Impact of neural noise on a sensory-motor pathway signaling impending collision

Reduced corticospinal drive to antagonist muscles of upper and lower limbs during hands-and-knees crawling in infants with cerebral palsy: Evidence from intermuscular EMG-EMG coherence

BACKGROUND

There is growing interest in understanding the central control of hands-and-knees crawling, especially as a significant motor developmental milestone for early assessment of motor dysfunction in infants with cerebral palsy (CP) who have not yet acquired walking ability. In particular, CP is known to be associated with walking dysfunctions caused by early damage and incomplete maturation of the corticospinal tract. However, the extent of damage to the corticospinal connections during crawling in infants with CP has not been fully clarified. Therefore, this study aimed to investigate the disparities in intermuscular EMG-EMG coherence, which serve as indicators of corticospinal drives to antagonist muscles in the upper and lower limbs during crawling, between infants with and without CP.

METHODS

This study involved 15 infants diagnosed with CP and 20 typically developing (TD) infants. Surface EMG recordings were obtained from two pairs of antagonist muscles in the upper limbs (triceps brachii (TB) and biceps brachii (BB)) and lower limbs (quadriceps femoris (QF) and hamstrings (HS)), while the infants performed hands-and-knees crawling at their self-selected velocity. Intermuscular EMG-EMG coherence was computed in two frequency bands, the beta band (15-30 Hz) and gamma band (30-60 Hz), which indicate corticospinal drive. Additionally, spatiotemporal parameters, including crawling velocity, cadence, duration, and the percentage of stance phase time, were calculated for comparison. Spearman rank correlations were conducted to assess the relationship between EMG-EMG coherence and crawling spatiotemporal parameters.

RESULTS

Infants with CP exhibited significantly reduced crawling velocity, decreased cadence, longer cycle duration, and a higher percentage of stance phase time compared to TD infants. Furthermore, CP infants demonstrated decreased coherence in the beta and gamma frequency bands (indicators of corticospinal drive) in both upper and lower limb muscles. Regarding limb-related differences in the beta and gamma coherence, significant disparities were found between upper and lower limb muscles in TD infants (p < 0.05), but not in infants with CP (p > 0.05). Additionally, significant correlations between coherence metrics and crawling spatiotemporal parameters were identified in the TD group (p < 0.05), while such correlations were not evident in the CP group.

CONCLUSIONS

Our findings suggest that the corticospinal drive may functionally influence the central control of antagonist muscles in the limbs during typical infant crawling. This functional role could be impaired by neurological conditions such as cerebral palsy. The neurophysiological markers of corticospinal drive, specifically intermuscular EMG-EMG coherence during crawling in infants with cerebral palsy, could potentially serve as a tool to assess developmental response to therapy.

Contrast polarity-specific mapping improves efficiency of neuronal computation for collision detection

Neurons receive information through their synaptic inputs, but the functional significance of how those inputs are mapped on to a cell's dendrites remains unclear. We studied this question in a grasshopper visual neuron that tracks approaching objects and triggers escape behavior before an impending collision. In response to black approaching objects, the neuron receives OFF excitatory inputs that form a retinotopic map of the visual field onto compartmentalized, distal dendrites. Subsequent processing of these OFF inputs by active membrane conductances allows the neuron to discriminate the spatial coherence of such stimuli. In contrast, we show that ON excitatory synaptic inputs activated by white approaching objects map in a random manner onto a more proximal dendritic field of the same neuron. The lack of retinotopic synaptic arrangement results in the neuron's inability to discriminate the coherence of white approaching stimuli. Yet, the neuron retains the ability to discriminate stimulus coherence for checkered stimuli of mixed ON/OFF polarity. The coarser mapping and processing of ON stimuli thus has a minimal impact, while reducing the total energetic cost of the circuit. Further, we show that these differences in ON/OFF neuronal processing are behaviorally relevant, being tightly correlated with the animal's escape behavior to light and dark stimuli of variable coherence. Our results show that the synaptic mapping of excitatory inputs affects the fine stimulus discrimination ability of single neurons and document the resulting functional impact on behavior.

