Abstract
Many types of neurons release endocannabinoids from their dendrites in response to elevation of intracellular calcium levels. Endocannabinoids then activate presynaptic cannabinoid receptors, thereby inhibiting neurotransmitter release for tens of seconds. A crucial step in understanding the physiological role of this retrograde signaling is to determine its sensitivity to elevations of postsynaptic calcium. Here we determine and compare the calcium dependence of endocannabinoid-mediated retrograde inhibition at three types of synapses onto cerebellar Purkinje cells. Previous studies have shown that Purkinje cell depolarization results in endocannabinoid-mediated retrograde inhibition of synapses received from climbing fibers, granule cell parallel fibers, and inhibitory interneurons. Using several calcium indicators with a range of affinities, we performed a series of in situ and in vitro calibrations to quantify calcium levels in Purkinje cells. We found that postsynaptic calcium levels of ∼15 μM are required for half-maximal retrograde inhibition at all of these synapses. In contrast, previous studies had suggested that endocannabinoid release could occur with slight elevations of calcium above resting levels, which implies that inhibition should be widespread and continuously modulated by subtle changes in intracellular calcium levels. However, our results indicate that such small changes in intracellular calcium are not sufficient to evoke endocannabinoid release. Instead, because of its high requirement for calcium, retrograde inhibition mediated by calcium-dependent endocannabinoid release from Purkinje cells will occur under more restricted conditions and with greater spatial localization than previously appreciated.
Introduction
Endocannabinoids act as retrograde signaling molecules in many brain regions, allowing postsynaptic cells to modulate their synaptic inputs (Kreitzer and Regehr, 2002; Wilson and Nicoll, 2002). Neurons release endocannabinoids such as anandamide and 2-arachidonylglycerol (Devane et al., 1992; Di Marzo et al., 1994; Stella et al., 1997), which modulate various ion channels (Mackie and Hille, 1992; Mackie et al., 1993; Twitchell et al., 1997), inhibit synaptic strength (Levenes et al., 1998; Takahashi and Linden, 2000), and alter the firing of presynaptic cells (Kreitzer et al., 2002) by activating receptors such as G-protein-coupled CB1 receptors (Matsuda et al., 1990; Devane et al., 1992). The synthesis and release of endocannabinoids involve the calcium-dependent cleavage of phospholipid precursors (Di Marzo et al., 1994; Stella et al., 1997) and do not require vesicle fusion (Ohno-Shosaku et al., 2001; Wilson and Nicoll, 2001). Determining the calcium sensitivity of endocannabinoid-mediated modulation is crucial for understanding the conditions under which endocannabinoids are released and will help determine the mechanisms and functional role of this signaling system.
Retrograde signaling by endocannabinoids occurs throughout the brain but has been studied most extensively in the hippocampus and cerebellum. Depolarization of either CA1 pyramidal cells or Purkinje cells releases endocannabinoids, which activate presynaptic CB1 receptors and transiently reduce the strength of inhibitory synapses for tens of seconds (Kreitzer and Regehr, 2001a; Ohno-Shosaku et al., 2001; Wilson and Nicoll, 2001; Diana et al., 2002). This depolarization-induced suppression of inhibition (DSI) was shown to be calcium dependent as indicated by a requirement for extracellular calcium (Ohno-Shosaku et al., 1998), a voltage dependence similar to that of voltage-gated calcium channels (Lenz et al., 1998), and its sensitivity both to the duration of depolarization (Ohno-Shosaku et al., 1998) and to changes in intracellular calcium buffering (Lenz and Alger, 1999; Glitsch et al., 2000). In addition to DSI, depolarization-induced suppression of excitation (DSE) has been described at both parallel fiber (PF) and climbing fiber (CF) synapses onto Purkinje cells (Kreitzer and Regehr, 2001b; Maejima et al., 2001). Retrograde inhibition of all of these types of synapses is prevented by dialyzing the postsynaptic cell with calcium chelators (Pitler and Alger, 1992; Kreitzer and Regehr, 2001b). Moreover, for hippocampal synapses, photolytic elevation of calcium in the absence of depolarization is sufficient to induce retrograde synaptic inhibition (Wang and Zucker, 2001; Wilson and Nicoll, 2001). Thus, elevation of postsynaptic calcium (Capost) triggers release of endocannabinoids that serve as retrograde messengers for DSE and DSI.
The calcium dependence of endocannabinoid signaling has important implications for the magnitude and spatial extent of endocannabinoid actions. Previous studies of DSI suggest that endocannabinoid release requires only small increases in Capost. In the cerebellum, an estimated increase of 40–200 nm in a Purkinje cell was sufficient to trigger retrograde inhibition (Glitsch et al., 2000). Such a low requirement for calcium suggests that substantial endocannabinoid release may occur tonically and that even modest activity levels could produce widespread synaptic inhibition. In contrast, studies of hippocampal DSI estimated that ∼4 μm Capost is required to significantly suppress inhibitory synapses onto pyramidal cells (Wang and Zucker, 2001). This higher estimate suggests that greater levels of activity may be required to evoke endocannabinoid release from pyramidal cells than from Purkinje cells. Thus, previous studies provide differing estimates of the calcium dependence of retrograde endocannabinoid signaling at inhibitory synapses, and no studies to date have examined calcium sensitivity of retrograde inhibition at excitatory synapses.
Here we measure the calcium sensitivity of endocannabinoid-mediated retrograde inhibition for climbing fiber, parallel fiber, and inhibitory synapses onto Purkinje cells. After taking a series of steps to determine absolute levels of Capost, we found that for all three types of synapses, retrograde inhibition produced by Purkinje cell depolarization required Capost of ∼15 μm. In contrast to previous reports that endocannabinoid signaling is highly sensitive to calcium, our findings demonstrate that very large calcium signals are required to activate the endocannabinoid signaling system.
