Intracellular Calcium Oscillations in Astrocytes: A Highly Plastic, Bidirectional Form of Communication between Neurons and Astrocytes In Situ

Activation of mGluRs induces [Ca²⁺]_i increases in different cell types from developing hippocampus and visual cortex

Acute brain slices from the hippocampus and the visual cortex were loaded with the fluorescent Ca²⁺ indicator indo-1 by using the cell-permeant acetoxymethyl derivative and were analyzed by laser scanning confocal microscopy. As illustrated in Figure1A, pyramidal neurons from CA1 hippocampal region were well loaded and can be distinguished clearly by their typical shapes and the large size of their somas. Small cells with a stellate shape, like cells labeled 1 and 2, also were well loaded with indo-1 but could not be classified unambiguously solely on the basis of their morphological features. The series of pseudocolor images in Figure 1A illustrates the effects on the [Ca²⁺]_i induced on these cells by application of the mGluR agonist t-ACPD (5 μm; Palmer et al., 1989). Small cells like cells 1 and 2, as well as two of the pyramidal neurons labeled 3 and 4, displayed a [Ca²⁺]_i transient on t-ACPD addition. The kinetics of the [Ca²⁺]_ichanges, as expressed by the ratio between indo-1 emission at 405 and 485 nm, on continuous exposure to t-ACPD are reported in Figure 1C (left) and reveal the presence of periodic [Ca²⁺]_i oscillations in small cells 1 and 2 (green and red traces), whereas the pyramidal neurons (cells 3 and 4) displayed a single transient peak (black traces). A delayed response also was observed in six additional neurons in the field. In two of these the pattern of the response was oscillatory (data not shown). In a series of comparable experiments from CA1 region, the pyramidal neurons responding to t-ACPD (ranging from 2 to 10 μmin the various experiments) were 68.7 ± 8.9% (mean ± SE; range 17–100), and small cells were 76.3 ± 3.7% (range 52–100). In a few experiments from both hippocampus and visual cortex, concentrations higher than 10 μm were tested also. Att-ACPD concentrations of 50–100 μm the probability of observing [Ca²⁺]_ioscillations decreased, and most of the cells displayed a [Ca²⁺]_i rise, followed by a slowly decreasing plateau, suggesting that [Ca²⁺]_i oscillations were critically dependent on the concentration of t-ACPD. In the experiments described below, for the analysis of [Ca²⁺]_i oscillations we usedt-ACPD concentrations between 2 and 10 μm.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Stimulation of mGluRs induces [Ca²⁺]_i oscillations in hippocampal cells. A, Time series of pseudocolor images of the [Ca²⁺]_i changes occurring in indo-1-loaded cells from CA1 hippocampal region of a young rat (at postnatal day 8) after perfusion of the slice with 5 μmt-ACPD. The sequence shows the [Ca²⁺]_i transient in two small-sized cells (labeled 1 and 2) and two pyramidal neurons (labeled 3 and 4). TheR405/485 is displayed as a pseudocolorscale. Sampling rate, 3 sec; scale bar, 10 μm.B, Pseudocolor images (a–d) from the same field illustrating the early, sustained [Ca²⁺]_i increase in neurons, including neurons labeled 3 and 4 inA, and the transient, delayed response in small cells, including cells labeled 1 and 2 inA, after bath application of 60 mm KCl. Symbols and conditions are as in A. C, Kinetics of the [Ca²⁺]_i changes in the cells labeled 1–4 after t-ACPD and KCl stimulation, as expressed by the ratio between indo-1 emission wavelength at 405 and 485 nm. Letters a–d correspond to images a–d in B.