Predicting individual neuron responses with anatomically constrained task optimization

Artificial neural networks trained to solve sensory tasks can develop statistical representations that match those in biological circuits. However, it remains unclear whether they can reproduce properties of individual neurons. Here, we investigated how artificial networks predict individual neuron properties in the visual motion circuits of the fruit fly Drosophila. We trained anatomically constrained networks to predict movement in natural scenes, solving the same inference problem as fly motion detectors. Units in the artificial networks adopted many properties of analogous individual neurons, even though they were not explicitly trained to match these properties. Among these properties was the split into ON and OFF motion detectors, which is not predicted by classical motion detection models. The match between model and neurons was closest when models were trained to be robust to noise. These results demonstrate how anatomical, task, and noise constraints can explain properties of individual neurons in a small neural network.

A model of feedforward, global, and lateral inhibition in the locust visual system predicts responses to looming stimuli

Detection of looming obstacles is a vital task for both natural and artificial systems. Locusts possess a visual nervous system with an extensively studied obstacle detection pathway, culminating in the lobula giant movement detector (LGMD) neuron. While numerous models of this system exist, none to date have incorporated recent data on the anatomy and function of feedforward and global inhibitory systems in the input network of the LGMD. Moreover, the possibility that global and lateral inhibition shape the feedforward inhibitory signals to the LGMD has not been investigated. To address these points, a novel model of feedforward inhibitory neurons in the locust optic lobe was developed based on the recent literature. This model also incorporated global and lateral inhibition into the afferent network of these neurons, based on their observed behaviour in existing data and the posited role of these mechanisms in the inputs to the LGMD. Tests with the model showed that it accurately replicates the behaviour of feedforward inhibitory neurons in locusts; the model accurately coded for stimulus angular size in an overall linear fashion, with decreasing response saturation and increasing linearity as stimulus size increased or approach velocity decreased. The model also exhibited only phasic responses to the appearance of a grating, along with sustained movement by it at constant speed. By observing the effects of altering inhibition schemes on these responses, it was determined that global inhibition serves primarily to normalize growing excitation as collision approaches, and keeps coding for subtense angle linear. Lateral inhibition was determined to suppress tonic responses to wide-field stimuli translating at constant speed. Based on these features being shared with characterizations of the LGMD input network, it was hypothesized that the feedforward inhibitory neurons and the LGMD share the same excitatory afferents; this necessitates further investigation.

Active membrane conductances and morphology of a collision detection neuron broaden its impedance profile and improve discrimination of input synchrony

How neurons filter and integrate their complex patterns of synaptic inputs is central to their role in neural information processing. Synaptic filtering and integration are shaped by the frequency-dependent neuronal membrane impedance. Using single and dual dendritic recordings in vivo, pharmacology, and computational modeling, we characterized the membrane impedance of a collision detection neuron in the grasshopper Schistocerca americana. This neuron, the lobula giant movement detector (LGMD), exhibits consistent impedance properties across frequencies and membrane potentials. Two common active conductances g_H and g_M, mediated respectively by hyperpolarization-activated cyclic nucleotide-gated (HCN) channels and by muscarine-sensitive M-type K⁺ channels, promote broadband integration with high temporal precision over the LGMD's natural range of membrane potentials and synaptic input frequencies. Additionally, we found that a model based on the LGMD's branching morphology increased the gain and decreased the delay associated with the mapping of synaptic input currents to membrane potential. More generally, this was true for a wide range of model neuron morphologies, including those of neocortical pyramidal neurons and cerebellar Purkinje cells. These findings show the unexpected role played by two widespread active conductances and by dendritic morphology in shaping synaptic integration.NEW & NOTEWORTHY Neuronal filtering and integration of synaptic input patterns depend on the electrochemical properties of dendrites. We used an identified collision detection neuron in grasshoppers to examine how its morphology and two conductances affect its membrane impedance in relation to the computations it performs. The neuronal properties examined are ubiquitous and therefore promote a general understanding of neuronal computations, including those in the human brain.

M current regulates firing mode and spike reliability in a collision-detecting neuron

