 nM, essentially no minis were observed. As expected, the maximum mini frequency increased with the free Ca²⁺ concentration. Surprisingly, though, the mean mini frequency was detectably greater than zero with as little as 50 nM internal free Ca²⁺.

View larger version (12K): [in this window] [in a new window]   FIG. 2. Release of perfusion-evoked minis depends on free Ca2+ concentration. Relationship between estimated free Ca2+ concentration in cell and average maximum frequency of minis in multiple cells examined at that concentration is shown. Number of cells used to construct each point: 4 (1 nM), 17 (50 nM), 15 (200 nM), and 13 (2 µM). Results are presented as means ± SE. [Ca2+]i, intracellular Ca2+ concentration.

Our results to this point suggest that our proposed experimental approach is valid; internal perfusion of buffered, low concentrations of Ca²⁺ can elicit minis in the absence of external Ca²⁺ and influx-induced microdomains of high Ca²⁺. We now test the synchronized release model by examining the mini amplitude distribution evoked by internal Ca²⁺ perfusion and comparing it with the mini amplitude distribution evoked by depolarization and Ca²⁺ influx across the plasma membrane.

Minis induced by internal perfusion of Ca²⁺ are highly variable in amplitude

Like depolarization-evoked minis, those evoked with internally perfused Ca²⁺ had a broad range of amplitudes, and frequently minis were observed with peak amplitudes much larger than the modal peak amplitude. To carefully compare perfusion-evoked minis with depolarization-evoked minis, six cells with an internal free Ca²⁺ concentration of 200 nM were chosen. These cells had recording periods sufficient to collect a reasonable number of minis (>60), but had release frequencies small enough to approximate the 1 Hz in our previous analysis of depolarization-evoked minis (Frerking et al. 1995).

As can be seen in Fig. 3A, the distributions of each of these six cells have mean amplitudes that vary widely, like those of previously described depolarization-evoked minis in these cells (Frerking et al. 1995) and others (Liu and Tsien 1995). This large intercell variability in mean amplitude prevents a comparison of mean amplitudes between experimental groups with any statistical power (but see below); however, neither the CV nor the skew was as variable between cells. The CVs of mini distributions evoked by internal Ca²⁺ perfusion (averaging 57 ± 10%, mean ± SE; Fig. 3B, open bars; n = 6) were clearly larger than the CVs expected if the mini distribution of each cell under these conditions was equivalent to 145 random samples from a Gaussian distribution with a CV of 30%, as would be implied by numerous quantal analyses if the synchronized release model is correct (averaging 31 ± 2%; Fig. 3B, black bars; n = 20; see METHODS); however, they were not resolvably different from the CVs of mini distributions evoked by depolarization (averaging 58 ± 11%; Fig. 3B, gray bars; n = 25). The distributions also retained a significant skew; as with the CV, the average skew of mini distributions evoked by internal Ca²⁺ perfusion (1.8 ± 0.3; Fig. 3B, open bars; n = 6) was much greater than expected if the underlying mini distribution was Gaussian (0.02 ± 0.10; Fig. 3B, black bars; n = 20), but not resolvably different from that of mini distributions evoked by depolarization (1.9 ± 0.9; Fig. 3B, gray bars; n = 25). Finally, five of the six perfusion-evoked mini amplitude distributions were significantly different from a Gaussian distribution; the one cell that was not significantly different from a Gaussian distribution had such a small sample size (63 minis) that this lack of significance is in fact expected (see METHODS).

View larger version (28K): [in this window] [in a new window]   FIG. 3. Mini amplitude distribution of perfusion-evoked minis has large variance and skew. A: mini distributions of 6 cells during internal perfusion of 200 nM free Ca2+-containing solutions are shown. Note that for all distributions, there is a pronounced skew toward large events, and in no case was the coefficient of variation (CV) less than ~50% (not shown). B: average CVs (left) compared for 3 groups: cells with minis evoked by internal Ca2+ perfusion (open bar; n = 6), multiple iterations of 145 random events drawn from a Gaussian distribution (black bar; n = 20), and cells with minis evoked by depolarization (gray bar; n = 25). Also shown are average standardized skews for the same groups (right). Results are shown as means ± SE. C1: same 6 distributions as shown in A, standardized with respect to their individual means and SDs. None of the 6 standardized distributions shown is significantly different from their average, or from any of the others. C2: the 6 distributions from C1 pooled together to produce estimated shape of mini distribution of perfusion-evoked minis (n = 870). Pooled standardized distribution of depolarization-evoked minis from Frerking et al. 1995 is also shown (n = 3,364); the 2 distributions are not significantly different despite large sample sizes in both cases, with maximum difference in cumulative relative frequency of <2%. Dotted line: shape of Gaussian distribution for comparison.

This comparison of skew and CV between experimental groups and simulated expectations rules out the possibility that the mini distributions evoked by internal Ca²⁺ perfusion are sampled from the uniquantal distribution implied by quantal analysis, and speaks strongly against the synchronized release model, suggesting that the mini distribution is not sensitive to the method by which minis are elicited; however, this latter point would be reinforced if the shapes of the mini distributions elicited under different conditions could be compared with high resolution. To compare the shapes of perfusion- and depolarization-evoked minis, we took advantage of the observation that depolarization-evoked mini distributions from different cells are not resolvably different when standardized to a common mean and variance (Frerking et al. 1995).

This standardization procedure is useful for two reasons. First, it allows a model-independent comparison of the shapes of the mini distributions between cells. Second, it provides a method by which minis from multiple cells can be pooled together to produce large sample sizes and high resolution of the shape of the mini distribution. We therefore tested to see whether, like amplitude distributions of depolarization-evoked minis, perfusion-evoked mini amplitude distributions from different cells could be standardized to a common shape. As shown in Fig. 3C1, none of the mini distributions shown in Fig. 3A were significantly different from the average of all of them, or from any of the other five, after standardization. This observation allows us to pool together the standardized minis from all six cells to produce a pooled standardized distribution with a much larger sample size (n = 870) and therefore better resolution of the shape underlying the mini distribution. When this pooled standardized distribution of perfusion-evoked minis is compared with that of depolarization-evoked minis described previously (data from Fig. 4C; n = 3,364) (Frerking et al. 1995), the two distributions are not significantly different (Fig. 3C2, ; P > 0.2). For comparison, the shape of a Gaussian distribution is also shown (Fig. 3C2, dotted line).

