The underlying neural basis of automatic action versus flexible movement exploration

The underlying neural basis of automatic action versus flexible movement exploration
Multiday relearning of an automatic skill. (a) Automated reach-to-grasp setup. After the door opens (signaled by a tone), the rat reaches through a slit to retrieve the pellet. M1 (green) and DLS (gray) locations in inset. (b) Relearning paradigm. Rats are overtrained to reach to location A (left). Then pellet moved to location B with continued training (middle, right). (c) Endpoint of first reaches. Top: example animal, histogram of endpoint x-position, across trials. Bottom: example reach trajectories and endpoint locations relative to pellet location (large black circle) for same sessions as (top). (d) Decay in reaches to A. Percentage of trials in a session with first reach (R1) to location A, as compared to low-amplitude (short, S) reaches and/or reaches to location B. (e) For those trials with multiple reaches, within-trial updating for reaches after the first were classified into short (S), old (A), or new (B) reaches. (f) Example x-trajectory of paw during a trial, with reach onset marked; r1 and r2 are the first and second reach onsets, respectively. g Reach type after first reach, for trials with multiple reaches. For all reaches after the first reach in a trial, proportion of A, B, or S reaches. Data are presented as mean values ± SEM. h Following an inaccurate trial, we classified the first reach type of the subsequent trial. i Reach type after inaccurate reach to A on previous trial. Data are presented as mean values ± SEM. *<0.05; **<0.01; ***<0.001, ANOVA, Tukey-Kramer post hoc test. Credit: Nature Communications (2022). DOI: 10.1038/s41467-022-30069-1

Neuroscientists aim to understand how the primary motor cortex (M1) can regulate precise movements while supporting behavioral exploration that encounters consistent errors. While consistency in the environment can help animals learn and automate highly reliable actions, they must remain flexible to adapt to environmental changes. In a new report now published in Nature Communications, Sravani Kondapavulur and a research team in neural engineering, at the University of California, San Francisco, U.S., developed a reach-to-grasp task in rats to monitor the neural variability of the primary motor cortex (M1) and dorsolateral striatum (DLS) regions. The results indicated how the motor and striatal areas transitioned between two modes by generating a reliable neural pattern for automatic and precise movements, compared to variable neural patterning observed during behavioral exploration.

How do movement circuits in our brain support both automatic action and flexible movement exploration?

While mobility and motor skills are defined as fast, accurate and consistent actions, they must also be flexible to adjust to a task at hand or to a changing environment. The link between the production of consistent and behavioral exploration for error correction currently remains elusive. In systems neuroscience, the is thought of as an engine for precise control by generating reliable sequential firing.

In this work, primary author and MD Ph.D. candidate Sravani Kondapavulur, along with the research team of Associate Professor Karunesh Ganguly at UCSF, described the movement circuits underlying behavioral flexibility in our brain. "Every day, most people use learned movements to accomplish daily tasks, such as turning on lights in a room or reaching for and grasping a coffee mug from a cabinet," she said.

"These movements are often automatic—produced in a fast, accurate, and consistent manner without much overt thought. Yet, a skill should also remain flexible to readily adjust to changes in the environment, such as moving to a different home with alternate light switch and cabinet placement." Since the process of automatic action and flexible movement are facilitated by circuits in our brain, Kondapavulur et al sought to examine the underlying mechanisms of the process in an . "We first overtrained rats to reach and grasp for a food reward in a singular location. We then moved the reward to a second, nearby location and observed how they re-learned where to retrieve the pellet," she said.

The underlying neural basis of automatic action versus flexible movement exploration
Breakdown of reach-locked M1 and DLS unit spiking during exploration. (a) Top: Rasters for a single M1 unit† across sessions (same channel across days). Bottom: trial-averaged firing rate (mean ± SEM) for −2 s to 2 s around first reach onset. (b) Same as (a), for a DLS unit.† (c) Equation for determining Fano factor per time bin for each unit, per session. (d) Five minimum Fano factors per unit for −0.5 s to 0.5 s around first reach onset, across representative sessions, for two example animals. (e) Session type re-categorization based on Fano factors: indigo = baseline, blue = automatic (post-switch, no Fano factor change), gray = variable (maximum increase in Fano factor), orange = relearned. Light blue/light orange = two sessions not analyzed because of data corruption during recording. (f) M1 unit peak trial-averaged modulation around reach onset for trials with first reach to A (blue dots) and for trials with first reach to B (orange dots), across session types; peak modulation was calculated after z-scoring unit firing across the session. (g) Same as (f), for DLS units. (h) Change in proportion of first reach to A in a session, relative to proportion of first reach to A in previous session; grouped by session type across all animals, all sessions. i Proportion of first reach to B in a session, grouped by session type across all animals, all sessions. †Unit may not be the same across sessions. *<0.05, **<0.01, ***<0.001, linear mixed effects model, Bonferroni-corrected significance. Credit: Nature Communications (2022). DOI: 10.1038/s41467-022-30069-1
Experimental food reward relocation

