P-17 Cortical-subcortical information flow reverses from naive to skilled movements

Background: Despite the dense interconnectivity of the brain’s
motor network, cortical and subcortical motor regions are associated
with distinct types of movements. Stable or habitual actions are classically
linked to subcortical regions, while more flexible or exploratory
control is commonly attributed to cortex. Nonetheless, no study
has investigated how motor-specific information is transmitted
between cortical and subcortical regions at different stages of
learning.
Objective: Studying how cortical and subcortical regions communicate
during the learning of a reach-to-grasp task by monitoring neural
signals in motor cortex (M1) and dorsolateral striatum (DLS) as
rats produce naive exploratory and stable skilled movements.
Methods: We used information theory to measure the timing of
information encoded about task-relevant movement features in M1
and DLS at different stages of learning. Then, we applied the recently
developed measure of Feature-specific Information Transfer (FIT) to
measure the motor-specific information flow between M1 and DLS.
FIT merges the Wiener-Granger causality principle of information
transmission with feature-specificity. In doing so, FIT extends previous
methodologies such as Directed Information (DI) which measures
the total, non-feature-specific, information transmission
between neural signals. By measuring both FIT and DI, we could dissociate
the transmission that was specifically about movement from
the overall information flow between M1 and DLS. Lastly, we developed
a simple network model to understand which type of modification
in the structure of the M1-DLS network could reproduce the
results we obtained from experimental data.
Results: We found that the timing of motor information in M1 and
DLS changed from naïve to skilled stages of learning. In particular,
the motor-information peaks in DLS were anticipated in time, while
they occurred later in M1, in the skilled compared to the naïve condition.
Crucially, using FIT, we measured that the directionality of
motor-specific information transmission between M1 and DLS
reversed with learning, with motor information flowing mainly from
M1 to DLS in the naïve and from DLS to M1 in the skilled condition.
This reversal was hidden when measuring overall neural information
transmission with DI, which bi-directionally increased between M1
and DLS with learning. We could reproduce the above results in a
network model by reversing the area that originated the motorspecific
information, going from a cortical-origin during naive movements
to a striatal-origin during skilled movements.
Conclusion: Our results provide evidence of a reversal in the area
originating, and then transmitting motor-specific information within
the M1-DLS network as rats learned a reach-to-grasp task. Such finding
provide essential support to the idea that cortical and subcortical
regions have a leading role in controlling, respectively, naïve and
automatized movements.