Reorganization of muscle synergies during multidirectional reaching in the horizontal plane with experimental muscle pain.

Reorganization of muscle synergies during multidirectional reaching in the horizontal plane with experimental muscle pain Silvia Muceli, Deborah Falla and Dario Farina

J Neurophysiol 111:1615-1630, 2014. First published 22 January 2014; doi:10.1152/jn.00147.2013 You might find this additional info useful... This article cites 42 articles, 22 of which can be accessed free at: /content/111/8/1615.full.html#ref-list-1 This article has been cited by 1 other HighWire hosted articles Effect of acute noxious stimulation to the leg or back on muscle synergies during walking Wolbert van den Hoorn, Paul W. Hodges, Jaap H. van Dieën and François Hug J Neurophysiol, January 1, 2015; 113 (1): 244-254. [Abstract] [Full Text] [PDF] Updated information and services including high resolution figures, can be found at: /content/111/8/1615.full.html

This information is current as of March 21, 2015.

Journal of Neurophysiology publishes original articles on the function of the nervous system. It is published 12 times a year (monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 2014 by the American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.

Downloaded from on March 21, 2015

Additional material and information about Journal of Neurophysiology can be found at: http://www.the-aps.org/publications/jn

J Neurophysiol 111: 1615–1630, 2014. First published January 22, 2014; doi:10.1152/jn.00147.2013.

Reorganization of muscle synergies during multidirectional reaching in the horizontal plane with experimental muscle pain Silvia Muceli,1,2 Deborah Falla,1,3 and Dario Farina1 1

Department of Neurorehabilitation Engineering, Bernstein Focus Neurotechnology Göttingen, Bernstein Center for Computational Neuroscience, University Medical Center Göttingen, Georg-August University, Göttingen, Germany; 2Center for Sensory-Motor Interaction, Department of Health Science and Technology, Aalborg University, Aalborg, Denmark; and 3 Pain Clinic, Center for Anesthesiology, Emergency and Intensive Care Medicine, University Hospital Göttingen, Göttingen, Germany Submitted 1 March 2013; accepted in final form 20 January 2014

muscle synergies; experimental muscle pain; reaching

induces a reorganization of motor strategies, although this reorganization may be hard to interpret (Hodges and Tucker 2011). Most experimental pain studies have reported decreased activity of the painful muscle (GravenNielsen et al. 2002). Inhibition of the painful muscle has been attributed to a decrease in the central drive that the muscle receives (Graven-Nielsen et al. 2002). However, this effect was task and muscle dependent (Ervilha et al. 2005; Falla et al. 2007). Muscles other than the painful one also change their activity to maintain the motor output, which is minimally altered (Ervilha et al. 2005). Therefore, it has been speculated that, in painful conditions, new synergies arise aimed at min-

IT IS KNOWN THAT MUSCLE PAIN

Address for reprint requests and other correspondence: D. Farina, Dept. of Neurorehabilitation Engineering, Bernstein Focus Neurotechnology (BFNT) Göttingen, Bernstein Center for Computational Neuroscience (BCCN), Univ. Medical Center Göttingen, Georg-August Univ., Von-Siebold-Str. 4, 37075 Göttingen, Germany (e-mail: [email protected]). www.jn.org

imizing the use of the painful muscle (Madeleine et al. 1999). However, most previous studies analyzed only a small group of muscles active during the task and focused on the analysis of each muscle independently. A long-lasting hypothesis is that the central nervous system (CNS) does not control each muscle individually, but adopts strategies that simplify the control by combining few motor modules (Bernstein 1967). Unit burst generators (Grillner 1981), spinal force fields (Giszter et al. 1993), and muscle synergies (Tresch et al. 1999) have been proposed as modular elements. However, it remains unclear how motor control modularity is influenced by pain. The time-invariant muscle synergy model (Tresch et al. 1999), which assumes that the activation pattern of the muscles active in a task can be decomposed as profiles of relative activations across a group of muscles (synergies) and the neural commands that the muscle synergies receive (activation signals), represents a suitable framework to investigate changes in muscle coordination and neural drive determined by muscle pain. The nonnegative matrix factorization (NMF) algorithm (Lee and Seung 2001) has been identified as a suitable tool to factorize muscle activation patterns in synergies and activation signals. In the present study, we exploit the above muscle synergy model to analyze the neural control of muscles during multijoint planar reaching tasks in an experimental model of muscle pain. Pain was experimentally induced by intramuscular injection of hypertonic saline (Kellgren 1938), which elicits the activity of group III and IV muscle afferents without compromising the contractile and electrophysiological proprieties of the muscle fibers (Farina et al. 2004; Graven-Nielsen et al. 2002). Muscular activation in the unperturbed task was previously described by a linear combination of one set of three muscle synergies common to multiple directions (Muceli et al. 2010). In this study, we exploit the time-invariant synergy model to investigate if 1) the reduction of dimensionality in control remains in the presence of muscle pain; and 2) if the baseline and altered-afferent conditions share the same muscle synergies. More specifically, we factorize the electromyographic (EMG) pattern in synergies and activation signals separately for the painful and the unperturbed task, and we compare the synergy structure and activation signals across the two conditions.

0022-3077/14 Copyright © 2014 the American Physiological Society

1615


Muceli S, Falla D, Farina D. Reorganization of muscle synergies during multidirectional reaching in the horizontal plane with experimental muscle pain. J Neurophysiol 111: 1615–1630, 2014. First published January 22, 2014; doi:10.1152/jn.00147.2013.—Muscle pain induces a complex reorganization of the motor strategy which cannot be fully explained by current theories. We tested the hypothesis that the neural control of muscles during reaching in the presence of nociceptive input is determined by a reorganization of muscle synergies with respect to control conditions. Muscle pain was induced by injection of hypertonic saline into the anterior deltoid muscle of eight men. Electromyographic (EMG) signals were recorded from 12 upper limb muscles as subjects performed a reaching task before (baseline) and after the injection of hypertonic (pain) saline, and after the pain sensation vanished. The EMG envelopes were factorized in muscle synergies, and activation signals extracted for each condition. Nociceptive stimulation resulted in a complex muscle reorganization without changes in the kinematic output. The anterior deltoid muscle activity decreased in all subjects while the changes in other muscles were subject specific. Three synergies sufficed to describe the EMG patterns in each condition, suggesting that reaching movements remain modular in the presence of experimental pain. Muscle reorganization in all subjects was accompanied by a change in the activation signals compatible with a change in the central drive to muscles. One, two or three synergies were shared between the baseline and painful conditions, depending on the subject. These results indicate that nociceptive stimulation may induce a reorganization of modular control in reaching. We speculate that such reorganization may be due to the recruitment of synergies specific to the painful condition.

1616

MODULAR MUSCLE CONTROL IN THE PRESENCE OF NOCICEPTIVE INPUT

METHODS

Subjects. Eight healthy men (age, 29.3 ⫾ 5.3 yr) volunteered for the experiment. All subjects were right-hand dominant. None of the subjects reported symptoms of neuromuscular disorders or musculoskeletal pain. Participants provided written, informed consent before participation, and the procedures were approved by the ethics committee of the Region North Jutland, Denmark (N-20080022) and conducted according to the Declaration of Helsinki.

Experimental procedures. The experiment consisted of multidirectional reaching movements in the horizontal plane (Muceli et al. 2010). The subjects sat in front of a table with the right arm supported by a custom-designed manipulandum which allowed flexion and extension of the shoulder and elbow. “Reaching” in the horizontal plane was defined as the task of moving the arm such that the final position of the wrist, which was projected on the table surface by a laser beam, was within a circle of 20-mm diameter (target). Audible cues were given to indicate the start and end of a movement. In an attempt to prevent changes due to learning, the subjects practiced the task until they were accustomed to it prior to the recordings (⬃15 min). The task consisted of reaching 12 targets evenly spaced along a circumference (Fig. 1A). The radius of the circumference was one-half the distance between the olecranon and styloid process of the ulna (⬃15 cm). The starting point was the center of the circle and corresponded to an elbow joint angle of 90° and a shoulder joint angle of 120°. Each target was reached once in a random sequence. The reaching movement had 1-s duration and was followed by 5-s rest at the target position before returning to the center point over 1 s and resting for a further 5 s. Thus the total task duration was 144 s. The task was performed in four conditions: at baseline, immediately after the injection of isotonic and hypertonic saline, and immediately after the painful sensation induced by the saline injection disappeared (Fig. 1B). Experimental muscle pain was induced by injection of 1 ml of sterile hypertonic saline (5.8%) into the right anterior deltoid muscle (DAN), close to the EMG electrode location. The DAN was selected because it has a large contribution in one of the three synergies that characterize the task (See RESULTS). Isotonic saline (1 ml, 0.9%) was used as a control injection in the same muscle. The participants were blinded to each injection and were told that one or both might be painful. The pain intensity profile normally includes a period with increasing pain intensity, a period with relatively high stable pain intensity, and a period with decreasing pain intensity (Falla et al. 2007). The rising phase is faster than the falling phase. Therefore, the experimental session was started immediately after the injection in both control and painful conditions to include the period where pain intensity was relatively stable. To limit the experimental session to this period, each target was reached once in the painful condition (total task duration: 144 s). For consistency, each target was reached once also in the other conditions. After the painful sensation vanished, the subjects executed the reaching task once again (post

A

12 (0°)

9 (270°)

Fig. 1. A: experimental setup. Top view of the manipulandum used to perform multijoint reaching in the horizontal plane, with starting point the center of the circle and targets around the circle. Vel, elbow joint angle; Vsh, shoulder joint angle. B: experimental protocol which includes the baseline, isotonic and hypertonic saline injections, and post conditions. Saline was injected into the right anterior deltoid (DAN) muscle. Each recording set consisted of reaching to the 12 targets depicted in A. The reaching movement had 1-s duration and was followed by 5-s rest at the target position before returning to the center point over 1 s.

