Research Article

A Selective Activity-Dependent Requirement for Dynamin 1 in Synaptic Vesicle Endocytosis

See allHide authors and affiliations

Science  27 Apr 2007:
Vol. 316, Issue 5824, pp. 570-574
DOI: 10.1126/science.1140621


Dynamin 1 is a neuron-specific guanosine triphosphatase thought to be critically required for the fission reaction of synaptic vesicle endocytosis. Unexpectedly, mice lacking dynamin 1 were able to form functional synapses, even though their postnatal viability was limited. However, during spontaneous network activity, branched, tubular plasma membrane invaginations accumulated, capped by clathrin-coated pits, in synapses of dynamin 1–knockout mice. Synaptic vesicle endocytosis was severely impaired during strong exogenous stimulation but resumed efficiently when the stimulus was terminated. Thus, dynamin 1–independent mechanisms can support limited synaptic vesicle endocytosis, but dynamin 1 is needed during high levels of neuronal activity.

Synaptic transmission is dependent on the continuous reformation of synaptic vesicles via local membrane recycling (1, 2). Although the precise mechanisms of synaptic vesicle reformation remain a matter of debate (37), there is strong evidence for a key role of the guanosine triphosphatase (GTPase) dynamin in this process (812), as well as in a variety of endocytic reactions in all cell types (9, 1316). Dynamin is thought to oligomerize at the neck of endocytic pits and to mediate neck constriction and fission (8, 9, 11). However, previous studies have addressed the action of dynamin at synapses through dominant-negative interference or pharmacological inhibition strategies, which may also elicit dominant-negative effects from the inactivated protein. Thus, we investigated the importance of dynamin in membrane traffic at synapses in dynamin 1–null mutants.

Mammals express three dynamins with different expression patterns (fig. S1) (17). Dynamin 1 is expressed exclusively in the brain, whereas dynamin 2 is ubiquitously expressed, and dynamin 3 is expressed selectively in brain and testis (fig. S1B) (18). In neurons, levels of dynamin 1 increase with synapse formation in parallel with the levels of synaptic vesicle proteins (fig. S1E). These and many other observations (9, 18, 19) strongly suggest that dynamin 1 plays a dedicated and essential role in the recycling of synaptic vesicles and, thus, a critical role in nervous system function.

Dynamin 1–KO mice appear normal at birth. A null allele of the mouse dynamin 1 gene was generated by deleting exon 1 (20) (fig. S1F). Heterozygous mice were viable, fertile, and without any apparent health defects. Their matings yielded wild-type (WT), heterozygous (Ht) and, surprisingly, knockout (KO) pups in the expected Mendelian ratio (table S1). At birth, KO mice breathed, moved, and suckled and were not distinguishable from their littermates (Fig. 1A). Thus, dynamin 1 is not required for either embryonic development or for the neurotransmission that supports perinatal life. However, a reduction in the ingestion of milk was apparent in KO pups within several hours after birth (Fig. 1A), and poor motor coordination became obvious over the following days. Overall, dynamin 1–KO pups failed to thrive and died within 2 weeks (fig. S3).

Fig. 1.

Dynamin 1–KO mice appear normal at birth. (A) HT and KO pups several hours after birth. Arrows highlight less milk in the stomach of the KO pup. (B) Immunoblot analysis of cell lysates from primary cortical neuron cultures (15 to 21 DIV) with dynamin isoform-specific antibodies and a pandynamin antibody. Clathrin LC, clathrin light chain.

Immunoblot analysis of brain tissue and cortical neuron cultures demonstrated the absence of dynamin 1 in KO mice and a dramatic decrease of total dynamin levels (Fig. 1B and fig. S2), confirming that dynamin 1 is by far the predominant dynamin in the nervous system. Levels of dynamin 2 and 3, as well as of a variety of proteins involved in synaptic transmission and endocytosis, were not changed (Fig. 1B and fig. S2).

