Multi-timescale neural adaptation underlying long-term musculoskeletal reorganization
- National Center of Neurology and Psychiatry, Department of Neurophysiology, Japan
- University of ElectroCommunications, Graduate School of Informatics and Engineering, Department of Mechanical and Intelligent Systems Engineering, Japan
- National Center of Neurology and Psychiatry, Department of Orthopaedic Surgery, Japan
- Western Institute for Neuroscience, University of Western Ontario, Canada
Figures
Figure 1
Muscle anatomy of the macaque forearm and the tendon transfer procedure.
A schematic of the primary forearm muscles involved in the study, showing both the dorsal and volar views. The diagram illustrates the surgical crossed tendon transfer of the extensor digitorum communis (EDC) and flexor digitorum superficialis (FDS) tendons. All labeled muscles were implanted with electromyography (EMG) electrodes. Muscle abbreviations: BRD: brachioradialis, ECR: extensor carpi radialis, ECU: extensor carpi ulnaris, ED2,3: extensor digitorum-2,3, ED4,5: extensor digitorum-4,5, EPL: extensor pollicis longus (not implanted), FCR: flexor carpi radialis, FCU: flexor carpi ulnaris, FDP: flexor digitorum profundus, PL: palmaris longus, PT: pronator teres (not implanted). (The deltoid [DEL] muscle was also implanted in Monkey B but is not shown as it is a shoulder muscle.) See also Supplementary file 1A for a complete list of all muscle abbreviations, their full names, and their assigned synergies.
Figure 2
Long-term confirmation of tendon surgery effectiveness to alter mechanical properties.
(A) Setup for the ultrasound measurement and video recordings of the stimulation-induced movements of the extensor digitorum communis (EDC) and flexor digitorum superficialis (FDS) tendons. (B) Sonogram of the FDS muscle and its intramuscular tendons. Left side (B-mode, i.e. brightness mode) shows the still image of the monkeys’ forearm at a given point in time. Right side (M-mode, i.e. motion mode) shows the staggered images of the FDS tendon displacement induced by muscle stimulation (50 mA). White arrows demarcate the FDS tendon which was used for the measurement. Grayscale gradations correspond to tissue densities: hyperechoic regions (white) denote denser structures like the surface of bones and tendons, while hypoechoic areas (black) signify less dense tissues such as adipose tissue and musculature. Inset demonstrates the area measurement. The area of the displacement waves was measured in the M-mode, representing the strength of muscle contraction. We measured the duration (a, s) and amplitude (b, cm) of three waves and calculated the average. Area = a*b/2(cm/s) for days 0, 7, and 105 after tendon transfer (TT). (C) Areas under the wave measured in the M-mode for 3 experimental days (0, 7, and 105 days post-TT) and regression lines in red and blue for FDS and EDC, respectively. R2>0.5 for FDS. The data suggested that muscle contractions induced by direct electrical stimulation were nearly constant. (D) Markers placed on the index, middle, and ring finger nails (A) were used to measure finger displacement in xyz-dimensions. We calculated the sum of the Euclidean distances of each marker from the origin of the 3D coordinate system as a scalar quantity. Observing the movement along the z-axis, it became reversed post-surgery, indicating a reversal from finger flexion to extension due to tendon transfer (D, blue = pre-TT at surgery day; dark brown = post-TT at surgery day; light brown = 1 week post-TT; red = 3 weeks post-TT). The scalar quantity of the fingers during muscle stimulation did not change much at day 0, 7, and 105 days (E), suggesting that there was no tendon rupture or slackening of the tendons postoperatively (EDC stimulation, left; FDS stimulation, right). Data were collected in Monkey A.
Figure 3
Experimental setup, task sequence, and typical electromyography (EMG) (Monkey A).