3 (90°)

6

Vel

Vsh

B

Isotonic injection

BASELINE RECORDING SET

5 min rest

CONTROL RECORDING SET

Hypertonic injection

5 min rest

J Neurophysiol • doi:10.1152/jn.00147.2013 • www.jn.org

PAIN RECORDING SET

Pain dur

POST RECORDING SET


We hypothesize that muscle pain results in altered muscle activity (Ervilha et al. 2005; Falla et al. 2007). In this study, we compare nonpainful and painful conditions during a two-dimensional reaching task. Previous investigation on the peripheral properties of the neuromuscular system showed that they are not affected by hypertonic saline (Farina et al. 2004, 2005), which implies that changes in muscle activity are due to centrally mediated mechanisms. Therefore, we predict that only the activation and not the structure of muscle synergies will change in the presence of pain. This hypothesis is in agreement with studies in animals and humans that investigated the consistency of motor modules in response to alterations in afferent input. Kargo et al. (2010) showed in a simulation study that deafferentation in frogs alters muscle synergy activation but not their structure. Accordingly, Cheung et al. (2005) found that intact and deafferented frogs share most muscle synergies and activate them with different amplitude and temporal patterns during locomotor behaviors. The same study also showed that some synergies were different in the two conditions but involved the same set of muscles in a different balance, indicating that synergies may be tuned by afferent feedback. A further possibility is that supplementary synergies are activated in the presence of nociceptive input. This hypothesis is suggested by the fact that variations in sensory inflow associated with different human postural configurations may result in recruitment of an additional synergy with maintenance of the synergies utilized during stance (Torres-Oviedo and Ting 2010). Finally, the possibility that modularity is disrupted in the painful condition also has to be considered.


lated at each time sample and low-pass filtered (4th order zero-lag Butterworth digital filter, cut-off frequency 1 Hz) to reduce instrumentation noise. The first derivative of the angle signals was calculated to estimate the angular velocity. To determine whether kinematics were affected by pain, we also analyzed endpoint kinematics using the following metrics: movement duration, maximum speed and its occurrence time, and trajectory length. Movement onset and movement end were identified as the times in which the speed profile exceeded 10% of its maximum (calculated across the four conditions). The interval between movement onset and end provided the movement duration. Maximum speed was expressed as percentage of the maximum speed value during the baseline condition. The occurrence time of the maximum speed was referred to the movement onset and was expressed as percentage of the movement duration. The trajectory length was calculated integrating the speed profile and was expressed as percentage of the maximum path length during the baseline condition. The processing of kinematic variables was performed for the purpose of signal segmentation and to compare them in the different conditions (baseline, control, pain, post). Moreover, the position of the markers in the two acromia was compared across conditions to determine whether subjects change posture since the trunk was not restrained. After every three targets during the reaching task in the painful condition, the subjects were asked to verbally rate their perceived pain intensity using a numerical rating scale (NRS). Participants were asked to rate their pain on a scale from 0 to 10, where 0 represented no pain and 10 represented the worst pain imaginable. The mean of these values was used to quantify the pain perceived during the task. Kinematics and muscle activity analysis. Visual inspection of the trajectories suggested that they were similar across conditions (see RESULTS). To confirm the results of visual inspection, the Friedman ANOVA for repeated measures was applied to compare movement duration, maximum speed and its occurrence time, and trajectory length in the baseline, control, pain, and post conditions (n ⫽ 8 subjects). When ANOVA was significant (P ⬍ 0.05), pairwise comparisons (baseline-control, baseline-pain, baseline-post) were obtained by the Wilcoxon matched pairs test (without post hoc corrections) to ascertain that there were not significant differences of the trajectories across conditions. We assessed the changes in muscle intensity level across the four conditions through the number of occurrence of a decrease of the peak value of the EMG envelope with respect to the value at baseline for each of the 24 movements (12 center-out and 12 out-center). As the peak amplitude of the EMG value varies among repetitions of the same movements, the absence of a trend would correspond to a percentage close to 50%, whereas a considerable difference with respect to this value may be interpreted as indicative of a consistent trend. Muscle synergy extraction. The first aim of the study was to investigate if the muscular activation during the reaching task remains modular with altered afferent input. A second aim was to compare the synergy structure among conditions. Therefore, synergies were extracted separately for each condition. It was assumed that the evolution of the activation of M muscles over time X (M ⫻ K) can be obtained combining a set of N ⬍ M time-invariant synergies S (M ⫻ N) by means of time-variant activation signals P (N ⫻ K), according to the following model: X⫽S·P

(1)

where K is the number of samples. In this study, X was estimated from rectified and low-pass filtered surface EMG signals. For the purpose of synergy extraction, muscle activation patterns (EMG envelopes) were resampled by interpolation to a sampling frequency of 40 Hz. A NMF algorithm (Lee and Seung 2001) was used to extract N muscle synergies from the EMG signals. The synergy matrix S and the activation signal matrix P were initialized with random nonnegative values. The muscular activation pattern was estimated as Xr ⫽ SP.



condition). As motor adaptation might occur while repeating the same task (Shadmehr and Mussa-Ivaldi 1994), the post condition was included in the experimental protocol to verify if eventual changes in the painful condition were due to alteration in the afferent input or to learning. If the effect is related to learning, it is expected that the post condition shows similar characteristics as the pain condition. Conversely, if the effect is associated to the afferent stimulation, the post condition should present similar results as the baseline and control conditions. Subjects were encouraged to execute the task at the same pace in all conditions. Kinematics and surface EMG signals from upper limb muscles were concurrently recorded while the subject executed the task. Pain intensity was also recorded during the isotonic/hypertonic saline conditions, as described below. Data collection and preprocessing. Surface EMG signals were recorded in bipolar derivations with pairs of Ag/AgCl electrodes (Ambu Neuroline 720 01-K/12, Ambu A/S, Ballerup, Denmark) with 22 mm of center-to-center spacing. Prior to electrode placement the skin was shaved, if necessary, and lightly abraded. The EMG signals were amplified with a gain of 2,000 (EMG-USB, OT Bioelettronica, Torino, Italy, cut-off frequencies 10 –750 Hz), sampled at 2,048 Hz, and analog-to-digital converted with 12-bit precision. A reference electrode was placed around the left wrist. Electrodes were positioned according to Hermens et al. (1999) and Muceli et al. (2010) over the following muscles on the right side: brachioradialis, anconeus (ANC), medial head of the biceps brachii (BME), lateral head of the biceps brachii, brachialis, lateral head of the triceps brachii, long head of the triceps brachii, medial deltoid (DME), pectoralis major (PEC), DAN, posterior deltoid, and latissimus dorsi. The EMG signal quality was visually inspected. Signals from PEC for subject 4 and from BME for subject 7 were discarded because their quality degraded during the experiment. The EMG signals were off-line band-pass filtered (4th order zero-lag Butterworth digital filter, pass-band 20 – 400 Hz) to attenuate DC offset, motion artifacts, and high-frequency noise (Hermens et al. 1999). The filtered signals were full-wave rectified and low-pass filtered (4th order zero-lag Butterworth digital filter, cut-off frequency 1 Hz) to obtain the muscle activation patterns (EMG envelopes). As the surface EMG signals recorded from the upper trunk (PEC and latissimus dorsi) were contaminated by electrocardiogram (ECG) artifact, a reference channel measuring the ECG signal was recorded from the left pectoralis muscle and used in a least mean squares adaptive digital transversal filter (50 taps) (Mesin et al. 2008) to eliminate the ECG artifact from the two contaminated channels. Only the time intervals corresponding to movements were analyzed. The center instant of a movement was defined as the time instant of maximum of the sum of the absolute values of the shoulder and elbow angular velocities, as measured from the kinematic data. Each movement was analyzed in an interval ranging from 1 s before to 1 s after the center instant. Both the center-out and out-center movements were analyzed. The peak of the EMG envelope was calculated for each muscle for each of the 12 targets in each of the four conditions (baseline, control, pain, post) to evaluate differences in EMG amplitude due to the saline injection. Reflective spherical markers (18-mm diameter) were placed over the left acromion (LAC), right acromion (RAC), lateral epicondyle of the humerus (LEP), and posterior midpoint between the styloid processes of the radius and ulna (MST). The positions of the markers were tracked with a motion analysis system (Qualisys Track Manager, Qualisys AB, Gothenburg, Sweden) with eight infrared digital video cameras (ProReflex MCU, Qualisys AB, Gothenburg, Sweden). The kinematic data were recorded with a sampling frequency of 240 Hz and synchronized with the EMG recordings for data analysis. The three-dimensional (3D) positions of the reflective markers were projected on the horizontal plane. The joint angles of the shoulder (LAC-RAC-LEP) and elbow (RAC-LEP-MST) were calcu-

1617

1618


reconstruction quality for muscle activity patterns was assessed through the R2 index. The nonnegative reconstruction method was run 10 times with different initial weights for the activation signals, and the solution that provide the best R2 was retained for the analysis. Results are reported as mean ⫾ SD. Assessing synergy similarity. The three indexes (NDP, CPA and R2) used to investigate the sharing of synergies across the baseline, control, painful and post conditions provide complementary information. To correctly interpret their meaning, it is necessary to analyze the synergy extraction problem from a geometric view. The activation of M ⫽ 12 muscles over time X can be seen as a collection of M-dimensional vectors. As the time-invariant synergy model assumes that the evolution of X can be obtained by combining a set of N ⬍ M synergies, the collection of vectors from which X is composed actually lies on a N-dimensional subspace, rather than on an M-dimensional subspace. A vector space (and thus a vector subspace) of dimension N can be generated by combing any set of N linearly independent vectors. Each of this set is said to provide a basis for the vector space. In our study, a basis set of N ⬍ M synergies was extracted separately from each experimental condition. If two bases are constituted by vectors (synergies) with similar components, i.e., whose NDP is close to 1, the subspaces spanned by them are similar. The converse is not true. Two different bases can also span a common subspace of dimension Q ⫽ N (complete overlap) or Q ⬍ N (partial overlap). In such a case, the cosine of the first Q principal angles is close to 1. Therefore, the information conveyed by the CPA is trivial only when all the N synergies extracted from two conditions are similar in structure. Conversely, we cannot draw information about the synergy structure from the CPA. Our analysis also included a direct test of whether or not each synergy matrix extracted from a certain condition was sufficient to describe the other three conditions. Again, a good reconstruction does not imply that the synergies are similar in structure, and a poor reconstruction does not provide direct information on how many synergies are actually shared by two conditions. Activation signals. The synergy model decomposes muscle activation patterns in activation signals and synergies. Studies on modular control of reaching identified muscle synergies as basic elements of neural control consistent across loads, forearm postures and endpoints (d’Avella et al. 2006; Muceli et al. 2010). The changes in shape of the activation waveforms of the individual muscles for the different movement conditions are obtained by tuning the activation signals. The injection of hypertonic saline was followed by a change in muscle activity (see RESULTS). A change in muscle activation pattern might be due to a change in the synergies, a change in the activation signals or both. Since the results showed that not all synergies were shared between the baseline and the painful condition, we further investigated if the activations signals were invariant across conditions. In fact, studies on rhythmic movements, such as locomotion, have identified activation signals as behavioral units which are invariant across walking speeds, loads, and directions (see Lacquaniti et al. 2012 for a review). Therefore, we compared the peak of activation signals corresponding to matched synergies through the peak of the normalized crosscorrelation function (c), which quantifies the similarity in shape of the activation signals. Since the 12 targets were reached in a random order, the activation signals were reordered before the comparison. As the cross-correlation parameter is insensitive to amplitude changes, we also compared the peak values of the activation signals. Similar to other matrix factorization algorithms, the results of NMF present ambiguities of scaling and permutation. Therefore, before the comparison, each synergy was normalized with respect to the level of activity of the most active muscle in that synergy, so that the maximum value in each synergy matrix was equal to 1. The corresponding activation signal was accordingly scaled in amplitude.



