A spinal cord injury (SCI) is a devastating event that, depending on the level and severity, impacts sensorimotor and autonomous function. In affected subjects the goal of rehabilitative interventions is the regaining of independence and thus a good quality of life. From the patients perspective this is probably best achieved by targeting restoration of bladder and bowel function, and in tetraplegic subjects upper limb function[
1]. However, recovery of locomotor ability is also of high priority by SCI subjects independently from the severity, time after injury and age at the time of injury[
2]. It is now widely accepted that the central nervous system is able to recover locomotor function following incomplete SCI with functional training on a treadmill combined with partial body weight support[
3‐
5]. However, the physiological requirements for training effects remain vague. A century of research into the organization of the neuronal processes underlying the control of locomotion in mammals has demonstrated that the basic neuronal circuitries responsible for generating efficient stepping patterns are embedded within the lumbosacral spinal cord[
6,
7]. These spinal locomotor circuitries appear to play a crucial role in stepping ability in animal models and in human SCI. This review covers the physiological basis of effective locomotor training after SCI. Several studies in animal models, especially in rats and cats, have unraveled the neuronal principles underlying neurorehabilitation of locomotion after SCI. The role of neuronal plasticity will be discussed in this review. After SCI neuroplasticity in the cortex, brainstem and spinal cord can be exploited by rehabilitative approaches. This review will mainly focus on studies of neuroplastic changes at the spinal level as the knowledge gained from these studies is used to create novel translatory neurorehabilitative approaches with the aim to restore and improve locomotor ability after human SCI.
Neuronal basis of human locomotion
The question, how does the central nervous system coordinate limb movements during locomotion in a seemingly “simple” and automatic manner challenges neuroscientists for more than a century. At the beginning of the last century (1911) Graham-Brown postulated his “half-center” hypothesis which demonstrated the intrinsic capacity of the mammalian spinal cord to generate rhythmic motor patterns without descending or sensory input[
8]. Subsequently Grillner called these spinal neuronal circuitries central pattern generators (CPGs). CPGs are embedded within the lumbosacral spinal segments and are capable of generating stepping-like activation patterns[
6]. However, a CPG alone does not appear to be sufficient for over-ground walking. Locomotion represents the interaction between the innate pattern and an appropriate modulation of leg muscle activation which has to continuously adapt to the present requirements, e.g., to the over-ground conditions. Feedback from a variety of sources, e.g., visual, vestibular and proprioceptive systems, is interpreted by and then integrated into the activity of the CPG[
9]. The CPG can open and close reflex pathways in a context- and task-dependent manner. The sensory feedback and the context-specific requirements of the motor task determine the mode of organization of muscle synergies[
10]. Additionally, supraspinal control is needed to provide both the drive for locomotion as well as the coordination to interact with a complex environment. Corticospinal access to locomotion control in humans is phase-dependent[
11]. Brain centers can initiate CPG activity but the fundamental rhythmicity is hard-wired. For example, in the cat, it was shown that application of clonidine, a substance mimicking the action of long descending pathways, results in distinct and consistent alternating bursts of electromyographic (EMG) activity which induce spinal stepping[
12]. In humans, neuroimaging methods have revealed that distinct cortical areas, e.g., both the medial primary sensory-motor cortices and the supplementary motor areas, become activated during locomotion[
13,
14] and the size of the activated areas is related to the subjects’ walking speed[
15]. In addition, more demanding tasks, such as walking over obstacles, require more cortical control, especially during swing phase of stepping in humans[
16] and the cat[
17].
Quadrupedal locomotion is characterized by the coordination of forelimb and hindlimb rhythmic activities generated by common spinal neuronal control mechanisms, i.e., long propriospinal neurons coupling the cervical and lumbar segments[
18,
19]. This neuronal coupling and coordination of upper and lower limbs that is present in quadrupedal locomotion is preserved in bipedal gait. Unilateral tibial nerve stimulation during locomotion, but not during sitting or standing, leads to reflex responses in leg muscles and in the proximal muscles of both arms[
20,
21]. Such task-dependent coupling of thoraco-lumbar and cervical locomotor centers is flexible and allows humans to use the upper limb for fine, skilled movements, or alternatively for locomotor tasks, such as swimming or crawling, or for the control of body equilibrium during stepping, e.g., arm swing as a residual function of quadrupedal locomotion[
22].
It is important that the neuronal mechanisms underlying human locomotor control in the normal and pathophysiological condition are understood, as it is only then that it is possible to maximize the recovery of locomotion in patients following central nervous system damage.
Focusing on neuronal plasticity brought about by locomotor training
Modern neurorehabilitation no longer aims to simply compensate for disabilities in SCI subjects, rather it aims to functionally regain locomotor ability by exploiting neural plasticity and/or neural repair. A first question to ask is, “What is neuronal plasticity?” An example of neuronal plasticity is the spontaneous reorganization that is observed after SCI at the cortical level which can occur over one year[
23]. Next to spontaneously occurring neuronal plasticity, it can also be induced by locomotor training at cortical level[
24]. Plastic changes in sensorimotor cortex activity are related to functional locomotor recovery after an SCI[
25]. Besides cortical plasticity it seems that other supraspinal centers, such as the cerebellum and brainstem are also important sites of neuronal plasticity in humans receiving locomotor training after SCI[
24]. In humans, it is assumed that supraspinal plasticity is associated with plasticity of spinal neuronal circuits, but the evidence for this is predominantly from animal models of SCI[
26].
Spinal neuronal circuits below the level of lesion can be activated by an appropriate afferent input, and this is considered important to sustain functional recovery after an SCI[
9]. In contrast, typical movement disorders after SCI, e.g., spastic movement disorder, are due to the defective utilization of afferent input in combination with secondary compensatory mechanisms[
27]. It has been shown that neuronal networks underlying the generation of locomotor patterns of cats[
28] and humans[
29,
30] have an impressively high level of flexibility after SCI. Rehabilitative interventions after SCI should therefore focus on exploiting the plasticity of neuronal circuits, i.e., at supraspinal and/or spinal level, rather than focusing on improving isolated clinical signs, such as muscle tone or reflex excitability.
The plasticity of spinal neuronal circuits is task-specific and use-dependent as shown in several earlier experiments in cats with complete SCI. For example, after several months of daily step training, spinal cats regained full weight-bearing locomotion on a treadmill[
31,
32]. If a spinal cat is intensively trained to stand it develops the ability to support its body weight for up to an hour, but stepping ability on the treadmill remains poor[
33]. These findings suggest that spinal neuronal circuits learn the sensorimotor task that is specifically practiced and trained[
34]. The repetitive activation of particular sensorimotor pathways by task-specific training can reinforce circuits and synapses used to successfully perform the practiced movement[
35,
36]. Therefore, the outcome of a neurorehabilitative approach strongly depends upon the type, the repetition and the quality of the trained motor function.
Neuronal plasticity is not always positive, anatomical and neurophysiological observations in animals suggest that after SCI severed axons degenerate and create free synaptic territories which could become re-occupied by sprouting of intraspinal fibres[
37]. The new neuronal circuits may be aberrant and can lead to inappropriate movement patterns or pain in rats[
38] and humans[
39]. This can largely be prevented by a combination of locomotor training, electrical and pharmacological stimulations of the region of the spinal cord that is deprived of supraspinal input which leads to a weight-bearing locomotor capacity in spinal rats[
38]. For human SCI, it will be important to make sure that future neurorehabilitatitive approaches direct the spontaneous and/or experimentally induced (e.g., stem cells or Nogo-A antibodies) neuronal plasticity towards functional synaptic connections that are associated with an improved locomotor performance.
The decline in supraspinal and peripheral input in humans after a severe SCI is suggested to be responsible for the development of a neuronal dysfunction below the level of lesion[
40,
41]. Characteristic changes in neuronal behavior occur up to approximately one year after SCI. At one year post-injury complete and severe but incomplete SCIs are characterized by an exhaustion of leg muscle activity during assisted locomotion which is associated with changes in polysynaptic spinal reflexes[
42]. Neuronal dysfunction after SCI is assumed to be dependent more on the subjects’ loss of mobility, i.e., decreased appropriate afferent information to the CPG, than on the completeness of the injury[
43]. Consequently, the better the stepping ability of SCI subjects, the less the neuronal dysfunction. The functional state of spinal locomotor circuitries is not fixed after an SCI and neuronal dysfunction can be improved by intensive locomotor training over one month, but only in incomplete SCI subjects[
43].