Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

The climbing fiber input to cerebellar Purkinje cells represents one of the strongest synaptic connections in the brain, generating unitary postsynaptic excitatory conductances so powerful that this single input generates a massive spike burst known as the complex spike, accompanied by widespread increases in intracellular calcium concentrations. The magnitude of this response alone implies that it represents a very important signal within the cerebellar system, and as such it has been proposed to have a range of important functions including acting as an error signal to trigger associative plasticity at parallel fiber synapses (Albus 1971; Ito 2001; Marr 1969). However, the climbing fiber signal is not only very large but is attributed with two other important features: firstly, under some circumstances, complex spikes can be periodic, suggesting that they represent a timing signal that regulates Purkinje cell output. Secondly, synchronous complex spike activity can occur across organized populations of Purkinje cells and so it is possible that climbing fiber input may assemble certain functional subpopulations. The source of both of these complex spike characteristics must reside in the properties of the neurons of the inferior olivary nuclei in the brainstem, which give rise to the climbing fibers.

The intimate association between the cerebellum and the inferior olivary (IO) nuclei of the brainstem was recognized long before electrophysiological recordings or detailed anatomical data were available, mostly from the studies where cerebellar lesions resulted in degeneration of this nucleus (Klimoff 1899; Keller 1901; Ramon y Cajal 1909). The substrate for this relationship became clear with the discovery that the climbing fiber pathway is composed of the axons of olivary neurons (Szentagothai and Rajkovits 1959; Eccles et al. 1966), and that activity in the cerebellar cortex is in turn fed back to the IO via the inhibitory output of the deep cerebellar nuclei (DCN) (de Zeeuw et al. 1989) in a topographically organized manner (Szentagothai and Rajkovits 1959; De Zeeuw et al. 1998; Apps and Garwicz 2005). Often referred to as the olivocerebellar system, one of its presumed functions is the generation of spatiotemporal patterns of neural activity (Jacobson et al. 2008) that, ultimately, modulates the output of the motor system to produce smooth movement. Underlying this view is the ever-increasing evidence that the IO has remarkable oscillatory properties, tempting the researchers to see the olive as a kind of clock that can provide a timing signal for coordination of cerebellar activity. This is a controversial view, however, and is not widely accepted. This chapter will review the evidence for the existence of such a clock, what makes it tick, and how it may direct cerebellar responses.

Basic Features of the Inferior Olive

Located bilaterally in the medulla oblongata, the inferior olives are composed of four subnuclei: the Principal Olive, the Medial and Dorsal Accessory Olives (Fig. 44.1a), and the Dorsal cap of Kooy (De Zeeuw et al. 1998). The neurons contained in these nuclei are almost exclusively projection neurons, and can be divided into two morphologically distinct populations defined by their dendritic arborizations: a phylogenetically more ancestral type carrying a small number of straight dendrites, and a type with a more complex dendritic tree characterized by multiple curling dendrites (Fig. 44.1b) (Scheibel and Scheibel 1955). The cells make no chemical synapses between each other, but are extensively electrotonically coupled by gap junctions (Llinas et al. 1974; Sotelo et al. 1974). These are thought to be composed principally of connexin 36 and to occur mainly between dendritic spines (De Zeeuw et al. 1995, 1997).

Fig. 44.1
figure 00441figure 00441

Anatomy of the inferior olive. (a) Nissl stain of a transverse section through the medulla oblongata showing the subnuclei of the inferior olive. MAO medial accessory olive, DAO dorsal accessory olive, PO principal olive, P posterior, A anterior (scale bar 500 μm). (b) Two adjacent olivary neurons filled with biocytin (scale bar 50 μm). (c) Olivary neurons are electrically connected via gap junctions between their spines, which also receive both excitatory and inhibitory input, forming a trisynaptic junction. (d) Whole cell patch clamp recording from a pair of coupled olivary neurons showing bidirectional transfer of slow voltage components in response to a current injection in one and then the other neuron, but little coupling of action potentials (seen here as rebound spikes)

Full size image

The axon of each olivary neuron ascends across the midline of the brainstem, where it ultimately branches to form approximately ten collaterals, each of which extends into the cerebellar cortex as a climbing fiber and innervates a single Purkinje cell. The climbing fibers from a given olivary neuron innervate a sagittally oriented band of cerebellar cortex called a microzone (Apps and Garwicz 2005) and the activity of Purkinje cells in this zone targets the DCN which in turn projects to the same area of the IO, thereby closing an olivocerebellar loop (De Zeeuw et al. 1989). Olivary axon collaterals also project directly to the DCN (De Zeeuw et al. 1997). The inferior olive receives direct excitatory input from spinal afferents and cortical areas, and these two sources are shown to converge in the IO (Pardoe et al. 2004). The principal and accessory olives receive considerable multimodal sensory input via these pathways (Armstrong 1974; Gellman et al. 1983). Both inhibitory and excitatory inputs innervate the dendritic spines of IO neurons frequently forming a trisynaptic glomerular structure containing a dendrodendritic gap junction as well as chemical synapses (Fig. 44.1c) (De Zeeuw et al. 1989). Importantly, the inferior olive also receives a dense serotonergic input from other brain stem nuclei (Takeuchi and Sano 1983; Bishop and Ho 1986).

Excitable Properties of IO Neurons

The excitable properties of olivary neurons result from a rich interaction between a number of different voltage-gated conductances: high-threshold calcium channels, calcium-activated potassium channels (BK), and hyperpolarization-activated cation channels (Ih) in the dendrites and low-threshold calcium channels, calcium-activated potassium channels (SK), and sodium channels in the soma (Llinas and Yarom 1981a, b; Bal and McCormick 1997). The relative availability of this complex variety of channels, in conjunction with the electrical coupling between cells, means that olivary neurons can display a range of activity states including quiescence, subthreshold oscillations (STOs), and rhythmic firing. Understanding what drives these cells to transition between these activity states is essential to determining the role of climbing fiber activity in cerebellar function.

When olivary neurons are depolarized to spike threshold, a complex somatic response occurs (Fig. 44.2a), consisting of an initial fast sodium spike followed by a broad after-depolarization (ADP) on which a number of high-frequency (130–450 Hz) spikelets are superimposed. This is usually followed by a prolonged after-hyperpolarization (AHP). Both the fast sodium spike and the spikelets of the plateau phase originate in the axon as full amplitude action potentials (Mathy et al. 2009; Crill 1970) and drive a burst of multiple excitatory postsynaptic potentials (EPSPs) in their Purkinje cell targets (Maruta et al. 2007; Mathy et al. 2009; Armstrong and Rawson 1979). High-threshold dendritic calcium currents generate the ADP (Llinas and Yarom 1981a) and, while they sustain depolarization at the soma, their effect is attenuated at the axon spike initiation zone (Mathy et al. 2009) Consequently, while somatic sodium channels remain largely inactivated during this phase, axonal sodium channels remain available at sufficient density to sustain a high-frequency burst of axonal spikes, manifest at the soma as small spikelets. The ADP is curtailed by large and small calcium-activated potassium channels that underlie the AHP (Llinas and Yarom 1981a; Benardo and Foster 1986; Lang et al. 1997). The presence of low-threshold (T-type) calcium channels at the soma, which become de-inactivated by the AHP, can give rise to a rebound depolarization which, if large enough, will trigger another spike and the cycle may be repeated, generating rhythmic firing. Ih regulates the rate of rise of the rebound depolarization (Bal and McCormick 1997) and so can modulate the firing frequency.

Fig 44.2
figure 00442figure 00442

Excitable properties of olivary neurons (a) Simultaneous patch clamp recordings from the soma (whole cell) and axon (cell attached) of an olivary neuron in vitro. Depolarization by a somatic current step triggers a large amplitude action potential followed by an after-depolarization upon which are superimposed small wavelets. In the axon, these are full amplitude action potentials (from Mathy et al. 2009 copyright 2009 Cell Press). (b) Paired recording from two olivary neurons in vitro showing synchronous sinusoidal STOs of the membrane potential. (c) Example of olivary recording in which the STOs show prominent intermittent amplitude modulation. (d) Recording from three olivary neurons in vitro, demonstrating diversity in shape and frequency of STOs

Full size image

Subthreshold Oscillations

Administration of harmaline, an alkaloid drug that induces a fine motor tremor at a frequency of around 10 Hz (de Montigny and Lamarre 1973, 1974; Llinas and Volkind 1973), was found to cause rhythmic activity at the same frequency in cerebellar cortex and in inferior olive. The activity in Purkinje cells was entirely composed of complex spikes with suppression of simple spikes and, since cerebellar but not olivary rhythmic activity was curtailed by surgical separation of the cerebellar cortex from the brain stem, the olivary neurons were confirmed to be the source of this activity. The concurrent tremor and rhythmic spike output was interpreted as evidence that oscillatory activity in the IO could be important in the generation of motor programs.

It is now well established that olivary neurons can spontaneously generate endogenous STOs (see Fig. 44.2b), which can also be induced by harmaline. STOs have been observed in vitro in isolated brain slices (Benardo and Foster 1986; Llinas and Yarom 1986; Bleasel and Pettigrew 1992; Bal and McCormick 1997; Placantonakis et al. 2000; Placantonakis and Welsh 2001; Devor and Yarom 2002a, b; Leznik et al. 2002; Long et al. 2002; Leznik and Llinas 2005; Placantonakis et al. 2006; Mathy et al. 2009; Choi et al. 2010; Welsh et al. 2011) across several species (guinea pig, rat, mouse, ferret, monkey) and, more recently, in vivo in anesthetized animals (Chorev et al. 2007; Khosrovani et al. 2007; Van Der Giessen et al. 2008). They occur at 1–10 Hz, are 1–20 mV in amplitude, are often sinusoidal in character, and are found in all olivary subnuclei except the Dorsal Cap of Kooy. The STOs are heterogeneous (Fig. 44.2c, d); in vivo they can be categorized into two main types (Devor and Yarom 2002a, b; Khosrovani et al. 2007): LTOs (low-threshold oscillations in the 1–3 Hz range) and SSTOs (sinusoidal subthreshold oscillations in the 6–9 Hz range) (Fig. 44.3a). They can be transient or sustained and can show continual amplitude modulation over time. How ubiquitous STOs are to olivary neurons is not yet clear. Their occurrence varies considerably between reports with some studies describing relatively infrequent STOs (Llinas and Yarom 1986) requiring hyperpolarizing current injections, synaptic stimulation, or increased extracellular potassium concentrations to trigger them, while others show that they occur in most neurons sampled (e.g., 90% Benardo and Foster 1986, 85% in (Khosrovani et al. 2007). These differences may be due to various factors including the species employed, the details of the slice preparation and anesthesia protocols used, as well as the criteria applied for positively identifying the presence of STOs. Importantly, the cells of Dorsal Cap of Kooy do not generate any subthreshold oscillations (Urbano et al. 2006), indicating that the floccular olivocerebellar system (involved in eye movements) may operate using different principles in comparison to the rest of this network. As well as variability in occurrence, the character of the oscillations is also somewhat controversial. In some cases, highly stable STOs are observed that persist throughout a recording (Khosrovani et al. 2007), while in others, oscillations are either frequently interrupted by periods of quiescence or they show considerable amplitude and frequency modulation (Devor and Yarom 2002b; Chorev et al. 2007). What underlies the STOs, their initiation, maintenance, amplitude, and frequency is fundamental to their proposed role of rhythm generators in the olivocerebellar system.

Fig. 44.3
figure 00443figure 00443

Oscillations, coupling, and spike output in the IO. (a) Different types of subthreshold activity and spontaneous firing observed using whole cell patch clamp recording in vivo (from Khosrovani et al. 2007 copyright 2007 National Academy of Sciences, USA): (1) a neuron showing both SSTOs and LTOs; (2) a neuron showing SSTOs alone; (3) LTO activity; (4) a neuron showing no STOs of either variety. Power spectra of the four recordings. (b) Action potential synchrony during oscillations in coupled olivary neurons. Paired whole cell recordings in vitro from wild-type (WT, left) and connexin-36 knockout mice. STOs are present in both mouse strains but are only synchronous in the wild-type. Lower panel shows that, in WT, spontaneous action potentials in one cell are accompanied by either junctional potentials (left) or action potentials (right) in the coupled cell. (c) Spike cross-correlogram (5 ms bin width) of the data shown in b shows a peak at 0 ms in WT indicating spike synchrony, but no peaks in the knockout. (d) Same data shown with finer temporal resolution (0.5 ms bin width) (reproduced with permission, Long et al. 2002)

Full size image

All of the voltage- and calcium-dependent conductances involved in generating olivary spike output could potentially be involved in generating these STOs. The original observations of STOs in inferior olivary slices demonstrated that voltage-gated sodium currents do not play a role in their generation they are resistant to the sodium channel antagonist tetrodotoxin. They are, however, sensitive to calcium channel blockers or removal of calcium from the bath (Benardo and Foster 1986; Llinas and Yarom 1986). Injection of either depolarizing or hyperpolarizing current injection does not block the STOs but changes their amplitude and shape. The relationship between amplitude and membrane potential is parabolic, with a maximum close to the resting potential (Benardo and Foster 1986; Long et al. 2002; Leznik and Llinas 2005). The smaller amplitudes at more depolarized and hyperpolarized potentials are consistent with a change in availability of the high- and low-threshold voltage-gated calcium channels present in olivary neurons.

By using two lines of knockout mice, one lacking the high-threshold P-/Q-type (Cav2.1) calcium channel and the other the low-threshold T-type (Cav3.1) channel it has recently been shown that, while both of these conductances are certainly involved in STO generation, lack of the T-type channel has the most deleterious effects by greatly reducing the number of oscillating cells and strongly suppressing the amplitude of remaining oscillations which also become voltage-independent (Choi et al. 2010). Harmaline administration to the Cav3.1 knockout mice fails to induce tremor, suggesting that the mechanism underlying the action of this drug is a reduction in the threshold for T-type calcium channel activation (Park et al. 2010). As with rhythmic firing, Ih also plays a key role in STO generation (Bal and McCormick 1997) both by keeping the membrane potential within the window for T-type calcium channel activation and by providing the initial depolarizing trajectory out of the hyperpolarizing phase to reach T-type channel threshold. This generates STOs of a higher frequency than could be achieved with calcium channels and calcium-activated potassium channels alone. In simple terms, STOs are principally generated by low-threshold calcium spikes that are regulated by SK-type calcium-activated potassium channels and Ih. When SK channels are blocked in the IO, the rate of complex spike activity in cerebellum increases, but it loses rhythmicity (Lang et al. 1997).

Gap Junctions and Synchrony

In addition to their reliance on intrinsic voltage-gated conductances, there is strong evidence that STOs are also a network phenomenon driven via the electrotonic coupling between IO cells. STOs are tightly synchronized among groups of coupled IO neurons (Benardo and Foster 1986; Devor and Yarom 2002a, b; Long et al. 2002) indicating that the intrinsically generated currents can also flow within the network. During development, oscillations emerge at the same stage as gap junctions, despite the earlier presence of the voltage-gated channel conductances that drive them (Bleasel and Pettigrew 1992). This leads to the question: Are STOs the result of summed network activity or are individual olivary neurons capable of generating STOs in isolation? Although depolarization and hyperpolarization reduce the amplitude and modulate the frequency of STOs, they cannot be blocked completely by changes in the membrane potential (Benardo and Foster 1986) which would be expected to suppress the voltage-gated channels involved in their generation. A gap junctional blocker, 18β-glycyrrhetinic acid (18β-GA), which substantially reduces dye-coupling between olivary neurons, enhances the voltage-dependent changes in amplitude and frequency of STOs (Leznik and Llinas 2005), leading to their absence at more positive and negative membrane potentials. In connexin 36 knockout mice, the oscillations are still present, displaying a similar voltage dependence to those in 18β-GA, but are not synchronized between cells (Long et al. 2002) (Fig. 44.3b–d). Possible compensatory mechanisms observed in these knockout mice, including changes in both morphology and excitability, makes the interpretation of the effects of removing connexin 36 problematic (De Zeeuw et al. 2003). However, the use of lentiviral-mediated knockdown of connexin 36 in the IO has confirmed that, in order to support STOs, olivary neurons do need to be coupled to a small ensemble of other IO cells (Placantonakis et al. 2006). It seems, therefore that, while olivary neurons are intrinsically oscillatory, sustained large amplitude oscillations may only occur when the cells are part of a coupled neuronal ensemble supporting intercellular current flow (Lampl and Yarom 1997; Manor et al. 2000).

The synchronization of oscillations in the IO results in temporal coherence of complex spike activity across populations of Purkinje cells beyond the microzonal organization of a single olivary projection (Lang et al. 1999; Schultz et al. 2009; Bell and Kawasaki 1972; Ozden et al. 2009). Voltage-sensitive dye imaging in IO slices has been used to characterize the spatial profile of STO synchrony across the olivary network with some mixed results. In one case (Leznik et al. 2002), after a brief electrical stimulus, the IO showed synchronous STO clusters, estimated to contain many hundreds of neurons, which were reliant on gap junction connectivity (Leznik and Llinas 2005). In another case (Devor and Yarom 2002b), patches of spontaneous or stimulus-evoked STOs were observed within which the oscillations were not spatially stationary but instead propagated rapidly in a wave-like manner throughout the local network resulting in the establishment of phase differences between cells. In agreement with this, in vivo multi-electrode array recordings of Purkinje cell complex spike activity in awake rats (Jacobson et al. 2009) suggest that, rather than being exactly in phase, stable phase differences can exist between neurons which oscillate at the same frequency. A networked compartmental model of olivary neurons (Schweighofer et al. 1999) demonstrated that these phase differences can be achieved by variations in coupling strength between the oscillating neurons.

It is still unresolved how large an oscillating cluster of cells is. A dye-coupling study suggests olive cells are directly coupled to a varying small number of cells (up to 35) contained within their dendritic field (Hoge et al. 2010). The voltage-sensitive dye studies, however, conclude that a synchronous cluster extends beyond this small neighborhood, and can contain between 100 and 2,000 cells (Leznik and Llinas 2005; Devor and Yarom 2002b). It is also known that olivary dendrites can sometimes extend across the midline to make contacts with the contralateral olive (De Zeeuw et al. 1996), which is reflected in the presence of bilateral climbing fiber synchrony in the cerebellum.

Oscillations and Olivary Output

What is the impact of STOs on olivary spike output? There is much support for the idea that they precisely and periodically time output (the “clock” hypothesis), with spikes occurring solely at the peak of any cycle, but this is controversial. While olivary neurons have been shown in vivo to preferentially fire at the crest of an oscillation (Chorev et al. 2007), spikes can also occur across a broader range of the oscillatory cycle (Khosrovani et al. 2007). These differences may be due, at least in part, to the variation in STO amplitude, with larger and faster rising STOs leading to peak-locked spikes and smaller, slower STOs showing more variation in spike timing. Little is known about the synaptic weight of excitatory inputs to olivary neurons, but synaptic diversity will also likely play a role in timing of output relative to STO phase. It is by now certainly clear that IO neurons do not function as simple, autonomous pacemaker cells. Block of glutamatergic input into the IO showed that in vivo firing rates drop by 50% (Lang 2001), indicating that spike output is both extrinsically and intrinsically generated. It has been argued, based on observations in vitro, that afferent-evoked responses are delayed until the peak of an oscillation (Lampl and Yarom 1993) but the mechanism for this is poorly understood and this is not consistently observed (Mathy et al. 2009). This idea is also contradicted by data showing that, in vivo, olivary neurons respond with short latencies to somatosensory stimuli (Gellman et al. 1985). Recent findings suggest that, rather than restricting spike output to occur only at the peak of the oscillation, the cycle phase at which excitation occurs can determine the number of axonal spikes initiated during the complex burst response of an olivary neuron (Fig. 44.4a). This could provide a means by which the cerebellar cortex could extract information about the cycle origin of climbing fiber input and, as a consequence, attribute new learning rules that incorporate this information (Mathy et al. 2009).

Fig. 44.4
figure 00444figure 00444

(a) Dependence of number of output spikes on STO phase. Top panel shows overlay of five somatic responses of an olivary neuron in vitro to synaptic stimulation during STOs generated by sinusoidal (5 Hz) current injection. EPSPs were triggered by single synaptic stimuli (arrow heads), timed to occur at different phases of the STO. Lower panel: same somatic responses on an expanded timescale and, below, simultaneously recorded axonal spikes from the same cell (from Mathy et al. 2009 copyright 2009 Cell Press). (b) In vivo phase response curves generated from neurons showing SSTOs with spontaneous (open circles) or stimulated (peripheral stimulation, closed circles) action potentials. The phase of the oscillation was reset when the spikes did not occur at the peak of the STOs. Inset shows recorded membrane potential (black) superimposed with a sinusoidal fit (red) to the data prior to the occurrence of the action potential (from Khosrovani et al. 2007 as above)

Full size image

Extrinsic Control of Oscillations

If, as many suspect, subthreshold oscillations are the substrate for generating motor rhythms, it is important to know what can modify and shape them. The variability of the oscillations within and between cells suggests that modulation occurs physiologically. The plausible loci for affecting the oscillations are the ion channels involved in rhythmogenesis, the afferent synaptic pathways into the nucleus, and the electrical coupling between the neurons.

It is discussed above how pharmacological and genetic approaches have dissected out the ion channels responsible for generating STOs. Whether modulation of these channels is engaged physiologically for shaping oscillations is as yet unknown, but plausible. For instance, it is known that the Ih current – important in the generation of oscillations – is extremely plastic in other neurons (van Welie et al. 2004). Recently, it was shown in monkeys that chronic alcohol abuse followed by abstinence can modulate both Ih and the low-threshold calcium current in the olive, and this might underlie the intention tremor that alcoholics can experience upon withdrawal (Welsh et al. 2011).

The IO receives a significant serotonergic input from the nucleus reticularis paragigantocellularis (Bishop and Ho 1984, 1986). In vitro application of serotonin suppresses oscillations (Placantonakis et al. 2000), thought to be via potentiation of Ih and suppression of the T-type calcium current. Interestingly, serotonin also strongly suppresses excitatory synaptic inputs in the IO (Best and Regehr 2008) through a retrograde, endocannabinoid mediated pathway. In vivo recordings have provided conflicting data as to whether serotonin decreases (Headley et al. 1976) or increases (Sugihara et al. 1995) rhythmicity in the olive.

Interestingly, activation of glutamate receptors has also been shown to interfere with oscillations in several different ways. Glutamatergic input will activate both AMPA and NMDA receptors in IO cells (Best and Regehr 2008). In vitro application of NMDA receptor antagonists has been shown to block STOs, while application of NMDA causes oscillations sensitive to L- and P-type calcium channel antagonists (Placantonakis and Welsh 2001) but not to TTX. This seems to implicate excitatory input in the generation of STOs via a calcium current. Blocking AMPA receptors with CNQX, on the other hand, stabilizes the oscillations (Devor and Yarom 2002b), which could plausibly be due to an increased input resistance and a reduced shunting of the intercellular currents through gap junctions (Fig. 44.5a). When a large excitatory input triggers a spike, the subthreshold oscillation resets to a given phase (Khosrovani et al. 2007; Llinas 2009). In other words, depending at which oscillatory phase the spike occurs, the oscillation is advanced so that after the spike the oscillation is always at the same phase (Fig. 44.4b). This is a network phenomenon affecting many cells simultaneously, and has been hypothesized to represent the retrieval of a new motor program in the olivocerebellar system.

Fig. 44.5
figure 00445figure 00445

Control of STOs by glutamate and GABA. (a) Bath application of CNQX stabilizes the frequency and amplitude of STOs in vitro. Two example patch clamp recordings in separate olivary slices (left, right) in the absence (top) and presence (bottom) of CNQX (40 μM). CNQX blocks the frequency modulation in both cases but does not prevent intermittence of oscillations (seen on the right). (b) GABA suppression of STOs: (1) Schematic illustration of the proposed shunting effect of GABA conductance on coupling between olivary neurons. Activation of GABAA receptors on gap junction (GJ) connected dendritic spines results in shunting of current flow via the open chloride-permeable channels, resulting in less current flow between the coupled cells (white arrows). (2) Whole cell recording from an olivary neuron shows that application of GABA can suppress STOs (reproduced with permission, Jacobson et al. 2008)

Full size image

The inhibitory pathway from the deep cerebellar nuclei to the IO is thought to play a key role in gating responses to excitatory input (Kim et al. 1998) and may do this to some extent by influencing the STOs (Fig. 44.5b). The peculiar synaptic organization within the olive, with GABAergic synapses placed adjacent to gap junctions, led to the suggestion in the 1970s that a role of inhibitory input from the DCN might be to uncouple the olivary neurons by shunting intercellular currents (Llinas et al. 1974), thereby reducing the size of oscillating neuronal assemblies and dampening the amplitude of the oscillations themselves (Jacobson et al. 2008). Subsequently it has been shown that, in vitro, blocking GABAA receptors creates more synchronous clusters of olive cells as visualized by voltage-sensitive dyes (Leznik et al. 2002), while local GABA application abolishes the STOs (Devor and Yarom 2000). It is not yet known whether physiological inhibitory input is sufficient either in time course or distribution to mediate these effects. If this is the case, then other physiological modifiers of gap junction coupling might be expected to influence both the STOs and their synchrony across the olivary network. Mammalian gap junctions are known to be modulated by intracellular pH, calcium and membrane potential, among other things (Connors and Long 2004). Intriguingly, thalamic neurons have recently been shown to undergo activity-dependent depression of intercellular coupling (Landisman and Connors 2005). Whether these phenomena are important within the IO requires further study.

Rhythmicity and Synchrony in the Climbing Fiber System

Determining how inferior olive spike output is timed relative to STOs is central to understanding the role played by the climbing fiber signal in cerebellar function. While the evidence for STOs in olivary neurons is compelling, the relationship between these and complex spike timing in the cerebellar cortex is not clear-cut. In the absence of direct intracellular recordings from the IO, the simplest predictor of olivary STOs would be the presence of rhythmic or periodic complex spike activity in cerebellar cortex but the prevalence of complex spike rhythmicity cannot be agreed on (Kitazawa and Wolpert 2005). Periodicity in complex spike activity has been demonstrated in awake animals during behavior and rest (Llinas and Sasaki 1989; Welsh et al. 1995; Lang et al. 1999), but this periodicity varies considerably across Purkinje cells, and is strongest when the behavior itself is rhythmic. In rats, olivary neurons fire rhythmically during the step cycle (Smith 1998), but rhythmic complex spikes are not seen during locomotion in cats (Armstrong et al. 1988). Another study in cat (Bloedel and Ebner 1984) found that, despite there being no rhythmicity in complex spikes in response to forepaw displacement, autocorrelograms of the PSTH over many trials show two to four peaks, indicating that there are rhythmic increases in excitability of IO neurons after a sensory input. This could be explained by the phase-resetting phenomenon discussed above. Importantly, recordings in awake monkeys and cats (Armstrong and Rawson 1979; Keating and Thach 1995, 1997) show no periodicity in complex spike activity in single Purkinje cells. A modeling study (Schweighofer et al. 2004) suggests that the firing of the olive is not periodic, but chaotic; so that the precise timing of the spikes can maximize the information transmitted through the olive despite occurring at low frequencies. It should be borne in mind that, while observed CS rhythmicity may indeed reflect the presence of STOs in the IO, it might also reflect rhythmic firing generated by another rhythmic input. Likewise, a lack of CS rhythmicity does not necessarily indicate the absence of STOs. Intracellular recordings from IO neurons are challenging and those that have been made, while showing clear STOs, have been carried out in anesthetized preparations. If oscillations are suppressed during action, testing the relationship between STOs and IO spike output while an animal performs a simple behavior may provide the missing piece of the puzzle.

Given the efficacy of afferent excitation in triggering olivary output, other brain areas may drive the olivocerebellar system to the beat of their own drum. It has been suggested, for example, that the IO acts as a resonating circuit, preferentially transmitting input within a certain bandwidth from the motor cortex to the cerebellum, and that this gating shapes the frequencies at which the cortex is able to generate motor output (Marshall and Lang 2004; Lang et al. 2006b). The peak of the frequency response curve resides close to 10 Hz, which is consistent with common frequencies of olivary STOs.

Physiologically relevant periodicity in olivary output may be disputed, but it has consistently been shown that groups of olivary neurons fire together: in the cerebellar cortex, Purkinje cells within a 500 μm sagittal strip discharge complex spikes in synchrony (Bell and Kawasaki 1972; Lang et al. 1999; Ozden et al. 2009; Schultz et al. 2009). A recent study suggests that even asynchronous climbing fibers fire at fixed time intervals with respect to each other (Jacobson et al. 2009) which is proposed to reflect fixed phase offsets between olivary neurons oscillating at the same frequency. Synchrony results from a combination of the electrotonic coupling between IO cells (Blenkinsop and Lang 2006), the subthreshold oscillations themselves (Lang et al. 1997), the branching of a single olivary axon into multiple climbing fibers (Lang et al. 2006a), and shared excitatory synaptic input between neighboring olivary neurons (Kistler et al. 2002). The fact that rodents and humans with impaired gap junction function show motor learning deficits is an indication that synchronous olivary activity is important for normal cerebellar function (Van Der Giessen et al. 2008; van Essen et al. 2010). This is supported by the finding that, in rats with impaired electrotonic coupling in inferior olive, there is a reduction in coherence across muscle groups during harmaline-induced tremor (Placantonakis et al. 2004).

An important feature of complex spike synchrony is that the populations involved are not fixed, but can change during behavior (Welsh et al. 1995). How this occurs is still open to question but synaptic input is a key candidate. Blocking either excitatory or inhibitory receptors in the inferior olive increases the synchronicity with which the cells fire (Lang 2002). This has been posited to be due to shunting of intercellular current by activated synapses located within the trisynaptic junctions. Furthermore, since the inhibitory afferents from the DCN is one arm of the olivocerebellar loop, a patch of IO could be regulating its own pattern of synchrony dynamically (Marshall and Lang 2009). This had led several groups to theorize that the olivocerebellar system is capable of dynamically generating different spatiotemporal patterns of climbing fiber activation relevant for the coordination of synergistic muscle activity (Jacobson et al. 2008; Llinas 2009). The rationale for taking this olive-centric view of the system is that the climbing fiber system has a much more dramatic effect on the output of the cerebellum (via the CF pause) (Davie et al. 2008), the modulation of PC bistability (Loewenstein et al. 2005), and rebound firing in the DCN (Hoebeek et al. 2010) than does the PF pathway, which can only weakly modulate the intrinsic firing of Purkinje cells.

Conclusions

This chapter reviews the remarkable properties of the inferior olive and its role in the motor system, showing that subthreshold oscillations present at frequencies of 1–10 Hz are intrinsic and reinforced by gap junction coupling, and that these oscillations can shape the climbing fiber output to the cerebellum by imposing rhythmicity and synchrony in sagittal bands in the cortex. Many other brain circuits employ oscillations (Buzsaki and Draguhn 2004), but the independence of extrinsic input is a rare feature of this system. Is the olive a clock? While this may be a good metaphor for, say, pacemaker cells in the sinoatrial node of the heart, it is likely too simplistic a description of the olive. While there is no doubt that, via its oscillatory activity, the olive can generate an intrinsic timing signal, it is also robustly activated by excitatory afferents responding vigorously and with short latency to minute peripheral stimuli (Gellman et al. 1985), especially when such stimuli are not expected by the animal. The effective silencing of the IO during expected sensory input suggests that signaling unanticipated events is a key function of this structure (Devor 2002).

Early theories of the cerebellum suggested that the main role of the climbing fiber input is to provide a teaching signal to adjust the weights of parallel fiber synapses in order to modulate Purkinje cell simple spike output (Marr 1969). Modern variants see the cerebellum as an adaptive filter for tuning motor output (Dean et al. 2010). Some of the best evidence for these theories comes from the floccular system for adjusting eye movements (Ito 1970), however, and as discussed above, the relevant climbing fibers originate in the Dorsal Cap of Kooy, an area of the olive which does not generate subthreshold oscillations (Urbano et al. 2006). This may therefore not represent a universal model for the olivocerebellar system – if indeed such a model can exist. It has been argued elsewhere (Mathy et al. 2009) that it is important that future cerebellar learning theories take into account the dynamic properties of the IO, since it is clear that precise timing must play a role in complex motor output. There is much evidence, however, to support the view that the olivocerebellar system behaves as a resonant circuit, which can amplify input signals within a relevant frequency range (Lampl and Yarom 1997; Lang et al. 2006a; Khosrovani et al. 2007). Whether, as has been suggested (De Zeeuw et al. 2011), it can also stably and reproducibly generate spatiotemporal patterns of activity in the cerebellum remains to be shown conclusively. Experimental interventions are needed that can precisely manipulate the output of the olivocerebellar system and quantify the effect on, for example, motor behavior. Furthermore, this hypothesis requires an elaborate learning process distributed across the network and a mechanism to select the correct activity pattern for the intended behavior – two aspects which require further theoretical refinement and experimental confirmation. More pressingly, we need to know how subthreshold oscillations relate to activity in the awake animal. Only then will neuroscientists be able to make sense of this elegant, yet perplexing, neuronal machine.