 |
||||
|
||||
|
||||
|
||||
The Journal of Neuroscience, July 9, 2003, 23(14):5984-5997
Previous Article | Next Article
The Representation of Retinal Blood Vessels in Primate Striate Cortex
Daniel L. Adams andJonathan C. Horton Beckman Vision Center, University of California, San Francisco, San Francisco, California 94143-0730
Abstract
Top
Abstract
Introduction
The blood vessels that nourish the inner retina cast shadows on photoreceptors, creating "angioscotomas" in the visual field. We have found the representations of angioscotomas in striate cortex of the squirrel monkey. They were detected in 9 of 12 normal adult animals by staining flatmounts for cytochrome oxidase activity after enucleation of one eye. They appeared as thin profiles in layer 4C radiating from the blind spot representation. Angioscotomas can be regarded as a local form of amblyopia. After birth, when light strikes the retina, photoreceptors beneath blood vessels are denied normal visual stimulation. This deprivation induces remodeling of geniculocortical afferents in a distribution that corresponds to the retinal vascular tree.Angioscotoma representations were most obvious in monkeys with fine ocular dominance columns and were invisible in monkeys with large, well segregated columns. In monkeys without columns, their width corresponded faithfully to the inducing retinal shadow, making it possible to calculate the minimum shadow required to produce a cortical representation. The "amblyogenic threshold" was calculated as the fraction of the pupil area eclipsed to trigger remodeling of geniculocortical afferents. It was found to be constant over retinal eccentricity, vessel size, and shadow size. Ambliogenic shadows only three to four cones wide were sufficient to generate a cortical representation, testifying to the remarkable precision of the cortical map. The representations of retinal blood vessels separated by only 0.65° were resolvable in the cortex, yielding an upper limit on cortical resolution of 340 µm in layer 4C.
Key words: amblyopia; deprivation; angioscotoma; ocular dominance column; cytochrome oxidase; retina
Introduction
Purkyn(1819
) discovered by transilluminating the sclera with candlelight that he could see his own retinal vessels. He took advantage of this phenomenon to become the first person to draw his own retinal vasculature. The easiest way to see one's own vessels is to hold a penlight at the lateral canthus and wiggle it over the closed eyelids. The retinal vessels become instantly visible as a dark pattern silhouetted against a red background. Constant movement is necessary, because stabilized retinal images fade quickly (Ditchburn and Ginsborg, 1953
von Helmholtz (1924
View larger version (174K):
[in this window]
[in a new window]
Figure 2. Monkey K. Fundus pictures showing the optic discs and retinal vessels. Below are single CO sections through layer 4C showing the representations of the optic discs pictured directly above. The left eye was enucleated, so the right blind spot representation (represented in the left cortex) appears pale. The thin lines radiating from the blind spot representations are the representations of angioscotomas from shadows cast by retinal vessels leaving the optic disc. Tissue between the arrows is shown in cross section in Figure 5. The leftmost vessel, an artery, is indicated below in the cortex by an arrowhead.
View larger version (5K):
[in this window]
[in a new window]
Figure 1. To detect one's own angioscotomas, close the right eye and stare at the X while holding the page 40 cm away. The oval will disappear into your blind spot. Slowly move the page a bit closer or farther away. The black dash above the blind spot will disappear when it falls into the angioscotoma created by your inferior temporal retinal vein. Next, shift your gaze methodically from one letter to the next. As you proceed leftward, you will lose the black dash again. This occurs when it falls in the angioscotoma of the inferior temporal artery.
View larger version (158K):
[in this window]
[in a new window]
Figure 5. Monkey K. Plastic section (1 µm) of the right eye, stained with Toluidine blue, passing between the arrows shown in Figure 2. The precise location was pinpointed by the junction between the vein and its branch, which can be seen in the retinal photograph, although it is a bit out of focus. Near the optic disc, large vessels are situated in the nerve fiber layer, which becomes especially thickened. The vein branch enters and disrupts the ganglion cell layer. The artery on the left, the angioscotoma representation of which is marked by an arrowhead in Figure 2, has a lumen diameter of 57 µm. This was calculated by measuring the lumen circumference and dividing by (we assumed the oval shape was a postmortem artifact). The small arrows denote four ellipsoids of cone inner segments, located underneath the artery. NFL, Nerve fiber layer; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; IS, inner segments; OS, outer segments; PE, pigment epithelium.
Geniculate inputs to striate cortex serving the left and right eyes are segregated into stripes within layer 4 that are known as ocular dominance columns (Hubel and Wiesel, 1969
For many years, the squirrel monkey was regarded as an anomaly among primates because it was shown repeatedly to lack ocular dominance columns (Tigges et al., 1977
This discovery has provided new insight into the physiological basis of angioscotomas. It has also yielded new data concerning the mechanisms underlying the formation of ocular dominance columns and the role of visual experience in the development of the cortex. Finally, it has revealed more clearly the precision of the visual field representation in layer 4C.