Astrocyte identification

Small cells displaying [Ca²⁺]_ioscillations from both visual cortex and hippocampal CA1 and stratum radiatum regions had an astrocyte-like morphology with a mean soma diameter of 5–10 μm and numerous radiating processes. These cells could be distinguished easily from pyramidal neurons. On the basis of pure morphological criteria, however, astrocytes hardly can be distinguished from small nonpyramidal neurons with a stellate or bipolar shape that are present in both areas (Ramon y Cajál, 1911). We first tried to identify astrocytes via glial fibrillary acidic protein (GFAP) immunostaining performed at the end of the recording session at the confocal microscope. Several cells with astrocyte-like morphology that displayed [Ca²⁺]_i oscillations in response tot-ACPD could, indeed, be identified retrospectively as astrocytes by anti-GFAP immunostaining. Because of the variability in the intensity of the GFAP staining in different slices and among cells and to the overall alteration in the cell morphology after fixation and permeabilization procedures, in <10% of the cells in the recording field was the identification unambiguous. An alternative approach thus was developed to distinguish functionally each astrocyte and neuron present in the recording field. The experiments described below refer to CA1 hippocampal region, although identical results were obtained in the visual cortex. We took advantage of the observation that, in mixed neuron–astrocyte cultures, neurons, as identified by immunocytochemical criteria, responded promptly with a [Ca²⁺]_i increase to depolarization induced by 60 mm K⁺, whereas none of the immunocytochemically identifiable astrocytes was sensitive to this treatment (data not shown). We thus applied the same protocol to brain slices to analyze whether neurons and presumed astrocytes responded differently, as in culture, to stimulation with 60 mmK⁺. As shown in Figure 1B, in response to this treatment all cells in the field showed large [Ca²⁺]_i increases. The onset of the response from the two cell populations was, however, clearly different. In the first population, mainly composed of cells with an evident pyramidal shape, the perfusion with high K⁺ induced a prompt [Ca²⁺]_i increase (Fig.1Bb). In the majority of neurons this first increase was followed by a further [Ca²⁺]_ielevation (Fig. 1C, right). In the second population, composed of small-sized cells with astrocyte-like morphology, the [Ca²⁺]_i increase occurred many seconds (13.3 ± 1.6 sec; mean ± SE;n = 5) after that of pyramidal neurons at approximately the same time of the second [Ca²⁺]_ipeak in neurons (Fig. 1Bc,C,right). A delay in the response to 60 mmK⁺ was never observed in cells having a pyramidal shape. The kinetics of the recovery were also different: in cells that displayed the prompt response to depolarization, the [Ca²⁺]_i remained elevated for several minutes and recovered to basal levels quite slowly, whereas in cells that displayed the delayed response the kinetics of the [Ca²⁺]_i decrease were faster, and basal [Ca²⁺]_i levels were recovered in ∼40 sec (Fig. 1Bd,C, right). Occasionally, these cells displayed repetitive [Ca²⁺]_i transients that resemble the typical t-ACPD-induced [Ca²⁺]_i oscillations (data not shown). A delayed response pattern is consistent with a secondary response to glutamate massively released by depolarized synaptic terminals. Indeed, the response of small-sized cells, but not that of pyramidal neurons, to the stimulation with 60 mm K⁺ was blocked in slices incubated for 40 min in TeNT (100 μg/ml), a well known, potent inhibitor of neuronal exocytosis (data not shown; see Fig. 7E). Given that all neurons express at least one subtype of functional voltage-gated Ca²⁺ channels, the delayed response to 60 mm K⁺ could be attributed only to non-neuronal cells, such as astrocytes, that either lack these channels or express voltage-gated Ca²⁺ channels at such low density (Barres et al., 1990) to induce negligible [Ca²⁺]_iincreases, at least under our experimental conditions. We also stimulated slices with 100 μm NMDA. Given that NMDA receptors in astrocytes are most likely absent or nonpermeable to Ca²⁺ (Müller et al., 1993), NMDA should induce a [Ca²⁺]_i increase in neurons, whereas astrocytes either should not respond or should display a delayed [Ca²⁺]_i increase because of a secondary release of glutamate. Stimulation with NMDA (100 μm) induced in the two populations a pattern of [Ca²⁺]_i increase similar to that observed after perfusion with 60 mm K⁺(Fig. 2). Interestingly, NMDA did not induce the biphasic response observed after depolarization with high K⁺. This latter finding supports the hypothesis that the second peak in the [Ca²⁺]_iincrease induced in neurons by high K⁺ stimulation is attributable to glutamate released by depolarized terminals.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Kinetics of [Ca²⁺]_i in cells from the visual cortex in response to NMDA. Presumed astrocytes (n= 7; solid lines) from the visual cortex of a 5-d-old rat display a delayed [Ca²⁺]_i increase to NMDA (100 μm) with respect to the prompt response of pyramidal neurons (n = 5; dashed lines).

To confirm that this type of response pattern can be used as a functional tool to identify astrocytes in situ, we first distinguished neurons and astrocytes on the basis of their biophysical properties by patch-clamp recording while indo-1 diffused into the cell through the patch pipette. Then the pipette was gently withdrawn, and the [Ca²⁺]_i response finally was analyzed at the confocal microscope. Figure3A illustrates an astrocyte and a pyramidal neuron (closed arrows) after intracellular injection of indo-1 (bright field is shown in Fig. 3B). Neurons were identified electrophysiologically by their action potential discharges on depolarizing current pulses (Fig.3D) and astrocytes by their absence (Fig. 3C) and highly negative resting potentials (<75 mV). In addition, indo-1 staining was observed not only in the patched astrocyte but also in other small cells, like those indicated in Figure 3A byopen arrows, indicating coupling via gap junctions. Small-sized cells were never stained with indo-1 when only neurons were injected (n = 12), excluding the existence of communication between these two types of cells, at least in the brain regions that were analyzed. As illustrated in the pseudocolor images of Figure 3Ea–c, after perfusion with 60 mm K⁺ the neuron displayed a prompt [Ca²⁺]_i increase (Fig.3Eb), whereas the response of the astrocyte appeared several seconds after that of the neuron (Fig. 3Ec). A similar response was detected in another small cell (open arrow) dye-coupled with the injected astrocyte. The kinetics of the [Ca²⁺]_i increase in the electrophysiologically classified neuron (dashed line) and astrocyte (continuous line) are reported in Figure3E (right). As previously observed in cells from slices loaded with indo-1/AM, the pyramidal neuron displayed a biphasic response to stimulation with 60 mm K⁺, and the second [Ca²⁺]_i increase occurred at the time of the [Ca²⁺]_iincrease in the astrocyte. After removal of KCl, the slice was perfused with t-ACPD (5 μm). The sequence of images in Figure 3F (left) and the kinetics of the [Ca²⁺]_i changes (Fig. 3F,right) demonstrate that the small cell responded tot-ACPD with [Ca²⁺]_ioscillations, whereas the neuron was unresponsive. Identical results were obtained from two additional experiments.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Functional identification of small-sized cells and pyramidal-shaped cells from CA1 hippocampal region as astrocytes and neurons, respectively. A, Pseudocolor image illustrating one pyramidal neuron and one astrocyte (white arrows) from CA1 hippocampal region of a 10-d-old rat injected with indo-1-free acid included in the patch pipette at 500 μm. This image was taken at the end of the recording session at the confocal microscope, using high laser power to permit visualization of four astrocytes (black arrows) dye-coupled with the injected astrocyte. This accounts for the saturation of the signal at the center of the cell bodies (black) in the injected astrocyte and neuron. Scale bar, 10 μm. B, Bright field from the same region as in A. The indo-1-injected neuron and astrocyte are indicated by black arrows. The latter cell lies on a slightly different focus plane. C, Absence of action potential discharges in the astrocyte after hyperpolarizing and depolarizing current pulses of increasing amplitude and 100 msec duration. D, Action potential discharge in the pyramidal neuron on a depolarizing current pulse of 200 pA and 100 msec duration.E, Pseudocolor images of the [Ca²⁺]_i changes after 60 mm KCl stimulation. Letters a–c in the plot refer to the images a–c. The laser power was set to a level that allowed us to visualize both the neuron and the less-loaded astrocyte. As a consequence, the fluorescence at 485 nm at the soma of the neuron reached saturation; accordingly, yellow does not correspond to the real value of the R405/485. The [Ca²⁺]_i change after both 60 mm KCl and 5 μmt-ACPD was, therefore, measured from a portion of the dendrite (dashed lines box). Sampling rate, 1 sec; scale bar, 10 μm.F, Time series of pseudocolor images of the [Ca²⁺]_i changes aftert-ACPD stimulation. The injected astrocyte displayed periodic [Ca²⁺]_i oscillations ont-ACPD stimulation (right). Sampling rate, 3 sec. One of the dye-coupled astrocytes also responded tot-ACPD (black arrow).

