Dendritic computation of direction selectivity by retinal ganglion cells

ON and OFF starburst amacrine cells are controlled by distinct cholinergic pathways

Cholinergic signaling in the retina is mediated by acetylcholine (ACh) released from starburst amacrine cells (SACs), which are key neurons for motion detection. SACs comprise ON and OFF subtypes, which morphologically show mirror symmetry to each other. Although many physiological studies on SACs have targeted ON cells only, the synaptic computation of ON and OFF SACs is assumed to be similar. Recent studies demonstrated that gene expression patterns and receptor types differed between ON and OFF SACs, suggesting differences in their functions. Here, we compared cholinergic signaling pathways between ON and OFF SACs in the mouse retina using the patch clamp technique. The application of ACh increased GABAergic feedback, observed as postsynaptic currents to SACs, in both ON and OFF SACs; however, the mode of GABAergic feedback differed. Nicotinic receptors mediated GABAergic feedback in both ON and OFF SACs, while muscarinic receptors mediated GABAergic feedback in ON SACs only in adults. Neither tetrodotoxin, which blocked action potentials, nor LY354740, which blocked neurotransmitter release from SACs, eliminated ACh-induced GABAergic feedback in SACs. These results suggest that ACh-induced GABAergic feedback in ON and OFF SACs is regulated by different feedback mechanisms in adults and mediated by non-spiking amacrine cells other than SACs.

Direct comparison reveals algorithmic similarities in fly and mouse visual motion detection

Evolution has equipped vertebrates and invertebrates with neural circuits that selectively encode visual motion. While similarities in the computations performed by these circuits in mouse and fruit fly have been noted, direct experimental comparisons have been lacking. Because molecular mechanisms and neuronal morphology in the two species are distinct, we directly compared motion encoding in these two species at the algorithmic level, using matched stimuli and focusing on a pair of analogous neurons, the mouse ON starburst amacrine cell (ON SAC) and Drosophila T4 neurons. We find that the cells share similar spatiotemporal receptive field structures, sensitivity to spatiotemporal correlations, and tuning to sinusoidal drifting gratings, but differ in their responses to apparent motion stimuli. Both neuron types showed a response to summed sinusoids that deviates from models for motion processing in these cells, underscoring the similarities in their processing and identifying response features that remain to be explained.

Two Sides of the Same Coin: Efficient and Predictive Neural Coding

Some visual properties are consistent across a wide range of environments, while other properties are more labile. The efficient coding hypothesis states that many of these regularities in the environment can be discarded from neural representations, thus allocating more of the brain's dynamic range to properties that are likely to vary. This paradigm is less clear about how the visual system prioritizes different pieces of information that vary across visual environments. One solution is to prioritize information that can be used to predict future events, particularly those that guide behavior. The relationship between the efficient coding and future prediction paradigms is an area of active investigation. In this review, we argue that these paradigms are complementary and often act on distinct components of the visual input. We also discuss how normative approaches to efficient coding and future prediction can be integrated.

Single dendritic neural classification with an effective spherical search-based whale learning algorithm

McCulloch-Pitts neuron-based neural networks have been the mainstream deep learning methods, achieving breakthrough in various real-world applications. However, McCulloch-Pitts neuron is also under longtime criticism of being overly simplistic. To alleviate this issue, the dendritic neuron model (DNM), which employs non-linear information processing capabilities of dendrites, has been widely used for prediction and classification tasks. In this study, we innovatively propose a hybrid approach to co-evolve DNM in contrast to back propagation (BP) techniques, which are sensitive to initial circumstances and readily fall into local minima. The whale optimization algorithm is improved by spherical search learning to perform co-evolution through dynamic hybridizing. Eleven classification datasets were selected from the well-known UCI Machine Learning Repository. Its efficiency in our model was verified by statistical analysis of convergence speed and Wilcoxon sign-rank tests, with receiver operating characteristic curves and the calculation of area under the curve. In terms of classification accuracy, the proposed co-evolution method beats 10 existing cutting-edge non-BP methods and BP, suggesting that well-learned DNMs are computationally significantly more potent than conventional McCulloch-Pitts types and can be employed as the building blocks for the next-generation deep learning methods.

Artificial Visual System for Orientation Detection

The human visual system is one of the most important components of the nervous system, responsible for visual perception. The research on orientation detection, in which neurons of the visual cortex respond only to a line stimulus in a particular orientation, is an important driving force of computer vision and biological vision. However, the principle underlying orientation detection remains a mystery. In order to solve this mystery, we first propose a completely new mechanism that explains planar orientation detection in a quantitative manner. First, we assume that there are planar orientation-detective neurons which respond only to a particular planar orientation locally and that these neurons detect local planar orientation information based on nonlinear interactions that take place on the dendrites. Then, we propose an implementation of these local planar orientation-detective neurons based on their dendritic computations, use them to extract the local planar orientation information, and infer the global planar orientation information from the local planar orientation information. Furthermore, based on this mechanism, we propose an artificial visual system (AVS) for planar orientation detection and other visual information processing. In order to prove the effectiveness of our mechanism and the AVS, we conducted a series of experiments on rectangular images which included rectangles of various sizes, shapes and positions. Computer simulations show that the mechanism can perfectly perform planar orientation detection regardless of their sizes, shapes and positions in all experiments. Furthermore, we compared the performance of both AVS and a traditional convolution neural network (CNN) on planar orientation detection and found that AVS completely outperformed CNN in planar orientation detection in terms of identification accuracy, noise resistance, computation and learning cost, hardware implementation and reasonability.

Parallel and Recurrent Cascade Models as a Unifying Force for Understanding Sub-cellular Computation

Neurons are very complicated computational devices, incorporating numerous non-linear processes, particularly in their dendrites. Biophysical models capture these processes directly by explicitly modelling physiological variables, such as ion channels, current flow, membrane capacitance, etc. However, another option for capturing the complexities of real neural computation is to use cascade models, which treat individual neurons as a cascade of linear and non-linear operations, akin to a multi-layer artificial neural network. Recent research has shown that cascade models can capture single-cell computation well, but there are still a number of sub-cellular, regenerative dendritic phenomena that they cannot capture, such as the interaction between sodium, calcium, and NMDA spikes in different compartments. Here, we propose that it is possible to capture these additional phenomena using parallel, recurrent cascade models, wherein an individual neuron is modelled as a cascade of parallel linear and non-linear operations that can be connected recurrently, akin to a multi-layer, recurrent, artificial neural network. Given their tractable mathematical structure, we show that neuron models expressed in terms of parallel recurrent cascades can themselves be integrated into multi-layered artificial neural networks and trained to perform complex tasks. We go on to discuss potential implications and uses of these models for artificial intelligence. Overall, we argue that parallel, recurrent cascade models provide an important, unifying tool for capturing single-cell computation and exploring the algorithmic implications of physiological phenomena.

