Synaptic contacts of somatostatin-immunoreactive amacrine cells in goldfish retina

Cellular patterns in the inner retina of adult zebrafish: quantitative analyses and a computational model of their formation

The mechanisms that control cellular pattern formation in the growing vertebrate central nervous system are poorly understood. In an effort to reveal mechanistic rules of cellular pattern formation in the central nervous system, quantitative spatial analysis and computational modeling techniques were applied to cellular patterns in the inner retina of the adult zebrafish. All the analyzed cell types were arrayed in nonrandom patterns tending toward regularity; specifically, they were locally anticlustered. Over relatively large spatial scales, only one cell type exhibited consistent evidence for pattern regularity, suggesting that cellular pattern formation in the inner retina is dominated by local anticlustering mechanisms. Cross-correlation analyses revealed independence between the patterns of different cell types, suggesting that cellular pattern formation may involve multiple, independent, homotypic anticlustering mechanisms. A computational model of cellular pattern formation in the growing zebrafish retina was developed, which featured an inhibitory, homotypic signaling mechanism, arising from differentiated cells, that controlled the spatial profile of cell fate decisions. By adjusting the spatial profile of this decaying-exponential signal, the model provided good estimates of all the cellular patterns that were observed in vivo, as objectively judged by quantitative spatial pattern analyses. The results support the hypothesis that cellular pattern formation in the inner retina of zebrafish is dominated by a set of anticlustering mechanisms that may control events at, or near, the spatiotemporal point of cell fate decision.

Synaptic organization of regenerated retina in the goldfish

In the adult goldfish, any manipulation that significantly depletes retinal neurons stimulates neurogenesis and the regeneration of nearly normal retina. We sought to determine the extent to which the regenerated neurons formed normal synaptic connections. We used qualitative and quantitative electron microscopy to compare the organization of the synaptic layers in regenerated and normal retinas. In eight eyes, a small patch of retina was surgically excised, stimulating regeneration of new retina in its place. Animals were killed 16-20 weeks after surgery. Qualitative comparisons of the synaptic architecture of photoreceptor terminals in the outer plexiform layer and quantitative comparisons of the synaptic organization in the inner plexiform layer were made between the patch of regenerated retina and an adjacent intact site. In the regenerated outer plexiform layer, cone pedicles and rod spherules were not arranged as regularly as normal, but they formed normal-appearing synaptic contacts. In the regenerated inner plexiform layer, with one exception, the quantitative descriptors of the synaptic organization in the normal and regenerate were not significantly different: The planimetric and numerical densities of the synapses, number of synapses/inner retinal neuron, and, with the exception of the bipolar terminals in the inner plexiform layer, and synapse depth profiles were similar. These data suggest that 1) relatively normal synaptic connections are recreated during regeneration, 2) the cellular mechanisms that guide synaptogenesis during development act during retinal regeneration, and 3) the physiological response properties of regenerated neurons should be comparable to that found in the normal retina.

Peptidergic neurons of teleost retinas

In retinas of teleost fish, neuropeptides typically have subtle, modulatory actions. The peptide effects typically have long latencies and durations, and, in some instances, they are known to be mediated by second messengers. Peptidergic neurons in teleost retinas have certain morphological features in common that are consistent with their function. Most peptidergic neurons are stratified amacrine cells with long, varicose processes; the processes of peptidergic centrifugal axons are also narrowly stratified and ramify extensively in the retina. The peptidergic amacrine cells are relatively infrequent, and, likewise, the centrifugal axons originate from a small number of perikarya in the brain. Cells that are so sparsely distributed and whose processes overlap so extensively are better-suited for modulation than for conveying detailed representations of visual space.

Light-and electron-microscopic studies of the somatostatin-immunoreactive plexus in the cat retina

Two monoclonal antibodies directed against somatostatin 14 were used to study immunoreactive neurons, their processes and their synapses in the cat retina. In retinal whole-mounts, a sparse population of wide-field displaced amacrine cells was observed predominantly in the ventral retina and near the retinal margin. Processes of these cells ramified mainly in two distinct strata within the inner plexiform layer: one near the inner nuclear layer (INL), and the other near the ganglion cell layer (GCL). The length of immunoreactive fibres within each plexus was measured: 232 +/- 32 mm/mm2 near the INL and 230 +/- 74 mm/mm2 near the GCL in all retinal regions. The immunoreactive processes were studied using electron-microscopic techniques; conventional and some ribbon-containing synapses ("dyads") were found. Immunolabelled processes received input synapses from other amacrine cell processes. These investigations provide further evidence that this cell population has a diffuse, regulatory or modulatory role for visual-information processing in the inner plexiform layer.

