Journal of Neurophysiology

Probing the Endogenous Ca2+ Buffers at the Presynaptic Terminals of the Crayfish Neuromuscular Junction

Jen-Wei Lin, Qinghao Fu, Tariq Allana


Ca2+ indicators of varying affinity and mobility were pressure injected into the presynaptic axon of the inhibitor of the crayfish neuromuscular junction (NMJ). Fluorescence transients recorded at a 2-kHz resolution were used to probe physiological parameters governing the decay of fluorescence transients within 100 ms after an action potential (early decay). Blocking Ca2+ extrusion or Ca2+ sequestration processes did not significantly alter early decay, arguing against a role for either mechanism. Fluorescence transients recorded with low mobility or fixed indicators exhibited early decay similar to that recorded with indicators of comparable affinity but high mobility, suggesting that early decay was not due to the rate of Ca2+-indicator diffusion. The extent of early decay correlated closely with the affinity, but not mobility, of the Ca2+ sensitive dyes tested. These results implicate intrinsic buffers with slow Ca2+ binding kinetics as the most likely determinants of early decay. However, computer simulations showed that intrinsic buffers with a slow binding rate are unlikely to be the only ones present in the system because the slow kinetics would be unable to buffer incoming Ca2+ during an action potential and would result in momentary indicator saturation. In fact, experimental data show that the peak amplitude of an action potential activated Ca2+ transient is about 20% of the maximal fluorescence intensity activated by prolonged Ca2+ influx. We conclude that endogenous buffering at the crayfish NMJ includes both fast and slow components, the former being fast enough to compete with fast Ca2+ indicators, and the latter dictating the early decay.


Calcium influx activated by a presynaptic action potential initiates a series of steps leading to the release of neurotransmitter. Having passed through calcium channels, calcium ions diffuse rapidly and bind to their target proteins. Calcium sensors for transmitter release have a fast on-rate, and Ca2+ binding initiates vesicular fusion within a fraction of a millisecond (Bollmann et al. 2000; Llinás et al. 1981; Schneggenburger and Neher 2000). Binding of Ca2+ to signal transduction proteins activates multi-step cascades that result in responses with longer delays and durations. For some proteins, the main function of Ca2+ binding seems to be one of buffering (Neher 1998). The mobility and binding kinetics of these proteins determine the dynamics of free calcium ions in the “nanodomain” within which Ca2+ channels and vesicles colocalize (Augustine et al. 2003). Current imaging technology is of insufficient resolution to monitor [Ca2+]i on a nanometer/submillisecond scale. However, it would be possible to infer the behavior of Ca2+ if the rate constants and mobility of calcium binding proteins were known. Experimental measurements of these parameters have been scarce, with the exception of studies carried out in chromaffin cells (Xu et al. 1997; Zhou and Neher 1993), hair cells (Roberts 1993), and cerebellar Purkinje cells (Schmidt et al. 2003a). A potential source for such information is Ca2+ transients recorded at a high time resolution, which have been used to infer the time-course of presynaptic ICa (Sabatini and Regehr 1996; Sinha et al. 1997), the kinetics of facilitation (Atluri and Regehr 1996), and the binding rates of endogenous buffers (Sinha et al. 1997). One way of expanding the use of this technique to probe the kinetics of intrinsic buffers would be to examine fluorescence transients recorded with low mobility indicators. The crayfish neuromuscular junction is an ideal preparation in which to implement this method. First, because there is no complex circuitry, drastic pharmacological manipulation can be carried out without causing uncontrolled network activity. Second, the accessibility of this preparation to pressure injection with sharp electrodes enables the introduction of dyes that cannot be delivered in the AM form. Third, pressure injection avoids uncertainties associated with washout of endogenous buffer that typically occurs with whole cell patch recordings (Edmonds et al. 2000; Neves et al. 2001; Zhou and Neher 1993). Finally, the crayfish neuromuscular junction is one of the best characterized model systems for the study of synaptic transmission (Zucker and Regehr 2002). It was one of the first preparations in which quantitative Ca2+ imaging provided estimates for the intrinsic Ca2+ buffer binding ratio and extrusion rate (Tank et al. 1995). Ongoing accumulation of data from imaging and electrophysiological studies has provided the basis for several generations of detailed mathematical models of presynaptic calcium dynamics (Matveev et al. 2002; Tang et al. 2000; Winslow et al. 1994; Yamada and Zucker 1992). However, the actual mobilities and rate constants of intrinsic buffers in this preparation are still a matter for speculation. In this report, we analyze the time-course of fluorescence transients recorded with calcium sensitive dyes with various affinities and mobilities. Our analyses provide new insights to the kinetic properties of intrinsic buffers.


Preparation and electrophysiology

Crayfish, Procambarus clarkii, were obtained from Carolina Biological (Burlington, NC). Animals, 4 cm length from head to tail, were maintained at 23°C, and experiments were performed at the same temperature. The opener muscle of the first walking leg was used for all experiments. A presynaptic electrode penetrated the primary branch point of the inhibitory axon (inhibitor) to record action potential and inject dye. The branch point was about 100–300 μm from the terminals on a central muscle fiber from which fluorescence transients were measured. A suction electrode was used to stimulate the inhibitor. One postsynaptic electrode, 5 MΩ with 3 M KCl, penetrated a muscle fiber to monitor inhibitory postsynaptic potentials (IPSPs).

Control saline contained (in mM) 195 NaCl, 5.4 KCl, 13.5 CaCl2, 2.6 MgCl2, and 10 HEPES, titrated to pH 7.4 by NaOH. When tetraethylammonium (TEA) chloride was added to the control saline, an equal amount of NaCl was removed. Unless indicated otherwise, all chemicals were purchased from Sigma.