Materials and Methods
Experimental animals. These experiments were performed on 12 adult squirrel monkeys (Saimiri sciureus) from an indoor colony at the California Regional Primate Research Center (Davis, CA). All procedures were approved by the University of California San Francisco Committee on Animal Research. Each animal was normal, verified by complete ophthalmological examination under ketamine anesthesia (15 mg/kg, i.m.). During this examination, the ocular fundi were photographed with a model TRC-FE camera (Topcon Medical Systems, Paramus, NJ) mounted on a platform that allowed easy pivoting around the center of the optical axis on the corneal front surface. These photographs were montaged using Photoshop 6.0 (Adobe Systems, San Jose, CA).After the photographic montages of the fundi were prepared (usually
The animal was placed in a stereotaxic frame mounted on a model 413 professional tripod (Gitzo, Créteil, France). The tripod allowed us to orient the stereotaxic frame to align the eye's visual axis perpendicular to the center of a 6 x 9 foot tangent screen located 57 inches away. The pupils were dilated with 2.5% neosynephrine HCl and 0.125% scopolamine HCl drops. A hard 7.5 mm diameter contact lens was used to prevent corneal drying.
The fundus montage (prepared in advance) was used to select a prominent retinal landmark (e.g., a vessel bifurcation). The landmark was then identified through the fundus camera in the anesthetized, paralyzed animal. The crosshair of the camera was focused on the landmark and locked in place. Next, a mirror was placed flush against the barrel of the objective lens, and the shutter was tripped with the back of the camera open. This reflected a small circle of light back to the tangent screen at a position corresponding to the retinal landmark. With practice, retinal landmarks in the central 30° could be projected with an accuracy of ±0.1°. After plotting
) on the tangent screen from the foveal projection point to degrees eccentricity (
) by using the formula
After finishing the calibration process for each retina, we enucleated one eye using sterile technique. This procedure was necessary to reveal the angioscotomas (Fig. 1) in the cortex with CO. The animal was then reawakened and returned to its cage. A potent analgesic, Buprenorphine HCl (0.03 mg/kg, i.m.), was given every 8 hr until the animal had recovered completely. In many animals, we recorded from striate cortex for 24 hr before enucleating one eye. These data will be presented in another paper.
Measurements of newborn squirrel monkey eyes. Angioscotoma representations form in striate cortex soon after birth, induced by the shadows cast by retinal blood vessels (Adams and Horton, 2002
In infants, the pupil is smaller, less reactive to light, and less variable in size than in adults (Isenberg et al., 1989
The distance (d) from the pupil to the retina was measured with a Humphrey-Allergan Model 820 A-ultrasound unit (Zeiss Humphrey Systems, Dublin, CA) in three newborn squirrel monkeys. The mean value was 8.4 ± 0.12 mm SEM. Mean axial length was 10.1 ± 0.12 mm SEM.
These measurements provided figures for the size of the entrance pupil and the position of the real pupil in the baby squirrel monkey. Calculation of angioscotoma dimensions depends on the size and position of the exit pupil. Vakkur and Bishop (1963
To determine the size of a vascular shadow, one must also know the retinal blood vessel lumen diameter (c) and the distance (s) from the blood vessel to the cone inner/outer segment junction. Measurements of these parameters were made in adult eyes that were postfixed in 2% paraformaldehyde and 2% glutaraldehyde, followed by embedding in Epon-Araldite. These procedures induce a mean tissue shrinkage of 18% (Perry and Cowey, 1985
Histological procedures. We waited a minimum of 10 d to maximize changes in cortical CO activity induced by monocular enucleation. We then killed the animals with an injection of sodium pentobarbital (150 mg/kg). Each monkey was perfused through the left ventricle with 1 l of normal saline followed by 1l of 1% paraformaldehyde in 0.1 M phosphate buffer. Flatmounts of striate cortex were prepared from each hemisphere (Horton and Hocking, 1996a
Data analysis. Cortical CO flatmounts were imaged at 600 dots per inch on an Arcus II flatbed scanner (Agfa, Mortsel, Belgium) fitted with a transparency adapter to prepare montages of layer 4C
. Autoradio-graphs, retina, and high-power cortex pictures were taken with a Spot RT Slider camera (Diagnostic Instruments, Sterling Heights, MI) mounted on an SZH10 (Olympus Optical Co., Tokyo, Japan) or an Axioskop microscope (Zeiss, Thornwood, NY). Images were imported into Illustrator 9.0 or Photoshop 6.0 (Adobe Systems) to prepare the Figures. Measurements were made using CADtools (Hotdoor Inc., Grass Valley, CA) for Illustrator. Adjustments in brightness and contrast were made of individual sections used to assemble cortical montages with Photoshop. In addition, blood vessels were removed using the "dust and scratches" noise filter. It assigned the white pixels in an empty blood vessel lumen to the mean local color value in surrounding cortex. This step removed the distracting profiles of cortical blood vessels from the montages, making it slightly easier to see the angioscotoma representations.
