Corresponding author V. N. Murthy: Department of Molecular and Cellular Biology, Harvard University, 16 Divinity Avenue, Cambridge, MA 02138, USA. Email: firstname.lastname@example.org
We investigated the roles of two Rab-family proteins, Rab3a and Rab5a, in hippocampal synaptic transmission using real-time fluorescence imaging. During synaptic activity, Rab3a dissociated from synaptic vesicles and dispersed into neighbouring axonal regions. Dispersion required calcium-dependent exocytosis and was complete before the entire vesicle pool turned over. In contrast, even prolonged synaptic activity produced limited dispersion of Rab5a. A GTPase-deficient mutant, Rab3a (Q81L), dispersed more slowly than wild-type Rab3a, and decreased the rate of exocytosis and the size of the recycling pool of vesicles. While overexpression of Rab3a did not affect vesicle recycling, overexpression of Rab5a reduced the recycling pool size by 50%. We propose that while Rab3a preferentially associates with recycling synaptic vesicles and modulates their trafficking, Rab5a is largely excluded from recycling vesicles.
Studies in neurosecretory cells have suggested a potentially important role for Rab3a in synaptic vesicle exocytosis (Chung et al. 1999; Schluter et al. 2002; Thiagarajan et al. 2004). Therefore, it was somewhat surprising that loss of all four isoforms of Rab3 in mice, which leads to neonatal lethality, affected transmitter release only slightly (Schluter et al. 2004). Although more subtle synaptic phenotypes have yet to be examined in these quadruple knockouts, earlier studies in Rab3a-null mice revealed a severe impairment of long-term potentiation at mossy fibre terminals in the hippocampus (Castillo et al. 1997). Behavioural defects have also been found in Rab3a knockouts (D'Adamo et al. 2004) as well as in a different Rab3a mutant mouse with lower levels of Rab3a (Kapfhamer et al. 2002). Therefore, it appears that Rab3a does have an important regulatory role in synaptic transmission that remains to be elucidated.
Here, we provide insight into the roles of Rab3a and Rab5a in the vesicle cycle by examining their real-time dynamics within living presynaptic terminals.
Cultures and transfection
Hippocampal neurones were dissociated from 1- to 2-day-old rats using methods previously described (Li & Murthy, 2001). Neonatal rats (P1–2) were anaesthetized with CO2 and decapitated. The hippocampus was removed and incubated in buffered salt solution containing papain for 20 min. The papain solution was removed and the tissue was washed with MEM (Gibco) containing 10% fetal bovine serum, and then grown in MEM containing 10% horse serum. Dissociated neurones were plated on glass coverslips treated with poly d-lysine and rat tail collagen with or without an established layer of astrocytes. Cells were grown at 37°C with 5% CO2. All animal experiments were approved by Harvard University's standing committee on the use of animals in research and training.
Enhanced green fluorescent protein (EGFP)–Rab3a (rat), vesicle-associated membrane protein (VAMP)-2–EGFP (rat) and EGFP–Rab5a (human) were introduced into neurones on day 7 using the calcium phosphate transfection method. Experiments were performed at 14–21 days in vitro. All experiments were done at room temperature (20–22°C). Site-directed mutagenesis using the Quikchange kit (Stratagene, La Jolla, CA, USA) was done to create the two Rab3a nucleotide mutants Q81L and T36N.
For tetanus toxin experiments, cells were incubated with 10 nm tetanus toxin (List Biological Laboratories Inc.; reconstituted in distilled water) for 16 h in the incubator.
Cells cultured on glass coverslips were washed in Hepes buffered salt solution (HBS) prior to a 15 min incubation in 4% paraformaldehyde (Electron Microscopy Science, Fort Washington, PA, USA). After a series of washing steps with phosphate buffered saline (PBS), the cells were permeabilized with 0.1% Triton X-100 in PBS for 5 min. Blocking was done with 10% bovine serum albumin (BSA) for 20 min at room temperature. Primary antibodies: Rab3a monoclonal, C142.2 (Synaptic Systems, Göttingen, Germany) dilution 1/100; anti-synaptotagmin I (Sigma-RBI, St Louis, MO, USA) dilution 1/1000. Secondary antibodies: Alexa 488-conjugated anti-IgG, dilution 1/500; Alexa 568-conjugated anti IgG, dilution 1/500 (Molecular Probes, Eugene, OR, USA). All antibodies were diluted with PBS containing 1% BSA. Incubation with primary antibodies was overnight at 4°C; for secondary antibodies, the incubation time was 1 h at room temperature. Coverslips were mounted with Slowfade antifade agent (Molecular Probes).
