Presynaptic Regulation of Quantal Size by the Vesicular Glutamate Transporter VGLUT1

Loading in excitatory synaptic vesicles depends on glutamate concentration and number of available transporters

To characterize the nature by which VGLUT1 expression might impact synaptic transmission, it was first important to demonstrate that varying the parameters of glutamate transport could actually modulate the extent of vesicle filling. A possible scenario is that although different transport conditions might influence the speed of filling, vesicles eventually fill to a “set point” that is dictated by some parameter independent of transport, such as vesicle size. In contrast, if the amount of glutamate loaded into vesicles was to be specified by a dynamic equilibrium process that could be stoichiometrically influenced by the components of the loading reaction, it would allow for the modulation of vesicle filling and the consequent release of glutamate by endogenous characteristics such as transporter expression.

Previous work has assayed the loading of glutamate into PC12 cells (Bellocchio et al., 2000), the BON line of human endocrine cells (Takamori et al., 2000), and even purified synaptic vesicles at specific external glutamate concentrations (Maycox et al., 1988; Wolosker et al., 1996). However, given a recent set of insightful studies that has begun to elucidate the previously unknown range of the cytosolic glutamate concentration (Ishikawa et al., 2002; Yamashita et al., 2003), we were interested in systematically measuring the glutamate uptake equilibrium around variants of this physiological range of conditions as well as assessing the influence of transporter number in defining this equilibrium.

We therefore quantified the time course and final equilibrium of glutamate transport under different conditions using VGLUT1 synaptic vesicles isolated from rat cerebral cortex and hippocampus (Fig. 1). Observing the uptake of the vesicles of 50 μm ³H-glutamate (Fig. 1A) revealed that vesicular transport in our system reached an equilibrium balance between transmitter influx and efflux along a similar time course to that reported previously (Maycox et al., 1988; Wolosker et al., 1996). To examine whether the number of transporters participating in loading could influence its final equilibrium, we also compared our equilibrium uptake to that occurring in the presence of Rose Bengal, an antagonist of VGLUT1 loading that is believed to interact allosterically with the transporter (Ogita et al., 2001). Because the drug is thought to noncompetitively disrupt glutamate transport at the site of loading, it offered a pharmacological means for reducing the effective number of transporters participating in uptake to examine whether the size of the available transporter pool can help determine the final amount of vesicle filling. Loading in the presence of Rose Bengal indeed exhibited a reduced equilibrium of filling (Fig. 1A), a property that was also observed under treatment with trypan blue, the most potent known competitive antagonist of vesicular glutamate transport (Roseth et al., 1998). Both drugs were applied at approximate IC₅₀ concentrations, whereby 2 μm at either 5 or 10 min preincubation achieved complete inhibition of loading (data not shown).

By repeating these treatments under a more physiological concentration of extravesicular glutamate (1 mm) (Fig. 1B), we were next able to observe whether transporter activity can also influence vesicle filling at a glutamate concentration closer to what has endogenously been estimated in functional studies (Ishikawa et al., 2002). At this higher concentration of glutamate, the vesicles now filled to a higher steady-state concentration, indicating that the final extent of VGLUT1 filling depends directly on the concentration of glutamate around the vesicle and does not have an independent set point. Equilibrium filling was again reduced by the presence of either RB or TB (Fig. 1B), demonstrating that, also at this higher concentration, the equilibrium can be shifted by reducing the number of active transporters contributing to the loading process.

Varying the external glutamate concentration along a continuum while measuring 10 min equilibrium uptake levels (Fig. 1C) reveals that maximal uptake increases linearly up to 1 mm exogenous glutamate and begins to saturate at ∼4-6 mm. Even at these concentrations, the extent of transporter availability is critical in determining the equilibrium value, because RB attenuates the final concentration reached for all values of external glutamate in a similar manner to TB, although competitive antagonism of TB is eventually overwhelmed as glutamate concentration is increased. Together, the results lead us to conclude that VGLUT1 vesicles are subject to a flexible degree of filling that can be influenced by both the concentration of glutamate outside of the vesicle as well as the effective number of available transporters. Aligning the time courses of loading observed under various glutamate concentrations (Fig. 1D) demonstrates that both the magnitude and, to a lesser degree, the rate of uptake are influenced by alterations in transport conditions.

We thus observe three properties of glutamate uptake into synaptic vesicles at physiological terminal glutamate concentrations. First, additional increases to external glutamate concentrations can enhance the amount of transmitter loaded, as predicted by the aforementioned functional work (Ishikawa et al., 2002; Yamashita et al., 2003). Second, the number of transporters available for loading can influence the final loading equilibrium, even at glutamate concentrations above the physiological range. Third, vesicle filling is a slow process, requiring minutes to reach completion, which could indicate an importance of changes to the vesicle loading rate, in addition to the loading amount, in determining physiological synaptic transmission.

Transporter overexpression in PC12 cells and hippocampal synapses

If VGLUT1 vesicles do not merely accumulate transmitter until full but rather have their speed and amount of filling determined by an equilibrium process that is subject to the available concentration of transmitter and transporter, we were interested in examining whether this equilibrium could also be shifted in a more physiological context. Therefore, we next attempted to vary the amount of vesicular transporters expressed in the synaptic terminals of neurons to examine the resulting consequences for glutamate release in endogenous synaptic transmission.

