Glutamate-induced Exocytosis of Glutamate from Astrocytes*

Recent studies indicate that astrocytes can play a much more active role in neuronal circuits than previously believed, by releasing neurotransmitters such as glutamate and ATP. Here we report that local application of glutamate or glutamine synthetase inhibitors induces astrocytic release of glutamate, which activates a slowly decaying transient inward current (SIC) in CA1 pyramidal neurons and a transient inward current in astrocytes in hippocampal slices. The occurrence of SICs was accompanied by an appearance of large vesicles around the puffing pipette. The frequency of SICs was positively correlated with [glutamate]o. EM imaging of anti-glial fibrillary acid protein-labeled astrocytes showed glutamate-induced large astrocytic vesicles. Imaging of FM 1-43 fluorescence using two-photon laser scanning microscopy detected glutamate-induced formation and fusion of large vesicles identified as FM 1-43-negative structures. Fusion of large vesicles, monitored by collapse of vesicles with a high intensity FM 1-43 stain in the vesicular membrane, coincided with SICs. Glutamate induced two types of large vesicles with high and low intravesicular [Ca2+]. The high [Ca2+] vesicle plays a major role in astrocytic release of glutamate. Vesicular fusion was blocked by infusing the Ca2+ chelator, 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid, or the SNARE blocker, tetanus toxin, suggesting Ca2+- and SNARE-dependent fusion. Infusion of the vesicular glutamate transport inhibitor, Rose Bengal, reduced astrocytic glutamate release, suggesting the involvement of vesicular glutamate transports in vesicular transport of glutamate. Our results demonstrate that local [glutamate]o increases induce formation and exocytotic fusion of glutamate-containing large astrocytic vesicles. These large vesicles could play important roles in the feedback control of neuronal circuits and epileptic seizures.

Whole-cell Patch Clamp Recording-Cells were visualized with a ϫ60/0.90 water immersion lens on an Olympus BX51 upright microscope (Olympus Optical Co., New York) equipped with IR differential inference contrast (DIC) optics. Patch electrodes with resistances of 5-10 M⍀ were pulled from KG-33 glass capillaries (inner diameter 1.0 mm, outer diameter. 1.5 mm; Garner Glass Co., Claremont, CA) using a P-97 electrode puller (Sutter Instrument Co., Novato, CA). Pyramidal neurons in the CA1 pyramidal layer and astrocytes in stratum radiatum were identified by their distinct DIC morphology and electrophysiological properties as described previously (9). Pyramidal neurons were voltage-clamped at Ϫ60 mV or current-clamped without holding currents (for measuring V m ), and astrocytes were voltage-clamped at Ϫ80 mV. Cells with a seal resistance Ͻ5 gigaohms, a holding current of more than Ϫ200 pA, or changes in the series resistance Ͼ10% of control were rejected from further analysis. The pipette filling solution for neuronal whole-cell recording contained the following (in mM): 123 potassium gluconate, 10 KCl, 1 MgCl 2 , 10 HEPES, 0.1 EGTA, 0.01 CaCl 2 , 1 ATP, 0.2 GTP, and 4 glucose (pH adjusted to 7.2 with KOH). The pipette solution for whole-cell recording in astrocytes contained the following (in mM): 123 potassium gluconate, 10 KCl, 1 MgCl 2 , 10 HEPES, 5 EGTA, 0.5 CaCl 2 , 1 ATP, 0.2 GTP, and 4 glucose (pH adjusted to 7.2 with KOH). Recorded signals were filtered through an 8-pole Bessel low pass filter with 2-kHz cut-off frequency and sampled using the PCLAMP 9.0 acquisition program (Axon Instruments Inc.) with an analog-digital point sample interval of 200 s.
Fluorescence Imaging-A customized two-photon laser scanning Olympus BX61WI microscope with a ϫ60/0.90 water immersion lens was used to detect fluorescence signals. A Mai-Tai TM laser (Solid-State Laser Co., Mountain View, CA) tuned to 830 or 890 nm was used for excitation. Image acquisition was controlled by Olympus Fluoview FV300 software (Olympus America INC, Melville, NY). In the transfluorescence pathway, a 565 nm dichroic mirror was used to separate green and red fluorescence. HQ525/50 and HQ605/50 or HQ525/50 and 710 nm high pass filters were placed in the "green" and "red" pathways, respectively, to eliminate transmitted or reflected excitation light (Chroma Technology Corp., Rockingham, VT). Fluorescence images were scanned in the X-Y-T or X-Y-Z mode with intervals of 2 or 4 s. Alexa Fluor-594 or FM 1-43 fluorescence was detected via the red pathway with the HQ605/50 filter. FM 4-64 was detected via the red pathway with the 710 nm filter. Fluo-4 fluorescence was detected using the green pathway. Base-line fluorescence (F 0 ) was the average of four images during control, and ⌬F/F was calculated as (⌬F/F)(t) ϭ (F(t) Ϫ F 0 )/F 0 . The position shift in the X-Y section during 5 min of scanning was 0.5 Ϯ 0.3 m (mean Ϯ S.D., n ϭ 30 cells).