Materials and Methods
Electrophysiology. Parasagittal slices (300 μm thick) were cut from the cerebellar vermis of 10- to 12-d-old Sprague Dawley rats. Dissections were performed in ice-cold sucrose solution containing (in mm): 78 NaCl, 26 NaHCO3, 68 sucrose, 25 glucose, 2.5 KCl, 1.25 NaH2PO4-H2O, 7 MgCl2, 0.5 CaCl2. Slices were incubated for 30 min at 34°C in sucrose solution and then transferred to saline solution containing (in mm): 125 NaCl, 26 NaHCO3, 1.25 NaH2PO4-H2O, 2.5 KCl, 1 MgCl2, 2 CaCl2, 25 glucose, and allowed to cool to room temperature. All solutions were bubbled with 95% O2/5% CO2.
Whole-cell voltage-clamp recordings of Purkinje cells were obtained using glass electrodes (1.5–2.5 MΩ) filled with an intracellular solution containing (in mm): 135 CsMeSO4, 15 CsCl, 15 HEPES, 0.2 EGTA, 1 MgCl2, 2 MgATP, 0.3 NaGTP, 10 phosphocreatine, 2 QX-314, and 20 TEA. The holding potential was -60 mV. This standard internal solution was supplemented with calcium indicators as described in Results. Series resistance was compensated by 75–90% during recordings, which were made using an Axon Multiclamp 700A amplifier (Axon Instruments, Union City, CA). Data were filtered at 2–5 kHz and sampled at 10 kHz. Access resistance and leak currents were monitored continuously, and experiments were rejected if these parameters changed significantly during recording. Flow rates were 1–2 ml/min. Experiments were performed at room temperature.
Parallel fibers were stimulated with brief pulses (100–300 μsec, 50–500 μA) using a glass theta electrode (World Precision Instruments, Sarasota, FL) filled with saline and positioned in the molecular layer. Bicuculline (20 μm) was used to block GABAA-mediated synaptic currents. Climbing fibers were activated by placing a stimulus electrode in the granule cell layer. Individual climbing fiber responses were identified by a large all-or-none response with a sharp stimulus threshold. In addition to bicuculline (20 μm), 2,3-Dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX) (250–500 nm) was added to reduce climbing fiber EPSCs by 25–75% (Foster and Regehr, 2002) and to reduce voltage-clamp errors. For experiments measuring DSI, interneurons were stimulated by placing the stimulus electrode in the molecular layer in the presence of 10 μm NBQX.
Calibration of calcium indicators. To convert fluorescence ratios to calcium concentrations, we used the relationship: 1 where KD is the experimentally measured dissociation constant of the indicator for calcium, Rmin and Rmax are the fluorescence ratios at 0 and saturating calcium concentrations, respectively, and β is defined as the ratio of calcium-free versus calcium-bound fluorescence at the wavelength used as the denominator in R (Grynkiewicz et al., 1985). Because we have chosen the isosbestic point of the indicators for this wavelength, β = 1.
Measurement of indicator dissociation constants (KD). Previous studies have shown that the properties of indicators can change both as a result of buffer composition and from cytoplasmic interactions within cells. For example, the KD of calcium indicators can increase three- to fourfold when loaded into some types of cells (Baylor and Hollingworth, 1988; Konishi et al., 1988; Harkins et al., 1993; Kurebayashi et al., 1993). The extent to which such a KD shift occurs in Purkinje cells is not known and therefore introduces uncertainty to measurements of Capost within Purkinje cells. We therefore performed experiments to determine the KD of indicators within Purkinje cells. We began by including BSA (20 mg/ml) in our internal solution to mimic protein binding that could occur in vivo (Konishi et al., 1988; Harkins et al., 1993).
The KD of calcium indicators was measured in a spectrofluorimeter (Spex Fluorolog, Edison, NJ). Calibration solutions were prepared according to the technique of Tsien and Pozzan (1989). Briefly, calcium was buffered to known levels with 10 mm EGTA to which varying amounts of CaCl2 were added. Measurements were made at room temperature in a buffer that consisted of (in mm): 135 CsMeSO4, 15 CsCl, 15 HEPES, and 10 TEA, pH 7.30, which are the same concentrations as in our intracellular recording solution. Several minor components of the intracellular recording solution that could interfere with KD measurements were omitted from the calibration buffer. EGTA-buffered solutions were used to clamp calcium at concentrations up to 31 μm. Above the buffering range of EGTA, CaCl2 was added directly to calibration solutions to obtain free calcium concentrations of 0.1–10.0 mm. Fits of the Hill equation to fluorescence values, with the Hill coefficient constrained to 1, were used to determine KD. The KD of mag-fura-5 for magnesium was measured at 3.7 mm.
Addition of BSA to the calibration buffer increased the KD of several of the calcium indicators. As shown for fura-2FF in Figure 1Aa, BSA increased the KD by fourfold, from 8 to 32 μm. Similarly, BSA increased the KD of fura-2 and mag-fura-5 by 3.3-fold and 4.0-fold, respectively (see Table 1). Although there is uncertainty regarding the cytoplasmic protein concentration in Purkinje cells, the shift that we observe in KD suggests that the affinity of these indicators may also change when loaded in a Purkinje cell. Because of the possibility of such a shift in affinity, using cuvette KD values to estimate calcium concentrations with these indicators could cause underestimation of calcium levels by a factor of 3–4.
Fluo-5N, however, is not prone to such large uncertainties. We found that BSA did not significantly alter the affinity of fluo-5N (Fig. 1 Ab), which suggests that the dissociation constant determined in a cuvette is appropriate for quantifying calcium levels within a cell, regardless of the extent to which fluo-5N is bound to protein within the cell. Our measurements of KD and the effects of BSA are summarized in Table 1.