FM4-64 labelling For labelling of synaptic vesicles, we mounted coverslips in a closed, small-volume, custom-built laminar flow chamber. A Grass SD9 stimulator was used to evoke action potentials, using brief voltage pulses (1.5 ms, 50 V, bipolar) applied to platinum wires. Synaptic vesicles were labelled with the styryl dye FM4-64 (Molecular Probes) by bathing the cells in HBS (136 mm NaCl, 2.5 nm KCl, 10 mm Hepes, 10 mmd-glucose, 2 mm CaCl2, 1.3 mm MgCl2, pH 7.3) containing 10 μm FM and delivering an electrical stimulation of 900 action potentials at 10 Hz. Following stimulation the dye was present for an additional 90 s to allow for the completion of endocytosis. Cultures were then perfused with dye-free HBS for 10 min to remove surface dye. After initial images of labelled synapses were collected, cultures were subjected to electrical stimulation to evoke the release of FM from vesicles, which is viewed as a loss of fluorescence. The HBS contained 50 μm 2-amino-5-phosphonovaleric acid (APV) and 10 μm 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) to block recurrent activity. The specific loading and destaining stimuli are given in the figure legends associated with each experiment. Images were acquired using a cooled CCD camera (PCO Sensicam, Cooke Corporation, Auburn Hills, MI, USA) on an Olympus IX-70 inverted microscope (60×, 1.35 NA oil lens). Light from the mercury lamp was shuttered using a Uniblitz shutter (Vincent Associates, Rochester, NY, USA). Excitation light for FM4-64 was at 530 ± 20 nm and emitted light was collected using a 590 nm long-pass filter. Typically, images consisting of the entire chip (1280 × 1024 pixels) were collected and stored for offline analysis.
Dispersion of presynaptic proteins The dispersion of EGFP-labelled synaptic proteins was imaged using the same set up as for FM4-64 labelling. The perfusion medium contained 50 μm APV and 10 μm CNQX to block recurrent activity. The excitation light was 480–490 nm and the emitted light was bandpass filtered at 500–550 nm.
Image analysis was done using custom-written routines in MATLAB (The MathWorks Inc., Natick, MA, USA). The calculated fluorescence intensity is an average of the fluorescence values for all pixels contained in a region of interest (ROI) drawn around a synaptic bouton. FM analysis was identical to previously published work (Murthy et al. 1997). For dispersion analysis along the axon, an ROI was drawn approximately 1 μm from the centre of intensity of the synaptic bouton and along the axon. Data are presented as the mean ±s.e.m. For the particular settings we used, photobleaching of EGFP was best fitted to an exponential with a frame constant of 300 frames, which had a negligible effect on estimates of dispersion. Corrections for photobleaching were applied only for measurements of reclustering (Fig. 2, for example).
In experiments involving two-colour imaging, we used the EGFP channel to identify presynaptic boutons and determined the intensity of corresponding puncta in the red channel. In the red channel, we also identified additional puncta to determine the average intensity of control boutons that were EGFP negative. We found that a systematic bias is introduced when comparing intensities of FM4-64 or immunolabel at EGFP positive boutons with control boutons selected as described above. Since EGFP boutons have much better signal-to-noise ratio, even faint synapses not readily identified from FM4-64 or antibody staining can be selected using the EGFP signal. For example, when comparing the amount of FM4-64 label (900 action potentials (APs) at 10 Hz staining stimulus), cytosolic EGFP positive presynaptic boutons had an average intensity of 0.89 ± 0.016 compared to control boutons (1.0 ± 0.011). Therefore, in all our analysis, we normalized all data to the average intensity of FM4-64 in cytosolic EGFP positive boutons (which we refer to as control).
EGFP–Rab3a is targeted to presynaptic terminals and disperses with stimulation
First, we investigated whether light microscopy could be used to detect activity-dependent dissociation and reassociation of Rab3a with synaptic vesicles (Fischer von Mollard et al. 1991). We labelled synapses in culture using an antibody to Rab3a under resting conditions, after neural activity and after recovery from activity. At rest, Rab3a is clustered at the synapse and assumes a punctate distribution along the axons (Fig. 1). In neurones fixed immediately after stimulation (600 APs at 10 Hz), Rab3a is more dispersed and significant labelling is found in the axonal regions between synaptic boutons (Fig. 1A). If a 10 min recovery period is allowed after the 600-AP stimulation, endogenous Rab3a reclusters at the synapse (Fig. 1A). These observations are compatible with the hypothesis that synaptic activity solubilizies Rab3a, which then disperses into the axonal regions.
To facilitate real time observation of Rab3a dynamics, we expressed EGFP–Rab3a in neurones. The low efficiency of transfection in cultures allowed us to compare synapses expressing EGFP–Rab3a with neighbouring control synapses. EGFP–Rab3a puncta colocalize with antibodies to synaptotagmin I confirming synaptic localization of the fusion protein (Fig. 1B). The intensity of synaptotagmin staining at EGFP–Rab3a positive synapses was comparable to that in neighbouring control synapses (0.86 ± 0.041 versus 1.0 ± 0.016, n= 98 and 293 boutons, respectively, P < 0.01). The slightly smaller average intensity is probably due to a small selection bias, since it also occurs when other presynaptic proteins such as VAMP–EGFP, or even cytosolic EGFP, are expressed (see Methods). Immunostaining with Rab3a antibodies indicated that EGFP–Rab3a positive terminals have 1.35 ± 0.23 times as much Rab3a as neighbouring synapses not expressing EGFP–Rab3a (Fig. 1C).
