Synaptic vesicles segregate into functionally diverse subpopulations within presynaptic terminals, yet there is no information about how this may occur. Here we demonstrate that a distinct subgroup of vesicles within individual glutamatergic, mossy fiber terminals contain vesicular zinc that is critical for the rapid release of a subgroup of synaptic vesicles during increased activity in mice. In particular, vesicular zinc dictates the Ca2+ sensitivity of release during high-frequency firing. Intense synaptic activity alters the subcellular distribution of zinc in presynaptic terminals and decreases the number of zinc-containing vesicles. Zinc staining also appears in endosomes, an observation that is consistent with the preferential replenishment of zinc-enriched vesicles by bulk endocytosis. We propose that functionally diverse vesicle pools with unique membrane protein composition support different modes of transmission and are generated via distinct recycling pathways.
Glutamatergic synaptic vesicles within the same terminal can be sorted into various pools based on their release probability (Rizzoli and Betz, 2005). Recent observations indicate that distinct pools contribute to different types of physiological activity (Prange and Murphy, 1999; Sara et al., 2005; Groemer and Klingauf, 2007; Fredj and Burrone, 2009). The mechanisms that sort vesicles into distinct pools and enable certain vesicles to participate in a specific physiological activity remain uncertain. One compelling possibility is that synaptic vesicles destined for distinct pools are generated through different adaptor-dependent mechanisms. In support of this idea, recycling pathways involving adaptor protein 2 (AP2) or AP3 lead to the generation of synaptic vesicles with different molecular composition (Faúndez et al., 1998; Shi et al., 1998; Salazar et al., 2004a,b). Here we hypothesized that specific vesicular proteins segregate vesicles within synaptic terminals and that this segregation is physiologically important for regulating information transfer.
Thus, we investigated transmission at the excitatory mossy fiber (MF) synapses onto CA3 pyramidal neurons. Electrophysiological and theoretical studies suggest that the dentate gyrus is involved in pattern separation and rate remapping, and the unique physiological features of MFs may be important for the encoding process involved in these functions (Treves and Rolls, 1992; Nakazawa et al., 2002; Leutgeb et al., 2007; McHugh et al., 2007). Moreover, contribution of dentate granule cells to information coding is primarily influenced by the dynamics of transmitter release at the MF synapses.
Hippocampal MF synapses terminating on pyramidal cells release glutamate from several release sites (Amaral and Witter, 1989; Acsády et al., 1998). Increases in presynaptic firing frequency lead to rapid and robust augmentation in synaptic strength, such as pronounced paired-pulse and frequency facilitation (Salin et al., 1996; Nicoll and Schmitz, 2005). Synaptic transmission and short-term plasticity are supported by a large pool of vesicles that can be rapidly recycled (Hallermann et al., 2003; Rollenhagen et al., 2007). Another unique feature of MF terminals is their high vesicular zinc content that colocalized with glutamate (Cole et al., 1999). Exocytosis of zinc during synaptic activity has been convincingly demonstrated (Kay, 2003; Qian and Noebels, 2005), but its actual release into the synaptic cleft and hence its regulatory role in postsynaptic functions remains controversial (Kay and Tóth, 2008; Paoletti et al., 2009; Tóth, 2011). Studies that have examined the physiological role of vesicular zinc mainly focused on its possible involvement in postsynaptic regulation (Vogt et al., 2000; Molnár and Nadler, 2001; Lopantsev et al., 2003; Mott et al., 2008; Besser et al., 2009), whereas its regulatory role in presynaptic release mechanisms remains unexplored.
Here we investigated the role of vesicular zinc in synaptic function at hippocampal MF terminals. Our data show that, in the absence of vesicular zinc, the release of a subpopulation of vesicles becomes slower. In addition, the release of these vesicles only becomes prevalent during intense synaptic activity. We propose that vesicular zinc has an important role in the dynamics of transmitter release and could therefore influence information coding in the hippocampal network.
Materials and Methods
Hippocampal slice preparation
ZnT3+/+ and ZnT3−/− mice (P16–P27 or 3–8 months, either sex) were anesthetized by isoflurane inhalation and decapitated. The brains were quickly removed, and horizontal slices (300 or 400 μm) were prepared in ice-cold solution containing the following (in mm): 75 NaCl, 25 NaHCO3, 1.25 NaH2PO4, 4 KCl, 25 glucose, 100 sucrose, 0.5 CaCl2, and 3 MgCl2, pH 7.4 (equilibrated with 95% O2 and 5% CO2). Sections were cut using a VT1000S microtome (Leica Microsystems), then transferred to a holding chamber containing normal ACSF at 34°C for 30 min, and subsequently stored at room temperature. All animal procedures were approved by the Animal Protection Committee of Laval University.
Unless stated otherwise, all recordings were performed with a physiological extracellular solution (ACSF) containing the following (in mm): 130 NaCl, 25 NaHCO3, 3.5 KCl, 1.25 NaH2PO4, 1.5 MgCl2, 2.5 CaCl2, and 10 glucose, pH 7.4 (equilibrated with 95% O2 and 5% CO2). Bicuculline methiodine (1 μm) was routinely added to the extracellular solution to isolate excitatory synaptic events.
For whole-cell recordings, pipettes were filled with a solution containing the following (in mm): 100 Cs-gluconate, 0.6 EGTA, 5 MgCl2, 8 NaCl, 2 MgATP, 0.3 NaGTP, 40 HEPES, and 1 QX-314, pH 7.3.