To characterize the functional relationship between VGLUT1 expression and excitatory synaptic transmission, we attempted to overexpress the transporter at hippocampal synaptic terminals. We created two expression constructs for delivery into neurons: one for wild-type VGLUT1 (WT) and another tagging VGLUT1 with the reporter GFP at its C terminus. To verify the effectiveness of our constructs in enhancing VGLUT1 expression, PC12 cells were transiently transfected as described previously for VGLUT2 (Varoqui et al., 2002) and the recently identified VGLUT3 (Schafer et al., 2002). A Western blot analysis (Fig. 2A) using a guinea pig antibody specific for VGLUT1 (Schafer et al., 2002) reveals that although mock-transfected cells (M) did not express detectable levels of VGLUT1, single immunoreactive bands of appropriate size were observed for PC12 cells transfected with either the wild-type VGLUT1 or the VGLUT-GFP fusion (+GFP) constructs. The fusion protein had an ∼26 kb higher molecular weight than wild-type VGLUT1 because of the additional mass of the fused EGFP. To check the functionality of these newly synthesized proteins, we observed whether transfection of our constructs into PC12 cells gave rise to glutamate transport by assaying for radioactive glutamate uptake in the presence of magnesium-ATP and chloride ions to activate transporter function (Fig. 2B). Here, both WT and GFP-tagged (+GFP) constructs demonstrated successful transport of ³H-glutamate into a light population of membranes, a function that was found to be inhibited by the proton uncoupler CCCP. CCCP-sensitive uptake in this and similar preparations of vesicles is considered a valid indicator of the H⁺-dependent uptake of glutamate (Takamori et al., 2000, 2002). Although notable levels of uptake were not observed in mock-transfected cells (Mock), uptake capacity for both constructs was found to be comparable with those reported previously for VGLUT1 (Bellocchio et al., 2000; Takamori et al., 2000) as well as for VGLUT2 and VGLUT3 (Schafer et al., 2002; Varoqui et al., 2002). We therefore concluded that our constructs were effective for increasing VGLUT1 protein levels and enhancing glutamate transport.

We next examined whether VGLUT1 protein levels could be enhanced at the synaptic terminals of hippocampal neurons. This requires not only an increase in the production of VGLUT1 protein in transfected neurons but also that the VGLUT1 generated from our constructs contain the correct targeting sequence to be transferred to synaptic terminals effectively. We examined VGLUT1 targeting by transfecting neurons with the VGLUT1-GFP fusion construct and visualizing protein localization. Here, confocal imaging of the fluorescent-tagged protein (Fig. 2C, top) showed that new protein was successfully targeted to the synaptic terminals, where it accumulated in discrete puncta colocalized with the presynaptic bouton label FM4-64 and was not observable in dendrites. In contrast, when GFP was cotransfected separately from VGLUT1 to fully label neurons overexpressing the protein for later experiments (Fig. 2C, bottom), the GFP diffused evenly throughout the cell, with no specific targeting to synaptic terminals. We therefore conclude that overexpressed VGLUT1 can be targeted to hippocampal synapses.

Although the fusion construct proved useful for establishing the targeting of overexpressed protein to synaptic terminals, we chose to use wild-type VGLUT1 for the subsequent functional characterizations to avoid any modification of VGLUT1 structure that could perturb insertion at the vesicular membrane or disrupt potential endogenous regulation at the terminal. It was then necessary to verify that our VGLUT1/GFP cotransfection also succeeded in increasing total VGLUT1 protein levels at the synaptic terminal. Figure 2D shows immunolabeled VGLUT1 occurring within synaptic terminals marked with synapsin I along axons cotransfected with GFP and VGLUT1 (top images) or transfected with GFP alone (bottom images). We found that VGLUT1 puncta appeared more intense in synapses occurring on the GFP-labeled transfected axon than in neighboring puncta occurring off the GFP axon (top images). As an additional control demonstrating that GFP transfection alone did not enhance VGLUT1 levels per synaptic terminal, the above comparison was repeated with GFP axons lacking VGLUT1 cotransfection, again versus neighboring GFP-negative terminals (Fig. 2D, bottom images). In this case, there did not appear to be an appreciable difference in the size or intensities of VGLUT1 puncta occurring on or off the GFP-labeled axon.