Mechanism of Motion Direction Detection Based on Barlow’s Retina Inhibitory Scheme in Direction-Selective Ganglion Cells

Previous studies have reported that directionally selective ganglion cells respond strongly in their preferred direction, but are only weakly excited by stimuli moving in the opposite null direction. Various studies have attempted to elucidate the mechanisms underlying direction selectivity with cellular basis. However, these studies have not elucidated the mechanism underlying motion direction detection. In this study, we propose the mechanism based on Barlow’s inhibitory scheme for motion direction detection. We described the local motion-sensing direction-selective neurons. Next, this model was used to construct the two-dimensional multi-directional detection neurons which detect the local motion directions. The information of local motion directions was finally used to infer the global motion direction. To verify the validity of the proposed mechanism, we conducted a series of experiments involving a dataset with a number of images. The proposed mechanism exhibited good performance in all experiments with high detection accuracy. Furthermore, we compare the performance of our proposed system and traditional Convolution Neural Network (CNN) on motion direction prediction. It is found that the performance of our system is much better than that of CNN in terms of accuracy, calculation speed and cost.

A Simple Dendritic Neural Network Model-Based Approach for Daily PM2.5 Concentration Prediction

Air pollution in cities has a massive impact on human health, and an increase in fine particulate matter (PM2.5) concentrations is the main reason for air pollution. Due to the chaotic and intrinsic complexities of PM2.5 concentration time series, it is difficult to utilize traditional approaches to extract useful information from these data. Therefore, a neural model with a dendritic mechanism trained via the states of matter search algorithm (SDNN) is employed to conduct daily PM2.5 concentration forecasting. Primarily, the time delay and embedding dimensions are calculated via the mutual information-based method and false nearest neighbours approach to train the data, respectively. Then, the phase space reconstruction is performed to map the PM2.5 concentration time series into a high-dimensional space based on the obtained time delay and embedding dimensions. Finally, the SDNN is employed to forecast the PM2.5 concentration. The effectiveness of this approach is verified through extensive experimental evaluations, which collect six real-world datasets from recent years. To the best of our knowledge, this study is the first attempt to utilize a dendritic neural model to perform real-world air quality forecasting. The extensive experimental results demonstrate that the SDNN offers very competitive performance relative to the latest prediction techniques.

The functional organization of excitation and inhibition in the dendrites of mouse direction-selective ganglion cells

Recent studies indicate that the precise timing and location of excitation and inhibition (E/I) within active dendritic trees can significantly impact neuronal function. How synaptic inputs are functionally organized at the subcellular level in intact circuits remains unclear. To address this issue, we took advantage of the retinal direction-selective ganglion cell circuit, where directionally tuned inhibition is known to shape non-directional excitatory signals. We combined two-photon calcium imaging with genetic, pharmacological, and single-cell ablation methods to examine the extent to which inhibition ‘vetoes’ excitation at the level of individual dendrites of direction-selective ganglion cells. We demonstrate that inhibition shapes direction selectivity independently within small dendritic segments (<10µm) with remarkable accuracy. The data suggest that the parallel processing schemes proposed for direction encoding could be more fine-grained than previously envisioned.

Topographic Mapping as a Basic Principle of Functional Organization for Visual and Prefrontal Functional Connectivity

The organization of region-to-region functional connectivity has major implications for understanding information transfer and transformation between brain regions. We extended connective field mapping methodology to 3-D anatomic space to derive estimates of corticocortical functional organization. Using multiple publicly available human (both male and female) resting-state fMRI data samples for model testing and replication analysis, we have three main findings. First, we found that the functional connectivity between early visual regions maintained a topographic relationship along the anterior-posterior dimension, which corroborates previous research. Higher order visual regions showed a pattern of connectivity that supports convergence and biased sampling, which has implications for their receptive field properties. Second, we demonstrated that topographic organization is a fundamental aspect of functional connectivity across the entire cortex, with higher topographic connectivity between regions within a functional network than across networks. The principle gradient of topographic connectivity across the cortex resembled whole-brain gradients found in previous work. Last but not least, we showed that the organization of higher order regions such as the lateral prefrontal cortex demonstrate functional gradients of topographic connectivity and convergence. These organizational features of the lateral prefrontal cortex predict task-based activation patterns, particularly visual specialization and higher order rules. In sum, these findings suggest that topographic input is a fundamental motif of functional connectivity between cortical regions for information processing and transfer, with maintenance of topography potentially important for preserving the integrity of information from one region to another.

Classes of dendritic information processing

Dendrites are much more than passive neuronal components. Mounting experimental evidence and decades of computational work have decisively shown that dendrites leverage a host of nonlinear biophysical phenomena and actively participate in sophisticated computations, at the level of the single neuron and at the level of the network. However, a coherent view of their processing power is still lacking and dendrites are largely neglected in neural network models. Here, we describe four classes of dendritic information processing and delineate their implications at the algorithmic level. We propose that beyond the well-known spatiotemporal filtering of their inputs, dendrites are capable of selecting, routing and multiplexing information. By separating dendritic processing from axonal outputs, neuron networks gain a degree of freedom with implications for perception and learning.