(A) Schematic of the task object using a rod requiring Monkey A to perform a controlled grasp. (B) Schematic of the behavioral sequence (hook → grasp → release). (C) Typical EMG traces of a control session (high-pass filtered) for all recorded muscles. Gray boxes represent the task sequence. Obj 1 ON: start of the hold period of object 1. Obj 1 OFF: end of the hold period of object 1, i.e., object release. Obj 2 ON: start of the hold period of object 2. Obj 2 OFF: end of the hold period of object 2, i.e., object release. Tendons of the muscles marked with * were cross-transferred. (D) Rectified and smoothed EMG for all recorded muscles (average for one recording session; amplitude [μV] over task sequence [%]). Horizontal bars illustrate the corresponding behavioral periods; red vertical lines indicate peak amplitude for each muscle. (E) The time the monkey spent on the left side of the yellow dotted line while moving from object 1 to object 2 was measured and used to quantify the maladaptive behavior.
Figure 4
Experimental setup, task sequence, and typical electromyography (EMG) (Monkey B).
(A) Schematic of the task requiring Monkey B to pick up food from a groove allowing for a more natural grasp. (B) Schematic of the task sequence (picking up food). (C) Typical EMG traces of a control session (high-pass filtered) for all recorded muscles. Gray boxes represent the task. Obj 1 ON: start of the hold period of object 1. Obj 1 OFF: end of object 1’s hold period, i.e., object release. LED ON: approximate start of the food touch. LED OFF: approximate time of food retrieval. Tendons of the muscles marked with * were cross-transferred. (D) Rectified and smoothed EMG for all recorded muscles (average for one recording session; amplitude [μV] over task sequence [%]). Horizontal bars illustrate the corresponding behavioral periods; red vertical lines indicate peak amplitude for each muscle. (E) Example of maladaptive behavior in Monkey B. The time the monkey spent in contact with or behind the object plate was measured and used to quantify the maladaptive behavior.
Figure 5
Behavioral and kinematic metrics of motor recovery.
(A, D) Grip formation times (mean ± SD; n=20 trials) for Monkey A (A) and Monkey B (D). (B, E) Duration of off-target reaching movements (mean ± SD; n=10 trials) for Monkey A (B) and Monkey B (E). (C) Pull time duration for Monkey A. (F) Grasp aperture size for Monkey B. Filled squares indicate significant difference from pre-tendon transfer (TT) baseline (p<0.05, two-sample t-test). All data are plotted over days relative to TT.
Figure 6 with 4 supplements
Temporal electromyography (EMG) profiles and cross-correlation analysis.
Temporal EMG profiles for Monkey A (left, A–J) and Monkey B (right, K–Q). (A–F, K–O) Average EMG activity profiles aligned to task events (0%; object release for A, food touch for B). Shaded envelopes represent standard deviations. Triangles indicate peak activity during extension (▼) or flexion (▽). Colored traces denote post-surgery landmark days. (A–D, K–M) Comparison of transferred muscles. Note the temporal shift in the post-surgery profile (B, L) relative to the pre-tendon transfer (TT) baseline (dashed lines). (E–F, N–O) Profiles of non-transferred muscles. (G–J, P–Q) Zero-lag cross-correlation coefficients between post-surgery EMG profiles and the pre-TT baseline profiles plotted over time. (H, I, P) Correlation coefficients calculated against the muscle’s own pre-TT baseline. (G, J, Q) Correlation coefficients calculated against the antagonist’s pre-TT baseline (e.g. post-EDC vs. pre-FDS). Black dashed lines on the right y-axis indicate behavioral error metrics (off-target reaching time for Monkey A; contact duration for Monkey B; gray shading represents SD). The // represents the recovery period. EDC, extensor digitorum communis; FDS, flexor digitorum superficialis.
Figure 6—figure supplement 1
Evolution of time lag at peak cross-correlation between transferred muscle electromyographies (EMGs).
Quantification of the optimal time lag (in % task range) yielding the maximum cross-correlation between individual muscle EMG profiles. (A–D) Monkey A; (E–F) Monkey B. Each panel plots the time lag calculated between the recorded EMG profile and the average pre-tendon transfer (TT) baseline (specific comparisons, e.g. ‘post-EDC vs. pre-EDC’, are indicated in titles). Left y-axis: time lag. Blue traces: pre-surgery data; red traces: post-surgery data. Right y-axis (gray dashed traces): behavioral error metrics (off-target reaching duration for Monkey A; contact time for Monkey B). A positive lag value indicates that the muscle activity is delayed relative to the pre-surgery baseline; a negative value indicates it is advanced (occurs earlier). The dotted horizontal line indicates zero lag. Shaded envelopes represent standard deviations. Black triangles indicate landmark days. EDC, extensor digitorum communis.
