A complete transection of the spinal cord, or spinalization, abolishes all descending inputs from structures rostral to the lesion (Fig. 3b). Complete SCI animal models have been instrumental in our understanding of the spinal control of locomotion. Numerous studies have investigated how the hindlimb locomotion recovers with time, training, and/or drug injections in kittens (32–37) and adult cats (38–56). With locomotor training, which is discussed later on, hindlimb walking (i.e., spinal locomotion) can recover after a minimum of 3–4 weeks. Figure 4 shows an example of kinematics and EMGs during treadmill locomotion at 0.4 m/s before (gray) and after (black) spinalization in the same cat (54). After spinalization, step length (Fig. 4a, b) and the duration of the locomotor cycle are generally reduced. In this cat, the duration of the cycle was 1,079 and 684 ms before and after spinalization, respectively. Thus, to maintain the same treadmill speed, the hindlimbs of the spinal cat must walk with a faster rhythm. Maximal angular excursions are generally reduced during spinal locomotion (Fig. 4c). The timing of most muscles is generally similar before and after spinalization, whereas the amplitude of the activity in flexor muscles typically increases while that of extensors decreases, although this varies across extensors and between animals (Fig. 4d). However, there are some notable changes in timing that are frequently observed. In particular, the delay in the onset of the knee flexor St activity with the hip flexor Srt is reduced, and both muscles can discharge simultaneously.

Dorsal lesions of the spinal cord disrupt the dorsal columns and the cortico- and rubrospinal tracts, depending on the extent and location of the lesion (Fig. 3c). Dorsal lesions of the spinal cord have been used in numerous studies to evaluate the recovery of quadrupedal locomotion (57–61). In adult cats, voluntary quadrupedal locomotion recovers within 10 days after dorsal lesions (61). Bilateral damage to the dorsal/dorsolateral spinal cord at low thoracic levels results in longer steps, paw drag during the early swing phase, and an impaired ability to step over obstacles. Paw drag can persist in cats with more extensive lesions. With large lesions, the duration of the cycle and of the swing phase can increase. Posture and the coupling between hindlimbs are only transiently impaired following a dorsal lesion. The delay in the onset of the knee flexor St activity with the hip flexor Srt is reduced and both muscles can discharge simultaneously, similar to what is observed following spinalization.

Ventral lesions of the spinal cord disrupt vestibulospinal and reticulospinal tracts, as well as serotoninergic and noradrenergic pathways. Ventral lesions have been studied in fewer studies (62–66) because they are harder to perform selectively. Following damage to the ventral/ventrolateral spinal cord at low thoracic levels, there is an initial period with loss of hindlimb walking and weight support that depends on the extent of the lesion (63–65). After a few days to a few weeks, hindlimb locomotion and weight support gradually recover. Despite the recovery of quadrupedal locomotion, lateral stability remains poor. Moreover, the coupling between the fore- and hindlimbs was altered following ventral lesions, as the frequency of the locomotor rhythm differed at both girdles in the most severely injured cats. In addition, the deficits were more pronounced during incline treadmill walking.

A lateral hemisection is made by sectioning half of the spinal cord on the left or right side at a specific segment, which is akin to a Brown-Séquard SCI (see Fig. 3e). The lateral hemisection is a popular model to investigate the recovery of hindlimb locomotion in cats (67–80). During the first few days after lateral hemisection, the hindlimb on the side of injury displays flaccid paresis and drags on a moving treadmill belt (70, 71, 75, 76, 80). If the lateral hemisection is performed at cervical levels, both the forelimb and hindlimb on the side of the lesion become paretic for approximately 1 week (70, 71). Animals require assistance for hindquarter support and body equilibrium during locomotion, although the deficits can vary between animals. Within 2 weeks, hindlimb weight bearing recovers and the affected hindlimb progresses from a passive crutch with minimal limb excursion to an active locomotor pattern (75, 76). With larger lesions, the coordination between the fore- and hindlimbs (i.e., intergirdle coordination) can be permanently lost, which is characterized by a different rhythm at the two girdles (80). The strategy used to adapt quadrupedal locomotion after lateral hemisection varies between cats and appears to depend on the presence or loss of intergirdle coordination.