View larger version (15K): [in this window] [in a new window]   FIG. 4. Microdomains of Ca2+ are spatially restricted by exogenous buffer. Compartment model of bouton after 2 ms of Ca2+ flux through a single open Ca2+-induced Ca2+ release (CICR) channel in the center of the bouton was used to estimate increase in cytosolic Ca2+ caused by a single "blip" of CICR. In the presence of simulated "native" buffer [50 µM bis-(o-aminophenoxy)-N,N,N,N-tetraacetic acid (BAPTA) equivalent], Ca2+ concentration exceeds 100 µM everywhere within the bouton (). In the presence of 2 mM partially loaded Br2BAPTA, free Ca2+ concentration exceeds 2 µM only in a region <150 nm from the site of Ca2+ flux (). Because the Kd of all known synaptic release processes is much higher than 2 µM, release outside this 150-nm radius area is highly improbable even with a large flux through a single CICR channel (roughly 10 times that of voltage-gated Ca2+ channel).

We have emphasized the results obtained with 200 nM internal free Ca²⁺, but the results obtained with this Ca²⁺ concentration were reproduced at other internal free Ca²⁺ concentrations, and neither 50 nM internal free Ca²⁺ nor 2 µM internal free Ca²⁺ produced pooled standardized mini distributions that were significantly different from that at 200 nM (not shown). These results strongly suggest that synchronized release does not generate the shape or variance of the mini distribution.

Could Ca²⁺ release from internal stores synchronize release?

The experiments in which internally perfused buffered Ca²⁺ was used rule out the possibility that transient Ca²⁺ microdomains inside the cell could be generated by influx across the plasma membrane. It is nevertheless possible that Ca²⁺ release from internal stores could create Ca²⁺ microdomains in the absence of external Ca²⁺. An alternative model, then, which would be consistent with our data, might propose that our internal perfusion of Ca²⁺ does not provide a sufficient concentration of Ca²⁺ to allow quantal release, but minis occur during our internal perfusion of Ca²⁺ because the perfused Ca²⁺ is sufficient to cause occasional single channel "blips" of CICR from internal stores, similar to those observed elsewhere (Cheng et al. 1993; Parker and Yao 1996). Because these Ca²⁺ blips would generate microdomains of high Ca²⁺ in the absence of extracellular Ca²⁺, the minimum concentration of Ca²⁺ required to elicit minis could be several orders of magnitude higher than our results suggest, and the mini distribution might still contain Ca²⁺ microdomain-dependent multiquantal minis that increase the skew and variance of the mini distribution.

This alternative model is difficult to address experimentally. No specific inhibitors of release from all types of internal stores exist, nor is there agreement about the number of internal Ca²⁺ stores. Additionally, we have no assay by which we could reliably confirm that an inhibitor of CICR blocks CICR during our experiments. However, the buffer present in the cell during our experiments should restrict the range of any Ca²⁺ microdomain, regardless of its source. If the range of the microdomain is sufficiently restricted, neighboring release sites would be effectively isolated from each other even if individual CICR sites were tightly colocalized with release sites.

The degree to which the buffer restricts any possible CICR-induced microdomains can be estimated by modeling the diffusion of Ca²⁺ into cytoplasm loaded with buffer. To maximally restrict the diffusion of Ca²⁺ away from release sites and simplify modeling, we assumed that release sites are located in diffusionally isolated spherical synaptic boutons with a diameter of 1 µm, smaller than GABAergic boutons from other cells and comparable in size with the dendrite diameter in these cells (Frerking et al. 1995; Korn et al. 1994), and that CICR occurs within the center of these boutons. These conditions maximize homogeneity in Ca²⁺ concentration, thereby maximizing the ability of the release sites within the bouton to act as a single cooperative unit.

The spatial range of the increase in free Ca²⁺ at the end of the simulated CICR blip, a 2-ms Ca²⁺ current of 8 pA, was modeled under two separate conditions: in the presence of our conservatively estimated native buffer (equivalent to 50 µM BAPTA) and in the presence of 2 mM partially loaded Br₂BAPTA, which roughly corresponds to the minimum expected concentration of exogenous buffer present in our experimental conditions. Figure 4 shows the expected free Ca²⁺ concentrations under the two conditions after 2 ms of Ca²⁺ influx. In the presence of only native buffer, the expected free Ca²⁺ concentration everywhere within the bouton rose to a concentration >100 µM (Fig. 4, ). In the presence of exogenous buffer (Fig. 4, ), the expected free Ca²⁺ concentration was only >2 µM within a <150-nm radius of the CICR site. This result indicates that Ca²⁺ microdomains would be highly restricted by the presence of our exogenous buffers and implies that, to be compatible with synchronized release, multiple release sites would have to be tightly clustered around a single CICR site, all contained within a small fraction (5-10%) of the total volume of the bouton. Such a close colocalization of all of the required components seems implausible, but we cannot rigorously rule this possibility out. We have therefore also examined minis elicited by Ca²⁺-independent means.

Ca²⁺-independent minis are highly variable in amplitude

The active component of black widow spider venom, -latrotoxin, is known to cause both Ca²⁺-dependent and Ca²⁺-independent release of neurotransmitter in a wide range of central and peripheral neurons, as well as synaptosomes (reviewed in Geppert et al. 1994; Petrenko 1993). The Ca²⁺-dependent actions of latrotoxin are probably generated by the actions of this secretagogue as a cation ionophore (Filippov et al. 1994; Wanke et al. 1986), but latrotoxin can cause transmitter release even in the presence of monoclonal antibodies that block this effect (Pashkov et al. 1993) and can cause transmitter release in the absence of extracellular Ca²⁺ (reviewed in Petrenko 1993; Storchak et al. 1994) by an unknown mechanism. We therefore examined the Ca²⁺-independent effects of this toxin on minis.

