undefined
Formation and Function of Retinal Center-Surround Receptive Fields
Todd Anderson
9.3.2002
C26
Introduction
Although the human retina contains hundreds of millions of photoreceptors, only a million fibers exist in the optic nerve to carry their signals towards higher processing centers in the cortex. Thus the retina must be processing much of the information obtained at the level of the photoreceptors, sending only this purified and relevant data on through the optic nerve. One of the main processing methods employed by the retina is the center-surround field – the subject of this paper. I will first present the methods of studying and characteristics of center-surround fields, and then explain how these characteristics relates to their higher-level function. I will then describe the mechanism by which center-surround fields are formed, including the molecular mechanisms of signal transmission and regulation.
Characteristics and Behavior
A center-surround field is a circular area of the retina, usually with a diameter of less than 1 mm, in which all of the photoreceptors are connected to a ganglion cell via complex retinal circuitry consisting of horizontal and bipolar cells (see figure three). The ganglion cell is stimulated to fire rapidly if the center of this area is stimulated (either by light, in an ON-center cell, or by the absence of light, in an OFF-center cell), and is inhibited from firing if the same stimulus is presented in the area around the center (the surround).
Ganglion cell center-surround fields were first described by Stephen Kuffler in 1953. Using a small spot of light (less than 100 microns in diameter), he stimulated a part of the retina which turned out to be the center of an ON-center receptive field – the ganglion cell from which he was recording began to rapidly fire action potentials. The frequency of firing slowed slightly with continued stimulus, but remained above the normal background frequency of the ganglion cell. Upon removing the stimulus, the cell stopped firing for a few moments before returning to normal background firing levels. When he moved the spot about 0.5 mm away from the initial stimulus point, the cell ceased to fire until the stimulus was removed, at which point it fired rapidly for a short time before returning to normal background firing frequency (see Figure 1a & 1b).
Kuffler reported that the change between stimulus-activation and stimulus-inhibition of the cell was quite sudden – this point marks the boundary between the center and the surround of the center-surround field (see Figure 1c). Knowing the boundaries of the center and surround, it is possible to achieve maximal activation/inhibition of the ganglion cell by stimulating the center of the field with a center-sized spot (of either light or dark). In a similar fashion, a doughnut-shaped spot sized to perfectly fit the surround can maximally activate/inhibit the ganglion cell. A large spot, the size of the center spot and the doughnut put together (whole-field stimulation), both activates and inhibits the cell – the effects nearly cancel out, but the center generally has more effect on the ganglion cell output. Thus if the ganglion is an ON-center cell, it will fire faster than background during whole-field light stimulation, and less than background during dark stimulation of the same area. The pattern is reversed with OFF-center cells (see Figure 2).
Recent research by Solange Brown et al. (2000) has shown that not all receptive fields are of the smooth dome-shaped variety depicted in Figure 1c. Though most ganglion cells with small receptive fields (those found in the fovea, for instance) did have dome-shaped sensitivity profiles, those with larger receptive fields showed irregular profiles with several points of peak sensitivity. To carry out this research, Brown’s group also developed a new technique for mapping the receptive field boundaries. Instead of using a very small spot of light, they projected a 15x15 grid of blocks with randomly-generated intensities. They switched to a new grid every 14.2 ms, generating a movie which was than processed to find the level of correlation between stimulus intensity at specific points on the receptive field and activation of the ganglion cell. The resulting map is very similar to one generated using the older single spot protocol, though this new method can map an entire receptive field in 7.5 minutes, versus the 73 minutes required for the older method.
Function
The possible functions of center-surround fields are surprisingly numerous considering their relatively simple concept. Contrast is the thread which unites all the functions, as center-surround fields are specialized to detect local contrast. Because the center is completely stimulated and the surround partially not stimulated, an ON center-surround field which happens to be located so that the edge between an area of lightness and an area of darkness falls just on the boundary between the center and the surround (with the light part covering the center) will be more activated than its fully-covered neighbors. Likewise, on ON-center field in which the center is just covered by a dark stimulus will be more inhibited than its neighbors which are fully covered by the dark, as only the inhibitory surround is stimulated (see Figure 3).
Using this contrast-detection ability, center-surround fields serve to sharpen the image transduced by the photoreceptors. In effect, when light parts are next to dark parts, light parts are perceived as lighter and dark parts are perceived as darker. Figure 4 shows an application of this technique in photographic image processing. The cactus picture on the left is unenhanced, while the one on the right was run through Adobe Photoshop’s unsharp mask filter, which is based on the center-surround premise. The detail in the right hand image has been brought out by the filter, almost as if a layer of grime has been washed off the photograph.
