Retinal Circuits Research Articles

Fast Retinomorphic Event-Driven Representations for Video Gameplay and Action Recognition

Good temporal representations are crucial for video understanding, and the state-of-the-art video recognition framework is based on two-stream networks. In such framework, besides the regular ConvNets responsible for RGB frame inputs, a second network is introduced to handle the temporal representation, usually the optical flow (OF). However, OF or other task-oriented flow is computationally costly, and is thus typically pre-computed. Critically, this prevents the two-stream approach from being applied to reinforcement learning (RL) applications such as video game playing, where the next state depends on current state and action choices. Inspired by the early vision systems of mammals and insects, we propose a fast event-driven representation (EDR) that models several major properties of early retinal circuits: (1) logarithmic input response, (2) multi-timescale temporal smoothing to filter noise, and (3) bipolar (ON/OFF) pathways for primitive event detection. Trading off the directional information for fast speed ( $>$ 9000 fps), EDR enables fast real-time inference/learning in video applications that require interaction between an agent and the world such as game-playing, virtual robotics, and domain adaptation. In this vein, we use EDR to demonstrate performance improvements over state-of-the-art reinforcement learning algorithms for Atari games, something that has not been possible with pre-computed OF. Moreover, with UCF-101 video action recognition experiments, we show that EDR performs near state-of-the-art in accuracy while achieving a 1,500x speedup in input representation processing, as compared to optical flow.

Retinal Circuits for Seeing in the Dark

How can we relate processing in the retina to an animal's behavior? In this issue of Neuron, Smeds etal. (2019) report that when "every photon counts," mice trade sensitivity for reliability to master visual tasks.

Response Latency Tuning by Retinal Circuits Modulates Signal Efficiency

In the visual system, retinal ganglion cells (RGCs) of various subtypes encode preprocessed photoreceptor signals into a spike output which is then transmitted towards the brain through parallel feature pathways. Spike timing determines how each feature signal contributes to the output of downstream neurons in visual brain centers, thereby influencing efficiency in visual perception. In this study, we demonstrate a marked population-wide variability in RGC response latency that is independent of trial-to-trial variability and recording approach. RGC response latencies to simple visual stimuli vary considerably in a heterogenous cell population but remain reliable when RGCs of a single subtype are compared. This subtype specificity, however, vanishes when the retinal circuitry is bypassed via direct RGC electrical stimulation. This suggests that latency is primarily determined by the signaling speed through retinal pathways that provide subtype specific inputs to RGCs. In addition, response latency is significantly altered when GABA inhibition or gap junction signaling is disturbed, which further supports the key role of retinal microcircuits in latency tuning. Finally, modulation of stimulus parameters affects individual RGC response delays considerably. Based on these findings, we hypothesize that retinal microcircuits fine-tune RGC response latency, which in turn determines the context-dependent weighing of each signal and its contribution to visual perception.

Direct effects of the light environment on daily neuroendocrine control.

Endocrine systems function as key mediators of adaptive responses to the external environment. As a reliable predictor of many salient variations in the external world, the light environment thus constitutes an influential source of control over neuroendocrine function. Accordingly, the vast majority of endocrine systems display 24hr variations in activity that are aligned to daily changes in external illumination. While the neural mechanisms responsible for driving these rhythms are still incompletely understood, circadian and light-dependent signals relayed via the suprachiasmatic nucleus of the hypothalamus (SCN) play a key role. Retinal projections to the SCN provide information from rods, cones and melanopsin, which, together, encode variations in the amount and spectral content of ambient light over the solar day. This sensory input, in turn, drives acute modulations in SCN cellular activity and aligns daily rhythms in the electrophysiological output of individual clock neurons. Neural outputs from the SCN can therefore convey both rapid and longer-term information about the light environment to other hypothalamic nuclei responsible for neuroendocrine control. In this review we summarises current understanding of the specific neural pathways by which the light environment influences key neuroendocrine axes, with a particular focus on the retinal and SCN-dependent circuits involved and their known sensory properties.