Simulated Saccadic Stimuli Suppress ON-Type Direction-Selective Retinal Ganglion Cells via Glycinergic Inhibition

Two types of mammalian direction-selective ganglion cells (DSGCs), ON and ONOFF, operate over different speed ranges. The directional axes of the ON-DSGCs are thought to align with the axes of the vestibular system and provide sensitivity at rotational velocities that are too slow to activate the semicircular canals. ONOFF-DSGCs respond to faster image velocities. Using natural images that simulate the natural visual inputs to freely moving animals, we show that simulated visual saccades suppress responses in ON-DSGCs but not ONOFF-DSGCs recorded in retinas of domestic rabbits of either gender. Analysis of the synaptic inputs shows that this saccadic suppression results from glycinergic inputs that are specific to ON-DSGCs and are absent in ONOFF-DSGCs. When this glycinergic input is blocked, both cell types respond similarly to visual saccades and display essentially identical speed tuning. The results demonstrate that glycinergic circuits within the retina can produce saccadic suppression of retinal ganglion cell activity. The cell-type-specific targeting of the glycinergic circuits further supports the proposed physiological roles of ON-DSGCs in retinal-image stabilization and of ONOFF-DSGCs in detecting local object motion and signaling optical flow.SIGNIFICANCE STATEMENT In the mammalian retina, ON direction-selective ganglion cells (DSGCs) respond preferentially to slow image motion, whereas ONOFF-DSGCs respond better to rapid motion. The mechanisms producing this different speed tuning remain unclear. Here we show that simulated visual saccades suppress ON-DSGCs, but not ONOFF-DSGCs. This selective saccadic suppression is because of the selective targeting of glycinergic inhibitory synaptic inputs to ON-DSGCs. The different saccadic suppression in the two cell types points to different physiological roles, consistent with their projections to distinct areas within the brain. ON-DSGCs may be critical for providing the visual feedback signals that contribute to stabilizing the image on the retina, whereas ONOFF-DSGCs may be important for detecting the onset of saccades or for signaling optical flow.

Retinal direction selectivity in the absence of asymmetric starburst amacrine cell responses

In the mammalian retina, direction-selectivity is thought to originate in the dendrites of GABAergic/cholinergic starburst amacrine cells, where it is first observed. However, here we demonstrate that direction selectivity in downstream ganglion cells remains remarkably unaffected when starburst dendrites are rendered non-directional, using a novel strategy combining a conditional GABA_A α2 receptor knockout mouse with optogenetics. We show that temporal asymmetries between excitation/inhibition, arising from the differential connectivity patterns of starburst cholinergic and GABAergic synapses to ganglion cells, form the basis for a parallel mechanism generating direction selectivity. We further demonstrate that these distinct mechanisms work in a coordinated way to refine direction selectivity as the stimulus crosses the ganglion cell’s receptive field. Thus, precise spatiotemporal patterns of inhibition and excitation that determine directional responses in ganglion cells are shaped by two ‘core’ mechanisms, both arising from distinct specializations of the starburst network.

A Mixed-Signal Structured AdEx Neuron for Accelerated Neuromorphic Cores

Here, we describe a multicompartment neuron circuit based on the adaptive-exponential I&F (AdEx) model, developed for the second-generation BrainScaleS hardware. Based on an existing modular leaky integrate-and-fire (LIF) architecture designed in 65-nm CMOS, the circuit features exponential spike generation, neuronal adaptation, intercompartmental connections as well as a conductance-based reset. The design reproduces a diverse set of firing patterns observed in cortical pyramidal neurons. Further, it enables the emulation of sodium and calcium spikes, as well as N-methyl-D-aspartate plateau potentials known from apical and thin dendrites. We characterize the AdEx circuit extensions and exemplify how the interplay between passive and nonlinear active signal processing enhances the computational capabilities of single (but structured) on-chip neurons.

Non-uniform weighting of local motion inputs underlies dendritic computation in the fly visual system

The fly visual system offers a unique opportunity to explore computations performed by single neurons. Two previous studies characterized, in vivo, the receptive field (RF) of the vertical system (VS) cells of the blowfly (calliphora vicina), both intracellularly in the axon, and, independently using Ca²⁺ imaging, in hundreds of distal dendritic branchlets. We integrated this information into detailed passive cable and compartmental models of 3D reconstructed VS cells. Within a given VS cell type, the transfer resistance (TR) from different branchlets to the axon differs substantially, suggesting that they contribute unequally to the shaping of the axonal RF. Weighting the local RFs of all dendritic branchlets by their respective TR yielded a faithful reproduction of the axonal RF. The model also predicted that the various dendritic branchlets are electrically decoupled from each other, thus acting as independent local functional subunits. The study suggests that single neurons in the fly visual system filter dendritic noise and compute the weighted average of their inputs.

Neural Mechanisms Mediating Motion Sensitivity in Parasol Ganglion Cells of the Primate Retina

Considerable theoretical and experimental effort has been dedicated to understanding how neural circuits detect visual motion. In primates, much is known about the cortical circuits that contribute to motion processing, but the role of the retina in this fundamental neural computation is poorly understood. Here, we used a combination of extracellular and whole-cell recording to test for motion sensitivity in the two main classes of output neurons in the primate retina-midget (parvocellular-projecting) and parasol (magnocellular-projecting) ganglion cells. We report that parasol, but not midget, ganglion cells are motion sensitive. This motion sensitivity is present in synaptic excitation and disinhibition from presynaptic bipolar cells and amacrine cells, respectively. Moreover, electrical coupling between neighboring bipolar cells and the nonlinear nature of synaptic release contribute to the observed motion sensitivity. Our findings indicate that motion computations arise far earlier in the primate visual stream than previously thought.

A neurociência computacional no estudo dos processos cognitivos

Resumo Nas últimas décadas o estudo de processos cognitivos vem sendo influenciado por duas tendências: a legitimação de diversas formas e níveis de estudo e a tentativa de integração multidisciplinar. A primeira teve grande importância na segunda metade do século XX, quando linhas de pesquisa na psicologia cognitiva e nas neurociências fortaleceram-se. Nesse sentido, destacam-se os três níveis de Marr (computacional, algorítmico e implementacional) como forma de estruturar o estudo dos processos cognitivos. A segunda tendência é mais recente e busca, apoiada na primeira, aprofundar o entendimento dos processos cognitivos em suas diversas escalas e integrar diversos paradigmas de estudos, buscando consiliência teórica. O intento deste artigo é apresentar a neurociência computacional e suas possíveis contribuições para a psicologia cognitiva, articulando, por meio dos três níveis de Marr, uma base teórica que explicite o papel de cada uma das disciplinas e as suas possíveis interações.

