Connectivity underlying motor cortex activity during goal-directed behaviour

Neural representations of information are shaped by long-range input and local network interactions. Previous studies linking neural coding and cortical connectivity have focused on input-driven activity in the sensory cortex^1-3. Here we studied neural activity in the motor cortex while mice gathered rewards with multidirectional tongue reaching. This behaviour does not require training, allowing us to probe neural coding and connectivity before activity is shaped by extended learning. Motor cortex neurons were tuned to target location and reward outcome, and typically responded during and after movements. We studied the underlying network interactions in vivo by estimating causal neural connections using an all-optical method^3-6. Mapping connectivity between more than 20,000,000 excitatory neuron pairs showed a multi-scale columnar architecture in layer 2/3 of the motor cortex. Neurons displayed local (less than 100 µm) like-to-like excitatory connectivity according to target-location tuning, and inhibition over longer spatial scales. Connectivity patterns comprised a continuum, with abundant sparsely connected neurons and rare densely connected neurons that function as network hubs. Hub neurons were weakly tuned to target location and reward outcome but influenced more neighbouring neurons. This network of neurons, encoding location and outcome of movements to different motor goals, may be a general substrate for rapid learning of complex, goal-directed behaviours.