Magnesium sensitivity of the indicators. For some indicators, sensitivity to magnesium can complicate the quantification of calcium levels. KD values for magnesium are 5.6 mm for fura-2 (Grynkiewicz et al., 1985) and 3.7 mm for mag-fura-5 (our data), whereas fura-2FF does not respond to Mg changes that are <1 m (Xu-Friedman and Regehr, 1999). Because the fractional changes in Mg are usually small and the indicators used in this study are much more sensitive to Ca than to Mg, it is unlikely that small changes in Mg influence the quantification of calcium levels. However, it is possible that resting levels of magnesium, which are much higher than those of calcium, can influence the resting fluorescence ratio and make it difficult to quantify small changes in calcium with an indicator such as mag-fura-5.
Indicator pairs. The insensitivity of the KD of fluo-5N to proteins suggested a strategy to determine the appropriate calibration parameters for fura-2, fura-2FF, and mag-fura-5. It is possible to estimate calcium levels simultaneously with fluo-5N and one of these UV-excited indicators as illustrated in Figure 1 B, in which excitation spectra of a combination of fura-2FF and fluo-5N are shown. Our goal was to test the accuracy of calibration parameters by comparing simultaneous calcium measurements from two indicators. Using the wavelengths indicated in Figure 1 B, we calculated the fluorescence ratio of fura-2FF as λ2/λ1. We also treated fluo-5N as a ratiometric indicator by calculating its fluorescence ratio asλ3/λ1, using its 490 nm peak (λ3) and the isosbestic point of fura-2FF (λ1). This approach, with slight variations in wavelengths (see Table 1), was taken to calculate fluorescence ratios for pairs of indicators in each of the following three combinations (at the indicated concentrations): (1) fluo-5N (83 μm) with fura-2 (830 μm), (2) fluo-5N (250 μm) with fura-2FF (500 μm), and (3) fluo-5N (250 μm) with mag-fura-5 (500 μm).
For calibration and imaging experiments using indicator pairs, a filter cube containing a 505 nm dichroic and 515 long-pass emission filter was used. For experiments with either fura-2FF or mag-fura-5 alone, a 455 nm dichroic and 455 nm long-pass emission filter were used.
In situ measurement of Rmax. We first used this approach of measuring calcium with a pair of indicators to make in situ measurements of the calibration parameter Rmax for fura-2, fura-2FF, and mag-fura-5 directly in Purkinje cell dendrites. This parameter is subject to change when indicators are introduced into a cell. As calcium levels become sufficiently high to cause the ratio to approach Rmax (the fluorescence ratio at saturating Capost), measurements of Capost become extremely sensitive to Rmax. Hence, errors in measuring Rmax greatly distort measurements of calcium at concentrations above the KD of the indicator. To measure Rmaxin situ, we elevated calcium in the cell by bath applying the calcium ionophore ionomycin (Erdahl et al., 1995; Wang et al., 1998) in the presence of 10 mm calcium (Fig. 1C). By comparing fluorescence ratios of fura-2FF and fluo-5N (Fig. 1Ca,Cb), it is apparent that the plateau of the fura-2FF fluorescence ratio indicates that the indicator is saturated, because the fluorescence ratio of the lower-affinity fluo-5N continues to increase, indicating that calcium continues to rise. These results indicate that the fluorescence ratio of fura-2FF has reached its maximum, thereby providing us with an in situ measurement of Rmax. In this manner, using fluo-5N to verify that the higher-affinity indicator reached saturation, we measured Rmaxin situ for fura-2, fura-2FF, and mag-fura-5.
It was not possible to elevate calcium levels within cells sufficiently to saturate the fluo-5N response and directly measure Rmax for this indicator. However, fluo-5N has such a low affinity (Table 1) that calcium levels in our experiments do not lead to ratio changes that approach Rmax for fluo-5N, and as a result slight changes in Rmax have minor effects on Capost measurements with fluo-5N. We therefore used cuvette values of Rmax to convert fluo-5N fluorescence to calcium concentrations.
We also estimated the calibration parameter Rmin within cells by bath applying ionomycin in the presence of ∼10 nm external calcium (4 mm EGTA with no added Ca). We found that changes in Rmin from loading in cells were small. Moreover, estimates of the large calcium levels required to evoke endocannabinoid release were not highly sensitive to Rmin.
Cuvette measurements of Rmin and Rmax were made using the experimental imaging system with a 30 μm path length cuvette placed on the stage of the microscope. Our standard cesium intracellular solution was supplemented with indicators. Rmin was measured in the presence of 6 mm EGTA and no added calcium. Rmax was measured with 10 mm CaCl2.
For measuring cellular calcium transients with fura-2, fura-2FF, and mag-fura-5, we used the Rmin value obtained in a cuvette and the Rmax value obtained in situ. Because we could not experimentally verify that fluo-5N was saturating during ionomycin application, cuvette values of Rmin and Rmax were used.
Cellular imaging. Imaging was conducted using a Cooke Sensicam QE mounted on an Olympus BX51 upright microscope equipped with a 60×, 0.9 numerical aperture objective. The imaging protocols for measurement of calcium transients evoked by voltage steps consisted of sets of three images (20 msec exposure) acquired in succession at the three appropriate excitation wavelengths (λ1, λ2, λ3) (Fig. 1 B, Table 1) using a rapid wavelength switching monochromator (Till Photonics, Grafelfing, Germany). Groups of three images were collected every 250 msec for 10 sec. There was a 6 msec delay between successive images at each time point. The voltage step began after 1 sec of imaging. Fluorescence excitation was restricted to a small area including the Purkinje cell dendrites and an adjacent cell-free region that was used for background correction. Care was taken to exclude the soma from the area of excitation. Photo-bleaching of indicators during imaging protocols was <1%. Data analysis was performed with Vision Software (Till Photonics) and Igor (Wave-metrics, Lake Oswego, OR).
Calcium indicators were obtained from Molecular Probes (Eugene, OR) with the exception of fura-2FF, which was from Teflabs (Austin, TX). Ionomycin was from Molecular Probes. NBQX was from Tocris (Ellisville, MO), and bicuculline was from Sigma (St. Louis, MO).