Synaptic responses were evoked by stimulation of the granule cell layer or the stratum lucidum (CA3c) via a constant-current isolation unit (A360; World Precision Instruments) connected to a bipolar tungsten electrode. Stimulation intensity that elicited 50–60% of maximum fiber volley was used for all experiments. For extracellular recordings, a glass pipette (4–6 MΩ resistance) filled with recording ACSF was placed in stratum lucidum (CA3b) to record MF field EPSPs (fEPSPs). Responses were recorded using a MultiClamp 700A amplifier (Molecular Devices) operating in the current-clamp mode and filtered at 3 kHz. At the end of each recording (2S,2′R,3′R)-2-(2′,3′-dicarboxycyclopropyl)glycine (DCG-IV) (1 μm), DNQX (20 μm), and TTX (1 μm) were sequentially applied to determine whether the signal recorded originated from MF terminals and to discriminate between a synaptic and nonsynaptic signal.
Recording patch pipettes were pulled from thin-walled borosilicate glass tubing (World Precision Instruments) and had resistance of 2–5 MΩ. Whole-cell patch-clamp recordings were made using an Axopatch 200B amplifier or a MultiClamp 700A amplifier (Molecular Devices) operating in the voltage-clamp mode. Data were acquired with pClamp8 or 9.2 software (Molecular Devices) at an acquisition rate of 20 kHz and filtered at 3 kHz. Recordings were made at room temperature (22–24°C) at a holding potential of −64 mV, unless mentioned otherwise. Uncompensated series resistance and input resistance (Ri) were monitored by the delivery of −10 mV voltage steps throughout the experiment, and recordings were discontinued after changes of >15%.
Synaptic responses for whole-cell recordings were evoked by low-intensity stimulation (100 μs duration) of the stratum lucidum (MF inputs) via a constant-current isolation unit (A360; World Precision Instruments) connected to a patch electrode filled with oxygenated extracellular solution referenced to a bath ground. For initial searching for fibers that elicited EPSCs, stimulation occurred at 20–110 μA at a frequency of 0.1 Hz. Once evoked EPSCs were detected, the stimulation frequency was reduced to 0.05 Hz.
Mice were injected intraperitoneally with 25 mg/kg kainic acid dissolved in 0.9% NaCl. The behavior of animals was monitored, and animals were killed after their first or second stage 5 seizure (Racine, 1972).
Bicuculline methiodine, CaEDTA, diethyldithiocarbamate (DEDTC) (Sigma-Aldrich), kainic acid (Ocean Produce), DCG-IV, DNQX (Tocris Bioscience), TTX (Alomone Labs), and EGTA-AM (AnaSpec) were used in the experiments. These reagents were prepared as stock solutions and stored as recommended. The chelators were diluted in properly oxygenated solutions at pH 7.4.
Animals used for the electron microscopy were deeply anesthetized and transcardially perfused first with a buffered sodium sulfide solution (12 g of Na2S·9H2O and 12 g of NaH2PO4·H2O in 1000 ml of distilled water, pH 7.4; 0.05 m) for 1 min, then with a buffered 3% glutaraldehyde solution in 0.12 m PBS, pH 7.4, for 20 min, and finally with the sodium sulfide solution again for 15 min. Brains were removed from the skull, postfixed in the 3% buffered glutaraldehyde solution for 2 h, and sectioned with a vibratome at 50 μm. Free-floating sections were washed with Tris buffer, pH 7.4, for 5 min periods to eliminate adsorbed phosphate ions, which would react with silver ions, causing an unwanted precipitation. Thereafter, sections were placed in the physical developer containing sodium tungstate as protective colloid, hydroquinone as reducing agent, sodium acetate and acetic acid to adjust the pH, and silver nitrate (for additional details, see Seress and Gallyas, 2000). The process of development was stopped by placing in the sections into 1% sodium thiosulfate for 1 min. Next, the sections were washed with Tris buffer for 5 min, then osmificated with 1% OsO4 for 1 h, dehydrated, and flat embedded in Durcupan according to a routine electron microscopic procedure. After microscopic examination, the area of interest was cut, reembedded, and thin sectioned. Thin sections were stained with uranyl acetate and lead citrate. A Tecnai electron microscope was used throughout analysis.
Quantification of zinc staining
Black particles indicating zinc were counted manually in standard-sized boxes on the micrographs.
Preparation of mossy fiber synaptosomes
Crude mossy fiber synaptosomes (MFSs) were prepared as described previously (Bancila et al., 2009). Briefly, mice were anesthetized by isoflurane inhalation and decapitated. Brains were rapidly removed and placed in ice-cold modified mammalian Krebs' medium containing the following (mm): 136 NaCl, 3 KCl, 1.2 MgCl2, 2.2 CaCl2, 16.2 NaHCO3, 5.5 glucose, and 1.2 Na/Na2 phosphate buffer, pH 7.4. The two hippocampi were dissected out, cut into smaller pieces, and placed in 300 μl of the medium. This preparation was gently homogenized by repeated pipetting (50–60 times) using a 200 μl tip until a homogenous suspension was obtained. This suspension was made up to a volume of 2 ml and passed through a nylon filter (100 μm mesh size). The filtrate was incubated in ice for 45 min, at the end of which a pellet (250 μl approximate volume) was formed by gravity sedimentation. The supernatant was removed without disturbing the pellet, and 250 μl of fresh, oxygenated medium at room temperature was added and incubated at room temperature for 10 min. In the appropriate cases, EGTA-AM (100 μm final concentration) was added during this incubation period to the samples. Protein estimation was done by Bradford assay.