To determine if EGFP–Rab3a undergoes a reversible cycle of dispersion during synaptic activity, we imaged living neurones before, during and after action potential stimulation. At rest, the majority of the EGFP–Rab3a fluorescence is concentrated in clusters at the presynaptic terminal. Upon stimulation, EGFP–Rab3a rapidly disperses into neighbouring axonal regions, and recovers over time after the cessation of stimulation (Fig. 2A). The dispersion and reassociation of EGFP–Rab3a at the synapse was quantified by measuring the change in fluorescence intensity of the synaptic bouton and neighbouring axonal regions over time. The fluorescence intensity at the centre of the bouton decreases rapidly upon stimulation, with a concomitant increase in the fluorescence intensity of neighbouring axonal regions. The summed intensity over a large region including many boutons and axonal segments is not altered significantly (Fig. 2C), indicative of redistribution of fluorescence. When the stimulus is terminated, the fluorescence intensity of the bouton increases over time as EGFP–Rab3a returns to synaptic vesicles and the vesicles recluster at the synapse (Fig. 2). We find that reclustering is more than 90% complete by 4 min. Similar dynamics have been reported for synapsin, another peripherally associated synaptic vesicle protein (Chi et al. 2001).
Dispersion of EGFP–Rab3a is faster than that of vesicular membrane-bound proteins
Is the dispersion of EGFP–Rab3a into axonal regions due to diffusion of cytosolic EGFP–Rab3a (presumably bound to Rab GDP dissociation inhibitor (RabGDI)) or the lateral movement of vesicular membrane that is introduced into the plasma membrane (Fig. 3A)? To distinguish between these two possibilities, we compared the rate of dispersion of Rab3a with that of VAMP, an integral vesicular membrane protein. Previous studies have shown that VAMP–EGFP localizes to presynaptic vesicle clusters at rest, disperses laterally in an activity-dependent manner along the plasma membrane and reclusters after termination of the stimulus (Sankaranarayanan & Ryan, 2000; Li & Murthy, 2001). We measured the appearance of fluorescence in the axonal regions rather than the loss of fluorescence in the synaptic regions to avoid complications due to pH sensitivity of VAMP-EGFP. During a stimulus of 300 APs at 10 Hz, the rate of dispersion for EGFP–Rab3a is faster than that for VAMP–EGFP (Fig. 3). This difference is most parsimoniously explained by dissolution of EGFP–Rab3a into the cytoplasm through binding to RabGDI, and subsequent diffusion in solution. VAMP–EGFP diffusion is in the plane of the membrane and is expected to be slower. In a subset of experiments, we also imaged a different vesicular protein, synaptophysin, and Rab3a in the same synapses by using spectrally distinct fusion proteins and found similar results (Supplemental material).
Reclustering of Rab3a follows time course of vesicle reclustering
Previous studies have shown that synaptic vesicles can be recaptured all along the axonal membrane and the reclustering of these vesicles occurs over several minutes (Li & Murthy, 2001). We wondered whether Rab3a is reattached to the vesicular membrane as these vesicles return to the cluster. If so, the kinetics of the loss of fluorescence in the axonal regions after cessation of activity should be similar to that of integral vesicular components such as VAMP–EGFP. Fluorescence imaging revealed that this was indeed the case (Fig. 3B). Therefore, the reattachment of Rab3a to synaptic vesicles must occur before or during return of endocytosed vesicles back to the synaptic cluster. Simultaneous imaging of Rab3a and VAMP also confirmed these observations (Supplemental material).
Dispersion of Rab3a is complete before the turnover of all vesicles
We observed that during a stimulus of 900 APs at 10 Hz, the dispersion of EGFP–Rab3a reaches a plateau phase within the first 300 APs (Fig. 4A). We wondered if this plateau occurs because (i) all Rab3a dissociates from vesicles, (ii) only a certain fraction of vesicles are labelled with Rab3a, or (iii) if a certain fraction of Rab3a is resistant to dissociation – for example, if it is present on resting vesicles that do not recycle under conditions of our stimulus (Murthy & Stevens, 1999; Sudhof, 2000; Harata et al. 2001a). We can distinguish between these possibilities by comparing the distribution of EGFP–Rab3a to that of a cytosolic protein such as EGFP. The ratio of fluorescence in the bouton to that in the axon for cytosolic EGFP was around 2, reflecting the larger volume of the bouton compared to the axon and the subresolution dimension of the axon (diameter of less than 200 nm). This ratio did not change upon stimulation (Fig. 4B). For EGFP–Rab3a, this ratio is high at rest since most of the fluorescence is in the bouton, where the vesicles are located. Upon stimulation, the ratio approaches the value for cytosolic EGFP within about 300 APs at 10 Hz (Fig. 4B). The observation that the dispersion of EGFP–Rab3a is complete during this stimulation is surprising since less than half of the vesicles at the synapse undergo recycling even under sustained stimulation and the other vesicles remain at rest (Murthy & Stevens, 1999; Harata et al. 2001a; Rizzoli & Betz, 2005).