Results
Our strategy for determining the calcium sensitivity of retrograde inhibition was to combine calcium measurements with voltage-clamp recordings of synaptic responses to quantify the relationship between postsynaptic calcium levels and retrograde endocannabinoid inhibition. This approach required two preliminary steps. First we needed to establish recording conditions for controlled elevation of calcium. This is difficult because Purkinje cell dendrites tend to fire calcium action potentials, which result in large calcium transients that do not depend on the length of the depolarization. Second, we needed to quantify postsynaptic calcium levels. Although there have been many studies of calcium signaling in Purkinje cells (Ross and Werman, 1987; Tank et al., 1988; Khodakhah and Armstrong, 1997; Finch and Augustine, 1998; Takechi et al., 1998; Wang et al., 2000), we needed to develop methods to determine absolute levels of calcium before we could ultimately determine the calcium dependence of retrograde signaling by endocannabinoids.
Using voltage steps to produce graded calcium transients
Our first goal was to control calcium influx. We took several steps to achieve voltage control sufficient to prevent spiking and permit regulation of the duration of calcium currents. First, we used a cesium-based internal solution with added TEA and QX-314 to block potassium and sodium channels. Second, we recorded from Purkinje cells from postnatal day (P) 10–P12 rats, which have electrically compact dendritic trees. The size of the dendritic arbor of the Purkinje cells used in this study is indicated by the fluorescence image of a cell loaded with the calcium-indicator fura-2FF (Fig. 2A). It was difficult to control calcium entry into Purkinje cells from animals older than P12, because brief depolarizations evoked dendritic calcium spikes. Finally, delivery of a 200 msec voltage ramp from -90 to -60 mV after the voltage step helped to suppress spiking (Fig. 2B, top). Under these conditions, voltage steps generally did not produce dendritic spikes and uncontrolled calcium entry (Fig. 2B, bottom). Recordings were terminated if cells spiked during voltage steps. Calcium transients were measured in dendritic regions selected from Figure 2A (right) and are shown here as ΔF/F (Fig. 2C). The amplitudes of calcium transients were graded, progressively increasing with the duration of the voltage step. These recording methods were used in subsequent experiments for indicator calibration and measurements of retrograde inhibition.
Quantification of dendritic calcium transients in Purkinje cells
We addressed two issues to quantify calcium levels with fluoro-metric indicators. First, we needed to select an indicator with appropriate calcium sensitivity. If the affinity were too high, the indicator would become saturated during calcium transients and it would not provide accurate estimates of Capost. If the affinity were too low, the indicator would have poor sensitivity. A second issue is that the properties of indicators can change both as a result of buffer composition and from cytoplasmic interactions within cells (Konishi et al., 1988; Harkins et al., 1993; Kurebayashi et al., 1993). Therefore we could not rely exclusively on in vitro calibrations, and we needed to determine the properties of the indicators within Purkinje cells.
As described in Materials and Methods, several steps were taken to address these concerns. Indicators were chosen for calibration on the basis of their affinities and spectral properties (Table 1). We selected three ratiometric indicators (fura-2, fura-2FF, and mag-fura-5) that can be calibrated using Equation 1 (Grynkiewicz et al., 1985). These indicators have very similar excitation peaks in the UV portion of the spectrum and similar emission peaks. However, they differ in their calcium affinities (131 nm, 7.7 μm, 38 μm, respectively, in our internal solution) and therefore are suited to measuring calcium over different concentration ranges. We also used fluo-5N, a single-wavelength dye, to evaluate the three ratiometric indicators. Fluo-5N has several useful properties. First, its extremely low affinity for calcium (KD = 410 μm) makes it unlikely to become saturated during dendritic calcium transients. Second, in cuvette experiments (see Materials and Methods) we found that protein binding did not significantly alter the affinity of fluo-5N (Fig. 1Ab), which suggests that the dissociation constant determined in a cuvette is appropriate for quantifying calcium levels within a cell, regardless of the extent to which fluo-5N is bound to protein. Finally, its spectral properties (excitation peak at 490 nm) allow it to be loaded simultaneously with the other indicators but excited independently. This allowed us to use fluo-5N to evaluate fura-2, fura2-FF, and mag-fura-2 by comparing independent measurements obtained from pairs of indicators (fura-2/fluo-5N, fura-2FF/fluo-5N, and mag-fura-5/fluo-5N), as shown in Figure 3.
Experiments in which Purkinje cells were loaded with fura-2 and fluo-5N revealed that the affinity of fura-2 is too high for our purposes. Fluorescence ratios for both indicators were measured simultaneously during a series of voltage steps of 50–1000 msec (Fig. 3Aa,Ab). Fura-2 exhibited nonsaturating ratio changes during voltage steps of 50 or 100 msec duration (Fig. 3Aa, gray traces), but because of its high affinity for calcium (KD = 131 nm), fura-2 saturated completely during voltage-step durations of 250–1000 msec (Fig. 3Aa). Fluo-5N, because of its low affinity for calcium (KD = 410 μm), did not exhibit a detectable fluorescence change during voltage steps of 50–100 msec (Fig. 3Ab, gray traces). Comparison of the normalized ratio changes indicates that fura-2 saturates completely with calcium signals that cause a <1% change in the fluorescence ratio of fluo-5N (Fig. 3Ac). Fluorescence ratios from fura-2 and fluo-5N were converted to calcium concentrations (Fig. 3Ad,Ae). Calcium measurements from these two indicators are difficult to compare because their dissociation constants differ by a factor of several thousand (Fig. 3Af). It is apparent, however, that fura-2 readily saturates and is not suitable for measuring dendritic calcium transients during voltage steps >100 msec.
