| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Anatomy and Cell Biology, Canadian Institute of Health Research Group in Sensory-Motor Systems, Centre for Neuroscience Studies, Queen's University, Kingston, Canada
Submitted 1 February 2006; accepted in final form 16 September 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Scott 2004; Todorov 2000). This complexity is exemplified by the dissociation of a muscle's function and anatomical action (Buchanan et al. 1989; Hoffman and Strick 1999; Jongen et al. 1989; Nozaki et al. 2005; van Zuylen et al. 1988). For example, the elbow extensor triceps brachii lacks any mechanical action at the wrist but is recruited during wrist supination to counter the elbow flexor torque of biceps brachii, an elbow flexor and wrist supinator (Buchanan et al. 1989). Although wrist supination could be produced using only single-joint muscles (supinators) the nervous system uses all muscles with supinating action, even those acting at multiple joints or having multiple degrees of freedom at a single joint. A similar strategy is predicted by models that optimize a global measure of muscle activity (Collins et al.1995; Crowninshield and Brand 1981; Dul et al. 1984; Fagg et al. 2002; van Bolhuis and Gielen 1999).
We recently observed a similar dissociation of muscle function and action in an animal model of postural control (Kurtzer et al. 2006). Monkeys trained to maintain a constant arm posture recruited their shoulder monoarticulars when countering elbow loads and elbow monoarticulars when countering shoulder loads. These single-joint muscles were most active when producing shoulder-extension/elbow-flexion torque and shoulder-flexion/elbow-extension torque consistent with the optimization of muscle activity.
Inferring the function of muscular activity during movement is considerably more complicated. The limb's intersegmental dynamics allow single-joint motion to arise from multijoint torque and multijoint motion to arise from single-joint torque (Graham et al. 2003a; Hollerbach and Flash 1982; Zajac and Gordon 1989). Moreover, a muscle's ability to generate force strongly depends on its length and velocity (Joyce et al. 1969; Scott et al. 1996). These complexities can be addressed by calculating the inverse dynamics associated with movement and using a muscle model. Another approach is to experimentally enforce similar movement patterns under different load conditions (van Bolhuis et al. 1998). Thereby, one can "factor out" the complex transformation of muscle activity to joint torque and relate the difference in activity specifically to the difference in load between conditions, i.e., torque-related activity.
The present study used a loaded-reach paradigm to examine muscular activity during center-out planar reaching. Similar reaching patterns were enforced while a robotic exoskeleton applied mechanical loads at the shoulder, elbow, both, or neither joint. During the unloaded condition, we observed that upper limb muscles were primarily activated during movements either toward the body or away from the body (i.e., the fore–aft axis). This strong bias was not observed in earlier studies that used handheld manipulanda and less constrained arm movements, but is prominent during the more controlled conditions we used. Comparing muscular activity across the load conditions revealed a bias in torque-related activity toward shoulder-extension/elbow-flexion torque and shoulder-flexion/elbow-extension torque. Although dissociations between muscular function and anatomy have been observed during posture (Buchanan et al. 1989; Jongen et al. 1989; Nozaki et al. 2005) and movement tasks (Hoffman and Strick 1999) this is the first movement task that experimentally controlled for the activity during unloaded reaching. Moreover, we reproduced these reach-related and torque-related biases with a model that optimized a global measure of muscle activity.
Such data are not only valuable for understanding how muscles are used for motor function, but also provide a foundation for interpreting how regions like primary motor cortex (M1) are involved in controlling movement. M1 is the principal cortical region supporting voluntary motor execution and has a strong association with the motor periphery in its somatosensory input and segmental output (Hepp-Reymond 1988; Porter and Lemon 1993; Scott 2003). In fact, here we report that arm-related M1 neurons recorded from the same monkeys exhibited both a reach-related bias and a torque-related bias similar to those of the arm muscles. Thus we provide a link between behavioral goals, neural function, and the musculoskeletal apparatus during planar reaching.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Four male rhesus monkeys (Macaca mulatta, 6–12 kg) performed upper limb motor tasks according to guidelines of the Queen's University Animal Care Committee. The monkey's right forelimb was supported by a robotic exoskeleton (KINARM; BKIN Technologies, Kingston, Ontario, Canada) that permitted horizontal shoulder and elbow motion, monitored those joint motions, and could apply mechanical loads to the two joints (Scott 1999). This mechanical system was coupled to a virtual reality display so that monkeys viewed their entire limb and visual targets in the horizontal plane of movement.
