Ribbon synapses of sensory neurons are able to sustain high rates of exocytosis in response to maintained depolarization, but it is not known how this is achieved. Using the capacitance technique, we have found that Ca2+ regulates the supply of releasable vesicles at the ribbon synapse of depolarizing bipolar cells from the retina of goldfish. Ca2+ had two actions that could be differentiated by introduction of the Ca2+chelator EGTA; one action stimulated refilling of the rapidly releasable pool of vesicles from a reserve pool, and a second action stimulated recruitment of vesicles to the reserve pool. The capacity of the reserve pool was ∼3500 vesicles, which is similar to the number that can attach to the ribbons. These results suggest that continuous exocytosis at ribbon synapses is maintained by the Ca2+-dependent translocation of vesicles from the cytoplasm, through the ribbon, to release sites on the plasma membrane.
- synaptic depression
- depolarizing bipolar cell
- ribbon synapse
The processes that make synaptic vesicles available for exocytosis play an important role in determining the efficiency of synaptic transmission during ongoing activity. For instance, many synapses exhibit short-term depression after a period of activity, and this is at least partly because of a decrease in the number of rapidly releasable vesicles docked at the plasma membrane (Del Castillo and Katz, 1954; Betz, 1970; Zucker, 1989; Rosenmund and Stevens, 1996; Dittman and Regehr, 1998). The question of how vesicles become available for exocytosis is of particular importance at ribbon synapses, which can support continuous high rates of exocytosis in response to maintained depolarization (Dowling and Ripps, 1973; Parsons et al., 1994; Lagnado et al., 1996; Rieke and Schwartz, 1996). Ribbon synapses are found in sensory neurons that generate graded voltage signals rather than action potentials, including photoreceptors and bipolar cells in the retina (Gray and Pease, 1971; Rao-Mirotznik et al., 1995; von Gersdorff et al., 1996) and auditory and vestibular hair cells in the ear (Jacobs and Hudspeth, 1990; Lenzi et al., 1999). Electron micrographs of these neurons show that the regions in which synaptic vesicles dock to the plasma membrane are associated with an electron-dense structure, called a ribbon or synaptic body, to which vesicles attach by short filaments. This arrangement suggests that vesicles attached to the ribbon form a reserve pool that supplies the active zone on the plasma membrane (Gray and Pease, 1971), although it has been difficult to test this idea experimentally (Burns and Augustine, 1995; Lenzi et al., 1999). Recently, our knowledge of the structure of ribbon synapses has been complemented by functional studies of exocytosis using capacitance and optical techniques. In particular, we have a relatively large amount of information about vesicle cycling at the ribbon synapse of depolarizing bipolar cells from the goldfish retina (Tachibana et al., 1993; Heidelberger et al., 1994; von Gersdorff and Matthews, 1994; Lagnado et al., 1996). An important feature of this synapse is that the size of the rapidly releasable pool of vesicles (RRP) can be directly measured using the capacitance technique (Mennerick and Matthews, 1996; Neves and Lagnado, 1999).
We have investigated the refilling of the RRP in goldfish bipolar cells during different patterns of stimulation and after introduction of the Ca2+ chelator EGTA. Our results indicate that Ca2+ entering through Ca2+channels has two effects that stimulate the supply of vesicles to the RRP. Over the short-term, Ca2+ stimulated refilling of the RRP from a reserve pool, and over the long-term, Ca2+ stimulated the recruitment of vesicles to the reserve pool. The ability of residual Ca2+ to stimulate refilling of the RRP for periods of 5–10 sec indicates that this action occurs at free Ca2+ concentrations in the range of 1 μm or below. The capacity of the reserve pool was estimated to be ∼3500 vesicles, which is similar to the total number attached to the ribbons, calculated from serial EM reconstructions (von Gersdorff et al., 1996). These results therefore provide functional evidence for the idea that the ribbon acts to supply vesicles to the active zone on the plasma membrane. Ca2+-dependent trafficking of vesicles through the ribbon to the RRP may act to maintain high rates of exocytosis that occur during maintained depolarizing responses (Lagnado et al., 1996;Rouze and Schwartz, 1998; Neves and Lagnado, 1999).