To remove the Ca²⁺-dependent effects of latrotoxin, we examined its effects in the absence of external Ca²⁺. As shown by a representative cell in Fig. 5A, bath application of 1 nM latrotoxin in the absence of external Ca²⁺ caused, after roughly 90 s, a large increase in the frequency of minis in 10 of 15 cells. These latrotoxin-evoked minis eventually became infrequent after a few minutes, as expected if latrotoxin depleted the available supply of vesicles (Hurlbut et al. 1990). The ability of latrotoxin to evoke Ca²⁺-independent minis provides us with a powerful method for testing the synchronized release model.

Like depolarization- and perfusion-evoked mini distributions, the CVs and skews of latrotoxin-evoked mini amplitude distributions (averaging 54 ± 7% and 1.7 ± 0.6, respectively; Fig. 5B, open bars; n = 7) were not resolvably different from depolarization-evoked mini distribution CVs and skews (averaging 58 ± 11% and 1.9 ± 0.9, respectively; Fig. 5B, gray bars), but were much larger than the CVs and skews expected if the distributions were Gaussian with a CV of 30% (averaging 31 ± 2% and 0.02 ± 0.10, respectively; Fig. 5B, black bars). Also like depolarization- and perfusion-evoked minis, the differences in mini amplitude distributions between cells were removed by standardizing the mini distributions with respect to their individual means and SDs (data not shown). None of the seven standardized mini distributions examined was significantly different from the average of all of them, and when the mini distributions of individual cells were compared, 19 of the 21 possible paired comparisons showed no significant differences. As with perfusion-evoked minis, the pooled standardized distribution of latrotoxin-evoked minis was not significantly different from that previously generated from depolarization-evoked minis (Fig. 5C). Finally, all of the individual cells examined, as well as the pooled results from all seven, were significantly different from a Gaussian distribution. We conclude that the shape and variance of the mini distribution are independent of transient Ca²⁺ fluxes into the cytoplasm.

Although latrotoxin-evoked minis occur in the absence of external Ca²⁺, it might be supposed that latrotoxin could somehow get across the plasma membrane into the cytoplasm and, acting as a cation ionophore, evoke minis by the release of calcium from internal stores. This possibility could invalidate our claim that latrotoxin-evoked minis are Ca²⁺ independent. We think it unlikely that latrotoxin could translocate completely across the plasma membrane, because this protein is large (130 kD) and hydrophilic; however, to test this possibility, we reason that if the ability of latrotoxin to evoke transmitter release in the absence of external Ca²⁺ depends on the action of latrotoxin as a Ca²⁺ ionophore, then we would expect other Ca²⁺ ionophores that are thought to cause internal Ca²⁺ release (Dolmetsch and Lewis 1994) to similarly cause transmitter release in the absence of external Ca²⁺. We therefore examined the ability of ionomycin to induce release in the absence of external Ca²⁺, on the grounds that this compound is smaller and more hydrophobic than latrotoxin and therefore at least as likely to cross the plasma membrane and permeabilize internal membranes. Bath application of ionomycin (5 µM) was not competent to cause transmitter release in the absence of external Ca²⁺ over a period of up to 3-4 min, although ionomycin could generate voltage-independent minis on addition of external Ca²⁺, as expected (n = 5; data not shown). This result suggests that the actions of latrotoxin in the absence of external Ca²⁺ cannot be accounted for by release from internal Ca²⁺ stores.

Mini amplitude depends on the external divalent composition

Our evidence to this point suggests that the skew and variance of the mini distribution are not generated by synchronized release. However, one unsatisfactory feature of our data analysis is that the standardization procedure that allows us to compare mini amplitudes between different cells removes information about the mean and variance of the mini distribution. The CV of the mini distribution is not widely variable between cells, allowing this parameter to be compared between experimental groups independent of our analysis of the standardized mini distribution shape, but the mean mini amplitude is too variable for a comparison between the small groups of cells examined here to have any resolving power. To remedy this shortcoming, we have compared the mean mini amplitudes of minis evoked by two different methods in the same cell.

The difficulties in altering the internal perfusate precluded a direct comparison of mean mini amplitudes evoked by internal perfusion with those evoked by other methods, but in four cells (see Fig. 5A) minis were evoked by depolarization, and then this stimulus was blocked by replacing external Ca²⁺ with Mg²⁺ and minis were evoked by bath application of latrotoxin. Surprisingly, the mean mini amplitude of latrotoxin-evoked minis was significantly smaller than that of depolarization-evoked minis (Fig. 6A; rank-sum test) in three of the four cells examined, even though the standardized mini distributions were not resolvably different between stimuli for any of the four cells examined (Fig. 6B; Kolmogorov-Smirnov test). On average, mini amplitudes evoked by latrotoxin were 78 ± 13% of the mini amplitudes evoked by depolarization (mean ± SE; Fig. 6E).

