alpha-Latrotoxin, acting via two Ca2+-dependent pathways, triggers exocytosis of two pools of synaptic vesicles.

alpha-Latrotoxin stimulates three types of [(3)H]gamma-aminobutyric acid and [(14)C]glutamate release from synaptosomes. The Ca(2+)-independent component (i) is insensitive to SNAP-25 cleavage or depletion of vesicle contents by bafilomycin A1 and represents transmitter efflux mediated by alpha-latrotoxin pores. Two other components of release are Ca(2+)-dependent and vesicular but rely on distinct mechanisms. The fast receptor-mediated pathway (ii) involves intracellular Ca(2+) stores and acts upon sucrose-sensitive readily releasable vesicles; this mechanism is insensitive to inhibition of phosphatidylinositol 4-kinase (PI 4-kinase). The delayed pore-dependent exocytotic component (iii) is stimulated by Ca(2+) entering through alpha-latrotoxin pores; it requires PI 4-kinase and occurs mainly from depot vesicles. Lanthanum perturbs alpha-latrotoxin pores and blocks the two pore-mediated components (i, iii) but not the receptor-mediated release (ii). alpha-Latrotoxin mutant (LTX(N4C)) cannot form pores and stimulates only the Ca(2+)-dependent receptor-mediated amino acid exocytosis (ii) (detectable biochemically and electrophysiologically). These findings explain experimental data obtained by different laboratories and implicate the toxin receptors in the regulation of the readily releasable pool of synaptic vesicles. Our results also suggest that, similar to noradrenergic vesicles, amino acid-containing vesicles at some point in their cycle require PI 4-kinase.


␣-Latrotoxin stimulates three types of [ 3 H]␥-aminobutyric acid and [ 14 C]glutamate release from synaptosomes. The Ca 2؉ -independent component (i) is insensitive to SNAP-25 cleavage or depletion of vesicle contents by bafilomycin A1 and represents transmitter efflux mediated by ␣-latrotoxin pores. Two other components of release are Ca 2؉ -dependent and vesicular but rely on distinct mechanisms. The fast receptor-mediated pathway (ii) involves intracellular Ca 2؉ stores and acts upon sucrose-sensitive readily releasable vesicles; this mechanism is insensitive to inhibition of phosphatidylinositol 4-kinase (PI 4-kinase). The delayed pore-dependent exocytotic component (iii) is stimulated by Ca 2؉ entering through ␣-latrotoxin pores; it requires PI 4-kinase and occurs mainly from depot vesicles. Lanthanum per-
turbs ␣-latrotoxin pores and blocks the two pore-mediated components (i, iii) but not the receptor-mediated release (ii). ␣-Latrotoxin mutant (LTX N4C ) cannot form pores and stimulates only the Ca 2؉ -dependent receptormediated amino acid exocytosis (ii) (detectable biochemically and electrophysiologically). These findings explain experimental data obtained by different laboratories and implicate the toxin receptors in the regulation of the readily releasable pool of synaptic vesicles. Our results also suggest that, similar to noradrenergic vesicles, amino acid-containing vesicles at some point in their cycle require PI 4-kinase.
As we demonstrated recently, the toxin forms pores by assembling into cyclical tetramers, which insert into lipid bilayers (17). By using a LTX preparation that formed tetramers/ pores only in the presence of Mg 2ϩ , we selectively studied the receptor-mediated LTX action in synaptosomes (8,18,19) and found that LTX-receptor interaction in the absence of pore formation causes norepinephrine (NE) exocytosis that requires both extracellular and stored Ca 2ϩ . On the other hand, release of amino acid (AA) transmitters, which mediate fast neurotransmission in the brain, could involve other mechanisms.
Therefore, the current work had two main objectives: to investigate the mechanism of LTX-induced release of L-glutamic acid (Glu) and ␥-aminobutyric acid (GABA) and to distinguish the pore-and receptor-dependent LTX actions. To avoid the toxin pore formation, we employed La 3ϩ (which disrupts toxin tetramers) and a mutant LTX (which is unable to form such tetramers). The use of Botulinum neurotoxin type E (BoNT/E) and bafilomycin A1 (BAF) allowed us to identify true SV exocytosis. We demonstrate that, in the absence of Ca 2ϩ , LTX causes predominantly nonvesicular AA efflux. Toxinevoked vesicular AA release is detectable biochemically only in the presence of Ca 2ϩ and is based on two pharmacologically and temporally distinct mechanisms: (i) fast receptor-mediated phase that acts upon readily releasable SVs and (ii) delayed pore-mediated Ca 2ϩ entry that stimulates depot vesicles. Our study confirms the differential regulation of the two vesicular pools and demonstrates the usefulness of mutant LTX for indepth studies of AA release.
