Cellular and synaptic mechanisms controlling cerebellar output

Abstract

The deep cerebellar nuclei (DCN) provide the major output of the cerebellum, projecting to premotor nuclei as well as non-motor areas in the brain. The dominant input to the DCN is provided by the output cells of cerebellar cortex, Purkinje cells (PCs). The synapse between PCs and DCN cells is GABAergic and yet, both PCs and DCN cells basally fire at tens of hertz in vivo. In this unusual situation, the output of the cerebellum is generated. We aimed to clarify the cellular and synaptic mechanisms that influence how DCN cells encode their PC inputs. At the synaptic level, we found that loss of the high-affinity calcium-binding protein, Doc2b, does not affect synaptic transmission or plasticity at PC to DCN synapses. PC to DCN plasticity is distinctive in that a balance between presynaptic facilitation and depression results in constant steady-state synaptic strength across afferent firing frequencies. To understand how this frequency-invariant mode of transmission would respond to a change in neurotransmitter release properties, we tested how synaptic modulation affects frequency-invariance. Although modulation occurs in young animals, we found that PC to DCN synapses are strikingly unresponsive to modulation in mature animals when frequency-invariant plasticity has been developed. We also studied how the firing activity of PC inputs influence DCN cells. In awake mice, climbing fiber activation of one PC reduced simple spike activity in neighboring PCs for several milliseconds through an ephaptic mechanism. Dynamic clamp and in vivo recordings indicated that DCN cells sensitively encode such brief pauses in PC activity with a prominent increase in firing. To understand how the DCN responds to more prolonged changes in input activity, we performed in vivo and in vitro studies that revealed a surprisingly slow form of spike frequency adaptation lasting up to over 20 seconds in DCN cells. This adaptation mechanism is largely calcium-independent and likely involves a sodium-activated potassium conductance mediated by SLICK/SLACK channels. Altogether, these findings help explain how the DCN can flexibly encode rapid changes in PC activity while also maintaining long-term stability at the synaptic and intrinsic firing levels.