III. Bipolar and Amacrine Cells to Retinal Ganglion Cells
As I detailed in the previous section, bipolar cells receive input from photoreceptors in the outer plexiform layer of the retina. The retinal layer in which they pass this information onwards is called the inner plexiform layer (IPL), where the receiving processes of retinal ganglion cells (RGC) are located. RGCs are much bigger than photoreceptors or bipolar cells, and their size is part of the reason why they function using action potentials. Their axons make up the innermost layer of the retina, but for the thin internal limiting membrane. They collect into what is known as the optic nerve (Cranial Nerve II), which exits the eye through the optic disc and travels into the deeper regions of the brain.
Like photoreceptors, bipolar cells use glutamate to communicate with RGCs. There are approximately 35.68 million bipolar cells, connecting to about 1.12 to 2.22 million RGCs. This difference in numbers indicates a degree of convergence; more than one bipolar cell must connect to a solitary retinal ganglion cell.
A diagram showing numerical convergence as signals traverse the layers of the retina
The IPL also contains the dendrites of a second type of cell that feeds onto RGCs; the amacrine cell. Amacrine cells are a type of interneuron (a relay cell of the nervous system). A cell of this kind can be assigned to one of at least 33 subtypes, depending on its location, the shape it takes, and the role that it plays. Except for a few (like the starburst amacrine cell) most of these amacrine subtypes are not yet well-understood at all. What I can say here, in very broad terms, is that the amacrines modulate the signalling taking place between bipolars and RGCs, releasing different kinds of excitatory and inhibitory neurotransmitters at the synapses they see fit to do so.
Of course, the current state of scientific understanding where amacrines are concerned isn't quite so vague as that. Generally, it's thought that their inputs create units of functional specialization. One such unit, for example, created by networks of starburst amacrines, contributes to the foundation underlying our ability to detect an object's direction of motion.
A collection of just five amacrine cells, to give you an idea of how much they can differ in their morphologies
Although RGCs are much more straightforward to characterize anatomically, because their axons rope backward through the eye and into the brain, a microscope will show you that they are actually very varied in appearance. In vertebrates, so far, 5 main classes are recognized. Each class of RGC requires a different mix of inputs from amacrine and bipolar cells to be stimulated into action. Therefore, it's supposed that each RGC type participates in a different circuit, sending different types of information - like depth, color, or movement - separately into the brain. This idea, that the retina separates out information streams and sends it out via specialized RGC classes, is called parallel processing.
Like most neurons, RGCs exhibit a constant level of background firing, even when they’re at rest. The circuits they take part in recognize and act on relative increases and decreases in this firing rate.
A series of accurate graphs showing the firing rates of the hypothetical RGC, Stuart
The road that has led to our current understanding of RGCs - before we knew there were five distinct classes of these cells, before we even knew what role they served - was begun by famous experiments performed on anesthetized cats in the 1950s.
At that time, a scientist called Kuffler decided to shine a small circular spot of light onto a particular region of a cat’s retina. When he directed the beam onto just the right retinal region, he noted a burst of action potentials in the retinal ganglion cell he’d stuck a micro-electrode into. This cell he described as an ON-center ganglion cell.
When he expanded this circle of light, however, the rate at which action potentials were produced fell below the baseline. Tinkering around some more, Kuffler found that the signals the cell gave off were almost completely shut off when the central region of the spot of light was darkened, leaving only a surrounding annular ring of light. This pointed to the presence of an antagonistic surround; a ring surrounding the ON-center which could inhibit the signal produced by that center if provoked by the same stimulus (light, in this case).
Portrait of Kuffler with a cat being shone into his eye
Kuffler had discovered and coined the concept of the receptive fields. This is the area within the eye's field of vision that an individual cell in the visual pathway (an RGC in this case) is sensitive to. If you've read the previous section, you might realize something that Kuffler couldn't, due to technical limitations at the time: that these RGC receptive fields have something to do with H and D bipolar cells, which connect photoreceptors to RGCs.
If the response generated by an RGC originates with cells located levels below it in the retinal hierarchy, this means that D bipolar cells must also have ON-center receptive fields, like the ON-center receptive fields of the RGCs I've described and pictured below. It also means there must exist an OFF-center ganglion cell which responds best to a small dark spot placed in its receptive field center and is inhibited by darkness produced on the surrounding ring of space; and that it must be integrally connected to H bipolars.
In fact, these two types of ganglion cells - the ON- and OFF-center RGCs - do exist, and are present in approximately equal numbers in the innermost retinal layer.
The electrical responses of an ON-center RGC
How are the effects of that portion of the receptive field appearing above as an annular ring - otherwise known as the 'antagonistic surround' - produced?
It's currently thought that a key role is played by horizontal cells, which connect photoreceptor cells making up the receptive field underneath the annular ring surround to the photoreceptors under the central spot. Horizontal cells absorb glutamate from many rods and cones, and depend on its stimulating effects to release their own neurotransmitter, the inhibitory GABA. They use GABA to provide negative feedback onto rods, cones, and bipolar cells, modulating the signals each is able to release. They're also connected to each other by 'easy-transmission portals' of communication called gap junctions, thus forming networks among themselves which span the area of the retina.
To understand the 'antagonistic surround' and how horizontal cells create it, let me describe one such surround in an OFF-center RGC. Light which falls in the center of such an OFF-center RGC receptive field causes the cell to become inhibited; when it falls on its surrounding ring, it causes excitation. Suppose a horizontal cell links all the photoreceptors of just that surrounding ring in the manner pictured below, and connects at the same time to a H bipolar cell located underneath the center spot. As we know, should light fall on the ring area, the photoreceptors in it will stop releasing glutamate. The horizontal cell, which depends on glutamate, stops releasing GABA onto the H bipolar. This disinhibits the H bipolar, causing it to fire. The connections that the H bipolar has with photoreceptors comprising the central spot and now releasing glutamate, enhances this effect.
On the other hand, when light falls in a concentrated way on the center spot, the horizontal cell is uninhibited and steadily suppresses the H bipolar with its streams of GABA release. The connections the H bipolar has with photoreceptors in the center contribute to its suppression, since they release no glutamate when light is on them.
If you carefully consider this scenario, it becomes evident that when there's diffuse light falling across both the center spot and the annular ring, a 'summation' of signals must occur. This kind of diffuse light stimulus is unlikely to affect the base firing rate of the H bipolar cell very much at all.
An OFF-center RGC receptive field formed by a H bipolar cell. For convenience, I've shown only part of the annular ring on the right as illuminated. The center ring finds itself in relative darkness. This causes a fall in the GABA volume released by the right horizontal cell and therefore a disinhibition and firing of the H bipolar cell.
As is pictured above, the soma - cell body - of a retinal ganglion cell is located in the geometric center of the ring of light that describes its receptive field. The receptive field size itself depends on the number of photoreceptors that comprise it, and thus on where in the retina the RGC is located. The smallest receptive fields exist in the fovea, the region where visual acuity is the highest and in which a bipolar cell may be connected to a solitary cone. Receptive fields get larger as you move into the retinal periphery.
The cell bodies of different categories of RGCs each form a regular mosaic. The dendritic trees that project from these cell bodies cover the retinal area in a very energetically efficient manner, without leaving gaps, but also without creating significant overlap. The different categories form such retinal mosaic fields individually, creating at each location on the retina those independent - ‘parallel’ - information pathways that I mentioned above.
Here is the mosaic formed by one such pathway, the ‘ON-center’ RGC channel:
An ON-center RGC mosaic created in a cat's retina. Notice how there is no significant overlap and the cells are regularly dispersed, leaving no gaps
Almost all RGCs have a receptive field in the shape of a circle, like the kinds I described and pictured above. However, some RGCs differ consistently from others in what information they receive, how they are shaped, and where they connect to. These differences have given rise to the following 5 RGC classes, some of which you'll see again in coming sections:
1) midget RGCs (P cells) - As a group, these form the largest subtype, encompassing nearly 80% of all RGCs in a given retina; but individually, they're one of the smallest sized cells, which is where they get their name from. They conduct their signals relatively slowly and get information from bipolar cells that connect to only a few photoreceptors located in the fovea, the area of the retina giving the most clearly defined images. Their receptive fields are the circles I've described above, and can come in either the ON- or OFF-center flavors.
2) parasol RGCs (M cells) - These RGCs are big and have unfurled dendritic trees which make them look somewhat like an umbrella. They constitute roughly 10% of RGCs. Their receptive fields are much larger than those of the midget cells, and consequently need many more rods and cones to form them, giving them quite low definition. Like P cells, they use both ON- and OFF-center responsive cells.
3) Bistratified RGCs (K cells) - These also constitute around 10% of all RGCs. K stands for koniocellular, which means 'small as dust', and is the reason why they're only a recent discovery. They have very large receptive fields, but are different in that they have no annular rings around them. In fact, K cells always seem to have an ON center receptive field when found above a group of blue cones, and an OFF center field above green and red cones.
4) Intrinsically photosensitive RGCs (ipRGCs) - These are a pretty cool population of RCGs, though a very small one, comprising just 2 percent of the total number. They have their own photosensitive pigment called melanopsin, allowing them to respond to light without any connection to rods or cones. They are involved in a non-visual pathway that is nonetheless dependent on light - the pathway that helps align the circadian rhythms. They may be a part of the reason why organisms born without any photoreceptors at all still exhibit regular sleep-wake cycles.
5) RGCs linked to the saccades - These cells also don't take part in a pathway that gives you your conscious world-image. Instead, they project to the superior colliculus, a brain area which is involved in the dorsal stream of vision and in blindsight, both of which topics I'll cover in upcoming sections.
Put together, the different classes of RGCs have receptive field networks that work separately, collecting their own kinds of information to form visual and non-visual, light-dependent neural circuits.