Amino Acid Release from Synaptosomes-Rat cerebrocortical synaptosomes, prepared as described (20), were suspended in physiological buffer (PB, containing, in mM: 125 NaCl, 5 KCl, 1 MgCl 2 , 10 glucose, 0.1 aminooxyacetic acid, and 25 HEPES, pH 7.4) supplemented with 2 mM CaCl 2 and kept on ice until required. All buffers and suspensions were constantly oxygenated. In the short loading procedure, synaptosomes were incubated with 2-5 Ci/ml [ 3 H]GABA or [ 14 C]Glu at 37°C for 5 min, washed, and stimulated 20 min later to reproduce conditions used by others (21). In the normal loading protocol, terminals were incubated for 5 min with [ 14 C]Glu and for 60 min with [ 3 H]GABA, washed, and incubated for 1 h at 37°C in PB containing 0.1 mM EGTA (20).
To evoke release, PB containing a stimulus (with Ca 2ϩ or 0.1 mM EGTA) was added to synaptosomes at 24°C; basal Ca 2ϩ -independent release was determined using stimulus-free PB/EGTA. For quick and thorough mixing and to prevent NT reuptake (20,25), synaptosomes (10 -25 g of protein) were diluted by the stimulus 8-fold. Using high K ϩ , maximal release was achieved at 25 mM K ϩ . A23187 in dimethyl sulfoxide was diluted into Ca 2ϩ -containing PB 10 s before use. For comparison with superfusion studies where terminals were exposed to LTX for 1 min and release was evaluated over a longer period (21), some samples were stimulated for 1 min and then diluted 5-fold, resulting in subnanomolar LTX present during release measurements. Hypertonic stimulation was done by adding sucrose in PB/EGTA to a final concentration of 150 mM, which induced maximal vesicular release (identified using BAF); higher [sucrose] only intensified nonvesicular NT efflux.
To terminate Ca 2ϩ -dependent release, EGTA was added to 0.1 mM free concentration. LTX-evoked Ca 2ϩ -independent release was stopped by 0.1 mM LaCl 3 . Sucrose-stimulated samples were returned to isotonic conditions by adding low sodium PB. Prestimulation with high K ϩ was stopped by dilution with potassium-free PB. Samples were then centrifuged for 30 s, and the supernatant was recovered immediately. The radioactivity of the supernatant and pellet was determined by liquid scintillation counting. Phase 1 of LTX-evoked AA release was measured during the first 2 min for [ 14 C]Glu and 1 min for [ 3 H]GABA or over 5 min in the presence of La 3ϩ for both AAs (both methods gave same results). Phase 2 was determined by subtracting phase 1 from the aggregate Ca 2ϩ -dependent release over 5 min (phases 1 and 2).
Production of Recombinant Toxins-LTX WT and LTX N4C for baculovirus expression (26) were made similar to those described previously (27), except no poly-His tags were introduced (for better resemblance of native LTX). The recombinant toxins were purified from culture medium by affinity chromatography on an anti-LTX monoclonal antibody (kind gift of Prof. E. V. Grishin) attached to CNBr-activated Sepharose (Amersham Pharmacia Biotech). The toxins were eluted with 1 M MgCl 2 and dialyzed against PB.
Electron Microscopy and Image Processing-Cryo-electron microscopy (cryo-EM) and data analysis were carried out as described (17).
LTX Pore Formation-LTX-stimulated influx of 45 Ca 2ϩ into synaptosomes was measured as described (8). For electrophysiological detection of LTX channels (9), baby hamster kidney (BHK) cells were transiently transfected with LPH (10) or an empty vector, mixed with a plasmid encoding green fluorescent protein (used to identify transfected cells). Whole-cell current recordings were performed 24 h later on isolated cells, using wide-tipped patch pipettes (1.5-3 megohms). Where indicated in figures, cells were perfused at 70 l/min, using a local multichannel superfusion system. pClamp 6 software was used for experimental protocols and Biopatch for analysis. Whole-cell current traces are shown without leak subtraction, after low pass filtering.
Electrophysiological Recordings in Hippocampal Slices-Immediately after extraction, 10 -12-day-old rat brains were placed in ice-cold oxygenated artificial cerebrospinal fluid (also used for recordings

RESULTS
Loading Synaptosomes with Radiolabeled Transmitters-To detect synaptosomal release of AAs, we chose a batch (versus superfusion) method, because it allows simultaneous and fast measurement of multiple samples and improves statistical evaluation of data. This assay yielded results comparable with fast superfusion (6 ml/min) (25).
When synaptosomes were incubated with [ 14 C]Glu at 37°C, a maximal radioactivity uptake was achieved in 5 min (data not shown). Labeled NT, taken up into the cytosol, requires some time to equilibrate with the vesicles. Therefore, in such shortloaded synaptosomes, vesicular release evoked by high K ϩ in a Ca 2ϩ -dependent manner was initially small but gradually increased upon further incubation at 37°C and then remained stable for several hours (Fig. 1A). In contrast, a Ca 2ϩ -independent (probably cytosolic) component decreased with time (Fig.  1B). For [ 3 H]GABA, the optimal loading/post-loading period was 120 min (data not shown). This optimized ("normal") loading protocol was used in most experiments below.
These results, in agreement with the earlier findings (28,29), suggest that radiolabeled cytosolic NT first enters a limited number of vesicles but, with time, accumulates in a larger SV population.