The absolute amount of VGLUT1 per synapse is determined by both the number of synaptic vesicles and the average amount of VGLUT1 protein per vesicle. It was important to discern whether any increase in VGLUT1 levels that we observed was indeed attributable to a specific change in the concentration of VGLUT1 across the synapse and not just a general increase in the size or density of the vesicle pool. Because synapsin I intensity provides a measure of the concentration of a second protein associated with vesicles, we used the ratio of VGLUT1 to synapsin I to determine the available amount of transporter protein independently of fluctuations in the number of vesicles or the synaptic area. Figure 2E demonstrates a quantification of this measure, which utilizes the IOD of VGLUT1 and synapsin I to compare the expression levels of the two proteins. Synapses from axons transfected with both GFP and VGLUT1 show a nearly twofold increase in the VGLUT1/synapsin I IOD ratio compared with neighboring synaptic terminals (Fig. 2E, GFP+VGLUT, left plot). GFP alone axons, in contrast, did not show noticeably larger VGLUT1/synapsin I IOD ratios than in neighboring terminals (Fig. 2E, GFP Alone, right plot). Repeating this measure across multiple coverslips demonstrates a consistent increase in the targeting of VGLUT1 protein to synapses as a result of GFP plus VGLUT1 cotransfection (Fig. 2F), with no increase in VGLUT1 protein at neighboring synapses in the same neuronal network. We therefore conclude that our constructs were capable of effectively targeting elevated levels of VGLUT1 protein to the synaptic terminals of our hippocampal cultures.

VGLUT1 overexpression enhances AMPA receptor-mediated responses and decreases evoked failure rate

Having demonstrated a successful elevation of VGLUT1 protein levels in synaptic terminals, we examined the physiological consequences for functional synaptic transmission. Traditional electrophysiological methods record events postsynaptically, such that postsynaptic genetic manipulations have the possibility of influencing all recorded events. Presynaptic modifications, however, are more difficult to assay, because they require strict differentiation over which terminals are contributing to measured events. To solve this problem, we used dual whole-cell patch clamp, evoking an action potential in a presynaptic cell under current clamp while recording the resulting EPSCs mediated by AMPA receptors in a neighboring postsynaptic cell under voltage clamp (Fig. 3A, traces). Using confocal microscopy (Fig. 3A, image), we could distinguish presynaptic cells that were transfected with VGLUT1 by their coexpressed GFP label, as opposed to control neurons that did not express GFP. A common paradigm was to elicit transmission from a VGLUT1 neuron to a control neuron and then, as an independent control, evoke synaptic transmission under reverse configuration to compare reciprocal transmission. Figure 3B shows evoked synaptic transmission in a reciprocally connected pair of neurons (top), where stimulating a VGLUT-enhanced neuron evoked an EPSC that was an average of 92% larger than that evoked from the control neuron in the reverse direction (left traces). Normalizing these traces to the same peak amplitude (Fig. 3B, right traces) revealed a similar average time course. Visualizing the distribution of events evoked from this reciprocal pair (Fig. 3C) reveals a rightward shift in the VGLUT+ distribution relative to control (p < 0.0001; Kolmogorov-Smirnov test), and, on average, control events were consistently exceeded by their VGLUT+ counterparts in EPSC amplitude (12.5 ± 2.9 vs 20.5 ± 3.4, respectively; p < 0.05, unpaired t test; n = 7 pairs control, 7 pairs VGLUT+) (Fig. 3D).

Of potential interest, VGLUT+ transmission also exhibited an unexpected decrease in the failure rate of synaptic transmission (0.54 ± 0.11 vs 0.18 ± 0.13, respectively; p = 0.03 unpaired t test) (Fig. 3E), suggesting that transporter overexpression leads to an increase in either of the following: (1) the number of release events that are large enough to elicit detectable AMPA receptor-mediated responses, or (2) the probability that a vesicle will be released after an action potential.

VGLUT1 overexpression enhances quantal synaptic transmission

Increases in evoked synaptic transmission after VGLUT1 overexpression could be a result of changes in the number of synaptic terminals, the probability of release per terminal, or the individual quantal size of response following each release. Therefore, we next examined whether quantal transmission is impacted by VGLUT1 overexpression by monitoring miniature synaptic transmission under bath application of tetrodotoxin to block action potentials and ensure spontaneous, single-vesicle release in control versus VGLUT1-overexpressing neurons. The difficulty in interpreting this experiment, of course, is that not all presynaptic terminals in a recorded neuron can be successfully transfected. Our strategy then was to rely first on a high-transfection rate (∼25%) and, from this population, select areas of the coverslip unusually dense in transfected cells and their axons (Fig. 4A, bottom image). Under this condition, for a given recorded neuron, a perceptible proportion of presynaptic terminals should contain higher levels of VGLUT1. Comparing responses observed here (VGLUT+) with responses collected in a completely nontransfected coverslip (control), we attempted to assay for detectable changes in quantal transmission.