MATERIALS AND METHODS
Isolated synaptic terminals of depolarizing bipolar cells from the goldfish retina were obtained by enzymatic dissociation (Burrone and Lagnado, 1997). Terminals were voltage-clamped using the perforated patch technique, and capacitance measurements were made using the piecewise linear method (Neher and Marty, 1982). Our implementation of this method is described by Neves and Lagnado (1999). Briefly, a sinusoidal command voltage (50 mV peak-to-peak, 2 kHz) was superimposed onto a holding potential of −70 mV. The output of the patch-clamp amplifier was analyzed by a two-phase lock-in amplifier, the phase of which was set using the “capacitance dither” technique, allowing independent measurement of changes in membrane capacitance and conductance (Gillis, 1995). The capacitance signal was calibrated by dithering the capacitance by 100 fF at the beginning of the recording episode. The change in capacitance elicited by the stimulus (ΔC m) was measured by subtracting the capacitance signal before the stimulus from that measured after. The “before” measurement was averaged over a period of 50–300 msec, ending 20 msec before depolarization. The “after” measurement was averaged over a period of 50–300 msec, beginning 20 msec after repolarization.
The capacity of the RRP was measured after various stimulation patterns by measuring the capacitance increase elicited by a 20 msec depolarization to 0 mV. At this potential, the Ca2+current is maximally activated. A 20 msec stimulus was sufficient to ensure that the whole of the RRP was exhausted, because it is normally released with a time constant of 2–4 msec (see Results). This duration stimulus also provided a safety margin that ensured depletion of the RRP when the Ca2+ current was partially inactivated, as sometimes occurred immediately after the introduction of a large Ca2+ load by prolonged depolarization.
The standard Ringer’s solution contained (in mm): 120 NaCl, 2.5 CaCl2, 2.5 KCl, 1 MgCl2, 10 glucose, and 10 HEPES, pH 7.3. The solution in the patch pipette contained (in mm): 110 caesium gluconate, 4 MgCl2, 3 Na2ATP, 1 Na2GTP, 10 TEACl, 0.4 BAPTA, and 20 HEPES, with 250 μg/ml nystatin (260 mOsm/l, pH 7.2).
In some experiments terminals were loaded with the Ca2+ chelator EGTA by incubation in 0.1 or 0.2 mm EGTA-AM (0.1% DMSO; Molecular Probes, Eugene, OR) for 15–20 min.
All measurements are given as mean ± SEM.
Measuring the rapidly releasable pool of vesicles
In depolarizing bipolar cells, there is a clear kinetic distinction between the first and second phases of exocytosis elicited by strongly activating the Ca2+ current (Mennerick and Matthews, 1996; Neves and Lagnado, 1999). Here, we define the pool of vesicles involved in the first phase as the RRP. The approach we used to measure the size of the RRP is derived from the behavior shown in Figure 1. Figure 1 Ashows the capacitance response to two pairs of depolarizing stimuli to −10 mV, a potential at which the Ca2+ current is maximally activated. The interval between the stimuli was 100 msec. The duration of the first stimulus was varied between 0.5 and 20 msec so as to release a varying proportion of the RRP. These responses are plotted as the filled circles in Figure 1 B. In this example, the maximum amplitude of the first phase of exocytosis was 45 fF, equivalent to ∼1700 vesicles, assuming the capacitance of a single vesicle is 26 aF (Neves and Lagnado, 1999). The RRP was exhausted with a time constant of 4 msec, so a 20 msec depolarization was more than sufficient to release all of it. Here, we term a 20 msec depolarization to a potential that maximally activates the Ca2+ current (−10 or 0 mV) the “emptying” stimulus, and we used it to probe the size of the RRP after various patterns of stimulation.
Evidence that this approach does indeed measure changes in the size of the RRP is shown in Figure 1 B in which the open circles plot the response to the emptying stimulus applied after variable amounts of the RRP were released. The decline in the measured size of the RRP mirrored the amount of rapid exocytosis elicited by the first stimulus. The dashed line shows the time course over which the RRP was expected to decline, assuming that the RRP was equivalent to 45 fF at the start of each trial and that there was negligible refilling in the 100 msec period between stimuli. The decline in the measured size of the RRP mirrored the amount of rapid exocytosis elicited by the first stimulus in similar experiments on eight other cells.
