A comparison of spatio-temporal receptive fields of ganglion cells in the retinas of the tadpole and adult frog

Receptive fields formed by ganglion cells were measured simultaneously in time and space in the adult and tadpole retinas. The spatio-temporal receptive-fields (STRFs) were measured by cross-correlating the spatio-temporal white-noise stimulus with the cells' spike discharges. Crosscorrelation was made photographically to extract the first order STRF kernel (approximately linear component of the STRFs). The time-course of STRF formed by the frog ganglion cells was twice as fast as that of the tadpole cells. The STRF formed by the tadpole ganglion cells was either a center-brightening or a center-dimming type whereas in the frog there was another class of cells which produced either complex STRFs or did not show any linear component in their STRFs. Size of the STRFs in frog and tadpole was similar in both frog and tadpole retinas.

Spatio-temporal visual receptive fields as revealed by spatio-temporal random noise

A means was devised to visualize the retinal receptive fields in time and space using the noise on unused television channels as spatio-temporal inputs and performing correlation between the input and output photographically. The method was applied to characterize the receptive fields of catfish retinal ganglion cells. The results were 1) there were two major types of receptive fields, circular and elliptical, 2) shapes and sizes of the field components changed with time (latency), and 3) a field's surround was often localized as hot spots.

Spatial organizations of catfish retinal neurons. II. Circular stimulus

1. Temporal dynamics of receptive-field components were identified by use of circular stimuli whose diameter was modulated in white-noise fashion. 2. A thin ring of light evoked a complex response from the horizontal-cell soma but not from its axon. 3. All bipolar cells had a biphasic receptive field whose incremental sensitivities were comparable for the field's center and surround. 4. Type-N cells had a biphasic field whose two components were segregated both in time and space. 5. Type-C cells showed nonlinearities, which did not depend on the types of light inputs and which were intrinsic to the cell. 6. Ganglion cells could be classified roughly into two classes: small- and large-field cells. Their main nonlinearity was rectification. 7. Our studies on the catfish retna have shown that the distal cells are essentially linear in time and space and can be identified functionally by any type of input. The proximal cells, however, have a complex functional repertoire whose identification poses considerable difficulty.

Spatial organization of catfish retinal neurons. I. Single- and random-bar stimulation

1. Receptive-field profiles of catfish (Ictalurus punctatus) retinal neurons were produced by a moving single bar or a moving random grating, which was swept across the cell's receptive field at a constant speed. 2. Bipolar cells form either an on- or an off-center biphasic field and are approximately linear in time and space. 3. Type-C or transient cells form predominantly monotonic receptive fields. We find two subclasses, one slow and the other fast transient cells. They can be identified functionally as well as morphologically. 4. Type-N or sustained cells form a biphasic receptive field, which is revealed by a bar of light. The monotonic field found by a spot or an annulus of light represents activity of the cell's field center. 5. There are two ganglion-cell types, small-field cells and large-field cells. It appears as if small-field cells copy signals in the bipolar cells and large-field cells, signals in the type-N cells. We suggest, however, that this observation represents the limitation imposed by our stimuli rather than an overall functional characteristic of catfish ganglion cells.

Spatiotemporal testing and modeling of catfish retinal neurons

The responses of retinal neurons depend on the interaction of both temporal and spatial aspects of a light stimulus. We developed a linear spatiotemporal model of receptor and horizontal cell layers in the catfish retina based on reciprocal interactions between both layers and coupling within each. Horizontal cell transfer properties were measured experimentally using white-noise intensity modulated light spots of different diameters and were compared with analytical predictions based on the model. Good agreement was obtained with a reasonable choice of model space-constants and feedback parameters. Furthermore, the same set of parameter values determined from spot experiments enabled accurate prediction of experimental horizontal cell responses to traveling gratings. The proposed feedback connections from horizontal cells to receptors quicken the time-course of responses in both layers and sharpen receptive fields.

Functional organization of catfish retina

1. The basic organization of the biphasic (or concentric) receptive field is established in the bipolar cells as the result of an interaction between two signals, one local representing the activity of a small number of receptors, and the other integrating (19, 20) or global (28) coming from the S space or a lamina formed by the horizontal cells (8, 14, 22, 29). 2. Bipolar-ganglion cell pairs are segregated into two types; A (on center) and B (off center) pairs. A depolarization of a bipolar cell produces spike discharges from ganglion cells of the same type and a hyperpolarization depresses their discharges. I haven't detected any cross talk between the types A and B pairs. Bipolar and ganglion cells must be interfaced by the classical chemical synapses, the only such kind in the catfish retina. 3. Horizontal and type N neurons form two lateral transmission systems, one distal and the other proximal (19, 20). Signals in the lateral systems are shared by the two receptive-field types and are not excitatory or inhibitory in themselves; it is incumbent upon the postsynaptic neurons to decide the polarity of the synaptic transmission. The horizontal cell participates directly in the formation of biphasic receptive fields of bipolar cells by providing their surrounding, whereas type N neuron seems to modify the receptive-field organization established in the bipolar cells. 4. Type N neurons are amacrine cells because they do not produce spike discharges (2, 18, 21) and because they influence the activity of both A and B receptive fields. 5. The function of the type C neuron is as unique as its structure (21) and is not fully clear as yet. It is not a conventional amacrine cell as the type N appears to be, nor is it a classical ganglion cell which forms either a type A or B receptive field (2). 6. Type Y neurons are a class of ganglion cells which forms either a type A or B receptive field.