Photometric measurement of calcium transients

The inhibitory axon was penetrated at the primary branch point by an electrode containing 2–5 mM of calcium-sensitive dye. Dyes were dissolved in 400 mM potassium methanesulfonate and 20 mM KHEPES (pH = 7.4) and resulted in a final electrode resistance of 20–50 MΩ. Indicators were pressure injected in most experiments. The injection typically lasted for 10 min, with 100-ms pressure pulses at 70 psi and delivered at a repeating rate of 0.1 Hz. Stained terminals were visible within minutes after the injection started. Injection was stopped when varicosities close to the injection site were bright. Experiments commenced after dye distribution had equilibrated, typically within 30 min after dye injection. The preparation remained stable for >4 h after a good injection.

The final concentration of an injected dye was estimated to be 200–300 μM. Dye concentration was estimated by visually comparing the fluorescence intensity of an injected axon with a dye-filled capillary tube of similar diameter. The capillary tube was filled with known concentrations of dye in a calibration solution prepared from Calcium Calibration Buffer Kit with 1 mM Mg2+ (Molecular Probes C-3721). The calibration solution, with a free [Ca2+]i of 100 nM, contained (mM) 4 CaEGTA, 6 K2EGTA, 100 KCl, 30 MOPS, and 1 MgCl2, pH = 7.2. The concentration of indicator in the axon was determined when its fluorescent intensity was bracketed by two calibration solutions differing by 50 μM in indicator concentration. Because the presynaptic axon is sensitive to the volume and ionic strength of the pressure-injected solutions, our experience was that the axon was typically unhealthy if the injected dye concentration was >10% of that in the pipette. This constraint automatically prevented us from overloading the axon with Ca2+ buffers and resulted in consistent concentrations of injected indicators. When multiple compounds were pressure injected, we assumed that the final concentration ratio in the cytoplasm would be identical to that in the injection pipettes. Therefore with the estimated final concentration of a fluorescent dyes, the concentration of coinjected compound would also be known.

Fluorescence signal measurement of Ca2+ transients in this preparation has been described before (Vyshedskiy et al. 2000). Briefly, a photomultiplier tube (HC124-06, Hamamatsu) was used to record fluorescence transients on an upright microscope (Zeiss Axioskop) with a ×40 or ×60 water immersion lens. The output of the photomultiplier tube was filtered at fc = 2 kHz and digitized at 20 kHz. In some experiments, a photodiode (Hammamatsu S5973) was used as the light sensor. In this case, a single channel headstage (Axon CV-5-100GU) attached to a GenClamp 500B was used to measure the photocurrent. This arrangement, when tested with a LED and filtered at 10 kHz, was able to follow a step increase in light intensity with a 10–90% rise time of 100 μs (data not shown). A 100-W tungsten lamp, powered by a stabilized power supply (Kepco, ATE 15-15DM), or a 150-W Xenon lamp (Optiquip 1600 Power supply with 770 Lamphouse), was used to illuminate the preparation. Illumination was gated by a shutter (Uniblitz, Vincent Associates) with a typical duration of 600 ms and repeated at 0.2 Hz. The field of illumination was restricted to an area of 20 ∼ 50 μm diam, which typically encompassed approximately five varicosities on the upper surface of a central muscle fiber. The fluorescent dyes used in this report were Magnesium orange (MgOrg), Calcium orange (CaOrg), Rhod-2, dextran-coupled Oregon green 488 BAPTA-1-70 kDa (Dex-OgnGr), Fura-2, and Fura2 coupled with hydrophobic tail (FFP18). FFP18 and Fura-2 were purchased from Teflabs (Austin, TX); all other indicators were from Molecular Probes (Eugene, OR).

Fluorescence transients are presented as ΔF/F = [F(t) − Frest)/Frest × 100%, where Frest represents the fluorescence intensity of stained varicosities in the absence of activity. The averaged background fluorescence level, in regions without stained structures, was about 59 ± 16% (n = 6) of Frest for MgOrg. This background fluorescence has been subtracted in the averaged MgOrg transient.

Fluorescence transients recorded from individual preparations were typically the average of 50–100 trials. Fluorescence transients from different preparations were aligned according to the rising phase of presynaptic action potentials before averages across preparations were taken. All the statistical values represent mean ± SE, and statistical significance was carried out with Student’s t-test.

Computer simulations

Calcium calculator, versions 4.97 to 5.0.3, was used for simulation of buffer-Ca2+ interactions in three dimensions (Matveev et al. 2002). Some computation was performed on a desktop PC, whereas most calculations were performed on an SGI Origin2000 Cluster or Intel Pentium III Linux Cluster maintained by Boston University Scientific Computing and Visualization group.

The geometry of the terminals was identical to that published previously—a cube of 0.8 × 0.8 × 1 μm with four Ca2+ channels located at one corner of the cube (Matveev et al. 2002; Tang et al. 2000) (Table 1). The channels were arranged in a square grid, 30 nm from the edge and 60 nm apart (Matveev et al. 2002; Tang et al. 2000). This arrangement is equivalent to one-quarter of a fourfold symmetric active zone (Govind et al. 1995). The simulated space was divided into a grid of 34 × 34 × 40 with a stretch factor of 1.07. The mobilities of Ca2+ (0.2 μm2/ms) and indicators (0.118 μm2/ms) were identical to those typically used in similar simulations in this and other preparations (Meinrenken et al. 2002; Tang et al. 2000). On-rates of the MgOrg and Fura-2 are known to be fast (Hollingworth et al. 1992; Xu et al. 1997) and were assumed to be 0.27 μM−1 · ms−1. The total binding ratio of endogenous buffer (600) and extrusion rate of Ca2+ (0.05 μmol/ms) have been measured previously (Tank et al. 1995). The kinetic properties of endogenous buffer are unknown; we initially assumed the presence of a single class of endogenous buffer with a fast on-rate of 0.5 μM−1 · ms−1 (Tang et al. 2000; Xu et al. 1997) (see Table 1 for the range of Kd investigated). The magnitude of Ca2+ influx activated by broad action potentials has been shown to be about 10 times larger than that activated by narrow action potentials (Vyshedskiy and Lin 2000). By adopting the amplitude of single channel current used in previous simulation studies (Tang et al. 2000), we prolonged the influx during the action potential and the tail current to 5 and 3.5 ms, respectively. These changes approximate the time-course of broadened action potentials and result in a 10-fold increase in total influx. To evaluate the results of simulation, we compared the time course of the spatially averaged concentration of Ca2+-bound buffer (ACaB) with experimental data.

