Mitochondria Exert a Negative Feedback on the Propagation of Intracellular Ca²⁺ Waves in Rat Cortical Astrocytes

All chemicals were obtained from Sigma Chemical Co. unless stated otherwise. All tissue plastics and tissue culture media were obtained from GIBCO BRL. All fluorescence indicators were from Molecular Probes Inc.

The procedure is essentially as described by Peuchen et al. (1996b) with slight modifications. A sucrose-based separation step was added to the procedure to aid in the removal of cell debris and dense myelin, thereby promoting cell attachment and growth. Earle's balanced salt solution (EBSS) was used throughout the isolation protocol for incubations and washes.

Cerebra were taken from 5- to 8-wk-old Sprague-Dawley rats. The cortex was separated from the other brain structures, the meninges and blood vessels were removed. The tissue, kept at 4°C, was chopped into small pieces, resuspended in EBSS, and forced through a 280-μm metal mesh (tissue pan; Sigma Chemicals Co). The filtered suspension was washed by centrifugation at 400 g for 4 min at room temperature. The pellet was resuspended in a Ca²⁺, Mg²⁺–free EBSS solution containing 50,000 U/ml trypsin (EC 3.4.21.4, isolated from porcine pancreas), 1.033 U/ml collagenase (EC 3.4.24.3, type XI, isolated from Clostridium histolyticum), 336 U/ml deoxyribonucleate 5′-oligonucleotidohydrolase (EC 3.1.21.1, type IV, isolated from bovine pancreas). The enzymatic digestion was stopped after 15 min at 36°C by the addition of 10% heat-inactivated FBS. The suspension was gravity-filtered through a 140-μm metal mesh and centrifuged at 400 g for 6 min. The pellet, resuspended in EBSS, was layered over a 0.4-M sucrose solution (molecular biology grade; BDH Chemicals Ltd.) and centrifuged at 400 g for 10 min. The resulting pellet was washed twice in EBSS, followed by one wash in d-valine–based minimum essential medium (MEM) with Earle's salts. The pellet was resuspended in d-valine MEM, which inhibits the growth of fibroblasts and endothelial cells, supplemented with 5% FBS, 2 mM glutamine, and 1 mM malate, and transferred to tissue culture flasks precoated with 0.01% poly- d-lysine. The flasks were placed in an incubator (95% air, 5% CO₂, at 36°C) for 20–40 h, after which the medium was refreshed. After 6 d in culture, the supplemented d-valine MEM was replaced with MEM supplemented with l-valine. The cells reached confluency at 12–14 d in vitro and were harvested and reseeded onto 24-mm-diam glass coverslips (BDH) precoated with 0.01% poly-d-lysine for fluorescence measurements.

Three or more days before experiment, astrocytes were plated onto glass coverslips with an initial plating density of ∼7,000 cells/cm². Cells were washed with a physiological saline containing 156 mM NaCl, 3 mM KCl, 2 mM MgSO₄, 1.25 mM KH₂PO₄, 2 mM CaCl₂, 10 mM glucose, and 7.5 mM Hepes at pH 7.35.

For the qualitative measurement of [Ca²⁺]_m and [Ca²⁺]_n the cells were loaded with 4.5 μM rhod-2/AM in the presence of 0.005% Pluronics 127 for 30 min at room temperature and washed thoroughly. Rhod-2 fluorescence was excited at 560 nm and emitted fluorescence was collected through a 590-nm-long pass barrier filter. High resolution imaging enabled identification of individual mitochondria for analysis.

To demonstrate that rhod-2 could be used to monitor changes in nuclear Ca²⁺, the cells were loaded with 4.5 μM rhod-2/AM and 4.5 μM fura-2/AM in the presence of 0.005% Pluronics 127 for 30 min at room temperature. The dyes were excited sequentially at 490 (for rhod-2), 340, and 380 nm (for fura-2) using specific filters housed in a computer-controlled filter wheel (Cairn Research Ltd.) and fluorescence at wavelengths longer than 520 nm was measured.

For analysis of [Ca²⁺]_cyt waves in single astrocytes, the cells were loaded with 4.5 μM fluo-3/AM in the presence of 0.005% Pluronics 127 for 30 min at room temperature. Fluo-3 was excited at 490 nm and emitted fluorescence ≥ 530 nm was collected.

For analysis of changes in mitochondrial membrane potential (Δψ_m), cells were loaded with 3 μM of tetramethylrhodamine ethyl ester (TMRE) for 15 min at room temperature. TMRE fluorescence was excited at 546 nm and emitted fluorescence was collected through a 590-nm-long pass barrier filter.

Fluorescence measurements were obtained via an epifluorescence inverted microscope (Nikon Diaphot) equipped with an oil immersion 63× fluorite objective (NA 1.3). Fluorescence was excited using a 75 W xenon arc lamp and emitted light projected onto the face of a slow scan, frame transfer 800 × 600 pixel, 12 bit cooled CCD camera (Digital Pixel Ltd.), acquisition rate ∼2 frames per second full frame, binning pixels 3 × 3. To resolve the rising phases of both nuclear and [Ca²⁺]_m signals and to monitor wave propagation better, fast imaging was carried out using a fast readout, cooled interline transfer CCD camera (model 4880; Hamamatsu Phototonics) allowing image acquisition rates up to 50 images per second with pixels binning at 4 × 4. All imaging data were acquired and analyzed using software from Kinetic Imaging. Because rhod-2, fluo-3, and TMRE are single wavelength fluorescent indicators, it was not possible to apply the ratiometric method for quantitative determination of [Ca²⁺]_m, [Ca²⁺]_cyt, and Δψ_m, respectively. Therefore, the information derived from the measurements of rhod-2, fluo-3, and TMRE fluorescence was normalized as a function of the first image. Nuclear Ca²⁺ ([Ca²⁺]_n) monitored by fura-2 was expressed as the ratio of fura-2 fluorescence after excitation at 340 and 380 nm. All imaging experiments were carried out at room temperature. Greater care was taken for the wave propagation experiments with an average temperature of 21 ± 1°C to avoid any artifact due to the temperature-dependence of Ca²⁺ wave kinetics (Lechleiter and Clapham, 1992). Excitatory light was kept to a minimum with neutral density filters and a computer-controlled shutter to minimize photobleaching and photodynamic injury to cells.