New Types of Experiments Reveal that a Neuron Functions as Multiple Independent Threshold Units

Neurons are the computational elements that compose the brain and their fundamental principles of activity are known for decades. According to the long-lasting computational scheme, each neuron sums the incoming electrical signals via its dendrites and when the membrane potential reaches a certain threshold the neuron typically generates a spike to its axon. Here we present three types of experiments, using neuronal cultures, indicating that each neuron functions as a collection of independent threshold units. The neuron is anisotropically activated following the origin of the arriving signals to the membrane, via its dendritic trees. The first type of experiments demonstrates that a single neuron’s spike waveform typically varies as a function of the stimulation location. The second type reveals that spatial summation is absent for extracellular stimulations from different directions. The third type indicates that spatial summation and subtraction are not achieved when combining intra- and extra- cellular stimulations, as well as for nonlocal time interference, where the precise timings of the stimulations are irrelevant. Results call to re-examine neuronal functionalities beyond the traditional framework, and the advanced computational capabilities and dynamical properties of such complex systems.

A retinal code for motion along the gravitational and body axes

Self-motion triggers complementary visual and vestibular reflexes supporting image-stabilization and balance. Translation through space produces one global pattern of retinal image motion (optic flow), rotation another. We show that each subtype of direction-selective ganglion cell (DSGC) adjusts its direction preference topographically to align with specific translatory optic flow fields, creating a neural ensemble tuned for a specific direction of motion through space. Four cardinal translatory directions are represented, aligned with two axes of high adaptive relevance: the body and gravitational axes. One subtype maximizes its output when the mouse advances, others when it retreats, rises, or falls. ON-DSGCs and ON-OFF-DSGCs share the same spatial geometry but weight the four channels differently. Each subtype ensemble is also tuned for rotation. The relative activation of DSGC channels uniquely encodes every translation and rotation. Though retinal and vestibular systems both encode translatory and rotatory self-motion, their coordinate systems differ.

Visual Circuits for Direction Selectivity

Images projected onto the retina of an animal eye are rarely still. Instead, they usually contain motion signals originating either from moving objects or from retinal slip caused by self-motion. Accordingly, motion signals tell the animal in which direction a predator, prey, or the animal itself is moving. At the neural level, visual motion detection has been proposed to extract directional information by a delay-and-compare mechanism, representing a classic example of neural computation. Neurons responding selectively to motion in one but not in the other direction have been identified in many systems, most prominently in the mammalian retina and the fly optic lobe. Technological advances have now allowed researchers to characterize these neurons' upstream circuits in exquisite detail. Focusing on these upstream circuits, we review and compare recent progress in understanding the mechanisms that generate direction selectivity in the early visual system of mammals and flies.

Orientation selectivity in the visual cortex of the nine-banded armadillo

Orientation selectivity in primary visual cortex (V1) has been proposed to reflect a canonical computation performed by the neocortical circuitry. Although orientation selectivity has been reported in all mammals examined to date, the degree of selectivity and the functional organization of selectivity vary across mammalian clades. The differences in degree of orientation selectivity are large, from reports in marsupials that only a small subset of neurons are selective to studies in carnivores, in which it is rare to find a neuron lacking selectivity. Furthermore, the functional organization in cortex varies in that the primate and carnivore V1 is characterized by an organization in which nearby neurons share orientation preference while other mammals such as rodents and lagomorphs either lack or have only extremely weak clustering. To gain insight into the evolutionary emergence of orientation selectivity, we examined the nine-banded armadillo, a species within the early placental clade Xenarthra. Here we use a combination of neuroimaging, histological, and electrophysiological methods to identify the retinofugal pathways, locate V1, and for the first time examine the functional properties of V1 neurons in the armadillo (Dasypus novemcinctus) V1. Individual neurons were strongly sensitive to the orientation and often the direction of drifting gratings. We uncovered a wide range of orientation preferences but found a bias for horizontal gratings. The presence of strong orientation selectivity in armadillos suggests that the circuitry responsible for this computation is common to all placental mammals.NEW & NOTEWORTHY The current study shows that armadillo primary visual cortex (V1) neurons share the signature properties of V1 neurons of primates, carnivorans, and rodents. Furthermore, these neurons exhibit a degree of selectivity for stimulus orientation and motion direction similar to that found in primate V1. Our findings in armadillo visual cortex suggest that the functional properties of V1 neurons emerged early in the mammalian lineage, near the time of the divergence of marsupials.

Direction-Selective Circuits Shape Noise to Ensure a Precise Population Code

Neural responses are noisy, and circuit structure can correlate this noise across neurons. Theoretical studies show that noise correlations can have diverse effects on population coding, but these studies rarely explore stimulus dependence of noise correlations. Here, we show that noise correlations in responses of ON-OFF direction-selective retinal ganglion cells are strongly stimulus dependent, and we uncover the circuit mechanisms producing this stimulus dependence. A population model based on these mechanistic studies shows that stimulus-dependent noise correlations improve the encoding of motion direction 2-fold compared to independent noise. This work demonstrates a mechanism by which a neural circuit effectively shapes its signal and noise in concert, minimizing corruption of signal by noise. Finally, we generalize our findings beyond direction coding in the retina and show that stimulus-dependent correlations will generally enhance information coding in populations of diversely tuned neurons.

NMDA Receptors Multiplicatively Scale Visual Signals and Enhance Directional Motion Discrimination in Retinal Ganglion Cells

Postsynaptic responses in many CNS neurons are typically small and variable, often making it difficult to distinguish physiologically relevant signals from background noise. To extract salient information, neurons are thought to integrate multiple synaptic inputs and/or selectively amplify specific synaptic activation patterns. Here, we present evidence for a third strategy: directionally selective ganglion cells (DSGCs) in the mouse retina multiplicatively scale visual signals via a mechanism that requires both nonlinear NMDA receptor (NMDAR) conductances in DSGC dendrites and directionally tuned inhibition provided by the upstream retinal circuitry. Postsynaptic multiplication enables DSGCs to discriminate visual motion more accurately in noisy visual conditions without compromising directional tuning. These findings demonstrate a novel role for NMDARs in synaptic processing and provide new insights into how synaptic and network features interact to accomplish physiologically relevant neural computations.