The task involved center-out movements of the monkey's right forelimb (Fig. 1A). Hand movements were initiated near the center of the hand's workspace (about 30° shoulder angle and 90° elbow angle) where the passive stiffness of the limb was relatively small (Graham et al. 2003b). Peripheral targets were 1.2 cm wide and spaced 6.0 cm from the central target. The different monkeys were exposed to different targets as follows. Monkeys A and B reached to eight targets distributed roughly uniformly in joint-torque space, but unevenly in hand space: 45, 67.5, 90, 180, 247.5, 270, 315, and 337.5°, where 0° is to the right and positive rotation is anticlockwise. In contrast, monkeys C and D reached to eight targets distributed uniformly in hand space: 0, 45, 90, 135, 180, 225, 270, and 315°.
|
Targets were presented randomly within a block of the same load condition. This pseudorandom block design allowed stable performance after the first reach to each target that was excluded from the analysis. Blocks were repeated six times (monkeys A, B, and D) or ten times (monkey C) for a total of 48 or 20 trials. Further, similar hand trajectories were implicitly enforced across the different load conditions by requiring temporal and spatial accuracy. The monkey's hand had to move between the start and peripheral radii within 220–350 ms. Note that that hand's acceleration/deceleration within these target radii led to total movement times within 500–600 ms.
Muscle recording
Electromyographic (EMG) activity was recorded with both acute and chronic techniques (Kurtzer et al. 2006; Loeb and Gans 1986; Scott and Kalaska 1997). Acute recordings (monkeys A–D) involved inserting two single-strand wires percutaneously into the muscle belly followed by microstimulation to verify the electrode placement. Chronic recordings (monkeys A and C) involved surgically implanting bipolar multistrand electrodes in the superficial portion of the muscle belly. For both techniques EMG signals were recorded (1,000 Hz for Monkey A; 4,000 Hz for monkeys B–D), full-wave rectified, and integrated into 5-ms bins.
EMG was collected from 11 different muscles. Each muscle could be sampled multiple times (maximum = 6) given the two monkeys with acute recordings only and two monkeys with both acute and chronic recordings. To be selected a muscle recording needed to score 
3 on a subjective rating of signal quality between 1 (poor) and 5 (excellent) and also exhibit directionally tuned muscle activity during unloaded reaching (bootstrap, P < 0.01). Directionally tuned activity was determined with the plate method, a nonparametric technique for characterizing the "mass distribution" of activity (Gribble and Scott 2002a).
The 35 samples that met this criterion are shown in Table 1. Only four sample muscles had scores 3 but lacked significant directional tuning. These few samples—three biarticulars and one monoarticular—exhibited phasic modulation during the reaching period along with large modulation with the final limb position; the complex interaction was not captured by a simple measure of directional tuning.
|
Standard techniques were used for extracellular recordings in the shoulder/elbow region of M1 (Scott and Kalaska 1997; Scott et al. 2001). A general analysis of these data was previously reported (Gribble and Scott 2002b; Scott et al. 2001). Here we restrict our analyses to neurons that exhibited directionally tuned activity during unloaded reaching (bootstrap, P < 0.01) and were recorded during a load condition; 95 neurons from all four monkeys met the criteria.
Data analysis
HAND SPACE. Analyses focused on the period between movement onset (10% of the peak hand velocity) and peak hand velocity. Muscle activity was quantified as the mean rectified voltage from movement onset to peak hand velocity. Likewise, joint torque, joint position, and joint velocity were quantified as their mean values from movement onset to peak hand velocity. We examined M1 activity from 75 ms before movement onset to 75 ms before peak hand velocity to approximate the delay between M1 and muscular activity (Cheney and Fetz 1988; Morrow and Miller 2003). Neural activity was quantified as the mean spike rate in this time window.
We examined two aspects of hand kinematics–the peak tangential velocity of the hand and the direction of movement at peak velocity, to confirm similar movement patterns across load conditions. Movement similarity was tested between the loaded and unloaded conditions with one-way ANOVAs (P < 0.01) for each session.
Several analyses focused on the muscle and neural activity related to the hand motion during unloaded reaching, i.e., reach-related activity. The plate method determined the movement direction with maximal activity, i.e., the preferred hand direction (PHD). We examined the PHD for each sampled muscle and the average PHD of the six relevant muscle groups: shoulder flexors, shoulder extensors, elbow flexors, elbow extensors, biarticular flexors, and biarticular extensors. Note that this working definition of muscle group includes muscles with a similar mechanical action (Table 1).