Confocal microscopy was used for colocalization studies and for high spatial resolution of mitochondrial depolarization waves. In brief, astrocytes were dual-loaded with rhod-2/AM as described above and with 20 nM MitoFluor Green/AM for 30 min at room temperature. Cells were imaged using a confocal laser scanning microscope (model 510; Carl Zeiss, Inc.). Cells were illuminated using the 488-nm emission line of an argon laser and fluorescence was collected simultaneously between 505 and 530 nm for MitoFluor Green and at λ ≥ 585 nm for rhod-2. To spatially resolve waves of mitochondrial depolarization, astrocytes were loaded with TMRE as described above and were illuminated with the 488-nm emission line. Emitted fluorescence was collected at λ ≥ 585 nm.

After testing for normality and variance homogeneity, the data were subjected to the unpaired t test. When the variance was heterogeneous, the Mann-Whitney U test was used. The 0.05 level was selected as the point of minimal statistical significance. All the curves shown in figures are representative of at least three measurements performed on four separate cell preparations.

Adult rat cortical astrocytes were loaded with rhod-2/AM, a Ca²⁺-sensitive fluorescent indicator (Minta et al., 1989) that accumulates in mitochondria (Hajnoczky et al., 1995; Jou et al., 1996; Babcock et al., 1997) in response to Δψ_m. Fig. 1 a illustrates a typical image of a rhod-2–loaded adult rat cortical astrocyte showing the localization of rhod-2 within rod-shaped, brightly fluorescent structures of variable organization, reminiscent of mitochondria. However, a diffuse small cytosolic signal and an accumulation of the dye in the nucleus (with rhod-2 also labeling nucleoli) were usually noticed (Trollinger et al., 1997). To confirm that the main rhod-2 fluorescence signal did originate from mitochondria, astrocytes were dual-loaded with rhod-2/AM and MitoFluor Green/AM, an indicator that labels mitochondria independently of mitochondrial function (Poot et al., 1996). The small overlap of the excitation spectra of rhod-2 and MitoFluor Green allows the simultaneous use of both the dyes (Fig. 1, a and b). Colocalization of MitoFluor Green (Fig. 1 b) and rhod-2 within the cell (Fig. 1 a) was confirmed by merging the two images (Fig. 1 c) and by the analysis of colocalization shown in Fig. 1 d. It should be stated that the relative compartmental loading varied substantially between batches of rhod-2 and between preparations, but the difference was quantitative, not qualitative (i.e., some batches of dye gave much clearer localization into mitochondria, others gave a brighter relative cytosolic [and nuclear] signal, but the partitioning was still between cytosol/nucleus and mitochondria).

Physiological activation of the P_2U purinoceptor by ATP in adult rat cortical astrocytes has been shown previously to raise [Ca²⁺]_cyt via an IP3-mediated pathway (Peuchen et al., 1996a). We took advantage of the concomitant localization of rhod-2 within the nucleus (a mitochondrion-free volume of the cell) to confirm and monitor changes in [Ca²⁺]_cyt in response to ATP challenge, as [Ca²⁺]_n follows the changes in [Ca²⁺]_cyt by diffusion of Ca²⁺ from one compartment into the other (al Mohanna et al., 1994; Brini et al., 1994). The temporal relationship between [Ca²⁺]_n and [Ca²⁺]_m after a brief pulse of 100 μM ATP is shown in the time sequence of Fig. 2 a. The sequential quantitative changes in [Ca²⁺]_n and [Ca²⁺]_m, measured along the longitudinal axis of the cell as indicated by the line in the first image of Fig. 2 a, are shown in Fig. 2, b and c. Before agonist application, the rhod-2 fluorescence was relatively dim and the mitochondrial signal was not obviously brighter than much of the cytosol (Fig. 2 a). Immediately after the mobilization of intracellular Ca²⁺ stores by ATP, an overall increase over both the nucleus (Fig. 2, asterisk) and the cytosol was observed, followed by a delayed increase in signal over mitochondria (selected close to nucleus) that remained elevated long after [Ca²⁺]_n recovery. The rise in [Ca²⁺]_m presented a wavelike pattern in that mitochondria at one end of the cell tended to respond before those at the other (Fig. 2 a, indicated by black arrows in a and c).

[Ca²⁺]_n rose rapidly (time to reach half the peak amplitude, t_1/2 = 1.95 ± 1.1 s, n = 26), reaching a normalized fluorescence peak of 1.75 ± 0.36 (n = 57), and returned to its resting level within a minute (Fig. 2 a and Fig. 3, a and b, solid circle). A quantitative analysis showed that [Ca²⁺]_n returned to resting values with a decay time constant of 4.84 ± 1.75 s (n = 41). In contrast, [Ca²⁺]_m rose slowly (t_1/2 = 3.17 ± 1.3 s, n = 26) and reached a fluorescence [Ca²⁺]_m peak corresponding to a 2 ± 0.57-fold increase (n = 63). It remained at a high plateau within the experimental time scale (55 s, Fig. 2). Long-term recording of the ATP-induced changes in [Ca²⁺]_m indicated that full recovery to the resting [Ca²⁺]_m took 28.6 ± 10.5 min (n = 24, Fig. 3 b).