Figure 6—figure supplement 2
Electromyography (EMG) activity profiles for all recorded muscles across all sessions.
EMG activity profiles [μV] for all recorded muscles in Monkey A (A–T) and Monkey B (U–AO). (A–J, U–AF) Average pre-tendon transfer (TT) profiles (± SD in blue). (K–T, AG–AO) Post-surgery profiles for each recording day, aligned on object release (Monkey A) or food touch (Monkey B). Lighter colors indicate later recording days (see color bar). Muscles with surgically transferred tendons are marked with an asterisk (*).
Figure 6—figure supplement 3
Electromyography (EMG) activity profiles of ‘landmark days’.
Monkey A: EMG activity profiles [μV] of selected ‘landmark days’ for all recorded muscles (pre- and post-tendon transfer [TT], A–J and K–T, respectively). Profiles are aligned on hold off-set (object 1) at zero percent task range [%] (dashed lines). Horizontal gray bars mark the three behavioral periods (hook, grasp, and release, respectively). Monkey B: pre- and post-TT, U–AF and AG–AO, respectively. Profiles are aligned on LED on-set (food touch) at zero percent task range [%] (dashed lines). Horizontal gray bars mark the four behavioral periods (finger extension, food touch, food pick-up, and food transport, respectively).
Figure 6—figure supplement 4
Cross-correlation analysis for individual muscles.
Cross-correlation of each muscle’s electromyography (EMG) activity profile against its own pre-tendon transfer (TT) control data. The plots show the evolution of the correlation coefficient over post-surgery days for Monkey A (A–J) and Monkey B (K–S), quantitatively illustrating the time course of adaptation for each muscle.
Figure 7 with 4 supplements
Spatial structure and temporal activation of primary synergies.
Analysis of the two primary synergies: Synergy A (flexor) and Synergy B (extensor). (A, B, E, F) Spatial synergy weights (W) showing the contribution of each muscle to the synergy. Bar plots represent the average weights across all recording days. (C, D, G, H) Temporal activation profiles (C) aligned to task events (0%). Dashed lines with shaded tubes indicate the average pre-tendon transfer (TT) electromyography (EMG) profiles of the key contributing muscles (flexor digitorum superficialis [FDS], extensor digitorum communis [EDC], flexor digitorum profundus [FDP]) for visual comparison with the synergy profile. Symbols and alignment are as described in Figure 6. (I, J) Quantification of spatial stability. Cosine similarity of spatial synergy weights (W) calculated between individual recording days and the pre-TT average. Blue markers indicate pre-TT control days; red markers indicate post-TT days. The horizontal gray shaded region (0.95–1.0) denotes the range of high baseline stability.
Figure 7—figure supplement 1
5 Variance accounted for (VAF).
(A–B) Cumulative VAF plotted against the number of synergies for Monkey A (bottom) and Monkey B (top). Blue lines: original data; red lines: shuffled data. The black horizontal line indicates the 80% threshold used to determine the number of synergies.
Figure 7—figure supplement 2
Synergy weights of all experimental sessions.
(A–D, Monkey A) Synergy weights for each extracted muscle synergy. Each recording day was plotted individually before (left column) and after tendon transfer (TT) surgery (right column). Red boxes indicate the main contributing muscles to each of the corresponding synergy. (E–F, Monkey B). (A, E) Synergy A, (B, F) Synergy B, (C, G) Synergy C, and (D, H) Synergy D.
Figure 7—figure supplement 3
Time varying activation profiles of muscle synergies.