We also determined whether the global trend of all muscular (or neural) PHDs was best described as one, two, or four symmetrically distributed clusters (Rayleigh test; Baschelet 1981). To test for a unimodal, bimodal, and quadrimodal distribution we multiplied the PHDs by 1, 2, or 4. Next, we summed the PHDs tip to tail and normalized by their absolute summed length. This transformation produces a mean vector whose length ranged from 0 to 1 for perfectly random to perfectly similar PHDs, i.e., a circular correlation. The orientation of this vector is the mean orientation of the distribution. Note that the resulting mean vector is not identical to the mode vector so that smaller subclusters remain unidentified. We used this same technique to examine the bimodal distribution [preferred hand axis (PHA)] of opposing kinematic/kinetic variables such as elbow flexor and extensor torque.
The PHD distributions of muscles and neurons were compared using a random bootstrap with replacement (n = 1,000). By repeatedly sampling each distribution's mean vector we estimated its 95% confidence interval. If the top or bottom 2.5% of the resampled means did not contain the "observed" mean of the other distribution then the two distributions were considered significantly different at P < 0.05.
Another test examined the relative increase (Max
) and decrease (Min) from the prereach activity during unloaded reaching
 |
TORQUE SPACE.
A second set of analyses focused on the change in (muscular and neural) activity with respect to the change in joint torque between the unloaded and loaded conditions (i.e., torque-related activity). Joint torques at the shoulder and elbow were calculated using the equations of motion for a planar two-segment arm including the robotic exoskeleton (Scott 1999). Changes in shoulder and elbow joint torque were subsequently calculated as the difference in joint torque between loaded and unloaded conditions.
Torque-related activity was assessed by a planar regression (F-statistic, P < 0.01). This method can readily handle the nonuniform distribution of torque direction and torque magnitude that results from the nonuniform relation between hand motion, joint velocity, and joint torque (Graham et al. 2003b). The shoulder (S) and elbow (E) slope coefficients from the plane fit describe the relative sensitivity to shoulder and elbow joint torque–preferred torque direction, atan (E/S) (Cabel et al. 2001; Kurtzer et al. 2005). Preferred torque direction (PTD) was measured counterclockwise from shoulder flexion so shoulder flexor, elbow flexor, shoulder extensor, and elbow extensor PTDs occur at 0, 90, 180, and 270°, respectively.
Applying a separate planar regression to each repeat block assessed intertrial reliability of the PTDs. The circular correlation and number of blocks together define a 95% angular confidence interval about the measured PTD (Baschelet 1981). The measured PTD was compared with the anatomical action of a muscle as determined from a previous study (Graham et al. 2003a). If the anatomical action was outside the confidence interval, then the measured muscle PTD is considered significantly deviated from expectation.
Three additional statistics examined trends in torque-related activity: 1) A Rayleigh's test determined whether the entire sample of PTDs had a significant unimodal, bimodal, or quadrimodal distribution (Baschelet 1981). 2) A loaded excitation–inhibition score (loaded-EI score) determined whether load-related changes were primarily an increase or decrease from the activity during unloaded reaching. The different sets of loaded-EI scores were compared with a Wilcoxon rank-sum test. 3) Bootstrap tests compared the PTDs of muscles and neurons.
Models of muscle activation
We examined several different models of muscle activation. These models included two parts: muscle properties and recruitment strategy. Muscle properties included the muscle's morphometry and intrinsic mechanics. We used a lumped representation for the six muscle groups using known values of the monkey's moment arms, fascicle lengths (l), and physiological cross-sectional area (PSCA) (Cheng and Scott 2000; Graham and Scott 2003b). Each lumped muscle group was represented four times, one for each monkey.
Muscle properties included passive stiffness and velocity-dependent force production. Stiffness was estimated as a one-sided cubic function (Graham et al. 2003a). Damping was estimated as a dissipating exponential for muscle shortening and saturating exponential for muscle lengthening (Scott et al. 1996). The mean joint kinematics for each condition and each monkey determined the length and velocity of the muscle fascicles of each muscle group.
The recruitment strategy determined how muscles were activated to achieve a target torque. The target torques for all models was the mean torque from movement onset to peak hand velocity measured for each condition and each monkey. We examined two basic recruitment strategies: "direct activation" and "optimal activation." The "direct activation" strategy recruited a muscle proportional to the angular similarity (dot-product) of the target torque and the muscle's moment arm. This strategy does not account for any interactions among muscles; i.e., it will not achieve the target torque when all muscles are activated. It is presented for didactic purposes.
