Compartmentalization of muscles and their motor nuclei: the partitioning hypothesis.

Physical therapists possess a working knowledge of muscle function and kinesiology that is exemplary in rehabilitation. The analysis of movement dysfunction is the cornerstone of good physical therapy practice. Accordingly, a contemporary perspective on the composition and functioning of human muscles is a necessary requirement to enhance the analytic skills of physical therapists in detecting and treating movement abnormalities. For more than a century, the substrates on which the central nervous system (CNS) exerts its influence in the control of movement have been clearly recognized. Individual muscles are the anatomical elements in this plan. They all have well-defined bony attachments, and through those origins and insertions they exert their effects on movement. Each muscle is usually innervated by a single distinct muscle nerve. Significant components of this nerve are the axons of motoneurons, the cell bodies of which form a spatially segregated collection known as a "motor nucleus." Control of movement by the CNS is thus via control of excitation of these motor nuclei.

Although individual muscles and their motor nuclei might form the anatomical substrate for the control of movement, functional substrates are often thought to involve combinations of these anatomical elements, commonly referred to as synergists and antagonists. Indeed, studies of the connectivity between muscle afferents and motoneurons and of that between descending systems and motoneurons most often emphasize the groupings of motoneuron pools into functional synergists or antagonists. The most common of such groupings is that proposed by Sherrington,[1] which classifies muscles (and their motor nuclei) as either flexors or extensors. Within this broad collection of functional elements are more restrictive aggregates. For example, the triceps surae (and motor nuclei), medial gastrocnemius (MG), lateral gastrocnemius (LG), and soleus (SOL) muscles are often referred to as ankle extensor synergists.

A combination of factors, including studies of muscle architecture, muscle histochemistry, joint mechanics, and reflex connectivity, have stimulated a renewed interest in defining the substrates for the neural control of movement. Indeed, examining these factors has led us to understand that the control of purposeful movement is exceptionally complex. This review will summarize recent work in this area. It will focus mainly on the notion that traditionally defined individual muscles and their motor nuclei do not provide a comprehensive description of the anatomical elements of the neuromuscular system. We will attempt to identify results that contribute to a new concept of neuromuscular specificity. This concept is derived from the results of a number of studies and has gone by different names. Because it embodies the principles of what Stuart and colleagues[2] have termed "neuromuscular partitioning" as well as part of what they have termed "central partitioning," we have chosen to call the concept the partitioning hypothesis. We also present results that fit this hypothesis only approximately. The applicability of this concept to the design of human muscles--how the concept of partitioned muscles might contribute to the formation of new ideas concerning muscle function, especially with respect to synergism, and how these concepts have an impact on therapeutic endeavors--will also be discussed, as this issue is the real health care significance of this recent work.

The Partitioning Hypothesis

The notion that individual muscles might be subdivided, at least functionally, is not new. Many textbooks describe parts of muscles, such as the human deltoid and pectoralis major muscles, which are said to subserve different functions. Most practitioners of clinical muscle testing utilize this functional localization. The anatomical basis for and the details of this partitioning, however, have only recently been investigated rigorously.

One of the earliest and most influential studies in this field was that of Cohen.[3] He showed that stretching a small strip of cat rectus femoris muscle resulted in the reflex contraction of that strip and not other parts of the muscle. This observation was perceived as an even more precise localization of the stretch reflex than that described by Liddel and Sherrington.[4] Beginning in the late 1970s, a second piece of evidence suggested a subdivision of individual muscles. Researchers in several laboratories working on different muscles showed that the histochemical fiber type composition of a single muscle was not uniform throughout the muscle.[5-7] Different regions of a muscle contain different proportions of fast and slow, or oxidative and glycolytic, fibers. Because it is well known that individual alpha motoneurons innervate muscle fibers of only single histochemical types,[8] these histochemical results meant that the innervation territories of single alpha motoneurons are not uniformly distributed in a muscle.

The significance of these observations was perceived early by Letbetter,[9] who examined the innervation patterns of the cat ankle extensor muscles MG, LG, and SOL. Although he did not publish extensively, his work lays most of the intellectual framework for the partitioning hypothesis. This hypothesis states simply that most mammalian skeletal muscles (and their motor nuclei) are composed of smaller elements called neuromuscular compartments. For the ankle extensor muscles, English and Letbetter[10] described these elements as the portion of a muscle innervated by a primary muscle nerve branch. As a muscle nerve enters a muscle, it branches, and those branches course to different regions of the muscle. In the cat and rat LG and MG muscles, these naturally occurring first-order branches of the muscle nerve are usually the last major subdivision of the nerve before it divides into multiple small fascicles, which pass to the end-plate region of the muscle (Gatesy SM, English AW; unpublished observations).[10] Using the method of glycogen depletion, English and Letbetter[10] described the innervation territories of these compartment branches as distinct and virtually nonoverlapping territories Fig. 1A). Stimulating isolated single motor units (an alpha motoneuron and the muscle fibers it innervates) in the LG muscle, English and Weeks were able to show, using both evoked electromyographic EMG) activity mapping[11] and glycogen depletion,[12] that the innervation territories of single units correspond quite precisely with those of the compartment branches. That is, the muscle fibers innervated by a single motoneuron are not widely distributed in the muscle, but are confined to the innervation territories of a neuromuscular compartment. The axon of any motoneuron thus branches into one and only one compartmental nerve branch. Individual motor units are partitioned, and the organizational plan of those partitions is centered on the compartment branches of the muscle nerve.