mEPSCs recorded from neurons in areas rich in VGLUT+ terminals were larger than those recorded in cells devoid of enhanced terminals. Comparing distributions from sample VGLUT+ and control neurons indicated a rightward shift of amplitudes in the VGLUT+ cultures (Fig. 4B) (p < 0.001; Kolmogorov-Smirnov test), whereas aligning sample and average traces (Fig. 4C) further emphasized the difference. On average, the size of mEPSCs from the neurons with VGLUT1 transfection was 31% larger than that of neurons in the control group (13.6 ± 0.6 vs 17.8 ± 2.2, respectively; p = 0.04 unpaired t test; n = 5 cells control, 8 cells VGLUT+) (Fig. 4D). In Figure 4D, we also compared the effects of VGLUT transfection in miniature versus evoked synaptic transmission. Control mEPSCs and EPSCs were similar in amplitude, suggesting that evoked EPSCs most likely originated from single synapses. Interestingly, although EPSCs evoked from transfected terminals were 64% larger than control, VGLUT+ mEPSCs only increased by 31%. This difference is expected from the fact that the VGLUT+ mean mEPSC amplitude was calculated from a population of events that included both VGLUT-enhanced and control terminals, and also given likely differences in detectability between evoked and spontaneous transmission (Zhou et al., 2000). Examining mEPSC frequency between the two pools of cells revealed no significant change (1.89 ± 0.54 vs 1.55 ± 0.56, respectively; p = 0.33; unpaired t test) (Fig. 4E). Similarly to what we observed in evoked transmission, no change was observed in the time course of control and VGLUT-enhanced spontaneous transmission (data not shown), suggesting that VGLUT1 overexpression did not induce detectable changes in postsynaptic receptor properties.

VGLUT-enhanced transmission results from increased glutamate release per vesicle

Changes in evoked and quantal transmission after VGLUT1 overexpression could be a result of either presynaptic changes in cleft glutamate concentration or postsynaptic changes in receptor number or properties. Therefore, we next investigated whether the observed enhancement to evoked and miniature excitatory transmission following VGLUT1 overexpression was a result of an increased amount of glutamate released from synaptic vesicles. To assess the extent of glutamate release during synaptic transmission, we made use of a previously established pharmacological approach (Liu et al., 1999; Wadiche and Jahr, 2001). The low-affinity antagonist γ-DGG competes with glutamate for postsynaptic AMPA receptors, such that it partially attenuates glutamatergic release events at hippocampal synapses. Because of the competitive nature of its interaction with glutamate, the degree of its attenuation is inversely related to the amount of glutamate being released, such that it can be used as a quasidirect indicator of the extent of glutamate release (Liu et al., 1999). By recording EPSCs and determining the extent of γ-DGG attenuation, we could obtain a relative metric of the amount of glutamate involved in a population of synaptic release events.

Therefore, we repeated our dual whole-cell patch-clamp procedure (Fig. 3), this time with perforated patches to obtain long-duration recordings. We evoked EPSCs from either control or VGLUT-enhanced presynaptic neurons and then applied γ-DGG to measure the impact of the antagonist on the transmission. By repeatedly evoking transmission at a regular interval, it was possible to establish a baseline amplitude that could then be reliably attenuated by the perfusion of 600 μm γ-DGG (Fig. 5A), a concentration that we found optimal for partial attenuation of typical transmission (Fig. 5B). A comparison of average traces from before and after antagonist application (Fig. 5C) at both control synapses (left traces) and VGLUT-enhanced synapses (right traces) depicts sample attenuation of control and VGLUT-enhanced transmission. As shown, the efficacy of γ-DGG in EPSC attenuation was weaker at synapses containing higher levels of VGLUT1. On the whole, γ-DGG reduced the size of control EPSCs by 51%, in contrast to 31% in VGLUT-transfected pairs (Fig. 5D) (control vs VGLUT+; 51 ± 5 vs 31 ± 11%, respectively; p < 0.05, unpaired t test; n = 10 pairs control, 6 pairs VGLUT+). This line of experiments provides direct evidence that the VGLUT-enhanced synaptic transmission that we observed is indeed a result of enhanced presynaptic glutamate release from synaptic terminals.

The amount of glutamate released by synaptic terminals can be varied by changes to the amount of glutamate released per vesicle, which is likely because VGLUT1 can impact vesicle loading (Fig. 1). However, we cannot rule out the possibility that VGLUT1 overexpression is instead altering the probability of release of the terminals, which in turn could impact the likelihood of multiple vesicles fusing simultaneously at the same terminal (multivesicular release) (Tong and Jahr, 1994; Oertner et al., 2002). To directly distinguish between these two possibilities, we repeated the above γ-DGG attenuation experiments in a low probability of release condition, in which divalent cation concentrations were systematically varied until the observed transmission evoked not only approximated the quantal size (Fig. 4D) but also exhibited a high-failure rate. If a low proportion of stimuli results in the release of vesicles, then the proportion of stimuli that fosters multivesicular release will be exceedingly low. If vesicles release independently of one another as is widely assumed and synaptic boutons could be comprised of ∼10 vesicles (Schikorski and Stevens, 1997), then the probability of a given EPSC involving two or more vesicles can be determined using binomial statistics (see Materials and Methods), and MVR will be minimized when probability of release is low and failure rates are high. Evoked synaptic transmission in control and VGLUT+ pairs was thus tuned for a high level of baseline failures in each individual pair using different ratios of calcium and magnesium concentrations, as depicted in Figure 5E. As shown in Figure 5F, the average failure rate of transmission used in these experiments was 0.64 for control pairs and 0.79 for VGLUT+ pairs. Therefore, during transmission in these experiments, an upper bound for the incidence of MVR (assuming synapses with 10 sites) was ∼7 and 3%, respectively, and the majority of synaptic transmission should have been mediated by single vesicle release events. If the synapses in our study had even fewer release sites, then the likelihood of MVR would have been still further reduced.

