The mammalian cortex plays a critical role in integrating multiple streams of information to guide adaptive behavior. For example, head direction (HD) information is combined with visual and spatial input in the mouse retrosplenial cortex (RSC). Accurate integration of these signals is a necessary component of navigation: recognizing a distant landmark while facing north vs. facing south has very different interpretations for one's position and future actions. However, the mechanisms by which any cortical association area integrates different inputs at the level of individual neurons during behavior is unknown. RSC is therefore a compelling model system in which to test general associative computations during a complex behavior: the combination of visual and HD information during navigation. Anatomical evidence suggests that HD inputs computed in the anterior thalamus make their synapses at distal apical dendrites in RSC, while visual and motor synapses are located closer to the somas of RSC principal neurons. This arrangement suggests that nonlinear dendritic integration may be used by RSC to combine HD with other inputs. Active dendritic integration is theorized to allow single neurons to respond flexibly to different combinations of input, where the state of one input nonlinearly influences the impact of another input. Our overarching hypothesis is that such mechanisms could work in concert with neural circuit computations to implement context-dependent cortical computations. Congruent with this idea, RSC neurons in navigating rats exhibit complex conjunctive receptive fields, a feature that is lacking from commonly studied primary sensory cortices. RSC is therefore an ideal area to evaluate the role of dendrites in associative computations during navigation. However, current methods are not well-suited to this level of investigation: they either allow mice to behave freely or they achieve sub-cellular resolution. This has led to a critical gap in our understanding of navigation, and by extension, associative cortex function. We have recently developed technology that bridges this gap: an animal-actuated rotating headpost that allows mice to engage in 2-D navigation by freely turning their head during conventional 2-photon imaging. We will use this new approach to test the hypothesis that neurons in RSC use sub-cellular processing to flexibly combine HD and visual information during navigation behavior. These experiments will provide new insights into cellular- and circuit-level mechanisms of navigation, and of associative cortical function in general. Results from this project will be valuable for understanding brain disease states as well as for building biologically-inspired artificial neural networks.
The brain is a network of staggering complexing, yet each of its elementary units, single neurons, can themselves be seen as a network of sub-modules, performing complex functions such as learning to flexibly integrate multiple inputs. These within-cell computations may play an important role across in a variety of disease states including schizophrenia, ADHD, and Alzheimer?s disease, though, their role in neural circuit function remains poorly understood. This project will apply new tools and approaches to establish the role of dendritic computations during behavior, providing novel insight into general mechanisms of associative cortical function.
