Sensory processing has conventionally been studied on restrained or anaesthetised animals. However, in natural settings, sensory perception is not as controlled or restricted; instead, it is a dynamic process involving continuous interactions between an animal’s movements and sensory inputs. This process of active sensing and sensorimotor integration is vital for adaptive behaviour. Thus, studying animals in more naturalistic settings, where they are freely moving, is crucial for better understanding sensory processing. The overall aim of our lab has been to do this by combining electrophysiological recordings from freely moving mice with advanced methods for 3D tracking of body state. Here, we specifically focus on comparing the function of primary vs associational sensory areas, using the mouse whisker system as a model. Previous work on immobilised animals has shown that neurons in the secondary somatosensory cortex (S2) have larger and more complex receptive fields compared to the primary somatosensory cortex (S1) but no previous study has compared their function in freely moving animals. To investigate this, we chronically implanted Neuropixels probe into the S1 and S2 and recorded the neural activity in both regions simultaneously while the animal was freely moving and exploring objects in an open-field arena.

All the behavioural experiments were conducted in darkness under infrared illumination, with four cameras capturing the freely moving mouse behaviour. We tracked 11 landmarks on the head and body using DeepLabCut and reconstructed their 3D coordinates using a custom triangulation algorithm. Snout-to-surface distance (SSD), or the distance between the snout and the closest point on any of the arena surfaces, served as a proxy for whisker-surface touch. We also characterised the body state by extracting parameters such as 3D whole-body velocity, 3D allocentric head angles, their temporal derivatives, and principal components of the body shape from the 3D landmarks.

A supervised learning algorithm (XGBoost) was trained to predict the firing rates of S1 and S2 neurons based on SSD and body state parameters. To assess the extent of body state encoding in S1 and S2 and to exclude the possible correlation between whisker touch and body state, all the experiments were repeated after severing the infra-orbital nerve to eliminate whisker afferent input. We found that SSD was a statistically significant predictor of firing rate for the majority of S1 and S2 neurons. Incorporating body state into the prediction significantly improved the accuracy in both areas. These results suggest that body state modulation is widespread, affecting both primary and associative sensory cortices.