Once such a low probability of release (P_r) condition had been achieved in each connection, we measured γ-DGG attenuation. In this new condition, favoring the release of single vesicles, γ-DGG attenuation in control versus VGLUT+ pairs were again significantly disparate (Fig. 5D) (control, low P_r vs VGLUT+, low P_r, 52 ± 7 vs 29 ± 14%, respectively; p < 0.05, unpaired t test; n = 5 pairs control, 5 pairs VGLUT+). Notably, even when the failure rates were made to differ dramatically between control and control, low P_r (0.41 ± 0.12 vs 0.64 ± 0.05; p < 0.05; unpaired t test) (Fig. 5F, left bars) and between VGLUT+ and VGLUT+, low P_r (0.30 ± 0.10 vs 0.79 ± 0.05; p < 0.01; unpaired t test) (Fig. 5F, right bars), similar degrees of γ-DGG attenuation were observed regardless of these manipulations to failure rate. Thus, VGLUT1 overexpression impacts the amount of glutamate released from single vesicles.

VGLUT1 expression is developmentally regulated

Although the above results demonstrate the potential of VGLUT1 regulation to alter synaptic transmission and establish the molecule as a limiting step for controlling glutamate release, we were interested in ascertaining whether this control point was used endogenously as a physiological mechanism for regulating synaptic transmission. In particular, our previous work (Renger et al., 2001) indicates that developing synapses exhibit several presynaptic functional differences from mature ones, including a reduced number of docking vesicles and less active vesicle cycling, that are rectified during maturation. Given that VGLUT1 can influence transmitter release (Fig. 5), we were interested in determining whether an additional maturational change might involve a developmental increase in VGLUT1 expression.

To monitor the developmental regulation of VGLUT1, quantitative immunohistochemistry was used to determine the amount of transporter expressed at each stage of development (Fig. 6). Immunostaining in mature hippocampal cultures (>14 DIV) depicts the transporter as discrete puncta that are tightly colocalized with synaptotagmin (Fig. 6A). Repeating this procedure across various stages of development reveals that expression of the transporter appears to markedly increase at later stages compared with earlier ones. This maturation results in an increased proportion of synapses expressing detectable levels of VGLUT1 (Fig. 6B) (n = 11 images for each group; 6 vs 9 DIV, p < 0.001, unpaired t test; 9 vs 20 DIV, p < 0.01; no significant change between 20 and 27 DIV). Of the synapses expressing detectable levels of the transporter, the ratio of VGLUT1 and synaptotagmin IODs per synapse also exhibits a graded increase across development (Fig. 6C) (6 vs 9 DIV, p < 0.01, unpaired t test; 9 vs 20 DIV, p < 0.01; no significant change between 20 and 27 DIV), indicating that the amount of VGLUT1 protein expressed per synaptic terminal is developmentally upregulated.

Given that VGLUT1 is upregulated within synapses across development, we were interested in identifying functional changes that could potentially be related to its gradual expression. One of our observations is that the peak amplitude of quantal excitatory synaptic currents seemed to undergo a strengthening as synapses matured during development (Fig. 6D), a trend that was significant across all populations of recorded neurons. mEPSC mean amplitude increased from 7.9 ± 1.3 pA at 8 DIV to 11.1 ± 0.9 pA at 10 DIV to 14.3 ± 0.9 pA by 15 DIV (Fig. 6E) (n = 8, 8, and 6 cells, respectively; p < 0.05; unpaired t test), thus mirroring a developmental increase that has been reported in the brainstem for AMPA receptor-mediated mEPSCs (Ishikawa et al., 2002). Although this effect could also coincide with changes in the number and conductance properties of postsynaptic receptors, the significant increase of VGLUT1 protein expressed in synaptic terminals during this period (Fig. 6C) could also contribute, at least in part, to this observed maturational change. Although a significant developmental increase was observed for mEPSC frequency between 10 and 15 DIV (8 DIV, 0.20 ± 0.004 Hz; 10 DIV, 0.31 ± 0.10 Hz; 15 DIV, 1.20 ± 0.45 Hz; p < 0.05; unpaired t test) (Fig. 6F), given our findings on the lack of influence of VGLUT1 expression on mEPSC frequency (Fig. 4E), we cannot necessarily conclude that VGLUT1 participates in this phenomenon.

VGLUT1 expression is coordinated with the onset of functional vesicle cycling

Previous work in our group (Renger et al., 2001) has shown that silent synapses can arise in part from incomplete activation of postsynaptic AMPA receptors by a diminished presynaptic release of glutamate. The work demonstrated that conversion from silent to functional transmission could correspond to the developmentally regulated onset of “functional” vesicle cycling and the uptake of the FM1-43 styryl dye (Betz and Bewick, 1992). Thus, there is a developmental delay in FM uptake at early stages of development and now a delay in VGLUT1 expression as well (Fig. 6). We therefore asked whether the delays were coordinated, and whether the developmental time course of VGLUT1 expression might be synchronized with the maturation of vesicle cycling. It seemed sensible that as vesicle cycling matures, the functional capacity of transmitter loading needs to be upregulated accordingly to prevent the release of synaptic vesicles with insignificant transmitter content.

