Published in: Michael A. Arbib (Ed.): The Handbook of Brain Theory and Neural Networks (second ed). MIT Press, 2002.

Retina

Robert G. Smith

Department of Neuroscience University of Pennsylvania Philadelphia, PA 19104-6058

email: rob<at>retina.anatomy.upenn.edu

INTRODUCTION

At the most basic level, the retina transduces spatial and temporal variations in light intensity and transmits them to the brain. However instead of directly coding intensity, the retina transforms visual signals in a multitude of ways to code properties of the visual world such as contrast, color, and motion. The purpose of this chapter is to develop a conceptual theory to explain why the retina codes visual signals and how the structure of the retina is related to its coding function.

The vertebrate retina reliably responds to light contrast as low as 1% (Shapley and Enroth-Cugell, 1984). Yet as the delicate visual signal is amplified in its passage through the retina, the biological limitations of neural processing add distortion and noise. The ease with which we see fine details in the presence of such biological limitations suggests that one function of retinal circuitry is to maintain the signal's quality by removing redundant signals to enhance signal quality (Laughlin, 1994). This hypothesis predicts that much of the retina's signal coding and structural detail is derived from the need to optimally amplify the signal and eliminate noise.

STRUCTURE

Layers and cell classes

The retina is a thin (100-200 um) tissue at the rear surface of the eye consisting of 3 layers of neurons and glial cells (Figure 1; see Dowling, 1987; Rodieck, 1998; Sterling, 1997). Neurons in the "outer nuclear layer" (ONL) are exclusively photoreceptors. The "inner nuclear layer" (INL) (i.e. the middle layer) contains the cell bodies of horizontal cells (H), bipolar cells (B), and amacrine cells (A). Between these two layers lies the "outer plexiform layer" (OPL) in which bipolar and horizontal cells extend dendritic processes laterally to receive synaptic contacts from photoreceptors. The innermost cell layer, called the "ganglion cell layer" (GCL), contains cell bodies of ganglion cells and amacrine cells. Between the INL and GCL lies the "inner plexiform layer" (IPL) where bipolar, amacrine and ganglion cells are synaptically connected. Ganglion cells send their output to the brain through axons that lie on the inner surface of the retina.


Cell types: specificity in form and function

Each class of neuron described above comprises several cell types, and overall the retina comprises several dozen (Sterling, 1997; Rodieck 1998). A cell type is defined by a distinctive morphology, distribution, synaptic connection pattern, physiology, and/or immunocytochemical staining pattern (Rodieck, 1998). That distinct cell types exist suggests that each has a specific function. Although the retina of one species may contain cell types not present in another, the same 5 retinal cell classes exist in all vertebrate species (Dowling, 1987; Sterling, 1997, Masland, 1988; Rodieck, 1998; Kandel et al., 2000). Therefore all vertebrates likely share similar neural circuit organization.

Receptive fields and connectivity

Neurons of each type are spaced in a regular array across the retina (see Figure 1), so the key to understanding retinal function is to identify the processing strategies of repeating functional circuits or modules (Sterling, 1997). To understand a retinal neuron's physiological function, investigators measure its "receptive field", the region in space and time over which it responds to light. Receptive fields of retinal neurons consist of a sensitive circular region in visual space called the "center", and a larger but weaker antagonistic region concentric with the center, called the "surround" (Rodieck, 1998), which are determined by intrinsic and presynaptic mechanisms. For example, a ganglion cell's receptive field shape and properties reflect its morphology and biophysical properties (Kandel et al., 2000), and also receptive field properties of its presynaptic bipolar and amacrine cells, which in turn originate to some extent in the receptive field properties of photoreceptors and horizontal cells (Dowling, 1987; Sterling, 1997; Rodieck 1998).

While receptive field analysis is a powerful method for studying the function of a neural circuit (see Rodieck, 1998; Shapley and Enroth-Cugell et al, 1984), the origin of a receptive field in a circuit that includes several layers of neurons is difficult to grasp. The difficulty is to separate the effects of cell morphology, synaptic connectivity, and membrane channels on the receptive field (e.g. see RETINAL DIRECTION SELECTIVITY). However, by computationally simulating these biophysical details based even on partial knowledge, it is possible to test specific hypotheses about neural circuit connectivity (Teeters and Arbib, 1991; Smith, 1995).

