Spinal reflexes: a potential target for treating hemiparetic gait. Editorial Focus on “Changes in Gait Asymmetry May Be Caused by Adaptation of Spinal Reflexes”

The ability to adapt and modify walking patterns is a critical feature of our nervous systems. It allows us to traverse uneven surfaces like a rocky hillside or the shifting footing of a sandy beach. Understanding how the nervous system successfully performs locomotion in such a wide variety of environmental and task-specific circumstances is critical to designing more effective interventions for patients for whom walking has become impaired. To this end, locomotor adaptation has been extensively studied in humans using split-belt treadmills. In this paradigm, people walk on a treadmill with two belts, one under each leg, that are moving at different speeds. Initially, the discordant belt speeds cause people to walk with an asymmetric limp, i.e. the leg stepping on the faster belt takes smaller steps due to the increased speed. Over the course of minutes, people gradually change the spatial and temporal characteristics of their gait to become more symmetric, i.e., once again, their legs begin to swing out of phase with one another with each leg taking similar step sizes. Thus, this paradigm has become a popular method to study error-driven sensorimotor adaptation in locomotion. Furthermore, it has been investigated as a potential rehabilitative approach for hemiparetic gait in stroke survivors. Hemiparetic gait is characterized by asymmetric weakness which itself leads to step-length asymmetry. Here, split-belt treadmill training can be applied to exaggerate step-length asymmetry; thus prompting corrective adaptation and more symmetrical gait. Through repeated training with this paradigm, some individuals will retain this more symmetrical gait for months after training has ceased (Reisman et al. 2013). However, like many neurorehabilitative interventions, the effectiveness of this training varies across individuals; some trainees will show no lasting improvement, i.e., nonresponders. Thus, there is a need to identify the neural mechanisms underlying split-belt sensorimotor training in order to improve its generalizability or identify which patients it can benefit (i.e., responders).

Studies of animal locomotion, particularly using cat models, have been crucial in understanding the neural mechanisms of locomotor control. Thomas Graham Brown’s seminal study in decerebrated and deafferented cats demonstrated that rhythmogenesis and muscle patterning could be generated without cortical or afferent input (Brown 1914). These observations led to the description of a central pattern generator (CPG) which maintains inter- and intra-limb coordination via a network of spinal interneurons. Subsequent studies affirmed the existence of CPGs in a number of vertebrate species and, while there is some debate, studies of people with spinal cord injuries and spinal stimulation offer indirect evidence that human locomotion is also generated via CPGs (Minassian et al. 2017). While rhythmogenesis and patterning appear to be due to CPGs, other regions of the central nervous system have been found to contribute to locomotion; these include several brainstem regions (e.g. mesencephalic locomotor region, subthalamic locomotor region, reticular and vestibular nuclei), the motor cortex, the cerebellum, and spinal reflex pathways. Split-belt walking has also been examined in cat models. It displays some interesting differences from what is observed in humans. Decerebrate cats can perform split-belt treadmill walking, suggesting that supraspinal input is not essential for modifying the asymmetric gait pattern. Notably, cats do not gradually become more symmetrical during split-belt treadmill walking as is seen in humans (Kuczynski et al. 2017). This may be due to differing biomechanical requirements of bipedal vs. quadrupedal interlimb coordination. For example, cats frequently use gait patterns with non-alternating leg swings such as galloping. Thus, it is still unknown how humans alter limb kinematic variables – such as step length asymmetry – during split-belt locomotion. In this issue of the Journal of Neurophysiology, Refy et al. address this question.

Refy et al. employ a neuromuscular model to infer a causal relationship between spinal reflex modulation and step length adaptation in split-belt treadmill walking. This work builds on their previous demonstration that adaptive changes in step-length asymmetry during split-belt treadmill locomotion are associated with changes in the H-reflex gains of leg muscles (Refy et al. 2023). Their neuromuscular model defines a simplified, 2D mechanical model of the lower body actuated with 14 muscles that are controlled, in part, by spinal reflexes elicited by proprioceptive, vestibular, and mechano-receptors. To determine whether there is a causal relationship between the observed reduction in H-reflex gains and step-length adaptation, the authors use this model to perform two simulation experiments. Both experiments simulate a split-belt walking paradigm in which the speed of one belt is reduced by 30%. The first experiment compares the resultant step-length asymmetries when: (1) reflex gains are kept the same throughout; or (2) reflex gains are reduced at the onset of the change in belt speed. Only when reflex gains are reduced does the model predict the observed step-length asymmetry that occurs in early split-belt treadmill walking. The second experiment compares the resultant step-length asymmetries when the reduced reflex gains: (1) recover gradually; or (2) remain reduced. When reflex gain is permitted to recover (as was observed in their previous study), step-length asymmetry decreases in a manner consistent with split-belt treadmill adaptation. In sum, their modeling supports the conclusion that spinal reflex modulation underlies step length adaptation in split-belt treadmill walking.

Afferent activity has long been recognized as a modifier of motoneuron activity during locomotion in cats (Yang et al. 1991). Spinal reflexes drive changes in bilateral phase durations, which are modified during split-belt walking in both humans and cats (Frigon et al. 2017). The authors’ findings indicate that spinal reflexes also play an important role in the gradual return of step-length symmetry observed in humans during split-belt locomotion. Interestingly, when the model’s reflex gains are reduced without recovery, the model displays step-length asymmetry patterns similar to those found in patients with cerebellar degeneration. This similarity suggests that the cerebellum modulates step lengths by changing spinal reflex gains. Importantly, the study provides mechanistic insight into a behavioral paradigm that is used not only in a research setting, but also has shown promise as a rehabilitative intervention. The study may help to explain the variable effectiveness of split-belt training for hemiparetic gait, i.e., responders vs. non-responders. If a person with hemiparetic gait after a stroke also has an impaired ability to modulate reflex gain, then they may struggle to adapt to split-belt walking or to retain the adaptation. Studies in clinical populations could test this hypothesis. Furthermore, by identifying spinal reflex changes as a mechanism of step-length adaptations, this study suggests therapies that might improve split-belt treadmill adaptations. For example, spinal reflex gains might be changed by reflex operant conditioning or other noninvasive stimulation paradigms that target reflex gain changes that enhance the benefit of split-belt training (e.g., Thompson et al. 2013).

The limitations of this study should also be considered. Mathematical models of biological processes are inherently simplifications, and as famously declared by George Box, “All models are wrong, but some are useful.” The neuromuscular model employed by Refy et al. is no exception; it describes a 2D mechanical and neural control system that is much less complex than the biological system it models. Furthermore, determining reflex gains via a simple Bayesian optimization, as the model does, means that the results of the model may not be the only solution. That said, the strength of this study remains. It uses a deterministic model of underlying processes to gain new mechanistic insight into split-belt treadmill adaptation, insight that is significant both scientifically and clinically (Refy et al. 2023).

In summary, the paper by Refy and colleagues is an important contribution to understanding the neural mechanisms of split-belt treadmill training and expands the role of spinal reflexes in locomotion. Continued research to elucidate mechanisms of locomotor control and adaptation hold great promise for improving rehabilitation for a wide range of musculoskeletal and neurological disorders.