Glycine-receptor immunoreactivity in retinal bipolar cells is postsynaptic to glycinergic and GABAergic amacrine cell synapses

Glycinergic innervation of the synaptic terminals of mixed rod-cone bipolar cells in the goldfish retina was investigated by electron microscopical immunocytochemistry with presynaptic and postsynaptic markers for glycinergic neurons: a monoclonal antibody (mAb 7A) against the 93 kDa subunit of the strychnine-sensitive glycine receptor and polyclonal antisera against a glycine/BSA conjugate. Conventional "glycinergic" synaptic contacts, made by amacrine cell processes, accounted for 7-10% of the input to the bipolar cell terminals, whether determined by glycine receptor immunoreactivity (GlyR-IR) or glycine-IR. In addition to the conventional synapses, the large bipolar cell terminals in the proximal inner plexiform layer (type Mb) gave rise to spinules (spine-like protrusions) that invaginated into presynaptic amacrine cell processes. Although 85% of the spinules were GlyR-IR, no spinules were postsynaptic to glycine-IR processes; yet 86% of the spinules were postsynaptic to GAD-IR processes, suggesting that the GlyR-IR spinules were postsynaptic to GABAergic terminals. Furthermore, a single amacrine cell process could make two synapses with an Mb terminal: a GlyR-IR contact onto a spinule and a conventional synapse that was not GlyR-IR. We suggest that glycinergic innervation of bipolar cell terminals involves conventional glycinergic synapses as well as an unconventional situation in which GABA and glycine may interact in as yet undetermined manner, perhaps by potentiation.

Synaptic organization of two types of amacrine cells with CRF-like immunoreactivity in the turtle retina

The ultrastructural features and synaptic contacts of two amacrine cell types with corticotropin-releasing factor-like immunoreactivity in the turtle retina were examined using electron immunocytochemistry. Type A cells were found only in the visual streak and had elongated dendritic arborizations that ran parallel to the visual streak. These cells arborized primarily in stratum 1 and near the border of strata 2 and 3 of the inner plexiform layer, with some processes extending into stratum 5. Type B cells were found only ventral to the visual streak and arborized primarily in a wide band in strata 4 and 5, with sparse dendritic arborizations in stratum 1. There was a diffuse cytoplasmic reaction product within each cell type; however, large labeled vesicles were rarely observed. Type A amacrine cells received many conventional synaptic contacts from amacrine cells in stratum 1 and at the border of strata 2 and 3, but only a small number of contacts in stratum 5. Bipolar synaptic contacts onto type A amacrine cells were observed in strata 1 and at the border of strata 2 and 3. The only positively identified synaptic outputs of type A cells were conventional synapses onto amacrine cells in strata 1 and at the border of 2 and 3. Type B amacrine cells received synaptic contacts from amacrine cells in strata 1 and 5, and bipolar cell synaptic input in stratum 5. They made conventional synapses onto amacrine cells in strata 1 and 5, and onto bipolar cells in stratum 5. We also found conventional synaptic contacts between unlabeled amacrine cells and type B amacrine cells outside of the primary layers of stratification. In addition, there were specialized junctions observed between type A cell profiles in stratum 1 and between type B cell profiles in stratum 5. The unique regional distributions of the type A and B cells, as well as their differences in synaptic connectivity, suggested that these amacrine cells play distinct physiological roles although they contain the same neuropeptide.

Glycine high-affinity uptake labels a subpopulation of somatostatin-like immunoreactive cells in the Rana pipiens retina

