Postembedding silver-intensified immunogold procedures reveal high levels of glutamate immunoreactivity in "vertical" elements of the goldfish retina: (1) Red-sensitive and green-sensitive cones display strong glutamate immunoreactivity, especially in their synaptic terminals, but blue-sensitive cones are poorly immunoreactive. (2) All type Mb (on-center) and Ma (off-center) mixed rod-cone bipolar cells and all identifiable cone bipolar cells are highly glutamate immunoreactive. We find no evidence for bipolar cells that lack glutamate immunoreactivity. (3) The majority of the somas in the ganglion cell layer and certain large cells of the amacrine cell layer resembling displaced ganglion cells are strongly glutamate immunoreactive. (4) Despite their high affinity symport of acidic amino acids, the endogenous levels of glutamate in Müller's cells are among the lowest in the retina. (5) GABAergic neurons possess intermediate levels of glutamate immunoreactivity. Quantitative immunocytochemistry coupled with digital image analysis allows estimates of intracellular glutamate levels. Photoreceptors and bipolar and ganglion cells contain from 1 to 10 mM glutamate. The bipolar and ganglion cell populations maintain high intracellular glutamate concentrations, averaging about 5 mM, whereas red-sensitive and green-sensitive cones apparently maintain lower levels. Importantly, photoreceptor glutamate levels are extremely volatile, and in vitro maintenance is required to preserve cone glutamate immunoreactivity in the goldfish. GABAergic horizontal and amacrine cells contain about 0.3-0.7 mM glutamate, which matches the values predicted from the Km of glutamic acid decarboxylase. Müller's cells and non-GABAergic amacrine cells contain less than 0.1 mM glutamate. Though Müller's cells are known to possess potent glutamate symport, they clearly possess equally potent mechanisms for maintaining low intracellular glutamate concentrations.

Autoradiography of goldfish retinas incubated in micromolar levels of 3H-serotonin displayed 3 kinds of labeled somas in the inner nuclear layer: S1 amacrine cells with heavy labeling, large somas, and a sparse distribution (approximately 93/mm2); S2 amacrine cells with moderate labeling, smaller somas, and a denser distribution (approximately 500/mm2); and a subset of bipolar cells with light labeling, small somas, and a very dense distribution (approximately 4000/mm2). Serotonin-like immunoreactivity was observed only in S1 amacrine cells and their synaptic terminals. Radiolabeled terminals in the inner plexiform layer formed 4 strata that were differentially assigned to the 3 cell types. S1 amacrine cells arborized in sublayers 1 and 5, received inputs from type a1 bipolar cells and amacrine cells, and made synapses on other amacrine cells, type a1 bipolar cells and unidentified processes. Thus, S1 amacrine cells seem to receive significant input from "off-center" pathways. S2 amacrine cells arborized in sublayer 3 and made synapses onto amacrine cells. Labeled bipolar cell terminals were exclusively located in sublayer 2 and were identified as type a2 mixed rod-cone bipolar cells. We conclude that the S1 amacrine cell is truly serotonergic and that radiolabeling of S2 amacrine cells and type a2 bipolar cells is due to cross-specificity for another carrier or processes unrelated to their neurochemical identities. These observations partially reconcile many previous observations on the types, numbers, and synaptologies of teleost retinal neurons identified by different markers for indoleaminergic transmission.