LTX Evokes Three Distinct Types of AA Release-In synaptosomes loaded by the normal procedure, LTX triggered both Ca 2ϩ -dependent and -independent AA release ( Fig. 1, C and D). To accurately determine the amount and properties of these types of release, secretion had to be terminated rapidly and selectively prior to the separation step (see "Experimental Procedures"). Addition of EGTA instantly stopped the Ca 2ϩ -dependent effect of LTX ( Fig. 1C) but could not prevent its Ca 2ϩindependent action. We found, however, that 0.1 mM La 3ϩ blocked the latter release immediately and completely ( Fig. 1D) (3,30). When mixed with LTX before synaptosomal intoxication, La 3ϩ also prevented all evoked Ca 2ϩ -independent release (Fig. 2, A and B). Measured using this method, the Ca 2ϩindependent AA secretion increased with LTX dose but always occurred at a constant rate.
In contrast, the Ca 2ϩ -dependent LTX-evoked release was not linear with time ( Fig. 1C) and displayed two temporally separated phases ( To compare the rates of the fast Ca 2ϩ -dependent and Ca 2ϩindependent LTX actions, we used 5 nM LTX and determined a higher resolution time course. Under these conditions, phase 1 was complete in 30 s, with up to 2.9 Ϯ 0.9% of total [ 14 C]Glu being released within the first 15 s. Importantly, no Ca 2ϩindependent release could be detected at these times, indicating that phase 1 of the Ca 2ϩ -dependent component is the fastest type of LTX-stimulated secretion. La 3ϩ Perturbs the Pore-dependent Release by Affecting LTX Tetramers-La 3ϩ blocks two types of LTX-evoked AA release: Ca 2ϩ -independent and delayed Ca 2ϩ -dependent. In noradrenergic terminals, these release types require the presence of LTX pores (19). Together, these findings suggest that La 3ϩ inhibits AA release by affecting LTX pores. To examine any direct interaction of this cation with toxin pores, while excluding the involvement of nonreceptor neuronal proteins, we used BHK cells. In such cells transfected with LTX receptors, the toxin induces large pores (9). Addition of La 3ϩ to the bath quickly inhibited conductance of these pores (Fig. 3A). This inhibitory effect was fully reversible upon washout, indicating that La 3ϩ acts extracellularly and does not permanently damage LTX.
LTX pores are formed by toxin tetramers inserting into the membrane (17), and La 3ϩ is likely to block such pores by perturbing the tetramers. We studied the effect of La 3ϩ on LTX structure by cryo-EM and found that only Ͻ10% of all molecules in a La 3ϩ -treated LTX sample were tetrameric, compared with 90 -95% in control toxin (Fig. 3B). Furthermore, the few tetramers still found in 0.1 mM La 3ϩ were grossly deformed; the shapes and relative positions of monomers became clearly altered, resulting in a nearly complete closure of the central pore (10 -25 Å in control and Յ3 Å in the presence of La 3ϩ ) (Fig. 3C).
Combined, these results indicate that La 3ϩ inhibits some LTX actions by affecting the toxin pores and that both the Ca 2ϩ -independent and the delayed Ca 2ϩ -dependent components of LTX-evoked AA release rely on the presence of such pores. In contrast, the fast Ca 2ϩ -dependent phase does not require LTX pore formation.
Only Ca 2ϩ -dependent LTX-evoked Release Is Vesicular-Because LTX pores are permeable to NTs (10), we checked whether any LTX-evoked release of radiolabeled AAs was caused by leakage through such pores. To identify vesicular secretion, we used two drugs: (i) BoNT/E, which cleaves SNAP 25, a SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) protein essential for exocytosis; and (ii) BAF, which blocks the vacuolar-type proton pump and prevents NT loading into SVs, thus decreasing measured exocytosis (22). Both reagents inhibited the Ca 2ϩ -dependent AA release stimulated by high K ϩ or Ca 2ϩ ionophore, A23187 (Fig.  4, A-D). Similarly, these drugs blocked the Ca 2ϩ -dependent component of LTX action (Fig. 4, E and F). Moreover, both the slow and fast Ca 2ϩ -dependent components of LTX-evoked release were abolished (Fig. 4, I-L), confirming their exocytotic nature.
In contrast, the Ca 2ϩ -independent LTX-evoked AA release was unperturbed in botulinized or BAF-treated terminals (Fig.  4, G and H), implying that the bulk of such release does not occur via conventional exocytosis. Because this type of release strictly depends on LTX pore formation (Figs. 2 and 3), we term it below "LTX-evoked pore-mediated efflux." Ca 2ϩ -dependent Components of LTX-evoked Release Require Intracellular Ca 2ϩ (Ca 2ϩ i )-To further characterize the two types of Ca 2ϩ -dependent LTX-stimulated exocytosis, we asked whether Ca 2ϩ acted as an extracellular co-factor for LTX or was required in the cytosol. Preloading synaptosomes with BAPTA substantially reduced the Ca 2ϩ -dependent AA release triggered by high K ϩ or by A23187 (Fig. 5, A-D), and inhibited both components of Ca 2ϩ -dependent LTX-evoked release (Fig. 4, E, F, H, and I). Thus, cytosolic Ca 2ϩ is necessary for all Ca 2ϩ -dependent LTX actions.