View this table:

Kinetic parameters used in simulation


Calcium transients

Figure 1 shows typical recordings of fluorescence transients obtained with MgOrg (A) and Rhod-2 (B). The top traces in Fig. 1, A and B, are Ca2+ transients recorded before (—) and after ( · · · ) K+ channels were blocked with 20 mM TEA and 1 mM 4-AP. Presynaptic action potentials recorded simultaneously are shown in the bottom panels. The insets show that fluorescence signals rise during the course of broad action potentials and exhibit a monophasic trajectory. The Ca2+ transients activated by broad action potentials are about 10 times larger than those activated by individual action potentials recorded in control saline (Vyshedskiy and Lin 2000). The decay of fluorescence transients also follows a near monophasic trajectory over the course of 100 ms. These characteristics were observed consistently, in >100 preparations. Because it is difficult to resolve the time-course of Ca2+ transients activated by narrow potentials, all data analyzed here were collected with broad action potentials. Previous studies have shown that the magnitude and decay time course of fluorescence transients activated by a single broad action potential are similar to those of transients evoked by a burst of 10 action potentials at 100 Hz (Vyshedskiy and Lin 2000; Vyshedskiy et al. 2000).

FIG. 1.

Time-course of fluorescence transients recorded with low and high affinity Ca2+ indicators. A: fluorescence transients recorded from MgOrg before (—) and after ( · · · ) K+ channels were blocked by 20 mM tetraethylammonium (TEA) and 1 mM 4-AP. B: fluorescence transients recorded from Rhod-2 before (—) and after ( · · · ) K+ channels are blocked. Insets in A and B show the rising phase of the fluorescence transient activated by broad action potentials. Fluorescence transient traces are averages of 100 trials. C and D: averaged fluorescence transients with SE envelopes recorded from MgOrg (C, n = 15) and Rhod-2 (D, n = 5). Arrow in C identifies the time-point at which early decay is measured. E: decay of a Ca2+ transient correlates negatively with dissociation constants of the Ca2+ indicator. Degree of fluorescence transient decay was measured at 100 ms after an action potential and is expressed as the percentage of its peak amplitude. Continuous line is drawn by hand. Affinities of the Ca2+ indicators used, as listed in the Molecular Probes catalog, are (in μM) Fura-2, (0.145); FFP18, (0.15); Calcium orange (0.185); Oregon green-BAPTA (0.225); Rhod-2 (0.57); and magnesium orange (12). Arrow and arrowhead identify averaged results form dextran-coupled Oregon green-BAPTA and FFP18, respectively. Number of preparations used to calculate individual data points are Fura-2 (9); FFP18 (10); Calcium orange (6); Oregon green-BAPTA (7); Rhod-2 (5); and Magnesium orange (15).

The averaged MgOrg transient has a peak amplitude of 6% (Fig. 1C), which is significantly lower than the maximal level of fluorescence intensity (33.8 ± 10.9%; n = 5) activated by a 5-s presynaptic depolarization to −10 mV. Therefore the Ca2+ influx activated by a typical broad action potential does not saturate MgOrg.

Figure 1, C and D, shows the averaged recordings, with SE envelopes, of fluorescence transients recorded with MgOrg (n = 15) and Rhod-2 (n = 5), respectively. The rate of fluorescence transient decay is typically rapid with low affinity dyes. To quantitate the extent of decay in fluorescence transient using different indicators, we normalize the amplitude measured at 100 ms after presynaptic action potential by its peak (early decay; Fig. 1C, arrow). This simple quantitation allows us to incorporate a large data set that includes traces with recording durations of 100 ms. Figure 1E shows that, when similar concentrations of indicators are injected, the fraction of remaining fluorescence at 100 ms correlates closely with the affinity of the indicator.

Computer simulation analysis of early decay

We explored the process underlying early decay initially by performing simulations of Ca2+-buffer interactions in three dimensions. Assuming the presence of a single class of mobile endogenous buffer (Kd = 1 μM) with the same mobility and fast on-rate as Fura-2, the spatially ACaB, i.e., MgOrg-Ca, exhibits a small initial spike followed by a plateau (Fig. 2A, —). The initial spike is due to equilibration of endogenous buffer and Ca2+. The plateau is due to the slow extrusion process that is unable to significantly reduce [Ca2+]i within 100 ms. Varying the affinity of endogenous buffer from 0.5 to 100 μM altered the magnitude but not the slope of the plateau.

FIG. 2.

Kinetic parameters that control early decay. A: time-course of spatially averaged MgOrg-Ca concentration calculated by assuming: a mobile (black traces) or fixed (shaded traces) endogenous buffer. With mobile endogenous buffer, Kd = 1 μM, averaged concentration of Ca2+-bound buffer (ACaB) exhibits an initial spike followed by a plateau (—). It is necessary to increase the Ca2+ extrusion rate by 100-fold to create a decay qualitatively similar to early decay (- - -). Also included for comparison is the experimentally obtained MgOrg transient shown in Fig. 1C ( · · · ). If the endogenous buffer is fixed, with a Kd of 1 μM, the time-course of ACaB exhibits a fast initial spike followed by a plateau (shaded —). Setting the mobility of the Ca2+ indicator to 0 slows down the decay of the initial spike (shaded - - -), but not enough to approximate experimental data. Peak amplitudes of all traces have been scaled to the same height for comparison. Actual peak concentrations are (μM) 8.1 (—), 4.6 (- - -), 16.4 (—shaded), and 6.0 (- - - shaded). B: time-course of [Ca2+]i calculated for [0.09,0.09,0.01] (—), [0.09,0.09,0.10] ( · · · ), and [0.60,0.60,0.75] (- - -), where the numbers in the brackets represent x, y, and z coordinates (in μm). The first 2 locations are, respectively, 10 and 100 nm above a Ca2+ channel with x-y coordinates of [0.09,0.09]. Results calculated with mobile endogenous buffer are shown as thin black traces, whereas results calculated with fixed endogenous buffer are shown as thick gray traces. Step increase in [Ca2+]i shortly before the influx ends at 20 ms is due to the tail-current component of Ca2+ influx. C: effects of indicator mobility on ACaB. The indicator chosen in this case has an affinity comparable to Fura-2 with mobility set to 0.118 (—), 0.01 ( · · · ), and 0 (- - -) μm2/ms. In the presence of fixed endogenous buffer, Kd = 1 μM, decreasing indicator mobility slows down the rising phase of ACaB.