Electron Microscopy-Slices (300 m) were moved from normal ACSF into ACSF containing 50 mM [glutamate] five times for 5 s at 2-min intervals, with washes in normal ACSF inbetween. Slices were then fixed with 2.5% glutaraldehyde in 0.1 mol/liter sodium cacodylate buffer (pH 7.4) overnight, and then re-sliced into 100-m sections with the vibratome. Sections were pretreated with 10% normal goat serum for 4 h at 4°C. Slices were incubated in the identical solution containing anti-GFAP antibody (1:2000, purified anti-mouse monoclonal GFAP, 2E1; BD Biosciences) for 24 h at 4°C, and then incubated with the biotinylated secondary antibody (1:200, Histostain-Plus Bulk kit, mouse IgG; Zymed Laboratories Inc.) for 24 h at 4°C, and with avidin-biotin complex overnight at 4°C. Immu- showing the recorded pyramidal neuron (Pyr) and puffing pipette (p2) in A before puffing (panel i) and after repetitive puffs (panel ii). There were multiple blebs around the puffing pipette tip in a circular area (ii, Ͼ). C, current clamp recording (V m ) showed normal RMP (Ϫ65 mV) and SIC-triggered overshooting action potentials (AP) after eight puffs. D, whole-cell recording when puffing ACSF plus 50 mM NaCl (ACSFϩ50 mM NaCl). E, whole-cell recording of SICs (SIC) from a pyramidal neuron during control (panel i), after four puffs (panel ii), 1 min (panel iii) or 40 min (panel iv) after removing the puffing pipette. F, DIC images at the time when each trace was recorded (panels i-iv). No blebs were observed 40 min after removing the puffing pipette (panel iv). Data in C-E are representative chosen from five experiments. Scale bars in B, and F, 5 m. noperoxidase labeling was visualized by incubating slices with chromogen (DAB-H 2 O 2 ) in 1% OsO 4 for 1 h at room temperature. Slices were dehydrated and embedded in Epon 812. Finally, ultra-thin sections (70 nm) were cut and stained with uranyl I acetate and lead citrate and then observed under a transmission electron microscope (HT 7100, Hitachi, Japan).
Local Application-Glass pipettes with resistances of 10 -14 M⍀ were used to rapidly pressure-eject (puff) glutamate or MSO dissolved in ACSF. Puffing pressure (3-5 p.s.i.) and puffing duration (100 ms) were controlled by a Picospritzer III (Parker Hannifin Co., Cleveland, OH), and intervals (60 -120 s) were controlled by a Master-8 stimulator (A.M.P.I., Jerusalem, Israel). Puffing pipettes were placed in stratum radiatum of the CA1 area 50 -80 m below the slice surface. To control for osmotic effects, the same concentration of NaCl (50 mM) was added to ACSF as the osmolarity-control puff solution.

Local Application of Glutamate Induces SICs in CA1 Pyramidal
Neurons-To study glutamate-induced astrocytic glutamate release, we locally applied glutamate in stratum radiatum of the CA1 area in hippocampal slices through a puffing pipette. Whole-cell recording in pyramidal neurons was performed to detect SICs that are iGluR-mediated currents activated by transient astrocytic glutamate release (7,8,26). Experiments were performed in the presence of TTX (1 M) to block neuronal action potential (AP)-dependent synaptic release. Glutamate was puffed every 1-2 min into the region near apical dendrites of the recorded pyramidal neuron (Fig.  1B, p2). During a control period before placing puffing pipettes into slices, virtually no spontaneous SICs were observed, and after placing puffing pipettes, SICs were detected occasionally, suggesting little leakage of glutamate from puffing pipettes. Puffing glutamate (50 mM) directly evoked an inward current in CA1 pyramidal neurons (Fig. 1A, I puff ). After 3-5 puffs, spontaneous SICs began to appear (Fig.  1A). Puffing glutamate induced an increase in SICs in 64 of 64 tested pyramidal neurons. SICs were blocked by the N-methyl-D-aspartic acid receptor antagonist APV (50 M) plus the ␣-amino-3-hydroxy-5-methyl-4-isoxazole propionate/kainate receptor antagonist CNQX (20 M) (Fig. 1A, APV/CNQX), confirming that glutamate-induced SICs are iGluR-mediated currents. The decay constant (mean Ϯ S.D., 394 Ϯ 21 ms, n ϭ 101 events) and the amplitude (105 Ϯ 19 pA) of spontaneous SICs were similar to SICs reported previously (7,8,26). When SICs occurred in pyramidal neurons, we always (64/64 neurons) observed, under DIC optics, an appearance of blebs around the puffing pipette tip (Fig. 1B, panel ii, Ͼ). Eight puffs of glutamate did not change neuronal resting membrane potential (RMP, Ϫ62.6 Ϯ 1.0 mV, n ϭ 10 cells) and ability to fire overshooting action potentials (Fig. 1C, V m ). SICs were not a result of high osmolarity of the glutamate solution, because a glutamate-free puff solution with similar osmolarity (ACSF added with 50 mM NaCl) did not induce SICs (Fig. 1D, ACSF ϩ NaCl). Blebs and SICs induced by puffing glutamate were reversible if five or less puffs had been applied. After removing the puffing pipette from the extracellular space, SICs