Direction-selective units in goldfish retina and tectum opticum - review and new aspects

The output units of fish retina, i.e., the retinal ganglion cells (detectors), send highly processed information to the primary visual centers of the brain, settled in the midbrain formation tectum opticum (TO). Axons of different fish motion detectors terminate in different tectal levels. In the superficial layer of TO, axons of direction-selective ganglion cells (DS GCs) are terminated. Single unit responses of the DS GCs were recorded in intact fish from their axon terminals in TO. Goldfish DS GCs projecting to TO were shown to comprise six physiological types according to their selectivity to sign of stimulus contrast (ON and OFF units) and their preferred directions: three directions separated by 120[Formula: see text]. These units, characterized by relatively small receptive fields and remarkable spatial resolution should be classified as local motion detectors. In addition to the retinal DS GCs, other kinds of DS units were extracellularly recorded in the superficial and deep sublaminae of tectum. Some features of their responses suggested that they originated from tectal neurons (TNs). Contrary to DS GCs which are characterized by small RFs and use separate ON and OFF channels, DS TNs have extra-large RFs and ON-OFF type responses. DS TNs were shown to select four preferred directions. Three of them are compatible with those already selected on the retinal level. Complementary to them, the fourth DS TN type with rostro-caudal preference (lacking in the retina) has been revealed. Possible functional interrelations between DS GCs and DS TNs are discussed.

Counting on dis-inhibition: a circuit motif for interval counting and selectivity in the anuran auditory system

Information can be encoded in the temporal patterning of spikes. How the brain reads these patterns is of general importance and represents one of the greatest challenges in neuroscience. We addressed this issue in relation to temporal pattern recognition in the anuran auditory system. Many species of anurans perform mating decisions based on the temporal structure of advertisement calls. One important temporal feature is the number of sound pulses that occur with a species-specific interpulse interval. Neurons representing this pulse count have been recorded in the anuran inferior colliculus, but the mechanisms underlying their temporal selectivity are incompletely understood. Here, we construct a parsimonious model that can explain the key dynamical features of these cells with biologically plausible elements. We demonstrate that interval counting arises naturally when combining interval-selective inhibition with pulse-per-pulse excitation having both fast- and slow-conductance synapses. Interval-dependent inhibition is modeled here by a simple architecture based on known physiology of afferent nuclei. Finally, we consider simple implementations of previously proposed mechanistic explanations for these counting neurons and show that they do not account for all experimental observations. Our results demonstrate that tens of millisecond-range temporal selectivities can arise from simple connectivity motifs of inhibitory neurons, without recourse to internal clocks, spike-frequency adaptation, or appreciable short-term plasticity.

Identification of Synaptic Patterns of NMDA Receptor Subtypes Upon Direction-Selective Rabbit Retinal Ganglion Cells

PURPOSE

The objective of this study was to identify anisotropies that contribute to the directional preference of direction-selective retinal ganglion cells (DS RGCs) in the rabbit retina. We investigated the distributions of N-methyl-d-aspartate receptor 1 (NMDAR1), NMDAR2A and NMDAR2B receptor subunits in the dendritic arbors of rabbit DS RGCs.

METHODS

The distributions of the NMDAR subunits on the DS RGCs were determined using immunocytochemistry. DS RGCs were injected with Lucifer yellow, and the cells were identified by their characteristic morphology. The triple-labeled images of dendrites, kinesin II and NMDARs were visualized using confocal microscopy and were reconstructed from high-resolution confocal images.

RESULTS

We found no evidence of asymmetry in any of the NMDAR subunits examined on the dendritic arbors of both the ON and OFF layers of DS RGCs.

CONCLUSIONS

Our results indicate that direction selectivity appears to lie in the neuronal circuitry afferent to the DS RGCs.

General features of the retinal connectome determine the computation of motion anticipation

Motion anticipation allows the visual system to compensate for the slow speed of phototransduction so that a moving object can be accurately located. This correction is already present in the signal that ganglion cells send from the retina but the biophysical mechanisms underlying this computation are not known. Here we demonstrate that motion anticipation is computed autonomously within the dendritic tree of each ganglion cell and relies on feedforward inhibition. The passive and non-linear interaction of excitatory and inhibitory synapses enables the somatic voltage to encode the actual position of a moving object instead of its delayed representation. General rather than specific features of the retinal connectome govern this computation: an excess of inhibitory inputs over excitatory, with both being randomly distributed, allows tracking of all directions of motion, while the average distance between inputs determines the object velocities that can be compensated for.

DOI:http://dx.doi.org/10.7554/eLife.06250.001

The retina is a structure at the back of the eye that converts light into nerve impulses, which are then processed in the brain to produce the images that we see. It normally takes about one-tenth of a second for the retina to send a signal to the brain after an object first moves into view. This is about the same time it takes a tennis ball to travel several meters during a tennis match, yet we are still able to see where the moving tennis ball is in real time. This is because a process called ‘motion anticipation’ is able to compensate for the delay in processing the position of a moving object. However, it was not known precisely how motion anticipation occurs.

Inside the retina, cells called photoreceptors detect light and ultimately send signals (via some intermediate cell types) to nerve cells known as retinal ganglion cells. These signals can either excite a retinal ganglion cell to cause it to send an electrical signal to the brain, or inhibit it, which temporarily prevents electrical activity. Each cell receives signals from several photoreceptors, which each connect to a different site along branch-like structures called dendrites that project out of the retinal ganglion cells.

Johnston and Lagnado have now investigated how motion anticipation occurs in the retina by using electrical recordings of the activity in the retinas of goldfish combined with computer simulations of this activity. This revealed inhibitory signals, sent from photoreceptors to retinal ganglion cells via a type of intermediate cell (called amacrine cells), play a key role in motion anticipation. The ability to track motion effectively in all directions requires more inhibitory signals to be sent to the dendrites of a retinal ganglion cell than excitatory signals. These two types of input must also be randomly distributed across the cell. Furthermore, it is the density of these input sites on a dendrite that determines how well the retina can compensate for the motion of a fast-moving object. The building blocks required for motion anticipation in the retina are also found in visual areas higher in the brain. Therefore, further work may reveal that higher visual areas also use this mechanism to predict the future location of moving objects.

