High-dimensional bidirectional population coding in pontine mossy fiber inputs to the cerebellar cortex.
TL;DR
The results establish the population-level properties of the pontine input to the cerebellar cortex and show that their neural representations provide a surprisingly high-dimensional substrate for cerebellar expansion recoding.
Abstract
The cerebellum gathers sensorimotor information from the neocortex via the pontine nucleus, mixing and expanding it to form associations for predicting movements and their sensory consequences. However, the functional properties of this major pontine input to the cerebellar cortex remain largely unexplored. Using 3D two-photon calcium imaging, we show that pontine mossy fibers utilize a high-dimensional bidirectional population code. However, only a fraction of the neural activity dimensions were required to encode specific locomotion and whisking-related variables. Curiously, axons with opposing response characteristics were intermingled within the granule cell layer. Experimentally constrained models of the circuit show that innervation of granule cells with different combinations of opposite-signed inputs enables effective propagation of high-dimensional bidirectional population codes to downstream circuits. Our results establish the population-level properties of the pontine input to the cerebellar cortex and show that their neural representations provide a surprisingly high-dimensional substrate for cerebellar expansion recoding.