PROJECT SUMMARY Stroke is the leading cause of motor disability in the United States. While brain stimulation to enhance motor function after stroke has shown promise in small studies, two recent large stroke trials did not find evidence for significant benefits. A key uncertainty is about how to exactly tailor brain stimulation to effectively modulate neural dynamics associated with movement preparation and control. Our recent studies in rats (Ramanathan et al., Nature Medicine 2018; Lemke et al., Nature Neuroscience, 2019) demonstrated that population dynamics linked to low-frequency oscillatory activity (0.5-4Hz “LFO”) are essential for movement control and can serve as a target for modulation using electrical stimulation. More specifically, cortical stimulation was found to both boost LFO power and augment motor function. We now also have substantial evidence in a non-human primate model that such an approach can be effective in more complex brains. However, it is essential to further optimize the delivery of such stimulation to specifically target cortical dynamics. We thus propose to optimize parameters for epidural stimulation to selectively modulate population dynamics in the intact motor network. Our approach entails simultaneous recording of single neurons in the non-human primate motor network along with electrical stimulation using a customized “ring” of epidural cranial screw electrodes. Moreover, we will use computational analysis to determine how task-related neural dynamics in a reach-to-grasp task are modulated by electrical stimulation. More specifically, we will optimize and develop principles for large-scale electrical stimulation to selectively enhance “neural modes” isolated to M1 or PMd or joint across both areas. This approach is built on the growing consensus that motor networks perform computations through coordinated ensemble activity or “neural modes”, i.e. patterns of neural covariation measured with dimensionality reduction methods. Activation of neural modes (i.e. Neural Model Activation or NMA) appear to constitute building blocks for computations underlying movement control. Our specific aims are: 1) Determine optimal ACS parameters that increases both local and cross-area NMA between M1 and PMd during a reach-grasp task; 2) Determine optimal ACS parameters that increases both local and cross-area NMA between M1 and S1 during a reach-grasp task; 3) Determine parameters for ACS to enhance task NMA during time periods away from the task. Completion of these aims will provide critical information for designing therapeutic stimulation that selectively targets population dynamics in the distributed motor network. The information gained may also help improve methods for non- invasive brain stimulation.