Happy to see the story on motor cortex finally out in @natureneuro today!

This was a fun project that spanned my entire time in grad school. Very grateful for the guidance of @BOlveczky, collaborators extraordinaire @Jack_W_Lindsey @seanescola, and the rest of the Ölveczky lab. Now on to the postdoc!

Remarkably, this architecture and learning rule naturally gives rise to the interference effect described above, where simultaneously learning flexible and automatic sequences prevents the consolidation of the automatic behavior.

The subcortical network follows a reinforcement learning algorithm that allows it to consolidate automatic motor sequences initially learned by the motor cortex. However, it lacks the input information required to execute flexible sequences, leaving those MC-dependent.

We then teamed up with @Jack_W_Lindsey and @seanescola, who built a biologically inspired neural network model, consisting of a motor cortical and subcortical component, to perform a virtual motor sequence task.

Thus, demands for the flexible re-use of motor elements that are part of an overtrained motor sequence interfered with its subcortical consolidation, rendering it MC dependent.

Could additional demands for flexibility interfere with the subcortical consolidation of an automatic sequence? To test this, we trained a cohort to do *both* flexible and automatic sequences in separate sessions. Remarkably, MC was now required for the automatic behavior!

But thus far, we've only considered scenarios where motor sequences are either overtrained or generated flexibly. In many cases, the motor elements that make up an automatic sequence are also re-used in flexible contexts (such as the keypresses in your password).

So BG are required for automatic but not flexible sequences, while MC is required for flexible but not automatic ones, suggesting distinct roles for these two important motor regions.

Now what about cue-guided sequences? Lesioning MC in rats trained on such flexible cue-guided sequences disrupted both the high-level sequence structure and the learned kinematics, implicating MC in flexible sequence execution.

But first, we needed to test whether our earlier results generalized to our piano-playing rats. We lesioned MC in rats overtrained on a single sequence. Just as Risa Kawai had seen a decade earlier, the learned motor patterns were intact after lesion!

While not required for overtrained stereotyped motor sequences, perhaps MC might be necessary for orchestrating cue-guided sequences of basic movements and actions – i.e., movements that are controlled subcortically.

We had previously shown that MC is not required for automatic behaviors. Here’s a video of a rat performing the very same motor skill before and after MC lesion! You can read about that study here: https://t.co/HS8Dmy3EQm.

However, this led us to ask: if not the BG, what part of the brain is sequencing flexible behaviors? One possibility was motor cortex (MC). We probed its role in our next study, which the preprint of is up now!

Surprisingly, animals could produce cue-guided sequences even after lesions to the BG. This mirrors results from Parkinsonian patients, whose actions can be aided and improved when provided with external cues. Here’s a patient receiving visual cues to help walk:

In our first study, we found that basal ganglia (BG) helped shape low-level movement kinematics for both types of sequences and were required for the high-level sequencing of automatic behaviors.

To study these distinct forms of sequence execution, we taught rats to play the piano, having them perform either lever-press sequences instructed by external cues, or a single sequence overtrained to the point of automaticity. See video below!

To recap, we wanted to understand sequences guided by external cues (e.g., a pianist reading from sheet music), differ from automatic ones, developed through repeated practice (e.g., performing a piece in concert).

Also, the first half of the story is now out! https://t.co/ux7OXB3Vn7 See our previous tweets about it here:

Ever wonder how the pianist can perform a new piece from sheet music, or effortlessly play a well-practiced sonata in a concert? We study the underlying neural circuits in ‘piano-playing’ rats and show that motor cortex and basal ganglia play different roles.