Overview

Chemical synaptic communication critically depends on specific calcium ion (Ca2+) signals on both sides of the synapse. The plasma membranes of the pre- and post-synaptic cellular compartments provide a diffusion barrier between the high Ca2+ concentration ([Ca2+]) in the extracellular space (∼1 mM range) and the low resting cytosolic [Ca2+] (∼20 to 50 nM [1]). The plasma membranes of these compartments contain the Ca2+ regulatory mechanisms specialized to control passive, “down the gradient,” or active, and energy consuming “against the gradient,” trans-membrane Ca2+ fluxes. Cerebellar Purkinje neurons (PNs) are a well studied and instructive example of a neuron that receives a variety of different synaptic inputs with distinct properties, all of which influence cerebellar function. In this review, we will explore how the different available Ca2+ regulatory mechanisms influence the unique properties of these different types of PN synapses.

Each of the approximately hundred thousand synapses formed with every PN in the mammalian cerebellum is a hotspot of chemical communication that relies on Ca2+-dependent events. Pre-synaptic Ca2+ initiates neurotransmitter release [2], while Ca2+ has several functions at the post-synaptic membrane. Being a charge carrier, Ca2+ flux across the PN membrane generates a depolarizing inward current that contributes directly to the electrical responsiveness of the post-synaptic neuron. Furthermore, Ca2+ influx exerts indirect effects on the electrical behaviour of the neuron by activating Ca2+-dependent outward currents (carried by Ca2+-dependent K+ channels and discussed in more detail by Anwar et al., in the accompanying review). In addition to these immediate (“real time”) actions, Ca2+ also acts as a chemical messenger to initiate changes in long-term synaptic efficacy, where the amplitude, dynamics, and location of the Ca2+ signal can influence whether the synapses strengthen or weaken (discussed in more detail by Finch and Augustine, in the accompanying review).

Here, we aim first to describe the repertoire of Ca2+ regulatory mechanisms that generate and regulate the amplitude and timing of Ca2+ fluxes during synaptic transmission (the “Ca2+ toolbox”). After describing how these basic mechanisms act in isolation, we will then explore how they interact to generate the unique functional properties of the following important PN synapses; the excitatory parallel fibre (PF) and climbing fibre (CF) synapses as well as the inhibitory synapses (IS) from molecular layer stellate and basket cells (see Fig. 1).

Fig. 1.
figure 1

a Schematic of a cerebellar Purkinje neuron (PN, blue) with its main synapses, the excitatory climbing fibre input from the inferior olive (CF-PN, red), the excitatory parallel fibre input from the cerebellar granule cells (PF-PN, light red), and the inhibitory synapses (IS-PN, green) from the stellate and basket cell inhibitory interneurones. b Calcium regulatory mechanisms within the different synapse compartments are shown; colours correspond to the colour used for the different synapses in a. VGCC voltage-gated calcium channel, PMCA plasma membrane calcium ATPase, NCX sodium calcium exchanger, NCKX sodium calcium potassium dependent exchanger, RyR ryanodine receptor, InsP 3 -R inositol trisphosphate receptor, NMDA-R N-methyl-d-aspartate receptor, mGluR metabotropic glutamate receptor sub-type, CB calbindin, PV parvalbumin, Clr calretinin

Full size image

The Components of the Ca2+ Toolbox at Cerebellar Synapses

The Plasma Membrane

As the main barrier to external high [Ca2+], the components of the Ca2+ toolkit that reside within the pre- and post-synaptic plasma membranes are critical for providing controlled entry and extrusion of Ca2+.

Voltage-Gated Ca2+ Channels

Voltage-gated Ca2+ channels (VGCC) are the route for practically all Ca2+ entering the pre-synaptic and post-synaptic compartments of PNs (reviewed also in [3]). VGCCs can be separated into the high-voltage activated channels (HVA), L-type, P/Q-type, and N-type channels and the low-voltage activated channels, or T-type. The HVA channels consist of the ion channel alpha1 subunit accompanied by the accessory alpha2delta and beta subunits to create functional and highly modifiable channels (for a review see [4]). Functional diversity of HVA channels arises through differences in the alpha1 subunit sequence to generate four main families of channels (Table 1). The PN richly expresses P/Q (alpha 1A; Cav2.1) type channels, called this following their first discovery in PNs [5] where, in contrast to a wide variety of other neurons, they are abundant in the post-synaptic dendrites [6, 7]. These P/Q channels contribute most to the PN’s HVA Ca2+ current, and although some dihydropyridine-sensitive L-type current (mediated by alpha 1C; Cav1.2) also contribute to this, there is little contribution from fast N-type currents (alpha 1B; Cav2.2) [8].There is some evidence for functional expression of R-type channels (alpha 1E subunit containing; Cav2.3) [9], particularly their contribution with T-type channels to the low threshold PN Ca2+ spike [10].

