INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Traditionally, translation of electricity to biochemistry is thought to take place at the nerve terminal of a neuron where electricity (the action potential) is translated into biochemistry via calcium mediated vesicular release of neurotransmitters. Prior to the nerve terminal, the axon has been regarded as a high-speed conduit whose major role is transmission of electricity rather than translation of electricity to biochemistry. This picture of axons is no longer tenable, as recent studies have firmly established activity-dependent calcium influx into mammalian axons, including unmyelinated axons from rat neocortical pyramidal neurons (Schiller et al. 1995), cerebellar Purkinje cells (Callewaert et al. 1996), cultured dorsal root ganglion cells (Lüscher et al. 1996), neonatal rat optic nerve (Sun and Chiu 1999), and rat vagus nerves (Wächtler et al. 1998). Activity-dependent calcium signals have also been detected from myelinated axons in the rat optic nerves (Lev-Ram and Grinvald 1987). The physiological role of axonal calcium influx remains unknown. In developing axons of the rat cerebellar interneurons, Forti et al. (2000) observed local calcium hot spots, which they speculated, might represent functional clusters of voltage-dependent Ca²⁺ channels marking the future position of presynaptic terminals. A local rise in calcium may trigger actin/cytoskeletal assembly (Bentley and O'Connor 1994; Lankford et al. 1996). Edwards and Cline (1999) hypothesize that local axonal calcium influx may regulate the addition of new branch points in growing retinal axons. However, in mature axons such as the mammalian optic nerves where varicosities or branches are absent, the functional role of calcium influx along the entire length of axon remains unclear. Pathologically, excessive elevation of axonal calcium is injurious to axons (for review, see Stys et al. 1995a,b). One of the best-studied mammalian CNS white-matter tracts where calcium has been implicated in axonal pathology is the mammalian optic nerve (Stys et al. 1995). This nerve can be isolated without neuronal contaminations and is highly amenable to experimental studies, including calcium imaging (Kriegler and Chiu 1993; Sun and Chiu 1999), electrophysiology and ischemia (Stys et al. 1995). These studies have led to the identical of physiological and pathological calcium-influx pathways. In particular, the reversal of the Na/Ca exchanger in ischemic axons has been proposed to produce damaging calcium influx into axons (Stys et al. 1995).

In this report, we examined the mechanisms for calcium clearance in axons of optic nerves. Because of the technical difficulty of staining axons in these nerves with calcium dyes for direct visualization of axonal calcium signals, very little is known about the calcium clearance mechanisms in axons of mammalian optic nerves under physiological conditions, including such basic issues as how these axons clear a calcium load following repetitive nerve activity and how the clearance is regulated. In this paper, we removed a major technical difficulty in studying calcium signaling in the optic nerve axons by devising a method to selectively stain pure population of axons in these nerves with calcium indicators, thereby allowing the first direct visualization of axonal calcium signals in these nerves in physiological time scales. We then applied this technique to examine some very basic properties of calcium clearance in these axons. We found a strong coupling between axonal [Na]_i and axonal [Ca²⁺]_i and suggest that this coupling is best explained if the Na/Ca exchanger is a key player in calcium extrusion following physiological nerve activity.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Selective axonal staining and calcium imaging

Optic nerves, excised between the eye and the chiasm, were freshly obtained from mice of two age groups, neonatal (P5-P10) and adult (3-8 wk), and laid down on the bottom of an experimental perfusion chamber. The distal end of the nerve trunk was loosely drawn into a stimulating pipette, and the proximal end drawn tightly into a recording pipette (Fig. 1A). The nerves were allowed to stabilize for 60 min before dye loading began. For dye loading, the normal saline solution in the recording pipette (the one with the cut end of the nerve tightly drawn in) was replaced a high-K (140 mM), low-calcium (0 calcium plus 1 mM EGTA) solution containing 4-5 µl of the cell impermeant form of various calcium dyes (2-8 mM). The axons were stained by diffusion of the dye from the cut end. A standard loading time of 3 h was allowed before experiments began. At the end of the loading period, there was typically a spatial gradient of dye along the axis of the nerve, with the resting dye fluorescence highest at the loading end, and smallest at the other end (the stimulating end). We typically performed calcium imaging in the middle of the nerve trunk, ~500-1,000 µm away from the dye-loading pipette. The dyes were allowed to remain in the loading pipette (also serve as the recording pipette) during the entire experiment. During a typical 1- to 2-h experiment, the calcium fluorescence baseline measured at the middle of the nerve was quite stable in most studies (see Fig. 11). We believed that this stable baseline was due partly to our use of the high-molecular-weight, dextran-conjugated Oregon Green bis-(o-aminophenoxy)-N,N,N',N'-tetraacetic acid (BAPTA)-1, which has restricted dye mobility and excellent retention in the axons without leaking. While a slowly changing baseline fluorescence was sometimes observed in some experiments, these slow baseline changes can be easily distinguished from fast calcium fluorescence changes induced by the various pharmacological (veratridine, ouabain, or monensin) used in this study. In most of the figures, we reported the calcium responses as either F/F₀ or F/F₀, which normalizes differences in background fluorescence due to differences in dye loading.

View larger version (60K): [in this window] [in a new window]   Fig. 1. Selective staining of axonal populations in mouse optic nerves. A: schematics of the experimental setup. An excised optic nerve is loosely drawn into a stimulating pipette at one end (left) and tightly drawn into a recording pipette at the other end (right). The tight recording pipette also contains a calcium dye solution (shaded) with EGTA present to keep the cut ends open to facilitate diffusion of the impermeant form of the dye into the axons. A ×40 objective in the upright configuration provides the pathway for both excitation of the calcium dyes and the emission of calcium fluorescence. Acquisition of both the calcium fluorescence and the compound action potential is under computer control. B: selective staining of axonal populations in this study. Right: staining of axons and glial cells by the method of Kriegler and Chiu (1993). Cell-permeant form of the calcium dye was injected into the optic nerve trunk and taken up and retained in both axons and glial cells. Left: selective staining of axonal populations in this study. Cell-impermeant form of the calcium dye was loaded into axons via diffusion from the cut ends of the optic nerve (see METHODS). Note the absence of glial staining. C: comparison of whole-field calcium response to spot calcium responses. Calcium responses were evoked by a single action potential (). Whole field responses were obtained from an area of 150 ×100 µm. Spot responses were obtained from 5 spots (50 µm2) selectively randomly from the same field of view. P8 optic nerves. Calcium images for both A and B obtained from the middle of the nerve trunk.