Not surprisingly, the neuromuscular compartments of cat and rat LG muscles contain different proportions of muscle fibers of different types.[13,14] Each compartment is innervated by a different subset of motoneurons of the LG muscle's motor nucleus. If different proportions of slow-twitch versus fast-twitch motor units are found in different compartments, then the muscle fiber type composition of the different compartments would be expected to differ. Using this same logic, we would argue that the finding that many muscles are not uniform with respect to muscle fiber type supports the argument that the motor units in those muscles are partitioned into neuromuscular compartments.

Similarly, the suggestion of reflex partitioning derived from Cohen's work[3] might be explained by this hypothesis. The connectivity of large group Ia afferents onto motoneurons is weighted in favor of compartments. The amplitude of both the aggregate (population) and unitary (single axon) group Ia excitatory postsynaptic potentials recorded from motoneurons is larger if both the motoneuron and the afferent innervate the same compartment than if one or the other innervates a different compartment of the same muscle.[15-17] It should be noted, however, that studies of the connectivity of other afferent systems have been unable to demonstrate similar weighting.[18] The interested reader may consult the excellent review by Windhorst et al[19] on this topic.

Partitioning of Motor Nuclei

In addition to the partitioning of the innervation of individual muscles, the motor nuclei innervating these muscles are partitioned. Within the motor nucleus to the LG and MG muscles, different compartment nuclei have been described, using retrograde tracer horseradish peroxidase (HRP) applied to the cut stumps of the compartment nerve branches. The HRP is taken up by the cut axons, is transported retrogradely, and accumulates in the cell bodies of these motor axons in the spinal cord, where it can be visualized in tissue sections. By careful analysis of these sections, the location of compartment nuclei within the LG or MG muscle's motor nuclei has been studied.[20,21] Each compartment nucleus extends over a substantial part of the rostrocaudal extent of the motor nucleus, but most of the cells are aggregated in distinct rostrocaudal positions (Fig. 1B). The motoneurons innervating the most proximal compartments of both muscles lie mainly in the rostral poles of their respective motor nuclei, whereas motoneurons innervating distal compartments tend to lie in more caudal parts of the nucleus.[20,21] Because it has been known for some time that a similar construct exists for the MG (rostral), SOL (middle), and LG (caudal) components of the triceps surae muscle motor column,[22] we believe that the topographiclike structural organization of compartment nuclei within motor nuclei might contribute to the identity of the targets of those motoneurons. We speculate that such an identity might underlie the observed weighting of homonymous group Ia muscle spindle afferent connectivity to motoneurons in different compartment nuclei.[15-17] The finding that LG muscle afferents are arranged at their entry to the spinal cord in a spatial order similar to that of the position of compartment nuclei serves to strengthen this speculation.[23]

Functions of Neuromuscular Compartments

The anatomical organization of neuromuscular compartments would suggest a substrate for a localized function as well. The evidence for a partitioning of muscle function about the compartments, however, is less clear than the anatomical data. For muscles with broad attachments, such as the cat biceps femoris (BF) or the pig masseter, the relationship between anatomical partitions and localized function is reasonably logical. Caudal compartments of the BF muscle are strong knee flexors, whereas rostral compartments are almost pure hip extensors. This functional partitioning is clear from the afferent connectivity to different motoneuron pools,[24] from their activation during locomotion,[25,26] and from their response to different postural perturbations.[27] Similarly, Herring and co-workers[28,29] have provided strong evidence that differential activation of regions of the pig masseter muscle are correlated with functional masticatory movements.

In compartmentalized muscles that attach via long tendons, a suggestion of the same type of functional partitioning has been made, but the biomechanical explanation is a little less clear. In the cat LG muscle, EMG records made during locomotion indicate that activity recorded from compartments containing many slow-twitch motor units is usually more robust, at slow stepping speeds, than activity recorded from compartments without many of these small motor units. Sometimes, however, this pattern is reversed.[30] These observations lead to the conclusion that motor units in different compartments might be activated according to different patterns, at least during locomotion, How this differential activation might translate to a different functional conclusion is not clear, as it is assumed that transmission of all forces through a common tendon must lead to a common mechanical action.[31]

Two recent studies have brought into question this notion that there is a common mechanical action of all portions of long tendons. Examination of the forces produced by activity of single motor units in cat extensor carpi ulnaris muscles shows that these forces are not distributed equally to different parts of this single tendon.[32] Stimulation of individual motor units produces forces on all parts of the tendon, but the relative contribution of forces on the different tendon parts varied from unit to unit. Even though the compartmental organization of this muscle is not described, these findings are consistent with the notion that the cat extensor carpi ulnaris is functionally partitioned. Similarly, analysis of the different contributions of cat leg muscles to ankle joint torques revealed that the gastrocnemius muscles (LG and MG), which are thought to be plantar flexors of the foot, exert very substantial abduction (toe-out) torques as well.[33] Activation of distal compartments of the MG muscle, through the crossed-extensor reflex, produce nearly pure plantar-flexion torques, whereas cutaneous (sural) nerve stimulation, which recruits motor units in proximal as well as distal compartments of the MG muscle, produces both plantar-flexion and abduction torques.[34] Thus, even when a compartmentalized muscle attaches via a single, long tendon, the capacity for significant functional or mechanical partitioning may exist. Examination of more of the details of this capacity are a clear goal of research on the functions of neuromuscular compartments. In addition, the results of these studies have implications for how cutaneous or reflex-induced stimuli might activate different partitions of human muscles in a selective manner.

