Once projected onto the retina, the image of our world stimulates photoreceptors and thereby initiates a cascade of processes that culminates in visual perception. The absorption of photons at the photoreceptor layer is signaled to bipolar cells (BCs), and transmitted from BCs to ganglion cells (GCs), which ultimately form the output of the eye to the brain. Before reaching the brain, however, this visual signal has to be compressed such as to transmit a maximum of information with a minimum of wiring.

BCs are not mere relay stations between photoreceptors and GCs. Complex calculations already take place at the photoreceptor-BC synapse, which enable the visual system to perform optimally from dusk to dawn, when light levels change dramatically. Among the many BC types, mixed-input BCs are specially suited for bridging the gap between high scotopic and high photopic levels, due to the fact that they receive signals from both rods and cones. These cells and their mechanisms for coding visual signals in the twilight zone are the object of study of this thesis.

In Chapter 1, a brief review of literature data concerning retinal anatomy and physiology is presented. Emphasis is given to the different mixed-input BC types and the visual tasks these cells are thought to participate in, such as the segregation of visual signals in two complementary channels (the ON- and OFF- pathways), chromatic processing, and light-dark adaptation.

In Chapter 2, the localization of metabotropic glutamate receptors (mGluRs) in the goldfish outer plexiform layer (OPL) is studied. The receptor responsible for rod-driven light responses in ON BCs, mGluR6 (a group III mGluR), is not the only one found at BC dendrites invaginating into rod spherules. Processes of putative mixed-input ON BCs in this synapse are also positively labeled for mGluR1 a, whose function remains obscure. Lateral elements in the rod triads, probably belonging to rod-driven horizontal cells (HCs), are sometimes immunoreactive for mGluR1 a and mGluR6 as well. In the cone pedicles, mGluRs of all three groups are found at horizontal cell (HC) dendrites, but none is found at BC processes. The function of these mGluRs at the HC level is unknown. They might be involved in shaping light responses, since the light-driven conductance in HCs is mediated by AMPA/KA receptors. Finally, mGluR4 (another group III mGluR) is found in about 80% of the rods, and in no cones. This pre-synaptic localization of mGluR4 points to a possible function in controlling glutamate release, a hypothesis that is further investigated in Chapter 6 .

In Chapter 3, electrical coupling is demonstrated in both mixed-input ON and OFF BCs. This lateral integration of signals makes the receptive field center of these cells much larger than their dendritic trees. As a consequence, light responses to spatially small stimuli that do not cover the whole receptive field are less sensitive than those to full-field stimulation. Further, coupling decreases the input resistance ( R in ) of BCs dramatically, limits effective dialysis with the pipette solution and shifts the apparent reversal potential ( V rev ) of light-driven conductances. The gap junctions responsible for this electrical communication are neither permeable to Lucifer Yellow (LY), nor are they sensitive to pH manipulations or to commonly used gap junction blockers. This indicates that the connexins forming these gap junctions are distinct from those responsible for electrical coupling at the HC level. The function of coupling as a mechanism to improve signal-to-noise ratio at the expense of some loss in sensitivity is discussed. The costs that these large receptive fields might represent for visual acuity are dealt with in Chapter 6 .

In Chapter 4, spectral opponency in mixed-input BCs is shown to be a result of antagonistic interactions between rods and cones, or between different spectral types of cones. Because the polarity of these spectrally coded responses also changes with the intensity of the stimulus, it cannot be the underlying mechanism for color vision, which does not show the same intensity dependency. Rather, these BCs seem to be ideal intensity change detectors. The antagonist inputs make the intensity-response relations of opponent BCs much steeper than those of photoreceptors. As a consequence, small changes in stimulus intensity evoke a large change in response amplitude in these cells. These results indicate that goldfish BCs seem to use the same coding scheme as cone-driven HCs to compress information, via a non-opponent, broadband channel (the non-opponent mixed-input ON and OFF BCs), and opponent channels. Which glutamate receptors could be responsible for such intensity-dependent behavior is discussed in Chapter 6 .

In Chapter 5, the presence and function of voltage-gated K + channels at the tips of the dendrites of mixed-input ON BCs is investigated through electrophysiological experiments and model simulations using NEURON. The anatomy of the dendritic terminals and the selective distribution of voltage-gated channels in these dendritic boutons lead to the rectification of light-driven IV relations, and can serve as a gain control mechanism at the rod-BC synapse. In the scotopic range, they can speed up synaptic transmission and generate transience by accelerating BC repolarization. This fast repolarization restores the high gain of the rod-BC synapse, allowing subsequent rod-driven signals to drive the cell efficiently. As light levels increase, tonic suppression of the rod input leads to the opening of many of these voltage-gated channels, shunting the rod pathway and decreasing the gain of the rod-BC synapse. Under this condition, the transmission of cone-driven signals is favored, and the balance between rod- and cone-driven signals at the BC level shifts from rod-dominated to cone-dominated.

In Chapter 6, unsolved issues such as the function of such a multitude of glutamatergic receptors at the first synapse, as well as the costs and benefits of large spatial summation for visual acuity and sensitivity are addressed.