Shining Light on the Nervous System: From Biomaterials to Bioelectronics

The dichotomy between advanced materials and brain has driven the curiosity of scientists to explore the wonders of the brain, as well as motivated the continued innovations of novel technologies based on advances in materials science and engineering to understand the brain. To improve treatments of brain-related diseases will require new tools and methods to map and to repair the brain with precision and biocompatibility. Current treatments of pain heavily rely on opioids, resulting in significant side effects such as addiction, tolerance, leading to the Opioid Overdose Crisis as we know of today. Smart drug delivery systems may provide an effective solution. Here I present the development of polymer-based externally-triggerable drug delivery systems for on-demand, repeatable and adjustable local anesthesia, where the timing, duration, and intensity of nerve block can be controlled through external energy triggers such as light. In addition to the new pharmacological approaches, bioelectronic platforms to enhance our insights into the eye and will also be discussed. The restoration of light response with complex spatiotemporal features in retinal degenerative diseases towards retinal prosthesis has proven to be a considerable challenge over the past decades. Herein, inspired by the structure and function of photoreceptors in retinas, I develop artificial retina based on gold nanoparticle-decorated titania nanowire arrays, for restoration of visual responses in the blind mice with degenerated photoreceptors. Green, blue and near UV light responses in the retinal ganglion cells (RGCs) are restored with a spatial resolution better than 100 µm. ON responses in RGCs are blocked by glutamatergic antagonists, suggesting functional preservation of the remaining retinal circuits. Moreover, neurons in the primary visual cortex respond to light after subretinal implant of nanowire arrays. Improvement in pupillary light reflex suggests the behavioral recovery of light sensitivity. My study will shed light on the development of a new generation of optoelectronic toolkits for subretinal prosthetic devices. Through pharmacological, optical, and electrical toolsets, I aim to develop effective therapeutic solutions to neurological disease states. These results, along with a discussion of future neural interfaces, aim to improve our understanding of the nervous system and to inform new therapeutic approaches for biomaterials and bioelectronics.

Activation of Rod Input in a Model of Retinal Degeneration Reverses Retinal Remodeling and Induces Formation of Functional Synapses and Recovery of Visual Signaling in the Adult Retina.

A major cause of human blindness is the death of rod photoreceptors. As rods degenerate, synaptic structures between rod and rod bipolar cells disappear and the rod bipolar cells extend their dendrites and occasionally make aberrant contacts. Such changes are broadly observed in blinding disorders caused by photoreceptor cell death and are thought to occur in response to deafferentation. How the remodeled retinal circuit affects visual processing following rod rescue is not known. To address this question, we generated male and female transgenic mice wherein a disrupted cGMP-gated channel (CNG) gene can be repaired at the endogenous locus and at different stages of degeneration by tamoxifen-inducible cre-mediated recombination. In normal rods, light-induced closure of CNG channels leads to hyperpolarization of the cell, reducing neurotransmitter release at the synapse. Similarly, rods lacking CNG channels exhibit a resting membrane potential that was ~10 mV hyperpolarized compared to WT rods, indicating diminished glutamate release. Retinas from these mice undergo stereotypic retinal remodeling as a consequence of rod malfunction and degeneration. Upon tamoxifen-induced expression of CNG channels, rods recovered their structure and exhibited normal light responses. Moreover, we show that the adult mouse retina displays a surprising degree of plasticity upon activation of rod input. Wayward bipolar cell dendrites establish contact with rods to support normal synaptic transmission, which is propagated to the retinal ganglion cells. These findings demonstrate remarkable plasticity extending beyond the developmental period and support efforts to repair or replace defective rods in patients blinded by rod degeneration.SIGNIFICANCE STATEMENT Current strategies for treatment of neurodegenerative disorders are focused on the repair of the primary affected cell type. However, the defective neurons function within a complex neural circuitry, which also becomes degraded during disease. It is not known whether rescued neurons and the remodeled circuit will establish communication to regain normal function. We show that the adult mammalian neural retina exhibits a surprising degree of plasticity following rescue of rod photoreceptors. The wayward dendrites of rod bipolar cells re-establish contact with rods to support normal synaptic transmission, which is propagated to the retinal ganglion cells. These findings support efforts to repair or replace defective rods in patients blinded by rod cell loss.

Subcutaneous delivery of tauroursodeoxycholic acid rescues the cone photoreceptors in degenerative retina: A promising therapeutic molecule for retinopathy

Inherited retinal degeneration (RD) comprises a heterogeneous group of retinopathies that rank among the main causes of blindness. Tauroursodeoxycholic acid (TUDCA) is taurine conjugate hydrophilic bile acid that demonstrates profound protective effects against a series of neurodegenerative diseases related to oxidative stress. This study sought to evaluate the TUDCA induced effects of on a pharmacologically induced RD animal model by electroretinogram (ERG) examination, behavior tests, morphological analysis and immunochemistry assay. Massive photoreceptor degeneration in mice retina was induced by an intraperitoneal administration of N-methyl-N-nitrosourea(MNU). Subcutaneous delivery of TUDCA inhibits effectively the photoreceptor loss and visual impairments in the MNU administered mice. In the retinal flat-mounts of TUDCA treated mice, the cone photoreceptors were efficiently preserved. Furthermore, the multi-electrodes array (MEA) was used to detect the firing activities of retinal ganglion cells within the inner retinal circuits. TUDCA therapy could restrain the spontaneous firing response, enhance the light induced firing response, and preserve the basic configurations of ON-OFF signal pathway in degenerative retinas. Our MEA assay provided an example to evaluate the potency of pharmacological compounds on retinal plasticity. TUDCA affords these protective effects by modulating apoptosis and alleviating oxidative stress in the degenerative retina. In conclusion, TUDCA therapy can ameliorate the photoreceptor degeneration and rectify the abnormities in visual signal transmission. These findings suggest that TUDCA might act as a potential medication for these retinopathies with progressive photoreceptor degeneration.

Adapting Laplacian based filtering in digital image processing to a retina-inspired analog image processing circuit

In this paper, a unique biologically inspired retina circuit architecture providing Laplacian filter based analog image processing has been suggested. A digital image filtering method is utilized for this aim. Convolution theory and masking technique have an important place among digital image processing methods. These two mathematical operations can be easily done with basic electronic circuit structures. We use current mirrors and current subtractor circuit for the purpose of performing convolution by the use of masking technique on any image. The concept of human retina is able to be mimicked by the help of using silicon circuits. A retina construction can be thought as a group of pixel structures. Because of this reason, we first design a novel pixel circuit as a subcircuit for the retina structure. Our new retina-inspired neuromorphic pixel consists of only 8 MOS transistors. Then, 10 k identical pixel circuits are united together with the help of proper subcircuit connections to achieve a retina structure of size 100 × 100 pixels which enables edge detection feature on images thanks to Laplacian filtering. We compare the analysis results of our grid retina circuit with the theoretical Laplacian filter method used in digital image processing. We obtain analysis results of four different grayscale images that agree well with the expected theoretical results for Laplacian filtering.

The Retinal Basis of Vertebrate Color Vision.

The jawless fish that were ancestral to all living vertebrates had four spectral cone types that were probably served by chromatic-opponent retinal circuits. Subsequent evolution of photoreceptor spectral sensitivities is documented for many vertebrate lineages, giving insight into the ecological adaptation of color vision. Beyond the photoreceptors, retinal color processing is best understood in mammals, especially the blueON system, which opposes short- against long-wavelength receptor responses. For other vertebrates that often have three or four types of cone pigment, new findings from zebrafish are extending older work on teleost fish and reptiles to reveal rich color circuitry. Here, horizontal cells establish diverse and complex spectral responses even in photoreceptor outputs. Cone-selective connections to bipolar cells then set up color-opponent synaptic layers in the inner retina, which lead to a large variety of color-opponent channels for transmission to the brain via retinal ganglion cells.

A simulation study on the effect of ionic currents on transmission from cones to retinal OFF type cone bipolar cells

The retinal cone bipolar cells are interneurons which receive inputs from cone photoreceptors and send outputs to retinal ganglion cells. Several subtypes of bipolar cells have been identified by morphology and electrophysiology in the mammalian retina, which convey distinct visual information to higher order neurons in parallel. The neural circuit in the retina not only converts light information to neural . information, but also performs visual information preprocessing that has not yet been fully understood. Recently, it has been revealed that the neural circuits in retinas of higher vertebrates, such as mammals and primates, have various biophysical properties arising from being composed of ionic channels, ionic pumps, and neurotransmitter receptors. Analysis using a mathematical model based on their ionic mechanisms is essential to understand the visual information processing in the retinal neural circuit of the higher vertebrates.
 The cones and the bipolar cells respond to continuous variation of light with a graded potential, in an analog manner. Especially, glutamate is continuously released from a cone synapse in the dark and is decreased by hyperpolarization of the cone that receives the light stimulus. The alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and kainate type ionotropic glutamate receptors (iGluRs) of the OFF type bipolar cells (OFF-BCs) exhibit partial or nearly complete desensitization in the sustained presence of glutamate. In the dark, glutamate concentration in the synaptic cleft of the cone pedicle rises to 0.1–0.5 mM.5,6 The baseline glutamate concentration depends on a sustained hyperpolarization of the cone by light. Thus, for understanding the working of the OFF-BCs, it is important to elucidate the mechanisms of synaptic transmission from cones to OFF-BCs via iGluRs, which undergo desensitization in the various background light conditions. Furthermore, there are various kinds of ionic channels in OFF-BCs that mediate membrane potential responses. It is considered that information transmitted from cones to OFF-BCs is modulated by the intrinsic ionic currents. We analyzed how ionic currents of OFF-BCs contribute to the transmission of light responses by developing a mathematical model.

Neural mechanisms of contextual modulation in the retinal direction selective circuit

Contextual modulation of neuronal responses by surrounding environments is a fundamental attribute of sensory processing. In the mammalian retina, responses of On–Off direction selective ganglion cells (DSGCs) are modulated by motion contexts. However, the underlying mechanisms are unknown. Here, we show that posterior-preferring DSGCs (pDSGCs) are sensitive to discontinuities of moving contours owing to contextually modulated cholinergic excitation from starburst amacrine cells (SACs). Using a combination of synapse-specific genetic manipulations, patch clamp electrophysiology and connectomic analysis, we identified distinct circuit motifs upstream of On and Off SACs that are required for the contextual modulation of pDSGC activity for bright and dark contrasts. Furthermore, our results reveal a class of wide-field amacrine cells (WACs) with straight, unbranching dendrites that function as “continuity detectors” of moving contours. Therefore, divergent circuit motifs in the On and Off pathways extend the information encoding of On-Off DSGCs beyond their direction selectivity during complex stimuli.

Partial Cone Loss Triggers Synapse-Specific Remodeling and Spatial Receptive Field Rearrangements in a Mature Retinal Circuit

SUMMARYResilience of neural circuits has been observed in the persistence of function despite neuronal loss. In vision, acuity and sensitivity can be retained after 50% loss of cones. While neurons in the cortex can remodel after input loss, the contributions of cell-type-specific circuits to resilience are unknown. Here, we study the effects of partial cone loss in mature mouse retina where cell types and connections are known. At first-order synapses, bipolar cell dendrites remodel and synaptic proteins diminish at sites of input loss. Sites of remaining inputs preserve synaptic proteins. Second-order synapses between bipolar and ganglion cells remain stable. Functionally, ganglion cell spatio-temporal receptive fields retain center-surround structure following partial cone loss. We find evidence for slower temporal filters and expanded receptive field surrounds, derived mainly from inhibitory inputs. Surround expansion is absent in partially stimulated control retina. Results demonstrate functional resilience to input loss beyond pre-existing mechanisms in control retina.

A Retinal Circuit Generating a Dynamic Predictive Code for Oriented Features

SummarySensory systems must reduce the transmission of redundant information to function efficiently. One strategy is to continuously adjust the sensitivity of neurons to suppress responses to common features of the input while enhancing responses to new ones. Here we image the excitatory synaptic inputs and outputs of retinal ganglion cells to understand how such dynamic predictive coding is implemented in the analysis of spatial patterns. Synapses of bipolar cells become tuned to orientation through presynaptic inhibition, generating lateral antagonism in the orientation domain. Individual ganglion cells receive excitatory synapses tuned to different orientations, but feedforward inhibition generates a high-pass filter that only transmits the initial activation of these inputs, removing redundancy. These results demonstrate how a dynamic predictive code can be implemented by circuit motifs common to many parts of the brain.

Bio-Inspired Nanowire Arrays As Artificial Photoreceptors for Retinal Prosthesis

The restoration of light response with complex spatiotemporal features in retinal degenerative diseases towards retinal prosthesis has proven to be a considerable challenge over the past decades. Herein, inspired by the structure and function of photoreceptors in retinas, we develop artificial photoreceptors based on gold nanoparticle-decorated titania nanowire arrays, for restoration of visual responses in the blind mice with degenerated photoreceptors. Green, blue and near UV light responses in the retinal ganglion cells (RGCs) are restored with a spatial resolution better than 100 µm. ON responses in RGCs are blocked by glutamatergic antagonists, suggesting functional preservation of the remaining retinal circuits. Moreover, neurons in the primary visual cortex respond to light after subretinal implant of nanowire arrays. Improvement in pupillary light reflex suggests the behavioral recovery of light sensitivity. Our study will shed light on the development of a new generation of optoelectronic toolkits for subretinal prosthetic devices. Figure 1

Site-specific abnormalities in the visual system of a mouse model of CDKL5 deficiency disorder.

CDKL5 deficiency disorder (CDD) is a neurodevelopmental disorder characterized by a severe global developmental delay and early-onset seizures. Notably, patients show distinctive visual abnormalities often clinically diagnosed as cortical visual impairment. However, the involvement of cerebral cortical dysfunctions in the origin of the symptoms is poorly understood. CDD mouse models also display visual deficits, and cortical visual responses can be used as a robust biomarker in CDKL5 mutant mice. A deeper understanding of the circuits underlying the described visual deficits is essential for directing preclinical research and translational approaches. Here, we addressed this question in two ways: first, we performed an in-depth morphological analysis of the visual pathway, from the retina to the primary visual cortex (V1), of CDKL5 null mice. We found that the lack of CDKL5 produced no alteration in the organization of retinal circuits. Conversely, CDKL5 mutants showed reduced density and altered morphology of spines and decreased excitatory synapse marker PSD95 in the dorsal lateral geniculate nucleus and in V1. An increase in the inhibitory marker VGAT was selectively present in V1. Second, using a conditional CDKL5 knockout model, we showed that selective cortical deletion of CDKL5 from excitatory cells is sufficient to produce abnormalities of visual cortical responses, demonstrating that the normal function of cortical circuits is dependent on CDKL5. Intriguingly, these deficits were associated with morphological alterations of V1 excitatory and inhibitory synaptic contacts. In summary, this work proposes cortical circuit structure and function as a critically important target for studying CDD.

Regulation effects of Ngn2 gene transfection on retinal neuron differentiation in three-dimensional optic vesicle of mice

Objective To investigate the regulation effects of Ngn2 gene transfection on retinal neuron differnetion in three-dimentional optic vesicle (OV) of mice. Methods OV was cultured in vitro using mouse induced pluripotent stem cells (iPSC) under specific conditions.During OV culture, it was transfected multiple times by lentivirus-mediated Ngn2 gene and then it was induced after maturation.The cells were specificly differentiated toward retinal nerve cells in OV.Using the green fluorescent protein (EGFP) gene as control, the differentiation of retinal nerve cells in OV was detected by immunohistochemistry.Reverse transcription PCR and Western blot were used to quantitatively detect the expressions of retinal neuron-specific proteins Pax6, Islet1 and Brn3b. Results The mouse iPS-derived OV was successfully cultured.The number of neural cells in the OV transfected with the Ngn2 gene was increased by the lentiviral-mediated lentivirus.The expressions of PAX6, Islet1 and Brn3b in the Ngn2 transfection group were significantly higher at the gene and protein levels than those in the control group, with significant differences between the two groups (P<0.05). Conclusions The Ngn2 gene can effectively increase the number of retinal neuron differentiation in OV and make in vitro cultured OV more mature and form a more perfect retinal cell neural circuit. Key words: Neuroretina; Ngn2 genes; Optic vesicle; differentiation; Gene transfection

Optimized nonionic emulsifier for the efficient delivery of astaxanthin nanodispersions to retina: in vivo and ex vivo evaluations

