The only thing so far agreed upon is that it is the horizontal cells which are the mediators of the antagonism. Horizontal cells are probably the largest of all retinal neurons and were originally considered not to be nerve cells at all but a specialized form of neuroglia (nerve putty) to provide an additional element of mechanical stabilization of the retina. Horizontal cells are further extensively coupled by so called gap junctions, tiny pores in the cells membrane in areas of cell to cell contacts through which electrical signals and second messengers (substances that carry signals within the cells) are exchanged between coupled cells. These gap junctions are permeable for a variety of substances up to a molecular weight of 1000 D including a number of fluorescent dyes, e.g. lucifer yellow. Thus, in order to see the entire network (also called syncytium) of these cells, one only needs to inject one of the cells and let the dye pass through the junctions to the others, that is, if the junctions are open.

Photomicrograph of a whole mounted turtle retina in which one horizontal cell was dye injected with lucifer yellow. The smaller, round spots are the somata of the cells which bear rather fragile lateral processes. The thick, convoluted processes are long extensions of the cells which form a separate signalling units (called horizontal cell axon terminals) and are functionally and electrically isolated from their somata. Through this design, the cells are actually split into two parts (one might say their number is doubled without actually having to support and maintain the additional cells). However, among each other, they are connected via gap junctions which permit passage of dye from cell to cell (under dark adapted conditions) as well as the communication of electrical signals.

As mentioned above, horizontal cells utilize the neurotransmitter GABA, which inhibits neuronal activity by shunting the postsynaptic cells (it opens Cl- channels on the membrane which leads to influx of the negatively charged ion, consequently, positive charges elicited by excitation cannot result in depolarization because they are compensated by the negative charge. Thus, the postsynaptic cell remains inactive or is inhibited).
So, there are two observations, the first being that horizontal cells mediate center-surround organization, and the second, that horizontal cells are GABAergic, that is they use GABA as transmitter. Consequently, one must assume that GABA, indeed is the substance that generates the inhibitory influence of the surround, or not?
Up to now, the most widely accepted models for center-surround organization rely on this assumption. Both models propose that under steady state conditions, horizontal cells constantly release low levels of GABA. Now if a photoreceptor is hit by a flash of light, it is hyperpolarized, that is, it is inhibited (strangely enough) and stops releasing its excitatory transmitter glutamate. Consequently, the horizontal cell further down in the signalling pathway is no longer excited and stops releasing its own (inhibitory) transmitter GABA. Horizontal cells, however, span over a wide field and thus, all other cells which fall into this field experience the sensation of annihilation of their inhibitory influence. In other words, they are disinhibited. As mentioned just above, photoreceptors use the excitatory transmitter glutamate and are inhibited by light but excited by darkness. Therefore, in a simple word, glutamate which is released by photoreceptors, is a signal for darkness. Now, if a photoreceptor that resides in darkness is further disinhibited, and releases even more glutamate, this will signal darker than dark to all other neurons further down the visual pathway.
The difference between the two models is that the actual targets for GABA are assumed to be the photoreceptors (in the first model) or the bipolar cells (second model). Since the first model proposes a photoreceptor to horizontal cell and back to photoreceptor signalling pathway, it has been called the “feedback model”. The second model proposes a photoreceptor to horizontal cell and further to bipolar cell signalling pathway and since these cells are serially connected, the term “feed forward model” was coined.

Schematic representation of the feedforward and the feedback models. Feedforward means that the inhibition by horizontal cells acts forward, that is on the further downstream located bipolar cells, whereas the feedback model implies that the inhibition is "feeding back" to the photoreceptors. In the latter case, this would provide a tonic level of inhibition of the photoreceptors which would be interrupted by a flash of light absorbed by a photoreceptor within the field of the horizontal cell. As a consequence, the horizontal cell would stop releasing GABA, and all other photoreceptors within its reach would be disinhibited. This would cause increased release of glutamate by photoreceptors which causes the sensation of darker than dark. A bit complicated, I agree.

There are only two problems with both models. First, if one assumes that the stop of GABA release is the signal that mediates the surround response, then, just bathing an isolated retina in GABA should completely inhibit this effect. However, reality proves otherwise because GABA does not affect the surround response of bipolar cells, and second, unlike other neurons, horizontal cells do not possess the presynaptic specializations necessary for even release of GABA. Of course, one can always propose that there should be “the other inhibitory transmitter”, only, nobody has been able so far to identify any suitable substance for this role. Excellent reviews on this subject were published by Dwight Burkhard and Marco Piccolino

On the other hand, one must not forget that GABA is also the endproduct of the metabolic glutamate inactivation pathway. So, if horizontal cells are inhibitory and contain GABA, they might actually function by removing glutamate (excitation) and the safest way to do this would be to convert it into an inhibitory substance which has very little toxic side effects: GABA. If one spins this thought a little further, it sounds plausible that, instead of only inhibiting glutamate release by their own release of GABA, these cells might do something much easier, that is, they simply remove glutamate at a constant rate to keep the level of excitation at a reasonable rate and then convert it into a harmless substance: GABA, similar to a Peltier element which removes heat and substitutes cold in return.