Refilling of the RRP after a brief stimulus
To assess how the RRP was refilled after a brief stimulus, two emptying stimuli were applied with varying delays, as shown in Figure2 A. After a delay of 100 msec, there was only a small response to the second stimulus (Fig.2 A, top left), averaging 6% of the first. The Ca2+ currents flowing during each stimulus were very similar (Fig. 2 A, inset), so depression of the second response was not a result of inactivation of the Ca2+ current. As the interval between stimuli was increased, the second response recovered, reflecting refilling of the RRP (Fig. 2 A, middle,right). Results from this type of experiment are collected in Figure 2 B, which plots the refilling of the RRP as a function of the interval between stimuli. There were two phases of refilling: 30% of the RRP was refilled with a time constant of 0.64 sec, and the remainder with a time constant of 31 sec.
Behavior qualitatively similar to that shown in Figure 2 has also been observed at synapses from the CNS of mammals in which there are fast and slow phases of recovery from synaptic depression. The fast phase can be inhibited by the Ca2+ chelator EGTA, indicating that it reflects the transient acceleration of refilling by residual Ca2+ (Dittman and Regehr, 1998; Stevens and Wesseling, 1998; Wang and Kaczmarek, 1998). Below, we present evidence that residual Ca2+ also stimulates refilling of the RRP in the synaptic terminal of bipolar cells. However, at this ribbon synapse, Ca2+ appears to stimulate refilling of the RRP by two actions, one of which is more sensitive to EGTA.
Refilling of the RRP was stimulated by calcium
When the duration of the depolarizing stimulus was increased, refilling of the RRP was accelerated above the basal rate for a longer period. The protocol used to measure refilling after a longer stimulus is shown in Figure 3 A,Test. First, the emptying stimulus was applied to measure the initial size of the RRP. Next, a 500 msec depolarization to 0 mV was delivered after a delay of 200 msec. Finally, the emptying stimulus was applied again to measure the RRP after a variable delay. In the example shown in Figure 3 A, 91% of the RRP was refilled 10 sec after a 500 msec depolarization (Test), but only ∼44% was refilled 10 sec after the emptying stimulus delivered alone (Control). The time course of refilling after depolarizations lasting 20 (filled circles) and 500 (open circles) msec is compared in Figure3 C. The basal rate constant of refilling, measured from the slow phase after a brief stimulus, was 0.03/sec (Fig.2 B). Figure 3 C shows that after a 500 msec depolarization, refilling was accelerated above the basal rate for 5–10 sec until the RRP was almost completely replenished. In comparison, it took ∼60 sec to refill 90% of the RRP after a 20 msec depolarization (Fig. 2 B).
Although the net effect of a longer stimulus was to potentiate refilling of the RRP, at short delays the RRP was significantly smaller (Fig. 3 C). For instance, in the example shown in Figure3 B (same terminal as Fig. 3 A), there did not appear to be any refilling of the RRP 1 sec after a 500 msec depolarization, whereas ∼19% was refilled 1 sec after a 20 msec depolarization. The delayed growth in the size of the RRP after a 500 msec stimulus is probably because residual Ca2+continued to stimulate exocytosis after closure of Ca2+ channels. Capacitance and optical measurements in bipolar cell terminals have demonstrated that residual Ca2+ can continue to drive exocytosis for periods of 1 sec or more, at rates of 400–800 vesicles/sec (Neves and Lagnado, 1999). We would expect that vesicles accumulate in the RRP only after their removal by exocytosis has ceased, which can take ∼1 sec after a 500 msec depolarization to 0 mV. Consistent with the results in Figure3 C, exocytosis does not continue after a 20 msec stimulus that introduces a smaller amount of Ca2+ (Neves and Lagnado, 1999).
An idea of the changes in free cytoplasmic [Ca2+] that occurred after stimulation could be obtained using Ca2+-activated conductances in the plasma membrane that are activated by residual Ca2+ and generate a slowly decaying tail current (Okada et al., 1995). Measurements of cytoplasmic [Ca2+] have demonstrated that the amplitude of this current is correlated with the spatially averaged [Ca2+], and it declines with a time constant that is similar to the fall in [Ca2+] (Neves and Lagnado, 1999). Figure4 A shows that the Ca2+-activated tail current was much larger and longer after a 500 msec depolarization than after a 20 msec depolarization.