Table 1 Variety of sub-types of the alpha1 subunit of voltage-gated calcium channels expressed and functional at cerebellar PN synapses
Full size table

While opening these channels provides the major route for Ca2+ flux across the plasma membranes of the pre- and post-synaptic sites, the amount of Ca2+ entry is also controlled by the kinetics of channel activation and inactivation. The rich contribution from the fast-activating and rapidly inactivating P/Q channels in the PN dendrites provides for sharp Ca2+ transients and contrasts with the slower and more prolonged Ca2+ entry that accompanies the opening of L-type VGCCs in other neurons. The reliance on P/Q type channels and paucity of N-type channels may provide a degree of signalling simplicity in the PN dendrite. Opening of both types of channels is not only controlled by membrane voltage but also through interaction with Ca2+-sensing proteins such as calmodulin [11]. The details of these interactions differ between the channel types but their general effect is to limit further Ca2+ entry when calcium rises to a certain level. Furthermore, both types of channels can be modulated by the activation of a variety of G-protein-coupled receptors, influencing both their open kinetics and their insertion into membranes.

Ca2+ Pumps and Exchangers

Ca2+ influx is balanced by mechanisms that return, or extrude, Ca2+ to the extracellular space (or to intracellular compartments, see below “Pre-Synaptic Ca2+ Control Mechanisms”). On the long term, the net balance of Ca2+ influx and efflux has to be zero in order to maintain low intracellular [Ca2+]. Influx can however exceed efflux on a time scale faster than a few seconds to cause a transient rise in intracellular [Ca2+] (Ca2+ transients). The main routes for Ca2+ efflux across the plasma membrane are via the plasma membrane Ca2+ ATPase (PMCA), often called the Ca2+ pump, and the sodium (Na+) Ca2+ exchangers (NCX and NCKX). These Ca2+ regulatory mechanisms transport Ca2+ against their concentration gradient from the cytosol into the extracellular space using energy obtained directly from ATP, in the case of the pump, or indirectly using the Na+ gradient, in the case of the exchangers.

All four PMCA isoforms (1–4) are present in the cerebellum, but PMCA2 is the most highly expressed [12]. PMCA2 is expressed throughout the cerebellar cortex where it is enriched in the molecular layer and the PN soma, dendrites, and dendritic spines [13]. Activation of PMCAs most often requires the binding of Ca2+ to the high-affinity Ca2+ sensing protein calmodulin. Ca2+ calmodulin then binds to the PMCA and relieves the intrinsic autoinhibition resident within its C-terminal tail. At physiological levels of calmodulin (up to micromolar levels in neurons), the apparent affinity (k 0.5) of PMCA2 for Ca2+ is 60 nM, close to resting [Ca2+] in the PN. This means that Ca2+ transport by PMCA2 can adjust even small fluctuations in cytosolic [Ca2+] [14]. A faster response to changes in [Ca2+] is achieved by PMCA2a, the shortened splice form of PMCA2 that lacks the full calmodulin-binding site. With an apparent rate constant for activation of 0.07 s−1 (in the presence of 500nM Ca2+) PMCA2a is one of the fastest activating PMCA Ca2+ transporters [15]. Interestingly, this splice variant is highly expressed in PN spines and dendrites [16].