The strategy of "optimal activation" achieved each target torque while minimizing a global measure of muscle activity; i.e., it accounts for interactions among muscles. We focused on one measure of muscle activity, the sum of muscle stress squared {[sumn](fi/PCSAi)2}. "Optimal activation" was determined by an iterative procedure previously used for our posture task (Kurtzer et al. 2006). The procedure achieved each target torque under the constraint that muscles could only pull and the muscle parameters were constant (fmincon function, MATLAB; The MathWorks, Natick, MA). We analyzed the simulated activity of the models with the same tools used for our real data set.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The animals' unloaded reaching movements exhibited the canonical pattern of gently curved hand trajectories with bell-shaped velocity profiles (Morasso 1981) (Fig. 1B). Hand paths showed minimal systematic curvature across all targets (Fig. 2A). Moreover, the hand velocities were largely similar across different targets (Fig. 2B). The slowest hand velocity across different targets was 80% of the fastest hand velocity; the slowest movements were often directed to targets near the fore–aft axis. This variation was within our strict requirement on temporal and spatial accuracy (see METHODS, Task and apparatus).
|
The described movement patterns were conserved across different load conditions; 90% of the individual loaded trials had a movement direction within 15° of the mean direction during unloaded trials (Fig. 3A). Likewise, 88% of the individual loaded trials had a peak velocity within 20% of the mean velocity during unloaded trials (Fig. 3B). Only 18 and 13% of the comparisons had significant differences in movement direction and peak velocity, respectively (one-way ANOVAs, P < 0.01).
|
|
Reach-related activity of shoulder and elbow muscles
During unloaded reaching the upper arm muscles were maximally active for a single direction of hand motion (muscle's PHD = 271°, P < 0.01) (Fig. 5A). Such tuning was required for our selection criterion, but was also a typical pattern (Flanders et al. 1991; Hoffman and Strick 1999; Scott and Kalaska 1997; Turner et al. 1995). Elbow flexors and extensors expressed PHDs strongly oriented toward (mean PHD = 268°) and away from (mean PHD = 98°) the body, respectively (Fig. 6). The PHDs of shoulder flexors and extensors were oriented mostly away-left (mean PHD = 114°) and mostly toward-right (mean PHD = 305°) (Fig. 6). Last, biarticular extensors and flexors formed two clusters of PHDs oriented toward-left (mean PHD = 228°) and toward-right (mean PHD = 305°) rather than the opposing PHDs observed in single-joint antagonists (Fig. 6). The entire set of sampled muscles had a weak unimodal distribution (P = 0.05) and a negligible quadrimodal distribution (P > 0.2). In contrast, the group expressed a strong bimodal distribution that approached the fore–aft axis (PHA = 102–282°, Rayleigh test, P < 0.001) (Fig. 7A).
|
|
|
Arm muscles typically displayed a different pattern of activity between unloaded and loaded conditions. The interrelation between movement, torque, and EMG activity is elaborated for a sample muscle (elbow flexor) (Fig. 8, A–D). Here we focus on the target nearest its unloaded PHD where its load-related modulation is greatest. During the viscous-shoulder and viscous-elbow conditions the monkey generated additional shoulder-extension and elbow-flexion torques to achieve a similar movement pattern. The muscle's EMG was elevated from its activity during unloaded reaching in both cases and was greater still during the viscous-both condition. Consequently, a planar regression of the change in activity versus the change in torque indicated a preferred torque direction of 130 
8° (P < 0.01).
|
|
|
The sample neuron of Fig. 5B displays several features commonly observed in M1 neurons (and arm muscles). Foremost, the neuron exhibited significant directional tuning during unloaded reaching (PHD = 90°, P < 0.01). Roughly 70% of neurons expressed unimodal tuning from a larger sample (n = 493) selected under relatively liberal criteria—modulated by passive movements, postural loads, and perturbations during posture or reaching movements. Moreover, the PHDs of neurons (like muscles) formed a strongly bimodal distribution bias that approached the fore–aft axis (PHA = 109–289°, Rayleigh test, P < 0.001) (Fig. 7B) (Scott et al. 2001); the distribution was not significantly unimodal (P > 0.2) or quadrimodal (P > 0.1). Bootstrap tests indicated no significant difference between the bimodal spatial orientation of arm muscles and arm-related neurons (P > 0.05).