These findings offer a possible mechanical explanation for any differential activation of motor units in the triceps surae muscles. Activation of motor units in different compartments may result in differences in both the magnitude and direction of force applied at the tendo calcaneus. Differential activation of motor units in different compartments thus may be a means of regulating the direction of overall force application in muscles with long tendons. At present, the evidence for such differential activation of compartments in such muscles is not abundant. One clear goal of future research will be to investigate this area.

The partitioning hypothesis thus suggests that the elements of the CNS control of movement are not whole muscles but subsets of muscles. Regulation of movement is thus focused onto the control of excitability of motoneuron aggregates based on compartment nuclei rather than whole motor nuclei.

How Universally Applicable is the Partitioning Hypothesis?

The principle underlying the partitioning hypothesis, that muscles are composed of smaller elements, has received relatively broad support. A number of cat and rat hind-limb, neck, and trunk muscles and the diaphragm have been studied and shown to be partitioned,[35-41] as have the masticatory muscles of pigs, humans, and rodents (Timmis DP, English AW; unpublished research).[42,43] Forelimb muscles of cats[44] and upper-extremity muscles of humans[45,46] are also thought to be partitioned, but they have been studied less widely. Although many of these studies agree with our original observations on cat ankle extensors, not all studies have agreed as to the nature of the partitioning and of the organizational principle subserving it. Three examples of these different opinions will be discussed in an effort to provide a sense of the diversity of views currently available: (1) the nature of the compartmental nerve branches, (2) the spatial organization of compartmental motor nuclei, and (3) the compartmentalization of human muscles.

The Nature of Compartment Nerve Branches

The most common element of the partitioning hypothesis that is discussed is the meaning of compartmental nerve branches. In the cat and rat ankle extensor muscles, primary or first-order muscle nerve branches provide the exclusive innervation of neuromuscular compartments, because axons of motoneurons enter one and only one of these branches. The motor axon populations of higher-order branches of these muscle nerve branches (eg, secondary, tertiary, and so forth), when they exist, do not contain this exclusivity. Most or all of the motor axons in these higher-order branches contain branch points more proximally. Thus, branches of the same motor axons are likely to be found in more than one branch of the primary muscle nerve branch. Higher-order branches of a muscle nerve should thus be considered subcompartmental (Fig. 2). Because of this pattern of axon branch distribution, we consider a compartment to be the smallest portion of muscle receiving an exclusive innervation by a set of motoneurons.[25]

We also recognize that the organization of the ankle extensor muscles into compartments about the primary muscle nerve branches may be a simple example of the partitioning hypothesis' In other muscles, the organization plan may be more complex. For example, in the cat hamstring muscles, the muscle nerves recognized by most electrophysiologists since the time of Sherrington clearly branch to innervate different muscle territories. These territories are arranged in parallel in the BF and semimembranosus muscles and in series in the semitendinosus muscle, and, because they are innervated by what appear to be primary branches of the hamstring nerves, they have been treated as compartments.[25] The large size and histochemical heterogeneity of these muscular regions, however, prompted English and Weeks[25] to consider them aggregates of several compartments, which they termed "parts." We believe that it is likely the branches of these supra-compartmental nerve branches to these "parts" that are compartmental (Fig. 2). A more detailed analysis of these "secondary" branch patterns in cat BF muscles has shown that they innervate distinct subvolumes of muscle, which are arranged in parallel to one another.[27] Higher-order branches were numerous, quite fine, and distributed to the End-plate regions of the muscle. This organization is strikingly similar to that noted for compartment branches of the ankle extensors. Analysis of their responses to postural perturbations has shown a differential activation of the subvolumes,[47] suggestive of a differential distribution of motor units.

Similar circumstances exist for other mammalian limb, neck, and masticatory muscles, but the common design feature to the muscle is that at some level of branching of the muscle nerve, the motor units are partitioned into compartmental territories. The distribution of innervation territories of alpha motoneurons appears to be fundamental to defining a neuromuscular compartment. In many muscles, the distribution of these motor unit territories can be approximated by careful consideration of the nerve branch patterns, as has been described. As Pratt and Loeb[26] have suggested, however, the clear identification of the boundaries of a compartment may require "extensive experimentation." Because it is almost certainly through compartmental assemblies of motor units that the CNS regulates muscle contraction, such experimentation has to be considered worthwhile.

These remarks should not be construed as attempting to provide a universal explanation for all variations in neuromuscular partitioning. The carefull and thorough work of Richmond and her collaborators on the cat sartorius muscle[48,49] and of Herring and colleagues on the pig masseter muscle[28] have provided but two pieces of evidence that the organization of some muscles may be at least more complex than that of the muscles previously mentioned. Motor units in cat anterior sartorius muscles are partitioned around a complex architecture and innervation pattern that is at best a variant of the partitioning hypothesis but that may also be a quite different model. Motor units in the pig masseter muscle also appear to be partitioned, but the organization of the partitions is not about any obvious muscle nerve branches. The possible extensive intermingling of the muscle fibers of motor units in the innervation territories of different masseter nerve branches is actually more like that described for cat anterior sartorius muscles than for other muscles. The resolution of these issues will no doubt emerge from the continued high-quality work of these researchers.

