Quantal Size and Variation Determined by Vesicle Size in Normal and Mutant Drosophila Glutamatergic Synapses

Normal occurrence of synaptic vesicle size difference

Although recent results from several studies have indicated that quantal size variation may be attributable to variation in the amount of neurotransmitter released from synaptic vesicles (Frerking and Wilson, 1996; Hanse and Gustafsson, 2001), a model of differential transmitter release has been lacking. Possibilities include differences in vesicle size and differences in duration of fusion pore opening or release efficiency (Choi et al., 2000; Elhamdani et al., 2001; Renger et al., 2001). Via ultrastructural observations of Drosophila synapses, we observed vesicle size differences that could account for mean quantal size and variance. We measured quantal size and synaptic ultrastructural features in two identified Drosophila motor neurons to examine presynaptic and postsynaptic structural features related to quantal amplitude. The two neurons, designated RP3 and 6/7b (Keshishian et al., 1993), innervate the larval ventral longitudinal muscles 6 and 7 and supply synaptic boutons that differ structurally (Atwood et al., 1993; Lnenicka and Keshishian, 2000); the boutons of the RP3 neuron (type 1b) are, on average, larger and possess more synapses, active zones, and mitochondria than those of the common excitor neuron 6/7b (type 1s boutons). Also, the enveloping subsynaptic reticulum (SSR) is more voluminous around type 1b boutons (Atwood et al., 1993) (Fig. 2a). Both neurons use glutamate as their primary neurotransmitter substance (Johansen et al., 1989). In electron micrographs from CS larvae we discovered that synaptic vesicles differ significantly in size in the two bouton types, whereas synaptic contact areas do not (Figs.2, 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Synaptic vesicle measurements for CS type 1b boutons, illustrating the form of the distribution and the effect on the estimated mean diameter of applying two corrections: that ofFroesch (1973)* and that of Parsons et al. (1995)**. For all samples the latter correction produced a larger displacement of the mean value. Note that a substantial fraction of the values appears to the right side of the median, indicating nonuniformity in synaptic vesicle size. Observed and corrected (Froesch, 1973) SDs also are indicated by thehorizontal lines; the correction has a small effect on this value.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Ultrastructure of synapses at CS,dlg^m52 mutant, and rescueddlg neuromuscular junctions. a, Neuromuscular synapses formed by 1b and 1s boutons in CS larva.b, c, Synapses in adlg^m52 mutant larva (b, 1b; c, 1s). d, Synapses of 1b and 1s boutons in a mutant rescued larva (dlg res.). Synapses (SY) are delimited by arrowheads; presynaptic dense bodies (active zone structures) are delimited by indicated arrows. Each synapse occurs in association with subsynaptic reticulum (SSR); synaptic vesicles (SV) are plentiful in all terminals. Note that densely stained synaptic membranes are much more extensive and synaptic vesicles are larger in dlg^m52 mutant boutons than in control (CS and dlg res.) counterparts. Scale bar: (in d) a–d, 1.0 μm.e, f, Serial reconstructions of type 1b boutons from CS (e) and dlg(f) larvae to illustrate a major difference in synaptic structure. Boutons of dlg larvae have larger synaptic areas than controls (red regions); synapses are often confluent. The dense bars are shown in yellow. Scale bars, 1 μm. g, Histogram of the size of synaptic areas in CS (1b, 129 synapses, 15 boutons; 1s, 61 synapses, 15 boutons), dlg^m52 (1b, 3 synapses, 3 boutons; 1s, 37 synapses, 10 boutons), and dlg res. (1b, 10 synapses, 2 boutons; 1s, 4 synapses, 2 boutons). The estimated CV is presented for each bouton type at the top of the corresponding bar.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

The 1s boutons contain larger vesicles than 1b boutons, and dlg^m52 mutants contain larger vesicles than controls. Electron micrographs show vesicles contained within 1b and 1s boutons in CS (1b, 198 vesicles; 1s, 254 vesicles), dlg^m52 (1b, 215 vesicles; 1s, 247 vesicles), and dlg res. larvae (1b, 319 vesicles; 1s, 325 vesicles). Synaptic vesicles are larger on average in 1s boutons for all cases. Scale bar (for all micrographs), 0.1 μm.