and blebs gradually disappeared (Fig. 1, E and F, panels iii and iv), suggesting that glutamateinduced SICs and blebs are reversible if there are not too many applied puffs. When stopping puffs but keeping the puff pipette in slices, SICs continued to occur with low frequency, probably resulting from leaking of glutamate from the puff pipette. Repetitively puffing 5 mM [glutamate] induced only a few SICs ( Fig. 2A), whereas 2 mM [glutamate] did not induce SICs ( Fig. 2A, 2 mM Glut). The frequency of SICs was calculated from a 10-min recording period (SICs during which the I puffs were excluded), starting at the fifth puff, and was positively correlated with the concentration of glutamate in the puffing pipette (Fig. 2B,  In a previous study (26), we found that astrocytic glutamate release also activates a transient inward current in astrocytes (aTC), which synchronizes with neuronal SICs. Here we tested whether puffing glutamate also induces aTCs in astrocytes. Astrocytes near the puffing pipette were patched with the pipette solution without glutamate. As before, experiments were performed in the presence of TTX. Puffing 50 mM [glutamate] directly evoked an inward current (Fig. 2C, I puff ) and induced an increase in spontaneous aTCs in astrocytes (Fig. 2C, 2). The decay constant of aTCs (442 Ϯ 93 ms, mean Ϯ S.D., n ϭ 60 events) was similar to the decay of SICs (394 Ϯ 21 ms). In the passive type of astrocytes (30,31), aTCs may be composed of glutamate transport currents and iGluRmediated currents (32). After repeated puffing of glutamate and appearance of aTCs, perfusion of slices with the glutamate transporter inhibitor, TBOA (200 M), the ␣-amino-3-hydroxy-5-methyl-4-isoxazole propionate/kainate receptor antagonist CNQX (20 M), and the N-methyl-D-aspartic acid receptor antagonist APV (50 M) blocked aTCs (Fig. 2, C and E, FIGURE 3. Glutamate induces large vesicles in astrocytes. A-E, EM pictures showing large vesicles (v) in glutamate-treated astrocytes. Slices were exposed to 50 mM [glutamate] and then fixed. Anti-GFAP antibody was used to label astrocytes. A, astrocyte; m, mitochondria. F, astrocytes from control slices showing small vesicles. Scale bars, 1 m. G, left panel, percentage of astrocytes that showed vesicles in control (Con) and glutamate-treated slices (Glut). Middle panel, the mean diameter (Size) of vesicles in control (Con) or glutamatetreated astrocytes (Glut). ***, p Ͻ 0.001 compared with control, Student's t test. Right panel, relative frequency of vesicular diameters in control (dashed line, n ϭ 259 vesicles) or glutamate-treated astrocytes (solid line, n ϭ 201 vesicles). H, glutamate did not induce enlargement in apical dendrites of pyramidal neurons. Pyramidal neurons were loaded with Alexa Fluor-594 by electroporation. A puffing pipette containing glutamate (50 mM) was placed among labeled dendrites. Con, X-Y-Z projection of multiple dendrites before puffing glutamate. Puff, two X-Y scanning sections with 2 m distance (panels i and ii) and X-Y-Z projection (panel iii) after puffing glutamate showing formation of Alexa Fluor-594-negative large vesicles (v in the circled area) without changes in apical dendrites of pyramidal neurons. Data are representative chosen from five experiments. I, glutamate induced swollen processes of GFAP-eGFP-labeled astrocytes. A puff pipette containing 50 mM glutamate was placed near a GFAP-eGFP-labeled astrocyte. Con, the astrocyte (Ast) before puffing. Puff, two scanning sections with 2 m distance (panels i and ii) and Glutamate Induces Large Vesicles in Astrocytes-To understand the nature of glutamate-induced blebs, we used transmission EM to examine the ultrastructure of glutamate-induced blebs. Slices were transiently exposed five times to ACSF containing 50 mM [glutamate] for 5 s at intervals of 2 min. Astrocytes were identified by anti-glial fibrillary acid protein (GFAP) antibody staining. Large vesicles were observed in GFAP-positive astrocytes (Fig. 3, A-E, v) in glutamate-treated slices. Vesicles were identified by their single layer membrane without internal structures, different from mitochondria that were double layers of membrane with cristae (Fig. 3, A and B, m). Astrocytes in untreated control slices showed small vesicles (Fig. 3F). Vesicles were present in 80% (20/25) of glutamate-treated astrocytes (Fig. 3G, left panel, Glut) and 58% (23/40) of control astrocytes (Fig. 3G, left panel, Con). The mean size (diameter) of vesicles in glutamate-treated astrocytes (Fig. 3G, Size, Glut) was significantly larger than in control astrocytes (Fig. 3G, Size, Con, p Ͻ 0.001, Student's t test). Cumulative distribution curve of the vesicle diameter (Fig. 3G, right panel) indicated that glutamate treatment (Fig. 3G, right panel, solid line) increased the number of large vesicles in astrocytes. These results suggest that glutamate induces large vesicles in astrocytes.
