CRCNS: Multiple Clocks for the Encoding of Time in Corticostriatal Circuits

The ability to predict when external events will occur, such as anticipating the actions of a predator or the availability of food, is critical for survival. Converging computational and experimental work suggests that dynamically changing patterns of neural activity, including neural sequences, underlie temporal prediction and temporal processing. It is increasingly clear that timing and temporal prediction are highly distributed computations, however, there has been little effort to systematically contrast and understand the computational tradeoffs between how time is encoded in different brain areas. Furthermore, while converging evidence suggests neural sequences in the striatum play a central role in timing, the mechanisms underlying the generation of neural sequences remains elusive. Critically, it is not known whether neural sequences are actively generated within the striatum or are “driven” by neural sequences present in corticostriatal inputs. We propose to address these major gaps in understanding with a combination of innovative experimental and computational approaches. Our key hypotheses are that: 1) neural sequences in the striatum provide a flexible dynamical regime that allows for temporal scaling, i.e., speeding-up or slowing-down of motor responses, 2) cortical input shapes neural sequence formation in the striatum, 3) local inhibitory circuits serve to refine the quality of these sequences in the striatum, and 4) neural dynamics encoding time are widely distributed throughout the brain but are more accurate in certain areas such as the striatum. Our project is anchored in a two-interval timing task in which mice learn to associate two cues with different reward delays, and has three major aims. Guided by large-scale neural recordings in multiple brain areas we will first develop cortical and striatal recurrent neural network models with the goal of understanding which circuit motifs are best suited to generate neural sequences, and determining which models best capture the experimentally observed activity patterns. Second, we will integrate neural recordings and optogenetic perturbations, together with computational approaches, to determine whether neural sequences in the striatum are driven by cortical input and refined by local inhibition, or in contrast actively generated within the striatum. Third, we will carry out a high-throughput electrophysiological survey of neural activity in multiple brain areas, to identify which areas contain the most accurate temporal codes as well as the potential computational tradeoffs between different codes. RELEVANCE (See instructions): By integrating advanced computational and experimental approaches, this collaborative project will provide fundamentally new insights about how the mammalian brain is able to predict when external events will occur, enabling animals to produce appropriately timed movements that are critical in daily life. This work will reveal which brain circuits are most strongly implicated in timing, which is often impaired in neurological disorders such as Parkinson’s and Huntington’s disease.