The two bouton types, embedded in the SSR, are found close together on the muscle fiber surface (Fig. 2); thus the conditions of fixation would be the same for both. We measured synaptic vesicle diameters (inner and outer) in paired boutons and calculated the corresponding vesicle volumes. Synaptic vesicles in 1s boutons were 18.0% larger in mean corrected outer diameter than those in 1b boutons (p < 0.001) (Fig.3, Table 1). This difference translates into a 67.7% difference in mean vesicle volume (p < 0.001; Table 1). The measurements of uncorrected vesicle outer diameter in 1b boutons agree with values reported previously (Zhang et al., 1998). Thus the two neurons normally possess synaptic vesicles of different mean volume that could, in principle, influence quantal size.

Genetic influence on synaptic vesicle size in dlgmutant larvae

We observed that synaptic vesicles of the dlg null mutant, dlg^m52 , known for its postsynaptic structural abnormalities (Lahey et al., 1994; Budnik et al., 1996), are larger than normal in diameter (p < 0.001 for both 1b and 1s boutons) (Fig. 3, Tables 1, 2) and volume (p < 0.001 for both 1b and 1s boutons; Tables1, 2). Mutant rescues, in which dlg expression was driven both presynaptically and postsynaptically, exhibited reduced vesicle size in both 1b and 1s boutons (Fig. 3, Tables 1, 2). In fact, vesicles were slightly smaller than in controls (p < 0.001 for both 1b and 1s boutons for both diameter and volume); this may relate to the dosage levels of dlg used to drive expression in nerve and muscle tissue. The difference in vesicle size between type 1b and 1s boutons is retained in controls, mutants, and rescues (Fig. 3, Tables 1, 2).

Synaptic contact areas in CS and dlg mutant larvae

To assess whether postsynaptic morphological factors relate to quantal size differences, we made serial reconstructions of CS (Fig.2e), dlg^m52 (Fig.2f), and dlg^m52 mutant rescue boutons, and we measured synaptic contact areas (Fig.2g). In CS larvae the mean synaptic area is not significantly different between 1b and 1s boutons (p = 0.101), in agreement with previous studies (Atwood et al., 1993; Stewart et al., 1996).

Individual dlg^m52 boutons exhibited unusual features, previously unreported; dlg synapses are significantly larger than in controls (1b, p = 0.048; 1s, p = 0.005) and often include “giant” synapses that envelop a larger than normal percentage of the surface area of the bouton (Fig. 2b,c,f). These large synapses could represent separate synapses that have become fused and enlarged. Three reconstructed dlg 1b boutons and three of 10 dlg1s boutons were enveloped by a single confluent synapse. AtDrosophila neuromuscular junctions DLG has important effects on postsynaptic components: clustering of shakerK⁺ channels and the cell adhesion molecule Fas II, and regulating the extent of the SSR (Koh et al., 2000). The enlargement of synaptic contacts represents an additional effect on synaptic structure. In mutant rescues the synaptic areas returned to near-control sizes for 1b boutons (p = 0.97) but were smaller than controls for 1s boutons (p = 0.016) (Fig. 2). Thus both presynaptic structures (synaptic vesicles) and postsynaptic structures (SSR and synaptic contact area) are influenced by the dlg gene. Relationships among synapse size, vesicle size, and quantal size can be tested experimentally for CS and dlg mutant synapses.

