Robert G. Smith - US grants

The long term goal is to determine how neural circuits in mammalian retina solve problems of signal processing. The present project concerns the circuit for night vision which is well defined. The input range spans 3-4 log units. At the low end each quantal event evokes a burst of 2-3 spikes in a ganglion cell (up to 20 events/second); above this level gain is controlled and varies inversely with mean luminance. The circuit's feedforward structure is known (1500 rods -> 100 rod bipolar -> 5 AII amacrine -> 4 b1 bipolar -> beta ganglion cell), and so are three of its feedback loops. This project addresses two questions: 1) By what mechanism does the circuit protect a quantal signal against noise? Lacking such a mechanism, the continuous dark noise from 1500 rods would tend to accumulate in the ganglion cell (as 1500) and obliterate the tiny signal. Noise might be removed by "thresholding" mechanisms at the first two stages of the circuit (where most convergence occurs). Candidate neurons to accomplish thresholding are, respectively, the rod horizontal cell and the A17 amacrine cell. 2) What is the mechanism for gain control? Candidate neurons for gain control are the rod horizontal and the interplexiform cells. To investigate these questions the project will: 1) Gather additional structural data regarding the feedback loops (measure fine features of the horizontal cell, quantitate gap junctions between AII amacrine cells, determine synaptic connections of the interplexiform cell). 2) Construct a compartmental model of each stage of the circuit (constrained by known structure and physiology). A model includes on the order of 103 neurons (104 compartments) and is constructed using a high-level language (based on "C") invented for this purpose. 3) Simulate the response of each stage at different light intensities to explore the dynamics and determine whether the mechanisms proposed for noise removal and gain control are plausible. 4) Simulate the overall circuit (constrained by results from individual stages and the known physiology of the ganglion cell) to explore whether the models of separate stages are compatible. Simulation of this multi-stage circuit, plus its several layers of feedback, should advance basic knowledge regarding the mechanisms of night vision, possibly identifying which sites in the circuit are most vulnerable to deterioration. Simulation should also help understand mechanisms that maintain stability (i.e., oppose seizures) in complex neural circuits. Also, simulating a realistic circuit with 103 neurons provides a start toward the larger scale simulations that will ultimately be needed to understand the brain.

My long term goal is to discover how neural circuits solve problems of signal processing. The key problems are that optical images are noisy (because natural scenes have low contrast) and that retinal circuits are also noisy (because they employ Poisson processes with relatively small numbers: few transmitter quanta, few channels). General strategies to improve and protect signal/noise ratio (SNR) are known, such as signal averaging, bandwidth compression, and gain control. But how these are implemented in specific neural circuits is not established. I propose to study circuits to the beta (x) ganglion cell in cat and the midget (P) ganglion cell in monkey. These cell types are critical to fine spatial vision, contributing respectively 50% and 90% of axons in the optic nerve. The "schematic wiring diagrams" for these circuits are virtually complete, including numbers of converging rods, cones, and bipolar cells, number Of synapses at each stage, sites of electrical coupling, sites of lateral connectivity, identity of neural transmitters and postsynaptic receptors. The responses of most individual neuron types are known, including for some the signal and noise amplitudes. Finally, computational models (compartmental) of several components of the overall circuits have been established (cone-horizontal cell, bipolar-ganglion cell) and "tuned" to reproduce the known responses such as receptive field extent and amplitude. These are large-scale models (up to 50,000 compartments) governed by an established simulator (NeuronC). I propose to use the existing models, extending them where necessary, and simulate different levels of photon, synaptic, and channel noise. I will evaluate the respective contributions of these noise sources for each stage in the circuits and the effects of specific circuit features in improving/maintaining SNR. Specifically, I will: l) compare the SNR at the beta cell from rod signals transmitted via rod bipolar in high photon noise (starlight) to that from rod signals transmitted by coupling to the cone bipolar circuit in moderate photon noise (twilight). 2) compare the SNR at bipolar cell input stage when controlled by feedforward or/and feedback inhibition. 3) compare SNRs of bipolar cell output stage when controlled by inhibitory amacrine synapses where the bipolar and amacrine inputs to ganglion cell are uncorrelated or correlated (as at "dyad" synapse). 4) evaluate noise in ganglion cell contributed by each circuit component, including the spike generator. 5) perform "sensitivity analysis" to assess which parameters in the circuit have greatest affect on SNR at the retinal output and the costs of improving SNR in terms of speed, reliability, and retinal thickness. The project will help understand how neural circuits contribute to spatial acuity in human vision.