Confocal fluorescence imaging of axonal calcium

Calcium images were viewed with a ×40 (Olympus) objective lens on a Noran Odyssey confocal system (Madison, WI). Throughout the course of our study, we tried four membrane-impermeant calcium indicators: Oregon Green 488 BAPTA-1 (MW = 1114) with or without conjugation to the high-molecular-weight dextran [MW = 10,000; K_d = 170 nM, K_d values from Handbook of Fluorescence Probes (6th ed.), Molecular Probes], Oregon Green 488 BAPTA-2 (K_d = 580 nM), Magnesium Green (K_d = 6,000 nM), and Oregon Green 488 BAPTA-5N (K_d = 20,000 nM). We finally chose the high-affinity, dextran-conjugated Oregon Green BAPTA-1 due to its excellent dye retention and signal-to-noise ratio. The disadvantage of this high-affinity dye is potential distortion of calcium signal kinetics due to dye saturation. The low-affinity dyes were used in this study mostly as a control to assess dye-saturation as a potential source of error in the interpretation of our data. All dyes were used at a concentration of 5-10 mM in the loading pipette. The actual axonal dye concentration at the site of optical recording was estimated to be ~0.013 of that at the loading pipette (see RESULTS). Most of the experiments were performed with the Noran Odyssey System where the dyes were excited with an argon laser at 488 nm and confocal fluorescence images monitored with a 500-nm long-pass emission filter. The average fluorescence signal from the whole field (15,000 µm² area) was collected on-line at near video rate (30 Hz) and stored for off-line analysis; image acquisition and on-line calculations were controlled through the Metamorph software (Universal Imaging). The disadvantage of the Noran system is poor time resolution (the calcium signal was sampled every 30-40 ms). To achieve higher time resolution where needed, some experiments were performed with a Praire Technology System (NED) that collects nonconfocal calcium signals every 500 µs. In most figures, intracellular calcium concentration was reported as F/F₀ or F/F₀ without calibration for absolute values. Experiments were either done at room temperature (22°C) or at 36°C. Temperature was controlled by a DC-thermistor-based system.

Electrophysiology

Compound action potentials were evoked by a 125% supra-maximal stimulus applied via the suction electrode to the cut end and recorded from a second suction electrode at the other cut end. Compound action potential (CAP) data were analyzed using pClamp 6.0 software (Axon Instruments, Foster City, CA).

Solution and drugs

The optic nerves were normally bathed in a Ringer solution that contained (in mM) 129 NaCl, 3 KCl, 1.2 NaH₂PO₄, 2.4 CaCl₂, 1.3 MgSO₄, 3 HEPES, 20 NaHCO₃, and 10 glucose. Calcium-free solutions were prepared by replacing Ca²⁺ with Mg²⁺ and by adding EGTA (1 mM); pH was adjusted to 7.4 with NaOH or HCl as necessary. All compounds were from Sigma.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This paper consists of two sections. In the first section, we will describe a technique for selective staining of axons in the mouse optic nerves and provide arguments as well as experimental evidence that we have indeed succeeded in selectively staining axons. This section is an important technical prerequisite to the second section, which uses this staining technique to examine several basic properties of calcium clearance mechanism in the mouse optic nerve axons.

Technique for selective staining of axonal population

In the rat optic nerve, the mean diameter of the axons remains ~0.2 µm for the first postnatal week and increases rapidly to ~1 µm in adult (>P28 days after birth) when the axons are fully myelinated (Foster et al. 1982). No morphometric data exist for the mouse optic nerves, the nerves used in this paper, and the developmental profile for the axon diameters is assumed to be similar to rats. An important prerequisite for a rigorous analysis of axonal calcium homeostasis is selective staining of axons with calcium indicators. In our previous studies, the cell-permeant form of the dye was injected into the extracellular space in the nerve trunk, which was taken up by both glial cells and axons (Kriegler and Chiu 1993; Sun and Chiu 1999). This staining method is useful for analysis of axonal signals only for short electrical stimulations that evoke only axonal calcium transients. For prolonged electrical stimulations, glial calcium responses are evoked (Kriegler and Chiu 1993) that contaminate the axonal signal, rendering conclusive analysis of axonal calcium homeostasis impossible. In this study, we selectively stained axons by allowing the cell-impermeant form of the calcium dye to diffuse into axons via a nerve cut end for 3 h before experiments began (Fig. 1A). Figure 1B compared optic nerve staining obtained with Kriegler and Chiu (1993)'s method, which stained both axons and glia (left), and that obtained with our current method (right). Our current staining method resulted in no glial cell staining (Fig. 1B, right), both in neonatal and in adult optic nerves. When we first started this study, we used Oregon Green BAPTA-1 (MW = 1,114). One concern we had was that this low-molecular-weight dye may cross gap junctions, allowing dyes that enter damage glial cells at the cut end to spread through glial network to reach the site of optical recording, which is typically 500-1,000 µm away from the site of dye loading. However, we did not observe staining of glial cells at the optical recording site, suggesting that this source of contamination of the axonal signal did not occur. To further exclude dye transfer via glial gap junctions, we switched (in later experiments) to calcium dyes conjugated to the high-molecular-weight dextran (MW = 10,000; Oregon Green BAPTA-1), which cannot penetrate gap junctions (Eckert et al. 1999). Because the site of optical recording is 2.5-5 times longer than the length of the average longitudinal processes of glial cells in mammalian optic nerves (~200 µm) (see Butt and Ransom 1993), transfer of the dextran-conjugated dye through glial processes to the recording site is unlikely.