The other major active Ca2+ transporter located at the plasma membrane is the Na+/Ca2+ exchanger (NCX), also highly expressed in the cerebellum. Isoforms NCX 1 and 3 are expressed in the cerebellar cortex in both granule cell and molecular layers and at pre- and post-synaptic membranes [17]. NCX exchanges three Na+ for one Ca2+, and the direction of Ca2+ flux depends on the relative electrochemical driving force of the Na+ and Ca2+ gradients. Working in its forward mode, the exchanger utilizes the inward Na+ gradient to extrude Ca2+ to the extracellular space. Unlike the PMCA, activation of NCX mediated Ca2+ efflux does not require calmodulin. Instead, intracellular Ca2+ (and Na+) binds to and activates the exchanger directly. Furthermore, the apparent affinity of Ca2+ binding (k 0.5) changes with internal [Ca2+]. At 500 nM [Ca2+], k 0.5 of the exchanger’s efflux activity in the squid giant axon is 22 µM [18], an approximately 500-fold lower apparent affinity for Ca2+ than PMCA2. Under conditions where the Na+ gradient is weakened (that is when intracellular Na+ is high), the NCX exchanger works in reverse mode to extrude Na+ and accumulate cytosolic Ca2+ [19]. In contrast, the Na+–Ca2+ potassium (K+) exchanger (NCKX), normally works in reverse mode to bring Ca2+ into cells but will work in forward mode as a Ca2+ extrusion mechanism when intracellular [Ca2+] is >300 µM. Furthermore, the forward mode of NCKX2 exhibits faster kinetics than the reverse (influx) mode [20]. NCKX is functional in cerebellar granule cells [21], so it is possible that this Ca2+ extrusion mechanism contributes to Ca2+ dynamics at the PF-PN synapse under certain conditions.

All together, these very different mechanisms provide for well-regulated Ca2+ efflux over a range of [Ca2+] at cerebellar synapses. We predict that the PMCA2 pump will be readily recruited by small fluctuations in intracellular Ca2+ but that, during larger intracellular Ca2+ excursions, when PMCA becomes saturated, Ca2+ efflux can continue via the NCX and possibly even the NCKX. However, the relative contribution of PMCA, NCX, and NCKX at the different compartments of cerebellar synapses and under different functional conditions is yet to be determined (see also below).

The Cytosol

Ca2+ Buffers

Cytosolic proteins that bind Ca2+ in the physiological [Ca2+] range and are expressed at concentrations of hundreds of micro- or millimolars can act as Ca2+ buffers. The prominent Ca2+ buffers expressed within the cerebellum are calretinin, calbindin, and parvalbumin, the latter two both richly expressed within the PN cytosol [22,] whereas inhibitory stellate and basket cells only express parvalbumin. Calretinin, expressed by granule cells and therefore present in the pre-synaptic PF terminals [23], has the fastest apparent Ca2+-binding kinetics of >>108 s−1 and is often referred to as a “fast buffer.” The Ca2+-binding kinetics of calbindin and parvalbumin are rather slower (but see also “Climbing Fiber Post-Synapse: CF Stimulation Leads to a Very Large Post-Synaptic Dendritic [Ca2+] Signal That Is Heavily Influenced by a Variety of Ca2+ Regulatory Mechanisms”). All three of the Ca2+ buffers express multiple copies of the EF hand Ca2+-binding motif and have an affinity for Ca2+ between 10 nM and 10 µM [24]. Ca2+ buffers have the capacity to efficiently dampen the effect of short-lasting but massive Ca2+ influx (that is a smaller rise in free [Ca2+] than expected from the actual amount of Ca2+ influx). Their limitation is that they can be saturated, or overwhelmed, during prolonged periods of strong Ca2+ influx, but the extent to which this occurs depends upon their expression levels. Interestingly, Ca2+ buffer expression and hence buffer capacity increases during cerebellar development [25]. Conversely, PN calbindin levels decrease during aging [26] and also before the onset of cerebellar ataxia [27], perhaps in the latter case as a pretext to cell death as occurs in several forms of human ataxias.

Intracellular, ER Ca2+ “Stores”

In addition to binding of Ca2+ by buffers, Ca2+ can also be removed from the cytosol by uptake (sequestration) into intracellular organelles. The major organelle used for this purpose in neurons is the endoplasmic reticulum (ER) where Ca2+ is bound to proteins such as calsequestrin within the lumen [28]. Specific mechanisms also allow Ca2+ in the ER to be mobilized (returned to the cytosol) so that the ER provides bidirectional control of Ca2+ dynamics: either to provide extra cytosolic [Ca2+] or to act as a sink for excessive cytosolic [Ca2+]. Ca2+ uptake into the ER is driven by the sarcoplasmic/endoplasmic reticulum Ca2+ ATPase (SERCA), another family of Ca2+ ATPase pump that exists as a variety of isoforms in the cerebellum. This protein shares significant homology with the PMCA, although SERCA2b’s apparent affinity, k 0.5, for Ca2+ is approximately 300 nM, rather higher than PMCA2 (see above). SERCAs [29], like the PMCAs, also have a fast apparent rate constant for activation by Ca2+ (see above discussion on fast PMCA2a, “Ca2+ Pumps and Exchangers”). The SERCA isoform 2 is widely distributed in the cerebellum [30], and more recently, SERCA has been localized to the PNs, the molecular layer interneurons, and also in pre- and post-synaptic structures [31]. The SERCA 3 isoform, which invariably co-expresses with the SERCA2b isoform [32], is also enriched within PNs [33]. SERCA3’s lower apparent affinity for Ca2+ (k 0.5 = approximately 1 µM) [32] is predicted to enable Ca2+ uptake into the ER during very large excursions in cytosolic [Ca2+]. Indeed, pharmacological studies showed that SERCA makes the largest contribution to somatic [Ca2+] recovery following a massive PN depolarization [34], making SERCA effective at [Ca2+] concentrations ranging between that of Ca2+ buffer binding and the activation of NCX (see above), with a contribution exceeding that of PMCA, at least at high, peak [Ca2+] in the PN soma.