DESCRIPTION (Verbatim from applicant's abstract): We propose to study temporal coding in the circuit to the brisk-sustained ganglion cell that is a crucial link in the visual system's reliability. Performance of the ganglion cell is limited by noise in the visual input and from several sources in the presynaptic circuit, but how the ganglion cell integrates a noisy signal for transmission to the brain is unknown. The standard theory for synaptic integration has been a passive dendritic tree coupled to an integrate-and-fire generator to code the signal as a spike rate. Two new facts imply a different theory: (1) dendrites of the ganglion cell express sodium channels, which amplify synaptic noise and are an integral part of the spike generator setting its spike frequency gain, and (2) the spike generator emphasizes high frequencies and responds to transient stimuli with high temporal precision, consistent with a "timing code" which has a greater efficiency than a rate code. We hypothesize that these observations are interrelated because enhancement of the high frequencies present in the noise of synaptic input may improve the spike generator's temporal precision. We propose to test this hypothesis by applying an "ideal observer" to the responses of a real cell and a model. The ideal observer is a computer program that discriminates between two stimuli using all information in a neural system's response. It calculates how reliably contrasts can be discriminated using a likelihood rule, and is therefore an appropriate method to compare fundamental performance of real cell, model, and human behavior. With data sets recorded from ganglion cells in an intact retina preparation and an existing computational model that includes stochastic voltage-gated channels and synaptic release, we propose to determine how the ganglion cell codes synaptic inputs and what factors limit its precision. We will compare the performance of real cell and model to determine what temporal features are coded most reliably, and we will determine how these features are coded by modifying features (e.g., channel kinetics, spike adaptation, slope and threshold of spike frequency vs. input curve) of the computational model. We will evaluate the contribution of voltage-gated channels to amplification of dendritic PSPs, and what spatial and temporal patterns enhance spike precision. Finally, we will determine the function of excitatory and inhibitory mPSPs, due to correlations among and between them. The project will help understand how neural circuits contribute to a precise temporal code in human vision.

bioimaging /biomedical imaging; computer graphics /printing; computer program /software

[unreadable] DESCRIPTION (provided by applicant): We propose to study the precision of coding in the retina where correlated visual signals are processed before being passed to ganglion cells for transmission to the brain. Performance of retinal circuits is limited by noise because the visual signal has a large (10 log unit) dynamic range but is carried by discrete stochastic events such as vesicle release, channel opening, and spikes. Therefore the retina takes advantage of correlated features of the visual environment such as extended objects, velocity, or direction of motion to code these features with specific circuits, improving their signal/noise ratio. But exactly how retinal circuits accomplish this is unknown. One standard theory is that noise from synaptic release and voltage-gated channels is removed by integrating over an extended time. However, the presence of nonlinearities in retinal circuitry suggests that encoding is more complex. For example, the All amacrine cells and bipolar cells contain voltage-gated channels that may amplify and provide adaptation, and they also contain gap junctions that detect correlated signals and remove noise. The dendrites of many ganglion cells are active and may boost postsynaptic potentials nonlinearly to generate a reliable signal. We hypothesize that these neural elements are poised to specifically amplify fast spatially-correlated signals, creating a coincidence detector that imparts salience to visual signals. We propose to test this hypothesis by applying an ideal observer to the responses of real and model neurons. The ideal observer is a computer program that discriminates using the likelihood rule between the responses to a pair of stimuli to measure the precision with which a neuron signals e.g. motion or contrast. This analysis provides the number of gray levels, a fundamental measure of information capacity. We will record from live bipolar, amacrine, and ganglion cells, construct realistic computer models of these neurons and their circuits, and measure the precision of real neurons and model with the ideal observer. Tracking the precision of transient, sustained, and directional selective visual signals from one layer to the next, we will discover where in the visual pathway information is lost and preserved, and gain a better understanding of how information is coded. This work will help to understand how the eye functions, and this knowledge will help clinical researchers determine what has gone wrong in many types of eye disease such as night blindness and other retinal dystrophies. [unreadable] [unreadable]

