Cerebellar activity is triggered by reach endpoint during learning of a complex locomotor task

Andry Andrianarivelo, Heike Stein, Jeremy Gabillet, Clarisse Batifol, Abdelali Jalil, N. Alex Cayco-Gajic, Michael GraupnerbioRxiv 2023 · DOI

TL;DR

This study introduces "LocoReach," a novel task where mice learn to walk on a runged treadmill, requiring precise paw placement. They discovered that cerebellar interneurons (MLIs) exhibit activity spikes precisely timed to critical moments in the paw's step cycle, particularly at the end of a swing phase. This suggests the cerebellum plays a crucial role in fine-tuning motor control for precise locomotion, especially during learning and in challenging environments.

What Problem Does This Paper Address?

Navigating complex environments and maintaining balance during locomotion demands incredibly precise control over individual limb movements. The cerebellum is known to be vital for this, but how its internal circuits, specifically molecular layer interneurons (MLIs), encode the dynamic and often irregular movements required for tasks like walking on uneven surfaces or over obstacles remains unclear. Existing research often focuses on simpler, flat-surface locomotion. This paper aims to bridge that gap by investigating cerebellar activity during a more challenging, adaptive locomotor task that mimics the need for precise paw placement and coordination.

Methods

The researchers developed a new behavioral paradigm called LocoReach. In this task, head-fixed mice walk on a motorized, runged treadmill in complete darkness. This setup allows for:

  • High-resolution paw tracking: Using infrared cameras and sophisticated algorithms, they precisely track the movement of each paw.
  • Locomotor learning: Mice gradually improve their performance over several days, making fewer, longer, and more accurate strides.
  • Cerebellar manipulation: They used two methods to investigate the cerebellum's role:
  • Pharmacological inactivation: Infusing muscimol into the cerebellar lobule simplex to broadly suppress activity.
  • Optogenetic inhibition: Using a closed-loop system to precisely inhibit MLI activity during specific phases of the paw's swing.
  • Electrophysiological recordings: They performed loose cell-attached recordings of MLIs in the cerebellar lobule simplex to capture their neural activity during the task.
  • Sophisticated data analysis: They employed techniques like peristimulus time histograms (PSTHs) and a two-step generalized linear model (GLM) to correlate neural activity with specific behavioral events and parameters.

Datasets

The study introduces the LocoReach dataset, which consists of behavioral data (paw kinematics) and neural recordings (MLI activity) from mice trained on the runged treadmill task over multiple sessions. This dataset is novel and specifically designed to capture learning-related changes in locomotion and corresponding cerebellar activity.

Key Insights and Ideas

  • Cerebellum is event-driven, not just continuous: Contrary to some assumptions about continuous kinematic encoding, MLI activity is tightly locked to discrete events in the step cycle, particularly the "reach endpoint" (end of swing). This suggests a role in precise timing rather than just smooth trajectory tracking.
  • Learning amplifies neural signals: As mice improve at the task, the neural responses in MLIs become stronger, indicating increased engagement and finer tuning of motor control.
  • Interneuron activity is key to motor precision: Inhibiting cerebellar output during the swing phase directly impacts paw movement, shortening the swing. This highlights the direct causal role of cerebellar interneurons in controlling limb kinematics.
  • Beyond ipsilateral control: While MLIs are most responsive to the ipsilateral front paw, a significant portion also encodes events from other paws, suggesting a role in coordinating multi-limb movements.

Key Results

  • Behavioral learning: Mice significantly improved performance on LocoReach over sessions, showing fewer strides, longer swing lengths, and fewer missteps.
  • MLI activity linked to step cycle events: A large majority of MLIs (74.2%) showed significant firing rate modulations around swing and stance transitions. These modulations were most prominent for the front left paw.
  • Optogenetic inhibition reduces swing length: Disrupting cerebellar output during the front left paw's swing phase led to significantly shorter swings of that paw.
  • Learning increases MLI engagement: As mice learned, more MLIs became responsive to step cycle events, and a greater proportion showed mixed selectivity for multiple paws.
  • Swing length drives modulation: The magnitude of MLI firing rate modulations was correlated with swing length, suggesting increased cerebellar involvement for more demanding steps.

Limitations and Future Work

  • Head-fixed setup: While enabling neural recordings, the head-fixed nature of the task might differ from natural locomotion. Future work could explore free-walking paradigms.
  • MLI subtypes: The study recorded from MLIs broadly. Future research could investigate potential differences in function between specific MLI subtypes (e.g., stellate vs. basket cells).
  • Compensation for perturbation: Mice adapted to optogenetic stimulation over time, showing no behavioral differences after several sessions. Understanding the mechanisms of this adaptation (e.g., synaptic plasticity) is a key future direction.
  • Cerebellar output pathways: While the study implicates lobule simplex, further investigation into how MLI activity influences downstream cerebellar nuclei and ultimately motor output is needed.
  • Other cerebellar regions: The study focused on lobule simplex. Exploring other cerebellar regions involved in locomotion would provide a more complete picture.

Summary updated

Abstract

Locomotion in complex environments depends on the precise timing and active control of single paw movements in order to adapt steps to surface structure and coordinate paws. Such motor control crucially depends on the cerebellum, which is thought to support moment-to-moment prediction and correction of paw trajectories. Although cerebellar activity has been linked to limb kinematics on flat surfaces, it remains unknown how cerebellar cortical neurons encode paw movements when gait becomes irregular and the requirements for precise motor control vary dynamically. To address this question, we developed LocoReach: a new task which combines continuous and discrete aspects of motor control by requiring mice to walk on a runged treadmill, where each step involves reaching for the next rung. Over several days of learning, mice became increasingly proficient at LocoReach, so that they made fewer, longer strides with faster swings and fewer missteps. Through real-time optogenetic disruption of cerebellar processing, we shortened the swing of the perturbed paw highlighting the online contribution of lobule simplex to precise limb control the task. We next investigated the role of the cerebellar lobule simplex during LocoReach learning using electrophysiological recordings, with particular focus on molecular layer interneurons (MLIs) that shape the timing and gain of Purkinje cell (PC) output. When analyzing behaviorally-evoked responses in MLIs and PCs, we found sharp changes in activity around paw-specific transitions from swing to stance and vice versa. Cells in lobule simplex showed clear behavioral specificity: most neurons were tuned to swing-stance transitions of the ipsilateral paw, a large proportion encoded transitions of other or even multiple paws. Specifically MLIs exhibited larger amplitude firing-rate changes during longer strides acquired through learning, indicating increased engagement during higher-demand steps. These results show that cerebellar activity is tightly aligned to defined events in the step cycle, providing a mechanism through which cerebellar cortical circuits can contribute to the precise control of paw placement during adaptive locomotion.
BibTeX
@misc{andrianarivelo2023cerebellar,
  author = {Andry Andrianarivelo and Heike Stein and Jeremy Gabillet and Clarisse Batifol and Abdelali Jalil and N. Alex Cayco-Gajic and Michael Graupner},
  title = {{Cerebellar activity is triggered by reach endpoint during learning of a complex locomotor task}},
  year = {2023},
  doi = {10.1101/2023.10.10.561690},
  howpublished = {bioRxiv},
}
Figure