When loaded with Ca2+, the ER store can provide Ca2+ to amplify the effect of Ca2+ influx. The mechanism is called Ca2+-induced Ca2+ release (CICR) and is triggered by cytosolic Ca2+ and mediated by ryanodine (Ry) or inositol trisphosphate (InsP3)-sensitive channels in the ER membrane. In addition to CICR, Ca2+ can be mobilized from filled ER stores by high concentrations of InsP3 generated in the cytosol following activation of metabotropic glutamate receptors (sub-type 1, mGluR1). These G-protein-coupled receptors are expressed at very high levels in PN dendrites. Both a and b mGluR1 splice variants (mGluR1a and mGluR1b) are localized in the post-synaptic membrane around PF synapses (peri-synaptic localization) and at extra-synaptic sites [35]. This localization explains the requirement of pooling and spillover of synaptically released glutamate to activate mGluR1 [36].

The PNs are unique among neurons, as they express the skeletal muscle type of ryanodine channel/receptor, RyR1 [37], while most other neurons (including the PN) express the cardiac isoform RyR2 [38]. While RyR1 mediates direct depolarisation-induced Ca2+ release [39] from the sarcoplasmic reticulum, as occurring at the skeletal muscle triad, there is no physiological evidence to indicate that this direct excitation-coupling occurs in PNs [40]. PNs also express some of the highest levels of the type 1 InsP3 receptor in the central nervous system [41, 42].

In the PN, both InsP3Rs and RyRs seem to access the same intracellular Ca2+ pool [43], although removal of InsP3 type 1 receptors from PNs in InsP3-1 knockout mice had little effect upon the magnitude of PN CICR activated by caffeine [44].

Regardless of the route, the effectiveness and contribution of the ER to pre- and post-synaptic transmission, whether behaving as a sink or a source of Ca2+, will clearly depend upon a variety of factors, such as the fill state, the balance between release and refilling during synaptic activity and the kinetics of the different Ca2+ release channels. For example, important questions remain as to whether the ER is primed to release or take up Ca2+ in different states and whether the store state under these conditions varies between the different synapses in the cerebellum.

While our summary of the Ca2+ regulatory mechanisms that operate in the cerebellum is not complete, we have covered the key components required to understand how the integration of these mechanisms contributes to cerebellar synapse function. In the following paragraphs, we will explore how these mechanisms work together to provide flexible-controlled generation of pre- and post-synaptic Ca2+ transients for effective neurotransmission at PN synapses.

Pre-synaptic Ca2+ Control Mechanisms

Excitatory PF Pre-synapse: At this Low Release Probability Synapse, Small and Accumulating Ca2+ Signals are Fine-Tuned by Ca2+ Regulatory Mechanisms to Control Pre-synaptic Ca2+ Transients and Transmitter Release

Entry of Ca2+ into pre-synaptic PF terminals occurs primarily through highly expressed P/Q-type (Cav2.1) channels [7], since omega-Aga-IVA toxin has the greatest inhibitory effect upon synaptic transmission at this synapse. However, N-type (Cav2.2) as well as possibly R-type (Cav2.3) VGCCs also make a contribution [45–47].