FUNCTIONAL CIRCUITS

Photoreceptors and adaptation

The outer segment (OS) of a vertebrate photoreceptor transduces light via a multi-step biochemical cascade (Rodieck, 1998; Kandel et al, 2000) into an electrical signal that is conducted through the photoreceptor's axon to its terminal in the OPL. In response to a flash of light, ion channels close, hyperpolarizing the photoreceptor proportionately over a limited range of stimulus intensity. The advantage of this coding function is that a photoreceptor responds well to low contrast signals common in the visual world. The disadvantage is that outside this limited range the photoreceptor responds poorly. At lower intensities, the photoreceptor's transduction gain (i.e. proportion of change in its output signal to a change in light) tends to be insufficient, and at higher intensities, the photoreceptor's response tends to saturate. To solve such saturation problems, the photoreceptor adjusts its gain in a process called "adaptation", which in some species can modulate transduction gain by up to 4 log units.

The two classes of photoreceptors, rods and cones (Rodieck, 1998), differ in that rods are sensitive to single photons and are bleached by daylight, but cones are less sensitive and can regenerate their pigment in daylight (photopic intensity range). At twilight (the mesopic intensity range), rods are coupled via gap junctions to neighboring cones, causing the rod signal to pass directly into cones where it is carried by the lower-gain cone pathway (Rodieck, 1998). At night (scotopic intensity range), a special "rod bipolar" pathway (RB in Figure 1) carries quantal "single photon" signals, removes dark noise, and adapts over an extra 3 log units of intensity (Sterling, 1997, Rodieck, 1998; Smith and van Rossum, 1998).

Outer plexiform layer

The axon terminal of a cone transmits its signal to bipolar cells with a chemical synapse, increasing signal gain at the cost of extra noise and a reduction the intensity range over which the bipolar cell can respond (Laughlin, 1994). To reduce the tendency of the cone signal to saturate its synapse, the OPL filters the signal (Laughlin, 1994). The filter consists of 2 components, a spatial low-pass filter constructed from lateral electrical connections (gap junctions) between cones, and a spatio-temporal high-pass filter constructed by horizonal cells. Cone coupling removes uncorrelated noise from the cone's response, and consequently causes cone synaptic release to be more correlated. Although the coupling also causes "neural blur", it is useful to provide an "anti-aliasing" filter for the next stage of processing in the IPL.

The OPL's high-pass filter is constructed by subtracting a "local average" from the cone. Horizontal cells, also coupled laterally by gap junctions, sum inputs from many cone terminals and provide negative feedback to each via a feedback synapse. The negative feedback mechanism in some cases is a GABA-ergic synapse (Dowling, 1987; Sterling, 1997), but has also recently been postulated to be a form of electrical feedback. The synaptic structure that performs this function, called a "triad", has both feedforward and feedback contacts so it is termed "reciprocal" (Dowling, 1987; Rodieck, 1998). This type of coding has been termed "predictive" (Laughlin, 1994) because the ideal signal to subtract would be a "local average" of signals from neighboring cones (Smith, 1995).

Synaptic function and noise: signal-processing mechanisms

The glutamatergic synapse that transmits a cone's signal to bipolar and horizontal cells adds noise originating in the random fluctuation of synaptic vesicle release rate (Kandel et al., 2000; Sterling, 1997). To reduce the amount of noise relative to the signal, vesicles are released at a high rate by a ribbon that functions as a docking site and reservoir for vesicles (Kandel et al, 2000; Rodieck, 1998; Sterling, 1997). The synapse that relays rod signals to rod-bipolar cells in starlight has a special challenge because noise generated by the rod's transduction cascade and synapse would swamp the tiny single-photon signal. A computer simulation suggested the solution (recently verified by in vitro recordings): a nonlinear threshold in the postsynaptic second-messenger system removes the noise (Smith and van Rossum, 1998).

IPL: bipolar and amacrine circuits

The retina contains about 10 types of bipolar cell and more than 20 types of amacrine cell (Sterling, 1997; Kolb et al, 1981, Masland, 1988; Rodieck, 1998). Bipolar cell dendrites arborize in the OPL to receive multiple synaptic contacts from photoreceptors, and their axons terminate in the IPL. Amacrine cells extend their dendrites laterally in the IPL to contact bipolar, amacrine, and ganglion cells.

Bipolar cells respond as photoreceptors do with a voltage proportional to light intensity, but their response range is narrower and they adapt over a wider range of stimuli. Adaptation occurs at the dendritic tip from changes in gain at a second-messenger biochemical cascade, at the membrane by voltage-gated ion channels, or at the axon where gain of output synapses is regulated in several ways by feedback. Bipolar cells contact ganglion cells with glutamatergic ribbon synapses to allow high release rates and reduce noise. A bipolar cell may contact several ganglion cell types, each with a different characteristic number of synapses, which implies a specific coding of the bipolar signal (Sterling, 1997; Teeters and Arbib, 1991).

