The ability to modulate ongoing walking gait with precise, voluntary adjustments is what allows animals to navigate complex terrains. However, how the nervous system generates the signals to precisely control the limbs while simultaneously maintaining locomotion is poorly understood. One potential strategy is to distribute the neural activity related to these two functions into distinct cortical activity coactivation subspaces so that both may be conducted simultaneously without disruptive interference. To investigate this hypothesis, we recorded the activity of primary motor cortex in male nonhuman primates during obstacle avoidance on a treadmill. We found that the same neural population was active during both basic unobstructed locomotion and volitional obstacle avoidance movements. We identified the neural modes spanning the subspace of the low-dimensional dynamics in primary motor cortex and found a subspace that consistently maintains the same cyclic activity throughout obstacle stepping, despite large changes in the movement itself. All of the variance corresponding to this large change in movement during the obstacle avoidance was confined to its own distinct subspace. Furthermore, neural decoders built for ongoing locomotion did not generalize to decoding obstacle avoidance during locomotion. Our findings suggest that separate underlying subspaces emerge during complex locomotion that coordinates ongoing locomotor-related neural dynamics with volitional gait adjustments. These findings may have important implications for the development of brain–machine interfaces.SIGNIFICANCE STATEMENTLocomotion and precise, goal-directed movements are two distinct movement modalities with known differing requirements of motor cortical input. Previous studies have characterized the cortical activity during obstacle avoidance while walking in rodents and felines, but, to date, no such studies have been completed in primates. Additionally, in any animal model, it is unknown how these two movements are represented in primary motor cortex (M1) low-dimensional dynamics when both activities are performed at the same time, such as during obstacle avoidance. We developed a novel obstacle avoidance paradigm in freely moving nonhuman primates and discovered that the rhythmic locomotion-related dynamics and the voluntary, gait-adjustment movement separate into distinct subspaces in M1 cortical activity. Our analysis of decoding generalization may also have important implications for the development of brain–machine interfaces.

Detection of neural signatures related to pathological behavioral states could enable adaptive deep brain stimulation (DBS), a potential strategy for improving efficacy of DBS for neurological and psychiatric disorders. This approach requires identifying neural biomarkers of relevant behavioral states, a task best performed in ecologically valid environments. Here, in human participants with obsessive-compulsive disorder (OCD) implanted with recording-capable DBS devices, we synchronized chronic ventral striatum local field potentials with relevant, disease-specific behaviors. We captured over 1,000 h of local field potentials in the clinic and at home during unstructured activity, as well as during DBS and exposure therapy. The wide range of symptom severity over which the data were captured allowed us to identify candidate neural biomarkers of OCD symptom intensity. This work demonstrates the feasibility and utility of capturing chronic intracranial electrophysiology during daily symptom fluctuations to enable neural biomarker identification, a prerequisite for future development of adaptive DBS for OCD and other psychiatric disorders. The identification of candidate neural biomarkers of obsessive-compulsive disorder symptom intensity in ecologically valid environments.

EditorialControl of MovementHighlights from the 30th Annual Meeting of the Society for the Neural Control of MovementMarta Russo, Nofar Ozeri-Engelhard, Kathleen Hupfeld, Caroline Nettekoven, Simon Thibault, Ehsan Sedaghat-Nejad, Daniela Buchwald, David Xing, Omid Zobeiri, Konstantina Kilteni, Scott T. Albert, and Giacomo ArianiMarta RussoDepartment of Neurology, Tor Vergata Polyclinic, Rome, ItalyDepartment of Biology, Northeastern University, Boston, Massachusetts, Nofar Ozeri-EngelhardWM Keck Center for Collaborative Neuroscience, Rutgers, The State University of New Jersey, Piscataway, New Jersey, Kathleen HupfeldDepartment of Applied Physiology and Kinesiology, University of Florida, Gainesville, Florida, Caroline NettekovenWellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UKDepartment of Psychiatry, School of Clinical Medicine, University of Cambridge, Cambridge, United Kingdom, Simon ThibaultImpAct team, Lyon Neuroscience Research Center, Inserm U1028, CNRS UMR5292, University of Lyon 1, Lyon, France, Ehsan Sedaghat-NejadLaboratory for Computational Motor Control, Department of Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, Maryland, Daniela BuchwaldOttobock SE & Co. KGaA, Software & Electronics Engineering, Duderstadt, Germany, David XingDepartment of Neurobiology, Northwestern University, Evanston, Illinois, Omid ZobeiriDepartment of Biomedical Engineering, McGill University, Montreal, Quebec, Canada, Konstantina KilteniDepartment of Neurology, Tor Vergata Polyclinic, Rome, Italy, Scott T. AlbertLaboratory for Computational Motor Control, Department of Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, Maryland, and Giacomo ArianiThe Brain and Mind Institute, Western University, London, Ontario, CanadaDepartment of Computer Science, Western University, London, Ontario, CanadaPublished Online:23 Sep 2021https://doi.org/10.1152/jn.00334.2021This is the final version - click for previous versionMoreSectionsPDF (1 MB)Download PDFDownload PDFPlus ToolsExport citationAdd to favoritesGet permissionsTrack citations