To examine whether such features of presynaptic maturation might be coordinated, it was necessary to concurrently compare the functional capacity of presynaptic terminals with their structural protein content. We made use of AM1-43, a fixable version of FM1-43, to label the functional presynaptic terminals, and followed up with an immunostaining procedure to check the level of desired presynaptic proteins (Renger et al., 2001). It has been shown that the amount FM dye taken up during stimulation can provide an index of the functional capacity of presynaptic terminals. The ability to visualize a fixable version together with immunolabeled protein allowed us to simultaneously measure both VGLUT1 expression levels across development and the developmental onset of functional vesicle cycling. As shown by imaging VGLUT1, AM1-43, and the synaptic vesicle protein synapsin I (Syn I) at the same synaptic terminals (Fig. 7A), AM1-43 uptake begins appearing at a similar stage of development as VGLUT1 expression. Figure 7B-D quantifies these developmental changes in a population of synapses, again using IOD as a measure of protein expression. Despite a relatively constant level of synapsin I staining across development (Fig. 7B), VGLUT1 expression (Fig. 7C) appeared to increase in tandem with the onset of AM1-43 labeled functional vesicle cycling (Fig. 7D). In general, the ratio of VGLUT1/Syn I increases from 8 to 15 DIV (p < 0.001; unpaired t test) and from 10 to 15 DIV (p < 0.001) in parallel with the ratio of AM1-43/Syn I (p < 0.01; P < 0.05) measured at the same synapses (Fig. 7E). We thus conclude that the maturation of transporter expression and delivery is jointly coordinated with functional vesicle cycling, a coupling that could be designed to help guarantee the efficient loading of transmitter as vesicle cycling becomes more active.

Activity-dependent plasticity of glutamate release and VGLUT1 expression in mature circuits

Exciting work by others (Turrigiano et al., 1998; Burrone et al., 2002; Thiagarajan et al., 2002) has demonstrated recently that excitatory synapses undergo activity-dependent plasticity in quantal size in response to global changes in network activity. Mechanisms have been proposed to begin to explain this change in quantal size, including changes in postsynaptic receptor number (O'Brien et al., 1998; Watt et al., 2000; Wierenga et al., 2005). Given that molecules responsible for specifying the quantal size postsynaptically, such as AMPA receptors, appear to be critically regulated during activity-dependent synaptic scaling, we decided to examine whether VGLUT1 expression might provide a complementary presynaptic mechanism by which quantal size can be coordinately regulated via endogenous, homeostatic control.

We therefore measured VGLUT1 expression levels per synaptic terminal by again taking the IOD ratio of VGLUT1 to synapsin I at each immunolabeled synapse as before but now comparing between untreated control cultures versus cultures that were preincubated for 48 h in 20 μm bicuculline (BIC) for the purpose of chronically increasing activity levels by blocking GABA_A receptor-mediated inhibitory currents (Turrigiano et al., 1998) (Fig. 8A). We observed that chronic increases in activity levels (BIC) led to a significant reduction in synaptic VGLUT1 expression relative to control values (0.75 ± 0.06 control vs 0.95 ± 0.03 BIC; p = 0.02; unpaired t test; n = 8 images, 4482 synapses control; n = 7 images, 4464 synapses BIC) (Fig. 8B). We then attempted the exact opposite condition, chronically blocking AMPA and NMDA-type excitatory receptors by applying either 5 μm NBQX or 50 μm AP-5 to effectively decrease activity levels through each one of these pathways (Thiagarajan et al., 2002; Liu, 2004) and measuring the resulting expression levels of VGLUT1. Here, although we observed a robust increase in VGLUT1 in response to chronic NMDA receptor blockade (Fig. 8A, B), a less prominent increase and a surprising variability were disclosed in response to AMPA receptor blockade (supplementary data, available at www.jneurosci.org as supplemental material). This dichotomy was confirmed by replicating the experiments for NBQX and AP-5 many times. NMDA receptor blockade, however, resulted in increased VGLUT1 expression across every instance tested and, in the experiment presented in Figure 8, A and B, increased VGLUT1/Syn I levels by 35 ± 4% relative to controls (p < 0.001; unpaired t test; n = 10 images, 6120 synapses control; n = 8 images, 6532 synapses AP-5).

