Strontium is capable of supporting synaptic transmission, but release is dramatically different from that evoked in calcium. By measuring presynaptic strontium levels, we gain insight into the actions of strontium, which has implications for the identification of molecules involved in different aspects of synaptic transmission. We examined presynaptic divalent levels and synaptic release at the granule cell to stellate cell synapse in mouse cerebellar slices. We find that the prolonged duration of release and paired-pulse facilitation in the presence of strontium can be accounted for by the slower removal of strontium from the presynaptic terminal. Phasic and delayed release are both driven by strontium less effectively than by calcium, indicating that a heightened sensitivity to strontium is not a feature of the binding sites involved in facilitation and delayed release. We also find that the cooperativity for phasic release is 1.7 for strontium compared with 3.2 for calcium, suggesting that differential binding may help to identify the calcium sensor involved in phasic release.
Strontium can effectively substitute for calcium in driving synaptic transmission (Miledi, 1966; Dodge et al., 1969). This makes strontium a useful tool in studying calcium-dependent aspects of release. In this paper, we examine three such phenomena, phasic release, delayed release, and paired-pulse facilitation (PPF), each of which is perturbed significantly by the substitution of strontium for calcium.
Studies at many synapses have found that the EPSC peak is greatly reduced in strontium (Dodge et al., 1969; Mellow et al., 1982;Augustine and Eckert, 1984; Bain and Quastel, 1992; Goda and Stevens, 1994; Abdul-Ghani et al., 1996). The dominant component of the EPSC at such early times is attributable to phasic release, which is driven by a high presynaptic calcium concentration (10–100 μm) local to the calcium channel pore (Calocal) (Fogelson and Zucker, 1985; Simon and Llinas, 1985; Roberts et al., 1990; Heidelberger et al., 1994). Phasic release has a steep, power law dependence on the calcium concentration, which is thought to be attributable to Calocaltriggering vesicle fusion by binding to multiple low-affinity calcium sensors (Dodge and Rahamimoff, 1967). Because no significant change has been reported in the power law dependence in strontium (Meiri and Rahamimoff, 1971; Augustine and Eckert, 1984; Goda and Stevens, 1994), the decreased phasic component has been ascribed to lower affinity of the sensor driving phasic release for strontium.
The most noticeable effect of strontium is that individual release events continue for hundreds of milliseconds after presynaptic stimulation. This appears to be an enhancement of delayed release, which consists of late release events driven by the low concentrations (≪1 μm) of residual calcium (Cares) that persist in the terminal after the calcium channels close (Barrett and Stevens, 1972; Rahamimoff and Yaari, 1973; Zengel and Magleby, 1981; Zucker and Lara-Estrella, 1983;Van Der Kloot and Molgo, 1993). Delayed release is thought to result from Cares acting on a high-affinity sensor (Goda and Stevens, 1994; Atluri and Regehr, 1998; Zucker, 1999). One hypothesis for the enhancement of delayed release in strontium is that the sensor has a higher affinity for strontium (Goda and Stevens, 1994). This hypothesis has influenced the search for candidate molecules that act as high-affinity sensors (Li et al., 1995). An alternative hypothesis is that enhanced delayed release could arise from differences in presynaptic strontium regulation (Goda and Stevens, 1994; Rumpel and Behrends, 1999; Yawo, 1999). Recent measurements of presynaptic residual strontium levels (Srres) have supported this view (Xu-Friedman and Regehr, 1999), but this study was limited by the inability to detect quantal events to determine release rates.
Strontium also enhances PPF (Zengel and Magleby, 1980). PPF is a form of short-term plasticity in which a presynaptic terminal stimulated twice at short intervals yields an enhanced second EPSC. PPF is similar to delayed release in that both are dependent on Cares (Van Der Kloot and Molgo, 1993; Kamiya and Zucker, 1994; Atluri and Regehr, 1996; Cummings et al., 1996), they have similar time courses (Zengel and Magleby, 1981), and both are enhanced in strontium. Thus, it has been proposed that they share the same underlying mechanism (Zucker and Lara-Estrella, 1983; Van Der Kloot and Molgo, 1993).