Comparison of temporal activation profiles (y-axis: amplitude [a.u.]) for muscle synergies (A–D) between the pooled pre-surgery control period (blue traces; mean ± SD) and the final post-surgery recording day (red traces; mean ± SD). (Top row) Monkey A. (Bottom row) Monkey B. To identify specific phases where the recovered motor plan differed from baseline, a point-by-point Wilcoxon rank-sum test was performed (Bonferroni-corrected for multiple comparisons). Black markers at the top of each panel indicate time points where the activation amplitude remained statistically significantly different (p<0.05) from the pre-surgery baseline. The widespread presence of these significant differences, even where temporal shapes appear similar, confirms that the final motor state is a functional approximation (‘good enough’) rather than a perfect restoration of the original control strategy.
Figure 7—figure supplement 4
Quantitative comparison of pre-surgery and final post-surgery synergy activation profiles.
Figure 8
Analysis of secondary muscle synergies.
(A–H) Analysis of secondary Synergies C (wrist flexor) and D (wrist extensor). (A, B, E, F) Spatial synergy weights (W) showing the contribution of each muscle. (C, D, G, H) Temporal activation profiles (C) aligned to task events (0%). Colored traces denote post-surgery landmark days. Layout and symbols are as described in Figure 6. (I, J) Quantification of spatial stability for Synergies C and D. Cosine similarity of spatial weights calculated between individual recording days and the pre-tendon transfer (TT) average. Blue markers indicate pre-TT control days; red markers indicate post-TT days. The horizontal gray shaded region (0.95–1.0) denotes the range of high baseline stability.
Figure 9 with 1 supplement
Cross-correlation analysis of primary synergy activation.
Zero-lag cross-correlation coefficients plotted over post-surgery days for Monkey A (A–D) and Monkey B (E–H). Activation patterns of the primary flexor (Synergy A) and primary extensor (Synergy B) were cross-correlated with pre-tendon transfer (TT) baseline profiles. (Top row: A, B, E, F) Correlations calculated against the pre-TT extensor synergy (Synergy B, blue traces). (Bottom row: C, D, G, H) Correlations calculated against the pre-TT flexor synergy (Synergy A, red traces). (C, G) Correlation of Synergy A with its own pre-TT baseline. (B, F) Correlation of Synergy B with its own pre-TT baseline. (A, E, D, H) Cross-correlations between antagonistic synergies (e.g. A is Synergy A vs. pre-Synergy B). Black dashed lines on the right y-axis indicate behavioral error metrics (gray shading represents SD). The // represents the recovery period. Triangles indicate landmark days.
Figure 9—figure supplement 1
Evolution of time lag at peak cross-correlation between primary synergies.
Quantification of the optimal time lag yielding the maximum cross-correlation between the primary flexor (Synergy A) and extensor (Synergy B) synergy activation profiles. (A–D) Monkey A; (E–H) Monkey B. Plots show the lag between the recorded synergy activation and the pre-tendon transfer (TT) baseline (specific comparisons indicated in titles). Left y-axis: time lag. Blue traces: pre-surgery data; red traces: post-surgery data. Right y-axis (gray dashed traces): behavioral error metrics (off-target reaching duration for Monkey A; contact time for Monkey B). Positive values indicate the synergy activation is delayed relative to the baseline; negative values indicate it is advanced. Note the temporal correspondence between high behavioral cost (peaks in gray traces) and large fluctuations in synergy timing during the early adaptation phase. Shaded envelopes represent standard deviations.
Figure 10
Cross-correlation analysis of secondary synergy activation.
Zero-lag cross-correlation coefficients for the secondary synergies (C and D) plotted over post-surgery days for Monkey A (A–D) and Monkey B (E–H). Activation patterns were cross-correlated with the pre-tendon transfer (TT) profiles of the primary synergies to assess changing affiliations. (Top row: A, B, E, F) Correlations calculated against the pre-TT extensor synergy (Synergy B, blue traces). (Bottom row: C, D, G, H) Correlations calculated against the pre-TT flexor synergy (Synergy A, red traces). (Left columns: A, C, E, G) Synergy C correlations. (Right columns: B, D, F, H) Synergy D correlations. Black dashed lines on the right y-axis indicate behavioral error metrics.