The Ca 2ϩ -independent LTX-evoked AA efflux has no Ca 2ϩ requirement, and BAPTA-loaded terminals actually exhibited a larger amount of such release (Fig. 4, K and L).
Only the Fast Phase of Ca 2ϩ -dependent LTX-evoked Release

Requires Ca 2ϩ Stores and Phospholipase C (PLC)-Cytosolic
Ca 2ϩ that participates in LTX actions may, at least in part, come from intracellular Ca 2ϩ stores known to play an important role in LTX-triggered release of NE (8,18,19). Therefore, prior to stimulation with LTX, we depleted Ca 2ϩ stores with TG. As a result, the fast phase of Ca 2ϩ -dependent LTX-evoked [ 14 C]Glu release was greatly attenuated (Fig. 5G), whereas the delayed phase was not affected (Fig. 5J). Some intracellular Ca 2ϩ stores can be mobilized by inositol trisphosphate, a product of PLC action. Indeed, inhibition of PLC by U73122 blocked the fast (Fig. 5G), but not the slow (Fig. 5J), phase of [ 14 C]Glu release. Importantly, neither TG nor U73122 acted by simply causing depletion of vesicles, which could still be stimulated by sucrose (both AAs; data not shown). Thus, phase 1 may involve PLC-mediated mobilization of Ca 2ϩ i , whereas phase 2 is probably triggered directly by Ca 2ϩ entering through the LTX pore.
As expected, the Ca 2ϩ -independent LTX pore-mediated release was unaffected by TG or U73122, indicating that this pathway does not involve Ca 2ϩ stores or PLC (Fig. 5M).

Non-pore-forming LTX Mutant Evokes Fast Ca 2ϩ -dependent
Exocytosis-Thus, the fast phase of LTX-stimulated release did not involve pore formation but required mobilization of Ca 2ϩ i , a process that can be induced by receptor signaling (8,27). However, a possibility remained that some LTX pores were not completely blocked by La 3ϩ . Therefore, to ascertain that phase 1 involves LTX receptor signaling, a toxin was needed that lacked any ability to form membrane pores. Coincidentally, an interesting LTX mutant (LTX N4C ) had been described that bound to both LTX receptors and activated PLC but did not cause any Ca 2ϩ -independent secretion from synaptosomes (27).
Because LTX pores are responsible for Ca 2ϩ -independent release (Figs. 2 and 3), we hypothesized that LTX N4C was inactive in the absence of Ca 2ϩ because it could not form pores. To test this hypothesis, we compared the pore-forming activities of LTX N4C and recombinant wild type toxin (LTX WT ) by measuring influx of 45 Ca 2ϩ into synaptosomes. Addition of LTX WT (or native LTX; data not shown) caused 45 Ca 2ϩ accumulation in the terminals (Fig. 6A). In contrast, LTX N4C did not induce any Ca 2ϩ influx (Fig. 6A). This result was confirmed by single-channel patch-clamp recordings in BHK cells expressing exogenous receptors. LTX N4C was incapable of forming pores in these cells (Fig. 6B), whereas LTX WT , added after the mutant, induced pores very efficiently (Fig. 6B), indicating that the inability to form pores was the feature of LTX N4C and not the cells.
We then compared the actions of both recombinant toxins on AA release. LTX WT mimicked the native toxin and triggered all three components of release in synaptosomes (Fig. 6, C-H). In contrast, LTX N4C failed to induce both the delayed phase of Ca 2ϩ -dependent secretion and the Ca 2ϩ -independent pore-mediated efflux (Fig. 6, E-H). Strikingly, the mutant was equipotent with LTX WT in stimulating the fast phase of Ca 2ϩ -dependent AA exocytosis (Fig. 6, C and D). Patch-clamp recordings in

FIG. 6. Mutant LTX N4C fails to form pores but still evokes the Ca 2؉ -dependent receptor-mediated exocytosis.