View larger version (26K): [in this window] [in a new window]   FIG. 6. Mean mini amplitude is affected by divalent composition of extracellular solution. A: amplitude distributions for minis evoked under 2 different conditions in same cell: first, minis were evoked by depolarization in presence of extracellular Ca2+, and subsequently, minis were evoked by latrotoxin when this Ca2+ was replaced with Mg2+. Latrotoxin-evoked minis are significantly smaller than depolarization-evoked minis. B: although minis evoked under the 2 experimental conditions in A have different mean amplitudes, standardized mini amplitude distributions of same minis are not resolvably different. C: bath application of 3 µM -aminobutyric acid (GABA) evokes currents sensitive to extracellular divalent composition. GABA evokes whole cell inward currents in a representative cell held at 60 mV (). On replacement of extracellular Ca2+ with Mg2+, currents evoked by a 2nd application of GABA are reduced (· · ·). Reduction is reversible on readdition of Ca2+ (- - -). Currents have been aligned at their onsets (). Scale bar: 20 pA, 1 s. D1: replacement of extracellular Ca2+ with Mg2+ reversibly reduces amplitude of whole cell responses to continuously applied 3 µM GABA. Scale bar: 40 pA, 5 s. D2: replacement of extracellular Ca2+ with Mg2+ does not alter baseline currents in the absence of GABA. Same scale as in D1. E: reduction in mean mini amplitude seen when minis evoked by latrotoxin in the absence of external Ca2+ are compared with those evoked by depolarization in the presence of external Ca2+ is not significantly different from reduction in mean mini amplitude seen between minis evoked by same method (internal perfusion) in the absence and presence of extracellular Ca2+, or from reduction in whole cell GABA peak and steady-state responses caused by removal of extracellular Ca2+. This suggests that reduction in mini amplitude seen in A is due to a postsynaptic effect of altered divalent composition, rather than to a difference in mechanism underlying minis evoked by latrotoxin and depolarization.

This discrepancy in mean amplitudes might seem to favor the synchronized release model, but in fact its origin lies in an effect of divalent cations on postsynaptic currents as shown in experiments comparing whole cell currents evoked by bath application of GABA in the presence and absence of external Ca²⁺. Bath application of 3 µM GABA generated inward currents that were completely blocked by bicucculine (unpublished observation). Peak responses to repeat applications of GABA were compared in the presence and absence of extracellular Ca²⁺, and the peak response was consistently found to be smaller in the absence of extracellular Ca²⁺ (Fig. 6C; n = 6). The peak GABA current in the absence of extracellular Ca²⁺ was, on average, 72 ± 4% (mean ± SE) of the peak GABA current in the presence of extracellular Ca²⁺ (Fig. 6E), and this reduction in GABA current on removal of extracellular Ca²⁺ is not significantly different from the reduction in mini amplitude seen above. Moreover, we found that in the continued presence of 3 µM GABA, removal of extracellular Ca²⁺ also reversibly reduced the GABA currents (Fig. 6D1; n = 6). As expected, removal of extracellular Ca²⁺ had no appreciable effect on baseline currents in the absence of GABA (Fig. 6D2; n = 6). The steady-state GABA current in the absence of extracellular Ca²⁺ was 73 ± 3% (mean ± SE) of the steady-state GABA current in the presence of extracellular Ca²⁺, and this reduction was also not significantly different from that of the peak GABA currents or from that of minis evoked by latrotoxin in the absence of extracellular Ca²⁺ when compared with depolarization-evoked minis in the presence of extracellular Ca²⁺ (Fig. 6E).

These experiments demonstrate that GABA-evoked currents are sensitive to the divalent composition of the extracellular solution, and moreover that the magnitude of this effect is large enough to account entirely for the discrepancy in mini amplitudes described above, but to be certain that mini amplitude can be altered postsynaptically by the divalent composition we examined perfusion-evoked minis in the presence and absence of extracellular Ca²⁺ when cells were held at 70 mV to prevent Ca²⁺ influx-evoked minis. In all four cells examined, the mean amplitude of perfusion-evoked minis in the absence of external Ca²⁺ was smaller than the mean mini amplitude of perfusion-evoked minis from the same cell in the presence of external Ca²⁺ (rank-sum test; not shown). Mean perfusion-evoked mini amplitude was reduced to 69 ± 5% (mean ± SE) of control values when external Ca²⁺ was replaced with Mg²⁺ (Fig. 6E), a value not significantly different from any other effects described above. We conclude that removal of extracellular Ca²⁺ (or alternatively an increase in extracellular Mg²⁺) causes a reduction in postsynaptic responses to GABA by an unknown mechanism. The magnitude of this effect is similar to that seen when comparing latrotoxin-evoked minis in the absence of extracellular Ca²⁺ with depolarization-evoked minis in the presence of extracellular Ca²⁺, suggesting that when these postsynaptic effects of altering the divalent concentration are taken into account, the latrotoxin-evoked minis are likely to have the same mean amplitude as depolarization-evoked minis in the same cells.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

Low concentrations of internal free Ca²⁺ can evoke transmitter release

Our experimental observations suggest that a detectable level of transmitter release can be induced at intracellular free Ca²⁺ concentrations in the range of 10-100 nM. The concentrations of free Ca²⁺ that detectably evoked minis in our experiments are similar in magnitude to the estimated resting concentration of free Ca²⁺ in the cytoplasm (Borges et al. 1996). It is frequently argued that spontaneous minis are Ca²⁺ independent because the removal of external Ca²⁺ or the addition of voltage-gated Ca²⁺ channel blockers does not prevent spontaneous release (reviewed in Edwards 1995b); however, this correspondence of measured resting Ca²⁺ levels and the level of intracellular Ca²⁺ required to evoke release suggests that spontaneous minis may be Ca²⁺ dependent even if they are not Ca²⁺ influx dependent.

Latrotoxin-evoked minis are indistinguishable from depolarization-evoked minis

We have found that, as in other central neurons in culture, latrotoxin can evoke transmitter release even in the absence of extracellular Ca²⁺ (Capogna et al. 1996a,b; Geppert et al. 1994; Grasso and Mercanti-Ciotti 1993). The minis evoked by latrotoxin are highly variable, and our results suggest that these minis are indistinguishable from minis evoked by depolarization. At the neuromuscular junction there is a roughly one-to-one correspondence between the number of vesicles released by latrotoxin and the number of minis evoked by latrotoxin (Hurlbut et al. 1990), and if central neurons have an analogous site of action for latrotoxin, then it seems plausible that a mini in our cells is generated by the release of a single vesicle, although this rough correspondence does not have sufficient resolution to state unambiguously that all minis are due to release of single vesicles. We also note that depolarization-evoked release, spontaneous minis, and minis evoked by latrotoxin can all be blocked by botulinum toxin (Capogna et al. 1996b), suggesting that all of these types of release have a similar biochemical cascade in the final steps preceding vesiclefusion.

