Muscle synergies during locomotion in the cat: a model for motor cortex control

Abstract

It is well established that the motor cortex makes an important contribution to the control of visually guided gait modifications, such as those required to step over an obstacle. However, it is less clear how the descending cortical signal interacts with the interneuronal networks in the spinal cord to ensure that precise changes in limb trajectory are appropriately incorporated into the base locomotor rhythm. Here we suggest that subpopulations of motor cortical neurones, active sequentially during the step cycle, may regulate the activity of small groups of synergistic muscles, likewise active sequentially throughout the step cycle. These synergies, identified by a novel associative cluster analysis, are defined by periods of muscle activity that are coextensive with respect to the onset and offset of the EMG activity. Moreover, the synergies are sparse and are frequently composed of muscles acting around more than one joint. During gait modifications, we suggest that subpopulations of motor cortical neurones may modify the magnitude and phase of the EMG activity of all muscles contained within a given synergy. Different limb trajectories would be produced by differentially modifying the activity in each synergy thus providing a flexible substrate for the control of intralimb coordination during locomotion.

Introduction

Navigation on foot through the crocodile-infested billabongs of the Top End requires a certain combination of nerve and naivety, together with the ability to quickly modify one's gait should the situation require it. These gait modifications need to be smoothly, and most definitely accurately, incorporated into the gait cycle to avoid any obstacles that appear in one's path when taking evasive action.

The ability to make such gait adaptations is the result of the integrated action of a large number of central structures. However, in this symposium report, we will address the functions of only one of these structures, the motor cortex. It has long been known that lesion of the pyramidal tract leads to an inability to walk over challenging surfaces (Liddell & Phillips, 1944). More recent work has shown that inactivation of the motor cortex (Beloozerova & Sirota, 1988; Drew et al. 1996) or damage to the corticospinal tract (Drew et al. 2002) equally leads to deficits in visually guided gait modifications. In addition, single-unit recording studies over the last three decades have shown that neurones in the motor cortex, including pyramidal tract neurones (PTNs), show large increases in discharge frequency and, in many cases, their phase of activity, in situations in which visual information is used to modify the base locomotor gait to account for more challenging terrain (Beloozerova & Sirota, 1988, 1993; Drew, 1988, 1993; Amos et al. 1990; Marple-Horvat et al. 1993; Widajewicz et al. 1994; Drew et al. 1996, 2002). However, changes in motor cortical activity are observed only in the short period prior to, and during, the gait modification suggesting that the motor cortex contributes primarily to the execution of the gait modifications rather than to their planning (Drew et al. 2008).

Motor cortical activity during voluntary gait modifications

In experiments from this laboratory, we examined the discharge characteristics of PTNs in a task in which cats were trained to step over obstacles attached to a moving treadmill belt (Drew, 1988, 1991a, 1993; Widajewicz et al. 1994; Drew et al. 1996). In this situation, cats step smoothly over an obstacle without touching it and without undue interruption of their forward progress. Moreover, the limb trajectory is adjusted to the shape of the obstacle so that a wide obstacle elicits a corresponding increase in the length of the step over the obstacle and a high obstacle elicits a corresponding increase in its height (Drew, 1988, 1991b). Analysis of the electromyographic (EMG) activity during the gait modification shows that the smooth limb trajectory is the result of a complex pattern of EMG activity in which the amplitude, duration and relative timing of different muscle groups is modified at different periods of the step over the obstacle (Drew, 1993). Moreover, many muscles are activated during only a brief period of the swing phase of locomotion and different muscle groups are activated sequentially throughout this period. This is illustrated in Fig. 1A for four muscles selected to emphasize both the phasic and sequential nature of this pattern of activity. A major challenge for the nervous system is to produce spatiotemporal patterns of muscle activity that are appropriately modified in phase and magnitude to produce limb trajectories that are adapted to the size and shape of a given obstacle.

Figure 1. Sequential activation of EMG and neuronal activity during locomotion.

A, averaged EMG activity of 4 muscles during the swing phase of a gait modification when the limb was the first to pass over an obstacle (lead limb). B, modifications in activity of 4 pyramidal tract neurones (PTNs) recorded simultaneously with the muscles shown in A. Note that these units and associated muscle activity were recorded in 4 different experiments. Traces are aligned to the onset of activity in the brachialis (Br). Values to the right of each trace indicate the time of peak activity of each trace. The thin black line in each trace represents activity during unobstructed locomotion and the coloured line the activity during the voluntary gait modifications. The arrow emphasizes the sequential nature of the EMG and neuronal activity. Abbreviations: ECR, extensor carpi radialis (brevis); EDC extensor digitorum communis; TrM, teres major. Modified from Drew et al. (2008); reproduced with permission from Elsevier.