Somatostatin-like immunoreactivity (Som-LI) and glycine high-affinity uptake have been characterized in the Rana pipiens retina. These labels are found in both the outer and inner plexiform layers (OPL and IPL), suggesting that interplexiform cells (IPCs) contain both Som and glycine in this retina. In double-label experiments these labels colocalize to an abundant population of cells in the mid-inner nuclear layer (INL), in the second or third cell layer distal from the IPL. These cells have medium sized spherical or oval somas, each with a single thin descending dendrite which ramifies in the distal IPL. Processes ascending from cells at this location were not visualized by immunocytochemistry, but could be seen by autoradiography of tissue processed for glycine high-affinity uptake. In autoradiographs apparent IPCs were the most intensely labeled cell type in this retina. Som-LI is also found in two types of probable amacrine cells in the proximal INL adjacent to the IPL, neither of which is labeled by glycine high-affinity uptake. One of these is rare (about 10 cells/mm2), and has a large pyriform soma with a thick dendrite that branches in the proximal IPL. The other type is more common (324 +/- 20 cells/mm2), has medium-sized spherical or horizontally elongated elliptical somas, and has multiple thin dendrites projecting into the distal IPL. In addition to the above cell types, faint Som-LI was seen in cells of the ganglion cell layer, possibly indicating the presence of somatostatinergic ganglion cells or displaced amacrine cells.

Morphology and distribution of synapses onto a type of large field ganglion cell in the retina of the goldfish

The morphology and dendritic distribution of terminals that synapse onto a type of large-field ganglion cell in the retina of the goldfish are described. Electron microscopy was combined with retrograde labelling of cells with horseradish peroxidase (HRP). Synapses from both amacrine (four types) and bipolar cells contacted the dendrites (all orders) of these cells. In contrast to a recent report describing the synaptic organization of large-field ganglion cells in the catfish (Sakai et al., '86), the synapses were relatively evenly distributed throughout the dendritic arbor, not clustered at discrete sites, and no presynaptic specializations were seen in the dendrites of the ganglion cells.

Light and electron microscopic immunocytochemistry of neurons in the blowfly optic lobe reacting with antisera to RFamide and FMRFamide

Different antisera to the molluscan cardioexcitatory peptide FMRFamide, and its fragment, RFamide (Arg-Phe-NH2), label a distinct population of neurons in the optic lobe of the blowfly, Calliphora erythrocephala. Seven morphological types of RFamide/FMRFamide-like immunoreactive neurons could be distinguished in the optic lobes based on the locations of their cell bodies, their axonal projections and the distribution of their processes. Of these, two types could be resolved in their entire extent, the others were labeled only in their cell bodies and terminal processes or were partly obscured by other immunoreactive processes. The RF-like immunoreactive neurons in the optic lobes are of two main classes: (1) two types of large field projection neurons and (2) five types of local neurons. One type of projection neurons (five in each lobe) connects the entire projected retinal mosaic of the medulla and lobula in the optic lobe with protocerebral centres associated with the mushroom body calyx. The other type (2-3 invading each lobe) has cell bodies in the protocerebrum and contralateral processes invading optic lobes. Of the class of local neurons there are two amacrine RF-like immunoreactive neurons in each medulla. Each of these amacrines supplies the entire mosaic with fine processes. The remaining local RF-like immunoreactive neurons are present in relatively large numbers (one type in more than 2000 copies in each medulla) and-supply the medulla, lobula and lobula plate neuropils with fine varicose processes. In the medulla the RF-like immunoreactive processes are arranged in strict layers whereas in the lobula complex the distribution is diffuse. Electron microscopic immunocytochemistry, using both pre-embedding immuno peroxidase-antiperoxidase and post-embedding protein A-gold labeling, was employed for analysis of cytology and synaptic connections of RF-like immunoreactive neurons in the medulla. The varicosities of the processes of the large field projection neurons were not found to make chemical synapses with other neurons in the medulla. The spines of the RF-like immunoreactive processes of the large medulla amacrines, however, make pre- and postsynaptic contacts with other neural elements. Our findings indicate that an RFamide/FMRFamide-like substance may be used as a neurotransmitter or neuromodulator by optic lobe neurons of different types. The local and projection RF-like immunoreactive pathways probably play different roles in visual processing.