This simulation result is qualitatively different from experimental recordings of MgOrg transients, which exhibit a continuous decay (Fig. 2A, · · · ). One way to approximate early decay is to increase the extrusion rate. For example, while maintaining the same kinetic parameters for the endogenous buffer, increasing the extrusion rate by 100-fold results in a decline in ACaB comparable with that of early decay (Fig. 2A, - - -).

If endogenous buffer is fixed, ACaB takes on the shape of an initial spike, which decays within 10 ms and is followed by a plateau (Fig. 2A, gray ). The initial spike is larger than that calculated with mobile endogenous buffer because fixed buffer retards the movement of Ca2+ and causes extremely high local [Ca2+]i and a high ACaB signal during Ca2+ influx (Roberts 1994). The diffusion of Ca2+ away from Ca2+ channels after Ca2+ influx ends contributes to the decay of the initial spike. The plateau is again due to the slow extrusion rate. Again, varying the Kd of fixed endogenous buffer, Kd = 0.5–100 μM, did not change the basic shape of ACaB decay, although the relative heights of the spike and plateau were altered. In the example shown in Fig. 2A (gray ), the affinity was set to 1 μM to allow the relative level of the plateau to be comparable with the final level of early decay. To evaluate the role of indicator mobility on early decay, we repeated the simulation by setting indicator mobility to zero (Fig. 2A, gray - - -). This modification slows the decay of the initial spike but not enough to approximate early decay. Nevertheless, it is clear that the mobility of a Ca2+ indicator may also shape early decay.

Although the mobility of endogenous buffer has little impact on the plateau phase of ACaB, it does influence the distribution of Ca2+ in space. When a high mobility is assigned to the endogenous buffer, Ca2+ and buffers reach spatial equilibration within 10 ms after Ca2+ influx ends. Figure 2B shows the rapid spatial equilibration by showing [Ca2+]i at 10 (—), 100 ( · · · ), and 978 nm (- - -) from one of the four Ca2+ channels. The second stepwise increase in [Ca2+]i at both 10 and 100 nm, shortly before 20 ms, is caused by the tail current component of Ca2+ influx. In contrast, spatial equilibration of [Ca2+]i continues for 100 ms if the endogenous buffer is fixed (Fig. 2B, shaded traces). This difference in spatial distribution is not detected by MgOrg during early decay because of the spatial averaging nature of ACaB and because of the low affinity of MgOrg for Ca2+. In other words, local differences in [Ca2+]i within the submicromolar range cannot be discriminated by spatially averaged MgOrg signals. For this reason, when we approximated the diffusion of Ca2+ and indicators into the axons that connect varicosities by increasing the terminal volume, the diffusion and dilution of Ca2+-indicator into the larger volume were not reflected in the period of early decay, which remained essentially flat.

We also performed simulations to investigate early decay resulting from a high affinity indicator, Kd = 0.145 μM. To compare results with data obtained from low mobility indicators, the impact of the mobility of high affinity indicators on early decay was also examined. Only results simulating fixed endogenous buffer are shown, because the impact of indicator mobility will be minimal if the dominating endogenous buffer is mobile. Assuming an affinity of 1 μM for the endogenous buffer, Fig. 2C shows that ACaB calculated for a mobile high affinity indicator gives rise to a shallow initial spike followed by a plateau (—). Decreasing indicator mobility to one-tenth ( · · · ) or to zero (- - -) eliminates the initial spike and results in a slow rise in ACaB during the period of early decay. Therefore decreasing the mobility of high affinity calcium indicators slows the rising phase and eliminates early decay of ACaB. Because an exhaustive exploration of parameters related to buffer/Ca2+ mobility is beyond the scope of this report, we did not scan related parameters systematically.

These simulations suggest that it is not possible to reconstruct the early decay observed experimentally by adopting published extrusion rates and assuming a fast on-rate for endogenous buffer. Adjusting the affinity and mobility of endogenous buffer failed to bend the plateau enough to approximate early decay. Two manipulations shown here to be hypothetically capable of creating continuous decay within 100 ms of an action potential are 1) dramatically increasing the extrusion rate and 2) decreasing indicator mobility. In the rest of this study, we investigate whether known Ca2+ extrusion and sequestration mechanisms could operate at a rate rapid enough to account for early decay. Furthermore, we explore the potential impact of indicator mobility on early decay.

Principal extrusion processes contribute minimally to early decay

Because Na+/Ca2+ exchange is known to be capable of extruding Ca2+ at a high rate (Philipson and Nicoll 2000), we first examined its effects on early decay by completely substituting Na+ with Li+. Despite the complete removal of Na+, action potentials are still supported under these conditions (Fig. 3A, bottom), although with a noticeably smaller amplitude. This is presumably because Na+ channels have high permeability for Li+ (Hille 1975). There is a decrease in the amplitude of the fluorescence transient to 65% (Fig. 3A, middle), but the decay time course is only slightly altered. IPSP amplitude, on the other hand, is nearly identical, despite a substantial decrease in total Ca2+ influx (Fig. 3A, top). This was a consistent finding; IPSP amplitude in Li+ was 105 ± 16% of control (n = 4). The issue of IPSP amplitude was not further pursued.

FIG. 3.