Synaptic transmission in dynamin 1–KO neurons. Whole-cell voltage-clamp recordings from primary cortical cultures were carried out to study the impact of the loss of dynamin 1 on synaptic transmission. Recordings of miniature excitatory and inhibitory postsynaptic currents (mEPSCs and mIPSCs, respectively) revealed a large increase (Fig. 2, A and B), possibly due to increased vesicle size (see below). Next, evoked synaptic transmission was analyzed in paired recordings from low-density cortical cultures. Both EPSCs and IPSCs elicited by single presynaptic stimuli (Fig. 2C) were recorded, and a statistically significant reduction in peak amplitude was observed for IPSCs at KO synapses. Other characteristics of IPSCs were unchanged (table S2).

Fig. 2.

Synaptic transmission in primary cultures of dynamin 1–KO cortical neurons. Cumulative probability histograms of peak mEPSC (A) and mIPSC (B) amplitudes from WT (black lines) and KO cultures (red lines). The mean mEPSC amplitude was increased from 17.0 ± 1.8 pA in WT neurons to 24.8 ± 2.5 pA in the dynamin 1–KO neurons (t test, P = 0.011, n = 2495 and 2435 respectively). A similar increase in amplitude was observed for mIPSCs (mean for WT = 17.3 ± 0.8 pA, n = 2369 versus 22.2 ± 1.0 pA, n = 2731, t test, P = 0.0006). (C) Modest reduction in single evoked (2-ms depolarization to +30 mV) EPSC amplitudes [WT n = 45, KO n = 40, P = 0.0967 (t test)] and significant reduction in single IPSC amplitudes in KO neuron pairs [*P <0.001 (t test, n = 39, 38 for WT, KO)]. (D) Examples of evoked IPSCs for WT and KO respectively. Shown are the 1st, 3rd, and 7th IPSC responses to stimulation at 10 Hz. Presynaptic stimulation is indicated as membrane voltage (gray dotted lines). (E) IPSC depression curves for control (black) and KO (red) during a sustained 10-Hz stimulation. Averaged data points represent 16 (control) and 7 (KO) cell pairs, binned into nonoverlapping groups of 10 responses. (F) Slower recovery of IPSCs from depression induced by 1000 action potentials at 10 Hz (shown in E) as assessed by 0.1-Hz stimulation. Averages represent unbinned data from 12 control and 6 KO cell pairs. The recovery (mean of the last five points) was markedly slower in the KO (t test, P = 0.0024).

IPSCs were further recorded during trains of 1000 stimuli. Peak amplitudes of IPSCs tended to depress faster in KO neurons than in control (WT/Ht) neurons (time constants = 8.5 ± 2.5 s for control versus 3.9 ± 1.2 s for KO; P = 0.26; t test) (Fig. 2E). Furthermore, given the smaller initial peak amplitudes in KO synapses, synaptic transmission failed at earlier times in these synapses. IPSCs recovered (as assessed by test stimuli delivered at 0.1 Hz) within 100 s to 46.0 ± 6.6% of the initial response in control cultures, but to only 11.2 ± 3.2% of their initial IPSC value in KO cultures (Fig. 2F). Thus, dynamin 1 is not essential for synaptic transmission, but is required for efficient and sustained evoked release.

Synaptic vesicles form without dynamin 1. Unexpectedly, in view of the predicted essential role of dynamin 1 in synaptic vesicle recycling, but consistent with the presence of a functional nervous system in KO mice, electron microscopy (EM) revealed an abundance of synaptic vesicles in both control and KO synapses in tissue sections examined at postnatal day 6 (fig. S5), as well as in primary cultures of cortical neurons at all ages examined (Fig. 3, A to E, and fig. S6). However, EM micrographs of cultured cortical neurons [mean age 21 days in vitro (DIV)] revealed a modest reduction (∼20%) in synaptic vesicle numbers (38.9 ± 2.3 per active zone profile in WT versus 31.3 ± 2.1 in KO, P = 0.018, t test). In addition, synaptic vesicles had a more heterogeneous and generally slightly larger size in KO synapses (mean external diameter = 43.13 ± 0.19 nm for WT versus 47.27 ± 0.25 nm for KO; P < 0.0001, t test) (Fig. 3G, see Fig. 3C and fig. S5C) in agreement with the increased quantal size detected electrophysiologically (Fig. 2A).

