Abstract
Mitochondria and endoplasmic reticulum (ER) are important modulators of intracellular calcium signaling pathways, but the role of these organelles in shaping synaptic calcium transients in dendrites of pyramidal neurons remains speculative. We have measured directly the concentrations of total Ca (bound plus free) within intracellular compartments of proximal dendrites of CA3 hippocampal neurons at times after synaptic stimulation corresponding to the peak of the cytoplasmic free Ca2+ transient (1 sec), to just after its decay (30 sec), and to well after its return to prestimulus levels (180 sec). Electron probe microanalysis of cryosections from rapidly frozen slice cultures has revealed that afferent mossy fiber stimulation evokes large, rapid elevations in the concentration of total mitochondrial Ca ([Ca]mito) in depolarized dendrites. A single tetanus (50 Hz/1 sec) elevated [Ca]mito more than fivefold above characteristically low basal levels within 1 sec of stimulation and >10-fold by 30 sec after stimulation. This strong Ca accumulation was reversible, because [Ca]mito had recovered by 180 sec after the tetanus. Ca sequestered within mitochondria was localized to small inclusions that were distributed heterogeneously within, and probably among, individual mitochondria. By 30 sec after stimulation an active subpopulation of ER cisterns had accumulated more Ca than had mitochondria despite a ∼1 sec delay before the onset of accumulation. Active ER cisterns retained their Ca load much longer (>3 min) than mitochondria. The complementary time courses of mitochondrial versus ER Ca2+ uptake and release suggest that these organelles participate in a choreographed interplay, each shaping dendritic Ca2+ signals within characteristic regimes of cytosolic Ca2+ concentration and time.
- calcium
- endoplasmic reticulum
- mitochondria
- hippocampus
- synaptic activity
- electron probe x-ray microanalysis
Mitochondria and the endoplasmic reticulum (ER) are the main organellar components of a cellular calcium regulatory system that is essential for shaping intracellular calcium signals. Although the importance of the ER has been recognized for decades (for review, see Berridge, 1998; Meldolesi, 2001), our understanding of the role of mitochondrial Ca2+ transport recently has undergone considerable evolution (Friel, 2000; Pozzan and Rizzuto, 2000). There is now accumulating evidence that in a variety of cell types, including neurons, elevation of the free cytosolic Ca2+ concentration ([Ca2+]i) leads to mitochondrial Ca2+ uptake and that this has a major impact on Ca2+ signaling (Babcock and Hille, 1998; Duchen, 1999; Pozzan and Rizzuto, 2000). Thus, somatic mitochondria in several neuronal cell types buffer physiological calcium loads elicited by depolarization and/or action potentials (Thayer and Miller, 1990; Friel and Tsien, 1994; White and Reynolds, 1995; Herrington et al., 1996; Babcock et al., 1997). Under pathological conditions, e.g., excessive glutamate exposure, mitochondria of CNS neurons accumulate large amounts of Ca (Budd and Nicholls, 1996; Stout et al., 1998; Brocard et al., 2001), ultimately leading to mitochondrial dysfunction and cell death (Reynolds, 1999;Nicholls and Budd, 2000).
Regarding synaptic Ca2+ transients, mitochondrial Ca2+ transport contributes to the regulation of [Ca2+]i in presynaptic terminals during normal transmission (David et al., 1998;Peng, 1998) and post-tetanic potentiation [Tang and Zucker (1997), but see Zenisek and Matthews (2000)]. To date, however, the role of mitochondria in modulating spatiotemporal patterns of synaptically evoked postsynaptic Ca2+ transients, for example in dendrites of CNS pyramidal neurons, is unclear. Yet dendritic mitochondria are potentially of great importance because, given the subcellular architecture of dendrites, they are exposed to substantial [Ca2+]i elevations after synaptic activity (Petrozzino et al., 1995; Helmchen et al., 1996).
In a previous study we found no elevation of mitochondrial total Ca concentration ([Ca]mito) in hippocampal CA3 pyramidal dendrites at 3 min after synaptic stimulation, although a subset of ER contained large amounts of Ca at this and even later times (Pozzo-Miller et al., 1997). These results revealed a remarkably long persistence for ER Ca sequestration but did not address the subcellular distribution of Ca at times close to synaptically driven [Ca2+]itransients, which peak and decay within seconds of synaptic stimulation (Regehr et al., 1989; Jaffe et al., 1992; Magee et al., 1995;Petrozzino et al., 1995; Spruston et al., 1995). We now have determined the distribution of Ca within the ER and mitochondria of proximal CA3 dendrites at the peak of the free Ca2+transient (1 sec), just after its decay (30 sec), and well after its return to prestimulus levels (180 sec). A large rise and fall of [Ca]mito was observed during the first minutes after synaptic activity in parallel with a delayed but sustained Ca accumulation in a subset of ER. These results indicate that in pyramidal dendrites both ER and mitochondria reversibly accumulate Ca but with characteristically different sequestration mechanisms and complementary temporal properties that are likely to be important in the modulation of dendritic Ca2+signals.
