Rehabilitative Training in Animal Models of Spinal Cord Injury

Training of hindlimb function

Rehabilitative training in animal models of thoracic SCI has generally focused on locomotor function. These approaches are based on the principal of promoting repetitive movement of the hindlimbs in one specific task, including treadmill walking, biking, running on a wheel, swimming, or over-ground walking training. Treadmill training is one of the most common methods to provide locomotor rehabilitation in SCI in both pre-clinical ⁶⁰ and clinical ⁶¹ settings. This is based on the discovery that spinalized cats could be trained on a treadmill to recover weight-supported stepping of the hindlimbs ^62,63 by promoting plasticity of the spinal circuit pattern generator (CPG).

For pre-clinical research, treadmill walking allows one to adjust the training for different injury types and severities as well as the option to control intensity by modulating the belt speed and/or the incline. Limitations include the fact that training (in general) is labor-intensive, and it is difficult to isolate training to the hindlimbs because forelimbs are also engaged, ⁶⁴ and the need to provide weight support for animals with severe lesions. Some of these limitations have been overcome by adding technological adaptations to the task, such as integrated weight support systems, ^65,66 forelimb platforms, ⁶⁷ or even robotics to allow assisted walking and biped posture in rats. ^{68 –70} However, the mechanism by which training promotes recovery might be different when these modifications are introduced. ^64,71

An alternative to treadmill training in rodents is wheel running, which requires the animals to actively participate in the task. ^72,73 Wheels are generally installed in the home cage where motor activity can be easily assessed and where intensity can be modulated by changing the wheel resistance. ⁷⁴ However, to accurately track running wheel usage for each animal, animals normally need be housed separately, thus precluding the use of multiple animals per cage. For social animals like rodents, housing individually is not recommended for ethical reasons. Introducing microchips that measure an individual's movements might be a solution to allow multiple animals per cage. The wheels themselves also pose a challenge given that they require large cages. Nonetheless, wheel training has been shown to promote over-ground locomotor recovery. ⁷⁵ Further, wheels can be motorized, which allows for change in speed (i.e., intensity of training) that can be adjusted for the abilities of mild and moderately injured animals. ⁷⁶ Another alternative similar to wheel running is ball training. Although less used, placing animals for a limited amount of time in a ball and allowing them to free run over a big surface have been used to promote recovery after SCI. ⁷⁷ Nonetheless, wheel and ball running limitations are similar to treadmill training in the sense that forelimb engagement and residual hindlimb movements are required. An interesting difference between wheel and treadmill training is the free or ad libitum training provided by incorporating wheels to the animal's home cage, which will be discussed later.

Another possibility to potentially train locomotor circuitry of rodents with SCI is bike training, where the feet are mounted on pedals, which are mechanically moved. ^10,78,79 This is different to FES (functional electric stimulation) biking in humans, where muscles are activated to move the pedals. ⁸⁰ By delivering rhythmic cycling movement to the hindlimbs, training can be isolated from the forelimbs and no residual motor control is required. These movements will thus provide orchestrated sensory feedback to CPGs, induce cortical reorganization, prevent muscle atrophy, and improve cardiovascular function. ^{10,78,81 –83} Considering that up to now it is unclear whether training has to be task specific or whether exercise in general is beneficial (e.g., by increasing BDNF levels) ²⁵ makes bike training an interesting training alternative, especially for animals with severe injuries. Although it is easy to regulate training intensity, the passive nature of the training (i.e., no intentional movements required by the animal) is a potential limitation for promoting functional rewiring. Further, for rodents biking is an artificial task, making this a useful task to study the importance of task-specific training and the transferability of training effects to untrained tasks.

Another approach to train locomotor function is swimming, where animals are placed in a tank of water to encourage hindlimb use. Depending on water levels, swimming can easily turn into walking with weight support (because of the buoyancy). Thus, swimming or locomotor training in water can overcome some of the restrictions of treadmill and wheel training by providing a controlled amount of weight support to perform hindlimb movement. ^84,85 In order to provide sensory feedback for swimming animals, a substrate can be added so the paddling legs would touch it when extended. ⁸⁶ Swim training has various challenges: the significant amount of stress the animals are exposed to, and the fact that with severe injuries rodents will either switch from the use of hindlimbs to forelimbs or abandon active swimming altogether. ⁸⁷ Few studies on swimming training have been performed, and our current knowledge of its efficacy is slim.

