Distributed correlates of visually-guided behavior across the mouse brain

Background

The study of neurophysiological activity during behavioural tasks has a long and informative history; however, previous technology had limited the scope of what we could discover. Many studies have focused on the role of a single brain region in a task. This work has revealed that multiple brain regions are active throughout each stage of visual guided behaviour. Many of the regions involved are reciprocally connected, and so fully understanding their role involves understanding their activity in coordination, not just in isolation.

Existing research methods have demanded a compromise between how many locations could be measured simultaneously and the spatial resolution at which they could be measured. Neurophysiology has long focused on the activity of individual neurons, but it has been a significant challenge to record from multiple single neurons in multiple brain regions with traditional techniques. Brain imaging techniques such as fMRI enables brain-wide recording, but the signal from each voxel is a proxy for the activity of a great number of neurons. Neuropixels silicon probes are a step forwards in resolving this compromise. They enable the simultaneous recording and individual resolution of hundreds of neurons across multiple brain regions.

In this study, Steinmetz and colleagues from UCL have used Neuropixels probes to simultaneously record from hundreds of neurons in mice performing a visual discrimination task. Based on the neural dynamics, they outline a behaviour-related network spread across multiple brain regions.

Key findings

Using Neuropixels silicon probes, almost 30,000 neurons were recorded across 42 brain regions during a head-fixed visually-guided behavioural task (92 probe insertions over 39 sessions in 10 mice, average of 747+/-38 neurons per session). Mice were trained in a two-alternative unforced choice task, which involved turning a wheel to the left or right to indicate the location of a visual grating with higher contrast, or to withhold a response in the absence of any stimulus.

Neuronal activity in nearly all recorded regions increased following trial onset, first in visual regions contralateral to the stimulus, then to other regions, including regions ipsilateral to the target stimulus. Passive presentation of the stimuli outside of the behavioural task produced activation that was more limited in scope and contained solely within the contralateral hemisphere. This suggests the widespread activation was specific to task engagement.

Further activity specific to task engagement could be seen in the pre-stimulus activity. In the 0.2 s prior to stimulus onset, firing rates in many regions were substantially different between passive presentation and task conditions. Pre-stimulus firing rates were lower in visual cortex and visual thalamus during task engagement, but higher in regions including the basal ganglia and midbrain structures. Interestingly, pre-stimulus activity on test trials when the animal failed to respond resembled passive presentation trials more closely than successful test trials.

Although many neurons across several regions responded throughout various stages of the task (stimulus presentation, movement, reward presentation), the authors found relatively few neurons that discriminated between the choice to move in one direction or the other. Those neurons that did discriminate accounted for a small subset of neurons in a limited set of regions – frontal cortex (Mos, PL, and Mop), striatum (CP) and midbrain (SNr, SCm, MRN, and ZI). Neurons in the forebrain and midbrain encoded choice differently – neurons in the midbrain almost exclusively increased firing during choices in the contralateral direction, while only a slight majority of choice-selective neurons in the forebrain preferred contralateral choices. Further, activity in the majority of choice-selective midbrain neurons was suppressed prior to their non-preferred choice, while this was only true of a much smaller proportion of choice-selective forebrain neurons.

Why is it important?

This study provides an extensive picture of the coordinated activity of individual neurons in a widespread network of brain regions during a behavioural task. Although the activity of individual cells in a range of brain regions have been studied before, this study leverages the Neuropixels silicon probes to simultaneously record a great number of neurons from multiple regions. These simultaneous recordings offer convincing evidence that task-related activity is distributed across a wide range of brain regions in a way that separated recordings do not. These simultaneous recordings allow the authors to suggest a model of circuit dynamics across multiple brain regions that is supported more directly than if the conclusions were built upon the combination of several experiments.

Further, the dataset is due to be shared by the investigators upon publication of the paper. This will allow other researchers to contribute by analysing the data in new ways and in greater detail.

Questions arising

How is the activity in each of these regions causally related to behaviour during the task? The authors refer to other studies where optogenetic inhibition of selected brain regions impair task performance.

How is activity in each brain region causally related to activity in others and how well can the system resist perturbation at each point?

The authors describe the responses as “near-simultaneous” – can we resolve the temporal profile of responses in each region relative to others and determine the flow of activity through the network?

What system modulates the different pre-stimulus activity that reflects ‘task engagement’? The authors suggest dopamine as a neuromodulatory agent in the basal ganglia. Acetylcholine is another contender, as it has been shown to change baseline activity and signal-noise ratio during attentive engagement.

 

doi: https://doi.org/10.1242/prelights.6488

Read preprint (No Ratings Yet)