Glutamate release assay
Glutamate release was estimated using a luciferase assay. The assay is based on a coupled reaction in which glutamate released from MFS produces nicotinamide adenine dinucleotide, which was monitored in “real-time” by a chemiluminescent reaction (Fosse et al., 1986; Helme-Guizon et al., 1998; Bancila et al., 2009) using a Packard Fusion microplate reader. MFS equivalent to 125 μg (∼5 μl) of total protein was resuspended to a final volume of 30 μl in the Krebs' buffer. To this, 20 μl of the enzymatic mix to produce the chemiluminescence was added, which was freshly prepared every time as follows: 50 μl of β-nicotinamide adenine dinucleotide coenzyme (16.6 mg dissolved in 2.5 ml of 0.2 m Tris buffer, pH 8), 5 μl of flavin mononucleotide (FMN; 1 mg in 8.3 ml of H2O), 20 μl of NADPH–FMN oxidoreductase (2.4 mg in 1 ml of H2O), 16 μl of bacterial luciferase (25 mg in 1 ml of H2O), and 10 μl of glutamate dehydrogenase (2.66 U/μl).
Synaptosomes were depolarized using either 25 mm KCl or 20 pm glutamate in the presence of 0.5 mm calcium to trigger glutamate release. Luminescence from each sample was read for 0.5 s without any time delay between samples.
All chemicals were purchased from Sigma-Aldrich except luciferase, NADPH–FMN oxidoreductase, and glutamate dehydrogenase, which were purchased from Roche Diagnostics. For protein estimation, Coomassie plus reagent was purchased from Pierce (Thermo Fisher Scientific), and quantifications were performed using an Eppendorf BioPhotometer plus UV/Vis Photometer. All experiments were performed by a blinded investigator.
Synaptic events were analyzed offline using Clampfit 9.2 software (Molecular Devices). A template search was used for event detection, and all single events were visually inspected. Templates were created using a minimum of 10 events aligned by their rising edges.
For the cumulative probability plots, equal number of events were used from each cell (n = 250 for sEPSCs; n = 50 for mEPSCs), and the Kolmogorov–Smirnov test was used to determine whether the event amplitudes from various conditions were significantly different.
Decay time constants of evoked EPSCs were measured from individual and 5–15 averaged traces using Clampfit 9.2 software (Molecular Devices). Traces with spontaneous activity within 75 ms of maximal amplitude were discarded. Briefly, the decay of evoked events was fitted with a single- or double-exponential function using the Chebyshev algorithm. An event was considered to be fitted by a single exponential if the correlation value was <2% of the one obtained when fitted with a double exponential. Unless stated otherwise, data are presented as mean ± SEM.
Amplitude distribution of spontaneous (s) and miniature (m) EPSCs recorded from CA3 pyramidal cells are known to be skewed toward larger values (Haug, 1967; Jonas et al., 1993) (Fig. 1B,C). The vast majority of the large-amplitude events were shown to originate from MFs (Henze et al., 1997). Although the prominent skewed distribution of sEPSCs is generally thought to result from action-potential-mediated synchronization of different release sites, explanation of similar distribution pattern of mEPSCs is more challenging. Calcium release from intracellular stores underlies large-amplitude mIPSCs and presynaptic receptor-activated mEPSCs (Llano et al., 2000; Sharma and Vijayaraghavan, 2003; Gordon and Bains, 2005; Sharma et al., 2008). However, it is not known how large-amplitude mEPSCs are generated under control conditions. Unlike mIPSCs (Llano et al., 2000), large mEPSCs recorded from CA3 pyramidal cells are insensitive to manipulations of intracellular and extracellular Ca2+ (Henze et al., 2002). Alternatively, the skew in miniature amplitude distribution can result from variations in vesicular glutamate content (Bruns et al., 2000; Karunanithi et al., 2002). If this is the case, cube root transformation of amplitude or charge measurements should yield a distribution pattern similar to the distribution pattern of vesicle radius measurements. Cube root transformation of mEPSCs amplitude recorded from CA3 pyramidal cells showed a skewed distribution (Fig. 1D), although distribution of vesicle radius was Gaussian (Fig. 1F) (R2 = 0.96, n = 255). These data indicate that variation in vesicle size is not sufficient to explain the mechanism by which larger-amplitude mEPSCs are generated by MFs. Therefore, we tested whether other native features of MF terminals could influence the generation of large-amplitude events. MFs are known to have zinc sequestered in the presynaptic vesicles, but the exact physiological role of vesicular zinc is unclear (Kay and Tóth, 2008; Paoletti et al., 2009). Our goal was to determine whether vesicular zinc can influence synaptic transmission at hippocampal MFs by modulating glutamate release.
Vesicular zinc influences sEPSCs and mEPSCs recorded in CA3 pyramidal cells
To determine the importance of zinc in neurotransmitter release on basal synaptic activity, we compared sEPSCs recorded in CA3 pyramidal cells in wild-type (WT) and ZnT3 knock-out (KO) animals that cannot load zinc into synaptic vesicles (Cole et al., 1999). A significant shift in amplitude distribution was observed in KO mice. When we compared the cumulative probability of amplitude distributions, a significant change was observed between KO and WT with amplitudes in KO mice shifted toward smaller values (Fig. 2A). The shift in amplitude distribution was attributable to the diminished number of larger-amplitude events observed in the absence of vesicular zinc. We also used an alternative method to eliminate zinc from the vesicles to confirm that the lack of vesicular zinc in KO animals is indeed behind the observed changes. We used the membrane-permeable zinc chelator DEDTC to eliminate chelatable zinc from acute ZnT3 WT slices. Slices were perfused with DEDTC (200 μm) for 15 min. This treatment also produced a significant change in the amplitude distribution of sEPSCs (Fig. 2B) similar to the one obtained with the comparison of ZnT3 WT and KO animals. The observed difference in amplitude distribution is most prominent among events >100 pA. Next, we aimed to determine whether the kinetic properties of these larger events are influenced by the absence of zinc. Kinetic analysis of large (>100 pA) amplitude events showed that large sEPSCs in ZnT3 KO animals have significantly slower decay kinetics (6.51 ± 0.28 vs 8.68 ± 0.22 ms; n = 7, p < 0.001) and rise time (0.71 ± 0.06 vs 1.03 ± 0.06 ms; n = 7, p < 0.001). In addition, although the amplitude and decay time constant of events recorded from WT animals did not show any correlation, these two parameters showed an increased correlation in ZnT3 KOs (Fig. 2C). In contrast, when we compared the properties of small-amplitude (20–40 pA) events, no significant difference was observed between WT and KO animals (Fig. 2D).
