The retina is a thin, filmy piece of brain tissue (in development, the retina arises as outpouching of the embryonic forebrain) that lines the inside of the eyeball (figure 6.1). It consists of millions of closely packed nerve cells arranged in layers with synaptic neuropil between the layers. The retina is the most important part of the eye, for it contains both the sensory neurons that are responsive to light and the first stages of image processing via intricate neural circuits. These circuits construct an electrical message concerning the visual scene that can be sent to the brain for further processing and visual perception. All this takes only a fraction of a second. Vision is truly a miracle of neural processing. If the retina is damaged or degenerates owing to disease, the brain centers cannot be stimulated, and the person or animal is blind.

Figure 6.1.  

Schematic diagram of the eyeball cut in half longitudinally to show the various layers and structure of the human eye. The retina lines the back of the eye against the pigment epithelium and choroid. The blood supply to the retina radiates from the optic nerve; 3.5 mm temporal from the optic nerve lies the fovea which is the center of focus. The white box shows a section of retina enlarged and stained and redrawn in Figures 6.4 and 6.5.

The retina is organized the reverse of the way one might intuitively expect. The sensory cells, the photoreceptors, lie at the very back of the retina, and light rays have to pass all the way through the three layers of cells, two layers of neuropils, and the length of the photoreceptors themselves before finally finding pigment molecules to excite. The reason that the photoreceptors lie at the last level of the retina in terms of light reaching them (although we call it the first level of excitation) is that some of these important pigment-bearing membranes of the photoreceptor, known as outer segment disks, have to be in contact with the pigment epithelial layer of the eye (also brain-derived tissue). An amazing exchange of molecules has to take place between the photoreceptors and the pigment epithelium for vision to occur. The vital molecule retinal, or vitamin A, has to be passed from the pigment epithelium to the opsin molecule in the photoreceptor outer segment membranes to form the photoactive molecules in rods and cones. The vitamin A comes from the blood system, so the pigment epithelium has to be provided with a rich blood source via the choroid of the eye. Retinal lies embedded in the center of the opsin molecule, and only the complete rhodopsin molecule is reactive to photons of light (figure 6.2). A cis-trans isomerization occurs in each of the millions of rhodopsin molecules occurs on exposure to light rays, and the retinal molecule is shifted back to the pigment epithelium in a different form and is there recycled to return to the opsin molecule to form new rhodopsin, ready for more photic activation. The pigment epithelium is usually a very black layer of cells owing to melanin granules contained in them (except in albino people or animals), and these pigment granules have a protective role in both absorbing stray photons that bypass the photoreceptor outer segments and masking the outer segments from too much constant light exposure. If the black retinal pigment epithelium layer and its necessary blood supply, the choroid, were to lie over photoreceptors at the front of the retina, not much light would penetrate at all to excite the photoreceptors and successive chains of neurons in the retina.

Figure 6.2.  

A, Cone photoreceptors of the monkey retina are stained by using a fluorescein-conjugated antibody to GCAP (guanylate cyclase activating protein) contained within them. B, The outer segments of the rods and cones (boxed area) are enlarged to show their interior structure of stacked doubled-over membrane disks. C, The disks contain thousands of rhodopsin molecules embedded in the lipid bilayers of the disks. Each rhodopsin molecule consists of seven transmembrane portions of the protein opsin surrounding the chromophore (11-cis-retinal).

All vertebrate species have retinas that contain at least two types of photoreceptor. Simply put, rods are the photoreceptors for low-light vision, and cones are the photoreceptor type for daylight, bright-colored vision. Animal species have adapted their eyes and retinal design according to the environment in which they live. Most nonmammalian species have very well-developed cone types as their photoreceptors of choice. Most fish, frogs, turtles, and birds have very good color vision too because they have retinas that are designed to take advantage of daylight. When the time of the dinosaurs was over, possible owing to climate changes and extremely long times of darkness because the earth's atmosphere was covered with ash and dark clouds, tiny fur-covered nocturnal mammals evolved that were able to generate their own heat by well-designed blood supplies. The earliest mammals were almost certainly nocturnal and developed visual systems that were most sensitive to dim light conditions. Their visual systems were dependent on rod photoreceptor–dominated retinas. Rodents such as rats and mice today still have the early rod photoreceptor–driven retinal design. Their cones are small and slender and form only 3–5% of the photoreceptor numbers.

Most other mammals have a preponderance of rods in their retinas too, but often cones are organized in higher numbers in specialized areas of the retina to deal with aspects of their visual environment that are important for their survival. Most mammals have their eyes at the side of the head with very little binocular overlap and thus little depth perception. So most mammals have either totally unspecialized areas of retina that are nocturnal in design (rats and mice) or a partial binocularity for daylight vision with a focusing of the image to a central specialized area (area centralis) where cones and cone pathways predominate (cats and dogs), or they have an elongated horizontal strip of specialized retina called a visual streak over which movement and fast actions of predators can be detected (rabbits, squirrels, and turtles). The ultimate in frontal projecting eyes and complete binocular overlap is achieved in some birds and primates. In these cases, the eyes are specialized for good daylight vision, color, and very fine detail discrimination. Thus, primates and raptor birds have a fovea and a foveate design of the retina.

