Afferent input plays an important role in rehabilitation after SCI. For people with SCI, locomotor training is based on providing sensory cues consistent with normal walking, especially from lower limb stretch- and load-sensitive mechanoreceptors [62], although a plethora of other inputs are also activated during the exercise [107]. In this context, of particular interest are cutaneous mechanoreceptors from plantar foot afferents because of their accessibility and simple integration with the functional task.

Plantar cutaneous feedback contributes to maintaining balance and ensuring a stable walking pattern throughout the step cycle [90, 102]. This has been shown in animal models, of cats [26] and rats [118]. In cats, the loss of cutaneous input does not produce a strong effect on stable walking. [114] observed that, after complete cutaneous denervation of the hindlimbs, cats were still able to walk in a pattern similar to that in intact cats. Using gait analyses instrumentation, subsequent studies detected small changes in limb kinematics and muscle activity [41]. However, in challenging walking conditions, such as ladder or incline walking, or when lateral stability was compromised by unexpected perturbations, the kinematics and locomotor muscle activity were significantly modified, with compromised stability [11, 14]. Walking disruption was even more evident in spinalized cats, suggesting that the remaining intact input of muscles, joints and other receptors was not sufficient in these animals to compensate for the loss of cutaneous information [13]. In rats, which are plantigrade animals as are humans, the elimination of cutaneous input by hypothermic anaesthesia modified hindlimb kinematics to produce less efficient locomotion, with a larger ankle and hip excursion and decreased distance between the hip and the ground, leading to a significant functional deficit [130].

As in the animal models, in humans information from plantar cutaneous afferents plays an important role in postural and walking stabilization. These have been shown to determine postural responses in ankle muscles, to stabilize stance and gait [4, 102, 127, 147]. For example, after a temporary plantar desensitization, motor activity and ground reaction forces during a balance recovery task were reduced [120, 126] and postural adjustments prior to step initiation were compromised in a way that could not be compensated by other inputs [79].

Furthermore, [146] found that in humans, electrical stimulation of plantar cutaneous afferents during walking generates withdrawal responses during swing and stabilizing responses during stance; this phase-dependent behaviour indicates an integration of cutaneous afferent feedback into spinal locomotor networks. In fact, the contribution of plantar afferent feedback to locomotion is not limited to reactive responses: it extends from the modulation of spinal reflex excitability [52] to the selection of motor patterns and the alterations of central pattern generator (CPG) frequency [107].

After SCI, the loss of descending input from supraspinal structures increases the importance of sensory feedback from the periphery [88]. The processing of sensory input within the SC is altered [51] and considerable changes in reflex pathways develop over time as a consequence of the injury [25, 26]. Locomotor training after SCI is based on the fact that appropriate peripheral sensory input can reactivate and reorganize the spinal locomotor circuitry [109].

Critical input comes from proprioceptive signals from the muscles. For example, in incomplete SCI rats, locomotion rehabilitation based on swimming training improves locomotion [118]. However, the therapy is more effective when swimming is potentiated by cutaneous feedback provided by a matrix of buoyant tubes suspended from the bottom of the pool [118]. A similar result was previously observed by [98], using the same protocol to rehabilitate locomotion of hemisected chicks. In this case, the phasic feedback from cutaneous plantar receptors was sufficient to permanently increase limb extension during swimming.

Cutaneous input is also critical for the recovery of stepping during walking training on a treadmill. For example, in cats, the chemical deactivation of the nerves from the low-threshold mechanoreceptors of the hind paws consistently decreased coordination between forelimbs and hind limbs, and between left and right legs, reducing walking stability and efficiency [117].

In humans, some neuromodulation protocols already take advantage of the strong interaction between feedback from plantar cutaneous afferents and spinal locomotor circuits, to restore a pathological or absent phase-dependent H-reflex modulation pattern or stimulate plastic changes to CNS microcircuits following injury [102]. For example, [52] applied electrical impulses to the sole of the foot to decrease the soleus H-reflex, showing that it is possible to modulate this reflex in both neurologically intact subjects and in subjects with SCI during walking and standing. These results were extended by [71], who used electrical stimuli to enhance sensory feedback during walking in subjects with SCI, producing phase-dependent normalization of H-reflex and tibialis anterior flexion reflexes.

