Modular neural circuit architecture for optic flow processing in the larval zebrafish


March 4, 2015 - 1:00pm
NW 243
About the Speaker
Eva Naumann (Engert Lab)

To study the neural computations underlying visual motion processing, we examined visually evoked behavior in freely swimming larval zebrafish using a closed-loop assay to precisely control the visual environment and monitor behavioral output. Analysis of behavioral responses to combinations of monocular and whole-field motion stimuli revealed the principles of information integration across eyes as well as general visuo-motor transformations that are used by fish for appropriately directed locomotion. Eye- and direction-specific behavioral kinematics predicted multiple, separate and overlapping neural circuit modules fed by distinct information channels emerging from each eye. Together, these modules account not only for the observed linear integration of congruent binocular information but also the stabilization of orienting behavior in ambiguous visual environments and the separate modulation of swim bout frequency and orientation change. With whole-brain two-photon calcium imaging, we found that primary information processing takes place in retino-recipient, lateralized pretectal nuclei that integrate motion binocularly and enable necessary reciprocal suppression via an interhemispheric connection. Subsequent locomotor instructions are then distributed to specific premotor areas that encode turn amplitude and bout frequency separately. Furthermore, cluster analysis across tens of thousands of neurons revealed that the majority fall within a limited set of functional classes that match the circuit modules predicted from behavior. Together, these results motivated a ­­feed-forward rate code model composed of overrepresented functional cell classes, illustrating a modular brain architecture efficiently controlling sensorimotor transformations.