Dispersion of EGFP–Rab3a occurs at a wide range of frequencies
It has been suggested that the mode of synaptic vesicle release may differ depending on the frequency of stimulation (Pyle et al. 2000; Richards et al. 2000; Richards et al. 2003; Rizzoli & Betz, 2004). For example, higher stimulus frequency may induce rapid local reuse of synaptic vesicles (Pyle et al. 2000; Sara et al. 2002). Alternately, at low frequencies ‘kiss-and-run’ exocytosis with the rapid opening and closing of a fusion pore may occur (Gandhi & Stevens, 2003). These different proposed modes of exocytosis may involve differential activation of Rab3a. Therefore, we investigated whether dispersion of EGFP–Rab3a occurs at different stimulus frequencies. In response to 300 APs delivered at 1, 10 or 20 Hz, EGFP–Rab3a dispersed robustly (Fig. 4C), suggesting that even if different modes of exocytosis exist at hippocampal synapses during physiological stimulation, dissociation of Rab3a from vesicles occurs with high probability.
We note that the rapid dispersion of EGFP–Rab3a is unlikely to be due to overexpression since there was no correlation between the kinetics of dispersion and the level of expression of EGFP–Rab3a (Fig. 4D).
Dispersion of Rab3a occurs only during calcium-triggered exocytosis
We used two experimental approaches to investigate the signals that induce the dispersion of Rab3a into neighbouring axonal regions during synaptic activity. Two possibilities are exocytosis itself and an increase in calcium concentration during synaptic activity. We separated these two processes by either (i) inhibiting vesicle fusion without affecting calcium influx through the addition of tetanus toxin, or (ii) inducing exocytosis in the absence of calcium by hyperosmotic stimulation.
Cells expressing EGFP–Rab3a were pretreated with 10 nm tetanus toxin, which cleaves VAMP and inhibits neurotransmitter release. In cells pretreated with tetanus toxin, there was no discernable dispersion of EGFP–Rab3a from the synapse into neighbouring axonal regions (Fig. 5A). This lack of dispersion suggests the protein remains associated with synaptic vesicles during the influx of calcium, and that calcium signalling alone is not sufficient to induce the release of Rab3a from synaptic vesicles. Matched control neurones expressing the EGFP–Rab3a construct and taken from the same culture plate underwent dispersion to the expected degree (Fig. 5A). The slow downward trend in tetanus toxin treated synapses is in part due to bleaching and slight focus drift (bleach control is shown in Fig. 5A, continuous line). We confirmed the activity of the tetanus toxin by the lack of vesicle recycling monitored using FM4-64 (data not shown).
In complementary experiments, we addressed the role of vesicle fusion in triggering the dispersion of Rab3a, in the absence of calcium influx. Hypertonic solutions evoke exocytosis in a calcium-independent manner and release is drawn from the same pool of readily releasable vesicles (Rosenmund & Stevens, 1996). Previous studies have shown that the fusion of the readily releasable pool in response to hypertonic sucrose is equivalent to a stimulation of 40 APs at 20 Hz (Schikorski & Stevens, 2001). In response to a stimulus of 40 APs at 20 Hz, EGFP–Rab3a undergoes a clearly discernible dispersion, easily seen as a drop in the fluorescence in the centre of the synaptic bouton (Fig. 5B). In contrast, addition of a hypertonic solution for several seconds, which releases the readily releasable pool of vesicles, causes no discernible dispersion of EGFP–Rab3a from the synaptic cluster (Fig. 5C). This suggests that vesicle fusion alone, in the absence of calcium entry, is not sufficient to induce the dispersion of Rab3a.
GTP hydrolysis mutant (Q81L) disperses more slowly and impairs release
To investigate the physiological relevance of the GTP hydrolysis event in the Rab3a cycle and in synaptic vesicle release, we generated a point mutation Q81L in EGFP–Rab3a, which impairs the rate of hydrolysis of GTP, keeping Rab3a in the GTP-bound or active form longer. The Q81L point mutation fixes Rab3a in the GTP-bound form and prevents its dissociation from synaptic vesicle membranes (Brondyk et al. 1993). Expression of EGFP–Rab3a Q81L in hippocampal neurones did not cause any noticeable change in the morphology of axons or presynaptic boutons. Upon stimulation, EGFP–Rab3a Q81L disperses into the axonal regions, but at a rate substantially slower than that of wild-type Rab3a (Fig. 6A). Even a 900 AP stimulus at 10 Hz does not induce the maximal dispersion seen at 300 APs for wild-type Rab3a.