Plotting changes in [Ca²⁺]_m as a function of the changes in [Ca²⁺]_n revealed four phases of intracellular Ca²⁺ handling triggered by the IP3-induced mobilization of intracellular Ca²⁺ stores (Fig. 3 c). Immediately after ATP application (phase 1), [Ca²⁺]_n rose very rapidly without any important change in [Ca²⁺]_m. [Ca²⁺]_m started increasing together with [Ca²⁺]_n (phase 2). In the following 70 s (phase 3), [Ca²⁺]_m fell slowly despite the rapid return of [Ca²⁺]_n to its resting value. Finally, [Ca²⁺]_m progressively and slowly decreased (phase 4) to its resting levels.

We cannot exclude that the measured rise in the extranuclear fluorescence probably reflects a combination of cytosolic and mitochondrial Ca²⁺ signals. However, neither the peak nor the recovery of the [Ca²⁺]_m signal appeared to be contaminated by the [Ca²⁺]_cyt signal that peaked and returned to basal level far more quickly than [Ca²⁺]_m (see below for time courses of [Ca²⁺]_cyt and [Ca²⁺]_n). In addition, the flat morphology of astrocytes in culture (Sheppard et al., 1997; Simpson et al., 1998) allowed the resolution and selection of individual mitochondria for image analysis.

Mitochondrial calcium uptake depends on Δψ_m and dissipation of the potential provides the easiest way to examine the consequences of [Ca²⁺]_m accumulation for cell Ca²⁺ signaling. To this end, we chose to use the mitochondrial protonophore p-trifluoromethoxy-phenylhydrazone (FCCP), a mitochondrial uncoupler that dissipates the proton gradient across the inner mitochondrial membrane, hence abolishing the Δψ_m. Under such conditions, activation of the mitochondrial ATPase in reverse mode may lead to the rapid consumption of ATP (Leyssens et al., 1996). Therefore, FCCP (1 μM) was always applied in combination with 2.5 μg/ml oligomycin, to inhibit mitochondrial consumption of cellular ATP.

The time series shown in Fig. 4 a, and more particularly the corresponding surface plot (Fig. 4 c) and line image (Fig. 4 d), illustrate the changes in [Ca²⁺]_cyt and [Ca²⁺]_n upon FCCP application, followed by 100 μM ATP. FCCP initially induced a transient rise in [Ca²⁺]_cyt localized to the perinuclear (Fig. 4, a and b, p; Fig. 4 c and d, arrows) and the cytosolic (Fig. 4, a and b, c; Fig. 4 c and d, arrows) regions, most likely reflecting Ca²⁺ release from mitochondria that was only followed later by a rise in [Ca²⁺]_n (Fig. 4, a and b, n; Fig. 4, c and d, asterisk). In the experiment illustrated in Fig. 4 b, the FCCP-induced change in rhod-2 fluorescence was ∼1.2-fold. The mean increase in rhod-2 signal in response to FCCP application was 1.76 ± 0.43 (n = 29) fold (see also below).

Subsequent application of ATP elicited a diffuse, transient rise in nuclear and cytosolic rhod-2 fluorescence (Fig. 4 a), without any apparent mitochondrial rhod-2 signal, demonstrating the effective inhibition of Ca²⁺ uptake in deenergized mitochondria. The relative change in the nuclear Ca²⁺ signal appeared far bigger than that in the cytosol (Fig. 4). This apparent nuclear amplification is likely to occur because of an underestimation of [Ca²⁺]_cyt changes resulting from the creation of a pool of indicator sequestered in mitochondria that was insensitive to changes in [Ca²⁺]_cyt after the dissipation of Δψ_m by FCCP (al Mohanna et al., 1994).

Collapse of Δψ_m by FCCP did not significantly alter the peak amplitude of the ATP-induced [Ca²⁺]_n transient (Fig. 4 b, control cells, peak = 1.75 ± 0.36, n = 57; FCCP-treated cells, peak = 1.63 ± 0.33, n = 38; P = 0.11, t test) or the time required to reach half the peak amplitude of [Ca²⁺]_n (control cells, t_1/2 = 1.95 ± 1.1 s, n = 26; FCCP-treated cells, t_1/2 = 1.93 ± 0.95 s, n = 38; P = 0.93, t test). However, the decay phase of the [Ca²⁺]_n was significantly longer in the presence of FCCP (control cells, t = 4.84 ± 1.75 s, n = 41; FCCP-treated cells, t = 8.92 ± 2.89 s, n = 44; P < 0.0001, Mann-Whitney U test). This result strongly suggests that mitochondrial Ca²⁺ buffering plays a substantial role in hastening the restoration of [Ca²⁺]_n and, hence, [Ca²⁺]_cyt to basal levels. The [Ca²⁺]_cyt transient had a very similar time course to that of the [Ca²⁺]_n (Fig. 4 b), strengthening the assumption that changes in [Ca²⁺]_n are a true index of changes in [Ca²⁺]_cyt.

Dissipation of Δψ_m by FCCP reverses the activity of the Ca²⁺ uniporter, which becomes a Ca²⁺ efflux pathway (Kröner, 1992). We applied 1 μM FCCP to resting astrocytes dual-loaded with rhod-2 and fura-2, in the presence of 2.5 μg/ml oligomycin. Data are illustrated in Fig. 5. Short application of FCCP induced a steep rise in [Ca²⁺]_n monitored by rhod-2 (solid circle) that returned to basal level within a minute (mean peak amplitude = 1.76 ± 0.43, n = 29).