In some studies, muscle activation patterns are normalized before synergy extraction, i.e., to the maximal value observed for each muscle across all the recorded trials (e.g., d’Avella et al. 2006; Tresch et al. 1999), whereas in others (e.g., Cheung et al. 2005; Muceli et al. 2010) they are not. Although many muscles are active during planar reaching, some contribute more than others to the task, resulting in higher amplitude of the corresponding EMG signals. This information is lost with prenormalization which makes equal the contribution of all muscles to the variance of the entire data set. However, for those muscles with low activity, prenormalization amplifies the noise as much as the muscular signal, which may be adverse for the synergy extraction. In the present work, we conducted the analysis with both normalized and nonnormalized data. We present more details for the nonnormalized data for consistency with our laboratory’s previous work (Muceli et al. 2010). However, the adoption of a normalization procedure should not have had a major influence on the conclusions, since the results have been obtained by comparison to control conditions (see Comparison between muscle synergies). Cross talk is not accounted for in the synergy model (Eq. 1). However, studies on reaching observed that synergies do not substantially change when eliminating the muscles mostly affected by potential cross talk from the analysis (d’Avella et al. 2006, 2008). Dimensionality analysis. The estimated muscular activation pattern Xr was compared with the recorded pattern X by the multivariate coefficient of determination R2 as in d’Avella et al. (2003, 2006). To minimize the probability of finding local minima, the NMF algorithm was run 10 times with different initial weights for the synergy matrix and the activation signals, and the solution that provided the highest R2 was retained for further analysis. The reconstruction performance was analyzed as a function of the factorization dimension to select the number of synergies needed to describe the tasks. The coefficient of determination was displayed as a function of the number of synergies (range 1 to 6). The optimal dimensionality was associated to the number of synergies corresponding to a change in slope of the association between the coefficient of determination and the number of synergies (d’Avella et al. 2003). To test for the level of performance which was due to chance, random activation patterns were created from exponential distributions with a mean value of 10 and low-pass filtered (4th order zero-lag Butterworth digital filter, cut-off frequency 1 Hz) to match the frequency content of the experimental activation patterns (Muceli et al. 2010). “Synergies” were then extracted from such generated random activation patterns to test if simulated data could be reconstructed with decreased dimensionality. Comparison between muscle synergies. Synergies extracted from the baseline condition data were compared with synergies extracted from the control, the pain, and the post conditions using the similarity indices reported by Cheung et al. (2005, 2009) and by Roh et al. (2011). According to the first index, similarity in the structure of two pairs of synergies was quantified through the normalized dot product (NDP), i.e., the best-matching scalar product of the two vectors divided by the product of the norms. This value ranges between 0 and 1 because vectors of synergies are nonnegative. NDP ⬍ 0.9 was considered as indicative of synergy dissimilarity (Cheung et al. 2005). According to the second index, the degree of overlap between the two subspaces spanned by the synergies extracted from two conditions was assessed by means of the cosine of the principal angles (CPA) between subspaces spanned by them (Golub and Van Loan 1996). This second index indicates if two sets of synergies may generate the same muscle spaces. A threshold of 0.9 was selected for the CPA (Cheung et al. 2009), as for NDP. The dimension of the common subspace was given by the number of CPAs above threshold. To test more directly whether or not synergies were condition specific, the synergy matrix extracted from a certain condition was fixed and used to reconstruct the spatio-temporal muscle activation patterns recorded during a different condition using the nonnegative reconstruction method, as proposed by Muceli et al. (2010). The

MODULAR MUSCLE CONTROL IN THE PRESENCE OF NOCICEPTIVE INPUT RESULTS

activity for all of the subjects characterized by a consistent decrease of the DAN EMG amplitude with pain across subjects. Table 2 displays the number of occurrences of decrease of the EMG amplitude peak value with respect to the baseline for each target and each subject. The high number of occurrences of decrease of DAN activity in the painful condition [151/192, 192 ⫽ 12 targets ⫻ 2 directions (center-out and out-center) ⫻ 8 subjects] is indicative of such trend. Moreover, there was a consistent decrease of the EMG activity of the ANC, although less pronounced (139/192). Figure 4 depicts tuning curves of the EMG envelope peak values averaged across subjects (n ⫽ 8). The decrease in EMG amplitude of the DAN and ANC is evident from the shrinking of the tuning curves. Other muscles also changed their activity, but the direction of change was not consistent across subjects. Therefore, the average tuning curves of these muscles did not change substantially. An example is provided by the long head of the triceps brachii, whose activity increased in subjects 1, 4 and 5, while it decreased in subjects 2, 3 and 7 (Table 2), so that on average it was unchanged. The EMG amplitude of all of the muscles was similar between the baseline and the control or post condition (about even number of occurrences of decrease and increase of EMG activity, Table 2, and similar tuning curves, Fig. 4). In summary, nociceptive stimulation resulted in decreased activation of the painful muscle (DAN) without changes in the kinematic output (Table 1), in agreement with other experimental pain studies on both isometric and dynamic conditions (Farina et al. 2004; Falla et al. 2007; Graven-Nielsen et al. 2002).

12

9

3

Fig. 2. Representative example of endpoint trajectories recorded from subject 1 during the 4 conditions. Kinematics were similar across conditions (see also Table 1).

Baseline Control Pain Post 6 J Neurophysiol • doi:10.1152/jn.00147.2013 • www.jn.org


Motor pattern and kinematics. The hypertonic saline injection elicited a painful sensation (NRS: 3.8 ⫾ 1.2) which lasted 5.3 ⫾ 1.8 min after the completion of the task. Subjects did not report pain during the control condition apart from the pricking sensation due to the needle insertion, which quickly disappeared once the needle was removed. Figure 2 shows an example of endpoint trajectory for subject 1. Statistical analysis on the four endpoint variables showed that the kinematics was similar in the four conditions (P ⬎ 0.05), with only 5 out of 288 [4 variables ⫻ 12 targets ⫻ 2 directions (center-out and out-center) ⫻ 3 comparisons (baseline-control, baseline-pain, baseline-post)] performed, resulting in a significant difference between baseline and control or painful or post conditions (P ⬍ 0.05, Table 1). This result indicates that the subjects were able to complete the reaching task with the same speed and precision, despite the painful sensation during the painful condition. The similarity in kinematics across the four investigated conditions was likely due to the fact that movements were directed by auditory cues which helped the subjects to maintain the established pace. The shoulder position was also similar across conditions, indicating that posture was not altered. Pain induced altered activity of the DAN as well as other muscles. Figure 3 displays an example of the muscle activation patterns recorded from subject 1 while performing the task at baseline and during the painful condition. In this example, the most evident effect of pain is a reduction of DAN EMG amplitude. Pain resulted in a dynamic reorganization of muscle