To further confirm that we have in fact selectively stained axons, we performed the experiment in Fig. 2. We first loaded the calcium impermeant form of the calcium dye (dextran-conjugated Oregon Green) through the cut ends of the nerve with a tight suction pipette, the standard loading method in this paper. After 3 h (A), we evoked a whole-field calcium response with a single compound action potential (). The nerve was allowed to rest for 10 min, after which adenosine was applied to the bath (Fig. 2A). The important observation is that no adenosine response was seen. The rationale behind the use of adenosine is that glial cells are known to express adenosine receptors, and previous calcium imaging studies from our laboratory (Kriegler and Chiu 1993) have demonstrated adenosine-evoked calcium responses in glial cells of mammalian optic nerves in situ. The fact that no adenosine-evoked calcium response was seen following our standard dye loading suggests that glial cells were not stained in our current study, and that the action potential evoked response () is purely axonal in origin. To show that glial cells are indeed viable and can respond to adenosine, we stained glial cells in the same nerve with a second dye loading (Fig. 2B) using the method of Kriegler and Chiu (1993). This was done by injecting into the middle of the nerve trunk, the calcium-permeant form of Oregon Green, which was taken up by both glial cells and axons. Two hours after this second dye loading, glial cells were stained and clearly visible. Further, the background fluorescence was greatly increased, due presumably to additional loading of the axonal population following the second dye loading. To compare the calcium responses following the second loading to that of the first one, we normalized the calcium fluorescence signal to the prestimulation baseline signal. Confirming our previous studies (Kriegler and Chiu 1993), this second dye loading, which stained glial cells, produced a robust adenosine response (Fig. 2B). To further exclude the possibility that our standard axonal loading paradigm (i.e., the 1st dye loading), if allowed to stain for an additional 2 h after the first 3 h (the addition time needed for the 2nd loading), would by itself lead to gradual glial staining, we did the control experiment in Fig. 2, C and D. Here we employed only the standard axonal loading method (putting the dye in the loading pipette), and applied adenosine twice, one at 3 h (C), and another at 5 h (D) after loading. In neither case was an adenosine response seen (C and D), suggesting our axonal loading method did not lead to glial loading even after extended periods.

View larger version (20K): [in this window] [in a new window]   Fig. 2. Verification of axonal staining with adenosine. In this experiment, we relied on the known presence of adenosine-induced calcium response in glial cells to verify selective axonal staining in our paper. A: 3-h loading with our standard axonal loading method produced an action potential evoked calcium response () but no adenosine response. B: the same nerve in A was subjected to a second dye loading using the method of Kriegler and Chiu (1993) in which both glial cells, in addition to axons, were stained. This resulted in an adenosine response, consistent with staining of glial cells. C and D: evoked calcium response and adenosine response assayed at 3 (C) and 5 h (D) after only axonal loading. No adenosine response was detected in both cases. P4 (C and D) and P7 (A and B) nerve. See text for details.

Our loading method relies on diffusion of dyes from the cut ends along the axoplasm of axons. Is it reasonable to expect that a dextran-conjugated molecule with a large MW of 10,000 can produce significant diffusion in only 3 h to allow calcium imaging in our studies. For example, McClellan et al. (1994) found that injection of dextran-conjugated Calcium Green in spinal cords resulted in retrograde labeling of neurons 5-14 mm away but only after 4 days. How might it be possible that we observed detectable dye loading in axons 3 h after dye loading? One explanation is that the mode of dye entry may be different. In McClellan et al. (1994), the dextran dye, which is impermeant, will have to be taken up by the cells, presumably via pinocytosis. In our case, we cut the axons, which allowed direct dye entry through the cut ends. Further, our loading pipette is filled with a calcium-free, EGTA-buffered solution intended to keep the cut ends of the axons from re-sealing. Hence, our in vitro loading method may allow faster dye diffusion into axons than in vivo loading. We further estimated the coefficient of diffusion for the dextran dye in our experiments. At a distance of ~540 µm from the loading site (the typical site of optical recording in experiments involving dextran dyes), the resting fluorescence is ~0.013 (n = 3) of the fluorescence in the loading pipette (which contains 2-8 mM of the dye) after a 3-h loading period. From this we estimated a diffusion coefficient (Crank 1956) of ~6.8 µm²/s at room temperature. This is considerably smaller than the coefficient of diffusion of 102 µm²/s measured for fura-2 (Gabso et al. 1997), which is a much smaller molecule. Our estimate suggests a very sluggish diffusion of dextran in our experiments. However, the use of a high dye concentration in the loading pipette evidently allowed sufficient dye diffusion to a recording site ~540 µm away to allow optical recording to be performed after a 3-h loading period.

In most of the experiments, we measured the whole-field calcium response (area ~15,000 µm²) elicited from a large number of axons. One concern is whether the whole-field response misses local heterogeneity and whether it faithfully represents response at the single axon level. We therefore compared the whole-field response simultaneously with several spot responses selected randomly from the same nerve, with each spot having an area that is ~300 times less than the whole field. Figure 1C shows the shape of the whole field (smooth trace) and spot (nosier traces) responses are virtually identical, suggesting the calcium response is spatially homogeneous and that the whole-field response faithfully captures the essence of the response at the single fiber level. Given that whole-field response has excellent signal-to-noise ratio, this measurement was routinely adopted as the primary tool for analysis of axonal calcium homeostasis in this paper.

Prior to our study, Ren et al. (2000) published a method of selective loading of axons in the mammalian optic nerves with calcium indicators by incubating the nerve in a solution containing the cell-impermeant form of the dye. Both that work and our work are in agreement that axons are selectively stained. Collectively, we believe that we have achieved staining of a pure population of axons in the optic nerve for rigorous analysis of axonal calcium homeostasis in this study. In the following section, we will apply this technique to examine several basic properties of axonal calcium clearance in the mouse optic nerves.