Fig. 3.

Ultrastructural defects in dynamin 1–KO synapses of cultured neurons. (A) WT synapse. (B and C) KO synapses revealing an abundance of clathrin-coated profiles (arrows highlight stalks connecting clathrin-coated buds), an increase in the average synaptic vesicle size, and the presence (C) of several abnormally large vesicles. (D) A KO synapse with a massive accumulation of interconnected clathrin-coated buds (arrowheads) and only a small cluster (arrow) of heterogeneously sized synaptic vesicles. An asterisk indicates an evagination of an adjacent cell into this nerve terminal. (Inset) A plasma membrane connected network of clathrin-coated pits observed in a serial section from this same synapse (∼200 nm away). (E) Accessibility of clathrin-coated profiles (arrowheads) in KO neurons to cholera toxin-HRP (10 μg/ml for 5 min on ice) supports their connection to the plasma membrane (the arrow indicates the location of the synaptic vesicle cluster within this synapse). (F) Partial reconstruction of a dynamin 1–KO synapse from electron tomography data shows three branched tubular networks (pale green, blue, and yellow) capped by clathrin-coated pits (white arrows) that are connected to the plasma membrane (green) in close proximity to two synaptic vesicle clusters (SVs, blue). (G) Quantification of synaptic vesicle external diameter (black, WT; red, KO; 10-nm bins). Vesicles exceeding 80 nm in diameter were 5.4 times as abundant in KO as in WT synapses (87 vesicles in KO versus 16 vesicles in WT) but were excluded from the analysis presented in (G) as their identification as synaptic vesicles remained questionable. (H) Histogram of clathrin-coated profile (CCP) counts from 75 WT and 87 KO synapses. WT = 0.2 ± 0.05, KO = 4.7 ± 0.89 CCPs/synapse (means ± SEM, P < 0.0001, t test). Data derived from three independent experiments. Horizontal black line is the mean. Scale bars, 200 nm.

Accumulation of clathrin-coated buds. A characteristic feature of KO synapses relative to controls was a much higher occurrence, striking at some synapses, of clathrin-coated vesicular profiles (Fig. 3, B to F and H). In ultrathin sections, many of the coated profiles appeared to be interconnected clathrin-coated buds rather than free coated vesicles (Fig. 3D, inset). Accordingly, they were accessible to extracellular tracers, such as horseradish peroxidase (HRP, a fluid phase tracer) (Fig. 4G) or HRP-labeled cholera toxin (a membrane tracer), even when cultures were labeled with the tracer on ice (Fig. 3E and fig. S6). Electron tomography unambiguously demonstrated that these structures were buds connected to the plasma membrane by narrow branched tubules (Fig. 3F). Thus, the fission of clathrin-coated vesicles was impaired.

Fig. 4.

Activity-dependent synaptic vesicle recycling defects in dynamin 1–KO synapses. (A) Immunofluorescence for clathrin light chain and dynamin 3 reveals a predominantly diffuse distribution in cultured WT neurons, but a punctate and overlapping distribution in KO neurons (see also fig. S4). Treatment with TTX (1 μM, 16 to 24 hours) caused clathrin and dynamin 3 in the KO neurons to redistribute to a diffuse localization resembling the localization of these proteins in untreated WT neurons (scale bar, 10 μm). (B) Morphometric analysis of EM images demonstrating that the accumulation of clathrin-coated profiles in KO synapses was reversed following TTX treatment (1 μM, 16 to 24 hours). (C to I) EM analysis of synapses from cortical cultures of WT and KO brains incubated with the extracellular tracer HRP (10 mg/ml) in control Tyrode's buffer (90 s), following a 90-s stimulation with 90 mM KCl and then a further 10-min recovery period in Tyrode's buffer. (C) Quantification of changes in total synaptic vesicle number and in the number of vesicles positive for the extracellular tracer HRP. (D to I) Representative examples of HRP uptake by WT and KO synapses under the conditions described above. At rest, HRP-labeled clathrin-coated buds emerging from the labeled plasma membrane invagination are visible in the KO synapse (G). Following stimulation, the WT nerve terminal (E) contains labeled and unlabeled vesicles, whereas in the KO synapse (H) synaptic vesicles are nearly depleted (long arrow) and labeled clathrin-coated buds (short arrows) are visible. After recovery, synaptic vesicles, including labeled vesicles, are present in both genotypes (F and I). It is expected that, at this concentration, only a fraction of the endocytic vesicles should take up HRP. EM scale bar, 200 nm.