Results
Representation of angioscotomas in striate cortex
Figure 2 shows the optic discs and cortical blind spot representations from monkey K. This animal had only a faint trace of ocular dominance columns, visible in a limited region of the cortex near the monocular crescent representation. No ocular dominance columns were visible in the cortex surrounding the blind spot representations. By the term "blind spot" representation, we are referring to a region of cortex corresponding to the visual field location of the blind spot. For example, the blind spot of the left eye is represented in the right cortex, and it receives input derived exclusively from the right (ipsilateral) eye.Figure 2 shows raw data, namely, single cortical sections before application of the dust and scratches filter and the preparation of montages. The sections pass mostly through layer 4C
In the right cortex, approximately eight thin dark threads radiated out from the blind spot representation of the left eye. These were the cortical representations of angioscotomas created by blood vessels emanating from the left optic disc. They were dark, like the blind spot representation, because they received input only from the intact right eye. In the left cortex, the representations of the angioscotomas were also visible; however, they were pale rather than dark, because their input derived exclusively from the enucleated left eye. Fewer angioscotomas are visible in this section because portions skimmed layer 5.
The angioscotomas were seen best in layer 4C
, but lower in contrast. They could not be detected in any other cortical layer. Figure 3A shows the array of CO patches in layer 3. No hint of the angioscotomas or the blind spot representation was apparent. Moreover, the trajectory of the angioscotomas in layer 4C
View larger version (177K):
[in this window]
[in a new window]
Figure 3. Monkey K. A, Single section of layer 3, cut through the same region of the right cortex shown in Figure 2 (see bottom left panel) but 210 µm more superficial. CO patches are present throughout the area. No effect of monocular enucleation is visible. The blind spot representation is not apparent, because the upper layers in squirrel monkey contain highly binocular cells. B, Same section as in A, with the locations of the blind spot and angioscotomas from layer 4 outlined with dots. The patches neither skirt nor follow the path of the angioscotomas. C, Single section grazing layer 4A, immediately adjacent to the section shown above. The region overlying the blind spot representation appears slightly darker than surrounding cortex, but the angioscotomas are not visible. D, Single section cut mostly through layer 5, 105 µm below the section shown from layer 4C in Figure 2, bottom left. No hint of the blind spot or angioscotomas is seen.
These findings highlight a fundamental difference between the organization of striate cortex in macaque monkeys and squirrel monkeys. In macaques, the patches outside layer 4C are organized in rows aligned with an ocular dominance column (Horton and Hubel, 1981
View larger version (87K):
[in this window]
[in a new window]
Figure 4. Macaque monkey. A, Single CO section through layer 4C, after enucleation of the contralateral eye. This experiment was done in connection with an unrelated project and is shown here simply for comparison with Figure 2, bottom left panel. The blind spot representation appears dark, as in the squirrel monkey. The striking difference is that the surrounding cortex is organized into high-contrast, large, sharply defined ocular dominance columns. B, Section from the same region, 315 µm more superficial, passing through layer 3. The patches in every other row appear pale, as opposed to the squirrel monkey, in which all patches retain the same level of CO activity after monocular enucleation. In addition, the blind spot representation is visible as an oval dark zone matching its shape in layer 4C. The blind spot representation is visible in all layers because the cortex outside layer 4 retains greater monocularity than in the squirrel monkey.
In the squirrel monkey, the patches have no relationship to ocular dominance columns in layer 4. All patches in the upper layers are labeled after a tracer injection into one eye (Horton and Hocking, 1996b
Shadows cast by retinal blood vessels
Figure 5 shows a 1 µm cross section from the right eye of monkey K that extends between the two arrows in Figure 2. The section contains two prominent arteries, flanking a vein, in the nerve fiber layer. Large retinal vessels emanating from the optic disc pass directly into the nerve fiber layer. They descend into the ganglion cell layer as their distance from the optic disc increases (Snodderly and Weinhaus, 1990
Angioscotomas arise from light absorption by hemoglobin within red blood cells, resulting in a shadow cast on photoreceptors beneath. The blood vessel wall is virtually transparent to light. Therefore, only lumen diameter is pertinent to the following optical calculations. Figure 6A shows a schematic drawing with rays of light traced from the edge of the pupil to a retinal vessel. Any object (e.g., a retinal vessel) illuminated by a non-point source (e.g., the pupil) creates a shadow composed of a central umbra surrounded by a penumbra (Fig. 6B). The width of these shadows depends on four variables: p = exit pupil diameter; d = distance from exit pupil aperture to retinal blood vessel; s = distance from blood vessel to the junction between cone inner/outer segments; c = retinal blood vessel lumen diameter.
View larger version (64K):
[in this window]
[in a new window]
Figure 6. A, Schematic scale drawing of the squirrel monkey eye illustrating the paths of light from the pupil to the retina. The dimensions used to calculated the size of shadows cast by retina blood vessels are labeled: p, pupil diameter; d, pupil aperture to blood vessel distance. B, The boxed region of retina from A shown at higher magnification. The rays of light are shown superimposed onto a 1 µm plastic section of squirrel monkey retina. A single artery of lumen diameter c is present in the retinal ganglion cell layer at a distance s from the photoreceptor inner/outer segment transition plane (the shadow plane). The shadow cast by this vessel comprises a central total shadow (umbra) bordered by a peripheral partial shadow (penumbra).