All animals must detect impending collisions to escape and reliably discriminate them from nonthreatening stimuli, thus preventing false alarms. Therefore, it is no surprise that animals have evolved highly selective and sensitive neurons dedicated to such tasks. We examined a well-studied collision-detection neuron in the grasshopper ( Schistocerca americana) using in vivo electrophysiology, pharmacology, and computational modeling. This lobula giant movement detector (LGMD) neuron is excitable by inputs originating from each ommatidia of the compound eye. It possesses many intrinsic properties that increase its selectivity to objects approaching on a collision course, including switching between burst and nonburst firing. In this study, we demonstrate that the LGMD neuron exhibits a large M current, generated by noninactivating K⁺ channels, that shortens the temporal window of dendritic integration, regulates a firing mode switch between burst and isolated spiking, increases the precision of spike timing, and increases the reliability of spike propagation to downstream motor centers. By revealing how the M current increases the LGMD's ability to detect impending collisions, our results suggest that similar channels may play an analogous role in other collision detection circuits. NEW & NOTEWORTHY The ability to reliably detect impending collisions is a critical survival skill. The nervous systems of many animals have developed dedicated neurons for accomplishing this task. We used a mix of in vivo electrophysiology and computational modeling to investigate the role of M potassium channels within one such collision-detecting neuron and show that through regulation of burst firing and enhancement of spiking reliability, the M current increases the ability to detect impending collisions.

Feedforward Inhibition Conveys Time-Varying Stimulus Information in a Collision Detection Circuit

Feedforward inhibition is ubiquitous as a motif in the organization of neuronal circuits. During sensory information processing, it is traditionally thought to sharpen the responses and temporal tuning of feedforward excitation onto principal neurons. As it often exhibits complex time-varying activation properties, feedforward inhibition could also convey information used by single neurons to implement dendritic computations on sensory stimulus variables. We investigated this possibility in a collision-detecting neuron of the locust optic lobe that receives both feedforward excitation and inhibition. We identified a small population of neurons mediating feedforward inhibition, with wide visual receptive fields and whose responses depend both on the size and speed of moving stimuli. By studying responses to simulated objects approaching on a collision course, we determined that they jointly encode the angular size of expansion of the stimulus. Feedforward excitation, on the other hand, encodes a function of the angular velocity of expansion and the targeted collision-detecting neuron combines these two variables non-linearly in its firing output. Thus, feedforward inhibition actively contributes to the detailed firing-rate time course of this collision-detecting neuron, a feature critical to the appropriate execution of escape behaviors. These results suggest that feedforward inhibition could similarly convey time-varying stimulus information in other neuronal circuits.

Burst Firing in a Motion-Sensitive Neural Pathway Correlates with Expansion Properties of Looming Objects that Evoke Avoidance Behaviors

The locust visual system contains a well-defined motion-sensitive pathway that transfers visual input to motor centers involved in predator evasion and collision avoidance. One interneuron in this pathway, the descending contralateral movement detector (DCMD), is typically described as using rate coding; edge expansion of approaching objects causes an increased rate of neuronal firing that peaks after a certain retinal threshold angle is exceeded. However, evidence of intrinsic DCMD bursting properties combined with observable oscillations in mean firing rates and tight clustering of spikes in raw traces, suggest that bursting may be important for motion detection. Sensory neuron bursting provides important timing information about dynamic stimuli in many model systems, yet no studies have rigorously investigated if bursting occurs in the locust DCMD during object approach. We presented repetitions of 30 looming stimuli known to generate behavioral responses to each of 20 locusts in order to identify and quantify putative bursting activity in the DCMD. Overall, we found a bimodal distribution of inter-spike intervals (ISI) with peaks of more frequent and shorter ISIs occurring from 1–8 ms and longer less frequent ISIs occurring from 40–50 ms. Subsequent analysis identified bursts and isolated single spikes from the responses. Bursting frequency increased in the latter phase of an approach and peaked at the time of collision, while isolated spiking was predominant during the beginning of stimulus approach. We also found that the majority of inter-burst intervals (IBIs) occurred at 40–50 ms (or 20–25 bursts/s). Bursting also occurred across varied stimulus parameters and suggests that burst timing may be a key component of looming detection. Our findings suggest that the DCMD uses two modes of coding to transmit information about looming stimuli and that these modes change dynamically with a changing stimulus at a behaviorally-relevant time.

Dendritic Pooling of Noisy Threshold Processes Can Explain Many Properties of a Collision-Sensitive Visual Neuron

Power laws describe brain functions at many levels (from biophysics to psychophysics). It is therefore possible that they are generated by similar underlying mechanisms. Previously, the response properties of a collision-sensitive neuron were reproduced by a model which used a power law for scaling its inhibitory input. A common characteristic of such neurons is that they integrate information across a large part of the visual field. Here we present a biophysically plausible model of collision-sensitive neurons with η-like response properties, in which we assume that each information channel is noisy and has a response threshold. Then, an approximative power law is obtained as a result of pooling these channels. We show that with this mechanism one can successfully predict many response characteristics of the Lobula Giant Movement Detector Neuron (LGMD). Moreover, the results depend critically on noise in the inhibitory pathway, but they are fairly robust against noise in the excitatory pathway.