We tested whether residual Ca2+ might be the stimulus that accelerated refilling after a longer stimulus by loading terminals with the Ca2+ chelator EGTA so as to reduce the rise in cytoplasmic [Ca2+]. Terminals were loaded by incubation in the AM ester (see Materials and Methods). Buffering the rise in [Ca2+] greatly reduced the Ca2+-activated tail current (Okada et al., 1995). This was most clearly observed after a 500 msec depolarization (Fig.4 A,C), but it was also apparent after a 20 msec depolarization that caused a small activation of the tail current (Fig. 4 B). The tail current averaged 50 ± 6 pA (n = 20) immediately after a 500 msec depolarization in a control terminal but only 10 ± 2 pA (n = 12) after loading with EGTA. Experiments in which we have altered the internal concentration of Ca2+ buffers through a patch pipette indicate that at least 2 mm EGTA is required to cause a similar inhibition of the tail current (J. Burrone and L. Lagnado, unpublished observations; Okada et al., 1995).
EGTA shortened the accelerated phase of refilling that followed a long stimulus. Examples of these measurements are shown in Figure5 A, and averaged results for delays of 10 and 5 sec are plotted as the open squaresin Figure 3 C to allow a comparison with cells that did not contain any exogenous Ca2+ buffers. After a 500 msec depolarization, the acceleration of refilling above its basal rate normally continued for ∼10 sec (open circles), but buffering the rise in cytoplasmic [Ca2+] blocked refilling at delays longer than 5 sec (open squares). EGTA did not, however, block the initial accelerated phase of refilling. Figure 3 C shows that 43 ± 4% of the RRP was refilled over the first 5 sec, although there was no further refilling over the next 5 sec. The refilling of 43% of the RRP in the first 5 sec indicates an average rate constant of 0.09/sec over this period. In comparison, the basal rate constant of refilling was 0.03/sec in control terminals, as measured from the second phase of the double-exponential fit in Figure 2 B. The initial rate constant of refilling in EGTA-loaded terminals was therefore at least three times that measured in control terminals at resting levels of Ca2+.
The effects of EGTA on the first and second phases of exocytosis
We compared the effects of EGTA on refilling of the RRP with its effects on the different kinetic components of exocytosis. EGTA had a small effect on exocytosis from the RRP. The first response to a 20 msec depolarization averaged 55.5 ± 4.7 fF (n = 58) in control terminals and 40.6 ± 3.6 fF (n = 38) in terminals loaded with EGTA (Fig.5 B). The somewhat smaller response in the presence of EGTA might reflect a slowing of exocytosis from the RRP and/or a decrease in the initial size of the pool. The latter possibility seems more likely because the kinetics of fast exocytosis are little affected by EGTA (Mennerick and Matthews, 1996), whereas EGTA inhibited refilling of the RRP over long time scales (see below). Our estimate of the initial size of the RRP in EGTA-loaded terminals is therefore likely to be a lower limit, especially in view of the fact that isolated bipolar cell terminals can spontaneously depolarize, triggering exocytosis (Burrone and Lagnado, 1997). In an effort to ensure that estimates of the size of the RRP were made before exocytosis had been stimulated to any appreciable extent, cells were loaded with EGTA in the absence of external Ca2+, and Ca2+ was replaced only after Ca2+ channels were held closed by voltage-clamping the terminal.
EGTA had a stronger inhibitory effect on the second, slower, phase of exocytosis. Figure 5 A compares responses to a 20 and 500 msec stimulus, first under control conditions and then in a terminal loaded with EGTA. Considering only the first stimulus episode in a given terminal, a 500 msec depolarization normally released 1.56 ± 0.24 (n = 22) times as many vesicles as the preceding emptying stimulus. However, after loading with EGTA, this ratio was reduced to 0.73 ± 0.17 (n = 10) (Fig.5 C). We propose that these two actions of EGTA, inhibition of the second phase of exocytosis and reduced refilling after a long stimulus, reflects the involvement of the same set of vesicles, which undergo Ca2+-dependent transfer to the RRP (see Discussion).
Recruitment to a reserve pool was stimulated by calcium
To investigate refilling over longer time scales, the RRP was repeatedly exhausted by delivering trains of emptying stimuli at a frequency of 0.1 Hz for durations of 3 min or more. The second response in a train normally released 40–50% as many vesicles as the first (Fig.6 A,B), as would be expected from the results obtained with a double-pulse protocol (Fig. 2 B). The amplitude of the capacitance response elicited by the emptying stimulus then remained constant, indicating that the same fraction of the RRP was refilled after each emptying stimulus and that the average rate of refilling was fixed (Fig. 6 A). The cumulative response during these stimulus trains, relative to the initial size of the RRP, is plotted as the filled circles in Figure 6 C. The first response averaged 50 ± 15 fF (n = 18). The RRP could be refilled at least ten times over, equivalent to the transfer of at least 20,000 vesicles, without any apparent depletion of the reservoir from which it was supplied.