Mean pupil diameter (p) was 1.9 mm and mean distance from the exit pupil to the retina (d) was 8.4 mm (see Materials and Methods).
The distance s was sampled in 1 µm plastic cross sections throughout the retina from three fixed adult specimens. Figure 7A shows a series of aligned sections, which are stacked to show retinal blood vessels in cross section from the fovea to the optic disc. One can identify individual vessels in the fundus photograph and find their corresponding cross section in the 1 µm plastic sections to determine s and c at any point (Fig. 7C,D). For vessels visible in the photograph, s ranged between 165 and 253 µm. In the macula, s was large because the ganglion cell and Henle fiber layers were thick. Near the optic disc and along the major vascular arcades, s was large because the nerve fiber layer was thick. In peripheral nasal retina, s was small because the retinal layers were thin; it measured only 115 µm at 50° (angioscotomas were visible rarely in the cortex beyond this eccentricity).
View larger version (154K):
[in this window]
[in a new window]
Figure 7. Monkey Q, right retina. A, Stacked series of 20 plastic cross sections (1 µm), each 125 µm apart, compiled to measure s (distance from center of blood vessel lumen to cone inner/outer segment junction) for individually identified vessels at different locations. Boxed vessel is shown in panels below. B, Photograph of right fundus, outlining the retina illustrated histologically (A) with a rectangle. The fine vertical white lines indicate the positions of cross sections shown below. C, High-power view of venule only 28 µm in lumen diameter (calculated assuming circular cross section), the representation of which was visible in the cortex. The vessel was located at an eccentricity of 7.5° and cast only penumbra, equal to 83 µm (gradient between thin black lines), on the basis of values s = 243 µm, p = 1.9 mm, d = 8.4 mm. D, The same venule, but at an eccentricity of 4.1°, cast only penumbra, measuring 64 µm, on the basis of the indicated parameters. This shadow did not induce a visible representation in the cortex. OD, Optic disc.
The width of vessel shadows can be calculated using similar triangle geometry in Figure 6 (Applegate et al., 1990
Figure 8A is a graph of vessel diameter versus shadow width calculated from the above formulas using fixed values of s = 200 µm, d = 8.4 mm, and four different pupil diameters. The shadow contains two components: the umbra (solid line) and the penumbra. Total shadow width (dotted line) equals the umbra plus twice the penumbra. The graph provides shadow widths cast by blood vessels of various lumen diameter. For example, a vessel of diameter 40 µm illuminated through a 1.0 mm pupil casts a shadow 65 µm wide, the umbra of which occupies the central 17 µm. If the vessel is small enough, or the pupil large enough, the shadow will consist of penumbra alone. For example, the same 40 µm diameter vessel illuminated through a 3 mm pupil will cast no umbra on the photoreceptors but will cast a penumbra measuring 111 µm wide.
View larger version (29K):
[in this window]
[in a new window]
Figure 8. A, Graph showing the widths of shadows cast onto photoreceptors by retinal vessels of different lumen (c) at various pupil diameters (p). Solid lines represent the width of the umbra alone, and dotted lines represent the width of the whole shadow (umbra + twice penumbra). Pupil diameters of 1, 3, and 4 mm are shown, plus 1.9 mm, the experimentally determined value for an infant squirrel monkey. The values of s and d were 200 µm and 8.4 mm, respectively. B, Graph showing the widths of shadows cast onto photoreceptors by retinal vessels of different lumen (c), at various distances (s) from the photoreceptor layer. Solid lines represent the width of the umbra alone, and dotted lines represent the width of the total shadow. Five values of s are shown: 100–300 µm. For this graph, p and d were fixed at 1.9 and 8.4 mm, respectively.
Figure 8A shows shadow widths for a pupil diameter of 1.9 mm, the mean pupil size in infant squirrel monkeys, as well as for diameters of 1.0, 3.0, and 4.0 mm. From inspection of the graph, it is evident that most vessels in the ganglion cell layer of the infant monkey cast only a penumbra on photoreceptors. In fact, vessels <44 µm in diameter cast no umbra when the pupil diameter equals 1.9 mm. Only the largest vessels, which form the main vascular arcades of the retina, give rise to an umbra. For example, the large artery in Figure 5 (and Fig. 2, top right panel) has a lumen diameter of 57 µm, producing an umbra of 13 µm and a total shadow of 104 µm.
With spontaneous changes in pupil diameter, shadows from blood vessels will fluctuate in size and density. As described in a later section, the representation of vessels with a lumen as small as 28 µm can be detected in the cortex. An umbra forms transiently under such small vessels only during moments of extreme pupil constriction. Therefore, we suspect that in many cases penumbra alone is sufficient to induce an angioscotoma representation in the cortex.
