Neurons Process Information Research Articles

Olfactory system structure and function in newly hatched and adult locusts

An important question in neuroscience is how sensory systems change as animals grow and interact with the environment. Exploring sensory systems in animals as they develop can reveal how networks of neurons process information as the neurons themselves grow and the needs of the animal change. Here we compared the structure and function of peripheral parts of the olfactory pathway in newly hatched and adult locusts. We found that populations of olfactory sensory neurons (OSNs) in hatchlings and adults responded with similar tunings to a panel of odors. The morphologies of local neurons (LNs) and projection neurons (PNs) in the antennal lobes (ALs) were very similar in both age groups, though they were smaller in hatchlings, they were proportional to overall brain size. The odor evoked responses of LNs and PNs were also very similar in both age groups, characterized by complex patterns of activity including oscillatory synchronization. Notably, in hatchlings, spontaneous and odor-evoked firing rates of PNs were lower, and LFP oscillations were lower in frequency, than in the adult. Hatchlings have smaller antennae with fewer OSNs; removing antennal segments from adults also reduced LFP oscillation frequency. Thus, consistent with earlier computational models, the developmental increase in frequency is due to increasing intensity of input to the oscillation circuitry. Overall, our results show that locusts hatch with a fully formed olfactory system that structurally and functionally matches that of the adult, despite its small size and lack of prior experience with olfactory stimuli.

Model-Based Explainable Deep Learning for Light-Field Microscopy Imaging.

In modern neuroscience, observing the dynamics of large populations of neurons is a critical step of understanding how networks of neurons process information. Light-field microscopy (LFM) has emerged as a type of scanless, high-speed, three-dimensional (3D) imaging tool, particularly attractive for this purpose. Imaging neuronal activity using LFM calls for the development of novel computational approaches that fully exploit domain knowledge embedded in physics and optics models, as well as enabling high interpretability and transparency. To this end, we propose a model-based explainable deep learning approach for LFM. Different from purely data-driven methods, the proposed approach integrates wave-optics theory, sparse representation and non-linear optimization with the artificial neural network. In particular, the architecture of the proposed neural network is designed following precise signal and optimization models. Moreover, the network's parameters are learned from a training dataset using a novel training strategy that integrates layer-wise training with tailored knowledge distillation. Such design allows the network to take advantage of domain knowledge and learned new features. It combines the benefit of both model-based and learning-based methods, thereby contributing to superior interpretability, transparency and performance. By evaluating on both structural and functional LFM data obtained from scattering mammalian brain tissues, we demonstrate the capabilities of the proposed approach to achieve fast, robust 3D localization of neuron sources and accurate neural activity identification.

Thermotaxis in an apolar, non-neuronal animal.

Neuronal circuits are hallmarks of complex decision-making processes in the animal world. How animals without neurons process information and respond to environmental cues promises a new window into studying precursors of neuronal control and origin of the nervous system as we know it today. Robust decision making in animals, such as in chemotaxis or thermotaxis, often requires internal symmetry breaking (such as anterior-posterior (AP) axis) provided naturally by a given body plan of an animal. Here we report the discovery of robust thermotaxis behaviour in Trichoplax adhaerens, an early-divergent, enigmatic animal with no anterior-posterior symmetry breaking (apolar) and no known neurons or muscles. We present a quantitative and robust behavioural response assay in Placozoa, which presents an apolar flat geometry. By exposing T. adhaerens to a thermal gradient under a long-term imaging set-up, we observe robust thermotaxis that occurs over timescale of hours, independent of any circadian rhythms. We quantify that T. adhaerens can detect thermal gradients of at least 0.1°C cm-1. Positive thermotaxis is observed for a range of baseline temperatures from 17°C to 22.5°C, and distributions of momentary speeds for both thermotaxis and control conditions are well described by single exponential fits. Interestingly, the organism does not maintain a fixed orientation while performing thermotaxis. Using natural diversity in size of adult organisms (100 µm to a few millimetres), we find no apparent size-dependence in thermotaxis behaviour across an order of magnitude of organism size. Several transient receptor potential (TRP) family homologues have been previously reported to be conserved in metazoans, including in T. adhaerens. We discover naringenin, a known TRPM3 antagonist, inhibits thermotaxis in T. adhaerens. The discovery of robust thermotaxis in T. adhaerens provides a tractable handle to interrogate information processing in a brainless animal. Understanding how divergent marine animals process thermal cues is also critical due to rapid temperature rise in our oceans.