Astaxanthin (AST) is a naturally occurring carotenoid with potent anti-oxidative and anti-inflammatory potency against chronic diseases. In this study, we suspended AST in different nonionic emulsifiers to produce nanodispersions. The basic physicochemical properties of the produced AST nanodispersions were verified to select the optimized nonionic emulsifier. Among the tested emulsifiers, Polysorbate 20 produced the AST nanoemulsions with smaller particle diameters, narrower size distributions, and higher AST contents among these emulsifiers. The N-methyl-N-nitrosourea (MNU) administered mouse is a chemically induced retinal degeneration (RD) model with rapid progress rate. AST suspended in Polysorbate 20 was demonstrated to ameliorate the dramatic consequences of MNU on retina architectures and function in several different tests encompassing from electrophysiology to histology and molecular tests. Furthermore, the multi-electrodes array (MEA) was used to detect the firing activities of retinal ganglion cells within the inner retinal circuits. We found that AST nanodispersions could restrain the spontaneous firing response, enhance the light induced firing response, and preserve the basic configurations of visual signal pathway in degenerative retinas. The MEA assay provided an appropriate example to evaluate the potency of pharmacological compounds on retinal plasticity. In summary, emulsifier type affects the basic physicochemical characteristic of AST nanodispersions. Polysorbate 20 acts as an optimized nonionic emulsifier for the efficient delivery of AST nanodispersions to retina. AST nanodispersions can alleviate the photoreceptor loss and rectify the abnormities in visual signal transmission.

Light-dependent pathways for dopaminergic amacrine cell development and function.

Retinal dopamine is a critical modulator of high acuity, light-adapted vision and photoreceptor coupling in the retina. Dopaminergic amacrine cells (DACs) serve as the sole source of retinal dopamine, and dopamine release in the retina follows a circadian rhythm and is modulated by light exposure. However, the retinal circuits through which light influences the development and function of DACs are still unknown. Intrinsically photosensitive retinal ganglion cells (ipRGCs) have emerged as a prime target for influencing retinal dopamine levels because they costratify with DACs in the inner plexiform layer and signal to them in a retrograde manner. Surprisingly, using genetic mouse models lacking specific phototransduction pathways, we find that while light influences the total number of DACs and retinal dopamine levels, this effect does not require ipRGCs. Instead, we find that the rod pathway is a critical modulator of both DAC number and retinal dopamine levels.

Light Prior to Eye Opening Promotes Retinal Waves and Eye-Specific Segregation

Retinal waves are bursts of correlated activity that occur prior to eye opening and provide a critical source of activity that drives the refinement of retinofugal projections. Retinal waves are thought to be initiated spontaneously with their spatiotemporal features dictated by immature neural circuits. Here we demonstrate that, during the second postnatal week in mice, changes in light intensity dictate where and when a subset of retinal waves are triggered via activation of conventional photoreceptors. Propagation properties of triggered waves are indistinguishable from spontaneous waves, indicating that theyare activating the same retinal circuits. Using whole-brain imaging techniques, we demonstrate that light deprivation prior to eye opening diminishes eye-specific segregation of the retinal projections to the dorsolateral geniculate nucleus of the thalamus, but not other retinal targets. These data indicate that light that passes through the closed eyelids plays a critical role in the development of the image-forming visual system.

Light propagation and capture in cone photoreceptors.

The light capturing properties of cone photoreceptors create the elementary signals that form the basis of vision. Variation in the amplitude of individual cone signals has been found physiologically as part of normal retinal circuit processing. Less well characterized is how cone signals may vary due to purely optical properties. We present a model of light propagation in cones using a finite difference beam propagation method to simulate how light from a small stimulus travels through a cone plus its immediate neighbors. The model calculates the amount of light absorbed in the cone outer segments, from which an estimate of the photoresponse can be made. We apply the method to adaptive optics microstimulation to find the optimum optical conditions that will confine the most light into a single cone in the human retina. We found that light capture is especially sensitive to beam size at the pupil and to the cone diameter itself, with the two factors having a complex relationship leading to sizable variation in light capture. Model predictions were validated with two types of psychophysical data. The model can be employed with arbitrary stimuli and photoreceptor parameters, making it a useful tool for studying photoreceptor function in normal or diseased conditions.