Transition from predictable to variable motor cortex and striatal ensemble patterning during behavioral exploration

Sravani Kondapavulur, Stefan Lemke, David Darevsky, Ling Guo, Preeya Khanna, Karunesh GangulyNature Communications 2022 · DOI

TL;DR

This study reveals how the brain flexibly switches between precise, automatic movements and exploratory behavior when faced with errors. By monitoring rat brains during a reach-to-grasp task, researchers found that increased neural variability in the motor cortex (M1) and striatum signals behavioral exploration, allowing for adaptation to new task demands.

What Problem Does This Paper Address?

We learn to perform many actions automatically and precisely, like riding a bike or typing. However, our brains must also be flexible enough to adapt when things go wrong or when the environment changes. It's not fully understood how the brain, specifically the primary motor cortex (M1) and the dorsolateral striatum (DLS) – key areas for motor control and learning – can support both consistent, automatic movements and the exploration needed to correct errors or learn new strategies. This paper investigates this transition, particularly when an established, automatic skill encounters consistent errors.

Methods

The researchers used a reach-to-grasp task in rats. They simultaneously recorded neural activity (electrophysiology) from M1 and DLS.

  • Behavioral Paradigm: Rats were trained to reach for a pellet at a specific location (Location A). Once this skill was automated, the pellet location was switched to a new spot (Location B), requiring the rats to adapt. This adaptation process was observed over several days.
  • Neural Recording: Simultaneous electrophysiological recordings were made from M1 and DLS to capture the activity of individual neurons and local field potentials.
  • Neural Variability Analysis: A key metric used was the Fano factor, which measures the trial-to-trial variability of neural spiking. Higher Fano factors indicate more variability.
  • Ensemble Patterning: Techniques like Canonical Correlation Analysis (CCA) were used to assess how activity patterns in M1 and DLS were coordinated. They also analyzed the consistency of single-trial neural patterns compared to an average "template" for the session.
  • Causal Manipulation: To directly test the role of neural variability, researchers used intracranial microstimulation (ICMS) in M1. They compared the effects of reliably patterned stimulation (predictable) versus variably patterned stimulation (unpredictable) on movement variability.

Datasets

Not explicitly detailed as separate datasets for training/evaluation. The study generated its own behavioral and neural data from the experimental setup with rats.

Key Insights and Ideas

  • Exploration is a "Variable" State: The study strongly suggests that behavioral exploration, especially when adapting to errors, is not just a minor adjustment but a distinct neural state characterized by increased variability in M1 and DLS.
  • Tandem Variability: This increased neural variability isn't isolated to one brain area; it occurs simultaneously in both M1 and DLS, suggesting a coordinated shift.
  • Flexibility Between Modes: The brain can flexibly switch between two modes: a "reliable pattern generation" mode for automatic, precise movements and a "variable neural patterning" mode for exploration and adaptation.
  • ICMS as a Causal Link: The ICMS experiments provide strong evidence that variability in M1 output can directly lead to variability in movement, supporting the idea that M1's output pattern is key to controlling movement precision.

Key Results

  • Multiday Adaptation: Rats took several days to adapt to the new pellet location, indicating that overcoming consistent errors is a gradual learning process, not rapid within-session adaptation.
  • Loss of Modulation During Exploration: During the "variable" sessions (when rats were exploring the new location), there was a significant decrease in the trial-averaged modulation of M1 and DLS neural activity locked to reach onset.
  • Increased Fano Factors: The Fano factor, a measure of trial-to-trial variability, significantly increased in M1 and DLS during these variable sessions.
  • Return to Consistency: As rats relearned the task and became proficient at reaching the new location, M1 and DLS neural activity regained consistent task-aligned modulation and decreased variability.
  • Patterned vs. Random Stimulation: Variable ICMS patterns in M1 led to significantly greater movement endpoint variability compared to reliably patterned stimulation.

Limitations and Future Work

  • Directionality of Influence: While the study shows tandem variability in M1 and DLS, it doesn't definitively establish whether M1 variability drives DLS variability or vice versa, or if other brain areas are involved in coordinating this. Future work could involve simultaneous recordings from more brain regions (e.g., prefrontal cortex, secondary motor cortex) and targeted inactivation experiments.
  • Cell-Type Specificity: The study used general electrophysiology. Future research could explore the roles of different neuron types within M1 and DLS (e.g., IT vs. PT pathways) in this flexible control.
  • Motor Noise and Skill Crystallization: The authors suggest that individual differences in learning speed might relate to "motor noise" and the degree to which a skill is "crystallized." Further investigation into how sleep and consolidation processes influence this motor noise could be fruitful.
  • Habit vs. Skill Distinction: While the study touches on the distinction between motor skills and habits, future work could more directly explore how these neural mechanisms differ when a behavior transitions from a goal-directed skill to a rigid habit.

Summary updated

Abstract

Abstract Animals can capitalize on invariance in the environment by learning and automating highly consistent actions; however, they must also remain flexible and adapt to environmental changes. It remains unclear how primary motor cortex (M1) can drive precise movements, yet also support behavioral exploration when faced with consistent errors. Using a reach-to-grasp task in rats, along with simultaneous electrophysiological monitoring in M1 and dorsolateral striatum (DLS), we find that behavioral exploration to overcome consistent task errors is closely associated with tandem increases in M1 and DLS neural variability; subsequently, consistent ensemble patterning returns with convergence to a new successful strategy. We also show that compared to reliably patterned intracranial microstimulation in M1, variable stimulation patterns result in significantly greater movement variability. Our results thus indicate that motor and striatal areas can flexibly transition between two modes, reliable neural pattern generation for automatic and precise movements versus variable neural patterning for behavioral exploration.
BibTeX
@inproceedings{kondapavulur2022transition,
  author = {Sravani Kondapavulur and Stefan Lemke and David Darevsky and Ling Guo and Preeya Khanna and Karunesh Ganguly},
  title = {{Transition from predictable to variable motor cortex and striatal ensemble patterning during behavioral exploration}},
  booktitle = {Nature Communications},
  year = {2022},
  doi = {10.1038/s41467-022-30069-1},
}
Figure