Next, we recorded mEPSCs from CA3 pyramidal cells in the presence of TTX (1 μm) in the recording solution from WT and KO animals to determine whether action-potential-independent transmitter release is influenced by the presence of vesicular zinc. Similarly to sEPSCs, the amplitude distribution of mEPSCs is shifted toward smaller events when vesicular zinc is absent in either ZnT3 KO animals (Fig. 3A) or DEDTC-treated slices (Fig. 3B). Although the rise time of events >50 pA are not different between WT and KO (0.92 ± 0.10 vs 0.94 ± 0.07 ms; n = 5), the decay time is significantly slower in KO animals (WT, 9.03 ± 0.48 ms; KO, 10.3 ± 0.54 ms; n = 5, p < 0.01) (Fig. 3C).
MF short- and long-term plasticity are maintained in ZnT3 KO mice
Changes in mEPSC frequency are generally thought to result from an altered probability of release. If the absence of vesicular zinc leads to a diminished release probability at hippocampal MFs, we would expect to see robust changes in short- and long-term plasticity. The role of vesicular zinc in synaptic plasticity was also supported by a recent study demonstrating a zinc-mediated transactivation of TrKB receptors (Huang et al., 2008). Therefore, we next compared the short-term plasticity properties of MF-evoked EPSCs in WT and KO mice. Field recordings from CA3 synapses of adult ZnT3 KO mice show no difference in the basal stimulus–response relationships for afferent fiber volleys or fEPSP amplitudes (Fig. 4A,B). The excitability of the CA3 region was investigated using three stimulus paradigms designed to probe presynaptic function. Frequency facilitation was assessed using stimulations of increasing frequency from 0.05 to 20 Hz with 20 pulses per frequency. Paired-pulse ratios were probed by applying two stimuli (10 pairs) at intervals ranging from 2 s to 50 ms (0.5–20 Hz). Finally, a train of five pulses was delivered at 25 Hz. The absence of vesicular zinc did not affect the paired-pulse measures, frequency facilitation, or the train stimuli (Fig. 4C,E). These results suggest that vesicular zinc does not play a significant role in short-term plasticity at MF synapses.
Contradictory data has been reported as to the requirement for vesicular zinc for the NMDAR-independent presynaptic form of LTP expressed at MF–CA3 synapses (Paoletti et al., 2009). Here we evaluated the role of vesicular zinc by comparing LTP evoked with high-frequency stimulation (HFS) (3 × 100 Hz, 10 s interval) of MFs between slices from ZnT3 WT and KO mice. KO mice showed substantial posttetanic potentiation that is significantly higher than controls (KO, 596.2 ± 106.8%, n = 7; WT, 368.7 ± 25.1%, n = 8; p = 0.03) but attained similar fEPSP values within 2 min after HFS. Moreover, LTP recorded from KO mice (161.3 ± 11.8%, n = 7) was not significantly different when compared with WTs (168.8 ± 14.4%, n = 8; Fig. 4F,G). A 10 min bath application of DCG-IV (1 μm) effectively decreased the response from both genotypes (WT, 51.8 ± 5.15%; KO, 58.6 ± 6.01%), indicating that it was generated by MFs. Considered together, our findings indicate that vesicular zinc does not play a critical role in the maintenance of long-term potentiation at MF synapses nor is vesicular zinc necessary to observe short-term frequency facilitation of MF release.
The slow membrane-permeant Ca2+ chelator EGTA-AM selectively affects ZnT3 KO MF-evoked response
Our data show that, whereas the lack in ZnT3 does not modify short- or long-term synaptic plasticity at MF/CA3 synapses, spontaneous release is influenced by the presence of zinc in glutamatergic vesicles. Spontaneous and evoked releases have been shown to use distinct pools of vesicles (Sara et al., 2005; Fredj and Burrone, 2009; but see Prange and Murphy, 1999; Groemer and Klingauf, 2007) and be differentially regulated by presynaptic calcium dynamics (Glitsch, 2006). Next, we examined the relationship between the dynamics of intraterminal calcium rise and the presence of the ZnT3/zinc in the vesicles. Fast synchronous release is known to be triggered by a rapid Ca2+ rise in presynaptic terminals (Bischofberger et al., 2002), and the slow calcium chelators EGTA/EGTA-AM were successfully used to alter the peak current of evoked responses in a concentration-dependent manner (Borst and Sakmann, 1996). Although a high concentration of EGTA/EGTA-AM attenuates the synchronous component of the evoked events (Castillo et al., 1996; Salin et al., 1996), lower concentrations of this chelator can be used to effectively differentiate between synchronous and asynchronous release (Chen and Regehr, 1999; Maximov and Südhof, 2005; Iremonger and Bains, 2007). We made whole-cell patch-clamp recordings from CA3 pyramidal cells and applied EGTA-AM while evoking MF synaptic events in WT and KO animals. MFs were stimulated at three different frequencies (Fig. 5A1,A2). The baseline was set as 0.05 Hz (10 pulses), after which stimulation frequency was first increased to 0.2 Hz (20 pulses) and then to 1 Hz (30 pulses). No significant difference was observed for basal response amplitudes (WT, 47.2 ± 6.68 pA; KO, 41.7 ± 8.48 pA) or for frequency facilitation capacity between the two genotypes (0.2 Hz, 78.1 ± 11.7 vs 70.0 ± 17.9 pA; 1 Hz, 182.2 ± 31.3 vs 239.2 ± 38.3 pA, WT and KO, respectively). This experimental paradigm was then repeated after a 10 min bath application of EGTA-AM (100 μm). Although EGTA-AM treatment did not significantly affect the amplitude of EPSCs evoked in ZnT3 WT mice, ZnT3 KO mice showed a selective sensitivity to EGTA-AM (Fig. 5A1,A2). In ZnT3 KO mice, application of the calcium chelator decreased the baseline value by ∼45%, and this effect became more prominent as the stimulation frequency increased (>55% for 0.2 and 1 Hz). The decrease observed in KO mice could partially be explained by an increase in failure rate. However, we did not observe a significant difference in failure rate at any frequency, not even at 1 Hz at which the EGTA-AM effect is the most striking (data not shown). There is also no difference in failure rates between genotypes (0.05 Hz, 31.3 ± 11.0 vs 41.1 ± 10.8%; 0.2 Hz, 30.0 ± 10.9 vs 32.8 ± 9.43%; 1 Hz, 12.9 ± 6.47 vs 17.0 ± 8.75%, WT and KO, respectively).
