Connectivity underlying motor cortex activity during goal-directed behaviour

Steven

Connectivity underlying motor cortex activity during goal-directed behaviour

Summary

This Nature paper maps how microcircuit connectivity in mouse motor cortex shapes neural activity during a goal-directed, multidirectional tongue-reaching task. The authors combined large-scale volumetric two-photon calcium imaging (tens of thousands of neurons) with targeted two-photon photostimulation to infer causal connections among excitatory neurons in layer 2/3 while animals performed behaviourally relevant movements. They analysed tuning to target location and time, assessed reward-related modulation beyond simple licking, and used GLMs and modelling to link measured causal connectivity to observed tuning and noise correlations across behavioural states. The team make their behavioural and neurophysiology data (NWB) and analysis code publicly available (DANDI and GitHub).

Key Points

  • Large-scale volumetric imaging and targeted two-photon photostimulation were used to measure causal connectivity and activity in mouse motor cortex during a multidirectional tongue-reaching task.
  • Neurons showed robust location–temporal tuning to target location and timing; tuning stability was quantified across trials and sessions (dataset includes >82,000 imaged neurons).
  • Two-photon photostimulation evoked reliable trial-by-trial responses in target and causally connected neurons, with positive correlations that persist after removing dominant population components.
  • Causal connection strength predicts tuning similarity and noise correlations beyond what is explained by anatomical distance, and can predict correlations across behavioural states (rest vs task).
  • Reward-outcome modulation in many neurons cannot be explained solely by changes in lick-rate; GLM analyses show reward terms contribute beyond time and licking predictors.
  • All behavioural/neurophysiological data are available in NWB on DANDI and custom MATLAB code and models are on GitHub (links in the paper).

Content summary

The study trained mice on a multidirectional tongue-reaching task and recorded large populations of excitatory layer 2/3 neurons while simultaneously applying targeted two-photon photostimulation to selected neurons. The authors characterised temporal and spatial tuning of neurons to target location and movement timing, quantified tuning stability, and examined reward-related activity using GLMs to separate licking effects from genuine reward signals.

Using photostimulation, they mapped causal influence from stimulated neurons to responders across the imaged volume. Photostimulation-evoked responses showed trial-to-trial covariation between targets and connected neurons that remained after removing major shared components, indicating functional interactions rather than only global fluctuations. Analyses demonstrated that measured causal connectivity relates to tuning similarity and noise correlations, even when controlling for pairwise distance, and that causal maps can predict correlation structure across states. The paper includes extensive extended data describing behaviour, tuning metrics, photostimulation protocol, controls and statistical tests.

Context and relevance

This work links synaptic-level causal connectivity to population activity patterns that underlie precise, goal-directed motor actions. By combining all-optical perturbation with large-scale imaging and careful statistical controls, the study advances our understanding of how microcircuit wiring shapes task-relevant representations and variability in motor cortex. The open dataset and code also provide an important resource for testing models of cortical computation and recurrent circuit dynamics.

Why should I read this?

Want to know how the wiring of motor cortex actually drives movement-related signals — not just correlations but causal links? This paper does exactly that, using big imaging datasets and targeted photostimulation. If you care about how microcircuits produce reliable behaviour, or you want an open dataset and code to play with, this will save you time and give you solid leads.

Author style

Punchy and method-driven: the team pair rigorous all-optical perturbation with large-scale recordings and clear statistical controls. If you work on cortical circuits, motor control or network dynamics, the methods and open data are especially worth diving into.

Source

Source: https://www.nature.com/articles/s41586-025-09758-6

Data & code

Behavioural and neurophysiology NWB data: https://dandiarchive.org/dandiset/001612. Analysis and modelling code: https://github.com/arsenyf/Finkelstein_et_al_Nature_2025.