DOI:http://dx.doi.org/10.7554/eLife.06250.002

Spike-timing prediction in cortical neurons with active dendrites

A complete single-neuron model must correctly reproduce the firing of spikes and bursts. We present a study of a simplified model of deep pyramidal cells of the cortex with active dendrites. We hypothesized that we can model the soma and its apical dendrite with only two compartments, without significant loss in the accuracy of spike-timing predictions. The model is based on experimentally measurable impulse-response functions, which transfer the effect of current injected in one compartment to current reaching the other. Each compartment was modeled with a pair of non-linear differential equations and a small number of parameters that approximate the Hodgkin-and-Huxley equations. The predictive power of this model was tested on electrophysiological experiments where noisy current was injected in both the soma and the apical dendrite simultaneously. We conclude that a simple two-compartment model can predict spike times of pyramidal cells stimulated in the soma and dendrites simultaneously. Our results support that regenerating activity in the apical dendritic is required to properly account for the dynamics of layer 5 pyramidal cells under in-vivo-like conditions.

Pyramidal cell development: postnatal spinogenesis, dendritic growth, axon growth, and electrophysiology

Here we review recent findings related to postnatal spinogenesis, dendritic and axon growth, pruning and electrophysiology of neocortical pyramidal cells in the developing primate brain. Pyramidal cells in sensory, association and executive cortex grow dendrites, spines and axons at different rates, and vary in the degree of pruning. Of particular note is the fact that pyramidal cells in primary visual area (V1) prune more spines than they grow during postnatal development, whereas those in inferotemporal (TEO and TE) and granular prefrontal cortex (gPFC; Brodmann's area 12) grow more than they prune. Moreover, pyramidal cells in TEO, TE and the gPFC continue to grow larger dendritic territories from birth into adulthood, replete with spines, whereas those in V1 become smaller during this time. The developmental profile of intrinsic axons also varies between cortical areas: those in V1, for example, undergo an early proliferation followed by pruning and local consolidation into adulthood, whereas those in area TE tend to establish their territory and consolidate it into adulthood with little pruning. We correlate the anatomical findings with the electrophysiological properties of cells in the different cortical areas, including membrane time constant, depolarizing sag, duration of individual action potentials, and spike-frequency adaptation. All of the electrophysiological variables ramped up before 7 months of age in V1, but continued to ramp up over a protracted period of time in area TE. These data suggest that the anatomical and electrophysiological profiles of pyramidal cells vary among cortical areas at birth, and continue to diverge into adulthood. Moreover, the data reveal that the “use it or lose it” notion of synaptic reinforcement may speak to only part of the story, “use it but you still might lose it” may be just as prevalent in the cerebral cortex.

Unsupervised learnable neuron model with nonlinear interaction on dendrites

Recent researches have provided strong circumstantial support to dendrites playing a key and possibly essential role in computations. In this paper, we propose an unsupervised learnable neuron model by including the nonlinear interactions between excitation and inhibition on dendrites. The model neuron self-adjusts its synaptic parameters, so that the synapse to dendrite, according to a generalized delta-rule-like algorithm. The model is used to simulate directionally selective cells by the unsupervised learning algorithm. In the simulations, we initialize the interaction and dendrite of the neuron randomly and use the generalized delta-rule-like unsupervised learning algorithm to learn the two-dimensional multi-directional selectivity problem without an external teacher's signals. Simulation results show that the directionally selective cells can be formed by unsupervised learning, acquiring the required number of dendritic branches, and if needed, enhanced and if not, eliminated. Further, the results show whether a synapse exists; if it exists, where and what type (excitatory or inhibitory) of synapse it is. This leads us to believe that the proposed neuron model may be considerably more powerful on computations than the McCulloch-Pitts model because theoretically a single neuron or a single layer of such neurons is capable of solving any complex problem. These may also lead to a completely new technique for analyzing the mechanisms and principles of neurons, dendrites, and synapses.

Excitatory synaptic inputs to mouse on-off direction-selective retinal ganglion cells lack direction tuning

Direction selectivity represents a fundamental visual computation. In mammalian retina, On-Off direction-selective ganglion cells (DSGCs) respond strongly to motion in a preferred direction and weakly to motion in the opposite, null direction. Electrical recordings suggested three direction-selective (DS) synaptic mechanisms: DS GABA release during null-direction motion from starburst amacrine cells (SACs) and DS acetylcholine and glutamate release during preferred direction motion from SACs and bipolar cells. However, evidence for DS acetylcholine and glutamate release has been inconsistent and at least one bipolar cell type that contacts another DSGC (On-type) lacks DS release. Here, whole-cell recordings in mouse retina showed that cholinergic input to On-Off DSGCs lacked DS, whereas the remaining (glutamatergic) input showed apparent DS. Fluorescence measurements with the glutamate biosensor intensity-based glutamate-sensing fluorescent reporter (iGluSnFR) conditionally expressed in On-Off DSGCs showed that glutamate release in both On- and Off-layer dendrites lacked DS, whereas simultaneously recorded excitatory currents showed apparent DS. With GABA-A receptors blocked, both iGluSnFR signals and excitatory currents lacked DS. Our measurements rule out DS release from bipolar cells onto On-Off DSGCs and support a theoretical model suggesting that apparent DS excitation in voltage-clamp recordings results from inadequate voltage control of DSGC dendrites during null-direction inhibition. SAC GABA release is the apparent sole source of DS input onto On-Off DSGCs.

Network oscillations drive correlated spiking of ON and OFF ganglion cells in the rd1 mouse model of retinal degeneration

Following photoreceptor degeneration, ON and OFF retinal ganglion cells (RGCs) in the rd-1/rd-1 mouse receive rhythmic synaptic input that elicits bursts of action potentials at ∼ 10 Hz. To characterize the properties of this activity, RGCs were targeted for paired recording and morphological classification as either ON alpha, OFF alpha or non-alpha RGCs using two-photon imaging. Identified cell types exhibited rhythmic spike activity. Cross-correlation of spike trains recorded simultaneously from pairs of RGCs revealed that activity was correlated more strongly between alpha RGCs than between alpha and non-alpha cell pairs. Bursts of action potentials in alpha RGC pairs of the same type, i.e. two ON or two OFF cells, were in phase, while bursts in dissimilar alpha cell types, i.e. an ON and an OFF RGC, were 180 degrees out of phase. This result is consistent with RGC activity being driven by an input that provides correlated excitation to ON cells and inhibition to OFF cells. A2 amacrine cells were investigated as a candidate cellular mechanism and found to display 10 Hz oscillations in membrane voltage and current that persisted in the presence of antagonists of fast synaptic transmission and were eliminated by tetrodotoxin. Results support the conclusion that the rhythmic RGC activity originates in a presynaptic network of electrically coupled cells including A2s via a Na(+)-channel dependent mechanism. Network activity drives out of phase oscillations in ON and OFF cone bipolar cells, entraining similar frequency fluctuations in RGC spike activity over an area of retina that migrates with changes in the spatial locus of the cellular oscillator.