Function of amacrine cells

Amacrine cells are a diverse group in both morphology (Kolb et al, 1981; Rodieck, 1998) and neurochemistry (Masland, 1988). Many have a large (0.5-2 mm) but sparse dendritic field with very fine dendritic processes (0.2 um dia) that stretch between small swellings, called "varicosities", where synaptic connections are made (Kolb, et al, 1981; Dowling, 1987). Most amacrine cells contain voltage-gated Na+ channels and fire action potentials which allows them to transmit signals laterally over the extent of their dendritic field (Masland, 1988). Amacrine cells are generally either GABAergic (Kandel et al, 2000; Rodieck, 1998) or glycinergic, which implies that they perform subtractive or shunting control functions. Some, e.g. the cholinergic "starburst" amacrine, are involved in temporal processing, and respond transiently to light (Masland, 1988). Amacrine circuitry is thought to be responsible for accentuating the surround, directional selectivity in ganglion cells (see RETINAL DIRECTION SELECTIVITY), excitatory transient and peripheral effects, and several types of gain control (Shapley and Enroth-Cugell, 1984; Dowling, 1987; Rodieck, 1998).

Amacrine cells receive synaptic contacts from bipolar cells at a "dyad" where a bipolar ribbon synapse contacts two postsynaptic neurons, usually ganglion and amacrine cells (Rodieck, 1998). The similarity between the synaptic "dyad" in the IPL and the "triad" in the OPL is striking (Rodieck, 1998). Both contain synaptic "ribbons", and both include reciprocal feedback from a lateral neuron. The reason may be the identical problem of noise. The reciprocal feedback from an amacrine varicosity to its presynaptic bipolar cell can process the signal reducing its dynamic range before transmission to ganglion cells (Masland, 1988; Dowling, 1987; Rodieck, 1998).

Gap junction coupling in amacrine and bipolar cells

Amacrine and bipolar cells, like many types of neuron in the brain, are widely coupled by gap junctions to their neighbors. Bipolar cell coupling, like cone coupling, correlates neighboring cells' signals to enhance synchronous vesicle release. Since many amacrines fire action potentials, one possibility is that gap junctions allow them to synchronize their firing. But their diversity emphasizes the complexity of retinal circuitry (Masland, 1988; Rodieck, 1998). The AII amacrine cell is small-field and carries rod signals from the rod bipolar to cone bipolars at night (Kolb et al., 1981, Rodieck, 1998). To grasp the function of the AII has been a special challenge because it is coupled by gap junctions to its AII and bipolar cell neighbors, and these two types of electrical coupling are controlled by independently-modulated second-messenger systems. The AII also contains voltage-gated Na+ channels and generates action-potential-like transients. These specialized biophysical properties elegantly solve a signal-processing challenge: in starlight, the AII collects single photon signals from an array of presynaptic rod bipolars, but synaptic noise is collected even when photons are rare. The AII's strategy, therefore, is to reduce noise by electrical coupling with neighbors, and to nonlinearly amplify the single-photon signal with voltage-gated channels (Smith and Vardi, 1995; Sterling, 1997) removing noise and reshaping the signal before passing it on.

DIVERSITY OF CODING

Ganglion cell coding

Ganglion cells have exquisite sensitivity to low contrast stimuli over a wide range of light intensity (Kandel et al, 2000). They are specialized into diverse types that code different properties of the visual world (see RETINAL DIRECTION SELECTIVITY, Kolb et al., 1981; Maturana et al, 1960; Rodieck, 1998). Some give a "tonic" response to stationary stimuli (e.g., the X or W cells of cat retina), and others give a more "phasic" response to signal the presence of flashing or moving stimuli (e.g., the Y cell of cat retina). Many species (lower vertebrates but also mammals) possess ganglion cells with more complex receptive fields, e.g. they respond only to small or large moving objects (Maturana et al, 1960; Teeters and Arbib, 1991). In many species "color-opponent" ganglion cells provide excellent color vision (Rodieck, 1998; Dowling, 1987).