The minimum vessel size of 44 µm for umbra formation calculated above used a value of 200 µm for the distance (s) from the blood vessel to the photoreceptor layer. Because s is a critical factor, we plotted shadow width as a function of s, setting pupil diameter (p) to 1.9 mm (Fig. 8B). Measurements in retinal cross sections from three animals revealed that s is never <150 µm for retinal blood vessels in the central 24°, whose angioscotomas are visible in the cortex. For p = 1.9 mm, s = 150 µm, the smallest vessel casting an umbra would have a lumen of 33 µm. As noted above, smaller vessels are represented in the cortex. These calculations, based on conservative values for p and s, confirm that many angioscotomas are induced by penumbra, without umbra.
We propose that angioscotoma representations are generated by the same process that leads to shrinkage of ocular dominance columns after early monocular eyelid suture or cataract (Hubel et al., 1977
According to this hypothesis, angioscotomas mapped in adults do not arise from vascular shadows (although they were induced by them during childhood) but represent a secondary phenomenon from amblyopia. This would explain why angioscotomas tend to be wider than the diameters of the blood vessels themselves (Remky et al., 1996
Mapping angioscotoma representations in striate cortex
Figure 9 shows the angioscotoma representations in the right cortex of monkey L. This animal had rudimentary ocular dominance columns. The montages of layer 4C were slightly uneven because of seams from the montaging process and fluctuations in CO density with varying depth in layer 4C
View larger version (105K):
[in this window]
[in a new window]
Figure 9. Monkey L, right hemisphere. A, CO flatmount montage, showing dark blind spot and angioscotoma representations after enucleation of the left eye. Except for some fluctuations in CO density near the monocular crescent, ocular dominance columns were essentially absent in this animal. B, Photomontage of the left fundus. C, Drawing of angioscotomas. D, Drawing of retina, with color-coded vessels represented in the cortex.
The angioscotoma representations were thinnest in animals with rudimentary columns, as in monkey L or monkey K (Fig. 2). Their pattern was drawn by eye (Fig. 9C) to estimate their total area and perimeter. In the right cortex of monkey L, the angioscotomas had an area of 20.8 mm2 and a perimeter of 240.6 mm. Mean angioscotoma width, obtained by dividing the area by half the perimeter, was 172 µm. In the left cortex, the angioscotomas had an area of 13.1 mm2 and a perimeter of 148.7 mm. These values yielded a mean angioscotoma width of 176 µm.
In monkey Q, the next animal, the ocular dominance columns were better developed than in monkey L (Fig. 10A, compare with Fig. 9A). Nonetheless, they were too indistinct to draw accurately. To gain an impression of their periodicity, the spatial Fourier spectrum of the CO montage was analyzed using Matlab 5.0 (Math-works, Natick, MA). The power spectrum had a peak at 1.94 cycles per millimeter, corresponding to a single column width of 258 µm. For the sake of illustration, we bandpass filtered the CO montage (Fig. 10A) to create a pattern similar in periodicity to the ocular dominance columns. This pattern was used to provide the background columns in the schematic drawing of the cortical angioscotomas (Fig. 10C).
View larger version (131K):
[in this window]
[in a new window]
Figure 10. Monkey Q, right hemisphere. A, Layer 4C CO montage, showing blind spot with 12 radiating angioscotoma representations. They are pale because the right eye was enucleated. The angioscotomas are particularly extensive and obvious in this animal. A fine pattern of rather indistinct ocular dominance columns is present. B, Retinal montage, compiled from fundus pictures taken in vivo. C, Color-coded drawing of angioscotomas visible in A. A few angioscotomas of reversed (dark) CO contrast, induced by vessels in the temporal retina of the right eye, are shown in black. D, Drawing showing 12 vessels in nasal retina represented in the right cortex.
Monkey Q had an unusually clear and complete pattern of cortical angioscotomas. They appeared so extensive, as we propose in the next few pages, because the periodicity of the animals' ocular dominance columns was ideal for their visualization. In monkey Q's right cortex, the angioscotomas had an area of 55.3 mm2, a perimeter of 358.1 mm, and a width of 310 µm. In the left cortex, the angioscotomas had an area of 82.2 mm2 and a perimeter of 528.7 mm, yielding a mean angioscotoma width of 310 µm. Compared with monkey L, the angioscotomas were much wider (310 vs 174 µm) and more fully represented.
Monkey E (Fig. 11) had clearly defined ocular dominance columns throughout striate cortex in both hemispheres. The columns were coarser than those in monkeys L and Q. Fourier analysis yielded an average single column width of 452 µm. Representations of only a few of the largest fundus vessels were visible. Combining data from both hemispheres, we found that the angioscotoma representations occupied just 36 mm2. They had a mean width of 344 µm.