When terminals were loaded with EGTA, the rate of refilling was not affected at first, because the second response was still ∼50% of the first (Fig. 6 A,B). Thus, EGTA did not directly affect the transfer of vesicles to the RRP after a brief stimulus. Over the long term, however, the rate of refilling gradually declined during a train, until no more vesicles could be transferred to the RRP and the capacitance response was abolished (Fig.6 A,B). The cumulative response during these stimulus trains is plotted as the open circles in Figure 6 C. In the presence of EGTA, the maximum number of vesicles transferred to the RRP averaged 2.7 times its initial size, or ∼3500 ± 400 vesicles (the amplitude of the first capacitance response in this set of EGTA-loaded terminals was 33.4 ± 3.5 fF; n = 16). In other words, only ∼3500 vesicles could be rapidly transferred from a reserve pool to the RRP in a terminal loaded with EGTA. A simple explanation for this result is that refilling of the RRP occurs from a reserve pool, which has an initial capacity of ∼3500 vesicles, and EGTA blocks the transfer of vesicles to this pool, causing it to gradually empty (see Discussion).
If EGTA inhibits refilling of the reserve pool by buffering Ca2+ transients, then it should be possible to overcome this effect by introducing a larger Ca2+load into the terminal. A demonstration that this is the case is shown in Figure 7. In this experiment, both the RRP and reserve pool were completely exhausted by applying a train of emptying stimuli, as in Figure 6. After a 150 sec period of rest, application of the emptying stimulus elicited a response of only 2 fF, confirming that the RRP was refilled very slowly at resting levels of Ca2+. A larger number of vesicles were released by a 500 msec depolarization, and refilling of the RRP was stimulated, as demonstrated by the much larger response to the emptying stimulus applied after a delay of 10 sec. Similar behavior was observed in four other terminals. These results provide further support for the idea that Ca2+ stimulates the transfer of vesicles to the reserve pool and then to the RRP.
Two basic observations indicate that Ca2+stimulates refilling of the RRP in the synaptic terminal of depolarizing bipolar cells. First, accelerated refilling of the RRP was prolonged by introducing a larger Ca2+ load into the terminal (Fig. 3 C), and this effect could be inhibited by buffering the rise in cytoplasmic [Ca2+] with EGTA (Fig. 5). Second, EGTA limited the number of vesicles that could be released by repetitive brief stimuli (Fig. 6), and this effect could be counteracted by introducing a larger Ca2+ load into the terminal (Fig. 7).
Vesicle pools in the bipolar cell terminal
After brief depolarizations, EGTA blocked the Ca2+-dependent transfer of all but a small fraction of available vesicles, although it had no immediate effect on refilling of the RRP (Fig. 6). The differential action of EGTA allows us to provide a functional definition for two pools of vesicles involved in refilling of the RRP: a reserve pool of ∼3500 vesicles that refills the RRP over the short term, and a much larger reservoir that maintains the supply of vesicles over the long term. Figure8 shows a simple model of vesicle transfer in the bipolar cell terminal that seeks to integrate these results with previous anatomical and physiological work. Features of this model are discussed below.
It is possible to define at least three pools of vesicles in the bipolar cell terminal. When the Ca2+ current is activated maximally, the most rapid phase of exocytosis involves the release of 1200–1800 vesicles within ∼10 msec (Heidelberger et al., 1994; Mennerick and Matthews, 1996; Neves and Lagnado, 1999). A similar number of vesicles are docked to the plasma membrane beneath the ribbon, suggesting that these correspond to the RRP (von Gersdorff et al., 1996). The second phase of exocytosis involves the release of ∼4400 vesicles over 1 sec (Neves and Lagnado, 1999). Two pieces of evidence indicate that these vesicles constitute the reserve pool that is used to refill the RRP after closure of Ca2+channels. First, the reserve pool contained a similar number of vesicles, ∼3500 (Fig. 6). Second, EGTA inhibited the second phase of exocytosis and also reduced refilling of the RRP stimulated by residual Ca2+ (Fig. 5). The anatomy of the bipolar cell terminal immediately suggests that the reserve pool of vesicles is attached to the ribbon behind the plasma membrane, and this idea is supported by EM measurements indicating that there are ∼4800 vesicles attached to the ribbons (von Gersdorff et al., 1996). The third component of exocytosis observed during maintained depolarization occurs at a rate of ∼1000 vesicles/sec but does not fatigue for periods of many tens of seconds (Lagnado et al., 1996; Rouze and Schwartz, 1998; Neves and Lagnado, 1999). Measurements made with FM1–43 indicate that this continuous process of exocytosis is supported by a reservoir of ∼750,000 vesicles that is efficiently replenished by endocytosis (Lagnado et al., 1996). EM measurements indicate that there are a similar number of vesicles in the cytoplasm, free of the ribbons (von Gersdorff et al., 1996).