The Spatial Organization of Compartment Nuclei

In addition to considerations regarding the nature of compartmental nerve branches, a number of questions have arisen regarding the spinal organization of the motoneurons innervating different neuromuscular compartments (compartment nuclei). For cat ankle extensor muscles, we showed that a topographylike organization of compartment nuclei exists.[20,21] Proximal compartments in the muscle are innervated by motoneurons in the rostral part of the muscles' motor nucleus, and distal compartments are innervated by caudal motoneurons. believe that this topographylike organization might help define the identity of the compartment nuclei. We also speculate that it might underlie the topographically weighted connectivity known to exist between muscle spindle primary afferents and compartmentalized motoneurons. Some studies have supported this finding of a topographylike relationship between compartment nuclei and compartments,[35,36,39] but others, using similar methods, have found a less spatially distinct distribution of motoneurons.[49,50] In some muscles, such as the cat anterior sartorius muscle,[49] these findings must be viewed with some caution, because the nature of the compartmentalization relative to muscle nerve branches is not yet clear. In muscles such as the diaphragm or the deep neck muscles, however, in which the innervation territories are reasonably clear, the finding of no obvious topographylike relationship between compartment nuclei and their peripheral targets seems more secure. Such observations have led to the conclusion that motoneuron position may not be significant, either functionally or in terms of defining the identity of the motoneuron.[51] At this time, a means of reconciling these two differing sets of observations is not clear.

In a different view of the central partitioning of motoneurons, Laskowski and Sanes[52] and DeSantis et al[53] have argued that muscles are divided into segmentally defined territories; different portions of a muscle are controlled by motoneurons in distinct spinal segments. This organization is clear for strictly segmentally arranged muscles such as the intercostal muscles[36] and, perhaps, some of the neck muscles. This organization is also true, however, in muscles, such as the serratus anterior muscle and the diaphragm, in which innervation is derived from several spinal segments. Some sectors of these muscles are innervated by motoneurons in rostral segments, and others are innervated by motoneurons in caudal segments.[52] This segmental specificity is preserved after reinnervation.[53]

The relationship between this segmental topography and the organization of compartment nuclei is not clear. Within the motor nuclei innervating the intercostal muscles, there seems to be a compartmental organization of motoneurons; proximal compartments are innervated by dorsally placed motoneurons.[36] In the rat diaphragm, the sectors of muscle that are distinctively innervated by motoneurons from a single spinal segment or part of a segment are beautifully arranged from one edge of the muscle to another.[52] These sector's appear not to be fully congruent with the compartments of the cat diaphragm, as defined by the innervation territories of the branches of the phrenic nerve,[41] and the motoneurons innervating the compartmental branches of the cat phrenic nerve display no coherent segmental topography.[51] Thus, we suspect that segmental and compartmental topography represent two types of partitioning of spinal motor nuclei. Providing explanations of the relationship between "compartmental" and "segmental" modes of partitioning will form an important area for further study.

Compartmentalization of Human Muscles

Anatomical and physiological evidence. The notion of partitioning of human muscles is not a very radical concept. As stated earlier in this review, anatomists, physiologists, and clinicians have long acknowledged the complexity of muscles such as the pectoralis major and deltoideus muscles. Less obvious potential examples of partitioning, however, have largely been ignored until recently. For example, ter Haar Romeny et al[54] used needle EMG recordings to distinguish parallel functional zones within the long head of the biceps brachii muscle. These zones contain motor units that have critical firing thresholds correlated to particular force vectors. Segal's dissections of the biceps brachii[46] provide an anatomical basis for the partitioning suggested by ter Haar Romeny et al, but, coupled with very preliminary evidence from EMG recordings from our laboratory, we believe that the partitioning of this muscle may be more extensive than considered previously.[55] Most studies of human muscle have not taken into account the possibility of partitioning beyond the grossly visible anatomical heads. For example, Huijing,[56] while carrying out detailed studies of the functional consequences of architectural arrangement of the gastrocnemius muscle, only divided that muscle into medial and lateral heads. Recent studies in our laboratory have suggested that the LG muscle alone contains three architecturally different heads.[45] The relationship of these heads to the branching patterns of the muscle nerve is less clear, however, suggesting that the LG muscle may be partitioned even more extensively than into three heads.

Our laboratory has found anatomical evidence to suggest that several human muscles (extensor carpi radialis longus [ECRL],[45] flexor carpi radialis [FCR],[45] LG,[45] and biceps brachii[46]) may be partitioned. Based on the results of our research, which focused on the gross architecture and branch patterns of the muscle nerves, we conclude that the nerve branch pattern in human muscles may be strikingly similar to that of cats and rats, both in terms of the regional distribution of branches within a muscle and with respect to architectural subdivisions of a muscle.[45] Unfortunately, methods used to date do not allow for either a precise definition of the boundaries of these proposed neuromuscular compartments or for analysis of the pattern of branching of axons with respect to nerve branches, both of which are important to defining compartments. Consequently, we refer to these subdivisions of human muscles by the general term "partitions."