Basic properties of axonal calcium clearance mechanisms

4-AP INCREASES EVOKED CALCIUM TRANSIENTS. To facilitate analysis of calcium responses with acceptable signal-to-noise, in some experiments we applied 4-AP, a blocker of fast K channels, to increase the amplitude of the evoked calcium transients. Figure 3 shows the effects of 4-AP on the calcium transient and compound action potential evoked by a single stimulation for a P6 (top) and an adult (bottom) optic nerve. In both nerves, 4-AP increased the amplitude and duration of the compound action potential (B and E) and concomitantly increased the area of the evoked calcium transient (C and F). Interestingly, 4-AP has a much more dramatic enhancing effect on the evoked calcium load (area of the calcium transient) in the adult than in the neonatal nerve. For example, 1 mM 4-AP enhanced the calcium load per action potential ~80 times in adult (D) versus ~13 times (D) in the neonatal nerve. In experiments where 4-AP was used to increase the calcium transient, a concentration of 1 mM was used.

View larger version (30K): [in this window] [in a new window]   Fig. 3. Effects of 4-aminopyridine (4-AP) on action potentials and evoked axonal calcium transients on neonatal (top) and adult (bottom) optic nerves. A and D: dose-response curve for relative calcium load per action potential and 4-AP concentrations. Calcium load is calculated as the area under the entire calcium transient evoked by a single action potential. The dose-response data were fitted with a sigmoidal curve with a Kd of ~65 µM in A. B and E: compound action potentials at various 4-AP concentrations. C and F: corresponding evoked calcium transients. Each calcium transient was evoked by a single action potential. P6 (top) and 10-wk-old adult (bottom) optic nerves. Calcium dye is Oregon Green 488 BAPTA-1.

View larger version (16K): [in this window] [in a new window]   Fig. 4. Comparison of evoked axonal calcium transients in neonatal and adult optic nerves. A: averaged axonal calcium transients evoked by a single action potential in P7 and adult (7 wk) optic nerves. B: comparison of the shape of the neonatal and adult responses from A. Temperature 22°C

GRADED CALCIUM DECAY FROM PROGRESSIVELY HIGHER CALCIUM LEVEL. We started our analysis of calcium homeostasis by examining the time course of axonal calcium restoration following repetitive nerve activity. Figure 5A shows superimposed axonal calcium transients evoked in a P27 optic nerve (without 4-AP) by a single action potential and a train of 20 action potentials (10 Hz). On termination of the stimulation, the axonal calcium decays toward baseline (A). Interestingly, the calcium decay is graded according to the calcium level at which the decay began, being slower following the 20 action potentials (which elevated axonal calcium to a higher level) than following the single action potential (B). The calcium decay is not a single exponential, either following the single or 20 action potentials.

View larger version (19K): [in this window] [in a new window]   Fig. 5. Dependence of posttetanus calcium decay on the calcium level reached during a tetanus. A: superimposed axonal calcium responses from a single action potential and a train of 20 action potentials at 10 Hz. The calcium fluorescence is in arbitrary units. B: normalized axonal calcium decay from data in A. The calcium decay was normalized with respect to the calcium level at the end of the stimulation at which the decay started. P27 optic nerves without 4-AP treatment. Temperature 22°C.

CALCIUM DECAY DEPENDS ON THE NUMBER OF ACTION POTENTIALS. In the preceding studies, both the calcium level at the start of the calcium decay and the number of action potentials during the tetanus were varied. It is possible that the number of action potentials itself (via axonal Na loading) is an important determinant of the posttetanus calcium decline. This issue was examined in Fig. 6. Figure 6A shows superimposed calcium responses to 1, 20, and 100 action potentials (at 5 Hz) at 22°C for a P8 nerve. The 20-action potential train elevates calcium to a higher level than the 1-action potential before the calcium decay began. However, for the 100-action potential train, the calcium response during the tetanus first reached a peak then slowly declined, so that by the end of the tetanus (at which the calcium decay began), the axonal calcium level is actually lower than that at the end of the 20-action potential train. The decline of the calcium response during the 100-action potential train may have various causes. First, there may be calcium-induced calcium inactivation, so that the increment of calcium influx per action potential is reduced during the tetanus. Second, there may be gradual axonal Na loading so that the amplitude of the compound action potential is gradually reduced during the tetanus, which leads secondarily to a reduction in the calcium influx per action potential. This gradual reduction of the compound action potential amplitude during a prolonged tetanus is frequently observed (also see Fig. 14), which is consistent with Na loading and is unlikely due entirely to refractory period. When we compared the posttetanus calcium decay following the 1-, 20- and 100-action potential train, we found a progressive slowing of the posttetanus decay (Fig. 6B). This is in spite of the fact the calcium decay started at a lower level at the end of the 100-action potential than at the 20-action potential train. Figure 6C shows posttetanus calcium decay from another experiment at 35°C where the calcium level at the start of the calcium decay actually reduced (from F/F₀ = 2.8 to 2.6 to 1.13) as the action potential number during the tetanus (5 Hz) increased from 20 to 100 to 300 respectively. Figure 6D demonstrates a similar phenomenon using two tetani of different frequency. The first tetanus is brief (4 action potentials/5 Hz), which brought the axonal calcium level to F/F₀ ~ 3 before the calcium decay. This is followed by a second tetanus that is considerably longer but at a lower frequency (40 action potentials/1 Hz), which elevated calcium to a slightly lower level (F/F₀ < 3) before the decay began. The second decay is slower than the first one (Fig. 6D).