Major cellular structures around the puffing pipette in CA1 stratum radiatum (Fig. 1B) are apical dendrites of pyramidal neurons, processes of astrocytes, interneurons, and Shaffer collateral fibers. To exclude the occurrence of large vesicles in apical dendrites of pyramidal neurons, we loaded multiple pyramidal neurons with Alexa Fluor-594 by extracellular electroporation through a pipette containing 5 mM Alexa Fluor-594 in the pyramidal layer (33). After five positive pulses of 15 V, 30-ms duration, stimuli at intervals of 2 s, multiple pyramidal neurons and their dendrites were loaded with Alexa Fluor-594 (Fig. 3H, Con). Whole-cell recording in Alexa Fluor-594loaded neurons showed a normal range of resting membrane potentials (Ϫ56 to Ϫ62 mV, n ϭ 4 neurons) and were able to fire action potentials. Glutamate (50 mM) was puffed through a puffing pipette placed among labeled dendrites. After 5-10 puffs, Alexa Fluor-594-negative large vesicles appeared (Fig.  3H, Puff, panels i, ii, and v). X-Y-Z projection images showed that dendrites of pyramidal neurons after puffing glutamate (Fig. 3H, Puff, panel iii) were similar to those before puffing (Fig.  3H, Con). No remarkable enlargement of dendrites was found during or after puffing, suggesting that glutamate does not induce the large vesicle in dendrites of pyramidal neurons.
To further confirm that glutamate-induced large vesicles originated from astrocytes, we used hippocampal slices prepared from GFAP-driven GFP-expressing mice (GFAP-eGFP). GFP fluorescence imaging showed that astrocytes were specifically labeled with GFP ( Fig. 3I, Con, Ast) in these animals. Puffing glutamate (50 mM) into CA1 stratum radiatum induced discontinued swollen astrocytic processes in a circled area (Fig. 3I, Puff, panels i-iii, circled area). The enlargement of astrocytic processes was probably because of the formation of large vesicles that occupied the internal space of astrocytic processes and forced internal contents, including GFAP, into the rest space, making processes swollen. Glutamate-induced morphological changes in the process of astrocytes further support that glutamate induces large vesicles in astrocytes.
Fusion of Large Vesicles Coincides with SICs-To test whether astrocytes release glutamate through fusion of large vesicles, we puffed glutamate (50 mM) together with FM 1-43 (20 M). FM 1-43 has a hydrophilic head with two positive charges that limit its translocation through the bilayer membrane and a hydrophobic tail with a high affinity for lipid membranes (34,35). This property allows FM 1-43 to bind and stay on the nonpolar region (lipid tail) of the bilayer membrane only when the nonpolar region is exposed to FM 1-43. During vesicular fusion, the fused membrane (broken membrane) near the fusion pore exposes its nonpolar region to the extracellular medium where FM 1-43 is applied. When bound to membrane, FM 1-43 enhances its fluorescence remarkably. Therefore, appearance of an FM 1-43 fluorescence stain with high intensity in the vesicular membrane during vesicular collapse indicates formation of the exocytotic fusion pore. Before puffing, two-photon imaging showed a very small diffusion volume of FM 1-43 fluorescence (Fig. 4A, 0 s), because of leakage from the small pipette tip (pipette resistance14 -15 M⍀). FM 1-43 fluorescence in the extracellular space was detectable but was washed out quickly. After puffing glutamate once, a few large vesicles were observed (Fig. 4A, 120 s). Repeated puffing glutamate induced more large vesicles around the puffing pipette (Fig. 4A, 600 s, v). An FM 1-43-negative large vesicle was identified as a round structure with low fluorescence that was surrounded by FM 1-43 fluorescence (Fig. 4A, 600 s, v). The FM 1-43 stain, a small structure with the high intensity of FM 1-43 fluorescence, appeared after repeated puffs (Fig. 4A, 600 s, 1). As a control, in the presence of TTX, puffing glutamate-free ACSF with FM 1-43 induced neither FM 1-43-negative large vesicles nor FM 1-43 stains (Fig. 4B). Perfusion of slices with the iGluR antagonists, CNQX/APV, significantly reduced the number of glutamate-induced large vesicles (Fig.  4C, 600 s), suggesting that iGluRs are involved in the formation of large vesicles. Fig. 4D showed that two vesicles (v1 and v2) were collapsing, whereas a nearby vesicle (v3) was enlarging (supplemental movie 1). When vesicle 1 was collapsing, an FM 1-43 stain appeared in the vesicular membrane (Fig. 4D,  v1 and arrow), indicating the formation of the fusion pore.