We were concerned that the change in the rhod-2 signal might reflect some form of redistribution of dye and dequench (i.e., as seen with TMRE, although this is not expected on theoretical grounds; see below) rather than a true reflection of changing [Ca²⁺]_n. We have recorded rhod-2 fluorescence simultaneously with a second widely used Ca²⁺-sensitive fluorescent indicator, fura-2. FCCP caused a rise in fura-2 ratio (open circle, mean peak amplitude = 2 ± 0.6, n = 21) sharing kinetic characteristics that could be superimposed on rhod-2. In principle, only the nonfluorescent AM ester form of rhod-2 exhibits a net positive charge, enabling its sequestration into mitochondria in response to Δψ_m. Once the AM ester is cleaved within mitochondria, the resulting rhod-2 is hydrophilic and negatively charged and is unlikely to cross (mitochondrial) membranes or redistribute upon mitochondrial depolarization. As FCCP interferes with nonmitochondrial stores (Jensen and Rehder, 1991) in some preparations, we repeated these experiments using antimycin A1 (an inhibitor of mitochondrial respiratory complex III) in the presence of oligomycin. Qualitatively, similar results were observed under these conditions (data not shown). Therefore, these data suggest that the rise in rhod-2 signal upon mitochondrial depolarization is an accurate reflection of changes in [Ca²⁺]_cyt and that Ca²⁺ is released from mitochondria in these resting astrocytes.

To investigate the spatio-temporal organization of the [Ca²⁺]_cyt changes in astrocytes, we examined Ca²⁺ waves in fluo-3–loaded astrocytes at higher time resolution (∼7–14 Hz). Single-cell stimulation was performed either chemically by pressure application from a micropipette filled with ATP (20 μM) or mechanically with a micromanipulator-driven glass micropipette filled with the recording saline. We chose an ATP concentration of 20 μM to temporally resolve the Ca²⁺ wave as stimulation with 100 μM induced a rapid wave that could not be resolved (data not shown).

After either chemical or mechanical stimulation, a wave of [Ca²⁺]_cyt was routinely observed throughout the stimulated cell, usually spreading in directions away from the point of stimulation. Fig. 6 a shows some sample images from a sequence to illustrate the propagation of an intracellular [Ca²⁺]_cyt wave after ATP application. The time course of propagation was defined by selecting a line along the wave path on the image series. The intensity profile for the line for each frame was plotted as a two-dimensional line image (the post hoc equivalent of a line scan on a confocal system) as shown in Fig. 6 c. The wave propagation is seen as a gradual increase of the fluorescent signal across the cell from one side to the opposite side as a function of time. The wave velocity was readily measured simply as the slope of distance with time across the cytosol (Fig. 6 c). The mean value for the velocity of the [Ca²⁺]_cyt wave was 21.6 ± 11.5 μm/s (Table I, n = 133) and 24.4 ± 10.7 μm/s (Table I, n = 129) under chemical and mechanical stimulation, respectively. These velocities were not significantly different, suggesting that once initiated, the waves were propagated by the same mechanism. The mean rate given by all these data was 22.9 ± 11.2 μm/s (Table I, n = 262). The time course of the amplitude changes in [Ca²⁺]_cyt signal for six regions of the cell during the propagation of the wave as a function of time is shown in Fig. 6 b. The peak amplitude of the wave was sustained throughout the regions of propagation although the rate of rise tended to decrease with propagation of the wave. Similar results were found for mechanically induced Ca²⁺ waves (data not shown). The [Ca²⁺]_cyt at this sustained level (∼1.8-fold increase from resting values) remained within the measurable range. Indeed, saturation of the fluo-3 signal by application of 20 μM ionomycin showed a fluorescence enhancement of 5.7 ± 3.3-fold (n = 77) in comparison to resting [Ca²⁺]_cyt. Determination of the diffusion rate of Ca²⁺ across the nucleus of ATP-stimulated cells showed a significantly higher rate than the Ca²⁺ wave velocity calculated over the cytosol (Table I, 34.6 ± 26.7 μm/s, n = 27, Mann-Whitney U test, P < 0.05) corresponding to a 60% increase.

We asked about the consequences of mitochondrial calcium uptake, for mitochondrial function and, as described below, for calcium signaling. Upon chemical and mechanical stimulation, waves of [Ca²⁺]_m propagating across single astrocytes could be observed (Fig. 2, ATP challenge). However, the features of a [Ca²⁺]_m wave front were hard to characterize given the spatially irregular punctate rhod-2 signal. In addition, the rhod-2 signal arising from the cytosol obscured the initial rising phase of the [Ca²⁺]_m wave characteristics. To overcome this problem, we followed changes in Δψ_m, as influx of Ca²⁺ into mitochondria induces a small transient mitochondrial depolarization (Duchen, 1992; Loew et al., 1994; Duchen et al., 1998; Peuchen et al., 1996b). Indeed, monitoring changes in Δψ_m tends to amplify the response to mitochondrial calcium uptake and provides a sensitive index of a small calcium flux during mitochondrial Ca²⁺ uptake (Duchen et al., 1998).

To follow changes in Δψ_m, astrocytes were loaded with the potentiometric dye TMRE that selectively accumulates in mitochondria. The concentration of TMRE into polarized mitochondria causes autoquenching of TMRE fluorescence that is relieved by mitochondrial depolarization. Therefore, TMRE dequench is associated with a rise in fluorescence. Induction of a Ca²⁺ wave with a brief deformation of the plasma membrane initiated a wave of mitochondrial depolarization spreading from the stimulation point across the cell seen clearly in the confocal image series shown in Fig. 7 a. A line image (Fig. 7 c) constructed from a similar image series but obtained using the fast readout CCD camera (Fig. 7 b) clearly showed the initiation sites of the propagating wave as mitochondria (the bright lines on the line image), whereas TMRE diffused slowly into the nucleus (asterisk). Determination of the wave velocity gave a mean value of 9.7 ± 4.2 μm/s (n = 89). Fig. 7 b illustrates the changes in Δψ_m recorded in six regions of the cell as a function of time. The peak amplitude of the transient mitochondrial depolarization was sustained over long distances throughout the regions of wave propagation (Fig. 7 b). Indeed, roughly the same peak amplitude of Δψ_m signals was seen at the cytosolic stimulation point 1 and the more distal point 6, suggesting again the regenerative nature of the propagating wave. In addition, the rate of rise tended to decrease with propagation of the wave, strikingly mirroring the gradual decrease of the rate of rise of the [Ca²⁺]_cyt wavefont. Thus, the propagation of the Ca²⁺ wave causes a wave of mitochondrial depolarization in its wake.