Light-Field Microscopy for Optical Imaging of Neuronal Activity: When Model-Based Methods Meet Data-Driven Approaches.

Understanding how networks of neurons process information is one of the key challenges in modern neuroscience. A necessary step to achieve this goal is to be able to observe the dynamics of large populations of neurons over a large area of the brain. Light-field microscopy (LFM), a type of scanless microscope, is a particularly attractive candidate for high-speed three-dimensional (3D) imaging. It captures volumetric information in a single snapshot, allowing volumetric imaging at video frame-rates. Specific features of imaging neuronal activity using LFM call for the development of novel machine learning approaches that fully exploit priors embedded in physics and optics models. Signal processing theory and wave-optics theory could play a key role in filling this gap, and contribute to novel computational methods with enhanced interpretability and generalization by integrating model-driven and data-driven approaches. This paper is devoted to a comprehensive survey to state-of-the-art of computational methods for LFM, with a focus on model-based and data-driven approaches.

Neural Circuit Mechanism Underlying the Feeding Controlled by Insula-Central Amygdala Pathway.

SummaryThe Central nucleus of amygdala (CeA) contains distinct populations of neurons that play opposing roles in feeding. The circuit mechanism of how CeA neurons process information sent from their upstream inputs to regulate feeding is still unclear. Here we show that activation of the neural pathway projecting from insular cortex neurons to the CeA suppresses food intake. Surprisingly, we find that the inputs from insular cortex form excitatory connections with similar strength to all types of CeA neurons. To reconcile this puzzling result, and previous findings, we developed a conductance-based dynamical systems model for the CeA neuronal network. Computer simulations showed that both the intrinsic electrophysiological properties of individual CeA neurons and the overall synaptic organization of the CeA circuit play a functionally significant role in shaping CeA neural dynamics. We successfully identified a specific CeA circuit structure that reproduces the desired circuit output consistent with existing experimentally observed feeding behaviors.

Universality of SNQ P Systems Using One Type of Spikes and Restrictive Rule Application

We investigate the spiking neural P systems with communication on request (SNQ P systems) that are devices in the area of neural like P systems abstracting the way in which neurons work and process information. Here we discuss the SNQ P systems using the rule application strategy as defined by Linqiang Pan and collaborators and we are able to improve their result of universality of such systems using two types of spikes. In the current work, we prove that only one type of spikes is sufficient for reaching the computational power of Turing Machines for these devices, bringing closer to implementation such a device. The result holds both in maximum parallel manner application of the rules as well as the maximum-sequentiality application of rules.

Information transmission of mean and variance coding in integrate-and-fire neurons.

Neurons process information by translating continuous signals into patterns of discrete spike times. An open question is how much information these spike times contain about signals which modulate either the mean or the variance of the somatic currents in neurons, as is observed experimentally. Here we calculate the exact information contained in discrete spike times about a continuous signal in both encoding strategies. We show that the information content about mean modulating signals is generally substantially larger than about variance modulating signals for biological parameters. Our analysis further reveals that higher information transmission is associated with a larger proportion of nonlinear signal encoding. Our study measures the complete information content of mean and variance coding and provides a method to determine what fraction of the total information is linearly decodable.

Gain control with A-type potassium current: IA as a switch between divisive and subtractive inhibition.