The accumulation of assembled clathrin coats in KO neurons was also reflected in a modified distribution of clathrin and clathrin adaptors (e.g., α-adaptin subunit of AP-2) as shown by immunofluorescence (Fig. 4A and fig. S7). Although the abundance of these proteins was unchanged (Fig. 1B and fig. S2), their immunoreactivity had a much more punctate pattern in KO cultures. A similar change was observed for dynamin 3, which strongly colocalized with clathrin puncta in KO neurons (Fig. 4A and fig. S4); however, the dynamin 2 staining pattern was weak and diffuse in both KO and controls (fig. S4). These effects, together with the accumulation of coated intermediates revealed by EM, were more prominent when neuron density or culture age was increased. Hence, we suspected that the accumulation of coated intermediates reflected a backup of endocytic traffic in response to spontaneous network activity. Indeed, following blockade of action potential firing with tetrodotoxin (TTX), clustering of clathrin and dynamin 3 immunoreactivity and accumulation of clathrin-coated profiles were no longer observed (Fig. 4, A and B, respectively).

Impaired vesicle recycling during stimulation. To measure stimulation-dependent synaptic vesicle recycling directly, neurons that had been acutely stimulated in the presence of the extracellular tracer HRP were analyzed by EM. After a 90-s stimulation with high-potassium buffer, numerous labeled synaptic vesicles were observed in WT synapses (Fig. 4E). In contrast, in KO synapses, mainly clathrin-coated profiles and other endocytic intermediates were labeled, and the total number of synaptic vesicles was more steeply reduced (Fig. 4, C and H). However, after a recovery period of 10 min in HRP-containing control buffer, numerous labeled synaptic vesicles were visible in synapses of both genotypes (Fig. 4, C, F, and I). On average, recovery in KO synapses lagged behind that of the controls (Fig. 4C), consistent with the delayed recovery from depression observed electrophysiologically (Fig. 2F); nevertheless, robust dynamin 1–independent vesicle formation was observed in some KO synapses (Fig. 4I).

Frequency-dependent endocytic blockade. To gain further quantitative and temporal insight into the synaptic vesicle recycling defect, we performed dynamic imaging assays of exo- and endocytosis. Cultured cortical neurons were transfected with synapto-pHluorin, a chimeric synaptic vesicle protein whose fluorescence is low in the acidic environment within the lumen of a synaptic vesicle and high when exocytosis exposes it to the cell surface (21). In WT neurons, the peak synapto-pHluorin fluorescence in response to a given number of action potentials was lower for a 10-Hz stimulation than for a 30-Hz stimulation, presumably because the ability of endocytosis to keep up with ongoing exocytosis was impaired with an increase of the stimulus frequency to 30 Hz (22) (fig. S8, see also Fig. 5E). In contrast, in dynamin 1–KO neurons, the extent of accumulation was almost identical following both stimulation conditions (fig. S8), which suggested that endocytic capacity was already saturated by stimulation at 10 Hz. Note, however, that the recovery kinetics after the end of the stimulus train were similar in WT and KO synapses (fig. S8).

Fig. 5.

