Enhanced Muscle Afferent Signals during Motor Learning in Humans Michael Dimitriou  Current Biology  Volume 26, Issue 8, Pages 1062-1068 (April 2016) DOI: 10.1016/j.cub.2016.02.030 Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 1 Methods and Exemplar Single-Subject Responses (A) The subjects used their right hand to perform a classic visuomotor learning task. Hand movements at the wrist controlled the 2D location of a cursor displayed on a monitor. The start position corresponded to a semi-pronated posture of the hand that was unrestrained throughout the task. On each trial of the task, the subjects were instructed to swiftly bring the visual cursor inside a highlighted peripheral target. Hand kinematics, wrist muscle electromyography (EMG), and single afferent impulses from the radial nerve were simultaneously recorded. (B) Movements were performed with and without a 45° counter-clockwise rotation of cursor direction. (C) Top row: the hand trajectories over the entirety of the visuomotor learning task (96 trials) performed by a single subject. Bottom row: the initial directional error (IDE) associated with each trial. (D) Calculated Extensor Carpi Radialis (ECR) muscle length and responses from a primary (type Ia) spindle afferent from this muscle during four exemplary trials from (C) (signified by thicker trajectory lines in top row and black circles in the second row of panels). In addition to any burst of afferent impulses early during movement (in response to muscle stretch), a tonic increase in firing rate was visible at the early-exposure and washout stage where adaptation (exponential decay of IDE) was evident, compared to when there was no adaptation (i.e., at the baseline and late-exposure stages). Current Biology 2016 26, 1062-1068DOI: (10.1016/j.cub.2016.02.030) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 2 Visuomotor Adaptation across Subjects The mean initial directional error (IDE) across individuals with whom a single spindle afferent was recorded. Shaded colored areas indicate ±1 SEM. Exponential curves could significantly account for most of the progression of mean IDE across trials in the early-exposure and washout stages of the task, but no such fit was possible in the baseline and late-exposure stages of the task. Current Biology 2016 26, 1062-1068DOI: (10.1016/j.cub.2016.02.030) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 3 Factors Influencing Spindle Afferent Output in the Visuomotor Learning Task (A) Normalized regression coefficients (beta) averaged across afferents (n = 12; one regression performed per afferent). In addition to the continuous variables representing muscle length, its derivatives and ECR EMG (known predictors of muscle spindle output), a categorical/dummy variable representing adaptation state was also used as a predictor (i.e., baseline and late-exposure stages assigned a value of 0, whereas early-exposure and washout stages were assigned a value of 1). The data of continuous variables used in the regression analyses were generated directly from raw data (of the kind displayed in Figure 1D) using a moving-average window of 50 ms, with sampling covering a period 100 ms before the onset of movement until 100 ms after movement cessation. The schematic (inset) below indicates the movement directions included in the regression analyses (six of eight; correspond to substantial lengthening or shortening of the ECR muscle). (B) Normalized (z-transformed) afferent firing rates, collapsed across movement directions and task stages (baseline and late exposure versus early exposure and washout; see also Figure S1). Individual lines represent relevant single afferent data. (C) There was no effect of adaptation state on normalized EMG and kinematic variables, indicating that the effect of adaptation state on afferent signals was not epiphenomenal to differences in kinematics or skeletomotor activity in the spindle-bearing ECR muscle. Throughout, error bars represent 95% confidence intervals. See also Figure S1. Current Biology 2016 26, 1062-1068DOI: (10.1016/j.cub.2016.02.030) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 4 Afferent Population Responses and Spindle Encoding across Task Stages (A) Spindle afferent population responses (mean across afferents), associated kinematics, and ECR EMG at the four stages of the task (i.e., initial block of trials in each stage). The data are aligned on initial peak velocity (time 0). Red and blue arrows indicate the onset and direction of presumed fusimotor effects on spindle output. The blue arrow in parenthesis indicates a possible increase in static fusimotor drive that can cause a decrease in afferent response to stretch (see also B). Error bars are omitted in (A) for visual clarity, but variances across afferents at the six periods of interest (gray background bars) are contrasted in (B) (firing rates) and Figure S2 (kinematics and EMG). (B) Mean firing rate and 95% confidence intervals across afferents for the latter three stages of the task, after subtracting the associated baseline firing rate at the level of single afferents. Single afferent data points used in these analyses represent mean firing rates over each 50-ms period. Asterisks indicate significance (p < 0.05) following one-sample t tests. (C) Data points represent eight single type Ia afferents (each recorded from a different subject) during the early-exposure stage of the task. (D) Adaptation rate in early exposure (taken directly from the exponential fit of IDE per subject) versus the above-baseline firing of all 12 afferents in periods 4 and 6 (i.e., average across late stretch and late shortening). (E) Left panel: afferent population firing rates as seen in (A) (i.e., ±250 ms; down-sampled using a 50-ms moving average window) could represent increases in muscle length in the washout stage of the task (r = 0.87, p = 0.002) but not in the other three stages (p > 0.05). Right panel: as the left panel but referring to muscle shortening movements (blue dots: r = 0.97, p < 10−5). The apparent control of spindle sensors during washout therefore induced linearization of their afferent signal with respect to muscle length. As expected, the linearization effect came at the cost of velocity encoding: in contrast to the other stages of the task (i.e., all p < 0.05), there was no significant relationship between afferent firing rates and muscle velocity in the washout stage (muscle lengthening: r = 0.39, p = 0.27; muscle shortening: r = 0.32, p = 0.36). See also Figures S2–S4. Current Biology 2016 26, 1062-1068DOI: (10.1016/j.cub.2016.02.030) Copyright © 2016 Elsevier Ltd Terms and Conditions