We then examined the effect of expression of wild-type and mutant Rab3a on synaptic vesicle recycling using FM4-64 imaging. Previous studies using neuroendocrine and PC12 cells have shown that expression of Rab3a severely impairs the calcium-dependent release of secretory vesicles (Chung et al. 1999; Schluter et al. 2002; Thiagarajan et al. 2004). To examine if a similar effect occurs in hippocampal neurones, we measured vesicle recycling parameters at synapses expressing EGFP–Rab3a and neighbouring control synapses in the same cultures. Synapses were labelled with FM4-64 using a stimulus of 900 APs at 10 Hz to label all recycling vesicles. The size of the recycling pool of vesicles was calculated by measuring the initial fluorescence intensity of synapses minus the fluorescence intensity remaining after exhaustive stimulation to release all labelled vesicles. The normalized recycling pool size at synapses expressing EGFP–Rab3a was 0.94 ± 0.028 (176 boutons, 10 experiments) compared to 1.0 ± 0.018 (128 boutons, 4 experiments) at EGFP-expressing control synapses (P < 0.001) (Fig. 6C). The slightly smaller pool size is likely to be due to the selection bias discussed earlier, since it is also observed with other presynaptic proteins. The rate of exocytosis was also measured for the first few hundred APs delivered at 10 Hz – EGFP–Rab3a synapses were not different from control synapses (Fig. 6D). The time constant of destaining was 67.8 ± 1.2 s (n= 96) for control boutons and 67.6 ± 1.8 s (n= 36) for Rab3a positive boutons. This suggests that moderate over-expression of EGFP–Rab3a does not have an effect on vesicle recycling in hippocampal neurones.
Unlike synapses expressing EGFP–Rab3a, synapses expressing EGFP–Rab3a Q81L had impaired vesicle exocytosis. At Q81L positive synapses, the total pool of recycling vesicles labelled with 900 APs was smaller than for wild-type EGFP–Rab3a synapses (0.75 ± 0.032 versus 0.94 ± 0.028, n= 94 and 176 boutons, respectively, 10 experiments; values normalized to control EGFP-expressing synapses, P < 0.001, Fig. 6C). In addition, measurement of the kinetics of release showed that the Q81L positive synapses had a significantly slower rate of exocytosis (Fig. 6D). The time constant of exocytosis was 106.1 ± 4.2 s (n= 45) for Q81L positive boutons, significantly slower than control or Rab3a positive boutons (P < 0.001). Labelling with synaptotagmin antibodies indicated that the total number of vesicles at EGFP–Rab3a (Q81L) positive boutons was similar to those at EGFP–Rab3a positive boutons (0.94 ± 0.051 versus 1.0 ± 0.048, over 100 boutons each). Therefore, expression of Q81L alters the functional properties of synapses, without altering the total number of vesicles.
We were unable to test whether EGFP–Rab3a T36N, a mutant that is thought to retain Rab3a in the GDP-bound form (Schluter et al. 2002), has altered dynamics. Neurones expressing this mutant did not survive beyond two days post transfection and had a uniform distribution of fluorescence. It is likely that the T36N mutant is in a nucleotide free state, since the off-rate for GDP binding is high (Burstein et al. 1992). This nucleotide-free state of Rab3a-T36N might have wide-spread, non-specific cellular effects.
EGFP–Rab5a localizes to the presynaptic terminal and undergoes minimal dispersion during evoked vesicle release
To investigate the role of Rab5a in the synaptic vesicle cycle, we expressed EGFP–Rab5a (Rosenfeld et al. 2001) in cultured hippocampal neurones. EGFP–Rab5a clusters along the axon, forming green puncta separated by axonal regions of fainter fluorescence (Fig. 7A). Although EGFP–Rab5a was also present in the somatodendritic compartments, by restricting our imaging to regions 200 μm or more away from the transfected cell body, we were able to examine axonal and presynaptic regions. Immunostaining of EGFP–Rab5a expressing neurones with antibodies to synaptotagmin demonstrated that the EGFP–Rab5a puncta are localized to presynaptic terminals (Fig. 7A). The intensity of SytI staining in EGFP–Rab5a positive synapses was lower than control synapses, even after taking into account the slight selection bias (Rab5a: 0.74 ± 0.031, n= 81 boutons versus Rab3a: 0.86 ± 0.041, 98 boutons, P < 0.01). Evidently, synapses that express EGFP–Rab5a have fewer vesicles. To determine if Rab5a undergoes activity-dependent dispersion at the presynaptic terminal we stimulated synaptic boutons expressing EGFP–Rab5a for 900 APs at 10 Hz. We saw minimal dispersion of Rab5a during sustained stimulation, far less than that seen with EGFP–Rab3a (Fig. 7B). The small amount of dispersion observed could be due to a few vesicles that carry Rab5a flattening into the plasma membrane upon exocytosis during sustained stimulation and undergoing lateral dispersion.
Comparison of the fraction of Rab5a or Rab3a appearing in the axonal regions upon stimulation confirms that most of the Rab5a remains within the vesicle clusters even after sustained stimuli (Fig. 7C). While Rab3a increases nearly 100% above resting levels in the axons upon stimulation, Rab5a levels rise only to about 20%. A parsimonious explanation is that Rab5a is not associated with vesicles that undergo recycling. This is reinforced by the large amount of a vesicular protein, VAMP, appearing in the axonal regions during the same stimulation (Fig. 7C). Simultaneous imaging of synaptophysin–ECFP and EYFP–Rab5a confirmed the above results (Supplemental material).
