Abstract
This study used fluorescent indicator dyes to measure changes in cytosolic and mitochondrial [Ca2+] produced by physiological stimulation of lizard motor nerve terminals. During repetitive action potential discharge at 10–50 Hz, the increase in average cytosolic [Ca2+] reached plateau at levels that increased with increasing stimulus frequency. This stabilization of cytosolic [Ca2+] was caused mainly by mitochondrial Ca2+ uptake, because drugs that depolarize mitochondria greatly increased the stimulation-induced elevation of cytosolic [Ca2+], whereas blockers of other Ca2+ clearance routes had little effect. Surprisingly, during this sustained Ca2+ uptake the free [Ca2+] in the mitochondrial matrix never exceeded a plateau level of ∼1 μm, regardless of stimulation frequency or pattern. When stimulation ceased, matrix [Ca2+] decreased over a slow (∼10 min) time course consisting of an initial plateau followed by a return to baseline. These measurements demonstrate that sustained mitochondrial Ca2+ uptake is not invariably accompanied by progressive elevation of matrix free [Ca2+]. Both the plateau of matrix free [Ca2+] during stimulation and its complex decay after stimulation could be accounted for by a model incorporating reversible formation of an insoluble Ca salt. This mechanism allows mitochondria to sequester large amounts of Ca2+ while maintaining matrix free [Ca2+] at levels sufficient to activate Ca2+-dependent mitochondrial dehydrogenases, but below levels that activate the permeability transition pore.
- mitochondria
- mitochondrial calcium uptake
- presynaptic terminal
- motor nerve terminal
- mitochondrial matrix
- calcium indicator dyes
- calcium buffering
- calcium sequestration
Isolated mitochondria take up Ca2+ via a uniporter in the inner mitochondrial membrane (for review, see Gunter and Pfeiffer, 1990). Application of drugs that depolarize mitochondria has demonstrated that mitochondrial Ca2+ uptake is important for buffering/sequestering moderate-to-large Ca2+ loads in neuronal somata, axons, presynaptic terminals, and secretory cells (Alnaes and Rahamimoff, 1975; Baker and Schlaepfer, 1978; Åkerman and Nicholls, 1981;Martı́nez-Serrano and Satrústegui, 1992; Stuenkel, 1994; Friel and Tsien, 1994; Werth and Thayer, 1994; White and Reynolds, 1995; Herrington et al., 1996; Park et al., 1996; Sidky and Baimbridge, 1997). In support of this idea, fluorescent Ca2+ indicator dyes in the mitochondrial matrix report an increase in free [Ca2+] in depolarized bovine adrenal chromaffin cells (Babcock et al., 1997). In lizard motor nerve terminals, mitochondrial Ca2+ uptake contributes importantly to limiting the increase in average cytosolic [Ca2+] during brief stimulus trains (David et al., 1998).
The present study used this motor terminal preparation to study how mitochondria influence nerve terminal Ca2+metabolism during and after longer stimulus trains. If mitochondrial Ca2+ uptake is limited, occurring only during initial phases of stimulation, then other plasma membrane/endoplasmic reticular Ca2+extrusion/sequestration mechanisms would be expected to contribute importantly to Ca2+ handling during longer stimulus trains. If instead mitochondria continue to take up Ca2+ during prolonged stimulus trains, then Ca2+ in the mitochondrial matrix would have to be buffered or otherwise sequestered to prevent dissipation of the gradient permitting passive Ca2+ entry. Indeed, the bound/free ratio estimated for calcium in matrix is ∼3000–4000, compared with ∼100 estimated in cytosol (Babcock et al. 1997; Magnus and Keizer, 1997). If matrix Ca2+ is buffered primarily by conventional buffers, one would expect matrix free [Ca2+] to continue to rise throughout stimulation.
Mitochondria can also contain precipitated Ca salts. Electron microscopic studies have documented the formation of electron-dense structures, including granules, within the matrix of mitochondria exposed to large Ca2+ loads (LeFurgey et al., 1988), including mitochondria within nerve terminals subjected to intense stimulation (Párducz and Jóo, 1976; Jóo et al., 1980). An important question is whether such salt formation contributes to buffering of matrix Ca2+(for review, see Nicholls and Åkerman, 1982) or occurs only when mitochondria are exposed to pathologically high Ca2+ loads and/or certain fixation procedures (Somlyo et al., 1979) (for review, see Brown et al., 1985;Carafoli, 1987). If such salt formation occurs in stimulated nerve terminals, one would expect matrix free [Ca2+] to rise at first but then reach plateau at a level near the salt’s solubility product.
This study tested these hypotheses by measuring changes in cytosolic and matrix free [Ca2+] in lizard motor nerve terminals stimulated under physiological conditions. I demonstrate that matrix free [Ca2+] increases early during a stimulus train but then stops increasing after ∼100 action potentials, at a level that is relatively independent of stimulation frequency. Pharmacological evidence indicates that mitochondrial Ca2+ uptake continues, although matrix free [Ca2+] stops increasing. When stimulation stops, matrix free [Ca2+] exhibits a prolonged plateau followed by a decay to baseline over a time course of minutes. The pattern of changes in cytosolic and mitochondrial free [Ca2+] during and after a stimulus train can be accounted for by a model that includes reversible precipitation of a Ca salt in the mitochondrial matrix.
