Filter, Detector, Predictor: The expanding repertoire of retinal computation in vertebrates.

Retinal anatomy: Comparison of primate (including human) and mouse retina.

Cell types and cell circuits in human and non-human primate retina.

The primate retina communicates visual information to the brain via a set of parallel pathways that originate from at least 22 anatomically distinct types of retinal ganglion cells. Knowledge of the physiological properties of these ganglion cell types is of critical importance for understanding the functioning of the primate visual system. Nonetheless, the physiological properties of only a handful of retinal ganglion cell types have been studied in detail. Here we show, using a newly developed multielectrode array system for the large-scale recording of neural activity, the existence of a physiologically distinct population of ganglion cells in the primate retina with distinctive visual response properties. These cells, which we will refer to as upsilon cells, are characterized by large receptive fields, rapid and transient responses to light, and significant nonlinearities in their spatial summation. Based on the measured properties of these cells, we speculate that they correspond to the smooth/large radiate cells recently identified morphologically in the primate retina and may therefore provide visual input to both the lateral geniculate nucleus and the superior colliculus. We further speculate that the upsilon cells may be the primate retina's counterparts of the Y-cells observed in the cat and other mammalian species.

Visual information is conveyed to the brain by the retinal ganglion cells. Midget ganglion cells serve fine spatial vision by summing excitation from a receptive field 'centre', receiving input from a single cone in the central retina, with lateral inhibition from a receptive field 'surround', receiving input from many surrounding cones. Midget ganglion cells are also thought to serve colour opponent vision because the centre excitation is from a cone of one spectral type, while the surround inhibition is from cones of the other type. The two major cone types, middle(M)- and long-(L)wavelength sensitive, are equally numerous and randomly distributed in the primate central retina, so a spectrally homogeneous surround requires that the cells mediating lateral interactions (horizontal or amacrine cells) receive selective input from only one cone type. Horizontal cells cannot do this because they receive input indiscriminately from M and L cones. Here we report that the amacrine cells connected to midget ganglion cells are similarly indiscriminate. The absence of spectral specificity in the inhibitory wiring raises doubt about the involvement of midget ganglion cells in colour vision and suggest that colour opponency may instead be conveyed by a different type of ganglion cell.

Evidence that Circuits for Spatial and Color Vision Segregate at the First Retinal Synapse

In this paper, we extend our previous light microscopic (LM) study of substance P (SP)-containing amacrine and ganglion cell types of the human retina (Cuenca et al. [1995] J. Comp. Neurol. 356:491-504) to an electron microscopic (EM) and confocal-imaging study in order to reveal synaptic circuitry and putative input and output neurons. SP-immunoreactive (-IR) amacrine cells in primate retina are typically wide-field cells with large cell bodies occurring in normal or displaced positions relative to the inner plexiform layer (IPL). Their main dendrites bear many spines and are monostratified in stratum 3 (S3) of the IPL. Axon-like processes arise from dendrites close to the cell body and run for hundreds of microns at the same level as the dendrites, thus forming a relatively dense plexus in S3 of the IPL. SP-IR axon processes also climb to S1 to surround some amacrine cell bodies, and others pass into the outer plexiform layer (OPL). Still other axons run down to the ganglion cell layer, where they encircle SP-IR ganglion cells and pass on to end in the nerve fiber layer. The SP-IR ganglion cell types have large cell bodies (20-22 microm diameter) and dendrites that costratify in S3 among the SP-IR amacrine cell processes. Double immunostaining and study by confocal microscopy reveals that SP-IR amacrine cells in the monkey colocalize gamma-aminobutyric acid (GABA). Their main plexus of dendrites in S3 of the IPL is skirted on the S2/S3 border by cone bipolar axons that stain for calbindin but intermingles primarily with glycinergic bipolar cell types of S3 and S3-S4. Strongly GABA-IR/weakly glycine-IR amacrine cell bodies, in addition to the SP-IR large-bodied ganglion cell type, are targets of encircling SP-IR axon processes. EM study of the human SP-IR amacrine cell indicates that input synapses to its dendrites are from bipolar cell axons of the S2/S3 border, S3, and the S3/S4 border of the IPL neuropil (33% of the synaptic input) and from amacrine cell processes (67% of the synaptic input). The input amacrine cells are of at least two distinct types based on cytological criteria. Synaptic output from the SP-IR amacrine cell dendrites is to bipolar cell axons as reciprocal synapses (31%), to amacrine cells (40%), and to ganglion cell profiles, primarily in S3 (29%) of the IPL. The SP-IR axons synapse upon SP-IR ganglion cell bodies and axons, upon normally placed and displaced amacrine cell bodies, and upon bipolar cell dendrites in the OPL. In addition, they appear to synapse among themselves. We shall discuss a wiring diagram and the possible role of SP-IR amacrine cells in the primate retina.