To clarify the differences between calcium- and strontium-evoked release, we measure Srres, synaptic currents, and quantal release rates at cerebellar granule cell to stellate cell synapses. We find that the effects of strontium on delayed release and the time course of PPF are most easily explained by higher Srres and slower extrusion from the terminal. Its effects on phasic release and PPF amplitude are most likely attributable to lower affinity and cooperativity of strontium for the sensors involved in phasic release. One implication of these results is that strontium stoichiometry, but not binding affinity, may help in the identification of molecules involved in different aspects of synaptic transmission.
MATERIALS AND METHODS
The methods used are similar to those described previously (Xu-Friedman and Regehr, 1999). Briefly, we cut 300 μm transverse slices from the vermis of the cerebellum of 14- to 21-d-old mice (ICR). Slices were incubated for 1 hr at 33°C in artificial CSF (ACSF), composed of (in mm): 125 NaCl, 26 NaHCO3, 1.25 NaH2PO4, 2.5 KCl, 25 glucose, 1 MgCl2, and 2 CaCl2 bubbled with 95% O2–5% CO2 (310 mOsm) pH 7.4. EPSCs were measured in whole-cell voltage-clamp recordings at 24°C from stellate cells in the distal half of the molecular layer, using 2–2.5 MΩ electrodes containing (in mm): 35 CsF, 100 CsCl, 10 HEPES, 10 EGTA, and 0.1 D-600 (290 mOsm), pH 7.2. Bicuculline (20 μm) was added to the bath to block spontaneous inhibitory activity. We measured presynaptic residual divalent levels from parallel fibers using the Ca-sensitive indicator magnesium green, as described previously (Xu-Friedman and Regehr, 1999). Changes in divalent levels were quantified as the percent change in fluorescence (% ΔF/F). In some experiments, we observed a distinct, small, second peak 10–20 msec after the first in the average EPSC or in the fluorescence transient, and these experiments were discarded from analysis.
To measure residual divalent levels or release in the presence of different calcium concentrations, standard ACSF was replaced by a similar bathing solution, except it contained decreased concentrations of CaCl2 and increased MgCl2, so that the total concentration of divalents was maintained at 3 mm. For simplicity, we refer to these bathing solutions by their free calcium concentrations, i.e., 2 Cae, 1 Cae, etc. To examine release in the presence of strontium, a different bathing solution was used to buffer contaminating calcium, composed of (in mm): 120 NaCl, 34 NaHCO3, 1.25 Na2HPO4, 2.5 KCl, and 2 EGTA. SrCl2 and MgCl2 were then added to bring total divalents to 5 mm. EGTA has a very high affinity for Ca2+(K d of ∼100 nm), which eliminates contaminating Ca2+ (∼6 μm). Most of the remaining EGTA binds Sr2+(K d of ∼30 μm), and very little binds Mg2+ because EGTA has a very low affinity for Mg2+(K d of ∼15 mm). Thus, adding 4 mmSrCl2 and 1 mmMgCl2 leaves a free Sr2+ concentration of 2 mm and free Mg2+concentration of 1 mm. For simplicity, we refer to these bathing solutions by their free Sr2+ concentration, i.e., 2 Sre, 1 Sre, etc. Most experiments were started in 2 Cae and then switched to one or more different calcium or strontium concentrations. Averages were computed after normalizing to the EPSC or peak ΔF/F in 2 Cae. Six physiology experiments that switched from 2 Sreto 1, 1.5, or 3 Sre were normalized by the average peak in 2 Sre. Physiology experiments were used for quantification only if they showed >75% recovery after washout.
The fluorescence transients recorded in Srereflect residual strontium levels. Calcium released from internal stores is unlikely to contribute to these signals for several reasons. First, experiments using thapsigargin and ryanodine have no effect on Cae fluorescence transients in rat parallel fibers (Sabatini and Regehr, 1995). Second, the rise times of the transients are fast, reaching a peak in a few milliseconds, whereas in systems in which calcium stores can be triggered by divalent influx, the rise times are slower and the peaks are broader (Lipp and Niggli, 1994). Third, there is close agreement in estimates of Srres based on indicators with differentK d-Sr/K d-Ca(Xu-Friedman and Regehr, 1999); this would not be the case if there were significant calcium contamination.