Figure 11
Aggregated and averaged electromyography (aaEMG).
aaEMG activities for the main contributing muscles of each synergy for Monkey A (A–H) and Monkey B (I–P). (A–D, I–L) Time course of aaEMG activity (summed within ±15% task range) plotted over post-surgery days. Black dashed lines on the right y-axis indicate behavioral error metrics. (E–H, M–P) Bar plots showing the mean (± SEM) aaEMG for the pre-tendon transfer (TT) period (‘pre’) and the five selected landmark days. Vertical colored bars on the time-series plots indicate the corresponding landmark days. Asterisks indicate significant difference from the pre-TT control period (*p<0.01, **p<0.001, ***p<0.0001; two-sample t-test with Bonferroni correction).
Figure 12 with 1 supplement
Kinematic analysis of joint angles (Monkey B).
Changes in joint angles (mean of 20 trials ± SD) for each landmark day. (A) Metacarpophalangeal (MCP) joint angle. (B) Wrist joint angle. Asterisks indicate significant difference from pre-tendon transfer (TT) baseline (**p<0.001, ***p<0.0001; ANOVA). (C) Schematic indicating the timing of the kinematic snapshot relative to the task timeline (dotted line; 83 ms before food touch), capturing the hand configuration during the pre-shaping phase.
Figure 12—figure supplement 1
Joint angle measurement.
This figure shows an illustration of joint angle measurement using a still picture of the monkey while performing the task. See Methods section for further details.
Figure 13
Kinematic analysis reveals gradual refinement of the compensatory tenodesis strategy over time in Monkey B.
(A) Each subplot shows the trial-by-trial relationship between wrist angle (x-axis) and metacarpophalangeal (MCP) angle (y-axis) for a single recording day (n=20 trials per day). Points are color-coded based on the day relative to surgery (color bar). Pre-tendon transfer (TT) (day –4), no correlation exists (R2=0.00). Post-surgery, a negative correlation emerges and strengthens over time, peaking around day 56 (R2=0.58), indicating the learned exploitation of the tenodesis effect where wrist flexion predicts finger extension. The tightening of the scatter plots and increase in R2 over weeks provide direct evidence for a gradual motor skill learning process. (B) All data points combined.
Figure 14
A proposed model of multi-timescale adaptation following tendon transfer.
This schematic illustrates the hypothesized interaction between fast and slow adaptive processes driving recovery. The initial tendon transfer triggers a rapid but maladaptive ‘swap’ of motor commands (fast adaptation 1), leading to a maladaptive state. During this phase, two slower processes are hypothesized to occur in parallel: a costly ‘arms race’ within the conflicted synergy (slow process A, red curve) and the gradual development of a functional compensatory strategy (slow process B, green curve). When the ‘arms race’ reaches a threshold of unsustainable cost (dashed blue line), a second fast adaptation (‘switch-back’, 2) is triggered. This allows for the abandonment of the flawed strategy and the adoption of a stable, ‘good enough’ solution, which is now supported by the newly learned compensatory strategy. The gray line represents the observed neural data (e.g. cross-correlation coefficients of electromyography and temporal activation profiles of muscle synergies), which reflects this two-phase process.
Videos
Video 1
Control behavior (Monkey A – pre-tendon transfer [TT]).
This video demonstrates Monkey A’s baseline performance on the reaching task before the crossed TT procedure. It illustrates the typical, coordinated movements exhibited by the monkey in its unaltered state, serving as a control for comparison with post-surgical behavior. Observe the smooth and accurate reaching and grasping motions as the monkey performs the task.
Video 2
29 days post crossed tendon transfer (Monkey A).
This video documents Monkey A’s attempts to perform the reaching task 29 days after undergoing the crossed tendon transfer procedure. At this early-stage post-surgery, the video clearly shows the significant impact of the procedure on the monkey’s motor control. Observe the marked malcoordination in the monkey’s reaching movements. The reaching attempts are furthermore characterized by ‘explorative’ finger movements over the object. The monkey’s reliance on the experimenter for support highlights the difficulty it experiences in performing the task independently.
Video 3
42 days post crossed tendon transfer (Monkey A).
