Abstract

Background

Acquired injuries to primary motor cortex (M1) contribute to motor impairment and disability. While reorganization that occurs during the post-acute period following injury has been characterized at static endpoints and correlated with behavioral outcomes extensively in animal models, there is a relative lack of information about how the role of cortex as a controller of movement evolves concurrently. Importantly, this same period is considered a critical period for rehabilitation.

Objective

The purpose of this study was to characterize injury-related changes in movement-related features of the neural activity in bilateral motor cortex that occurs on the timescale of milliseconds to seconds immediately before, during, and after a skilled upper-extremity behavior.

Method

Bilateral microwire arrays were chronically embedded in layer 5 of rat cortical motor areas in order to record extracellular field potentials during reward pellet retrievals, daily, over the first four weeks following a focal vasoconstrictive ischemia targeted to the rat M1 homolog (caudal forelimb area; CFA) located contralateral to the preferred reaching forelimb. Multi-unit spiking activity was analyzed at the levels of individual channels, by area, and as an aggregate proxy for control dynamics that commonly occur during motor behavior. Details.

Result

When the reach-to-grasp was broken down into an advance and retract phase, it was found that spike counts in the distinct rostral forelimb area (RFA) located in the secondary motor cortex of the ipsilesional hemisphere exhibited a small but statistically significant decrease in average spike counts just prior to the advance phase when the ischemia group was compared to intact control rats. By contrast, during the retract phase, spike counts were subtly but significantly higher in RFA of rats from the ischemia group, although a reduction in the duration of this phase was better able to explain longitudinal changes. Similarly, there was no significant difference in distribution of units with predictive power for determining successful pellet retrievals when comparing contralesional CFA to ipsilesional RFA, or in longitudinal trends. Surprisingly, when bilateral population-level features were regressed to recover linearized first-order dynamics using standard multiple least-squares (MLS) regression, individual rats exhibited relatively direct longitudinal trends between the ability to accurately regress dynamics and the time from surgery. Furthermore, when regression was constrained such that matrix optimization imposed a specific kind of fixed point with rotatory structure (i.e. jPCA), it was observed that accurate MLS regression was only predictive of a recovery phenotype when the skew-symmetric regression served as a reasonable MLS approximation. Critically, this result suggests that the presence of rotatory dynamics in the population structure of multi-unit spiking is indicative of computations that are optimal for accurate movement. By contrast, when only MLS regression could accurately describe dynamics it was predictive of reduced functional outcomes (likelihood of successful pellet retrieval). We speculate that the poor behavioral outcomes may be related to filter instability, as is qualitatively suggested by the fixed points obtained in such cases. Applying the Kalman formulation, it is hypothesized that such instability could either arise due to an erroneous transformation on sensory feedback (i.e. measurement noise), or due to the inability of motor areas to accurately predict the future limb state (i.e. process noise). Future work to test which of these scenarios is more likely will be critical to the success of defining impairment at the level of the cortical controller with respect to rehabilitation from stroke. Details.