View larger version (114K):
[in this window]
[in a new window]
Figure 11. Monkey E, right hemisphere. A, CO flatmount montage, showing dark blind spot and angioscotoma representations after enucleation of the left eye. A pattern of "intermediate" ocular dominance columns can be seen throughout the cortex except in the representation of the blind spot and monocular crescent. Only the largest blood vessels are represented within the ocular dominance mosaic. B, Single fundus photograph including the optic disk, major retinal vessels, and fovea. C, Drawing of angioscotomas. Dotted line represents striate cortex corresponding to the extent of the fundus photograph. D, Drawing of retina, with color-coded vessels represented in the cortex. Star, Fovea.
Monkey N had well segregated ocular dominance columns (Fig. 12). For a squirrel monkey, they were extraordinarily large. Mean single column width by Fourier analysis was 608 µm. No representations of retinal blood vessels were recognizable in the cortex; however, the blind spot and monocular crescent representations were seen easily.
View larger version (189K):
[in this window]
[in a new window]
Figure 12. Monkey N, right cortex. A highly segregated pattern of coarse ocular dominance columns (reminiscent of macaque and human striate cortex) precludes the formation of angioscotoma representations.
Among normal squirrel monkeys, there is enormous inherent variability in the expression of ocular dominance columns (Adams and Horton, 2003a
View larger version (115K):
[in this window]
[in a new window]
Figure 13. Schematic illustration of the relationship between ocular dominance column periodicity and angioscotoma visibility in the cortex. The pattern of geniculate afferents from the left and right eye is represented as black and white. The region of striate cortex shown here contains the superior half of the blind spot representation. The four panels show examples of different column periodicities: A, rudimentary columns (monkeys L, K); B, fine columns (monkeys C, Q); C, intermediate columns (monkey E); D, coarse columns (monkey N). Note that angioscotomas are best seen in animals with fine columns, because they are highlighted by columns serving the other eye. Angioscotomas are not well seen in animals with coarser, better segregated columns, because the columns override the angioscotoma pattern.
In animals with rudimentary columns, the width of angioscotoma representations is a faithful reflection of shadow width, shadow density, and local cortical magnification factor. In animals with better developed ocular dominance columns, the situation is more complex. Assume that ocular dominance columns in squirrel monkey, when they occur, are present at birth, as they are in macaques. Under these circumstances, ocular dominance columns must be remodeled in the cortex for angioscotomas to become visible. Territory belonging to the eye in which photoreceptors are obscured is taken over by the eye that has an unencumbered view of the visual scene. Consequently, solid channels develop, cutting swaths through the mosaic of ocular dominance columns. Enucleation reveals the angioscotomas because they are tantamount to remodeled ocular dominance columns. These remodeled ocular dominance columns are centered on the retinotopic position of retinal blood vessels but are not necessarily equal to their retinotopic width.
Because angioscotoma representations are simply ocular dominance columns, their width appears to be strongly influenced by the inherent periodicity of the ocular dominance columns present in any given animal. The angioscotoma representations are wider in animals with fine or intermediate columns (Figs. 10, 11) than in animals with rudimentary columns (Fig. 9). Their enhanced width makes them easier to detect. This helps explains why more of the retinal vascular tree is represented in monkey Q than in monkey L. It appears that the cortex strings together "pixels" about the size of ocular dominance columns to create the angioscotoma representations.
As the periodicity of columns increases beyond an optimal range (200–300 µm single column width), fewer and fewer retinal blood vessels are represented. In monkey E (Fig. 11), an animal with a mean column width of 450 µm (calculated from the peak in the spatial Fourier spectrum of the CO montage), only the largest retinal vessels were represented. They appeared similar to the underlying mosaic, and in many places, their paths became disrupted by the column pattern. In monkey N, whose mean column width was 608 µm, no angioscotomas could be discerned at all (Fig. 12). We suspect that in such animals columns are unable to remodel to form angioscotoma representations. This may reflect the tendency for larger columns to be better segregated. In such animals, greater "repellency" between the geniculocortical afferents serving the two eyes may render it impossible for the afferents of one eye to carve a narrow channel (corresponding to an angioscotoma) through a large ocular dominance column serving the other eye.
In animals with fine or intermediate columns, the cortical angioscotoma representations were flanked by zones of opposing contrast, producing an optical density profile resembling an inverted Mexican hat (Adams and Horton, 2002
Why are the angioscotoma representations flanked by zones of opposite contrast in monkeys with fine columns? In such animals, an angioscotoma representation consists of a continuous channel created by remodeled ocular dominance columns. As explained previously, this channel is larger than the retinotopic width of the vessel. Consequently, the angioscotoma representation also incorporates slivers of visual field flanking the blood vessel. To balance the local representation of each eye, the angioscotoma representation becomes bracketed by cortex serving the other eye. This means, essentially, that an exchange of cortical territory must occur during early development. One eye takes over a solid swath of territory, corresponding to the location of the blood vessel of the other eye, plus some surrounding visual field. In return, it abandons all territory along a strip flanking either side of the angioscotoma representation.
