The Ca_v3–K_v4 Complex Acts as a Calcium Sensor to Maintain Inhibitory Charge Transfer during Extracellular Calcium Fluctuations

The Ca_v3–K_v4 complex is modulated by a range of [Ca]_o

Measurements of [Ca]_o in the cerebellar molecular layer in vivo have revealed a decrease in [Ca]_o from 1.2 mm to as low as 0.8 mm during 5 Hz climbing fiber stimulation (Stöckle and ten Bruggencate, 1978). We showed previously that blocking Ca_v3 calcium current causes a hyperpolarizing shift in the half-inactivation voltage (V_h) of stellate cell I_A, effectively reducing the availability of I_A (Anderson et al., 2010b). However, the ability for the Ca_v3–K_v4 complex to respond to the shifts in [Ca]_o that occur during physiological rates of activity is unknown.

We measured the effects of bath perfusing various concentrations of [Ca]_o (0.1–2.2 mm) on the inactivation voltage of I_A in stellate cells. I_A was isolated pharmacologically by blocking sodium, hyperpolarization-activated cyclic nucleotide-gated cation (HCN), and noninactivating potassium currents through bath application of 100 nm TTX, 2 mm Cs⁺, and 5 mm TEA, respectively. Calcium channel blockers were specifically excluded to allow low voltage-activated Ca_v3-mediated calcium influx (I_T) during voltage commands. Under these conditions, the primary current recorded is a LVA, fast inactivating I_A (Fig. 1A). The V_h of I_A was determined using 1 s conditioning steps from −110 to 0 mV in 10 mV steps followed by a test pulse to −30 mV. A test pulse of −30 mV was used, as it corresponds to the peak of I_T activation and is within the range of I_A activation, but below that of high voltage-activated (HVA) calcium currents. In this regard, previous work has shown that stellate cells exhibit little HVA calcium current (Anderson et al., 2010b), and that none of the HVA calcium channels modulate the K_v4–KChIP3 interaction even when overexpressed in a heterologous system (Anderson et al., 2010b).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

A physiologically relevant decrease of [Ca]_o shifts the V_h of I_A. A, Inactivation plots of stellate cell I_A recorded in 1.5 or 0.1 mm [Ca]_o from a holding potential of −110 mV, followed by a series of preconditioning voltages of 10 mV steps to 0 mV, and current amplitude tested with a test pulse to −30 mV (see inset). B, The dose–response curve of the relationship between [Ca]_o and mean I_A V_h in stellate cells. Voltage commands are as in A. C, Dose–response curve between [Ca]_o and K_v4.2 V_h expressed in tsA-201 cells for a Ca_v3–K_v4 complex in the presence or absence of expressed KChIP3 protein. Voltage commands are as in A. D, The effects of varying [Ca]_o on the density of I_A and I_T in two different stellate cells. I_A was first evoked by a step from −80 to −45 mV in 1.5 or 1.1 mm [Ca]_o, and I_T was then isolated in the same cells by perfusion of 20 mm 4-AP. On the right are plots of the mean densities of I_A and I_T as a function of [Ca]_o. Current densities are averages of 10 traces and normalized to membrane capacitance.

Under control conditions of 1.5 mm [Ca]_o, stellate cells exhibit an I_A with a V_h of −69.5 ± 1.9 mV (n = 7). Decreasing bath [Ca]_o to 0.1 mm resulted in an approximately −10 mV hyperpolarizing shift in V_h for I_A to −80.8 ± 2.0 mV (n = 7) (Fig. 1A). In contrast, increasing [Ca]_o to 2.2 mm did not shift the V_h of I_A (Fig. 1B; n = 5, p > 0.05, unpaired t-test), indicating that the V_h recorded under control conditions could not be further shifted in the depolarizing direction by increasing calcium conductance. As [Ca]_o was decreased to levels between 1.5 and 0.1 mm [Ca]_o, the V_h exhibited a graded leftward hyperpolarizing shift, reaching a stable saturation point of −80.1 ± 1.2 mV at ∼0.8 mm [Ca]_o (n = 5, p < 0.001; Fig. 1B). The relationship between V_h and [Ca]_o was readily fit with the Hill equation (k = 1.1 ± 0.1 mm; Hill coefficient, N = 8.3 ± 2.4) with a midpoint of −75.3 ± 1.3 mV at 1.1 mm [Ca]_o (n = 6, p < 0.05; Fig. 1B). The effects of low [Ca]_o on I_A availability were not due to a change in divalent charge screening, as changes in [Ca]_o were always offset by a corresponding change in [Mg]_o to yield a total divalent cation concentration of 3 mm. We further confirmed that the mean V_h of I_A in stellate cells was not significantly different in 1.1 mm [Ca]_o if [Mg]_o was not elevated (n = 5, p > 0.05, unpaired t-test; data not shown), and that elevating [Mg]_o by 0.4 mm did not affect Ca_v3.3 calcium current measured in tsA-201 cells (n = 7; data not shown). The effects of decreasing [Ca]_o on K_v4 V_h are thus interpreted as a decrease of calcium influx through Ca_v3 T-type calcium channels.

To further examine the relationship between [Ca]_o and the Ca_v3–K_v4 interaction, we recorded K_v4.2 current in tsA-201 cells coexpressing cDNA for Ca_v3.3, K_v4.2, KChIP3, and DPP10c as a representative complement of the Ca_v3–K_v4 complex in stellate cells (Anderson et al., 2010b). tsA-201 cells were bathed in an external medium containing 1.5 mm [Ca]_o at rest. Under these conditions, the K_v4.2 V_h was −57.4 ± 0.6 mV (n = 30), a baseline value lower than that encountered in stellate cells, but not unexpected for differences in absolute voltage dependence when comparing cDNA expressed in a heterologous system. Importantly, varying [Ca]_o between 0.1 and 2.2 mm again shifted the voltage dependence of K_v4 inactivation over a range of ∼10 mV, similar to that found in stellate cells (Fig. 1C). K_v4.2 V_h reached a maximum hyperpolarized value of −66.5 ± 1.3 mV (n = 6) for a [Ca]_o of 0.8 mm, with a V_h of −62.9 ± 0.9 mV at 1.1 mm [Ca]_o (n = 5; k = 1.2 ± 0.1 mm; N = 11.2 ± 5.4). Importantly, the Ca_v3–K_v4 interaction has been shown to depend on KChIP3, an accessory calcium-sensing protein of the K_v4 complex (Anderson et al., 2010b). When cDNA encoding for KChIP3 proteins was omitted from the transfection medium, the dose–response relationship between [Ca]_o and K_v4.2 V_h was abolished, with K_v4.2 V_h remaining stable at approximately −70 mV for all values of [Ca]_o (Fig. 1C).

These data indicate that the effects of a decrease in [Ca]_o on I_A availability derives at least in part from an interaction between I_T and KChIP3. Our understanding of the interaction between Ca_v3 and K_v4 channels suggests that the dose-dependent response of K_v4 V_h to [Ca]_o reflects a shift in V_h due to a calcium-dependent KChIP3 interaction. The prediction is that K_v4 current density should plateau once all KChIP3 proteins are maximally bound as calcium influx increases with [Ca]_o. To test this we measured the peak density of I_A and I_T under different [Ca]_o (Fig. 1D). I_A density was assessed with I_T left intact, and then I_T was isolated in the same cell by bath application of 4-AP (20 mm) and evoked using the same step command (Fig. 1D). Control tests established that 20 mm 4-AP has no effect on the current–voltage relationship of either Ca_v3.2 or Ca_v3.3 channels, the two Ca_v3 channel isoforms expressed by stellate cells (n = 6; data not shown) (Molineux et al., 2006). While I_A could be recorded over a range of 0.8–2.2 mm [Ca]_o, I_T could not be reliably resolved below 1.1 mm [Ca]_o, restricting I_T density measurements to 1.1–2.2 mm [Ca]_o. These tests showed that I_T peak density exhibited a linear increase over the full range of 1.1 to 2.2 mm [Ca]_o, as predicted by an increased driving force and the number of permeant ions as [Ca]_o was raised. In contrast, I_A peak density followed the shift in V_h established for different levels of [Ca]_o in Figure 1B, reaching a plateau of maximal density for [Ca]_o levels of 1.5 mm or greater (Fig. 1D).

These data establish that K_v4 availability can be modulated within the range of shifts in [Ca]_o that have been reported to occur under physiological conditions in vivo. The results also indicate that reductions in [Ca]_o will decrease Ca_v3-mediated calcium influx and reduce the availability of K_v4 current in a manner consistent with an action through the Ca_v3–K_v4 complex.

Climbing fiber input decreases stellate cell I_T and I_A

To be physiologically relevant, the low calcium-induced change in K_v4 availability should be reproduced by direct activation of synaptic inputs in the molecular layer. In this regard, climbing fibers present an especially interesting case for potential regulation of the Ca_v3–K_v4 complex. Climbing fiber boutons have been shown to overlap with the soma and dendritic processes of stellate cells (Nishiyama et al., 2007; Brown et al., 2012), while ultrastructural analyses reveal boutons that are tightly apposed to stellate cell membranes, but without evidence for vesicular release machinery (Hámori and Szentágothai, 1980; Kollo et al., 2006). In fact, the locations at which climbing fiber boutons align with stellate cells closely matches the distribution pattern of K_v4 channel immunolabel in stellate cell membranes, a highly unusual relationship for which there was no known function (Kollo et al., 2006). Although the distribution of Ca_v3 channels has not been determined at the ultrastructural level, the known association between Ca_v3 and K_v4 channels within a functional nanodomain (<50 nm) (Anderson et al., 2010b) suggests that Ca_v3 channels will be found in close proximity to some number of the K_v4 channels at these junctions. Climbing fiber activation is further interesting in reportedly generating a reduction in [Ca]_o (Stöckle and ten Bruggencate, 1980) in the course of generating a calcium-dependent complex spike in Purkinje cells that conducts throughout the dendritic tree (Llinás and Sugimori, 1980b; Kitamura and Häusser, 2011), where stellate cells are embedded.

Sensing changes in calcium driving force

We wished to test whether repetitive climbing fiber stimulation could evoke a decrease in [Ca]_o that in turn alters the availability of I_A through the Ca_v3–K_v4 interaction. To date, the only measure of a change in [Ca]_o during climbing fiber stimulation was obtained using calcium-selective microelectrodes in vivo (Heinemann et al., 1977; Nicholson et al., 1978; Stöckle and ten Bruggencate, 1980; McCreery and Agnew, 1983). The ability to evoke a decrease in [Ca]_o in vitro during repetitive synaptic input must also contend with the technical necessity of perfusing a fixed level of calcium in the bathing medium. We note, however, that the recordings of I_T in stellate cells in Figure 1D indicate that a change in [Ca]_o in the bathing medium can be detected as a change in driving force, as reflected by a linear reduction in I_T density as [Ca]_o is lowered, with an ∼30% decrease in I_T when [Ca]_o was lowered from 1.5 to 1.1 mm. We thus tested whether repetitive climbing fiber stimulation could decrease I_T in a manner consistent with a change in the driving force for calcium. In this way we could effectively use the stellate cell itself as a delicate sensor for any changes in [Ca]_o in the surrounding microenvironment during in vitro recording conditions.

To examine the effect of climbing fiber stimulation on I_T in stellate cells, we simultaneously recorded from the soma of a Purkinje cell and that of a stellate cell visualized directly above in the molecular layer, and thus presumably within or near the dendritic arbor of the recorded Purkinje cell. In some cases this could be confirmed by tracking the course of a Purkinje cell dendritic tree through DIC-IR optics to the position of the targeted stellate cell. We then stimulated climbing fiber input to simultaneously monitor complex spike discharge in the Purkinje cell and I_T in the stellate cell. I_T was isolated by including in the electrode 5 mm QX-314 to block sodium channels and 5 mm 4-AP and 5 mm TEA to block I_A and any delayed rectifying potassium channels. Although internal QX-314 can affect some calcium currents (Talbot and Sayer, 1996), we confirmed that we could record I_T with a stable baseline amplitude of 8.4 ± 1.9 pA (n = 6; maximum 20 pA) that reliably recovered to control levels following trains of synaptic stimulation. Blockade of any remaining potassium channels was ensured by using a CsCl-based internal electrolyte. Synaptic activation of stellate cells (Jörntell and Ekerot, 2003; Szapiro and Barbour, 2007; Barmack and Yakhnitsa, 2008) was blocked by internal perfusion of 100 μm philanthotoxin to block AMPA receptors (Nilsen and England, 2007) and 1 mm MK-801 to block NMDA receptors, and bath application of 1 μm MPEP, 1.5 μm JNJ-16295685, and 2.5 μm AM-251 to block potential activation of group I metabotropic glutamate or CB1 receptors (Gatley et al., 1996; Gasparini et al., 1999; Crepel and Daniel, 2007; Knöpfel, 2007). Under these conditions, the majority of synaptic currents were blocked with only occasional fast spontaneous synaptic currents in four of six cells (mean of −57.8 ± 20.0 pA amplitude; n = 4). Spontaneous synaptic events were not associated with climbing fiber stimulation and were readily distinguishable from I_T, but any records with overlapping spontaneous synaptic currents were discarded. In this way, I_T could be recorded in stellate cells while preserving climbing fiber-evoked complex spike discharge in nearby Purkinje cells.

Voltage pulses were delivered to stellate cells from a holding potential of −110 to −45 mV (50 ms) at a 5 Hz base rate (Fig. 2A) to measure I_T near its peak voltage on I–V plots. After 2.5 s of control recording, a climbing fiber stimulus was interleaved with step commands by stimulating at 10 Hz (10 pulses), and recordings of I_T continued for at least 5 s after stimulation was terminated. Data from recordings were accepted when I_T density was sufficient to evoke a clearly resolved transient inward current, a complex spike was reliably evoked in the Purkinje cell, and recovery of I_T peak amplitude was apparent after synaptic stimulation. I_T peak amplitude was stable during control pulses, but declined rapidly during climbing fiber stimuli (Fig. 2A), reducing from an original peak value of 8.4 ± 1.9 pA to 5.5 ± 1.8 pA (n = 6) by the end of the 10 pulse stimulus train, reflecting a 42.2 ± 11.8% decrease in amplitude (n = 6, p < 0.05, one-way repeated measures ANOVA, Tukey's pairwise comparison). Within 1 s following a stimulus train, I_T peak amplitude had recovered to 8.2 ± 1.5 pA (n = 6, p = 0.98), a level that was not significantly different from control. Full recovery was apparent by 5 s (8.45 ± 1.92 pA, n = 6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Repetitive 10 Hz climbing fiber stimulation evokes a decrease in stellate cell I_T. A, Dual recordings of I_T in a stellate cell (SC) and spike discharge in a Purkinje cell (PC) soma recorded in the cell layer directly below the SC. Synaptic responses are pharmacologically blocked in the SC to allow activation of complex spikes in the Purkinje cell by climbing fiber (CF; arrows) stimulation. I_T is evoked by a step from −110 to −45 mV (5 Hz), and CF stimuli are interleaved with the step commands using a 10 pulse 10 Hz CF stimulus. Recordings are shown for control (left), every second CF stimulus during the 10 pulse train (middle), and 1.0 s after the end of the train (Recovery; right). The dashed line below recordings marks the control I_T amplitude for comparison. B, Mean values for I_T amplitude normalized and color coded with respect to the mean of control pulses preceding CF stimulation (n = 6). The period of 10 Hz CF stimulation is shown bounded by a box. C, Expanded and superimposed representative records of I_T in control, immediately post CF stimulation, and recovery 1.0 s after the stimulus train, as for experiments in A and B. D, Bar plots of the mean change in I_T amplitude immediately following CF stimulation (Post CF) and 1.0 s following the end of the stimulus train. For display purposes, stimulus and capacitative artifacts are digitally reduced in A and C and a slight shift in baseline inward current during CF stimulation is removed from I_T recordings in A. Sample values are shown in brackets within the bar plot. *p < 0.05.

A reduction in I_T density could come about through a change in Ca_v3 channel insertion in the membrane, a reduction in conductance through the action of a ligand or a shift in voltage-dependent properties, or a reduction in calcium driving force. It seems highly unlikely that the density of Ca_v3 channels in the plasma membrane could be reduced and restored to control levels within ∼1 s time in relation to climbing fiber stimulation. We note that we cannot fully exclude the possibility that an unknown ligand (such as that producing occasional spontaneous synaptic events observed in stellate cells under our conditions) could have affected Ca_v3 channel conductance, although we did not detect any synaptic responses evoked by climbing fiber stimuli per se. We also have no evidence to support a shift in the voltage dependence of Ca_v3 channels, a parameter that would be difficult to test in the short time frame associated with changes in I_T density recorded here. Rather, the results presented in Figure 1D indicate that even subtle decreases in [Ca]_o are sufficient to substantially alter I_T density, and in a linear fashion as predicted for a shift in calcium driving force. In fact, comparing the 40% mean reduction in I_T density recorded during climbing fiber stimulation (Fig. 2D) to that recorded during bath perfusion of different levels of [Ca]_o predicts a potential climbing fiber-induced change in [Ca]_o of 0.63 mm (from 1.5 mm down to 0.87 mm). These data are important in revealing that climbing fiber stimulation reduces I_T in a manner consistent with a decrease in [Ca]_o in the microenvironment surrounding stellate cells even under in vitro recording conditions.

Stimulus-evoked reductions in I_A

We next repeated these tests to examine the effect of repetitive climbing fiber stimulation on the amplitude of I_A in stellate cells. We thus obtained dual recordings from the soma of a Purkinje cell and a stellate cell positioned directly above in the molecular layer. I_A was isolated in stellate cells by including 5 mm QX-314 and 5 mm TEA in the electrode using a KCl-based internal electrolyte, and complex spike discharge was recorded in the Purkinje cell in current-clamp mode (Fig. 3A). The same set of internal and external synaptic blockers were applied as for recordings of I_T. In a separate set of experiments, we verified that QX-314 had no significant effect on the magnitude of the shift in K_v4 V_h induced by perfusing 1.1 mm [Ca]_o (n = 6, p > 0.05, unpaired t-test; data not shown). I_A was evoked by a voltage step from −85 to −30 mV (50 ms) at a 5 Hz base rate to test I_A over the voltage range relevant to a shift in inactivation voltage. Control recordings of I_A were obtained over 1.5 s before interleaving a 10 Hz climbing fiber stimulus at 10 Hz (10 pulses). I_A peak amplitude was stable during the control pulses, but showed a progressive decrease during climbing fiber evoked complex spike discharge in a nearby Purkinje cell, reaching an average 15.8 ± 3.2% reduction after 10 stimuli (maximum 30%; n = 10, p < 0.001; Fig. 3A,B). At the end of the climbing fiber stimulus train, I_A amplitude gradually recovered, returning to 92.3 ± 1.6% of its original value by 2.4 s (n = 10, p < 0.01), and was fully recovered at 5 s to 1.0 ± 0.02 of control (n = 8, p > 0.05; Fig. 3A–C). Moreover, the reduction in I_A could be evoked in a reversible manner on repeated occasions at 5 min intervals, ruling out any significant washout over time.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Repetitive 10 Hz climbing fiber stimulation evokes a decrease in stellate cell I_A. A, Dual recordings of I_A in a stellate cell (SC) and spike discharge in a Purkinje cell (PC) soma recorded in the cell layer directly below. Synaptic responses are pharmacologically blocked in the SC to allow selective activation of complex spikes in the Purkinje cell through climbing fiber (CF; arrows) stimulation. I_A is evoked by a step from −85 to −30 mV (5 Hz), and CF stimuli are interleaved with the step commands using a 10 pulse, 10 Hz CF stimulus. Recordings are shown for control (left), every second CF stimulus during the 10 pulse train (middle), and after 2.4 s recovery (right). The dashed line above recordings marks the control I_A amplitude for comparison. B, Mean values for I_A amplitude normalized and color coded with respect to the mean of six control pulses preceding CF stimulation when recorded in 1.5 mm [Ca]_o (n = 10) or 2.2 mm [Ca]_o (n = 4). The period of 10 Hz CF stimulation is shown bounded by a box. C, Expanded and superimposed representative records of I_A in 1.5 or 2.2 mm [Ca]_o in control, during CF stimulation, and recovery 2.4 s after the stimulus train, as for experiments in A and B. D, Bar plots of the mean change in I_A amplitude immediately following the CF stimulus train in 1.5 or 2.2 mm [Ca]_o. Note the rapid decrease in peak I_A during CF stimulation in 1.5 mm [Ca]_o and the lack of effect in 2.2 mm [Ca]_o to offset stimulus-induced changes in [Ca]_o. For display purposes, recordings are truncated in A. Stimulus and capacitative artifacts were digitally reduced in A and C. Sample values are shown in brackets within bar plots. ***p < 0.001.

Based on the V_h [Ca]_o dose–response curve (Fig. 1B), we predicted that conducting these tests in 2.2 mm [Ca]_o should prevent significant V_h shifts by offsetting any reduction in [Ca]_o during climbing fiber stimulation. Therefore, in a subset of cells, we subsequently bath applied 2.2 mm [Ca]_o (n = 4 of 10) for 5 min and retested the effects of climbing fiber stimulation on I_A amplitude. No significant decrease in I_A was detected in the presence of 2.2 mm [Ca]_o (0.8 ± 2.2% decrease, n = 4, p = 0.95; Fig. 3B–D), even though the preceding trains in this group of cells decreased I_A by 10.6 ± 2.9% when delivered in 1.5 mm [Ca]_o (n = 4, p < 0.05).

Collectively, these results are consistent with the interpretation that repetitive stimulation of complex spike discharge in Purkinje cells promotes a decrease in the driving force of Ca_v3-mediated calcium influx, which is accompanied by a reduction in the availability of I_A in stellate cells, as predicted for actions of the Ca_v3–K_v4 complex.

Physiological changes in [Ca]_o alter stellate cell excitability

Outward current generated by I_A is capable of controlling several aspects of spike output, including spike latency, threshold, and frequency (Connor and Stevens, 1971a, b; Hoffman et al., 1997; Molineux et al., 2005; Khaliq and Bean, 2008; Anderson et al., 2010b; Norris and Nerbonne, 2010; Norris et al., 2010; Carrasquillo et al., 2012). We next examined the effect of decreasing [Ca]_o from a resting level of 1.5 to 1.1 mm on firing frequency using on-cell recordings to avoid cell dialysis inherent to whole-cell recordings. Stellate cells exhibited a resting level of spontaneous discharge between 5 and 10 Hz (6.6 ± 0.5 Hz, n = 35) in 1.5 mm [Ca]_o. Decreasing [Ca]_o to 1.1 mm resulted in a nearly twofold increase in the rate of spontaneous discharge (n = 6) that was reversible when [Ca]_o was returned to 1.5 mm (Fig. 4A). It is possible that the observed elevation in firing frequency in 1.1 mm [Ca]_o could involve a decreased activation of calcium-dependent potassium channels. However, neither the SK channel blocker apamin (100 nm, n = 6, p > 0.05) nor the BK channel blocker paxilline (100 nm, n = 5, p > 0.05) affected the rate of spontaneous on-cell firing frequency.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Low [Ca]_o increases the gain and lowers the threshold of stellate cell firing. A, B, Lowering [Ca]_o from 1.5 to 1.1 mm significantly increases the frequency of spontaneous stellate cell spike discharge in on-cell recordings and gain of intracellular spike discharge evoked by square wave current pulses to calculate current–frequency plots in B. C, Bar plots of the effect of 1.1 mm [Ca]_o on gain of firing and its occlusion by pretreatment with mibefradil (Mibef; 500 nm) or infusion of an antibody against KChIP3 (1:100) through the patch electrode, but not by an antibody against KChIP1. D, Recordings and scatter plot reveal a significant decrease in the absolute voltage threshold for spike discharge (dashed lines) in 1.1 mm [Ca]_o. Sample values are shown in brackets. *p < 0.05; **p < 0.01; ***p < 0.001.

A second issue to consider is that repetitive activation of synaptic inputs also elevates the extracellular concentration of potassium ([K]_o). In the molecular layer, 5 Hz climbing fiber discharge evoked by peripheral nerve stimulation in vivo produces an increase in [K]_o of ∼0.7 mm from resting conditions that could affect E_K and the driving force for I_A (Bruggencate et al., 1976; Stöckle and ten Bruggencate, 1980). However, we found that a 0.75 mm elevation of [K]_o did not significantly alter either the rate of on-cell spike discharge (n = 6, p > 0.05) or the amplitude of I_A (n = 10, p > 0.05). A combined change in [Ca]_o and [K]_o reduced I_A amplitude (n = 6, p < 0.05), but it was not significantly different from that of 1.1 mm [Ca]_o alone (n = 6, p > 0.05). These tests reveal that the increases in [K]_o reported during physiological levels of climbing fiber stimulation in vivo are not sufficient to account for the effects detected here for associated decreases in [Ca]_o. Therefore, we restricted further tests to changes in [Ca]_o relevant to actual peripheral stimulation in vivo.

We showed previously that pharmacological blockade of Ca_v3 calcium channels and its interaction with the K_v4 channel complex increased the gain of firing in stellate cells (Anderson et al., 2010b). We next determined the effect of different levels of [Ca]_o on stellate cell gain used whole-cell recordings. Applying a decrease in [Ca]_o from 1.5 to 1.1 mm to reduce I_A availability revealed an ∼65% increase in the gain of stellate cell firing on current–frequency plots (n = 5, p < 0.001; Fig. 4B,C). The increase in gain was not related to any secondary regulation of internal calcium stores, as the gain of stellate cell firing increased in a similar manner in the presence of internal heparin (4 mg/ml) and bath-applied 6 μm cyclopiazonic acid to block any IP3-mediated calcium release and calcium ATPases (n = 8, p < 0.05). Pretreating slices with the Ca_v3 channel blocker mibefradil (500 nm) further occluded the effects of 1.1 mm [Ca]_o on the gain of firing frequency (n = 8), supporting the interpretation that the increase in gain is mediated by Ca_v3 calcium influx (Fig. 4C).

One means of interrupting the Ca_v3-KChIP3 interaction is by internally infusing a monoclonal KChIP3 antibody (Anderson et al., 2010b). We thus infused KChIP antibodies (1:100 dilution) through the electrode to test the ability for low [Ca]_o to induce a change in the gain of firing. Frequency–current plots were constructed at 10 min intervals for up to 30 min following entry in the whole-cell configuration to allow full dialysis of KChIP antibodies. These tests revealed that infusion of a KChIP3 antibody occluded the ability of 1.1 mm [Ca]_o to increase the gain of firing (n = 7; Fig. 4C). By comparison, infusion of a monoclonal antibody against KChIP1 (1:100 dilution), a protein not expressed in stellate cells, did not prevent an increase in gain of firing by 1.1 mm [Ca]_o (n = 7, p < 0.001; Fig. 4C).

A closer examination of the process of spike firing revealed that the increase in gain of firing in 1.1 mm [Ca]_o was associated with a significant decrease in spike threshold voltage, with the point of membrane potential transition to spike output shifting in a negative direction by 4–10 mV (n = 5, p < 0.05; Fig. 4D). The increase in spike frequency in low [Ca]_o could also be accompanied by a decrease in the amplitude of the AHP, although the magnitude of these changes were variable and not statistically significant (data not shown). The loss of I_A and outward current during the interspike interval in low [Ca]_o can thus reduce spike threshold, identifying one factor contributing to the increase in firing frequency.

The Ca_v3–K_v4 complex adjusts firing rate in response to fluctuations in [Ca]_o

To determine the extent to which a selective shift in the V_h of stellate cell I_A can alter the gain or threshold of stellate cell firing, we applied dynamic clamp. For this we used the established biophysical properties of I_A and I_T in stellate cells (Molineux et al., 2005), the dose–response relationship between [Ca]_o and I_A V_h values (Fig. 1B), and the inactivation curves in 1.5 versus 1.1 mm [Ca]_o to estimate the h_∞ values reflecting I_A availability. These data were then used to iteratively simulate either condition of [Ca]_o and dynamically modulate the I_A available at each voltage sampled during a spike response (25 μs sampling interval). In this way, a cell could be maintained in either 1.5 or 1.1 mm [Ca]_o and dynamic clamp used to simulate a shift in I_A availability typical of a change in [Ca]_o without altering the driving force for calcium entry. Practically, this equated to calculating the I_A predicted to be available for either [Ca]_o and then injecting or subtracting the difference current (Fig. 5). Simulating a shift from an actual level of 1.5 mm [Ca]_o in the bath to a “virtual” 1.1 mm [Ca]_o thus predicted a net inward current that acted in a depolarizing manner during the time course of an evoked spike (Fig. 5B). Conversely, if a cell was held in 1.1 mm [Ca]_o, we simulated the condition relevant to I_A availability for a virtual increase of [Ca]_o to 1.5 mm by injecting an outward difference current during the course of a spike (Fig. 5C). Thus, the availability of I_A that would be characteristic of a given [Ca]_o could be applied on-line without directly manipulating [Ca]_o or I_T.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Simulating a selective shift in I_A voltage for inactivation through dynamic clamp. A, The stellate cell I_A V_h [Ca]_o dose–response curve (top) and shift in I_A h_∞ for 1.5 or 1.1 mm [Ca]_o (bottom). The dashed lines (top) represent the calculated V_h for I_A in the presence of 1.5 mm [Ca]_o (blue) or 1.1 mm [Ca]_o (green; Fig. 1B). Fits of the mean inactivation curves of stellate cell I_A (bottom) for 1.5 and 1.1 mm [Ca]_o are used to iteratively calculate the h_∞ values and difference currents for I_A (I_diff). Dashed lines indicate the h_∞ values calculated for either condition at −70 mV. I_diff is then added or subtracted to simulate conditions of 1.5 or 1.1 mm [Ca]_o in real time during the spike (25 μs sampling interval). B, C, Recordings of single spike responses (top row) performed in the presence of 1.5 mm [Ca]_o or 1.1 mm [Ca]_o, but modulated through dynamic clamp according to the predicted availability of I_A at the alternate [Ca]_o. The second row represents the I_A predicted for the [Ca]_o indicated, the third row the I_A predicted to be available at the alternate [Ca]_o, and the fourth row the difference current calculated to dynamically add or subtract to simulate the alternate [Ca²⁺]_o condition. The I_diff was inward in B and had a depolarizing effect on membrane potential, and outward in C with a hyperpolarizing effect on membrane potential.

The effects of dynamically modulating I_A availability are shown in Figure 6. Recordings were first performed in 1.5 mm [Ca]_o, and the gain of firing frequency measured using current pulses under conventional current clamp to construct a frequency–current plot. Subtraction of I_A using dynamic clamp to simulate a left shift in V_h relevant to 1.1 mm [Ca]_o caused a substantial increase in gain of firing (n = 6, p < 0.05; Fig. 6A,C). We note that an additional increase in the initial rate of firing on the frequency–current plot was apparent for the dynamic clamp subtraction of I_A (Fig. 6A), likely reflecting the depolarizing action of I_T that remains intact under these conditions. Recordings from the same neurons were then performed in the presence of 1.1 mm [Ca]_o in the bath to reduce the Ca_v3–K_v4 interaction and increase gain in the expected manner, as evident by an increased rate of firing and slope of the baseline frequency–current plot (Fig. 6B). Dynamic injection of I_A that would be expected in the presence of 1.5 mm [Ca]_o then restored excitability by decreasing the gain of firing to within control levels (n = 6, p < 0.05; Fig. 6B,C). Consistent with results in bath-applied low [Ca]_o, a reduction of I_A via dynamic clamp for cells bathed in 1.5 mm [Ca]_o also resulted in a significant decrease in spike threshold (n = 6, p < 0.01; Fig. 6C). Conversely, dynamic addition of I_A in cells perfused initially with 1.1 mm [Ca]_o significantly increased spike threshold (n = 6, p < 0.05; Fig. 6C). Collectively, these results demonstrate that a selective shift in the V_h for I_A (without affecting calcium conductance directly) is sufficient to account for the changes in gain of stellate cell firing during fluctuations of [Ca]_o, again implicating a key role for the Ca_v3–K_v4 complex.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Dynamic subtraction or addition of I_A modulates the gain and threshold of stellate cell firing. A, B, A comparison of the effects of recording spike output in a stellate cell maintained in 1.5 mm [Ca]_o and then subtracting I_A to simulate a shift to 1.1 mm [Ca]_o (A), and the same cell now perfused with 1.1 mm [Ca]_o before (black) and after dynamic clamp injection of I_A (red) according to the predicted availability of I_A in 1.5 mm [Ca]_o (B). Representative traces of current-evoked spike output and the associated current–frequency plots are shown. C, Plots of the average effect of either dynamically subtracting I_A in cells perfused with 1.5 mm [Ca]_o (left column) or dynamically adding I_A to cells perfused in 1.1 mm [Ca]_o (right column) on the gain of firing frequency and spike voltage threshold. Lines on current–frequency plots represent best-fit calculations to the data, and sample values are shown at the base of bar plots. *p < 0.05; **p < 0.01.

The Ca_v3–K_v4 complex maintains inhibitory synaptic efficacy during reductions of [Ca]_o

Synaptic transmission depends critically on calcium influx into presynaptic terminals to cause vesicle fusion and transmitter release. Any change in [Ca]_o would then be expected to reduce release probability within the local environment. This is important, as a decrease in release probability could reduce the net inhibitory charge transferred to Purkinje cells when excitatory synaptic input lowers [Ca]_o. We thus tested the effects of low [Ca]_o on synaptic transmission between stellate and Purkinje cells. To understand the effects of low [Ca]_o on the probability of transmitter release, we recorded miniature IPSC (mIPSC) frequency in either 1.5 or 1.1 mm [Ca]_o in the presence of TTX (200 nm) to block all spike-associated transmitter release from presynaptic cells. Shifting from 1.5 to 1.1 mm [Ca]_o lowered mIPSC frequency from 15 ± 2.9 Hz to 9.7 ± 1.5 Hz (n = 10, p < 0.001), reflecting an ∼35% decrease in release probability at inhibitory synapses onto Purkinje cells (Fig. 7A,B), which is consistent with previous studies (King et al., 2001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Inhibitory charge transfer of sIPSCs to Purkinje cells is maintained in low [Ca]_o. A, mIPSCs recorded from a Purkinje cell in the presence of 200 nm TTX to block presynaptic spike discharge in 1.5 and 1.1 mm [Ca]_o. B, Bar plots of average mIPSC frequency reveal a significant decrease in mIPSC frequency in low [Ca]_o. C, sIPSCs recorded from a Purkinje cell in 1.5 mm [Ca]_o and 1.1 mm [Ca]_o. The internal electrolyte was a CsCl-based solution, and DNQX and dl-AP5 were bath applied to block excitatory synaptic transmission. Lowering [Ca]_o to 1.1 mm increases the amplitude of sIPSCs, while subsequent perfusion of the GABA_A blocker picrotoxin (PTX; 50 μm) abolishes all sIPSCs. D, A cumulative distribution function of sIPSC amplitudes indicating an increase in the proportion of large-amplitude sIPSCs with application of 1.1 mm [Ca]_o compared to 1.5 mm [Ca]_o. E, Mean bar plot of sIPSC frequency measured in 1.5 and 1.1 mm [Ca]_o and after perfusion of picrotoxin indicate that sIPSC frequency is maintained in low [Ca]_o and significantly decreases in the presence of picrotoxin. F, Bar plots of the total charge transfer in Purkinje cell recordings in the presence of 1.5 or 1.1 mm [Ca]_o and after perfusion of picrotoxin. Sample values are shown within bar plots. *p < 0.05; **p < 0.01; ***p < 0.001.