During the experiments, the team was specifically interested in the underlying role of M1 (primary motor cortex) and DLS (dorsolateral striatum) to modify skills, while still largely maintaining the original action. They accomplished this by "recording from both the primary motor area (M1) and the downstream motor area (DLS)." Kondapavulur further evaluated their findings: "We found that re-learning was a multi-day process, with many errors of reaching to the old location early on." She continued, "interestingly, on the day with the greatest breakdown of automatic movement errors," they noticed "a switch in M1 and DLS brain activity from consistent patterns during the task, to significantly increased variability from trial to trial."

The team notably observed a close association between the established behavioral exploration; transitioning from the highly predictable ensemble activation in the two neural regions of interest, to a variable state in both regions. Kondapavulur et al. continued the experiments for subsequent days to highlight how modifying an automatic skill was a multi-day process.

The underlying neural basis of automatic action versus flexible movement exploration
Changes in M1 and DLS single-trial ensemble reliability. (a) Top: leave-one-out trial-averaged spiking pattern of M1 units for reach period, −0.5 s to 0.5 s around first reach onset, across session types. Bottom: left-out trial correlation with session template. (b) Trial-to-template correlation across trials, with A trial-to-template correlations shown for the baseline session, and B trial-to-template correlations shown for subsequent sessions. Baseline (BL), automatic (auto), variable (var), and relearned (rel). (c) Comparison of M1 trial-to-template correlations for all trials by session type. d Same as (a) for DLS units in example sessions. (e) DLS trial-to-template correlations for all trials by session type. *<0.05; **<0.01; ***<0.001, linear mixed effects model, Bonferroni-corrected significance. Credit: Nature Communications (2022). DOI: 10.1038/s41467-022-30069-1

Re-establishing behavioral flexibility, in time

The scientists next sought to understand the of re-establishing behavioral flexibility. They observed a loss of the previously stable neural spiking behavior relationship across the two regions, and increased neural variability, which coincided with goal-directed behavioral explorations, and behavioral changes during the switch process. According to Kondapavulur et al, while examining a potential link between variable neural activity and the process of behavioral exploration, they showed how the "timing of maximal 'communication' between the two regions was no longer consistently associated with timing of the reaching behavior." Further experiments from stimulation in M1 showed that "consistent versus variable timing of electrical activity in the primary motor area could produce correspondingly consistent and variable movements."

The underlying neural basis of automatic action versus flexible movement exploration
Changes in M1-DLS cross-area subspace modulation. (a) Canonical correlation analysis (CCA): identification of axis with maximal correlation between M1 and DLS activity (purple dotted line), with example value of projected population activity at time t1 (solid line on shared axis) for M1 population activity (left, green) and DLS population activity (right, black). (b) Example bootstrap shuffled distribution of R2 values for identification of significant canonical variable (CV) axes. (c) Top: single-trial M1 spiking activity. Middle: single-trial DLS spiking activity. Bottom: corresponding single-trial M1 (green) and DLS (black) activations along CV1. Pre-reach period (gray) is −1 to −0.5 s before first reach onset. Reach period (blue-orange) is −0.1 s to 0.4 s around first reach onset. (d) M1 and DLS cross-area activity before (black) and during (color) reach, for A trials (top row) and B trials (bottom row) across session types for example animal. (e) Equation for relative modulation index (RMI) of CCA activation, comparing reach and pre-reach periods. (f) RMI for trials with first reach to A, for M1 activations (left, green) and DLS activations (right, green), across automatic and variable sessions. R1 to A, M1 and DLS RMI: auto, n = 220 trials; var, n = 95 trials. Data are presented as box plots with 25th, 50th, and 75th percentiles. (g) Same as (f), for trials with first reach to B, across variable and relearned sessions. R1 to B, M1 and DLS RMI: var, n = 150 trials; rel, n = 464 trials. Data are presented as box plots with 25th, 50th, and 75th percentiles. *<0.05; **<0.01; ***<0.001, linear mixed effects model, Bonferroni-corrected significance. Credit: Nature Communications (2022). DOI: 10.1038/s41467-022-30069-1