Spatial distribution of excitatory synapses on the dendrites of ganglion cells in the mouse retina

Excitatory glutamatergic inputs from bipolar cells affect the physiological properties of ganglion cells in the mammalian retina. The spatial distribution of these excitatory synapses on the dendrites of retinal ganglion cells thus may shape their distinct functions. To visualize the spatial pattern of excitatory glutamatergic input into the ganglion cells in the mouse retina, particle-mediated gene transfer of plasmids expressing postsynaptic density 95-green fluorescent fusion protein (PSD95-GFP) was used to label the excitatory synapses. Despite wide variation in the size and morphology of the retinal ganglion cells, the expression of PSD95 puncta was found to follow two general rules. Firstly, the PSD95 puncta are regularly spaced, at 1–2 µm intervals, along the dendrites, whereby the presence of an excitatory synapse creates an exclusion zone that rules out the presence of other glutamatergic synaptic inputs. Secondly, the spatial distribution of PSD95 puncta on the dendrites of diverse retinal ganglion cells are similar in that the number of excitatory synapses appears to be less on primary dendrites and to increase to a plateau on higher branch order dendrites. These observations suggest that synaptogenesis is spatially regulated along the dendritic segments and that the number of synaptic contacts is relatively constant beyond the primary dendrites. Interestingly, we also found that the linear puncta density is slightly higher in large cells than in small cells. This may suggest that retinal ganglion cells with a large dendritic field tend to show an increased connectivity of excitatory synapses that makes up for their reduced dendrite density. Mapping the spatial distribution pattern of the excitatory synapses on retinal ganglion cells thus provides explicit structural information that is essential for our understanding of how excitatory glutamatergic inputs shape neuronal responses.

Direction selectivity is computed by active dendritic integration in retinal ganglion cells

Active dendritic integration is thought to enrich the computational power of central neurons. However, a direct role of active dendritic processing in the execution of defined neuronal computations in intact neural networks has not been established. Here we used multi-site electrophysiological recording techniques to demonstrate that active dendritic integration underlies the computation of direction selectivity in rabbit retinal ganglion cells. Direction-selective retinal ganglion cells fire action potentials in response to visual image movement in a preferred direction. Dendritic recordings revealed that preferred-direction moving-light stimuli led to dendritic spike generation in terminal dendrites, which were further integrated and amplified as they spread through the dendritic arbor to the axon to drive action potential output. In contrast, when light bars moved in a null direction, synaptic inhibition vetoed neuronal output by directly inhibiting terminal dendritic spike initiation. Active dendritic integration therefore underlies a physiologically engaged circuit-based computation in the retina.

Emergence of orientation selectivity in the Mammalian visual pathway

Orientation selectivity is a property of mammalian primary visual cortex (V1) neurons, yet its emergence along the visual pathway varies across species. In carnivores and primates, elongated receptive fields first appear in V1, whereas in lagomorphs such receptive fields emerge earlier, in the retina. Here we examine the mouse visual pathway and reveal the existence of orientation selectivity in lateral geniculate nucleus (LGN) relay cells. Cortical inactivation does not reduce this orientation selectivity, indicating that cortical feedback is not its source. Orientation selectivity is similar for LGN relay cells spiking and subthreshold input to V1 neurons, suggesting that cortical orientation selectivity is inherited from the LGN in mouse. In contrast, orientation selectivity of cat LGN relay cells is small relative to subthreshold inputs onto V1 simple cells. Together, these differences show that although orientation selectivity exists in visual neurons of both rodents and carnivores, its emergence along the visual pathway, and thus its underlying neuronal circuitry, is fundamentally different.

Mathematical analysis and modeling of motion direction selectivity in the retina

Motion detection is one of the most important and primitive computations performed by our visual system. Specifically in the retina, ganglion cells producing motion direction-selective responses have been addressed by different disciplines, such as mathematics, neurophysiology and computational modeling, since the beginnings of vision science. Although a number of studies have analyzed theoretical and mathematical considerations for such responses, a clear picture of the underlying cellular mechanisms is only recently emerging. In general, motion direction selectivity is based on a non-linear asymmetric computation inside a receptive field differentiating cell responses between preferred and null direction stimuli. To what extent can biological findings match these considerations? In this review, we outline theoretical and mathematical studies of motion direction selectivity, aiming to map the properties of the models onto the neural circuitry and synaptic connectivity found in the retina. Additionally, we review several compartmental models that have tried to fill this gap. Finally, we discuss the remaining challenges that computational models will have to tackle in order to fully understand the retinal motion direction-selective circuitry.

The first stage of cardinal direction selectivity is localized to the dendrites of retinal ganglion cells

Inferring the direction of image motion is a fundamental component of visual computation and essential for visually guided behavior. In the retina, the direction of image motion is computed in four cardinal directions, but it is not known at which circuit location along the flow of visual information the cardinal direction selectivity first appears. We recorded the concerted activity of the neuronal circuit elements of single direction-selective (DS) retinal ganglion cells at subcellular resolution by combining GCaMP3-functionalized transsynaptic viral tracing and two-photon imaging. While the visually evoked activity of the dendritic segments of the DS cells were direction selective, direction-selective activity was absent in the axon terminals of bipolar cells. Furthermore, the glutamate input to DS cells, recorded using a genetically encoded glutamate sensor, also lacked direction selectivity. Therefore, the first stage in which extraction of a cardinal motion direction occurs is the dendrites of DS cells.