Humans and monkeys have a retina that is specialized to have the cone-predominant daylight vision in the fovea and the rod-predominant vision for night-time sensitivity to poor lighting in the extrafoveal and peripheral parts of the retina (figure 6.3). We have what is called a duplex retina, and we can make good visual discriminations in all lighting conditions. Our retina has the ability to adapt to different lighting conditions, from using our rods at night to perceive the slightest glimmer of photons to our cones taking over in sunlight, allowing us to make color and the finest spatial discriminations. Three separate cone types in primates (red-sensitive, green-sensitive, and blue-sensitive) and two types in most mammals (green-sensitive and blue-sensitive) sense wavelength, allowing the visual system to detect color. We can see with our cone vision from gray dawn to the extreme dazzle condition of high noon with the sun burning down onto white sand. The daylight adaptation to brighter and brighter conditions takes place in the cone photoreceptors themselves initially and then by exclusive circuitry through the retina. Dark adaptation to lower and lower light conditions takes place in the rod photoreceptor–initiated neural circuitry through the retina. These tasks of the retina are placed on millions of nerve cells that are specifically connected into specialized neural chains that are able to influence the output of ganglion cells under constantly varying light stimuli.

Figure 6.3.  

The human retina as seen by an ophthalmologist. The optic nerve head has an array of arteries and veins radiating from it to nourish every part of the retina. Toward the temporal side of the optic disk is the fovea, which is the center of vision and binocular overlap, specialized for high-acuity daylight vision using only cone photoreceptors.

The retina has four to six types of photoreceptor (dependent on species) in the photoreceptor layer, one to four types of horizontal cell and 11 types of bipolar cell in the second (inner) nuclear layer, 22 to 30 types of amacrine cell in the inner nuclear layer, and about 20 types of ganglion cells in the ganglion cell layer sending the visual messages to the brain through over a million optic nerve fibers. The photoreceptors synapse with bipolar and horizontal cell dendrites in the outer plexiform layer of neuropil, and the bipolar, amacrine, and ganglion cells talk to each other in the inner plexiform layer neuropil. To understand the shapes and sizes of the different cell types, we have had to use different staining techniques, from old-style Golgi silver staining employed originally over a hundred years ago by Ramon y Cajal to modern-day immunocytochemical or gene gun techniques. We have been able to understand how the different morphological cell types synapse on each other by examining the synaptic neuropil by electron microscopy to visualize the actual synapses. A great advance in our understanding came with the electron microscope description of the retinal synapses by Dowling and Boycott in 1966. We could recognize different cell-specific synapses made by photoreceptor terminals on horizontal and bipolar cells and by bipolar cells on amacrine and ganglion cells. This arrangement of synapses has been extended now to include staining techniques to reveal gap (electrical) junctions and the neurotransmitter receptor molecules and neurotransmitter uptake transporters. So now we know that the neurotransmitter of the vertical pathways through the retina (photoreceptors, bipolar cells, and ganglion cells) is glutamate and the neurotransmitters of the laterally extending horizontal and amacrine cells are various excitatory and inhibitory amino acids, catecholamines, peptides, and nitric oxide (figures 6.4 and 6.5).

Figure 6.4.  

Immunostained monkey retina close to the fovea. Some neurons of each of the layers are immunolabeled with antibodies against GCAP (photoreceptors), calbindin (horizontal cells and some bipolar cells), calretinin (AII amacrine cells and two other varieties of amacrine cells), and parvalbumin (ganglion cells). Photo, photoreceptor layer—rods and cones; OPL, outer plexiform layer; bc, bipolar cells; hc, horizontal cells; INL, inner nuclear layer; amac, amacrine cells; IPL, inner plexiform layer; GCL, ganglion cell layer; gc, ganglion cells.

Figure 6.5.  

A drawing of a slice of the human retina showing all the nerve cells we currently understand on the basis of their shape, function, and neurocircuitry. The photoreceptors lie deep at the back of the retina against the pigment epithelial cells (top of drawing), and the ganglion cells lie at the superficial surface of the retina (bottom of drawing). Bipolar cells and horizontal and amacrine cells pack the middle of the retina with two plexiform layers dividing them, where synaptic interactions take place.