MATERIALS AND METHODS
Preparation. External intercostal neuromuscular preparations were dissected from lizards (Anolis sagrei) that were killed by pithing and decapitation after ether anesthesia. Preparations were mounted and viewed as described in David et al. (1998). The physiological saline contained (in mm): 157 NaCl, 4 KCl, 2 CaCl2, 2 MgCl2, and 1 HEPES. The motor nerve was stimulated by applying brief, suprathreshold depolarizing pulses (0.1–0.2 msec) via a suction electrode. Muscle contractions were usually blocked by adding carbachol (150–200 μm) or d-tubocurare (10 mg/l) to the bath. Experiments were performed at room temperature (20–25°C). In a few experiments (see Fig. 1B), end-plate potentials were monitored via a KCl-filled microelectrode inserted into the muscle fiber in the end-plate region [as in David et al. (1997)].
In some experiments (as noted in the figure legends), the visibility of the imaged terminal was improved by crushing the ends of the underlying muscle fiber with a sharp glass micropipette in saline containing trypsin (1 mg/ml, washed out 2–4 min after crushing the muscle). In some muscle fibers this treatment clarified the cytoplasm in the end-plate region, improving the visibility of motor nerve terminals synapsing on them (see Fig. 2A) and ensuring that fluorescence from mitochondria in the end-plate region of the muscle fiber (David et al., 1998) did not affect the measurements of mitochondrial [Ca2+] in nerve terminals. Stimulation-induced changes in cytosolic and mitochondrial [Ca2+] in trypsin-treated terminals were similar to those measured in preparations with intact muscle fibers.
Fluorometric measurement of cytosolic and mitochondrial [Ca2+]. Changes in cytosolic [Ca2+] were monitored using Oregon green-BAPTA 5N (OG-5N) loaded ionophoretically as the K salt via a microelectrode inserted into the motor axon (David et al., 1997). This form of OG-5N is membrane-impermeable and hence does not enter organelles.
Changes in mitochondrial [Ca2+] were monitored in terminals bath-loaded with the membrane-permeable acetoxymethylester (AM) forms of dihydrorhod-2, rhod-2, or OG-5N [5–10 μg/ml for 2–3 hr, prepared from 1000× stock solutions in dimethylsulfoxide (DMSO)]. Preparations were then washed with indicator-free medium for ≥3 hr before the onset of imaging. Dihydrorhod-2 fluoresces only after it is oxidized to rhod-2, which occurs preferentially within mitochondria (Hajnóczky et al., 1995). The AM forms of fluorescent indicator dyes can cross not only the plasma membrane but also the membranes surrounding intracellular compartments such as mitochondria, and the AM moiety can be cleaved by esterases in both cytosol and intracellular compartments. In this preparation, dyes loaded from the bath in their AM form using the protocol described above tended to localize within mitochondria, as judged by four criteria. First, fluorescence was clustered in the terminal rather than distributed continuously throughout the cytosol of the terminal and axon (Fig. 2, compare photographs in A,B). Second, after the onset of stimulation the increase in fluorescence began with a delay, in contrast to the immediate increase measured with dyes injected ionophoretically into the cytosol. [This differential delay is not evident on the time scales shown here but was evident with the faster sampling used in David et al. (1998).] Third, after stimulation ceased, fluorescence decayed with a very slow time course relative to the initial fast decay of cytosolic [Ca2+] (Fig. 2, compare A,B). Fourth, the stimulation-induced increase in fluorescence was blocked by agents that depolarize mitochondria [carbonyl cyanide m-chlorophenylhydrazone (CCCP), antimycin A1, rotenone; see Fig.2A] but not by agents that release Ca2+ from endoplasmic reticulum (ER) (Fig.4B).
A likely reason why AM-loaded dyes compartmentalized within organelles is that during the ≥3 hr washout period, dye in the terminal cytosol was diluted by diffusion into the axon. (The perineurial and myelin sheaths prevented axonal uptake of dye from the bath.) Thus the only dye remaining in the terminal was that contained within relatively nonmobile intracellular compartments. We cannot exclude the possibility that dye trapped within the ER made some contribution to the resting fluorescence, but the pharmacological manipulations mentioned above indicated that the main compartment exhibiting stimulation-induced changes in fluorescence was that of the mitochondria, which are abundant in these motor terminals (Walrond and Reese, 1985). Dye localization in this preparation therefore depended strongly on the loading technique, with ionophoretic injection of the salt filling primarily the cytosol, and bath-loading of the AM form filling primarily organelles such as mitochondria.
In experiments like that in Figure 1 involving simultaneous measurement of mitochondrial and cytosolic [Ca2+], mitochondrial rhod-2 and cytosolic OG-5N were excited with 488 nm light, and a dichroic mirror was used to separate the emitted light into a red component from rhod-2 (>570 nm) and a green component from OG-5N (535 ± 20 nm). Measurements were corrected for “cross-talk” between rhod-2 and OG-5N signals as described in David et al. (1998).