Definition of quantal events at individual boutons

Using a focal macropatch electrode to record synaptic activity extracellularly, we compared quantal size at individual visualized boutons. Type 1b and 1s boutons were identified easily for selective recording (Fig. 4a). However, recording from identified, physically isolated boutons in this manner does not ensure complete electrical isolation, because the SSR in which the boutons are embedded often conducts contaminating signals from adjacent boutons (Wong et al., 1999). “Chemical isolation” of boutons by recording in a Ca²⁺-free solution while retaining Ca²⁺ in the macropatch pipette is not feasible, because spontaneous quantal events continue to occur in Ca²⁺-free solutions. Accordingly, statistical procedures have been developed (Wong et al., 1999) to separate with good confidence the spontaneously occurring quantal signals at a selected bouton from contaminants generated at nearby boutons (Fig. 4b).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Separation of quantal events recorded for 1s control bouton into signals and contaminants. a, String of 1b and 1s boutons viewed under DiOC₂(5) fluorescence (left). The macropatch electrode is placed on the muscle surface to enclose the 1s bouton from which mEJCs were recorded, seen under Nomarski optics (right). The circlerepresents the 1s bouton from which recordings were made. Scale bar, 4 μm. b, Distribution of observations separated into signals (open circles) and contaminants (filled circles). Simultaneous records of quantal currents and their respective potentials were plotted, and contaminants were identified as described in Materials and Methods, usingP_s = 0.5 as a cutoff for statistical definition. Inset, Representative mEJCs (top traces) and simultaneously recorded mEJPs (bottom traces). Vertical calibration, 0.4 mV for mEJCs and 3.0 mV for mEJPs; horizontal calibration, 10 msec. c, Angular distribution of the points shown in b. Signals (open bars) reside at shallower angles (66 ± 0.9°; n = 63) than contaminants (filled bars; 86.5 ± 0.2°,n = 57). d, Amplitude–frequency distribution of mEJCs separated into signals (open bars; 0.49 ± 0.04 mV; n = 63) and contaminants (filled bars; 0.09 ± 0.0 mV;n = 57). The distribution of signal amplitudes is skewed toward larger values. Frequency of occurrence of signals is lower after (0.046/sec) than before (0.087/sec) the separation of contaminants. e, Amplitude–frequency distribution of mEJPs belonging to the signal group (mean, 1.20 ± 0.010 mV;n = 63). f, Plot of the rise times versus mEJC amplitudes for both signals (open circles) and contaminants (filled circles). Contaminants are smaller in amplitude and possess longer rise times than signals.Inset, The percentage of cumulative frequency distribution of mEJC rise times belonging to signals (solid line; mean 0.89 ± 0.03 msec; n = 63) and contaminants (dotted line; mean, 1.30 ± 0.07 msec; n = 57). The K–S test indicates that the two distributions are significantly different; rise times of contaminants are slower than those of signals.

Spontaneously occurring quantal current events (mEJCs) at individual boutons in CS and dlg mutant larvae were matched with their corresponding mEJPs recorded intracellularly. Plotting mEJC amplitudes against the corresponding mEJPs usually showed two classes of event. The bouton-specific mEJC signals display a quasi-linear covariation of amplitude with corresponding mEJPs (Fig. 4b, open circles); this covariation is absent in contaminants (Fig.4b, filled circles). The two classes can be distinguished statistically when the observations are reduced to one dimension by calculating the angle (α) each observation makes with the mEJC axis. Signals reside at shallower angles (Fig. 4c,open bars; 4b, open circles) than contaminants (Fig. 4c, hatched bars; 4b, filled circles). The amplitude–frequency distribution of signal mEJCs (Fig. 4d, open bars) displays larger values than the contaminant mEJC distribution (Fig.4d, filled bars). By matching mEJCs with their corresponding mEJPs, we also can generate an amplitude–frequency distribution of signal mEJPs (Fig. 4e). The rise times of mEJCs identified as signals (Fig. 4f, solid line, inset) are significantly faster than those of contaminants (Fig.4f, broken line, inset). Plots of rise time against mEJC amplitude reveal that signals (Fig. 4f,open circles) generally combine larger amplitudes and faster rise times than contaminants (Fig. 4f,filled circles). Thus contaminants can be separated from signals because of their different properties. This method, which reduces the misclassification of contaminants as signals, avoids the variable underestimation of the mean amplitude of signal mEJCs that would result from the inclusion of contaminants. In the remainder of this paper the analysis will be confined to signals.