Morphological and functional identifications of catfish retinal neurons. I. Classical morphology

The morphology of the catfish horizontal cells is comparable to that in other fish retinas. The external horizontal cells contact cone receptors and are stellate in shape; the intermediate horizontal cells are even more so and contact rod receptors. The internal horizontal cells constitute the most proximal layer of the inner nuclear layer and may possibly be, in reality, extended processes from the other two horizontal cell types. Bipolar cells resemble those in other teleost retinas: the size and shape of their dendritic tree encompass a continuous spectrum ranging from what is known as the small to the large bipolar cells. The accepted definition of amacrine cells is sufficiently vague to justify our originating a more descriptive and less inferential name for the (axonless) neurons in the inner nuclear layer which radiate processes throughout the inner synaptic layer. These starbust and spaghetti cells vary considerably in the character and extent of their dendritic spread, but correlates exist in other vertebrate retinas. Ganglion cells are found not only in the classical ganglion layer but displaced into the inner nuclear layer as well. Several types can be distinguished on the basis of cell geometry and by the properties of their dendritic tree. Not all of the categorization corresponds with previous descriptions; our findings suggest that some reorganization may be necessary in the accepted classification of cells in the proximal areas of the vertebrate retina. A subtle yet remarkable pattern underlies the entire structure of the catfish retina; there exists a definite gradient of size within a particular class of cells, and of configuration among the subclasses of a specific cell type. It remains to be seen if these morphological spectra bear any functional consequences. The fact that the structure of the catfish retina most closely resembles those of other phylogenetically ancient animals, such as the skate and the dogfish shark, testifies to its primitive organization; morphological and functional mechanisms discernible in this simple system may, therefore, be applicable to the retinas of higher ordered vertebrates.

Morphological and functional identifications of catfish retinal neurons. II. Morphological identification

In this study the morphological origins of the responses from the catfish retinal neurons evoked by step inputs were determined by injecting intracellularly a dye, Procion yellow. A method was devised to view the dye-injected neurons in flat mount to study their dendritic expansion; later the same neurons could be sectioned radially to locate the levels of their somata or dendritic expansion. The results of this study show the inherent danger of identifying dye-injected neurons only in a radial or tangential view. Bipolar cells could be identified functionally without any ambiguity by changing widely the stimulus parameters, because the stimulation of their receptive-field center and surround gave rise to responses of opposing polarity. We found no exception to this rule. The neurons in the proximal layers produced a large variety of responses which could not be segregated into two such classes as the amacrine and ganglion cells. In this part II they were classified into three broad categories: neurons giving rise to sustained, transient, and spiking responses. The demarcation among the three types, morphologywise and functionwise, was vague and not well established. The sustained responses were recoreded from the starburst and spaghetti neurons (part I (9)) which correspond to Ramón y Cajal's (2) amacrine cells. The transient responses, whose patterns were largely invariant of the changes in the stimulus parameters, were recorded from a class of neurons with spindle-shaped somata in the INL. We do not know whether they had axons or not, but we will not be surprised if a future study defines them as a class of ganglion cells. Responses with or without spike discharges were recorded from a class of neurons which were identified as ganglion cells. Observations made on a large number of Procion-injected neurons in both flat-mount preparations and radial sections show that finer dendritic arborizations were not seen in the dye-injected neurons although the presence of such branches was proved in the Golgi preparations. Probably this was due to the weak contrast of the Procion-injected cell against the tissue background, rather than the failure of the dye to diffuse into finer branches. We recognize the severe difficulty involved in the traditional approach of identifying a class of neurons based on typical but subjectively selected functional and structural samples. Neurons have to be classified statistically according to their (quantitative) parameters. (cont'd)

White-noise analysis of a neuron chain: an application of the Wiener theory

The Wiener theory of nonlinear system identification was applied to a three-stage neuron chain in the catfish retina in order to determine the functional relationship between the artificial polarization of the horizontal cell membrane potential and the resulting discharge of the ganglion cell. A mathematical model was obtained that can predict quantitatively, with reasonable accuracy, the nonlinear, dynamic behavior of the neuron chain. The applicability of the method is discussed. We conclude that this is a very powerful method in the analysis of information transfer in the central nervous system.