Vesicle Endocytosis Requires Dynamin-Dependent GTP Hydrolysis at a Fast CNS Synapse

See allHide authors and affiliations

Science  07 Jan 2005:
Vol. 307, Issue 5706, pp. 124-127
DOI: 10.1126/science.1103631


Molecular dependence of vesicular endocytosis was investigated with capacitance measurements at the calyx of Held terminal in brainstem slices. Intraterminal loading of botulinum toxin E revealed that the rapid capacitance transient implicated as “kiss-and-run” was unrelated to transmitter release. The release-related capacitance change decayed with an endocytotic time constant of 10 to 25 seconds, depending on the magnitude of exocytosis. Presynaptic loading of the nonhydrolyzable guanosine 5′-triphosphate (GTP) analog GTPgS or dynamin-1 proline-rich domain peptide abolished endocytosis. These compounds had no immediate effect on exocytosis, but caused a use-dependent rundown of exocytosis. Thus, the guanosine triphosphatase dynamin-1 is indispensable for vesicle endocytosis at this fast central nervous system (CNS) synapse.

Endocytosis of distinct kinetics and mechanisms are thought to operate in parallel at nerve terminals (1) and to be critically involved in the maintenance of synaptic transmission (2). However, there is no direct information about the molecular dependence of endocytosis at the nerve terminal. The monomeric guanosine triphosphatase (GTPase) dynamin is indispensable for Drosophila neuromuscular transmission (3, 4) and is thought to play an essential role in the fission reaction of clathrin-coated vesicles during endocytosis (5). However, at goldfish retinal ribbon synapses, fast endocytosis depends entirely on adenosine triphosphatases (ATPases), not on GTPases (6). Thus, it is not known to what extent dynamin is essential for endocytosis at the CNS synapse. At hippocampal synapses (7, 8) and secretory cells (9), the flickering fusion pore called “kiss-and-run” (10) has been proposed to serve as rapid vesicle retrieval. At the giant nerve terminal calyx of Held, a rapid capacitance change, observed with spontaneous or evoked synaptic responses (11), has been implicated as kiss-and-run (12). At this presynaptic terminal, a blockade of GTP hydrolysis causes a use-dependent rundown of excitatory post-synaptic currents (EPSCs), indicating that GTPases are essential for vesicle recycling (13). However, neither the type of GTPases involved nor the GTP-dependent step in the recycling pathway is clarified.

To address these questions, we made simultaneous pre- and postsynaptic recordings combined with capacitance measurements at the calyx of Held synapse in rat brainstem slices (14). Presynaptic Ca2+ currents, evoked by a depolarizing pulse, elicited EPSCs, increased presynaptic membrane capacitance, and transiently decreased presynaptic membrane resistance without changing series resistance (Fig. 1A) (15). To determine whether the capacitance change is correlated with synaptic transmission, we loaded botulinum toxin E (BoNT/E; 200 nM), which specifically synaptosome-associated protein of 25 kD (SNAP-25) (16), directly into the calyceal terminal through a patch pipette. Evoked EPSCs were gradually diminished and eventually abolished 10 to 15 min after loading BoNT/E at 32° to 34°C (Fig. 1B), whereas spontaneous miniature EPSCs remained. Concomitantly, the capacitance change was blocked but for a rapidly decaying component. This component (±SEM) was 64.3 ± 18.5 pF (n = 5) in amplitude and 227 ± 62 ms (n = 5) in weighted mean decay time constant, similar in both magnitude and kinetics to that implicated as kiss-and-run exo-endocytosis (11, 12). This capacitance component remained stable in both kinetics and amplitude during repeated stimulation in the presence of BoNT/E (fig. S1A) (17), but was blocked by Cd2+ (0.25 mM, Fig. 1B) as previously reported (11). A similar BoNT/E-resistant capacitance change could also be evoked by a brief (1-ms) depolarizing pulse (11) (fig. S1B). These results suggest that the rapid capacitance change implicated as kiss-and-run is unrelated to synaptic transmission, but might arise from SNAP-25–independent nontransmitter or extrasynaptic vesicle exocytosis or from a nonvesicular artifact.

Fig. 1.