To determine the impact of mitochondrial Ca²⁺ uptake on the spatio-temporal characteristics of calcium signaling, fluo-3–loaded astrocytes were treated either with 2.5 μg/ml antimycin A1 or with 10 μM rotenone (an inhibitor of mitochondrial respiratory complex I), both in association with 2.5 μg/ml oligomycin. This protocol induced a complete collapse of Δψ_m (assessed with TMRE, data not shown), preventing any subsequent Ca²⁺ uptake into mitochondria. Whatever the mitochondrial inhibitor used, mechanical or chemical stimulation elicited a typical [Ca²⁺]_cyt wave spreading away from the stimulation point, very similar to control cells in terms of pattern and regenerative nature of the propagation. Fig. 8 shows a typical example of Ca²⁺ waves spreading across an astrocyte upon ATP challenge either in control conditions (a) or after treatment with antimycin A1/oligomycin (b). The differentiated line image (corresponding to a time series of iteratively subtracted images, only showing changes occurring between consecutive frames) given in the lower part of each frame showed the wavefront propagating through the cell. It is clear that the Ca²⁺ wave traveled faster in the antimycin-treated cells (Fig. 8 b).

Determination of the wave velocity in antimycin-treated cells showed a mean value of 35.6 ± 23 μm/s (n = 121) in response to ATP application and 31.1 ± 10.5 μm/s (n = 66) to mechanical stimulation (Table I). These values were not significantly different and the pooled data gave an average Ca²⁺ wave rate in antimycin-treated cells of 34 ± 19.6 μm/s (n = 187). Rotenone-treated cells exhibited a Ca²⁺ wave rate of 40.4 ± 28.6 μm/s (n = 50) to ATP challenge and of 30.6 ± 12.4 μm/s (n = 35) to mechanical stimulation (Table I). The averaged Ca²⁺ wave rate was 36.4 ± 23.7 μm/s (n = 85), remarkably similar to the diffusion rate of Ca²⁺ through the mitochondrion-free nucleus (see above). This corresponds to a significant 48% and 58% increase (P < 0.0001, Mann-Whitney U test) of the [Ca²⁺]_cyt wave velocity in antimycin- and rotenone-treated astrocytes, respectively. Statistical analysis of the Ca²⁺ wave velocities from antimycin- and rotenone-treated cells revealed no significant difference. These results clearly demonstrate that mitochondrial calcium uptake exerts a negative feedback action upon the kinetics of Ca²⁺ wave propagation in astrocytes.

In the present study, we have addressed several related questions: (a) Do mitochondria take up Ca²⁺ during physiological Ca²⁺ signaling in adult rat cortical astrocytes? (b) If so, what are the temporal characteristics of the mitochondrial response? (c) What are the consequences for mitochondrial potential and (d) what are the consequences for the spatio-temporal characteristics of the Ca²⁺ signal? Our data show that mitochondria in these cells take up Ca²⁺ during physiological Ca²⁺ signaling, that they tend to retain that Ca²⁺ for prolonged periods of time, and that the uptake not only shapes [Ca²⁺]_cyt signals but also exerts a negative feedback on the rate of propagation of Ca²⁺ waves, by acting as high capacity fixed Ca²⁺ buffers.

Little information is available about the functional consequences of [Ca²⁺]_m accumulation with respect to [Ca²⁺]_cyt handling in glial cells. Stimulation of IP3-mediated ER Ca²⁺ release by ATP in astrocytes routinely induced a simultaneous rise in [Ca²⁺]_n and in [Ca²⁺]_m. However, these signals exhibited quite distinct temporal features: the [Ca²⁺]_n transient recovered within 1 min. In contrast, the mitochondrial response rose 1.6-fold more slowly than the nuclear signal, and full recovery took several minutes. A longer time of recovery to basal level for [Ca²⁺]_m than for [Ca²⁺]_cyt also has been described in the glial cell line RBA-1 (Jou et al., 1996), in oligodendrocytes (Simpson and Russell, 1998), in hepatocytes (Hajnoczky et al., 1995), in myocytes (Trollinger et al., 1997; Zhou et al., 1998), and in adrenal chromaffin cells (Herrington et al., 1996), although the specific rates varied widely between these different cell types.

These observations differ strikingly from results obtained with aequorin targeted to mitochondria in HeLa cells (Rizzuto et al., 1992). This discrepancy could result from the use of different cell types, different methodologies, different temperatures, the relative inaccuracy of aequorin at [Ca²⁺] values below 300–400 nM (K_d for Ca²⁺ = 1 μM), or alteration of the apparent kinetics of the [Ca²⁺]_m signal by rhod-2 given its relatively high affinity to Ca²⁺ compared with the low affinity aequorin. The mitochondrial model elaborated by Magnus and Keizer (1998) also predicts a delay between increases in [Ca²⁺]_cyt and [Ca²⁺]_m. This delay may result from the positive cooperativity of uniporter activation by extramitochondrial Ca²⁺ (Hill coefficient ∼2) (Gunter and Pfeiffer, 1990). Mitochondria start accumulating cytosolic Ca²⁺ from [Ca²⁺]_cyt ∼500 nM but do not saturate readily (Gunter, 1994). Rates of [Ca²⁺]_m increase after intracellular Ca²⁺ mobilization have been demonstrated to be more than one order of magnitude faster than that predicted on the basis of the amplitude of the mean rise in [Ca²⁺]_cyt (Rizzuto et al., 1993). This has been explained by proposing that mitochondria closely apposed to Ca²⁺ release sites would be exposed to local concentrations of Ca²⁺ much higher than those measured in the bulk of the cytoplasm (Rizzuto et al., 1993, 1998). Such a close localization of mitochondria to IP3 receptors recently has been observed in rat astrocytes in culture (Simpson et al., 1998), and is also suggested by the appearance of transient mitochondrial depolarizations due to focal calcium release from the sarcoplasmic reticulum in cardiac myocytes (Duchen et al., 1998).