The sample neuron of Fig. 5B also showed changes in activity related to the change in torque (PTD = 292 15°, P < 0.01). Torque-related changes were common in M1 neurons and represented a significant fraction of the activity during unloaded reaching; the maximum change was nearly 76% of the maximum unloaded activity on average. Moreover, the entire sample of M1 neurons showed a strongly bimodal distribution of PTDs toward shoulder-extension/elbow-flexion torque and shoulder-flexion/elbow-extension torque (125–305°, Rayleigh test, P < 0.01) (Fig. 10B); the distribution was not significantly unimodal (P > 0.2) or quadrimodal (P > 0.2). Further, bootstrap tests indicated no significant difference between the bimodal torque orientation of muscles and neurons (P > 0.05).
Although M1 neurons and muscles shared several key features, they did not exhibit identical patterns. For example, during unloaded reaching the neurons showed greater decreases from prereach activity than observed in muscles (median unloaded-EI score: muscles = 0.97 vs. neurons = 0.73; Wilcoxon test, P < 0.05). Torque-related activity was also less common than that in muscles (proportion: muscle = 83% vs. neuron = 50%; difference of proportions, P < 0.01). Last, the torque-related changes in M1 neurons involved equal increases and decreases relative to the activity during unloaded reaching, whereas muscles showed significantly more increases than decreases (median loaded-EI score: muscles = 0.75 vs. neurons = 0.03; Wilcoxon test, P < 0.01).
Models of muscle activation
We found that both muscle properties and the recruitment strategy could affect the modeled PHDs. The simplest case is the "direct activation" model without any length- and velocity-dependent muscle properties. This arrangement predicts that the PHDs of single-joint muscles mirror the torque of their spanned joint (compare Fig. 2C with Fig. 11A). Likewise, biarticular PHDs reflect a combination of shoulder and elbow torque in proportion to their anatomical contribution. Therefore the "direct activation" model incorrectly predicts 1) that the PHDs of all muscle groups are biased to the left–right axis (PHA = 174–354°) and 2) that biarticulars have nearly opposite PHDs (Fig. 11, A and B).
|
We modeled the muscle activation during both unloaded and loaded conditions. Similar to our observations the modeled muscles showed increases in activity with increased load that were nearly collinear with the unloaded baseline (compare Fig. 6A with Fig. 12A). We examined the models' PTDs under a variety of conditions: with/without muscle intrinsic properties, with identical or measured movement patterns, and with/without the optimal muscle activation. The most complex "direct activation" model—with muscle intrinsic properties and measured movement patterns—did not show a significant PTD rotation toward shoulder-flexion/elbow-extension torque and shoulder-extension/elbow-flexion torque (mean = –2°, t-test, P > 0.05). In fact, significant PTD rotation occurred only with the optimal activation strategy (mean = 17°, t-test, P < 0.01) (compare Fig. 10C with Fig. 12B). This effect was robust over a range of simulated movement speeds (halved and doubled) and different cost functions.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
A major concern in motor control is to understand the muscle patterns that underlie movement (Zajac and Gordon 1989). In studies of reaching movements, researchers often implicitly vary the muscular requirement of the task (for review, see Berardelli et al. 1996). Experimental manipulations include different movement speeds (Corcos et al. 1989), movement distances (Brown and Cooke 1981), movement directions (Flanders et al. 1991; Karst and Hasan 1991; Scott and Kalaska 1997; Turner et al. 1995), and applied loads (Gottlieb 1996). Such studies provide insight into the flexibility of activation patterns but cannot (nor are intended to) examine how muscle activity is specifically related to joint torque.
In the present study we enforced similar limb movements during the application of different joint-based loads. This paradigm is analogous to muscular studies of postural control where the limb's configuration is constant during torque production (Buchanan et al. 1989; Kurtzer et al. 2006; Nozaki et al. 2005; Theeuwen et al. 1994). Muscle activation during these two tasks is fundamentally different because posture involves tonic activity and movement requires complex spatiotemporal patterns correlated to the extrinsic load and intrinsic muscle properties. However, enforcing similar movements with and without loads allowed us to peel away the unloaded reach-related activity and examine how muscle activity is specifically related to multijoint torque.
The loaded-reaching paradigm has several limitations arising from the velocity dependency of muscle and use of velocity-dependent loads. First, the impact on muscle force production resulting from shortening velocity varies across muscles because of their differing fascicle lengths, fiber type, and moment arms (Graham et al. 2003b; Singh et al. 2002). However, the velocity profile of each muscle will be approximately constant across loads if the movement pattern is conserved. Second, to reliably obtain similar limb trajectories across load conditions we enforced a movement speed below the animal's unloaded preference and above their loaded preference. This resulted in consistent performance but without a strong antagonist burst. Our modeling results suggest that similar patterns of torque-related activity occur over a range of movement speeds. Last, the load paradigm applied a nonuniform distribution of joint torques arising from the nonuniform relation between joint velocity and hand motion. The most frequent and largest torques involved shoulder-extension/elbow-flexion torque and shoulder-flexion/elbow-extension torque rather than flexion/flexion or extension/extension. We partly addressed this problem with a planar regression because it can readily handle nonuniform and unequal samples. Our modeling effort further confirmed that our uneven sampling per se did not systematically misestimate the PTDs because biases did not occur with the direct activation model. Future studies using a different applied load (such as a constant load) may allow faster movements with an antagonist burst and more uniform sampling of joint torque.