MATERIALS AND METHODS
Stimulation and rapid freezing of organotypic hippocampal slice cultures. Slice cultures of hippocampus were prepared from postnatal day 7 rats in the manner of Stoppini et al. (1991). Such organotypic slice cultures are widely accepted experimental models (for review, see Gähwiler et al., 1997) and are essential for the present experiments because they provide healthy superficial neuronal cell bodies, dendrites, and synapses that are close enough to the surface to allow for rapid freezing with excellent structural preservation (Pozzo-Miller et al., 1993) and no elemental redistribution. The specific experimental setup is essentially similar to that described and schematically diagrammed in Pozzo-Miller et al. (1997), with one significant modification: two insulated nichrome wires (20 μm in diameter; California Fine Wire, Grover Beach, CA) for afferent fiber stimulation were attached permanently to custom-modified 12 mm Millicell-CM filter inserts (Millipore, Bedford, MA). The slice grew above one exposed uninsulated section of the wire that was fixed flush against the filter insert membrane, such that the wire ran under the dentate gyrus. The presence of the wires had no discernible effect on the growth, anatomy, or synaptic circuitry of the cultures. Just before a slice was transferred to a recording chamber, the exposed uninsulated section of the second wire was lowered onto the slice parallel to the lower wire, with a spacing of ∼1 mm, and was fixed to the edge of the insert with silicone vacuum grease. The recording chamber was perfused continuously with oxygenated artificial CSF (aCSF) containing (in mm) 124 NaCl, 2 KCl, 1.24 KH2PO4, 1.3 MgSO4, 17.6 NaHCO3, 2.5 CaCl2, and 10 d-glucose; the solution was bubbled with 95% O2/5% CO2. Field EPSPs (fEPSPs) were evoked at low frequency by single pulses (100 μsec duration) delivered to the top wire by a multichannel stimulator and isolator set-up (Master-8 and ISO-Flex, AMPI, Jerusalem, Israel) and were recorded with a bridge amplifier (Axoclamp-1A, Foster City, CA), using an extracellular glass microelectrode filled with aCSF (2 MΩ final resistance) positioned in CA3 stratum lucidum. This arrangement was effective for generating fEPSPs along the apical CA3 dendritic field and permitted selection of the stimulus strength that was necessary to evoke subthreshold fEPSPs (∼0.5 mV, without superimposed population spike) in each culture individually.
For slices with a 1 sec interval between stimulation and freezing, cultures with the wires still in position were positioned briefly under a dissecting microscope to wick off adventitious surface fluid and to place (with a fine needle) three delicate copier toner marks that later would be used to triangulate the CA3 region in frozen specimens. Cultures then were mounted on custom-made freezing stages that adapted the Millicell-CM inserts to a custom-modified rapid freezing machine (LifeCell CF-100; The Woodlands, TX). These steps required ∼10 sec; in no case was the tissue touched or otherwise disturbed. The stimulator was used to trigger the freezing machine as well as the stimulus train, allowing rapid freezing at accurately programmable intervals as short as a few milliseconds between the delivery of a stimulus to mossy fibers and the instant of impact on the freezing block. The stimulus train used here, 50 Hz/1 sec, is a standard high-frequency train and is identical to that described as the “single train” in previous work (Pozzo-Miller et al., 1997). For 30 sec (and longer) experiments the interval is long enough that cultures can be stimulated alternatively in the recording chamber (and therefore can use cultures grown with or without embedded stimulating wires) and then mounted on stages and rapidly frozen. Experiments were performed to evaluate all permutations, and no differences were found among the various procedures. Controls, i.e., nonstimulated cultures, were subjected to all experimental manipulations except delivery of the tetanus. Again, no differences were observed between controls frozen directly from the recording chamber and those premounted on freezing stages.
Experiments were performed on slices 6–8 d in vitro. This relatively short time avoids one disadvantage of slice cultures, namely, that aberrant sprouting by intrinsic fibers may occupy vacant postsynaptic sites, leading to hyperexcitable slices. This isolation-induced sprouting becomes a significant problem only after 2 weeks in vitro (Pozzo-Miller et al., 1994).
