Effect of motor control on sensory coding in the awake, freely moving mouse

Textbook knowledge of how neurons respond to sensory stimuli (visual, tactile etc) comes mainly from a passive experimental paradigm where sensory stimuli are presented to the immobilised sense organ of an anaesthetised animal (e.g., spot of light projected on the retina). However, most species actively seek information by active sensation - they coordinate motor control of their sense organs (e.g., eyes, hands, whiskers) with sensory processing. For example, during natural, sensory-guided actions such as tea-making, humans make coordinated movements of torso, hands and eyes. Studies of neural mechanisms in behaving rodents have revealed that the activity of neurons in the primary sensory areas of the cerebral cortex, classically considered to be purely sensory areas (V1, S1, A1), depends on motor state (running speed, whisker movement, pupil size). The surprising, new picture emerging from these findings is that sensation and action are integrated even in the brain's early sensory systems.

Whilst the head-fixed preparation has important experimental benefits, it substantially limits the range of movements that the animal can make. Head-fixed mice can walk/run on a treadmill but this is just one of many elementary motor actions in the behavioural repertoire of freely moving mice; others include head rotations along different axes, rearing, body bending and elongation. These motor actions combine to generate a wide repertoire of behavioural motifs that cannot occur when under head-fixation. Thus, the question of exactly how sensory coding depends on motor state in early sensory systems remains unresolved.

The ideal way to approach our objectives is to record neural activity from mice which are free to move so that they can exhibit as many different motor actions as possible. Our approach is to image the behaviour of mice as they explore objects with their whiskers, using a system of video cameras. We have recently developed powerful machine vision algorithms that can automatically extract the 3D pose of a mouse and the 3D shape of its whiskers. We have also recently established an electrophysiological system for state-of-the-art wireless recording of neural activity. Putting these methods together, we are ideally placed to contribute to the field.

The ideal way to address our objectives is to perform experiments where: (1) mice are free to move, (2) mouse/whisker state can be precisely quantified and (3) neural activity can be measured with cellular resolution. Such experiments were difficult in the past but new methods make them feasible and timely. First, we have recently developed automatic systems for tracking body/whisker state from video. Second, we have established a cutting-edge wireless recording technique to measure neural activity from freely moving mice whilst minimally constraining their behavior.

To address O1, we will image mouse behavior as they explore objects with their whiskers. Using multiple cameras allows us (via out new tracker algorithms) to reconstruct both body and whisker posture in 3D. This will result in time series that describe both the 3D motor state of the mouse (whiskers, head and torso) and sensory input to the whisker system (whisker-object touch). Application of clustering analysis to these data, will allow us to identify recurrent, behavioural motifs in body-head-whisker actions.

To address O2-3, we will record neural activity from freely moving mice whilst they explore objects with their whiskers. This will be done using a state-of-the-art wireless recording system. In O2, activity will be recorded from the whisker-related thalamic relay nucleus (VPM); in O3 from the primary somatosensory cortex (S1). By applying our tracker algorithms to the video data, we will thereby obtain data that consists of time series of the whisker-body state of the mouse (as defined in O1) in register with spike trains of neurons in primary somatosensory cortex (S1). The rival hypotheses (action-specific modulation or global modulation) have distinct implications for how well neural activity can be predicted from mouse motor state. We will test these implications by fitting statistical models to the neural data.