Electrophysiological investigations of the retina started 60 years ago. Studies of optic nerve discharges showed that indeed the optic nerve fibers could be stimulated to give traditional depolarizing spikes. However, the first recordings in the retina by Svaetichin in the 1950s showed very odd responses to light. The neurons of the outer retina (it was not immediately clear which cells were being recorded from) responded in a slow hyperpolarizing manner and not as depolarizing spikes. These “S potentials” are now known to originate with the photoreceptor and to be transmitted with relatively unchanged waveform to horizontal cells and one set of bipolar cell. The membrane hyperpolarization starts at light ON, follows the time course of the light flash, and then returns to the baseline value at light OFF. We now know that the photoreceptors, both rods and cones, release neurotransmitter during the dark, because under dark conditions, the membrane of the sensory neuron is in a depolarized state. Cyclic GMP–gated channels are open to sodium influx in the dark state. On light exposure, the rhodopsin molecules undergo their conformational change as mentioned above, and a resulting phototransduction cascade closes the membrane channels, sodium is kept out, and the membrane of the whole cell goes into a hyperpolarized state for as long as the light is present. The hyperpolarizing response can be recorded both in the outer segment of the photoreceptor by suction electrodes and in the cell body or synaptic region of the photoreceptor by sharp microelectrodes. The hyperpolarizing response of a cone has a small area over which it responds that is not much bigger than the diameter of the cone. This space over which the cone gives its response is known as its receptive field (figure 6.6).

Figure 6.6.  

Physiological response of single cone photoreceptors. Suction electrodes record the response of the outer segment area. A brief light flash momentarily hyperpolarizes (by 1–2 millivolts) the cone cell's membrane. The intracellular response to a longer flash of light can be recorded in the cone cell body as a slow hyperpolarization (20 or more millivolts) that lasts as long as the light flash. Input from neighboring rods that are coupled to the cone by gap junctions can also be recorded in the cone response under certain stimulating conditions (rod input). The receptive field of a cone is very narrow and is a hyperpolarization (downward direction) of the cell's membrane potential.

Both rods and cones respond to light with the slow hyperpolarizing response described above, yet rods and cones report different image properties. Rod vision typically deals with a slow type of feature detection in which dim light against dark is detected. Cones deal with bright signals and can detect rapid light fluctuations. Cone system signals are revealed in forms of feature detection in which bright against dark (or vice versa) colors or edges are being detected. Thus, photoreceptors are the first neurons in the visual chain to decompose the image into separate parts. However, now the image has to be differentiated into further component elements. This happens at the first synapses of the visual pathway: the synapses between photoreceptors and bipolar cells. Here, different bipolar cell types selectively express different types of receptors for glutamate, allowing each bipolar type to respond to photoreceptor input in a different way. Some bipolar cells are tuned to faster and some to slower fluctuations in the visual signal. Electron microscopy shows that bipolar cell dendrites make different types of contact with the cone or rod synaptic region, either beneath the synaptic ribbon or at more distant basal contacts.

The types of bipolar cells that make the basal contacts express either rapidly desensitizing, rapidly resensitizing AMPA receptors or slowly resensitizing kainate receptors. Both of these types of receptor are excitatory and are called ionotropic glutamate receptors (iGluRs). But the most extraordinary difference between bipolar cells is that another, separate group of bipolar cells express inhibitory glutamate receptors. Inhibitory glutamate receptors known as metabotropic glutamate receptors (mGluRs) are unique to the vertebrate retinas. Typically, bipolar cells that make the central element, ribbon-related contacts with the photoreceptor synaptic terminal, use these inhibitory mGluRs. Together, these iGluR- and mGluR-expressing bipolar cells initiate a set of parallel visual pathways connecting photoreceptors to ganglion cells, for shadow and for highlight detection. These are known as OFF (dark-on-light) and ON (light-on-dark) pathways, respectively (figure 6.7).

Figure 6.7.  

The photoreceptor endings (rod, cone pedicle) contact second-order neurons, with horizontal and bipolar cell dendrites, at specialized synapses. The detailed structure of the photoreceptor synapse (boxed area) can be understood only after electron microscopic investigations. The actual glutamate receptors that are known to be associated with each type of contact at the synaptic ribbon synapse (black bar) are indicated. Ionotropic glutamate receptors (iGluRs) are of two basic types, AMPA and kainate (KA), and are associated with excitatory fast transmission to OFF bipolar cells (OFF bc) and horizontal cells (hc). Metabotropic glutamate receptors (mGluR6) receptors are on ON bipolar cell dendrites (ON bc) and are associated with slow inhibitory transmission in which G proteins and second messengers are involved in transduction. Horizontal cell processes (hc) are thought to feed electrical information back to the cone synaptic area at hemi gap junctions (hemi, crosses) and to bipolar cell dendrites at GABA receptors (γ, arrows).

This parallel set of visual channels for ON and OFF qualities of the image are fundamental to our seeing. Our vision consists of the contrast of one image against a different background. For example, we read black letters against a white background, actually thereby using the OFF channels started in the retina. In the retina, the parallel bipolar channels are maintained by segregated and parallel inputs to ganglion cells. The architecture of the inner plexiform layer in fact becomes demarcated early in development for the segregation of synaptic input to parallel ON and OFF ganglion cell pathways. In the upper inner plexiform layer (called sublamina a), connections occur only between OFF iGluR-bearing bipolar cells and OFF ganglion cells; in the inner part of the inner plexiform layer (sublamina b), the ON mGluR-bearing bipolar cells contact solely ON ganglion cells (figure 6.8).