To count individual quantal events underlying evoked release, we examined the first derivative of each trace, and took events whose initial slopes exceeded some threshold. At low release frequency, this method successfully captured most events, as determined by visual inspection and by comparing the average EPSC against the convolution of event times with the average quantal event. Release rates were too high to detect all events during the first 5 msec and are therefore blanked. For the average delayed release time course, we averaged data from eight experiments in 2 Cae by first subtracting the spontaneous release rate and then normalizing each experiment by the total number of detected events. Of those eight experiments, three had accompanying data in 2 Sre, which were scaled by the Cae data recorded in the same experiment. In addition, we included seven further 2 Sreexperiments that had no accompanying estimates of relative release rates in Cae, by scaling them to the average number of release events in Sre, and then computing an overall average.
In studying phasic release, we used measurements of peak ΔF/F to estimate divalent influx in different concentrations of Cae and Sre. For this approach to work, several preconditions must be met. First, the number of fibers stimulated in each condition must remain constant. Field potential measurements of the presynaptic action potential volley propagating along the parallel fibers have shown that this does not change significantly upon changing to Sre (Xu-Friedman and Regehr, 1999). Second, the ΔF/F transient must be proportional to residual divalent levels without dye saturation, which is satisfied because peak residual divalent levels (<500 nm for calcium, <1 μm for strontium) are much less than the binding affinities for magnesium green (K d-Ca of 6 μmand K d-Sr of 33 μm). Third, peak residual divalent levels must be proportional to influx, i.e., the endogenous calcium buffer is not saturated. The endogenous buffer appears to have sufficiently low affinity for calcium that saturation is not significant (Sabatini and Regehr, 1998), and it has yet lower affinity for strontium (Xu-Friedman and Regehr, 1999). Thus, complications caused by fiber excitability, indicator saturation, and buffer saturation appear to be small, and we use the peak ΔF/F as an estimate of influx.
To examine the power law relationship between release and external divalent levels, we fit the data using a function of the following form: EPSC = k([divalent]e)n. Thus, to statistically compare the cooperativities (n) for strontium and calcium, we used the logarithmic form of the function and compared the slopes of the regression lines using a t test (Glantz, 1997). This approach was also applied to the power law between release and internal divalent level. The contribution of measurement errors in divalent influx was considered negligible.
We examined the effects of strontium on synaptic transmission at the cerebellar granule cell to stellate cell synapse in 14- to 21-d-old mice. This preparation has the following advantages: there is no recurrent excitation (Palay and Chan-Palay, 1974), spontaneous inhibitory responses can be blocked with bicuculline (Llano et al., 1991), individual excitatory release events are easily detected (Atluri and Regehr, 1998; Chen and Regehr, 1999), spontaneous release is low, and presynaptic residual divalent levels can be measured (Regehr and Atluri, 1995; Xu-Friedman and Regehr, 1999).
We considered three aspects of synaptic transmission: phasic release, delayed release, and paired-pulse facilitation (Fig.1). Phasic release is prominent in 2 mm Cae, as reflected in the fast decay time of the average EPSC evoked by stimulating the parallel fibers (Fig. 1 A, top trace). There is a contribution from delayed release, and individual quantal release events can be seen in single trials (Fig. 1 A,bottom traces). When the external calcium is replaced with Sre, phasic release is reduced (Fig.1 B, top trace), on average to 7.9 ± 0.9% of control (mean ± SE, n = 10 experiments). However, delayed release is increased, as reflected in the slower decay of the average EPSC, and in the marked frequency of evoked quantal release for hundreds of milliseconds in individual trials (Fig.1 B, bottom traces). Additionally, when two stimuli are delivered 20 msec apart in Cae, the second EPSC is facilitated (Fig. 1 C, left). PPF was quantified as A2/A1 − 1, whereA1 and A2 are the peak amplitudes of the first and second EPSCs, respectively. In Cae, PPF was 1.4 ± 0.1 (n = 28), but in Sre, PPF is larger (Fig. 1 C,right), on average 3.3 ± 0.2 (n = 22).
Thus, there are three major effects of strontium: delayed release is increased, phasic release is reduced, and paired-pulse facilitation is increased. These effects have been found previously at other synapses. However, at this synapse, we are able to measure strontium levels to provide insight into the underlying mechanisms.
To examine the effects of strontium on delayed release, we changed from 2 Cae to 2 Sre. The decrease in phasic release and the increase in delayed release are fully reversible (Fig.2 A,B). We quantified the rate of release by detecting individual release events. For individual trials, we recorded the times for all detected quanta and compiled them into a raster plot (Fig.2 B). To measure the average release rate over time, these events were binned into a peristimulus time histogram (PSTH) (Fig. 2 C). The raster and PSTH plots both show that the evoked release rate is elevated over the spontaneous rate in Cae for ∼100 msec and in Sre for >500 msec. After ∼10 msec, evoked release rates in Sre are considerably higher than in Cae.