In conclusion, on the basis of these observations, the delayed increase in the response to 60 mm K⁺ reasonably can be used as a criterion to distinguish astrocytes from neuronsin situ. In the various experiments that will be described below, at the end of each recording session astrocytes and neurons were identified according to the different kinetics of their response to stimulation with 60 mm K⁺. It cannot be excluded that other non-neuronal cells such as oligodendrocytes might respond to the various stimuli with a pattern similar to that of astrocytes, but their number in the brain regions we analyzed is known to be much lower than that of the astrocytes.

Repetitive activation of the mGluR induces long-term changes in [Ca²⁺]_i oscillations in astrocytes

We previously reported that astrocytes from the visual cortex in culture displayed a long-term modification of their response, i.e., an increased frequency of [Ca²⁺]_ioscillations, on repetitive stimulation with l-glutamate (Pasti et al., 1995). We thus investigated whether repetitive stimulation with t-ACPD can induce a similar potentiation in the response of astrocytes from acute brain slices. Oscillations in neuronal cells will be considered separately. Figure4A shows the oscillatory response from one astrocyte on three successive 5 μmt-ACPD stimulations applied with a time interval of 10 min. The second stimulation induced [Ca²⁺]_i oscillations of increased frequency, although their amplitude and shape did not change significantly (Fig. 4A). A third stimulation resulted in a further increase in oscillation frequency. The potentiation of the response was observed mainly in astrocytes displaying a low frequency of oscillations during the first challenge with t-ACPD, as in the case reported in Figure 4A, whereas those oscillating initially at high frequency (Fig. 4B) were, in general, not potentiated. Results are summarized in Figure4C, which reports the relative change in oscillation frequency occurring in each cell on the second (filled symbols) and third (open symbols) t-ACPD stimulations as a function of the oscillation frequency on the firstt-ACPD stimulation. The average increase in oscillation frequency on repetitive stimulation obtained from a subpopulation of cells comprising astrocytes oscillating initially at a frequency ≤1 was more pronounced than that obtained from the whole population of cells (Fig. 4D, Table1). Similar results were obtained from visual cortical astrocytes (Table 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Long-term changes of the astrocyte response tot-ACPD. A, Progressive increase in the frequency of [Ca²⁺]_i oscillations on three successive stimulations with 5 μmt-ACPD in one astrocyte oscillating at low frequency after the first stimulation. The continuous line at thebottom of the traces indicates the application oft-ACPD. The time interval between stimulations was 10 min. B, The frequency of [Ca²⁺]_i oscillations on three successive bath applications of 5 μmt-ACPD did not increase in one astrocyte oscillating at high frequency during the first stimulation. Conditions and labels are as in A. C, The frequency of oscillations in each cell, as measured during the first t-ACPD pulse, is plotted as a function of the relative change in oscillation frequency in the second (filled symbols) and third (open symbols) with respect to the first pulse.D, Average frequency of oscillations after the threet-ACPD stimulations (I, II, andIII) from all astrocytes (open bars) and from a subpopulation of astrocytes comprising cells oscillating at a frequency ≤1 (striped bars); *p < 0.05; **p < 0.001 (paired t test). The frequency of [Ca²⁺]_i oscillations in this as well as in the other figures is expressed as the number of [Ca²⁺]_i peaks per minute.

In contrast to the results obtained in culture, in calcium-free medium astrocytes in situ failed to respond to t-ACPD with an oscillatory pattern and displayed, in most cases, a single [Ca²⁺]_i rise (data not shown). Only a few cells (5 of 30) exhibited two or three [Ca²⁺]_i transients under these conditions. The subsequent addition in the perfusate of 2 mm Ca²⁺ resulted in an immediate increase in the [Ca²⁺]_i that recovered to basal levels in a relatively short time. At this extracellular Ca²⁺ concentration the normal response tot-ACPD stimulation was restored (n = 13).