Rhod-2 binds Ca2+ with aKd of ∼0.5 μm. The Kd of OG-5N is much higher: ∼60 μm (see below). Thus changes in OG-5N fluorescence were approximately linearly related to changes in cytosolic and mitochondrial [Ca2+]. The exact concentration of dyes in the cytosol and mitochondrial matrix could not be measured, but OG-5N has such a low affinity that it was unlikely to buffer a significant amount of Ca2+ in cytosol or matrix. Also, fluorescence transients recorded in mitochondria filled with the high-affinity rhod-2 had time courses similar to those recorded in mitochondria filled with OG-5N. Thus the presence of dyes as exogenous buffers did not seriously distort the phenomena under study.
Fluorescence was monitored using a confocal laser-scanning microscope (Odyssey XL, Noran Instruments, Middleton, WI). Images were collected (≥0.266 sec/image) using an Indy workstation (Silicon Graphics) with Noran InterVision software, and stored and analyzed as described inDavid et al. (1997). The number of images collected in a given experiment was limited by the available computer memory and by the need to limit overall laser exposure to avoid tissue damage. Thus images were collected either frequently for a short period, or infrequently over longer periods. Fluorescence was plotted in one of two ways: (1) as ΔF/Frest (abbreviated as ΔF/F), where ΔF is the change in fluorescence and Frest is resting (prestimulation) fluorescence, or (2) as net fluorescence (after subtraction of background), to assess the slow decay of mitochondrial [Ca2+] and effects of drugs on resting [Ca2+]. Fluorescence was averaged over regions of interest that included all dye-labeled terminal regions. In dual-imaging studies the regions of interest defined for OG-5N were also used to analyze rhod-2 fluorescence.
OG-5N fluorescence changes were converted into estimates of [Ca2+] changes as described in David et al. (1997), using the following equation: [Ca2+] =Kd[(ΔF/Frest+ 1 − 1/β)/(Fmax/Frest− ΔF/Frest − 1)], where β =Frest/Fmin= (Kd + ([Ca2+]rest)(Fmax/Fmin))/(Kd+ [Ca2+]rest).
The Kd for OG-5N was 50–70 μm, and the ratioFmax/Fminranged from 20 to 30 in in vitro measurements made over the pH range 6.5–7.5 (Ying Wang and Dr. W. Glenn L. Kerrick, personal communication). Calculations assumed a resting free [Ca2+] of 100 nmfor both cytosol and mitochondria (for review, see Babcock and Hille, 1998).
Preparations were checked for stability by administering the same stimulus train several times at 15 min intervals. Under control conditions, ΔF/F transients were quite stable (Figs. 1C, 2 C), displaying little evidence of the “rundown” often observed in internally dialyzed cells. Terminals exhibiting abrupt increases in Frestwere not analyzed further.
Changes in mitochondrial membrane potential were monitored using 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetra-ethylbenzimidazolylcarbocyanine iodide (JC-1), 20 min exposure to 5 μg/ml, followed by wash (DiLisa et al., 1995). The excitation wavelength was 488 nm; an increase in the ratio of green (<570 nm) to red (>570 nm) emissions indicates depolarization.
Reagents. Indicator dyes were purchased from Molecular Probes (Eugene, OR). CGP 37157 was the kind gift of Dr. Anna Suter of Ciba-Geigy Pharmaceuticals (Basel, Switzerland). Ru360 was from Calbiochem (San Diego, CA). All other reagents were from Sigma (St. Louis, MO). Antimycin A1 and oligomycin were added from 1000× stock solutions in DMSO and ethanol, respectively. Cyclopiazonic acid (CPA) was added from 50 mm stock in DMSO, thapsigargin from 10 mm stock in DMSO, and caffeine was dissolved fresh in H2O. Digitonin solution was prepared from 10 mm stock in DMSO.
RESULTS
Mitochondrial free [Ca2+] attains a plateau during prolonged stimulation
Figure 1A plots stimulation-induced ΔF/F transients measured in a motor nerve terminal containing OG-5N in the cytosol and rhod-2 in mitochondria (see Materials and Methods). The terminal was stimulated with a 50 Hz train lasting 10 sec, with a prolonged 25 Hz train, and with an alternating pattern (50 Hz for 1 sec followed by 1 sec rest). During sustained stimulation at 50 and 25 Hz, both cytosolic and mitochondrial [Ca2+] increased rapidly at first but later stabilized at a plateau level. The amplitude of the cytosolic plateau was greater during 50 Hz than during 25 Hz stimulation [as also reported by Ravin et al. (1997); David et al. (1998); Ohnuma et al. (1999)], but the amplitude of the mitochondrial plateau was similar for both stimulation frequencies. During the intermittent stimulation pattern, cytosolic [Ca2+] fell rapidly during rest periods, but mitochondrial [Ca2+] did not. Rather, the amplitude of the mitochondrial fluorescence transient was similar for both steady and intermittent patterns of stimulation. The finding that the initial decay of mitochondrial [Ca2+] was much slower than that for cytosolic [Ca2+] agrees with the findings of Babcock et al. (1997) after depolarizing pulses in adrenal chromaffin cells.