Analytical electron microscopy. The principles and techniques for cryosectioning, electron microscopy, and electron probe x-ray microanalysis (EPMA) for measuring total (free plus bound) concentrations of diffusible elements at the subcellular level are well established (Kitazawa et al., 1983; Somlyo, 1985) (for review, seeRoomans and Von Euler, 1996; Meldolesi and Grohovaz, 2001). Techniques for cryosectioning and EPMA within organelles of pyramidal neurons in slice cultures are described in Pozzo-Miller et al. (1997). Briefly, cryosections (80 nm nominal thickness) were prepared from the well frozen specimen face by means of a Leica Ultracut S/FCS ultracryomicrotome (Deerfield, IL). Sections were mounted on carbon- and Formvar-coated grids and cryo-transferred into an EM912 Omega electron microscope (LEO Electron Microscopy, Thornwood, NY) equipped with a Linksystem Pentafet energy-dispersive x-ray (EDX) detector (Oxford Instruments, Concord, MA) and a ProScan HSSC-1 slow-scan CCD camera (1024 × 1024) interfaced to AnalySIS software (Soft-Imaging Software GmbH, Munster, Germany). Sections were freeze dried in the microscope at approximately –100°C and then recooled to approximately –170°C for imaging and x-ray analysis. Individual x-ray spectra were recorded for 100 sec at ∼4 nA probe current, using a focused probe with a diameter of 100 nm for mitochondria and cytoplasm and ≤50 nm for ER cisterns.
Spectra subsequently were processed and quantified using the program DeskTop Spectrum Analyzer (DTSA) for the Macintosh [C. E. Fiori, C. R. Swyt, R. L. Myklebust (1993) Office of Standard Reference Data, National Institute of Standards and Technology, Gaithersburg, MD]. A simplex fitting routine and the peak/continuum method (Kitazawa et al., 1983) were used to quantify the concentrations of the following elements in units of mmol/kg dry weight: Na, Mg, P, Cl, K, and Ca. Methods for converting concentrations to mmol/kg wet weight (the units for data presentation in Figs. 2B,4) are described by Roomans and Von Euler (1996) and by Pozzo-Miller et al. (2000). Relative dry mass fractions required for this calculation were derived from x-ray continuum counts (Buchanan et al., 1993) specifically from these experiments. Mass fractions were 0.15 ± 0.02, 0.27 ± 0.02, and 0.45 ± 0.03 for cytoplasm, ER, and mitochondria, respectively; they were distributed normally and are consistent with published data (Pozzo-Miller et al., 1997, 2000). Throughout this work elemental symbols with the oxidation state specified, e.g., Ca2+, are used conventionally to indicate ions in the free, water-solvated state, whereas symbols without such specification, e.g., Ca, refer to the element without regard to whether it is free or bound and are used mainly to indicate total concentrations as measured by EPMA.
Sampling, data analysis, and statistics. Individual x-ray spectra were recorded from the major Ca2+-regulating compartments of CA3 pyramidal dendrites, namely, cytoplasm, mitochondria, and ER. The term “ER” is used by convention throughout this paper to refer to structurally identified smooth-membrane cisterns and may include some elements that were not in fact a part of the true ER network, e.g., endocytic vesicles; this is, however, unlikely at the locations that were analyzed and would not affect materially any conclusions, as further discussed in Results. Measurements were performed at a distance of 50–100 μm from the cell soma and therefore were from unbranched primary dendrites, each arising from a different cell. Duplicate sampling, i.e., two dendrites from the same cell or two measurements from the same dendrite, was not likely. The afferent stimulation used here depolarizes most, but not all, CA3 neurons in a culture and most, but not all, dendrites of a given neuron. Because the aim of the present experiments was to compare Ca2+buffering in various dendritic compartments after Ca2+ influx, dendrites that did not respond to the stimulus train [that is, dendrites that did not show the characteristic elemental changes known to accompany dendritic depolarization and Ca2+ influx and in which the elemental composition of all compartments was statistically (ANOVA) indistinguishable from controls] were excluded from the data set. Typically, the fraction of dendrites that was excluded was small, ∼10%.
Our sampling strategy aimed to acquire five analyses per dendrite for each compartment from five different stimulated dendrites per culture. In practice, the actual number of analyses per section varied because of an abundance or paucity of organelles or dendrites but in no case was fewer than three compartments or dendrites. The final data set was unbalanced and was treated as such for statistical analysis. Nested ANOVA indicated that in all cases there were no differences between slice cultures from the same experimental group; therefore, data from individual cultures (control, n = 4; 1 sec,n = 3; 30 sec, n = 5; 180 sec,n = 3) were pooled to calculate group means and medians. For all elements except Ca, statistical differences among experimental groups were determined using one-way ANOVA, followed by the Dunnett's test for multiple comparisons (vs controls). Because much of the critical data for Ca was not distributed normally, as determined by the Kolmogorov–Smirnov normality test, Ca concentrations are given as both means and medians in Table1. Similarly, and for the sake of clarity, convenience, and consistency, Ca concentrations are presented throughout the text as both means and medians. However, Ca data in all figures are medians. Statistical differences between Ca concentrations in various experimental groups were determined using the Kruskal–Wallis nonparametric ANOVA with Dunn's test for multiple comparisons of unbalanced data. Analysis was performed by means of InStat software (GraphPad Software, San Diego, CA).