Thus, VGLUT1 expression appears to be capable of bidirectional endogenous regulation in response to changes in activity patterns, which could provide a mechanism for presynaptic terminals to coordinate functionally with activity-dependent changes to their postsynaptic partners. To verify that such changes to expression levels indeed had functional implications, we next measured whether glutamate release was also changing as a result of the activity-dependent VGLUT1 regulation. To examine activity-driven changes in glutamate release, we repeated our competitive antagonism protocol applied previously (Fig. 5), again in mature hippocampal cultures, except that in this case, one set of cultures was preincubated for 48 h in either 20 μm BIC or 50 μm AP-5. In these experiments, the cleft glutamate concentration at nontreated synapses appeared to be similar to that observed previously (Fig. 5D), with an ∼50% reduction in evoked postsynaptic current in response to 600 μm γ-DGG (Fig. 8C, gray symbols, D, middle traces). However, bicuculline-treated synapses were significantly more sensitive to the antagonist (Fig. 8C, white symbols, D, left traces) and led to 47% more attenuation than control (76 ± 4 vs 51 ± 3%; p = 0.03; unpaired t test; n = 4 pairs BIC, 4 pairs control) (Fig. 8E), indicating that the bicuculline-treated terminals released less glutamate, thus rendering their transmission more susceptible to the effects of the antagonist. To examine whether a comparable change in glutamate release was also observed in the opposite direction following a treatment that upregulates VGLUT1, we repeated this comparison with chronically treated AP-5 cultures. Here, we observed that the antagonist was significantly less effective on AP-5-treated transmission (Fig. 8C, black symbols, D, right traces), whereby the γ-DGG block was reduced by 65% in AP-5 synapses compared with control transmission (20 ± 8% AP-5 vs 61 ± 5% control; p < 0.01; unpaired t test; n = 6 pairs AP-5, 6 pairs control) (Fig. 8E).

Because the activity-dependent changes that altered vesicular glutamate release (Fig. 8C-E) also impacted VGLUT1 expression (Fig. 8A, B), and because our previous data demonstrate a causal relationship between VGLUT1 expression and glutamate release, we interpret our observed changes in glutamate release under BIC and AP-5 treatments to occur primarily as a direct consequence of the altered VGLUT1 expression. However, if synaptic terminals can at times undergo multivesicular release from single release sites, then the cleft glutamate concentration could also be altered in part by changes to the number of vesicles released from single terminals. This is possible given evidence that the probability of release, and thus the prevalence of MVR-mediated changes to cleft glutamate concentration, can be modified by perturbations to network activity levels (Murthy et al., 2001). As an additional control to further verify that these changes correspond to changes to the loading of single vesicles, we also measured the cleft glutamate concentration during spontaneous, miniature EPSCs in the bicuculline-treated transmission (Fig. 8F). Here, we observed that median γ-DGG attenuation was significantly increased relative to controls (Fig. 8F) (25 ± 4% control vs 39 ± 2% BIC; p = 0.02; unpaired t test), although the attenuations observed in the mEPSCs were less than those in the EPSCs, presumably because of the detection problems associated with spontaneous transmission when it is pharmacologically reduced (Zhou et al., 2000), which likely serves to underestimate the attenuation in the control condition and, to a greater degree, the smaller amplitude bicuculline-treated condition. Even with this detection problem, significant changes in the amount of glutamate released from single vesicles were observed between the two treatments.

To verify that activity-induced changes to VGLUT1 expression were also accompanied by changes to quantal synaptic efficacy, we quantified the amplitude of spontaneous mEPSCs. Here, we observed that in bicuculline-treated transmission, mEPSC amplitude was significantly reduced compared with control transmission (Fig. 8G, top trace, H) (n = 6 control; n = 4 BIC; p = 0.04; unpaired t test), whereas in AP-5-treated transmission, mEPSC amplitude was increased by 52% from control transmission (Fig. 8G, bottom trace, H) (n = 7 control; n = 6 AP-5; p = 0.03; unpaired t test). Thus, in both cases, activity-induced changes to VGLUT1 expression are accompanied by changes to quantal synaptic efficacy.

As such, chronic changes to either levels of activity or calcium flux alter vesicular transporter expression in a manner that is accompanied by relevant changes to physiological transmission. Endogenous control over VGLUT1 provides one mechanism by which an observed presynaptic scaling of glutamate release could be coordinated with a concomitant scaling of postsynaptic sensitivity (Watt et al., 2000).

Vesicle loading depends on the number of available VGLUT1 transporters

It is imaginable that vesicles set in a low concentration of glutamate might simply continue filling until they have reached a set point level of saturation (Sulzer and Pothos, 2000), such that adding transporters could influence the speed but not the final concentration of glutamate uptake. The work presented here indicates that VGLUT1 vesicles do not fill to an invariant set point, and that reducing the pool of participating transporters (Fig. 1) reduces the equilibrium value of uptake, although it still cannot be resolved whether this is accomplished by changing the intravesicular concentration of transmitter or, instead, the structural capacity of the vesicle (Bruns et al., 2000; Colliver et al., 2000; Sulzer and Edwards, 2000; Karunanithi et al., 2002).

VGLUT1 overexpression enhances excitatory synaptic transmission via an increase in glutamate released per vesicle

We then found that it was possible to traffic additional functional VGLUT1 protein to synaptic terminals (Fig. 2), where we consequently observed larger evoked and spontaneous excitatory currents (Figs. 3, 4) as a result of an increased cleft glutamate concentration (Fig. 5). We provide direct verification that vesicular transport is able to exploit receptor nonsaturation at the level of single vesicles via a shift in the cleft glutamate concentration. An additional finding of some surprise was that VGLUT1 overexpression resulted in a decrease in evoked failure rate (Fig. 3E). This could suggest that in addition to enhancing the amount of glutamate deposited per release event, VGLUT1 expression might also be able to influence the likelihood that such release events will occur. Interestingly, another recent study reports that loss of VGLUT1 expression results in a substantially depleted pool of synaptic vesicles at excitatory terminals (Fremeau et al., 2004). Such a relationship between VGLUT1 expression and likelihood of release could depend on either the presence of the transporter itself or on the extent of filling that it facilitates.