Similar experiments revealed that fura-2FF is well suited to measuring the Capost transients that produce retrograde inhibition. Fluorescence ratios obtained during voltage steps of 50–1000 msec are shown for fura-2FF (Fig. 3Ba) and fluo-5N (Fig. 3Bb). The plot of the fluorescence ratio of fura-2FF versus the ratio of fluo-5N shows that there is some saturation of the fura-2FF response during calcium transients that cause large fluorescence signals in fluo-5N (Fig. 3Bc). We then compared the calcium concentrations estimated with fura-2FF (Fig. 3Bd) and fluo-5N (Fig. 3Be). Because peak calcium levels during long voltage steps saturated fura-2FF (Fig. 3Ba,Bc), we discarded calcium measurements that exceeded 77 μm (10-fold above the KD of fura-2FF). In this concentration range, calcium measurements made with fura-2FF and fluo-5N were in close agreement (Fig. 3Bf), as indicated by the linear regression of fura-2FF calcium values versus fluo-5N calcium values, which had a slope of 1.03. In a group of eight cells that were loaded with fura-2FF and fluo-5N, the slope of this regression was 0.99 ± 0.03. We conclude that fura-2FF and fluo-5N report calcium concentrations that are in agreement up to ∼80 μm.
Similar experiments suggested that mag-fura-5 is also suitable for quantifying calcium levels arising from Purkinje cell depolarization (Fig. 3C). After conversion of fluorescence ratios (Fig. 3Ca,Cb) to calcium concentrations (Fig. 3Cd,Ce), we compared calcium estimates obtained with mag-fura-5 and fluo-5N, considering values up to 10-fold greater than the KD of mag-fura-5, or 380 μm. Because of its low affinity (KD = 38 μm), mag-fura-5 does not saturate during calcium transients that cause large fluorescence changes in fluo-5N (Fig. 3Cc). For the cell shown in Figure 3C, the slope of the linear regression of calcium concentrations measured with fluo-5N versus mag-fura-5 had a slope of 1.04 (Fig. 3Cf). For a group of 10 cells examined in this manner, the slope of a regression of calcium measurements from magfura-5 versus fluo-5N was 1.02 ± 0.03. We conclude that measurements of calcium with mag-fura-5 and fluo-5N are in agreement at concentrations up to 380 μm. Moreover, despite its sensitivity to magnesium (see Materials and Methods), magfura-5 appears to be a suitable indicator for measuring calcium transients in Purkinje cells.
On the basis of these experiments, we conclude that fura-2FF and mag-fura-5 are suitable for measuring absolute calcium concentrations during depolarization of Purkinje cells. This conclusion rests on the close agreement of these two indicators with fluo-5N when the indicators were simultaneously loaded and imaged. On the basis of its fivefold lower KD, fura-2FF is better suited to the smaller calcium levels produced by brief depolarizations. Fura-2, however, readily saturates during brief depolarizations, and because of its high affinity it could not resolve calcium transients during depolarizations >100 msec. The agreement between the simultaneously loaded indicators suggests that despite the shift in KD seen for fura-2FF and mag-fura-5 caused by addition of BSA (Fig. 1, Table 1), such a shift does not occur in the cell. If we had been unable to determine the KD values of these indicators within Purkinje cells, our estimates of Capost on the basis of in vitro KD values would be lower bounds subject to the uncertainty of the extent to which the KD shifted to higher values within cells. Also, the linear relation between calcium measurements obtained with the simultaneously loaded dyes indicates that the experiments with ionomycin provide an accurate measurement of Rmax. For subsequent experiments that measured the calcium dependence of retrograde inhibition, we used a single calibrated indicator, either fura-2FF or mag-fura-5, to minimize the duration of fluorescence excitation and bleaching.
Capost dependence of DSE
We next combined synaptic recordings with calcium imaging to determine the relationship between peak dendritic Capost and the extent of retrograde inhibition by endocannabinoids at synapses onto Purkinje cells. Our experimental approach was to deliver a series of voltage steps to the Purkinje cell to elevate Capost to different levels and then determine the relationship between Capost and the resulting synaptic inhibition. Recent studies have suggested that metabotropic receptor activation may modulate the calcium dependence of cannabinoid release or even evoke calcium-independent cannabinoid release (Maejima et al., 2001; Ohno-Shosaku et al., 2001; Varma et al., 2001; Kim et al., 2002; Doherty and Dingledine, 2003). Here we focus on the calcium-dependent release of endocannabinoids in the absence of metabotropic receptor activation at CF, PF, and inhibitory synapses onto Purkinje cells.
Parallel fiber DSE
We first examined DSE at the parallel fiber to Purkinje cell synapse (Fig. 4). Each Purkinje cell receives inputs from thousands of parallel fibers that form synapses onto distal portions of its dendrites (Palay and Chan-Palay, 1974). Parallel fiber synapses were stimulated by placing a stimulus electrode in the molecular layer. PF EPSCs were evoked at 0.5 Hz, allowing us to monitor the onset and recovery of DSE after depolarizations. Short-term facilitation, the predominant form of plasticity at this synapse, decays with a time course of 200 msec and does not alter the amplitude of EPSCs evoked at this stimulus rate (Atluri and Regehr, 1996). In this initial series of experiments, Purkinje cells were loaded with fura-2FF. Calcium transients resulting from voltage steps of 100–1000 msec were imaged in distal portions of Purkinje cell dendrites in the region close to the stimulus electrode (Fig. 4A). Fluorescence ratios obtained from fura-2FF were converted to Capost using the calibration parameters described previously. After 10 stimuli to measure a stable baseline EPSC amplitude, voltage steps of varying durations were delivered to evoke DSE (Fig. 4B). Parallel fiber stimulation resumed 1 sec after the end of the voltage step. Inhibition of parallel fiber EPSCs was maximal at 3–7 sec after the voltage step (Fig. 4B, top panel, vertical dashed lines). To quantify DSE, we averaged three EPSCs preceding the depolarization (“pre”) and three EPSCs evoked 3–7 sec after the depolarization (“post”). In addition, three EPSCs were averaged at 55–59 sec after the depolarization (“rec”) (Fig. 4B, bottom panel). Peak DSE was defined as 1 - (post/pre) and ranges from 0 (no inhibition) to 1 (complete inhibition of EPSCs). Depolarizations ranging from 100 to 1000 msec evoked progressively larger dendritic calcium transients (Fig. 4A) and greater amounts of DSE (Fig. 4B). From each depolarization trial we obtained a value of peak Capost and a corresponding value of peak parallel fiber DSE. By repeating this protocol with varying durations of depolarization, we were able to plot the relationship between peak DSE and Capost for a single cell (Fig. 4C). By fitting the data points obtained in this manner to the Hill equation, we determined the Capost that produces half-maximal DSE (Ca0.5) and the Hill coefficient that describes the cooperativity of the relationship between Capost and DSE. For the cell shown in Figure 4, Ca0.5 was 14.2 μm, and the Hill coefficient was 2.5.
