The distribution of GABA-like immunoreactivity in the cat retina was studied through the use of preembedding immunocytochemistry for light microscopy and by postembedding immunogold techniques for electron microscopy. Staining was observed in both inner and outer plexiform layers. Approximately 30% of the somata in the amacrine portion of the inner nuclear layer were immunoreactive and included amacrine and interplexiform cells. Horizontal cells and a subpopulation of cone bipolar cells were also stained. In the ganglion cell layer, staining was observed in both small- and medium-sized neurons. GABA-labeled amacrine cells were presynaptic to somata of amacrine cells and to dendrites of amacrine, bipolar, and ganglion cells. Bipolar cells were a major target, receiving more than 60% of all labeled synapses in the inner plexiform layer. Many of these contacts were reciprocal synapses. These findings support a major role for GABA-labeled amacrines in providing feedback inhibition to bipolar cells in the inner retina.

The present study examined GABA immunoreactivity within the retinopetal nucleus isthmo-opticus (NIO) of the pigeon centrifugal visual system (CVS) using light- (immunohistofluorescence, peroxidase anti-peroxidase: PAP) and electron- (postembedding GABA immunogold) microscopic techniques. In some double-labeling experiments, the retrograde transport of the fluorescent dye rhodamine β−isothiocyanate (RITC) after its intraocular injection was combined with GABA immunohistofluorescence. GABA-immunoreactive (-ir) somata were demonstrated within the neuropilar zone of the NIO adjacent to the centrifugal cell laminae whereas the centrifugal neurons were always immunonegative. A quantitative ultrastructural analysis was performed which distinguished five categories of axon terminal profiles (P1–5) on the basis of various cytological criteria: type of synaptic contact (symmetrical or asymmetrical); shape, size, and density of synaptic vesicles as well as the immunolabeling (positive or negative), size of profile and appearance of hyaloplasm. Numerous GABA-ir afferents to centrifugal neurons via axon terminal types P2a, P2c, and P3 were observed which comprised 47.1% of the total input. Moreover, the data suggest that some of the P2a terminals, which make up 26.4% of the input, stem from the intrinsic GABA-ir interneurons, whereas the latter receive P1, P3, but also P2 terminal input, indicating that interneurons may contact other interneurons via type P2a axon terminals. The results also suggest that the GABA-ir P3 or the immunonegative P1b and P5 axon terminals are of extrinsic origin arising from cells in the optic tectum whereas the P2c and P4 axon terminals are associated with extra-tectal input to the NIO. The GABAergic innervation of centrifugal neurons within the NIO may be the basis for the demonstrated facilitatory effect of the centrifugal output upon ganglion cell responses. This is relevant to hypotheses regarding CVS involvement in attentional mechanisms through selective enhancement of retinal sensitivity depending on the location of meaningful or novel stimuli.