We have previously shown that disabling forward-mode Na+-Ca2+ exchange in amacrine cells greatly prolongs the depolarization-induced release of transmitter. To investigate the mechanism for this, we imaged [Ca2+]i in segments of dendrites during depolarization. Removal of [Na+]o produced no immediate effect on resting [Ca2+]i but did prolong [Ca2+]i transients induced by brief depolarization in both voltage-clamped and unclamped cells. In some cells, depolarization gave rise to stable patterns of higher and lower [Ca2+] over micrometer-length scales that collapsed once [Na+]o was restored. Prolongation of [Ca2+]i transients by removal of [Na+]o is not due to reverse mode operation of Na+-Ca2+ exchange but is instead a consequence of Ca2+ release from endoplasmic reticulum (ER) stores over which Na+-Ca2+exchange normally exercises control. Even in normal [Na+]o, hotspots for [Ca2+] could be seen following depolarization, that are attributable to local Ca2+-induced Ca2+ release. Hotspots were seen to be labile, probably reflecting the state of local stores or their Ca2+ release channels. When ER stores were emptied of Ca2+ by thapsigargin, [Ca2+] transients in dendrites were greatly reduced and unaffected by the removal of [Na+]o implying that even when Na+-Ca2+ exchange is working normally, the majority of the [Ca2+]i increase by depolarization is due to internal release rather than influx across the plasma membrane. Na+-Ca2+ exchange has an important role in controlling [Ca2+] dynamics in amacrine cell dendrites chiefly by moderating the positive feedback of the Ca2+ amplifier.
Calcium entering neurons can be temporarily sequestered within Ca2+ stores or chelated by the many native Ca2+ buffer molecules. Eventually however, almost all excess Ca2+ is expelled via the plasmalemmal Ca2+ ATPase (PMCA) or the plasmalemmal Na+-Ca2+ exchanger. The relative importance of these mechanisms and their interactions are not well understood, particularly at synapses where Ca2+ plays several roles, including the triggering of transmitter release. To the extent that it has been looked at, it appears that the relative importance of different mechanisms of Ca2+ clearance varies from preparation to preparation. In the retina, the ribbon synapses at the photoreceptor and bipolar cell terminals are thought to chiefly use the PMCA (Krizaj and Copenhagen 1998; Morgans et al. 1998; Zenisek and Matthews 2000). In cultured retinal amacrine cells, however, the plasmalemmal Na+-Ca2+ exchanger seems to play a dominant role in clearing Ca2+ loads because preventing the normal operation of this exchanger greatly prolongs transmission following Ca2+ influx through voltage-gated channels (Gleason et al. 1994). Similar results have been found with cultured hippocampal neurons (Reuter and Porzig 1995).
Many types of retinal neuron, including many types of amacrine cell, signal with graded potentials and release transmitter asynchronously without the need for action potentials (Bieda and Copenhagen 1999; Gleason et al. 1993). An important characteristic of asynchronous transmission is that, unlike fast synchronous transmission, it requires only a modest increase in cytoplasmic [Ca2+]. Retinal amacrine cells from the chick release transmitter even at resting [Ca2+]i(Frerking et al. 1997), and in both retinal photoreceptors (Rieke and Schwartz 1996) and bipolar cells (Lagnado et al. 1996; Rouze and Schwartz 1998) continuous transmitter release requires only low-micromolar [Ca2+]. Because of this, the mechanism of Ca2+ clearance is likely to be of particular significance at these synapses.
The mechanisms controlling asynchronous transmission are not well known, but there is some evidence suggesting that internal Ca2+ stores in the terminal boutons of hippocampal pyramidal neurons play a significant role in providing the Ca2+ required for this form of transmission (Emptage et al. 2001). A similar result has been found in cultured rat ganglion cells where the continuous release of quanta is made possible by Ca2+ release from the endoplasmic reticulum (ER) through IP3Rs (Han et al. 2001). Interestingly, there is some evidence in neurons and other preparations, suggesting a close association between the Na+-Ca2+ exchanger and internal Ca2+ stores. This association is well established in the case of smooth muscle where the Na+-Ca2+exchanger has an intimate and perhaps molecular linkage to the superficial sarcoplasmic reticulum such that Ca2+ release from the SR through IP3Rs or ryanodine receptors (RyRs) is closely coupled to its removal to the extracellular space by the exchanger (Nazer and van Breemen 1998). In neurons, there is evidence for a spatial association of the exchanger with ER stores (Blaustein et al. 1996), but the functional significance of this is largely unknown. Among retinal neurons, catfish horizontal cells have been examined with respect to the interaction between the exchanger and the caffeine-sensitive Ca2+ store. In this preparation, reverse mode operation of the exchanger refills the ER once it is depleted of Ca2+ (Micci and Christensen 1998).
In the work described here, we investigate the relationship between Na+-Ca2+ exchange and Ca2+ release from internal stores with the aim of understanding how Na+-Ca2+ exchange might influence synaptic transmission from amacrine cell dendrites. Unexpectedly, we find that the increase in [Ca2+] produced by depolarization is chiefly due to Ca2+ released internally, triggered by a relatively small influx through voltage-gated Ca2+channels. The role of Na+-Ca2+ exchange appears to be that of moderating the gain of this Ca2+ amplifier mechanism.
Chick retinae were dissociated on embryonic day 8 and cultured on coverslips at low density as previously described (Gleason et al. 1993). Solitary, isolated neurons identified as amacrine cells were used after 7–12 days in culture.
Cells were loaded for ∼1 h at room temperature with 10 μM Fura-2 AM in normal external solution (see Solutions) with 0.02% wt/vol pluronic. A second set of experiments employed confocal imaging, chiefly in linescan mode, and for these, the AM salt of Oregon Green Bapta-1 (OGB-1, Molecular Probes, K d= 170 nM) was loaded at 5 μM in a similar manner. As described in the following text, a few additional experiments were conducted in which the dextran-conjugated (10,000 MW) form of Fura-2 (Molecular Probes) was loaded from a patch pipette.
Fura-2-AM-loaded cells were voltage clamped using the nystatin perforated-patch technique previously described (Gleason et al. 1995). Voltage-clamped cells were stimulated with 500 ms depolarizations from −60 or −70 to 0 mV, which is sufficient to activate the Ca2+ currents in these cells close to maximally (Gleason et al. 1995). Unclamped cells were stimulated with a 2-s puff of 100 mM K+ external solution, applied from a pipette with a tip diameter of ∼1 μm positioned over the cell body. For both the confocal and the video imaging experiments, coverslips with cells were mounted in a 40-μL Plexiglas chamber (model RC-24, Warner Instruments, Hampden, CT) in which gravity fed solution changes could be made within ∼4 s. In puff experiments, leakage from the pipette tip was unavoidable; so to minimize its effects, the puff pipette was parked a long way from the cell and brought up to the cell just before the 2- to 6-kPa pressure step and withdrawn immediately after its application. Pipette movement was automated by programming two positions into a modified Eppendorf 7171 micromanipulator (Eppendorf, Westbury, NY) and triggering movement with a TTL pulse. In some experiments such as that shown in Fig.2 B, two solutions had to be applied rapidly one after the other. In those experiments, two manipulators were employed, each carrying its own puff pipette.
Pipette solution was initially made up as follows (in mM) 10 CsCl, 150 CsAc, 3 NaCl, 2 MgCl2, 0.1 CaCl2, 10 HEPES, 1.1 EGTA, and 5 TEAAc. pH was adjusted to 7.4 with CsOH. Empirically we found that an osmolarity of 230 mosM, ∼20% hypotonic to the external solution, produced the longest recordings from neurons without any swelling or shrinking. External solutions used in patch-clamp and unclamped experiments were as follows (in mM):Normal: 5.3 KCl, 116.9 NaCl, 20 TEACl, 3 CaCl2, 0.41 MgCl2, 3 HEPES, 5.6 glucose. 0 Ca: 5.3 KCl, 116.9 NaCl, 20 TEACl, 3.41 MgCl2, 3 HEPES, 5.6 glucose.High K: 100 KCl, 22.2 NaCl, 20 TEACl, 3 CaCl2, 0.41 MgCl2, 3 HEPES, 5.6 glucose. 0 Na: 5.3 KCl, 116.9 LiCl, 20 TEACl, 3 CaCl2, 0.41 MgCl2, 3 HEPES, 5.6 glucose. Mn: 1 mM in high-K solution.
All external solutions had 300 nM TTX, and were adjusted for pH to 7.4 with NaOH or LiOH (for 0 Na+ solution). In some experiments, bicuculline methiodide (3 μM) was added to suppress autaptic currents. Drugs were used at the following concentrations: thapsigargin, 1–1.6 μM (Calbiochem); sodium orthovanadate, 1 mM (Sigma); FCCP, 1 μM (Sigma); ionomycin, 10 μM (Calbiochem); thimerosal; 10 μM (Sigma); Digitonin, 20 μM (in 0 Ca).
For video imaging of Fura-2-loaded cells, mounted coverslips were placed on an inverted Nikon Diaphot microscope and visualized with an oil-immersion objective (1.3 NA, ×100). UV excitation was generated with a short arc Xenon lamp (XBO 150W/CR, Osram), filtered through 40-nm band pass UV interference filters (peak, 340 and 380 nm, Chroma Technology) and introduced into the microscope by a liquid light guide. Switching between the two wavelengths was achieved rapidly (<2 ms) by means of a galvanometer mirror (DX-1000, Solamere Technology Group, Salt Lake City, UT). The intensity of excitation light at the two wavelengths could be set separately. This was a useful feature when imaging thin dendrites, as it was often necessary to adjust the excitation intensities to obtain acceptable signal-to-noise ratios without leaving the linear range of the imaging setup. As a consequence, fluorescence ratios had to be corrected for excitation intensities to be consistent with the settings used for calibration. To measure the relative intensities at the two excitation wavelengths, a photodiode (UV-360BQ, EG&G Canada) was introduced into the excitation path following every experiment.
Fluorescence images (>510 nm) were projected to a GenIII intensified CCD camera (Stanford Photonics) and digitized to 12 bits at video rate using the Axon Imaging Workbench 2.0 and 4.0 software and Image Lightning 2000 Frame Grabber (Axon Instruments, Union City, CA). In most experiments, ratio imaging used frame averaging at alternating wavelengths to give a rate of one ratio image every 0.429 s. Because the volume of a dendrite is much smaller than the volume of a cell body, the light scattered from a cell body could swamp the signal in a nearby dendrite. To avoid this, the field of the excitation light was restricted to a circle of 32 μm diam by placing a pinhole on the end of the light guide, conjugate with the object plane.
Confocal experiments were run on an Olympus FLUOVIEW confocal microscope using the 488-nm line of an Argon Laser. Line scanning utilized a mode whereby the user could draw a single line on the image of a dendrite that could then be repeatedly scanned. Control experiments confirmed that there was no movement of the line relative to the dendrite and the deviation of the drawn line from the scanned line was nowhere >3 pixel widths (0.17 μm).
Calculation of free [Ca2+] and correction of imaging artifacts
The steps in the calculation of free [Ca2+] from ratio images were performed in Matlab using in-house software summarized here. Analysis of images was restricted to the region within the boundaries of the dendrite as drawn by hand but guided by the Laplacian transform of the average of 40 consecutive phase images, saved prior to the fluorescence imaging. An exact registration between the phase and fluorescence images assured us that the dendrite had not moved during the experiment. The processing steps outlined below result in a time series of ratio images. From this, [Ca2+] was calculated pixel by pixel using the following equation (Grynkiewicz et al. 1985; Poenie 1990) Where R is the ratio between corrected images at the two wavelengths, K eff is the effective dissociation constant and K D is the dissociation constant of the dye, I F and I B are the background corrected fluorescences of the bound and free forms of Fura-2 with 380 nm excitation. The parametersR min, R max, andI B/I F were obtained from in vitro calibrations.
In calculating [Ca2+], several corrections were applied to raw images. Raw images were first smoothed with a Gaussian having a 2 pixel SD before application of a shading correction (also smoothed). An adjustment was applied to correct a small amount of bleeding between the two channels due to imperfect synchrony between the switching of excitation wavelengths and frame initiation in the camera. Backgrounds, chiefly due to thermal noise in the intensified CCD, were determined for every cell and subtracted from raw images. Autofluorescence in dendrites was found to be very small, 5–10% of the total background, and was not routinely measured.
We performed in vitro calibrations on a routine basis. Microcapillaries (100-μm thick) were filled with solutions containing 50 μM Fura-2 and either 5 mM CaCl2 (Max [Ca2+] solution) or 0 [Ca2+] 5 mM EGTA (Min [Ca2+] solution), pH 7.2. From background subtracted ratios we obtainedR max, R min, andS b/S f. ForK D, we used a typical value reported in the literature, K D = 252 nM, and we applied a viscosity correction factor of 0.7 to R max andR min (Poenie 1990).
Are Ca gradients real?
One of the conclusions of the work described here is that local differences in [Ca2+] may exist within a dendrite over distances on the order of 1 μm. Numerous errors and artifacts are associated with the use of [Ca2+] indicator dyes, especially those loaded as membrane-permeant AM esters, and we briefly consider whether these might have artifactually generated the observed gradients.
The use of a high-affinity dye, because it acts to buffer Ca2+, inevitably distorts the kinetics of [Ca2+] transients in three significant ways. [Ca2+] is expected to rise and fall more slowly, reach lower peak values, and diffuse more quickly than normal. This last effect follows from the likely fact that native buffers within the cell are mostly immobile, whereas the indicator dye is able to carry Ca2+ from high to low concentration regions. None of these three consequences, however, is expected to accentuate spatial gradients for [Ca2+]; rather they would all tend to lessen gradients, leading us to suppose that gradients might be more common and larger in unperturbed cells than those measured here. Two pieces of evidence stemming directly from our data indicate that [Ca2+] gradients are not the effect of imaging or dye artifacts. The first is that in a given cell, gradients could change sign over time, a hot spot becoming cool and vice versa. Second, resting [Ca2+]i was uniform and gradients appeared only after stimulation (see Fig. 7 A), yet gradients should be present in both situations if the cause was an optical or dye artifact. For instance, if dye was compartmentalized, the effect should be more evident in the resting state, because [Ca2+] in intracellular compartments is high enough to saturate the dye and would contrast sharply with the low resting cytosolic [Ca2+].
Aside from these observations, an experimental test was designed to validate the non-artifactual origin of gradients. The experiment is based on the observation that repeated high-K+ stimuli can, in some cells, induce long-lasting (>10 min) [Ca2+] gradients in unclamped dendrites (data not shown). A similar phenomenon has been reported in pyramidal neurons (Connor et al. 1988). If the gradients are due to Ca2+ release and uptake from intracellular sources and sinks, then disabling the Ca2+ handling mechanisms of the cell should bring the cell to a uniform [Ca2+]i. For this purpose, we used an external solution with 1 mM vanadate (an inhibitor of plasma membrane ATPases), 1 μM thapsigargin (an specific inhibitor of SERCA pumps), 0 Na+ (Li+ replacement) to disable the Na+-Ca2+ exchanger, 1 μM FCCP (a protonophore that collapses the mitochondrial membrane potential), 10 μM ionomycin, and 800 nM [Ca2+]/EGTA. In two cells where we were able to induce gradients, bathing the cell in this external solution collapsed [Ca2+]i gradients.
A second test consisted in loading cells (n = 5) with the high-molecular-mass dextran conjugate of Fura-2 from a pipette in the whole cell configuration. The pipette was then retracted after partial filling of the cell, allowing the plasma membrane to seal and the cell to recover for ∼15 min before imaging. The dextran conjugate of Fura-2 does not cross intracellular membranes neither does it bind to immobile proteins or suffer from deesterification problems. A series of high-K+ puffs was then applied from a pipette near the cell body as described in the preceding text. In one cell, we established a clear gradient that lasted for >10 min and was abolished by 0 Ca2+ external solution (data not shown).
We also performed quantitative tests to assess the importance of the known imaging artifacts including dye compartmentalization in intracellular organelles, incomplete deesterification of the AM form (Oakes et al. 1988), and binding of dye to immobile proteins without loss of fluorescence. We performed the following tests to examine these effects.
To examine compartmentalization, we adapted the method described inKao (1994) in which digitonin and Triton X-100 are applied sequentially. Over the majority of a dendrite, compartmentalization was negligible, and averaged over entire dendrites the percentage of fluorescence compartmentalized was typically 3%, in certain locations; however, it could reach 15%.
As a test for deesterification, we used the Mn2+-quenching method (McCarthy et al. 1994). Cells were loaded with Fura-2 AM as described previously and Ca-independent fluorescence was assessed using the isocoefficient method (Neher 2000;Zhou and Neher 1993). A measurement was done after 1 h AM loading, with the cell bathed in 0 Ca2+external solution, and a second one after quenching by bathing the cell for 5 min in a 0 Ca2+ solution containing 1 mM Mn2+ and high (100 mM) K+. Comparing the fluorescence between the two we found that residual fluorescence after quenching amounted to 11 ± 2% (mean ± SD within a dendrite) of the initial fluorescence. Because this residual value includes not only unquenched Fura-2 but also autofluorescence, and probably compartmentalized dye, the small variability within a dendrite indicates that quenching was uniform. This experiment was repeated in four cells with similar results.
Preventing forward mode exchange prolongs [Ca2+]i recovery after depolarization
Synapses between cultured amacrine cells are mostly dendro-dendritic. For this reason, we have looked closely at Ca2+ handling within dendrites to see if Na+-Ca2+ exchange is present there and if so, how it controls [Ca2+]i. In other neurons, there is evidence that dendrites and cell bodies have significantly different Ca2+-handling mechanisms (Thayer and Miller 1990), consistent with their different functions as well as their different surface to volume ratios.
Isolated, solitary amacrine cells were held at −60 or −70 mV in perforated whole cell patch-clamp and stepped to 0 mV for 500 ms. A change in Fura-2 absorbtion ratio indicated a sharp increase in [Ca2+] in dendrites accompanying an inward current carried by Ca2+, mostly through L-type Ca2+channels (Gleason et al. 1994, 1995). The magnitude of this [Ca2+] increase varied considerably between cells, and even within a single dendrite was often very nonuniform. Typically peak [Ca2+]i values, as illustrated in Fig.1, ranged from 200 nM to 1 μM. Recovery back to baseline occurred with several time constants but was >80% complete within 10 s. In a typical experiment (Fig. 1), normal external solution was rapidly replaced with one in which either Li+ or N-methylglucamine (NMG), neither of which support Na+-Ca2+ exchange, was substituted for Na+ o. These two agents gave indistinguishable results. Transition to Na+-free solution produced no change in resting dendritic [Ca2+]; but when the voltage step was repeated in the Na+-free solution, the peak [Ca2+]i was usually a little higher than in normal solution and the return of [Ca2+] to baseline was always prolonged.
As illustrated in Fig. 1, considerable variability between cells was seen with respect to the kinetics of [Ca2+] recovery. In an initial survey of 46 cells, 50% showed a prolongation of the decline but no change to its monotonic nature. In 33% of cells, [Ca2+]i fell, or in some cases rose, slowly, to reach a plateau value. In these cases, as shown in Fig. 1,B and C, washing with normal solution permitted recovery of dendritic [Ca2+] with a time course similar to that following the original voltage step in normal solution. The remaining 17% of cells showed oscillatory or chaotic behavior of [Ca2+]i that also quickly returned to baseline in normal external solution. A second voltage step in normal solution was always applied to verify that changes in [Ca2+]i kinetics were reversible.
Although we have here classified the observed dendritic [Ca2+] as falling into three kinetic patterns, these are not natural classes and there is, in reality, a full range of intermediate conditions. A commonly observed feature in many cells was the sudden increase in [Ca2+]i by, typically, one to a few hundred nanomolar. Often these events could be seen to occur on the falling phases of [Ca2+]i (Fig.1 E), but sometimes they occurred “spontaneously” without any obvious timelocking to the stimulus (Fig. 1, D andE). Because these cells were voltage-clamped, the origin of these events does not lie with voltage-gated Ca2+ channels.
Is reverse mode exchange causing Ca2+ influx?
In catfish horizontal cells (Micci and Christensen 1998) and other neurons (Hoyt et al. 1998), it has been shown that the Na+-Ca2+ exchanger can effect a rapid Ca2+ influx through reverse mode operation. Because the removal of Na+ o, by reversing the normal Na+ gradient, would favor this mode of exchange, we looked closely to see if this mechanism of Ca2+ influx was contributing to the changes in [Ca2+]idynamics seen in Fig. 1.
Comparison of the dynamics of dendritic [Ca2+] in pairs of records with and without Na+ o (Fig.2 A) showed that the rate of rise in [Ca2+]i was very similar, whether or not Na+ o was present. Were reverse mode exchange generating a Ca2+ influx, the [Ca2+]i rise rate in 0 Na+ o solution might be expected to be greater. A second consistent feature of this comparison was that in the absence of Na+ o, dendritic [Ca2+] was seen to continue to increase, reaching a peak later than in the presence of Na+ o. Following the peak, the decay of [Ca2+]i, as already described, was slower when Na+ o was absent. The longer time to peak in the absence of Na+ o does not immediately suggest reverse mode exchange but would be consistent with reverse mode exchange continuing at a high rate after the closure of Ca2+ channels. If so, then removal of Ca2+ o during this phase ought to accelerate the decline of [Ca2+]i. This possibility was ruled out experimentally. Depolarization in these experiments was achieved by a 2-s-long, high-K+ puff applied to unclamped cells immediately followed by the removal of Ca2+ o by blowing 0 Ca2+ solution over a cell. As shown in Fig. 2 B, no change in the kinetics of [Ca2+]i decline was seen in the trials in which Ca2+ o was removed (time to half decay, 6.88 ± 3.51 s in 0 Na+ o, 7.24 ± 2.73 in 0 Na+ o/0 Ca2+ o, no significant difference,P > 0.4, paired t-test, n = 6 cells. Taken together, these results suggest that reverse mode exchange does not contribute significantly to the kinetics of Ca2+ handling in dendrites.
Disabling forward-mode exchange generates [Ca2+] gradients in dendrites
A striking feature of our experiments was that, in the absence of Na+ o [Ca2+] within dendrites was frequently seen to be inhomogeneous over periods of tens of seconds, following a depolarization. In those cells in which a plateau was reached, long time averaging increased signal-to-noise ratio to the extent that it was possible to resolve stable hotspots and cool spots for Ca2+ with separations of as little as 1 μm. As seen in Fig. 3 A, adjacent hot- and cool spots differed in [Ca2+] by as much as several hundred nanomolar. On return to normal external solution, spatial gradients collapsed and were absent prior to depolarization. In addition to steady-state gradients we frequently observed gradients that changed dynamically when the exchanger was disabled (Fig.3 B). Because extensive temporal averaging is precluded in these cases, spatial resolution is not so good; nevertheless it is possible to see that over a few micrometers [Ca2+] appears to be independently regulated over a time course of seconds or tens of seconds. In some cases, [Ca2+]ioscillations were seen in one part of the dendrite, whereas in other parts of the dendrite, these oscillations were extremely damped, and in some cases, uncorrelated fluctuations were seen.
If the Ca2+ entering a dendrite upon depolarization was able to diffuse freely, either in its free form or else bound to a mobile buffer, [Ca2+]i would quickly become homogeneous within the dendrite. Assuming that Fura-2 with a diffusion coefficient of 100 μm2 s−1 (Gabso et al. 1997; Murthy et al. 2000) is the main buffer in our cells, we calculate that homogeneity of a 10-μm stretch of dendrite would be achieved within 500 ms. The fact that [Ca2+]i gradients are maintained for considerably longer periods than is compatible with diffusive mixing implies that there must be Ca2+ sources within a dendrite that continue to release Ca2+ after the closure of Ca2+ channels. Additionally it implies that there must also be Ca2+ sinks operating within a dendrite. Together, sources and sinks are able to set up the stable or dynamic gradients seen in Fig. 3. The sources within dendrites are linked to the ER, as shown in the following text, but the nature of the sinks is unclear and might be intradendritic organelles or the PMCA or a combination of these. The one process that cannot be part of the sink mechanism though is the plasmalemmal Na+-Ca2+ exchanger.
Ca2+ is released from internal stores
A diagnostic criterion for Ca2+ induced Ca2+ release (CICR) from ER is that it should be abolished once ER stores are empty of Ca2+. Thapsigargin (tg) is a membrane-permeant, specific inhibitor of the SERCA pump that refills the ER with Ca2+ from the cytoplasm. Sufficiently long incubation in tg should deplete the ER of Ca2+ and reduce or eliminate the effect on the [Ca2+]itransient of removing Na+ from the external solution.
To determine the duration of treatment with tg that would be sure to empty the ER stores completely, we tried different incubation times before challenging cells with 10 μM thimerosal. Thimerosal is a non-specific sulfhydryl reagent that promotes Ca2+ efflux from the ER through both RyRs and InsP3 receptors (Elferink 1999). Because it also has effects on plasma membrane channels, we performed the challenge in the absence of Ca2+ o. Based on preliminary experiments with 12 cells exposed to tg for various times, we found that 1 h of tg was sufficient to deplete the dendritic ER totally, although interestingly cell bodies retained some releasable Ca2+ even after this treatment (Fig. 4).
Using linescan imaging of dendrites loaded with OGB-1, we replicated qualitatively the results of Fig. 1 using a 2-s K+depolarization of unclamped cells. As expected, Na+ removal prolonged the Ca2+ transient, measured as time to half decay (P < 0.031, n = 5, Wilcoxon signed-rank sum test). On average the increase int 1/2 was 80% (range 28–207%) in 0 Na+ versus normal (Fig. 5). After treatment with tg, results were different in two ways. First, Na+ removal no longer extended the duration of the transients (P > 0.3, n = 5, Wilcoxon signed-rank sum test, average change int 1/2 = 5 ± 10%). Second, the [Ca2+]i transient, even in normal external solution, was shorter and smaller than in untreated dendrites (P < 0.025, Mann-Whitney test, n = 5). The difference in fluorescence change between tg-pretreated cells and controls was not due to dye saturation or nonlinear behavior due to an increase in basal [Ca2+]i with tg. From Fura-2 imaging, we found that tg not only failed to increase resting [Ca2+]i, when measured after 1 h of treatment, but caused a slight but significant decrease compared to control cells (35 ± 7 vs. 71 ± 30 nM in controls;P < 0.005; n = 6 Mann-Whitney sum rank test).
[Ca2+]i gradients exist when exchange is enabled
Because the [Ca2+]i transient in normal solution is much smaller following tg pretreatment, Ca2+release from the ER is not a phenomenon solely associated with the disabling of the exchanger but occurs in normal conditions as well. It is reasonable then to ask to what extent the local [Ca2+]i gradients seen when exchange is disabled (Fig. 3), are also found in normal conditions. Fura-2 imaging indicates, as shown in Fig. 3, that at rest there are generally no detectable [Ca2+]i gradients within dendrites. Following depolarization, however, transient local Ca2+ gradients were often seen in both the Fura-2 imaging and the OGB-1 linescan experiments. In Fig. 5 (1st trial in normal solution, untreated cell) depolarization causes a substantial increase in fluorescence that is non-uniformly distributed along the dendrite. One region (red spot) shows an unusually prolonged [Ca2+]i increase lasting tens of seconds. This region of the dendrite is interesting because it has sharp spatial boundaries on either side, implying active sources of Ca2+within the region and active sinks for Ca2+ at its margin. Toward the end of this first trial, another region of the dendrite (blue spot) shows a transient increase in [Ca2+]i not seen elsewhere in the dendrite. In a second trial, in the absence of Na+ o, [Ca2+]i recovery is, as expected, greatly slowed down. In the third trial, back in normal external solution, neither the region marked with the red spot nor the region marked with the blue spot show higher [Ca2+]i than the rest of the dendrite. The implication of these observations is that not only are [Ca2+]i gradients normally present after depolarization but also the sources and sinks that give rise to them are labile and behave in a history-dependent manner. This general conclusion was supported by experiments (not shown) in which [Ca2+]i hotspots identified in linescans of dendrites during depolarization were compared with the [Ca2+]i profile seen when thimerosal was applied in Ca2+-free medium so as to release Ca2+ from all ER stores. Many of the hotspots identified in depolarization were present in thimerosal-evoked release but some were absent, perhaps implying that stores at these sites were empty.
As shown in Fig. 1, brief spontaneous increases in [Ca2+]i could often be seen in normal external solution. The same observation was made using linescan of dendrites in which the spatial and temporal resolution was increased. Spontaneous events usually had a clear locus of origin from which Ca2+ spread in both directions (Fig.6). Generally the total extent of an event was <10 μm, and the time course was ∼1 s, although we frequently saw events originating at the same site following so closely after each other as to be difficult to separate. The similarity of these [Ca2+]i events seen here in amacrine cells to elementary [Ca2+]i events seen in cultured hippocampal neurons (Koizumi et al. 1999) leads us to suppose that these may be the basic units from the CICR responses are built.
Interaction of Na+-Ca2+ exchange and CICR
Exchanger molecules are thought to be inhomogeneously distributed on the plasma membrane of neurons (Reuter and Porzig 1995), and as we have described, Ca2+ can be very locally distributed in a dendrite; it seems possible therefore that the exchanger might have a close functional association with the Ca2+ sensor of CICR, as is the case for the smooth muscle (Nazer and van Breemen 1998). On this view, the exchanger might not actually contribute very much to the clearance of Ca2+ from the bulk of the dendritic cytoplasm. To investigate this possibility, we performed experiments to monitor the activity of the Na+-Ca2+ exchanger while simultaneously imaging [Ca2+]i. Exchange was monitored by recording the small inward current generated by the exchanger as a consequence of the fact that three positive charges, carried by Na+, enter the cell for every two charges, carried by Ca2+, that leave the cell. We have shown previously that following a depolarizing step in ionic conditions similar to those used here, the slow “tail” current seen at the holding voltage is dominated by the exchange current (Gleason et al. 1995). A complication in this approach is that the exchange current measured in our patch-clamp configuration contains contributions from the cell body as well as the dendrites. Because direct observation (data not shown) indicated that the timecourse of [Ca2+]i in the cell body can be significantly different from that in the dendrites, the correlation of dendritic [Ca2+] with whole cell exchange current is unlikely to be very meaningful. To avoid this problem, we used local perfusion of a dendrite by a small-tipped pipette to enable exchange along the dendrite under observation while it was disabled elsewhere in the cell.
Experiments of this kind were automated so that the several solution changes could be carried out reliably and the entire sequence repeated in order that the currents and [Ca2+]iestimates could be averaged. One experiment is illustrated in Fig.7. Figure 7 A shows currents from the whole cell along with [Ca2+] in a segment of dendrite. Leak-subtracted Ca2+ current is shown for a 500-ms depolarization from −70 to 0 mV. The tail current, which is seen on return to −70 mV and decays over 2 s, is largely generated by Na+-Ca2+ exchange (Gleason et al. 1995). As previously described, [Ca2+] in the dendrite under examination shows a rapid rise in concentration that has almost returned to baseline within 10 s. In the ubiquitous absence of Ca2+ o (B), depolarization was ineffective in raising dendritic [Ca2+] and elicited only a small outward current (not shown) that has been used as a measure of the leak current and subtracted from all records. The absence of a [Ca2+]i transient in these trials confirms the need for Ca2+ entry in promoting Ca2+ release from the ER and rules out the possibility of direct conformational coupling between L channels and RyRs as is well established in skeletal muscle. When Ca2+ was removed from all of the cell except for the perfused dendrite (C), Ca2+ current was smaller, about one-fifth the size of the current entering the whole cell. Despite this, the magnitude of the [Ca2+] transient in the dendrite was almost as great as it was when Ca2+ bathed the entire cell, implying that lateral diffusion along the dendrite is not a significant factor in determining [Ca2+] following its influx. When Na+ was removed from the entire cell (D), the [Ca2+] transient in the dendrite was slightly larger than in normal solution and, as described earlier, the peak occurred later and the fall of [Ca2+]i was slower. Without Na+ o, the inward exchange current was suppressed leaving only a very small slow component to the tail that we have previously shown is due to Ca2+-activated Cl− and K+ channels (Gleason et al. 1995). In E, Ca2+ was present over the whole cell but Na+ was applied only to the dendrite under observation. The [Ca2+] transient seen in the dendrite resembles that seen in normal external solution, implying that Na+-Ca2+ exchange locally within the dendrite is able to remove Ca2+ from that dendrite. Surprisingly though, the tail currents seen in this condition were not measurably different (shown enlarged in Fig. 8) from those seen when Na+ was removed from the entire cell (D) and in none of 10 cells examined was a difference in tail currents resolvable.
This result is unexpected because on the one hand, comparison of the Ca2+ transients in D and E shows that Na+-Ca2+ exchange is effective in shaping Ca2+ dynamics in the dendrites but on the other hand, evidence of this exchange cannot be had by looking for its signature tail current. Because the Ca2+ current seen in Cis about one-fifth the size of the Ca2+ current seen for the whole cell in A, the tail current produced by the perfused dendrite would be expected to be about one-fifth the size of that of the whole cell if Ca2+ current and exchange current scaled proportionally. This is not the case and, within the limits set by noise in our data, the exchange current cannot be more than 1/20 of the magnitude for the whole cell. From these experiments, we conclude that Ca2+ current and exchange current do not scale proportionately in the cell body and the dendrites. In dendrites, a small amount of Na+-Ca2+ exchange, so small as to be electrically undetectable, is apparently able to exert a strong influence on dendritic [Ca2+] by controlling Ca2+ release from internal stores.
The experiments described here allow a rejection of the simple picture that Ca2+ entering through open VGCCs elevates dendritic [Ca2+] that is then reduced by the action of Na+-Ca2+ exchange and other processes. Brief depolarizing steps in voltage clamp, or K+ depolarizations of unclamped cells, produce dendritic [Ca2+] transients in which a substantial fraction, the majority as indicated in Fig. 5, is contributed by Ca2+ released from the ER by CICR, rather than Ca2+ entering through Ca2+ channels. This arrangement can be usefully thought of as a Ca2+amplifier, taking the small amount of Ca2+ entering through Ca2+ channels and boosting it with internally stored Ca2+. Positive feedback, provided by the released Ca2+, tends to increase the gain of the amplifier while Na+-Ca2+ exchange works to reduce the gain.
Na+-Ca2+ exchange and the Ca2+ amplifier
Na+-Ca2+ exchange clearly has a major influence on dendritic [Ca2+] as evidenced by the, often dramatic, effect that removal of Na+ o has on dendritic [Ca2+] dynamics. The results shown here are similar in several ways to those described by Reuter and Porzig (1995) for presynaptic terminals of hippocampal neurons, although in that study neurons were not voltage clamped. In particular, hippocampal neurons showed prolonged but variable [Ca2+]i dynamics in the absence of [Na+]o and like amacrine cells could show stable plateau [Ca2+]i following depolarization. A major point of difference though is in the effect of thapsigargin which is reported by Reuter and Porzig (1995) to have no significant effect, whereas in our study, it clearly produces a large reduction in both the amplitude and duration of [Ca2+]i following depolarization, providing sufficient time is allowed for the ER stores to empty completely.
The experiments in which forward-mode Na+-Ca2+exchange was disabled employed the removal of Na+ o and the substitution of a non-transportable ion. This does not, of course, prevent reverse mode exchange in which the Na+-Ca2+ exchanger, being thermodynamically reversible, operates so as to import Ca2+while expelling cytoplasmic Na+. However, because removal of [Ca2+]o in the experiments illustrated in Fig. 2 B has no effect on the falling phase of [Ca2+]i transients, we conclude that reverse mode exchange contributes little to dendritic [Ca2+]i dynamics. Because forward-mode exchange has a consistent, and in many cells dramatic, influence on [Ca2+]i dynamics, this conclusion might seem surprising. Two considerations reconcile these observations. First, although thermodynamics requires that exchange is reversible, the rate of exchange in reverse mode might well be low. Second, as discussed in the following text, Na+-Ca2+ exchange influences [Ca2+]i in an indirect and nonlinear way.
A striking feature of our results is the variability between cells seen in experiments in which [Na+]o was removed (see Fig. 1). In some cells, [Ca2+]i dynamics were drastically affected, whereas in a minority, the effect was relatively subtle. We cannot rule out the possibility that these differences characterize different cell types because our cultures may include more than one type of amacrine cell; but we have noticed no correlation with cell morphology. In contrast to this variability, the [Ca2+]i dynamics seen in normal [Na+]o were similar in cells drastically affected by [Na+]o removal and those much less affected. Very likely the implication of this is that in a minority of cells, some mechanism other than Na+-Ca2+ exchange has a leading role in moderating the gain of the Ca2+ amplifier.
The precise way in which Na+-Ca2+ exchange moderates Ca2+ release from ER stores remains to be elucidated. Several pieces of evidence suggest, however, that the interaction is more intimate than simply that the exchanger reduces [Ca2+]i in a well-stirred cytoplasmic compartment in which the ER is subject to CICR. As shown in Fig. 7, Na+-Ca2+ exchange is able to reduce dendritic [Ca2+]i transients, even though the summed activity of the exchanger in a dendrite, shown by its signature current, is undetectably small. One interpretation of this, illustrated diagrammatically in Fig. 9, is that the exchanger is not the chief mechanism for clearance of Ca2+from a dendrite but instead operates close to the Ca2+sensor of CICR where it can have a disproportionate effect on [Ca2+]i. Implicit in this interpretation is that the exchanger sees a higher [Ca2+] than is measured in the bulk of the dendrite. This consideration would resolve a paradox concerning the apparent concentration range over which the exchanger is effective. On the one hand, disabling the exchanger had no effect on resting [Ca2+]i, as would be expected because the K 1/2 for exchange is ∼1 μM (Blaustein and Lederer 1999) and below this, in other systems, exchange is found to be ineffective (Gall et al. 1999). On the other hand, exchange clearly exerted an effect following depolarization even though [Ca2+] in bulk dendritic cytoplasm was below 1 μM, as seen in most panels of Fig. 1.
An alternative explanation of how the exchanger might control Ca2+ release is that the exchanger competes with the SERCA pump for Ca2+ during a [Ca2+]itransient. When the exchanger is disabled, the internal store is able to continuously replenish itself and thereby sustain a longer and larger [Ca2+]i transient. This explanation implies that the Ca2+ stores are relatively small.
Local control of Ca2+ in a dendrite
Our results with both linescan and video imaging show that depolarization can give rise to local hotspots and cool spots for [Ca2+]i in amacrine cell dendrites that can persist for many seconds. These local differences are not the consequence of an inhomogeneous distribution of Ca2+channels because local [Ca2+]i gradients only become apparent after Ca2+ channels have closed. Instead, local [Ca2+]i differences reflect release from internal stores. The idea that small regions of dendrites can exert local control of [Ca2+]i via internal stores now has strong experimental support in the case of postsynaptic spines of cerebellar Purkinje cells and pyramidal neurons of hippocampus and neocortex (Finch and Augustine 1998;Koester and Sakmann 1998; Takechi et al. 1998; Yuste and Denk 1995). A smaller body of evidence shows that, similar to the results shown here, CICR may also be seen in presynaptic structures such as presynaptic boutons of pyramidal cells (Emptage et al. 2001), Purkinje neurons (Llano et al. 2000), and presynaptic terminals of sympathetic neurons (Peng 1996).
A surprising feature of our results though, is that long-lasting [Ca2+]i gradients of several hundred nanomolar can exist in smooth dendrites without spines or boutons to act as physical barriers to diffusion. Some evidence that this is a normal feature of amacrine cell physiology comes from a recent study using 2 photon microscopy to view cells in the intact retina (Denk and Detwiler 1999) in which amacrine cell dendrites are reported to show very local differences in [Ca2+]i when receiving normal synaptic input. Another finding of that study is that, as we report, local spontaneous increases in [Ca2+]i are sometimes seen. At least some of these probably represent elementary events from which more long-lasting [Ca2+]i transients are built up. Inspection of long-lasting [Ca2+]itransients, like those in Fig. 5, usually fails to reveal a temporal structure consistent with this idea but closer investigation is required to resolve this question.