Figure 1B shows that end-plate potentials continued to be evoked in the underlying muscle fiber throughout prolonged stimulus trains. Because evoked transmitter release requires Ca2+ entry through depolarization-activated Ca2+ channels, this result demonstrates that the failure of cytosolic and mitochondrial free [Ca2+] to continue rising during maintained stimulation was not caused by cessation of Ca2+ entry across the plasma membrane.
The plateau of mitochondrial [Ca2+] during stimulation is not an artifact of dye saturation
The high Ca2+ affinity of rhod-2 raises the possibility that the plateau of its fluorescence during prolonged stimulation might have been caused by dye saturation. Figure1C presents evidence that this was not the case: the maximal fluorescence of mitochondrial rhod-2 during 50 Hz stimulation was less than that measured when the terminal was subsequently exposed to ionomycin (a Ca2+ ionophore) in the presence of 2 mm bath Ca2+.
To test further whether the fluorescence increase of mitochondrial indicator dyes was limited by dye saturation, mitochondria were instead AM-loaded with the lower-affinity OG-5N, whose fluorescence is unlikely to saturate at any [Ca2+] achieved within living cells (Fig.2A). The OG-5N ΔF/F transients in Figure 2Aconfirm its mitochondrial localization: the stimulation-induced fluorescence increase declined slowly (rather than rapidly) after stimulation and was blocked by antimycin A1, which collapses the membrane potential across the inner mitochondrial membrane by irreversibly inhibiting complex III in the electron transport chain.
In contrast, when OG-5N was injected ionophoretically into the axonal cytosol, its fluorescence was visible within the axon as well as the terminal (Fig. 2B). The stimulation-induced ΔF/F transient was increased by antimycin A1 and showed an initial steep decline when stimulation stopped (as in Fig. 1A, a,c). Rotenone (5 μm), which inhibits electron transport at complex I, produced effects similar to those of antimycin A1 (data not shown). Oligomycin, which inhibits ATP synthase but does not depolarize mitochondria, had no effect on either cytosolic or mitochondrial ΔF/F transients (Fig.2A,B), indicating that the changes produced by antimycin A1 and rotenone were caused by mitochondrial depolarization rather than inhibition of mitochondrial ATP synthesis.
Figure 2C shows that the fluorescence of OG-5N in mitochondria did not saturate during repetitive stimulation, because a larger fluorescence increase was measured when terminals were exposed to 2 mm bath Ca2+ in the presence of the detergent digitonin. Similar results were seen in another terminal permeabilized with ionomycin. Thus results in Figures1D and 2C, combined with the fact that the plateau of mitochondrial fluorescence during prolonged stimulation was measured with low- as well as high-affinity dyes, suggest that this plateau was caused by stabilization of free [Ca2+] in the matrix rather than saturation of the indicator dyes.
A possible concern about the experiments with ionomycin and digitonin is that some of the increase in fluorescence measured in AM-loaded preparations with 2 mm bath Ca2+ might have come from dye in ER, but any contribution from dye in the ER was probably negligible, because application of agents that deplete ER Ca2+produced no change in resting fluorescence. Also, other conditions that would have been expected to cause ER Ca2+to rise, e.g., stimulation in the presence of mitochondrial inhibitors, did not increase the fluorescence of AM-loaded dyes (Fig.2A).
Mitochondrial free [Ca2+] plateaus at ∼1 μm during repetitive stimulation
Figure 2D shows mitochondrial OG-5N ΔF/F transients recorded during repetitive stimulation at 50, 25, and 10 Hz. The rate of rise of mitochondrial [Ca2+] increased as stimulation frequency increased, but the final steady level was similar for all three frequencies (see also mitochondrial rhod-2 transients in Fig. 1A, d–f).
During 50 Hz stimulation, the maximal ΔF/F for cytosolic OG-5N averaged 0.20 ± 0.06 (SD, n = 18 terminals), whereas that for mitochondrial OG-5N averaged 0.43 ± 0.20 (n = 12). Assuming that OG-5N in cytosol and matrix has properties similar to those measured in vitro, these ΔF/F values suggest an increase in cytosolic [Ca2+] within the range 0.35–0.74 μm and an increase in matrix [Ca2+] within the range 0.75–1.62 μm (see Materials and Methods). The latter estimate might be too low if some of the resting fluorescence of AM-loaded dyes originates from intracellular compartments other than mitochondria.
Repetitive nerve stimulation produces no detectable mitochondrial depolarization
Figure 3 shows measurements of the emissions ratio of JC-1, an index of mitochondrial depolarization, during 50 Hz, 10 sec stimulus trains (Fig. 3A) and after application of oligomycin and antimycin A1 (Fig. 3B). The stimulus train, similar to that applied in Figures1A and 2A, produced no detectable depolarization. The mitochondrial membrane potential was also maintained in oligomycin but depolarized when antimycin A1 was added (Fig. 3B).
