A fundamental question in neuroscience is how entire neural circuits generate behavior, and adapt it to changes in the environment. Because large populations of neurons, across multiple brain areas, are involved in implementing sensorimotor transformations, it is desirable to record activity from many of them during behavior. Here, we introduce a new paradigm for recording neural activity, at the single neuron level, throughout the entire brain of the larval zebrafish, from several thousands of neurons at a time, during fictive behavior. A two-photon microscope scans over a fish expressing a genetically-encoded calcium indicator in almost all neurons, while the animal is engaged in a virtual environment. The fish is actually paralyzed, to completely stabilize the brain for imaging, and intended motor output - called fictive swimming - is measured by electrodes recording from motor neuron axons in the tail.
This setup is used to investigate the dynamics of large neuronal populations during motor learning, during which the fish adapts its motor output to changes in the strength of visual feedback following a motor command, much like learning in the vestibulo-ocular reflex. Upon changing the feedback gain in a closed-loop virtual environment, the fictively behaving animals quickly adapt their motor output to compensate for the change: an increase in feedback gain leads to a reduction in swim power, and vice-versa. Imaging throughout the brain during this behavior, in combination with dimensionality-reduction methods for the analysis of the activity of large populations of neurons, reveals multiple brain areas with activity correlating with multiple aspects of the behavior. In particular, the cerebellum and the inferior olive are implicated, potentially forging a link between motor adaptation in fish and in mammals. Lesioning the inferior olive removes the ability of the animals to adapt their motor output, showing that this nucleus is necessary for the implementation of flexible motor programs.
The fictive virtual-reality paradigm is then extended from one to two-dimensions, so that the fish can fictively swim around arbitrary visual environments, opening the possibility of recording and perturbing whole-brain activity at the single neuron level also during more complex behaviors.