Corticostriatal plasticity is necessary for learning intentional neuroprosthetic skills

Abstract

The ability to learn new skills and perfect them with practice applies not only to physical skills but also to abstract skills, like motor planning or neuroprosthetic actions. Although plasticity in corticostriatal circuits has been implicated in learning physical skills, it remains unclear if similar circuits or processes are required for abstract skill learning. Here we use a novel behavioural task in rodents to investigate the role of corticostriatal plasticity in abstract skill learning. Rodents learned to control the pitch of an auditory cursor to reach one of two targets by modulating activity in primary motor cortex irrespective of physical movement. Degradation of the relation between action and outcome, as well as sensory-specific devaluation and omission tests, demonstrate that these learned neuroprosthetic actions are intentional and goal-directed, rather than habitual. Striatal neurons change their activity with learning, with more neurons modulating their activity in relation to target-reaching as learning progresses. Concomitantly, strong relations between the activity of neurons in motor cortex and the striatum emerge. Specific deletion of striatal NMDA receptors impairs the development of this corticostriatal plasticity, and disrupts the ability to learn neuroprosthetic skills. These results suggest that corticostriatal plasticity is necessary for abstract skill learning, and that neuroprosthetic movements capitalize on the neural circuitry involved in natural motor learning.

Figure 1. Volitional modulation of M1 neural activity in awake behaving rats

a. Task schematic. M1 unit activity was entered into an online transform algorithm that related ensemble activity to the pitch of an auditory cursor. Two opposing ensembles were chosen, with activity of one ensemble increasing the cursor pitch and activity of the other ensemble decreasing the cursor pitch. Constant auditory feedback about cursor location was supplied to rodents, and distinct rewards were supplied when rodents brought M1 activity into one of two target states. b. Mean M1 ensemble firing rates for units in ensemble 1 (green), ensemble 2 (blue), and M1 units not used in the transform (black) in relation to the achievement of target 1 (top) or target 2 (bottom). Representative waveforms recorded from M1 in rats are shown on the right, with shaded regions denoting the standard deviation.

Figure 2. Learning intentional neuroprosthetic actions independently of movement

a. Mean percentage of correct responses for all rats across days 1–11 of learning. Shaded regions denote the range of days from which the early and late learning analyses were performed. b. The percentage of correct responses for all rats increased significantly from early (light blue) to late (dark blue) in learning. c. Percentage of correct responses in late learning is significantly greater than expected by chance. d. Representative accelerometer traces show no gross motor behavior leading to target achievement (time zero), but clear deflections as animals initiate movement to retrieve reward. e. Representative EMG traces show no muscle activity in the mystacial pad before target achievement, but clear deflections as animals retrieve and consume reward. f. Mean performance in all rats when lidocaine was injected into the whisker pad before a behavioral session late in learning (red) compared to performance during a no-lidocaine session (dark blue). g. Significant reduction in response rate when the causal relation between target achievement and reward delivery was degraded (dark blue). When contingency was reinstated, performance rapidly returned to pre-degradation levels (red). h. Percentage of total correct trials that were directed at the target associated with pellet reward (blue) or sucrose solution reward (red) during choice sessions where rats had free access to pellets (left; “Pellets Devalued”), or to sucrose before the session (right; “Sucrose Devalued”). i. Percentage of total trials that involved responses toward target 1 (blue), target 2 (red), or response omissions (black) when omission tests were performed for target 1 (left) or target 2 (right).

Figure 3. Learning abstract skills is accompanied by corticostriatal plasticity

a. Mean normalized firing rates in DS increased significantly from early (light blue) to late (dark blue) learning. Representative waveforms recorded from the DS are shown on the right (shaded regions denote standard deviation). b. Z-scored firing rates for individual DS units in relation to target achievement (time zero) showing marked modulation before target achievement in late learning. c. The percentage of DS units exhibiting target-related firing rate modulation increased significantly with learning. d. Cross-correlation histograms in late learning for M1 spiking activity in relation to DS spikes (left), and DS spiking activity in relation to M1 spikes (right), showing oscillatory coupling between the two regions. e. Coherence between M1 spikes and DS spikes in early (left) and late (right) learning shows a clear increase in low frequency coherence from early to late learning. f. Significant increase in mean coherence in the theta range in late (dark blue) versus early (light blue) learning. Shaded regions denote standard error.

Figure 4. Selective deletion of NMDA receptors in the striatum impairs BMI learning

a. RGS9L-Cre/Nr1^f/f mice (red) exhibit no significant increase in the percentage of correct trials over the course of learning, despite clear performance improvement in littermate controls (blue). b. Accelerometer traces from control mice showing no clear oscillation before target achievement, but clear deflections as mice retrieve reward. c. There is no significant increase in DS firing rates in RGS9L-Cre/Nr1^f/f mice (red) from early to late learning, although DS firing rates increase markedly in control mice (blue). d. DS units of controls exhibit strong target-related firing rate modulations, including both excitation (left) and inhibition (right). e. The percentage of DS units showing significant target-related firing rate modulations increases significantly across learning in control mice (blue), but not in RGS9L-Cre/Nr1^f/f mutants (red). f. Coherograms showing coherence between M1 spikes and DS spikes in early (left) and late (right) learning for both control mice (top) or RGS9L-Cre/Nr1^f/f mice (bottom). Coherence in low frequency bands increases from early to late learning in control mice, but not in RGS9LCre/Nr1^f/f mice. g. Mean coherence in the theta range for control (top) and knockout (bottom) mice., There is a significant increase in coherence from early (light blue) to late (dark blue) learning in control mice but not in mutant mice (early learning, light red; late learning- dark red).