To begin to assess the compartmental nature of the partitions of human muscles, we have recorded EMG activity from carefully placed fine-wire electrodes in the different partitions during the performance of several different motor tasks. Electrodes were placed simultaneously within different partitions in the proximal and distal portions of the FCR and ECRL muscles, and rectified and integrated EMG activity normalized to that produced during maximal voluntary contractions was recorded during the performance of different tasks. Analysis of these data revealed that significant differences in activity between proximal and distal muscle partitions exists for several, but not all, tasks. The Table shows tasks for elbow and wrist movements in which the limb was either supported or unsupported that yield significant differences in muscle activity within each muscle group. Under most movement conditions, significantly more activity could be recorded from the proximal FCR muscle than from the distal FCR muscle. The same pattern held true for the ECRL muscle, but there were even more tasks for which proximal ECRL muscle motor unit activity was significantly greater than that of the distal ECRL muscle. As might be expected, more activity could be recorded from electrodes within either proximal or distal partitions of each muscle during the wrist flexion or extension movements for which that muscle is named than during abduction (radial deviation) movements. During abduction, activity recorded from the proximal ECRL muscle was always the greatest of the four recording sites (Fig. 3).

In addition to these population EMG studies, we were able to use the technique of spike-triggered averaging to begin to study the localization of motor unit activity within whole muscles. Isolated single motor units were recorded from the electrodes placed in different partitions of the FCR and ECRL muscles. The recorded spikes were used to trigger a computer to average the activity recorded from electrodes placed into other partitions of the respective muscles. The average activity recorded in this manner, therefore, will be activity that is time-locked to the discharge of a motor unit recorded from a more distant part of the same muscle. Any significant time-locked activity could be due to (1) activity of the triggering motor unit noted because of the proximity of active muscle fibers to the recording electrode (ie, the muscle fibers of the unit are widely dispersed), (2) activity of the triggering motor unit noted because it is conducted passively in the extracellular spaces of the muscle (volume conduction or "cross talk"), or (3) simultaneous activation of two or more motor units located in different parts of the same muscle. Because the techniques we used do not allow us to differentiate between these scenarios, we have chosen to define any time-locked activity found using spike-triggered averaging by the general term "synchronization."

There was no evidence for synchronization of motor unit discharges between proximal and distal recording sites in all six FCR muscle motor units examined or in 16 of 20 motor units isolated in the ECRL muscle.[57] When found, the magnitude of this synchronization is quite small. Although the number of motor units analyzed is small, the marked lack of synchronization observed provides evidence at the level of single motor units that different partitions of the FCR and ECRL muscles in humans contain different populations of motor units. These data do not allow us to evaluate precisely whether the partitions we have identified both anatomically and electrophysiogically are a single neuromuscular compartment or more than one compartment ("part," Fig. 2). Coupled with the results of analysis of EMG intensity, however, these data do establish that motor units in human muscles are compartmentalized. These preliminary findings are being extended in an ongoing study.[58]

We have also made recordings from locations within the LG muscle that correspond to the partitions we described anatomically (Segal RL, Cheng LS, Catlin PA, et al; unpublished research).[59,60] Using analysis of normalized EMG activity, we found that different magnitudes of activity could be recorded from different compartments, depending on the task performed. For example, during a step-up exercise, overall motor unit activity in the distal portion of the A head of the LG muscle was greater than that recorded from any other partition, including an electrode placed in the proximal portion of the A head. In addition to conclusions similar to those drawn from study of the FCR and ECRL muscles, these findings suggest that motor unit activity even within a single anatomical partition of the LG muscle may be heterogeneous (ie, the partition may contain more than one compartment). These results might also be used to suggest that motor units in different partitions of the human LG muscle might contribute differently to the production of different parts of a movement.

The activity recorded from these sites might very well represent the recruitment of motor units uniquely housed within each partition. Using spike-triggered averaging aimed across the different electrode sites among 64 motor units obtained from three different subjects, we found evidence for synchronization in only 13 motor units (21%) (Segal RL, Cheng LS, Catlin PA, et al; unpublished research).[59,60] As noted in Figure 4, when such synchronized activity was recorded, it was proportionately quite small. Thus, as noted earlier for the FCR and ECRL muscles, the EMG activity recorded from each partition of the LG muscle is primarily the result of activation of motor units in which territories are confined to that recording location. It is of considerable interest that such findings are nearly identical to those described earlier for cat LG muscle motor units.[12]

If future efforts to define compartments based on anatomical, physiological, and histochemical criteria replicate these findings and extend them to additional muscles, the implications for rehabilitation are profound. If we assume that subvolumes of muscle are called on in a task-specific manner, then the response characteristics may dictate how the muscle should be stimulated or retrained. For example, we have preliminary data analyses from 276 motor units within the three heads of the LG muscle and from the SOL muscle recruited under a series of task-specific ankle and knee movements.[61] Recruitment thresholds during ankle plantar flexion against a fixed load with the knee extended were highest for motor units within the C head. Although not conclusive, such data suggest that either higher-threshold, and perhaps more phasic, motor units might be housed within this compartment and activated with greater loading in this task or the C head may function selectively in tasks that include knee flexion, as stated earlier (Segal RL, Cheng LS, Catlin PA, unpublished research).