Electrical Spinal Cord Stimulation (SCS) is an invasive neuromodulation technique that entails the application of minute electrical impulses to the spinal cord to modulate segmental spinal and/or brain stem/spinal pathways. These techniques have been used to treat motor dysfunction associated with several degenerative diseases of the CNS and has been shown to improve a wide range of functions including motor function, posture, walking, and bladder control [24, 131]. SCS may be applied by epidural electrical stimulation (EES) or intraspinal microstimulation (ISMS).

EES is applied to the dorsal surface of the spinal cord by surgically implanted electrodes, to activate the dorsal root entry zones or dorsal columns. Its principal use has been for the treatment of neurologic pain [2, 23, 73, 74, 128]. However, when applied in association with partial body-weight support training, it has the potential to initiate and sustain locomotion in people with chronic incomplete SCI, even in cases where partial body weight support alone was not able to achieve functional over-ground ambulation [16, 63].

EES, in combination with physical therapy, has also been used in individuals with chronic complete SCI. For example, [3] showed that four subjects with complete SCI (AIS A or B) were able to recover voluntary movement soon after the implantation of the stimulation device, and their condition improved progressively in the following months with the combination of EES and body weight supported standing exercises or manual therapist-assisted locomotor training on a treadmill [3, 60]. The same therapies performed before the implantation of the EES device were not able to produce functional recovery in these patients.

Together, these results suggest that EESs effect on the spinal pathways may enhance the central excitatory drive to the motor neurons; at the same time, task-specific training may promote force generation and accuracy [3]. In addition, the fact that stimulating specific SC regions (L2 in humans, L5 in cats and between L2 and S1 in rats [53]) can induce or facilitate stepping, has been attributed to the activation of central pattern generators circuits (CPG) in the SC, i.e., neural circuits that generate coordinated alternating flexor-extensor neuromotor patterns in the absence of supraspinal or sensory modulation [50]. Nevertheless, it has not yet been shown whether muscle activity generated with EES will be sufficient in individuals with SCI to support the body during stepping, and what the long-term effects of such a treatment will be [32].

A similar and more invasive technique is ISMS [94]. In this method, arrays of microwire electrodes are implanted in the spinal cord to target the intermediate and ventral gray matter for activation of local circuits. ISMS potential for rehabilitation comes from its capacity to evoke, in the isolated SC, specific functional movements in the hindlimbs and forelimbs. This effect depends on the level and characteristics of the stimulation. For example, standing and stepping movements can be evoked by the application of ISMS to the lumbar SC in spinal cats [7, 59, 99], while forelimb movements can be evoked when ISMS is delivered to the cervical spinal cord in non-human primates [95, 148].

Since these SCS techniques require the use of implanted electrodes near to (EES) or within (ISMS) the CNS, complications may arise during and after the surgical implantation [24]. Principal intra-operatory complications include direct damage to the neuraxial structures and risks associated with the anaesthesia and surgery; in the post-operatory interval, principal risks include pain, seroma, lead migration and hardware failure, and infections [24, 70, 128].

A recent review of documented cases of EES stimulation provided an estimation of the risk of complication associated with EES [91]. Of the 787 patients considered in the review, those who were initially submitted to the EES initial trial lead placement, 11 % reported emergence of pain at the stimulation site and 4.5 % developed documented infections. After the test phase, 200 patients were excluded from the pulse generator implant, with an average implant-to-trial ratio of 75 %. For 38.1 % of the remaining 512 patients who were implanted with the stimulator, hardware related complications (lead migration, lead connection failure and lead breakage) occurred.

ISMS has not yet been tested in humans. The principal challenges of ISMS relate to the safety and durability of the implanted electrodes, since adverse tissue responses are still common following implantation of currently available electrodes within the CNS [92].

Pharmacological treatments have also shown potential for activating locomotor circuits in rats and mice [5, 76]. [27] proposed an alternative approach to neuromodulation by combining specific pharmacological and electrical stimulation interventions, together with locomotor training [27, 129]. Combining injection of serotonin antagonists (), stimulation with EES and locomotor gait training induced full weight bearing bipedal treadmill locomotion in complete SC-transected rats. In other studies, [12, 133], it was shown that by automatically modulating electrochemical stimulation parameters, it is possible with a closed-loop controller to precisely control leg kinematics in rats during movement. This method allows the animals to voluntarily initiate bipedal locomotion from a resting position, to walk for extended periods of time, to pass obstacles, and to climb a staircase. These results inspire hope for future neuroprosthetic devices as well as for promoting recovery and motor function after severe lesions.

Electrical and pharmacological stimulation of spinal neural networks is still in an experimental stage. Despite promising results in animal models, further research is required before widespread application of these techniques in humans will be possible. Principal concerns are the selection of the appropriate neuromodulators to be used in humans [37] and the timed control of pharmacological and electrical stimulation. Moreover, in animals the techniques have been applied only targeting the hindlimbs. In the future, refinement of animal studies may contribute to greater translational success [32]. Thus, the combination of pharmacological and epidural stimulation to enhance motor function remains a promising research direction yet to be fully explored.

Starting in the 1980s [139–141] and continuing to the present [122–124], Wolpaw and colleagues have shown that descending supraspinal activity may be used to modulate specific spinal behaviours. These studies involved the operant conditioning of spinal reflexes. In the first studies, in monkeys, an operant conditioning protocol was used to train the animals to up-or down-regulate their spinal stretch reflex or its electrical analogue, the H-reflex [139, 140]. The animals were rewarded if they changed the reflex amplitude in the correct direction over 50 days. Approximately 80–90 % of the test animals were able to successfully modulate the reflex in the desired direction [140]. Changes were observable immediately after onset of training and persisted even during inactive periods after cessation of the reward protocol. The protocol had three key features. First, it required maintenance of both a certain elbow angle and a certain level of biceps EMG activity as the animal opposed a constant extension torque in order to generate a stimulus producing the reflex. Second, it based reward on the size of the recorded reflex. Finally, the reward criterion (i.e., requiring successful up- or down-conditioning in the animals) remained constant over days and weeks. In sum, the protocol was designed to induce and maintain a long-term change in descending influence over the spinal arc of the reflex, and to thereby change the spinal cord. Subsequent studies in which the spinal stretch reflex or H-reflex was conditioned in rats [21, 22], mice [20], and humans [43, 112] confirmed the first results with monkeys and served to further elucidate the anatomical, physiological, and biochemical mechanisms of this conditioning (see [124] for review).

Interestingly, in rat studies [22], analysis of the effect of soleus H-reflex conditioning on locomotor kinematics showed that successful conditioning in the intended direction of the reflex produced changes not only in this reflex, but also in the kinematics of locomotion. However, the changes did not alter key aspects of locomotion (such as gait speed, gait symmetry or step length) but rather provoked compensatory changes in other pathways (i.e., the quadriceps reflex) [21]. These compensatory changes demonstrated the ubiquity of plastic changes in the SC and how newly acquired behaviours can cohabitate with existing ones, leaving the latter unaffected despite the change in the soleus reflex [21, 137]. In addition, [22] found that rats with partial SCI undergoing soleus H-reflex up-conditioning could improve locomotion parameters such as soleus burst and locomotion symmetry, providing the first evidence of the potential functional impact of this kind of intervention.

The experimental accessibility of the SC, the relative simplicity of the conditioned reflex (largely monosynaptic), the well-defined descending pathways from the brain, and the direct association to a specific observable behaviour, make H-reflex operant conditioning a unique tool for understanding the basic mechanisms of spinal cord plasticity and how it changes behavior. Moreover, by taking advantage of the targeted modulation of specific spinal pathways, it offers the possibility of new approaches to neurorehabilitative intervention.

An adaptation of the stretch-reflex and H-reflex conditioning paradigm was also tested in humans [44, 122]. Whereas the animal studies included several thousand trials per day, 7 days per week, over 50 days of conditioning [141], the humans performed only 225 trials per one-hour session three times per week [122]. For the human studies, the electrodes used for stimulation and recording were placed on the skin rather than implanted as in the animal studies. In the animal studies, the animals were rewarded with food for correct performance; in the human studies, the subjects received visual feedback showing performance on a computer display. In these human studies, a small number of trials at the beginning of each session were acquired as control trials: in these trials the subject was not asked to change the reflex and received no feedback [122]. These control trials served to check for intra- vs. inter-session H-reflex changes. These studies showed that 70–90 % of normal human subjects were able to modify the conditioned reflex in the correct direction. Over 24 conditioning sessions H-reflex size increased to an average of 140 % of initial value in up-conditioned subjects and decreased to an average of 69 % in down-conditioned subjects. Reflex changes occurred in two phases: they began (Phase 1) with a rapid small change, in the correct direction, that was attributable to rapid task-dependent adaptations, and then (Phase 2), over days and weeks, changed gradually further in the correct direction (long-term change) [122]. Phase 1 appears to reflect rapid mode-appropriate change in descending influence over the spinal arc of the reflex, while Phase 2 appears to reflect gradual spinal cord plasticity produced by the chronic continuation of the descending input responsible for Phase 1 [122]. These bi-phasic results were consistent with the animal studies previously performed by [142].

A more recent study was carried out in people with incomplete SCI [123]. The experimental procedure was similar to the one with healthy subjects. Success rate and magnitude of reflex change were comparable to those in people with intact SC. As in the previously cited studies of locomotion in rats, the comparison of functional capacities before and after the conditioning showed that appropriate H-reflex conditioning (i.e., down-conditioning in this human study) was associated with faster and more symmetrical locomotion. The improvement was evident both in quantitative testing and, most important, in subjective observations by the subjects themselves: they spontaneously reported improvements in walking speed and endurance, and in other aspects of motor function (e.g., balance). Interestingly, the subjects with SCI showed long-term H-reflex changes that were greater than the changes observed in healthy subjects (24 % vs. 16 %). In contrast, task-adaptation changes (i.e., reflex changes observed within sessions but not maintained between consecutive sessions) were smaller in subjects with SCI (7 % vs. 15 %). Another recent study [82] also found that appropriate H-reflex conditioning had beneficial effects on locomotion in people with incomplete SCI.

The smaller long-term changes in healthy subjects as compared to people with SCI suggests that, in healthy subjects, H-reflex changes are mainly occurring during the experimental sessions. That is, for healthy subjects, the H-reflex changes are unlikely to provide any beneficial effects in terms of functional outcomes outside the experimental sessions (in locomotion for example), and they may even interfere with the already well-functioning circuits for locomotion and other behaviours in the intact SC. Therefore, in these healthy subjects, over the conditioning sessions, additional plasticity may occur to maintain SC equilibrium that ensures that locomotion and other important behaviours are normal . On the other hand, as was the case with the SC injured rats [22], appropriate H-reflex conditioning in people with SCI led to improvements in lower-limb functions (such as locomotion). Since the H-reflex change was beneficial, the SC retained the neurophysiological change resulting from the experimental protocol. In other words, the injured SC was guided towards a new and more optimal equilibrium [124, 137].

Taken together, these animal and human experiments indicate that reflex conditioning protocols can improve recovery after chronic incomplete SCI, and possibly in other neurological and physical disorders. Mrachacz-Kersting and colleagues are currently exploring ways to up-condition the soleus stretch reflex in healthy people. The main goal in this work is to enhance ankle stiffness in athletes who are prone to suffer ankle sprain. Preliminary data indicate that up-regulating the reflex is possible that this reflex change is associated with enhancement of the intrinsic and total stiffness of the ankle joint and an improved stability after drop landings [96].