Could enhanced reflex function contribute to improving locomotion after spinal cord repair?

Abstract

Although numerous treatments have been found to improve locomotion in spinal cord injured mammals, the underlying mechanisms are very poorly understood. Some of the main possibilities are: (1) regeneration of axons across the injury site and the re-establishment of descending pathways needed to voluntarily initiate and maintain stepping in the hind legs, (2) enhanced effectiveness of undamaged neurons in preparations with incomplete transections of the cord, (3) non-specific facilitation of reflexes and intrinsic spinal networks by transmitters released from regenerated axons and/or by substances introduced by the treatment, and (4) enhanced trunk movements close to the injury site strengthening the mechanical coupling of the trunk to the hind legs via spinal reflexes. In addition, any procedure that even slightly improves stepping may be further enhanced by use-dependent modification of reflex pathways and interneuronal networks in the lumbar cord. The emphasis of this review is on the contribution of spinal reflexes to the patterning of motor activity for walking, and how enhancing reflex function may contribute to the improvement of locomotion by treatments aimed at restoring locomotion after complete transection of the spinal cord.

Restoration of locomotor function after damage to the spinal cord is an enormously challenging problem (McDonald, 1999; Bunge, 2000). The major difficulty is that the axons of damaged neurons do not regenerate in the spinal cord under normal conditions. Although considerable progress has been made in promoting axonal regeneration in experimental animals (Bregman, 1998), no growth-promoting procedure has yet been developed that could be safely and effectively applied to humans with chronic spinal cord injury. Currently, rehabilitation of locomotion in patients with chronic spinal cord injury relies either on intense physical therapy using a treadmill (Wernig et al. 1995, 1999; Dietz et al. 1995, 1998) or functional electrical stimulation of leg muscles (Kralj & Bajd, 1989; Wieler et al. 1999). An encouraging sign is the flurry of recent reports describing procedures promoting survival, growth and regeneration of neurons in the spinal cord and an associated improvement of locomotor function in animals with injuries of the spinal cord (Table 1). Although none of these studies reported a full recovery of locomotor function, the modest gains have led to optimism that effective restorative procedures are feasible and might be developed in the near future.

Table 1.

Treatments improving locomotion after spinal cord injury

Animal	Lesion	Treatment	Reference
Neonatal rat	Complete	Embryonic cord segment	Iwashita et al. 1994
	Complete	Fetal cord	Miya et al. 1997
	Complete	Fetal cord	Giszter et al. 1998
	Hemisection	Fetal cord	Kunkel-Bagden & Bregman, 1990
Adult rat	Complete	Peripheral nerve/acidic FGF	Cheng et al. 1996
	Complete	Peripheral nerve/acidic FGF	Cheng et al. 1997
	Complete	Schwann cells	Guest et al. 1997
	Complete	Macrophages	Rapalino et al. 1998
	Complete	Olfactory ensheathing cells	Ramon-Cueto et al. 2000
	Complete	Raphe neurons	Giménez y Ribotta et al. 2000
	Hemisection	IN-1	Bregman et al. 1995
	Partial	NT-3	Grill et al. 1997
	Contusion	BDNF	Jakeman et al. 1998
	Contusion	Embryonic stem cells	McDonald et al. 1999
	Contusion	Myelin basic protein	Hauben et al. 2000
Mouse	Contusion	CM101	Wamil et al. 1998
Kitten	Complete	Fetal cord	Howland et al. 1995