We averaged data from several such experiments (Fig.2 D). First, we normalized all experiments to the same total number of events in Cae and then scaled the accompanying trials in Sre accordingly. Delayed release shows a double-exponential decay in Cae(τ1 of 6.4 msec, τ2 of 64 msec; n = 8) and Sre(τ1 of 20 msec, τ2 of 89 msec; n = 10). After the first 50 msec, the rate of release in Sre is ∼10 times greater than that in Cae.
Release in strontium versus calcium was analyzed quantitatively in a previous study, which described it as a combination of two components, a fast component (equivalent to phasic release) that was greater in calcium, and a slow component (equivalent to delayed release) that was greater in strontium (Goda and Stevens, 1994). One explanation for these findings was that the affinity of the calcium sensors driving release may be different for strontium (Goda and Stevens, 1994). In particular, the low-affinity sensor responsible for phasic release may have a lower affinity for strontium, and the high-affinity sensor responsible for delayed release may have a higher affinity for strontium. An alternative mechanism that could also influence the time course of delayed release is that Srres dynamics could differ considerably from Cares, because of differences in buffering and extrusion (Barrett and Stevens, 1972; Goda and Stevens, 1994; Xu-Friedman and Regehr, 1999). The first explanation has gained dominance in the literature. However, to distinguish between these alternatives, it is important to take into account measurements of the time course of Cares and Srres.
We have previously applied conventional fluorometric Cares measurement techniques toward quantifying Srres (Xu-Friedman and Regehr, 1999). Here, we used magnesium green as the fluorophore (Zhao et al., 1996) for two reasons. First, it has a low affinity for strontium and calcium compared with the amplitudes of Cares and Srres after single stimuli (see Materials and Methods) (Regehr and Atluri, 1995;Xu-Friedman and Regehr, 1999), such that the time course of fluorescence transients linearly reflects the time course of divalent transients. Second, magnesium green shows large percent fluorescence changes upon stimulation, yielding a high signal-to-noise ratio. Contamination of the fluorescence signal by magnesium is not significant, because ΔF/F measurements for the low-affinity calcium indicator fura-2FF, which has no measurable magnesium affinity, did not differ significantly in their time course (Xu-Friedman and Regehr, 1999).
When the bathing solution is changed from Cae to Sre, there is a drop in the peak fluorescence change in response to stimulation (Fig.3 A, B,left). We converted these fluorescence transients into the relative concentrations of residual divalents by taking into account the binding affinities of magnesium green for calcium versus strontium (i.e., by multiplying by 5.5 =K d-Sr/K d-Ca) (Xu-Friedman and Regehr, 1999). After this correction, it is clear that Srres is greater than Cares, and persists much longer in the terminal (Fig. 3 B, right). We averaged together several such experiments and found that peak Srres is 1.60 ± 0.02 (n = 16) times peak Cares (Fig. 3 C), which is probably a result of a combination of greater presynaptic influx and lower affinity of the endogenous buffer for strontium compared with calcium. Srres then decays approximately five times more slowly than Cares (Srreshalf decay time, 165 ± 5 msec, n = 8; Cares, 31 ± 0.4 msec, n = 17), probably because of lower efficacy of extrusion.
We show the average delayed release rate as a function of average residual divalent concentration in Figure4. In Cae, two components of delayed release with different dependencies on calcium are evident (Fig. 4 B) (Atluri and Regehr, 1998), whereas only one component is visible in strontium. This is probably because release rates have declined too much at low strontium concentrations to adequately quantify a second component. If release rates depended solely on instantaneous divalent levels, it should be possible to use these curves to determine the power law dependence of release and to compare release driven by strontium and calcium. However, previous work has shown that release rates also depend on the kinetics of a process driven by divalent binding (Atluri and Regehr, 1998). Therefore, fits have not been included, and it is difficult to interpret the apparent difference in the steepness of divalent dependence for Sre and Cae.
It is clear, however, that overall concentrations of divalent, release rates in Cae are 5–10 times greater than in Sre. Thus, strontium is less effective at driving release. The higher levels of delayed release observed in Figure2 D are most simply explained by the high and prolonged presence of Srres in the presynaptic terminal.