View larger version (30K): [in this window] [in a new window]   Fig. 6. Dependence of the posttetanus calcium decay on the action potential number in a tetanus. A: superimposed calcium responses to 1-, 20-, and 100-action potential trains at 5 Hz. B: normalized calcium decay after the train from data in A. Note that decay after the 100-action potential train is slower than after the 20-action potential train even though the decay started from a lower level calcium level for the 100-action potential train. P7 optic nerve at 22°C. C: normalized calcium decays from another nerve at 35°C after 20-, 100-, and 300-action potential trains. The F/F0 values indicate the calcium level at the start of each decay. P5 optic nerve. D: posttetanus calcium decay after 2 tetani of different frequency. The 1st tetanus is 4 action potentials at 5 Hz (inset). The second tetanus is 40 action potentials at 1 Hz (inset). The normalized calcium decays show that even though both decays started from a similar calcium level, the 40-action potential train retarded the decay more than the 4-action potential train. P8 optic nerves treated with 1 mM 4-AP. E: residual component in the calcium recovery following long tetanus. An adult optic was stimulated with the sequence: brief tetanus - long tetanus - brief tetanus with ~18 min between tetani. The brief tetanus is 20 action potential/10 Hz. The long tetanus is 94 action potentials/1 Hz. The graph shows the axonal calcium signal over a ~9 min baseline and a ~12 min posttetanus period. The calcium response was normalized with respect to the calcium level at the end of the tetanus that coincided with the start of the calcium decay. The calcium response during the tetanus has been omitted for clarity of presentation. For most of the recording, the calcium signal was briefly sampled every ~1.5 min to avoid dye bleaching. Note the long tetanus induced a reversible, persistent component in the calcium recovery.

RESIDUAL COMPONENT IN CALCIUM CLEARANCE. Figure 6, A and C, shows that after a long tetanus, a residual, ultra-slow component in the calcium recovery appeared. We further examined this residual component, particularly with respect to its reversibility, over long time scales in Fig. 6E. To reliably measure the persistent component in the recovery, a 200-s baseline was obtained before applying tetanus stimulation. The control tetanus was brief (20 action potentials at 10 Hz) and produced no significant residual posttetanus component. As the tetanus was increased to 94 action potentials (at 1 Hz), a residual posttetanus component developed that lasted ~10 min before slowly returning to baseline. When the nerve was re-stimulated with the control brief tetanus 18 min later, the residual posttetanus component disappeared, showing reversibility in the development of the persistent component. This argues against irreversible axonal damage due to the long tetanus stimulation. In this plot, only the pretetanus baseline and the posttetanus calcium response were shown; the calcium response during the tetanus deleted for clarity of presentation. The residual component in the calcium recovery could be demonstrated in nerves with or without treatment with 4-AP.

PHARMACOLOGICAL INCREASES IN AXONAL Na IMPEDES POSTTETANUS CALCIUM CLEARANCE. The sensitivity of the posttetanus calcium decay on the number of action potentials suggests that axonal Na loading during the tetanus may impede posttetanus calcium clearance. To further test this idea, we examined the effects of pharmacologically increasing axonal [Na]_i on the posttetanus calcium clearance. To increase axonal [Na]_i, we bath applied monensin (4-50 µM, a Na-ionophore) (Senatorov et al. 2000), ouabain (20-30 µM, a Na/K-ATPase inhibitor), and veratridine (a modifier of Na channel inactivation).

View larger version (18K): [in this window] [in a new window]   Fig. 7. The Na-ionophore monensin retards posttetanus calcium decay. A and B: simultaneous recordings of compound action potential (A) and evoked axonal calcium response (B) before, during, and after bath application of 4 µM monensin. C: normalized posttetanus calcium decay from the data in B. The tetanus used was 20 action potentials/10 Hz. Note monensin retards the posttetanus calcium decay that is partially reversible on washing. P8 optic nerves treated with 1 mM 4-AP. Temperature 22°C.

View larger version (19K): [in this window] [in a new window]   Fig. 8. Inhibition of Na/K-ATPase (A and B) and Na channel inactivation (C) retard posttetanus calcium decay. A and B: plot of the half-time of posttetanus calcium decay before and after K-free (A) or 20 µM ouabain (B) application. The posttetanus calcium decay was assayed with a tetanus applied every 5 min. Inset: superimposed, normalized calcium responses before and after switching to the respective test solutions. Tetanus was 50 action potentials/10 Hz in A and 20 action potentials/5 Hz in B. Temperature was 35°C in A and 22°C in B. Age of optic nerve was 7 wk in A and P8 in B. C: effect of veratridine (5 µM) on axonal calcium signal evoked by a single action potential. The calcium response before and after veratridine was normalized with respect to the calcium level right after the action potential. P7 optic nerves. Temperature was 22°C.

POSTTETANUS CALCIUM CLEARANCE IS NOT AFFECTED BY K-DEPOLARIZATION. A simple explanation for the impeding effect of monensin (or ouabain) on the posttetanus clearance is that [Na]_i loading inhibits the Na/Ca exchanger that normally plays a role in posttetanus calcium extrusion. However, an alternative explanation might be that the membrane depolarization produced by monensin or ouabain (Leppanen and Stys 1997) directly inhibits the Na/Ca exchanger, which is electrogenic. This depolarization may be minimal in low dosage of monensin (4 µM) since the action potential amplitude was not appreciably affected (Fig. 7A). However, in higher doses of ionophore (or ouabain), there was a ~30-50% reduction of action potential that accompanies the retardation of the posttetanus calcium decay. We therefore tested if a K-mediated membrane depolarization can retard the posttetanus calcium clearance. To achieve a similar depolarization of the resting membrane as in the case of ionophore (or ouabain), the extracellular K was increased by 15 mM to produce a ~50% reduction in the action potential. Figure 9 shows that tetanizing the nerves in a solution with K elevated by 15 mM did not result in a slowing of the posttetanus calcium decline (B), suggesting that membrane depolarization cannot explain the retardation of posttetanus decay seen in Na loading experiments.