Many different animals (from insects to primates) try to escape from collision threats, because it is very possible that the approaching object is a predator. The corresponding neurons in the various nervous systems must therefore detect such threats and signal when it is time to escape. Surprisingly, the neurons of different animals which selectively respond to approaching objects have very similar properties. It is therefore worthwhile to understand their underlying computational principles. A common characteristic of such neurons is that they receive (or integrate) information from the whole visual field. The integration process is carried out by the dendritic tree of the neuron. Here we present a computational model in which we assume that each of the input signals is contaminated by noise, as well as having a response threshold (which has to be crossed in order to evoke a response). Then, dendritic integration approximates a mathematical function (a power law) which is essential in our model for explaining the response characteristics of collision-sensitive neurons. Thus, noise is used in a constructive way for computing collision-sensitive responses. Power laws are furthermore found in many different contexts, and may consequently hint at the presence of noise and thresholds.

Representation of Naturalistic Image Structure in the Primate Visual Cortex

The perception of complex visual patterns emerges from neuronal activity in a cascade of areas in the primate cerebral cortex. We have probed the early stages of this cascade with "naturalistic" texture stimuli designed to capture key statistical features of natural images. Humans can recognize and classify these synthetic images and are insensitive to distortions that do not alter the local values of these statistics. The responses of neurons in the primary visual cortex, V1, are relatively insensitive to the statistical information in these textures. However, in the area immediately downstream, V2, cells respond more vigorously to these stimuli than to matched control stimuli. Humans show blood-oxygen-level-dependent functional magnetic resonance imaging (BOLD fMRI responses in V1 and V2) that are consistent with the neuronal measurements in macaque. These fMRI measurements, as well as neurophysiological work by others, show that true natural scenes become a more prominent driving feature of cortex downstream from V2. These results suggest a framework for thinking about how information about elementary visual features is transformed into the specific representations of scenes and objects found in areas higher in the visual pathway.

Synaptic connections of first-stage visual neurons in the locust Schistocerca gregaria extend evolution of tetrad synapses back 200 million years

The small size of some insects, and the crystalline regularity of their eyes, have made them ideal for large-scale reconstructions of visual circuits. In phylogenetically recent muscomorph flies, like Drosophila, precisely coordinated output to different motion-processing pathways is delivered by photoreceptors (R cells), targeting four different postsynaptic cells at each synapse (tetrad). Tetrads were linked to the evolution of aerial agility. To reconstruct circuits for vision in the larger brain of a locust, a phylogenetically old, flying insect, we adapted serial block-face scanning electron microscopy (SBEM). Locust lamina monopolar cells, L1 and L2, were the main targets of the R cell pathway, L1 and L2 each fed a different circuit, only L1 providing feedback onto R cells. Unexpectedly, 40% of all locust R cell synapses onto both L1 and L2 were tetrads, revealing the emergence of tetrads in an arthropod group present 200 million years before muscomorph flies appeared, coinciding with the early evolution of flight.

Impact and sources of neuronal variability in the fly's motion vision pathway

Nervous systems encode information about dynamically changing sensory input by changes in neuronal activity. Neuronal activity changes, however, also arise from noise sources within and outside the nervous system or from changes of the animal's behavioral state. The resulting variability of neuronal responses in representing sensory stimuli limits the reliability with which animals can respond to stimuli and may thus even affect the chances for survival in certain situations. Relevant sources of noise arising at different stages along the motion vision pathway have been investigated from the sensory input to the initiation of behavioral reactions. Here, we concentrate on the reliability of processing visual motion information in flies. Flies rely on visual motion information to guide their locomotion. They are among the best established model systems for the processing of visual motion information allowing us to bridge the gap between behavioral performance and underlying neuronal computations. It has been possible to directly assess the consequences of noise at major stages of the fly's visual motion processing system on the reliability of neuronal signals. Responses of motion sensitive neurons and their variability have been related to optomotor movements as indicators for the overall performance of visual motion computation. We address whether and how noise already inherent in the stimulus, e.g. photon noise for the visual system, influences later processing stages and to what extent variability at the output level of the sensory system limits behavioral performance. Recent advances in circuit analysis and the progress in monitoring neuronal activity in behaving animals should now be applied to understand how the animal meets the requirements of fast and reliable manoeuvres in naturalistic situations.