Synergies may be between muscles partitions. Muscle synergies have traditionally been thought of as coordinated activities between muscles defined by their origins and insertions, as described in anatomy textbooks. For example, the biceps brachii, brachialis, and brachioradialis muscles can work together to perform elbow flexion. More complex synergies have also been identified, such as the triceps brachii muscle preventing elbow flexion while the biceps brachii (or perhaps part of it) participates in supination.[62] Missing from these analyses, however, is the overlaying of partitioning as a principle of muscle organization. For example, the ECRL and FCR muscles are antagonists in the flexion-extension plane, but they are synergists in the abduction-adduction plane. Does the purported proximal/deep partition of the ECRL muscle (more easily recruited for abduction) form a synergy with a portion of the FCR muscle during abduction of the wrist? That is, can partitions of different muscles form a synergy? Does a portion of the biceps brachii muscle's long head form a synergy with the supinator for supination? The next functional questions to ask may revolve around (1) whether anatomically defined partitions of human muscles are functional partitions and (2) whether partitions of muscles are linked with partitions of other muscles to form functional synergies (see also Chanaud and Macpherson[47]).

Therapeutic Significance of Partitioning

Because of the unique knowledge base and skills possessed by physical therapists in the analysis and amelioration of dysfunctional movement, the notion that partitions of volitionally controlled muscles may influence the precision and purpose of limb or body activity must have some exciting therapeutic implications. At the same time, recognizing the importance of muscle partitioning may require that we reassess our understanding of how muscles truly operate. Using only the traditional perspective of muscle origin and insertion as the sole criterion upon which we base action and direction of movement may limit optimal evaluation and treatment of dysfunction within the neuromuscular system.

Physical therapists may now find themselves rethinking some aspects of therapeutic regimens such as the size and location of stimulating electrodes over specific muscles, the intensity and duration of stimulation, and the positioning and circumstances (tasks) for providing stimulation. We know, for example, that the partition in the LG muscle we call the "C" head[45,59] has a specific role during lengthening of this muscle in ascending stairs. If weakness prevails in performing this activity, then stimulation of this partition, either during the task or in a simulation of it, might yield a more specific and justifiable intervention. Should we discover in future studies that this partition has a particular or predominant histochemical profile that differs from its counterparts in the LG muscle, then that profile will determine factors such as intensity and duration of contraction with electrical stimulation.

As another example of specification of movement circumstance, we might identify a muscle thought to produce different directional torques, such as the ECRL muscle. If a portion of the muscle produces an extension torque while another portion produces an abduction torque, and the desired goal is to yield an electrically induced extension, then we must place stimulating electrodes over that portion yielding extension.

In addition, therapeutic applications to train or monitor muscles would require a reevaluation of surface electrode placements about both the muscle end-plate orientation within each partition to ensure the best "pick-up" site and the recording electrode size appropriate for that portion of the muscles from which recordings are sought during a particular task. For example, if a muscle has partitions, electrodes for biofeedback would have to be placed over the muscle region generating the appropriate force. Thus, from a kinesiological point of view, selective differences in EMG activity from different partitions may become respecified by situation rather than action of the "entire" muscle. Such an observation would help to define which portions of a muscle are optimally used for a given situation and render a more legitimate approach to muscle retraining. Recording from the medial portion of the long head of the biceps brachii muscle, for example, may be necessary to best isolate the supination component from the elbow flexion component of this muscular head.[54] In any event, the reality that muscle partitions exist and may be functionally relevant provides reasons for us to reconsider our understanding of muscle anatomy and how this knowledge will assist us in refining our treatment approaches.

Acknowledgments

We would like to thank our colleague, T Richard Nichols, for reading drafts of this article. Thanks are also due to Pamela A Cadin, Edd, PT, the Division of Physical Therapy at Emory University, and the Emory University Graduate Program in Neuroscience and to Li-Shu Cheng for providing the data for Figure 4.