Project Summary/Abstract Retinal bipolar cells are the key link between photoreceptors and ganglion cells. One bipolar cell type, the rod bipolar cell, transmits the dim light signal at night, while about 10 types of cone bipolar cells transmit the detailed information of the visual image in daylight. Because the visual image contains information from various features (contrast, spatial, temporal, color, etc.), each cone bipolar type extracts certain features and transmits them optimally. The largest class of bipolar cells, the ON class, conveys positive contrast with responses that are mediated by a transduction cascade. When whole-cell patched, their light responses runs down rapidly. Consequently, information about the physiological properties of different ON cone bipolar cell types is scarce. Recently, a new calcium indicator protein (GCaMP3) was developed, and it can specifically be targeted to ON bipolar cells (under control of mGluR6 promoter) or to the closely connected AII amacrine cells (under control of mGluR1 promoter). We here propose to image this indicator with two-photon microscopy and combined it with electrophysiology to investigate the physiology and visual contribution of these cells. Aim 1 will investigate the rod bipolar cell's adaptation mechanism that critically depends on calcium accumulation to lower the response gain. Retinas will be stimulated with ascending light intensities and calcium signal will be recorded in rod bipolar dendrites and axon terminals. Input-output functions will determine the amount of calcium that causes adaptation. The source of calcium will be determined by either emptying calcium stores, blocking intracellular calcium channels, or blocking TRPM1 transduction channels. Aim 2 will determine the physiological differences among the types of ON cone bipolar cells in two ways. First, the retina will be stimulated with flashing or temporally modulated sinusoidal light with varying intensities, and the calcium responses of different cone bipolar types will be recorded by imaging axon terminals that reside in all ON layers of the inner plexiform layer. Second, an AII cell will be depolarized, and the strength of its coupling to the cone bipolar types will be measured by calcium imaging. In order to reveal the cell type identity of the imaged terminals, at the end of the recording session, dye will be injected into multiple cells with a microelectrode. Aim 3 will measure the dynamics of coupling and noise within the AII network under different light intensities using two complementary methods. First, AII amacrine cells will be infected with channelrhodopsin fused to GFP; an AII cell will be patched with whole cell configuration; channelrhodopsin at various distances from the patched cell will be stimulated; and the resulting voltage in the cell will be recorded. Second, AII amacrine cells will be infected with GCaMP3; current will be injected into a cell that is whole-cell patched; and the resulting calcium response in neighboring AII cells will be measured. These experiments will be repeated after blocking gap junctions and/or Na+ channels. The proposed experiments will greatly facilitate our understanding of retinal circuits and parallel processing and they will help apply this knowledge to efforts in restoring vision.