Coding by the spike generator

To transmit a signal to the brain, the ganglion cell codes its intracellular voltage ("the generator potential") as the firing rate of action potentials along its axon (Kandel et al., 2000). Like synaptic coding, this process is limited by noise and dynamic range, so optimal coding of information is at a premium. The ganglion cell's spike generator, consisting of voltage- and ion-gated channels traditionally thought to be located in the axon hillock and soma, is responsible for the coding properties. However, ganglion cells have voltage-gated channels in their dendritic membrane, and recently it was shown by simulation that the dendritic tree must contain these channels at sufficient densities to conduct action potentials, for without them the spike rate becomes too high (Fohlmeister and Miller, 1997). Dendritic morphology and slowly activated K+ channels (Kandel et al., 2000) are also involved in shaping the ganglion cell's response. Simulations have also shown that noise in the spike generator causes variability in the spike rate, and that a significant portion of the information available in the ganglion cells' generator potential is lost in the process.

IPL: specific circuits in sublayers

One problem faced by the spike generator is inherent: it cannot respond well to hyperpolarizations below a certain threshold, and just above threshold, spiking is noisy. To cope with this problem, the retina contains two subclasses of ganglion cell, called "on" and "off", that respond with opposite polarity to a light stimulus. The "on" cell increases its firing rate to a flash of light and the "off" cell reduces its firing rate. Responses of on- and off-ganglion cells in many species are symmetric which allows the retina to code bright and dark objects without much distortion or noise.

To supply on- and off-ganglion cells with appropriate signals, the IPL is organized into on- and off-layers ("sublaminae"). Two bipolar cell subclasses, "on", and "off", respond oppositely to glutamate released by cones. Bipolar and amacrine cell types are divided roughly equally between the two layers, although some arborize in both. The on- and off-layers are in turn organized into specific sub-layers defined by microcircuits comprising bipolar, amacrine, and ganglion cells, each generating a specific spatio-temporal code (Sterling, 1997).

On-bipolar dendrites contain "metabotropic" receptors, which, when bound by glutamate released by a photoreceptor, signal a cytoplasmic second-messenger to turn off the synapse's ionic channels (Sterling, 1997; Dowling, 1987; Rodieck, 1998). Thus an on-bipolar depolarizes when the photoreceptor decreases its glutamate release (i.e. in response to light). An off-bipolar contains "ionotropic" glutamate receptors that directly open an ion channel and hyperpolarize to light. Each off-bipolar type contains glutamate receptors with different kinetic parameters which are the first step in generating a specific temporal code. Some bipolar cells code stimulus velocity, direction, or color (Haverkamp, et. al, 1999; Rodieck, 1998). These specializations increase signal fidelity which is an advantage for a visual signal that is destined to pass through a noisy channel to the brain.

DISCUSSION

Several reasons explain the diversity of retinal circuitry. By discarding part of the information it receives, a neuron specializes in coding specific properties of the signal, i.e. contrast, motion, bright, dark, or colored light flashes, etc. The exact details of the coding scheme are probably related to the ecological niche occupied by the organism. Rod signals because of their quantal nature are qualitatively different from cone signals so there is an advantage in a separate rod pathway. Such specialization in coding increases the signal/noise ratio and makes better use of the limited dynamic range of neurons, synapses, and the spike train in the ganglion cell axon (Laughlin, 1994). Specialization in coding also simplifies the task of brain circuitry in visual "segmentation" (Kandel et al, 2000), which implies a function for retinal circuit structure.

Receptive fields of many retinal neurons, and ganglion cells in particular, share important properties: their center-surround organization, high sensitivity to contrast, and wide-ranging adaptation. To the extent that each retinal circuit amplifies the signal, it adapts to reduce the signal's dynamic range, which implies that the retina's high sensitivity is achieved at the cost of complexity. For example, the net effect of the OPL circuit is to create for the photoreceptor a receptive field with a broad center region and a wide antagonistic surround (Sterling, 1997; Rodieck 1998; Kandel et. al., 2000) that adapts temporally and spatially. By removing information about absolute light intensity, the OPL circuit transmits what is left, i.e. information about contrast.

In turn, the IPL circuit removes more information about light intensity and contrast, shaping the signal in time to code transients, and accentuating the spatial center-surround receptive field in bipolar and amacrine cells (Rodieck, 1998; Dowling, 1987). This process further regulates the visual signal's gain to improve discrimination of low-contrast objects from noise and to prevent saturation at high contrast (Shapley and Enroth-Cugell, 1984). The result of these operations is that retinal receptive fields change with background intensity to maximize information transfer (Laughlin, 1994), and the consequence of this processing is the familiar center and surround of the ganglion cell. Thus it appears that circuits along the retinal pathway all contribute to the ganglion cell's receptive field properties for a similar reason: to prevent noise or saturation from degrading the signal (Laughlin, 1994).