In animals with rudimentary columns, the angioscotoma representations faithfully represent the amblyogenic width of shadows cast by retinal vessels. They are not flanked by columns serving the other eye. Consequently, they are more difficult to detect in the cortex, because of the lack of "highlighting" by the other eye and their finer caliber. Animals with fine columns, like monkey Q, have the most extensive pattern of cortical angioscotomas because their columns amplify the width of the angioscotomas and create a frosting effect. Animals with large columns (Figs. 11, 12) have fewer cortical angioscotomas because their columns are less capable of remodeling. Thus, there is an optimal width of ocular dominance column (
Precision of the cortical retinotopic map
The retinal vascular tree is transformed by cortical magnification (M) to form the pattern of angioscotoma representations in the cortex. The largest vessels in the squirrel monkey retina are the arteries and veins emerging from the optic disc. They follow an arcuate course into the temporal fundus, maintaining an eccentricity ofAs retinal vessels branch toward the fovea, they rapidly diminish in caliber. Their taper is offset by a rise in M, maintaining cortical angioscotomas at a fairly constant width. At an eccentricity of 4–5°, however, angioscotoma representations disappear abruptly in the cortex. Do they vanish because retinal vessels become too thin to cast an amblyogenic shadow or because their cortical representations become too small to see?
To address this issue, we determined the expected width of the cortical representation of vessels coursing toward the fovea. For example, the penumbra of the vessel featured in Figure 7, C and D, was calculated to be 83 µm at 7.5° and 64 µm at 4.1°. In the squirrel monkey the equation for M along an isoeccentricity ring is Me = 188(6.45 + E)-1.85, where E = eccentricity in degrees (Adams and Horton, 2003b
As vessels flow into peripheral nasal retina, they become smaller; in addition, M diminishes. Therefore, few angioscotomas are represented in the cortex at eccentricities >30°. Only a single animal, monkey Q, had angioscotoma representations that reached the monocular crescent (Adams and Horton, 2002
Thus two distinct phenomena conspire to concentrate angioscotoma representations in a belt of cortex extending from 5° to 30°. At <5°, shadow contrast is too low to deprive retinal photoreceptors. At >30°, vessel size and M are too small to create a cortical representation large enough to be detected with CO. Between 5° and 30°, vessel diameter and M are balanced to produce angioscotomas of quite constant width.
The largest vessels represented in the cortex are the veins that emanate from the optic disc. They attain a diameter in squirrel monkey retina of nearly 100 µm (0.62°). What is the smallest retinal vessel represented in the cortex? As discussed in the previous section, angioscotoma representations are a direct product of shadow width only in animals with rudimentary ocular dominance columns. To answer this question, therefore, we examined monkey K, an animal with virtually no ocular dominance columns. Figure 14 shows the cortical representation of a venule in the left retina. It was one of the smallest vessels with a detectable cortical representation in this animal. Because the venule crossed a prominent artery, its location could be pinpointed in the retinal whole mount. It had a lumen diameter of 30 µm, corresponding to 11.2 min of arc, at an eccentricity of 12.5°. The cone outer segments in this portion of the retina had a diameter of 7–8 µm. Therefore, the vessel, when projected directly onto the cone mosaic, had a width of approximately four cones (Fig. 14E); it cast, of course, a much wider shadow. The vessel lumen was centered 180 µm above the cones. Inserting values of s = 180 µm, c = 30 µm into the graph in Figure 8B yields a shadow 72 µm (0.45°) in width that consists entirely of penumbra.
View larger version (97K):
[in this window]
[in a new window]
Figure 14. Monkey K. A, Diagram showing vessels represented in cortex of monkey K, an animal with only rudimentary ocular dominance columns. A photograph of the left optic disc is shown in Figure 2. The boxed region is shown in C. B, Diagram showing the angioscotomas visible in the right cortex, color-coded to vessels in A. The angioscotomas near the optic disc are shown in Figure 2. C, View of a small venule (red arrows) crossing the inferotemporal artery at an eccentricity of 12.5°. D, Single CO-stained section, showing representation of the venule, which measured 102 µm in the cortex. E, Photomicrograph of the unstained retinal whole mount, showing the boxed region in C where the venule crosses the artery. At this location the venule had a lumen diameter of 30 µm. The cobblestone pattern of cones is shown for comparison with the size of the blood vessels. This figure was prepared by photographing the retinal layers containing the cones and vessels separately. Photoshop 6.0 was then used to superimpose the vessels on the cone mosaic. Cones underneath the blood vessels are obscured by red cells. F, Graph showing the fraction of pupil occlusion as a function of distance on the retina from the center of the penumbra cast by the venule in E. The shadow eclipsed a maximum of 86% of the pupil. The central 20 µm (between vertical dashed lines) of the 72 µm penumbra had sufficient contrast to produce a representation in the cortex, corresponding to an amblyogenic threshold of 66% pupil occlusion. Pupil occlusion is illustrated schematically by red vessel eclipsing gray pupil at five different positions.