The differences between gross and skilled locomotion (i.e., locomotion on irregular or space-limited surfaces such as horizontal ladder or a narrow beam that requires fine motor control) are valuable given that they may target different pathways. For example, a frequently studied neuronal tract is the corticospinal tract (CST), which is important in humans and rodents in fine motor control and is easily accessible for tracing and anatomical analysis in animals. Although CST plasticity has recently been demonstrated to contribute to recovery of voluntary locomotor function in rodents, ^21,68 it is generally acknowledged that CST function is less important for over-ground locomotion in these animals. ^{88 –90} This is highlighted by the fact that rats with CST-specific lesions (e.g., pyramidal lesion) can walk very well. ⁸⁸ Consequently, one could argue that locomotor training is not the best avenue to promote post-injury CST plasticity. This, however, can likely be achieved with training skilled walking, such as walking on a horizontal ladder ^{91 –94} or by spacing the rungs of a running wheel further apart. ⁴¹ Here, the correct paw placement is based on forelimb hindlimb coupling and CST integrity. ^90,95 Ladder walking has been utilized, for example, to test the time window of opportunity for successful training after SCI in rats. ⁹² Other advantages of training to walk along horizontal rungs is that this type of stepping is not performed in the cage, thus limiting self-training, and that relatively small SCI models can be utilized. Conversely, a limitation of this task is that animals with severe injuries cannot be trained.

Training of forelimb function

After cervical SCI, rehabilitative training of the affected forelimbs is commonly delivered by reaching and grasping tasks, including the tray and other multi-reaching task, the Montoya staircase task, and the single pellet reaching and grasping (SPG) task. The SPG task consists of presenting a food pellet to the animal beyond a slit. To complete the task, the animals must reach through the slit, grasp the pellet, and bring it to the mouth. Once the animal moves away from the slit, a new pellet is presented. Although originally designed to test forelimb movements, ^{96 –98} SPG has been successfully used to deliver rehabilitative training in rodents after SCI. ^{15 –17,20,99,100} Derivations of this task are the pull ^101,102 and the supination ^103,104 task. Here, the animal has to operate the target (usually a handle) by applying a pulling or supinating force to get rewarded, allowing for isolation of certain components of the forelimb function. Nevertheless, the usability of those tasks as training paradigm have yet to be explored. Multi-reaching tasks, in contrary, allow animals to maintain their position in front of a tray ^96,105 or other dispenser and repeatedly reach for and grasp pellets through one or more slits. Although this can increase the number of repetitions, because there are multiple potential targets each reach might be different to the previous one, reducing the specificity of the task. A different approach is the staircase method reported by Montoya and colleagues in 1991. ¹⁰⁶ This method was also originally developed to test forelimb function, but has also been successfully applied as a rehabilitative training task. ¹⁰⁷ For this, animals are placed in an enclosure with floor openings to individual compartments for the left and right forelimb to reach for food pellets or seeds. In each compartment, a stair with wells containing pellets/seeds are placed in such a way that the closest pellets/seeds to the animal are rostral to its shoulder. When animals have reached for the closest pellets/seeds, they must reach down a bit farther to get new pellets/seeds. Finally, another method to train forelimb function is to use holes in the floor of an enclosure and allow animals to reach and grasp for food pellets or seeds located in the holes. Training using this approach has been shown to promote functional recovery, especially when combined with a pharmaceutical treatment. ^14,108

A common characteristic of all these tasks is the necessity of a significant amount of residual motor function allowing the animal to be trained. At the same time, the volitional nature of forelimb function in comparison to locomotion, which is orchestrated by spinal circuitry, makes this training especially challenging. Animals have to be introduced to the task before injury, and slight variations in the lesion severity or location can profoundly affect the capacity of the animal to perform the task. ^109,110 Nonetheless, reaching and grasping training, contrary to locomotor training, allows for better isolation of the training effect by providing rehabilitation only in one specific task, by limiting the effects of exercise and cardiovascular exertion, as well as self-training in the home cage. Which task is more appropriate for forelimb training depends on the purpose of the study and functional deficits of the used injury model. Severe injuries preventing shoulder mobility cannot be rehabilitated using SPG (nor pull or supination task) and tray or holes in the floor should be chosen. Contrary, if the goal is the recovery of fine control of the reaching and grasping function, SPG, pull, or supination tasks are better suited.

Enriched environment

Rodents in the laboratory setting are generally housed in extremely deprived environments consisting of only one or two interactive and/or social enhancing components (e.g., a tube and nesting material), if any. Enriching animal housing by adding a variety of devices and/or objects with which the animals can interact “normalizes” their environment and increases their overall motor activity. ^49,58 Thus, EE provides a form of ad libitum self-training. EE might include different housing levels with ropes, beams, tubs, ladders, wheels, pellet reaching devices, etc., encouraging various forms of activity or training. These systems are useful by not limiting training to a specific schedule or task, especially given that rodents are nocturnal. However, the major downside is the high variability between animals, probably caused by differences in motivation and hierarchical social dynamics. The amount of training of the individual animals can be measured by using microchips and readers in the devices and can be controlled by limiting the access of each animal to the enriched system when a desired training limit is reached. ⁴⁹ Although this might reduce variability, controlling the amount of time and training each animal gets in a specific task is challenging. Even more important, training in multiple tasks during the same period can interfere with the recovery with increased performance in highly trained task(s) and poor performance in the un- or less trained tasks. Despite these limitations, EE has been shown to increase recovery of fore- and hindlimb function in animals with SCI, ^14,49,58,111 proving its versatility to cater to specific injury types, locations, and severities. If EE is chosen as the training method, we recommend studying the individual activity of each animal and the possible interactions between the different tasks included in the environment.

Self-training and excessive testing

Depending on lesion severity, a certain degree of recovery occurs in both humans and experimental models of SCI. ^113,114 A big difference, and a great opportunity to study the role of motor activity, is that rodents with severe thoracic injuries are able to partially recover because of self-training. Self-training occurs when animals are capable to repeatedly trigger or attempt specific movements (or movement patterns) in the home cage. Animals will be able to drag their hindlimbs and the resulting combination of sensory feedback with the hip extension can trigger flexion movements, the first stage in the recovery of hindlimb stepping. In rats with thoracic SCI, when hindlimb movements are restricted using a “wheelchair,” self-training is nearly eliminated and recovery is dramatically reduced. ³⁰ In contrast, enhancing hindlimb movement in the home cages by using an EE or with running wheels improves the recovery of over-ground locomotion. ^49,75 This suggests that self-training, and the resulting engagement of residual motor function, is a key component of spontaneous recovery. This poses a challenge in the study of rehabilitative training focusing on locomotor function given that it is difficult to exceed the spontaneous recovery achieved by self-training. For instance, treadmill training of rats with incomplete SCI has been found to be ineffective, likely because the intensity of the treadmill training was negligible in comparison to the self-training of the animals in the home cage as previously suggested. ^112,115,116 In addition, insufficient training intensity and repeated testing (which can act as training) have also been proposed for the absence of recovery by treadmill training. ^112,115 Testing for the treated function must therefore be considered a training session. When testing occurs too frequently, the cumulative training effect can enhance recovery, even in animals without an intended training regime, thus underestimating the real effect of training. How much testing is too much is debatable. Testing three times a week of a reaching task ¹¹⁰ might be excessive, and we do not know whether testing once a week is already sufficient to induce a training effect. A solution, although cumbersome, is determining the minimal frequency and dosage of a chosen training/testing task to induce improvement beyond spontaneous recovery before carrying further research.

Onset of training

In humans and animal models, spontaneous recovery occurs within the first few months and weeks respectively after SCI. ^117,118 After this period, further functional gains are minimal, suggesting that CNS plasticity decreases over time. Accordingly, starting training sooner rather than later after SCI increase the chances for functional improvements both in experimental ^11,60,77,92 and clinical ¹¹⁹ settings. Despite this general finding, the question of when to start the training is not clear. An early onset, within 1–2 weeks, of treadmill training is critical for locomotor recovery after thoracic SCI. ⁶⁰ Likewise, the development of neuropathic pain after SCI is prevented by early wheel training (starting day 5 after injury), ¹²⁰ but not when training is delayed for 2 weeks. ¹²¹ In rats with a thoracic contusion, delaying ball training form 1 to 7 days after injury suppresses the improved over-ground locomotion. ⁷⁷ Similarly, starting horizontal ladder training within a week after a dorsolateral thoracic SCI showed better recovery than starting the same training 3 months after injury. ⁹² After cervical SCI in rats, recovery of the forelimb function can be achieved by early (4 days after injury) onset of reaching training. ^20,122 Conversely, delaying the onset of the same training months after injury, recovery is limited or nonexistent. ¹⁵ Thus, a common motto for when to initiate training would be: “the sooner, the better.” But, how soon is too soon? This question is difficult to address after traumatic SCI, because the onset of intensive motor training after injury in individuals is extremely variable depending, for example, on injury severity and trauma to bones, organs, or soft tissue. In the field of stroke rehabilitation, data suggest that starting training within a few days after injury can be detrimental. ^123,124 Although the negative effect and the underlying mechanisms of immediate training after injury are still under debate, a “too early” effect might also exist after SCI. Starting treadmill training in rats with SCI before 72 h after injury is ineffective or event prevents recovery in rats. ¹²⁵ Similarly, starting reaching training 4 days after SCI in rats results functional recovery in the training task but affects performance in another nontrained task. ²⁰ Importantly, this downside of an early training on other tasks might be preventable by delaying the onset of training for a few days without affecting the recovery in the training task. ¹²² Additionally, delaying passive hindlimb cycling training for 1–5 weeks after thoracic transection in rats prevents the increase in the number of autonomic dysreflexia events induced by early training, without affecting the cardiovascular function gained by training.

Nevertheless, in terms of improving the trained task, evidence suggests that delaying training onset weeks or months after SCI hampers the success of the treatment. In general, starting within 2 weeks after injury is better for maximizing recovery, although the window of training onset must be determined for each training and injury model.

Duration of training

Training duration can be defined as how much time per session or for how long (weeks and months) the training regime should be applied. Here, we refer to the latter, leaving the former to the next section. A systematic revision of locomotor training in rodent models of SCI found that the duration of training impacts functional recovery. ⁶⁰ Five days of treadmill training starting a week after SCI is enough to influence soleus muscle function and over-ground locomotion. ¹² Three weeks of treadmill training after a thoracic hemisection in mice induces functional recovery ¹²⁶ and synaptic changes in spinal interneurons, ¹²⁷ changes that were more prominent after 6 weeks of training. Similarly, while 3 weeks of treadmill training in mice reduce muscle atrophy, 9 weeks of training was more effective. ¹²⁸ After cervical SCI, 4 weeks of reach training is sufficient to increase expression of the growth-associated protein 43 in the contralateral cortex to the lesion site, translated to functional recovery after 6 weeks of training. ²⁰ Taken together, these findings suggest that changes in the CNS can be achieved by a few days (or weeks) of training, while perceptible functional improvement might need longer training periods. Nevertheless, the minimal training duration to maximize recovery likely differs for each injury type, training modality, and animal model.

Another important consideration of any treatment is the retention effect (i.e., if the benefits of such treatment last after the therapy stopped). Rehabilitative training should be provided for enough time to induce recovery, but also to retain it after training stops. Unfortunately, only a few studies have tested for retention of the recovered function after experimental SCI. The combination of reaching training and amphetamine ⁵⁰ enhanced the qualitative reaching function in the SPG task after cervical contusion in rats, even 4 weeks after both drug treatment and training were terminated. In other experiments, the functional recovery obtained by 13 weeks of electrical stimulation combined with training after cervical SCI was retained 3 weeks after the intervention stopped. ¹²⁹ Conversely, the gain in cardiac and hemodynamic function induced by 4 weeks of passive hindlimb cycling is reduced 4 weeks after training has stopped. ¹¹ Although that might be related to the cardiovascular component of the training that is reduced when the training stops, it shows how some beneficial aspects of training diminish after treatment. Beyond the obvious cardiovascular benefits of physical activity, the exercise component of training might favor the retention effect through BDNF. Rats with access a running wheel for 3 weeks performed better than sedentary animals in a memory test with an associated increased level of BDNF for weeks after the exercise stopped. ⁷² Considering that training after SCI is accompanied with increased BDNF expression and the role of BDNF in the formation, guidance, and maintenance of neuronal circuitries, ^23,24,130 the retention of a trained function could be associated to the capacity of the task to increase BDNF levels and maintain those after training is finished.

When applying rehabilitative training, we recommend determining the minimal duration of the treatment to foster recovery (probably weeks to months) and to test for the retention of functional gain once the intervention stops.

Dosage and intensity of training

The amount of rehabilitative training per session of training (number of repetitions of the training task during a session) and its intensity (the number of repetitions per time unit) have been associated with locomotor recovery after stroke ^131,132 and, more recently, after SCI ^133,134 in humans, but remains poorly studied in pre-clinical experimental models of SCI. ⁶⁰

The number of steps during treadmill training in rats with SCI determine the extent of recovery. ^66,135 Likewise, functional recovery using EE correlates on the individual amount of activity. ⁴⁹ In a cervical SCI model in rats, delivering a low amount of rehabilitative training in the SPG task (20-30 pellets per 10 min) soon after injury results in motor recovery. ^16,20,122 However, starting the same training months after injury only show benefits in recovering grasping if intensity is increased by 3-4 times. ¹⁵ This may reflect the importance of increased training intensity as a way to deliver enough repetitions to promote plasticity, and thus functional improvements. Nevertheless, increasing the number of repetitions per time unit might also be detrimental; prolonged sessions (hours) of high-intensity reaching training in rats can cause bone pathologies ¹³⁶ and repetition-associated musculoskeletal disorders. ¹³⁷ Thus, determining intensity of training to maximize recovery while preventing side effects is necessary. In this regard, comparing pre-clinical studies that incorporate training faces a problem: the lack of reporting training intensity. Studies that showed recovery induced by reaching training after cervical SCI ^{14,107,108,138} do not report the number of repetitions per session, and only reported the session duration. Similarly, not all locomotor training studies report intensity as number of steps or speed per session. ⁶⁰ This paucity in reporting makes it difficult to determine the influence of training dosage and intensity on recovery, and we strongly encourage researchers to measure and describe training details when publishing their findings.

Passive, active, and ad libitum training

Active training can be described as volitional, which stands in contrast to passive training, which can be performed without the intention to move. Two examples are unassisted walking versus walking in a fully automated exoskeleton, or bicycling by FES biking where muscles are activated by stimulators. ⁸⁰ It has to be recognized that both active and passive training approaches have beneficial effects, and that passive training is frequently applied for reasons other than to improve descending motor control (e.g., cardiovascular, muscle, or motoneuron function). It is also fairly accepted that following Hebb's idea of “firing together wiring together” successful training of motor tasks requires concomitant activity of descending pathways and spinal circuitry. ⁶⁸ Nevertheless, even passive training showed powerful effects in modulating spinal circuitry, which becomes obvious from training animals with complete spinal lesions. In these cases, in which all descending inputs are cut, spinal circuitry can be recruited to adjust weight support and to produce rhymical stepping, thus illustrating the capacity of the spinal cord to learn. ^62,63 Similarly, bike training was found to be beneficial for muscle and motoneuron function. ^10,82

Another consideration when selecting a training approach is the difference of forcing training or letting the animals choose when to train. Some training tasks require the animals to be placed in a specific apparatus, thus limiting training to a pre-determined schedule. In other approaches, training can be provided by systems attached to the animal's home cage allowing for ad libitum or free training, such as with EE, ^49,52,139 an automatic pellet dispenser, ^140,141 or running wheels. ⁷² Letting animals choose when to train can be more practical given that rodents will usually train more at night when they are more active, less researcher time is needed during the training process, and training can be “continuous” instead of just a few minutes to hours a day. ¹⁴⁰ In addition, the training task is self-motivated given that activities in the home cage have to be interesting enough that animals will participate, ^49,58 probably reducing the stress that might cause daily sessions of forced training. There are also limitations that can influence the success of free training. The individual dosage of training is more difficult to control during ad libitum rehabilitation, introducing variability within experimental groups. Further, if animals are group housed, hierarchic structures might influence the accessibility of a task; and if animals are housed individually, stress and depression-like behaviors ¹⁴² might affect training efficacy. With the development of computer-aided methods, the constraints of free training can be reduced. For instance, similar to EE, ⁴⁹ access of an animal to a pellet in a reaching and grasping task or to a running wheel could be restricted to a certain amount of time or repetitions a day if the system incorporates an automatic identification of the animals. Even so, ad libitum training and its comparison to methods where training is administered according to a specific schedule are still insufficiently studied.

Motivating the training task

Animals, as humans, when exposed to a repetitive task will lose motivation over time. This is especially true after injury, when it is difficult to achieve success in the task. To keep animals' interest in the training task, some enforcement or reinforcement is, sometimes, required. The use of uncomfortable, but painless, electrical shocks to keep animals walking on a treadmill, ¹⁴³ restricting the food in the home cage hours before a task with a food-reward–based task such as the SPG, ^15,20 or simply coupling the success of a task with a tasty reward ¹⁰¹ are some of the techniques to encourage training. As much as motivating animals to train might be unavoidable, how motivation is achieved can affect the success of training. Food restriction, for example, to encourage reaching for food pellets raises the number of reaches performed per session, but if animals are too hungry, they may show anxiety-like behavior, eagerly repeating the task without focus and ultimately reducing reaching success. ¹⁴⁴ On the other hand, feeding sucrose pellets to motivate reaching can result in high sugar intake, which on its own can have side effects. Other reinforcement techniques may directly interfere with learning of the training task. Electrical shocks to the tail and/or the plantar surface of the foot to encourage animals to continuously walk on a treadmill ¹⁴³ might cause aberrant afferent information to the spinal cord, modulating plasticity. ³⁴ Thus, the way training is motivated can have serious implications and should be carefully considered when interpreting or comparing the results.

Pre-training matters

Another challenge in the design of rehabilitative training experiments is the potential influence of pre-injury experience. Natural tasks, such as walking, running, or swimming, require less exposure to the task preceding the SCI compared to a more specialized task such as SPG. Thus, as a trained task gets more similar to a “natural” task/behavior there is a concomitant increase in the risk of self-training in the home cage. This also means that learning of a new skill is not necessary for natural tasks and the neuronal circuitry to support such motor function is already present. Contrarily, exposure to a specific task may require learning of the task, which thus requires changes in the connectivity of the CNS. ^24,145 Consequently, one could argue that the degree of pre-injury training might influence the outcome of rehabilitative training after injury. For example, the amount of SPG training and how well the animal is trained before injury determines how fast animals can effectively train thereafter. ¹⁴⁶ Further, exposure to a task before an injury can interfere with post-injury training ¹⁴⁶ or accelerate motor relearning. ¹⁴⁷ Therefore, previous experiences with a given task before injury may modify the outcome of training. When applying training, individual experience of each animal to the training task preceding the injury must be quantified (i.e., pre-injury baseline) and recovery should be referred to that baseline. Additionally, pre-injury experience to the task can serve a balancing factor to distribute variability between experimental groups.

Transferability of the trained task

A common question when applying task-specific training is the transferability (the capacity to convey a learned or regained skill to a new context) of the trained task to other tasks. Transferability can be achieved if a learned skill is transferred to a different nontrained task. When training promotes execution of the trained task, but does not affect different tasks, then the effects are considered task specific. An example of task specificity comes from treadmill, swim, and wheel training, which does not always transfer to over-ground stepping. ^{72,84 –86,115} Others, however, have reported transferability between a variety of training paradigm and nontrained task. ^{14,122,148 –151} Contrarily, training in a task might interfere with the performance of another task. ^20,152 It is not clear whether transferability is dependent on the similarity of the tasks. For example, in a cervical CST injury, rats, training in a task to retrieve pellets through holes in the floor, failed to induce recovery in the Montoya-type staircase test, but did in the SPG task, ¹⁴ where it could be argued that the staircase test is closer related to the trained task. Interestingly, delivering training in a very similar way, by reaching for seeds through holes in the floor, subsequent to a selective CST and rubrospinal tract injury, the same researchers failed to find improvements in the SPG task. ¹⁰⁸ The researchers then proposed that this training alone did not promote any recovery. However, given that the researchers did not test performance in the trained task, the lack of effect might be based on a lack of transfer. Therefore, slight differences in training protocols and the injured neural circuitry might influence transferability. In a different set of experiments, horizontal ladder or SPG training (two fairly different tasks) induced recovery in the trained task, transferred reciprocally (i.e., SPG trained animals improved in ladder walking and vice versa), and even enhanced performance in a nontrained staircase test in animals with a CST injury. ¹⁵⁰ A potential explanation for transferability and specificity is a training specific rewiring of the neural resources, favoring or competing with other functions. ¹⁵³ SPG training after incomplete cervical SCI induces reorganization of injured and noninjured motor pathways, ²⁰ probably explaining the alteration of function in the unaffected forelimb. ¹⁵⁴ Similarly, the reorganization of spared spinal interneuronal networks is likely promoting recovery of hindlimb function induced by engaging forelimb during training. ⁶⁴ Therefore, testing for the trained and nontrained task should be systematically included in studies with rehabilitative training to ponder its benefits and side effects.

Compensation, qualitative measures, kinematics, and longitudinal studies

In order to translate pre-clinical results to the clinic, functional testing must be carried out carefully to delineate compensatory mechanisms from true recovery. ¹¹² For example, during recovery of forelimb function induced by reaching training, rats can develop compensatory movements such as dragging the pellet (a.k.a. scooping) instead of a precision grip. ^14,15,155 Similarly, increasing trunk muscle tone and biomechanical stability, rather than just improving function in the hindlimbs, contributes to locomotor recovery. ^{156 –158} By quantitative analysis, such as the success rate in the reaching test (i.e., number of successful eaten pellets in proportion of number of reaches) or maximal walking speed or distance, one might conclude that training induced recovery. However, it will remain unknown whether this recovery occurred by compensatory or restorative mechanisms. Thus, the addition of qualitative measures, such as kinematic and movement pattern analysis, ^{97,156,159 –161} that can inform about the nature of success (e.g., better wrist supination, higher intra- and interlimb coordination, etc.) is warranted. In addition, slight modifications of the trained task designed to prevent the use of compensatory strategies can be useful to quantify the influence of compensation in the recovered function. For instance, the introduction of a space or gap between the pellet and the animal during the single-pellet reaching task prevents the animal from scooping pellets. ¹⁵ By comparing the trained task with the task that prevents compensation, it is possible to account for how much recovery is attributed to that compensation.

Another important consideration is studying how functional recovery evolves over time. Measuring outcomes of recovery at one single time point, normally at the end of the experimental period, might not represent the full benefit of training. Longitudinal studies, where outcomes are measured over a period of time, increase our knowledge of the effects of training. The number of time points where outcome measures are taken and the timeline of the measurements can affect our interpretation of the temporal dynamics of recovery. If there is too much time between outcome measures, the temporal resolution is low, resulting in lost information. On the other hand, testing/measuring animals too often is cumbersome, time-consuming, and might introduce excessive testing phenomenon. Taken together, the combination of both quantitative and qualitative measures in longitudinal studies will improve our understanding on how much recovery is achieved, how recovery occurred, and the temporal dynamics of that recovery.

Measuring other beneficial and adverse side effects

So far, we have discussed the benefits of training in recovering a targeted motor function and its effects on transferability to other motor functions. Yet, other beneficial effects of training have been described including lower heart rate, ¹⁶² increase bone density, ¹⁶³ modify muscle composition, ^10,12 normalize blood pressure, ^11,162 reduce pain, ^120,164,165 and ameliorate spasticity. ³⁷ Conversely, negative side effects have also been attributed to training after SCI, such as an increase of AD ^11,32 and bone and musculoskeletal pathologies, ^136,137 when applied excessively. Therefore, measuring other potential secondary effects of training, such as AD, neuropathic pain, cardiovascular function, sensory processing, or muscle mass, might unveil the consequences of training beyond the targeted motor function.

Enable training

Voluntary training post-SCI requires sufficient residual function to allow the training to be performed. Neuromodulatory approaches (i.e., methods to change neuronal excitability/activity) have been used to elevate residual activity after SCI, enabling training that would otherwise be difficult, or impossible. ¹⁶⁶ Electrical and/or chemical stimulation of the lumbosacral spinal cord can facilitate the induction of hindlimb movement and stepping, ^{68,69,167 –171} allowing training in animals with SCI. ^68,172 Also, neuromodulation by electrical stimulation of the cervical spinal cord restores forelimb muscle activation after SCI ^173,174 and facilitates reaching and grasping function for minutes to hours after stimulation. ^174,175 Another approach frequently used to activate spinal circuitry to enable training/stepping is the stimulation of the perineal area in rats and cats with a spinal cord transection. ^176,177 All this research indicates that stimulating the CNS can reveal residual function, enabling training. This approach is currently explored in SCI patients. ^170,175

Increase training efficacy

Strategies to maximize or enhance training efficacy are promising options to promote recovery given that they are based on a naturally occurring recovery process. As mentioned above, neuromodulation facilitates activity to allow training, but has also been successfully implemented to increase training efficacy. Regular electrical stimulation, by itself, can restructure neural circuitry. ^178,179 Whether the goal is to enable training or to enhance neuroplasticity, modulation contributes to the rewiring of spared neuronal substrate driven by training. For instance, Alam and colleagues ¹⁷⁴ promoted long-lasting enhancement of reaching and grasping function, but these gains were only observed when the stimulation was turned on. The researchers suggested that plastic changes make the system dependent on stimulation. Conversely, Van den Brand and colleagues ⁶⁸ showed that combined electrical stimulation of the spinal cord with locomotor training improves voluntary control of over-ground stepping even when stimulation is discontinued, suggesting that adding training to the neuromodulatory treatment is essential to shape the circuitry in a meaningful manner. McPherson and colleagues combined forelimb training with a closed-loop feedback system with forelimb extensor muscle electromyography as input and intraspinal electrical stimulation as output after a cervical SCI, to enhance training efficacy maintained even weeks after treatment stopped. ¹²⁹ Cortical electrical stimulation paired with training also enhanced training-induced recovery, ¹⁸⁰ and cervical electrical stimulation delivering electro-magnetical stimulation (EMS) to the spinal cord above the injury after a cervical SCI in combination to treadmill training reduced spasticity and improved gait. ³⁶ In another study, EMS was effective at increasing swimming and ball training efficacy in promoting hindlimb recovery after thoracic SCI. ¹⁸¹ Training efficacy to enhance function can also be achieved by chemical neuromodulation. Serotoninergic agonists can raise the excitability of the CNS ¹⁸² and induce plasticity that can be driven by training to improve recovery. ^79,183,184 In addition, both electrical and chemical neuromodulation can be administered together to potentiate spinal excitability even further, maximizing the benefits of training. ^68,172

Another successful approach for enhancing training efficacy is combining training with pharmacological, plasticity-promoting therapies. After cervical SCI, for example, the application of chABC endorsed plasticity and enhanced the recovery induced by forelimb reaching training. ^14,107 In a similar context, the promotion of plasticity by preventing myelin-associated glycoproteins (inhibitors of axonal extension) or Nogo from binding to their receptor was successfully combined with training to enhance recovery after SCI. ⁹⁵ Nonetheless, blocking plasticity inhibiting pathways is not always effective. The degradation of keratan sulphates, another inhibitor of axonal growth, plus training did not translate to a significant functional improvement compared to training alone. ⁹⁹ Plasticity can be boosted not only by reducing the inhibitory environment for axons to extend, but also by treatments that promote or enhance intrinsic neuronal plasticity. For instance, the overexpression of neurotrophins such as BDNF, in combination with reaching training, increases forelimb function after cervical SCI. ¹⁷ Using the same injury and training paradigm, recovery can be further improved by delivering a phosphokinase A (PKA) inhibitor to the cortex. ¹⁶ By blocking PKA, it is likely that more cAMP is accessible to exchange protein directly activated by cAMP, ultimately boosting neurite extension and neuroplasticity. Both approaches, inhibiting the inhibition and boosting intrinsic neuronal plasticity, have been used together to promote locomotor recovery after SCI by combining chABC with a neurotrophin cocktail and treadmill training. ¹⁸⁵ Even further, neuromodulatory and plasticity-based approaches can be combined with training to facilitate recovery. EMS of the spinal cord plus intraspinal administration of the neurotrophin, NT-3, together with training outperformed animals with EMS and training only. ¹⁸¹ Plasticity can also be induced by neuronal extrinsic processes. It has been shown that axonal sprouting after SCI is dependent on the inflammatory reaction induced by injury. In animals lacking an immune response, axonal sprouting after SCI is reduced, and only when inflammation is exogenously induced by systemic injection of lipopolysaccharide does sprouting recover, ¹⁸⁶ enhancing training efficacy. ¹⁵ Importantly, the combined effect of plasticity-inducing treatments and training can be promoted weeks ¹⁰⁷ or even months ¹⁵ after injury. Thus, treatments designed to increase or prolong the period of heightened plasticity can be used to enhance the effects of training. The key factor here is the interaction between augmenting plasticity and training. Training alone might not be sufficient to induce robust recovery, especially in the chronic scenario, ^14,15 and promoting plasticity without training can have little ^14,187 or even adverse ^15,33 functional effects. Therefore, training is essential to drive plasticity to meaningful changes and recovery. Importantly, this concept seems to translate well to the clinical setting. For example, it was shown that pairing intermittent hypoxia with ladder walking produced greater improvements in ladder walking than either treatment alone in rats with incomplete SCI. ¹⁸⁸ Studies replicated features of these rehabilitative training paradigms and found similar benefits in humans with incomplete SCI. ¹⁸⁹

Test therapies in a clinically relevant context

Another form of combinatorial treatment comes from the idea that rehabilitative training is regularly performed in the clinic, and thus any pre-clinical treatment has to be tested under clinical conditions, including rehabilitative training, before its translation is considered. Unfortunately, that is far from the reality. Only a few potential treatments for SCI (other than the ones designed to enhance training efficacy) have been combined with training in animal models, dismissing the interactions between treatments. For example, bone marrow stromal/stem cell transplantation (BMSC) combined with swim training or neural stem cell graft and treadmill training combination after SCI improve functional recovery over the individual treatments. ^190,191 Even further, cell grafting combined with serotonergic agonist treatment (meta-chlorophenylpiperazine or quipazine) and cycling training promoted better recovery than the cell therapy alone. ^183,184 These results are in contrast to Yoshihara and colleagues, that failed to find any benefits of combining BMSC transplantation and cycling training after SCI. ¹⁹² That could be the consequence of the lack of a beneficial effect of the BMSC graft by its own in this experiment, although have been proven by others. ^190,193 But, even when stem cell transplantation promotes recovery, the addition of training can hamper the beneficial effects of the grafts. ¹⁹⁴ Besides cell therapy, the combination of polypyrrole/iodine nanoparticles implantation in a contusion injury, a neuroprotective treatment, and treadmill training improved recovery. ¹⁹⁵ The combination of the antidepressant fluoxetine with treadmill training enhanced function after contusive SCI beyond either treatment alone. ¹⁹⁶ Interestingly, treating the depression-like behavior caused by SCI might also increase the mood of the animals and their motivation to perform the training. All together, this raises the question of how many experimental treatments have been diminished or even discarded for not combining them with training. Therefore, understanding the mutual influence of a non-rehabilitative-based treatment and training is essential to maximize clinical translation.

Factors influencing combination

Combining treatments is challenging. The interaction between treatments might reveal unknown effects, prompt synergisms, induce mutual annulation, or be completely reversal, and these effects can be time-, dose-, and training modality-dependent. For instance, the individual application of either locomotor treadmill training or the anti-Nogo-A antibody induced motor recovery after SCI, but improvements were not observed when the treatments were applied simultaneously. ¹⁹⁷ Maier and colleagues suggested that mechanisms of both treatments might interfere with each other and therefore the simultaneous application might cancel their effect. Indeed, SCI animals treated sequentially by applying anti-Nogo-A first, followed by treadmill training recovered beyond the motor improvement achieved by each treatment alone. ¹⁹⁸ A similar phenomenon has also been observed after stroke in rats. ¹⁹⁹ Collectively, these studies reveal a temporal pattern in which multi-intervention treatments might have to be orchestrated in order to be effective. Yet, further research will be necessary to determine the best chronology of each treatment to maximize recovery. This is of special importance for interventions with mechanisms of action that might conflict with training. For instance, ChABC can boost plasticity by degrading CSPGs, a potent axonal growth inhibitor, but also necessary for the stability of synapses. ²⁰⁰ Although training can drive the ChABC-increased plasticity to proper targets, continuous digestion of CSPGs might prevent the proper stabilization of those new connections. Thus, determining the optimal schedule for combinatory treatments is an important factor to consider. In addition, the dosage of training to provide with combined treatments is also important. The combined effect of an LPS-induced plasticity and rehabilitation is dependent on the training intensity, ¹⁵ pointing to the necessity of a minimal amount of training to drive the therapeutically induced plasticity. Therefore, using combinatorial approaches requires to not only optimize each individual treatment, but also to adjust the timing and dosage of their combination.