Supercomputer-based Models of Motoneurons for Estimating Their Synaptic Inputs in Humans

All motor commands flow through motoneurons in the spinal cord and brainstem. As for inputs to neural circuits throughout the CNS, these commands comprise three main components: two types of ionotropic input (excitation and inhibition) and a set of G-protein coupled inputs (neuromodulation). Lack of understanding of how these components produce output constitutes a fundamental uncertainty at the foundation of the neural control of movement. Fortunately, motor output in humans can be studied at the level of single neurons. Motoneuron action potentials are 1-to-1 with those of their muscle fibers, forming motor units whose action potentials can be recorded relatively easily in muscles. The potential for using these motor unit firing patterns for understanding motor commands has long been appreciated. Our goal is to maximize this potential by developing supercomputer-based techniques for reverse engineering motor unit firing patterns to identify the amplitudes and patterns of the excitatory, inhibitory and neuromodulatory inputs underlying motor commands in humans. Recent advances that allow simultaneous recording of many motor units have allowed us to identify distinctive nonlinear behaviors in motor unit firing patterns. Our development of realistic models of motoneurons show that these nonlinearities arise from complex interactions between input components. We plan to use these models as the core of a reverse engineering (RE) approach that estimates these three components from nonlinear human motor unit firing patterns. Our premise is that implementation of our models on supercomputers at Argonne National Laboratories will allow systematic exploration of the firing patterns generated by many thousands of input combinations. Those input organizations that accurately recreate a measured set of firing patterns will then be considered to be part of the “solution space” for that particular motor output. The key problem for this analysis is redundancy. If the same motor output can be produced by many input combinations, then reverse engineering will reveal huge solution spaces that provide little insight into motor commands. Overall motor outputs like force and EMG suffer from this problem. Our concept, however, is that measuring motor output at the single neuron level, via motor unit recordings, allows for effective reverse engineering. We have 3 aims: 1) to develop and evaluate supercomputer-based reverse engineering techniques for analysis of motor unit firing patterns. 2) to deploy RE to investigate the mechanisms of muscle-specific differences in populations of motor unit firing patterns. And 3) to deploy RE to investigate whether inhibitory-neuromodulation interactions that are specific for each muscle are relatively fixed, or instead are continuously adapted for different motor tasks. The development of supercomputer-based analysis techniques provides an ideal complement to emergence of techniques to measure firing patterns of large populations of motor units. Our novel reverse engineering method have the potential to transform our understanding of the synaptic organization of motor commands in humans.