Subcortical Contributions to Flexible Spatial Navigation

Abstract

Spatial navigation embodies many cognitive processes. Specifically, relevant sensory information needs to be extracted from the environment, and representations and decisions need to be formed and oftentimes maintained in order to execute appropriate navigational behavior. This process is additionally modulated by the internal state and past experiences of the agent, and the sensorimotor transformations must be flexibly adapted to changing reward contingencies or different behavioral needs of the agent. While cortical areas are thought to drive these adaptive processes, subcortical structures could also contribute meaningfully to flexible behavior and their involvement remain poorly understood. In this work, we assess two types of flexible navigation behavior: 1) the ability to reversibly adapt cue-action mappings in an environment with implicit rule-switches, and 2) the update of neural representations despite unchanged external environment and behavioral output. In the first project, we investigated how higher-order thalamus supports flexible rule-switching during navigation. The thalamus is known primarily as a relay structure owing to studies on first-order thalamic nuclei, but recent advances have begun to show that higher-order thalamic nuclei are capable of more complex processing than passively relaying information. In this study, we focus on mediodorsal (MD) and lateral posterior (LP) thalamus, which form the strongest bidirectional connections with the cortical areas important for spatial navigation and flexible behavior. We trained mice to perform a rule-switching navigation task in virtual reality, in which the rewarded cue-action mapping switched multiple times within a session without explicit signals. Using Neuropixels 1.0, we recorded extracellular activity during behavior and found that LP has strong encoding of visual cues and choice in the maze, while MD shows the strongest outcome encoding during the inter-trial-interval (ITI). Using optogenetic inhibition, we found that LP is causally involved in the sensory-motor interface to associate visual cues with behavioral choices, whereas MD is important for evaluating the correctness of actions to enable rule relearning. These results highlight that MD and LP contribute to flexible navigation, providing complementary functionality to cortical areas. In the second project, we examined how internal states modulate hippocampal spatial codes. The hippocampus is thought to automatically form and maintain place codes by combining sensory and self-motion signals. In this study, we took advantage of voluntary changes in a mouse’s engagement in a goal-directed navigation task in virtual reality and recorded concurrent hippocampal CA1 activity using two-photon calcium imaging. Strikingly, we observed an extensive degradation of hippocampal place code when mice voluntarily disengaged from a virtual navigation task, even as they continued to traverse the identical environment. Our results reveal that internal states can strongly gate spatial maps and reorganize hippocampal activity, and sensory and self-motion information alone are insufficient for forming a reliable spatial map. Across these two projects, a common principle emerges: subcortical structures do not provide fixed or passive contributions to navigation but instead carry signals that are dynamically shaped by the animal's behavioral needs and internal states. Hippocampal spatial representations are gated by behavioral engagement rather than automatically tracking position, and thalamic coding of sensorimotor associations and outcome are causally involved during distinct phases of rule-guided behavior. Together, these findings argue that flexible navigation arises from the coordinated action of cortical and subcortical circuits, and that understanding this interaction is essential for explaining different aspects of flexible behavior.