In the model shown in Figure 8, the rate constantsK 1 and K 2 are both increased by Ca2+, but during brief periods of Ca2+ influx, EGTA is more effective at preventing the increase in K 2. When the Ca2+ current is activated maximally, the second phase of exocytosis occurs at a rate of ∼4500 vesicles/sec (Neves and Lagnado, 1999), so the maximum rate constant for the transfer of vesicles from the reserve pool to the RRP,K 1, is ∼1.3/sec (assuming the capacity of the reserve pool is 3500 vesicles). As might be expected, the rate of refilling that we measured after closure of Ca2+channels was somewhat slower, occurring with a rate constant of ∼0.5/sec in the few hundred milliseconds after a 20 msec depolarization (Fig. 2 B). It appears that Ca2+ could modulate the rate of refilling over a range of ∼40-fold, because the basal value ofK 1 at resting levels of Ca2+was ∼0.03/sec (Fig. 2 B). After the first two pools of vesicles have been exhausted, continuous exocytosis occurs at a rate of ∼1000 vesicles/sec (Neves and Lagnado, 1999), indicating that the maximum rate constant for the transfer of vesicles from the reservoir to the reserve pool, K 2, is ∼1.3 × 10−3/sec.
Two actions of Ca2+
The differential effect of EGTA on vesicle transfer to the RRP and reserve pool (Figs. 6, 7) might be explained if the Ca2+-binding molecule regulatingK 2 was further from Ca2+channels than the Ca2+-binding molecule regulatingK 1. During brief periods of Ca2+ influx, the molecule regulatingK 2 would be subject to smaller rises in [Ca2+] that are more effectively buffered by EGTA (Neher, 1998).
The EGTA concentrations we introduced into the terminals were sufficient to strongly buffer Ca2+ at the Ca2+-dependent conductances in the plasma membrane (Fig. 4). Two factors might explain why they were not sufficient to inhibit refilling of the RRP after a brief stimulus (Fig. 6). First, the Ca2+-sensor for refilling may be closer to Ca2+ channels than the Ca2+-dependent conductances, so that it senses a larger rise in [Ca2+] while Ca2+ channels are open. Second, the Ca2+-sensor for refilling may have a higher affinity for Ca2+ than the Ca2+-dependent conductances. The latter idea is supported by the observation that residual Ca2+ stimulated refilling for ∼10 sec after a 500 msec depolarization (Fig. 3 C), whereas the Ca2+-activated tail current had decayed completely after 2 sec (Fig. 4). Neves and Lagnado (1999) used furaptra, a Ca2+-indicator of relatively low affinity, to measure the cytoplasmic [Ca2+] in the bipolar cell terminal. Strongly activating the Ca2+ current for 1 sec caused the spatially averaged [Ca2+] to reach ∼25 μm. After repolarization, the [Ca2+] fell to undetectably low levels with a time constant of ∼0.7 sec, causing the Ca2+-activated tail current to decay similarly (Neves and Lagnado, 1999, their Fig. 10). Together, these measurements indicate that residual Ca2+ stimulated the transfer of vesicles at concentrations that did not activate Ca2+-dependent conductances at the plasma membrane and that were close to the limit of detection with furaptra. In the experiments of Neves and Lagnado (1999), it was quite possible to detect Ca2+ levels of 1.5 μm with furaptra, so this figure provides an upper limit for the [Ca2+] sufficient to stimulate the transfer of vesicles to the RRP.
What is the relationship between transient and continuous exocytosis?