Based on these results, we predicted that sIPSCs would show a decrease in amplitude and frequency due to the decreased probability of release. We recorded sIPSCs in Purkinje cells in the presence of DNQX and dl-AP5 to block excitatory synaptic events and CGP55845 to block GABA_B receptors. The use of a high CsCl internal solution to block potassium currents reversed the chloride gradient, revealing sIPSCs as inward current deflections. sIPSCs recorded in the slice preparation at 1.5 mm [Ca]_o had variable amplitudes, attributed to synaptic events occurring at high frequency and originating from varying sites within the dendritic tree that lead to overlap and summation (Fig. 7C). Counter to our predictions, bath application of 1.1 mm [Ca]_o substantially increased the proportion of large-amplitude sIPSCs (n = 7, p < 0.001, Kolmogorov–Smirnov test; Fig. 7D). Furthermore, the measured value of sIPSC frequency was not significantly different between 1.5 mm [Ca]_o (26.6 ± 2.6 Hz, n = 7) and 1.1 mm [Ca]_o (25 ± 4.7 Hz, n = 7, p > 0.05; Fig. 7E). Subsequent application of picrotoxin (50 μm) blocked nearly all synaptic currents (Fig. 7D,E), indicating that the majority of sIPSC activity was mediated by GABA_A receptors.

In considering these results, we noted that large-amplitude events that may constitute simultaneous input from multiple sources would be difficult to distinguish from single events, potentially decreasing sIPSC frequency and increasing the measured proportion of large-amplitude events. We also reasoned that because probability of release decreased in 1.1 mm [Ca]_o (Fig. 7A,B), the proportional increase in large-amplitude sIPSCs did not reflect an increase in spike-associated transmitter release. Our previous experiments also determined that stellate cell firing frequency increases in the presence of low [Ca]_o (Fig. 4A). Therefore, we interpret these data to indicate that the increase in large-amplitude sIPSCs in 1.1 mm [Ca]_o can be explained by an increase in interneuron firing frequency, resulting in a larger number of overlapping and summated events that could not be resolved to accurately determine net sIPSC frequency.

To determine the combined effect of these changes on Purkinje cell inhibitory tone, we measured the average charge transfer per second associated with sIPSCs (see Materials and Methods) (Fig. 6F). Under control conditions of 1.5 mm [Ca]_o, Purkinje cells received an average of 7.7 ± 1.1 pC (n = 7) from sIPSCs, with application of picrotoxin significantly reducing charge transfer (1.7 ± 0.3 pC, n = 7; Fig. 7F), indicating the primary role of GABAergic inputs. In contrast, application of 1.1 mm [Ca]_o did not significantly change charge transfer (9.9 ± 2.6 pC, n = 7, p > 0.05; Fig. 7F). Therefore, these results suggest that charge transfer is maintained despite a decrease in release probability via a compensatory increase in stellate cell discharge.

To more closely examine the relationship between presynaptic spike discharge and postsynaptic inhibition in different levels of [Ca]_o, we performed dual recordings of synaptically connected stellate and Purkinje cells (Fig. 8). Presynaptic spikes in the recorded stellate cell that were correlated with postsynaptic IPSCs were identified by detecting an IPSC of at least 5 pA within 5 ms of a presynaptic spike peak (Fig. 8A). Spike-associated IPSCs were referred to as evoked IPSCs (eIPSCs) and used to create a spike-triggered average. If a presynaptic spike was not followed by an IPSC it was taken to reflect a failure of presynaptic transmission, as expected for the probabilistic nature of vesicle fusion and transmitter release. Finally, IPSCs with no clear association with spike discharge in the recorded presynaptic stellate cell were considered to be derived from other stellate or basket cells synapsing on the recorded Purkinje cell. Cell pairs that exhibited correlated presynaptic spikes and eIPSCs were encountered with a success rate of ∼10–15% (n = 5/42 cells), as judged in post hoc analysis of spike-triggered responses in Purkinje cells. To improve calculations of the net influence of presynaptic spike discharge on postsynaptic inhibition in low [Ca]_o, stellate cells were held below threshold to prevent spontaneous discharge, and 20 current pulses were delivered to evoke repetitive firing (5–8 pA, 1 s duration, 5 s intervals; Fig. 8A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Inhibitory charge transfer of evoked IPSCs to Purkinje cells (PCs) is maintained in low [Ca]_o. Dual recordings of stellate (SC) and Purkinje cells were used to record presynaptic spike discharge and postsynaptic IPSCs in 1.5 or 1.1 mm [Ca]_o in the presence of synaptic blockers. Stellate cells were held below threshold and presynaptic spike discharge evoked using 20 1 s current pulses while holding Purkinje cells at −65 mV to record inward IPSCs. A, Recordings from a synaptically connected SCs and PCs. Dashed lines indicate stellate cell spikes giving rise to eIPSCs. B, Superimposed spike-triggered averages of IPSCs and bar plots of eIPSC amplitudes in 1.5 or 1.1 mm [Ca]_o. Top traces include only eIPSCs directly associated with presynaptic spikes, while lower traces are all responses including transmission failures. SEM is indicated by shaded regions. C–E, Bar plots of spike and eIPSC properties from identified stellate–Purkinje cell pairs in 1.5 and 1.1 mm [Ca]_o. Reducing [Ca]_o from 1.5 to 1.1 mm significantly increases eIPSC failure rate in C and increases current-evoked stellate cell spike count in D. E, Mean values for spike train charge transfer from stellate to Purkinje cells in 1.5 and 1.1 mm [Ca]_o, calculated as the product of spike count in D and fits (inset) to the mean eIPSCs shown in lower records of B. Charge transfer in 1.1 mm [Ca]_o was calculated for both the actual spike count and for a hypothetical situation (1.1 hypoth) where spike count did not change from that recorded in 1.5 mm [Ca]_o. F, Bar plots of Purkinje cell spontaneous firing rates during on-cell recordings. The rate of Purkinje cell firing is not significantly different during perfusion of 1.5 or 1.1 mm [Ca]_o. Addition of picrotoxin (50 μm) reveals that the level of Purkinje cell excitability would be significantly higher in 1.1 mm [Ca]_o without the associated increase in GABAergic inhibition in the network. Sample values are shown in brackets within bar plots. *p < 0.05; **p < 0.01.