The PF input to the PN is a low release probability synapse, but release probability increases (facilitates) for a short period following initial activation. This phenomenon is termed paired pulse facilitation (or frequency facilitation in the case of repetitive activity) and has been explained by the effect of residual pre-synaptic [Ca2+] that adds to the second response in a paired stimulation paradigm [48]. Release probability and frequency facilitation are directly related to the amplitude and decay rate of the pre-synaptic [Ca2+] transient. The amplitude of the Ca2+ signal is directly regulated by pre-synaptic auto- and hetero-receptors (GABAB, cannabinoid, CB1, and mGluR4), acting on pre-synaptic VGCCs [49, 50]. The control of the decay of the pre-synaptic [Ca2+] transient, and hence the amount of residual pre-synaptic Ca2+, falls to Ca2+ extrusion mechanisms [51] and Ca2+ buffers [52, 53]. Somewhat surprisingly, the behaviour of paired pulse facilitation at the PF-PN synapse was not significantly altered in the calretinin knockout mouse even though the mice exhibit a lack of motor coordination when challenged [54]. A possible reason for this may lie in the relatively small contribution from calretinin alone as an endogenous buffer during a paired pulse stimulation paradigm. More recently, a rather heterogeneous distribution of buffer capacity within individual PF boutons has been shown [55] and may arise through different expression levels of endogenous Ca2+ buffers between and possibly even within boutons. As yet, the contribution of factors that influence buffer capacity of individual PF boutons are not fully understood, despite the fact that their relative contributions will influence the release properties and facilitation of this important pre-synaptic compartment (see also below). Most recently, heterogeneity of individual [Ca2+] transients in PF pre-synaptic boutons was ascribed to differences in pre-synaptic action of auto- and hetero-receptors [56].

Physiological evidence indicates that Ca2+ extrusion via PMCA2 normally shapes the decay of the pre-synaptic PF [Ca2+] transient since the rate of [Ca2+] recovery is delayed in PFs of PMCA2 knockout mice [51]. Furthermore, this slowed recovery in the PMCA2 PFs is expressed as a prolongation of paired pulse facilitation at this synapse. This indicates that an accumulation of residual [Ca2+] is enhanced in the absence of PMCA2. Modelling studies [57] also indicate that NCX may play a similar role, but this has not been experimentally tested to date. It remains possible that some of the heterogeneity between individual PF boutons arises from different expression levels and even types of Ca2+ extrusion mechanisms (PMCA vs NCX isoforms, see above, “Ca2+ Pumps and Exchangers”)

The ER Ca2+ stores seem to play very little, if any, part in the control of PF pre-synaptic [Ca2+] under physiological conditions, as revealed by the lack of sensitivity of paired pulse facilitation to thapsigargin (a SERCA inhibitor that empties ER stores) at this synapse [58]. This finding is consistent with the idea that release of Ca2+ from the stores is really only needed when amplification of the [Ca2+] is necessary. Presumably, given the low release probability of the PF-PN synapse, additional amplification of pre-synaptic [Ca2+] is not needed since accumulation is sufficient. It is even possible that restricted space within the small volume of the PF terminal might preclude the participation of the stores, although more recently, a contribution from Ry-sensitive Ca2+ stores to mGluR-dependent alterations of release probability at mature PF synapses has been described [59]. This raises the possibility that, under special circumstances, ER Ca2+ stores can contribute to the facilitatory properties of PFs.

Excitatory CF Pre-synapse: Where Ca2+ Dynamics in the Pre-synapse are Dominated by Ca2+ Influx to Ensure Reliable Transmission

The importance of the CF input to the PNs is reflected by its strength and reliability. The high probability of glutamate release at each of numerous release sites [60] distributed over a very large area of PN dendrite ensures that there is no synaptic failure.

At the CF pre-synapse, the specific toxin omega-Aga-IVA largely reduces the synaptic response, indicating the importance of P/Q-type VGCCs, although N-type channels also contribute to this reduction [47, 61]. The [Ca2+] transient in the CF pre-synaptic compartment has not been systematically measured, but the high release probability of the synapse suggests that Ca2+ influx will dominate. We anticipate that during high frequency repetitive stimulation of the CF, accumulation of Ca2+ occurs but that depletion of vesicles dictates that synapse strength decreases (exhibits paired-pulse depression). The paucity of literature on the [Ca2+] dynamics within the CF pre-synaptic compartment is presumably because the behaviour of this synapse is dominated by the anatomical specialisation of its release machinery. We could predict, however, that the CF pre-synapse may behave rather similarly to the calyx of Held, another high release probability, depressing synapse where repetitive stimulation leads to an increase in the amplitude of the pre-synaptic Ca2+ transient per action potential (presumably via accumulation) even as the post-synaptic response decreases [62].