The primate central retina is characterized by an avascular fovea and well-defined perifoveal capillary plexus. Neither blood vessels nor their accompanying astrocytes enter the fovea during any stage of retinal development; a balance of angiogenic and angiostatic factors probably maintains foveal avascularity throughout life. The aim of this study was to identify potentially angiorepulsive factors involved in the development of the avascular primate retinal fovea. Retinas of newborn, juvenile, and adult Callithrix jacchus and Macaca fascicularis monkeys and control human retinas were studied to determine the localization of angiostatin relative to III β-tubulin, glial fibrillary acidic protein, vascular endothelial growth factor (VEGF), platelet endothelial cell adhesion molecule-1 (PECAM), and the angiostatin receptor αvβ3-integrin in the foveal, macular, and peripheral retina. Expression studies were performed using immunohistochemistry (IHC) on retinal whole-mount and paraffin sections, and Western blotting on frozen material. The complex network of the main retinal cell types was identified by IHC of retinal whole mounts. In general, lifetime expression of angiostatin was found in all retinas. Colabeling with different markers revealed retinal ganglion cells as the main source of angiostatin expression in the primate retina, whereas PECAM-immunopositive blood capillaries expressed the angiostatin receptor αvβ3-integrin, and capillary-associated astrocytes expressed VEGF. This study provides the first evidence of angiostatin expression in the primate retina; the expression of angiostatin in the avascular foveal region and the peripheral retina suggests that angiostatin may play a role in the regulation of retinal vascularization, providing a possible explanation for the development and persistence of an avascular fovea.

There are many features common to the the retinae of different mammals. There are, for example, often two distinct types of horizontal cell (Cajal, 1933) and there is a distinct class of ganglion cell which can be visualised by neurofibrillar staining (Peichl et al., 1987). Yet it has become increasingly evident over the last few years that the primate retina possesses unique features, which distinguish it from retinae of other mammals such as the cat. One feature is the presence of three distinct cone types. Although other mammals may possess a rudimentary ability to distinguish wavelength differences, simian primates are unique in their well-developed colour vision. How far the primate retina has been remodelled to process spectral as well as spatial composition is uncertain, but interpretation of results from primate retina are coloured by the fact that wavelength as well as intensity is being encoded.

Anatomical and physiological approaches are beginning to reveal the synaptic origins of parallel ON- and OFF-pathway retinal circuits for the transmission of short (S-) wavelength sensitive cone signals in the primate retina. Anatomical data suggest that synaptic output from S-cones is largely segregated; central elements of synaptic triads arise almost exclusively from the "blue-cone" bipolar cell, a presumed ON bipolar, whereas triad-associated contacts derive primarily from the "flat" midget bipolar cell, a hyperpolarizing, OFF bipolar. Similarly, horizontal cell connectivity is also segregated, with only the H2 cell-type receiving numerous contacts from S-cones. Negative feedback from long (L-) and middle (M-) wavelength sensitive cones via the H2 horizontal cells elicits an antagonistic surround in S-cones demonstrating that S versus L + M or "blue-yellow" opponency is first established in the S-cone. However, the S-cone output utilizes distinct synaptic mechanisms to create color opponency at the ganglion cell level. The blue-cone bipolar cell is presynaptic to the small bistratified, "blue-ON" ganglion cell. S versus L + M cone opponency arises postsynaptically by converging S-ON and LM-OFF excitatory bipolar inputs to the ganglion cell's bistratified dendritic tree. The common L + M cone surrounds of the parallel S-ON and LM-OFF cone bipolar inputs appear to cancel resulting in "blue-yellow" antagonism without center-surround spatial opponency. By contrast, in midget ganglion cells, opponency arises by the differential weighting of cone inputs to the receptive field center versus surround. In the macula, the "private-line" connection from a midget ganglion cell to a single cone predicts that S versus L + M opponency is transmitted from the S-cone to the S-OFF midget bipolar and ganglion cell. Beyond the macula, OFF-midget ganglion cell dendritic trees enlarge and collect additional input from multiple L and M cones. Thus S-OFF opponency via the midget pathway would be expected to become more complex in the near retinal periphery as L and/or M and S cone inputs sum to the receptive field center. An important goal for further investigation will be to explore the hypothesis that distinct bistratified S-ON versus midget S-OFF retinal circuits are the substrates for human psychophysical detection mechanisms attributed to S-ON versus S-OFF perceptual channels.

Morphological and electrophysiological properties of dissociated primate retinal cells

Visually guided behaviour at its sensitivity limit relies on single-photon responses originating in a small number of rod photoreceptors. For decades, researchers have debated the neural mechanisms and noise sources that underlie this striking sensitivity. To address this question, we need to understand the constraints arising from the retinal output signals provided by distinct retinal ganglion cell types. It has recently been shown in the primate retina that On and Off parasol ganglion cells, the cell types likely to underlie light detection at the absolute visual threshold, differ fundamentally not only in response polarity, but also in the way they handle single-photon responses originating in rods. The On pathway provides the brain with a thresholded, low-noise readout and the Off pathway with a noisy, linear readout. We outline the mechanistic basis of these different coding strategies and analyse their implications for detecting the weakest light signals. We show that high-fidelity, nonlinear signal processing in the On pathway comes with costs: more single-photon responses are lost and their propagation is delayed compared with the Off pathway. On the other hand, the responses of On ganglion cells allow better intensity discrimination compared with the Off ganglion cell responses near visual threshold.This article is part of the themed issue ‘Vision in dim light’.