The change in oscillation frequency is a relatively long-lasting phenomenon. In fact, we observed a significant increase in oscillation frequency when the second t-ACPD stimulation was applied after a time interval of 3 hr (see Table 1).

t-ACPD is known to activate all the mGluR subtypes, although with different affinity. To identify the mGluR subtype that mediates [Ca²⁺]_i oscillations in astrocytes, we used several known blockers of group 1 mGluRs. The competitive antagonist 4CPG (10–500 μm; Watkins and Collingridge, 1994) did not block t-ACPD-induced oscillations in a total of 30 astrocytes from the hippocampus and 11 from the visual cortex. The mGluR1/5 antagonist AIDA (50–200 μm;n = 14, two experiments; Pellicciari et al., 1995), as well as the noncompetitive antagonist l-AP3 (30 μm; n = 11, two experiments), was also ineffective. In contrast, the nonspecific mGluR antagonist MCPG at 1 mm concentration blocked t-ACPD-induced [Ca²⁺]_i increases.

Neuronal stimulation induces [Ca²⁺]_i oscillations in astrocytes

By applying current pulses to afferent projections through a bipolar tungsten electrode, we next investigated whether stimulation of neuronal afferents could trigger [Ca²⁺]_i oscillations in astrocytes. The sequence of images in Figure5A corresponds to the portion of the trace highlighted by the dashed lines box in Figure5B and illustrates the somatic [Ca²⁺]_i transients of a pyramidal neuron (cell 1) in response to stimulation at 0.16 Hz of Schaffer collaterals. Between stimuli, [Ca²⁺]_i recovered to baseline levels (see also top trace in Fig. 5B). A [Ca²⁺]_i rise out of synchrony with the stimulus was observed in cell 2 that was identified retrospectively as an astrocyte. On continuous neuronal stimulation this cell displayed repetitive transients with a relatively regular periodicity (Fig.5B). The [Ca²⁺]_i increase in astrocytes occurred, in general, with a delay of 10–16 sec with respect to that of adjacent neurons, although a delay of <2 sec was, in some cases, observed when the train of stimuli was applied at a frequency of 1 Hz. In all cases, the [Ca²⁺]_i oscillations in the astrocytes were clearly out of synchrony with respect to the timing of the electrical stimulation. Interestingly, [Ca²⁺]_i oscillations with a frequency higher than that at the soma were observed at the level of the astrocyte process labeled 3 in Figure 5A (see bottom trace in Fig. 5B). Figure 5, C andD, illustrates an additional example. The sequence of images (Fig. 5C) shows that, on stimulation of Schaffer collaterals at 0.16 Hz, the [Ca²⁺]_i increase initially was restricted to the process only (arrow), whereas the second [Ca²⁺]_i increase spread through the whole cell body (see also inset of Fig.5D). The kinetics of the [Ca²⁺]_i changes from the process and the soma are compared in Figure 5D (the trace within the dashed lines box corresponds to the sequence of images in C). Oscillations triggered by stimulation of afferent fibers were observed in a low number of astrocytes corresponding to ∼15% of indo-1-loaded astrocytes.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Neuronal activity-dependent [Ca²⁺]_i oscillations in astrocytes.A, Time series of pseudocolor images illustrating the [Ca²⁺]_i changes in one pyramidal neuron (labeled 1) and one adjacent astrocyte (labeled2) from CA1 hippocampal region of a 8-d-old rat after neuronal stimulation at 0.16 Hz, i.e., a series of six pulses at 30 Hz applied every 6 sec. Label 3 indicates one of the astrocyte processes. The sequence of images (time interval, 2 sec) corresponds to the portion of the traces shown in B and is highlighted by the dashed lines box. Because the pyramidal neuron and the astrocyte were localized at a different depth and the plane of focus was set to visualize the astrocyte, the neuron looks smaller than the astrocyte. The real mean diameter of the neuron was 16.9 μm, whereas that of the astrocyte was 10 μm. Scale bar, 10 μm. B, Kinetics of the [Ca²⁺]_i changes in the cells and the process shown in A after two successive episodes of neuronal stimulation applied with 5 min intervals. The second episode of stimulation was performed in the presence of MK801 and NBQX, both at 50 μm. C, Time series of pseudocolor images illustrating the [Ca²⁺]_ichanges in an astrocyte as measured at the level of one process (filled arrow) and the cell body after neuronal stimulation. The sequence of images (time interval, 2 sec) corresponds to the portion of the trace highlighted by the dashed lines box in D. Scale bar, 10 μm. D, Kinetics of the [Ca²⁺]_i oscillations in the process and the soma of the astrocyte shown in Cduring neuronal stimulation at 0.16 Hz. In the inset, the sequence of points representing the R405/485 values at the process (filled symbols) and the soma (open symbols) corresponds to the sequence of images inC.

These results suggest that glutamate released by synaptic terminals is responsible for the [Ca²⁺]_i increase in astrocytes. However, the stimulation of presynaptic fibers can trigger action potential discharges in target neurons and results in a secondary release of the neurotransmitter. The [Ca²⁺]_i change in astrocytes might, therefore, originate also from glutamate released at synapses that belong to intrinsic connections among target neurons. To investigate this point, we blocked the activation of postsynaptic neurons by perfusing slices with the AMPAR antagonist NBQX (50 μm) and the NMDAR open channel blocker MK801 (50 μm; Watkins et al., 1990). Under these conditions the electrical stimulus failed to produce any [Ca²⁺]_i change in neurons that were responsive before the application of the blockers, whereas astrocytes were still responsive with a pattern of oscillations similar to that observed in the absence of glutamate receptor antagonists (Fig.5B). On the contrary, perfusion with the Na⁺ channel blocker tetrodotoxin (TTX; 5 μm) abolished the response from both types of cells (15 neurons and 7 astrocytes in CA1 region from two hippocampal slices; 10 neurons and 4 astrocytes from two visual cortical slices). The perfusion for 15 min with the mGluR antagonist MCPG (1 mm) abolished the astrocyte response to electrical stimulation of afferents, whereas that from neurons was unchanged (14 neurons and 8 astrocytes from four hippocampal slices).