Inhibitory Basket Cell Pre-synapse: Here, the Pre-synaptic [Ca2+] Transient is Dominated by Amplification of [Ca2+] via Release from Intracellular Stores Resulting in a High Transmitter Release Probability

Like the CF-PN synapse, the release of GABA from basket/stellate cells onto PNs is a high release probability synapse [63], reflecting the importance of feed forward synaptic inhibition at the granule cell to PN pathway for cerebellar function.

Ca2+ entry into the GABA-ergic pre-synaptic terminals uses P/Q-type (Cav2.1) channels present on GABA-ergic stellate and basket cell terminals within the molecular layer of mouse cerebellar cortex. Opening of these channels is required for GABA release, as determined by the sensitivity of PN spontaneous GABA-mediated inhibitory post-synaptic currents (IPSCs) to 200 nM omega-Aga-IVA ([64]; but see also [65], and below). However, there is little or no contribution from N-, R-, or T-type (Cav3.1) VGCCs to GABA release from stellate and basket cell terminals (but see also [47]). More recently, NMDA receptor-mediated Ca2+ influx and depolarisation of stellate cell dendrites has been shown to release GABA from stellate cell axon varicosities [66]. In contrast, brief activation of AMPA receptors by glutamate suppresses Ca2+ entry by acting at pre-synaptic VGCCs on the basket cell terminals [67].

Measurements of Ca2+ transients in the large basket cell axonal varicosities during GABAergic transmission revealed large spontaneous [Ca2+] transients (SCaTs) coincident [65] with the giant IPSCs in the PNs. These events persist in the absence of pre-synaptic action potentials [68] and are driven by very large excursions in [Ca2+] amplified by additional Ca2+ released from RyR-sensitive ER Ca2+ stores [65]. By utilizing Ca2+ from ER stores in this way, the synapse is able to generate the size of Ca2+ signal that is required. Presumably, the large size of the varicosity means that stores of sufficient capacity can also be accommodated. We predict that the fill state of the ER stores should influence the amplitude and distribution of SCaTs, but whether these ER stores can also act as a sink for [Ca2+] is not known. Moreover, at this synapse, the SCaTs persisted in the presence of a high concentration of the P/Q/N-type Ca2+ channel antagonist, omega-conotoxin MVIIC, suggesting that Ca2+ influx through VGCC is not required as a trigger [65]. Nevertheless, other pre-synaptic receptors may provide additional ways to modify [Ca2+] dynamics at this synapse [69].

More recently, coupling between activation of glutamate operated pre-synaptic NMDA receptors and RyR-sensitive ER stores at basket cell terminals has been shown during depolarisation-induced potentiation of inhibition. In this mode, [Ca2+] elevation in the PN drives retrograde activation of pre-synaptic NMDA receptors to elevate [Ca2+] in the pre-synaptic terminal through release from intracellular stores [70].

The main mechanism that limits the large rises in [Ca2+] in these pre-synaptic terminals is endogenous Ca2+ buffers. The basket and stellate cells express high concentrations of parvalbumin (PV) to ensure a very high endogenous Ca2+ buffer capacity. The high action potential firing frequency that is typical for basket cells may well demand this. Indeed, PV is known to be required for the control of release probability at the basket cell to PN synapse; without PV in PV−/− mice, the synapse switches from being a high release probability depressing synapse to a facilitating one [71]. PV is also critical for shaping the pre-synaptic Ca2+ transient and GABA release, where it recovers the initial part of the [Ca2+] transient. However, PV also acts more slowly to release previously bound Ca2+ as a way to sustain pre-synaptic [Ca2+] during high-frequency action potential firing. In this way, GABA release is sustained by the terminals even between burst firing of the stellate and basket cells [25].

Whether the PMCA and NCX make a physiological contribution to Ca2+ recovery mechanisms at the basket cell to PN synapse is not known, although both proteins are present [16, 17]. The very fast decay of action potential-evoked [Ca2+] transients in basket cell axon varicosities under conditions of minimal Ca2+ buffering (from the Ca2+ indicator) predicts that a highly active and efficient Ca2+ extrusion [65] mechanism is at work.

Post-synaptic Ca2+ Control Mechanisms

Climbing Fiber Post-synapse: CF Stimulation Leads to a Very Large Post-synaptic Dendritic [Ca2+] Signal that is Heavily Influenced by a Variety of Ca2+ Regulatory Mechanisms

The fail-safe behaviour of the pre-synaptic Ca2+ dynamics at this synapse (see “Excitatory CF Pre-synapse: Where Ca2+ Dynamics in the Pre-synapse are Dominated by Ca2+ Influx to Ensure Reliable Transmission” above) is repeated at its post-synapse. The massive and distributed glutamate release associated with activation of the CF causes an exceptionally powerful post-synaptic response known as the complex spike [72]. The complex spike is a large, active depolarizing response triggered by the AMPA receptor-mediated post-synaptic CF current. This large depolarization triggers a burst of fast action potentials generated by somatic and axonal sodium channels [73] and involves the activation of VGCC in the dendrites [74, 75], although the latter are not critical for the electrical appearance of the complex spike at the level of the axon and cell body [73] their activity is required for a dendritic Ca2+ signal such as required during PN synapse plasticity (see Finch and Augustine this issue]. Morphologically, the CF pre-synaptic terminal is restricted to the proximal portion of the PN dendrite [76], but the complex spike spreads electrotonically into distal dendrites and activates VGCC in spines, too [77–79]. More recently, a contribution from glutamate NMDA receptor-mediated post-synaptic Ca2+ influx to the CF response was identified in more mature PNs [80, 81].

The [Ca2+] rise associated with a complex spike has been quantified in several studies. Taking into account the kinetics of the different fluorescent [Ca2+] indicators used [74, 77], peak amplitudes of the complex spike [Ca2+] transient tend to be larger in spines (∼400 nM) as compared to dendritic shafts (∼200 nM). The rising phase of this transient is determined by the duration of the depolarizing response (20–50 ms), while Ca2+ buffers and extrusion mechanisms determine the bi-exponential decay time course of the Ca2+ signal (20–30 ms and 120–150 ms) that is only slightly faster in spines compared with dendritic shafts [77]. Furthermore, calbindin, the fastest among the PN main buffers, acts as a shuttle for Ca2+ transfer between the spines and dendrite [82].

Given the large CF stimulated [Ca2+] rise, it is not surprising that PNs have a very high capacity to buffer Ca2+, both in their soma and dendrites. The PN’s capacity to buffer [Ca2+] increases with developmental age as the expression levels of endogenous Ca2+ buffers increases [1]. Studies to remove calbindin and parvalbumin showed that, together, these slower Ca2+ buffers influence both the peak and the biphasic decay of the CF-induced [Ca2+] transient in PNs [77]. Removal of both Ca2+-binding proteins allowed peak Ca2+ to rise higher and remain elevated for longer. Individually, it is calbindin that exerts most influence on the peak and early phase of the [Ca2+] transient, whereas parvalbumin, with its slower kinetics, has a greater influence on the slower phase of Ca2+ recovery. Furthermore, removal of these important endogenous Ca2+ buffers influences the firing behaviour of the PNs; in two separate studies on knockout of calbindin and knockout of calretinin, the most significant effect was an increase in the frequency of simple spike fast firing and a reduced duration and post-spike pause of the complex spike [83, 54].

Ca2+ buffers, while very effective at limiting [Ca2+] rises during brief periods of large Ca2+ influx, will saturate during large and longer lasting periods of Ca2+ influx. Under these conditions, it is only the active extrusion of Ca2+ from the cytosol that can maintain [Ca2+] homeostasis, although active extrusion by PMCA and NCX does still influence the shape of short-lasting [Ca2+] transients. The first indication of this came from an elegant and controlled study using pharmacological tools to block extrusion mechanisms [34]. When the PN was stimulated with a large [Ca2+] load, both PMCA and NCX contributed to the decay of the Ca2+ transient: the pump by about 6% and the exchanger around 18%. More recently, in PMCA2 heterozygous knockout mice, where PMCA2 expression in the cerebellum is reduced by half, we observed a doubling of the recovery time of the [Ca2+] transient in the PN dendrite. This greater involvement of PMCA2 in the dendrite than expected from the earlier measurements from the cell body reflects the larger Ca2+ channel density and the larger surface to volume ratio of the PN dendrites (compared with the soma) [84]. Since PMCA pump activity is also under control of the biochemical environment [85], its role in the PN dendrites may have consequences for post-synaptic PN plasticity under certain conditions.

The post-synaptic CF-induced Ca2+ transients are also amplified by the release of Ca2+ from RyR-sensitive intracellular stores [40] and perhaps also by InsP3-mediated rises in intracellular [Ca2+] following mGluR activation ([86] but see [36]). While the high density of fast-activating (and inactivating) P/Q channels in the PN dendrite provides for the initial fast depolarisation of the dendrite and the accompanying fast rise in [Ca2+] during the complex spike, it is the slower propagation of the CF [Ca2+] signal throughout the PN dendrite [87] that utilises the slower release of Ca2+ from the intracellular stores. Indeed, the stores, their fill state, and the type of release may even provide for heterogeneity or compartmentalisation of the CF-induced Ca2+ signal between dendrites and even within individual compartments of the same dendrites.

PF Post-Synapse: Here, the Small Unitary EPSPs Need Both Spatial and Temporal Summation to Engage VGCCs and mGluRs to Amplify the Ca2+ Response via Ca2+ Release from Intracellular Stores

Single PF-PC synapses produce small excitatory post-synaptic potentials (EPSPs) (∼2 mV) [88] that are insufficient to significantly activate VGCC. Only when a significant number of PFs are activated together can spatial summation lead to EPSPs large enough to activate VGCC and Ca2+ influx [89]. Modelling indicates that P-type channels are important for the amplification of spontaneous unsynchronized excitatory synaptic inputs to the PN distal dendrite [90]. More recently, a contribution from T-type currents to PF-PN dendrite/spine excitability and their Ca2+ dynamics has been revealed using both pharmacology [88] and genetic deletion of Cav 3.1 channels [91].

Early work indicated that responses to exogenously applied glutamate or an agonist of mGluR-evoked large [Ca2+] rises in PN dendrites. These Ca2+ transients persisted in the absence of extracellular Ca2+, suggesting that their source was an intracellular Ca2+ store [92, 93]. Repetitive stimulation of PFs causes glutamate release and activation of mGluR1 receptors to initiate a slow depolarisation and mobilisation of Ca2+ from internal stores via rises in InsP3 [94–96]. Furthermore, since the PF-mediated Ca2+ rise in single spines is rather restricted [95], a close apposition of a Ca2+ efflux mechanism with this Ca2+ amplification mechanism is suggested. Indeed, mGluR1, the scaffold protein Homer3 and InsP3 receptors together form a complex with PMCA2 [97], while T-type Ca2+ channels are modulated by co-localized mGluR1 at PF post-synapses [91].

Inhibitory Post-synapse: Strong Post-synaptic Inhibition Shunts Post-synaptic VGCC Activation to Help Control PN Output

The large inhibitory synaptic input from the stellate and basket cells provides a modifying influence on the excitability of the dendrites of the PN. Stimulation of the molecular layer to activate feed-forward inhibition simultaneous with CF activation leads to a reduction in the CF-induced [Ca2+] transient in the dendrites and a curtailment of the Ca2+ spike [98]. The strength and timing of this feed-forward inhibition is necessary to control the precision of PN spike timing and is also sufficient to reduce the effectiveness of summation of excitatory synaptic inputs [99] presumably because the membrane becomes shunted. Modeling also predicts the importance of feed-forward inhibition for the integration of asynchronous synaptic activity in the PN dendrite [90].

Concluding Remarks

Although clearly not completely comprehensive, we hope that this review has highlighted how synaptic transmission to PNs relies on a multitude of complex and interacting Ca2+ regulatory mechanisms. The different PN synapses use diverse mechanisms within their pre- and post-synaptic compartments in order to generate Ca2+ signals that are specialized for specific modes of neurotransmitter release and post-synaptic PN behaviour. Furthermore, these mechanisms can operate in different modes in response to alterations in the [Ca2+] within the compartment. The resulting calcium dynamics will in turn be influenced by the timing of the electrical activity that initially triggered the Ca2+ rise, by the physical geometry, and by the combination of molecular components within the very different pre- and post-synaptic compartments. As we have seen, significant advances in fluorescence based [Ca2+] measurement technology has revealed rather precise contributions from a number of regulatory mechanisms, but remaining contributors await investigation before a full picture can emerge. Furthermore, the interaction of the Ca2+ regulatory mechanisms brings another level of complexity as compared to the properties of each mechanism in isolation, and this particular challenge will benefit from the development and application of computational models to quantitative physiological data. The importance of the interaction of [Ca2+] regulation mechanisms at PN synapses is highlighted by the fact that removal of even one component of the Ca2+ toolkit has deleterious consequences for cerebellar function, such as loss of motor coordination. To help solve this complex puzzle, genetic modification or deletion of the individual pieces of the Ca2+ toolkit will continue to provide a powerful approach. With its variety of synaptic inputs all geared to one important output and the emergence of PN-specific knockout technology, the PN will continue to provide an excellent model to improve our understanding of how Ca2+ regulatory mechanisms influence neuronal function.