Accounting for quantal size and variance at 1b and 1s boutons

Quantal size has not been compared previously for 1b and 1s boutons. In macropatch recordings of mEJCs from individual identified 1b and 1s boutons, quantal size is 53.1% larger for 1s boutons (0.47 ± 0.06 mV) than for 1b boutons (0.3 ± 0.02 mV) (Fig.5a). We attempted to account for the factors that generated differences in quantal size between the two bouton types. Specifically, we asked whether quantal size is linked to differences in synapse or vesicle size. In recent studies the immunolabeling of mammalian synapses has shown that the postsynaptic receptor number is related linearly to synaptic area (Nusser et al., 1997; Mackenzie et al., 1999; Oleskevich et al., 1999; Takumi et al., 1999). If this were to hold true at 1b and 1s boutons, the difference in quantal size between the two bouton types could result from differences in the number of postsynaptic receptors available for activation. Alternatively, quantal size differences could result from differences in the amount of transmitter released at 1b and 1s synapses, possibly related to differences in vesicle volume.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Quantal size differs with input and phenotype, whereas variance is similar for all cases. a, Summary of the mean signal mEJC amplitude parameters from the five types of bouton that were investigated (1b, n = 10; 1sn = 11; dlg 1s,n = 10; dlg 1s res.,n = 8; lap 1b, n= 7). A one-way ANOVA test revealed significant differences among the groups (p < 0.0001). Pairwise comparisons of signals via a Student's t test revealed significant differences in mean mEJC amplitudes between 1b and 1s boutons (p = 0.020), 1b and dlg 1s boutons (p < 0.001), 1s anddlg 1s boutons (p = 0.021), 1b and dlg 1s res. (p = 0.030), 1b and lap 1b (p < 0.001), 1s and lap 1b (p = 0.014), 1b and dlg 1s res. (p= 0.030), and dlg 1s res. and lap 1b (p = 0.0016). There was no significant difference between 1s and dlg 1s res. (p = 0.49) and dlg 1s andlap 1b (p = 0.73).b, The quantal variance of the mEJCs (CV_mEJC). A one-way ANOVA test revealed no significant difference among the five bouton types (p = 0.44). For all graphs, error bars indicate SEM.

Synaptic areas were not significantly different between 1b and 1s boutons and therefore could not account for the quantal size differences if receptor density is similar for both types of synapse. However, in 1s boutons outer and inner vesicle volumes were greater than in 1b boutons (see Tables 1, 2) and therefore potentially could explain the quantal size differences.

Another method of testing between the two possibilities was to determine whether variation in quantal amplitude at both bouton types corresponds with variation in synaptic areas or vesicle volume. The coefficient of variation (CV = SD/mean) for mEJC amplitude (or quantal variance, CV_mEJC) has been used in previous studies to identify the locus of variation in quantal amplitude. When quantal variance is greater than the variance attributable to transmitter–receptor interactions (5–15%), postsynaptic receptors likely are unsaturated by a quantum of released transmitter (Faber et al., 1992; Frerking and Wilson, 1996; Liu et al., 1999; McAllister and Stevens, 2000). Quantal variances at 1b and 1s boutons were large, but not significantly different (p = 0.31) (Fig. 5b), indicating that both bouton types share a similar source of quantal variability. We attempted to identify this source of variability and tested whether it resulted from variation in synapse size or vesicle volume.

The standardized cumulative frequency distributions (Nusser et al., 1997) of mEJCs and synaptic areas (Fig.6a,b) were compared by using the K–S statistical test. This test revealed a significant difference between the two distributions for 1s boutons, but not for 1b boutons. Thus the variation in mEJC amplitude probably is not determined by variation in synaptic areas at 1s boutons, although it could be at 1b boutons. If the variance arises from a common source, the very similar CV_mEJC values of the two bouton types exclude synaptic area variation as the main factor.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Quantal variance matched to variations in synaptic area and vesicle volume (inner and outer). Shown is a comparison of the amplitude distribution of mEJCs with the synaptic areas for 1b and 1s boutons (a, b) and with the distribution of the vesicle volumes for 1b, 1s, dlg 1s, anddlg res. 1s boutons (c–f).a, b, The standardized distributions of mEJC amplitudes (solid lines) pooled from all experiments (1b, 423 events, 10 experiments; 1s, 484 events, 11 experiments) compared with the standardized distribution of synaptic areas (gray line) for 1b boutons (a; n = 129) and 1s boutons (b; n = 61). The distributions were significantly different (K–S test) for 1s boutons, but not for 1b boutons. c–f, The standardized distributions of mEJC amplitudes (solid lines) pooled from all experiments (1b, 423 events, 10 experiments; 1s, 484 events, 11 experiments;dlg 1s, 376 events, 10 experiments; dlg1s res., 687 events, 8 experiments) compared with the standardized distribution of outer (gray lines) and inner (broken lines) vesicle volumes for each bouton type (1b, 198 vesicles; 1s, 254 vesicles; dlg 1s, 247 vesicles;dlg 1s res., 325 vesicles). The mEJC and vesicle volume distributions were not significantly different for any bouton type (K–S test).