The selective effect of EGTA-AM on ZnT3 KO mice raises the question whether the observed changes are caused by the absence of vesicular zinc or by the absence of the zinc transporter. To differentiate between these two possibilities, we investigated whether chemical zinc chelation has similar effects to the genetic deletion of ZnT3. Slices were acutely perfused with DEDTC (200 μm) for at least 15 min. The amplitude of EPSCs evoked in DEDTC-treated slices had values similar to those obtained in KO mice (0.05 Hz, 43.0 ± 14.0 pA; 0.2 Hz, 90.4 ± 21.6 pA; 1 Hz, 260.7 ± 49.4 pA) (Fig. 5B2). EGTA-AM treatment did not modify the amplitude of EPSCs evoked at 0.05 Hz, but it decreased the amplitude of the responses by ∼47% at 0.2 Hz and by ∼55% at 1 Hz. Similarly to KO mice, application of EGTA-AM did not significantly change failure rates (data not shown). These data show that both chemical zinc chelation and genetic elimination of the vesicular zinc transporter renders MF EPSCs evoked at higher frequencies EGTA-AM sensitive.
Our recordings from ZnT3 WT mice are in agreement with previous findings showing only a modest impact on synchronous release by EGTA-AM (Salin et al., 1996). In contrast, EPSCs evoked in KO mice and DEDTC-treated slices were severely attenuated in the presence of the slow calcium chelator. These data indicate that some vesicles recruited by increased synaptic activity in mice lacking vesicular zinc are selectively sensitive to slow calcium chelators.
ZnT3 KO mice exhibit prolonged MF EPSC decay kinetics
Sensitivity to EGTA-AM could indicate that presynaptic Ca2+ channels are not as tightly coupled to Ca2+ sensors in the absence of ZnT3. If this is the case, we would expect to see slower decay time constants in ZnT3 KO mice. Therefore, we compared the decay kinetics (τ values) of EPSCs evoked from WT and KO animals. In control conditions, WT-evoked events at different frequencies could all be fitted with a single exponential. In contrast, although KO events could almost all be fitted with a single exponential at 0.05 Hz, a slower component emerged at higher stimulation frequencies and events were better fit with two exponentials (Fig. 6A1,A2). The slower component was prominent at 1 Hz at which 55.6% of the cells were better fit using a biexponential function.
When evoked events decayed monoexponentially, ZnT3 WT and KO mice exhibited similar decay time constants (Fig. 6B) and showed no change in their decay kinetics after the application of EGTA-AM. Events that decayed biexponentially presented a higher sensitivity to EGTA-AM, especially those recorded from KO mice. Indeed, although approximately half of KO events in control condition could be fitted with two exponentials, EGTA-AM treatment eliminated the slower component in a high number of the recordings (Fig. 6A3).
Although the lack of ZnT3 causes a change in decay kinetics, it does not influence other kinetic properties, such as the rise time of evoked EPSCs. Likewise, EGTA-AM application did not change the rise time of EPSCs recorded from slices prepared from WT or KO animals at any stimulation frequency tested (data not shown).
Glutamate release is EGTA-AM sensitive in ZnT3 KO animals
To provide direct evidence that in our experiments the observed EGTA-AM effect is the result of the altered glutamate release, we used isolated synaptosomes and measured glutamate release directly with a chemiluminescence assay. We compared the effect of EGTA-AM on depolarization-induced and glutamate-induced glutamate release from hippocampus MFSs. We added 25 mm KCl to induce depolarization (Zoccarato et al., 1999) and 20 pm glutamate to activate presynaptic receptors on MF terminals (Schmitz et al., 2001; Rebola et al., 2008; but see Kwon and Castillo, 2008). Addition of KCl to the synaptosomes lead to large glutamate release, whereas 20 pm glutamate evokes significantly smaller responses. Addition of 1 or 10 nm glutamate to the buffer in the absence of synaptosomes was not detectable under our experimental conditions (Fig. 7A). Total glutamate release evoked either with KCl or glutamate from synaptosomes was not significantly altered by EGTA-AM in WT animals (KCl, 28.2 ± 3.28 vs 20.5 ± 3.96 ng/μg; glutamate, 7.07 ± 0.90 vs 6.10 ± 1.13 ng/μg). In contrast, EGTA-AM has significantly reduced glutamate release from ZnT3 KO synaptosomes (KCl, 24.9 ± 2.70 vs 17.4 ± 0.61 ng/μg; glutamate, 8.08 ± 0.79 vs 5.28 ± 0.65 ng/μg) (Fig. 7B,C). These results demonstrate that glutamate release from MF terminals is governed by different calcium dynamics in ZnT3 KO animals.