Depolarization of the mitochondrial potential with FCCP completely prevented [Ca²⁺]_m loading after a challenge with ATP. In addition, the rate of decay of the [Ca²⁺]_n signal was significantly slowed, suggesting that mitochondrial Ca²⁺ uptake plays a significant role in the clearance of [Ca²⁺]_cyt loads in these cells. The effects of the mitochondrial uncoupler on Ca²⁺ clearance were almost certainly not attributable to ATP depletion, as the mitochondrial ATPase was inhibited throughout by oligomycin to minimize mitochondrial consumption of cellular ATP. In contrast to our results, FCCP exposure did not influence the kinetic parameters of the depolarization-triggered [Ca²⁺]_cyt transient in oligodendrocytes (Kirischuk et al., 1995b), suggesting that [Ca²⁺]_m accumulation does not play an important role in shaping Ca²⁺ signals in oligodendroglia. It seems likely that differences in such contributions of mitochondria to calcium signaling may vary simply as a function of the relative mitochondrial density or intracellular location in different cell types.

The [Ca²⁺]_n signal in the astrocytes after ATP stimulation recovered with a monoexponential time course. In a number of cell types, [Ca²⁺]_cyt decay characteristically follows a biphasic sequence in which a fast initial decay is followed by a much slower secondary decay that appears to be maintained by the reequilibration of [Ca²⁺]_m by the mitochondrial Na⁺/Ca²⁺ exchange (rat gonadotropes, Hehl et al., 1996; sensory neurons, Thayer and Miller, 1990). The absence of such a plateau phase in the astrocyte model seems consistent with the very prolonged retention of Ca²⁺ in the mitochondria of these cells. This may reflect a low [Na⁺] content of astrocytes and the absence of Na⁺ influx during this form of stimulation (in contrast to depolarization-induced transients in the excitable sensory neurons, gonadotropes, and chromaffin cells). Indeed, while the accumulated [Ca²⁺]_m is clearly released back into the cytosol, this occurs at such a low rate that it does not cause a significant change in [Ca²⁺]_cyt. Therefore, the [Ca²⁺]_m efflux pathway does not significantly shape the [Ca²⁺]_cyt signal in these cells.

We noted that depolarization of the mitochondrial potential routinely released Ca²⁺ into the cytosol within quiescent cells, suggesting that mitochondria in this preparation were often Ca²⁺-loaded at rest. Given the very prolonged time course of [Ca²⁺]_m reequilibration, perhaps this is not very surprising as only occasional spontaneous activity would keep the mitochondria primed with a significant Ca²⁺ content. A substantial Ca²⁺ pool has been found in mitochondria of oligodendrocytes under resting conditions (Kirischuk et al., 1995a,b), in contrast to the very modest [Ca²⁺]_m content in neurons (Werth and Thayer, 1994), T lymphocytes (Hoth et al., 1997), rat gonadotropes (Hehl et al., 1996), and rat chromaffin cells (Herrington et al., 1996).

In astrocytes, mitochondria seem to contribute to cytosolic Ca²⁺ clearance during the [Ca²⁺]_n decay only but not in the fast rising phase of the [Ca²⁺]_n transient. In addition, dissipation of Δψ_m by FCCP did not appreciably affect the peak amplitude of the ATP-induced [Ca²⁺]_n response. These results are similar to those described in adrenal chromaffin cells (Babcock et al., 1997), T lymphocytes (Hoth et al., 1997), rat gonadotropes (Hehl et al., 1996), and HeLa cells (Rizzuto et al., 1992). Simpson et al. (1998) found that FCCP pretreatment caused variable effects on the norepinephrine-evoked [Ca²⁺]_cyt responses in astrocytes, ranging from complete block of the response to minor effects. In contrast, we consistently found no significant alteration in the amplitude of the Ca²⁺ peak induced by ATP in FCCP (+ oligomycin)-treated astrocytes. This disparity could be partly explained by the absence of oligomycin in most of their experiments, leading to ATP run-down in the course of the norepinephrine stimulation.

Stimulation of astrocytes triggers an increase in [Ca²⁺]_cyt that can propagate as a wave through the cytoplasm of individual cells and through astrocytic networks (Cornell-Bell et al., 1990; Verkhratsky et al., 1998). Thus, we were interested in determining the role of mitochondria upon the spatio-temporal characteristics of these Ca²⁺ waves. Both mechanical and chemical stimuli were used to trigger Ca²⁺ waves. In addition to triggering Ca²⁺ influx, mechanical stimulation recently has been shown to activate phospholipase C in astrocytes (Venance et al., 1997). The propagation velocities of these Ca²⁺ waves were nearly identical regardless of the initiating stimulus (pooled control rate = 22.9 ± 11.2 μm/s, n = 262), suggesting that once initiated, the waves were propagated by the same mechanism. This range of Ca²⁺ wave velocities correlates well with those found in hippocampal astrocytes (Cornell-Bell et al., 1990) and retinal astrocytes (Newman and Zahs, 1997).