Neuronal stimulation modulates the frequency of oscillations in astrocytes

Having demonstrated that neuronal activity can trigger [Ca²⁺]_i oscillations in astrocytes from both visual cortex and hippocampus, we next investigated whether the frequency of oscillations could change according to the firing rates of neuronal afferents. We analyzed the oscillatory response of astrocytes after a train of stimuli applied first at low frequency (0.1–0.2 Hz) and low intensity (50–100 pA) and then at either increased frequency (0.3–1 Hz) or higher intensity (200–500 pA). Figure 6 shows that an increase in either the frequency (A) or intensity (B) of the stimulus resulted in a clear increase in the frequency of [Ca²⁺]_i oscillations in astrocytes. The mean frequency before and after the change in the stimulus is reported in Figure 6, A and B (right). When the frequency of the stimulus was increased, the average frequency of oscillations (± SE) changed from 1.41 ± 0.23 to 2.16 ± 0.24 peaks/min. When the percentage of increase in oscillation frequency in the second with respect to the first stimulation from each individual astrocyte was taken into account, the average increase (± SE) corresponded to 115 ± 44.1%. A similar increase was obtained when the stimulus was applied at higher intensity (mean ± SE; 106 ± 27.5%, n = 20). It is noteworthy that changes in the pattern of the electrical stimulus induced either an increased amplitude of the [Ca²⁺]_irise in neurons that were already responsive, as in the case of the two neurons in A and B, or the appearance of [Ca²⁺]_i increased in neurons and astrocytes that were initially unresponsive (data not shown), suggesting the recruitment of additional presynaptic inputs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Neuronal activity-dependent modulation of [Ca²⁺]_i oscillation frequency in astrocytes. A, [Ca²⁺]_ioscillations in one astrocyte (thick line) after neuronal stimulation. The frequency of [Ca²⁺]_i oscillations rapidly increased after the shift in the frequency of neuronal stimulation from 0.2 to 1 Hz. The response of an adjacent neuron to each pulse is reported also (thin line). The bar histogram reports the average oscillation frequency from a data base of 10 astrocytes in the first (stimulus frequency range, 0.1–0.2 Hz) and second (stimulus frequency range, 0.3–1 Hz) phase of stimulation; **p < 0.001. B, [Ca²⁺]_ioscillations in one astrocyte (thick line) after neuronal stimulation at 1 Hz. The frequency of [Ca²⁺]_i oscillations rapidly increased after the shift in the intensity of the stimulus from 100 to 200 pA. The increase in the [Ca²⁺]_i from an adjacent neuron is reported also (thin line). The bar histogram reports the average oscillation frequency from a data base of 20 astrocytes in the first and second phase of stimulation. In this second phase the frequency of the stimulus was increased from the initial 100 pA to a minimum of 200 pA and a maximum of 500 pA in the various experiments; **p < 0.001.C, The frequency of [Ca²⁺]_i oscillations that follows a first episode of neuronal stimulation increased (1.05 vs 2.17 peaks/min) after a second episode of stimulation at unchanged intensity (100 pA) and frequency (0.5 Hz) applied after a time interval of 10 min. D, The frequency of [Ca²⁺]_i oscillations in each cell, as measured in the first episode of neuronal stimulation, is plotted as a function of its relative change in the second episode of stimulation. On the right, the bar histogram summarizes data at the first (I) and second (II) stimulation from a total of 22 astrocytes; *p< 0.05. E, Relative change in the frequency of [Ca²⁺]_i oscillations in the second with respect to the first t-ACPD stimulation in astrocytes that displayed (filled symbols andbars) or did not display (open symbolsand bars) [Ca²⁺]_ioscillations during an episode of neuronal stimulation applied between the two t-ACPD stimulations. The bar histogram reports the average oscillation frequency in the two astrocyte populations; **p < 0.001.

In the experiment presented in Figure 4, we showed that successive t-ACPD stimulations resulted in an increase in the frequency of [Ca²⁺]_i oscillations in astrocytes. The question then arises as to whether a similar form of potentiation in astrocytes can be elicited by repetitive stimulation of afferent fibers. Figure 6C shows the response from a single astrocyte for which the frequency of [Ca²⁺]_i oscillations changed from 1.0 at the first to 2.1 at the second series of pulses. Similar to what was observed with repetitive t-ACPD stimulation, the potentiation of the response was observed mainly in cells having a low frequency of oscillations at the first pulse (Fig.6D), although the average increase in oscillation frequency (65.4 ± 18.3%; Fig. 6D,right) was more pronounced than that observed on two successive t-ACPD applications (see Table 1). The higher efficacy of neuronal stimulation in inducing the potentiation of the astrocyte response was investigated further. The experimental protocol was as follows: after the first and before the second t-ACPD stimulation, we applied a series of 0.5–1 Hz stimuli to CA1 region afferent projections. As discussed above, only a percentage of the astrocytes displayed oscillations after neuronal stimulation. We thus had the ability to compare in the same slice the increase in oscillations frequency that follows the second t-ACPD stimulation in two astrocyte populations: the first composed of astrocytes that displayed oscillations on neuronal stimulation between the two successive t-ACPD applications and the second composed of astrocytes that were not responsive to neuronal stimulation. Figure 6E reports the relative change in frequency in each astrocyte (left) and the mean values from the two subpopulations (right). The increase in the frequency of [Ca²⁺]_i oscillations after the second t-ACPD application was higher in those astrocytes that showed oscillations on neuronal stimulation (Fig.6E, closed symbols and bars) with respect to that of astrocytes that did not respond to this latter challenge (open symbols and bars). It is noteworthy that the average increase at the second t-ACPD application in astrocytes responsive to neuronal stimulation was higher than that observed after three successive t-ACPD applications (114% vs 44.7%; Table 1).