We examined the effect of strontium on phasic release by determining the power law relationship between the peak EPSC and the magnitude of presynaptic divalent influx. Measurements of peak Cares or Srres in different Cae or Sre were used as an indication of influx (see Materials and Methods). We began each experiment in 2 Cae and switched to test solutions containing different concentrations of Cae or Sre (Fig.5 A,B). To compute average results from the test solutions, they were first normalized by the peak in 2 Cae. The relationship between Cae and peak Careswas sublinear, indicating saturation of influx (Fig.5 A,C) (Mintz et al., 1995). In Sre, however, we saw no evidence of influx saturation (Fig. 5 B,C).
We estimated peak release rates using the EPSC peak amplitude in different Cae and Sre (Fig.6 A,B). These data were fit to a power law relationship: EPSC =k([divalent]e)n (Fig.6 C). The values of n for calcium (2.1 ± 0.2, n = 11) and strontium (1.9 ± 0.1,n = 28) did not differ significantly (t= 0.54, df = 35, p > 0.5). However, this traditional analysis can be misleading because influx and external divalent concentrations are not linearly related (Fig. 5 C). When we use our measure of relative divalent influx and fit the data using the revised relationship (EPSC = k([divalent]i)n), the power law exponent changes significantly: 3.2 ± 0.2 for calcium versus 1.7 ± 0.1 for strontium (Fig. 6 D) (t = 2.49, df = 35, p < 0.02).
We examined the effect of substituting strontium for calcium on the amplitude and time course of PPF. The amplitude of PPF increases when the external solution is changed from Cae to Sre (Fig. 1 C). One possible explanation for this is that a special calcium sensor is responsible for facilitation and that Srres binds to it more effectively than to Cares, thereby producing prominent facilitation. Another possible explanation is that a decrease in the initial probability of release enhances PPF, as has been found in many studies (Fig. 7 A) (Feng, 1941; Rahamimoff, 1968; Creager et al., 1980; McNaughton, 1982;Manabe et al., 1993; Dobrunz and Stevens, 1997). Experiments in the presence of strontium change both the residual divalent signal available to drive facilitation and the initial probability of release. To determine which of these factors leads to the enhancement of PPF in Sre, we manipulated residual divalent levels and release probability by changing the concentrations of Cae and Sre.
We measured the amplitude of PPF using pairs of stimulation pulses separated by 20 msec. In 2 Cae, PPF was 1.4 ± 0.1 (n = 28) (Fig. 7 A, left), whereas in 2 Sre, PPF was much larger, 3.3 ± 0.2 (n = 22) (Fig. 7 B, left). However, when the divalent concentration was reduced, PPF in 1 Cae increased significantly to 3.1 ± 0.5 (n = 4) (Fig. 7 A, right), and PPF in 1 Sre remained unchanged (2.7 ± 0.3,n = 8) (Fig. 7 B, right). Thus, although there is a considerable difference between the magnitude of PPF in 2 Cae versus 2 Sre, this difference decreases as the concentration of divalents is reduced (Fig. 7 C). A plot of the magnitude of PPF as a function of peak residual divalent concentration (Fig. 7 D) reveals that, as peak Cares increases, PPF in Cae decreases, but as peak Srres increases, PPF in Sreremains unchanged. These findings indicate that the amplitude of PPF at this synapse does not simply reflect differences in the residual divalent signals in Sre and Cae, and that the initial probability of release must also be considered.
When the initial probability of release is taken into account (Fig.7 E), there is no fundamental difference between the magnitude of PPF in Sre and Cae. The magnitude of PPF is ∼3 when the probability of release is low, whether in 1 Caeor in 1 to 3 mm Sre. The magnitude of PPF decreases only when the probability of release increases, as when Cae is larger than 1 mm. Thus, the high levels of PPF in Sre can be accounted for by the low initial probability of release in Sre. It not necessary to hypothesize that the calcium sensor involved in facilitation is particularly sensitive to strontium.
We next examined the time course of PPF in the presence of strontium. Previous studies have shown that PPF is prolonged in Sre compared with Cae(Zengel and Magleby, 1980; Van Der Kloot and Molgo, 1993). Although the reason for the protracted time course was not clear, several possible explanations are apparent after considering the role of calcium in facilitation. Comparisons between the decay of PPF and of Cares have suggested that PPF is produced by calcium binding to a high-affinity receptor but with slow kinetics (Atluri and Regehr, 1996). Thus, the time course of facilitation in Cae reflects in part the time course of Cares and in part a slow calcium-driven process, which produces a lag between the decay of Caresand of PPF. The prolongation of PPF in Sre could arise from slower kinetics of strontium binding to the sensors responsible for facilitation or from slower extrusion of Srres from the terminal compared with Cares.