The plateau of mitochondrial [Ca2+] is not caused by accelerated activity of non-mitochondrial Ca2+ sequestration/extrusion mechanisms
One possible explanation for the finding that both cytosolic and mitochondrial [Ca2+] attain steady states (rather than continuing to increase) during prolonged stimulation is a “relay race” hypothesis in which the Ca2+ that enters stimulated terminals is sequestered first by cytosolic buffers, second by mitochondrial uptake, and subsequently by other transport mechanisms in the ER and/or plasma membrane. Figure 4Ashows that the magnitude of the stimulation-induced increase in mitochondrial [Ca2+] was not altered when the plasma membrane Na+/Ca2+exchanger was inhibited by substituting Li+ for Na+in the bathing solution. Figure4B,C shows that CPA (20 μm), an inhibitor of ER Ca2+-ATPase, did not alter the magnitude of mitochondrial or cytosolic [Ca2+] transients, respectively. Caffeine (10 mm) (Fig.4B), ryanodine, and thapsigargin (each at 10 μm; data not shown), which deplete certain ER Ca2+ stores, also produced no detectable change. CPA, caffeine, and ryanodine produced muscle contractures; the illustrated records were taken after these contractures subsided. Figure 4D shows that inhibition of plasma membrane Ca2+-ATPase activity with alkaline pH (Milanick, 1990) was similarly without effect on the magnitude of cytosolic [Ca2+] transients. These treatments to inhibit ER Ca2+sequestration and plasma membrane Ca2+extrusion also had little effect on resting cytosolic or mitochondrial [Ca2+] over the time course studied here (data not shown).
In contrast, after addition of CCCP (1 μm), a protonophore that depolarizes mitochondria and thus inhibits mitochondrial Ca2+ uptake, cytosolic [Ca2+] continued to increase throughout the stimulus train (Fig. 4C,D). These results, as well as those in Figure 2A,B, suggest that mitochondria continue to play a dominant role in limiting the increase in cytosolic [Ca2+] during prolonged stimulation, even after the mitochondrial free [Ca2+] reported by intramitochondrial dyes has stopped increasing.
I also tried to test for continued mitochondrial Ca2+ uptake using Ru360, reported to block selectively the mitochondrial Ca2+uniporter in heart cells (Matlib et al., 1998). In motor nerve terminals, however, Ru360 (10–100 μm) produced a reversible, concentration-dependent reduction in stimulation-induced cytosolic [Ca2+] transients (data not shown), suggesting that Ru360 reduced Ca2+entry via plasma membrane Ca2+ channels (N-type) (David et al., 1997). Thus the effects of Ru360 were not specific for the uniporter in this preparation.
If during sustained stimulation mitochondria stabilize matrix free [Ca2+] by increasing extrusion of Ca2+ via the mitochondrial Na+/Ca2+exchanger, inhibition of this exchanger with CGP 37157 (Cox and Matlib, 1993) would be expected to reduce the cytosolic [Ca2+] measured during stimulation. Figure 5A shows that CGP 37157 did not have this effect but did reduce a slowly decaying component of the post-stimulation cytosolic [Ca2+] transient thought to reflect Ca2+extrusion from mitochondria, as also reported by Babcock et al. (1997). Cyclosporin A, which inhibits the mitochondrial permeability transition pore, another possible route of mitochondrial Ca2+ extrusion, had no effect on cytosolic [Ca2+] transients during or after sustained stimulation (data not shown). These results thus suggest that the plateau of mitochondrial [Ca2+] during prolonged stimulation is not caused by enhanced mitochondrial Ca2+ extrusion into the cytosol.
Some of the Ca stored in mitochondria is readily releasable
The above results suggest that during prolonged stimulation matrix free [Ca2+] stabilizes, although mitochondria continue to take up Ca2+. Such a result might occur if a Ca salt precipitates within the mitochondrial matrix. If so, one might wonder how readily the Ca within such a precipitate could be released. Figure 5B shows that application of CCCP after a brief stimulus train (100 stimuli at 50 Hz) produced no detectable elevation of cytosolic [Ca2+], but that a similar CCCP application after a longer train (500 stimuli) did elevate cytosolic [Ca2+]. However, the elevation of cytosolic [Ca2+] produced by post-train CCCP was small in comparison with the severalfold increase in cytosolic [Ca2+] recorded when CCCP was appliedbefore the stimulus train (Fig. 4C,D). The smaller post-train elevation of cytosolic [Ca2+] suggests that elevating cytosolic Ca2+ via efflux from mitochondria is slower than elevating cytosolic Ca2+ via depolarization-activated Ca2+ channels in the plasma membrane. A more quantitative assessment of the effect of train duration on CCCP-evoked ΔF/F transients was not possible because of the muscle contractions often evoked by CCCP application and the long intertrain intervals required to reverse CCCP effects.