We next determined whether vesicle size distributions could account for the quantal variances at both bouton types, assuming that all vesicles contain the same concentration of glutamate. We compared the standardized cumulative frequency distributions of mEJC amplitudes and vesicle volumes (Bekkers et al., 1990). For both 1b and 1s synapses the K–S test revealed no significant difference between the two distributions (Fig. 6c,d). The observed CV_mEJC values could be attributed to variation in transmitter content from individual synaptic vesicles related to vesicle volume. Thus the larger mean quantal size recorded at 1s boutons (Fig. 5a) correlates with their larger synaptic vesicles, whereas quantal variance matches that of the vesicle size distribution (Figs. 5b, 6c,d).

Variation in vesicle size could account for quantal variance indlg mutants

Records of mEJCs were analyzed for type 1s boutons ofdlg^m52 mutants. Type 1b boutons indlg preparations were usually difficult to record from in our specimens, being sandwiched between muscles 6 and 7, so they were not included in this analysis. Mean quantal size of dlg 1s boutons was significantly larger (51.0%) than for CS 1s boutons (Fig.5a). In dlg 1s boutons the synapses and corrected outer (inner) vesicle volumes were 258.3 and 47.8% (81.5%), respectively, larger than in CS 1s boutons. The increase in quantal size correlated well with the increase in vesicle volume, but not with that of synapse area, indicating that the amount of transmitter released is the more likely determinant of quantal size atdlg 1s boutons. Similar CV_mEJC values were found for dlg 1s boutons and CS 1s boutons (p = 0.95) (Fig. 5b), indicating a common source of variability. Standardized cumulative frequency histograms of mEJC amplitudes and vesicle volumes were compared. Atdlg 1s boutons the K–S test revealed no significant difference between the two distributions (Fig. 6e).

In dlg rescues (dlg res.) the quantal size was restored to control values (Fig. 5a). The K–S test revealed no significant difference between standardized cumulative frequency plots of quantal amplitude and vesicle volume (Fig.6f). Because CV_mEJC values for CS Is, dlg 1s, and dlg 1s res. were similar (Fig.5b), despite large differences in synapse area (Fig.2g), the common source of variability most likely arises from variability in vesicle volume. Thus the larger vesicles ofdlg 1s boutons and their size variation can account for the larger quantal currents and their observed variation. These results indicate that the dlg^m52 mutation most likely increases quantal size via an effect on the size of synaptic vesicles.

To ascertain further whether vesicle size is the likely determinant of quantal size, we examined the lap mutant for which the 1b boutons contain vesicles similar in mean uncorrected outer diameter (49 nm) (Zhang et al., 1998) to those of dlg 1s boutons (lap 1b vesicles are ∼14.7% smaller in uncorrected outer volume than dlg 1s vesicles). Quantal size in laphas been measured by Zhang et al. (1998) by using whole-cell voltage clamp, which does not distinguish the boutons from which mEJCs originate. In the present study we selected 1b boutons to compare their mEJCs with those of dlg^m52 1s boutons. Quantal size (Fig. 5a) and variance (p = 0.091) (Fig. 5b) were not significantly different for the two bouton types. Again, the similar CV_mEJC values indicate that quantal variance is determined predominantly by variation in vesicle volume. Furthermore, if receptor subcomposition or number (Petersen et al., 1997) were different between 1b and 1s bouton types, similar-sized vesicles containing similar amounts of transmitter should produce different quantal sizes in the lap anddlg^m52 comparisons; however, this is not the case. Thus quantal size and variance seem to be controlled by presynaptic factors, namely, vesicle size.

General relationship between quantal size and vesicle volume

A linear relationship between vesicle volume and quantal amplitude appears when mean values for both measurements are plotted for all of the different genotypes of the present study (Fig.7). This relationship further supports the hypothesis that quantal amplitude is determined to a large extent by synaptic vesicle volume and transmitter content. If quantal size indirectly represents the glutamate that has been released, the linear relationship in Figure 7 indicates that the concentration of glutamate (mol/vol) in synaptic vesicles is constant. Because this relationship was established by recording from different inputs and genotypes, it further indicates that glutamate concentration is regulated homeostatically to a constant value, although vesicle size varies among inputs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

A plot of mean quantal amplitude against the mean outer (a) and inner (b) vesicle volumes (calculated from both outer and inner vesicle diameters, respectively) for five bouton types. A linear fit is applied to the data points (uncorrected outer volumes: p = 0.0031, r = 0.98, slope = 0.01 mV/zL,filledcircles; corrected outer volumes:p = 0.0057, r = 0.96, slope = 0.008 mV/zL, open circles; uncorrected inner volumes: p = 0.043, r = 0.86, slope = 0.041 mV/zL, filled squares; corrected inner volumes: p = 0.033,r = 0.88, slope = 0.035 mV/zL, open squares). The constant slopes imply that the concentration of transmitter is the same for all vesicles.