Only a subpopulation of vesicles is zinc positive
Based on the complex effect of EGTA-AM on MF EPSCs in ZnT3 KO animals, we propose that two functionally distinct vesicle populations exist in the presynaptic terminal: one that expresses ZnT3 and hence contains vesicular zinc, and the other without ZnT3 and intravesicular zinc. We tested this hypothesis by directly visualizing zinc in individual vesicles at the electron microscopic level and to determine whether all vesicles are stained or only a subpopulation is labeled. We used a modified Timm's method (Seress and Gallyas, 2000) in which the silver end product labeling zinc has a slightly smaller diameter than a synaptic vesicle. We calculated the percentage of vesicles labeled with Timm's staining (Fig. 8A) and found that only a small population of the vesicles showed positive staining (16.7 ± 3.4%; 30 terminals from three adult rats). To ensure that this low level of staining did not reflect suboptimal development time during the staining protocol, we allowed for the chemical reaction responsible for silver crystal formation to proceed longer. If this is an important factor, increasing development times would lead to more and more positively labeled vesicles in the presynaptic terminals. In contrast, if this parameter does not significantly influence the number of positively labeled elements, the size of the individual crystals would increase with longer development times, but the percentage of labeled vesicles would not. We found that the percentage of positively labeled vesicles was not significantly different in samples in which we developed the electron-dense silver end products for 12, 18, or 21 min (15.6 ± 2.5, 17.5 ± 2.1, and 17.8 ± 3.2%, respectively; n = 15 at each conditions). Moreover, these values were not significantly different from the analysis detailed above, in which we used 15 min for this reaction (p < 0.05). These results confirm that MF terminals contain molecularly diverse subpopulations of vesicles, providing an anatomical substrate for the observed differential zinc sensitivity of presynaptic function. This heterogeneity results from the fact that, in a single presynaptic terminal, not all vesicles are composed of the same set of vesicle proteins, which in turn leads to a diverse vesicular content. Next, we investigated whether vesicle volume is affected by the presence of zinc. We compared the size of the synaptic vesicles in WT and ZnT3 KO animals. In WT animals, MF terminals contain both zinc-positive and zinc-negative vesicles, whereas in the ZnT3 KO animals, all vesicles are zinc negative. Our data show that the distribution of vesicle size was not significantly different between the two groups (diameter, 28.6 ± 2.8 vs 27.6 ± 2.5 nm, n = 250), indicating that vesicular zinc does not play a role in the determination of vesicle volume (Fig. 8C).
Increased synaptic activity alters zinc distribution in MF boutons
The slow calcium chelator EGTA-AM drastically attenuated the amplitudes of events evoked at higher frequencies in ZnT3 KO animals but not in WT animals. This could result from the preferential release of zinc-containing vesicles at higher presynaptic firing rates in WT animals. In this case, high-intensity synaptic activity is expected to change the composition of the synaptic vesicle pool such that a smaller percentage of zinc-containing vesicles would remain in the terminal after a bout of high-frequency activity. We investigated this possibility by comparing the percentage of zinc-labeled vesicles in control and kainate-injected animals. After kainate injection, animals had seizures and were killed after their second major seizure. We found that the ratio of zinc-positive vesicles was significantly smaller (8.4 ± 1.6%) in kainate-injected animals when compared with control subjects (22.5 ± 6.3%; n = 5, p < 0.01).
AP3 is responsible for the recruitment of ZnT3 into synaptic vesicles (Salazar et al., 2004a). Vesicles containing AP3 are recycled via bulk endocytosis (Voglmaier et al., 2006). Based on this, we predict that, as activity increases, more and more zinc-containing vesicles would go through bulk endocytosis. Thus, we quantitatively assessed the distribution of zinc in various structural elements of the presynaptic terminals in control and kainate-injected animals. In control animals, the majority of the staining was confined to synaptic vesicles with an average diameter of ∼30 nm (Fig. 9A,B). In contrast, kainate-injected animals showed a significantly lower percentage of labeled vesicles, and zinc staining was apparent in larger endosome-like structures (Fig. 9C,D). This morphological evidence demonstrating a shift in the localization of zinc after high-intensity activity suggests that our conclusion regarding the preferential recycling of zinc-containing vesicles via bulk endocytosis involving endosomes.
In the present study, we show that zinc is only present in a subpopulation of vesicles and that these zinc-containing vesicles are preferentially released during increased synaptic activity. Although vesicular zinc minimally affects synaptic plasticity, it significantly influences the dynamics of spontaneous and evoked release. MF-evoked EPSCs show a slower component that becomes prominent as stimulation frequency increases. The selective sensitivity of ZnT3 KO mice to slow calcium chelator suggests an alteration of their Ca2+-dependent release mechanisms.