Mitochondrial [Ca2+] decays with a slow and complex time course after stimulation
Figure 6A shows that the elevation of mitochondrial [Ca2+] produced by a train of 500 stimuli decayed with a complex time course lasting ∼10 min. Immediately after the train, mitochondrial [Ca2+] decayed very slowly, followed by a more rapid phase of decay to baseline (see also Fig. 6C). The prolonged and complex time course of decay was not caused by light damage or dye bleaching induced by repeated laser scanning, because a similarly prolonged decay was evident when imaging was minimized during the decay (Fig. 6B). Also, post-train decays were similar (within experimental error) before and after an interval without imaging (Fig. 6C). The slow decay was not caused by the intramitochondrial dye acting as an exogenous buffer, because the same slow decay was seen in mitochondria filled with the low-affinity dye OG-5N (data not shown), which would be expected to bind less Ca2+ than the higher affinity rhod-2. This slow release of mitochondrial Ca2+contributes a slow component to the decay of cytosolic [Ca2+] and hence to short-term synaptic memory (Tang and Zucker, 1997). It is unclear whether matrix [Ca2+] ever falls as low in vivo as the baseline values measured here, because intercostal motor nerve terminals in a breathing animal might never experience 15 min periods of inactivity.
Another interesting feature of the fluorescence transients shown in Figure 6B,C is that the increment in mitochondrial free [Ca2+] produced by a 50 Hz train (1–10 sec) was larger when a stimulus train was applied to terminals that had been inactive for at least 15 min before stimulation than when the same stimulus train was applied at briefer intertrain intervals. The larger increase in mitochondrial free [Ca2+] produced by the initial train was not caused by a greater elevation of cytosolic [Ca2+], because the cytosolic fluorescence transient evoked by the initial train was similar to that evoked by subsequent stimulus trains (Fig. 6D). The increase in mitochondrial free [Ca2+] produced by a given train may depend on the initial value of mitochondrial free [Ca2+], with large increases when the initial value is low (after a long rest) and smaller increases when the initial value is closer to the “ceiling” level of mitochondrial free [Ca2+] measured in continuously stimulated terminals (Figs. 1, 2, 4, 6).
Changes in mitochondrial free [Ca2+] during and after stimulation are consistent with reversible precipitation of a Ca salt in the mitochondrial matrix
One hypothesis to explain the stability of matrix free [Ca2+] during continued mitochondrial Ca2+ uptake is precipitation of Ca salt(s) in the matrix. This hypothesis predicts that mitochondrial Ca2+ uptake will be accompanied by increases in matrix free [Ca2+] during the early stages of stimulation, but that as stimulation and mitochondrial Ca2+ uptake continue, matrix free [Ca2+] will stop increasing. Thus at these later stages matrix free [Ca2+] becomes independent of total mitochondrial Ca content (Nicholls and Åkerman, 1982). Figure 7Adiagrams such a model, and Figure7B,C shows simulated cytosolic and mitochondrial [Ca2+] transients, demonstrating that this model can account (at least qualitatively) for several features of the measured transients. For example, Figure7B (a,b,d,e) shows that as stimulation frequency increases from 25 to 50 Hz, the model predicts transients similar to those measured in Figures1A and 2D, i.e., an increase in the maximal level of cytosolic [Ca2+] and in the initial rate of rise of mitochondrial [Ca2+] but no increase in the maximal level of matrix free [Ca2+].
During the early post-train period, mitochondria extruded Ca2+ into the cytosol, as evidenced by the effect of inhibiting the mitochondrial Na+/Ca2+exchanger with CGP 37157 (Fig. 5A). However, this initial Ca2+ extrusion sometimes produced little or no reduction in matrix free [Ca2+] (Fig. 6A,C). The Ca salt precipitation hypothesis accounts for the delayed fall of matrix free [Ca2+] after stimulation by predicting that matrix free [Ca2+] will remain elevated as long as the Ca salt is dissolving and fall only after all salt has dissolved (Fig. 7B, f).
During prolonged inactivity, matrix free [Ca2+] fell to low levels (Fig.6A,B). In the model (Fig.7C, a,d), as in the measurements of Figure 6B–D, intermittent stimulation delivered after a prolonged rest produced mitochondrial [Ca2+] transients that were larger for the first few stimulus trains, although cytosolic [Ca2+] transients were similar for all trains. An explanation for these findings is that total mitochondrial Ca2+ uptake was similar for all stimulus trains, but large increments in matrix free [Ca2+] occurred only until the salt solubility product was exceeded, with further stimulus trains producing only small additional increases in matrix free [Ca2+]. The model simulated the effects of mitochondrial depolarizing agents (antimycin A1 as in Fig.2A,B; CCCP as in Fig.4C,D) by blocking the uniporter (Fig.7C, b,e).