1619

1620


Table 1. Kinematics variables Target 2

3

4

5

6

7

8

9

10

11

12

0.89 ⫾ 0.15 0.83 ⫾ 0.19 0.88 ⫾ 0.16 0.82 ⫾ 0.22

0.80 ⫾ 0.20 0.76 ⫾ 0.26 0.74 ⫾ 0.21 0.68 ⫾ 0.29

0.76 ⫾ 0.20 0.74 ⫾ 0.23 0.83 ⫾ 0.18 0.68 ⫾ 0.27

0.81 ⫾ 0.20 0.75 ⫾ 0.28 0.98 ⫾ 0.18 0.78 ⫾ 0.26

0.85 ⫾ 0.15 0.89 ⫾ 0.25 0.80 ⫾ 0.24 0.87 ⫾ 0.09

0.86 ⫾ 0.32 0.83 ⫾ 0.39 0.90 ⫾ 0.26 0.91 ⫾ 0.19

0.89 ⫾ 0.20 0.88 ⫾ 0.24 0.93 ⫾ 0.17 0.89 ⫾ 0.16

0.93 ⫾ 0.19 0.85 ⫾ 0.21 0.86 ⫾ 0.11 0.74 ⫾ 0.23

0.82 ⫾ 0.08 0.80 ⫾ 0.20 0.86 ⫾ 0.17 0.81 ⫾ 0.27

0.84 ⫾ 0.23 0.79 ⫾ 0.24 0.90 ⫾ 0.20 0.81 ⫾ 0.20

0.95 ⫾ 0.18 0.93 ⫾ 0.23 0.88 ⫾ 0.34 0.87 ⫾ 0.18

1.02 ⫾ 0.17 0.98 ⫾ 0.15 1.01 ⫾ 0.22 0.87 ⫾ 0.13

0.87 ⫾ 0.13 0.80 ⫾ 0.30 0.88 ⫾ 0.18 0.91 ⫾ 0.14

0.89 ⫾ 0.18 0.93 ⫾ 0.29 0.87 ⫾ 0.25 0.75 ⫾ 0.25*

0.82 ⫾ 0.19 0.91 ⫾ 0.11 0.96 ⫾ 0.12 0.83 ⫾ 0.14

0.83 ⫾ 0.24 0.84 ⫾ 0.37 0.91 ⫾ 0.13 0.77 ⫾ 0.25

0.83 ⫾ 0.34 0.88 ⫾ 0.31 0.84 ⫾ 0.20 0.85 ⫾ 0.20

0.98 ⫾ 0.25 0.79 ⫾ 0.36 0.86 ⫾ 0.24 0.85 ⫾ 0.32

0.94 ⫾ 0.37 0.80 ⫾ 0.34 0.82 ⫾ 0.31 0.83 ⫾ 0.19

0.79 ⫾ 0.25 0.75 ⫾ 0.27 0.76 ⫾ 0.30 0.78 ⫾ 0.19

0.78 ⫾ 0.24 0.80 ⫾ 0.27 0.85 ⫾ 0.17 0.77 ⫾ 0.20

0.84 ⫾ 0.25 0.84 ⫾ 0.28 0.86 ⫾ 0.28 0.75 ⫾ 0.27*

0.92 ⫾ 0.13 0.89 ⫾ 0.13 0.87 ⫾ 0.20 0.89 ⫾ 0.14

0.96 ⫾ 0.14 0.91 ⫾ 0.17 1.00 ⫾ 0.31 0.93 ⫾ 0.13

75.3 ⫾ 10.6 68.3 ⫾ 15.1 69.5 ⫾ 13.2 75.8 ⫾ 18.3

71.0 ⫾ 16.7 70.7 ⫾ 20.6 74.4 ⫾ 22.1 78.8 ⫾ 27.6

82.6 ⫾ 13.8 72.4 ⫾ 24.1 68.9 ⫾ 24.3 78.8 ⫾ 27.5

69.7 ⫾ 15.2 71.8 ⫾ 14.9 65.9 ⫾ 20.0 73.5 ⫾ 16.0

71.5 ⫾ 12.8 64.8 ⫾ 16.4 69.9 ⫾ 20.8 80.0 ⫾ 16.8

65.5 ⫾ 17.1 63.4 ⫾ 18.5 64.8 ⫾ 22.1 69.1 ⫾ 8.5

70.7 ⫾ 20.1 61.4 ⫾ 16.1 66.4 ⫾ 21.1 82.9 ⫾ 29.9

79.6 ⫾ 15.4 72.9 ⫾ 18.0 79.7 ⫾ 20.0 77.3 ⫾ 20.2

77.7 ⫾ 13.3 72.0 ⫾ 20.0 72.8 ⫾ 13.5 75.3 ⫾ 16.9

76.1 ⫾ 16.8 71.0 ⫾ 15.0 68.4 ⫾ 10.3 70.2 ⫾ 17.7

66.6 ⫾ 11.0 64.4 ⫾ 22.4 63.4 ⫾ 24.4 73.2 ⫾ 16.7

70.0 ⫾ 14.2 68.1 ⫾ 11.3 66.7 ⫾ 17.1 75.5 ⫾ 18.3

77.3 ⫾ 17.5 72.6 ⫾ 23.4 71.4 ⫾ 20.3 73.3 ⫾ 16.9

64.8 ⫾ 16.1 67.9 ⫾ 17.9 68.4 ⫾ 21.9 71.0 ⫾ 21.3

81.2 ⫾ 8.7 75.6 ⫾ 14.7 72.2 ⫾ 17.9 82.2 ⫾ 10.3

70.3 ⫾ 15.2 68.5 ⫾ 24.4 73.9 ⫾ 15.9 73.7 ⫾ 20.4

68.8 ⫾ 24.7 67.3 ⫾ 22.3 75.1 ⫾ 17.6 80.2 ⫾ 12.1

61.8 ⫾ 12.4 63.9 ⫾ 21.7 65.2 ⫾ 17.1 65.7 ⫾ 18.1

66.9 ⫾ 21.3 61.9 ⫾ 24.0 68.0 ⫾ 21.9 71.9 ⫾ 13.5

71.8 ⫾ 24.4 74.2 ⫾ 25.9 72.7 ⫾ 29.6 83.5 ⫾ 17.6

70.4 ⫾ 16.2 68.7 ⫾ 25.8 69.3 ⫾ 16.3 77.5 ⫾ 18.4

73.3 ⫾ 20.5 64.7 ⫾ 15.1 65.8 ⫾ 12.1 70.7 ⫾ 18.3

72.6 ⫾ 14.3 74.1 ⫾ 21.0 74.4 ⫾ 21.9 78.9 ⫾ 16.5

71.1 ⫾ 17.8 71.4 ⫾ 18.2 75.8 ⫾ 32.8 89.7 ⫾ 45.0

45.6 ⫾ 19.3 47.4 ⫾ 12.0 51.6 ⫾ 17.7 50.0 ⫾ 12.8

47.9 ⫾ 9.7 46.4 ⫾ 10.4 51.9 ⫾ 10.6 49.7 ⫾ 13.6

47.1 ⫾ 12.4 57.6 ⫾ 9.3 61.2 ⫾ 12.6 53.6 ⫾ 8.6

47.9 ⫾ 18.3 53.0 ⫾ 12.6 54.2 ⫾ 18.1 49.3 ⫾ 8.6

51.0 ⫾ 9.8 44.1 ⫾ 8.1 44.1 ⫾ 13.8 51.0 ⫾ 8.0

48.8 ⫾ 9.5 53.8 ⫾ 11.9 47.8 ⫾ 14.6 49.2 ⫾ 9.6

50.8 ⫾ 17.2 48.4 ⫾ 19.7 53.8 ⫾ 18.1 43.7 ⫾ 16.3

41.3 ⫾ 7.6 49.4 ⫾ 10.9 48.9 ⫾ 11.6 49.2 ⫾ 10.4

46.7 ⫾ 6.3 43.8 ⫾ 6.8 46.6 ⫾ 12.5 46.7 ⫾ 10.7

41.5 ⫾ 13.8 39.4 ⫾ 9.8 38.6 ⫾ 6.2 44.3 ⫾ 13.1

45.6 ⫾ 16.5 47.5 ⫾ 12.9 39.6 ⫾ 14.2 44.5 ⫾ 11.5

43.1 ⫾ 21.0 40.7 ⫾ 11.9 43.5 ⫾ 16.2 45.7 ⫾ 14.7

50.1 ⫾ 10.7 47.3 ⫾ 14.3 55.6 ⫾ 9.7 48.2 ⫾ 12.9

55.1 ⫾ 14.0 43.6 ⫾ 13.3 43.2 ⫾ 17.6 49.4 ⫾ 12.0

51.3 ⫾ 11.9 48.9 ⫾ 12.8 42.6 ⫾ 8.8 52.1 ⫾ 11.7

48.2 ⫾ 9.7 50.6 ⫾ 15.4 49.2 ⫾ 13.5 51.7 ⫾ 10.7

49.2 ⫾ 11.4 44.6 ⫾ 15.0 49.1 ⫾ 3.2 51.3 ⫾ 14.5

55.4 ⫾ 22.5 57.2 ⫾ 10.5 47.3 ⫾ 12.7 51.3 ⫾ 17.8

50.4 ⫾ 12.5 51.6 ⫾ 9.6 44.4 ⫾ 9.3 55.2 ⫾ 16.1

48.0 ⫾ 15.5 59.3 ⫾ 11.7* 59.5 ⫾ 7.5* 52.8 ⫾ 11.5

52.2 ⫾ 6.5 53.7 ⫾ 13.8 52.4 ⫾ 13.9 55.5 ⫾ 12.2

59.4 ⫾ 11.0 57.5 ⫾ 13.9 57.0 ⫾ 14.9 50.6 ⫾ 9.2*

49.9 ⫾ 11.4 57.7 ⫾ 16.9 56.0 ⫾ 14.1 55.6 ⫾ 10.1

50.7 ⫾ 12.5 54.7 ⫾ 7.9 51.4 ⫾ 14.9 56.8 ⫾ 15.1

80.0 ⫾ 17.7 81.3 ⫾ 8.1 87.8 ⫾ 12.7 85.9 ⫾ 11.6

82.6 ⫾ 9.1 85.5 ⫾ 11.3 85.9 ⫾ 12.6 77.4 ⫾ 9.0

73.4 ⫾ 16.0 89.0 ⫾ 6.0 78.4 ⫾ 14.7 80.2 ⫾ 13.2

82.3 ⫾ 8.7 80.3 ⫾ 13.2 86.3 ⫾ 10.6 75.3 ⫾ 10.1

73.3 ⫾ 15.3 79.2 ⫾ 10.8 84.3 ⫾ 8.2 81.3 ⫾ 14.6

80.2 ⫾ 11.2 81.8 ⫾ 11.6 79.0 ⫾ 16.9 83.8 ⫾ 12.3

79.1 ⫾ 10.4 86.4 ⫾ 13.1 85.2 ⫾ 9.3 85.7 ⫾ 7.7

83.0 ⫾ 13.3 79.0 ⫾ 13.8 83.9 ⫾ 13.0 79.7 ⫾ 8.8

75.9 ⫾ 11.4 81.9 ⫾ 13.1 74.3 ⫾ 8.2 82.2 ⫾ 12.1

75.3 ⫾ 12.2 83.3 ⫾ 11.4 87.4 ⫾ 9.7 78.8 ⫾ 9.7

76.5 ⫾ 11.8 75.6 ⫾ 10.5 79.3 ⫾ 14.9 78.3 ⫾ 11.0

79.8 ⫾ 15.7 77.2 ⫾ 7.6 89.2 ⫾ 7.7 74.4 ⫾ 11.7

85.0 ⫾ 16.2 79.3 ⫾ 12.9 81.1 ⫾ 9.1 77.2 ⫾ 12.9

82.6 ⫾ 8.8 81.6 ⫾ 15.5 80.9 ⫾ 12.0 77.8 ⫾ 18.2

82.1 ⫾ 12.2 84.3 ⫾ 9.0 81.3 ⫾ 13.2 79.3 ⫾ 9.7

83.8 ⫾ 10.8 80.4 ⫾ 16.3 76.8 ⫾ 11.3 84.1 ⫾ 14.3

83.3 ⫾ 13.4 93.8 ⫾ 5.4 87.7 ⫾ 8.6 115.8 ⫾ 100.8

89.4 ⫾ 10.3 80.2 ⫾ 8.5 84.7 ⫾ 9.7 90.9 ⫾ 8.3

76.0 ⫾ 11.1 82.3 ⫾ 14.0 82.6 ⫾ 12.0 78.2 ⫾ 11.1

83.1 ⫾ 8.0 90.3 ⫾ 8.9 80.9 ⫾ 13.8 77.8 ⫾ 13.7

82.9 ⫾ 14.9 84.1 ⫾ 13.2 83.3 ⫾ 15.8 80.7 ⫾ 12.5

82.9 ⫾ 11.9 94.0 ⫾ 8.5 86.5 ⫾ 14.0 80.3 ⫾ 12.8

77.1 ⫾ 10.2 76.0 ⫾ 16.2 81.2 ⫾ 13.0 83.2 ⫾ 14.0

76.1 ⫾ 7.2 80.5 ⫾ 11.1 86.4 ⫾ 6.0 77.5 ⫾ 11.4

Movement duration (s), maximum speed (% of the maximum speed in the baseline condition) and its occurrence time (% of the movement duration), and trajectory length (% of the maximum length in the baseline condition) in the baseline (B), control (C), pain (P), and post (A) conditions for each of the 12 center-out and out-center movements (n ⫽ 8 subjects) are shown. *Significant differences with respect to the baseline condition are indicated (P ⬍ 0.05).

