Synaptic dynamics as a tunable substrate shaping neuronal activity sequences

Across brain regions and behaviors, neural population activity unfolds as temporally structured sequences that underlie perception, memory, and precisely timed actions1-10. However, how neural circuits transform continuous information streams into transient patterns of activity over time remains poorly understood. A long-standing hypothesis for cerebellar learning posits that the granule cell (GC) layer segments sensory and motor information arriving via mossy fibers (MFs) into temporal basis sets that enable precisely timed motor and cognitive commands11-15. Measurements of such basis sets have been elusive. Using high-speed multiphoton calcium imaging of MF and GC responses to whisker air puff stimulation, we show that prolonged MF activity is transformed into temporally sharpened GC responses that form a sparse population sequence tiling the sensory event in time. Temporal sparsity of GC sequences varied between cerebellar regions. By combining in vivo glutamate imaging with ex vivo synaptic recordings, we identify heterogeneous MF-GC synaptic strength and short-term plasticity as the mechanisms underlying region-specific temporal sparsification. Mathematical modeling predicted that region-specific MF-GC synaptic dynamics generate temporally sparse GC sequences with distinct statistics specifically suited for learning across different timescales. Thus, heterogeneous synaptic dynamics provide a biological substrate for shaping population activity in time, setting the temporal precision of sensorimotor associations underlying adaptive behavior. One-sentence summaryDiverse short-term synaptic dynamics transform input activity patterns into temporally sparse neural sequences in the cerebellar cortex, providing a mechanistic basis for precise temporal learning.