View larger version (19K): [in this window] [in a new window]   Fig. 9. Potassium-mediated depolarization does not retard posttetanus calcium decay. A: evoked axonal calcium response before and 5 min after the bath K was elevated by 15 mM. Calcium response was evoked using a tetanus of 20 action potentials at 5 Hz. The compound action potential was reduced (data not shown) with corresponding reduction in the evoked calcium response. B: normalized calcium responses showing that K-mediated depolarization did not retard the posttetanus calcium decay. P7 optic nerve. Temperature 22°C.

Na-free solutions and bepridil

Because we postulate that the Na/Ca exchanger is involved in calcium extrusion after a tetanus, we should be able to slow the extrusion by tetanizing nerves in a Na-free solution. The extrusion efficacy of this exchanger is coupled to the transmembrane Na concentration, which is normally high outside and low inside. We therefore tetanized nerves in solutions in which Na was replaced by lithium (Li). Li is permeable to Na channels and should support action potentials. We found that the compound action potentials (both in neonatal and in adult nerves) were gradually abolished in the Li-saline solutions. However, at 10-15 min after Li application (before a complete block of the action potentials), we found a approximate twofold slowing of the posttetanus calcium decay, consistent with the exchanger playing a role in the calcium extrusion (Fig. 10, A and B). Next, we applied bepridil (10 µM), which has been suggested to inhibit Na/Ca exchangers in ischemic optic nerves (Stys et al. 1995a,b). The compound action potentials also gradually declined in bepridil. However, we found an ~1.5-fold slowing of the posttetanus calcium decay after bepridil application (Fig. 10, A and B). Thus both the bepridil and Li experiments suggest that the Na/Ca exchanger is involved in the posttetanus calcium clearance.

View larger version (28K): [in this window] [in a new window]   Fig. 10. Summary data for pharmacological manipulation of posttetanus calcium decay (A and B) and resting calcium level (C and D) for neonatal and adult optic nerves. Top: plot of normalized changes in the half-time of posttetanus calcium decay in axons before and after various drug application for P5-P8 (A) and P60 (B) optic nerves. A: drug concentration: 20 µM (ouabain), 4 µM (monensin), 10 µM (bepridil), 5 µM (veratridine). For Li, bath Na was completely replaced by lithium (Li). B: same as A except veratridine was used at a lower concentration of 2 µM because adult nerve action potentials were rapidly blocked by 5 µM. In each drug assay, evoked calcium transients were sampled every 5 min for 4 times to obtain a baseline before drugs were applied. For each drug, posttetanus calcium decay was measured at a fixed time (10-30 min) after application. Calcium transients were evoked with 20 action potentials/5 Hz for the neonates (A) and 20 action potentials/10 Hz in the adult (B) in all drugs except veratridine. In the case of veratridine, a single action potential was used in the neonates and 10 action potentials/10 Hz for the adult. Bottom: plot of the relative changes in the resting axonal calcium level before and after drug application for the neonates (C) and the adult (D). Relative changes in the resting axonal level measured 10-30 min after drug application. *, the drug data are significantly different from control (P <0.05). Temperature 22°C

Figure 10 (A and B) summarizes effects of the various drug treatments on the posttetanus calcium decay according to two age groups, neonate (P5-P8) and adult (P60). In both age groups, pharmacological manipulations used to elevate axonal [Na]_i (metabolic blocker ouabain, Na ionophore monensin, Na channel inactivation inhibitor veratridine) or to inhibit the Na/Ca exchanger (bepridil, Li⁺ replacement of bath Na⁺) all result in a retardation of the posttetanus calcium clearance.

ELEVATION OF AXONAL Na CAUSES AN INCREASE IN RESTING CALCIUM INFLUX INTO AXONS. The preceding studies suggest a coupling between axonal [Na]_i and calcium clearance following nerve activity. In this section, we examined if axonal [Na]_i is also coupled to the resting axonal calcium level. As noted in the preceding text with regard to ouabain and monensin, application of these agents, in addition to retarding the posttetanus calcium decline, also caused a rise in the resting axonal [Ca²⁺]_i. What are the mechanisms underlying this rise in the resting calcium level? There are at least three possibilities. First, increasing axonal [Na]_i may cause a depolarization, presumably by increasing the sodium conductance of the membrane. This membrane depolarization could cause calcium channel activation, leading to calcium influx into axons. Second, increasing axonal [Na]_i triggers calcium release from internal stores (Mulkey and Zucker 1992). Third, increasing axonal [Na]_i causes influx of calcium through the Na/Ca exchanger when it is driven by intracellular Na accumulation to operate in the reversed mode. To raise axonal [Na]_i, we used ouabain and monensin (as used in preceding studies), as well as veratridine (which activates a resting Na influx by shifting the inactivation of the Na channels so that they are open at the resting potential).

View larger version (18K): [in this window] [in a new window]   Fig. 11. Axonal Na loading induces axonal calcium influx. A: bath application of Na-ionophore monensin (50 µM) causes an elevation of resting axonal [Ca2+]i. B: calcium elevation induced by Na loading is due to calcium influx. The bath solution was first switched to a calcium-free solution for 20 min before 5 µM veratridine was applied to induce axonal Na loading. Subsequent solutions had calcium either present or absent as indicated. P8 optic nerves. Temperature 22°C.

View larger version (19K): [in this window] [in a new window]   Fig. 12. Resting calcium influx induced by Na loading is not due to voltage-gated calcium channels. Simultaneous recordings of compound action potential (A) and resting axonal calcium signal (B) during various test solution applications. The resting calcium signal was obtained from the baseline calcium signal prior to the calcium transient evoked by each single action potential. Note that B only displays the resting calcium level and not the evoked calcium transient. The first test solution was one in which K was elevated by 15 mM. This was followed by a wash before 5 µM veratridine was applied. P9 optic nerve. Temperature 22°C.