Spike-triggered averages were calculated both with and without transmission failures to determine the effect on lowering [Ca]_o on eIPSCs. When failures were not included, eIPSC amplitudes decreased from −92.8 ± 4.1 pA in 1.5 mm [Ca]_o to −62.8 ± 10.6 pA in 1.1 mm [Ca]_o (n = 5, p < 0.05; Fig. 8B, top traces). When all postsynaptic responses (regardless of failure) were included in the average as a measure of overall inhibitory influence, it was revealed that lowering [Ca]_o decreased the net amplitude of spike-related postsynaptic inhibitory events from −62.8 ± 10.6 pA in 1.5 mm [Ca]_o to −15.6 ± 4.6 pA 1.1 mm [Ca]_o (n = 5976 spikes over 5 cells, p < 0.01; Fig. 8B, right, bottom traces). This large decrease in eIPSC amplitude was due to an increase in failure rate at the stellate cell to Purkinje cell synapse from 34.6 ± 7.7% in 1.5 mm [Ca]_o to 62.8 ± 8.0% in 1.1 mm [Ca]_o (n = 5, p < 0.01; Fig. 8C).

The increase in failure rate of stellate cell synaptic transmission in 1.1 mm [Ca]_o would predict that the level of stellate cell-mediated inhibition will be diminished during repetitive synaptic input that decreases [Ca]_o in vivo. On the other hand, the gain of stellate firing is increased in 1.1 mm [Ca]_o through the action of the Ca_v3–K_v4 complex, suggesting a mechanism to counteract decreased synaptic efficacy at stellate to Purkinje cell synapses. Indeed, the average number of spikes evoked by a 1 s current step in stellate cells increased from 10.9 ± 1.3 spikes in control to 24.6 ± 3.3 spikes in 1.1 mm [Ca]_o (Fig. 8D; n = 5, p < 0.01). We thus examined the combined effects of an increased failure rate of synaptic events, decreased IPSC amplitude, and an increase in spike count on the overall charge transfer. Charge transfer was compared under three conditions: normal gain of stellate cell output in 1.5 mm [Ca]_o, increased gain of stellate cell output in 1.1 mm [Ca]_o, and a hypothetical case where stellate cell gain was unchanged in 1.1 mm [Ca]_o. To calculate charge transfer, we used simulated IPSCs for 1.5 and 1.1 mm [Ca]_o conditions, created by fitting the spike-triggered averages (including failures) from a representative cell. The charge transfer for a 1 s presynaptic spike train was calculated by multiplying the area under the simulated IPSC by the average number of spikes per second for each of the three conditions.

With an average of 11 spikes for a 1 s current step in 1.5 mm [Ca]_o (n = 5), charge transfer was predicted to be 9.7 ± 1.2 pC (Fig. 8E). Repeating these calculations for 1.1 mm [Ca]_o revealed that spike train charge transfer was maintained in the lower calcium condition when the number of spikes increased to 25 spikes/s (7.8 ± 1.0 pC, n = 5, p > 0.05; Fig. 8E). However, in a hypothetical situation where the number of spikes per unit time was unchanged in 1.1 mm [Ca]_o from that in 1.5 mm [Ca]_o (11 spikes/s), the predicted spike train charge transfer was greatly reduced to 3.4 ± 0.4 pC (n = 5; Fig. 8E), a value that is significantly different from both the 1.5 and 1.1 mm [Ca]_o conditions (p < 0.01 for both comparisons). Collectively, these data indicate that decreased synaptic efficacy in low [Ca]_o is overcome by way of a Ca_v3–K_v4 complex-mediated increase in stellate cell gain of firing.

An adaptive response of the inhibitory network restrains Purkinje cell firing

These results would predict that at the network level, the rate of Purkinje cell firing should be maintained in homeostatic manner despite fluctuations in extracellular ionic concentrations during synaptic input. To directly test this, we used on-cell recordings to monitor the rate of spontaneous Purkinje cell firing in response to modulations of [Ca]_o and [K]_o characteristic of physiologically relevant inputs. In all of these recordings, dl-AP5 (25 μm) and DNQX (10 μm) were included in the bath to block excitatory synaptic transmission. Cells were recorded from animals at P18–P20 to reduce the probability of inducing trimodal activity (McKay and Turner, 2005), and were excluded if trimodal activity was detected. Cells were recorded in an initial control solution of 1.5 mm [Ca]_o and then perfused with 1.1 mm [Ca]_o, and finally 50 μm picrotoxin to block all GABAergic inputs to assess the relative influence of the inhibitory network on Purkinje cell firing rate under each condition. In 1.5 mm [Ca]_o Purkinje cells discharged at a rate of 43.6 ± 7.8 Hz (n = 8), with no significant increase in firing rate in the presence of lowered [Ca]_o (Fig. 8F). Perfusing picrotoxin on cells maintained in 1.5 mm [Ca]_o revealed a modest but significant mean increase in firing rate of ∼5 Hz. However, in the presence of 1.1 mm [Ca]_o, picrotoxin unmasked a significant increase in Purkinje cell firing of ∼20–25 Hz over the baseline rate of firing (p < 0.05; Fig. 8F). To fully test the influence of ionic factors that can change during repetitive synaptic inputs, these tests were repeated for exposures to both 1.1 mm [Ca]_o and 4.0 mm [K]_o, with the same results (data not shown).

These data indicate that the increase in activity of inhibitory interneurons in 1.1 mm [Ca]_o acts to lower the excitability and firing rate of Purkinje cells in the order of ∼40% compared to the level that would be encountered without the inhibitory network. The reason for the increase in firing rate of Purkinje cells in the absence of GABAergic inhibition was not fully determined here. However, it does not relate to a Ca_v3–K_v4 interaction in Purkinje cells, as rat Purkinje cells have not been found to express K_v4 current (our unpublished observations). It could instead be due to an increase in excitability caused by decreased activation of calcium-dependent AHPs in Purkinje cells (Hosy et al., 2011; Engbers et al., 2012). These data are then consistent with a compensatory increase in the firing rate of stellate cells via the Ca_v3–K_v4 complex to suppress an increase in Purkinje cell excitability that would occur in low [Ca]_o. This adaptive form of inhibitory control then acts in a homeostatic fashion to limit nonspecific increases in Purkinje cell excitability during repetitive afferent activity.