To determine the calcium dependence of PF DSE, this experimental approach was repeated on five neurons loaded with fura-2FF. The relation between peak Capost and peak DSE is shown in Figure 5A. Maximum DSE (DSEmax) after long voltage steps was 0.89 ± 0.02. Fits of the Hill equation have been made to data from each cell. For clarity, DSE values were normalized to the maximum DSE obtained in each cell so that data range between 0 and 1. For these experiments, Ca0.5 = 15.3 ± 1.1 μm, and the Hill coefficient was 2.3 ± 0.2.
To confirm these measurements of the calcium dependence and cooperativity of PF DSE, we performed further experiments using a different calcium indicator, mag-fura-5. Using the same experimental technique as illustrated in Figure 4, we measured the relationship between Capost and PF DSE in five neurons using mag-fura-5 to measure Capost (Fig. 5B). In these cells, Ca0.5 = 17.4 ± 1.0 μm, and the Hill coefficient was 1.2 ± 0.1. DSEmax was 0.91 ± 0.02. Measurements of Ca0.5 obtained with fura-2FF and mag-fura-5 for DSE at the parallel fiber synapse are statistically indistinguishable (p = 0.36; t test); however, estimates of the Hill coefficient obtained with fura-2FF and mag-fura-5 differed (p < 0.01, t test). The lower estimate of cooperativity obtained with mag-fura-5 may result from the difficulty of resolving small calcium transients, less than one-tenth of the KD of mag-fura-5 (<∼4 μm), which are critical for accurately measuring cooperativity. Fura-2FF on the other hand was able to resolve these small calcium signals and reported a higher degree of cooperativity in the relationship between postsynaptic Capost and parallel fiber DSE.
Climbing fiber DSE
We next determined whether the calcium dependence of DSE was similar at the other excitatory synapse onto Purkinje cells, the climbing fiber synapse (Fig. 6). Purkinje cells receive a single powerful climbing fiber synapse that consists of hundreds of release sites located in proximal regions of Purkinje cell dendrites (Palay and Chan-Palay, 1974; Xu-Friedman et al., 2001). Climbing fiber synapses are subject to inhibition by endocannabinoids (Kreitzer and Regehr, 2001b). We determined the calcium dependence of CF DSE using an approach that was similar to that used for PF DSE, but with slight modifications to take into account distinct properties of the synapses. Because prominent synaptic depression is observed with repetitive stimulation of CF synapses (Eccles et al., 1966), we stimulated the CF synapse at 30 sec intervals to allow full recovery from depression between successive EPSCs. This enabled us to measure inhibition of CF EPSCs that resulted from retrograde endocannabinoid inhibition and not from other forms of short-term plasticity such as vesicle depletion. Voltage steps were delivered to voltage-clamped Purkinje cells loaded with fura-2FF, and resulting calcium transients were imaged in proximal dendritic regions corresponding to the location of climbing fiber synapses (Fig. 6A). The voltage step was timed to precede the test stimulus by 5 sec (Fig. 6B, top). Voltage steps of 200–2000 msec produced progressively greater Capost and greater inhibition of CF EPSCs (Fig. 6B, bottom). We determined the baseline EPSC amplitude averaging the three EPSCs preceding the voltage step (pre) and used the EPSC after the voltage step (post) to quantify CF DSE as 1 - (post/pre).
To determine the relationship between peak Capost and DSE for climbing fiber synapses, we repeated this protocol of delivering voltage steps and measuring Capost and DSE (Fig. 6C). For the neuron shown in Figure 6A–C, half-maximal DSE occurred at Capost = 16.0 μm, and the Hill coefficient was 2.6. This procedure was repeated on six neurons (Fig. 6D). Each cell is indicated with a different plot symbol. For these cells, DSE at the climbing fiber synapse was half-maximal at Capost = 18.5 ± 0.7 and had a Hill coefficient of 2.0 ± 0.2. The values of Ca0.5 and the Hill coefficient at CF synapses are very similar to those described above for PF synapses. DSEmax at the climbing fiber synapse was 0.47 ± 0.08. This value is roughly half of the maximum DSE observed at parallel fiber synapses. Thus DSE at CF and PF synapses shows a similar low sensitivity to Capost and high cooperativity, although the maximal extent of endocannabinoid inhibition is less at CF compared with PF synapses.
Calcium dependence of DSI
Thus far we have seen that extremely high levels of calcium are required for DSE, almost 100 times as large as has been reported previously for DSI. This raises the question of whether there are large differences in the calcium levels needed to produce retrograde suppression of excitatory and inhibitory synapses or whether previous studies may have underestimated the calcium levels required for DSI. We therefore determined the Capost dependence of DSI.
Purkinje cells receive inhibitory synaptic inputs from stellate and basket cells that are located in the molecular layer of the cerebellum (Palay and Chan-Palay, 1974) and are sensitive to inhibition by endocannabinoids. Previous studies of cerebellar DSI have shown that endocannabinoids inhibit transmitter release from inhibitory nerve terminals by modulating presynaptic calcium channels (Kreitzer and Regehr, 2001a; Diana et al., 2002), altering miniature IPSC frequency (Takahashi and Linden, 2000) and reducing spontaneous firing of interneurons (Kreitzer et al., 2002). To facilitate comparison between different synapses, we determined the calcium dependence of DSI using techniques similar to those used for measurement of CF and PF DSE.