A, LTX WT forms cation-permeable membrane pores, whereas LTX N4C does not. Synaptosomes were equilibrated with 2 mM 45 Ca 2ϩ /Ca 2ϩ and exposed to 3 nM LTX WT or LTX N4C ; radioactive uptake was determined at different times. B, electrophysiological detection of channel formation by LTX WT . Whole-cell voltage-clamp current recordings were done on BHK cells expressing LPH (see Fig. 3). Main trace, 1 nM LTX N4C or LTX WT was added by superfusion as indicated by bars; hatched bars indicate wash periods. Inset, expanded main trace at the outset of LTX WT pore formation (two distinct channels are visible). C-H, components of NT release triggered by LTX WT and LTX N4C . Release from loaded synaptosomes was determined as in Fig. 2; both recombinant toxins were used at 2.5 nM for [ 3 H]GABA and 4 nM for [ 14 C]Glu. The Ca 2ϩ -independent efflux was measured over 5 min. I, LTX N4C causes a massive increase in the frequency of mEPSCs in rat hippocampal slices. Glu-mediated mEPSCs were recorded from hippocampal sections in the presence of 2 mM Ca 2ϩ before or 1 min after the addition of 0.5 nM LTX N4C . J, cumulative probability distributions of mEPSC amplitudes, demonstrating that LTX N4C does not change the amplitude of mini-events. K, cumulative probability distributions of inter-event intervals, showing that LTX N4C dramatically shortens the intervals between mEPSCs. The data in A are from a representative experiment done in triplicate; other data are from sextuplet determinations in two (C, E, G) or one (D, F, and H) independent experiments (three preparations of recombinant toxins); B and I are representative traces (n ϭ 5); J, Kolmogorov-Smirnov's twotailed p Ͼ 0.5, n ϭ 502 events for both samples; K, p Ͻ 0.00001, n ϭ 500 intervals.
CA3 neurons of rat hippocampal slices unequivocally demonstrated that, in the presence of Ca 2ϩ , LTX N4C causes a strong increase in the frequency of Glu-mediated asynchronous miniature events without altering their amplitudes (Fig. 6, I-K).
Our data indicate that LTX N4C is capable of triggering AA exocytosis (phase 1) without forming pores but by interacting with the toxin receptor(s).
The Delayed Ca 2ϩ -dependent LTX-evoked Release Requires Active PI 4-Kinase-Because the two Ca 2ϩ -dependent phases had different pharmacological properties and time courses, they could involve different SV pools. Some NE-containing vesicles require active PI 4-kinase, and Ca 2ϩ -dependent LTXevoked NE release is sensitive to an inhibitor of this enzyme, PAO (19,21,31). If AAs are released via a similar mechanism, their secretion must also be perturbed by PAO.
When testing this drug, we found that it considerably inhibited Ca 2ϩ -dependent [ 14 C]Glu release evoked by high K ϩ or A23187 (Fig. 7, A and B). As reported previously (21), PAO had no effect on the Ca 2ϩ -independent release triggered by sucrose (Fig. 7C) or LTX (Fig. 7, H and I).
However, when comparing the sensitivities to PAO of the two Ca 2ϩ -dependent phases of LTX-evoked AA release, we ob-served an interesting effect; surprisingly, the fast phase was unaltered (Fig. 7, D and E), whereas phase 2 was blocked completely (Fig. 7, F and G), indicating that these types of Ca 2ϩ -dependent release occur from vesicular pools with different sensitivities to PAO.
The Fast Ca 2ϩ -dependent LTX-evoked Release Occurs from a Fast-loading Pool of Vesicles-Our results employing PAO agreed with previous findings for endogenous Glu secretion (31), whereas some authors did not observe any PAO-sensitive release of radiolabeled AAs (21). We noted, however, that the latter group had labeled synaptosomes for only 5 min, while our normal protocol required long post-loading periods (Fig. 1). We speculated that the length of loading might differentiate between vesicular pools with fast and slow loading kinetics. Therefore, we tested the effect of LTX on the "short-loaded" synaptosomes. Interestingly, short loading did not affect significantly phase 1 but abolished phase 2 of Ca 2ϩ -dependent LTX-evoked AA exocytosis (Fig. 7, D-G). Evidently, the latter phase occurs from a distinct SV pool that slowly accumulates radioactive AAs from the cytosol. The Ca 2ϩ -independent LTX pore-mediated AA efflux was similar in terminals loaded by either procedure (Fig. 7, H and I), further demonstrating that the Ca 2ϩ -dependent and -independent types of release occur from different NT pools.
We then tested whether the SV pool involved in the fast Ca 2ϩ -dependent LTX action retained its pharmacological features in such short-loaded terminals. As after the normal loading procedure, PAO still did not affect the fast Ca 2ϩ -dependent LTX exocytosis, whereas TG greatly attenuated it (Fig. 7E). As expected, the slow Ca 2ϩ -dependent component remained nonexistent (Fig. 7G), and neither drug affected the Ca 2ϩ -independent LTX pore-mediated Glu efflux (Fig. 7I). The same results were obtained in similar experiments with GABA (data not shown).
These findings indicate that the fast Ca 2ϩ -dependent LTXevoked exocytosis occurs from a fast-loading vesicular pool that may require intact Ca 2ϩ stores but not PI 4-kinase.