Elemental concentrations in dendritic compartments of hippocampal CA3 pyramidal neurons after synaptic activity
RESULTS
Afferent synaptic stimulation evokes calcium uptake by dendritic mitochondria
CA3 pyramidal neurons in hippocampal slice cultures respond to high-frequency stimulation of afferent mossy fibers with micromolar elevations in dendritic [Ca2+]i that reach a maximum in ∼1 sec and decay to baseline in <20 sec (Pozzo-Miller et al., 1993; Petrozzino et al., 1995). To investigate correlated changes in the concentration and distribution of intracellular total Ca, we rapidly froze slice cultures at 1 or 30 sec after a single 1 sec/50 Hz tetanus. This afferent stimulation, identical to the “single train” stimulus used in previous studies (Pozzo-Miller et al., 1997), depolarizes most of the dendrites on most of the CA3 neurons in slice culture (for acute slice experiments, see Jaffe et al., 1992; Magee et al., 1995; Spruston et al., 1995; Johnston et al., 1996; Helmchen, 1999). Subsequently, concentrations of total Ca, as well as Na, Mg, P, Cl, and K within the three principal subcellular compartments of CA3 proximal apical dendrites, namely, mitochondria, ER, and cytoplasm (Fig. 1), were measured by EPMA of freeze-dried cryosections prepared from frozen slice cultures. Because the aim of the present experiments was to characterize the Ca2+ buffering behavior of the various dendritic compartments after Ca2+ influx, measurements were performed only in those dendrites that responded to the stimulus train as indicated by the characteristic elemental changes known to accompany depolarization-induced Ca2+ influx.
Ultrastructural organization of proximal dendrites in CA3 hippocampal neurons. Shown are representative digital transmission electron micrographs of freeze-dried cryosections prepared from rapidly frozen hippocampal slice cultures, illustrating the subcellular structure of proximal dendritic compartments and demonstrating that the three compartments targeted for EPMA (ER, mitochondria, and cytoplasm) are readily identifiable. Images were recorded using standard low-dose techniques in a LEO 912 Omega cryoanalytical electron microscope at −170°C by means of a ProScan slow-scan CCD camera (1024 × 1024). Survey view (A) shows the general appearance of the proximal dendritic field of CA3 neuropil in unstained cryosections from slice cultures. Three primary dendrites (asterisks) course across the field; cytoplasm, characterized by longitudinal bundles of microtubules, and elongated mitochondria (adjacent toasterisks) are evident. At higher magnification (B) both mitochondria and ER [essentially the only intracellular organelles present in proximal apical dendrites at the distance that was sampled (50–100 μm from the soma)] are apparent. The latter is seen as a network of smooth-surfaced cisterns, often in clusters (arrowheads). Together, the ER, mitochondria, and cytosol comprise essentially all of the dendritic volume while also including all important components of the Ca2+ regulatory system. Scale bars, 1 μm.
The results revealed a rapid, graded, and reversible elevation in the total Ca concentration within dendritic mitochondria ([Ca]mito). Afferent tetanic stimulation induced a significant elevation of [Ca]mito at both 1 sec [to 7.9 mmol/kg dry weight (mean); 2.4 mmol/kg (median)] and 30 sec [to 17.4 mmol/kg (mean); 5.8 mmol/kg (median)] after synaptic activation, compared with the characteristically low [Ca]mito [1.1 mmol/kg (mean); 1.2 mmol/kg (median)] in dendrites of control cultures that were grown, mounted, and frozen identically but received only low-frequency test stimuli (Table 1; Fig. 2A,right panel). (Because measurements of Ca concentrations were not distributed normally under conditions of Ca2+ uptake, as discussed below, Ca concentrations are given throughout the text as both means and medians; see Materials and Methods for details.) As an indication of how robust mitochondrial Ca accumulation could be, individual local measurements of [Ca]mito ranged as high as 60.0 mmol/kg at 1 sec and 220 mmol/kg at 30 sec. To place these observations in a biochemical perspective, the mean concentrations given above are equivalent (assuming that mitochondria are 75% protein by weight) to increases from resting levels of ∼1.5 nmol/mg protein to ∼10.5 and ∼23.2 nmol/mg protein, respectively. This elevation of mitochondrial Ca is transient, as indicated by the finding, consistent with previous observations (Pozzo-Miller et al., 1997) that under identical conditions [Ca]mito had recovered completely 3 min after stimulation.