DESCRIPTION (provided by applicant): We propose to study the precision with which retinal circuits code different types of stimulus features. This will be important for understanding how retinal circuits inform the brain about tasks important for survival such as distinguishing a fleeing animal from one approaching. The retina's ability to signal the visual world is limited by biological mechanisms, because the retinal output to the brain is limited by saturation and noise. To cope with this problem, the retina removes the background level using a variety of adaptation mechanisms, allowing ganglion cell signals to code for fine details. However the inevitable cost of these mechanisms is the addition of synaptic noise to the signal which limits fine details'visibility. Although much is known about circuits presynaptic to ganglion cells, what is lacking beyond important details is an understanding of the rationale behind their signal processing mechanisms. For example, it is unknown how a ganglion cell's presynaptic circuit shapes its neural code, the pattern of response that can inform about a stimulus, nor is it known what signal processing tradeoffs make necessary such circuit mechanisms as pooling of the receptive field center signal by convergence and gap junction coupling, and receptive field surround subtraction from amacrine and horizontal cell feedback. Using in vitro live recordings from characterized ganglion cells and horizontal cells and realistic computational models of them, we propose to test several hypotheses about the role of circuits presynaptic to the ganglion cell in its sensitivity and neural code. We hypothesize that a ganglion cell can simultaneously distinguish several stimuli that differ in contrast, size, or location, because these stimuli are conveyed by different neural codes. We will analyze the neural responses with an ideal observer, a computer program that discriminates using the likelihood rule between the responses to a pair of stimuli in a behaviorally-relevant task to measure the precision with which a neuron signals e.g. contrast or motion, and to measure the neural code. Using the ideal observer to analyze single and multiple recordings from retinal neurons, we will determine the sensitivity and neural code for discriminating multiple visual features. Next, we hypothesize that the retinal signal that relays the background level modulates the maintained synaptic release rate and signal-to-noise ratio (SNR) of the receptive field center, and that these are also modulated by the surround. Using live recordings and models, we will study how the ganglion cell's presynaptic circuitry for center and surround control its SNR. Last, we hypothesize that reciprocal inhibition between starburst amacrine cells generates positive feedback to amplify the directional signal for the direction-selective ganglion cell. Using live recordings and models, we will test the hypothesis that reciprocal synaptic feedback helps the starburst amacrine network maximize sensitivity to direction of motion and reduce noise in the direction- selective ganglion cell. These proposed studies are new and important and will provide knowledge about circuit function relevant to a basic understanding of the brain, its behavior, and clinical testing of disease. PUBLIC HEALTH RELEVANCE: The proposed studies of retinal circuitry will provide new knowledge about how the retina functions to reliably detect features of the visual environment. The use of ideal observer analysis allows comparing the performance of one or several neurons to the performance of a person looking at the same stimulus. The results provided by this method are relevant to public health because it will help scientists and eye doctors to determine the neural circuits that are responsible for vision in health and disease.

DESCRIPTION (provided by applicant): This project proposes to study mechanisms of synaptic processing within specific ganglion cell types in the mammalian retina, both by direct neurophysiological recording and through the use of realistic computer models. Our visual system functions under a wide range of light conditions from night to day, and the retina adapts to prevent saturation, so that the output is largely invariant to changes in the illumination level. The synaptic mechanisms that accomplish adaptation and signal transmission introduce noise, which, coupled with the limited dynamic range of neurons, reduces the fidelity of the visual signal. To cope with this problem, the retina segments the visual world using different types of ganglion cells that each code specific visual features with high fidelity. This project focuses on a specific type of retinal ganglion cell that signals directional motion, called the direction-selective ganglion cell (DSGC). Using a live in-vitro isolated rabbit retina, we will record responses of neurons to light stimuli, and construct computational models of the responses to determine the biophysical mechanisms present. The study comprises three sections. Aim 1 examines the function of the starburst amacrine cell (SBAC), essential for generating the direction selective signal for the DSGC. This aim tests several hypotheses relating to specific biophysical mechanisms intrinsic to the cell, such as voltage-gated channels, that generate its directional output. A realistic computer model of the SBAC will help to determine which mechanisms are present. Aim 2 tests the hypothesis that inhibition between adjacent cells within the network of SBACs is crucial for amplifying directional signals. The experimental results will be used to develop and test a computational model, derived from the results of Aim 1 that contains several SBACs with their network interactions. Aim 3 examines noise and precision in the spiking output of the direction-selective ganglion cell, and will account for its spiking properties using a computational model based on physiological results from all three Aims. The final model will represent a detailed and essentially complete representation of directional signaling in the mammalian retina. Overall, the proposed research will improve our understanding of the complex circuitry of the adult retina; the knowledge gained will inform continuing efforts to develop treatments and visual prosthetic devices that restore vision loss from a range of eye diseases.

DESCRIPTION (provided by applicant): Retinal circuits for local synaptic processing In this application, three widely-known laboratories with complementary expertise that specialize in studying function of retinal circuitry propose to investigate mechanisms of signal processing within On- and Off-bipolar cell types in the mammalian retina. Bipolar cells are essential for the retina because they transmit signals from photoreceptors in the outer retina to amacrine and ganglion cells in the inner retina. They are thought to be non-spiking neurons that relay signals via graded depolarization. However, recently bipolar cells in several species have been reported to generate spike-like calcium transients that may generate bursts of neurotransmitter release similar to those found at most central synapses. In preliminary studies, we showed that mouse bipolar cells also display such spike-like transient events and that independent calcium events can be generated in different axon terminal branches. Using a combination of physiological and computational methods, this project will determine whether the calcium transients in bipolar cell axon terminals represent full- blown spikes, and what computational functions they can perform. Using a live ex vivo mouse retina, we will record responses of bipolar cells and ganglion cells to light stimuli, and construct computational models of the responses to determine the biophysical mechanisms present. The project comprises three specific aims. Aim 1 will determine whether individual bipolar cell axon terminal branches can respond to light independently. This aim will use two-photon calcium imaging in whole-mounted retina to test several hypotheses about the signals that are generated and relayed by graded potentials and spikes in the bipolar cell axon terminal to postsynaptic ganglion cells. Aim 2 will determine the influence of inhibitory feedback to bipolar cells on their spike-like events, using whole-cell patch clamp in retinal slices and whole-mounts. Integrating the results from Aims 1 and 2, Aim 3 will develop realistic computational models that will enable us to test novel hypotheses about local processing functions performed by spikes and graded potentials in bipolar cell terminals. We will develop computer models of identified types of bipolar cell with different axon terminal morphologies. This will allow us to determine the critical parameters for independence of signals in bipolar cell axon terminals. In particular, we will explore the role of inhibitory feedback from amacrine cells in modulating spikes and graded signals in bipolar cell axon terminals, to support signal independence between terminals and control their vesicle release with precise timing. Overall, the project will reveal how mammalian bipolar cells transform the graded potentials they receive from photoreceptors with high fidelity into spike-like transient events in their axon terminals, and how these signals interact with inhibitory feedback to modulate bipolar synaptic output. The proposed research will improve our understanding of the signal processing of the adult retina. As bipolar cells are critical targets for stimulation by visual prostheses and genetic approaches to restoring vision loss from a range of eye diseases, the knowledge gained here will guide the further development of such devices and treatments.