Expression of EGFP–Rab5a reduces synaptic vesicle recycling
To measure vesicle recycling in synapses expressing EGFP–Rab5a, boutons were labelled with FM4-64 using a stimulus of 900 APs at 10 Hz. Expression of EGFP–Rab5a in hippocampal neurones impairs the uptake of FM4-64 as shown in Fig. 8A. We determined that the fluorescence intensity of synapses expressing EGFP–Rab5a was about half that in EGFP-expressing control cells (0.64 ± 0.032 versus 1.0 ± 0.018, 68 and 128 boutons each, P < 0.01, Fig. 8B). This indicates that substantially fewer vesicles participate in recycling in EGFP–Rab5a expressing synapses than in control synapses or synapses expressing EGFP–Rab3a. To determine if expression of Rab5a also alters the rate of exocytosis, we quantified the rate of loss of FM dye from terminals during 10 Hz stimulation (Fig. 8C). We found that EGFP–Rab5a expressing synapses have a slightly slower rate of destaining compared to control synapses (time constant: 72.5 ± 3.9 s versus 64.9 ± 4.6 s, P < 0.05, 163 and 193 boutons, respectively).
To investigate the physiological relevance of GTP hydrolysis in the role of Rab5a in the synaptic vesicle cycle we transfected cells with a GTP bound form of EGFP–Rab5a (Q79L) (Hoffenberg et al. 1995). Neurones expressing EGFP–Rab5a Q79L did not survive in culture past 2 days post-transfection, and we did not observe distinct clusters of EGFP–Rab5a Q79L. Therefore, we were unable to characterize dispersion of the mutant Rab5a or investigate synaptic vesicle recycling using FM4-64.
The mechanisms regulating the progression of a vesicle through the synaptic vesicle cycle are largely unknown (Murthy & De Camilli, 2003). Recent studies have uncovered multiple nested vesicle recycling pathways within the synaptic terminal, suggesting a complex regulation of postendocytic traffic (Pyle et al. 2000; Gandhi & Stevens, 2003; Rizzoli & Betz, 2004, 2005). There are at least three functionally distinct synaptic vesicle pools at the hippocampal nerve terminal; the majority of vesicles comprise the resting or reserve pool, less than 50% belong to an actively recycling pool, and the readily releasable pool is less than 5% of the total vesicle pool size (Sudhof, 2000; Rizzoli & Betz, 2004; Li et al. 2005; Rizzoli & Betz, 2005). Although the actively recycling vesicles differ from those in the reserve pool in terms of mobilization and endocytosis, they are scattered throughout the nerve terminal with only a small bias in their location (Schikorski & Stevens, 2001; Rizzoli & Betz, 2004). In this study, we have investigated the roles of Rab3a and Rab5a in the synaptic vesicle cycle using real-time imaging. The findings presented here raise the possibility that synaptic vesicles may be heterogeneous in their composition of Rab proteins, which in turn might confer distinct identities to vesicles clustered within the nerve terminal.
The first important finding of our study is that the dissociation of Rab3a from synaptic vesicles is complete within a few hundred action potentials. This is a surprising result since less than half the total vesicle pool at a synapse participates in recycling, with the remaining vesicles constituting a resting or reserve pool of vesicles (Murthy et al. 1997; Harata et al. 2001b; Rizzoli & Betz, 2005). This result suggests that only a subset of the synaptic vesicle pool is labelled with Rab3a. We ruled out the possibility that complete dispersion is due to dissociation of Rab3a from all vesicles in a manner independent of exocytosis, but dependent on intraterminal calcium. We find that Rab3a dispersion is dependent on vesicle exocytosis since treatment of neurones with tetanus toxin prevents stimulus-evoked dispersion. Exocytosis without calcium entry, however, is not effective in dispersing Rab3a. Since the dispersion of Rab3a is faster than that of VAMP, Rab3a does not appear to stay with the vesicular membrane after fusion. Based on these findings, we suggest that Rab3a is associated preferentially with those vesicles that participate actively in release and recycling. Vesicle repriming in hippocampal synapses can occur within minutes and a substantial portion of newly endocytosed vesicles can be re-released in 2–3 min (Ryan & Smith, 1995). Our measurements indicate that recovery of Rab3a at synapses can be substantial enough to support such rapid repriming.
A second interesting finding of our study is that expression of EGFP–Rab3a does not have an effect on synaptic vesicle recycling. The total recycling pool measured with FM4-64 labelling was very similar in EGFP–Rab3a positive and negative synapses, and there was no difference in the rate of FM release during stimulation. This observation was somewhat unexpected considering the dramatic effect seen with the overexpression of Rab3a in chromaffin and PC12 cells (Chung et al. 1999; Schluter et al. 2002). Overexpression of Rab3a in PC12 cells results in the virtual abolition of evoked synaptic vesicle release, and an increase in the level of constitutive vesicle release (Schluter et al. 2002). The lack of observed effect in our neurones may be because PC12 cells are imperfect models for neuronal exocytosis, or because the level of expression of EGFP–Rab3a is not as high in our system as in these non-neuronal cells. The amount of Rab3a at hippocampal synapses expressing EGFP–Rab3a in our experiment was only 1.35 times that of control synapses.