It is also important to consider whether VGLUT1 expression could contribute to synaptic efficacy by influencing not only the amount of filling but also the rate (Sulzer and Pothos, 2000). This is especially plausible given two recent studies into the nature of vesicle cycling (Aravanis et al., 2003; Gandhi and Stevens, 2003), which confirmed a “kiss-and-run” type of release that involves the brief fusion and rapid reuse of synaptic vesicles. Such efficiency in cycling would place high demands on vesicle filling to prevent the “firing of blanks” via the exocytosis of “empty” vesicles. Although the precise speed of vesicle filling in the endogenous context can only be estimated, loading in vesicles isolated from the brain (Fig. 1) (Maycox et al., 1988; Wolosker et al., 1996) as well as loading in BON and PC12 cell lines (Bellocchio et al., 2000; Takamori et al., 2000) have been shown to involve a relatively slow process (several minutes), compared with the rapid process of vesicle reuse (∼20 s) (Aravanis et al., 2003). A rapid turnover of vesicles could thus result in a depleted synaptic response, and an increasing degree of transporter expression may be important for ensuring a match between the speed of transmitter loading and the functional capacity of the release machinery. Indeed, VGLUT1 is upregulated in close register with the increase of vesicle cycling that occurs during synaptic maturation (Fig. 7) and downregulated during activity-dependent scaling (Fig. 8) in parallel with a reduction in probability of vesicle release (Murthy et al., 2001).

Vesicular transporter expression is endogenously regulated

We found that transporter levels are modified endogenously across development (Fig. 6) and by levels of activity (Fig. 8). If vesicular transporters do contribute meaningfully to synaptic transmission, it would not be surprising to find them amid a molecular ensemble of similar versatility to that in the postsynaptic density (PSD) (Sheng and Kim, 2002). Molecular studies at inhibitory synapses, for example, have recently uncovered a direct association between the vesicular GABA transporter, VGAT, and the synthetic enzyme GAD65 (Jin et al., 2003), suggesting that transmitter synthesis may be functionally coupled to vesicle loading. The fact that VGLUT1 expression is coordinated with the onset of functional vesicle cycling (Fig. 7) could, in turn, suggest a conjunctive signal that couples fusion with adequate filling to ensure that increasingly active vesicles are able to release meaningful quantities of transmitter, perhaps in coordination with a growing pool of receptors postsynaptically.

Similarly, a negative-feedback relationship between certain forms of synaptic activity and VGLUT1 expression, which adjusts glutamate release in response to either circuit excitability or calcium influx (Fig. 8), could provide a presynaptic means for maintaining a stable degree of activity in the face of changing input levels (Turrigiano and Nelson, 2000) in a manner congruent with other recently uncovered postsynaptic mechanisms (O'Brien et al., 1998; Watt et al., 2000; Thiagarajan et al., 2002). Chronic blockade or enhancement of synaptic activity has been shown previously to lead to compensatory shifts in the size of the readily releasable pool and the number of presynaptic active zones (Murthy et al., 2001). This observation has supported a “size principle” (Harris and Stevens, 1989; Pierce and Lewin, 1994; Schikorski and Stevens, 1997) in which functional changes in morphology can correspond to changes in the size of the presynaptic terminal or PSD. It is possible that an activity-driven scaling of presynaptic terminal size (Murthy et al., 2001) is similarly accompanied by alterations in transporter expression per vesicle (Fig. 8) to provide a presynaptic shift in quantal efficacy that coincides with the known activity-driven changes in probability of release. In contrast, although blocking inhibitory receptors was effective at scaling VGLUT1, we were surprised to find that blocking NMDA receptors triggered a more profound regulation than that observed while blocking AMPA receptors. This suggests that VGLUT1 regulation, at least in hippocampal synapses, may be more sensitive to calcium flux than postsynaptic membrane depolarization, although other regimes of regulation may be more prominent in cortical synapses (Watt et al., 2000; Pratt et al., 2003) depending on the layer of cortex and which vesicular transporter is expressed.

Most likely, additional work will elucidate a range of distinct modes by which vesicular transport contributes to the dynamics of synaptic efficacy as well as distinct isoforms by which those contributions are made. Independent regulation of VGLUT1 and its recently identified isoforms VGLUT2 (Fremeau et al., 2001; Takamori et al., 2001; Varoqui et al., 2002) and VGLUT3 (Fremeau et al., 2002; Schafer et al., 2002) could offer, in particular, exciting clinical possibilities. The first mutations to the VGLUT1 ortholog eat-4 in Xenopus have revealed the importance of vesicular transport in both short-term (Rankin and Wicks, 1995) and long-term (Rose et al., 2002) memory. Future studies into the regulation of transmitter loading and release should thus discern new principles by which adaptations on both sides of the synapse are coordinated to ensure a cohesive robustness of memory.