The Fast Phase of LTX-evoked Release Occurs from Sucroseand Depolarization-sensitive Vesicles-Being PAO-insensitive, phase 1 of Ca 2ϩ -dependent LTX exocytosis must occur from SVs that have already undergone PI 4-kinase-dependent priming. These vesicles may belong to the readily releasable pool (RRP) distinguished by its sensitivity to sucrose (32,33). We found that 150 mM sucrose stimulated Ca 2ϩ -independent [ 14 C]Glu release from synaptosomes (ϳ9% of total AA content; e.g. Fig. 7C, control); this secretion was largely vesicular because pretreatment of terminals with BAF decreased it by 65 Ϯ15%. Interestingly, the amount of AAs released by sucrose did not depend on the loading time or the length of stimulation (3-90 s; data not shown). However, the length of such stimulation had a profound effect on the efficacy of a second stimulus applied within 4 min (see Fig. 8, top panel, for experimental design): terminals stimulated first for 3 s produced the same release as untreated synaptosomes when challenged with sucrose for a second time (Fig. 8, A and B). In contrast, a prolonged (90 s) first stimulation strongly inhibited further sucrose-induced release (Fig. 8, A and B). Sensitivity to hypertonic solutions recovered slowly after 4 min.
This temporary depression of the RRP, similar to that observed previously (34 -36), allowed us to test whether sucrose and Ca 2ϩ -dependent LTX mechanisms acted on the same SV pool. We found that a 90-s prestimulation with sucrose drastically reduced the subsequent Ca 2ϩ -dependent LTX phase 1 of [ 14 C]Glu exocytosis (measured with or without La 3ϩ ), whereas phase 2 remained unperturbed (Fig. 8, C and D). In a reverse experiment, to avoid irreversible pore formation by native LTX, we used LTX N4C (in the presence of Ca 2ϩ ). The mutant strongly inhibited subsequent Ca 2ϩ -independent release triggered by sucrose (Fig. 8F). Sucrose and LTX N4C also reciprocally perturbed each other's action in the short-loaded synaptosomes, in which only the fast-loading vesicles contained [ 14 C]Glu (Fig. 8,  G, H, and J). Very similar data were obtained with [ 3 H]GABA (data not shown). Combined, these results strongly suggest that sucrose and the fast LTX mechanism act on the same vesicular pool.
Interestingly, predepolarization of synaptosomes with high K ϩ inhibited the stimulation of this pool by LTX or sucrose (Fig. 8, K and N). In contrast, pretreatment with sucrose or high K ϩ had no effect on either the slow Ca 2ϩ -dependent (Fig.  8, D, H, and L) or the Ca 2ϩ -independent (Fig. 8, E, I, and M) LTX-evoked AA release, indicating that these types of release do not involve the RRP. DISCUSSION Pore-dependent and -independent LTX Actions-Of the two modes of LTX action (pore formation and receptor stimulation), the former is likely to cause the strongest effect. Indeed, large LTX pores (10 -25 Å) (17), permeable to cations and NTs (7,8,13,14), must perturb membrane potential, energy level, and cation balance in excitable cells (7,13,37) and cause cell/ terminal death 2 (15,16). Nevertheless, many studies have concluded that LTX interaction with its receptor(s) can facilitate exocytosis (2,4,38,39). Obviously, more reliable results can be obtained if pore formation is precluded. This has been achieved here by blocking LTX pores with La 3ϩ and by using LTX N4C , which cannot form pores.
Trivalent cations can block LTX-evoked release, and this rapid, extracellular, and reversible inhibitory action (Figs. 1D and 3A) has been always attributed to the blockade of LTX pores (3,7,30). We have now revealed the molecular mechanism of this inhibition; La 3ϩ , likely residing in the pore center, induces conformational changes in LTX tetramers/pores, leading to the closure of the central channel and tetramer dissociation (Fig. 3, B and C). As a result, this cation selectively inhibits such AA release that depends on LTX pores: (i) exocytosis evoked by Ca 2ϩ entry (Fig. 2, E and F) and (ii) nonvesicular efflux of cytosolic AAs (Fig. 2, A and B). In contrast, LTX binding to the receptors persists in 0.1 mM La 3ϩ (27,30) and could lead to receptor activation. Consistent with this idea and in agreement with the earlier findings (2, 40), some LTX actions are not blocked by La 3ϩ (Fig. 2, C-F).
To avoid any direct effect of La 3ϩ on synaptosomal secretion, we employed the LTX mutant. LTX N4C has ideal qualities; it binds to LPH and NRX and stimulates hydrolysis of phosphoinositides (27), but does not form pores (Fig. 6, A and B) or trigger Ca 2ϩ -independent NT release (27) (Fig. 6, G and H). We have now demonstrated 3 that LTX N4C does not form pores because it cannot tetramerize. Consistent with that, LTX N4C fails to stimulate those events that require pore formation (Fig.  6, E-H) but causes receptor-mediated secretion (Fig. 6, C, D, and I).
Three Types of LTX-stimulated Release-We have identified here at least three components of LTX action in synaptosomes: one Ca 2ϩ -independent and two Ca 2ϩ -dependent pathways (Fig.  2). These release types have very different properties and principal mechanisms and require a systematic analysis (as detailed below).