Shared neural activity

In its neural architecture, the primary motor cortex and dorsolateral striatum regions (M1 and DLS) are structurally monosynaptically connected; while DLS also receives inputs from a variety of other cortical and subcortical regions. Kondapavulur et al therefore explored if, despite the loss of consistent timing to the task, the population spiking remained correlated to each other in the two regions. They accomplished this by examining cross-area communication between the two neural regions and how the areas changed during variable sessions via canonical correlation analysis to represent maximally correlated activity across areas, to identify shared neural activity. They found that while communication axes could still be identified during the variable state, the maximum shared activity was no longer consistently reach-related, also demonstrating time-related variability.

The underlying neural basis of automatic action versus flexible movement exploration
The role stimulation variability in driving variability of upper-limb movements. a Location of stimulation, via microwire array to M1 for identification of regions eliciting forelimb movement. b Patterns of stimulation delivery, with balance of total charge delivered. Patterned burst stimulation (top, PB) consisted of 333 Hz triplets, at 10 Hz triplet frequency. Random pulse stimulation (bottom, RP) consisted of single pulses, randomly timed at 30 Hz. Stimulation was delivered for 1 s per session, for a total of 30 pulses per session. c Stimulation pulse parameters, with 200 μA biphasic pulse, 200 μs per phase with 100 μs inter-phase interval. d Lateral view of paw: limb markers labeled for supervised kinematic tracking: tip of digit 4 (D4, blue), tip of digit 3 (D3, purple), and center of paw (center, black). Endpoint location of each trial (Δx, Δy) was calculated relative to start location in trial window and normalized by animal across stimulation conditions. e Example animal histograms of normalized Δx (top) and Δy (bottom) for center of paw across patterned burst stimulation trials (left) and random pulse stimulation trials (right). Gray boxes show an example 300 ms window with identical number of pulses. Kolmogorov-Smirnov test for two samples in this example animal, center of paw, p = 1.8e−11; *<0.05; **<0.01; ***<0.001, Kolmogorov-Smirnov test for two samples. Credit: Nature Communications (2022). DOI: 10.1038/s41467-022-30069-1


The team wrote: "We found that transitioning from automatic, consistent movements to enable flexible exploration involved a transition from consistent to variable motor area activity." They found the outcome to be significant since "it demonstrates another distinct method of movement learning, separate from novel skill learning, such as playing the piano, when compared to the process of adaptation, i.e., updating fast movements in the face of errors."

In this way, Sravani Kondapavulur and colleagues sought to understand the underlying basis of behavioral flexibility during automatic performance of a complex motor skill. They showed that neural regions of the primary motor cortex (M1) in tandem with dorsolateral striatum (DLS) regions enabled flexible transitions between behaviorally rigid and exploratory states. The team explored the behavioral flexibility of reach and grasp with a rodent model at a new location in a multiday process. The work outlined the relation of a motor skill and a habit to show how automaticity or unconscious performance of an action were applicable to both instances. The outcomes support the underlying basis of motor brain activity during the proposed actions.

More information: Sravani Kondapavulur et al, Transition from predictable to variable motor cortex and striatal ensemble patterning during behavioral exploration, Nature Communications (2022). DOI: 10.1038/s41467-022-30069-1

Andrew J. Peters et al, Emergence of reproducible spatiotemporal activity during motor learning, Nature (2014). DOI: 10.1038/nature13235

Journal information: Nature Communications , Nature

© 2022 Science X Network

Citation: The underlying neural basis of automatic action versus flexible movement exploration (2022, May 16) retrieved 13 July 2024 from
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

Explore further

Ready, set…GO! Scientists discover a brain circuit that triggers the execution of planned movement


Feedback to editors