Lagnado et al. (1996) and Rouze and Schwartz (1998) have demonstrated that continuous exocytosis at the bipolar cell synapse can be driven by Ca2+ concentrations of the order of 1 μm. Refilling of the RRP probably occurs at similar low levels of Ca2+, suggesting that continuous exocytosis might occur by the trafficking of vesicles from the bulk cytoplasm, through the reserve pool, to the RRP (Fig. 8). The initial rapid phase of exocytosis requires much higher levels of Ca2+; it is half-maximal at ∼200 μmCa2+ (Heidelberger et al., 1994). So, if the model presented in Figure 8 is correct, the final Ca2+-dependent step during the first phase of release is probably different from that controlling the later phases. This might occur if vesicles docked at release sites could undergo exocytosis from one of two states: state 1 in which fusion requires Ca2+ to bind to a sensor of high affinity (or does not require Ca2+ at all), and state 2 in which fusion requires Ca2+ to bind a sensor of low affinity. If all docked vesicles were in state 2 after a period of rest, the first phase of exocytosis measured experimentally would have a requirement for high levels of Ca2+. But if vesicles arrived at release sites in state 1, the continuous component of exocytosis would occur at lower Ca2+concentrations. The idea of exocytosis from two different states is not new. In other neurons in which the kinetics of exocytosis have been characterized, the fast component of exocytosis synchronous with the action potential is followed by a slower asynchronous component that appears to be driven by residual Ca2+ (Zucker, 1989;Delaney and Tank, 1994; Goda and Stevens, 1994). Geppert et al. (1994)found that the synchronous component of exocytosis was blocked in hippocampal neurons cultured from mice in which synaptotagmin I had been knocked out, although the asynchronous component appeared intact.
At the moment, we do not have any information about the proteins that are involved in these vesicle movements, although Heidelberger (1998)has demonstrated that the hydrolysis of ATP is required for the refilling of the RRP in the terminal of bipolar cells.
A comparison with other synapses
The ribbon synapses of hair cells and photoreceptors also support continuous exocytosis at rates that indicate that release sites are efficiently replenished (Parsons et al., 1994; Rieke and Schwartz, 1996). Recently, Lenzi et al. (1999) used electron tomography to investigate the structure of ribbon synapses in saccular hair cells and found that it was not possible to reduce the number of ribbon-associated vesicles by continuous depolarization, leading to the conclusion that the transfer of vesicles to the ribbon was not rate-limiting during continuous exocytosis. It may therefore be that the Ca2+-dependent transport of vesicles through ribbons also occurs in hair cells.
In neurons that fire action potentials, residual Ca2+ accelerates recovery from synaptic depression after repetitive high-frequency stimulation (Dittman and Regehr, 1998;Stevens and Wesseling, 1998; Wang and Kaczmarek, 1998). This effect can be completely blocked by EGTA, which, as pointed out by Wang and Kaczmarek (1998), suggests that the reserve pool of vesicles is some way from the RRP. In contrast, EGTA did not inhibit the transfer of vesicles to the RRP at the ribbon synapse of bipolar cells (Figs. 6,7). This functional difference might reflect the introduction of lower concentrations of EGTA into the bipolar cell terminal, or it might be caused by the different anatomical arrangement of the reserve pool of vesicles in “classical” synapses and ribbon synapses. At a typical synapse in the CNS, there may be 100–200 vesicles per active zone, of which 10–20 may be docked, while the remainder form a cluster extending ∼500 nm from the presynaptic membrane in which they are linked to each other and the cytoskeleton (Burns and Augustine, 1995). At ribbon synapses, the reserve pool of vesicles appears to be more organized (Gray and Pease, 1971; Jacobs and Hudspeth, 1990;Rao-Mirotznik et al., 1995; Lenzi et al., 1999). In goldfish bipolar cells, vesicles dock to the plasma membrane in two rows of 11, on either side of the ribbon, and each face of the ribbon has approximately five rows of 11 vesicles attached by short filaments (von Gersdorff et al., 1996). The ribbon is only one vesicle diameter from the presynaptic membrane, and extends ∼150 nm. This arrangement may hold vesicles of the reserve pool closer to the RRP and Ca2+ channels than is the case at other synapses.
A.G. was supported by the European Community. We thank Guilherme Neves for helpful discussions.
Correspondence should be addressed to Leon Lagnado, Medical Research Council Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK.