Direction-selective Ganglion Cells Research Articles

A novel artificial visual system for motion direction detection in color images

In recent years, convolutional neural networks (CNNs) have dominated the field of computer vision. Compared to traditional methods, these neural network algorithms exhibit strong biomimetic performance advantages in complex visual tasks due to their brain-like structure. However, because some necessary neural characteristics are ignored, these neural network algorithms differ greatly from the computational mechanisms of the brain. This paper starts with extracting basic visual features such as motion direction information from the brain and abstracts, generalizes and models a novel artificial visual system (AVS) for detecting object motion direction in color images based on existing relevant neurophysiological knowledge. We propose a mathematical model and quantification mechanism for each component neuron that generates motion direction selectivity in the visual system using dendritic neuron models, spiking neural network concepts and neurophysiological knowledge of retinal direction-selective ganglion cell pathways. The experiment is based on one million instances of object motion under different environments of noise-free, static and dynamic random noise, dynamic salt-and-pepper noise, dynamic Gaussion noise, and dynamic light changing. In comparison with 4 famous CNNs, LeNet-5, EfficientNetB0, ResNet18, and RegNetX-200MF, we test and verify AVS’s effectiveness, efficiency and strong generalization ability as well as other biomimetic performance advantages including high biological rationality, learning-free capability, interpretability and ease-of-use etc. AVS demonstrates that neuroscience still has important implications for guiding and promoting the development of artificial intelligence technology. Furthermore, AVS firstly provides a successful quantitative reference case study for further understanding motion direction selectivity and other primary cortical encoding characteristics in the brain.

Predation without direction selectivity

Across the animal kingdom, visual predation relies on motion-sensing neurons in the superior colliculus (SC) and its orthologs. These neurons exhibit complex stimulus preferences, including direction selectivity, which is thought to be critical for tracking the unpredictable escape routes of prey. The source of direction selectivity in the SC is contested, and its contributions to predation have not been tested experimentally. Here, we use type-specific cell removal to show that narrow-field (NF) neurons in the mouse SC guide predation. In vivo recordings demonstrate that direction-selective responses of NF cells are independent of recently reported stimulus-edge effects. Monosynaptic retrograde tracing reveals that NF cells receive synaptic input from direction-selective ganglion cells. When we eliminate direction selectivity in the retina of adult mice, direction-selective responses in the SC, including in NF cells, are lost. However, eliminating retinal direction selectivity does not affect the hunting success or strategies of mice, even when direction selectivity is removed after mice have learned to hunt, and despite abolishing the gaze-stabilizing optokinetic reflex. Thus, our results identify the retinal source of direction selectivity in the SC. They show that NF cells in the SC guide predation, an essential spatial orienting task, independent of their direction selectivity, revealing behavioral multiplexing of complex neural feature preferences and highlighting the importance of feature-selective manipulations for neuroethology.

Dopamine and motion perception: Unveiling retinal connections in Parkinson's disease

Aims/Purpose: Parkinson's disease (PD) is characterized by degeneration of dopaminergic neurons, which occurs not only in the pars compacta of the sustantia nigra of the midbrain, but also in the retina, where dopaminergic amacrine cells (DAC) have been well characterized. Indeed, parkinsonian patients display visual impairments in early stages of the disease, including motion perception deterioration. Retinal cells responsible of motion perception are the directional selective ganglion cells (DSGC) which receive inhibitory synapses from starburst amacrine cells (SAC). Consequences of DAC degeneration in PD on the functionality of SAC and DSGC is unknown. In this work, we studied the functional connections between DAC and SAC in order to understand the pathophysiological mechanisms that determine the visual dysfunction in PD patients.Methods: Using immunohistochemistry and confocal imaging, we studied the retinal connections between DAC and SAC in wild‐type mice. To determine the type of Dopamine (DA) receptor expressed by SAC, we used transgenic mice whose cells expressing D1 or D2 receptors also express fluorescent markers. Intravitreal injections of 6‐OHDA were performed in a series of animals, followed by immunohistochemistry and optokinetic test.Results: We observed that DAC and SAC have direct synaptic contacts, and that the SAC located in the OFF sublamina of the inner plexiform layer express D2 receptors (D2R) but not D1 receptors. Moreover, depletion of DA in the 6‐OHDA injected mice induced a significant reduction of motion perception.Conclusions: This study shows for the first time in mice that the amacrine retinal cells responsible for perceiving directional selective moving stimuli express D2R. These findings provide new ways of detecting visual impairments in preclinical models of retinal DA degeneration.

Cadherin-13 Maintains Retinotectal Synapses via Transneuronal Interactions.

Maintaining precise synaptic contacts between neuronal partners is critical to ensure the proper functioning of the mammalian central nervous system (CNS). Diverse cell recognition molecules, such as classic cadherins (Cdhs), are part of the molecular machinery mediating synaptic choices during development and synaptic maintenance. Yet, the principles governing neuron-neuron wiring across diverse CNS neuron types remain largely unknown. The retinotectal synapses, connections from the retinal ganglion cells (RGCs) to the superior collicular (SC) neurons, offer an ideal experimental system to reveal molecular logic underlying synaptic choices and formation. This is due to the retina's unidirectional and laminar-restricted projections to the SC and the large databases of presynaptic RGC subtypes and postsynaptic SC neuronal types. Here, we focused on determining the role of Type II Cdhs in wiring the retinotectal synapses. We surveyed Cdhs expression patterns at neuronal resolution and revealed that Cdh13 is enriched in the wide-field neurons in the superficial SC (sSC). In either the Cdh13 null mutant or selective adult deletion within the wide-field neurons, there is a significant reduction of spine densities in the distal dendrites of these neurons in both sexes. Additionally, Cdh13 removal from presynaptic RGCs reduced dendritic spines in the postsynaptic wide-field neurons. Cdh13-expressing RGCs use differential mechanisms than αRGCs and On-Off Direction-Selective Ganglion Cells (ooDSGCs) to form specific retinotectal synapses. The results revealed a selective transneuronal interaction mediated by Cdh13 to maintain proper retinotectal synapses in vivo.

Neural Circuits Underlying Multifeature Extraction in the Retina.

Classic ON-OFF direction-selective ganglion cells (DSGCs) that encode the four cardinal directions were recently shown to also be orientation-selective. To clarify the mechanisms underlying orientation selectivity, we employed a variety of electrophysiological, optogenetic, and gene knock-out strategies to test the relative contributions of glutamate, GABA, and acetylcholine (ACh) input that are known to drive DSGCs, in male and female mouse retinas. Extracellular spike recordings revealed that DSGCs respond preferentially to either vertical or horizontal bars, those that are perpendicular to their preferred-null motion axes. By contrast, the glutamate input to all four DSGC types measured using whole-cell patch-clamp techniques was found to be tuned along the vertical axis. Tuned glutamatergic excitation was heavily reliant on type 5A bipolar cells, which appear to be electrically coupled via connexin 36 containing gap junctions to the vertically oriented processes of wide-field amacrine cells. Vertically tuned inputs are transformed by the GABAergic/cholinergic "starburst" amacrine cells (SACs), which are critical components of the direction-selective circuit, into distinct patterns of inhibition and excitation. Feed-forward SAC inhibition appears to "veto" preferred orientation glutamate excitation in dorsal/ventral (but not nasal/temporal) coding DSGCs "flipping" their orientation tuning by 90° and accounts for the apparent mismatch between glutamate input tuning and the DSGC's spiking response. Together, these results reveal how two distinct synaptic motifs interact to generate complex feature selectivity, shedding light on the intricate circuitry that underlies visual processing in the retina.

An ON-type direction-selective ganglion cell in primate retina

To maintain a stable and clear image of the world, our eyes reflexively follow the direction in which a visual scene is moving. Such gaze-stabilization mechanisms reduce image blur as we move in the environment. In non-primate mammals, this behaviour is initiated by retinal output neurons called ON-type direction-selective ganglion cells (ON-DSGCs), which detect the direction of image motion and transmit signals to brainstem nuclei that drive compensatory eye movements1. However, ON-DSGCs have not yet been identified in the retina of primates, raising the possibility that this reflex is mediated by cortical visual areas. Here we mined single-cell RNA transcriptomic data from primate retina to identify a candidate ON-DSGC. We then combined two-photon calcium imaging, molecular identification and morphological analysis to reveal a population of ON-DSGCs in the macaque retina. The morphology, molecular signature and GABA (γ-aminobutyric acid)-dependent mechanisms that underlie direction selectivity in primate ON-DSGCs are highly conserved with those in other mammals. We further identify a candidate ON-DSGC in human retina. The presence of ON-DSGCs in primates highlights the need to examine the contribution of subcortical retinal mechanisms to normal and aberrant gaze stabilization in the developing and mature visual system.

Asymmetric connections with starburst amacrine cells underlie the upward motion selectivity of J-type retinal ganglion cells.

Motion is an important aspect of visual information. The directions of visual motion are encoded in the retina by direction-selective ganglion cells (DSGCs). ON-OFF DSGCs and ON DSGCs co-stratify with starburst amacrine cells (SACs) in the inner plexiform layer and depend on SACs for their direction selectivity. J-type retinal ganglion cells (J-RGCs), a type of OFF DSGCs in the mouse retina, on the other hand, do not co-stratify with SACs, and how direction selectivity in J-RGCs emerges has not been understood. Here, we report that both the excitatory and inhibitory synaptic inputs to J-RGCs are direction-selective (DS), with the inhibitory inputs playing a more important role for direction selectivity. The DS inhibitory inputs come from SACs, and the functional connections between J-RGCs and SACs are spatially asymmetric. Thus, J-RGCs and SACs form functionally important synaptic contacts even though their dendritic arbors show little overlap. These findings underscore the need to look beyond the neurons' stratification patterns in retinal circuit studies. Our results also highlight the critical role of SACs for retinal direction selectivity.

A Novel Artificial Visual System for Motion Direction Detection with Completely Modeled Retinal Direction-Selective Pathway

Some fundamental visual features have been found to be fully extracted before reaching the cerebral cortex. We focus on direction-selective ganglion cells (DSGCs), which exist at the terminal end of the retinal pathway, at the forefront of the visual system. By utilizing a layered pathway composed of various relevant cells in the early stage of the retina, DSGCs can extract multiple motion directions occurring in the visual field. However, despite a considerable amount of comprehensive research (from cells to structures), a definitive conclusion explaining the specific details of the underlying mechanisms has not been reached. In this paper, leveraging some important conclusions from neuroscience research, we propose a complete quantified model for the retinal motion direction selection pathway and elucidate the global motion direction information acquisition mechanism from DSGCs to the cortex using a simple spiking neural mechanism. This mechanism is referred to as the artificial visual system (AVS). We conduct extensive testing, including one million sets of two-dimensional eight-directional binary object motion instances with 10 different object sizes and random object shapes. We also evaluate AVS’s noise resistance and generalization performance by introducing random static and dynamic noises. Furthermore, to thoroughly validate AVS’s efficiency, we compare its performance with two state-of-the-art deep learning algorithms (LeNet-5 and EfficientNetB0) in all tests. The experimental results demonstrate that due to its highly biomimetic design and characteristics, AVS exhibits outstanding performance in motion direction detection. Additionally, AVS possesses biomimetic computing advantages in terms of hardware implementation, learning difficulty, and parameter quantity.

TARPγ2 Is Required for Normal AMPA Receptor Expression and Function in Direction-Selective Circuits of the Mammalian Retina.

AMPA receptors (AMPARs) are the major mediators of fast excitatory neurotransmission in the retina as in other parts of the brain. In most neurons, the synaptic targeting, pharmacology, and function of AMPARs are influenced by auxiliary subunits including the transmembrane AMPA receptor regulatory proteins (TARPs). However, it is unclear which TARP subunits are present at retinal synapses and how they influence receptor localization and function. Here, we show that TARPɣ2 (stargazin) is associated with AMPARs in the synaptic layers of the mouse, rabbit, macaque, and human retina. In most species, TARPɣ2 expression was high where starburst amacrine cells (SACs) ramify and transcriptomic analyses suggest correspondingly high gene expression in mouse and human SACs. Synaptic expression of GluA2, GluA3, and GluA4 was significantly reduced in a mouse mutant lacking TARPɣ2 expression (stargazer mouse; stg), whereas GluA1 levels were unaffected. AMPAR-mediated light-evoked EPSCs in ON-SACs from stg mice were ∼30% smaller compared with heterozygous littermates. There was also loss of a transient ON pathway-driven GABAergic input to ON-SACs in stg mutants. Direction-selective ganglion cells in the stg mouse showed normal directional tuning, but their surround inhibition and thus spatial tuning was reduced. Our results indicate that TARPɣ2 is required for normal synaptic expression of GluA2, GluA3, and GluA4 in the inner retina. The presence of residual AMPAR expression in the stargazer mutant suggests that other TARP subunits may compensate in the absence of TARPɣ2.

Hierarchical retinal computations rely on hybrid chemical-electrical signaling.

Bipolar cells (BCs) are integral to the retinal circuits that extract diverse features from the visual environment. They bridge photoreceptors to ganglion cells, the source of retinal output. Understanding how such circuits encode visual features requires an accounting of the mechanisms that control glutamate release from bipolar cell axons. Here, we demonstrate orientation selectivity in a specific genetically identifiable type of mouse bipolar cell-type 5A (BC5A). Their synaptic terminals respond best when stimulated with vertical bars that are far larger than their dendritic fields. We provide evidence that this selectivity involves enhanced excitation for vertical stimuli that requires gap junctional coupling through connexin36. We also show that this orientation selectivity is detectable postsynaptically in direction-selective ganglion cells, which were not previously thought to be selective for orientation. Together, these results demonstrate how multiple features are extracted by a single hierarchical network, engaging distinct electrical and chemical synaptic pathways.

Direction selectivity of the retinotectal system of fish: Findings based on microelectrode extracellular recordings of the tectum opticum

Vision in fish plays an important role in different forms of visually guided behavior. The visual system of fish is available for research by different methods; it is a convenient experimental model for studying and understanding the mechanisms of vision in general. Responses of retinal direction-selective (DS) ganglion cells (GCs) are recorded extracellularly from their axon terminals in the superficial layers of the tectum opticum (TO). They can be divided into three distinct groups according to the preferred directions of stimulus movement: caudorostral, dorsoventral and ventrodorsal. Each of these groups comprises both ON and OFF units in equal proportions. Relatively small receptive fields (3-8?) and fine spatial resolution characterize retinal DS units as local motion detectors. Conversely, the responses of direction-selective tectal neurons (DS TNs) are recorded at two different tectal levels, deeper than the zone of retinal DS afferents. They are characterized by large receptive fields (up to 60?) and are indifferent to any sign of contrast, i.e., they can be considered as ON-OFF-type units. Four types of ON-OFF DS TNs preferring different directions of motion have been recorded. The preferred directions of three types of DS TNs match the preferred directions of three types of DS GCs. Matching the three preferred directions of ON and OFF DS GCs and ON-OFF DS TNs has allowed us to hypothesize that the GCs with caudorostral, ventrodorsal and dorsoventral preferences are input neurons for the corresponding types of DS TNs. On the other hand, the rostrocaudal preference in the fourth type of DS TNs, recorded exclusively in the deep tectal zone, is an emergent property of the TO. In this review, our findings are compared with the results of other authors examining direction selectivity in the fish retinotectal system.

A Novel Bio-Inspired Motion Direction Detection Mechanism in Binary and Grayscale Background

The visual system plays a vital role in the daily life of humans, as more than 90 percent of the external information received by the human brain throughout the day comes from the visual system. However, how the human brain processes the received visual information remains a mystery. The information received from the external through the visual system can be divided into three main categories, namely, shape features, color features, and motion features. Of these, motion features are considered the key to deciphering the secrets of the visual system due to their independence and importance. In this paper, we propose a novel bio-inspired motion direction detection mechanism using direction-selective ganglion cells to explore the mystery of motion information extraction and analysis. The mechanism proposed in this paper is divided into two parts: local motion direction detection neurons and global motion direction detection neurons; the former is used to extract motion direction information from the local area, while the latter infers global motion direction from the local motion direction information. This mechanism is more consistent with the biological perception of the human natural visual system than the previously proposed model and has a higher biological plausibility and greater versatility. It is worth mentioning that we have overcome the problem in which the previous motion direction detection model could only be applied in the binary background by introducing the horizontal cells. Through the association formed by horizontal cells and bipolar cells, this model can be applied to recognizing problems of motion direction detection on a grayscale background. To further validate the effectiveness of the proposed model, a series of experiments with objects of different sizes, shapes, and positions are conducted by computer simulation. According to the simulation results, this model has been proven to have high accuracy rates regardless of objects’ sizes, shapes, and positions in all experiments. Furthermore, the proposed model is verified to own more stable accuracy rates and stronger noise immunity by comparing it with the recognized superior classical convolutional neural network in a background of different percentage noise.

The transcription factor Tbx5 regulates direction-selective retinal ganglion cell development and image stabilization

The diversity of visual input processed by the mammalian visual system requires the generation of many distinct retinal ganglion cell (RGC) types, each tuned to a particular feature. The molecular code needed to generate this cell-type diversity is poorly understood. Here, we focus on the molecules needed to specify one type of retinal cell: the upward-preferring ON direction-selective ganglion cell (up-oDSGC) of the mouse visual system. Single-cell transcriptomic profiling of up- and down-oDSGCs shows that the transcription factor Tbx5 is selectively expressed in up-oDSGCs. The loss of Tbx5 in up-oDSGCs results in a selective defect in the formation of up-oDSGCs and a corresponding inability to detect vertical motion. A downstream effector of Tbx5, Sfrp1, is also critical for vertical motion detection but not up-oDSGC formation. These results advance our understanding of the molecular mechanisms that specify a rare retinal cell type and show how disrupting this specification leads to a corresponding defect in neural circuitry and behavior.

A Novel Artificial Visual System for Motion Direction Detection in Grayscale Images

How specific features of the environment are represented in the mammalian brain is an important unexplained mystery in neuroscience. Visual information is considered to be captured most preferentially by the brain. As one of the visual information elements, motion direction in the receptive field is thought to be collected already at the retinal direction-selective ganglion cell (DSGC) layer. However, knowledge of direction-selective (DS) mechanisms in the retina has remained only at a cellular level, and there is a lack of complete direction-sensitivity understanding in the visual system. Previous studies of DS models have been limited to the stage of one-dimensional black-and-white (binary) images or still lack biological rationality. In this paper, we innovatively propose a two-dimensional, eight-directional motion direction detection mechanism for grayscale images called the artificial visual system (AVS). The structure and neuronal functions of this mechanism are highly faithful to neuroscientific perceptions of the mammalian retinal DS pathway, and thus highly biologically reasonable. In particular, by introducing the horizontal contact pathway provided by horizontal cells (HCs) in the retinal inner nuclear layer and forming a functional collaboration with bipolar cells (BCs), the limitation that previous DS models can only recognize object motion directions in binary images is overcome; the proposed model can solve the recognizing problem of object motion directions in grayscale images. Through computer simulation experiments, we verified that AVS is effective and has high detection accuracy, and it is not affected by the shape, size, and location of objects in the receptive field. Its excellent noise immunity was also verified by adding multiple types of noise to the experimental data set. Compared to a classical convolutional neural network (CNN), it was verified that AVS is completely significantly better in terms of effectiveness and noise immunity, and has various advantages such as high interpretability, no need for learning, and easy hardware implementation. In addition, activation characteristics of neurons in AVS are highly consistent with those real in the retinal DS pathway, with strong neurofunctional similarity and brain-like superiority. Moreover, AVS will also provide a novel perspective and approach to understanding and analyzing mechanisms as well as principles of mammalian retinal direction-sensitivity in face of a cognitive bottleneck on the DS pathway that has persisted for nearly 60 years.

Functional convergence of on-off direction-selective ganglion cells in the visual thalamus

In the mouse visual system, multiple types of retinal ganglion cells (RGCs) each encode distinct features of the visual space. A clear understanding of how this information is parsed in their downstream target, the dorsal lateral geniculate nucleus (dLGN), remains elusive. Here, we characterized retinogeniculate connectivity in Cart-IRES2-Cre-D and BD-CreER2 mice, which labels subsets of on-off direction-selective ganglion cells (ooDSGCs) tuned to the vertical directions and to only ventral motion, respectively. Our immunohistochemical, electrophysiological, and optogenetic experiments reveal that only a small fraction (<15%) of thalamocortical (TC) neurons in the dLGN receives primary retinal drive from these subtypes of ooDSGCs. The majority of the functionally identifiable ooDSGC inputs in the dLGN are weak and converge together with inputs from other RGC types. Yet our modeling indicates that this mixing is not random: BD-CreER+ ooDSGC inputs converge less frequently with ooDSGCs tuned to the opposite direction than with non-CART-Cre+ RGC types. Taken together, these results indicate that convergence of distinct information lines in dLGN follows specific rules of organization.

Sodium Hyaluronate-Induced Ocular Hypertension in Rats Damages the Direction-Selective Circuit and Inner/Outer Retinal Plexiform Layers.

PurposeTo assess the changes in retinal morphology in a rat model of chronic glaucoma induced by ocular hypertension.MethodsIntraocular pressure (IOP) was surgically increased through weekly injections of sodium hyaluronate (HYA) in the anterior eye chamber of the left eye of male Wistar rats, whereas the right eyes were sham operated (salt solution). During the 10-week experimental period, IOP was measured weekly with a rebound tonometer. Retinal cryosections were prepared for histological/immunohistochemical analysis and morphometry.ResultsIOP was higher in HYA-treated eyes than in sham-operated eyes along the 10-week period, which was significant from the fourth to the nineth week. Ocular hypertension in HYA-treated eyes was associated with morphologic and morphometric changes in bipolar cells, ON-OFF direction-selective ganglion cells, ON/OFF starburst amacrine cells, and inner plexiform layer sublamina.ConclusionsSerial HYA treatment in the rat anterior eye chamber results in mild-to-moderate elevated and sustained IOP and ganglion cell death, which mimics most human open-angle glaucoma hallmarks. The reduced number of direction-selective ganglion cells and starburst amacrine cells accompanied by a deteriorated ON/OFF plexus in this glaucoma model could lend insight to the abnormalities in motion perception observed in patients with glaucoma.

Visual Stimulation Induces Distinct Forms of Sensitization of On-Off Direction-Selective Ganglion Cell Responses in the Dorsal and Ventral Retina.

Experience-dependent modulation of neuronal responses is a key attribute in sensory processing. In the mammalian retina, the On-Off direction-selective ganglion cell (DSGC) is well known for its robust direction selectivity. However, how the On-Off DSGC light responsiveness dynamically adjusts to the changing visual environment is underexplored. Here, we report that On-Off DSGCs tuned to posterior motion direction [i.e. posterior DSGCs (pDSGCs)] in mice of both sexes can be transiently sensitized by prior stimuli. Notably, distinct sensitization patterns are found in dorsal and ventral pDSGCs. Although responses of both dorsal and ventral pDSGCs to dark stimuli (Off responses) are sensitized, only dorsal cells show the sensitization of responses to bright stimuli (On responses). Visual stimulation to the dorsal retina potentiates a sustained excitatory input from Off bipolar cells, leading to tonic depolarization of pDSGCs. Such tonic depolarization propagates from the Off to the On dendritic arbor of the pDSGC to sensitize its On response. We also identified a previously overlooked feature of DSGC dendritic architecture that can support dendritic integration between On and Off dendritic layers bypassing the soma. By contrast, ventral pDSGCs lack a sensitized tonic depolarization and thus do not exhibit sensitization of their On responses. Our results highlight a topographic difference in Off bipolar cell inputs underlying divergent sensitization patterns of dorsal and ventral pDSGCs. Moreover, substantial crossovers between dendritic layers of On-Off DSGCs suggest an interactive dendritic algorithm for processing On and Off signals before they reach the soma.SIGNIFICANCE STATEMENT Visual neuronal responses are dynamically influenced by the prior visual experience. This form of plasticity reflects the efficient coding of the naturalistic environment by the visual system. We found that a class of retinal output neurons, On-Off direction-selective ganglion cells, transiently increase their responsiveness after visual stimulation. Cells located in dorsal and ventral retinas exhibit distinct sensitization patterns because of different adaptive properties of Off bipolar cell signaling. A previously overlooked dendritic morphologic feature of the On-Off direction-selective ganglion cell is implicated in the cross talk between On and Off pathways during sensitization. Together, these findings uncover a topographic difference in the adaptive encoding of upper and lower visual fields and the underlying neural mechanism in the dorsal and ventral retinas.