[Ca²⁺]_i oscillations in astrocytes are followed by [Ca²⁺]_i oscillations in neurons

Besides the effects on astrocytes, t-ACPD induced in a number of CA1 pyramidal neurons either a single [Ca²⁺]_i transient or [Ca²⁺]_i oscillations. These [Ca²⁺]_i changes could be attributable to (1) direct stimulation of neuronal mGluRs (Stratton et al., 1990) and (2) t-ACPD-induced release of glutamate from astrocytes with activation of neuronal ionotropic glutamate receptors (iGluR), i.e., AMPA and NMDARs. Indeed, cultured astrocytes have been reported to release glutamate and excite adjacent neurons (Parpura et al., 1994a; Jeftinija et al., 1996). Figure 7,A and B, shows a typical oscillatory response induced by 10 μmt-ACPD in one neuron (A) and one adjacent astrocyte (B). On washout of the antagonist, the [Ca²⁺]_i returned to basal levels in both cell types. A second challenge with t-ACPD (10 μm) was applied in the presence of NBQX andd-AP5 (both at 50 μm), specific antagonists of AMPAR and NMDARs, respectively. Under these conditions, in this as well as in a number of other neurons (14 of 49), the response tot-ACPD was inhibited completely but recovered on washout of the antagonists (Fig. 7A), whereas the astrocyte response was not changed significantly (Fig. 7B). In the remaining 35 neurons, the response to the second t-ACPD stimulation, performed in the presence of NBQX and d-AP5, either was decreased significantly in amplitude, as in the case reported in Figure7C, or was unchanged. As a mean, the amplitude of the [Ca²⁺]_i elevation induced in neurons by 10 μmt-ACPD applied in the presence of NBQX/d-AP5 was reduced significantly (Fig. 7D). The relative change in the amplitude of the [Ca²⁺]_i increase in each neuron at the second (in the presence of the AMPAR and NMDAR blockers) with respect to the first t-ACPD stimulation is reported in Figure7D. It is noteworthy that no reductions of the response after two successive t-ACPD stimulations were observed in control neurons (n = 32; Fig. 7D). These results indicate that [Ca²⁺]_ielevations in neurons can be elicited, partially or totally, by glutamate released in response to the t-ACPD challenge. To clarify whether glutamate released after t-ACPD stimulation derived from afferent fibers, neurons, or astrocytes, we incubated the slice with TeNT (100 μg/ml). TeNT is known to be a highly specific blocker of neurotransmitter secretion in neurons (Calabresi et al., 1989; Schiavo et al., 1992). Its action is exerted on the vesicle-associated membrane protein VAMP/synaptobrevin (Schiavo et al., 1992), one of the components of the neuroexocytosis apparatus, after cell internalization probably via the recycling of synaptic vesicles (Matteoli et al., 1996). Although VAMP/synaptobrevin is expressed in many cell types besides neurons, the lack of TeNT effects on other cell types probably depends on the absence of a receptor for the neurotoxin (Parpura et al., 1994b; Rossetto et al., 1996). By recording EPSCs from CA1 neurons in response to stimulation of the Schaffer collaterals before and after perfusion of the slice with TeNT (100 μg/ml), we obtained evidence for the inhibitory effects of TeNT on synaptic transmission. As shown in Figure 7E, the EPSC first was reduced in amplitude and then abolished after 20 min of TeNT treatment. Despite changing the position of the stimulating electrode and increasing the stimulus intensity, we obtained no response from this and three other neurons of the same slice, as well as from neurons of an additional slice, tested after 40 min of incubation with TeNT. In toxin-treated slices the effects of t-ACPD on both neurons and astrocytes was unchanged, as compared with controls: it stimulated [Ca²⁺]_i oscillations in astrocytes and [Ca²⁺]_i elevations in neurons. In 6 of 35 t-ACPD-responsive neurons, the response was inhibited completely by NBQX/d-AP5 application, whereas in the remaining 29 neurons the amplitude was either reduced significantly or unchanged. As a mean, the reduction of the response of neurons to t-ACPD by NBQX and d-AP5 in toxin-treated slices was not significantly different as compared with that obtained from neurons in toxin-untreated slices (Fig.7D, right). The relative change in the amplitude of the [Ca²⁺]_i increase in each neuron after the second with respect to the first t-ACPD stimulation also is reported in Figure 7D(right).

The inhibitory effect of TeNT on neurotransmitter exocytosis was confirmed in each slice used at the confocal microscope by the following experimental observations (data not shown in figures): (1) electrical stimulation of Schaffer collaterals failed to produce [Ca²⁺]_i increases in both neurons and astrocytes; (2) small cells, i.e., presumed astrocytes, did not display any [Ca²⁺]_i increase after 60 mm K⁺ stimulation; and (3) pyramidal neurons still responded to K⁺-induced depolarization with an early [Ca²⁺]_i increase, but their response did not display the biphasic pattern observed in controls (see Figs. 1C, 3E) and recovered to [Ca²⁺]_i basal levels much faster than in neurons from toxin-untreated slices.