A third finding is that overexpression of the GTPase deficient mutant EGFP–Rab3a Q81L perturbs presynaptic vesicle recycling. Expression of EGFP–Rab3a Q81L results in a decrease in the total number of vesicles that undergo recycling as well as the rate of exocytosis. This impairment of synaptic vesicle recycling, combined with the observation that the dispersion of EGFP–Rab3a Q81L is reduced, suggests that the hydrolysis of GTP-Rab3a is necessary for the efficient release of synaptic vesicles. Synapses expressing the GTPase deficient mutant Rab3a still undergo synaptic vesicle recycling. This could be partly due to the presence of endogenous wild-type Rab3a. Alternately, it is possible that GTP hydrolysis merely accelerates an intrinsic basal release probability and in its absence there is still a basal level of release. It remains to be seen if the effect of Rab3a Q81L is greater in the absence of endogenous wild-type Rab3a.
The results from our study of the expression of EGFP–Rab3a in hippocampal neurones are consistent with those of a study done on evoked acetylcholine release from Aplysia synapses. Expression of recombinant Aplysia Rab3 had little effect on vesicle release; however, expression of the GTPase-deficient Rab3 (Q80L) reduced neurotransmitter release (Doussau et al. 1998). Ultrastructural analysis of synaptosomes isolated from Rab3a-null mutant mice showed a reduction in the activity-dependent recruitment of vesicles to the active zone, indicating a role for Rab3a in synaptic vesicle trafficking (Leenders et al. 2001). In Caenorhabditis elegans, deletion of Rab3 resulted in a reduction in the number of vesicles at the active zone, but an increase in the number of vesicles within the perisynaptic region (Nonet et al. 1997). Other studies have also shown that introduction of GDPβS, a non-hydrolysable analogue of GDP, into the presynaptic terminal results in the slow inhibition of neurotransmitter release and the loss of vesicles attached to the plasma membrane, and addition of GTPγS enhances the number of docked vesicles (Hess et al. 1993; Augustine et al. 1999). These effects could have been, at least in part, due to disruption of the Rab3a GTPase cycle. Kavalali and colleagues have recently used a novel assay using permeabilized synapses to demonstrate a role for GTP in vesicle recruitment (Mozhayeva et al. 2004). Although they find that GTP and GTPγS have similar facilitatory roles, the rate of exocytosis in their case is substantially slower than the rates at intact synapses (Mozhayeva et al. 2004). Their study points to a clear role of GTP binding in vesicle recruitment, which could potentially involve Rab3a. It is worth noting that Rab proteins have been hypothesized to regulate the assembly of the core complex of SNARE proteins and may do so through interaction with effector molecules in a GTP-dependent fashion (Zerial & McBride, 2001). In the GTP bound form Rab3a interacts with various effector proteins, notably rabphilin 3A (Li et al. 1994) and RIM (Wang et al. 1997).
Our final set of findings relate to Rab5a, whose real-time dynamics were imaged here for the first time in presynaptic terminals of living mammalian synapses. An earlier study in cultured hippocampal neurones demonstrated that overexpressed Rab5a is targeted to synaptic terminals and labels vesicular as well as endosomal structures, but that study did not examine real-time dynamics of the protein in synapses (de Hoop et al. 1994). We find that EGFP–Rab5a is targeted to presynaptic boutons, but undergoes minimal dispersion during evoked vesicle release. This is in line with biochemical observations that Rab5a does not undergo a cycle of dissociation and association with synaptic vesicles (Fischer von Mollard et al. 1991, 1994). Any appearance of Rab5a in the axonal region is likely to be due to diffusion within the plasma membrane following the insertion of vesicles. Even so, the amount of dispersion of EGFP–Rab5a is significantly smaller than that of vesicular components such as VAMP-EGFP. Our working model to explain these findings is that vesicles that undergo exocytosis have little Rab5a associated with them, which would imply that a great proportion of Rab5a is associated with resting vesicles, some of which might represent recycling compartments.
How do our findings suggesting differential localization of Rab proteins on synaptic vesicles compare with previous biochemical analysis of Rab3a and Rab5a (Fischer von Mollard et al. 1991, 1994)? Experiments with synaptosomes demonstrated that Rab3a undergoes a cycle of association and dissociation with synaptic vesicles; however, Rab5a is found in the GDP-bound form in association with vesicular membrane (Fischer von Mollard et al. 1991, 1994). Experiments using vesicles isolated by density gradients cannot easily differentiate between resting and recycling vesicles. Considering that there are over 100 vesicles in the average hippocampal resting pool and only about 25 vesicles in the actively recycling pool (Sudhof, 2000), it is conceivable that a subset of Rab5a negative vesicles could be missed in any biochemical analysis. In addition, a surprising number of Rab proteins were found on clathrin coated vesicles from brain using tandem mass spectrometry (Blondeau et al. 2004), pointing to the complexities of defining composition of vesicles. We also note that a recent study indicated that most of the Rab5 at hippocampal synapses is present at postsynaptic sites (Brown et al. 2005). Our hypothesis of selective localization of Rab proteins is further supported by the observation that actively recycling synaptic vesicles taken up through clathrin-mediated endo-cytosis lack both Rab3a (Maycox et al. 1992) and Rab5a (Fischer von Mollard et al. 1994). Such a deficiency is predicted by the demonstration that Rab3a does not remain with the synaptic vesicle membrane after fusion, and that there is minimal trafficking of Rab5a containing vesicles during synaptic transmission. This issue of the selective localization of Rab proteins on synaptic vesicles awaits further study through the examination of the ultrastructural localization of Rab proteins in the presynaptic terminal.