Regulation of neurotransmitter content and release of synaptic vesicles
Our data indicate that the genetic deletion of ZnT3 leads to the selective attenuation of sEPSCs and mEPSCs of large amplitude in CA3 pyramidal cells. Modification of several presynaptic mechanisms could underlie the observed differences between WT and KO animals. First, zinc-containing vesicles could have higher glutamate concentration (Liu, 2003). Uptake of glutamate into synaptic vesicles is mediated by vesicular glutamate transporters (VGLUT). VGLUT1 is expressed in hippocampal MFs and is co-localized with VGLUT2 during development (Herzog et al., 2006). Quantal size and vesicular glutamate load is influenced by several factors. A positive correlation has been established between the number of transporter molecules per vesicle and quantal size (Wojcik et al., 2004; Wilson et al., 2005). Cl− content of endocytosed synaptic vesicles could also be a major determinant of glutamate load (Schenck et al., 2009). ZnT3 and VGLUT1 are cotargeted to the same vesicle population and can reciprocally regulate their transport mechanisms. The presence of ZnT3 on the vesicles increases the vesicular uptake of glutamate in a zinc-dependent manner (Salazar et al., 2005). Therefore, vesicular zinc might regulate the neurotransmitter content of ZnT3-containing vesicles. Second, vesicular quantal content could also be influenced by vesicle size. It has been proposed that giant EPSCs are generated by monovesicular release of large glutamatergic vesicles (Henze et al., 2002; Hallermann et al., 2003). However, our data demonstrating that the diameter of synaptic vesicles is identical in both genotypes strongly suggest that vesicular zinc does not play a critical role in the determination of vesicle volume. Third, our findings could indicate that synchronization of multiple active zones within the presynaptic terminal is altered in KO mice. Presynaptic calcium stores have been shown to mediate synchronous release from several release sites, leading to large miniature events (Llano et al., 2000). It has also been shown that store-derived calcium is the trigger for a large fraction of mEPSCs (Emptage et al., 2001). Therefore, the release machinery of MF vesicles lacking zinc could have a lower affinity for Ca2+ that would decrease their release capacity. In our experiments, the amplitude of events evoked at higher frequencies was attenuated by EGTA-AM in KO mice, whereas responses from WT animals were unaltered. These data indicate that, in the absence of vesicular zinc, certain vesicles become EGTA sensitive, suggesting altered calcium cooperativity. Diverse mechanisms downstream of the calcium influx could underlie this change in KO mice, such as a looser spatial coupling between the calcium channels and the vesicles, or a different calcium sensitivity of their release machinery (Augustine, 2001). Neurotransmitter release relies on a source of calcium and its sensors (Sudhof, 2004). Classical and recent works provide evidence for a close spatial proximity (few nanometers) of calcium channels and vesicular release sites (Adler et al., 1991; Stanley, 1991; Bucurenciu et al., 2008; Wang et al., 2008). This tight coupling has been shown to increase the efficacy, speed, and temporal precision of transmitter release (Fedchyshyn and Wang, 2005; Bucurenciu et al., 2008). In the presynaptic terminal, calcium signals are detected by calcium-sensing proteins, and synchronous release in excitatory synapses of the hippocampus is triggered by Ca2+ binding to synaptotagmin 1 (Geppert et al., 1994; Fernández-Chacón et al., 2001). Whether all glutamatergic vesicles in the MF boutons have the same Ca2+ sensor is still an unresolved question.
Synaptic plasticity and zinc
The role of endogenous zinc in various properties of synaptic plasticity has been studied previously in different animal models. Some studies showed that chelation of extracellular zinc or decrease in vesicular zinc (mocha mice) did not affect MF paired-pulse and frequency facilitation, which is in agreement with our current findings (Vogt et al., 2000; Lopantsev et al., 2003). Results from our experiments using a wide-range of stimulus protocols indicate that MF short-term plasticity is unaltered in the complete absence of vesicular zinc. These results reinforce previous reports stating that zinc is not responsible for the unusual short-term physiological properties of the MF synapses.
Another unique physiological property of MF/CA3 synapses is their NMDA-independent form of LTP. Several studies have used bath application of different zinc chelators (membrane impermeable or permeable) to assess the physiological role of zinc in this phenomenon (Lu et al., 2000; Li et al., 2001; Huang et al., 2008). Synaptically released zinc has been proposed to be necessary and sufficient for the induction of MF LTP (Li et al., 2001). In our study, LTP was unaltered in the complete absence of vesicular zinc. Recently, it has been proposed that, with the elimination of vesicular zinc, presynaptic LTP is diminished while a postsynaptic form of MF LTP is unmasked (Pan et al., 2011). The combination of these two opposing effects could lead to the zero net difference we observed between the facilitation capacity of WT and ZnT3 KO animals.
AP3-dependent bulk endocytosis
Synaptic vesicle recycling is required to sustain the fidelity of transmission as well as to maintain the integrity of the presynaptic membrane. Multiple mechanisms of recycling have been proposed, including the formation of deep invaginations of plasma membrane or biogenesis from an endosomal intermediate. Unlike classic clathrin-mediated endocytosis (CME), vesicles budding from these endosomes require AP-3 (Faúndez et al., 1998; Shi et al., 1998; Faúndez and Kelly, 2000). Different synaptic terminals show endosome-like structures after strong synaptic stimulation, suggesting that this form of endocytosis is an important mechanism to recover exocytosed vesicles (Heuser and Reese, 1973; Koenig and Ikeda, 1996; de Lange et al., 2003; Paillart et al., 2003; Clayton et al., 2008). Although synaptic vesicle recycling is thought to be mainly mediated through kiss-and-run and CME at hippocampal terminals (Klingauf et al., 1998; Pyle et al., 2000; Aravanis et al., 2003; Gandhi and Stevens, 2003; Zhu et al., 2009), our data support the existence of bulk endocytosis at MF terminals. The kinetic evidence of this form of recycling was elegantly demonstrated in the calyx of Held using capacitance measurements. Bulk endocytosis is a frequency-dependent form of recycling, as capacitance shifts increase with stimulation intensity (Wu and Wu, 2007). Our data demonstrating a change in the subcellular localization of chelatable zinc from synaptic vesicles to endosome-like structures after increased synaptic activity suggest the preferential recycling of zinc-containing vesicles through this pathway.