Figure 6.8.  

The synaptic contact of the bipolar cell dendrites and the cone synaptic pedicle determines whether the signal carried by the bipolar cell stream is detecting light-on-dark (ON pathway) or dark-on-light (OFF pathway) to the ganglion cells. The former pathways are initiated at mGluRs and the latter at iGluRs with the cone synapse. The mGluR-containing bipolar cells send their axons to lower sublamina b of the inner plexiform layer, while the iGluR-containing bipolar cells have axons ending in upper sublamina a of the inner plexiform layer, thereby continuing segregation of the pathways into the connections with ganglion cells. The receptive fields of mGluR containing cells are ON center (depolarizing, upward-pointing response) with a surround inhibition that is OFF (hyperpolarizing, downward indentation). The receptive fields of iGluR bipolar cells is OFF center (hyperpolarizing, downward pointing) and has a surround of the opposite or ON polarity.

Thus, the parallel series of ganglion cells have been developed to receive those segregated bipolar inputs. This has happened particularly strikingly in the mammalian retina, where the same morphological type of ganglion cell has been split into two subtypes: one for the ON pathway and the other for the OFF pathway. In the cat retina, these are known as ON or OFF center alpha cells and ON or OFF center beta cells. In the human (primate), these ganglion cells are known as ON or OFF center P cells (because they project to the parvocellular layers of the lateral geniculate nucleus) and ON or OFF center M cells (because they project to the magnocellular layers of the lateral geniculate nucleus). The beta and P cells are for carrying ON and OFF views of the image to the cortex for fine detail discrimination, while the alpha and M cells inform on larger-sized, fast action ON and OFF images.

If the retina were simply to inform the brain concerning these opposite contrast images, one could imagine the resultant vision to be rather coarse-grained and blurry. How do we get our precise edges to the images and our ability to read and focus on the finest detail? This process of honing of the image and putting boundaries on it also starts in the retina and even at the first synaptic level. There are horizontal cells at the outer plexiform layer that are making their play at the ribbon synapse of the photoreceptor terminal. Here the horizontal cells receive their excitatory input from the photoreceptors; in actual fact in all species, the horizontal cells receive only input from cones. And they receive input from a lot of cones, so their collection area or receptive field is very large. Their collective input gives them a large hyperpolarizing slow potential response following the time course of the light ON. The size of their receptive field is very large, not only because of the large number of cones with their individual small receptive fields summating but also because horizontal cells are joined, one to another, at electrical junctions known as gap junctions. Thus, a whole sheet of cells have their membranes potential sitting at the same hyperpolarized level, and their response to light is consequently very large in area. [The receptive field is orders of magnitude large that that of a single cone and even that of the single bipolar cell, which receives input from a handful of cones and thus has a medium-size receptive field.] Remember that the bipolar cell receives either excitatory input and thus responds like the photoreceptor and horizontal cell and has a hyperpolarizing response (due to iGluRs) or gets an inhibitory input (due to mGluRs) and gives a depolarizing response to light.

So a single bipolar cells with its hyperpolarizing (OFF) or depolarizing (ON) light response would carry a fairly blurry, large-field response to its ganglion cell were it not for the horizontal cells adding an opponent surround that is spatially constrictive, puts an edge around the field, and gives the bipolar cell what is known as a center surround organization (figure 6.9). The bipolar center would be of one sign, i.e., either ON in the center, or OFF in the center, and the horizontal cell by a feedback mechanism adds an OFF or ON surround, respectively. There are two means by which the horizontal cell can add the opponent surround: either by synapsing, directly on the bipolar cell at a chemical synapse, which seems to occur in some species, or by feeding back information to the cone photoreceptor itself, and this information then feeds forward to the different varieties of bipolar cell making contact with that cone. Feedback to the cone itself is now thought to take place by a very novel electrical synapse consisting of half a gap junction. Hemi gap junctions are thought to change the ionic environment at the photoreceptor ribbon synapse and cause membrane changes in the cone photoreceptor and thence in the bipolar cell dendrite—a complicated circuit that still is a subject of hot debate in the retinal research community. Horizontal cell function in general has occupied many vision researchers for decades, and much is now known of the role of these cells in the organization of the visual message. Horizontal cells are influenced by more than photoreceptors that have input to them though. There is neuronal feedback from inner to outer plexiform layer influencing horizontal cell activity as well. These feedback signals are mostly chemical, coming from neuroactive substances such as dopamine, nitric oxide, and even retinoic acid. The end result is that horizontal cells modulate the photoreceptor signal under different lighting conditions in addition to shaping the receptive field of the bipolar cell response. In species in which color signals are carried by ganglion cells, the horizontal cells have a major influence on the bipolar cells, often making them color coded; again, this is all thought to take place through feedback circuits to the cones at the first synaptic level.

Figure 6.9.  

Cat retina (and most mammalian retinas) contains two morphological types of horizontal cell (A-type HC and B-type HC), but they serve the same purpose of interconnecting and modulating responses of photoreceptors and bipolar cells. Receptive fields of horizontal cells are very wide owing to electrical coupling between cells at gap junctions, so the spread of membrane potential is hyperpolarizing (downward pointing) and large in extent. The model shows how horizontal cells feed back their wide field responses (black arrows, minus signs) to the cones to influence the bipolar cell response in the form of an inhibitory surround response to their OFF or ON center photoresponse.

Now we have learned that the horizontal cell is responsible for adding the surround mechanism to the bipolar cell receptive field. This mechanism allows the two sets of cone bipolar cell channels (i.e., ON center and OFF center) to transmit their center-surround receptive field organization to the ganglion cells with which they synapse in the inner plexiform layer. The ON and OFF center ganglion cells thus have concentric receptive field organization that is often modeled as the sum of two Gaussian curves in a “Mexican hat” shape, where membrane potential and center size are the “peak” and the much wider surround of the opposite membrane potential direction is the wide “brim” of the hat (figure 6.10).

Figure 6.10.  

Ganglion cells are morphologically similar but are split into two subtypes or paramorphic pairs in mammalian retinas. The types that branches in sublamina a of the inner plexiform layer receive input from iGluR-containing bipolar cells and transmit messages concerning dark on light (OFF responses). The subtypes of ganglion cell that branch in sublamina b of the inner plexiform layer receive input from mGluR-bearing bipolar cells and transmit a message concerning light on dark images (ON responses). The message that is transmitted to the brain is a burst of spikes when the light is present for ON center ganglion cells or a burst of spikes when the light terminates for OFF center ganglion cells (middle traces). Receptive fields are the “Mexican hat” shape with depolarizing membrane potentials (upward going) for ON center cells and hyperpolarizing (downward going) for OFF center cells. Both ON and OFF ganglion cells have large and strong inhibitory surrounds (downward or upward responses) in the “brim.”

Recent research has shown that amacrine cell circuitry in the inner plexiform layer also adds information to the surround of the ganglion cell, possibly sharpening the boundary between center and surround even further than the horizontal cell input does. The pair of ON- and OFF-center, concentrically organized types of the P (beta) and M (alpha) ganglion cells are highly developed in retinas of mammals with area centrales or foveas. In the human retina, these two types of ganglion cell are extremely well developed and form the major output of the retina to the higher visual centers (figure 6.11). The ganglion cells of the fovea are the ultimate type of P cell. They are called midget ganglion cells because they have the minutest dendritic trees in a one-to-one connection with a single midget bipolar cell. The midget bipolar to midget ganglion cell channel carries information concerning a single cone to the brain. Because only a single cone is involved, we know that the center response of such midget bipolar and ganglion cells will be either ON or OFF center to red or green cone messages. As in the case of bipolar cells that collect from several cones (diffuse bipolar cells), midget bipolar cells come in the ON and OFF variety. Thus, from the center of our focus, the fovea, a dark on light (OFF) or light on dark (ON) message is sent to the brain for each cone. If there are 200,000 cones in the central fovea, then 400,000 midget ganglion cells are carrying their message to the brain. And the message carries both spatial and spectral information of the finest resolution, because the message from each cone is both spectrally and spatially opponent.

Figure 6.11.  

A drawing, based on an original from Polyak (1941), showing the neurocircuitry of the fovea in the primate retina. Midget or P cell pathways consist of a single cone, two midget bipolar cells, and two midget ganglion cells. Because P cells carry information from only one cone, it will also be spectrally tuned. Red and green cones pass either ON center/OFF surround information or OFF center/ON surround information concerning which are both spectrally and spatially opponent (small red and green circles and rings). Blue cones have their own pathway through a dedicated blue ON center bipolar cell feeding to the lower dendrites of a bistratified blue/yellow ganglion cell type. The yellow message carried to the top tier of the bistratified ganglion cells dendrites comes from a diffuse bipolar cell (yellow) that contacts green and red cones. M ganglion cells of the fovea carry a message from diffuse ON center or OFF center bipolar cells (orange and brown bipolar cells) and form the parallel OFF and ON center, achromatic channels (gray and white circles and rings) concerned with movement and contrast to the brain.

Electrical recordings indicate that several varieties of ganglion cell do not appear to have the concentric organization of those described above, particularly in retinas with a nonfoveate organization. These would be in most nonmammalian species and in mammalian species that have a visual streak organization to their retinas. These latter species have a great deal more feature detection going on in the retina itself rather than postponing this finer honing of the visual message to the brain. Such species have really well developed directionally selective, motion selective edge detectors and dimming detectors already in their retinal ganglion cell responses. Also, it will be noted by the perceptive that blue cones have not been mentioned yet. The message concerning blue light is carried by a special pathway of bipolar cells to a bistratified ganglion cell type in the retina. For some reason, the blue cones are not part of the dual ON or OFF pair of midget bipolar/ganglion cell channels described above. The blue system is an older system in evolutionary terms. Blue cones are found in just about all species retinas. The absorption peak (428 nm) is very different from that of the red 563 nm and green 535 nm cones and so are the opsins. Color contrast between blue and the others is very strong. By contrast, the red and green are similar, and this recent evolutionary split permits fine color discrimination in the appropriate spectral regions. Even the red and green opsins are so similar in molecular design that we cannot yet make an antibody against each to separate them by immunohistochemical staining techniques. In mammals, blue and green cones are the common cone types. Color vision in most nonprimate mammals consists of contrasting blue (light on dark) against orange/green or dark shapes (dark on light) so the blue system has a more spread apart distribution and a convergent and divergent set of neural interactions. The blue system ganglion cell responds with a blue ON response and a large receptive field and gives a yellow OFF response in a spatially coextensive broad receptive field. In other words, one ganglion cell is carrying an opposite but superimposed message to the brain concerning blue and yellow—quite a different organization from the red and green midget systems!

To understand more about the organization of the ganglion cell receptive field, whether it be about concentric organization or direction and motion detection, we need to study the organization of the inner plexiform layer in detail. We need to find out what are the roles of those 22-plus different varieties of amacrine cell making their synaptic connections with 20 or more different types of ganglion cells here. Although it was clear from the time of Cajal's description in the nineteenth century that amacrine, ganglion, and bipolar cell dendrites and axons were organized into layers of intermeshing processes (Cajal divided the inner plexiform layer neuropil into five different strata of layering of processes), we could not immediately figure out what this meant and what sort of synapses were going on between the layered and opposed processes. Using the electron microscope, we began to unravel this neurocircuitry. We now know something of the input/output relationships of nine different types of bipolar cell, 14 different types of amacrine cell, and eight different types of ganglion cell, so we are halfway to the goal of understanding the neural interplay between all the nerve cells of the retina. In the neuropil of the inner plexiform layer with the higher magnification of the electron microscope, we can recognize bipolar cell axons by their containing a synaptic ribbon and the amacrine cells on their vesicle clusters at synaptic sites (figure 6.12).

Figure 6.12.  

The synaptic connections of the different nerve cell types in the inner plexiform layer has to be studied by electron microscopy. Bipolar cells (BC) are detected by their containing a synaptic ribbon pointing to synaptic output sites, commonly consisting of a ganglion cell dendrite (GC) and an amacrine cell process (Am). Amacrine cells synapse on bipolar cells, other amacrine cells, and ganglion cells (clusters of vesicles at synaptic sites). The types of receptors for the neurotransmitter glutamate in bipolar cells and glycine and GABA in amacrine cells have now been described for such synaptic circuitry. For glutamate transmission, mGluRs and the iGluRs AMPA and NMDA are present on ganglion cell and amacrine cell dendrites. For amacrine cells, glycine receptors and GABA_A, GABA_B, and GABA_C have been detected on bipolar, amacrine, and ganglion cell dendrites.

Much has been learned of the types of neurochemicals that are contained within different amacrine cells and the organization of receptors at the different synapses. Glutamate receptors between bipolar cells and ganglion cells are both NMDA and AMPA types, and amacrine cells are about equally divided between those that use glycine and those that use GABA inhibitory neurotransmitters.

Glycinergic amacrine cells are typically small field. Their processes are usually full of appendages and lobules that are able to spread interactions vertically across several of the five strata in the inner plexiform layer. Glycinergic amacrine cells mostly make a lot of synapses between bipolar axons in either the OFF layer or the ON layer and feed forward synapses to ganglion cell dendrites and other bipolar and amacrine cells. Some glycinergic amacrine cells cross the two major OFF and ON sublaminae of the inner plexiform to provide interconnections between ON and OFF systems of bipolar and ganglion cells. The most famous of these small bilaminar, glycinergic amacrine cells is called the AII cell. The AII cell together with a wide-field GABAergic amacrine cell called A17 is pivotal in the connectivity of the rod pathways to ganglion cells and in the circuitry of rod or dim light vision in the mammalian retina. Both these “rod” amacrine cell types are not found in nonmammalian species or in mammalian species that are diurnal and contain very few rods in their retinas (squirrels, for example).

When we considered the ON and OFF channels and their separate connectivity through different receptor contacts at the cone synapses (mGluR- and iGluR-driven bipolar cells) and their spatial segregation to the ON and OFF ganglion cell types, we neglected to talk about the connections of the mGluR-driven rod bipolar cells. The reason is that, unlike the direct pipeline from cone to ganglion cell for the cone-driven ON or OFF channels, rod bipolar cells do not synapse with ganglion cells at all. These bipolar cells, and there is only one type that contains an mGluR receptor and hence is an ON type, uses the glycinergic AII and the GABAergic A17 amacrine cells as intermediaries to get rod information to ganglion cells. The small field AII cell collects rod messages from a group of approximately 30 axon terminals in the ON sublamina of the inner plexiform layer and passes this rod sensitivity depolarizing message through a gap junction with ON cone bipolar cells to ON ganglion cells. At the same time, the AII passes rod information to the OFF system cone bipolars and ganglion cells via direct chemical inhibitory synapses at their lobular appendages in the OFF sublamina of the inner plexiform layer. The AII amacrine cell seems to have been developed in the rod-dominated parts of mammalian retinas as an afterthought to the original direct cone-to-ganglion cell architecture. The rod system has inserted the new amacrine cells to piggyback on the original cone system connections (figure 6.13).

Figure 6.13.  

Summary of the rod pathways through the mammalian retina and how the AII amacrine cell piggybacks on the cone pathways because the rod bipolar cells do not have direct synapses on a ganglion cell of any type. The AII amacrine cell receives input from rod bipolar axons (ON rb) and passes that information to OFF cone bipolar (OFF cb) and OFF ganglion cells (OFF GC) at glycinenergic chemical synapses (large open arrows). The AII amacrine cells contacts ON cone bipolar cells (ON cb) at gap junctions (large arrow, asterisk), and the electrical message is passed to the ganglion cells that those cone bipolar cells contact (ON GC, large arrow).

At the same time, the A17 amacrine cell is also collecting rod messages from, in this case, thousands of rod bipolar axons to amplify and feed this information back into the AII to cone bipolar–ganglion cell route. How it does this is not completely understood yet, although we know this GABAergic wide-field amacrine type uses the novel GABAc (rho) receptor to feed back on rod bipolar axons, thus presumably influencing the whole state of the rod system. The rod pathway with its series of convergent and then divergent intermediary neurons in the retina is clearly well designed to collect and amplify on widely scattered vestiges of light in night and twilight vision. The rod pathways are solely ON system neural chains that is, designed to detect light on dark only.

GABA is commonly the neurotransmitter used by wide-field amacrine cells that stretch laterally across the inner plexiform layers for hundreds of microns and can interact with hundreds of bipolar cells and many ganglion cells. Such amacrine cells are usually stratified in one of the five different strata of the retina in beautifully organized mosaics of elegant meshworks of dendrites (figure 6.14).

Figure 6.14.  

A, An immunostained image of rod bipolar cells immunostained with antibodies against protein kinase c (PKC). B, Small-field bistratified AII amacrine cells are immunostained with antibodies to parvalbumin (PV). C, Dopamine-containing cells are immunostained with antibodies to tyrosine hydroxylase (TOH) as seen in a flat mount of the retina. Thousands of dopamine cell processes cross each other and make a dense network of processes in the top part of the inner plexiform layer, to synapse on various cell types, among them the AII amacrine cell. D, Two mirror symmetric amacrine cell populations, known as starburst cells, are immunostained for their acetylcholine neurotransmitter (ChAT) and seen in flat mount of retina. One set of starburst cells sits in the ganglion cell layer, and the other sits in the amacrine cell layer. Their respective dendritic plexi run and synapse in sublamina b and sublamina a. Starburst amacrine cells are thought to influence ganglion cells to be able to transmit messages concerning direction of movement in the visual field. These cells are particularly well developed in animals with visual streaks in their retinas. E, A17 amacrine cells immunolabeled with antibodies to serotonin (Ser) and also to GABA. A17 cells connect rod bipolar axon terminals in reciprocal GABAc receptor–activated circuits across the entire retina. (E from Vaney DI: Many diverse types of retinal neurons show tracer coupling when injected with biocytin or Neurobiotin. Vision Res 1998; 38:1359–1369.)

Their synaptic interactions would be with other neurons branching in the same layer or stratum. Frequently, GABAergic amacrine cells with rather simple, sparsely branching dendritic trees connect to neighboring homologous amacrine cells by gap junctions, thus increasing their field of influence and speed of transmission of signals across large areas of retina. In many instances, wide-field GABAergic amacrine cells send out even farther-reaching axonlike processes to other layers of the inner plexiform or to the inner nuclear and outer plexiform layers and the blood vessels of the retina. Most GABAergic amacrine cells in the retina contain at least one other neuroactive substance besides GABA. The secondary neuroactive substance is usually acting as a neuromodulator rather than a fast neurotransmitter. The peptides substance P, somatostatin, vasointestinal peptide, and cholecystokinin have been associated with such amacrine cells, as have serotonin, dopamine, acetylcholine, and adenososine. Nitric oxide has been localized to certain wide-field amacrines as well. Peptidergic, nicotinic, and muscarinic receptors in addition to different varieties of GABA_A, GABA_B, and GABA_C receptors have all been found on ganglion and bipolar cells, indicating that such neuroactive agents in amacrine cells are influencing the organization of the ganglion cell receptive field. Most of these neuromodulators are not released at conventional synapses, though, and their release is thought to influence neurons even at a distance by diffusion in a “paracrine” manner. We have discovered that frequently, the role of such neuromodulators in the retina is to influence the retinal circuitry under changing lighting conditions or even to stabilize activity to different times of day in the circadian clock.

Dopamine is released from a specialized amacrine cell on stimulation of the retina by intermittent light flashing. Dopamine influences horizontal cell activity by uncoupling the gap junctions between these cell types, thus reducing their effective receptive field sizes and consequent expression of surround size on bipolar cells. Dopamine also directly affects the glutamate receptor on horizontal cells so that the amplitude of the photoresponse is reduced under different light conditions. In the inner plexiform layer, dopamine is particularly effective on gap junctions again, but this time on the gap junctions that join the AII amacrine cell in large coupled networks in the dark state. Light causes dopamine release, which by diffusion to the lower inner plexiform neuropil affects the gap junctions between neighboring AII cell distal processes. This uncoupling of AIIs from their dark network makes the effective field of influence of the rod system amacrine cell much less significant in the light. Any large-field rod pathway interference in the direct cone pathways is thereby removed. Similarly, another wide-field amacrine cell branching in the center of the inner plexiform neuropil releases nitric oxide, which uncouples the AII cell and the ON cone bipolar cell system at that particularly gap junction (figure 6.15). Again this has an additive effect to that of dopamine of removing the rod system neurons from the cone system direct pipeline to ganglion cells, thereby helping to make the operative image in bright light conditions a high-contrast, high-acuity signal (figure 6.16).

Figure 6.15.  

Nitric oxide–containing amacrine cells form a beautiful network of processes across the middle of the inner plexiform layer to influence various nerve cells penetrating this network and making synapses or gap junctions in this area of neuropil. The NO-containing amacrine cells also contain GABA as their conventional inhibitory chemical messenger.

Figure 6.16.  

The retinal circuits that are thought to influence the ganglion cell message to the brain concerning low light vision and bright light vision. The AII amacrine cell is driven in the dark to pass messages to both ON and OFF center ganglion cells about very low light levels. The whole network of involved neurons is designed to be very wide-field and to collect every glimmer of light through convergence, divergence, and further convegence of information. Gap junctions between AII cells allows the spread of rod-driven messages over large areas of retina together with the A17 cells' convergent input (not shown in this figure). In the light, the dopamine amacrine cell releases dopamine, and this uncouples the AII cell from its electrical network; thus, information through the cone bipolar to ganglion cell pathways is less convergent, narrower in field, with sharp edge contrast and concerned more with only bright, cone-initiated messages. At the same time, amacrine cells containing nitric oxide (NO) as a neurotransmitter release this nitric oxide to sever the gap junction between AII amacrine cells and ON cone bipolar cells, thus contributing to the high-contrast, narrow cone-driven messages reaching ganglion cells.

So, as can be seen from the above broad sketch of retinal circuitry and modes of action of the nerve cells into systematically organized pathways, the retina is surprisingly more complex than was initially thought, and its function takes on an increasingly more active role in visual perception than was originally thought. Although we do not fully understand the neural code that is sent as trains of spikes through the ganglion cell axons to the central brain areas, we are coming close to understanding how consorts of ganglion cells respond differently to different aspects of the visual scene and how receptive fields of influence on particular ganglion cells are constructed. It seems that much construction of the visual image does take place in the retina itself, although the final consequence, “seeing,” is indisputably done by the brain. Are we finished with the retina? Do we know all that we need to know to understand the first stages if vision, or are there more surprises on the horizon? Earlier surprises included finding that so much information transfer rested on electrical connections over standard chemical synapses. The major rod pathway is dependent in large part on electrical connections to get information through the retina. Some other fast-acting amacrine pathways also use direct access to ganglion cell at gap junctions. Neuromodulators and gases change the milieu of circuits of neurons but act from a far distance by diffusion rather than at closely apposed synapses. Again this is a new and surprising concept compared to our older view that all neural interactions took place at specialized isolated patches of membrane apposition known as synapses by packaged neurotransmitter substances. The most recent surprise has been the revelation of a new ganglion cell type in the retina that appears to be able to function as a giant photoreceptor itself and needs no rod or cone bipolar inputs to drive it. This ganglion cell's membrane contains photoactive molecules known as melanopsins, so this ganglion cell can respond to light in the absence of neural circuitry.

In conclusion, we have certainly come far in our understanding of the organization of this piece of the brain, the retina, in the last half century. However, we cannot rest on our laurels. We still undoubtedly have much more to learn about how the retina creates the first steps of our visual perception—“seeing.”