This video shows Monkey A’s progress 42 days post-surgery. While an improvement in motor control is evident compared to the 29-day mark, the monkey still exhibits some residual deficits. Observe the monkey’s attempts to perform the task autonomously. Although it can now perform the task fully without experimenter assistance, it continues to rely on the support of its unaffected arm, suggesting ongoing challenges with coordination and strength.
Video 4
100 days post crossed tendon transfer (Monkey A).
This video demonstrates Monkey A’s performance 100 days after the crossed tendon transfer. At this point, the monkey has achieved substantial recovery and performs the reaching task with near-normal proficiency. Observe the smooth, coordinated movements, and the monkey’s ability to execute the task independently and accurately. This footage showcases the significant recovery of motor function following the procedure.
Video 5
Control behavior (Monkey B – pre-tendon transfer [TT]).
This video establishes the baseline performance for Monkey B prior to the crossed TT procedure. It shows the monkey’s typical, coordinated reaching behavior in its unaltered state, providing a control for comparison with its post-surgical performance. Observe the precision and fluidity of the monkey’s movements as it executes the task.
Video 6
29 days post crossed tendon transfer (Monkey B).
This video documents Monkey B’s performance 29 days after the crossed tendon transfer. Like Monkey A, Monkey B exhibits significant motor deficits. The video clearly demonstrates the malcoordinated reaching, characterized by overshooting the target and bumping into objects.
Video 7
68 days post crossed tendon transfer (Monkey B).
This video shows Monkey B’s recovery 68 days post-surgery. At this time point, Monkey B has achieved a full recovery and performs the reaching task with accuracy and coordination comparable to its pre-surgical baseline. Observe the smooth and efficient movements, demonstrating the successful recovery of motor function.
Tables
Key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Strain (Macaca fuscata) | Macaque subjects | National BioResource Project (NBRP) | N/A | Two purpose-bred male subjects (Monkey A: 7.8 kg; Monkey B: 9.9 kg) |
| Other | AlphaLab SnR system | Alpha Omega Engineering | https://www.alphaomega-eng.com/ | Multi-channel data acquisition system for multi-unit EMG signal recording |
| Other | SONIMAGE MX1 | Konica Minolta, Inc | https://www.konicaminolta.com/jp-ja/index.html | High-resolution ultrasound scanning system for real-time muscle and tendon tracking |
| Other | DS8R | Digitimer | RRID:SCR_024845 | Constant current stimulator used for percutaneous muscle activation |
| Software, algorithm | MATLAB | MathWorks | R2024a; RRID:SCR_001622 | Core computing platform for signal filtering, rectification, cross-correlation, and NMF synergy factorization |
| Software, algorithm | DeepLabCut | Mathis et al., 2018 | RRID:SCR_021391 | Deep learning framework used for markerless 3D tracking of finger and wrist joint coordinates |
| Software, algorithm | Kinovea | Kinovea (https://www.kinovea.org/) | N/A | Video analysis software utilized for precision behavioral event time detection |
Additional files
-
Supplementary file 1
Muscle abbreviations, functions, and synergy groupings.
Table A: Muscle abbreviations, functions, and synergy groupings. This table provides a reference for all muscles recorded in the study. Columns show the muscle abbreviation, full name, primary functional group, and the main synergy (A, B, C, or D) to which each muscle was assigned for Monkey A and Monkey B. (place table here) Bolded muscles (EDC and FDS) were the targets of the tendon transfer surgery. (¹): EMG signal for this muscle was lost post-surgery in Monkey B and was excluded from the synergy analysis. (N/A): Muscle not recorded or included in the analysis for that monkey. Table B: Statistical comparison of pre-surgery vs. final post-surgery synergy profiles. (place table here) (Note: Global amplitude refers to a comparison of the distribution of mean activation values across the entire task cycle. n.s.=not significant).
- https://cdn.elifesciences.org/articles/108684/elife-108684-supp1-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/108684/elife-108684-mdarchecklist1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Multi-timescale neural adaptation underlying long-term musculoskeletal reorganization
eLife 14:RP108684.
https://doi.org/10.7554/eLife.108684.3