Exo- and endocytosis of synaptic vesicles in simultaneous pre- and postsynaptic recordings at the calyx of Held. (A) Ca2+ currents (Ipre), EPSCs (Ipost), and capacitance (Cm) change of presynaptic terminals induced by a depolarizing command pulse (Vcom) (10 ms, from –80 to 0 mV). Membrane resistance (Rm) and series resistance (Rs) are simultaneously recorded. Baseline levels of Cm, Rm, and Rs were 24.0 pF, 5.5 giga-ohms (GΩ), and 14.6 mega-ohms (MΩ), respectively. (B) BoNT/E (200 nM) included in the presynaptic patch pipette abolished EPSCs (Ipost) but spared changes in Cm and Rm, both of which were blocked by CdCl2 (0.25 mM, thin traces, superimposed). Baseline levels of Cm, Rm, and Rs were 26.9 pF, 1.3 giga-ohms, and 10.3 mega-ohms, respectively. (C) The relationship between the amplitude of the release-related Cm change (ΔCmt) and the charge of presynaptic Ca2+ currents (QCa). Ca2+ currents were induced by 2- to 10-ms depolarizing pulses or AP-eq stimulation (14) (sample Cm traces in inset). Data points (n = 6 to 14) were fitted by a single exponential curve. (D) The relationship between the ΔCmt decay time constant (τendo) and the ΔCmt amplitude (n = 6 to 8, averaged Cm traces in inset). The ΔCmt decay was fitted by a single (AP-eq) or double (10-ms) exponential curve (in gray, superimposed). Arrows indicate the onset of depolarizing pulse. Error bars show mean ± SEM.

Given that the BoNT/E-resistant capacitance and resistance changes lasted 300 to 400 ms after depolarization, we restricted our analysis to the capacitance change 450 to 500 ms after depolarization (ΔCmt) (mt indicates transmitter release–related membrane capacitance change). As the duration of depolarizing pulse was shortened, the magnitude of ΔCmt decreased exponentially with the charge of Ca2+ current (QCa) (Fig. 1C). ΔCmt could also be evoked by a weak stimulation, which is comparable in QCa to that evoked by an action potential waveform (18) (AP-eq, Fig. 1C). The ΔCmt decayed with a single or double exponential time course; its time constant (τendo) became shorter in linear proportion to ΔCmt (Fig. 1D). The τendo (±SEM) for ΔCmt evoked by a 10-ms pulse was 24.7 ± 3.3 s (n = 8), compared with 10.4 ± 1.4 s (n = 6) for that evoked by AP-eq stimulation. These results are consistent with the previous reports that the time required for endocytosis is proportional to the amount of exocytosis (11, 19, 20).

We next tested whether guanine nucleotide analogs might affect exoor endocytosis. Intraterminal loading of 3 mM guanosine 5′-O-(2-thiodiphosphate) (GDPβS), which inhibits GTPase activity (13), slowed τendo (10 ms) twofold (51.7 ± 7.6 s, n = 6, P < 0.05, Fig. 2A), whereas it had no effect on ΔCmt evoked by the first stimulus (fig. S2A). During repeated 10-ms stimulation at 1- or 2-min intervals, the ΔCmt amplitude decreased, more rapidly for the shorter interstimulus interval (Fig. 2B). These results are consistent with the finding that intraterminal loading of GDPβS has no effect on basal transmission, but slows recovery from short-term depression in a frequency-dependent manner (13), suggesting together that GTPase activity is essential for accelerating recruitment of synaptic vesicles.

Fig. 2.

Effects of guanine nucleotide analogs on vesicle endocytosis and recycling. (A) Averaged ΔCmt traces evoked by 10-ms (left) or AP-eq (right) depolarizing pulse in the presence of GTP (0.3 mM, black, n = 5 to 6), GDPβS (3 mM, blue, n = 6), or GTPγS (0.2 mM, red, n = 4 to 5) in the presynaptic pipette. Averaged ΔCmt traces are superimposed after normalizing the amplitude at 500 to 550 or 600 to 650 ms after depolarization. Arrows indicate the onset of depolarizing pulse. (B) ΔCmt induced by a 10-ms pulse applied at 1-min intervals (open circles) or 2-min intervals (filled circles) in the presence of GTP, GDPβS, or GTPγS (n = 4 to 6) in the presynaptic pipette. Upper panels indicate presynaptic Ca2+ current charge (QCa). Both ΔCmt and QCa are normalized to the initial value. Error bars show mean ± SEM.

Intraterminal loading of a 0.2 mM solution of the nonhydrolyzable GTP analog guanosine 5′-O-(3-thiotriphosphate) (GTPγS) completely abolished endocytosis after 10-ms stimulation, except for the early component (up to 4 s; Fig. 2A), and completely abolished that following AP-eq stimulation. When stimulated repeatedly by a 10-ms pulse, the ΔCmt amplitude decreased in a use-dependent manner; the rate of decline was the same for the 1- or 2-min interstimulus interval (Fig. 2B). GTPγS had no clear effect on the initial amplitude of ΔCmt (21) (fig. S2). These results indicate that GTP hydrolysis is indispensable for synaptic vesicle endocytosis at the calyx of Held.