Cartoon of the glutamate uptake mode as an alternative for generating center surround organization. Photoreceptors constantly release a certain amount of glutamate (green arrows) which results in an intermediate level in the extracellular space contained within the cone pedicle. Horizontal cells take up this glutamate and convert it into GABA (red). Now, all one has to postulate is that the uptake of glutamate is regulated, that is, it stops once the cells are hyperpolarized whereas it is highest when the cells are excited.
If a flash of light falls upon a photoreceptor, it will stop releasing glutamate. This results in inhibition of the horizontal cells connected to this photoreceptor. The horizontal cell will stop taking up glutamate at all of its connections to photoreceptors. Consequently, the glutamate levels at the synapse of nearby, unilluminated cones must rise instantaneously and generate a "darker than dark" signal
Note the opposite responses of the two bipolar cells connected to the two photoreceptors when the light comes on.

Sounds almost too simple to be true and further, how could one ever prove that such a mechanism actually exists?
There is no real way of unambiguously proving this, however, we were able to show some rather strong evidence for the faisability of our model. The first thing that is essential for this model is that, under normal conditions, horizontal cells do not contain any glutamate and we were able to confirm this notion already published by some colleagues (*****) by performing an immunoreaction against this transmitter. As shown in the above mentioned studies, photoreceptors, as well as bipolar cells and retinal ganglion cells contain an abundance of this excitatory transmitter, whereas it is absent or non-detectable from horizontal and amacrine cells.

This is a photomicrograph of a 20 µm thick frozen section of a turtle retina, stained for glutamate using an immunoreaction coupled to a fluorescent tag. The cells that show the strongest staining are the bipolar (long thin arrows) cells but also photoreceptors (curved arrows) and retinal ganglion cells are labeled (opem arrows). Note that, however, there is a complete absence of labeling from the horizontal cells (big white arrows). Small arrows in the IPL identify stratum #2.

So, if we consider this “non-detectable” level of glutamate in horizontal cells as the baseline, all we had to do was to block the catabolism of glutamate into GABA (which is done by the enzyme glutamate dehydrogenase) and, after a short period of time we would be able to see whether the cells had, in fact taken up glutamate by using the same immunoreaction as shown above. If our assumption of a regulated uptake of glutamate held, we would further have to look at the effect of either excitatory or inhibitory input to these cells over a certain period of time to see whether there was any difference in the amount of accumulated glutamate, i. e. in the intensity of the immunoreaction. To recapitulate the idea, what we assumed was that, upon hyperpolarizaton (inihibition), the horizontal cells would stop taking up glutamate. In the closed environment of the photoreceptor to bipolar cell / horizontal cell synapse, this would lead to an immediate increase of the free glutamate levels which is equivalent to a darker than dark signal to all higher order neurons of the visual pathway.

This picture table shows the outcome of four experiments that were done in order to prove that there is a regulated uptake of glutamate by retinal horizontal cells. Starting from the baseline distribution (shown in the previous picture), all retinas were incubated in 1 mM aminooxyacetic acid which blocks the enzyme glutamate dehydrogenase (GAD) responsible for the conversion of glutamate into GABA. At the same time, different drugs were added, 10 µM kainic acid (a very strong glutamate agonist, extracted from Japanese seaweed) to excite horizontal cells in panel "a", or 1 mM piperidine dicarboxylic acid (a strong glutamate antagonist) to hyperpolarize (inhibit) horizontal cells in panel "b", both for 20 minutes. The effect is quite obvious, excitation of horizontal cells (marked by the asterisks) leads to a dramatic accumulation of glutamate by these cells and their axon-terminals (arrows). Hyperpolarization appears to completely block this uptake since there is no detectable glutamate in these cells and they are visible only as black holes (indicated by the white asterisks) between the labeled bipolar cells and photoreceptors. Black asterisks at the bottom of the micrograph mark labeled retinal ganglion cell axon bundles (the fibers that transmit the information to the brain)
As additional controls, we ran two more experiments, in which we used the conditions of panel "a", that is excitation of horizontal cells but this time, we inhibited any release of glutamate from retinal neurons by removing all extracellular calcium (which is necessary for the release to work). This was necessary to show that, indeed the proposed uptak e and not some suddenly activated synthesis of glutamate within the horizontal cells was the source of the glutamate levels detected in figure "a" (Fig. c). In other words, if there is no extracellular glutamate available, then horizontal cells could also not be expected to take it up.
As last control, we added D,L -threo-aspartate which blocks the glutamate transporter in neurons to the kainic acid solution. So, in this case, there was enough extracel lular glutamate available, but the transporter molecule was blocked and we expected some kind of washout of glutamate from the entire retina. (Fig. d). All experiments worked exactly as planned without any ambiguity since there was no detectable glutamate in horizontal cells except under the conditions described in Fig. "a"

The experimental findings strongly suggest that there is, at least, some validity to our model, even though there are still some other problems (under construction).

There is also a fourth model now, proposed by Maarten Kamermans from the University of Amster dam which is too complicated for me to explain and besides it is not my achievement but maybe he'll take some time and put a page up, too

As a summary, I might state that there is no unequivocal proof for any of the models proposed, neither does it appear plausible that one single model is responsible for the generation of the antagonistic surround. Most likely, a combination of all four models is what really generates the surround responses in second order cells.

next page (under construction)