In those studies in which the experimental procedure promoted axonal regeneration across the injury site, a key issue is the extent to which the improved locomotor performance was due to the regenerated axons forming functional connections to the neuronal networks generating the locomotor movements (Fig. 1A). This issue remains unresolved. The confounding factors are that other mechanisms could have been involved in improving locomotion, and the tests for assessing locomotor function were not designed to establish the mechanism(s) for functional improvement. When considering the mechanisms responsible for the improvement in locomotion an important fact to be kept in mind is that the neuronal circuitry for generating the motor pattern for stepping of the hind legs of quadrupeds is located mainly in the lumbar segments of the cord. In cats and dogs it has been know for about a century that stepping movements can be produced in the hind legs after spinal cord transection (Sherrington, 1910), while more recent work has demonstrated that stepping in spinal cats can be facilitated by training (Lovely et al. 1986; Barbeau & Rossignol, 1987; De Leon et al. 1998). Adult spinal rats do not spontaneously develop strong stepping movements but these movements can be produced by training (London et al. 2000). Another important fact is that implantation of serotonergic neurons into the lumber cord of adult spinal rats can enhance hindlimb stepping to a degree that closely resembles normal stepping (Giménez y Ribotta et al. 2000). Similarly, local application of serotonin to the isolated lumbar cord enables the generation of vigorous locomotor activity (Feraboli-Lohnherr et al. 1999). The primary lessons from these recent investigations on rats is that training and the local application of neuromodulatory agents can powerfully facilitate the generation of locomotor activity in the adult rat spinal cord independently of any input from rostral structures. Thus it is conceivable that the functional improvements in locomotion occurring after enhancing growth of axons across an injury site may be due either to tonic release of neuromodulators from the regenerated neurons, or to the actions of the implanted agents (or their metabolites) onto lumbar locomotor networks (Fig. 1B). The functioning of these networks could then be further enhanced by use-dependent modifications of reflex pathways and/or interneuronal circuits involved in the patterning of locomotor activity.

Figure 1. Schematic diagrams illustrating three possible mechanisms for improved locomotion by treatments (shaded areas) bridging a complete transection of the spinal cord.

A, descending axons regenerate (dashed lines) into the lumbar cord and re-establish functional connections with the neuronal networks patterning motor activity for stepping. B, regenerating axons and/or substances derived from the bridge non-specifically increase the excitability of the pattern-generating network. C, substances derived from the bridge facilitate voluntary activation of axial muscles that are then mechanically coupled to the pattern-generating network via segmental reflexes.

Another method for indirectly engaging the locomotor pattern generation network in the spinal cord has been suggested by a recent study of motor cortex reorganization following the implantation of fetal spinal cord into spinalized neonatal rats (Giszter et al. 1998). In the animals receiving the transplants, the size of the cortical representation of the trunk region was significantly larger than that in control animals. No representation of the hind legs emerged in either group. However, locomotor activity of the hind legs of the transplanted animals improved. These and other findings led to the suggestion ‘that operated rats that achieve weight support may primarily use the axial muscles to steer the pelvis and hindlimbs indirectly rather than use explicit hindlimb control during weight-supported locomotion’ (Giszter et al. 1998). One way for coupling the axial and hindlimbs movements could be via reflex pathways regulating the locomotor pattern-generating network (Fig. 1C). It is conceivable that influences of the transplant on rostral spinal segments enabled animals to learn a pattern of axial movements that more effectively engaged spinal reflexes involved in the production of stepping movement. The expansion of the cortical representation of the trunk could be a reflection of this learning process. If this were true then spinal reflexes would have to influence strongly the spinal network generating the locomotor pattern. Although data on afferent regulation of stepping in the rat are meagre (Fouad & Pearson, 1997a) some important aspects are similar to those in the cat, an animal in which we have a great deal of knowledge about the afferent control of stepping. Therefore, before assessing in more detail the possible role of spinal reflexes in the functional improvement of locomotion in treated spinal animals, the main features of the afferent control of stepping in mammals will be summarized briefly.

Afferent regulation of stepping

Over the past decade it has become increasingly clear that afferent feedback plays an essential role in the patterning of motor activity for stepping in mammals. During normal unperturbed walking both the timing and magnitude of activity in leg muscles is regulated by afferent signals (Pearson, 1995; Pearson et al. 1998; Duysens et al. 2000). In addition, injury-induced adaptive modifications of the motor programme for walking may also depend on altered afferent signals from leg receptors (Pearson, 2000a, b). The following sections describe each of these points separately.

Regulation of the timing of stepping by afferent signals

One of the clearest, and simplest, demonstrations that afferent signals control the timing of stepping is that rats, cats and human infants, when suspended above a treadmill, adapt their rate of stepping to the treadmill speed (Grillner & Rossignol, 1978; Fouad & Pearson, 1997a; Yang et al. 1998). The matching of stepping rate to treadmill speed depends on sensory afferents signalling leg extension and muscle unloading near the end of stance. This information is then used to control the timing of the transition between stance and swing. Another simple phenomenon, which also illustrates the importance of afferent information in controlling the step cycle, is that impeding leg extension during stance inhibits the initiation of swing until the leg is allowed to move to an extended position (Grillner & Rossignol, 1978; Fouad & Pearson, 1997a; Pang & Yang, 2000).

The nature of the afferent signals controlling the timing of the stance-to-swing transition has been reviewed in detail in other articles (Pearson, 1995; Pearson et al. 1998; Duysens et al. 2000). The main point to note in the context of the present article is that feedback from the Golgi tendon organs in the ankle extensor muscles has a particularly potent inhibitory action on flexor burst generation (Duysens & Pearson, 1980; Conway et al. 1987; Pearson et al. 1992; Pearson & Collins, 1993; Gossard et al. 1994). These receptors signal the force in the ankle extensors during stance and thus provide the nervous system with information about the load carried by the leg. The functional consequence of the inhibitory action of the Golgi tendon organ feedback on flexor burst generation is that the initiation of the swing phase is prevented until the load carried by the leg is reduced. Unloading normally occurs during the latter part of the stance phase as weight is transferred to other legs. The shortening of the ankle extensors in the latter half of stance also acts to reduce force production thus further contributing to the reduction in the inhibitory signal from the tendon organs. Thus any bodily movements that alter the load on a leg can alter the time the swing phase of that leg is initiated. Hence under some circumstances the coordination of stepping in different legs could depend on mechanical coupling between different limbs independently of neuronal pathways in the spinal cord.

Regulation of extensor burst magnitude by afferent feedback

In humans and cats feedback from muscle proprioceptors during stance contributes significantly to the generation of burst activity in leg extensor muscles. The magnitude of the afferent contribution to the activation of ankle extensors has been estimated to be about 30 % during early stance in humans (Yang et al. 1991) and more than 50 % during most of stance in decerebrate cats (Hiebert & Pearson, 1999). A consequence of the large afferent contribution is that unloading of extensor muscles, such as stepping into a hole, supporting the weight of the body or assisting muscle shortening, causes a marked reduction in the magnitude of extensor activity (Gorassini et al. 1994; Hiebert & Pearson, 1999; Sinkjaer et al. 2000). On the other hand stretching and loading the extensor muscles enhances extensor activity (Hiebert & Pearson, 1999; Stephens & Yang, 1999; Misiaszek et al. 2000; Sinkjaer et al. 2000) and this increase in activity contributes substantially to force production (Stein et al. 2000).

The primary function of the proprioceptive regulation of the extensor activity during stance appears to be to automatically adjust motor commands to the muscle to a level appropriate for the load carried by the leg. Additional loading of the leg produces an increase to motor activity thus helping to maintain constancy in the ongoing movement. Appropriately, therefore, one of the afferent signals contributing to enhancing extensor activity in cats is from the force-sensitive Golgi tendon organs (Pearson & Collins, 1993). Whether or not Golgi tendon organs function to enhance extensor activity in walking rodents and humans remains uncertain (Fouad & Pearson, 1997a; Sinkjaer et al. 2000), although afferents other than group I afferents may contribute to enhancing extensor activity in humans (Sinkjaer et al. 2000).

Plasticity in afferent pathways regulating stepping

It has been known for some time that the recovery of stepping in the hind legs of spinalized cats can be greatly enhanced by regular training on a treadmill (Lovely et al. 1986; Barbeau & Rossignol, 1987). Apart from demonstrating a remarkable plasticity in the locomotor pattern-generating network, this finding demonstrates that adaptive modifications in the network are driven by sensory signals from the moving legs. The nature of these adaptive modifications have not been established at a cellular level, although a decrease in glycinergic inhibition appears to be one factor (De Leon et al. 1999). Another factor may be use-dependent changes in afferent pathways involved in the production of the stepping motor pattern. This speculation is based on the fact that alterations in afferent pathways have been found to be associated with functional recovery of stepping after damage to the spinal cord and peripheral nerves (Goldberger & Murray, 1982; Helgren & Goldberger, 1993; Whelan et al. 1995; Whelan & Pearson, 1997; Fouad & Pearson, 1997b; Pearson & Misiaszek, 2000). The topic of plasticity in afferent pathways controlling stepping has been reviewed elsewhere in detail (Pearson, 2000a,b; Bouyer & Rossignol, 2000).

A recent finding relevant to the issue of functional recovery after spinal cord repair is that reflexes regulating the timing of the stance-to-swing transition and reinforcing the generation of extensor activity are enhanced by partial denervation of ankle extensor muscles (Whelan & Pearson, 1997; Pearson & Misiaszek, 2000; Gritsenko et al. 2001). These increases in reflex strength take days to weeks to fully develop, and they are established while animals are behaving normally. The latter makes it difficult to be confident about the location of the sites of the adaptive changes, but there are strong indications that some modifications occur at the level of the spinal cord. For example, the amplitudes of group I field potentials from an innervated synergist muscle are increased in the intermediate nucleus of lumbar segments (Fouad & Pearson, 1997b), and an increase in the effectiveness of group I input from the same muscle in controlling extensor burst duration is maintained in some animals after spinal cord transection (Whelan & Pearson, 1997). Whether or not these adaptive changes can be induced in spinal animals has not been established, although some observations do indicate that reflexes reinforcing extensor burst generation can be enhanced in these animals (Rossignol et al. 1997).

An older, but very important finding indicating that plasticity in spinal reflexes is a factor in functional recovery of stepping after nervous system damage is that the terminals of sensory afferents sprout following hemisection of the spinal cord in adult cats (Goldberger & Murray, 1982; Helgren & Goldberger, 1993). Sprouting of afferent terminals in the lumbar cord has also been observed after partial denervation of the cord (Zhang et al. 1995). These observations, together with the recent findings of functional modification of spinal reflexes, strongly indicate that the properties of spinal reflex pathways can be easily modified by injury to the central or peripheral nervous system, and these modification are probably involved in functional improvement in locomotion.

Improved locomotion after repair of complete spinal transections: a role for reflexes?

Improved stepping of the hind legs after treatments promoting regeneration of axons across a complete transection of the spinal cord in rats and kittens seems, at first sight, a compelling reason for believing that the regenerated axons form functional connections. Certainly it is a stronger indication than improvements in locomotion after partial transections or contusion injuries since functional gains with these injuries could be due to enhanced function of undamaged axons. But is it sufficient to be completely confident that axons regenerating across a treated injury site are participating in the voluntary initiation of locomotion or the coordination of stepping in the fore and hind legs? Numerous behavioural measures have been used to document improved locomotion with treatments bridging complete transections of the spinal cord. These include the number of weight-bearing steps of the hind legs, the ability to walk up inclined grids, and the coordination of fore and hind leg stepping (see references listed in Table 1). Unfortunately none of these measures, as they have been applied, unambiguously demonstrates re-establishment of functional descending pathways for initiating and maintaining hind leg stepping and/or of propriospinal pathways coordinating activity in the fore and hind legs. This is not to say that these connections have not been formed in some situations, but only that there are alternative explanations that can as easily explain the behavioural data. For instance, neuromodulatory substances (such as serotonin) released non-specifically from regenerated axons may facilitate activity in lumbar pattern-generating networks and the reflex pathways controlling these networks.

A contribution of enhanced reflex function in improving locomotor performance is indicated by recent observations on adult spinal rats treated with olfactory ensheathing cells (Ramon-Cueto et al. 2000). Improved locomotion on an inclined grid was strongly correlated to the appearance of cutaneous and proprioceptive reflexes thus leading to the suggestion that ‘…reestablishment of proprioception and light touch responses might be necessary to properly perform the climbing test…’. It should be noted that climbing on an inclined grid is exactly the situation that would benefit most by enhanced reflex function. In this situation the hind legs are more loaded than normal thus activating more strongly load-dependent reflex pathways involved in the regulation of the timing and magnitude of locomotor activity (Fouad & Pearson, 1997a). If transplanted animals had the tendency to spontaneously produce more rhythmic hind leg movements due to a non-specific effect of the treatment (perhaps from regenerated axons) on interneuronal networks, then the enhanced influence of proprioceptive feedback onto these networks may allow the animals to succeed in climbing a grid. Interestingly it appears that the treated animals did not improve their walking on horizontal surfaces. The difference in performance on a horizontal surface and a grid could be due to a relatively low level of afferent feedback when traversing the former.

One mechanism that could underlie enhanced reflex function is suggested by investigations on 5HT agonists on stretch reflexes in spinal cats (Miller et al. 1996) and on locomotor function in spinal rats (Kim et al. 1999). When the spinal cord is transected in decerebrate cats the stretch reflex in ankle extensor muscles is greatly attenuated. However, treatment of the spinalized animals with a 5HT₂ receptor agonist enhances extensor excitability and facilitates the stretch reflex (Miller et al. 1996). We also know that 5HT receptors are up-regulated in cats and rats after spinal cord transection (Giroux et al. 1999; Kim et al. 1999) and that locomotion in spinal rats treated with fetal cord grafts is enhanced by treatment with 5HT₂ receptor agonists (Kim et al. 1999). Thus it is conceivable that functional improvements in locomotion produced by treatments that promote regeneration of serotonergic neurons across injury sites (Bregman et al. 1995; Cheng et al. 1996; Miya et al. 1997; Giszter et al. 1998; Ramon-Cueto et al. 2000) are due, in part, to 5HT released from the regenerated axons facilitating spinal reflexes controlling locomotion.

Spinal reflexes may also contribute to functional improvement by treatments that allow animals to gain greater control of movements of the trunk either by a neuroprotective influence on rostral neurons or enhanced in-growth of dorsal root afferents near the transection site. In this situation hind leg movements could be mechanically coupled to movements of the forelimbs and trunk via spinal reflexes controlling the lumbar pattern-generating network (Fig. 1C). The only evidence that this may occur comes from a study in neonatal rats treated with fetal cord tissue (Giszter et al. 1998). However, improved coordination of stepping in kittens treated with fetal cord (Howland et al. 1995) may also depend to some extent on enhanced mechanical linkage between foreleg and hind leg movements via the trunk and segmental reflexes. This hypothesis was not explored but, given the absence of any compelling evidence for specificity of regeneration of propriospinal pathways, it seems more plausible at the present time. The possibility that enhanced mechanical linkage of foreleg and hind leg movements contributes to functional improvement in locomotion must also make us cautious about the interpretation of observations made after a second transection of the cord. If the second transection is rostral to the initial transection, as has been the case in studies to date (Miya et al. 1997; Rapalino et al. 1998), the abolition of any functional gains may be due to the loss of trunk function near the site of the original transection and not due to severing regenerated descending axons.

Conclusions

Although the emphasis of this article has been on a possible role of spinal reflexes in the improvement of locomotion by treatments promoting neuronal survival and axonal regeneration following spinal cord injury, it is quite apparent that many different mechanisms could contribute to functional recovery. Whatever the mechanisms, they will not be clearly understood until electrophysiological procedures allowing examination of the functional properties of neuronal circuits and the patterns of muscle activation are used more frequently. Some of the most obvious directions for future investigations using these procedures are alterations in the excitability of central pattern-generating networks, modifications of segmental reflexes influencing pattern-generating networks, changes in the pattern of movements of the trunk and in the coordination of trunk and hind leg movements, and an evaluation of the capacity of brainstem centres to evoke locomotor activity. Information gained from these investigations should help in developing strategies for the rational treatment of spinal cord injury.