To demonstrate that fusion of large vesicles resulted in glutamate release, we examined temporal coincidence of vesicular fusion with SICs in pyramidal neurons. Fig. 5A showed that the collapse of a large vesicle (Fig. 5A, panels i-iv, v) was accompanied by a high intensity FM 1-43 stain in the membrane of the collapsing vesicle (Fig. 5A, panels ii-iv, arrow, and supplemental movie 2). Traces of whole-cell recording in the patched pyramidal neuron (Fig. 5B, Pyr) and FM 1-43 fluorescence in the large vesicle (Fig.  5B, FM) showed that the start of the increase in fluorescence coincided with a SIC (Fig. 5B, Pyr, SIC). Coincident SIC was defined as a SIC that occurred within 4 s (imaging sample interval) following the start of vesicular collapse. Fig. 5C shows another experiment in which eight large vesicles underwent collapse (v1-v8; supplemental movie 3). The start of collapse of all eight large vesicles coincided with SICs in the recorded pyramidal neuron (Fig. 5C,  Pyr). The coincidence rate of fusion events with SICs in paired recordings (simultaneously recording FM 1-43 fluorescence and SICs) was 89% (49/55 fusion events), significantly different from unpaired recordings (SICs and FM fluorescence were separately recorded at different time, representing the random rate) (Fig. 5D, Coincidence, p Ͻ 0.01, 2 test), suggesting that fusion of large vesicles elicits SICs. The mean diameter of large vesicles measured before collapse was 3.8 Ϯ 0.1 m (Fig. 5D, Size, range 2-7 m, n ϭ 68). Above results suggest that glutamate induces formation of large vesicles that fuse with cytoplasmic membrane to exocytose glutamate into the extracellular space.
Inhibiting Glutamine Synthetase Induced Large Vesicles and SICs-To test whether endogenous astrocytic glutamate induces large vesicles and SICs, we used glutamine synthetase inhibitor, MSO, to block glutamate metabolism in astrocytes. In the central nervous system, glutamine synthetase is located in astrocytes (36). Blocking the glutamine synthetase by MSO leads to accumulation of glutamate in astrocytes. To balance the reduction in extracellular glutamine because of inhibition of astrocytic production of glutamine, 4 mM glutamine was added to the puffing solution that contained 20 mM MSO. Experiments were first performed in the absence of TTX, and 5 mM QX-314 was added to the pipette solution to block APs internally. Single application of MSO did not have direct effects on membrane currents and sEPSCs in pyramidal neurons (Fig.  6A, whole-cell). Repetitively puffing MSO slightly hyperpolarized neurons from Ϫ61.6 Ϯ 0.5 to Ϫ63.5 Ϯ 0.6 mV (p Ͻ 0.01, paired t test, n ϭ 10 cells) but did not change neuronal firing (Fig. 6A, V m ). After 20 puffs of MSO (40 min), synaptic release remained functional (Fig. 6A, eEPSC). MSO significantly increased the frequency and amplitude of SICs (Fig. 6, B and H,  MSO). Imaging of FM 1-43 fluorescence showed that MSO also induced formation of large vesicles (Fig. 6C, 960 s, v). Fusion of large vesicles was observed and coincided with SICs (Fig. 6, D and E). The coincidence rate of vesicular fusion with SICs was 82% (Fig. 6I, Coincidence, 18/22 fusion events), further supporting the idea that fusion of large vesicles causes SICs. Perfusion of slices with TTX before puffing MSO significantly reduced the number of MSO-induced large vesicles (Fig.  6, F and I, TTX 3 MSO) and SICs (Fig. 6, G and H, TTX 3 MSO), suggesting that AP-dependent synaptic release of glutamate significantly contributes to MSO-induced increase in endogenous [glutamate]. If first puffing MSO eight times to form large vesicles and then perfusing TTX, TTX did not significantly reduce the frequency and amplitude of SICs (Fig. 6H, MSO 3 TTX). These results suggest that APdependent synaptic release is only involved in the formation of large vesicles and storage of glutamate in vesicles but does not directly activate SICs.

4-AP Induces Ca 2ϩ -and Alexa Fluor-594-containing Large Vesicles in Astrocytes-
We have previously reported that bath application of the K ϩ -channel blocker, 4-AP, induces Ca 2ϩ -and Alexa Fluor-594-containing large vesicles in astrocytes (26). Here we describe more data on the appearance of Ca 2ϩ -and Alexa Fluor-594-containing vesicles in astrocytes induced by 4-AP. Simultaneous whole-cell recordings from pairs of pyramidal neurons and astrocytes showed that in the presence of TTX continuous perfusion of slices with 100 M 4-AP induced SICs in pyramidal neurons (Fig. 7A, Pyr, SIC) and concurrent aTCs in astrocytes (Fig. 7A,  Ast, aTC). To detect Ca 2ϩ -and Alexa Fluor-594-containing vesicles, we patched astrocytes with pipettes containing a filling solution with the Ca 2ϩ indicator, Fluo-4 potassium (50 M), Alexa Fluor-594 (100 M), and 50 mM glutamate. During application of 4-AP, we observed 27 Ca 2ϩ -and Alexa Fluor-594containing large vesicles in 18 of 38 patched astrocytes that showed sudden decreases (Ͼ0.5 ⌬F/F) in Ca 2ϩ and Alexa Fluor-594 fluorescence within 2 s, indicative of fusion events. In a representative cell (Fig. 7B), the intensity of both Alexa Fluor-594 (Fig. 7B, panels i-iv, red, arrow) and Ca 2ϩ fluorescence (Fig.  7B, panels i-iv, green, arrow) in a vesicle (Fig. 7B) increased along with enlargement of the vesicle, attained their highest value (Fig. 7B, panel iv, arrow), and then suddenly and remarkably decreased (Fig. 7B, panel v, arrow), suggesting the occurrence of fusion. Whole-cell recording detected an aTC (Fig. 7C, black trace, aTC) that immediately followed the disappearance of the vesicle within the imaging sample interval (2 s), suggesting that the disappearing vesicle contained glutamate. Threedimensional images of Ca 2ϩ and Alexa Fluor-594 fluorescence from another representative cell showed that two large vesicles were in the process of the astrocyte before fusion (Fig. 7D, panel i, 1) and disappeared 5 min later (Fig. 7D, panel ii, 1). All fusion events were examined by X-Y-Z scanning, and changes because of slice moving were excluded. The coincidence of vesicular fusion with aTCs was 75% (Fig. 7E, Coincidence). 25% of vesicular fusion did not coincide with aTCs, suggesting that a small number of large vesicles might contain low [glutamate] under these dye-loading conditions. The mean diameter of Alexa Fluor-594-loaded vesicles before fusion was 3.3 Ϯ 0.3 m (Fig. 7E, size), similar to the size of glutamate-induced large vesicles (Fig. 5D, size, 3.8 Ϯ 0.1 m).
