Calcium-binding proteins such as calretinin are abundantly expressed in distinctive patterns in the CNS, but their physiological function remains poorly understood. Calretinin is expressed in cerebellar granule cells, which provide the major excitatory input to Purkinje cells through parallel fibers. Calretinin-deficient mice exhibit dramatic alterations in motor coordination and Purkinje cell firing recorded in vivo through unknown mechanisms. In the present study, we used patch-clamp recording techniques in acute slice preparation to investigate the effect of a null mutation of the calretinin gene on the intrinsic electroresponsiveness of cerebellar granule cells at a mature developmental stage. Calretinin-deficient granule cells exhibit faster action potentials and generate repetitive spike discharge showing an enhanced frequency increase with injected currents. These alterations disappear when 0.15 mm of the exogenous fast-calcium buffer BAPTA is infused in the cytosol to restore the calcium-buffering capacity. A proposed mathematical model demonstrates that the observed alterations of granule cell excitability can be explained by a decreased cytosolic calcium-buffering capacity resulting from the absence of calretinin. This result suggests that calcium-binding proteins modulate intrinsic neuronal excitability and may therefore play a role in information processing in the CNS.
Calcium regulates a large variety of neuronal functions, including neurotransmitter release, ionic channel permeability, enzyme activity, and gene transcription. Hence, the cytosolic calcium concentration must be tightly regulated. Cytoplasmic calcium-binding proteins play a key role in this regulation leading to specific adjustments of neuronal signaling. Among these, calretinin is the only calcium-binding protein known to be expressed in cerebellar granule cells (Résibois and Rogers, 1992; Marini et al., 1997), whereas the structurally related calbindin is exclusively expressed in Purkinje cells. Granule cells form the largest neuronal population in the mammalian brain. They process information entering into the cerebellar cortex through the mossy fibers (Ito, 1984) and convey the major excitatory afference to Purkinje cells through the parallel fibers. The involvement of the cerebellum in motor coordination has long been recognized and, interestingly, it has been shown recently that calretinin-deficient mice (Cr-/-) were impaired in motor coordination tests and displayed alterations in Purkinje cell activity recorded in vivo (Schiffmann et al., 1999). Because calretinin is not expressed in Purkinje cells, the latter study lacked direct evidence for intrinsic cerebellar electrophysiological alteration at the cellular level, resulting from the Cr-/- mutation. This is particularly relevant, because granule cells show a calcium-dependent regulation on their intrinsic excitability (D'Angelo et al., 1997, 1998, 2001; Gabbiani et al., 1994), so that alteration in calcium buffering is expected to affect action potential generation. Thus, in addition to addressing a question specific to cerebellar physiology, Cr-/- mice may provide a model for understanding the regulation of calcium-dependent control of neuronal excitability.
In this study, we directly address this question by recording granule cells using the perforated patch configuration of the patch-clamp technique in brain slices. We show that the electroresponsiveness of granule cells from Cr-/- mice is altered. Furthermore, we show that intrinsic excitability of granule cells from Cr-/- mice can be restored to its control level by the infusion of an exogenous calcium buffer in the cytosol by means of the whole-cell configuration. Finally, we present a mathematical model providing a link between the observed alterations in granule cell electroresponsiveness and the decreased cytosolic Ca2+-buffering capacity, resulting from the absence of calretinin. Our results suggest a critical role for calretinin in regulating granule cell excitability and signal coding at the input stage of the cerebellum.
Materials and Methods
Mice. Sex- and age-matched wild-type (WT) and Cr-/- mice back-crossed on a C57BL/6 genetic background (f7) were used in all experiments (Schurmans et al., 1997).