The results from our single unit recording experiments (Drew, 1993; Widajewicz et al. 1994; Drew et al. 2002) showed that many PTNs likewise show discrete phasic periods of activity at different times during the gait modifications. Examples are shown in Fig. 1B for four PTNs that were recorded simultaneously with the muscles illustrated to the left in Fig. 1A. In each of these examples, the neurones maintained a constant temporal relationship with the EMG activity of the illustrated muscle, irrespective of the shape and form of the limb trajectory. We have proposed that these subpopulations of PTNs act through the neuronal circuits in the spinal cord that produce the basic locomotor pattern (Drew, 1991b; Rho et al. 1999). In an attempt to explain how the descending signal may be integrated with the spinal circuits responsible for generating the base locomotor rhythm, we devised a conceptual model (Fig. 2A) based on the unit pattern generator hypothesis of Grillner (1982). In this model (Drew, 1991b), we proposed that subpopulations of PTNs would project preferentially to a single module and would thus exert their primary influence on the muscles influenced by that module. However, the composition of the putative modules and the mode of the integration were left open.

Figure 2. Integration of the decending signal from the motor cortex with the CPG.

A, we have previously suggested that subpopulations of PTNs in the motor cortex act via interneuronal networks that form part of, or are influenced by, the central pattern generator (CPG) for locomotion. We suggested that these interneurones are organized in a modular form (M1–M4), representing unit pattern generators (Grillner, 1982) acting on muscles acting around different joints (Drew, 1991b). Differential projections to the different modules (M1–M₄) provide a means of specifying activity in muscles acting around different joints while the connections between modules, responsible for generating and coordinating locomotion, ensure that the gait modification is appropriately integrated into the step cycle. B, a modified version of this conceptual model based on synergies (S1 – S_N). In this view, the modules activate muscles acting around multiple joints. A modified from Drew (1991b); reproduced with permission from Elsevier.

Muscle synergies during unobstructed treadmill locomotion

One possibility raised by recent work is that the descending commands from the motor cortex may act by modifying the activity of spinal primitives or synergies. This concept is one that has been recently developed in some detail by Bizzi and his collaborators (Giszter et al. 1993; Tresch et al. 1999, 2002; Bizzi et al. 2000; Saltiel et al. 2001; d'Avella et al. 2003; Hart & Giszter, 2004). The approach taken in most of these studies has been one in which mathematical decomposition methods have been used to determine the minimal number of synergies that can define the greatest range of motor patterns. In this approach each synergy contributes to the final level of activity in all of the muscles included in the database according to a weighting matrix that determines the sign and the efficacy of its contribution to each muscle. A similar approach has been used to study the organization of synergies during human locomotion (Ivanenko et al. 2004, 2005) as well as the production of postural responses to a perturbation (Ting & Macpherson, 2005; Torres-Oviedo et al. 2006).

As an initial examination of this proposition, we have recently undertaken a re-examination of the organization of the muscle activity patterns in the fore- and hindlimbs with a particular view to determining the extent to which different muscles, acting across the entire forelimb, function as synergists (Krouchev et al. 2006). In this study, we used a relatively simple definition of a synergy as a group of muscles whose activity is coextensive during a particular phase of the locomotor step cycle. We defined our synergies by measuring the time of onset and offset of discrete periods of muscle activity during locomotion and expressing these times as a phase of the step cycle (the latter was defined as the time between two successive periods of activity in the flexor muscles, cleidobrachialis (ClB) or brachialis (Br)). We then plotted the phase of offset of EMG activity as a function of the phase of onset and used a novel associative cluster method to determine the periods of EMG activity that were coextensive. By applying this analysis to a large number (26) of periods of EMG activity recorded from the forelimb during normal treadmill locomotion, we were able to define 11 clusters of activity spread throughout the step cycle (Fig. 3A). Most of these clusters (9/11) were active during the swing phase of locomotion. Several features of these synergies should be emphasized. First, several of the synergies include muscles that act around different joints. Second, muscles may be included in more than one synergy if they show more than one period of activity during locomotion. Third, each synergy contains only a small subset of the total muscles in the dataset. Fourth, each synergy can be related to discrete behavioural events during the gait modification (Fig. 3B).

Figure 3. Muscle synergies during unobstructed locomotion.

A, clusters defining groups of synergistic muscles during unobstructed treadmill locomotion. Each data point plots the phase of offset of a burst of muscle activity as a function of its onset. For each muscle we plot data from ∼20 step cycles. An associative cluster analysis groups together muscles whose activity is coincident. Colours and numbers identify 11 clusters of synergists. The red rectangle identifies the swing phase of locomotion. B, these groups of synergists (colour codes identical in A and B) can be related to different behavioural epochs of the step cycle. Abbreviations: AcD, acromiodeltoideus; Bic, biceps brachii; BrR, brachioradialis; ClB, cleidobrachialis; ClT, cleidotrapezius; ECU, extensor carpi ulnaris; LtD, latissimus dorsi; LvS, levator scapulae (ventralis); PaL, palmaris longus; PrT, pronator teres; SpD, spinodeltoideus; SSp, supraspinatus; Tri, triceps brachii, long head; TriL, triceps brachii, lateral head. Numbers in parentheses following the abbreviations indicate different periods of activity in a given muscle. From Krouchev et al. (2006), reproduced with permission from the American Physiological Society.