In our experiments, expression of EGFP–Rab5a has a significant effect on synaptic vesicle recycling as it significantly reduces the uptake of FM4-64. Although we were unable to determine the exact level of overexpression of Rab5a due to lack of antibodies that can distinguish pre- and postsynaptic Rab5a, we expect it to be similar to that for Rab3a, since similar expression vectors were used. The reduction in FM4-64 uptake in neurones expressing EGFP–Rab5a suggests that the number of vesicles that participate in the synaptic vesicle cycle is reduced. We note that the reduction in pool size cannot fully explain the substantially lower dispersion observed with EGFP–Rab5a, since even synapses with small recycling pools exhibit robust dispersion of EGFP–Rab3a. The rate of exocytosis of synapses overexpressing EGFP–Rab5a, estimated from the rate of loss FM4-64 during 10 Hz stimulation is lower than their untransfected counterparts. The fractional rate of dye loss was only slightly smaller, but since EGFP–Rab5a synapses have smaller recycling pools, the absolute rate of vesicle release is likely to be significantly smaller. Although the simplest explanation for our results is that overexpression of Rab5a impairs exocytosis, it remains possible that EGFP–Rab5a somehow selectively localizes to smaller, less efficient synapses without altering their properties. This latter possibility, however, is not supported by an examination of synapses simultaneously expressing synaptophysin–ECFP and EYFP–Rab5a – the two probes show almost perfect colocalization.
A recent study at the Drosophila larval neuromuscular junction suggested that overexpression of Rab5a enhances synaptic transmission and its absence results in reduction of vesicle exocytosis (Wucherpfennig et al. 2003). Another study using Drosophila photoreceptor synapses has suggested a role for Rab5 in preventing homotypic fusion between synaptic vesicles and therefore in maintaining the synaptic vesicle size (Shimizu et al. 2003). Since vesicle pools were not measured in the fly studies and release probability was not measured in our study, a simple comparison between the studies may not be appropriate. It is possible that Rab5 has different roles in different species or in different types of synapses. Alternatively, a long-term change in Rab5a expression might alter signalling pathways (Brown et al. 2005) or axonal traffic (Kanaani et al. 2004), either of which might indirectly alter synaptic structure and function.
In addition to the interesting individual findings regarding Rab3a and Rab5a, taken together our results also suggest the potential for a dramatic difference between these two small GTPases in terms of their association with functionally distinct vesicles at presynaptic boutons. Based on the observation that Rab3a and Rab5a have very different dispersion profiles in response to synaptic activity, we speculate that Rab3a may associate largely with recycling vesicles, whereas Rab5a does not. Since the discovery that a large number of synaptic vesicles at central synapses appear to not actively recycle (Murthy et al. 1997; Harata et al. 2001a), the source of vesicle heterogeneity has been a puzzle. An underlying molecular heterogeneity might underlie this functional heterogeneity. Only one molecule, syntaxin, has previously been suggested to have differential distribution (Mitchell & Ryan, 2004). Our study adds two more candidates – Rab3a for recycling vesicles and Rab5a for resting vesicles.
In summary, we have determined that a majority of the Rab3a at a synapse dissociates from the synaptic vesicle membrane even though a large number of vesicles remain at rest. Our data suggest that vesicles that actively recycle carry most of the Rab3a in synaptic terminals, and that the resting vesicles do not. A GTP hydrolysis mutant, EGFP–Rab3a Q81L, shows slower dispersion kinetics and impairs synaptic vesicle release, suggesting that the hydrolysis of Rab3a–GTP may be a rate-limiting step in synaptic vesicle exocytosis. Rab5a undergoes minimal dispersion during evoked exocytosis, and expression of EGFP–Rab5a significantly impairs synaptic vesicle recycling. We propose that Rab5a may be largely excluded from the actively recycling pool, instead labelling resting vesicles or endo-somal structures that may resemble synaptic vesicles morphologically.
We thank Dr Juan Burrone for critical discussions and comments, and Dr Brian Knoll for EGFP–Rab5a and EGFP-Rab5a(Q79L). This work was supported by a grant from the NIH. Research in V.N.'s laboratory was also supported by grants from the NSF, the EJLB Foundation, the Pew Scholars Program and the Klingenstein Fund.