Visual stimulation reverses the directional preference of direction-selective retinal ganglion cells

Direction selectivity in the retina is mediated by direction-selective ganglion cells. These cells are part of a circuit in which they are asymmetrically wired to inhibitory neurons. Thus, they respond strongly to an image moving in the preferred direction and weakly to an image moving in the opposite (null) direction. Here, we demonstrate that adaptation with short visual stimulation of a direction-selective ganglion cell using drifting gratings can reverse this cell's directional preference by 180°. This reversal is robust, long lasting, and independent of the animal's age. Our findings indicate that, even within circuits that are hardwired, the computation of direction can be altered by dynamic circuit mechanisms that are guided by visual stimulation.

Ionotropic glutamate receptors mediate OFF responses in light-adapted ON bipolar cells

Previous studies have suggested that photoreceptor synaptic inputs to depolarizing bipolar cells (DBCs or ON bipolar cells) are mediated by mGluR6 receptors and those to hyperpolarizing bipolar cells (HBCs or OFF bipolar cells) are mediated by AMPA/kainate receptors. Here we show that in addition to mGluR6 receptors which mediate the sign-inverting, depolarizing light responses, subpopulations of cone-dominated and rod/cone mixed DBCs use GluR4 AMPA receptors to generate a transient sign-preserving OFF response under light adapted conditions. These AMPA receptors are located at the basal junctions postsynaptic to rods and they are silent under dark-adapted conditions, as tonic glutamate release in darkness desensitizes these receptors. Light adaptation enhances rod-cone coupling and thus allows cone photocurrents with an abrupt OFF depolarization to enter the rods. The abrupt rod depolarization triggers glutamate activation of unoccupied AMPA receptors, resulting in a transient OFF response in DBCs. It has been widely accepted that the DNQX-sensitive, OFF transient responses in retinal amacrine cells and ganglion cells are mediated exclusively by HBCs. Our results suggests that this view needs revision as AMPA receptors in subpopulations of DBCs are likely to significantly contribute to the DNQX-sensitive OFF transient responses in light-adapted third- and higher-order visual neurons.

Synaptic Pattern of KA1 and KA2 upon the Direction-Selective Ganglion Cells in Developing and Adult Mouse Retina

The detection of image motion is important to vision. Direction-selective retinal ganglion cells (DS-RGCs) respond strongly to stimuli moving in one direction of motion and are strongly inhibited by stimuli moving in the opposite direction. In this article, we investigated the distributions of kainate glutamate receptor subtypes KA1 and KA2 on the dendritic arbors of DS-RGCs in developing (5, 10) days postnatal (PN) and adult mouse retina to search for anisotropies. The distribution of kainate receptor subtypes on the DS-RGCs was determined using antibody immunocytochemistry. To identify their characteristic morphology, DS-RGCs were injected with Lucifer yellow. The triple-labeled images of dendrites, kinesin II, and receptors were visualized by confocal microscopy and were reconstructed from high-resolution confocal images. We found no evidence of asymmetry in any of the kainate receptor subunits examined on the dendritic arbors of both the On and Off layers of DS-RGCs in all periods of developing and adult stage that would predict direction selectivity.

Direction selectivity in the retina: symmetry and asymmetry in structure and function

Visual information is processed in the retina to a remarkable degree before it is transmitted to higher visual centres. Several types of retinal ganglion cells (the output neurons of the retina) respond preferentially to image motion in a particular direction, and each type of direction-selective ganglion cell (DSGC) is comprised of multiple subtypes with different preferred directions. The direction selectivity of the cells is generated by diverse mechanisms operating within microcircuits that rely on independent neuronal processing in individual dendrites of both the DSGCs and the presynaptic neurons that innervate them.

Parallel mechanisms encode direction in the retina

In the retina, presynaptic inhibitory mechanisms that shape directionally selective (DS) responses in output ganglion cells are well established. However, the nature of inhibition-independent forms of directional selectivity remains poorly defined. Here, we describe a genetically specified set of ON-OFF DS ganglion cells (DSGCs) that code anterior motion. This entire population of DSGCs exhibits asymmetric dendritic arborizations that orientate toward the preferred direction. We demonstrate that morphological asymmetries along with nonlinear dendritic conductances generate a centrifugal (soma-to-dendrite) preference that does not critically depend upon, but works in parallel with the GABAergic circuitry. We also show that in symmetrical DSGCs, such dendritic DS mechanisms are aligned with, or are in opposition to, the inhibitory DS circuitry in distinct dendritic subfields where they differentially interact to promote or weaken directional preferences. Thus, pre- and postsynaptic DS mechanisms interact uniquely in distinct ganglion cell populations, enabling efficient DS coding under diverse conditions.

Imperfect space clamp permits electrotonic interactions between inhibitory and excitatory synaptic conductances, distorting voltage clamp recordings

The voltage clamp technique is frequently used to examine the strength and composition of synaptic input to neurons. Even accounting for imperfect voltage control of the entire cell membrane ("space clamp"), it is often assumed that currents measured at the soma are a proportional indicator of the postsynaptic conductance. Here, using NEURON simulation software to model somatic recordings from morphologically realistic neurons, we show that excitatory conductances recorded in voltage clamp mode are distorted significantly by neighboring inhibitory conductances, even when the postsynaptic membrane potential starts at the reversal potential of the inhibitory conductance. Analogous effects are observed when inhibitory postsynaptic currents are recorded at the reversal potential of the excitatory conductance. Escape potentials in poorly clamped dendrites reduce the amplitude of excitatory or inhibitory postsynaptic currents recorded at the reversal potential of the other conductance. In addition, unclamped postsynaptic inhibitory conductances linearize the recorded current-voltage relationship of excitatory inputs comprising AMPAR and NMDAR-mediated components, leading to significant underestimation of the relative contribution by NMDARs, which are particularly sensitive to small perturbations in membrane potential. Voltage clamp accuracy varies substantially between neurons and dendritic arbors of different morphology; as expected, more reliable recordings are obtained from dendrites near the soma, but up to 80% of the synaptic signal on thin, distant dendrites may be lost when postsynaptic interactions are present. These limitations of the voltage clamp technique may explain how postsynaptic effects on synaptic transmission could, in some cases, be attributed incorrectly to presynaptic mechanisms.