The use of in vitro preparations of primate retina provides new perspectives on the mosaic organization and physiological properties of three ganglion cell types that project to the lateral geniculate nucleus: the parasol, midget and small bistratified cells. Dendritic field sizes and coverage for the three types suggest that their relative densities vary with eccentricity. Of the total ganglion cells in the human fovea, midget cells constitute about 90%, parasol cells about 5%, and small bistratified cells about 1%. In the periphery, midget cells make up about 40-45%, parasol cells about 20% and small bistratified cells about 10% of the total. Thus from peripheral to central retina the number of midget ganglion cells progressively increases relative to the parasol and small bistratified types. Physiological properties of these cells have recently been studied in macaque (Macaca nemestrina) retina by combining intracellular recording and dye injection. As expected, parasol cells, projecting to geniculate magnocellular layers, give phasic, non-opponent light responses. Midget cells, which project to geniculate parvocellular layers, show opponent responses sensitive to only mid and long wavelengths; no evidence of short-wavelength-sensitive cone (S-cone) input to any midget ganglion cell has been found. However, the small bistratified cells, which also project to the parvocellular geniculate layers, give a strong blue-ON response to stimuli designed to modulate S-cones. Thus, S-cone and medium- or long-wavelength-sensitive cone opponent signals arise from morphologically distinct ganglion cell types that project in parallel to the lateral geniculate nucleus.

Small bistratified (blue-ON) ganglion cells in the primate retina are involved in processing short wavelength sensitive cone signals. These ganglion cells stratify in both the ON- and OFF-sublamina of the inner plexiform layer. We investigated the origin of synaptic input to the small bistratified ganglion cell in the retina of a New World primate, the marmoset Callithrix jacchus. Two small bistratified cells from peripheral retina were intracellularly filled with Lucifer Yellow, subsequently photoconverted and processed for electron microscopy. Serial ultrathin sections were cut through portions of each cell, and these were analysed in the electron microscope. The majority of synaptic input (about 84%) to both the inner and outer tier of dendrites was from amacrine cells. Both dendritic tiers also received bipolar cell input. These findings are consistent with predictions from physiological studies that synaptic input to the inner and outer tier of small bistratified cells should be excitatory. However, the tiny fraction of total input supplied from bipolar cells to the outer tier is not consistent with the strong excitatory OFF response in cells of this pathway.

The dendritic field size, the distribution of the dendrites relative to the cell body, and the overall shape of the dendritic field of type I ganglion cells in the rat retina were analyzed. These features of neuronal structure were related to the topography of the rat retina. As in the cat, the cell bodies of type I ganglion cells are arranged in a nonrandom mosaic. Previous work has demonstrated that the density of type I cells in the rat retina does not covary with the density of all ganglion cells. Type I dendritic field size varies over the retina; the increase in dendritic field size is accounted for better by the decrease in type I density than by the decrease in overall ganglion cell density. The center of the dendritic field of most type I cells is displaced in the plane of the retina from the cell body. Unlike in carnivore retina (Schall and Leventhal: J. Comp. Neurol. 257:149-159, '87), the dendritic fields in the rat are not displaced down the ganglion cell density gradient. Rather, there is a tendency for the dendritic trees, especially in temporal retina, to be displaced toward dorsal retina. Most of the dendritic fields are elongated, but the degree of elongation is less than that observed in carnivore or primate retina. Unlike in carnivore and primate retina (Leventhal and Schall: J. Comp. Neurol. 220:465-475, '83; Schall et al.: Brain Res. 368:18-23, '86), there is no relationship between dendritic tree orientation and position relative to any point on the retina in the rat.(ABSTRACT TRUNCATED AT 250 WORDS)

Sensory neurons have been hypothesized to efficiently encode signals from the natural environment subject to resource constraints. The predictions of this efficient coding hypothesis regarding the spatial filtering properties of the visual system have been found consistent with human perception, but they have not been compared directly with neural responses. Here, we analyze the information that retinal ganglion cells transmit to the brain about the spatial information in natural images subject to three resource constraints: the number of retinal ganglion cells, their total response variances, and their total synaptic strengths. We derive a model that optimizes the transmitted information and compare it directly with measurements of complete functional connectivity between cone photoreceptors and the four major types of ganglion cells in the primate retina, obtained at single-cell resolution. We find that the ganglion cell population exhibited 80% efficiency in transmitting spatial information relative to the model. Both the retina and the model exhibited high redundancy (~30%) among ganglion cells of the same cell type. A novel and unique prediction of efficient coding, the relationships between projection patterns of individual cones to all ganglion cells, was consistent with the observed projection patterns in the retina. These results indicate a high level of efficiency with near-optimal redundancy in visual signaling by the retina.