Calcium accumulation in dendritic compartments of CA3 hippocampal neurons. Comparison of calcium concentrations at different time points after synaptic stimulation in three dendritic compartments. In certain instances data were not distributed normally; therefore, all results are given as medians. In A, data are in primary units of mmol/kg dry weight. In B, Ca concentrations have been converted to mmol/kg wet weight (as described in Materials and Methods) to facilitate quantitative comparisons of Ca concentrations between compartments. Uncertainties introduced by this conversion are discussed in Results. Statistical significance (p values) relative to corresponding resting Ca concentrations is given above the bars; other significant differences are indicated by bracketed pairs. Statistical tests are described in Materials and Methods. A, At 1 sec after afferent synaptic stimulation [Ca]mito and [Ca]cyto, but not [Ca]ER, were elevated compared with resting levels. By 30 sec after stimulation Ca concentrations were elevated in all three compartments. (For ER, only data from the subpopulation that actively sequesters Ca in responsive dendrites, as presented in Table 1and discussed in Results, are included.) By 180 sec after stimulation only [Ca]mito had recovered. Note the differenty-axes. B, At rest, ER Ca levels are significantly higher than those in mitochondria and cytosol (compareblack bars). The ER also contains by far the highest Ca concentrations at 30 and 180 sec (compare light gray andwhite bars), but not at 1 sec (dark gray bars).
Among other intracellular elements, only sodium was elevated dramatically as a result of the afferent tetanus. Total Na elevations occurred not just in mitochondria but in all three dendritic compartments and were especially large at 30 sec (Table 1). Comparable changes previously were observed in CA3 dendrites in the same preparation after strong stimulation with four successive tetani and were attributed to the expected enhanced activity of Na+ channels and Na+/Ca2+exchangers (Pozzo-Miller et al., 1997). Stimulus-induced Cl increases also were detected; as expected, these were smaller and appeared to progress sequentially from cytosol to mitochondria, consistent with counter-ion flow through the cell. Like Ca, the Na and Cl elevations were completely reversible, returning to prestimulus levels in all compartments by 3 min after the tetanus (Table 1).
Calcium sequestration in dendritic ER is slower than in mitochondria but persists longer
The ER of responsive dendrites (as described in the previous section) also exhibited elevated levels of total Ca ([Ca]ER) at 30 sec after stimulation, but, in contrast to mitochondria, not at 1 sec (Table 1; Fig.2A, middle panel). [Ca]ER was elevated only in a responsive subset of ER cisterns, defined as those in which [Ca]ER exceeded 15 mmol/kg dry weight. This population, ∼40% of analyzed cisterns with a mean [Ca]ER of 41.3 mmol/kg dry weight (median, 30.7 mmol/kg), was qualitatively and quantitatively similar to the subset of Ca-sequestering ER previously characterized in these slice cultures (Pozzo-Miller et al., 1997). The remaining ∼60% of smooth membrane elements within the same responsive dendrites evidently do not accumulate Ca, because they belong to a normally distributed population with a mean [Ca]ER of ∼6 mmol/kg, statistically indistinguishable from control ER (Table 1). These cisterns are very likely also elements of the true endoplasmic reticulum, because other smooth-membraned cisterns or vesicles are quite rare at the sampled level of primary hippocampal dendrites. We note that if the nonaccumulating ER pool were contaminated adventitiously by non-ER elements, this would have the effect of deflating the estimate of the responsive ER fraction but would not affect quantitative comparisons of Ca accumulation between organelles.
At 3 min after stimulation [Ca]ER was still elevated (Table 1; Fig. 2A, middle panel), indicating that Ca2+uptake by the ER continued, or was at least sustained (depending on the magnitude and rate of Ca2+ release from ER), for several minutes beyond the termination of Ca2+ entry. The ratio of responsive/nonresponsive cisterns essentially was unchanged, arguing against late recruitment to the responsive pool.
The fractional water and dry mass contents of mitochondria, ER, and cytoplasm differ substantially. Consequently, to compare the distribution of Ca between compartments, it is necessary to convert EPMA-derived Ca concentrations to hydration-compensated units such as mmol/kg wet weight (which is also approximately equivalent to mmol/l hydrated cell volume). Median data in Figure 2B have been converted as described in Materials and Methods. Although this conversion inevitably introduces additional uncertainties attributable, for example, to errors in the estimation of dry mass fractions for each compartment, it serves to factor out the effects of compartmental mass differences so as to convey a realistic sense of the quantitative differences in the amounts of sequestered Ca in different organelles. [Ca]ER in control slices was much higher on a wet-weight basis than [Ca]mito or cytoplasmic total Ca ([Ca]cyto) (Fig.2B), indicating that under basal conditions the ER is the most powerful Ca2+ buffer. This observation is also consistent with a role for dendritic ER as a releasable Ca2+ store (Seymour-Laurent and Barish, 1995; Pozzo-Miller et al., 1996; Garaschuk et al., 1997;Emptage et al., 1999; Nakamura et al., 1999; Rae et al., 2000). At 30 sec after tetanic stimulation, when both ER and mitochondria are accumulating Ca avidly, [Ca]ER is substantially larger than [Ca]mito (Fig.2B). The difference is even greater at 3 min after stimulation, at which time [Ca]ER is still elevated significantly, but [Ca]mito has recovered completely. [Ca]ER was not significantly different from the control at 1 sec post-tetanus, indicating a delay of at least 1 sec before net ER Ca2+ uptake begins.
[Ca]cyto in CA3 dendrites was elevated maximally within 1 sec of stimulation (Table 1; Fig.2A, left panel). Although the absolute increase in the cytoplasmic Ca concentration is modest, ∼5 mmol/kg dry weight, and much less than in ER or mitochondria, the cytoplasmic volume fraction is correspondingly larger than that of either organelle. Dendritic cytoplasm thus represents a significant Ca2+-binding reservoir, approximately equivalent by amount to organelle sequestration (see also Pozzo-Miller et al., 2000).
Spatial heterogeneity of synaptically evoked mitochondrial Ca uptake
The values of [Ca]mito presented in Table1 and Figure 2 are means or medians of measurements obtained with randomly placed probes (diameter, ∼100 nm) in each mitochondria in a given dendrite and therefore represent spatially averaged concentrations. (A similar approach with ∼50 nm diameter probes was used for measurements of [Ca]ER.) This sampling strategy provides a reasonable estimate of spatially averaged [Ca]mito at the level of individual dendrites, although sequestered Ca is not distributed uniformly within, and perhaps among, individual mitochondria. Under resting conditions individual measurements of [Ca]mito were always low; as expected for brain mitochondria (Somlyo et al., 1985; Andrews et al., 1988). The distribution of these measurements is normal (Fig.3A, top), with a tight dispersion consistent with that expected from statistical errors of EPMA analysis. After afferent fiber stimulation this distribution became skewed toward higher [Ca]mito by the emergence of a fraction of measurements with elevated [Ca]mito (Fig. 3A,middle, bottom). The most extreme members of this subset had exceedingly high [Ca]mito, often >100 mmol/kg dry weight.
Heterogeneous distribution of [Ca]mito within individual mitochondria after synaptic stimulation. A, Mean [Ca]mito in resting dendrites is low, and the distribution of individual measurements is normal (top panel), as indicated by a good fit (R2 = 0.99) to a single Gaussian with a width (ς = 2.5 mmol/kg dry weight) expected because of statistical uncertainties in EDX analysis. At 1 sec (middle panel) and 30 sec (bottom panel) the distribution becomes progressively skewed toward higher [Ca]mito as a result of a sequestration mechanism that concentrates Ca in small inclusions within the mitochondrial matrix.B, Distribution of single [Ca]mitomeasurements from six representative mitochondria with elevated Ca levels that were analyzed at three separate locations using a 100 nm probe. Large differences in [Ca]mito within each of these mitochondria reflect the heterogeneous distribution of calcium within individual mitochondria.
The skewed distribution of [Ca]mito after stimulation arises at least partly from spatial heterogeneity of accumulated Ca within individual mitochondria, as reflected in the widely dispersed values for single measurements within the same mitochondrion (Fig. 3B). The basis for this form of spatial heterogeneity lies in the chemical nature of mitochondrial Ca sequestration. We and others have reported evidence for the formation, in intact stimulated neurons, of discrete calcium- and phosphorus-rich complexes that are thought to reflect a reversible, high-capacity Ca storage mechanism (David, 1999; Pivovarova et al., 1999). In frog sympathetic neurons depolarized with high K+ such complexes appear as one or more small (∼10 nm) punctate inclusions within the mitochondrial matrix. A similar mechanism of mitochondrial Ca sequestration in CA3 dendrites would account for the frequency distributions of [Ca]mito observed here (Fig. 3A), in that some fraction of Ca-accumulating mitochondria nonetheless are expected to give low individual measurements of [Ca]mito because the local sites of Ca sequestration were missed by randomly placed probes. An additional level of heterogeneity, caused by variability among mitochondria, is also plausible and perhaps likely (see Discussion).
DISCUSSION
Mitochondria in a wide variety of cell types, including central pyramidal neurons (White and Reynolds, 1995; Wang and Thayer, 1996), accumulate Ca2+ after micromolar elevations in [Ca2+]i (Pozzan and Rizzuto, 2000), but a significant role for dendritic mitochondrial Ca2+ uptake after synaptic activity-induced Ca2+ elevations has not been demonstrated. Most available data on mitochondrial Ca2+ transport in neurons come from somatic measurements and are therefore of uncertain relevance to dendritic Ca2+ signaling. The present experiments show that mitochondria in proximal dendrites of CA3 pyramidal neurons rapidly and transiently accumulate large amounts of Ca in the seconds after afferent fiber stimulation. Presumably, rapid Ca2+ uptake occurs because dendritic [Ca2+]i reaches the supramicromolar levels (Petrozzino et al., 1995) consistent with strong activation of the mitochondrial uniporter (for isolated cardiac and liver mitochondria, EC50 ∼10–20 μm;Gunter and Pfeiffer, 1990) and possibly the mitochondrial rapid uptake mode (RaM; Gunter et al., 2000). The Ca2+ uptake rate for dendritic mitochondria over the first second after synaptic stimulation is 9.0 ± 2.9 nmol/mg protein/sec (estimated from data in Table 1 as the difference between [Ca]mito at 1 sec and at rest). This value compares reasonably with the maximal velocity of the uniporter in isolated mitochondria (10–30 nmol/mg protein/sec; Gunter and Pfeiffer, 1990) as well as with Ca2+uptake rates estimated in mitochondria of chromaffin cells (Babcock et al., 1997) and sympathetic neurons (Pivovarova et al., 1999). The large amounts of mitochondrially sequestered Ca found here are also consistent with data from physiologically stimulated nonexcitable cells (Montero et al., 2000).