Electrophoresis and immunoblotting. One dimensional polyacrylamide gel electrophoresis was performed as described previously (Galas et al., 2000). Cerebellum from WT and Cr-/- were homogenized in lysis buffer (10 μl/mg of tissue) (M-PER; Pierce, Rockfold, IL) containing a protease inhibitor mixture (Complete; Roche Molecular Biochemicals, Vilvorde, Belgium). Samples were stored at -20°C until analysis. Protein concentrations were determined using MicroBCA Protein Assay (Pierce). Equal amounts of protein (40 μg or 250 μg for the densitometrical analysis) were denatured in 2× Laemmli buffer at 100°C for 5 min and then separated on an SDS-PAGE. Proteins were transferred to supported nitrocellulose (Bio-Rad, Hercules, CA) at 250 mA for 90 min at 4°C. The membrane was blocked with 5% BSA and 0.1% Tween 20 in PBS buffer. The membrane was then incubated with the primary antibody anti-calretinin (1:10000; Swant, Bellinzona, Switzerland) or the primary antibody anti-actin (1:2000; Sigma, St. Louis, MO) overnight at 4°C. After washing in 0.1% Tween 20 containing PBS buffer, the membrane was incubated with the HRP-labeled secondary antibody (anti-rabbit IgG; NEN, Boston, MA) at a concentration of 0.1 μg/ml for 60 min at room temperature. Immunoreactive bands were visualized by chemiluminescent ECL Plus western blotting reagents (Amersham Biosciences, Buckinghamshire, UK). For the densitometrical analysis, the bands on the ECL hyperfilm (Amersham Biosciences) corresponding to calretinin and actin were scanned and quantified using the public domain NIH Image β 4.0.2 program (National Institutes of Health, Bethesda, MD). The ratio between calretinin and actin in the cerebellum is considered 100%.
Immunocytochemistry. Mice were killed by decapitation after a slight ether anesthesia, and brains were removed and fixed in 0.1 m phosphate-buffered 4% paraformaldehyde freshly prepared for 24 hr at 4°C. Brains were then rinsed in several changes of 0.1 m PBS, pH 7.4, and cryoprotected in sucrose. Cryostat-cut 30 μm thick brain coronal sections were processed free-floating. After quenching of the endogenous peroxydases, the floating sections were incubated for 1 hr in 10% swine normal serum (Hormonologie Laboratoire, Marloie, Belgium). Sections were then incubated overnight at room temperature with anti-calretinin polyclonal antiserum (1/1000; Swant). Thereafter, they were successively incubated with swine anti-rabbit gammaglobulins (1/30; Dako, Glostrup, Denmark) and rabbit PAP complex (1/300; Dako). The peroxydase activity was revealed by diaminobenzidine in the presence of hydrogen peroxide.
Histology. Four wild-type and four calretinin-deficient mice were killed, and their brains were removed and fixed overnight in 4% paraformaldehyde. After a wash in 0.1 m PBS and consecutive 24 hr incubations in 10, 20, and 30% sucrose solutions, brains were frozen in isopentane. Sagittal sections (15 μm) were performed at the level of the vermis using a cryotome (Leica, Wetzlar, Germany) and stained using a Nissl procedure. For each mouse, four digitized fields (100 × 100 μm) were randomly generated from four adjacent stained sections at a 63× magnification using a confocal microscope (Bio-Rad). On each field, we measured the surface of 10 individual granule cells using the Scion-NIH Image software. The mean apparent area of granule cells was then determined for each mouse, and the difference between genotypes was compared using unpaired Student's t test.
Whole-cell patch-clamp recordings. Cerebellar granule cells were recorded in acute cerebellar slices obtained from 15- to 31-d-old mice, when the excitable response has assumed its mature pattern (D'Angelo et al., 1997). Slice preparations were performed as described previously (D'Angelo et al., 1998). Current-clamp recordings were taken from granule cells in the internal granular layer at room temperature using the “blind patch” approach. All recordings were made with an EPC-8 amplifier (Heka Elektronik, Lambrecht/Pfalz, Germany) in the fast current-clamp mode. Membrane potential signal was filtered using an 8-pole Bessel low-pass filter (Frequency Devices, Haverhill, MA) at a cutoff frequency of 2 KHz and subsequently digitized at 10 kHz using the acquisition software Pulse (Heka) in combination with an ITC-16 analog-to-digital converter (InstruTech, Port Washington, NY). Patch pipettes were pulled from borosilicate glass capillaries (Harvard Apparatus, Edenbridge, UK) and had a resistance of ∼6 MΩ. For whole-cell recordings, the pipette solution contained the following (in mm): 126 KGluconate, 0.05 CaCl2, 0.15 BAPTA, 4 NaCl, 1 MgSO4, 15 glucose, 5 HEPES, 3 MgATP, and 0.1 GTP, pH adjusted at 7.2 with KOH. Recordings were started 5 min after obtaining the whole-cell configuration to allow diffusion of the calcium buffer BAPTA into the cytosol. In some experiments, the perforated patch configuration of the patch-clamp technique (Horn and Marty, 1988) was used to maintain the endogenous Ca2+ buffers. The pipette solution contained the following (in mm): 80 K2SO4, 10 NaCl, 15 glucose, and 5 HEPES, pH adjusted at 7.2 with KOH, and 100 μg/ml nystatin. The extracellular solution contained the following (in mm): 120 NaCl, 2 KCl, 2 CaCl2, 1.19 MgSO4, 26 NaHCO3, 1.18 KH2PO4, and 11 glucose equilibrated with 95% O2/5% CO2, pH 7.4. Intrinsic excitability was investigated by setting resting membrane potential at -80 mV and injecting 1 sec steps of depolarizing current (from 2-30 pA in a 2 pA increment). Action potential frequency was measured by dividing the number of interspike intervals by the time interval between the first and the last spike. The amplitude of the action potential undershoot was estimated as the difference between the threshold of spike prepotential and the minimal potential during spike afterhyperpolarization. When EPSPs were measured, the extracellular solution was supplemented with 25 μm picrotoxin to block GABAergic inhibition generated by Golgi cell synapses (Armano et al., 2000). The mossy fibers were stimulated at a frequency of 0.1 Hz with a bipolar tungsten electrode via a stimulus isolation unit. In our EPSP-recording protocol, 200 μs step current pulses were applied to activate the mossy fibers while measuring EPSPs from a resting potential of -70 mV. The maximal effective stimulus intensity was used in all EPSP recordings. The time to peak was estimated as the interval between the beginning of the stimulating pulse and the maximum of the spike upstroke during the EPSP. The time constant of EPSP relaxation, τdecay, was estimated by monoexponential fitting using the Levenberg-Marquardt algorithm implemented in IGOR (WaveMetrics, Lake Oswego, OR). Only EPSPs giving rise to action potentials were used to evaluate the number of spikes per EPSP. Passive cellular parameters were extracted in voltage clamp by analyzing current relaxation induced by a 10 mV step change from a holding potential of -70 mV as described previously (D'Angelo et al., 1995). In the perforated patch configuration, access resistance was monitored to ensure that voltage attenuation in current-clamp mode was always <10%. Data are reported as means ± SEM, and statistical comparisons were performed using unpaired Student's t test [p > 0.05, not significant (NS)].
Minimal model of the granule cell. Calcium dynamics have a profound influence on Ca2+-activated K+ current (IK-Ca) activation, thereby regulating spike discharge. A typical approach to investigate Ca2+ dynamics (Traub and Llinas, 1979) is to model Ca2+ and IK-Ca and to adapt Ca2+ dilution and removal to match the firing pattern. This approach has been applied previously to cerebellar granule cell model (Gabbiani et al., 1994; Maex and Schutter, 1998; D'Angelo et al., 2001). Here, to focus our attention on Ca2+ dynamics, we reduced the number of gating variables and currents involved in action potential generation. We also modified the equation governing the Ca2+ dynamics to take into account variations in the concentration of endogenous Ca2+ buffers.
Because granule cells have a compact electrotonic structure (Silver et al., 1992; D'Angelo et al., 1993, 1995, 2001), a single compartment model was used. Following the classical Hodgkin and Huxley (1952) approach, the membrane can be considered as a leaky capacitor, and the membrane potential dynamics are governed by the following current balance equation: 1 where Cm is cell capacitance, INa is a voltage-dependent Na+ current, IK-V is a delayed rectifier K+ current, ICa is a high-threshold voltage-dependent Ca2+ current, and IK-Ca is a Ca2+-activated K+ current. These ionic currents have been shown to be the core of action potential generation in cerebellar granule cells (D'Angelo et al., 1998) when the excitable response has assumed its mature pattern (D'Angelo et al., 1997). The interplay between INa and IK-V is the basic mechanism, giving rise to the action potentials, and the presence of ICa and IK-Ca allows coupling between intracellular calcium dynamics and membrane potential dynamics. We also reduced the number of gating variables by exploiting the difference in their time scale compared with the time scale of the evolution of V. The complete expressions for the different ionic currents are given by the following: 2 3 4 5 where the evolution of the gating variables h, s, and a is governed by the following relationships: 6 7 8 For a given gating variable, x, the steady state function, x∞, and the relaxation time constant, τx, are related to the corresponding first-order rate functions, αx and βx, by the following relationships: 9 Channel parameter values are given in Table 1. The gating parameters for INa, IK-V, ICa, and the cell capacitance, (Cm = 3.14 pF) are the same as those in Maex and Schutter (1998). The conductances of the different currents and the parameters related to the calcium dependence of IK-Ca were adapted to obtain current intensities during the action potential similar to those in the models of Gabbiani et al. (1994) and D'Angelo et al. (2001).
To complete the model, the following balance equation gives the evolution of the free calcium concentration (in μm) in a submembrane shell of thickness d in a cell of surface area A. 10 where the dimensionless parameter f represents the calcium-buffering capacity of the cytosol inside the submembrane shell, resulting from the presence of fast calcium-binding proteins (binding is assumed to be instantaneous). In particular, calretinin is a fast-calcium buffer, the mean free lifetime for a Ca2+ ion in the presence of a physiological concentration of the protein of the μs order (Edmonds et al., 2000), third order of magnitude faster than the time scale of the evolution of the system variables. Therefore, any alteration in the calretinin level corresponds to a modification of f. Moreover, because this parameter can be seen as the ratio of free Ca2+ concentration to bound Ca2+ concentration (Chay and Keizer, 1983; Gall and Susa, 1999; Gall et al., 1999), a decrease in cytosolic Ca2+-buffering capacity, attributable to the absence of calretinin, can be mimicked by an increase in the value of f. In the absence of data for cerebellar granule cells, the numerical values for the cytosolic Ca2+-buffering capacity (in the order of 0.01) have been set in agreement with values reported for other neuronal types (Tatsumi and Katayama, 1993; Stuenkel, 1994; Helmchen et al., 1996), with the notable exception of cerebellar Purkinje cells, which show a calcium-binding ratio of an order of magnitude higher (Fierro and Llano, 1996). The parameter βCa describes Ca2+ removal corresponding to diffusion, action of ionic pumps, and slow buffers. Calcium dynamics were adapted to yield Ca2+ transients in the μm range, similar to those reported from Gabbiani et al. (1994). Parameter values are A = 314 (μm)2, d = 0.084 μm, βCa = 10 msec-1, and F is the Faraday constant.
Equations 1, 6, 7, 8, and 10 constitute our minimal model for the cerebellar granule cell. The equations of the model are numerically solved using a fourth-order Runge-Kutta method, as implemented in the subroutine D02BBF of the Numerical Algorithms Group library. Computations are performed on an SGI R12000 workstation (Silicon Graphics, Mountain View, CA). To test the accuracy of our minimal model, simulations were also run with similar results (data not shown), with models by Maex and Schutter (1998) and D'Angelo et al. (2001) modified to include Ca2+ dynamics as described in Equation 10.
Enrichment of calretinin in the mouse cerebellum
Calcium-binding proteins are differentially expressed depending on the neuronal type. Figure 1A shows that calretinin is particularly enriched in the mice cerebellum as compared with other brain areas such as the hippocampus (Schurmans et al., 1997). In this study, the role of calretinin was investigated by using calretinin-deficient mice. Calretinin was revealed by Western blotting with anti-calretinin antiserum. No calretinin signal could be detected in the cerebellum of Cr-/- mice as expected for a null mutation (Fig. 1B). In addition, immunocytochemistry revealed no calretinin immunoreactivity in the cerebellar granule cell layer of Cr-/- mice (Fig. 1C).
Alteration of granule cell intrinsic membrane excitability
Unless otherwise stated, recordings were made using the perforated patch technique to minimize unwanted alteration of the endogenous Ca2+-buffering capacity. No difference in membrane resistance or capacitance could be observed in Cr-/- compared with wild-type granule cells (Table 2), indicating the absence of granule cell morphological alterations. This was confirmed by the quantitative data we obtained using histological staining and confocal microscopy showing that the mean granule cell-apparent area was similar in both WT (21.8 ± 1.05 μm2; n = 4) and calretinin-deficient mice (20.69 ± 1.28 μm2; n = 4; NS). Moreover, no difference in subthreshold currents measured at -70 and -80 mV could be detected in Cr-/- mice (-1.9 ± 2.0 pA and -7.8 ± 2.0 pA, respectively; n = 8) as compared with wild-type granule cells (-2.2 ± 5.7 pA and -7.9 ± 4.8 pA, respectively; n = 5; NS).
Intrinsic granule cell electroresponsiveness was investigated in current-clamp recordings. The resting potential of WT and Cr-/- granule cells was not significantly different (-64.8 ± 5.5 mV, n = 5 vs -65.9 ± 5.3 mV, n = 8; NS). Neither WT nor Cr-/- granule cells generated spontaneous action potentials. Active cell membrane properties were evaluated by measuring the voltage response while injecting steps of depolarizing current of increasing intensities in the granule cell soma. Above a critical value of the injected current (4.4 ± 1.6 pA for WT, n = 5 vs 4.9 ± 1.0 pA for Cr-/-, n = 8; NS), fast repetitive spiking was obtained with a threshold of spike prepotential of -58.0 ± 2.3 mV for WT granule cells (n = 5) and -58.7 ± 2.0 mV for Cr-/- granule cells (n = 8; NS). Action potentials occurred in regular trains showing little or no adaptation (Fig. 2A), and frequency increased with the intensity of the injected current. The average frequency was measured over the entire duration of current injection (1 sec) and was used to construct current-frequency plots (Fig. 2B). At low current intensities, the current-frequency plots were interpolated with a straight line. Because the threshold current and maximal frequency varied substantially from cell to cell, the evaluation of the slope factor of the linear part of current-frequency plots was used as a normalized measure of excitability. Using such analysis, we observed a significant increase in the slope of the current-frequency plots (4.8 ± 0.2 Hz·pA-1 for WT, n = 5 and 6.6 ± 0.7 Hz·pA-1 for Cr-/-, n = 8; p < 0.05), indicating that the excitability of Cr-/- granule cells was increased (Fig. 2C). In addition, Cr-/- granule cells showed a 23% decrease in the action potential half-width (Fig. 2D) evaluated at the threshold potential in which fast repetitive spiking is obtained (1.02 ± 0.06 msec for WT, n = 5 and 0.78 ± 0.06 msec for Cr-/-, n = 8; p < 0.05). Besides changes observed in Cr-/- mice, it should be noted that both the spike shape and frequency of spike discharge in WT mice were similar to those reported previously (D'Angelo et al., 1998). The action potential undershoot amplitude was deeper for Cr-/- than WT granule cells, although this change did not reach statistical significance (-4.0 ± 0.7 mV for WT, n = 5 and -6.4 ± 1.7 mV for Cr-/-, n = 7; p = 0.2406; NS). The fact that the latter change fails to be statistically significant probably reflects the higher sensitivity of this parameter toward cell-to-cell variations, compared with more robust temporal parameters such as the slope factor of current-frequency relationship or the action potential half-width.
Restoration of neuronal excitability by an exogenous buffer
The lack of calretinin in Cr-/- mice is likely to alter cytosolic Ca2+ buffering in cerebellar granule cells. To test this hypothesis, we measured the ability of an exogenous Ca2+ buffer to restore similar levels of excitability in the mutant mice as in the WT mice. Because calretinin and BAPTA have similar Ca2+-binding kinetics (Roberts, 1993; Edmonds et al., 2000), we perfused the cells with 0.15 mm BAPTA using the whole-cell configuration of the patch-clamp technique. In these conditions, no differences between the mutant and the WT mice could be found (Fig. 3) either concerning spike-discharge frequency (4.4 ± 0.3 Hz·pA-1 for WT, n = 5 and 4.4 ± 0.4 Hz·pA-1 for Cr-/-, n = 5) or action potential half-width (0.59 ± 0.07 msec for WT, n = 50 and 0.64 ± 0.06 msec for Cr-/-, n = 5). The difference in action potential half-width observed in WT granule cells using the whole-cell configuration, compared with the corresponding value measured using the perforated patch method, is probably a consequence of the difference between the two recording conditions. Indeed, the higher series resistance obtained in perforated patch configuration determines a voltage attenuation (D'Angelo et al., 1995) that could slightly alter the spike shape but not the spike count. Therefore, we can compare the slope factors obtained in the two recording conditions. Our results show that the presence of 0.15 mm BAPTA in the pipette is sufficient to restore an excitability level in the Cr-/- granule cell similar to the one measured in the WT using the perforated patch method. This indicates that the endogenous cytosolic Ca2+-buffering capacity in the cerebellar granule cell is approximately equivalent to 0.15 mm BAPTA.
We also studied the effect of a decreased calcium-buffering capacity on granule cell excitability by performing whole-cell recording with low BAPTA concentration (0.01 mm) on wild-type mice. In these conditions, we observe an important spike-frequency accommodation (data not shown) in all of our recordings (n = 7). The absence of a stable repetitive firing pattern prevented us from estimating accurately the average action potential frequency at a given injected current and thus to construct current-frequency plots. Therefore, the intrinsic excitability cannot be evaluated in a similar way as the previous recordings. In addition, the spike-frequency accommodation alters the action potential kinetics, preventing us from accurately measuring the action potential half-width. Such modulation of spike-frequency accommodation by changes in calcium-buffering capacity has been shown to occur in hippocampal neurons (Lancaster and Nicoll, 1987). The absence of a similar increase in spike-frequency accommodation in the perforated patch recording on the Cr-/- granule cells indicates the existence of compensatory mechanisms at work in the mutant.
Relationship between calretinin deficiency and altered excitability
To investigate the role of cytosolic Ca2+ buffering on intrinsic granule cell excitability, we used a mathematical model (see Materials and Methods). This model should be seen as a minimal model allowing us to qualitatively understand the impact of variations in cytosolic Ca2+-buffering capacity on the electroresponsiveness of an excitable cell. Nevertheless, parameter values have been chosen to reproduce the relevant aspects of the cerebellar granule cell electroresponsiveness. Figure 4A shows the model response to an intrasomatic injection of a 14 pA depolarizing current. In our model, calcium-buffering capacity, resulting from the action of fast-calcium buffers, is represented by the dimensionless parameter f giving the ratio of free Ca2+ concentration to bound Ca2+ concentration (Chay and Keizer, 1983; Gall and Susa, 1999; Gall et al., 1999). A decrease in cytosolic Ca2+-buffering capacity, resulting from the absence of calretinin, can be simulated by an increase in the value of f. The observed alterations in Cr-/- mice can be mimicked by a fourfold decrease in the cytosolic Ca2+-buffering capacity, raising the action potential frequency from 55.1 (f = 0.01) to 105.7 Hz (f = 0.04) in good agreement with the experimental data (Fig. 2A). The linear slope increase in current-frequency plots from 3.0 Hz·pA-1 for f = 0.01 to 5.2 Hz·pA-1 for f = 0.04 (Fig. 4B) is also close to the experimental values. Figure 5 shows the effect of a fourfold decrease of the cytosolic Ca2+-buffering capacity on membrane voltage, submembrane Ca2+ concentration, and membrane currents during an action potential. The action potential half-width undergoes a 41% decrease when f is increased from 0.01 to 0.04. This action potential shortening reflects a more pronounced activation of IK-Ca because of faster Ca2+ dynamics taking place when the cytosolic Ca2+-buffering capacity is decreased. This greater IK-Ca activation also leads to a 28% increase in the action potential undershoot. Therefore, the model suggests that the faster Ca2+ dynamics can fully explain the increased excitability observed in Cr-/- mice through increased activation of Ca2+-activated K+ current. It is important to note that a change in the dynamics leading to the activation of IK-Ca is needed to obtain hyperexcitability in the model. A simple increase in the maximal conductance, attributable to Ca2+-activated K+ channels (ḡK-Ca), leads to the opposite effect, lowering the spike frequency (data not shown). Finally, because our model is a qualitative one, we do not claim that the fourfold variation in f represents the actual variation of the calcium-buffering capacity resulting from the absence of calretinin. A quantitative analysis would require a detailed knowledge of the mechanisms regulating the Ca2+ homeostasis of the cerebellar granule cell. Such a complete set of quantitative data is not available at the present time. Similar results (date not shown) were obtained with models by Maex and Schutter (1998) and D'Angelo et al. (2001), which were modified to include Ca2+ dynamics as described by Equation 10.
Unaltered synaptic transmission
We have also examined whether the absence of calretinin also affects synaptic transmission at the mossy fiber-granule cell synapse. Granule cells are excited by glutamatergic mossy fiber synapses and inhibited by GABAergic Golgi cell synapses. Without GABAA receptor inhibitors, mossy fiber stimulation in sagittal cerebellar slices causes strong granule cell inhibition through direct activation of Golgi cells by the mossy fibers (Armano et al., 2000). Therefore, to measure the EPSPs, we stimulated mossy fibers in the presence of the GABAA receptor-blocker picrotoxin (25 μm). In these conditions, a similar increase in the intrinsic excitability was observed, because the slope of the current-frequency plots was significantly higher in mutant mice (3.4 ± 0.6 Hz·pA-1 for WT, n = 6 and 7.3 ± 1.0 Hz·pA-1 for Cr-/-, n = 5; p < 0.05). However, no differences in the EPSP amplitude or kinetics could be observed (Fig. 6, Table 3).
This study shows that the absence of calretinin increases cerebellar granule cell excitability without altering granule cell morphology or passive electrical membrane properties. Calretinin is a Ca2+-binding protein member of the “EF-hand” family (Persechini et al., 1989). On the basis of available functional data, two types of EF-hand Ca2+-binding proteins can be distinguished. Proteins like calmodulin change their conformation after Ca2+ binding, thereby modulating the activity of various enzymes (Cheung, 1980) and ionic channels (Peterson et al., 1999; Qin et al., 1999; Zuhlke et al., 1999). In contrast, proteins like parvalbumin, calbindin-D28K, and calretinin are believed to play the role of passive-buffering systems, limiting the rise in intracellular free Ca2+ concentration.
We used a mathematical model to examine whether the observed alterations of Cr-/- cerebellar granule cell electroresponsiveness could be linked to the decreased Ca2+-buffering capacity determined by the absence of calretinin. When challenged with a decreased cytosolic Ca2+-buffering capacity, the model correctly predicts all of the changes in Cr-/- granule cell electroresponsiveness that are observed experimentally. Thus, calretinin seems to affect excitability primarily by acting as a passive Ca2+ buffer. A direct confirmation of this conclusion is provided by the fact that 0.15 mm BAPTA is able to restore normal excitability levels in Cr-/- granule cells. A consequence of this observation is that, because calretinin and BAPTA have similar Ca2+-binding kinetics and because calretinin is the only calcium-binding protein detected so far in cerebellar granule cells, 0.15 mm BAPTA may effectively mimic the contribution of calretinin to the cytosolic Ca2+-buffering capacity of the granule cell. Finally, the similar intrinsic excitability in the wild-type and Cr-/- mice observed in the presence of 0.15 mm BAPTA supports the notion that the membrane properties involved in the action potential generation are similar in the two strains. Thus, the possibility that the increased excitability observed in Cr-/- mice in the perforated patch configuration is a consequence of long-term changes in the electrical membrane properties resulting from the mutation can be ruled out.
Our model should not be seen as a complex model of the granule cell, similar to the one proposed by D'Angelo et al. (2001), but rather as an abstract model focusing on a specific cell property, cytosolic Ca2+ buffering. The purpose of this model is to demonstrate the basic mechanism linking alteration of the intrinsic excitability in Cr-/- mice to alterations in the cytosolic Ca2+-buffering capacity. Thus, on a broader perspective, the conclusions drawn from our simulations can be applied to other neuronal types, providing that the mechanisms of excitability are essentially the same as in cerebellar granule cells and that the conductance of the Ca2+-activated K+ channels is sufficient to obtain a strong coupling between excitability and Ca2+ dynamics during the spike generation. Although it is general knowledge that increasing Ca2+-activated K+ current slows down the firing rate, faster Ca2+ dynamics through reduced Ca2+ buffering has the opposite effect. The model suggests that this is because of direct control of IK-Ca by the Ca2+ transient, speeding up spike repolarization when the calcium-buffering capacity is decreased. Interestingly, experimental data obtained from hippocampal neurons perfused with BAPTA support our results (Lancaster and Nicoll, 1987; Storm, 1987). In this view, in addition to their obvious role in Ca2+ homeostasis, Ca2+-binding proteins could play an active role in modulating neuronal intrinsic excitability and therefore neuronal plasticity. Although information storage is usually believed to be mediated by long-term modifications in the strength of synaptic transmission, activity-dependent changes in the neuronal intrinsic excitability also take place, causing forms of nonsynaptic plasticity. Such activity-dependent changes in intrinsic excitability have been shown to occur in cerebellar granule cells (Armano et al., 2000) and neurons of the deep cerebellar nuclei (Aizenman and Linden, 2000). It would be interesting to know whether activity-dependent modifications in the localization or in the level of expression of Ca2+-binding proteins are involved. Although expression changes are still controversial (for review, see Baimbridge et al., 1992), changes in the localization of calretinin have been shown to occur in neurons (Hack et al., 2000).
Our results shed light on previous data obtained by Schiffmann et al. (1999), which show an increased simple-spike firing in Purkinje cells of alert Cr-/- mice. Because calretinin is not expressed in Purkinje cells, their firing alterations have to be an indirect consequence of the mutation. Indeed, because granule cells excite Purkinje cells through the parallel fibers, the increased Purkinje cell-firing rate observed in vivo can be explained by the increased granule cell excitability. Indeed, neither any increase in synaptic excitation was observed at the mossy fiber granule cell synapse nor excitatory transmission or short-term plasticity were altered at the parallel fiber Purkinje cell synapse in Cr-/- cerebellar slices (Schiffmann et al., 1999). Thus, altered intrinsic excitability in granule cells is probably the most important factor causing the in vivo alterations in Cr-/- mice.
The possibility that calretinin deficiency might increase excitability in other cerebellar neurons cannot be ruled out at present. Indeed, in rodents, calretinin is also present in two other neuronal subtypes of the granule cell layer: unipolar brush cells and Lugaro cells (Résibois and Rogers, 1992). Lugaro cells are believed to exert a disinhibitory influence on Purkinje cells via a glycinergic inhibition of Golgi cells (Dieudonné and Dumoulin, 2000). Therefore, an increased excitability of the Lugaro cells induced by the calretinin deficiency could lead to an increased activity at the level of the Purkinje cell. The involvement of unipolar brush cells may also cause a direct excitatory enhancement of Purkinje cells. However, it should be noted that unipolar brush cells are restricted to a subset of cerebellar lobes (Diño et al., 2000), whereas the increased Purkinje cell firing is ubiquitous in the cerebellar cortex of Cr-/- mice (G. Cheron, D. Gall, and S. N. Schiffman, personal communication).
In conclusion, this study demonstrates that calretinin deficiency increases the intrinsic excitability of cerebellar granule cells. The increased granule cell electroresponsiveness may explain the electrophysiological and behavioral alterations observed in vivo on alert Cr-/- mice, indicating the critical role of granule cells in information processing in the cerebellar cortex. On a broader perspective, we suggest that modulation of neuronal excitability by Ca2+-binding proteins could play a functional role in the control of information coding and storage in the CNS.
This work was supported by the Belgian Fonds pour la Recherche dans l'Industrie et l'Agriculture (C.R., B.B.), Queen Elisabeth Medical Foundation (Neurobiology 99-01 and 02-04), Fund for Medical Scientific Research (Belgium 3.4551.98/3.4507.02), Fondation David et Alice Van Buren, Action de Recherche Concertée (2002-2005), and the European Commission's Fifth Framework Program for Research (Cerebellum BIO4CT98-0182 and Spikeforce IST35271). We thank Philippe Lebrun and Geneviève Dupont for helpful comments. We also thank Albert Goldbeter for providing computing facilities. D.G. is a postdoctoral researcher at the Belgian Fonds National de la Recherche Scientifique (FNRS). M.C.G. is a researcher at the French Centre National de la Recherche Scientifique and a postdoctoral fellow at the Belgian FNRS, and D.B. is a postdoctoral fellow at the Belgian FNRS.
Correspondence should be addressed to Dr. Serge Schiffmann, Laboratoire de Neurophysiologie (CP601), Université Libre de Bruxelles, Route de Lennik 808, B-1070 Bruxelles, Belgium. E-mail:.
Copyright © 2003 Society for Neuroscience 0270-6474/03/239320-08$15.00/0