Distribution of synaptic inputs onto goldfish retinal ganglion cell dendrites

Retinal ganglion cells in the goldfish were labeled by retrograde transport of horseradish peroxidase, and areas near the optic disk where the dendrites appeared to be completely filled were analyzed by electron microscopy. Only 6% of their inputs were ribbon synapses from bipolar cells; the other 94% of the inputs were conventional synapses mostly or entirely from amacrine cells. There were three strata of the inner plexiform layer with high densities of inputs to ganglion cells, the first centered at approx. 50% and the second at approx. 80% of the inner plexiform layer depth, as measured from the ganglion cell layer to the inner nuclear layer. These two strata comprised 25% of the volume but contained 41% of the inputs to ganglion cells. There were also two strata with very low densities of ganglion cell inputs located near the boundaries of the inner plexiform layer, from 0- to 15% and 90- to 100% of the depth. These strata, which also comprised 25% of the volume, contained only 7% of the inputs to retinal ganglion cells. These strata near the boundaries of the inner plexiform layer also contained 81% of the processes with large, dense-cored vesicles characteristic of peptidergic neurons. We concluded that each of the two sublaminae, a and b, identified previously by physiological criteria, could be further divided into at least two strata, one near the boundary of the inner plexiform layer with abundant peptidergic terminals and very few ganglion cell synapses and another near the center of the inner plexiform layer with numerous ganglion cell synapses. We also propose a hypothesis that could explain the functions of these additional strata.

Amacrine cells with neurotensin- and somatostatin-like immunoreactivities in three species of teleosts with different color vision

Neurotensin- and somatostatin-like immunoreactivities were localized by pre-embedding techniques in retinal whole-mounts and radial sections of a monochromatic glass catfish (Kryptopterus bicirrhis), a dichromatic cichlid species (Aequidens pulcher), and the tetrachromatic roach (Rutilus rutilus). Both neuropeptides were observed in perikarya and processes of amacrine cells. For a precise identification of cell types, tangential and radial views were correlated with Golgi-impregnated material. The dendritic pattern defining the morphological subtype of amacrine cells was determined by the given neuropeptide or by the species-specific degrees of complexity of retinal structure and function. Neurotensin-like immunoreactivity was localized in amacrine cells of intermediate size, radial symmetry and dendrites with numerous varicosities; they were monostratified in sublayer 3 of the inner plexiform layer. This cell type was common to all three species. In the mono- and dichromatic retinas, a single type of amacrine cell with somatostatin-like immunoreactivity was found with radially oriented, varicose dendrites in sublayer 5. In the tetrachromatic roach retina, two somatostatin-positive amacrine cell types were found with very different patterns of ramification; furthermore, both of these types occurred in more than one sublayer. Possible functional implications for color vision of neuropeptide-specific amacrine cells with uniform morphology in all three species and those with a more varied morphology in the tetrachromatic roach are discussed.

Neurocircuitry of two types of neurotensin-containing amacrine cells in the turtle retina

The ultrastructural features and synaptic contacts of two types of neurotensin-containing amacrine cells in turtle retina were examined by electron immunocytochemistry, and the retinal peptides were characterized using radioimmunoassay and high-pressure liquid chromatography. The two types of cell were distinguished on the basis of their sizes, dendritic arborizations, synaptic connections and cytoplasmic staining characteristics. Type A cells had lightly labeled cytoplasm and large vertically elongated cell bodies which gave rise to a single primary process which in turn branched and ramified as smooth tapering processes in stratum 3 of the inner plexiform layer. Type A cells received approximately equal synaptic input from amacrine and bipolar cells. Type A amacrines had much more overall synaptic input than synaptic output, and they made conventional synaptic contacts onto bipolar, amacrine, and ganglion cells. Type B cells had a much darker-staining cytoplasm and a smaller cell body which gave rise to numerous delicate beaded dendrites which arborized in strata 3, 4 and 5 of the inner plexiform layer. Type B cells received primarily amacrine and some bipolar cell input. Type B cells had equal amounts of synaptic input and output and they made conventional synaptic contacts onto amacrine, bipolar, and ganglion cells. Whereas there were numerous large vesicles (120 nm diameter) that stained for neurotensin in both types of cells, conventional synaptic vesicles (60 nm diameter) were not labeled. In several cases these large labeled vesicles appeared to fuse with the cell membrane in non-synaptic regions and release their contents into extracellular space, which suggested a non-synaptic release of the neurotensin from type A neurons. Immunochemical and chromatographic studies demonstrated that the neurotensin-related material in retina was indistinguishable from neurotensin found in brain. These results are consistent with a neuroactive role for the neurotensin present within the large vesicles. The differences in the synaptic contacts and dendritic arborizations of the two amacrine cell types suggest they play distinctive functions in visual processing.