EFFECT OF MITOCHONDRIAL BLOCKERS. Besides various calcium extrusion mechanisms located on the plasma membrane (such as Na/Ca exchanger, Ca-ATPase), cytosolic calcium buffers such as mitochondria may influence the decay kinetics of the evoked calcium response. For example, Colegrove et al. (2000) have recently shown that mitochondria in sympathetic neurons participate in buffering fairly modest elevation in cytoplasmic calcium induced by depolarizations. We therefore examined the role of mitochondria in the optic nerves by applying the mitochondrial blocker carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP, 1 µM) or carbonyl cyanide m-chlorophenylhydrazone (CCCP, 1 µM) (Colegrove et al. 2000; Fierro et al. 1998).

View larger version (20K): [in this window] [in a new window]   Fig. 13. Effects of the mitochondrial blocker carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP) on the resting and evoked axonal calcium signal. A and B: effects of bath application of FCCP on the amplitude of the compound action potential (A) and the corresponding resting axonal calcium fluorescence (B). C: effects of FCCP on the posttetanus calcium decay. The amplitude of the peak calcium response was normalized to compare the posttetanus calcium decay. The response was evoked with a tetanus (20 action potentials at 10 Hz) before, 15 min after FCCP, and 40 min after wash. D: application of FCCP in calcium-free bath solutions (0 calcium plus 1.0 mM EGTA) elicited no elevation in the resting axonal calcium fluorescence. P10 nerve. Temperature 22°C.

EFFECTS OF POSTTETANUS CALCIUM ELEVATION ON EXCITABILITY. Our analysis of axonal calcium reveals a prolonged phase of calcium elevation following a long tetanus (Fig. 6). An interesting issue is whether this protracted posttetanus calcium elevation has any physiological consequences on nerve excitability. We therefore measured posttetanus recovery of compound action potentials with or without calcium in the bath (Fig. 14). Mg²⁺ was added to replace the omitted calcium to minimize effects on the nerve excitability due to screening of surface charges. To produce significant posttetanus calcium elevation, we used a 5-min tetanus at 5 Hz. The amplitude of the compound action potentials was reduced at the end of the tetanus, to 49 and 39% of the control values for Ca-free and normal solutions, respectively (Fig. 14). The posttetanus recovery of the compound action potential is calcium dependent. In the case where calcium is present, the recovery occurred in an initial fast phase (~1 min) followed by a slow one that took ~20 min. In calcium-free bath solutions (i.e., no activity-dependent axonal calcium elevation), the two recovery phases are still present. However, fully recovery to the pretetanus level is achieved faster (~10 min). This suggests that the posttetanus axonal calcium elevation demonstrated in our calcium analysis modulates posttetanus excitability.

View larger version (25K): [in this window] [in a new window]   Fig. 14. Calcium-dependent posttetanus excitability changes in optic nerves. Plot of peak amplitude of compound action potential vs. time before and after tetanus. Both optic nerves from each mouse were dissected with 1 nerve tetanized in normal solution and the contralateral nerve tetanized in calcium-free Ringer solution (0 calcium plus 1 mM EGTA). Each tetanus is 5 min/5 Hz. Only the pretetanus baseline and the posttetanus response are shown. Note that 2 mM Mg2+ was added to the calcium free solution to keep the total divalent cation concentration constant in the calcium-free and the normal solutions. Averaged results from P6 mice (n = 5). Temperature 22°C.

DOES INTRODUCTION OF CALCIUM DYES DISTORT AXONAL CALCIUM SIGNALING? One potential source of uncertainty in this study is that the calcium dyes introduced in the axons might distort calcium signaling. We addressed this issue by using different concentrations of calcium dye in the loading, and by using low-affinity calcium dyes.

View larger version (21K): [in this window] [in a new window]   Fig. 15. Control experiments to examine possible distortion of calcium data by calcium dyes. A: control experiments with low-affinity calcium dyes. Tetanus-dependent retardation of posttetanus calcium decay demonstrated with the low-affinity calcium indicator Magnesium Green (Kd = 6000 nM). Superimposed axonal calcium responses from 20-, 100- and 300-action potential trains (10 Hz). Inset: normalized posttetanus calcium decays. B: monensin-induced retardation of posttetanus decay demonstrated with low-affinity calcium indicator Oregon Green 488 BAPTA-5N (Kd = 20,000 nM). The superimposed calcium response traces were normalized with respect to the calcium level at the start of the decay (10 action potentials/5 Hz). 10-wk optic nerve in A and P5 optic nerve in B. Temperature 22°C. C: comparison of the time course of evoked calcium transients with different dye concentrations in the loading pipette. Optic nerves were loaded either with 8 mM or 250 µM of dextran-conjugated Oregon Green in the pipette for 3 and 20 h, respectively. The evoked calcium transients (20 action potentials at 10 Hz) were normalized with respect to the peak amplitude so that the time course can be compared. P22 nerve for the 250 µM loading and P27 nerve for the 8 mM loading.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this paper, we reported a method for selective staining of axonal populations in the mouse optic nerves with calcium dyes and argued, based on the use of a dextran-conjugated dye that is known to be impermeable to gap junctions, and on the lack of an adenosine-induced calcium response (the hallmark of glia), that we have indeed successfully achieved selective axonal staining. We then applied this method to examine several basic properties of calcium clearance mechanisms in these axons and obtained two key observations regarding the clearance mechanisms. The first observation is tetanus-dependent calcium decay. A tetanus that elevates axonal calcium to a higher level leads to a slower calcium clearance. Further, even from a similar calcium level, the posttetanus calcium clearance is slower if it is preceded by a larger number of action potentials. The second observation is that inducing axonal Na loading with pharmacological agents retards posttetanus calcium clearance as well as induces a resting calcium influx into axons. We believe that our observations can be explained by evoking the Na/Ca exchanger as a key player in calcium homeostasis in the axons of the mouse optic nerves; the coupling of calcium homeostasis to axonal [Na]_i simply reflects the dependence of the operation of the exchanger on the Na gradient across the axolemma. Our postulated role for the Na/Ca exchanger makes excellent functional sense given the intense staining of the exchanger in the axons of the mammalian optic nerves (Steffensen et al. 1997).