Neurons process and convey information by transforming barrages of synaptic inputs into spiking activity. Synaptic inhibition typically suppresses the output firing activity of a neuron, and is commonly classified as having a subtractive or divisive effect on a neuron’s output firing activity. Subtractive inhibition can narrow the range of inputs that evoke spiking activity by eliminating responses to non-preferred inputs. Divisive inhibition is a form of gain control: it modifies firing rates while preserving the range of inputs that evoke firing activity. Since these two “modes” of inhibition have distinct impacts on neural coding, it is important to understand the biophysical mechanisms that distinguish these response profiles. In this study, we use simulations and mathematical analysis of a neuron model to find the specific conditions (parameter sets) for which inhibitory inputs have subtractive or divisive effects. Significantly, we identify a novel role for the A-type Potassium current (IA). In our model, this fast-activating, slowly-inactivating outward current acts as a switch between subtractive and divisive inhibition. In particular, if IA is strong (large maximal conductance) and fast (activates on a time-scale similar to spike initiation), then inhibition has a subtractive effect on neural firing. In contrast, if IA is weak or insufficiently fast-activating, then inhibition has a divisive effect on neural firing. We explain these findings using dynamical systems methods (plane analysis and fast-slow dissection) to define how a spike threshold condition depends on synaptic inputs and IA. Our findings suggest that neurons can “self-regulate” the gain control effects of inhibition via combinations of synaptic plasticity and/or modulation of the conductance and kinetics of A-type Potassium channels. This novel role for IA would add flexibility to neurons and networks, and may relate to recent observations of divisive inhibitory effects on neurons in the nucleus of the solitary tract.

Layer-Dependent Short-Term Synaptic Plasticity Between Excitatory Neurons in the C2 Barrel Column of Mouse Primary Somatosensory Cortex.

Neurons process information through spatiotemporal integration of synaptic input. Synaptic transmission between any given pair of neurons is typically a dynamic process with presynaptic action potentials(APs) evoking depressing or facilitating postsynaptic potentials when presynaptic APs occur within hundreds of milliseconds of each other. In order to understand neocortical function, it is therefore important to investigate such short-term synaptic plasticity at synapses between different types of neocortical neurons. Here, we examine short-term synaptic dynamics between excitatory neurons in different layers of the mouse C2 barrel column through in vitro whole-cell recordings. We find layer-dependent short-term plasticity, with depression being dominant at many synaptic connections. Interestingly, however, presynaptic layer 2 neurons predominantly give rise to facilitating excitatory synaptic output at short interspike intervals of 10 and 30 ms. Previous studies have found prominent burst firing of excitatory neurons in supragranular layers of awake mice. The facilitation we observed in the synaptic output of layer 2 may, therefore, be functionally relevant, possibly serving to enhance the postsynaptic impact of burst firing.

Bilinearity in spatiotemporal integration of synaptic inputs.

Neurons process information via integration of synaptic inputs from dendrites. Many experimental results demonstrate dendritic integration could be highly nonlinear, yet few theoretical analyses have been performed to obtain a precise quantitative characterization analytically. Based on asymptotic analysis of a two-compartment passive cable model, given a pair of time-dependent synaptic conductance inputs, we derive a bilinear spatiotemporal dendritic integration rule. The summed somatic potential can be well approximated by the linear summation of the two postsynaptic potentials elicited separately, plus a third additional bilinear term proportional to their product with a proportionality coefficient . The rule is valid for a pair of synaptic inputs of all types, including excitation-inhibition, excitation-excitation, and inhibition-inhibition. In addition, the rule is valid during the whole dendritic integration process for a pair of synaptic inputs with arbitrary input time differences and input locations. The coefficient is demonstrated to be nearly independent of the input strengths but is dependent on input times and input locations. This rule is then verified through simulation of a realistic pyramidal neuron model and in electrophysiological experiments of rat hippocampal CA1 neurons. The rule is further generalized to describe the spatiotemporal dendritic integration of multiple excitatory and inhibitory synaptic inputs. The integration of multiple inputs can be decomposed into the sum of all possible pairwise integration, where each paired integration obeys the bilinear rule. This decomposition leads to a graph representation of dendritic integration, which can be viewed as functionally sparse.