ShareShare onFacebookTwitterLinkedInEmail INTRODUCTIONThe 30th meeting of the Society for the Neural Control of Movement (NCM) was originally scheduled to take place in Dubrovnik, Croatia, in April of 2020. Due to the COVID-19 pandemic, the in-person meeting was canceled and replaced by a virtual symposium showcasing the work of the society’s 2020 scholarship winners (https://ncm-society.org/symposium/). By the spring of 2021 (April 20th-22nd), the annual meeting was ready to return, although moving for the first time to a completely virtual setting on the online platforms Pheedloop (https://pheedloop.com) and Gathertown (https://gather.town) (1).More than ever before, this virtual format facilitated participation and social media engagement (e.g., the popular hashtag #NCM2021) from the research community all around the world (Fig. 1). Indeed, the forced choice of moving online had several positive side effects. First, rebounding from a concerning dip in academic participation from female researchers during the COVID-19 pandemic (2), with respect to previous in-person meetings, NCM 2021 showed an uptick in the percentage of female panelists (34% female attendees, 44% female speakers, and 30% female submissions). Second, early career researchers had more opportunities to present their work thanks to the addition of “data blitz” sessions on top of the traditional posters, individual talks, and panel discussions. This novel category allowed 5 min for presentation and 1 min for questions, followed by separate breakout rooms. Third, to enhance discussions during the meeting, panels, and posters were prerecorded and posted online in advance. Live recordings of every session allowed delegates to catch up on missed talks or rewatch talks later. These were welcome additions, as indicated by many respondents to the meeting’s feedback survey.Figure 1.World countries color-coded by the number of people attending Neural Control of Movement (NCM) 2021 affiliated with a university located in that country. Colors are represented on a log10 scale. The number insets in each country show the exact number of attendees.Download figureDownload PowerPointAs done in previous years (3–5), here we present highlights from the meeting. These highlights revolve around four central themes: 1) neuroplasticity, 2) complex motor skills, 3) multimodal sensory integration, and 4) the role of descending spinal tracts in motor control.NEUROPLASTICITY FOLLOWING ALTERED SENSORY INPUTA particular focus of NCM 2021 was on studies using multimodal neuroimaging to understand brain plasticity resultant from exposure to body augmentation technologies or abnormal sensory input. Tamar Makin, the 2020 Early Career Award winner, was commended for their innovative work examining the drivers and limitations of plasticity in the human brain as well as their contributions to the Society—including the organization of the first all-female NCM panel in 2018. Investigating embodiment of prostheses using fMRI, Makin and colleagues found that two-handed participants represented cosmetic prostheses more like hands and functional prostheses more like tools. However, among prosthesis users, both functional and cosmetic prostheses were not represented either as hands or as tools, but instead constituted a separate category (6, 7). Makin and collaborators explored a similar effect in another group of expert tool users, London litter pickers (7). Similar to amputees, litter pickers viewed their tool as a separate entity and not as an extension of their hand. Together, these results suggest that the human brain is plastic enough to create representations that are distinct from those shaped by evolution.A PhD student in Makin’s laboratory, Elena Amoruso, explicitly tested this hypothesis in their “third thumb” project (8). Participants wore an artificial thumb controlled by their toes. Coordination rapidly improved with practice and led to successful learning transfer when the controllers were switched (e.g., the toe that controlled flexion/extension now controlled abduction/adduction). When using local anesthesia to block proprioceptive and somatosensory input of the toes controlling the artificial thumb, early training was not affected, but retention of learning on the second day was smaller compared to a sham anesthesia control group, suggesting that sensory input played a critical role in learning to control the additional finger. The third thumb was also tested in a real-world context (9), where participants wore the device for ∼4.5 h a day for 5 days. Participants rapidly improved their performance and reported increasing sensations that the thumb was part of their body. Participants thus were able to develop a sense of proprioception of the thumb’s position relative to their biological fingers. Together, Makin and colleagues’ work demonstrates the remarkable ability of the human nervous system to undergo plastic changes—an ability Makin believes we should leverage in the development of prostheses. Similarly, Robert Nickl characterized the stability of neuronal responses to movement intention in a patient with incomplete tetraplegia who was bilaterally implanted with multi-unit electrode arrays in the primary motor (M1) and primary somatosensory (S1) cortices (10). Across 12 sessions, the number of active units was more stable in S1 than M1. However, in both areas, the number of overall active units showed a nearly exponential decline. With respect to recorded neuronal activity, in a single channel stability declined within minutes to hours. Interestingly, contralateral activity was more stable than activity in ipsilateral units. This characterization of the stability of neural activity is important for the development of brain-machine interface devices capable of decoding this representation into motor output.The brain’s ability to form new representations can also be probed by studying the neuroplastic changes driven by exposure to a completely novel environment. For example, Grant Tays investigated the impact of microgravity on fifteen astronauts who spent ∼6 months onboard the International Space Station (11–13). Participants completed MRI and behavioral testing multiple times pre- and postflight. Their data suggested little cognitive change from pre- to postflight but pronounced postflight impairments to mobility, balance, and bimanual coordination (14). These performance declines extended beyond only peripheral changes such as disuse muscle atrophy, suggesting that centrally mediated processes might also contribute to these effects. Further probing central nervous system changes with spaceflight, Kathleen Hupfeld discussed their work on vestibular processing in astronauts. Hupfeld and colleagues applied vestibular stimulation (pneumatic cheekbone taps; 15) to measure brain activity during vestibular processing at multiple times pre- and postflight. As previously demonstrated on Earth (12, 16), preflight, vestibular stimulation elicited activation of the parietal opercular area (i.e., the so-called “vestibular cortex”) and deactivation of somatosensory and visual cortices. Postflight, astronauts showed widespread reductions in somatosensory and visual cortical deactivation. In addition, greater reductions in the deactivation of visual brain regions were associated with smaller declines in standing balance. These findings suggest that microgravity exposure results in cortical plasticity in the form of sensory reweighting, i.e., down-weighting of vestibular inputs (due to this system’s altered signaling in the absence of gravity), and concurrent up-weighting of other sensory processing regions, such as the somatosensory and visual cortices. This reweighting may facilitate more adaptable postflight standing balance when crewmembers readjust to normal vestibular inputs on Earth.To better characterize the mechanisms underlying cortical plasticity, Caroline Nettekoven presented their data on the relation between motor cortical γ-aminobutyric acid (GABA), previously shown to play a role in motor learning and visuomotor adaptation (17–19). Although participants adapted to a stepwise increasing rotation (or performed a control task) in the MRI scanner, Nettekoven and colleagues measured GABA concentration in the left M1 hand area. GABA concentration before adapting predicted retention of the adapted movement but not the extent of adaptation, suggesting a role for M1 GABA in maintaining but not acquiring the adapted state. This relationship between GABA and retention of the adapted movement was mediated by the change in functional connectivity between the left M1 hand area and the right cerebellar hand area. Participants with higher M1 GABA concentration before adapting showed a decrease in functional connectivity between M1 and the cerebellum during adaptation, and they better retained the adaptive movement. These findings imply a link between motor performance, motor network connectivity, and cortical inhibition, and shed light on the neurochemical bases of human motor adaptation (20). Chris Horton and colleagues provided additional evidence that region-specific cortical GABA concentrations predict aspects of human motor performance. Measuring GABA concentrations in S1 and thalamus during “go” and “stop” tasks, they found no association between GABA and “stopping” performance. However, in the “go” task, higher ipsilateral thalamic GABA correlated with faster reaction time. These data suggest that thalamic GABA concentration supports speeded selection and execution of cued choice responses.Together, these lines of research contribute to our understanding of how the brain adapts to perturbations or novel sensory conditions (e.g., prolonged prosthesis or tool use, microgravity, and visuomotor perturbations) and have numerous applications for improving human health, such as designing more effective prostheses and maintaining astronaut health during future missions to Mars.EXPLORING COMPLEX MOTOR SKILLS BEYOND UNIMANUAL REACHINGAnother emergent theme of NCM 2021 was a renewed push to study more complex motor skills. Although the definition of a “complex motor skill” may seem arbitrary, here we consider studies that examined naturalistic multijoint movements that tend to involve interactions with tools and/or the control of many degrees of freedom (DOFs).One example is sophisticated finger control. As Tamar Makin’s work shows, besides prompting questions about artificial limb embodiment and neural plasticity, augmentation technologies can open new avenues for studying complex behaviors by enabling previously impossible actions. In addition, within the clinical setting, we can leverage assistive technologies to measure hand function and monitor the progress of rehabilitation. For example, Jing Xu introduced a novel device, the HAND (Hand Actuation Neural-training Device), that is equipped with force sensors capable of detecting even the smallest isometric forces from a near-plegic hand in 3-D (21). They used this device to characterize finger coactivation patterns in healthy participants and stroke patients. The importance of such research is most evident in manual activities that demand finger individuation, such as producing a chord when playing the piano. By comparing naïve participants and expert musicians using a foot-controlled supernumerary robotic thumb to play the piano, Aldo Faisal asked what determines our ability to learn and use augmentation in skilled tasks. They showed that foot dexterity (and not task-relevant piano expertise) is the best predictor of future performance (22). In addition, the observation of highly idiosyncratic learning curves prompted new questions for future research: can everyone be augmented equally? Or should we design personalized training? Regardless, it is clear that we need real-world complex tasks to improve the training of real-world sensorimotor skills. According to Ilana Nisky, 2021 Early Career Award winner, robot-assisted surgery can bridge the gap between laboratory-based research and real-life applications. Currently, surgical training is not optimized, partly due to a lack of haptic feedback and partly due to limited knowledge on how to measure surgical skills (23–27). Nisky’s research used a teleoperated needle driving task and integrated data measuring the dynamics of the robotic manipulandum and modeling human kinematics to describe the quality of surgical skill throughout the learning process. A key aspect of surgical skill is that it requires extremely precise control of external tools via the coordination of multiple effectors. Similarly, watchmaking is a complex craft that includes bimanual control of 44 DOFs. To understand and model such dexterity, Aude Billard and colleagues examined cohorts of apprentices and expert watchmakers using a combination of motion capture and tactile sensing systems. They found that experts consistently used distinctive hand poses that optimized manipulability and made use of longer preparation times to reduce possible mistakes during execution time (28, 29). Moreover, their research revealed that the two hands can work together distributing control of different variables to achieve better precision than a single hand.Combining tool use and whole body movements is another way to increase realism and complexity in the study of motor skills. For example, Antonella Maselli presented their work on ball throwing in which they applied spatiotemporal principal component analysis and Hessian-based decomposition to whole body kinematics to obtain compact descriptions of unconstrained throwing. They used these descriptions to quantitatively characterize performance, individual strategies, gender differences, and common patterns from a heterogeneous sample of nontrained throwers (30–32). Zhaoran Zhang also investigated ball throwing but compared movements in a virtual and a real set-up. Tolerance-Noise-Covariation decomposition revealed distinct stages of learning, indicating that subjects reached the stage of fine-tuning throwing variability in the real but not in the virtual task. These findings resonate with the reported problems in transferring therapeutic benefits from virtual to real environments (33). Expanding the research on tool use to the more exotic example of manipulating a bullwhip, Marta Russo investigated how humans can achieve dexterity in manipulating the wave dynamics of the whip's infinite DOFs. Their experimental and simulation results suggested that humans may represent control of this prodigiously complex dynamic object in terms of low-dimensional dynamic primitives. Thus, in the same set of studies, Moses Nah tested whether a distant target could be reached with a whip using a controller composed of only motor primitives. This approach was able to manage 54 DOFs by means of a single submovement in joint space. A detailed model of the whip dynamics was not needed for this approach. This may be a key simplification that humans leverage to learn complex motor skills, avoiding the need to internalize the detailed dynamic properties of the object being manipulated (34, 58).If details regarding objects’ properties do not need to be internalized, then what are the neural mechanisms underlying tool use? Simon Thibault showed that there is a functional overlap between tool-use planning and complex syntactic processing in the basal ganglia. Behaviorally, this is reflected by bidirectional cross-domain learning transfer, where tool use benefits syntax and vice-versa (35). In addition, Raeed Chowdhury studied the neural underpinnings of highly feedback-driven tasks, such as balancing a stick on a palm. During the task, monkeys displayed multiple control schemes within each trial, suggesting that they might have had multiple goals, and thus corresponding neural strategies, in different phases of the arm motion.Primates are not the only species that use strategies to control objects. Indeed, it is possible to perform complex tool manipulations without the benefit of specialized hands. New Caledonian crows are an excellent nonprimate animal model for studying complex object manipulation, as shown by Christian Rutz’s work. These animals exhibit a striking degree of dexterity with their beak when manufacturing tools from raw plant materials, using these tools to extract insect prey from hiding places in deadwood, and storing tools for future use in holes or behind tree bark (36).Such sophisticated control of our bodies extends beyond hand and upper-limb dexterity and is also exemplified by walking. Jacqueline Palmer presented their results on the relationship between motor cortical activity and circuit-specific cortico-cortical interactions during a whole body dual-task involving balance and cognition in older adults. Consistent with findings in younger adults, their results support motor cortical β activity as a potential biomarker for individual levels of balance challenge in older adults (37). To capture individual differences in gait dynamics, Taniel Winner used a data-driven dynamical model: a recurrent neural network (RNN) with long short-term memory. Measuring how the internal parameters of the model discriminated individuals with or without stroke, their work showed that using advanced models over discrete summary variables increased the accuracy of group classification. The ability to discriminate between different individuals may lead to the development of individually tailored rehabilitation to improve balance and gait in elderly or impaired individuals.Finally, speech is another example of a complex motor skill that requires the fine control of several muscles to produce a sound. To better understand this complexity, two talks focused on the effects of different perturbations during speech production. First, Zoe Swann assessed how a startling acoustic stimulus might affect word repetition in individuals with poststroke aphasia and apraxia. Startle exposure resulted in faster and louder speech. These results were analogous to the finding that a startle during upper extremity movement produced a higher probability of muscle activity onset in severe poststroke subjects who were unable to activate their arm muscles on their own (38). Second, Ding-lan Tang examined movement variability during speech production perturbed by auditory feedback. Motor variability increased with auditory perturbation, and this higher variability persisted even after removing the perturbation.The breadth of research highlighted in this section confirms a renewed interest in understanding the neural control of complex motor skills across species, body parts, interacting tools, and artificial limbs. Future work should look into transferring these abilities to robotic devices, as part of a continuing effort to close the loop between biological capability in humans and technical capability in robots.MULTIMODAL SENSORIMOTOR INTEGRATION IN HEALTH AND INJURYIn motor tasks, sensory and motor circuits interact to adjust motor commands to changes in the environment and to modulate the sensory experience to optimize task performance. Here we discuss talks from this year’s meeting that investigated sensory processing and sensorimotor interactions in both healthy and injured systems.Sliman Bensmaia and colleagues tackled the question of how multimodal sensory information is represented in the cuneate nucleus, a brainstem structure that receives sensory input from primary afferents in the forelimbs. They recorded single-unit activity in the S1, cuneate, and cutaneous primary afferents in response to skin stimulation and found that responses of cuneate neurons resembled those of S1, more so than those of primary afferents. Moreover, by using their novel simulation model (39), Bensmaia’s group was able to use the activity of 5–9 primary afferents from different cutaneous modalities to faithfully predict the response of single cells in the cuneate to skin stimulation. This study demonstrates that integration of multimodal sensory information occurs at the level of the cuneate, well before sensory input reaches the brain. A study led by Nofar Ozeri-Engelhard also demonstrated that multimodal integration of sensory inputs happens in the spinal cord. Using intersectional genetics, they isolated the parvalbumin-expressing interneurons located in the deep dorsal horn of the spinal cord (dPVs) to study their role in sensory processing and motor performance. With functional and histological assays, they provided evidence that dPVs form a circuit that integrates multimodal sensory information to directly communicate with motor neurons. To test whether this circuit plays a role in motor performance, they ablated dPVs and showed that mouse locomotion was perturbed. These results suggest that peripheral sensory circuits directly modulate motor output in the spinal cord to adjust motor performance to changes in the sensory environment.As we have seen, sensory processing affects movement. However, the opposite is also true: motor circuits interact with sensory pathways to modify sensory input, and consequently, the sensory experience. Kazuhiko Seki tested the idea that a copy of the motor command (i.e., efference copy) modulates sensory-evoked potentials (SEPs) recorded in the cuneate nucleus of monkeys, in response to sensory stimulation of primary afferents innervating the forelimb (40). SEPs were recorded during active, passive, and no movement (i.e., hold) conditions. SEP amplitude was attenuated during active movement compared to hold, suggesting that the motor command’s efference copy modulated cuneate sensory responses. Although to a lesser degree, attenuation was also observed during passive movements, demonstrating that other descending sensory inputs are involved in modulating the sensory response. In support of these conclusions, Seki presented anatomical evidence that the cuneate nuclei receive top-down projections from both the somatosensory and motor (new M1) cortices, proposing these descending projections as the source of attenuation.Eiman Azim and colleagues studied the role of descending cortico-cuneate pathways in the execution of tactile-guided movements. Using genetic tools in mice, they identified local inhibitory neurons in the cuneate that bidirectionally regulate the activity of cuneolemniscal neurons (which project from the cuneate to the thalamus). Optogenetic manipulation of these neurons altered the gain of the activity of the cuneoleminiscal neurons, and accordingly, the performance of dexterous movements (41). Anatomical experiments showed that both the inhibitory neurons and the cuneolemniscal neurons receive input from the sensory cortex. This supports Azim’s hypothesis that the cortex indirectly disinhibits, and directly inhibits cuneolemniscal neurons to augment sensory information necessary for optimal task performance while attenuating unnecessary information. In contrast to Seki’s findings, Azim’s work showed that descending pathways mostly originated from sensory areas. This discrepancy might be due to differences in species since new M1, which projects to the cuneate in non-human primates, does not exist in mice.In the cerebellum, the efference copy of motor commands is believed to be used by internal models to predict the sensory consequences of active movements. Sensory prediction can be used to distinguish between a sensory state arising from active (i.e., self-generated) versus passive (i.e., externally generated) movement. For example, although vestibular-spinal reflexes are essential for maintaining balance in response to a passive perturbation, these are counterproductive during active movements. Indeed, Kathleen Cullen’s group showed that the responses of vestibulospinal neurons in deep cerebellar nuclei (DCN) are suppressed during active compared to passive movement (42). Omid Zobeiri investigated whether Purkinje cells in the vestibular cerebellum, which inhibit DCN, can predict the sensory consequences of efference copy and in turn suppress DCN responses during active head movements. Although single Purkinje cells did not encode sensory prediction, they showed heterogeneous responses to vestibular and proprioceptive sensory inputs and motor efference copy. Simulation data suggested that combining the responses of ∼40 Purkinje cells is sufficient to generate sensory predictions that suppress DCN responses during active movements, providing evidence that cerebellar internal models are encoded by Purkinje cell sub-populations.During voluntary movements, when sensory prediction matches motor output (i.e., no sensory prediction error, SPE), DCN downregulate their responses to externally applied perturbations. But how does DCN sensitivity change in the presence of an SPE? Robyn Mildren studied how varying degrees of SPE impact the suppressed DCN response during active head movements. DCN responses were recorded from the rostral fastigial nucleus of one rhesus monkey. Different magnitudes of assistive and resistive torques were externally applied to introduce SPEs. As SPE increased, suppression of DCN responses during active movements gradually decreased, suggesting a gradual shift in encoding from self-generated to externally applied motion.In addition to head movements, primates use eye movements to sample the visual environment. Ehsan Sedaghat-Nejad investigated the role of the oculomotor cerebellum in encoding SPEs by recording Purkinje cell activity while marmosets performed saccades to a target presented in random locations. Occasionally, the target location was quickly changed, to introduce a prediction error. They showed that Purkinje cells can be divided into subpopulations, according to their encoding of SPEs, resulting in a population activity that was predictive of the saccade termination (43).Our brain’s ability to perform two sequential saccades has been long studied by Michael Goldberg, commended for his lifelong contributions to science and to the NCM community, and selected by NCM for the distinguished career award lecture. Although the first saccadic eye movement is an “easy” task for the brain, the second requires an updated representation of the initial position from which the movement starts. This position could be estimated from proprioceptive information about the eye, or from the efference copy of the first movement. To distinguish between these possibilities, the Goldberg laboratory investigated the lateral intraparietal area (LIP), a region that evokes saccades. In a series of experiments, they showed that although LIP receives proprioceptive information regarding eye position, this information arrives too late to alter second saccade planning (44). However, they found that LIP neurons fired before the second saccade, even though the second movement was not in their initial receptive field (45). This observation implies that the efference copy of the first saccade re-maps LIP receptive fields to generate the second saccade. What then is the role of proprioception in saccadic eye movements? Interestingly, with an increase in the number of sequential saccades, the brain shifts from an efference copy to proprioceptive-based encoding (46, 47).Modulation of bottom-up sensory input by top-down signals is important to maintain movement accuracy. But what role do these interactions serve in functional recovery? Corinna Darian-Smith and their laboratory investigated this question by performing incomplete lesions of dorsal roots innervating forelimb fingers. Although these lesions produced severe deficits in prehension tasks, monkeys exhibited an impressive recovery over 1 to 3 mo. To uncover the mechanism underlying recovery, the Darian-Smith laboratory

The dynamical systems view of movement generation in motor cortical areas has emerged as an effective way to explain the firing properties of populations of neurons recorded from these regions. Recently, many studies have focused on finding low-dimensional representations of these dynamical systems during voluntary reaching and grasping behaviors carried out by the forelimbs. One such model, the Poisson linear-dynamical-system (PLDS) model, has been shown to extract dynamics which can be used to decode reaching kinematics. However, few have investigated these dynamics, especially in non-human primates, during behaviors such as locomotion, which may involve motor cortex to a lesser degree. Here, we focused on unconstrained quadrupedal locomotion, and investigated whether unsupervised latent state-space models can extract low-dimensional dynamics while preserving information about hind-limb kinematics. Spiking activity from the leg area of primary motor cortex of rhesus macaques was recorded simultaneously with hind-limb joint positions during ambulation across a corridor, ladder, and on a treadmill at various speeds. We found that PLDS models can extract stereotyped low-dimensional neural trajectories from these neurons phase-locked to the gait cycle, and that distinct trajectories emerge depending on the speed and class of behavior. Additionally, it was possible to decode both the hind-limb kinematics and the gait phase from these inferred trajectories just as well or better than from the full neural population (18-80 neurons) with only 12 dimensions. Our results demonstrate that kinematics and gait phase during various locomotion tasks are well represented in low-dimensional latent dynamics inferred from motor cortex population activity.

Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms Corresponding author: Grégoire Courtine, PhD, Professor, International Paraplegic Foundation Chair in Spinal Cord Repair, Center for Neuroprosthetics and Brain Mind Institute, SWISS FEDERAL INSTITUTE OF TECHNOLOGY (EPFL), CH-1015 Lausanne, gregoire.courtine@epfl.ch. *, £, &contributed equally to this work Author Contributions. M.C., T.M. and D.B. contributed equally to this work. S.M., E.B. and J.B. contributed equally to this work. M.C. developed the spinal cord stimulation protocols and the routines for the identification of flexion and extension hotspots. T.M. developed the brain decoder and the decoder calibration routines. D.B. developed the experimental platform. M.C., T.M., F.W. and E.M.M. performed all the behavioural experiments with help from D.B., J.G., Y.J. and G.C.; M.C., T.M. and F.W. analysed the data, with input from E.M.M., J.B.M. and D.X.; M.C., T.M., F.W., E.M.M. and J.G. developed the real-time software application. N.B. and T.D. developed the Neural Research Programmer, with input from M.C., D.B., T.M., F.W. and J.G. Q.B. and E.R. processed the anatomical data. Y.J. trained all the monkeys. W.K.D.K., Q.L. and E.B. managed experimental protocols and procedures. P.D. developed and produced the spinal implants from design by M.C., D.B., J.B. and G.C.; J.B., D.B., Q.L. and G.C. performed the surgeries. G.C., S.M., E.B., J.B., and P.D. secured funding for the study. G.C. conceived and supervised the study. G.C. wrote the paper with M.C., T.M. and F.W., and all the authors contributed to its editing. Author Information. Data that supports the findings and software routines developed for the data analysis will be made available upon reasonable request to the corresponding author at gregoire.courtine@epfl.ch. Reprints and permissions information is available at www.nature.com/reprints. The authors declare competing financial interests: G.C., D.B., M.C., S.M., E.M.M. and J.B. hold various patents in relation with the present work. T.D. and N.B are Medtronic employees. In review of the manuscript they contributed to technical accuracy but did not influence the results or the content of the manuscript. E.B. reports personal fees from Motac Neuroscience Ltd UK and is a shareholder of Motac Holding UK and Plenitudes SARL France. G.C., S.M. and J.B. are founders and shareholders of G–Therapeutics BV. Europe PMC Funders Group Author Manuscript Nature. Author manuscript; available in PMC 2017 May 09. Published in final edited form as: Nature. 2016 November 10; 539(7628): 284–288. doi:10.1038/nature20118. E uope PM C Fuders A uhor M ancripts E uope PM C Fuders A uhor M ancripts Spinal cord injury disrupts the communication between the brain and the spinal circuits that orchestrate movement. To bypass the lesion, brain–computer interfaces1–3 have directly linked cortical activity to electrical stimulation of muscles, which have restored grasping abilities after hand paralysis1,4. Theoretically, this strategy could also restore control over leg muscle activity for walking5. However, replicating the complex sequence of individual muscle activation patterns underlying natural and adaptive locomotor movements poses formidable conceptual and technological challenges6,7. Recently, we showed in rats that epidural electrical stimulation of the lumbar spinal cord can reproduce the natural activation of synergistic muscle groups producing locomotion8–10. Here, we interfaced leg motor cortex activity with epidural electrical stimulation protocols to establish a brain–spinal interface that alleviated gait deficits after a spinal cord injury in nonhuman primates. Rhesus monkeys were implanted with an intracortical microelectrode array into the leg area of motor cortex; and a spinal cord stimulation system composed of a spatially selective epidural implant and a pulse generator with real-time triggering capabilities. We designed and implemented wireless control systems that linked online neural decoding of extension and flexion motor states with stimulation protocols promoting these movements. These systems allowed the monkeys to behave freely without any restrictions or constraining tethered electronics. After validation of the brain–spinal interface in intact monkeys, we performed a unilateral corticospinal tract lesion at the thoracic level. As early as six days post-injury and without prior training of the monkeys, the brain–spinal interface restored weight-bearing locomotion of the paralyzed leg on a treadmill and overground. The implantable components integrated in the brain– spinal interface have all been approved for investigational applications in similar human research, suggesting a practical translational pathway for proof-of-concept studies in people with spinal cord injury. A century of research in spinal cord physiology has demonstrated that the circuits embedded in lumbar segments of mammals can produce coordinated patterns of leg motor activity without brain input11,12. Various neuromodulation approaches have been developed to activate these circuits after injury to reestablish locomotion8,13–17. For example, epidural electrical stimulation (EES) of lumbar segments restored adaptive locomotion in paralyzed rats8. Recent studies showed that EES is also capable of activating lumbar spinal circuits in people with paraplegia14,16. These empirical observations prompted us to develop an evidenced-based framework to understand the interactions between EES and spinal circuits8–10. We aimed to exploit this knowledge to optimize stimulation protocols for clinical applications. Computational modelling and functional experiments revealed that EES engages spinal circuits through the modulation of proprioceptive feedback circuits10. This framework guided the design of spatiotemporal neuromodulation therapies that not only activate but also control the activity of spinal circuits engaging synergistic muscle groups8–10, enabling robust modulation of locomotor movements in rats whose spinal cords were void of brain input. However, volitional locomotion requires the brain to control the activity of spinal circuits. Brain–computer interface technologies1–4,18 provide the tools to link the intended motor states to EES protocols19–21 to reestablish voluntary control of locomotion after injury. For these developments, nonhuman primates are more appropriate models than rodents since Capogrosso et al. Page 2 Nature. Author manuscript; available in PMC 2017 May 09. E uope PM C Fuders A uhor M ancripts E uope PM C Fuders A uhor M ancripts they exhibit cortical engagement during locomotion similar to humans22, analogous recovery mechanisms from injury23, and comparable technological requirements24. Here, we decoded motor states from leg motor cortex activity to trigger EES protocols facilitating extension and flexion of the corresponding leg. We show that this brain–spinal interface alleviated gait deficits after spinal cord injury in nonhuman primates. To support the development of the brain–spinal interface, we established a wireless recording and stimulation platform in freely behaving, unconstrained and untethered nonhuman primates (Fig. 1 and Supplementary Video 1). Rhesus monkeys (Supplementary Table 1) were implanted with a microelectrode array into the leg area of the left motor cortex to record spiking activity from neuronal ensembles. Electromyographic signals were monitored using bipolar electrodes implanted into antagonist muscles spanning each joint of the right leg. Wireless modules enabled transmission of neural (20kHz) and electromyographic (2kHz) signals to external receivers25. We simultaneously acquired video recordings (100Hz) to reconstruct whole-body kinematics23. To deliver EES, we used technologies previously developed in rats9, which we adapted to the characteristics of spinal segments and vertebras measured in three monkeys (Extended Data Fig.1). These spinal implants were inserted into the epidural space over lumbar segments, and connected to an implantable pulse generator commonly used for deep brain stimulation therapy. We engineered wireless communication modules that enabled control over the spatial and temporal parameters of EES with a latency of about 100ms (Extended Data Fig.2). We first used well-established methods9,26 to identify the natural spatiotemporal pattern of motoneuron activation underlying locomotion. Our aim was to reproduce this pattern after injury. We conducted an anatomical tracing to identify the spatial distribution of motoneuron pools innervating antagonist muscles spanning each joint of the leg (Fig. 2a). We then projected the muscle activity recorded during locomotion onto motoneuron locations to visualize the spatiotemporal maps of motoneuron activation (Fig. 2c). These maps showed that locomotion involves the successive activation of well-defined hotspots located in specific regions of the spinal cord that were reproducible across monkeys (Extended Data Fig. 3). The most intense hotspots emerged in the caudal (L6/L7) and rostral (L1/L2) compartments of lumbar segments around the transitions between stance and swing phases. We labelled these hotspots extension and flexion hotspots, respectively. EES activates motoneurons through the recruitment of large-diameter proprioceptive fibres within the dorsal roots10,27. To access the extension and flexion hotspots, we targeted the dorsal roots projecting to spinal segments containing these hotspots. We reconstructed the spatial trajectory of the dorsal roots innervating each lumbar segment, and integrated this information together with motoneuron distribution into a unified library (Fig. 2a). We utilized the entry points of the dorsal roots as the targeted anatomical landmarks that guided th

A wireless brain–spine interface is presented that enables macaques with a spinal cord injury to regain locomotor movements of a paralysed leg. Grégoire Courtine and colleagues show that a fully implantable, wireless brain–spine interface can be used to improve locomotion after a unilateral spinal lesion in monkeys without training. The authors implanted monkeys with an electrode array in the leg area of the motor cortex and a stimulator in the lumbar spinal cord, enabling real-time decoding and stimulation. Decoded activity from the motor cortex was used to stimulate 'hotspot' locations in the lumbar spinal cord that control hindlimb flexion and extension during locomotion. Stimulating these hotspots enhanced flexion and extension of the target muscles during locomotion in intact monkeys and restored weight-bearing locomotion of the paralysed leg in monkeys with a unilateral spinal cord lesion six days after the injury. This proof-of-principle study shows that a similar system may improve or restore locomotion in people with spinal cord injury. Spinal cord injury disrupts the communication between the brain and the spinal circuits that orchestrate movement. To bypass the lesion, brain–computer interfaces1,2,3 have directly linked cortical activity to electrical stimulation of muscles, and have thus restored grasping abilities after hand paralysis1,4. Theoretically, this strategy could also restore control over leg muscle activity for walking5. However, replicating the complex sequence of individual muscle activation patterns underlying natural and adaptive locomotor movements poses formidable conceptual and technological challenges6,7. Recently, it was shown in rats that epidural electrical stimulation of the lumbar spinal cord can reproduce the natural activation of synergistic muscle groups producing locomotion8,9,10. Here we interface leg motor cortex activity with epidural electrical stimulation protocols to establish a brain–spine interface that alleviated gait deficits after a spinal cord injury in non-human primates. Rhesus monkeys (Macaca mulatta) were implanted with an intracortical microelectrode array in the leg area of the motor cortex and with a spinal cord stimulation system composed of a spatially selective epidural implant and a pulse generator with real-time triggering capabilities. We designed and implemented wireless control systems that linked online neural decoding of extension and flexion motor states with stimulation protocols promoting these movements. These systems allowed the monkeys to behave freely without any restrictions or constraining tethered electronics. After validation of the brain–spine interface in intact (uninjured) monkeys, we performed a unilateral corticospinal tract lesion at the thoracic level. As early as six days post-injury and without prior training of the monkeys, the brain–spine interface restored weight-bearing locomotion of the paralysed leg on a treadmill and overground. The implantable components integrated in the brain–spine interface have all been approved for investigational applications in similar human research, suggesting a practical translational pathway for proof-of-concept studies in people with spinal cord injury.