Ca 2ϩ -independent LTX-evoked Release-Electrophysiologically detected quantal secretion caused by LTX in the absence of Ca 2ϩ is weaker than the Ca 2ϩ -dependent exocytosis (2,11,12). In synaptosomes, however, the Ca 2ϩ -independent LTX action is inconsistently strong (27) and must (at least in part) come from the cytosol. Indeed, the cytosol contains Ͼ80% of endogenous synaptic AAs (37) that are used to refill SVs (41); in short-loaded synaptosomes, the radiolabeled cytosolic pool is even larger (see Fig. 1, A and B). Furthermore, the Ca 2ϩindependent LTX-evoked release of several transmitters (NE, Glu, and GABA) biochemically measured in synaptosomes is insensitive to inhibitors of exocytosis: (i) clostridial neurotoxins (8) (Fig. 4), (ii) BAF (Fig. 4), or (iii) hypertonic or hyperkalemic overstimulation (Fig. 8). Together, these results indicate that the Ca 2ϩ -independent LTX effect in synaptosomes is largely nonvesicular. Being strictly dependent on the NT-permeable LTX pores, Ca 2ϩ -independent AA release is likely to represent leakage through such pores.
This seems to contradict electrophysiological recordings in tissue preparations clearly demonstrating vesicular release triggered by LTX in "zero" Ca 2ϩ (e.g. Ref. 1). However, unlike release observed in synaptosomes, this secretion is blocked by clostridial toxins (2). In addition, Ca 2ϩ -independent quantal release evoked by LTX is 500 -1000 times lower than that induced by sucrose (2). Therefore, although sucrose-stimulated secretion is detectable biochemically (ϳ8 -10% of the total synaptosomal pool; Figs. 7C and 8 (A and B)), the Ca 2ϩ -independent LTX-evoked exocytosis (ϳ0.02%) is below the detection limit of radioassay.
Two Phases of Ca 2ϩ -dependent LTX Action-In our experiments, the aggregate Ca 2ϩ -dependent LTX-stimulated release over 5 min constitutes (percentage of total content) 10.3 Ϯ 2.0 for Glu and 11.8 Ϯ 1.9 for GABA, which is compatible with the maximal amount of endogenous Glu that could be released from synaptosomes in a Ca 2ϩ -dependent fashion (8 -16%) (29,42). As verified by BoNT and BAF treatment (Fig. 4), the Ca 2ϩ -dependent components of LTX-triggered release are vesicular.
Our findings corroborate and extend the previously published data (21,27,43). (i) In addition to the Ca 2ϩ -independent AA release seen by these authors, we have now detected the Ca 2ϩ -dependent component, always observed electrophysiologically; (ii) having confirmed that LTX N4C is inactive in the absence of Ca 2ϩ , we have also found that it causes Ca 2ϩ -dependent exocytosis; (iii) having explained how La 3ϩ inhibits the Ca 2ϩ -independent release (27), we have now demonstrated that the cation does not block the fast Ca 2ϩ -dependent action (as reported previously (40)). Several factors have allowed us to observe the Ca 2ϩ -dependent LTX actions. (i) Long loading of synaptosomes labeled the depot vesicles and revealed phase 2 (Figs. 2 (E-H) and 7F); (ii) instantaneous termination of release allowed us to detect the fast phase 1 of LTX action, usually lost with slow superfusion (Ͻ1 ml/min) (25); (iii) avoiding prestimulation of synaptosomes with high K ϩ before adding LTX preserved phase 1 (Fig. 8K).
Remarkably, we have also been able to divide the Ca 2ϩ -dependent LTX-evoked exocytosis into two temporally distinct phases. These phases have different pharmacological characteristics and, consequently, different underlying mechanisms. Phase 1 (i) is complete within 30 -60 s after the addition of LTX (Fig. 2, C-F), (ii) requires Ͻ0.4 mM Ca 2ϩ e (Fig. 2, G and H), (iii) is independent of the toxin pore (not blocked by La 3ϩ ), (iv) depends on internal Ca 2ϩ stores and active PLC (Fig. 5G), and (v) is stimulated by LTX N4C (Fig. 6, C and D). Phase 2 (i) is delayed, producing detectable release only 2-4 min after LTX addition (Fig. 2, E and F), (ii) requires Ͼ1 mM Ca 2ϩ e , (iii) is sensitive to La 3ϩ , (iv) is insensitive to disruption of Ca 2ϩ stores and inhibition of PLC (Fig. 5J), and (v) is not evoked by LTX N4C (Fig. 6, E and F). Phase 1, therefore, does not involve LTX pore formation, whereas phase 2 is probably caused by pore-mediated Ca 2ϩ accumulation in the cytosol (Fig. 6A).
An important question arises about the nature of the fast phase. Is it evoked by toxin's partial internalization and direct activation of some exocytotic proteins (44) or is it mediated by the receptor(s)? The former hypothesis seems to be unlikely because (i) toxin internalization cannot be faster than pore formation and (ii) the mutant toxin, which is unable to internalize (44), still stimulates this phase of release upon binding to the receptors (Fig. 6). Therefore, this release must involve some receptor-transduced signaling.