DESCRIPTION (provided by applicant): In this application, an expert in the function of retinal circuitry proposes to investigate mechanisms of signal processing and adaptation within the rod pathway for night vision in the mammalian retina. Rod and cone bipolar cells are essential for night and day vision because they transmit and signals from photoreceptors in the outer retina for processing to the inner retina. Understanding how the rod pathway encodes information is a fundamental problem that applies to all sensory pathways, including cone bipolar pathways and cortical circuits. The rod bipolar makes a synaptic ribbon contact onto the A17 amacrine cell, which then makes a reciprocal inhibitory feedback contact onto the rod bipolar cell. Over the past decade, this synaptic connection has been studied by many laboratories, producing a wealth of biophysical details relevant to its function. However, the neurophysiological detail appears so complex that its functional role is difficult to grasp. Recent studies have discovered several mechanisms in the rod bipolar ribbon synapse that cause it to adapt to the background level and to contrast. However the feedback inhibition from the A17 amacrine cell is not thought to contribute to this adaptation. Several other signal processing mechanisms in the A17 have been identified that regulate its feedback. These mechanisms are fundamental and significant because they are similar to those found in many other neurons in the brain. We hypothesize that at night, a divisive receptive field surround from the A17 amacrine cell regulates synaptic release by the rod bipolar cell to improve its signal quality, and that the A17 regulates the laterl extent of the surround according to the background level. We propose to study the effect of amacrine feedback on the synaptic processing performed by the rod and cone bipolar cells. Using realistic computational models of retinal circuitry, we will delineate the possible roles of feedback and feedforward mechanisms involved at the rod bipolar - A17 reciprocal synapse. In Aim 1, we will develop a detailed model of the presynaptic and postsynaptic biophysical mechanisms excluding the details of morphology. We will examine how the known mechanisms for modulating vesicle release by the rod bipolar ribbon can limit or enhance the information content of its signal. Aim 2 will test the hypothesis that negative feedback to the rod bipolar cel generates a divisive spatial surround that enhances the contrast response to twilight signals. In this aim, we will take the model of feedback from Aim 1 and add details of the fine radiating dendrites of the A17 amacrine, including its voltage-gated sodium and potassium channels. Overall, the proposed research will improve our understanding of the signal processing at different background levels performed by visual pathways of the retina. As the rod and cone pathways are critical for vision, the research will improve our understanding of a range of eye diseases. Because these pathways are critical targets for stimulation by visual prostheses and genetic approaches to restoring vision loss from a range of eye diseases, the knowledge gained here will help to advance the development of such devices and treatments.

Project summary This project proposes to study mechanisms of synaptic processing within specific ganglion cell types in the mammalian retina, both by direct electrophysiological recording and through the use of realistic computer models. Retinal ganglion cells are the output neurons of the retina. There are around 20-30 types of retinal ganglion cell in a typical mammalian retina, with each type optimized to detect different features in the visual scene. Each retinal ganglion cell type is present as an orderly array of cells that cover the entire retina, and therefore the concerted activity of each ganglion cell type represents a separate version of the visual image. Thus, the brain simultaneously receives 20-30 distinct images that it combines within central visual areas to produce a continuous, coherent model of the visual world. This project focuses on a specific type of retinal ganglion cell that signals direction of motion, called the direction-selective ganglion cell (DSGC). Recordings from DSGCs from in-vitro isolated retina preparations in mouse and rabbit will be used to characterize the electrical and morphological properties of these cells. Realistic computational models of the neural circuitry will be generated based on this information. Recapitulation of the real responses by the model system will be used to test our understanding of the underlying neural circuits. The study comprises three sections. Aim 1 examines the function of the starburst amacrine cell (SBAC), an interneuron that is the source of direction selective inhibitory inputs to the DSGC. This aim tests the hypothesis that several mechanisms in SBAC dendrites generate and amplify the direction-selective inhibitory signal. The computer model of the SBAC circuit will include sodium and calcium channels in a network of SBACs that are interconnected by reciprocal inhibition. Aim 2 tests the hypothesis that surround inhibition modulates the strength and spatial and temporal resolution of the synaptic input to the DSGC. The experimental results will be used to extend the computer model to take into account the inhibitory inputs from amacrine cells that integrate information over larger regions of the visual scene surrounding the DSGC. Aim 3 examines the reliability in the spiking output of the DSGC, and how this is affected by the presence of ambiguities and noise in the visual input. The data obtained will be used to further develop the computer model to simulate DSGC spike responses. The final model, based on physiological results from all three Aims, will represent a detailed, and essentially complete representation of the neural mechanisms involved in directional signalling in the mammalian retina. If successful, the model should reproduce realistic spiking output for any visual stimulus. Overall, the proposed research will improve our understanding of the complex circuitry of the adult retina; the knowledge gained will inform continuing efforts to develop treatments and visual prosthetic devices that restore vision loss from a range of eye diseases.