Figure 1 a & b: Schematic drawing of the two types of center-surround field, evenly illuminated with a medium (gray shading) stimulus except for small spots of light (white circles). The pluses indicated where light stimulus will elicit a positive response (in the ganglion cell), and the minuses where a dark stimulus will elicit a negative response. These responses are represented as traces of action potentials fired by the ganglion cell before, during, and after the stimulus. c: The arrow on the left shows the path of a light stimulus spot across a ON-center receptive field. The arrow on the right shows the state of the ganglion at corresponding points, showing the sensitivity profile of the receptive field (above the arrow is activation, below is inhibition).
Figure 2: The firing frequency of both ON- and OFF- center ganglion cells in response to center stimulation, surround stimulation, and whole-field (full) stimulation with both light and dark stimuli. During light center stimulation (with a medium gray projected onto the surround), ON-center cells are quite active while OFF-center cells are quite inhibited. Upon stimulus removal, ON cells are inhibited for a short time while OFF cells are active. During dark center stimulation, this pattern is reversed, as it is when the surround is stimulated (with a medium gray projected onto the center) with both light and dark. With whole-field light stimulation, the ON cell is slightly more active than the OFF cell, though the firing frequency is not very high. The reverse is true with whole-field dark stimulation.
Figure 3: Contrast detection by center-surround fields. Field A is fully stimulated in the center and not fully stimulated in the surround, so its ganglion is more activated than that of its neighbor field B which is equally stimulated in the center and in the surround. Thus the response curve (showing the activation of ganglia from left to right) jumps up just before the edge between the light and dark parts of the stimulus, corresponding to the location of field A. Similarly, field C is not at all stimulated in its center, but is partially stimulated in its surround. This means field C is more inhibited than its neighbor field D, which is equally stimulated in its center and surround. There is a corresponding dip in the response curve at field C. Because the center has the greatest effect on ganglion output, the left portion of the response curve is elevated as field B is stimulated in the center. Similarly on the right, field D is not stimulated in the center, so the response curve is depressed. The bottom box shows a dramatization of the perceived stimulus – that is, the perception of the viewer of the stimulus in the upper box.
Figure 4: A photographic image sharpened using Adobe System’s Unsharp Mask filter for Photoshop. The filtering software achieves an effect similar to that of the center-surround fields in the retina – when dark parts are next to light parts, the dark parts get darker and the light parts get lighter. Details are highlighted in the sharpened image, an effect especially obvious along the blades of the cactus.
Another function of center-surround fields is spatial filtering. Groups of fields, the ganglion cells of which are connected either in the retina by amacrine cells or in higher processing centers like the lateral geniculate nucleus, can be selectively activated by certain spatial frequencies (see Figure 5). Spatial frequencies are patterns of blurry-looking, alternating light and dark bands – the bands are separated by a characteristic distance (usually expressed in degrees or in mm across the retina), the inverse of which is the frequency of the pattern (bands per degree/mm). When the spatial frequency matches the alternating center-surround layout of a certain group of ganglion cells, those cells will be either exceedingly activated or exceedingly inhibited (depending on whether the phase of the pattern is such that light parts overly the excitatory parts of the center-surround fields).
Figure 5: Spatial filtering function of center-surround cells. The group A fields are sized and arranged in such a way that their excitatory centers fall on the bright areas, and their inhibitory surrounds mostly on the dark areas, of the spatial pattern. Thus, the ganglions connected to the fields in group A are nearly maximally active for this spatial frequency at this phase. In contrast, the group B fields are arranged to be sensitive for another have spatial frequency. For this frequency, some ganglion cells connected to the fields in group B are activated (the 1st from the left), some are neutral (the 2nd from the left), and some are inhibited (the 2nd from the right).
Because they detect local contrast, center-surround fields are also used in light adaptation. They allow vision over a wide range of light levels, as contrast detection relies on the relation between light and dark parts in a local area, not the absolute levels of lightness or darkness. For example, a patch of gray surrounded by a lighter shade will appear darker than that same patch of gray surrounded by a darker shade (see Figure 6). This is mainly due to the surround antagonism characteristic of center-surround fields, where stimulation in the surround inhibits activation of the ganglion. Thus if one patch is surrounded by a darker patch, the center-surround fields will be stimulated in the surround more than in the center, inhibiting the ganglion and making the first patch seem darker. Similarly, if the surrounding patch is darker than the middle patch, the surround will not be stimulated and the as center, so the ganglion will not be inhibited and the middle patch will seem brighter.
