Summary, Overall (Thalamus in the middle: computations in multi-regional neural circuits) This collaborative project aims to uncover the logic of signal routing from subcortical areas to the frontal cortex through the thalamus. The frontal cortex displays rich patterns of neural activity, which can be decomposed into “activity modes” that correspond to specific aspects of behavior. Examples include the persistent activity correlated with short-term memory and motor planning, and the rapidly oscillating activity during voluntary movements. In this dynamical systems perspective of neural computation, complex behaviors correspond to distinct sequences of cortical activity modes. However, the cortex does not generate these activity modes in isolation, but instead is strongly and bidirectionally coupled to the thalamus, the central hub of the forebrain. Most of thalamus is non- sensory (‘higher-order’), receiving subcortical input from the cerebellum, midbrain, and hippocampus. Our central theory is that these subcortical signals flow through higher-order thalamus to reach the frontal cortex, where they enable activity modes, update activity modes, and cause switching between modes, akin to the ‘update’ and ‘reset’ signals in Long Short-Term Memory networks in machine learning. However, most of what we know about thalamus comes from sensory systems, and our knowledge of subcortex→ thalamus→ frontal cortex circuits is nascent. We still have only a rudimentary understanding of the input and output circuits of higher-order thalamus, the morphology and molecular properties of thalamic neurons, the circuit motifs that link subcortical input to cortical activity, and the engagement of these networks across the frontal cortex. We bring together a team with expertise in modern high-throughput anatomy (Project 1, 2), molecular neuroscience (Project 2, Molecular Science Core), cellular and synaptic neurophysiology (Project 3), large-scale neurophysiology in mice performing behaviors that require short-term memory and decision-making (Project 3, 4), and theory and computation (Project 5, Data Science Core). We will collaborate to uncover how information flows from subcortical areas, through thalamus, to control cortical activity modes and thereby shape behavior. Individual projects are guided by a conceptual framework of multi-regional neural computation, placing the thalamus in the middle of a multi-regional neural network. Together, our work will have broad implications for the understanding of neural computation in subcortex→ thalamus→ cortex circuits and will produce anatomy- guided multi-regional circuit models of cognitive function. We will also produce paradigm-shifting community resources, including quantitative anatomy, novel genetic reagents, neurophysiological data, and a rich modeling framework, upon which future studies of thalamic circuits will be built.

Narrative, Overall (Thalamus in the middle: computations in multi-regional neural circuits) The thalamus is the main conduit for subcortical information flow to the cortex. Despite decades of work in the sensory thalamus, we know little about non-sensory (or 'higher-order') thalamus, even though higher-order thalamus is the largest part of the thalamus and damaging it can cause cognitive and motor disorders, as well as coma and hypersomnia. This collaboration will map the architecture of subcortex → thalamus → cortex circuits, measure functional relationships of these brain regions with cell type-specificity during multiple cognitive behaviors, and synthesize these data in conceptual and computational models for a multi-regional neural circuit theory of cognition.