P1: Sources and Mechanisms of Sequential Activity

点击翻译按钮获取中文摘要

Project Summary/Abstract: Project 1, Sources and Mechanisms of Sequential Activity Sequential activity is widespread and predominant across the mouse brain during an evidence-accumulation decision task and in other tasks as well. Such activity may form a “temporal scaffold,” on top of which other variables are encoded in the amplitude of these sequentially active responses. This activity, different from the ramps and persistent activity often studied in perceptual decisions, could be driven by navigation. The first aim will be to test this idea by identifying conditions that produce sequential representations, such as the task’s timing structure, navigation through spatial locations, or visual stimulation. To distinguish these possibilities, we will record neural activity from regions containing sequential activity during tasks that isolate these key features: active navigation, passive navigation, and visual stimulation. This work will establish whether removal of task features eliminates sequential activity, producing ramps or persistence. The second aim will be to use focal cooling to test a potential role of the striatum as a master temporal scaffold. Medium spiny neurons of dorsal medial striatum (DMS) show sequential activity. Inactivation of this region in the cue period has large effects on choice, yet few DMS sequences are choice-specific in this period. We propose instead that DMS generates a temporal scaffold that controls the timing of choice and evidence-encoding sequences in neocortex and hippocampus. To test this hypothesis, we will use focal cooling to slow striatal neural dynamics, while recording in neocortex and hippocampus. These results will constrain models of sequence generation and reveal the mechanistic foundations of sequential neural activity. The third aim will be to identify network architectures that could underlie the observed data, by building models with three architectures that generate choice-selective sequences. In the moving bump attractor model, activity location in a population jointly encodes position and evidence. In the competing-chains model, evidence is encoded in amplitude, while position is encoded in activity location, in two competing sequences. In the position-evidence multiplicative model, evidence is accumulated in classic ramping activity that controls the gain of activity that is sequentially activated with position. These models will make testable experimental predictions to help us distinguish these network architectures. The fourth aim will be to compare ultrastructural anatomical connectivity with neural coding. We will use cellular-resolution imaging of neural activity during behavior, followed by serial-section electron microscopy of the same neurons in the dorsal hippocampus and neocortex, to empirically test network models of sequential activity. Together, the results of this project will identify task features, brain regions, neural architectures, and microscale anatomy underlying the appearance, timing, and function of sequential activity. We expect that the experiments and models in this project will substantially advance three priority areas of the BRAIN Initiative: the brain in action, demonstrating causality, and identifying fundamental principles.