The light microscopic localization of substance P- and somatostatin-like immunoreactive cells in the larval tiger salamander retina

Light microscopic immunocytochemistry was utilized to localize the populations of substance P (SP)- and somatostatin (SOM)-like immunoreactive cells in the larval tiger salamander retina. Of 104 SP-immunostained cells observed, 82% were Type 1 amacrine cells. Another 8% of the SP-cells were classified as Type 2 amacrine cells, while 10% of the SP-cells had their cell bodies located in the ganglion cell layer and were designated as displaced amacrine cells. Each type of SP-like immunoreactive cell was observed in the central and peripheral retina. SP-immunopositive processes were observed in the inner plexiform layer as a sparse plexus in sublamina 1 and as a denser network of fibers in sublamina 5. Seventy-eight percent of the 110 somatostatin-immunopositive cells observed were designated as Type 1 amacrine cells. Another 12% of SOM-cells were classified as displaced amacrine cells, while only two SOM-immunopositive Type 2 amacrine cells were observed. Nine percent of the SOM-cells were designated as interplexiform cells, based on their giving rise to processes distributing in the outer plexiform layer as well as processes ramifying in the inner plexiform layer. Each type of SOM-immunoreactive cell was observed in the central and peripheral retina, with the exception of the Type 2 amacrine cells, whose somas were only found in the central retina. Lastly, SOM-immunopositive processes in the inner plexiform layer appeared as a fine plexus in sublamina 1 and as a somewhat denser network of fibers in sublamina 5.

Somatostatin-immunoreactive amacrine cells of chicken retina: retinal mosaic, ultrastructural features, and light-driven variations in peptide metabolism

Somatostatin-like immunoreactive amacrine cells of the chicken retina have been characterized by immunohistochemistry at the light and electron microscope levels. The cell bodies were set back from the junction of the inner nuclear and inner plexiform layers, and prominent fibre plexuses were found in sublaminas 1 and 3-5 of the inner plexiform layer. The cells were distributed across the retinal surface with a centroperipheral gradient of cell density. Locally, the cells were organized in a non-random mosaic. Ultrastructurally, immunohistochemical reaction product was found throughout the cytoplasm of the cell bodies, particularly associated with membranous structures, including the cytoplasmic surfaces of the Golgi apparatus, and within large dense-core vesicles. In dendritic varicosities in the inner plexiform layer, reaction product was associated with the external surfaces of small, clear synaptic vesicles. The synaptic relationships of the somatostatin-immunoreactive terminals in sublamina 1 were distinct from those in sublaminas 3-5. Those in sublamina 1 received input predominantly, possibly exclusively, from bipolar cells. Feedback synapses onto bipolar terminals or to the other amacrine cell process at a synaptic dyad were observed. In sublaminas 3-5, input came predominantly, possibly exclusively, from other, non-immunoreactive amacrine cells, and output was primarily onto other amacrine cells. No synaptic contacts with ganglion cells or with other somatostatin-immunoreactive amacrine cells were identified. Changes in levels of somatostatin-like immunoreactivity in retinas of chicks kept on 12:12 light:dark cycles were detected by radioimmunoassay, and by light and electron microscopic immunohistochemistry. Levels of retinal somatostatin-like immunoreactivity increased in the light and decreased in the dark. The changes appear to be light-driven rather than circadian, since with prolonged exposure to light or dark, the levels of somatostatin-like immunoreactivity continued to increase or decrease until plateaus were reached. The light-driven change in levels of somatostatin-like immunoreactivity may be related to the predominance of bipolar input to the immunoreactive processes in sublamina 1 of the inner plexiform layer. The reduction in peptide levels in the dark may indicate greater release of somatostatin-like immunoreactivity from the amacrine cells in the dark, resulting in an inability of peptide synthesis to keep pace with breakdown. In the light, release of somatostatin-like immunoreactivity may be lower, leading to a net synthesis of peptide.

An immunohistochemical study on the river lamprey retina

The localization of structures immunoreactive to various polypeptides and proteins in the retina of the river lamprey (Lampetra japonica) was investigated by means of an indirect immunohistofluorescence method. In the majority of frozen sections, a subpopulation of amacrine cells showed the immunoreactivity (IR) to one of the examined antisera against corticotropin-releasing factor, glucagon, neuropeptide Y, somatostatin, visinin and 5-hydroxytryptamine, respectively. A few fibers in the inner plexiform layer were immunoreactive to cholecystokinin or substance P antiserum. The visinin-like IR was also found in two types of bipolar cells. The IR was positive in Müller cells against glutamine synthetase but not to glial fibrillary acidic protein. In some flat-mounted preparations, glucagon- and serotonin-like reactive amacrine cells and visinin-like immunoreactive bipolar cells (distally located) could be observed. The results obtained suggest that in the lamprey retina the cytoplasmic property of some cells is similar to, but that of others is different from, vertebrate retinal cells.