In contrast to the stereotypical localization of excitatory PF and CF synapses on Purkinje cells, the location of inhibitory synapses is more varied. Stellate cells form synapses over a wide area of the dendritic tree, and basket cell terminals form specialized arbors in axosomatic regions. In our experiments, IPSCs were evoked at 0.5 Hz with a stimulus electrode placed in the molecular layer. DSI was quantified by measuring inhibition of evoked IPSCs (eIPSCs) evoked 3–7 sec after the depolarization in the same manner as measurements of PF DSE. We measured calcium transients (Fig. 7A) and inhibition of eIPSCs (Fig. 7B) in Purkinje cells after depolarizations of different durations. Dendritic calcium transients were measured near the location of the stimulus electrode, but the location of the activated synapses was not certain. Data from one cell are shown in Figure 7A–C. Depolarizations that ranged from 150 to 1500 msec produced increasing DSI (Fig. 7B). As in previous experiments, voltage step protocols were repeated to determine the relationship between Capost and DSI. Results from a single neuron are shown in Figure 7C. Half-maximal DSI in this cell occurred at Capost = 14.5 μm, and the Hill coefficient was 1.0. Maximal DSI was 0.80. This experiment was repeated on four neurons (Fig. 7D). For these cells, DSI was half-maximal at Capost = 16.0 ± 0.7 μm and had a Hill coefficient of 1.3 ± 0.1. The maximal extent of DSI was 0.82 ± 0.01. We conclude that the calcium sensitivity of DSI is very similar to that of both PF and CF DSE.
Discussion
Our primary finding is that half-maximal retrograde inhibition by endocannabinoids occurs with postsynaptic calcium levels of ∼15 μm at excitatory and inhibitory synapses onto Purkinje cells. In addition, DSE at PF and CF synapses had Hill coefficients of 2–2.3, indicating a moderate degree of cooperativity between Capost and retrograde endocannabinoid inhibition. The high requirement for calcium shown in this study suggests that release of endocannabinoids by Purkinje cells may occur only in localized regions of dendrites that experience large elevations of calcium.
Quantifying calcium levels within Purkinje cells
The approach that we took to quantify dendritic calcium transients in Purkinje cells has general implications for the fluoro-metric measurements of calcium levels for cells in brain slices. The key steps were to perform both cuvette and in situ calibrations and to assess the responses of ratiometric indicators by simultaneously monitoring responses of the single-wavelength indicator fluo-5N. First, we selected the indicators with the appropriate sensitivities and found that fura-2FF and mag-fura-5 were both appropriate for measuring calcium transients evoked by depolarizations of sufficient duration to produce retrograde synaptic inhibition. In contrast, fura-2 was not appropriate because it readily saturated during brief voltage steps as a consequence of its high affinity. The dissociation constant of fura-2FF (7.7 μm) makes it well suited to measuring calcium transients <5 μm that enabled us to resolve the steep cooperativity of the relationship between Capost and DSE. Second, we determined Rmaxin situ for fura-2, fura2-FF, and mag-fura-5 by using a calcium ionophore to increase Capost and used fluo-5N to ensure that Capost had been elevated sufficiently. Finally, we assessed whether binding to protein might affect the sensitivity of indicators within Purkinje cells. Although in cuvette measurements the addition of protein increased the KD of fura-2, fura-2FF, and mag-fura-5, the KD of fluo-5N was unaffected. We took advantage of this insensitivity of fluo-5N to determine the appropriate dissociation constants for fura-2 FF and mag-fura-5 in Purkinje cells. Our results suggest that neither of these indicators undergoes a shift in affinity within Purkinje cells. By eliminating a three- to fourfold uncertainty in the KD, we were better able to determine Capost.
Calcium dependence of retrograde inhibition
We found similarities and differences in the Capost dependence of retrograde inhibition at the three synapses that we studied. Although Ca0.5 was similar at all of the synapses (∼15 μm), the Hill coefficients and extent of retrograde inhibition differed. For CF and PF synapses the Hill coefficient was 2–2.3, whereas it was only 1.3 for inhibitory synapses. The maximal extent of inhibition ranged from a reduction of 89% for PF synapses to 82% for inhibitory synapses and 47% for CF synapses.
In considering the factors underlying such differences in the extent and Capost dependence of retrograde inhibition, it is necessary to consider the many steps between the initial elevation in Capost and the reduction in synaptic strength. (1) Endocannabinoids are synthesized and released in a calcium-dependent manner (Devane et al., 1992; Di Marzo et al., 1994; Stella et al., 1997). (2) After their liberation, endocannabinoids must diffuse to the presynaptic terminal. Factors determining the amplitude and duration of endocannabinoid signal available to activate presynaptic receptors include diffusion, uptake, and degradation (Cravatt et al., 1996; Beltramo et al., 1997; Di Marzo et al., 1998). (3) Endocannabinoids then bind to presynaptic CB1 receptors, which leads to (4) the inhibition of presynaptic calcium channels (Kreitzer and Regehr, 2001b; Diana et al., 2002; Howlett et al., 2002), which in turn (5) reduces the release of neurotransmitter (Levenes et al., 1998; Takahashi and Linden, 2000; Kreitzer and Regehr, 2001a,b), ultimately leading to a reduction in the postsynaptic response. Although little is known about steps 1–3, much is known about the relationship between presynaptic calcium influx (Capre) and EPSC amplitude at PF and CF synapses. The relationship between Capre and EPSC amplitude at PF and CF synapses is different. Notably, at the PF synapse there is a power-law dependence of the EPSC on Capre (Mintz et al., 1995; Sabatini and Regehr, 1997), whereas at the CF synapse this relationship exhibits saturation (Fig. 8Aa,Ba) (Foster et al., 2002). These differences in the dependence of EPSCs on Capre will differentially influence the manner in which modulation of Capre by endocannabinoids ultimately affects the EPSC at PF and CF synapses.