We next examined whether dynamin might be involved in endocytosis at the calyx of Held. The proline-rich domain (PRD) peptide of dynamin-1 disrupts interactions between dynamin and amphiphysin (22). Injection of this peptide into a reticulospinal presynaptic axon of lamprey increases the number of unfissioned coated vesicles (23). When we loaded the PRD peptide (1 mM) into the calyceal terminal, endocytosis following exocytosis induced by AP-eq stimulation was completely abolished and endocytosis following a 10-ms pulse was blocked, except for the early component (Fig. 3A), whereas ΔCmt at the first stimulus was unchanged (fig. S2). Upon repetitive 10-ms stimulation, ΔCmt underwent a use-dependent rundown (Fig. 3B). A scramble control peptide had no such effect (Fig. 3, A and B). Thus, the effect of PRD peptide was essentially the same as GTPγS, but for the relatively slower rundown of ΔCmt compared with that of GTPγS. This difference may arise from other GTPases such as Rab3 (24) or Arf (25), which may also be involved in vesicle recycling steps.

Fig. 3.

Effects of dynamin-1 PRD peptide on vesicle endocytosis and recycling. (A) Averaged ΔCmt evoked by 10-ms (left) or AP-eq (right) pulse in the presence of PRD peptide (open circles, n = 4 to 5) or scramble peptide (filled circles, n = 5) in the presynaptic pipette (both at 1 mM). Arrows indicate the onset of depolarizing pulse. (B) ΔCmt induced by a 10-ms pulse applied at 1-min intervals (open circles) or 2-min intervals (filled circles) in the presence of PRD peptide (left, n = 4 to 5) or scramble peptide (n = 4). After establishing whole-cell recording, before stimulation, we allowed at least 13 min for peptides to diffuse into calyces. Error bars show mean ± SEM.

To examine whether GTPase-independent recycling mechanisms might additionally operate at the calyx of Held, we carried out continuous 10-ms stimulation at 1-min intervals in the presence of GTPγS while monitoring ΔCmt (Fig. 4A) or EPSCs (Fig. 4B). After 30 to 35 stimuli, ΔCmt became undetectable (<10 fF, n = 4), whereas EPSCs were still observed (26). After 40 stimuli, however, EPSCs were completely abolished (n = 4). Because GTPγS reduces Ca2+ currents by activating trimeric G proteins (27), we compensated for it by increasing extracellular Ca2+ concentration (to 6 mM, with Mg2+ reduced to 0.1 mM). Even in this condition, no EPSC was evoked (Fig. 4B, n = 3), suggesting that vesicle recycling entirely depends upon GTP hydrolysis at this synapse (28).

Fig. 4.

Depletion of recycling vesicles by GTPγS in the presynaptic terminal. (A) Summary data of ΔCmt (four calyces) evoked by the presynaptic Ca2+ currents every 1 min in the presence of GTPγS in the presynaptic pipette. (B) Top panel: EPSCs evoked every 1 min in the presence of GTPγS sampled from different epochs (1 to 3, bottom panel). After the 47th stimulation, extacellular Ca2+ concentration was raised to 6 mM (Mg2+, 0.1 mM, bar), but no EPSC was evoked (3).

From the total vesicles released during repetitive stimulation in the presence of GTPγS, the number of recycling vesicles (±SEM) can be estimated as 42,000 ± 6900 (n = 4) (14). This number is twice as large as that estimated from styryl dye–uptake experiments (29) and corresponds to 20% of total synaptic vesicles at the calyx of Held (29). Also, the maximal releasable pool size can be estimated, from the number of vesicles released by a 10-ms depolarizing pulse, as 4470 ± 400 (n = 14). This number, corresponding to 11% of recycling vesicles, is similar to those previously reported (15, 30, 31).

Synaptic vesicles fused into plasma membrane are reused through endocytosis and recycling (1, 2). Among various recycling routes with distinct kinetics (1), a subsecond capacitance change evoked by a brief presynaptic Ca2+ current has been implicated as kiss-and-run exo-endocytosis at the calyx of Held (11, 12). Although our results obtained with BoNT/E do not support this implication, they do not necessarily preclude the presence of dynamin-dependent kiss-and-run (32) at this nerve terminal. Furthermore, abolishment of endocytosis by GTPγS or dynamin-1 PRD peptide was incomplete when exocytosis was evoked by a 10-ms depolarizing pulse, whereas it was complete when evoked by AP-eq stimulation. Although our results suggest that the dynamin-1–dependent vesicular fission mechanism predominantly mediates endocytosis at the calyx of Held, there may additionally be a dynamin-independent endocytotic mechanism such as bulk endocytosis (1), which may operate after massive exocytosis (29). However, complete block of transmitter release after prolonged stimulation in the presence of GTPγS indicates that GTP hydrolysis is indispensable for vesicle recycling.

Supporting Online Material

Materials and Methods

Figs. S1 and S2


References and Notes

Stay Connected to Science

Navigate This Article