Distinct inhibitory pathways control velocity and directional tuning in the mouse retina

The sensory periphery is responsible for detecting ethologically relevant features of the external world, using compact, predominantly feedforward circuits. Visual motion is a particularly prevalent sensory feature, the presence of which can be a signal to enact diverse behaviors ranging from gaze stabilization reflexes to predator avoidance or prey capture. To understand how the retina constructs the distinct neural representations required for these behaviors, we investigated two circuits responsible for encoding different aspects of image motion: ON and ON-OFF direction-selective ganglion cells (DSGCs). Using a combination of two-photon targeted whole-cell electrophysiology, pharmacology, and conditional knockout mice, we show that distinct inhibitory pathways independently control tuning for motion velocity and motion direction in these two cell types. We further employ dynamic clamp and numerical modeling techniques to show that asymmetric inhibition provides a velocity-invariant mechanism of directional tuning, despite the strong velocity dependence of classical models of direction selectivity. We therefore demonstrate that invariant representations of motion features by inhibitory interneurons act as computational building blocks to construct distinct, behaviorally relevant signals at the earliest stages of the visual system.

Emergence of Direction-Selective Retinal Cell Types in Task-Optimized Deep Learning Models.

Convolutional neural networks (CNNs), a class of deep learning models, have experienced recent success in modeling sensory cortices and retinal circuits through optimizing performance on machine learning tasks, otherwise known as task optimization. Previous research has shown task-optimized CNNs to be capable of providing explanations as to why the retina efficiently encodes natural stimuli and how certain retinal cell types are involved in efficient encoding. In our work, we sought to use task-optimized CNNs as a means of explaining computational mechanisms responsible for motion-selective retinal circuits. We designed a biologically constrained CNN and optimized its performance on a motion-classification task. We drew inspiration from psychophysics, deep learning, and systems neuroscience literature to develop a toolbox of methods to reverse engineer the computational mechanisms learned in our model. Through reverse engineering our model, we proposed a computational mechanism in which direction-selective ganglion cells and starburst amacrine cells, both experimentally observed retinal cell types, emerge in our model to discriminate among moving stimuli. This emergence suggests that direction-selective circuits in the retina are ecologically designed to robustly discriminate among moving stimuli. Our results and methods also provide a framework for how to build more interpretable deep learning models and how to understand them.

Parallel processing in active dendrites during periods of intense spiking activity.

A neuron's ability to perform parallel computations throughout its dendritic arbor substantially improves its computational capacity. However, during natural patterns of activity, the degree to which computations remain compartmentalized, especially in neurons with active dendritic trees, is not clear. Here, we examine how the direction of moving objects is computed across the bistratified dendritic arbors of ON-OFF direction-selective ganglion cells (DSGCs) in the mouse retina. We find that although local synaptic signals propagate efficiently throughout their dendritic trees, direction-selective computations in one part of the dendritic arbor have little effect on those being made elsewhere. Independent dendritic processing allows DSGCs to compute the direction of moving objects multiple times as they traverse their receptive fields, enabling them to rapidly detect changes in motion direction on a sub-receptive-field basis. These results demonstrate that the parallel processing capacity of neurons can be maintained even during periods of intense synaptic activity.