Neurotransmitter release by certain neuropeptides in the chicken retina

The ability of certain neuropeptides (glucagon, somatostatin, leu-enkephalin and neurotensin) to release known neurotransmitters (glycine, GABA, dopamine and 5-hydroxytryptamine) was tested in the chicken retina. Tritiated neurotransmitters were injected intravitreally in chicken eyes. After excision, the retina was stimulated in vitro with the neuropeptide in micromolar concentrations while monitoring the efflux of radioactivity from the retina. A rise of the efflux represents a stimulus dependent release. Neurotensin release [3H] glycine, [3H]dopamine and [3H]5-hydroxytryptamine. Leu-enkephalin released [3H]dopamine and somatostatin released [3H]5-hydroxytryptamine. Glucagon was without effect. [3H]GABA was not released by any of the neuropeptides.

Varicosity-associated antigens define neuropile subfields in the leech central nervous system

A panel of twelve monoclonal antibodies raised against homogenates of leech nerve cords and four polyclonal antisera raised against purified neurotransmitters were used to label varicosities immunocytologically in the central neuropile of leech segmental ganglia. These immunoreactive varicosities occur in distinct patterns, some of which have a simple geometry. Three antibodies label immunoreactive varicosities distributed in a single dorsoventrally-oriented plane, two label varicosities distributed in lateral hemi-neuropiles (leaving void a central cephalocaudal passageway), and five label varicosities distributed throughout the neuropile. Six antibodies tested label varicosities across leech species, and five of these varicosity populations are distributed in patterns conserved across leech species. Immunocytologically-defined neuropile subfields do not correspond with previously identified histological and ultrastructural features of leech segmental ganglia. Analysis of immunocytologically-defined subfields is extended to include identification of sets of neurons which appear to project into these subfields, and to include an intracellular characterization of one of these neuron sets.

Ultrastructure and synaptic contacts of enkephalinergic amacrine cells in the retina of turtle (Pseudemys scripta)

Bistratified amacrine cells of the turtle retina containing enkephalin-like immunoreactivity were examined with the electron microscope with the aid of peroxidase immunocytochemical techniques. Our goal was to determine the nature and the location of the synaptic contacts of these cells and the intracellular localization of the immunoreactivity. There was a diffuse reaction product throughout the cytoplasm which coated the surfaces of all the organelles and a dense reaction product which filled the core of some large cytoplasmic vesicles (130 nm in dia.). These labeled amacrine cells received conventional synaptic contacts from other unlabeled amacrine cells and ribbon synaptic contacts from unlabeled bipolar cells, in both the proximal and distal inner plexiform layer. These enkephalin-positive amacrine cells made conventional synaptic contacts containing unlabeled synaptic vesicles (60 nm in dia.), with ganglion cells in the proximal inner plexiform layer and with bipolar cells in the distal inner plexiform layer. These results suggest that enkephalin-like material coexists with another neurotransmitter within these neurons and that these amacrine cells are able to integrate information from both amacrine cells and bipolar cells and provide synaptic input to bipolar cells, ganglion cells, and possibly other amacrine cells.

Two types of pyriform Ab amacrine cells in the goldfish retina: an EM analysis of [3H]GABA uptake and somatostatin-like immunoreactivity

Pyriform Ab amacrine cells in the goldfish retina take up [3H]GABA and show somatostatin-like immunoreactivity (SLIR), leading to the question of whether these two markers are labeling the same or different types of Ab amacrine cells. We used a double-label radio/immuno-technique at the electron microscopical level to visualize the comparative location of [3H]GABA uptake and SLIR in Ab amacrine cells and their processes. SLIR was restricted to large dense-cored vesicles in processes of Ab amacrine cells. In no case were processes labeled with SLIR observed to take up [3H]GABA. Thus, there are at least two types of pyriform Ab amacrine cells: one that takes up [3H]GABA and one that shows SLIR. These may correspond to the two types of Ab amacrine cells described by Cajal in his classic studies on the retina.