We measured the time course of PPF at the granule cell to stellate cell synapse in 2 Cae and 2 Sre(Fig. 8 A) by varying the interval (Δt) between pairs of pulses. In Cae, PPF returns to initial values by Δt = 500 msec (Fig. 8 A,top), but in Sre it remains elevated until after Δt = 750 msec (Fig. 8 A,bottom). On average, PPF was higher in Sre than in Cae for at least 1 sec (Fig. 8 B). This is not just a result of the higher levels of PPF observed in 2 Sre, because when the time course of facilitation is normalized for peak PPF, facilitation persists longer in Sre (Fig.8 C). Fitting the PPF curves with an exponential function, the τdecay in Cae was 175 ± 24 msec (n = 3) and in 2 Sre was 480 ± 60 msec (n = 5).
To determine the basis for this prolongation, we directly compare the time courses of residual divalent levels and of PPF (Fig.8 D). In Cae, we observe the expected lag between the decay of PPF and of Cares (Fig. 8 D,top). In Sre, the time courses of PPF and Srres match each other much more closely (Fig. 8 D, bottom). This suggests that the primary determinant of the time course of PPF in Sre is the time course of Srres, which is prolonged by the slower removal of strontium from the presynaptic terminal.
Our studies have provided new insights into calcium-driven aspects of synaptic transmission. Contrary to the accepted view, we find that the prominent delayed release in Sre, a characteristic that has been widely used in the study of synaptic transmission, is a consequence of less efficient buffering and extrusion of strontium from the presynaptic terminal. Thus, high sensitivity to strontium is not a hallmark of the calcium sensor involved in facilitation and delayed release. Our studies of the strontium dependence of phasic release confirm that strontium triggers phasic release less effectively than does calcium. However, in contrast to previous studies, by measuring presynaptic strontium influx we find that strontium also triggers phasic release with a lower cooperativity than calcium. This feature will be important in identifying and characterizing the calcium sensor involved in phasic release.
Strontium dependence of phasic release
Our studies of the strontium dependence of phasic release provide important insights into the calcium sensors that trigger fast synaptic transmission. In agreement with studies of other synapses, we found that the EPSC peak is greatly reduced in strontium. Traditionally, this has been interpreted as being attributable to the sensor having lower affinity for strontium than for calcium. However, our measurements of presynaptic strontium influx additionally show that strontium triggers phasic release with a lower cooperativity than does calcium (n = 1.7 for strontium and n = 3.2 for calcium). This indicates that the effect of strontium on phasic release is more complicated than was previously envisioned.
To understand the implications of this difference in cooperativity, it is useful to consider the power law relationship between calcium and release. A widespread feature of fast synaptic transmission is that release is steeply dependent on extracellular calcium levels, with a power law relationship of n = 2–4 at most synapses (Dodge and Rahamimoff, 1967; Augustine et al., 1985; Lando and Zucker, 1994; Mintz et al., 1995; Borst and Sakmann, 1996; Sabatini and Regehr, 1997). This power law relationship is usually interpreted as multiple calcium ions binding to the sensors that trigger release. Most previous studies did not reveal differences in the cooperativity for calcium and strontium (Meiri and Rahamimoff, 1971; Augustine and Eckert, 1984; Goda and Stevens, 1994), suggesting that the same number of strontium ions as calcium ions are required to bind to trigger release. However, technical limitations in these experiments may have obscured differences in the power laws. For example, most of these studies were unable to take into account the relationship between the extracellular divalent concentration and influx. We found that it was necessary to measure intracellular calcium and strontium levels to estimate influx to adequately estimate the power law relationships, because external divalent concentrations and influx are not necessarily linearly related (Fig. 5 C).
