While exploring the functions of individual cells in a circuit, we can often overlook the importance of specialized subcellular compartments organized within one neuron. We probe the former with techniques like optogenetics, optical imaging, and (multi) cell recordings, while the latter can be clouded with mystery that have only begun to resolve in recent years. We explore these seemingly remote and difficult-to-record-from places in a well-mapped circuit in the cerebellum that is also relevant to behavior. Granule cells of the cerebellar cortex carry information associated with the context of a motor movement to the Purkinje cells, which as the sole output of the cerebellar cortex, is a major site of motor learning. Traditionally thought to be inhibitory, GABAA receptors (GABAARs) in granule cells, although inhibitory in the soma, have been found to be excitatory in the axons. These axons, the parallel fibers, are too thin to record from, and methods to study them include population recordings like calcium imaging and fiber volley detection, computational modeling, and voltage sensitive dyes.

We first studied the presynaptic terminals of parallel fibers using calcium imaging, and showed that GABAAR activation leads to increased amplitude of stimulus-evoked calcium transients. This excitation appears to bring axons closer to firing threshold, recruiting addition fibers as opposed to increasing calcium levels on a per fiber basis. Blocking the transporter that accumulates chloride reduces the effect, suggesting that high chloride concentration in the axons is the mechanism behind this effect.

To gain better temporal resolution, we used GABA uncaging to probe extracellular fiber volleys and found that GABA not only increases fiber volley amplitudes, but also increases conduction velocity on parallel fibers. We showed that δ-subunits are not required for GABAAR-mediated excitation and that endogenous GABA in a slice is sufficient to excite parallel fibers. Using a computational model of the granule cell, we showed how GABA-mediated excitation on the axons can influence cell spiking, and that properties of sodium channel inactivation determine whether there's an excitatory effect of GABA on threshold for spiking and on conduction velocity.

Finally we described a novel method for detecting voltage in subcellular compartments utilizing fluorescence transfer between a lipophilic tracer dye and a voltage sensitive compound. This two-component system is seen to produce some of the biggest voltage sensing signals in the literature, and is capable of recording submillisecond voltage fluctuations in subcompartments of a Purkinje cell. We propose to use this system to record voltage from parallel fibers and determine the effect of GABA on action potential shape and subthreshold voltage fluctuations.

The cerebellum has long been hypothesized to be involved in the storage and expression of motor memories. Work presented in this thesis demonstrates that optogenetically conditioned motor memories are stored in the cortical and nuclear regions of the cerebellum and expressed via disinhibition-mediated bursting of the cerebellar nuclei. Building on this observation we sought to determine the impact of preventing disinhibition of the cerebellar nuclei when cueing an optogenetically conditioned forelimb movement. By transiently increasing inhibition of cerebellar nuclei during high-speed video recordings of mice responding to a previously conditioned auditory cue, we demonstrated that preventing disinhibition of cerebellar nuclei prevents the expression of the learned forelimb movements. We conclude that disinhibition-mediated bursting of the cerebellar nuclei is necessary for retrieving the motor memory and the subsequent forelimb movement. Furthermore, this mechanism could be relevant for the expression of other kinds of motor memories.

The cerebellum plays a fundamental role in motor movement coordination and motor error correction. By using (and associating) sensory cues present during motor errors, the cerebellum learns to anticipate motor error mistakes and execute correct movements.

Damage to the cerebellum, either by trauma or disease, can impair or eliminate this ability, resulting in poor coordination and an inability to adapt or correct motor movements. In order to assist these patients, a detailed understanding of the mechanisms involved during error correction and adaptation learning is needed.

Many groups have worked to understand how this learning occurs. One identified area that requires more understanding is at the output of the cerebellum, the synapses made by Purkinje neurons on Deep Cerebellar Nucleus neurons.

The climbing fiber input to the cerebellum is believed to serve as a teaching signal during associative, cerebellum-dependent forms of motor learning. In Purkinje neurons the climbing fiber generates a complex spike, and a post-complex spike pause, that interrupts Purkinje neuron pacemaker firing. Because Purkinje neurons form

inhibitory (GABAergic) synapses onto DCN neurons, coordinated pauses in Purkinje neuron firing lead to transient periods of dis-inhibition at the DCN, allowing them to be active. We hypothesize that windows of dis-inhibition relay information about the presence of a motor error to the DCN, since the DCN neurons do not actually see the climbing fiber-evoked complex spike that the Purkinje neurons see.

In addition to facilitating DCN neurons to activate their downstream targets (including the Red Nucleus and other motor targets), we propose that this relief from inhibition allows for learning-related plasticity mechanisms, like mossy fiber long term potentiation (LTP), to occur at the DCN. It is not entirely understood, however, how the

climbing fiber signal is coordinating changes in cerebellar circuitry during learning.

The post-complex spike pause is the final component of the climbing fiber signal transmitted by Purkinje neurons to the DCN following a motor error. The pause itself may therefore play an important role in the motor learning process.

In in vitro experiments described in Chapter 3, the post-complex spike pause is reliably prolonged by two different drugs acting by two distinct, and opposing, mechanisms.

The same drugs delivered in vivo, during classical eye-blink conditioning experiments described in Chapter 4, facilitated the onset of learning in this paradigm: rats receiving the drugs learned the behavioral association faster than their control

counterparts. These results elucidate an unappreciated aspect of the climbing fiber teaching signal, the post-complex spike pause, and support a model in which synchrony of postcomplex spike pauses drives learning-related plasticity in the deep cerebellar nucleus.