Determinants of the shape of the posttetanus calcium decline

A key observation in this paper is that the posttetanus calcium decline is multi-exponential, and its shape is dependent on the calcium level and the action potential numbers during the tetanus. Posttetanus slowing of the calcium decline is observed with both high and low affinity calcium indicators, suggesting that it is not related to dye saturation artifacts. Bi-phasic calcium decline following a stimulus-induced calcium load has been observed in various neuronal preparations, including terminals of pyramidal cells of rat neocortex (Koester and Sakmann 2000), mouse cerebellar Purkinje cells (Maeda et al. 1999), and presynaptic terminals of cerebellar granule cells (Regehr 1997).

What determines the shape of the calcium decline after a calcium load? In general, uptake by mitochrondria, binding to endogenous axonal calcium buffers, extrusion through axolemmal Ca-ATPase and Na/Ca exchanger will jointly determine the shape of the posttetanus calcium decline. As discussed by Koester and Sakmann (2000), several factors may produce multi-exponential calcium decay. One is the existence of multiple species of endogenous calcium buffers. For example, Maeda et al. (1999) suggested that the bi-phasic calcium decline in Purkinje neurons can be explained by the presence of two endogenous buffers, one with high affinity and the other with low affinity. In their model, raising the calcium level to different levels by increasing the intensity of the tetanus will lead to various degree of saturation of the endogenous buffers, producing a calcium decline whose shape is tetanus-dependent. In the optic nerve axons, we found that increasing the tetanus duration makes the bi-phasic calcium decline more prominent, particularly after 4-AP treatment, which dramatically increases the calcium influx per action potential. The complex calcium clearance time course in the mouse optic nerve axons could be explained by the saturation of several distinct as-yet-unidentified endogenous calcium buffer species. In neuronal systems, strong candidates for endogenous buffers include the calcium-binding proteins calbindin-D and parvalbumin. The nature of endogenous buffers in the axons of the mammalian optic nerves has not been explored.

Another factor that can influence the shape of the calcium decline at the site of influx is diffusion along the axoplasm to regions of low calcium. Such diffusion can act as spatial buffering to restore calcium level at the site of entry. After an axon has been fully myelinated, the tiny nodal gap (~1 µm) is presumably the site of local calcium influx. The two long internodes flanking the node might act as a large "low-calcium" sink to spatially dissipate a high calcium increase at the node. The diffusion of calcium ions away from the node to the internode, coupled with calcium extrusion at the node, may produce a multi-exponential calcium decline. However, this spatial diffusion is unlikely to explain the multi-exponential calcium decay seen also in premyelinated axons (P5-P7) where the calcium influx is likely to be uniform over the entire length of the axon.

Besides endogenous buffers and axial diffusion of calcium, saturation of calcium extrusion mechanisms might also produce tetanus-dependent calcium decay (Blaustein 1988). Of the two extrusion pathways, Ca-ATPase and Na/Ca exchanger, which one might produce tetanus-dependent calcium decline? Regehr (1997) examined this issue with computer modeling of calcium decline in the granule cell terminals. He found that Ca-ATPase extrusion does not produce tetanus-dependent calcium decline. That is, increasing the calcium level in the tetanus by increasing the action potential numbers will not affect the shape of the posttetanus calcium decline if Ca-ATPase is the sole source for calcium extrusion. This is contrary to our observation. On the other hand, Regehr (1997) shows that extrusion through the Na/Ca exchanger will produce a tetanus-dependent calcium decline. According to Regher's model, following a calcium load, the Na/Ca exchanger will operate in the extrusion mode to extrude calcium, which will cause the initial fast decline in calcium. However, this extrusion also simultaneously brings in Na, which causes axonal [Na]_i to rise above the normal, pretetanus level. The Na/Ca exchanger will rapidly reach a new equilibrium, with a slightly higher axonal [Na]_i level and a higher axonal [Ca]_i level than the pretetanus level. As the calcium load is progressively increased (as happens after 4-AP treatment), the posttetanus axonal [Ca]_i will recover to a higher level. This predicts a persistent posttetanus calcium elevation whose amplitude should be proportional to the calcium level at the start of the decline as indeed observed in Fig. 6. With the participation of the Ca-ATPase in the extrusion process, this posttetanus, persistent calcium elevation should be slowly dissipated. Thus the behavior of the posttetanus calcium decline observed in this study fits qualitatively with Regehr's model (Regehr 1997), and argues for a role for Na/Ca exchanger in calcium extrusion following nerve activity. This conclusion is clearly in line with recent immunohistochemical data showing heavy staining of the Na/Ca exchangers in axons of the rat optic nerves (Steffensen et al. 1997).

Na accumulation retards posttetanus calcium clearance

Regher's model (Regehr 1997) cannot account for our observation that increasing the number of action potentials also can impede posttetanus decline, even under conditions where there is no increase in the calcium level from which the decline begins. We hypothesize that Na accumulation during tetanus retards posttetanus calcium clearance. This would explain why under certain conditions, the posttetanus decline is dependent on the number of action potentials during the train and not on the calcium level reached at the end of the train. This hypothesis is supported by the observation that elevation of axonal [Na]_i with pharmacological agents does indeed retard posttetanus calcium decay. What kind of calcium clearance mechanism might be sensitive to axonal [Na]_i? Extrusion via the Na/Ca exchanger would exhibit such a property. Calcium extrusion is coupled to the Na gradient, and elevation of internal Na is known to inhibit the exchanger (for review, see Stys et al. 1995). Alternatively, elevation of axonal Na might weaken the buffering capacity of endogenous buffers or inhibit calcium uptake into organelles, thus retarding the fall of free calcium in the axon. However, such an effect of Na has not been described. Finally, calcium clearance might be directly inhibited by the calcium elevation that accompanies pharmacological elevation of axonal Na. It is unclear what kind of calcium clearance mechanism might be inhibi