The definition of these synergies during locomotion leads us to modify our original conceptual model from one in which motor cortical neurones activate modules based on unit pattern generators acting around different joints to one in which the modules are based on multiarticular synergists. This modified concept is illustrated in Fig. 2B. In this view, subpopulations of motor cortical neurones are still considered to have restricted projection patterns to one or a few modules, allowing specificity of control over a limited number of muscles. However, in this model, each module represents a sparse group of functionally synergistic muscles that may act either around a single joint, or more commonly around multiple joints (see also Stein & Smith, 1997). By definition, the muscles in each synergy are active only during a limited part of the step cycle.

In contrast to the complicated recombinatorial process required to produce complex motor activities with the decomposition methods (see above), the sparse nature of our synergies means that each synergy can be controlled as an independent unit without the need for weighting matrices and without the need to modulate the activity in other synergies. This provides a more flexible substrate by which gait can be adapted to different sizes and shapes of obstacles by differentially and independently modifying the activity in the different groups of synergists. This concept may be better understood if one transforms the clusters shown in Fig. 3A into a family of basis units or direct components (DCs) by fitting a Gaussian to the centroid that defines each cluster (Fig. 4). Each DC then represents the ensemble activity of the group of muscles in the synergy. Each synergy is activated sequentially at a different phase of the step cycle and is active for only a short period of the entire step cycle. To produce gait modifications that are appropriately scaled to the size and shape of a given obstacle, the motor cortex needs to modify the magnitude and phase of these synergies. For example, the discrete phasic and sequential changes in motor cortical activity illustrated in Fig. 1 would specify the changes in magnitude and phase of the DCs representing the activity of the different synergies. This is illustrated schematically in Fig. 4B in the inset for one of the synergies.

Figure 4. Muscle synergies expressed as direct components (DCs).

A, the data in Fig. 3A are replotted to illustrate the centroids of each cluster. B, these centroids were used to create Gaussians (B) based on the mean and standard deviation of the phase of onset and offset of each centroid (method in Krouchev et al. 2006). We refer to these Gaussians as direct components (DCs). Inset schematically illustrates the type of changes in magnitude and phase that might be expected in DC3 when the forelimb was the lead or trail limb during steps over an obstacle. Similar changes would occur in each of the other DCs to modify limb trajectory according to the requirements of the task.

Conclusions

The characteristics of the motor cortical discharge and the functional connectivity of the corticospinal system with the spinal cord are fully compatible with the model of control that we propose. For example, electrophysiological and anatomical studies (Shinoda et al. 1976, 1986; Futami et al. 1979; Li & Martin, 2002) have shown that individual corticospinal axons in the cat branch widely in the spinal cord and probably innervate multiple motoneurone pools. In addition, it has been shown that low strength microstimulation within the cat motor cortex often activates multiple muscles, frequently acting around more than one joint (Armstrong & Drew, 1985a, b). However, the concept that the motor cortex regulates gait modifications by modifying the activity of sparse, multiarticular muscle synergies remains to be more thoroughly tested. For example, it is necessary to show that the composition of the synergies identified during unobstructed treadmill locomotion remain stable during gait modifications. Ideally, one would also show that the magnitude of the changes in different muscles comprising a synergy show similar changes. Last, one would also expect that appropriate analysis of the changes of phase and magnitude of population of PTNs recorded during gait modifications would show clusters of PTNs compatible with the changes in phase and magnitude that one would expect to observe in the muscle synergies (e.g. inset in Fig. 4B).

The concepts raised in this short review also raise questions concerning the organization of the spinal circuits that comprise or are influenced by the CPG. One of the rationales for this work was the hypothesis that a major driving force behind the evolution of the spinal cord was a need to produce and coordinate locomotion. We suggest that the synergies that we have described represent this base function and that these synergies may be hard-wired into the pattern of connectivity of interneuronal networks involved in locomotor control. As such, the challenge for central structures such as the motor cortex, which developed later in evolution in parallel with a more complex behavioural repertoire, is to appropriately use and modify these synergies. In relatively simple activities, such as locomotion, control over the magnitude and time of activation of these synergies is probably sufficient to allow adaptation of the limb trajectory to a wide range of terrains. However, for more complex and discrete movements, especially of the type that are used by non-human primates, it is probable that there would also be a need to modify the composition of the synergies. In this respect, as suggested by several others (see, e.g. Lemon & Griffiths, 2005), a major function of the motor cortex may be one of inhibiting synergies to permit fractionated movements.