The well-known antagonistic center-surround and adaptation properties of the ganglion cell receptive field, therefore, seem driven by the goal of preserving signal quality. To accomplish this, the circuitry of both OPL and IPL increase the receptive field's lateral extent. But the need for high visual acuity mandates that OPL and IPL circuits not extend laterally too far. Thus the retina is shaped to compensate for biological limitations by a compromise between spatial acuity and accuracy of coding.

Testing the theory

Although knowledge of the biophysical components of retinal circuitry and its receptive fields is progressing at an accelerating pace, such knowledge does not guarantee a useful theory. For example, whole-cell patch recordings allow the biophysical properties and visual responses of bipolar and amacrine cells presynaptic to a ganglion cell to be measured, and these presynaptic responses contribute to the ganglion cell's receptive field. Yet such knowledge alone cannot answer the question of function in design: what function the individual components add to the circuit and therefore why they exist. The answer can only be derived from synthetic models that integrate details of the retina's neural circuitry with the noise and dynamic range limitations inherent to neurobiology.

Computational modeling promises to help find the answers (Teeters and Arbib, 1991; Fohlmeister and Miller, 1997; Smith, 1995; Smith and Vardi, 1995; Haverkamp et al, 1999). Once the basic signal flow and function in a retinal circuit have been established, simulations can help determine overall strategies, and with information theory can find what biological limitations are most serious to the circuit (Laughlin, 1994). The effect of noise on the retina's performance can be tested by simulating noise from all the sources in the signal pathway, and comparing the resulting signal/noise ratios as a measure of signal quality.

ACKNOWLEDGEMENTS

Thanks to V. Ciaramitaro and Drs. M. Freed, K. Morigiwa, N. Vardi and L. Palmer for their comments. Supported by MH48168.

REFERENCES

* = resource reference.

*Dowling, J.E., 1987, The Retina: An Approachable Part of the Brain, Cambridge, MA: Harvard University Press.

Fohlmeister, J.F., and Miller, R.F., 1997, Mechanisms by which cell geometry controls repetitive impulse firing in ganglion cells, J. Neurophysiol., 78: 1948-1964.

*Kandel, E.R., Schwartz, J. H., and T.M. Jessel, 2000, Principles of Neural Science, 4th Ed., New York, NY: McGraw-Hill.

Kolb, H., Nelson, R., and Mariani, A., 1981, Amacrine cells, bipolar cells, and ganglion cells of the cat retina: a Golgi study, Vision Res., 21:1081-1114.

Haverkamp, S., Moeckel, W., and Ammermuller, J., 1999, Different types of synapses with different spectral types of cones underlie color opponency in a bipolar cell of the turtle retina, Vis. Neurosci., 16:801-9, 1999.

*Laughlin, S.B., 1994, Matching coding, circuits, cells, and molecules to signals: General synaptic principles of retinal design in the fly's eye, Prog. in Retinal Eye Res., 13: 165-196.

*Masland R H., 1988, Amacrine cells, Trends Neurosci, 11:405-410.

Maturana, H.R., Lettvin, J.Y., and McCulloch, W.S., 1960. Anatomy and physiology of vision in the frog (rana pipiens), J. Gen. Physiol., 43:129-175.

*Rodieck, R.W., 1998, The First Steps in Seeing, Sunderland MA: Sinauer.

*Shapley, R.M., and Enroth-Cugell, C., 1984, Visual adaptation and retinal gain controls, Progr. in Retinal Res., 3:263-346.

Smith, R.G., 1995, Simulation of an anatomically defined local circuit: The cone-horizontal cell network in cat retina, Vis. Neurosci., 12: 545-561.

Smith, R.G., and van Rossum, M.C.W., 1998, Noise removal at the rod synapse of mammalian retina, Vis. Neurosci., 15: 809-821.

Smith, R.G., and Vardi, N., 1995, Simulation of the AII amacrine cell of mammalian retina: Functional consequences of electrical coupling and regenerative membrane properties, Vis. Neurosci., 12: 851-860.

*Sterling, P. 1997, Retina, in The Synaptic Organization of the Brain, Fourth ed., (Gordon M. Shepherd, ed.), New York: Oxford Univ. Press.

Teeters J. L., and Arbib, M.A., 1991, A model of anuran retina relating interneurons to ganglion cell responses, Biol Cybern., 64:197-207.