Dimensionality. For all examined conditions (baseline, control, pain, and post), the curve of the percent variance accounted for by combinations of one to six synergies showed a change of slope in correspondence of three synergies. Three synergies led to an average R2 across the subjects of 0.95 ⫾ 0.02 (baseline), 0.94 ⫾ 0.03 (control), 0.91 ⫾ 0.03 (pain), and 0.93 ⫾ 0.02 (post). Additional synergies only captured a small amount of variation due to noise. For example, the inclusion of another synergy determined a reconstruction quality improvement of only 0.03 ⫾ 0.02 (baseline), 0.04 ⫾ 0.02 (control), 0.06 ⫾ 0.03 (pain), and 0.04 ⫾ 0.01 (post). Therefore, it was assumed that the dimensionality of the muscular activation patterns was 3 for all conditions. The results from the baseline condition are in agreement with our laboratory’s previous study (Muceli et al. 2010). The observation that the data set recorded during the painful condition could also be described by a low dimensional set of activation patterns indicates that the control dimensionality is reduced also in the presence of upper limb muscle pain. For comparison, random data could be reconstructed with an R2 of 0.38 ⫾ 0.05 using the same number of synergies. The result that simulated data could not be reconstructed with decreased dimensionality lends support to

the fact that a low dimensionality is embedded in the recorded dataset rather than being an artifact of the NMF algorithm. Comparison of muscle synergies. Figure 5 displays the synergies extracted from the baseline condition (first bar, blue) and those extracted from the control (second bar, magenta), painful (third bar, red if shared with the baseline condition, white otherwise), and post conditions (fourth bar, green) for the eight participants (subjects 1– 8). The synergies extracted from the baseline condition were ordered across subjects so that the first one primarily controlled DAN, the second one was characterized by a high contribution of ANC, and the third one involved the activity of muscles similarly involved in shoulder and elbow tasks. For subject 2, the synergy structure was different with respect to subjects 1 and 3– 8, with the ANC activity coupled with the DAN activity in synergy 1 and absent in synergy 2. In this case, the order showed in Fig. 5 was adopted because in synergy 1 DAN is still the main muscle and in synergy 3 DME, lateral head of the triceps brachii and posterior deltoid are the key muscles, whose activity was mainly taken into account by synergy 3 in subjects 1 and 3– 8. The synergies extracted from the control, the pain,



Movement duration, s Center-out B C P A Out-center B C P A Speed peak, % Center-out B C P A Out-center B C P A Time to peak, % Center-out B C P A Out-center B C P A Path length, % Center-out B C P A Out-center B C P A

1


8 (240°) out

in

9 (270°) out in

1621

10 (300°) out in

BIO

ANC

BME

BLA

TLA

TLO

DME

PEC

DAN

DPO

LAT

Baseline

Pain

and the post conditions were ordered so that they best matched with the synergies extracted from the baseline condition. From Fig. 5, it is evident that the subjects presented different changes in the synergies due to pain, and these adjustments could be grouped in three categories: 1) two synergies differed among conditions (subjects 1 and 2); 2) one synergy was condition specific (subjects 3–5); and 3) the three synergies were preserved (subjects 6 – 8). Synergy 3 was not maintained in the presence of pain for subjects 1 and 5. For subjects 1– 4, the synergy 1 was not preserved during the painful condition

(NDP ⬍ 0.9). For synergy 1 the relative weight of the DAN decreased in the painful condition in favor of BME and lateral head of the biceps brachii for subject 1, ANC for subject 2, and DME for subjects 3 and 4. It is worth noting that synergies represent the relative levels of muscle activation. Therefore, although for example for subject 4, the DAN still has the maximal activity in synergy 1, the increase in the weight of DME means that the relative activation between the two muscles is changed. Nevertheless, the total DAN activity results from the sum of the DAN activation level in each of the



Fig. 3. Representative example of muscle activation patterns recorded from subject 1 during the baseline and the pain condition. See METHODS for the 12 monitored muscles. Values were normalized with respect to the maximum value assumed by each of the muscle activation pattern during the baseline condition. Time support was 2 s centered at the time instants of maximum angular velocity. The columns refer to some of the directions of the center-out and out-center movements (8 –10) as illustrated in Fig. 1. Reaching such targets requires high activation of the DAN muscle, whose activity decreased during the painful condition. BIO, brachioradialis; ANC, anconeus; BME, medial head of the biceps brachii; BLA, lateral head of the biceps brachii; BIA, brachialis; TLA, lateral head of the triceps brachii; TLO, long head of the triceps brachii; DME, medial deltoid; PEC, pectoralis major; DPO, posterior deltoid; LAT, latissimus dorsi.

BIA

1622


Table 2. Muscle activity ANC

8 24 11 15 15 13 9 11 106/192 55.2

12 19 16 19 7 18 10 16 117/192 60.9

8 17 16 14 10 16 15 18 114/192 59.4

18 20 20 14 11 21 15 20 139/192 72.4

7 9 12 13 8 17 12 17 95/192 49.5

9 16 16 9 11 15 8 20 104/192 54.2

BME

BLA

BIA

TLA

TLO

DME

PEC

DAN

DPO

LAT

7 17 16 9 13 10

11 12 16 17 16 6 3 15 96/192 50.0

14 15 13 13 8 11 8 12 94/192 49.0

11 17 12 12 6 9 3 14 84/192 43.8

18 16 14 7 5 8 11 12 91/192 47.4

10 18 7 8 6 10 10 13 82/192 42.7

12 14 10

13 14 11 11 9 11 8 8 85/192 44.3

8 18 11 12 16 14 5 14 98/192 51.0

5 10 24 17 9 8 14 13 100/192 52.1

12 24 17 9 12 7 12 17 110/192 57.3

11 11 20 10 8 17 8 19 104/192 54.2

7 22 16 6 9 15 13 19 107/192 55.7

6 24 22 0 4 12 18 11 97/192 50.5

15 22 13 1 14 23 23 9 120/192 62.5

24 24 15 13 11 22 24 18 151/192 78.6

9 24 20 3 21 16 22 15 130/192 67.7

19 8 24 12 9 3 13 17 105/192 54.7

16 19 17 13 12 6 13 19 115/192 59.9

10 11 10 10 5 10 19 18 93/192 48.4

11 18 12 8 4 11 10 17 91/192 47.4

10 18 9 8 1 8 18 12 84/192 43.8

10 19 10 14 12 18 23 12 118/192 61.5

22 24 19 3 9 7 24 18 126/192 65.6

6 17 5 9 15 16 21 16 105/192 54.7

19 19 24 17 5 5 22 15 126/192 65.6

14 86/168 51.2 4 24 15 5 4 7 14 73/168 43.5 10 19 16 11 7 11 17 91/168 54.2

3 17 10 15 81/168 48.2 17 18 18 6 6 18 0 83/168 49.4 8 16 10 5 10 8 20 77/168 45.8

Values are no. of occurrences of a decrease of the EMG envelope peak during the control, the painful, and the post conditions with respect to the baseline condition for all subjects [192 ⫽ 12 targets ⫻ 2 directions (center-out and out-center) ⫻ 8 subjects, 168 ⫽ 12 targets ⫻ 2 directions (center-out and out-center) ⫻ 7 subjects]. All of the changes were considered either as an increase or a decrease with respect to baseline; therefore, a decrease of ⬃50% implies an equal number of occurrences of an increase and decrease of the EMG envelope peak. BIO, brachioradialis; ANC, anconeus; BME, medial head of the biceps brachii; BLA, lateral head of the biceps brachii; BIA, brachialis; TLA, lateral head of the triceps brachii; TLO, long head of the triceps brachii; DME, medial deltoid; PEC, pectoralis major; DAN, right anterior deltoid muscle; DPO, posterior deltoid; LAT, latissimus dorsi.

three synergies weighted by the corresponding activation signals. Therefore, the synergy weights themselves should not be considered as indicative of the absolute level of muscle activity. Subject 1 preserved only the baseline synergy 2, and subject 2 only synergy 3 across the pain condition. For subjects 1–5, the CPA of the first two angles was ⬎0.9, and the CPA of the third angle was ⬍0.9 (indicating that the two synergy sets spanned a common two-dimensional, and not 3D, subspace), while the three CPAs resulted in ⬎0.9 for subjects 6 – 8 (indicating that the two synergy sets spanned the same complete 3D subspace, Fig. 6). On the other hand, the control, baseline, and post conditions always shared all three synergies. Only for subject 3, the CPA of the third angle between control and baseline conditions (0.88) was slightly below the imposed threshold, but the synergies were similar as indicated by the NDPs and the CPA was anyway very close to the threshold. In summary, from the analysis of similarity indexes (Fig. 6), the injection of hypertonic saline had a major influence on synergy 1 (average NDP across n ⫽ 8 subjects 0.75, average CPA for the third angle 0.71). The large standard deviation is due to the fact that four out of eight subjects (1– 4) exhibited such behavior, as observed in Fig. 5. The analysis also showed that the synergy 3 was shared between the baseline and the painful condition in six out of eight subjects. This implies that

the coupling that naturally occurs between the elbow and shoulder joints during reaching remained invariant in the presence of noxious stimulation. In all except one subject, the synergy 2 was also preserved with pain, as expected, since the pain experienced by the subjects was localized to the proximal region of the upper limb, while the synergy 2 had a maximal activation direction opposite the direction of maximal activation of DAN. The synergy 1, however, was not maintained during the painful condition in four out of the eight subjects. This was due to a reorganization of shoulder muscle activity resulting from the decrease in activity of the DAN muscle following nociceptive stimulation. On the other hand, the injection of isotonic saline did not have any influence on the recruited synergies, which maintained the same structure (average NDP across n ⫽ 8 subjects ⬎ 0.97 for each of the three synergies) and could thus span the same subspace as the synergies extracted from the baseline condition (average CPA, n ⫽ 8 subjects, across baseline and control conditions ⬎ 0.96 for each of the three angles). The fact that no change in EMG responses or synergies (Figs. 4 and 5) was observed during the control condition with respect to the baseline condition confirms that any changes observed during the painful injection were due to the painful stimulation (the volume of isotonic and hypertonic saline, as well as the



B/C Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 Subject 6 Subject 7 Subject 8 Total Percentage B/P Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 Subject 6 Subject 7 Subject 8 Total Percentage B/A Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 Subject 6 Subject 7 Subject 8 Total Percentage

BIO


BIO

ANC

BME

BLA

BIA

TLA

TLO

DME

PEC

DAN

DPO

LAT

Control (out)

Pain (out)

Post (out)

BIO

ANC

BME

BLA

BIA

TLA

TLO

DME

PEC

DAN

DPO

LAT

Baseline (in)

Control (in)

Pain (in)

injection procedures were identical). In the post condition, the synergies were similar to the baseline condition (average NDP across n ⫽ 8 subjects ⬎ 0.96 for each of the three synergies; average CPA ⬎ 0.96 for each of the three angles). This result indicates that the synergies did not change throughout the experiment due to learning and returned to baseline values soon after pain passed. When the analysis was repeated on normalized data, conclusions would not change if the threshold of the CPA was set

Fig. 4. Directional tuning of the EMG envelope peak value during the 4 conditions. See METHODS for the 12 monitored muscles. For each subject and muscle, the EMG envelope was normalized to the maximal value across the four conditions. Peak values were then averaged across subjects (n ⫽ 8). Both center-out and out-center movements were considered in this figure. The “shrinking” of the pain curves of the DAN muscle was due to a consistent decrease of the EMG activity of this muscle across subjects. Others muscle also changed their activity, but the direction of change was different across subjects.