References

[1] Sherrington CG. Flexion reflex of the limb, crossed extension reflex and reflex stepping and standing. J Physiol (Lond). 1910;40:28-121. [2] Stuart DG, Hamm TM, Vanden Noven S. Partitioning of monosynaptic la excitation to motoneurons according to neuromuscular topography: generality and functional implications. Prog Neurobiology. 1988;30:437-447. [3] Cohen IA Localization of stretch reflex. J Neurophysiol. 1953;16:272-285. [4] Liddel EGT, Sherrington CG. Reflexes in response to stretch (myotatic reflexes). Proc R Soc Long [Biol]. 1924;96:212-242. [5] Gonyea WJ, Ericson GC. Morphological and histochemical organization of the flexor carpi radialis muscle in the cat. Am J Anat 1977; 148: 329-344. [6] Richmond IUF, Abrahams VC. Morphology and enzyme histochemistry of dorsal muscles of the cat neck J Neurophysiol. 1975;38:1312-1321. [7] Chanaud CM, Pratt CA, Loeb GE. Functionally complex muscles of the cat hindlimb, V: the roles of histochemical fiber-type regionalization and mechanical heterogeneity in differential Initial muscle activation. Exp Brain Res. 1991; 85:300-313. [8] Burke RE, Levine DN, Tsairis P, Zajac FE. Physiological types and histochemical profiles in motor units of the cat gastrocnemius. J Phsiol (Lond). 1973;234:723-748. [9] Letbetter WD. Influence of intramuscular nerve branching on motor unit organization in medial gastrocnemius muscle. Anat Rec. 1974; 178:402. [10] English AW, Letbetter WD. Anatomy and innervation patterns of cat lateral gastrocnemius and plantaris muscles. Am J Anat. 1982; 164:67-77. [11] English AW, Weeks OI. Electromyographic "cross-talk" within a compartmentalized muscle. J Physiol. 1989;416:327-336. [12] English AW, Weeks OI. Compartmentalization of single muscle units in cat lateral gastrocnemius muscle. Exp Brain Res. 1984;56: 361-368. [13] English AW, Letbetter WD. A histochemical analysis of identified compartments of cat lateral gastrocnemius muscle. Anat Rec. 1992;204: 123-130. [14] English AW Schwartz GA. Postnatal changes in muscle fiber type composition. Amer Zool. 1989;29:14A. Abstract. [15] Lucas SM, Binder MD. Topographic factors in distribution of homonymous group Ia-afferent input to cat medial gastrocnemius motoneurons. J Neurophysiol. 1984;51:50-63. [16] Lucas SM, Cope TC, Binder MD. Analysis of individual Ia-afferent EPSPs in a homonymous motoneuron pool with respect to muscle topography J Neuropbysiol. 1984;51:64-74. [17] Vanden-Noven S, Hamm TM, Stuart DG. Partitioning of monosynaptic la excitatory postsynaptic potentials in the motor nucleus of the cat lateral gastrocnemius muscle. J Neurophysiol. 1986;55:569-586. [18] Powers RK, Binder MD. Distribution of oligosynaptic group I input to the cat medial gastrocnemius motoneuron pool. J Neurophsiol. 1985;53:497-517. [19] Windhorst U, Hamm TM, Stuart DG. On the function of muscle and reflex partitioning. Behav Brain Sci. 1989; 12:629-681. [20] Weeks OI, English AW. Compartmentalization of the cat lateral gastrocnemius motor nucleus. J Comp Neurol. 1985;235:255-267. [21] Weeks 01, English AW. Cat triceps surae motor nuclei are topologically organized. Exp Neurol. 1987;96:163-167. [22] Romanes GJ. The motor cell columns of the lumbosacral spinal cord of the cat. J Comp Neurol. 1951;94:313-363. [23] Weeks 01, English AW. Topography of afferents entering the spinal cord from cat neuromuscular compartments. Abstracts of social Neuroscience. 1985;11:405. Abstract. [24] Botterman BR, Hamm TM, Reinking RM, Stuart DG. Localization of monosynaptic la excitatory post-synaptic potentials in the motor nucleus of the cat biceps femoris muscle. J Physiol (Lond). 1983;338:355-377. [25] English AW, Weeks OI. An anatomical and functional analysis of cat biceps femoris and semitendinosus muscles J Morphol. 1987;191: 161-176. [26] Pratt CA, Loeb GE. Functionally complex muscles of the cat hindlimb, I: patterns of activation across sartorius. Exp Brain Res, 1991;85: 243-256. [27] Chanaud CM, Pratt CA, Loeb GE. Functionally complex muscles of the cat hindlimb, 11: mechanical and architectural heterogeneity within the biceps femoris. Exp Brain Res. 1991;85:257-270. [28] Herring SW, Wineski IE, Anapol FC. Neural organization of the masseter muscle in the pig. J Comp Neurol. 1989;280:563-576. [29] Herring SW, Anapol FC, Wineski LE. Motorunit territories in the masseter muscle of infant pigs. Arch Oral Biol. 1991;36:867-873. [30] English AW. An electromyographic analysis of compartments of cat lateral gastrocnemius muscle during unrestrained locomotion. J Neuropbysiol. 1984;52:114-125. [31] Windhorst U. A possible partitioning of segmental muscle stretch reflex into incompletely decoupled parallel loops. Biol Cybern. 1979;34:205-215. [32] Fritz N, Schmidt C. Contractile properties of single motor units in two multitendoned muscles of the cat distal forelimb. Exp Brain Res. 1992;88:401-410. [33] Lawrence JH, Nichols TR, English AW. Cat hindlimb muscles exert substantial torques outside the sagittal plane. J Neurophysiol. 1993; 69:282-285. [34] Bonasera SJ, Lawrence JH, Nichols TR Selective recruitment within triceps surae of the decerebrate cat alters the direction of ankle torque profiles. Abstracts of Social Neuroscience. 1992;18:1549. Abstract. [35] Iliya AR, Dum RP. Somatotopic relations between the motor nucleus and its innervated muscle fibers in the cat tibialis anterior. Exp Neurol. 1984;86:272-292. [36] Harriman VJ, Brown MC. Spatial organization within rat motoneuron pools. Neurosci Lett. 1985;60:325-329. [37] Samojla BG, Rebeta PJ, Glenn U. Physiological and histochemical characterization of the rectus femoris muscle in the cat. Abstracts of Social Neuroscience. 1987;13:1213. Abstract. [38] Ballice-Gordon R, Thompson WJ. The organization and development of compartmentalized innervation in rat extensor digitorum longus muscle. J Physiol (Lond). 1987;395:211-232. [39] Richmond FJR, MacGillis DRR, Scott DA. Muscle fiber compartmentalization in cat splenius muscles. J Neurophysiol. 1985;53:868-885. [40] Armstrong JB, Rose PK, Vanner S, et al. Compartmentalization of motor units in the cat neck muscle, biventer cervicis. J Neurophysiol. 1988;60:30-45. [41] Hammond CG, Gordon DC, Fisher JT, Richmond FJR. Motor unit territories supplied by primary branches of the phrenic nerve. J Appl Physiol. 1989;66:61-71. [42] McMillan AS, Hannam AG. Motor-unit territory in the human masseter muscle. Arch Oral Biol. 1991;36:435-441. [43] Lev-Tov A, Tal M. The organization and activity patterns of the anterior and posterior heads of the guinea pig digastric muscle. J Neurophysiol. 1987;58:496-509. [44] Galvas PE Gonyea WJ. Motor-end-plate and nerve distribution in a histochemically

compartmentalized pennate muscle in the cat. Am J Anat. 1980;159:147-156. [45] Segal R, Wolf SL, DeCamp M, et al. Anatomical partitioning of three multiarticular human muscles. Acta Anat. 1991;142:261-267. [46] Segal RL. Neuromuscular compartments in the human biceps brachii muscle. Neurosci Lett. 1992;40:98-102. [47] Chanaud CM, Macpherson JM. Functionally complex muscles of the cat hindlimb, III: differential activation within biceps femoris during postural perturbations. Exp Brain Res. 1991;85:271-280. [48] Thomson DB, Scott SH, Richmond FJR. Neuromuscular organization of feline anterior sartorius, I: asymmetric distribution of motor units. J Morphol. 1991;210:147-162. [49] Gordon DC, Loeb GE, Richmond FJR. Distribution of motoneurons supplying cat sartorius and tensor fasciae latae, demonstrated by retrograde multiple-labelling methods. J Comp Neurol. 1991;304:357-372. [50] Gordon DC, Richmond FJR. Distribution of motoneurons supplying dorsal suboccipital and intervertebral muscles in the cat neck. J Comp Neurol. 1991;304:3433-3456. [51] Gordon DC, Richmond FJR. Topography in the phrenic motoneuron nucleus demonstrated by retrograde multiple-labelling techniques. J Comp Neurol. 1991;292:424-434. [52] Laskowski MB Sanes JP, Topographic mapping of motor pools onto skeletal muscles. J Neurosci. 1987;7:252-260. [53] DeSantis M, Berger PK, Laskowski MB, Norton AS. Regeneration by skeletomotor axons in neonatal rats is topographically selective at an early stage of reinnervation. Exp Neurol. 1992;116:229-239. [54] ter Haar Romeny BM, Denier van der Gon JJ, Gielen CCAM. Relation between location of a motor unit in the human biceps brachii and its critical firing levels for different tasks. Exp Neurol. 1984;85:632-650. [55] Bloom K, Schaet K, Stiffey S. Parallel Effects of Conditioning the Stretch Reflex of One Neuromuscular Partition of the Long Head of Biceps Brachii on Other Partitions Atlanta, Ga: Emory University; 1993. Master's thesis project. [56] Huijing PA. Architecture of the human gastrocnemius muscle and some functional consequences. Acta Anat. 1985;123:101-117. [57] McMahon T, Pianta T, Couch I, et al. Normalized electromyographic activity patterns in human extensor carpi radialis longus and flexor carpi radialis muscles: differential activity. In: Anderson PA, Hobart DJ, Danoff JV, eds. Electromyographic Kinesiology. New York, NY: Elsevier Science Publishing Co Inc; 1991:39-44. [58] Brideau A, Cupps I, Paulsen E, Thomasson E. Functional Partitions in Human Extensor Carpi Radialis Longus Muscle as Defined by Recruitment Threshold, Instantaneous Firing Rate, and Spike-Triggered Averaging. Atlanta, Ga: Emory University; 1993. Master's thesis project. [59] Wolf SL, Segal RL, English AW. Task-oriented EMG activity recorded from partitions in human lateral gastrocnemius muscle. J Electromyogr Kinesiol. 1993;3:87-94. [60] Richardson-Bond C, Coratti I, Mast A, et al. Patterns of electromyographic activity in the human lateral gastrocnemius muscle during weightbearing and non-weightbearing tasks. In: Anderson PA, Hobart DJ, Danoff JV, eds. Electromyographic Kinesiology New York, NY Elsevier Science Publishing Co Inc; 1991:65-68. [61] Hartney C, Taylor D, Weismantel D. Electromyographic Analysis of Human Lateral Gastrocnemius and Soleus Motor Units During Various Tasks in Healthy Subjects. Atlanta, Ga: Emory University; 1992. Master's thesis project. [62] Soderberg GL. Kinesiology: Application to Pathological Motion. Baltimore, Md: Williams & Wilkins; 1986.

AW English, PhD, is Professor, Neuroscience Program and Department of Anatomy and Cell Biology, and Associate Professor, Department of Rehabilitation Medicine, Emory University School of Medicine, 1248 Pierce Dr, Atlanta, GA 30322 (USA). Address all correspondence to Dr English.

SL Wolf, PhD, PT, FAPTA, is Professor and Director of Research, Department of Rehabilitation Medic Professor, Department of Internal Medicine, and Associate Professor, Department of Anatomy and Cell Biology, Emory University School of Medicine.

RL Segal, PhD, PT, is Associate Professor, Division of Physical Therapy, Department of Rehabilita Medicine, and Assistant Professor, Department of Anatomy and Cell Biology, and Neuroscience Program, Emory University School of Medicine.