We used -latrotoxin as a tool to examine the synchronized release model, but our experiments using this toxin have additional implications for the nature of minis. It has recently been proposed that minis might be the result of a lower concentration of transmitter than stimulus-evoked events, if stimulus-evoked events were due to full vesicle fusion but minis were due to "kiss-and-run" exocytosis, in which incomplete vesicle fusion causes only partial release of vesicular contents (Clements 1996). If the concentration of transmitter during stimulation was saturating but the concentration during spontaneous release was not, this might provide a novel explanation for the discrepancy between the variance of minis and the results of quantal analysis (Clements 1996). However, this explanation requires that minis generated by full vesicle fusion must have less variance than those generated under more "conventional" conditions. Latrotoxin is known to induce full vesicle fusion, because latrotoxin depletes the presynaptic terminal of morphologically identifiable synaptic vesicles (Hurlbut et al. 1990; Matteoli et al. 1988). A specific prediction of this model, then, is that minis evoked by latrotoxin should be less variable than minis evoked by other methods. Contrary to this prediction, both the CV and shape of the amplitude distribution of minis evoked by latrotoxin were indistinguishable from those of minis evoked by other methods. We conclude that incomplete vesicle fusion is not a significant source of variance in mini amplitude.

Ca²⁺ influx is not required to produce the skew or variance of the mini distribution

The ability of our internal solutions to elicit minis in the absence of Ca²⁺ influx has allowed us to determine whether Ca²⁺ influx during depolarization-evoked release synchronizes adjacent release sites to generate a large number of multiquantal events. If this were the case, we would expect to see the mini distribution evoked by internal Ca²⁺ perfusion or latrotoxin in the absence of external Ca²⁺ collapse into a Gaussian distribution with a CV equal to that implied by quantal analysis (<30%). Contrary to this expectation, mini distributions under these conditions are significantly skewed, have a CV averaging >50%, and are indistinguishable in shape, skew, and CV from depolarization-evoked mini distributions, making it unlikely that a Ca²⁺-influx dependent release of multiple quanta at different sites could account for the skew and variance of the mini distribution.

An alternative explanation for the skew of the mini distribution in the absence of external Ca²⁺ could be that the microdomains necessary to synchronize release are generated under these circumstances by Ca²⁺ release from internal stores instead of by Ca²⁺ influx across the plasma membrane. Four considerations make it unlikely that such microdomains could explain our results. The first is that anatomic studies have indicated that small synaptic boutons lack endoplasmic reticulum, making it unlikely that there is Ca²⁺ release from internal stores in the first place (reviewed in Edwards 1995b).

A second argument against this idea stems from our observation that ionomycin does not cause an increase in the frequency of minis if extracellular Ca²⁺ is removed. Because it has been reported that this ionophore can release Ca²⁺ from intracellular stores (Dolmetsch and Lewis 1994), the lack of effect of ionomycin on mini frequency suggests that either Ca²⁺ stores are not colocalized with release sites well enough for the released Ca²⁺ to support vesicle fusion, or alternatively that the Ca²⁺ stores drain rapidly in the absence of external Ca²⁺ to replenish them and are therefore empty under these conditions. In support of this second possibility, internal Ca²⁺ stores in these cells seem to have a small capacity (Borges et al. 1996).

A third argument against the involvement of internal Ca²⁺ stores is that in our experiments with internally perfused Ca²⁺, the concentration of buffer present in our internal solution restricts the spatial profile of the microdomain to such a degree that a coordination of transmitter release from multiple sites could only be achieved if the single Ca²⁺ release site and multiple active zones were all within ~300 nm of each other with no diffusional barriers. For reference, this distance is roughly equal to the diameter of a single postsynaptic density in CA1 cells of the hippocampus (Harris and Stevens 1989) or six synaptic vesicles lying side by side, and such a close apposition of multiple transmitter release sites with a Ca²⁺ release site is highly improbable, even in boutons with multiple release sites (see Sorra and Harris 1993).

The final argument against an involvement of internal Ca²⁺ stores in generating variance in mini amplitude is that the minis evoked by latrotoxin appear to be Ca²⁺ independent, yet latrotoxin-evoked mini amplitude distributions have a large CV and a prominent skew. We note that, if our assessment of latrotoxin-evoked minis as Ca²⁺ independent is correct, our results rule out not only multiquantal minis coordinated by Ca²⁺ at adjacent release sites, but also the recent proposal that the variance and skew in the mini distribution is due to a Ca²⁺-dependent coordination of multivesicular release at single release sites (Bennett et al. 1995).

Our results suggest strongly that Ca²⁺ influx does not coordinate the release of quanta from multiple release sites, but our experiments do not rule out a Ca²⁺-independent coordination of release. However, three considerations limit enthusiasm for this possibility. The first is that the only endogenous signal known to directly promote quantal transmitter release is Ca²⁺. Second, the time window over which this putative signal would have to be synthesized, diffuse, and exert its actions is maximally 2 ms, making a biochemical event unlikely. Third, production of this putative signal would have to be stimulated similarly by all three methods we have used to evoke minis, even though these methods are mechanistically unrelated. Because these considerations make it unlikely that multiquantal minis are generated by a Ca²⁺-independent mechanism, and our data suggest that multiquantal minis are not generated by Ca²⁺ influx, we suggest that the mini amplitude distribution is uniquantal.

Our results indicate that minis are uniquantal, but because a quantum is defined as the output of a single release site and not as the output of a single vesicle at a single release site, we cannot say with certainty that minis are responses to the release of single vesicles of transmitter. There is currently no direct measure of the number of vesicles that make up a quantum, and it is beyond the scope of this paper to determine whether or not minis are univesicular; however, we note that even if multivesicular release at a single site occurs, it could not contribute significantly to the variance in mini amplitude unless postsynaptic receptors were not saturated during quantal transmitter release and the output of single release sites was highly variable. Our conclusion that uniquantal events are highly variable is therefore independent of the number of vesicles that comprise a quantum.