We therefore assessed how the relationship between Capre and EPSCs influences the Hill coefficient, the Capost sensitivity, and the extent of retrograde inhibition at PF and CF synapses. For PF and CF synapses, we used the known relationship between Capre and EPSC amplitude (Fig. 8Aa,Ba) and our experimentally determined measurement of the dependence of DSE on Capost (Fig. 8Ab,Bb) to express modulation of Capre as a function of Capost during retrograde inhibition (Fig. 8Ac,Bc). Although multiple mechanisms may contribute to endocannabinoid modulation, at the PF and CF synapses the effects of endocannabinoids reflect a reduction of presynaptic calcium influx (Kreitzer and Regehr, 2001b; S. Brown and W. Regehr, unpublished observations), and there is no evidence for cannabinoid-mediated effects downstream of calcium influx (Takahashi and Linden, 2000). Therefore, this analysis allows us to examine the ability of endocannabinoids to directly modulate presynaptic calcium influx. Although DSEmax was 89 and 47% for PF and CF synapses, respectively, the underlying retrograde inhibition of Capre was 52 and 42% for PF and CF synapses. This indicates that although inhibition of presynaptic calcium is similar at PF and CF synapses, the difference in the relationship between Capre and EPSC amplitudes produces large differences between DSEmax at PF and CF synapses.
We also found that the relationship between Capre and EPSC amplitude at PF versus CF synapses altered the apparent Capost sensitivity of retrograde inhibition. At PF synapses, Ca0.5 for inhibition of EPSCs and for inhibition of Capre was 15.3 and 20.9 μm, respectively. At CF synapses, Ca0.5 for inhibition of EPSCs and Capre was 18.5 and 12.3 μm, respectively. This comparison of postsynaptic calcium required for half-maximal inhibition of EPSCs versus Capre revealed that at a saturating synapse such as CF synapses, the Capost required for half-maximal inhibition of Capre compared with inhibition of EPSCs shifts to lower values and at a nonsaturating synapse such as PF synapses the shift is to higher values. Thus, the relationship between Capre and EPSC amplitude alters the Capost dependence of retrograde inhibition, but the effects are small.
The final issue that we addressed was the source of cooperativity in endocannabinoid-mediated retrograde inhibition. We considered the possibility that the steep relationship between Capre and the EPSC might account for steep Hill coefficients of 2–2.3 for the relationship between Capost and DSE. On the basis of the analysis in Figure 8, however, we found that the cooperativity changes only slightly when we consider the effects of Capost on Capre, shifting the Hill coefficients to 1.8–2.0, thus indicating that the cooperativity of DSE occurs at a stage before presynaptic calcium influx. This is consistent with previous studies that found varying degrees of cooperativity (Hill coefficients of 1.2–3) in channel modulation by neurotransmitters (Ito et al., 1992; Krapivinsky et al., 1995; Sodickson and Bean, 1996, 1998). The calcium dependence of endocannabinoid synthesis and release may also be cooperative, contributing to the steep dependence of retrograde inhibition on postsynaptic calcium.
The Hill coefficient that we measured for the relationship between Capost and DSI was 1.3, consistent with values of 1.3 and 1 reported previously for DSI in hippocampus and cerebellum, respectively (Glitsch et al., 2000; Wang and Zucker, 2001). This suggests the intriguing possibility that there may be less cooperativity at inhibitory relative to excitatory synapses. However, because the locations of interneuron synapses are difficult to determine, there is ambiguity regarding the precise region of dendrites in which to measure calcium. This may introduce uncertainty to the measurements of calcium dependence of DSI.
Functional implications
The high Capost levels required for retrograde inhibition and the Hill coefficients of ∼2 for DSE at PF and CF synapses indicate that calcium-dependent retrograde inhibition at excitatory synapses has a high activation threshold. For example, an elevation of Capost to 5 μm results in only 6% inhibition of PF EPSCs, whereas an elevation of Capost to 25 μm inhibits PF EPSCs by 67%.
Additional factors may also be important under physiological conditions. To accurately determine the Capost dependence of retrograde inhibition, we performed experiments at room temperature in young animals in the absence of metabotropic receptor activation. Considering the many steps between elevation of Capost and retrograde inhibition, it is likely that the Capost dependence of retrograde inhibition differs somewhat at physiological temperatures. The Capost dependence may also be developmentally regulated. Last, recent studies point to a role for metabotropic glutamate and acetylcholine receptors in potentiating Capost-dependent release in evoking release through calcium-independent mechanisms (Varma et al., 2001; Kim et al., 2002; Ohno-Shosaku et al., 2002; Yoshida et al., 2002).
To more fully appreciate the functional implications of the Capost dependence of endocannabinoid release, it is necessary to know the magnitude and spatial location of dendritic calcium signals under different physiological conditions. Numerous studies have shown that calcium signals in Purkinje cells arise from spontaneous activity and synaptic activation, both of which can open voltage-gated calcium channels and release calcium from internal stores (Tank et al., 1988; Mikawa et al., 1997; Finch and Augustine, 1998; Takechi et al., 1998). These studies indicate that calcium signals in Purkinje cell dendrites can be widespread and that there are also mechanisms that provide spatially localized signaling. Many of these studies, however, focused on spatial and temporal aspects of calcium signaling, and it remains an open question just how high Capost gets under physiological conditions and what means of Purkinje cell activation can elevate calcium levels sufficiently to reach the tens of micromolar levels sufficient to evoke endocannabinoid release.
Footnotes
This work was supported by National Institutes of Health Grant RO1-NS32405-01. We thank Michael Beierlein, Dawn Blitz, Solange Brown, Kelly Foster, Anatol Kreitzer, Patrick Safo, and Matthew Xu-Friedman for comments on this manuscript and Kim Irwin for technical assistance.
Correspondence should be addressed to Wade Regehr, 220 Longwood Avenue, Harvard Medical School, Boston, MA 02115. E-mail: wade_regehr{at}hms.harvard.edu.
Copyright © 2003 Society for Neuroscience 0270-6474/03/236373-12$15.00/0