Post (in)

to 0.8 instead of 0.9 (Fig. 6). That means that the data exhibited the same trend, but the numerical values of this particular metric were lower. The average NRS was 3.5 ⫾ 1.4 and 4.4 ⫾ 1.0 for subjects 1–5 and 6 – 8, respectively, with the group who shared all three synergies across conditions experiencing more pain. Motor pattern reconstruction. From the above results, the baseline condition shared all synergies with the control and post conditions and at least one (depending on the subject) with



Baseline (out)

1623


Pain (specific) 1

syn 1 0.5

0.5

0

0

1

1

syn 2 0.5

0.5

0

0

1

1

syn 3 0.5

0.5

0

0

1

syn 1 0.5

0.5

0

0 1 0.5

0

0

1

1

syn 3 0.5

0.5

0

0

1

syn 1 0.5

0.5

0

0

1

1

syn 2 0.5

0.5

0

0

1

1

syn 3 0.5

0.5

0

0

syn 1 0.5

0.5

0

0

1

1

syn 2 0.5

0.5

0

0

LAT

DPO

DAN

PEC

TLO

DME

TLA

BIA

BLA

0 BME

0 BIO

1 0.5 ANC

1

syn 3 0.5

LAT

1

DPO

sub 8

1

DAN

sub 7

PEC

sub 6

1

TLO

sub 5


1

syn 2 0.5

DME

sub 4

1

TLA

sub 3

BIA

sub 2

1

Post

BLA

Pain (shared)

BME

sub 1

Control

BIO

Baseline

ANC

1624

Fig. 5. Synergies underlying the baseline (first bar), control (second), pain (third), and post (fourth) conditions for subjects 1– 8. Synergy 1 is dominated by DAN activity, synergy 2 by ANC, and synergy 3 by muscles acting on both the shoulder and elbow joints. For subjects 1– 4, synergy 1 appears to be condition specific, while subjects 6 – 8 shared all of the 3 synergies in the two conditions. For subjects 1 and 2, only synergies 2 and 3, respectively, appears preserved in the painful condition. Synergies are white when dissimilar to the best matched synergies of the baseline condition.


MODULAR MUSCLE CONTROL IN THE PRESENCE OF NOCICEPTIVE INPUT Non-normalized envelopes

Normalized envelopes

NDP

Baseline - Control

Baseline - Pain 1

1

0.8

0.8

0.8

0.6

0.6

0.6

0.4

0.4

0.4

0.2

0.2

0.2

Syn 1

0

Syn 3

Syn 2

Syn 1

Syn 2

Baseline - Control

CPA

Baseline - Post

1

0

0

Syn 3

1

0.8

0.8

0.8

0.6

0.6

0.6

0.4

0.4

0.4

0.2

0.2

0.2

1D

2D

0

3D

the painful condition. It is thus expected that the muscle activity pattern of either the control or post condition, but not that of the painful condition, can be expressed by a linear combination of the synergies obtained for the baseline condition, and vice versa. Accordingly, the quality of reconstruction depended on the condition from which the synergies were extracted (Fig. 7). In particular, at least for those subjects characterized by different synergies in the baseline and painful conditions, the reconstruction of the baseline (control) datasets Non-normalized envelopes

Syn 3

1D

Fig. 6. Similarity indexes between the synergies extracted from the baseline and the other datasets. Individual values for each subject are color-coded (subjects 1– 8: red, black, green, cyan, brown, orange, magenta, and blue). Mean and SD (n ⫽ 8) for the nonnormalized (square) and normalized (circle) EMG envelopes are shown. Top: normalized dot product (NDP). Bottom: cosine of the principal angles (CPA). Synergies extracted from the prepain and post conditions are similar in structure and span the same threedimensional (3D) subspace, while baseline and pain conditions span a common twodimensional (2D) subspace. Synergy 1, where DAN activity is dominant, appears to be specific of the prepain conditions. This resulted in a bidimensional subspace common to the baseline and pain conditions. Results were similar in trend, irrespective of the prenormalization.

3D

2D

was poorer when performed from the synergies extracted from the painful condition than from either the control (baseline) or post conditions, as expected by the NDP and CPA values reported in Fig. 6. Similarly, the post dataset was better reconstructed from the synergies of either the baseline or control conditions than the ones of the painful condition. On the other hand, similar indexes were found for subjects with synergies shared across the conditions. It is worth noting that the R2 standard deviation was larger when synergies extracted

Normalized envelopes Reconstruction of

Baseline

Control

Pain

Post

1

1

1

1

0.9

0.9

0.9

0.9

0.8

0.8

0.8

0.8

0.7

0.7

0.7

0.7

0.6

0.6

0.6

0.6

0.5

0.5

0.5

0.5

0.4

0.4

0.4

0.4

Fig. 7. Performance index in the reconstruction of muscle activation patterns of one condition (see title) using 3 synergies extracted from the same or another condition (see xaxis labels). As in Fig. 6, individual values for each subject, mean and SD (n ⫽ 8) in case of the nonnormalized (square) and normalized (circle) EMG envelopes, are shown. Synergies extracted from the dataset collected during the pain condition allowed accurate reconstruction of the same dataset only, while synergies extracted from the baseline, control, and post datasets could reconstruct equally well the two prepain and the post datasets, but not the pain dataset.

R2

0.3 from synergies of

Baseline

0.3

Pain

Control

Post

Baseline

0.3

Pain

Control

Post

Baseline

0.3

Pain

Control

Post

Baseline

Pain

Control

Post



0

3D

2D

Syn 2 Baseline - Post

1

1D

Syn 1

Baseline - Pain

1

0

1625

1626


4). Since the synergies were invariant across the baseline and painful conditions for subjects 6 – 8, the observed difference in muscle activation was necessarily due to a change in the activation signals. In subject 6, muscle activity reorganization was mainly due to a variation in the activation signal amplitude (the peak decreased by ⬃14% and 14%, and increased by ⬃39% for the activation signals corresponding to the first, second, and third synergy respectively; the corresponding c values were 0.94, 0.94, and 0.92). In subject 7, c ⫽ 0.94, 0.93 and 0.92, with a peak decrease of 19% and increase of 12% and 10% for the three synergies. In subject 8, the shape of the activation signals was relatively less similar (c ⫽ 0.82, 0.83, and 0.81 for the three activation signals; the peak was reduced by ⬃60%, 40%, and 26%). Although the synergies were similar for the above-mentioned subjects, they were not equal (NDP ⫽ 1). Therefore, in theory, the small difference between the synergies could have influenced the comparison of the activation signals. For this reason, we also compared the activation signals of the baseline data to the activation signals of the dataset recorded during the painful condition using the synergies extracted from the baseline dataset. Information at a glance is provided by the tuning curves shown in Fig. 8 that shows the average results for subjects 6 – 8. The most evident result is a decrease in the activation of synergy 1. For subject 1–5, the activation signals changed as well (c ⫽ 0.81 ⫾ 0.21 average across three activation signals and five subjects), specifically for the case of shared and specific synergies (c ⫽ 0.88 ⫾ 0.05 and 0.73 ⫾ 0.29, respectively). Figure 9 shows activation signals corresponding to subject 4. Notice that the signals scaled in amplitude without much change in the waveform in several painful instances, while there was a substantial reshaping of the waveform in other such instances.

syn 1

syn 2

out

Fig. 8. Directional tuning of the activation signal peak value (average across subjects 6 – 8) extracted from baseline and pain conditions using the synergies of the baseline condition. Both center-out and out-center movements were considered in this figure.

in

Baseline

Pain


syn 3


from the pain dataset were used for the reconstruction. This again reflects a greater variability in performance in the painful condition across subjects. In general, variability of EMG responses for the baseline (painful) dataset was best accounted for by synergies extracted from the same dataset rather than those obtained from the painful (baseline) dataset (Fig. 7). Although the decrease in the quality of reconstruction was significant, the synergies underlying the two conditions could account for a large percentage of variability in muscle activity across the other condition (⬎0.83 across n ⫽ 8). This high value indicates the existence of a large degree of sharing in control between the two conditions. Analogous results were found when comparing the control and post conditions to the painful one. Since DAN was the muscle injected, it might be argued that an inferior performance when reconstructing the pain dataset may be due to the fact that only the DAN EMG pattern was not well reconstructed. Therefore, we also reconstructed the muscular pattern of all muscles apart from the DAN using the synergies extracted from the baseline condition (excluding the weights corresponding to the DAN). Performance was basically unchanged (R2 ⫽ 0.86 ⫾ 0.07), meaning that the activity of other muscles also changed. The baseline, control, and post conditions could be interchangeably reconstructed one from the other without a significant effect on the performance (average R2 ⱖ 0.9 and ⱖ 0.7 across n ⫽ 8 in case of nonnormalized and normalized data, Fig. 7). The trend of the results was the same when considering nonnormalized or normalized envelopes for the data analysis. The absolute values of the R2 was higher in the absence of normalization, because muscles with low activity provided small contribution to the variance of the entire dataset. Activation signals. All subjects demonstrated a change in the pattern of muscle activation during the painful condition (Fig.

MODULAR MUSCLE CONTROL IN THE PRESENCE OF NOCICEPTIVE INPUT 1 (30°)

2 (60°)

out

out

in

3 (90°) in

out

4 (120°) in

out

5 (150°) in

out

6 (180°) in

out

7 (210°) in

out

8 (240°) in

out

9 (270°) in

out

10 (300°) in

out

in

11 (330°) out

in

1627

12 (0°) out

in

activation signal 1

Fig. 9. Activation signals corresponding to synergies 1–3 for subject 4 during the baseline and pain conditions. Each activation signal has been normalized to the maximum across the two conditions. Time support was 2 s centered at the time instants of maximum angular velocity. The columns refer to the directions of the center-out and out-center movements, as illustrated in Fig. 1.

activation signal 2

activation signal 3

Baseline

DISCUSSION

spinal and supraspinal levels. Moreover, fear of pain may lead the subjects to diminish the use of the affected muscle voluntarily. For the remaining subjects, one or two synergies were also common between the two conditions, showing a certain extent of shared control between the two conditions. However, some synergies were specific to the painful condition. The presence of a synergy specific to the painful condition may be interpreted as 1) the activation of a synergy different from that recruited during the baseline planar reaching; or 2) the reorganization of the synergy activated during the normal condition by nociceptive feedback. The first hypothesis is compatible with the observation that deafferentiation does not alter force feedback in frogs (Loeb et al. 1993) and is supported by some studies (Kargo and Giszter 2000; Oliveira et al. 2012; TorresOviedo and Ting 2010), showing that modules from the normal condition plus the addition of new modules are necessary to explain the perturbed condition when afferent input is altered. For example, Kargo and Giszter (2000) observed during hindlimb wiping trajectories in frogs, that cutaneous feedback signaling obstacle collision triggers correction responses that are generated by the addition of a muscle synergy to the basic pattern. In our study, the afferent perturbed condition was explained by the replacement rather than the addition of some modules to the basic pattern. However, it should be noted that, in our experiment, the subjects maintained the same kinematics during the painful condition. Conversely, in the experiment by Kargo and Giszter (2000), the trajectory of the frog ankle deviated from the trajectory of the unimpeded movement when an obstacle was placed on the movement path of the wiping limb. Activation of a subset of the synergies underlying the normal condition and synergies specific to an afferent perturbed condition was also observed by Torres-Oviedo and Ting (2010) and Oliveira et al. (2012), albeit they reported a change in movement kinematics during the perturbed condition. In these studies, afferent feedback was altered by sudden postural perturbations. The first group found that muscle synergies are generally consistent across different biomechanical contexts. Postural configurations extremely different from the normal