Replacing Na+ by Li+ has a small effect on early decay. A: recordings from a representative experiment where complete substitution of extracellular Na+ (195 mM) by Li+, to inhibit Na+/Ca2+ exchange, reduces the amplitude of both the presynaptic action potential (bottom) and Ca2+ influx (middle). Inhibitory postsynaptic potential (IPSP) amplitude (top) is not significantly affected. B: Ca2+ transients averaged from 7 preparations before (—) and after (—) Li+ substitution. A scaled fluorescence transient recorded in Li+ saline ( · · · ) is also shown to illustrate that its early decay is only slightly affected by the block of Na+/Ca2+ exchange. See Table 2 for statistical values.

To show the consistency among preparations, averaged results from seven preparations are summarized in Fig. 3B. Here, MgOrg transients recorded before (—) and after (- - -) Li+ substitution have been superimposed. Also included is the scaled Ca2+ transient recorded in Li+ ( · · · ) to show that its decay rate is only slightly slower than that of the control Ca2+ transient. Averaged values of early decay before (29.1 ± 4.1%) and after (43.1 ± 6.0%) Li+ substitution are not statistically different (Table 2).

View this table:

Effects of Ca2+ extrusion blockers on early decay

The effects of additional blockers of Ca2+ extrusion/sequestration processes were also examined. The efficacy of these blockers was tested by examining their effects on post-tetanic potentiation (PTP) transmitter release or Ca2+ transients (see supplementary information III1 ). The averaged fluorescence transient recorded in the presence of both plasma membrane (KB R7943, 20 μM) and mitochondrial (CGP37157 25 μM) Na+/Ca2+ exchange blockers (Fig. 4A, · · ·; n = 5) exhibits a decay time-course identical to that recorded before the drugs were applied (—). Addition of an endoplasmic reticulum (ER) CaATPase blocker (thapsigargin, 2 μM), in the presence of the two above-mentioned blockers also failed to alter the time-course of the fluorescence transient (Fig. 4A, —; n = 4). A second blocker for ER CaATPase, cyclopiazonic acid (CPA), was tested separately and yielded the same results (see supplementary information III). The averaged MgOrg transient recorded after intra-axonal injection of plasma membrane CaATPase blocker (C28R2 at ∼10 μM; Fig. 4B, · · ·; n = 3) is essentially identical to that recorded under control conditions (—). Finally, injection of ruthenium red, known to block mitochondrial uniporter-mediated Ca2+ uptake (50 μM), also fails to change the decay phase of the fluorescence transient (Fig. 4C, · · ·; n = 3). Statistical results of these blockers on early decay are listed in Table 2. In summary, with the exception of Li+, early decay is not affected by the Ca2+ extrusion/sequestration blockers tested here.

FIG. 4.

Blockers of the Ca2+ extrusion processes do not affect early decay. A: averaged Ca2+ transient (n = 5) recorded in the presence of both KB R7943 at 25 μM (plasma membrane Na+/Ca2+ exchange blocker) and CGP 37157 at 2 μM (mitochondrial Na+/Ca2+ exchange blocker) ( · · · ) exhibits early decay identical to that recorded before the drugs were added (—). In the presence of the 2 blockers, addition of thapsigargin at 2 μM (—) (endoplasmic reticulum Ca2+ pump blocker) also fails to alter the decay time-course. Peak amplitudes of fluorescence transients have been scaled to the same height for comparison. Actual amplitudes, after background correction, were as follows: control, 6.1%; KB R7943 + CGP 37157, 5.5%; +thapsigargin, 5.1%. Only preparations exposed to all 3 blockers were used in the averages (n = 5). B: intra-axonal injection of peptide C28R2 (∼10 μM, a plasma membrane Ca2+ ATPase blocker), does not significantly alter the decay of the fluorescence transient. The Ca2+ transient measured after peptide injection ( · · · ) represents the average of 3 preparations. Since the peptide was injected simultaneously with MgOrg, there was no control fluorescence transient for individual preparations. Instead, the averaged Ca2+ transient from MgOrg shown in Fig. 1C (—) is displayed for comparison. The peak amplitude of the MgOrg transient after C28R2 injection was 6.9%, and was 6% for the control trace. C: blocking the mitochondrial Ca2+ uptake uniporter with ruthenium red (50 μM) does not alter early decay ( · · · ) monitored with MgOrg, averaged from 3 preparations. Also included for comparison is the averaged Ca2+ transient from MgOrg shown in Fig. 1C (—). The peak amplitude of the MgOrg transient after ruthenium red injection was 9.1%.

Early decay is independent of dye mobility

In an effort to evaluate the impact of dye mobility on early decay, we measured fluorescence transients obtained with a dextran-coupled dye, Dex-OgnGr. Its high molecular weight (70 kDa) and anticipated low mobility are consistent with the observation that it took ≤1 h, as opposed to a matter of minutes in the case of low molecular weight indicators, for terminals near the injection sites to become sufficiently bright for experiments. However, the early decay recorded with Dex-OgnGr is similar to those recorded using Rhod-2 or CaOrg. Figure 5A shows the averaged fluorescence transient recorded with Dex-OgnG from seven preparations ( · · · ), superimposed on the averaged transients from Rhod-2 (—, n = 5) and CaOrg (—, n = 6). All three fluorescence transients exhibit a similar rise as well as decay in their time course. The averaged early decay of Dex-OgnGr transients in Fig. 1E (arrow) follows the same general trend as those of the other highly mobile dyes, suggesting that early decay is dictated mainly by affinity and does not correlate with the mobility of Ca2+ indicators.

FIG. 5.