The amplitude of the Ca²⁺ signal from the wave initiation site could be maintained or could even increase during spread across the cell, strongly suggesting that the wave propagates as an actively regenerative process and not simple long-range passive diffusion following release from a point source. The propagating [Ca²⁺]_cyt wave was associated with a propagating wave of mitochondrial depolarization that traveled at around 10 μm/s following mechanical stimulation. Thus, during the propagation of an intracellular [Ca²⁺]_cyt wave, mitochondria take up Ca²⁺ as the wave travels across the cell. The discrepancy between the velocity of the [Ca²⁺]_cyt wave and of the wave of mitochondrial depolarization probably reflects fundamental differences in the behavior of the fluorescent dyes fluo-3 and TMRE. Whereas fluo-3 fluorescence rises immediately upon Ca²⁺ binding, the increase in TMRE fluorescence after mitochondrial depolarization results from dye egress from mitochondria, dequench of fluorescence and then diffusion of dye through the cytosol, inevitably underestimating the rate of progression of the wave of mitochondrial depolarization. Furthermore, the change in Δψ_m reflects the rate of Ca²⁺ flux, which tends to be slower at the distal end of the cell (Fig. 8) compared with the origin of the wave front. Nevertheless, the appearance of a wave of mitochondrial depolarization clearly demonstrates that mitochondria take up Ca²⁺ as a [Ca²⁺]_cyt wave travels across the cell.

To determine whether mitochondria play a functional role in the propagation of the Ca²⁺ wave in astrocytes, mitochondrial Ca²⁺ uptake was prevented by dissipation of the mitochondrial potential using antimycin A1 or rotenone (each in association with oligomycin). The Ca²⁺ waves presented the same regenerative pattern of propagation as in the control cells, but the wave velocity increased substantially. Thus, mitochondrial Ca²⁺ uptake appears to slow the rate of propagation of the cytosolic signal. Taken together, these results suggest that functional mitochondria exert a negative feedback on the propagation of intracellular Ca²⁺ waves in adult rat cortical astrocytes.

What could be the underlying mechanism of the negative feedback control exerted by mitochondria upon the propagating Ca²⁺ wave? Current models to account for the propagation of Ca²⁺ signals suggest that IP3 induces a wave by diffusing through the cell, priming IP3 receptors, and releasing Ca²⁺ as puffs from ER release sites (Finkbeiner, 1993; Venance et al., 1997; Wang et al., 1997). The rise in [Ca²⁺]_cyt generates additional IP3 through the Ca²⁺-dependent activation of phospholipase C (Berridge, 1993) providing a positive feedback step, whereas the diffusion of Ca²⁺ to neighboring IP3 receptors activates increasing numbers of IP3-primed IP3 receptors (Yao et al., 1995), as type 1 IP3 receptors display a bell-shaped sensitivity to Ca²⁺ that functions as a coagonist with IP3 to release stored Ca²⁺ (Bezprozvanny et al., 1991; Finch et al., 1991). The peak sensitivity to Ca²⁺ lies at ∼300 nM, whereas at higher concentrations Ca²⁺ becomes inhibitory. However, type 2 IP3 receptors, lack the Ca²⁺-dependent inactivation at high [Ca²⁺]_cyt (Ramos-Franco et al., 1998). Thus, depending on the type of receptors expressed by a cell, mitochondrial control of local [Ca²⁺]_cyt through Ca²⁺ uptake could result in either a positive or a negative control on the Ca²⁺ wave propagation. The importance of mitochondria in the regulation of propagating [Ca²⁺]_cyt waves was first demonstrated in oocytes of the amphibian Xenopus laevis (Jouaville et al., 1995) that express type 1 IP3 receptors (Kume et al., 1993). These authors found that IP3-mediated Ca²⁺ waves were synchronized and increased in both amplitude and velocity by addition of substrates for mitochondrial respiration. The additional substrate hyperpolarizes the Δψ_m, promoting [Ca²⁺]_m uptake and slowing IP3 receptor inactivation by reducing the Ca²⁺ concentration in the immediate vicinity of active IP3 receptors (Jouaville et al., 1995).

In the present study, we found that mitochondria exert a negative control on Ca²⁺ wave propagation. Rat astrocytes in culture express predominantly type 2 IP3 receptors (Sheppard et al., 1997; Boitier, E., S. Brind, and M.R. Duchen, unpublished observations). In addition, mitochondria in astrocytes are located near IP3 receptors, which correspond to the amplification sites for Ca²⁺ waves (Sheppard et al., 1997). One possible explanation would be that mitochondria in close proximity to IP3 receptors participate in a local Ca²⁺ buffering that alters gating kinetics of type 2 IP3 receptor channels. The fall in local [Ca²⁺]_cyt would decrease the affinity of the IP3 receptors to IP3, decreasing further IP3-mediated Ca²⁺ release. Less Ca²⁺ will be available to feedback onto the Ca²⁺ releasing IP3 receptors and to diffuse away to stimulate nearby IP3 receptors. This mechanism would completely account for the slowing of the intracellular Ca²⁺ wave (Wang et al., 1997). In this model, type 2 IP3 receptors will not be inherently self-limiting, because Ca²⁺ passing through an active type 2 channel cannot negatively feedback and turn the channel off (Ramos-Franco et al., 1998). This hypothesis is further supported by the observation that the diffusion rate of Ca²⁺ across the (mitochondrion-free) nucleus was of the same order of magnitude as the Ca²⁺ wave velocities in cells in which mitochondrial uptake had been abolished (∼34 μm/s). Such differences in the wave rates between nucleus and cytosol have been observed in other cell types (Donnadieu and Bourguignon, 1996; Koopman et al., 1997). This probably reflects the paucity of Ca²⁺ buffers in the nucleus, and strengthens the concept that mitochondria are predominant cytosolic Ca²⁺ buffers in astrocytes.