One general concern in evaluating our results is that our experiments have been performed on isolated neurons in culture, whereas much of the quantal analysis which underlies the synchronized release model was performed in brain slices. However, this difference in preparations limits the relevance of our conclusions only if the sources of variance in mini amplitude differ significantly between slices and culture. It is therefore of interest to note that many neurons appear to have mini distributions that are at least qualitatively similar in shape and CV to the minis we have seen in cultured amacrine cells (reviewed in Frerking and Wilson 1996b). This similarity appears to hold between cells of the same type in slices and in culture (Bekkers and Stevens 1995; Bekkers et al. 1990; Liu and Tsien 1995; Malgaroli and Tsien 1992; Manabe et al. 1992; Tong and Jahr 1994) and between synapses from many different brain regions and different transmitter types (Bekkers et al. 1990; De Koninck and Mody 1994; Edwards et al. 1990; Jonas et al. 1993; Silver et al. 1992), although it must be noted that not all central synapses show a skewed mini distribution (reviewed in Korn et al. 1994). This suggests that the sources of variance in mini amplitude may be conserved not only between slices and culture, but also among a large set of central neurons, lending physiological relevance to our conclusions.

Another concern when evaluating our data is that a comparison of minis from different cells was used in the majority of the experiments described here. It is plausible that, because of variability introduced by this cross-cell comparison, the CV of mini amplitudes for perfusion- or latrotoxin-evoked minis could be slightly smaller than that for depolarization-evoked minis, but this difference would be below our limit of statistical resolution. This would indicate that, in fact, some small fraction of minis might be multiquantal. By examining the properties of perfusion- and latrotoxin-evoked minis independent of depolarization-evoked minis, however, we can make several statements about the uniquantal distribution that are independent of this cross-cell comparison. Minimally, we have shown that the CV of the uniquantal distribution exceeds 40% and that, moreover, the uniquantal distribution is highly skewed and has a shape virtually identical to the mini distribution. Given a uniquantal distribution with these properties, it seems unlikely that quantal peaks could be resolved (Frerking and Wilson 1996b), and we therefore suggest that quantal analysis from cells with a mini distribution similar to that observed here should be interpreted with caution.

Our results differ from those observed in goldfish Mauthner cell lateral dendrites, where the skewed mini distribution seen in the presence of external Ca²⁺ collapses into a Gaussian distribution when this Ca²⁺ is reduced (Korn et al. 1993). This discrepancy may indicate that synchronized release accounts for some skewed mini distributions but not others; however, isolated results from other systems are at least suggestive that our results may be more representative of the "typical" central neuron. First, putative uniquantal events desynchronized by bath application of Sr²⁺ are not significantly different from minis in hippocampal CA1 pyramidal cells (Oliet et al. 1996). Second, adenosine and baclofen, which both inhibit release presynaptically, decrease non-N-methyl-D-aspartate-receptor-mediated mini frequency but do not affect mean mini amplitude in either hippocampal CA3 cells (Scanziani et al. 1992) or cerebellar Purkinje cells (Dittman and Regehr 1996). Third, latrotoxin-evoked minis have the same mean amplitude as spontaneous minis in CA3 cells (Capogna et al. 1996b). Finally, minis from zebrafish Mauthner cells retain a skewed amplitude distribution even in reduced Ca²⁺ (Legendre and Korn 1994). Each of these results is difficult to reconcile with the synchronized release model, but all are in agreement with our observations.

	ACKNOWLEDGEMENTS

  This work was supported by National Eye Institute Grant EY-04112 to M. Wilson.

	FOOTNOTES

   Present address of M. Frerking: Dept. of Pharmacology, University of California, San Francisco, CA 94143.

  Address reprint requests to M. Wilson.

  Received 6 February 1997; accepted in final form 7 May 1997.