We further examined whether fusion of large vesicles depended on cytoplasmic [Ca 2ϩ ], by adding the high affinity Ca 2ϩ chelator, BAPTA (20 mM), together with Fluo-4, Alexa Fluor-594, and 50 mM glutamate to the patch pipette filling solution for astrocytes. 4-AP-induced large vesicles were studied because local application of glutamate induces neighboring astrocytes to release glutamate that can be detected by BAPTAinfusing astrocytes. Because [BAPTA] in coupled neighbors by passing through gap-junction channels may not be high enough to clamp [Ca 2ϩ ], we applied the gap-junction blocker, carben-oxolone (100 M), in the bath before patching astrocytes, to inhibit the diffusion of glutamate from the patched cell into neighboring coupled cells. In the presence of intracellular BAPTA, the appearance of Alexa Fluor-594-loaded vesicles was still induced by 4-AP in 7 of 11 cells (Fig. 9C), but their fusion was blocked (Fig. 9C, red, 1-3), and whole-cell recordings showed no aTCs (Fig. 9D, black trace). Fusion rate (Fig. 9E, fusion rate, BAPTA) and the mean amplitudes of aTCs (Fig. 9E, aTC Amp, BAPTA versus Control) were significantly reduced by infusion of BAPTA, suggesting that astrocytic Ca 2ϩ is necessary for fusion of large vesicles with the plasma membrane and release of their contents, but not for vesicular formation and uptake of Alexa Fluor-594 into large vesicles. In the presence of BAPTA, Ca 2ϩ fluorescence in large vesicles still increased during their formation (Fig. 9, C and D, green). However, the green/red ratio (Fig. 9E, green/red, BAPTA) was significantly reduced, compared with that in the absence of BAPTA (Fig. 9E, green/red, Con), suggesting that the intensity of Ca 2ϩ fluorescence in large vesicles was attenuated by BAPTA (Fig. 9, C and D, green). The increase in vesicular [Ca 2ϩ ] in the presence of BAPTA suggests that, even though BAPTA buffered cytoplasmic free [Ca 2ϩ ], Ca 2ϩ was still actively transported into large vesicles against [Ca 2ϩ ] gradient.
In a previous study (26), we have demonstrated that fusion of IP 3 -induced large vesicles depends on SNARE proteins. Here we tested whether fusion of 4-AP-induced Ca 2ϩ -and Alexa Fluor-594-containing large vesicles also depends on SNARE proteins, by adding tetanus toxin (TeNT, 15 g/ml) to the patch pipette filling solution for astrocytes. In the presence of TeNT, Alexa Fluor-594-containing large vesicles were still induced by 4-AP in 4 of 7 cells (Fig. 10A, 1), but no fusion was observed (Fig. 10A, 400 s). Likewise, whole-cell recordings showed no aTCs in TeNT-infused astrocytes, consistent with blockade of an astrocytic SNARE-dependent fusion-release system responsible for astrocytic glutamate release (Fig. 10C, black trace). In summary, TeNT significantly reduced both the fusion rate (Fig. 10F, Fusion rate, TeNT) and mean amplitude of aTCs (Fig. 10F, aTC Amp, TeNT). Interestingly, TeNT also fully prevented the rise in [Ca 2ϩ ] fluorescence in Alexa Fluor-594containing large vesicles, including the [Ca 2ϩ ] increase during vesicular formation (Fig. 10, A and C, green, and F, green/red,  TeNT). As a control, scanning sections through the astrocytic soma showed that somatic cytoplasmic [Ca 2ϩ ] was not affected by TeNT (Fig. 10B). These results suggest that vesicular uptake of Ca 2ϩ and/or Fluo-4 is inhibited by TeNT. Because TeNT selectively cleaves vesicle-associated membrane proteins in SNARE complexes (39 -41), our data indicate that functional vesicle-associated membrane proteins are required for transport of Ca 2ϩ and/or Fluo-4 into high [Ca 2ϩ ] large vesicles.