This study provides the first analysis of muscle reorganization following nociceptive stimulation in humans within the time-invariant synergy framework. The results show that some synergies are shared in the nonpainful and painful conditions, while others are condition specific. Muscle activity in the presence of pain. The injected muscle (DAN) decreased its activity with pain. Altered activity of muscles other than the one injected with saline (DAN) was also observed, but the direction of change was not always consistent among subjects (Table 2 and Fig. 4). Large variability in the change in muscle activity with experimental and clinical pain is a common observation (e.g., Falla et al. 2011; Hodges and Tucker 2011). Shared and specific synergies. The same synergies of the baseline condition reappeared in the control and post conditions, lending support to the notion that the difference observed in the painful condition was due to a reorganization of modular control in such condition. Comparison of synergies between the baseline and painful conditions showed heterogeneous changes across different subjects (Fig. 5), as also observed for the EMG activity (Table 2). Overall, the results showed that the task could be performed during the painful condition using three synergies. Three of the eight participant subjects shared the three synergies across the baseline and pain conditions. In these cases, muscle activity reorganization could be obtained by a change in the activation signals. Specifically, there was a decrease in the activation of the synergy 1, which is consistent with a decrease in the central drive in the presence of pain, proposed by Graven-Nielsen et al. (2002). The change in muscle activation could be compensated by varying the activation of the other two synergies to maintain the kinematic output unvaried (Fig. 8). The fact that the same synergies were used during the baseline and perturbed tasks, despite the change in EMG activation, supports the notion that the control is modular in both conditions. A change in the drive that the motoneuron pool receives was predictable, as both the afferent and the supraspinal inputs change along with pain. In fact, nociceptive afferents project both at the

Pain

1628


Furthermore, it has to be taken into account that deafferentation affects all muscles involved in a task, whereas pain affected only some. Therefore, pain might require a reorganization only across the involved muscles. In addition, deafferentiation remove all sensory inputs, both nociceptive and somatosensory, while saline injection mainly alters the nociceptive input. It could be argued that the difference in muscle synergies observed in the pain condition might be attributed to adaptation due to the task repetition. In fact, it has been shown that animals who undergo training adapt motor programs either by modifying the weights of some synergies or adjusting the amplitude of synergy activations without changing the balance of muscle activities within the synergies (Kargo and Nitz 2003). However, this possibility can be ruled out in our study as the protocol included a post condition where the synergies were similar to those underlying the baseline dataset. In an attempt to prevent changes due to learning, the subjects indeed practiced the task until they were accustomed to it prior to the recordings. Overall, the fact that not all synergies were shared despite the similarity in kinematics lends support to the fact that a reduction in dimensionality with respect to the number of muscles active in a task does not reflect biomechanical constraints but rather a neural control strategy. Nevertheless, Ivanenko et al. (2003) and Grasso et al. (2004) found that spinal cord injury patients can be trained to produce foot kinematics similar to those of healthy subjects, despite a muscle pattern and loadings of the factor analysis different from the control group and variable across subjects. Functional significance. This study is the first that investigated muscle activity reorganization following local nociceptor stimulation within the muscle synergy framework. In the literature, there are numerous studies on the influence of muscle pain on muscle activity. Most have focused on changes in individual muscles or in a limited set of muscles, rather than a comprehensive assessment of muscles acting during the task investigated. As a result there is a large heterogeneity of results for the effect of pain on motor control, which cannot be fully explained by existing pain theories (Hodges and Tucker 2011). Our analysis supports the concept that groups of muscles modify their activity in concert rather than individually. It is likely that such changes appear inconsistent across different studies which analyzed few muscles, although all results can be unified under the synergy framework. Two major theories have been proposed in the literature to explain motor adaptation to pain, but neither is able to accommodate all varied outcomes observed in studies on pain (Hodges and Tucker 2011). The “vicious cycle” theory proposes that pain leads to an increase in the activity of painful muscles (Roland 1986). This model does not apply to the present work since in our experiment the activity of the DAN decreased. The pain adaptation model instead proposes that pain leads to an inhibition within the painful muscle accompanied by a hyperactivity of the antagonist muscles to minimize the use of the painful muscle (Graven-Nielsen et al. 1997; Lund et al. 1991). Accordingly, discharge rates of the motor units of a muscle injected with saline are reduced (Farina et al. 2004). Our results partly fit with the pain adaptation model in that the activity of the DAN was reduced, but we did not observe a consistent change in the activity of the muscles that



stance required the activation of a task-specific muscle synergy in addition to some of the synergies from the normal stance condition. Oliveira et al. (2012) found that recovery of balance during walking may be achieved using three out of the four synergies used during walking, together with a synergy specific to the perturbed conditions. Chvatal and Ting (2013) also showed muscle synergies specific to either walking or reactive balance during standing. Proprioceptive input is carried by group I and II afferents, while group III and IV afferents are involved in case of nociception. Since nociceptive and proprioceptive inputs are mediated by different pathways (spinothalamic and lemniscal, respectively), pain and somatosensory inputs are not necessarily modulated in a similar way. In fact, limb proprioception seems robust to experimental muscle pain (Matre et al. 2002). However, both the nociceptive sensory fibers (A ␦ and C) and the other sensory fibers (A ␣ and ␤) contact inhibitory interneurons in the spinal cord dorsal horn via collateral fibers. Therefore, the passage of pain is modulated by the interaction between these two types of fibers (gate control theory, Melzack and Wall 1965) and the modulation of the different sensory modalities cannot be completely independent. Therefore, it is conceivable that nociceptive input implies the recruitment of new synergies as it occurs for the somatosensory input. Furthermore, it is not completely clear how the sensory-motor integration between the nociceptive and other sensory inputs exactly occurs at both the spinal and supraspinal levels. The hypothesis that synergies different than those of the baseline condition are recruited during the painful condition could be confirmed by observing the same synergy which was seen during the painful condition in other movements. This approach has been used by Roh et al. (2011), who analyzed modularity in frog movements before and after transection at different levels of the neuraxis. They found that some synergies persisted after transection, and some were specific for a certain behavior in either the pretransection or the posttransection condition, but they appeared as shared synergies for other motor behaviors of the same animal. On the other hand, the possibility of reconfiguring preexisting muscle synergies is supported by other animal studies. For example, Cheung et al. (2005) showed that activities of individual muscles within centrally activated synergies may be fine-tuned by afferent feedback. Hart and Giszter (2004) observed by using independent component analysis that synergies of spinal and brainstem frogs share the presence of the dominant muscles but the synergies of spinal frogs also includes additional muscles with respect to the brain stem animals. A reconfiguration of the synergies would be compatible with the observation that motoneuron excitability may change with pain (Le Pera et al. 2001). However, this would imply that the muscle whose excitability is decreased will decrease its relative weight in all synergies, and vice versa in case of increased excitability. This conjecture is not compatible with our results, as can be seen observing, for example, the synergy structure of subject 1. Synergy 1 showed a decreased activity of the DAN muscle. The relative weight of the same muscle would have decreased also in the other synergies if this was due to a decreased excitability of its motoneuron pool, in contrast with the fact that the contribution of the same muscle to synergy 2 actually increased.


is not straightforward. The saline injection is an acute stimulus to which the system adapts immediately and immediately comes back to the initial condition after the stimulus stops (post-pain condition). Moreover, our subjects had some constraints on how to perform the task (e.g., speed, trajectory). It may be that they would have used another strategy if they were allowed to perform the task freely, and this will have an effect on the strategy they will develop to adapt to pain. ACKNOWLEDGMENTS The authors are sincerely grateful to Professor Andrea d’Avella and Professor Francesco Lacquaniti for useful suggestions and discussions on the manuscript.

GRANTS This work was supported by the European Research Council (ERC) via the ERC Advanced Grant DEMOVE (no. 267888).

DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS Author contributions: S.M., D. Falla, and D. Farina conception and design of research; S.M. and D. Falla performed experiments; S.M. and D. Farina analyzed data; S.M., D. Falla, and D. Farina interpreted results of experiments; S.M. prepared figures; S.M., D. Falla, and D. Farina drafted manuscript; S.M., D. Falla, and D. Farina edited and revised manuscript; S.M., D. Falla, and D. Farina approved final version of manuscript.

REFERENCES Bernstein NA. The Co-Ordination and Regulation of Movements. New York: Pergamon, 1967. Cheung VCK, d’Avella A, Tresch MC, Bizzi E. Central and sensory contributions to the activation and organization of muscle synergies during natural motor behaviors. J Neurosci 25: 6419 – 6434, 2005. Cheung VCK, Piron L, Agostini M, Silvoni S, Turolla A, Bizzi E. Stability of muscle synergies for voluntary actions after cortical stroke in humans. Proc Natl Acad Sci U S A 106: 19563–19568, 2009. Chvatal SA, Ting LH. Common muscle synergies for balance and walking. Front Comput Neurosci 7: 48, 2013. d’Avella A, Fernandez L, Portone A, Lacquaniti F. Modulation of phasic and tonic muscle synergies with reaching direction and speed. J Neurophysiol 100: 1433–1454, 2008. d’Avella A, Portone A, Fernandez L, Lacquaniti F. Control of fast-reaching movements by muscle synergy combinations. J Neurosci 26: 7791–7810, 2006. d’Avella A, Saltiel P, Bizzi E. Combinations of muscle synergies in the construction of a natural motor behavior. Nat Neurosci 6: 300 –308, 2003. Dominici N, Ivanenko YP, Cappellini G, d’Avella A, Mondì V, Cicchese M, Fabiano A, Silei T, Di Paolo A, Giannini C. Locomotor primitives in newborn babies and their development. Science 334: 997–999, 2011. Ervilha UF, Farina D, Arendt-Nielsen L, Graven-Nielsen T. Experimental muscle pain changes motor control strategies in dynamic contractions. Exp Brain Res 164: 215–224, 2005. Falla D, Farina D, Dahl MK, Graven-Nielsen T. Muscle pain induces task-dependent changes in cervical agonist/antagonist activity. J Appl Physiol 102: 601– 609, 2007. Falla D, O’Leary S, Farina D, Jull G. Association between intensity of pain and impairment in onset and activation of the deep cervical flexors in patients with persistent neck pain. Clin J Pain 27: 309 –314, 2011. Farina D, Arendt-Nielsen L, Graven-Nielsen T. Experimental muscle pain decreases voluntary EMG activity but does not affect the muscle potential evoked by transcutaneous electrical stimulation. Clin Neurophysiol 116: 1558 –1565, 2005.