Dendritic mitochondria still retain a significant Ca load 30 sec after stimulation (Fig. 2), well after [Ca2+]iessentially has returned to prestimulus levels. By this time the net uptake rate has fallen dramatically, which is explained by the decreasing velocity of Ca2+ uptake as [Ca2+]i declines and by the activation of mitochondrial Ca2+ extrusion mechanisms. The synaptically induced rise in [Ca]mito is completely reversible, as indicated by the observation that [Ca]mito has recovered to prestimulus levels 3 min after afferent stimulation (see also Pozzo-Miller et al., 1997). This implies a Ca2+ recovery half-time of 1–2 min, which is compatible with measured Ca2+ extrusion rates in other neuronal mitochondria (Colegrove et al., 2000) and also consistent with Ca2+ extrusion mainly via the Na+/Ca2+exchanger, the activity of which would, in turn, account for the elevated levels of mitochondrial Na+ found here. Last, mitochondrial Ca2+ uptake after a single tetanic stimulus is consistent with similarly evoked changes in the mitochondria membrane potential in pyramidal dendrites (Bindokas et al., 1998), changes that presumably reflect mitochondrial Ca accumulation.
It should be mentioned that the dynamic pattern of elemental changes described here, namely, sequential or parallel elevation and recovery of cytoplasmic and organelle Ca and Na, has been observed in other neuronal systems, for example in the somata of cerebellar granule cells (Kiedrowski et al., 1994) and sympathetic neurons (Pivovarova et al., 1999) after strong K+ depolarization, and appears to be a characteristic, physiologically normal consequence of Ca2+ entry and/or clearance mechanisms. In contrast to injurious stimuli, e.g., anoxia (Taylor et al., 1999) or high concentrations of glutamate (Kiedrowski et al., 1994), which evoke very large elevations of cytoplasmic Ca and Na, tetanus-evoked elevations of these elements are much smaller and completely reversible (Table 1) (Pozzo-Miller et al., 1997) and can be triggered repeatedly and reproducibly. Accumulating evidence suggests that Na+ spikes or transients in pyramidal dendrites may subserve important functions such as triggering Ca2+ entry (Jaffe et al., 1992; Spruston et al., 1995; Johnston et al., 1996) or even acting as an independent signaling pathway (Callaway and Ross, 1997; Rose and Konnerth, 2001). Similarly, large cytosolic Na+ elevations appear to be an important component of transmitter release mechanisms in presynaptic terminals (Zhong et al., 2001).
So far we have evaluated mitochondrial Ca2+ uptake on the basis of averaged measurements of [Ca]mito at random locations within individual mitochondria, but this approach is oversimplified because the actual spatial distribution of accumulated Ca is clearly heterogeneous. One major factor underlying this heterogeneity is the punctate distribution of sequestered Ca within the matrix of individual mitochondria. This distribution is a reflection of a major mechanism of Ca2+ buffering and sequestration that, as described above (Fig. 3), gives a skewed distribution of individual measurements of [Ca]mito. This occurrence obscures higher levels of heterogeneity such as intermitochondrial variations that might reflect variability in mitochondrial energetics (Buckman and Reynolds, 2001) or spatial distribution (Collins et al., 2002) or in local [Ca2+]i levels. In sympathetic neurons (Pivovarova et al., 1999), as well as in other excitable and nonexcitable cells (Montero et al., 2000; Collins et al., 2002), there is evidence for differences in Ca2+ uptake between individual mitochondria responding to the same stimulus in the same cell. These differences appear to depend at least partly on proximity to a local Ca2+ source. For mitochondria in dendrites of pyramidal neurons a similar heterogeneous response would be consistent with the skewed [Ca]mitodistribution, but evidence that this occurs is masked because the intramitochondrial Ca-rich inclusions affect the frequency distributions in the same way as would intermitochondrial heterogeneity. If, in fact, there are differences in [Ca]mito between mitochondria, this well may reflect proximity to Ca2+ channels of the dendritic plasma membrane, a situation for which there is evidence in sympathetic neurons (Pivovarova et al., 1999) and secretory cells (Tse et al., 1997). Spatially resolved [Ca]mitomeasurements, i.e., “calcium maps” (Meldolesi and Grohovaz, 2001), will be necessary to address this issue.