Frequency dependent impairment of synaptic vesicle endocytosis in dynamin 1–KO neurons (A) Representative traces from WT (left panel) and KO (right panel) synapto-pHluorin expressing neurons stimulated in the presence (blue circles) or absence (black squares) of bafilomycin (Baf). A 10-Hz field stimulation began at t = 0 and ended after 30 s (300 action potentials, no Baf) or 90 s (900 action potentials, Baf). (B) Brief application of extracellular, membrane impermeant acid rapidly quenches all surface synapto-pHluorin in the prestimulus period (KO neuron). Following a 30-s stimulus (end marked by arrow), the fluorescence is quenched to the same level as the prestimulus period. The average degree of quenching poststimulus was 94.0 ± 1.4% in WT (n = 5) and 94.6 ± 1.0% in KO (n = 8). (C) The pooled average kinetics of exocytosis (exo = ΔFBaf) from WT (blue) and KO (green) neurons after 900 action potentials (10 Hz in presence of bafilomycin) and the pooled average kinetics of endocytosis (endo = ΔFBaf – ΔFno Baf) from WT (red) and KO (cyan) neurons after 300 action potentials at 10 Hz (stimulus ends at first arrow). The dashed line indicates the extent of exocytosis at the 30-s time point where endocytosis to exocytosis ratios (endo/exo) are calculated. Error bars are shown at two time points on the endocytosis curves (n = 9 for both WT and KO). (D) The average endo/exo ratio after 300 action potentials at 10 Hz as determined by using either synapto-pHluorin (spH) or synaptotagmin1 (syt1)-pHluorin (gray bars). Rescue refers to dynamin 1–KO cells that were cotransfected with synapto-pHluorin and dynamin 1, dynamin 2, or dynamin 3. (E) Endo/exo ratios following 300 action potentials at different frequencies. (F) Mean Endo/exo ratios after 300 action potentials (10 Hz) in 0.75 mM [Ca2+]. The numbers shown in parentheses in (D, E, and F) represent the number of independent experiments and error bars in this figure show ± SEM.

This result was confirmed by a different experimental protocol involving both synapto-pHluorin and the H+-ATPase (adenosine triphosphatase) inhibitor bafilomycin, which block synaptic vesicle reacidification, thus quenching the pHluorin signal after endocytosis (23). At WT but not at KO synapses, the increase in fluorescence produced by 300 action potentials delivered at 10 Hz was higher in the presence of bafilomycin, the “endocytosis-blind” condition (Fig. 5, A and C). Acid-quench experiments excluded the possibility that this difference was due to a defect in acidification (Fig. 5B). Similarly, this difference could not be attributed to differences in exocytosis rates during stimulation (Fig. 5C). As observed above (fig. S8), when the stimulus was removed, fluorescence declined in both sets of synapses at remarkably similar rates. Nonetheless, as 33% of the membrane that had undergone exocytosis had already been recovered by endocytosis during the stimulus, recovery to prestimulus levels was faster in the control synapses (Fig. 5C). Similar results were obtained using synaptotagmin-pHluorin (24) as the reporter of synaptic vesicle recycling (Fig. 5D). The endocytic blockade during stimulation was fully rescued if dynamin 1 was cotransfected into the KO neurons along with synapto-pHluorin (Fig. 5D). When cotransfected under the same conditions, i.e., conditions that result in overexpression, dynamin 3 produced an efficient rescue but dynamin 2, only a partial rescue (Fig. 5D). This result demonstrates that all three dynamin isoforms can participate in synaptic vesicle endocytosis, but also reveals a greater functional similarity between dynamin 1 and 3. It further suggests that dynamin 1–independent recycling is mediated by dynamin 2 and/or 3, consistent with the complete block of synaptic vesicle endocytosis by pharmacological inhibition of dynamin that is not isoform-specific (12, 25).