	REFERENCES

Abstract Introduction Methods Results Discussion References

• ADLER, E., AUGUSTINE, G., DUFFY, S., CHARLTON, M. Alien intracellular calcium chelators attenuate neurotransmitter release at the squid giant synapse. J. Neurosci. 11: 1496-1507, 1991.[Abstract]
• BEKKERS, J. M., RICHERSON, G. B., STEVENS, C. F. Origin of variability in quantal size in cultured hippocampal neurons and hippocampal slices. Proc. Natl. Acad. Sci. USA 87: 5359-5362, 1990.[Abstract/Free Full Text]
• BEKKERS, J. M., STEVENS, C. F. The nature of quantal transmission at central excitatory synapses. In: Molecular and Cellular Mechanisms of Neurotransmitter Release, edited by L. Stjärne, P. Greengard, S. Grillner, T. Hökfelt, and D. Ottoson . New York: Raven, 1994, p. 261-273
• BEKKERS, J. M., STEVENS, C. F. Quantal analysis of EPSCs recorded from small numbers of synapses in hippocampal cultures. J. Neurophysiol. 73: 1145-1156, 1995.[Abstract/Free Full Text]
• BENNETT, M. R., GIBSON, W. G., ROBINSON, J. Probabilistic secretion of quanta: spontaneous release at active zones of varicosities, boutons, and endplates. Biophys. J. 69: 42-56, 1995.[Medline]
• BERS, D., PATTON, C., AND NUCCITELLI, R. A practical guide to the preparation of Ca²⁺ buffers. In: Methods in Cell Biology: A Practical Guide to the Study of Calcium in Living Cells, edited by R. Nuccitelli. San Diego, CA: Academic, 1994, p. 3-29.
• BEZPROZVANNY, I., EHRLICH, B. Inositol (1,4,5)-trisphosphate (InsP₃)-gated Ca channels from cerebellum: conduction properties for divalent cations and regulation by intraluminal calcium. J. Gen. Physiol. 104: 821-856, 1994.[Abstract/Free Full Text]
• BORGES, S., GLEASON, E., FRERKING, M., WILSON, M. Neurotensin induces calcium oscillations in cultured amacrine cells. Visual Neurosci. 13: 311-318, 1996.[Medline]
• BORGES, S., GLEASON, E., TURELLI, M., WILSON, M. The kinetics of quantal transmitter release from retinal amacrine cells. Proc. Natl. Acad. Sci. USA 92: 6896-6900, 1995.[Abstract/Free Full Text]
• BORST, J. G., HELMCHEN, F., SAKMANN, B. Pre- and postsynaptic whole-cell recordings in the medial nucleus of the trapezoid body of the rat. J. Physiol. (Lond.) 489: 825-840, 1995.[Abstract/Free Full Text]
• CAPOGNA, M., GAHWILER, B., THOMPSON, S. Presynaptic inhibition of calcium-dependent and -independent release elicited with ionomycin, gadolinium, and -latrotoxin in the hippocampus. J. Neurophysiol. 75: 2017-2028, 1996a.[Abstract/Free Full Text]
• CAPOGNA, M., GAHWILER, B., THOMPSON, S. Calcium-independent actions of -latrotoxin on spontaneous and evoked synaptic transmission in the hippocampus. J. Neurophysiol. 76: 3149-3158, 1996b.[Abstract/Free Full Text]
• CHENG, H., LEDERER, W., CANNELL, M. Calcium sparks: elementary events underlying excitation-contraction coupling in heart muscle. Science 262: 1419-1423, 1993.[Abstract/Free Full Text]
• CLEMENTS, J. D. Transmitter timecourse in the synaptic cleftits role in central synaptic function. Trends Neurosci. 19: 163-171, 1996.[Medline]
• DE KONINCK, Y., MODY, I. Noise analysis of miniature IPSCs in adult rat brain slices: properties and modulation of synaptic GABA_A receptor channels. J. Neurophysiol. 71: 1318-1335, 1994.[Abstract/Free Full Text]
• DEL CASTILLO, J., KATZ, B. Quantal components of the end-plate potential. J. Physiol. (Lond.) 124: 560-573, 1954.
• DITTMAN, J., REGEHR, W. Contributions of calcium-dependent and calcium-independent mechanisms to presynaptic inhibition at a cerebellar synapse. J. Neurosci. 16: 1623-1633, 1996.[Abstract/Free Full Text]
• DOLMETSCH, R., LEWIS, R. S. Signaling between intracellular Ca²⁺ stores and depletion-activated Ca²⁺ channels generates [Ca²⁺]_i oscillations in T lymphocytes. J. Gen. Physiol. 103: 365-388, 1994.[Abstract/Free Full Text]
• EDWARDS, F. A. LTPa structural model to explain the inconsistencies. Trends Neurosci. 18: 250-255, 1995a.[Medline]
• EDWARDS, F. A. Anatomy and electrophysiology of fast central synapses lead to a structural model for long-term potentiation. Physiol. Rev. 75: 759-787, 1995b.[Abstract/Free Full Text]
• EDWARDS, F. A., KONNERTH, A., SAKMANN, B. Quantal analysis of inhibitory synaptic transmission in the dentate gyrus of rat hippocampal slices: a patch-clamp study. J. Physiol. (Lond.) 430: 213-249, 1990.[Abstract/Free Full Text]
• FILIPPOV, A. K., TERTISHNIKOVA, S. M., ALEKSEEV, A. E., TSURUPA, G. P., PASHKOV, V. N., GRISHIN, E. V. Mechanism of -latrotoxin action as revealed by patch-clamp experiments on Xenopus oocytes injected with rat brain messenger RNA. Neuroscience 61: 179-189, 1994.[Medline]
• FRERKING, M., BORGES, S., WILSON, M. Variation in GABA mini amplitude is the consequence of variation in transmitter concentration. Neuron 15: 885-895, 1995.[Medline]
• FRERKING, M., WILSON, M. Saturation of postsynaptic receptors at central synapses? Curr. Opin. Neurobiol. 6: 395-403, 1996a.[Medline]
• FRERKING, M., WILSON, M. Effects of variance in mini amplitude on stimulus-evoked release: a comparison of two models. Biophys. J. 70: 2078-2091, 1996b.[Medline]
• GEPPERT, M., GODA, Y., HAMMER, R. E., LI, C., ROSAHL, T., STEVENS, C. F., SÜDHOF, T. C. Synaptotagmin I: a major Ca²⁺ sensor for transmitter release at a central synapse. Cell 79: 717-727, 1994.[Medline]
• GLEASON, E., BORGES, S., WILSON, M. Synaptic transmission between pairs of retinal amacrine cells in culture. J. Neurosci. 13: 2359-2370, 1993.[Abstract]
• GLEASON, E., BORGES, S., WILSON, M. Control of transmitter release from retinal amacrine cells by Ca²⁺ influx and efflux. Neuron 13: 1109-1117, 1994.[Medline]
• GRASSO, A., MERCANTI-CIOTTI, M. T. The secretion of amino acid transmitters from cerebellar primary cultures probed by -latrotoxin. Neuroscience 54: 595-604, 1993.[Medline]