Receptor-mediated Exocytosis-What could be the mechanism of LTX receptor-mediated Ca 2ϩ -dependent AA release? As with the Ca 2ϩ -dependent release of NE (8), this component is sensitive to TG and U73122 (Figs. 5G and 7E), implicating internal Ca 2ϩ stores and PLC. Activation of PLC by either LTX WT or LTX N4C (27) has features remarkably similar to those of the fast phase of AA release (Fig. 6, C and D): both require receptor binding and the presence of Ca 2ϩ e , and both are insensitive to La 3ϩ . It seems likely that inositol trisphosphate, a product of PLC action, could trigger mobilization of Ca 2ϩ from intracellular stores. Whether the initial local increase in [Ca 2ϩ ] i is sufficient to stimulate exocytosis (45) or whether it causes activation of store-operated Ca 2ϩ channels and subsequent amplification of free [Ca 2ϩ ] i remains the subject of future studies, as does the exact nature of these Ca 2ϩ stores.
Which of the two LTX receptors may be involved in the two phases of AA release? Activation of PLC and mobilization of Ca 2ϩ i , which underlie LTX-evoked NE secretion (8) and the fast AA release, can be mediated by the heptahelical LPH coupled to G␣ q (18). On the other hand, both LPH and NRX (or even their nonsignaling mutants) can mediate the pore-dependent phase 2 (9, 10); this explains the results used to speculate that LTX acts without receptor activation (43). The future use of LTX N4C should help to determine the role of each receptor in LTX action.
Two Distinct SV Pools-The most interesting distinction between the two phases of LTX-evoked Ca 2ϩ -dependent release is their differential sensitivity to PAO and loading time, implicating physiologically distinct vesicular pools.
When synaptosomes are loaded with radiolabeled AAs, the sucrose-sensitive RRP accumulates the label very quickly, in parallel with AA uptake into the cytosol. In contrast, the time necessary to load the second SV pool (Fig. 1A) appears surprisingly long (1-2 h) compared with ϳ20 s required for recovery of vesicles after stimulation (46). However, without stimulation, the endogenous NT present in some vesicles must prevent their loading with labeled cytosolic AAs. Thus, if a vesicular pool is emptied infrequently, it will be labeled only after a considerable delay. This consideration justifies our loading paradigm and supports the idea that the slow-loading SVs belong to the resting (depot) pool.
If synaptosomes are treated with PAO (inhibitor of PI 4-kinase, which participates in priming of vesicles for release) (31) and then loaded with radiolabeled AAs, release will be detected only from those vesicles that have passed the priming step prior to drug addition. Such SVs can be specifically released by sucrose (Fig. 7C), which only stimulates the RRP (32,33), and by LTX (fast Ca 2ϩ -dependent phase, Fig. 7, D and E). We demonstrate that these two types of release occur from a common SV pool (Fig. 8, C, F, G, and J). Thus, the fast-loading vesicles insensitive to PAO must belong to the RRP. The size of this pool determined here is 6 -8% of the total radioactive AA content, and this probably includes both the morphologically docked and near-membrane SVs (47).
On the other hand, Ca 2ϩ -dependent AA release induced by the delayed (pore-mediated) LTX action, by high K ϩ /Ca 2ϩ , or by A23187 is sensitive to PAO (Fig. 7, A, B, and F), suggesting that it occurs from SVs that still need priming by PI 4-kinase. However, such release is absent in short-loaded synaptosomes (Figs. 1A and 7G), implying that: (i) the PAO-sensitive vesicles do not accumulate radiolabeled AAs during a short loading, and (ii) LTX pore acts on an SV pool distinct from the RRP. Because this pool contains unprimed and slow-loading SVs, it probably represents the depot pool.
The two SV pools are differentially regulated by the two Ca 2ϩ -dependent LTX mechanisms. The receptor-mediated pathway triggers rapid fusion of the RRP (probably by increasing local [Ca 2ϩ ] i ) but does not induce fusion of the depot vesicles. The latter may require sufficient Ca 2ϩ entering via LTX pores (Fig. 6A). In gonadotrophs too, Ca 2ϩ release from stores causes fast exocytosis, whereas a universal rise in [Ca 2ϩ ] i induces only a slow granule fusion (45). Can the depot vesicles be released separately, or do they become readily releasable prior to exocytosis? After repeated 4-s applications of hypertonic solutions or high frequency electrical stimulation, the RRP is depressed for 3-4 min (36). Similarly, in our experiments, once the RRP is released by a 90-s treatment with sucrose or by LTX N4C , replacement vesicles sensitive to these stimuli are not immediately available (Fig. 8, A-C, F, G, and J). Importantly, in such overstimulated synaptosomes, the LTX pore can still induce delayed Ca 2ϩ -dependent release (e.g. Fig. 8D) of vesicles that are sensitive to PAO but not to sucrose, indicating that the depot vesicles can be released independently, without necessarily becoming readily releasable.
Conclusions-Our results indicate that LTX can be successfully used only when carefully separating out its diverse actions. We demonstrate that LTX N4C is a very useful probe for elucidating the receptor-mediated actions. An important implication of our findings is that the readily releasable vesicles can be selectively regulated via the LTX receptors; the mechanism of this regulation can now be studied using the mutant toxin.