Muscular bias in PTDs and PHDs
During loaded reaching, the arm muscles exhibited a bias in their PTDs toward shoulder-flexion/elbow-extension torque and shoulder-extension/elbow-flexion torque. The resulting bimodal distribution in torque space could be considered evidence of functional grouping of muscles such as whole-arm flexion and extension synergies. Similar synergies were previously suggested to underlie motor behavior (Bernstein 1967; d'Avella et al. 2003; Ivanenko et al. 2004; MacPherson 1986; Ting and MacPherson 2005). However, we could reproduce the PTD bias using an optimization model that lacked any explicit grouping of muscles (Fagg et al. 2002; Kurtzer et al. 2006; Nozaki et al. 2005; van Bolhuis and Gielen 1999). [Note that PTD biases can result from other constraints on the motor system (besides optimization) such as cosine tuning to torque (Nozaki et al. 2005).] By examining the model with and without biarticular muscles we discounted that the PTD bias was an artifact of nonuniformities in the torques and/or small differences in movement patterns between load conditions. Rather, PTD biases reflect the interactions within a redundant muscle system.
During unloaded center-out reaching, we observed a strong bias in the muscles' preferred hand direction to targets located toward and away from the body. The bias in PHDs likely reflects three separate factors: limb geometry, muscle mechanics, and the bias of PTDs. If arm muscles were pure torque generators, the required muscular torque would be biased to the left and right of the body (see Models of muscle activation in RESULTS). However, the limb's geometry ensures that the largest joint velocities (and thus muscle velocity and displacement) occur toward and away from the body (Graham et al. 2003a). This meant that the greatest compensation for intrinsic damping and stiffness (Joyce et al. 1969; Scott et al. 1996) occurred along the fore–aft orientation. Last, the bias of muscles' PTDs (Kurtzer et al. 2006) will similarly bias the PHDs because shoulder-flexion/elbow-extension torque and shoulder-extension/elbow-flexion torque (where the muscle activity is greatest) occur with movements toward and away from the body, respectively. The efficacy of each separate factor was confirmed with a model of muscle activation.
In contrast to the present study, previous investigations have not reported a strong bias in the muscle's distribution of PHDs. However, these previous studies required abduction/adduction and internal/external rotation of the limb in addition to shoulder and elbow flexion/extension (Georgopoulos et al. 1982; Kalaska et al. 1989; Moran and Schwartz 1999; Scott and Kalaska 1997; Sergio and Kalaska 2005). Use of these degrees of freedom will certainly involve different muscle patterns. Second, these studies used unsupported limb movements that require greater muscle activity for movements away from the body resulting from the influence of gravity (Moran and Schwartz 1999). Finally, previous studies often used handheld pendula (Kalaska et al. 1989; Scott and Kalaska 1997; Sergio and Kalaska 2005) that will inflate the required muscular torques for left–right movements versus fore–aft movements arising from different mechanical advantages in the two directions. In contrast, the robotic exoskeleton used in the present study constrained the entire limb to a single plane, removed the influence of gravity, and had an inertial distribution roughly aligned with the animal's limb (Scott 1999).
Comparing muscular and M1 activity patterns
The observed patterns of muscular activity during reaching provide a foundation for interpreting cortical function in the same task. In particular, numerous studies have shown an intimate relationship between M1 and the motor periphery, including a dense descending projection to segmental regions (Dum and Strick 1991; Porter and Lemon 1993), with some direct input to motor neurons (Fetz and Cheney 1987; Lemon and Griffiths 2005). Moreover, the discharges of individual M1 neurons often covary with the mechanical attributes of a task and underlying muscular activity (Ashe 1997; Drew et al. 2004; Evarts 1968; Holdefer and Miller 2002; Lamarre et al. 1981).