• HARRIS, K. M., STEVENS, J. K. Dendritic spines of CA1 pyramidal cells in the rat hippocampus: serial electron microscopy with reference to their biophysical characteristics. J. Neurosci. 9: 2982-2987, 1989.[Abstract]
• HESS, P., LANSMAN, J., TSIEN, R. W. Calcium channel selectivity for divalent and monovalent cations. J. Gen. Physiol. 88: 293-319, 1986.[Abstract/Free Full Text]
• HURLBUT, W., IEZZI, N., FESCE, R., CECCARELLI, B. Correlation between quantal secretion and vesicle loss at the frog neuromuscular junction. J. Physiol. (Lond.) 425: 501-526, 1990.[Abstract/Free Full Text]
• JONAS, P., MAJOR, G., SAKMANN, B. Quantal components of unitary EPSCs at the mossy fibre synapse on CA3 pyramidal cells of the rat hippocampus. J. Physiol. (Lond.) 472: 615-663, 1993.[Abstract/Free Full Text]
• KORN, H., BAUSELA, F., CHARPIER, S., FABER, D. S. Synaptic noise and multiquantal release at dendritic synapses. J. Neurophysiol. 70: 1249-1254, 1993.[Abstract/Free Full Text]
• KORN, H., SUR, C., CHARPIER, S., LEGENDRE, P., FABER, D. S. The one-vesicle hypothesis and multivesicular release. In: Molecular and Cellular Mechanisms of Neurotransmitter Release, edited by L. Stjärne, P. Greengard, S. Grillner, T. Hökfelt, and D. Ottoson . New York: Raven, 1994, p. 301-322
• LEGENDRE, P., KORN, H. Glycinergic inhibitory synaptic currents and related receptor channels in the zebrafish brain. Eur. J. Neurosci. 6: 1544-1557, 1994.[Medline]
• LEWIS, C. A., FABER, D. S. Inhibitory synaptic transmission in isolated patches of membrane from cultured rat spinal cord and medullary neurons. J. Neurophysiol. 76: 461-470, 1996.[Abstract/Free Full Text]
• LIU, G., TSIEN, R. W. Properties of synaptic transmission at single hippocampal synaptic boutons. Nature 375: 404-408, 1995.[Medline]
• MALGAROLI, A., TSIEN, R. W. Glutamate-induced long-term potentiation of the frequency of miniature synaptic currents in cultured hippocampal neurons. Nature 357: 134-139, 1992.[Medline]
• MANABE, T., RENNER, P., NICOLL, R. A. Postsynaptic contribution to long-term potentiation revealed by the analysis of miniature synaptic currents. Nature 355: 50-55, 1992.[Medline]
• MATTEOLI, M., HAIMANN, C., TORRI-TARELLI, F., POLAK, J., CECCARELLI, B., DE CAMILLI, P. Differential effect of -latrotoxin on exocytosis from small synaptic vesicles and from large dense-core vesicles containing calcitonin gene-related peptide at the frog neuromuscular junction. Proc. Natl. Acad. Sci. USA 85: 7366-7370, 1988.[Abstract/Free Full Text]
• OLIET, S.H.R., MALENKA, R. C., NICOLL, R. A. Bidirectional control of quantal size by synaptic activity in the hippocampus. Science 271: 1294-1297, 1996.[Abstract]
• PARKER, I., YAO, Y. Ca²⁺ transients associated with openings of inositol trisphosphate-gated channels in Xenopus oocytes. J. Physiol. (Lond.) 491: 663-668, 1996.[Abstract/Free Full Text]
• PASHKOV, V., GRICO, N., TSURUPA, G., STORCHAK, L., SHATURSKY, O., HIMMELREICH, N., GRISHIN, E. Monoclonal antibodies can uncouple the main -latrotoxin effects: toxin-induced Ca²⁺ influx and stimulated neurotransmitter release. Neuroscience 56: 695-701, 1993.[Medline]
• PETRENKO, A. G. -Latrotoxin receptor: implications in nerve terminal function. FEBS Lett. 325: 81-85, 1993.[Medline]
• ROBERTS, W. M. Localization of calcium signals by a mobile calcium buffer in frog saccular hair cells. J. Neurosci. 14: 3246-3262, 1994.[Abstract]
• ROPERT, N., MILES, R., KORN, H. Characteristics of miniature inhibitory postsynaptic currents in CA1 pyramidal neurons of rat hippocampus. J. Physiol. (Lond.) 428: 707-722, 1990.[Abstract/Free Full Text]
• SCANZIANI, M., CAPOGNA, M., GAHWILER, B., THOMPSON, S. Presynaptic inhibition of miniature excitatory synaptic currents by baclofen and adenosine in the hippocampus. Neuron 9: 919-927, 1992.[Medline]
• SILVER, R. A., TRAYNELIS, S., CULL-CANDY, S. Rapid-time-course miniature and evoked excitatory currents at cerebellar synapses in situ. Nature 355: 163-166, 1992.[Medline]
• SORRA, K., HARRIS, K. Occurrence and three-dimensional structure of multiple synapses between individual radiatum axons and their target pyramidal cells in hippocampal area CA1. J. Neurosci. 13: 3736-3748, 1993.[Abstract]
• STEVENS, C. F. Quantal release of neurotransmitter and long-term potentiation. Cell 72/Neuron 10, Suppl.: 55-63, 1993.
• STORCHAK, L. G., PASHKOV, V. N., POZDNYAKOVA, N. G., HIMMELREICH, N. H., GRISHIN, E. V. -Latrotoxin-stimulated GABA release can occur in Ca²⁺-free, Na⁺-free medium. FEBS Lett. 351: 267-270, 1994.[Medline]
• TINKER, A., WILLIAMS, A. J. Divalent cation conduction in the ryanodine receptor channel of sheep cardiac muscle sarcoplasmic reticulum. J. Gen. Physiol. 100: 479-493, 1992.[Abstract/Free Full Text]
• TONG, G., JAHR, C. Block of glutamate transporters potentiates postsynaptic excitation. Neuron 13: 1195-1203, 1994.[Medline]
• TSIEN, R. Y. New calcium indicators and buffers with high selectivity against magnesium and protons: design, synthesis, and properties of prototype structures. Biochemistry 19: 2396-2404, 1980.[Medline]
• ULRICH, D., LÜSCHER, H.-R. Miniature excitatory synaptic currents corrected for dendritic cable properties reveal quantal size and variance. J. Neurophysiol. 69: 1769-1773, 1993.[Abstract/Free Full Text]
• VOGT, K., LÜSCHER, H.-R., STREIT, J. Analysis of synaptic transmission at single identified boutons on rat spinal neurons in culture. Pflügers Arch. 430: 1022-1028, 1995.[Medline]
• WANKE, E., FERRONI, A., GATTANINI, P., MELDOLESI, J. Latrotoxin of the black widow spider venom opens a small, nonclosing cation channel. Biochem. Biophys. Res. Commun. 134: 320-325, 1986.[Medline]

0022-3077/97 $5.00 Copyright ©1997 The American Physiological Society