Vesicle size and synaptic differentiation

The normally occurring size difference between synaptic vesicles in two identified neurons constitutes a new feature of synaptic differentiation. Electron microscopic studies previously have revealed differences in size and shape between GABAergic and glutamatergic boutons in the mammalian CNS (Uchizono, 1965; Hámori et al., 1990) and in arthropod excitatory and inhibitory neurons (Uchizono, 1967; Atwood and Tse, 1993), but a vesicle size difference between two glutamatergic neurons innervating the same postsynaptic target cell has not been found previously. In Drosophila the functional outcome is a difference in quantal size; fewer quantal events would be needed to achieve a given level of excitation at 1s boutons, which in fact generate larger EJPs (Kurdyak et al., 1994;Lnenicka and Keshishian, 2000).

Vesicle size has been found to respond to the level of synaptic activity in several instances. Inactivity was shown to cause an enlargement of synaptic vesicles in the electroreceptor afferents of gymnotid fish, which can be reversed by stimulation (Maler and Mathieson, 1985), whereas prolonged high-frequency stimulation produces a reduction in vesicle size in lamprey reticulospinal axons (Wickelgren et al., 1985) and in the electric organ of Torpedo(Zimmerman and Whittaker, 1974). By recording motor pattern discharge (fictive locomotion; Cattaert and Birman, 2001) from either 1b or 1s boutons, we found that almost all of the synaptic activity was generated by 1b boutons, which contain the smaller vesicles; very little activity was recorded from 1s boutons containing the larger vesicles (our unpublished observations). Probably 1s boutons participate in less frequent, more vigorous motor responses. Thus our preliminary observations are consistent with those of previous studies: synapses experiencing more electrical activity have smaller synaptic vesicles.

Accounting for quantal variance

Synapse area and vesicle volume variations in control anddlg mutant boutons provided an opportunity to test whether quantal variance could be explained by either of these morphological features. Recent results for mammalian central synapses have not favored one alternative over the other consistently (Auger and Marty, 2000). The amplitude distribution of mEJCs correlates well with the third power of vesicle diameter (or vesicle volumes), as reported for hippocampal cultures and slices (Bekkers et al., 1990) and monoamine-secreting cells (Finnegan et al., 1996; Bruns et al., 2000). In contrast, Frerking et al. (1995) found that mIPSC amplitudes correlate with the sixth power of vesicle diameter distributions at dynapses in cultured amacrine cells; they concluded that variation in mIPSC amplitudes arises from variation in the amount of released transmitter and that the correlation with the sixth power of vesicle diameter relates to the requirement for two transmitter–receptor interactions for channel opening. At present, we are not able to account for the difference in the power dependence of GABAergic mIPSCs (Frerking et al., 1995) and glutamatergic EPSCs (Bekkers et al., 1990; this study). All of these studies support the implication that quantal amplitude variability arises from variation in the amount of glutamate released by a synaptic vesicle. For Drosophila the strong correlation between the distributions of mEJCs and vesicle volumes at all inputs and the linear relationship between quantal size and vesicle volume indicate that the vesicles empty their transmitter content abruptly during exocytosis; partial emptying of vesicles during exocytosis, as suggested at other synaptic types (Choi et al., 2000;Renger et al., 2001), may not contribute to the quantal variability atDrosophila synapses. The large and similar CV_mEJC values recorded in all five types of synapse indicate that postsynaptic receptors of these synapses are not saturated. This is consistent with observations in other systems (Silver et al., 1996; McAllister and Stevens, 2000; Ishikawa et al., 2002) and is in accord with the type of model proposed by Bartol et al. (1991) in which receptors close to the point of vesicle exocytosis are activated strongly by released transmitter, whereas those further away are less likely to be activated but could be recruited if more transmitter is released by a vesicle.