Ca2+ transients measured with low mobility dye are similar to those obtained using dyes of comparable affinity but high mobility. A: averaged fluorescence transient (n = 7) measured from Oregon green 488 BAPTA-1 (Kd = 0.225 μM) coupled to 70 kDa dextran ( · · · ) exhibits a similar time-course to those measured from CaOrg (- - -, Kd = 0.185 μM) or Rhod-2(—, Kd = 0.57 μM). The CaOrg transient is barely visible because it is largely masked by the noisier Oregon green trace. The peaks of Ca2+ transients have been scaled to the same height for comparison; their actual amplitudes are as follows: CaOrg, 23%; Oregon green 488 BAPTA-1, 5%; Rhod-2, 16%. B: early decay of the membrane bound FFP18 transient (—, n = 10) is similar to that recorded from diffusible Fura-2 (- - -, n = 9). Peaks of the Ca2+ transients have been scaled to the same heights for comparison. Actual amplitudes are 1.8% for Fura-2 and 1.4% for FFP18.

Although 70 kDa dextran is a large molecule compared with a typical Ca2+ indicator, of 800-1,500 Da, it is not completely immobile. We therefore tested FFP18, namely Fura-2 coupled to a hydrophobic carbon chain (Etter et al. 1994) (see supplementary information I for evidence indicating that FFP18 is indeed membrane-bound after injection; supplementary information II shows that FFP18 injection has no adverse effect on synaptic function1). Figure 5B shows that the early decay of the averaged FFP18 transient (n = 10, —) is similar to that recorded with diffusible Fura-2 (n = 9, - - -; see Fig. 1E, arrowhead, for the statistical summary of early decay measured from FFP18). In addition, the rising phase of FFP18 appears to exhibit a distinct slow component that is not observed in the Fura-2 transient. This difference in the rising phase, although indicative of fixed endogenous buffer as suggested in Fig. 2C, was not pursued further. Assuming FFP18 is predominately anchored to membranous compartments, and is thus truly immobile, our results suggest that the mobility of a given Ca2+ sensitive dye has no impact on early decay.

Dynamics of local [Ca2+]i in the presence of slow endogenous buffer

Because experimental data thus far suggest that Ca2+ extrusion/sequestration and indicator mobility have minimal impact on early decay, we next used computer simulation to examine whether a slow endogenous buffer could theoretically account for early decay. Figure 6A shows three examples of ACaB that closely approximate experimental data. These ACaBs were calculated on the basis of a single class of slow endogenous buffer, with Kd = 0.5 (—), 2 (- - -), or 20 (gray ) μM. The on-rate of the endogenous buffer was decreased by 1,000- to 10,000-fold from the original fast on-rate used in Fig. 2. Although it is obvious that multiple sets of buffer parameters can recreate early decay, local [Ca2+]i dynamics calculated from these parameter sets are essentially identical. Figure 6B shows [Ca2+]i at 10 (—), 100 ( · · · ), and 978 (- - -) nm from the Ca2+ cluster, estimated assuming endogenous buffer with a Kd of 0.5 (—gray) and 20 () μM. Intracellular [Ca2+] rises above 100 μM at both 10 and 100 nm from the Ca2+ cluster. Even at 978 nm, [Ca2+]i is as high as 10 μM at the end of Ca2+ influx. These high concentrations suggest that slow buffer alone cannot significantly buffer incoming Ca2+ during influx. Indeed, the peak amplitude of ACaB is ∼94 μM of a total MgOrg concentration of 200 μM. An indicator saturation of nearly 50% is significantly higher than our experimental estimate of ∼20% MgOrg binding at the peak of the Ca2+ transient. Therefore, although an endogenous buffer with slow binding kinetics can easily recreate early decay, the buffer alone leads to extremely high local [Ca2+]i and an unrealistically high level of indicator saturation. It would therefore be more reasonable to suggest the simultaneous presence of fast and slow endogenous buffers. The slow buffer would account for early decay, whereas the fast buffer would be capable of buffering incoming Ca2+ and competing with MgOrg.

FIG. 6.

Early decay approximated by a single class of endogenous buffer. A: 3 examples of ACaB that closely resemble the early decay observed experimentally. They were calculated assuming slow endogenous buffers with Kd of 0.5 (—), 2 (- - -), and 20 (—shaded) μM. The on-rates of these buffers are 0.00072, 0.0001, and 0.00001 μM−1 · ms−1, respectively. The buffers are assumed to be mobile, with mobility identical to that of Fura-2 (Table 1). Also shown for comparison is the averaged MgOrg transient ( · · · ) shown in Fig. 1C. B: local [Ca2+]i calculated in the presence of slow endogenous buffers with Kd of 0.5 (—shaded) and 20 μM (—). Here the time-courses of [Ca2+]i shown are for points 10 (—) and 100 ( · · · ) nm above a Ca2+ channel with the x-y coordinates of [0.09, 0.09] (μm). Also shown is [Ca2+]i at [0.6, 0.6, 0.75] (μm), at a distance of 978 nm from the Ca2+ channel (- - -). Local [Ca2+]i at the 3 locations calculated from the 2 slow buffers are identical. Also shown in this graph is the local [Ca2+]i calculated with fast endogenous buffer in combination with a 100-fold faster extrusion rate, so that ACaB exhibits a decay comparable to early decay, (○) 10, (□) 100, and (▵) 978 nm locations. See the trace (- - -) in Fig. 2A for ACaB. Although either a slow endogenous buffer alone or a fast buffer in combination with fast extrusion can approximate early decay, the local [Ca2+]i estimated in the presence of slow buffer is >10 times higher.

Also shown in Fig. 6B for comparison is local [Ca2+]i—10 (○), 100 (□), and 978 (▵) nm, calculated from a system where early decay is shaped by fast extrusion in the presence of fast endogenous buffer (Fig. 2A, —). This comparison shows that, although both fast extrusion and slow endogenous buffer can shape early decay correctly, local [Ca2+]i would differ by 10-fold depending the mechanisms involved.


Several observations reported here provide valuable and previously undefined insights to the properties of intrinsic calcium buffers. In the first place, we observed that the early decay of fluorescence transients was not affected by blockers of major Ca2+ extrusion/sequestration pathways. Furthermore, the magnitude of early decay correlated with the affinity but not mobility of indicators. The elimination of these mechanisms led to our suggestion that slow intrinsic buffers are the likely underlying process dictating early decay. Results from computer simulations suggest that the crayfish inhibitor should contain slow and fast buffers. A buffer slow enough to dictate early decay would not be able to absorb incoming Ca2+ during an action potential. As a result, even the low affinity indicators such as MgOrg would likely be saturated during a broad action potential. Experimental results suggest that MgOrg is far from being saturated, suggesting the need to propose a fast endogenous buffer capable of competing with MgOrg.

