The striatum is a critical brain area in the control and initiation of movement. Recently, the activity in striatal spiny projection neurons (SPNs) has also been shown to be different for different gross body movements, suggesting that activity in striatum is movement-specific and can bias which movements animals will perform. However, these larger actions were performed with different body parts, and may reflect somatotopy more than action specificity, and therefore it is unclear how fine the granularity of movement-specificity is. To probe this, we have developed an isometric force task that requires no overt movement, and imaged and analyzed the activity in striatum during learning and execution. From the optically recorded neural activity combined with simultaneous behavior, we trained classifiers that could predict action from activity across days, revealing stable ensembles encoded specific actions. Finally, using targeted two-photon stimulation, we perturbed small sub-populations of these ensembles in closed-loop during behavior, seemingly leading to small, but perceptible, changes in action.
|