Glutamate Is Transported into Large Vesicles by vGluTs-To test whether glutamate is transported into large vesicles by vGluTs, we added the vGluT inhibitor Rose Bengal (0.5 M) (13,42) to the patch pipette filling solution. In the presence of Rose Bengal, accumulation of Alexa Fluor-594 into large vesicles was still observed in 4 of 7 cells. Ca 2ϩ fluorescence in large vesicles was reduced by Rose Bengal by unknown mechanisms (Fig. 10,  D, green, and F, green/red, RB). Formation (Fig. 10D, red, Ͼ) and fusion (Fig. 10D, red, 1) of large vesicles were still observed, and multiple large vesicles in the astrocyte in Fig. 10D disappeared after a 10-min scanning (Fig. 10D, red, 600 s), but they were no longer associated with aTCs (Fig. 10E, black trace). Pooled data showed that vesicular fusion in the presence of Rose Bengal (Fig. 10F, Fusion rate, RB) was not significantly different from control events (Fig. 10F, Fusion rate, Con, p Ͼ 0.10, 2 test). However, both the coincidence rate of fusion events with aTCs (Fig. 10F, Coincidence, RB) and the amplitude of fusion-associated aTCs (Fig. 10F, aTC Amp, RB) in the presence of Rose Bengal were significantly reduced compared with controls (Fig. 10F, Coincidence and Amp, Con). These data suggest that glutamate is transported into large vesicles by Rose Bengal-sensitive vGluTs.

Extracellular [glutamate]
o plays an important role in mediating the interaction between neurons and astrocytes under both physiological or pathological conditions. Here we report that transient increases in local [glutamate] o induce SICs in CA1 pyramidal neurons through glutamate uptake and storage in large glutamate-containing vesicles in astrocytes and fusion and release of their contents. SICs can be induced in a number of ways, including high frequency stimulation of Shaffer collateral fibers (8), perfusion with Mg 2ϩ -free ACSF (7,8), perfusion with the A-type K ϩ channel blocker 4-AP (26), mechanical stimulation of astrocytes (7,38), perfusion with hypotonic solution (43), and glutamate transport inhibitors (7). These stimuli can all increase extracellular [glutamate] o , either by triggering neuronal release of glutamate or by inhibiting its uptake. Therefore, it is possible that glutamate-induced exocytosis of glutamate in this study could serve as a mechanism for induction of SICs by fiber stimulation, 4-AP, low Mg 2ϩ , and glutamate transporter inhibitors. However, it is also possible that SICs induced by mechanic stimulation, hypotonic solution, and other neurotransmitters might be mechanistically different.
Dose  (44,47), and release of a single vesicle increases the glutamate concentration in the synaptic cleft to about 2 mM (44 -46). Spatial and/or temporal summation of multivesicular releases (48,49) can produce a transient peak of synaptic cleft [glutamate] higher than 5 mM that could induce perisynaptic astrocytic processes to form large vesicles and release glutamate. In the absence of TTX, we occasionally observed spontaneous SICs during control periods (Fig. 6, B and C, Con), compared with no SICs in the presence of TTX (Fig. 6C, TTX), suggesting that synaptic release of glutamate can induce transient astrocytic glutamate release. The high frequency of SICs induced by [glutamate] o Ͼ20 mM probably only occurs under pathological conditions, such as epileptic seizures.
Local applications of 50 mM glutamate (eight puffs) did not change RMPs (Ϫ62.6 Ϯ 1.0 mV) or overshooting action potentials of pyramidal neurons (Fig. 1C, V m ), suggesting that pyramidal neurons can endure transient exposure to high glutamate. Whole-cell recording in astrocytes in the puffing area showed normal RMPs (Ϫ81 Ϯ 2.1 mV, n ϭ 5 cells) and relative constant responses to glutamate during repeated applications (Fig. 2C), suggesting that astrocytes can also endure transient exposure to high glutamate. Large vesicles and SICs were induced by puffing MSO (Fig. 6), further demonstrating that astrocytes can form large vesicles and then release glutamate in response to increased endogenous [glutamate].
Our results in this study demonstrated that extracellular glutamate can trigger astrocytes to exocytose glutamate through fusion of a large vesicle. Key pieces of evidence for this hypothesis include the following. 1) Local application of glutamate stimulated both the formation of large vesicles and the occurrence of SICs and aTCs. 2) EM studies showed that glutamate induced large vesicles in astrocytes. 3) Glutamate induced large vesicles in GFAP-eGFP-labeled astrocytic processes but not in apical dendrites of pyramidal neurons. 4) Concurrent vesicular collapse with the high intensity FM 1-43 stain on the membrane of collapsing vesicles indicated exocytotic fusion. 5) Fusion of large vesicles coincided with SICs. 6) Local application of the glutamine synthetase inhibitor, MSO, induced formation and fusion of large vesicles, and the latter coincides with SICs. 7) Glutamate induced the formation and fusion of high [Ca 2ϩ ] large vesicles in astrocytes. 8) Infusion of BAPTA or TeNT into astrocytes blocked both fusion of large vesicles and aTCs. 9) Infusion of vGluR inhibitor, Rose Bengal, inhibited aTCs but not formation and fusion of large vesicles. These data suggest that local increases in [glutamate] o can induce forma-tion of astrocytic glutamate-containing large vesicles that then fuse with the cytoplasmic membrane to exocytose glutamate. Our study also provides a useful method, using two-photon laser scanning microscopy and FM 1-43, to monitor the dynamic fusion process of large vesicles and formation of the fusion pore for research on vesicular release.