In the simulations of Figure 7, the poorly soluble matrix salt was assumed to behave like hydroxyapatite, the least soluble Ca phosphate salt at the slightly alkaline pH (7.3–7.5) of the mitochondrial matrix. Phosphate is abundant in the mitochondrial matrix, and uptake of Ca2+ into isolated mitochondria is greatly increased in the presence of phosphate (Mela and Hess, 1982). In the model, the free concentration of inorganic phosphate in the matrix was assumed to be maintained by a phosphate transporter. The inner mitochondrial membrane contains a fast phosphate transporter (Ligeti et al., 1985), and total mitochondrial phosphate increases with Ca2+ uptake (Greenawalt et al., 1964). When Ca2+ uptake was simulated with only non-phosphate buffers in the matrix (Fig. 7C,c,f), matrix free [Ca2+] rose to levels much higher than those measured, and the initial recovery of cytosolic [Ca2+] was much slower. Ligeti and Lukacs (1984) found that in isolated rat liver mitochondria the presence of phosphate not only facilitated Ca2+ uptake but also prevented depolarization of the mitochondrial membrane potential during Ca2+ uptake, consistent with the absence of detectable mitochondrial depolarization during the stimulation patterns studied here (Fig. 3).
An additional argument suggests that mitochondria take up buffer (such as phosphate) during prolonged stimulation. If total matrix Ca2+ buffer remained constant, then one would expect that during the sustained Ca2+ influx associated with prolonged stimulation, the rate of change of mitochondrial free [Ca2+] during stimulation would always be greater than zero. Yet measurements with both low- and high-affinity dyes showed that matrix free [Ca2+] stopped increasing during prolonged stimulus trains.
Absence of phosphate transport could not be achieved experimentally. The sulfhydryl reducing agents mersalyl and n-ethymaleimide block phosphate transport, but their effects are not specific. Measured fluorescence transients were similar with (Fig. 2A) or without inorganic phosphate in the bathing solution, but it is unlikely that phosphate-free bathing solutions produced a large reduction in intraterminal phosphate because of the reservoir of phosphate in the myelinated portion of the axon.
DISCUSSION
Motor nerve terminals allow the study of changes in cytosolic and mitochondrial [Ca2+] produced by controlled rates and patterns of physiological stimulation (action potential discharge) in a nondialyzed, stable preparation. The preparation permits loading of the mitochondrial matrix with either low- or high-affinity indicator dyes. Previous work demonstrated that mitochondrial [Ca2+] in this preparation begins to increase after trains of only 25–50 action potentials, and that this mitochondrial Ca2+ uptake limits the increase in cytosolic [Ca2+] during repetitive discharge (David et al., 1998). Work presented here shows that after trains of as few as 100–200 action potentials, mitochondrial free [Ca2+] stops increasing, although mitochondrial uptake continues to be a major route for removing Ca2+ from the cytosol. After stimulation, mitochondrial free [Ca2+] decays slowly (∼10 min) with a complex time course.
The plateau of mitochondrial free [Ca2+] during stimulation is not caused by saturation of a rapid uptake mode
Figure 7 presented a model suggesting that the plateau of matrix free [Ca2+] measured during prolonged stimulation results from formation of an insoluble Ca salt within the mitochondrial matrix. Another idea that might be proffered to explain this plateau is inactivation of a “rapid uptake mode” described in isolated liver and heart mitochondria (Sparagna et al., 1995; Gunter et al., 1998). This rapid uptake mode is hypothesized to operate at cytosolic [Ca2+] levels lower than those thought necessary to activate the mitochondrial uniporter, to inactivate rapidly, and to be “recharged” within a few seconds after cytosolic [Ca2+] falls. However, inactivation of such a rapid uptake mode would cause cytosolic [Ca2+] to keep increasing during maintained stimulation, whereas recordings showed that cytosolic [Ca2+] remained stable after matrix free [Ca2+] stopped increasing (Fig.1A). In contrast, when mitochondrial Ca2+ uptake was inhibited with CCCP or antimycin A1, cytosolic [Ca2+] increased throughout the stimulus train. Thus it appears that in normal saline lacking these inhibitors, mitochondrial uptake of Ca2+ did not inactivate or diminish during the later portions of the stimulus train. Another argument against the idea that the limited increase in mitochondrial free [Ca2+] was caused by inactivation of a rapid uptake mode is that the maximal level of mitochondrial free [Ca2+] did not increase when stimulus trains were separated by intervals long enough (5 min) to recharge the hypothesized rapid uptake mechanism (Fig. 6C).
The insoluble Ca salt hypothesis outlined in Figure 7 offers an explanation for not only the plateau of matrix free [Ca2+] during prolonged stimulation, but also for other aspects of the recorded fluorescence transients such as the complex time course of decay and the dependence of transient amplitude on the intertrain interval. Many uncertainties remain concerning this model: for example, the composition of the hypothesized poorly soluble Ca salt(s), possible effects of matrix or membrane components on the solubility of these salts, the time course of their precipitation and solvation, and other possible mechanisms (besides the mitochondrial Na+/Ca2+exchanger) for removing Ca2+ from mitochondria. Despite these limitations, the hypothesis that much of the Ca2+ taken up by motor terminal mitochondria during sustained stimulation is removed from solution in the matrix, and that this process is reversible and repeatable under physiological conditions, seems to provide the best explanation for presently available data.