Studying the impact of cytoplasmic Ca²⁺ buffering on the spatial and temporal characteristics of intercellular Ca²⁺ signals in astrocytes, Wang et al. (1997) found that pretreatment with an exogenous Ca²⁺ chelator such as BAPTA dramatically decreased the Ca²⁺ wave velocity within individual astrocytes. According to these authors, the positive feedback role of Ca²⁺ in wave propagation is primarily local, acting at the Ca²⁺ release sites, while Ca²⁺ diffusion constitutes an important factor as a rate-limiting step during the initiation and propagation of Ca²⁺ waves (Lechleiter et al., 1991; Wang and Thompson, 1995; Wang et al., 1997). One major consequence of the presence of an exogenous Ca²⁺ buffer might be to reduce the peak-free Ca²⁺ concentration at a puff site, thereby attenuating the regenerative potential of the wave. These authors suggested that the presence of endogenous cytosolic Ca²⁺ buffers with a relatively low affinity for Ca²⁺ would permit wave propagation but still restrict this mode of intracellular Ca²⁺ signaling to a very localized range (Wang et al., 1997). Ca²⁺ uptake into mitochondria is a low affinity, high capacity process (Gunter, 1994). In addition, a rapid mode of Ca²⁺ uptake recently has been discovered in liver mitochondria allowing [Ca²⁺]_m accumulation for low [Ca²⁺]_cyt (Sparagna et al., 1995). Taken together, the data indicate that Ca²⁺ wave kinetic parameters are controlled tightly by active Ca²⁺ buffering by mitochondria.

Working on the same cell model, Simpson et al. drew completely opposite conclusions (Simpson and Russell, 1996; Simpson et al, 1998), suggesting that mitochondria located close to IP3 receptors promote Ca²⁺ wave progression by preventing Ca²⁺-dependent inactivation of the IP3 receptor at high [Ca²⁺]_cyt. This hypothesis relies on the bell-shaped response of the IP3 receptor to Ca²⁺, characteristic of type 1 IP3 receptor (Bezprozvanny et al., 1991; Finch et al., 1991). However, they showed that astrocytes essentially possess type 2 IP3 receptors (Sheppard et al., 1997). The fact that mitochondria are located near IP3 receptors does not necessarily mean that they play a positive role in Ca²⁺ wave propagation. On the contrary, such a close location could represent a highly plastic mechanism whereby the cell could finely tune intracellular Ca²⁺ wave characteristics, thus modulating this newly discovered form of nonsynaptic long-range signaling in the brain (Cornell-Bell et al., 1990).

Under physiological conditions, mitochondrial status could contribute to the modulation of the intercellular propagation of Ca²⁺ waves within the astrocyte network. Indeed, modulation could simply rely upon supply of mitochondrial respiratory substrates to mitochondria or transient inhibition of mitochondrial respiration, regulating Δψ_m, and modulating mitochondrial Ca²⁺ uptake. In the astrocyte, functional active mitochondria would permit wave propagation but could restrict the extent of spread of this mode of intercellular Ca²⁺ signaling. Modulation of such a mitochondrial contribution would allow a differential activation of specific neurons in the brain and may directly participate in information processing in the CNS.

Brain mitochondria have been found to be impaired in some pathologies such as ischemia/reperfusion injury (Sims, 1995) and Parkinson's disease (Bowling and Beal, 1995). In such conditions, [Ca²⁺]_m uptake is impaired and mitochondria will no longer be able to regulate Ca²⁺ wave spread. One possible consequence might be an increased spread of intercellular Ca²⁺ waves in the astrocytic networks that might eventually lead to Ca²⁺ deregulation in the surrounding nervous system. The astrocytic Ca²⁺ wave may be important for the phenomenon known as Leao's spreading depression (SD) (Do Carmo and Somjen, 1994): a slow wave of depression of neuronal activity (Leao, 1944). It is accompanied generally by an increase in intracellular Ca²⁺ concentration (Martins-Ferreira and Ribeiro, 1995). More recently, a similar, if not identical, phenomenon has been shown to be induced by hypoxia (Somjen et al., 1993). A glial role was suggested in SD since this depression propagates through diverse heterogeneous brain regions with similar velocity (33–100 μm/s). The Ca²⁺ waves in the astrocytes studied therein, more particularly in those where mitochondrial Ca²⁺ uptake was prevented, bear similarities to SD such as their high rates of propagation. The fact that hypoxia seems to mimic such a depression may support our hypothesis of negative mitochondrial feedback on Ca²⁺ wave propagation. In such a condition, mitochondrial function would be disrupted, Δψ_m dissipated and finally mitochondrial Ca²⁺ uptake prevented. This would inexorably lead to further spreading of the glial Ca²⁺ wave, stimulating in passing neuronal networks that normally are not activated by these astrocytes, and may contribute to brain activity disorganization.

In conclusion, we have characterized an important role of mitochondria in glial Ca²⁺ signaling, i.e., a strong negative feedback control on intracellular Ca²⁺ wave propagation via their high Ca²⁺ buffering capacities. Physiological alteration of mitochondrial function would allow the fine tuning of the Ca²⁺ wave spreading through the astrocyte network, hence modulating their information processing functions (Wang et al., 1997). Besides, in pathological conditions, a mitochondrial dysfunction could contribute to the pathogenesis of numerous CNS disorders such as ischemia and epilepsy.

We thank Miss D.L. Patterson for cell culture, and Mr. Jake Jacobson, Dr. Julie Keelan, and Dr. Mart H. Mojet for critical appraisal of the manuscript and helpful discussion.

We thank the Wellcome Trust for financial support. R. Rea was supported by the Wellcome 4 Year Ph.D. Programme in Neuroscience.

[Ca²⁺]_cyt

cytosolic Ca²⁺

[Ca²⁺]_m

mitochondrial Ca²⁺

[Ca²⁺]_n

nuclear Ca²⁺

CNS

central nervous system

FCCP

carbonyl cyanide p-trifluoromethoxy-phenyl-hydrazone

IP3

inositol 1,4,5-trisphosphate

Δψ_m

mitochondrial membrane potential

SD

Leao's spreading depression

TMRE

tetramethylrhodamine ethyl ester