Because no large vesicles exist in deep layer astrocytes in control slices (Fig. 1, B and C, i), large vesicles were most likely incorporating from small undetectable vesicles, through enlargement and/or fusion during application of glutamate. Indeed, we observed large vesicles undergoing enlargement, which presumably requires addition of new membrane. The source of new membranes for the growth of vesicles may include fusion of multiple small vesicles (homotypic fusion) (50,51), and/or incorporation of new membrane into large vesicles by fusing with other types of organelles (heterotypic fusion). In some cases, we did observe fusion between two large vesicles, but in many other cases we did not, even though vesicles did enlarge, suggesting the incorporation of small invisible organelles to large vesicles.
An astrocytic vesicle with a small diameter (ϳ30 nm) similar to synaptic vesicles was previously reported to exocytose glutamate (12,17). However, the quantal of SICs is very large (30 -700 pA). There is no evidence showing large pools of these small vesicles in astrocytes. Therefore, it is unlikely that SICs are because of synchronized fusion of a great number of these small vesicles. Thus, even though the small astrocytic vesicle releases glutamate, it may not cause SICs. On the other hand, the large astrocytic vesicle (2-7 m) in this study explains well the large quantal of SICs.
In previous studies, IP 3  Transmitter-induced astrocytic release of glutamate has been reported previously to depend on astrocytic Ca 2ϩ signals (7, 8, 12, 13, 37, 38, 52). Here, the observation that fusion of large vesicles was blocked by intracellular BAPTA supports that Ca 2ϩ dependency of large vesicle fusion. The observation that high [Ca 2ϩ ] large vesicles are the major vesicle undergoing fusion and releasing glutamate implies that intravesicular Ca 2ϩ may be involved in fusion of these large vesicles, as it is for homotypic fusion between yeast vacuoles (53,54). By infusing the specific SNARE protein inhibitor, TeNT, into astrocytes, we demonstrated that fusion of high [Ca 2ϩ ] large vesicles also depends on astrocytic SNARE proteins. Therefore, glutamateinduced astrocytic release of glutamate is a Ca 2ϩ -and SNAREdependent process. By internal application of the vGluT inhibitor, Rose Bengal, into astrocytes, we found that Rose Bengal only blocked aTCs but not formation and fusion of large vesicles (Fig. 10, D and E). These results suggest that transporting glutamate into large vesicles and formation/fusion of large vesicles are separately controlled. Vesicular glutamate transport is controlled by vGluTs, whereas fusion of large vesicles depends on Ca 2ϩ and SNARE proteins; and formation of large vesicles may be controlled by glutamate receptors and other factors. Although we demonstrated that vGluTs were used to transport cytoplasmic glutamate into large vesicles, the driving force for vesicular transport of glutamate is still unknown. SICs induced by low Mg 2ϩ (7) or hypotonic solution (43) have been reported to be insensitive to the V-type H ϩ /ATPase inhibitor, bafilomycin A, leading to a thought of the channel-mediated mechanism underlying SICs (43). However, large astrocytic vesicles may be different from synaptic vesicles in the driving force for transporting glutamate into vesicles. Large astrocytic vesicles have the Ca 2ϩ pump (Fig. 9C) that can build up an electrical gradient across vesicular membrane for transporting glutamate into large vesicles. In supporting this possibility, in contrast to synaptic vesicles that are acridine orangepositive (55), all FM 1-43-negative large vesicles were acridine orange-negative, 3 implying that large astrocytic vesicles are not acidic.
Our results suggest that glutamate is a key factor for stimulating astrocytes to form glutamate-containing large vesicles, fusion of which causes SICs. However, formation of glutamatecontaining large vesicles requires [glutamate] o higher than 5 mM. This high [glutamate] o may occur in local neuronal circuits when neurons are close to overexcitation. Transient astrocytic release of glutamate by large vesicles may activate GABAergic synaptic inputs (9,27) and serve as a negative feedback control to balance the overexcitation of local circuits. Recently, Fiacco et al. (43) have reported that stimulation of astrocytic Ca 2ϩ signals by activating a G q -coupled receptor that is specifically expressed in astrocytes does not induce SICs. There are two possible mechanisms underlying their observation. One is that formation of glutamate-containing large vesicles in astrocytes is a precondition for inducing SICs, and without stimulating formation of glutamate-containing large vesicles with high [glutamate] o , astrocytic Ca 2ϩ signals alone cannot induce SICs. Another possibility is that SICs are not directly triggered by astrocytic cytoplasmic Ca 2ϩ but triggered by vesicular Ca 2ϩ that depends on astrocytic cytoplasmic Ca 2ϩ .
In this study, we demonstrated a glutamate-stimulated transient astrocytic glutamate release that is through fusion of a glutamate-containing large vesicle. This glutamate-stimulated astrocytic release of glutamate is well controlled by extracellular [glutamate] o and may serve as a negative feedback control of neuronal circuits by activating GABAergic synapses, but may, when the GABAergic inhibitory system is impaired, contribute to epileptic seizures.