This differential cooperativity will be important in identifying the calcium sensor involved in phasic release. Further experiments are required to evaluate possible contributions to differential cooperativity, including endogenous buffer saturation near the channel, lower strontium permeability through channels more strongly coupled to release, and differences in the kinetics of binding to the release sensor. However, the simplest interpretation of the lower cooperativity for strontium than calcium is that fewer strontium ions bind to the sensors that trigger release, and when they do bind they are less effective at triggering release. Although this hypothesis needs to be investigated further, a decrease in the number of strontium ions binding is consistent with preliminary structural data of the C2A domain of synaptotagmin I, which has been hypothesized to serve as the calcium sensor involved in phasic release. This domain binds three calcium ions but only one strontium ion (J. Garcia, T.C. Südhof, and J. Rizo, personal communication), which would be difficult to reconcile with calcium and strontium having the same cooperativity for phasic release. Similar studies of strontium and calcium binding to other presynaptic calcium binding proteins could provide further insight into the identity of the calcium sensors involved in phasic release. In addition, if the binding of fewer strontium ions can trigger release, this will provide an important means of studying how divalent ion binding produces the conformational changes involved in vesicle fusion. Other possible influences on cooperativity should also be investigated.
Mechanism responsible for prominent delayed release in strontium
The most widespread use of strontium in the study of synaptic transmission is to desynchronize release to allow examination of the amplitude of evoked quantal events (Abdul-Ghani et al., 1996; Oliet et al., 1996; Choi and Lovinger, 1997; Morishita and Alger, 1997; Otis et al., 1997; Behrends and ten Bruggencate, 1998;Lévénès et al., 1998; Bekkers and Clements, 1999;Bellingham and Walmsley, 1999). Our studies suggest that the prevailing hypothesis for this aspect of synaptic transmission needs to be revised. Previously, prominent delayed release in the presence of strontium had been ascribed to a specialized calcium sensor that binds strontium with higher affinity than it binds calcium (Goda and Stevens, 1994). Although we see a similar enhancement of delayed release, we find that, when the residual divalent levels in the presynaptic terminals are taken into account, delayed release is actually much lower for strontium compared with calcium over all concentrations. These high sustained levels of strontium are likely a consequence of less efficient buffering and extrusion of strontium from the presynaptic terminal.
Mechanism responsible for effects on PPF in the presence of strontium
Our measurements of residual strontium levels and the initial probability of release suggest that the explanation for the effects of strontium on the amplitude and time course of facilitation must also be revised. In agreement with previous studies (Zengel and Magleby, 1980;Van Der Kloot and Molgo, 1993), we found that the time course of facilitation was prolonged in the presence of strontium compared with calcium. Just as for delayed release, this is likely a consequence of the prolonged elevation of strontium in the presynaptic terminal because of the close agreement between the time courses of presynaptic strontium and facilitation.
We also observed an increase in the amplitude of PPF. One interpretation of this result is that the increased amplitude of PPF in the presence of strontium reflects the preferential binding of strontium to the calcium sensors responsible for facilitation. Our results suggest an alternative interpretation that is based on the widely observed contribution of the initial probability of release to the amplitude of facilitation. Facilitation is much more prominent in strontium than in calcium, except when calcium levels are low. These results suggest that the major effect of strontium on the amplitude of facilitation is not a direct consequence of strontium binding preferentially to the release sites involved in facilitation. Rather, it is a consequence of strontium reducing the initial probability of release, which indirectly increases the amplitude of facilitation.
One implication of these findings is that, as for delayed release, strontium acts through the same machinery that calcium does but with lower affinity. Because many calcium-dependent biological processes are driven less well by strontium, this is not likely to be a useful criterion for identifying the high-affinity calcium sensors involved in facilitation.
Overall, this study highlights the advantages of a comprehensive approach in the study of synaptic transmission. This includes the measurement of divalent levels, their influx and time course, and consideration of the interaction between different aspects of transmitter release, such as between PPF and the initial probability of release. This approach was invaluable here in understanding the varied effects of strontium on synaptic transmission.
This work was supported by National Institutes of Health (NIH) Postdoctoral Training Grant 5T32 NS07112–19 to M.A.X.-F. and NIH Grant R01-NS32405–01 to W.G.R. We thank Bruce Bean, Adam Carter, Chinfei Chen, Jeremy Dittman, Anatol Kreitzer, and Kaspar Vogt for helpful comments. We also thank Michael Escobar and Raji Balasubramanian for statistical help.
Correspondence should be addressed to Dr. Wade G. Regehr, Department of Neurobiology, Harvard Medical School, 220 Longwood Avenue, Boston, MA 02115. E-mail:.