In the great majority of neurons (16 of 20) from untreated and TeNT-treated slices for which the [Ca²⁺]_i increase was entirely attributable to the t-ACPD-induced release of glutamate, the pattern of the response was clearly oscillatory (see, for example, Fig.7A).

When a second t-ACPD challenge was applied in the presence of the mGluR antagonist MCPG (1 mm), the [Ca²⁺]_i elevations observed after the first t-ACPD stimulation were abolished in all responsive astrocytes (n = 8) and neurons (n = 12).

Long-term changes in oscillation frequency mediated by t-ACPD

We previously reported that successive stimulations of the mGluR in cultured astrocytes induced a long-lasting increase in [Ca²⁺]_i oscillation frequency (Pasti et al., 1995). Here we demonstrate that repetitive activations of mGluRs by t-ACPD induce a similar potentiation in the [Ca²⁺]_i response of astrocytes in acute brain slices from both CA1 hippocampal region and visual cortex. The potentiation was induced rapidly and was relatively long-lasting. In contrast to the results obtained in cultured astrocytes (Pasti et al., 1995), in the absence of extracellular Ca²⁺, astrocytes either failed to oscillate or displayed rapidly fading oscillations. The effects of t-ACPD on astrocytes probably are mediated by the mGluR5 subtype (Romano et al., 1995). Antagonists of group 1 mGluRs with a preference specificity for mGluR1, failed, however, to affect t-ACPD-induced oscillations, whereas MCPG, a nonspecific mGluR antagonist, blocked the response. The mGluR subtype responsible for [Ca²⁺]_ioscillations thus remains to be established.

Modulation of the astrocyte response by neuronal activity

One of the most striking observations of this study is that astrocytes are extremely sensitive to synaptic activity. Indeed, the pattern of their [Ca²⁺]_i oscillations in response to neuronal stimulation changed according to the level of synaptic activity; when the frequency or intensity of the stimulus applied to presynaptic afferents was increased, the frequency of astrocyte oscillations was increased also. It is noteworthy that the increase in the stimulus rate determines a higher firing rate of neuronal afferents, whereas the increase in intensity may result in the recruitment of additional fibers that were not stimulated initially. The change in stimulus intensity or frequency probably causes an increased glutamate concentration in the extrasynaptic space and/or the activation of a higher number of localized [Ca²⁺]_i transients along processes of individual astrocyte that may account for the increased oscillation frequency in the astrocytes. These results provide a mechanism for a highly regulated and dynamic control on [Ca²⁺]_i oscillation frequency that depends on the integration of the Ca²⁺ signal deriving from these multiple sites of activation.

The inhibition of astrocyte [Ca²⁺]_ioscillations by TTX excludes that the response of astrocytes after the application of the stimulus to neuronal afferents could be attributable to a direct mechanical or electrical stimulation of the glial cells (Charles et al., 1991; Nedergaard, 1994). On the other hand, the finding that the mGluR antagonist MCPG blocked the astrocyte response to neuronal stimulation provides convincing evidence that synaptically released glutamate is responsible for astrocyte [Ca²⁺]_i oscillations. Further support for the synaptic origin of the stimulus inducing the [Ca²⁺]_i changes in astrocytes derives from the observation that in the presence of iGluR blockers the electrical stimulus failed to produce any response from neurons, whereas it did induce an oscillatory response in astrocytes. This result also suggests that activation of afferent fibers is sufficient for inducing astrocyte oscillations, thereby excluding the hypothesis that effects secondary to postsynaptic neuron activation are critical for astrocyte responsiveness.

Oscillations in astrocytes occurred with a delay with respect to the [Ca²⁺]_i increase observed in neurons. One possible explanation for this delay may be the time required for the diffusion of glutamate away from the synaptic cleft. With respect to the receptors at the neuronal postsynaptic membrane, those at the astrocyte membrane are relatively far away from the site of neurotransmitter release and are probably activated only when the concentration of glutamate in the perisynaptic space reaches a threshold level. Theoretical models, however, suggest that glutamate can diffuse from the site of release for several micrometers and reaches concentrations of >10 μm in the perisynaptic space within 0.5–5 msec (Clements, 1996). This explanation can, therefore, hardly account for the delay observed in the astrocyte response. The delay may be attributable to the time necessary for the [Ca²⁺]_i change to spread from the site of activation, presumably in proximity of the synaptic cleft, to the cell body and to the occurrence of a spatial–temporal integration of the Ca²⁺ signal. Indeed, astrocyte membranes have been identified within 1 μm from the spine synapses of the cerebellum (Chaudhry et al., 1995), and a similar anatomical pattern has been described in CA1 hippocampal region (Rothstein et al., 1994).

Long-term changes in [Ca²⁺]_ioscillation frequency mediated by neuronal activity

A second intriguing property of astrocytes is the plasticity of their response to neuronal stimulation. When successive stimulations were applied to neuronal afferents, astrocytes adjacent to stimulated neurons displayed an increased oscillation frequency rather similar to that observed after repetitive t-ACPD stimulation. Apparently, the astrocyte response can be potentiated according to previous episodes of activity occurring at synapses in close proximity. This represents, therefore, an activity-dependent change previously considered an exclusive feature of neuronal cells. It is reminiscent of the activity-dependent increase in synaptic efficacy of excitatory synaptic transmission, the so-called long-term potentiation (LTP) (Bliss and Lomo, 1973). LTP is believed to represent, at the cellular level, certain aspects of learning and memory phenomena (Bliss and Collingridge, 1993). Although many questions still remain unresolved, substantial understanding exists on the cellular and molecular basis of LTP (Bliss and Collingridge, 1993; Kullmann and Singelbaum, 1995). In contrast, very little is known about the long-term change in the astrocyte response. Thus it may seem premature to compare the two phenomena. Nevertheless, besides the fact that glutamate is the principal mediator of both events, the potentiation of the astrocyte response appears to share with neurons at least two of the formal properties that characterize LTP, such as saturation and persistency. As to the first, we observed the failure of inducing potentiation in cells oscillating at high frequency at the very beginning. As to the second property, i.e., persistency, we demonstrated that the potentiation of the astrocyte response persisted for at least 3 hr. In contrast to these common aspects, the kinetics of the induction mechanism are very different: LTP occurs within milliseconds, whereas the potentiation in astrocytes appears much slower. This latter observation suggests that the plasticity in the astrocyte response likely is involved in the slow modulation of the neuron–astrocyte network actions. The slowness of a signaling system, based on [Ca²⁺]_i oscillations and waves with respect to the rapidity of synaptic transmission, has been suggested to denote the modulatory role of astrocytes in brain function (Dani et al., 1992; Smith, 1994).

On the functional role of [Ca²⁺]_ioscillations in astrocytes

The final and key question is the functional role of [Ca²⁺]_i oscillations in astrocytes and the possible significance of their potentiation in response to repetitive episodes of neuronal activity. Our observation that [Ca²⁺]_i oscillations in astrocytes are accompanied by [Ca²⁺]_i elevations in adjacent neurons, together with the finding that this response could be blocked, at least in a number of neurons, by iGluR antagonists, suggests that astrocytes in situ can release glutamate or a glutamate analog efficiently. Indeed, in a number of neurons the [Ca²⁺]_i increase induced byt-ACPD was blocked by NBQX and d-AP5, indicating that this effect was totally dependent on the activation of AMPA and NMDA receptors. The insensitivity to TeNT of t-ACPD-induced [Ca²⁺]_i increase in these neurons indicates that the neurotransmitter is released by neither presynaptic terminals nor other neurons and thus points to the astrocytes as the cells responsible for this release. In other neurons, the activation of AMPARs and NMDARs by glutamate as well as the direct simulation byt-ACPD of mGluRs expressed at the neuronal membrane are responsible for the [Ca²⁺]_i elevation induced by t-ACPD. Active synaptic terminals may, therefore, not only rapidly excite postsynaptic target neurons but, by triggering [Ca²⁺]_i oscillations in astrocytes and governing their frequency, modulate the excitability of other neurons that are not activated synaptically but lay within the domain underlined by the actions of responsive astrocytes. Signaling transduction systems that are based on [Ca²⁺]_i elevations as well as typical manifestations of neuronal plasticity such as LTP, which critically depend on both the degree of activity of presynaptic afferents and the [Ca²⁺]_i rise in the postsynaptic neurons (Bliss and Collingridge, 1993), thus may be deeply modified by the activity of astrocytes. Interestingly, in GFAP mutant mice the induction of long-term depression of synaptic transmission in the cerebellum is impaired (Shibuki et al., 1996).

Stimuli that increase the [Ca²⁺]_i in cultured astrocytes have been demonstrated to cause a Ca²⁺-dependent release of glutamate from astrocytes that can affect the [Ca²⁺]_i in adjacent neurons (Parpura et al., 1994a). Our results demonstrated that glutamate-mediated [Ca²⁺]_i elevations were repetitive in most of the neurons (80%), thus suggesting that they may derive from a pulsating release of the neurotransmitter by astrocytes. The hypothesis could be proposed that each of the [Ca²⁺]_i transients in oscillating astrocytes in situ could trigger an episode of release. In such a case the increase in the frequency of [Ca²⁺]_i oscillations in astrocytes after repetitive episodes of neuronal stimulation ultimately might result in a higher glutamate release and thus in a higher or more extensive influence of astrocytes on neuronal excitability.

Interestingly, the increase in oscillation frequency was higher after repetitive episodes of neuronal stimulation than after repetitive stimulation with t-ACPD. The possibility that the physiological stimulus is more effective than the confined activation of mGluRs is supported by the observation that a higher potentiation in the response of the astrocytes on the second t-ACPD application was observed in astrocytes displaying [Ca²⁺]_i oscillations after neuronal stimulation applied between the two t-ACPD applications. Apparently, glutamate is a more powerful agent than t-ACPD in inducing the potentiation of the astrocyte response. The activation of other receptors besides the metabotropic GluR that probably follows the synaptic release of glutamate or a factor co-released with glutamate by synaptic terminals or derived from postsynaptic neurons, such as nitric oxide (Schuman and Madison, 1993), may account for this difference. The observation that in cultured astrocytes the application of nitric oxide synthase inhibitors inhibited the plasticity of the astrocyte response (Pasti et al., 1995) is in favor of this latter hypothesis.

The critical dependence of the [Ca²⁺]_ioscillation frequency in astrocytes on the pattern of neuronal activity, their long-lasting frequency change after repetitive stimulation, and the [Ca²⁺]_i increases in neurons that follow astrocyte activation suggest the existence of a glutamate-mediated bidirectional communication between neurons and astrocytes that may uncover unexpected roles of astrocyte [Ca²⁺]_i oscillations in synaptic transmission.