Three different types of dual lesion paradigms are illustrated in Fig. 5. A few studies in cats have used two spinal lesions separated in time (i.e., staggered) to determine what structures might be responsible for the recovery of hindlimb locomotion after SCI. For instance, Kato and colleagues (70, 71) performed a unilateral hemisection at C2 or T12, which was followed several weeks later by a second unilateral hemisection on the opposite side of the spinal cord at mid-thoracic levels (Fig. 5a). Such staggered lateral hemisections disrupt long supraspinal and propriospinal pathways. In all cats, a voluntary quadrupedal locomotion recovered within a few days to a few weeks. However, the coordination between the fore- and hindlimbs was permanently lost. Kato also performed a hemisection at L2–L3, which was followed a few weeks later by a caudal myelotomy (73, 74), which is a longitudinal split along the midline of the spinal cord (Figs. 3f and 5b). The hemisection followed by myelotomy isolates the spinal cord controlling one hindlimb. With such lesions, locomotion initially consists of a tripod gait involving the forelimbs and the nonisolated hindlimb. A few days to a few weeks later, the isolated hindlimb recovers its walking ability, although the coordination between hindlimbs is lost.

More recently, another type of dual spinal lesion paradigm was used (78–80), which consisted of a lateral hemisection at low thoracic levels followed several weeks later by a complete spinal transection (Fig. 5c). All cats recovered a voluntary quadrupedal locomotion following the hemisection. What was most remarkable was that following the spinalization, hindlimb locomotion spontaneously recovered within hours, a process that normally takes 3–4 weeks of intense treadmill training in chronic spinal cats. The underlying mechanisms involved in locomotor recovery with dual spinal lesion paradigms are discussed later on.

Therefore, different types of spinal lesions produce characteristic deficits that reflect in part damage to specific pathways. What is clear, however, is that no descending pathway is essential for the recovery of a voluntary quadrupedal locomotion following an incomplete SCI.

Arguably, the most important discovery of the twentieth century for the control of walking was the demonstration that the spinal cord could generate the basic rhythm of locomotion by itself, which was first done in the adult cat. Thomas Graham Brown, a Scottish physiologist, transected the spinal cord of adult cats at the 12th thoracic (T12) segment and performed an extensive deafferentation of the hindlimbs by sectioning muscle nerves, thus abolishing supraspinal and peripheral inputs to the spinal cord (81). Only the nerves to the two test muscles, the gastrocnemius (ankle extensor) and tibialis anterior (TA, ankle flexor), were left intact, and sensory inputs from these muscles to the spinal cord were abolished by sectioning their dorsal roots. Despite the absence of inputs to the spinal cord, rhythmic alternating activity of the ankle extensor and flexor muscles occurred, which was evaluated by recording movement of the tendons of both muscles. This groundbreaking demonstration indicated that the generation of a rhythmic locomotor-like pattern is intrinsic to the spinal cord. The work by Brown is summarized in an excellent review article (82).

Despite what now seems like a clear demonstration for a spinal origin of locomotion, Brown’s findings did not garner much support among his peers. The prevailing view that locomotion was of “reflex” origin (83), persisted into the 1960s. It was not until the work by Anders Lundberg and his colleagues in Sweden on spinalized cats (84–87) that Brown’s findings resurfaced and became widely accepted. The assumption that an autonomous locomotor network also exists within the human spinal cord has greatly impacted therapeutic interventions in human SCI patients, although a definitive experimental demonstration of a spinal locomotor CPG in humans will never occur. For reviews discussing the existence of a spinal locomotor CPG in humans, see (88–90).

The spinal CPG for locomotion is also largely genetically determined, as kittens spinalized early after birth, before performing any walking movements, will express spinal locomotion (32–34). However, some elements of the spinal locomotor CPG are malleable and most likely contribute to the recovery of hindlimb walking after SCI (91). Experimental data from cats have been used to ­conceptualize or model the organization of the spinal locomotor CPG (92–96). At present, the functional organization of the mammalian locomotor CPG remains elusive, and extensive work in rodents and cats is ongoing. Therefore, a system of interconnected interneurons within the spinal cord organizes the basic pattern of locomotion, which can function autonomously from supraspinal structures following a complete SCI to restore walking ability. However, it should be emphasized that the CPG after complete SCI interacts with tonic and phasic sensory inputs from peripheral receptors located in joints, muscles, and skin. Thus, a better term is spinal “pattern generator (PG)” to denote an intrinsic spinal neuronal network that produces locomotor activity with inputs from the periphery. As discussed later on, the spinal locomotor PG is also critical in the recovery of locomotion following an incomplete SCI, even though supraspinal inputs can still access the spinal circuitry.

Locomotor training is based on the premise that providing sensory cues consistent with walking induces activity-dependent plasticity in the spinal network that generates locomotion (reviewed in (97, 98)). Locomotor training after a complete SCI was developed in adult cats during the 1980s concurrently in the laboratories of Serge Rossignol at the Université de Montréal (41) and of Victor “Reggie” Edgerton at the University of California at Los Angeles (38).

In the first few days after complete SCI, the legs of the cat are paretic, with little to no reflex movements or muscle spasms. During this period, the legs drag on the treadmill in an extended position with weak or absent spontaneous activity. Locomotor training consists of two experimenters moving the hindlimbs in a pattern consistent with walking and placing the plantar surface of the paws on the treadmill belt while it moves. The forelimbs are positioned on a stationary platform. After a few days, reflex movements and muscle spasms emerge and locomotor-like activity can be triggered by stimulating the perineal region (i.e., the skin located below the tail). At this point, the locomotor pattern is disorganized, with inappropriate contacts of the foot dorsum with the treadmill belt, short step lengths, inconsistent bouts of activity, and poor weight support. A Plexiglas separator is placed between the hindlimbs to prevent crossing of the legs. After a minimum of 3–4 weeks of locomotor training, 3–5 times a week, cats regain the ability to generate an organized hindlimb walking pattern with plantar foot contacts and weight support (38, 41, 42). However, postural control is permanently lost and the experimenter must hold the tail of the animal to provide equilibrium. The time course of recovery and the quality of spinal locomotion vary from one cat to another.

It was shown, in the adult cat, that functional recovery after complete SCI is specific to the type of training (99). For instance, cats that were trained to stand performed poorly at walking, whereas cats that were trained to walk had weak standing capacity. However, based on personal observations, a combination of stand and locomotor training works bests for locomotor recovery (unpublished observations). Combining stand and locomotor training appears to act synergistically, most likely by strengthening hindlimb extensor muscles, which is key to recovering weight support. The Edgerton group also showed that the spinal cord can “remember.” For example, when spinal cats were locomotor-trained for 12 weeks and training was then stopped for 12 weeks, the capacity for hindlimb walking diminished considerably (99). However, if locomotor training was restarted, hindlimb walking returned much more rapidly than initially observed. Therefore, locomotor training can facilitate the recovery of hindlimb walking following complete SCI, training appears to be task-specific, and the spinal cord devoid of signals from supraspinal structures can “remember.”

Although Sherrington’s original hypothesis that locomotion was generated by a reflex mechanism was inaccurate, it is evident that inputs from reflex pathways are critical in shaping the locomotor pattern through dynamic interactions with the spinal locomotor CPG (comprehensively reviewed in (26)). Moreover, sensory feedback to the spinal cord after SCI is critical for functional recovery, which was first made evident in the cat. For instance, early after spinalization, stimulating the perineal region facilitates the emergence of locomotion in adult cats. The skin from the perineum is thought to provide general excitability to the spinal locomotor network. If sensory inputs to the spinal cord are abolished before SCI, either by sectioning muscle nerves (56, 100), cutaneous nerves (53), or dorsal roots (101), the recovery of spinal locomotion is compromised (e.g., poor weight support, disorganized locomotor pattern). Some animals denervated before spinalization will not recover spinal locomotion (56). It thus appears that the integrity of sensory pathways at the time of SCI is critical for functional recovery thereafter.

Some sensory cues have more important roles during locomotion in the cat than others. For example, the position of the hip at the end of stance (102) and the amount of loading on ankle extensors (23) influence the stance-to-swing transition. Proprioceptive inputs from hip flexors or ankle extensors regulate phase transitions and can entrain the rhythm during fictive locomotion in adult cats (103–111) and/or treadmill walking in unparalyzed decerebrate cats (112–117). Experimental studies in humans successfully translated some of the cat data, primarily for sensory inputs from ankle extensors (118–120), thus showing the importance of sensory cues during bipedal human walking. The importance of hip position and ankle loading is stressed during rehabilitative walking in human spinal cord-injured patients (98).

Inputs from the skin, particularly from the foot, also have specific functions during locomotion. For example, mechanically stimulating the dorsum of the foot during the swing phase evokes a coordinated response that removes the stimulated limb from the obstacle in intact and spinal cats (121, 122), as well as humans (123–125). Removing skin inputs from the foot by sectioning the five cutaneous nerves produces only transient deficits during treadmill locomotion in intact cats (126). However, after complete SCI, these same cats were unable to correctly place their paws on the plantar surface despite several weeks of locomotor training (53). In addition, if the cutaneous denervation was performed after spinalization, some locomotor deficits appeared, such as paw drag during swing, but only once all skin inputs from the foot were removed. Therefore, inputs from the skin of the foot become increasingly important for locomotion after SCI.