The endocytic defect in KO synapses during stimulation (as quantified by calculation of the endocytic/exocytic ratio), did not reach statistical significance following 300 action potentials at lower stimulation frequencies [2-Hz and 5-Hz stimulation (Fig. 5E)], and it was less severe during stimulation at 10 Hz when extracellular [Ca2+] was decreased to 0.75 mM (Fig. 5F). Under these conditions, the size of the exocytic responses and the accumulation of presynaptic [Ca2+] are both expected to be attenuated. Conversely, at a higher stimulation frequency (20 Hz), the endocytic/exocytic ratio was dramatically decreased also in WT synapses, as expected (22, 24) (Fig. 5E). Thus, the absence of dynamin 1 lowers the threshold of activity at which synaptic vesicle endocytosis becomes unable to compensate for exocytosis.

Discussion. Surprisingly, dynamin 1, the nervous system–specific dynamin, and by far the major dynamin in neurons, is largely dispensable for the biogenesis and endocytic recycling of synaptic vesicles. Dynamin 1 only becomes essential when an intense stimulus imposes a heavy load on endocytosis and only as long as the stimulus persists. The importance of dynamin 1 under these conditions is likely to be related to its abundance but also, at least in part, to its unique properties. Endocytosis occurs efficiently in both WT and KO synapses immediately after removal of the stimulus, when the endocytic load is still maximal. Thus, dynamin 1 may have a specific function within stimulated synapses. Strong stimulation produces a change in the state of phosphorylation of a variety of nerve terminal proteins (26). Several proteins implicated in endocytosis, including dynamin 1, undergo Ca2+-dependent dephosphorylation by calcineurin, and this process, which is rapidly reversed upon cessation of the stimulus, enhances their recruitment to endocytic sites (26). Given the ability of dynamin 3, when overexpressed, to efficiently rescue the dynamin 1–KO phenotype, it is noteworthy that dynamin 1 and 3 share conserved phosphorylation sites on key residues controlling protein-protein interactions (27, 28). Furthermore, the coclustering of dynamin 3 with clathrin in nerve terminals favors the hypothesis that dynamin 3 has overlapping functions with dynamin 1, but is insufficient to support endocytosis at high levels of activity.

The morphological changes observed in nerve terminals of KO synapses under conditions of basal network activity are consistent with a role of dynamin 1 in fission. The slightly larger and more heterogeneous size of synaptic vesicles raises the possibility that actions of dynamin 1 at the free edge of the clathrin coat of budding vesicles may help to control their size. The fidelity of this process may be disrupted when the overall dynamin content of the nerve terminal is drastically reduced. Indirect effects arising from the lack of dynamin 1 should also be considered because increases in synaptic vesicle size and size heterogeneity have been observed in neurons with other mutations that cause endocytosis defects (29, 30)

The most striking morphological change is the presence of deeply invaginated clathrin-coated endocytic pits, generally located at the tip of branched narrow tubules that are likely generated by the numerous clathrin accessory factors with membrane tubulating properties (31). This is in contrast to the collared but uncoated pits of the Drosophila Shibire mutant (a temperature-sensitive dynamin allele) synapses after stimulation at the nonpermissive temperature (32). Although the longstanding debate concerning the potential occurrence of two major pathways of synaptic vesicle recycling—kiss-and-run versus clathrin-mediated endocytosis—is ongoing (6, 7, 3337), our results strongly support an important role of clathrin-mediated endocytosis.

The selective requirement for dynamin 1 in stimulation-dependent synaptic vesicle reformation is in contrast with the powerful complete block of several forms of endocytosis produced by expression of mutant forms of either dynamin 1 or dynamin 2 in a variety of cell types (8, 13, 15, 16). These dominant-negative effects may reflect hetero-oligomerization of mutant dynamins with endogenous dynamin(s), leading to impairment of actions that require the coordinated function of subunits within a polymeric complex. Overexpressed mutant dynamins could additionally sequester critical binding partners in nonfunctional protein complexes. In conclusion, dynamin 1, one of the most abundant synaptic proteins, is specifically dedicated to control plastic adaptation of the synaptic vesicle recycling machinery to high levels of activity.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Table S1 to S3


References and Notes

View Abstract

Navigate This Article