Present results confirm and extend our previous observation that, within dendrites that have responded to synaptic stimulation with an increase in [Ca2+]i, a specific subset of ER has a profound capacity for accumulating and storing calcium. Figure 4 compares the magnitude and time courses of mitochondrial and ER Ca2+ uptake and release. It illustrates the fact that, in both dendritic organelles, the evoked Ca load persists well after decay of the Ca2+transient, but the ER Ca load is particularly long-lived; previously published observations (Pozzo-Miller et al., 1997) indicate that Ca-accumulating ER may sustain their Ca load for longer than 15 min. Another interesting new finding is that ER Ca2+ uptake appears to be delayed for at least 1 sec relative to the cytosolic Ca2+transient and to the onset of mitochondrial Ca2+ uptake. This may reflect the relative Ca2+ dependencies of the uptake mechanisms, i.e., the uniporter and RaM versus sarcoplasmic/endoplasmic reticulum Ca2+-ATPase (SERCA) pumps in the ER. Alternatively, it may indicate release from Ca stores, which is known to be evoked in CA3 pyramidal neurons by mossy fiber activation (Pozzo-Miller et al., 1996; Yeckel et al., 1999). The role of releasable stores in dendritic Ca2+signaling is an area of active investigation. Several recent studies indicate that Ca2+ release from ryanodine-sensitive and/or IP3-sensitive internal stores may be a significant mechanism for signal amplification and/or modulation in both dendrites and spines of hippocampal neurons (Seymour-Laurent and Barish, 1995; Pozzo-Miller et al., 1996; Garaschuk et al., 1997; Emptage et al., 1999; Nakamura et al., 1999, 2000;Yeckel et al., 1999; Rae et al., 2000; Kapur et al., 2001). In the present case, because the ER reloads and even overloads during later stages of net Ca accumulation, dendritic ER may act first as a Ca2+ source to amplify the original Ca2+ signal and later switch into a buffering mode. We have demonstrated recently such a switchover in the caffeine/ryanodine-sensitive Ca stores of sympathetic neurons (Albrecht et al., 2001; Hongpaisan et al., 2001).
Calcium dynamics in dendritic compartments after synaptic stimulation. Time courses for [Ca]cyto(opencircles), [Ca]mito(filled circles), and [Ca]ER(filled triangles; responsive subset only at 30 and 180 sec) illustrate complementary temporal relationship for activity-dependent Ca2+ uptake and release by dendritic mitochondria and ER. These organelles appear to transport Ca2+ within distinct, possibly cooperative, time domains. For example, [Ca]mito increases earlier and declines faster than [Ca]ER, which remained elevated for >10 min. Data are given as medians. Note the log scale on the abscissa.
Much recent work supports the idea that interactions between the Ca2+ transport mechanisms of ER and mitochondria determine how these organelles cooperate or compete to modulate cytoplasmic Ca2+ signals (Hajnoczky et al., 2000; Rizzuto et al., 2000). Transfer of Ca2+ from the ER to the mitochondrion is the general (but not universal; Hoth et al., 1997; Hongpaisan et al., 2001) direction of transport. As noted previously (Pozzo-Miller et al., 2000), comparison of the time courses of Ca2+ transport by dendritic mitochondria and ER shows that these organelles accumulate and release Ca2+ in distinct and complementary time domains (Fig. 4), raising the possibility of a cooperative relationship. Mutual interdependence also is suggested by experiments showing that thapsigargin inhibition of ER Ca2+ uptake greatly enhances mitochondrial Ca2+ uptake (Pozzo-Miller et al., 1997). It is not clear whether the interactions underlying ER/mitochondria coupling are direct, as proposed for nonexcitable cells (Rizzuto et al., 1998), or mediated via changes in global [Ca2+]i. In either case any Ca “shuttling” at longer times in CA3 dendrites must be from mitochondria to ER, because dendritic ER retains its Ca load much longer than do mitochondria. Although the details of such interactions remain to be fully understood, it is clear that the ER and mitochondria of pyramidal dendrites play a fundamental role in shaping dendritic calcium signals during synaptically driven neuronal activity.
Footnotes
↵* N.B.P. and L.D.P.-M. contributed equally to this work.
This work was supported by the National Institutes of Health Intramural Research Program. We thank Drs. C. A. Brantner, R. D. Leapman, and T. S. Reese for resource support, helpful discussions, and technical assistance.
Correspondence should be addressed to Brian Andrews, Building 36, Room 2A-21, Laboratory of Neurobiology, National Institute of Neurological Disorders and Stroke, National Institutes of Health, 36 Convent Drive, Bethesda, MD 20892-4062. E-mail: sba{at}helix.nih.gov.