The alternative explanation, variation in number of receptors at individual synapses with saturation or near-saturation of available receptors, is less well supported by our data, although it has been proposed in studies of mammalian synapses (Tang et al., 1994; Harris and Sultan, 1995; Nusser et al., 1997; Lim et al., 1999). If postsynaptic receptors were saturated, quantal variance would be small, arising from stochastic receptor–transmitter interactions (Frerking and Wilson, 1996). The studies of Liu et al. (1999) and McAllister and Stevens (2000) conducted on individual, cultured glutamatergic hippocampal boutons (which display large quantal variations) indicate that only a fraction of the quantal variance (10%) arises from stochastic transmitter–receptor interactions, but the studies fail to account for the factors producing the bulk of the quantal variance. Our work on naturally formed glutamatergic synapses indicates that most of the quantal variance arises from variation in vesicular glutamate content. Our data do not suggest roles for “off-center” release (Uteshev and Pennefather, 1996) or structural variability of the cleft (Stiles et al., 2001) as significant contribution factors.

Glutamate concentration of synaptic vesicles

For quantal amplitude to be proportional to vesicle volume (Fig.7), the concentration of glutamate would have to be equal in all vesicles. Studies in other systems support this proposition; for example, monoamine-secreting vesicles of PC12 cells (Colliver et al., 2000) and 5-HT-containing vesicles of cultured leech Retzius cells (Bruns et al., 2000) contain uniform transmitter concentrations. However, in developing cholinergic neuromuscular junctions ofXenopus, the overexpression of a vesicular acetylcholine transporter increased quantal size (Song et al., 1997); whether this effect was accompanied by changes in vesicle size is not known. The present result indicates that constancy of vesicular transmitter concentration holds for normally formed glutamatergic synapses. Probably glutamate concentration is regulated homeostatically in the same way at all inputs that have been examined in the present study. The genetic interventions (dlg and lap mutants) affect vesicle size, but probably not intravesicular glutamate concentration. The vesicular glutamate transporter may load vesicles to their maximal capacity, in which case larger vesicles would contain more glutamate (cf. Sulzer and Edwards, 2000), in accordance with a “set point” rather than a “steady-state” model of vesicle filling (Williams, 1997).

Genetic regulation of vesicle size

Occurrence of larger synaptic vesicles indlg^m52 indicates regulation of vesicle size by the dlg gene, perhaps via a mechanism different from that involving lap and stoned genes, which affect vesicle retrieval and recycling steps (Zhang et al., 1999). DLG is known to be expressed both presynaptically and postsynaptically inDrosophila synapses and has effects on synaptic structure (Lahey et al., 1994; Budnik et al., 1996; Guan et al., 1996; Thomas et al., 1997). The present data indicate that DLG affects vesicle size, but it has not been found in association with clathrin coats or synaptic vesicle membranes, unlike LAP and Stoned proteins. Nevertheless, other tumor suppressor genes are shown to affect endocytosis in vivo (Krishnan et al., 2001). The mammalian homolog PSD-95, detected presynaptically as well as postsynaptically (Aoki et al., 2001), transiently associates with perinuclear vesicles during sorting in hippocampal neurons (El-Husseini et al., 2000). InDrosophila epithelial cells the dlg,lgl, and scribble tumor suppresser genes have been implicated in targeting of transport vesicles to cellular compartments and plasma membranes to maintain cell polarity (Bilder et al., 2000). Furthermore, the lgl gene product was reported to coat transport vesicles (Peifer, 2000). Thus products of the tumor suppressor genes are associated with vesicles and could influence their size.

Physiological implications

Our work shows that quantal size differences are correlated with vesicle size among synaptic inputs. We postulate that larger vesicles release more transmitter and thereby provide greater weighting per stimulus in synaptic excitation of the postsynaptic cell. The shared source of quantal variation is postulated to be variation in transmitter content among vesicles as determined by vesicle volume.

It is also possible that vesicle size may influence release probability, as indicated by theoretical studies (Glavinovic and Rabie, 2001). Probability of transmitter release is higher than normal indlg mutants (Budnik et al., 1996) and higher at 1s synapses than at 1b synapses (Atwood et al., 1997); this order of release probabilities corresponds with the order of vesicle sizes.