oppose the painful movement. The outcome predicted by the pain adaptation model is that the force production and the range and velocity of movement are often reduced. In our study, we encouraged the subjects to execute the movements with the same speed and precision across conditions. This might explain why the outcome was only partly in agreement with this model. It cannot be excluded that if we did not impose a constraint on the pace, the subjects would have slowed down the movement and maybe used a different strategy. A recent finding that may account for the heterogeneous adaption in motor control to pain observed in different studies is that the activity of one muscle is not homogeneously inhibited or excited but rather redistributed within the same muscle other than between different muscles (Hodges and Tucker 2011). Tucker et al. (2009) observed that the same constant force could be generated with and without pain, with different populations of motor units recruited in the two conditions. These motor unit recruitment strategies could not be appreciated from surface EMG, whose amplitude remained unaltered. However, our results showed a net decrease in the surface EMG activity, suggesting that altered recruitment of motor units was not the mechanism responsible for maintaining unaltered kinematics of the movements in our study. Overall, our results showed a subject-dependent manner of reorganizing the motor control strategies to execute painful movements with kinematics and kinetics matched with the painless counterpart. We propose that the CNS adopts a modular control in both the altered and normal conditions, as suggested by the synergy sharing in subjects 6 – 8. A change to the drive of the motoneuron pool could account for the differences observed in muscle response. The remaining subjects adopted specific synergies in the painful conditions. We speculated that this may be due to the recruitment of a different synergy during pain rather than a tuning of the synergy used during the baseline condition by nociceptive feedback. Synergy organization implies that a group of muscles are activated in a fixed balance; thus pain-induced inhibition of one muscle will necessarily induce changes in the activity of other muscles grouped in the same synergy. Such mechanism is in agreement with the model proposed by Giszter et al. (2010) stating that low-level modules are innate, and new synergies are developed at later stages. In an attempt to minimize the use of the painful muscle, the CNS may use synergies abandoned throughout development or alter existing synergies or create novel synergies to accomplish a specific task during painful stimuli (Giszter et al. 2010). This possibility is consistent with the results of a recent study (Dominici et al. 2011) that showed that locomotor modules are retained and augmented by new modules throughout development. The fact that some subjects use the same muscle synergies in the nonpainful and painful conditions is consistent with the observation that some persons perform a particular task in a more stereotyped manner than others (Moseley and Hodges 2006). The strategy adopted by each individual to cope with pain may be based on personal experience, anthropometrics and posture assumed by the person during a certain task (Hodges 2011). Those individuals with less variable motor programs seem to be those more prone to develop pain as they overuse the same strategy rather than taking advantage of the redundancy of the motor system. It is worth noting that our study dealt with acute experimental pain, and an extension of the results to chronic pain patients

1629

1630

MODULAR MUSCLE CONTROL IN THE PRESENCE OF NOCICEPTIVE INPUT Lee DD, Seung HS. Algorithms for non-negative matrix factorization. Adv Neural Inf Process Syst 13: 556 –562, 2001. Le Pera D, Graven-Nielsen T, Valeriani M, Oliviero A, Di Lazzaro V, Tonali PA, Arendt-Nielsen L. Inhibition of motor system excitability at cortical and spinal level by tonic muscle pain. Clin Neurophysiol 112: 1633–1641, 2001. Loeb EP, Giszter SF, Borghesani P, Bizzi E. Effects of dorsal root cut on the forces evoked by spinal microstimulation in the spinalized frog. Somatosens Mot Res 10: 81–95, 1993. Lund JP, Donga R, Widmer CG, Stohler CS. The pain-adaptation model: a discussion of the relationship between chronic musculoskeletal pain and motor activity. Can J Physiol Pharmacol 69: 683– 694, 1991. Madeleine P, Lundager B, Voigt M, Arendt-Nielsen L. Shoulder muscle co-ordination during chronic and acute experimental neck-shoulder pain. An occupational pain study. Eur J Appl Physiol Occup Physiol 79: 127–140, 1999. Matre D, Arendt-Nielsen L, Knardahl S. Effects of localization and intensity of experimental muscle pain on ankle joint proprioception. Eur J Pain 6: 245–260, 2002. Melzack R, Wall PD. Pain mechanisms: a new theory. Science 150: 971–979, 1965. Mesin L, Boye A, Troiano A, Merletti R, Farina D. Real-time reduction of power line interference in m-channel surface EMG. In: Proceedings of the XVIIth Congress of the International Society of Electrophysiology and Kinesiology (ISEK), June 18 –21, 2008, Niagara Falls, Canada. Amsterdam: Elsevier Science, 2008. Moseley GL, Hodges PW. Reduced variability of postural strategy prevents normalization of motor changes induced by back pain: a risk factor for chronic trouble? Behav Neurosci 120: 474 – 476, 2006. Muceli S, Boye AT, d’Avella A, Farina D. Identifying representative synergy matrices for describing muscular activation patterns during multidirectional reaching in the horizontal plane. J Neurophysiol 103: 1532–1542, 2010. Oliveira ASC, Gizzi L, Kersting UG, Farina D. Modular organization of balance control following perturbations during walking. J Neurophysiol 108: 1895–1906, 2012. Roh J, Cheung VCK, Bizzi E. Modules in the brain stem and spinal cord underlying motor behaviors. J Neurophysiol 106: 1363–1378, 2011. Roland M. A critical review of the evidence for a pain-spasm-pain cycle in spinal disorders. Clin Biomech (Bristol, Avon) 1: 102–109, 1986. Shadmehr R, Mussa-Ivaldi FA. Adaptive representation of dynamics during learning of a motor task. J Neurosci 14: 3208 –3224, 1994. Torres-Oviedo G, Ting LH. Subject-specific muscle synergies in human balance control are consistent across different biomechanical contexts. J Neurophysiol 103: 3084 –3098, 2010. Tresch MC, Saltiel P, Bizzi E. The construction of movement by the spinal cord. Nat Neurosci 2: 162–167, 1999. Tucker K, Butler J, Graven-Nielsen T, Riek S, Hodges P. Motor unit recruitment strategies are altered during deep-tissue pain. J Neurosci 29: 10820 –10826, 2009.



Farina D, Arendt-Nielsen L, Merletti R, Graven-Nielsen T. Effect of experimental muscle pain on motor unit firing rate and conduction velocity. J Neurophysiol 91: 1250 –1259, 2004. Giszter SF, Hart CB, Silfies SP. Spinal cord modularity: evolution, development, and optimization and the possible relevance to low back pain in man. Exp Brain Res 200: 283–306, 2010. Giszter SF, Mussa-Ivaldi FA, Bizzi E. Convergent force fields organized in the frog’s spinal cord. J Neurosci 13: 467– 491, 1993. Golub GH, Van Loan CF. Matrix Computations. Baltimore, MD: Johns Hopkins University Press, 1996. Grasso R, Ivanenko YP, Zago M, Molinari M, Scivoletto G, Castellano V, Macellari V, Lacquaniti F. Distributed plasticity of locomotor pattern generators in spinal cord injured patients. Brain 127: 1019 –1034, 2004. Graven-Nielsen T, Lund H, Arendt-Nielsen L, Danneskiold-Samsøe B, Bliddal H. Inhibition of maximal voluntary contraction force by experimental muscle pain: a centrally mediated mechanism. Muscle Nerve 26: 708 – 712, 2002. Graven-Nielsen T, Svensson P, Arendt-Nielsen L. Effects of experimental muscle pain on muscle activity and co-ordination during static and dynamic motor function. Electroencephalogr Clin Neurophysiol 105: 156 –164, 1997. Grillner S. Control of locomotion in bipeds, tetrapods, and fish. In: Handbook of Physiology. The Nervous System. Motor Control. Bethesda, MD: Am. Physiol. Soc., 1981, sect. 1, vol. II, pt. 2, chapt. 26, p. 1179 –1236. Hart CB, Giszter SF. Modular premotor drives and unit bursts as primitives for frog motor behaviors. J Neurosci 24: 5269 –5282, 2004. Hermens H, Freriks B, Merletti R, Stegeman DF, Blok J, Rau C, Disselhorst-Klug C, Hagg G. European Recommendations for surface electromyography. Results of the SENIAM Project. Enschede, the Netherlands: Roessingh Research and Development, 1999. Hodges PW. Pain and motor control: from the laboratory to rehabilitation. J Electromyogr Kinesiol 21: 220 –228, 2011. Hodges PW, Tucker K. Moving differently in pain: a new theory to explain the adaptation to pain. Pain 152: S90 –S98, 2011. Ivanenko YP, Grasso R, Zago M, Molinari M, Scivoletto G, Castellano V, Macellari V, Lacquaniti F. Temporal components of the motor patterns expressed by the human spinal cord reflect foot kinematics. J Neurophysiol 90: 3555–3565, 2003. Kargo WJ, Giszter SF. Rapid correction of aimed movements by summation of force-field primitives. J Neurosci 20: 409 – 426, 2000. Kargo WJ, Nitz DA. Early skill learning is expressed through selection and tuning of cortically represented muscle synergies. J Neurosci 23: 11255– 11269, 2003. Kargo WJ, Ramakrishnan A, Hart CB, Rome LC, Giszter SF. A simple experimentally based model using proprioceptive regulation of motor primitives captures adjusted trajectory formation in spinal frogs. J Neurophysiol 103: 573–590, 2010. Kellgren J. Referred pains from muscle. Br Med J 1: 325–327, 1938. Lacquaniti F, Ivanenko YP, Zago M. Patterned control of human locomotion. J Physiol 590: 2189 –2199, 2012.

The effect of arm weight support on upper limb muscle synergies during reaching movements.

Reaching control of a full-torso, modelled musculoskeletal robot using muscle synergies emergent under reinforcement learning.

Reaching for the Unreachable: Reorganization of Reaching with Walking.

Muscle synergies evoked by microstimulation are preferentially encoded during behavior.

Sarcomere reorganization in mite muscle.

Segmental muscle vibration modifies muscle activation during reaching in chronic stroke: A pilot study.

Empirical Evaluation of Voluntarily Activatable Muscle Synergies.

Brain Connectivity Associated with Muscle Synergies in Humans.

Recruitment of muscle synergies is associated with endpoint force fluctuations during multi-directional isometric contractions.

Between-muscle differences in the adaptation to experimental pain.

Muscle contributions to frontal plane angular momentum during walking.

Electromyography Data Processing Impacts Muscle Synergies during Gait for Unimpaired Children and Children with Cerebral Palsy.

Motor Cortex Reorganization and Impaired Function in the Transition to Sustained Muscle Pain.

Inter-subject variability of muscle synergies during bench press in power lifters and untrained individuals.

Muscle path length in horizontal strabismus.

Muscle Synergies Facilitate Computational Prediction of Subject-Specific Walking Motions.

Scratch responses in normal cats: hindlimb kinematics and muscle synergies.

Muscle synergies may improve optimization prediction of knee contact forces during walking.

Effect of experimental jaw muscle pain on dynamic bite force during mastication.

Modulation of motor variability related to experimental muscle pain during elbow-flexion contractions.

Strength and muscle activities during the toe-gripping action: comparison of ankle angle in the horizontal plane between the sitting upright and standing positions.

Muscle Synergies of Untrained Subjects during 6 min Maximal Rowing on Slides and Fixed Ergometer.

Experimental jaw muscle pain increases pain scores and jaw movement variability in higher pain catastrophizers.

Muscle synergies: input or output variables for neural control?