Nerve terminal mitochondria accumulate Ca2+ at relatively low cytosolic [Ca2+]
Mitochondria in motor nerve terminals begin to take up Ca2+ when the average cytosolic [Ca2+] is ≤300 nm[stimulation-induced 200 nm increase (David et al., 1998) superimposed on an assumed resting level of ∼100 nm]. Mitochondrial [Ca2+] uptake at comparably low average cytosolic [Ca2+] has been demonstrated in other cell types (Rizzuto et al., 1994, their Fig. 4; Babcock et al., 1997). This concentration is below the threshold [Ca2+] needed to activate uniporter-mediated steady Ca2+ uptake in isolated liver and heart mitochondria. Various suggestions have been made to reconcile this discrepancy between measurements in isolated versus in situ mitochondria. One suggestion is that cytosolic factors (e.g., spermine) (Rustenbeck et al., 1993) increase the sensitivity of the mitochondrial uniporter. Another suggestion (discussed above) is that mitochondria have a rapid Ca2+ uptake mode, which has a higher affinity than the standard uniporter but inactivates during sustained exposure to cytosolic Ca2+.
It has also been hypothesized that mitochondrial Ca2+ uptake occurs mainly in high [Ca2+] domains within the cell, localized, for example, near IP3-activated Ca2+ release channels in ER or Ca2+ channels in the plasma membrane (Svichar et al., 1997; Peng and Greenamyre, 1998; Csordás et al., 1999) (for review, see Rutter et al., 1998). However, several arguments suggest that stimulation-induced Ca2+uptake into motor terminal mitochondria is not restricted to localized, high [Ca2+] domains. First, synaptic vesicles (rather than mitochondria) are localized near the plasma membrane voltage-activated Ca2+channels that face the synaptic cleft (Lichtman et al., 1989). Second, local Ca2+ domains near the plasma membrane of these motor terminals are very transient, dissipating within ∼15 msec after stimulation, whereas the increase in mitochondrial [Ca2+] lagged hundreds of milliseconds behind the increase in average cytosolic [Ca2+] (David et al., 1997, 1998). Local domains around ER are unlikely to be important in these terminals, because drugs that deplete ER Ca2+ stores did not alter the maximal amplitude of the stimulation-induced increases in cytosolic and mitochondrial [Ca2+] (Fig.4B,C) (see also Tang and Zucker, 1997). Studies reporting pharmacological evidence for ER Ca2+ stores in motor terminals (Narita et al., 1998) used stimulation patterns much more prolonged than those applied here. Thus there is no evidence that Ca2+ uptake into motor terminal mitochondria during moderate stimulation required [Ca2+] substantially greater than the average cytosolic values reported by indicator dyes.
Mitochondrial buffering of cytosolic Ca2+ may be especially prominent in synaptic terminals
Data presented here demonstrate that mitochondrial uptake is the dominant mechanism limiting the increase in cytosolic [Ca2+] in lizard motor nerve terminals subjected to stimulation exceeding 25–50 action potentials at 25–100 Hz, well within the physiological range of activation of these terminals. Mitochondria appear to be similarly dominant in other synaptic terminals and in adrenal chromaffin cells (Martı́nez-Serrano and Satrústegui, 1992; Babcock et al., 1997; Tang and Zucker, 1997). However, there is abundant evidence for important ER contributions to handling of Ca2+ loads in neuronal somata (Garaschuk et al., 1997; Fierro et al., 1998). Perhaps mitochondrial Ca2+ uptake is especially important in cellular regions with large surface-to-volume ratios that are regularly subjected to large, rapid Ca2+ influxes requiring rapid sequestration. Here the ability of mitochondria to take up large amounts of Ca2+ by rapid passive transport might have a pronounced advantage over the slower active transport required to pump Ca2+ into ER stores or across the plasma membrane.
In summary, evidence presented here demonstrates that during repetitive stimulation the free [Ca2+] in the mitochondrial matrix increases to up to ∼1 μm, with the same maximum observed over a range of stimulation frequencies. This concentration is sufficient to activate multiple Ca2+-sensitive mitochondrial dehydrogenases (for review, see McCormack et al., 1990), which are thought to help link energy production to energy demand. Mitochondria remain able to take up substantial additional Ca2+ after this maximal matrix free [Ca2+] is attained, most likely because of import of buffer (probably phosphate) and (reversible) precipitation of Ca salts within the matrix. The hypothesized salt formation would allow mitochondria to sequester temporarily the large Ca2+ loads associated with stimulation of synaptic terminals while preventing excessive run-down of the mitochondrial membrane potential or elevation of intramitochondrial free [Ca2+] to levels that might activate the mitochondrial permeability transition pore.
Footnotes
This work was supported by National Institutes of Health Grant RO1 NS 12404. I thank Drs. Ellen Barrett and John Barrett for help with this manuscript and valuable discussions, Ying Wang and Dr. Glenn Kerrick for measuring the Ca2+ affinity of Oregon green BAPTA-5N, and Dr. Anna Suter (Ciba-Geigy Pharmaceuticals) for providing CGP 37157.
Correspondence should be addressed to Dr. Gavriel David, Department of Physiology and Biophysics, R-430, University of Miami School of Medicine, P.O. Box 016430, Miami FL 33101.