The midbrain reticular formation in flexible visual decision-making

Abstract

A hallmark of mammalian behavior is the ability to rapidly remap actions in response to sensory stimuli depending on internal representations of the environment. Flexible visual decisions are thought to be computed within recurrent interactions across diverse brain circuitry in the frontal cortex, basal ganglia, and midbrain. However, the precise regions involved and their computations remain unclear. Previous work has found that a large and poorly understood structure of the midbrain, the midbrain reticular formation (MRF), contains similar task activity patterns to well-established decision regions and connects extensively with them. Open questions include whether the task dynamics in MRF reflect motor processing versus abstract task rules as well as how structure and function organize within the MRF. In this dissertation, we dissect the organization of MRF and demonstrate a role for MRF in abstract context computation using a novel flexible decision-making task, large-scale electrophysiological recordings, modeling of task behavior and neural population dynamics, and single neuron morphological reconstructions. In Chapter 1, we first review the approaches to studying flexible visual decision-making in the lab and the current understanding of the regions and computations involved. We then discuss what is currently known about MRF’s functional and anatomical characteristics and examine historical perspectives on its role in contextual processing. Next, we introduce our novel flexible decision-making task and show that mice achieved high levels of performance by integrating an abstract context belief variable with visual stimuli to remap stimulus-action associations (Chapter 2). We performed dense electrophysiological recordings within MRF and established nodes of the flexible-decision making circuitry and find that MRF, in a network with the superior colliculus (SC), secondary motor cortex (MOs), and caudoputamen (CP), maintained a baseline representation of context which putatively enabled flexible remapping by shifting action attractor dynamics between contexts (Chapter 3). In Chapter 4, we record from both trained and task-naive mice while passively presenting stimuli and find that task-specific visual stimulus representations appear throughout MRF as a result of task learning. Then, we examine the spatial distribution of context-coding neurons and cortical inputs across MRF, revealing that context-coding neurons and cortical axon terminals both contain non-uniform distribution patterns and are spatially aligned (Chapter 5). Altogether, our results establish MRF as a key node in the circuitry underlying flexible visual decisions and provide fundamental insights into how the brain processes contextual information to flexibly update responses to environmental stimuli. In Chapter 6, we integrate prior literature with the findings in this dissertation to update current theories of MRF and discuss future work to build upon and test these theories.