Significance of the time-course of Ca2+ transients

The photometric method used in this report measures changes in fluorescence intensity from the entire terminal volume: approximately five varicosities, each of which was 1–5 μm in diameter. The decay of a Ca2+ transient reflects the decrease in the number of calcium bound indicator molecules. It has been suggested that the decay of a Ca2+ transient reported by a low affinity indicator closely approximates the time course of spatially averaged [Ca2+]i in both small structures such as parallel fiber terminals (Sabatini and Regehr 1998) and large structures such as calyx of Held (Meinrenken et al. 2003). In small structures, e.g., synaptic terminals or dendritic spines, the distribution of free Ca2+ can reach spatial uniformity within 10 ms (Majewska et al. 2000; Sabatini and Regehr 1998). Therefore the time-course of early decay measured with MgOrg should mirror the decay of spatially uniform [Ca2+]i. The decay of [Ca2+]i, in turn, should be controlled mainly by extrusion, sequestration and buffering.

When high affinity indicators were used in this study, the degree of dye saturation was not quantitatively evaluated. As a result, the time-course of fluorescence transients recorded does not directly reflect that of intracellular Ca2+ transients. However, for the purpose of analyzing the dependence of early decay on indicator affinity, this is not an issue, provided that injected concentrations are similar for all indicators. In the case of dextran-coupled indicator, care was taken to ensure that the final concentration of injected indicator was comparable with that of highly mobile ones. The concentration of FFP18, however, is much more difficult to evaluate. This indicator partitions itself into two dimensional; membranous compartments and the dye concentration in two dimensions is difficult to determine. Nevertheless, given the injection condition used in this report, the total injected FFP18 concentration should be comparable with that of mobile dyes.

Processes underlying the early decay of fluorescence transients

The dynamics of intracellular Ca2+ within the time window of early decay has been closely examined in dendritic spines, where the Ca2+ pump on the ER plays a dominant role in regulating the rapid decline seen in some spines, whereas transients in other spines exhibit an additional slower decaying component attributed to diffusion to and from the dendritic shaft (Majewska et al. 2000). Although the time scale of early decay reported here corresponds well with the rapid decline in intraspine [Ca2+], it was not affected by any of the tested blockers of extrusion processes. Blockers tested in this report include those that block Ca2+ extrusion processes: KB R7943 for plasma membrane Na+/Ca2+ exchange and C28R2 for the plasma membrane Ca2+ pump. In addition, blockers of Ca2+ sequestration processes are also tested: thapsigargin and CPA for ER Ca2+ pump, CGP37157 for mitochondrial Na+/Ca2+ exchange, and ruthenium red for the mitochondrial uniporter. Some of these blockers (KB R7943, C28R2, ruthenium red) are effective on the time scale of PTP, i.e., minutes (Brenner and Wilkens 2001; Tang and Zucker 1997; Zhong et al. 2001). Although KB R7943 is also a Na+/Ca2+ exchange blocker, unlike Li+, it did not alter early decay. This was presumably because this drug mainly blocks the exchanger when it is operating in reverse mode, which does not happen during early decay.

Thapsigargin, CPA, and CGP37157 have been reported to be ineffective at changing PTP at the excitor (Tang and Zucker 1997; Zhong et al. 2001), and our test results extend this to the inhibitor. In addition, these blockers also have no effect on the release and Ca2+ transients evoked by 100-Hz trains of 0.1 to 5 s in duration. Thus it is likely that the mitochondrial Na+/Ca2+ exchanger and ER Ca2+ pump are functionally unimportant in crayfish terminals. Alternatively, it is possible that the blockers did not have access to presynaptic terminals. We think this is unlikely because the terminals are directly exposed to perfusing saline. In addition, we and others have obtained expected biological effects using various reagents dissolved in DMSO, e.g., Ca2+ buffers or drugs. Although the selection of blockers in this report was based mainly on their shown efficacy in previous studies, it remains possible that these blockers, which were developed in mammalian systems, cannot effectively block the intended targets in crayfish at the concentrations used here. Finally, there are additional agents that are known to be effective in mammalian tissues but remain untested in invertebrates. Therefore it is possible that blockers exist that are more effective than those used here. Nevertheless, based on the results reported above, it is reasonable to suggest that the major extrusion/sequestration pathways cannot account for early decay.

Lithium substitution represents the only exception in that it slowed early decay slightly. However, part of this change was probably due to a significant reduction in Ca2+ influx. Specifically, due to the [Ca2+]i-dependent nature of the buffering processes, the reduction in Ca2+ influx that resulted from the smaller action potential amplitude in Li+ is expected to result in a slower decay (Rozov et al. 2001). Therefore the real impact of the Li+ mediated slowing of early decay is likely to be smaller than that suggested by the traces in Fig. 3B. These negative results are consistent with a previous report showing that the main extrusion process at the crayfish NMJ removes [Ca2+]i, from a concentration of 1 ∼ 2 μM to the resting level of ∼100 nM, on a time scale of tens of seconds (Tank et al. 1995); this extrusion rate is too slow to account for the early decay reported here. Therefore we considered alternative mechanisms that could underlie early decay.

