Behavioural framework for studying visually-directed reaching in the mouse

Abstract

Reaching and grasping are fundamental behaviours that humans and other mammalian species depend on to perform useful interactions with our environment. To execute this behaviour, forelimb movements require to be accurately targeted based on the egocentric spatial position of objects of interest. As we navigate our environment, we first visually identify the location of nearby objects, then use this information to compute necessary movement trajectories. Decades of primate research has successfully utilised reaching tasks to identify and characterise a number of brain areas involved in visually-directed movement. Mice display a somewhat similar behavioural repertoire to humans, including the ability to reach and grasp, and can therefore be used as a model system to investigate neural control of movement. The improved genetic tractability of mice has expanded our ability to delineate cell type- and pathwayspecific functionality in the brain, and such research benefits from higher throughput and lower financial cost. However, due to the lack of a suitable behavioural task, to date there has been a distinct lack of research using mice to study visually-directed reaching. To address this problem, I aimed to design a novel task for mice that researchers can use to study this critical behaviour.

First, I designed and implemented a task in which freely moving mice learned to perform visually-directed reaches. During the task, one of two reach targets was illuminated and mice learned to reach and grasp the illuminated target to receive a water reward. Once trained, mice would perform an initial reach to the visual stimulus followed by up to 10 repetitive, self-initiated reaches. Expert mice had a high success rate and performed an average of 120 cued reaches per 30-minute session after 4 weeks of training.

Next, to take full advantage of the kinematic richness of the behaviour by enabling the task to be combined with high temporal resolution extracellular recordings, I developed a user-friendly data processing and analysis pipeline for electrophysiological data collected using Neuropixels silicone probes. The pipeline, named pixels and opensourced at https://github.com/DuguidLab/pixels, provides a simple Python API that enables visualisation and statistical testing of neuron population extracellular activity recorded during trial-based behavioural tasks. To enable high resolution examination of the spatiotemporally detailed movements performed during the task and how they relate to neural activity, pixels integrates closely with the DeepLabCut motion tracking toolbox (Mathis et al., 2018).

Lastly, as a proof-of-concept demonstrating the utility of the visually-directed reaching task, I performed extracellular recordings in expert mice and used pixels to examine neural correlates of behaviour. Using chronically-implanted Neuropixels silicone probes, I recorded population activity in posterior secondary motor cortex (pM2) and dorsomedial striatum (DMS), brain areas implicated in visuomotor control (Hwang et al., 2019; Itokazu et al., 2018) and goal-directed behaviour (Hwang et al., 2019; Wang et al., 2013), respectively. In deep layers of pM2, over half of pM2 neurons exhibited reproducible firing rate changes as mice initiated and executed visually-directed reaches. During a 300 ms pre-reach window, a small proportion of pM2 neurons displayed activity changes, suggestive of a role in movement preparation or initiation. During reach execution, reduced firing rates dominated pM2 population activity in both pyramidal cells and interneurons, while a small subset of pyramidal cells became more active. Two competing models may explain this finding. The small excitatory component may convey motor signals, with widespread inhibition possibly serving to improve the signal-to-noise ratio in recipient areas. Alternatively, movement-related information may be conveyed directly by the firing rate decreases, complementing the excitatory signals to provide bidirectional modulation of recipient areas. Only approximately half of pM2 neurons active during cued reaches remained active during repetitive reaches, suggestive of distinct motor pathways underlying these types of behaviours. Activity profiles in DMS mirrored those in pM2, indicative of a close functional relationship between the two areas. A quarter of DMS neurons active during cued reaches were also active during repetitive reaches. As repetitive reaches had lower average peak velocities, this finding is consistent with DMS regulating movement vigour according to expected reward value (Wang et al., 2013). By identifying novel correlates between activity in these brain areas and visually-directed reaching, this experiment demonstrates how the newly developed task can yield new insights into the neural dynamics underlying this behaviour.

In addition, I utilised pixels to analyse silicone probe recordings as part of collaborative projects to investigate the neural underpinnings of forelimb motor control in mice performing object manipulation tasks (Currie et al., 2022; Dacre et al., 2021). We found that cerebellar-thalamocortical signals are required to shape motor cortical activity and trigger movement in a behavioural context-dependent manner (Dacre et al., 2021). Next, we examined how motor cortex encodes movement when mice perform two distinct movements. We found that while movement-invariant responses dominated layer 5B projection neurons, movement-specific activity is displayed by a small, distributed population of layer 5B neurons and differentially between projection neuron classes (Currie et al., 2022).