Applying the formula for Me, one can determine that a shadow 0.45° wide at 12.5° eccentricity will produce an angioscotoma representation 364 µm wide in the cortex. In fact, the angioscotoma representation measured only 102 µm across (Fig. 14D). This discrepancy indicates that only the central portion (102/364 µm = 28%) of the shadow was dense enough to cause remodeling of geniculocortical afferents and hence to become represented in the cortex. This corresponds to the central 20 µm (20/72 µm = 28%) of the shadow. Figure 14F shows the density profile of the penumbra cast by this vessel. In this example, the vessel eclipsed a maximum of 86% of the pupil. At 10 µm from the shadow center, where the vessel eclipsed only 66% of the pupil, the shadow density fell below the threshold required to induce a cortical representation. In terms of optical density, a shadow ranging between 0.85 and 0.45 log units was required to cause amblyopia.
Angioscotoma representations are smaller than the visual angle subtended by the vessel shadow, because only the densest central portion of the shadow is amblyogenic. The proportion of the shadow represented in the cortex varies according to s, d, p, and c. It also depends on the shadow contrast required to induce deprivation. It is unknown whether this "amblyogenic threshold" varies throughout the retina. To address this point, the analysis was repeated for five other vessels at points that could be located reliably by vessel crossings or bifurcations (Table 1). A wide range of blood vessel diameters (30–78 µm) and eccentricities (12.5–20°) was sampled. Some of the blood vessels measured were sufficiently large and close enough to the photoreceptor layer to cast an umbra, whereas others were not. The amblyogenic threshold showed a narrow range, between 52 and 66% pupil eclipse.
View this table:
[in this window]
[in a new window]
Table 1. Pupil occlusion (amblyogenic threshold) corresponding to edges of the representations of six retinal vessels in monkey K
For the 30 µm venule in Figure 14, only 20 µm of the shadow was represented in the cortex. The cones in the vicinity of the venule (located at 12.5°) had a diameter of 6–7 µm. Thus, the visuotopic map in layer 4C is sufficiently precise to resolve patterns of CO activity created by amblyogenic shadows just three cones wide. The true grain of the map is probably even finer and may approach the discrete representation of individual cones.
Another measure of the fidelity of the cortical map can be derived by examining the representation in layer 4C of closely spaced retinal vessels. Figure 15 shows the representation of the superior artery and vein in the left retina of monkey C. As the vessels emerge from the optic disc and course along the retina, their separation gradually increases. At the top of Figure 15 (
View larger version (120K):
[in this window]
[in a new window]
Figure 15. Monkey C. A, Upper pole of the left optic disc, showing an emerging superior retinal artery and vein. The vessels (blue arrowheads) remain completely separate along their retinal course. B, Corresponding region from the right striate cortex, showing the cortical representation of the vessels in layer 4C (blue arrowheads). They merge
Previously, deoxyglucose and optical imaging have been used to derive estimates of cortical resolution (Tootell et al., 1988
Discussion
Helmholtz discovered 135 years ago that scotomas corresponding to the PurkynThe representation of angioscotomas in V1 provides an intriguing glimpse of the role played by visual experience in the development of the cortex. Its development proceeds according to an innate plan, which is refined by sensory stimulation after birth. Ocular dominance columns provide a classic example of this interplay between genetic instructions and visual experience. Their formation occurs without visual experience (Horton and Hocking, 1996a
Amblyopia supplies a classic example of how visual experience can influence the development of the cortex. Kittens and monkeys raised with monocular eyelid suture, to simulate congenital cataract, show expansion of the columns serving the open eye at the expense of those belonging to the deprived eye (Hubel et al., 1977
Our finding that angioscotomas are represented in the brain allows one to generalize from observations in lid-sutured animals to the process of cortical development in normal animals. In some squirrel monkeys, ocular dominance columns are so weakly expressed that vascular shadows have an unfettered ability to rearrange geniculocortical afferents. In these cases, the narrow channel carved through the cortex reflects faithfully the "amblyogenic threshold" of the shadow (Fig. 14F). In other animals, with better developed columns, shadows also induce rearrangement of geniculocortical afferents but do so by reallocating units approximately the size of ocular dominance columns. The angioscotoma representations thus appear widened, or amplified, because they subtend more visual field than the amblyogenic portion of the blood vessel shadow. Remarkably, they become surrounded on either side by columns serving the other eye. This frosting makes the angioscotomas more conspicuous in cortical flatmounts. The frosting phenomenon arises, we suspect, because the ocular dominance columns possess an intrinsic periodicity. Vascular shadows perturb the local balance of neural activity in the cortex, tipping column formation in favor of one eye. The other eye is compensated by reciprocal assignment of the nearest adjacent cortex. Such mechanisms are proposed, on theoretical grounds, to underlie the crystallization of patterns of ocular dominance columns in striate cortex (Willshaw and von der Malsburg, 1976
The discovery that angioscotomas are represented in the cortex demonstrates that neuronal activity can influence the pattern formed locally by ocular dominance columns; however, there are inherent limits to the spatial scale of this process. Squirrel monkeys manifest enormous natural heterogeneity in their expression of ocular dominance columns (Adams and Horton, 2003a
Another possi