Theoretically, the rapid diffusion of Ca2+-bound MgOrg to terminal branches that do not contain Ca2+ channels (Delaney et al. 1989; Vyshedskiy and Lin 2000), and the subsequent release of Ca2+ there, could cause fluorescence intensity to decrease on a time scale comparable with that of early decay (Majewska et al. 2000). This process would lead to the prediction that a buffer with low mobility would delay this spatial re-equilibration and slow the early decay. There are several lines of evidence arguing against this hypothesis. First, inhibitor terminals accounted for the majority of volume from which fluorescence transients were recorded (e.g., see images in Supplement II). As a result, the volume attributable to connective axons between varicosities was small and not likely to represent a significant sink for Ca2+-bound MgOrg. In addition, early decay correlated predominantly with the affinity of a dye rather than its mobility. Specifically, 70-kDa dextran-coupled Oregon green had a theoretical diffusion constant about seven times smaller than that of the other dyes used here (Schmidt et al. 2003a), but its early decay conformed to the trend defined by its affinity (Fig. 1E). Furthermore, FFP18 was likely completely fixed to the membrane, and yet its early decay was essentially identical to that of diffusible Fura-2. Therefore diffusion is unlikely to contribute to early decay.

Although endogenous buffers with a slow Ca2+ binding rate have not commonly been postulated, it should be noted that results shown here are not compromised by the uncertainty related to buffer washout that typically occurs in experiments using patch electrodes. Moreover, calcium binding proteins capable of buffering Ca2+ at a slow rate have been described. For example, reloading of chromaffin cells with purified parvalbumin (PV) (Lee et al. 2000) and imaging of PV knock-out mice (Schmidt et al. 2003b) reveal that this protein can act as a slow buffer. Specifically, PV has been shown to bind both Ca2+ and Mg2+ (Eberhard and Erne 1994) and a significant fraction of PV appears to bind Mg2+ at rest. As a result, the PV buffering of Ca2+ presumably involves release of Mg2+ before binding to Ca2+. This two-step buffering could potentially absorb Ca2+ on a time scale comparable with that of early decay (Lee et al. 2000; Schmidt et al. 2003b). Therefore, although there is ample evidence for the presence of fast intrinsic buffers (Neher 2000), it should not be surprising to find systems in which slow buffering processes dominate.

Fast and slow buffers coexist at the presynaptic terminals of the crayfish NMJ

Because a slow buffer might not effectively buffer incoming Ca2+ during the brief period of an action potential and given the large Ca2+ influx at crayfish terminals estimated previously (Tank et al. 1995), the crayfish terminal with only slow endogenous buffer is likely to experience a high [Ca2+]i during an action potential. We explored this possibility by examining whether MgOrg was saturated by the Ca2+ influx activated by a broad action potential. The simulations in Fig. 6 show that a terminal containing only a slow buffer appropriate for early decay would leave Ca2+ unbuffered during the course of a broad action potential. Low affinity indicators such as MgOrg, Kd = 12 μM, would be ∼50% Ca2+ bound in such a system. Because the peak amplitude of the averaged MgOrg transient was about 6%, whereas the maximal MgOrg fluorescence intensity was 34%, the indicator is far from being saturated under these conditions (see also Vyshedskiy and Lin 2000 for data obtained with Magnesium Green, Kd = 7 μM). These results suggest that fast intrinsic buffers capable of competing with MgOrg are likely to coexist with the slow buffers that dictate early decay. Functionally, a fast buffer should play an important role in controlling local [Ca2+]i during the time window of synaptic transmission, whereas a slow buffer would dictate the duration of elevated [Ca2+]i after action potentials.

With the assumption that fast and slow endogenous buffers coexist, the mobility of these buffers remains to be determined. Simulations shown in Fig. 2C suggest that Ca2+ indicators of different mobility may be useful tools for probing the mobility of endogenous buffers. Indeed, the small but clear difference between the rising phases of FFP18 and Fura-2 transients (Fig. 5B) suggests that the endogenous buffers could be fixed. However, given the uncertainty about the relative proportion of fast and slow buffer, it is premature to further consider the issue of buffer mobility with the data presented here.

Functional considerations of slow intrinsic buffers

A slow intrinsic buffer can be kinetically indistinguishable from a fast calcium dependent extrusion process as the cause of early decay. For example, one typical method for estimating the Ca2+ extrusion rate is to measure the decay time constants of [Ca2+]i after loading cells with different concentrations of an exogenous buffer. The slope obtained by plotting the decay time constants against the binding ratios of exogenous buffer yields an estimate of the extrusion rate (Neher and Augustine 1992; Tank et al. 1995; see Helmchen and Tank 1999 for review). However, a similar correlation could also be accounted for by the properties of slow intrinsic buffers. Specifically, injection of a high concentration of dye would reduce [Ca2+]i, which in turn would slow down the absorption of free Ca2+ by the intrinsic buffer (Rozov et al. 2001).

Although both hypotheses, a fast extrusion rate and a slow intrinsic buffer, could result in a similar interaction of Ca2+ with exogenous Ca2+ indicators, the two processes would have significantly different functional consequences. For example, free Ca2+ absorbed by a slow buffer would subsequently have to be released before it could be extruded. The re-release phase would result in a small but persistent elevation of [Ca2+]i that would be absent if Ca2+ had already been removed by a fast extrusion process (Tang and Zucker 1997; Zhong et al. 2001). In addition, the two mechanisms would give rise to different spatial and temporal dynamics of Ca2+ during the course of an action potential. A very high submembrane [Ca2+]i would be likely during and shortly after an action potential in a system dominated by slow buffer. In contrast, in a system in which a fast extrusion process dictated early decay, [Ca2+]i near the membrane would be rapidly removed and kept at a relatively low level by nearly 10-fold (Fig. 6).

In conclusion, based on the negative effects of extrusion/sequestration blockers and the lack of effect of indicator mobility on early decay, we propose that early decay could be due to slow Ca2+ binding kinetics of endogenous buffer. In addition, we propose the simultaneous presence of both fast and slow buffers at the crayfish NMJ. The coexistence of two buffers would reduce the binding ratio that has been attributed to fast buffers. The consequence of this suggestion is that the transient [Ca2+]i increase across the synaptic terminal may be higher than that estimated in previous modeling studies.


This work was supported by National Institute of Neurological Disorders and Stroke Grant NS-31707 to J.-W. Lin.


We thank V. Matveev for comments and N. Schweitzer for correcting our English.


  • 1 The Supplementary Material for this article is available on line at

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract