Portal de Periódicos da CAPES

Highlights from the 32nd Annual Meeting of the Society for the Neural Control of Movement

2023; American Physiological Society; Volume: 131; Issue: 1 Linguagem: Inglês

10.1152/jn.00428.2023

1522-1598

K Love, Di Cao, Joanna Chang, Lucas Rebelo Dal’Bello, Xuan Ma, Daniel J. O’Shea, Hunter R. Schone, Mahdiyar Shahbazi, Adam L. Smoulder,

Tactile and Sensory Interactions

EditorialHighlights from the 32nd Annual Meeting of the Society for the Neural Control of MovementKassia Love, Di Cao, Joanna C. Chang, Lucas R. Dal’Bello, Xuan Ma, Daniel J. O’Shea, Hunter R. Schone, Mahdiyar Shahbazi, and Adam SmoulderKassia LoveMassachusetts Eye and Ear, Boston, Massachusetts, United States, Di CaoDepartment of Mechanical Engineering, Johns Hopkins University, Baltimore, Maryland, United StatesCenter for Movement Studies, Kennedy Krieger Institute, Baltimore, Maryland, United States, Joanna C. ChangDepartment of Bioengineering, Imperial College London, London, United Kingdom, Lucas R. Dal’BelloLaboratory of Neuromotor Physiology, IRCCS Fondazione Santa Lucia, Rome, Italy, Xuan MaDepartment of Neuroscience, Northwestern University, Chicago, Illinois, United States, Daniel J. O’SheaDepartment of Bioengineering, Stanford University, Stanford, California, United States, Hunter R. SchoneRehabilitation and Neural Engineering Laboratory, University of Pittsburgh, Pittsburgh, Pennsylvania, United StatesDepartment of Physical Medicine and Rehabilitation, University of Pittsburgh, Pittsburgh, Pennsylvania, United States, Mahdiyar ShahbaziWestern Institute for Neuroscience, Western University, London, Ontario, Canada, andAdam SmoulderDepartment of Biomedical Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania, United StatesCenter for the Neural Basis of Cognition, Pittsburgh, Pennsylvania, United StatesPublished Online:16 Jan 2024https://doi.org/10.1152/jn.00428.2023This is the final version - click for previous versionMoreSectionsPDF (1 MB)Download PDF ToolsExport citationAdd to favoritesGet permissionsTrack citations ShareShare onFacebookTwitterLinkedInWeChat INTRODUCTIONThe Society for the Neural Control of Movement (NCM) held its 32nd Annual Meeting in beautiful Victoria, British Columbia, Canada in April of 2023. Over 350 researchers, clinicians, and students from 17 countries came together to discuss the latest findings in the neural underpinnings of motor control (Fig. 1). The presented research featured diverse perspectives across a broad range of species (from fruit flies to bats to humans) and approaches (from theoretical computational models to large-scale neural population recordings; Fig. 2). Altogether, the conference provided a platform to deepen our understanding of how the brain skillfully orchestrates movement, enabling us to navigate and interact with the world around us.Figure 1.Neural Control of Movement (NCM) 2023 attendance map based on country of primary affiliation (color bar log-scaled).Download figureDownload PowerPointFigure 2.The most common words in submitted abstracts of Neural Control of Movement (NCM) conference presenters.Download figureDownload PowerPointThe conference opened with a satellite meeting on the “Computations and neural code underlying the control of posture” and then moved on to a main meeting composed of panel discussions, individual talks, and poster presentations. Notably, Chris Miall was invited to speak as the recipient of this year’s Distinguished Career Award, for his work on the predictive role of the cerebellum. As the recipient of the Early Career Award, Juan Álvaro Gallego also delivered a keynote focusing on his research in neural manifolds for motor control. This year, NCM continued its efforts to increase gender representation, with around 40% of women for both its featured speakers and general membership.As a subset of the trainee scholarship award winners of NCM 2023, we wrote this article with the aim of highlighting the innovative research that was presented at this year’s conference. As highlighted in previous years (1–3), NCM plays a role in synthesizing the various aspects of neural motor control through interdisciplinary discussions and collaborations. In this article, the highlights are focused on seven themes: 1) evolutionary perspectives of motor control, 2) prediction and coordination in the cerebellum, 3) de novo learning and memory, 4) internal states and errors, 5) sensory feedback, 6) latent neural dynamics, and 7) translational efforts to understand and restore motor function. Through our article, we hope to convey the innovation and potential that we witnessed at the conference, encouraging further discussions and exploration in the years to come.EVOLUTIONARY PERSPECTIVE AND INSIGHTS FROM DIVERSE SPECIESBillions of years of evolutionary adaptations and continuous refinement mold behavior. As such, a panel at NCM emphasized the value of understanding behavior and its underlying neural processes through an evolutionary lens. Combining insights from theory, experiments, and history, the panel painted a picture of how movement of all creatures builds on the neural developments of its ancestors.In the panel, Paul Cisek (University of Montreal) provided an overview of how the nervous system and the way animals control their behavior have evolved over time, from the earliest chordates to primates (4). Examining this history highlighted how evolutionary changes build on the existing nervous system to increase the complexity of actions that different species can perform. As an example, Cisek examined the evolution of the dopaminergic system. Initially, dopamine (DA) likely acted as the governing element between exploratory versus exploitative food finding behavior. If the nutrient intake was high, dopamine levels would rise, thereby encouraging animals to stay in a local area with abundant resources. Conversely, when resources were depleted and hunger persisted, dopamine levels would decrease, motivating them to engage in exploratory movements to seek out new patches. Over time, the brain evolved specialized regions that led to more complex uses of dopamine. Brain areas that initially only involved movement selection (e.g., the subpallium) began to incorporate dopamine signals to reinforce rewarding actions, a process that resulted in the basal ganglia over millions of years of evolution. Through this example, Cisek demonstrated how evolutionary changes led to more sophisticated action-selection capabilities by slowly building upon the complexity of existing systems.Drawing parallels from a computational perspective, Terence Sanger (University of California, Irvine) described how increasingly more complex control systems can also be built by expanding on simpler systems. Starting from basic reflexes, they showed how incrementally including additional capabilities like reward prediction, sensory integration, and dynamic control can help animals produce more advantageous behaviors for their survival. Sanger further described how each capability may arise from different components of the nervous system. Thus, although there may be many possible computational models that describe animal behavior and the underlying neural data, Sanger and Cisek proposed that an evolutionary perspective allows us to pinpoint the models that are compatible with how the nervous system evolved.Expanding on this evolutionary framework, Leah Krubitzer (University of California, Davis) examined how evolution drove the development of different cortical maps for limb and hand representations in different species (5). By synthesizing intracortical stimulation experiments across mammalian species, roughly chronologically following the depth of the phylogenetic tree, they explored how brains from different species developed to accomplish unique behaviors.Beginning with opossums and rats, Krubitzer showed that a limited section of M1/S1 stimulation evokes gross movement, often involving many limbs. As they detailed experiments from tree shrews to early primates, species with greater hand/digit usage exhibited more brain areas (M1, PPC, 3a and 3b) evoking more precise movements (specific digits, wrist). Finally, cortical stimulation in nonhuman primates often used for laboratory studies (cebus, macaques) could evoke movement of individual digits and lead to specific grip types, naturally coinciding with the development of opposable thumbs and the need to use fingers for tool use. In contrast to these other mammals, bats have little need for dexterous finger movements. Instead, they have to fly and echolocate. In cortical stimulation of Egyptian fruit bats, Krubitzer observed movements evoked for the shoulders, hindlimbs, and tongue in huge swathes of cortex, precisely aligning with the movements needed for the bats’ most crucial behaviors. Together, these results show that the cortex evolved to support the movements that are most relevant to the dexterous needs of each species. Even though evolutionary history plays a large role in defining these cortical maps, Krubitzer noted that development also plays a key role, showing that mice reared in more natural environments exhibited greater forelimb representations compared with laboratory-reared mice.Whereas Krubitzer emphasized how species-specific behaviors are reflected in cortical influence, Andreas Kardamakis (Universitas Miguel Hernandez) probed how deeply evolutionarily conserved behaviors can be studied across species. The behavior they focused on was “knowing where to look” for orienting, attending, and avoiding different stimuli. Specifically, the neural circuitry of the superior colliculus/tectum has been demonstrated to be necessary for this process across a vast array of species. By comparing tectal function in lamprey with that of mammals, Andreas highlighted how activity driven by evolutionarily conserved inputs (i.e., from the retina, midbrain) can be altered or countered by cortical input in more developed species (6). Thus, although common principles govern visuomotor transformations and center-surround interactions, their specific circuit implementations might differ across species and understanding the evolutionary relationships between species can uncover these mechanisms. Outside of the evolution panel, Brandie Verdone (Johns Hopkins University) took a similar approach, describing how even though mice are nonfoveate creatures, their head orientation behavior and circuitry are similar to those of primates since the mechanisms are likely evolutionarily conserved. Further research into the motor circuits of a variety of species, like the descending neurons of fruit flies as discussed in Helen Yang’s (Harvard University) talk, could thus yield generalizable knowledge concerning motor control.Together, these results suggest that an evolutionary perspective may provide insight into the neural control of movement in several ways. Understanding how neural circuits evolved to achieve more complex behaviors grounds our understanding of the functions of different brain areas: some neural mechanisms in modern animals that may seem obscure or convoluted may be interpretable in the context of evolutionary constraints. By recognizing the evolutionary relationships between different animals, we can also identify mechanisms that may be conserved across species and translate findings across species. This idea aligns with a broader class of talks at NCM 2023 that emphasized the importance of studying motor control across diverse species. Although research presented at NCM is still mainly focused on a few species (namely humans, monkeys, mice, and rats), a growing trend of highlighting more species may yield greater insight through an evolutionary lens.CEREBELLAR CIRCUITS, PREDICTION, AND COORDINATIONChris Miall (University of Birmingham) delivered a keynote talk on internal models in the cerebellum, discussing extensive experimental evidence from experiments by their laboratory and others in nonhuman primates and humans. One kind of internal model, a forward model, is a neural mechanism that generates predictions about the future sensory outcomes of motor commands (7, 8). Learning a predictive forward model allows the brain to estimate the state of the body and the environment despite sensory delays and to optimize and adjust movements for more accurate and coordinated actions (9, 10). In macaques, Purkinje cell simple spikes represent both the anticipated sensory consequences of movement (11) as well as error signals derived from sensory feedback (12). Beyond its well-established role in movement, a growing body of evidence implicates the cerebellum in cognition, social behavior, and language processing. To study this, Miall and colleagues have also developed approaches to study cerebellar activity in humans. These studies suggest roles for cerebellar internal models in motor coordination, motor learning, and the consolidation of motor memories. Using the high sensitivity and temporal resolution of optically pumped magnetometer-based magnetoencephalography (MEG-OPM), Miall and colleagues have begun to study evoked signals in the cerebellum, opening the possibility of studying the role of complex spike signaling in sophisticated cognitive behaviors in humans. Finally, Miall presented a few studies using transcranial magnetic stimulation (TMS) and event-triggered transcranial direct-current stimulation (TDCS) to disrupt cerebellum during motor and language processing tasks. These effects demonstrate a causal contribution of the cerebellum to the adaptive processes that underlie prediction in motor and language domains and also highlight the need for a deeper understanding of cerebellar functions and its interactions with other brain systems.Megan Carey (Champalimaud Centre for the Unknown), Andrew Pruszynski (Western University), Abigail Person (University of Colorado), and Javier Medina (Baylor College of Medicine) explored these topics further in their panel titled “Internal models: From systems to circuits.” Pruszynski presented fascinating evidence that adaptive, predictive models of the arm’s dynamics transfer between feedforward and feedback control systems work. Because of the intersegmental coupling of the arm’s dynamics, rotation of the elbow induces torques at the shoulder; because of sensory delays, the motor system cannot wait for errors induced by elbow motion to propagate to the shoulder but instead learns to predictively compensate for these intersegmental dynamics. Pruszynski showed that when this coupling is artificially removed by locking rotation of the shoulder, learning of the new dynamics via perturbations (engaging feedback corrections) leads to reduced activation during voluntary control, indicating that both processes engage a shared internal. This work was spearheaded by the late Rodrigo Maeda, a brilliant and exceptionally generous young member of our community who passed away from stomach cancer in early 2021.Abigail Person discussed their laboratory’s efforts to map feedforward motor control mechanisms onto cerebellar circuitry. They presented recent work performed in mice during a forelimb reaching task. Neurons in the interposed anterior nucleus, an output pathway of the cerebellum, causally regulate reach deceleration (13). They showed that in the cerebellar cortex Purkinje cells display an inverse population code of reach velocity and developed a learning model to explore how synaptic plasticity could train a graded, precisely timed, adaptive suppression of cerebellar output neurons to achieve precisely aimed reaching movements (14).Megan Carey explored the role of predictive internal models in establishing skilled locomotion, which requires precise coordination of the entire body. Mouse models of cerebellar ataxia display specific coordination deficits during locomotion, including large, delayed, reactive movements of the tail, that are consistent with a failure to compensate for the predicted consequences of movement. Purkinje cells display widespread, diverse stride modulation during locomotion, often with mixed selectivity to multiple paws, as well as modulation with other behaviors like licking. This encoding scheme could allow for simple linear readouts of the Purkinje population activity to facilitate coordinated motor control in navigating dynamic environments. An important consideration is that producing skilled locomotion is not restricted to the cerebellum. Indeed, a separate talk by Abhilasha Joshi (University of California, San Francisco) highlighted the role of the hippocampus in limb coordination in freely behaving rats. As such, understanding how disease affects different brain regions and their contribution to limb coordination is crucial.Finally, Javier Medina reported an intriguing recurrent cerebellar circuit that may link together inverse and forward internal models. An inverse model computes the motor commands needed to achieve a specific motor goal, whereas a forward model predicts the motor and sensory effects of that action. Medina’s laboratory studied the rostral anterior interposed nucleus (rAIN) of the cerebellum in a classical conditioning task where a mouse learns to produce anticipatory eyeblinks (conditioned response or CR) to a light cue preceding an air puff that drives reflexive blinking (unconditioned response or UR). They presented evidence that a central CR module in the cerebellar cortex may act as the inverse model, inhibiting rAIN after the light cue and driving the conditioned blinking response. A recurrent circuit then couples the central CR module to surrounding cerebellar cortical modules, whose activity encodes predictions of eyelid position 10 ms in the future, suggesting that these surround modules act as forward models that compute the consequences of the motor commands produced by the inverse model in the CR module. The full circuit connectivity and the role of climbing fiber input in this recurrent circuit need to be elaborated; this work provides an exciting new direction in how cerebellar circuitry links action generation and prediction to learn and coordinate skilled movements.The cerebellum’s critical role in motor function implies that cerebellar injury or degeneration can lead to severe consequences. Cerebellar dysfunction encompasses a wide range of motor disorders symptoms, including impaired interjoint coordination, imprecise movements, balance problems, and the inability to perform repeated movements. However, we are still missing a link between the structural changes in the cerebellum that are caused by disease and the changes in motor function. Di Cao (Johns Hopkins University) offered insights into the impact of cerebellar ataxia on motor function, as the computational deficits in feedforward and feedback control pathways. Participants performed the hand tracking task in a virtual reality environment (further adapted from Ref. 15). Using a control-theoretic experimental approach, their work showed that damage to the cerebellum does cause two computational deficits: 1) cerebellar damage increases time delay differently on feedforward and feedback pathways, and the more severe ataxia, the greater feedback time delay; 2) cerebellar damage reduces the ability to extract useful information from time-lead (preview) information. Typically, cerebellar degeneration is characterized by the loss of cerebellar cells, such as Purkinje cells. In this light, Jovin Jacobs (Champalimaud Centre for the Unknown) presented recent findings on how the loss of Purkinje cells/granule cells in adult mice affects locomotion (walking) and locomotor learning (split-belt treadmill adaptation). Their research revealed 1) the extent to which ataxia and learning deficits scale with cell death magnitude and that 2) locomotion and locomotor learning may be more robust to loss of Purkinje cells than granule cells (16, 17).There are noticeable parallels in the symptoms exhibited by cerebellar ataxia and age-related motor dysfunction (such as problems with balance and coordination). Actually, both conditions lead to macroscopic changes in cerebellar structure, and they affect overlapping pieces of the cerebellar circuit. A panel titled “Age- and disease-related changes in the cerebellum impact motor function” aimed to bridge this knowledge gap by offering an overview of how changes in the cerebellum associated with age or disease impact motor function.Jean-Jacques Orban de Xivry (KU Leuven) gave a talk on the effect of aging on cerebellar motor tasks. Across various motor paradigms linked to internal model function (implicit motor adaptation and sensory attenuation), the performance of older participants was found to be at least as good as or even “better” than that of younger participants (18, 19). These results suggest that the cerebellum can maintain at least partially internal model function despite age-related degeneration. They further hypothesized the existence of a motor reserve, similar to the idea of a cognitive reserve, to explain these results. Alanna Watt (McGill University) discussed recent work from their laboratory demonstrating that similar cellular changes occur in Purkinje cells in the mouse cerebellum in both aging and ataxia. They investigated whether cerebellar alterations contribute to aging-related motor decline. They observed an age-dependent reduction in Purkinje cell firing rates across adult life span. Furthermore, their group used chemogenetics to reduce firing rates in Purkinje cells, which led to a decrease in motor coordination in young mice. These findings suggest that aging-related decreases in Purkinje cell firing rates contribute to declining motor coordination and that targeting these physiological changes may lead to future therapeutic avenues to improve motor aging (20).LEARNING AND MEMORYWe have a remarkable ability to learn and retain new motor skills, along with the neural substrates that underpin the behavior. NCM 2023 had talks on research that explored how motor skills are learned, specifically in the context of research on de novo learning, and how these newly formed motor skills are retained. To understand the neural substrates behind how we learn new skills, several researchers presented their work during a session on de novo learning, which is the learning of a task using a novel controller in contrast with adapting an already existing controller (21). The de novo learning approach is especially useful for understanding how a local neuron network adapts to support learning a new motor skill. The research on de novo learning has investigated the behavioral, computational, and neural levels that lend to our ability to expand our motor functionality. This then lays the groundwork for investigations on how newly learned motor skills are retained and the characteristics of these motor memories.Behavioral tasks have played a crucial role in deepening our understanding of the process of de novo learning. Andrea d’Avella (IRCCS Fondazione Santa Lucia/University of Messina) presented results from a myoelectric control task, where surface electromyograms (EMGs) from the arm were used to control the position of a cursor on a screen. By virtually modulating the force that each muscle generates in the movement of the cursor, they were able to create a compatible and an incompatible muscle synergy environment. They showed that de novo learning is expressed as learning new muscle synergies in a novel incompatible environment, in contrast to a compatible environment that only requires recombining existing synergies (22). On the first day of practice, task performance improved quicker in the compatible environment. However, after 3 days of training task performance improvement was comparable across the compatible and incompatible environments, suggesting that new muscle synergies had been learned. Therefore, it was concluded that learning of new muscle synergies occurs when the existing muscle synergies are not sufficient to do a task. Yet, this muscle synergy control scheme prompts further inquiry into the relationship between the newly learned controller and adaptation of a preexisting controller. Kahori Kita (Johns Hopkins University) presented their work with a bimanual kinematic cursor control task that explores this inquiry. Participants were tasked with moving both of their hands to control a cursor on a screen. One hand controlled horizontal movement of the cursor, and the other hand controlled vertical movement. They called this bimanual control mapping. They compared this control mapping to a baseline mapping, which mapped the location of the right hand to cursor position (23). When participants switched between these two mappings, there were minor performance costs, suggesting that control policies are stored distinctly. This means that the bimanual mapping was not learned through adaptation but through de novo learning.One aspect of de novo learning that was noted by both d’Avella and Kita is that de novo learning takes multiple days. One hypothesis as to why this mode of learning is slow is that it requires an exploration of the high-dimensional motor space. In other words, the various degrees of freedom of the human body, leading to a high-dimensional motor space, call for exploration of muscle activations to acquire the new motor skill. To support this hypothesis, Lucas Dal’Bello (University of Tsukuba) presented a computational model of error-based de novo learning, showing how exploration of the entire motor space can be leveraged to learn an internal model used for error correction. Learning this internal model then in turn is used to learn a motor task in an error-based way (24). Additionally, Dal’Bello showed how their model explains the results of other motor learning experiments (such as arm reaching and cursor control tasks), suggesting that motor exploration plays a key role in de novo learning.In light of the finding that de novo learning is a slow process, accelerating this learning process has important implications for technology that relays communication from the brain to external devices. Technology like brain-computer interfaces (BCIs) or prosthetics has great promise in increasing patients’ quality of life by restoring motor function. Therefore, it is crucial that adapting to new control mechanisms is efficient and effective. One possible approach to accelerate learning de novo was presented by Amy Orsborn (University of Washington) in their research using decoders that adapt motor commands required for a task based on the user’s performance. Decoders are algorithms that translate recorded neural activity into a command for the computer. Orsborn utilized a closed-loop decoder adaptation algorithm (CLDA) while subjects learned to control a BCI (25). This type of algorithm updates the decoder’s parameters through feedback performance of the assigned task. Amazingly, CLDA facilitates faster improvement in closed-loop BCI performance. This was attributed to the increased decoder update frequency by recording batches of neural activity in short time intervals. As such, the next generation of neural decoders need to consider user learning and the intrinsic plasticity of the sampled neurons.With this understanding of how motor skills can be learned de novo, there was a separate conference session that explored and characterized retention of newly learned motor skills and how the brain adapts to retain these motor memories. To explore the neuromuscular and kinematic space of learning and retention, Dagmar Sternad (Northeastern University) and Se-Woong Park (University of Texas at San Antonio) did a longitudinal study that had human subjects performing a polyrhythmic bimanual task (26). They tasked subjects to swing a pendulum in each hand at the same time but with different swinging frequencies for each hand. Their analysis included characterization of the learning but, more remarkably, the longitudinal retention of the task. Their participants could perform the learned task 6 mo and, for some, even as long as 8 yr after learning it. Despite these extensive gaps in time, participants maintained the key features of the skill and even demonstrated the ability to generalize the skill to other rhythms. This suggests that motor skills can remain stable over time, showcasing the robustness of the underlying kinematics of motor learning. However, it is unclear whether the neural substrates underlying these skills are also stable.Expanding on this idea, Hunter Schone (University of Pittsburgh) provided a strong demonstration for the long-term retention and stability of sensorimotor representations in the brain. Using functional (f)MRI, he longitudinally tracked patients with planned hand amputations to investigate the impact of amputation on cortical body representations. By asking amputees to move their fingers before amputation and their phantom fingers after amputation, he observed highly stable cortical hand and body representations, even up to 6 mo after amputation. This suggests that cortical body representations are highly stable over time, and even despite arm amputation (27).Understanding the neural substrates of motor learning is crucial for motor skill training and motor assistive technology. With a greater understanding of the neural circuits and motor representations in the brain, we can develop lasting therapies for neuro-motor pathologies. Steve Chase’s (Carnegie Mellon University) work has offered some insight into the neural substrates involved in motor learning. Specifically, they tested whether learning new skills impacts the neural activity of preexisting learned skills. Utilizing a BCI task, Chase trained rhesus macaques to control a cursor by modulating their own motor cortex neural activity by using an intuitive neural control mapping to control the BCI (28). The task they performed was a center-out reaching task where they would move the cursor from the middle of the screen to some peripheral target. Chase then perturbed the BCI neural mapping to create a novel mapping to investigate the adaptability and redundancy of existing motor representations on task performance. They observed that the novel mapping alters the neural activity for the familiar task without decreasing perfor