Functionally distinct vesicle pools
Synaptic vesicles are organized into functionally distinct pools based on their probability of release during neuronal activity. The readily releasable pool and the reserve pool recycle in response to neuronal activity, whereas the resting pool is defined as a pool that is not stained with FM dyes during stimulation. Spontaneous and evoked release is maintained by vesicles belonging to distinct vesicle pools, underlining the functional importance of these separate pools in the presynaptic terminal (Sara et al., 2005; Fredj and Burrone, 2009; but see Prange and Murphy, 1999; Groemer and Klingauf, 2007; Hua et al., 2010). Our study did not address the identity of pools for which the distinct types of vesicles (e.g., zinc-containing and zinc-negative vesicles) belong. Rather we show how the molecular composition of vesicles contributes to their functional properties and how various recycling pathways could allow the generation of a diverse pool. Adaptor proteins could play a key role in the generation of a molecularly diverse vesicle pool. Recruitment of different sets of proteins into distinct vesicle pools is possible by the involvement of various adaptor proteins in different recycling pathways. In the mocha mouse mutant, AP3, which is associated with bulk endocytosis, is missing. In these animals, ZnT3 levels are drastically reduced, indicating that AP3 is responsible for the recruitment of the zinc transporter into vesicles that are generated via bulk endocytosis (Kantheti et al., 1998; Salazar et al., 2004a). mocha mice also lack the tetanus neurotoxin-insensitive vesicle-associated membrane protein (TI-VAMP or VAMP7) and show the loss of asynchronous evoked release at hippocampal MFs (Scheuber et al., 2006). These studies demonstrate that AP3 is responsible for the recruitment of ZnT3 and VAMP7 into a subset of vesicles and therefore very likely that ZnT3 and VAMP7 are expressed on the same population of synaptic vesicles. VAMP7 expression only in a subpopulation of synaptic vesicles was elegantly demonstrated using VAMP7–HRP constructs (Hua et al., 2011). When compared with VGLUT1–pHluorin-labeled vesicles, VAMP7–pHluorin-labeled vesicles showed a significantly smaller rate of exocytosis (Hua et al., 2011); this finding is in good agreement with our data showing that zinc-containing vesicles are preferentially released during higher-intensity stimulation. In contrast, VAMP7-positive vesicles undergo a higher rate of spontaneous release than VGLUT1-containing vesicles (Hua et al., 2011). Similarly, our data show that spontaneous release is significantly reduced in ZnT3 KO animals. These similarities between release dynamics of VAMP7-containing vesicles and our current findings regarding the behavior of zinc-containing vesicles during various physiological stimuli further strengthen the possibility that ZnT3 and VAMP7 is expressed on the same vesicle pool that is generated via bulk endocytosis. Reserve pool vesicles were shown to be reluctant to release, and large endosomal intermediates of bulk endocytosis were demonstrated after high-intensity stimuli (Takei et al., 1996; de Lange et al., 2003). VAMP7 is expressed in both recycling and resting pools with higher levels in the resting pool (Hua et al., 2011). Therefore, it is possible that zinc-containing vesicles predominantly populate the reserve pool. Additional studies will be needed to answer this question unequivocally.
We used modified Timm's staining to visualize zinc in synaptic vesicles; we found that only a subpopulation of vesicles is stained with this technique. We also altered the development time of the silver end product and found that, with longer development times, only the size of the individual silver granules increased but not their overall number, suggesting that only 15–20% of the vesicles contain zinc. However, it is possible that we underestimate the number of zinc-positive vesicles because of the following reasons: (1) during perfusion, the Na2S-containing solution might not sufficiently saturate the entire tissue, and (2) during the staining process, silver intensification and conversion of zinc sulfide to silver sulfide might not occur in vesicles containing smaller amounts of zinc. This possibility is supported by the seemingly uniform staining produced by ZnT3 immunocytochemistry (Cole et al., 1999). However, in these experiments, diaminobenzidine, a diffuse end product was used as chromogen, making visualization of individual vesicles unlikely. Postembedding immunocytochemistry using gold particles could be used to visualize ZnT3 content of individual vesicles, but this technique will most likely underestimate the number of ZnT3-positive vesicles because of the unpredictable antibody-binding ratio by antigens. Future experiments will need to determine whether every vesicle contains ZnT3 and/or whether every expressed copy is functional.
In summary, our data suggest a novel perspective for vesicle heterogeneity. At a single terminal, vesicles differing in their protein composition could respond to different physiological stimuli. Moreover, this physiological heterogeneity could involve the generation of vesicles via separate recycling pathways.
This work was supported by Canadian Institutes of Health Research (CIHR) Operating Grants MOP-81142 (K.T.) and MOP-82718 (L.P.). N.L. was supported by Natural Sciences and Engineering Research Council of Canada and Centre de recherche sur le cerveau, le comportement et la neuropsychiatrie CRCN scholarships, and D.V.J. was a holder of a CIHR Banting Scholarship. We thank Philippe Lemieux for expert technical assistance, as well as Drs. Alan Kay (University of Iowa, Iowa City, IA), Kenneth A. Pelkey (National Institutes of Health/National Institute of Child Health and Human Development, Rockville, MD), and Jaideep S. Bains (University of Calgary, Calgary, AB, Canada) for the stimulating scientific discussions and critical reading of this manuscript.
- Correspondence should be addressed to Dr. Katalin Tóth, Robert Giffard Research Center of Laval University, 2601 chemin de la Canardière, Québec, QC, Canada, G1J 2G3.