Other studies have examined M1's relation to the motor periphery by considering all the muscles or neurons together. As a population, both arm muscles and arm-related M1 neurons exhibit a coupling between their onset time and magnitude of response (Scott 1997). Both rotate their PHDs (Caminiti et al. 1990; Scott and Kalaska 1997) and load sensitivities (Sergio and Kalaska 2003) with rotations about the shoulder (for rotations of preferred wrist direction, see Kakei et al. 1999). Both show similar temporal changes in their global preferred direction during reaching movements (Sergio and Kalaska 2005; Todorov 2000). Last, both muscles and M1 neurons are biased toward whole-arm flexion and whole-arm extension torques during a posture task (Cabel et al. 2001; Kurtzer et al. 2006).
The present data add to this list of population similarities in two ways. First, during unloaded center-out reaching the muscles' PHDs were strongly biased along an axis toward–away from the body as observed in the M1 neurons from the same monkeys (Scott et al. 2001). Second, the PTDs of arm muscles were biased toward shoulder-extension/elbow-flexion torque and shoulder-flexion/elbow-extension torque, again like M1 neurons from the same monkeys.
The similarities between M1 and muscular patterns do not imply a one-to-one mapping. Anatomically, M1 projects broadly to spinal lamina involved in different sensorimotor functions (Porter and Lemon 1993). Moreover, previous studies showed that M1 neurons show an impressive degree of variety, plasticity, and context dependency not present in limb muscles (Kakei et al. 1999; Kurtzer et al. 2005; Li et al. 2001; Paz et al. 2004; Sanes and Donoghue 2000; Scott 2003). In the present study, one clear difference between upper limb muscles and M1 neurons is the higher frequency of load-related activity in muscles; roughly half the neurons did not show load-related activity. Therefore many aspects of M1 function illustrate its role for converting global goals into detailed motor plans but also indicate it is not merely a collection of upper motor neurons. This association of low-level specificity with context-dependent processing is consistent with M1 playing a pivotal role within a motor system that behaves like an optimal feedback controller (Scott 2004; Todorov and Jordan 2002).
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: S. H. Scott, Centre for Neuroscience Studies, Botterell Hall, Rm 232, Queen's University, Kingston, Ontario, K7L 3N6, Canada (E-mail: steve{at}biomed.queensu.ca)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Batschelet E. Circular Statistics in Biology. New York: Academic Press, 1981.
Berardelli A, Hallett M, Rothwell JC, Agostino R, Manfredi M, Thompson PD, and Marsden CD. Single-joint rapid arm movements in normal subjects and in patients with motor disorders. Brain 119: 661–674, 1996.
Bernstein NA. The Coordination and Regulation of Movements. Oxford, UK: Pergamon Press, 1967.
Brown SH and Cooke JD. Amplitude- and instruction-dependent modulation of movement-related electromyogram activity in humans. J Physiol 316: 97–107, 1981.
Buchanan TS, Rovai GP, and Rymer WZ. Strategies for muscle activation during isometric torque generation at the human elbow. J Neurophysiol 62: 1201–1212, 1989.
Cabel DW, Cisek P, and Scott SH. Neural activity in primary motor cortex related to mechanical loads applied to the shoulder and elbow during a postural task. J Neurophysiol 86: 2102–2108, 2001.
Caminiti R, Johnson PB, and Urbano A. Making arm movements within different parts of space: dynamic aspects in the primate motor cortex. J Neurosci 10: 2039–2058, 1990.[Abstract]
Cheney PD, Mewes K, and Fetz EE. Encoding of motor parameters by corticomotoneuronal (CM) and rubromotoneuronal (RM) cells producing postspike facilitation of forelimb muscles in the behaving monkey. Behav Brain Res 28: 181–191, 1988.[CrossRef][ISI][Medline]
Cheng EJ and Scott SH. Morphometry of Macaca mulatta forelimb. I. Shoulder and elbow muscles and segment inertial parameters. J Morphol 245: 206–224, 2000.[CrossRef][ISI][Medline]
Collins JJ. The redundant nature of locomotor optimization laws. J Biomech 28: 251–267, 1995.[CrossRef][ISI][Medline]
Corcos DM, Gottlieb GL, and Agarwal GC. Organizing principles for single-joint movements. II. A speed-sensitive strategy. J Neurophysiol 62: 358–368, 1989.