Hotspots for [Ca2+] in amacrine dendrites very likely correspond to sites of Ca2+ release through RyRs and possibly InsP3Rs. The identity of cool spots is less certain but might include the SERCA pumps of the ER itself. SERCA pumps are thought not to be co-located with RyRs, and there is growing evidence showing that the ER is quite inhomogeneous with regard to its Ca2+handling molecules (Golovina and Blaustein 1997; reviewed in Meldolesi and Pozzan 1998;Pozzo-Miller et al. 2000). A recent study of Ca2+ handling in DRG neurons divides Ca2+handling by the ER at low and moderate loads into two modes (Hongpaisan et al. 2001). At the lowest Ca2+loads, the ER is a net importer of cytoplasmic Ca2+, whereas at higher loads the ER is a net exporter of Ca2+. We suggest that in those stable patterns such as Fig. 3 A,the ER is balanced between these two modes, possibly allowing Ca2+ tunneling between the sinks and sources for Ca2+ in the ER (Petersen et al. 1999).
Ca2+ flux implied by local gradients
When Na+-Ca2+ exchange is disabled, Ca2+ influx can generate stable spatial patterns of [Ca2+]. It might be thought that the existence of standing gradients like that illustrated in Fig. 3 A because they imply continuously active sources and sinks would place an unreasonable energetic burden on a cell and quickly empty internal stores. Neither of these suppositions is likely to be true because the fluxes required to sustain the observed gradients are small as shown in this order-of-magnitude calculation.
In the steady state, immobile buffers have no influence on flux, thus the total flux, J T, between a source and a sink is the sum of the flux of free Ca2+ plus the flux of Ca2+ bound to mobile buffer For simplicity, we neglect intrinsic mobile buffers and assume that Fura-2 is the only mobile buffer and is present at a total concentration, [B]T, of 500 μM with aK D of 200 nM and a diffusion coefficientD B = 100 μm2/s that we assume is identical for the bound and free forms of Fura-2.
For a constant flux, J T, in a one-dimensional dendrite, Fick's equation can be integrated and rearranged to yield Provided the binding kinetics of Fura-2 are rapid compared to diffusion (Smith et al. 1996; Wagner and Keizer 1994), Ca2+ will be in equilibrium with Fura-2 everywhere, and the concentration of Ca2+ bound to Fura-2, [CaB] can be obtained from Considering a typical [Ca2+] gradient of 200 nM over a distance, Δx, of 5 μm, then Δ[CaB] = 83 μM, and the associated fluxes, assumingD Ca = 300 μm2/s, are Note that the flux due to free Ca2+,J Ca, is <1% of the total flux. Most of the Ca2+ is carried by Fura-2. We can express the total Ca2+ flux in a more intuitive form, as the Ca2+current at the source where F is Faraday's constant.
Therefore, in a dendrite with a cross sectional area, A= 1 μm2, the total current at the source is ∼0.33 pA. This current is close to the estimate of 0.35 pA for the current through a single RyR (Mejı́a-Alvarez et al. 1999) and 44 times less than the peak current in a Ca2+ spark of skeletal muscle (Rı́os et al. 1999).
What function might be served by the local release and uptake of Ca2+ in amacrine dendrites? In the dendrites of hippocampal pyramidal cells and cerebellar Purkinje cells, local [Ca2+] regulation has been linked to long-term changes in synaptic efficacy (Nishiyama et al. 2000) and gene expression (Deisseroth et al. 1996). These mechanisms may also be relevant to amacrine cell dendrites but a peculiarity of amacrine cells is that, lacking an axon, dendrites act as pre- as well as postsynaptic structures, suggesting that local [Ca2+] release controls local release of transmitter. Strong support for this supposition is provided by experiments showing that disabling Na+-Ca2+ exchange extends the duration of asynchronous transmitter release (Gleason et al. 1994). Recent studies of both hippocampal CA3 pyramidal cell presynaptic boutons (Emptage et al. 2001), and Purkinje neurons from the cerebellum (Llano et al. 2000) suggest that in those neurons as well, Ca2+ release is coupled to transmitter release.
An implication of the work described here is that presynaptic sites, even when close together, might have different transmission characteristics. If, as we suggest here, the exchanger enjoys a privileged location relative to the ER, we speculate that local regulation of Na+-Ca2+ exchange could be used to locally influence transmitter release. Clearly the location of synapses relative to the ER and the exchanger will be an important factor in this and remains to be determined. Small alterations in the exchanger, by phosphorylation, or allosteric binding of ATP (DiPolo and Beaugé 1999), would be expected to change the gain of Ca2+ amplification and thereby affect the time course of synaptic transmission. A clear expectation is that, because the rate of Na+-Ca2+ exchange is voltage dependent (Dipolo et al. 1985; Hayashida et al. 1998; Kimura et al. 1987), CICR ought therefore to be voltage dependent in a way separate from the voltage dependence of Ca2+ influx. This possibility could lead to a multiplicative interaction between depolarization and Ca2+influx on transmitter release.
Synaptic transmission is usually thought of as a process in which widespread voltage changes in a neuron are coupled rather rigidly to the influx of Ca2+ and the subsequent release of transmitter. Our results here, because they identify a local Ca2+ amplification step, imply a looser coupling between membrane voltage and the Ca2+ available for transmitter release. Na+-Ca2+ exchange is one mechanism that can influence this step but very likely there are others, including local [Ca2+] increases brought about by the local action of transmitters acting on both ionotropic and metabotropic receptors. From this perspective, the mixing of pre and post synaptic sites along an amacrine dendrite would effectively allow a single amacrine cell to behave as multiple processing units loosely coupled to each other by membrane voltage but independent by virtue of their local [Ca2+] regulation.
We thank G. Benison for contributing to the initial experiments of this work and I. Pessah for advice and discussion during this work.
This work was funded by National Eye Institute Grants EY-04112 and EY-12576.
Address for reprint requests: M. Wilson, Section of Neurobiology, Physiology & Behavior, Division of Biological Sciences, UC Davis, Davis, CA 95616 (E-mail:).
- Copyright © 2002 The American Physiological Society