The Chronotron: A Neuron That Learns to Fire Temporally Precise Spike Patterns

In many cases, neurons process information carried by the precise timings of spikes. Here we show how neurons can learn to generate specific temporally precise output spikes in response to input patterns of spikes having precise timings, thus processing and memorizing information that is entirely temporally coded, both as input and as output. We introduce two new supervised learning rules for spiking neurons with temporal coding of information (chronotrons), one that provides high memory capacity (E-learning), and one that has a higher biological plausibility (I-learning). With I-learning, the neuron learns to fire the target spike trains through synaptic changes that are proportional to the synaptic currents at the timings of real and target output spikes. We study these learning rules in computer simulations where we train integrate-and-fire neurons. Both learning rules allow neurons to fire at the desired timings, with sub-millisecond precision. We show how chronotrons can learn to classify their inputs, by firing identical, temporally precise spike trains for different inputs belonging to the same class. When the input is noisy, the classification also leads to noise reduction. We compute lower bounds for the memory capacity of chronotrons and explore the influence of various parameters on chronotrons' performance. The chronotrons can model neurons that encode information in the time of the first spike relative to the onset of salient stimuli or neurons in oscillatory networks that encode information in the phases of spikes relative to the background oscillation. Our results show that firing one spike per cycle optimizes memory capacity in neurons encoding information in the phase of firing relative to a background rhythm.

Closed-Loop Measurements of Iso-Response Stimuli Reveal Dynamic Nonlinear Stimulus Integration in the Retina

Neurons often integrate information from multiple parallel signaling streams. How a neuron combines these inputs largely determines its computational role in signal processing. Experimental assessment of neuronal signal integration, however, is often confounded by cell-intrinsic nonlinear processes that arise after signal integration has taken place. To overcome this problem and determine how ganglion cells in the salamander retina integrate visual contrast over space, we used automated online analysis of recorded spike trains and closed-loop control of the visual stimuli to identify different stimulus patterns that give the same neuronal response. These iso-response stimuli revealed a threshold-quadratic transformation as a fundamental nonlinearity within the receptive field center. Moreover, for a subset of ganglion cells, the method revealed an additional dynamic nonlinearity that renders these cells particularly sensitive to spatially homogeneous stimuli. This function is shown to arise from a local inhibition-mediated dynamic gain control mechanism.

The chronotron: a neuron that learns to fire temporally-precise spike patterns

AbstractIn many cases, neurons process information carried by the precise timing of spikes. Here we show how neurons can learn to generate specific temporally-precise output spikes in response to input spike patterns, thus processing and memorizing information that is fully temporally coded, both as input and as output. We introduce two new supervised learning rules for spiking neurons with temporal coding of information (chronotrons), one that is analytically-derived and highly efficient, and one that has a high degree of biological plausibility. We show how chronotrons can learn to classify their inputs and we study their memory capacity.

Properties of the Force Exerted by Filopodia and Lamellipodia and the Involvement of Cytoskeletal Components

During neuronal differentiation, lamellipodia and filopodia explore the environment in search for the correct path to the axon's final destination. Although the motion of lamellipodia and filopodia has been characterized to an extent, little is known about the force they exert. In this study, we used optical tweezers to measure the force exerted by filopodia and lamellipodia with a millisecond temporal resolution. We found that a single filopodium exerts a force not exceeding 3 pN, whereas lamellipodia can exert a force up to 20 pN. Using metabolic inhibitors, we showed that no force is produced in the absence of actin polymerization and that development of forces larger than 3 pN requires microtubule polymerization. These results show that actin polymerization is necessary for force production and demonstrate that not only do neurons process information, but they also act on their environment exerting forces varying from tenths pN to tens of pN.

Shape of a Common Protein Module Suggests Role as Molecular Switch

A vital aspect of a neuron's job is deciding when to pass their cache of chemicals on to neighboring cells. To do this in a way that ensures effective communication, neurons must keep tight reins on their neurotransmitters—the chemical messengers they release to influence neighboring cells. Neurons quickly collect and then jettison these neurotransmitters, cycling through this process many times per second. Now a new study reveals the structure of a fragment of Munc-13—a key protein in this process—showing how it could act as a switch that neurons use to toggle quickly between storing and releasing their neurotransmitters. For this switching role, the key part of the protein is one of its C 2 domains (the C 2A domain). This domain is widespread, suggesting that other proteins may function by a similar mechanism. Neurons store their neurotransmitters in vesicles, tiny sacs that amass at the synapse between cells. There, they dock at the inside of the cell membrane and undergo one or more reactions that “prime” the vesicles to be ready to quickly unload their neurotransmitters outside the cell when the neuron fires. A cascade of protein interactions mediates these different steps. These protein cascades not only allow neurons to do their usual duty, but they also help neurons process information. If neurons react quickly or sluggishly, reliably or not, to a neuron firing, this affects the signal that gets passed on to neighboring neurons. And, if their reactions change with time, this plasticity can help the animal remember, learn, and adapt. To understand how the proteins interact, a key step is figuring out the shapes the proteins assume. Once they have the shapes, researchers can see how proteins, like pieces of a jigsaw puzzle, fit together. For example, this helps researchers make sense of chemical and biological data that suggest how proteins bind together and influence each other. Josep Rizo and colleagues targeted a protein called Munc13-1 since its family of proteins plays important roles in preparing vesicles to release neurotransmitters, and Munc13-1 itself is known to have a role in the plasticity of synapses. Rizo and colleagues homed in on one of the key zones of Munc13-1, called the C 2A domain. C 2 domains are well known for being able to bind to phospholipids, the fatty acids that make up cell and vesicle membranes. But the researchers found an unexpected role for the Munc13-1 C 2A domain: it can bind to other proteins as well. To figure out the shape of Munc13-1's C 2A domain, down to the atomic level, the researchers combined two methods: X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy. X-ray crystallography is the classic method for “solving” the mystery of a protein's shape. But not all proteins crystallize well, and so are not amenable to the method. Or if they do crystallize, their shape may be distorted from the form it normally takes when floating in a cell's watery interior. Rizo and colleagues combined the two methods to create a detailed picture of the protein's shape. They used NMR spectroscopy to check how well the protein behaves in solution and to optimize the crystallization, aiding the X-ray studies. Also, data from the NMR spectroscopy on the shape of the protein helped them interpret the X-ray results. In the end, they were able to figure out the natural shape of the protein's C 2A domain to atomic resolution. These structural data also helped Rizo and colleagues make sense of their experiments designed to determine Munc13-1's binding partners. The researchers found that, in a pure solution, Munc13-1 proteins bound to each other in pairs. But when mixed with RIM2α (another protein that plays a key role in neurotransmitter release and neuron plasticity), then Munc13-1 formed pairs with RIM2α instead. Combining the structural and binding data, the researchers were able to piece together a picture of how Munc13-1 could work as a switch. They found that the C 2 domain of Munc13-1 has two adjacent and partially overlapping regions that can bind to other proteins. So two C 2 domains could bind each other using one part of that domain, and another part of the domain could bind to RIM2α. But the C 2 domain can't bind both at the same time. The researchers could see, once they had the shapes of the proteins, that they simply can't all fit together at once. Thus it seems that there is competition for Munc13-1 to bind either to itself or to other proteins. This mechanism could prepare the vesicles to be ready to unload the neurotransmitters when the neuron fires. Since the C 2 domain is widespread among proteins—and is highly similar among distantly related species—the findings could hold clues to understanding more roles for this common sequence of amino acids.

Neural representation of probabilistic information.

It has been proposed that populations of neurons process information in terms of probability density functions (PDFs) of analog variables. Such analog variables range, for example, from target luminance and depth on the sensory interface to eye position and joint angles on the motor output side. The requirement that analog variables must be processed leads inevitably to a probabilistic description, while the limited precision and lifetime of the neuronal processing units lead naturally to a population representation of information. We show how a time-dependent probability density rho(x; t) over variable x, residing in a specified function space of dimension D, may be decoded from the neuronal activities in a population as a linear combination of certain decoding functions phi(i)(x), with coefficients given by the N firing rates a(i)(t) (generally with D << N). We show how the neuronal encoding process may be described by projecting a set of complementary encoding functions phi;(i)(x) on the probability density rho(x; t), and passing the result through a rectifying nonlinear activation function. We show how both encoders phi;(i)(x) and decoders phi(i)(x) may be determined by minimizing cost functions that quantify the inaccuracy of the representation. Expressing a given computation in terms of manipulation and transformation of probabilities, we show how this representation leads to a neural circuit that can carry out the required computation within a consistent Bayesian framework, with the synaptic weights being explicitly generated in terms of encoders, decoders, conditional probabilities, and priors.

Gain modulation and frequency locking under conductance noise

In the intact cortical network, neurons process information while receiving thousands of synaptic inputs. We investigated how this noise modulates the ability to resolve constant and oscillatory stimulation by using simplified neuron models. The cellular response was precise and reliable for a broad range of model parameters, while the output rate depended on the stimulus amplitude. The results suggest that both temporal and rate coding schemes are not exclusive but rather encode different aspects of the input at the same time. Conductance noise was found to play a modulating role, suggesting that high-conductance noise confers advantageous computational properties to neurons.

The effects of temperature on vesicular supply and release in autaptic cultures of rat and mouse hippocampal neurons.

Membrane fusion plays a central role in the synaptic vesicle cycle. While many of the pre- and postfusion events have been investigated at room temperature, few researchers have investigated these processes at more physiologically relevant temperatures. We have used autaptic cultures of hippocampal neurons to investigate changes in the size and refilling rate of the readily releasable pool (RRP) of synaptic vesicles brought about by an increase in temperature from 25 to 35 degrees C. We have also examined temperature-dependent changes in spontaneous and action potential (AP)-evoked release as well as the fraction of the RRP that is released during an AP. Although we found a threefold increase in the refilling rate of the RRP at the higher temperature, there was no apparent change in the size of the RRP with increased temperature. Moreover, we observed a slight but significant decrease in the quanta released during an AP. This increased refilling rate and decreased release probability resulted in a reduction of both the degree and time course of synaptic depression during high frequency stimulation at the higher temperature. This reduction in synaptic depression was accompanied by an increased maintenance of the synchronous component of release during high frequency stimulation. These findings indicate that the dynamics of vesicular supply and release in hippocampal neurons at room temperature are significantly different at near physiological temperatures and could affect our present understanding of the way in which individual neurons and networks of neurons process information.

Theta-Frequency Bursting and Resonance in Cerebellar Granule Cells: Experimental Evidence and Modeling of a Slow K+-Dependent Mechanism

Neurons process information in a highly nonlinear manner, generating oscillations, bursting, and resonance, enhancing responsiveness at preferential frequencies. It has been proposed that slow repolarizing currents could be responsible for both oscillation/burst termination and for high-pass filtering that causes resonance (Hutcheon and Yarom, 2000). However, different mechanisms, including electrotonic effects (Mainen and Sejinowski, 1996), the expression of resurgent currents (Raman and Bean, 1997), and network feedback, may also be important. In this study we report theta-frequency (3-12 Hz) bursting and resonance in rat cerebellar granule cells and show that these neurons express a previously unidentified slow repolarizing K(+) current (I(K-slow)). Our experimental and modeling results indicate that I(K-slow) was necessary for both bursting and resonance. A persistent (and potentially a resurgent) Na(+) current exerted complex amplifying actions on bursting and resonance, whereas electrotonic effects were excluded by the compact structure of the granule cell. Theta-frequency bursting and resonance in granule cells may play an important role in determining synchronization, rhythmicity, and learning in the cerebellum.

Motor learning in a recurrent network model based on the vestibulo-ocular reflex.

Most models of neural networks have assumed that neurons process information on a timescale of milliseconds and that the long-term modification of synaptic strengths underlies learning and memory. But neurons also have cellular mechanisms that operate on a timescale of tens or hundreds of milliseconds, such as a gradual rise in firing rate in response to injection of constant current or a rapid rise followed by a slower adaptation. These dynamic properties of neuronal responses are mediated by ion channels that are subject to modulation. We demonstrate here how a neural network with recurrent feedback connections can convert long-term modulation of neural responses that occur over these intermediate timescales into changes in the amplitude of the steady output from the system. This general principle may be relevant to many feedback systems in the brain. Here it is applied to the vestibulo-ocular reflex, whose amplitude is subject to long-term adaptive modification by visual inputs. The model reconciles apparently contradictory data on the neural locus of the cellular mechanisms that mediate this simple form of learning and memory.