Signal Processing in the Outer Retina

Light-evoked hyperpolarizing signals in rods and cones are transmitted to and processed by neurons in the rest of the retina through a complex, but highly organized network of electrical and chemical synapses. In the outer retina, photoreceptors are electrically coupled to one another, and rods and cones send output signals to second-order retinal cells, the horizontal cells (HCs) and bipolar cells (BCs), via chemical synapses. HCs make feedback synapses on cones, and send feedforward synaptic signals to BCs. BCs can be classified into two major types, the on-center (or depolarizing bipolar cells, DBCs) and the off-center (or hyperpolarizing bipolar cells, HBCs). Additionally, they can be further classified according to their relative rod/cone inputs, color opponency, axonal outputs and other parameters. BCs are the first neurons along the visual pathway that exhibit clear center-surround antagonistic receptive field (CSARF) organization, the basic code for spatial information processing in the visual system, and the synaptic circuitry underlying bipolar cell receptive fields will be discussed in this chapter.

Electrical synapses (coupling) between photoreceptors

Electrical coupling between photoreceptors was first discovered in the turtle retina by Baylor et al in 1971. Electron microscopic studies revealed that gap junctions exist between photoreceptors in many vertebrate species. In the primate retina, anatomical data have shown large gap junctions between cones and smaller junctions between rods and cones, suggesting that cone–cone coupling is stronger than rod–cone coupling. ^, In the turtle and ground squirrel retina, couplings are found between cones of the same spectral sensitivity and between rods; whereas in amphibians, rods are strongly coupled to each other, but cones are not. ^, Coupling between rods and cones has been observed in turtles, tiger salamander and mammals, but the coupling strength is weaker than that between cones (in turtles) and between rods (in tiger salamander). ^, Recent evidence has shown that gap junctions between photoreceptors are mediated by connexin 35/36 gap junction proteins ( Fig. 22.1 ) and double patch recordings have revealed detailed properties of electrical synapses between photoreceptors. For example, junctional conductance between salamander rods is linear, with an average value of 500 pS ( Fig. 22.2A ). In ground-squirrel, green–green cone pairs are coupled with an average conductance of 220 pS, whereas coupling is undetectable in blue–green cone pairs ( Fig. 22.2B ). In the monkey retina, coupling has been found between red-cones, between green-cones and between red- and green-cones ( Fig. 22.2C ).

Although most rods are weakly coupled with cones in the tiger salamander retina, a small fraction of rods (10–15%) are strongly coupled with adjacent cones. Voltage responses of these rods (named rod _C s) to current injection into a next-neighbor cone are about three to four times larger than those of the other rods. Rod _C s behave like hybrids of rods and cones, and one of their functions is to generate off overshoot responses in higher-order retinal neurons under light-adapted conditions.

Photoreceptor coupling decreases resolution (visual acuity) of the visual system by spatially averaging photoreceptor signals over some lateral distance in the retina. However, the advantage of this arrangement is that it improves the signal-to-noise ratio of the photoreceptor output when the retina is uniformly illuminated. This is especially important for the rods to detect dim images under dark-adapted conditions when the voltage noise is high. It can be shown mathematically, however, that photoreceptor coupling decreases the signal-to-noise ratio of the photoreceptor network if the number of illuminated photoreceptors is less than the square root of the number of photoreceptors that are effectively coupled ( Box 22.1 ).

Box 22.1

Photoreceptor Coupling Alters the Signal-to-Noise Ratio of the Photoreceptor Network

If N photoreceptors are coupled and k photoreceptors are uniformly illuminated, then the average (sample) signal is S = kS ₁ /N (where S ₁ is the light-evoked signal in one photoreceptor and σ ₁ is the noise, or the standard deviation of S ₁ from the mean), and the average (sample) variance:

$\begin{array}{c}\sigma ²=var\left(S\right)=var\left(\left(1/N\right)\left(S_1+S_2+...\right)\right)=\left(1/\left(N^2\right)\right)var\left(S_1+S_2+...\right)\\ =\left(1/\left(N^2\right)\right)\left({\sigma }_1+{\sigma }_2+...\right)=\left(\left({\sigma }_1\right)/N\right),\text{and thus}\\ \sigma =\left(\left({\sigma }_1\right)/\left(\surd N\right)\right)\end{array}$

The average (sample) signal-to-noise ratio:

$S/\sigma =\left(k{S}_1/N\right)/\left({\sigma }_1/\surd N\right)=\left(k/\surd N\right)\left(S_1/{\sigma }_1\right).$

Therefore when k > √ N , S/σ > S ₁ /σ ₁ , the average signal-to-noise ratio improves, and when k < √ N , S/σ < S ₁ /σ ₁ , the average signal-to-noise ratio worsens.

Rod and cone photoreceptors operate in different ranges of illumination, and they exhibit different spectral sensitivities. Electrical coupling between rods and cones broadens the operating ranges and spectral sensitivity spans of both types of photoreceptor by mixing their light-evoked signals. Additionally, under conditions when the signal in one type of photoreceptor is suppressed (e.g. rod response in the presence of background light), its output synapse can be used to transmit signal from the other type of photoreceptor (e.g. cones). It has been shown that rod–cone coupling in the tiger salamander retina is enhanced by background light. This allows transmission of cone signals through the rod output synapses. Such “synapse sharing” minimizes the amount of neural hardware and facilitates cone inputs in second-order retinal cells.

Glutamatergic synapses between photoreceptors and second-order retinal neurons

Vertebrate photoreceptors use glutamate as their neurotransmitter, which is released continuously in darkness as the photoreceptors are depolarized. Calcium enters photoreceptor synaptic terminals through voltage-gated calcium channels and triggers exocytosis of glutamatergic vesicles, releasing packages of glutamate, each of these packages activates a number of glutamate receptors that initiate transient postsynaptic currents termed spontaneous excitatory postsynaptic currents, or sEPSCs. ^, sEPSCs are seen in most HBCs, but analogous (sign inverted) discrete postsynaptic currents are generally not observed in the DBCs. ^, This probably reflects a difference in the kinetics of the glutamate receptors in DBCs and HBCs. Postsynaptic responses of the HBCs and horizontal cells (HCs) are mediated primarily by ionotropic glutamate receptors (kainate (KA) or α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors), and the responses of the DBCs are mediated mainly by metabotropic (mGluR6, or l -α-amino-4-phosphonobutyrate (L-AP4)-sensitive) receptors ( Fig. 22.3 ). Glutamate released from photoreceptors in darkness binds to the KA/AMPA receptors in HBCs and HCs and opens postsynaptic cation channels. It also binds to the mGluR6 receptors in DBCs, resulting in a G protein-mediated process, which closes cation channels. Light suppresses glutamate release from photoreceptors, closes AMPA/KA-mediated cation channels and results in hyperpolarization in HBCs and HCs, and it opens mGluR-mediated cation channels and results in depolarization in DBCs. In fish, cone transmission to DBCs has been shown to be mediated by a glutamate-activated chloride current (TBOA-sensitive) that is shut down in the light, resulting in membrane depolarization. ^,

It has been shown that the AMPA/KA receptors in HBCs are largely desensitized in darkness, leading to small postsynaptic currents and reduced response dynamic ranges. In the salamander retina, rod-dominated HBCs have been found to evade postsynaptic receptor desensitization by using synchronized multiquantal release at multiple invaginating ribbon junctions and GluR-4 AMPA receptors as the postsynaptic receptors. The large multiquantal events have fast decay time so that self-desensitization is avoided and they allow long intervals between events so that mutual desensitization is minimized. Consequently HBC _R s are not desensitized in darkness, allowing light signals to be encoded by the full operating range of the glutamate-gated postsynaptic currents.

The exact gating mechanism of the mGluR6 receptor-mediated cation channels is unclear. The mGluR6 receptors generate substantially slower conductance changes compared with the AMPA/KA receptors in HBCs, and thus the sEPSC events in DBCs are heavily filtered.

Detailed descriptions of ionotropic and metabotropic glutamate receptors and glutamate transporters in the retina are also given in Chapter 21 .