Figure 6: Light adaptation is partly due to center surround fields. In the top box, the center gray square on the left appears darker than the one on the right, while in the bottom box both squares appear the same. The shade of all smaller gray squares is equal.
Mechanism
What is the mechanism which accounts for these remarkable characteristics and functions? Connections from the photoreceptors to bipolar cells, from horizontal cells to bipolar cells, and from bipolar cells to ganglion cells give rise to these phenomena. All photoreceptors (rods and every type of cone) hyperpolarize in response to light, though they do not fire action potentials – photoreceptors exhibit a graded response, meaning the amount of light hitting the receptor is measured in a gradation rather than in the all-or-nothing manner of action potentials. In response to light, the membrane potential of the receptor cell becomes more negative due to the cessation of Ca2+ influx as voltage-gated Ca2+ channels close. In the dark, photoreceptors release glutamate at their synapses with both bipolar and horizontal cells (see figure 6). In the light, as they hyperpolarize, they release less glutamate. Both the bipolar and the horizontal cells exhibit glutamate-dependent responses; these responses will be dealt with in following paragraphs – for now, I’d like to concentrate on the global architecture of center-surround fields.
Bipolar cells also do not fire action potentials; they use graded potentials to transmit information through their sign-conserving synapse with a ganglion cell. There are two kinds of bipolar cells, as there are two kinds of ganglion cells: ON and OFF.
The ganglion cell fires action potentials in response to the graded potential input of the bipolar cell. Ganglion cells are the only cells in the center-surround system which fire action potentials; as they send visual information to the lateral geniculate nucleus and the occipital lobe – a much greater distance than the cells previously mentioned – they must use action potentials, which propagate signals over long distances much more reliably.
Figure 7: Wiring diagram retinal center-surround fields.
Bipolar Cells
What of these bipolar cells? Bipolar cells exist in two main categories, the same two categories as ganglion cells: ON-center and OFF-center. The difference pertains to the way each type cell reacts to the neurotransmitter glutamate. As mentioned above, photoreceptors release glutamate across their synapses with bipolar cells – ON-center bipolar cells depolarize in response to glutamate, while OFF-center cells hyperpolarize under the same conditions. Since a hyperpolarization of the photoreceptor elicits a depolarization in the ON bipolar cell, this synapse can be thought of as sign-inverting. In contrast, the same hyperpolarization in the photoreceptor elicits a hyperpolarization in the OFF bipolar cell, making this synapse sign-conserving (see figure 7).
The molecular basis for the behavior of ON-center bipolar cells is nearly the same as for phototransduction in rods and cones. Glutamate binds to a glutamate receptor on the bipolar cell membrane, which activates a G-protein. This G-protein binds to and activates cGMP phosphodiesterase, which converts cGMP into 5’-GMP. The Na+ channel in the ON bipolar cell is cGMP gated, so with low cGMP concentrations the Na+ channels close. When light hits the photoreceptors, they stop releasing glutamate, which means that the G-protein is no longer activated. Therefore, cGMP phosphodiesterase is also no longer active, and thus cGMP is no longer being converted. cGMP concentration rises and the Na+ channels open, leading to depolarization of the bipolar cell (see Figure 8).
The OFF bipolar cell has a much simpler molecular basis. OFF bipolar cells have glutamate-gated Na+ channels, which open in the presence of glutamate and hold the cell at a depolarized potential. When light hits the photoreceptor, glutamate release is staunched, and the Na+ channels close, leading to hyperpolarization (see Figure 9).
Figure 8: ON-center bipolar mechanism. The top portion of the figure shows the opposite responses of the photoreceptor and ON-bipolar cell to a light stimulus, with the photoreceptor hyperpolarizing and the bipolar cell depolarizing. The gray box shows a close up of the molecular machinery underlying this effect: in the absence of light, glutamate (GLU) is released into the synaptic cleft, where it binds the glutamate receptor (GLUr), which activates a G-protein (G-pro). G-pro binds to and activates cGMP phosphodiesterase (cGMP PDE), which converts cGMP to 5’-GMP. Since the Na+ channels in ON bipolar cells are cGMP-gated, in the absence of light (and hence of glutamate) the channels are closed and the cell hyperpolarizes. In the light, glutamate release is staunched and cGMP concentration rises, opening the Na+ channel and depolarizing the bipolar cell.
Figure 9: OFF bipolar mechanism. The top portion of the figure shows the similar hyperpolarizing response of the photoreceptor and bipolar cell to a light stimulus. The gray box shows a close-up of the molecular machinery underlying the bipolar cell response: in the absence of light, glutamate (GLU) is released into the synaptic cleft, where it binds the glutamate-gated Na+ channel. The channel opens, depolarizing the bipolar cell. In the presence of light, glutamate release is depressed, the Na+ channel closes, and the bipolar cell hyperpolarizes.
Horizontal Cells
Horizontal cells take input from many local photoreceptors in the surround and relay the information to bipolar cells. They do this indirectly by synapsing onto the center photoreceptors, which are connected to the bipolar cells. Horizontal cells give rise to lateral inhibition (or center-surround antagonism), meaning stimulus in both center and surround inhibits activation of the receptive field bipolar or ganglion cells. They do this using inhibitory neurotransmitters like glycine and GABA. In the dark, surround photoreceptors are semi depolarized, which means the horizontal cell is also semi-depolarized (as the synapse is a sign-conserving one). In this semi-excited state, the horizontal cell releases inhibitory neurotransmitter onto the center photoreceptor, but when light hits the surround photoreceptors, the resulting hyperpolarization of the horizontal cell reduces inhibitory transmitter release. Therefore, the center photoreceptors are less inhibited and thus tend towards depolarization, which is the opposite of the response they give when struck by light. This means that stimulating both the center and the surround of these systems with light will lead to a cancellation of response in the bipolar cell (and thus in the ganglion).
Horizontal cell wiring also gives rise to a mechanism of light adaptation. The range of light to which we are sensitive is enormous (about ten orders of magnitude), while the possible firing frequency response from our neurons is comparably miniscule (about two orders of magnitude). If absolute light levels were transmitted directly, the signal for "really dark" would not be much different than the signal for "somewhat dark", because such a great range of light levels would need to be encoded by the ganglion cell’s limited range of firing frequencies. Therefore, something besides absolute light levels must be transmitted. The signal from the center-surround receptive field ganglion will be most intense when the difference between light levels in the center and surround is greatest, not when the absolute level of light is greatest. Therefore, cortical areas receive information about the difference in light level, not about the light level itself. Small differences in light level can thus be transmitted to higher visual processing centers, differences which would be indiscernible if absolute information were sent.
Ganglion Cells
Ganglion cells translate the graded potential received from bipolar cells into action potentials. Ganglion receptive fields are traditionally thought of with dome-shaped sensitivity profiles, but recent work has shown more complicated sensitivity maps in wide-field beta ganglion cells from rabbit retinas (Brown et al). These wide-field cells are mostly located in the outer retina and are connected to many bipolar cells which are in turn connected to many rod photoreceptors; approaching the inner retina and the fovea, receptive fields are smaller and formed mostly from a few cone photoreceptors. In fact, some cones in the fovea are connected to bipolar cells which have no connections to horizontal cells or to other cones. These are called midget bipolar cells, and they are connected to midget ganglion cells, which are similarly not connected to any other bipolar cells. Thus light intensity data collected in the fovea can be transmitted directly to higher processing centers; the benefits of center-surround fields (contrast-detection, spatial filtering, light adaptation) are sacrificed in return for increased resolution.
Conclusion
Retinal center-surround receptive fields, through the intricate wiring of photoreceptor, bipolar, horizontal and ganglion cells, endow the visual system with vital retinal preprocessing of visual stimuli. Providing such functions as local contrast detection, spatial filtering, and light adaptation, center-surround fields supply higher visual processing centers the data we need to make sense of the world.
References
Brown S.P., He S., Masland R.H. (2000). Receptive field microstructure and dendritic geometry of retinal ganglion cells. Neuron 27: 371-383.
Dowling JE. Neurons and Networks. 2001. Harvard University Press, Cambridge, Ma.
Kuffer S.W. (1953). Discharge patterns and functional organization of the mammalian retina. Journal of Neurophysiology 16: 37-68.
Nelsen R (2002). "Visual responses of ganglion cells" at http//:webvision.med.utah.edu.
Schiller PH (1992). The ON and OFF channels of the visual system. Trends in Neuroscience 15: 86-91.