Crowninshield RD and Brand RA. A physiologically based criterion of muscle force prediction in locomotion. J Biomech 14: 793–801, 1981.[CrossRef][ISI][Medline]
d'Avella A, Saltiel P, and Bizzi E. Combinations of muscle synergies in the construction of a natural motor behavior. Nat Neurosci 6: 300–308, 2003.[CrossRef][ISI][Medline]
Drew T, Prentice S, and Schepens B. Cortical and brainstem control of locomotion. Prog Brain Res 143: 251–261, 2004.[ISI][Medline]
Dul J, Johnson GE, Shiavi R, and Townsend MA. Muscular synergism—II. A minimum-fatigue criterion for load sharing between synergistic muscles. J Biomech 17: 675–684, 1984.[CrossRef][ISI][Medline]
Dum RP and Strick PL. The origin of corticospinal projections from the premotor areas in the frontal lobe. J Neurosci 11: 667–689, 1991.[Abstract]
Evarts EV. Relation of pyramidal tract activity to force exerted during voluntary movement. J Neurophysiol 31: 14–27, 1968.
Fagg AH, Shah A, and Barto AG. A computational model of muscle recruitment for wrist movements. J Neurophysiol 88: 3348–3358, 2002.
Fetz EE and Cheney PD. Functional relations between primate motor cortex cells and muscles: fixed and flexible. Ciba Found Symp 132: 98–117, 1987.[Medline]
Flanders M. Temporal patterns of muscle activation for arm movements in three-dimensional space. J Neurosci 11: 2680–2693, 1991.[Abstract]
Georgopoulos AP, Kalaska JF, Caminiti R, and Massey JT. On the relations between the direction of two-dimensional arm movements and cell discharge in primate motor cortex. J Neurosci 2: 1527–1537, 1982.[Abstract]
Gottlieb GL. On the voluntary movement of compliant (inertial-viscoelastic) loads by parcellated control mechanisms. J Neurophysiol 76: 3207–3229, 1996.
Graham KM, Moore KD, Cabel DW, Gribble PL, Cisek P, and Scott SH. Kinematics and kinetics of multijoint reaching in nonhuman primates. J Neurophysiol 89: 2667–2677, 2003a.
Graham KM and Scott SH. Morphometry of Macaca mulatta forelimb. III. Moment arm of shoulder and elbow muscles. J Morphol 255: 301–314, 2003b.[CrossRef][ISI][Medline]
Gribble PL and Scott SH. Method for assessing directional characteristics of non-uniformly sampled neural activity. J Neurosci Methods 113: 187–197, 2002a.[CrossRef][ISI][Medline]
Gribble PL and Scott SH. Overlap of internal models in motor cortex for mechanical loads during reaching. Nature 417: 938–941, 2002b.[CrossRef][Medline]
Haruno M and Wolpert DM. Optimal control of redundant muscles in step-tracking wrist movements. J Neurophysiol 94: 4244–4255, 2005.
Hepp-Reymond M. Functional organization of motor cortex and its participation in voluntary movements. In: Comparative Primate Biology, edited by Alan R. New York: Liss, 1988, p. 501–624.
Hoffman DS and Strick PL. Step-tracking movements of the wrist. IV. Muscle activity associated with movements in different directions. J Neurophysiol 81: 319–333, 1999.
Holdefer RN and Miller LE. Primary motor cortical neurons encode functional muscle synergies. Exp Brain Res 146: 233–243, 2002.[CrossRef][ISI][Medline]
Hollerbach JM and Flash T. Dynamic interactions between limb segments during planar arm movement. Biol Cybern 44: 67–77, 1982.[CrossRef][ISI][Medline]
Ivanenko YP, Poppele RE, and Lacquaniti F. Five basic muscle activation patterns account for muscle activity during human locomotion. J Physiol 556: 267–282, 2004.
Jongen HA, Denier van der Gon JJ, and Gielen CC. Activation of human arm muscles during flexion/extension and supination/pronation tasks: a theory on muscle coordination. Biol Cybern 61: 1–9, 1989.[CrossRef][ISI][Medline]
Joyce GC, Rack PM, and Westbury DR. The mechanical properties of cat soleus muscle during controlled lengthening and shortening movements. J Physiol 204: 461–474, 1969.
Kakei S, Hoffman DS, and Strick PL. Muscle and movement representations in the primary motor cortex. Science 285: 2136–2139, 1999.
Kalaska JF, Cohen DA, Hyde ML, and Prud’homme M. A comparison of movement direction-related versus load direction-related activity in primate motor cortex, using a two-dimensional reaching task. J Neurosci 9: 2080–2102, 1989.[Abstract]
Karst GM and Hasan Z. Timing and magnitude of electromyographic activity for two-joint arm movements in different directions. J Neurophysiol 66: 1594–1604, 1991.
Kurtzer I, Herter TM, and Scott SH. Random change in cortical loa
